From 102712a32ff5107cdf9c029c75c805cc8f5aada5 Mon Sep 17 00:00:00 2001 From: Jeffy Chen Date: Thu, 19 Oct 2017 11:48:04 +0800 Subject: [PATCH 0001/1471] drm/rockchip: analogix_dp: Remove unnecessary init code Remove unnecessary init code, since we would do it in the power_on() callback. Also move of parse code to probe(). Fixes: 9e32e16e9e98 ("drm: rockchip: dp: add rockchip platform dp driver") Signed-off-by: Jeffy Chen Acked-by: Mark Yao Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/20171019034812.13768-3-jeffy.chen@rock-chips.com --- .../gpu/drm/rockchip/analogix_dp-rockchip.c | 27 +++++-------------- 1 file changed, 6 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c b/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c index 4d3f6ad0abdde..8cae5ad926cd0 100644 --- a/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c +++ b/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c @@ -269,7 +269,7 @@ static struct drm_encoder_funcs rockchip_dp_encoder_funcs = { .destroy = rockchip_dp_drm_encoder_destroy, }; -static int rockchip_dp_init(struct rockchip_dp_device *dp) +static int rockchip_dp_of_probe(struct rockchip_dp_device *dp) { struct device *dev = dp->dev; struct device_node *np = dev->of_node; @@ -303,19 +303,6 @@ static int rockchip_dp_init(struct rockchip_dp_device *dp) return PTR_ERR(dp->rst); } - ret = clk_prepare_enable(dp->pclk); - if (ret < 0) { - DRM_DEV_ERROR(dp->dev, "failed to enable pclk %d\n", ret); - return ret; - } - - ret = rockchip_dp_pre_init(dp); - if (ret < 0) { - DRM_DEV_ERROR(dp->dev, "failed to pre init %d\n", ret); - clk_disable_unprepare(dp->pclk); - return ret; - } - return 0; } @@ -361,10 +348,6 @@ static int rockchip_dp_bind(struct device *dev, struct device *master, if (!dp_data) return -ENODEV; - ret = rockchip_dp_init(dp); - if (ret < 0) - return ret; - dp->data = dp_data; dp->drm_dev = drm_dev; @@ -398,7 +381,6 @@ static void rockchip_dp_unbind(struct device *dev, struct device *master, rockchip_drm_psr_unregister(&dp->encoder); analogix_dp_unbind(dev, master, data); - clk_disable_unprepare(dp->pclk); } static const struct component_ops rockchip_dp_component_ops = { @@ -414,7 +396,7 @@ static int rockchip_dp_probe(struct platform_device *pdev) int ret; ret = drm_of_find_panel_or_bridge(dev->of_node, 1, 0, &panel, NULL); - if (ret) + if (ret < 0) return ret; dp = devm_kzalloc(dev, sizeof(*dp), GFP_KERNEL); @@ -422,9 +404,12 @@ static int rockchip_dp_probe(struct platform_device *pdev) return -ENOMEM; dp->dev = dev; - dp->plat_data.panel = panel; + ret = rockchip_dp_of_probe(dp); + if (ret < 0) + return ret; + /* * We just use the drvdata until driver run into component * add function, and then we would set drvdata to null, so -- GitLab From 02c9656b2f0d6997939933d8573c2ffb587427e6 Mon Sep 17 00:00:00 2001 From: Haneen Mohammed Date: Tue, 17 Oct 2017 22:30:07 -0600 Subject: [PATCH 0002/1471] drm: Move debug macros out of drmP.h This patch extract DRM_* debug macros from drmP.h to drm_print.h and move printing related functions used by these macros from drm_drv.[hc] to drm_print.[hc]. Signed-off-by: Haneen Mohammed Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/4020bc7c5ffad2af516919f78bb837c7f366b82b.1508297716.git.hamohammed.sa@gmail.com --- drivers/gpu/drm/drm_drv.c | 47 --------- drivers/gpu/drm/drm_print.c | 47 +++++++++ include/drm/drmP.h | 193 +--------------------------------- include/drm/drm_drv.h | 7 -- include/drm/drm_print.h | 199 ++++++++++++++++++++++++++++++++++++ 5 files changed, 247 insertions(+), 246 deletions(-) diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index c0292e5d72811..bc2f10adc5901 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -74,53 +74,6 @@ static bool drm_core_init_complete = false; static struct dentry *drm_debugfs_root; -#define DRM_PRINTK_FMT "[" DRM_NAME ":%s]%s %pV" - -void drm_dev_printk(const struct device *dev, const char *level, - unsigned int category, const char *function_name, - const char *prefix, const char *format, ...) -{ - struct va_format vaf; - va_list args; - - if (category != DRM_UT_NONE && !(drm_debug & category)) - return; - - va_start(args, format); - vaf.fmt = format; - vaf.va = &args; - - if (dev) - dev_printk(level, dev, DRM_PRINTK_FMT, function_name, prefix, - &vaf); - else - printk("%s" DRM_PRINTK_FMT, level, function_name, prefix, &vaf); - - va_end(args); -} -EXPORT_SYMBOL(drm_dev_printk); - -void drm_printk(const char *level, unsigned int category, - const char *format, ...) -{ - struct va_format vaf; - va_list args; - - if (category != DRM_UT_NONE && !(drm_debug & category)) - return; - - va_start(args, format); - vaf.fmt = format; - vaf.va = &args; - - printk("%s" "[" DRM_NAME ":%ps]%s %pV", - level, __builtin_return_address(0), - strcmp(level, KERN_ERR) == 0 ? " *ERROR*" : "", &vaf); - - va_end(args); -} -EXPORT_SYMBOL(drm_printk); - /* * DRM Minors * A DRM device can provide several char-dev interfaces on the DRM-Major. Each diff --git a/drivers/gpu/drm/drm_print.c b/drivers/gpu/drm/drm_print.c index 74c466aca6220..0b3bf476dc4b5 100644 --- a/drivers/gpu/drm/drm_print.c +++ b/drivers/gpu/drm/drm_print.c @@ -63,3 +63,50 @@ void drm_printf(struct drm_printer *p, const char *f, ...) va_end(args); } EXPORT_SYMBOL(drm_printf); + +#define DRM_PRINTK_FMT "[" DRM_NAME ":%s]%s %pV" + +void drm_dev_printk(const struct device *dev, const char *level, + unsigned int category, const char *function_name, + const char *prefix, const char *format, ...) +{ + struct va_format vaf; + va_list args; + + if (category != DRM_UT_NONE && !(drm_debug & category)) + return; + + va_start(args, format); + vaf.fmt = format; + vaf.va = &args; + + if (dev) + dev_printk(level, dev, DRM_PRINTK_FMT, function_name, prefix, + &vaf); + else + printk("%s" DRM_PRINTK_FMT, level, function_name, prefix, &vaf); + + va_end(args); +} +EXPORT_SYMBOL(drm_dev_printk); + +void drm_printk(const char *level, unsigned int category, + const char *format, ...) +{ + struct va_format vaf; + va_list args; + + if (category != DRM_UT_NONE && !(drm_debug & category)) + return; + + va_start(args, format); + vaf.fmt = format; + vaf.va = &args; + + printk("%s" "[" DRM_NAME ":%ps]%s %pV", + level, __builtin_return_address(0), + strcmp(level, KERN_ERR) == 0 ? " *ERROR*" : "", &vaf); + + va_end(args); +} +EXPORT_SYMBOL(drm_printk); diff --git a/include/drm/drmP.h b/include/drm/drmP.h index 7277783a4ff00..c6666cd09347f 100644 --- a/include/drm/drmP.h +++ b/include/drm/drmP.h @@ -75,6 +75,7 @@ #include #include #include +#include #include #include #include @@ -94,208 +95,16 @@ struct dma_buf_attachment; struct pci_dev; struct pci_controller; -/* - * The following categories are defined: - * - * CORE: Used in the generic drm code: drm_ioctl.c, drm_mm.c, drm_memory.c, ... - * This is the category used by the DRM_DEBUG() macro. - * - * DRIVER: Used in the vendor specific part of the driver: i915, radeon, ... - * This is the category used by the DRM_DEBUG_DRIVER() macro. - * - * KMS: used in the modesetting code. - * This is the category used by the DRM_DEBUG_KMS() macro. - * - * PRIME: used in the prime code. - * This is the category used by the DRM_DEBUG_PRIME() macro. - * - * ATOMIC: used in the atomic code. - * This is the category used by the DRM_DEBUG_ATOMIC() macro. - * - * VBL: used for verbose debug message in the vblank code - * This is the category used by the DRM_DEBUG_VBL() macro. - * - * Enabling verbose debug messages is done through the drm.debug parameter, - * each category being enabled by a bit. - * - * drm.debug=0x1 will enable CORE messages - * drm.debug=0x2 will enable DRIVER messages - * drm.debug=0x3 will enable CORE and DRIVER messages - * ... - * drm.debug=0x3f will enable all messages - * - * An interesting feature is that it's possible to enable verbose logging at - * run-time by echoing the debug value in its sysfs node: - * # echo 0xf > /sys/module/drm/parameters/debug - */ -#define DRM_UT_NONE 0x00 -#define DRM_UT_CORE 0x01 -#define DRM_UT_DRIVER 0x02 -#define DRM_UT_KMS 0x04 -#define DRM_UT_PRIME 0x08 -#define DRM_UT_ATOMIC 0x10 -#define DRM_UT_VBL 0x20 -#define DRM_UT_STATE 0x40 - /***********************************************************************/ /** \name DRM template customization defaults */ /*@{*/ -/***********************************************************************/ -/** \name Macros to make printk easier */ -/*@{*/ - -#define _DRM_PRINTK(once, level, fmt, ...) \ - do { \ - printk##once(KERN_##level "[" DRM_NAME "] " fmt, \ - ##__VA_ARGS__); \ - } while (0) - -#define DRM_INFO(fmt, ...) \ - _DRM_PRINTK(, INFO, fmt, ##__VA_ARGS__) -#define DRM_NOTE(fmt, ...) \ - _DRM_PRINTK(, NOTICE, fmt, ##__VA_ARGS__) -#define DRM_WARN(fmt, ...) \ - _DRM_PRINTK(, WARNING, fmt, ##__VA_ARGS__) - -#define DRM_INFO_ONCE(fmt, ...) \ - _DRM_PRINTK(_once, INFO, fmt, ##__VA_ARGS__) -#define DRM_NOTE_ONCE(fmt, ...) \ - _DRM_PRINTK(_once, NOTICE, fmt, ##__VA_ARGS__) -#define DRM_WARN_ONCE(fmt, ...) \ - _DRM_PRINTK(_once, WARNING, fmt, ##__VA_ARGS__) - -/** - * Error output. - * - * \param fmt printf() like format string. - * \param arg arguments - */ -#define DRM_DEV_ERROR(dev, fmt, ...) \ - drm_dev_printk(dev, KERN_ERR, DRM_UT_NONE, __func__, " *ERROR*",\ - fmt, ##__VA_ARGS__) -#define DRM_ERROR(fmt, ...) \ - drm_printk(KERN_ERR, DRM_UT_NONE, fmt, ##__VA_ARGS__) - -/** - * Rate limited error output. Like DRM_ERROR() but won't flood the log. - * - * \param fmt printf() like format string. - * \param arg arguments - */ -#define DRM_DEV_ERROR_RATELIMITED(dev, fmt, ...) \ -({ \ - static DEFINE_RATELIMIT_STATE(_rs, \ - DEFAULT_RATELIMIT_INTERVAL, \ - DEFAULT_RATELIMIT_BURST); \ - \ - if (__ratelimit(&_rs)) \ - DRM_DEV_ERROR(dev, fmt, ##__VA_ARGS__); \ -}) -#define DRM_ERROR_RATELIMITED(fmt, ...) \ - DRM_DEV_ERROR_RATELIMITED(NULL, fmt, ##__VA_ARGS__) - -#define DRM_DEV_INFO(dev, fmt, ...) \ - drm_dev_printk(dev, KERN_INFO, DRM_UT_NONE, __func__, "", fmt, \ - ##__VA_ARGS__) - -#define DRM_DEV_INFO_ONCE(dev, fmt, ...) \ -({ \ - static bool __print_once __read_mostly; \ - if (!__print_once) { \ - __print_once = true; \ - DRM_DEV_INFO(dev, fmt, ##__VA_ARGS__); \ - } \ -}) - -/** - * Debug output. - * - * \param fmt printf() like format string. - * \param arg arguments - */ -#define DRM_DEV_DEBUG(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_CORE, __func__, "", fmt, \ - ##args) -#define DRM_DEBUG(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_CORE, fmt, ##__VA_ARGS__) - -#define DRM_DEV_DEBUG_DRIVER(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_DRIVER, __func__, "", \ - fmt, ##args) -#define DRM_DEBUG_DRIVER(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_DRIVER, fmt, ##__VA_ARGS__) - -#define DRM_DEV_DEBUG_KMS(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_KMS, __func__, "", fmt, \ - ##args) -#define DRM_DEBUG_KMS(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_KMS, fmt, ##__VA_ARGS__) - -#define DRM_DEV_DEBUG_PRIME(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_PRIME, __func__, "", \ - fmt, ##args) -#define DRM_DEBUG_PRIME(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_PRIME, fmt, ##__VA_ARGS__) - -#define DRM_DEV_DEBUG_ATOMIC(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_ATOMIC, __func__, "", \ - fmt, ##args) -#define DRM_DEBUG_ATOMIC(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_ATOMIC, fmt, ##__VA_ARGS__) - -#define DRM_DEV_DEBUG_VBL(dev, fmt, args...) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_VBL, __func__, "", fmt, \ - ##args) -#define DRM_DEBUG_VBL(fmt, ...) \ - drm_printk(KERN_DEBUG, DRM_UT_VBL, fmt, ##__VA_ARGS__) - -#define _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, level, fmt, args...) \ -({ \ - static DEFINE_RATELIMIT_STATE(_rs, \ - DEFAULT_RATELIMIT_INTERVAL, \ - DEFAULT_RATELIMIT_BURST); \ - if (__ratelimit(&_rs)) \ - drm_dev_printk(dev, KERN_DEBUG, DRM_UT_ ## level, \ - __func__, "", fmt, ##args); \ -}) - -/** - * Rate limited debug output. Like DRM_DEBUG() but won't flood the log. - * - * \param fmt printf() like format string. - * \param arg arguments - */ -#define DRM_DEV_DEBUG_RATELIMITED(dev, fmt, args...) \ - DEV__DRM_DEFINE_DEBUG_RATELIMITED(dev, CORE, fmt, ##args) -#define DRM_DEBUG_RATELIMITED(fmt, args...) \ - DRM_DEV_DEBUG_RATELIMITED(NULL, fmt, ##args) -#define DRM_DEV_DEBUG_DRIVER_RATELIMITED(dev, fmt, args...) \ - _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, DRIVER, fmt, ##args) -#define DRM_DEBUG_DRIVER_RATELIMITED(fmt, args...) \ - DRM_DEV_DEBUG_DRIVER_RATELIMITED(NULL, fmt, ##args) -#define DRM_DEV_DEBUG_KMS_RATELIMITED(dev, fmt, args...) \ - _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, KMS, fmt, ##args) -#define DRM_DEBUG_KMS_RATELIMITED(fmt, args...) \ - DRM_DEV_DEBUG_KMS_RATELIMITED(NULL, fmt, ##args) -#define DRM_DEV_DEBUG_PRIME_RATELIMITED(dev, fmt, args...) \ - _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, PRIME, fmt, ##args) -#define DRM_DEBUG_PRIME_RATELIMITED(fmt, args...) \ - DRM_DEV_DEBUG_PRIME_RATELIMITED(NULL, fmt, ##args) - -/* Format strings and argument splitters to simplify printing - * various "complex" objects - */ - -/*@}*/ - /***********************************************************************/ /** \name Internal types and structures */ /*@{*/ #define DRM_IF_VERSION(maj, min) (maj << 16 | min) - /** * drm_drv_uses_atomic_modeset - check if the driver implements * atomic_commit() diff --git a/include/drm/drm_drv.h b/include/drm/drm_drv.h index ee06ecd6c01fc..0e90ef24214b9 100644 --- a/include/drm/drm_drv.h +++ b/include/drm/drm_drv.h @@ -592,13 +592,6 @@ struct drm_driver { int dev_priv_size; }; -__printf(6, 7) -void drm_dev_printk(const struct device *dev, const char *level, - unsigned int category, const char *function_name, - const char *prefix, const char *format, ...); -__printf(3, 4) -void drm_printk(const char *level, unsigned int category, - const char *format, ...); extern unsigned int drm_debug; int drm_dev_init(struct drm_device *dev, diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h index ca4d7c6321f28..41dd757c82f48 100644 --- a/include/drm/drm_print.h +++ b/include/drm/drm_print.h @@ -128,4 +128,203 @@ static inline struct drm_printer drm_debug_printer(const char *prefix) }; return p; } + +/* + * The following categories are defined: + * + * CORE: Used in the generic drm code: drm_ioctl.c, drm_mm.c, drm_memory.c, ... + * This is the category used by the DRM_DEBUG() macro. + * + * DRIVER: Used in the vendor specific part of the driver: i915, radeon, ... + * This is the category used by the DRM_DEBUG_DRIVER() macro. + * + * KMS: used in the modesetting code. + * This is the category used by the DRM_DEBUG_KMS() macro. + * + * PRIME: used in the prime code. + * This is the category used by the DRM_DEBUG_PRIME() macro. + * + * ATOMIC: used in the atomic code. + * This is the category used by the DRM_DEBUG_ATOMIC() macro. + * + * VBL: used for verbose debug message in the vblank code + * This is the category used by the DRM_DEBUG_VBL() macro. + * + * Enabling verbose debug messages is done through the drm.debug parameter, + * each category being enabled by a bit. + * + * drm.debug=0x1 will enable CORE messages + * drm.debug=0x2 will enable DRIVER messages + * drm.debug=0x3 will enable CORE and DRIVER messages + * ... + * drm.debug=0x3f will enable all messages + * + * An interesting feature is that it's possible to enable verbose logging at + * run-time by echoing the debug value in its sysfs node: + * # echo 0xf > /sys/module/drm/parameters/debug + */ +#define DRM_UT_NONE 0x00 +#define DRM_UT_CORE 0x01 +#define DRM_UT_DRIVER 0x02 +#define DRM_UT_KMS 0x04 +#define DRM_UT_PRIME 0x08 +#define DRM_UT_ATOMIC 0x10 +#define DRM_UT_VBL 0x20 +#define DRM_UT_STATE 0x40 + +__printf(6, 7) +void drm_dev_printk(const struct device *dev, const char *level, + unsigned int category, const char *function_name, + const char *prefix, const char *format, ...); +__printf(3, 4) +void drm_printk(const char *level, unsigned int category, + const char *format, ...); +/***********************************************************************/ +/** \name Macros to make printk easier */ +/*@{*/ + +#define _DRM_PRINTK(once, level, fmt, ...) \ + do { \ + printk##once(KERN_##level "[" DRM_NAME "] " fmt, \ + ##__VA_ARGS__); \ + } while (0) + +#define DRM_INFO(fmt, ...) \ + _DRM_PRINTK(, INFO, fmt, ##__VA_ARGS__) +#define DRM_NOTE(fmt, ...) \ + _DRM_PRINTK(, NOTICE, fmt, ##__VA_ARGS__) +#define DRM_WARN(fmt, ...) \ + _DRM_PRINTK(, WARNING, fmt, ##__VA_ARGS__) + +#define DRM_INFO_ONCE(fmt, ...) \ + _DRM_PRINTK(_once, INFO, fmt, ##__VA_ARGS__) +#define DRM_NOTE_ONCE(fmt, ...) \ + _DRM_PRINTK(_once, NOTICE, fmt, ##__VA_ARGS__) +#define DRM_WARN_ONCE(fmt, ...) \ + _DRM_PRINTK(_once, WARNING, fmt, ##__VA_ARGS__) + +/** + * Error output. + * + * \param fmt printf() like format string. + * \param arg arguments + */ +#define DRM_DEV_ERROR(dev, fmt, ...) \ + drm_dev_printk(dev, KERN_ERR, DRM_UT_NONE, __func__, " *ERROR*",\ + fmt, ##__VA_ARGS__) +#define DRM_ERROR(fmt, ...) \ + drm_printk(KERN_ERR, DRM_UT_NONE, fmt, ##__VA_ARGS__) + +/** + * Rate limited error output. Like DRM_ERROR() but won't flood the log. + * + * \param fmt printf() like format string. + * \param arg arguments + */ +#define DRM_DEV_ERROR_RATELIMITED(dev, fmt, ...) \ +({ \ + static DEFINE_RATELIMIT_STATE(_rs, \ + DEFAULT_RATELIMIT_INTERVAL, \ + DEFAULT_RATELIMIT_BURST); \ + \ + if (__ratelimit(&_rs)) \ + DRM_DEV_ERROR(dev, fmt, ##__VA_ARGS__); \ +}) +#define DRM_ERROR_RATELIMITED(fmt, ...) \ + DRM_DEV_ERROR_RATELIMITED(NULL, fmt, ##__VA_ARGS__) + +#define DRM_DEV_INFO(dev, fmt, ...) \ + drm_dev_printk(dev, KERN_INFO, DRM_UT_NONE, __func__, "", fmt, \ + ##__VA_ARGS__) + +#define DRM_DEV_INFO_ONCE(dev, fmt, ...) \ +({ \ + static bool __print_once __read_mostly; \ + if (!__print_once) { \ + __print_once = true; \ + DRM_DEV_INFO(dev, fmt, ##__VA_ARGS__); \ + } \ +}) + +/** + * Debug output. + * + * \param fmt printf() like format string. + * \param arg arguments + */ +#define DRM_DEV_DEBUG(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_CORE, __func__, "", fmt, \ + ##args) +#define DRM_DEBUG(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_CORE, fmt, ##__VA_ARGS__) + +#define DRM_DEV_DEBUG_DRIVER(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_DRIVER, __func__, "", \ + fmt, ##args) +#define DRM_DEBUG_DRIVER(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_DRIVER, fmt, ##__VA_ARGS__) + +#define DRM_DEV_DEBUG_KMS(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_KMS, __func__, "", fmt, \ + ##args) +#define DRM_DEBUG_KMS(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_KMS, fmt, ##__VA_ARGS__) + +#define DRM_DEV_DEBUG_PRIME(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_PRIME, __func__, "", \ + fmt, ##args) +#define DRM_DEBUG_PRIME(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_PRIME, fmt, ##__VA_ARGS__) + +#define DRM_DEV_DEBUG_ATOMIC(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_ATOMIC, __func__, "", \ + fmt, ##args) +#define DRM_DEBUG_ATOMIC(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_ATOMIC, fmt, ##__VA_ARGS__) + +#define DRM_DEV_DEBUG_VBL(dev, fmt, args...) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_VBL, __func__, "", fmt, \ + ##args) +#define DRM_DEBUG_VBL(fmt, ...) \ + drm_printk(KERN_DEBUG, DRM_UT_VBL, fmt, ##__VA_ARGS__) + +#define _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, level, fmt, args...) \ +({ \ + static DEFINE_RATELIMIT_STATE(_rs, \ + DEFAULT_RATELIMIT_INTERVAL, \ + DEFAULT_RATELIMIT_BURST); \ + if (__ratelimit(&_rs)) \ + drm_dev_printk(dev, KERN_DEBUG, DRM_UT_ ## level, \ + __func__, "", fmt, ##args); \ +}) + +/** + * Rate limited debug output. Like DRM_DEBUG() but won't flood the log. + * + * \param fmt printf() like format string. + * \param arg arguments + */ +#define DRM_DEV_DEBUG_RATELIMITED(dev, fmt, args...) \ + DEV__DRM_DEFINE_DEBUG_RATELIMITED(dev, CORE, fmt, ##args) +#define DRM_DEBUG_RATELIMITED(fmt, args...) \ + DRM_DEV_DEBUG_RATELIMITED(NULL, fmt, ##args) +#define DRM_DEV_DEBUG_DRIVER_RATELIMITED(dev, fmt, args...) \ + _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, DRIVER, fmt, ##args) +#define DRM_DEBUG_DRIVER_RATELIMITED(fmt, args...) \ + DRM_DEV_DEBUG_DRIVER_RATELIMITED(NULL, fmt, ##args) +#define DRM_DEV_DEBUG_KMS_RATELIMITED(dev, fmt, args...) \ + _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, KMS, fmt, ##args) +#define DRM_DEBUG_KMS_RATELIMITED(fmt, args...) \ + DRM_DEV_DEBUG_KMS_RATELIMITED(NULL, fmt, ##args) +#define DRM_DEV_DEBUG_PRIME_RATELIMITED(dev, fmt, args...) \ + _DRM_DEV_DEFINE_DEBUG_RATELIMITED(dev, PRIME, fmt, ##args) +#define DRM_DEBUG_PRIME_RATELIMITED(fmt, args...) \ + DRM_DEV_DEBUG_PRIME_RATELIMITED(NULL, fmt, ##args) + +/* Format strings and argument splitters to simplify printing + * various "complex" objects + */ + +/*@}*/ + #endif /* DRM_PRINT_H_ */ -- GitLab From 091756bbb1a961324bdadf10e273b677fcf4e74a Mon Sep 17 00:00:00 2001 From: Haneen Mohammed Date: Tue, 17 Oct 2017 22:31:22 -0600 Subject: [PATCH 0003/1471] drm/print: Update old comment style Remove old comment style used by doxygen. And remove comment left from commit 99cdb35e787b ("drm/doc: move printf helpers out of drmP.h") after refactoring drmP.h. Signed-off-by: Haneen Mohammed Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/4d89eef3945dbffd402d2ecdc130108b0565c158.1508297716.git.hamohammed.sa@gmail.com --- include/drm/drm_print.h | 27 ++++++++++----------------- 1 file changed, 10 insertions(+), 17 deletions(-) diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h index 41dd757c82f48..7b9c86a6ca3e8 100644 --- a/include/drm/drm_print.h +++ b/include/drm/drm_print.h @@ -179,9 +179,8 @@ void drm_dev_printk(const struct device *dev, const char *level, __printf(3, 4) void drm_printk(const char *level, unsigned int category, const char *format, ...); -/***********************************************************************/ -/** \name Macros to make printk easier */ -/*@{*/ + +/* Macros to make printk easier */ #define _DRM_PRINTK(once, level, fmt, ...) \ do { \ @@ -206,8 +205,8 @@ void drm_printk(const char *level, unsigned int category, /** * Error output. * - * \param fmt printf() like format string. - * \param arg arguments + * @dev: device pointer + * @fmt: printf() like format string. */ #define DRM_DEV_ERROR(dev, fmt, ...) \ drm_dev_printk(dev, KERN_ERR, DRM_UT_NONE, __func__, " *ERROR*",\ @@ -218,8 +217,8 @@ void drm_printk(const char *level, unsigned int category, /** * Rate limited error output. Like DRM_ERROR() but won't flood the log. * - * \param fmt printf() like format string. - * \param arg arguments + * @dev: device pointer + * @fmt: printf() like format string. */ #define DRM_DEV_ERROR_RATELIMITED(dev, fmt, ...) \ ({ \ @@ -249,8 +248,8 @@ void drm_printk(const char *level, unsigned int category, /** * Debug output. * - * \param fmt printf() like format string. - * \param arg arguments + * @dev: device pointer + * @fmt: printf() like format string. */ #define DRM_DEV_DEBUG(dev, fmt, args...) \ drm_dev_printk(dev, KERN_DEBUG, DRM_UT_CORE, __func__, "", fmt, \ @@ -301,8 +300,8 @@ void drm_printk(const char *level, unsigned int category, /** * Rate limited debug output. Like DRM_DEBUG() but won't flood the log. * - * \param fmt printf() like format string. - * \param arg arguments + * @dev: device pointer + * @fmt: printf() like format string. */ #define DRM_DEV_DEBUG_RATELIMITED(dev, fmt, args...) \ DEV__DRM_DEFINE_DEBUG_RATELIMITED(dev, CORE, fmt, ##args) @@ -321,10 +320,4 @@ void drm_printk(const char *level, unsigned int category, #define DRM_DEBUG_PRIME_RATELIMITED(fmt, args...) \ DRM_DEV_DEBUG_PRIME_RATELIMITED(NULL, fmt, ##args) -/* Format strings and argument splitters to simplify printing - * various "complex" objects - */ - -/*@}*/ - #endif /* DRM_PRINT_H_ */ -- GitLab From 9bf15d0542776c7cff98c0cf76124db058422f54 Mon Sep 17 00:00:00 2001 From: Sean Paul Date: Fri, 20 Oct 2017 13:25:48 -0400 Subject: [PATCH 0004/1471] drm/rockchip: Fix build warning in analogix_dp-rockchip.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I didn't catch this before applying, just right after (of course). Fixes: ../drivers/gpu/drm/rockchip/analogix_dp-rockchip.c: In function ‘rockchip_dp_of_probe’: ../drivers/gpu/drm/rockchip/analogix_dp-rockchip.c:276:6: warning: unused variable ‘ret’ [-Wunused-variable] int ret; ^~~ Fixes: 102712a32ff5 ("drm/rockchip: analogix_dp: Remove unnecessary init code") Cc: Jeffy Chen Cc: Mark Yao Cc: Sean Paul Cc: Heiko Stuebner Cc: dri-devel@lists.freedesktop.org Cc: linux-arm-kernel@lists.infradead.org Cc: linux-rockchip@lists.infradead.org Acked-by: Daniel Vetter Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/20171020172557.54900-1-seanpaul@chromium.org --- drivers/gpu/drm/rockchip/analogix_dp-rockchip.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c b/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c index 8cae5ad926cd0..39944c0f1c0be 100644 --- a/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c +++ b/drivers/gpu/drm/rockchip/analogix_dp-rockchip.c @@ -273,7 +273,6 @@ static int rockchip_dp_of_probe(struct rockchip_dp_device *dp) { struct device *dev = dp->dev; struct device_node *np = dev->of_node; - int ret; dp->grf = syscon_regmap_lookup_by_phandle(np, "rockchip,grf"); if (IS_ERR(dp->grf)) { -- GitLab From b7fa1bdb10359bf4330f30bb9f36ecde91e5e5a5 Mon Sep 17 00:00:00 2001 From: Harsha Sharma Date: Sun, 15 Oct 2017 01:58:23 +0530 Subject: [PATCH 0005/1471] drm/tinydrm: Replace list_for_each with list_for_each_entry MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace use of list_for_each with list_for_each_entry to simplify the code and remove variables that are used only in list_for_each. Done with following coccinelle patch: @r@ identifier fn,i,f,p; expression e; iterator name list_for_each, list_for_each_entry; type T; @@ fn(...) { ++ T *i; <+... - list_for_each(p,e) + list_for_each_entry(i,e,f) { ... - T *i = list_entry(p,T,f); ... } ...+> } @@ identifier r.fn,r.p; @@ fn(...) { ... - struct list_head *p; ... when != p } @@ identifier r.fn,r.i,r.f; expression r.e; statement S; @@ fn(...) { <... list_for_each_entry(i,e,f) - { S - } ...> } @s@ identifier i,f,p; expression e; type T; @@ - list_for_each(p,e) + list_for_each_entry(i,e,f) { ... when != T *i; - i = list_entry(p,T,f); ... } @@ identifier s.p; @@ - struct list_head *p; ... when != p @@ identifier s.i,s.f; expression s.e; statement S; @@ list_for_each_entry(i,e,f) - { S - } Signed-off-by: Harsha Sharma Reviewed-by: Sean Paul Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/20171014202823.29230-1-harshasharmaiitr@gmail.com --- drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c b/drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c index bd6cce093a850..bf96072d1b970 100644 --- a/drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c +++ b/drivers/gpu/drm/tinydrm/core/tinydrm-helpers.c @@ -414,11 +414,9 @@ tinydrm_dbg_spi_print(struct spi_device *spi, struct spi_transfer *tr, void _tinydrm_dbg_spi_message(struct spi_device *spi, struct spi_message *m) { struct spi_transfer *tmp; - struct list_head *pos; int i = 0; - list_for_each(pos, &m->transfers) { - tmp = list_entry(pos, struct spi_transfer, transfer_list); + list_for_each_entry(tmp, &m->transfers, transfer_list) { if (tmp->tx_buf) tinydrm_dbg_spi_print(spi, tmp, tmp->tx_buf, i, true); -- GitLab From 5427f207852d5b905e251a5a728c8604d3594d58 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 23 Oct 2017 22:32:34 +0100 Subject: [PATCH 0006/1471] drm/i915: Bump wait-times for the final CS interrupt before parking In the idle worker we drop the prolonged GT wakeref used to cover such essentials as interrupt delivery. (When a CS interrupt arrives, we also assert that the GT is awake.) However, it turns out that 10ms is not long enough to be assured that the last CS interrupt has been delivered, so bump that to 200ms, and move the entirety of that wait to before we take the struct_mutex to avoid blocking. As this is now a potentially long wait, restore the earlier behaviour of bailing out early when a new request arrives. v2: Break out the repeated check for new requests into its own little helper to try and improve the self-commentary. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Cc: Imre Deak Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171023213237.26536-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 37 +++++++++++++++++++++++---------- 1 file changed, 26 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 026cb52ece0b7..bb0e85043e010 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3276,13 +3276,20 @@ i915_gem_retire_work_handler(struct work_struct *work) } } +static inline bool +new_requests_since_last_retire(const struct drm_i915_private *i915) +{ + return (READ_ONCE(i915->gt.active_requests) || + work_pending(&i915->gt.idle_work.work)); +} + static void i915_gem_idle_work_handler(struct work_struct *work) { struct drm_i915_private *dev_priv = container_of(work, typeof(*dev_priv), gt.idle_work.work); - struct drm_device *dev = &dev_priv->drm; bool rearm_hangcheck; + ktime_t end; if (!READ_ONCE(dev_priv->gt.awake)) return; @@ -3291,14 +3298,21 @@ i915_gem_idle_work_handler(struct work_struct *work) * Wait for last execlists context complete, but bail out in case a * new request is submitted. */ - wait_for(intel_engines_are_idle(dev_priv), 10); - if (READ_ONCE(dev_priv->gt.active_requests)) - return; + end = ktime_add_ms(ktime_get(), 200); + do { + if (new_requests_since_last_retire(dev_priv)) + return; + + if (intel_engines_are_idle(dev_priv)) + break; + + usleep_range(100, 500); + } while (ktime_before(ktime_get(), end)); rearm_hangcheck = cancel_delayed_work_sync(&dev_priv->gpu_error.hangcheck_work); - if (!mutex_trylock(&dev->struct_mutex)) { + if (!mutex_trylock(&dev_priv->drm.struct_mutex)) { /* Currently busy, come back later */ mod_delayed_work(dev_priv->wq, &dev_priv->gt.idle_work, @@ -3310,13 +3324,14 @@ i915_gem_idle_work_handler(struct work_struct *work) * New request retired after this work handler started, extend active * period until next instance of the work. */ - if (work_pending(work)) - goto out_unlock; - - if (dev_priv->gt.active_requests) + if (new_requests_since_last_retire(dev_priv)) goto out_unlock; - if (wait_for(intel_engines_are_idle(dev_priv), 10)) + /* + * We are committed now to parking the engines, make sure there + * will be no more interrupts arriving later. + */ + if (!intel_engines_are_idle(dev_priv)) DRM_ERROR("Timeout waiting for engines to idle\n"); intel_engines_mark_idle(dev_priv); @@ -3330,7 +3345,7 @@ i915_gem_idle_work_handler(struct work_struct *work) gen6_rps_idle(dev_priv); intel_runtime_pm_put(dev_priv); out_unlock: - mutex_unlock(&dev->struct_mutex); + mutex_unlock(&dev_priv->drm.struct_mutex); out_rearm: if (rearm_hangcheck) { -- GitLab From ff320d6e72ff3600184a1c01ea75b689b4b8a16d Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 23 Oct 2017 22:32:35 +0100 Subject: [PATCH 0007/1471] drm/i915: Synchronize irq before parking each engine When we park the engine (upon idling), we kill the irq tasklet. However, to be sure that it is not restarted by a spurious interrupt after doing so, flush the interrupt handler before parking. As we only park the engines when we believe the system is idle, there should not be any interrupts to distrub us; so flushing the final in-flight interrupt should be sufficient. (However, we are still dependent on the HW behaving in an orderly and timely fashion, which we shall endeavour to improve upon later.) Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Cc: Imre Deak Link: https://patchwork.freedesktop.org/patch/msgid/20171023213237.26536-2-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_gem.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bb0e85043e010..bafe1cdd57d8d 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3327,6 +3327,19 @@ i915_gem_idle_work_handler(struct work_struct *work) if (new_requests_since_last_retire(dev_priv)) goto out_unlock; + /* + * Be paranoid and flush a concurrent interrupt to make sure + * we don't reactivate any irq tasklets after parking. + * + * FIXME: Note that even though we have waited for execlists to be idle, + * there may still be an in-flight interrupt even though the CSB + * is now empty. synchronize_irq() makes sure that a residual interrupt + * is completed before we continue, but it doesn't prevent the HW from + * raising a spurious interrupt later. To complete the shield we should + * coordinate disabling the CS irq with flushing the interrupts. + */ + synchronize_irq(dev_priv->drm.irq); + /* * We are committed now to parking the engines, make sure there * will be no more interrupts arriving later. -- GitLab From 4a118ecbe99c93cf9f9582e83a88d03f18d6cb84 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 23 Oct 2017 22:32:36 +0100 Subject: [PATCH 0008/1471] drm/i915: Filter out spurious execlists context-switch interrupts Back in commit a4b2b01523a8 ("drm/i915: Don't mark an execlists context-switch when idle") we noticed the presence of late context-switch interrupts. We were able to filter those out by looking at whether the ELSP remained active, but in commit beecec901790 ("drm/i915/execlists: Preemption!") that became problematic as we now anticipate receiving a context-switch event for preemption while ELSP may be empty. To restore the spurious interrupt suppression, add a counter for the expected number of pending context-switches and skip if we do not need to handle this interrupt to make forward progress. v2: Don't forget to switch on for preempt. v3: Reduce the counter to a on/off boolean tracker. Declare the HW as active when we first submit, and idle after the final completion event (with which we confirm the HW says it is idle), and track each source of activity separately. With a finite number of sources, it should aide us in debugging which gets stuck. Fixes: beecec901790 ("drm/i915/execlists: Preemption!") Signed-off-by: Chris Wilson Cc: Michal Winiarski Cc: Tvrtko Ursulin Cc: Arkadiusz Hiler Cc: Mika Kuoppala Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171023213237.26536-3-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_guc_submission.c | 3 ++ drivers/gpu/drm/i915/i915_irq.c | 6 ++-- drivers/gpu/drm/i915/intel_engine_cs.c | 5 ++-- drivers/gpu/drm/i915/intel_lrc.c | 27 ++++++++++++----- drivers/gpu/drm/i915/intel_ringbuffer.h | 34 ++++++++++++++++++++-- 5 files changed, 62 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index a2e8114b739d9..f84c267728fda 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -610,6 +610,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) execlists->first = rb; if (submit) { port_assign(port, last); + execlists_set_active(execlists, EXECLISTS_ACTIVE_USER); i915_guc_submit(engine); } spin_unlock_irq(&engine->timeline->lock); @@ -633,6 +634,8 @@ static void i915_guc_irq_handler(unsigned long data) rq = port_request(&port[0]); } + if (!rq) + execlists_clear_active(execlists, EXECLISTS_ACTIVE_USER); if (!port_isset(last_port)) i915_guc_dequeue(engine); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index b1296a55c1e4a..f8205841868bc 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1388,8 +1388,10 @@ gen8_cs_irq_handler(struct intel_engine_cs *engine, u32 iir, int test_shift) bool tasklet = false; if (iir & (GT_CONTEXT_SWITCH_INTERRUPT << test_shift)) { - __set_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted); - tasklet = true; + if (READ_ONCE(engine->execlists.active)) { + __set_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted); + tasklet = true; + } } if (iir & (GT_RENDER_USER_INTERRUPT << test_shift)) { diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index a47a9c6bea521..ab5bf4e2e28e2 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1548,8 +1548,8 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine) if (test_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted)) return false; - /* Both ports drained, no more ELSP submission? */ - if (port_request(&engine->execlists.port[0])) + /* Waiting to drain ELSP? */ + if (READ_ONCE(engine->execlists.active)) return false; /* ELSP is empty, but there are ready requests? */ @@ -1749,6 +1749,7 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) idx); } } + drm_printf(m, "\t\tHW active? 0x%x\n", execlists->active); rcu_read_unlock(); } else if (INTEL_GEN(dev_priv) > 6) { drm_printf(m, "\tPP_DIR_BASE: 0x%08x\n", diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 7f45dd7dc3e50..5b87d5284a844 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -575,7 +575,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine) * the state of the GPU is known (idle). */ inject_preempt_context(engine); - execlists->preempt = true; + execlists_set_active(execlists, + EXECLISTS_ACTIVE_PREEMPT); goto unlock; } else { /* @@ -683,8 +684,10 @@ static void execlists_dequeue(struct intel_engine_cs *engine) unlock: spin_unlock_irq(&engine->timeline->lock); - if (submit) + if (submit) { + execlists_set_active(execlists, EXECLISTS_ACTIVE_USER); execlists_submit_ports(engine); + } } static void @@ -696,6 +699,7 @@ execlist_cancel_port_requests(struct intel_engine_execlists *execlists) while (num_ports-- && port_isset(port)) { struct drm_i915_gem_request *rq = port_request(port); + GEM_BUG_ON(!execlists->active); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_PREEMPTED); i915_gem_request_put(rq); @@ -861,15 +865,21 @@ static void intel_lrc_irq_handler(unsigned long data) unwind_incomplete_requests(engine); spin_unlock_irq(&engine->timeline->lock); - GEM_BUG_ON(!execlists->preempt); - execlists->preempt = false; + GEM_BUG_ON(!execlists_is_active(execlists, + EXECLISTS_ACTIVE_PREEMPT)); + execlists_clear_active(execlists, + EXECLISTS_ACTIVE_PREEMPT); continue; } if (status & GEN8_CTX_STATUS_PREEMPTED && - execlists->preempt) + execlists_is_active(execlists, + EXECLISTS_ACTIVE_PREEMPT)) continue; + GEM_BUG_ON(!execlists_is_active(execlists, + EXECLISTS_ACTIVE_USER)); + /* Check the context/desc id for this event matches */ GEM_DEBUG_BUG_ON(buf[2 * head + 1] != port->context_id); @@ -892,6 +902,9 @@ static void intel_lrc_irq_handler(unsigned long data) /* After the final element, the hw should be idle */ GEM_BUG_ON(port_count(port) == 0 && !(status & GEN8_CTX_STATUS_ACTIVE_IDLE)); + if (port_count(port) == 0) + execlists_clear_active(execlists, + EXECLISTS_ACTIVE_USER); } if (head != execlists->csb_head) { @@ -901,7 +914,7 @@ static void intel_lrc_irq_handler(unsigned long data) } } - if (!execlists->preempt) + if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT)) execlists_dequeue(engine); intel_uncore_forcewake_put(dev_priv, execlists->fw_domains); @@ -1460,7 +1473,7 @@ static int gen8_init_common_ring(struct intel_engine_cs *engine) GT_CONTEXT_SWITCH_INTERRUPT << engine->irq_shift); clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted); execlists->csb_head = -1; - execlists->preempt = false; + execlists->active = 0; /* After a GPU reset, we may have requests to replay */ if (!i915_modparams.enable_guc_submission && execlists->first) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 17186f0674085..6a42ed618a28c 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -241,9 +241,17 @@ struct intel_engine_execlists { } port[EXECLIST_MAX_PORTS]; /** - * @preempt: are we currently handling a preempting context switch? + * @active: is the HW active? We consider the HW as active after + * submitting any context for execution and until we have seen the + * last context completion event. After that, we do not expect any + * more events until we submit, and so can park the HW. + * + * As we have a small number of different sources from which we feed + * the HW, we track the state of each inside a single bitfield. */ - bool preempt; + unsigned int active; +#define EXECLISTS_ACTIVE_USER 0 +#define EXECLISTS_ACTIVE_PREEMPT 1 /** * @port_mask: number of execlist ports - 1 @@ -525,6 +533,27 @@ struct intel_engine_cs { u32 (*get_cmd_length_mask)(u32 cmd_header); }; +static inline void +execlists_set_active(struct intel_engine_execlists *execlists, + unsigned int bit) +{ + __set_bit(bit, (unsigned long *)&execlists->active); +} + +static inline void +execlists_clear_active(struct intel_engine_execlists *execlists, + unsigned int bit) +{ + __clear_bit(bit, (unsigned long *)&execlists->active); +} + +static inline bool +execlists_is_active(const struct intel_engine_execlists *execlists, + unsigned int bit) +{ + return test_bit(bit, (unsigned long *)&execlists->active); +} + static inline unsigned int execlists_num_ports(const struct intel_engine_execlists * const execlists) { @@ -538,6 +567,7 @@ execlists_port_complete(struct intel_engine_execlists * const execlists, const unsigned int m = execlists->port_mask; GEM_BUG_ON(port_index(port, execlists) != 0); + GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_USER)); memmove(port, port + 1, m * sizeof(struct execlist_port)); memset(port + m, 0, sizeof(struct execlist_port)); -- GitLab From 64b80085dd3603d401fc05879f700b86a3a5c8e8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 24 Oct 2017 12:55:01 +0100 Subject: [PATCH 0009/1471] drm/i915/execlists: Remove the priority "optimisation" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Originally we set the priority to max upon inserting the request into the execlists queue (and removing it from the scheduler lists). We could then use the prio==INT_MAX as a shortcut within execlists_schedule() to detect the end of the dependency chain. Since commit 1f181225f8ec ("drm/i915/execlists: Keep request->priority for its lifetime") this is no longer true as we use the request completion as an indicator the schedule dependency chain is complete instead. (This allows us to then reschedule requests even when its context is in flight.) However, this makes the GEM_BUG_ON() inside execlists_schedule() racy as we may change the rq->prio at the same time. As the assertion is useful, let's keep the assertion and remove the micro-optimisation. Fixes: 1f181225f8ec ("drm/i915/execlists: Keep request->priority for its lifetime") Signed-off-by: Chris Wilson Cc: Michał Winiarski Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171024115501.21033-1-chris@chris-wilson.co.uk Reviewed-by: Michał Winiarski --- drivers/gpu/drm/i915/intel_lrc.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 5b87d5284a844..d36e256074354 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -734,7 +734,6 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) list_for_each_entry_safe(rq, rn, &p->requests, priotree.link) { INIT_LIST_HEAD(&rq->priotree.link); - rq->priotree.priority = INT_MAX; dma_fence_set_error(&rq->fence, -EIO); __i915_gem_request_submit(rq); @@ -891,7 +890,6 @@ static void intel_lrc_irq_handler(unsigned long data) execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); trace_i915_gem_request_out(rq); - rq->priotree.priority = INT_MAX; i915_gem_request_put(rq); execlists_port_complete(execlists, port); -- GitLab From 191f896085cf3b5d85920d58a759da4eea141721 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Tue, 24 Oct 2017 16:27:28 +0100 Subject: [PATCH 0010/1471] drm/i915/perf: fix perf enable/disable ioctls with 32bits userspace The compat callback was missing and triggered failures in 32bits userspace when enabling/disable the perf stream. We don't require any particular processing here as these ioctls don't take any argument. Signed-off-by: Lionel Landwerlin Fixes: eec688e1420 ("drm/i915: Add i915 perf infrastructure") Cc: linux-stable Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171024152728.4873-1-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_perf.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 1383a2995a698..59ee808f8fd97 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -2537,6 +2537,10 @@ static const struct file_operations fops = { .poll = i915_perf_poll, .read = i915_perf_read, .unlocked_ioctl = i915_perf_ioctl, + /* Our ioctl have no arguments, so it's safe to use the same function + * to handle 32bits compatibility. + */ + .compat_ioctl = i915_perf_ioctl, }; -- GitLab From 436009b578ab872619350a4c4f2b36c98c7e55a2 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Mon, 23 Oct 2017 10:39:20 -0700 Subject: [PATCH 0011/1471] drm/i915/cnl: Force DDI_A_4_LANES when needed. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As we faced in BXT, on CNL DDI_A_4_LANES is not set as expected when system is boot with multiple monitors connected. This result in wrong lane setup impacting the max data rate available and consequently blocking modeset on eDP, resulting in a blank screen. Most of CNL SKUs don't support DDI-E. The only SKU that supports DDI-E is the same that supports the full A/E split called DDI-F. Also when DDI-F is used DDI-E cannot be used because they share Interrupts. So DDI-E is almost useless. Anyways let's consider this is possible and rely on VBT for that. This patch was initialy start by Clint, but required many changes including full commit message. So Credits entirely to Clint for finding this. v2: Extract all messy conditions into a helper function as suggested by Ville. Along with simplification I removed the debug message on the working case since now all conditions are grouped. v3: Split the conditions even more as suggested by Ville. Get's cleaner and easier to add new cases in the future. Suggested-by: Clint Taylor Cc: Clint Taylor Cc: Mika Kahola Cc: Jani Nikula Cc: Ville Syrjälä Signed-off-by: Rodrigo Vivi Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171023173920.22890-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_ddi.c | 46 ++++++++++++++++++++++++-------- 1 file changed, 35 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 933c18fd4258a..08f04df08fa53 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2711,6 +2711,34 @@ intel_ddi_init_hdmi_connector(struct intel_digital_port *intel_dig_port) return connector; } +static bool intel_ddi_a_force_4_lanes(struct intel_digital_port *dport) +{ + struct drm_i915_private *dev_priv = to_i915(dport->base.base.dev); + + if (dport->port != PORT_A) + return false; + + if (dport->saved_port_bits & DDI_A_4_LANES) + return false; + + /* Broxton/Geminilake: Bspec says that DDI_A_4_LANES is the only + * supported configuration + */ + if (IS_GEN9_LP(dev_priv)) + return true; + + /* Cannonlake: Most of SKUs don't support DDI_E, and the only + * one who does also have a full A/E split called + * DDI_F what makes DDI_E useless. However for this + * case let's trust VBT info. + */ + if (IS_CANNONLAKE(dev_priv) && + !intel_bios_is_port_present(dev_priv, PORT_E)) + return true; + + return false; +} + void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) { struct intel_digital_port *intel_dig_port; @@ -2820,18 +2848,14 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) } /* - * Bspec says that DDI_A_4_LANES is the only supported configuration - * for Broxton. Yet some BIOS fail to set this bit on port A if eDP - * wasn't lit up at boot. Force this bit on in our internal - * configuration so that we use the proper lane count for our - * calculations. + * Some BIOS might fail to set this bit on port A if eDP + * wasn't lit up at boot. Force this bit set when needed + * so we use the proper lane count for our calculations. */ - if (IS_GEN9_LP(dev_priv) && port == PORT_A) { - if (!(intel_dig_port->saved_port_bits & DDI_A_4_LANES)) { - DRM_DEBUG_KMS("BXT BIOS forgot to set DDI_A_4_LANES for port A; fixing\n"); - intel_dig_port->saved_port_bits |= DDI_A_4_LANES; - max_lanes = 4; - } + if (intel_ddi_a_force_4_lanes(intel_dig_port)) { + DRM_DEBUG_KMS("Forcing DDI_A_4_LANES for port A\n"); + intel_dig_port->saved_port_bits |= DDI_A_4_LANES; + max_lanes = 4; } intel_dig_port->max_lanes = max_lanes; -- GitLab From 0aab201b4ad10fa530e4d12f8ea30b6f0a0540bd Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Mon, 23 Oct 2017 15:46:12 -0700 Subject: [PATCH 0012/1471] drm/i915/cnl: Get RC6 working. On CNL, individual wake rate limit was added to each engine. GT can only go to RC6 if both Render and Media engines are individually qualified. So we need to set their individual wake rate limit. +-----------------+---------------+--------------+--------------+ | | GT RC6 | Render C6 | Media C6 | +-----------------+---------------+--------------+--------------+ | Wake rate limit | 0xA09C[31:16] | 0xA09C[15:0] | 0xA0A0[15:0] | +-----------------+---------------+--------------+--------------+ v2: - Tune Render and Media wake rate values according to some extra info I got from HW engineers. Value can be tuned, but for now these are the recommended values. - Fix typos pointed by James. Cc: Nathan Ciobanu Cc: Wayne Boyer Cc: Joe Konno Cc: David Weinehall Signed-off-by: Rodrigo Vivi Reviewed-by: James Ausmus Reviewed-by: David Weinehall Link: https://patchwork.freedesktop.org/patch/msgid/20171023224612.27208-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_pm.c | 15 +++++++++++---- 2 files changed, 12 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 68a58cce6ab1c..f138eae82bf0c 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -7905,6 +7905,7 @@ enum { #define GEN6_RC1_WAKE_RATE_LIMIT _MMIO(0xA098) #define GEN6_RC6_WAKE_RATE_LIMIT _MMIO(0xA09C) #define GEN6_RC6pp_WAKE_RATE_LIMIT _MMIO(0xA0A0) +#define GEN10_MEDIA_WAKE_RATE_LIMIT _MMIO(0xA0A0) #define GEN6_RC_EVALUATION_INTERVAL _MMIO(0xA0A8) #define GEN6_RC_IDLE_HYSTERSIS _MMIO(0xA0AC) #define GEN6_RC_SLEEP _MMIO(0xA0B0) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index c42a65a93b3a9..1ead51754e8f5 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6604,12 +6604,19 @@ static void gen9_enable_rc6(struct drm_i915_private *dev_priv) I915_WRITE(GEN6_RC_CONTROL, 0); /* 2b: Program RC6 thresholds.*/ - - /* WaRsDoubleRc6WrlWithCoarsePowerGating: Doubling WRL only when CPG is enabled */ - if (IS_SKYLAKE(dev_priv)) + if (INTEL_GEN(dev_priv) >= 10) { + I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); + I915_WRITE(GEN10_MEDIA_WAKE_RATE_LIMIT, 150); + } else if (IS_SKYLAKE(dev_priv)) { + /* + * WaRsDoubleRc6WrlWithCoarsePowerGating:skl Doubling WRL only + * when CPG is enabled + */ I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 108 << 16); - else + } else { I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16); + } + I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ for_each_engine(engine, dev_priv, id) -- GitLab From b40c88fda1300b2dd65fcc3377d54b724edf47e6 Mon Sep 17 00:00:00 2001 From: Anusha Srivatsa Date: Tue, 3 Oct 2017 10:59:48 -0700 Subject: [PATCH 0013/1471] drm/i915/cnl: Update the DMC version on CNL The latest version of DMC on CNL is 1.06. Update the version so as to load the latest firmware. Release Notes: Version: 1.06 1. DDI and AUX IO related fix. v2: Improve the prefixes in commit message. Add Release Notes directly. (Rodrigo) Cc: Rodrigo Vivi Signed-off-by: Anusha Srivatsa Reviewed-by: Rodrigo Vivi Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/1507053588-677-1-git-send-email-anusha.srivatsa@intel.com --- drivers/gpu/drm/i915/intel_csr.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c index da9de47562b80..3e1f86d0c6cc2 100644 --- a/drivers/gpu/drm/i915/intel_csr.c +++ b/drivers/gpu/drm/i915/intel_csr.c @@ -37,8 +37,8 @@ #define I915_CSR_GLK "i915/glk_dmc_ver1_04.bin" #define GLK_CSR_VERSION_REQUIRED CSR_VERSION(1, 4) -#define I915_CSR_CNL "i915/cnl_dmc_ver1_04.bin" -#define CNL_CSR_VERSION_REQUIRED CSR_VERSION(1, 4) +#define I915_CSR_CNL "i915/cnl_dmc_ver1_06.bin" +#define CNL_CSR_VERSION_REQUIRED CSR_VERSION(1, 6) #define I915_CSR_KBL "i915/kbl_dmc_ver1_01.bin" MODULE_FIRMWARE(I915_CSR_KBL); -- GitLab From 2b58417ffbca9fafa1d54b9f1272965f98f456ca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:07 +0300 Subject: [PATCH 0014/1471] drm/i915: Clean up some cdclk switch statements MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Redo some switch statements in the cdclk code to use a common fall through for the default case. Makes everything look a bit more uniform Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Mika Kahola Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-2-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 68 +++++++++++++++--------------- 1 file changed, 34 insertions(+), 34 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index b2a6d62b71c04..4bffd31a89249 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -681,6 +681,13 @@ static void bdw_set_cdclk(struct drm_i915_private *dev_priv, val &= ~LCPLL_CLK_FREQ_MASK; switch (cdclk) { + default: + MISSING_CASE(cdclk); + /* fall through */ + case 337500: + val |= LCPLL_CLK_FREQ_337_5_BDW; + data = 2; + break; case 450000: val |= LCPLL_CLK_FREQ_450; data = 0; @@ -689,17 +696,10 @@ static void bdw_set_cdclk(struct drm_i915_private *dev_priv, val |= LCPLL_CLK_FREQ_54O_BDW; data = 1; break; - case 337500: - val |= LCPLL_CLK_FREQ_337_5_BDW; - data = 2; - break; case 675000: val |= LCPLL_CLK_FREQ_675_BDW; data = 3; break; - default: - WARN(1, "invalid cdclk frequency\n"); - return; } I915_WRITE(LCPLL_CTL, val); @@ -926,8 +926,6 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, u32 freq_select, pcu_ack; int ret; - WARN_ON((cdclk == 24000) != (vco == 0)); - mutex_lock(&dev_priv->pcu_lock); ret = skl_pcode_request(dev_priv, SKL_PCODE_CDCLK_CONTROL, SKL_CDCLK_PREPARE_FOR_CHANGE, @@ -942,6 +940,15 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, /* set CDCLK_CTL */ switch (cdclk) { + default: + WARN_ON(cdclk != dev_priv->cdclk.hw.ref); + WARN_ON(vco != 0); + /* fall through */ + case 308571: + case 337500: + freq_select = CDCLK_FREQ_337_308; + pcu_ack = 0; + break; case 450000: case 432000: freq_select = CDCLK_FREQ_450_432; @@ -951,12 +958,6 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, freq_select = CDCLK_FREQ_540; pcu_ack = 2; break; - case 308571: - case 337500: - default: - freq_select = CDCLK_FREQ_337_308; - pcu_ack = 0; - break; case 617143: case 675000: freq_select = CDCLK_FREQ_675_617; @@ -1110,6 +1111,7 @@ static int bxt_de_pll_vco(struct drm_i915_private *dev_priv, int cdclk) switch (cdclk) { default: MISSING_CASE(cdclk); + /* fall through */ case 144000: case 288000: case 384000: @@ -1134,6 +1136,7 @@ static int glk_de_pll_vco(struct drm_i915_private *dev_priv, int cdclk) switch (cdclk) { default: MISSING_CASE(cdclk); + /* fall through */ case 79200: case 158400: case 316800: @@ -1246,24 +1249,22 @@ static void bxt_set_cdclk(struct drm_i915_private *dev_priv, /* cdclk = vco / 2 / div{1,1.5,2,4} */ switch (DIV_ROUND_CLOSEST(vco, cdclk)) { - case 8: - divider = BXT_CDCLK_CD2X_DIV_SEL_4; - break; - case 4: - divider = BXT_CDCLK_CD2X_DIV_SEL_2; + default: + WARN_ON(cdclk != dev_priv->cdclk.hw.ref); + WARN_ON(vco != 0); + /* fall through */ + case 2: + divider = BXT_CDCLK_CD2X_DIV_SEL_1; break; case 3: WARN(IS_GEMINILAKE(dev_priv), "Unsupported divider\n"); divider = BXT_CDCLK_CD2X_DIV_SEL_1_5; break; - case 2: - divider = BXT_CDCLK_CD2X_DIV_SEL_1; + case 4: + divider = BXT_CDCLK_CD2X_DIV_SEL_2; break; - default: - WARN_ON(cdclk != dev_priv->cdclk.hw.ref); - WARN_ON(vco != 0); - - divider = BXT_CDCLK_CD2X_DIV_SEL_1; + case 8: + divider = BXT_CDCLK_CD2X_DIV_SEL_4; break; } @@ -1532,18 +1533,16 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, /* cdclk = vco / 2 / div{1,2} */ switch (DIV_ROUND_CLOSEST(vco, cdclk)) { - case 4: - divider = BXT_CDCLK_CD2X_DIV_SEL_2; - break; - case 2: - divider = BXT_CDCLK_CD2X_DIV_SEL_1; - break; default: WARN_ON(cdclk != dev_priv->cdclk.hw.ref); WARN_ON(vco != 0); - + /* fall through */ + case 2: divider = BXT_CDCLK_CD2X_DIV_SEL_1; break; + case 4: + divider = BXT_CDCLK_CD2X_DIV_SEL_2; + break; } switch (cdclk) { @@ -1592,6 +1591,7 @@ static int cnl_cdclk_pll_vco(struct drm_i915_private *dev_priv, int cdclk) switch (cdclk) { default: MISSING_CASE(cdclk); + /* fall through */ case 168000: case 336000: ratio = dev_priv->cdclk.hw.ref == 19200 ? 35 : 28; -- GitLab From 64600bd5b8280208dfbd33bcd7a5f32255058f07 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:08 +0300 Subject: [PATCH 0015/1471] drm/i915: Start tracking voltage level in the cdclk state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit For CNL we'll need to start considering the port clocks when we select the voltage level for the system agent. To that end start tracking the voltage in the cdclk state (since that already has to adjust it). v2: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-3-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/intel_cdclk.c | 31 +++++++++++++++++++------ drivers/gpu/drm/i915/intel_display.c | 8 +++---- drivers/gpu/drm/i915/intel_drv.h | 4 +++- drivers/gpu/drm/i915/intel_runtime_pm.c | 3 ++- 5 files changed, 34 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 54b5d4c582b61..2c64c6680ac73 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2228,6 +2228,7 @@ struct i915_oa_ops { struct intel_cdclk_state { unsigned int cdclk, vco, ref; + u8 voltage_level; }; struct drm_i915_private { diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 4bffd31a89249..6dad2efd5e276 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1690,17 +1690,34 @@ void cnl_uninit_cdclk(struct drm_i915_private *dev_priv) } /** - * intel_cdclk_state_compare - Determine if two CDCLK states differ + * intel_cdclk_needs_modeset - Determine if two CDCLK states require a modeset on all pipes * @a: first CDCLK state * @b: second CDCLK state * * Returns: - * True if the CDCLK states are identical, false if they differ. + * True if the CDCLK states require pipes to be off during reprogramming, false if not. */ -bool intel_cdclk_state_compare(const struct intel_cdclk_state *a, +bool intel_cdclk_needs_modeset(const struct intel_cdclk_state *a, const struct intel_cdclk_state *b) { - return memcmp(a, b, sizeof(*a)) == 0; + return a->cdclk != b->cdclk || + a->vco != b->vco || + a->ref != b->ref; +} + +/** + * intel_cdclk_changed - Determine if two CDCLK states are different + * @a: first CDCLK state + * @b: second CDCLK state + * + * Returns: + * True if the CDCLK states don't match, false if they do. + */ +bool intel_cdclk_changed(const struct intel_cdclk_state *a, + const struct intel_cdclk_state *b) +{ + return intel_cdclk_needs_modeset(a, b) || + a->voltage_level != b->voltage_level; } /** @@ -1714,15 +1731,15 @@ bool intel_cdclk_state_compare(const struct intel_cdclk_state *a, void intel_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state) { - if (intel_cdclk_state_compare(&dev_priv->cdclk.hw, cdclk_state)) + if (!intel_cdclk_changed(&dev_priv->cdclk.hw, cdclk_state)) return; if (WARN_ON_ONCE(!dev_priv->display.set_cdclk)) return; - DRM_DEBUG_DRIVER("Changing CDCLK to %d kHz, VCO %d kHz, ref %d kHz\n", + DRM_DEBUG_DRIVER("Changing CDCLK to %d kHz, VCO %d kHz, ref %d kHz, voltage_level %d\n", cdclk_state->cdclk, cdclk_state->vco, - cdclk_state->ref); + cdclk_state->ref, cdclk_state->voltage_level); dev_priv->display.set_cdclk(dev_priv, cdclk_state); } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index e2ac976844d80..1d2031a5f4bb2 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11931,16 +11931,16 @@ static int intel_modeset_checks(struct drm_atomic_state *state) * holding all the crtc locks, even if we don't end up * touching the hardware */ - if (!intel_cdclk_state_compare(&dev_priv->cdclk.logical, - &intel_state->cdclk.logical)) { + if (intel_cdclk_changed(&dev_priv->cdclk.logical, + &intel_state->cdclk.logical)) { ret = intel_lock_all_pipes(state); if (ret < 0) return ret; } /* All pipes must be switched off while we change the cdclk. */ - if (!intel_cdclk_state_compare(&dev_priv->cdclk.actual, - &intel_state->cdclk.actual)) { + if (intel_cdclk_needs_modeset(&dev_priv->cdclk.actual, + &intel_state->cdclk.actual)) { ret = intel_modeset_all_pipes(state); if (ret < 0) return ret; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 47d022d487181..a551aadb157bf 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1323,8 +1323,10 @@ void intel_init_cdclk_hooks(struct drm_i915_private *dev_priv); void intel_update_max_cdclk(struct drm_i915_private *dev_priv); void intel_update_cdclk(struct drm_i915_private *dev_priv); void intel_update_rawclk(struct drm_i915_private *dev_priv); -bool intel_cdclk_state_compare(const struct intel_cdclk_state *a, +bool intel_cdclk_needs_modeset(const struct intel_cdclk_state *a, const struct intel_cdclk_state *b); +bool intel_cdclk_changed(const struct intel_cdclk_state *a, + const struct intel_cdclk_state *b); void intel_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state); diff --git a/drivers/gpu/drm/i915/intel_runtime_pm.c b/drivers/gpu/drm/i915/intel_runtime_pm.c index 8af286c63d3b6..8315499452dc9 100644 --- a/drivers/gpu/drm/i915/intel_runtime_pm.c +++ b/drivers/gpu/drm/i915/intel_runtime_pm.c @@ -705,7 +705,8 @@ static void gen9_dc_off_power_well_enable(struct drm_i915_private *dev_priv, gen9_set_dc_state(dev_priv, DC_STATE_DISABLE); dev_priv->display.get_cdclk(dev_priv, &cdclk_state); - WARN_ON(!intel_cdclk_state_compare(&dev_priv->cdclk.hw, &cdclk_state)); + /* Can't read out voltage_level so can't use intel_cdclk_changed() */ + WARN_ON(intel_cdclk_needs_modeset(&dev_priv->cdclk.hw, &cdclk_state)); gen9_assert_dbuf_enabled(dev_priv); -- GitLab From 999c5766f39e3a21aeeb40b5f00ee8d3b48c4331 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:09 +0300 Subject: [PATCH 0016/1471] drm/i915: Use cdclk_state->voltage on VLV/CHV MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Store the punit DSPFREQUAR value into cdclk_state->voltage on VLV/CHV. Since we can actually read that out from the hardware this can give us a bit more cross checking between the hardware and software state. v2: Don't break waiting for cdclk change on VLV/CHV v3: Split out the cdclk sanity check in vlv_set_cdclk() (Rodrigo) v4: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-4-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 54 +++++++++++++++++++++--------- 1 file changed, 38 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 6dad2efd5e276..8bc0930825118 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -437,13 +437,45 @@ static int vlv_calc_cdclk(struct drm_i915_private *dev_priv, int min_cdclk) return 200000; } +static u8 vlv_calc_voltage_level(struct drm_i915_private *dev_priv, int cdclk) +{ + if (IS_VALLEYVIEW(dev_priv)) { + if (cdclk >= 320000) /* jump to highest voltage for 400MHz too */ + return 2; + else if (cdclk >= 266667) + return 1; + else + return 0; + } else { + /* + * Specs are full of misinformation, but testing on actual + * hardware has shown that we just need to write the desired + * CCK divider into the Punit register. + */ + return DIV_ROUND_CLOSEST(dev_priv->hpll_freq << 1, cdclk) - 1; + } +} + static void vlv_get_cdclk(struct drm_i915_private *dev_priv, struct intel_cdclk_state *cdclk_state) { + u32 val; + cdclk_state->vco = vlv_get_hpll_vco(dev_priv); cdclk_state->cdclk = vlv_get_cck_clock(dev_priv, "cdclk", CCK_DISPLAY_CLOCK_CONTROL, cdclk_state->vco); + + mutex_lock(&dev_priv->pcu_lock); + val = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ); + mutex_unlock(&dev_priv->pcu_lock); + + if (IS_VALLEYVIEW(dev_priv)) + cdclk_state->voltage_level = (val & DSPFREQGUAR_MASK) >> + DSPFREQGUAR_SHIFT; + else + cdclk_state->voltage_level = (val & DSPFREQGUAR_MASK_CHV) >> + DSPFREQGUAR_SHIFT_CHV; } static void vlv_program_pfi_credits(struct drm_i915_private *dev_priv) @@ -486,7 +518,7 @@ static void vlv_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state) { int cdclk = cdclk_state->cdclk; - u32 val, cmd; + u32 val, cmd = cdclk_state->voltage_level; /* There are cases where we can end up here with power domains * off and a CDCLK frequency other than the minimum, like when @@ -496,13 +528,6 @@ static void vlv_set_cdclk(struct drm_i915_private *dev_priv, */ intel_display_power_get(dev_priv, POWER_DOMAIN_PIPE_A); - if (cdclk >= 320000) /* jump to highest voltage for 400MHz too */ - cmd = 2; - else if (cdclk == 266667) - cmd = 1; - else - cmd = 0; - mutex_lock(&dev_priv->pcu_lock); val = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ); val &= ~DSPFREQGUAR_MASK; @@ -562,7 +587,7 @@ static void chv_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state) { int cdclk = cdclk_state->cdclk; - u32 val, cmd; + u32 val, cmd = cdclk_state->voltage_level; switch (cdclk) { case 333333: @@ -583,13 +608,6 @@ static void chv_set_cdclk(struct drm_i915_private *dev_priv, */ intel_display_power_get(dev_priv, POWER_DOMAIN_PIPE_A); - /* - * Specs are full of misinformation, but testing on actual - * hardware has shown that we just need to write the desired - * CCK divider into the Punit register. - */ - cmd = DIV_ROUND_CLOSEST(dev_priv->hpll_freq << 1, cdclk) - 1; - mutex_lock(&dev_priv->pcu_lock); val = vlv_punit_read(dev_priv, PUNIT_REG_DSPFREQ); val &= ~DSPFREQGUAR_MASK_CHV; @@ -1859,11 +1877,15 @@ static int vlv_modeset_calc_cdclk(struct drm_atomic_state *state) cdclk = vlv_calc_cdclk(dev_priv, min_cdclk); intel_state->cdclk.logical.cdclk = cdclk; + intel_state->cdclk.logical.voltage_level = + vlv_calc_voltage_level(dev_priv, cdclk); if (!intel_state->active_crtcs) { cdclk = vlv_calc_cdclk(dev_priv, 0); intel_state->cdclk.actual.cdclk = cdclk; + intel_state->cdclk.actual.voltage_level = + vlv_calc_voltage_level(dev_priv, cdclk); } else { intel_state->cdclk.actual = intel_state->cdclk.logical; -- GitLab From d7ffaeef96f39870fec47f79b9a81a501737b01c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:10 +0300 Subject: [PATCH 0017/1471] drm/i915: Use cdclk_state->voltage on BDW MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Track the system agent voltage we request from pcode in the cdclk state on BDW. Annoyingly we can't actually read out the current value since there's no pcode command to do that, so we'll have to just assume that it worked. v2: Keep the WARN_ON (Rodrigo) v3: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-5-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 35 +++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 8bc0930825118..c98fec333dd1f 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -639,6 +639,21 @@ static int bdw_calc_cdclk(int min_cdclk) return 337500; } +static u8 bdw_calc_voltage_level(int cdclk) +{ + switch (cdclk) { + default: + case 337500: + return 2; + case 450000: + return 0; + case 540000: + return 1; + case 675000: + return 3; + } +} + static void bdw_get_cdclk(struct drm_i915_private *dev_priv, struct intel_cdclk_state *cdclk_state) { @@ -657,13 +672,20 @@ static void bdw_get_cdclk(struct drm_i915_private *dev_priv, cdclk_state->cdclk = 337500; else cdclk_state->cdclk = 675000; + + /* + * Can't read this out :( Let's assume it's + * at least what the CDCLK frequency requires. + */ + cdclk_state->voltage_level = + bdw_calc_voltage_level(cdclk_state->cdclk); } static void bdw_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state) { int cdclk = cdclk_state->cdclk; - uint32_t val, data; + uint32_t val; int ret; if (WARN((I915_READ(LCPLL_CTL) & @@ -704,19 +726,15 @@ static void bdw_set_cdclk(struct drm_i915_private *dev_priv, /* fall through */ case 337500: val |= LCPLL_CLK_FREQ_337_5_BDW; - data = 2; break; case 450000: val |= LCPLL_CLK_FREQ_450; - data = 0; break; case 540000: val |= LCPLL_CLK_FREQ_54O_BDW; - data = 1; break; case 675000: val |= LCPLL_CLK_FREQ_675_BDW; - data = 3; break; } @@ -731,7 +749,8 @@ static void bdw_set_cdclk(struct drm_i915_private *dev_priv, DRM_ERROR("Switching back to LCPLL failed\n"); mutex_lock(&dev_priv->pcu_lock); - sandybridge_pcode_write(dev_priv, HSW_PCODE_DE_WRITE_FREQ_REQ, data); + sandybridge_pcode_write(dev_priv, HSW_PCODE_DE_WRITE_FREQ_REQ, + cdclk_state->voltage_level); mutex_unlock(&dev_priv->pcu_lock); I915_WRITE(CDCLK_FREQ, DIV_ROUND_CLOSEST(cdclk, 1000) - 1); @@ -1910,11 +1929,15 @@ static int bdw_modeset_calc_cdclk(struct drm_atomic_state *state) cdclk = bdw_calc_cdclk(min_cdclk); intel_state->cdclk.logical.cdclk = cdclk; + intel_state->cdclk.logical.voltage_level = + bdw_calc_voltage_level(cdclk); if (!intel_state->active_crtcs) { cdclk = bdw_calc_cdclk(0); intel_state->cdclk.actual.cdclk = cdclk; + intel_state->cdclk.actual.voltage_level = + bdw_calc_voltage_level(cdclk); } else { intel_state->cdclk.actual = intel_state->cdclk.logical; -- GitLab From 2aa97491da8aa237e792079541ee2fa57ae9d960 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:11 +0300 Subject: [PATCH 0018/1471] drm/i915: Use cdclk_state->voltage on SKL/KBL/CFL MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Track the system agent voltage we request from pcode in the cdclk state on SKL/KBL/CFL. Annoyingly we can't actually read out the current value since there's no pcode command to do that, so we'll have to just assume that it worked. v2: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-6-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 43 +++++++++++++++++++++++++----- 1 file changed, 36 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index c98fec333dd1f..38e31df265bda 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -785,6 +785,24 @@ static int skl_calc_cdclk(int min_cdclk, int vco) } } +static u8 skl_calc_voltage_level(int cdclk) +{ + switch (cdclk) { + default: + case 308571: + case 337500: + return 0; + case 450000: + case 432000: + return 1; + case 540000: + return 2; + case 617143: + case 675000: + return 3; + } +} + static void skl_dpll0_update(struct drm_i915_private *dev_priv, struct intel_cdclk_state *cdclk_state) { @@ -835,7 +853,7 @@ static void skl_get_cdclk(struct drm_i915_private *dev_priv, cdclk_state->cdclk = cdclk_state->ref; if (cdclk_state->vco == 0) - return; + goto out; cdctl = I915_READ(CDCLK_CTL); @@ -876,6 +894,14 @@ static void skl_get_cdclk(struct drm_i915_private *dev_priv, break; } } + + out: + /* + * Can't read this out :( Let's assume it's + * at least what the CDCLK frequency requires. + */ + cdclk_state->voltage_level = + skl_calc_voltage_level(cdclk_state->cdclk); } /* convert from kHz to .1 fixpoint MHz with -1MHz offset */ @@ -960,7 +986,7 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, { int cdclk = cdclk_state->cdclk; int vco = cdclk_state->vco; - u32 freq_select, pcu_ack; + u32 freq_select; int ret; mutex_lock(&dev_priv->pcu_lock); @@ -984,21 +1010,17 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, case 308571: case 337500: freq_select = CDCLK_FREQ_337_308; - pcu_ack = 0; break; case 450000: case 432000: freq_select = CDCLK_FREQ_450_432; - pcu_ack = 1; break; case 540000: freq_select = CDCLK_FREQ_540; - pcu_ack = 2; break; case 617143: case 675000: freq_select = CDCLK_FREQ_675_617; - pcu_ack = 3; break; } @@ -1014,7 +1036,8 @@ static void skl_set_cdclk(struct drm_i915_private *dev_priv, /* inform PCU of the change */ mutex_lock(&dev_priv->pcu_lock); - sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, pcu_ack); + sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, + cdclk_state->voltage_level); mutex_unlock(&dev_priv->pcu_lock); intel_update_cdclk(dev_priv); @@ -1093,6 +1116,7 @@ void skl_init_cdclk(struct drm_i915_private *dev_priv) if (cdclk_state.vco == 0) cdclk_state.vco = 8100000; cdclk_state.cdclk = skl_calc_cdclk(0, cdclk_state.vco); + cdclk_state.voltage_level = skl_calc_voltage_level(cdclk_state.cdclk); skl_set_cdclk(dev_priv, &cdclk_state); } @@ -1110,6 +1134,7 @@ void skl_uninit_cdclk(struct drm_i915_private *dev_priv) cdclk_state.cdclk = cdclk_state.ref; cdclk_state.vco = 0; + cdclk_state.voltage_level = skl_calc_voltage_level(cdclk_state.cdclk); skl_set_cdclk(dev_priv, &cdclk_state); } @@ -1968,12 +1993,16 @@ static int skl_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.logical.vco = vco; intel_state->cdclk.logical.cdclk = cdclk; + intel_state->cdclk.logical.voltage_level = + skl_calc_voltage_level(cdclk); if (!intel_state->active_crtcs) { cdclk = skl_calc_cdclk(0, vco); intel_state->cdclk.actual.vco = vco; intel_state->cdclk.actual.cdclk = cdclk; + intel_state->cdclk.actual.voltage_level = + skl_calc_voltage_level(cdclk); } else { intel_state->cdclk.actual = intel_state->cdclk.logical; -- GitLab From 2123f442ca42e9dce2fcb51e8d4fd8180339b4f0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:12 +0300 Subject: [PATCH 0019/1471] drm/i915: Use cdclk_state->voltage on BXT/GLK MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Track the system agent voltage we request from pcode in the cdclk state on BXT/GLK. Annoyingly we can't actually read out the current value since there's no pcode command to do that, so we'll have to just assume that it worked. v2: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-7-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 38e31df265bda..a40dc66dc7733 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1163,6 +1163,11 @@ static int glk_calc_cdclk(int min_cdclk) return 79200; } +static u8 bxt_calc_voltage_level(int cdclk) +{ + return DIV_ROUND_UP(cdclk, 25000); +} + static int bxt_de_pll_vco(struct drm_i915_private *dev_priv, int cdclk) { int ratio; @@ -1239,7 +1244,7 @@ static void bxt_get_cdclk(struct drm_i915_private *dev_priv, cdclk_state->cdclk = cdclk_state->ref; if (cdclk_state->vco == 0) - return; + goto out; divider = I915_READ(CDCLK_CTL) & BXT_CDCLK_CD2X_DIV_SEL_MASK; @@ -1263,6 +1268,14 @@ static void bxt_get_cdclk(struct drm_i915_private *dev_priv, } cdclk_state->cdclk = DIV_ROUND_CLOSEST(cdclk_state->vco, div); + + out: + /* + * Can't read this out :( Let's assume it's + * at least what the CDCLK frequency requires. + */ + cdclk_state->voltage_level = + bxt_calc_voltage_level(cdclk_state->cdclk); } static void bxt_de_pll_disable(struct drm_i915_private *dev_priv) @@ -1365,7 +1378,7 @@ static void bxt_set_cdclk(struct drm_i915_private *dev_priv, mutex_lock(&dev_priv->pcu_lock); ret = sandybridge_pcode_write(dev_priv, HSW_PCODE_DE_WRITE_FREQ_REQ, - DIV_ROUND_UP(cdclk, 25000)); + cdclk_state->voltage_level); mutex_unlock(&dev_priv->pcu_lock); if (ret) { @@ -1457,6 +1470,7 @@ void bxt_init_cdclk(struct drm_i915_private *dev_priv) cdclk_state.cdclk = bxt_calc_cdclk(0); cdclk_state.vco = bxt_de_pll_vco(dev_priv, cdclk_state.cdclk); } + cdclk_state.voltage_level = bxt_calc_voltage_level(cdclk_state.cdclk); bxt_set_cdclk(dev_priv, &cdclk_state); } @@ -1474,6 +1488,7 @@ void bxt_uninit_cdclk(struct drm_i915_private *dev_priv) cdclk_state.cdclk = cdclk_state.ref; cdclk_state.vco = 0; + cdclk_state.voltage_level = bxt_calc_voltage_level(cdclk_state.cdclk); bxt_set_cdclk(dev_priv, &cdclk_state); } @@ -2031,6 +2046,8 @@ static int bxt_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.logical.vco = vco; intel_state->cdclk.logical.cdclk = cdclk; + intel_state->cdclk.logical.voltage_level = + bxt_calc_voltage_level(cdclk); if (!intel_state->active_crtcs) { if (IS_GEMINILAKE(dev_priv)) { @@ -2043,6 +2060,8 @@ static int bxt_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.actual.vco = vco; intel_state->cdclk.actual.cdclk = cdclk; + intel_state->cdclk.actual.voltage_level = + bxt_calc_voltage_level(cdclk); } else { intel_state->cdclk.actual = intel_state->cdclk.logical; -- GitLab From 48469eced282155608a80a37626d28a0abd3c2e5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:13 +0300 Subject: [PATCH 0020/1471] drm/i915: Use cdclk_state->voltage on CNL MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Track the system agent voltage we request from pcode in the cdclk state on CNL. Annoyingly we can't actually read out the current value since there's no pcode command to do that, so we'll have to just assume that it worked. v2: s/voltage/voltage_level/ (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-8-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 47 ++++++++++++++++++++---------- 1 file changed, 31 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index a40dc66dc7733..210e79193fe65 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1503,6 +1503,19 @@ static int cnl_calc_cdclk(int min_cdclk) return 168000; } +static u8 cnl_calc_voltage_level(int cdclk) +{ + switch (cdclk) { + default: + case 168000: + return 0; + case 336000: + return 1; + case 528000: + return 2; + } +} + static void cnl_cdclk_pll_update(struct drm_i915_private *dev_priv, struct intel_cdclk_state *cdclk_state) { @@ -1536,7 +1549,7 @@ static void cnl_get_cdclk(struct drm_i915_private *dev_priv, cdclk_state->cdclk = cdclk_state->ref; if (cdclk_state->vco == 0) - return; + goto out; divider = I915_READ(CDCLK_CTL) & BXT_CDCLK_CD2X_DIV_SEL_MASK; @@ -1553,6 +1566,14 @@ static void cnl_get_cdclk(struct drm_i915_private *dev_priv, } cdclk_state->cdclk = DIV_ROUND_CLOSEST(cdclk_state->vco, div); + + out: + /* + * Can't read this out :( Let's assume it's + * at least what the CDCLK frequency requires. + */ + cdclk_state->voltage_level = + cnl_calc_voltage_level(cdclk_state->cdclk); } static void cnl_cdclk_pll_disable(struct drm_i915_private *dev_priv) @@ -1593,7 +1614,7 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, { int cdclk = cdclk_state->cdclk; int vco = cdclk_state->vco; - u32 val, divider, pcu_ack; + u32 val, divider; int ret; mutex_lock(&dev_priv->pcu_lock); @@ -1622,19 +1643,6 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, break; } - switch (cdclk) { - case 528000: - pcu_ack = 2; - break; - case 336000: - pcu_ack = 1; - break; - case 168000: - default: - pcu_ack = 0; - break; - } - if (dev_priv->cdclk.hw.vco != 0 && dev_priv->cdclk.hw.vco != vco) cnl_cdclk_pll_disable(dev_priv); @@ -1652,7 +1660,8 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, /* inform PCU of the change */ mutex_lock(&dev_priv->pcu_lock); - sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, pcu_ack); + sandybridge_pcode_write(dev_priv, SKL_PCODE_CDCLK_CONTROL, + cdclk_state->voltage_level); mutex_unlock(&dev_priv->pcu_lock); intel_update_cdclk(dev_priv); @@ -1745,6 +1754,7 @@ void cnl_init_cdclk(struct drm_i915_private *dev_priv) cdclk_state.cdclk = cnl_calc_cdclk(0); cdclk_state.vco = cnl_cdclk_pll_vco(dev_priv, cdclk_state.cdclk); + cdclk_state.voltage_level = cnl_calc_voltage_level(cdclk_state.cdclk); cnl_set_cdclk(dev_priv, &cdclk_state); } @@ -1762,6 +1772,7 @@ void cnl_uninit_cdclk(struct drm_i915_private *dev_priv) cdclk_state.cdclk = cdclk_state.ref; cdclk_state.vco = 0; + cdclk_state.voltage_level = cnl_calc_voltage_level(cdclk_state.cdclk); cnl_set_cdclk(dev_priv, &cdclk_state); } @@ -2085,6 +2096,8 @@ static int cnl_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.logical.vco = vco; intel_state->cdclk.logical.cdclk = cdclk; + intel_state->cdclk.logical.voltage_level = + cnl_calc_voltage_level(cdclk); if (!intel_state->active_crtcs) { cdclk = cnl_calc_cdclk(0); @@ -2092,6 +2105,8 @@ static int cnl_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.actual.vco = vco; intel_state->cdclk.actual.cdclk = cdclk; + intel_state->cdclk.actual.voltage_level = + cnl_calc_voltage_level(cdclk); } else { intel_state->cdclk.actual = intel_state->cdclk.logical; -- GitLab From 53e9bf5e8159765e0dc807567180afd0b389f149 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:14 +0300 Subject: [PATCH 0021/1471] drm/i915: Adjust system agent voltage on CNL if required by DDI ports MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit On CNL we may need to bump up the system agent voltage not only due to CDCLK but also when driving DDI port with a sufficiently high clock. To that end start tracking the minimum acceptable voltage for each crtc. We do the tracking via crtcs because we don't have any kind of encoder state. Also there's no downside to doing it this way, and it matches how we track cdclk requirements on account of pixel rate. v2: Allow disabled crtcs to use the min voltage Add IS_CNL check to intel_ddi_compute_min_voltage() since we're using CNL specific values there s/intel_compute_min_voltage/cnl_compute_min_voltage/ since the function makes hw specific assumptions about the voltage values v3: Drop the test hack leftovers from skl_modeset_calc_cdclk() v4: s/voltage/voltage_level/ (Rodrigo) Replace DPLL DVFS FIXMEs with an explanation why we don't do anything there (Rodrigo) Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Reviewed-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-9-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/intel_cdclk.c | 46 ++++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_ddi.c | 11 +++++++ drivers/gpu/drm/i915/intel_display.c | 11 +++++++ drivers/gpu/drm/i915/intel_dp_mst.c | 5 +++ drivers/gpu/drm/i915/intel_dpll_mgr.c | 16 +++++----- drivers/gpu/drm/i915/intel_drv.h | 7 ++++ 7 files changed, 89 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2c64c6680ac73..366ba74b0ad28 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2416,6 +2416,8 @@ struct drm_i915_private { unsigned int active_crtcs; /* minimum acceptable cdclk for each pipe */ int min_cdclk[I915_MAX_PIPES]; + /* minimum acceptable voltage level for each pipe */ + u8 min_voltage_level[I915_MAX_PIPES]; int dpio_phy_iosf_port[I915_NUM_PHYS_VLV]; diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 210e79193fe65..4ca4a34b7bfaf 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1665,6 +1665,12 @@ static void cnl_set_cdclk(struct drm_i915_private *dev_priv, mutex_unlock(&dev_priv->pcu_lock); intel_update_cdclk(dev_priv); + + /* + * Can't read out the voltage level :( + * Let's just assume everything is as expected. + */ + dev_priv->cdclk.hw.voltage_level = cdclk_state->voltage_level; } static int cnl_cdclk_pll_vco(struct drm_i915_private *dev_priv, int cdclk) @@ -1934,6 +1940,43 @@ static int intel_compute_min_cdclk(struct drm_atomic_state *state) return min_cdclk; } +/* + * Note that this functions assumes that 0 is + * the lowest voltage value, and higher values + * correspond to increasingly higher voltages. + * + * Should that relationship no longer hold on + * future platforms this code will need to be + * adjusted. + */ +static u8 cnl_compute_min_voltage_level(struct intel_atomic_state *state) +{ + struct drm_i915_private *dev_priv = to_i915(state->base.dev); + struct intel_crtc *crtc; + struct intel_crtc_state *crtc_state; + u8 min_voltage_level; + int i; + enum pipe pipe; + + memcpy(state->min_voltage_level, dev_priv->min_voltage_level, + sizeof(state->min_voltage_level)); + + for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) { + if (crtc_state->base.enable) + state->min_voltage_level[i] = + crtc_state->min_voltage_level; + else + state->min_voltage_level[i] = 0; + } + + min_voltage_level = 0; + for_each_pipe(dev_priv, pipe) + min_voltage_level = max(state->min_voltage_level[pipe], + min_voltage_level); + + return min_voltage_level; +} + static int vlv_modeset_calc_cdclk(struct drm_atomic_state *state) { struct drm_i915_private *dev_priv = to_i915(state->dev); @@ -2097,7 +2140,8 @@ static int cnl_modeset_calc_cdclk(struct drm_atomic_state *state) intel_state->cdclk.logical.vco = vco; intel_state->cdclk.logical.cdclk = cdclk; intel_state->cdclk.logical.voltage_level = - cnl_calc_voltage_level(cdclk); + max(cnl_calc_voltage_level(cdclk), + cnl_compute_min_voltage_level(intel_state)); if (!intel_state->active_crtcs) { cdclk = cnl_calc_cdclk(0); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 08f04df08fa53..28c25cb9eb2c9 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2542,6 +2542,13 @@ bool intel_ddi_is_audio_enabled(struct drm_i915_private *dev_priv, return false; } +void intel_ddi_compute_min_voltage_level(struct drm_i915_private *dev_priv, + struct intel_crtc_state *crtc_state) +{ + if (IS_CANNONLAKE(dev_priv) && crtc_state->port_clock > 594000) + crtc_state->min_voltage_level = 2; +} + void intel_ddi_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { @@ -2641,6 +2648,8 @@ void intel_ddi_get_config(struct intel_encoder *encoder, if (IS_GEN9_LP(dev_priv)) pipe_config->lane_lat_optim_mask = bxt_ddi_phy_get_lane_lat_optim_mask(encoder); + + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); } static bool intel_ddi_compute_config(struct intel_encoder *encoder, @@ -2667,6 +2676,8 @@ static bool intel_ddi_compute_config(struct intel_encoder *encoder, bxt_ddi_phy_calc_lane_lat_optim_mask(encoder, pipe_config->lane_count); + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); + return ret; } diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 1d2031a5f4bb2..b5fa643e1812b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -5938,6 +5938,7 @@ static void intel_crtc_disable_noatomic(struct drm_crtc *crtc, dev_priv->active_crtcs &= ~(1 << intel_crtc->pipe); dev_priv->min_cdclk[intel_crtc->pipe] = 0; + dev_priv->min_voltage_level[intel_crtc->pipe] = 0; } /* @@ -11289,6 +11290,8 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_CLOCK_FUZZY(base.adjusted_mode.crtc_clock); PIPE_CONF_CHECK_CLOCK_FUZZY(port_clock); + PIPE_CONF_CHECK_I(min_voltage_level); + #undef PIPE_CONF_CHECK_X #undef PIPE_CONF_CHECK_I #undef PIPE_CONF_CHECK_P @@ -11949,6 +11952,9 @@ static int intel_modeset_checks(struct drm_atomic_state *state) DRM_DEBUG_KMS("New cdclk calculated to be logical %u kHz, actual %u kHz\n", intel_state->cdclk.logical.cdclk, intel_state->cdclk.actual.cdclk); + DRM_DEBUG_KMS("New voltage level calculated to be logical %u, actual %u\n", + intel_state->cdclk.logical.voltage_level, + intel_state->cdclk.actual.voltage_level); } else { to_intel_atomic_state(state)->cdclk.logical = dev_priv->cdclk.logical; } @@ -12517,6 +12523,9 @@ static int intel_atomic_commit(struct drm_device *dev, if (intel_state->modeset) { memcpy(dev_priv->min_cdclk, intel_state->min_cdclk, sizeof(intel_state->min_cdclk)); + memcpy(dev_priv->min_voltage_level, + intel_state->min_voltage_level, + sizeof(intel_state->min_voltage_level)); dev_priv->active_crtcs = intel_state->active_crtcs; dev_priv->cdclk.logical = intel_state->cdclk.logical; dev_priv->cdclk.actual = intel_state->cdclk.actual; @@ -15027,6 +15036,8 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) } dev_priv->min_cdclk[crtc->pipe] = min_cdclk; + dev_priv->min_voltage_level[crtc->pipe] = + crtc_state->min_voltage_level; intel_pipe_config_sanity_check(dev_priv, crtc_state); } diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 772521440a9f1..3d62c63c07636 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -34,6 +34,7 @@ static bool intel_dp_mst_compute_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config, struct drm_connector_state *conn_state) { + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp_mst_encoder *intel_mst = enc_to_mst(&encoder->base); struct intel_digital_port *intel_dig_port = intel_mst->primary; struct intel_dp *intel_dp = &intel_dig_port->dp; @@ -87,6 +88,8 @@ static bool intel_dp_mst_compute_config(struct intel_encoder *encoder, pipe_config->dp_m_n.tu = slots; + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); + return true; } @@ -307,6 +310,8 @@ static void intel_dp_mst_enc_get_config(struct intel_encoder *encoder, intel_dp_get_m_n(crtc, pipe_config); intel_ddi_clock_get(&intel_dig_port->base, pipe_config); + + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); } static int intel_dp_mst_get_ddc_modes(struct drm_connector *connector) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index df808a94c5119..897fffe1ecd83 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2008,8 +2008,8 @@ static void cnl_ddi_pll_enable(struct drm_i915_private *dev_priv, * requirement, follow the Display Voltage Frequency Switching * Sequence Before Frequency Change * - * FIXME: (DVFS) is used to adjust the display voltage to match the - * display clock frequencies + * Note: DVFS is actually handled via the cdclk code paths, + * hence we do nothing here. */ /* 6. Enable DPLL in DPLL_ENABLE. */ @@ -2030,8 +2030,8 @@ static void cnl_ddi_pll_enable(struct drm_i915_private *dev_priv, * requirement, follow the Display Voltage Frequency Switching * Sequence After Frequency Change * - * FIXME: (DVFS) is used to adjust the display voltage to match the - * display clock frequencies + * Note: DVFS is actually handled via the cdclk code paths, + * hence we do nothing here. */ /* @@ -2055,8 +2055,8 @@ static void cnl_ddi_pll_disable(struct drm_i915_private *dev_priv, * requirement, follow the Display Voltage Frequency Switching * Sequence Before Frequency Change * - * FIXME: (DVFS) is used to adjust the display voltage to match the - * display clock frequencies + * Note: DVFS is actually handled via the cdclk code paths, + * hence we do nothing here. */ /* 3. Disable DPLL through DPLL_ENABLE. */ @@ -2077,8 +2077,8 @@ static void cnl_ddi_pll_disable(struct drm_i915_private *dev_priv, * requirement, follow the Display Voltage Frequency Switching * Sequence After Frequency Change * - * FIXME: (DVFS) is used to adjust the display voltage to match the - * display clock frequencies + * Note: DVFS is actually handled via the cdclk code paths, + * hence we do nothing here. */ /* 6. Disable DPLL power in DPLL_ENABLE. */ diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index a551aadb157bf..3de8d98baed76 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -386,6 +386,8 @@ struct intel_atomic_state { unsigned int active_crtcs; /* minimum acceptable cdclk for each pipe */ int min_cdclk[I915_MAX_PIPES]; + /* minimum acceptable voltage level for each pipe */ + u8 min_voltage_level[I915_MAX_PIPES]; struct intel_shared_dpll_state shared_dpll[I915_NUM_PLLS]; @@ -739,6 +741,9 @@ struct intel_crtc_state { */ uint8_t lane_lat_optim_mask; + /* minimum acceptable voltage level */ + u8 min_voltage_level; + /* Panel fitter controls for gen2-gen4 + VLV */ struct { u32 control; @@ -1293,6 +1298,8 @@ void intel_ddi_clock_get(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config); void intel_ddi_set_vc_payload_alloc(const struct intel_crtc_state *crtc_state, bool state); +void intel_ddi_compute_min_voltage_level(struct drm_i915_private *dev_priv, + struct intel_crtc_state *crtc_state); u32 bxt_signal_levels(struct intel_dp *intel_dp); uint32_t ddi_signal_levels(struct intel_dp *intel_dp); u8 intel_ddi_dp_voltage_max(struct intel_encoder *encoder); -- GitLab From 0c9f353f014e6d88a5af8b305503a5396fe63ff8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:15 +0300 Subject: [PATCH 0022/1471] drm/i915: Sanity check cdclk in vlv_set_cdclk() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit chv_set_cdclk() sanity checks that the cdclk frequency is one of the legal values. Do the same in the VLV function. Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-10-ville.syrjala@linux.intel.com Reviewed-by: Rodrigo Vivi --- drivers/gpu/drm/i915/intel_cdclk.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 4ca4a34b7bfaf..fedfe3c720b62 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -520,6 +520,18 @@ static void vlv_set_cdclk(struct drm_i915_private *dev_priv, int cdclk = cdclk_state->cdclk; u32 val, cmd = cdclk_state->voltage_level; + switch (cdclk) { + case 400000: + case 333333: + case 320000: + case 266667: + case 200000: + break; + default: + MISSING_CASE(cdclk); + return; + } + /* There are cases where we can end up here with power domains * off and a CDCLK frequency other than the minimum, like when * issuing a modeset without actually changing any display after -- GitLab From cfddadc98abc85f478c92187c944e0ee963f741b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 24 Oct 2017 12:52:16 +0300 Subject: [PATCH 0023/1471] drm/i915: Perform a central cdclk state sanity check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit WARN if the cdclk state doesn't match what we expect after programming. And let's remove the WARN from bdw_set_cdclk() that's trying to achieve the same thing in a more limite fashion. Also take the opportunity to refactor the code to use a common function for dumping out a cdclk state. Cc: Mika Kahola Cc: Manasi Navare Cc: Rodrigo Vivi Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171024095216.1638-11-ville.syrjala@linux.intel.com Reviewed-by: Rodrigo Vivi --- drivers/gpu/drm/i915/intel_cdclk.c | 30 ++++++++++++++++++---------- drivers/gpu/drm/i915/intel_display.c | 3 +++ drivers/gpu/drm/i915/intel_drv.h | 2 ++ 3 files changed, 24 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index fedfe3c720b62..51cd23dd86764 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -768,10 +768,6 @@ static void bdw_set_cdclk(struct drm_i915_private *dev_priv, I915_WRITE(CDCLK_FREQ, DIV_ROUND_CLOSEST(cdclk, 1000) - 1); intel_update_cdclk(dev_priv); - - WARN(cdclk != dev_priv->cdclk.hw.cdclk, - "cdclk requested %d kHz but got %d kHz\n", - cdclk, dev_priv->cdclk.hw.cdclk); } static int skl_calc_cdclk(int min_cdclk, int vco) @@ -1068,6 +1064,8 @@ static void skl_sanitize_cdclk(struct drm_i915_private *dev_priv) goto sanitize; intel_update_cdclk(dev_priv); + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); + /* Is PLL enabled and locked ? */ if (dev_priv->cdclk.hw.vco == 0 || dev_priv->cdclk.hw.cdclk == dev_priv->cdclk.hw.ref) @@ -1407,6 +1405,7 @@ static void bxt_sanitize_cdclk(struct drm_i915_private *dev_priv) u32 cdctl, expected; intel_update_cdclk(dev_priv); + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); if (dev_priv->cdclk.hw.vco == 0 || dev_priv->cdclk.hw.cdclk == dev_priv->cdclk.hw.ref) @@ -1713,6 +1712,7 @@ static void cnl_sanitize_cdclk(struct drm_i915_private *dev_priv) u32 cdctl, expected; intel_update_cdclk(dev_priv); + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); if (dev_priv->cdclk.hw.vco == 0 || dev_priv->cdclk.hw.cdclk == dev_priv->cdclk.hw.ref) @@ -1826,6 +1826,14 @@ bool intel_cdclk_changed(const struct intel_cdclk_state *a, a->voltage_level != b->voltage_level; } +void intel_dump_cdclk_state(const struct intel_cdclk_state *cdclk_state, + const char *context) +{ + DRM_DEBUG_DRIVER("%s %d kHz, VCO %d kHz, ref %d kHz, voltage level %d\n", + context, cdclk_state->cdclk, cdclk_state->vco, + cdclk_state->ref, cdclk_state->voltage_level); +} + /** * intel_set_cdclk - Push the CDCLK state to the hardware * @dev_priv: i915 device @@ -1843,11 +1851,15 @@ void intel_set_cdclk(struct drm_i915_private *dev_priv, if (WARN_ON_ONCE(!dev_priv->display.set_cdclk)) return; - DRM_DEBUG_DRIVER("Changing CDCLK to %d kHz, VCO %d kHz, ref %d kHz, voltage_level %d\n", - cdclk_state->cdclk, cdclk_state->vco, - cdclk_state->ref, cdclk_state->voltage_level); + intel_dump_cdclk_state(cdclk_state, "Changing CDCLK to"); dev_priv->display.set_cdclk(dev_priv, cdclk_state); + + if (WARN(intel_cdclk_changed(&dev_priv->cdclk.hw, cdclk_state), + "cdclk state doesn't match!\n")) { + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "[hw state]"); + intel_dump_cdclk_state(cdclk_state, "[sw state]"); + } } static int intel_pixel_rate_to_cdclk(struct drm_i915_private *dev_priv, @@ -2280,10 +2292,6 @@ void intel_update_cdclk(struct drm_i915_private *dev_priv) { dev_priv->display.get_cdclk(dev_priv, &dev_priv->cdclk.hw); - DRM_DEBUG_DRIVER("Current CD clock rate: %d kHz, VCO: %d kHz, ref: %d kHz\n", - dev_priv->cdclk.hw.cdclk, dev_priv->cdclk.hw.vco, - dev_priv->cdclk.hw.ref); - /* * 9:0 CMBUS [sic] CDCLK frequency (cdfreq): * Programmng [sic] note: bit[9:2] should be programmed to the number diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index b5fa643e1812b..7d7952b78a3b8 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -8857,7 +8857,9 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv) } intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); + intel_update_cdclk(dev_priv); + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); } /* @@ -14358,6 +14360,7 @@ void intel_modeset_init_hw(struct drm_device *dev) struct drm_i915_private *dev_priv = to_i915(dev); intel_update_cdclk(dev_priv); + intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); dev_priv->cdclk.logical = dev_priv->cdclk.actual = dev_priv->cdclk.hw; intel_init_clock_gating(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3de8d98baed76..3b4eafd39f558 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1336,6 +1336,8 @@ bool intel_cdclk_changed(const struct intel_cdclk_state *a, const struct intel_cdclk_state *b); void intel_set_cdclk(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state); +void intel_dump_cdclk_state(const struct intel_cdclk_state *cdclk_state, + const char *context); /* intel_display.c */ void i830_enable_pipe(struct drm_i915_private *dev_priv, enum pipe pipe); -- GitLab From 753bdbd001dca74f9f4528d7be415801f043748a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 24 Oct 2017 21:50:53 +0100 Subject: [PATCH 0024/1471] drm/i915: Call cond_resched() before repeating i915_gem_evict_something() Insert a breakpoint, a chance to escape back to the scheduler and run something else for a bit, if we find that the GGTT is full and needs to be idled in order to make some room. In practice, this should only be an issue in stress tests as the wait itself will normally give the chance for the scheduler to intervene and make progress. References: https://bugs.freedesktop.org/show_bug.cgi?id=103438 Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171024205053.7845-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_evict.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c index 8daa8a78cdc06..a6b769994d8df 100644 --- a/drivers/gpu/drm/i915/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/i915_gem_evict.c @@ -216,6 +216,7 @@ i915_gem_evict_something(struct i915_address_space *vm, if (ret) return ret; + cond_resched(); goto search_again; } -- GitLab From b1f9107e1b251b2cbd413062fad3636570d2ea20 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 24 Oct 2017 23:08:54 +0100 Subject: [PATCH 0025/1471] drm/i915/selftests: Don't try to queue a request with zero delay Instead of trying to create a timer with zero delay (i.e. with expires set to the current jiffies and not the future, an already expired timer), execute that request immediately. v2: Refactor list_del_init+signal into its own little function. v3: Reorder testing so as not to immediately signal a delayed request. Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171024220855.30155-1-chris@chris-wilson.co.uk Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/selftests/mock_engine.c | 37 +++++++++++++++----- 1 file changed, 28 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/mock_engine.c b/drivers/gpu/drm/i915/selftests/mock_engine.c index 331c2b09869ee..0aafa8a105b8f 100644 --- a/drivers/gpu/drm/i915/selftests/mock_engine.c +++ b/drivers/gpu/drm/i915/selftests/mock_engine.c @@ -32,6 +32,13 @@ static struct mock_request *first_request(struct mock_engine *engine) link); } +static void advance(struct mock_engine *engine, + struct mock_request *request) +{ + list_del_init(&request->link); + mock_seqno_advance(&engine->base, request->base.global_seqno); +} + static void hw_delay_complete(struct timer_list *t) { struct mock_engine *engine = from_timer(engine, t, hw_delay); @@ -39,15 +46,23 @@ static void hw_delay_complete(struct timer_list *t) spin_lock(&engine->hw_lock); - request = first_request(engine); - if (request) { - list_del_init(&request->link); - mock_seqno_advance(&engine->base, request->base.global_seqno); - } - + /* Timer fired, first request is complete */ request = first_request(engine); if (request) - mod_timer(&engine->hw_delay, jiffies + request->delay); + advance(engine, request); + + /* + * Also immediately signal any subsequent 0-delay requests, but + * requeue the timer for the next delayed request. + */ + while ((request = first_request(engine))) { + if (request->delay) { + mod_timer(&engine->hw_delay, jiffies + request->delay); + break; + } + + advance(engine, request); + } spin_unlock(&engine->hw_lock); } @@ -98,8 +113,12 @@ static void mock_submit_request(struct drm_i915_gem_request *request) spin_lock_irq(&engine->hw_lock); list_add_tail(&mock->link, &engine->hw_queue); - if (mock->link.prev == &engine->hw_queue) - mod_timer(&engine->hw_delay, jiffies + mock->delay); + if (mock->link.prev == &engine->hw_queue) { + if (mock->delay) + mod_timer(&engine->hw_delay, jiffies + mock->delay); + else + advance(engine, mock); + } spin_unlock_irq(&engine->hw_lock); } -- GitLab From 20ccd4d3f689ac14dce8632d76769be0ac952060 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 24 Oct 2017 23:08:55 +0100 Subject: [PATCH 0026/1471] drm/i915: Use same test for eviction and submitting kernel context During evict, we wish to idle the GPU if we see that the GGTT is full. However, our test for idle in i915_gem_evict_something() and in i915_gem_switch_to_kernel_context() do not match leading to disappointment - we never believe that we are idle and keep trying to flush the GGTT ad infinitum. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103438 Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Tvrtko Ursulin Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171024220855.30155-2-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_context.c | 7 +++---- drivers/gpu/drm/i915/i915_gem_evict.c | 3 ++- drivers/gpu/drm/i915/intel_engine_cs.c | 6 ++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 2 ++ 4 files changed, 13 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 5bf96a2585099..4f26f80b1b3eb 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -897,7 +897,7 @@ int i915_switch_context(struct drm_i915_gem_request *req) return do_rcs_switch(req); } -static bool engine_has_kernel_context(struct intel_engine_cs *engine) +static bool engine_has_idle_kernel_context(struct intel_engine_cs *engine) { struct i915_gem_timeline *timeline; @@ -913,8 +913,7 @@ static bool engine_has_kernel_context(struct intel_engine_cs *engine) return false; } - return (!engine->last_retired_context || - i915_gem_context_is_kernel(engine->last_retired_context)); + return intel_engine_has_kernel_context(engine); } int i915_gem_switch_to_kernel_context(struct drm_i915_private *dev_priv) @@ -931,7 +930,7 @@ int i915_gem_switch_to_kernel_context(struct drm_i915_private *dev_priv) struct drm_i915_gem_request *req; int ret; - if (engine_has_kernel_context(engine)) + if (engine_has_idle_kernel_context(engine)) continue; req = i915_gem_request_alloc(engine, dev_priv->kernel_context); diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c index a6b769994d8df..60ca4f05ae94a 100644 --- a/drivers/gpu/drm/i915/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/i915_gem_evict.c @@ -46,7 +46,7 @@ static bool ggtt_is_idle(struct drm_i915_private *i915) return false; for_each_engine(engine, i915, id) { - if (engine->last_retired_context != i915->kernel_context) + if (!intel_engine_has_kernel_context(engine)) return false; } @@ -73,6 +73,7 @@ static int ggtt_flush(struct drm_i915_private *i915) if (err) return err; + GEM_BUG_ON(!ggtt_is_idle(i915)); return 0; } diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index ab5bf4e2e28e2..f6cdc50d42371 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1585,6 +1585,12 @@ bool intel_engines_are_idle(struct drm_i915_private *dev_priv) return true; } +bool intel_engine_has_kernel_context(const struct intel_engine_cs *engine) +{ + return (!engine->last_retired_context || + i915_gem_context_is_kernel(engine->last_retired_context)); +} + void intel_engines_reset_default_submission(struct drm_i915_private *i915) { struct intel_engine_cs *engine; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 6a42ed618a28c..a2589aa891639 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -866,6 +866,8 @@ static inline u32 *gen8_emit_pipe_control(u32 *batch, u32 flags, u32 offset) bool intel_engine_is_idle(struct intel_engine_cs *engine); bool intel_engines_are_idle(struct drm_i915_private *dev_priv); +bool intel_engine_has_kernel_context(const struct intel_engine_cs *engine); + void intel_engines_mark_idle(struct drm_i915_private *i915); void intel_engines_reset_default_submission(struct drm_i915_private *i915); -- GitLab From 6d0dbd309687113009a276886628c7c74c848d3c Mon Sep 17 00:00:00 2001 From: Kees Cook Date: Tue, 24 Oct 2017 08:13:44 -0700 Subject: [PATCH 0027/1471] drm/i915/selftests: Convert timers to use timer_setup() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new timer_setup() and from_timer() to pass the timer pointer explicitly. Cc: Jani Nikula Cc: Joonas Lahtinen Cc: Rodrigo Vivi Cc: David Airlie Cc: Tvrtko Ursulin Cc: Chris Wilson Cc: intel-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Signed-off-by: Kees Cook Link: https://patchwork.freedesktop.org/patch/msgid/20171024151344.GA104417@beast Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/selftests/lib_sw_fence.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c index 3790fdf44a1ab..b26f07b55d861 100644 --- a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c +++ b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c @@ -49,9 +49,9 @@ void onstack_fence_fini(struct i915_sw_fence *fence) i915_sw_fence_fini(fence); } -static void timed_fence_wake(unsigned long data) +static void timed_fence_wake(struct timer_list *t) { - struct timed_fence *tf = (struct timed_fence *)data; + struct timed_fence *tf = from_timer(tf, t, timer); i915_sw_fence_commit(&tf->fence); } @@ -60,7 +60,7 @@ void timed_fence_init(struct timed_fence *tf, unsigned long expires) { onstack_fence_init(&tf->fence); - setup_timer_on_stack(&tf->timer, timed_fence_wake, (unsigned long)tf); + timer_setup_on_stack(&tf->timer, timed_fence_wake, 0); if (time_after(expires, jiffies)) mod_timer(&tf->timer, expires); -- GitLab From 02a1ca45cf33d23c7540b572aed5bf624ba968fe Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 25 Oct 2017 14:13:36 +0100 Subject: [PATCH 0028/1471] Revert "drm/i915/selftests: Convert timers to use timer_setup()" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This reverts commit 6d0dbd309687113009a276886628c7c74c848d3c. timer_setup_on_stack() does not yet exist: In file included from drivers/gpu/drm/i915/i915_sw_fence.c:517:0: drivers/gpu/drm/i915/selftests/lib_sw_fence.c: In function ‘timed_fence_init’: drivers/gpu/drm/i915/selftests/lib_sw_fence.c:63:2: error: implicit declaration of function ‘timer_setup_on_stack’; did you mean ‘hrtimer_init_on_stack’? [-Werror=implicit-function-declaration] timer_setup_on_stack(&tf->timer, timed_fence_wake, 0); Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171025131336.2584-1-chris@chris-wilson.co.uk Acked-by: Joonas Lahtinen --- drivers/gpu/drm/i915/selftests/lib_sw_fence.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c index b26f07b55d861..3790fdf44a1ab 100644 --- a/drivers/gpu/drm/i915/selftests/lib_sw_fence.c +++ b/drivers/gpu/drm/i915/selftests/lib_sw_fence.c @@ -49,9 +49,9 @@ void onstack_fence_fini(struct i915_sw_fence *fence) i915_sw_fence_fini(fence); } -static void timed_fence_wake(struct timer_list *t) +static void timed_fence_wake(unsigned long data) { - struct timed_fence *tf = from_timer(tf, t, timer); + struct timed_fence *tf = (struct timed_fence *)data; i915_sw_fence_commit(&tf->fence); } @@ -60,7 +60,7 @@ void timed_fence_init(struct timed_fence *tf, unsigned long expires) { onstack_fence_init(&tf->fence); - timer_setup_on_stack(&tf->timer, timed_fence_wake, 0); + setup_timer_on_stack(&tf->timer, timed_fence_wake, (unsigned long)tf); if (time_after(expires, jiffies)) mod_timer(&tf->timer, expires); -- GitLab From 22a8a4fc93b14b5a8cfc785edbdc6f7bd98bffb6 Mon Sep 17 00:00:00 2001 From: Joonas Lahtinen Date: Mon, 23 Oct 2017 18:32:09 +0300 Subject: [PATCH 0029/1471] drm/i915: Disable lazy PPGTT page table optimization for vGPU When running under virtualization (vGPU active), we must disable the lazy PPGTT page table initialization optimization introduced by commit 14826673247e ("drm/i915: Only initialize partially filled pagetables"). We must do this because GVT-g makes unduly assumptions about guest behaviour, which this optimization breaks. This results in following looking errors in the host: ERROR gvt: guest page write error -22, gfn 0x7ada8, pa 0x7ada89a8, var 0x6, len 1 The real fix is to not to depend on i915 driver behaviour, but instead either rely on only the contracts that i915 has with the hardware, or add some paravirtualization. While the real fix is en route, it won't be finished in time for 4.15, so the best option is to disable the optimization for now when vGPU is active to avoid breaking 4.15 guests in existing VM environments. Fixes: 14826673247e ("drm/i915: Only initialize partially filled pagetables") Suggested-by: Xiaolin Zhang Signed-off-by: Xiaolin Zhang [Joonas: Rewrote the commit message and added tags.] Signed-off-by: Joonas Lahtinen Cc: Zhenyu Wang Cc: Zhi Wang Cc: Chris Wilson Cc: Matthew Auld Cc: Joonas Lahtinen Cc: Jani Nikula Cc: Rodrigo Vivi Acked-by: Chris Wilson Acked-by: Zhenyu Wang Link: https://patchwork.freedesktop.org/patch/msgid/20171023153209.10527-1-joonas.lahtinen@linux.intel.com --- drivers/gpu/drm/i915/i915_gem_gtt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 527a2d2d62810..5eaa6893daaa1 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -1341,7 +1341,7 @@ static int gen8_ppgtt_alloc_pd(struct i915_address_space *vm, if (IS_ERR(pt)) goto unwind; - if (count < GEN8_PTES) + if (count < GEN8_PTES || intel_vgpu_active(vm->i915)) gen8_initialize_pt(vm, pt); gen8_ppgtt_set_pde(vm, pd, pt, pde); -- GitLab From 43037c86d10cea185f6518f797f6303a06e734f9 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 3 Oct 2017 15:31:42 -0700 Subject: [PATCH 0030/1471] drm/i915/cnl: Allow 2 pixel per clock on Cannonlake. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is heavily based on a initial patch provided by Ville plus all changes provided later by Ander. As Geminilake, Cannonlake also supports 2 pixels per clock. Different from Geminilake we are not implementing the 99% Wa. But we can revisit that decision later if we find out any limitation on later CNL SKUs. v2: Rebase on top of commit 'd305e0614601 ("drm/i915: Track minimum acceptable cdclk instead of "minimum dotclock")' v3: When fixing HDMI on CNL I noticed that I missed to convert back the doubled pixel rate to cdclk. Cc: Paulo Zanoni Cc: Ville Syrjälä Cc: Dhinakaran Pandiyan Cc: Jani Nikula Signed-off-by: Rodrigo Vivi Reviewed-by: Paulo Zanoni Link: https://patchwork.freedesktop.org/patch/msgid/20171003223142.26264-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_cdclk.c | 14 ++------------ drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_pm.c | 3 ++- 3 files changed, 5 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index 51cd23dd86764..e8884c2ade98b 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1866,12 +1866,7 @@ static int intel_pixel_rate_to_cdclk(struct drm_i915_private *dev_priv, int pixel_rate) { if (INTEL_GEN(dev_priv) >= 10) - /* - * FIXME: Switch to DIV_ROUND_UP(pixel_rate, 2) - * once DDI clock voltage requirements are - * handled correctly. - */ - return pixel_rate; + return DIV_ROUND_UP(pixel_rate, 2); else if (IS_GEMINILAKE(dev_priv)) /* * FIXME: Avoid using a pixel clock that is more than 99% of the cdclk @@ -2188,12 +2183,7 @@ static int intel_compute_max_dotclk(struct drm_i915_private *dev_priv) int max_cdclk_freq = dev_priv->max_cdclk_freq; if (INTEL_GEN(dev_priv) >= 10) - /* - * FIXME: Allow '2 * max_cdclk_freq' - * once DDI clock voltage requirements are - * handled correctly. - */ - return max_cdclk_freq; + return 2 * max_cdclk_freq; else if (IS_GEMINILAKE(dev_priv)) /* * FIXME: Limiting to 99% as a temporary workaround. See diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 7d7952b78a3b8..0e493a1fedab5 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12765,7 +12765,7 @@ skl_max_scale(struct intel_crtc *intel_crtc, struct intel_crtc_state *crtc_state crtc_clock = crtc_state->base.adjusted_mode.crtc_clock; max_dotclk = to_intel_atomic_state(crtc_state->base.state)->cdclk.logical.cdclk; - if (IS_GEMINILAKE(dev_priv)) + if (IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10) max_dotclk *= 2; if (WARN_ON_ONCE(!crtc_clock || max_dotclk < crtc_clock)) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 1ead51754e8f5..742d5455b2018 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3932,6 +3932,7 @@ skl_pipe_downscale_amount(const struct intel_crtc_state *crtc_state) int skl_check_pipe_max_pixel_rate(struct intel_crtc *intel_crtc, struct intel_crtc_state *cstate) { + struct drm_i915_private *dev_priv = to_i915(intel_crtc->base.dev); struct drm_crtc_state *crtc_state = &cstate->base; struct drm_atomic_state *state = crtc_state->state; struct drm_plane *plane; @@ -3974,7 +3975,7 @@ int skl_check_pipe_max_pixel_rate(struct intel_crtc *intel_crtc, crtc_clock = crtc_state->adjusted_mode.crtc_clock; dotclk = to_intel_atomic_state(state)->cdclk.logical.cdclk; - if (IS_GEMINILAKE(to_i915(intel_crtc->base.dev))) + if (IS_GEMINILAKE(dev_priv) || INTEL_GEN(dev_priv) >= 10) dotclk *= 2; pipe_max_pixel_rate = div_round_up_u32_fixed16(dotclk, pipe_downscale); -- GitLab From aba5e278586b16a4fbd377e7e8403aa585d0532a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 25 Oct 2017 15:39:41 +0100 Subject: [PATCH 0031/1471] drm/i915: Add a hook for making the engines idle (parking) and unparking MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In the next patch, we will want to install a callback when the engines (GT as a whole) become idle and similarly when they first become busy. To enable that callback, first rename intel_engines_mark_idle() to intel_engines_park() and provide the companion intel_engines_unpark(). Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Tvrtko Ursulin Cc: Michał Winiarski Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171025143943.7661-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 2 +- drivers/gpu/drm/i915/i915_gem_request.c | 2 ++ drivers/gpu/drm/i915/intel_engine_cs.c | 33 +++++++++++++++++++++++-- drivers/gpu/drm/i915/intel_lrc.c | 3 +++ drivers/gpu/drm/i915/intel_ringbuffer.c | 5 +++- drivers/gpu/drm/i915/intel_ringbuffer.h | 7 +++++- 6 files changed, 47 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bafe1cdd57d8d..8364304e9d2b2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3347,7 +3347,7 @@ i915_gem_idle_work_handler(struct work_struct *work) if (!intel_engines_are_idle(dev_priv)) DRM_ERROR("Timeout waiting for engines to idle\n"); - intel_engines_mark_idle(dev_priv); + intel_engines_park(dev_priv); i915_gem_timelines_mark_idle(dev_priv); GEM_BUG_ON(!dev_priv->gt.awake); diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index d140fcf5c6a39..e0d6221022a83 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -259,6 +259,8 @@ static void mark_busy(struct drm_i915_private *i915) if (INTEL_GEN(i915) >= 6) gen6_rps_busy(i915); + intel_engines_unpark(i915); + queue_delayed_work(i915->wq, &i915->gt.retire_work, round_jiffies_up_relative(HZ)); diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index f6cdc50d42371..fedb839dff611 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1600,19 +1600,48 @@ void intel_engines_reset_default_submission(struct drm_i915_private *i915) engine->set_default_submission(engine); } -void intel_engines_mark_idle(struct drm_i915_private *i915) +/** + * intel_engines_park: called when the GT is transitioning from busy->idle + * @i915: the i915 device + * + * The GT is now idle and about to go to sleep (maybe never to wake again?). + * Time for us to tidy and put away our toys (release resources back to the + * system). + */ +void intel_engines_park(struct drm_i915_private *i915) { struct intel_engine_cs *engine; enum intel_engine_id id; for_each_engine(engine, i915, id) { + if (engine->park) + engine->park(engine); + intel_engine_disarm_breadcrumbs(engine); - i915_gem_batch_pool_fini(&engine->batch_pool); tasklet_kill(&engine->execlists.irq_tasklet); + + i915_gem_batch_pool_fini(&engine->batch_pool); engine->execlists.no_priolist = false; } } +/** + * intel_engines_unpark: called when the GT is transitioning from idle->busy + * @i915: the i915 device + * + * The GT was idle and now about to fire up with some new user requests. + */ +void intel_engines_unpark(struct drm_i915_private *i915) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + + for_each_engine(engine, i915, id) { + if (engine->unpark) + engine->unpark(engine); + } +} + bool intel_engine_can_store_dword(struct intel_engine_cs *engine) { switch (INTEL_GEN(engine->i915)) { diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index d36e256074354..eeb3622803a8f 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1891,6 +1891,9 @@ static void execlists_set_default_submission(struct intel_engine_cs *engine) engine->cancel_requests = execlists_cancel_requests; engine->schedule = execlists_schedule; engine->execlists.irq_tasklet.func = intel_lrc_irq_handler; + + engine->park = NULL; + engine->unpark = NULL; } static void diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 8da1bde442dd9..05e01446b00b3 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -2028,12 +2028,15 @@ static void i9xx_set_default_submission(struct intel_engine_cs *engine) { engine->submit_request = i9xx_submit_request; engine->cancel_requests = cancel_requests; + + engine->park = NULL; + engine->unpark = NULL; } static void gen6_bsd_set_default_submission(struct intel_engine_cs *engine) { + i9xx_set_default_submission(engine); engine->submit_request = gen6_bsd_submit_request; - engine->cancel_requests = cancel_requests; } static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv, diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index a2589aa891639..a7c95f0c3101a 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -365,6 +365,9 @@ struct intel_engine_cs { void (*reset_hw)(struct intel_engine_cs *engine, struct drm_i915_gem_request *req); + void (*park)(struct intel_engine_cs *engine); + void (*unpark)(struct intel_engine_cs *engine); + void (*set_default_submission)(struct intel_engine_cs *engine); struct intel_ring *(*context_pin)(struct intel_engine_cs *engine, @@ -868,7 +871,9 @@ bool intel_engines_are_idle(struct drm_i915_private *dev_priv); bool intel_engine_has_kernel_context(const struct intel_engine_cs *engine); -void intel_engines_mark_idle(struct drm_i915_private *i915); +void intel_engines_park(struct drm_i915_private *i915); +void intel_engines_unpark(struct drm_i915_private *i915); + void intel_engines_reset_default_submission(struct drm_i915_private *i915); bool intel_engine_can_store_dword(struct intel_engine_cs *engine); -- GitLab From bcbd5c33a342bc6f4a25fed528dc3ed2a0b1c140 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 25 Oct 2017 15:39:42 +0100 Subject: [PATCH 0032/1471] drm/i915/guc: Always enable the breadcrumbs irq MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The execlists emulation on top of the GuC (used for scheduling and preemption) depends on the MI_USER_INTERRUPT for its notifications and tasklet action. As we always employ the irq, there is no advantage in ever disabling it while we are using the GuC, so allow us to arm the breadcrumb irq when enabling GuC submission and disarm upon disabling. The impact should be lessened by the delayed irq disabling we do (we only disable after receiving an interrupt for which no one was wanting), but allowing guc to explicitly manage the irq in relation to itself is simpler and prevents an issue with losing an interrupt for preemption as it is not coupled to an active request. Internally, we add a reference counter (breadcrumbs.irq_enabled) as a simple mechanism to allow GuC to keep the breadcrumb irq enabled. To improve upon always enabling the irq while guc is selected, we need to hook into the parking facility of intel_engines so that we only enable the breadcrumbs while the GT is active (one step better would be to individually park/unpark each engine). In effect, this means that we keep the breadcrumb irq always enabled for the entire duration the guc is busy, whereas before we would try to switch it off whenever we idled for more than interrupt with no associated waiters. The difference *should* be negligible in practice! v2: Stop abusing fence signaling (and its auxiliary data structures) to enable the breadcrumbs irqs. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Michał Winiarski , Cc: Tvrtko Ursulin Cc: Daniele Ceraolo Spurio Reviewed-by: Michał Winiarski , Link: https://patchwork.freedesktop.org/patch/msgid/20171025143943.7661-3-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_guc_submission.c | 39 ++++++++----------- drivers/gpu/drm/i915/i915_irq.c | 6 ++- drivers/gpu/drm/i915/intel_breadcrumbs.c | 45 +++++++++++++++++----- drivers/gpu/drm/i915/intel_ringbuffer.h | 5 ++- 4 files changed, 59 insertions(+), 36 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index f84c267728fda..141ed9df3d5ca 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -523,29 +523,6 @@ static void i915_guc_submit(struct intel_engine_cs *engine) } } -static void nested_enable_signaling(struct drm_i915_gem_request *rq) -{ - /* If we use dma_fence_enable_sw_signaling() directly, lockdep - * detects an ordering issue between the fence lockclass and the - * global_timeline. This circular dependency can only occur via 2 - * different fences (but same fence lockclass), so we use the nesting - * annotation here to prevent the warn, equivalent to the nesting - * inside i915_gem_request_submit() for when we also enable the - * signaler. - */ - - if (test_and_set_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, - &rq->fence.flags)) - return; - - GEM_BUG_ON(test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &rq->fence.flags)); - trace_dma_fence_enable_signal(&rq->fence); - - spin_lock_nested(&rq->lock, SINGLE_DEPTH_NESTING); - intel_engine_enable_signaling(rq, true); - spin_unlock(&rq->lock); -} - static void port_assign(struct execlist_port *port, struct drm_i915_gem_request *rq) { @@ -555,7 +532,6 @@ static void port_assign(struct execlist_port *port, i915_gem_request_put(port_request(port)); port_set(port, port_pack(i915_gem_request_get(rq), port_count(port))); - nested_enable_signaling(rq); } static void i915_guc_dequeue(struct intel_engine_cs *engine) @@ -1097,6 +1073,16 @@ static void guc_interrupts_release(struct drm_i915_private *dev_priv) rps->pm_intrmsk_mbz &= ~ARAT_EXPIRED_INTRMSK; } +static void i915_guc_submission_park(struct intel_engine_cs *engine) +{ + intel_engine_unpin_breadcrumbs_irq(engine); +} + +static void i915_guc_submission_unpark(struct intel_engine_cs *engine) +{ + intel_engine_pin_breadcrumbs_irq(engine); +} + int i915_guc_submission_enable(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; @@ -1154,6 +1140,9 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) execlists->irq_tasklet.func = i915_guc_irq_handler; clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted); tasklet_schedule(&execlists->irq_tasklet); + + engine->park = i915_guc_submission_park; + engine->unpark = i915_guc_submission_unpark; } return 0; @@ -1168,6 +1157,8 @@ void i915_guc_submission_disable(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; + GEM_BUG_ON(dev_priv->gt.awake); /* GT should be parked first */ + guc_interrupts_release(dev_priv); /* Revert back to manual ELSP submission */ diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f8205841868bc..ff00e462697ad 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1068,6 +1068,9 @@ static void notify_ring(struct intel_engine_cs *engine) struct drm_i915_gem_request *rq = NULL; struct intel_wait *wait; + if (!engine->breadcrumbs.irq_armed) + return; + atomic_inc(&engine->irq_count); set_bit(ENGINE_IRQ_BREADCRUMB, &engine->irq_posted); @@ -1101,7 +1104,8 @@ static void notify_ring(struct intel_engine_cs *engine) if (wakeup) wake_up_process(wait->tsk); } else { - __intel_engine_disarm_breadcrumbs(engine); + if (engine->breadcrumbs.irq_armed) + __intel_engine_disarm_breadcrumbs(engine); } spin_unlock(&engine->breadcrumbs.irq_lock); diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c index 48e1ba01ccf88..1bd0a461d6653 100644 --- a/drivers/gpu/drm/i915/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c @@ -123,7 +123,7 @@ static void intel_breadcrumbs_fake_irq(struct timer_list *t) */ spin_lock_irq(&b->irq_lock); - if (!__intel_breadcrumbs_wakeup(b)) + if (b->irq_armed && !__intel_breadcrumbs_wakeup(b)) __intel_engine_disarm_breadcrumbs(engine); spin_unlock_irq(&b->irq_lock); if (!b->irq_armed) @@ -171,15 +171,37 @@ void __intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine) lockdep_assert_held(&b->irq_lock); GEM_BUG_ON(b->irq_wait); + GEM_BUG_ON(!b->irq_armed); - if (b->irq_enabled) { + GEM_BUG_ON(!b->irq_enabled); + if (!--b->irq_enabled) irq_disable(engine); - b->irq_enabled = false; - } b->irq_armed = false; } +void intel_engine_pin_breadcrumbs_irq(struct intel_engine_cs *engine) +{ + struct intel_breadcrumbs *b = &engine->breadcrumbs; + + spin_lock_irq(&b->irq_lock); + if (!b->irq_enabled++) + irq_enable(engine); + GEM_BUG_ON(!b->irq_enabled); /* no overflow! */ + spin_unlock_irq(&b->irq_lock); +} + +void intel_engine_unpin_breadcrumbs_irq(struct intel_engine_cs *engine) +{ + struct intel_breadcrumbs *b = &engine->breadcrumbs; + + spin_lock_irq(&b->irq_lock); + GEM_BUG_ON(!b->irq_enabled); /* no underflow! */ + if (!--b->irq_enabled) + irq_disable(engine); + spin_unlock_irq(&b->irq_lock); +} + void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine) { struct intel_breadcrumbs *b = &engine->breadcrumbs; @@ -241,6 +263,9 @@ static bool __intel_breadcrumbs_enable_irq(struct intel_breadcrumbs *b) struct intel_engine_cs *engine = container_of(b, struct intel_engine_cs, breadcrumbs); struct drm_i915_private *i915 = engine->i915; + bool enabled; + + GEM_BUG_ON(!intel_irqs_enabled(i915)); lockdep_assert_held(&b->irq_lock); if (b->irq_armed) @@ -252,7 +277,6 @@ static bool __intel_breadcrumbs_enable_irq(struct intel_breadcrumbs *b) * the irq. */ b->irq_armed = true; - GEM_BUG_ON(b->irq_enabled); if (I915_SELFTEST_ONLY(b->mock)) { /* For our mock objects we want to avoid interaction @@ -273,14 +297,15 @@ static bool __intel_breadcrumbs_enable_irq(struct intel_breadcrumbs *b) */ /* No interrupts? Kick the waiter every jiffie! */ - if (intel_irqs_enabled(i915)) { - if (!test_bit(engine->id, &i915->gpu_error.test_irq_rings)) - irq_enable(engine); - b->irq_enabled = true; + enabled = false; + if (!b->irq_enabled++ && + !test_bit(engine->id, &i915->gpu_error.test_irq_rings)) { + irq_enable(engine); + enabled = true; } enable_fake_irq(b); - return true; + return enabled; } static inline struct intel_wait *to_wait(struct rb_node *node) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index a7c95f0c3101a..e5a0d489a304a 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -339,9 +339,9 @@ struct intel_engine_cs { struct timer_list hangcheck; /* detect missed interrupts */ unsigned int hangcheck_interrupts; + unsigned int irq_enabled; bool irq_armed : 1; - bool irq_enabled : 1; I915_SELFTEST_DECLARE(bool mock : 1); } breadcrumbs; @@ -848,6 +848,9 @@ unsigned int intel_engine_wakeup(struct intel_engine_cs *engine); #define ENGINE_WAKEUP_WAITER BIT(0) #define ENGINE_WAKEUP_ASLEEP BIT(1) +void intel_engine_pin_breadcrumbs_irq(struct intel_engine_cs *engine); +void intel_engine_unpin_breadcrumbs_irq(struct intel_engine_cs *engine); + void __intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine); -- GitLab From 9bdc3573a5a2824da7d7e55eac95ca5a014181d8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 18:25:19 +0100 Subject: [PATCH 0033/1471] drm/i915/guc: Initialize GuC before restarting engines MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that we're handling request resubmission the same way as regular submission (from the tasklet), we can move GuC initialization earlier, before restarting the engines. This way, we're no longer being in the state of flux during engine restart - we're already in user requested submission mode. Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Michal Wajdeczko Cc: Michel Thierry Cc: Mika Kuoppala Cc: Oscar Mateo Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025172519.10670-5-chris@chris-wilson.co.uk Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_gem.c | 10 +++++----- drivers/gpu/drm/i915/i915_guc_submission.c | 8 -------- drivers/gpu/drm/i915/intel_lrc.c | 2 +- 3 files changed, 6 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 8364304e9d2b2..d803ef5f4a7f2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4926,6 +4926,11 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) goto out; } + /* We can't enable contexts until all firmware is loaded */ + ret = intel_uc_init_hw(dev_priv); + if (ret) + goto out; + /* Need to do basic initialisation of all rings first: */ ret = __i915_gem_restart_engines(dev_priv); if (ret) @@ -4933,11 +4938,6 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) intel_mocs_init_l3cc_table(dev_priv); - /* We can't enable contexts until all firmware is loaded */ - ret = intel_uc_init_hw(dev_priv); - if (ret) - goto out; - out: intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); return ret; diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 141ed9df3d5ca..07289a4b5bfa5 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -1132,15 +1132,7 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) for_each_engine(engine, dev_priv, id) { struct intel_engine_execlists * const execlists = &engine->execlists; - /* The tasklet was initialised by execlists, and may be in - * a state of flux (across a reset) and so we just want to - * take over the callback without changing any other state - * in the tasklet. - */ execlists->irq_tasklet.func = i915_guc_irq_handler; - clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted); - tasklet_schedule(&execlists->irq_tasklet); - engine->park = i915_guc_submission_park; engine->unpark = i915_guc_submission_unpark; } diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index eeb3622803a8f..e821c1eba7a44 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1474,7 +1474,7 @@ static int gen8_init_common_ring(struct intel_engine_cs *engine) execlists->active = 0; /* After a GPU reset, we may have requests to replay */ - if (!i915_modparams.enable_guc_submission && execlists->first) + if (execlists->first) tasklet_schedule(&execlists->irq_tasklet); return 0; -- GitLab From f8c3dcf946bfb1e0519fb095d02f9b7def30a749 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 25 Oct 2017 17:15:46 -0700 Subject: [PATCH 0034/1471] drm/i915/cnl: Fix SSEU Device Status. CNL adds an extra register for slice/subslice information. Although no SKU is planed with an extra slice let's already handle this extra piece of information so we don't have the risk in future of getting a part that might have chosen this part of the die instead of other slices or anything like that. Also if subslice is disabled the information of eu ack for that is garbage, so let's skip checks for eu if subslice is disabled as we skip the subslice if slice is disabled. The rest is pretty much like gen9. v2: Remove IS_CANNONLAKE from gen9 status function. v3: Consider s_max = 6 and ss_max=4 to run over all possible slices and subslices possible by spec. Although no real hardware will have that many slices/subslices. To match with sseu info init. v4: Fix offset calculation for slices 4 and 5. Removed Oscar's rv-b since this change also needs review. v5: Let's consider only valid bits for SLICE*_PGCTL_ACK. This looks like wrong in Spec, but seems to be enough for now. Whenever Spec gets updated and fixed we come back and properly update the masks. Also add a FIXME, so we can revisit this later when we find some strange info on debugfs or when we noitce spec got updated. Cc: Oscar Mateo Signed-off-by: Rodrigo Vivi Reviewed-by: Lionel Landwerlin Link: https://patchwork.freedesktop.org/patch/msgid/20171026001546.28203-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/i915_debugfs.c | 61 ++++++++++++++++++++++++++++- drivers/gpu/drm/i915/i915_reg.h | 7 ++++ 2 files changed, 66 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index c65e381b85f39..34862aa8f9c0b 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -4448,6 +4448,61 @@ static void cherryview_sseu_device_status(struct drm_i915_private *dev_priv, } } +static void gen10_sseu_device_status(struct drm_i915_private *dev_priv, + struct sseu_dev_info *sseu) +{ + const struct intel_device_info *info = INTEL_INFO(dev_priv); + int s_max = 6, ss_max = 4; + int s, ss; + u32 s_reg[s_max], eu_reg[2 * s_max], eu_mask[2]; + + for (s = 0; s < s_max; s++) { + /* + * FIXME: Valid SS Mask respects the spec and read + * only valid bits for those registers, excluding reserverd + * although this seems wrong because it would leave many + * subslices without ACK. + */ + s_reg[s] = I915_READ(GEN10_SLICE_PGCTL_ACK(s)) & + GEN10_PGCTL_VALID_SS_MASK(s); + eu_reg[2 * s] = I915_READ(GEN10_SS01_EU_PGCTL_ACK(s)); + eu_reg[2 * s + 1] = I915_READ(GEN10_SS23_EU_PGCTL_ACK(s)); + } + + eu_mask[0] = GEN9_PGCTL_SSA_EU08_ACK | + GEN9_PGCTL_SSA_EU19_ACK | + GEN9_PGCTL_SSA_EU210_ACK | + GEN9_PGCTL_SSA_EU311_ACK; + eu_mask[1] = GEN9_PGCTL_SSB_EU08_ACK | + GEN9_PGCTL_SSB_EU19_ACK | + GEN9_PGCTL_SSB_EU210_ACK | + GEN9_PGCTL_SSB_EU311_ACK; + + for (s = 0; s < s_max; s++) { + if ((s_reg[s] & GEN9_PGCTL_SLICE_ACK) == 0) + /* skip disabled slice */ + continue; + + sseu->slice_mask |= BIT(s); + sseu->subslice_mask = info->sseu.subslice_mask; + + for (ss = 0; ss < ss_max; ss++) { + unsigned int eu_cnt; + + if (!(s_reg[s] & (GEN9_PGCTL_SS_ACK(ss)))) + /* skip disabled subslice */ + continue; + + eu_cnt = 2 * hweight32(eu_reg[2 * s + ss / 2] & + eu_mask[ss % 2]); + sseu->eu_total += eu_cnt; + sseu->eu_per_subslice = max_t(unsigned int, + sseu->eu_per_subslice, + eu_cnt); + } + } +} + static void gen9_sseu_device_status(struct drm_i915_private *dev_priv, struct sseu_dev_info *sseu) { @@ -4483,7 +4538,7 @@ static void gen9_sseu_device_status(struct drm_i915_private *dev_priv, sseu->slice_mask |= BIT(s); - if (IS_GEN9_BC(dev_priv) || IS_CANNONLAKE(dev_priv)) + if (IS_GEN9_BC(dev_priv)) sseu->subslice_mask = INTEL_INFO(dev_priv)->sseu.subslice_mask; @@ -4589,8 +4644,10 @@ static int i915_sseu_status(struct seq_file *m, void *unused) cherryview_sseu_device_status(dev_priv, &sseu); } else if (IS_BROADWELL(dev_priv)) { broadwell_sseu_device_status(dev_priv, &sseu); - } else if (INTEL_GEN(dev_priv) >= 9) { + } else if (IS_GEN9(dev_priv)) { gen9_sseu_device_status(dev_priv, &sseu); + } else if (INTEL_GEN(dev_priv) >= 10) { + gen10_sseu_device_status(dev_priv, &sseu); } intel_runtime_pm_put(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index f138eae82bf0c..8c775e96b4e4f 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -8037,11 +8037,18 @@ enum { #define CHV_EU311_PG_ENABLE (1<<1) #define GEN9_SLICE_PGCTL_ACK(slice) _MMIO(0x804c + (slice)*0x4) +#define GEN10_SLICE_PGCTL_ACK(slice) _MMIO(0x804c + ((slice) / 3) * 0x34 + \ + ((slice) % 3) * 0x4) #define GEN9_PGCTL_SLICE_ACK (1 << 0) #define GEN9_PGCTL_SS_ACK(subslice) (1 << (2 + (subslice)*2)) +#define GEN10_PGCTL_VALID_SS_MASK(slice) ((slice) == 0 ? 0x7F : 0x1F) #define GEN9_SS01_EU_PGCTL_ACK(slice) _MMIO(0x805c + (slice)*0x8) +#define GEN10_SS01_EU_PGCTL_ACK(slice) _MMIO(0x805c + ((slice) / 3) * 0x30 + \ + ((slice) % 3) * 0x8) #define GEN9_SS23_EU_PGCTL_ACK(slice) _MMIO(0x8060 + (slice)*0x8) +#define GEN10_SS23_EU_PGCTL_ACK(slice) _MMIO(0x8060 + ((slice) / 3) * 0x30 + \ + ((slice) % 3) * 0x8) #define GEN9_PGCTL_SSA_EU08_ACK (1 << 0) #define GEN9_PGCTL_SSA_EU19_ACK (1 << 2) #define GEN9_PGCTL_SSA_EU210_ACK (1 << 4) -- GitLab From 21e8860ef467001330368eceeb855fde46ab1dec Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:09 +0200 Subject: [PATCH 0035/1471] drm/i915/guc: Do not use 0 for GuC doorbell cookie MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Apparently, this value is reserved and may be interpreted as changing doorbell ownership. Even though we're not observing any side effects now, let's skip over it to be consistent with the spec. v2: Apply checkpatch (Sagar) Suggested-by: Sagar Arun Kamble Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-2-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_guc_submission.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 07289a4b5bfa5..0b61bf18c3f43 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -475,9 +475,12 @@ static void guc_ring_doorbell(struct i915_guc_client *client) /* pointer of current doorbell cacheline */ db = __get_doorbell(client); - /* we're not expecting the doorbell cookie to change behind our back */ + /* + * We're not expecting the doorbell cookie to change behind our back, + * we also need to treat 0 as a reserved value. + */ cookie = READ_ONCE(db->cookie); - WARN_ON_ONCE(xchg(&db->cookie, cookie + 1) != cookie); + WARN_ON_ONCE(xchg(&db->cookie, cookie + 1 ?: cookie + 2) != cookie); /* XXX: doorbell was lost and need to acquire it again */ GEM_BUG_ON(db->db_status != GUC_DOORBELL_ENABLED); -- GitLab From 89922d0145755b0d22906925e9bac9dc78524fc3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:10 +0200 Subject: [PATCH 0036/1471] drm/i915/guc: Extract GuC stage desc pool creation into a helper MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since it's a two-step process, we can have a cleaner error handling in the caller if we do the allocations in a helper. Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Daniele Ceraolo Spurio Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Daniele Ceraolo Spurio Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-3-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_guc_submission.c | 65 +++++++++++++--------- 1 file changed, 38 insertions(+), 27 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 0b61bf18c3f43..e195bdee04735 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -311,6 +311,37 @@ static void guc_proc_desc_init(struct intel_guc *guc, desc->priority = client->priority; } +static int guc_stage_desc_pool_create(struct intel_guc *guc) +{ + struct i915_vma *vma; + void *vaddr; + + vma = intel_guc_allocate_vma(guc, + PAGE_ALIGN(sizeof(struct guc_stage_desc) * + GUC_MAX_STAGE_DESCRIPTORS)); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + i915_vma_unpin_and_release(&vma); + return PTR_ERR(vaddr); + } + + guc->stage_desc_pool = vma; + guc->stage_desc_pool_vaddr = vaddr; + ida_init(&guc->stage_ids); + + return 0; +} + +static void guc_stage_desc_pool_destroy(struct intel_guc *guc) +{ + ida_destroy(&guc->stage_ids); + i915_gem_object_unpin_map(guc->stage_desc_pool->obj); + i915_vma_unpin_and_release(&guc->stage_desc_pool); +} + /* * Initialise/clear the stage descriptor shared with the GuC firmware. * @@ -953,47 +984,29 @@ static void guc_ads_destroy(struct intel_guc *guc) int i915_guc_submission_init(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; - struct i915_vma *vma; - void *vaddr; int ret; if (guc->stage_desc_pool) return 0; - vma = intel_guc_allocate_vma(guc, - PAGE_ALIGN(sizeof(struct guc_stage_desc) * - GUC_MAX_STAGE_DESCRIPTORS)); - if (IS_ERR(vma)) - return PTR_ERR(vma); - - guc->stage_desc_pool = vma; - - vaddr = i915_gem_object_pin_map(guc->stage_desc_pool->obj, I915_MAP_WB); - if (IS_ERR(vaddr)) { - ret = PTR_ERR(vaddr); - goto err_vma; - } - - guc->stage_desc_pool_vaddr = vaddr; + ret = guc_stage_desc_pool_create(guc); + if (ret) + return ret; ret = intel_guc_log_create(guc); if (ret < 0) - goto err_vaddr; + goto err_stage_desc_pool; ret = guc_ads_create(guc); if (ret < 0) goto err_log; - ida_init(&guc->stage_ids); - return 0; err_log: intel_guc_log_destroy(guc); -err_vaddr: - i915_gem_object_unpin_map(guc->stage_desc_pool->obj); -err_vma: - i915_vma_unpin_and_release(&guc->stage_desc_pool); +err_stage_desc_pool: + guc_stage_desc_pool_destroy(guc); return ret; } @@ -1001,11 +1014,9 @@ void i915_guc_submission_fini(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; - ida_destroy(&guc->stage_ids); guc_ads_destroy(guc); intel_guc_log_destroy(guc); - i915_gem_object_unpin_map(guc->stage_desc_pool->obj); - i915_vma_unpin_and_release(&guc->stage_desc_pool); + guc_stage_desc_pool_destroy(guc); } static void guc_interrupts_capture(struct drm_i915_private *dev_priv) -- GitLab From b8e5eb960b28bc120b03f174dfa3ac1e32e1d10f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:11 +0200 Subject: [PATCH 0037/1471] drm/i915/guc: Allocate separate shared data object for GuC communication MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We were using first page of kernel context render state for sharing data with GuC. While it's justified by the fact that those pages are not used (note, GuC still enforces this layout and refuses to work if we remove the extra page in front), it's also confusing (why are we using this particular page?). Let's allocate a separate object instead. v2: Drop kernel_context from GuC suspend/resume action handlers (Michel) Suggested-by: Daniele Ceraolo Spurio Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Daniele Ceraolo Spurio Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Daniele Ceraolo Spurio Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-4-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_guc_submission.c | 36 +++++++++++++++++++++- drivers/gpu/drm/i915/intel_guc.c | 14 ++------- drivers/gpu/drm/i915/intel_guc.h | 2 ++ 3 files changed, 39 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index e195bdee04735..d1a5613da24c6 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -437,6 +437,33 @@ static void guc_stage_desc_fini(struct intel_guc *guc, memset(desc, 0, sizeof(*desc)); } +static int guc_shared_data_create(struct intel_guc *guc) +{ + struct i915_vma *vma; + void *vaddr; + + vma = intel_guc_allocate_vma(guc, PAGE_SIZE); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + i915_vma_unpin_and_release(&vma); + return PTR_ERR(vaddr); + } + + guc->shared_data = vma; + guc->shared_data_vaddr = vaddr; + + return 0; +} + +static void guc_shared_data_destroy(struct intel_guc *guc) +{ + i915_gem_object_unpin_map(guc->shared_data->obj); + i915_vma_unpin_and_release(&guc->shared_data); +} + /* Construct a Work Item and append it to the GuC's Work Queue */ static void guc_wq_item_append(struct i915_guc_client *client, struct drm_i915_gem_request *rq) @@ -993,9 +1020,13 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) if (ret) return ret; + ret = guc_shared_data_create(guc); + if (ret) + goto err_stage_desc_pool; + ret = intel_guc_log_create(guc); if (ret < 0) - goto err_stage_desc_pool; + goto err_shared_data; ret = guc_ads_create(guc); if (ret < 0) @@ -1005,6 +1036,8 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) err_log: intel_guc_log_destroy(guc); +err_shared_data: + guc_shared_data_destroy(guc); err_stage_desc_pool: guc_stage_desc_pool_destroy(guc); return ret; @@ -1016,6 +1049,7 @@ void i915_guc_submission_fini(struct drm_i915_private *dev_priv) guc_ads_destroy(guc); intel_guc_log_destroy(guc); + guc_shared_data_destroy(guc); guc_stage_desc_pool_destroy(guc); } diff --git a/drivers/gpu/drm/i915/intel_guc.c b/drivers/gpu/drm/i915/intel_guc.c index 10037c0fdf952..f74d50fdaeb07 100644 --- a/drivers/gpu/drm/i915/intel_guc.c +++ b/drivers/gpu/drm/i915/intel_guc.c @@ -268,7 +268,6 @@ int intel_guc_auth_huc(struct intel_guc *guc, u32 rsa_offset) int intel_guc_suspend(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; - struct i915_gem_context *ctx; u32 data[3]; if (guc->fw.load_status != INTEL_UC_FIRMWARE_SUCCESS) @@ -276,14 +275,10 @@ int intel_guc_suspend(struct drm_i915_private *dev_priv) gen9_disable_guc_interrupts(dev_priv); - ctx = dev_priv->kernel_context; - data[0] = INTEL_GUC_ACTION_ENTER_S_STATE; /* any value greater than GUC_POWER_D0 */ data[1] = GUC_POWER_D1; - /* first page is shared data with GuC */ - data[2] = guc_ggtt_offset(ctx->engine[RCS].state) + - LRC_GUCSHR_PN * PAGE_SIZE; + data[2] = guc_ggtt_offset(guc->shared_data); return intel_guc_send(guc, data, ARRAY_SIZE(data)); } @@ -295,7 +290,6 @@ int intel_guc_suspend(struct drm_i915_private *dev_priv) int intel_guc_resume(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; - struct i915_gem_context *ctx; u32 data[3]; if (guc->fw.load_status != INTEL_UC_FIRMWARE_SUCCESS) @@ -304,13 +298,9 @@ int intel_guc_resume(struct drm_i915_private *dev_priv) if (i915_modparams.guc_log_level >= 0) gen9_enable_guc_interrupts(dev_priv); - ctx = dev_priv->kernel_context; - data[0] = INTEL_GUC_ACTION_EXIT_S_STATE; data[1] = GUC_POWER_D0; - /* first page is shared data with GuC */ - data[2] = guc_ggtt_offset(ctx->engine[RCS].state) + - LRC_GUCSHR_PN * PAGE_SIZE; + data[2] = guc_ggtt_offset(guc->shared_data); return intel_guc_send(guc, data, ARRAY_SIZE(data)); } diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index 418450b1ae27a..aa1583167b0ac 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -54,6 +54,8 @@ struct intel_guc { struct i915_vma *stage_desc_pool; void *stage_desc_pool_vaddr; struct ida stage_ids; + struct i915_vma *shared_data; + void *shared_data_vaddr; struct i915_guc_client *execbuf_client; -- GitLab From 4ddbe87a16d75e8f8cb59722fc385b3469a219c3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:12 +0200 Subject: [PATCH 0038/1471] drm/i915/guc: Add preemption action to GuC firmware interface MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We're using GuC action to request preemption. However, after requesting preemption we need to wait for GuC to finish its own post-processing before we start submitting our requests. Firmware is using shared context to report its status. Let's update GuC firmware interface with those new definitions. v2: Drop unused INTEL_GUC_PREEMPT_OPTION_IMMEDIATE Signed-off-by: Michał Winiarski Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Jeff McGee Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-5-michal.winiarski@intel.com --- drivers/gpu/drm/i915/intel_guc_fwif.h | 39 +++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_guc_fwif.h b/drivers/gpu/drm/i915/intel_guc_fwif.h index 80c5074354581..e24dbec2ead8c 100644 --- a/drivers/gpu/drm/i915/intel_guc_fwif.h +++ b/drivers/gpu/drm/i915/intel_guc_fwif.h @@ -544,9 +544,36 @@ union guc_log_control { u32 value; } __packed; +struct guc_ctx_report { + u32 report_return_status; + u32 reserved1[64]; + u32 affected_count; + u32 reserved2[2]; +} __packed; + +/* GuC Shared Context Data Struct */ +struct guc_shared_ctx_data { + u32 addr_of_last_preempted_data_low; + u32 addr_of_last_preempted_data_high; + u32 addr_of_last_preempted_data_high_tmp; + u32 padding; + u32 is_mapped_to_proxy; + u32 proxy_ctx_id; + u32 engine_reset_ctx_id; + u32 media_reset_count; + u32 reserved1[8]; + u32 uk_last_ctx_switch_reason; + u32 was_reset; + u32 lrca_gpu_addr; + u64 execlist_ctx; + u32 reserved2[66]; + struct guc_ctx_report preempt_ctx_report[GUC_MAX_ENGINES_NUM]; +} __packed; + /* This Action will be programmed in C180 - SOFT_SCRATCH_O_REG */ enum intel_guc_action { INTEL_GUC_ACTION_DEFAULT = 0x0, + INTEL_GUC_ACTION_REQUEST_PREEMPTION = 0x2, INTEL_GUC_ACTION_SAMPLE_FORCEWAKE = 0x6, INTEL_GUC_ACTION_ALLOCATE_DOORBELL = 0x10, INTEL_GUC_ACTION_DEALLOCATE_DOORBELL = 0x20, @@ -562,6 +589,18 @@ enum intel_guc_action { INTEL_GUC_ACTION_LIMIT }; +enum intel_guc_preempt_options { + INTEL_GUC_PREEMPT_OPTION_DROP_WORK_Q = 0x4, + INTEL_GUC_PREEMPT_OPTION_DROP_SUBMIT_Q = 0x8, +}; + +enum intel_guc_report_status { + INTEL_GUC_REPORT_STATUS_UNKNOWN = 0x0, + INTEL_GUC_REPORT_STATUS_ACKED = 0x1, + INTEL_GUC_REPORT_STATUS_ERROR = 0x2, + INTEL_GUC_REPORT_STATUS_COMPLETE = 0x4, +}; + /* * The GuC sends its response to a command by overwriting the * command in SS0. The response is distinguishable from a command -- GitLab From e12ab16907252ebc3ec956932eb40ed88105346b Mon Sep 17 00:00:00 2001 From: Dave Gordon Date: Thu, 26 Oct 2017 16:17:37 +0200 Subject: [PATCH 0039/1471] drm/i915/guc: Add a second client, to be used for preemption MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This second client is created with priority KMD_HIGH, and marked as preemptive. This will allow us to request preemption using GuC actions. v2: Extract clients creation into a helper, debugfs fixups. (Michał) Recreate doorbell on init. (Daniele) Move clients into an array. v3: And move clients back from an array, to get rid of the enum (Michał) v4: Use is_high_priority, move DRM_ERROR into __create_doorbell, move GEM_BUG_ON inside guc_clients_create (Michał) v5: Split the BUG_ON (Michał) v6: Cleanup after error during doorbell reinit (Michał) Signed-off-by: Dave Gordon Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Daniele Ceraolo Spurio Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Michal Wajdeczko Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171026141737.31656-1-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_debugfs.c | 2 + drivers/gpu/drm/i915/i915_guc_submission.c | 118 +++++++++++++++------ drivers/gpu/drm/i915/intel_guc.h | 1 + 3 files changed, 86 insertions(+), 35 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 34862aa8f9c0b..c1e5bba91bff8 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2484,6 +2484,8 @@ static int i915_guc_info(struct seq_file *m, void *data) seq_printf(m, "\nGuC execbuf client @ %p:\n", guc->execbuf_client); i915_guc_client_info(m, dev_priv, guc->execbuf_client); + seq_printf(m, "\nGuC preempt client @ %p:\n", guc->preempt_client); + i915_guc_client_info(m, dev_priv, guc->preempt_client); i915_guc_log_info(m, dev_priv); diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index d1a5613da24c6..b0f034f924319 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -33,10 +33,11 @@ * * GuC client: * A i915_guc_client refers to a submission path through GuC. Currently, there - * is only one of these (the execbuf_client) and this one is charged with all - * submissions to the GuC. This struct is the owner of a doorbell, a process - * descriptor and a workqueue (all of them inside a single gem object that - * contains all required pages for these elements). + * are two clients. One of them (the execbuf_client) is charged with all + * submissions to the GuC, the other one (preempt_client) is responsible for + * preempting the execbuf_client. This struct is the owner of a doorbell, a + * process descriptor and a workqueue (all of them inside a single gem object + * that contains all required pages for these elements). * * GuC stage descriptor: * During initialization, the driver allocates a static pool of 1024 such @@ -83,7 +84,8 @@ static inline bool is_high_priority(struct i915_guc_client* client) { - return client->priority <= GUC_CLIENT_PRIORITY_HIGH; + return (client->priority == GUC_CLIENT_PRIORITY_KMD_HIGH || + client->priority == GUC_CLIENT_PRIORITY_HIGH); } static int __reserve_doorbell(struct i915_guc_client *client) @@ -196,8 +198,11 @@ static int __create_doorbell(struct i915_guc_client *client) doorbell->cookie = 0; err = __guc_allocate_doorbell(client->guc, client->stage_id); - if (err) + if (err) { doorbell->db_status = GUC_DOORBELL_DISABLED; + DRM_ERROR("Couldn't create client %u doorbell: %d\n", + client->stage_id, err); + } return err; } @@ -363,6 +368,8 @@ static void guc_stage_desc_init(struct intel_guc *guc, memset(desc, 0, sizeof(*desc)); desc->attribute = GUC_STAGE_DESC_ATTR_ACTIVE | GUC_STAGE_DESC_ATTR_KERNEL; + if (is_high_priority(client)) + desc->attribute |= GUC_STAGE_DESC_ATTR_PREEMPT; desc->stage_id = client->stage_id; desc->priority = client->priority; desc->db_id = client->doorbell_id; @@ -763,14 +770,14 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) /* Now for every client (and not only execbuf_client) make sure their * doorbells are known by the GuC */ - //for (client = client_list; client != NULL; client = client->next) - { - ret = __create_doorbell(client); - if (ret) { - DRM_ERROR("Couldn't recreate client %u doorbell: %d\n", - client->stage_id, ret); - return ret; - } + ret = __create_doorbell(guc->execbuf_client); + if (ret) + return ret; + + ret = __create_doorbell(guc->preempt_client); + if (ret) { + __destroy_doorbell(guc->execbuf_client); + return ret; } /* Read back & verify all (used & unused) doorbell registers */ @@ -895,6 +902,50 @@ static void guc_client_free(struct i915_guc_client *client) kfree(client); } +static int guc_clients_create(struct intel_guc *guc) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + struct i915_guc_client *client; + + GEM_BUG_ON(guc->execbuf_client); + GEM_BUG_ON(guc->preempt_client); + + client = guc_client_alloc(dev_priv, + INTEL_INFO(dev_priv)->ring_mask, + GUC_CLIENT_PRIORITY_KMD_NORMAL, + dev_priv->kernel_context); + if (IS_ERR(client)) { + DRM_ERROR("Failed to create GuC client for submission!\n"); + return PTR_ERR(client); + } + guc->execbuf_client = client; + + client = guc_client_alloc(dev_priv, + INTEL_INFO(dev_priv)->ring_mask, + GUC_CLIENT_PRIORITY_KMD_HIGH, + dev_priv->preempt_context); + if (IS_ERR(client)) { + DRM_ERROR("Failed to create GuC client for preemption!\n"); + guc_client_free(guc->execbuf_client); + guc->execbuf_client = NULL; + return PTR_ERR(client); + } + guc->preempt_client = client; + + return 0; +} + +static void guc_clients_destroy(struct intel_guc *guc) +{ + struct i915_guc_client *client; + + client = fetch_and_zero(&guc->execbuf_client); + guc_client_free(client); + + client = fetch_and_zero(&guc->preempt_client); + guc_client_free(client); +} + static void guc_policy_init(struct guc_policy *policy) { policy->execution_quantum = POLICY_DEFAULT_EXECUTION_QUANTUM_US; @@ -1134,7 +1185,6 @@ static void i915_guc_submission_unpark(struct intel_engine_cs *engine) int i915_guc_submission_enable(struct drm_i915_private *dev_priv) { struct intel_guc *guc = &dev_priv->guc; - struct i915_guc_client *client = guc->execbuf_client; struct intel_engine_cs *engine; enum intel_engine_id id; int err; @@ -1152,28 +1202,28 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) sizeof(struct guc_wq_item) * I915_NUM_ENGINES > GUC_WQ_SIZE); - if (!client) { - client = guc_client_alloc(dev_priv, - INTEL_INFO(dev_priv)->ring_mask, - GUC_CLIENT_PRIORITY_KMD_NORMAL, - dev_priv->kernel_context); - if (IS_ERR(client)) { - DRM_ERROR("Failed to create GuC client for execbuf!\n"); - return PTR_ERR(client); - } - - guc->execbuf_client = client; + /* + * We're being called on both module initialization and on reset, + * until this flow is changed, we're using regular client presence to + * determine which case are we in, and whether we should allocate new + * clients or just reset their workqueues. + */ + if (!guc->execbuf_client) { + err = guc_clients_create(guc); + if (err) + return err; + } else { + guc_reset_wq(guc->execbuf_client); + guc_reset_wq(guc->preempt_client); } err = intel_guc_sample_forcewake(guc); if (err) - goto err_execbuf_client; - - guc_reset_wq(client); + goto err_free_clients; err = guc_init_doorbell_hw(guc); if (err) - goto err_execbuf_client; + goto err_free_clients; /* Take over from manual control of ELSP (execlists) */ guc_interrupts_capture(dev_priv); @@ -1187,9 +1237,8 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) return 0; -err_execbuf_client: - guc_client_free(guc->execbuf_client); - guc->execbuf_client = NULL; +err_free_clients: + guc_clients_destroy(guc); return err; } @@ -1204,6 +1253,5 @@ void i915_guc_submission_disable(struct drm_i915_private *dev_priv) /* Revert back to manual ELSP submission */ intel_engines_reset_default_submission(dev_priv); - guc_client_free(guc->execbuf_client); - guc->execbuf_client = NULL; + guc_clients_destroy(guc); } diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index aa1583167b0ac..6ca55c5bed3c2 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -58,6 +58,7 @@ struct intel_guc { void *shared_data_vaddr; struct i915_guc_client *execbuf_client; + struct i915_guc_client *preempt_client; DECLARE_BITMAP(doorbell_bitmap, GUC_NUM_DOORBELLS); /* Cyclic counter mod pagesize */ -- GitLab From a0991e1d2801c5d3e7d98abdefcb8032e09bde6d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:14 +0200 Subject: [PATCH 0040/1471] drm/i915/guc: Split guc_wq_item_append MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We're using a special preempt context for HW to preempt into. We don't want to emit any requests there, but we still need to wrap this context into a valid GuC work item. Let's cleanup the functions operating on GuC work items. We can extract guc_request_add - responsible for adding GuC work item and ringing the doorbell, and guc_wq_item_append - used by the function above, not tied to the concept of gem request. Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Chris Wilson Reviewed-by: Jeff McGee Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-7-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_guc_submission.c | 55 ++++++++++++---------- 1 file changed, 30 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index b0f034f924319..301c658a4cdea 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -71,7 +71,7 @@ * WQ_TYPE_INORDER is needed to support legacy submission via GuC, which * represents in-order queue. The kernel driver packs ring tail pointer and an * ELSP context descriptor dword into Work Item. - * See guc_wq_item_append() + * See guc_add_request() * * ADS: * The Additional Data Struct (ADS) has pointers for different buffers used by @@ -394,7 +394,7 @@ static void guc_stage_desc_init(struct intel_guc *guc, * submission or, in other words, not using a direct submission * model) the KMD's LRCA is not used for any work submission. * Instead, the GuC uses the LRCA of the user mode context (see - * guc_wq_item_append below). + * guc_add_request below). */ lrc->context_desc = lower_32_bits(ce->lrc_desc); @@ -473,22 +473,18 @@ static void guc_shared_data_destroy(struct intel_guc *guc) /* Construct a Work Item and append it to the GuC's Work Queue */ static void guc_wq_item_append(struct i915_guc_client *client, - struct drm_i915_gem_request *rq) + u32 target_engine, u32 context_desc, + u32 ring_tail, u32 fence_id) { /* wqi_len is in DWords, and does not include the one-word header */ const size_t wqi_size = sizeof(struct guc_wq_item); const u32 wqi_len = wqi_size / sizeof(u32) - 1; - struct intel_engine_cs *engine = rq->engine; - struct i915_gem_context *ctx = rq->ctx; struct guc_process_desc *desc = __get_process_desc(client); struct guc_wq_item *wqi; - u32 ring_tail, wq_off; + u32 wq_off; lockdep_assert_held(&client->wq_lock); - ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64); - GEM_BUG_ON(ring_tail > WQ_RING_TAIL_MAX); - /* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we * should not have the case where structure wqi is across page, neither * wrapped to the beginning. This simplifies the implementation below. @@ -510,15 +506,14 @@ static void guc_wq_item_append(struct i915_guc_client *client, /* Now fill in the 4-word work queue item */ wqi->header = WQ_TYPE_INORDER | (wqi_len << WQ_LEN_SHIFT) | - (engine->guc_id << WQ_TARGET_SHIFT) | + (target_engine << WQ_TARGET_SHIFT) | WQ_NO_WCFLUSH_WAIT; - - wqi->context_desc = lower_32_bits(intel_lr_context_descriptor(ctx, engine)); - + wqi->context_desc = context_desc; wqi->submit_element_info = ring_tail << WQ_RING_TAIL_SHIFT; - wqi->fence_id = rq->global_seqno; + GEM_BUG_ON(ring_tail > WQ_RING_TAIL_MAX); + wqi->fence_id = fence_id; - /* Postincrement WQ tail for next time. */ + /* Make the update visible to GuC */ WRITE_ONCE(desc->tail, (wq_off + wqi_size) & (GUC_WQ_SIZE - 1)); } @@ -551,6 +546,25 @@ static void guc_ring_doorbell(struct i915_guc_client *client) GEM_BUG_ON(db->db_status != GUC_DOORBELL_ENABLED); } +static void guc_add_request(struct intel_guc *guc, + struct drm_i915_gem_request *rq) +{ + struct i915_guc_client *client = guc->execbuf_client; + struct intel_engine_cs *engine = rq->engine; + u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(rq->ctx, engine)); + u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64); + + spin_lock(&client->wq_lock); + + guc_wq_item_append(client, engine->guc_id, ctx_desc, + ring_tail, rq->global_seqno); + guc_ring_doorbell(client); + + client->submissions[engine->id] += 1; + + spin_unlock(&client->wq_lock); +} + /** * i915_guc_submit() - Submit commands through GuC * @engine: engine associated with the commands @@ -562,10 +576,8 @@ static void i915_guc_submit(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; struct intel_guc *guc = &dev_priv->guc; - struct i915_guc_client *client = guc->execbuf_client; struct intel_engine_execlists * const execlists = &engine->execlists; struct execlist_port *port = execlists->port; - const unsigned int engine_id = engine->id; unsigned int n; for (n = 0; n < execlists_num_ports(execlists); n++) { @@ -579,14 +591,7 @@ static void i915_guc_submit(struct intel_engine_cs *engine) if (i915_vma_is_map_and_fenceable(rq->ring->vma)) POSTING_READ_FW(GUC_STATUS); - spin_lock(&client->wq_lock); - - guc_wq_item_append(client, rq); - guc_ring_doorbell(client); - - client->submissions[engine_id] += 1; - - spin_unlock(&client->wq_lock); + guc_add_request(guc, rq); } } } -- GitLab From df77cd83d5566ef6a04528bb08732ccbf829580e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:15 +0200 Subject: [PATCH 0041/1471] drm/i915: Extract "emit write" part of emit breadcrumb functions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Let's separate the "emit" part from touching any internal structures, this way we can have a generic "emit coherent GGTT write" function. We would like to reuse this functionality for emitting HWSP write, to confirm that preempt-to-idle has finished. v2: Reorder args to match emit_pipe_control, s/render/rcs (Chris) Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Tvrtko Ursulin Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-8-michal.winiarski@intel.com --- drivers/gpu/drm/i915/intel_lrc.c | 28 ++++++------------ drivers/gpu/drm/i915/intel_ringbuffer.h | 38 +++++++++++++++++++++++++ 2 files changed, 46 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index e821c1eba7a44..599c709fc5a76 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1794,10 +1794,8 @@ static void gen8_emit_breadcrumb(struct drm_i915_gem_request *request, u32 *cs) /* w/a: bit 5 needs to be zero for MI_FLUSH_DW address. */ BUILD_BUG_ON(I915_GEM_HWS_INDEX_ADDR & (1 << 5)); - *cs++ = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW; - *cs++ = intel_hws_seqno_address(request->engine) | MI_FLUSH_DW_USE_GTT; - *cs++ = 0; - *cs++ = request->global_seqno; + cs = gen8_emit_ggtt_write(cs, request->global_seqno, + intel_hws_seqno_address(request->engine)); *cs++ = MI_USER_INTERRUPT; *cs++ = MI_NOOP; request->tail = intel_ring_offset(request, cs); @@ -1807,24 +1805,14 @@ static void gen8_emit_breadcrumb(struct drm_i915_gem_request *request, u32 *cs) } static const int gen8_emit_breadcrumb_sz = 6 + WA_TAIL_DWORDS; -static void gen8_emit_breadcrumb_render(struct drm_i915_gem_request *request, +static void gen8_emit_breadcrumb_rcs(struct drm_i915_gem_request *request, u32 *cs) { /* We're using qword write, seqno should be aligned to 8 bytes. */ BUILD_BUG_ON(I915_GEM_HWS_INDEX & 1); - /* w/a for post sync ops following a GPGPU operation we - * need a prior CS_STALL, which is emitted by the flush - * following the batch. - */ - *cs++ = GFX_OP_PIPE_CONTROL(6); - *cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL | - PIPE_CONTROL_QW_WRITE; - *cs++ = intel_hws_seqno_address(request->engine); - *cs++ = 0; - *cs++ = request->global_seqno; - /* We're thrashing one dword of HWS. */ - *cs++ = 0; + cs = gen8_emit_ggtt_write_rcs(cs, request->global_seqno, + intel_hws_seqno_address(request->engine)); *cs++ = MI_USER_INTERRUPT; *cs++ = MI_NOOP; request->tail = intel_ring_offset(request, cs); @@ -1832,7 +1820,7 @@ static void gen8_emit_breadcrumb_render(struct drm_i915_gem_request *request, gen8_emit_wa_tail(request, cs); } -static const int gen8_emit_breadcrumb_render_sz = 8 + WA_TAIL_DWORDS; +static const int gen8_emit_breadcrumb_rcs_sz = 8 + WA_TAIL_DWORDS; static int gen8_init_rcs_context(struct drm_i915_gem_request *req) { @@ -1991,8 +1979,8 @@ int logical_render_ring_init(struct intel_engine_cs *engine) engine->init_hw = gen8_init_render_ring; engine->init_context = gen8_init_rcs_context; engine->emit_flush = gen8_emit_flush_render; - engine->emit_breadcrumb = gen8_emit_breadcrumb_render; - engine->emit_breadcrumb_sz = gen8_emit_breadcrumb_render_sz; + engine->emit_breadcrumb = gen8_emit_breadcrumb_rcs; + engine->emit_breadcrumb_sz = gen8_emit_breadcrumb_rcs_sz; ret = intel_engine_create_scratch(engine, PAGE_SIZE); if (ret) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index e5a0d489a304a..c15161e569645 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -869,6 +869,44 @@ static inline u32 *gen8_emit_pipe_control(u32 *batch, u32 flags, u32 offset) return batch + 6; } +static inline u32 * +gen8_emit_ggtt_write_rcs(u32 *cs, u32 value, u32 gtt_offset) +{ + /* We're using qword write, offset should be aligned to 8 bytes. */ + GEM_BUG_ON(!IS_ALIGNED(gtt_offset, 8)); + + /* w/a for post sync ops following a GPGPU operation we + * need a prior CS_STALL, which is emitted by the flush + * following the batch. + */ + *cs++ = GFX_OP_PIPE_CONTROL(6); + *cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_QW_WRITE; + *cs++ = gtt_offset; + *cs++ = 0; + *cs++ = value; + /* We're thrashing one dword of HWS. */ + *cs++ = 0; + + return cs; +} + +static inline u32 * +gen8_emit_ggtt_write(u32 *cs, u32 value, u32 gtt_offset) +{ + /* w/a: bit 5 needs to be zero for MI_FLUSH_DW address. */ + GEM_BUG_ON(gtt_offset & (1 << 5)); + /* Offset should be aligned to 8 bytes for both (QW/DW) write types */ + GEM_BUG_ON(!IS_ALIGNED(gtt_offset, 8)); + + *cs++ = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW; + *cs++ = gtt_offset | MI_FLUSH_DW_USE_GTT; + *cs++ = 0; + *cs++ = value; + + return cs; +} + bool intel_engine_is_idle(struct intel_engine_cs *engine); bool intel_engines_are_idle(struct drm_i915_private *dev_priv); -- GitLab From 3b8a8a30064d0b89d09e979adb4c975d892c21ef Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:16 +0200 Subject: [PATCH 0042/1471] drm/i915: Add information needed to track engine preempt state MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We shouldn't inspect ELSP context status (or any other bits depending on specific submission backend) when using GuC submission. Let's use another piece of HWSP for preempt context, to write its bit of information, meaning that preemption has finished, and hardware is now idle. Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Jeff McGee Cc: Michal Wajdeczko Cc: Oscar Mateo Reviewed-by: Chris Wilson Reviewed-by: Jeff McGee Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-9-michal.winiarski@intel.com --- drivers/gpu/drm/i915/intel_ringbuffer.h | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index c15161e569645..4a5a089853282 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -626,6 +626,8 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value) */ #define I915_GEM_HWS_INDEX 0x30 #define I915_GEM_HWS_INDEX_ADDR (I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT) +#define I915_GEM_HWS_PREEMPT_INDEX 0x32 +#define I915_GEM_HWS_PREEMPT_ADDR (I915_GEM_HWS_PREEMPT_INDEX << MI_STORE_DWORD_INDEX_SHIFT) #define I915_GEM_HWS_SCRATCH_INDEX 0x40 #define I915_GEM_HWS_SCRATCH_ADDR (I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT) @@ -778,6 +780,11 @@ static inline u32 intel_hws_seqno_address(struct intel_engine_cs *engine) return engine->status_page.ggtt_offset + I915_GEM_HWS_INDEX_ADDR; } +static inline u32 intel_hws_preempt_done_address(struct intel_engine_cs *engine) +{ + return engine->status_page.ggtt_offset + I915_GEM_HWS_PREEMPT_ADDR; +} + /* intel_breadcrumbs.c -- user interrupt bottom-half for waiters */ int intel_engine_init_breadcrumbs(struct intel_engine_cs *engine); -- GitLab From fa87271e5bc57ad30c19385e2d0d1d7fbd3c21c1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:17 +0200 Subject: [PATCH 0043/1471] drm/i915/guc: Keep request->priority for its lifetime MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We also want to support preemption with GuC submission backend. In order to do that, we need to remember the priority, like we do on execlists path. v2: Remove completed prio == INT_MAX optimization Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Jeff McGee Cc: Joonas Lahtinen Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-10-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_guc_submission.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 301c658a4cdea..c99ddd8457f3d 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -641,7 +641,6 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) } INIT_LIST_HEAD(&rq->priotree.link); - rq->priotree.priority = INT_MAX; __i915_gem_request_submit(rq); trace_i915_gem_request_in(rq, port_index(port, execlists)); -- GitLab From a4598d17551ab5ce5c9fd2ac7a7803e89b63260e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Wed, 25 Oct 2017 22:00:18 +0200 Subject: [PATCH 0044/1471] drm/i915: Rename helpers used for unwinding, use macro for can_preempt MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We would also like to make use of execlist_cancel_port_requests and unwind_incomplete_requests in GuC preemption backend. Let's rename the functions to use the correct prefixes, so that we can simply add the declarations in the following patch. Similar thing for applies for can_preempt, except we're introducing HAS_LOGICAL_RING_PREEMPTION macro instad, converting other users that were previously touching device info directly. v2: s/intel_engine/execlists and pass execlists to unwind (Chris) v3: use locked version for exporting, drop const qual (Chris) Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Michal Wajdeczko Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171025200020.16636-11-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_drv.c | 2 +- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/intel_engine_cs.c | 6 ++--- drivers/gpu/drm/i915/intel_lrc.c | 35 ++++++++++++++------------ 4 files changed, 25 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 3db5851756f06..7b871802ae360 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -372,7 +372,7 @@ static int i915_getparam(struct drm_device *dev, void *data, value |= I915_SCHEDULER_CAP_ENABLED; value |= I915_SCHEDULER_CAP_PRIORITY; - if (INTEL_INFO(dev_priv)->has_logical_ring_preemption && + if (HAS_LOGICAL_RING_PREEMPTION(dev_priv) && i915_modparams.enable_execlists && !i915_modparams.enable_guc_submission) value |= I915_SCHEDULER_CAP_PREEMPTION; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 366ba74b0ad28..61c155cbf9d71 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3140,6 +3140,8 @@ intel_info(const struct drm_i915_private *dev_priv) #define HAS_LOGICAL_RING_CONTEXTS(dev_priv) \ ((dev_priv)->info.has_logical_ring_contexts) +#define HAS_LOGICAL_RING_PREEMPTION(dev_priv) \ + ((dev_priv)->info.has_logical_ring_preemption) #define USES_PPGTT(dev_priv) (i915_modparams.enable_ppgtt) #define USES_FULL_PPGTT(dev_priv) (i915_modparams.enable_ppgtt >= 2) #define USES_FULL_48BIT_PPGTT(dev_priv) (i915_modparams.enable_ppgtt == 3) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index fedb839dff611..3ac876ca6cae2 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -620,7 +620,7 @@ int intel_engine_init_common(struct intel_engine_cs *engine) * Similarly the preempt context must always be available so that * we can interrupt the engine at any time. */ - if (INTEL_INFO(engine->i915)->has_logical_ring_preemption) { + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) { ring = engine->context_pin(engine, engine->i915->preempt_context); if (IS_ERR(ring)) { @@ -651,7 +651,7 @@ int intel_engine_init_common(struct intel_engine_cs *engine) err_breadcrumbs: intel_engine_fini_breadcrumbs(engine); err_unpin_preempt: - if (INTEL_INFO(engine->i915)->has_logical_ring_preemption) + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) engine->context_unpin(engine, engine->i915->preempt_context); err_unpin_kernel: engine->context_unpin(engine, engine->i915->kernel_context); @@ -679,7 +679,7 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) intel_engine_cleanup_cmd_parser(engine); i915_gem_batch_pool_fini(&engine->batch_pool); - if (INTEL_INFO(engine->i915)->has_logical_ring_preemption) + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) engine->context_unpin(engine, engine->i915->preempt_context); engine->context_unpin(engine, engine->i915->kernel_context); } diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 599c709fc5a76..b5d382ef8d85b 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -354,7 +354,7 @@ static void unwind_wa_tail(struct drm_i915_gem_request *rq) assert_ring_tail_valid(rq->ring, rq->tail); } -static void unwind_incomplete_requests(struct intel_engine_cs *engine) +static void __unwind_incomplete_requests(struct intel_engine_cs *engine) { struct drm_i915_gem_request *rq, *rn; struct i915_priolist *uninitialized_var(p); @@ -385,6 +385,17 @@ static void unwind_incomplete_requests(struct intel_engine_cs *engine) } } +static void +execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists) +{ + struct intel_engine_cs *engine = + container_of(execlists, typeof(*engine), execlists); + + spin_lock_irq(&engine->timeline->lock); + __unwind_incomplete_requests(engine); + spin_unlock_irq(&engine->timeline->lock); +} + static inline void execlists_context_status_change(struct drm_i915_gem_request *rq, unsigned long status) @@ -515,11 +526,6 @@ static void inject_preempt_context(struct intel_engine_cs *engine) elsp_write(ce->lrc_desc, elsp); } -static bool can_preempt(struct intel_engine_cs *engine) -{ - return INTEL_INFO(engine->i915)->has_logical_ring_preemption; -} - static void execlists_dequeue(struct intel_engine_cs *engine) { struct intel_engine_execlists * const execlists = &engine->execlists; @@ -567,7 +573,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) if (port_count(&port[0]) > 1) goto unlock; - if (can_preempt(engine) && + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915) && rb_entry(rb, struct i915_priolist, node)->priority > max(last->priotree.priority, 0)) { /* @@ -691,7 +697,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) } static void -execlist_cancel_port_requests(struct intel_engine_execlists *execlists) +execlists_cancel_port_requests(struct intel_engine_execlists * const execlists) { struct execlist_port *port = execlists->port; unsigned int num_ports = execlists_num_ports(execlists); @@ -718,7 +724,7 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) spin_lock_irqsave(&engine->timeline->lock, flags); /* Cancel the requests on the HW and clear the ELSP tracker. */ - execlist_cancel_port_requests(execlists); + execlists_cancel_port_requests(execlists); /* Mark all executing requests as skipped. */ list_for_each_entry(rq, &engine->timeline->requests, link) { @@ -858,11 +864,8 @@ static void intel_lrc_irq_handler(unsigned long data) if (status & GEN8_CTX_STATUS_ACTIVE_IDLE && buf[2*head + 1] == PREEMPT_ID) { - execlist_cancel_port_requests(execlists); - - spin_lock_irq(&engine->timeline->lock); - unwind_incomplete_requests(engine); - spin_unlock_irq(&engine->timeline->lock); + execlists_cancel_port_requests(execlists); + execlists_unwind_incomplete_requests(execlists); GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT)); @@ -1531,10 +1534,10 @@ static void reset_common_ring(struct intel_engine_cs *engine, * guessing the missed context-switch events by looking at what * requests were completed. */ - execlist_cancel_port_requests(execlists); + execlists_cancel_port_requests(execlists); /* Push back any incomplete requests for replay after the reset. */ - unwind_incomplete_requests(engine); + __unwind_incomplete_requests(engine); spin_unlock_irqrestore(&engine->timeline->lock, flags); -- GitLab From c41937fd994a449b4e862826d7eb9f366f6d33e9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Winiarski?= Date: Thu, 26 Oct 2017 15:35:58 +0200 Subject: [PATCH 0045/1471] drm/i915/guc: Preemption! With GuC MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pretty similar to what we have on execlists. We're reusing most of the GEM code, however, due to GuC quirks we need a couple of extra bits. Preemption is implemented as GuC action, and actions can be pretty slow. Because of that, we're using a mutex to serialize them. Since we're requesting preemption from the tasklet, the task of creating a workitem and wrapping it in GuC action is delegated to a worker. To distinguish that preemption has finished, we're using additional piece of HWSP, and since we're not getting context switch interrupts, we're also adding a user interrupt. The fact that our special preempt context has completed unfortunately doesn't mean that we're ready to submit new work. We also need to wait for GuC to finish its own processing. v2: Don't compile out the wait for GuC, handle workqueue flush on reset, no need for ordered workqueue, put on a reviewer hat when looking at my own patches (Chris) Move struct work around in intel_guc, move user interruput outside of conditional (Michał) Keep ring around rather than chase though intel_context v3: Extract WA for flushing ggtt writes to a helper (Chris) Keep work_struct in intel_guc rather than engine (Michał) Use ordered workqueue for inject_preempt worker to avoid GuC quirks. v4: Drop now unused INTEL_GUC_PREEMPT_OPTION_IMMEDIATE (Daniele) Drop stray newlines, use container_of for intel_guc in worker, check for presence of workqueue when flushing it, rather than enable_guc_submission modparam, reorder preempt postprocessing (Chris) v5: Make wq NULL after destroying it v6: Swap struct guc_preempt_work members (Michał) Signed-off-by: Michał Winiarski Cc: Chris Wilson Cc: Jeff McGee Cc: Joonas Lahtinen Cc: Oscar Mateo Cc: Tvrtko Ursulin Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171026133558.19580-1-michal.winiarski@intel.com --- drivers/gpu/drm/i915/i915_drv.c | 3 +- drivers/gpu/drm/i915/i915_gem.c | 10 + drivers/gpu/drm/i915/i915_guc_submission.c | 208 +++++++++++++++++++-- drivers/gpu/drm/i915/intel_guc.h | 8 + drivers/gpu/drm/i915/intel_lrc.c | 4 +- drivers/gpu/drm/i915/intel_lrc.h | 1 - drivers/gpu/drm/i915/intel_ringbuffer.h | 6 + 7 files changed, 222 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 7b871802ae360..af745749509c0 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -373,8 +373,7 @@ static int i915_getparam(struct drm_device *dev, void *data, value |= I915_SCHEDULER_CAP_PRIORITY; if (HAS_LOGICAL_RING_PREEMPTION(dev_priv) && - i915_modparams.enable_execlists && - !i915_modparams.enable_guc_submission) + i915_modparams.enable_execlists) value |= I915_SCHEDULER_CAP_PREEMPTION; } break; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d803ef5f4a7f2..c2506fb3a4836 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2921,6 +2921,16 @@ i915_gem_reset_prepare_engine(struct intel_engine_cs *engine) tasklet_kill(&engine->execlists.irq_tasklet); tasklet_disable(&engine->execlists.irq_tasklet); + /* + * We're using worker to queue preemption requests from the tasklet in + * GuC submission mode. + * Even though tasklet was disabled, we may still have a worker queued. + * Let's make sure that all workers scheduled before disabling the + * tasklet are completed before continuing with the reset. + */ + if (engine->i915->guc.preempt_wq) + flush_workqueue(engine->i915->guc.preempt_wq); + if (engine->irq_seqno_barrier) engine->irq_seqno_barrier(engine); diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index c99ddd8457f3d..3049a0781b885 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -565,6 +565,108 @@ static void guc_add_request(struct intel_guc *guc, spin_unlock(&client->wq_lock); } +/* + * When we're doing submissions using regular execlists backend, writing to + * ELSP from CPU side is enough to make sure that writes to ringbuffer pages + * pinned in mappable aperture portion of GGTT are visible to command streamer. + * Writes done by GuC on our behalf are not guaranteeing such ordering, + * therefore, to ensure the flush, we're issuing a POSTING READ. + */ +static void flush_ggtt_writes(struct i915_vma *vma) +{ + struct drm_i915_private *dev_priv = to_i915(vma->obj->base.dev); + + if (i915_vma_is_map_and_fenceable(vma)) + POSTING_READ_FW(GUC_STATUS); +} + +#define GUC_PREEMPT_FINISHED 0x1 +#define GUC_PREEMPT_BREADCRUMB_DWORDS 0x8 +static void inject_preempt_context(struct work_struct *work) +{ + struct guc_preempt_work *preempt_work = + container_of(work, typeof(*preempt_work), work); + struct intel_engine_cs *engine = preempt_work->engine; + struct intel_guc *guc = container_of(preempt_work, typeof(*guc), + preempt_work[engine->id]); + struct i915_guc_client *client = guc->preempt_client; + struct intel_ring *ring = client->owner->engine[engine->id].ring; + u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(client->owner, + engine)); + u32 *cs = ring->vaddr + ring->tail; + u32 data[7]; + + if (engine->id == RCS) { + cs = gen8_emit_ggtt_write_rcs(cs, GUC_PREEMPT_FINISHED, + intel_hws_preempt_done_address(engine)); + } else { + cs = gen8_emit_ggtt_write(cs, GUC_PREEMPT_FINISHED, + intel_hws_preempt_done_address(engine)); + *cs++ = MI_NOOP; + *cs++ = MI_NOOP; + } + *cs++ = MI_USER_INTERRUPT; + *cs++ = MI_NOOP; + + GEM_BUG_ON(!IS_ALIGNED(ring->size, + GUC_PREEMPT_BREADCRUMB_DWORDS * sizeof(u32))); + GEM_BUG_ON((void *)cs - (ring->vaddr + ring->tail) != + GUC_PREEMPT_BREADCRUMB_DWORDS * sizeof(u32)); + + ring->tail += GUC_PREEMPT_BREADCRUMB_DWORDS * sizeof(u32); + ring->tail &= (ring->size - 1); + + flush_ggtt_writes(ring->vma); + + spin_lock_irq(&client->wq_lock); + guc_wq_item_append(client, engine->guc_id, ctx_desc, + ring->tail / sizeof(u64), 0); + spin_unlock_irq(&client->wq_lock); + + data[0] = INTEL_GUC_ACTION_REQUEST_PREEMPTION; + data[1] = client->stage_id; + data[2] = INTEL_GUC_PREEMPT_OPTION_DROP_WORK_Q | + INTEL_GUC_PREEMPT_OPTION_DROP_SUBMIT_Q; + data[3] = engine->guc_id; + data[4] = guc->execbuf_client->priority; + data[5] = guc->execbuf_client->stage_id; + data[6] = guc_ggtt_offset(guc->shared_data); + + if (WARN_ON(intel_guc_send(guc, data, ARRAY_SIZE(data)))) { + execlists_clear_active(&engine->execlists, + EXECLISTS_ACTIVE_PREEMPT); + tasklet_schedule(&engine->execlists.irq_tasklet); + } +} + +/* + * We're using user interrupt and HWSP value to mark that preemption has + * finished and GPU is idle. Normally, we could unwind and continue similar to + * execlists submission path. Unfortunately, with GuC we also need to wait for + * it to finish its own postprocessing, before attempting to submit. Otherwise + * GuC may silently ignore our submissions, and thus we risk losing request at + * best, executing out-of-order and causing kernel panic at worst. + */ +#define GUC_PREEMPT_POSTPROCESS_DELAY_MS 10 +static void wait_for_guc_preempt_report(struct intel_engine_cs *engine) +{ + struct intel_guc *guc = &engine->i915->guc; + struct guc_shared_ctx_data *data = guc->shared_data_vaddr; + struct guc_ctx_report *report = + &data->preempt_ctx_report[engine->guc_id]; + + WARN_ON(wait_for_atomic(report->report_return_status == + INTEL_GUC_REPORT_STATUS_COMPLETE, + GUC_PREEMPT_POSTPROCESS_DELAY_MS)); + /* + * GuC is expecting that we're also going to clear the affected context + * counter, let's also reset the return status to not depend on GuC + * resetting it after recieving another preempt action + */ + report->affected_count = 0; + report->report_return_status = INTEL_GUC_REPORT_STATUS_UNKNOWN; +} + /** * i915_guc_submit() - Submit commands through GuC * @engine: engine associated with the commands @@ -574,8 +676,7 @@ static void guc_add_request(struct intel_guc *guc, */ static void i915_guc_submit(struct intel_engine_cs *engine) { - struct drm_i915_private *dev_priv = engine->i915; - struct intel_guc *guc = &dev_priv->guc; + struct intel_guc *guc = &engine->i915->guc; struct intel_engine_execlists * const execlists = &engine->execlists; struct execlist_port *port = execlists->port; unsigned int n; @@ -588,8 +689,7 @@ static void i915_guc_submit(struct intel_engine_cs *engine) if (rq && count == 0) { port_set(&port[n], port_pack(rq, ++count)); - if (i915_vma_is_map_and_fenceable(rq->ring->vma)) - POSTING_READ_FW(GUC_STATUS); + flush_ggtt_writes(rq->ring->vma); guc_add_request(guc, rq); } @@ -617,13 +717,32 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) bool submit = false; struct rb_node *rb; - if (port_isset(port)) - port++; - spin_lock_irq(&engine->timeline->lock); rb = execlists->first; GEM_BUG_ON(rb_first(&execlists->queue) != rb); - while (rb) { + + if (!rb) + goto unlock; + + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915) && port_isset(port)) { + struct guc_preempt_work *preempt_work = + &engine->i915->guc.preempt_work[engine->id]; + + if (rb_entry(rb, struct i915_priolist, node)->priority > + max(port_request(port)->priotree.priority, 0)) { + execlists_set_active(execlists, + EXECLISTS_ACTIVE_PREEMPT); + queue_work(engine->i915->guc.preempt_wq, + &preempt_work->work); + goto unlock; + } else if (port_isset(last_port)) { + goto unlock; + } + + port++; + } + + do { struct i915_priolist *p = rb_entry(rb, typeof(*p), node); struct drm_i915_gem_request *rq, *rn; @@ -653,7 +772,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) INIT_LIST_HEAD(&p->requests); if (p->priority != I915_PRIORITY_NORMAL) kmem_cache_free(engine->i915->priorities, p); - } + } while (rb); done: execlists->first = rb; if (submit) { @@ -661,6 +780,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) execlists_set_active(execlists, EXECLISTS_ACTIVE_USER); i915_guc_submit(engine); } +unlock: spin_unlock_irq(&engine->timeline->lock); } @@ -669,8 +789,6 @@ static void i915_guc_irq_handler(unsigned long data) struct intel_engine_cs * const engine = (struct intel_engine_cs *)data; struct intel_engine_execlists * const execlists = &engine->execlists; struct execlist_port *port = execlists->port; - const struct execlist_port * const last_port = - &execlists->port[execlists->port_mask]; struct drm_i915_gem_request *rq; rq = port_request(&port[0]); @@ -685,7 +803,19 @@ static void i915_guc_irq_handler(unsigned long data) if (!rq) execlists_clear_active(execlists, EXECLISTS_ACTIVE_USER); - if (!port_isset(last_port)) + if (execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT) && + intel_read_status_page(engine, I915_GEM_HWS_PREEMPT_INDEX) == + GUC_PREEMPT_FINISHED) { + execlists_cancel_port_requests(&engine->execlists); + execlists_unwind_incomplete_requests(execlists); + + wait_for_guc_preempt_report(engine); + + execlists_clear_active(execlists, EXECLISTS_ACTIVE_PREEMPT); + intel_write_status_page(engine, I915_GEM_HWS_PREEMPT_INDEX, 0); + } + + if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT)) i915_guc_dequeue(engine); } @@ -1059,6 +1189,51 @@ static void guc_ads_destroy(struct intel_guc *guc) i915_vma_unpin_and_release(&guc->ads_vma); } +static int guc_preempt_work_create(struct intel_guc *guc) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* + * Even though both sending GuC action, and adding a new workitem to + * GuC workqueue are serialized (each with its own locking), since + * we're using mutliple engines, it's possible that we're going to + * issue a preempt request with two (or more - each for different + * engine) workitems in GuC queue. In this situation, GuC may submit + * all of them, which will make us very confused. + * Our preemption contexts may even already be complete - before we + * even had the chance to sent the preempt action to GuC!. Rather + * than introducing yet another lock, we can just use ordered workqueue + * to make sure we're always sending a single preemption request with a + * single workitem. + */ + guc->preempt_wq = alloc_ordered_workqueue("i915-guc_preempt", + WQ_HIGHPRI); + if (!guc->preempt_wq) + return -ENOMEM; + + for_each_engine(engine, dev_priv, id) { + guc->preempt_work[id].engine = engine; + INIT_WORK(&guc->preempt_work[id].work, inject_preempt_context); + } + + return 0; +} + +static void guc_preempt_work_destroy(struct intel_guc *guc) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + for_each_engine(engine, dev_priv, id) + cancel_work_sync(&guc->preempt_work[id].work); + + destroy_workqueue(guc->preempt_wq); + guc->preempt_wq = NULL; +} + /* * Set up the memory resources to be shared with the GuC (via the GGTT) * at firmware loading time. @@ -1083,12 +1258,18 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) if (ret < 0) goto err_shared_data; + ret = guc_preempt_work_create(guc); + if (ret) + goto err_log; + ret = guc_ads_create(guc); if (ret < 0) - goto err_log; + goto err_wq; return 0; +err_wq: + guc_preempt_work_destroy(guc); err_log: intel_guc_log_destroy(guc); err_shared_data: @@ -1103,6 +1284,7 @@ void i915_guc_submission_fini(struct drm_i915_private *dev_priv) struct intel_guc *guc = &dev_priv->guc; guc_ads_destroy(guc); + guc_preempt_work_destroy(guc); intel_guc_log_destroy(guc); guc_shared_data_destroy(guc); guc_stage_desc_pool_destroy(guc); diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index 6ca55c5bed3c2..607e025002624 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -34,6 +34,11 @@ #include "i915_guc_reg.h" #include "i915_vma.h" +struct guc_preempt_work { + struct work_struct work; + struct intel_engine_cs *engine; +}; + /* * Top level structure of GuC. It handles firmware loading and manages client * pool and doorbells. intel_guc owns a i915_guc_client to replace the legacy @@ -60,6 +65,9 @@ struct intel_guc { struct i915_guc_client *execbuf_client; struct i915_guc_client *preempt_client; + struct guc_preempt_work preempt_work[I915_NUM_ENGINES]; + struct workqueue_struct *preempt_wq; + DECLARE_BITMAP(doorbell_bitmap, GUC_NUM_DOORBELLS); /* Cyclic counter mod pagesize */ u32 db_cacheline; diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index b5d382ef8d85b..6840ec8db0376 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -385,7 +385,7 @@ static void __unwind_incomplete_requests(struct intel_engine_cs *engine) } } -static void +void execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists) { struct intel_engine_cs *engine = @@ -696,7 +696,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) } } -static void +void execlists_cancel_port_requests(struct intel_engine_execlists * const execlists) { struct execlist_port *port = execlists->port; diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h index 689fde1a63a90..17182ce296740 100644 --- a/drivers/gpu/drm/i915/intel_lrc.h +++ b/drivers/gpu/drm/i915/intel_lrc.h @@ -107,7 +107,6 @@ intel_lr_context_descriptor(struct i915_gem_context *ctx, return ctx->engine[engine->id].lrc_desc; } - /* Execlists */ int intel_sanitize_enable_execlists(struct drm_i915_private *dev_priv, int enable_execlists); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 4a5a089853282..69ad875fd0111 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -557,6 +557,12 @@ execlists_is_active(const struct intel_engine_execlists *execlists, return test_bit(bit, (unsigned long *)&execlists->active); } +void +execlists_cancel_port_requests(struct intel_engine_execlists * const execlists); + +void +execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists); + static inline unsigned int execlists_num_ports(const struct intel_engine_execlists * const execlists) { -- GitLab From bad7b7e674cf94775cf86a7c4331bab61d79d7ea Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Tue, 24 Oct 2017 10:50:56 +0000 Subject: [PATCH 0046/1471] drm/i915/huc: Use helper function while waiting for DMA completion Waiting for DMA status register can be done with dedicated function. Lets use it as additional bonus will be smaller driver footprint. Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171024105056.43276-1-michal.wajdeczko@intel.com --- drivers/gpu/drm/i915/intel_huc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_huc.c b/drivers/gpu/drm/i915/intel_huc.c index c8a48cbc2b7da..98d17254593c8 100644 --- a/drivers/gpu/drm/i915/intel_huc.c +++ b/drivers/gpu/drm/i915/intel_huc.c @@ -151,7 +151,7 @@ static int huc_ucode_xfer(struct intel_uc_fw *huc_fw, struct i915_vma *vma) I915_WRITE(DMA_CTRL, _MASKED_BIT_ENABLE(HUC_UKERNEL | START_DMA)); /* Wait for DMA to finish */ - ret = wait_for((I915_READ(DMA_CTRL) & START_DMA) == 0, 100); + ret = intel_wait_for_register_fw(dev_priv, DMA_CTRL, START_DMA, 0, 100); DRM_DEBUG_DRIVER("HuC DMA transfer wait over with ret %d\n", ret); -- GitLab From 3c75de5b983a0a147ec93d1960f241a74d76b347 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 26 Oct 2017 12:50:48 +0100 Subject: [PATCH 0047/1471] drm/i915: Include RING_MODE when dumping the engine state Knowing the RING_MODE flags is useful for checking the state of the engine, such as whether the CS is idle after trying to stop the engines before reset. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171026115048.20144-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/intel_engine_cs.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 3ac876ca6cae2..f31f2d6384c35 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1723,9 +1723,14 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) drm_printf(m, "\tRING_TAIL: 0x%08x [0x%08x]\n", I915_READ(RING_TAIL(engine->mmio_base)) & TAIL_ADDR, rq ? rq->ring->tail : 0); - drm_printf(m, "\tRING_CTL: 0x%08x [%s]\n", + drm_printf(m, "\tRING_CTL: 0x%08x%s\n", I915_READ(RING_CTL(engine->mmio_base)), - I915_READ(RING_CTL(engine->mmio_base)) & (RING_WAIT | RING_WAIT_SEMAPHORE) ? "waiting" : ""); + I915_READ(RING_CTL(engine->mmio_base)) & (RING_WAIT | RING_WAIT_SEMAPHORE) ? " [waiting]" : ""); + if (INTEL_GEN(engine->i915) > 2) { + drm_printf(m, "\tRING_MODE: 0x%08x%s\n", + I915_READ(RING_MI_MODE(engine->mmio_base)), + I915_READ(RING_MI_MODE(engine->mmio_base)) & (MODE_IDLE) ? " [idle]" : ""); + } rcu_read_unlock(); -- GitLab From 39bf4de89ff7c87a2205e7ec5483e7424b86a1f6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 24 Oct 2017 19:15:47 +0100 Subject: [PATCH 0048/1471] drm/i915: Add -Wall -Wextra to our build, set warnings to full MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Recently W=1 on gcc-7.2 (-Wunused-const-variable) caught a regression that had been lurking for 6 months, so lets try enabling the full set of warnings for CI builds. This means more patches will be rejected early that contain trivial and sometimes not so trivial bugs. However, our code does not yet compile cleanly with W=1, so we have to apply a filter to the set of warnings until we can eliminate the mistakes. It also means that developers will have to be running the full gamut of gcc to ensure that as warnings come and go with gcc updates, we have the CI build prepared. v2: Use fine-grained -Wno overrides. Inside the makefile, we can specify CFLAGS on a per-object level, which allows us to limit the scope of any particular warning override. v3: Place per-file overrides after the main enabling block. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Jani Nikula Cc: Daniel Vetter Cc: Tomi Sarvela Cc: Michal Wajdeczko Cc: Ville Syrjälä Acked-by: Tomi Sarvela Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171024181547.27889-1-chris@chris-wilson.co.uk Acked-by: Jani Nikula --- drivers/gpu/drm/i915/Makefile | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 6c3b0481ef829..7750be8e27a6b 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -2,7 +2,26 @@ # Makefile for the drm device driver. This driver provides support for the # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher. -subdir-ccflags-$(CONFIG_DRM_I915_WERROR) := -Werror +# Add a set of useful warning flags and enable -Werror for CI to prevent +# trivial mistakes from creeping in. We have to do this piecemeal as we reject +# any patch that isn't warning clean, so turning on -Wall -Wextra (or W=1) we +# need to filter out dubious warnings. Still it is our interest +# to keep running locally with W=1 C=1 until we are completely clean. +# +# Note the danger in using -Wall -Wextra is that when CI updates gcc we +# will most likely get a sudden build breakage... Hopefully we will fix +# new warnings before CI updates! +subdir-ccflags-y := -Wall -Wextra +subdir-ccflags-y += $(call cc-option,-Wno-unused-parameter,) +subdir-ccflags-y += $(call cc-option,-Wno-type-limits,) +subdir-ccflags-y += $(call cc-option,-Wno-missing-field-initializers,) +subdir-ccflags-y += $(call cc-option,-Wno-implicit-fallthrough,) +subdir-ccflags-$(CONFIG_DRM_I915_WERROR) += -Werror + +# Fine grained warnings disable +CFLAGS_i915_pci.o = $(call cc-option,-Wno-override-init,) +CFLAGS_intel_fbdev.o = $(call cc-option,-Wno-override-init,) + subdir-ccflags-y += \ $(call as-instr,movntdqa (%eax)$(comma)%xmm0,-DCONFIG_AS_MOVNTDQA) -- GitLab From 886c6b8692ba5f71b578097524b3b082e2e02119 Mon Sep 17 00:00:00 2001 From: Manasi Navare Date: Thu, 26 Oct 2017 14:52:00 -0700 Subject: [PATCH 0049/1471] drm/i915: Cancel the modeset retry work during modeset cleanup During modeset cleanup on driver unload we may have a pending hotplug work. This needs to be canceled early during the teardown so that it does not fire after we have freed the connector. We do this after drm_kms_helper_poll_fini(dev) since this might trigger modeset retry work due to link retrain and before intel_fbdev_fini() since this work requires the lock from fbdev. If this is not done we may see something like: DEBUG_LOCKS_WARN_ON(mutex_is_locked(lock)) ------------[ cut here ]------------ WARNING: CPU: 4 PID: 5010 at kernel/locking/mutex-debug.c:103 mutex_destroy+0x4e/0x60 Modules linked in: i915(-) snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec snd_hwdep snd_hda_core snd_pcm vgem ax88179_178 +a usbnet mii x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel e1000e ptp pps_core prime_numbers i2c_hid +[last unloaded: snd_hda_intel] CPU: 4 PID: 5010 Comm: drv_module_relo Tainted: G U 4.14.0-rc3-CI-CI_DRM_3186+ #1 Hardware name: Intel Corporation CoffeeLake Client Platform/CoffeeLake S UDIMM RVP, BIOS CNLSFWX1.R00.X104.A03.1709140524 09/14/2017 task: ffff8803c827aa40 task.stack: ffffc90000520000 RIP: 0010:mutex_destroy+0x4e/0x60 RSP: 0018:ffffc90000523d58 EFLAGS: 00010292 RAX: 000000000000002a RBX: ffff88044fbef648 RCX: 0000000000000000 RDX: 0000000080000001 RSI: 0000000000000001 RDI: ffffffff810f0cf0 RBP: ffffc90000523d60 R08: 0000000000000001 R09: 0000000000000001 R10: 000000000f21cb81 R11: 0000000000000000 R12: ffff88044f71efc8 R13: ffffffffa02b3d20 R14: ffffffffa02b3d90 R15: ffff880459b29308 FS: 00007f5df4d6e8c0(0000) GS:ffff88045d300000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000055ec51f00a18 CR3: 0000000451782006 CR4: 00000000003606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: drm_fb_helper_fini+0xd9/0x130 intel_fbdev_destroy+0x12/0x60 [i915] intel_fbdev_fini+0x28/0x30 [i915] intel_modeset_cleanup+0x45/0xa0 [i915] i915_driver_unload+0x92/0x180 [i915] i915_pci_remove+0x19/0x30 [i915] i915_driver_unload+0x92/0x180 [i915] i915_pci_remove+0x19/0x30 [i915] pci_device_remove+0x39/0xb0 device_release_driver_internal+0x15d/0x220 driver_detach+0x40/0x80 bus_remove_driver+0x58/0xd0 driver_unregister+0x2c/0x40 pci_unregister_driver+0x36/0xb0 i915_exit+0x1a/0x8b [i915] SyS_delete_module+0x18c/0x1e0 entry_SYSCALL_64_fastpath+0x1c/0xb1 RIP: 0033:0x7f5df3286287 RSP: 002b:00007fff8e107cc8 EFLAGS: 00000246 ORIG_RAX: 00000000000000b0 RAX: ffffffffffffffda RBX: ffffffff81493a03 RCX: 00007f5df3286287 RDX: 0000000000000001 RSI: 0000000000000800 RDI: 0000564c7be02e48 RBP: ffffc90000523f88 R08: 0000000000000000 R09: 0000000000000080 R10: 00007f5df4d6e8c0 R11: 0000000000000246 R12: 0000000000000000 R13: 00007fff8e107eb0 R14: 0000000000000000 R15: 0000000000000000 Or a GPF like: general protection fault: 0000 [#1] PREEMPT SMP Modules linked in: i915(-) snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec snd_hwdep snd_hda_core snd_pcm vgem ax88179_178 +a usbnet mii x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel e1000e ptp pps_core prime_numbers i2c_hid +[last unloaded: snd_hda_intel] CPU: 0 PID: 82 Comm: kworker/0:1 Tainted: G U W 4.14.0-rc3-CI-CI_DRM_3186+ #1 Hardware name: Intel Corporation CoffeeLake Client Platform/CoffeeLake S UDIMM RVP, BIOS CNLSFWX1.R00.X104.A03.1709140524 09/14/2017 Workqueue: events intel_dp_modeset_retry_work_fn [i915] task: ffff88045a5caa40 task.stack: ffffc90000378000 RIP: 0010:drm_setup_crtcs+0x143/0xbf0 RSP: 0018:ffffc9000037bd20 EFLAGS: 00010202 RAX: 6b6b6b6b6b6b6b6b RBX: 0000000000000002 RCX: 0000000000000001 RDX: 0000000000000001 RSI: 0000000000000780 RDI: 00000000ffffffff RBP: ffffc9000037bdb8 R08: 0000000000000001 R09: 0000000000000001 R10: 0000000000000780 R11: 0000000000000000 R12: 0000000000000002 R13: ffff88044fbef4e8 R14: 0000000000000780 R15: 0000000000000438 FS: 0000000000000000(0000) GS:ffff88045d200000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000055ec51ee5168 CR3: 000000044c89d003 CR4: 00000000003606f0 Call Trace: drm_fb_helper_hotplug_event.part.18+0x7e/0xc0 drm_fb_helper_hotplug_event+0x1a/0x20 intel_fbdev_output_poll_changed+0x1a/0x20 [i915] drm_kms_helper_hotplug_event+0x27/0x30 intel_dp_modeset_retry_work_fn+0x77/0x80 [i915] process_one_work+0x233/0x660 worker_thread+0x206/0x3b0 kthread+0x152/0x190 ? process_one_work+0x660/0x660 ? kthread_create_on_node+0x40/0x40 ret_from_fork+0x27/0x40 Code: 06 00 00 45 8b 45 20 31 db 45 31 e4 45 85 c0 0f 8e 91 06 00 00 44 8b 75 94 44 8b 7d 90 49 8b 45 28 49 63 d4 44 89 f6 41 83 c4 01 <48> 8b 04 d0 44 +89 fa 48 8b 38 48 8b 87 a8 01 00 00 ff 50 20 01 RIP: drm_setup_crtcs+0x143/0xbf0 RSP: ffffc9000037bd20 ---[ end trace 08901ff1a77d30c7 ]--- v2: * Rename it to intel_hpd_poll_fini() and call drm_kms_helper_fini() inside it as the first step before cancel work (Chris Wilson) * Add GPF trace in commit message and make the function static (Maarten Lankhorst) Suggested-by: Maarten Lankhorst Suggested-by: Chris Wilson Fixes: 9301397a63b3 ("drm/i915: Implement Link Rate fallback on Link training failure") Cc: Chris Wilson Cc: Tony Cheng Cc: Harry Wentland Cc: Jani Nikula Cc: Daniel Vetter Cc: Ville Syrjala Cc: Manasi Navare Cc: Maarten Lankhorst Signed-off-by: Manasi Navare Link: https://patchwork.freedesktop.org/patch/msgid/1509054720-25325-1-git-send-email-manasi.d.navare@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_display.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0e493a1fedab5..a369b35d044d6 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15200,6 +15200,23 @@ void intel_connector_unregister(struct drm_connector *connector) intel_panel_destroy_backlight(connector); } +static void intel_hpd_poll_fini(struct drm_device *dev) +{ + struct intel_connector *connector; + struct drm_connector_list_iter conn_iter; + + /* First disable polling... */ + drm_kms_helper_poll_fini(dev); + + /* Then kill the work that may have been queued by hpd. */ + drm_connector_list_iter_begin(dev, &conn_iter); + for_each_intel_connector_iter(connector, &conn_iter) { + if (connector->modeset_retry_work.func) + cancel_work_sync(&connector->modeset_retry_work); + } + drm_connector_list_iter_end(&conn_iter); +} + void intel_modeset_cleanup(struct drm_device *dev) { struct drm_i915_private *dev_priv = to_i915(dev); @@ -15220,7 +15237,7 @@ void intel_modeset_cleanup(struct drm_device *dev) * Due to the hpd irq storm handling the hotplug work can re-arm the * poll handlers. Hence disable polling after hpd handling is shut down. */ - drm_kms_helper_poll_fini(dev); + intel_hpd_poll_fini(dev); /* poll work can call into fbdev, hence clean that up afterwards */ intel_fbdev_fini(dev_priv); -- GitLab From 28283f4f359cd7cfa9e65457bb98c507a2cd0cd0 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Thu, 19 Oct 2017 17:13:40 +0200 Subject: [PATCH 0050/1471] drm/i915: Do not rely on wm preservation for ILK watermarks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The original intent was to preserve watermarks as much as possible in intel_pipe_wm.raw_wm, and put the validated ones in intel_pipe_wm.wm. It seems this approach is insufficient and we don't always preserve the raw watermarks, so just use the atomic iterator we're already using to get a const pointer to all bound planes on the crtc. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102373 Signed-off-by: Maarten Lankhorst Cc: stable@vger.kernel.org #v4.8+ Acked-by: Ville Syrjälä Reviewed-by: Matt Roper Link: https://patchwork.freedesktop.org/patch/msgid/20171019151341.4579-1-maarten.lankhorst@linux.intel.com --- drivers/gpu/drm/i915/intel_drv.h | 1 - drivers/gpu/drm/i915/intel_pm.c | 51 +++++++++++++------------------- 2 files changed, 21 insertions(+), 31 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3b4eafd39f558..df966427232cc 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -501,7 +501,6 @@ struct intel_crtc_scaler_state { struct intel_pipe_wm { struct intel_wm_level wm[5]; - struct intel_wm_level raw_wm[5]; uint32_t linetime; bool fbc_wm_enabled; bool pipe_enabled; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 742d5455b2018..e7926973112e6 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2721,9 +2721,9 @@ static void ilk_compute_wm_level(const struct drm_i915_private *dev_priv, const struct intel_crtc *intel_crtc, int level, struct intel_crtc_state *cstate, - struct intel_plane_state *pristate, - struct intel_plane_state *sprstate, - struct intel_plane_state *curstate, + const struct intel_plane_state *pristate, + const struct intel_plane_state *sprstate, + const struct intel_plane_state *curstate, struct intel_wm_level *result) { uint16_t pri_latency = dev_priv->wm.pri_latency[level]; @@ -3043,28 +3043,24 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *cstate) struct intel_pipe_wm *pipe_wm; struct drm_device *dev = state->dev; const struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_plane *intel_plane; - struct intel_plane_state *pristate = NULL; - struct intel_plane_state *sprstate = NULL; - struct intel_plane_state *curstate = NULL; + struct drm_plane *plane; + const struct drm_plane_state *plane_state; + const struct intel_plane_state *pristate = NULL; + const struct intel_plane_state *sprstate = NULL; + const struct intel_plane_state *curstate = NULL; int level, max_level = ilk_wm_max_level(dev_priv), usable_level; struct ilk_wm_maximums max; pipe_wm = &cstate->wm.ilk.optimal; - for_each_intel_plane_on_crtc(dev, intel_crtc, intel_plane) { - struct intel_plane_state *ps; - - ps = intel_atomic_get_existing_plane_state(state, - intel_plane); - if (!ps) - continue; + drm_atomic_crtc_state_for_each_plane_state(plane, plane_state, &cstate->base) { + const struct intel_plane_state *ps = to_intel_plane_state(plane_state); - if (intel_plane->base.type == DRM_PLANE_TYPE_PRIMARY) + if (plane->type == DRM_PLANE_TYPE_PRIMARY) pristate = ps; - else if (intel_plane->base.type == DRM_PLANE_TYPE_OVERLAY) + else if (plane->type == DRM_PLANE_TYPE_OVERLAY) sprstate = ps; - else if (intel_plane->base.type == DRM_PLANE_TYPE_CURSOR) + else if (plane->type == DRM_PLANE_TYPE_CURSOR) curstate = ps; } @@ -3086,11 +3082,9 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *cstate) if (pipe_wm->sprites_scaled) usable_level = 0; - ilk_compute_wm_level(dev_priv, intel_crtc, 0, cstate, - pristate, sprstate, curstate, &pipe_wm->raw_wm[0]); - memset(&pipe_wm->wm, 0, sizeof(pipe_wm->wm)); - pipe_wm->wm[0] = pipe_wm->raw_wm[0]; + ilk_compute_wm_level(dev_priv, intel_crtc, 0, cstate, + pristate, sprstate, curstate, &pipe_wm->wm[0]); if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) pipe_wm->linetime = hsw_compute_linetime_wm(cstate); @@ -3100,8 +3094,8 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *cstate) ilk_compute_wm_reg_maximums(dev_priv, 1, &max); - for (level = 1; level <= max_level; level++) { - struct intel_wm_level *wm = &pipe_wm->raw_wm[level]; + for (level = 1; level <= usable_level; level++) { + struct intel_wm_level *wm = &pipe_wm->wm[level]; ilk_compute_wm_level(dev_priv, intel_crtc, level, cstate, pristate, sprstate, curstate, wm); @@ -3111,13 +3105,10 @@ static int ilk_compute_pipe_wm(struct intel_crtc_state *cstate) * register maximums since such watermarks are * always invalid. */ - if (level > usable_level) - continue; - - if (ilk_validate_wm_level(level, &max, wm)) - pipe_wm->wm[level] = *wm; - else - usable_level = level; + if (!ilk_validate_wm_level(level, &max, wm)) { + memset(wm, 0, sizeof(*wm)); + break; + } } return 0; -- GitLab From b6b178a77210055b153dbc175e4468bd3c7122df Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Thu, 19 Oct 2017 17:13:41 +0200 Subject: [PATCH 0051/1471] drm/i915: Calculate ironlake intermediate watermarks correctly, v2. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The watermarks it should calculate against are the old optimal watermarks. The currently active crtc watermarks are pure fiction, and are invalid in case of a nonblocking modeset, page flip enabling/disabling planes or any other reason. When the crtc is disabled or during a modeset the intermediate watermarks don't need to be programmed separately, and could be directly assigned to the optimal watermarks. Changes since v1: - Use intel_atomic_get_old_crtc_state. (ville) Signed-off-by: Maarten Lankhorst Reviewed-by: Ville Syrjälä Reviewed-by: Matt Roper Link: https://patchwork.freedesktop.org/patch/msgid/20171019151341.4579-2-maarten.lankhorst@linux.intel.com [mlankhorst: Add cc stable and bugzilla link, since previous patch doesn't fix issue by itself] Cc: stable@vger.kernel.org #v4.8+ Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102373 --- drivers/gpu/drm/i915/intel_pm.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index e7926973112e6..07118c0b69d37 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3124,7 +3124,11 @@ static int ilk_compute_intermediate_wm(struct drm_device *dev, struct intel_crtc_state *newstate) { struct intel_pipe_wm *a = &newstate->wm.ilk.intermediate; - struct intel_pipe_wm *b = &intel_crtc->wm.active.ilk; + struct intel_atomic_state *intel_state = + to_intel_atomic_state(newstate->base.state); + const struct intel_crtc_state *oldstate = + intel_atomic_get_old_crtc_state(intel_state, intel_crtc); + const struct intel_pipe_wm *b = &oldstate->wm.ilk.optimal; int level, max_level = ilk_wm_max_level(to_i915(dev)); /* @@ -3133,6 +3137,9 @@ static int ilk_compute_intermediate_wm(struct drm_device *dev, * and after the vblank. */ *a = newstate->wm.ilk.optimal; + if (!newstate->base.active || drm_atomic_crtc_needs_modeset(&newstate->base)) + return 0; + a->pipe_enabled |= b->pipe_enabled; a->sprites_enabled |= b->sprites_enabled; a->sprites_scaled |= b->sprites_scaled; -- GitLab From 0501a3b0eb01ac2209ef6fce76153e5d6b07034e Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Thu, 26 Oct 2017 17:29:31 +0300 Subject: [PATCH 0052/1471] drm/i915/edp: read edp display control registers unconditionally MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Per my reading of the eDP spec, DP_DPCD_DISPLAY_CONTROL_CAPABLE bit in DP_EDP_CONFIGURATION_CAP should be set if the eDP display control registers starting at offset DP_EDP_DPCD_REV are "enabled". Currently we check the bit before reading the registers, and DP_EDP_DPCD_REV is the only way to detect eDP revision. Turns out there are (likely buggy) displays that require eDP 1.4+ features, such as supported link rates and link rate select, but do not have the bit set. Read the display control registers unconditionally. They are supposed to read zero anyway if they are not supported, so there should be no harm in this. This fixes the referenced bug by enabling the eDP version check, and thus reading of the supported link rates. The panel in question has 0 in DP_MAX_LINK_RATE which is only supported in eDP 1.4+. Without the supported link rates method we default to RBR which is insufficient for the panel native mode. As a curiosity, the panel also has a bogus value of 0x12 in DP_EDP_DPCD_REV, but that passes our check for >= DP_EDP_14 (which is 0x03). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103400 Reported-and-tested-by: Nicolas P. Cc: Ville Syrjälä Cc: stable@vger.kernel.org Reviewed-by: Ville Syrjälä Reviewed-by: Manasi Navare Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171026142932.17737-1-jani.nikula@intel.com --- drivers/gpu/drm/i915/intel_dp.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index aa75f55eeb61b..158438bb03891 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3735,9 +3735,16 @@ intel_edp_init_dpcd(struct intel_dp *intel_dp) } - /* Read the eDP Display control capabilities registers */ - if ((intel_dp->dpcd[DP_EDP_CONFIGURATION_CAP] & DP_DPCD_DISPLAY_CONTROL_CAPABLE) && - drm_dp_dpcd_read(&intel_dp->aux, DP_EDP_DPCD_REV, + /* + * Read the eDP display control registers. + * + * Do this independent of DP_DPCD_DISPLAY_CONTROL_CAPABLE bit in + * DP_EDP_CONFIGURATION_CAP, because some buggy displays do not have it + * set, but require eDP 1.4+ detection (e.g. for supported link rates + * method). The display control registers should read zero if they're + * not supported anyway. + */ + if (drm_dp_dpcd_read(&intel_dp->aux, DP_EDP_DPCD_REV, intel_dp->edp_dpcd, sizeof(intel_dp->edp_dpcd)) == sizeof(intel_dp->edp_dpcd)) DRM_DEBUG_KMS("EDP DPCD : %*ph\n", (int) sizeof(intel_dp->edp_dpcd), -- GitLab From e6ed2a1b99051bc04abcc2daed3972efada0b7a9 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Thu, 26 Oct 2017 17:29:32 +0300 Subject: [PATCH 0053/1471] drm/i915/edp: clean up code and comments around eDP DPCD read MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Some minor drive-by cleanups. Reviewed-by: Ville Syrjälä Reviewed-by: Manasi Navare Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171026142932.17737-2-jani.nikula@intel.com --- drivers/gpu/drm/i915/intel_dp.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 158438bb03891..73b1ed8cff6b6 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3747,11 +3747,11 @@ intel_edp_init_dpcd(struct intel_dp *intel_dp) if (drm_dp_dpcd_read(&intel_dp->aux, DP_EDP_DPCD_REV, intel_dp->edp_dpcd, sizeof(intel_dp->edp_dpcd)) == sizeof(intel_dp->edp_dpcd)) - DRM_DEBUG_KMS("EDP DPCD : %*ph\n", (int) sizeof(intel_dp->edp_dpcd), + DRM_DEBUG_KMS("eDP DPCD: %*ph\n", (int) sizeof(intel_dp->edp_dpcd), intel_dp->edp_dpcd); - /* Intermediate frequency support */ - if (intel_dp->edp_dpcd[0] >= 0x03) { /* eDp v1.4 or higher */ + /* Read the eDP 1.4+ supported link rates. */ + if (intel_dp->edp_dpcd[0] >= DP_EDP_14) { __le16 sink_rates[DP_MAX_SUPPORTED_RATES]; int i; @@ -3775,6 +3775,10 @@ intel_edp_init_dpcd(struct intel_dp *intel_dp) intel_dp->num_sink_rates = i; } + /* + * Use DP_LINK_RATE_SET if DP_SUPPORTED_LINK_RATES are available, + * default to DP_MAX_LINK_RATE and DP_LINK_BW_SET otherwise. + */ if (intel_dp->num_sink_rates) intel_dp->use_rate_select = true; else -- GitLab From e26612aa0959a80f51b25ebda853af8264036142 Mon Sep 17 00:00:00 2001 From: Dhinakaran Pandiyan Date: Fri, 11 Aug 2017 11:10:08 -0700 Subject: [PATCH 0054/1471] drm/dp: Bit definition for D3 power state that keeps AUX fully powered DPCD 600h - SET_POWER & SET_DP_PWR_VOLTAGE defines power state 101 = Set Main-Link for local Sink device and all downstream Sink devices to D3 (power-down mode), keep AUX block fully powered, ready to reply within a Response Timeout period of 300us. This state is useful in a MST dock + MST monitor configuration that doesn't wake up from D3 state. v2: Use spaces instead of tabs (Jani) Reviewed-by: Harry Wentland Signed-off-by: Dhinakaran Pandiyan Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/1502475008-2035-1-git-send-email-dhinakaran.pandiyan@intel.com --- include/drm/drm_dp_helper.h | 1 + 1 file changed, 1 insertion(+) diff --git a/include/drm/drm_dp_helper.h b/include/drm/drm_dp_helper.h index 11c39f15f1b32..d9bcc0861fba8 100644 --- a/include/drm/drm_dp_helper.h +++ b/include/drm/drm_dp_helper.h @@ -618,6 +618,7 @@ # define DP_SET_POWER_D0 0x1 # define DP_SET_POWER_D3 0x2 # define DP_SET_POWER_MASK 0x3 +# define DP_SET_POWER_D3_AUX_ON 0x5 #define DP_EDP_DPCD_REV 0x700 /* eDP 1.2 */ # define DP_EDP_11 0x00 -- GitLab From 11caf5517c2c9eebd435c60e9343a3784d74e9a1 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 27 Oct 2017 10:43:11 +0100 Subject: [PATCH 0055/1471] drm/i915: Empty the ring before disabling An interesting snippet from Sandybridge's prm: "Although a Ring Buffer can be enabled in the non-empty state, it must not be disabled unless it is empty. Attempting to disable a Ring Buffer in the non-empty state is UNDEFINED." Let's avoid the undefined behaviour as we disable the rings prior to reset and resume. v2: Tell HEAD to catch up to TAIL (empty ring) first, then reset both to 0 (supposedly while stopped). Signed-off-by: Chris Wilson Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171027094311.30380-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_ringbuffer.c | 6 +++++- drivers/gpu/drm/i915/intel_uncore.c | 6 +++++- 2 files changed, 10 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 05e01446b00b3..47fadf8da84e9 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -480,10 +480,14 @@ static bool stop_ring(struct intel_engine_cs *engine) } } - I915_WRITE_CTL(engine, 0); + I915_WRITE_HEAD(engine, I915_READ_TAIL(engine)); + I915_WRITE_HEAD(engine, 0); I915_WRITE_TAIL(engine, 0); + /* The ring must be empty before it is disabled */ + I915_WRITE_CTL(engine, 0); + return (I915_READ_HEAD(engine) & HEAD_ADDR) == 0; } diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 20e3c65c0999f..96ee6b2754bed 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1387,10 +1387,14 @@ static void gen3_stop_engine(struct intel_engine_cs *engine) DRM_DEBUG_DRIVER("%s: timed out on STOP_RING\n", engine->name); - I915_WRITE_FW(RING_CTL(base), 0); + I915_WRITE_FW(RING_HEAD(base), I915_READ_FW(RING_TAIL(base))); + I915_WRITE_FW(RING_HEAD(base), 0); I915_WRITE_FW(RING_TAIL(base), 0); + /* The ring must be empty before it is disabled */ + I915_WRITE_FW(RING_CTL(base), 0); + /* Check acts as a post */ if (I915_READ_FW(RING_HEAD(base)) != 0) DRM_DEBUG_DRIVER("%s: ring head not parked\n", -- GitLab From bea6e987c1ff358224e7bef7084be7650f5d1c38 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 26 Oct 2017 14:00:31 +0100 Subject: [PATCH 0056/1471] drm/i915: Hold rcu_read_lock when iterating over the radixtree (objects) Kasan spotted [IGT] gem_tiled_pread_pwrite: exiting, ret=0 ================================================================== BUG: KASAN: use-after-free in __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] Read of size 8 at addr ffff8801359da310 by task kworker/3:2/182 CPU: 3 PID: 182 Comm: kworker/3:2 Tainted: G U 4.14.0-rc6-CI-Custom_3340+ #1 Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017 Workqueue: events __i915_gem_free_work [i915] Call Trace: dump_stack+0x68/0xa0 print_address_description+0x78/0x290 ? __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] kasan_report+0x23d/0x350 __asan_report_load8_noabort+0x19/0x20 __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] ? i915_gem_object_truncate+0x100/0x100 [i915] ? lock_acquire+0x380/0x380 __i915_gem_object_put_pages+0x30d/0x530 [i915] __i915_gem_free_objects+0x551/0xbd0 [i915] ? lock_acquire+0x13e/0x380 __i915_gem_free_work+0x4e/0x70 [i915] process_one_work+0x6f6/0x1590 ? pwq_dec_nr_in_flight+0x2b0/0x2b0 worker_thread+0xe6/0xe90 ? pci_mmcfg_check_reserved+0x110/0x110 kthread+0x309/0x410 ? process_one_work+0x1590/0x1590 ? kthread_create_on_node+0xb0/0xb0 ret_from_fork+0x27/0x40 Allocated by task 1801: save_stack_trace+0x1b/0x20 kasan_kmalloc+0xee/0x190 kasan_slab_alloc+0x12/0x20 kmem_cache_alloc+0xdc/0x2e0 radix_tree_node_alloc.constprop.12+0x48/0x330 __radix_tree_create+0x274/0x480 __radix_tree_insert+0xa2/0x610 i915_gem_object_get_sg+0x224/0x670 [i915] i915_gem_object_get_page+0xb5/0x1c0 [i915] i915_gem_pread_ioctl+0x822/0xf60 [i915] drm_ioctl_kernel+0x13f/0x1c0 drm_ioctl+0x6cf/0x980 do_vfs_ioctl+0x184/0xf30 SyS_ioctl+0x41/0x70 entry_SYSCALL_64_fastpath+0x1c/0xb1 Freed by task 37: save_stack_trace+0x1b/0x20 kasan_slab_free+0xaf/0x190 kmem_cache_free+0xbf/0x340 radix_tree_node_rcu_free+0x79/0x90 rcu_process_callbacks+0x46d/0xf40 __do_softirq+0x21c/0x8d3 The buggy address belongs to the object at ffff8801359da0f0 which belongs to the cache radix_tree_node of size 576 The buggy address is located 544 bytes inside of 576-byte region [ffff8801359da0f0, ffff8801359da330) The buggy address belongs to the page: page:ffffea0004d67600 count:1 mapcount:0 mapping: (null) index:0x0 compound_mapcount: 0 flags: 0x8000000000008100(slab|head) raw: 8000000000008100 0000000000000000 0000000000000000 0000000100110011 raw: ffffea0004b52920 ffffea0004b38020 ffff88015b416a80 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8801359da200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8801359da280: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8801359da300: fb fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc ^ ffff8801359da380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff8801359da400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ================================================================== Disabling lock debugging due to kernel taint which looks like the slab containing the radixtree iter was freed as we traversed the tree, taking the rcu read lock across the loop should prevent that (deferring all the frees until the end). Reported-by: Tomi Sarvela Fixes: 96d776345277 ("drm/i915: Use a radixtree for random access to the object's backing storage") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171026130032.10677-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_gem.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index c2506fb3a4836..e43688f778172 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2267,8 +2267,10 @@ static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj) struct radix_tree_iter iter; void __rcu **slot; + rcu_read_lock(); radix_tree_for_each_slot(slot, &obj->mm.get_page.radix, &iter, 0) radix_tree_delete(&obj->mm.get_page.radix, iter.index); + rcu_read_unlock(); } void __i915_gem_object_put_pages(struct drm_i915_gem_object *obj, -- GitLab From 547da76b5777859f98bb78e6b57f19463f803c04 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 26 Oct 2017 14:00:32 +0100 Subject: [PATCH 0057/1471] drm/i915: Hold rcu_read_lock when iterating over the radixtree (vma idr) Kasan spotted [IGT] gem_tiled_pread_pwrite: exiting, ret=0 ================================================================== BUG: KASAN: use-after-free in __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] Read of size 8 at addr ffff8801359da310 by task kworker/3:2/182 CPU: 3 PID: 182 Comm: kworker/3:2 Tainted: G U 4.14.0-rc6-CI-Custom_3340+ #1 Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017 Workqueue: events __i915_gem_free_work [i915] Call Trace: dump_stack+0x68/0xa0 print_address_description+0x78/0x290 ? __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] kasan_report+0x23d/0x350 __asan_report_load8_noabort+0x19/0x20 __i915_gem_object_reset_page_iter+0x15c/0x170 [i915] ? i915_gem_object_truncate+0x100/0x100 [i915] ? lock_acquire+0x380/0x380 __i915_gem_object_put_pages+0x30d/0x530 [i915] __i915_gem_free_objects+0x551/0xbd0 [i915] ? lock_acquire+0x13e/0x380 __i915_gem_free_work+0x4e/0x70 [i915] process_one_work+0x6f6/0x1590 ? pwq_dec_nr_in_flight+0x2b0/0x2b0 worker_thread+0xe6/0xe90 ? pci_mmcfg_check_reserved+0x110/0x110 kthread+0x309/0x410 ? process_one_work+0x1590/0x1590 ? kthread_create_on_node+0xb0/0xb0 ret_from_fork+0x27/0x40 Allocated by task 1801: save_stack_trace+0x1b/0x20 kasan_kmalloc+0xee/0x190 kasan_slab_alloc+0x12/0x20 kmem_cache_alloc+0xdc/0x2e0 radix_tree_node_alloc.constprop.12+0x48/0x330 __radix_tree_create+0x274/0x480 __radix_tree_insert+0xa2/0x610 i915_gem_object_get_sg+0x224/0x670 [i915] i915_gem_object_get_page+0xb5/0x1c0 [i915] i915_gem_pread_ioctl+0x822/0xf60 [i915] drm_ioctl_kernel+0x13f/0x1c0 drm_ioctl+0x6cf/0x980 do_vfs_ioctl+0x184/0xf30 SyS_ioctl+0x41/0x70 entry_SYSCALL_64_fastpath+0x1c/0xb1 Freed by task 37: save_stack_trace+0x1b/0x20 kasan_slab_free+0xaf/0x190 kmem_cache_free+0xbf/0x340 radix_tree_node_rcu_free+0x79/0x90 rcu_process_callbacks+0x46d/0xf40 __do_softirq+0x21c/0x8d3 The buggy address belongs to the object at ffff8801359da0f0 which belongs to the cache radix_tree_node of size 576 The buggy address is located 544 bytes inside of 576-byte region [ffff8801359da0f0, ffff8801359da330) The buggy address belongs to the page: page:ffffea0004d67600 count:1 mapcount:0 mapping: (null) index:0x0 compound_mapcount: 0 flags: 0x8000000000008100(slab|head) raw: 8000000000008100 0000000000000000 0000000000000000 0000000100110011 raw: ffffea0004b52920 ffffea0004b38020 ffff88015b416a80 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8801359da200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8801359da280: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8801359da300: fb fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc ^ ffff8801359da380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ffff8801359da400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc ================================================================== Disabling lock debugging due to kernel taint which looks like the slab containing the radixtree iter was freed as we traversed the tree, taking the rcu read lock across the loop should prevent that (deferring all the frees until the end). Reported-by: Tomi Sarvela Fixes: d1b48c1e7184 ("drm/i915: Replace execbuf vma ht with an idr") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171026130032.10677-2-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_gem_context.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 4f26f80b1b3eb..10affb35ac566 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -104,6 +104,7 @@ static void lut_close(struct i915_gem_context *ctx) kmem_cache_free(ctx->i915->luts, lut); } + rcu_read_lock(); radix_tree_for_each_slot(slot, &ctx->handles_vma, &iter, 0) { struct i915_vma *vma = rcu_dereference_raw(*slot); struct drm_i915_gem_object *obj = vma->obj; @@ -115,6 +116,7 @@ static void lut_close(struct i915_gem_context *ctx) __i915_gem_object_release_unless_active(obj); } + rcu_read_unlock(); } static void i915_gem_context_free(struct i915_gem_context *ctx) -- GitLab From 1dd3a060c7d01c00a3c2b18eff5a3428b9e2c28f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Thu, 26 Oct 2017 18:57:26 +0200 Subject: [PATCH 0058/1471] drm/gem: Remove trailing whitespace MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Remove two trailing spaces. Signed-off-by: Noralf Trønnes Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171026165731.5793-4-noralf@tronnes.org --- drivers/gpu/drm/drm_gem.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 55d6182555c73..4c84b23d37ccf 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -348,7 +348,7 @@ EXPORT_SYMBOL_GPL(drm_gem_dumb_map_offset); * @file: drm file-private structure to remove the dumb handle from * @dev: corresponding drm_device * @handle: the dumb handle to remove - * + * * This implements the &drm_driver.dumb_destroy kms driver callback for drivers * which use gem to manage their backing storage. */ @@ -365,7 +365,7 @@ EXPORT_SYMBOL(drm_gem_dumb_destroy); * @file_priv: drm file-private structure to register the handle for * @obj: object to register * @handlep: pointer to return the created handle to the caller - * + * * This expects the &drm_device.object_name_lock to be held already and will * drop it before returning. Used to avoid races in establishing new handles * when importing an object from either an flink name or a dma-buf. -- GitLab From 5448f53ffe62c75e2cf84d407510efadaa1aff8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 19 Oct 2017 16:37:12 +0300 Subject: [PATCH 0059/1471] drm/i915: Don't use encoder->type in intel_ddi_set_pipe_settings() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit encoder->type isn't reliable for DP/HDMI so instead extract the correct type from the crtc state in intel_ddi_set_pipe_settings(). Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171019133721.11794-2-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_ddi.c | 47 ++++++++++++++++---------------- 1 file changed, 24 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 28c25cb9eb2c9..c23d55c3344cc 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1504,33 +1504,34 @@ void intel_ddi_set_pipe_settings(const struct intel_crtc_state *crtc_state) { struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - struct intel_encoder *encoder = intel_ddi_get_crtc_encoder(crtc); enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; - int type = encoder->type; - uint32_t temp; + u32 temp; - if (type == INTEL_OUTPUT_DP || type == INTEL_OUTPUT_EDP || type == INTEL_OUTPUT_DP_MST) { - WARN_ON(transcoder_is_dsi(cpu_transcoder)); + if (!intel_crtc_has_dp_encoder(crtc_state)) + return; - temp = TRANS_MSA_SYNC_CLK; - switch (crtc_state->pipe_bpp) { - case 18: - temp |= TRANS_MSA_6_BPC; - break; - case 24: - temp |= TRANS_MSA_8_BPC; - break; - case 30: - temp |= TRANS_MSA_10_BPC; - break; - case 36: - temp |= TRANS_MSA_12_BPC; - break; - default: - BUG(); - } - I915_WRITE(TRANS_MSA_MISC(cpu_transcoder), temp); + WARN_ON(transcoder_is_dsi(cpu_transcoder)); + + temp = TRANS_MSA_SYNC_CLK; + switch (crtc_state->pipe_bpp) { + case 18: + temp |= TRANS_MSA_6_BPC; + break; + case 24: + temp |= TRANS_MSA_8_BPC; + break; + case 30: + temp |= TRANS_MSA_10_BPC; + break; + case 36: + temp |= TRANS_MSA_12_BPC; + break; + default: + MISSING_CASE(crtc_state->pipe_bpp); + break; } + + I915_WRITE(TRANS_MSA_MISC(cpu_transcoder), temp); } void intel_ddi_set_vc_payload_alloc(const struct intel_crtc_state *crtc_state, -- GitLab From 3a6d84e677d4e5a6366437b7007e6a2f97aba9d1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 19 Oct 2017 16:37:13 +0300 Subject: [PATCH 0060/1471] drm/i915: Pass crtc state to intel_prepare_dp_ddi_buffers() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Eliminate intel_prepare_dp_ddi_buffers()'s reliance on the encoder->type by passing in the crtc state. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171019133721.11794-3-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_ddi.c | 25 +++++++++---------------- 1 file changed, 9 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index c23d55c3344cc..0595824954965 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -811,7 +811,8 @@ static int intel_ddi_hdmi_level(struct drm_i915_private *dev_priv, enum port por * values in advance. This function programs the correct values for * DP/eDP/FDI use cases. */ -static void intel_prepare_dp_ddi_buffers(struct intel_encoder *encoder) +static void intel_prepare_dp_ddi_buffers(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); u32 iboost_bit = 0; @@ -819,23 +820,15 @@ static void intel_prepare_dp_ddi_buffers(struct intel_encoder *encoder) enum port port = intel_ddi_get_encoder_port(encoder); const struct ddi_buf_trans *ddi_translations; - switch (encoder->type) { - case INTEL_OUTPUT_EDP: + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_ANALOG)) + ddi_translations = intel_ddi_get_buf_trans_fdi(dev_priv, + &n_entries); + else if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_EDP)) ddi_translations = intel_ddi_get_buf_trans_edp(dev_priv, port, &n_entries); - break; - case INTEL_OUTPUT_DP: + else ddi_translations = intel_ddi_get_buf_trans_dp(dev_priv, port, &n_entries); - break; - case INTEL_OUTPUT_ANALOG: - ddi_translations = intel_ddi_get_buf_trans_fdi(dev_priv, - &n_entries); - break; - default: - MISSING_CASE(encoder->type); - return; - } /* If we're boosting the current, set bit 31 of trans1 */ if (IS_GEN9_BC(dev_priv) && @@ -937,7 +930,7 @@ void hsw_fdi_link_train(struct intel_crtc *crtc, for_each_encoder_on_crtc(dev, &crtc->base, encoder) { WARN_ON(encoder->type != INTEL_OUTPUT_ANALOG); - intel_prepare_dp_ddi_buffers(encoder); + intel_prepare_dp_ddi_buffers(encoder, crtc_state); } /* Set the FDI_RX_MISC pwrdn lanes and the 2 workarounds listed at the @@ -2201,7 +2194,7 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, else if (IS_GEN9_LP(dev_priv)) bxt_ddi_vswing_sequence(encoder, level, encoder->type); else - intel_prepare_dp_ddi_buffers(encoder); + intel_prepare_dp_ddi_buffers(encoder, crtc_state); intel_ddi_init_dp_buf_reg(encoder); if (!is_mst) -- GitLab From f49b44ab84035b0f9c1808770684432bb6804328 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 19 Oct 2017 16:37:14 +0300 Subject: [PATCH 0061/1471] drm/i915: Start using output_types for DPLL selection MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit encoder->type is not realiable for DP/HDMI so let's switch the DPLL selection over to using output_types. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171019133721.11794-4-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 36 ++++++++++----------------- 1 file changed, 13 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index 897fffe1ecd83..a83bf1c38e05f 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -813,15 +813,11 @@ hsw_get_dpll(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, memset(&crtc_state->dpll_hw_state, 0, sizeof(crtc_state->dpll_hw_state)); - if (encoder->type == INTEL_OUTPUT_HDMI) { + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) { pll = hsw_ddi_hdmi_get_dpll(clock, crtc, crtc_state); - - } else if (encoder->type == INTEL_OUTPUT_DP || - encoder->type == INTEL_OUTPUT_DP_MST || - encoder->type == INTEL_OUTPUT_EDP) { + } else if (intel_crtc_has_dp_encoder(crtc_state)) { pll = hsw_ddi_dp_get_dpll(encoder, clock); - - } else if (encoder->type == INTEL_OUTPUT_ANALOG) { + } else if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_ANALOG)) { if (WARN_ON(crtc_state->port_clock / 2 != 135000)) return NULL; @@ -1369,15 +1365,13 @@ skl_get_dpll(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, memset(&dpll_hw_state, 0, sizeof(dpll_hw_state)); - if (encoder->type == INTEL_OUTPUT_HDMI) { + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) { bret = skl_ddi_hdmi_pll_dividers(crtc, crtc_state, clock); if (!bret) { DRM_DEBUG_KMS("Could not get HDMI pll dividers.\n"); return NULL; } - } else if (encoder->type == INTEL_OUTPUT_DP || - encoder->type == INTEL_OUTPUT_DP_MST || - encoder->type == INTEL_OUTPUT_EDP) { + } else if (intel_crtc_has_dp_encoder(crtc_state)) { bret = skl_ddi_dp_set_dpll_hw_state(clock, &dpll_hw_state); if (!bret) { DRM_DEBUG_KMS("Could not set DP dpll HW state.\n"); @@ -1388,7 +1382,7 @@ skl_get_dpll(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, return NULL; } - if (encoder->type == INTEL_OUTPUT_EDP) + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_EDP)) pll = intel_find_shared_dpll(crtc, crtc_state, DPLL_ID_SKL_DPLL0, DPLL_ID_SKL_DPLL0); @@ -1812,14 +1806,12 @@ bxt_get_dpll(struct intel_crtc *crtc, struct intel_shared_dpll *pll; int i, clock = crtc_state->port_clock; - if (encoder->type == INTEL_OUTPUT_HDMI && + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI) && !bxt_ddi_hdmi_set_dpll_hw_state(crtc, crtc_state, clock, &dpll_hw_state)) return NULL; - if ((encoder->type == INTEL_OUTPUT_DP || - encoder->type == INTEL_OUTPUT_EDP || - encoder->type == INTEL_OUTPUT_DP_MST) && + if (intel_crtc_has_dp_encoder(crtc_state) && !bxt_ddi_dp_set_dpll_hw_state(clock, &dpll_hw_state)) return NULL; @@ -1828,7 +1820,7 @@ bxt_get_dpll(struct intel_crtc *crtc, crtc_state->dpll_hw_state = dpll_hw_state; - if (encoder->type == INTEL_OUTPUT_DP_MST) { + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) { struct intel_dp_mst_encoder *intel_mst = enc_to_mst(&encoder->base); intel_dig_port = intel_mst->primary; @@ -2345,15 +2337,13 @@ cnl_get_dpll(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, memset(&dpll_hw_state, 0, sizeof(dpll_hw_state)); - if (encoder->type == INTEL_OUTPUT_HDMI) { + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) { bret = cnl_ddi_hdmi_pll_dividers(crtc, crtc_state, clock); if (!bret) { DRM_DEBUG_KMS("Could not get HDMI pll dividers.\n"); return NULL; } - } else if (encoder->type == INTEL_OUTPUT_DP || - encoder->type == INTEL_OUTPUT_DP_MST || - encoder->type == INTEL_OUTPUT_EDP) { + } else if (intel_crtc_has_dp_encoder(crtc_state)) { bret = cnl_ddi_dp_set_dpll_hw_state(clock, &dpll_hw_state); if (!bret) { DRM_DEBUG_KMS("Could not set DP dpll HW state.\n"); @@ -2361,8 +2351,8 @@ cnl_get_dpll(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, } crtc_state->dpll_hw_state = dpll_hw_state; } else { - DRM_DEBUG_KMS("Skip DPLL setup for encoder %d\n", - encoder->type); + DRM_DEBUG_KMS("Skip DPLL setup for output_types 0x%x\n", + crtc_state->output_types); return NULL; } -- GitLab From 742745f1ee4f5a36e5cc5e8b360b519df45efa89 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 19 Oct 2017 16:37:15 +0300 Subject: [PATCH 0062/1471] drm/i915: Stop using encoder->type in intel_ddi_enable_transcoder_func() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit intel_ddi_enable_transcoder_func() already has the crtc state so we can use that instead of the untrustworthy encoder->type. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171019133721.11794-5-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_ddi.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 0595824954965..c721775687e72 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1550,7 +1550,6 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) enum pipe pipe = crtc->pipe; enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; enum port port = intel_ddi_get_encoder_port(encoder); - int type = encoder->type; uint32_t temp; /* Enable TRANS_DDI_FUNC_CTL for the pipe to work in HDMI mode */ @@ -1605,7 +1604,7 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) } } - if (type == INTEL_OUTPUT_HDMI) { + if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) { if (crtc_state->has_hdmi_sink) temp |= TRANS_DDI_MODE_SELECT_HDMI; else @@ -1615,19 +1614,15 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) temp |= TRANS_DDI_HDMI_SCRAMBLING_MASK; if (crtc_state->hdmi_high_tmds_clock_ratio) temp |= TRANS_DDI_HIGH_TMDS_CHAR_RATE; - } else if (type == INTEL_OUTPUT_ANALOG) { + } else if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_ANALOG)) { temp |= TRANS_DDI_MODE_SELECT_FDI; temp |= (crtc_state->fdi_lanes - 1) << 1; - } else if (type == INTEL_OUTPUT_DP || - type == INTEL_OUTPUT_EDP) { - temp |= TRANS_DDI_MODE_SELECT_DP_SST; - temp |= DDI_PORT_WIDTH(crtc_state->lane_count); - } else if (type == INTEL_OUTPUT_DP_MST) { + } else if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) { temp |= TRANS_DDI_MODE_SELECT_DP_MST; temp |= DDI_PORT_WIDTH(crtc_state->lane_count); } else { - WARN(1, "Invalid encoder type %d for pipe %c\n", - encoder->type, pipe_name(pipe)); + temp |= TRANS_DDI_MODE_SELECT_DP_SST; + temp |= DDI_PORT_WIDTH(crtc_state->lane_count); } I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder), temp); -- GitLab From 5161d058dff4d53c78a046350d64beff54e9a9f7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 16:43:48 +0300 Subject: [PATCH 0063/1471] drm/i915: Fix BXT lane latency optimal setting with MST MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Call the DDI .pre_pll_enable() hook from the MST code so that BXT gets the correct lane latency optimal setting applied. And we obviously need to compute the correct value, and read it out to keep the state checker happy. While at it drop the useless 'encoder' parameter to bxt_ddi_phy_calc_lane_lat_optim_mask() Cc: Maarten Lankhorst Cc: Imre Deak Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027134348.31190-1-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/i915_drv.h | 3 +-- drivers/gpu/drm/i915/intel_ddi.c | 3 +-- drivers/gpu/drm/i915/intel_dp_mst.c | 23 +++++++++++++++++++++++ drivers/gpu/drm/i915/intel_dpio_phy.c | 3 +-- 4 files changed, 26 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 61c155cbf9d71..c0a716e596bae 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -4179,8 +4179,7 @@ bool bxt_ddi_phy_is_enabled(struct drm_i915_private *dev_priv, enum dpio_phy phy); bool bxt_ddi_phy_verify_state(struct drm_i915_private *dev_priv, enum dpio_phy phy); -uint8_t bxt_ddi_phy_calc_lane_lat_optim_mask(struct intel_encoder *encoder, - uint8_t lane_count); +uint8_t bxt_ddi_phy_calc_lane_lat_optim_mask(uint8_t lane_count); void bxt_ddi_phy_set_lane_optim_mask(struct intel_encoder *encoder, uint8_t lane_lat_optim_mask); uint8_t bxt_ddi_phy_get_lane_lat_optim_mask(struct intel_encoder *encoder); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index c721775687e72..9c118e5305f71 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2662,8 +2662,7 @@ static bool intel_ddi_compute_config(struct intel_encoder *encoder, if (IS_GEN9_LP(dev_priv) && ret) pipe_config->lane_lat_optim_mask = - bxt_ddi_phy_calc_lane_lat_optim_mask(encoder, - pipe_config->lane_count); + bxt_ddi_phy_calc_lane_lat_optim_mask(pipe_config->lane_count); intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 3d62c63c07636..c34ffa959e900 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -88,6 +88,10 @@ static bool intel_dp_mst_compute_config(struct intel_encoder *encoder, pipe_config->dp_m_n.tu = slots; + if (IS_GEN9_LP(dev_priv)) + pipe_config->lane_lat_optim_mask = + bxt_ddi_phy_calc_lane_lat_optim_mask(pipe_config->lane_count); + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); return true; @@ -182,6 +186,20 @@ static void intel_mst_post_disable_dp(struct intel_encoder *encoder, DRM_DEBUG_KMS("active links %d\n", intel_dp->active_mst_links); } +static void intel_mst_pre_pll_enable_dp(struct intel_encoder *encoder, + const struct intel_crtc_state *pipe_config, + const struct drm_connector_state *conn_state) +{ + struct intel_dp_mst_encoder *intel_mst = enc_to_mst(&encoder->base); + struct intel_digital_port *intel_dig_port = intel_mst->primary; + struct intel_dp *intel_dp = &intel_dig_port->dp; + + if (intel_dp->active_mst_links == 0 && + intel_dig_port->base.pre_pll_enable) + intel_dig_port->base.pre_pll_enable(&intel_dig_port->base, + pipe_config, NULL); +} + static void intel_mst_pre_enable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *pipe_config, const struct drm_connector_state *conn_state) @@ -311,6 +329,10 @@ static void intel_dp_mst_enc_get_config(struct intel_encoder *encoder, intel_ddi_clock_get(&intel_dig_port->base, pipe_config); + if (IS_GEN9_LP(dev_priv)) + pipe_config->lane_lat_optim_mask = + bxt_ddi_phy_get_lane_lat_optim_mask(encoder); + intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); } @@ -582,6 +604,7 @@ intel_dp_create_fake_mst_encoder(struct intel_digital_port *intel_dig_port, enum intel_encoder->compute_config = intel_dp_mst_compute_config; intel_encoder->disable = intel_mst_disable_dp; intel_encoder->post_disable = intel_mst_post_disable_dp; + intel_encoder->pre_pll_enable = intel_mst_pre_pll_enable_dp; intel_encoder->pre_enable = intel_mst_pre_enable_dp; intel_encoder->enable = intel_mst_enable_dp; intel_encoder->get_hw_state = intel_dp_mst_enc_get_hw_state; diff --git a/drivers/gpu/drm/i915/intel_dpio_phy.c b/drivers/gpu/drm/i915/intel_dpio_phy.c index de38d014ed39b..63b76eac018fa 100644 --- a/drivers/gpu/drm/i915/intel_dpio_phy.c +++ b/drivers/gpu/drm/i915/intel_dpio_phy.c @@ -567,8 +567,7 @@ bool bxt_ddi_phy_verify_state(struct drm_i915_private *dev_priv, } uint8_t -bxt_ddi_phy_calc_lane_lat_optim_mask(struct intel_encoder *encoder, - uint8_t lane_count) +bxt_ddi_phy_calc_lane_lat_optim_mask(uint8_t lane_count) { switch (lane_count) { case 1: -- GitLab From 857c416e58dc306b15b7919117b6225544662bfa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 12:45:23 +0300 Subject: [PATCH 0064/1471] drm/i915: Improve DP downstream HPD handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit DP dongles may signal downstream HPD via short HPD pulses. Setting the sink to DPMS off apparently kills the downstream HPD (at least on my DP->VGA dongle), so skip the DPMS off for such dongles when we turn off the port. v2: Deal with DDI as well by moving the check into intel_dp_sink_dpms() (Dhinakaran) Cc: Dhinakaran Pandiyan Cc: David Woodhouse Cc: Imre Deak Cc: Pablo Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103472 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99114 Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027094523.9317-1-ville.syrjala@linux.intel.com Reviewed-by: Jani Nikula --- drivers/gpu/drm/i915/intel_dp.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 73b1ed8cff6b6..8d25a019b772d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2505,6 +2505,21 @@ static void ironlake_edp_pll_off(struct intel_dp *intel_dp) udelay(200); } +static bool downstream_hpd_needs_d0(struct intel_dp *intel_dp) +{ + /* + * DPCD 1.2+ should support BRANCH_DEVICE_CTRL, and thus + * be capable of signalling downstream hpd with a long pulse. + * Whether or not that means D3 is safe to use is not clear, + * but let's assume so until proven otherwise. + * + * FIXME should really check all downstream ports... + */ + return intel_dp->dpcd[DP_DPCD_REV] == 0x11 && + intel_dp->dpcd[DP_DOWNSTREAMPORT_PRESENT] & DP_DWN_STRM_PORT_PRESENT && + intel_dp->downstream_ports[0] & DP_DS_PORT_HPD; +} + /* If the sink supports it, try to set the power state appropriately */ void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode) { @@ -2515,6 +2530,9 @@ void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode) return; if (mode != DRM_MODE_DPMS_ON) { + if (downstream_hpd_needs_d0(intel_dp)) + return; + ret = drm_dp_dpcd_writeb(&intel_dp->aux, DP_SET_POWER, DP_SET_POWER_D3); } else { -- GitLab From be33be5df2a0d4b534ace0aaad65cba8144bfce8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 26 Oct 2017 18:14:04 +0300 Subject: [PATCH 0065/1471] drm/i915: Clean up the mess around hdmi_12bpc_possible() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Move the crtc state related 12bpc checks into hdmi_12bpc_possible() since that one already examines other parts of the crtc state. Note that we can drop the !force_dvi check since crtc_state->has_hdmi_sink already accounts for that. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171026151405.30710-1-ville.syrjala@linux.intel.com Reviewed-by: Chris Wilson --- drivers/gpu/drm/i915/intel_hdmi.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 5132dc8147884..aa486b8925cf5 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1336,6 +1336,12 @@ static bool hdmi_12bpc_possible(const struct intel_crtc_state *crtc_state) if (HAS_GMCH_DISPLAY(dev_priv)) return false; + if (crtc_state->pipe_bpp <= 8*3) + return false; + + if (!crtc_state->has_hdmi_sink) + return false; + /* * HDMI 12bpc affects the clocks, so it's only possible * when not cloning with other encoder types. @@ -1461,9 +1467,8 @@ bool intel_hdmi_compute_config(struct intel_encoder *encoder, * outputs. We also need to check that the higher clock still fits * within limits. */ - if (pipe_config->pipe_bpp > 8*3 && pipe_config->has_hdmi_sink && !force_dvi && - hdmi_port_clock_valid(intel_hdmi, clock_12bpc, true, force_dvi) == MODE_OK && - hdmi_12bpc_possible(pipe_config)) { + if (hdmi_12bpc_possible(pipe_config) && + hdmi_port_clock_valid(intel_hdmi, clock_12bpc, true, force_dvi) == MODE_OK) { DRM_DEBUG_KMS("picking bpc to 12 for HDMI output\n"); desired_bpp = 12*3; -- GitLab From 4e7b212075d9c4b71af9528bdaa81e642fab0e90 Mon Sep 17 00:00:00 2001 From: Aastha Gupta Date: Mon, 23 Oct 2017 22:08:06 +0530 Subject: [PATCH 0066/1471] drm: virtio: replace reference/unreference with get/put DRM core uses reference/unreference suffixes for refcounting functions, but kernel uses get/put (e.g. kref_get/put()). Replace reference/unreference with get/put for consistency and also it's shorter. The following cocci script was used to generate the patch: @@ expression e; @@ ( -drm_gem_object_reference(e); +drm_gem_object_get(e); | -drm_gem_object_unreference(e); +drm_gem_object_put(e); | -drm_gem_object_unreference_unlocked(e); +drm_gem_object_put_unlocked(e); ) Signed-off-by: Aastha Gupta Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/1508776686-29664-1-git-send-email-aastha.gupta4104@gmail.com --- drivers/gpu/drm/virtio/virtgpu_ioctl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c b/drivers/gpu/drm/virtio/virtgpu_ioctl.c index 0528edb4a2bfe..461f81aa1bbeb 100644 --- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c +++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c @@ -261,7 +261,7 @@ static int virtio_gpu_resource_create_ioctl(struct drm_device *dev, void *data, ret = virtio_gpu_object_attach(vgdev, qobj, res_id, NULL); } else { /* use a gem reference since unref list undoes them */ - drm_gem_object_reference(&qobj->gem_base); + drm_gem_object_get(&qobj->gem_base); mainbuf.bo = &qobj->tbo; list_add(&mainbuf.head, &validate_list); -- GitLab From 39c9ce70f13a3127a8057e30b48a239bf08f7c8e Mon Sep 17 00:00:00 2001 From: Pan Bian Date: Sun, 29 Oct 2017 13:55:12 +0800 Subject: [PATCH 0067/1471] gma500: mmu: unmap the correct address The function kunmap_atomatic() is called on the same variable twice, i.e. pt->v. In the second call, its parameter should be variable v rather than pt->v. Signed-off-by: Pan Bian Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/1509256512-5962-1-git-send-email-bianpan2016@163.com --- drivers/gpu/drm/gma500/mmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/gma500/mmu.c b/drivers/gpu/drm/gma500/mmu.c index 0eaf11c199395..ccb161c73a59b 100644 --- a/drivers/gpu/drm/gma500/mmu.c +++ b/drivers/gpu/drm/gma500/mmu.c @@ -395,7 +395,7 @@ static void psb_mmu_pt_unmap_unlock(struct psb_mmu_pt *pt) psb_mmu_clflush(pd->driver, (void *)&v[pt->index]); atomic_set(&pd->driver->needs_tlbflush, 1); } - kunmap_atomic(pt->v); + kunmap_atomic(v); spin_unlock(&pd->driver->lock); psb_mmu_free_pt(pt); return; -- GitLab From 6e8fbf8d19e48ca3b451233c8939b493a51d030d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 23:17:38 +0300 Subject: [PATCH 0068/1471] drm/i915/vbt: Fix HDMI level shifter and max data rate bitfield sizes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The HDMI level shifter value should be 5 bits and the max data rate 3 bits. Cc: Jani Nikula Reported-by: Jani Nikula Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027201738.3640-1-ville.syrjala@linux.intel.com Reviewed-by: Jani Nikula --- drivers/gpu/drm/i915/intel_vbt_defs.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_vbt_defs.h b/drivers/gpu/drm/i915/intel_vbt_defs.h index f225c288a121d..3c3c421e2e43c 100644 --- a/drivers/gpu/drm/i915/intel_vbt_defs.h +++ b/drivers/gpu/drm/i915/intel_vbt_defs.h @@ -342,8 +342,8 @@ struct child_device_config { u8 i2c_speed; u8 dp_onboard_redriver; /* 158 */ u8 dp_ondock_redriver; /* 158 */ - u8 hdmi_level_shifter_value:4; /* 169 */ - u8 hdmi_max_data_rate:4; /* 204 */ + u8 hdmi_level_shifter_value:5; /* 169 */ + u8 hdmi_max_data_rate:3; /* 204 */ u16 dtd_buf_ptr; /* 161 */ u8 edidless_efp:1; /* 161 */ u8 compression_enable:1; /* 198 */ -- GitLab From d6038611aa3d7d8080a8ae7f3122779fb4779a13 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 30 Oct 2017 16:57:02 +0200 Subject: [PATCH 0069/1471] drm/i915: Parse max HDMI TMDS clock from VBT MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Starting from version 204 VBT can specify the max TMDS clock we are allowed to use with HDMI ports. Parse that information and take it into account when filtering modes and computing a crtc state. Also take the opportunity to sort the platform check if ladder from new to old. v2: Add defines for the values into intel_vbt_defs.h (Jani) Don't fall back to 0 silently for unknown values (Jani) Skip the debug print for the 0 case (Jani) Cc: Jani Nikula Reviewed-by: Jani Nikula Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171030145702.23662-1-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/intel_bios.c | 24 +++++++++++++++++++++ drivers/gpu/drm/i915/intel_hdmi.c | 30 ++++++++++++++++++--------- drivers/gpu/drm/i915/intel_vbt_defs.h | 4 ++++ 4 files changed, 50 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index c0a716e596bae..4a7325c4189c5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1698,6 +1698,8 @@ enum modeset_restore { #define DDC_PIN_D 0x06 struct ddi_vbt_port_info { + int max_tmds_clock; + /* * This is an index in the HDMI/DVI DDI buffer translation table. * The special value HDMI_LEVEL_SHIFT_UNKNOWN means the VBT didn't diff --git a/drivers/gpu/drm/i915/intel_bios.c b/drivers/gpu/drm/i915/intel_bios.c index 5e122673d32a7..5c016b1f0f247 100644 --- a/drivers/gpu/drm/i915/intel_bios.c +++ b/drivers/gpu/drm/i915/intel_bios.c @@ -1227,6 +1227,30 @@ static void parse_ddi_port(struct drm_i915_private *dev_priv, enum port port, info->hdmi_level_shift = hdmi_level_shift; } + if (bdb_version >= 204) { + int max_tmds_clock; + + switch (child->hdmi_max_data_rate) { + default: + MISSING_CASE(child->hdmi_max_data_rate); + /* fall through */ + case HDMI_MAX_DATA_RATE_PLATFORM: + max_tmds_clock = 0; + break; + case HDMI_MAX_DATA_RATE_297: + max_tmds_clock = 297000; + break; + case HDMI_MAX_DATA_RATE_165: + max_tmds_clock = 165000; + break; + } + + if (max_tmds_clock) + DRM_DEBUG_KMS("VBT HDMI max TMDS clock for port %c: %d kHz\n", + port_name(port), max_tmds_clock); + info->max_tmds_clock = max_tmds_clock; + } + /* Parse the I_boost config for SKL and above */ if (bdb_version >= 196 && child->iboost) { info->dp_boost_level = translate_iboost(child->dp_iboost_level); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index aa486b8925cf5..38fe24565b4d1 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1224,24 +1224,34 @@ static void pch_post_disable_hdmi(struct intel_encoder *encoder, intel_disable_hdmi(encoder, old_crtc_state, old_conn_state); } -static int intel_hdmi_source_max_tmds_clock(struct drm_i915_private *dev_priv) +static int intel_hdmi_source_max_tmds_clock(struct intel_encoder *encoder) { - if (IS_G4X(dev_priv)) - return 165000; - else if (IS_GEMINILAKE(dev_priv)) - return 594000; - else if (IS_HASWELL(dev_priv) || INTEL_INFO(dev_priv)->gen >= 8) - return 300000; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + const struct ddi_vbt_port_info *info = + &dev_priv->vbt.ddi_port_info[encoder->port]; + int max_tmds_clock; + + if (IS_GEMINILAKE(dev_priv)) + max_tmds_clock = 594000; + else if (INTEL_GEN(dev_priv) >= 8 || IS_HASWELL(dev_priv)) + max_tmds_clock = 300000; + else if (INTEL_GEN(dev_priv) >= 5) + max_tmds_clock = 225000; else - return 225000; + max_tmds_clock = 165000; + + if (info->max_tmds_clock) + max_tmds_clock = min(max_tmds_clock, info->max_tmds_clock); + + return max_tmds_clock; } static int hdmi_port_clock_limit(struct intel_hdmi *hdmi, bool respect_downstream_limits, bool force_dvi) { - struct drm_device *dev = intel_hdmi_to_dev(hdmi); - int max_tmds_clock = intel_hdmi_source_max_tmds_clock(to_i915(dev)); + struct intel_encoder *encoder = &hdmi_to_dig_port(hdmi)->base; + int max_tmds_clock = intel_hdmi_source_max_tmds_clock(encoder); if (respect_downstream_limits) { struct intel_connector *connector = hdmi->attached_connector; diff --git a/drivers/gpu/drm/i915/intel_vbt_defs.h b/drivers/gpu/drm/i915/intel_vbt_defs.h index 3c3c421e2e43c..e3d7745a91518 100644 --- a/drivers/gpu/drm/i915/intel_vbt_defs.h +++ b/drivers/gpu/drm/i915/intel_vbt_defs.h @@ -304,6 +304,10 @@ struct bdb_general_features { #define DVO_PORT_MIPIC 23 /* 171 */ #define DVO_PORT_MIPID 24 /* 171 */ +#define HDMI_MAX_DATA_RATE_PLATFORM 0 /* 204 */ +#define HDMI_MAX_DATA_RATE_297 1 /* 204 */ +#define HDMI_MAX_DATA_RATE_165 2 /* 204 */ + #define LEGACY_CHILD_DEVICE_CONFIG_SIZE 33 /* DDC Bus DDI Type 155+ */ -- GitLab From e1214b95ed83c0bb903fe368c9a408dffd80491d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:23 +0300 Subject: [PATCH 0070/1471] drm/i915: Populate output_types from .get_config() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Rather than having the caller of .get_config() set output_types based on encoder->type, let's just have .get_config() itself populate output_types. This way we are isolated from encoder->type, which won't be useable for this purpose anyway soon (at least for DDI encoders). Cc: Maarten Lankhorst Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-2-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_crt.c | 2 ++ drivers/gpu/drm/i915/intel_ddi.c | 11 +++++++++++ drivers/gpu/drm/i915/intel_display.c | 5 +---- drivers/gpu/drm/i915/intel_dp.c | 5 +++++ drivers/gpu/drm/i915/intel_dp_mst.c | 2 ++ drivers/gpu/drm/i915/intel_dsi.c | 2 ++ drivers/gpu/drm/i915/intel_dvo.c | 2 ++ drivers/gpu/drm/i915/intel_hdmi.c | 2 ++ drivers/gpu/drm/i915/intel_lvds.c | 2 ++ drivers/gpu/drm/i915/intel_sdvo.c | 2 ++ drivers/gpu/drm/i915/intel_tv.c | 2 ++ 11 files changed, 33 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c index 437339f5d0981..9c000ac612da5 100644 --- a/drivers/gpu/drm/i915/intel_crt.c +++ b/drivers/gpu/drm/i915/intel_crt.c @@ -119,6 +119,8 @@ static unsigned int intel_crt_get_flags(struct intel_encoder *encoder) static void intel_crt_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { + pipe_config->output_types |= BIT(INTEL_OUTPUT_ANALOG); + pipe_config->base.adjusted_mode.flags |= intel_crt_get_flags(encoder); pipe_config->base.adjusted_mode.crtc_clock = pipe_config->port_clock; diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 9c118e5305f71..7e0b1a02912a6 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2595,12 +2595,23 @@ void intel_ddi_get_config(struct intel_encoder *encoder, pipe_config->hdmi_high_tmds_clock_ratio = true; /* fall through */ case TRANS_DDI_MODE_SELECT_DVI: + pipe_config->output_types |= BIT(INTEL_OUTPUT_HDMI); pipe_config->lane_count = 4; break; case TRANS_DDI_MODE_SELECT_FDI: + pipe_config->output_types |= BIT(INTEL_OUTPUT_ANALOG); break; case TRANS_DDI_MODE_SELECT_DP_SST: + if (encoder->type == INTEL_OUTPUT_EDP) + pipe_config->output_types |= BIT(INTEL_OUTPUT_EDP); + else + pipe_config->output_types |= BIT(INTEL_OUTPUT_DP); + pipe_config->lane_count = + ((temp & DDI_PORT_WIDTH_MASK) >> DDI_PORT_WIDTH_SHIFT) + 1; + intel_dp_get_m_n(intel_crtc, pipe_config); + break; case TRANS_DDI_MODE_SELECT_DP_MST: + pipe_config->output_types |= BIT(INTEL_OUTPUT_DP_MST); pipe_config->lane_count = ((temp & DDI_PORT_WIDTH_MASK) >> DDI_PORT_WIDTH_SHIFT) + 1; intel_dp_get_m_n(intel_crtc, pipe_config); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index a369b35d044d6..8f769e9b9342d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11562,10 +11562,8 @@ verify_crtc_state(struct drm_crtc *crtc, "Encoder connected to wrong pipe %c\n", pipe_name(pipe)); - if (active) { - pipe_config->output_types |= 1 << encoder->type; + if (active) encoder->get_config(encoder, pipe_config); - } } intel_crtc_compute_pixel_rate(pipe_config); @@ -14960,7 +14958,6 @@ static void intel_modeset_readout_hw_state(struct drm_device *dev) crtc_state = to_intel_crtc_state(crtc->base.state); encoder->base.crtc = &crtc->base; - crtc_state->output_types |= 1 << encoder->type; encoder->get_config(encoder, crtc_state); } else { encoder->base.crtc = NULL; diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 8d25a019b772d..30688a5d680db 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2621,6 +2621,11 @@ static void intel_dp_get_config(struct intel_encoder *encoder, enum port port = dp_to_dig_port(intel_dp)->port; struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + if (encoder->type == INTEL_OUTPUT_EDP) + pipe_config->output_types |= BIT(INTEL_OUTPUT_EDP); + else + pipe_config->output_types |= BIT(INTEL_OUTPUT_DP); + tmp = I915_READ(intel_dp->output_reg); pipe_config->has_audio = tmp & DP_AUDIO_OUTPUT_ENABLE && port != PORT_A; diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index c34ffa959e900..210ad0580a667 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -291,6 +291,8 @@ static void intel_dp_mst_enc_get_config(struct intel_encoder *encoder, enum transcoder cpu_transcoder = pipe_config->cpu_transcoder; u32 temp, flags = 0; + pipe_config->output_types |= BIT(INTEL_OUTPUT_DP_MST); + pipe_config->has_audio = intel_ddi_is_audio_enabled(dev_priv, crtc); diff --git a/drivers/gpu/drm/i915/intel_dsi.c b/drivers/gpu/drm/i915/intel_dsi.c index 83f15848098a9..2bff7ab25bf39 100644 --- a/drivers/gpu/drm/i915/intel_dsi.c +++ b/drivers/gpu/drm/i915/intel_dsi.c @@ -1243,6 +1243,8 @@ static void intel_dsi_get_config(struct intel_encoder *encoder, u32 pclk; DRM_DEBUG_KMS("\n"); + pipe_config->output_types |= BIT(INTEL_OUTPUT_DSI); + if (IS_GEN9_LP(dev_priv)) bxt_dsi_get_pipe_config(encoder, pipe_config); diff --git a/drivers/gpu/drm/i915/intel_dvo.c b/drivers/gpu/drm/i915/intel_dvo.c index 53c9b763f4ce2..754baa00bea92 100644 --- a/drivers/gpu/drm/i915/intel_dvo.c +++ b/drivers/gpu/drm/i915/intel_dvo.c @@ -159,6 +159,8 @@ static void intel_dvo_get_config(struct intel_encoder *encoder, struct intel_dvo *intel_dvo = enc_to_dvo(encoder); u32 tmp, flags = 0; + pipe_config->output_types |= BIT(INTEL_OUTPUT_DVO); + tmp = I915_READ(intel_dvo->dev.dvo_reg); if (tmp & DVO_HSYNC_ACTIVE_HIGH) flags |= DRM_MODE_FLAG_PHSYNC; diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 38fe24565b4d1..0ec30b4a7dcee 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -957,6 +957,8 @@ static void intel_hdmi_get_config(struct intel_encoder *encoder, u32 tmp, flags = 0; int dotclock; + pipe_config->output_types |= BIT(INTEL_OUTPUT_HDMI); + tmp = I915_READ(intel_hdmi->hdmi_reg); if (tmp & SDVO_HSYNC_ACTIVE_HIGH) diff --git a/drivers/gpu/drm/i915/intel_lvds.c b/drivers/gpu/drm/i915/intel_lvds.c index 38572d65e46ea..ef80499113ee3 100644 --- a/drivers/gpu/drm/i915/intel_lvds.c +++ b/drivers/gpu/drm/i915/intel_lvds.c @@ -125,6 +125,8 @@ static void intel_lvds_get_config(struct intel_encoder *encoder, struct intel_lvds_encoder *lvds_encoder = to_lvds_encoder(&encoder->base); u32 tmp, flags = 0; + pipe_config->output_types |= BIT(INTEL_OUTPUT_LVDS); + tmp = I915_READ(lvds_encoder->reg); if (tmp & LVDS_HSYNC_POLARITY) flags |= DRM_MODE_FLAG_NHSYNC; diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c index 7437944b388f3..42ec2d1f7a613 100644 --- a/drivers/gpu/drm/i915/intel_sdvo.c +++ b/drivers/gpu/drm/i915/intel_sdvo.c @@ -1429,6 +1429,8 @@ static void intel_sdvo_get_config(struct intel_encoder *encoder, u8 val; bool ret; + pipe_config->output_types |= BIT(INTEL_OUTPUT_SDVO); + sdvox = I915_READ(intel_sdvo->sdvo_reg); ret = intel_sdvo_get_input_timing(intel_sdvo, &dtd); diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c index a79a7591b2cfd..b18609cebe035 100644 --- a/drivers/gpu/drm/i915/intel_tv.c +++ b/drivers/gpu/drm/i915/intel_tv.c @@ -868,6 +868,8 @@ static void intel_tv_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { + pipe_config->output_types |= BIT(INTEL_OUTPUT_TVOUT); + pipe_config->base.adjusted_mode.crtc_clock = pipe_config->port_clock; } -- GitLab From 7e732cacb1ae27b2eb6902cabd93e9da086c54f0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:24 +0300 Subject: [PATCH 0071/1471] drm/i915: Stop frobbing with DDI encoder->type MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Currently the DDI encoder->type will change at runtime depending on what kind of hotplugs we've processed. That's quite bad since we can't really trust that that current value of encoder->type actually matches the type of signal we're trying to drive through it. Let's eliminate that problem by declaring that non-eDP DDI port will always have the encoder type as INTEL_OUTPUT_DDI. This means the code can no longer try to distinguish DP vs. HDMI based on encoder->type. We'll leave eDP as INTEL_OUTPUT_EDP, since it'll never change and there's a bunch of code that relies on that value to identify eDP encoders. We'll introduce a new encoder .compute_output_type() hook. This allows us to compute the full output_types before any encoder .compute_config() hooks get called, thus those hooks can rely on output_types being correct, which is useful for cloning on oldr platforms. For now we'll just look at the connector type and pick the correct mode based on that. In the future the new hook could be used to implement dynamic switching between LS and PCON modes for LSPCON. v2: Fix BXT/GLK PPS explosion with DSI/MST encoders v3: Avoid the PPS warn on pure HDMI/DVI DDI encoders by checking dp.output_reg v4: Rebase v5: Populate output_types in .get_config() rather than in the caller v5: Split out populating output_types in .get_config() (Maarten) Cc: Maarten Lankhorst Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-3-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/i915_debugfs.c | 2 +- drivers/gpu/drm/i915/intel_ddi.c | 32 ++++++++++++++++++++------- drivers/gpu/drm/i915/intel_display.c | 11 ++++++--- drivers/gpu/drm/i915/intel_dp.c | 19 +++++----------- drivers/gpu/drm/i915/intel_drv.h | 7 ++++-- drivers/gpu/drm/i915/intel_hdmi.c | 10 ++------- drivers/gpu/drm/i915/intel_opregion.c | 2 +- 7 files changed, 47 insertions(+), 36 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index c1e5bba91bff8..39883cd915db1 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3051,7 +3051,7 @@ static void intel_connector_info(struct seq_file *m, break; case DRM_MODE_CONNECTOR_HDMIA: if (intel_encoder->type == INTEL_OUTPUT_HDMI || - intel_encoder->type == INTEL_OUTPUT_UNKNOWN) + intel_encoder->type == INTEL_OUTPUT_DDI) intel_hdmi_info(m, intel_connector); break; default: diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 7e0b1a02912a6..e5dd281781fea 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -497,10 +497,8 @@ enum port intel_ddi_get_encoder_port(struct intel_encoder *encoder) switch (encoder->type) { case INTEL_OUTPUT_DP_MST: return enc_to_mst(&encoder->base)->primary->port; - case INTEL_OUTPUT_DP: case INTEL_OUTPUT_EDP: - case INTEL_OUTPUT_HDMI: - case INTEL_OUTPUT_UNKNOWN: + case INTEL_OUTPUT_DDI: return enc_to_dig_port(&encoder->base)->port; case INTEL_OUTPUT_ANALOG: return PORT_E; @@ -1534,6 +1532,7 @@ void intel_ddi_set_vc_payload_alloc(const struct intel_crtc_state *crtc_state, struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; uint32_t temp; + temp = I915_READ(TRANS_DDI_FUNC_CTL(cpu_transcoder)); if (state == true) temp |= TRANS_DDI_DP_VC_PAYLOAD_ALLOC; @@ -2652,21 +2651,36 @@ void intel_ddi_get_config(struct intel_encoder *encoder, intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); } +static enum intel_output_type +intel_ddi_compute_output_type(struct intel_encoder *encoder, + struct intel_crtc_state *crtc_state, + struct drm_connector_state *conn_state) +{ + switch (conn_state->connector->connector_type) { + case DRM_MODE_CONNECTOR_HDMIA: + return INTEL_OUTPUT_HDMI; + case DRM_MODE_CONNECTOR_eDP: + return INTEL_OUTPUT_EDP; + case DRM_MODE_CONNECTOR_DisplayPort: + return INTEL_OUTPUT_DP; + default: + MISSING_CASE(conn_state->connector->connector_type); + return INTEL_OUTPUT_UNUSED; + } +} + static bool intel_ddi_compute_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config, struct drm_connector_state *conn_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - int type = encoder->type; int port = intel_ddi_get_encoder_port(encoder); int ret; - WARN(type == INTEL_OUTPUT_UNKNOWN, "compute_config() on unknown output!\n"); - if (port == PORT_A) pipe_config->cpu_transcoder = TRANSCODER_EDP; - if (type == INTEL_OUTPUT_HDMI) + if (intel_crtc_has_type(pipe_config, INTEL_OUTPUT_HDMI)) ret = intel_hdmi_compute_config(encoder, pipe_config, conn_state); else ret = intel_dp_compute_config(encoder, pipe_config, conn_state); @@ -2815,6 +2829,7 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) drm_encoder_init(&dev_priv->drm, encoder, &intel_ddi_funcs, DRM_MODE_ENCODER_TMDS, "DDI %c", port_name(port)); + intel_encoder->compute_output_type = intel_ddi_compute_output_type; intel_encoder->compute_config = intel_ddi_compute_config; intel_encoder->enable = intel_enable_ddi; if (IS_GEN9_LP(dev_priv)) @@ -2868,9 +2883,10 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) max_lanes = 4; } + intel_dig_port->dp.output_reg = INVALID_MMIO_REG; intel_dig_port->max_lanes = max_lanes; - intel_encoder->type = INTEL_OUTPUT_UNKNOWN; + intel_encoder->type = INTEL_OUTPUT_DDI; intel_encoder->power_domain = intel_port_to_power_domain(port); intel_encoder->port = port; intel_encoder->crtc_mask = (1 << 0) | (1 << 1) | (1 << 2); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 8f769e9b9342d..737de251d0f85 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10579,7 +10579,7 @@ static const char * const output_type_str[] = { OUTPUT_TYPE(DP), OUTPUT_TYPE(EDP), OUTPUT_TYPE(DSI), - OUTPUT_TYPE(UNKNOWN), + OUTPUT_TYPE(DDI), OUTPUT_TYPE(DP_MST), }; @@ -10750,7 +10750,7 @@ static bool check_digital_port_conflicts(struct drm_atomic_state *state) switch (encoder->type) { unsigned int port_mask; - case INTEL_OUTPUT_UNKNOWN: + case INTEL_OUTPUT_DDI: if (WARN_ON(!HAS_DDI(to_i915(dev)))) break; case INTEL_OUTPUT_DP: @@ -10883,7 +10883,12 @@ intel_modeset_pipe_config(struct drm_crtc *crtc, * Determine output_types before calling the .compute_config() * hooks so that the hooks can use this information safely. */ - pipe_config->output_types |= 1 << encoder->type; + if (encoder->compute_output_type) + pipe_config->output_types |= + BIT(encoder->compute_output_type(encoder, pipe_config, + connector_state)); + else + pipe_config->output_types |= BIT(encoder->type); } encoder_retry: diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 30688a5d680db..f0c49962ffbeb 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -758,11 +758,16 @@ void intel_power_sequencer_reset(struct drm_i915_private *dev_priv) struct intel_dp *intel_dp; if (encoder->type != INTEL_OUTPUT_DP && - encoder->type != INTEL_OUTPUT_EDP) + encoder->type != INTEL_OUTPUT_EDP && + encoder->type != INTEL_OUTPUT_DDI) continue; intel_dp = enc_to_intel_dp(&encoder->base); + /* Skip pure DVI/HDMI DDI encoders */ + if (!i915_mmio_reg_valid(intel_dp->output_reg)) + continue; + WARN_ON(intel_dp->active_pipe != INVALID_PIPE); if (encoder->type != INTEL_OUTPUT_EDP) @@ -4733,8 +4738,6 @@ intel_dp_long_pulse(struct intel_connector *intel_connector) { struct drm_connector *connector = &intel_connector->base; struct intel_dp *intel_dp = intel_attached_dp(connector); - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_encoder *intel_encoder = &intel_dig_port->base; struct drm_device *dev = connector->dev; enum drm_connector_status status; u8 sink_irq_vector = 0; @@ -4767,9 +4770,6 @@ intel_dp_long_pulse(struct intel_connector *intel_connector) goto out; } - if (intel_encoder->type != INTEL_OUTPUT_EDP) - intel_encoder->type = INTEL_OUTPUT_DP; - if (intel_dp->reset_link_params) { /* Initial max link lane count */ intel_dp->max_link_lane_count = intel_dp_max_common_lane_count(intel_dp); @@ -4886,9 +4886,6 @@ intel_dp_force(struct drm_connector *connector) intel_dp_set_edid(intel_dp); intel_display_power_put(dev_priv, intel_dp->aux_power_domain); - - if (intel_encoder->type != INTEL_OUTPUT_EDP) - intel_encoder->type = INTEL_OUTPUT_DP; } static int intel_dp_get_modes(struct drm_connector *connector) @@ -5107,10 +5104,6 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd) struct drm_i915_private *dev_priv = to_i915(dev); enum irqreturn ret = IRQ_NONE; - if (intel_dig_port->base.type != INTEL_OUTPUT_EDP && - intel_dig_port->base.type != INTEL_OUTPUT_HDMI) - intel_dig_port->base.type = INTEL_OUTPUT_DP; - if (long_hpd && intel_dig_port->base.type == INTEL_OUTPUT_EDP) { /* * vdd off can generate a long pulse on eDP which diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index df966427232cc..629ebc73bb094 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -173,7 +173,7 @@ enum intel_output_type { INTEL_OUTPUT_DP = 7, INTEL_OUTPUT_EDP = 8, INTEL_OUTPUT_DSI = 9, - INTEL_OUTPUT_UNKNOWN = 10, + INTEL_OUTPUT_DDI = 10, INTEL_OUTPUT_DP_MST = 11, }; @@ -216,6 +216,9 @@ struct intel_encoder { enum port port; unsigned int cloneable; void (*hot_plug)(struct intel_encoder *); + enum intel_output_type (*compute_output_type)(struct intel_encoder *, + struct intel_crtc_state *, + struct drm_connector_state *); bool (*compute_config)(struct intel_encoder *, struct intel_crtc_state *, struct drm_connector_state *); @@ -1152,7 +1155,7 @@ enc_to_dig_port(struct drm_encoder *encoder) struct intel_encoder *intel_encoder = to_intel_encoder(encoder); switch (intel_encoder->type) { - case INTEL_OUTPUT_UNKNOWN: + case INTEL_OUTPUT_DDI: WARN_ON(!HAS_DDI(to_i915(encoder->dev))); case INTEL_OUTPUT_DP: case INTEL_OUTPUT_EDP: diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 0ec30b4a7dcee..fe46abfeba26a 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1627,12 +1627,9 @@ intel_hdmi_detect(struct drm_connector *connector, bool force) intel_hdmi_unset_edid(connector); - if (intel_hdmi_set_edid(connector)) { - struct intel_hdmi *intel_hdmi = intel_attached_hdmi(connector); - - hdmi_to_dig_port(intel_hdmi)->base.type = INTEL_OUTPUT_HDMI; + if (intel_hdmi_set_edid(connector)) status = connector_status_connected; - } else + else status = connector_status_disconnected; intel_display_power_put(dev_priv, POWER_DOMAIN_GMBUS); @@ -1643,8 +1640,6 @@ intel_hdmi_detect(struct drm_connector *connector, bool force) static void intel_hdmi_force(struct drm_connector *connector) { - struct intel_hdmi *intel_hdmi = intel_attached_hdmi(connector); - DRM_DEBUG_KMS("[CONNECTOR:%d:%s]\n", connector->base.id, connector->name); @@ -1654,7 +1649,6 @@ intel_hdmi_force(struct drm_connector *connector) return; intel_hdmi_set_edid(connector); - hdmi_to_dig_port(intel_hdmi)->base.type = INTEL_OUTPUT_HDMI; } static int intel_hdmi_get_modes(struct drm_connector *connector) diff --git a/drivers/gpu/drm/i915/intel_opregion.c b/drivers/gpu/drm/i915/intel_opregion.c index 1d946240e55f4..39714be3eb6b1 100644 --- a/drivers/gpu/drm/i915/intel_opregion.c +++ b/drivers/gpu/drm/i915/intel_opregion.c @@ -383,7 +383,7 @@ int intel_opregion_notify_encoder(struct intel_encoder *intel_encoder, case INTEL_OUTPUT_ANALOG: type = DISPLAY_TYPE_CRT; break; - case INTEL_OUTPUT_UNKNOWN: + case INTEL_OUTPUT_DDI: case INTEL_OUTPUT_DP: case INTEL_OUTPUT_HDMI: case INTEL_OUTPUT_DP_MST: -- GitLab From 0fce04c8764bd0d6ef9b4488460a5a880afb1c73 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:25 +0300 Subject: [PATCH 0072/1471] drm/i915: Nuke intel_ddi_get_encoder_port() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit encoder->port works for FDI, and it also works for MST (regardless of whether we're dealing with the "fake" MST encoder, or mst->primary). So let's eliminate intel_ddi_get_encoder_port(). Signed-off-by: Ville Syrjälä Reviewed-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-4-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_ddi.c | 50 +++++++++------------------ drivers/gpu/drm/i915/intel_drv.h | 1 - drivers/gpu/drm/i915/intel_opregion.c | 2 +- 3 files changed, 18 insertions(+), 35 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index e5dd281781fea..aaaca906c97fa 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -492,22 +492,6 @@ static const struct cnl_ddi_buf_trans cnl_ddi_translations_edp_1_05V[] = { { 0x2, 0x7F, 0x3F, 0x00, 0x00 }, /* 400 400 0.0 */ }; -enum port intel_ddi_get_encoder_port(struct intel_encoder *encoder) -{ - switch (encoder->type) { - case INTEL_OUTPUT_DP_MST: - return enc_to_mst(&encoder->base)->primary->port; - case INTEL_OUTPUT_EDP: - case INTEL_OUTPUT_DDI: - return enc_to_dig_port(&encoder->base)->port; - case INTEL_OUTPUT_ANALOG: - return PORT_E; - default: - MISSING_CASE(encoder->type); - return PORT_A; - } -} - static const struct ddi_buf_trans * bdw_get_buf_trans_edp(struct drm_i915_private *dev_priv, int *n_entries) { @@ -815,7 +799,7 @@ static void intel_prepare_dp_ddi_buffers(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); u32 iboost_bit = 0; int i, n_entries; - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; const struct ddi_buf_trans *ddi_translations; if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_ANALOG)) @@ -852,7 +836,7 @@ static void intel_prepare_hdmi_ddi_buffers(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); u32 iboost_bit = 0; int n_entries; - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; const struct ddi_buf_trans *ddi_translations; ddi_translations = intel_ddi_get_buf_trans_hdmi(dev_priv, &n_entries); @@ -1468,7 +1452,7 @@ static void bxt_ddi_clock_get(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; enum intel_dpll_id pll_id = port; pipe_config->port_clock = bxt_calc_pll_link(dev_priv, pll_id); @@ -1548,7 +1532,7 @@ void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state) struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); enum pipe pipe = crtc->pipe; enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; uint32_t temp; /* Enable TRANS_DDI_FUNC_CTL for the pipe to work in HDMI mode */ @@ -1644,7 +1628,7 @@ bool intel_ddi_connector_get_hw_state(struct intel_connector *intel_connector) struct drm_i915_private *dev_priv = to_i915(dev); struct intel_encoder *encoder = intel_connector->encoder; int type = intel_connector->base.connector_type; - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; enum pipe pipe = 0; enum transcoder cpu_transcoder; uint32_t tmp; @@ -1703,7 +1687,7 @@ bool intel_ddi_get_hw_state(struct intel_encoder *encoder, { struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; u32 tmp; int i; bool ret; @@ -1788,7 +1772,7 @@ void intel_ddi_enable_pipe_clock(const struct intel_crtc_state *crtc_state) struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); struct intel_encoder *encoder = intel_ddi_get_crtc_encoder(crtc); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; enum transcoder cpu_transcoder = crtc_state->cpu_transcoder; if (cpu_transcoder != TRANSCODER_EDP) @@ -1927,8 +1911,8 @@ static void cnl_ddi_vswing_program(struct intel_encoder *encoder, int level, enum intel_output_type type) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); const struct cnl_ddi_buf_trans *ddi_translations; + enum port port = encoder->port; int n_entries, ln; u32 val; @@ -1991,7 +1975,7 @@ static void cnl_ddi_vswing_sequence(struct intel_encoder *encoder, int level, enum intel_output_type type) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; int width, rate, ln; u32 val; @@ -2110,7 +2094,7 @@ static void intel_ddi_clk_select(struct intel_encoder *encoder, const struct intel_shared_dpll *pll) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; uint32_t val; if (WARN_ON(!pll)) @@ -2149,7 +2133,7 @@ static void intel_ddi_clk_select(struct intel_encoder *encoder, static void intel_ddi_clk_disable(struct intel_encoder *encoder) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; if (IS_CANNONLAKE(dev_priv)) I915_WRITE(DPCLKA_CFGCR0, I915_READ(DPCLKA_CFGCR0) | @@ -2167,7 +2151,7 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder, { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; struct intel_digital_port *dig_port = enc_to_dig_port(&encoder->base); bool is_mst = intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST); int level = intel_ddi_dp_level(intel_dp); @@ -2205,7 +2189,7 @@ static void intel_ddi_pre_enable_hdmi(struct intel_encoder *encoder, struct intel_digital_port *intel_dig_port = enc_to_dig_port(&encoder->base); struct intel_hdmi *intel_hdmi = &intel_dig_port->hdmi; struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; int level = intel_ddi_hdmi_level(dev_priv, port); struct intel_digital_port *dig_port = enc_to_dig_port(&encoder->base); @@ -2250,7 +2234,7 @@ static void intel_ddi_pre_enable(struct intel_encoder *encoder, static void intel_disable_ddi_buf(struct intel_encoder *encoder) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; bool wait = false; u32 val; @@ -2379,7 +2363,7 @@ static void intel_enable_ddi_dp(struct intel_encoder *encoder, { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; if (port == PORT_A && INTEL_GEN(dev_priv) < 9) intel_dp_stop_link_train(intel_dp); @@ -2398,7 +2382,7 @@ static void intel_enable_ddi_hdmi(struct intel_encoder *encoder, { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_digital_port *dig_port = enc_to_dig_port(&encoder->base); - enum port port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; intel_hdmi_handle_sink_scrambling(encoder, conn_state->connector, @@ -2674,7 +2658,7 @@ static bool intel_ddi_compute_config(struct intel_encoder *encoder, struct drm_connector_state *conn_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - int port = intel_ddi_get_encoder_port(encoder); + enum port port = encoder->port; int ret; if (port == PORT_A) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 629ebc73bb094..3828d9bee7f57 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1279,7 +1279,6 @@ void intel_ddi_fdi_post_disable(struct intel_encoder *intel_encoder, void hsw_fdi_link_train(struct intel_crtc *crtc, const struct intel_crtc_state *crtc_state); void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port); -enum port intel_ddi_get_encoder_port(struct intel_encoder *intel_encoder); bool intel_ddi_get_hw_state(struct intel_encoder *encoder, enum pipe *pipe); void intel_ddi_enable_transcoder_func(const struct intel_crtc_state *crtc_state); void intel_ddi_disable_transcoder_func(struct drm_i915_private *dev_priv, diff --git a/drivers/gpu/drm/i915/intel_opregion.c b/drivers/gpu/drm/i915/intel_opregion.c index 39714be3eb6b1..fc65f5e451dd0 100644 --- a/drivers/gpu/drm/i915/intel_opregion.c +++ b/drivers/gpu/drm/i915/intel_opregion.c @@ -367,7 +367,7 @@ int intel_opregion_notify_encoder(struct intel_encoder *intel_encoder, if (intel_encoder->type == INTEL_OUTPUT_DSI) port = 0; else - port = intel_ddi_get_encoder_port(intel_encoder); + port = intel_encoder->port; if (port == PORT_E) { port = 0; -- GitLab From bb911536f07e5ed9147e3acf55a2cd72dffff70d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:26 +0300 Subject: [PATCH 0073/1471] drm/i915: Eliminate pll->state usage from bxt_calc_pll_link() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We should be using the DPLL hw state we got from the current crtc state to determine the corresponding port clock frequency rather than getting it via the current state programmed into the DPLL. v2: Rebase due to intel_dpll_id changes Signed-off-by: Ville Syrjälä Reviewed-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-5-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_ddi.c | 17 +++++------------ 1 file changed, 5 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index aaaca906c97fa..8183304c7d342 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1423,19 +1423,16 @@ static void hsw_ddi_clock_get(struct intel_encoder *encoder, ddi_dotclock_get(pipe_config); } -static int bxt_calc_pll_link(struct drm_i915_private *dev_priv, - enum intel_dpll_id pll_id) +static int bxt_calc_pll_link(struct intel_crtc_state *crtc_state) { - struct intel_shared_dpll *pll; struct intel_dpll_hw_state *state; struct dpll clock; /* For DDI ports we always use a shared PLL. */ - if (WARN_ON(pll_id == DPLL_ID_PRIVATE)) + if (WARN_ON(!crtc_state->shared_dpll)) return 0; - pll = &dev_priv->shared_dplls[pll_id]; - state = &pll->state.hw_state; + state = &crtc_state->dpll_hw_state; clock.m1 = 2; clock.m2 = (state->pll0 & PORT_PLL_M2_MASK) << 22; @@ -1449,13 +1446,9 @@ static int bxt_calc_pll_link(struct drm_i915_private *dev_priv, } static void bxt_ddi_clock_get(struct intel_encoder *encoder, - struct intel_crtc_state *pipe_config) + struct intel_crtc_state *pipe_config) { - struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = encoder->port; - enum intel_dpll_id pll_id = port; - - pipe_config->port_clock = bxt_calc_pll_link(dev_priv, pll_id); + pipe_config->port_clock = bxt_calc_pll_link(pipe_config); ddi_dotclock_get(pipe_config); } -- GitLab From 1939ba51fd058b869c06aa084c0b9ae75bc721c6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:27 +0300 Subject: [PATCH 0074/1471] drm/i915: Pass a crtc state to ddi post_disable from MST code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Pass an old crtc state to intel_ddi_post_disable() from the MST code. Note that this crtc state won't necessaitly match the one that was passed to intel_ddi_pre_enable() if the first stream to be enabled isn't the last stream to be disabled. But this is fine since the states should be identical in every important way. This does mean people frobbing the DDI pre_enable/post_disable hooks have to pay attention in what parts of the state they consult. The alternative would be to inline the relevant code into the MST code. That is actually what we used to do for pre_enable before commit e081c8463ac9 ("drm/i915: Remove duplicate DDI enabling logic from MST path"). For post_disable we've always called the DDI hook. v2: Pimp up the comments explaining the MST issues Cc: Maarten Lankhorst Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-6-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_ddi.c | 37 ++++++++++++++++++++--------- drivers/gpu/drm/i915/intel_dp_mst.c | 6 ++--- 2 files changed, 29 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 8183304c7d342..7a6806d77678d 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2214,6 +2214,19 @@ static void intel_ddi_pre_enable(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); enum pipe pipe = crtc->pipe; + /* + * When called from DP MST code: + * - conn_state will be NULL + * - encoder will be the main encoder (ie. mst->primary) + * - the main connector associated with this port + * won't be active or linked to a crtc + * - crtc_state will be the state of the first stream to + * be activated on this port, and it may not be the same + * stream that will be deactivated last, but each stream + * should have a state that is identical when it comes to + * the DP link parameteres + */ + WARN_ON(crtc_state->has_pch_encoder); intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, true); @@ -2254,12 +2267,7 @@ static void intel_ddi_post_disable_dp(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_digital_port *dig_port = enc_to_dig_port(&encoder->base); struct intel_dp *intel_dp = &dig_port->dp; - /* - * old_crtc_state and old_conn_state are NULL when called from - * DP_MST. The main connector associated with this port is never - * bound to a crtc for MST. - */ - bool is_mst = !old_crtc_state; + bool is_mst = intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_DP_MST); /* * Power down sink before disabling the port, otherwise we end @@ -2303,12 +2311,19 @@ static void intel_ddi_post_disable(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { /* - * old_crtc_state and old_conn_state are NULL when called from - * DP_MST. The main connector associated with this port is never - * bound to a crtc for MST. + * When called from DP MST code: + * - old_conn_state will be NULL + * - encoder will be the main encoder (ie. mst->primary) + * - the main connector associated with this port + * won't be active or linked to a crtc + * - old_crtc_state will be the state of the last stream to + * be deactivated on this port, and it may not be the same + * stream that was activated last, but each stream + * should have a state that is identical when it comes to + * the DP link parameteres */ - if (old_crtc_state && - intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_HDMI)) + + if (intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_HDMI)) intel_ddi_post_disable_hdmi(encoder, old_crtc_state, old_conn_state); else diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 210ad0580a667..d523302e5081a 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -179,10 +179,10 @@ static void intel_mst_post_disable_dp(struct intel_encoder *encoder, intel_dp->active_mst_links--; intel_mst->connector = NULL; - if (intel_dp->active_mst_links == 0) { + if (intel_dp->active_mst_links == 0) intel_dig_port->base.post_disable(&intel_dig_port->base, - NULL, NULL); - } + old_crtc_state, NULL); + DRM_DEBUG_KMS("active links %d\n", intel_dp->active_mst_links); } -- GitLab From 35686a44e4b87f45ebd426a7e2b074a54486844f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 27 Oct 2017 22:31:28 +0300 Subject: [PATCH 0075/1471] drm/i915: Use intel_ddi_get_config() for MST MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Eliminate the partially duplicated DDI readout code from MST, and instead just call intel_ddi_get_config(). As a nice bonus we get more cross checking as intel_ddi_get_config() will populate output_types based on the actual mode of the DDI port. Additonally intel_ddi_get_config() must be changed to get the crtc from the passed in crtc state rather than from the encoder->crtc link. encoder->crtc really shouldn't be used anyway. v2: Rebased on BXT MST latency_optim fix Make intel_ddi_clock_get() static Cc: Maarten Lankhorst Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171027193128.14483-7-ville.syrjala@linux.intel.com Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_ddi.c | 6 ++-- drivers/gpu/drm/i915/intel_dp_mst.c | 50 +---------------------------- drivers/gpu/drm/i915/intel_drv.h | 2 -- 3 files changed, 4 insertions(+), 54 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 7a6806d77678d..eb8f4e3bb85f8 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1453,8 +1453,8 @@ static void bxt_ddi_clock_get(struct intel_encoder *encoder, ddi_dotclock_get(pipe_config); } -void intel_ddi_clock_get(struct intel_encoder *encoder, - struct intel_crtc_state *pipe_config) +static void intel_ddi_clock_get(struct intel_encoder *encoder, + struct intel_crtc_state *pipe_config) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); @@ -2533,7 +2533,7 @@ void intel_ddi_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(pipe_config->base.crtc); enum transcoder cpu_transcoder = pipe_config->cpu_transcoder; struct intel_digital_port *intel_dig_port; u32 temp, flags = 0; diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index d523302e5081a..6f11bb35f66f5 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -286,56 +286,8 @@ static void intel_dp_mst_enc_get_config(struct intel_encoder *encoder, { struct intel_dp_mst_encoder *intel_mst = enc_to_mst(&encoder->base); struct intel_digital_port *intel_dig_port = intel_mst->primary; - struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); - struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum transcoder cpu_transcoder = pipe_config->cpu_transcoder; - u32 temp, flags = 0; - - pipe_config->output_types |= BIT(INTEL_OUTPUT_DP_MST); - - pipe_config->has_audio = - intel_ddi_is_audio_enabled(dev_priv, crtc); - - temp = I915_READ(TRANS_DDI_FUNC_CTL(cpu_transcoder)); - if (temp & TRANS_DDI_PHSYNC) - flags |= DRM_MODE_FLAG_PHSYNC; - else - flags |= DRM_MODE_FLAG_NHSYNC; - if (temp & TRANS_DDI_PVSYNC) - flags |= DRM_MODE_FLAG_PVSYNC; - else - flags |= DRM_MODE_FLAG_NVSYNC; - - switch (temp & TRANS_DDI_BPC_MASK) { - case TRANS_DDI_BPC_6: - pipe_config->pipe_bpp = 18; - break; - case TRANS_DDI_BPC_8: - pipe_config->pipe_bpp = 24; - break; - case TRANS_DDI_BPC_10: - pipe_config->pipe_bpp = 30; - break; - case TRANS_DDI_BPC_12: - pipe_config->pipe_bpp = 36; - break; - default: - break; - } - pipe_config->base.adjusted_mode.flags |= flags; - - pipe_config->lane_count = - ((temp & DDI_PORT_WIDTH_MASK) >> DDI_PORT_WIDTH_SHIFT) + 1; - - intel_dp_get_m_n(crtc, pipe_config); - - intel_ddi_clock_get(&intel_dig_port->base, pipe_config); - if (IS_GEN9_LP(dev_priv)) - pipe_config->lane_lat_optim_mask = - bxt_ddi_phy_get_lane_lat_optim_mask(encoder); - - intel_ddi_compute_min_voltage_level(dev_priv, pipe_config); + intel_ddi_get_config(&intel_dig_port->base, pipe_config); } static int intel_dp_mst_get_ddc_modes(struct drm_connector *connector) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3828d9bee7f57..4498d743cebc3 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1295,8 +1295,6 @@ bool intel_ddi_is_audio_enabled(struct drm_i915_private *dev_priv, void intel_ddi_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config); -void intel_ddi_clock_get(struct intel_encoder *encoder, - struct intel_crtc_state *pipe_config); void intel_ddi_set_vc_payload_alloc(const struct intel_crtc_state *crtc_state, bool state); void intel_ddi_compute_min_voltage_level(struct drm_i915_private *dev_priv, -- GitLab From 79a5ad2fdb3cb48e9c4e9d0d7688f0401cdb85d5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 27 Oct 2017 12:06:02 +0100 Subject: [PATCH 0076/1471] drm: Enable pr_debug() for drm_printer pr_debug() is conditionally compiled and requires either dynamic-debugging to be enabled or for the code to opt-in using #define DEBUG. Since drm_print provides a central debugging facility using pr_debug(), make sure it will always produce output. Signed-off-by: Chris Wilson Cc: Rob Clark Cc: Daniel Vetter Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/20171027110602.31519-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/drm_print.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/drm_print.c b/drivers/gpu/drm/drm_print.c index 0b3bf476dc4b5..82ff327eb2df8 100644 --- a/drivers/gpu/drm/drm_print.c +++ b/drivers/gpu/drm/drm_print.c @@ -23,6 +23,8 @@ * Rob Clark */ +#define DEBUG /* for pr_debug() */ + #include #include #include -- GitLab From 4ab09d0ec3c81aceaebf8967efe111ba69a406d8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 30 Oct 2017 17:29:27 +0000 Subject: [PATCH 0077/1471] drm/i915: Replace "cc-option -Wno-foo" with "cc-disable-warning foo" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To quote kbuild/makefiles.txt: cc-disable-warning checks if gcc supports a given warning and returns the commandline switch to disable it. This special function is needed, because gcc 4.4 and later accept any unknown -Wno-* option and only warn about it if there is another warning in the source file. This is exactly what we were trying to achieve with cc-option -Wno-foo and failed miserably. Reported-by: kbuild-all@01.org Fixes: 39bf4de89ff7 ("drm/i915: Add -Wall -Wextra to our build, set warnings to full") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171030172927.18158-1-chris@chris-wilson.co.uk Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/Makefile | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 7750be8e27a6b..1bbc5440db40d 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -12,15 +12,15 @@ # will most likely get a sudden build breakage... Hopefully we will fix # new warnings before CI updates! subdir-ccflags-y := -Wall -Wextra -subdir-ccflags-y += $(call cc-option,-Wno-unused-parameter,) -subdir-ccflags-y += $(call cc-option,-Wno-type-limits,) -subdir-ccflags-y += $(call cc-option,-Wno-missing-field-initializers,) -subdir-ccflags-y += $(call cc-option,-Wno-implicit-fallthrough,) +subdir-ccflags-y += $(call cc-disable-warning, unused-parameter) +subdir-ccflags-y += $(call cc-disable-warning, type-limits) +subdir-ccflags-y += $(call cc-disable-warning, missing-field-initializers) +subdir-ccflags-y += $(call cc-disable-warning, implicit-fallthrough) subdir-ccflags-$(CONFIG_DRM_I915_WERROR) += -Werror # Fine grained warnings disable -CFLAGS_i915_pci.o = $(call cc-option,-Wno-override-init,) -CFLAGS_intel_fbdev.o = $(call cc-option,-Wno-override-init,) +CFLAGS_i915_pci.o = $(call cc-disable-warning, override-init) +CFLAGS_intel_fbdev.o = $(call cc-disable-warning, override-init) subdir-ccflags-y += \ $(call as-instr,movntdqa (%eax)$(comma)%xmm0,-DCONFIG_AS_MOVNTDQA) -- GitLab From 33b54ea1109721dcd07d3f7ee753c07482021eed Mon Sep 17 00:00:00 2001 From: Kees Cook Date: Tue, 24 Oct 2017 08:16:48 -0700 Subject: [PATCH 0078/1471] drm/vc4: Convert timers to use timer_setup() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new timer_setup() and from_timer() to pass the timer pointer explicitly. Cc: Eric Anholt Cc: David Airlie Cc: dri-devel@lists.freedesktop.org Signed-off-by: Kees Cook Signed-off-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/20171024151648.GA104538@beast Reviewed-by: Eric Anholt --- drivers/gpu/drm/vc4/vc4_bo.c | 9 +++------ drivers/gpu/drm/vc4/vc4_gem.c | 10 ++++------ 2 files changed, 7 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/vc4/vc4_bo.c b/drivers/gpu/drm/vc4/vc4_bo.c index 01a53ba304f89..3c3d112369109 100644 --- a/drivers/gpu/drm/vc4/vc4_bo.c +++ b/drivers/gpu/drm/vc4/vc4_bo.c @@ -674,10 +674,9 @@ void vc4_bo_dec_usecnt(struct vc4_bo *bo) mutex_unlock(&bo->madv_lock); } -static void vc4_bo_cache_time_timer(unsigned long data) +static void vc4_bo_cache_time_timer(struct timer_list *t) { - struct drm_device *dev = (struct drm_device *)data; - struct vc4_dev *vc4 = to_vc4_dev(dev); + struct vc4_dev *vc4 = from_timer(vc4, t, bo_cache.time_timer); schedule_work(&vc4->bo_cache.time_work); } @@ -1039,9 +1038,7 @@ int vc4_bo_cache_init(struct drm_device *dev) INIT_LIST_HEAD(&vc4->bo_cache.time_list); INIT_WORK(&vc4->bo_cache.time_work, vc4_bo_cache_time_work); - setup_timer(&vc4->bo_cache.time_timer, - vc4_bo_cache_time_timer, - (unsigned long)dev); + timer_setup(&vc4->bo_cache.time_timer, vc4_bo_cache_time_timer, 0); return 0; } diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c index e00ac2f3a264b..6c32c89a83a96 100644 --- a/drivers/gpu/drm/vc4/vc4_gem.c +++ b/drivers/gpu/drm/vc4/vc4_gem.c @@ -312,10 +312,10 @@ vc4_reset_work(struct work_struct *work) } static void -vc4_hangcheck_elapsed(unsigned long data) +vc4_hangcheck_elapsed(struct timer_list *t) { - struct drm_device *dev = (struct drm_device *)data; - struct vc4_dev *vc4 = to_vc4_dev(dev); + struct vc4_dev *vc4 = from_timer(vc4, t, hangcheck.timer); + struct drm_device *dev = vc4->dev; uint32_t ct0ca, ct1ca; unsigned long irqflags; struct vc4_exec_info *bin_exec, *render_exec; @@ -1154,9 +1154,7 @@ vc4_gem_init(struct drm_device *dev) spin_lock_init(&vc4->job_lock); INIT_WORK(&vc4->hangcheck.reset_work, vc4_reset_work); - setup_timer(&vc4->hangcheck.timer, - vc4_hangcheck_elapsed, - (unsigned long)dev); + timer_setup(&vc4->hangcheck.timer, vc4_hangcheck_elapsed, 0); INIT_WORK(&vc4->job_done_work, vc4_job_done_work); -- GitLab From 85b4587f8e94143bafb8b6a4003a5187b9a8753d Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Fri, 20 Oct 2017 03:28:45 +0300 Subject: [PATCH 0079/1471] drm/panel: rpi-touchscreen: propagate errors in rpi_touchscreen_i2c_read() There is one caller which checks whether rpi_touchscreen_i2c_read() returns negative error codes. Currently it can't because negative error codes are truncated to u8, but that's easy to fix if we change the type to int. Fixes: 2f733d6194bd ("drm/panel: Add support for the Raspberry Pi 7" Touchscreen.") Signed-off-by: Dan Carpenter Signed-off-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/20171020002845.kar2wg7gqxg7tzqi@mwanda Reviewed-by: Eric Anholt --- drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c b/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c index 890fd6ff397c4..d964d454e4aea 100644 --- a/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c +++ b/drivers/gpu/drm/panel/panel-raspberrypi-touchscreen.c @@ -221,7 +221,7 @@ static struct rpi_touchscreen *panel_to_ts(struct drm_panel *panel) return container_of(panel, struct rpi_touchscreen, base); } -static u8 rpi_touchscreen_i2c_read(struct rpi_touchscreen *ts, u8 reg) +static int rpi_touchscreen_i2c_read(struct rpi_touchscreen *ts, u8 reg) { return i2c_smbus_read_byte_data(ts->i2c, reg); } -- GitLab From 1de86387bc06db2259a0ff97145d5642f8d00fa8 Mon Sep 17 00:00:00 2001 From: Kees Cook Date: Tue, 24 Oct 2017 08:16:09 -0700 Subject: [PATCH 0080/1471] drm: gma500: Convert timers to use timer_setup() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new timer_setup() and from_timer() to pass the timer pointer explicitly. Cc: Patrik Jakobsson Cc: David Airlie Cc: dri-devel@lists.freedesktop.org Signed-off-by: Kees Cook Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171024151609.GA104501@beast --- drivers/gpu/drm/gma500/psb_lid.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/gma500/psb_lid.c b/drivers/gpu/drm/gma500/psb_lid.c index 1d2ebb5e530f9..be6dda58fcae3 100644 --- a/drivers/gpu/drm/gma500/psb_lid.c +++ b/drivers/gpu/drm/gma500/psb_lid.c @@ -23,9 +23,9 @@ #include "psb_intel_reg.h" #include -static void psb_lid_timer_func(unsigned long data) +static void psb_lid_timer_func(struct timer_list *t) { - struct drm_psb_private * dev_priv = (struct drm_psb_private *)data; + struct drm_psb_private *dev_priv = from_timer(dev_priv, t, lid_timer); struct drm_device *dev = (struct drm_device *)dev_priv->dev; struct timer_list *lid_timer = &dev_priv->lid_timer; unsigned long irq_flags; @@ -77,10 +77,8 @@ void psb_lid_timer_init(struct drm_psb_private *dev_priv) spin_lock_init(&dev_priv->lid_lock); spin_lock_irqsave(&dev_priv->lid_lock, irq_flags); - init_timer(lid_timer); + timer_setup(lid_timer, psb_lid_timer_func, 0); - lid_timer->data = (unsigned long)dev_priv; - lid_timer->function = psb_lid_timer_func; lid_timer->expires = jiffies + PSB_LID_DELAY; add_timer(lid_timer); -- GitLab From 9f446781616ccc317ba081d0f367f37ee94ed42f Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 30 Oct 2017 14:15:36 +0100 Subject: [PATCH 0081/1471] drm/TODO: Clarify DRM_DEV_* conversion Not everyone agrees this is the best thing, so make it really clear that maintainers need to be asked first, then the conversion. We've had a few newbies that did this the other way round, got their patches rejected, which isn't the best newbie experience. Cc: Sean Paul Acked-by: Sean Paul Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171030131536.11654-1-daniel.vetter@ffwll.ch --- Documentation/gpu/todo.rst | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst index 96f8ec7dbe4ed..e9840d693a863 100644 --- a/Documentation/gpu/todo.rst +++ b/Documentation/gpu/todo.rst @@ -179,6 +179,10 @@ don't do this, drivers used dev_info/warn/err to make this differentiation. We now have DRM_DEV_* variants of the drm print macros, so we can start to convert those drivers back to using drm-formwatted specific log messages. +Before you start this conversion please contact the relevant maintainers to make +sure your work will be merged - not everyone agrees that the DRM dmesg macros +are better. + Contact: Sean Paul, Maintainer of the driver you plan to convert Core refactorings -- GitLab From 37569925f4a71d45a90892418b18c1faea0ecf6b Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 30 Oct 2017 14:10:28 +0100 Subject: [PATCH 0082/1471] MAINTAINERS: Gustavo as drm-misc co-maintainer Gustavo volunteered to become a drm-misc co-maintainer, he'll take care of 4.16 to get started. Cc: Gustavo Padovan Cc: Daniel Vetter Cc: Sean Paul Cc: Jani Nikula Cc: Dave Airlie Acked-by: Sean Paul Acked-by: Jani Nikula Acked-by: Dave Airlie Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171030131028.11285-1-daniel.vetter@ffwll.ch --- MAINTAINERS | 1 + 1 file changed, 1 insertion(+) diff --git a/MAINTAINERS b/MAINTAINERS index fe121ea1ab99e..0d77f22fd38b8 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4498,6 +4498,7 @@ F: include/linux/vga* DRM DRIVERS AND MISC GPU PATCHES M: Daniel Vetter M: Jani Nikula +M: Gustavo Padovan M: Sean Paul W: https://01.org/linuxgraphics/gfx-docs/maintainer-tools/drm-misc.html S: Maintained -- GitLab From c777990fb45b574b967d982986dd2dea8bb85643 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 30 Oct 2017 16:39:37 +0100 Subject: [PATCH 0083/1471] drm/fb-helper: Handle function NULL argument MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Make functions tolerate that the drm_fb_helper argument is NULL. This is useful for drivers that continue probing when fbdev emulation fails and not having to do this check themselves. Update docs for functions that already handles this. Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171030153951.56269-2-noralf@tronnes.org --- drivers/gpu/drm/drm_fb_helper.c | 28 +++++++++++++++++----------- 1 file changed, 17 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 116d1f1337c7e..954cdd48de92e 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -150,6 +150,9 @@ int drm_fb_helper_add_one_connector(struct drm_fb_helper *fb_helper, { int err; + if (!fb_helper) + return 0; + mutex_lock(&fb_helper->lock); err = __drm_fb_helper_add_one_connector(fb_helper, connector); mutex_unlock(&fb_helper->lock); @@ -161,7 +164,7 @@ EXPORT_SYMBOL(drm_fb_helper_add_one_connector); /** * drm_fb_helper_single_add_all_connectors() - add all connectors to fbdev * emulation helper - * @fb_helper: fbdev initialized with drm_fb_helper_init + * @fb_helper: fbdev initialized with drm_fb_helper_init, can be NULL * * This functions adds all the available connectors for use with the given * fb_helper. This is a separate step to allow drivers to freely assign @@ -179,7 +182,7 @@ int drm_fb_helper_single_add_all_connectors(struct drm_fb_helper *fb_helper) struct drm_connector_list_iter conn_iter; int i, ret = 0; - if (!drm_fbdev_emulation) + if (!drm_fbdev_emulation || !fb_helper) return 0; mutex_lock(&fb_helper->lock); @@ -245,6 +248,9 @@ int drm_fb_helper_remove_one_connector(struct drm_fb_helper *fb_helper, { int err; + if (!fb_helper) + return 0; + mutex_lock(&fb_helper->lock); err = __drm_fb_helper_remove_one_connector(fb_helper, connector); mutex_unlock(&fb_helper->lock); @@ -484,7 +490,7 @@ static int restore_fbdev_mode(struct drm_fb_helper *fb_helper) /** * drm_fb_helper_restore_fbdev_mode_unlocked - restore fbdev configuration - * @fb_helper: fbcon to restore + * @fb_helper: driver-allocated fbdev helper, can be NULL * * This should be called from driver's drm &drm_driver.lastclose callback * when implementing an fbcon on top of kms using this helper. This ensures that @@ -498,7 +504,7 @@ int drm_fb_helper_restore_fbdev_mode_unlocked(struct drm_fb_helper *fb_helper) bool do_delayed; int ret; - if (!drm_fbdev_emulation) + if (!drm_fbdev_emulation || !fb_helper) return -ENODEV; if (READ_ONCE(fb_helper->deferred_setup)) @@ -883,7 +889,7 @@ EXPORT_SYMBOL(drm_fb_helper_alloc_fbi); /** * drm_fb_helper_unregister_fbi - unregister fb_info framebuffer device - * @fb_helper: driver-allocated fbdev helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * * A wrapper around unregister_framebuffer, to release the fb_info * framebuffer device. This must be called before releasing all resources for @@ -898,7 +904,7 @@ EXPORT_SYMBOL(drm_fb_helper_unregister_fbi); /** * drm_fb_helper_fini - finialize a &struct drm_fb_helper - * @fb_helper: driver-allocated fbdev helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * * This cleans up all remaining resources associated with @fb_helper. Must be * called after drm_fb_helper_unlink_fbi() was called. @@ -937,7 +943,7 @@ EXPORT_SYMBOL(drm_fb_helper_fini); /** * drm_fb_helper_unlink_fbi - wrapper around unlink_framebuffer - * @fb_helper: driver-allocated fbdev helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * * A wrapper around unlink_framebuffer implemented by fbdev core */ @@ -1138,7 +1144,7 @@ EXPORT_SYMBOL(drm_fb_helper_cfb_imageblit); /** * drm_fb_helper_set_suspend - wrapper around fb_set_suspend - * @fb_helper: driver-allocated fbdev helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * @suspend: whether to suspend or resume * * A wrapper around fb_set_suspend implemented by fbdev core. @@ -1155,7 +1161,7 @@ EXPORT_SYMBOL(drm_fb_helper_set_suspend); /** * drm_fb_helper_set_suspend_unlocked - wrapper around fb_set_suspend that also * takes the console lock - * @fb_helper: driver-allocated fbdev helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * @suspend: whether to suspend or resume * * A wrapper around fb_set_suspend() that takes the console lock. If the lock @@ -2568,7 +2574,7 @@ EXPORT_SYMBOL(drm_fb_helper_initial_config); /** * drm_fb_helper_hotplug_event - respond to a hotplug notification by * probing all the outputs attached to the fb - * @fb_helper: the drm_fb_helper + * @fb_helper: driver-allocated fbdev helper, can be NULL * * Scan the connectors attached to the fb_helper and try to put together a * setup after notification of a change in output configuration. @@ -2590,7 +2596,7 @@ int drm_fb_helper_hotplug_event(struct drm_fb_helper *fb_helper) { int err = 0; - if (!drm_fbdev_emulation) + if (!drm_fbdev_emulation || !fb_helper) return 0; mutex_lock(&fb_helper->lock); -- GitLab From 29ad20b22c8f3ab35e91c2f68b4c7956cee30fd0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 30 Oct 2017 16:39:38 +0100 Subject: [PATCH 0084/1471] drm: Add drm_device->fb_helper pointer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_fb_helper is *the* way of doing fbdev emulation so add a pointer to struct drm_device. This makes it possible to add callback helpers for .last_close and .output_poll_changed further reducing fbdev emulation footprint in drivers. The pointer is set by drm_fb_helper_init() and cleared by drm_fb_helper_fini(). Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171030153951.56269-3-noralf@tronnes.org --- drivers/gpu/drm/drm_fb_helper.c | 13 +++++++++++-- include/drm/drm_device.h | 9 +++++++++ 2 files changed, 20 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 954cdd48de92e..19aaca9fea06e 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -799,8 +799,10 @@ int drm_fb_helper_init(struct drm_device *dev, struct drm_mode_config *config = &dev->mode_config; int i; - if (!drm_fbdev_emulation) + if (!drm_fbdev_emulation) { + dev->fb_helper = fb_helper; return 0; + } if (!max_conn_count) return -EINVAL; @@ -835,6 +837,8 @@ int drm_fb_helper_init(struct drm_device *dev, i++; } + dev->fb_helper = fb_helper; + return 0; out_free: drm_fb_helper_crtc_free(fb_helper); @@ -913,7 +917,12 @@ void drm_fb_helper_fini(struct drm_fb_helper *fb_helper) { struct fb_info *info; - if (!drm_fbdev_emulation || !fb_helper) + if (!fb_helper) + return; + + fb_helper->dev->fb_helper = NULL; + + if (!drm_fbdev_emulation) return; cancel_work_sync(&fb_helper->resume_work); diff --git a/include/drm/drm_device.h b/include/drm/drm_device.h index e21af87a2f3c6..7c4fa32f3fc6f 100644 --- a/include/drm/drm_device.h +++ b/include/drm/drm_device.h @@ -17,6 +17,7 @@ struct drm_vblank_crtc; struct drm_sg_mem; struct drm_local_map; struct drm_vma_offset_manager; +struct drm_fb_helper; struct inode; @@ -185,6 +186,14 @@ struct drm_device { struct drm_vma_offset_manager *vma_offset_manager; /*@} */ int switch_power_state; + + /** + * @fb_helper: + * + * Pointer to the fbdev emulation structure. + * Set by drm_fb_helper_init() and cleared by drm_fb_helper_fini(). + */ + struct drm_fb_helper *fb_helper; }; #endif -- GitLab From 304a4f6accac4add4967fc062069fc8421c46aae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 30 Oct 2017 16:39:39 +0100 Subject: [PATCH 0085/1471] drm/fb-helper: Add .last_close and .output_poll_changed helpers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This adds helpers for the drm_driver->last_close and the drm_mode_config_funcs->output_poll_changed callbacks. Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171030153951.56269-4-noralf@tronnes.org --- drivers/gpu/drm/drm_fb_helper.c | 28 ++++++++++++++++++++++++++++ include/drm/drm_fb_helper.h | 11 +++++++++++ 2 files changed, 39 insertions(+) diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c index 19aaca9fea06e..defd6a76fef33 100644 --- a/drivers/gpu/drm/drm_fb_helper.c +++ b/drivers/gpu/drm/drm_fb_helper.c @@ -2633,6 +2633,34 @@ int drm_fb_helper_hotplug_event(struct drm_fb_helper *fb_helper) } EXPORT_SYMBOL(drm_fb_helper_hotplug_event); +/** + * drm_fb_helper_lastclose - DRM driver lastclose helper for fbdev emulation + * @dev: DRM device + * + * This function can be used as the &drm_driver->lastclose callback for drivers + * that only need to call drm_fb_helper_restore_fbdev_mode_unlocked(). + */ +void drm_fb_helper_lastclose(struct drm_device *dev) +{ + drm_fb_helper_restore_fbdev_mode_unlocked(dev->fb_helper); +} +EXPORT_SYMBOL(drm_fb_helper_lastclose); + +/** + * drm_fb_helper_output_poll_changed - DRM mode config \.output_poll_changed + * helper for fbdev emulation + * @dev: DRM device + * + * This function can be used as the + * &drm_mode_config_funcs.output_poll_changed callback for drivers that only + * need to call drm_fb_helper_hotplug_event(). + */ +void drm_fb_helper_output_poll_changed(struct drm_device *dev) +{ + drm_fb_helper_hotplug_event(dev->fb_helper); +} +EXPORT_SYMBOL(drm_fb_helper_output_poll_changed); + /* The Kconfig DRM_KMS_HELPER selects FRAMEBUFFER_CONSOLE (if !EXPERT) * but the module doesn't depend on any fb console symbols. At least * attempt to load fbcon to avoid leaving the system without a usable console. diff --git a/include/drm/drm_fb_helper.h b/include/drm/drm_fb_helper.h index 33fe959277428..877e5b395c02e 100644 --- a/include/drm/drm_fb_helper.h +++ b/include/drm/drm_fb_helper.h @@ -310,6 +310,9 @@ drm_pick_cmdline_mode(struct drm_fb_helper_connector *fb_helper_conn); int drm_fb_helper_add_one_connector(struct drm_fb_helper *fb_helper, struct drm_connector *connector); int drm_fb_helper_remove_one_connector(struct drm_fb_helper *fb_helper, struct drm_connector *connector); + +void drm_fb_helper_lastclose(struct drm_device *dev); +void drm_fb_helper_output_poll_changed(struct drm_device *dev); #else static inline void drm_fb_helper_prepare(struct drm_device *dev, struct drm_fb_helper *helper, @@ -507,6 +510,14 @@ drm_fb_helper_remove_one_connector(struct drm_fb_helper *fb_helper, return 0; } +static inline void drm_fb_helper_lastclose(struct drm_device *dev) +{ +} + +static inline void drm_fb_helper_output_poll_changed(struct drm_device *dev) +{ +} + #endif static inline int -- GitLab From 8ec47de21bfab96790c4202ae8cdb5092ad7ec33 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 30 Oct 2017 20:46:53 +0200 Subject: [PATCH 0086/1471] drm/i915: Pass around crtc and connector states for audio MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Explicitly pass the crtc and connector states into the audio code enable/disable hooks, and plumb them all the way down. This gets rid of almost all crtc->config and encoder->crtc uses. The one place where we still use them is i915_audio_component_sync_audio_rate() since that gets called from the audio driver and we don't have explicit states around then. Cc: Jani Nikula Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171030184654.17429-1-ville.syrjala@linux.intel.com Reviewed-by: Jani Nikula --- drivers/gpu/drm/i915/i915_drv.h | 10 +- drivers/gpu/drm/i915/intel_audio.c | 202 +++++++++++++++------------- drivers/gpu/drm/i915/intel_ddi.c | 6 +- drivers/gpu/drm/i915/intel_dp.c | 3 +- drivers/gpu/drm/i915/intel_dp_mst.c | 3 +- drivers/gpu/drm/i915/intel_drv.h | 4 +- drivers/gpu/drm/i915/intel_hdmi.c | 6 +- 7 files changed, 132 insertions(+), 102 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 4a7325c4189c5..c10fece58e867 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -726,10 +726,12 @@ struct drm_i915_display_funcs { void (*crtc_disable)(struct intel_crtc_state *old_crtc_state, struct drm_atomic_state *old_state); void (*update_crtcs)(struct drm_atomic_state *state); - void (*audio_codec_enable)(struct drm_connector *connector, - struct intel_encoder *encoder, - const struct drm_display_mode *adjusted_mode); - void (*audio_codec_disable)(struct intel_encoder *encoder); + void (*audio_codec_enable)(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state); + void (*audio_codec_disable)(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state); void (*fdi_link_train)(struct intel_crtc *crtc, const struct intel_crtc_state *crtc_state); void (*init_clock_gating)(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 0ddba16fde1bf..e565200372701 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -102,13 +102,13 @@ static const struct dp_aud_n_m dp_aud_n_m[] = { }; static const struct dp_aud_n_m * -audio_config_dp_get_n_m(struct intel_crtc *intel_crtc, int rate) +audio_config_dp_get_n_m(const struct intel_crtc_state *crtc_state, int rate) { int i; for (i = 0; i < ARRAY_SIZE(dp_aud_n_m); i++) { if (rate == dp_aud_n_m[i].sample_rate && - intel_crtc->config->port_clock == dp_aud_n_m[i].clock) + crtc_state->port_clock == dp_aud_n_m[i].clock) return &dp_aud_n_m[i]; } @@ -157,8 +157,10 @@ static const struct { }; /* get AUD_CONFIG_PIXEL_CLOCK_HDMI_* value for mode */ -static u32 audio_config_hdmi_pixel_clock(const struct drm_display_mode *adjusted_mode) +static u32 audio_config_hdmi_pixel_clock(const struct intel_crtc_state *crtc_state) { + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; int i; for (i = 0; i < ARRAY_SIZE(hdmi_audio_clock); i++) { @@ -179,9 +181,11 @@ static u32 audio_config_hdmi_pixel_clock(const struct drm_display_mode *adjusted return hdmi_audio_clock[i].config; } -static int audio_config_hdmi_get_n(const struct drm_display_mode *adjusted_mode, +static int audio_config_hdmi_get_n(const struct intel_crtc_state *crtc_state, int rate) { + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; int i; for (i = 0; i < ARRAY_SIZE(hdmi_aud_ncts); i++) { @@ -220,7 +224,9 @@ static bool intel_eld_uptodate(struct drm_connector *connector, return true; } -static void g4x_audio_codec_disable(struct intel_encoder *encoder) +static void g4x_audio_codec_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); uint32_t eldv, tmp; @@ -239,11 +245,12 @@ static void g4x_audio_codec_disable(struct intel_encoder *encoder) I915_WRITE(G4X_AUD_CNTL_ST, tmp); } -static void g4x_audio_codec_enable(struct drm_connector *connector, - struct intel_encoder *encoder, - const struct drm_display_mode *adjusted_mode) +static void g4x_audio_codec_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { - struct drm_i915_private *dev_priv = to_i915(connector->dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct drm_connector *connector = conn_state->connector; uint8_t *eld = connector->eld; uint32_t eldv; uint32_t tmp; @@ -279,16 +286,20 @@ static void g4x_audio_codec_enable(struct drm_connector *connector, } static void -hsw_dp_audio_config_update(struct intel_crtc *intel_crtc, enum port port, - const struct drm_display_mode *adjusted_mode) +hsw_dp_audio_config_update(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { - struct drm_i915_private *dev_priv = to_i915(intel_crtc->base.dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct i915_audio_component *acomp = dev_priv->audio_component; - int rate = acomp ? acomp->aud_sample_rate[port] : 0; - const struct dp_aud_n_m *nm = audio_config_dp_get_n_m(intel_crtc, rate); - enum pipe pipe = intel_crtc->pipe; + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + enum port port = encoder->port; + enum pipe pipe = crtc->pipe; + const struct dp_aud_n_m *nm; + int rate; u32 tmp; + rate = acomp ? acomp->aud_sample_rate[port] : 0; + nm = audio_config_dp_get_n_m(crtc_state, rate); if (nm) DRM_DEBUG_KMS("using Maud %u, Naud %u\n", nm->m, nm->n); else @@ -323,23 +334,26 @@ hsw_dp_audio_config_update(struct intel_crtc *intel_crtc, enum port port, } static void -hsw_hdmi_audio_config_update(struct intel_crtc *intel_crtc, enum port port, - const struct drm_display_mode *adjusted_mode) +hsw_hdmi_audio_config_update(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { - struct drm_i915_private *dev_priv = to_i915(intel_crtc->base.dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct i915_audio_component *acomp = dev_priv->audio_component; - int rate = acomp ? acomp->aud_sample_rate[port] : 0; - enum pipe pipe = intel_crtc->pipe; - int n; + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + enum port port = encoder->port; + enum pipe pipe = crtc->pipe; + int n, rate; u32 tmp; + rate = acomp ? acomp->aud_sample_rate[port] : 0; + tmp = I915_READ(HSW_AUD_CFG(pipe)); tmp &= ~AUD_CONFIG_N_VALUE_INDEX; tmp &= ~AUD_CONFIG_PIXEL_CLOCK_HDMI_MASK; tmp &= ~AUD_CONFIG_N_PROG_ENABLE; - tmp |= audio_config_hdmi_pixel_clock(adjusted_mode); + tmp |= audio_config_hdmi_pixel_clock(crtc_state); - n = audio_config_hdmi_get_n(adjusted_mode, rate); + n = audio_config_hdmi_get_n(crtc_state, rate); if (n != 0) { DRM_DEBUG_KMS("using N %d\n", n); @@ -363,20 +377,22 @@ hsw_hdmi_audio_config_update(struct intel_crtc *intel_crtc, enum port port, } static void -hsw_audio_config_update(struct intel_crtc *intel_crtc, enum port port, - const struct drm_display_mode *adjusted_mode) +hsw_audio_config_update(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { - if (intel_crtc_has_dp_encoder(intel_crtc->config)) - hsw_dp_audio_config_update(intel_crtc, port, adjusted_mode); + if (intel_crtc_has_dp_encoder(crtc_state)) + hsw_dp_audio_config_update(encoder, crtc_state); else - hsw_hdmi_audio_config_update(intel_crtc, port, adjusted_mode); + hsw_hdmi_audio_config_update(encoder, crtc_state); } -static void hsw_audio_codec_disable(struct intel_encoder *encoder) +static void hsw_audio_codec_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); - enum pipe pipe = intel_crtc->pipe; + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); + enum pipe pipe = crtc->pipe; uint32_t tmp; DRM_DEBUG_KMS("Disable audio codec on pipe %c\n", pipe_name(pipe)); @@ -389,7 +405,7 @@ static void hsw_audio_codec_disable(struct intel_encoder *encoder) tmp |= AUD_CONFIG_N_PROG_ENABLE; tmp &= ~AUD_CONFIG_UPPER_N_MASK; tmp &= ~AUD_CONFIG_LOWER_N_MASK; - if (intel_crtc_has_dp_encoder(intel_crtc->config)) + if (intel_crtc_has_dp_encoder(old_crtc_state)) tmp |= AUD_CONFIG_N_VALUE_INDEX; I915_WRITE(HSW_AUD_CFG(pipe), tmp); @@ -402,14 +418,14 @@ static void hsw_audio_codec_disable(struct intel_encoder *encoder) mutex_unlock(&dev_priv->av_mutex); } -static void hsw_audio_codec_enable(struct drm_connector *connector, - struct intel_encoder *intel_encoder, - const struct drm_display_mode *adjusted_mode) +static void hsw_audio_codec_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { - struct drm_i915_private *dev_priv = to_i915(connector->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(intel_encoder->base.crtc); - enum pipe pipe = intel_crtc->pipe; - enum port port = intel_encoder->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_connector *connector = conn_state->connector; + enum pipe pipe = crtc->pipe; const uint8_t *eld = connector->eld; uint32_t tmp; int len, i; @@ -448,17 +464,19 @@ static void hsw_audio_codec_enable(struct drm_connector *connector, I915_WRITE(HSW_AUD_PIN_ELD_CP_VLD, tmp); /* Enable timestamps */ - hsw_audio_config_update(intel_crtc, port, adjusted_mode); + hsw_audio_config_update(encoder, crtc_state); mutex_unlock(&dev_priv->av_mutex); } -static void ilk_audio_codec_disable(struct intel_encoder *intel_encoder) +static void ilk_audio_codec_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state) { - struct drm_i915_private *dev_priv = to_i915(intel_encoder->base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(intel_encoder->base.crtc); - enum pipe pipe = intel_crtc->pipe; - enum port port = intel_encoder->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); + enum pipe pipe = crtc->pipe; + enum port port = encoder->port; uint32_t tmp, eldv; i915_reg_t aud_config, aud_cntrl_st2; @@ -485,7 +503,7 @@ static void ilk_audio_codec_disable(struct intel_encoder *intel_encoder) tmp |= AUD_CONFIG_N_PROG_ENABLE; tmp &= ~AUD_CONFIG_UPPER_N_MASK; tmp &= ~AUD_CONFIG_LOWER_N_MASK; - if (intel_crtc_has_dp_encoder(intel_crtc->config)) + if (intel_crtc_has_dp_encoder(old_crtc_state)) tmp |= AUD_CONFIG_N_VALUE_INDEX; I915_WRITE(aud_config, tmp); @@ -497,14 +515,15 @@ static void ilk_audio_codec_disable(struct intel_encoder *intel_encoder) I915_WRITE(aud_cntrl_st2, tmp); } -static void ilk_audio_codec_enable(struct drm_connector *connector, - struct intel_encoder *intel_encoder, - const struct drm_display_mode *adjusted_mode) +static void ilk_audio_codec_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { - struct drm_i915_private *dev_priv = to_i915(connector->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(intel_encoder->base.crtc); - enum pipe pipe = intel_crtc->pipe; - enum port port = intel_encoder->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_connector *connector = conn_state->connector; + enum pipe pipe = crtc->pipe; + enum port port = encoder->port; uint8_t *eld = connector->eld; uint32_t tmp, eldv; int len, i; @@ -568,36 +587,36 @@ static void ilk_audio_codec_enable(struct drm_connector *connector, tmp &= ~AUD_CONFIG_N_VALUE_INDEX; tmp &= ~AUD_CONFIG_N_PROG_ENABLE; tmp &= ~AUD_CONFIG_PIXEL_CLOCK_HDMI_MASK; - if (intel_crtc_has_dp_encoder(intel_crtc->config)) + if (intel_crtc_has_dp_encoder(crtc_state)) tmp |= AUD_CONFIG_N_VALUE_INDEX; else - tmp |= audio_config_hdmi_pixel_clock(adjusted_mode); + tmp |= audio_config_hdmi_pixel_clock(crtc_state); I915_WRITE(aud_config, tmp); } /** * intel_audio_codec_enable - Enable the audio codec for HD audio - * @intel_encoder: encoder on which to enable audio + * @encoder: encoder on which to enable audio * @crtc_state: pointer to the current crtc state. * @conn_state: pointer to the current connector state. * * The enable sequences may only be performed after enabling the transcoder and * port, and after completed link training. */ -void intel_audio_codec_enable(struct intel_encoder *intel_encoder, +void intel_audio_codec_enable(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) { - struct drm_encoder *encoder = &intel_encoder->base; - const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; - struct drm_connector *connector; - struct drm_i915_private *dev_priv = to_i915(encoder->dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct i915_audio_component *acomp = dev_priv->audio_component; - enum port port = intel_encoder->port; - enum pipe pipe = to_intel_crtc(crtc_state->base.crtc)->pipe; + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_connector *connector = conn_state->connector; + const struct drm_display_mode *adjusted_mode = + &crtc_state->base.adjusted_mode; + enum port port = encoder->port; + enum pipe pipe = crtc->pipe; - connector = conn_state->connector; - if (!connector || !connector->eld[0]) + if (!connector->eld[0]) return; DRM_DEBUG_DRIVER("ELD on [CONNECTOR:%d:%s], [ENCODER:%d:%s]\n", @@ -609,19 +628,20 @@ void intel_audio_codec_enable(struct intel_encoder *intel_encoder, connector->eld[6] = drm_av_sync_delay(connector, adjusted_mode) / 2; if (dev_priv->display.audio_codec_enable) - dev_priv->display.audio_codec_enable(connector, intel_encoder, - adjusted_mode); + dev_priv->display.audio_codec_enable(encoder, + crtc_state, + conn_state); mutex_lock(&dev_priv->av_mutex); - intel_encoder->audio_connector = connector; + encoder->audio_connector = connector; /* referred in audio callbacks */ - dev_priv->av_enc_map[pipe] = intel_encoder; + dev_priv->av_enc_map[pipe] = encoder; mutex_unlock(&dev_priv->av_mutex); if (acomp && acomp->audio_ops && acomp->audio_ops->pin_eld_notify) { /* audio drivers expect pipe = -1 to indicate Non-MST cases */ - if (intel_encoder->type != INTEL_OUTPUT_DP_MST) + if (encoder->type != INTEL_OUTPUT_DP_MST) pipe = -1; acomp->audio_ops->pin_eld_notify(acomp->audio_ops->audio_ptr, (int) port, (int) pipe); @@ -629,36 +649,41 @@ void intel_audio_codec_enable(struct intel_encoder *intel_encoder, intel_lpe_audio_notify(dev_priv, pipe, port, connector->eld, crtc_state->port_clock, - intel_encoder->type == INTEL_OUTPUT_DP); + encoder->type == INTEL_OUTPUT_DP); } /** * intel_audio_codec_disable - Disable the audio codec for HD audio - * @intel_encoder: encoder on which to disable audio + * @encoder: encoder on which to disable audio + * @crtc_state: pointer to the old crtc state. + * @conn_state: pointer to the old connector state. * * The disable sequences must be performed before disabling the transcoder or * port. */ -void intel_audio_codec_disable(struct intel_encoder *intel_encoder) +void intel_audio_codec_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state) { - struct drm_encoder *encoder = &intel_encoder->base; - struct drm_i915_private *dev_priv = to_i915(encoder->dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct i915_audio_component *acomp = dev_priv->audio_component; - enum port port = intel_encoder->port; - struct intel_crtc *crtc = to_intel_crtc(encoder->crtc); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); + enum port port = encoder->port; enum pipe pipe = crtc->pipe; if (dev_priv->display.audio_codec_disable) - dev_priv->display.audio_codec_disable(intel_encoder); + dev_priv->display.audio_codec_disable(encoder, + old_crtc_state, + old_conn_state); mutex_lock(&dev_priv->av_mutex); - intel_encoder->audio_connector = NULL; + encoder->audio_connector = NULL; dev_priv->av_enc_map[pipe] = NULL; mutex_unlock(&dev_priv->av_mutex); if (acomp && acomp->audio_ops && acomp->audio_ops->pin_eld_notify) { /* audio drivers expect pipe = -1 to indicate Non-MST cases */ - if (intel_encoder->type != INTEL_OUTPUT_DP_MST) + if (encoder->type != INTEL_OUTPUT_DP_MST) pipe = -1; acomp->audio_ops->pin_eld_notify(acomp->audio_ops->audio_ptr, (int) port, (int) pipe); @@ -793,10 +818,9 @@ static int i915_audio_component_sync_audio_rate(struct device *kdev, int port, int pipe, int rate) { struct drm_i915_private *dev_priv = kdev_to_i915(kdev); - struct intel_encoder *intel_encoder; - struct intel_crtc *crtc; - struct drm_display_mode *adjusted_mode; struct i915_audio_component *acomp = dev_priv->audio_component; + struct intel_encoder *encoder; + struct intel_crtc *crtc; int err = 0; if (!HAS_DDI(dev_priv)) @@ -806,23 +830,19 @@ static int i915_audio_component_sync_audio_rate(struct device *kdev, int port, mutex_lock(&dev_priv->av_mutex); /* 1. get the pipe */ - intel_encoder = get_saved_enc(dev_priv, port, pipe); - if (!intel_encoder || !intel_encoder->base.crtc) { + encoder = get_saved_enc(dev_priv, port, pipe); + if (!encoder || !encoder->base.crtc) { DRM_DEBUG_KMS("Not valid for port %c\n", port_name(port)); err = -ENODEV; goto unlock; } - /* pipe passed from the audio driver will be -1 for Non-MST case */ - crtc = to_intel_crtc(intel_encoder->base.crtc); - pipe = crtc->pipe; - - adjusted_mode = &crtc->config->base.adjusted_mode; + crtc = to_intel_crtc(encoder->base.crtc); /* port must be valid now, otherwise the pipe will be invalid */ acomp->aud_sample_rate[port] = rate; - hsw_audio_config_update(crtc, port, adjusted_mode); + hsw_audio_config_update(encoder, crtc->config); unlock: mutex_unlock(&dev_priv->av_mutex); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index eb8f4e3bb85f8..ace674cd79b92 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -2425,7 +2425,8 @@ static void intel_disable_ddi_dp(struct intel_encoder *encoder, struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); intel_edp_drrs_disable(intel_dp, old_crtc_state); intel_psr_disable(intel_dp, old_crtc_state); @@ -2437,7 +2438,8 @@ static void intel_disable_ddi_hdmi(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); intel_hdmi_handle_sink_scrambling(encoder, old_conn_state->connector, diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index f0c49962ffbeb..d27c0145ac91d 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2708,7 +2708,8 @@ static void intel_disable_dp(struct intel_encoder *encoder, struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); /* Make sure the panel is off before trying to change the mode. But also * ensure that we have vdd while we switch off the panel. */ diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 6f11bb35f66f5..653ca39789b8f 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -149,7 +149,8 @@ static void intel_mst_disable_dp(struct intel_encoder *encoder, DRM_ERROR("failed to update payload %d\n", ret); } if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); } static void intel_mst_post_disable_dp(struct intel_encoder *encoder, diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 4498d743cebc3..00b4886880427 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1311,7 +1311,9 @@ void intel_init_audio_hooks(struct drm_i915_private *dev_priv); void intel_audio_codec_enable(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state); -void intel_audio_codec_disable(struct intel_encoder *encoder); +void intel_audio_codec_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state); void i915_audio_component_init(struct drm_i915_private *dev_priv); void i915_audio_component_cleanup(struct drm_i915_private *dev_priv); void intel_audio_init(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index fe46abfeba26a..fa1c793a21efd 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1206,7 +1206,8 @@ static void g4x_disable_hdmi(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); intel_disable_hdmi(encoder, old_crtc_state, old_conn_state); } @@ -1216,7 +1217,8 @@ static void pch_disable_hdmi(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { if (old_crtc_state->has_audio) - intel_audio_codec_disable(encoder); + intel_audio_codec_disable(encoder, + old_crtc_state, old_conn_state); } static void pch_post_disable_hdmi(struct intel_encoder *encoder, -- GitLab From 9f846643c788bbdf4d4b4fe587661705d07de8ef Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 30 Oct 2017 20:46:54 +0200 Subject: [PATCH 0087/1471] drm/i915: Remove most encoder->type uses from the audio code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit encoder->type isn't genreally safe around DDI ports, so let's replace some uses in the audio code with the crtc state's output_types instead. Actually in these cases encoder->type would work since the DP SST case is only relevant for VLV/CHV and encoder->type==DP is a thing on those platforms. The DP MST cases would work as well since MST encoder->type==DP_MST always. But I think it's best to try and minimize the encoder->type use in general to avoid showing a bad example to people. Cc: Jani Nikula Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171030184654.17429-2-ville.syrjala@linux.intel.com Reviewed-by: Jani Nikula --- drivers/gpu/drm/i915/intel_audio.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index e565200372701..4705194b19929 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -641,7 +641,7 @@ void intel_audio_codec_enable(struct intel_encoder *encoder, if (acomp && acomp->audio_ops && acomp->audio_ops->pin_eld_notify) { /* audio drivers expect pipe = -1 to indicate Non-MST cases */ - if (encoder->type != INTEL_OUTPUT_DP_MST) + if (!intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) pipe = -1; acomp->audio_ops->pin_eld_notify(acomp->audio_ops->audio_ptr, (int) port, (int) pipe); @@ -649,7 +649,7 @@ void intel_audio_codec_enable(struct intel_encoder *encoder, intel_lpe_audio_notify(dev_priv, pipe, port, connector->eld, crtc_state->port_clock, - encoder->type == INTEL_OUTPUT_DP); + intel_crtc_has_dp_encoder(crtc_state)); } /** @@ -683,7 +683,7 @@ void intel_audio_codec_disable(struct intel_encoder *encoder, if (acomp && acomp->audio_ops && acomp->audio_ops->pin_eld_notify) { /* audio drivers expect pipe = -1 to indicate Non-MST cases */ - if (encoder->type != INTEL_OUTPUT_DP_MST) + if (!intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_DP_MST)) pipe = -1; acomp->audio_ops->pin_eld_notify(acomp->audio_ops->audio_ptr, (int) port, (int) pipe); -- GitLab From d9b99ffcb5aa113d175a3df845c72eac0d4624e0 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Fri, 27 Oct 2017 15:32:07 -0700 Subject: [PATCH 0088/1471] drm/i915/cnl: Remove unnecessary check in cnl_setup_private_ppat There is no need check if PPGTT is disabled because that not possible in CNL. Execlists and GuC submission modes rely on at least aliasing PPGTT and even intel_sanitize_enable_ppgtt says: "We don't allow disabling PPGTT for gen9+ as it's a requirement for execlists, the sole mechanism available to submit work." Suggested-by: Daniele Ceraolo Spurio Cc: Rodrigo Vivi Signed-off-by: Michel Thierry Reviewed-by: Daniele Ceraolo Spurio Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171027223207.7869-1-michel.thierry@intel.com --- drivers/gpu/drm/i915/i915_gem_gtt.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 5eaa6893daaa1..0684d5df97d9d 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -3178,12 +3178,6 @@ static void cnl_setup_private_ppat(struct intel_ppat *ppat) ppat->match = bdw_private_pat_match; ppat->clear_value = GEN8_PPAT_WB | GEN8_PPAT_LLCELLC | GEN8_PPAT_AGE(3); - /* XXX: spec is unclear if this is still needed for CNL+ */ - if (!USES_PPGTT(ppat->i915)) { - __alloc_ppat_entry(ppat, 0, GEN8_PPAT_UC); - return; - } - __alloc_ppat_entry(ppat, 0, GEN8_PPAT_WB | GEN8_PPAT_LLC); __alloc_ppat_entry(ppat, 1, GEN8_PPAT_WC | GEN8_PPAT_LLCELLC); __alloc_ppat_entry(ppat, 2, GEN8_PPAT_WT | GEN8_PPAT_LLCELLC); -- GitLab From 1d033beb20d6d5885587a02a393b6598d766a382 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 31 Oct 2017 10:36:07 +0000 Subject: [PATCH 0089/1471] drm/i915: Check incoming alignment for unfenced buffers (on i915gm) In case the object has changed tiling between calls to execbuf, we need to check if the existing offset inside the GTT matches the new tiling constraint. We even need to do this for "unfenced" tiled objects, where the 3D commands use an implied fence and so the object still needs to match the physical fence restrictions on alignment (only required for gen2 and early gen3). In commit 2889caa92321 ("drm/i915: Eliminate lots of iterations over the execobjects array"), the idea was to remove the second guessing and only set the NEEDS_MAP flag when required. However, the entire check for an unusable offset for fencing was removed and not just the secondary check. I.e. /* avoid costly ping-pong once a batch bo ended up non-mappable */ if (entry->flags & __EXEC_OBJECT_NEEDS_MAP && !i915_vma_is_map_and_fenceable(vma)) return !only_mappable_for_reloc(entry->flags); was entirely removed as the ping-pong between execbuf passes was fixed, but its primary purpose in forcing unaligned unfenced access to be rebound was forgotten. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103502 Fixes: 2889caa92321 ("drm/i915: Eliminate lots of iterations over the execobjects array") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171031103607.17836-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 3d7190764f101..5b8213b1a8c7e 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -343,6 +343,10 @@ eb_vma_misplaced(const struct drm_i915_gem_exec_object2 *entry, (vma->node.start + vma->node.size - 1) >> 32) return true; + if (flags & __EXEC_OBJECT_NEEDS_MAP && + !i915_vma_is_map_and_fenceable(vma)) + return true; + return false; } -- GitLab From e5330ac1f50b897d245753828e8887f297f69dd0 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 31 Oct 2017 12:22:35 +0000 Subject: [PATCH 0090/1471] drm/i915: Check that the breadcrumb wasn't disarmed automatically before parking MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We will disarm the breadcrumb interrupt if we see a user interrupt whilst no one is waiting. This may race with the call to intel_engine_disarm_breadcrumbs() triggering an assert that we aren't trying to do the same job twice. Prevent this by checking that the irq is still armed after flushing the interrupt (for the irq spinlock). Fixes: bcbd5c33a342 ("drm/i915/guc: Always enable the breadcrumbs irq") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Michał Winiarski Link: https://patchwork.freedesktop.org/patch/msgid/20171031122235.1395-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_breadcrumbs.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c index 1bd0a461d6653..4de054f8c1baf 100644 --- a/drivers/gpu/drm/i915/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c @@ -219,7 +219,8 @@ void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine) spin_lock(&b->irq_lock); first = fetch_and_zero(&b->irq_wait); - __intel_engine_disarm_breadcrumbs(engine); + if (b->irq_armed) + __intel_engine_disarm_breadcrumbs(engine); spin_unlock(&b->irq_lock); rbtree_postorder_for_each_entry_safe(wait, n, &b->waiters, node) { -- GitLab From 680273879d125d644831b8de42c66576e6290378 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 27 Oct 2017 12:06:16 +0100 Subject: [PATCH 0091/1471] drm/i915: Move parking-while-active warning to intel_engines_park() We will want to break this down to give detailed per-engine warnings as to why we still think we are active as we attempt to park the engines. For the first step, just move the warning verbatim from the idle-worker to intel_engines_park(). Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171027110617.31745-3-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_gem.c | 7 ------- drivers/gpu/drm/i915/intel_engine_cs.c | 7 +++++++ 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e43688f778172..9470ba0c1930e 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3352,13 +3352,6 @@ i915_gem_idle_work_handler(struct work_struct *work) */ synchronize_irq(dev_priv->drm.irq); - /* - * We are committed now to parking the engines, make sure there - * will be no more interrupts arriving later. - */ - if (!intel_engines_are_idle(dev_priv)) - DRM_ERROR("Timeout waiting for engines to idle\n"); - intel_engines_park(dev_priv); i915_gem_timelines_mark_idle(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index f31f2d6384c35..9767586e22898 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1613,6 +1613,13 @@ void intel_engines_park(struct drm_i915_private *i915) struct intel_engine_cs *engine; enum intel_engine_id id; + /* + * We are committed now to parking the engines, make sure there + * will be no more interrupts arriving later. + */ + if (!intel_engines_are_idle(dev_priv)) + DRM_ERROR("Timeout waiting for engines to idle\n"); + for_each_engine(engine, i915, id) { if (engine->park) engine->park(engine); -- GitLab From 3265124a2d3744d789ede58452ab6f8a9b454be8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 27 Oct 2017 12:06:17 +0100 Subject: [PATCH 0092/1471] drm/i915: Give more details for the active-when-parking warning for the engines If the we think the engine is still active when we attempt to park it, we want more details -- so dump the engine state. References: https://bugs.freedesktop.org/show_bug.cgi?id=103479 Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171027110617.31745-4-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/intel_engine_cs.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 9767586e22898..6895a90af0083 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1613,14 +1613,20 @@ void intel_engines_park(struct drm_i915_private *i915) struct intel_engine_cs *engine; enum intel_engine_id id; - /* - * We are committed now to parking the engines, make sure there - * will be no more interrupts arriving later. - */ - if (!intel_engines_are_idle(dev_priv)) - DRM_ERROR("Timeout waiting for engines to idle\n"); - for_each_engine(engine, i915, id) { + /* + * We are committed now to parking the engines, make sure there + * will be no more interrupts arriving later and the engines + * are truly idle. + */ + if (!intel_engine_is_idle(engine)) { + struct drm_printer p = drm_debug_printer(__func__); + + DRM_ERROR("%s is not idle before parking\n", + engine->name); + intel_engine_dump(engine, &p); + } + if (engine->park) engine->park(engine); -- GitLab From 6e7406db8c3a6998e84a605d3e1e04b166e795d1 Mon Sep 17 00:00:00 2001 From: Mika Kahola Date: Wed, 1 Nov 2017 12:08:50 +0200 Subject: [PATCH 0093/1471] drm/i915/cnl: Symmetric scalers for each pipe For Cannonlake the number of scalers for each pipe is 2. Let's increase the number of scalers for pipe C. v2: Use INTEL_GEN() instead of IS_CANNONLAKE() Signed-off-by: Mika Kahola Reviewed-by: Rodrigo Vivi Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/1509530930-24960-1-git-send-email-mika.kahola@intel.com --- drivers/gpu/drm/i915/intel_device_info.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c index 875d428ea75f2..db03d179fc854 100644 --- a/drivers/gpu/drm/i915/intel_device_info.c +++ b/drivers/gpu/drm/i915/intel_device_info.c @@ -347,7 +347,10 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) struct intel_device_info *info = mkwrite_device_info(dev_priv); enum pipe pipe; - if (INTEL_GEN(dev_priv) >= 9) { + if (INTEL_GEN(dev_priv) >= 10) { + for_each_pipe(dev_priv, pipe) + info->num_scalers[pipe] = 2; + } else if (INTEL_GEN(dev_priv) == 9) { info->num_scalers[PIPE_A] = 2; info->num_scalers[PIPE_B] = 2; info->num_scalers[PIPE_C] = 1; -- GitLab From 7ce66118aaca8c9ed23cd44d629b8037a4117c11 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Wed, 1 Nov 2017 19:21:58 -0400 Subject: [PATCH 0094/1471] drm/amd: Update kgd_kfd interface for resuming SDMA queues Add wptr and mm parameters to hqd_sdma_load and pass these parameters from device_queue_manager through the mqd_manager. SDMA doesn't support polling while the engine believes it's idle. The driver must update the wptr. The new parameters will be used for looking up the updated value from the specified mm when SDMA queues are resumed after being disabled. Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 6 ++++-- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c | 4 +++- drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 3 ++- 4 files changed, 13 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 1e3e9be7d77ec..a55d7948075bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -105,7 +105,8 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd); +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); @@ -375,7 +376,8 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd) +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm) { struct amdgpu_device *adev = get_amdgpu_device(kgd); struct cik_sdma_rlc_registers *m; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 056929b8ccd04..1017ff51c957c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -64,7 +64,8 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd); +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); @@ -358,7 +359,8 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, return 0; } -static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd) +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm) { return 0; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 4728fad3fd742..ea02bfa3bda3d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -160,7 +160,9 @@ static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, uint32_t queue_id, struct queue_properties *p, struct mm_struct *mms) { - return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd); + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + (uint32_t __user *)p->write_ptr, + mms); } static int update_mqd(struct mqd_manager *mm, void *mqd, diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index f516fd10e6ba7..c6d4e64531c71 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -184,7 +184,8 @@ struct kfd2kgd_calls { uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); - int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd); + int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm); bool (*hqd_is_occupied)(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); -- GitLab From fd0f0762dc590736a6692707860af3ce793f5028 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Wed, 1 Nov 2017 19:21:59 -0400 Subject: [PATCH 0095/1471] drm/amdgpu: Add support for resuming SDMA queues w/o HWS Save wptr in hqd_sdma_destroy, restore it in hqd_sdma_load. Also read updated wptr from user mode when resuming an SDMA queue. Signed-off-by: Jay Cornwall Signed-off-by: Yong Zhao Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 30 ++++++++++++------- .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 7 ++--- 2 files changed, 23 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index a55d7948075bd..14333af92ed9d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -412,10 +412,17 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, WREG32(mmSDMA0_GFX_CONTEXT_CNTL, data); } - WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, - m->sdma_rlc_doorbell); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, 0); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, 0); + data = REG_SET_FIELD(m->sdma_rlc_doorbell, SDMA0_RLC0_DOORBELL, + ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, data); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, m->sdma_rlc_rb_rptr); + + if (read_user_wptr(mm, wptr, data)) + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, data); + else + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, + m->sdma_rlc_rb_rptr); + WREG32(sdma_base_addr + mmSDMA0_RLC0_VIRTUAL_ADDR, m->sdma_rlc_virtual_addr); WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, m->sdma_rlc_rb_base); @@ -425,8 +432,10 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, m->sdma_rlc_rb_rptr_addr_lo); WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_HI, m->sdma_rlc_rb_rptr_addr_hi); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, - m->sdma_rlc_rb_cntl); + + data = REG_SET_FIELD(m->sdma_rlc_rb_cntl, SDMA0_RLC0_RB_CNTL, + RB_ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, data); return 0; } @@ -577,7 +586,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, struct cik_sdma_rlc_registers *m; uint32_t sdma_base_addr; uint32_t temp; - int timeout = utimeout; + unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies; m = get_sdma_mqd(mqd); sdma_base_addr = get_sdma_base_addr(m); @@ -590,10 +599,9 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT) break; - if (timeout <= 0) + if (time_after(jiffies, end_jiffies)) return -ETIME; - msleep(20); - timeout -= 20; + usleep_range(500, 1000); } WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0); @@ -601,6 +609,8 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL) | SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK); + m->sdma_rlc_rb_rptr = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 1017ff51c957c..c0c8cc3b23b0a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -514,7 +514,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, struct cik_sdma_rlc_registers *m; uint32_t sdma_base_addr; uint32_t temp; - int timeout = utimeout; + unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies; m = get_sdma_mqd(mqd); sdma_base_addr = get_sdma_base_addr(m); @@ -527,10 +527,9 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT) break; - if (timeout <= 0) + if (time_after(jiffies, end_jiffies)) return -ETIME; - msleep(20); - timeout -= 20; + usleep_range(500, 1000); } WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0); -- GitLab From 9807c3668500b0964e8ee59d68083675971bc224 Mon Sep 17 00:00:00 2001 From: Philip Cox Date: Wed, 1 Nov 2017 19:22:00 -0400 Subject: [PATCH 0096/1471] drm/amdgpu: Implement amdgpu SDMA functions for VI Signed-off-by: Philip Cox Signed-off-by: shaoyun liu Signed-off-by: Yong Zhao Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 87 ++++++++++++++++--- drivers/gpu/drm/amd/amdgpu/vid.h | 2 + drivers/gpu/drm/amd/include/vi_structs.h | 2 + 3 files changed, 80 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index c0c8cc3b23b0a..1d989e4915b11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -45,7 +45,7 @@ enum hqd_dequeue_request_type { RESET_WAVES }; -struct cik_sdma_rlc_registers; +struct vi_sdma_mqd; /* * Register access functions @@ -269,9 +269,15 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) return 0; } -static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers *m) +static inline uint32_t get_sdma_base_addr(struct vi_sdma_mqd *m) { - return 0; + uint32_t retval; + + retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET + + m->sdma_queue_id * KFD_VI_SDMA_QUEUE_OFFSET; + pr_debug("kfd: sdma base address: 0x%x\n", retval); + + return retval; } static inline struct vi_mqd *get_mqd(void *mqd) @@ -279,9 +285,9 @@ static inline struct vi_mqd *get_mqd(void *mqd) return (struct vi_mqd *)mqd; } -static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) +static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd) { - return (struct cik_sdma_rlc_registers *)mqd; + return (struct vi_sdma_mqd *)mqd; } static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, @@ -362,6 +368,63 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct vi_sdma_mqd *m; + unsigned long end_jiffies; + uint32_t sdma_base_addr; + uint32_t data; + + m = get_sdma_mqd(mqd); + sdma_base_addr = get_sdma_base_addr(m); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, + m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK)); + + end_jiffies = msecs_to_jiffies(2000) + jiffies; + while (true) { + data = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); + if (data & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK) + break; + if (time_after(jiffies, end_jiffies)) + return -ETIME; + usleep_range(500, 1000); + } + if (m->sdma_engine_id) { + data = RREG32(mmSDMA1_GFX_CONTEXT_CNTL); + data = REG_SET_FIELD(data, SDMA1_GFX_CONTEXT_CNTL, + RESUME_CTX, 0); + WREG32(mmSDMA1_GFX_CONTEXT_CNTL, data); + } else { + data = RREG32(mmSDMA0_GFX_CONTEXT_CNTL); + data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL, + RESUME_CTX, 0); + WREG32(mmSDMA0_GFX_CONTEXT_CNTL, data); + } + + data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA0_RLC0_DOORBELL, + ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, data); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, m->sdmax_rlcx_rb_rptr); + + if (read_user_wptr(mm, wptr, data)) + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, data); + else + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, + m->sdmax_rlcx_rb_rptr); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_VIRTUAL_ADDR, + m->sdmax_rlcx_virtual_addr); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, m->sdmax_rlcx_rb_base); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE_HI, + m->sdmax_rlcx_rb_base_hi); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_LO, + m->sdmax_rlcx_rb_rptr_addr_lo); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_HI, + m->sdmax_rlcx_rb_rptr_addr_hi); + + data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA0_RLC0_RB_CNTL, + RB_ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, data); + return 0; } @@ -390,7 +453,7 @@ static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct cik_sdma_rlc_registers *m; + struct vi_sdma_mqd *m; uint32_t sdma_base_addr; uint32_t sdma_rlc_rb_cntl; @@ -511,7 +574,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, unsigned int utimeout) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - struct cik_sdma_rlc_registers *m; + struct vi_sdma_mqd *m; uint32_t sdma_base_addr; uint32_t temp; unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies; @@ -525,7 +588,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, while (true) { temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); - if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT) + if (temp & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK) break; if (time_after(jiffies, end_jiffies)) return -ETIME; @@ -533,9 +596,11 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, } WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, 0); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, 0); - WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, 0); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, + RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL) | + SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK); + + m->sdmax_rlcx_rb_rptr = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/vid.h b/drivers/gpu/drm/amd/amdgpu/vid.h index dbf3703cbd1b8..19ddd2312e00d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vid.h +++ b/drivers/gpu/drm/amd/amdgpu/vid.h @@ -27,6 +27,8 @@ #define SDMA1_REGISTER_OFFSET 0x200 /* not a register */ #define SDMA_MAX_INSTANCE 2 +#define KFD_VI_SDMA_QUEUE_OFFSET 0x80 /* not a register */ + /* crtc instance offsets */ #define CRTC0_REGISTER_OFFSET (0x1b9c - 0x1b9c) #define CRTC1_REGISTER_OFFSET (0x1d9c - 0x1b9c) diff --git a/drivers/gpu/drm/amd/include/vi_structs.h b/drivers/gpu/drm/amd/include/vi_structs.h index 20234820194bd..717fbae1d3620 100644 --- a/drivers/gpu/drm/amd/include/vi_structs.h +++ b/drivers/gpu/drm/amd/include/vi_structs.h @@ -153,6 +153,8 @@ struct vi_sdma_mqd { uint32_t reserved_125; uint32_t reserved_126; uint32_t reserved_127; + uint32_t sdma_engine_id; + uint32_t sdma_queue_id; }; struct vi_mqd { -- GitLab From 97b9ad12ba08353b3e19d85e65761b4cd936c12c Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Wed, 1 Nov 2017 19:22:01 -0400 Subject: [PATCH 0097/1471] drm/amdkfd: Use ASIC-specific SDMA MQD type Signed-off-by: shaoyun liu Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c | 13 +++++-------- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 5 +++++ drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 2 -- 3 files changed, 10 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index ea02bfa3bda3d..9873929ab4efa 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -36,6 +36,11 @@ static inline struct cik_mqd *get_mqd(void *mqd) return (struct cik_mqd *)mqd; } +static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) +{ + return (struct cik_sdma_rlc_registers *)mqd; +} + static int init_mqd(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) @@ -362,14 +367,6 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, return 0; } -struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd) -{ - struct cik_sdma_rlc_registers *m; - - m = (struct cik_sdma_rlc_registers *)mqd; - - return m; -} struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, struct kfd_dev *dev) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index 4ea854f9007b7..dc9249709e9bc 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -38,6 +38,11 @@ static inline struct vi_mqd *get_mqd(void *mqd) return (struct vi_mqd *)mqd; } +static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd) +{ + return (struct vi_sdma_mqd *)mqd; +} + static int init_mqd(struct mqd_manager *mm, void **mqd, struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, struct queue_properties *q) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 9e4134c5b4819..47504737ab4a2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -643,8 +643,6 @@ int kgd2kfd_resume(struct kfd_dev *kfd); int kfd_init_apertures(struct kfd_process *process); /* Queue Context Management */ -struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd); - int init_queue(struct queue **q, const struct queue_properties *properties); void uninit_queue(struct queue *q); void print_queue_properties(struct queue_properties *q); -- GitLab From 5aaf2befd4eceb2efa16c63ccb64c8e13785594a Mon Sep 17 00:00:00 2001 From: Philip Cox Date: Wed, 1 Nov 2017 19:22:02 -0400 Subject: [PATCH 0098/1471] drm/amdkfd: Implement amdkfd SDMA functions for VI Signed-off-by: Philip Cox Signed-off-by: shaoyun liu Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- .../gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 103 +++++++++++++++++- 1 file changed, 102 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index dc9249709e9bc..a117d2b7721f5 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -30,7 +30,7 @@ #include "vi_structs.h" #include "gca/gfx_8_0_sh_mask.h" #include "gca/gfx_8_0_enum.h" - +#include "oss/oss_3_0_sh_mask.h" #define CP_MQD_CONTROL__PRIV_STATE__SHIFT 0x8 static inline struct vi_mqd *get_mqd(void *mqd) @@ -239,6 +239,101 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, return retval; } +static int init_mqd_sdma(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + int retval; + struct vi_sdma_mqd *m; + + + retval = kfd_gtt_sa_allocate(mm->dev, + sizeof(struct vi_sdma_mqd), + mqd_mem_obj); + + if (retval != 0) + return -ENOMEM; + + m = (struct vi_sdma_mqd *) (*mqd_mem_obj)->cpu_ptr; + + memset(m, 0, sizeof(struct vi_sdma_mqd)); + + *mqd = m; + if (gart_addr != NULL) + *gart_addr = (*mqd_mem_obj)->gpu_addr; + + retval = mm->update_mqd(mm, m, q); + + return retval; +} + +static void uninit_mqd_sdma(struct mqd_manager *mm, void *mqd, + struct kfd_mem_obj *mqd_mem_obj) +{ + kfd_gtt_sa_free(mm->dev, mqd_mem_obj); +} + +static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + struct queue_properties *p, struct mm_struct *mms) +{ + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + (uint32_t __user *)p->write_ptr, + mms); +} + +static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct vi_sdma_mqd *m; + + m = get_sdma_mqd(mqd); + m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1) + << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | + q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | + 6 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT; + + m->sdmax_rlcx_rb_base = lower_32_bits(q->queue_address >> 8); + m->sdmax_rlcx_rb_base_hi = upper_32_bits(q->queue_address >> 8); + m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits((uint64_t)q->read_ptr); + m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits((uint64_t)q->read_ptr); + m->sdmax_rlcx_doorbell = + q->doorbell_off << SDMA0_RLC0_DOORBELL__OFFSET__SHIFT; + + m->sdmax_rlcx_virtual_addr = q->sdma_vm_addr; + + m->sdma_engine_id = q->sdma_engine_id; + m->sdma_queue_id = q->sdma_queue_id; + + q->is_active = (q->queue_size > 0 && + q->queue_address != 0 && + q->queue_percent > 0); + + return 0; +} + +/* + * * preempt type here is ignored because there is only one way + * * to preempt sdma queue + */ +static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); +} + +static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); +} + + + struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, struct kfd_dev *dev) { @@ -272,6 +367,12 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, mqd->is_occupied = is_occupied; break; case KFD_MQD_TYPE_SDMA: + mqd->init_mqd = init_mqd_sdma; + mqd->uninit_mqd = uninit_mqd_sdma; + mqd->load_mqd = load_mqd_sdma; + mqd->update_mqd = update_mqd_sdma; + mqd->destroy_mqd = destroy_mqd_sdma; + mqd->is_occupied = is_occupied_sdma; break; default: kfree(mqd); -- GitLab From 6d5646006f66880574c75c4e142ef370b4503af3 Mon Sep 17 00:00:00 2001 From: Jeff McGee Date: Wed, 1 Nov 2017 15:16:30 -0700 Subject: [PATCH 0099/1471] drm/i915/guc: Clear terminated attribute bit on GuC preemption context If GuC firmware performs an engine reset while that engine had a preemption pending, it will set the terminated attribute bit on our preemption stage descriptor. GuC firmware retains all pending work items for a high-priority GuC client, unlike the normal-priority GuC client where work items are dropped. It wants to make sure the preempt- to-idle work doesn't run when scheduling resumes, and uses this bit to inform its scheduler and presumably us as well. Our job is to clear it for the next preemption after reset, otherwise that and future preemptions will never complete. We'll just clear it every time. Signed-off-by: Jeff McGee Link: https://patchwork.freedesktop.org/patch/msgid/20171101221630.25086-1-jeff.mcgee@intel.com Reviewed-by: Michel Thierry Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_guc_submission.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 3049a0781b885..d14c1342f09dc 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -590,6 +590,7 @@ static void inject_preempt_context(struct work_struct *work) struct intel_guc *guc = container_of(preempt_work, typeof(*guc), preempt_work[engine->id]); struct i915_guc_client *client = guc->preempt_client; + struct guc_stage_desc *stage_desc = __get_stage_desc(client); struct intel_ring *ring = client->owner->engine[engine->id].ring; u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(client->owner, engine)); @@ -623,6 +624,20 @@ static void inject_preempt_context(struct work_struct *work) ring->tail / sizeof(u64), 0); spin_unlock_irq(&client->wq_lock); + /* + * If GuC firmware performs an engine reset while that engine had + * a preemption pending, it will set the terminated attribute bit + * on our preemption stage descriptor. GuC firmware retains all + * pending work items for a high-priority GuC client, unlike the + * normal-priority GuC client where work items are dropped. It + * wants to make sure the preempt-to-idle work doesn't run when + * scheduling resumes, and uses this bit to inform its scheduler + * and presumably us as well. Our job is to clear it for the next + * preemption after reset, otherwise that and future preemptions + * will never complete. We'll just clear it every time. + */ + stage_desc->attribute &= ~GUC_STAGE_DESC_ATTR_TERMINATED; + data[0] = INTEL_GUC_ACTION_REQUEST_PREEMPTION; data[1] = client->stage_id; data[2] = INTEL_GUC_PREEMPT_OPTION_DROP_WORK_Q | -- GitLab From cb20a3c056b04044c5e92c6d5d4485ba46ab4f72 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Mon, 30 Oct 2017 11:56:14 -0700 Subject: [PATCH 0100/1471] drm/i915/guc: Rename the function that resets the GuC intel_guc_reset sounds more like the microcontroller is the one performing a reset, while in this case is the opposite. intel_reset_guc not only makes it clearer, it follows the other intel_reset functions available. v2: Print error message in English (Tvrtko). Cc: Tvrtko Ursulin Signed-off-by: Michel Thierry Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171030185616.32836-2-michel.thierry@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.h | 2 +- drivers/gpu/drm/i915/intel_uc.c | 4 ++-- drivers/gpu/drm/i915/intel_uncore.c | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index c10fece58e867..05425bb7565b6 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3324,7 +3324,7 @@ extern int i915_reset_engine(struct intel_engine_cs *engine, unsigned int flags); extern bool intel_has_reset_engine(struct drm_i915_private *dev_priv); -extern int intel_guc_reset(struct drm_i915_private *dev_priv); +extern int intel_reset_guc(struct drm_i915_private *dev_priv); extern void intel_engine_init_hangcheck(struct intel_engine_cs *engine); extern void intel_hangcheck_init(struct drm_i915_private *dev_priv); extern unsigned long i915_chipset_val(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_uc.c b/drivers/gpu/drm/i915/intel_uc.c index 25bd162f38d25..aec295470e0d9 100644 --- a/drivers/gpu/drm/i915/intel_uc.c +++ b/drivers/gpu/drm/i915/intel_uc.c @@ -33,9 +33,9 @@ static int __intel_uc_reset_hw(struct drm_i915_private *dev_priv) int ret; u32 guc_status; - ret = intel_guc_reset(dev_priv); + ret = intel_reset_guc(dev_priv); if (ret) { - DRM_ERROR("GuC reset failed, ret = %d\n", ret); + DRM_ERROR("Failed to reset GuC, ret = %d\n", ret); return ret; } diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 96ee6b2754bed..d629b2e40013d 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1803,7 +1803,7 @@ bool intel_has_reset_engine(struct drm_i915_private *dev_priv) i915_modparams.reset >= 2); } -int intel_guc_reset(struct drm_i915_private *dev_priv) +int intel_reset_guc(struct drm_i915_private *dev_priv) { int ret; -- GitLab From 6acbea89efbdc6aef56bc6f364d4d8b72a5fc145 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Tue, 31 Oct 2017 15:53:09 -0700 Subject: [PATCH 0101/1471] drm/i915/guc: Add support for reset engine using GuC commands This patch adds per engine reset and recovery (TDR) support when GuC is used to submit workloads to GPU. In the case of i915 directly submission to ELSP, driver manages hang detection, recovery and resubmission. With GuC submission these tasks are shared between driver and GuC. i915 is still responsible for detecting a hang, and when it does it only requests GuC to reset that Engine. GuC internally manages acquiring forcewake and idling the engine before resetting it. Once the reset is successful, i915 takes over again and handles the resubmission. The scheduler in i915 knows which requests are pending so after resetting a engine, pending workloads/requests are resubmitted again. v2: s/i915_guc_request_engine_reset/i915_guc_reset_engine/ to match the non-guc function names. v3: Removed debug message about engine restarting from which request, since the new baseline do it regardless of submission mode. (Chris) v4: Rebase. v5: Do not pass unnecessary reporting flags to the fw (Jeff); tasklet_schedule(&execlists->irq_tasklet) handles the resubmit; rebase. v6: Rename the existing reset engine function and share a similar interface between guc and non-guc paths (Chris). Signed-off-by: Michel Thierry Cc: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171031225309.10888-1-michel.thierry@intel.com Reviewed-by: Jeff McGee Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.c | 14 ++++++++++++-- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/intel_guc.c | 24 ++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_guc_fwif.h | 1 + drivers/gpu/drm/i915/intel_uncore.c | 5 ----- 5 files changed, 39 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index af745749509c0..e7e9e061073b7 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1950,6 +1950,12 @@ void i915_reset(struct drm_i915_private *i915, unsigned int flags) goto finish; } +static inline int intel_gt_reset_engine(struct drm_i915_private *dev_priv, + struct intel_engine_cs *engine) +{ + return intel_gpu_reset(dev_priv, intel_engine_flag(engine)); +} + /** * i915_reset_engine - reset GPU engine to recover from a hang * @engine: engine to reset @@ -1984,10 +1990,14 @@ int i915_reset_engine(struct intel_engine_cs *engine, unsigned int flags) goto out; } - ret = intel_gpu_reset(engine->i915, intel_engine_flag(engine)); + if (!engine->i915->guc.execbuf_client) + ret = intel_gt_reset_engine(engine->i915, engine); + else + ret = intel_guc_reset_engine(&engine->i915->guc, engine); if (ret) { /* If we fail here, we expect to fallback to a global reset */ - DRM_DEBUG_DRIVER("Failed to reset %s, ret=%d\n", + DRM_DEBUG_DRIVER("%sFailed to reset %s, ret=%d\n", + engine->i915->guc.execbuf_client ? "GuC " : "", engine->name, ret); goto out; } diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 05425bb7565b6..72bb5b51035a0 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3325,6 +3325,8 @@ extern int i915_reset_engine(struct intel_engine_cs *engine, extern bool intel_has_reset_engine(struct drm_i915_private *dev_priv); extern int intel_reset_guc(struct drm_i915_private *dev_priv); +extern int intel_guc_reset_engine(struct intel_guc *guc, + struct intel_engine_cs *engine); extern void intel_engine_init_hangcheck(struct intel_engine_cs *engine); extern void intel_hangcheck_init(struct drm_i915_private *dev_priv); extern unsigned long i915_chipset_val(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_guc.c b/drivers/gpu/drm/i915/intel_guc.c index f74d50fdaeb07..9678630a1c705 100644 --- a/drivers/gpu/drm/i915/intel_guc.c +++ b/drivers/gpu/drm/i915/intel_guc.c @@ -24,6 +24,7 @@ #include "intel_guc.h" #include "i915_drv.h" +#include "i915_guc_submission.h" static void gen8_guc_raise_irq(struct intel_guc *guc) { @@ -283,6 +284,29 @@ int intel_guc_suspend(struct drm_i915_private *dev_priv) return intel_guc_send(guc, data, ARRAY_SIZE(data)); } +/** + * intel_guc_reset_engine() - ask GuC to reset an engine + * @guc: intel_guc structure + * @engine: engine to be reset + */ +int intel_guc_reset_engine(struct intel_guc *guc, + struct intel_engine_cs *engine) +{ + u32 data[7]; + + GEM_BUG_ON(!guc->execbuf_client); + + data[0] = INTEL_GUC_ACTION_REQUEST_ENGINE_RESET; + data[1] = engine->guc_id; + data[2] = 0; + data[3] = 0; + data[4] = 0; + data[5] = guc->execbuf_client->stage_id; + data[6] = guc_ggtt_offset(guc->shared_data); + + return intel_guc_send(guc, data, ARRAY_SIZE(data)); +} + /** * intel_guc_resume() - notify GuC resuming from suspend state * @dev_priv: i915 device private diff --git a/drivers/gpu/drm/i915/intel_guc_fwif.h b/drivers/gpu/drm/i915/intel_guc_fwif.h index e24dbec2ead8c..6a10aa6f04d35 100644 --- a/drivers/gpu/drm/i915/intel_guc_fwif.h +++ b/drivers/gpu/drm/i915/intel_guc_fwif.h @@ -574,6 +574,7 @@ struct guc_shared_ctx_data { enum intel_guc_action { INTEL_GUC_ACTION_DEFAULT = 0x0, INTEL_GUC_ACTION_REQUEST_PREEMPTION = 0x2, + INTEL_GUC_ACTION_REQUEST_ENGINE_RESET = 0x3, INTEL_GUC_ACTION_SAMPLE_FORCEWAKE = 0x6, INTEL_GUC_ACTION_ALLOCATE_DOORBELL = 0x10, INTEL_GUC_ACTION_DEALLOCATE_DOORBELL = 0x20, diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index d629b2e40013d..0564d87ad416e 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1792,14 +1792,9 @@ bool intel_has_gpu_reset(struct drm_i915_private *dev_priv) return intel_get_gpu_reset(dev_priv) != NULL; } -/* - * When GuC submission is enabled, GuC manages ELSP and can initiate the - * engine reset too. For now, fall back to full GPU reset if it is enabled. - */ bool intel_has_reset_engine(struct drm_i915_private *dev_priv) { return (dev_priv->info.has_reset_engine && - !dev_priv->guc.execbuf_client && i915_modparams.reset >= 2); } -- GitLab From 7fb50b957be10e631f649eca8bd66a0b89795149 Mon Sep 17 00:00:00 2001 From: Liviu Dudau Date: Wed, 1 Nov 2017 14:44:58 +0000 Subject: [PATCH 0102/1471] drm/drm_vma_manager.c: Remove useless goto statement Commit db2395eccf08i ("drm: Convert drm_vma_manager to embedded interval-tree in drm_mm") removed a line in drm_vma_offset_add() function that makes checking the result of calling drm_mm_insert_node() and the goto call redundant. Rework the function (as suggested by Chris Wilson) to eliminate the need for the goto and associated label. v2: rewrite function to remove all goto statements. Fixes: db2395eccf08i ("drm: Convert drm_vma_manager to embedded interval-tree in drm_mm") Cc: Chris Wilson Signed-off-by: Liviu Dudau Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171101144458.5353-1-Liviu.Dudau@arm.com --- drivers/gpu/drm/drm_vma_manager.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/drm_vma_manager.c b/drivers/gpu/drm/drm_vma_manager.c index 28f1226576f8c..23c749c05b5aa 100644 --- a/drivers/gpu/drm/drm_vma_manager.c +++ b/drivers/gpu/drm/drm_vma_manager.c @@ -203,21 +203,16 @@ EXPORT_SYMBOL(drm_vma_offset_lookup_locked); int drm_vma_offset_add(struct drm_vma_offset_manager *mgr, struct drm_vma_offset_node *node, unsigned long pages) { - int ret; + int ret = 0; write_lock(&mgr->vm_lock); - if (drm_mm_node_allocated(&node->vm_node)) { - ret = 0; - goto out_unlock; - } + if (!drm_mm_node_allocated(&node->vm_node)) + ret = drm_mm_insert_node(&mgr->vm_addr_space_mm, + &node->vm_node, pages); - ret = drm_mm_insert_node(&mgr->vm_addr_space_mm, &node->vm_node, pages); - if (ret) - goto out_unlock; - -out_unlock: write_unlock(&mgr->vm_lock); + return ret; } EXPORT_SYMBOL(drm_vma_offset_add); -- GitLab From 7130630323c562597191653560963e61c5bd0f57 Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Thu, 2 Nov 2017 11:48:36 +0200 Subject: [PATCH 0103/1471] drm/i915: Use fallback forcewake if primary ack missing There is a possibility on gen9 hardware to miss the forcewake ack message. The recommended workaround is to use another free bit and toggle it until original bit is successfully acknowledged. Some future gen9 revs might or might not fix the underlying issue but using fallback forcewake bit dance can be considered as harmless: without the ack timeout we never reach the fallback bit forcewake. Thus as of now we adopt a blanket approach for all gen9 and leave the bypassing the fallback bit approach for future patches if corresponding hw revisions do appear. Commit 83e3337204b2 ("drm/i915: Increase maximum polling time to 50ms for forcewake request/clear ack") did increase the forcewake timeout. If the issue was a delayed ack, future work could include finding a suitable timeout value both for primary ack and reserve toggle to reduce the worst case latency. v2: use bit 15, naming, comment (Chris), only wait fallback ack v3: fix return on fallback, backoff after fallback write (Chris) v4: udelay on first pass, grammar (Chris) v4: s/reserve/fallback References: HSDES #1604254524 References: https://bugs.freedesktop.org/show_bug.cgi?id=102051 Cc: Chris Wilson Cc: Rodrigo Vivi Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Signed-off-by: Mika Kuoppala Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171102094836.2506-1-mika.kuoppala@linux.intel.com --- drivers/gpu/drm/i915/i915_reg.h | 5 +- drivers/gpu/drm/i915/intel_uncore.c | 137 ++++++++++++++++++++++++++-- 2 files changed, 130 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 8c775e96b4e4f..f0f8f60596528 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -7774,8 +7774,9 @@ enum { #define FORCEWAKE_ACK_MEDIA_GEN9 _MMIO(0x0D88) #define FORCEWAKE_ACK_RENDER_GEN9 _MMIO(0x0D84) #define FORCEWAKE_ACK_BLITTER_GEN9 _MMIO(0x130044) -#define FORCEWAKE_KERNEL 0x1 -#define FORCEWAKE_USER 0x2 +#define FORCEWAKE_KERNEL BIT(0) +#define FORCEWAKE_USER BIT(1) +#define FORCEWAKE_KERNEL_FALLBACK BIT(15) #define FORCEWAKE_MT_ACK _MMIO(0x130040) #define ECOBUS _MMIO(0xa180) #define FORCEWAKE_MT_ENABLE (1<<5) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 0564d87ad416e..0a6d952a2df14 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -69,17 +69,104 @@ fw_domain_arm_timer(struct intel_uncore_forcewake_domain *d) HRTIMER_MODE_REL); } +static inline int +__wait_for_ack(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d, + const u32 ack, + const u32 value) +{ + return wait_for_atomic((__raw_i915_read32(i915, d->reg_ack) & ack) == value, + FORCEWAKE_ACK_TIMEOUT_MS); +} + +static inline int +wait_ack_clear(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d, + const u32 ack) +{ + return __wait_for_ack(i915, d, ack, 0); +} + +static inline int +wait_ack_set(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d, + const u32 ack) +{ + return __wait_for_ack(i915, d, ack, ack); +} + static inline void fw_domain_wait_ack_clear(const struct drm_i915_private *i915, const struct intel_uncore_forcewake_domain *d) { - if (wait_for_atomic((__raw_i915_read32(i915, d->reg_ack) & - FORCEWAKE_KERNEL) == 0, - FORCEWAKE_ACK_TIMEOUT_MS)) + if (wait_ack_clear(i915, d, FORCEWAKE_KERNEL)) DRM_ERROR("%s: timed out waiting for forcewake ack to clear.\n", intel_uncore_forcewake_domain_to_str(d->id)); } +enum ack_type { + ACK_CLEAR = 0, + ACK_SET +}; + +static int +fw_domain_wait_ack_with_fallback(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d, + const enum ack_type type) +{ + const u32 ack_bit = FORCEWAKE_KERNEL; + const u32 value = type == ACK_SET ? ack_bit : 0; + unsigned int pass; + bool ack_detected; + + /* + * There is a possibility of driver's wake request colliding + * with hardware's own wake requests and that can cause + * hardware to not deliver the driver's ack message. + * + * Use a fallback bit toggle to kick the gpu state machine + * in the hope that the original ack will be delivered along with + * the fallback ack. + * + * This workaround is described in HSDES #1604254524 + */ + + pass = 1; + do { + wait_ack_clear(i915, d, FORCEWAKE_KERNEL_FALLBACK); + + __raw_i915_write32(i915, d->reg_set, + _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL_FALLBACK)); + /* Give gt some time to relax before the polling frenzy */ + udelay(10 * pass); + wait_ack_set(i915, d, FORCEWAKE_KERNEL_FALLBACK); + + ack_detected = (__raw_i915_read32(i915, d->reg_ack) & ack_bit) == value; + + __raw_i915_write32(i915, d->reg_set, + _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL_FALLBACK)); + } while (!ack_detected && pass++ < 10); + + DRM_DEBUG_DRIVER("%s had to use fallback to %s ack, 0x%x (passes %u)\n", + intel_uncore_forcewake_domain_to_str(d->id), + type == ACK_SET ? "set" : "clear", + __raw_i915_read32(i915, d->reg_ack), + pass); + + return ack_detected ? 0 : -ETIMEDOUT; +} + +static inline void +fw_domain_wait_ack_clear_fallback(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d) +{ + if (likely(!wait_ack_clear(i915, d, FORCEWAKE_KERNEL))) + return; + + if (fw_domain_wait_ack_with_fallback(i915, d, ACK_CLEAR)) + fw_domain_wait_ack_clear(i915, d); +} + static inline void fw_domain_get(struct drm_i915_private *i915, const struct intel_uncore_forcewake_domain *d) @@ -88,16 +175,25 @@ fw_domain_get(struct drm_i915_private *i915, } static inline void -fw_domain_wait_ack(const struct drm_i915_private *i915, - const struct intel_uncore_forcewake_domain *d) +fw_domain_wait_ack_set(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d) { - if (wait_for_atomic((__raw_i915_read32(i915, d->reg_ack) & - FORCEWAKE_KERNEL), - FORCEWAKE_ACK_TIMEOUT_MS)) + if (wait_ack_set(i915, d, FORCEWAKE_KERNEL)) DRM_ERROR("%s: timed out waiting for forcewake ack request.\n", intel_uncore_forcewake_domain_to_str(d->id)); } +static inline void +fw_domain_wait_ack_set_fallback(const struct drm_i915_private *i915, + const struct intel_uncore_forcewake_domain *d) +{ + if (likely(!wait_ack_set(i915, d, FORCEWAKE_KERNEL))) + return; + + if (fw_domain_wait_ack_with_fallback(i915, d, ACK_SET)) + fw_domain_wait_ack_set(i915, d); +} + static inline void fw_domain_put(const struct drm_i915_private *i915, const struct intel_uncore_forcewake_domain *d) @@ -119,7 +215,27 @@ fw_domains_get(struct drm_i915_private *i915, enum forcewake_domains fw_domains) } for_each_fw_domain_masked(d, fw_domains, i915, tmp) - fw_domain_wait_ack(i915, d); + fw_domain_wait_ack_set(i915, d); + + i915->uncore.fw_domains_active |= fw_domains; +} + +static void +fw_domains_get_with_fallback(struct drm_i915_private *i915, + enum forcewake_domains fw_domains) +{ + struct intel_uncore_forcewake_domain *d; + unsigned int tmp; + + GEM_BUG_ON(fw_domains & ~i915->uncore.fw_domains); + + for_each_fw_domain_masked(d, fw_domains, i915, tmp) { + fw_domain_wait_ack_clear_fallback(i915, d); + fw_domain_get(i915, d); + } + + for_each_fw_domain_masked(d, fw_domains, i915, tmp) + fw_domain_wait_ack_set_fallback(i915, d); i915->uncore.fw_domains_active |= fw_domains; } @@ -1142,7 +1258,8 @@ static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv) } if (INTEL_GEN(dev_priv) >= 9) { - dev_priv->uncore.funcs.force_wake_get = fw_domains_get; + dev_priv->uncore.funcs.force_wake_get = + fw_domains_get_with_fallback; dev_priv->uncore.funcs.force_wake_put = fw_domains_put; fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER, FORCEWAKE_RENDER_GEN9, -- GitLab From 50f365cde4ffb5ae70c3f02384bbb46698aba65c Mon Sep 17 00:00:00 2001 From: Boris BREZILLON Date: Wed, 1 Nov 2017 10:57:31 +0100 Subject: [PATCH 0104/1471] drm/vc4: Fix wrong printk format in vc4_bo_stats_debugfs() vc4->purgeable.size and vc4->purgeable.purged_size are size_t fields and should be printed with a %zd specifier. Fixes: b9f19259b84d ("drm/vc4: Add the DRM_IOCTL_VC4_GEM_MADVISE ioctl") Signed-off-by: Boris Brezillon Reviewed-by: Gustavo Padovan Reviewed-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/20171101095731.14878-1-boris.brezillon@free-electrons.com --- drivers/gpu/drm/vc4/vc4_bo.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/vc4/vc4_bo.c b/drivers/gpu/drm/vc4/vc4_bo.c index 3c3d112369109..4ae45d7dac428 100644 --- a/drivers/gpu/drm/vc4/vc4_bo.c +++ b/drivers/gpu/drm/vc4/vc4_bo.c @@ -88,11 +88,11 @@ int vc4_bo_stats_debugfs(struct seq_file *m, void *unused) mutex_lock(&vc4->purgeable.lock); if (vc4->purgeable.num) - seq_printf(m, "%30s: %6dkb BOs (%d)\n", "userspace BO cache", + seq_printf(m, "%30s: %6zdkb BOs (%d)\n", "userspace BO cache", vc4->purgeable.size / 1024, vc4->purgeable.num); if (vc4->purgeable.purged_num) - seq_printf(m, "%30s: %6dkb BOs (%d)\n", "total purged BO", + seq_printf(m, "%30s: %6zdkb BOs (%d)\n", "total purged BO", vc4->purgeable.purged_size / 1024, vc4->purgeable.purged_num); mutex_unlock(&vc4->purgeable.lock); -- GitLab From 820c5bbbf418fba41149fdeb870d623e21be2463 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 1 Nov 2017 20:21:49 +0000 Subject: [PATCH 0105/1471] drm/i915: Flush the irq and tasklets before asserting engine is idle Before we assert that the engine is idle, make sure we flush any residual tasklet. After that point, if the engine is not idle, more work may be queued despite us trying to park the engine and go to sleep. References: https://bugs.freedesktop.org/show_bug.cgi?id=103479 Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171101202149.32493-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/intel_engine_cs.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 6895a90af0083..ddbe5c9bf45ab 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1614,6 +1614,10 @@ void intel_engines_park(struct drm_i915_private *i915) enum intel_engine_id id; for_each_engine(engine, i915, id) { + /* Flush the residual irq tasklets first. */ + intel_engine_disarm_breadcrumbs(engine); + tasklet_kill(&engine->execlists.irq_tasklet); + /* * We are committed now to parking the engines, make sure there * will be no more interrupts arriving later and the engines @@ -1630,9 +1634,6 @@ void intel_engines_park(struct drm_i915_private *i915) if (engine->park) engine->park(engine); - intel_engine_disarm_breadcrumbs(engine); - tasklet_kill(&engine->execlists.irq_tasklet); - i915_gem_batch_pool_fini(&engine->batch_pool); engine->execlists.no_priolist = false; } -- GitLab From 21be91549c6b6792a37cfe9c0409d0635df220f4 Mon Sep 17 00:00:00 2001 From: Liviu Dudau Date: Wed, 1 Nov 2017 14:04:36 +0000 Subject: [PATCH 0106/1471] drm: Spelling fixes Minor spelling fix for 'monster' and replace 'on' with 'own' in comments. Signed-off-by: Liviu Dudau Reviewed-by: Alex Deucher Signed-off-by: Gustavo Padovan Link: https://patchwork.freedesktop.org/patch/msgid/20171101140436.2743-1-Liviu.Dudau@arm.com --- drivers/gpu/drm/drm_atomic.c | 2 +- drivers/gpu/drm/drm_mm.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index 6c9c4a8e09af7..8ec7d7bade8a3 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -1808,7 +1808,7 @@ int drm_atomic_debugfs_init(struct drm_minor *minor) #endif /* - * The big monstor ioctl + * The big monster ioctl */ static struct drm_pending_vblank_event *create_vblank_event( diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c index 61a1c8ea74bc5..eb86bc3f753b3 100644 --- a/drivers/gpu/drm/drm_mm.c +++ b/drivers/gpu/drm/drm_mm.c @@ -92,7 +92,7 @@ * some basic allocator dumpers for debugging. * * Note that this range allocator is not thread-safe, drivers need to protect - * modifications with their on locking. The idea behind this is that for a full + * modifications with their own locking. The idea behind this is that for a full * memory manager additional data needs to be protected anyway, hence internal * locking would be fully redundant. */ -- GitLab From c138782d895d52aa3831aa8c468b33d68fa5c954 Mon Sep 17 00:00:00 2001 From: Liviu Dudau Date: Wed, 1 Nov 2017 14:06:30 +0000 Subject: [PATCH 0107/1471] dma-buf: Cleanup comments on dma_buf_map_attachment() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Mappings need to be unmapped by calling dma_buf_unmap_attachment() and not by calling again dma_buf_map_attachment(). Also fix some spelling mistakes. Signed-off-by: Liviu Dudau Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Gustavo Padovan Link: https://patchwork.freedesktop.org/patch/msgid/20171101140630.2884-1-Liviu.Dudau@arm.com --- drivers/dma-buf/dma-buf.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c index bc1cb284111cb..1792385405f0e 100644 --- a/drivers/dma-buf/dma-buf.c +++ b/drivers/dma-buf/dma-buf.c @@ -351,13 +351,13 @@ static inline int is_dma_buf_file(struct file *file) * * 2. Userspace passes this file-descriptors to all drivers it wants this buffer * to share with: First the filedescriptor is converted to a &dma_buf using - * dma_buf_get(). The the buffer is attached to the device using + * dma_buf_get(). Then the buffer is attached to the device using * dma_buf_attach(). * * Up to this stage the exporter is still free to migrate or reallocate the * backing storage. * - * 3. Once the buffer is attached to all devices userspace can inniate DMA + * 3. Once the buffer is attached to all devices userspace can initiate DMA * access to the shared buffer. In the kernel this is done by calling * dma_buf_map_attachment() and dma_buf_unmap_attachment(). * @@ -617,7 +617,7 @@ EXPORT_SYMBOL_GPL(dma_buf_detach); * Returns sg_table containing the scatterlist to be returned; returns ERR_PTR * on error. May return -EINTR if it is interrupted by a signal. * - * A mapping must be unmapped again using dma_buf_map_attachment(). Note that + * A mapping must be unmapped by using dma_buf_unmap_attachment(). Note that * the underlying backing storage is pinned for as long as a mapping exists, * therefore users/importers should not hold onto a mapping for undue amounts of * time. -- GitLab From 0a2adb02d71e68e7e00913394ca2d9c5d9fe5eb7 Mon Sep 17 00:00:00 2001 From: Liviu Dudau Date: Wed, 1 Nov 2017 14:04:45 +0000 Subject: [PATCH 0108/1471] drm/drm_mm.h: Fix the name of the referenced function in comment MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_mm_insert_node_generic() is a simplified version of drm_mm_insert_node_in_range(), update comment to reflect correct function name. Signed-off-by: Liviu Dudau Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Gustavo Padovan Link: https://patchwork.freedesktop.org/patch/msgid/20171101140445.2798-1-Liviu.Dudau@arm.com --- include/drm/drm_mm.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/drm/drm_mm.h b/include/drm/drm_mm.h index 8d10fc97801ce..101f566ae43d0 100644 --- a/include/drm/drm_mm.h +++ b/include/drm/drm_mm.h @@ -386,7 +386,7 @@ int drm_mm_insert_node_in_range(struct drm_mm *mm, * @color: opaque tag value to use for this node * @mode: fine-tune the allocation search and placement * - * This is a simplified version of drm_mm_insert_node_in_range_generic() with no + * This is a simplified version of drm_mm_insert_node_in_range() with no * range restrictions applied. * * The preallocated node must be cleared to 0. -- GitLab From 0338f0d0d70a26adc2236148bb5a78f1fa8c7f99 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 20:35:33 +0200 Subject: [PATCH 0109/1471] drm: Print the fb dimensions when the src coord check fails MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When debugging bad plane source coordinates it helps to have an idea of what the framebuffer dimensions are. Signed-off-by: Ville Syrjälä Reviewed-by: Daniel Vetter Signed-off-by: Gustavo Padovan Link: https://patchwork.freedesktop.org/patch/msgid/20171101183533.28466-1-ville.syrjala@linux.intel.com --- drivers/gpu/drm/drm_atomic.c | 5 +++-- drivers/gpu/drm/drm_framebuffer.c | 5 +++-- 2 files changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index 8ec7d7bade8a3..fd96c3dad2fbd 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -907,11 +907,12 @@ static int drm_atomic_plane_check(struct drm_plane *plane, state->src_h > fb_height || state->src_y > fb_height - state->src_h) { DRM_DEBUG_ATOMIC("Invalid source coordinates " - "%u.%06ux%u.%06u+%u.%06u+%u.%06u\n", + "%u.%06ux%u.%06u+%u.%06u+%u.%06u (fb %ux%u)\n", state->src_w >> 16, ((state->src_w & 0xffff) * 15625) >> 10, state->src_h >> 16, ((state->src_h & 0xffff) * 15625) >> 10, state->src_x >> 16, ((state->src_x & 0xffff) * 15625) >> 10, - state->src_y >> 16, ((state->src_y & 0xffff) * 15625) >> 10); + state->src_y >> 16, ((state->src_y & 0xffff) * 15625) >> 10, + state->fb->width, state->fb->height); return -ENOSPC; } diff --git a/drivers/gpu/drm/drm_framebuffer.c b/drivers/gpu/drm/drm_framebuffer.c index af279844d7ceb..40e6565105acb 100644 --- a/drivers/gpu/drm/drm_framebuffer.c +++ b/drivers/gpu/drm/drm_framebuffer.c @@ -78,11 +78,12 @@ int drm_framebuffer_check_src_coords(uint32_t src_x, uint32_t src_y, src_h > fb_height || src_y > fb_height - src_h) { DRM_DEBUG_KMS("Invalid source coordinates " - "%u.%06ux%u.%06u+%u.%06u+%u.%06u\n", + "%u.%06ux%u.%06u+%u.%06u+%u.%06u (fb %ux%u)\n", src_w >> 16, ((src_w & 0xffff) * 15625) >> 10, src_h >> 16, ((src_h & 0xffff) * 15625) >> 10, src_x >> 16, ((src_x & 0xffff) * 15625) >> 10, - src_y >> 16, ((src_y & 0xffff) * 15625) >> 10); + src_y >> 16, ((src_y & 0xffff) * 15625) >> 10, + fb->width, fb->height); return -ENOSPC; } -- GitLab From 49e43ef7c6e533c110a7ff8fa8096291ec1c0e0e Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Thu, 2 Nov 2017 12:18:27 +0000 Subject: [PATCH 0110/1471] drm/i915: ensure oa config uuid is null terminated Because dev_priv is 0-ed it's not currently an issue, but since we have dev_priv->perf.oa.test_config.uuid size at uuid + 1, we could just copy the null character. v2: Use strlcpy instead of strncpy (Chris) Signed-off-by: Lionel Landwerlin Cc: Rodrigo Vivi Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171102121827.436-1-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_oa_bdw.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_bxt.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_cflgt2.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_chv.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_glk.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_hsw.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_kblgt2.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_kblgt3.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_sklgt2.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_sklgt3.c | 4 ++-- drivers/gpu/drm/i915/i915_oa_sklgt4.c | 4 ++-- 11 files changed, 22 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_oa_bdw.c b/drivers/gpu/drm/i915/i915_oa_bdw.c index abdf4d0abcceb..4abd2e8b50839 100644 --- a/drivers/gpu/drm/i915/i915_oa_bdw.c +++ b/drivers/gpu/drm/i915/i915_oa_bdw.c @@ -85,9 +85,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_bdw(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "d6de6f55-e526-4f79-a6a6-d7315c09044e", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_bxt.c b/drivers/gpu/drm/i915/i915_oa_bxt.c index b69b900de0fe1..cb6f304ec16a3 100644 --- a/drivers/gpu/drm/i915/i915_oa_bxt.c +++ b/drivers/gpu/drm/i915/i915_oa_bxt.c @@ -83,9 +83,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_bxt(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "5ee72f5c-092f-421e-8b70-225f7c3e9612", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_cflgt2.c b/drivers/gpu/drm/i915/i915_oa_cflgt2.c index 368c87d7ee9a9..8641ae30e343a 100644 --- a/drivers/gpu/drm/i915/i915_oa_cflgt2.c +++ b/drivers/gpu/drm/i915/i915_oa_cflgt2.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_cflgt2(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "74fb4902-d3d3-4237-9e90-cbdc68d0a446", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_chv.c b/drivers/gpu/drm/i915/i915_oa_chv.c index 322a3f94cd167..556febb2c3c87 100644 --- a/drivers/gpu/drm/i915/i915_oa_chv.c +++ b/drivers/gpu/drm/i915/i915_oa_chv.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_chv(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "4a534b07-cba3-414d-8d60-874830e883aa", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_glk.c b/drivers/gpu/drm/i915/i915_oa_glk.c index 4ee527e4c9265..971db587957c6 100644 --- a/drivers/gpu/drm/i915/i915_oa_glk.c +++ b/drivers/gpu/drm/i915/i915_oa_glk.c @@ -83,9 +83,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_glk(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "dd3fd789-e783-4204-8cd0-b671bbccb0cf", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_hsw.c b/drivers/gpu/drm/i915/i915_oa_hsw.c index 56b03773bb9d7..434a9b96d7abd 100644 --- a/drivers/gpu/drm/i915/i915_oa_hsw.c +++ b/drivers/gpu/drm/i915/i915_oa_hsw.c @@ -113,9 +113,9 @@ show_render_basic_id(struct device *kdev, struct device_attribute *attr, char *b void i915_perf_load_test_config_hsw(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "403d8832-1a27-4aa6-a64e-f5389ce7b212", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_render_basic; diff --git a/drivers/gpu/drm/i915/i915_oa_kblgt2.c b/drivers/gpu/drm/i915/i915_oa_kblgt2.c index b6e7cc7741369..2fa98a40bbc84 100644 --- a/drivers/gpu/drm/i915/i915_oa_kblgt2.c +++ b/drivers/gpu/drm/i915/i915_oa_kblgt2.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_kblgt2(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "baa3c7e4-52b6-4b85-801e-465a94b746dd", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_kblgt3.c b/drivers/gpu/drm/i915/i915_oa_kblgt3.c index 5576afdd9a7e4..f3cb6679a1bcf 100644 --- a/drivers/gpu/drm/i915/i915_oa_kblgt3.c +++ b/drivers/gpu/drm/i915/i915_oa_kblgt3.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_kblgt3(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "f1792f32-6db2-4b50-b4b2-557128f1688d", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_sklgt2.c b/drivers/gpu/drm/i915/i915_oa_sklgt2.c index 890d55879946a..bf8b8cd8a50d6 100644 --- a/drivers/gpu/drm/i915/i915_oa_sklgt2.c +++ b/drivers/gpu/drm/i915/i915_oa_sklgt2.c @@ -83,9 +83,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_sklgt2(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "1651949f-0ac0-4cb1-a06f-dafd74a407d1", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_sklgt3.c b/drivers/gpu/drm/i915/i915_oa_sklgt3.c index 85e51addf86aa..ae534c7c8135e 100644 --- a/drivers/gpu/drm/i915/i915_oa_sklgt3.c +++ b/drivers/gpu/drm/i915/i915_oa_sklgt3.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_sklgt3(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "2b985803-d3c9-4629-8a4f-634bfecba0e8", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; diff --git a/drivers/gpu/drm/i915/i915_oa_sklgt4.c b/drivers/gpu/drm/i915/i915_oa_sklgt4.c index bce031ee4445f..817fba2d82df3 100644 --- a/drivers/gpu/drm/i915/i915_oa_sklgt4.c +++ b/drivers/gpu/drm/i915/i915_oa_sklgt4.c @@ -84,9 +84,9 @@ show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) void i915_perf_load_test_config_sklgt4(struct drm_i915_private *dev_priv) { - strncpy(dev_priv->perf.oa.test_config.uuid, + strlcpy(dev_priv->perf.oa.test_config.uuid, "882fa433-1f4a-4a67-a962-c741888fe5f5", - UUID_STRING_LEN); + sizeof(dev_priv->perf.oa.test_config.uuid)); dev_priv->perf.oa.test_config.id = 1; dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; -- GitLab From ebcaa1ff8b59097805d548fe7a676f194625c033 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 31 Oct 2017 10:23:25 +0000 Subject: [PATCH 0111/1471] drm/i915: Reject unknown syncobj flags We have to reject unknown flags for uAPI considerations, and also because the curent implementation limits their i915 storage space to two bits. v2: (Chris Wilson) * Fix fail in ABI check. * Added unknown flags and BUILD_BUG_ON. v3: * Use ARCH_KMALLOC_MINALIGN instead of alignof. (Chris Wilson) Signed-off-by: Tvrtko Ursulin Fixes: cf6e7bac6357 ("drm/i915: Add support for drm syncobjs") Cc: Jason Ekstrand Cc: Chris Wilson Cc: Jani Nikula Cc: Joonas Lahtinen Cc: Rodrigo Vivi Cc: David Airlie Cc: intel-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171031102326.9738-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 ++++++++ include/uapi/drm/i915_drm.h | 1 + 2 files changed, 9 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 5b8213b1a8c7e..435ed95df144c 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -2104,6 +2104,11 @@ get_fence_array(struct drm_i915_gem_execbuffer2 *args, goto err; } + if (fence.flags & __I915_EXEC_FENCE_UNKNOWN_FLAGS) { + err = -EINVAL; + goto err; + } + syncobj = drm_syncobj_find(file, fence.handle); if (!syncobj) { DRM_DEBUG("Invalid syncobj handle provided\n"); @@ -2111,6 +2116,9 @@ get_fence_array(struct drm_i915_gem_execbuffer2 *args, goto err; } + BUILD_BUG_ON(~(ARCH_KMALLOC_MINALIGN - 1) & + ~__I915_EXEC_FENCE_UNKNOWN_FLAGS); + fences[n] = ptr_pack_bits(syncobj, fence.flags, 2); } diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 125bde7d95045..ac3c6503ca27f 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -839,6 +839,7 @@ struct drm_i915_gem_exec_fence { #define I915_EXEC_FENCE_WAIT (1<<0) #define I915_EXEC_FENCE_SIGNAL (1<<1) +#define __I915_EXEC_FENCE_UNKNOWN_FLAGS (-(I915_EXEC_FENCE_SIGNAL << 1)) __u32 flags; }; -- GitLab From 2d7514152b1c82c81a363f8a171237567ebb7e66 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 3 Nov 2017 09:05:38 +0000 Subject: [PATCH 0112/1471] drm/i915: Warn in debug builds of incorrect usages of ptr_pack_bits GEM_BUG_ON if the packed bits do not fit into the specified width. v2: Avoid using the macro argument twice. v3: Drop unnecessary braces. (Joonas) Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson (v1) Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171103090538.14474-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_utils.h | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h index af3d7cc53fa13..8d07764887ec2 100644 --- a/drivers/gpu/drm/i915/i915_utils.h +++ b/drivers/gpu/drm/i915/i915_utils.h @@ -83,8 +83,11 @@ (typeof(ptr))(__v & -BIT(n)); \ }) -#define ptr_pack_bits(ptr, bits, n) \ - ((typeof(ptr))((unsigned long)(ptr) | (bits))) +#define ptr_pack_bits(ptr, bits, n) ({ \ + unsigned long __bits = (bits); \ + GEM_BUG_ON(__bits & -BIT(n)); \ + ((typeof(ptr))((unsigned long)(ptr) | __bits)); \ +}) #define page_mask_bits(ptr) ptr_mask_bits(ptr, PAGE_SHIFT) #define page_unmask_bits(ptr) ptr_unmask_bits(ptr, PAGE_SHIFT) -- GitLab From 136109c67f67ebe449032950ddc3ad5118934635 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 2 Nov 2017 13:14:30 +0000 Subject: [PATCH 0113/1471] drm/i915: Set up mocs tables before restarting the engines After a reset, we may immediately begin executing requests on restarting the engines. Ergo this has to be last step with all re-initialisation completed beforehand. The mocs setup was after we started executing the requests; do it earlier! Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171102131430.22328-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 9470ba0c1930e..6a71805be3899 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4936,13 +4936,10 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) if (ret) goto out; - /* Need to do basic initialisation of all rings first: */ - ret = __i915_gem_restart_engines(dev_priv); - if (ret) - goto out; - intel_mocs_init_l3cc_table(dev_priv); + /* Only when the HW is re-initialised, can we replay the requests */ + ret = __i915_gem_restart_engines(dev_priv); out: intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); return ret; -- GitLab From de7e8bd01a241103c141801479e0bda786b76290 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Sat, 4 Nov 2017 22:20:08 +0100 Subject: [PATCH 0114/1471] drm: via: use ktime_get() instead of do_gettimeofday We want to remove uses of do_gettimeofday() from the kernel since the resulting timeval structure overflows in 2038. This is not a problem for this particular use, but do_gettimeofday() is also not an appropriate method for measuring time intervals, since it requires a conversion into microseconds and is complicated to work with. ktime_get() is a better replacement, as it works with the monontonic kernel timebase and requires a minimum of computation. I'm slightly changing the output from microseconds to nanoseconds here, to avoid introducing a new division operation. This should be fine since the value is only used for debugging. Signed-off-by: Arnd Bergmann Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171104212131.2939989-1-arnd@arndb.de --- drivers/gpu/drm/via/via_drv.h | 4 ++-- drivers/gpu/drm/via/via_irq.c | 21 +++++++-------------- 2 files changed, 9 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/via/via_drv.h b/drivers/gpu/drm/via/via_drv.h index 9873942ca8f48..6d1ae834484ca 100644 --- a/drivers/gpu/drm/via/via_drv.h +++ b/drivers/gpu/drm/via/via_drv.h @@ -74,9 +74,9 @@ typedef struct drm_via_private { volatile uint32_t *last_pause_ptr; volatile uint32_t *hw_addr_ptr; drm_via_ring_buffer_t ring; - struct timeval last_vblank; + ktime_t last_vblank; int last_vblank_valid; - unsigned usec_per_vblank; + ktime_t nsec_per_vblank; atomic_t vbl_received; drm_via_state_t hc_state; char pci_buf[VIA_PCI_BUF_SIZE]; diff --git a/drivers/gpu/drm/via/via_irq.c b/drivers/gpu/drm/via/via_irq.c index ea8172c747a26..24e71578af4df 100644 --- a/drivers/gpu/drm/via/via_irq.c +++ b/drivers/gpu/drm/via/via_irq.c @@ -88,13 +88,6 @@ static int via_num_unichrome = ARRAY_SIZE(via_unichrome_irqs); static int via_irqmap_unichrome[] = {-1, -1, -1, 0, -1, 1}; -static unsigned time_diff(struct timeval *now, struct timeval *then) -{ - return (now->tv_usec >= then->tv_usec) ? - now->tv_usec - then->tv_usec : - 1000000 - (then->tv_usec - now->tv_usec); -} - u32 via_get_vblank_counter(struct drm_device *dev, unsigned int pipe) { drm_via_private_t *dev_priv = dev->dev_private; @@ -111,7 +104,7 @@ irqreturn_t via_driver_irq_handler(int irq, void *arg) drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; u32 status; int handled = 0; - struct timeval cur_vblank; + ktime_t cur_vblank; drm_via_irq_t *cur_irq = dev_priv->via_irqs; int i; @@ -119,18 +112,18 @@ irqreturn_t via_driver_irq_handler(int irq, void *arg) if (status & VIA_IRQ_VBLANK_PENDING) { atomic_inc(&dev_priv->vbl_received); if (!(atomic_read(&dev_priv->vbl_received) & 0x0F)) { - do_gettimeofday(&cur_vblank); + cur_vblank = ktime_get(); if (dev_priv->last_vblank_valid) { - dev_priv->usec_per_vblank = - time_diff(&cur_vblank, - &dev_priv->last_vblank) >> 4; + dev_priv->nsec_per_vblank = + ktime_sub(cur_vblank, + dev_priv->last_vblank) >> 4; } dev_priv->last_vblank = cur_vblank; dev_priv->last_vblank_valid = 1; } if (!(atomic_read(&dev_priv->vbl_received) & 0xFF)) { - DRM_DEBUG("US per vblank is: %u\n", - dev_priv->usec_per_vblank); + DRM_DEBUG("nsec per vblank is: %llu\n", + ktime_to_ns(dev_priv->nsec_per_vblank)); } drm_handle_vblank(dev, 0); handled = 1; -- GitLab From d36caeea4b0f76bf691c7d0cddc4a71ed93783c5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 5 Nov 2017 12:45:50 +0000 Subject: [PATCH 0115/1471] drm/i915: Assert vma->flags are updated correctly during binding As we bind, and unbind on error, we want to be sure that the vma->flags are updated to reflect the binding state so that on the next invocation all is well. v2: Take two. v3: Take three; vma-misplaced is checking map-and-fenceable so keep it last! Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171105124550.32715-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_vma.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index fc77e8191eb55..7cdf34800549c 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -633,15 +633,17 @@ int __i915_vma_do_pin(struct i915_vma *vma, if (ret) goto err_unpin; } + GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); ret = i915_vma_bind(vma, vma->obj->cache_level, flags); if (ret) goto err_remove; + GEM_BUG_ON((vma->flags & I915_VMA_BIND_MASK) == 0); + if ((bound ^ vma->flags) & I915_VMA_GLOBAL_BIND) __i915_vma_set_map_and_fenceable(vma); - GEM_BUG_ON(!drm_mm_node_allocated(&vma->node)); GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags)); return 0; @@ -649,6 +651,7 @@ int __i915_vma_do_pin(struct i915_vma *vma, if ((bound & I915_VMA_BIND_MASK) == 0) { i915_vma_remove(vma); GEM_BUG_ON(vma->pages); + GEM_BUG_ON(vma->flags & I915_VMA_BIND_MASK); } err_unpin: __i915_vma_unpin(vma); -- GitLab From 69ea47a5a98b198c9d36fe157a3986748a9e2554 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 25 Oct 2017 16:32:07 +0100 Subject: [PATCH 0116/1471] drm/i915/selftests: Hide dangerous tests Some tests are designed to exercise the limits of the HW and may trigger unintended side-effects making the machine unusable. This should not be executed by default, but are still useful for early platform validation. References: https://bugs.freedesktop.org/show_bug.cgi?id=103453 Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171025153207.9589-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/Kconfig.debug | 14 ++++++++++++++ drivers/gpu/drm/i915/selftests/intel_uncore.c | 8 ++++---- 2 files changed, 18 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug index aed7d207ea846..19c77c6feb245 100644 --- a/drivers/gpu/drm/i915/Kconfig.debug +++ b/drivers/gpu/drm/i915/Kconfig.debug @@ -90,6 +90,20 @@ config DRM_I915_SELFTEST If in doubt, say "N". +config DRM_I915_SELFTEST_BROKEN + bool "Enable broken and dangerous selftests" + depends on DRM_I915_SELFTEST + depends on BROKEN + default n + help + This option enables the execution of selftests that are "dangerous" + and may trigger unintended HW side-effects as they break strict + rules given in the HW specification. For science. + + Recommended for masochistic driver developers only. + + If in doubt, say "N". + config DRM_I915_LOW_LEVEL_TRACEPOINTS bool "Enable low level request tracing events" depends on DRM_I915 diff --git a/drivers/gpu/drm/i915/selftests/intel_uncore.c b/drivers/gpu/drm/i915/selftests/intel_uncore.c index 3cac22eb47ce1..f52a4ab9aa986 100644 --- a/drivers/gpu/drm/i915/selftests/intel_uncore.c +++ b/drivers/gpu/drm/i915/selftests/intel_uncore.c @@ -120,10 +120,10 @@ static int intel_uncore_check_forcewake_domains(struct drm_i915_private *dev_pri !IS_CHERRYVIEW(dev_priv)) return 0; - if (IS_VALLEYVIEW(dev_priv)) /* XXX system lockup! */ - return 0; - - if (IS_BROADWELL(dev_priv)) /* XXX random GPU hang afterwards! */ + /* + * This test may lockup the machine or cause GPU hangs afterwards. + */ + if (!IS_ENABLED(CONFIG_DRM_I915_SELFTEST_BROKEN)) return 0; valid = kzalloc(BITS_TO_LONGS(FW_RANGE) * sizeof(*valid), -- GitLab From f991c492aa55fb1c6834882c5d786d5bb3b25f07 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 6 Nov 2017 11:15:08 +0000 Subject: [PATCH 0117/1471] drm/i915: Lock llist_del_first() vs llist_del_all() An oversight in commit 87701b4b5593 ("drm/i915: Only free the oldest stale object before a fresh allocation") was that not only do we have to serialise concurrent users of llist_del_first(), but we also have to lock llist_del_first() vs llist_del_all(). From llist.h, * This can be summarized as follows: * * | add | del_first | del_all * add | - | - | - * del_first | | L | L * del_all | | | - * * Where, a particular row's operation can happen concurrently with a column's * operation, with "-" being no lock needed, while "L" being lock is needed. This should hopefully explain: <4>[ 89.287106] general protection fault: 0000 [#1] PREEMPT SMP <4>[ 89.287126] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic x86_pkg_temp_thermal intel_powerclamp coretemp i915 crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core r8169 mii mei_me mei snd_pcm prime_numbers i2c_hid pinctrl_geminilake pinctrl_intel <4>[ 89.287226] CPU: 2 PID: 23 Comm: ksoftirqd/2 Tainted: G U 4.14.0-rc8-CI-CI_DRM_3315+ #1 <4>[ 89.287247] Hardware name: Intel Corp. Geminilake/GLK RVP2 LP4SD (07), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017 <4>[ 89.287270] task: ffff88017ab34ec0 task.stack: ffffc90000128000 <4>[ 89.287290] RIP: 0010:llist_add_batch+0x4/0x20 <4>[ 89.287301] RSP: 0018:ffffc9000012bdb8 EFLAGS: 00010296 <4>[ 89.287314] RAX: ffffffff811017ad RBX: 6e468801a1560000 RCX: ef3e53fceecdeb81 <4>[ 89.287330] RDX: 6e468801a1566130 RSI: ffff880103d73d98 RDI: ffff880103d73d98 <4>[ 89.287346] RBP: ffffc9000012bdb8 R08: ffff88017ab35780 R09: 0000000000000000 <4>[ 89.287361] R10: ffffc9000012bd68 R11: 00000000abb18c3d R12: ffffffffa01369e0 <4>[ 89.287377] R13: ffff88017fd1b8f8 R14: ffff88017ab34ec0 R15: 000000000000000a <4>[ 89.287393] FS: 0000000000000000(0000) GS:ffff88017fd00000(0000) knlGS:0000000000000000 <4>[ 89.287411] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 <4>[ 89.287424] CR2: 00007ff0c0755018 CR3: 000000016df9b000 CR4: 00000000003406e0 <4>[ 89.287440] Call Trace: <4>[ 89.287511] __i915_gem_free_object_rcu+0x20/0x40 [i915] <4>[ 89.287527] rcu_process_callbacks+0x27a/0x730 <4>[ 89.287544] __do_softirq+0xc0/0x4ae <4>[ 89.287559] ? smpboot_thread_fn+0x2d/0x280 <4>[ 89.287571] run_ksoftirqd+0x1f/0x70 <4>[ 89.287582] smpboot_thread_fn+0x18a/0x280 <4>[ 89.287595] kthread+0x114/0x150 <4>[ 89.287605] ? sort_range+0x30/0x30 <4>[ 89.287615] ? kthread_create_on_node+0x40/0x40 <4>[ 89.287628] ret_from_fork+0x27/0x40 <4>[ 89.287641] Code: 0d 48 83 ea 01 4c 89 c1 48 83 fa ff 74 12 48 23 0c d7 74 ed 48 c1 e2 06 48 0f bd c9 48 8d 04 0a 5d c3 90 90 90 90 90 55 48 89 e5 <48> 8b 0a 48 89 0e 48 89 c8 f0 48 0f b1 3a 48 39 c1 75 ed 48 85 <1>[ 89.287774] RIP: llist_add_batch+0x4/0x20 RSP: ffffc9000012bdb8 <4>[ 89.287826] ---[ end trace e775d15174d8ae02 ]--- (Lockless lists are only easy (and lockless) when only using llist_add/llist_del_all!) Fixes: 87701b4b5593 ("drm/i915: Only free the oldest stale object before a fresh allocation") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171106111508.11941-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_gem.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 6a71805be3899..889ae8810d5f5 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4636,11 +4636,17 @@ static void __i915_gem_free_work(struct work_struct *work) * unbound now. */ + spin_lock(&i915->mm.free_lock); while ((freed = llist_del_all(&i915->mm.free_list))) { + spin_unlock(&i915->mm.free_lock); + __i915_gem_free_objects(i915, freed); if (need_resched()) - break; + return; + + spin_lock(&i915->mm.free_lock); } + spin_unlock(&i915->mm.free_lock); } static void __i915_gem_free_object_rcu(struct rcu_head *head) -- GitLab From 856efd21c6f7139a74d340cdb6e0de5d8e226134 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 6 Nov 2017 11:48:33 +0000 Subject: [PATCH 0118/1471] drm/i915/guc: Assert guc->stage_desc_pool is allocated Silence smatch by demonstrating that guc->stage_desc_pool is allocated following a successful guc_stage_desc_pool_create(), drivers/gpu/drm/i915/i915_guc_submission.c:1293 i915_guc_submission_init() error: we previously assumed 'guc->stage_desc_pool' could be null (see line 1261) Signed-off-by: Chris Wilson Cc: Oscar Mateo Cc: Daniele Ceraolo Spurio Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171106114833.31199-1-chris@chris-wilson.co.uk Reviewed-by: Michal Wajdeczko --- drivers/gpu/drm/i915/i915_guc_submission.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index d14c1342f09dc..0ba2fc04fe9c1 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -1264,10 +1264,16 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) ret = guc_stage_desc_pool_create(guc); if (ret) return ret; + /* + * Keep static analysers happy, let them know that we allocated the + * vma after testing that it didn't exist earlier. + */ + GEM_BUG_ON(!guc->stage_desc_pool); ret = guc_shared_data_create(guc); if (ret) goto err_stage_desc_pool; + GEM_BUG_ON(!guc->shared_data); ret = intel_guc_log_create(guc); if (ret < 0) @@ -1276,10 +1282,12 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) ret = guc_preempt_work_create(guc); if (ret) goto err_log; + GEM_BUG_ON(!guc->preempt_wq); ret = guc_ads_create(guc); if (ret < 0) goto err_wq; + GEM_BUG_ON(!guc->ads_vma); return 0; -- GitLab From 1c5a907180365d0842c17fb15977ad426d429c14 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Mon, 6 Nov 2017 13:51:54 +0000 Subject: [PATCH 0119/1471] drm/i915/guc: Assert ctch->vma is allocated Silence smatch by demonstrating that ctch->vma is allocated following a successful chch_init() drivers/gpu/drm/i915/intel_guc_ct.c:204 ctch_open() error: we previously assumed 'ctch->vma' could be null (see line 197) Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171106135154.52520-1-michal.wajdeczko@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_ct.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/intel_guc_ct.c b/drivers/gpu/drm/i915/intel_guc_ct.c index c4cbec140101b..24ad557523965 100644 --- a/drivers/gpu/drm/i915/intel_guc_ct.c +++ b/drivers/gpu/drm/i915/intel_guc_ct.c @@ -198,6 +198,7 @@ static int ctch_open(struct intel_guc *guc, err = ctch_init(guc, ctch); if (unlikely(err)) goto err_out; + GEM_BUG_ON(!ctch->vma); } /* vma should be already allocated and map'ed */ -- GitLab From 7d41ef3479a6bc3ba2ddcf49e365916611e4fa39 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Thu, 26 Oct 2017 17:36:55 +0000 Subject: [PATCH 0120/1471] drm/i915: Add Guc/HuC firmware details to error state Include GuC and HuC firmware details in captured error state to provide additional debug information. To reuse existing uc firmware pretty printer, introduce new drm-printer variant that works with our i915_error_state_buf output. Also update uc firmware pretty printer to accept const input. v2: don't rely on current caps (Chris) dump correct fw info (Michal) v3: simplify capture of custom paths (Chris) v4: improve 'why' comment (Joonas) trim output if no fw path (Michal) group code around uc error state (Michal) v5: use error in cleanup_uc (Michal) Suggested-by: Chris Wilson Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-1-michal.wajdeczko@intel.com [ickle: allow printing uc_fw after allocation failure] Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.h | 5 +++ drivers/gpu/drm/i915/i915_gpu_error.c | 64 +++++++++++++++++++++++++++ drivers/gpu/drm/i915/intel_uc_fw.c | 2 +- drivers/gpu/drm/i915/intel_uc_fw.h | 2 +- 4 files changed, 71 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 72bb5b51035a0..632249f90b91a 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -913,6 +913,11 @@ struct i915_gpu_state { struct intel_device_info device_info; struct i915_params params; + struct i915_error_uc { + struct intel_uc_fw guc_fw; + struct intel_uc_fw huc_fw; + } uc; + /* Generic register state */ u32 eir; u32 pgtbl_er; diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 653fb69e7ecb0..bc264000361f1 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -30,6 +30,8 @@ #include #include #include +#include + #include "i915_drv.h" static const char *engine_str(int engine) @@ -175,6 +177,21 @@ static void i915_error_puts(struct drm_i915_error_state_buf *e, #define err_printf(e, ...) i915_error_printf(e, __VA_ARGS__) #define err_puts(e, s) i915_error_puts(e, s) +static void __i915_printfn_error(struct drm_printer *p, struct va_format *vaf) +{ + i915_error_vprintf(p->arg, vaf->fmt, *vaf->va); +} + +static inline struct drm_printer +i915_error_printer(struct drm_i915_error_state_buf *e) +{ + struct drm_printer p = { + .printfn = __i915_printfn_error, + .arg = e, + }; + return p; +} + #ifdef CONFIG_DRM_I915_COMPRESS_ERROR struct compress { @@ -589,6 +606,20 @@ static void err_print_pciid(struct drm_i915_error_state_buf *m, pdev->subsystem_device); } +static void err_print_uc(struct drm_i915_error_state_buf *m, + const struct i915_error_uc *error_uc) +{ + struct drm_printer p = i915_error_printer(m); + const struct i915_gpu_state *error = + container_of(error_uc, typeof(*error), uc); + + if (!error->device_info.has_guc) + return; + + intel_uc_fw_dump(&error_uc->guc_fw, &p); + intel_uc_fw_dump(&error_uc->huc_fw, &p); +} + int i915_error_state_to_str(struct drm_i915_error_state_buf *m, const struct i915_gpu_state *error) { @@ -773,6 +804,7 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m, err_print_capabilities(m, &error->device_info); err_print_params(m, &error->params); + err_print_uc(m, &error->uc); if (m->bytes == 0 && m->err) return m->err; @@ -831,6 +863,14 @@ static __always_inline void free_param(const char *type, void *x) kfree(*(void **)x); } +static void cleanup_uc_state(struct i915_gpu_state *error) +{ + struct i915_error_uc *error_uc = &error->uc; + + kfree(error_uc->guc_fw.path); + kfree(error_uc->huc_fw.path); +} + void __i915_gpu_state_free(struct kref *error_ref) { struct i915_gpu_state *error = @@ -870,6 +910,8 @@ void __i915_gpu_state_free(struct kref *error_ref) I915_PARAMS_FOR_EACH(FREE); #undef FREE + cleanup_uc_state(error); + kfree(error); } @@ -1559,6 +1601,26 @@ static void i915_capture_pinned_buffers(struct drm_i915_private *dev_priv, error->pinned_bo = bo; } +static void capture_uc_state(struct i915_gpu_state *error) +{ + struct drm_i915_private *i915 = error->i915; + struct i915_error_uc *error_uc = &error->uc; + + /* Capturing uC state won't be useful if there is no GuC */ + if (!error->device_info.has_guc) + return; + + error_uc->guc_fw = i915->guc.fw; + error_uc->huc_fw = i915->huc.fw; + + /* Non-default firmware paths will be specified by the modparam. + * As modparams are generally accesible from the userspace make + * explicit copies of the firmware paths. + */ + error_uc->guc_fw.path = kstrdup(i915->guc.fw.path, GFP_ATOMIC); + error_uc->huc_fw.path = kstrdup(i915->huc.fw.path, GFP_ATOMIC); +} + static void i915_gem_capture_guc_log_buffer(struct drm_i915_private *dev_priv, struct i915_gpu_state *error) { @@ -1710,6 +1772,8 @@ static int capture(void *data) I915_PARAMS_FOR_EACH(DUP); #undef DUP + capture_uc_state(error); + i915_capture_gen_state(error->i915, error); i915_capture_reg_state(error->i915, error); i915_gem_record_fences(error->i915, error); diff --git a/drivers/gpu/drm/i915/intel_uc_fw.c b/drivers/gpu/drm/i915/intel_uc_fw.c index 973888e94cba6..4bc82d3005ff3 100644 --- a/drivers/gpu/drm/i915/intel_uc_fw.c +++ b/drivers/gpu/drm/i915/intel_uc_fw.c @@ -299,7 +299,7 @@ void intel_uc_fw_fini(struct intel_uc_fw *uc_fw) * * Pretty printer for uC firmware. */ -void intel_uc_fw_dump(struct intel_uc_fw *uc_fw, struct drm_printer *p) +void intel_uc_fw_dump(const struct intel_uc_fw *uc_fw, struct drm_printer *p) { drm_printf(p, "%s firmware: %s\n", intel_uc_fw_type_repr(uc_fw->type), uc_fw->path); diff --git a/drivers/gpu/drm/i915/intel_uc_fw.h b/drivers/gpu/drm/i915/intel_uc_fw.h index 1329036693918..5394d9d1e683e 100644 --- a/drivers/gpu/drm/i915/intel_uc_fw.h +++ b/drivers/gpu/drm/i915/intel_uc_fw.h @@ -116,6 +116,6 @@ int intel_uc_fw_upload(struct intel_uc_fw *uc_fw, int (*xfer)(struct intel_uc_fw *uc_fw, struct i915_vma *vma)); void intel_uc_fw_fini(struct intel_uc_fw *uc_fw); -void intel_uc_fw_dump(struct intel_uc_fw *uc_fw, struct drm_printer *p); +void intel_uc_fw_dump(const struct intel_uc_fw *uc_fw, struct drm_printer *p); #endif -- GitLab From 0397ac13dd47bb0c99b10a11b756aafb72018fca Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Thu, 26 Oct 2017 17:36:56 +0000 Subject: [PATCH 0121/1471] drm/i915: Make GuC log part of the uC error state We keep details of GuC and HuC in separate error state struct. Make GuC log part of it to group all related data together. Since we are printing uC details at the end, with this change GuC log will be moved there too. v2: comment on new placement of the log (Chris) Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-2-michal.wajdeczko@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.h | 2 +- drivers/gpu/drm/i915/i915_gpu_error.c | 18 +++--------------- 2 files changed, 4 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 632249f90b91a..fe93115c4caa9 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -916,6 +916,7 @@ struct i915_gpu_state { struct i915_error_uc { struct intel_uc_fw guc_fw; struct intel_uc_fw huc_fw; + struct drm_i915_error_object *guc_log; } uc; /* Generic register state */ @@ -941,7 +942,6 @@ struct i915_gpu_state { struct intel_overlay_error_state *overlay; struct intel_display_error_state *display; struct drm_i915_error_object *semaphore; - struct drm_i915_error_object *guc_log; struct drm_i915_error_engine { int engine_id; diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index bc264000361f1..6e840db28bec9 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -618,6 +618,7 @@ static void err_print_uc(struct drm_i915_error_state_buf *m, intel_uc_fw_dump(&error_uc->guc_fw, &p); intel_uc_fw_dump(&error_uc->huc_fw, &p); + print_error_obj(m, NULL, "GuC log buffer", error_uc->guc_log); } int i915_error_state_to_str(struct drm_i915_error_state_buf *m, @@ -794,8 +795,6 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m, print_error_obj(m, NULL, "Semaphores", error->semaphore); - print_error_obj(m, NULL, "GuC log buffer", error->guc_log); - if (error->overlay) intel_overlay_print_error_state(m, error->overlay); @@ -869,6 +868,7 @@ static void cleanup_uc_state(struct i915_gpu_state *error) kfree(error_uc->guc_fw.path); kfree(error_uc->huc_fw.path); + i915_error_object_free(error_uc->guc_log); } void __i915_gpu_state_free(struct kref *error_ref) @@ -897,7 +897,6 @@ void __i915_gpu_state_free(struct kref *error_ref) } i915_error_object_free(error->semaphore); - i915_error_object_free(error->guc_log); for (i = 0; i < ARRAY_SIZE(error->active_bo); i++) kfree(error->active_bo[i]); @@ -1619,17 +1618,7 @@ static void capture_uc_state(struct i915_gpu_state *error) */ error_uc->guc_fw.path = kstrdup(i915->guc.fw.path, GFP_ATOMIC); error_uc->huc_fw.path = kstrdup(i915->huc.fw.path, GFP_ATOMIC); -} - -static void i915_gem_capture_guc_log_buffer(struct drm_i915_private *dev_priv, - struct i915_gpu_state *error) -{ - /* Capturing log buf contents won't be useful if logging was disabled */ - if (!dev_priv->guc.log.vma || (i915_modparams.guc_log_level < 0)) - return; - - error->guc_log = i915_error_object_create(dev_priv, - dev_priv->guc.log.vma); + error_uc->guc_log = i915_error_object_create(i915, i915->guc.log.vma); } /* Capture all registers which don't fit into another category. */ @@ -1780,7 +1769,6 @@ static int capture(void *data) i915_gem_record_rings(error->i915, error); i915_capture_active_buffers(error->i915, error); i915_capture_pinned_buffers(error->i915, error); - i915_gem_capture_guc_log_buffer(error->i915, error); error->overlay = intel_overlay_capture_error_state(error->i915); error->display = intel_display_capture_error_state(error->i915); -- GitLab From 84a20a8a360d5d3322d976672427fae800722038 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Thu, 26 Oct 2017 17:36:57 +0000 Subject: [PATCH 0122/1471] drm/i915: Handle error-state modparams in dedicated functions Capturing and cleanup and modparams in error state requires some macro tricks. Move that code into separated functions for easier maintenance. Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171026173657.49648-3-michal.wajdeczko@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_gpu_error.c | 26 +++++++++++++++++--------- 1 file changed, 17 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 6e840db28bec9..5c2d83a838d8c 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -862,6 +862,13 @@ static __always_inline void free_param(const char *type, void *x) kfree(*(void **)x); } +static void cleanup_params(struct i915_gpu_state *error) +{ +#define FREE(T, x, ...) free_param(#T, &error->params.x); + I915_PARAMS_FOR_EACH(FREE); +#undef FREE +} + static void cleanup_uc_state(struct i915_gpu_state *error) { struct i915_error_uc *error_uc = &error->uc; @@ -905,10 +912,7 @@ void __i915_gpu_state_free(struct kref *error_ref) kfree(error->overlay); kfree(error->display); -#define FREE(T, x, ...) free_param(#T, &error->params.x); - I915_PARAMS_FOR_EACH(FREE); -#undef FREE - + cleanup_params(error); cleanup_uc_state(error); kfree(error); @@ -1746,6 +1750,14 @@ static __always_inline void dup_param(const char *type, void *x) *(void **)x = kstrdup(*(void **)x, GFP_ATOMIC); } +static void capture_params(struct i915_gpu_state *error) +{ + error->params = i915_modparams; +#define DUP(T, x, ...) dup_param(#T, &error->params.x); + I915_PARAMS_FOR_EACH(DUP); +#undef DUP +} + static int capture(void *data) { struct i915_gpu_state *error = data; @@ -1756,11 +1768,7 @@ static int capture(void *data) ktime_to_timeval(ktime_sub(ktime_get(), error->i915->gt.last_init_time)); - error->params = i915_modparams; -#define DUP(T, x, ...) dup_param(#T, &error->params.x); - I915_PARAMS_FOR_EACH(DUP); -#undef DUP - + capture_params(error); capture_uc_state(error); i915_capture_gen_state(error->i915, error); -- GitLab From 6d566930257df564d0aaacad0f02855d9c85aecd Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 6 Nov 2017 14:52:27 -0500 Subject: [PATCH 0123/1471] drm/amdkfd: Hardware DWORD size is 4 bytes Don't use sizeof(uint32_t) or similar types for hardware or firmware DWORD size. The hardware and firmware don't care about Linux types. Signed-off-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c | 14 +++++--------- drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c | 10 ++++------ drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 9 ++++----- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 2 +- 5 files changed, 15 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c index c407f6bd99565..afb26f205d297 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_dbgdev.c @@ -95,7 +95,7 @@ static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev, ib_packet->bitfields3.ib_base_hi = largep->u.high_part; ib_packet->control = (1 << 23) | (1 << 31) | - ((size_in_bytes / sizeof(uint32_t)) & 0xfffff); + ((size_in_bytes / 4) & 0xfffff); ib_packet->bitfields5.pasid = pasid; @@ -126,8 +126,7 @@ static int dbgdev_diq_submit_ib(struct kfd_dbgdev *dbgdev, rm_packet->header.opcode = IT_RELEASE_MEM; rm_packet->header.type = PM4_TYPE_3; - rm_packet->header.count = sizeof(struct pm4__release_mem) / - sizeof(unsigned int) - 2; + rm_packet->header.count = sizeof(struct pm4__release_mem) / 4 - 2; rm_packet->bitfields2.event_type = CACHE_FLUSH_AND_INV_TS_EVENT; rm_packet->bitfields2.event_index = @@ -652,8 +651,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev, packets_vec[0].header.opcode = IT_SET_UCONFIG_REG; packets_vec[0].header.type = PM4_TYPE_3; packets_vec[0].bitfields2.reg_offset = - GRBM_GFX_INDEX / (sizeof(uint32_t)) - - USERCONFIG_REG_BASE; + GRBM_GFX_INDEX / 4 - USERCONFIG_REG_BASE; packets_vec[0].bitfields2.insert_vmid = 0; packets_vec[0].reg_data[0] = reg_gfx_index.u32All; @@ -661,8 +659,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev, packets_vec[1].header.count = 1; packets_vec[1].header.opcode = IT_SET_CONFIG_REG; packets_vec[1].header.type = PM4_TYPE_3; - packets_vec[1].bitfields2.reg_offset = SQ_CMD / (sizeof(uint32_t)) - - AMD_CONFIG_REG_BASE; + packets_vec[1].bitfields2.reg_offset = SQ_CMD / 4 - AMD_CONFIG_REG_BASE; packets_vec[1].bitfields2.vmid_shift = SQ_CMD_VMID_OFFSET; packets_vec[1].bitfields2.insert_vmid = 1; @@ -678,8 +675,7 @@ static int dbgdev_wave_control_diq(struct kfd_dbgdev *dbgdev, packets_vec[2].ordinal1 = packets_vec[0].ordinal1; packets_vec[2].bitfields2.reg_offset = - GRBM_GFX_INDEX / (sizeof(uint32_t)) - - USERCONFIG_REG_BASE; + GRBM_GFX_INDEX / 4 - USERCONFIG_REG_BASE; packets_vec[2].bitfields2.insert_vmid = 0; packets_vec[2].reg_data[0] = reg_gfx_index.u32All; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 8b0c0645d7c05..5dc6567d4a138 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -218,7 +218,7 @@ static int acquire_packet_buffer(struct kernel_queue *kq, rptr = *kq->rptr_kernel; wptr = *kq->wptr_kernel; queue_address = (unsigned int *)kq->pq_kernel_addr; - queue_size_dwords = kq->queue->properties.queue_size / sizeof(uint32_t); + queue_size_dwords = kq->queue->properties.queue_size / 4; pr_debug("rptr: %d\n", rptr); pr_debug("wptr: %d\n", wptr); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 9873929ab4efa..efed6ef6dad8f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -154,7 +154,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, uint32_t pipe_id, { /* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */ uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); - uint32_t wptr_mask = (uint32_t)((p->queue_size / sizeof(uint32_t)) - 1); + uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1); return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, @@ -183,8 +183,7 @@ static int update_mqd(struct mqd_manager *mm, void *mqd, * Calculating queue size which is log base 2 of actual queue size -1 * dwords and another -1 for ffs */ - m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int)) - - 1 - 1; + m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); @@ -209,7 +208,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, struct cik_sdma_rlc_registers *m; m = get_sdma_mqd(mqd); - m->sdma_rlc_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1) + m->sdma_rlc_rb_cntl = (ffs(q->queue_size / 4) - 1) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | @@ -350,8 +349,7 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, * Calculating queue size which is log base 2 of actual queue * size -1 dwords */ - m->cp_hqd_pq_control |= ffs(q->queue_size / sizeof(unsigned int)) - - 1 - 1; + m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index a117d2b7721f5..85e1b676b1d7e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -103,7 +103,7 @@ static int load_mqd(struct mqd_manager *mm, void *mqd, { /* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */ uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); - uint32_t wptr_mask = (uint32_t)((p->queue_size / sizeof(uint32_t)) - 1); + uint32_t wptr_mask = (uint32_t)((p->queue_size / 4) - 1); return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, (uint32_t __user *)p->write_ptr, @@ -121,8 +121,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd, m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT | atc_bit << CP_HQD_PQ_CONTROL__PQ_ATC__SHIFT | mtype << CP_HQD_PQ_CONTROL__MTYPE__SHIFT; - m->cp_hqd_pq_control |= - ffs(q->queue_size / sizeof(unsigned int)) - 1 - 1; + m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control); m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); @@ -152,7 +151,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd, * is safe, giving a maximum field value of 0xA. */ m->cp_hqd_eop_control |= min(0xA, - ffs(q->eop_ring_buffer_size / sizeof(unsigned int)) - 1 - 1); + ffs(q->eop_ring_buffer_size / 4) - 1 - 1); m->cp_hqd_eop_base_addr_lo = lower_32_bits(q->eop_ring_buffer_address >> 8); m->cp_hqd_eop_base_addr_hi = @@ -288,7 +287,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, struct vi_sdma_mqd *m; m = get_sdma_mqd(mqd); - m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / sizeof(unsigned int)) - 1) + m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / 4) - 1) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index 16da8ad02d8be..69c147a835918 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -45,7 +45,7 @@ static unsigned int build_pm4_header(unsigned int opcode, size_t packet_size) header.u32All = 0; header.opcode = opcode; - header.count = packet_size/sizeof(uint32_t) - 2; + header.count = packet_size / 4 - 2; header.type = PM4_TYPE_3; return header.u32All; -- GitLab From 115c8c4104bab4c435e0eaf7ffd5aabaf2adbbd2 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 6 Nov 2017 14:52:28 -0500 Subject: [PATCH 0124/1471] drm/amdkfd: Use order_base_2 to get log2 of buffes sizes Replace (ffs(size) - 1) with order_base_2(size) as a more straight forward way to get log2 of buffer sizes. Signed-off-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c | 6 +++--- drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index efed6ef6dad8f..7aa57abebbcd0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -183,7 +183,7 @@ static int update_mqd(struct mqd_manager *mm, void *mqd, * Calculating queue size which is log base 2 of actual queue size -1 * dwords and another -1 for ffs */ - m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; + m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1; m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); @@ -208,7 +208,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, struct cik_sdma_rlc_registers *m; m = get_sdma_mqd(mqd); - m->sdma_rlc_rb_cntl = (ffs(q->queue_size / 4) - 1) + m->sdma_rlc_rb_cntl = order_base_2(q->queue_size / 4) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | @@ -349,7 +349,7 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, * Calculating queue size which is log base 2 of actual queue * size -1 dwords */ - m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; + m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1; m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index 85e1b676b1d7e..2ba7cea7b99bf 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -121,7 +121,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd, m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT | atc_bit << CP_HQD_PQ_CONTROL__PQ_ATC__SHIFT | mtype << CP_HQD_PQ_CONTROL__MTYPE__SHIFT; - m->cp_hqd_pq_control |= ffs(q->queue_size / 4) - 1 - 1; + m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1; pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control); m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); @@ -151,7 +151,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd, * is safe, giving a maximum field value of 0xA. */ m->cp_hqd_eop_control |= min(0xA, - ffs(q->eop_ring_buffer_size / 4) - 1 - 1); + order_base_2(q->eop_ring_buffer_size / 4) - 1); m->cp_hqd_eop_base_addr_lo = lower_32_bits(q->eop_ring_buffer_address >> 8); m->cp_hqd_eop_base_addr_hi = @@ -287,7 +287,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, struct vi_sdma_mqd *m; m = get_sdma_mqd(mqd); - m->sdmax_rlcx_rb_cntl = (ffs(q->queue_size / 4) - 1) + m->sdmax_rlcx_rb_cntl = order_base_2(q->queue_size / 4) << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | -- GitLab From 5684514ba9dc6d7aa932cc53d97d866b2386221f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 6 Nov 2017 21:11:28 +0000 Subject: [PATCH 0125/1471] drm/i915: Deconstruct struct sgt_dma initialiser MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit gcc-4.4 complains about: struct sgt_dma iter = { .sg = vma->pages->sgl, .dma = sg_dma_address(iter.sg), .max = iter.dma + iter.sg->length, }; drivers/gpu/drm/i915/i915_gem_gtt.c: In function ‘gen8_ppgtt_insert_4lvl’: drivers/gpu/drm/i915/i915_gem_gtt.c:938: error: ‘iter.sg’ is used uninitialized in this function drivers/gpu/drm/i915/i915_gem_gtt.c:939: error: ‘iter.dma’ is used uninitialized in this function and worse generates invalid code that triggers a GPF: BUG: unable to handle kernel NULL pointer dereference at 0000000000000010 IP: gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915] PGD 0 Oops: 0000 [#1] SMP Modules linked in: snd_aloop nf_conntrack_ipv6 nf_defrag_ipv6 nf_log_ipv6 ip6table_filter ip6_tables ctr ccm xt_state nf_log_ipv4 nf_log_common xt_LOG xt_limit xt_recent xt_owner xt_addrtype iptable_filter ipt_MASQUERADE nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack libcrc32c ip_tables dm_mod vhost_net macvtap macvlan vhost tun kvm_intel kvm irqbypass uas usb_storage hid_multitouch btusb btrtl uvcvideo videobuf2_v4l2 videobuf2_core videodev media videobuf2_vmalloc videobuf2_memops sg ppdev dell_wmi sparse_keymap mei_wdt sd_mod iTCO_wdt iTCO_vendor_support rtsx_pci_ms memstick rtsx_pci_sdmmc mmc_core dell_smm_hwmon hwmon dell_laptop dell_smbios dcdbas joydev input_leds hci_uart btintel btqca btbcm bluetooth parport_pc parport i2c_hid intel_lpss_acpi intel_lpss pcspkr wmi int3400_thermal acpi_thermal_rel dell_rbtn mei_me mei snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic ahci libahci acpi_pad xhci_pci xhci_hcd snd_hda_intel snd_hda_codec snd_hda_core snd_hwdep snd_seq snd_seq_device snd_pcm snd_timer snd soundcore int3403_thermal arc4 e1000e ptp pps_core i2c_i801 iwlmvm mac80211 rtsx_pci iwlwifi cfg80211 rfkill intel_pch_thermal processor_thermal_device int340x_thermal_zone intel_soc_dts_iosf i915 video fjes CPU: 2 PID: 2408 Comm: X Not tainted 4.10.0-rc5+ #1 Hardware name: Dell Inc. Latitude E7470/0T6HHJ, BIOS 1.11.3 11/09/2016 task: ffff880219fe4740 task.stack: ffffc90005f98000 RIP: 0010:gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915] RSP: 0018:ffffc90005f9b8c8 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff8802167d8000 RCX: 0000000000000001 RDX: 00000000ffff7000 RSI: ffff880219f94140 RDI: ffff880228444000 RBP: ffffc90005f9b948 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000080 R13: 0000000000000001 R14: ffffc90005f9bcd7 R15: ffff88020c9a83c0 FS: 00007fb53e1ee920(0000) GS:ffff88024dd00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000010 CR3: 000000022ef95000 CR4: 00000000003406e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: ppgtt_bind_vma+0x40/0x50 [i915] i915_vma_bind+0xcb/0x1c0 [i915] __i915_vma_do_pin+0x6e/0xd0 [i915] i915_gem_execbuffer_reserve_vma+0x162/0x1d0 [i915] i915_gem_execbuffer_reserve+0x4fc/0x510 [i915] ? __kmalloc+0x134/0x250 ? i915_gem_wait_for_error+0x25/0x100 [i915] ? i915_gem_wait_for_error+0x25/0x100 [i915] i915_gem_do_execbuffer+0x2df/0xa00 [i915] ? drm_malloc_gfp.clone.0+0x42/0x80 [i915] ? path_put+0x22/0x30 ? __check_object_size+0x62/0x1f0 ? terminate_walk+0x44/0x90 i915_gem_execbuffer2+0x95/0x1e0 [i915] drm_ioctl+0x243/0x490 ? handle_pte_fault+0x1d7/0x220 ? i915_gem_do_execbuffer+0xa00/0xa00 [i915] ? handle_mm_fault+0x10d/0x2a0 vfs_ioctl+0x18/0x30 do_vfs_ioctl+0x14b/0x3f0 SyS_ioctl+0x92/0xa0 entry_SYSCALL_64_fastpath+0x1a/0xa9 RIP: 0033:0x7fb53b4fcb77 RSP: 002b:00007ffe0c572898 EFLAGS: 00003246 ORIG_RAX: 0000000000000010 RAX: ffffffffffffffda RBX: 00007fb53e17c038 RCX: 00007fb53b4fcb77 RDX: 00007ffe0c572900 RSI: 0000000040406469 RDI: 000000000000000b RBP: 00007fb5376d67e0 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000028 R11: 0000000000003246 R12: 0000000000000000 R13: 0000000000000000 R14: 000055eecb314d00 R15: 000055eecb315460 Code: 0f 84 5d ff ff ff eb a2 0f 1f 84 00 00 00 00 00 55 48 89 e5 41 57 41 56 41 55 41 54 53 48 83 ec 58 0f 1f 44 00 00 31 c0 89 4d b0 <4c> 8b 60 10 44 8b 70 0c 48 89 d0 4c 8b 2e 48 c1 e8 27 25 ff 01 RIP: gen8_ppgtt_insert_4lvl+0x1b/0x1e0 [i915] RSP: ffffc90005f9b8c8 CR2: 0000000000000010 Recent gccs, such as 4.9, 6.3 or 7.2, do not generate the warning nor do they explode on use. If we manually create the struct using locals from the stack, this should eliminate this issue, and does not alter code generation with gcc-7.2. Fixes: 894ccebee2b0 ("drm/i915: Micro-optimise gen8_ppgtt_insert_entries()") Reported-by: Kelly French Signed-off-by: Chris Wilson Cc: Kelly French Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171106211128.12538-1-chris@chris-wilson.co.uk Tested-by: Kelly French Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_gem_gtt.c | 25 +++++++++---------------- 1 file changed, 9 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 0684d5df97d9d..2847a6b41c168 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -958,10 +958,14 @@ static void gen8_ppgtt_clear_4lvl(struct i915_address_space *vm, } } -struct sgt_dma { +static inline struct sgt_dma { struct scatterlist *sg; dma_addr_t dma, max; -}; +} sgt_dma(struct i915_vma *vma) { + struct scatterlist *sg = vma->pages->sgl; + dma_addr_t addr = sg_dma_address(sg); + return (struct sgt_dma) { sg, addr, addr + sg->length }; +} struct gen8_insert_pte { u16 pml4e; @@ -1042,11 +1046,7 @@ static void gen8_ppgtt_insert_3lvl(struct i915_address_space *vm, u32 unused) { struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); - struct sgt_dma iter = { - .sg = vma->pages->sgl, - .dma = sg_dma_address(iter.sg), - .max = iter.dma + iter.sg->length, - }; + struct sgt_dma iter = sgt_dma(vma); struct gen8_insert_pte idx = gen8_insert_pte(vma->node.start); gen8_ppgtt_insert_pte_entries(ppgtt, &ppgtt->pdp, &iter, &idx, @@ -1158,11 +1158,7 @@ static void gen8_ppgtt_insert_4lvl(struct i915_address_space *vm, u32 unused) { struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); - struct sgt_dma iter = { - .sg = vma->pages->sgl, - .dma = sg_dma_address(iter.sg), - .max = iter.dma + iter.sg->length, - }; + struct sgt_dma iter = sgt_dma(vma); struct i915_page_directory_pointer **pdps = ppgtt->pml4.pdps; if (vma->page_sizes.sg > I915_GTT_PAGE_SIZE) { @@ -1869,13 +1865,10 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm, unsigned act_pt = first_entry / GEN6_PTES; unsigned act_pte = first_entry % GEN6_PTES; const u32 pte_encode = vm->pte_encode(0, cache_level, flags); - struct sgt_dma iter; + struct sgt_dma iter = sgt_dma(vma); gen6_pte_t *vaddr; vaddr = kmap_atomic_px(ppgtt->pd.page_table[act_pt]); - iter.sg = vma->pages->sgl; - iter.dma = sg_dma_address(iter.sg); - iter.max = iter.dma + iter.sg->length; do { vaddr[act_pte] = pte_encode | GEN6_PTE_ADDR_ENCODE(iter.dma); -- GitLab From bb3fb5e515df44c750fc641970445b61c09a72b6 Mon Sep 17 00:00:00 2001 From: Philippe CORNU Date: Thu, 26 Oct 2017 12:04:50 +0200 Subject: [PATCH 0126/1471] drm/stm: dsi: Rename driver name Rename the driver name from "dw_mipi_dsi-stm" to "stm32-display-dsi" for a better readability in /sys/bus/platform/drivers entries. Signed-off-by: Philippe Cornu Reviewed-by: Gustavo Padovan Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1509012290-16906-1-git-send-email-philippe.cornu@st.com --- drivers/gpu/drm/stm/dw_mipi_dsi-stm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c b/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c index e5b6310240fe6..10b2b779e72e6 100644 --- a/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c +++ b/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c @@ -342,7 +342,7 @@ static struct platform_driver dw_mipi_dsi_stm_driver = { .remove = dw_mipi_dsi_stm_remove, .driver = { .of_match_table = dw_mipi_dsi_stm_dt_ids, - .name = "dw_mipi_dsi-stm", + .name = "stm32-display-dsi", }, }; -- GitLab From 0163d1f62f511c6f85bb84cbc55e84cedc3eab88 Mon Sep 17 00:00:00 2001 From: Philippe CORNU Date: Thu, 26 Oct 2017 12:24:33 +0200 Subject: [PATCH 0127/1471] drm/stm: checkpatch strict minor updates Minor fixes detected with "scripts/checkpatch.pl --strict" Signed-off-by: Philippe Cornu Reviewed-by: Gustavo Padovan Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1509013473-18365-1-git-send-email-philippe.cornu@st.com --- drivers/gpu/drm/stm/dw_mipi_dsi-stm.c | 4 ++-- drivers/gpu/drm/stm/ltdc.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c b/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c index 10b2b779e72e6..3e8b9ed7e8b3c 100644 --- a/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c +++ b/drivers/gpu/drm/stm/dw_mipi_dsi-stm.c @@ -129,7 +129,7 @@ static int dsi_pll_get_params(int clkin_khz, int clkout_khz, int fvco_min, fvco_max, delta, best_delta; /* all in khz */ /* Early checks preventing division by 0 & odd results */ - if ((clkin_khz <= 0) || (clkout_khz <= 0)) + if (clkin_khz <= 0 || clkout_khz <= 0) return -EINVAL; fvco_min = LANE_MIN_KBPS * 2 * ODF_MAX; @@ -155,7 +155,7 @@ static int dsi_pll_get_params(int clkin_khz, int clkout_khz, for (o = ODF_MIN; o <= ODF_MAX; o *= 2) { n = DIV_ROUND_CLOSEST(i * o * clkout_khz, clkin_khz); /* Check ndiv according to vco range */ - if ((n < n_min) || (n > n_max)) + if (n < n_min || n > n_max) continue; /* Check if new delta is better & saves parameters */ delta = dsi_pll_get_clkout_khz(clkin_khz, i, n, o) - diff --git a/drivers/gpu/drm/stm/ltdc.c b/drivers/gpu/drm/stm/ltdc.c index 735c9081202a8..7be6710a2450b 100644 --- a/drivers/gpu/drm/stm/ltdc.c +++ b/drivers/gpu/drm/stm/ltdc.c @@ -556,7 +556,7 @@ static int ltdc_plane_atomic_check(struct drm_plane *plane, src_h = state->src_h >> 16; /* Reject scaling */ - if ((src_w != state->crtc_w) || (src_h != state->crtc_h)) { + if (src_w != state->crtc_w || src_h != state->crtc_h) { DRM_ERROR("Scaling is not supported"); return -EINVAL; } -- GitLab From 2ac9d363376a3bae4854ddc1089442d63bf6c3b0 Mon Sep 17 00:00:00 2001 From: Philippe CORNU Date: Thu, 26 Oct 2017 13:48:08 +0200 Subject: [PATCH 0128/1471] dt-bindings: display: stm32: add a 2nd endpoint ltdc can have up to 2 endpoints: - dpi external gpios: for rgb panels or external bridge ICs. - dpi internal ios: connected internally to dsi. Note: Refer to the reference manual to know if the dsi is present on your device. Signed-off-by: Philippe Cornu Acked-by: Rob Herring Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1509018489-19641-2-git-send-email-philippe.cornu@st.com --- Documentation/devicetree/bindings/display/st,stm32-ltdc.txt | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/Documentation/devicetree/bindings/display/st,stm32-ltdc.txt b/Documentation/devicetree/bindings/display/st,stm32-ltdc.txt index 74b5ac7b26d65..029252253ad43 100644 --- a/Documentation/devicetree/bindings/display/st,stm32-ltdc.txt +++ b/Documentation/devicetree/bindings/display/st,stm32-ltdc.txt @@ -10,7 +10,11 @@ - "lcd" for the clock feeding the output pixel clock & IP clock. - resets: reset to be used by the device (defined by use of RCC macro). Required nodes: - - Video port for RGB output. + - Video port for DPI RGB output: ltdc has one video port with up to 2 + endpoints: + - for external dpi rgb panel or bridge, using gpios. + - for internal dpi input of the MIPI DSI host controller. + Note: These 2 endpoints cannot be activated simultaneously. * STMicroelectronics STM32 DSI controller specific extensions to Synopsys DesignWare MIPI DSI host controller -- GitLab From c188d7ebbebd0b21a071777ac937c0743e2e7a64 Mon Sep 17 00:00:00 2001 From: Philippe CORNU Date: Thu, 26 Oct 2017 13:48:09 +0200 Subject: [PATCH 0129/1471] drm/stm: ltdc: add a 2nd endpoint ltdc can have up to 2 endpoints: - dpi external gpios: for rgb panels or external bridge ICs. - dpi internal ios: connected internally to dsi. Note: Refer to the reference manual to know if the dsi is present on your device. Signed-off-by: Philippe Cornu Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1509018489-19641-3-git-send-email-philippe.cornu@st.com --- drivers/gpu/drm/stm/ltdc.c | 64 +++++++++++++++++++++++++++----------- 1 file changed, 45 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/stm/ltdc.c b/drivers/gpu/drm/stm/ltdc.c index 7be6710a2450b..35e884239f60b 100644 --- a/drivers/gpu/drm/stm/ltdc.c +++ b/drivers/gpu/drm/stm/ltdc.c @@ -33,6 +33,8 @@ #define MAX_IRQ 4 +#define MAX_ENDPOINTS 2 + #define HWVER_10200 0x010200 #define HWVER_10300 0x010300 #define HWVER_20101 0x020101 @@ -856,18 +858,33 @@ int ltdc_load(struct drm_device *ddev) struct ltdc_device *ldev = ddev->dev_private; struct device *dev = ddev->dev; struct device_node *np = dev->of_node; - struct drm_bridge *bridge; - struct drm_panel *panel; + struct drm_bridge *bridge[MAX_ENDPOINTS] = {NULL}; + struct drm_panel *panel[MAX_ENDPOINTS] = {NULL}; struct drm_crtc *crtc; struct reset_control *rstc; struct resource *res; - int irq, ret, i; + int irq, ret, i, endpoint_not_ready = -ENODEV; DRM_DEBUG_DRIVER("\n"); - ret = drm_of_find_panel_or_bridge(np, 0, 0, &panel, &bridge); - if (ret) - return ret; + /* Get endpoints if any */ + for (i = 0; i < MAX_ENDPOINTS; i++) { + ret = drm_of_find_panel_or_bridge(np, 0, i, &panel[i], + &bridge[i]); + + /* + * If at least one endpoint is ready, continue probing, + * else if at least one endpoint is -EPROBE_DEFER and + * there is no previous ready endpoints, defer probing. + */ + if (!ret) + endpoint_not_ready = 0; + else if (ret == -EPROBE_DEFER && endpoint_not_ready) + endpoint_not_ready = -EPROBE_DEFER; + } + + if (endpoint_not_ready) + return endpoint_not_ready; rstc = devm_reset_control_get_exclusive(dev, NULL); @@ -928,19 +945,25 @@ int ltdc_load(struct drm_device *ddev) DRM_INFO("ltdc hw version 0x%08x - ready\n", ldev->caps.hw_version); - if (panel) { - bridge = drm_panel_bridge_add(panel, DRM_MODE_CONNECTOR_DPI); - if (IS_ERR(bridge)) { - DRM_ERROR("Failed to create panel-bridge\n"); - ret = PTR_ERR(bridge); - goto err; + /* Add endpoints panels or bridges if any */ + for (i = 0; i < MAX_ENDPOINTS; i++) { + if (panel[i]) { + bridge[i] = drm_panel_bridge_add(panel[i], + DRM_MODE_CONNECTOR_DPI); + if (IS_ERR(bridge[i])) { + DRM_ERROR("panel-bridge endpoint %d\n", i); + ret = PTR_ERR(bridge[i]); + goto err; + } } - } - ret = ltdc_encoder_init(ddev, bridge); - if (ret) { - DRM_ERROR("Failed to init encoder\n"); - goto err; + if (bridge[i]) { + ret = ltdc_encoder_init(ddev, bridge[i]); + if (ret) { + DRM_ERROR("init encoder endpoint %d\n", i); + goto err; + } + } } crtc = devm_kzalloc(dev, sizeof(*crtc), GFP_KERNEL); @@ -968,7 +991,8 @@ int ltdc_load(struct drm_device *ddev) return 0; err: - drm_panel_bridge_remove(bridge); + for (i = 0; i < MAX_ENDPOINTS; i++) + drm_panel_bridge_remove(bridge[i]); clk_disable_unprepare(ldev->pixel_clk); @@ -978,10 +1002,12 @@ int ltdc_load(struct drm_device *ddev) void ltdc_unload(struct drm_device *ddev) { struct ltdc_device *ldev = ddev->dev_private; + int i; DRM_DEBUG_DRIVER("\n"); - drm_of_panel_bridge_remove(ddev->dev->of_node, 0, 0); + for (i = 0; i < MAX_ENDPOINTS; i++) + drm_of_panel_bridge_remove(ddev->dev->of_node, 0, i); clk_disable_unprepare(ldev->pixel_clk); } -- GitLab From 2dff0a0ce52bd4c07f3f8f843c25b6250a599361 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 10:41:31 +0000 Subject: [PATCH 0130/1471] drm/selftests/mm: Insert cond_resched() between insert modes kbuilder has begun running the selftests and reported a soft-lockup inside __igt_insert(), so break up the test loop over different modes with another call to cond_resched(). Reported-by: Fengguang Wu Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Signed-off-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171107104131.5923-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/selftests/test-drm_mm.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/gpu/drm/selftests/test-drm_mm.c b/drivers/gpu/drm/selftests/test-drm_mm.c index 86eb4c185a285..7cc935d7b7aaa 100644 --- a/drivers/gpu/drm/selftests/test-drm_mm.c +++ b/drivers/gpu/drm/selftests/test-drm_mm.c @@ -682,6 +682,8 @@ static int __igt_insert(unsigned int count, u64 size, bool replace) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + + cond_resched(); } ret = 0; @@ -944,6 +946,8 @@ static int __igt_insert_range(unsigned int count, u64 size, u64 start, u64 end) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + + cond_resched(); } ret = 0; @@ -1068,6 +1072,7 @@ static int igt_align(void *ignored) drm_mm_for_each_node_safe(node, next, &mm) drm_mm_remove_node(node); DRM_MM_BUG_ON(!drm_mm_clean(&mm)); + cond_resched(); } -- GitLab From f7da778522ef15201783678e1dc01a59ba462cdb Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:20:57 +0200 Subject: [PATCH 0131/1471] drm/edid: use macros for ELD offsets and values MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We have the macros, use them. No functional changes. Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/36f00f018280a04d2db806574d337f7512986c86.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/drm_edid.c | 27 ++++++++++++++------------- 1 file changed, 14 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 00ddabfbf9804..6229735ecc15f 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -3756,8 +3756,8 @@ drm_parse_hdmi_vsdb_audio(struct drm_connector *connector, const u8 *db) { u8 len = cea_db_payload_len(db); - if (len >= 6) - connector->eld[5] |= (db[6] >> 7) << 1; /* Supports_AI */ + if (len >= 6 && (db[6] & (1 << 7))) + connector->eld[DRM_ELD_SAD_COUNT_CONN_TYPE] |= DRM_ELD_SUPPORTS_AI; if (len >= 8) { connector->latency_present[0] = db[8] >> 7; connector->latency_present[1] = (db[8] >> 6) & 1; @@ -3865,17 +3865,18 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) return; } - mnl = get_monitor_name(edid, eld + 20); + mnl = get_monitor_name(edid, &eld[DRM_ELD_MONITOR_NAME_STRING]); + DRM_DEBUG_KMS("ELD monitor %s\n", &eld[DRM_ELD_MONITOR_NAME_STRING]); - eld[4] = (cea[1] << 5) | mnl; - DRM_DEBUG_KMS("ELD monitor %s\n", eld + 20); + eld[DRM_ELD_CEA_EDID_VER_MNL] = cea[1] << DRM_ELD_CEA_EDID_VER_SHIFT; + eld[DRM_ELD_CEA_EDID_VER_MNL] |= mnl; - eld[0] = 2 << 3; /* ELD version: 2 */ + eld[DRM_ELD_VER] = DRM_ELD_VER_CEA861D; - eld[16] = edid->mfg_id[0]; - eld[17] = edid->mfg_id[1]; - eld[18] = edid->prod_code[0]; - eld[19] = edid->prod_code[1]; + eld[DRM_ELD_MANUFACTURER_NAME0] = edid->mfg_id[0]; + eld[DRM_ELD_MANUFACTURER_NAME1] = edid->mfg_id[1]; + eld[DRM_ELD_PRODUCT_CODE0] = edid->prod_code[0]; + eld[DRM_ELD_PRODUCT_CODE1] = edid->prod_code[1]; if (cea_revision(cea) >= 3) { int i, start, end; @@ -3896,14 +3897,14 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) /* Audio Data Block, contains SADs */ sad_count = min(dbl / 3, 15 - total_sad_count); if (sad_count >= 1) - memcpy(eld + 20 + mnl + total_sad_count * 3, + memcpy(&eld[DRM_ELD_CEA_SAD(mnl, total_sad_count)], &db[1], sad_count * 3); total_sad_count += sad_count; break; case SPEAKER_BLOCK: /* Speaker Allocation Data Block */ if (dbl >= 1) - eld[7] = db[1]; + eld[DRM_ELD_SPEAKER] = db[1]; break; case VENDOR_BLOCK: /* HDMI Vendor-Specific Data Block */ @@ -3915,7 +3916,7 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) } } } - eld[5] |= total_sad_count << 4; + eld[DRM_ELD_SAD_COUNT_CONN_TYPE] |= total_sad_count << DRM_ELD_SAD_COUNT_SHIFT; eld[DRM_ELD_BASELINE_ELD_LEN] = DIV_ROUND_UP(drm_eld_calc_baseline_block_size(eld), 4); -- GitLab From 1d1c36650752b7fb81cee515a9bba4131cac4b7c Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:20:58 +0200 Subject: [PATCH 0132/1471] drm/edid: set ELD connector type in drm_edid_to_eld() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since drm_edid_to_eld() knows the connector type, we can set the type in ELD while at it. Most connectors this gets called on are not DP encoders, and with the HDMI type being 0, this does not change behaviour for non-DP. For i915 having this in place earlier would have saved a considerable amount of debugging that lead to the fix 2d8f63297b9f ("drm/i915: always update ELD connector type after get modes"). I don't see other drivers, even the ones calling drm_edid_to_eld() on DP connectors, setting the connector type in ELD. Cc: Alex Deucher Cc: Christian König Cc: Archit Taneja Cc: Andrzej Hajda Cc: Russell King Cc: CK Hu Cc: Philipp Zabel Cc: Ben Skeggs Cc: Mark Yao Cc: Benjamin Gaignard Cc: Vincent Abriou Cc: Thierry Reding Cc: Eric Anholt Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/d527b31619528c477c2c136f25cdf118bc0cfc1d.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/drm_edid.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 6229735ecc15f..3139c2e90e32a 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -3835,8 +3835,7 @@ EXPORT_SYMBOL(drm_edid_get_monitor_name); * @edid: EDID to parse * * Fill the ELD (EDID-Like Data) buffer for passing to the audio driver. The - * Conn_Type, HDCP and Port_ID ELD fields are left for the graphics driver to - * fill in. + * HDCP and Port_ID ELD fields are left for the graphics driver to fill in. */ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) { @@ -3918,6 +3917,12 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) } eld[DRM_ELD_SAD_COUNT_CONN_TYPE] |= total_sad_count << DRM_ELD_SAD_COUNT_SHIFT; + if (connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort || + connector->connector_type == DRM_MODE_CONNECTOR_eDP) + eld[DRM_ELD_SAD_COUNT_CONN_TYPE] |= DRM_ELD_CONN_TYPE_DP; + else + eld[DRM_ELD_SAD_COUNT_CONN_TYPE] |= DRM_ELD_CONN_TYPE_HDMI; + eld[DRM_ELD_BASELINE_ELD_LEN] = DIV_ROUND_UP(drm_eld_calc_baseline_block_size(eld), 4); -- GitLab From 42750d39ad51691dc363cad2c221209449b24884 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:21:00 +0200 Subject: [PATCH 0133/1471] drm/edid: abstract connector ELD clearing MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Preparation for future work. No functional changes. Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/95e83e7a62566b65a56c8cb5038ddf0c6b8e5e95.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/drm_edid.c | 21 +++++++++++++-------- 1 file changed, 13 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 3139c2e90e32a..3162ea58e4500 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -3829,6 +3829,18 @@ void drm_edid_get_monitor_name(struct edid *edid, char *name, int bufsize) } EXPORT_SYMBOL(drm_edid_get_monitor_name); +static void clear_eld(struct drm_connector *connector) +{ + memset(connector->eld, 0, sizeof(connector->eld)); + + connector->latency_present[0] = false; + connector->latency_present[1] = false; + connector->video_latency[0] = 0; + connector->audio_latency[0] = 0; + connector->video_latency[1] = 0; + connector->audio_latency[1] = 0; +} + /** * drm_edid_to_eld - build ELD from EDID * @connector: connector corresponding to the HDMI/DP sink @@ -3846,14 +3858,7 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) int mnl; int dbl; - memset(eld, 0, sizeof(connector->eld)); - - connector->latency_present[0] = false; - connector->latency_present[1] = false; - connector->video_latency[0] = 0; - connector->audio_latency[0] = 0; - connector->video_latency[1] = 0; - connector->audio_latency[1] = 0; + clear_eld(connector); if (!edid) return; -- GitLab From c945b8c14bb7cefe8ac31c63eaf8bad521df643e Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:21:01 +0200 Subject: [PATCH 0134/1471] drm/edid: build ELD in drm_add_edid_modes() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Call drm_edid_to_eld() from drm_add_edid_modes() to fill in the ELD automatically. There's no harm in doing this for connectors that do not support audio. Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/551b2e1cf19be04c510e7865d7539cfc2f54ea89.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/drm_edid.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 3162ea58e4500..4e3ef3d91b956 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -4612,8 +4612,8 @@ static int add_displayid_detailed_modes(struct drm_connector *connector, * @edid: EDID data * * Add the specified modes to the connector's mode list. Also fills out the - * &drm_display_info structure in @connector with any information which can be - * derived from the edid. + * &drm_display_info structure and ELD in @connector with any information which + * can be derived from the edid. * * Return: The number of modes added or 0 if we couldn't find any. */ @@ -4623,9 +4623,11 @@ int drm_add_edid_modes(struct drm_connector *connector, struct edid *edid) u32 quirks; if (edid == NULL) { + clear_eld(connector); return 0; } if (!drm_edid_is_valid(edid)) { + clear_eld(connector); dev_warn(connector->dev->dev, "%s: EDID invalid.\n", connector->name); return 0; @@ -4633,6 +4635,8 @@ int drm_add_edid_modes(struct drm_connector *connector, struct edid *edid) quirks = edid_get_quirks(edid); + drm_edid_to_eld(connector, edid); + /* * CEA-861-F adds ycbcr capability map block, for HDMI 2.0 sinks. * To avoid multiple parsing of same block, lets parse that map -- GitLab From d471ed04b487c6e66a406bf3763efbfed56baa5b Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:21:02 +0200 Subject: [PATCH 0135/1471] drm/drivers: drop redundant drm_edid_to_eld() calls MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_add_edid_modes() now fills in the ELD automatically, so the calls to drm_edid_to_eld() are redundant. Remove them. All the other places are obvious, but nv50 has detached drm_edid_to_eld() from the drm_add_edid_modes() call. Cc: Alex Deucher Cc: Christian König Cc: Archit Taneja Cc: Andrzej Hajda Cc: Russell King Cc: CK Hu Cc: Philipp Zabel Cc: Ben Skeggs Cc: Mark Yao Cc: Benjamin Gaignard Cc: Vincent Abriou Cc: Thierry Reding Cc: Eric Anholt Acked-by: Eric Anholt Acked-by: Archit Taneja Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/0959ca02b983afc9e74dd9acd190ba6e25f21678.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c | 1 - drivers/gpu/drm/bridge/analogix-anx78xx.c | 2 -- drivers/gpu/drm/bridge/synopsys/dw-hdmi.c | 2 -- drivers/gpu/drm/i2c/tda998x_drv.c | 1 - drivers/gpu/drm/i915/intel_dp.c | 1 - drivers/gpu/drm/i915/intel_modes.c | 1 - drivers/gpu/drm/mediatek/mtk_hdmi.c | 1 - drivers/gpu/drm/nouveau/nv50_display.c | 5 +---- drivers/gpu/drm/radeon/radeon_connectors.c | 1 - drivers/gpu/drm/radeon/radeon_dp_mst.c | 1 - drivers/gpu/drm/rockchip/cdn-dp-core.c | 4 +--- drivers/gpu/drm/sti/sti_hdmi.c | 1 - drivers/gpu/drm/tegra/output.c | 1 - drivers/gpu/drm/vc4/vc4_hdmi.c | 1 - 14 files changed, 2 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index f51b41f094ef2..126be9e0f05f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -358,7 +358,6 @@ static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) if (amdgpu_connector->edid) { drm_mode_connector_update_edid_property(connector, amdgpu_connector->edid); ret = drm_add_edid_modes(connector, amdgpu_connector->edid); - drm_edid_to_eld(connector, amdgpu_connector->edid); return ret; } drm_mode_connector_update_edid_property(connector, NULL); diff --git a/drivers/gpu/drm/bridge/analogix-anx78xx.c b/drivers/gpu/drm/bridge/analogix-anx78xx.c index 9385eb0b1ee40..ed12a7ddd64ae 100644 --- a/drivers/gpu/drm/bridge/analogix-anx78xx.c +++ b/drivers/gpu/drm/bridge/analogix-anx78xx.c @@ -977,8 +977,6 @@ static int anx78xx_get_modes(struct drm_connector *connector) } num_modes = drm_add_edid_modes(connector, anx78xx->edid); - /* Store the ELD */ - drm_edid_to_eld(connector, anx78xx->edid); unlock: mutex_unlock(&anx78xx->lock); diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c index bf14214fa4640..a64ce71122881 100644 --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c @@ -1910,8 +1910,6 @@ static int dw_hdmi_connector_get_modes(struct drm_connector *connector) drm_mode_connector_update_edid_property(connector, edid); cec_notifier_set_phys_addr_from_edid(hdmi->cec_notifier, edid); ret = drm_add_edid_modes(connector, edid); - /* Store the ELD */ - drm_edid_to_eld(connector, edid); kfree(edid); } else { dev_dbg(hdmi->dev, "failed to get edid\n"); diff --git a/drivers/gpu/drm/i2c/tda998x_drv.c b/drivers/gpu/drm/i2c/tda998x_drv.c index 4d1f45acf2cdb..60981505763cb 100644 --- a/drivers/gpu/drm/i2c/tda998x_drv.c +++ b/drivers/gpu/drm/i2c/tda998x_drv.c @@ -1100,7 +1100,6 @@ static int tda998x_connector_get_modes(struct drm_connector *connector) drm_mode_connector_update_edid_property(connector, edid); n = drm_add_edid_modes(connector, edid); - drm_edid_to_eld(connector, edid); kfree(edid); diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 887953c0f495d..6fce08209b4f6 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -5865,7 +5865,6 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, if (drm_add_edid_modes(connector, edid)) { drm_mode_connector_update_edid_property(connector, edid); - drm_edid_to_eld(connector, edid); } else { kfree(edid); edid = ERR_PTR(-EINVAL); diff --git a/drivers/gpu/drm/i915/intel_modes.c b/drivers/gpu/drm/i915/intel_modes.c index 951e834dd2744..b39846613e3c7 100644 --- a/drivers/gpu/drm/i915/intel_modes.c +++ b/drivers/gpu/drm/i915/intel_modes.c @@ -42,7 +42,6 @@ int intel_connector_update_modes(struct drm_connector *connector, drm_mode_connector_update_edid_property(connector, edid); ret = drm_add_edid_modes(connector, edid); - drm_edid_to_eld(connector, edid); return ret; } diff --git a/drivers/gpu/drm/mediatek/mtk_hdmi.c b/drivers/gpu/drm/mediatek/mtk_hdmi.c index 3ff502771ba2f..b787910619839 100644 --- a/drivers/gpu/drm/mediatek/mtk_hdmi.c +++ b/drivers/gpu/drm/mediatek/mtk_hdmi.c @@ -1222,7 +1222,6 @@ static int mtk_hdmi_conn_get_modes(struct drm_connector *conn) drm_mode_connector_update_edid_property(conn, edid); ret = drm_add_edid_modes(conn, edid); - drm_edid_to_eld(conn, edid); kfree(edid); return ret; } diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 2dbf62a2ac413..d8ddb73967216 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -2688,7 +2688,6 @@ nv50_audio_enable(struct drm_encoder *encoder, struct drm_display_mode *mode) if (!drm_detect_monitor_audio(nv_connector->edid)) return; - drm_edid_to_eld(&nv_connector->base, nv_connector->edid); memcpy(args.data, nv_connector->base.eld, sizeof(args.data)); nvif_mthd(disp->disp, 0, &args, @@ -3064,10 +3063,8 @@ nv50_mstc_get_modes(struct drm_connector *connector) mstc->edid = drm_dp_mst_get_edid(&mstc->connector, mstc->port->mgr, mstc->port); drm_mode_connector_update_edid_property(&mstc->connector, mstc->edid); - if (mstc->edid) { + if (mstc->edid) ret = drm_add_edid_modes(&mstc->connector, mstc->edid); - drm_edid_to_eld(&mstc->connector, mstc->edid); - } if (!mstc->connector.display_info.bpc) mstc->connector.display_info.bpc = 8; diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c index 2f642cbefd8ea..9e45127329d76 100644 --- a/drivers/gpu/drm/radeon/radeon_connectors.c +++ b/drivers/gpu/drm/radeon/radeon_connectors.c @@ -392,7 +392,6 @@ static int radeon_ddc_get_modes(struct drm_connector *connector) if (radeon_connector->edid) { drm_mode_connector_update_edid_property(connector, radeon_connector->edid); ret = drm_add_edid_modes(connector, radeon_connector->edid); - drm_edid_to_eld(connector, radeon_connector->edid); return ret; } drm_mode_connector_update_edid_property(connector, NULL); diff --git a/drivers/gpu/drm/radeon/radeon_dp_mst.c b/drivers/gpu/drm/radeon/radeon_dp_mst.c index ebdf1b859cb6c..1768926d2ebc0 100644 --- a/drivers/gpu/drm/radeon/radeon_dp_mst.c +++ b/drivers/gpu/drm/radeon/radeon_dp_mst.c @@ -196,7 +196,6 @@ static int radeon_dp_mst_get_ddc_modes(struct drm_connector *connector) if (radeon_connector->edid) { drm_mode_connector_update_edid_property(&radeon_connector->base, radeon_connector->edid); ret = drm_add_edid_modes(&radeon_connector->base, radeon_connector->edid); - drm_edid_to_eld(&radeon_connector->base, radeon_connector->edid); return ret; } drm_mode_connector_update_edid_property(&radeon_connector->base, NULL); diff --git a/drivers/gpu/drm/rockchip/cdn-dp-core.c b/drivers/gpu/drm/rockchip/cdn-dp-core.c index 275844d0d0ec2..ec999d9f15f6d 100644 --- a/drivers/gpu/drm/rockchip/cdn-dp-core.c +++ b/drivers/gpu/drm/rockchip/cdn-dp-core.c @@ -276,11 +276,9 @@ static int cdn_dp_connector_get_modes(struct drm_connector *connector) dp->sink_has_audio = drm_detect_monitor_audio(edid); ret = drm_add_edid_modes(connector, edid); - if (ret) { + if (ret) drm_mode_connector_update_edid_property(connector, edid); - drm_edid_to_eld(connector, edid); - } } mutex_unlock(&dp->lock); diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index 30f02d2fdd034..d1902750a85da 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -976,7 +976,6 @@ static int sti_hdmi_connector_get_modes(struct drm_connector *connector) count = drm_add_edid_modes(connector, edid); drm_mode_connector_update_edid_property(connector, edid); - drm_edid_to_eld(connector, edid); kfree(edid); return count; diff --git a/drivers/gpu/drm/tegra/output.c b/drivers/gpu/drm/tegra/output.c index 595d1ec3e02e8..1480f6aaffe49 100644 --- a/drivers/gpu/drm/tegra/output.c +++ b/drivers/gpu/drm/tegra/output.c @@ -36,7 +36,6 @@ int tegra_output_connector_get_modes(struct drm_connector *connector) if (edid) { err = drm_add_edid_modes(connector, edid); - drm_edid_to_eld(connector, edid); kfree(edid); } diff --git a/drivers/gpu/drm/vc4/vc4_hdmi.c b/drivers/gpu/drm/vc4/vc4_hdmi.c index fa37a1c07cf69..9a9a6b4acccfd 100644 --- a/drivers/gpu/drm/vc4/vc4_hdmi.c +++ b/drivers/gpu/drm/vc4/vc4_hdmi.c @@ -287,7 +287,6 @@ static int vc4_hdmi_connector_get_modes(struct drm_connector *connector) drm_mode_connector_update_edid_property(connector, edid); ret = drm_add_edid_modes(connector, edid); - drm_edid_to_eld(connector, edid); kfree(edid); return ret; -- GitLab From 79436a1c9bccf5e38cb6ea26e4e4b9283baf2e20 Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Wed, 1 Nov 2017 16:21:03 +0200 Subject: [PATCH 0136/1471] drm/edid: make drm_edid_to_eld() static MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is no longer needed outside of drm_edid.c. Reviewed-by: Ville Syrjälä Reviewed-by: Alex Deucher Acked-by: Thierry Reding Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/9c0be7b97d0144ed0419c87ac42b30f5835ca7e6.1509545641.git.jani.nikula@intel.com --- drivers/gpu/drm/drm_edid.c | 5 ++--- include/drm/drm_edid.h | 1 - include/drm/drm_modeset_helper_vtables.h | 3 --- 3 files changed, 2 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 4e3ef3d91b956..749d07a017729 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -3841,7 +3841,7 @@ static void clear_eld(struct drm_connector *connector) connector->audio_latency[1] = 0; } -/** +/* * drm_edid_to_eld - build ELD from EDID * @connector: connector corresponding to the HDMI/DP sink * @edid: EDID to parse @@ -3849,7 +3849,7 @@ static void clear_eld(struct drm_connector *connector) * Fill the ELD (EDID-Like Data) buffer for passing to the audio driver. The * HDCP and Port_ID ELD fields are left for the graphics driver to fill in. */ -void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) +static void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) { uint8_t *eld = connector->eld; u8 *cea; @@ -3934,7 +3934,6 @@ void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid) DRM_DEBUG_KMS("ELD size %d, SAD count %d\n", drm_eld_size(eld), total_sad_count); } -EXPORT_SYMBOL(drm_edid_to_eld); /** * drm_edid_to_sad - extracts SADs from EDID diff --git a/include/drm/drm_edid.h b/include/drm/drm_edid.h index 6f35909b8add3..9e4e23524840e 100644 --- a/include/drm/drm_edid.h +++ b/include/drm/drm_edid.h @@ -333,7 +333,6 @@ struct drm_encoder; struct drm_connector; struct drm_display_mode; -void drm_edid_to_eld(struct drm_connector *connector, struct edid *edid); int drm_edid_to_sad(struct edid *edid, struct cea_sad **sads); int drm_edid_to_speaker_allocation(struct edid *edid, u8 **sadb); int drm_av_sync_delay(struct drm_connector *connector, diff --git a/include/drm/drm_modeset_helper_vtables.h b/include/drm/drm_modeset_helper_vtables.h index 16646c44b7df7..3e76ca805b0f4 100644 --- a/include/drm/drm_modeset_helper_vtables.h +++ b/include/drm/drm_modeset_helper_vtables.h @@ -801,9 +801,6 @@ struct drm_connector_helper_funcs { * resolution can call drm_add_modes_noedid(), and mark the preferred * one using drm_set_preferred_mode(). * - * Finally drivers that support audio probably want to update the ELD - * data, too, using drm_edid_to_eld(). - * * This function is only called after the @detect hook has indicated * that a sink is connected and when the EDID isn't overridden through * sysfs or the kernel commandline. -- GitLab From 0ffba1fc98e8ec35caae8d50b657296ebb9a9a51 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 15:40:55 +0000 Subject: [PATCH 0137/1471] drm/i915: Silence smatch for cmdparser drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:808:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:811:23: error: not an lvalue drivers/gpu/drm/i915/i915_cmd_parser.c:814:23: error: not an lvalue If we move the shift into each case not only do we kill the warning from smatch, but we shrink the code slightly: text data bss dec hex filename 1267906 20587 3168 1291661 13b58d before 1267890 20587 3168 1291645 13b57d after Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171107154055.19460-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld Reviewed-by: Gabriel Krisman Bertazi --- drivers/gpu/drm/i915/i915_cmd_parser.c | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index 8ba932b22f7c8..b11629beeb634 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -798,22 +798,15 @@ struct cmd_node { */ static inline u32 cmd_header_key(u32 x) { - u32 shift; - switch (x >> INSTR_CLIENT_SHIFT) { default: case INSTR_MI_CLIENT: - shift = STD_MI_OPCODE_SHIFT; - break; + return x >> STD_MI_OPCODE_SHIFT; case INSTR_RC_CLIENT: - shift = STD_3D_OPCODE_SHIFT; - break; + return x >> STD_3D_OPCODE_SHIFT; case INSTR_BC_CLIENT: - shift = STD_2D_OPCODE_SHIFT; - break; + return x >> STD_2D_OPCODE_SHIFT; } - - return x >> shift; } static int init_hash_table(struct intel_engine_cs *engine, -- GitLab From 2f59f1b33a84447247b21be60ecf05080ce2982a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 14:53:34 +0000 Subject: [PATCH 0138/1471] drm/i915: Silence compiler for csr_load_work_fn() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit gcc-4.7 is not very smart and can not tell that "si" is guarded by size being 0. So it complains, drivers/gpu/drm/i915/intel_csr.c: In function ‘csr_load_work_fn’: drivers/gpu/drm/i915/intel_csr.c:204:3: warning: ‘si’ may be used uninitialized in this function [-Wmaybe-uninitialized] drivers/gpu/drm/i915/intel_csr.c:190:30: note: ‘si’ was declared in Give in and mark si as NULL. Signed-off-by: Chris Wilson Cc: Anusha Srivatsa Cc: Rodrigo Vivi Cc: Imre Deak Link: https://patchwork.freedesktop.org/patch/msgid/20171107145334.27154-1-chris@chris-wilson.co.uk Reviewed-by: Gabriel Krisman Bertazi Reviewed-by: Michal Wajdeczko --- drivers/gpu/drm/i915/intel_csr.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c index 3e1f86d0c6cc2..77d8b3d483ca2 100644 --- a/drivers/gpu/drm/i915/intel_csr.c +++ b/drivers/gpu/drm/i915/intel_csr.c @@ -198,6 +198,7 @@ intel_get_stepping_info(struct drm_i915_private *dev_priv) si = bxt_stepping_info; } else { size = 0; + si = NULL; } if (INTEL_REVID(dev_priv) < size) -- GitLab From 0e870bf872929a026c58a74608734b89a95dda74 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 13:53:24 +0000 Subject: [PATCH 0139/1471] drm/i915: Simplify onion for bxt_ddi_phy_init() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Older compilers (gcc-4.9) are not as able to track uninitialised variables as well as more recent compilers. In particular, drivers/gpu/drm/i915/intel_dpio_phy.c: In function ‘bxt_ddi_phy_init’: drivers/gpu/drm/i915/intel_dpio_phy.c:482:25: warning: ‘was_enabled’ may be used uninitialized in this function [-Wmaybe-uninitialized] In this case, we can rearrange code slightly to make the control flow clearer to the reader, as well as the compiler. That is we only call uninit using the same predicate as calling init Signed-off-by: Chris Wilson Cc: Ander Conselvan de Oliveira Cc: Imre Deak Link: https://patchwork.freedesktop.org/patch/msgid/20171107135324.28300-1-chris@chris-wilson.co.uk Reviewed-by: Gabriel Krisman Bertazi --- drivers/gpu/drm/i915/intel_dpio_phy.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpio_phy.c b/drivers/gpu/drm/i915/intel_dpio_phy.c index 63b76eac018fa..490a42d9a3e8b 100644 --- a/drivers/gpu/drm/i915/intel_dpio_phy.c +++ b/drivers/gpu/drm/i915/intel_dpio_phy.c @@ -466,21 +466,21 @@ void bxt_ddi_phy_init(struct drm_i915_private *dev_priv, enum dpio_phy phy) lockdep_assert_held(&dev_priv->power_domains.lock); - if (rcomp_phy != -1) { + was_enabled = true; + if (rcomp_phy != -1) was_enabled = bxt_ddi_phy_is_enabled(dev_priv, rcomp_phy); - /* - * We need to copy the GRC calibration value from rcomp_phy, - * so make sure it's powered up. - */ - if (!was_enabled) - _bxt_ddi_phy_init(dev_priv, rcomp_phy); - } + /* + * We need to copy the GRC calibration value from rcomp_phy, + * so make sure it's powered up. + */ + if (!was_enabled) + _bxt_ddi_phy_init(dev_priv, rcomp_phy); _bxt_ddi_phy_init(dev_priv, phy); - if (rcomp_phy != -1 && !was_enabled) - bxt_ddi_phy_uninit(dev_priv, phy_info->rcomp_phy); + if (!was_enabled) + bxt_ddi_phy_uninit(dev_priv, rcomp_phy); } static bool __printf(6, 7) -- GitLab From 1a1f12872edcd5e425b668a35fb23548cfa918ef Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 14:03:38 +0000 Subject: [PATCH 0140/1471] drm/i915: Prevent unbounded wm results in g4x_compute_wm() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Smatch warns of drivers/gpu/drm/i915/intel_pm.c:1161 g4x_compute_wm() warn: signedness bug returning '(-33554430)' which is a result of it believing that wm may be INT_MAX following g4x_tlb_miss_wa(). Just declaring g4x_tlb_miss_wa() as returning an unsigned integer is not sufficient, we need to tell smatch that wm itself is unsigned for it to not worry. So mark up the locals we expect to be non-negative, and so silence smatch. v2: Mark up vlv_compute_wm_level() as unsigned similarly. Signed-off-by: Chris Wilson Cc: Ville Syrjälä Cc: Dan Carpenter Reviewed-by: Ville Syrjälä #v1 Link: https://patchwork.freedesktop.org/patch/msgid/20171107140338.13748-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_pm.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 07118c0b69d37..46440e2ecb33b 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -928,7 +928,7 @@ static void pineview_update_wm(struct intel_crtc *unused_crtc) * and the size of 8 whole lines. This adjustment is always performed * in the actual pixel depth regardless of whether FBC is enabled or not." */ -static int g4x_tlb_miss_wa(int fifo_size, int width, int cpp) +static unsigned int g4x_tlb_miss_wa(int fifo_size, int width, int cpp) { int tlb_miss = fifo_size * 64 - width * cpp * 8; @@ -1105,8 +1105,8 @@ static uint16_t g4x_compute_wm(const struct intel_crtc_state *crtc_state, struct drm_i915_private *dev_priv = to_i915(plane->base.dev); const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; - int clock, htotal, cpp, width, wm; - int latency = dev_priv->wm.pri_latency[level] * 10; + unsigned int latency = dev_priv->wm.pri_latency[level] * 10; + unsigned int clock, htotal, cpp, width, wm; if (latency == 0) return USHRT_MAX; @@ -1145,7 +1145,7 @@ static uint16_t g4x_compute_wm(const struct intel_crtc_state *crtc_state, level == G4X_WM_LEVEL_NORMAL) { wm = intel_wm_method1(clock, cpp, latency); } else { - int small, large; + unsigned int small, large; small = intel_wm_method1(clock, cpp, latency); large = intel_wm_method2(clock, htotal, width, cpp, latency); @@ -1158,7 +1158,7 @@ static uint16_t g4x_compute_wm(const struct intel_crtc_state *crtc_state, wm = DIV_ROUND_UP(wm, 64) + 2; - return min_t(int, wm, USHRT_MAX); + return min_t(unsigned int, wm, USHRT_MAX); } static bool g4x_raw_plane_wm_set(struct intel_crtc_state *crtc_state, @@ -1602,7 +1602,7 @@ static uint16_t vlv_compute_wm_level(const struct intel_crtc_state *crtc_state, struct drm_i915_private *dev_priv = to_i915(plane->base.dev); const struct drm_display_mode *adjusted_mode = &crtc_state->base.adjusted_mode; - int clock, htotal, cpp, width, wm; + unsigned int clock, htotal, cpp, width, wm; if (dev_priv->wm.pri_latency[level] == 0) return USHRT_MAX; @@ -1628,7 +1628,7 @@ static uint16_t vlv_compute_wm_level(const struct intel_crtc_state *crtc_state, dev_priv->wm.pri_latency[level] * 10); } - return min_t(int, wm, USHRT_MAX); + return min_t(unsigned int, wm, USHRT_MAX); } static bool vlv_need_sprite0_fifo_workaround(unsigned int active_planes) -- GitLab From f6d03042b9ad514077175b5cc53069e0a364d0ec Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 11:05:59 +0000 Subject: [PATCH 0141/1471] drm/i915/selftests: Skip mixed page exhaustion if only small pages available If we only have 4k pages, we can't mix together different combinations of hugepages to see if the world burns. However, as the loops did nothing, we never set err to 0 and reported ENODEV aborting the test. Teach the test to skip instead. Signed-off-by: Chris Wilson Cc: Matthew Auld Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171107110559.6098-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/selftests/huge_pages.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index 5cc8101bb2b1a..01af540b6ef9a 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -1159,6 +1159,9 @@ static int igt_ppgtt_exhaust_huge(void *arg) int n, i; int err = -ENODEV; + if (supported == I915_GTT_PAGE_SIZE_4K) + return 0; + /* * Sanity check creating objects with a varying mix of page sizes -- * ensuring that our writes lands in the right place. -- GitLab From c29ccb9f426e6e4e91c9901bb6ee8dbe6abc48eb Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 11:40:51 +0000 Subject: [PATCH 0142/1471] drm/i915/selftests: Take rpm wakeref around GGTT lowlevel tests The vma routines are responsible for acquiring the device rpm wakeref before they poke the HW. However, some of the selftests bypass the higher level vma routines in order to poke directly at the lowlevel GGTT functions; these are then responsible for managing rpm themselves. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171107114051.10583-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index 9da0c9f999167..5812968605392 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -267,7 +267,9 @@ static int lowlevel_hole(struct drm_i915_private *i915, mock_vma.node.size = BIT_ULL(size); mock_vma.node.start = addr; + intel_runtime_pm_get(i915); vm->insert_entries(vm, &mock_vma, I915_CACHE_NONE, 0); + intel_runtime_pm_put(i915); } count = n; @@ -1047,6 +1049,7 @@ static int igt_ggtt_page(void *arg) goto out_remove; } + intel_runtime_pm_get(i915); for (n = 0; n < count; n++) { u64 offset = tmp.start + order[n] * PAGE_SIZE; u32 __iomem *vaddr; @@ -1086,6 +1089,7 @@ static int igt_ggtt_page(void *arg) break; } } + intel_runtime_pm_put(i915); kfree(order); out_remove: -- GitLab From 693b1ccabe43d96a1eb57716ac032632882de20f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 11:56:53 +0000 Subject: [PATCH 0143/1471] drm/i915/selftests: Take rpm wakeref around partial tiling tests Since the partial tiling tests are poking into the GGTT to watch the fence registers in operation, it itself needs the device rpm wakeref in order for the GGTT to remain accessible. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171107115653.10716-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/selftests/i915_gem_object.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c index 1b8774a42e487..f32aa6bb79e29 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c @@ -317,6 +317,7 @@ static int igt_partial_tiling(void *arg) } mutex_lock(&i915->drm.struct_mutex); + intel_runtime_pm_get(i915); if (1) { IGT_TIMEOUT(end); @@ -418,6 +419,7 @@ next_tiling: ; } out_unlock: + intel_runtime_pm_put(i915); mutex_unlock(&i915->drm.struct_mutex); i915_gem_object_unpin_pages(obj); out: -- GitLab From 846c7dfc1193eb2f9866fe2fa0ae7d45c72f95b9 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 1 Nov 2017 16:04:33 +0100 Subject: [PATCH 0144/1471] drm/atomic: Try to preserve the crtc enabled state in drm_atomic_remove_fb, v2. This introduces a slight behavioral change to rmfb. Instead of disabling a crtc when the primary plane is disabled, we try to preserve it. Apart from old versions of the vmwgfx xorg driver, there is nothing depending on rmfb disabling a crtc. Vmwgfx' and simple kms helper atomic implementation rejects CRTC enabled without plane, so we can do this safely. If the atomic commit is rejected by the driver then we will still fall back to the old behavior and turn off the crtc. Changes since v1: - Restart completely when rmfb with crtc on fails (Sean Paul). Signed-off-by: Maarten Lankhorst Cc: Sean Paul Cc: Daniel Vetter Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171101150433.10777-1-maarten.lankhorst@linux.intel.com Reviewed-by: Daniel Vetter --- drivers/gpu/drm/drm_framebuffer.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/drm_framebuffer.c b/drivers/gpu/drm/drm_framebuffer.c index 40e6565105acb..2305e7800b45a 100644 --- a/drivers/gpu/drm/drm_framebuffer.c +++ b/drivers/gpu/drm/drm_framebuffer.c @@ -765,14 +765,18 @@ static int atomic_remove_fb(struct drm_framebuffer *fb) struct drm_plane *plane; struct drm_connector *conn; struct drm_connector_state *conn_state; - int i, ret = 0; + int i, ret; unsigned plane_mask; + bool disable_crtcs = false; - state = drm_atomic_state_alloc(dev); - if (!state) - return -ENOMEM; - +retry_disable: drm_modeset_acquire_init(&ctx, 0); + + state = drm_atomic_state_alloc(dev); + if (!state) { + ret = -ENOMEM; + goto out; + } state->acquire_ctx = &ctx; retry: @@ -793,7 +797,7 @@ static int atomic_remove_fb(struct drm_framebuffer *fb) goto unlock; } - if (plane_state->crtc->primary == plane) { + if (disable_crtcs && plane_state->crtc->primary == plane) { struct drm_crtc_state *crtc_state; crtc_state = drm_atomic_get_existing_crtc_state(state, plane_state->crtc); @@ -818,6 +822,7 @@ static int atomic_remove_fb(struct drm_framebuffer *fb) plane->old_fb = plane->fb; } + /* This list is only filled when disable_crtcs is set. */ for_each_new_connector_in_state(state, conn, conn_state, i) { ret = drm_atomic_set_crtc_for_connector(conn_state, NULL); @@ -840,9 +845,15 @@ static int atomic_remove_fb(struct drm_framebuffer *fb) drm_atomic_state_put(state); +out: drm_modeset_drop_locks(&ctx); drm_modeset_acquire_fini(&ctx); + if (ret == -EINVAL && !disable_crtcs) { + disable_crtcs = true; + goto retry_disable; + } + return ret; } -- GitLab From cc6b741c6f637484316188ae7f3be211b2f2b33a Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Wed, 8 Nov 2017 14:50:14 +0100 Subject: [PATCH 0145/1471] drm: sti: remove useless fields from vtg structure Drivers are registered on platform bus so the private list could be replace by a call to of_find_device_by_node(). Changing this also makes dev, np and link fields useless in vtg structure. Acked-by: Daniel Vetter Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1510146255-24982-1-git-send-email-benjamin.gaignard@linaro.org --- drivers/gpu/drm/sti/Kconfig | 1 + drivers/gpu/drm/sti/sti_vtg.c | 32 ++++++++------------------------ 2 files changed, 9 insertions(+), 24 deletions(-) diff --git a/drivers/gpu/drm/sti/Kconfig b/drivers/gpu/drm/sti/Kconfig index acd72865feacc..cca4b3c9aeb56 100644 --- a/drivers/gpu/drm/sti/Kconfig +++ b/drivers/gpu/drm/sti/Kconfig @@ -8,5 +8,6 @@ config DRM_STI select DRM_PANEL select FW_LOADER select SND_SOC_HDMI_CODEC if SND_SOC + select OF help Choose this option to enable DRM on STM stiH4xx chipset diff --git a/drivers/gpu/drm/sti/sti_vtg.c b/drivers/gpu/drm/sti/sti_vtg.c index 2dcba1d3a1228..5f75aea7b8fa6 100644 --- a/drivers/gpu/drm/sti/sti_vtg.c +++ b/drivers/gpu/drm/sti/sti_vtg.c @@ -9,6 +9,7 @@ #include #include +#include #include #include @@ -72,8 +73,6 @@ #define AWG_DELAY_ED (-8) #define AWG_DELAY_SD (-7) -static LIST_HEAD(vtg_lookup); - /* * STI VTG register offset structure * @@ -123,42 +122,31 @@ struct sti_vtg_sync_params { /** * STI VTG structure * - * @dev: pointer to device driver - * @np: device node * @regs: register mapping * @sync_params: synchronisation parameters used to generate timings * @irq: VTG irq * @irq_status: store the IRQ status value * @notifier_list: notifier callback * @crtc: the CRTC for vblank event - * @link: List node to link the structure in lookup list */ struct sti_vtg { - struct device *dev; - struct device_node *np; void __iomem *regs; struct sti_vtg_sync_params sync_params[VTG_MAX_SYNC_OUTPUT]; int irq; u32 irq_status; struct raw_notifier_head notifier_list; struct drm_crtc *crtc; - struct list_head link; }; -static void vtg_register(struct sti_vtg *vtg) -{ - list_add_tail(&vtg->link, &vtg_lookup); -} - struct sti_vtg *of_vtg_find(struct device_node *np) { - struct sti_vtg *vtg; + struct platform_device *pdev; - list_for_each_entry(vtg, &vtg_lookup, link) { - if (vtg->np == np) - return vtg; - } - return NULL; + pdev = of_find_device_by_node(np); + if (!pdev) + return NULL; + + return (struct sti_vtg *)platform_get_drvdata(pdev); } static void vtg_reset(struct sti_vtg *vtg) @@ -397,9 +385,6 @@ static int vtg_probe(struct platform_device *pdev) if (!vtg) return -ENOMEM; - vtg->dev = dev; - vtg->np = pdev->dev.of_node; - /* Get Memory ressources */ res = platform_get_resource(pdev, IORESOURCE_MEM, 0); if (!res) { @@ -428,10 +413,9 @@ static int vtg_probe(struct platform_device *pdev) return ret; } - vtg_register(vtg); platform_set_drvdata(pdev, vtg); - DRM_INFO("%s %s\n", __func__, dev_name(vtg->dev)); + DRM_INFO("%s %s\n", __func__, dev_name(dev)); return 0; } -- GitLab From 71469385dbfde768b3431820606f9f0845c9d56c Mon Sep 17 00:00:00 2001 From: Benjamin Gaignard Date: Wed, 8 Nov 2017 14:50:45 +0100 Subject: [PATCH 0146/1471] drm: sti: remove vtg_remove vtg_remove does nothing just remove it Acked-by: Daniel Vetter Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/1510146367-25329-1-git-send-email-benjamin.gaignard@linaro.org --- drivers/gpu/drm/sti/sti_vtg.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/gpu/drm/sti/sti_vtg.c b/drivers/gpu/drm/sti/sti_vtg.c index 5f75aea7b8fa6..3aa2fa6f22284 100644 --- a/drivers/gpu/drm/sti/sti_vtg.c +++ b/drivers/gpu/drm/sti/sti_vtg.c @@ -420,11 +420,6 @@ static int vtg_probe(struct platform_device *pdev) return 0; } -static int vtg_remove(struct platform_device *pdev) -{ - return 0; -} - static const struct of_device_id vtg_of_match[] = { { .compatible = "st,vtg", }, { /* sentinel */ } @@ -438,7 +433,6 @@ struct platform_driver sti_vtg_driver = { .of_match_table = vtg_of_match, }, .probe = vtg_probe, - .remove = vtg_remove, }; MODULE_AUTHOR("Benjamin Gaignard "); -- GitLab From 58ecd9d58939382a352b10d43f06f5a23a36db08 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 5 Nov 2017 13:49:05 +0000 Subject: [PATCH 0147/1471] drm/i915: Read ilk FDI PLL frequency once during initialisation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit During intel_atomic_check(), we do not take the intel_runtime_pm_get() wakeref and so should do the atomic modeset precalculations without referring to the HW. However, on Ironlake we see <7>[ 23.487557] [drm:intel_atomic_check [i915]] [CONNECTOR:47:VGA-1] checking for sink bpp constrains <7>[ 23.487615] [drm:intel_atomic_check [i915]] clamping display bpp (was 36) to default limit of 24 <4>[ 23.487621] RPM wakelock ref not held during HW access <4>[ 23.487652] ------------[ cut here ]------------ <4>[ 23.487697] WARNING: CPU: 0 PID: 1343 at drivers/gpu/drm/i915/intel_drv.h:1813 gen5_read32+0x183/0x200 [i915] <4>[ 23.487701] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic i915 intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul snd_hda_intel ghash_clmulni_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm lpc_ich e1000e mei_me ptp mei pps_core prime_numbers <4>[ 23.487784] CPU: 0 PID: 1343 Comm: debugfs_test Tainted: G W 4.14.0-rc7-CI-Trybot_1378+ #1 <4>[ 23.487788] Hardware name: Hewlett-Packard HP Compaq 8100 Elite SFF PC/304Ah, BIOS 786H1 v01.13 07/14/2011 <4>[ 23.487793] task: ffff8801f90aa6c0 task.stack: ffffc900013ec000 <4>[ 23.487838] RIP: 0010:gen5_read32+0x183/0x200 [i915] <4>[ 23.487842] RSP: 0018:ffffc900013efb58 EFLAGS: 00010292 <4>[ 23.487849] RAX: 000000000000002a RBX: ffff880205c00000 RCX: 0000000000000006 <4>[ 23.487854] RDX: 000000000000140a RSI: ffffffff81d0eb14 RDI: ffffffff81cc26f6 <4>[ 23.487857] RBP: ffffc900013efb80 R08: ffff8801f90aaff8 R09: 0000000000000000 <4>[ 23.487861] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001 <4>[ 23.487865] R13: 0000000000046000 R14: ffff88020ffaba78 R15: ffff88020b109bf8 <4>[ 23.487870] FS: 00007f53b5e40a40(0000) GS:ffff88021bc00000(0000) knlGS:0000000000000000 <4>[ 23.487874] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 <4>[ 23.487878] CR2: 000055e41900c0e8 CR3: 00000001fa0d6005 CR4: 00000000000206f0 <4>[ 23.487882] Call Trace: <4>[ 23.487931] intel_atomic_check+0x745/0x1290 [i915] <4>[ 23.487948] drm_atomic_check_only+0x459/0x560 <4>[ 23.487956] ? drm_atomic_set_crtc_for_connector+0xc9/0x100 <4>[ 23.488025] drm_atomic_commit+0x18/0x50 <4>[ 23.488035] restore_fbdev_mode_atomic+0x190/0x1f0 <4>[ 23.488059] restore_fbdev_mode+0x32/0x120 <4>[ 23.488072] drm_fb_helper_restore_fbdev_mode_unlocked+0x50/0xa0 <4>[ 23.488139] intel_fbdev_restore_mode+0x34/0x90 [i915] <4>[ 23.488194] i915_driver_lastclose+0xe/0x10 [i915] <4>[ 23.488208] drm_lastclose+0x39/0xf0 <4>[ 23.488219] drm_release+0x30c/0x3c0 <4>[ 23.488236] __fput+0xb9/0x200 <4>[ 23.488252] ____fput+0xe/0x10 <4>[ 23.488264] task_work_run+0x89/0xc0 <4>[ 23.488278] exit_to_usermode_loop+0x83/0x90 <4>[ 23.488290] syscall_return_slowpath+0xd0/0x110 <4>[ 23.488304] entry_SYSCALL_64_fastpath+0xaf/0xb1 <4>[ 23.488312] RIP: 0033:0x7f53b4317560 <4>[ 23.488320] RSP: 002b:00007ffca7e70748 EFLAGS: 00000246 ORIG_RAX: 0000000000000003 <4>[ 23.488333] RAX: 0000000000000000 RBX: 0000000000000001 RCX: 00007f53b4317560 <4>[ 23.488340] RDX: 0000000000000005 RSI: 00007ffca7e70640 RDI: 0000000000000004 <4>[ 23.488347] RBP: 000055e417783900 R08: 000055e418f9e290 R09: 0000000000000000 <4>[ 23.488356] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000001 <4>[ 23.488363] R13: 00007f53b4302c40 R14: 0000000000000000 R15: 0000000000000000 <4>[ 23.488384] Code: b5 f2 f2 e0 0f ff e9 c5 fe ff ff 80 3d 0e 4b 10 00 00 0f 85 c6 fe ff ff 48 c7 c7 30 73 29 a0 c6 05 fa 4a 10 00 01 e8 8e f2 f2 e0 <0f> ff e9 ac fe ff ff e8 51 9d f3 e0 85 c0 0f 85 01 ff ff ff 48 <4>[ 23.488780] ---[ end trace 6bc72ce7f1596190 ]--- <7>[ 23.488844] [drm:intel_atomic_check [i915]] checking fdi config on pipe A, lanes 1 <7>[ 23.488911] [drm:intel_atomic_check [i915]] hw max bpp: 36, pipe bpp: 24, dithering: 0 due to intel_fdi_link_freq() poking at FDI_PLL_BIOS_0. Avoid this by recording the fdi pll frequency during device initiailisation. v2: Also extract the static FDI PLL frequencies for Sandybridge and Ivybridge. Signed-off-by: Chris Wilson Cc: Ville Syrjälä Cc: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171107214713.18704-1-chris@chris-wilson.co.uk Reviewed-by: Maarten Lankhorst Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/intel_display.c | 21 ++++++++++++++++++--- 2 files changed, 19 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index fe93115c4caa9..5e6938d0a9033 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2349,6 +2349,7 @@ struct drm_i915_private { unsigned int max_dotclk_freq; unsigned int rawclk_freq; unsigned int hpll_freq; + unsigned int fdi_pll_freq; unsigned int czclk_freq; struct { diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 737de251d0f85..3dec70fa1ee84 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -219,10 +219,8 @@ intel_fdi_link_freq(struct drm_i915_private *dev_priv, { if (HAS_DDI(dev_priv)) return pipe_config->port_clock; /* SPLL */ - else if (IS_GEN5(dev_priv)) - return ((I915_READ(FDI_PLL_BIOS_0) & FDI_PLL_FB_CLOCK_MASK) + 2) * 10000; else - return 270000; + return dev_priv->fdi_pll_freq; } static const struct intel_limit intel_limits_i8xx_dac = { @@ -14454,6 +14452,22 @@ static void sanitize_watermarks(struct drm_device *dev) drm_modeset_acquire_fini(&ctx); } +static void intel_update_fdi_pll_freq(struct drm_i915_private *dev_priv) +{ + if (IS_GEN5(dev_priv)) { + u32 fdi_pll_clk = + I915_READ(FDI_PLL_BIOS_0) & FDI_PLL_FB_CLOCK_MASK; + + dev_priv->fdi_pll_freq = (fdi_pll_clk + 2) * 10000; + } else if (IS_GEN6(dev_priv) || IS_IVYBRIDGE(dev_priv)) { + dev_priv->fdi_pll_freq = 270000; + } else { + return; + } + + DRM_DEBUG_DRIVER("FDI PLL freq=%d\n", dev_priv->fdi_pll_freq); +} + int intel_modeset_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = to_i915(dev); @@ -14541,6 +14555,7 @@ int intel_modeset_init(struct drm_device *dev) } intel_shared_dpll_init(dev); + intel_update_fdi_pll_freq(dev_priv); intel_update_czclk(dev_priv); intel_modeset_init_hw(dev); -- GitLab From c400cc2a1aea6c6c0d60d22521fa4b7261106964 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 15:22:11 +0000 Subject: [PATCH 0148/1471] drm/i915: Include intel_engine_is_idle() status in engine pretty-printer Upon parking, if we discover an active engine we dump its state. Follow that state with an indication of whether the engine was idle. References: https://bugs.freedesktop.org/show_bug.cgi?id=103479 Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171107152211.19930-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_engine_cs.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index ddbe5c9bf45ab..6997306be0d28 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1835,6 +1835,7 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) } spin_unlock_irq(&b->rb_lock); + drm_printf(m, "Idle? %s\n", yesno(intel_engine_is_idle(engine))); drm_printf(m, "\n"); } -- GitLab From 2f6a3783833dde63f1c08982943a8b2229b97afb Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 8 Nov 2017 09:44:00 +0000 Subject: [PATCH 0149/1471] drm/i915: Idle the GPU before shinking everything The handling of contexts are peculiar. Instead of tieing their vma to activity, we pin the context. This means that we cannot simply unbind the context object itself at will (which would normally cause us to wait for the vma to be idle), but must manually idle the GPU and retire requests first. A consequence of this peculiarity is when doing a last desperate attempt to recover memory. If the memory is tied up inside active context objects, we will fail to recover any memory simply by trying to unbind the objects without first doing a wait-for-idle. A side-effect of removing the call to shrinker_lock_uninterruptible() from i915_gem_shrinker_oom() was that we removed an unlocked wait-for-idle, and so lost the "natural" shrinkage of context objects. By replacing that with a locked wait from inside i915_gem_shrink(), we not only replace it with the ability to recover all context objects, but do so for all i915_gem_shrink_all() callers. v2: Switching requires request allocation, which is not permitted from inside the shrinker as it only uses ordinary allocations. References: https://bugs.freedesktop.org/show_bug.cgi?id=102936 Fixes: f2123818ffad ("drm/i915: Move dev_priv->mm.[un]bound_list to its own lock") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171108094400.1386-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_shrinker.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c index eb31f8aa5c216..3770e3323fc8e 100644 --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c @@ -162,6 +162,18 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, if (!shrinker_lock(dev_priv, &unlock)) return 0; + /* + * When shrinking the active list, also consider active contexts. + * Active contexts are pinned until they are retired, and so can + * not be simply unbound to retire and unpin their pages. To shrink + * the contexts, we must wait until the gpu is idle. + * + * We don't care about errors here; if we cannot wait upon the GPU, + * we will free as much as we can and hope to get a second chance. + */ + if (flags & I915_SHRINK_ACTIVE) + i915_gem_wait_for_idle(dev_priv, I915_WAIT_LOCKED); + trace_i915_gem_shrink(dev_priv, target, flags); i915_gem_retire_requests(dev_priv); -- GitLab From 1ab22356b37ab08a391d6f007fda4c822bef9fb5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 22:06:56 +0000 Subject: [PATCH 0150/1471] drm/i915: Prune the reservation shared fence array The shared fence array is not autopruning and may continue to grow as an object is shared between new timelines. Take the opportunity when we think the object is idle (we have to confirm that any external fence is also signaled) to decouple all the fences. We apply a similar trick after waiting on an object, see commit e54ca9774777 ("drm/i915: Remove completed fences after a wait") v2: No longer need to handle the batch pool as a special case. v3: Need to trylock from within i915_vma_retire as this may be called form the shrinker - and we may later try to allocate underneath the reservation lock, so a deadlock is possible. References: https://bugs.freedesktop.org/show_bug.cgi?id=102936 Fixes: d07f0e59b2c7 ("drm/i915: Move GEM activity tracking into a common struct reservation_object") Fixes: 80b204bce8f2 ("drm/i915: Enable multiple timelines") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171107220656.5020-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_vma.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index 7cdf34800549c..b7ac84db1d2e6 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -54,6 +54,13 @@ i915_vma_retire(struct i915_gem_active *active, if (--obj->active_count) return; + /* Prune the shared fence arrays iff completely idle (inc. external) */ + if (reservation_object_trylock(obj->resv)) { + if (reservation_object_test_signaled_rcu(obj->resv, true)) + reservation_object_add_excl_fence(obj->resv, NULL); + reservation_object_unlock(obj->resv); + } + /* Bump our place on the bound list to keep it roughly in LRU order * so that we don't steal from recently used but inactive objects * (unless we are forced to ofc!) -- GitLab From f72b84c677d61f201b869223a8d6e389c7bb7d3d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 8 Nov 2017 15:35:55 +0200 Subject: [PATCH 0151/1471] drm/i915: Move init_clock_gating() back to where it was MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Apparently setting up a bunch of GT registers before we've properly initialized the rest of the GT hardware leads to these setting being lost. So looks like I broke HSW with commit b7048ea12fbb ("drm/i915: Do .init_clock_gating() earlier to avoid it clobbering watermarks") by doing init_clock_gating() too early. This should actually affect other platforms as well, but apparently not to such a great degree. What I was ultimately after in that commit was to move the ilk_init_lp_watermarks() call earlier. So let's undo the damage and move init_clock_gating() back to where it was, and call ilk_init_lp_watermarks() just before the watermark state readout. This highlights how fragile and messed up our init order really is. I wonder why we even initialize the display before gem. The opposite order would make much more sense to me... v2: Keep WaRsPkgCStateDisplayPMReq:hsw early as it really must be done before all planes might get disabled. Cc: stable@vger.kernel.org Cc: Chris Wilson Cc: Mark Janes Cc: Maarten Lankhorst Cc: Daniel Vetter Cc: Joonas Lahtinen Cc: Oscar Mateo Cc: Mika Kuoppala Reported-by: Mark Janes Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103549 Fixes: b7048ea12fbb ("drm/i915: Do .init_clock_gating() earlier to avoid it clobbering watermarks") References: https://lists.freedesktop.org/archives/intel-gfx/2017-November/145432.html Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171108133555.14091-1-ville.syrjala@linux.intel.com Tested-by: Chris Wilson Reviewed-by: Chris Wilson --- drivers/gpu/drm/i915/intel_display.c | 14 +++++++-- drivers/gpu/drm/i915/intel_pm.c | 44 ++++++++++++---------------- 2 files changed, 30 insertions(+), 28 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 3dec70fa1ee84..84817ccc53053 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3674,6 +3674,7 @@ void intel_finish_reset(struct drm_i915_private *dev_priv) intel_pps_unlock_regs_wa(dev_priv); intel_modeset_init_hw(dev); + intel_init_clock_gating(dev_priv); spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display.hpd_irq_setup) @@ -14363,8 +14364,6 @@ void intel_modeset_init_hw(struct drm_device *dev) intel_update_cdclk(dev_priv); intel_dump_cdclk_state(&dev_priv->cdclk.hw, "Current CDCLK"); dev_priv->cdclk.logical = dev_priv->cdclk.actual = dev_priv->cdclk.hw; - - intel_init_clock_gating(dev_priv); } /* @@ -15094,6 +15093,15 @@ intel_modeset_setup_hw_state(struct drm_device *dev, struct intel_encoder *encoder; int i; + if (IS_HASWELL(dev_priv)) { + /* + * WaRsPkgCStateDisplayPMReq:hsw + * System hang if this isn't done before disabling all planes! + */ + I915_WRITE(CHICKEN_PAR1_1, + I915_READ(CHICKEN_PAR1_1) | FORCE_ARB_IDLE_PLANES); + } + intel_modeset_readout_hw_state(dev); /* HW state is read out, now we need to sanitize this mess. */ @@ -15191,6 +15199,8 @@ void intel_modeset_gem_init(struct drm_device *dev) intel_init_gt_powersave(dev_priv); + intel_init_clock_gating(dev_priv); + intel_setup_overlay(dev_priv); } diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 46440e2ecb33b..e09377df590d8 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5754,12 +5754,30 @@ void vlv_wm_sanitize(struct drm_i915_private *dev_priv) mutex_unlock(&dev_priv->wm.wm_mutex); } +/* + * FIXME should probably kill this and improve + * the real watermark readout/sanitation instead + */ +static void ilk_init_lp_watermarks(struct drm_i915_private *dev_priv) +{ + I915_WRITE(WM3_LP_ILK, I915_READ(WM3_LP_ILK) & ~WM1_LP_SR_EN); + I915_WRITE(WM2_LP_ILK, I915_READ(WM2_LP_ILK) & ~WM1_LP_SR_EN); + I915_WRITE(WM1_LP_ILK, I915_READ(WM1_LP_ILK) & ~WM1_LP_SR_EN); + + /* + * Don't touch WM1S_LP_EN here. + * Doing so could cause underruns. + */ +} + void ilk_wm_get_hw_state(struct drm_device *dev) { struct drm_i915_private *dev_priv = to_i915(dev); struct ilk_wm_values *hw = &dev_priv->wm.hw; struct drm_crtc *crtc; + ilk_init_lp_watermarks(dev_priv); + for_each_crtc(dev, crtc) ilk_pipe_wm_get_hw_state(crtc); @@ -8213,18 +8231,6 @@ static void g4x_disable_trickle_feed(struct drm_i915_private *dev_priv) } } -static void ilk_init_lp_watermarks(struct drm_i915_private *dev_priv) -{ - I915_WRITE(WM3_LP_ILK, I915_READ(WM3_LP_ILK) & ~WM1_LP_SR_EN); - I915_WRITE(WM2_LP_ILK, I915_READ(WM2_LP_ILK) & ~WM1_LP_SR_EN); - I915_WRITE(WM1_LP_ILK, I915_READ(WM1_LP_ILK) & ~WM1_LP_SR_EN); - - /* - * Don't touch WM1S_LP_EN here. - * Doing so could cause underruns. - */ -} - static void ilk_init_clock_gating(struct drm_i915_private *dev_priv) { uint32_t dspclk_gate = ILK_VRHUNIT_CLOCK_GATE_DISABLE; @@ -8258,8 +8264,6 @@ static void ilk_init_clock_gating(struct drm_i915_private *dev_priv) (I915_READ(DISP_ARB_CTL) | DISP_FBC_WM_DIS)); - ilk_init_lp_watermarks(dev_priv); - /* * Based on the document from hardware guys the following bits * should be set unconditionally in order to enable FBC. @@ -8372,8 +8376,6 @@ static void gen6_init_clock_gating(struct drm_i915_private *dev_priv) I915_WRITE(GEN6_GT_MODE, _MASKED_FIELD(GEN6_WIZ_HASHING_MASK, GEN6_WIZ_HASHING_16x4)); - ilk_init_lp_watermarks(dev_priv); - I915_WRITE(CACHE_MODE_0, _MASKED_BIT_DISABLE(CM0_STC_EVICT_DISABLE_LRA_SNB)); @@ -8600,8 +8602,6 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv) I915_GTT_PAGE_SIZE_2M); enum pipe pipe; - ilk_init_lp_watermarks(dev_priv); - /* WaSwitchSolVfFArbitrationPriority:bdw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); @@ -8652,8 +8652,6 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv) static void hsw_init_clock_gating(struct drm_i915_private *dev_priv) { - ilk_init_lp_watermarks(dev_priv); - /* L3 caching of data atomics doesn't work -- disable it. */ I915_WRITE(HSW_SCRATCH1, HSW_SCRATCH1_L3_DATA_ATOMICS_DISABLE); I915_WRITE(HSW_ROW_CHICKEN3, @@ -8697,10 +8695,6 @@ static void hsw_init_clock_gating(struct drm_i915_private *dev_priv) /* WaSwitchSolVfFArbitrationPriority:hsw */ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL); - /* WaRsPkgCStateDisplayPMReq:hsw */ - I915_WRITE(CHICKEN_PAR1_1, - I915_READ(CHICKEN_PAR1_1) | FORCE_ARB_IDLE_PLANES); - lpt_init_clock_gating(dev_priv); } @@ -8708,8 +8702,6 @@ static void ivb_init_clock_gating(struct drm_i915_private *dev_priv) { uint32_t snpcr; - ilk_init_lp_watermarks(dev_priv); - I915_WRITE(ILK_DSPCLK_GATE_D, ILK_VRHUNIT_CLOCK_GATE_DISABLE); /* WaDisableEarlyCull:ivb */ -- GitLab From 0a60797a0efbc495f514304d83eb289bb55990a6 Mon Sep 17 00:00:00 2001 From: Rafael Antognolli Date: Fri, 3 Nov 2017 11:30:27 -0700 Subject: [PATCH 0152/1471] drm/i915: Implement ReadHitWriteOnlyDisable. The workaround for this is described as: "if RenderSurfaceState.Num_Multisamples > 1, disable RCC clock gating if RenderSurfaceState.Num_Multisamples == 1, set 0x7010[14] = 1" Further documentation in the internal bug referenced by the bspec suggest that any of the above suggestions should suffice to fix the issue. We are going with disabling RCC clock gating. Unfortunately, what we are doing doesn't match the name of the workaround, but at least it matches its description. This change improves CNL stability by avoiding some of the hangs seen in the platform. v2: Only disable RCC clock gating. Signed-off-by: Rafael Antognolli Reviewed-by: Rodrigo Vivi Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171103183027.5051-1-rafael.antognolli@intel.com --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_engine_cs.c | 3 +++ 2 files changed, 4 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index f0f8f60596528..6ef33422f7629 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3837,6 +3837,7 @@ enum { */ #define SLICE_UNIT_LEVEL_CLKGATE _MMIO(0x94d4) #define SARBUNIT_CLKGATE_DIS (1 << 5) +#define RCCUNIT_CLKGATE_DIS (1 << 7) /* * Display engine regs diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 6997306be0d28..6cb8e3ed97e4d 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1320,6 +1320,9 @@ static int cnl_init_workarounds(struct intel_engine_cs *engine) WA_SET_FIELD_MASKED(GEN8_CS_CHICKEN1, GEN9_PREEMPT_GPGPU_LEVEL_MASK, GEN9_PREEMPT_GPGPU_COMMAND_LEVEL); + /* ReadHitWriteOnlyDisable: cnl */ + WA_SET_BIT_MASKED(SLICE_UNIT_LEVEL_CLKGATE, RCCUNIT_CLKGATE_DIS); + /* WaEnablePreemptionGranularityControlByUMD:cnl */ I915_WRITE(GEN7_FF_SLICE_CS_CHICKEN1, _MASKED_BIT_ENABLE(GEN9_FFSC_PERCTX_PREEMPT_CTRL)); -- GitLab From 2ceddb189306443a8d0d300de5b2bdf496749548 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 3 Nov 2017 15:18:12 +0000 Subject: [PATCH 0153/1471] drm/i915/guc: Split GuC firmware xfer function into clear steps Transfer of GuC firmware requires few steps that currently are implemented in two large functions. Split this code into smaller functions to make these steps small and clear. Also be prepared for potential DMA xfer step failure. v2: rename function steps (Sagar) Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-1-michal.wajdeczko@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_fw.c | 172 +++++++++++++++++----------- 1 file changed, 103 insertions(+), 69 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index ef67a36354c56..c4f4526e3b770 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -97,23 +97,55 @@ int intel_guc_fw_select(struct intel_guc *guc) return 0; } -/* - * Read the GuC status register (GUC_STATUS) and store it in the - * specified location; then return a boolean indicating whether - * the value matches either of two values representing completion - * of the GuC boot process. - * - * This is used for polling the GuC status in a wait_for() - * loop below. - */ -static inline bool guc_ucode_response(struct drm_i915_private *dev_priv, - u32 *status) +static void guc_prepare_xfer(struct intel_guc *guc) { - u32 val = I915_READ(GUC_STATUS); - u32 uk_val = val & GS_UKERNEL_MASK; - *status = val; - return (uk_val == GS_UKERNEL_READY || - ((val & GS_MIA_CORE_STATE) && uk_val == GS_UKERNEL_LAPIC_DONE)); + struct drm_i915_private *dev_priv = guc_to_i915(guc); + + /* Enable MIA caching. GuC clock gating is disabled. */ + I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE); + + /* WaDisableMinuteIaClockGating:bxt */ + if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { + I915_WRITE(GUC_SHIM_CONTROL, (I915_READ(GUC_SHIM_CONTROL) & + ~GUC_ENABLE_MIA_CLOCK_GATING)); + } + + /* WaC6DisallowByGfxPause:bxt */ + if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) + I915_WRITE(GEN6_GFXPAUSE, 0x30FFF); + + if (IS_GEN9_LP(dev_priv)) + I915_WRITE(GEN9LP_GT_PM_CONFIG, GT_DOORBELL_ENABLE); + else + I915_WRITE(GEN9_GT_PM_CONFIG, GT_DOORBELL_ENABLE); + + if (IS_GEN9(dev_priv)) { + /* DOP Clock Gating Enable for GuC clocks */ + I915_WRITE(GEN7_MISCCPCTL, (GEN8_DOP_CLOCK_GATE_GUC_ENABLE | + I915_READ(GEN7_MISCCPCTL))); + + /* allows for 5us (in 10ns units) before GT can go to RC6 */ + I915_WRITE(GUC_ARAT_C6DIS, 0x1FF); + } +} + +/* Copy RSA signature from the fw image to HW for verification */ +static int guc_xfer_rsa(struct intel_guc *guc, struct i915_vma *vma) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + struct intel_uc_fw *guc_fw = &guc->fw; + struct sg_table *sg = vma->pages; + u32 rsa[UOS_RSA_SCRATCH_MAX_COUNT]; + int i; + + if (sg_pcopy_to_buffer(sg->sgl, sg->nents, rsa, sizeof(rsa), + guc_fw->rsa_offset) != sizeof(rsa)) + return -EINVAL; + + for (i = 0; i < UOS_RSA_SCRATCH_MAX_COUNT; i++) + I915_WRITE(UOS_RSA_SCRATCH(i), rsa[i]); + + return 0; } /* @@ -122,29 +154,17 @@ static inline bool guc_ucode_response(struct drm_i915_private *dev_priv, * Architecturally, the DMA engine is bidirectional, and can potentially even * transfer between GTT locations. This functionality is left out of the API * for now as there is no need for it. - * - * Note that GuC needs the CSS header plus uKernel code to be copied by the - * DMA engine in one operation, whereas the RSA signature is loaded via MMIO. */ -static int guc_ucode_xfer_dma(struct drm_i915_private *dev_priv, - struct i915_vma *vma) +static int guc_xfer_ucode(struct intel_guc *guc, struct i915_vma *vma) { - struct intel_uc_fw *guc_fw = &dev_priv->guc.fw; + struct drm_i915_private *dev_priv = guc_to_i915(guc); + struct intel_uc_fw *guc_fw = &guc->fw; unsigned long offset; - struct sg_table *sg = vma->pages; - u32 status, rsa[UOS_RSA_SCRATCH_MAX_COUNT]; - int i, ret = 0; - - /* where RSA signature starts */ - offset = guc_fw->rsa_offset; - /* Copy RSA signature from the fw image to HW for verification */ - sg_pcopy_to_buffer(sg->sgl, sg->nents, rsa, sizeof(rsa), offset); - for (i = 0; i < UOS_RSA_SCRATCH_MAX_COUNT; i++) - I915_WRITE(UOS_RSA_SCRATCH(i), rsa[i]); - - /* The header plus uCode will be copied to WOPCM via DMA, excluding any - * other components */ + /* + * The header plus uCode will be copied to WOPCM via DMA, excluding any + * other components + */ I915_WRITE(DMA_COPY_SIZE, guc_fw->header_size + guc_fw->ucode_size); /* Set the source address for the new blob */ @@ -162,33 +182,57 @@ static int guc_ucode_xfer_dma(struct drm_i915_private *dev_priv, /* Finally start the DMA */ I915_WRITE(DMA_CTRL, _MASKED_BIT_ENABLE(UOS_MOVE | START_DMA)); + return 0; +} + +/* + * Read the GuC status register (GUC_STATUS) and store it in the + * specified location; then return a boolean indicating whether + * the value matches either of two values representing completion + * of the GuC boot process. + * + * This is used for polling the GuC status in a wait_for() + * loop below. + */ +static inline bool guc_ready(struct intel_guc *guc, u32 *status) +{ + struct drm_i915_private *dev_priv = guc_to_i915(guc); + u32 val = I915_READ(GUC_STATUS); + u32 uk_val = val & GS_UKERNEL_MASK; + + *status = val; + return (uk_val == GS_UKERNEL_READY) || + ((val & GS_MIA_CORE_STATE) && (uk_val == GS_UKERNEL_LAPIC_DONE)); +} + +static int guc_wait_ucode(struct intel_guc *guc) +{ + u32 status; + int ret; + /* - * Wait for the DMA to complete & the GuC to start up. + * Wait for the GuC to start up. * NB: Docs recommend not using the interrupt for completion. * Measurements indicate this should take no more than 20ms, so a * timeout here indicates that the GuC has failed and is unusable. * (Higher levels of the driver will attempt to fall back to * execlist mode if this happens.) */ - ret = wait_for(guc_ucode_response(dev_priv, &status), 100); - - DRM_DEBUG_DRIVER("DMA status 0x%x, GuC status 0x%x\n", - I915_READ(DMA_CTRL), status); + ret = wait_for(guc_ready(guc, &status), 100); + DRM_DEBUG_DRIVER("GuC status %#x\n", status); if ((status & GS_BOOTROM_MASK) == GS_BOOTROM_RSA_FAILED) { DRM_ERROR("GuC firmware signature verification failed\n"); ret = -ENOEXEC; } - DRM_DEBUG_DRIVER("returning %d\n", ret); - return ret; } /* * Load the GuC firmware blob into the MinuteIA. */ -static int guc_ucode_xfer(struct intel_uc_fw *guc_fw, struct i915_vma *vma) +static int guc_fw_xfer(struct intel_uc_fw *guc_fw, struct i915_vma *vma) { struct intel_guc *guc = container_of(guc_fw, struct intel_guc, fw); struct drm_i915_private *dev_priv = guc_to_i915(guc); @@ -198,34 +242,24 @@ static int guc_ucode_xfer(struct intel_uc_fw *guc_fw, struct i915_vma *vma) intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL); - /* Enable MIA caching. GuC clock gating is disabled. */ - I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE); - - /* WaDisableMinuteIaClockGating:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { - I915_WRITE(GUC_SHIM_CONTROL, (I915_READ(GUC_SHIM_CONTROL) & - ~GUC_ENABLE_MIA_CLOCK_GATING)); - } - - /* WaC6DisallowByGfxPause:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) - I915_WRITE(GEN6_GFXPAUSE, 0x30FFF); + guc_prepare_xfer(guc); - if (IS_GEN9_LP(dev_priv)) - I915_WRITE(GEN9LP_GT_PM_CONFIG, GT_DOORBELL_ENABLE); - else - I915_WRITE(GEN9_GT_PM_CONFIG, GT_DOORBELL_ENABLE); - - if (IS_GEN9(dev_priv)) { - /* DOP Clock Gating Enable for GuC clocks */ - I915_WRITE(GEN7_MISCCPCTL, (GEN8_DOP_CLOCK_GATE_GUC_ENABLE | - I915_READ(GEN7_MISCCPCTL))); + /* + * Note that GuC needs the CSS header plus uKernel code to be copied + * by the DMA engine in one operation, whereas the RSA signature is + * loaded via MMIO. + */ + ret = guc_xfer_rsa(guc, vma); + if (ret) + DRM_WARN("GuC firmware signature xfer error %d\n", ret); - /* allows for 5us (in 10ns units) before GT can go to RC6 */ - I915_WRITE(GUC_ARAT_C6DIS, 0x1FF); - } + ret = guc_xfer_ucode(guc, vma); + if (ret) + DRM_WARN("GuC firmware code xfer error %d\n", ret); - ret = guc_ucode_xfer_dma(dev_priv, vma); + ret = guc_wait_ucode(guc); + if (ret) + DRM_ERROR("GuC firmware xfer error %d\n", ret); intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); @@ -247,5 +281,5 @@ static int guc_ucode_xfer(struct intel_uc_fw *guc_fw, struct i915_vma *vma) */ int intel_guc_fw_upload(struct intel_guc *guc) { - return intel_uc_fw_upload(&guc->fw, guc_ucode_xfer); + return intel_uc_fw_upload(&guc->fw, guc_fw_xfer); } -- GitLab From a86af2314a867804bb1531d514aa25c264fe5cb9 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 3 Nov 2017 15:18:13 +0000 Subject: [PATCH 0154/1471] drm/i915/guc: Wait for ucode DMA transfer completion We silently assumed that DMA transfer will be completed within assumed timeout and thus we were waiting only at last step for GuC to become ready. Add intermediate wait to catch unexpected delays in DMA transfer. Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-2-michal.wajdeczko@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_fw.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index c4f4526e3b770..74a61fe9448cc 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -160,6 +160,8 @@ static int guc_xfer_ucode(struct intel_guc *guc, struct i915_vma *vma) struct drm_i915_private *dev_priv = guc_to_i915(guc); struct intel_uc_fw *guc_fw = &guc->fw; unsigned long offset; + u32 status; + int ret; /* * The header plus uCode will be copied to WOPCM via DMA, excluding any @@ -182,7 +184,12 @@ static int guc_xfer_ucode(struct intel_guc *guc, struct i915_vma *vma) /* Finally start the DMA */ I915_WRITE(DMA_CTRL, _MASKED_BIT_ENABLE(UOS_MOVE | START_DMA)); - return 0; + /* Wait for DMA to finish */ + ret = __intel_wait_for_register_fw(dev_priv, DMA_CTRL, START_DMA, 0, + 2, 100, &status); + DRM_DEBUG_DRIVER("GuC DMA status %#x\n", status); + + return ret; } /* -- GitLab From e6843363768cc639a590917b509303f8e22dd6de Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 3 Nov 2017 15:18:14 +0000 Subject: [PATCH 0155/1471] drm/i915/guc: Drop legacy workarounds from guc_prepare_xfer We don't keep the workarounds for pre-production hardware (see intel_detect_preproduction_hw) thus we can drop some extra steps during firmware upload needed only for unsupported platforms. Suggested-by: Chris Wilson Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-3-michal.wajdeczko@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_fw.c | 10 ---------- 1 file changed, 10 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index 74a61fe9448cc..a63b5cf7421a2 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -104,16 +104,6 @@ static void guc_prepare_xfer(struct intel_guc *guc) /* Enable MIA caching. GuC clock gating is disabled. */ I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE); - /* WaDisableMinuteIaClockGating:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { - I915_WRITE(GUC_SHIM_CONTROL, (I915_READ(GUC_SHIM_CONTROL) & - ~GUC_ENABLE_MIA_CLOCK_GATING)); - } - - /* WaC6DisallowByGfxPause:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) - I915_WRITE(GEN6_GFXPAUSE, 0x30FFF); - if (IS_GEN9_LP(dev_priv)) I915_WRITE(GEN9LP_GT_PM_CONFIG, GT_DOORBELL_ENABLE); else -- GitLab From b814b0744a1434aa46c72bd02c3e9c8b1263f120 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 3 Nov 2017 15:18:15 +0000 Subject: [PATCH 0156/1471] drm/i915/guc: Simplify programming of GUC_SHIM_CONTROL We can program GUC_SHIM_CONTROL register with all expected bits without use of extra macro defined in fwif.h v2: rebased without pre-prod code v3: fixed typo Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171103151816.62048-4-michal.wajdeczko@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_guc_reg.h | 7 ------- drivers/gpu/drm/i915/intel_guc_fw.c | 9 +++++++-- 2 files changed, 7 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_reg.h b/drivers/gpu/drm/i915/i915_guc_reg.h index 35cf9918d09a4..bc1ae7d8f424d 100644 --- a/drivers/gpu/drm/i915/i915_guc_reg.h +++ b/drivers/gpu/drm/i915/i915_guc_reg.h @@ -102,13 +102,6 @@ #define GUC_ENABLE_MIA_CLOCK_GATING (1<<15) #define GUC_GEN10_SHIM_WC_ENABLE (1<<21) -#define GUC_SHIM_CONTROL_VALUE (GUC_DISABLE_SRAM_INIT_TO_ZEROES | \ - GUC_ENABLE_READ_CACHE_LOGIC | \ - GUC_ENABLE_MIA_CACHING | \ - GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA | \ - GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA | \ - GUC_ENABLE_MIA_CLOCK_GATING) - #define GUC_SEND_INTERRUPT _MMIO(0xc4c8) #define GUC_SEND_TRIGGER (1<<0) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index a63b5cf7421a2..69ba015995757 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -101,8 +101,13 @@ static void guc_prepare_xfer(struct intel_guc *guc) { struct drm_i915_private *dev_priv = guc_to_i915(guc); - /* Enable MIA caching. GuC clock gating is disabled. */ - I915_WRITE(GUC_SHIM_CONTROL, GUC_SHIM_CONTROL_VALUE); + /* Must program this register before loading the ucode with DMA */ + I915_WRITE(GUC_SHIM_CONTROL, GUC_DISABLE_SRAM_INIT_TO_ZEROES | + GUC_ENABLE_READ_CACHE_LOGIC | + GUC_ENABLE_MIA_CACHING | + GUC_ENABLE_READ_CACHE_FOR_SRAM_DATA | + GUC_ENABLE_READ_CACHE_FOR_WOPCM_DATA | + GUC_ENABLE_MIA_CLOCK_GATING); if (IS_GEN9_LP(dev_priv)) I915_WRITE(GEN9LP_GT_PM_CONFIG, GT_DOORBELL_ENABLE); -- GitLab From 2e2b96ef7a9b71e1bf08439e363c18cf700bafec Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 8 Nov 2017 21:30:07 +0100 Subject: [PATCH 0157/1471] drm: Update docs for legacy kms state Point at the equivalent atomic state and explain that atomic drivers shouldn't really depend upon legacy state. Motivated by questions from Manasi about how this all is supposed to work. Cc: Manasi Navare Reviewed-by: Manasi Navare Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171108203007.12274-1-daniel.vetter@ffwll.ch --- drivers/gpu/drm/drm_atomic_helper.c | 6 ++++++ include/drm/drm_connector.h | 9 +++++++-- include/drm/drm_encoder.h | 6 +++++- include/drm/drm_plane.h | 12 ++++++++++-- 4 files changed, 28 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 27dc6838533ab..15cc189ed1a33 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -907,6 +907,12 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state) * * Drivers can use this for building their own atomic commit if they don't have * a pure helper-based modeset implementation. + * + * Since these updates are not synchronized with lockings, only code paths + * called from &drm_mode_config_helper_funcs.atomic_commit_tail can look at the + * legacy state filled out by this helper. Defacto this means this helper and + * the legacy state pointers are only really useful for transitioning an + * existing driver to the atomic world. */ void drm_atomic_helper_update_legacy_modeset_state(struct drm_device *dev, diff --git a/include/drm/drm_connector.h b/include/drm/drm_connector.h index b34904dc8b9be..b81b7fb846f9b 100644 --- a/include/drm/drm_connector.h +++ b/include/drm/drm_connector.h @@ -699,7 +699,6 @@ struct drm_cmdline_mode { * @force: a DRM_FORCE_ state for forced mode sets * @override_edid: has the EDID been overwritten through debugfs for testing? * @encoder_ids: valid encoders for this connector - * @encoder: encoder driving this connector, if any * @eld: EDID-like data, if present * @latency_present: AV delay info from ELD, if found * @video_latency: video latency info from ELD, if found @@ -869,7 +868,13 @@ struct drm_connector { #define DRM_CONNECTOR_MAX_ENCODER 3 uint32_t encoder_ids[DRM_CONNECTOR_MAX_ENCODER]; - struct drm_encoder *encoder; /* currently active encoder */ + /** + * @encoder: Currently bound encoder driving this connector, if any. + * Only really meaningful for non-atomic drivers. Atomic drivers should + * instead look at &drm_connector_state.best_encoder, and in case they + * need the CRTC driving this output, &drm_connector_state.crtc. + */ + struct drm_encoder *encoder; #define MAX_ELD_BYTES 128 /* EDID bits */ diff --git a/include/drm/drm_encoder.h b/include/drm/drm_encoder.h index 8d8245ec0181d..c3ca2278ea6e3 100644 --- a/include/drm/drm_encoder.h +++ b/include/drm/drm_encoder.h @@ -88,7 +88,6 @@ struct drm_encoder_funcs { * @head: list management * @base: base KMS object * @name: human readable name, can be overwritten by the driver - * @crtc: currently bound CRTC * @bridge: bridge associated to the encoder * @funcs: control functions * @helper_private: mid-layer private data @@ -166,6 +165,11 @@ struct drm_encoder { */ uint32_t possible_clones; + /** + * @crtc: Currently bound CRTC, only really meaningful for non-atomic + * drivers. Atomic drivers should instead check + * &drm_connector_state.crtc. + */ struct drm_crtc *crtc; struct drm_bridge *bridge; const struct drm_encoder_funcs *funcs; diff --git a/include/drm/drm_plane.h b/include/drm/drm_plane.h index 82a217bd77f05..68bf66473faa6 100644 --- a/include/drm/drm_plane.h +++ b/include/drm/drm_plane.h @@ -474,8 +474,6 @@ enum drm_plane_type { * @format_types: array of formats supported by this plane * @format_count: number of formats supported * @format_default: driver hasn't supplied supported formats for the plane - * @crtc: currently bound CRTC - * @fb: currently bound fb * @old_fb: Temporary tracking of the old fb while a modeset is ongoing. Used by * drm_mode_set_config_internal() to implement correct refcounting. * @funcs: helper functions @@ -512,7 +510,17 @@ struct drm_plane { uint64_t *modifiers; unsigned int modifier_count; + /** + * @crtc: Currently bound CRTC, only really meaningful for non-atomic + * drivers. Atomic drivers should instead check &drm_plane_state.crtc. + */ struct drm_crtc *crtc; + + /** + * @fb: Currently bound framebuffer, only really meaningful for + * non-atomic drivers. Atomic drivers should instead check + * &drm_plane_state.fb. + */ struct drm_framebuffer *fb; struct drm_framebuffer *old_fb; -- GitLab From 94dec87159af6f3dcc0b78d3f909aefa9e29c01a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Nov 2017 08:55:40 +0000 Subject: [PATCH 0158/1471] drm/i915: Reorder context-close to avoid calling i915_vma_close() under RCU When we close the VMA, we unbind it from the ppgtt and tear down the page directory pointing at it. That may trigger us to return WC pages back to the system, requiring conversion back to WB which itself may sleep. That makes i915_vma_close() unsuitable for use inside the RCU read lock, which we need to hold to iterate the radixtree. The fix is quite simple, we can close all the VMA as we close the ppgtt, we only need to do that instead of closing them during destruction of the LUT. v2: Order between closing the LUT and the ppgtt is important; we use the vma inside the LUT as a means of retrieving the object, and so we must clear the LUT before freeing the VMA when closing the ppgtt. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103638 Fixes: 547da76b5777 ("drm/i915: Hold rcu_read_lock when iterating over the radixtree (vma idr)") Fixes: d1b48c1e7184 ("drm/i915: Replace execbuf vma ht with an idr") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171109085540.32264-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_gem_context.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 10affb35ac566..c05c3d7d21a5e 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -107,14 +107,9 @@ static void lut_close(struct i915_gem_context *ctx) rcu_read_lock(); radix_tree_for_each_slot(slot, &ctx->handles_vma, &iter, 0) { struct i915_vma *vma = rcu_dereference_raw(*slot); - struct drm_i915_gem_object *obj = vma->obj; radix_tree_iter_delete(&ctx->handles_vma, &iter, slot); - - if (!i915_vma_is_ggtt(vma)) - i915_vma_close(vma); - - __i915_gem_object_release_unless_active(obj); + __i915_gem_object_release_unless_active(vma->obj); } rcu_read_unlock(); } @@ -200,6 +195,11 @@ static void context_close(struct i915_gem_context *ctx) { i915_gem_context_set_closed(ctx); + /* + * The LUT uses the VMA as a backpointer to unref the object, + * so we need to clear the LUT before we close all the VMA (inside + * the ppgtt). + */ lut_close(ctx); if (ctx->ppgtt) i915_ppgtt_close(&ctx->ppgtt->base); -- GitLab From adc103047e583497f675f0db32b758c8c2069633 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:14 +0200 Subject: [PATCH 0159/1471] drm/i915: Eliminate some encoder->crtc usage from DP code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Extract the current crtc from the crtc state rather than via the legacy encoder->crtc pointer whenever possible. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-2-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_dp.c | 63 +++++++++++++++------------------ 1 file changed, 29 insertions(+), 34 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index d27c0145ac91d..4f64d83537d98 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -129,10 +129,12 @@ static struct intel_dp *intel_attached_dp(struct drm_connector *connector) return enc_to_intel_dp(&intel_attached_encoder(connector)->base); } -static void intel_dp_link_down(struct intel_dp *intel_dp); +static void intel_dp_link_down(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state); static bool edp_panel_vdd_on(struct intel_dp *intel_dp); static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync); -static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp); +static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state); static void vlv_steal_power_sequencer(struct drm_device *dev, enum pipe pipe); static void intel_dp_unset_edid(struct intel_dp *intel_dp); @@ -1858,7 +1860,7 @@ static void intel_dp_prepare(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); enum port port = dp_to_dig_port(intel_dp)->port; - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); const struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; intel_dp_set_link_params(intel_dp, pipe_config->port_clock, @@ -2491,10 +2493,10 @@ static void ironlake_edp_pll_on(struct intel_dp *intel_dp, udelay(200); } -static void ironlake_edp_pll_off(struct intel_dp *intel_dp) +static void ironlake_edp_pll_off(struct intel_dp *intel_dp, + const struct intel_crtc_state *old_crtc_state) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_crtc *crtc = to_intel_crtc(intel_dig_port->base.base.crtc); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); assert_pipe_disabled(dev_priv, crtc->pipe); @@ -2624,7 +2626,7 @@ static void intel_dp_get_config(struct intel_encoder *encoder, struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); enum port port = dp_to_dig_port(intel_dp)->port; - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); if (encoder->type == INTEL_OUTPUT_EDP) pipe_config->output_types |= BIT(INTEL_OUTPUT_EDP); @@ -2723,12 +2725,10 @@ static void g4x_disable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) { - struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - intel_disable_dp(encoder, old_crtc_state, old_conn_state); /* disable the port before the pipe on g4x */ - intel_dp_link_down(intel_dp); + intel_dp_link_down(encoder, old_crtc_state); } static void ilk_disable_dp(struct intel_encoder *encoder, @@ -2754,33 +2754,29 @@ static void ilk_post_disable_dp(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; - intel_dp_link_down(intel_dp); + intel_dp_link_down(encoder, old_crtc_state); /* Only ilk+ has port A */ if (port == PORT_A) - ironlake_edp_pll_off(intel_dp); + ironlake_edp_pll_off(intel_dp, old_crtc_state); } static void vlv_post_disable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) { - struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - - intel_dp_link_down(intel_dp); + intel_dp_link_down(encoder, old_crtc_state); } static void chv_post_disable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) { - struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - intel_dp_link_down(intel_dp); + intel_dp_link_down(encoder, old_crtc_state); mutex_lock(&dev_priv->sb_lock); @@ -2909,7 +2905,7 @@ static void intel_enable_dp(struct intel_encoder *encoder, struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); uint32_t dp_reg = I915_READ(intel_dp->output_reg); enum pipe pipe = crtc->pipe; @@ -2919,7 +2915,7 @@ static void intel_enable_dp(struct intel_encoder *encoder, pps_lock(intel_dp); if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) - vlv_init_panel_power_sequencer(intel_dp); + vlv_init_panel_power_sequencer(encoder, pipe_config); intel_dp_enable_port(intel_dp, pipe_config); @@ -3047,13 +3043,13 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, } } -static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) +static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_encoder *encoder = &intel_dig_port->base; + struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); lockdep_assert_held(&dev_priv->pps_mutex); @@ -3084,7 +3080,7 @@ static void vlv_init_panel_power_sequencer(struct intel_dp *intel_dp) intel_dp->pps_pipe = crtc->pipe; DRM_DEBUG_KMS("initializing pipe %c power sequencer for port %c\n", - pipe_name(intel_dp->pps_pipe), port_name(intel_dig_port->port)); + pipe_name(intel_dp->pps_pipe), port_name(encoder->port)); /* init power sequencer on this pipe and port */ intel_dp_init_panel_power_sequencer(dev, intel_dp); @@ -3624,13 +3620,13 @@ void intel_dp_set_idle_link_train(struct intel_dp *intel_dp) } static void -intel_dp_link_down(struct intel_dp *intel_dp) +intel_dp_link_down(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_crtc *crtc = to_intel_crtc(intel_dig_port->base.base.crtc); - enum port port = intel_dig_port->port; - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); + enum port port = encoder->port; uint32_t DP = intel_dp->DP; if (WARN_ON(HAS_DDI(dev_priv))) @@ -5493,7 +5489,6 @@ static void intel_dp_set_drrs_state(struct drm_i915_private *dev_priv, dig_port = dp_to_dig_port(intel_dp); encoder = &dig_port->base; - intel_crtc = to_intel_crtc(encoder->base.crtc); if (!intel_crtc) { DRM_DEBUG_KMS("DRRS: intel_crtc not initialized\n"); -- GitLab From 3c0628f8ce220f35a545de40f0426b9b999ce02d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:15 +0200 Subject: [PATCH 0160/1471] drm/i915: Eliminate some encoder->crtc usage from DSI code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Extract the current crtc from the crtc state rather than via the legacy encoder->crtc pointer whenever possible. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-3-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_dsi.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dsi.c b/drivers/gpu/drm/i915/intel_dsi.c index 2bff7ab25bf39..f09474b0c4d35 100644 --- a/drivers/gpu/drm/i915/intel_dsi.c +++ b/drivers/gpu/drm/i915/intel_dsi.c @@ -662,11 +662,11 @@ static void vlv_dsi_clear_device_ready(struct intel_encoder *encoder) } } -static void intel_dsi_port_enable(struct intel_encoder *encoder) +static void intel_dsi_port_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder->base); enum port port; @@ -705,7 +705,7 @@ static void intel_dsi_port_enable(struct intel_encoder *encoder) if (IS_BROXTON(dev_priv)) temp |= LANE_CONFIGURATION_DUAL_LINK_A; else - temp |= intel_crtc->pipe ? + temp |= crtc->pipe ? LANE_CONFIGURATION_DUAL_LINK_B : LANE_CONFIGURATION_DUAL_LINK_A; } @@ -875,7 +875,7 @@ static void intel_dsi_pre_enable(struct intel_encoder *encoder, intel_dsi_vbt_exec_sequence(intel_dsi, MIPI_SEQ_DISPLAY_ON); - intel_dsi_port_enable(encoder); + intel_dsi_port_enable(encoder, pipe_config); } intel_panel_enable_backlight(pipe_config, conn_state); @@ -1082,7 +1082,7 @@ static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder, struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; struct drm_display_mode *adjusted_mode_sw; - struct intel_crtc *intel_crtc; + struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); struct intel_dsi *intel_dsi = enc_to_intel_dsi(&encoder->base); unsigned int lane_count = intel_dsi->lane_count; unsigned int bpp, fmt; @@ -1093,8 +1093,7 @@ static void bxt_dsi_get_pipe_config(struct intel_encoder *encoder, crtc_hblank_start_sw, crtc_hblank_end_sw; /* FIXME: hw readout should not depend on SW state */ - intel_crtc = to_intel_crtc(encoder->base.crtc); - adjusted_mode_sw = &intel_crtc->config->base.adjusted_mode; + adjusted_mode_sw = &crtc->config->base.adjusted_mode; /* * Atleast one port is active as encoder->get_config called only if -- GitLab From 463320ae538bb02fd16ca9ff9ed39a0b35d920e2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:16 +0200 Subject: [PATCH 0161/1471] drm/i915: Eliminate some encoder->crtc usage from SDVO code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Extract the current crtc from the crtc state rather than via the legacy encoder->crtc pointer whenever possible. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-4-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_sdvo.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c index 42ec2d1f7a613..2b8764897d686 100644 --- a/drivers/gpu/drm/i915/intel_sdvo.c +++ b/drivers/gpu/drm/i915/intel_sdvo.c @@ -1512,7 +1512,7 @@ static void intel_disable_sdvo(struct intel_encoder *encoder, { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_sdvo *intel_sdvo = to_sdvo(encoder); - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); u32 temp; intel_sdvo_set_active_outputs(intel_sdvo, 0); @@ -1571,7 +1571,7 @@ static void intel_enable_sdvo(struct intel_encoder *encoder, struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_sdvo *intel_sdvo = to_sdvo(encoder); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(pipe_config->base.crtc); u32 temp; bool input1, input2; int i; -- GitLab From a7f519ba706b3be9280c063e7e231da138c202de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:17 +0200 Subject: [PATCH 0162/1471] drm/i915: Eliminate some encoder->crtc usage from TV code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Extract the current crtc from the crtc state rather than via the legacy encoder->crtc pointer whenever possible. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-5-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_tv.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c index b18609cebe035..b3dabc219e6a1 100644 --- a/drivers/gpu/drm/i915/intel_tv.c +++ b/drivers/gpu/drm/i915/intel_tv.c @@ -822,7 +822,7 @@ intel_enable_tv(struct intel_encoder *encoder, /* Prevents vblank waits from timing out in intel_tv_detect_type() */ intel_wait_for_vblank(dev_priv, - to_intel_crtc(encoder->base.crtc)->pipe); + to_intel_crtc(pipe_config->base.crtc)->pipe); I915_WRITE(TV_CTL, I915_READ(TV_CTL) | TV_ENC_ENABLE); } @@ -982,7 +982,7 @@ static void intel_tv_pre_enable(struct intel_encoder *encoder, const struct drm_connector_state *conn_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(pipe_config->base.crtc); struct intel_tv *intel_tv = enc_to_tv(encoder); const struct tv_mode *tv_mode = intel_tv_mode_find(conn_state); u32 tv_ctl; -- GitLab From 2e1029c672c252ad5f34c0bcd0ebacf4a6412cd0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:18 +0200 Subject: [PATCH 0163/1471] drm/i915: Pass crtc state to DPIO PHY functions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Rather than digging through encoder->crtc and crtc->config in the DPIO PHY functions, pass down the correct crtc state from the caller. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-6-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/i915_drv.h | 19 ++++-- drivers/gpu/drm/i915/intel_dp.c | 16 ++--- drivers/gpu/drm/i915/intel_dpio_phy.c | 87 +++++++++++++-------------- drivers/gpu/drm/i915/intel_hdmi.c | 17 +++--- 4 files changed, 72 insertions(+), 67 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5e6938d0a9033..19b679ca8211e 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -4200,18 +4200,25 @@ void chv_set_phy_signal_level(struct intel_encoder *encoder, u32 deemph_reg_value, u32 margin_reg_value, bool uniq_trans_scale); void chv_data_lane_soft_reset(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, bool reset); -void chv_phy_pre_pll_enable(struct intel_encoder *encoder); -void chv_phy_pre_encoder_enable(struct intel_encoder *encoder); +void chv_phy_pre_pll_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state); +void chv_phy_pre_encoder_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state); void chv_phy_release_cl2_override(struct intel_encoder *encoder); -void chv_phy_post_pll_disable(struct intel_encoder *encoder); +void chv_phy_post_pll_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state); void vlv_set_phy_signal_level(struct intel_encoder *encoder, u32 demph_reg_value, u32 preemph_reg_value, u32 uniqtranscale_reg_value, u32 tx3_demph); -void vlv_phy_pre_pll_enable(struct intel_encoder *encoder); -void vlv_phy_pre_encoder_enable(struct intel_encoder *encoder); -void vlv_phy_reset_lanes(struct intel_encoder *encoder); +void vlv_phy_pre_pll_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state); +void vlv_phy_pre_encoder_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state); +void vlv_phy_reset_lanes(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state); int intel_gpu_freq(struct drm_i915_private *dev_priv, int val); int intel_freq_opcode(struct drm_i915_private *dev_priv, int val); diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 4f64d83537d98..0151c8d18b79e 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -2781,7 +2781,7 @@ static void chv_post_disable_dp(struct intel_encoder *encoder, mutex_lock(&dev_priv->sb_lock); /* Assert data lane reset */ - chv_data_lane_soft_reset(encoder, true); + chv_data_lane_soft_reset(encoder, old_crtc_state, true); mutex_unlock(&dev_priv->sb_lock); } @@ -3091,7 +3091,7 @@ static void vlv_pre_enable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *pipe_config, const struct drm_connector_state *conn_state) { - vlv_phy_pre_encoder_enable(encoder); + vlv_phy_pre_encoder_enable(encoder, pipe_config); intel_enable_dp(encoder, pipe_config, conn_state); } @@ -3102,14 +3102,14 @@ static void vlv_dp_pre_pll_enable(struct intel_encoder *encoder, { intel_dp_prepare(encoder, pipe_config); - vlv_phy_pre_pll_enable(encoder); + vlv_phy_pre_pll_enable(encoder, pipe_config); } static void chv_pre_enable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *pipe_config, const struct drm_connector_state *conn_state) { - chv_phy_pre_encoder_enable(encoder); + chv_phy_pre_encoder_enable(encoder, pipe_config); intel_enable_dp(encoder, pipe_config, conn_state); @@ -3123,14 +3123,14 @@ static void chv_dp_pre_pll_enable(struct intel_encoder *encoder, { intel_dp_prepare(encoder, pipe_config); - chv_phy_pre_pll_enable(encoder); + chv_phy_pre_pll_enable(encoder, pipe_config); } static void chv_dp_post_pll_disable(struct intel_encoder *encoder, - const struct intel_crtc_state *pipe_config, - const struct drm_connector_state *conn_state) + const struct intel_crtc_state *old_crtc_state, + const struct drm_connector_state *old_conn_state) { - chv_phy_post_pll_disable(encoder); + chv_phy_post_pll_disable(encoder, old_crtc_state); } /* diff --git a/drivers/gpu/drm/i915/intel_dpio_phy.c b/drivers/gpu/drm/i915/intel_dpio_phy.c index 490a42d9a3e8b..5958d3d8b90e4 100644 --- a/drivers/gpu/drm/i915/intel_dpio_phy.c +++ b/drivers/gpu/drm/i915/intel_dpio_phy.c @@ -733,11 +733,12 @@ void chv_set_phy_signal_level(struct intel_encoder *encoder, } void chv_data_lane_soft_reset(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state, bool reset) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); enum dpio_channel ch = vlv_dport_to_channel(enc_to_dig_port(&encoder->base)); - struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); enum pipe pipe = crtc->pipe; uint32_t val; @@ -776,17 +777,16 @@ void chv_data_lane_soft_reset(struct intel_encoder *encoder, } } -void chv_phy_pre_pll_enable(struct intel_encoder *encoder) +void chv_phy_pre_pll_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = - to_intel_crtc(encoder->base.crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); enum dpio_channel ch = vlv_dport_to_channel(dport); - enum pipe pipe = intel_crtc->pipe; + enum pipe pipe = crtc->pipe; unsigned int lane_mask = - intel_dp_unused_lane_mask(intel_crtc->config->lane_count); + intel_dp_unused_lane_mask(crtc_state->lane_count); u32 val; /* @@ -802,7 +802,7 @@ void chv_phy_pre_pll_enable(struct intel_encoder *encoder) mutex_lock(&dev_priv->sb_lock); /* Assert data lane reset */ - chv_data_lane_soft_reset(encoder, true); + chv_data_lane_soft_reset(encoder, crtc_state, true); /* program left/right clock distribution */ if (pipe != PIPE_B) { @@ -832,7 +832,7 @@ void chv_phy_pre_pll_enable(struct intel_encoder *encoder) val |= CHV_PCS_USEDCLKCHANNEL; vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW8(ch), val); - if (intel_crtc->config->lane_count > 2) { + if (crtc_state->lane_count > 2) { val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW8(ch)); val |= CHV_PCS_USEDCLKCHANNEL_OVRRIDE; if (pipe != PIPE_B) @@ -857,16 +857,15 @@ void chv_phy_pre_pll_enable(struct intel_encoder *encoder) mutex_unlock(&dev_priv->sb_lock); } -void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) +void chv_phy_pre_encoder_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct intel_digital_port *dport = dp_to_dig_port(intel_dp); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = - to_intel_crtc(encoder->base.crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); enum dpio_channel ch = vlv_dport_to_channel(dport); - int pipe = intel_crtc->pipe; + enum pipe pipe = crtc->pipe; int data, i, stagger; u32 val; @@ -877,16 +876,16 @@ void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) val &= ~DPIO_LANEDESKEW_STRAP_OVRD; vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW11(ch), val); - if (intel_crtc->config->lane_count > 2) { + if (crtc_state->lane_count > 2) { val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW11(ch)); val &= ~DPIO_LANEDESKEW_STRAP_OVRD; vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW11(ch), val); } /* Program Tx lane latency optimal setting*/ - for (i = 0; i < intel_crtc->config->lane_count; i++) { + for (i = 0; i < crtc_state->lane_count; i++) { /* Set the upar bit */ - if (intel_crtc->config->lane_count == 1) + if (crtc_state->lane_count == 1) data = 0x0; else data = (i == 1) ? 0x0 : 0x1; @@ -895,13 +894,13 @@ void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) } /* Data lane stagger programming */ - if (intel_crtc->config->port_clock > 270000) + if (crtc_state->port_clock > 270000) stagger = 0x18; - else if (intel_crtc->config->port_clock > 135000) + else if (crtc_state->port_clock > 135000) stagger = 0xd; - else if (intel_crtc->config->port_clock > 67500) + else if (crtc_state->port_clock > 67500) stagger = 0x7; - else if (intel_crtc->config->port_clock > 33750) + else if (crtc_state->port_clock > 33750) stagger = 0x4; else stagger = 0x2; @@ -910,7 +909,7 @@ void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) val |= DPIO_TX2_STAGGER_MASK(0x1f); vlv_dpio_write(dev_priv, pipe, VLV_PCS01_DW11(ch), val); - if (intel_crtc->config->lane_count > 2) { + if (crtc_state->lane_count > 2) { val = vlv_dpio_read(dev_priv, pipe, VLV_PCS23_DW11(ch)); val |= DPIO_TX2_STAGGER_MASK(0x1f); vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW11(ch), val); @@ -923,7 +922,7 @@ void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) DPIO_TX1_STAGGER_MULT(6) | DPIO_TX2_STAGGER_MULT(0)); - if (intel_crtc->config->lane_count > 2) { + if (crtc_state->lane_count > 2) { vlv_dpio_write(dev_priv, pipe, VLV_PCS23_DW12(ch), DPIO_LANESTAGGER_STRAP(stagger) | DPIO_LANESTAGGER_STRAP_OVRD | @@ -933,7 +932,7 @@ void chv_phy_pre_encoder_enable(struct intel_encoder *encoder) } /* Deassert data lane reset */ - chv_data_lane_soft_reset(encoder, false); + chv_data_lane_soft_reset(encoder, crtc_state, false); mutex_unlock(&dev_priv->sb_lock); } @@ -949,10 +948,11 @@ void chv_phy_release_cl2_override(struct intel_encoder *encoder) } } -void chv_phy_post_pll_disable(struct intel_encoder *encoder) +void chv_phy_post_pll_disable(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state) { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum pipe pipe = to_intel_crtc(encoder->base.crtc)->pipe; + enum pipe pipe = to_intel_crtc(old_crtc_state->base.crtc)->pipe; u32 val; mutex_lock(&dev_priv->sb_lock); @@ -990,7 +990,7 @@ void vlv_set_phy_signal_level(struct intel_encoder *encoder, struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); enum dpio_channel port = vlv_dport_to_channel(dport); - int pipe = intel_crtc->pipe; + enum pipe pipe = intel_crtc->pipe; mutex_lock(&dev_priv->sb_lock); vlv_dpio_write(dev_priv, pipe, VLV_TX_DW5(port), 0x00000000); @@ -1008,15 +1008,14 @@ void vlv_set_phy_signal_level(struct intel_encoder *encoder, mutex_unlock(&dev_priv->sb_lock); } -void vlv_phy_pre_pll_enable(struct intel_encoder *encoder) +void vlv_phy_pre_pll_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = - to_intel_crtc(encoder->base.crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); enum dpio_channel port = vlv_dport_to_channel(dport); - int pipe = intel_crtc->pipe; + enum pipe pipe = crtc->pipe; /* Program Tx lane resets to default */ mutex_lock(&dev_priv->sb_lock); @@ -1036,15 +1035,15 @@ void vlv_phy_pre_pll_enable(struct intel_encoder *encoder) mutex_unlock(&dev_priv->sb_lock); } -void vlv_phy_pre_encoder_enable(struct intel_encoder *encoder) +void vlv_phy_pre_encoder_enable(struct intel_encoder *encoder, + const struct intel_crtc_state *crtc_state) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); struct intel_digital_port *dport = dp_to_dig_port(intel_dp); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); enum dpio_channel port = vlv_dport_to_channel(dport); - int pipe = intel_crtc->pipe; + enum pipe pipe = crtc->pipe; u32 val; mutex_lock(&dev_priv->sb_lock); @@ -1066,14 +1065,14 @@ void vlv_phy_pre_encoder_enable(struct intel_encoder *encoder) mutex_unlock(&dev_priv->sb_lock); } -void vlv_phy_reset_lanes(struct intel_encoder *encoder) +void vlv_phy_reset_lanes(struct intel_encoder *encoder, + const struct intel_crtc_state *old_crtc_state) { struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - struct intel_crtc *intel_crtc = - to_intel_crtc(encoder->base.crtc); + struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); enum dpio_channel port = vlv_dport_to_channel(dport); - int pipe = intel_crtc->pipe; + enum pipe pipe = crtc->pipe; mutex_lock(&dev_priv->sb_lock); vlv_dpio_write(dev_priv, pipe, VLV_PCS_DW0(port), 0x00000000); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index fa1c793a21efd..1f2258dec09e0 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1683,10 +1683,9 @@ static void vlv_hdmi_pre_enable(struct intel_encoder *encoder, const struct drm_connector_state *conn_state) { struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - vlv_phy_pre_encoder_enable(encoder); + vlv_phy_pre_encoder_enable(encoder, pipe_config); /* HDMI 1.0V-2dB */ vlv_set_phy_signal_level(encoder, 0x2b245f5f, 0x00002000, 0x5578b83a, @@ -1707,7 +1706,7 @@ static void vlv_hdmi_pre_pll_enable(struct intel_encoder *encoder, { intel_hdmi_prepare(encoder, pipe_config); - vlv_phy_pre_pll_enable(encoder); + vlv_phy_pre_pll_enable(encoder, pipe_config); } static void chv_hdmi_pre_pll_enable(struct intel_encoder *encoder, @@ -1716,14 +1715,14 @@ static void chv_hdmi_pre_pll_enable(struct intel_encoder *encoder, { intel_hdmi_prepare(encoder, pipe_config); - chv_phy_pre_pll_enable(encoder); + chv_phy_pre_pll_enable(encoder, pipe_config); } static void chv_hdmi_post_pll_disable(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) { - chv_phy_post_pll_disable(encoder); + chv_phy_post_pll_disable(encoder, old_crtc_state); } static void vlv_hdmi_post_disable(struct intel_encoder *encoder, @@ -1731,7 +1730,7 @@ static void vlv_hdmi_post_disable(struct intel_encoder *encoder, const struct drm_connector_state *old_conn_state) { /* Reset lanes to avoid HDMI flicker (VLV w/a) */ - vlv_phy_reset_lanes(encoder); + vlv_phy_reset_lanes(encoder, old_crtc_state); } static void chv_hdmi_post_disable(struct intel_encoder *encoder, @@ -1744,7 +1743,7 @@ static void chv_hdmi_post_disable(struct intel_encoder *encoder, mutex_lock(&dev_priv->sb_lock); /* Assert data lane reset */ - chv_data_lane_soft_reset(encoder, true); + chv_data_lane_soft_reset(encoder, old_crtc_state, true); mutex_unlock(&dev_priv->sb_lock); } @@ -1757,7 +1756,7 @@ static void chv_hdmi_pre_enable(struct intel_encoder *encoder, struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - chv_phy_pre_encoder_enable(encoder); + chv_phy_pre_encoder_enable(encoder, pipe_config); /* FIXME: Program the support xxx V-dB */ /* Use 800mV-0dB */ -- GitLab From c249f1f423eac64b7aaeaa369935e9c1e6c9f4d1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:19 +0200 Subject: [PATCH 0164/1471] drm/i915: Eliminate crtc->config usage from CRT code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace crtc->config usage with the passed down crtc state. Also take the opportunity for some s/pipe_config/crtc_state/ while at it. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-7-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_crt.c | 44 ++++++++++++++------------------ 1 file changed, 19 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c index 9c000ac612da5..9f31aea51dfff 100644 --- a/drivers/gpu/drm/i915/intel_crt.c +++ b/drivers/gpu/drm/i915/intel_crt.c @@ -219,11 +219,9 @@ static void hsw_disable_crt(struct intel_encoder *encoder, const struct intel_crtc_state *old_crtc_state, const struct drm_connector_state *old_conn_state) { - struct drm_crtc *crtc = old_crtc_state->base.crtc; - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - WARN_ON(!intel_crtc->config->has_pch_encoder); + WARN_ON(!old_crtc_state->has_pch_encoder); intel_set_pch_fifo_underrun_reporting(dev_priv, PIPE_A, false); } @@ -247,46 +245,42 @@ static void hsw_post_disable_crt(struct intel_encoder *encoder, } static void hsw_pre_pll_enable_crt(struct intel_encoder *encoder, - const struct intel_crtc_state *pipe_config, + const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) { - struct drm_crtc *crtc = pipe_config->base.crtc; - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - WARN_ON(!intel_crtc->config->has_pch_encoder); + WARN_ON(!crtc_state->has_pch_encoder); intel_set_pch_fifo_underrun_reporting(dev_priv, PIPE_A, false); } static void hsw_pre_enable_crt(struct intel_encoder *encoder, - const struct intel_crtc_state *pipe_config, + const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) { - struct drm_crtc *crtc = pipe_config->base.crtc; - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + enum pipe pipe = crtc->pipe; - WARN_ON(!intel_crtc->config->has_pch_encoder); + WARN_ON(!crtc_state->has_pch_encoder); intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false); - dev_priv->display.fdi_link_train(intel_crtc, pipe_config); + dev_priv->display.fdi_link_train(crtc, crtc_state); } static void hsw_enable_crt(struct intel_encoder *encoder, - const struct intel_crtc_state *pipe_config, + const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) { - struct drm_crtc *crtc = pipe_config->base.crtc; - struct drm_i915_private *dev_priv = to_i915(crtc->dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + enum pipe pipe = crtc->pipe; - WARN_ON(!intel_crtc->config->has_pch_encoder); + WARN_ON(!crtc_state->has_pch_encoder); - intel_crt_set_dpms(encoder, pipe_config, DRM_MODE_DPMS_ON); + intel_crt_set_dpms(encoder, crtc_state, DRM_MODE_DPMS_ON); intel_wait_for_vblank(dev_priv, pipe); intel_wait_for_vblank(dev_priv, pipe); @@ -295,10 +289,10 @@ static void hsw_enable_crt(struct intel_encoder *encoder, } static void intel_enable_crt(struct intel_encoder *encoder, - const struct intel_crtc_state *pipe_config, + const struct intel_crtc_state *crtc_state, const struct drm_connector_state *conn_state) { - intel_crt_set_dpms(encoder, pipe_config, DRM_MODE_DPMS_ON); + intel_crt_set_dpms(encoder, crtc_state, DRM_MODE_DPMS_ON); } static enum drm_mode_status -- GitLab From 005b5bc694cf7e0a899fd5d02a06459df191e1ce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:20 +0200 Subject: [PATCH 0165/1471] drm/i915: Replace dig_port->port with encoder port for BXT DPLL selection MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace dig_port->port with encoder->port in the BXT DPLL selection. We can do this because both the master encoder and the fake MST encoders have the same encoder->port value, whereas using dig_port->port only worked for the master encoder since the fake encoders were't derived from intel_digital_port. This eliminates the DP MST special case. Do this by hand because spatch is having problems with the control flow due to the dig_port assignment happening in two different branches. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-8-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index a83bf1c38e05f..be74d4767c8a4 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -1802,7 +1802,6 @@ bxt_get_dpll(struct intel_crtc *crtc, { struct intel_dpll_hw_state dpll_hw_state = { }; struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - struct intel_digital_port *intel_dig_port; struct intel_shared_dpll *pll; int i, clock = crtc_state->port_clock; @@ -1820,15 +1819,8 @@ bxt_get_dpll(struct intel_crtc *crtc, crtc_state->dpll_hw_state = dpll_hw_state; - if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_DP_MST)) { - struct intel_dp_mst_encoder *intel_mst = enc_to_mst(&encoder->base); - - intel_dig_port = intel_mst->primary; - } else - intel_dig_port = enc_to_dig_port(&encoder->base); - /* 1:1 mapping between ports and PLLs */ - i = (enum intel_dpll_id) intel_dig_port->port; + i = (enum intel_dpll_id) encoder->port; pll = intel_get_shared_dpll_by_id(dev_priv, i); DRM_DEBUG_KMS("[CRTC:%d:%s] using pre-allocated %s\n", -- GitLab From 8f4f279707236e273cd50b095a06797fe0ed6ccc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 9 Nov 2017 17:24:34 +0200 Subject: [PATCH 0166/1471] drm/i915: Nuke intel_digital_port->port MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Remove intel_digital_port->port and replace its users with intel_encoder->port. intel_encoder->port is a superset of intel_digital_port->port, and it works correctly even for MST encoders. v2: Eliminate a few dp_to_dig_port()->base.port cases too (DK) Performed with cocci: @@ @@ struct intel_digital_port { ... - enum port port; ... } @@ struct intel_digital_port *D; expression E; @@ - D->port = E; @@ struct intel_digital_port *D; @@ - D->port + D->base.port @ expression E; @@ ( - dp_to_dig_port(E)->port + dp_to_dig_port(E)->base.port | - enc_to_dig_port(E)->port + to_intel_encoder(E)->port ) @@ expression E; @@ - to_intel_encoder(&E->base) + E @@ struct intel_digital_port *D; identifier I, M; @@ I = &D->base <... ( - D->base.M + I->M | - &D->base + I ) ...> @@ identifier D; expression E; identifier M; @@ D = enc_to_dig_port(&E->base) <... ( - D->base.M + E->M | - &D->base + E ) ...> @@ identifier D, DP; expression E; identifier M; @@ DP = enc_to_intel_dp(&E->base) <... ( - dp_to_dig_port(DP)->base.M + E->M | - &dp_to_dig_port(DP)->base + E ) ...> @@ expression E; identifier M; @@ ( - enc_to_dig_port(&E->base)->base.M + E->M | - enc_to_dig_port(&E->base)->base + E | - enc_to_mst(&E->base)->primary->base.port + E->port ) @@ expression E; identifier D; @@ - struct intel_digital_port *D = E; ... when != D Cc: Dhinakaran Pandiyan Reviewed-by: Dhinakaran Pandiyan Acked-by: Jani Nikula Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171109152434.32074-1-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/i915_debugfs.c | 2 +- drivers/gpu/drm/i915/intel_ddi.c | 13 ++-- drivers/gpu/drm/i915/intel_display.c | 13 ++-- drivers/gpu/drm/i915/intel_dp.c | 103 +++++++++++++------------- drivers/gpu/drm/i915/intel_dp_mst.c | 6 +- drivers/gpu/drm/i915/intel_dpio_phy.c | 12 ++- drivers/gpu/drm/i915/intel_drv.h | 5 +- drivers/gpu/drm/i915/intel_hdmi.c | 17 ++--- drivers/gpu/drm/i915/intel_pipe_crc.c | 4 +- drivers/gpu/drm/i915/intel_psr.c | 4 +- 10 files changed, 87 insertions(+), 92 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 39883cd915db1..d89321f0468c0 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3603,7 +3603,7 @@ static int i915_dp_mst_info(struct seq_file *m, void *unused) continue; seq_printf(m, "MST Source Port %c\n", - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); drm_dp_mst_dump_topology(m, &intel_dig_port->dp.mst_mgr); } drm_connector_list_iter_end(&conn_iter); diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index ace674cd79b92..65ab55496ab74 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1801,8 +1801,8 @@ static void skl_ddi_set_iboost(struct intel_encoder *encoder, int level, enum intel_output_type type) { struct intel_digital_port *intel_dig_port = enc_to_dig_port(&encoder->base); - struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev); - enum port port = intel_dig_port->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum port port = encoder->port; uint8_t iboost; if (type == INTEL_OUTPUT_HDMI) @@ -2470,7 +2470,7 @@ void intel_ddi_prepare_link_retrain(struct intel_dp *intel_dp) struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; uint32_t val; bool wait = false; @@ -2698,7 +2698,7 @@ static struct intel_connector * intel_ddi_init_dp_connector(struct intel_digital_port *intel_dig_port) { struct intel_connector *connector; - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; connector = intel_connector_alloc(); if (!connector) @@ -2717,7 +2717,7 @@ static struct intel_connector * intel_ddi_init_hdmi_connector(struct intel_digital_port *intel_dig_port) { struct intel_connector *connector; - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; connector = intel_connector_alloc(); if (!connector) @@ -2733,7 +2733,7 @@ static bool intel_ddi_a_force_4_lanes(struct intel_digital_port *dport) { struct drm_i915_private *dev_priv = to_i915(dport->base.base.dev); - if (dport->port != PORT_A) + if (dport->base.port != PORT_A) return false; if (dport->saved_port_bits & DDI_A_4_LANES) @@ -2836,7 +2836,6 @@ void intel_ddi_init(struct drm_i915_private *dev_priv, enum port port) intel_encoder->suspend = intel_dp_encoder_suspend; intel_encoder->get_power_domains = intel_ddi_get_power_domains; - intel_dig_port->port = port; intel_dig_port->saved_port_bits = I915_READ(DDI_BUF_CTL(port)) & (DDI_BUF_PORT_REVERSAL | DDI_A_4_LANES); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 84817ccc53053..435dc18aa2394 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1701,7 +1701,7 @@ void vlv_wait_port_ready(struct drm_i915_private *dev_priv, u32 port_mask; i915_reg_t dpll_reg; - switch (dport->port) { + switch (dport->base.port) { case PORT_B: port_mask = DPLL_PORTB_READY_MASK; dpll_reg = DPLL(0); @@ -1723,7 +1723,8 @@ void vlv_wait_port_ready(struct drm_i915_private *dev_priv, dpll_reg, port_mask, expected_mask, 1000)) WARN(1, "timed out waiting for port %c ready: got 0x%x, expected 0x%x\n", - port_name(dport->port), I915_READ(dpll_reg) & port_mask, expected_mask); + port_name(dport->base.port), + I915_READ(dpll_reg) & port_mask, expected_mask); } static void ironlake_enable_pch_transcoder(struct drm_i915_private *dev_priv, @@ -4481,7 +4482,7 @@ intel_trans_dp_port_sel(struct intel_crtc *crtc) for_each_encoder_on_crtc(dev, &crtc->base, encoder) { if (encoder->type == INTEL_OUTPUT_DP || encoder->type == INTEL_OUTPUT_EDP) - return enc_to_dig_port(&encoder->base)->port; + return encoder->port; } return -1; @@ -7632,7 +7633,7 @@ static void ironlake_init_pch_refclk(struct drm_i915_private *dev_priv) break; case INTEL_OUTPUT_EDP: has_panel = true; - if (enc_to_dig_port(&encoder->base)->port == PORT_A) + if (encoder->port == PORT_A) has_cpu_edp = true; break; default: @@ -10755,7 +10756,7 @@ static bool check_digital_port_conflicts(struct drm_atomic_state *state) case INTEL_OUTPUT_DP: case INTEL_OUTPUT_HDMI: case INTEL_OUTPUT_EDP: - port_mask = 1 << enc_to_dig_port(&encoder->base)->port; + port_mask = 1 << encoder->port; /* the same port mustn't appear more than once */ if (used_ports & port_mask) @@ -10765,7 +10766,7 @@ static bool check_digital_port_conflicts(struct drm_atomic_state *state) break; case INTEL_OUTPUT_DP_MST: used_mst_ports |= - 1 << enc_to_mst(&encoder->base)->primary->port; + 1 << encoder->port; break; default: break; diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 0151c8d18b79e..9b0c6dea40bc2 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -223,7 +223,7 @@ intel_dp_set_source_rates(struct intel_dp *intel_dp) { struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - enum port port = dig_port->port; + enum port port = dig_port->base.port; const int *source_rates; int size; u32 voltage; @@ -479,11 +479,11 @@ vlv_power_sequencer_kick(struct intel_dp *intel_dp) if (WARN(I915_READ(intel_dp->output_reg) & DP_PORT_EN, "skipping pipe %c power seqeuncer kick due to port %c being active\n", - pipe_name(pipe), port_name(intel_dig_port->port))) + pipe_name(pipe), port_name(intel_dig_port->base.port))) return; DRM_DEBUG_KMS("kicking pipe %c power sequencer for port %c\n", - pipe_name(pipe), port_name(intel_dig_port->port)); + pipe_name(pipe), port_name(intel_dig_port->base.port)); /* Preserve the BIOS-computed detected bit. This is * supposed to be read-only. @@ -610,7 +610,7 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) DRM_DEBUG_KMS("picked pipe %c power sequencer for port %c\n", pipe_name(intel_dp->pps_pipe), - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); /* init power sequencer on this pipe and port */ intel_dp_init_panel_power_sequencer(dev, intel_dp); @@ -706,7 +706,7 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; lockdep_assert_held(&dev_priv->pps_mutex); @@ -966,7 +966,7 @@ static uint32_t ilk_get_aux_clock_divider(struct intel_dp *intel_dp, int index) * like to run at 2MHz. So, take the cdclk or PCH rawclk value and * divide by 2000 and use that */ - if (intel_dig_port->port == PORT_A) + if (intel_dig_port->base.port == PORT_A) return DIV_ROUND_CLOSEST(dev_priv->cdclk.hw.cdclk, 2000); else return DIV_ROUND_CLOSEST(dev_priv->rawclk_freq, 2000); @@ -977,7 +977,7 @@ static uint32_t hsw_get_aux_clock_divider(struct intel_dp *intel_dp, int index) struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev); - if (intel_dig_port->port != PORT_A && HAS_PCH_LPT_H(dev_priv)) { + if (intel_dig_port->base.port != PORT_A && HAS_PCH_LPT_H(dev_priv)) { /* Workaround for non-ULT HSW */ switch (index) { case 0: return 63; @@ -1447,7 +1447,7 @@ static void intel_aux_reg_init(struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); enum port port = intel_aux_port(dev_priv, - dp_to_dig_port(intel_dp)->port); + dp_to_dig_port(intel_dp)->base.port); int i; intel_dp->aux_ch_ctl_reg = intel_aux_ctl_reg(dev_priv, port); @@ -1465,7 +1465,7 @@ static void intel_dp_aux_init(struct intel_dp *intel_dp) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; intel_aux_reg_init(intel_dp); drm_dp_aux_init(&intel_dp->aux); @@ -1635,7 +1635,7 @@ intel_dp_compute_config(struct intel_encoder *encoder, struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; struct intel_crtc *intel_crtc = to_intel_crtc(pipe_config->base.crtc); struct intel_connector *intel_connector = intel_dp->attached_connector; struct intel_digital_connector_state *intel_conn_state = @@ -1859,7 +1859,7 @@ static void intel_dp_prepare(struct intel_encoder *encoder, struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); const struct drm_display_mode *adjusted_mode = &pipe_config->base.adjusted_mode; @@ -2074,7 +2074,7 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) intel_display_power_get(dev_priv, intel_dp->aux_power_domain); DRM_DEBUG_KMS("Turning eDP port %c VDD on\n", - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); if (!edp_have_panel_power(intel_dp)) wait_panel_power_cycle(intel_dp); @@ -2094,7 +2094,7 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) */ if (!edp_have_panel_power(intel_dp)) { DRM_DEBUG_KMS("eDP port %c panel power wasn't enabled\n", - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); msleep(intel_dp->panel_power_up_delay); } @@ -2120,7 +2120,7 @@ void intel_edp_panel_vdd_on(struct intel_dp *intel_dp) pps_unlock(intel_dp); I915_STATE_WARN(!vdd, "eDP port %c VDD already requested on\n", - port_name(dp_to_dig_port(intel_dp)->port)); + port_name(dp_to_dig_port(intel_dp)->base.port)); } static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) @@ -2140,7 +2140,7 @@ static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) return; DRM_DEBUG_KMS("Turning eDP port %c VDD off\n", - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); pp = ironlake_get_pp_control(intel_dp); pp &= ~EDP_FORCE_VDD; @@ -2200,7 +2200,7 @@ static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) return; I915_STATE_WARN(!intel_dp->want_panel_vdd, "eDP port %c VDD not forced on", - port_name(dp_to_dig_port(intel_dp)->port)); + port_name(dp_to_dig_port(intel_dp)->base.port)); intel_dp->want_panel_vdd = false; @@ -2223,11 +2223,11 @@ static void edp_panel_on(struct intel_dp *intel_dp) return; DRM_DEBUG_KMS("Turn eDP port %c panel power on\n", - port_name(dp_to_dig_port(intel_dp)->port)); + port_name(dp_to_dig_port(intel_dp)->base.port)); if (WARN(edp_have_panel_power(intel_dp), "eDP port %c panel power already on\n", - port_name(dp_to_dig_port(intel_dp)->port))) + port_name(dp_to_dig_port(intel_dp)->base.port))) return; wait_panel_power_cycle(intel_dp); @@ -2282,10 +2282,10 @@ static void edp_panel_off(struct intel_dp *intel_dp) return; DRM_DEBUG_KMS("Turn eDP port %c panel power off\n", - port_name(dp_to_dig_port(intel_dp)->port)); + port_name(dp_to_dig_port(intel_dp)->base.port)); WARN(!intel_dp->want_panel_vdd, "Need eDP port %c VDD to turn off panel\n", - port_name(dp_to_dig_port(intel_dp)->port)); + port_name(dp_to_dig_port(intel_dp)->base.port)); pp = ironlake_get_pp_control(intel_dp); /* We need to switch off panel power _and_ force vdd, for otherwise some @@ -2437,7 +2437,7 @@ static void assert_dp_port(struct intel_dp *intel_dp, bool state) I915_STATE_WARN(cur_state != state, "DP port %c state assertion failure (expected %s, current %s)\n", - port_name(dig_port->port), + port_name(dig_port->base.port), onoff(state), onoff(cur_state)); } #define assert_dp_port_disabled(d) assert_dp_port((d), false) @@ -2570,7 +2570,7 @@ static bool intel_dp_get_hw_state(struct intel_encoder *encoder, enum pipe *pipe) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); u32 tmp; @@ -2625,7 +2625,7 @@ static void intel_dp_get_config(struct intel_encoder *encoder, u32 tmp, flags = 0; struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); if (encoder->type == INTEL_OUTPUT_EDP) @@ -2794,7 +2794,7 @@ _intel_dp_set_link_train(struct intel_dp *intel_dp, struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; if (dp_train_pat & DP_TRAINING_PATTERN_MASK) DRM_DEBUG_KMS("Using DP training pattern TPS%d\n", @@ -2969,7 +2969,7 @@ static void g4x_pre_enable_dp(struct intel_encoder *encoder, const struct drm_connector_state *conn_state) { struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = encoder->port; intel_dp_prepare(encoder, pipe_config); @@ -3002,7 +3002,7 @@ static void vlv_detach_power_sequencer(struct intel_dp *intel_dp) * from a port. */ DRM_DEBUG_KMS("detaching pipe %c power sequencer from port %c\n", - pipe_name(pipe), port_name(intel_dig_port->port)); + pipe_name(pipe), port_name(intel_dig_port->base.port)); I915_WRITE(pp_on_reg, 0); POSTING_READ(pp_on_reg); @@ -3026,7 +3026,7 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, continue; intel_dp = enc_to_intel_dp(&encoder->base); - port = dp_to_dig_port(intel_dp)->port; + port = dp_to_dig_port(intel_dp)->base.port; WARN(intel_dp->active_pipe == pipe, "stealing pipe %c power sequencer from active (e)DP port %c\n", @@ -3178,7 +3178,7 @@ uint8_t intel_dp_voltage_max(struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = dp_to_dig_port(intel_dp)->base.port; if (INTEL_GEN(dev_priv) >= 9) { struct intel_encoder *encoder = &dp_to_dig_port(intel_dp)->base; @@ -3197,7 +3197,7 @@ uint8_t intel_dp_pre_emphasis_max(struct intel_dp *intel_dp, uint8_t voltage_swing) { struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = dp_to_dig_port(intel_dp)->base.port; if (INTEL_GEN(dev_priv) >= 9) { switch (voltage_swing & DP_TRAIN_VOLTAGE_SWING_MASK) { @@ -3531,7 +3531,7 @@ void intel_dp_set_signal_levels(struct intel_dp *intel_dp) { struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = to_i915(dev); uint32_t signal_levels, mask = 0; @@ -3591,7 +3591,7 @@ void intel_dp_set_idle_link_train(struct intel_dp *intel_dp) struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); struct drm_device *dev = intel_dig_port->base.base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; uint32_t val; if (!HAS_DDI(dev_priv)) @@ -4490,7 +4490,7 @@ static bool ibx_digital_port_connected(struct drm_i915_private *dev_priv, { u32 bit; - switch (port->port) { + switch (port->base.port) { case PORT_B: bit = SDE_PORTB_HOTPLUG; break; @@ -4501,7 +4501,7 @@ static bool ibx_digital_port_connected(struct drm_i915_private *dev_priv, bit = SDE_PORTD_HOTPLUG; break; default: - MISSING_CASE(port->port); + MISSING_CASE(port->base.port); return false; } @@ -4513,7 +4513,7 @@ static bool cpt_digital_port_connected(struct drm_i915_private *dev_priv, { u32 bit; - switch (port->port) { + switch (port->base.port) { case PORT_B: bit = SDE_PORTB_HOTPLUG_CPT; break; @@ -4524,7 +4524,7 @@ static bool cpt_digital_port_connected(struct drm_i915_private *dev_priv, bit = SDE_PORTD_HOTPLUG_CPT; break; default: - MISSING_CASE(port->port); + MISSING_CASE(port->base.port); return false; } @@ -4536,7 +4536,7 @@ static bool spt_digital_port_connected(struct drm_i915_private *dev_priv, { u32 bit; - switch (port->port) { + switch (port->base.port) { case PORT_A: bit = SDE_PORTA_HOTPLUG_SPT; break; @@ -4555,7 +4555,7 @@ static bool g4x_digital_port_connected(struct drm_i915_private *dev_priv, { u32 bit; - switch (port->port) { + switch (port->base.port) { case PORT_B: bit = PORTB_HOTPLUG_LIVE_STATUS_G4X; break; @@ -4566,7 +4566,7 @@ static bool g4x_digital_port_connected(struct drm_i915_private *dev_priv, bit = PORTD_HOTPLUG_LIVE_STATUS_G4X; break; default: - MISSING_CASE(port->port); + MISSING_CASE(port->base.port); return false; } @@ -4578,7 +4578,7 @@ static bool gm45_digital_port_connected(struct drm_i915_private *dev_priv, { u32 bit; - switch (port->port) { + switch (port->base.port) { case PORT_B: bit = PORTB_HOTPLUG_LIVE_STATUS_GM45; break; @@ -4589,7 +4589,7 @@ static bool gm45_digital_port_connected(struct drm_i915_private *dev_priv, bit = PORTD_HOTPLUG_LIVE_STATUS_GM45; break; default: - MISSING_CASE(port->port); + MISSING_CASE(port->base.port); return false; } @@ -4599,7 +4599,7 @@ static bool gm45_digital_port_connected(struct drm_i915_private *dev_priv, static bool ilk_digital_port_connected(struct drm_i915_private *dev_priv, struct intel_digital_port *port) { - if (port->port == PORT_A) + if (port->base.port == PORT_A) return I915_READ(DEISR) & DE_DP_A_HOTPLUG; else return ibx_digital_port_connected(dev_priv, port); @@ -4608,7 +4608,7 @@ static bool ilk_digital_port_connected(struct drm_i915_private *dev_priv, static bool snb_digital_port_connected(struct drm_i915_private *dev_priv, struct intel_digital_port *port) { - if (port->port == PORT_A) + if (port->base.port == PORT_A) return I915_READ(DEISR) & DE_DP_A_HOTPLUG; else return cpt_digital_port_connected(dev_priv, port); @@ -4617,7 +4617,7 @@ static bool snb_digital_port_connected(struct drm_i915_private *dev_priv, static bool ivb_digital_port_connected(struct drm_i915_private *dev_priv, struct intel_digital_port *port) { - if (port->port == PORT_A) + if (port->base.port == PORT_A) return I915_READ(DEISR) & DE_DP_A_HOTPLUG_IVB; else return cpt_digital_port_connected(dev_priv, port); @@ -4626,7 +4626,7 @@ static bool ivb_digital_port_connected(struct drm_i915_private *dev_priv, static bool bdw_digital_port_connected(struct drm_i915_private *dev_priv, struct intel_digital_port *port) { - if (port->port == PORT_A) + if (port->base.port == PORT_A) return I915_READ(GEN8_DE_PORT_ISR) & GEN8_PORT_DP_A_HOTPLUG; else return cpt_digital_port_connected(dev_priv, port); @@ -5109,12 +5109,12 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd) * "vdd off -> long hpd -> vdd on -> detect -> vdd off -> ..." */ DRM_DEBUG_KMS("ignoring long hpd on eDP port %c\n", - port_name(intel_dig_port->port)); + port_name(intel_dig_port->base.port)); return IRQ_HANDLED; } DRM_DEBUG_KMS("got hpd irq on port %c - %s\n", - port_name(intel_dig_port->port), + port_name(intel_dig_port->base.port), long_hpd ? "long" : "short"); if (long_hpd) { @@ -5364,7 +5364,7 @@ intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, u32 pp_on, pp_off, pp_div, port_sel = 0; int div = dev_priv->rawclk_freq / 1000; struct pps_registers regs; - enum port port = dp_to_dig_port(intel_dp)->port; + enum port port = dp_to_dig_port(intel_dp)->base.port; const struct edp_power_seq *seq = &intel_dp->pps_delays; lockdep_assert_held(&dev_priv->pps_mutex); @@ -5965,9 +5965,9 @@ intel_dp_init_connector_port_info(struct intel_digital_port *intel_dig_port) struct intel_encoder *encoder = &intel_dig_port->base; struct intel_dp *intel_dp = &intel_dig_port->dp; - encoder->hpd_pin = intel_hpd_pin(intel_dig_port->port); + encoder->hpd_pin = intel_hpd_pin(encoder->port); - switch (intel_dig_port->port) { + switch (encoder->port) { case PORT_A: intel_dp->aux_power_domain = POWER_DOMAIN_AUX_A; break; @@ -5985,7 +5985,7 @@ intel_dp_init_connector_port_info(struct intel_digital_port *intel_dig_port) intel_dp->aux_power_domain = POWER_DOMAIN_AUX_D; break; default: - MISSING_CASE(intel_dig_port->port); + MISSING_CASE(encoder->port); } } @@ -6021,7 +6021,7 @@ intel_dp_init_connector(struct intel_digital_port *intel_dig_port, struct intel_encoder *intel_encoder = &intel_dig_port->base; struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_dig_port->port; + enum port port = intel_encoder->port; int type; /* Initialize the work for modeset in case of link train failure */ @@ -6191,7 +6191,6 @@ bool intel_dp_init(struct drm_i915_private *dev_priv, intel_encoder->disable = g4x_disable_dp; } - intel_dig_port->port = port; intel_dig_port->dp.output_reg = output_reg; intel_dig_port->max_lanes = 4; diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c index 653ca39789b8f..c3de0918ee13f 100644 --- a/drivers/gpu/drm/i915/intel_dp_mst.c +++ b/drivers/gpu/drm/i915/intel_dp_mst.c @@ -209,7 +209,7 @@ static void intel_mst_pre_enable_dp(struct intel_encoder *encoder, struct intel_digital_port *intel_dig_port = intel_mst->primary; struct intel_dp *intel_dp = &intel_dig_port->dp; struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; struct intel_connector *connector = to_intel_connector(conn_state->connector); int ret; @@ -253,7 +253,7 @@ static void intel_mst_enable_dp(struct intel_encoder *encoder, struct intel_digital_port *intel_dig_port = intel_mst->primary; struct intel_dp *intel_dp = &intel_dig_port->dp; struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); - enum port port = intel_dig_port->port; + enum port port = intel_dig_port->base.port; int ret; DRM_DEBUG_KMS("active links %d\n", intel_dp->active_mst_links); @@ -552,7 +552,7 @@ intel_dp_create_fake_mst_encoder(struct intel_digital_port *intel_dig_port, enum intel_encoder->type = INTEL_OUTPUT_DP_MST; intel_encoder->power_domain = intel_dig_port->base.power_domain; - intel_encoder->port = intel_dig_port->port; + intel_encoder->port = intel_dig_port->base.port; intel_encoder->crtc_mask = 0x7; intel_encoder->cloneable = 0; diff --git a/drivers/gpu/drm/i915/intel_dpio_phy.c b/drivers/gpu/drm/i915/intel_dpio_phy.c index 5958d3d8b90e4..76473e9836c60 100644 --- a/drivers/gpu/drm/i915/intel_dpio_phy.c +++ b/drivers/gpu/drm/i915/intel_dpio_phy.c @@ -586,9 +586,8 @@ bxt_ddi_phy_calc_lane_lat_optim_mask(uint8_t lane_count) void bxt_ddi_phy_set_lane_optim_mask(struct intel_encoder *encoder, uint8_t lane_lat_optim_mask) { - struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct drm_i915_private *dev_priv = to_i915(dport->base.base.dev); - enum port port = dport->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum port port = encoder->port; enum dpio_phy phy; enum dpio_channel ch; int lane; @@ -613,9 +612,8 @@ void bxt_ddi_phy_set_lane_optim_mask(struct intel_encoder *encoder, uint8_t bxt_ddi_phy_get_lane_lat_optim_mask(struct intel_encoder *encoder) { - struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct drm_i915_private *dev_priv = to_i915(dport->base.base.dev); - enum port port = dport->port; + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); + enum port port = encoder->port; enum dpio_phy phy; enum dpio_channel ch; int lane; @@ -641,7 +639,7 @@ void chv_set_phy_signal_level(struct intel_encoder *encoder, { struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_digital_port *dport = enc_to_dig_port(&encoder->base); - struct intel_crtc *intel_crtc = to_intel_crtc(dport->base.base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(encoder->base.crtc); enum dpio_channel ch = vlv_dport_to_channel(dport); enum pipe pipe = intel_crtc->pipe; u32 val; diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 00b4886880427..a0a5cead868c9 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1056,7 +1056,6 @@ struct intel_lspcon { struct intel_digital_port { struct intel_encoder base; - enum port port; u32 saved_port_bits; struct intel_dp dp; struct intel_hdmi hdmi; @@ -1088,7 +1087,7 @@ struct intel_dp_mst_encoder { static inline enum dpio_channel vlv_dport_to_channel(struct intel_digital_port *dport) { - switch (dport->port) { + switch (dport->base.port) { case PORT_B: case PORT_D: return DPIO_CH0; @@ -1102,7 +1101,7 @@ vlv_dport_to_channel(struct intel_digital_port *dport) static inline enum dpio_phy vlv_dport_to_phy(struct intel_digital_port *dport) { - switch (dport->port) { + switch (dport->base.port) { case PORT_B: case PORT_C: return DPIO_PHY0; diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 1f2258dec09e0..2d95db64cdf28 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -186,7 +186,7 @@ static bool g4x_infoframe_enabled(struct drm_encoder *encoder, if ((val & VIDEO_DIP_ENABLE) == 0) return false; - if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->port)) + if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->base.port)) return false; return val & (VIDEO_DIP_ENABLE_AVI | @@ -245,7 +245,7 @@ static bool ibx_infoframe_enabled(struct drm_encoder *encoder, if ((val & VIDEO_DIP_ENABLE) == 0) return false; - if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->port)) + if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->base.port)) return false; return val & (VIDEO_DIP_ENABLE_AVI | @@ -362,7 +362,7 @@ static bool vlv_infoframe_enabled(struct drm_encoder *encoder, if ((val & VIDEO_DIP_ENABLE) == 0) return false; - if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->port)) + if ((val & VIDEO_DIP_PORT_MASK) != VIDEO_DIP_PORT(intel_dig_port->base.port)) return false; return val & (VIDEO_DIP_ENABLE_AVI | @@ -535,7 +535,7 @@ static void g4x_set_infoframes(struct drm_encoder *encoder, struct intel_hdmi *intel_hdmi = &intel_dig_port->hdmi; i915_reg_t reg = VIDEO_DIP_CTL; u32 val = I915_READ(reg); - u32 port = VIDEO_DIP_PORT(intel_dig_port->port); + u32 port = VIDEO_DIP_PORT(intel_dig_port->base.port); assert_hdmi_port_disabled(intel_hdmi); @@ -686,7 +686,7 @@ static void ibx_set_infoframes(struct drm_encoder *encoder, struct intel_hdmi *intel_hdmi = &intel_dig_port->hdmi; i915_reg_t reg = TVIDEO_DIP_CTL(intel_crtc->pipe); u32 val = I915_READ(reg); - u32 port = VIDEO_DIP_PORT(intel_dig_port->port); + u32 port = VIDEO_DIP_PORT(intel_dig_port->base.port); assert_hdmi_port_disabled(intel_hdmi); @@ -782,7 +782,7 @@ static void vlv_set_infoframes(struct drm_encoder *encoder, struct intel_hdmi *intel_hdmi = enc_to_intel_hdmi(encoder); i915_reg_t reg = VLV_TVIDEO_DIP_CTL(intel_crtc->pipe); u32 val = I915_READ(reg); - u32 port = VIDEO_DIP_PORT(intel_dig_port->port); + u32 port = VIDEO_DIP_PORT(intel_dig_port->base.port); assert_hdmi_port_disabled(intel_hdmi); @@ -1545,7 +1545,7 @@ intel_hdmi_dp_dual_mode_detect(struct drm_connector *connector, bool has_edid) { struct drm_i915_private *dev_priv = to_i915(connector->dev); struct intel_hdmi *hdmi = intel_attached_hdmi(connector); - enum port port = hdmi_to_dig_port(hdmi)->port; + enum port port = hdmi_to_dig_port(hdmi)->base.port; struct i2c_adapter *adapter = intel_gmbus_get_adapter(dev_priv, hdmi->ddc_bus); enum drm_dp_dual_mode_type type = drm_dp_dual_mode_detect(adapter); @@ -2015,7 +2015,7 @@ void intel_hdmi_init_connector(struct intel_digital_port *intel_dig_port, struct intel_encoder *intel_encoder = &intel_dig_port->base; struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - enum port port = intel_dig_port->port; + enum port port = intel_encoder->port; DRM_DEBUG_KMS("Adding HDMI connector on port %c\n", port_name(port)); @@ -2135,7 +2135,6 @@ void intel_hdmi_init(struct drm_i915_private *dev_priv, if (IS_G4X(dev_priv)) intel_encoder->cloneable |= 1 << INTEL_OUTPUT_HDMI; - intel_dig_port->port = port; intel_dig_port->hdmi.hdmi_reg = hdmi_reg; intel_dig_port->dp.output_reg = INVALID_MMIO_REG; intel_dig_port->max_lanes = 4; diff --git a/drivers/gpu/drm/i915/intel_pipe_crc.c b/drivers/gpu/drm/i915/intel_pipe_crc.c index 899839f2f7c6f..61641d479b938 100644 --- a/drivers/gpu/drm/i915/intel_pipe_crc.c +++ b/drivers/gpu/drm/i915/intel_pipe_crc.c @@ -269,7 +269,7 @@ static int i9xx_pipe_crc_auto_source(struct drm_i915_private *dev_priv, case INTEL_OUTPUT_DP: case INTEL_OUTPUT_EDP: dig_port = enc_to_dig_port(&encoder->base); - switch (dig_port->port) { + switch (dig_port->base.port) { case PORT_B: *source = INTEL_PIPE_CRC_SOURCE_DP_B; break; @@ -281,7 +281,7 @@ static int i9xx_pipe_crc_auto_source(struct drm_i915_private *dev_priv, break; default: WARN(1, "nonexisting DP port %c\n", - port_name(dig_port->port)); + port_name(dig_port->base.port)); break; } break; diff --git a/drivers/gpu/drm/i915/intel_psr.c b/drivers/gpu/drm/i915/intel_psr.c index 6e3b430fccdc7..a1ad85fa5c1ae 100644 --- a/drivers/gpu/drm/i915/intel_psr.c +++ b/drivers/gpu/drm/i915/intel_psr.c @@ -163,7 +163,7 @@ static void hsw_psr_enable_sink(struct intel_dp *intel_dp) [3] = 1 - 1, [4] = DP_SET_POWER_D0, }; - enum port port = dig_port->port; + enum port port = dig_port->base.port; u32 aux_ctl; int i; @@ -376,7 +376,7 @@ void intel_psr_compute_config(struct intel_dp *intel_dp, * ones. Since by Display design transcoder EDP is tied to port A * we can safely escape based on the port A. */ - if (HAS_DDI(dev_priv) && dig_port->port != PORT_A) { + if (HAS_DDI(dev_priv) && dig_port->base.port != PORT_A) { DRM_DEBUG_KMS("PSR condition failed: Port not supported\n"); return; } -- GitLab From 46bd8383d8c940516196f9c28422e127f3b01dc1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 31 Oct 2017 22:51:22 +0200 Subject: [PATCH 0167/1471] drm/i915: Clean up PPS code calling conventions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit No need to pass 'dev' or 'dev_priv' when the function already takes 'intel_dp'. Also let's prefer passing 'dev_priv' instead of 'dev' when we have to pass one or the other. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171031205123.13123-10-ville.syrjala@linux.intel.com Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_dp.c | 99 ++++++++++++++------------------- 1 file changed, 42 insertions(+), 57 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 9b0c6dea40bc2..37bfe8b272218 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -135,7 +135,7 @@ static bool edp_panel_vdd_on(struct intel_dp *intel_dp); static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync); static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state); -static void vlv_steal_power_sequencer(struct drm_device *dev, +static void vlv_steal_power_sequencer(struct drm_i915_private *dev_priv, enum pipe pipe); static void intel_dp_unset_edid(struct intel_dp *intel_dp); @@ -429,14 +429,12 @@ static void intel_dp_unpack_aux(uint32_t src, uint8_t *dst, int dst_bytes) } static void -intel_dp_init_panel_power_sequencer(struct drm_device *dev, - struct intel_dp *intel_dp); +intel_dp_init_panel_power_sequencer(struct intel_dp *intel_dp); static void -intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, - struct intel_dp *intel_dp, +intel_dp_init_panel_power_sequencer_registers(struct intel_dp *intel_dp, bool force_disable_vdd); static void -intel_dp_pps_init(struct drm_device *dev, struct intel_dp *intel_dp); +intel_dp_pps_init(struct intel_dp *intel_dp); static void pps_lock(struct intel_dp *intel_dp) { @@ -580,9 +578,8 @@ static enum pipe vlv_find_free_pps(struct drm_i915_private *dev_priv) static enum pipe vlv_power_sequencer_pipe(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); enum pipe pipe; lockdep_assert_held(&dev_priv->pps_mutex); @@ -605,7 +602,7 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) if (WARN_ON(pipe == INVALID_PIPE)) pipe = PIPE_A; - vlv_steal_power_sequencer(dev, pipe); + vlv_steal_power_sequencer(dev_priv, pipe); intel_dp->pps_pipe = pipe; DRM_DEBUG_KMS("picked pipe %c power sequencer for port %c\n", @@ -613,8 +610,8 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) port_name(intel_dig_port->base.port)); /* init power sequencer on this pipe and port */ - intel_dp_init_panel_power_sequencer(dev, intel_dp); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, true); + intel_dp_init_panel_power_sequencer(intel_dp); + intel_dp_init_panel_power_sequencer_registers(intel_dp, true); /* * Even vdd force doesn't work until we've made @@ -628,9 +625,7 @@ vlv_power_sequencer_pipe(struct intel_dp *intel_dp) static int bxt_power_sequencer_idx(struct intel_dp *intel_dp) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); lockdep_assert_held(&dev_priv->pps_mutex); @@ -651,7 +646,7 @@ bxt_power_sequencer_idx(struct intel_dp *intel_dp) * Only the HW needs to be reprogrammed, the SW state is fixed and * has been setup during connector init. */ - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, false); + intel_dp_init_panel_power_sequencer_registers(intel_dp, false); return 0; } @@ -703,9 +698,8 @@ vlv_initial_pps_pipe(struct drm_i915_private *dev_priv, static void vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); enum port port = intel_dig_port->base.port; lockdep_assert_held(&dev_priv->pps_mutex); @@ -733,8 +727,8 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) DRM_DEBUG_KMS("initial power sequencer for port %c: pipe %c\n", port_name(port), pipe_name(intel_dp->pps_pipe)); - intel_dp_init_panel_power_sequencer(dev, intel_dp); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, false); + intel_dp_init_panel_power_sequencer(intel_dp); + intel_dp_init_panel_power_sequencer_registers(intel_dp, false); } void intel_power_sequencer_reset(struct drm_i915_private *dev_priv) @@ -790,10 +784,10 @@ struct pps_registers { i915_reg_t pp_div; }; -static void intel_pps_get_registers(struct drm_i915_private *dev_priv, - struct intel_dp *intel_dp, +static void intel_pps_get_registers(struct intel_dp *intel_dp, struct pps_registers *regs) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); int pps_idx = 0; memset(regs, 0, sizeof(*regs)); @@ -816,8 +810,7 @@ _pp_ctrl_reg(struct intel_dp *intel_dp) { struct pps_registers regs; - intel_pps_get_registers(to_i915(intel_dp_to_dev(intel_dp)), intel_dp, - ®s); + intel_pps_get_registers(intel_dp, ®s); return regs.pp_ctrl; } @@ -827,8 +820,7 @@ _pp_stat_reg(struct intel_dp *intel_dp) { struct pps_registers regs; - intel_pps_get_registers(to_i915(intel_dp_to_dev(intel_dp)), intel_dp, - ®s); + intel_pps_get_registers(intel_dp, ®s); return regs.pp_stat; } @@ -1947,8 +1939,7 @@ static void intel_dp_prepare(struct intel_encoder *encoder, #define IDLE_CYCLE_MASK (PP_ON | PP_SEQUENCE_MASK | PP_CYCLE_DELAY_ACTIVE | PP_SEQUENCE_STATE_MASK) #define IDLE_CYCLE_VALUE (0 | PP_SEQUENCE_NONE | 0 | PP_SEQUENCE_STATE_OFF_IDLE) -static void intel_pps_verify_state(struct drm_i915_private *dev_priv, - struct intel_dp *intel_dp); +static void intel_pps_verify_state(struct intel_dp *intel_dp); static void wait_panel_status(struct intel_dp *intel_dp, u32 mask, @@ -1960,7 +1951,7 @@ static void wait_panel_status(struct intel_dp *intel_dp, lockdep_assert_held(&dev_priv->pps_mutex); - intel_pps_verify_state(dev_priv, intel_dp); + intel_pps_verify_state(intel_dp); pp_stat_reg = _pp_stat_reg(intel_dp); pp_ctrl_reg = _pp_ctrl_reg(intel_dp); @@ -3009,15 +3000,14 @@ static void vlv_detach_power_sequencer(struct intel_dp *intel_dp) intel_dp->pps_pipe = INVALID_PIPE; } -static void vlv_steal_power_sequencer(struct drm_device *dev, +static void vlv_steal_power_sequencer(struct drm_i915_private *dev_priv, enum pipe pipe) { - struct drm_i915_private *dev_priv = to_i915(dev); struct intel_encoder *encoder; lockdep_assert_held(&dev_priv->pps_mutex); - for_each_intel_encoder(dev, encoder) { + for_each_intel_encoder(&dev_priv->drm, encoder) { struct intel_dp *intel_dp; enum port port; @@ -3046,9 +3036,8 @@ static void vlv_steal_power_sequencer(struct drm_device *dev, static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, const struct intel_crtc_state *crtc_state) { + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); lockdep_assert_held(&dev_priv->pps_mutex); @@ -3069,7 +3058,7 @@ static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, * We may be stealing the power * sequencer from another port. */ - vlv_steal_power_sequencer(dev, crtc->pipe); + vlv_steal_power_sequencer(dev_priv, crtc->pipe); intel_dp->active_pipe = crtc->pipe; @@ -3083,8 +3072,8 @@ static void vlv_init_panel_power_sequencer(struct intel_encoder *encoder, pipe_name(intel_dp->pps_pipe), port_name(encoder->port)); /* init power sequencer on this pipe and port */ - intel_dp_init_panel_power_sequencer(dev, intel_dp); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, true); + intel_dp_init_panel_power_sequencer(intel_dp); + intel_dp_init_panel_power_sequencer_registers(intel_dp, true); } static void vlv_pre_enable_dp(struct intel_encoder *encoder, @@ -5062,7 +5051,7 @@ void intel_dp_encoder_reset(struct drm_encoder *encoder) if (intel_dp_is_edp(intel_dp)) { /* Reinit the power sequencer, in case BIOS did something with it. */ - intel_dp_pps_init(encoder->dev, intel_dp); + intel_dp_pps_init(intel_dp); intel_edp_panel_vdd_sanitize(intel_dp); } @@ -5202,13 +5191,13 @@ static void intel_dp_init_panel_power_timestamps(struct intel_dp *intel_dp) } static void -intel_pps_readout_hw_state(struct drm_i915_private *dev_priv, - struct intel_dp *intel_dp, struct edp_power_seq *seq) +intel_pps_readout_hw_state(struct intel_dp *intel_dp, struct edp_power_seq *seq) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp_on, pp_off, pp_div = 0, pp_ctl = 0; struct pps_registers regs; - intel_pps_get_registers(dev_priv, intel_dp, ®s); + intel_pps_get_registers(intel_dp, ®s); /* Workaround: Need to write PP_CONTROL with the unlock key as * the very first thing. */ @@ -5252,13 +5241,12 @@ intel_pps_dump_state(const char *state_name, const struct edp_power_seq *seq) } static void -intel_pps_verify_state(struct drm_i915_private *dev_priv, - struct intel_dp *intel_dp) +intel_pps_verify_state(struct intel_dp *intel_dp) { struct edp_power_seq hw; struct edp_power_seq *sw = &intel_dp->pps_delays; - intel_pps_readout_hw_state(dev_priv, intel_dp, &hw); + intel_pps_readout_hw_state(intel_dp, &hw); if (hw.t1_t3 != sw->t1_t3 || hw.t8 != sw->t8 || hw.t9 != sw->t9 || hw.t10 != sw->t10 || hw.t11_t12 != sw->t11_t12) { @@ -5269,10 +5257,9 @@ intel_pps_verify_state(struct drm_i915_private *dev_priv, } static void -intel_dp_init_panel_power_sequencer(struct drm_device *dev, - struct intel_dp *intel_dp) +intel_dp_init_panel_power_sequencer(struct intel_dp *intel_dp) { - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct edp_power_seq cur, vbt, spec, *final = &intel_dp->pps_delays; @@ -5282,7 +5269,7 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, if (final->t11_t12 != 0) return; - intel_pps_readout_hw_state(dev_priv, intel_dp, &cur); + intel_pps_readout_hw_state(intel_dp, &cur); intel_pps_dump_state("cur", &cur); @@ -5356,11 +5343,10 @@ intel_dp_init_panel_power_sequencer(struct drm_device *dev, } static void -intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, - struct intel_dp *intel_dp, +intel_dp_init_panel_power_sequencer_registers(struct intel_dp *intel_dp, bool force_disable_vdd) { - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp_on, pp_off, pp_div, port_sel = 0; int div = dev_priv->rawclk_freq / 1000; struct pps_registers regs; @@ -5369,7 +5355,7 @@ intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, lockdep_assert_held(&dev_priv->pps_mutex); - intel_pps_get_registers(dev_priv, intel_dp, ®s); + intel_pps_get_registers(intel_dp, ®s); /* * On some VLV machines the BIOS can leave the VDD @@ -5441,16 +5427,15 @@ intel_dp_init_panel_power_sequencer_registers(struct drm_device *dev, I915_READ(regs.pp_div)); } -static void intel_dp_pps_init(struct drm_device *dev, - struct intel_dp *intel_dp) +static void intel_dp_pps_init(struct intel_dp *intel_dp) { - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) { vlv_initial_power_sequencer_setup(intel_dp); } else { - intel_dp_init_panel_power_sequencer(dev, intel_dp); - intel_dp_init_panel_power_sequencer_registers(dev, intel_dp, false); + intel_dp_init_panel_power_sequencer(intel_dp); + intel_dp_init_panel_power_sequencer_registers(intel_dp, false); } } @@ -5864,7 +5849,7 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, pps_lock(intel_dp); intel_dp_init_panel_power_timestamps(intel_dp); - intel_dp_pps_init(dev, intel_dp); + intel_dp_pps_init(intel_dp); intel_edp_panel_vdd_sanitize(intel_dp); pps_unlock(intel_dp); -- GitLab From 2f7734770cfbc337fbd842c64a551cddc04a41c2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 9 Nov 2017 17:27:58 +0200 Subject: [PATCH 0168/1471] drm/i915: Clean up DP code local variables and calling conventions MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Eliminate a ton of pointless 'dev' variables in the DP code, and pass around 'dev_priv' instead of 'dev'. v2: Rebase Acked-by: Jani Nikula Reviewed-by: Dhinakaran Pandiyan Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171109152758.32257-1-ville.syrjala@linux.intel.com --- drivers/gpu/drm/i915/intel_dp.c | 151 ++++++++++++-------------------- 1 file changed, 54 insertions(+), 97 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 37bfe8b272218..efe18269115f4 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -438,10 +438,7 @@ intel_dp_pps_init(struct intel_dp *intel_dp); static void pps_lock(struct intel_dp *intel_dp) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_encoder *encoder = &intel_dig_port->base; - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); /* * See vlv_power_sequencer_reset() why we need @@ -454,10 +451,7 @@ static void pps_lock(struct intel_dp *intel_dp) static void pps_unlock(struct intel_dp *intel_dp) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_encoder *encoder = &intel_dig_port->base; - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); mutex_unlock(&dev_priv->pps_mutex); @@ -467,8 +461,8 @@ static void pps_unlock(struct intel_dp *intel_dp) static void vlv_power_sequencer_kick(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_i915_private *dev_priv = to_i915(intel_dig_port->base.base.dev); enum pipe pipe = intel_dp->pps_pipe; bool pll_enabled, release_cl_override = false; enum dpio_phy phy = DPIO_PHY(pipe); @@ -733,7 +727,6 @@ vlv_initial_power_sequencer_setup(struct intel_dp *intel_dp) void intel_power_sequencer_reset(struct drm_i915_private *dev_priv) { - struct drm_device *dev = &dev_priv->drm; struct intel_encoder *encoder; if (WARN_ON(!IS_VALLEYVIEW(dev_priv) && !IS_CHERRYVIEW(dev_priv) && @@ -750,7 +743,7 @@ void intel_power_sequencer_reset(struct drm_i915_private *dev_priv) * should use them always. */ - for_each_intel_encoder(dev, encoder) { + for_each_intel_encoder(&dev_priv->drm, encoder) { struct intel_dp *intel_dp; if (encoder->type != INTEL_OUTPUT_DP && @@ -832,8 +825,7 @@ static int edp_notify_handler(struct notifier_block *this, unsigned long code, { struct intel_dp *intel_dp = container_of(this, typeof(* intel_dp), edp_notifier); - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); if (!intel_dp_is_edp(intel_dp) || code != SYS_RESTART) return 0; @@ -863,8 +855,7 @@ static int edp_notify_handler(struct notifier_block *this, unsigned long code, static bool edp_have_panel_power(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); lockdep_assert_held(&dev_priv->pps_mutex); @@ -877,8 +868,7 @@ static bool edp_have_panel_power(struct intel_dp *intel_dp) static bool edp_have_panel_vdd(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); lockdep_assert_held(&dev_priv->pps_mutex); @@ -892,8 +882,7 @@ static bool edp_have_panel_vdd(struct intel_dp *intel_dp) static void intel_dp_check_edp(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); if (!intel_dp_is_edp(intel_dp)) return; @@ -909,9 +898,7 @@ intel_dp_check_edp(struct intel_dp *intel_dp) static uint32_t intel_dp_aux_wait_done(struct intel_dp *intel_dp, bool has_aux_irq) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); i915_reg_t ch_ctl = intel_dp->aux_ch_ctl_reg; uint32_t status; bool done; @@ -1478,8 +1465,7 @@ static void intel_dp_set_clock(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); const struct dp_link_dpll *divisor = NULL; int i, count = 0; @@ -1848,8 +1834,7 @@ void intel_dp_set_link_params(struct intel_dp *intel_dp, static void intel_dp_prepare(struct intel_encoder *encoder, const struct intel_crtc_state *pipe_config) { - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); enum port port = encoder->port; struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); @@ -1945,8 +1930,7 @@ static void wait_panel_status(struct intel_dp *intel_dp, u32 mask, u32 value) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); i915_reg_t pp_stat_reg, pp_ctrl_reg; lockdep_assert_held(&dev_priv->pps_mutex); @@ -2022,8 +2006,7 @@ static void edp_wait_backlight_off(struct intel_dp *intel_dp) static u32 ironlake_get_pp_control(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 control; lockdep_assert_held(&dev_priv->pps_mutex); @@ -2044,9 +2027,8 @@ static u32 ironlake_get_pp_control(struct intel_dp *intel_dp) */ static bool edp_panel_vdd_on(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); u32 pp; i915_reg_t pp_stat_reg, pp_ctrl_reg; bool need_to_disable = !intel_dp->want_panel_vdd; @@ -2116,8 +2098,7 @@ void intel_edp_panel_vdd_on(struct intel_dp *intel_dp) static void edp_panel_vdd_off_sync(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); u32 pp; @@ -2203,8 +2184,7 @@ static void edp_panel_vdd_off(struct intel_dp *intel_dp, bool sync) static void edp_panel_on(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp; i915_reg_t pp_ctrl_reg; @@ -2262,8 +2242,7 @@ void intel_edp_panel_on(struct intel_dp *intel_dp) static void edp_panel_off(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp; i915_reg_t pp_ctrl_reg; @@ -2311,9 +2290,7 @@ void intel_edp_panel_off(struct intel_dp *intel_dp) /* Enable backlight in the panel power control. */ static void _intel_edp_backlight_on(struct intel_dp *intel_dp) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp; i915_reg_t pp_ctrl_reg; @@ -2356,8 +2333,7 @@ void intel_edp_backlight_on(const struct intel_crtc_state *crtc_state, /* Disable backlight in the panel power control. */ static void _intel_edp_backlight_off(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u32 pp; i915_reg_t pp_ctrl_reg; @@ -2560,10 +2536,9 @@ void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode) static bool intel_dp_get_hw_state(struct intel_encoder *encoder, enum pipe *pipe) { + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); enum port port = encoder->port; - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); u32 tmp; bool ret; @@ -2612,10 +2587,9 @@ static bool intel_dp_get_hw_state(struct intel_encoder *encoder, static void intel_dp_get_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config) { + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); u32 tmp, flags = 0; - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); enum port port = encoder->port; struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); @@ -2782,9 +2756,8 @@ _intel_dp_set_link_train(struct intel_dp *intel_dp, uint32_t *DP, uint8_t dp_train_pat) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); enum port port = intel_dig_port->base.port; if (dp_train_pat & DP_TRAINING_PATTERN_MASK) @@ -2868,8 +2841,7 @@ _intel_dp_set_link_train(struct intel_dp *intel_dp, static void intel_dp_enable_port(struct intel_dp *intel_dp, const struct intel_crtc_state *old_crtc_state) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); /* enable with pattern 1 (as per spec) */ @@ -2893,9 +2865,8 @@ static void intel_enable_dp(struct intel_encoder *encoder, const struct intel_crtc_state *pipe_config, const struct drm_connector_state *conn_state) { + struct drm_i915_private *dev_priv = to_i915(encoder->base.dev); struct intel_dp *intel_dp = enc_to_intel_dp(&encoder->base); - struct drm_device *dev = encoder->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); struct intel_crtc *crtc = to_intel_crtc(pipe_config->base.crtc); uint32_t dp_reg = I915_READ(intel_dp->output_reg); enum pipe pipe = crtc->pipe; @@ -3519,10 +3490,9 @@ gen7_edp_signal_levels(uint8_t train_set) void intel_dp_set_signal_levels(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); enum port port = intel_dig_port->base.port; - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); uint32_t signal_levels, mask = 0; uint8_t train_set = intel_dp->train_set[0]; @@ -3577,9 +3547,8 @@ intel_dp_program_link_training_pattern(struct intel_dp *intel_dp, void intel_dp_set_idle_link_train(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); enum port port = intel_dig_port->base.port; uint32_t val; @@ -4303,11 +4272,11 @@ intel_dp_retrain_link(struct intel_dp *intel_dp) static void intel_dp_check_link_status(struct intel_dp *intel_dp) { + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_encoder *intel_encoder = &dp_to_dig_port(intel_dp)->base; - struct drm_device *dev = intel_dp_to_dev(intel_dp); u8 link_status[DP_LINK_STATUS_SIZE]; - WARN_ON(!drm_modeset_is_locked(&dev->mode_config.connection_mutex)); + WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex)); if (!intel_dp_get_link_status(intel_dp, link_status)) { DRM_ERROR("Failed to get link status\n"); @@ -4353,8 +4322,7 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) static bool intel_dp_short_pulse(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct intel_encoder *intel_encoder = &dp_to_dig_port(intel_dp)->base; + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); u8 sink_irq_vector = 0; u8 old_sink_count = intel_dp->sink_count; bool ret; @@ -4393,13 +4361,13 @@ intel_dp_short_pulse(struct intel_dp *intel_dp) DRM_DEBUG_DRIVER("CP or sink specific irq unhandled\n"); } - drm_modeset_lock(&dev->mode_config.connection_mutex, NULL); + drm_modeset_lock(&dev_priv->drm.mode_config.connection_mutex, NULL); intel_dp_check_link_status(intel_dp); - drm_modeset_unlock(&dev->mode_config.connection_mutex); + drm_modeset_unlock(&dev_priv->drm.mode_config.connection_mutex); if (intel_dp->compliance.test_type == DP_TEST_LINK_TRAINING) { DRM_DEBUG_KMS("Link Training Compliance Test requested\n"); /* Send a Hotplug Uevent to userspace to start modeset */ - drm_kms_helper_hotplug_event(intel_encoder->base.dev); + drm_kms_helper_hotplug_event(&dev_priv->drm); } return true; @@ -4463,8 +4431,7 @@ intel_dp_detect_dpcd(struct intel_dp *intel_dp) static enum drm_connector_status edp_detect(struct intel_dp *intel_dp) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); enum drm_connector_status status; status = intel_panel_detect(dev_priv); @@ -4720,22 +4687,21 @@ intel_dp_unset_edid(struct intel_dp *intel_dp) } static int -intel_dp_long_pulse(struct intel_connector *intel_connector) +intel_dp_long_pulse(struct intel_connector *connector) { - struct drm_connector *connector = &intel_connector->base; - struct intel_dp *intel_dp = intel_attached_dp(connector); - struct drm_device *dev = connector->dev; + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); + struct intel_dp *intel_dp = intel_attached_dp(&connector->base); enum drm_connector_status status; u8 sink_irq_vector = 0; - WARN_ON(!drm_modeset_is_locked(&connector->dev->mode_config.connection_mutex)); + WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex)); - intel_display_power_get(to_i915(dev), intel_dp->aux_power_domain); + intel_display_power_get(dev_priv, intel_dp->aux_power_domain); /* Can't disconnect eDP, but you can close the lid... */ if (intel_dp_is_edp(intel_dp)) status = edp_detect(intel_dp); - else if (intel_digital_port_connected(to_i915(dev), + else if (intel_digital_port_connected(dev_priv, dp_to_dig_port(intel_dp))) status = intel_dp_detect_dpcd(intel_dp); else @@ -4806,7 +4772,7 @@ intel_dp_long_pulse(struct intel_connector *intel_connector) intel_dp->aux.i2c_defer_count = 0; intel_dp_set_edid(intel_dp); - if (intel_dp_is_edp(intel_dp) || intel_connector->detect_edid) + if (intel_dp_is_edp(intel_dp) || connector->detect_edid) status = connector_status_connected; intel_dp->detect_done = true; @@ -4829,7 +4795,7 @@ intel_dp_long_pulse(struct intel_connector *intel_connector) if (status != connector_status_connected && !intel_dp->is_mst) intel_dp_unset_edid(intel_dp); - intel_display_power_put(to_i915(dev), intel_dp->aux_power_domain); + intel_display_power_put(dev_priv, intel_dp->aux_power_domain); return status; } @@ -4996,9 +4962,7 @@ void intel_dp_encoder_suspend(struct intel_encoder *intel_encoder) static void intel_edp_panel_vdd_sanitize(struct intel_dp *intel_dp) { - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); lockdep_assert_held(&dev_priv->pps_mutex); @@ -5086,8 +5050,7 @@ enum irqreturn intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd) { struct intel_dp *intel_dp = &intel_dig_port->dp; - struct drm_device *dev = intel_dig_port->base.base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); enum irqreturn ret = IRQ_NONE; if (long_hpd && intel_dig_port->base.type == INTEL_OUTPUT_EDP) { @@ -5546,8 +5509,7 @@ static void intel_dp_set_drrs_state(struct drm_i915_private *dev_priv, void intel_edp_drrs_enable(struct intel_dp *intel_dp, const struct intel_crtc_state *crtc_state) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); if (!crtc_state->has_drrs) { DRM_DEBUG_KMS("Panel doesn't support DRRS\n"); @@ -5582,8 +5544,7 @@ void intel_edp_drrs_enable(struct intel_dp *intel_dp, void intel_edp_drrs_disable(struct intel_dp *intel_dp, const struct intel_crtc_state *old_crtc_state) { - struct drm_device *dev = intel_dp_to_dev(intel_dp); - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); if (!old_crtc_state->has_drrs) return; @@ -5766,7 +5727,7 @@ void intel_edp_drrs_flush(struct drm_i915_private *dev_priv, /** * intel_dp_drrs_init - Init basic DRRS work and mutex. - * @intel_connector: eDP connector + * @connector: eDP connector * @fixed_mode: preferred mode of panel * * This function is called only once at driver load to initialize basic @@ -5778,12 +5739,10 @@ void intel_edp_drrs_flush(struct drm_i915_private *dev_priv, * from VBT setting). */ static struct drm_display_mode * -intel_dp_drrs_init(struct intel_connector *intel_connector, - struct drm_display_mode *fixed_mode) +intel_dp_drrs_init(struct intel_connector *connector, + struct drm_display_mode *fixed_mode) { - struct drm_connector *connector = &intel_connector->base; - struct drm_device *dev = connector->dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); struct drm_display_mode *downclock_mode = NULL; INIT_DELAYED_WORK(&dev_priv->drrs.work, intel_edp_drrs_downclock_work); @@ -5799,8 +5758,8 @@ intel_dp_drrs_init(struct intel_connector *intel_connector, return NULL; } - downclock_mode = intel_find_panel_downclock - (dev_priv, fixed_mode, connector); + downclock_mode = intel_find_panel_downclock(dev_priv, fixed_mode, + &connector->base); if (!downclock_mode) { DRM_DEBUG_KMS("Downclock mode is not found. DRRS not supported\n"); @@ -5817,11 +5776,9 @@ intel_dp_drrs_init(struct intel_connector *intel_connector, static bool intel_edp_init_connector(struct intel_dp *intel_dp, struct intel_connector *intel_connector) { - struct drm_connector *connector = &intel_connector->base; - struct intel_digital_port *intel_dig_port = dp_to_dig_port(intel_dp); - struct intel_encoder *intel_encoder = &intel_dig_port->base; - struct drm_device *dev = intel_encoder->base.dev; + struct drm_device *dev = intel_dp_to_dev(intel_dp); struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_connector *connector = &intel_connector->base; struct drm_display_mode *fixed_mode = NULL; struct drm_display_mode *alt_fixed_mode = NULL; struct drm_display_mode *downclock_mode = NULL; @@ -5839,7 +5796,7 @@ static bool intel_edp_init_connector(struct intel_dp *intel_dp, * eDP and LVDS bail out early in this case to prevent interfering * with an already powered-on LVDS power sequencer. */ - if (intel_get_lvds_encoder(dev)) { + if (intel_get_lvds_encoder(&dev_priv->drm)) { WARN_ON(!(HAS_PCH_IBX(dev_priv) || HAS_PCH_CPT(dev_priv))); DRM_INFO("LVDS was detected, not registering eDP\n"); -- GitLab From 32e823c63e90f7535ea1cc5311d25c0233e1456d Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 20 Sep 2017 15:59:34 -0700 Subject: [PATCH 0169/1471] drm/vc4: Reject HDMI modes with too high of clocks. Peter Robinson reported issues on Fedora with 4k monitors not having their modes filtered down to 1920x1080 on Raspberry Pi. v2: Fix vc5 typo in place of vc4. Cc: Peter Robinson Signed-off-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/20170920225935.14566-1-eric@anholt.net Acked-by: Daniel Vetter (v1) --- drivers/gpu/drm/vc4/vc4_hdmi.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/drivers/gpu/drm/vc4/vc4_hdmi.c b/drivers/gpu/drm/vc4/vc4_hdmi.c index 9a9a6b4acccfd..ab3bcbe0bb93f 100644 --- a/drivers/gpu/drm/vc4/vc4_hdmi.c +++ b/drivers/gpu/drm/vc4/vc4_hdmi.c @@ -693,7 +693,22 @@ static void vc4_hdmi_encoder_enable(struct drm_encoder *encoder) } } +static enum drm_mode_status +vc4_hdmi_encoder_mode_valid(struct drm_encoder *crtc, + const struct drm_display_mode *mode) +{ + /* HSM clock must be 108% of the pixel clock. Additionally, + * the AXI clock needs to be at least 25% of pixel clock, but + * HSM ends up being the limiting factor. + */ + if (mode->clock > HSM_CLOCK_FREQ / (1000 * 108 / 100)) + return MODE_CLOCK_HIGH; + + return MODE_OK; +} + static const struct drm_encoder_helper_funcs vc4_hdmi_encoder_helper_funcs = { + .mode_valid = vc4_hdmi_encoder_mode_valid, .disable = vc4_hdmi_encoder_disable, .enable = vc4_hdmi_encoder_enable, }; -- GitLab From 19d814cc0765b409e2b90ed7a5514a15ce6ecf7b Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 2 Nov 2017 22:03:33 +0200 Subject: [PATCH 0170/1471] drm/syncobj: Mark up the fence as an RCU protected pointer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We take advantage of that syncobj->fence is an RCU-protected pointer, and so sparse complains that it is lacking annotation. Cc: Dave Airlie Cc: Jason Ekstrand Cc: linaro-mm-sig@lists.linaro.org Cc: linux-media@vger.kernel.org Cc: Alex Deucher Cc: Christian König Cc: Sumit Semwal Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171102200336.23347-2-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Acked-by: Christian König Signed-off-by: Ville Syrjälä --- include/drm/drm_syncobj.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/drm/drm_syncobj.h b/include/drm/drm_syncobj.h index c00fee5398224..4556606732599 100644 --- a/include/drm/drm_syncobj.h +++ b/include/drm/drm_syncobj.h @@ -49,7 +49,7 @@ struct drm_syncobj { * This field should not be used directly. Use drm_syncobj_fence_get * and drm_syncobj_replace_fence instead. */ - struct dma_fence *fence; + struct dma_fence __rcu *fence; /** * @cb_list: * List of callbacks to call when the fence gets replaced -- GitLab From 5f72db59160c697d1dec3f6074dba72bd834c695 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 2 Nov 2017 22:03:34 +0200 Subject: [PATCH 0171/1471] dma-buf/fence: Sparse wants __rcu on the object itself MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In order to silence sparse in dma_fence_get_rcu_safe(), we need to mark the incoming fence object as being RCU protected and not the pointer to the object. Cc: Dave Airlie Cc: Jason Ekstrand Cc: linaro-mm-sig@lists.linaro.org Cc: linux-media@vger.kernel.org Cc: Alex Deucher Cc: Christian König Cc: Sumit Semwal Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171102200336.23347-3-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Acked-by: Christian König Acked-by: Sumit Semwal [vsyrjala: s/silent/silence/ in commit message] Signed-off-by: Ville Syrjälä --- include/linux/dma-fence.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h index efdabbb64e3c7..4c008170fe653 100644 --- a/include/linux/dma-fence.h +++ b/include/linux/dma-fence.h @@ -242,7 +242,7 @@ static inline struct dma_fence *dma_fence_get_rcu(struct dma_fence *fence) * The caller is required to hold the RCU read lock. */ static inline struct dma_fence * -dma_fence_get_rcu_safe(struct dma_fence * __rcu *fencep) +dma_fence_get_rcu_safe(struct dma_fence __rcu **fencep) { do { struct dma_fence *fence; -- GitLab From 563eaf53317f2511f78217a1f8d785b8c8ed1cad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 2 Nov 2017 22:03:35 +0200 Subject: [PATCH 0172/1471] drm/syncobj: Use proper methods for accessing rcu protected pointers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use rcu_dereference_protected() and rcu_assign_pointer() for accessing the rcu protected syncobj->fence pointer. This eliminates several sparse warnings. Cc: Dave Airlie Cc: Jason Ekstrand Cc: linaro-mm-sig@lists.linaro.org Cc: linux-media@vger.kernel.org Cc: Alex Deucher Cc: Christian König Cc: Sumit Semwal Cc: Chris Wilson Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171102200336.23347-4-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Acked-by: Christian König --- drivers/gpu/drm/drm_syncobj.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c index 7081ae601d6f6..7943c215a03ce 100644 --- a/drivers/gpu/drm/drm_syncobj.c +++ b/drivers/gpu/drm/drm_syncobj.c @@ -106,7 +106,8 @@ static int drm_syncobj_fence_get_or_add_callback(struct drm_syncobj *syncobj, * callback when a fence has already been set. */ if (syncobj->fence) { - *fence = dma_fence_get(syncobj->fence); + *fence = dma_fence_get(rcu_dereference_protected(syncobj->fence, + lockdep_is_held(&syncobj->lock))); ret = 1; } else { *fence = NULL; @@ -168,8 +169,9 @@ void drm_syncobj_replace_fence(struct drm_syncobj *syncobj, spin_lock(&syncobj->lock); - old_fence = syncobj->fence; - syncobj->fence = fence; + old_fence = rcu_dereference_protected(syncobj->fence, + lockdep_is_held(&syncobj->lock)); + rcu_assign_pointer(syncobj->fence, fence); if (fence != old_fence) { list_for_each_entry_safe(cur, tmp, &syncobj->cb_list, node) { @@ -613,7 +615,8 @@ static void syncobj_wait_syncobj_func(struct drm_syncobj *syncobj, container_of(cb, struct syncobj_wait_entry, syncobj_cb); /* This happens inside the syncobj lock */ - wait->fence = dma_fence_get(syncobj->fence); + wait->fence = dma_fence_get(rcu_dereference_protected(syncobj->fence, + lockdep_is_held(&syncobj->lock))); wake_up_process(wait->task); } -- GitLab From ad46d7b89371c5920abdcad1ed04fbae68c9c8a9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 2 Nov 2017 22:03:36 +0200 Subject: [PATCH 0173/1471] dma-buf: Use rcu_assign_pointer() to set rcu protected pointers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use rcu_assign_pointer() when setting an rcu protected pointer. This gets rid of another sparse warning. Cc: Dave Airlie Cc: Jason Ekstrand Cc: linaro-mm-sig@lists.linaro.org Cc: linux-media@vger.kernel.org Cc: Alex Deucher Cc: Christian König Cc: Sumit Semwal Cc: Chris Wilson Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171102200336.23347-5-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Reviewed-by: Christian König . Acked-by: Sumit Semwal --- drivers/dma-buf/reservation.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c index b44d9d7db347b..d90333e0b6d5f 100644 --- a/drivers/dma-buf/reservation.c +++ b/drivers/dma-buf/reservation.c @@ -318,7 +318,7 @@ int reservation_object_copy_fences(struct reservation_object *dst, continue; } - dst_list->shared[dst_list->shared_count++] = fence; + rcu_assign_pointer(dst_list->shared[dst_list->shared_count++], fence); } } else { dst_list = NULL; -- GitLab From bccd3b831185e75c4138bc3fd5201f3214dfeb3d Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Nov 2017 14:30:19 +0000 Subject: [PATCH 0174/1471] drm/i915: Use trace_printk to provide a death rattle for GEM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Trying to enable printk debugging for GEM is fraught with the issue of spam; interactions with HW are very frequent and often boring. However, one instance where they are not so boring is just before a BUG; here ftrace provides a facility to dump its ringbuffer on an oops. So for CI let's enable trace_printk() to capture the last exchanges with HW as a death rattle. For example, [ 79.234110] ------------[ cut here ]------------ [ 79.234137] kernel BUG at drivers/gpu/drm/i915/intel_lrc.c:907! [ 79.234145] invalid opcode: 0000 [#1] SMP [ 79.234153] Dumping ftrace buffer: [ 79.234158] --------------------------------- ... [ 79.314044] gem_conc-1059 1..s1 79203443us : intel_lrc_irq_handler: bcs0 out[0]: ctx=5.2, seqno=145 [ 79.314089] gem_conc-1059 1..s. 79220800us : intel_lrc_irq_handler: bcs0 csb[1/1]: status=0x00000018:0x00000005 [ 79.314133] gem_conc-1059 1..s. 79220803us : intel_lrc_irq_handler: bcs0 out[0]: ctx=5.1, seqno=145 [ 79.314177] gem_conc-1062 2..s1 79230458us : intel_lrc_irq_handler: bcs0 in[0]: ctx=8.1, seqno=146 [ 79.314220] gem_conc-1062 2..s1 79230515us : intel_lrc_irq_handler: bcs0 in[0]: ctx=8.2, seqno=147 [ 79.314265] gem_conc-1059 1..s1 79230951us : intel_lrc_irq_handler: bcs0 csb[2/3]: status=0x00000012:0x00000008 [ 79.314309] gem_conc-1059 1..s1 79230954us : intel_lrc_irq_handler: bcs0 out[0]: ctx=8.2, seqno=147 [ 79.314353] gem_conc-1059 1..s1 79230954us : intel_lrc_irq_handler: bcs0 csb[3/3]: status=0x00008002:0x00000008 [ 79.314396] gem_conc-1059 1..s1 79230955us : intel_lrc_irq_handler: bcs0 out[0]: ctx=8.1, seqno=147 [ 79.314402] --------------------------------- v2: Tweak the formatting to be more consistent between in/out. v3: do {} while (0) stub macro protection Suggested-by: Michał Winiarski Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Joonas Lahtinen Cc: Tvrtko Ursulin Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171109143019.16568-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/Kconfig.debug | 14 ++++++++++++++ drivers/gpu/drm/i915/i915_gem.h | 6 ++++++ drivers/gpu/drm/i915/intel_lrc.c | 17 +++++++++++++++++ 3 files changed, 37 insertions(+) diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug index 19c77c6feb245..9e53edbc713b9 100644 --- a/drivers/gpu/drm/i915/Kconfig.debug +++ b/drivers/gpu/drm/i915/Kconfig.debug @@ -28,6 +28,7 @@ config DRM_I915_DEBUG select SW_SYNC # signaling validation framework (igt/syncobj*) select DRM_I915_SW_FENCE_DEBUG_OBJECTS select DRM_I915_SELFTEST + select DRM_I915_TRACE_GEM default n help Choose this option to turn on extra driver debugging that may affect @@ -49,6 +50,19 @@ config DRM_I915_DEBUG_GEM If in doubt, say "N". +config DRM_I915_TRACE_GEM + bool "Insert extra ftrace output from the GEM internals" + select TRACING + default n + help + Enable additional and verbose debugging output that will spam + ordinary tests, but may be vital for post-mortem debugging when + used with /proc/sys/kernel/ftrace_dump_on_oops + + Recommended for driver developers only. + + If in doubt, say "N". + config DRM_I915_SW_FENCE_DEBUG_OBJECTS bool "Enable additional driver debugging for fence objects" depends on DRM_I915 diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h index ee54597465b60..ff42b5f0e9812 100644 --- a/drivers/gpu/drm/i915/i915_gem.h +++ b/drivers/gpu/drm/i915/i915_gem.h @@ -44,6 +44,12 @@ #define GEM_DEBUG_BUG_ON(expr) #endif +#if IS_ENABLED(CONFIG_DRM_I915_TRACE_GEM) +#define GEM_TRACE(...) trace_printk(__VA_ARGS__) +#else +#define GEM_TRACE(...) do { } while (0) +#endif + #define I915_NUM_ENGINES 5 #endif /* __I915_GEM_H__ */ diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 6840ec8db0376..020ca7c9c75f9 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -466,6 +466,11 @@ static void execlists_submit_ports(struct intel_engine_cs *engine) port_set(&port[n], port_pack(rq, count)); desc = execlists_update_context(rq); GEM_DEBUG_EXEC(port[n].context_id = upper_32_bits(desc)); + + GEM_TRACE("%s in[%d]: ctx=%d.%d, seqno=%x\n", + engine->name, n, + rq->ctx->hw_id, count, + rq->global_seqno); } else { GEM_BUG_ON(!n); desc = 0; @@ -520,6 +525,7 @@ static void inject_preempt_context(struct intel_engine_cs *engine) ce->ring->tail &= (ce->ring->size - 1); ce->lrc_reg_state[CTX_RING_TAIL+1] = ce->ring->tail; + GEM_TRACE("\n"); for (n = execlists_num_ports(&engine->execlists); --n; ) elsp_write(0, elsp); @@ -832,6 +838,10 @@ static void intel_lrc_irq_handler(unsigned long data) head = execlists->csb_head; tail = READ_ONCE(buf[write_idx]); } + GEM_TRACE("%s cs-irq head=%d [%d], tail=%d [%d]\n", + engine->name, + head, GEN8_CSB_READ_PTR(readl(dev_priv->regs + i915_mmio_reg_offset(RING_CONTEXT_STATUS_PTR(engine)))), + tail, GEN8_CSB_WRITE_PTR(readl(dev_priv->regs + i915_mmio_reg_offset(RING_CONTEXT_STATUS_PTR(engine))))); while (head != tail) { struct drm_i915_gem_request *rq; @@ -859,6 +869,9 @@ static void intel_lrc_irq_handler(unsigned long data) */ status = READ_ONCE(buf[2 * head]); /* maybe mmio! */ + GEM_TRACE("%s csb[%dd]: status=0x%08x:0x%08x\n", + engine->name, head, + status, buf[2*head + 1]); if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK)) continue; @@ -886,6 +899,10 @@ static void intel_lrc_irq_handler(unsigned long data) GEM_DEBUG_BUG_ON(buf[2 * head + 1] != port->context_id); rq = port_unpack(port, &count); + GEM_TRACE("%s out[0]: ctx=%d.%d, seqno=%x\n", + engine->name, + rq->ctx->hw_id, count, + rq->global_seqno); GEM_BUG_ON(count == 0); if (--count == 0) { GEM_BUG_ON(status & GEN8_CTX_STATUS_PREEMPTED); -- GitLab From 15e4cda950ea660837f6733426691c8e467010a6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Nov 2017 21:34:49 +0000 Subject: [PATCH 0175/1471] drm/i915: Mark vm_free_page() as a potential sleeper agent vm_free_page() may call down into set_pages_array_wb() (which itself sleeps, on x86 at least) but only if on !llc and the caches overflow. Since this is unlikely, we only rarely trigger the error in practice, and so to make CI detection of this sleeping bug possible we want to mark the common vm_free_page() as a potential sleep. Suggested-by: Tvrtko Ursulin References: https://bugs.freedesktop.org/show_bug.cgi?id=103638 Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Matthew Auld Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171109213450.13875-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem_gtt.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 2847a6b41c168..1e40eeb31f9da 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -454,6 +454,14 @@ static void vm_free_pages_release(struct i915_address_space *vm, static void vm_free_page(struct i915_address_space *vm, struct page *page) { + /* + * On !llc, we need to change the pages back to WB. We only do so + * in bulk, so we rarely need to change the page attributes here, + * but doing so requires a stop_machine() from deep inside arch/x86/mm. + * To make detection of the possible sleep more likely, use an + * unconditional might_sleep() for everybody. + */ + might_sleep(); if (!pagevec_add(&vm->free_pages, page)) vm_free_pages_release(vm, false); } -- GitLab From 7f017b19fbce6200a993305425bf930f38fcd9e6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Nov 2017 21:34:50 +0000 Subject: [PATCH 0176/1471] drm/i915: Mark up i915_vma_unbind() as a potential sleeper Whenever we want to unbind a vma, we must wait on all GPU activity to complete first. (This is what gives us the ability to do fine grained eviction and purging by only having to wait on the VMA that we need to unbind to proceed; though if pushed we can make it a rule that we are only allowed to unbind already idle VMA and move the burden of the work and organising the sleep onto the caller.) Currently, we might only sleep if the vma is still active on the GPU, but in principle i915_vma_unbind() always implies a sleep, so mark it up with a might_sleep(). Suggested-by: Tvrtko Ursulin References: https://bugs.freedesktop.org/show_bug.cgi?id=103638 Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Matthew Auld Cc: Daniel Vetter Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171109213450.13875-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_vma.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index b7ac84db1d2e6..bf6d8d1eaabeb 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -743,6 +743,7 @@ int i915_vma_unbind(struct i915_vma *vma) /* First wait upon any activity as retiring the request may * have side-effects such as unpinning or even unbinding this vma. */ + might_sleep(); active = i915_vma_get_active(vma); if (active) { int idx; -- GitLab From 25be4a154504d60293e345535eeab5d6a022b4fa Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Thu, 9 Nov 2017 15:18:04 -0800 Subject: [PATCH 0177/1471] drm/i915: Update DRIVER_DATE to 20171109 Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/i915/i915_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 19b679ca8211e..7a8f5d7638f01 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -80,8 +80,8 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20171023" -#define DRIVER_TIMESTAMP 1508748913 +#define DRIVER_DATE "20171109" +#define DRIVER_TIMESTAMP 1510269484 /* Use I915_STATE_WARN(x) and I915_STATE_WARN_ON() (rather than WARN() and * WARN_ON()) for hw state sanity checks to check for unexpected conditions -- GitLab From e25dcf6ed8f5406b12689543698df98bd4c48cc7 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Thu, 9 Nov 2017 15:30:06 -0800 Subject: [PATCH 0178/1471] drm/i915: Update DRIVER_DATE to 20171109 Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/i915/i915_drv.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 7a8f5d7638f01..c94f34f5cee3c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -81,7 +81,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" #define DRIVER_DATE "20171109" -#define DRIVER_TIMESTAMP 1510269484 +#define DRIVER_TIMESTAMP 1510270206 /* Use I915_STATE_WARN(x) and I915_STATE_WARN_ON() (rather than WARN() and * WARN_ON()) for hw state sanity checks to check for unexpected conditions -- GitLab From b1a1e5d2cebf4630d971628d20b3034dfe9c2c40 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 10:11:10 +0000 Subject: [PATCH 0179/1471] drm/i915/selftests: Reduce the volume of the timeout message Originally it was anticipated that timeouts would be a rare event, and so merit a warning that the test was incomplete. However, for igt we keep the timeout low, and hitting the timeout is intentional. It no longer necessitates a warning, but to be expected. Signed-off-by: Chris Wilson Cc: Matthew Auld Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171110101110.12042-1-chris@chris-wilson.co.uk Reviwed-by: Matthew Auld Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_selftest.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_selftest.h b/drivers/gpu/drm/i915/i915_selftest.h index 78e1a1b168ff6..9766e806dce63 100644 --- a/drivers/gpu/drm/i915/i915_selftest.h +++ b/drivers/gpu/drm/i915/i915_selftest.h @@ -99,6 +99,6 @@ __printf(2, 3) bool __igt_timeout(unsigned long timeout, const char *fmt, ...); #define igt_timeout(t, fmt, ...) \ - __igt_timeout((t), KERN_WARNING pr_fmt(fmt), ##__VA_ARGS__) + __igt_timeout((t), KERN_NOTICE pr_fmt(fmt), ##__VA_ARGS__) #endif /* !__I915_SELFTEST_H__ */ -- GitLab From 7bac534e027fee318ded13512d12cfaf7d3a937f Mon Sep 17 00:00:00 2001 From: Lepton Wu Date: Wed, 8 Nov 2017 10:42:09 -0800 Subject: [PATCH 0180/1471] drm/virtio: add create_handle support. Add create_handle support to virtio fb. Without this, screenshot tool in chromium OS can't work. Signed-off-by: Lepton Wu Link: http://patchwork.freedesktop.org/patch/msgid/20171108184209.46211-1-ytht.net@gmail.com Signed-off-by: Gerd Hoffmann --- drivers/gpu/drm/virtio/virtgpu_display.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/drivers/gpu/drm/virtio/virtgpu_display.c b/drivers/gpu/drm/virtio/virtgpu_display.c index 41b0930f79681..19114a3c5ee4f 100644 --- a/drivers/gpu/drm/virtio/virtgpu_display.c +++ b/drivers/gpu/drm/virtio/virtgpu_display.c @@ -71,7 +71,19 @@ virtio_gpu_framebuffer_surface_dirty(struct drm_framebuffer *fb, return virtio_gpu_surface_dirty(virtio_gpu_fb, clips, num_clips); } +static int +virtio_gpu_framebuffer_create_handle(struct drm_framebuffer *fb, + struct drm_file *file_priv, + unsigned int *handle) +{ + struct virtio_gpu_framebuffer *virtio_gpu_fb = + to_virtio_gpu_framebuffer(fb); + + return drm_gem_handle_create(file_priv, virtio_gpu_fb->obj, handle); +} + static const struct drm_framebuffer_funcs virtio_gpu_fb_funcs = { + .create_handle = virtio_gpu_framebuffer_create_handle, .destroy = virtio_gpu_user_framebuffer_destroy, .dirty = virtio_gpu_framebuffer_surface_dirty, }; -- GitLab From adbf6eb3cd703398934a274ccb689344cfceb0cb Mon Sep 17 00:00:00 2001 From: Lepton Wu Date: Wed, 8 Nov 2017 10:55:37 -0800 Subject: [PATCH 0181/1471] drm/cirrus: add create_handle support. Add create_handle support to cirrus fb. Without this, screenshot tool in chromium OS can't work. Signed-off-by: Lepton Wu Link: http://patchwork.freedesktop.org/patch/msgid/20171108185537.56167-1-ytht.net@gmail.com Signed-off-by: Gerd Hoffmann --- drivers/gpu/drm/cirrus/cirrus_main.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/drivers/gpu/drm/cirrus/cirrus_main.c b/drivers/gpu/drm/cirrus/cirrus_main.c index b5f5285439562..26df1e8cd490d 100644 --- a/drivers/gpu/drm/cirrus/cirrus_main.c +++ b/drivers/gpu/drm/cirrus/cirrus_main.c @@ -13,6 +13,14 @@ #include "cirrus_drv.h" +static int cirrus_create_handle(struct drm_framebuffer *fb, + struct drm_file* file_priv, + unsigned int* handle) +{ + struct cirrus_framebuffer *cirrus_fb = to_cirrus_framebuffer(fb); + + return drm_gem_handle_create(file_priv, cirrus_fb->obj, handle); +} static void cirrus_user_framebuffer_destroy(struct drm_framebuffer *fb) { @@ -24,6 +32,7 @@ static void cirrus_user_framebuffer_destroy(struct drm_framebuffer *fb) } static const struct drm_framebuffer_funcs cirrus_fb_funcs = { + .create_handle = cirrus_create_handle, .destroy = cirrus_user_framebuffer_destroy, }; -- GitLab From 4013e44ffaf7eaed60fa39750a2944b774b9ea3f Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Tue, 12 Sep 2017 18:34:22 +0100 Subject: [PATCH 0182/1471] drm: qxl: remove pr_info message, stops log spamming Simply mmap'ing /dev/dri/card0 repeatedly will spam the kernel log with qxl_mmap information messages. The following example code illustrates this: int main(void) { int fd = open("/dev/dri/card0", O_RDONLY); if (fd == -1) err(1, "open failed"); for (;;) { void *m = mmap(NULL, 4096, PROT_READ, MAP_SHARED, fd, 0); if (m != MAP_FAILED) munmap(m, 4096); } } Stop the spamming by removing the pr_info message. Since the mmap'ing returns -EINVAL there is no need for the pr_info message. Signed-off-by: Colin Ian King Link: http://patchwork.freedesktop.org/patch/msgid/20170912173422.14062-1-colin.king@canonical.com Signed-off-by: Gerd Hoffmann --- drivers/gpu/drm/qxl/qxl_ttm.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c index 7ecf8a4b9fe6a..0f50c0b25a585 100644 --- a/drivers/gpu/drm/qxl/qxl_ttm.c +++ b/drivers/gpu/drm/qxl/qxl_ttm.c @@ -123,11 +123,8 @@ int qxl_mmap(struct file *filp, struct vm_area_struct *vma) struct qxl_device *qdev; int r; - if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET)) { - pr_info("%s: vma->vm_pgoff (%ld) < DRM_FILE_PAGE_OFFSET\n", - __func__, vma->vm_pgoff); + if (unlikely(vma->vm_pgoff < DRM_FILE_PAGE_OFFSET)) return -EINVAL; - } file_priv = filp->private_data; qdev = file_priv->minor->dev->dev_private; -- GitLab From a86f25518e3ab7aec2d44b195835fd8d22d0eada Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Miros=C5=82aw?= Date: Thu, 31 Aug 2017 20:06:12 +0200 Subject: [PATCH 0183/1471] virtgpu: remove redundant task_comm copying MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Remove redundant task name copying because: 1. task->comm is already NUL-terminated 2. virtio_gpu_context_create() trims passed dbgname anyway Signed-off-by: Michał Mirosław Link: http://patchwork.freedesktop.org/patch/msgid/588612bbc672e8b598e10f79fcbc041122e844ba.1504202471.git.mirq-linux@rere.qmqm.pl Signed-off-by: Gerd Hoffmann --- drivers/gpu/drm/virtio/virtgpu_kms.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/virtio/virtgpu_kms.c b/drivers/gpu/drm/virtio/virtgpu_kms.c index 6400506a06b07..65060c08522d7 100644 --- a/drivers/gpu/drm/virtio/virtgpu_kms.c +++ b/drivers/gpu/drm/virtio/virtgpu_kms.c @@ -272,20 +272,18 @@ int virtio_gpu_driver_open(struct drm_device *dev, struct drm_file *file) struct virtio_gpu_device *vgdev = dev->dev_private; struct virtio_gpu_fpriv *vfpriv; uint32_t id; - char dbgname[64], tmpname[TASK_COMM_LEN]; + char dbgname[TASK_COMM_LEN]; /* can't create contexts without 3d renderer */ if (!vgdev->has_virgl_3d) return 0; - get_task_comm(tmpname, current); - snprintf(dbgname, sizeof(dbgname), "%s", tmpname); - dbgname[63] = 0; /* allocate a virt GPU context for this opener */ vfpriv = kzalloc(sizeof(*vfpriv), GFP_KERNEL); if (!vfpriv) return -ENOMEM; + get_task_comm(dbgname, current); virtio_gpu_context_create(vgdev, strlen(dbgname), dbgname, &id); vfpriv->ctx_id = id; -- GitLab From c16c4ba7250b681e103126c54481a8b17c073758 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 7 Nov 2017 10:20:03 +0000 Subject: [PATCH 0184/1471] drm/i915: Move irqs enabled assertion deeper for mock breadcrumbs In order to allow the mock breadcrumbs tests to run without device irqs being enabled, move the intel_irqs_enabled() assert deeper to just before we commit to enabling the HW irq. v2: Add a FIXME explaining that placing the assertion so deep is not ideal, but a compromise for mock breadcrumbs. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171107102003.1802-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_breadcrumbs.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c index 4de054f8c1baf..4c4fbf5f20f9b 100644 --- a/drivers/gpu/drm/i915/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c @@ -145,6 +145,14 @@ static void intel_breadcrumbs_fake_irq(struct timer_list *t) static void irq_enable(struct intel_engine_cs *engine) { + /* + * FIXME: Ideally we want this on the API boundary, but for the + * sake of testing with mock breadcrumbs (no HW so unable to + * enable irqs) we place it deep within the bowels, at the point + * of no return. + */ + GEM_BUG_ON(!intel_irqs_enabled(engine->i915)); + /* Enabling the IRQ may miss the generation of the interrupt, but * we still need to force the barrier before reading the seqno, * just in case. @@ -266,8 +274,6 @@ static bool __intel_breadcrumbs_enable_irq(struct intel_breadcrumbs *b) struct drm_i915_private *i915 = engine->i915; bool enabled; - GEM_BUG_ON(!intel_irqs_enabled(i915)); - lockdep_assert_held(&b->irq_lock); if (b->irq_armed) return false; -- GitLab From af0ab55ffe2467fab82bc0909e469cba7701fac7 Mon Sep 17 00:00:00 2001 From: Hans de Goede Date: Thu, 19 Oct 2017 13:16:19 +0200 Subject: [PATCH 0185/1471] x86/platform/intel/iosf_mbi: Add unlocked PMIC bus access notifier unregister For race free unregistration drivers may need to acquire PMIC bus access through iosf_mbi_punit_acquire() and then (un)register the notifier without dropping the lock. This commit adds an unlocked variant of iosf_mbi_unregister_pmic_bus_access_notifier for this use case. Acked-by: Ingo Molnar Signed-off-by: Hans de Goede Link: https://patchwork.freedesktop.org/patch/msgid/20171019111620.26761-2-hdegoede@redhat.com --- arch/x86/include/asm/iosf_mbi.h | 25 +++++++++++++++++++++++++ arch/x86/platform/intel/iosf_mbi.c | 19 +++++++++++++++++-- 2 files changed, 42 insertions(+), 2 deletions(-) diff --git a/arch/x86/include/asm/iosf_mbi.h b/arch/x86/include/asm/iosf_mbi.h index c313cac36f564..eb959cd77994f 100644 --- a/arch/x86/include/asm/iosf_mbi.h +++ b/arch/x86/include/asm/iosf_mbi.h @@ -145,6 +145,18 @@ int iosf_mbi_register_pmic_bus_access_notifier(struct notifier_block *nb); */ int iosf_mbi_unregister_pmic_bus_access_notifier(struct notifier_block *nb); +/** + * iosf_mbi_unregister_pmic_bus_access_notifier_unlocked - Unregister PMIC bus + * notifier, unlocked + * + * Like iosf_mbi_unregister_pmic_bus_access_notifier(), but for use when the + * caller has already called iosf_mbi_punit_acquire() itself. + * + * @nb: notifier_block to unregister + */ +int iosf_mbi_unregister_pmic_bus_access_notifier_unlocked( + struct notifier_block *nb); + /** * iosf_mbi_call_pmic_bus_access_notifier_chain - Call PMIC bus notifier chain * @@ -153,6 +165,11 @@ int iosf_mbi_unregister_pmic_bus_access_notifier(struct notifier_block *nb); */ int iosf_mbi_call_pmic_bus_access_notifier_chain(unsigned long val, void *v); +/** + * iosf_mbi_assert_punit_acquired - Assert that the P-Unit has been acquired. + */ +void iosf_mbi_assert_punit_acquired(void); + #else /* CONFIG_IOSF_MBI is not enabled */ static inline bool iosf_mbi_available(void) @@ -196,12 +213,20 @@ int iosf_mbi_unregister_pmic_bus_access_notifier(struct notifier_block *nb) return 0; } +static inline int +iosf_mbi_unregister_pmic_bus_access_notifier_unlocked(struct notifier_block *nb) +{ + return 0; +} + static inline int iosf_mbi_call_pmic_bus_access_notifier_chain(unsigned long val, void *v) { return 0; } +static inline void iosf_mbi_assert_punit_acquired(void) {} + #endif /* CONFIG_IOSF_MBI */ #endif /* IOSF_MBI_SYMS_H */ diff --git a/arch/x86/platform/intel/iosf_mbi.c b/arch/x86/platform/intel/iosf_mbi.c index a952ac1997414..6f37a2137a795 100644 --- a/arch/x86/platform/intel/iosf_mbi.c +++ b/arch/x86/platform/intel/iosf_mbi.c @@ -218,14 +218,23 @@ int iosf_mbi_register_pmic_bus_access_notifier(struct notifier_block *nb) } EXPORT_SYMBOL(iosf_mbi_register_pmic_bus_access_notifier); +int iosf_mbi_unregister_pmic_bus_access_notifier_unlocked( + struct notifier_block *nb) +{ + iosf_mbi_assert_punit_acquired(); + + return blocking_notifier_chain_unregister( + &iosf_mbi_pmic_bus_access_notifier, nb); +} +EXPORT_SYMBOL(iosf_mbi_unregister_pmic_bus_access_notifier_unlocked); + int iosf_mbi_unregister_pmic_bus_access_notifier(struct notifier_block *nb) { int ret; /* Wait for the bus to go inactive before unregistering */ mutex_lock(&iosf_mbi_punit_mutex); - ret = blocking_notifier_chain_unregister( - &iosf_mbi_pmic_bus_access_notifier, nb); + ret = iosf_mbi_unregister_pmic_bus_access_notifier_unlocked(nb); mutex_unlock(&iosf_mbi_punit_mutex); return ret; @@ -239,6 +248,12 @@ int iosf_mbi_call_pmic_bus_access_notifier_chain(unsigned long val, void *v) } EXPORT_SYMBOL(iosf_mbi_call_pmic_bus_access_notifier_chain); +void iosf_mbi_assert_punit_acquired(void) +{ + WARN_ON(!mutex_is_locked(&iosf_mbi_punit_mutex)); +} +EXPORT_SYMBOL(iosf_mbi_assert_punit_acquired); + #ifdef CONFIG_IOSF_MBI_DEBUG static u32 dbg_mdr; static u32 dbg_mcr; -- GitLab From a5266db4d31410fbfb4f40118e58085994e83dbc Mon Sep 17 00:00:00 2001 From: Hans de Goede Date: Thu, 19 Oct 2017 13:16:20 +0200 Subject: [PATCH 0186/1471] drm/i915: Acquire PUNIT->PMIC bus for intel_uncore_forcewake_reset() intel_uncore_forcewake_reset() does forcewake puts and gets as such we need to make sure that no-one tries to access the PUNIT->PMIC bus (on systems where this bus is shared) while it runs, otherwise bad things happen. Normally this is taken care of by the i915_pmic_bus_access_notifier() which does an intel_uncore_forcewake_get(FORCEWAKE_ALL) when some other driver tries to access the PMIC bus, so that later forcewake gets are no-ops (for the duration of the bus access). But intel_uncore_forcewake_reset gets called in 3 cases: 1) Before registering the pmic_bus_access_notifier 2) After unregistering the pmic_bus_access_notifier 3) To reset forcewake state on a GPU reset In all 3 cases the i915_pmic_bus_access_notifier() protection is insufficient. This commit fixes this race by calling iosf_mbi_punit_acquire() before calling intel_uncore_forcewake_reset(). In the case where it is called directly after unregistering the pmic_bus_access_notifier, we need to hold the punit-lock over both calls to avoid a race where intel_uncore_fw_release_timer() may execute between the 2 calls. Reviewed-by: Imre Deak Acked-by: Ingo Molnar Signed-off-by: Hans de Goede Link: https://patchwork.freedesktop.org/patch/msgid/20171019111620.26761-3-hdegoede@redhat.com --- drivers/gpu/drm/i915/intel_uncore.c | 17 +++++++++++++---- drivers/gpu/drm/i915/selftests/intel_uncore.c | 3 +++ 2 files changed, 16 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 0a6d952a2df14..211acee7c31da 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -345,6 +345,7 @@ intel_uncore_fw_release_timer(struct hrtimer *timer) return HRTIMER_NORESTART; } +/* Note callers must have acquired the PUNIT->PMIC bus, before calling this. */ static void intel_uncore_forcewake_reset(struct drm_i915_private *dev_priv, bool restore) { @@ -353,6 +354,8 @@ static void intel_uncore_forcewake_reset(struct drm_i915_private *dev_priv, int retry_count = 100; enum forcewake_domains fw, active_domains; + iosf_mbi_assert_punit_acquired(); + /* Hold uncore.lock across reset to prevent any register access * with forcewake not set correctly. Wait until all pending * timers are run before holding. @@ -532,14 +535,18 @@ static void __intel_uncore_early_sanitize(struct drm_i915_private *dev_priv, GT_FIFO_CTL_RC6_POLICY_STALL); } + iosf_mbi_punit_acquire(); intel_uncore_forcewake_reset(dev_priv, restore_forcewake); + iosf_mbi_punit_release(); } void intel_uncore_suspend(struct drm_i915_private *dev_priv) { - iosf_mbi_unregister_pmic_bus_access_notifier( + iosf_mbi_punit_acquire(); + iosf_mbi_unregister_pmic_bus_access_notifier_unlocked( &dev_priv->uncore.pmic_bus_access_nb); intel_uncore_forcewake_reset(dev_priv, false); + iosf_mbi_punit_release(); } void intel_uncore_resume_early(struct drm_i915_private *dev_priv) @@ -1419,12 +1426,14 @@ void intel_uncore_init(struct drm_i915_private *dev_priv) void intel_uncore_fini(struct drm_i915_private *dev_priv) { - iosf_mbi_unregister_pmic_bus_access_notifier( - &dev_priv->uncore.pmic_bus_access_nb); - /* Paranoia: make sure we have disabled everything before we exit. */ intel_uncore_sanitize(dev_priv); + + iosf_mbi_punit_acquire(); + iosf_mbi_unregister_pmic_bus_access_notifier_unlocked( + &dev_priv->uncore.pmic_bus_access_nb); intel_uncore_forcewake_reset(dev_priv, false); + iosf_mbi_punit_release(); } static const struct reg_whitelist { diff --git a/drivers/gpu/drm/i915/selftests/intel_uncore.c b/drivers/gpu/drm/i915/selftests/intel_uncore.c index f52a4ab9aa986..2f63676431713 100644 --- a/drivers/gpu/drm/i915/selftests/intel_uncore.c +++ b/drivers/gpu/drm/i915/selftests/intel_uncore.c @@ -148,7 +148,10 @@ static int intel_uncore_check_forcewake_domains(struct drm_i915_private *dev_pri for_each_set_bit(offset, valid, FW_RANGE) { i915_reg_t reg = { offset }; + iosf_mbi_punit_acquire(); intel_uncore_forcewake_reset(dev_priv, false); + iosf_mbi_punit_release(); + check_for_unclaimed_mmio(dev_priv); (void)I915_READ(reg); -- GitLab From 30b29406d9374989f34bce0eadaa630813049808 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 11:25:50 +0000 Subject: [PATCH 0187/1471] drm/i915: Restore the wait for idle engine after flushing interrupts So it appears that commit 5427f207852d ("drm/i915: Bump wait-times for the final CS interrupt before parking") was a little over optimistic in its belief that it had successfully waited for all residual activity on the engines before parking. Numerous sightings in CI since then of <7>[ 52.542886] [IGT] core_auth: executing <3>[ 52.561013] [drm:intel_engines_park [i915]] *ERROR* vcs0 is not idle before parking <7>[ 52.561215] intel_engines_park vcs0 <7>[ 52.561229] intel_engines_park current seqno 98, last 98, hangcheck 0 [-247449 ms], inflight 0 <7>[ 52.561238] intel_engines_park Reset count: 0 <7>[ 52.561266] intel_engines_park Requests: <7>[ 52.561363] intel_engines_park RING_START: 0x00000000 [0x00000000] <7>[ 52.561377] intel_engines_park RING_HEAD: 0x00000000 [0x00000000] <7>[ 52.561390] intel_engines_park RING_TAIL: 0x00000000 [0x00000000] <7>[ 52.561406] intel_engines_park RING_CTL: 0x00000000 <7>[ 52.561422] intel_engines_park RING_MODE: 0x00000200 [idle] <7>[ 52.561442] intel_engines_park ACTHD: 0x00000000_00000000 <7>[ 52.561459] intel_engines_park BBADDR: 0x00000000_00000000 <7>[ 52.561474] intel_engines_park Execlist status: 0x00000301 00000000 <7>[ 52.561489] intel_engines_park Execlist CSB read 5 [5 cached], write 5 [5 from hws], interrupt posted? no <7>[ 52.561500] intel_engines_park ELSP[0] idle <7>[ 52.561510] intel_engines_park ELSP[1] idle <7>[ 52.561519] intel_engines_park HW active? 0x0 <7>[ 52.561608] intel_engines_park Idle? yes <7>[ 52.561617] intel_engines_park on Braswell, which indicates that the engine just needs that little bit longer after flushing the tasklet to settle. So give it a few more milliseconds before declaring an err and applying the emergency brake. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103479 Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171110112550.28909-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/intel_engine_cs.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 6cb8e3ed97e4d..87778f03393b3 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1626,11 +1626,12 @@ void intel_engines_park(struct drm_i915_private *i915) * will be no more interrupts arriving later and the engines * are truly idle. */ - if (!intel_engine_is_idle(engine)) { + if (wait_for(intel_engine_is_idle(engine), 10)) { struct drm_printer p = drm_debug_printer(__func__); - DRM_ERROR("%s is not idle before parking\n", - engine->name); + dev_err(i915->drm.dev, + "%s is not idle before parking\n", + engine->name); intel_engine_dump(engine, &p); } -- GitLab From 9511ce1ce7299029acc985ce5f96a1e2e703f084 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 15:19:19 +0000 Subject: [PATCH 0188/1471] drm/i915/selftests: Initialise mock_i915->mm.obj_lock lockdep spotted that the mock tests were using the i915->mm.obj_lock without first initialiasing it: >[ 1303.217043] [IGT] drv_selftest: starting subtest mock_objects <4>[ 1303.240898] Setting dangerous option mock_selftests - tainting kernel <6>[ 1303.253665] i915: Performing mock selftests with st_random_seed=0xd87ea6c6 st_timeout=1000 <4>[ 1303.254812] INFO: trying to register non-static key. <4>[ 1303.254816] the code is fine but needs lockdep annotation. <4>[ 1303.254818] turning off the locking correctness validator. <4>[ 1303.254820] CPU: 4 PID: 13112 Comm: drv_selftest Tainted: G U W 4.14.0-rc8-CI-Patchwork_7058+ #1 <4>[ 1303.254823] Hardware name: MSI MS-7924/Z97M-G43(MS-7924), BIOS V1.12 02/15/2016 <4>[ 1303.254825] Call Trace: <4>[ 1303.254829] dump_stack+0x68/0x9f <4>[ 1303.254832] register_lock_class+0x3fd/0x580 <4>[ 1303.254835] ? ___slab_alloc.constprop.29+0x157/0x3d0 <4>[ 1303.254837] ? ___slab_alloc.constprop.29+0x157/0x3d0 <4>[ 1303.254840] ? sg_kmalloc+0x1e/0x50 <4>[ 1303.254842] ? debug_smp_processor_id+0x17/0x20 <4>[ 1303.254845] __lock_acquire+0xa4/0x1b00 <4>[ 1303.254884] ? __i915_gem_object_set_pages+0x116/0x1f0 [i915] <4>[ 1303.254887] ? __this_cpu_preempt_check+0x13/0x20 <4>[ 1303.254889] ? sg_kmalloc+0x1e/0x50 <4>[ 1303.254891] lock_acquire+0xb0/0x200 <4>[ 1303.254893] ? lock_acquire+0xb0/0x200 <4>[ 1303.254917] ? __i915_gem_object_set_pages+0x116/0x1f0 [i915] <4>[ 1303.254920] _raw_spin_lock+0x32/0x50 <4>[ 1303.254944] ? __i915_gem_object_set_pages+0x116/0x1f0 [i915] <4>[ 1303.254967] __i915_gem_object_set_pages+0x116/0x1f0 [i915] <4>[ 1303.254991] i915_gem_object_get_pages_phys+0x286/0x2b0 [i915] <4>[ 1303.255015] ____i915_gem_object_get_pages+0x20/0x60 [i915] <4>[ 1303.255039] i915_gem_object_attach_phys+0x137/0x1a0 [i915] <4>[ 1303.255063] igt_phys_object+0x45/0x120 [i915] <4>[ 1303.255094] __i915_subtests+0x40/0xd0 [i915] <4>[ 1303.255099] ? work_on_cpu_safe+0x60/0x60 <4>[ 1303.255131] i915_gem_object_mock_selftests+0x34/0x50 [i915] Signed-off-by: Chris Wilson Cc: Matthew Auld Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171110151919.18451-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/selftests/mock_gem_device.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index 04eb9362f4f87..d2bf2729c3314 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -180,6 +180,7 @@ struct drm_i915_private *mock_gem_device(void) I915_GTT_PAGE_SIZE_2M; spin_lock_init(&i915->mm.object_stat_lock); + spin_lock_init(&i915->mm.obj_lock); mock_uncore_init(i915); init_waitqueue_head(&i915->gpu_error.wait_queue); -- GitLab From 1803fcbca2e444f7972430c4dc1c3e98c6ee1bc9 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 10 Nov 2017 14:26:27 +0000 Subject: [PATCH 0189/1471] drm/i915: Define an engine class enum for the uABI We want to be able to report back to userspace details about an engine's class, and in return for userspace to be able to request actions regarding certain classes of engines. To isolate the uABI from any variations between hw generations, we define an abstract class for the engines and internally map onto the hw. v2: Remove MAX from the uABI; keep it internal if we need it, but don't let userspace make the mistake of using it themselves. v3: s/OTHER/INVALID/ The use of OTHER is ill-defined, so remove it from the uABI as any future new type of engine can define a class to suit it. But keep a reserved value for an invalid class, so that we can always unambiguously express when something doesn't belong to the classification. Signed-off-by: Tvrtko Ursulin Signed-off-by: Chris Wilson Cc: Lionel Landwerlin Reviewed-by: Joonas Lahtinen #v2 Reviewed-by: Lionel Landwerlin Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 10 +++++++++- drivers/gpu/drm/i915/intel_ringbuffer.h | 5 ++++- include/uapi/drm/i915_drm.h | 16 ++++++++++++++++ 3 files changed, 29 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 87778f03393b3..bded9c40dbd53 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -50,6 +50,8 @@ struct engine_class_info { const char *name; int (*init_legacy)(struct intel_engine_cs *engine); int (*init_execlists)(struct intel_engine_cs *engine); + + u8 uabi_class; }; static const struct engine_class_info intel_engine_classes[] = { @@ -57,21 +59,25 @@ static const struct engine_class_info intel_engine_classes[] = { .name = "rcs", .init_execlists = logical_render_ring_init, .init_legacy = intel_init_render_ring_buffer, + .uabi_class = I915_ENGINE_CLASS_RENDER, }, [COPY_ENGINE_CLASS] = { .name = "bcs", .init_execlists = logical_xcs_ring_init, .init_legacy = intel_init_blt_ring_buffer, + .uabi_class = I915_ENGINE_CLASS_COPY, }, [VIDEO_DECODE_CLASS] = { .name = "vcs", .init_execlists = logical_xcs_ring_init, .init_legacy = intel_init_bsd_ring_buffer, + .uabi_class = I915_ENGINE_CLASS_VIDEO, }, [VIDEO_ENHANCEMENT_CLASS] = { .name = "vecs", .init_execlists = logical_xcs_ring_init, .init_legacy = intel_init_vebox_ring_buffer, + .uabi_class = I915_ENGINE_CLASS_VIDEO_ENHANCE, }, }; @@ -213,13 +219,15 @@ intel_engine_setup(struct drm_i915_private *dev_priv, WARN_ON(snprintf(engine->name, sizeof(engine->name), "%s%u", class_info->name, info->instance) >= sizeof(engine->name)); - engine->uabi_id = info->uabi_id; engine->hw_id = engine->guc_id = info->hw_id; engine->mmio_base = info->mmio_base; engine->irq_shift = info->irq_shift; engine->class = info->class; engine->instance = info->instance; + engine->uabi_id = info->uabi_id; + engine->uabi_class = class_info->uabi_class; + engine->context_size = __intel_engine_context_size(dev_priv, engine->class); if (WARN_ON(engine->context_size > BIT(20))) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 69ad875fd0111..f3dbfe7ae6e4f 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -289,11 +289,14 @@ struct intel_engine_execlists { struct intel_engine_cs { struct drm_i915_private *i915; char name[INTEL_ENGINE_CS_MAX_NAME]; + enum intel_engine_id id; - unsigned int uabi_id; unsigned int hw_id; unsigned int guc_id; + u8 uabi_id; + u8 uabi_class; + u8 class; u8 instance; u32 context_size; diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index ac3c6503ca27f..1f7dfb22a7c21 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -86,6 +86,22 @@ enum i915_mocs_table_index { I915_MOCS_CACHED, }; +/* + * Different engines serve different roles, and there may be more than one + * engine serving each role. enum drm_i915_gem_engine_class provides a + * classification of the role of the engine, which may be used when requesting + * operations to be performed on a certain subset of engines, or for providing + * information about that group. + */ +enum drm_i915_gem_engine_class { + I915_ENGINE_CLASS_RENDER = 0, + I915_ENGINE_CLASS_COPY = 1, + I915_ENGINE_CLASS_VIDEO = 2, + I915_ENGINE_CLASS_VIDEO_ENHANCE = 3, + + I915_ENGINE_CLASS_INVALID = -1 +}; + /* Each region is a minimum of 16k, and there are at most 255 of them. */ #define I915_NR_TEX_REGIONS 255 /* table size 2k - maximum due to use -- GitLab From ae6c4574782dbfebcbf1f7e3620bcaf58ceb69e3 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:28 +0000 Subject: [PATCH 0190/1471] drm/i915: Force the switch to the i915->kernel_context In the next few patches, we will have a hard requirement that we emit a context-switch to the perma-pinned i915->kernel_context (so that we can save the HW state using that context-switch). As the first context itself may be classed as a kernel context, we want to be explicit in our comparison. For an extra-layer of finesse, we can check the last unretired context on the engine; as well as the last retired context when idle. v2: verbose verbosity v3: Always force the switch, even when the engine is idle, and update the assert that this happens before suspend. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen #v1 Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 10 ++++++---- drivers/gpu/drm/i915/intel_engine_cs.c | 26 ++++++++++++++++++++++++-- 2 files changed, 30 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 889ae8810d5f5..824515556733b 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4693,14 +4693,16 @@ void __i915_gem_object_release_unless_active(struct drm_i915_gem_object *obj) i915_gem_object_put(obj); } -static void assert_kernel_context_is_current(struct drm_i915_private *dev_priv) +static void assert_kernel_context_is_current(struct drm_i915_private *i915) { + struct i915_gem_context *kernel_context = i915->kernel_context; struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, dev_priv, id) - GEM_BUG_ON(engine->last_retired_context && - !i915_gem_context_is_kernel(engine->last_retired_context)); + for_each_engine(engine, i915, id) { + GEM_BUG_ON(__i915_gem_active_peek(&engine->timeline->last_request)); + GEM_BUG_ON(engine->last_retired_context != kernel_context); + } } void i915_gem_sanitize(struct drm_i915_private *i915) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index bded9c40dbd53..6d0b38189a7d7 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1596,10 +1596,32 @@ bool intel_engines_are_idle(struct drm_i915_private *dev_priv) return true; } +/** + * intel_engine_has_kernel_context: + * @engine: the engine + * + * Returns true if the last context to be executed on this engine, or has been + * executed if the engine is already idle, is the kernel context + * (#i915.kernel_context). + */ bool intel_engine_has_kernel_context(const struct intel_engine_cs *engine) { - return (!engine->last_retired_context || - i915_gem_context_is_kernel(engine->last_retired_context)); + const struct i915_gem_context * const kernel_context = + engine->i915->kernel_context; + struct drm_i915_gem_request *rq; + + lockdep_assert_held(&engine->i915->drm.struct_mutex); + + /* + * Check the last context seen by the engine. If active, it will be + * the last request that remains in the timeline. When idle, it is + * the last executed context as tracked by retirement. + */ + rq = __i915_gem_active_peek(&engine->timeline->last_request); + if (rq) + return rq->ctx == kernel_context; + else + return engine->last_retired_context == kernel_context; } void intel_engines_reset_default_submission(struct drm_i915_private *i915) -- GitLab From f58d13d5717938d4dfcc82a2eeba0a6d7644f6e5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:29 +0000 Subject: [PATCH 0191/1471] drm/i915: Move GT powersaving init to i915_gem_init() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit GT powersaving is tightly coupled to the request infrastructure. To avoid complications with the order of initialisation in the next patch (where we want to send requests to hw during GEM init) move the powersaving initialisation into the purview of i915_gem_init(). Signed-off-by: Chris Wilson Cc: Ville Syrjälä Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-3-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 2 ++ drivers/gpu/drm/i915/intel_display.c | 2 -- drivers/gpu/drm/i915/intel_pm.c | 2 -- 3 files changed, 2 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 824515556733b..37586f703c1e2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5021,6 +5021,8 @@ int i915_gem_init(struct drm_i915_private *dev_priv) if (ret) goto out_unlock; + intel_init_gt_powersave(dev_priv); + ret = i915_gem_init_hw(dev_priv); if (ret == -EIO) { /* Allow engine initialisation to fail by marking the GPU as diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 435dc18aa2394..da11ba8986217 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15198,8 +15198,6 @@ void intel_modeset_gem_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = to_i915(dev); - intel_init_gt_powersave(dev_priv); - intel_init_clock_gating(dev_priv); intel_setup_overlay(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index e09377df590d8..48a127ac6de53 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7918,7 +7918,6 @@ void intel_init_gt_powersave(struct drm_i915_private *dev_priv) intel_runtime_pm_get(dev_priv); } - mutex_lock(&dev_priv->drm.struct_mutex); mutex_lock(&dev_priv->pcu_lock); /* Initialize RPS limits (for userspace) */ @@ -7960,7 +7959,6 @@ void intel_init_gt_powersave(struct drm_i915_private *dev_priv) rps->boost_freq = rps->max_freq; mutex_unlock(&dev_priv->pcu_lock); - mutex_unlock(&dev_priv->drm.struct_mutex); intel_autoenable_gt_powersave(dev_priv); } -- GitLab From cc6a818ad6bdb0d3008314cbd0fc9c9a2cd02695 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:30 +0000 Subject: [PATCH 0192/1471] drm/i915: Move intel_init_clock_gating() to i915_gem_init() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Despite its name intel_init_clock_gating applies both display clock gating workarounds; GT mmio workarounds and the occasional GT power context workaround. Worse, sometimes it includes a context register workaround which we need to apply before we record the default HW state for all contexts. Signed-off-by: Chris Wilson Cc: Ville Syrjälä Cc: Joonas Lahtinen Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-4-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 17 +++++++++++++++-- drivers/gpu/drm/i915/intel_display.c | 2 -- 2 files changed, 15 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 37586f703c1e2..ed335612be250 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5024,6 +5024,21 @@ int i915_gem_init(struct drm_i915_private *dev_priv) intel_init_gt_powersave(dev_priv); ret = i915_gem_init_hw(dev_priv); + if (ret) + goto out_unlock; + + /* + * Despite its name intel_init_clock_gating applies both display + * clock gating workarounds; GT mmio workarounds and the occasional + * GT power context workaround. Worse, sometimes it includes a context + * register workaround which we need to apply before we record the + * default HW state for all contexts. + * + * FIXME: break up the workarounds and apply them at the right time! + */ + intel_init_clock_gating(dev_priv); + +out_unlock: if (ret == -EIO) { /* Allow engine initialisation to fail by marking the GPU as * wedged. But we only want to do this where the GPU is angry, @@ -5035,8 +5050,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv) } ret = 0; } - -out_unlock: intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); mutex_unlock(&dev_priv->drm.struct_mutex); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index da11ba8986217..63cc09ade9d19 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15198,8 +15198,6 @@ void intel_modeset_gem_init(struct drm_device *dev) { struct drm_i915_private *dev_priv = to_i915(dev); - intel_init_clock_gating(dev_priv); - intel_setup_overlay(dev_priv); } -- GitLab From d378a3efb819e6d1992127122d957337571b4594 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:31 +0000 Subject: [PATCH 0193/1471] drm/i915: Inline intel_modeset_gem_init() intel_modeset_gem_init() now only sets up the legacy overlay, so let's remove the function and call the setup directly during driver load. This should help us find a better point in the initialisation sequence for it later. Signed-off-by: Chris Wilson Reviewed-by: Joonas Lahtinen Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-5-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_drv.c | 2 +- drivers/gpu/drm/i915/i915_drv.h | 1 - drivers/gpu/drm/i915/intel_display.c | 7 ------- 3 files changed, 1 insertion(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index e7e9e061073b7..1b440f2b90a5e 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -676,7 +676,7 @@ static int i915_load_modeset_init(struct drm_device *dev) if (ret) goto cleanup_uc; - intel_modeset_gem_init(dev); + intel_setup_overlay(dev_priv); if (INTEL_INFO(dev_priv)->num_pipes == 0) return 0; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index c94f34f5cee3c..0be9a918697a3 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -4124,7 +4124,6 @@ void intel_device_info_dump(struct drm_i915_private *dev_priv); /* modesetting */ extern void intel_modeset_init_hw(struct drm_device *dev); extern int intel_modeset_init(struct drm_device *dev); -extern void intel_modeset_gem_init(struct drm_device *dev); extern void intel_modeset_cleanup(struct drm_device *dev); extern int intel_connector_register(struct drm_connector *); extern void intel_connector_unregister(struct drm_connector *); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 63cc09ade9d19..5c7540f3f5dcd 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15194,13 +15194,6 @@ void intel_display_resume(struct drm_device *dev) drm_atomic_state_put(state); } -void intel_modeset_gem_init(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = to_i915(dev); - - intel_setup_overlay(dev_priv); -} - int intel_connector_register(struct drm_connector *connector) { struct intel_connector *intel_connector = to_intel_connector(connector); -- GitLab From f4e15af7e21861445821d5f09922ef7e695269a1 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:32 +0000 Subject: [PATCH 0194/1471] drm/i915: Mark the context state as dirty/written In the next few patches, we will want to both copy out of the context image and write a valid image into a new context. To be completely safe, we should then couple in our domain tracking to ensure that we don't have any issues with stale data remaining in unwanted cachelines. Historically, we omitted the .write=true from the call to set-gtt-domain in i915_switch_context() in order to avoid a stall between every request as we would want to wait for the previous context write from the gpu. Since then, we limit the set-gtt-domain to only occur when we first bind the vma, so once in use we will never stall, and we are sure to flush the context following a load from swap. Equally we never applied the lessons learnt from ringbuffer submission to execlists; so time to apply the flush of the lrc after load as well. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Acked-by: Joonas Lahtinen Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-6-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_lrc.c | 32 ++++++++++++++++++------- drivers/gpu/drm/i915/intel_ringbuffer.c | 6 ++--- 2 files changed, 27 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 020ca7c9c75f9..1baaab35905c8 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1077,12 +1077,34 @@ static void execlists_schedule(struct drm_i915_gem_request *request, int prio) spin_unlock_irq(&engine->timeline->lock); } +static int __context_pin(struct i915_gem_context *ctx, struct i915_vma *vma) +{ + unsigned int flags; + int err; + + /* + * Clear this page out of any CPU caches for coherent swap-in/out. + * We only want to do this on the first bind so that we do not stall + * on an active context (which by nature is already on the GPU). + */ + if (!(vma->flags & I915_VMA_GLOBAL_BIND)) { + err = i915_gem_object_set_to_gtt_domain(vma->obj, true); + if (err) + return err; + } + + flags = PIN_GLOBAL | PIN_HIGH; + if (ctx->ggtt_offset_bias) + flags |= PIN_OFFSET_BIAS | ctx->ggtt_offset_bias; + + return i915_vma_pin(vma, 0, GEN8_LR_CONTEXT_ALIGN, flags); +} + static struct intel_ring * execlists_context_pin(struct intel_engine_cs *engine, struct i915_gem_context *ctx) { struct intel_context *ce = &ctx->engine[engine->id]; - unsigned int flags; void *vaddr; int ret; @@ -1099,11 +1121,7 @@ execlists_context_pin(struct intel_engine_cs *engine, } GEM_BUG_ON(!ce->state); - flags = PIN_GLOBAL | PIN_HIGH; - if (ctx->ggtt_offset_bias) - flags |= PIN_OFFSET_BIAS | ctx->ggtt_offset_bias; - - ret = i915_vma_pin(ce->state, 0, GEN8_LR_CONTEXT_ALIGN, flags); + ret = __context_pin(ctx, ce->state); if (ret) goto err; @@ -1123,9 +1141,7 @@ execlists_context_pin(struct intel_engine_cs *engine, ce->lrc_reg_state[CTX_RING_BUFFER_START+1] = i915_ggtt_offset(ce->ring->vma); - ce->state->obj->mm.dirty = true; ce->state->obj->pin_global++; - i915_gem_context_get(ctx); out: return ce->ring; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 47fadf8da84e9..7e2a671882fb5 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1363,12 +1363,13 @@ static int context_pin(struct i915_gem_context *ctx) struct i915_vma *vma = ctx->engine[RCS].state; int ret; - /* Clear this page out of any CPU caches for coherent swap-in/out. + /* + * Clear this page out of any CPU caches for coherent swap-in/out. * We only want to do this on the first bind so that we do not stall * on an active context (which by nature is already on the GPU). */ if (!(vma->flags & I915_VMA_GLOBAL_BIND)) { - ret = i915_gem_object_set_to_gtt_domain(vma->obj, false); + ret = i915_gem_object_set_to_gtt_domain(vma->obj, true); if (ret) return ret; } @@ -1445,7 +1446,6 @@ intel_ring_context_pin(struct intel_engine_cs *engine, if (ret) goto err; - ce->state->obj->mm.dirty = true; ce->state->obj->pin_global++; } -- GitLab From d2b4b97933f5adacfba42dc3b9200d0e21fbe2c4 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:33 +0000 Subject: [PATCH 0195/1471] drm/i915: Record the default hw state after reset upon load MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Take a copy of the HW state after a reset upon module loading by executing a context switch from a blank context to the kernel context, thus saving the default hw state over the blank context image. We can then use the default hw state to initialise any future context, ensuring that each starts with the default view of hw state. v2: Unmap our default state from the GTT after stealing it from the context. This should stop us from accidentally overwriting it via the GTT (and frees up some precious GTT space). Testcase: igt/gem_ctx_isolation Signed-off-by: Chris Wilson Cc: Ville Syrjälä Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-7-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gvt/scheduler.c | 2 - drivers/gpu/drm/i915/i915_debugfs.c | 1 - drivers/gpu/drm/i915/i915_drv.c | 3 + drivers/gpu/drm/i915/i915_gem.c | 115 ++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_gem_context.c | 55 +++--------- drivers/gpu/drm/i915/i915_gem_context.h | 4 +- drivers/gpu/drm/i915/intel_engine_cs.c | 17 ++++ drivers/gpu/drm/i915/intel_lrc.c | 38 +++++--- drivers/gpu/drm/i915/intel_ringbuffer.c | 45 +++++++--- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 + include/uapi/drm/i915_drm.h | 15 ++++ 11 files changed, 224 insertions(+), 73 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index f6ded475bb2cc..42cc61230ca7d 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -723,8 +723,6 @@ int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu) if (IS_ERR(vgpu->shadow_ctx)) return PTR_ERR(vgpu->shadow_ctx); - vgpu->shadow_ctx->engine[RCS].initialised = true; - bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES); return 0; diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index d89321f0468c0..533ba096b9a63 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1974,7 +1974,6 @@ static int i915_context_status(struct seq_file *m, void *unused) struct intel_context *ce = &ctx->engine[engine->id]; seq_printf(m, "%s: ", engine->name); - seq_putc(m, ce->initialised ? 'I' : 'i'); if (ce->state) describe_obj(m, ce->state->obj); if (ce->ring) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 1b440f2b90a5e..d97fe9c9439a4 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -406,6 +406,9 @@ static int i915_getparam(struct drm_device *dev, void *data, */ value = 1; break; + case I915_PARAM_HAS_CONTEXT_ISOLATION: + value = intel_engines_has_context_isolation(dev_priv); + break; case I915_PARAM_SLICE_MASK: value = INTEL_INFO(dev_priv)->sseu.slice_mask; if (!value) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index ed335612be250..4bf1183040864 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4972,6 +4972,120 @@ bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value) return true; } +static int __intel_engines_record_defaults(struct drm_i915_private *i915) +{ + struct i915_gem_context *ctx; + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err; + + /* + * As we reset the gpu during very early sanitisation, the current + * register state on the GPU should reflect its defaults values. + * We load a context onto the hw (with restore-inhibit), then switch + * over to a second context to save that default register state. We + * can then prime every new context with that state so they all start + * from the same default HW values. + */ + + ctx = i915_gem_context_create_kernel(i915, 0); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + for_each_engine(engine, i915, id) { + struct drm_i915_gem_request *rq; + + rq = i915_gem_request_alloc(engine, ctx); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto out_ctx; + } + + err = i915_switch_context(rq); + if (engine->init_context) + err = engine->init_context(rq); + + __i915_add_request(rq, true); + if (err) + goto err_active; + } + + err = i915_gem_switch_to_kernel_context(i915); + if (err) + goto err_active; + + err = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED); + if (err) + goto err_active; + + assert_kernel_context_is_current(i915); + + for_each_engine(engine, i915, id) { + struct i915_vma *state; + + state = ctx->engine[id].state; + if (!state) + continue; + + /* + * As we will hold a reference to the logical state, it will + * not be torn down with the context, and importantly the + * object will hold onto its vma (making it possible for a + * stray GTT write to corrupt our defaults). Unmap the vma + * from the GTT to prevent such accidents and reclaim the + * space. + */ + err = i915_vma_unbind(state); + if (err) + goto err_active; + + err = i915_gem_object_set_to_cpu_domain(state->obj, false); + if (err) + goto err_active; + + engine->default_state = i915_gem_object_get(state->obj); + } + + if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) { + unsigned int found = intel_engines_has_context_isolation(i915); + + /* + * Make sure that classes with multiple engine instances all + * share the same basic configuration. + */ + for_each_engine(engine, i915, id) { + unsigned int bit = BIT(engine->uabi_class); + unsigned int expected = engine->default_state ? bit : 0; + + if ((found & bit) != expected) { + DRM_ERROR("mismatching default context state for class %d on engine %s\n", + engine->uabi_class, engine->name); + } + } + } + +out_ctx: + i915_gem_context_set_closed(ctx); + i915_gem_context_put(ctx); + return err; + +err_active: + /* + * If we have to abandon now, we expect the engines to be idle + * and ready to be torn-down. First try to flush any remaining + * request, ensure we are pointing at the kernel context and + * then remove it. + */ + if (WARN_ON(i915_gem_switch_to_kernel_context(i915))) + goto out_ctx; + + if (WARN_ON(i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED))) + goto out_ctx; + + i915_gem_contexts_lost(i915); + goto out_ctx; +} + int i915_gem_init(struct drm_i915_private *dev_priv) { int ret; @@ -5038,6 +5152,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv) */ intel_init_clock_gating(dev_priv); + ret = __intel_engines_record_defaults(dev_priv); out_unlock: if (ret == -EIO) { /* Allow engine initialisation to fail by marking the GPU as diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index c05c3d7d21a5e..2db0406950356 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -418,8 +418,8 @@ i915_gem_context_create_gvt(struct drm_device *dev) return ctx; } -static struct i915_gem_context * -create_kernel_context(struct drm_i915_private *i915, int prio) +struct i915_gem_context * +i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio) { struct i915_gem_context *ctx; @@ -473,7 +473,7 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv) ida_init(&dev_priv->contexts.hw_ida); /* lowest priority; idle task */ - ctx = create_kernel_context(dev_priv, I915_PRIORITY_MIN); + ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN); if (IS_ERR(ctx)) { DRM_ERROR("Failed to create default global context\n"); err = PTR_ERR(ctx); @@ -487,7 +487,7 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv) dev_priv->kernel_context = ctx; /* highest priority; preempting task */ - ctx = create_kernel_context(dev_priv, INT_MAX); + ctx = i915_gem_context_create_kernel(dev_priv, INT_MAX); if (IS_ERR(ctx)) { DRM_ERROR("Failed to create default preempt context\n"); err = PTR_ERR(ctx); @@ -522,28 +522,6 @@ void i915_gem_contexts_lost(struct drm_i915_private *dev_priv) engine->context_unpin(engine, engine->last_retired_context); engine->last_retired_context = NULL; } - - /* Force the GPU state to be restored on enabling */ - if (!i915_modparams.enable_execlists) { - struct i915_gem_context *ctx; - - list_for_each_entry(ctx, &dev_priv->contexts.list, link) { - if (!i915_gem_context_is_default(ctx)) - continue; - - for_each_engine(engine, dev_priv, id) - ctx->engine[engine->id].initialised = false; - - ctx->remap_slice = ALL_L3_SLICES(dev_priv); - } - - for_each_engine(engine, dev_priv, id) { - struct intel_context *kce = - &dev_priv->kernel_context->engine[engine->id]; - - kce->initialised = true; - } - } } void i915_gem_contexts_fini(struct drm_i915_private *i915) @@ -718,9 +696,6 @@ static inline bool skip_rcs_switch(struct i915_hw_ppgtt *ppgtt, if (to->remap_slice) return false; - if (!to->engine[RCS].initialised) - return false; - if (ppgtt && (intel_engine_flag(engine) & ppgtt->pd_dirty_rings)) return false; @@ -795,11 +770,14 @@ static int do_rcs_switch(struct drm_i915_gem_request *req) return ret; } - if (!to->engine[RCS].initialised || i915_gem_context_is_default(to)) - /* NB: If we inhibit the restore, the context is not allowed to - * die because future work may end up depending on valid address - * space. This means we must enforce that a page table load - * occur when this occurs. */ + if (i915_gem_context_is_kernel(to)) + /* + * The kernel context(s) is treated as pure scratch and is not + * expected to retain any state (as we sacrifice it during + * suspend and on resume it may be corrupted). This is ok, + * as nothing actually executes using the kernel context; it + * is purely used for flushing user contexts. + */ hw_flags = MI_RESTORE_INHIBIT; else if (ppgtt && intel_engine_flag(engine) & ppgtt->pd_dirty_rings) hw_flags = MI_FORCE_RESTORE; @@ -843,15 +821,6 @@ static int do_rcs_switch(struct drm_i915_gem_request *req) to->remap_slice &= ~(1<engine[RCS].initialised) { - if (engine->init_context) { - ret = engine->init_context(req); - if (ret) - return ret; - } - to->engine[RCS].initialised = true; - } - return 0; } diff --git a/drivers/gpu/drm/i915/i915_gem_context.h b/drivers/gpu/drm/i915/i915_gem_context.h index 44688e22a5c22..4bfb72f8e1cb2 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.h +++ b/drivers/gpu/drm/i915/i915_gem_context.h @@ -157,7 +157,6 @@ struct i915_gem_context { u32 *lrc_reg_state; u64 lrc_desc; int pin_count; - bool initialised; } engine[I915_NUM_ENGINES]; /** ring_size: size for allocating the per-engine ring buffer */ @@ -292,6 +291,9 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data, int i915_gem_context_reset_stats_ioctl(struct drm_device *dev, void *data, struct drm_file *file); +struct i915_gem_context * +i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio); + static inline struct i915_gem_context * i915_gem_context_get(struct i915_gem_context *ctx) { diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 6d0b38189a7d7..868c07a693b55 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -687,6 +687,9 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) intel_engine_cleanup_cmd_parser(engine); i915_gem_batch_pool_fini(&engine->batch_pool); + if (engine->default_state) + i915_gem_object_put(engine->default_state); + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) engine->context_unpin(engine, engine->i915->preempt_context); engine->context_unpin(engine, engine->i915->kernel_context); @@ -1705,6 +1708,20 @@ bool intel_engine_can_store_dword(struct intel_engine_cs *engine) } } +unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + unsigned int which; + + which = 0; + for_each_engine(engine, i915, id) + if (engine->default_state) + which |= BIT(engine->uabi_class); + + return which; +} + static void print_request(struct drm_printer *m, struct drm_i915_gem_request *rq, const char *prefix) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 1baaab35905c8..0c93f27f36eef 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1180,7 +1180,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request) struct intel_engine_cs *engine = request->engine; struct intel_context *ce = &request->ctx->engine[engine->id]; u32 *cs; - int ret; GEM_BUG_ON(!ce->pin_count); @@ -1194,14 +1193,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request) if (IS_ERR(cs)) return PTR_ERR(cs); - if (!ce->initialised) { - ret = engine->init_context(request); - if (ret) - return ret; - - ce->initialised = true; - } - /* Note that after this point, we have committed to using * this request as it is being used to both track the * state of engine initialisation and liveness of the @@ -2133,7 +2124,6 @@ static void execlists_init_reg_state(u32 *regs, CTX_REG(regs, CTX_CONTEXT_CONTROL, RING_CONTEXT_CONTROL(engine), _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH | - CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT | (HAS_RESOURCE_STREAMER(dev_priv) ? CTX_CTRL_RS_CTX_ENABLE : 0))); CTX_REG(regs, CTX_RING_HEAD, RING_HEAD(base), 0); @@ -2210,6 +2200,7 @@ populate_lr_context(struct i915_gem_context *ctx, struct intel_ring *ring) { void *vaddr; + u32 *regs; int ret; ret = i915_gem_object_set_to_cpu_domain(ctx_obj, true); @@ -2226,11 +2217,31 @@ populate_lr_context(struct i915_gem_context *ctx, } ctx_obj->mm.dirty = true; + if (engine->default_state) { + /* + * We only want to copy over the template context state; + * skipping over the headers reserved for GuC communication, + * leaving those as zero. + */ + const unsigned long start = LRC_HEADER_PAGES * PAGE_SIZE; + void *defaults; + + defaults = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (IS_ERR(defaults)) + return PTR_ERR(defaults); + + memcpy(vaddr + start, defaults + start, engine->context_size); + i915_gem_object_unpin_map(engine->default_state); + } + /* The second page of the context object contains some fields which must * be set up prior to the first execution. */ - - execlists_init_reg_state(vaddr + LRC_STATE_PN * PAGE_SIZE, - ctx, engine, ring); + regs = vaddr + LRC_STATE_PN * PAGE_SIZE; + execlists_init_reg_state(regs, ctx, engine, ring); + if (!engine->default_state) + regs[CTX_CONTEXT_CONTROL + 1] |= + _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT); i915_gem_object_unpin_map(ctx_obj); @@ -2283,7 +2294,6 @@ static int execlists_context_deferred_alloc(struct i915_gem_context *ctx, ce->ring = ring; ce->state = vma; - ce->initialised |= engine->init_context == NULL; return 0; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 7e2a671882fb5..464dc58af27b8 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1384,11 +1384,34 @@ alloc_context_vma(struct intel_engine_cs *engine) struct drm_i915_private *i915 = engine->i915; struct drm_i915_gem_object *obj; struct i915_vma *vma; + int err; obj = i915_gem_object_create(i915, engine->context_size); if (IS_ERR(obj)) return ERR_CAST(obj); + if (engine->default_state) { + void *defaults, *vaddr; + + vaddr = i915_gem_object_pin_map(obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + err = PTR_ERR(vaddr); + goto err_obj; + } + + defaults = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (IS_ERR(defaults)) { + err = PTR_ERR(defaults); + goto err_map; + } + + memcpy(vaddr, defaults, engine->context_size); + + i915_gem_object_unpin_map(engine->default_state); + i915_gem_object_unpin_map(obj); + } + /* * Try to make the context utilize L3 as well as LLC. * @@ -1410,10 +1433,18 @@ alloc_context_vma(struct intel_engine_cs *engine) } vma = i915_vma_instance(obj, &i915->ggtt.base, NULL); - if (IS_ERR(vma)) - i915_gem_object_put(obj); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto err_obj; + } return vma; + +err_map: + i915_gem_object_unpin_map(obj); +err_obj: + i915_gem_object_put(obj); + return ERR_PTR(err); } static struct intel_ring * @@ -1449,16 +1480,6 @@ intel_ring_context_pin(struct intel_engine_cs *engine, ce->state->obj->pin_global++; } - /* The kernel context is only used as a placeholder for flushing the - * active context. It is never used for submitting user rendering and - * as such never requires the golden render context, and so we can skip - * emitting it when we switch to the kernel context. This is required - * as during eviction we cannot allocate and pin the renderstate in - * order to initialise the context. - */ - if (i915_gem_context_is_kernel(ctx)) - ce->initialised = true; - i915_gem_context_get(ctx); out: diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index f3dbfe7ae6e4f..3372228591669 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -306,6 +306,7 @@ struct intel_engine_cs { struct intel_ring *buffer; struct intel_timeline *timeline; + struct drm_i915_gem_object *default_state; struct intel_render_state *render_state; atomic_t irq_count; @@ -932,6 +933,7 @@ void intel_engines_park(struct drm_i915_private *i915); void intel_engines_unpark(struct drm_i915_private *i915); void intel_engines_reset_default_submission(struct drm_i915_private *i915); +unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915); bool intel_engine_can_store_dword(struct intel_engine_cs *engine); diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 1f7dfb22a7c21..6c02ced663f81 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -466,6 +466,21 @@ typedef struct drm_i915_irq_wait { */ #define I915_PARAM_HAS_EXEC_FENCE_ARRAY 49 +/* + * Query whether every context (both per-file default and user created) is + * isolated (insofar as HW supports). If this parameter is not true, then + * freshly created contexts may inherit values from an existing context, + * rather than default HW values. If true, it also ensures (insofar as HW + * supports) that all state set by this context will not leak to any other + * context. + * + * As not every engine across every gen support contexts, the returned + * value reports the support of context isolation for individual engines by + * returning a bitmask of each engine class set to true if that class supports + * isolation. + */ +#define I915_PARAM_HAS_CONTEXT_ISOLATION 50 + typedef struct drm_i915_getparam { __s32 param; /* -- GitLab From 7c2fa7faf18f80f9fbbe7fcad8072604304db5dd Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 14:26:34 +0000 Subject: [PATCH 0196/1471] drm/i915: Stop caching the "golden" renderstate As we now record the default HW state and so only emit the "golden" renderstate once to prepare the HW, there is no advantage in keeping the renderstate batch around as it will never be used again. Signed-off-by: Chris Wilson Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-8-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_drv.h | 1 - drivers/gpu/drm/i915/i915_gem_render_state.c | 141 +++++++------------ drivers/gpu/drm/i915/i915_gem_render_state.h | 4 +- drivers/gpu/drm/i915/intel_engine_cs.c | 9 +- drivers/gpu/drm/i915/intel_lrc.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.c | 5 +- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 - 7 files changed, 54 insertions(+), 109 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 0be9a918697a3..40012b6daea27 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -67,7 +67,6 @@ #include "i915_gem_fence_reg.h" #include "i915_gem_object.h" #include "i915_gem_gtt.h" -#include "i915_gem_render_state.h" #include "i915_gem_request.h" #include "i915_gem_timeline.h" diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c index 3703dc91eedab..c2723a06fbb4e 100644 --- a/drivers/gpu/drm/i915/i915_gem_render_state.c +++ b/drivers/gpu/drm/i915/i915_gem_render_state.c @@ -26,10 +26,12 @@ */ #include "i915_drv.h" +#include "i915_gem_render_state.h" #include "intel_renderstate.h" struct intel_render_state { const struct intel_renderstate_rodata *rodata; + struct drm_i915_gem_object *obj; struct i915_vma *vma; u32 batch_offset; u32 batch_size; @@ -40,6 +42,9 @@ struct intel_render_state { static const struct intel_renderstate_rodata * render_state_get_rodata(const struct intel_engine_cs *engine) { + if (engine->id != RCS) + return NULL; + switch (INTEL_GEN(engine->i915)) { case 6: return &gen6_null_state; @@ -74,17 +79,16 @@ static int render_state_setup(struct intel_render_state *so, struct drm_i915_private *i915) { const struct intel_renderstate_rodata *rodata = so->rodata; - struct drm_i915_gem_object *obj = so->vma->obj; unsigned int i = 0, reloc_index = 0; unsigned int needs_clflush; u32 *d; int ret; - ret = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush); + ret = i915_gem_obj_prepare_shmem_write(so->obj, &needs_clflush); if (ret) return ret; - d = kmap_atomic(i915_gem_object_get_dirty_page(obj, 0)); + d = kmap_atomic(i915_gem_object_get_dirty_page(so->obj, 0)); while (i < rodata->batch_items) { u32 s = rodata->batch[i]; @@ -112,7 +116,7 @@ static int render_state_setup(struct intel_render_state *so, goto err; } - so->batch_offset = so->vma->node.start; + so->batch_offset = i915_ggtt_offset(so->vma); so->batch_size = rodata->batch_items * sizeof(u32); while (i % CACHELINE_DWORDS) @@ -160,9 +164,9 @@ static int render_state_setup(struct intel_render_state *so, drm_clflush_virt_range(d, i * sizeof(u32)); kunmap_atomic(d); - ret = i915_gem_object_set_to_gtt_domain(obj, false); + ret = i915_gem_object_set_to_gtt_domain(so->obj, false); out: - i915_gem_obj_finish_shmem_access(obj); + i915_gem_obj_finish_shmem_access(so->obj); return ret; err: @@ -173,112 +177,61 @@ static int render_state_setup(struct intel_render_state *so, #undef OUT_BATCH -int i915_gem_render_state_init(struct intel_engine_cs *engine) +int i915_gem_render_state_emit(struct drm_i915_gem_request *rq) { - struct intel_render_state *so; - const struct intel_renderstate_rodata *rodata; - struct drm_i915_gem_object *obj; - int ret; + struct intel_engine_cs *engine = rq->engine; + struct intel_render_state so = {}; /* keep the compiler happy */ + int err; - if (engine->id != RCS) + so.rodata = render_state_get_rodata(engine); + if (!so.rodata) return 0; - rodata = render_state_get_rodata(engine); - if (!rodata) - return 0; - - if (rodata->batch_items * 4 > PAGE_SIZE) + if (so.rodata->batch_items * 4 > PAGE_SIZE) return -EINVAL; - so = kmalloc(sizeof(*so), GFP_KERNEL); - if (!so) - return -ENOMEM; - - obj = i915_gem_object_create_internal(engine->i915, PAGE_SIZE); - if (IS_ERR(obj)) { - ret = PTR_ERR(obj); - goto err_free; - } + so.obj = i915_gem_object_create_internal(engine->i915, PAGE_SIZE); + if (IS_ERR(so.obj)) + return PTR_ERR(so.obj); - so->vma = i915_vma_instance(obj, &engine->i915->ggtt.base, NULL); - if (IS_ERR(so->vma)) { - ret = PTR_ERR(so->vma); + so.vma = i915_vma_instance(so.obj, &engine->i915->ggtt.base, NULL); + if (IS_ERR(so.vma)) { + err = PTR_ERR(so.vma); goto err_obj; } - so->rodata = rodata; - engine->render_state = so; - return 0; - -err_obj: - i915_gem_object_put(obj); -err_free: - kfree(so); - return ret; -} - -int i915_gem_render_state_emit(struct drm_i915_gem_request *req) -{ - struct intel_render_state *so; - int ret; - - lockdep_assert_held(&req->i915->drm.struct_mutex); - - so = req->engine->render_state; - if (!so) - return 0; - - /* Recreate the page after shrinking */ - if (!i915_gem_object_has_pages(so->vma->obj)) - so->batch_offset = -1; - - ret = i915_vma_pin(so->vma, 0, 0, PIN_GLOBAL | PIN_HIGH); - if (ret) - return ret; + err = i915_vma_pin(so.vma, 0, 0, PIN_GLOBAL | PIN_HIGH); + if (err) + goto err_vma; - if (so->vma->node.start != so->batch_offset) { - ret = render_state_setup(so, req->i915); - if (ret) - goto err_unpin; - } + err = render_state_setup(&so, rq->i915); + if (err) + goto err_unpin; - ret = req->engine->emit_flush(req, EMIT_INVALIDATE); - if (ret) + err = engine->emit_flush(rq, EMIT_INVALIDATE); + if (err) goto err_unpin; - ret = req->engine->emit_bb_start(req, - so->batch_offset, so->batch_size, - I915_DISPATCH_SECURE); - if (ret) + err = engine->emit_bb_start(rq, + so.batch_offset, so.batch_size, + I915_DISPATCH_SECURE); + if (err) goto err_unpin; - if (so->aux_size > 8) { - ret = req->engine->emit_bb_start(req, - so->aux_offset, so->aux_size, - I915_DISPATCH_SECURE); - if (ret) + if (so.aux_size > 8) { + err = engine->emit_bb_start(rq, + so.aux_offset, so.aux_size, + I915_DISPATCH_SECURE); + if (err) goto err_unpin; } - i915_vma_move_to_active(so->vma, req, 0); + i915_vma_move_to_active(so.vma, rq, 0); err_unpin: - i915_vma_unpin(so->vma); - return ret; -} - -void i915_gem_render_state_fini(struct intel_engine_cs *engine) -{ - struct intel_render_state *so; - struct drm_i915_gem_object *obj; - - so = fetch_and_zero(&engine->render_state); - if (!so) - return; - - obj = so->vma->obj; - - i915_vma_close(so->vma); - __i915_gem_object_release_unless_active(obj); - - kfree(so); + i915_vma_unpin(so.vma); +err_vma: + i915_vma_close(so.vma); +err_obj: + __i915_gem_object_release_unless_active(so.obj); + return err; } diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.h b/drivers/gpu/drm/i915/i915_gem_render_state.h index 87481845799d3..86369520482e4 100644 --- a/drivers/gpu/drm/i915/i915_gem_render_state.h +++ b/drivers/gpu/drm/i915/i915_gem_render_state.h @@ -26,8 +26,6 @@ struct drm_i915_gem_request; -int i915_gem_render_state_init(struct intel_engine_cs *engine); -int i915_gem_render_state_emit(struct drm_i915_gem_request *req); -void i915_gem_render_state_fini(struct intel_engine_cs *engine); +int i915_gem_render_state_emit(struct drm_i915_gem_request *rq); #endif /* _I915_GEM_RENDER_STATE_H_ */ diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 868c07a693b55..a0f9d0eb4bce9 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -641,21 +641,15 @@ int intel_engine_init_common(struct intel_engine_cs *engine) if (ret) goto err_unpin_preempt; - ret = i915_gem_render_state_init(engine); - if (ret) - goto err_breadcrumbs; - if (HWS_NEEDS_PHYSICAL(engine->i915)) ret = init_phys_status_page(engine); else ret = init_status_page(engine); if (ret) - goto err_rs_fini; + goto err_breadcrumbs; return 0; -err_rs_fini: - i915_gem_render_state_fini(engine); err_breadcrumbs: intel_engine_fini_breadcrumbs(engine); err_unpin_preempt: @@ -682,7 +676,6 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) else cleanup_status_page(engine); - i915_gem_render_state_fini(engine); intel_engine_fini_breadcrumbs(engine); intel_engine_cleanup_cmd_parser(engine); i915_gem_batch_pool_fini(&engine->batch_pool); diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 0c93f27f36eef..58d050a9a8665 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -136,6 +136,7 @@ #include #include #include "i915_drv.h" +#include "i915_gem_render_state.h" #include "intel_mocs.h" #define RING_EXECLIST_QFULL (1 << 0x2) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 464dc58af27b8..3321b801e77d7 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -28,9 +28,12 @@ */ #include + #include -#include "i915_drv.h" #include + +#include "i915_drv.h" +#include "i915_gem_render_state.h" #include "i915_trace.h" #include "intel_drv.h" diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 3372228591669..ef22c994038bb 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -165,7 +165,6 @@ struct i915_ctx_workarounds { }; struct drm_i915_gem_request; -struct intel_render_state; /* * Engine IDs definitions. @@ -307,7 +306,6 @@ struct intel_engine_cs { struct intel_timeline *timeline; struct drm_i915_gem_object *default_state; - struct intel_render_state *render_state; atomic_t irq_count; unsigned long irq_posted; -- GitLab From a0a8b1cf93383592fdfa86588beb8dcc4e32f036 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Nov 2017 14:06:44 +0000 Subject: [PATCH 0197/1471] drm/i915: Kerneldoc typo s/rps/rps_client/ Update the kerneldoc parameter name to match the real parameter name. Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171109140644.10805-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 4bf1183040864..a2e5a54128c10 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -538,7 +538,7 @@ i915_gem_object_wait_priority(struct drm_i915_gem_object *obj, * @obj: i915 gem object * @flags: how to wait (under a lock, for all rendering or just for writes etc) * @timeout: how long to wait - * @rps: client (user process) to charge for any waitboosting + * @rps_client: client (user process) to charge for any waitboosting */ int i915_gem_object_wait(struct drm_i915_gem_object *obj, -- GitLab From 9c52d1c816baa5b8c97485b20d95af29c98d26ee Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 10 Nov 2017 23:24:47 +0000 Subject: [PATCH 0198/1471] drm/i915/selftests: Yet another forgotten mock_i915->mm initialiser Move all of the i915->mm initialisation to a private function that can be reused by the mock i915 device to save forgetting any more steps. For example, <7>[ 1542.046332] [IGT] drv_selftest: starting subtest mock_objects <4>[ 1542.123924] Setting dangerous option mock_selftests - tainting kernel <6>[ 1542.167941] i915: Performing mock selftests with st_random_seed=0x246f5ab5 st_timeout=1000 <4>[ 1542.178012] INFO: trying to register non-static key. <4>[ 1542.178027] the code is fine but needs lockdep annotation. <4>[ 1542.178032] turning off the locking correctness validator. <4>[ 1542.178041] CPU: 3 PID: 6008 Comm: kworker/3:7 Tainted: G U 4.14.0-rc8-CI-CI_DRM_3332+ #1 <4>[ 1542.178049] Hardware name: /NUC6CAYB, BIOS AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017 <4>[ 1542.178144] Workqueue: events __i915_gem_free_work [i915] <4>[ 1542.178152] Call Trace: <4>[ 1542.178163] dump_stack+0x68/0x9f <4>[ 1542.178170] register_lock_class+0x3fd/0x580 <4>[ 1542.178177] ? unwind_next_frame+0x14/0x20 <4>[ 1542.178184] ? __save_stack_trace+0x73/0xd0 <4>[ 1542.178191] __lock_acquire+0xa4/0x1b00 <4>[ 1542.178254] ? __i915_gem_free_work+0x28/0xa0 [i915] <4>[ 1542.178261] ? __lock_acquire+0x4ab/0x1b00 <4>[ 1542.178268] lock_acquire+0xb0/0x200 <4>[ 1542.178273] ? lock_acquire+0xb0/0x200 <4>[ 1542.178336] ? __i915_gem_free_work+0x28/0xa0 [i915] <4>[ 1542.178344] _raw_spin_lock+0x32/0x50 <4>[ 1542.178405] ? __i915_gem_free_work+0x28/0xa0 [i915] <4>[ 1542.178468] __i915_gem_free_work+0x28/0xa0 [i915] <4>[ 1542.178476] process_one_work+0x221/0x650 <4>[ 1542.178483] worker_thread+0x4e/0x3c0 <4>[ 1542.178489] kthread+0x114/0x150 <4>[ 1542.178494] ? process_one_work+0x650/0x650 <4>[ 1542.178499] ? kthread_create_on_node+0x40/0x40 <4>[ 1542.178506] ret_from_fork+0x27/0x40 v2: Fish out i915->mm.object_stat_lock which was being inited over in i915_drv.c (Matthew) Signed-off-by: Chris Wilson Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110232447.21618-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_drv.c | 1 - drivers/gpu/drm/i915/i915_gem.c | 26 ++++++++++++------- .../gpu/drm/i915/selftests/mock_gem_device.c | 8 +----- 3 files changed, 18 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index d97fe9c9439a4..f8600cc96f2af 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -894,7 +894,6 @@ static int i915_driver_init_early(struct drm_i915_private *dev_priv, mutex_init(&dev_priv->backlight_lock); spin_lock_init(&dev_priv->uncore.lock); - spin_lock_init(&dev_priv->mm.object_stat_lock); mutex_init(&dev_priv->sb_lock); mutex_init(&dev_priv->modeset_restore_lock); mutex_init(&dev_priv->av_mutex); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index a2e5a54128c10..82e6e69c222cb 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5218,6 +5218,22 @@ i915_gem_load_init_fences(struct drm_i915_private *dev_priv) i915_gem_detect_bit_6_swizzle(dev_priv); } +static void i915_gem_init__mm(struct drm_i915_private *i915) +{ + spin_lock_init(&i915->mm.object_stat_lock); + spin_lock_init(&i915->mm.obj_lock); + spin_lock_init(&i915->mm.free_lock); + + init_llist_head(&i915->mm.free_list); + + INIT_LIST_HEAD(&i915->mm.unbound_list); + INIT_LIST_HEAD(&i915->mm.bound_list); + INIT_LIST_HEAD(&i915->mm.fence_list); + INIT_LIST_HEAD(&i915->mm.userfault_list); + + INIT_WORK(&i915->mm.free_work, __i915_gem_free_work); +} + int i915_gem_load_init(struct drm_i915_private *dev_priv) { @@ -5259,15 +5275,7 @@ i915_gem_load_init(struct drm_i915_private *dev_priv) if (err) goto err_priorities; - INIT_WORK(&dev_priv->mm.free_work, __i915_gem_free_work); - - spin_lock_init(&dev_priv->mm.obj_lock); - spin_lock_init(&dev_priv->mm.free_lock); - init_llist_head(&dev_priv->mm.free_list); - INIT_LIST_HEAD(&dev_priv->mm.unbound_list); - INIT_LIST_HEAD(&dev_priv->mm.bound_list); - INIT_LIST_HEAD(&dev_priv->mm.fence_list); - INIT_LIST_HEAD(&dev_priv->mm.userfault_list); + i915_gem_init__mm(dev_priv); INIT_DELAYED_WORK(&dev_priv->gt.retire_work, i915_gem_retire_work_handler); diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index d2bf2729c3314..80f152aaedf9b 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -179,9 +179,8 @@ struct drm_i915_private *mock_gem_device(void) I915_GTT_PAGE_SIZE_64K | I915_GTT_PAGE_SIZE_2M; - spin_lock_init(&i915->mm.object_stat_lock); - spin_lock_init(&i915->mm.obj_lock); mock_uncore_init(i915); + i915_gem_init__mm(i915); init_waitqueue_head(&i915->gpu_error.wait_queue); init_waitqueue_head(&i915->gpu_error.reset_queue); @@ -190,11 +189,6 @@ struct drm_i915_private *mock_gem_device(void) if (!i915->wq) goto put_device; - INIT_WORK(&i915->mm.free_work, __i915_gem_free_work); - init_llist_head(&i915->mm.free_list); - INIT_LIST_HEAD(&i915->mm.unbound_list); - INIT_LIST_HEAD(&i915->mm.bound_list); - mock_init_contexts(i915); INIT_DELAYED_WORK(&i915->gt.retire_work, mock_retire_work_handler); -- GitLab From bf38b0550359ee7a78244d17ed49d3c557cceb01 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:37 +0100 Subject: [PATCH 0199/1471] drm/vma-manager: drm_vma_node_start() constify argument MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Constify argument so functions calling into this take a const argument. Reviewed-by: Ville Syrjälä Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-2-noralf@tronnes.org --- include/drm/drm_vma_manager.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/drm/drm_vma_manager.h b/include/drm/drm_vma_manager.h index d84d52f6d2b11..8758df94e9a0f 100644 --- a/include/drm/drm_vma_manager.h +++ b/include/drm/drm_vma_manager.h @@ -152,7 +152,7 @@ static inline void drm_vma_node_reset(struct drm_vma_offset_node *node) * Start address of @node for page-based addressing. 0 if the node does not * have an offset allocated. */ -static inline unsigned long drm_vma_node_start(struct drm_vma_offset_node *node) +static inline unsigned long drm_vma_node_start(const struct drm_vma_offset_node *node) { return node->vm_node.start; } -- GitLab From 6ff1086e21e5b1d6239e2c47267d21dc19b6e801 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:38 +0100 Subject: [PATCH 0200/1471] drm/framebuffer: drm_framebuffer_read_refcount() constify argument MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Constify argument so functions calling into this take a const argument. Reviewed-by: Ville Syrjälä Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-3-noralf@tronnes.org --- include/drm/drm_framebuffer.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/drm/drm_framebuffer.h b/include/drm/drm_framebuffer.h index b6996ddb19d6e..6cce22e1a0f27 100644 --- a/include/drm/drm_framebuffer.h +++ b/include/drm/drm_framebuffer.h @@ -263,7 +263,7 @@ static inline void drm_framebuffer_unreference(struct drm_framebuffer *fb) * * This functions returns the framebuffer's reference count. */ -static inline uint32_t drm_framebuffer_read_refcount(struct drm_framebuffer *fb) +static inline uint32_t drm_framebuffer_read_refcount(const struct drm_framebuffer *fb) { return kref_read(&fb->base.refcount); } -- GitLab From bf6234a294c5b8bcc6f47ecdb646f41ad5efddd3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:39 +0100 Subject: [PATCH 0201/1471] drm/print: Add drm_printf_indent() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add drm_printf_indent() that adds tab indentation according to argument. Indentation overflow is marked with an X. Reviewed-by: Daniel Vetter Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-4-noralf@tronnes.org --- include/drm/drm_print.h | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h index 7b9c86a6ca3e8..7dbfdebec973e 100644 --- a/include/drm/drm_print.h +++ b/include/drm/drm_print.h @@ -80,6 +80,14 @@ void __drm_printfn_debug(struct drm_printer *p, struct va_format *vaf); __printf(2, 3) void drm_printf(struct drm_printer *p, const char *f, ...); +/** + * drm_printf_indent - Print to a &drm_printer stream with indentation + * @printer: DRM printer + * @indent: Tab indentation level (max 5) + * @fmt: Format string + */ +#define drm_printf_indent(printer, indent, fmt, ...) \ + drm_printf((printer), "%.*s" fmt, (indent), "\t\t\t\t\tX", ##__VA_ARGS__) /** * drm_seq_file_printer - construct a &drm_printer that outputs to &seq_file -- GitLab From 45d58b40292b16ab847497dcd299e315a2ad7956 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:40 +0100 Subject: [PATCH 0202/1471] drm/framebuffer: Add framebuffer debugfs file MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add debugfs file that dumps info about the framebuffers and its planes. Also dump info about any connected gem object(s). Reviewed-by: Daniel Vetter Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-5-noralf@tronnes.org --- drivers/gpu/drm/drm_debugfs.c | 6 ++++ drivers/gpu/drm/drm_framebuffer.c | 59 +++++++++++++++++++++++++++++++ drivers/gpu/drm/drm_gem.c | 17 +++++++++ drivers/gpu/drm/drm_internal.h | 7 ++++ include/drm/drm_drv.h | 15 ++++++++ 5 files changed, 104 insertions(+) diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c index c1807d5754b2a..550f29de6c1fc 100644 --- a/drivers/gpu/drm/drm_debugfs.c +++ b/drivers/gpu/drm/drm_debugfs.c @@ -158,6 +158,12 @@ int drm_debugfs_init(struct drm_minor *minor, int minor_id, } } + ret = drm_framebuffer_debugfs_init(minor); + if (ret) { + DRM_ERROR("Failed to create framebuffer debugfs file\n"); + return ret; + } + if (dev->driver->debugfs_init) { ret = dev->driver->debugfs_init(minor); if (ret) { diff --git a/drivers/gpu/drm/drm_framebuffer.c b/drivers/gpu/drm/drm_framebuffer.c index 2305e7800b45a..0153000373ae8 100644 --- a/drivers/gpu/drm/drm_framebuffer.c +++ b/drivers/gpu/drm/drm_framebuffer.c @@ -25,7 +25,9 @@ #include #include #include +#include +#include "drm_internal.h" #include "drm_crtc_internal.h" /** @@ -967,3 +969,60 @@ int drm_framebuffer_plane_height(int height, return fb_plane_height(height, fb->format, plane); } EXPORT_SYMBOL(drm_framebuffer_plane_height); + +void drm_framebuffer_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_framebuffer *fb) +{ + struct drm_format_name_buf format_name; + unsigned int i; + + drm_printf_indent(p, indent, "refcount=%u\n", + drm_framebuffer_read_refcount(fb)); + drm_printf_indent(p, indent, "format=%s\n", + drm_get_format_name(fb->format->format, &format_name)); + drm_printf_indent(p, indent, "modifier=0x%llx\n", fb->modifier); + drm_printf_indent(p, indent, "size=%ux%u\n", fb->width, fb->height); + drm_printf_indent(p, indent, "layers:\n"); + + for (i = 0; i < fb->format->num_planes; i++) { + drm_printf_indent(p, indent + 1, "size[%u]=%dx%d\n", i, + drm_framebuffer_plane_width(fb->width, fb, i), + drm_framebuffer_plane_height(fb->height, fb, i)); + drm_printf_indent(p, indent + 1, "pitch[%u]=%u\n", i, fb->pitches[i]); + drm_printf_indent(p, indent + 1, "offset[%u]=%u\n", i, fb->offsets[i]); + drm_printf_indent(p, indent + 1, "obj[%u]:%s\n", i, + fb->obj[i] ? "" : "(null)"); + if (fb->obj[i]) + drm_gem_print_info(p, indent + 2, fb->obj[i]); + } +} + +#ifdef CONFIG_DEBUG_FS +static int drm_framebuffer_info(struct seq_file *m, void *data) +{ + struct drm_info_node *node = m->private; + struct drm_device *dev = node->minor->dev; + struct drm_printer p = drm_seq_file_printer(m); + struct drm_framebuffer *fb; + + mutex_lock(&dev->mode_config.fb_lock); + drm_for_each_fb(fb, dev) { + drm_printf(&p, "framebuffer[%u]:\n", fb->base.id); + drm_framebuffer_print_info(&p, 1, fb); + } + mutex_unlock(&dev->mode_config.fb_lock); + + return 0; +} + +static const struct drm_info_list drm_framebuffer_debugfs_list[] = { + { "framebuffer", drm_framebuffer_info, 0 }, +}; + +int drm_framebuffer_debugfs_init(struct drm_minor *minor) +{ + return drm_debugfs_create_files(drm_framebuffer_debugfs_list, + ARRAY_SIZE(drm_framebuffer_debugfs_list), + minor->debugfs_root, minor); +} +#endif diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 4c84b23d37ccf..01f8d9481211f 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -40,6 +40,7 @@ #include #include #include +#include #include "drm_internal.h" /** @file drm_gem.c @@ -1040,3 +1041,19 @@ int drm_gem_mmap(struct file *filp, struct vm_area_struct *vma) return ret; } EXPORT_SYMBOL(drm_gem_mmap); + +void drm_gem_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_gem_object *obj) +{ + drm_printf_indent(p, indent, "name=%d\n", obj->name); + drm_printf_indent(p, indent, "refcount=%u\n", + kref_read(&obj->refcount)); + drm_printf_indent(p, indent, "start=%08lx\n", + drm_vma_node_start(&obj->vma_node)); + drm_printf_indent(p, indent, "size=%zu\n", obj->size); + drm_printf_indent(p, indent, "imported=%s\n", + obj->import_attach ? "yes" : "no"); + + if (obj->dev->driver->gem_print_info) + obj->dev->driver->gem_print_info(p, indent, obj); +} diff --git a/drivers/gpu/drm/drm_internal.h b/drivers/gpu/drm/drm_internal.h index fbc3f308fa199..430ce3fe4f3b1 100644 --- a/drivers/gpu/drm/drm_internal.h +++ b/drivers/gpu/drm/drm_internal.h @@ -106,6 +106,8 @@ int drm_gem_open_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv); void drm_gem_open(struct drm_device *dev, struct drm_file *file_private); void drm_gem_release(struct drm_device *dev, struct drm_file *file_private); +void drm_gem_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_gem_object *obj); /* drm_debugfs.c drm_debugfs_crc.c */ #if defined(CONFIG_DEBUG_FS) @@ -173,3 +175,8 @@ int drm_syncobj_reset_ioctl(struct drm_device *dev, void *data, struct drm_file *file_private); int drm_syncobj_signal_ioctl(struct drm_device *dev, void *data, struct drm_file *file_private); + +/* drm_framebuffer.c */ +void drm_framebuffer_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_framebuffer *fb); +int drm_framebuffer_debugfs_init(struct drm_minor *minor); diff --git a/include/drm/drm_drv.h b/include/drm/drm_drv.h index 0e90ef24214b9..1536e5bddeaf0 100644 --- a/include/drm/drm_drv.h +++ b/include/drm/drm_drv.h @@ -39,6 +39,7 @@ struct drm_minor; struct dma_buf_attachment; struct drm_display_mode; struct drm_mode_create_dumb; +struct drm_printer; /* driver capabilities and requirements mask */ #define DRIVER_USE_AGP 0x1 @@ -428,6 +429,20 @@ struct drm_driver { */ void (*gem_close_object) (struct drm_gem_object *, struct drm_file *); + /** + * @gem_print_info: + * + * If driver subclasses struct &drm_gem_object, it can implement this + * optional hook for printing additional driver specific info. + * + * drm_printf_indent() should be used in the callback passing it the + * indent argument. + * + * This callback is called from drm_gem_print_info(). + */ + void (*gem_print_info)(struct drm_printer *p, unsigned int indent, + const struct drm_gem_object *obj); + /** * @gem_create_object: constructor for gem objects * -- GitLab From f02b604b07d5314d01968eda4e199e6c46352a6d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:41 +0100 Subject: [PATCH 0203/1471] drm/atomic: Use drm_framebuffer_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use drm_framebuffer_print_info() to print framebuffer info in drm_atomic_plane_print_state(). This will give optional GEM info as well. Reviewed-by: Daniel Vetter Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-6-noralf@tronnes.org --- drivers/gpu/drm/drm_atomic.c | 18 +++--------------- 1 file changed, 3 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c index fd96c3dad2fbd..bf8ca515c9588 100644 --- a/drivers/gpu/drm/drm_atomic.c +++ b/drivers/gpu/drm/drm_atomic.c @@ -33,6 +33,7 @@ #include #include "drm_crtc_internal.h" +#include "drm_internal.h" void __drm_crtc_commit_free(struct kref *kref) { @@ -935,21 +936,8 @@ static void drm_atomic_plane_print_state(struct drm_printer *p, drm_printf(p, "plane[%u]: %s\n", plane->base.id, plane->name); drm_printf(p, "\tcrtc=%s\n", state->crtc ? state->crtc->name : "(null)"); drm_printf(p, "\tfb=%u\n", state->fb ? state->fb->base.id : 0); - if (state->fb) { - struct drm_framebuffer *fb = state->fb; - int i, n = fb->format->num_planes; - struct drm_format_name_buf format_name; - - drm_printf(p, "\t\tformat=%s\n", - drm_get_format_name(fb->format->format, &format_name)); - drm_printf(p, "\t\t\tmodifier=0x%llx\n", fb->modifier); - drm_printf(p, "\t\tsize=%dx%d\n", fb->width, fb->height); - drm_printf(p, "\t\tlayers:\n"); - for (i = 0; i < n; i++) { - drm_printf(p, "\t\t\tpitch[%d]=%u\n", i, fb->pitches[i]); - drm_printf(p, "\t\t\toffset[%d]=%u\n", i, fb->offsets[i]); - } - } + if (state->fb) + drm_framebuffer_print_info(p, 2, state->fb); drm_printf(p, "\tcrtc-pos=" DRM_RECT_FMT "\n", DRM_RECT_ARG(&dest)); drm_printf(p, "\tsrc-pos=" DRM_RECT_FP_FMT "\n", DRM_RECT_FP_ARG(&src)); drm_printf(p, "\trotation=%x\n", state->rotation); -- GitLab From 8d25ccebef55b8fc47b674f24ccdccbef08e2f32 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:42 +0100 Subject: [PATCH 0204/1471] drm/cma-helper: Turn to_drm_gem_cma_obj() into a macro MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This allows the argument to be a const. The other option was to keep it an inline function and make the argument a const: static inline struct drm_gem_cma_object * to_drm_gem_cma_obj(const struct drm_gem_object *gem_obj) { return container_of(gem_obj, struct drm_gem_cma_object, base); } This will happily return a non-const pointer to the drm_gem_cma_object based on a const pointer to the contained drm_gem_object, thus creating const-safety problems. There was an attempt to fix the problem in the container_of() macro itself (see https://lkml.org/lkml/2017/5/19/381) but the patch seems to have fallen through the cracks. It would require turning this inline function into a macro. By making this a macro now, we will benefit from a possible future enhancement of container_of(). We don't loose type checking by doing this, container_of() takes care of that. Suggested-by: Laurent Pinchart Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-7-noralf@tronnes.org --- include/drm/drm_gem_cma_helper.h | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/include/drm/drm_gem_cma_helper.h b/include/drm/drm_gem_cma_helper.h index 58a739bf15f1f..7a3dcf0cf2896 100644 --- a/include/drm/drm_gem_cma_helper.h +++ b/include/drm/drm_gem_cma_helper.h @@ -20,11 +20,8 @@ struct drm_gem_cma_object { void *vaddr; }; -static inline struct drm_gem_cma_object * -to_drm_gem_cma_obj(struct drm_gem_object *gem_obj) -{ - return container_of(gem_obj, struct drm_gem_cma_object, base); -} +#define to_drm_gem_cma_obj(gem_obj) \ + container_of(gem_obj, struct drm_gem_cma_object, base) #ifndef CONFIG_MMU #define DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ -- GitLab From d68920120d712136cc1538e5f4f8e119f2d4f43a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:43 +0100 Subject: [PATCH 0205/1471] drm/cma-helper: Add drm_gem_cma_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add drm_gem_cma_print_info() for debugfs printing struct drm_gem_cma_object specific info. Reviewed-by: Daniel Vetter Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-8-noralf@tronnes.org --- drivers/gpu/drm/drm_gem_cma_helper.c | 19 +++++++++++++++++++ include/drm/drm_gem_cma_helper.h | 3 +++ 2 files changed, 22 insertions(+) diff --git a/drivers/gpu/drm/drm_gem_cma_helper.c b/drivers/gpu/drm/drm_gem_cma_helper.c index 020e7668dfaba..29f7b0fd3ad4f 100644 --- a/drivers/gpu/drm/drm_gem_cma_helper.c +++ b/drivers/gpu/drm/drm_gem_cma_helper.c @@ -423,6 +423,25 @@ void drm_gem_cma_describe(struct drm_gem_cma_object *cma_obj, EXPORT_SYMBOL_GPL(drm_gem_cma_describe); #endif +/** + * drm_gem_cma_print_info() - Print &drm_gem_cma_object info for debugfs + * @p: DRM printer + * @indent: Tab indentation level + * @gem: GEM object + * + * This function can be used as the &drm_driver->gem_print_info callback. + * It prints paddr and vaddr for use in e.g. debugfs output. + */ +void drm_gem_cma_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_gem_object *obj) +{ + const struct drm_gem_cma_object *cma_obj = to_drm_gem_cma_obj(obj); + + drm_printf_indent(p, indent, "paddr=%pad\n", &cma_obj->paddr); + drm_printf_indent(p, indent, "vaddr=%p\n", cma_obj->vaddr); +} +EXPORT_SYMBOL(drm_gem_cma_print_info); + /** * drm_gem_cma_prime_get_sg_table - provide a scatter/gather table of pinned * pages for a CMA GEM object diff --git a/include/drm/drm_gem_cma_helper.h b/include/drm/drm_gem_cma_helper.h index 7a3dcf0cf2896..40435f3480e5d 100644 --- a/include/drm/drm_gem_cma_helper.h +++ b/include/drm/drm_gem_cma_helper.h @@ -91,6 +91,9 @@ unsigned long drm_gem_cma_get_unmapped_area(struct file *filp, void drm_gem_cma_describe(struct drm_gem_cma_object *obj, struct seq_file *m); #endif +void drm_gem_cma_print_info(struct drm_printer *p, unsigned int indent, + const struct drm_gem_object *obj); + struct sg_table *drm_gem_cma_prime_get_sg_table(struct drm_gem_object *obj); struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, -- GitLab From 80c7f714c1f720c77d9cc381adf9e1cd42351433 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:44 +0100 Subject: [PATCH 0206/1471] drm/arc: Use drm_gem_cma_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is a new core debugfs file that prints fb/gem info: /dri//framebuffer Use drm_gem_cma_print_info() to provide info to that output instead of using drm_fb_cma_debugfs_show(). Cc: Alexey Brodkin Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Acked-by: Alexey Brodkin Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-9-noralf@tronnes.org --- drivers/gpu/drm/arc/arcpgu_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/arc/arcpgu_drv.c b/drivers/gpu/drm/arc/arcpgu_drv.c index 074fd4ea7ece4..f067de4e1e829 100644 --- a/drivers/gpu/drm/arc/arcpgu_drv.c +++ b/drivers/gpu/drm/arc/arcpgu_drv.c @@ -155,7 +155,6 @@ static int arcpgu_show_pxlclock(struct seq_file *m, void *arg) static struct drm_info_list arcpgu_debugfs_list[] = { { "clocks", arcpgu_show_pxlclock, 0 }, - { "fb", drm_fb_cma_debugfs_show, 0 }, }; static int arcpgu_debugfs_init(struct drm_minor *minor) @@ -180,6 +179,7 @@ static struct drm_driver arcpgu_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_free_object_unlocked = drm_gem_cma_free_object, + .gem_print_info = drm_gem_cma_print_info, .gem_vm_ops = &drm_gem_cma_vm_ops, .gem_prime_export = drm_gem_prime_export, .gem_prime_import = drm_gem_prime_import, -- GitLab From b92c44fa239ddb6fe86fcf97ad5b59c97d74f288 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:45 +0100 Subject: [PATCH 0207/1471] drm/arm/hdlcd: Use drm_gem_cma_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is a new core debugfs file that prints fb/gem info: /dri//framebuffer Use drm_gem_cma_print_info() to provide info to that output instead of using drm_fb_cma_debugfs_show(). Cc: Liviu Dudau Cc: Brian Starkey Signed-off-by: Noralf Trønnes Acked-by: Liviu Dudau Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-10-noralf@tronnes.org --- drivers/gpu/drm/arm/hdlcd_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/arm/hdlcd_drv.c b/drivers/gpu/drm/arm/hdlcd_drv.c index 764d0c83710ca..59b21bdc0c30d 100644 --- a/drivers/gpu/drm/arm/hdlcd_drv.c +++ b/drivers/gpu/drm/arm/hdlcd_drv.c @@ -230,7 +230,6 @@ static int hdlcd_show_pxlclock(struct seq_file *m, void *arg) static struct drm_info_list hdlcd_debugfs_list[] = { { "interrupt_count", hdlcd_show_underrun_count, 0 }, { "clocks", hdlcd_show_pxlclock, 0 }, - { "fb", drm_fb_cma_debugfs_show, 0 }, }; static int hdlcd_debugfs_init(struct drm_minor *minor) @@ -252,6 +251,7 @@ static struct drm_driver hdlcd_driver = { .irq_postinstall = hdlcd_irq_postinstall, .irq_uninstall = hdlcd_irq_uninstall, .gem_free_object_unlocked = drm_gem_cma_free_object, + .gem_print_info = drm_gem_cma_print_info, .gem_vm_ops = &drm_gem_cma_vm_ops, .dumb_create = drm_gem_cma_dumb_create, .prime_handle_to_fd = drm_gem_prime_handle_to_fd, -- GitLab From fbf65b7efe4b8ced3c597512ad8fa71062233d2a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:46 +0100 Subject: [PATCH 0208/1471] drm/tilcdc: Use drm_gem_cma_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is a new core debugfs file that prints fb/gem info: /dri//framebuffer Use drm_gem_cma_print_info() to provide info to that output instead of using drm_fb_cma_debugfs_show(). Cc: Jyri Sarha Cc: Tomi Valkeinen Signed-off-by: Noralf Trønnes Acked-by: Jyri Sarha Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-11-noralf@tronnes.org --- drivers/gpu/drm/tilcdc/tilcdc_drv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/tilcdc/tilcdc_drv.c b/drivers/gpu/drm/tilcdc/tilcdc_drv.c index 72ce063aa0d88..bc4feb3a84b9a 100644 --- a/drivers/gpu/drm/tilcdc/tilcdc_drv.c +++ b/drivers/gpu/drm/tilcdc/tilcdc_drv.c @@ -507,7 +507,6 @@ static int tilcdc_mm_show(struct seq_file *m, void *arg) static struct drm_info_list tilcdc_debugfs_list[] = { { "regs", tilcdc_regs_show, 0 }, { "mm", tilcdc_mm_show, 0 }, - { "fb", drm_fb_cma_debugfs_show, 0 }, }; static int tilcdc_debugfs_init(struct drm_minor *minor) @@ -541,6 +540,7 @@ static struct drm_driver tilcdc_driver = { .lastclose = tilcdc_lastclose, .irq_handler = tilcdc_irq, .gem_free_object_unlocked = drm_gem_cma_free_object, + .gem_print_info = drm_gem_cma_print_info, .gem_vm_ops = &drm_gem_cma_vm_ops, .dumb_create = drm_gem_cma_dumb_create, -- GitLab From beed8313be00c26cc7130ce1878b1ead3bf36511 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:47 +0100 Subject: [PATCH 0209/1471] drm/tinydrm: Use drm_gem_cma_print_info() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is a new core debugfs file that prints fb/gem info: /dri//framebuffer Use drm_gem_cma_print_info() to provide info to that output instead of using drm_fb_cma_debugfs_show(). Reviewed-by: Daniel Vetter Signed-off-by: Noralf Trønnes Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-12-noralf@tronnes.org --- drivers/gpu/drm/tinydrm/mipi-dbi.c | 8 +------- include/drm/tinydrm/tinydrm.h | 1 + 2 files changed, 2 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/tinydrm/mipi-dbi.c b/drivers/gpu/drm/tinydrm/mipi-dbi.c index d43e992ab432f..347f9b226f267 100644 --- a/drivers/gpu/drm/tinydrm/mipi-dbi.c +++ b/drivers/gpu/drm/tinydrm/mipi-dbi.c @@ -961,10 +961,6 @@ static const struct file_operations mipi_dbi_debugfs_command_fops = { .write = mipi_dbi_debugfs_command_write, }; -static const struct drm_info_list mipi_dbi_debugfs_list[] = { - { "fb", drm_fb_cma_debugfs_show, 0 }, -}; - /** * mipi_dbi_debugfs_init - Create debugfs entries * @minor: DRM minor @@ -987,9 +983,7 @@ int mipi_dbi_debugfs_init(struct drm_minor *minor) debugfs_create_file("command", mode, minor->debugfs_root, mipi, &mipi_dbi_debugfs_command_fops); - return drm_debugfs_create_files(mipi_dbi_debugfs_list, - ARRAY_SIZE(mipi_dbi_debugfs_list), - minor->debugfs_root, minor); + return 0; } EXPORT_SYMBOL(mipi_dbi_debugfs_init); diff --git a/include/drm/tinydrm/tinydrm.h b/include/drm/tinydrm/tinydrm.h index 4774fe3d42739..423828922e5ab 100644 --- a/include/drm/tinydrm/tinydrm.h +++ b/include/drm/tinydrm/tinydrm.h @@ -46,6 +46,7 @@ pipe_to_tinydrm(struct drm_simple_display_pipe *pipe) */ #define TINYDRM_GEM_DRIVER_OPS \ .gem_free_object = tinydrm_gem_cma_free_object, \ + .gem_print_info = drm_gem_cma_print_info, \ .gem_vm_ops = &drm_gem_cma_vm_ops, \ .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ -- GitLab From b5e821bb86071bb37624bafe9e4e8f345be89f45 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 7 Nov 2017 20:13:48 +0100 Subject: [PATCH 0210/1471] drm/cma-helper: Remove drm_fb_cma_debugfs_show() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_fb_cma_debugfs_show() and drm_gem_cma_describe() are superseded by drm_framebuffer_debugfs_init() and drm_gem_cma_print_info(). Cc: Laurent Pinchart Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Reviewed-by: Laurent Pinchart Link: https://patchwork.freedesktop.org/patch/msgid/20171107191348.17555-13-noralf@tronnes.org --- drivers/gpu/drm/drm_fb_cma_helper.c | 37 ---------------------------- drivers/gpu/drm/drm_gem_cma_helper.c | 26 ------------------- include/drm/drm_fb_cma_helper.h | 6 ----- include/drm/drm_gem_cma_helper.h | 4 --- 4 files changed, 73 deletions(-) diff --git a/drivers/gpu/drm/drm_fb_cma_helper.c b/drivers/gpu/drm/drm_fb_cma_helper.c index 0e3c14174d08d..35b56dfba9294 100644 --- a/drivers/gpu/drm/drm_fb_cma_helper.c +++ b/drivers/gpu/drm/drm_fb_cma_helper.c @@ -130,43 +130,6 @@ dma_addr_t drm_fb_cma_get_gem_addr(struct drm_framebuffer *fb, } EXPORT_SYMBOL_GPL(drm_fb_cma_get_gem_addr); -#ifdef CONFIG_DEBUG_FS -static void drm_fb_cma_describe(struct drm_framebuffer *fb, struct seq_file *m) -{ - int i; - - seq_printf(m, "fb: %dx%d@%4.4s\n", fb->width, fb->height, - (char *)&fb->format->format); - - for (i = 0; i < fb->format->num_planes; i++) { - seq_printf(m, " %d: offset=%d pitch=%d, obj: ", - i, fb->offsets[i], fb->pitches[i]); - drm_gem_cma_describe(drm_fb_cma_get_gem_obj(fb, i), m); - } -} - -/** - * drm_fb_cma_debugfs_show() - Helper to list CMA framebuffer objects - * in debugfs. - * @m: output file - * @arg: private data for the callback - */ -int drm_fb_cma_debugfs_show(struct seq_file *m, void *arg) -{ - struct drm_info_node *node = (struct drm_info_node *) m->private; - struct drm_device *dev = node->minor->dev; - struct drm_framebuffer *fb; - - mutex_lock(&dev->mode_config.fb_lock); - drm_for_each_fb(fb, dev) - drm_fb_cma_describe(fb, m); - mutex_unlock(&dev->mode_config.fb_lock); - - return 0; -} -EXPORT_SYMBOL_GPL(drm_fb_cma_debugfs_show); -#endif - static int drm_fb_cma_mmap(struct fb_info *info, struct vm_area_struct *vma) { return dma_mmap_writecombine(info->device, vma, info->screen_base, diff --git a/drivers/gpu/drm/drm_gem_cma_helper.c b/drivers/gpu/drm/drm_gem_cma_helper.c index 29f7b0fd3ad4f..88ad88719605d 100644 --- a/drivers/gpu/drm/drm_gem_cma_helper.c +++ b/drivers/gpu/drm/drm_gem_cma_helper.c @@ -397,32 +397,6 @@ unsigned long drm_gem_cma_get_unmapped_area(struct file *filp, EXPORT_SYMBOL_GPL(drm_gem_cma_get_unmapped_area); #endif -#ifdef CONFIG_DEBUG_FS -/** - * drm_gem_cma_describe - describe a CMA GEM object for debugfs - * @cma_obj: CMA GEM object - * @m: debugfs file handle - * - * This function can be used to dump a human-readable representation of the - * CMA GEM object into a synthetic file. - */ -void drm_gem_cma_describe(struct drm_gem_cma_object *cma_obj, - struct seq_file *m) -{ - struct drm_gem_object *obj = &cma_obj->base; - uint64_t off; - - off = drm_vma_node_start(&obj->vma_node); - - seq_printf(m, "%2d (%2d) %08llx %pad %p %zu", - obj->name, kref_read(&obj->refcount), - off, &cma_obj->paddr, cma_obj->vaddr, obj->size); - - seq_printf(m, "\n"); -} -EXPORT_SYMBOL_GPL(drm_gem_cma_describe); -#endif - /** * drm_gem_cma_print_info() - Print &drm_gem_cma_object info for debugfs * @p: DRM printer diff --git a/include/drm/drm_fb_cma_helper.h b/include/drm/drm_fb_cma_helper.h index 023f052a58731..a613ff022e6cf 100644 --- a/include/drm/drm_fb_cma_helper.h +++ b/include/drm/drm_fb_cma_helper.h @@ -35,11 +35,5 @@ dma_addr_t drm_fb_cma_get_gem_addr(struct drm_framebuffer *fb, struct drm_plane_state *state, unsigned int plane); -#ifdef CONFIG_DEBUG_FS -struct seq_file; - -int drm_fb_cma_debugfs_show(struct seq_file *m, void *arg); -#endif - #endif diff --git a/include/drm/drm_gem_cma_helper.h b/include/drm/drm_gem_cma_helper.h index 40435f3480e5d..76ded75f02cca 100644 --- a/include/drm/drm_gem_cma_helper.h +++ b/include/drm/drm_gem_cma_helper.h @@ -87,10 +87,6 @@ unsigned long drm_gem_cma_get_unmapped_area(struct file *filp, unsigned long flags); #endif -#ifdef CONFIG_DEBUG_FS -void drm_gem_cma_describe(struct drm_gem_cma_object *obj, struct seq_file *m); -#endif - void drm_gem_cma_print_info(struct drm_printer *p, unsigned int indent, const struct drm_gem_object *obj); -- GitLab From 556fe36d09da5f82879e92bafa0371b4b79f7d6f Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:53 +0100 Subject: [PATCH 0211/1471] drm/i915: Update watermark state correctly in sanitize_watermarks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We no longer use intel_crtc->wm.active for watermarks any more, which was incorrect. But this uncovered a bug in sanitize_watermarks(), which meant that we wrote the correct watermarks, but the next update would still use the wrong hw watermarks for calculating. This caused all further updates to fail with -EINVAL and the log would reveal an error like the one below: [ 10.043902] [drm:ilk_validate_wm_level.part.8 [i915]] Sprite WM0 too large 56 (max 0) [ 10.043960] [drm:ilk_validate_pipe_wm [i915]] LP0 watermark invalid [ 10.044030] [drm:intel_crtc_atomic_check [i915]] No valid intermediate pipe watermarks are possible Signed-off-by: Maarten Lankhorst Fixes: b6b178a77210 ("drm/i915: Calculate ironlake intermediate watermarks correctly, v2.") Cc: stable@vger.kernel.org #v4.8+ Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-1-maarten.lankhorst@linux.intel.com Signed-off-by: Maarten Lankhorst Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5c7540f3f5dcd..63c8b9ccb31e9 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -14443,6 +14443,8 @@ static void sanitize_watermarks(struct drm_device *dev) cs->wm.need_postvbl_update = true; dev_priv->display.optimize_watermarks(intel_state, cs); + + to_intel_crtc_state(crtc->state)->wm = cs->wm; } put_state: -- GitLab From 37cd33006d02d84e08caedea2db48662c18e499e Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 12 Nov 2017 11:27:38 +0000 Subject: [PATCH 0212/1471] drm/i915: Remove redundant intel_autoenable_gt_powersave() Now that we always execute a context switch upon module load, there is no need to queue a delayed task for doing so. The purpose of the delayed task is to enable GT powersaving, for which we need the HW state to be valid (i.e. having loaded a context and initialised basic state). We used to defer this operation as historically it was slow (due to slow register polling, fixed with commit 1758b90e38f5 ("drm/i915: Use a hybrid scheme for fast register waits")) but now we have a requirement to save the default HW state. v2: Load the kernel context (to provide the power context) upon resume. Signed-off-by: Chris Wilson Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171112112738.1463-3-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_drv.c | 11 +----- drivers/gpu/drm/i915/i915_drv.h | 1 - drivers/gpu/drm/i915/i915_gem.c | 31 +++++++++++---- drivers/gpu/drm/i915/intel_drv.h | 1 - drivers/gpu/drm/i915/intel_pm.c | 66 -------------------------------- 5 files changed, 24 insertions(+), 86 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index f8600cc96f2af..9df7b5d59a94b 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1683,8 +1683,6 @@ static int i915_drm_resume(struct drm_device *dev) intel_csr_ucode_resume(dev_priv); - i915_gem_resume(dev_priv); - i915_restore_state(dev_priv); intel_pps_unlock_regs_wa(dev_priv); intel_opregion_setup(dev_priv); @@ -1705,12 +1703,7 @@ static int i915_drm_resume(struct drm_device *dev) drm_mode_config_reset(dev); - mutex_lock(&dev->struct_mutex); - if (i915_gem_init_hw(dev_priv)) { - DRM_ERROR("failed to re-initialize GPU, declaring wedged!\n"); - i915_gem_set_wedged(dev_priv); - } - mutex_unlock(&dev->struct_mutex); + i915_gem_resume(dev_priv); intel_guc_resume(dev_priv); @@ -1745,8 +1738,6 @@ static int i915_drm_resume(struct drm_device *dev) intel_opregion_notify_adapter(dev_priv, PCI_D0); - intel_autoenable_gt_powersave(dev_priv); - enable_rpm_wakeref_asserts(dev_priv); return 0; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 40012b6daea27..07b9e78bbe966 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1392,7 +1392,6 @@ struct intel_gen6_power_mgmt { struct intel_rps rps; struct intel_rc6 rc6; struct intel_llc_pstate llc_pstate; - struct delayed_work autoenable_work; }; /* defined intel_pm.c */ diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 82e6e69c222cb..fa9e5130c9863 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4806,23 +4806,38 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) return ret; } -void i915_gem_resume(struct drm_i915_private *dev_priv) +void i915_gem_resume(struct drm_i915_private *i915) { - struct drm_device *dev = &dev_priv->drm; + WARN_ON(i915->gt.awake); - WARN_ON(dev_priv->gt.awake); + mutex_lock(&i915->drm.struct_mutex); + intel_uncore_forcewake_get(i915, FORCEWAKE_ALL); - mutex_lock(&dev->struct_mutex); - i915_gem_restore_gtt_mappings(dev_priv); - i915_gem_restore_fences(dev_priv); + i915_gem_restore_gtt_mappings(i915); + i915_gem_restore_fences(i915); /* As we didn't flush the kernel context before suspend, we cannot * guarantee that the context image is complete. So let's just reset * it and start again. */ - dev_priv->gt.resume(dev_priv); + i915->gt.resume(i915); - mutex_unlock(&dev->struct_mutex); + if (i915_gem_init_hw(i915)) + goto err_wedged; + + /* Always reload a context for powersaving. */ + if (i915_gem_switch_to_kernel_context(i915)) + goto err_wedged; + +out_unlock: + intel_uncore_forcewake_put(i915, FORCEWAKE_ALL); + mutex_unlock(&i915->drm.struct_mutex); + return; + +err_wedged: + DRM_ERROR("failed to re-initialize GPU, declaring wedged!\n"); + i915_gem_set_wedged(i915); + goto out_unlock; } void i915_gem_init_swizzling(struct drm_i915_private *dev_priv) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index a0a5cead868c9..10bec8796239b 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1880,7 +1880,6 @@ void intel_init_gt_powersave(struct drm_i915_private *dev_priv); void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv); void intel_sanitize_gt_powersave(struct drm_i915_private *dev_priv); void intel_enable_gt_powersave(struct drm_i915_private *dev_priv); -void intel_autoenable_gt_powersave(struct drm_i915_private *dev_priv); void intel_disable_gt_powersave(struct drm_i915_private *dev_priv); void intel_suspend_gt_powersave(struct drm_i915_private *dev_priv); void gen6_rps_busy(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 48a127ac6de53..c1a56809f1431 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7959,8 +7959,6 @@ void intel_init_gt_powersave(struct drm_i915_private *dev_priv) rps->boost_freq = rps->max_freq; mutex_unlock(&dev_priv->pcu_lock); - - intel_autoenable_gt_powersave(dev_priv); } void intel_cleanup_gt_powersave(struct drm_i915_private *dev_priv) @@ -7985,9 +7983,6 @@ void intel_suspend_gt_powersave(struct drm_i915_private *dev_priv) if (INTEL_GEN(dev_priv) < 6) return; - if (cancel_delayed_work_sync(&dev_priv->gt_pm.autoenable_work)) - intel_runtime_pm_put(dev_priv); - /* gen6_rps_idle() will be called later to disable interrupts */ } @@ -8146,65 +8141,6 @@ void intel_enable_gt_powersave(struct drm_i915_private *dev_priv) mutex_unlock(&dev_priv->pcu_lock); } -static void __intel_autoenable_gt_powersave(struct work_struct *work) -{ - struct drm_i915_private *dev_priv = - container_of(work, - typeof(*dev_priv), - gt_pm.autoenable_work.work); - struct intel_engine_cs *rcs; - struct drm_i915_gem_request *req; - - rcs = dev_priv->engine[RCS]; - if (rcs->last_retired_context) - goto out; - - if (!rcs->init_context) - goto out; - - mutex_lock(&dev_priv->drm.struct_mutex); - - req = i915_gem_request_alloc(rcs, dev_priv->kernel_context); - if (IS_ERR(req)) - goto unlock; - - if (!i915_modparams.enable_execlists && i915_switch_context(req) == 0) - rcs->init_context(req); - - /* Mark the device busy, calling intel_enable_gt_powersave() */ - i915_add_request(req); - -unlock: - mutex_unlock(&dev_priv->drm.struct_mutex); -out: - intel_runtime_pm_put(dev_priv); -} - -void intel_autoenable_gt_powersave(struct drm_i915_private *dev_priv) -{ - if (IS_IRONLAKE_M(dev_priv)) { - ironlake_enable_drps(dev_priv); - intel_init_emon(dev_priv); - } else if (INTEL_INFO(dev_priv)->gen >= 6) { - /* - * PCU communication is slow and this doesn't need to be - * done at any specific time, so do this out of our fast path - * to make resume and init faster. - * - * We depend on the HW RC6 power context save/restore - * mechanism when entering D3 through runtime PM suspend. So - * disable RPM until RPS/RC6 is properly setup. We can only - * get here via the driver load/system resume/runtime resume - * paths, so the _noresume version is enough (and in case of - * runtime resume it's necessary). - */ - if (queue_delayed_work(dev_priv->wq, - &dev_priv->gt_pm.autoenable_work, - round_jiffies_up_relative(HZ))) - intel_runtime_pm_get_noresume(dev_priv); - } -} - static void ibx_init_clock_gating(struct drm_i915_private *dev_priv) { /* @@ -9435,8 +9371,6 @@ void intel_pm_setup(struct drm_i915_private *dev_priv) { mutex_init(&dev_priv->pcu_lock); - INIT_DELAYED_WORK(&dev_priv->gt_pm.autoenable_work, - __intel_autoenable_gt_powersave); atomic_set(&dev_priv->gt_pm.rps.num_waiters, 0); dev_priv->runtime_pm.suspended = false; -- GitLab From 1b790cd9bfa8dc897710b5cdc24895cfda3d5f09 Mon Sep 17 00:00:00 2001 From: Oscar Mateo Date: Wed, 8 Nov 2017 15:59:44 -0800 Subject: [PATCH 0213/1471] drm/i915: Remove Gen9 WAs with no effect GEN8_CONFIG0 (0xD00) is a protected by a lock (bit 31) which is set by the BIOS, so there is no way we can enable the three chicken bits mandated by the WA (the BIOS should be doing it instead). v2: Rebased v3: Standalone patch References: b033bb6d5d3a ("drm/i915/gen9: Enable must set chicken bits in config0 reg") Signed-off-by: Oscar Mateo Cc: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/1510185589-9100-2-git-send-email-oscar.mateo@intel.com Reviewed-by: Mika Kuoppala Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_reg.h | 3 --- drivers/gpu/drm/i915/intel_pm.c | 3 --- 2 files changed, 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 6ef33422f7629..4e92db28af9c8 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -355,9 +355,6 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define ECOCHK_PPGTT_WT_HSW (0x2<<3) #define ECOCHK_PPGTT_WB_HSW (0x3<<3) -#define GEN8_CONFIG0 _MMIO(0xD00) -#define GEN9_DEFAULT_FIXES (1 << 3 | 1 << 2 | 1 << 1) - #define GAC_ECO_BITS _MMIO(0x14090) #define ECOBITS_SNB_BIT (1<<13) #define ECOBITS_PPGTT_CACHE64B (3<<8) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index c1a56809f1431..8e7f02e6008aa 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -75,9 +75,6 @@ static void gen9_init_clock_gating(struct drm_i915_private *dev_priv) I915_WRITE(CHICKEN_PAR1_1, I915_READ(CHICKEN_PAR1_1) | SKL_EDP_PSR_FIX_RDWRAP); - I915_WRITE(GEN8_CONFIG0, - I915_READ(GEN8_CONFIG0) | GEN9_DEFAULT_FIXES); - /* WaEnableChickenDCPR:skl,bxt,kbl,glk,cfl */ I915_WRITE(GEN8_CHICKEN_DCPR_1, I915_READ(GEN8_CHICKEN_DCPR_1) | MASK_WAKEMEM); -- GitLab From f9bab55ee60d6c885eb30dab1f20827053e23355 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:54 +0100 Subject: [PATCH 0214/1471] drm/i915: Remove bogus ips_enabled check. The flag just tells us IPS can be enabled, if the primary plane is not enabled it means IPS might not be. This never triggered in CI because we don't have a haswell ULT there, but can be reproduced easily with kms_atomic_transitions.plane-all-modeset-transition Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-2-maarten.lankhorst@linux.intel.com [mlankhorst: Remove from haswell_get_pipe_config too. (danvet)] Reviewed-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_display.c | 8 -------- 1 file changed, 8 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 63c8b9ccb31e9..2ecdd20ea7728 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9233,10 +9233,6 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, ironlake_get_pfit_config(crtc, pipe_config); } - if (IS_HASWELL(dev_priv)) - pipe_config->ips_enabled = hsw_crtc_supports_ips(crtc) && - (I915_READ(IPS_CTL) & IPS_ENABLE); - if (pipe_config->cpu_transcoder != TRANSCODER_EDP && !transcoder_is_dsi(pipe_config->cpu_transcoder)) { pipe_config->pixel_multiplier = @@ -11259,10 +11255,6 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_CLOCK_FUZZY(pixel_rate); } - /* BDW+ don't expose a synchronous way to read the state */ - if (IS_HASWELL(dev_priv)) - PIPE_CONF_CHECK_I(ips_enabled); - PIPE_CONF_CHECK_I(double_wide); PIPE_CONF_CHECK_P(shared_dpll); -- GitLab From d640bf7986cf49ef99aee72b6bb1433d6527f687 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:55 +0100 Subject: [PATCH 0215/1471] drm/i915: Check boolean options in intel_pipe_config_compare with its own macro Add PIPE_CONF_CHECK_BOOL for boolean options, which are printed with yesno. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-3-maarten.lankhorst@linux.intel.com Reviewed-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_display.c | 30 ++++++++++++++++++---------- 1 file changed, 20 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 2ecdd20ea7728..f56e7d29ed696 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11092,6 +11092,15 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, ret = false; \ } +#define PIPE_CONF_CHECK_BOOL(name) \ + if (current_config->name != pipe_config->name) { \ + pipe_config_err(adjust, __stringify(name), \ + "(expected %s, found %s)\n", \ + yesno(current_config->name), \ + yesno(pipe_config->name)); \ + ret = false; \ + } + #define PIPE_CONF_CHECK_P(name) \ if (current_config->name != pipe_config->name) { \ pipe_config_err(adjust, __stringify(name), \ @@ -11177,7 +11186,7 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_I(cpu_transcoder); - PIPE_CONF_CHECK_I(has_pch_encoder); + PIPE_CONF_CHECK_BOOL(has_pch_encoder); PIPE_CONF_CHECK_I(fdi_lanes); PIPE_CONF_CHECK_M_N(fdi_m_n); @@ -11209,17 +11218,17 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_I(base.adjusted_mode.crtc_vsync_end); PIPE_CONF_CHECK_I(pixel_multiplier); - PIPE_CONF_CHECK_I(has_hdmi_sink); + PIPE_CONF_CHECK_BOOL(has_hdmi_sink); if ((INTEL_GEN(dev_priv) < 8 && !IS_HASWELL(dev_priv)) || IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) - PIPE_CONF_CHECK_I(limited_color_range); + PIPE_CONF_CHECK_BOOL(limited_color_range); - PIPE_CONF_CHECK_I(hdmi_scrambling); - PIPE_CONF_CHECK_I(hdmi_high_tmds_clock_ratio); - PIPE_CONF_CHECK_I(has_infoframe); - PIPE_CONF_CHECK_I(ycbcr420); + PIPE_CONF_CHECK_BOOL(hdmi_scrambling); + PIPE_CONF_CHECK_BOOL(hdmi_high_tmds_clock_ratio); + PIPE_CONF_CHECK_BOOL(has_infoframe); + PIPE_CONF_CHECK_BOOL(ycbcr420); - PIPE_CONF_CHECK_I(has_audio); + PIPE_CONF_CHECK_BOOL(has_audio); PIPE_CONF_CHECK_FLAGS(base.adjusted_mode.flags, DRM_MODE_FLAG_INTERLACE); @@ -11245,7 +11254,7 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_I(pipe_src_w); PIPE_CONF_CHECK_I(pipe_src_h); - PIPE_CONF_CHECK_I(pch_pfit.enabled); + PIPE_CONF_CHECK_BOOL(pch_pfit.enabled); if (current_config->pch_pfit.enabled) { PIPE_CONF_CHECK_X(pch_pfit.pos); PIPE_CONF_CHECK_X(pch_pfit.size); @@ -11255,7 +11264,7 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_CLOCK_FUZZY(pixel_rate); } - PIPE_CONF_CHECK_I(double_wide); + PIPE_CONF_CHECK_BOOL(double_wide); PIPE_CONF_CHECK_P(shared_dpll); PIPE_CONF_CHECK_X(dpll_hw_state.dpll); @@ -11293,6 +11302,7 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, #undef PIPE_CONF_CHECK_X #undef PIPE_CONF_CHECK_I +#undef PIPE_CONF_CHECK_BOOL #undef PIPE_CONF_CHECK_P #undef PIPE_CONF_CHECK_FLAGS #undef PIPE_CONF_CHECK_CLOCK_FUZZY -- GitLab From 4493e098d71996bde2f2fbee8a3d92af21e9bcc8 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:56 +0100 Subject: [PATCH 0216/1471] drm/i915: Handle adjust better in intel_pipe_config_compare Some parameters use CHECK_BOOL, but should really use CHECK_BOOL_INCOMPLETE. We cannot currently check whether the inherited infoframes and audio are set up correctly. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-4-maarten.lankhorst@linux.intel.com [mlankhorst: Add danvet's comment about why this is needed.] Reviewed-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_display.c | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index f56e7d29ed696..0ebf3f283b877 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11073,6 +11073,9 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, bool adjust) { bool ret = true; + bool fixup_inherited = adjust && + (current_config->base.mode.private_flags & I915_MODE_FLAG_INHERITED) && + !(pipe_config->base.mode.private_flags & I915_MODE_FLAG_INHERITED); #define PIPE_CONF_CHECK_X(name) \ if (current_config->name != pipe_config->name) { \ @@ -11101,6 +11104,22 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, ret = false; \ } +/* + * Checks state where we only read out the enabling, but not the entire + * state itself (like full infoframes or ELD for audio). These states + * require a full modeset on bootup to fix up. + */ +#define PIPE_CONF_CHECK_BOOL_INCOMPLETE(name) \ + if (!fixup_inherited || (!current_config->name && !pipe_config->name)) { \ + PIPE_CONF_CHECK_BOOL(name); \ + } else { \ + pipe_config_err(adjust, __stringify(name), \ + "unable to verify whether state matches exactly, forcing modeset (expected %s, found %s)\n", \ + yesno(current_config->name), \ + yesno(pipe_config->name)); \ + ret = false; \ + } + #define PIPE_CONF_CHECK_P(name) \ if (current_config->name != pipe_config->name) { \ pipe_config_err(adjust, __stringify(name), \ @@ -11225,10 +11244,10 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, PIPE_CONF_CHECK_BOOL(hdmi_scrambling); PIPE_CONF_CHECK_BOOL(hdmi_high_tmds_clock_ratio); - PIPE_CONF_CHECK_BOOL(has_infoframe); + PIPE_CONF_CHECK_BOOL_INCOMPLETE(has_infoframe); PIPE_CONF_CHECK_BOOL(ycbcr420); - PIPE_CONF_CHECK_BOOL(has_audio); + PIPE_CONF_CHECK_BOOL_INCOMPLETE(has_audio); PIPE_CONF_CHECK_FLAGS(base.adjusted_mode.flags, DRM_MODE_FLAG_INTERLACE); @@ -11303,6 +11322,7 @@ intel_pipe_config_compare(struct drm_i915_private *dev_priv, #undef PIPE_CONF_CHECK_X #undef PIPE_CONF_CHECK_I #undef PIPE_CONF_CHECK_BOOL +#undef PIPE_CONF_CHECK_BOOL_INCOMPLETE #undef PIPE_CONF_CHECK_P #undef PIPE_CONF_CHECK_FLAGS #undef PIPE_CONF_CHECK_CLOCK_FUZZY -- GitLab From 10bf0a38c4084c14c57df5d635a5ba0867a3c260 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:58 +0100 Subject: [PATCH 0217/1471] drm/i915: Handle locking better in i915_sink_crc. Lock the bare minimum, instead of the entire world, and use interruptible locking because we can. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-6-maarten.lankhorst@linux.intel.com Reviewed-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_debugfs.c | 40 +++++++++++++++++++++++------ 1 file changed, 32 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 533ba096b9a63..cb77aeab4ee26 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2735,39 +2735,63 @@ static int i915_sink_crc(struct seq_file *m, void *data) struct intel_connector *connector; struct drm_connector_list_iter conn_iter; struct intel_dp *intel_dp = NULL; + struct drm_modeset_acquire_ctx ctx; int ret; u8 crc[6]; - drm_modeset_lock_all(dev); + drm_modeset_acquire_init(&ctx, DRM_MODESET_ACQUIRE_INTERRUPTIBLE); + drm_connector_list_iter_begin(dev, &conn_iter); + for_each_intel_connector_iter(connector, &conn_iter) { struct drm_crtc *crtc; + struct drm_connector_state *state; - if (!connector->base.state->best_encoder) + if (connector->base.connector_type != DRM_MODE_CONNECTOR_eDP) continue; - crtc = connector->base.state->crtc; - if (!crtc->state->active) +retry: + ret = drm_modeset_lock(&dev->mode_config.connection_mutex, &ctx); + if (ret) + goto err; + + state = connector->base.state; + if (!state->best_encoder) continue; - if (connector->base.connector_type != DRM_MODE_CONNECTOR_eDP) + crtc = state->crtc; + ret = drm_modeset_lock(&crtc->mutex, &ctx); + if (ret) + goto err; + + if (!crtc->state->active) continue; - intel_dp = enc_to_intel_dp(connector->base.state->best_encoder); + intel_dp = enc_to_intel_dp(state->best_encoder); ret = intel_dp_sink_crc(intel_dp, crc); if (ret) - goto out; + goto err; seq_printf(m, "%02x%02x%02x%02x%02x%02x\n", crc[0], crc[1], crc[2], crc[3], crc[4], crc[5]); goto out; + +err: + if (ret == -EDEADLK) { + ret = drm_modeset_backoff(&ctx); + if (!ret) + goto retry; + } + goto out; } ret = -ENODEV; out: drm_connector_list_iter_end(&conn_iter); - drm_modeset_unlock_all(dev); + drm_modeset_drop_locks(&ctx); + drm_modeset_acquire_fini(&ctx); + return ret; } -- GitLab From a54b19f17746b9574c30acb8cefca75ed5de0f9f Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:39 +0000 Subject: [PATCH 0218/1471] drm/i915/perf: complete whitelisting for OA programming on HSW We were missing some registers and also can name one for which we only had the offset. Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-2-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_perf.c | 3 ++- drivers/gpu/drm/i915/i915_reg.h | 14 ++++++++++++++ 2 files changed, 16 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 59ee808f8fd97..45aef15b9e7cd 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -3023,7 +3023,8 @@ static bool hsw_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { return gen7_is_valid_mux_addr(dev_priv, addr) || (addr >= 0x25100 && addr <= 0x2FF90) || - addr == 0x9ec0; + (addr >= HSW_MBVID2_NOA0.reg && addr <= HSW_MBVID2_NOA9.reg) || + addr == HSW_MBVID2_MISR0.reg; } static bool chv_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 4e92db28af9c8..0f09ba2e823b3 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -1117,6 +1117,20 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) /* RPC unit config (Gen8+) */ #define RPM_CONFIG _MMIO(0x0D08) +/* NOA (HSW) */ +#define HSW_MBVID2_NOA0 _MMIO(0x9E80) +#define HSW_MBVID2_NOA1 _MMIO(0x9E84) +#define HSW_MBVID2_NOA2 _MMIO(0x9E88) +#define HSW_MBVID2_NOA3 _MMIO(0x9E8C) +#define HSW_MBVID2_NOA4 _MMIO(0x9E90) +#define HSW_MBVID2_NOA5 _MMIO(0x9E94) +#define HSW_MBVID2_NOA6 _MMIO(0x9E98) +#define HSW_MBVID2_NOA7 _MMIO(0x9E9C) +#define HSW_MBVID2_NOA8 _MMIO(0x9EA0) +#define HSW_MBVID2_NOA9 _MMIO(0x9EA4) + +#define HSW_MBVID2_MISR0 _MMIO(0x9EC0) + /* NOA (Gen8+) */ #define NOA_CONFIG(i) _MMIO(0x0D0C + (i) * 4) -- GitLab From 4407eaa9b0ccbc55408884475301f02280aec402 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:40 +0000 Subject: [PATCH 0219/1471] drm/i915/perf: add support for Coffeelake GT3 We can enable GT3 as well as GT2. Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-3-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_drv.h | 2 + drivers/gpu/drm/i915/i915_oa_cflgt3.c | 109 ++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_oa_cflgt3.h | 34 ++++++++ drivers/gpu/drm/i915/i915_perf.c | 3 + 5 files changed, 150 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/i915/i915_oa_cflgt3.c create mode 100644 drivers/gpu/drm/i915/i915_oa_cflgt3.h diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 1bbc5440db40d..3c419455b0afb 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -162,7 +162,8 @@ i915-y += i915_perf.o \ i915_oa_kblgt2.o \ i915_oa_kblgt3.o \ i915_oa_glk.o \ - i915_oa_cflgt2.o + i915_oa_cflgt2.o \ + i915_oa_cflgt3.o ifeq ($(CONFIG_DRM_I915_GVT),y) i915-y += intel_gvt.o diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 07b9e78bbe966..d511bf948cd63 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3057,6 +3057,8 @@ intel_info(const struct drm_i915_private *dev_priv) (INTEL_DEVID(dev_priv) & 0x00F0) == 0x00A0) #define IS_CFL_GT2(dev_priv) (IS_COFFEELAKE(dev_priv) && \ (dev_priv)->info.gt == 2) +#define IS_CFL_GT3(dev_priv) (IS_COFFEELAKE(dev_priv) && \ + (dev_priv)->info.gt == 3) #define IS_ALPHA_SUPPORT(intel_info) ((intel_info)->is_alpha_support) diff --git a/drivers/gpu/drm/i915/i915_oa_cflgt3.c b/drivers/gpu/drm/i915/i915_oa_cflgt3.c new file mode 100644 index 0000000000000..42ff06fe54a3a --- /dev/null +++ b/drivers/gpu/drm/i915/i915_oa_cflgt3.c @@ -0,0 +1,109 @@ +/* + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + * + * + * Copyright (c) 2015 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#include + +#include "i915_drv.h" +#include "i915_oa_cflgt3.h" + +static const struct i915_oa_reg b_counter_config_test_oa[] = { + { _MMIO(0x2740), 0x00000000 }, + { _MMIO(0x2744), 0x00800000 }, + { _MMIO(0x2714), 0xf0800000 }, + { _MMIO(0x2710), 0x00000000 }, + { _MMIO(0x2724), 0xf0800000 }, + { _MMIO(0x2720), 0x00000000 }, + { _MMIO(0x2770), 0x00000004 }, + { _MMIO(0x2774), 0x00000000 }, + { _MMIO(0x2778), 0x00000003 }, + { _MMIO(0x277c), 0x00000000 }, + { _MMIO(0x2780), 0x00000007 }, + { _MMIO(0x2784), 0x00000000 }, + { _MMIO(0x2788), 0x00100002 }, + { _MMIO(0x278c), 0x0000fff7 }, + { _MMIO(0x2790), 0x00100002 }, + { _MMIO(0x2794), 0x0000ffcf }, + { _MMIO(0x2798), 0x00100082 }, + { _MMIO(0x279c), 0x0000ffef }, + { _MMIO(0x27a0), 0x001000c2 }, + { _MMIO(0x27a4), 0x0000ffe7 }, + { _MMIO(0x27a8), 0x00100001 }, + { _MMIO(0x27ac), 0x0000ffe7 }, +}; + +static const struct i915_oa_reg flex_eu_config_test_oa[] = { +}; + +static const struct i915_oa_reg mux_config_test_oa[] = { + { _MMIO(0x9840), 0x00000080 }, + { _MMIO(0x9888), 0x11810000 }, + { _MMIO(0x9888), 0x07810013 }, + { _MMIO(0x9888), 0x1f810000 }, + { _MMIO(0x9888), 0x1d810000 }, + { _MMIO(0x9888), 0x1b930040 }, + { _MMIO(0x9888), 0x07e54000 }, + { _MMIO(0x9888), 0x1f908000 }, + { _MMIO(0x9888), 0x11900000 }, + { _MMIO(0x9888), 0x37900000 }, + { _MMIO(0x9888), 0x53900000 }, + { _MMIO(0x9888), 0x45900000 }, + { _MMIO(0x9888), 0x33900000 }, +}; + +static ssize_t +show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "1\n"); +} + +void +i915_perf_load_test_config_cflgt3(struct drm_i915_private *dev_priv) +{ + strncpy(dev_priv->perf.oa.test_config.uuid, + "577e8e2c-3fa0-4875-8743-3538d585e3b0", + UUID_STRING_LEN); + dev_priv->perf.oa.test_config.id = 1; + + dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; + dev_priv->perf.oa.test_config.mux_regs_len = ARRAY_SIZE(mux_config_test_oa); + + dev_priv->perf.oa.test_config.b_counter_regs = b_counter_config_test_oa; + dev_priv->perf.oa.test_config.b_counter_regs_len = ARRAY_SIZE(b_counter_config_test_oa); + + dev_priv->perf.oa.test_config.flex_regs = flex_eu_config_test_oa; + dev_priv->perf.oa.test_config.flex_regs_len = ARRAY_SIZE(flex_eu_config_test_oa); + + dev_priv->perf.oa.test_config.sysfs_metric.name = "577e8e2c-3fa0-4875-8743-3538d585e3b0"; + dev_priv->perf.oa.test_config.sysfs_metric.attrs = dev_priv->perf.oa.test_config.attrs; + + dev_priv->perf.oa.test_config.attrs[0] = &dev_priv->perf.oa.test_config.sysfs_metric_id.attr; + + dev_priv->perf.oa.test_config.sysfs_metric_id.attr.name = "id"; + dev_priv->perf.oa.test_config.sysfs_metric_id.attr.mode = 0444; + dev_priv->perf.oa.test_config.sysfs_metric_id.show = show_test_oa_id; +} diff --git a/drivers/gpu/drm/i915/i915_oa_cflgt3.h b/drivers/gpu/drm/i915/i915_oa_cflgt3.h new file mode 100644 index 0000000000000..c13b5aac01b9b --- /dev/null +++ b/drivers/gpu/drm/i915/i915_oa_cflgt3.h @@ -0,0 +1,34 @@ +/* + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + * + * + * Copyright (c) 2015 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#ifndef __I915_OA_CFLGT3_H__ +#define __I915_OA_CFLGT3_H__ + +extern void i915_perf_load_test_config_cflgt3(struct drm_i915_private *dev_priv); + +#endif diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 45aef15b9e7cd..7271debe04174 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -207,6 +207,7 @@ #include "i915_oa_kblgt3.h" #include "i915_oa_glk.h" #include "i915_oa_cflgt2.h" +#include "i915_oa_cflgt3.h" /* HW requires this to be a power of two, between 128k and 16M, though driver * is currently generally designed assuming the largest 16M size is used such @@ -2934,6 +2935,8 @@ void i915_perf_register(struct drm_i915_private *dev_priv) } else if (IS_COFFEELAKE(dev_priv)) { if (IS_CFL_GT2(dev_priv)) i915_perf_load_test_config_cflgt2(dev_priv); + if (IS_CFL_GT3(dev_priv)) + i915_perf_load_test_config_cflgt3(dev_priv); } if (dev_priv->perf.oa.test_config.id == 0) -- GitLab From ba6b7c1ab26f223dd9b19c73df7813e944444d36 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:41 +0000 Subject: [PATCH 0220/1471] drm/i915/perf: refactor perf setup Gen8/9 aren't very different and we can merge some of this code. Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-4-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_perf.c | 48 +++++++++++++++++--------------- 1 file changed, 25 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 7271debe04174..802928c54f066 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -3423,41 +3423,46 @@ void i915_perf_init(struct drm_i915_private *dev_priv) * worth the complexity to maintain now that BDW+ enable * execlist mode by default. */ - dev_priv->perf.oa.ops.is_valid_b_counter_reg = - gen7_is_valid_b_counter_addr; - dev_priv->perf.oa.ops.is_valid_mux_reg = - gen8_is_valid_mux_addr; - dev_priv->perf.oa.ops.is_valid_flex_reg = - gen8_is_valid_flex_addr; + dev_priv->perf.oa.oa_formats = gen8_plus_oa_formats; dev_priv->perf.oa.ops.init_oa_buffer = gen8_init_oa_buffer; - dev_priv->perf.oa.ops.enable_metric_set = gen8_enable_metric_set; - dev_priv->perf.oa.ops.disable_metric_set = gen8_disable_metric_set; dev_priv->perf.oa.ops.oa_enable = gen8_oa_enable; dev_priv->perf.oa.ops.oa_disable = gen8_oa_disable; dev_priv->perf.oa.ops.read = gen8_oa_read; dev_priv->perf.oa.ops.oa_hw_tail_read = gen8_oa_hw_tail_read; - dev_priv->perf.oa.oa_formats = gen8_plus_oa_formats; - - if (IS_GEN8(dev_priv)) { - dev_priv->perf.oa.ctx_oactxctrl_offset = 0x120; - dev_priv->perf.oa.ctx_flexeu0_offset = 0x2ce; + if (IS_GEN8(dev_priv) || IS_GEN9(dev_priv)) { + dev_priv->perf.oa.ops.is_valid_b_counter_reg = + gen7_is_valid_b_counter_addr; + dev_priv->perf.oa.ops.is_valid_mux_reg = + gen8_is_valid_mux_addr; + dev_priv->perf.oa.ops.is_valid_flex_reg = + gen8_is_valid_flex_addr; - dev_priv->perf.oa.timestamp_frequency = 12500000; - - dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<25); if (IS_CHERRYVIEW(dev_priv)) { dev_priv->perf.oa.ops.is_valid_mux_reg = chv_is_valid_mux_addr; } - } else if (IS_GEN9(dev_priv)) { - dev_priv->perf.oa.ctx_oactxctrl_offset = 0x128; - dev_priv->perf.oa.ctx_flexeu0_offset = 0x3de; - dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16); + dev_priv->perf.oa.ops.enable_metric_set = gen8_enable_metric_set; + dev_priv->perf.oa.ops.disable_metric_set = gen8_disable_metric_set; + + if (IS_GEN8(dev_priv)) { + dev_priv->perf.oa.ctx_oactxctrl_offset = 0x120; + dev_priv->perf.oa.ctx_flexeu0_offset = 0x2ce; + + dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<25); + } else { + dev_priv->perf.oa.ctx_oactxctrl_offset = 0x128; + dev_priv->perf.oa.ctx_flexeu0_offset = 0x3de; + + dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16); + } switch (dev_priv->info.platform) { + case INTEL_BROADWELL: + dev_priv->perf.oa.timestamp_frequency = 12500000; + break; case INTEL_BROXTON: case INTEL_GEMINILAKE: dev_priv->perf.oa.timestamp_frequency = 19200000; @@ -3468,9 +3473,6 @@ void i915_perf_init(struct drm_i915_private *dev_priv) dev_priv->perf.oa.timestamp_frequency = 12000000; break; default: - /* Leave timestamp_frequency to 0 so we can - * detect unsupported platforms. - */ break; } } -- GitLab From 5888576b0b5fedaaee278879b98b66c21984c4e6 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:42 +0000 Subject: [PATCH 0221/1471] drm/i915: fix register naming This name was added with the whitelisting of registers for building up OA configs. It is contained in a range gen8 whitelist : addr >= RPM_CONFIG0.reg && addr <= NOA_CONFIG(8).reg Hence why the name isn't used anywhere. v2: Fix register name again RPC->RCP (Matthew) Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-5-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_reg.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 0f09ba2e823b3..92f0c497fac85 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -1114,8 +1114,8 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define RPM_CONFIG0 _MMIO(0x0D00) #define RPM_CONFIG1 _MMIO(0x0D04) -/* RPC unit config (Gen8+) */ -#define RPM_CONFIG _MMIO(0x0D08) +/* RCP unit config (Gen8+) */ +#define RCP_CONFIG _MMIO(0x0D08) /* NOA (HSW) */ #define HSW_MBVID2_NOA0 _MMIO(0x9E80) -- GitLab From 95690a02fb5d963e62aa16c3796af3dde01f63c9 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:43 +0000 Subject: [PATCH 0222/1471] drm/i915/perf: enable perf support on CNL This adds new registers to the whitelist to configs emitted from userspace. Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-6-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_oa_cnl.c | 121 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_oa_cnl.h | 34 ++++++++ drivers/gpu/drm/i915/i915_perf.c | 41 +++++++++- drivers/gpu/drm/i915/i915_reg.h | 5 ++ 5 files changed, 202 insertions(+), 2 deletions(-) create mode 100644 drivers/gpu/drm/i915/i915_oa_cnl.c create mode 100644 drivers/gpu/drm/i915/i915_oa_cnl.h diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 3c419455b0afb..f7afd44214b51 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -163,7 +163,8 @@ i915-y += i915_perf.o \ i915_oa_kblgt3.o \ i915_oa_glk.o \ i915_oa_cflgt2.o \ - i915_oa_cflgt3.o + i915_oa_cflgt3.o \ + i915_oa_cnl.o ifeq ($(CONFIG_DRM_I915_GVT),y) i915-y += intel_gvt.o diff --git a/drivers/gpu/drm/i915/i915_oa_cnl.c b/drivers/gpu/drm/i915/i915_oa_cnl.c new file mode 100644 index 0000000000000..ff0ac3627cc4b --- /dev/null +++ b/drivers/gpu/drm/i915/i915_oa_cnl.c @@ -0,0 +1,121 @@ +/* + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + * + * + * Copyright (c) 2015 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#include + +#include "i915_drv.h" +#include "i915_oa_cnl.h" + +static const struct i915_oa_reg b_counter_config_test_oa[] = { + { _MMIO(0x2740), 0x00000000 }, + { _MMIO(0x2710), 0x00000000 }, + { _MMIO(0x2714), 0xf0800000 }, + { _MMIO(0x2720), 0x00000000 }, + { _MMIO(0x2724), 0xf0800000 }, + { _MMIO(0x2770), 0x00000004 }, + { _MMIO(0x2774), 0x0000ffff }, + { _MMIO(0x2778), 0x00000003 }, + { _MMIO(0x277c), 0x0000ffff }, + { _MMIO(0x2780), 0x00000007 }, + { _MMIO(0x2784), 0x0000ffff }, + { _MMIO(0x2788), 0x00100002 }, + { _MMIO(0x278c), 0x0000fff7 }, + { _MMIO(0x2790), 0x00100002 }, + { _MMIO(0x2794), 0x0000ffcf }, + { _MMIO(0x2798), 0x00100082 }, + { _MMIO(0x279c), 0x0000ffef }, + { _MMIO(0x27a0), 0x001000c2 }, + { _MMIO(0x27a4), 0x0000ffe7 }, + { _MMIO(0x27a8), 0x00100001 }, + { _MMIO(0x27ac), 0x0000ffe7 }, +}; + +static const struct i915_oa_reg flex_eu_config_test_oa[] = { +}; + +static const struct i915_oa_reg mux_config_test_oa[] = { + { _MMIO(0xd04), 0x00000200 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x17060000 }, + { _MMIO(0x9840), 0x00000000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x13034000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x07060066 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x05060000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x0f080040 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x07091000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x0f041000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x1d004000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x35000000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x49000000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x3d000000 }, + { _MMIO(0x9884), 0x00000007 }, + { _MMIO(0x9888), 0x31000000 }, +}; + +static ssize_t +show_test_oa_id(struct device *kdev, struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "1\n"); +} + +void +i915_perf_load_test_config_cnl(struct drm_i915_private *dev_priv) +{ + strncpy(dev_priv->perf.oa.test_config.uuid, + "db41edd4-d8e7-4730-ad11-b9a2d6833503", + UUID_STRING_LEN); + dev_priv->perf.oa.test_config.id = 1; + + dev_priv->perf.oa.test_config.mux_regs = mux_config_test_oa; + dev_priv->perf.oa.test_config.mux_regs_len = ARRAY_SIZE(mux_config_test_oa); + + dev_priv->perf.oa.test_config.b_counter_regs = b_counter_config_test_oa; + dev_priv->perf.oa.test_config.b_counter_regs_len = ARRAY_SIZE(b_counter_config_test_oa); + + dev_priv->perf.oa.test_config.flex_regs = flex_eu_config_test_oa; + dev_priv->perf.oa.test_config.flex_regs_len = ARRAY_SIZE(flex_eu_config_test_oa); + + dev_priv->perf.oa.test_config.sysfs_metric.name = "db41edd4-d8e7-4730-ad11-b9a2d6833503"; + dev_priv->perf.oa.test_config.sysfs_metric.attrs = dev_priv->perf.oa.test_config.attrs; + + dev_priv->perf.oa.test_config.attrs[0] = &dev_priv->perf.oa.test_config.sysfs_metric_id.attr; + + dev_priv->perf.oa.test_config.sysfs_metric_id.attr.name = "id"; + dev_priv->perf.oa.test_config.sysfs_metric_id.attr.mode = 0444; + dev_priv->perf.oa.test_config.sysfs_metric_id.show = show_test_oa_id; +} diff --git a/drivers/gpu/drm/i915/i915_oa_cnl.h b/drivers/gpu/drm/i915/i915_oa_cnl.h new file mode 100644 index 0000000000000..fb918b1311058 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_oa_cnl.h @@ -0,0 +1,34 @@ +/* + * Autogenerated file by GPU Top : https://github.com/rib/gputop + * DO NOT EDIT manually! + * + * + * Copyright (c) 2015 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#ifndef __I915_OA_CNL_H__ +#define __I915_OA_CNL_H__ + +extern void i915_perf_load_test_config_cnl(struct drm_i915_private *dev_priv); + +#endif diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 802928c54f066..00be015e01dfc 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -208,6 +208,7 @@ #include "i915_oa_glk.h" #include "i915_oa_cflgt2.h" #include "i915_oa_cflgt3.h" +#include "i915_oa_cnl.h" /* HW requires this to be a power of two, between 128k and 16M, though driver * is currently generally designed assuming the largest 16M size is used such @@ -1852,7 +1853,7 @@ static int gen8_enable_metric_set(struct drm_i915_private *dev_priv, * be read back from automatically triggered reports, as part of the * RPT_ID field. */ - if (IS_GEN9(dev_priv)) { + if (IS_GEN9(dev_priv) || IS_GEN10(dev_priv)) { I915_WRITE(GEN8_OA_DEBUG, _MASKED_BIT_ENABLE(GEN9_OA_DEBUG_DISABLE_CLK_RATIO_REPORTS | GEN9_OA_DEBUG_INCLUDE_CLK_RATIO)); @@ -1885,6 +1886,16 @@ static void gen8_disable_metric_set(struct drm_i915_private *dev_priv) } +static void gen10_disable_metric_set(struct drm_i915_private *dev_priv) +{ + /* Reset all contexts' slices/subslices configurations. */ + gen8_configure_all_contexts(dev_priv, NULL, false); + + /* Make sure we disable noa to save power. */ + I915_WRITE(RPM_CONFIG1, + I915_READ(RPM_CONFIG1) & ~GEN10_GT_NOA_ENABLE); +} + static void gen7_oa_enable(struct drm_i915_private *dev_priv) { /* @@ -2937,6 +2948,8 @@ void i915_perf_register(struct drm_i915_private *dev_priv) i915_perf_load_test_config_cflgt2(dev_priv); if (IS_CFL_GT3(dev_priv)) i915_perf_load_test_config_cflgt3(dev_priv); + } else if (IS_CANNONLAKE(dev_priv)) { + i915_perf_load_test_config_cnl(dev_priv); } if (dev_priv->perf.oa.test_config.id == 0) @@ -3022,6 +3035,12 @@ static bool gen8_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) (addr >= RPM_CONFIG0.reg && addr <= NOA_CONFIG(8).reg); } +static bool gen10_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) +{ + return gen8_is_valid_mux_addr(dev_priv, addr) || + (addr >= OA_PERFCNT3_LO.reg && addr <= OA_PERFCNT4_HI.reg); +} + static bool hsw_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { return gen7_is_valid_mux_addr(dev_priv, addr) || @@ -3475,6 +3494,26 @@ void i915_perf_init(struct drm_i915_private *dev_priv) default: break; } + } else if (IS_GEN10(dev_priv)) { + dev_priv->perf.oa.ops.is_valid_b_counter_reg = + gen7_is_valid_b_counter_addr; + dev_priv->perf.oa.ops.is_valid_mux_reg = + gen10_is_valid_mux_addr; + dev_priv->perf.oa.ops.is_valid_flex_reg = + gen8_is_valid_flex_addr; + + dev_priv->perf.oa.ops.enable_metric_set = gen8_enable_metric_set; + dev_priv->perf.oa.ops.disable_metric_set = gen10_disable_metric_set; + + dev_priv->perf.oa.ctx_oactxctrl_offset = 0x128; + dev_priv->perf.oa.ctx_flexeu0_offset = 0x3de; + + dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16); + + /* Default frequency, although we need to read it from + * the register as it might vary between parts. + */ + dev_priv->perf.oa.timestamp_frequency = 12000000; } } diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 92f0c497fac85..31e4543ca9422 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -1106,6 +1106,10 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define OA_PERFCNT1_HI _MMIO(0x91BC) #define OA_PERFCNT2_LO _MMIO(0x91C0) #define OA_PERFCNT2_HI _MMIO(0x91C4) +#define OA_PERFCNT3_LO _MMIO(0x91C8) +#define OA_PERFCNT3_HI _MMIO(0x91CC) +#define OA_PERFCNT4_LO _MMIO(0x91D8) +#define OA_PERFCNT4_HI _MMIO(0x91DC) #define OA_PERFMATRIX_LO _MMIO(0x91C8) #define OA_PERFMATRIX_HI _MMIO(0x91CC) @@ -1113,6 +1117,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) /* RPM unit config (Gen8+) */ #define RPM_CONFIG0 _MMIO(0x0D00) #define RPM_CONFIG1 _MMIO(0x0D04) +#define GEN10_GT_NOA_ENABLE (1 << 9) /* RCP unit config (Gen8+) */ #define RCP_CONFIG _MMIO(0x0D08) -- GitLab From dab91783338bd3dd42638f89b5f7e34c57773207 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 10 Nov 2017 19:08:44 +0000 Subject: [PATCH 0223/1471] drm/i915: expose command stream timestamp frequency to userspace We use to have this fixed per generation, but starting with CNL userspace cannot tell just off the PCI ID. Let's make this information available. This is particularly useful for performance monitoring where much of the normalization work is done using those timestamps (this include pipeline statistics in both GL & Vulkan as well as OA reports). v2: Use variables for 24MHz/19.2MHz values (Ewelina) Renamed function & coding style (Sagar) v3: Fix frequency read on Broadwell (Sagar) Fix missing divide by 4 on <= gen4 (Sagar) Signed-off-by: Lionel Landwerlin Tested-by: Rafael Antognolli Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171110190845.32574-7-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_debugfs.c | 2 + drivers/gpu/drm/i915/i915_drv.c | 3 + drivers/gpu/drm/i915/i915_drv.h | 2 + drivers/gpu/drm/i915/i915_reg.h | 21 +++++ drivers/gpu/drm/i915/intel_device_info.c | 107 +++++++++++++++++++++++ include/uapi/drm/i915_drm.h | 6 ++ 6 files changed, 141 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index cb77aeab4ee26..82ff186108f18 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3269,6 +3269,8 @@ static int i915_engine_info(struct seq_file *m, void *unused) yesno(dev_priv->gt.awake)); seq_printf(m, "Global active requests: %d\n", dev_priv->gt.active_requests); + seq_printf(m, "CS timestamp frequency: %llu Hz\n", + dev_priv->info.cs_timestamp_frequency); p = drm_seq_file_printer(m); for_each_engine(engine, dev_priv, id) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 9df7b5d59a94b..42813f4247e29 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -419,6 +419,9 @@ static int i915_getparam(struct drm_device *dev, void *data, if (!value) return -ENODEV; break; + case I915_PARAM_CS_TIMESTAMP_FREQUENCY: + value = INTEL_INFO(dev_priv)->cs_timestamp_frequency; + break; default: DRM_DEBUG("Unknown parameter %d\n", param->param); return -EINVAL; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d511bf948cd63..b538df740ac39 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -885,6 +885,8 @@ struct intel_device_info { /* Slice/subslice/EU info */ struct sseu_dev_info sseu; + u64 cs_timestamp_frequency; + struct color_luts { u16 degamma_lut_size; u16 gamma_lut_size; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 31e4543ca9422..05e33a41fcc77 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -1116,9 +1116,24 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) /* RPM unit config (Gen8+) */ #define RPM_CONFIG0 _MMIO(0x0D00) +#define GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_SHIFT 3 +#define GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_MASK (1 << GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_SHIFT) +#define GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_19_2_MHZ 0 +#define GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_24_MHZ 1 +#define GEN10_RPM_CONFIG0_CTC_SHIFT_PARAMETER_SHIFT 1 +#define GEN10_RPM_CONFIG0_CTC_SHIFT_PARAMETER_MASK (0x3 << GEN10_RPM_CONFIG0_CTC_SHIFT_PARAMETER_SHIFT) + #define RPM_CONFIG1 _MMIO(0x0D04) #define GEN10_GT_NOA_ENABLE (1 << 9) +/* GPM unit config (Gen9+) */ +#define CTC_MODE _MMIO(0xA26C) +#define CTC_SOURCE_PARAMETER_MASK 1 +#define CTC_SOURCE_CRYSTAL_CLOCK 0 +#define CTC_SOURCE_DIVIDE_LOGIC 1 +#define CTC_SHIFT_PARAMETER_SHIFT 1 +#define CTC_SHIFT_PARAMETER_MASK (0x3 << CTC_SHIFT_PARAMETER_SHIFT) + /* RCP unit config (Gen8+) */ #define RCP_CONFIG _MMIO(0x0D08) @@ -8863,6 +8878,12 @@ enum skl_power_gate { #define ILK_TIMESTAMP_HI _MMIO(0x70070) #define IVB_TIMESTAMP_CTR _MMIO(0x44070) +#define GEN9_TIMESTAMP_OVERRIDE _MMIO(0x44074) +#define GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_SHIFT 0 +#define GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_MASK 0x3ff +#define GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_SHIFT 12 +#define GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_MASK (0xf << 12) + #define _PIPE_FRMTMSTMP_A 0x70048 #define PIPE_FRMTMSTMP(pipe) \ _MMIO_PIPE2(pipe, _PIPE_FRMTMSTMP_A) diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c index db03d179fc854..78bf7374fbdde 100644 --- a/drivers/gpu/drm/i915/intel_device_info.c +++ b/drivers/gpu/drm/i915/intel_device_info.c @@ -329,6 +329,108 @@ static void broadwell_sseu_info_init(struct drm_i915_private *dev_priv) sseu->has_eu_pg = 0; } +static u64 read_reference_ts_freq(struct drm_i915_private *dev_priv) +{ + u32 ts_override = I915_READ(GEN9_TIMESTAMP_OVERRIDE); + u64 base_freq, frac_freq; + + base_freq = ((ts_override & GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_MASK) >> + GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_SHIFT) + 1; + base_freq *= 1000000; + + frac_freq = ((ts_override & + GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_MASK) >> + GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_SHIFT); + if (frac_freq != 0) + frac_freq = 1000000 / (frac_freq + 1); + + return base_freq + frac_freq; +} + +static u64 read_timestamp_frequency(struct drm_i915_private *dev_priv) +{ + u64 f12_5_mhz = 12500000; + u64 f19_2_mhz = 19200000; + u64 f24_mhz = 24000000; + + if (INTEL_GEN(dev_priv) <= 4) { + /* PRMs say: + * + * "The value in this register increments once every 16 + * hclks." (through the “Clocking Configuration” + * (“CLKCFG”) MCHBAR register) + */ + return (dev_priv->rawclk_freq * 1000) / 16; + } else if (INTEL_GEN(dev_priv) <= 8) { + /* PRMs say: + * + * "The PCU TSC counts 10ns increments; this timestamp + * reflects bits 38:3 of the TSC (i.e. 80ns granularity, + * rolling over every 1.5 hours). + */ + return f12_5_mhz; + } else if (INTEL_GEN(dev_priv) <= 9) { + u32 ctc_reg = I915_READ(CTC_MODE); + u64 freq = 0; + + if ((ctc_reg & CTC_SOURCE_PARAMETER_MASK) == CTC_SOURCE_DIVIDE_LOGIC) { + freq = read_reference_ts_freq(dev_priv); + } else { + freq = IS_GEN9_LP(dev_priv) ? f19_2_mhz : f24_mhz; + + /* Now figure out how the command stream's timestamp + * register increments from this frequency (it might + * increment only every few clock cycle). + */ + freq >>= 3 - ((ctc_reg & CTC_SHIFT_PARAMETER_MASK) >> + CTC_SHIFT_PARAMETER_SHIFT); + } + + return freq; + } else if (INTEL_GEN(dev_priv) <= 10) { + u32 ctc_reg = I915_READ(CTC_MODE); + u64 freq = 0; + u32 rpm_config_reg = 0; + + /* First figure out the reference frequency. There are 2 ways + * we can compute the frequency, either through the + * TIMESTAMP_OVERRIDE register or through RPM_CONFIG. CTC_MODE + * tells us which one we should use. + */ + if ((ctc_reg & CTC_SOURCE_PARAMETER_MASK) == CTC_SOURCE_DIVIDE_LOGIC) { + freq = read_reference_ts_freq(dev_priv); + } else { + u32 crystal_clock; + + rpm_config_reg = I915_READ(RPM_CONFIG0); + crystal_clock = (rpm_config_reg & + GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_MASK) >> + GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_SHIFT; + switch (crystal_clock) { + case GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_19_2_MHZ: + freq = f19_2_mhz; + break; + case GEN9_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_24_MHZ: + freq = f24_mhz; + break; + } + } + + /* Now figure out how the command stream's timestamp register + * increments from this frequency (it might increment only + * every few clock cycle). + */ + freq >>= 3 - ((rpm_config_reg & + GEN10_RPM_CONFIG0_CTC_SHIFT_PARAMETER_MASK) >> + GEN10_RPM_CONFIG0_CTC_SHIFT_PARAMETER_SHIFT); + + return freq; + } + + DRM_ERROR("Unknown gen, unable to compute command stream timestamp frequency\n"); + return 0; +} + /* * Determine various intel_device_info fields at runtime. * @@ -450,6 +552,9 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) else if (INTEL_GEN(dev_priv) >= 10) gen10_sseu_info_init(dev_priv); + /* Initialize command stream timestamp frequency */ + info->cs_timestamp_frequency = read_timestamp_frequency(dev_priv); + DRM_DEBUG_DRIVER("slice mask: %04x\n", info->sseu.slice_mask); DRM_DEBUG_DRIVER("slice total: %u\n", hweight8(info->sseu.slice_mask)); DRM_DEBUG_DRIVER("subslice total: %u\n", @@ -465,4 +570,6 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) info->sseu.has_subslice_pg ? "y" : "n"); DRM_DEBUG_DRIVER("has EU power gating: %s\n", info->sseu.has_eu_pg ? "y" : "n"); + DRM_DEBUG_DRIVER("CS timestamp frequency: %llu\n", + info->cs_timestamp_frequency); } diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 6c02ced663f81..b579859295536 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -481,6 +481,12 @@ typedef struct drm_i915_irq_wait { */ #define I915_PARAM_HAS_CONTEXT_ISOLATION 50 +/* Frequency of the command streamer timestamps given by the *_TIMESTAMP + * registers. This used to be fixed per platform but from CNL onwards, this + * might vary depending on the parts. + */ +#define I915_PARAM_CS_TIMESTAMP_FREQUENCY 51 + typedef struct drm_i915_getparam { __s32 param; /* -- GitLab From ce453b3e426f33eb56e5425f0b839f75eed621d9 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Fri, 10 Nov 2017 16:44:47 -0800 Subject: [PATCH 0224/1471] drm/i915: Clear per-engine fault register as early as possible From gen6, the hardware tracks address lookup failures and we should clear those registers upon startup to prevent false positives. However, this was happening before we have the engines defined (intel_uncore_init()) and the for_each_engine loop was just a nop. The earliest we can call this is inside intel_engines_init_mmio(). Suggested-by: Chris Wilson Signed-off-by: Michel Thierry Cc: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171111004448.12360-1-michel.thierry@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_engine_cs.c | 2 ++ drivers/gpu/drm/i915/intel_uncore.c | 2 -- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index a0f9d0eb4bce9..70bbe8ef8f54f 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -289,6 +289,8 @@ int intel_engines_init_mmio(struct drm_i915_private *dev_priv) device_info->num_rings = hweight32(mask); + i915_check_and_clear_faults(dev_priv); + return 0; cleanup: diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 211acee7c31da..a78ceafcc8254 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1420,8 +1420,6 @@ void intel_uncore_init(struct drm_i915_private *dev_priv) iosf_mbi_register_pmic_bus_access_notifier( &dev_priv->uncore.pmic_bus_access_nb); - - i915_check_and_clear_faults(dev_priv); } void intel_uncore_fini(struct drm_i915_private *dev_priv) -- GitLab From b03ec3d67ab840f674eb61a3467fd971c9552aa4 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Mon, 13 Nov 2017 09:36:28 -0800 Subject: [PATCH 0225/1471] drm/i915: There is only one fault register from GEN8 onwards Until Haswell/Baytrail, the hardware used to have a per engine fault register (e.g. 0x4094 - render fault register, 0x4194 - media fault register and so on). But since Broadwell, all these registers were combined into a singe one and the engine id stored in bits 14:12. Not only we should not been reading (and writing to) registers that do not exist, in platforms with VCS2 (SKL), the address that would belong this engine (0x4494, VCS2_HW = 4) is already assigned to other register. v2: use less controversial function names (Chris). v3: make non-exported functions static, remove now obsolete check for engine presence before posting_read (Chris). References: IHD-OS-BDW-Vol 2c-11.15, page 75. References: IHD-OS-SKL-Vol 2c-05.16, page 350. Signed-off-by: Michel Thierry Cc: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171113173628.11689-1-michel.thierry@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_gem_gtt.c | 57 ++++++++++++++++++++------- drivers/gpu/drm/i915/i915_gpu_error.c | 8 ++-- drivers/gpu/drm/i915/i915_reg.h | 2 + 3 files changed, 49 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 1e40eeb31f9da..3c3a699436c92 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -2256,35 +2256,62 @@ static bool needs_idle_maps(struct drm_i915_private *dev_priv) return IS_GEN5(dev_priv) && IS_MOBILE(dev_priv) && intel_vtd_active(); } -void i915_check_and_clear_faults(struct drm_i915_private *dev_priv) +static void gen6_check_and_clear_faults(struct drm_i915_private *dev_priv) { struct intel_engine_cs *engine; enum intel_engine_id id; - - if (INTEL_INFO(dev_priv)->gen < 6) - return; + u32 fault; for_each_engine(engine, dev_priv, id) { - u32 fault_reg; - fault_reg = I915_READ(RING_FAULT_REG(engine)); - if (fault_reg & RING_FAULT_VALID) { + fault = I915_READ(RING_FAULT_REG(engine)); + if (fault & RING_FAULT_VALID) { DRM_DEBUG_DRIVER("Unexpected fault\n" "\tAddr: 0x%08lx\n" "\tAddress space: %s\n" "\tSource ID: %d\n" "\tType: %d\n", - fault_reg & PAGE_MASK, - fault_reg & RING_FAULT_GTTSEL_MASK ? "GGTT" : "PPGTT", - RING_FAULT_SRCID(fault_reg), - RING_FAULT_FAULT_TYPE(fault_reg)); + fault & PAGE_MASK, + fault & RING_FAULT_GTTSEL_MASK ? "GGTT" : "PPGTT", + RING_FAULT_SRCID(fault), + RING_FAULT_FAULT_TYPE(fault)); I915_WRITE(RING_FAULT_REG(engine), - fault_reg & ~RING_FAULT_VALID); + fault & ~RING_FAULT_VALID); } } - /* Engine specific init may not have been done till this point. */ - if (dev_priv->engine[RCS]) - POSTING_READ(RING_FAULT_REG(dev_priv->engine[RCS])); + POSTING_READ(RING_FAULT_REG(dev_priv->engine[RCS])); +} + +static void gen8_check_and_clear_faults(struct drm_i915_private *dev_priv) +{ + u32 fault = I915_READ(GEN8_RING_FAULT_REG); + + if (fault & RING_FAULT_VALID) { + DRM_DEBUG_DRIVER("Unexpected fault\n" + "\tAddr: 0x%08lx\n" + "\tEngine ID: %d\n" + "\tSource ID: %d\n" + "\tType: %d\n", + fault & PAGE_MASK, + GEN8_RING_FAULT_ENGINE_ID(fault), + RING_FAULT_SRCID(fault), + RING_FAULT_FAULT_TYPE(fault)); + I915_WRITE(GEN8_RING_FAULT_REG, + fault & ~RING_FAULT_VALID); + } + + POSTING_READ(GEN8_RING_FAULT_REG); +} + +void i915_check_and_clear_faults(struct drm_i915_private *dev_priv) +{ + /* From GEN8 onwards we only have one 'All Engine Fault Register' */ + if (INTEL_GEN(dev_priv) >= 8) + gen8_check_and_clear_faults(dev_priv); + else if (INTEL_GEN(dev_priv) >= 6) + gen6_check_and_clear_faults(dev_priv); + else + return; } void i915_gem_suspend_gtt_mappings(struct drm_i915_private *dev_priv) diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 5c2d83a838d8c..7481c8e1b5a8f 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1217,11 +1217,13 @@ static void error_record_engine_registers(struct i915_gpu_state *error, if (INTEL_GEN(dev_priv) >= 6) { ee->rc_psmi = I915_READ(RING_PSMI_CTL(engine->mmio_base)); - ee->fault_reg = I915_READ(RING_FAULT_REG(engine)); - if (INTEL_GEN(dev_priv) >= 8) + if (INTEL_GEN(dev_priv) >= 8) { gen8_record_semaphore_state(error, engine, ee); - else + ee->fault_reg = I915_READ(GEN8_RING_FAULT_REG); + } else { gen6_record_semaphore_state(engine, ee); + ee->fault_reg = I915_READ(RING_FAULT_REG(engine)); + } } if (INTEL_GEN(dev_priv) >= 4) { diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 05e33a41fcc77..e52e74db4c7c3 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -2360,6 +2360,8 @@ enum i915_power_well_id { #define ARB_MODE_SWIZZLE_BDW (1<<1) #define RENDER_HWS_PGA_GEN7 _MMIO(0x04080) #define RING_FAULT_REG(engine) _MMIO(0x4094 + 0x100*(engine)->hw_id) +#define GEN8_RING_FAULT_REG _MMIO(0x4094) +#define GEN8_RING_FAULT_ENGINE_ID(x) (((x) >> 12) & 0x7) #define RING_FAULT_GTTSEL_MASK (1<<11) #define RING_FAULT_SRCID(x) (((x) >> 3) & 0xff) #define RING_FAULT_FAULT_TYPE(x) (((x) >> 1) & 0x3) -- GitLab From 5d276a1acaa80f4b7a76577510a2b1835ce01f4f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Thu, 9 Nov 2017 09:59:05 +0100 Subject: [PATCH 0226/1471] dma-buf: add reservation_object_lock_interruptible() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit That's the only wrapper function missing and necessary to cleanup TTM. Reviewed-and-Tested-by: Michel Dänzer Signed-off-by: Christian König Signed-off-by: Alex Deucher Link: https://patchwork.freedesktop.org/patch/msgid/20171109085909.1653-3-christian.koenig@amd.com --- include/linux/reservation.h | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/include/linux/reservation.h b/include/linux/reservation.h index 21fc84d82d415..02166e815afb9 100644 --- a/include/linux/reservation.h +++ b/include/linux/reservation.h @@ -166,6 +166,29 @@ reservation_object_lock(struct reservation_object *obj, return ww_mutex_lock(&obj->lock, ctx); } +/** + * reservation_object_lock_interruptible - lock the reservation object + * @obj: the reservation object + * @ctx: the locking context + * + * Locks the reservation object interruptible for exclusive access and + * modification. Note, that the lock is only against other writers, readers + * will run concurrently with a writer under RCU. The seqlock is used to + * notify readers if they overlap with a writer. + * + * As the reservation object may be locked by multiple parties in an + * undefined order, a #ww_acquire_ctx is passed to unwind if a cycle + * is detected. See ww_mutex_lock() and ww_acquire_init(). A reservation + * object may be locked by itself by passing NULL as @ctx. + */ +static inline int +reservation_object_lock_interruptible(struct reservation_object *obj, + struct ww_acquire_ctx *ctx) +{ + return ww_mutex_lock_interruptible(&obj->lock, ctx); +} + + /** * reservation_object_trylock - trylock the reservation object * @obj: the reservation object -- GitLab From 0c51ef8684e30333ee2a7cef359258d91727166b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Tue, 14 Nov 2017 00:12:01 +0100 Subject: [PATCH 0227/1471] drm/debugfs: Fix framebuffer debugfs file init MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The introduction of: drm/framebuffer: Add framebuffer debugfs file broke vgem. That patch assumed that all drivers had initialized the dev->mode_config.fb_lock mutex which happens in drm_mode_config_init(). vgem doesn't need to call drm_mode_config_init(). Fix this by only creating the framebuffer debugfs file for modesetting drivers. Fixes: 45d58b40292b ("drm/framebuffer: Add framebuffer debugfs file") Suggested-by: Chris Wilson Signed-off-by: Noralf Trønnes Reviewed-by: Dave Airlie Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171113231201.26177-1-noralf@tronnes.org --- drivers/gpu/drm/drm_debugfs.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c index 550f29de6c1fc..b2482818fee8c 100644 --- a/drivers/gpu/drm/drm_debugfs.c +++ b/drivers/gpu/drm/drm_debugfs.c @@ -158,10 +158,12 @@ int drm_debugfs_init(struct drm_minor *minor, int minor_id, } } - ret = drm_framebuffer_debugfs_init(minor); - if (ret) { - DRM_ERROR("Failed to create framebuffer debugfs file\n"); - return ret; + if (drm_core_check_feature(dev, DRIVER_MODESET)) { + ret = drm_framebuffer_debugfs_init(minor); + if (ret) { + DRM_ERROR("Failed to create framebuffer debugfs file\n"); + return ret; + } } if (dev->driver->debugfs_init) { -- GitLab From 40c7ae457cf969b2558a8a8f3827e8399e68a986 Mon Sep 17 00:00:00 2001 From: Lucas De Marchi Date: Mon, 13 Nov 2017 16:46:38 -0800 Subject: [PATCH 0228/1471] drm/i915: Fix function name in comment Commit 78597996370c (drm/i915/bxt: Fix PPS lost state after suspend breaking eDP link training) renamed the function to intel_power_sequencer_reset() but forgot to update comment. Cc: Imre Deak Signed-off-by: Lucas De Marchi Signed-off-by: Imre Deak Reviewed-by: Imre Deak Link: https://patchwork.freedesktop.org/patch/msgid/20171114004638.5186-1-lucas.demarchi@intel.com --- drivers/gpu/drm/i915/intel_dp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index efe18269115f4..c9c416389d0ef 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -441,7 +441,7 @@ static void pps_lock(struct intel_dp *intel_dp) struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); /* - * See vlv_power_sequencer_reset() why we need + * See intel_power_sequencer_reset() why we need * a power domain reference here. */ intel_display_power_get(dev_priv, intel_dp->aux_power_domain); -- GitLab From 274b2462a0493ba73957346fe5fca168e3190b53 Mon Sep 17 00:00:00 2001 From: Tina Zhang Date: Tue, 14 Nov 2017 10:25:12 +0000 Subject: [PATCH 0229/1471] drm/i915: Object w/o backing storage is banned by -ENXIO -ENXIO should be returned when operations are banned from changing backing storage of objects without backing storage. v4: - update "Reviewed-by". (Joonas) v3: - separate this patch from "Introduce GEM proxy" patch-set. (Joonas) v2: - update the patch description and subject to just mention objects w/o backing storage, instead of "GEM proxy". (Joonas) Signed-off-by: Tina Zhang Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-1-git-send-email-tina.zhang@intel.com Reviewed-by: Joonas Lahtinen Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index fa9e5130c9863..a70fbfe8555ec 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1725,7 +1725,7 @@ i915_gem_mmap_ioctl(struct drm_device *dev, void *data, */ if (!obj->base.filp) { i915_gem_object_put(obj); - return -EINVAL; + return -ENXIO; } addr = vm_mmap(obj->base.filp, 0, args->size, -- GitLab From a03f395ad78f883df490234366dd4e4fc922d174 Mon Sep 17 00:00:00 2001 From: Tina Zhang Date: Tue, 14 Nov 2017 10:25:13 +0000 Subject: [PATCH 0230/1471] drm/i915: Introduce GEM proxy GEM proxy is a kind of GEM, whose backing physical memory is pinned and produced by guest VM and is used by host as read only. With GEM proxy, host is able to access guest physical memory through GEM object interface. As GEM proxy is such a special kind of GEM, a new flag I915_GEM_OBJECT_IS_PROXY is introduced to ban host from changing the backing storage of GEM proxy. v3: - update "Reviewed-by". (Joonas) v2: - return -ENXIO when pin and map pages of GEM proxy to kernel space. (Chris) Here are the histories of this patch in "Dma-buf support for Gvt-g" patch-set: v14: - return -ENXIO when gem proxy object is banned by ioctl. (Chris) (Daniel) v13: - add comments to GEM proxy. (Chris) - don't ban GEM proxy in i915_gem_sw_finish_ioctl. (Chris) - check GEM proxy bar after finishing i915_gem_object_wait. (Chris) - remove GEM proxy bar in i915_gem_madvise_ioctl. v6: - add gem proxy barrier in the following ioctls. (Chris) i915_gem_set_caching_ioctl i915_gem_set_domain_ioctl i915_gem_sw_finish_ioctl i915_gem_set_tiling_ioctl i915_gem_madvise_ioctl Signed-off-by: Tina Zhang Reviewed-by: Joonas Lahtinen #v1 Reviewed-by: Chris Wilson Cc: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/1510555798-21079-2-git-send-email-tina.zhang@intel.com Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171114102513.22269-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 31 ++++++++++++++++++++++++-- drivers/gpu/drm/i915/i915_gem_object.h | 11 +++++++-- drivers/gpu/drm/i915/i915_gem_tiling.c | 9 ++++++++ 3 files changed, 47 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index a70fbfe8555ec..a7979b74ce21e 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1619,7 +1619,19 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data, if (err) goto out; - /* Flush and acquire obj->pages so that we are coherent through + /* + * Proxy objects do not control access to the backing storage, ergo + * they cannot be used as a means to manipulate the cache domain + * tracking for that backing storage. The proxy object is always + * considered to be outside of any cache domain. + */ + if (i915_gem_object_is_proxy(obj)) { + err = -ENXIO; + goto out; + } + + /* + * Flush and acquire obj->pages so that we are coherent through * direct access in memory with previous cached writes through * shmemfs and that our cache domain tracking remains valid. * For example, if the obj->filp was moved to swap without us @@ -1675,6 +1687,11 @@ i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data, if (!obj) return -ENOENT; + /* + * Proxy objects are barred from CPU access, so there is no + * need to ban sw_finish as it is a nop. + */ + /* Pinned buffers may be scanout, so flush the cache */ i915_gem_object_flush_if_display(obj); i915_gem_object_put(obj); @@ -2669,7 +2686,8 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj, void *ptr; int ret; - GEM_BUG_ON(!i915_gem_object_has_struct_page(obj)); + if (unlikely(!i915_gem_object_has_struct_page(obj))) + return ERR_PTR(-ENXIO); ret = mutex_lock_interruptible(&obj->mm.lock); if (ret) @@ -3888,6 +3906,15 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data, if (!obj) return -ENOENT; + /* + * The caching mode of proxy object is handled by its generator, and + * not allowed to be changed by userspace. + */ + if (i915_gem_object_is_proxy(obj)) { + ret = -ENXIO; + goto out; + } + if (obj->cache_level == level) goto out; diff --git a/drivers/gpu/drm/i915/i915_gem_object.h b/drivers/gpu/drm/i915/i915_gem_object.h index 63ce38c1cce9c..19fb28c177d80 100644 --- a/drivers/gpu/drm/i915/i915_gem_object.h +++ b/drivers/gpu/drm/i915/i915_gem_object.h @@ -53,8 +53,9 @@ struct i915_lut_handle { struct drm_i915_gem_object_ops { unsigned int flags; -#define I915_GEM_OBJECT_HAS_STRUCT_PAGE BIT(0) -#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(1) +#define I915_GEM_OBJECT_HAS_STRUCT_PAGE BIT(0) +#define I915_GEM_OBJECT_IS_SHRINKABLE BIT(1) +#define I915_GEM_OBJECT_IS_PROXY BIT(2) /* Interface between the GEM object and its backing storage. * get_pages() is called once prior to the use of the associated set @@ -361,6 +362,12 @@ i915_gem_object_is_shrinkable(const struct drm_i915_gem_object *obj) return obj->ops->flags & I915_GEM_OBJECT_IS_SHRINKABLE; } +static inline bool +i915_gem_object_is_proxy(const struct drm_i915_gem_object *obj) +{ + return obj->ops->flags & I915_GEM_OBJECT_IS_PROXY; +} + static inline bool i915_gem_object_is_active(const struct drm_i915_gem_object *obj) { diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 1294cf695df05..b85d7ebd9beea 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -345,6 +345,15 @@ i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data, if (!obj) return -ENOENT; + /* + * The tiling mode of proxy objects is handled by its generator, and + * not allowed to be changed by userspace. + */ + if (i915_gem_object_is_proxy(obj)) { + err = -ENXIO; + goto err; + } + if (!i915_tiling_ok(obj, args->tiling_mode, args->stride)) { err = -EINVAL; goto err; -- GitLab From 4036c78ccf6bf411d09dbf3ec9a9cc7c1838b7b2 Mon Sep 17 00:00:00 2001 From: James Ausmus Date: Mon, 13 Nov 2017 10:11:28 -0800 Subject: [PATCH 0231/1471] drm/i915/glk: Refactor handling of PLANE_COLOR_CTL for GLK+ MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since GLK, some plane configuration settings have moved to the PLANE_COLOR_CTL register. Refactor handling of the register to work like PLANE_CTL. This also allows us to fix the set/read of the plane Alpha Mode for GLK+. v2: Adjust ordering of platform checks to be newest->oldest, drop redundant comment about alpha blending. (Ville) v3: Move Alpha Mode bits out of skl_plane_ctl_format into skl_plane_ctl_alpha, and drop glk_plane_ctl_format, drop initialization of state->color_ctl on platforms that don't use it, and drop color_ctl local var. (Ville) v4: Consolidate skl_plane_ctl_format switch statement on formats that return the same settings. (Ville) Signed-off-by: James Ausmus Cc: Paulo Zanoni Cc: Ville Syrjälä Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171113181128.2926-1-james.ausmus@intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_reg.h | 12 +++-- drivers/gpu/drm/i915/intel_display.c | 73 ++++++++++++++++++++++------ drivers/gpu/drm/i915/intel_drv.h | 5 ++ drivers/gpu/drm/i915/intel_sprite.c | 11 ++--- 4 files changed, 76 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index e52e74db4c7c3..d7e24426120d0 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -6297,7 +6297,7 @@ enum { #define _PLANE_CTL_2_A 0x70280 #define _PLANE_CTL_3_A 0x70380 #define PLANE_CTL_ENABLE (1 << 31) -#define PLANE_CTL_PIPE_GAMMA_ENABLE (1 << 30) +#define PLANE_CTL_PIPE_GAMMA_ENABLE (1 << 30) /* Pre-GLK */ #define PLANE_CTL_FORMAT_MASK (0xf << 24) #define PLANE_CTL_FORMAT_YUV422 ( 0 << 24) #define PLANE_CTL_FORMAT_NV12 ( 1 << 24) @@ -6307,7 +6307,7 @@ enum { #define PLANE_CTL_FORMAT_AYUV ( 8 << 24) #define PLANE_CTL_FORMAT_INDEXED ( 12 << 24) #define PLANE_CTL_FORMAT_RGB_565 ( 14 << 24) -#define PLANE_CTL_PIPE_CSC_ENABLE (1 << 23) +#define PLANE_CTL_PIPE_CSC_ENABLE (1 << 23) /* Pre-GLK */ #define PLANE_CTL_KEY_ENABLE_MASK (0x3 << 21) #define PLANE_CTL_KEY_ENABLE_SOURCE ( 1 << 21) #define PLANE_CTL_KEY_ENABLE_DESTINATION ( 2 << 21) @@ -6320,13 +6320,13 @@ enum { #define PLANE_CTL_YUV422_VYUY ( 3 << 16) #define PLANE_CTL_DECOMPRESSION_ENABLE (1 << 15) #define PLANE_CTL_TRICKLE_FEED_DISABLE (1 << 14) -#define PLANE_CTL_PLANE_GAMMA_DISABLE (1 << 13) +#define PLANE_CTL_PLANE_GAMMA_DISABLE (1 << 13) /* Pre-GLK */ #define PLANE_CTL_TILED_MASK (0x7 << 10) #define PLANE_CTL_TILED_LINEAR ( 0 << 10) #define PLANE_CTL_TILED_X ( 1 << 10) #define PLANE_CTL_TILED_Y ( 4 << 10) #define PLANE_CTL_TILED_YF ( 5 << 10) -#define PLANE_CTL_ALPHA_MASK (0x3 << 4) +#define PLANE_CTL_ALPHA_MASK (0x3 << 4) /* Pre-GLK */ #define PLANE_CTL_ALPHA_DISABLE ( 0 << 4) #define PLANE_CTL_ALPHA_SW_PREMULTIPLY ( 2 << 4) #define PLANE_CTL_ALPHA_HW_PREMULTIPLY ( 3 << 4) @@ -6366,6 +6366,10 @@ enum { #define PLANE_COLOR_PIPE_GAMMA_ENABLE (1 << 30) #define PLANE_COLOR_PIPE_CSC_ENABLE (1 << 23) #define PLANE_COLOR_PLANE_GAMMA_DISABLE (1 << 13) +#define PLANE_COLOR_ALPHA_MASK (0x3 << 4) +#define PLANE_COLOR_ALPHA_DISABLE (0 << 4) +#define PLANE_COLOR_ALPHA_SW_PREMULTIPLY (2 << 4) +#define PLANE_COLOR_ALPHA_HW_PREMULTIPLY (3 << 4) #define _PLANE_BUF_CFG_1_A 0x7027c #define _PLANE_BUF_CFG_2_A 0x7037c #define _PLANE_NV12_BUF_CFG_1_A 0x70278 diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0ebf3f283b877..ed6a4a8d92734 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3432,20 +3432,11 @@ static u32 skl_plane_ctl_format(uint32_t pixel_format) case DRM_FORMAT_RGB565: return PLANE_CTL_FORMAT_RGB_565; case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ABGR8888: return PLANE_CTL_FORMAT_XRGB_8888 | PLANE_CTL_ORDER_RGBX; case DRM_FORMAT_XRGB8888: - return PLANE_CTL_FORMAT_XRGB_8888; - /* - * XXX: For ARBG/ABGR formats we default to expecting scanout buffers - * to be already pre-multiplied. We need to add a knob (or a different - * DRM_FORMAT) for user-space to configure that. - */ - case DRM_FORMAT_ABGR8888: - return PLANE_CTL_FORMAT_XRGB_8888 | PLANE_CTL_ORDER_RGBX | - PLANE_CTL_ALPHA_SW_PREMULTIPLY; case DRM_FORMAT_ARGB8888: - return PLANE_CTL_FORMAT_XRGB_8888 | - PLANE_CTL_ALPHA_SW_PREMULTIPLY; + return PLANE_CTL_FORMAT_XRGB_8888; case DRM_FORMAT_XRGB2101010: return PLANE_CTL_FORMAT_XRGB_2101010; case DRM_FORMAT_XBGR2101010: @@ -3465,6 +3456,33 @@ static u32 skl_plane_ctl_format(uint32_t pixel_format) return 0; } +/* + * XXX: For ARBG/ABGR formats we default to expecting scanout buffers + * to be already pre-multiplied. We need to add a knob (or a different + * DRM_FORMAT) for user-space to configure that. + */ +static u32 skl_plane_ctl_alpha(uint32_t pixel_format) +{ + switch (pixel_format) { + case DRM_FORMAT_ABGR8888: + case DRM_FORMAT_ARGB8888: + return PLANE_CTL_ALPHA_SW_PREMULTIPLY; + default: + return PLANE_CTL_ALPHA_DISABLE; + } +} + +static u32 glk_plane_color_ctl_alpha(uint32_t pixel_format) +{ + switch (pixel_format) { + case DRM_FORMAT_ABGR8888: + case DRM_FORMAT_ARGB8888: + return PLANE_COLOR_ALPHA_SW_PREMULTIPLY; + default: + return PLANE_COLOR_ALPHA_DISABLE; + } +} + static u32 skl_plane_ctl_tiling(uint64_t fb_modifier) { switch (fb_modifier) { @@ -3521,7 +3539,8 @@ u32 skl_plane_ctl(const struct intel_crtc_state *crtc_state, plane_ctl = PLANE_CTL_ENABLE; - if (!IS_GEMINILAKE(dev_priv) && !IS_CANNONLAKE(dev_priv)) { + if (INTEL_GEN(dev_priv) < 10 && !IS_GEMINILAKE(dev_priv)) { + plane_ctl |= skl_plane_ctl_alpha(fb->format->format); plane_ctl |= PLANE_CTL_PIPE_GAMMA_ENABLE | PLANE_CTL_PIPE_CSC_ENABLE | @@ -3540,6 +3559,20 @@ u32 skl_plane_ctl(const struct intel_crtc_state *crtc_state, return plane_ctl; } +u32 glk_plane_color_ctl(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) +{ + const struct drm_framebuffer *fb = plane_state->base.fb; + u32 plane_color_ctl = 0; + + plane_color_ctl |= PLANE_COLOR_PIPE_GAMMA_ENABLE; + plane_color_ctl |= PLANE_COLOR_PIPE_CSC_ENABLE; + plane_color_ctl |= PLANE_COLOR_PLANE_GAMMA_DISABLE; + plane_color_ctl |= glk_plane_color_ctl_alpha(fb->format->format); + + return plane_color_ctl; +} + static int __intel_display_resume(struct drm_device *dev, struct drm_atomic_state *state, @@ -8426,7 +8459,7 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - u32 val, base, offset, stride_mult, tiling; + u32 val, base, offset, stride_mult, tiling, alpha; int pipe = crtc->pipe; int fourcc, pixel_format; unsigned int aligned_height; @@ -8448,9 +8481,16 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, goto error; pixel_format = val & PLANE_CTL_FORMAT_MASK; + + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { + alpha = I915_READ(PLANE_COLOR_CTL(pipe, 0)); + alpha &= PLANE_COLOR_ALPHA_MASK; + } else { + alpha = val & PLANE_CTL_ALPHA_MASK; + } + fourcc = skl_format_to_fourcc(pixel_format, - val & PLANE_CTL_ORDER_RGBX, - val & PLANE_CTL_ALPHA_MASK); + val & PLANE_CTL_ORDER_RGBX, alpha); fb->format = drm_format_info(fourcc); tiling = val & PLANE_CTL_TILED_MASK; @@ -12853,6 +12893,9 @@ intel_check_primary_plane(struct intel_plane *plane, state->ctl = i9xx_plane_ctl(crtc_state, state); } + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + state->color_ctl = glk_plane_color_ctl(crtc_state, state); + return 0; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 10bec8796239b..e9b66e0cb6477 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -425,6 +425,9 @@ struct intel_plane_state { /* plane control register */ u32 ctl; + /* plane color control register */ + u32 color_ctl; + /* * scaler_id * = -1 : not using a scaler @@ -1503,6 +1506,8 @@ static inline u32 intel_plane_ggtt_offset(const struct intel_plane_state *state) return i915_ggtt_offset(state->vma); } +u32 glk_plane_color_ctl(const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state); u32 skl_plane_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state); u32 skl_plane_stride(const struct drm_framebuffer *fb, int plane, diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 4fcf80ca91dd4..ce615704982ae 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -263,13 +263,9 @@ skl_update_plane(struct intel_plane *plane, spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); - if (IS_GEMINILAKE(dev_priv) || IS_CANNONLAKE(dev_priv)) { + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) I915_WRITE_FW(PLANE_COLOR_CTL(pipe, plane_id), - PLANE_COLOR_PIPE_GAMMA_ENABLE | - PLANE_COLOR_PIPE_CSC_ENABLE | - PLANE_COLOR_PLANE_GAMMA_DISABLE); - } - + plane_state->color_ctl); if (key->flags) { I915_WRITE_FW(PLANE_KEYVAL(pipe, plane_id), key->min_value); I915_WRITE_FW(PLANE_KEYMAX(pipe, plane_id), key->max_value); @@ -978,6 +974,9 @@ intel_check_sprite_plane(struct intel_plane *plane, state->ctl = g4x_sprite_ctl(crtc_state, state); } + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) + state->color_ctl = glk_plane_color_ctl(crtc_state, state); + return 0; } -- GitLab From 3657e92762c42b098cb7057ac86b1c7cd30336e5 Mon Sep 17 00:00:00 2001 From: Mika Kahola Date: Thu, 9 Nov 2017 10:37:50 +0200 Subject: [PATCH 0232/1471] drm/i915: Generalize transcoder looping MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit To make looping through transcoders in intel_ddi.c more generic, let's switch to use 'for_each_pipe()' macro to do this. v2: Add a notion that we are dealing with transcoders instead of pipes (Jani) Reviewed-by: Ville Syrjälä Signed-off-by: Mika Kahola Link: https://patchwork.freedesktop.org/patch/msgid/1510216670-16848-1-git-send-email-mika.kahola@intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_ddi.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c index 65ab55496ab74..eff3b51872ebc 100644 --- a/drivers/gpu/drm/i915/intel_ddi.c +++ b/drivers/gpu/drm/i915/intel_ddi.c @@ -1681,8 +1681,8 @@ bool intel_ddi_get_hw_state(struct intel_encoder *encoder, struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); enum port port = encoder->port; + enum pipe p; u32 tmp; - int i; bool ret; if (!intel_display_power_get_if_enabled(dev_priv, @@ -1717,15 +1717,17 @@ bool intel_ddi_get_hw_state(struct intel_encoder *encoder, goto out; } - for (i = TRANSCODER_A; i <= TRANSCODER_C; i++) { - tmp = I915_READ(TRANS_DDI_FUNC_CTL(i)); + for_each_pipe(dev_priv, p) { + enum transcoder cpu_transcoder = (enum transcoder) p; + + tmp = I915_READ(TRANS_DDI_FUNC_CTL(cpu_transcoder)); if ((tmp & TRANS_DDI_PORT_MASK) == TRANS_DDI_SELECT_PORT(port)) { if ((tmp & TRANS_DDI_MODE_SELECT_MASK) == TRANS_DDI_MODE_SELECT_DP_MST) goto out; - *pipe = i; + *pipe = p; ret = true; goto out; -- GitLab From f577a03ba920df1a9163221d4823627999c59bfd Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Mon, 13 Nov 2017 23:34:53 +0000 Subject: [PATCH 0233/1471] drm/i915: fix 64bit divide MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ERROR: "__udivdi3" [drivers/gpu/drm/i915/i915.ko] undefined! ERROR: "__divdi3" [drivers/gpu/drm/i915/i915.ko] undefined! Store the frequency in kHz and drop 64bit divisions. v2: Use div64_u64 (Matthew) v3: store frequency in kHz to avoid 64bit divs (Chris/Ville) Fixes: dab9178333 ("drm/i915: expose command stream timestamp frequency to userspace") Reported-by: Matthew Auld Signed-off-by: Lionel Landwerlin Link: https://patchwork.freedesktop.org/patch/msgid/20171113233455.12085-3-lionel.g.landwerlin@intel.com Reviewed-by: Ewelina Musial Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_debugfs.c | 4 ++-- drivers/gpu/drm/i915/i915_drv.c | 2 +- drivers/gpu/drm/i915/i915_drv.h | 2 +- drivers/gpu/drm/i915/intel_device_info.c | 29 ++++++++++++------------ 4 files changed, 18 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 82ff186108f18..6ba08b0c1c226 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3269,8 +3269,8 @@ static int i915_engine_info(struct seq_file *m, void *unused) yesno(dev_priv->gt.awake)); seq_printf(m, "Global active requests: %d\n", dev_priv->gt.active_requests); - seq_printf(m, "CS timestamp frequency: %llu Hz\n", - dev_priv->info.cs_timestamp_frequency); + seq_printf(m, "CS timestamp frequency: %u kHz\n", + dev_priv->info.cs_timestamp_frequency_khz); p = drm_seq_file_printer(m); for_each_engine(engine, dev_priv, id) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 42813f4247e29..171b86f37878c 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -420,7 +420,7 @@ static int i915_getparam(struct drm_device *dev, void *data, return -ENODEV; break; case I915_PARAM_CS_TIMESTAMP_FREQUENCY: - value = INTEL_INFO(dev_priv)->cs_timestamp_frequency; + value = 1000 * INTEL_INFO(dev_priv)->cs_timestamp_frequency_khz; break; default: DRM_DEBUG("Unknown parameter %d\n", param->param); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index b538df740ac39..2158a758a17d7 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -885,7 +885,7 @@ struct intel_device_info { /* Slice/subslice/EU info */ struct sseu_dev_info sseu; - u64 cs_timestamp_frequency; + u32 cs_timestamp_frequency_khz; struct color_luts { u16 degamma_lut_size; diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c index 78bf7374fbdde..f3e4940fed499 100644 --- a/drivers/gpu/drm/i915/intel_device_info.c +++ b/drivers/gpu/drm/i915/intel_device_info.c @@ -329,29 +329,28 @@ static void broadwell_sseu_info_init(struct drm_i915_private *dev_priv) sseu->has_eu_pg = 0; } -static u64 read_reference_ts_freq(struct drm_i915_private *dev_priv) +static u32 read_reference_ts_freq(struct drm_i915_private *dev_priv) { u32 ts_override = I915_READ(GEN9_TIMESTAMP_OVERRIDE); - u64 base_freq, frac_freq; + u32 base_freq, frac_freq; base_freq = ((ts_override & GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_MASK) >> GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DIVIDER_SHIFT) + 1; - base_freq *= 1000000; + base_freq *= 1000; frac_freq = ((ts_override & GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_MASK) >> GEN9_TIMESTAMP_OVERRIDE_US_COUNTER_DENOMINATOR_SHIFT); - if (frac_freq != 0) - frac_freq = 1000000 / (frac_freq + 1); + frac_freq = 1000 / (frac_freq + 1); return base_freq + frac_freq; } -static u64 read_timestamp_frequency(struct drm_i915_private *dev_priv) +static u32 read_timestamp_frequency(struct drm_i915_private *dev_priv) { - u64 f12_5_mhz = 12500000; - u64 f19_2_mhz = 19200000; - u64 f24_mhz = 24000000; + u32 f12_5_mhz = 12500; + u32 f19_2_mhz = 19200; + u32 f24_mhz = 24000; if (INTEL_GEN(dev_priv) <= 4) { /* PRMs say: @@ -360,7 +359,7 @@ static u64 read_timestamp_frequency(struct drm_i915_private *dev_priv) * hclks." (through the “Clocking Configuration” * (“CLKCFG”) MCHBAR register) */ - return (dev_priv->rawclk_freq * 1000) / 16; + return dev_priv->rawclk_freq / 16; } else if (INTEL_GEN(dev_priv) <= 8) { /* PRMs say: * @@ -371,7 +370,7 @@ static u64 read_timestamp_frequency(struct drm_i915_private *dev_priv) return f12_5_mhz; } else if (INTEL_GEN(dev_priv) <= 9) { u32 ctc_reg = I915_READ(CTC_MODE); - u64 freq = 0; + u32 freq = 0; if ((ctc_reg & CTC_SOURCE_PARAMETER_MASK) == CTC_SOURCE_DIVIDE_LOGIC) { freq = read_reference_ts_freq(dev_priv); @@ -389,7 +388,7 @@ static u64 read_timestamp_frequency(struct drm_i915_private *dev_priv) return freq; } else if (INTEL_GEN(dev_priv) <= 10) { u32 ctc_reg = I915_READ(CTC_MODE); - u64 freq = 0; + u32 freq = 0; u32 rpm_config_reg = 0; /* First figure out the reference frequency. There are 2 ways @@ -553,7 +552,7 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) gen10_sseu_info_init(dev_priv); /* Initialize command stream timestamp frequency */ - info->cs_timestamp_frequency = read_timestamp_frequency(dev_priv); + info->cs_timestamp_frequency_khz = read_timestamp_frequency(dev_priv); DRM_DEBUG_DRIVER("slice mask: %04x\n", info->sseu.slice_mask); DRM_DEBUG_DRIVER("slice total: %u\n", hweight8(info->sseu.slice_mask)); @@ -570,6 +569,6 @@ void intel_device_info_runtime_init(struct drm_i915_private *dev_priv) info->sseu.has_subslice_pg ? "y" : "n"); DRM_DEBUG_DRIVER("has EU power gating: %s\n", info->sseu.has_eu_pg ? "y" : "n"); - DRM_DEBUG_DRIVER("CS timestamp frequency: %llu\n", - info->cs_timestamp_frequency); + DRM_DEBUG_DRIVER("CS timestamp frequency: %u kHz\n", + info->cs_timestamp_frequency_khz); } -- GitLab From 34991bd48c927712678d0cea77628328f9046923 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sat, 11 Nov 2017 10:03:36 +0000 Subject: [PATCH 0234/1471] drm/i915: Unify SLICE_UNIT_LEVEL_CLKGATE w/a for cnl gem_workarounds reports that the SLICE_UNIT_LEVEL_CLKGATE write isn't sticking. Commit 0a60797a0efb ("drm/i915: Implement ReadHitWriteOnlyDisable.") presumes that SLICE_UNIT_LEVEL_CLKGATE is a masked register in the context image, but commit 90007bca6162 ("drm/i915/cnl: Introduce initial Cannonlake Workarounds.") lists it as an ordering unmasked register. The masked write will be losing the default settings if we trust the original commit. That gem_workarounds reports the value is lost entirely is more worrying though -- but it clearly suggests that it is not a masked register in the context image, so unify both w/a to use the original rmw. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103705 Fixes: 0a60797a0efb ("drm/i915: Implement ReadHitWriteOnlyDisable.") References: 90007bca6162 ("drm/i915/cnl: Introduce initial Cannonlake Workarounds.") Signed-off-by: Chris Wilson Cc: Rafael Antognolli Cc: Rodrigo Vivi Cc: Oscar Mateo Cc: Mika Kuoppala Cc: Jani Nikula Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171111100336.11020-1-chris@chris-wilson.co.uk Reviewed-by: Rafael Antognolli --- drivers/gpu/drm/i915/intel_engine_cs.c | 3 --- drivers/gpu/drm/i915/intel_pm.c | 8 +++++--- 2 files changed, 5 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 70bbe8ef8f54f..1bd9462d206f6 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1326,9 +1326,6 @@ static int cnl_init_workarounds(struct intel_engine_cs *engine) WA_SET_FIELD_MASKED(GEN8_CS_CHICKEN1, GEN9_PREEMPT_GPGPU_LEVEL_MASK, GEN9_PREEMPT_GPGPU_COMMAND_LEVEL); - /* ReadHitWriteOnlyDisable: cnl */ - WA_SET_BIT_MASKED(SLICE_UNIT_LEVEL_CLKGATE, RCCUNIT_CLKGATE_DIS); - /* WaEnablePreemptionGranularityControlByUMD:cnl */ I915_WRITE(GEN7_FF_SLICE_CS_CHICKEN1, _MASKED_BIT_ENABLE(GEN9_FFSC_PERCTX_PREEMPT_CTRL)); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 8e7f02e6008aa..8c69ec9eb6eec 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -8471,11 +8471,13 @@ static void cnl_init_clock_gating(struct drm_i915_private *dev_priv) I915_WRITE(DISP_ARB_CTL, I915_READ(DISP_ARB_CTL) | DISP_FBC_MEMORY_WAKE); + val = I915_READ(SLICE_UNIT_LEVEL_CLKGATE); + /* ReadHitWriteOnlyDisable:cnl */ + val |= RCCUNIT_CLKGATE_DIS; /* WaSarbUnitClockGatingDisable:cnl (pre-prod) */ if (IS_CNL_REVID(dev_priv, CNL_REVID_A0, CNL_REVID_B0)) - I915_WRITE(SLICE_UNIT_LEVEL_CLKGATE, - I915_READ(SLICE_UNIT_LEVEL_CLKGATE) | - SARBUNIT_CLKGATE_DIS); + val |= SARBUNIT_CLKGATE_DIS; + I915_WRITE(SLICE_UNIT_LEVEL_CLKGATE, val); /* Display WA #1133: WaFbcSkipSegments:cnl */ val = I915_READ(ILK_DPFC_CHICKEN); -- GitLab From f3e2b2c540503fb6087d779b0633a10c79cb6d5c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 13:43:39 +0000 Subject: [PATCH 0235/1471] drm/i915: Remove pre-production Broxton register workarounds We've begun excluding pre-production Broxton machines since commit 0102ba1fd8af ("drm/i915: Add early BXT sdv to the list of preproduction machines"), now remove the list of workaround register values for those early machines. Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20170927093325.24206-1-chris@chris-wilson.co.uk Reviewed-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171114134340.5439-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 56 +------------------------- 1 file changed, 1 insertion(+), 55 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 1bd9462d206f6..580e79058433a 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1020,22 +1020,6 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine) WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3, GEN9_DISABLE_OCL_OOB_SUPPRESS_LOGIC); - /* WaDisableDgMirrorFixInHalfSliceChicken5:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) - WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5, - GEN9_DG_MIRROR_FIX_ENABLE); - - /* WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { - WA_SET_BIT_MASKED(GEN7_COMMON_SLICE_CHICKEN1, - GEN9_RHWO_OPTIMIZATION_DISABLE); - /* - * WA also requires GEN9_SLICE_COMMON_ECO_CHICKEN0[14:14] to be set - * but we do that in per ctx batchbuffer as there is an issue - * with this register not getting restored on ctx restore - */ - } - /* WaEnableYV12BugFixInHalfSliceChicken7:skl,bxt,kbl,glk,cfl */ /* WaEnableSamplerGPGPUPreemptionSupport:skl,bxt,kbl,cfl */ WA_SET_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN7, @@ -1051,11 +1035,6 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine) WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5, GEN9_CCS_TLB_PREFETCH_ENABLE); - /* WaDisableMaskBasedCammingInRCC:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) - WA_SET_BIT_MASKED(SLICE_ECO_CHICKEN0, - PIXEL_MASK_CAMMING_DISABLE); - /* WaForceContextSaveRestoreNonCoherent:skl,bxt,kbl,cfl */ WA_SET_BIT_MASKED(HDC_CHICKEN0, HDC_FORCE_CONTEXT_SAVE_RESTORE_NON_COHERENT | @@ -1085,8 +1064,7 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine) /* WaDisableSamplerPowerBypassForSOPingPong:skl,bxt,kbl,cfl */ if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv) || - IS_COFFEELAKE(dev_priv) || - IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) + IS_COFFEELAKE(dev_priv)) WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3, GEN8_SAMPLER_POWER_BYPASS_DIS); @@ -1216,17 +1194,6 @@ static int bxt_init_workarounds(struct intel_engine_cs *engine) if (ret) return ret; - /* WaStoreMultiplePTEenable:bxt */ - /* This is a requirement according to Hardware specification */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) - I915_WRITE(TILECTL, I915_READ(TILECTL) | TILECTL_TLBPF); - - /* WaSetClckGatingDisableMedia:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { - I915_WRITE(GEN7_MISCCPCTL, (I915_READ(GEN7_MISCCPCTL) & - ~GEN8_DOP_CLOCK_GATE_MEDIA_ENABLE)); - } - /* WaDisableThreadStallDopClockGating:bxt */ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, STALL_DOP_GATING_DISABLE); @@ -1237,27 +1204,6 @@ static int bxt_init_workarounds(struct intel_engine_cs *engine) _MASKED_BIT_ENABLE(GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE)); } - /* WaDisableSbeCacheDispatchPortSharing:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) { - WA_SET_BIT_MASKED( - GEN7_HALF_SLICE_CHICKEN1, - GEN7_SBE_SS_CACHE_DISPATCH_PORT_SHARING_DISABLE); - } - - /* WaDisableObjectLevelPreemptionForTrifanOrPolygon:bxt */ - /* WaDisableObjectLevelPreemptionForInstancedDraw:bxt */ - /* WaDisableObjectLevelPreemtionForInstanceId:bxt */ - /* WaDisableLSQCROPERFforOCL:bxt */ - if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) { - ret = wa_ring_whitelist_reg(engine, GEN9_CS_DEBUG_MODE1); - if (ret) - return ret; - - ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4); - if (ret) - return ret; - } - /* WaProgramL3SqcReg1DefaultForPerf:bxt */ if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER)) { u32 val = I915_READ(GEN8_L3SQCREG1); -- GitLab From 70a84f3c6075031dbf004a1610ca2471f4c528aa Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 13:43:40 +0000 Subject: [PATCH 0236/1471] drm/i915: Unconditionally apply the Broxton register workaround set Having removed the preproduction Broxton support (see commit 0102ba1fd8af ("drm/i915: Add early BXT sdv to the list of preproduction machines")), we know we then always need the production Broxton workaround set and do not need a predicate upon revision. Signed-off-by: Chris Wilson Reviewed-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171114134340.5439-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_engine_cs.c | 29 +++++++++++--------------- 1 file changed, 12 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 580e79058433a..a42b738e79e77 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1188,6 +1188,7 @@ static int skl_init_workarounds(struct intel_engine_cs *engine) static int bxt_init_workarounds(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; + u32 val; int ret; ret = gen9_init_workarounds(engine); @@ -1199,29 +1200,23 @@ static int bxt_init_workarounds(struct intel_engine_cs *engine) STALL_DOP_GATING_DISABLE); /* WaDisablePooledEuLoadBalancingFix:bxt */ - if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER)) { - I915_WRITE(FF_SLICE_CS_CHICKEN2, - _MASKED_BIT_ENABLE(GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE)); - } + I915_WRITE(FF_SLICE_CS_CHICKEN2, + _MASKED_BIT_ENABLE(GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE)); /* WaProgramL3SqcReg1DefaultForPerf:bxt */ - if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER)) { - u32 val = I915_READ(GEN8_L3SQCREG1); - val &= ~L3_PRIO_CREDITS_MASK; - val |= L3_GENERAL_PRIO_CREDITS(62) | L3_HIGH_PRIO_CREDITS(2); - I915_WRITE(GEN8_L3SQCREG1, val); - } + val = I915_READ(GEN8_L3SQCREG1); + val &= ~L3_PRIO_CREDITS_MASK; + val |= L3_GENERAL_PRIO_CREDITS(62) | L3_HIGH_PRIO_CREDITS(2); + I915_WRITE(GEN8_L3SQCREG1, val); /* WaToEnableHwFixForPushConstHWBug:bxt */ - if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER)) - WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2, - GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION); + WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2, + GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION); /* WaInPlaceDecompressionHang:bxt */ - if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER)) - I915_WRITE(GEN9_GAMT_ECO_REG_RW_IA, - (I915_READ(GEN9_GAMT_ECO_REG_RW_IA) | - GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS)); + I915_WRITE(GEN9_GAMT_ECO_REG_RW_IA, + (I915_READ(GEN9_GAMT_ECO_REG_RW_IA) | + GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS)); return 0; } -- GitLab From df49ec8223050d510bd243d85c51139fc101029f Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 10 Nov 2017 16:03:19 -0800 Subject: [PATCH 0237/1471] drm/i915: Display WA #1185 WaDisableDARBFClkGating:cnl, glk MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Display is not sending a PMRsp when a PMReq is received at the same time that all planes are turned off. State machine in the dcprunit is stuck in the WAIT4DONE state which means that there is no fill_done. WA: disable arbiter clock gating, set bit [27] of 0x46530 v2: As Ville pointed out, based on the description the issue can happen when disabling the planes, similar to WaRsPkgCStateDisplayPMReq:hsw Also description of the issue was updated on commit message to make it more clear that we need this earlier. v3: Restore comment about possibility to system hang to where we are sure about it, without speculation. (Ville). v4: Remove doubled sob. Actually do v3 changes :/ Cc: Radhakrishna Sripada Cc: Imre Deak Cc: Ville Syrjälä Signed-off-by: Rodrigo Vivi Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171111000319.5040-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_display.c | 27 ++++++++++++++++++--------- 2 files changed, 19 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index d7e24426120d0..cfdf4f821ac3a 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3852,6 +3852,7 @@ enum { * GEN9 clock gating regs */ #define GEN9_CLKGATE_DIS_0 _MMIO(0x46530) +#define DARBF_GATING_DIS (1 << 27) #define PWM2_GATING_DIS (1 << 14) #define PWM1_GATING_DIS (1 << 13) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ed6a4a8d92734..23aa7191024ef 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15148,6 +15148,23 @@ get_encoder_power_domains(struct drm_i915_private *dev_priv) } } +static void intel_early_display_was(struct drm_i915_private *dev_priv) +{ + /* Display WA #1185 WaDisableDARBFClkGating:cnl,glk */ + if (IS_CANNONLAKE(dev_priv) || IS_GEMINILAKE(dev_priv)) + I915_WRITE(GEN9_CLKGATE_DIS_0, I915_READ(GEN9_CLKGATE_DIS_0) | + DARBF_GATING_DIS); + + if (IS_HASWELL(dev_priv)) { + /* + * WaRsPkgCStateDisplayPMReq:hsw + * System hang if this isn't done before disabling all planes! + */ + I915_WRITE(CHICKEN_PAR1_1, + I915_READ(CHICKEN_PAR1_1) | FORCE_ARB_IDLE_PLANES); + } +} + /* Scan out the current hw modeset state, * and sanitizes it to the current state */ @@ -15161,15 +15178,7 @@ intel_modeset_setup_hw_state(struct drm_device *dev, struct intel_encoder *encoder; int i; - if (IS_HASWELL(dev_priv)) { - /* - * WaRsPkgCStateDisplayPMReq:hsw - * System hang if this isn't done before disabling all planes! - */ - I915_WRITE(CHICKEN_PAR1_1, - I915_READ(CHICKEN_PAR1_1) | FORCE_ARB_IDLE_PLANES); - } - + intel_early_display_was(dev_priv); intel_modeset_readout_hw_state(dev); /* HW state is read out, now we need to sanitize this mess. */ -- GitLab From 4d9c62e8ce69d0b0a834282a34bff5ce8eeacb1d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Tue, 14 Nov 2017 15:24:35 +0100 Subject: [PATCH 0238/1471] dma-buf: keep only not signaled fence in reservation_object_add_shared_replace v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The amdgpu issue to also need signaled fences in the reservation objects should be fixed by now. Optimize the list by keeping only the not signaled yet fences around. v2: temporary put the signaled fences at the end of the new container v3: put the old fence at the end of the new container as well. Signed-off-by: Christian König Reviewed-by: Chris Wilson Tested-by: Chris Wilson Signed-off-by: Alex Deucher Link: https://patchwork.freedesktop.org/patch/msgid/20171114142436.1360-1-christian.koenig@amd.com --- drivers/dma-buf/reservation.c | 38 +++++++++++++++++++---------------- 1 file changed, 21 insertions(+), 17 deletions(-) diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c index d90333e0b6d5f..737885e79b741 100644 --- a/drivers/dma-buf/reservation.c +++ b/drivers/dma-buf/reservation.c @@ -145,8 +145,7 @@ reservation_object_add_shared_replace(struct reservation_object *obj, struct reservation_object_list *fobj, struct dma_fence *fence) { - unsigned i; - struct dma_fence *old_fence = NULL; + unsigned i, j, k; dma_fence_get(fence); @@ -162,24 +161,21 @@ reservation_object_add_shared_replace(struct reservation_object *obj, * references from the old struct are carried over to * the new. */ - fobj->shared_count = old->shared_count; - - for (i = 0; i < old->shared_count; ++i) { + for (i = 0, j = 0, k = fobj->shared_max; i < old->shared_count; ++i) { struct dma_fence *check; check = rcu_dereference_protected(old->shared[i], reservation_object_held(obj)); - if (!old_fence && check->context == fence->context) { - old_fence = check; - RCU_INIT_POINTER(fobj->shared[i], fence); - } else - RCU_INIT_POINTER(fobj->shared[i], check); - } - if (!old_fence) { - RCU_INIT_POINTER(fobj->shared[fobj->shared_count], fence); - fobj->shared_count++; + if (check->context == fence->context || + dma_fence_is_signaled(check)) + RCU_INIT_POINTER(fobj->shared[--k], check); + else + RCU_INIT_POINTER(fobj->shared[j++], check); } + fobj->shared_count = j; + RCU_INIT_POINTER(fobj->shared[fobj->shared_count], fence); + fobj->shared_count++; done: preempt_disable(); @@ -192,10 +188,18 @@ reservation_object_add_shared_replace(struct reservation_object *obj, write_seqcount_end(&obj->seq); preempt_enable(); - if (old) - kfree_rcu(old, rcu); + if (!old) + return; - dma_fence_put(old_fence); + /* Drop the references to the signaled fences */ + for (i = k; i < fobj->shared_max; ++i) { + struct dma_fence *f; + + f = rcu_dereference_protected(fobj->shared[i], + reservation_object_held(obj)); + dma_fence_put(f); + } + kfree_rcu(old, rcu); } /** -- GitLab From ca25fe5efe4ab43cc5b4f3117a205c281805a5ca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Christian=20K=C3=B6nig?= Date: Tue, 14 Nov 2017 15:24:36 +0100 Subject: [PATCH 0239/1471] dma-buf: try to replace a signaled fence in reservation_object_add_shared_inplace MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The amdgpu issue to also need signaled fences in the reservation objects should be fixed by now. Optimize the handling by replacing a signaled fence when adding a new shared one. Signed-off-by: Christian König Reviewed-by: Chris Wilson Signed-off-by: Alex Deucher Link: https://patchwork.freedesktop.org/patch/msgid/20171114142436.1360-2-christian.koenig@amd.com --- drivers/dma-buf/reservation.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c index 737885e79b741..b759a569b7b8f 100644 --- a/drivers/dma-buf/reservation.c +++ b/drivers/dma-buf/reservation.c @@ -104,7 +104,8 @@ reservation_object_add_shared_inplace(struct reservation_object *obj, struct reservation_object_list *fobj, struct dma_fence *fence) { - u32 i; + struct dma_fence *signaled = NULL; + u32 i, signaled_idx; dma_fence_get(fence); @@ -126,17 +127,28 @@ reservation_object_add_shared_inplace(struct reservation_object *obj, dma_fence_put(old_fence); return; } + + if (!signaled && dma_fence_is_signaled(old_fence)) { + signaled = old_fence; + signaled_idx = i; + } } /* * memory barrier is added by write_seqcount_begin, * fobj->shared_count is protected by this lock too */ - RCU_INIT_POINTER(fobj->shared[fobj->shared_count], fence); - fobj->shared_count++; + if (signaled) { + RCU_INIT_POINTER(fobj->shared[signaled_idx], fence); + } else { + RCU_INIT_POINTER(fobj->shared[fobj->shared_count], fence); + fobj->shared_count++; + } write_seqcount_end(&obj->seq); preempt_enable(); + + dma_fence_put(signaled); } static void -- GitLab From dadcc5e02f675947b22bfd2df14dbf77f6888613 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 24 Aug 2017 22:10:58 +0300 Subject: [PATCH 0240/1471] drm: Fix modifiers_property kernel doc MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The member is called 'modifiers_property' instead of 'modifiers'. Adjust the kernel docs to match. Cc: dri-devel@lists.freedesktop.org Cc: Ben Widawsky Cc: Jason Ekstrand Cc: Daniel Stone Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20170824191100.10949-11-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter --- include/drm/drm_mode_config.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/drm/drm_mode_config.h b/include/drm/drm_mode_config.h index 1b37368416c85..6040c4b73e6d5 100644 --- a/include/drm/drm_mode_config.h +++ b/include/drm/drm_mode_config.h @@ -758,7 +758,7 @@ struct drm_mode_config { bool allow_fb_modifiers; /** - * @modifiers: Plane property to list support modifier/format + * @modifiers_property: Plane property to list support modifier/format * combination. */ struct drm_property *modifiers_property; -- GitLab From 1fbdb980a9f32441638d7571fa35f36841fe93ee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 14 Nov 2017 21:10:19 +0200 Subject: [PATCH 0241/1471] drm/gem-cma-helper: Fix kerneldocs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use the correct name for the function argument in the docs. Cc: Noralf Trønnes Cc: Daniel Vetter Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171114191021.15591-1-ville.syrjala@linux.intel.com Acked-by: Noralf Trønnes Reviewed-by: Alex Deucher --- drivers/gpu/drm/drm_gem_cma_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/drm_gem_cma_helper.c b/drivers/gpu/drm/drm_gem_cma_helper.c index 88ad88719605d..ce67d7d42c665 100644 --- a/drivers/gpu/drm/drm_gem_cma_helper.c +++ b/drivers/gpu/drm/drm_gem_cma_helper.c @@ -401,7 +401,7 @@ EXPORT_SYMBOL_GPL(drm_gem_cma_get_unmapped_area); * drm_gem_cma_print_info() - Print &drm_gem_cma_object info for debugfs * @p: DRM printer * @indent: Tab indentation level - * @gem: GEM object + * @obj: GEM object * * This function can be used as the &drm_driver->gem_print_info callback. * It prints paddr and vaddr for use in e.g. debugfs output. -- GitLab From 3df674585fd174977655f273b471d1aa1e0cefee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 14 Nov 2017 21:10:21 +0200 Subject: [PATCH 0242/1471] drm: Fix kerneldocs for drm_plane modifiers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add the missing kerneldoc for modifiers and modifier_count. Cc: Ben Widawsky Cc: Daniel Vetter Suggested-by: Daniel Vetter Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171114191021.15591-3-ville.syrjala@linux.intel.com Reviewed-by: Alex Deucher --- include/drm/drm_plane.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/include/drm/drm_plane.h b/include/drm/drm_plane.h index 68bf66473faa6..c062d797b9f99 100644 --- a/include/drm/drm_plane.h +++ b/include/drm/drm_plane.h @@ -474,6 +474,8 @@ enum drm_plane_type { * @format_types: array of formats supported by this plane * @format_count: number of formats supported * @format_default: driver hasn't supplied supported formats for the plane + * @modifiers: array of modifiers supported by this plane + * @modifier_count: number of modifiers supported * @old_fb: Temporary tracking of the old fb while a modeset is ongoing. Used by * drm_mode_set_config_internal() to implement correct refcounting. * @funcs: helper functions -- GitLab From b20cd0df152cf4a64e6ae5a1be5f7f346a27eedd Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 14 Nov 2017 16:41:17 -0500 Subject: [PATCH 0243/1471] drm/amdkfd: Cleanup qpd.pqm initialization The PQM doesn't change after process creation. So initialize it in kfd_create_process_device_data. Signed-off-by: Felix Kuehling Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 1 + drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 4 +--- 2 files changed, 2 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 1f5ccd28bd414..1bb9b2643d5af 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -348,6 +348,7 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, INIT_LIST_HEAD(&pdd->qpd.queues_list); INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); pdd->qpd.dqm = dev->dqm; + pdd->qpd.pqm = &p->pqm; pdd->process = p; pdd->bound = PDD_UNBOUND; pdd->already_dequeued = false; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index a3f1e62c60ba9..eeb7726b62991 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -178,10 +178,8 @@ int pqm_create_queue(struct process_queue_manager *pqm, return retval; if (list_empty(&pdd->qpd.queues_list) && - list_empty(&pdd->qpd.priv_queue_list)) { - pdd->qpd.pqm = pqm; + list_empty(&pdd->qpd.priv_queue_list)) dev->dqm->ops.register_process(dev->dqm, &pdd->qpd); - } pqn = kzalloc(sizeof(*pqn), GFP_KERNEL); if (!pqn) { -- GitLab From 449fea612690c3a5ae69d5c09fa7d2ce5773e736 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 14 Nov 2017 16:41:18 -0500 Subject: [PATCH 0244/1471] drm/amdkfd: Add trap handler for CWSR The trap handler is like an interrupt handler running on the GPU compute unit. It is needed for supporting CWSR (compute wave save/restore). This file defines an array with the pre-compiled GFXv8 shader ISA. The assembly code is included for reference in #if 0 ... #endif. Signed-off-by: Shaoyun.liu Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Signed-off-by: Oded Gabbay --- .../drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm | 1384 +++++++++++++++++ 1 file changed, 1384 insertions(+) create mode 100644 drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm new file mode 100644 index 0000000000000..997a383dcb8b7 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm @@ -0,0 +1,1384 @@ +/* + * Copyright 2015-2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#if 0 +HW (VI) source code for CWSR trap handler +#Version 18 + multiple trap handler + +// this performance-optimal version was originally from Seven Xu at SRDC + +// Revison #18 --... +/* Rev History +** #1. Branch from gc dv. //gfxip/gfx8/main/src/test/suites/block/cs/sr/cs_trap_handler.sp3#1,#50, #51, #52-53(Skip, Already Fixed by PV), #54-56(merged),#57-58(mergerd, skiped-already fixed by PV) +** #4. SR Memory Layout: +** 1. VGPR-SGPR-HWREG-{LDS} +** 2. tba_hi.bits.26 - reconfigured as the first wave in tg bits, for defer Save LDS for a threadgroup.. performance concern.. +** #5. Update: 1. Accurate g8sr_ts_save_d timestamp +** #6. Update: 1. Fix s_barrier usage; 2. VGPR s/r using swizzle buffer?(NoNeed, already matched the swizzle pattern, more investigation) +** #7. Update: 1. don't barrier if noLDS +** #8. Branch: 1. Branch to ver#0, which is very similar to gc dv version +** 2. Fix SQ issue by s_sleep 2 +** #9. Update: 1. Fix scc restore failed issue, restore wave_status at last +** 2. optimize s_buffer save by burst 16sgprs... +** #10. Update 1. Optimize restore sgpr by busrt 16 sgprs. +** #11. Update 1. Add 2 more timestamp for debug version +** #12. Update 1. Add VGPR SR using DWx4, some case improve and some case drop performance +** #13. Integ 1. Always use MUBUF for PV trap shader... +** #14. Update 1. s_buffer_store soft clause... +** #15. Update 1. PERF - sclar write with glc:0/mtype0 to allow L2 combine. perf improvement a lot. +** #16. Update 1. PRRF - UNROLL LDS_DMA got 2500cycle save in IP tree +** #17. Update 1. FUNC - LDS_DMA has issues while ATC, replace with ds_read/buffer_store for save part[TODO restore part] +** 2. PERF - Save LDS before save VGPR to cover LDS save long latency... +** #18. Update 1. FUNC - Implicitly estore STATUS.VCCZ, which is not writable by s_setreg_b32 +** 2. FUNC - Handle non-CWSR traps +*/ + +var G8SR_WDMEM_HWREG_OFFSET = 0 +var G8SR_WDMEM_SGPR_OFFSET = 128 // in bytes + +// Keep definition same as the app shader, These 2 time stamps are part of the app shader... Should before any Save and after restore. + +var G8SR_DEBUG_TIMESTAMP = 0 +var G8SR_DEBUG_TS_SAVE_D_OFFSET = 40*4 // ts_save_d timestamp offset relative to SGPR_SR_memory_offset +var s_g8sr_ts_save_s = s[34:35] // save start +var s_g8sr_ts_sq_save_msg = s[36:37] // The save shader send SAVEWAVE msg to spi +var s_g8sr_ts_spi_wrexec = s[38:39] // the SPI write the sr address to SQ +var s_g8sr_ts_save_d = s[40:41] // save end +var s_g8sr_ts_restore_s = s[42:43] // restore start +var s_g8sr_ts_restore_d = s[44:45] // restore end + +var G8SR_VGPR_SR_IN_DWX4 = 0 +var G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 = 0x00100000 // DWx4 stride is 4*4Bytes +var G8SR_RESTORE_BUF_RSRC_WORD1_STRIDE_DWx4 = G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 + + +/*************************************************************************/ +/* control on how to run the shader */ +/*************************************************************************/ +//any hack that needs to be made to run this code in EMU (either because various EMU code are not ready or no compute save & restore in EMU run) +var EMU_RUN_HACK = 0 +var EMU_RUN_HACK_RESTORE_NORMAL = 0 +var EMU_RUN_HACK_SAVE_NORMAL_EXIT = 0 +var EMU_RUN_HACK_SAVE_SINGLE_WAVE = 0 +var EMU_RUN_HACK_SAVE_FIRST_TIME = 0 //for interrupted restore in which the first save is through EMU_RUN_HACK +var EMU_RUN_HACK_SAVE_FIRST_TIME_TBA_LO = 0 //for interrupted restore in which the first save is through EMU_RUN_HACK +var EMU_RUN_HACK_SAVE_FIRST_TIME_TBA_HI = 0 //for interrupted restore in which the first save is through EMU_RUN_HACK +var SAVE_LDS = 1 +var WG_BASE_ADDR_LO = 0x9000a000 +var WG_BASE_ADDR_HI = 0x0 +var WAVE_SPACE = 0x5000 //memory size that each wave occupies in workgroup state mem +var CTX_SAVE_CONTROL = 0x0 +var CTX_RESTORE_CONTROL = CTX_SAVE_CONTROL +var SIM_RUN_HACK = 0 //any hack that needs to be made to run this code in SIM (either because various RTL code are not ready or no compute save & restore in RTL run) +var SGPR_SAVE_USE_SQC = 1 //use SQC D$ to do the write +var USE_MTBUF_INSTEAD_OF_MUBUF = 0 //because TC EMU currently asserts on 0 of // overload DFMT field to carry 4 more bits of stride for MUBUF opcodes +var SWIZZLE_EN = 0 //whether we use swizzled buffer addressing + +/**************************************************************************/ +/* variables */ +/**************************************************************************/ +var SQ_WAVE_STATUS_INST_ATC_SHIFT = 23 +var SQ_WAVE_STATUS_INST_ATC_MASK = 0x00800000 +var SQ_WAVE_STATUS_SPI_PRIO_MASK = 0x00000006 + +var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT = 12 +var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE = 9 +var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT = 8 +var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE = 6 +var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT = 24 +var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE = 3 //FIXME sq.blk still has 4 bits at this time while SQ programming guide has 3 bits + +var SQ_WAVE_TRAPSTS_SAVECTX_MASK = 0x400 +var SQ_WAVE_TRAPSTS_EXCE_MASK = 0x1FF // Exception mask +var SQ_WAVE_TRAPSTS_SAVECTX_SHIFT = 10 +var SQ_WAVE_TRAPSTS_MEM_VIOL_MASK = 0x100 +var SQ_WAVE_TRAPSTS_MEM_VIOL_SHIFT = 8 +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK = 0x3FF +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT = 0x0 +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE = 10 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK = 0xFFFFF800 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT = 11 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE = 21 + +var SQ_WAVE_IB_STS_RCNT_SHIFT = 16 //FIXME +var SQ_WAVE_IB_STS_RCNT_SIZE = 4 //FIXME +var SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT = 15 //FIXME +var SQ_WAVE_IB_STS_FIRST_REPLAY_SIZE = 1 //FIXME +var SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK_NEG = 0x00007FFF //FIXME + +var SQ_BUF_RSRC_WORD1_ATC_SHIFT = 24 +var SQ_BUF_RSRC_WORD3_MTYPE_SHIFT = 27 + + +/* Save */ +var S_SAVE_BUF_RSRC_WORD1_STRIDE = 0x00040000 //stride is 4 bytes +var S_SAVE_BUF_RSRC_WORD3_MISC = 0x00807FAC //SQ_SEL_X/Y/Z/W, BUF_NUM_FORMAT_FLOAT, (0 for MUBUF stride[17:14] when ADD_TID_ENABLE and BUF_DATA_FORMAT_32 for MTBUF), ADD_TID_ENABLE + +var S_SAVE_SPI_INIT_ATC_MASK = 0x08000000 //bit[27]: ATC bit +var S_SAVE_SPI_INIT_ATC_SHIFT = 27 +var S_SAVE_SPI_INIT_MTYPE_MASK = 0x70000000 //bit[30:28]: Mtype +var S_SAVE_SPI_INIT_MTYPE_SHIFT = 28 +var S_SAVE_SPI_INIT_FIRST_WAVE_MASK = 0x04000000 //bit[26]: FirstWaveInTG +var S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT = 26 + +var S_SAVE_PC_HI_RCNT_SHIFT = 28 //FIXME check with Brian to ensure all fields other than PC[47:0] can be used +var S_SAVE_PC_HI_RCNT_MASK = 0xF0000000 //FIXME +var S_SAVE_PC_HI_FIRST_REPLAY_SHIFT = 27 //FIXME +var S_SAVE_PC_HI_FIRST_REPLAY_MASK = 0x08000000 //FIXME + +var s_save_spi_init_lo = exec_lo +var s_save_spi_init_hi = exec_hi + + //tba_lo and tba_hi need to be saved/restored +var s_save_pc_lo = ttmp0 //{TTMP1, TTMP0} = {3??h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]} +var s_save_pc_hi = ttmp1 +var s_save_exec_lo = ttmp2 +var s_save_exec_hi = ttmp3 +var s_save_status = ttmp4 +var s_save_trapsts = ttmp5 //not really used until the end of the SAVE routine +var s_save_xnack_mask_lo = ttmp6 +var s_save_xnack_mask_hi = ttmp7 +var s_save_buf_rsrc0 = ttmp8 +var s_save_buf_rsrc1 = ttmp9 +var s_save_buf_rsrc2 = ttmp10 +var s_save_buf_rsrc3 = ttmp11 + +var s_save_mem_offset = tma_lo +var s_save_alloc_size = s_save_trapsts //conflict +var s_save_tmp = s_save_buf_rsrc2 //shared with s_save_buf_rsrc2 (conflict: should not use mem access with s_save_tmp at the same time) +var s_save_m0 = tma_hi + +/* Restore */ +var S_RESTORE_BUF_RSRC_WORD1_STRIDE = S_SAVE_BUF_RSRC_WORD1_STRIDE +var S_RESTORE_BUF_RSRC_WORD3_MISC = S_SAVE_BUF_RSRC_WORD3_MISC + +var S_RESTORE_SPI_INIT_ATC_MASK = 0x08000000 //bit[27]: ATC bit +var S_RESTORE_SPI_INIT_ATC_SHIFT = 27 +var S_RESTORE_SPI_INIT_MTYPE_MASK = 0x70000000 //bit[30:28]: Mtype +var S_RESTORE_SPI_INIT_MTYPE_SHIFT = 28 +var S_RESTORE_SPI_INIT_FIRST_WAVE_MASK = 0x04000000 //bit[26]: FirstWaveInTG +var S_RESTORE_SPI_INIT_FIRST_WAVE_SHIFT = 26 + +var S_RESTORE_PC_HI_RCNT_SHIFT = S_SAVE_PC_HI_RCNT_SHIFT +var S_RESTORE_PC_HI_RCNT_MASK = S_SAVE_PC_HI_RCNT_MASK +var S_RESTORE_PC_HI_FIRST_REPLAY_SHIFT = S_SAVE_PC_HI_FIRST_REPLAY_SHIFT +var S_RESTORE_PC_HI_FIRST_REPLAY_MASK = S_SAVE_PC_HI_FIRST_REPLAY_MASK + +var s_restore_spi_init_lo = exec_lo +var s_restore_spi_init_hi = exec_hi + +var s_restore_mem_offset = ttmp2 +var s_restore_alloc_size = ttmp3 +var s_restore_tmp = ttmp6 //tba_lo/hi need to be restored +var s_restore_mem_offset_save = s_restore_tmp //no conflict + +var s_restore_m0 = s_restore_alloc_size //no conflict + +var s_restore_mode = ttmp7 + +var s_restore_pc_lo = ttmp0 +var s_restore_pc_hi = ttmp1 +var s_restore_exec_lo = tma_lo //no conflict +var s_restore_exec_hi = tma_hi //no conflict +var s_restore_status = ttmp4 +var s_restore_trapsts = ttmp5 +var s_restore_xnack_mask_lo = xnack_mask_lo +var s_restore_xnack_mask_hi = xnack_mask_hi +var s_restore_buf_rsrc0 = ttmp8 +var s_restore_buf_rsrc1 = ttmp9 +var s_restore_buf_rsrc2 = ttmp10 +var s_restore_buf_rsrc3 = ttmp11 + +/**************************************************************************/ +/* trap handler entry points */ +/**************************************************************************/ +/* Shader Main*/ + +shader main + asic(VI) + type(CS) + + + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) //hack to use trap_id for determining save/restore + //FIXME VCCZ un-init assertion s_getreg_b32 s_save_status, hwreg(HW_REG_STATUS) //save STATUS since we will change SCC + s_and_b32 s_save_tmp, s_save_pc_hi, 0xffff0000 //change SCC + s_cmp_eq_u32 s_save_tmp, 0x007e0000 //Save: trap_id = 0x7e. Restore: trap_id = 0x7f. + s_cbranch_scc0 L_JUMP_TO_RESTORE //do not need to recover STATUS here since we are going to RESTORE + //FIXME s_setreg_b32 hwreg(HW_REG_STATUS), s_save_status //need to recover STATUS since we are going to SAVE + s_branch L_SKIP_RESTORE //NOT restore, SAVE actually + else + s_branch L_SKIP_RESTORE //NOT restore. might be a regular trap or save + end + +L_JUMP_TO_RESTORE: + s_branch L_RESTORE //restore + +L_SKIP_RESTORE: + + s_getreg_b32 s_save_status, hwreg(HW_REG_STATUS) //save STATUS since we will change SCC + s_andn2_b32 s_save_status, s_save_status, SQ_WAVE_STATUS_SPI_PRIO_MASK //check whether this is for save + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + s_and_b32 s_save_trapsts, s_save_trapsts, SQ_WAVE_TRAPSTS_SAVECTX_MASK //check whether this is for save + s_cbranch_scc1 L_SAVE //this is the operation for save + + // ********* Handle non-CWSR traps ******************* +if (!EMU_RUN_HACK) + /* read tba and tma for next level trap handler, ttmp4 is used as s_save_status */ + s_load_dwordx4 [ttmp8,ttmp9,ttmp10, ttmp11], [tma_lo,tma_hi], 0 + s_waitcnt lgkmcnt(0) + s_or_b32 ttmp7, ttmp8, ttmp9 + s_cbranch_scc0 L_NO_NEXT_TRAP //next level trap handler not been set + s_setreg_b32 hwreg(HW_REG_STATUS), s_save_status //restore HW status(SCC) + s_setpc_b64 [ttmp8,ttmp9] //jump to next level trap handler + +L_NO_NEXT_TRAP: + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + s_and_b32 s_save_trapsts, s_save_trapsts, SQ_WAVE_TRAPSTS_EXCE_MASK // Check whether it is an exception + s_cbranch_scc1 L_EXCP_CASE // Exception, jump back to the shader program directly. + s_add_u32 ttmp0, ttmp0, 4 // S_TRAP case, add 4 to ttmp0 + s_addc_u32 ttmp1, ttmp1, 0 +L_EXCP_CASE: + s_and_b32 ttmp1, ttmp1, 0xFFFF + s_setreg_b32 hwreg(HW_REG_STATUS), s_save_status //restore HW status(SCC) + s_rfe_b64 [ttmp0, ttmp1] +end + // ********* End handling of non-CWSR traps ******************* + +/**************************************************************************/ +/* save routine */ +/**************************************************************************/ + +L_SAVE: + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_save_s + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? +end + + //check whether there is mem_viol + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + s_and_b32 s_save_trapsts, s_save_trapsts, SQ_WAVE_TRAPSTS_MEM_VIOL_MASK + s_cbranch_scc0 L_NO_PC_REWIND + + //if so, need rewind PC assuming GDS operation gets NACKed + s_mov_b32 s_save_tmp, 0 //clear mem_viol bit + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_MEM_VIOL_SHIFT, 1), s_save_tmp //clear mem_viol bit + s_and_b32 s_save_pc_hi, s_save_pc_hi, 0x0000ffff //pc[47:32] + s_sub_u32 s_save_pc_lo, s_save_pc_lo, 8 //pc[31:0]-8 + s_subb_u32 s_save_pc_hi, s_save_pc_hi, 0x0 // -scc + +L_NO_PC_REWIND: + s_mov_b32 s_save_tmp, 0 //clear saveCtx bit + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_SAVECTX_SHIFT, 1), s_save_tmp //clear saveCtx bit + + s_mov_b32 s_save_xnack_mask_lo, xnack_mask_lo //save XNACK_MASK + s_mov_b32 s_save_xnack_mask_hi, xnack_mask_hi //save XNACK must before any memory operation + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS, SQ_WAVE_IB_STS_RCNT_SHIFT, SQ_WAVE_IB_STS_RCNT_SIZE) //save RCNT + s_lshl_b32 s_save_tmp, s_save_tmp, S_SAVE_PC_HI_RCNT_SHIFT + s_or_b32 s_save_pc_hi, s_save_pc_hi, s_save_tmp + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS, SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT, SQ_WAVE_IB_STS_FIRST_REPLAY_SIZE) //save FIRST_REPLAY + s_lshl_b32 s_save_tmp, s_save_tmp, S_SAVE_PC_HI_FIRST_REPLAY_SHIFT + s_or_b32 s_save_pc_hi, s_save_pc_hi, s_save_tmp + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS) //clear RCNT and FIRST_REPLAY in IB_STS + s_and_b32 s_save_tmp, s_save_tmp, SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK_NEG + + s_setreg_b32 hwreg(HW_REG_IB_STS), s_save_tmp + + /* inform SPI the readiness and wait for SPI's go signal */ + s_mov_b32 s_save_exec_lo, exec_lo //save EXEC and use EXEC for the go signal from SPI + s_mov_b32 s_save_exec_hi, exec_hi + s_mov_b64 exec, 0x0 //clear EXEC to get ready to receive + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_sq_save_msg + s_waitcnt lgkmcnt(0) +end + + if (EMU_RUN_HACK) + + else + s_sendmsg sendmsg(MSG_SAVEWAVE) //send SPI a message and wait for SPI's write to EXEC + end + + L_SLEEP: + s_sleep 0x2 // sleep 1 (64clk) is not enough for 8 waves per SIMD, which will cause SQ hang, since the 7,8th wave could not get arbit to exec inst, while other waves are stuck into the sleep-loop and waiting for wrexec!=0 + + if (EMU_RUN_HACK) + + else + s_cbranch_execz L_SLEEP + end + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_spi_wrexec + s_waitcnt lgkmcnt(0) +end + + /* setup Resource Contants */ + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_SAVE_SINGLE_WAVE)) + //calculate wd_addr using absolute thread id + v_readlane_b32 s_save_tmp, v9, 0 + s_lshr_b32 s_save_tmp, s_save_tmp, 6 + s_mul_i32 s_save_tmp, s_save_tmp, WAVE_SPACE + s_add_i32 s_save_spi_init_lo, s_save_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_save_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_save_spi_init_hi, s_save_spi_init_hi, CTX_SAVE_CONTROL + else + end + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_SAVE_SINGLE_WAVE)) + s_add_i32 s_save_spi_init_lo, s_save_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_save_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_save_spi_init_hi, s_save_spi_init_hi, CTX_SAVE_CONTROL + else + end + + + s_mov_b32 s_save_buf_rsrc0, s_save_spi_init_lo //base_addr_lo + s_and_b32 s_save_buf_rsrc1, s_save_spi_init_hi, 0x0000FFFF //base_addr_hi + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE + s_mov_b32 s_save_buf_rsrc2, 0 //NUM_RECORDS initial value = 0 (in bytes) although not neccessarily inited + s_mov_b32 s_save_buf_rsrc3, S_SAVE_BUF_RSRC_WORD3_MISC + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_ATC_MASK + s_lshr_b32 s_save_tmp, s_save_tmp, (S_SAVE_SPI_INIT_ATC_SHIFT-SQ_BUF_RSRC_WORD1_ATC_SHIFT) //get ATC bit into position + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, s_save_tmp //or ATC + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_MTYPE_MASK + s_lshr_b32 s_save_tmp, s_save_tmp, (S_SAVE_SPI_INIT_MTYPE_SHIFT-SQ_BUF_RSRC_WORD3_MTYPE_SHIFT) //get MTYPE bits into position + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, s_save_tmp //or MTYPE + + //FIXME right now s_save_m0/s_save_mem_offset use tma_lo/tma_hi (might need to save them before using them?) + s_mov_b32 s_save_m0, m0 //save M0 + + /* global mem offset */ + s_mov_b32 s_save_mem_offset, 0x0 //mem offset initial value = 0 + + + + + /* save HW registers */ + ////////////////////////////// + + L_SAVE_HWREG: + // HWREG SR memory offset : size(VGPR)+size(SGPR) + get_vgpr_size_bytes(s_save_mem_offset) + get_sgpr_size_bytes(s_save_tmp) + s_add_u32 s_save_mem_offset, s_save_mem_offset, s_save_tmp + + + s_mov_b32 s_save_buf_rsrc2, 0x4 //NUM_RECORDS in bytes + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset) //M0 + + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_SAVE_FIRST_TIME)) + s_add_u32 s_save_pc_lo, s_save_pc_lo, 4 //pc[31:0]+4 + s_addc_u32 s_save_pc_hi, s_save_pc_hi, 0x0 //carry bit over + s_mov_b32 tba_lo, EMU_RUN_HACK_SAVE_FIRST_TIME_TBA_LO + s_mov_b32 tba_hi, EMU_RUN_HACK_SAVE_FIRST_TIME_TBA_HI + end + + write_hwreg_to_mem(s_save_pc_lo, s_save_buf_rsrc0, s_save_mem_offset) //PC + write_hwreg_to_mem(s_save_pc_hi, s_save_buf_rsrc0, s_save_mem_offset) + write_hwreg_to_mem(s_save_exec_lo, s_save_buf_rsrc0, s_save_mem_offset) //EXEC + write_hwreg_to_mem(s_save_exec_hi, s_save_buf_rsrc0, s_save_mem_offset) + write_hwreg_to_mem(s_save_status, s_save_buf_rsrc0, s_save_mem_offset) //STATUS + + //s_save_trapsts conflicts with s_save_alloc_size + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + write_hwreg_to_mem(s_save_trapsts, s_save_buf_rsrc0, s_save_mem_offset) //TRAPSTS + + write_hwreg_to_mem(s_save_xnack_mask_lo, s_save_buf_rsrc0, s_save_mem_offset) //XNACK_MASK_LO + write_hwreg_to_mem(s_save_xnack_mask_hi, s_save_buf_rsrc0, s_save_mem_offset) //XNACK_MASK_HI + + //use s_save_tmp would introduce conflict here between s_save_tmp and s_save_buf_rsrc2 + s_getreg_b32 s_save_m0, hwreg(HW_REG_MODE) //MODE + write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset) + write_hwreg_to_mem(tba_lo, s_save_buf_rsrc0, s_save_mem_offset) //TBA_LO + write_hwreg_to_mem(tba_hi, s_save_buf_rsrc0, s_save_mem_offset) //TBA_HI + + + + /* the first wave in the threadgroup */ + // save fist_wave bits in tba_hi unused bit.26 + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK // extract fisrt wave bit + //s_or_b32 tba_hi, s_save_tmp, tba_hi // save first wave bit to tba_hi.bits[26] + s_mov_b32 s_save_exec_hi, 0x0 + s_or_b32 s_save_exec_hi, s_save_tmp, s_save_exec_hi // save first wave bit to s_save_exec_hi.bits[26] + + + /* save SGPRs */ + // Save SGPR before LDS save, then the s0 to s4 can be used during LDS save... + ////////////////////////////// + + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_save_mem_offset) + // TODO, change RSRC word to rearrange memory layout for SGPRS + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_save_alloc_size, s_save_alloc_size, 1 + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 4 //Number of SGPRs = (sgpr_size + 1) * 16 (non-zero value) + + if (SGPR_SAVE_USE_SQC) + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 2 //NUM_RECORDS in bytes + else + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 8 //NUM_RECORDS in bytes (64 threads) + end + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // backup s_save_buf_rsrc0,1 to s_save_pc_lo/hi, since write_16sgpr_to_mem function will change the rsrc0 + //s_mov_b64 s_save_pc_lo, s_save_buf_rsrc0 + s_mov_b64 s_save_xnack_mask_lo, s_save_buf_rsrc0 + s_add_u32 s_save_buf_rsrc0, s_save_buf_rsrc0, s_save_mem_offset + s_addc_u32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0 + + s_mov_b32 m0, 0x0 //SGPR initial index value =0 + L_SAVE_SGPR_LOOP: + // SGPR is allocated in 16 SGPR granularity + s_movrels_b64 s0, s0 //s0 = s[0+m0], s1 = s[1+m0] + s_movrels_b64 s2, s2 //s2 = s[2+m0], s3 = s[3+m0] + s_movrels_b64 s4, s4 //s4 = s[4+m0], s5 = s[5+m0] + s_movrels_b64 s6, s6 //s6 = s[6+m0], s7 = s[7+m0] + s_movrels_b64 s8, s8 //s8 = s[8+m0], s9 = s[9+m0] + s_movrels_b64 s10, s10 //s10 = s[10+m0], s11 = s[11+m0] + s_movrels_b64 s12, s12 //s12 = s[12+m0], s13 = s[13+m0] + s_movrels_b64 s14, s14 //s14 = s[14+m0], s15 = s[15+m0] + + write_16sgpr_to_mem(s0, s_save_buf_rsrc0, s_save_mem_offset) //PV: the best performance should be using s_buffer_store_dwordx4 + s_add_u32 m0, m0, 16 //next sgpr index + s_cmp_lt_u32 m0, s_save_alloc_size //scc = (m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_SGPR_LOOP //SGPR save is complete? + // restore s_save_buf_rsrc0,1 + //s_mov_b64 s_save_buf_rsrc0, s_save_pc_lo + s_mov_b64 s_save_buf_rsrc0, s_save_xnack_mask_lo + + + + + /* save first 4 VGPR, then LDS save could use */ + // each wave will alloc 4 vgprs at least... + ///////////////////////////////////////////////////////////////////////////////////// + + s_mov_b32 s_save_mem_offset, 0 + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // VGPR Allocated in 4-GPR granularity + +if G8SR_VGPR_SR_IN_DWX4 + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + buffer_store_dwordx4 v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE // reset const stride to 4 bytes +else + buffer_store_dword v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + buffer_store_dword v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256 + buffer_store_dword v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2 + buffer_store_dword v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3 +end + + + + /* save LDS */ + ////////////////////////////// + + L_SAVE_LDS: + + // Change EXEC to all threads... + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) //lds_size + s_and_b32 s_save_alloc_size, s_save_alloc_size, 0xFFFFFFFF //lds_size is zero? + s_cbranch_scc0 L_SAVE_LDS_DONE //no lds used? jump to L_SAVE_DONE + + s_barrier //LDS is used? wait for other waves in the same TG + //s_and_b32 s_save_tmp, tba_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK //exec is still used here + s_and_b32 s_save_tmp, s_save_exec_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK //exec is still used here + s_cbranch_scc0 L_SAVE_LDS_DONE + + // first wave do LDS save; + + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 6 //LDS size in dwords = lds_size * 64dw + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 2 //LDS size in bytes + s_mov_b32 s_save_buf_rsrc2, s_save_alloc_size //NUM_RECORDS in bytes + + // LDS at offset: size(VGPR)+SIZE(SGPR)+SIZE(HWREG) + // + get_vgpr_size_bytes(s_save_mem_offset) + get_sgpr_size_bytes(s_save_tmp) + s_add_u32 s_save_mem_offset, s_save_mem_offset, s_save_tmp + s_add_u32 s_save_mem_offset, s_save_mem_offset, get_hwreg_size_bytes() + + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + s_mov_b32 m0, 0x0 //lds_offset initial value = 0 + + +var LDS_DMA_ENABLE = 0 +var UNROLL = 0 +if UNROLL==0 && LDS_DMA_ENABLE==1 + s_mov_b32 s3, 256*2 + s_nop 0 + s_nop 0 + s_nop 0 + L_SAVE_LDS_LOOP: + //TODO: looks the 2 buffer_store/load clause for s/r will hurt performance.??? + if (SAVE_LDS) //SPI always alloc LDS space in 128DW granularity + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 // first 64DW + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:256 // second 64DW + end + + s_add_u32 m0, m0, s3 //every buffer_store_lds does 256 bytes + s_add_u32 s_save_mem_offset, s_save_mem_offset, s3 //mem offset increased by 256 bytes + s_cmp_lt_u32 m0, s_save_alloc_size //scc=(m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_LDS_LOOP //LDS save is complete? + +elsif LDS_DMA_ENABLE==1 && UNROLL==1 // UNROOL , has ichace miss + // store from higest LDS address to lowest + s_mov_b32 s3, 256*2 + s_sub_u32 m0, s_save_alloc_size, s3 + s_add_u32 s_save_mem_offset, s_save_mem_offset, m0 + s_lshr_b32 s_save_alloc_size, s_save_alloc_size, 9 // how many 128 trunks... + s_sub_u32 s_save_alloc_size, 128, s_save_alloc_size // store from higheset addr to lowest + s_mul_i32 s_save_alloc_size, s_save_alloc_size, 6*4 // PC offset increment, each LDS save block cost 6*4 Bytes instruction + s_add_u32 s_save_alloc_size, s_save_alloc_size, 3*4 //2is the below 2 inst...//s_addc and s_setpc + s_nop 0 + s_nop 0 + s_nop 0 //pad 3 dw to let LDS_DMA align with 64Bytes + s_getpc_b64 s[0:1] // reuse s[0:1], since s[0:1] already saved + s_add_u32 s0, s0,s_save_alloc_size + s_addc_u32 s1, s1, 0 + s_setpc_b64 s[0:1] + + + for var i =0; i< 128; i++ + // be careful to make here a 64Byte aligned address, which could improve performance... + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:0 // first 64DW + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:256 // second 64DW + + if i!=127 + s_sub_u32 m0, m0, s3 // use a sgpr to shrink 2DW-inst to 1DW inst to improve performance , i.e. pack more LDS_DMA inst to one Cacheline + s_sub_u32 s_save_mem_offset, s_save_mem_offset, s3 + end + end + +else // BUFFER_STORE + v_mbcnt_lo_u32_b32 v2, 0xffffffff, 0x0 + v_mbcnt_hi_u32_b32 v3, 0xffffffff, v2 // tid + v_mul_i32_i24 v2, v3, 8 // tid*8 + v_mov_b32 v3, 256*2 + s_mov_b32 m0, 0x10000 + s_mov_b32 s0, s_save_buf_rsrc3 + s_and_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, 0xFF7FFFFF // disable add_tid + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, 0x58000 //DFMT + +L_SAVE_LDS_LOOP_VECTOR: + ds_read_b64 v[0:1], v2 //x =LDS[a], byte address + s_waitcnt lgkmcnt(0) + buffer_store_dwordx2 v[0:1], v2, s_save_buf_rsrc0, s_save_mem_offset offen:1 glc:1 slc:1 +// s_waitcnt vmcnt(0) + v_add_u32 v2, vcc[0:1], v2, v3 + v_cmp_lt_u32 vcc[0:1], v2, s_save_alloc_size + s_cbranch_vccnz L_SAVE_LDS_LOOP_VECTOR + + // restore rsrc3 + s_mov_b32 s_save_buf_rsrc3, s0 + +end + +L_SAVE_LDS_DONE: + + + /* save VGPRs - set the Rest VGPRs */ + ////////////////////////////////////////////////////////////////////////////////////// + L_SAVE_VGPR: + // VGPR SR memory offset: 0 + // TODO rearrange the RSRC words to use swizzle for VGPR save... + + s_mov_b32 s_save_mem_offset, (0+256*4) // for the rest VGPRs + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_save_alloc_size, s_save_alloc_size, 1 + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 2 //Number of VGPRs = (vgpr_size + 1) * 4 (non-zero value) //FIXME for GFX, zero is possible + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 8 //NUM_RECORDS in bytes (64 threads*4) + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // VGPR Allocated in 4-GPR granularity + +if G8SR_VGPR_SR_IN_DWX4 + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + s_mov_b32 m0, 4 // skip first 4 VGPRs + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc0 L_SAVE_VGPR_LOOP_END // no more vgprs + + s_set_gpr_idx_on m0, 0x1 // This will change M0 + s_add_u32 s_save_alloc_size, s_save_alloc_size, 0x1000 // because above inst change m0 +L_SAVE_VGPR_LOOP: + v_mov_b32 v0, v0 // v0 = v[0+m0] + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + + + buffer_store_dwordx4 v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + s_add_u32 m0, m0, 4 + s_add_u32 s_save_mem_offset, s_save_mem_offset, 256*4 + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc1 L_SAVE_VGPR_LOOP //VGPR save is complete? + s_set_gpr_idx_off +L_SAVE_VGPR_LOOP_END: + + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE // reset const stride to 4 bytes +else + // VGPR store using dw burst + s_mov_b32 m0, 0x4 //VGPR initial index value =0 + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc0 L_SAVE_VGPR_END + + + s_set_gpr_idx_on m0, 0x1 //M0[7:0] = M0[7:0] and M0[15:12] = 0x1 + s_add_u32 s_save_alloc_size, s_save_alloc_size, 0x1000 //add 0x1000 since we compare m0 against it later + + L_SAVE_VGPR_LOOP: + v_mov_b32 v0, v0 //v0 = v[0+m0] + v_mov_b32 v1, v1 //v0 = v[0+m0] + v_mov_b32 v2, v2 //v0 = v[0+m0] + v_mov_b32 v3, v3 //v0 = v[0+m0] + + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_store_format_x v0, v0, s_save_buf_rsrc0, s_save_mem_offset format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_store_dword v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + buffer_store_dword v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256 + buffer_store_dword v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2 + buffer_store_dword v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3 + end + + s_add_u32 m0, m0, 4 //next vgpr index + s_add_u32 s_save_mem_offset, s_save_mem_offset, 256*4 //every buffer_store_dword does 256 bytes + s_cmp_lt_u32 m0, s_save_alloc_size //scc = (m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_VGPR_LOOP //VGPR save is complete? + s_set_gpr_idx_off +end + +L_SAVE_VGPR_END: + + + + + + + /* S_PGM_END_SAVED */ //FIXME graphics ONLY + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_SAVE_NORMAL_EXIT)) + s_and_b32 s_save_pc_hi, s_save_pc_hi, 0x0000ffff //pc[47:32] + s_add_u32 s_save_pc_lo, s_save_pc_lo, 4 //pc[31:0]+4 + s_addc_u32 s_save_pc_hi, s_save_pc_hi, 0x0 //carry bit over + s_rfe_b64 s_save_pc_lo //Return to the main shader program + else + end + +// Save Done timestamp +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_save_d + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_save_mem_offset) + s_add_u32 s_save_mem_offset, s_save_mem_offset, G8SR_DEBUG_TS_SAVE_D_OFFSET + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? + // Need reset rsrc2?? + s_mov_b32 m0, s_save_mem_offset + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + s_buffer_store_dwordx2 s_g8sr_ts_save_d, s_save_buf_rsrc0, m0 glc:1 +end + + + s_branch L_END_PGM + + + +/**************************************************************************/ +/* restore routine */ +/**************************************************************************/ + +L_RESTORE: + /* Setup Resource Contants */ + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) + //calculate wd_addr using absolute thread id + v_readlane_b32 s_restore_tmp, v9, 0 + s_lshr_b32 s_restore_tmp, s_restore_tmp, 6 + s_mul_i32 s_restore_tmp, s_restore_tmp, WAVE_SPACE + s_add_i32 s_restore_spi_init_lo, s_restore_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_restore_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_restore_spi_init_hi, s_restore_spi_init_hi, CTX_RESTORE_CONTROL + else + end + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_restore_s + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? + // tma_lo/hi are sgpr 110, 111, which will not used for 112 SGPR allocated case... + s_mov_b32 s_restore_pc_lo, s_g8sr_ts_restore_s[0] + s_mov_b32 s_restore_pc_hi, s_g8sr_ts_restore_s[1] //backup ts to ttmp0/1, sicne exec will be finally restored.. +end + + + + s_mov_b32 s_restore_buf_rsrc0, s_restore_spi_init_lo //base_addr_lo + s_and_b32 s_restore_buf_rsrc1, s_restore_spi_init_hi, 0x0000FFFF //base_addr_hi + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, S_RESTORE_BUF_RSRC_WORD1_STRIDE + s_mov_b32 s_restore_buf_rsrc2, 0 //NUM_RECORDS initial value = 0 (in bytes) + s_mov_b32 s_restore_buf_rsrc3, S_RESTORE_BUF_RSRC_WORD3_MISC + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_ATC_MASK + s_lshr_b32 s_restore_tmp, s_restore_tmp, (S_RESTORE_SPI_INIT_ATC_SHIFT-SQ_BUF_RSRC_WORD1_ATC_SHIFT) //get ATC bit into position + s_or_b32 s_restore_buf_rsrc3, s_restore_buf_rsrc3, s_restore_tmp //or ATC + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_MTYPE_MASK + s_lshr_b32 s_restore_tmp, s_restore_tmp, (S_RESTORE_SPI_INIT_MTYPE_SHIFT-SQ_BUF_RSRC_WORD3_MTYPE_SHIFT) //get MTYPE bits into position + s_or_b32 s_restore_buf_rsrc3, s_restore_buf_rsrc3, s_restore_tmp //or MTYPE + + /* global mem offset */ +// s_mov_b32 s_restore_mem_offset, 0x0 //mem offset initial value = 0 + + /* the first wave in the threadgroup */ + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_FIRST_WAVE_MASK + s_cbranch_scc0 L_RESTORE_VGPR + + /* restore LDS */ + ////////////////////////////// + L_RESTORE_LDS: + + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on //be consistent with SAVE although can be moved ahead + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) //lds_size + s_and_b32 s_restore_alloc_size, s_restore_alloc_size, 0xFFFFFFFF //lds_size is zero? + s_cbranch_scc0 L_RESTORE_VGPR //no lds used? jump to L_RESTORE_VGPR + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 6 //LDS size in dwords = lds_size * 64dw + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 2 //LDS size in bytes + s_mov_b32 s_restore_buf_rsrc2, s_restore_alloc_size //NUM_RECORDS in bytes + + // LDS at offset: size(VGPR)+SIZE(SGPR)+SIZE(HWREG) + // + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, get_hwreg_size_bytes() //FIXME, Check if offset overflow??? + + + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + s_mov_b32 m0, 0x0 //lds_offset initial value = 0 + + L_RESTORE_LDS_LOOP: + if (SAVE_LDS) + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 // first 64DW + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:256 // second 64DW + end + s_add_u32 m0, m0, 256*2 // 128 DW + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*2 //mem offset increased by 128DW + s_cmp_lt_u32 m0, s_restore_alloc_size //scc=(m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_RESTORE_LDS_LOOP //LDS restore is complete? + + + /* restore VGPRs */ + ////////////////////////////// + L_RESTORE_VGPR: + // VGPR SR memory offset : 0 + s_mov_b32 s_restore_mem_offset, 0x0 + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on //be consistent with SAVE although can be moved ahead + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 1 + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 2 //Number of VGPRs = (vgpr_size + 1) * 4 (non-zero value) + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 8 //NUM_RECORDS in bytes (64 threads*4) + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + +if G8SR_VGPR_SR_IN_DWX4 + get_vgpr_size_bytes(s_restore_mem_offset) + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, G8SR_RESTORE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + s_mov_b32 m0, s_restore_alloc_size + s_set_gpr_idx_on m0, 0x8 // Note.. This will change m0 + +L_RESTORE_VGPR_LOOP: + buffer_load_dwordx4 v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 + s_waitcnt vmcnt(0) + s_sub_u32 m0, m0, 4 + v_mov_b32 v0, v0 // v[0+m0] = v0 + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + s_cmp_eq_u32 m0, 0x8000 + s_cbranch_scc0 L_RESTORE_VGPR_LOOP + s_set_gpr_idx_off + + s_and_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, S_RESTORE_BUF_RSRC_WORD1_STRIDE // const stride to 4*4 bytes + +else + // VGPR load using dw burst + s_mov_b32 s_restore_mem_offset_save, s_restore_mem_offset // restore start with v1, v0 will be the last + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + s_mov_b32 m0, 4 //VGPR initial index value = 1 + s_set_gpr_idx_on m0, 0x8 //M0[7:0] = M0[7:0] and M0[15:12] = 0x8 + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 0x8000 //add 0x8000 since we compare m0 against it later + + L_RESTORE_VGPR_LOOP: + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_load_format_x v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 + buffer_load_dword v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256 + buffer_load_dword v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*2 + buffer_load_dword v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*3 + end + s_waitcnt vmcnt(0) //ensure data ready + v_mov_b32 v0, v0 //v[0+m0] = v0 + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + s_add_u32 m0, m0, 4 //next vgpr index + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 //every buffer_load_dword does 256 bytes + s_cmp_lt_u32 m0, s_restore_alloc_size //scc = (m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_RESTORE_VGPR_LOOP //VGPR restore (except v0) is complete? + s_set_gpr_idx_off + /* VGPR restore on v0 */ + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_load_format_x v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 + buffer_load_dword v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256 + buffer_load_dword v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*2 + buffer_load_dword v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*3 + end + +end + + /* restore SGPRs */ + ////////////////////////////// + + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 16*4 // restore SGPR from S[n] to S[0], by 16 sgprs group + // TODO, change RSRC word to rearrange memory layout for SGPRS + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 1 + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 4 //Number of SGPRs = (sgpr_size + 1) * 16 (non-zero value) + + if (SGPR_SAVE_USE_SQC) + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 2 //NUM_RECORDS in bytes + else + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 8 //NUM_RECORDS in bytes (64 threads) + end + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + /* If 112 SGPRs ar allocated, 4 sgprs are not used TBA(108,109),TMA(110,111), + However, we are safe to restore these 4 SGPRs anyway, since TBA,TMA will later be restored by HWREG + */ + s_mov_b32 m0, s_restore_alloc_size + + L_RESTORE_SGPR_LOOP: + read_16sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset) //PV: further performance improvement can be made + s_waitcnt lgkmcnt(0) //ensure data ready + + s_sub_u32 m0, m0, 16 // Restore from S[n] to S[0] + + s_movreld_b64 s0, s0 //s[0+m0] = s0 + s_movreld_b64 s2, s2 + s_movreld_b64 s4, s4 + s_movreld_b64 s6, s6 + s_movreld_b64 s8, s8 + s_movreld_b64 s10, s10 + s_movreld_b64 s12, s12 + s_movreld_b64 s14, s14 + + s_cmp_eq_u32 m0, 0 //scc = (m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc0 L_RESTORE_SGPR_LOOP //SGPR restore (except s0) is complete? + + /* restore HW registers */ + ////////////////////////////// + L_RESTORE_HWREG: + + +if G8SR_DEBUG_TIMESTAMP + s_mov_b32 s_g8sr_ts_restore_s[0], s_restore_pc_lo + s_mov_b32 s_g8sr_ts_restore_s[1], s_restore_pc_hi +end + + // HWREG SR memory offset : size(VGPR)+size(SGPR) + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + + + s_mov_b32 s_restore_buf_rsrc2, 0x4 //NUM_RECORDS in bytes + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + read_hwreg_from_mem(s_restore_m0, s_restore_buf_rsrc0, s_restore_mem_offset) //M0 + read_hwreg_from_mem(s_restore_pc_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //PC + read_hwreg_from_mem(s_restore_pc_hi, s_restore_buf_rsrc0, s_restore_mem_offset) + read_hwreg_from_mem(s_restore_exec_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //EXEC + read_hwreg_from_mem(s_restore_exec_hi, s_restore_buf_rsrc0, s_restore_mem_offset) + read_hwreg_from_mem(s_restore_status, s_restore_buf_rsrc0, s_restore_mem_offset) //STATUS + read_hwreg_from_mem(s_restore_trapsts, s_restore_buf_rsrc0, s_restore_mem_offset) //TRAPSTS + read_hwreg_from_mem(xnack_mask_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //XNACK_MASK_LO + read_hwreg_from_mem(xnack_mask_hi, s_restore_buf_rsrc0, s_restore_mem_offset) //XNACK_MASK_HI + read_hwreg_from_mem(s_restore_mode, s_restore_buf_rsrc0, s_restore_mem_offset) //MODE + read_hwreg_from_mem(tba_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //TBA_LO + read_hwreg_from_mem(tba_hi, s_restore_buf_rsrc0, s_restore_mem_offset) //TBA_HI + + s_waitcnt lgkmcnt(0) //from now on, it is safe to restore STATUS and IB_STS + + s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff //pc[47:32] //Do it here in order not to affect STATUS + + //for normal save & restore, the saved PC points to the next inst to execute, no adjustment needs to be made, otherwise: + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) + s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 8 //pc[31:0]+8 //two back-to-back s_trap are used (first for save and second for restore) + s_addc_u32 s_restore_pc_hi, s_restore_pc_hi, 0x0 //carry bit over + end + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_RESTORE_NORMAL)) + s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 4 //pc[31:0]+4 // save is hack through s_trap but restore is normal + s_addc_u32 s_restore_pc_hi, s_restore_pc_hi, 0x0 //carry bit over + end + + s_mov_b32 m0, s_restore_m0 + s_mov_b32 exec_lo, s_restore_exec_lo + s_mov_b32 exec_hi, s_restore_exec_hi + + s_and_b32 s_restore_m0, SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK, s_restore_trapsts + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE), s_restore_m0 + s_and_b32 s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK, s_restore_trapsts + s_lshr_b32 s_restore_m0, s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE), s_restore_m0 + //s_setreg_b32 hwreg(HW_REG_TRAPSTS), s_restore_trapsts //don't overwrite SAVECTX bit as it may be set through external SAVECTX during restore + s_setreg_b32 hwreg(HW_REG_MODE), s_restore_mode + //reuse s_restore_m0 as a temp register + s_and_b32 s_restore_m0, s_restore_pc_hi, S_SAVE_PC_HI_RCNT_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, S_SAVE_PC_HI_RCNT_SHIFT + s_lshl_b32 s_restore_m0, s_restore_m0, SQ_WAVE_IB_STS_RCNT_SHIFT + s_mov_b32 s_restore_tmp, 0x0 //IB_STS is zero + s_or_b32 s_restore_tmp, s_restore_tmp, s_restore_m0 + s_and_b32 s_restore_m0, s_restore_pc_hi, S_SAVE_PC_HI_FIRST_REPLAY_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, S_SAVE_PC_HI_FIRST_REPLAY_SHIFT + s_lshl_b32 s_restore_m0, s_restore_m0, SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT + s_or_b32 s_restore_tmp, s_restore_tmp, s_restore_m0 + s_and_b32 s_restore_m0, s_restore_status, SQ_WAVE_STATUS_INST_ATC_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, SQ_WAVE_STATUS_INST_ATC_SHIFT + s_setreg_b32 hwreg(HW_REG_IB_STS), s_restore_tmp + + s_and_b64 exec, exec, exec // Restore STATUS.EXECZ, not writable by s_setreg_b32 + s_and_b64 vcc, vcc, vcc // Restore STATUS.VCCZ, not writable by s_setreg_b32 + s_setreg_b32 hwreg(HW_REG_STATUS), s_restore_status // SCC is included, which is changed by previous salu + + s_barrier //barrier to ensure the readiness of LDS before access attempts from any other wave in the same TG //FIXME not performance-optimal at this time + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_restore_d + s_waitcnt lgkmcnt(0) +end + +// s_rfe_b64 s_restore_pc_lo //Return to the main shader program and resume execution + s_rfe_restore_b64 s_restore_pc_lo, s_restore_m0 // s_restore_m0[0] is used to set STATUS.inst_atc + + +/**************************************************************************/ +/* the END */ +/**************************************************************************/ +L_END_PGM: + s_endpgm + +end + + +/**************************************************************************/ +/* the helper functions */ +/**************************************************************************/ + +//Only for save hwreg to mem +function write_hwreg_to_mem(s, s_rsrc, s_mem_offset) + s_mov_b32 exec_lo, m0 //assuming exec_lo is not needed anymore from this point on + s_mov_b32 m0, s_mem_offset + s_buffer_store_dword s, s_rsrc, m0 glc:1 + s_add_u32 s_mem_offset, s_mem_offset, 4 + s_mov_b32 m0, exec_lo +end + + +// HWREG are saved before SGPRs, so all HWREG could be use. +function write_16sgpr_to_mem(s, s_rsrc, s_mem_offset) + + s_buffer_store_dwordx4 s[0], s_rsrc, 0 glc:1 + s_buffer_store_dwordx4 s[4], s_rsrc, 16 glc:1 + s_buffer_store_dwordx4 s[8], s_rsrc, 32 glc:1 + s_buffer_store_dwordx4 s[12], s_rsrc, 48 glc:1 + s_add_u32 s_rsrc[0], s_rsrc[0], 4*16 + s_addc_u32 s_rsrc[1], s_rsrc[1], 0x0 // +scc +end + + +function read_hwreg_from_mem(s, s_rsrc, s_mem_offset) + s_buffer_load_dword s, s_rsrc, s_mem_offset glc:1 + s_add_u32 s_mem_offset, s_mem_offset, 4 +end + +function read_16sgpr_from_mem(s, s_rsrc, s_mem_offset) + s_buffer_load_dwordx16 s, s_rsrc, s_mem_offset glc:1 + s_sub_u32 s_mem_offset, s_mem_offset, 4*16 +end + + + +function get_lds_size_bytes(s_lds_size_byte) + // SQ LDS granularity is 64DW, while PGM_RSRC2.lds_size is in granularity 128DW + s_getreg_b32 s_lds_size_byte, hwreg(HW_REG_LDS_ALLOC, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) // lds_size + s_lshl_b32 s_lds_size_byte, s_lds_size_byte, 8 //LDS size in dwords = lds_size * 64 *4Bytes // granularity 64DW +end + +function get_vgpr_size_bytes(s_vgpr_size_byte) + s_getreg_b32 s_vgpr_size_byte, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_vgpr_size_byte, s_vgpr_size_byte, 1 + s_lshl_b32 s_vgpr_size_byte, s_vgpr_size_byte, (2+8) //Number of VGPRs = (vgpr_size + 1) * 4 * 64 * 4 (non-zero value) //FIXME for GFX, zero is possible +end + +function get_sgpr_size_bytes(s_sgpr_size_byte) + s_getreg_b32 s_sgpr_size_byte, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_sgpr_size_byte, s_sgpr_size_byte, 1 + s_lshl_b32 s_sgpr_size_byte, s_sgpr_size_byte, 6 //Number of SGPRs = (sgpr_size + 1) * 16 *4 (non-zero value) +end + +function get_hwreg_size_bytes + return 128 //HWREG size 128 bytes +end + + +#endif + +static const uint32_t cwsr_trap_gfx8_hex[] = { + 0xbf820001, 0xbf820123, + 0xb8f4f802, 0x89748674, + 0xb8f5f803, 0x8675ff75, + 0x00000400, 0xbf850011, + 0xc00a1e37, 0x00000000, + 0xbf8c007f, 0x87777978, + 0xbf840002, 0xb974f802, + 0xbe801d78, 0xb8f5f803, + 0x8675ff75, 0x000001ff, + 0xbf850002, 0x80708470, + 0x82718071, 0x8671ff71, + 0x0000ffff, 0xb974f802, + 0xbe801f70, 0xb8f5f803, + 0x8675ff75, 0x00000100, + 0xbf840006, 0xbefa0080, + 0xb97a0203, 0x8671ff71, + 0x0000ffff, 0x80f08870, + 0x82f18071, 0xbefa0080, + 0xb97a0283, 0xbef60068, + 0xbef70069, 0xb8fa1c07, + 0x8e7a9c7a, 0x87717a71, + 0xb8fa03c7, 0x8e7a9b7a, + 0x87717a71, 0xb8faf807, + 0x867aff7a, 0x00007fff, + 0xb97af807, 0xbef2007e, + 0xbef3007f, 0xbefe0180, + 0xbf900004, 0xbf8e0002, + 0xbf88fffe, 0xbef8007e, + 0x8679ff7f, 0x0000ffff, + 0x8779ff79, 0x00040000, + 0xbefa0080, 0xbefb00ff, + 0x00807fac, 0x867aff7f, + 0x08000000, 0x8f7a837a, + 0x877b7a7b, 0x867aff7f, + 0x70000000, 0x8f7a817a, + 0x877b7a7b, 0xbeef007c, + 0xbeee0080, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8fa1605, 0x807a817a, + 0x8e7a867a, 0x806e7a6e, + 0xbefa0084, 0xbefa00ff, + 0x01000000, 0xbefe007c, + 0xbefc006e, 0xc0611bfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611c3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611c7c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611cbc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611cfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611d3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xb8f5f803, + 0xbefe007c, 0xbefc006e, + 0xc0611d7c, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xbefe007c, 0xbefc006e, + 0xc0611dbc, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xbefe007c, 0xbefc006e, + 0xc0611dfc, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xb8eff801, 0xbefe007c, + 0xbefc006e, 0xc0611bfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611b3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611b7c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0x867aff7f, + 0x04000000, 0xbef30080, + 0x8773737a, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8f51605, 0x80758175, + 0x8e758475, 0x8e7a8275, + 0xbefa00ff, 0x01000000, + 0xbef60178, 0x80786e78, + 0x82798079, 0xbefc0080, + 0xbe802b00, 0xbe822b02, + 0xbe842b04, 0xbe862b06, + 0xbe882b08, 0xbe8a2b0a, + 0xbe8c2b0c, 0xbe8e2b0e, + 0xc06b003c, 0x00000000, + 0xc06b013c, 0x00000010, + 0xc06b023c, 0x00000020, + 0xc06b033c, 0x00000030, + 0x8078c078, 0x82798079, + 0x807c907c, 0xbf0a757c, + 0xbf85ffeb, 0xbef80176, + 0xbeee0080, 0xbefe00c1, + 0xbeff00c1, 0xbefa00ff, + 0x01000000, 0xe0724000, + 0x6e1e0000, 0xe0724100, + 0x6e1e0100, 0xe0724200, + 0x6e1e0200, 0xe0724300, + 0x6e1e0300, 0xbefe00c1, + 0xbeff00c1, 0xb8f54306, + 0x8675c175, 0xbf84002c, + 0xbf8a0000, 0x867aff73, + 0x04000000, 0xbf840028, + 0x8e758675, 0x8e758275, + 0xbefa0075, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8fa1605, 0x807a817a, + 0x8e7a867a, 0x806e7a6e, + 0x806eff6e, 0x00000080, + 0xbefa00ff, 0x01000000, + 0xbefc0080, 0xd28c0002, + 0x000100c1, 0xd28d0003, + 0x000204c1, 0xd1060002, + 0x00011103, 0x7e0602ff, + 0x00000200, 0xbefc00ff, + 0x00010000, 0xbe80007b, + 0x867bff7b, 0xff7fffff, + 0x877bff7b, 0x00058000, + 0xd8ec0000, 0x00000002, + 0xbf8c007f, 0xe0765000, + 0x6e1e0002, 0x32040702, + 0xd0c9006a, 0x0000eb02, + 0xbf87fff7, 0xbefb0000, + 0xbeee00ff, 0x00000400, + 0xbefe00c1, 0xbeff00c1, + 0xb8f52a05, 0x80758175, + 0x8e758275, 0x8e7a8875, + 0xbefa00ff, 0x01000000, + 0xbefc0084, 0xbf0a757c, + 0xbf840015, 0xbf11017c, + 0x8075ff75, 0x00001000, + 0x7e000300, 0x7e020301, + 0x7e040302, 0x7e060303, + 0xe0724000, 0x6e1e0000, + 0xe0724100, 0x6e1e0100, + 0xe0724200, 0x6e1e0200, + 0xe0724300, 0x6e1e0300, + 0x807c847c, 0x806eff6e, + 0x00000400, 0xbf0a757c, + 0xbf85ffef, 0xbf9c0000, + 0xbf8200ca, 0xbef8007e, + 0x8679ff7f, 0x0000ffff, + 0x8779ff79, 0x00040000, + 0xbefa0080, 0xbefb00ff, + 0x00807fac, 0x8676ff7f, + 0x08000000, 0x8f768376, + 0x877b767b, 0x8676ff7f, + 0x70000000, 0x8f768176, + 0x877b767b, 0x8676ff7f, + 0x04000000, 0xbf84001e, + 0xbefe00c1, 0xbeff00c1, + 0xb8f34306, 0x8673c173, + 0xbf840019, 0x8e738673, + 0x8e738273, 0xbefa0073, + 0xb8f22a05, 0x80728172, + 0x8e728a72, 0xb8f61605, + 0x80768176, 0x8e768676, + 0x80727672, 0x8072ff72, + 0x00000080, 0xbefa00ff, + 0x01000000, 0xbefc0080, + 0xe0510000, 0x721e0000, + 0xe0510100, 0x721e0000, + 0x807cff7c, 0x00000200, + 0x8072ff72, 0x00000200, + 0xbf0a737c, 0xbf85fff6, + 0xbef20080, 0xbefe00c1, + 0xbeff00c1, 0xb8f32a05, + 0x80738173, 0x8e738273, + 0x8e7a8873, 0xbefa00ff, + 0x01000000, 0xbef60072, + 0x8072ff72, 0x00000400, + 0xbefc0084, 0xbf11087c, + 0x8073ff73, 0x00008000, + 0xe0524000, 0x721e0000, + 0xe0524100, 0x721e0100, + 0xe0524200, 0x721e0200, + 0xe0524300, 0x721e0300, + 0xbf8c0f70, 0x7e000300, + 0x7e020301, 0x7e040302, + 0x7e060303, 0x807c847c, + 0x8072ff72, 0x00000400, + 0xbf0a737c, 0xbf85ffee, + 0xbf9c0000, 0xe0524000, + 0x761e0000, 0xe0524100, + 0x761e0100, 0xe0524200, + 0x761e0200, 0xe0524300, + 0x761e0300, 0xb8f22a05, + 0x80728172, 0x8e728a72, + 0xb8f61605, 0x80768176, + 0x8e768676, 0x80727672, + 0x80f2c072, 0xb8f31605, + 0x80738173, 0x8e738473, + 0x8e7a8273, 0xbefa00ff, + 0x01000000, 0xbefc0073, + 0xc031003c, 0x00000072, + 0x80f2c072, 0xbf8c007f, + 0x80fc907c, 0xbe802d00, + 0xbe822d02, 0xbe842d04, + 0xbe862d06, 0xbe882d08, + 0xbe8a2d0a, 0xbe8c2d0c, + 0xbe8e2d0e, 0xbf06807c, + 0xbf84fff1, 0xb8f22a05, + 0x80728172, 0x8e728a72, + 0xb8f61605, 0x80768176, + 0x8e768676, 0x80727672, + 0xbefa0084, 0xbefa00ff, + 0x01000000, 0xc0211cfc, + 0x00000072, 0x80728472, + 0xc0211c3c, 0x00000072, + 0x80728472, 0xc0211c7c, + 0x00000072, 0x80728472, + 0xc0211bbc, 0x00000072, + 0x80728472, 0xc0211bfc, + 0x00000072, 0x80728472, + 0xc0211d3c, 0x00000072, + 0x80728472, 0xc0211d7c, + 0x00000072, 0x80728472, + 0xc0211a3c, 0x00000072, + 0x80728472, 0xc0211a7c, + 0x00000072, 0x80728472, + 0xc0211dfc, 0x00000072, + 0x80728472, 0xc0211b3c, + 0x00000072, 0x80728472, + 0xc0211b7c, 0x00000072, + 0x80728472, 0xbf8c007f, + 0x8671ff71, 0x0000ffff, + 0xbefc0073, 0xbefe006e, + 0xbeff006f, 0x867375ff, + 0x000003ff, 0xb9734803, + 0x867375ff, 0xfffff800, + 0x8f738b73, 0xb973a2c3, + 0xb977f801, 0x8673ff71, + 0xf0000000, 0x8f739c73, + 0x8e739073, 0xbef60080, + 0x87767376, 0x8673ff71, + 0x08000000, 0x8f739b73, + 0x8e738f73, 0x87767376, + 0x8673ff74, 0x00800000, + 0x8f739773, 0xb976f807, + 0x86fe7e7e, 0x86ea6a6a, + 0xb974f802, 0xbf8a0000, + 0x95807370, 0xbf810000, +}; + -- GitLab From 373d7080896a3cb3b28ae3a2abdafb7bb87552b1 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 14 Nov 2017 16:41:19 -0500 Subject: [PATCH 0245/1471] drm/amdkfd: Add CWSR support This hardware feature allows the GPU to preempt shader execution in the middle of a compute wave, save the state and restore it later to resume execution. Memory for saving the state is allocated per queue in user mode and the address and size passed to the create_queue ioctl. The size depends on the number of waves that can be in flight simultaneously on a given ASIC. Signed-off-by: Shaoyun.liu Signed-off-by: Yong Zhao Signed-off-by: Felix Kuehling Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 7 +- drivers/gpu/drm/amd/amdkfd/kfd_device.c | 20 ++++- .../drm/amd/amdkfd/kfd_device_queue_manager.c | 6 ++ drivers/gpu/drm/amd/amdkfd/kfd_module.c | 4 + .../gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 27 ++++++ drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 31 ++++++- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 87 ++++++++++++++++++- include/uapi/linux/kfd_ioctl.h | 3 +- 8 files changed, 179 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 505d39156acdb..2a4612d8437ae 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -117,7 +117,7 @@ static int kfd_open(struct inode *inode, struct file *filep) return -EPERM; } - process = kfd_create_process(current); + process = kfd_create_process(filep); if (IS_ERR(process)) return PTR_ERR(process); @@ -206,6 +206,7 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, q_properties->ctx_save_restore_area_address = args->ctx_save_restore_address; q_properties->ctx_save_restore_area_size = args->ctx_save_restore_size; + q_properties->ctl_stack_size = args->ctl_stack_size; if (args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE || args->queue_type == KFD_IOC_QUEUE_TYPE_COMPUTE_AQL) q_properties->type = KFD_QUEUE_TYPE_COMPUTE; @@ -1088,6 +1089,10 @@ static int kfd_mmap(struct file *filp, struct vm_area_struct *vma) KFD_MMAP_EVENTS_MASK) { vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_EVENTS_MASK; return kfd_event_mmap(process, vma); + } else if ((vma->vm_pgoff & KFD_MMAP_RESERVED_MEM_MASK) == + KFD_MMAP_RESERVED_MEM_MASK) { + vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_RESERVED_MEM_MASK; + return kfd_reserved_mem_mmap(process, vma); } return -EFAULT; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 621a3b53a0384..4f05eacca786a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -27,6 +27,7 @@ #include "kfd_priv.h" #include "kfd_device_queue_manager.h" #include "kfd_pm4_headers_vi.h" +#include "cwsr_trap_handler_gfx8.asm" #define MQD_SIZE_ALIGNED 768 @@ -38,7 +39,8 @@ static const struct kfd_device_info kaveri_device_info = { .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, - .mqd_size_aligned = MQD_SIZE_ALIGNED + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = false, }; static const struct kfd_device_info carrizo_device_info = { @@ -49,7 +51,8 @@ static const struct kfd_device_info carrizo_device_info = { .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, - .mqd_size_aligned = MQD_SIZE_ALIGNED + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = true, }; struct kfd_deviceid { @@ -212,6 +215,17 @@ static int iommu_invalid_ppr_cb(struct pci_dev *pdev, int pasid, return AMD_IOMMU_INV_PRI_RSP_INVALID; } +static void kfd_cwsr_init(struct kfd_dev *kfd) +{ + if (cwsr_enable && kfd->device_info->supports_cwsr) { + BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE); + + kfd->cwsr_isa = cwsr_trap_gfx8_hex; + kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex); + kfd->cwsr_enabled = true; + } +} + bool kgd2kfd_device_init(struct kfd_dev *kfd, const struct kgd2kfd_shared_resources *gpu_resources) { @@ -286,6 +300,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, goto device_iommu_pasid_error; } + kfd_cwsr_init(kfd); + if (kfd_resume(kfd)) goto kfd_resume_error; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index e202921c150e3..5c065024e2855 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -173,6 +173,9 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm, *allocated_vmid = qpd->vmid; q->properties.vmid = qpd->vmid; + q->properties.tba_addr = qpd->tba_addr; + q->properties.tma_addr = qpd->tma_addr; + if (q->properties.type == KFD_QUEUE_TYPE_COMPUTE) retval = create_compute_queue_nocpsch(dqm, q, qpd); else if (q->properties.type == KFD_QUEUE_TYPE_SDMA) @@ -846,6 +849,9 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, } dqm->asic_ops.init_sdma_vm(dqm, q, qpd); + + q->properties.tba_addr = qpd->tba_addr; + q->properties.tma_addr = qpd->tma_addr; retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, &q->gart_mqd_addr, &q->properties); if (retval) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index f744caeaee049..ee8adf654cd00 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -50,6 +50,10 @@ module_param(sched_policy, int, 0444); MODULE_PARM_DESC(sched_policy, "Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)"); +int cwsr_enable = 1; +module_param(cwsr_enable, int, 0444); +MODULE_PARM_DESC(cwsr_enable, "CWSR enable (0 = Off, 1 = On (Default))"); + int max_num_of_queues_per_device = KFD_MAX_NUM_OF_QUEUES_PER_DEVICE_DEFAULT; module_param(max_num_of_queues_per_device, int, 0444); MODULE_PARM_DESC(max_num_of_queues_per_device, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index 2ba7cea7b99bf..00e1f1a9728b2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -89,6 +89,28 @@ static int init_mqd(struct mqd_manager *mm, void **mqd, if (q->format == KFD_QUEUE_FORMAT_AQL) m->cp_hqd_iq_rptr = 1; + if (q->tba_addr) { + m->compute_tba_lo = lower_32_bits(q->tba_addr >> 8); + m->compute_tba_hi = upper_32_bits(q->tba_addr >> 8); + m->compute_tma_lo = lower_32_bits(q->tma_addr >> 8); + m->compute_tma_hi = upper_32_bits(q->tma_addr >> 8); + m->compute_pgm_rsrc2 |= + (1 << COMPUTE_PGM_RSRC2__TRAP_PRESENT__SHIFT); + } + + if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) { + m->cp_hqd_persistent_state |= + (1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT); + m->cp_hqd_ctx_save_base_addr_lo = + lower_32_bits(q->ctx_save_restore_area_address); + m->cp_hqd_ctx_save_base_addr_hi = + upper_32_bits(q->ctx_save_restore_area_address); + m->cp_hqd_ctx_save_size = q->ctx_save_restore_area_size; + m->cp_hqd_cntl_stack_size = q->ctl_stack_size; + m->cp_hqd_cntl_stack_offset = q->ctl_stack_size; + m->cp_hqd_wg_state_offset = q->ctl_stack_size; + } + *mqd = m; if (gart_addr) *gart_addr = addr; @@ -167,6 +189,11 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd, 2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT; } + if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) + m->cp_hqd_ctx_save_control = + atc_bit << CP_HQD_CTX_SAVE_CONTROL__ATC__SHIFT | + mtype << CP_HQD_CTX_SAVE_CONTROL__MTYPE__SHIFT; + q->is_active = (q->queue_size > 0 && q->queue_address != 0 && q->queue_percent > 0); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 47504737ab4a2..a668764679958 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -41,6 +41,7 @@ #define KFD_MMAP_DOORBELL_MASK 0x8000000000000 #define KFD_MMAP_EVENTS_MASK 0x4000000000000 +#define KFD_MMAP_RESERVED_MEM_MASK 0x2000000000000 /* * When working with cp scheduler we should assign the HIQ manually or via @@ -62,6 +63,15 @@ #define KFD_MAX_NUM_OF_PROCESSES 512 #define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024 +/* + * Size of the per-process TBA+TMA buffer: 2 pages + * + * The first page is the TBA used for the CWSR ISA code. The second + * page is used as TMA for daisy changing a user-mode trap handler. + */ +#define KFD_CWSR_TBA_TMA_SIZE (PAGE_SIZE * 2) +#define KFD_CWSR_TMA_OFFSET PAGE_SIZE + /* * Kernel module parameter to specify maximum number of supported queues per * device @@ -78,6 +88,8 @@ extern int max_num_of_queues_per_device; /* Kernel module parameter to specify the scheduling policy */ extern int sched_policy; +extern int cwsr_enable; + /* * Kernel module parameter to specify whether to send sigterm to HSA process on * unhandled exception @@ -131,6 +143,7 @@ struct kfd_device_info { size_t ih_ring_entry_size; uint8_t num_of_watch_points; uint16_t mqd_size_aligned; + bool supports_cwsr; }; struct kfd_mem_obj { @@ -200,6 +213,11 @@ struct kfd_dev { /* Debug manager */ struct kfd_dbgmgr *dbgmgr; + + /* CWSR */ + bool cwsr_enabled; + const void *cwsr_isa; + unsigned int cwsr_isa_size; }; /* KGD2KFD callbacks */ @@ -332,6 +350,9 @@ struct queue_properties { uint32_t eop_ring_buffer_size; uint64_t ctx_save_restore_area_address; uint32_t ctx_save_restore_area_size; + uint32_t ctl_stack_size; + uint64_t tba_addr; + uint64_t tma_addr; }; /** @@ -439,6 +460,11 @@ struct qcm_process_device { uint32_t num_gws; uint32_t num_oac; uint32_t sh_hidden_private_base; + + /* CWSR memory */ + void *cwsr_kaddr; + uint64_t tba_addr; + uint64_t tma_addr; }; @@ -563,7 +589,7 @@ struct amdkfd_ioctl_desc { void kfd_process_create_wq(void); void kfd_process_destroy_wq(void); -struct kfd_process *kfd_create_process(const struct task_struct *); +struct kfd_process *kfd_create_process(struct file *filep); struct kfd_process *kfd_get_process(const struct task_struct *); struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid); @@ -577,6 +603,9 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, struct kfd_process *p); +int kfd_reserved_mem_mmap(struct kfd_process *process, + struct vm_area_struct *vma); + /* Process device data iterator */ struct kfd_process_device *kfd_get_first_process_device_data( struct kfd_process *p); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 1bb9b2643d5af..39f4c19aaf617 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -28,6 +28,7 @@ #include #include #include +#include struct mm_struct; @@ -53,6 +54,8 @@ struct kfd_process_release_work { static struct kfd_process *find_process(const struct task_struct *thread); static struct kfd_process *create_process(const struct task_struct *thread); +static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep); + void kfd_process_create_wq(void) { @@ -68,9 +71,10 @@ void kfd_process_destroy_wq(void) } } -struct kfd_process *kfd_create_process(const struct task_struct *thread) +struct kfd_process *kfd_create_process(struct file *filep) { struct kfd_process *process; + struct task_struct *thread = current; if (!thread->mm) return ERR_PTR(-EINVAL); @@ -101,6 +105,8 @@ struct kfd_process *kfd_create_process(const struct task_struct *thread) up_write(&thread->mm->mmap_sem); + kfd_process_init_cwsr(process, filep); + return process; } @@ -168,6 +174,11 @@ static void kfd_process_wq_release(struct work_struct *work) amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid); list_del(&pdd->per_device_list); + + if (pdd->qpd.cwsr_kaddr) + free_pages((unsigned long)pdd->qpd.cwsr_kaddr, + get_order(KFD_CWSR_TBA_TMA_SIZE)); + kfree(pdd); } @@ -260,6 +271,46 @@ static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = { .release = kfd_process_notifier_release, }; +static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep) +{ + int err = 0; + unsigned long offset; + struct kfd_process_device *temp, *pdd = NULL; + struct kfd_dev *dev = NULL; + struct qcm_process_device *qpd = NULL; + + mutex_lock(&p->mutex); + list_for_each_entry_safe(pdd, temp, &p->per_device_data, + per_device_list) { + dev = pdd->dev; + qpd = &pdd->qpd; + if (!dev->cwsr_enabled || qpd->cwsr_kaddr) + continue; + offset = (dev->id | KFD_MMAP_RESERVED_MEM_MASK) << PAGE_SHIFT; + qpd->tba_addr = (int64_t)vm_mmap(filep, 0, + KFD_CWSR_TBA_TMA_SIZE, PROT_READ | PROT_EXEC, + MAP_SHARED, offset); + + if (IS_ERR_VALUE(qpd->tba_addr)) { + pr_err("Failure to set tba address. error -%d.\n", + (int)qpd->tba_addr); + err = qpd->tba_addr; + qpd->tba_addr = 0; + qpd->cwsr_kaddr = NULL; + goto out; + } + + memcpy(qpd->cwsr_kaddr, dev->cwsr_isa, dev->cwsr_isa_size); + + qpd->tma_addr = qpd->tba_addr + KFD_CWSR_TMA_OFFSET; + pr_debug("set tba :0x%llx, tma:0x%llx, cwsr_kaddr:%p for pqm.\n", + qpd->tba_addr, qpd->tma_addr, qpd->cwsr_kaddr); + } +out: + mutex_unlock(&p->mutex); + return err; +} + static struct kfd_process *create_process(const struct task_struct *thread) { struct kfd_process *process; @@ -535,3 +586,37 @@ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) return p; } + +int kfd_reserved_mem_mmap(struct kfd_process *process, + struct vm_area_struct *vma) +{ + struct kfd_dev *dev = kfd_device_by_id(vma->vm_pgoff); + struct kfd_process_device *pdd; + struct qcm_process_device *qpd; + + if (!dev) + return -EINVAL; + if ((vma->vm_end - vma->vm_start) != KFD_CWSR_TBA_TMA_SIZE) { + pr_err("Incorrect CWSR mapping size.\n"); + return -EINVAL; + } + + pdd = kfd_get_process_device_data(dev, process); + if (!pdd) + return -EINVAL; + qpd = &pdd->qpd; + + qpd->cwsr_kaddr = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, + get_order(KFD_CWSR_TBA_TMA_SIZE)); + if (!qpd->cwsr_kaddr) { + pr_err("Error allocating per process CWSR buffer.\n"); + return -ENOMEM; + } + + vma->vm_flags |= VM_IO | VM_DONTCOPY | VM_DONTEXPAND + | VM_NORESERVE | VM_DONTDUMP | VM_PFNMAP; + /* Mapping pages to user process */ + return remap_pfn_range(vma, vma->vm_start, + PFN_DOWN(__pa(qpd->cwsr_kaddr)), + KFD_CWSR_TBA_TMA_SIZE, vma->vm_page_prot); +} diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h index 6e80501368aee..f7563ef2e883d 100644 --- a/include/uapi/linux/kfd_ioctl.h +++ b/include/uapi/linux/kfd_ioctl.h @@ -58,7 +58,8 @@ struct kfd_ioctl_create_queue_args { __u64 eop_buffer_address; /* to KFD */ __u64 eop_buffer_size; /* to KFD */ __u64 ctx_save_restore_address; /* to KFD */ - __u64 ctx_save_restore_size; /* to KFD */ + __u32 ctx_save_restore_size; /* to KFD */ + __u32 ctl_stack_size; /* to KFD */ }; struct kfd_ioctl_destroy_queue_args { -- GitLab From d7b9bd2248d794275b53d34e665f7c5a08c4b396 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 14 Nov 2017 16:41:20 -0500 Subject: [PATCH 0246/1471] drm/amdkfd: Add support for user-mode trap handlers A second-level user mode trap handler can be installed. The CWSR trap handler jumps to the secondary trap handler conditionally for any conditions not handled by it. This can be used e.g. for debugging or catching math exceptions. When CWSR is disabled, the user mode trap handler is installed as first level trap handler. Signed-off-by: Shaoyun.liu Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 37 ++++++++++++++++++- .../drm/amd/amdkfd/kfd_device_queue_manager.c | 22 +++++++++++ .../drm/amd/amdkfd/kfd_device_queue_manager.h | 5 +++ include/uapi/linux/kfd_ioctl.h | 12 +++++- 4 files changed, 74 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 2a4612d8437ae..cc61ec2898802 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -432,6 +432,38 @@ static int kfd_ioctl_set_memory_policy(struct file *filep, return err; } +static int kfd_ioctl_set_trap_handler(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_set_trap_handler_args *args = data; + struct kfd_dev *dev; + int err = 0; + struct kfd_process_device *pdd; + + dev = kfd_device_by_id(args->gpu_id); + if (dev == NULL) + return -EINVAL; + + mutex_lock(&p->mutex); + + pdd = kfd_bind_process_to_device(dev, p); + if (IS_ERR(pdd)) { + err = -ESRCH; + goto out; + } + + if (dev->dqm->ops.set_trap_handler(dev->dqm, + &pdd->qpd, + args->tba_addr, + args->tma_addr)) + err = -EINVAL; + +out: + mutex_unlock(&p->mutex); + + return err; +} + static int kfd_ioctl_dbg_register(struct file *filep, struct kfd_process *p, void *data) { @@ -980,7 +1012,10 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = { kfd_ioctl_set_scratch_backing_va, 0), AMDKFD_IOCTL_DEF(AMDKFD_IOC_GET_TILE_CONFIG, - kfd_ioctl_get_tile_config, 0) + kfd_ioctl_get_tile_config, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_SET_TRAP_HANDLER, + kfd_ioctl_set_trap_handler, 0), }; #define AMDKFD_CORE_IOCTL_COUNT ARRAY_SIZE(amdkfd_ioctls) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 5c065024e2855..8447810c9a1e8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -1116,6 +1116,26 @@ static bool set_cache_memory_policy(struct device_queue_manager *dqm, return retval; } +static int set_trap_handler(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + uint64_t tba_addr, + uint64_t tma_addr) +{ + uint64_t *tma; + + if (dqm->dev->cwsr_enabled) { + /* Jump from CWSR trap handler to user trap */ + tma = (uint64_t *)(qpd->cwsr_kaddr + KFD_CWSR_TMA_OFFSET); + tma[0] = tba_addr; + tma[1] = tma_addr; + } else { + qpd->tba_addr = tba_addr; + qpd->tma_addr = tma_addr; + } + + return 0; +} + static int process_termination_nocpsch(struct device_queue_manager *dqm, struct qcm_process_device *qpd) { @@ -1247,6 +1267,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) dqm->ops.create_kernel_queue = create_kernel_queue_cpsch; dqm->ops.destroy_kernel_queue = destroy_kernel_queue_cpsch; dqm->ops.set_cache_memory_policy = set_cache_memory_policy; + dqm->ops.set_trap_handler = set_trap_handler; dqm->ops.process_termination = process_termination_cpsch; break; case KFD_SCHED_POLICY_NO_HWS: @@ -1262,6 +1283,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) dqm->ops.initialize = initialize_nocpsch; dqm->ops.uninitialize = uninitialize; dqm->ops.set_cache_memory_policy = set_cache_memory_policy; + dqm->ops.set_trap_handler = set_trap_handler; dqm->ops.process_termination = process_termination_nocpsch; break; default: diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index 5b77cb69f732d..8752edf9cd9b3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -123,6 +123,11 @@ struct device_queue_manager_ops { void __user *alternate_aperture_base, uint64_t alternate_aperture_size); + int (*set_trap_handler)(struct device_queue_manager *dqm, + struct qcm_process_device *qpd, + uint64_t tba_addr, + uint64_t tma_addr); + int (*process_termination)(struct device_queue_manager *dqm, struct qcm_process_device *qpd); }; diff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.h index f7563ef2e883d..f4cab5b3ba9aa 100644 --- a/include/uapi/linux/kfd_ioctl.h +++ b/include/uapi/linux/kfd_ioctl.h @@ -262,6 +262,13 @@ struct kfd_ioctl_get_tile_config_args { */ }; +struct kfd_ioctl_set_trap_handler_args { + uint64_t tba_addr; /* to KFD */ + uint64_t tma_addr; /* to KFD */ + uint32_t gpu_id; /* to KFD */ + uint32_t pad; +}; + #define AMDKFD_IOCTL_BASE 'K' #define AMDKFD_IO(nr) _IO(AMDKFD_IOCTL_BASE, nr) #define AMDKFD_IOR(nr, type) _IOR(AMDKFD_IOCTL_BASE, nr, type) @@ -322,7 +329,10 @@ struct kfd_ioctl_get_tile_config_args { #define AMDKFD_IOC_GET_TILE_CONFIG \ AMDKFD_IOWR(0x12, struct kfd_ioctl_get_tile_config_args) +#define AMDKFD_IOC_SET_TRAP_HANDLER \ + AMDKFD_IOW(0x13, struct kfd_ioctl_set_trap_handler_args) + #define AMDKFD_COMMAND_START 0x01 -#define AMDKFD_COMMAND_END 0x13 +#define AMDKFD_COMMAND_END 0x14 #endif -- GitLab From 7469c62cb6549cf72bfd4d63cb3fc5f34eb0b2d0 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 13:03:00 +0000 Subject: [PATCH 0247/1471] drm/i915: Resume GuC before using GEM Resuming GEM presumes it can talk to hw, in particular to ensure the kernel context is loaded upon resume for powersaving. If the GuC is still asleep at this point, we upset the HW. Rearrange the resume such that we restore the original order of init-hw, resume-guc, use-gem. Fixes: 37cd33006d02 ("drm/i915: Remove redundant intel_autoenable_gt_powersave()") References: a1c419941453 ("drm/i915/guc: Add host2guc notification for suspend and resume") Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Alex Dai Cc: Sagar Arun Kamble Cc: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/20171114130300.25677-2-chris@chris-wilson.co.uk Reviewed-by: Sagar Arun Kamble --- drivers/gpu/drm/i915/i915_drv.c | 2 -- drivers/gpu/drm/i915/i915_gem.c | 2 ++ 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 171b86f37878c..1ae7031eedaa4 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1708,8 +1708,6 @@ static int i915_drm_resume(struct drm_device *dev) i915_gem_resume(dev_priv); - intel_guc_resume(dev_priv); - intel_modeset_init_hw(dev); spin_lock_irq(&dev_priv->irq_lock); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index a7979b74ce21e..bf8fea7920480 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4852,6 +4852,8 @@ void i915_gem_resume(struct drm_i915_private *i915) if (i915_gem_init_hw(i915)) goto err_wedged; + intel_guc_resume(i915); + /* Always reload a context for powersaving. */ if (i915_gem_switch_to_kernel_context(i915)) goto err_wedged; -- GitLab From 6e1281412ab9e6772d8f6c26e592181fcdd2ae0c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 22:33:46 +0000 Subject: [PATCH 0248/1471] drm/i915/selftests: Always initialise err smatch does not track initialised values as well as gcc, and this triggers many warnings by smatch not presented by gcc. Silence smatch by initialising the error values to -ENODEV, which we use to denote internal errors. (If we see a selftest fail with a silent -ENODEV, we know smatch was right!) v2: smatch was right about igt_create_vma(), it may unlikely fail on the first object allocation which we want to be loud about. Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171114223346.25958-1-chris@chris-wilson.co.uk Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/selftests/i915_gem_context.c | 2 +- drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 8 ++++---- drivers/gpu/drm/i915/selftests/i915_gem_request.c | 2 +- drivers/gpu/drm/i915/selftests/i915_gem_timeline.c | 2 +- drivers/gpu/drm/i915/selftests/i915_syncmap.c | 6 +++--- drivers/gpu/drm/i915/selftests/i915_vma.c | 2 +- 6 files changed, 11 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c index def5052862aec..c82780a9d4556 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c @@ -325,7 +325,7 @@ static int igt_ctx_exec(void *arg) LIST_HEAD(objects); unsigned long ncontexts, ndwords, dw; bool first_shared_gtt = true; - int err; + int err = -ENODEV; /* Create a few different contexts (with different mm) and write * through each ctx/mm using the GPU making sure those writes end diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index 5812968605392..d9560d8a6cc87 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -699,7 +699,7 @@ static int drunk_hole(struct drm_i915_private *i915, unsigned int *order, count, n; struct i915_vma *vma; u64 hole_size; - int err; + int err = -ENODEV; hole_size = (hole_end - hole_start) >> size; if (hole_size > KMALLOC_MAX_SIZE / sizeof(u32)) @@ -958,7 +958,7 @@ static int exercise_ggtt(struct drm_i915_private *i915, u64 hole_start, hole_end, last = 0; struct drm_mm_node *node; IGT_TIMEOUT(end_time); - int err; + int err = -ENODEV; mutex_lock(&i915->drm.struct_mutex); restart: @@ -1164,7 +1164,7 @@ static int igt_gtt_reserve(void *arg) struct drm_i915_gem_object *obj, *on; LIST_HEAD(objects); u64 total; - int err; + int err = -ENODEV; /* i915_gem_gtt_reserve() tries to reserve the precise range * for the node, and evicts if it has to. So our test checks that @@ -1355,7 +1355,7 @@ static int igt_gtt_insert(void *arg) }, *ii; LIST_HEAD(objects); u64 total; - int err; + int err = -ENODEV; /* i915_gem_gtt_insert() tries to allocate some free space in the GTT * to the node, evicting if required. diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_request.c b/drivers/gpu/drm/i915/selftests/i915_gem_request.c index a999161e8db1e..6bce99050e947 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_request.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_request.c @@ -332,7 +332,7 @@ static int live_nop_request(void *arg) struct intel_engine_cs *engine; struct live_test t; unsigned int id; - int err; + int err = -ENODEV; /* Submit various sized batches of empty requests, to each engine * (individually), and wait for the batch to complete. We can check diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c index 4795877abe561..3000e6a7d82d3 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_timeline.c @@ -79,7 +79,7 @@ static int igt_sync(void *arg) }, *p; struct intel_timeline *tl; int order, offset; - int ret; + int ret = -ENODEV; tl = mock_timeline(0); if (!tl) diff --git a/drivers/gpu/drm/i915/selftests/i915_syncmap.c b/drivers/gpu/drm/i915/selftests/i915_syncmap.c index bcab3d00a7859..47f4ae18a1ef0 100644 --- a/drivers/gpu/drm/i915/selftests/i915_syncmap.c +++ b/drivers/gpu/drm/i915/selftests/i915_syncmap.c @@ -333,7 +333,7 @@ static int igt_syncmap_join_below(void *arg) { struct i915_syncmap *sync; unsigned int step, order, idx; - int err; + int err = -ENODEV; i915_syncmap_init(&sync); @@ -402,7 +402,7 @@ static int igt_syncmap_neighbours(void *arg) I915_RND_STATE(prng); IGT_TIMEOUT(end_time); struct i915_syncmap *sync; - int err; + int err = -ENODEV; /* * Each leaf holds KSYNCMAP seqno. Check that when we create KSYNCMAP @@ -447,7 +447,7 @@ static int igt_syncmap_compact(void *arg) { struct i915_syncmap *sync; unsigned int idx, order; - int err; + int err = -ENODEV; i915_syncmap_init(&sync); diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c index 2e86ec136b355..eb89e301b602f 100644 --- a/drivers/gpu/drm/i915/selftests/i915_vma.c +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c @@ -150,7 +150,7 @@ static int igt_vma_create(void *arg) IGT_TIMEOUT(end_time); LIST_HEAD(contexts); LIST_HEAD(objects); - int err; + int err = -ENOMEM; /* Exercise creating many vma amonst many objections, checking the * vma creation and lookup routines. -- GitLab From ce30560c80dead91e98a03d90fb8791e57a9b69d Mon Sep 17 00:00:00 2001 From: Hans de Goede Date: Fri, 10 Nov 2017 16:03:01 +0100 Subject: [PATCH 0249/1471] drm/i915: Fix false-positive assert_rpm_wakelock_held in i915_pmic_bus_access_notifier v2 assert_rpm_wakelock_held is triggered from i915_pmic_bus_access_notifier even though it gets unregistered on (runtime) suspend, this is caused by a race happening under the following circumstances: intel_runtime_pm_put does: atomic_dec(&dev_priv->pm.wakeref_count); pm_runtime_mark_last_busy(kdev); pm_runtime_put_autosuspend(kdev); And pm_runtime_put_autosuspend calls intel_runtime_suspend from a workqueue, so there is ample of time between the atomic_dec() and intel_runtime_suspend() unregistering the notifier. If the notifier gets called in this windowd assert_rpm_wakelock_held falsely triggers (at this point we're not runtime-suspended yet). This commit adds disable_rpm_wakeref_asserts and enable_rpm_wakeref_asserts calls around the intel_uncore_forcewake_get(FORCEWAKE_ALL) call in i915_pmic_bus_access_notifier fixing the false-positive WARN_ON. Changes in v2: -Reword comment explaining why disabling the wakeref asserts is ok and necessary Cc: stable@vger.kernel.org Reported-by: FKr Reviewed-by: Imre Deak Signed-off-by: Hans de Goede Link: https://patchwork.freedesktop.org/patch/msgid/20171110150301.9601-2-hdegoede@redhat.com --- drivers/gpu/drm/i915/intel_uncore.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index a78ceafcc8254..6e01c1898aea0 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -1364,8 +1364,15 @@ static int i915_pmic_bus_access_notifier(struct notifier_block *nb, * bus, which will be busy after this notification, leading to: * "render: timed out waiting for forcewake ack request." * errors. + * + * The notifier is unregistered during intel_runtime_suspend(), + * so it's ok to access the HW here without holding a RPM + * wake reference -> disable wakeref asserts for the time of + * the access. */ + disable_rpm_wakeref_asserts(dev_priv); intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL); + enable_rpm_wakeref_asserts(dev_priv); break; case MBI_PMIC_BUS_ACCESS_END: intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); -- GitLab From bedf4d79c3654921839b62246b0965ddb308b201 Mon Sep 17 00:00:00 2001 From: Hans de Goede Date: Tue, 14 Nov 2017 14:55:17 +0100 Subject: [PATCH 0250/1471] drm/i915: Re-register PMIC bus access notifier on runtime resume intel_uncore_suspend() unregisters the uncore code's PMIC bus access notifier and gets called on both normal and runtime suspend. intel_uncore_resume_early() re-registers the notifier, but only on normal resume. Add a new intel_uncore_runtime_resume() function which only re-registers the notifier and call that on runtime resume. Cc: stable@vger.kernel.org Reported-by: Imre Deak Reviewed-by: Imre Deak Signed-off-by: Hans de Goede Link: https://patchwork.freedesktop.org/patch/msgid/20171114135518.15981-2-hdegoede@redhat.com --- drivers/gpu/drm/i915/i915_drv.c | 2 ++ drivers/gpu/drm/i915/intel_uncore.c | 6 ++++++ drivers/gpu/drm/i915/intel_uncore.h | 1 + 3 files changed, 9 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 1ae7031eedaa4..f4588add15db2 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -2621,6 +2621,8 @@ static int intel_runtime_resume(struct device *kdev) ret = vlv_resume_prepare(dev_priv, true); } + intel_uncore_runtime_resume(dev_priv); + /* * No point of rolling back things in case of an error, as the best * we can do is to hope that things will still work (and disable RPM). diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 6e01c1898aea0..b4621271e7a2a 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -557,6 +557,12 @@ void intel_uncore_resume_early(struct drm_i915_private *dev_priv) i915_check_and_clear_faults(dev_priv); } +void intel_uncore_runtime_resume(struct drm_i915_private *dev_priv) +{ + iosf_mbi_register_pmic_bus_access_notifier( + &dev_priv->uncore.pmic_bus_access_nb); +} + void intel_uncore_sanitize(struct drm_i915_private *dev_priv) { i915_modparams.enable_rc6 = diff --git a/drivers/gpu/drm/i915/intel_uncore.h b/drivers/gpu/drm/i915/intel_uncore.h index 582771251b57a..9ce079b5dd0d8 100644 --- a/drivers/gpu/drm/i915/intel_uncore.h +++ b/drivers/gpu/drm/i915/intel_uncore.h @@ -134,6 +134,7 @@ bool intel_uncore_arm_unclaimed_mmio_detection(struct drm_i915_private *dev_priv void intel_uncore_fini(struct drm_i915_private *dev_priv); void intel_uncore_suspend(struct drm_i915_private *dev_priv); void intel_uncore_resume_early(struct drm_i915_private *dev_priv); +void intel_uncore_runtime_resume(struct drm_i915_private *dev_priv); u64 intel_uncore_edram_size(struct drm_i915_private *dev_priv); void assert_forcewakes_inactive(struct drm_i915_private *dev_priv); -- GitLab From 01c799c995bb8a87878b0d64edfd33014968da8e Mon Sep 17 00:00:00 2001 From: Hans de Goede Date: Tue, 14 Nov 2017 14:55:18 +0100 Subject: [PATCH 0251/1471] drm/i915: Call uncore_suspend before platform suspend handlers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Quoting Ville: "the forcewake timer might still be active until the uncore suspend, and having active forcewakes while we've already told the GT wake stuff to stop acting normally doesn't seem quite right to me." Reported-by: Ville Syrjälä Suggested-by: Imre Deak Reviewed-by: Imre Deak Signed-off-by: Hans de Goede Link: https://patchwork.freedesktop.org/patch/msgid/20171114135518.15981-3-hdegoede@redhat.com --- drivers/gpu/drm/i915/i915_drv.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index f4588add15db2..3423d873123a7 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -2526,6 +2526,8 @@ static int intel_runtime_suspend(struct device *kdev) intel_runtime_pm_disable_interrupts(dev_priv); + intel_uncore_suspend(dev_priv); + ret = 0; if (IS_GEN9_LP(dev_priv)) { bxt_display_core_uninit(dev_priv); @@ -2538,6 +2540,8 @@ static int intel_runtime_suspend(struct device *kdev) if (ret) { DRM_ERROR("Runtime suspend failed, disabling it (%d)\n", ret); + intel_uncore_runtime_resume(dev_priv); + intel_runtime_pm_enable_interrupts(dev_priv); enable_rpm_wakeref_asserts(dev_priv); @@ -2545,8 +2549,6 @@ static int intel_runtime_suspend(struct device *kdev) return ret; } - intel_uncore_suspend(dev_priv); - enable_rpm_wakeref_asserts(dev_priv); WARN_ON_ONCE(atomic_read(&dev_priv->runtime_pm.wakeref_count)); -- GitLab From 1f2d9bdc9085a4477a997344b17c40ba2a28d4e4 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Fri, 10 Nov 2017 11:53:12 +0100 Subject: [PATCH 0252/1471] drm/atomic-helper: always track connector commits, too MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It's useful for syncing async connector work like link retraining. v2: Make it work (Manasi&Ville) Cc: Manasi Navare Cc: Maarten Lankhorst Cc: Ville Syrjälä Signed-off-by: Daniel Vetter Reviewed-by: Maarten Lankhorst Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110105313.16718-1-daniel.vetter@ffwll.ch --- drivers/gpu/drm/drm_atomic_helper.c | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 15cc189ed1a33..07d4058d97d0d 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -1793,11 +1793,8 @@ int drm_atomic_helper_setup_commit(struct drm_atomic_state *state, !try_wait_for_completion(&old_conn_state->commit->flip_done)) return -EBUSY; - /* commit tracked through new_crtc_state->commit, no need to do it explicitly */ - if (new_conn_state->crtc) - continue; - - commit = crtc_or_fake_commit(state, old_conn_state->crtc); + /* Always track connectors explicitly for e.g. link retraining. */ + commit = crtc_or_fake_commit(state, new_conn_state->crtc ?: old_conn_state->crtc); if (!commit) return -ENOMEM; @@ -1811,10 +1808,7 @@ int drm_atomic_helper_setup_commit(struct drm_atomic_state *state, !try_wait_for_completion(&old_plane_state->commit->flip_done)) return -EBUSY; - /* - * Unlike connectors, always track planes explicitly for - * async pageflip support. - */ + /* Always track planes explicitly for async pageflip support. */ commit = crtc_or_fake_commit(state, new_plane_state->crtc ?: old_plane_state->crtc); if (!commit) return -ENOMEM; -- GitLab From 764b9f2c4d252fd893a07b39b4ab14ae8044c9cb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Tue, 14 Nov 2017 21:11:27 +0200 Subject: [PATCH 0253/1471] drm/i915: Fix kerneldocs for intel_audio.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix copy/paste fail in kerneldocs for intel_audio_codec_disable(). Cc: Daniel Vetter Suggested-by: Daniel Vetter Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171114191127.16188-1-ville.syrjala@linux.intel.com Reviewed-by: Dhinakaran Pandiyan --- drivers/gpu/drm/i915/intel_audio.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_audio.c b/drivers/gpu/drm/i915/intel_audio.c index 4705194b19929..f1502a0188eb2 100644 --- a/drivers/gpu/drm/i915/intel_audio.c +++ b/drivers/gpu/drm/i915/intel_audio.c @@ -655,8 +655,8 @@ void intel_audio_codec_enable(struct intel_encoder *encoder, /** * intel_audio_codec_disable - Disable the audio codec for HD audio * @encoder: encoder on which to disable audio - * @crtc_state: pointer to the old crtc state. - * @conn_state: pointer to the old connector state. + * @old_crtc_state: pointer to the old crtc state. + * @old_conn_state: pointer to the old connector state. * * The disable sequences must be performed before disabling the transcoder or * port. -- GitLab From db7fb60593e42824ea217ff8ed14b74376ac98c8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 2 Nov 2017 17:17:35 +0200 Subject: [PATCH 0254/1471] drm/i915: Check if the stolen memory "reserved" area is enabled or not MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Apparently there are some machines that put semi-sensible looking values into the stolen "reserved" base and size, except those values are actually outside the stolen memory. There is a bit in the register which supposedly could tell us whether the reserved area is even enabled or not. Let's check for that before we go trusting the base and size. Cc: Tomi Sarvela Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171102151737.23336-1-ville.syrjala@linux.intel.com Reviewed-by: Paulo Zanoni --- drivers/gpu/drm/i915/i915_gem_stolen.c | 30 ++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_reg.h | 2 ++ 2 files changed, 32 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index 03e7abc7e043a..44a5dbc8c23b1 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -294,6 +294,12 @@ static void g4x_get_stolen_reserved(struct drm_i915_private *dev_priv, ELK_STOLEN_RESERVED); dma_addr_t stolen_top = dev_priv->mm.stolen_base + ggtt->stolen_size; + if ((reg_val & G4X_STOLEN_RESERVED_ENABLE) == 0) { + *base = 0; + *size = 0; + return; + } + *base = (reg_val & G4X_STOLEN_RESERVED_ADDR2_MASK) << 16; WARN_ON((reg_val & G4X_STOLEN_RESERVED_ADDR1_MASK) < *base); @@ -313,6 +319,12 @@ static void gen6_get_stolen_reserved(struct drm_i915_private *dev_priv, { uint32_t reg_val = I915_READ(GEN6_STOLEN_RESERVED); + if ((reg_val & GEN6_STOLEN_RESERVED_ENABLE) == 0) { + *base = 0; + *size = 0; + return; + } + *base = reg_val & GEN6_STOLEN_RESERVED_ADDR_MASK; switch (reg_val & GEN6_STOLEN_RESERVED_SIZE_MASK) { @@ -339,6 +351,12 @@ static void gen7_get_stolen_reserved(struct drm_i915_private *dev_priv, { uint32_t reg_val = I915_READ(GEN6_STOLEN_RESERVED); + if ((reg_val & GEN6_STOLEN_RESERVED_ENABLE) == 0) { + *base = 0; + *size = 0; + return; + } + *base = reg_val & GEN7_STOLEN_RESERVED_ADDR_MASK; switch (reg_val & GEN7_STOLEN_RESERVED_SIZE_MASK) { @@ -359,6 +377,12 @@ static void chv_get_stolen_reserved(struct drm_i915_private *dev_priv, { uint32_t reg_val = I915_READ(GEN6_STOLEN_RESERVED); + if ((reg_val & GEN6_STOLEN_RESERVED_ENABLE) == 0) { + *base = 0; + *size = 0; + return; + } + *base = reg_val & GEN6_STOLEN_RESERVED_ADDR_MASK; switch (reg_val & GEN8_STOLEN_RESERVED_SIZE_MASK) { @@ -387,6 +411,12 @@ static void bdw_get_stolen_reserved(struct drm_i915_private *dev_priv, uint32_t reg_val = I915_READ(GEN6_STOLEN_RESERVED); dma_addr_t stolen_top; + if ((reg_val & GEN6_STOLEN_RESERVED_ENABLE) == 0) { + *base = 0; + *size = 0; + return; + } + stolen_top = dev_priv->mm.stolen_base + ggtt->stolen_size; *base = reg_val & GEN6_STOLEN_RESERVED_ADDR_MASK; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index cfdf4f821ac3a..107e2d7c9fba7 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -379,6 +379,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define GEN8_STOLEN_RESERVED_2M (1 << 7) #define GEN8_STOLEN_RESERVED_4M (2 << 7) #define GEN8_STOLEN_RESERVED_8M (3 << 7) +#define GEN6_STOLEN_RESERVED_ENABLE (1 << 0) /* VGA stuff */ @@ -3431,6 +3432,7 @@ enum i915_power_well_id { #define ELK_STOLEN_RESERVED _MMIO(MCHBAR_MIRROR_BASE + 0x48) #define G4X_STOLEN_RESERVED_ADDR1_MASK (0xFFFF << 16) #define G4X_STOLEN_RESERVED_ADDR2_MASK (0xFFF << 4) +#define G4X_STOLEN_RESERVED_ENABLE (1 << 0) /* Memory controller frequency in MCHBAR for Haswell (possible SNB+) */ #define DCLK _MMIO(MCHBAR_MIRROR_BASE_SNB + 0x5e04) -- GitLab From da1dd0dbe0244db849395df5e00cd62a6d8b0ed4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 2 Nov 2017 17:17:36 +0200 Subject: [PATCH 0255/1471] drm/i915: Make the report about a bogus stolen reserved area an error MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that we should be properly filtering out the cases when the stolen reserved area is disabled, let's convert the debug message about a misplaced reserved area into an error. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171102151737.23336-2-ville.syrjala@linux.intel.com Reviewed-by: Paulo Zanoni --- drivers/gpu/drm/i915/i915_gem_stolen.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index 44a5dbc8c23b1..4f9377b5f4aed 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -503,9 +503,9 @@ int i915_gem_init_stolen(struct drm_i915_private *dev_priv) if (reserved_base < dev_priv->mm.stolen_base || reserved_base + reserved_size > stolen_top) { dma_addr_t reserved_top = reserved_base + reserved_size; - DRM_DEBUG_KMS("Stolen reserved area [%pad - %pad] outside stolen memory [%pad - %pad]\n", - &reserved_base, &reserved_top, - &dev_priv->mm.stolen_base, &stolen_top); + DRM_ERROR("Stolen reserved area [%pad - %pad] outside stolen memory [%pad - %pad]\n", + &reserved_base, &reserved_top, + &dev_priv->mm.stolen_base, &stolen_top); return 0; } -- GitLab From b099a4459d99fbc032b046e55ca77c8577f5a214 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 2 Nov 2017 17:17:37 +0200 Subject: [PATCH 0256/1471] drm/i915: Use ELK stolen memory reserved detection for ILK MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit While I have no solid proof that ILK follows the ELK path when it comes to the stolen memory reserved area, there are some hints that it might be the case. Unfortunately my ILK doesn't have this enabled, and no way to enable it via the BIOS it seems. So let's have ILK use the ELK code path, and let's toss in a WARN into the code to see if we catch anyone with an ILK that has this enabled to further analyze the situation. Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171102151737.23336-3-ville.syrjala@linux.intel.com Acked-by: Paulo Zanoni --- drivers/gpu/drm/i915/i915_gem_stolen.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index 4f9377b5f4aed..1877ae9a1d9b4 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -300,6 +300,12 @@ static void g4x_get_stolen_reserved(struct drm_i915_private *dev_priv, return; } + /* + * Whether ILK really reuses the ELK register for this is unclear. + * Let's see if we catch anyone with this supposedly enabled on ILK. + */ + WARN(IS_GEN5(dev_priv), "ILK stolen reserved found? 0x%08x\n", reg_val); + *base = (reg_val & G4X_STOLEN_RESERVED_ADDR2_MASK) << 16; WARN_ON((reg_val & G4X_STOLEN_RESERVED_ADDR1_MASK) < *base); @@ -466,14 +472,12 @@ int i915_gem_init_stolen(struct drm_i915_private *dev_priv) case 3: break; case 4: - if (IS_G4X(dev_priv)) - g4x_get_stolen_reserved(dev_priv, - &reserved_base, &reserved_size); - break; + if (!IS_G4X(dev_priv)) + break; + /* fall through */ case 5: - /* Assume the gen6 maximum for the older platforms. */ - reserved_size = 1024 * 1024; - reserved_base = stolen_top - reserved_size; + g4x_get_stolen_reserved(dev_priv, + &reserved_base, &reserved_size); break; case 6: gen6_get_stolen_reserved(dev_priv, -- GitLab From fb4e14860b5e24645926ff46d3fe7237e97acc0f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 19:18:42 +0000 Subject: [PATCH 0257/1471] drm/i915/selftests: Markup __iomem for igt_gem_coherency Silence sparse warnings by using __iomem markup and io accessors. Signed-off-by: Chris Wilson Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171114191842.19063-1-chris@chris-wilson.co.uk --- .../gpu/drm/i915/selftests/i915_gem_coherency.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c index 35d778d70626b..7a0d1e17c1ad0 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c @@ -33,7 +33,7 @@ static int cpu_set(struct drm_i915_gem_object *obj, { unsigned int needs_clflush; struct page *page; - typeof(v) *map; + u32 *map; int err; err = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush); @@ -59,7 +59,7 @@ static int cpu_get(struct drm_i915_gem_object *obj, { unsigned int needs_clflush; struct page *page; - typeof(v) map; + u32 *map; int err; err = i915_gem_obj_prepare_shmem_read(obj, &needs_clflush); @@ -82,7 +82,7 @@ static int gtt_set(struct drm_i915_gem_object *obj, u32 v) { struct i915_vma *vma; - typeof(v) *map; + u32 __iomem *map; int err; err = i915_gem_object_set_to_gtt_domain(obj, true); @@ -98,7 +98,7 @@ static int gtt_set(struct drm_i915_gem_object *obj, if (IS_ERR(map)) return PTR_ERR(map); - map[offset / sizeof(*map)] = v; + iowrite32(v, &map[offset / sizeof(*map)]); i915_vma_unpin_iomap(vma); return 0; @@ -109,7 +109,7 @@ static int gtt_get(struct drm_i915_gem_object *obj, u32 *v) { struct i915_vma *vma; - typeof(v) map; + u32 __iomem *map; int err; err = i915_gem_object_set_to_gtt_domain(obj, false); @@ -125,7 +125,7 @@ static int gtt_get(struct drm_i915_gem_object *obj, if (IS_ERR(map)) return PTR_ERR(map); - *v = map[offset / sizeof(*map)]; + *v = ioread32(&map[offset / sizeof(*map)]); i915_vma_unpin_iomap(vma); return 0; @@ -135,7 +135,7 @@ static int wc_set(struct drm_i915_gem_object *obj, unsigned long offset, u32 v) { - typeof(v) *map; + u32 *map; int err; err = i915_gem_object_set_to_wc_domain(obj, true); @@ -156,7 +156,7 @@ static int wc_get(struct drm_i915_gem_object *obj, unsigned long offset, u32 *v) { - typeof(v) map; + u32 *map; int err; err = i915_gem_object_set_to_wc_domain(obj, false); -- GitLab From 4667c2d54cff89d6413a49f60c784a7d17d3e7a7 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 15 Nov 2017 13:17:05 +0000 Subject: [PATCH 0258/1471] drm/i915: Initialise entry in intel_ppat_get() for older compilers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit gcc-4.7.3 is confused by the guards inside intel_ppat_get() and reports: drivers/gpu/drm/i915/i915_gem_gtt.c: In function ‘intel_ppat_get’: drivers/gpu/drm/i915/i915_gem_gtt.c:3044:27: warning: ‘entry’ may be used uninitialized in this function [-Wmaybe-uninitialized] Forgive the compiler this once, and rearrange the code so that entry is always initialised. v2: Flavour with a bit of NULL (instead of ERR_PTR(-ENOSPC)) Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Cc: Zhi Wang Cc: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171115131705.16341-1-chris@chris-wilson.co.uk Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_gem_gtt.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 3c3a699436c92..f92a39fc511cf 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -3076,7 +3076,7 @@ const struct intel_ppat_entry * intel_ppat_get(struct drm_i915_private *i915, u8 value) { struct intel_ppat *ppat = &i915->ppat; - struct intel_ppat_entry *entry; + struct intel_ppat_entry *entry = NULL; unsigned int scanned, best_score; int i; @@ -3099,7 +3099,7 @@ intel_ppat_get(struct drm_i915_private *i915, u8 value) } if (scanned == ppat->max_entries) { - if (!best_score) + if (!entry) return ERR_PTR(-ENOSPC); kref_get(&entry->ref); -- GitLab From 24fd018aaeaecdba79f85e2232ca37a412e2754b Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 15 Nov 2017 15:12:03 +0000 Subject: [PATCH 0259/1471] drm/i915/selftests: Increase size for mock ringbuffer We don't actually emit any commands into the ringbuffer, so we set it very small. However, an upcoming change centralises the wait-for-space into i915_gem_request_alloc() and that imposes a minimum size upon all ringbuffers (mock or real) of MIN_SPACE_FOR_ADD_REQUEST. Grow the mock ringbuffer such that we allocate a single page for the struct+buffer, satisfying the new condition without wasting too much space. Signed-off-by: Chris Wilson Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171115151204.8105-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/mock_engine.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/selftests/mock_engine.c b/drivers/gpu/drm/i915/selftests/mock_engine.c index 0aafa8a105b8f..55c0e2c15782f 100644 --- a/drivers/gpu/drm/i915/selftests/mock_engine.c +++ b/drivers/gpu/drm/i915/selftests/mock_engine.c @@ -124,9 +124,11 @@ static void mock_submit_request(struct drm_i915_gem_request *request) static struct intel_ring *mock_ring(struct intel_engine_cs *engine) { - const unsigned long sz = roundup_pow_of_two(sizeof(struct intel_ring)); + const unsigned long sz = PAGE_SIZE / 2; struct intel_ring *ring; + BUILD_BUG_ON(MIN_SPACE_FOR_ADD_REQUEST > sz); + ring = kzalloc(sizeof(*ring) + sz, GFP_KERNEL); if (!ring) return NULL; -- GitLab From fd13821219dda093e402c5849e5d4525bb64b4f3 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 15 Nov 2017 15:12:04 +0000 Subject: [PATCH 0260/1471] drm/i915: Make request's wait-for-space explicit At the start of building a request, we would wait for roughly enough space to fit the average request (to reduce the likelihood of having to wait and abort partway through request construction). To achieve we would try to begin a 0-length command packet, this just adds extra confusion so make the wait-for-space explicit, as in the next patch we want to move it from the backend to the i915_gem_request_alloc() so it can ensure that the wait-for-space is the first operation in building a new request. Signed-off-by: Chris Wilson Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171115151204.8105-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_lrc.c | 8 ++-- drivers/gpu/drm/i915/intel_ringbuffer.c | 56 ++++++++++++++++--------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + 3 files changed, 41 insertions(+), 24 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 58d050a9a8665..ebd9596fe83b7 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1180,7 +1180,7 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request) { struct intel_engine_cs *engine = request->engine; struct intel_context *ce = &request->ctx->engine[engine->id]; - u32 *cs; + int ret; GEM_BUG_ON(!ce->pin_count); @@ -1190,9 +1190,9 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request) */ request->reserved_space += EXECLISTS_REQUEST_SIZE; - cs = intel_ring_begin(request, 0); - if (IS_ERR(cs)) - return PTR_ERR(cs); + ret = intel_ring_wait_for_space(request->ring, request->reserved_space); + if (ret) + return ret; /* Note that after this point, we have committed to using * this request as it is being used to both track the diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 3321b801e77d7..12e734b29463d 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1578,7 +1578,7 @@ void intel_legacy_submission_resume(struct drm_i915_private *dev_priv) static int ring_request_alloc(struct drm_i915_gem_request *request) { - u32 *cs; + int ret; GEM_BUG_ON(!request->ctx->engine[request->engine->id].pin_count); @@ -1588,37 +1588,24 @@ static int ring_request_alloc(struct drm_i915_gem_request *request) */ request->reserved_space += LEGACY_REQUEST_SIZE; - cs = intel_ring_begin(request, 0); - if (IS_ERR(cs)) - return PTR_ERR(cs); + ret = intel_ring_wait_for_space(request->ring, request->reserved_space); + if (ret) + return ret; request->reserved_space -= LEGACY_REQUEST_SIZE; return 0; } -static noinline int wait_for_space(struct drm_i915_gem_request *req, - unsigned int bytes) +static noinline int wait_for_space(struct intel_ring *ring, unsigned int bytes) { - struct intel_ring *ring = req->ring; struct drm_i915_gem_request *target; long timeout; - lockdep_assert_held(&req->i915->drm.struct_mutex); + lockdep_assert_held(&ring->vma->vm->i915->drm.struct_mutex); if (intel_ring_update_space(ring) >= bytes) return 0; - /* - * Space is reserved in the ringbuffer for finalising the request, - * as that cannot be allowed to fail. During request finalisation, - * reserved_space is set to 0 to stop the overallocation and the - * assumption is that then we never need to wait (which has the - * risk of failing with EINTR). - * - * See also i915_gem_request_alloc() and i915_add_request(). - */ - GEM_BUG_ON(!req->reserved_space); - list_for_each_entry(target, &ring->request_list, ring_link) { /* Would completion of this request free enough space? */ if (bytes <= __intel_ring_space(target->postfix, @@ -1642,6 +1629,22 @@ static noinline int wait_for_space(struct drm_i915_gem_request *req, return 0; } +int intel_ring_wait_for_space(struct intel_ring *ring, unsigned int bytes) +{ + GEM_BUG_ON(bytes > ring->effective_size); + if (unlikely(bytes > ring->effective_size - ring->emit)) + bytes += ring->size - ring->emit; + + if (unlikely(bytes > ring->space)) { + int ret = wait_for_space(ring, bytes); + if (unlikely(ret)) + return ret; + } + + GEM_BUG_ON(ring->space < bytes); + return 0; +} + u32 *intel_ring_begin(struct drm_i915_gem_request *req, unsigned int num_dwords) { @@ -1681,7 +1684,20 @@ u32 *intel_ring_begin(struct drm_i915_gem_request *req, } if (unlikely(total_bytes > ring->space)) { - int ret = wait_for_space(req, total_bytes); + int ret; + + /* + * Space is reserved in the ringbuffer for finalising the + * request, as that cannot be allowed to fail. During request + * finalisation, reserved_space is set to 0 to stop the + * overallocation and the assumption is that then we never need + * to wait (which has the risk of failing with EINTR). + * + * See also i915_gem_request_alloc() and i915_add_request(). + */ + GEM_BUG_ON(!req->reserved_space); + + ret = wait_for_space(ring, total_bytes); if (unlikely(ret)) return ERR_PTR(ret); } diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index ef22c994038bb..15a15cb876a63 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -660,6 +660,7 @@ void intel_legacy_submission_resume(struct drm_i915_private *dev_priv); int __must_check intel_ring_cacheline_align(struct drm_i915_gem_request *req); +int intel_ring_wait_for_space(struct intel_ring *ring, unsigned int bytes); u32 __must_check *intel_ring_begin(struct drm_i915_gem_request *req, unsigned int n); -- GitLab From 998fb1a0f478b83492220ff79583bf9ad538bdd8 Mon Sep 17 00:00:00 2001 From: Liviu Dudau Date: Fri, 10 Nov 2017 13:33:10 +0000 Subject: [PATCH 0261/1471] drm: gem_cma_helper.c: Allow importing of contiguous scatterlists with nents > 1 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_gem_cma_prime_import_sg_table() will fail if the number of entries in the sg_table > 1. However, you can have a device that uses an IOMMU engine and can map a discontiguous buffer with multiple entries that have consecutive sg_dma_addresses, effectively making it contiguous. Allow for that scenario by testing the entries in the sg_table for contiguous coverage. Reviewed-by: Laurent Pinchart Signed-off-by: Liviu Dudau Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/20171110133310.1225-1-Liviu.Dudau@arm.com --- drivers/gpu/drm/drm_gem_cma_helper.c | 22 ++++++++++++++++++++-- include/drm/drm_gem_cma_helper.h | 4 +++- 2 files changed, 23 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/drm_gem_cma_helper.c b/drivers/gpu/drm/drm_gem_cma_helper.c index ce67d7d42c665..80a5115c38469 100644 --- a/drivers/gpu/drm/drm_gem_cma_helper.c +++ b/drivers/gpu/drm/drm_gem_cma_helper.c @@ -475,8 +475,26 @@ drm_gem_cma_prime_import_sg_table(struct drm_device *dev, { struct drm_gem_cma_object *cma_obj; - if (sgt->nents != 1) - return ERR_PTR(-EINVAL); + if (sgt->nents != 1) { + /* check if the entries in the sg_table are contiguous */ + dma_addr_t next_addr = sg_dma_address(sgt->sgl); + struct scatterlist *s; + unsigned int i; + + for_each_sg(sgt->sgl, s, sgt->nents, i) { + /* + * sg_dma_address(s) is only valid for entries + * that have sg_dma_len(s) != 0 + */ + if (!sg_dma_len(s)) + continue; + + if (sg_dma_address(s) != next_addr) + return ERR_PTR(-EINVAL); + + next_addr = sg_dma_address(s) + sg_dma_len(s); + } + } /* Create a CMA GEM buffer. */ cma_obj = __drm_gem_cma_create(dev, attach->dmabuf->size); diff --git a/include/drm/drm_gem_cma_helper.h b/include/drm/drm_gem_cma_helper.h index 76ded75f02cca..f191227f60c0f 100644 --- a/include/drm/drm_gem_cma_helper.h +++ b/include/drm/drm_gem_cma_helper.h @@ -8,7 +8,9 @@ * struct drm_gem_cma_object - GEM object backed by CMA memory allocations * @base: base GEM object * @paddr: physical address of the backing memory - * @sgt: scatter/gather table for imported PRIME buffers + * @sgt: scatter/gather table for imported PRIME buffers. The table can have + * more than one entry but they are guaranteed to have contiguous + * DMA addresses. * @vaddr: kernel virtual address of the backing memory */ struct drm_gem_cma_object { -- GitLab From 34cc9efc27e2623c76a69d2ad1fa2b972e27a2c1 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 13:51:16 +0000 Subject: [PATCH 0262/1471] drm/i915: Remove pre-production pooled-EU w/a for Broxton WaEnablePooledEuFor2x6 only applies to preproduction models, unsupported since commit 0102ba1fd8af ("drm/i915: Add early BXT sdv to the list of preproduction machines"). Signed-off-by: Chris Wilson Cc: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/20171114135116.30036-1-chris@chris-wilson.co.uk Reviewed-by: David Weinehall --- drivers/gpu/drm/i915/intel_device_info.c | 10 ---------- 1 file changed, 10 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_device_info.c b/drivers/gpu/drm/i915/intel_device_info.c index f3e4940fed499..02f8bf101ccd7 100644 --- a/drivers/gpu/drm/i915/intel_device_info.c +++ b/drivers/gpu/drm/i915/intel_device_info.c @@ -235,16 +235,6 @@ static void gen9_sseu_info_init(struct drm_i915_private *dev_priv) #define IS_SS_DISABLED(ss) (!(sseu->subslice_mask & BIT(ss))) info->has_pooled_eu = hweight8(sseu->subslice_mask) == 3; - /* - * There is a HW issue in 2x6 fused down parts that requires - * Pooled EU to be enabled as a WA. The pool configuration - * changes depending upon which subslice is fused down. This - * doesn't affect if the device has all 3 subslices enabled. - */ - /* WaEnablePooledEuFor2x6:bxt */ - info->has_pooled_eu |= (hweight8(sseu->subslice_mask) == 2 && - IS_BXT_REVID(dev_priv, 0, BXT_REVID_B_LAST)); - sseu->min_eu_in_pool = 0; if (info->has_pooled_eu) { if (IS_SS_DISABLED(2) || IS_SS_DISABLED(0)) -- GitLab From a58c38aa6cdf0d00727cafd85cd2354dec46401a Mon Sep 17 00:00:00 2001 From: Xiaolin Zhang Date: Fri, 8 Sep 2017 21:37:48 +0800 Subject: [PATCH 0263/1471] drm/i915/gvt: Add support for opregion virtualization opregion emulated with a copy from host which leads to some display bugs such as guest resolution adjustment failure due to host opregion fail to claim port D support. with a fake opregion table provided to fully emulate opregion to meet guest port requirement. v1 - initial patch v2 - reforamt opregion arrary with 0x02x output v3 - opregion array removed with opregion generation on host initizaiton v4 - rebased v3 patch from stable branch to staging branch which also has different struct child_device_config and addressed v3 review comments. Signed-off-by: Xiaolin Zhang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/opregion.c | 180 ++++++++++++++++++++++++---- drivers/gpu/drm/i915/gvt/reg.h | 3 + 2 files changed, 159 insertions(+), 24 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c index 311799136d7f6..db8abac5faad4 100644 --- a/drivers/gpu/drm/i915/gvt/opregion.c +++ b/drivers/gpu/drm/i915/gvt/opregion.c @@ -25,9 +25,133 @@ #include "i915_drv.h" #include "gvt.h" +/* + * Note: Only for GVT-g virtual VBT generation, other usage must + * not do like this. + */ +#define _INTEL_BIOS_PRIVATE +#include "intel_vbt_defs.h" + +#define OPREGION_SIGNATURE "IntelGraphicsMem" +#define MBOX_VBT (1<<3) + +/* device handle */ +#define DEVICE_TYPE_CRT 0x01 +#define DEVICE_TYPE_EFP1 0x04 +#define DEVICE_TYPE_EFP2 0x40 +#define DEVICE_TYPE_EFP3 0x20 +#define DEVICE_TYPE_EFP4 0x10 + +#define DEV_SIZE 38 + +struct opregion_header { + u8 signature[16]; + u32 size; + u32 opregion_ver; + u8 bios_ver[32]; + u8 vbios_ver[16]; + u8 driver_ver[16]; + u32 mboxes; + u32 driver_model; + u32 pcon; + u8 dver[32]; + u8 rsvd[124]; +} __packed; + +struct bdb_data_header { + u8 id; + u16 size; /* data size */ +} __packed; + +struct vbt { + /* header->bdb_offset point to bdb_header offset */ + struct vbt_header header; + struct bdb_header bdb_header; + + struct bdb_data_header general_features_header; + struct bdb_general_features general_features; + + struct bdb_data_header general_definitions_header; + struct bdb_general_definitions general_definitions; + struct child_device_config child0; + struct child_device_config child1; + struct child_device_config child2; + struct child_device_config child3; + + struct bdb_data_header driver_features_header; + struct bdb_driver_features driver_features; +}; + +static void virt_vbt_generation(struct vbt *v) +{ + int num_child; + + memset(v, 0, sizeof(struct vbt)); + + v->header.signature[0] = '$'; + v->header.signature[1] = 'V'; + v->header.signature[2] = 'B'; + v->header.signature[3] = 'T'; + + /* there's features depending on version! */ + v->header.version = 155; + v->header.header_size = sizeof(v->header); + v->header.vbt_size = sizeof(struct vbt) - sizeof(v->header); + v->header.bdb_offset = offsetof(struct vbt, bdb_header); + + strcpy(&v->bdb_header.signature[0], "BIOS_DATA_BLOCK"); + v->bdb_header.version = 198; /* child_dev_size = 38 */ + v->bdb_header.header_size = sizeof(v->bdb_header); + + v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header) + - sizeof(struct bdb_header); + + /* general features */ + v->general_features_header.id = BDB_GENERAL_FEATURES; + v->general_features_header.size = sizeof(struct bdb_general_features); + v->general_features.int_crt_support = 0; + v->general_features.int_tv_support = 0; + + /* child device */ + num_child = 4; /* each port has one child */ + v->general_definitions_header.id = BDB_GENERAL_DEFINITIONS; + /* size will include child devices */ + v->general_definitions_header.size = + sizeof(struct bdb_general_definitions) + num_child * DEV_SIZE; + v->general_definitions.child_dev_size = DEV_SIZE; + + /* portA */ + v->child0.handle = DEVICE_TYPE_EFP1; + v->child0.device_type = DEVICE_TYPE_DP; + v->child0.dvo_port = DVO_PORT_DPA; + v->child0.aux_channel = DP_AUX_A; + + /* portB */ + v->child1.handle = DEVICE_TYPE_EFP2; + v->child1.device_type = DEVICE_TYPE_DP; + v->child1.dvo_port = DVO_PORT_DPB; + v->child1.aux_channel = DP_AUX_B; + + /* portC */ + v->child2.handle = DEVICE_TYPE_EFP3; + v->child2.device_type = DEVICE_TYPE_DP; + v->child2.dvo_port = DVO_PORT_DPC; + v->child2.aux_channel = DP_AUX_C; + + /* portD */ + v->child3.handle = DEVICE_TYPE_EFP4; + v->child3.device_type = DEVICE_TYPE_DP; + v->child3.dvo_port = DVO_PORT_DPD; + v->child3.aux_channel = DP_AUX_D; + + /* driver features */ + v->driver_features_header.id = BDB_DRIVER_FEATURES; + v->driver_features_header.size = sizeof(struct bdb_driver_features); + v->driver_features.lvds_config = BDB_DRIVER_FEATURE_NO_LVDS; +} + static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa) { - u8 *buf; int i; if (WARN((vgpu_opregion(vgpu)->va), @@ -35,26 +159,11 @@ static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa) vgpu->id)) return -EINVAL; - vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL | - __GFP_ZERO, - get_order(INTEL_GVT_OPREGION_SIZE)); - - if (!vgpu_opregion(vgpu)->va) - return -ENOMEM; - - memcpy(vgpu_opregion(vgpu)->va, vgpu->gvt->opregion.opregion_va, - INTEL_GVT_OPREGION_SIZE); + vgpu_opregion(vgpu)->va = vgpu->gvt->opregion.opregion_va; for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++) vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i; - /* for unknown reason, the value in LID field is incorrect - * which block the windows guest, so workaround it by force - * setting it to "OPEN" - */ - buf = (u8 *)vgpu_opregion(vgpu)->va; - buf[INTEL_GVT_OPREGION_CLID] = 0x3; - return 0; } @@ -139,7 +248,8 @@ int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa) */ void intel_gvt_clean_opregion(struct intel_gvt *gvt) { - memunmap(gvt->opregion.opregion_va); + free_pages((unsigned long)gvt->opregion.opregion_va, + get_order(INTEL_GVT_OPREGION_SIZE)); gvt->opregion.opregion_va = NULL; } @@ -152,17 +262,39 @@ void intel_gvt_clean_opregion(struct intel_gvt *gvt) */ int intel_gvt_init_opregion(struct intel_gvt *gvt) { + u8 *buf; + struct opregion_header *header; + struct vbt v; + gvt_dbg_core("init host opregion\n"); - pci_read_config_dword(gvt->dev_priv->drm.pdev, INTEL_GVT_PCI_OPREGION, - &gvt->opregion.opregion_pa); + gvt->opregion.opregion_va = (void *)__get_free_pages(GFP_KERNEL | + __GFP_ZERO, + get_order(INTEL_GVT_OPREGION_SIZE)); - gvt->opregion.opregion_va = memremap(gvt->opregion.opregion_pa, - INTEL_GVT_OPREGION_SIZE, MEMREMAP_WB); if (!gvt->opregion.opregion_va) { - gvt_err("fail to map host opregion\n"); - return -EFAULT; + gvt_err("fail to get memory for virt opregion\n"); + return -ENOMEM; } + + /* emulated opregion with VBT mailbox only */ + header = (struct opregion_header *)gvt->opregion.opregion_va; + memcpy(header->signature, OPREGION_SIGNATURE, + sizeof(OPREGION_SIGNATURE)); + header->mboxes = MBOX_VBT; + + /* for unknown reason, the value in LID field is incorrect + * which block the windows guest, so workaround it by force + * setting it to "OPEN" + */ + buf = (u8 *)gvt->opregion.opregion_va; + buf[INTEL_GVT_OPREGION_CLID] = 0x3; + + /* emulated vbt from virt vbt generation */ + virt_vbt_generation(&v); + memcpy(gvt->opregion.opregion_va + INTEL_GVT_OPREGION_VBT_OFFSET, + &v, sizeof(struct vbt)); + return 0; } diff --git a/drivers/gpu/drm/i915/gvt/reg.h b/drivers/gpu/drm/i915/gvt/reg.h index 7d01c77a0f7ac..83f2f63d7eeba 100644 --- a/drivers/gpu/drm/i915/gvt/reg.h +++ b/drivers/gpu/drm/i915/gvt/reg.h @@ -51,6 +51,9 @@ #define INTEL_GVT_OPREGION_PAGES 2 #define INTEL_GVT_OPREGION_SIZE (INTEL_GVT_OPREGION_PAGES * PAGE_SIZE) +#define INTEL_GVT_OPREGION_VBT_OFFSET 0x400 +#define INTEL_GVT_OPREGION_VBT_SIZE \ + (INTEL_GVT_OPREGION_SIZE - INTEL_GVT_OPREGION_VBT_OFFSET) #define VGT_SPRSTRIDE(pipe) _PIPE(pipe, _SPRA_STRIDE, _PLANE_STRIDE_2_B) -- GitLab From 54cff6479fd87e4a941e3686d93faa734586922b Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 16:40:04 +0800 Subject: [PATCH 0264/1471] drm/i915/gvt: Make elsp_dwords in the right order The context descriptors in elsp_dwords are stored in a reversed order and the definition of context descriptor is also reversed. The revesred stuff is hard to be used and might cause misunderstanding. Make them in the right oder for following code re-factoring. Tested on my SKL NUC. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 22 +++++++++++----------- drivers/gpu/drm/i915/gvt/execlist.h | 8 ++++---- drivers/gpu/drm/i915/gvt/handlers.c | 2 +- 3 files changed, 16 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 4427be18e4a93..9f02072051179 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -511,8 +511,8 @@ static int prepare_execlist_workload(struct intel_vgpu_workload *workload) if (!workload->emulate_schedule_in) return 0; - ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1); - ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0); + ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0); + ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1); ret = emulate_execlist_schedule_in(&vgpu->execlist[ring_id], ctx); if (!ret) @@ -770,21 +770,21 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) { struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id]; - struct execlist_ctx_descriptor_format desc[2]; + struct execlist_ctx_descriptor_format *desc[2]; int i, ret; - desc[0] = *get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1); - desc[1] = *get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0); + desc[0] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0); + desc[1] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1); - if (!desc[0].valid) { + if (!desc[0]->valid) { gvt_vgpu_err("invalid elsp submission, desc0 is invalid\n"); goto inv_desc; } for (i = 0; i < ARRAY_SIZE(desc); i++) { - if (!desc[i].valid) + if (!desc[i]->valid) continue; - if (!desc[i].privilege_access) { + if (!desc[i]->privilege_access) { gvt_vgpu_err("unexpected GGTT elsp submission\n"); goto inv_desc; } @@ -792,9 +792,9 @@ int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) /* submit workload */ for (i = 0; i < ARRAY_SIZE(desc); i++) { - if (!desc[i].valid) + if (!desc[i]->valid) continue; - ret = submit_context(vgpu, ring_id, &desc[i], i == 0); + ret = submit_context(vgpu, ring_id, desc[i], i == 0); if (ret) { gvt_vgpu_err("failed to submit desc %d\n", i); return ret; @@ -805,7 +805,7 @@ int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) inv_desc: gvt_vgpu_err("descriptors content: desc0 %08x %08x desc1 %08x %08x\n", - desc[0].udw, desc[0].ldw, desc[1].udw, desc[1].ldw); + desc[0]->udw, desc[0]->ldw, desc[1]->udw, desc[1]->ldw); return -EINVAL; } diff --git a/drivers/gpu/drm/i915/gvt/execlist.h b/drivers/gpu/drm/i915/gvt/execlist.h index 7eced40a1e309..427e40e64d41e 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.h +++ b/drivers/gpu/drm/i915/gvt/execlist.h @@ -36,10 +36,6 @@ #define _GVT_EXECLIST_H_ struct execlist_ctx_descriptor_format { - union { - u32 udw; - u32 context_id; - }; union { u32 ldw; struct { @@ -54,6 +50,10 @@ struct execlist_ctx_descriptor_format { u32 lrca : 20; }; }; + union { + u32 udw; + u32 context_id; + }; }; struct execlist_status_format { diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index a5bed2e71b926..820d266705411 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1453,7 +1453,7 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, execlist = &vgpu->execlist[ring_id]; - execlist->elsp_dwords.data[execlist->elsp_dwords.index] = data; + execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data; if (execlist->elsp_dwords.index == 3) { ret = intel_vgpu_submit_execlist(vgpu, ring_id); if(ret) -- GitLab From 874b6a910e6cc094629bd2634d14061cf5eb7690 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 20:08:18 +0800 Subject: [PATCH 0265/1471] drm/i915/gvt: Rename intel_vgpu_{init, clean}_gvt_context() To move workload related functions into scheduler.c, an expected way is to collect all the init/clean functions related to vGPU workload submission into fewer functions. Rename intel_vgpu_{init, clean}_gvt_context() for above usage in future. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/scheduler.c | 21 +++++++++++++++++++-- drivers/gpu/drm/i915/gvt/scheduler.h | 4 ++-- drivers/gpu/drm/i915/gvt/vgpu.c | 6 +++--- 3 files changed, 24 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 42cc61230ca7d..5913bcb7b73e5 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -709,12 +709,29 @@ int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt) return ret; } -void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu) +/** + * intel_vgpu_clean_submission - free submission-related resource for vGPU + * @vgpu: a vGPU + * + * This function is called when a vGPU is being destroyed. + * + */ +void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) { i915_gem_context_put(vgpu->shadow_ctx); } -int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu) +/** + * intel_vgpu_setup_submission - setup submission-related resource for vGPU + * @vgpu: a vGPU + * + * This function is called when a vGPU is being created. + * + * Returns: + * Zero on success, negative error code if failed. + * + */ +int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) { atomic_set(&vgpu->running_workload_num, 0); diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 2d694f6c09076..c216aefaa73e1 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -137,9 +137,9 @@ void intel_gvt_clean_workload_scheduler(struct intel_gvt *gvt); void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu); -int intel_vgpu_init_gvt_context(struct intel_vgpu *vgpu); +int intel_vgpu_setup_submission(struct intel_vgpu *vgpu); -void intel_vgpu_clean_gvt_context(struct intel_vgpu *vgpu); +void intel_vgpu_clean_submission(struct intel_vgpu *vgpu); void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx); #endif diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 02c61a1ad56a2..3d69871d28e9e 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -254,7 +254,7 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu) idr_remove(&gvt->vgpu_idr, vgpu->id); intel_vgpu_clean_sched_policy(vgpu); - intel_vgpu_clean_gvt_context(vgpu); + intel_vgpu_clean_submission(vgpu); intel_vgpu_clean_execlist(vgpu); intel_vgpu_clean_display(vgpu); intel_vgpu_clean_opregion(vgpu); @@ -376,7 +376,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, if (ret) goto out_clean_display; - ret = intel_vgpu_init_gvt_context(vgpu); + ret = intel_vgpu_setup_submission(vgpu); if (ret) goto out_clean_execlist; @@ -389,7 +389,7 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, return vgpu; out_clean_shadow_ctx: - intel_vgpu_clean_gvt_context(vgpu); + intel_vgpu_clean_submission(vgpu); out_clean_execlist: intel_vgpu_clean_execlist(vgpu); out_clean_display: -- GitLab From 9a9829e9eb8bc4b4e870ce15a8904a32991608d5 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 20:28:09 +0800 Subject: [PATCH 0266/1471] drm/i915/gvt: Move workload cache init/clean into intel_vgpu_{setup, clean}_submission() Move vGPU workload cache initialization/de-initialization into intel_vgpu_{setup, clean}_submission() since they are not specific to execlist stuffs. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 15 +-------------- drivers/gpu/drm/i915/gvt/scheduler.c | 24 +++++++++++++++++++++++- 2 files changed, 24 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 9f02072051179..0ee40a5ee1923 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -856,14 +856,12 @@ void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) struct intel_engine_cs *engine; clean_workloads(vgpu, ALL_ENGINES); - kmem_cache_destroy(vgpu->workloads); for_each_engine(engine, vgpu->gvt->dev_priv, i) { kfree(vgpu->reserve_ring_buffer_va[i]); vgpu->reserve_ring_buffer_va[i] = NULL; vgpu->reserve_ring_buffer_size[i] = 0; } - } #define RESERVE_RING_BUFFER_SIZE ((1 * PAGE_SIZE)/8) @@ -872,19 +870,8 @@ int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) enum intel_engine_id i; struct intel_engine_cs *engine; - /* each ring has a virtual execlist engine */ - for_each_engine(engine, vgpu->gvt->dev_priv, i) { + for_each_engine(engine, vgpu->gvt->dev_priv, i) init_vgpu_execlist(vgpu, i); - INIT_LIST_HEAD(&vgpu->workload_q_head[i]); - } - - vgpu->workloads = kmem_cache_create("gvt-g_vgpu_workload", - sizeof(struct intel_vgpu_workload), 0, - SLAB_HWCACHE_ALIGN, - NULL); - - if (!vgpu->workloads) - return -ENOMEM; /* each ring has a shadow ring buffer until vgpu destroyed */ for_each_engine(engine, vgpu->gvt->dev_priv, i) { diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 5913bcb7b73e5..81952139b00cc 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -719,6 +719,7 @@ int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt) void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) { i915_gem_context_put(vgpu->shadow_ctx); + kmem_cache_destroy(vgpu->workloads); } /** @@ -733,7 +734,9 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) */ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) { - atomic_set(&vgpu->running_workload_num, 0); + enum intel_engine_id i; + struct intel_engine_cs *engine; + int ret; vgpu->shadow_ctx = i915_gem_context_create_gvt( &vgpu->gvt->dev_priv->drm); @@ -742,5 +745,24 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES); + vgpu->workloads = kmem_cache_create("gvt-g_vgpu_workload", + sizeof(struct intel_vgpu_workload), 0, + SLAB_HWCACHE_ALIGN, + NULL); + + if (!vgpu->workloads) { + ret = -ENOMEM; + goto out_shadow_ctx; + } + + for_each_engine(engine, vgpu->gvt->dev_priv, i) + INIT_LIST_HEAD(&vgpu->workload_q_head[i]); + + atomic_set(&vgpu->running_workload_num, 0); + return 0; + +out_shadow_ctx: + i915_gem_context_put(vgpu->shadow_ctx); + return ret; } -- GitLab From 1406a14b0ed977fc18f43398b391e4bb5d744174 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 21:15:18 +0800 Subject: [PATCH 0267/1471] drm/i915/gvt: Introduce intel_vgpu_submission Introduce intel_vgpu_submission to hold all members related to submission in struct intel_vgpu before. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 30 ++++++++----- drivers/gpu/drm/i915/gvt/gvt.h | 17 +++++--- drivers/gpu/drm/i915/gvt/handlers.c | 2 +- drivers/gpu/drm/i915/gvt/kvmgt.c | 2 +- drivers/gpu/drm/i915/gvt/render.c | 9 ++-- drivers/gpu/drm/i915/gvt/scheduler.c | 65 +++++++++++++++------------- drivers/gpu/drm/i915/gvt/scheduler.h | 2 +- drivers/gpu/drm/i915/gvt/vgpu.c | 4 +- 8 files changed, 75 insertions(+), 56 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 0ee40a5ee1923..f118454d2eabb 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -362,7 +362,7 @@ static void free_workload(struct intel_vgpu_workload *workload) { intel_vgpu_unpin_mm(workload->shadow_mm); intel_gvt_mm_unreference(workload->shadow_mm); - kmem_cache_free(workload->vgpu->workloads, workload); + kmem_cache_free(workload->vgpu->submission.workloads, workload); } #define get_desc_from_elsp_dwords(ed, i) \ @@ -401,7 +401,8 @@ static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx) struct intel_vgpu_workload, wa_ctx); int ring_id = workload->ring_id; - struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx; + struct intel_vgpu_submission *s = &workload->vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; struct drm_i915_gem_object *ctx_obj = shadow_ctx->engine[ring_id].state->obj; struct execlist_ring_context *shadow_ring_context; @@ -474,6 +475,7 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) static int prepare_execlist_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; struct execlist_ctx_descriptor_format ctx[2]; int ring_id = workload->ring_id; int ret; @@ -514,7 +516,7 @@ static int prepare_execlist_workload(struct intel_vgpu_workload *workload) ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0); ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1); - ret = emulate_execlist_schedule_in(&vgpu->execlist[ring_id], ctx); + ret = emulate_execlist_schedule_in(&s->execlist[ring_id], ctx); if (!ret) goto out; else @@ -533,7 +535,8 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; int ring_id = workload->ring_id; - struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id]; + struct intel_vgpu_submission *s = &vgpu->submission; + struct intel_vgpu_execlist *execlist = &s->execlist[ring_id]; struct intel_vgpu_workload *next_workload; struct list_head *next = workload_q_head(vgpu, ring_id)->next; bool lite_restore = false; @@ -652,6 +655,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, struct execlist_ctx_descriptor_format *desc, bool emulate_schedule_in) { + struct intel_vgpu_submission *s = &vgpu->submission; struct list_head *q = workload_q_head(vgpu, ring_id); struct intel_vgpu_workload *last_workload = get_last_workload(q); struct intel_vgpu_workload *workload = NULL; @@ -689,7 +693,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, gvt_dbg_el("ring id %d begin a new workload\n", ring_id); - workload = kmem_cache_zalloc(vgpu->workloads, GFP_KERNEL); + workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL); if (!workload) return -ENOMEM; @@ -738,7 +742,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, } if (emulate_schedule_in) - workload->elsp_dwords = vgpu->execlist[ring_id].elsp_dwords; + workload->elsp_dwords = s->execlist[ring_id].elsp_dwords; gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n", workload, ring_id, head, tail, start, ctl); @@ -748,7 +752,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, ret = prepare_mm(workload); if (ret) { - kmem_cache_free(vgpu->workloads, workload); + kmem_cache_free(s->workloads, workload); return ret; } @@ -769,7 +773,8 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) { - struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id]; + struct intel_vgpu_submission *s = &vgpu->submission; + struct intel_vgpu_execlist *execlist = &s->execlist[ring_id]; struct execlist_ctx_descriptor_format *desc[2]; int i, ret; @@ -811,7 +816,8 @@ int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id) { - struct intel_vgpu_execlist *execlist = &vgpu->execlist[ring_id]; + struct intel_vgpu_submission *s = &vgpu->submission; + struct intel_vgpu_execlist *execlist = &s->execlist[ring_id]; struct execlist_context_status_pointer_format ctx_status_ptr; u32 ctx_status_ptr_reg; @@ -833,6 +839,7 @@ static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id) static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) { + struct intel_vgpu_submission *s = &vgpu->submission; struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; struct intel_engine_cs *engine; struct intel_vgpu_workload *pos, *n; @@ -841,12 +848,11 @@ static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) /* free the unsubmited workloads in the queues. */ for_each_engine_masked(engine, dev_priv, engine_mask, tmp) { list_for_each_entry_safe(pos, n, - &vgpu->workload_q_head[engine->id], list) { + &s->workload_q_head[engine->id], list) { list_del_init(&pos->list); free_workload(pos); } - - clear_bit(engine->id, vgpu->shadow_ctx_desc_updated); + clear_bit(engine->id, s->shadow_ctx_desc_updated); } } diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 9c2e7c0aa38fb..c3f84f26090aa 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -142,6 +142,15 @@ struct vgpu_sched_ctl { int weight; }; +struct intel_vgpu_submission { + struct intel_vgpu_execlist execlist[I915_NUM_ENGINES]; + struct list_head workload_q_head[I915_NUM_ENGINES]; + struct kmem_cache *workloads; + atomic_t running_workload_num; + struct i915_gem_context *shadow_ctx; + DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES); +}; + struct intel_vgpu { struct intel_gvt *gvt; int id; @@ -161,16 +170,12 @@ struct intel_vgpu { struct intel_vgpu_gtt gtt; struct intel_vgpu_opregion opregion; struct intel_vgpu_display display; - struct intel_vgpu_execlist execlist[I915_NUM_ENGINES]; - struct list_head workload_q_head[I915_NUM_ENGINES]; - struct kmem_cache *workloads; - atomic_t running_workload_num; + struct intel_vgpu_submission submission; /* 1/2K for each reserve ring buffer */ void *reserve_ring_buffer_va[I915_NUM_ENGINES]; int reserve_ring_buffer_size[I915_NUM_ENGINES]; DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); - struct i915_gem_context *shadow_ctx; - DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES); + #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT) struct { diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 820d266705411..00893532394ac 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1451,7 +1451,7 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, if (WARN_ON(ring_id < 0 || ring_id > I915_NUM_ENGINES - 1)) return -EINVAL; - execlist = &vgpu->execlist[ring_id]; + execlist = &vgpu->submission.execlist[ring_id]; execlist->elsp_dwords.data[3 - execlist->elsp_dwords.index] = data; if (execlist->elsp_dwords.index == 3) { diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c index 96060920a6fea..fc6878bb24968 100644 --- a/drivers/gpu/drm/i915/gvt/kvmgt.c +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c @@ -1188,7 +1188,7 @@ hw_id_show(struct device *dev, struct device_attribute *attr, struct intel_vgpu *vgpu = (struct intel_vgpu *) mdev_get_drvdata(mdev); return sprintf(buf, "%u\n", - vgpu->shadow_ctx->hw_id); + vgpu->submission.shadow_ctx->hw_id); } return sprintf(buf, "\n"); } diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c index 6d066cf354789..db4d091be60b7 100644 --- a/drivers/gpu/drm/i915/gvt/render.c +++ b/drivers/gpu/drm/i915/gvt/render.c @@ -261,14 +261,15 @@ static void restore_mocs(struct intel_vgpu *vgpu, int ring_id) static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id) { struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; - struct render_mmio *mmio; - u32 v; - int i, array_size; - u32 *reg_state = vgpu->shadow_ctx->engine[ring_id].lrc_reg_state; + struct intel_vgpu_submission *s = &vgpu->submission; + u32 *reg_state = s->shadow_ctx->engine[ring_id].lrc_reg_state; u32 ctx_ctrl = reg_state[CTX_CONTEXT_CONTROL_VAL]; u32 inhibit_mask = _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT); i915_reg_t last_reg = _MMIO(0); + struct render_mmio *mmio; + u32 v; + int i, array_size; if (IS_SKYLAKE(vgpu->gvt->dev_priv) || IS_KABYLAKE(vgpu->gvt->dev_priv)) { diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 81952139b00cc..864a2bc06e45c 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -57,7 +57,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) struct intel_vgpu *vgpu = workload->vgpu; struct intel_gvt *gvt = vgpu->gvt; int ring_id = workload->ring_id; - struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx; + struct i915_gem_context *shadow_ctx = vgpu->submission.shadow_ctx; struct drm_i915_gem_object *ctx_obj = shadow_ctx->engine[ring_id].state->obj; struct execlist_ring_context *shadow_ring_context; @@ -249,12 +249,13 @@ void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) */ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) { + struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; int ring_id = workload->ring_id; - struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx; - struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv; struct intel_engine_cs *engine = dev_priv->engine[ring_id]; struct drm_i915_gem_request *rq; - struct intel_vgpu *vgpu = workload->vgpu; struct intel_ring *ring; int ret; @@ -267,7 +268,7 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) shadow_ctx->desc_template |= workload->ctx_desc.addressing_mode << GEN8_CTX_ADDRESSING_MODE_SHIFT; - if (!test_and_set_bit(ring_id, vgpu->shadow_ctx_desc_updated)) + if (!test_and_set_bit(ring_id, s->shadow_ctx_desc_updated)) shadow_context_descriptor_update(shadow_ctx, dev_priv->engine[ring_id]); @@ -326,9 +327,11 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) static int dispatch_workload(struct intel_vgpu_workload *workload) { + struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; int ring_id = workload->ring_id; - struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx; - struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv; struct intel_engine_cs *engine = dev_priv->engine[ring_id]; int ret = 0; @@ -414,7 +417,7 @@ static struct intel_vgpu_workload *pick_next_workload( gvt_dbg_sched("ring id %d pick new workload %p\n", ring_id, workload); - atomic_inc(&workload->vgpu->running_workload_num); + atomic_inc(&workload->vgpu->submission.running_workload_num); out: mutex_unlock(&gvt->lock); return workload; @@ -424,8 +427,9 @@ static void update_guest_context(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; struct intel_gvt *gvt = vgpu->gvt; + struct intel_vgpu_submission *s = &vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; int ring_id = workload->ring_id; - struct i915_gem_context *shadow_ctx = workload->vgpu->shadow_ctx; struct drm_i915_gem_object *ctx_obj = shadow_ctx->engine[ring_id].state->obj; struct execlist_ring_context *shadow_ring_context; @@ -491,15 +495,14 @@ static void update_guest_context(struct intel_vgpu_workload *workload) static void complete_current_workload(struct intel_gvt *gvt, int ring_id) { struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler; - struct intel_vgpu_workload *workload; - struct intel_vgpu *vgpu; + struct intel_vgpu_workload *workload = + scheduler->current_workload[ring_id]; + struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; int event; mutex_lock(&gvt->lock); - workload = scheduler->current_workload[ring_id]; - vgpu = workload->vgpu; - /* For the workload w/ request, needs to wait for the context * switch to make sure request is completed. * For the workload w/o request, directly complete the workload. @@ -536,7 +539,7 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) } mutex_lock(&dev_priv->drm.struct_mutex); /* unpin shadow ctx as the shadow_ctx update is done */ - engine->context_unpin(engine, workload->vgpu->shadow_ctx); + engine->context_unpin(engine, s->shadow_ctx); mutex_unlock(&dev_priv->drm.struct_mutex); } @@ -548,7 +551,7 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) list_del_init(&workload->list); workload->complete(workload); - atomic_dec(&vgpu->running_workload_num); + atomic_dec(&s->running_workload_num); wake_up(&scheduler->workload_complete_wq); if (gvt->scheduler.need_reschedule) @@ -637,14 +640,15 @@ static int workload_thread(void *priv) void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu) { + struct intel_vgpu_submission *s = &vgpu->submission; struct intel_gvt *gvt = vgpu->gvt; struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler; - if (atomic_read(&vgpu->running_workload_num)) { + if (atomic_read(&s->running_workload_num)) { gvt_dbg_sched("wait vgpu idle\n"); wait_event(scheduler->workload_complete_wq, - !atomic_read(&vgpu->running_workload_num)); + !atomic_read(&s->running_workload_num)); } } @@ -718,8 +722,10 @@ int intel_gvt_init_workload_scheduler(struct intel_gvt *gvt) */ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) { - i915_gem_context_put(vgpu->shadow_ctx); - kmem_cache_destroy(vgpu->workloads); + struct intel_vgpu_submission *s = &vgpu->submission; + + i915_gem_context_put(s->shadow_ctx); + kmem_cache_destroy(s->workloads); } /** @@ -734,35 +740,36 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) */ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) { + struct intel_vgpu_submission *s = &vgpu->submission; enum intel_engine_id i; struct intel_engine_cs *engine; int ret; - vgpu->shadow_ctx = i915_gem_context_create_gvt( + s->shadow_ctx = i915_gem_context_create_gvt( &vgpu->gvt->dev_priv->drm); - if (IS_ERR(vgpu->shadow_ctx)) - return PTR_ERR(vgpu->shadow_ctx); + if (IS_ERR(s->shadow_ctx)) + return PTR_ERR(s->shadow_ctx); - bitmap_zero(vgpu->shadow_ctx_desc_updated, I915_NUM_ENGINES); + bitmap_zero(s->shadow_ctx_desc_updated, I915_NUM_ENGINES); - vgpu->workloads = kmem_cache_create("gvt-g_vgpu_workload", + s->workloads = kmem_cache_create("gvt-g_vgpu_workload", sizeof(struct intel_vgpu_workload), 0, SLAB_HWCACHE_ALIGN, NULL); - if (!vgpu->workloads) { + if (!s->workloads) { ret = -ENOMEM; goto out_shadow_ctx; } for_each_engine(engine, vgpu->gvt->dev_priv, i) - INIT_LIST_HEAD(&vgpu->workload_q_head[i]); + INIT_LIST_HEAD(&s->workload_q_head[i]); - atomic_set(&vgpu->running_workload_num, 0); + atomic_set(&s->running_workload_num, 0); return 0; out_shadow_ctx: - i915_gem_context_put(vgpu->shadow_ctx); + i915_gem_context_put(s->shadow_ctx); return ret; } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index c216aefaa73e1..3ca0087f10b20 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -122,7 +122,7 @@ struct intel_shadow_bb_entry { }; #define workload_q_head(vgpu, ring_id) \ - (&(vgpu->workload_q_head[ring_id])) + (&(vgpu->submission.workload_q_head[ring_id])) #define queue_workload(workload) do { \ list_add_tail(&workload->list, \ diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 3d69871d28e9e..35a5ec2012062 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -226,7 +226,7 @@ void intel_gvt_deactivate_vgpu(struct intel_vgpu *vgpu) vgpu->active = false; - if (atomic_read(&vgpu->running_workload_num)) { + if (atomic_read(&vgpu->submission.running_workload_num)) { mutex_unlock(&gvt->lock); intel_gvt_wait_vgpu_idle(vgpu); mutex_lock(&gvt->lock); @@ -293,7 +293,7 @@ struct intel_vgpu *intel_gvt_create_idle_vgpu(struct intel_gvt *gvt) vgpu->gvt = gvt; for (i = 0; i < I915_NUM_ENGINES; i++) - INIT_LIST_HEAD(&vgpu->workload_q_head[i]); + INIT_LIST_HEAD(&vgpu->submission.workload_q_head[i]); ret = intel_vgpu_init_sched_policy(vgpu); if (ret) -- GitLab From 91d5d85442b2a65e5f4e1726565c1c1a8ba9976f Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 21:33:20 +0800 Subject: [PATCH 0268/1471] drm/i915/gvt: Move tlb_handle_pending into intel_vgpu_submission Move tlb_handle_pending into intel_vgpu_submssion since it belongs to a part of vGPU submission stuffs Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gvt.h | 2 +- drivers/gpu/drm/i915/gvt/handlers.c | 2 +- drivers/gpu/drm/i915/gvt/render.c | 3 ++- drivers/gpu/drm/i915/gvt/scheduler.c | 1 + drivers/gpu/drm/i915/gvt/vgpu.c | 1 - 5 files changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index c3f84f26090aa..93ff530eee303 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -149,6 +149,7 @@ struct intel_vgpu_submission { atomic_t running_workload_num; struct i915_gem_context *shadow_ctx; DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES); + DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); }; struct intel_vgpu { @@ -174,7 +175,6 @@ struct intel_vgpu { /* 1/2K for each reserve ring buffer */ void *reserve_ring_buffer_va[I915_NUM_ENGINES]; int reserve_ring_buffer_size[I915_NUM_ENGINES]; - DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 00893532394ac..c78e45058219d 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1526,7 +1526,7 @@ static int gvt_reg_tlb_control_handler(struct intel_vgpu *vgpu, default: return -EINVAL; } - set_bit(id, (void *)vgpu->tlb_handle_pending); + set_bit(id, (void *)vgpu->submission.tlb_handle_pending); return 0; } diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c index db4d091be60b7..e16c3551b4a3e 100644 --- a/drivers/gpu/drm/i915/gvt/render.c +++ b/drivers/gpu/drm/i915/gvt/render.c @@ -147,6 +147,7 @@ static u32 gen9_render_mocs_L3[32]; static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id) { struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + struct intel_vgpu_submission *s = &vgpu->submission; enum forcewake_domains fw; i915_reg_t reg; u32 regs[] = { @@ -160,7 +161,7 @@ static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id) if (WARN_ON(ring_id >= ARRAY_SIZE(regs))) return; - if (!test_and_clear_bit(ring_id, (void *)vgpu->tlb_handle_pending)) + if (!test_and_clear_bit(ring_id, (void *)s->tlb_handle_pending)) return; reg = _MMIO(regs[ring_id]); diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 864a2bc06e45c..7cb1cf4223ed4 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -766,6 +766,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) INIT_LIST_HEAD(&s->workload_q_head[i]); atomic_set(&s->running_workload_num, 0); + bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES); return 0; diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 35a5ec2012062..1c9818d9f1d85 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -346,7 +346,6 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, vgpu->handle = param->handle; vgpu->gvt = gvt; vgpu->sched_ctl.weight = param->weight; - bitmap_zero(vgpu->tlb_handle_pending, I915_NUM_ENGINES); intel_vgpu_init_cfg_space(vgpu, param->primary); -- GitLab From bf4097ea5762bd90d836df9904594eb4822fefa7 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 21:36:21 +0800 Subject: [PATCH 0269/1471] drm/i915/gvt: Fix a memory leak in cmd_parser.c The pointer points to the original memory can never take the return value of krealloc(). Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 701a3c6f16696..9405a214760c8 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -2620,14 +2620,16 @@ static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload) gma_top = workload->rb_start + guest_rb_size; if (workload->rb_len > vgpu->reserve_ring_buffer_size[ring_id]) { - void *va = vgpu->reserve_ring_buffer_va[ring_id]; + void *va, *p; + /* realloc the new ring buffer if needed */ - vgpu->reserve_ring_buffer_va[ring_id] = - krealloc(va, workload->rb_len, GFP_KERNEL); - if (!vgpu->reserve_ring_buffer_va[ring_id]) { + va = vgpu->reserve_ring_buffer_va[ring_id]; + p = krealloc(va, workload->rb_len, GFP_KERNEL); + if (!p) { gvt_vgpu_err("fail to alloc reserve ring buffer\n"); return -ENOMEM; } + vgpu->reserve_ring_buffer_va[ring_id] = p; vgpu->reserve_ring_buffer_size[ring_id] = workload->rb_len; } -- GitLab From 8cf80a2e4b313d1aba7d10ce6ebfbb44a119c66c Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 21:46:06 +0800 Subject: [PATCH 0270/1471] drm/i915/gvt: Rename reserved ring buffer "reserved" means reserve something from somewhere. Actually they are buffers used by command scanner. Rename it to ring_scan_buffer. v2: - Remove the usage of an extra variable. (Zhenyu) Fixes: 0a53bc07f044 ("drm/i915/gvt: Separate cmd scan from request allocation") Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 16 ++++++++-------- drivers/gpu/drm/i915/gvt/execlist.c | 22 +++++++++++----------- drivers/gpu/drm/i915/gvt/gvt.h | 6 +++--- 3 files changed, 22 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 9405a214760c8..a2b970fb3c6fd 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -2619,21 +2619,21 @@ static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload) gma_tail = workload->rb_start + workload->rb_tail; gma_top = workload->rb_start + guest_rb_size; - if (workload->rb_len > vgpu->reserve_ring_buffer_size[ring_id]) { - void *va, *p; + if (workload->rb_len > vgpu->ring_scan_buffer_size[ring_id]) { + void *p; /* realloc the new ring buffer if needed */ - va = vgpu->reserve_ring_buffer_va[ring_id]; - p = krealloc(va, workload->rb_len, GFP_KERNEL); + p = krealloc(vgpu->ring_scan_buffer[ring_id], workload->rb_len, + GFP_KERNEL); if (!p) { - gvt_vgpu_err("fail to alloc reserve ring buffer\n"); + gvt_vgpu_err("fail to re-alloc ring scan buffer\n"); return -ENOMEM; } - vgpu->reserve_ring_buffer_va[ring_id] = p; - vgpu->reserve_ring_buffer_size[ring_id] = workload->rb_len; + vgpu->ring_scan_buffer[ring_id] = p; + vgpu->ring_scan_buffer_size[ring_id] = workload->rb_len; } - shadow_ring_buffer_va = vgpu->reserve_ring_buffer_va[ring_id]; + shadow_ring_buffer_va = vgpu->ring_scan_buffer[ring_id]; /* get shadow ring buffer va */ workload->shadow_ring_buffer_va = shadow_ring_buffer_va; diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index f118454d2eabb..5e4f35f9e290a 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -864,9 +864,9 @@ void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) clean_workloads(vgpu, ALL_ENGINES); for_each_engine(engine, vgpu->gvt->dev_priv, i) { - kfree(vgpu->reserve_ring_buffer_va[i]); - vgpu->reserve_ring_buffer_va[i] = NULL; - vgpu->reserve_ring_buffer_size[i] = 0; + kfree(vgpu->ring_scan_buffer[i]); + vgpu->ring_scan_buffer[i] = NULL; + vgpu->ring_scan_buffer_size[i] = 0; } } @@ -881,21 +881,21 @@ int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) /* each ring has a shadow ring buffer until vgpu destroyed */ for_each_engine(engine, vgpu->gvt->dev_priv, i) { - vgpu->reserve_ring_buffer_va[i] = + vgpu->ring_scan_buffer[i] = kmalloc(RESERVE_RING_BUFFER_SIZE, GFP_KERNEL); - if (!vgpu->reserve_ring_buffer_va[i]) { - gvt_vgpu_err("fail to alloc reserve ring buffer\n"); + if (!vgpu->ring_scan_buffer[i]) { + gvt_vgpu_err("fail to alloc ring scan buffer\n"); goto out; } - vgpu->reserve_ring_buffer_size[i] = RESERVE_RING_BUFFER_SIZE; + vgpu->ring_scan_buffer_size[i] = RESERVE_RING_BUFFER_SIZE; } return 0; out: for_each_engine(engine, vgpu->gvt->dev_priv, i) { - if (vgpu->reserve_ring_buffer_size[i]) { - kfree(vgpu->reserve_ring_buffer_va[i]); - vgpu->reserve_ring_buffer_va[i] = NULL; - vgpu->reserve_ring_buffer_size[i] = 0; + if (vgpu->ring_scan_buffer_size[i]) { + kfree(vgpu->ring_scan_buffer[i]); + vgpu->ring_scan_buffer[i] = NULL; + vgpu->ring_scan_buffer_size[i] = 0; } } return -ENOMEM; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 93ff530eee303..5b723fa00fcb5 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -172,9 +172,9 @@ struct intel_vgpu { struct intel_vgpu_opregion opregion; struct intel_vgpu_display display; struct intel_vgpu_submission submission; - /* 1/2K for each reserve ring buffer */ - void *reserve_ring_buffer_va[I915_NUM_ENGINES]; - int reserve_ring_buffer_size[I915_NUM_ENGINES]; + /* 1/2K for each engine */ + void *ring_scan_buffer[I915_NUM_ENGINES]; + int ring_scan_buffer_size[I915_NUM_ENGINES]; #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT) -- GitLab From 325eb94a33451db138e8da7393ad8e9a0e22dd18 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 21:58:11 +0800 Subject: [PATCH 0271/1471] drm/i915/gvt: Move ring scan buffers into intel_vgpu_submission Move ring scan buffers into intel_vgpu_submission since they belongs to a part of vGPU submission stuffs. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 11 ++++++----- drivers/gpu/drm/i915/gvt/execlist.c | 23 +++++++++++++---------- drivers/gpu/drm/i915/gvt/gvt.h | 7 +++---- 3 files changed, 22 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index a2b970fb3c6fd..5b033c950a31f 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -2604,6 +2604,7 @@ static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; unsigned long gma_head, gma_tail, gma_top, guest_rb_size; void *shadow_ring_buffer_va; int ring_id = workload->ring_id; @@ -2619,21 +2620,21 @@ static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload) gma_tail = workload->rb_start + workload->rb_tail; gma_top = workload->rb_start + guest_rb_size; - if (workload->rb_len > vgpu->ring_scan_buffer_size[ring_id]) { + if (workload->rb_len > s->ring_scan_buffer_size[ring_id]) { void *p; /* realloc the new ring buffer if needed */ - p = krealloc(vgpu->ring_scan_buffer[ring_id], workload->rb_len, + p = krealloc(s->ring_scan_buffer[ring_id], workload->rb_len, GFP_KERNEL); if (!p) { gvt_vgpu_err("fail to re-alloc ring scan buffer\n"); return -ENOMEM; } - vgpu->ring_scan_buffer[ring_id] = p; - vgpu->ring_scan_buffer_size[ring_id] = workload->rb_len; + s->ring_scan_buffer[ring_id] = p; + s->ring_scan_buffer_size[ring_id] = workload->rb_len; } - shadow_ring_buffer_va = vgpu->ring_scan_buffer[ring_id]; + shadow_ring_buffer_va = s->ring_scan_buffer[ring_id]; /* get shadow ring buffer va */ workload->shadow_ring_buffer_va = shadow_ring_buffer_va; diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 5e4f35f9e290a..7c7ab63a7bb8a 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -864,15 +864,18 @@ void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) clean_workloads(vgpu, ALL_ENGINES); for_each_engine(engine, vgpu->gvt->dev_priv, i) { - kfree(vgpu->ring_scan_buffer[i]); - vgpu->ring_scan_buffer[i] = NULL; - vgpu->ring_scan_buffer_size[i] = 0; + struct intel_vgpu_submission *s = &vgpu->submission; + + kfree(s->ring_scan_buffer[i]); + s->ring_scan_buffer[i] = NULL; + s->ring_scan_buffer_size[i] = 0; } } #define RESERVE_RING_BUFFER_SIZE ((1 * PAGE_SIZE)/8) int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) { + struct intel_vgpu_submission *s = &vgpu->submission; enum intel_engine_id i; struct intel_engine_cs *engine; @@ -881,21 +884,21 @@ int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) /* each ring has a shadow ring buffer until vgpu destroyed */ for_each_engine(engine, vgpu->gvt->dev_priv, i) { - vgpu->ring_scan_buffer[i] = + s->ring_scan_buffer[i] = kmalloc(RESERVE_RING_BUFFER_SIZE, GFP_KERNEL); - if (!vgpu->ring_scan_buffer[i]) { + if (!s->ring_scan_buffer[i]) { gvt_vgpu_err("fail to alloc ring scan buffer\n"); goto out; } - vgpu->ring_scan_buffer_size[i] = RESERVE_RING_BUFFER_SIZE; + s->ring_scan_buffer_size[i] = RESERVE_RING_BUFFER_SIZE; } return 0; out: for_each_engine(engine, vgpu->gvt->dev_priv, i) { - if (vgpu->ring_scan_buffer_size[i]) { - kfree(vgpu->ring_scan_buffer[i]); - vgpu->ring_scan_buffer[i] = NULL; - vgpu->ring_scan_buffer_size[i] = 0; + if (s->ring_scan_buffer_size[i]) { + kfree(s->ring_scan_buffer[i]); + s->ring_scan_buffer[i] = NULL; + s->ring_scan_buffer_size[i] = 0; } } return -ENOMEM; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 5b723fa00fcb5..49fe548700003 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -150,6 +150,9 @@ struct intel_vgpu_submission { struct i915_gem_context *shadow_ctx; DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES); DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); + /* 1/2K for each engine */ + void *ring_scan_buffer[I915_NUM_ENGINES]; + int ring_scan_buffer_size[I915_NUM_ENGINES]; }; struct intel_vgpu { @@ -172,10 +175,6 @@ struct intel_vgpu { struct intel_vgpu_opregion opregion; struct intel_vgpu_display display; struct intel_vgpu_submission submission; - /* 1/2K for each engine */ - void *ring_scan_buffer[I915_NUM_ENGINES]; - int ring_scan_buffer_size[I915_NUM_ENGINES]; - #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT) struct { -- GitLab From 8652a8aca614cc2576cf5d4c73d3b87e871ce004 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 10 Sep 2017 22:01:10 +0800 Subject: [PATCH 0272/1471] drm/i915/gvt: Do not allocate initial ring scan buffer Theoretically, the largest bulk of commands in the ring buffer of an engine might be the first submission, which usually contains a lot of commands to initialize the HW. After removing the initial allocation of the ring scan buffer and let krealloc() do everything we need, we still have a big chance to get the buffer of suitable size in the first submission. Tested on my SKL NUC. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 21 --------------------- drivers/gpu/drm/i915/gvt/gvt.h | 1 - 2 files changed, 22 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 7c7ab63a7bb8a..adf7668f8cc1c 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -872,36 +872,15 @@ void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) } } -#define RESERVE_RING_BUFFER_SIZE ((1 * PAGE_SIZE)/8) int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) { - struct intel_vgpu_submission *s = &vgpu->submission; enum intel_engine_id i; struct intel_engine_cs *engine; for_each_engine(engine, vgpu->gvt->dev_priv, i) init_vgpu_execlist(vgpu, i); - /* each ring has a shadow ring buffer until vgpu destroyed */ - for_each_engine(engine, vgpu->gvt->dev_priv, i) { - s->ring_scan_buffer[i] = - kmalloc(RESERVE_RING_BUFFER_SIZE, GFP_KERNEL); - if (!s->ring_scan_buffer[i]) { - gvt_vgpu_err("fail to alloc ring scan buffer\n"); - goto out; - } - s->ring_scan_buffer_size[i] = RESERVE_RING_BUFFER_SIZE; - } return 0; -out: - for_each_engine(engine, vgpu->gvt->dev_priv, i) { - if (s->ring_scan_buffer_size[i]) { - kfree(s->ring_scan_buffer[i]); - s->ring_scan_buffer[i] = NULL; - s->ring_scan_buffer_size[i] = 0; - } - } - return -ENOMEM; } void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu, diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 49fe548700003..7a770b1c99d6a 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -150,7 +150,6 @@ struct intel_vgpu_submission { struct i915_gem_context *shadow_ctx; DECLARE_BITMAP(shadow_ctx_desc_updated, I915_NUM_ENGINES); DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); - /* 1/2K for each engine */ void *ring_scan_buffer[I915_NUM_ENGINES]; int ring_scan_buffer_size[I915_NUM_ENGINES]; }; -- GitLab From 5c56883a9531cd89561fb9a11a33697f2847c82a Mon Sep 17 00:00:00 2001 From: fred gao Date: Wed, 20 Sep 2017 05:36:47 +0800 Subject: [PATCH 0273/1471] drm/i915/gvt: Change the return type during command scan Generally, there are 3 types of errors during command scan: a) some commands might be unknown with EBADRQC; b) some cmd access invalid address with EFAULT; c) some unexpected force nonpriv cmd with EPERM. later the healthy state can be judged through the return error. v2: - remove some internal i915 errors rating. (Zhenyu) v3: - the healthy state is judged through the internal defined return error. (Zhenyu) - force non priv cmd error can be ignored. (Kevin) v4: - reuse standard defined errno instead of recreate, e.g EBADRQC for unknown cmd, EFAULT for invalid address, EPERM for nonpriv. (Zhenyu) v5: - remove some irrelevant code for the patch. - fix typo of vgpu_is_vm_unhealthy. (Zhenyu) v6: - move the healthy check and failsafe code into another patch. (Zhenyu) v7: - polish title and commit message. (Zhenyu) Signed-off-by: fred gao Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 78 ++++++++++++++++++--------- drivers/gpu/drm/i915/gvt/scheduler.c | 2 +- 2 files changed, 54 insertions(+), 26 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 5b033c950a31f..885cccf708ac7 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -825,7 +825,7 @@ static int force_nonpriv_reg_handler(struct parser_exec_state *s, if (!intel_gvt_in_force_nonpriv_whitelist(gvt, data)) { gvt_err("Unexpected forcenonpriv 0x%x LRI write, value=0x%x\n", offset, data); - return -EINVAL; + return -EPERM; } return 0; } @@ -839,7 +839,7 @@ static int cmd_reg_handler(struct parser_exec_state *s, if (offset + 4 > gvt->device_info.mmio_size) { gvt_vgpu_err("%s access to (%x) outside of MMIO range\n", cmd, offset); - return -EINVAL; + return -EFAULT; } if (!intel_gvt_mmio_is_cmd_access(gvt, offset)) { @@ -854,8 +854,8 @@ static int cmd_reg_handler(struct parser_exec_state *s, } if (is_force_nonpriv_mmio(offset) && - force_nonpriv_reg_handler(s, offset, index)) - return -EINVAL; + force_nonpriv_reg_handler(s, offset, index)) + return -EPERM; if (offset == i915_mmio_reg_offset(DERRMR) || offset == i915_mmio_reg_offset(FORCEWAKE_MT)) { @@ -894,11 +894,14 @@ static int cmd_handler_lri(struct parser_exec_state *s) i915_mmio_reg_offset(DERRMR)) ret |= 0; else - ret |= (cmd_reg_inhibit(s, i)) ? -EINVAL : 0; + ret |= (cmd_reg_inhibit(s, i)) ? + -EBADRQC : 0; } if (ret) break; ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lri"); + if (ret) + break; } return ret; } @@ -912,11 +915,15 @@ static int cmd_handler_lrr(struct parser_exec_state *s) if (IS_BROADWELL(s->vgpu->gvt->dev_priv)) ret |= ((cmd_reg_inhibit(s, i) || (cmd_reg_inhibit(s, i + 1)))) ? - -EINVAL : 0; + -EBADRQC : 0; if (ret) break; ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrr-src"); + if (ret) + break; ret |= cmd_reg_handler(s, cmd_reg(s, i + 1), i, "lrr-dst"); + if (ret) + break; } return ret; } @@ -934,15 +941,19 @@ static int cmd_handler_lrm(struct parser_exec_state *s) for (i = 1; i < cmd_len;) { if (IS_BROADWELL(gvt->dev_priv)) - ret |= (cmd_reg_inhibit(s, i)) ? -EINVAL : 0; + ret |= (cmd_reg_inhibit(s, i)) ? -EBADRQC : 0; if (ret) break; ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrm"); + if (ret) + break; if (cmd_val(s, 0) & (1 << 22)) { gma = cmd_gma(s, i + 1); if (gmadr_bytes == 8) gma |= (cmd_gma_hi(s, i + 2)) << 32; ret |= cmd_address_audit(s, gma, sizeof(u32), false); + if (ret) + break; } i += gmadr_dw_number(s) + 1; } @@ -958,11 +969,15 @@ static int cmd_handler_srm(struct parser_exec_state *s) for (i = 1; i < cmd_len;) { ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "srm"); + if (ret) + break; if (cmd_val(s, 0) & (1 << 22)) { gma = cmd_gma(s, i + 1); if (gmadr_bytes == 8) gma |= (cmd_gma_hi(s, i + 2)) << 32; ret |= cmd_address_audit(s, gma, sizeof(u32), false); + if (ret) + break; } i += gmadr_dw_number(s) + 1; } @@ -1116,7 +1131,7 @@ static int gen8_decode_mi_display_flip(struct parser_exec_state *s, v = (dword0 & GENMASK(21, 19)) >> 19; if (WARN_ON(v >= ARRAY_SIZE(gen8_plane_code))) - return -EINVAL; + return -EBADRQC; info->pipe = gen8_plane_code[v].pipe; info->plane = gen8_plane_code[v].plane; @@ -1136,7 +1151,7 @@ static int gen8_decode_mi_display_flip(struct parser_exec_state *s, info->surf_reg = SPRSURF(info->pipe); } else { WARN_ON(1); - return -EINVAL; + return -EBADRQC; } return 0; } @@ -1185,7 +1200,7 @@ static int skl_decode_mi_display_flip(struct parser_exec_state *s, default: gvt_vgpu_err("unknown plane code %d\n", plane); - return -EINVAL; + return -EBADRQC; } info->stride_val = (dword1 & GENMASK(15, 6)) >> 6; @@ -1348,10 +1363,13 @@ static unsigned long get_gma_bb_from_cmd(struct parser_exec_state *s, int index) { unsigned long addr; unsigned long gma_high, gma_low; - int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd; + struct intel_vgpu *vgpu = s->vgpu; + int gmadr_bytes = vgpu->gvt->device_info.gmadr_bytes_in_cmd; - if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) + if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) { + gvt_vgpu_err("invalid gma bytes %d\n", gmadr_bytes); return INTEL_GVT_INVALID_ADDR; + } gma_low = cmd_val(s, index) & BATCH_BUFFER_ADDR_MASK; if (gmadr_bytes == 4) { @@ -1374,16 +1392,16 @@ static inline int cmd_address_audit(struct parser_exec_state *s, if (op_size > max_surface_size) { gvt_vgpu_err("command address audit fail name %s\n", s->info->name); - return -EINVAL; + return -EFAULT; } if (index_mode) { if (guest_gma >= GTT_PAGE_SIZE / sizeof(u64)) { - ret = -EINVAL; + ret = -EFAULT; goto err; } } else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) { - ret = -EINVAL; + ret = -EFAULT; goto err; } @@ -1439,7 +1457,7 @@ static inline int unexpected_cmd(struct parser_exec_state *s) gvt_vgpu_err("Unexpected %s in command buffer!\n", s->info->name); - return -EINVAL; + return -EBADRQC; } static int cmd_handler_mi_semaphore_wait(struct parser_exec_state *s) @@ -1588,22 +1606,26 @@ static int find_bb_size(struct parser_exec_state *s) /* get the start gm address of the batch buffer */ gma = get_gma_bb_from_cmd(s, 1); + if (gma == INTEL_GVT_INVALID_ADDR) + return -EFAULT; + cmd = cmd_val(s, 0); info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id); if (info == NULL) { gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n", cmd, get_opcode(cmd, s->ring_id)); - return -EINVAL; + return -EBADRQC; } do { - copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm, - gma, gma + 4, &cmd); + if (copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm, + gma, gma + 4, &cmd) < 0) + return -EFAULT; info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id); if (info == NULL) { gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n", cmd, get_opcode(cmd, s->ring_id)); - return -EINVAL; + return -EBADRQC; } if (info->opcode == OP_MI_BATCH_BUFFER_END) { @@ -1634,11 +1656,13 @@ static int perform_bb_shadow(struct parser_exec_state *s) /* get the start gm address of the batch buffer */ gma = get_gma_bb_from_cmd(s, 1); + if (gma == INTEL_GVT_INVALID_ADDR) + return -EFAULT; /* get the size of the batch buffer */ bb_size = find_bb_size(s); if (bb_size < 0) - return -EINVAL; + return bb_size; /* allocate shadow batch buffer */ entry_obj = kmalloc(sizeof(*entry_obj), GFP_KERNEL); @@ -1710,13 +1734,13 @@ static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s) if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) { gvt_vgpu_err("Found MI_BATCH_BUFFER_START in 2nd level BB\n"); - return -EINVAL; + return -EFAULT; } second_level = BATCH_BUFFER_2ND_LEVEL_BIT(cmd_val(s, 0)) == 1; if (second_level && (s->buf_type != BATCH_BUFFER_INSTRUCTION)) { gvt_vgpu_err("Jumping to 2nd level BB from RB is not allowed\n"); - return -EINVAL; + return -EFAULT; } s->saved_buf_addr_type = s->buf_addr_type; @@ -2430,7 +2454,7 @@ static int cmd_parser_exec(struct parser_exec_state *s) if (info == NULL) { gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n", cmd, get_opcode(cmd, s->ring_id)); - return -EINVAL; + return -EBADRQC; } s->info = info; @@ -2465,6 +2489,10 @@ static inline bool gma_out_of_range(unsigned long gma, return (gma > gma_tail) && (gma < gma_head); } +/* Keep the consistent return type, e.g EBADRQC for unknown + * cmd, EFAULT for invalid address, EPERM for nonpriv. later + * works as the input of VM healthy status. + */ static int command_scan(struct parser_exec_state *s, unsigned long rb_head, unsigned long rb_tail, unsigned long rb_start, unsigned long rb_len) @@ -2487,7 +2515,7 @@ static int command_scan(struct parser_exec_state *s, s->ip_gma, rb_start, gma_bottom); parser_exec_state_dump(s); - return -EINVAL; + return -EFAULT; } if (gma_out_of_range(s->ip_gma, gma_head, gma_tail)) { gvt_vgpu_err("ip_gma %lx out of range." diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 7cb1cf4223ed4..0771b715f825c 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -84,7 +84,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) GTT_PAGE_SHIFT)); if (context_gpa == INTEL_GVT_INVALID_ADDR) { gvt_vgpu_err("Invalid guest context descriptor\n"); - return -EINVAL; + return -EFAULT; } page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i); -- GitLab From e011c6ce2b4fc7c577ade41485d74431a4e6ea1a Mon Sep 17 00:00:00 2001 From: fred gao Date: Tue, 19 Sep 2017 15:11:28 +0800 Subject: [PATCH 0274/1471] drm/i915/gvt: Add VM healthy check for workload_thread When a scan error occurs in dispatch_workload, this patch is to check the healthy state and free all the queued workloads before the failsafe mode is entered. Signed-off-by: fred gao Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gvt.h | 6 ++++++ drivers/gpu/drm/i915/gvt/handlers.c | 4 +++- drivers/gpu/drm/i915/gvt/scheduler.c | 7 +++++++ 3 files changed, 16 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 7a770b1c99d6a..be1bb6dbcbd27 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -193,6 +193,10 @@ struct intel_vgpu { #endif }; +/* validating GM healthy status*/ +#define vgpu_is_vm_unhealthy(ret_val) \ + (((ret_val) == -EBADRQC) || ((ret_val) == -EFAULT)) + struct intel_gvt_gm { unsigned long vgpu_allocated_low_gm_size; unsigned long vgpu_allocated_high_gm_size; @@ -497,6 +501,7 @@ int intel_vgpu_emulate_opregion_request(struct intel_vgpu *vgpu, u32 swsci); void populate_pvinfo_page(struct intel_vgpu *vgpu); int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload); +void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason); struct intel_gvt_ops { int (*emulate_cfg_read)(struct intel_vgpu *, unsigned int, void *, @@ -519,6 +524,7 @@ struct intel_gvt_ops { enum { GVT_FAILSAFE_UNSUPPORTED_GUEST, GVT_FAILSAFE_INSUFFICIENT_RESOURCE, + GVT_FAILSAFE_GUEST_ERR, }; static inline void mmio_hw_access_pre(struct drm_i915_private *dev_priv) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index c78e45058219d..acc1cf4fa6f53 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -157,7 +157,7 @@ static int render_mmio_to_ring_id(struct intel_gvt *gvt, unsigned int reg) (num * 8 + i915_mmio_reg_offset(FENCE_REG_GEN6_LO(0))) -static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason) +void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason) { switch (reason) { case GVT_FAILSAFE_UNSUPPORTED_GUEST: @@ -165,6 +165,8 @@ static void enter_failsafe_mode(struct intel_vgpu *vgpu, int reason) break; case GVT_FAILSAFE_INSUFFICIENT_RESOURCE: pr_err("Graphics resource is not enough for the guest\n"); + case GVT_FAILSAFE_GUEST_ERR: + pr_err("GVT Internal error for the guest\n"); default: break; } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 0771b715f825c..02af140233836 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -634,6 +634,13 @@ static int workload_thread(void *priv) FORCEWAKE_ALL); intel_runtime_pm_put(gvt->dev_priv); + if (ret && (vgpu_is_vm_unhealthy(ret))) { + mutex_lock(&gvt->lock); + intel_vgpu_clean_execlist(vgpu); + mutex_unlock(&gvt->lock); + enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); + } + } return 0; } -- GitLab From c9214008b52b20ba2a88dad6ba92f1a19c533db6 Mon Sep 17 00:00:00 2001 From: fred gao Date: Tue, 19 Sep 2017 15:11:29 +0800 Subject: [PATCH 0275/1471] drm/i915/gvt: Add VM healthy check for submit_context When a scan error occurs in submit_context, this patch is to decrease the mm ref count and free the workload struct before the workload is abandoned. v2: - submit_context related code should be combined together. (Zhenyu) v3: - free all the unsubmitted workloads. (Zhenyu) v4: - refine the clean path. (Zhenyu) v5: - polish the title. (Zhenyu) Signed-off-by: fred gao Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index adf7668f8cc1c..1347f61fd7091 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -360,7 +360,6 @@ static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist, static void free_workload(struct intel_vgpu_workload *workload) { - intel_vgpu_unpin_mm(workload->shadow_mm); intel_gvt_mm_unreference(workload->shadow_mm); kmem_cache_free(workload->vgpu->submission.workloads, workload); } @@ -540,7 +539,7 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) struct intel_vgpu_workload *next_workload; struct list_head *next = workload_q_head(vgpu, ring_id)->next; bool lite_restore = false; - int ret; + int ret = 0; gvt_dbg_el("complete workload %p status %d\n", workload, workload->status); @@ -581,17 +580,12 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) if (lite_restore) { gvt_dbg_el("next context == current - no schedule-out\n"); - free_workload(workload); - return 0; + goto out; } ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc); - if (ret) - goto err; out: - free_workload(workload); - return 0; -err: + intel_vgpu_unpin_mm(workload->shadow_mm); free_workload(workload); return ret; } @@ -762,13 +756,22 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, if (list_empty(workload_q_head(vgpu, ring_id))) { intel_runtime_pm_get(dev_priv); mutex_lock(&dev_priv->drm.struct_mutex); - intel_gvt_scan_and_shadow_workload(workload); + ret = intel_gvt_scan_and_shadow_workload(workload); mutex_unlock(&dev_priv->drm.struct_mutex); intel_runtime_pm_put(dev_priv); } - queue_workload(workload); - return 0; + if (ret == 0) + queue_workload(workload); + else { + free_workload(workload); + if (vgpu_is_vm_unhealthy(ret)) { + intel_vgpu_clean_execlist(vgpu); + enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); + } + } + return ret; + } int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) -- GitLab From 24f8a29af4afe7c53e08f4afa0c3fa9eb3791b89 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Tue, 19 Sep 2017 16:55:34 +0100 Subject: [PATCH 0276/1471] drm/i915/gvt: ensure -ve return value is handled correctly An earlier fix changed the return type from find_bb_size however the integer return is being assigned to a unsigned int so the -ve error check will never be detected. Make bb_size an int to fix this. Detected by CoverityScan CID#1456886 ("Unsigned compared against 0") Fixes: 1e3197d6ad73 ("drm/i915/gvt: Refine error handling for perform_bb_shadow") Signed-off-by: Colin Ian King Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 885cccf708ac7..cd2681a23fcd5 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -1650,7 +1650,7 @@ static int perform_bb_shadow(struct parser_exec_state *s) struct intel_shadow_bb_entry *entry_obj; struct intel_vgpu *vgpu = s->vgpu; unsigned long gma = 0; - uint32_t bb_size; + int bb_size; void *dst = NULL; int ret = 0; -- GitLab From a34e8def4d9152123582a37a3660c2092e37ceed Mon Sep 17 00:00:00 2001 From: Shuo Liu Date: Thu, 21 Sep 2017 08:02:31 +0800 Subject: [PATCH 0277/1471] drm/i915/gvt: Use dyndbg for gvt debug info It's better enable/disable and classify gvt debug info dynamically. This patch change it to dyndbg so can be dynamically enable/disable each item. All gvt log can be enabled by, $ echo 'file *gvt* +p' > /sys/kernel/debug/dynamic_debug/control Signed-off-by: Shuo Liu Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/debug.h | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/debug.h b/drivers/gpu/drm/i915/gvt/debug.h index b0cff4dc26847..d98d5a0b928c7 100644 --- a/drivers/gpu/drm/i915/gvt/debug.h +++ b/drivers/gpu/drm/i915/gvt/debug.h @@ -25,41 +25,41 @@ #define __GVT_DEBUG_H__ #define gvt_err(fmt, args...) \ - DRM_ERROR("gvt: "fmt, ##args) + pr_err("gvt: "fmt, ##args) #define gvt_vgpu_err(fmt, args...) \ do { \ if (IS_ERR_OR_NULL(vgpu)) \ - DRM_DEBUG_DRIVER("gvt: "fmt, ##args); \ + pr_debug("gvt: "fmt, ##args); \ else \ - DRM_DEBUG_DRIVER("gvt: vgpu %d: "fmt, vgpu->id, ##args);\ + pr_debug("gvt: vgpu %d: "fmt, vgpu->id, ##args);\ } while (0) #define gvt_dbg_core(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: core: "fmt, ##args) + pr_debug("gvt: core: "fmt, ##args) #define gvt_dbg_irq(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: irq: "fmt, ##args) + pr_debug("gvt: irq: "fmt, ##args) #define gvt_dbg_mm(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: mm: "fmt, ##args) + pr_debug("gvt: mm: "fmt, ##args) #define gvt_dbg_mmio(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: mmio: "fmt, ##args) + pr_debug("gvt: mmio: "fmt, ##args) #define gvt_dbg_dpy(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: dpy: "fmt, ##args) + pr_debug("gvt: dpy: "fmt, ##args) #define gvt_dbg_el(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: el: "fmt, ##args) + pr_debug("gvt: el: "fmt, ##args) #define gvt_dbg_sched(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: sched: "fmt, ##args) + pr_debug("gvt: sched: "fmt, ##args) #define gvt_dbg_render(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: render: "fmt, ##args) + pr_debug("gvt: render: "fmt, ##args) #define gvt_dbg_cmd(fmt, args...) \ - DRM_DEBUG_DRIVER("gvt: cmd: "fmt, ##args) + pr_debug("gvt: cmd: "fmt, ##args) #endif -- GitLab From 21527a8dafc40fc499ae57492c1c5d0098cbcf08 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 12 Sep 2017 21:42:09 +0800 Subject: [PATCH 0278/1471] drm/i915/gvt: Factor out vGPU workload creation/destroy Factor out vGPU workload creation/destroy functions since they are not specific to execlist emulation. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 27 ++++----------- drivers/gpu/drm/i915/gvt/scheduler.c | 51 ++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/scheduler.h | 5 +++ 3 files changed, 62 insertions(+), 21 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 1347f61fd7091..a3bdb037a9feb 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -358,12 +358,6 @@ static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist, return 0; } -static void free_workload(struct intel_vgpu_workload *workload) -{ - intel_gvt_mm_unreference(workload->shadow_mm); - kmem_cache_free(workload->vgpu->submission.workloads, workload); -} - #define get_desc_from_elsp_dwords(ed, i) \ ((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2])) @@ -586,7 +580,7 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc); out: intel_vgpu_unpin_mm(workload->shadow_mm); - free_workload(workload); + intel_vgpu_destroy_workload(workload); return ret; } @@ -687,10 +681,6 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, gvt_dbg_el("ring id %d begin a new workload\n", ring_id); - workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL); - if (!workload) - return -ENOMEM; - /* record some ring buffer register values for scan and shadow */ intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + RING_CTX_OFF(rb_start.val), &start, 4); @@ -699,13 +689,10 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4); - INIT_LIST_HEAD(&workload->list); - INIT_LIST_HEAD(&workload->shadow_bb); - - init_waitqueue_head(&workload->shadow_ctx_status_wq); - atomic_set(&workload->shadow_ctx_active, 0); + workload = intel_vgpu_create_workload(vgpu); + if (IS_ERR(workload)) + return PTR_ERR(workload); - workload->vgpu = vgpu; workload->ring_id = ring_id; workload->ctx_desc = *desc; workload->ring_context_gpa = ring_context_gpa; @@ -715,9 +702,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, workload->rb_ctl = ctl; workload->prepare = prepare_execlist_workload; workload->complete = complete_execlist_workload; - workload->status = -EINPROGRESS; workload->emulate_schedule_in = emulate_schedule_in; - workload->shadowed = false; if (ring_id == RCS) { intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + @@ -764,7 +749,7 @@ static int submit_context(struct intel_vgpu *vgpu, int ring_id, if (ret == 0) queue_workload(workload); else { - free_workload(workload); + intel_vgpu_destroy_workload(workload); if (vgpu_is_vm_unhealthy(ret)) { intel_vgpu_clean_execlist(vgpu); enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); @@ -853,7 +838,7 @@ static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) list_for_each_entry_safe(pos, n, &s->workload_q_head[engine->id], list) { list_del_init(&pos->list); - free_workload(pos); + intel_vgpu_destroy_workload(pos); } clear_bit(engine->id, s->shadow_ctx_desc_updated); } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 02af140233836..10ccb05d0e8d5 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -781,3 +781,54 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) i915_gem_context_put(s->shadow_ctx); return ret; } + +/** + * intel_vgpu_destroy_workload - destroy a vGPU workload + * @vgpu: a vGPU + * + * This function is called when destroy a vGPU workload. + * + */ +void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload) +{ + struct intel_vgpu_submission *s = &workload->vgpu->submission; + + if (workload->shadow_mm) + intel_gvt_mm_unreference(workload->shadow_mm); + + kmem_cache_free(s->workloads, workload); +} + +/** + * intel_vgpu_create_workload - create a vGPU workload + * @vgpu: a vGPU + * + * This function is called when creating a vGPU workload. + * + * Returns: + * struct intel_vgpu_workload * on success, negative error code in + * pointer if failed. + * + */ +struct intel_vgpu_workload * +intel_vgpu_create_workload(struct intel_vgpu *vgpu) +{ + struct intel_vgpu_submission *s = &vgpu->submission; + struct intel_vgpu_workload *workload; + + workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL); + if (!workload) + return ERR_PTR(-ENOMEM); + + INIT_LIST_HEAD(&workload->list); + INIT_LIST_HEAD(&workload->shadow_bb); + + init_waitqueue_head(&workload->shadow_ctx_status_wq); + atomic_set(&workload->shadow_ctx_active, 0); + + workload->status = -EINPROGRESS; + workload->shadowed = false; + workload->vgpu = vgpu; + + return workload; +} diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 3ca0087f10b20..3fe4702cda06b 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -141,5 +141,10 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu); void intel_vgpu_clean_submission(struct intel_vgpu *vgpu); +struct intel_vgpu_workload * +intel_vgpu_create_workload(struct intel_vgpu *vgpu); + +void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload); + void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx); #endif -- GitLab From 497aa3f5e3bdb6bea5994f7075e2f2df2377d70e Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 12 Sep 2017 21:51:10 +0800 Subject: [PATCH 0279/1471] drm/i915/gvt: Factor out prepare_workload() Factor out prepare_workload() for the following re-factor. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/scheduler.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 10ccb05d0e8d5..3d1435f55c7bf 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -325,6 +325,16 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) return ret; } +static int prepare_workload(struct intel_vgpu_workload *workload) +{ + int ret = 0; + + if (workload->prepare) + ret = workload->prepare(workload); + + return ret; +} + static int dispatch_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; @@ -344,12 +354,10 @@ static int dispatch_workload(struct intel_vgpu_workload *workload) if (ret) goto out; - if (workload->prepare) { - ret = workload->prepare(workload); - if (ret) { - engine->context_unpin(engine, shadow_ctx); - goto out; - } + ret = prepare_workload(workload); + if (ret) { + engine->context_unpin(engine, shadow_ctx); + goto out; } out: -- GitLab From d8235b5e55845de19983cec38af245cc200b81e2 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 12 Sep 2017 22:06:39 +0800 Subject: [PATCH 0280/1471] drm/i915/gvt: Move common workload preparation into prepare_workload() Move common workload preparation into prepare_workload() in scheduler.c, as they are not specific to execlist emulation. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 154 +-------------------------- drivers/gpu/drm/i915/gvt/scheduler.c | 152 +++++++++++++++++++++++++- 2 files changed, 155 insertions(+), 151 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index a3bdb037a9feb..45cb342500b79 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -361,110 +361,6 @@ static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist, #define get_desc_from_elsp_dwords(ed, i) \ ((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2])) -static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload) -{ - const int gmadr_bytes = workload->vgpu->gvt->device_info.gmadr_bytes_in_cmd; - struct intel_shadow_bb_entry *entry_obj; - - /* pin the gem object to ggtt */ - list_for_each_entry(entry_obj, &workload->shadow_bb, list) { - struct i915_vma *vma; - - vma = i915_gem_object_ggtt_pin(entry_obj->obj, NULL, 0, 4, 0); - if (IS_ERR(vma)) { - return PTR_ERR(vma); - } - - /* FIXME: we are not tracking our pinned VMA leaving it - * up to the core to fix up the stray pin_count upon - * free. - */ - - /* update the relocate gma with shadow batch buffer*/ - entry_obj->bb_start_cmd_va[1] = i915_ggtt_offset(vma); - if (gmadr_bytes == 8) - entry_obj->bb_start_cmd_va[2] = 0; - } - return 0; -} - -static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx) -{ - struct intel_vgpu_workload *workload = container_of(wa_ctx, - struct intel_vgpu_workload, - wa_ctx); - int ring_id = workload->ring_id; - struct intel_vgpu_submission *s = &workload->vgpu->submission; - struct i915_gem_context *shadow_ctx = s->shadow_ctx; - struct drm_i915_gem_object *ctx_obj = - shadow_ctx->engine[ring_id].state->obj; - struct execlist_ring_context *shadow_ring_context; - struct page *page; - - page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN); - shadow_ring_context = kmap_atomic(page); - - shadow_ring_context->bb_per_ctx_ptr.val = - (shadow_ring_context->bb_per_ctx_ptr.val & - (~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma; - shadow_ring_context->rcs_indirect_ctx.val = - (shadow_ring_context->rcs_indirect_ctx.val & - (~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma; - - kunmap_atomic(shadow_ring_context); - return 0; -} - -static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) -{ - struct i915_vma *vma; - unsigned char *per_ctx_va = - (unsigned char *)wa_ctx->indirect_ctx.shadow_va + - wa_ctx->indirect_ctx.size; - - if (wa_ctx->indirect_ctx.size == 0) - return 0; - - vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL, - 0, CACHELINE_BYTES, 0); - if (IS_ERR(vma)) { - return PTR_ERR(vma); - } - - /* FIXME: we are not tracking our pinned VMA leaving it - * up to the core to fix up the stray pin_count upon - * free. - */ - - wa_ctx->indirect_ctx.shadow_gma = i915_ggtt_offset(vma); - - wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1); - memset(per_ctx_va, 0, CACHELINE_BYTES); - - update_wa_ctx_2_shadow_ctx(wa_ctx); - return 0; -} - -static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) -{ - /* release all the shadow batch buffer */ - if (!list_empty(&workload->shadow_bb)) { - struct intel_shadow_bb_entry *entry_obj = - list_first_entry(&workload->shadow_bb, - struct intel_shadow_bb_entry, - list); - struct intel_shadow_bb_entry *temp; - - list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb, - list) { - i915_gem_object_unpin_map(entry_obj->obj); - i915_gem_object_put(entry_obj->obj); - list_del(&entry_obj->list); - kfree(entry_obj); - } - } -} - static int prepare_execlist_workload(struct intel_vgpu_workload *workload) { struct intel_vgpu *vgpu = workload->vgpu; @@ -473,36 +369,6 @@ static int prepare_execlist_workload(struct intel_vgpu_workload *workload) int ring_id = workload->ring_id; int ret; - ret = intel_vgpu_pin_mm(workload->shadow_mm); - if (ret) { - gvt_vgpu_err("fail to vgpu pin mm\n"); - goto out; - } - - ret = intel_vgpu_sync_oos_pages(workload->vgpu); - if (ret) { - gvt_vgpu_err("fail to vgpu sync oos pages\n"); - goto err_unpin_mm; - } - - ret = intel_vgpu_flush_post_shadow(workload->vgpu); - if (ret) { - gvt_vgpu_err("fail to flush post shadow\n"); - goto err_unpin_mm; - } - - ret = prepare_shadow_batch_buffer(workload); - if (ret) { - gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n"); - goto err_unpin_mm; - } - - ret = prepare_shadow_wa_ctx(&workload->wa_ctx); - if (ret) { - gvt_vgpu_err("fail to prepare_shadow_wa_ctx\n"); - goto err_shadow_batch; - } - if (!workload->emulate_schedule_in) return 0; @@ -510,18 +376,11 @@ static int prepare_execlist_workload(struct intel_vgpu_workload *workload) ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1); ret = emulate_execlist_schedule_in(&s->execlist[ring_id], ctx); - if (!ret) - goto out; - else + if (ret) { gvt_vgpu_err("fail to emulate execlist schedule in\n"); - - release_shadow_wa_ctx(&workload->wa_ctx); -err_shadow_batch: - release_shadow_batch_buffer(workload); -err_unpin_mm: - intel_vgpu_unpin_mm(workload->shadow_mm); -out: - return ret; + return ret; + } + return 0; } static int complete_execlist_workload(struct intel_vgpu_workload *workload) @@ -538,11 +397,6 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) gvt_dbg_el("complete workload %p status %d\n", workload, workload->status); - if (!workload->status) { - release_shadow_batch_buffer(workload); - release_shadow_wa_ctx(&workload->wa_ctx); - } - if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) { /* if workload->status is not successful means HW GPU * has occurred GPU hang or something wrong with i915/GVT, diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 3d1435f55c7bf..a7a67cc65a071 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -325,13 +325,157 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) return ret; } +static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload) +{ + struct intel_gvt *gvt = workload->vgpu->gvt; + const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd; + struct intel_shadow_bb_entry *entry_obj; + + /* pin the gem object to ggtt */ + list_for_each_entry(entry_obj, &workload->shadow_bb, list) { + struct i915_vma *vma; + + vma = i915_gem_object_ggtt_pin(entry_obj->obj, NULL, 0, 4, 0); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + /* FIXME: we are not tracking our pinned VMA leaving it + * up to the core to fix up the stray pin_count upon + * free. + */ + + /* update the relocate gma with shadow batch buffer*/ + entry_obj->bb_start_cmd_va[1] = i915_ggtt_offset(vma); + if (gmadr_bytes == 8) + entry_obj->bb_start_cmd_va[2] = 0; + } + return 0; +} + +static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx) +{ + struct intel_vgpu_workload *workload = container_of(wa_ctx, + struct intel_vgpu_workload, + wa_ctx); + int ring_id = workload->ring_id; + struct intel_vgpu_submission *s = &workload->vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; + struct drm_i915_gem_object *ctx_obj = + shadow_ctx->engine[ring_id].state->obj; + struct execlist_ring_context *shadow_ring_context; + struct page *page; + + page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN); + shadow_ring_context = kmap_atomic(page); + + shadow_ring_context->bb_per_ctx_ptr.val = + (shadow_ring_context->bb_per_ctx_ptr.val & + (~PER_CTX_ADDR_MASK)) | wa_ctx->per_ctx.shadow_gma; + shadow_ring_context->rcs_indirect_ctx.val = + (shadow_ring_context->rcs_indirect_ctx.val & + (~INDIRECT_CTX_ADDR_MASK)) | wa_ctx->indirect_ctx.shadow_gma; + + kunmap_atomic(shadow_ring_context); + return 0; +} + +static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) +{ + struct i915_vma *vma; + unsigned char *per_ctx_va = + (unsigned char *)wa_ctx->indirect_ctx.shadow_va + + wa_ctx->indirect_ctx.size; + + if (wa_ctx->indirect_ctx.size == 0) + return 0; + + vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL, + 0, CACHELINE_BYTES, 0); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + /* FIXME: we are not tracking our pinned VMA leaving it + * up to the core to fix up the stray pin_count upon + * free. + */ + + wa_ctx->indirect_ctx.shadow_gma = i915_ggtt_offset(vma); + + wa_ctx->per_ctx.shadow_gma = *((unsigned int *)per_ctx_va + 1); + memset(per_ctx_va, 0, CACHELINE_BYTES); + + update_wa_ctx_2_shadow_ctx(wa_ctx); + return 0; +} + +static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) +{ + /* release all the shadow batch buffer */ + if (!list_empty(&workload->shadow_bb)) { + struct intel_shadow_bb_entry *entry_obj = + list_first_entry(&workload->shadow_bb, + struct intel_shadow_bb_entry, + list); + struct intel_shadow_bb_entry *temp; + + list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb, + list) { + i915_gem_object_unpin_map(entry_obj->obj); + i915_gem_object_put(entry_obj->obj); + list_del(&entry_obj->list); + kfree(entry_obj); + } + } +} + static int prepare_workload(struct intel_vgpu_workload *workload) { + struct intel_vgpu *vgpu = workload->vgpu; int ret = 0; - if (workload->prepare) + ret = intel_vgpu_pin_mm(workload->shadow_mm); + if (ret) { + gvt_vgpu_err("fail to vgpu pin mm\n"); + return ret; + } + + ret = intel_vgpu_sync_oos_pages(workload->vgpu); + if (ret) { + gvt_vgpu_err("fail to vgpu sync oos pages\n"); + goto err_unpin_mm; + } + + ret = intel_vgpu_flush_post_shadow(workload->vgpu); + if (ret) { + gvt_vgpu_err("fail to flush post shadow\n"); + goto err_unpin_mm; + } + + ret = prepare_shadow_batch_buffer(workload); + if (ret) { + gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n"); + goto err_unpin_mm; + } + + ret = prepare_shadow_wa_ctx(&workload->wa_ctx); + if (ret) { + gvt_vgpu_err("fail to prepare_shadow_wa_ctx\n"); + goto err_shadow_batch; + } + + if (workload->prepare) { ret = workload->prepare(workload); + if (ret) + goto err_shadow_wa_ctx; + } + return 0; +err_shadow_wa_ctx: + release_shadow_wa_ctx(&workload->wa_ctx); +err_shadow_batch: + release_shadow_batch_buffer(workload); +err_unpin_mm: + intel_vgpu_unpin_mm(workload->shadow_mm); return ret; } @@ -557,6 +701,12 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) scheduler->current_workload[ring_id] = NULL; list_del_init(&workload->list); + + if (!workload->status) { + release_shadow_batch_buffer(workload); + release_shadow_wa_ctx(&workload->wa_ctx); + } + workload->complete(workload); atomic_dec(&s->running_workload_num); -- GitLab From 6d76303553bab75ffc53993c56aad06251d8de60 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 12 Sep 2017 22:33:12 +0800 Subject: [PATCH 0281/1471] drm/i915/gvt: Move common vGPU workload creation into scheduler.c Move common vGPU workload creation functions into scheduler.c since they are not specific to execlist emulation. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 156 +-------------------- drivers/gpu/drm/i915/gvt/scheduler.c | 196 ++++++++++++++++++++++++--- drivers/gpu/drm/i915/gvt/scheduler.h | 3 +- 3 files changed, 183 insertions(+), 172 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 45cb342500b79..c753cb5bfb589 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -438,179 +438,29 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) return ret; } -#define RING_CTX_OFF(x) \ - offsetof(struct execlist_ring_context, x) - -static void read_guest_pdps(struct intel_vgpu *vgpu, - u64 ring_context_gpa, u32 pdp[8]) -{ - u64 gpa; - int i; - - gpa = ring_context_gpa + RING_CTX_OFF(pdp3_UDW.val); - - for (i = 0; i < 8; i++) - intel_gvt_hypervisor_read_gpa(vgpu, - gpa + i * 8, &pdp[7 - i], 4); -} - -static int prepare_mm(struct intel_vgpu_workload *workload) -{ - struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc; - struct intel_vgpu_mm *mm; - struct intel_vgpu *vgpu = workload->vgpu; - int page_table_level; - u32 pdp[8]; - - if (desc->addressing_mode == 1) { /* legacy 32-bit */ - page_table_level = 3; - } else if (desc->addressing_mode == 3) { /* legacy 64 bit */ - page_table_level = 4; - } else { - gvt_vgpu_err("Advanced Context mode(SVM) is not supported!\n"); - return -EINVAL; - } - - read_guest_pdps(workload->vgpu, workload->ring_context_gpa, pdp); - - mm = intel_vgpu_find_ppgtt_mm(workload->vgpu, page_table_level, pdp); - if (mm) { - intel_gvt_mm_reference(mm); - } else { - - mm = intel_vgpu_create_mm(workload->vgpu, INTEL_GVT_MM_PPGTT, - pdp, page_table_level, 0); - if (IS_ERR(mm)) { - gvt_vgpu_err("fail to create mm object.\n"); - return PTR_ERR(mm); - } - } - workload->shadow_mm = mm; - return 0; -} - -#define get_last_workload(q) \ - (list_empty(q) ? NULL : container_of(q->prev, \ - struct intel_vgpu_workload, list)) - static int submit_context(struct intel_vgpu *vgpu, int ring_id, struct execlist_ctx_descriptor_format *desc, bool emulate_schedule_in) { struct intel_vgpu_submission *s = &vgpu->submission; - struct list_head *q = workload_q_head(vgpu, ring_id); - struct intel_vgpu_workload *last_workload = get_last_workload(q); struct intel_vgpu_workload *workload = NULL; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; - u64 ring_context_gpa; - u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx; - int ret; - - ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, - (u32)((desc->lrca + 1) << GTT_PAGE_SHIFT)); - if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) { - gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca); - return -EINVAL; - } - - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(ring_header.val), &head, 4); - - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(ring_tail.val), &tail, 4); - - head &= RB_HEAD_OFF_MASK; - tail &= RB_TAIL_OFF_MASK; - if (last_workload && same_context(&last_workload->ctx_desc, desc)) { - gvt_dbg_el("ring id %d cur workload == last\n", ring_id); - gvt_dbg_el("ctx head %x real head %lx\n", head, - last_workload->rb_tail); - /* - * cannot use guest context head pointer here, - * as it might not be updated at this time - */ - head = last_workload->rb_tail; - } - - gvt_dbg_el("ring id %d begin a new workload\n", ring_id); - - /* record some ring buffer register values for scan and shadow */ - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(rb_start.val), &start, 4); - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(rb_ctrl.val), &ctl, 4); - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4); - - workload = intel_vgpu_create_workload(vgpu); + workload = intel_vgpu_create_workload(vgpu, ring_id, desc); if (IS_ERR(workload)) return PTR_ERR(workload); - workload->ring_id = ring_id; - workload->ctx_desc = *desc; - workload->ring_context_gpa = ring_context_gpa; - workload->rb_head = head; - workload->rb_tail = tail; - workload->rb_start = start; - workload->rb_ctl = ctl; workload->prepare = prepare_execlist_workload; workload->complete = complete_execlist_workload; workload->emulate_schedule_in = emulate_schedule_in; - if (ring_id == RCS) { - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4); - intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + - RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4); - - workload->wa_ctx.indirect_ctx.guest_gma = - indirect_ctx & INDIRECT_CTX_ADDR_MASK; - workload->wa_ctx.indirect_ctx.size = - (indirect_ctx & INDIRECT_CTX_SIZE_MASK) * - CACHELINE_BYTES; - workload->wa_ctx.per_ctx.guest_gma = - per_ctx & PER_CTX_ADDR_MASK; - workload->wa_ctx.per_ctx.valid = per_ctx & 1; - } - if (emulate_schedule_in) workload->elsp_dwords = s->execlist[ring_id].elsp_dwords; - gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n", - workload, ring_id, head, tail, start, ctl); - gvt_dbg_el("workload %p emulate schedule_in %d\n", workload, emulate_schedule_in); - ret = prepare_mm(workload); - if (ret) { - kmem_cache_free(s->workloads, workload); - return ret; - } - - /* Only scan and shadow the first workload in the queue - * as there is only one pre-allocated buf-obj for shadow. - */ - if (list_empty(workload_q_head(vgpu, ring_id))) { - intel_runtime_pm_get(dev_priv); - mutex_lock(&dev_priv->drm.struct_mutex); - ret = intel_gvt_scan_and_shadow_workload(workload); - mutex_unlock(&dev_priv->drm.struct_mutex); - intel_runtime_pm_put(dev_priv); - } - - if (ret == 0) - queue_workload(workload); - else { - intel_vgpu_destroy_workload(workload); - if (vgpu_is_vm_unhealthy(ret)) { - intel_vgpu_clean_execlist(vgpu); - enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); - } - } - return ret; - + queue_workload(workload); + return 0; } int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index a7a67cc65a071..69893f29ff6dc 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -792,13 +792,8 @@ static int workload_thread(void *priv) FORCEWAKE_ALL); intel_runtime_pm_put(gvt->dev_priv); - if (ret && (vgpu_is_vm_unhealthy(ret))) { - mutex_lock(&gvt->lock); - intel_vgpu_clean_execlist(vgpu); - mutex_unlock(&gvt->lock); + if (ret && (vgpu_is_vm_unhealthy(ret))) enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); - } - } return 0; } @@ -957,9 +952,90 @@ void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload) kmem_cache_free(s->workloads, workload); } +static struct intel_vgpu_workload * +alloc_workload(struct intel_vgpu *vgpu) +{ + struct intel_vgpu_submission *s = &vgpu->submission; + struct intel_vgpu_workload *workload; + + workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL); + if (!workload) + return ERR_PTR(-ENOMEM); + + INIT_LIST_HEAD(&workload->list); + INIT_LIST_HEAD(&workload->shadow_bb); + + init_waitqueue_head(&workload->shadow_ctx_status_wq); + atomic_set(&workload->shadow_ctx_active, 0); + + workload->status = -EINPROGRESS; + workload->shadowed = false; + workload->vgpu = vgpu; + + return workload; +} + +#define RING_CTX_OFF(x) \ + offsetof(struct execlist_ring_context, x) + +static void read_guest_pdps(struct intel_vgpu *vgpu, + u64 ring_context_gpa, u32 pdp[8]) +{ + u64 gpa; + int i; + + gpa = ring_context_gpa + RING_CTX_OFF(pdp3_UDW.val); + + for (i = 0; i < 8; i++) + intel_gvt_hypervisor_read_gpa(vgpu, + gpa + i * 8, &pdp[7 - i], 4); +} + +static int prepare_mm(struct intel_vgpu_workload *workload) +{ + struct execlist_ctx_descriptor_format *desc = &workload->ctx_desc; + struct intel_vgpu_mm *mm; + struct intel_vgpu *vgpu = workload->vgpu; + int page_table_level; + u32 pdp[8]; + + if (desc->addressing_mode == 1) { /* legacy 32-bit */ + page_table_level = 3; + } else if (desc->addressing_mode == 3) { /* legacy 64 bit */ + page_table_level = 4; + } else { + gvt_vgpu_err("Advanced Context mode(SVM) is not supported!\n"); + return -EINVAL; + } + + read_guest_pdps(workload->vgpu, workload->ring_context_gpa, pdp); + + mm = intel_vgpu_find_ppgtt_mm(workload->vgpu, page_table_level, pdp); + if (mm) { + intel_gvt_mm_reference(mm); + } else { + + mm = intel_vgpu_create_mm(workload->vgpu, INTEL_GVT_MM_PPGTT, + pdp, page_table_level, 0); + if (IS_ERR(mm)) { + gvt_vgpu_err("fail to create mm object.\n"); + return PTR_ERR(mm); + } + } + workload->shadow_mm = mm; + return 0; +} + +#define same_context(a, b) (((a)->context_id == (b)->context_id) && \ + ((a)->lrca == (b)->lrca)) + +#define get_last_workload(q) \ + (list_empty(q) ? NULL : container_of(q->prev, \ + struct intel_vgpu_workload, list)) /** * intel_vgpu_create_workload - create a vGPU workload * @vgpu: a vGPU + * @desc: a guest context descriptor * * This function is called when creating a vGPU workload. * @@ -969,24 +1045,108 @@ void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload) * */ struct intel_vgpu_workload * -intel_vgpu_create_workload(struct intel_vgpu *vgpu) +intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, + struct execlist_ctx_descriptor_format *desc) { struct intel_vgpu_submission *s = &vgpu->submission; - struct intel_vgpu_workload *workload; + struct list_head *q = workload_q_head(vgpu, ring_id); + struct intel_vgpu_workload *last_workload = get_last_workload(q); + struct intel_vgpu_workload *workload = NULL; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + u64 ring_context_gpa; + u32 head, tail, start, ctl, ctx_ctl, per_ctx, indirect_ctx; + int ret; - workload = kmem_cache_zalloc(s->workloads, GFP_KERNEL); - if (!workload) - return ERR_PTR(-ENOMEM); + ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, + (u32)((desc->lrca + 1) << GTT_PAGE_SHIFT)); + if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) { + gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca); + return ERR_PTR(-EINVAL); + } - INIT_LIST_HEAD(&workload->list); - INIT_LIST_HEAD(&workload->shadow_bb); + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(ring_header.val), &head, 4); - init_waitqueue_head(&workload->shadow_ctx_status_wq); - atomic_set(&workload->shadow_ctx_active, 0); + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(ring_tail.val), &tail, 4); - workload->status = -EINPROGRESS; - workload->shadowed = false; - workload->vgpu = vgpu; + head &= RB_HEAD_OFF_MASK; + tail &= RB_TAIL_OFF_MASK; + + if (last_workload && same_context(&last_workload->ctx_desc, desc)) { + gvt_dbg_el("ring id %d cur workload == last\n", ring_id); + gvt_dbg_el("ctx head %x real head %lx\n", head, + last_workload->rb_tail); + /* + * cannot use guest context head pointer here, + * as it might not be updated at this time + */ + head = last_workload->rb_tail; + } + + gvt_dbg_el("ring id %d begin a new workload\n", ring_id); + + /* record some ring buffer register values for scan and shadow */ + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(rb_start.val), &start, 4); + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(rb_ctrl.val), &ctl, 4); + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(ctx_ctrl.val), &ctx_ctl, 4); + + workload = alloc_workload(vgpu); + if (IS_ERR(workload)) + return workload; + + workload->ring_id = ring_id; + workload->ctx_desc = *desc; + workload->ring_context_gpa = ring_context_gpa; + workload->rb_head = head; + workload->rb_tail = tail; + workload->rb_start = start; + workload->rb_ctl = ctl; + + if (ring_id == RCS) { + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(bb_per_ctx_ptr.val), &per_ctx, 4); + intel_gvt_hypervisor_read_gpa(vgpu, ring_context_gpa + + RING_CTX_OFF(rcs_indirect_ctx.val), &indirect_ctx, 4); + + workload->wa_ctx.indirect_ctx.guest_gma = + indirect_ctx & INDIRECT_CTX_ADDR_MASK; + workload->wa_ctx.indirect_ctx.size = + (indirect_ctx & INDIRECT_CTX_SIZE_MASK) * + CACHELINE_BYTES; + workload->wa_ctx.per_ctx.guest_gma = + per_ctx & PER_CTX_ADDR_MASK; + workload->wa_ctx.per_ctx.valid = per_ctx & 1; + } + + gvt_dbg_el("workload %p ring id %d head %x tail %x start %x ctl %x\n", + workload, ring_id, head, tail, start, ctl); + + ret = prepare_mm(workload); + if (ret) { + kmem_cache_free(s->workloads, workload); + return ERR_PTR(ret); + } + + /* Only scan and shadow the first workload in the queue + * as there is only one pre-allocated buf-obj for shadow. + */ + if (list_empty(workload_q_head(vgpu, ring_id))) { + intel_runtime_pm_get(dev_priv); + mutex_lock(&dev_priv->drm.struct_mutex); + ret = intel_gvt_scan_and_shadow_workload(workload); + mutex_unlock(&dev_priv->drm.struct_mutex); + intel_runtime_pm_put(dev_priv); + } + + if (ret && (vgpu_is_vm_unhealthy(ret))) { + enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR); + intel_vgpu_destroy_workload(workload); + return ERR_PTR(ret); + } return workload; } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 3fe4702cda06b..79199693697bd 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -142,7 +142,8 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu); void intel_vgpu_clean_submission(struct intel_vgpu *vgpu); struct intel_vgpu_workload * -intel_vgpu_create_workload(struct intel_vgpu *vgpu); +intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, + struct execlist_ctx_descriptor_format *desc); void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload); -- GitLab From d0d51282b88e11a1aa71040b21a9a8cae584a1d4 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 12 Sep 2017 22:39:08 +0800 Subject: [PATCH 0282/1471] drm/i915/gvt: Remove one extra declaration in scheduler.h Now the function has been moved into scheduler.c. The extra declaration is not necessary. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/scheduler.h | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 79199693697bd..2dc729320a61b 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -147,5 +147,4 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, void intel_vgpu_destroy_workload(struct intel_vgpu_workload *workload); -void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx); #endif -- GitLab From ad1d36369b07f6b9db81897802ee5d8764eaa922 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Wed, 13 Sep 2017 00:31:29 +0800 Subject: [PATCH 0283/1471] drm/i915/gvt: Introduce vGPU submission ops Introduce vGPU submission ops to support easy switching submission mode of one vGPU between different OSes. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 10 ++++-- drivers/gpu/drm/i915/gvt/gvt.h | 14 ++++++++ drivers/gpu/drm/i915/gvt/handlers.c | 16 +++++++-- drivers/gpu/drm/i915/gvt/scheduler.c | 53 ++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/scheduler.h | 6 ++++ drivers/gpu/drm/i915/gvt/vgpu.c | 13 ++----- 6 files changed, 98 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index c753cb5bfb589..e633ac4991e89 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -548,7 +548,7 @@ static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) } } -void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) +void clean_execlist(struct intel_vgpu *vgpu) { enum intel_engine_id i; struct intel_engine_cs *engine; @@ -564,7 +564,7 @@ void intel_vgpu_clean_execlist(struct intel_vgpu *vgpu) } } -int intel_vgpu_init_execlist(struct intel_vgpu *vgpu) +int init_execlist(struct intel_vgpu *vgpu) { enum intel_engine_id i; struct intel_engine_cs *engine; @@ -586,3 +586,9 @@ void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu, for_each_engine_masked(engine, dev_priv, engine_mask, tmp) init_vgpu_execlist(vgpu, engine->id); } + +const struct intel_vgpu_submission_ops intel_vgpu_execlist_submission_ops = { + .name = "execlist", + .init = init_execlist, + .clean = clean_execlist, +}; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index be1bb6dbcbd27..1be41505c4ba2 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -142,6 +142,17 @@ struct vgpu_sched_ctl { int weight; }; +enum { + INTEL_VGPU_EXECLIST_SUBMISSION = 1, + INTEL_VGPU_GUC_SUBMISSION, +}; + +struct intel_vgpu_submission_ops { + const char *name; + int (*init)(struct intel_vgpu *vgpu); + void (*clean)(struct intel_vgpu *vgpu); +}; + struct intel_vgpu_submission { struct intel_vgpu_execlist execlist[I915_NUM_ENGINES]; struct list_head workload_q_head[I915_NUM_ENGINES]; @@ -152,6 +163,9 @@ struct intel_vgpu_submission { DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES); void *ring_scan_buffer[I915_NUM_ENGINES]; int ring_scan_buffer_size[I915_NUM_ENGINES]; + const struct intel_vgpu_submission_ops *ops; + int virtual_submission_interface; + bool active; }; struct intel_vgpu { diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index acc1cf4fa6f53..5438dafea7a49 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1471,9 +1471,11 @@ static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { + struct intel_vgpu_submission *s = &vgpu->submission; u32 data = *(u32 *)p_data; int ring_id = render_mmio_to_ring_id(vgpu->gvt, offset); bool enable_execlist; + int ret; write_vreg(vgpu, offset, p_data, bytes); @@ -1495,8 +1497,18 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, (enable_execlist ? "enabling" : "disabling"), ring_id); - if (enable_execlist) - intel_vgpu_start_schedule(vgpu); + if (!enable_execlist) + return 0; + + if (s->active) + return 0; + + ret = intel_vgpu_select_submission_ops(vgpu, + INTEL_VGPU_EXECLIST_SUBMISSION); + if (ret) + return ret; + + intel_vgpu_start_schedule(vgpu); } return 0; } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 69893f29ff6dc..f3be88fa88dda 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -884,6 +884,7 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) { struct intel_vgpu_submission *s = &vgpu->submission; + intel_vgpu_select_submission_ops(vgpu, 0); i915_gem_context_put(s->shadow_ctx); kmem_cache_destroy(s->workloads); } @@ -935,6 +936,58 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) return ret; } +/** + * intel_vgpu_select_submission_ops - select virtual submission interface + * @vgpu: a vGPU + * @interface: expected vGPU virtual submission interface + * + * This function is called when guest configures submission interface. + * + * Returns: + * Zero on success, negative error code if failed. + * + */ +int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu, + unsigned int interface) +{ + struct intel_vgpu_submission *s = &vgpu->submission; + const struct intel_vgpu_submission_ops *ops[] = { + [INTEL_VGPU_EXECLIST_SUBMISSION] = + &intel_vgpu_execlist_submission_ops, + }; + int ret; + + if (WARN_ON(interface >= ARRAY_SIZE(ops))) + return -EINVAL; + + if (s->active) { + s->ops->clean(vgpu); + s->active = false; + gvt_dbg_core("vgpu%d: de-select ops [ %s ] \n", + vgpu->id, s->ops->name); + } + + if (interface == 0) { + s->ops = NULL; + s->virtual_submission_interface = 0; + gvt_dbg_core("vgpu%d: no submission ops\n", vgpu->id); + return 0; + } + + ret = ops[interface]->init(vgpu); + if (ret) + return ret; + + s->ops = ops[interface]; + s->virtual_submission_interface = interface; + s->active = true; + + gvt_dbg_core("vgpu%d: activate ops [ %s ]\n", + vgpu->id, s->ops->name); + + return 0; +} + /** * intel_vgpu_destroy_workload - destroy a vGPU workload * @vgpu: a vGPU diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 2dc729320a61b..8652acda6436c 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -141,6 +141,12 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu); void intel_vgpu_clean_submission(struct intel_vgpu *vgpu); +int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu, + unsigned int interface); + +extern const struct intel_vgpu_submission_ops +intel_vgpu_execlist_submission_ops; + struct intel_vgpu_workload * intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, struct execlist_ctx_descriptor_format *desc); diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 1c9818d9f1d85..e63abd48bceef 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -255,7 +255,6 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu) idr_remove(&gvt->vgpu_idr, vgpu->id); intel_vgpu_clean_sched_policy(vgpu); intel_vgpu_clean_submission(vgpu); - intel_vgpu_clean_execlist(vgpu); intel_vgpu_clean_display(vgpu); intel_vgpu_clean_opregion(vgpu); intel_vgpu_clean_gtt(vgpu); @@ -371,26 +370,20 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, if (ret) goto out_clean_gtt; - ret = intel_vgpu_init_execlist(vgpu); - if (ret) - goto out_clean_display; - ret = intel_vgpu_setup_submission(vgpu); if (ret) - goto out_clean_execlist; + goto out_clean_display; ret = intel_vgpu_init_sched_policy(vgpu); if (ret) - goto out_clean_shadow_ctx; + goto out_clean_submission; mutex_unlock(&gvt->lock); return vgpu; -out_clean_shadow_ctx: +out_clean_submission: intel_vgpu_clean_submission(vgpu); -out_clean_execlist: - intel_vgpu_clean_execlist(vgpu); out_clean_display: intel_vgpu_clean_display(vgpu); out_clean_gtt: -- GitLab From 06bb372f9ace47296aeaaca8e130d948ea2855cf Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Wed, 13 Sep 2017 01:41:35 +0800 Subject: [PATCH 0284/1471] drm/i915/gvt: Introduce intel_vgpu_reset_submission Introduce an generic API to reset vGPU virtual submission interface. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 20 ++++++++------------ drivers/gpu/drm/i915/gvt/gvt.h | 1 + drivers/gpu/drm/i915/gvt/scheduler.c | 20 ++++++++++++++++++++ drivers/gpu/drm/i915/gvt/scheduler.h | 3 +++ drivers/gpu/drm/i915/gvt/vgpu.c | 4 ++-- 5 files changed, 34 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index e633ac4991e89..107eeda82e9dd 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -564,18 +564,7 @@ void clean_execlist(struct intel_vgpu *vgpu) } } -int init_execlist(struct intel_vgpu *vgpu) -{ - enum intel_engine_id i; - struct intel_engine_cs *engine; - - for_each_engine(engine, vgpu->gvt->dev_priv, i) - init_vgpu_execlist(vgpu, i); - - return 0; -} - -void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu, +void reset_execlist(struct intel_vgpu *vgpu, unsigned long engine_mask) { struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; @@ -587,8 +576,15 @@ void intel_vgpu_reset_execlist(struct intel_vgpu *vgpu, init_vgpu_execlist(vgpu, engine->id); } +int init_execlist(struct intel_vgpu *vgpu) +{ + reset_execlist(vgpu, ALL_ENGINES); + return 0; +} + const struct intel_vgpu_submission_ops intel_vgpu_execlist_submission_ops = { .name = "execlist", .init = init_execlist, + .reset = reset_execlist, .clean = clean_execlist, }; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 1be41505c4ba2..b7253d7daf725 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -151,6 +151,7 @@ struct intel_vgpu_submission_ops { const char *name; int (*init)(struct intel_vgpu *vgpu); void (*clean)(struct intel_vgpu *vgpu); + void (*reset)(struct intel_vgpu *vgpu, unsigned long engine_mask); }; struct intel_vgpu_submission { diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index f3be88fa88dda..88ce57116a4c1 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -889,6 +889,26 @@ void intel_vgpu_clean_submission(struct intel_vgpu *vgpu) kmem_cache_destroy(s->workloads); } + +/** + * intel_vgpu_reset_submission - reset submission-related resource for vGPU + * @vgpu: a vGPU + * @engine_mask: engines expected to be reset + * + * This function is called when a vGPU is being destroyed. + * + */ +void intel_vgpu_reset_submission(struct intel_vgpu *vgpu, + unsigned long engine_mask) +{ + struct intel_vgpu_submission *s = &vgpu->submission; + + if (!s->active) + return; + + s->ops->reset(vgpu, engine_mask); +} + /** * intel_vgpu_setup_submission - setup submission-related resource for vGPU * @vgpu: a vGPU diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index 8652acda6436c..e0b5730a30187 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -139,6 +139,9 @@ void intel_gvt_wait_vgpu_idle(struct intel_vgpu *vgpu); int intel_vgpu_setup_submission(struct intel_vgpu *vgpu); +void intel_vgpu_reset_submission(struct intel_vgpu *vgpu, + unsigned long engine_mask); + void intel_vgpu_clean_submission(struct intel_vgpu *vgpu); int intel_vgpu_select_submission_ops(struct intel_vgpu *vgpu, diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index e63abd48bceef..841b95c6c2313 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -492,10 +492,10 @@ void intel_gvt_reset_vgpu_locked(struct intel_vgpu *vgpu, bool dmlr, mutex_lock(&gvt->lock); } - intel_vgpu_reset_execlist(vgpu, resetting_eng); - + intel_vgpu_reset_submission(vgpu, resetting_eng); /* full GPU reset or device model level reset */ if (engine_mask == ALL_ENGINES || dmlr) { + intel_vgpu_select_submission_ops(vgpu, 0); /*fence will not be reset during virtual reset */ if (dmlr) { -- GitLab From e2c43c0111d54d4857d052fcfca9f3f16bf1b1b2 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Wed, 13 Sep 2017 01:58:35 +0800 Subject: [PATCH 0285/1471] drm/i915/gvt: Move clean_workloads() into scheduler.c Move clean_workloads() into scheduler.c since it's not specific to execlist. v2: - Remove clean_workloads in intel_vgpu_select_submission_ops. (Zhenyu) Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 41 +--------------------------- drivers/gpu/drm/i915/gvt/scheduler.c | 37 +++++++++++++++++++++++++ 2 files changed, 38 insertions(+), 40 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 107eeda82e9dd..5c966edfeefb6 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -46,8 +46,6 @@ #define same_context(a, b) (((a)->context_id == (b)->context_id) && \ ((a)->lrca == (b)->lrca)) -static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask); - static int context_switch_events[] = { [RCS] = RCS_AS_CONTEXT_SWITCH, [BCS] = BCS_AS_CONTEXT_SWITCH, @@ -397,23 +395,8 @@ static int complete_execlist_workload(struct intel_vgpu_workload *workload) gvt_dbg_el("complete workload %p status %d\n", workload, workload->status); - if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) { - /* if workload->status is not successful means HW GPU - * has occurred GPU hang or something wrong with i915/GVT, - * and GVT won't inject context switch interrupt to guest. - * So this error is a vGPU hang actually to the guest. - * According to this we should emunlate a vGPU hang. If - * there are pending workloads which are already submitted - * from guest, we should clean them up like HW GPU does. - * - * if it is in middle of engine resetting, the pending - * workloads won't be submitted to HW GPU and will be - * cleaned up during the resetting process later, so doing - * the workload clean up here doesn't have any impact. - **/ - clean_workloads(vgpu, ENGINE_MASK(ring_id)); + if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) goto out; - } if (!list_empty(workload_q_head(vgpu, ring_id))) { struct execlist_ctx_descriptor_format *this_desc, *next_desc; @@ -529,32 +512,11 @@ static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id) vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw; } -static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) -{ - struct intel_vgpu_submission *s = &vgpu->submission; - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; - struct intel_engine_cs *engine; - struct intel_vgpu_workload *pos, *n; - unsigned int tmp; - - /* free the unsubmited workloads in the queues. */ - for_each_engine_masked(engine, dev_priv, engine_mask, tmp) { - list_for_each_entry_safe(pos, n, - &s->workload_q_head[engine->id], list) { - list_del_init(&pos->list); - intel_vgpu_destroy_workload(pos); - } - clear_bit(engine->id, s->shadow_ctx_desc_updated); - } -} - void clean_execlist(struct intel_vgpu *vgpu) { enum intel_engine_id i; struct intel_engine_cs *engine; - clean_workloads(vgpu, ALL_ENGINES); - for_each_engine(engine, vgpu->gvt->dev_priv, i) { struct intel_vgpu_submission *s = &vgpu->submission; @@ -571,7 +533,6 @@ void reset_execlist(struct intel_vgpu *vgpu, struct intel_engine_cs *engine; unsigned int tmp; - clean_workloads(vgpu, engine_mask); for_each_engine_masked(engine, dev_priv, engine_mask, tmp) init_vgpu_execlist(vgpu, engine->id); } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 88ce57116a4c1..391690c0c28c1 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -644,6 +644,25 @@ static void update_guest_context(struct intel_vgpu_workload *workload) kunmap(page); } +static void clean_workloads(struct intel_vgpu *vgpu, unsigned long engine_mask) +{ + struct intel_vgpu_submission *s = &vgpu->submission; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + struct intel_engine_cs *engine; + struct intel_vgpu_workload *pos, *n; + unsigned int tmp; + + /* free the unsubmited workloads in the queues. */ + for_each_engine_masked(engine, dev_priv, engine_mask, tmp) { + list_for_each_entry_safe(pos, n, + &s->workload_q_head[engine->id], list) { + list_del_init(&pos->list); + intel_vgpu_destroy_workload(pos); + } + clear_bit(engine->id, s->shadow_ctx_desc_updated); + } +} + static void complete_current_workload(struct intel_gvt *gvt, int ring_id) { struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler; @@ -707,6 +726,23 @@ static void complete_current_workload(struct intel_gvt *gvt, int ring_id) release_shadow_wa_ctx(&workload->wa_ctx); } + if (workload->status || (vgpu->resetting_eng & ENGINE_MASK(ring_id))) { + /* if workload->status is not successful means HW GPU + * has occurred GPU hang or something wrong with i915/GVT, + * and GVT won't inject context switch interrupt to guest. + * So this error is a vGPU hang actually to the guest. + * According to this we should emunlate a vGPU hang. If + * there are pending workloads which are already submitted + * from guest, we should clean them up like HW GPU does. + * + * if it is in middle of engine resetting, the pending + * workloads won't be submitted to HW GPU and will be + * cleaned up during the resetting process later, so doing + * the workload clean up here doesn't have any impact. + **/ + clean_workloads(vgpu, ENGINE_MASK(ring_id)); + } + workload->complete(workload); atomic_dec(&s->running_workload_num); @@ -906,6 +942,7 @@ void intel_vgpu_reset_submission(struct intel_vgpu *vgpu, if (!s->active) return; + clean_workloads(vgpu, engine_mask); s->ops->reset(vgpu, engine_mask); } -- GitLab From c5d71cb31723135dbaa54b32bf046342fe538210 Mon Sep 17 00:00:00 2001 From: fred gao Date: Thu, 28 Sep 2017 11:03:02 +0800 Subject: [PATCH 0286/1471] drm/i915/gvt: Move vGPU type related code into gvt file In this patch, all the vGPU type related code will be merged into same gvt file and the common interface will be exposed to both XenGT and KvmGT. v2: - remove the useless mdev_* gvt_ops. add get_gvt_attr ops for MPT module. intel_gvt_{init,cleanup}_vgpu_type_groups are initialized in gvt part. (Wang, Zhi) - set gvt_vgpu_type_groups[i] to NULL. (Zhang,Xiong) Signed-off-by: fred gao Reviewed-by: Zhi Wang Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gvt.c | 134 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/gvt.h | 4 + 2 files changed, 138 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c index aaa347f8620cc..ed313811fa761 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.c +++ b/drivers/gpu/drm/i915/gvt/gvt.c @@ -36,6 +36,8 @@ #include "i915_drv.h" #include "gvt.h" +#include +#include struct intel_gvt_host intel_gvt_host; @@ -44,6 +46,129 @@ static const char * const supported_hypervisors[] = { [INTEL_GVT_HYPERVISOR_KVM] = "KVM", }; +static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt, + const char *name) +{ + int i; + struct intel_vgpu_type *t; + const char *driver_name = dev_driver_string( + &gvt->dev_priv->drm.pdev->dev); + + for (i = 0; i < gvt->num_types; i++) { + t = &gvt->types[i]; + if (!strncmp(t->name, name + strlen(driver_name) + 1, + sizeof(t->name))) + return t; + } + + return NULL; +} + +static ssize_t available_instances_show(struct kobject *kobj, + struct device *dev, char *buf) +{ + struct intel_vgpu_type *type; + unsigned int num = 0; + void *gvt = kdev_to_i915(dev)->gvt; + + type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj)); + if (!type) + num = 0; + else + num = type->avail_instance; + + return sprintf(buf, "%u\n", num); +} + +static ssize_t device_api_show(struct kobject *kobj, struct device *dev, + char *buf) +{ + return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING); +} + +static ssize_t description_show(struct kobject *kobj, struct device *dev, + char *buf) +{ + struct intel_vgpu_type *type; + void *gvt = kdev_to_i915(dev)->gvt; + + type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj)); + if (!type) + return 0; + + return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n" + "fence: %d\nresolution: %s\n" + "weight: %d\n", + BYTES_TO_MB(type->low_gm_size), + BYTES_TO_MB(type->high_gm_size), + type->fence, vgpu_edid_str(type->resolution), + type->weight); +} + +static MDEV_TYPE_ATTR_RO(available_instances); +static MDEV_TYPE_ATTR_RO(device_api); +static MDEV_TYPE_ATTR_RO(description); + +static struct attribute *gvt_type_attrs[] = { + &mdev_type_attr_available_instances.attr, + &mdev_type_attr_device_api.attr, + &mdev_type_attr_description.attr, + NULL, +}; + +static struct attribute_group *gvt_vgpu_type_groups[] = { + [0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL, +}; + +static bool intel_get_gvt_attrs(struct attribute ***type_attrs, + struct attribute_group ***intel_vgpu_type_groups) +{ + *type_attrs = gvt_type_attrs; + *intel_vgpu_type_groups = gvt_vgpu_type_groups; + return true; +} + +static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt) +{ + int i, j; + struct intel_vgpu_type *type; + struct attribute_group *group; + + for (i = 0; i < gvt->num_types; i++) { + type = &gvt->types[i]; + + group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL); + if (WARN_ON(!group)) + goto unwind; + + group->name = type->name; + group->attrs = gvt_type_attrs; + gvt_vgpu_type_groups[i] = group; + } + + return true; + +unwind: + for (j = 0; j < i; j++) { + group = gvt_vgpu_type_groups[j]; + kfree(group); + } + + return false; +} + +static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt) +{ + int i; + struct attribute_group *group; + + for (i = 0; i < gvt->num_types; i++) { + group = gvt_vgpu_type_groups[i]; + gvt_vgpu_type_groups[i] = NULL; + kfree(group); + } +} + static const struct intel_gvt_ops intel_gvt_ops = { .emulate_cfg_read = intel_vgpu_emulate_cfg_read, .emulate_cfg_write = intel_vgpu_emulate_cfg_write, @@ -54,6 +179,8 @@ static const struct intel_gvt_ops intel_gvt_ops = { .vgpu_reset = intel_gvt_reset_vgpu, .vgpu_activate = intel_gvt_activate_vgpu, .vgpu_deactivate = intel_gvt_deactivate_vgpu, + .gvt_find_vgpu_type = intel_gvt_find_vgpu_type, + .get_gvt_attrs = intel_get_gvt_attrs, }; /** @@ -202,6 +329,7 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv) intel_gvt_free_firmware(gvt); intel_gvt_hypervisor_host_exit(&dev_priv->drm.pdev->dev, gvt); + intel_gvt_cleanup_vgpu_type_groups(gvt); intel_gvt_clean_vgpu_types(gvt); idr_destroy(&gvt->vgpu_idr); @@ -292,6 +420,12 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) if (ret) goto out_clean_thread; + ret = intel_gvt_init_vgpu_type_groups(gvt); + if (ret == false) { + gvt_err("failed to init vgpu type groups: %d\n", ret); + goto out_clean_types; + } + ret = intel_gvt_hypervisor_host_init(&dev_priv->drm.pdev->dev, gvt, &intel_gvt_ops); if (ret) { diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index b7253d7daf725..35c8d87151ddf 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -533,6 +533,10 @@ struct intel_gvt_ops { void (*vgpu_reset)(struct intel_vgpu *); void (*vgpu_activate)(struct intel_vgpu *); void (*vgpu_deactivate)(struct intel_vgpu *); + struct intel_vgpu_type *(*gvt_find_vgpu_type)(struct intel_gvt *gvt, + const char *name); + bool (*get_gvt_attrs)(struct attribute ***type_attrs, + struct attribute_group ***intel_vgpu_type_groups); }; -- GitLab From 6aa23ced915f94e23cd7bc3c1d005e39ef39bb2b Mon Sep 17 00:00:00 2001 From: fred gao Date: Thu, 28 Sep 2017 11:03:03 +0800 Subject: [PATCH 0287/1471] drm/i915/gvt: Refactor vGPU type code in kvmgt part all the vGPU type related code in kvmgt will be moved into gvt.c/gvt.h files while the common vGPU type related interfaces will be called. v2: - intel_gvt_{init,cleanup}_vgpu_type_groups are initialized in gvt part. (Wang, Zhi) Signed-off-by: fred gao Reviewed-by: Zhi Wang Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/kvmgt.c | 128 ++----------------------------- 1 file changed, 8 insertions(+), 120 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c index fc6878bb24968..110f07e8bcfb9 100644 --- a/drivers/gpu/drm/i915/gvt/kvmgt.c +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c @@ -248,120 +248,6 @@ static void gvt_cache_destroy(struct intel_vgpu *vgpu) } } -static struct intel_vgpu_type *intel_gvt_find_vgpu_type(struct intel_gvt *gvt, - const char *name) -{ - int i; - struct intel_vgpu_type *t; - const char *driver_name = dev_driver_string( - &gvt->dev_priv->drm.pdev->dev); - - for (i = 0; i < gvt->num_types; i++) { - t = &gvt->types[i]; - if (!strncmp(t->name, name + strlen(driver_name) + 1, - sizeof(t->name))) - return t; - } - - return NULL; -} - -static ssize_t available_instances_show(struct kobject *kobj, - struct device *dev, char *buf) -{ - struct intel_vgpu_type *type; - unsigned int num = 0; - void *gvt = kdev_to_i915(dev)->gvt; - - type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj)); - if (!type) - num = 0; - else - num = type->avail_instance; - - return sprintf(buf, "%u\n", num); -} - -static ssize_t device_api_show(struct kobject *kobj, struct device *dev, - char *buf) -{ - return sprintf(buf, "%s\n", VFIO_DEVICE_API_PCI_STRING); -} - -static ssize_t description_show(struct kobject *kobj, struct device *dev, - char *buf) -{ - struct intel_vgpu_type *type; - void *gvt = kdev_to_i915(dev)->gvt; - - type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj)); - if (!type) - return 0; - - return sprintf(buf, "low_gm_size: %dMB\nhigh_gm_size: %dMB\n" - "fence: %d\nresolution: %s\n" - "weight: %d\n", - BYTES_TO_MB(type->low_gm_size), - BYTES_TO_MB(type->high_gm_size), - type->fence, vgpu_edid_str(type->resolution), - type->weight); -} - -static MDEV_TYPE_ATTR_RO(available_instances); -static MDEV_TYPE_ATTR_RO(device_api); -static MDEV_TYPE_ATTR_RO(description); - -static struct attribute *type_attrs[] = { - &mdev_type_attr_available_instances.attr, - &mdev_type_attr_device_api.attr, - &mdev_type_attr_description.attr, - NULL, -}; - -static struct attribute_group *intel_vgpu_type_groups[] = { - [0 ... NR_MAX_INTEL_VGPU_TYPES - 1] = NULL, -}; - -static bool intel_gvt_init_vgpu_type_groups(struct intel_gvt *gvt) -{ - int i, j; - struct intel_vgpu_type *type; - struct attribute_group *group; - - for (i = 0; i < gvt->num_types; i++) { - type = &gvt->types[i]; - - group = kzalloc(sizeof(struct attribute_group), GFP_KERNEL); - if (WARN_ON(!group)) - goto unwind; - - group->name = type->name; - group->attrs = type_attrs; - intel_vgpu_type_groups[i] = group; - } - - return true; - -unwind: - for (j = 0; j < i; j++) { - group = intel_vgpu_type_groups[j]; - kfree(group); - } - - return false; -} - -static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt) -{ - int i; - struct attribute_group *group; - - for (i = 0; i < gvt->num_types; i++) { - group = intel_vgpu_type_groups[i]; - kfree(group); - } -} - static void kvmgt_protect_table_init(struct kvmgt_guest_info *info) { hash_init(info->ptable); @@ -441,7 +327,7 @@ static int intel_vgpu_create(struct kobject *kobj, struct mdev_device *mdev) pdev = mdev_parent_dev(mdev); gvt = kdev_to_i915(pdev)->gvt; - type = intel_gvt_find_vgpu_type(gvt, kobject_name(kobj)); + type = intel_gvt_ops->gvt_find_vgpu_type(gvt, kobject_name(kobj)); if (!type) { gvt_vgpu_err("failed to find type %s to create\n", kobject_name(kobj)); @@ -1212,8 +1098,7 @@ static const struct attribute_group *intel_vgpu_groups[] = { NULL, }; -static const struct mdev_parent_ops intel_vgpu_ops = { - .supported_type_groups = intel_vgpu_type_groups, +static struct mdev_parent_ops intel_vgpu_ops = { .mdev_attr_groups = intel_vgpu_groups, .create = intel_vgpu_create, .remove = intel_vgpu_remove, @@ -1229,17 +1114,20 @@ static const struct mdev_parent_ops intel_vgpu_ops = { static int kvmgt_host_init(struct device *dev, void *gvt, const void *ops) { - if (!intel_gvt_init_vgpu_type_groups(gvt)) - return -EFAULT; + struct attribute **kvm_type_attrs; + struct attribute_group **kvm_vgpu_type_groups; intel_gvt_ops = ops; + if (!intel_gvt_ops->get_gvt_attrs(&kvm_type_attrs, + &kvm_vgpu_type_groups)) + return -EFAULT; + intel_vgpu_ops.supported_type_groups = kvm_vgpu_type_groups; return mdev_register_device(dev, &intel_vgpu_ops); } static void kvmgt_host_exit(struct device *dev, void *gvt) { - intel_gvt_cleanup_vgpu_type_groups(gvt); mdev_unregister_device(dev); } -- GitLab From bc7b0be316aebac42eb9e8e54c984609555944da Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Tue, 26 Sep 2017 16:19:13 +0800 Subject: [PATCH 0288/1471] drm/i915/gvt: Add basic debugfs infrastructure MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We need debugfs entry to expose some debug information of gvt and vGPUs. The first tool will be added is mmio-diff, which help to find the difference values of host and vGPU mmio. It's useful for platform enabling. This patch just add a basic debugfs infrastructure, each vGPU has its own sub-folder. Two simple attributes are created as a template. . ├── num_tracked_mmio ├── vgpu1 | └── active └── vgpu2 └── active Signed-off-by: Changbin Du Reviewed-by: Zhi Wang Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/Makefile | 2 +- drivers/gpu/drm/i915/gvt/debugfs.c | 97 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/gvt.c | 5 ++ drivers/gpu/drm/i915/gvt/gvt.h | 12 +++- drivers/gpu/drm/i915/gvt/handlers.c | 2 - drivers/gpu/drm/i915/gvt/vgpu.c | 7 +++ 6 files changed, 121 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/i915/gvt/debugfs.c diff --git a/drivers/gpu/drm/i915/gvt/Makefile b/drivers/gpu/drm/i915/gvt/Makefile index f5486cb94818c..7100240a4a81a 100644 --- a/drivers/gpu/drm/i915/gvt/Makefile +++ b/drivers/gpu/drm/i915/gvt/Makefile @@ -1,7 +1,7 @@ GVT_DIR := gvt GVT_SOURCE := gvt.o aperture_gm.o handlers.o vgpu.o trace_points.o firmware.o \ interrupt.o gtt.o cfg_space.o opregion.o mmio.o display.o edid.o \ - execlist.o scheduler.o sched_policy.o render.o cmd_parser.o + execlist.o scheduler.o sched_policy.o render.o cmd_parser.o debugfs.o ccflags-y += -I$(src) -I$(src)/$(GVT_DIR) i915-y += $(addprefix $(GVT_DIR)/, $(GVT_SOURCE)) diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c new file mode 100644 index 0000000000000..02acd58426055 --- /dev/null +++ b/drivers/gpu/drm/i915/gvt/debugfs.c @@ -0,0 +1,97 @@ +/* + * Copyright(c) 2011-2017 Intel Corporation. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include +#include "i915_drv.h" +#include "gvt.h" + + +/** + * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU + * @vgpu: a vGPU + * + * Returns: + * Zero on success, negative error code if failed. + */ +int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu) +{ + struct dentry *ent; + char name[10] = ""; + + sprintf(name, "vgpu%d", vgpu->id); + vgpu->debugfs = debugfs_create_dir(name, vgpu->gvt->debugfs_root); + if (!vgpu->debugfs) + return -ENOMEM; + + ent = debugfs_create_bool("active", 0444, vgpu->debugfs, + &vgpu->active); + if (!ent) + return -ENOMEM; + + return 0; +} + +/** + * intel_gvt_debugfs_remove_vgpu - remove debugfs entries of a vGPU + * @vgpu: a vGPU + */ +void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu) +{ + debugfs_remove_recursive(vgpu->debugfs); + vgpu->debugfs = NULL; +} + +/** + * intel_gvt_debugfs_init - register gvt debugfs root entry + * @gvt: GVT device + * + * Returns: + * zero on success, negative if failed. + */ +int intel_gvt_debugfs_init(struct intel_gvt *gvt) +{ + struct drm_minor *minor = gvt->dev_priv->drm.primary; + struct dentry *ent; + + gvt->debugfs_root = debugfs_create_dir("gvt", minor->debugfs_root); + if (!gvt->debugfs_root) { + gvt_err("Cannot create debugfs dir\n"); + return -ENOMEM; + } + + ent = debugfs_create_ulong("num_tracked_mmio", 0444, gvt->debugfs_root, + &gvt->mmio.num_tracked_mmio); + if (!ent) + return -ENOMEM; + + return 0; +} + +/** + * intel_gvt_debugfs_clean - remove debugfs entries + * @gvt: GVT device + */ +void intel_gvt_debugfs_clean(struct intel_gvt *gvt) +{ + debugfs_remove_recursive(gvt->debugfs_root); + gvt->debugfs_root = NULL; +} diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c index ed313811fa761..cef06d59884e3 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.c +++ b/drivers/gpu/drm/i915/gvt/gvt.c @@ -318,6 +318,7 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv) if (WARN_ON(!gvt)) return; + intel_gvt_debugfs_clean(gvt); clean_service_thread(gvt); intel_gvt_clean_cmd_parser(gvt); intel_gvt_clean_sched_policy(gvt); @@ -441,6 +442,10 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) } gvt->idle_vgpu = vgpu; + ret = intel_gvt_debugfs_init(gvt); + if (ret) + gvt_err("debugfs registeration failed, go on.\n"); + gvt_dbg_core("gvt device initialization is done\n"); dev_priv->gvt = gvt; return 0; diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 35c8d87151ddf..f08d194f639fa 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -190,6 +190,8 @@ struct intel_vgpu { struct intel_vgpu_display display; struct intel_vgpu_submission submission; + struct dentry *debugfs; + #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT) struct { struct mdev_device *mdev; @@ -253,7 +255,7 @@ struct intel_gvt_mmio { unsigned int num_mmio_block; DECLARE_HASHTABLE(mmio_info_table, INTEL_GVT_MMIO_HASH_BITS); - unsigned int num_tracked_mmio; + unsigned long num_tracked_mmio; }; struct intel_gvt_firmware { @@ -301,6 +303,8 @@ struct intel_gvt { struct task_struct *service_thread; wait_queue_head_t service_thread_wq; unsigned long service_request; + + struct dentry *debugfs_root; }; static inline struct intel_gvt *to_gvt(struct drm_i915_private *i915) @@ -619,6 +623,12 @@ static inline bool intel_gvt_mmio_has_mode_mask( return gvt->mmio.mmio_attribute[offset >> 2] & F_MODE_MASK; } +int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu); +void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu); +int intel_gvt_debugfs_init(struct intel_gvt *gvt); +void intel_gvt_debugfs_clean(struct intel_gvt *gvt); + + #include "trace.h" #include "mpt.h" diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 5438dafea7a49..408600befc9f4 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -2928,8 +2928,6 @@ int intel_gvt_setup_mmio_info(struct intel_gvt *gvt) gvt->mmio.mmio_block = mmio_blocks; gvt->mmio.num_mmio_block = ARRAY_SIZE(mmio_blocks); - gvt_dbg_mmio("traced %u virtual mmio registers\n", - gvt->mmio.num_tracked_mmio); return 0; err: intel_gvt_clean_mmio_info(gvt); diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 841b95c6c2313..99dbefadfe919 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -252,6 +252,7 @@ void intel_gvt_destroy_vgpu(struct intel_vgpu *vgpu) WARN(vgpu->active, "vGPU is still active!\n"); + intel_gvt_debugfs_remove_vgpu(vgpu); idr_remove(&gvt->vgpu_idr, vgpu->id); intel_vgpu_clean_sched_policy(vgpu); intel_vgpu_clean_submission(vgpu); @@ -378,10 +379,16 @@ static struct intel_vgpu *__intel_gvt_create_vgpu(struct intel_gvt *gvt, if (ret) goto out_clean_submission; + ret = intel_gvt_debugfs_add_vgpu(vgpu); + if (ret) + goto out_clean_sched_policy; + mutex_unlock(&gvt->lock); return vgpu; +out_clean_sched_policy: + intel_vgpu_clean_sched_policy(vgpu); out_clean_submission: intel_vgpu_clean_submission(vgpu); out_clean_display: -- GitLab From 5e86ccefa3c4e2d12acd7f5fc417ea120edbca21 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 26 Sep 2017 15:02:21 +0800 Subject: [PATCH 0289/1471] drm/i915/gvt: Use BIT() to make klockwork happy Replace the plain bit usage with BIT() to make klockwork happy. Cc: Deng Hongyi Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 2801d70579d8c..77da3ecfc4cc6 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -351,7 +351,7 @@ static bool gen8_gtt_test_pse(struct intel_gvt_gtt_entry *e) return false; e->type = get_entry_type(e->type); - if (!(e->val64 & (1 << 7))) + if (!(e->val64 & BIT(7))) return false; e->type = get_pse_type(e->type); @@ -369,12 +369,12 @@ static bool gen8_gtt_test_present(struct intel_gvt_gtt_entry *e) || e->type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) return (e->val64 != 0); else - return (e->val64 & (1 << 0)); + return (e->val64 & BIT(0)); } static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e) { - e->val64 &= ~(1 << 0); + e->val64 &= ~BIT(0); } /* -- GitLab From 58facf8c46160706ccb3b7b3768c0f0e29407548 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Fri, 22 Sep 2017 21:12:03 +0800 Subject: [PATCH 0290/1471] drm/i915/gvt: Refine find_bb_size() Returns the error code if something is wrong and the size of batch buffer is passed through the pointer. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 30 +++++++++++++-------------- 1 file changed, 14 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index cd2681a23fcd5..3355eb4fab753 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -1594,23 +1594,23 @@ static int batch_buffer_needs_scan(struct parser_exec_state *s) return 1; } -static int find_bb_size(struct parser_exec_state *s) +static int find_bb_size(struct parser_exec_state *s, unsigned long *bb_size) { unsigned long gma = 0; struct cmd_info *info; - int bb_size = 0; uint32_t cmd_len = 0; - bool met_bb_end = false; + bool bb_end = false; struct intel_vgpu *vgpu = s->vgpu; u32 cmd; + *bb_size = 0; + /* get the start gm address of the batch buffer */ gma = get_gma_bb_from_cmd(s, 1); if (gma == INTEL_GVT_INVALID_ADDR) return -EFAULT; cmd = cmd_val(s, 0); - info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id); if (info == NULL) { gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x\n", @@ -1629,20 +1629,18 @@ static int find_bb_size(struct parser_exec_state *s) } if (info->opcode == OP_MI_BATCH_BUFFER_END) { - met_bb_end = true; + bb_end = true; } else if (info->opcode == OP_MI_BATCH_BUFFER_START) { - if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0) { + if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0) /* chained batch buffer */ - met_bb_end = true; - } + bb_end = true; } cmd_len = get_cmd_length(info, cmd) << 2; - bb_size += cmd_len; + *bb_size += cmd_len; gma += cmd_len; + } while (!bb_end); - } while (!met_bb_end); - - return bb_size; + return 0; } static int perform_bb_shadow(struct parser_exec_state *s) @@ -1650,7 +1648,7 @@ static int perform_bb_shadow(struct parser_exec_state *s) struct intel_shadow_bb_entry *entry_obj; struct intel_vgpu *vgpu = s->vgpu; unsigned long gma = 0; - int bb_size; + unsigned long bb_size; void *dst = NULL; int ret = 0; @@ -1660,9 +1658,9 @@ static int perform_bb_shadow(struct parser_exec_state *s) return -EFAULT; /* get the size of the batch buffer */ - bb_size = find_bb_size(s); - if (bb_size < 0) - return bb_size; + ret = find_bb_size(s, &bb_size); + if (ret) + return ret; /* allocate shadow batch buffer */ entry_obj = kmalloc(sizeof(*entry_obj), GFP_KERNEL); -- GitLab From f52c380a48f527930c86ea6fd7242873c93ba682 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sun, 24 Sep 2017 21:53:03 +0800 Subject: [PATCH 0291/1471] drm/i915/gvt: Refine shadow batch buffer 1) Use standard i915 GEM object sequence to access the shadow batch buffer. 2) Manage i915 vma life cycle to solve one FIXME. v2: - Refine code structure. - Refine the usage of GEM APIs. - Add the missing lock/unlock in release_shadow_batch_buffer. Test on my SKL NuC. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 76 +++++++++++------------ drivers/gpu/drm/i915/gvt/scheduler.c | 88 ++++++++++++++++++--------- drivers/gpu/drm/i915/gvt/scheduler.h | 7 ++- 3 files changed, 100 insertions(+), 71 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 3355eb4fab753..0204430836810 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -1645,11 +1645,10 @@ static int find_bb_size(struct parser_exec_state *s, unsigned long *bb_size) static int perform_bb_shadow(struct parser_exec_state *s) { - struct intel_shadow_bb_entry *entry_obj; struct intel_vgpu *vgpu = s->vgpu; + struct intel_vgpu_shadow_bb *bb; unsigned long gma = 0; unsigned long bb_size; - void *dst = NULL; int ret = 0; /* get the start gm address of the batch buffer */ @@ -1657,51 +1656,51 @@ static int perform_bb_shadow(struct parser_exec_state *s) if (gma == INTEL_GVT_INVALID_ADDR) return -EFAULT; - /* get the size of the batch buffer */ ret = find_bb_size(s, &bb_size); if (ret) return ret; - /* allocate shadow batch buffer */ - entry_obj = kmalloc(sizeof(*entry_obj), GFP_KERNEL); - if (entry_obj == NULL) + bb = kzalloc(sizeof(*bb), GFP_KERNEL); + if (!bb) return -ENOMEM; - entry_obj->obj = - i915_gem_object_create(s->vgpu->gvt->dev_priv, - roundup(bb_size, PAGE_SIZE)); - if (IS_ERR(entry_obj->obj)) { - ret = PTR_ERR(entry_obj->obj); - goto free_entry; + bb->obj = i915_gem_object_create(s->vgpu->gvt->dev_priv, + roundup(bb_size, PAGE_SIZE)); + if (IS_ERR(bb->obj)) { + ret = PTR_ERR(bb->obj); + goto err_free_bb; } - entry_obj->len = bb_size; - INIT_LIST_HEAD(&entry_obj->list); - dst = i915_gem_object_pin_map(entry_obj->obj, I915_MAP_WB); - if (IS_ERR(dst)) { - ret = PTR_ERR(dst); - goto put_obj; - } + ret = i915_gem_obj_prepare_shmem_write(bb->obj, &bb->clflush); + if (ret) + goto err_free_obj; - ret = i915_gem_object_set_to_cpu_domain(entry_obj->obj, false); - if (ret) { - gvt_vgpu_err("failed to set shadow batch to CPU\n"); - goto unmap_src; + bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB); + if (IS_ERR(bb->va)) { + ret = PTR_ERR(bb->va); + goto err_finish_shmem_access; } - entry_obj->va = dst; - entry_obj->bb_start_cmd_va = s->ip_va; + if (bb->clflush & CLFLUSH_BEFORE) { + drm_clflush_virt_range(bb->va, bb->obj->base.size); + bb->clflush &= ~CLFLUSH_BEFORE; + } - /* copy batch buffer to shadow batch buffer*/ ret = copy_gma_to_hva(s->vgpu, s->vgpu->gtt.ggtt_mm, gma, gma + bb_size, - dst); + bb->va); if (ret < 0) { gvt_vgpu_err("fail to copy guest ring buffer\n"); - goto unmap_src; + ret = -EFAULT; + goto err_unmap; } - list_add(&entry_obj->list, &s->workload->shadow_bb); + INIT_LIST_HEAD(&bb->list); + list_add(&bb->list, &s->workload->shadow_bb); + + bb->accessing = true; + bb->bb_start_cmd_va = s->ip_va; + /* * ip_va saves the virtual address of the shadow batch buffer, while * ip_gma saves the graphics address of the original batch buffer. @@ -1710,17 +1709,17 @@ static int perform_bb_shadow(struct parser_exec_state *s) * buffer's gma in pair. After all, we don't want to pin the shadow * buffer here (too early). */ - s->ip_va = dst; + s->ip_va = bb->va; s->ip_gma = gma; - return 0; - -unmap_src: - i915_gem_object_unpin_map(entry_obj->obj); -put_obj: - i915_gem_object_put(entry_obj->obj); -free_entry: - kfree(entry_obj); +err_unmap: + i915_gem_object_unpin_map(bb->obj); +err_finish_shmem_access: + i915_gem_obj_finish_shmem_access(bb->obj); +err_free_obj: + i915_gem_object_put(bb->obj); +err_free_bb: + kfree(bb); return ret; } @@ -1762,7 +1761,6 @@ static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s) if (ret < 0) return ret; } - return ret; } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 391690c0c28c1..f2d4c90ea1d47 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -325,31 +325,46 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) return ret; } +static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload); + static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload) { struct intel_gvt *gvt = workload->vgpu->gvt; const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd; - struct intel_shadow_bb_entry *entry_obj; + struct intel_vgpu_shadow_bb *bb; + int ret; - /* pin the gem object to ggtt */ - list_for_each_entry(entry_obj, &workload->shadow_bb, list) { - struct i915_vma *vma; + list_for_each_entry(bb, &workload->shadow_bb, list) { + bb->vma = i915_gem_object_ggtt_pin(bb->obj, NULL, 0, 0, 0); + if (IS_ERR(bb->vma)) { + ret = PTR_ERR(bb->vma); + goto err; + } - vma = i915_gem_object_ggtt_pin(entry_obj->obj, NULL, 0, 4, 0); - if (IS_ERR(vma)) - return PTR_ERR(vma); + /* relocate shadow batch buffer */ + bb->bb_start_cmd_va[1] = i915_ggtt_offset(bb->vma); + if (gmadr_bytes == 8) + bb->bb_start_cmd_va[2] = 0; - /* FIXME: we are not tracking our pinned VMA leaving it - * up to the core to fix up the stray pin_count upon - * free. - */ + /* No one is going to touch shadow bb from now on. */ + if (bb->clflush & CLFLUSH_AFTER) { + drm_clflush_virt_range(bb->va, bb->obj->base.size); + bb->clflush &= ~CLFLUSH_AFTER; + } - /* update the relocate gma with shadow batch buffer*/ - entry_obj->bb_start_cmd_va[1] = i915_ggtt_offset(vma); - if (gmadr_bytes == 8) - entry_obj->bb_start_cmd_va[2] = 0; + ret = i915_gem_object_set_to_gtt_domain(bb->obj, false); + if (ret) + goto err; + + i915_gem_obj_finish_shmem_access(bb->obj); + bb->accessing = false; + + i915_vma_move_to_active(bb->vma, workload->req, 0); } return 0; +err: + release_shadow_batch_buffer(workload); + return ret; } static int update_wa_ctx_2_shadow_ctx(struct intel_shadow_wa_ctx *wa_ctx) @@ -410,22 +425,37 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload) { - /* release all the shadow batch buffer */ - if (!list_empty(&workload->shadow_bb)) { - struct intel_shadow_bb_entry *entry_obj = - list_first_entry(&workload->shadow_bb, - struct intel_shadow_bb_entry, - list); - struct intel_shadow_bb_entry *temp; - - list_for_each_entry_safe(entry_obj, temp, &workload->shadow_bb, - list) { - i915_gem_object_unpin_map(entry_obj->obj); - i915_gem_object_put(entry_obj->obj); - list_del(&entry_obj->list); - kfree(entry_obj); + struct intel_vgpu *vgpu = workload->vgpu; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + struct intel_vgpu_shadow_bb *bb, *pos; + + if (list_empty(&workload->shadow_bb)) + return; + + bb = list_first_entry(&workload->shadow_bb, + struct intel_vgpu_shadow_bb, list); + + mutex_lock(&dev_priv->drm.struct_mutex); + + list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) { + if (bb->obj) { + if (bb->accessing) + i915_gem_obj_finish_shmem_access(bb->obj); + + if (bb->va && !IS_ERR(bb->va)) + i915_gem_object_unpin_map(bb->obj); + + if (bb->vma && !IS_ERR(bb->vma)) { + i915_vma_unpin(bb->vma); + i915_vma_close(bb->vma); + } + __i915_gem_object_release_unless_active(bb->obj); } + list_del(&bb->list); + kfree(bb); } + + mutex_unlock(&dev_priv->drm.struct_mutex); } static int prepare_workload(struct intel_vgpu_workload *workload) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.h b/drivers/gpu/drm/i915/gvt/scheduler.h index e0b5730a30187..e4a9f9acd4a95 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.h +++ b/drivers/gpu/drm/i915/gvt/scheduler.h @@ -112,13 +112,14 @@ struct intel_vgpu_workload { struct intel_shadow_wa_ctx wa_ctx; }; -/* Intel shadow batch buffer is a i915 gem object */ -struct intel_shadow_bb_entry { +struct intel_vgpu_shadow_bb { struct list_head list; struct drm_i915_gem_object *obj; + struct i915_vma *vma; void *va; - unsigned long len; u32 *bb_start_cmd_va; + unsigned int clflush; + bool accessing; }; #define workload_q_head(vgpu, ring_id) \ -- GitLab From 7d1e5cdf01789729aff2da4005f51f58b491040c Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Fri, 29 Sep 2017 02:47:55 +0800 Subject: [PATCH 0292/1471] drm/i915/gvt: Factor intel_vgpu_page_track As the data structure of "intel_vgpu_guest_page" will become much heavier in future, it's better to factor out the guest memory page track mechnisim as early as possible. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 154 ++++++++++++++++++-------------- drivers/gpu/drm/i915/gvt/gtt.h | 30 +++---- drivers/gpu/drm/i915/gvt/mmio.c | 32 +++---- drivers/gpu/drm/i915/gvt/mpt.h | 34 +++---- 4 files changed, 136 insertions(+), 114 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 77da3ecfc4cc6..bd3dc209cd892 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -494,7 +494,7 @@ static inline int ppgtt_spt_get_entry( return -EINVAL; ret = ops->get_entry(page_table, e, index, guest, - spt->guest_page.gfn << GTT_PAGE_SHIFT, + spt->guest_page.track.gfn << GTT_PAGE_SHIFT, spt->vgpu); if (ret) return ret; @@ -516,7 +516,7 @@ static inline int ppgtt_spt_set_entry( return -EINVAL; return ops->set_entry(page_table, e, index, guest, - spt->guest_page.gfn << GTT_PAGE_SHIFT, + spt->guest_page.track.gfn << GTT_PAGE_SHIFT, spt->vgpu); } @@ -537,86 +537,101 @@ static inline int ppgtt_spt_set_entry( spt->shadow_page.type, e, index, false) /** - * intel_vgpu_init_guest_page - init a guest page data structure + * intel_vgpu_init_page_track - init a page track data structure * @vgpu: a vGPU - * @p: a guest page data structure + * @t: a page track data structure * @gfn: guest memory page frame number - * @handler: function will be called when target guest memory page has + * @handler: the function will be called when target guest memory page has * been modified. * - * This function is called when user wants to track a guest memory page. + * This function is called when a user wants to prepare a page track data + * structure to track a guest memory page. * * Returns: * Zero on success, negative error code if failed. */ -int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *p, +int intel_vgpu_init_page_track(struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t, unsigned long gfn, int (*handler)(void *, u64, void *, int), void *data) { - INIT_HLIST_NODE(&p->node); + INIT_HLIST_NODE(&t->node); - p->writeprotection = false; - p->gfn = gfn; - p->handler = handler; - p->data = data; - p->oos_page = NULL; - p->write_cnt = 0; + t->tracked = false; + t->gfn = gfn; + t->handler = handler; + t->data = data; - hash_add(vgpu->gtt.guest_page_hash_table, &p->node, p->gfn); + hash_add(vgpu->gtt.tracked_guest_page_hash_table, &t->node, t->gfn); return 0; } -static int detach_oos_page(struct intel_vgpu *vgpu, - struct intel_vgpu_oos_page *oos_page); - /** - * intel_vgpu_clean_guest_page - release the resource owned by guest page data - * structure + * intel_vgpu_clean_page_track - release a page track data structure * @vgpu: a vGPU - * @p: a tracked guest page + * @t: a page track data structure * - * This function is called when user tries to stop tracking a guest memory - * page. + * This function is called before a user frees a page track data structure. */ -void intel_vgpu_clean_guest_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *p) +void intel_vgpu_clean_page_track(struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t) { - if (!hlist_unhashed(&p->node)) - hash_del(&p->node); + if (!hlist_unhashed(&t->node)) + hash_del(&t->node); - if (p->oos_page) - detach_oos_page(vgpu, p->oos_page); - - if (p->writeprotection) - intel_gvt_hypervisor_unset_wp_page(vgpu, p); + if (t->tracked) + intel_gvt_hypervisor_disable_page_track(vgpu, t); } /** - * intel_vgpu_find_guest_page - find a guest page data structure by GFN. + * intel_vgpu_find_tracked_page - find a tracked guest page * @vgpu: a vGPU * @gfn: guest memory page frame number * - * This function is called when emulation logic wants to know if a trapped GFN - * is a tracked guest page. + * This function is called when the emulation layer wants to figure out if a + * trapped GFN is a tracked guest page. * * Returns: - * Pointer to guest page data structure, NULL if failed. + * Pointer to page track data structure, NULL if not found. */ -struct intel_vgpu_guest_page *intel_vgpu_find_guest_page( +struct intel_vgpu_page_track *intel_vgpu_find_tracked_page( struct intel_vgpu *vgpu, unsigned long gfn) { - struct intel_vgpu_guest_page *p; + struct intel_vgpu_page_track *t; - hash_for_each_possible(vgpu->gtt.guest_page_hash_table, - p, node, gfn) { - if (p->gfn == gfn) - return p; + hash_for_each_possible(vgpu->gtt.tracked_guest_page_hash_table, + t, node, gfn) { + if (t->gfn == gfn) + return t; } return NULL; } +static int init_guest_page(struct intel_vgpu *vgpu, + struct intel_vgpu_guest_page *p, + unsigned long gfn, + int (*handler)(void *, u64, void *, int), + void *data) +{ + p->oos_page = NULL; + p->write_cnt = 0; + + return intel_vgpu_init_page_track(vgpu, &p->track, gfn, handler, data); +} + +static int detach_oos_page(struct intel_vgpu *vgpu, + struct intel_vgpu_oos_page *oos_page); + +static void clean_guest_page(struct intel_vgpu *vgpu, + struct intel_vgpu_guest_page *p) +{ + if (p->oos_page) + detach_oos_page(vgpu, p->oos_page); + + intel_vgpu_clean_page_track(vgpu, &p->track); +} + static inline int init_shadow_page(struct intel_vgpu *vgpu, struct intel_vgpu_shadow_page *p, int type) { @@ -664,6 +679,9 @@ static inline struct intel_vgpu_shadow_page *find_shadow_page( return NULL; } +#define page_track_to_guest_page(ptr) \ + container_of(ptr, struct intel_vgpu_guest_page, track) + #define guest_page_to_ppgtt_spt(ptr) \ container_of(ptr, struct intel_vgpu_ppgtt_spt, guest_page) @@ -697,7 +715,7 @@ static void ppgtt_free_shadow_page(struct intel_vgpu_ppgtt_spt *spt) trace_spt_free(spt->vgpu->id, spt, spt->shadow_page.type); clean_shadow_page(spt->vgpu, &spt->shadow_page); - intel_vgpu_clean_guest_page(spt->vgpu, &spt->guest_page); + clean_guest_page(spt->vgpu, &spt->guest_page); list_del_init(&spt->post_shadow_list); free_spt(spt); @@ -713,22 +731,24 @@ static void ppgtt_free_all_shadow_page(struct intel_vgpu *vgpu) ppgtt_free_shadow_page(shadow_page_to_ppgtt_spt(sp)); } -static int ppgtt_handle_guest_write_page_table_bytes(void *gp, +static int ppgtt_handle_guest_write_page_table_bytes( + struct intel_vgpu_guest_page *gpt, u64 pa, void *p_data, int bytes); -static int ppgtt_write_protection_handler(void *gp, u64 pa, +static int ppgtt_write_protection_handler(void *data, u64 pa, void *p_data, int bytes) { - struct intel_vgpu_guest_page *gpt = (struct intel_vgpu_guest_page *)gp; + struct intel_vgpu_page_track *t = data; + struct intel_vgpu_guest_page *p = page_track_to_guest_page(t); int ret; if (bytes != 4 && bytes != 8) return -EINVAL; - if (!gpt->writeprotection) + if (!t->tracked) return -EINVAL; - ret = ppgtt_handle_guest_write_page_table_bytes(gp, + ret = ppgtt_handle_guest_write_page_table_bytes(p, pa, p_data, bytes); if (ret) return ret; @@ -768,7 +788,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_alloc_shadow_page( goto err; } - ret = intel_vgpu_init_guest_page(vgpu, &spt->guest_page, + ret = init_guest_page(vgpu, &spt->guest_page, gfn, ppgtt_write_protection_handler, NULL); if (ret) { gvt_vgpu_err("fail to initialize guest page for spt\n"); @@ -856,7 +876,7 @@ static int ppgtt_invalidate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) int v = atomic_read(&spt->refcount); trace_spt_change(spt->vgpu->id, "die", spt, - spt->guest_page.gfn, spt->shadow_page.type); + spt->guest_page.track.gfn, spt->shadow_page.type); trace_spt_refcount(spt->vgpu->id, "dec", spt, v, (v - 1)); @@ -878,7 +898,7 @@ static int ppgtt_invalidate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) } release: trace_spt_change(spt->vgpu->id, "release", spt, - spt->guest_page.gfn, spt->shadow_page.type); + spt->guest_page.track.gfn, spt->shadow_page.type); ppgtt_free_shadow_page(spt); return 0; fail: @@ -895,6 +915,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry( struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_vgpu_ppgtt_spt *s = NULL; struct intel_vgpu_guest_page *g; + struct intel_vgpu_page_track *t; int ret; if (WARN_ON(!gtt_type_is_pt(get_next_pt_type(we->type)))) { @@ -902,8 +923,9 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry( goto fail; } - g = intel_vgpu_find_guest_page(vgpu, ops->get_pfn(we)); - if (g) { + t = intel_vgpu_find_tracked_page(vgpu, ops->get_pfn(we)); + if (t) { + g = page_track_to_guest_page(t); s = guest_page_to_ppgtt_spt(g); ppgtt_get_shadow_page(s); } else { @@ -915,7 +937,8 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry( goto fail; } - ret = intel_gvt_hypervisor_set_wp_page(vgpu, &s->guest_page); + ret = intel_gvt_hypervisor_enable_page_track(vgpu, + &s->guest_page.track); if (ret) goto fail; @@ -923,7 +946,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_populate_shadow_page_by_guest_entry( if (ret) goto fail; - trace_spt_change(vgpu->id, "new", s, s->guest_page.gfn, + trace_spt_change(vgpu->id, "new", s, s->guest_page.track.gfn, s->shadow_page.type); } return s; @@ -953,7 +976,7 @@ static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) int ret; trace_spt_change(spt->vgpu->id, "born", spt, - spt->guest_page.gfn, spt->shadow_page.type); + spt->guest_page.track.gfn, spt->shadow_page.type); if (gtt_type_is_pte_pt(spt->shadow_page.type)) { for_each_present_guest_entry(spt, &ge, i) { @@ -1082,7 +1105,7 @@ static int sync_oos_page(struct intel_vgpu *vgpu, index++) { ops->get_entry(oos_page->mem, &old, index, false, 0, vgpu); ops->get_entry(NULL, &new, index, true, - oos_page->guest_page->gfn << PAGE_SHIFT, vgpu); + oos_page->guest_page->track.gfn << PAGE_SHIFT, vgpu); if (old.val64 == new.val64 && !test_and_clear_bit(index, spt->post_shadow_bitmap)) @@ -1132,8 +1155,9 @@ static int attach_oos_page(struct intel_vgpu *vgpu, struct intel_gvt *gvt = vgpu->gvt; int ret; - ret = intel_gvt_hypervisor_read_gpa(vgpu, gpt->gfn << GTT_PAGE_SHIFT, - oos_page->mem, GTT_PAGE_SIZE); + ret = intel_gvt_hypervisor_read_gpa(vgpu, + gpt->track.gfn << GTT_PAGE_SHIFT, + oos_page->mem, GTT_PAGE_SIZE); if (ret) return ret; @@ -1152,7 +1176,7 @@ static int ppgtt_set_guest_page_sync(struct intel_vgpu *vgpu, { int ret; - ret = intel_gvt_hypervisor_set_wp_page(vgpu, gpt); + ret = intel_gvt_hypervisor_enable_page_track(vgpu, &gpt->track); if (ret) return ret; @@ -1200,7 +1224,7 @@ static int ppgtt_set_guest_page_oos(struct intel_vgpu *vgpu, gpt, guest_page_to_ppgtt_spt(gpt)->guest_page_type); list_add_tail(&oos_page->vm_list, &vgpu->gtt.oos_page_list_head); - return intel_gvt_hypervisor_unset_wp_page(vgpu, gpt); + return intel_gvt_hypervisor_disable_page_track(vgpu, &gpt->track); } /** @@ -1335,10 +1359,10 @@ int intel_vgpu_flush_post_shadow(struct intel_vgpu *vgpu) return 0; } -static int ppgtt_handle_guest_write_page_table_bytes(void *gp, +static int ppgtt_handle_guest_write_page_table_bytes( + struct intel_vgpu_guest_page *gpt, u64 pa, void *p_data, int bytes) { - struct intel_vgpu_guest_page *gpt = (struct intel_vgpu_guest_page *)gp; struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt); struct intel_vgpu *vgpu = spt->vgpu; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; @@ -2032,7 +2056,7 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu) struct intel_vgpu_gtt *gtt = &vgpu->gtt; struct intel_vgpu_mm *ggtt_mm; - hash_init(gtt->guest_page_hash_table); + hash_init(gtt->tracked_guest_page_hash_table); hash_init(gtt->shadow_page_hash_table); INIT_LIST_HEAD(>t->mm_list_head); diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index 30a4c8d160266..bd9b298b87778 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -193,18 +193,16 @@ struct intel_vgpu_scratch_pt { unsigned long page_mfn; }; - struct intel_vgpu_gtt { struct intel_vgpu_mm *ggtt_mm; unsigned long active_ppgtt_mm_bitmap; struct list_head mm_list_head; DECLARE_HASHTABLE(shadow_page_hash_table, INTEL_GVT_GTT_HASH_BITS); - DECLARE_HASHTABLE(guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS); - atomic_t n_write_protected_guest_page; + DECLARE_HASHTABLE(tracked_guest_page_hash_table, INTEL_GVT_GTT_HASH_BITS); + atomic_t n_tracked_guest_page; struct list_head oos_page_list_head; struct list_head post_shadow_list_head; struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX]; - }; extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu); @@ -228,12 +226,16 @@ struct intel_vgpu_shadow_page { unsigned long mfn; }; -struct intel_vgpu_guest_page { +struct intel_vgpu_page_track { struct hlist_node node; - bool writeprotection; + bool tracked; unsigned long gfn; int (*handler)(void *, u64, void *, int); void *data; +}; + +struct intel_vgpu_guest_page { + struct intel_vgpu_page_track track; unsigned long write_cnt; struct intel_vgpu_oos_page *oos_page; }; @@ -258,22 +260,16 @@ struct intel_vgpu_ppgtt_spt { struct list_head post_shadow_list; }; -int intel_vgpu_init_guest_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *guest_page, +int intel_vgpu_init_page_track(struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t, unsigned long gfn, int (*handler)(void *gp, u64, void *, int), void *data); -void intel_vgpu_clean_guest_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *guest_page); - -int intel_vgpu_set_guest_page_writeprotection(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *guest_page); - -void intel_vgpu_clear_guest_page_writeprotection(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *guest_page); +void intel_vgpu_clean_page_track(struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t); -struct intel_vgpu_guest_page *intel_vgpu_find_guest_page( +struct intel_vgpu_page_track *intel_vgpu_find_tracked_page( struct intel_vgpu *vgpu, unsigned long gfn); int intel_vgpu_sync_oos_pages(struct intel_vgpu *vgpu); diff --git a/drivers/gpu/drm/i915/gvt/mmio.c b/drivers/gpu/drm/i915/gvt/mmio.c index 1e1310f50289a..4ea0feb5f04dc 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.c +++ b/drivers/gpu/drm/i915/gvt/mmio.c @@ -117,18 +117,18 @@ static void failsafe_emulate_mmio_rw(struct intel_vgpu *vgpu, uint64_t pa, else memcpy(pt, p_data, bytes); - } else if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) { - struct intel_vgpu_guest_page *gp; + } else if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) { + struct intel_vgpu_page_track *t; /* Since we enter the failsafe mode early during guest boot, * guest may not have chance to set up its ppgtt table, so * there should not be any wp pages for guest. Keep the wp * related code here in case we need to handle it in furture. */ - gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT); - if (gp) { + t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT); + if (t) { /* remove write protection to prevent furture traps */ - intel_vgpu_clean_guest_page(vgpu, gp); + intel_vgpu_clean_page_track(vgpu, t); if (read) intel_gvt_hypervisor_read_gpa(vgpu, pa, p_data, bytes); @@ -170,17 +170,17 @@ int intel_vgpu_emulate_mmio_read(struct intel_vgpu *vgpu, uint64_t pa, return ret; } - if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) { - struct intel_vgpu_guest_page *gp; + if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) { + struct intel_vgpu_page_track *t; - gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT); - if (gp) { + t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT); + if (t) { ret = intel_gvt_hypervisor_read_gpa(vgpu, pa, p_data, bytes); if (ret) { gvt_vgpu_err("guest page read error %d, " "gfn 0x%lx, pa 0x%llx, var 0x%x, len %d\n", - ret, gp->gfn, pa, *(u32 *)p_data, + ret, t->gfn, pa, *(u32 *)p_data, bytes); } mutex_unlock(&gvt->lock); @@ -267,17 +267,17 @@ int intel_vgpu_emulate_mmio_write(struct intel_vgpu *vgpu, uint64_t pa, return ret; } - if (atomic_read(&vgpu->gtt.n_write_protected_guest_page)) { - struct intel_vgpu_guest_page *gp; + if (atomic_read(&vgpu->gtt.n_tracked_guest_page)) { + struct intel_vgpu_page_track *t; - gp = intel_vgpu_find_guest_page(vgpu, pa >> PAGE_SHIFT); - if (gp) { - ret = gp->handler(gp, pa, p_data, bytes); + t = intel_vgpu_find_tracked_page(vgpu, pa >> PAGE_SHIFT); + if (t) { + ret = t->handler(t, pa, p_data, bytes); if (ret) { gvt_err("guest page write error %d, " "gfn 0x%lx, pa 0x%llx, " "var 0x%x, len %d\n", - ret, gp->gfn, pa, + ret, t->gfn, pa, *(u32 *)p_data, bytes); } mutex_unlock(&gvt->lock); diff --git a/drivers/gpu/drm/i915/gvt/mpt.h b/drivers/gpu/drm/i915/gvt/mpt.h index f0e5487e66886..c436e20ea59ef 100644 --- a/drivers/gpu/drm/i915/gvt/mpt.h +++ b/drivers/gpu/drm/i915/gvt/mpt.h @@ -154,51 +154,53 @@ static inline unsigned long intel_gvt_hypervisor_virt_to_mfn(void *p) } /** - * intel_gvt_hypervisor_set_wp_page - set a guest page to write-protected + * intel_gvt_hypervisor_enable - set a guest page to write-protected * @vgpu: a vGPU - * @p: intel_vgpu_guest_page + * @t: page track data structure * * Returns: * Zero on success, negative error code if failed. */ -static inline int intel_gvt_hypervisor_set_wp_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *p) +static inline int intel_gvt_hypervisor_enable_page_track( + struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t) { int ret; - if (p->writeprotection) + if (t->tracked) return 0; - ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, p->gfn); + ret = intel_gvt_host.mpt->set_wp_page(vgpu->handle, t->gfn); if (ret) return ret; - p->writeprotection = true; - atomic_inc(&vgpu->gtt.n_write_protected_guest_page); + t->tracked = true; + atomic_inc(&vgpu->gtt.n_tracked_guest_page); return 0; } /** - * intel_gvt_hypervisor_unset_wp_page - remove the write-protection of a + * intel_gvt_hypervisor_disable_page_track - remove the write-protection of a * guest page * @vgpu: a vGPU - * @p: intel_vgpu_guest_page + * @t: page track data structure * * Returns: * Zero on success, negative error code if failed. */ -static inline int intel_gvt_hypervisor_unset_wp_page(struct intel_vgpu *vgpu, - struct intel_vgpu_guest_page *p) +static inline int intel_gvt_hypervisor_disable_page_track( + struct intel_vgpu *vgpu, + struct intel_vgpu_page_track *t) { int ret; - if (!p->writeprotection) + if (!t->tracked) return 0; - ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, p->gfn); + ret = intel_gvt_host.mpt->unset_wp_page(vgpu->handle, t->gfn); if (ret) return ret; - p->writeprotection = false; - atomic_dec(&vgpu->gtt.n_write_protected_guest_page); + t->tracked = false; + atomic_dec(&vgpu->gtt.n_tracked_guest_page); return 0; } -- GitLab From 62a6a53786fc4b4e7543cc63b704dbb3f7df4c0f Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Sat, 30 Sep 2017 17:42:20 +0800 Subject: [PATCH 0293/1471] drm/i915/gvt: Export intel_gvt_render_mmio_to_ring_id() Since many emulation logic needs to convert the offset of ring registers into ring id, we export it for other caller which might need it. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 21 +++++++++++++++------ drivers/gpu/drm/i915/gvt/mmio.h | 2 ++ 2 files changed, 17 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 408600befc9f4..bb16ef5c85104 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -137,17 +137,26 @@ static int new_mmio_info(struct intel_gvt *gvt, return 0; } -static int render_mmio_to_ring_id(struct intel_gvt *gvt, unsigned int reg) +/** + * intel_gvt_render_mmio_to_ring_id - convert a mmio offset into ring id + * @gvt: a GVT device + * @offset: register offset + * + * Returns: + * Ring ID on success, negative error code if failed. + */ +int intel_gvt_render_mmio_to_ring_id(struct intel_gvt *gvt, + unsigned int offset) { enum intel_engine_id id; struct intel_engine_cs *engine; - reg &= ~GENMASK(11, 0); + offset &= ~GENMASK(11, 0); for_each_engine(engine, gvt->dev_priv, id) { - if (engine->mmio_base == reg) + if (engine->mmio_base == offset) return id; } - return -1; + return -ENODEV; } #define offset_to_fence_num(offset) \ @@ -1445,7 +1454,7 @@ static int mmio_read_from_hw(struct intel_vgpu *vgpu, static int elsp_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { - int ring_id = render_mmio_to_ring_id(vgpu->gvt, offset); + int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset); struct intel_vgpu_execlist *execlist; u32 data = *(u32 *)p_data; int ret = 0; @@ -1473,7 +1482,7 @@ static int ring_mode_mmio_write(struct intel_vgpu *vgpu, unsigned int offset, { struct intel_vgpu_submission *s = &vgpu->submission; u32 data = *(u32 *)p_data; - int ring_id = render_mmio_to_ring_id(vgpu->gvt, offset); + int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset); bool enable_execlist; int ret; diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h index 32cd64ddad266..dbc04ad2c7a1e 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.h +++ b/drivers/gpu/drm/i915/gvt/mmio.h @@ -65,6 +65,8 @@ struct intel_gvt_mmio_info { struct hlist_node node; }; +int intel_gvt_render_mmio_to_ring_id(struct intel_gvt *gvt, + unsigned int reg); unsigned long intel_gvt_get_device_type(struct intel_gvt *gvt); bool intel_gvt_match_device(struct intel_gvt *gvt, unsigned long device); -- GitLab From 9556e118889293f6d5d226b64688ee2adfd8964c Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 10 Oct 2017 13:51:32 +0800 Subject: [PATCH 0294/1471] drm/i915/gvt: Use I915_GTT_PAGE_SIZE As there is already an I915_GTT_PAGE_SIZE marco in i915, let GVT-g use it as well. Also this patch re-names some GTT marcos with additional prefix. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 13 +++---- drivers/gpu/drm/i915/gvt/gtt.c | 51 ++++++++++++++------------- drivers/gpu/drm/i915/gvt/gtt.h | 7 ++-- drivers/gpu/drm/i915/gvt/reg.h | 3 +- drivers/gpu/drm/i915/gvt/scheduler.c | 14 ++++---- 5 files changed, 45 insertions(+), 43 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 0204430836810..8c8514ee73ac0 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -1396,7 +1396,7 @@ static inline int cmd_address_audit(struct parser_exec_state *s, } if (index_mode) { - if (guest_gma >= GTT_PAGE_SIZE / sizeof(u64)) { + if (guest_gma >= I915_GTT_PAGE_SIZE / sizeof(u64)) { ret = -EFAULT; goto err; } @@ -1563,10 +1563,10 @@ static int copy_gma_to_hva(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm, return -EFAULT; } - offset = gma & (GTT_PAGE_SIZE - 1); + offset = gma & (I915_GTT_PAGE_SIZE - 1); - copy_len = (end_gma - gma) >= (GTT_PAGE_SIZE - offset) ? - GTT_PAGE_SIZE - offset : end_gma - gma; + copy_len = (end_gma - gma) >= (I915_GTT_PAGE_SIZE - offset) ? + I915_GTT_PAGE_SIZE - offset : end_gma - gma; intel_gvt_hypervisor_read_gpa(vgpu, gpa, va + len, copy_len); @@ -2540,7 +2540,7 @@ static int scan_workload(struct intel_vgpu_workload *workload) int ret = 0; /* ring base is page aligned */ - if (WARN_ON(!IS_ALIGNED(workload->rb_start, GTT_PAGE_SIZE))) + if (WARN_ON(!IS_ALIGNED(workload->rb_start, I915_GTT_PAGE_SIZE))) return -EINVAL; gma_head = workload->rb_start + workload->rb_head; @@ -2589,7 +2589,8 @@ static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) wa_ctx); /* ring base is page aligned */ - if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma, GTT_PAGE_SIZE))) + if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma, + I915_GTT_PAGE_SIZE))) return -EINVAL; ring_tail = wa_ctx->indirect_ctx.size + 3 * sizeof(uint32_t); diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index bd3dc209cd892..c9cbd71be4c58 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -94,12 +94,12 @@ int intel_gvt_ggtt_index_g2h(struct intel_vgpu *vgpu, unsigned long g_index, u64 h_addr; int ret; - ret = intel_gvt_ggtt_gmadr_g2h(vgpu, g_index << GTT_PAGE_SHIFT, + ret = intel_gvt_ggtt_gmadr_g2h(vgpu, g_index << I915_GTT_PAGE_SHIFT, &h_addr); if (ret) return ret; - *h_index = h_addr >> GTT_PAGE_SHIFT; + *h_index = h_addr >> I915_GTT_PAGE_SHIFT; return 0; } @@ -109,12 +109,12 @@ int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index, u64 g_addr; int ret; - ret = intel_gvt_ggtt_gmadr_h2g(vgpu, h_index << GTT_PAGE_SHIFT, + ret = intel_gvt_ggtt_gmadr_h2g(vgpu, h_index << I915_GTT_PAGE_SHIFT, &g_addr); if (ret) return ret; - *g_index = g_addr >> GTT_PAGE_SHIFT; + *g_index = g_addr >> I915_GTT_PAGE_SHIFT; return 0; } @@ -382,7 +382,7 @@ static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e) */ static unsigned long gma_to_ggtt_pte_index(unsigned long gma) { - unsigned long x = (gma >> GTT_PAGE_SHIFT); + unsigned long x = (gma >> I915_GTT_PAGE_SHIFT); trace_gma_index(__func__, gma, x); return x; @@ -494,7 +494,7 @@ static inline int ppgtt_spt_get_entry( return -EINVAL; ret = ops->get_entry(page_table, e, index, guest, - spt->guest_page.track.gfn << GTT_PAGE_SHIFT, + spt->guest_page.track.gfn << I915_GTT_PAGE_SHIFT, spt->vgpu); if (ret) return ret; @@ -516,7 +516,7 @@ static inline int ppgtt_spt_set_entry( return -EINVAL; return ops->set_entry(page_table, e, index, guest, - spt->guest_page.track.gfn << GTT_PAGE_SHIFT, + spt->guest_page.track.gfn << I915_GTT_PAGE_SHIFT, spt->vgpu); } @@ -649,7 +649,7 @@ static inline int init_shadow_page(struct intel_vgpu *vgpu, INIT_HLIST_NODE(&p->node); - p->mfn = daddr >> GTT_PAGE_SHIFT; + p->mfn = daddr >> I915_GTT_PAGE_SHIFT; hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn); return 0; } @@ -659,7 +659,7 @@ static inline void clean_shadow_page(struct intel_vgpu *vgpu, { struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev; - dma_unmap_page(kdev, p->mfn << GTT_PAGE_SHIFT, 4096, + dma_unmap_page(kdev, p->mfn << I915_GTT_PAGE_SHIFT, 4096, PCI_DMA_BIDIRECTIONAL); if (!hlist_unhashed(&p->node)) @@ -818,7 +818,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_find_shadow_page( ((spt)->vgpu->gvt->device_info.gtt_entry_size_shift) #define pt_entries(spt) \ - (GTT_PAGE_SIZE >> pt_entry_size_shift(spt)) + (I915_GTT_PAGE_SIZE >> pt_entry_size_shift(spt)) #define for_each_present_guest_entry(spt, e, i) \ for (i = 0; i < pt_entries(spt); i++) \ @@ -1101,8 +1101,8 @@ static int sync_oos_page(struct intel_vgpu *vgpu, old.type = new.type = get_entry_type(spt->guest_page_type); old.val64 = new.val64 = 0; - for (index = 0; index < (GTT_PAGE_SIZE >> info->gtt_entry_size_shift); - index++) { + for (index = 0; index < (I915_GTT_PAGE_SIZE >> + info->gtt_entry_size_shift); index++) { ops->get_entry(oos_page->mem, &old, index, false, 0, vgpu); ops->get_entry(NULL, &new, index, true, oos_page->guest_page->track.gfn << PAGE_SHIFT, vgpu); @@ -1156,8 +1156,8 @@ static int attach_oos_page(struct intel_vgpu *vgpu, int ret; ret = intel_gvt_hypervisor_read_gpa(vgpu, - gpt->track.gfn << GTT_PAGE_SHIFT, - oos_page->mem, GTT_PAGE_SIZE); + gpt->track.gfn << I915_GTT_PAGE_SHIFT, + oos_page->mem, I915_GTT_PAGE_SIZE); if (ret) return ret; @@ -1439,7 +1439,7 @@ static int gen8_mm_alloc_page_table(struct intel_vgpu_mm *mm) mm->shadow_page_table = mem + mm->page_table_entry_size; } else if (mm->type == INTEL_GVT_MM_GGTT) { mm->page_table_entry_cnt = - (gvt_ggtt_gm_sz(gvt) >> GTT_PAGE_SHIFT); + (gvt_ggtt_gm_sz(gvt) >> I915_GTT_PAGE_SHIFT); mm->page_table_entry_size = mm->page_table_entry_cnt * info->gtt_entry_size; mem = vzalloc(mm->page_table_entry_size); @@ -1761,8 +1761,8 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma) gma_ops->gma_to_ggtt_pte_index(gma)); if (ret) goto err; - gpa = (pte_ops->get_pfn(&e) << GTT_PAGE_SHIFT) - + (gma & ~GTT_PAGE_MASK); + gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT) + + (gma & ~I915_GTT_PAGE_MASK); trace_gma_translate(vgpu->id, "ggtt", 0, 0, gma, gpa); return gpa; @@ -1814,8 +1814,8 @@ unsigned long intel_vgpu_gma_to_gpa(struct intel_vgpu_mm *mm, unsigned long gma) } } - gpa = (pte_ops->get_pfn(&e) << GTT_PAGE_SHIFT) - + (gma & ~GTT_PAGE_MASK); + gpa = (pte_ops->get_pfn(&e) << I915_GTT_PAGE_SHIFT) + + (gma & ~I915_GTT_PAGE_MASK); trace_gma_translate(vgpu->id, "ppgtt", 0, mm->page_table_level, gma, gpa); @@ -1883,7 +1883,7 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, if (bytes != 4 && bytes != 8) return -EINVAL; - gma = g_gtt_index << GTT_PAGE_SHIFT; + gma = g_gtt_index << I915_GTT_PAGE_SHIFT; /* the VM may configure the whole GM space when ballooning is used */ if (!vgpu_gmadr_is_valid(vgpu, gma)) @@ -1946,7 +1946,7 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, { struct intel_vgpu_gtt *gtt = &vgpu->gtt; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; - int page_entry_num = GTT_PAGE_SIZE >> + int page_entry_num = I915_GTT_PAGE_SIZE >> vgpu->gvt->device_info.gtt_entry_size_shift; void *scratch_pt; int i; @@ -1970,7 +1970,7 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, return -ENOMEM; } gtt->scratch_pt[type].page_mfn = - (unsigned long)(daddr >> GTT_PAGE_SHIFT); + (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT); gtt->scratch_pt[type].page = virt_to_page(scratch_pt); gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n", vgpu->id, type, gtt->scratch_pt[type].page_mfn); @@ -2013,7 +2013,7 @@ static int release_scratch_page_tree(struct intel_vgpu *vgpu) for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { if (vgpu->gtt.scratch_pt[i].page != NULL) { daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn << - GTT_PAGE_SHIFT); + I915_GTT_PAGE_SHIFT); dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); __free_page(vgpu->gtt.scratch_pt[i].page); vgpu->gtt.scratch_pt[i].page = NULL; @@ -2310,7 +2310,8 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) return -ENOMEM; } gvt->gtt.scratch_ggtt_page = virt_to_page(page); - gvt->gtt.scratch_ggtt_mfn = (unsigned long)(daddr >> GTT_PAGE_SHIFT); + gvt->gtt.scratch_ggtt_mfn = (unsigned long)(daddr >> + I915_GTT_PAGE_SHIFT); if (enable_out_of_sync) { ret = setup_spt_oos(gvt); @@ -2337,7 +2338,7 @@ void intel_gvt_clean_gtt(struct intel_gvt *gvt) { struct device *dev = &gvt->dev_priv->drm.pdev->dev; dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_ggtt_mfn << - GTT_PAGE_SHIFT); + I915_GTT_PAGE_SHIFT); dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index bd9b298b87778..80b2deb097ad9 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -34,9 +34,8 @@ #ifndef _GVT_GTT_H_ #define _GVT_GTT_H_ -#define GTT_PAGE_SHIFT 12 -#define GTT_PAGE_SIZE (1UL << GTT_PAGE_SHIFT) -#define GTT_PAGE_MASK (~(GTT_PAGE_SIZE-1)) +#define I915_GTT_PAGE_SHIFT 12 +#define I915_GTT_PAGE_MASK (~(I915_GTT_PAGE_SIZE - 1)) struct intel_vgpu_mm; @@ -245,7 +244,7 @@ struct intel_vgpu_oos_page { struct list_head list; struct list_head vm_list; int id; - unsigned char mem[GTT_PAGE_SIZE]; + unsigned char mem[I915_GTT_PAGE_SIZE]; }; #define GTT_ENTRY_NUM_IN_ONE_PAGE 512 diff --git a/drivers/gpu/drm/i915/gvt/reg.h b/drivers/gpu/drm/i915/gvt/reg.h index 83f2f63d7eeba..d4f7ce6dc1d73 100644 --- a/drivers/gpu/drm/i915/gvt/reg.h +++ b/drivers/gpu/drm/i915/gvt/reg.h @@ -74,6 +74,7 @@ #define RB_HEAD_OFF_MASK ((1U << 21) - (1U << 2)) #define RB_TAIL_OFF_MASK ((1U << 21) - (1U << 3)) #define RB_TAIL_SIZE_MASK ((1U << 21) - (1U << 12)) -#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + GTT_PAGE_SIZE) +#define _RING_CTL_BUF_SIZE(ctl) (((ctl) & RB_TAIL_SIZE_MASK) + \ + I915_GTT_PAGE_SIZE) #endif diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index f2d4c90ea1d47..7a1ffaa9ae06d 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -81,7 +81,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) while (i < context_page_num) { context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, (u32)((workload->ctx_desc.lrca + i) << - GTT_PAGE_SHIFT)); + I915_GTT_PAGE_SHIFT)); if (context_gpa == INTEL_GVT_INVALID_ADDR) { gvt_vgpu_err("Invalid guest context descriptor\n"); return -EFAULT; @@ -90,7 +90,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i); dst = kmap(page); intel_gvt_hypervisor_read_gpa(vgpu, context_gpa, dst, - GTT_PAGE_SIZE); + I915_GTT_PAGE_SIZE); kunmap(page); i++; } @@ -120,7 +120,7 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload) sizeof(*shadow_ring_context), (void *)shadow_ring_context + sizeof(*shadow_ring_context), - GTT_PAGE_SIZE - sizeof(*shadow_ring_context)); + I915_GTT_PAGE_SIZE - sizeof(*shadow_ring_context)); kunmap(page); return 0; @@ -635,7 +635,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload) while (i < context_page_num) { context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, (u32)((workload->ctx_desc.lrca + i) << - GTT_PAGE_SHIFT)); + I915_GTT_PAGE_SHIFT)); if (context_gpa == INTEL_GVT_INVALID_ADDR) { gvt_vgpu_err("invalid guest context descriptor\n"); return; @@ -644,7 +644,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload) page = i915_gem_object_get_page(ctx_obj, LRC_HEADER_PAGES + i); src = kmap(page); intel_gvt_hypervisor_write_gpa(vgpu, context_gpa, src, - GTT_PAGE_SIZE); + I915_GTT_PAGE_SIZE); kunmap(page); i++; } @@ -669,7 +669,7 @@ static void update_guest_context(struct intel_vgpu_workload *workload) sizeof(*shadow_ring_context), (void *)shadow_ring_context + sizeof(*shadow_ring_context), - GTT_PAGE_SIZE - sizeof(*shadow_ring_context)); + I915_GTT_PAGE_SIZE - sizeof(*shadow_ring_context)); kunmap(page); } @@ -1198,7 +1198,7 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id, int ret; ring_context_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, - (u32)((desc->lrca + 1) << GTT_PAGE_SHIFT)); + (u32)((desc->lrca + 1) << I915_GTT_PAGE_SHIFT)); if (ring_context_gpa == INTEL_GVT_INVALID_ADDR) { gvt_vgpu_err("invalid guest context LRCA: %x\n", desc->lrca); return ERR_PTR(-EINVAL); -- GitLab From 22115cef0869b7d08cb2aa60a65d4d978598644c Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 10 Oct 2017 14:34:11 +0800 Subject: [PATCH 0295/1471] drm/i915/gvt: Let the caller choose if a shadow page should be put into hash table As we want to re-use intel_vgpu_shadow_page in buidling scrach page table and we don't want to put scrach page table page into hash table, a new param is introduced to give the caller a choice to decide if a shadow page should be put into hash table. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 25 +++++++++++++------------ drivers/gpu/drm/i915/gvt/gtt.h | 4 ++-- 2 files changed, 15 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index c9cbd71be4c58..912d63b2ff447 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -633,7 +633,7 @@ static void clean_guest_page(struct intel_vgpu *vgpu, } static inline int init_shadow_page(struct intel_vgpu *vgpu, - struct intel_vgpu_shadow_page *p, int type) + struct intel_vgpu_shadow_page *p, int type, bool hash) { struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev; dma_addr_t daddr; @@ -650,7 +650,8 @@ static inline int init_shadow_page(struct intel_vgpu *vgpu, INIT_HLIST_NODE(&p->node); p->mfn = daddr >> I915_GTT_PAGE_SHIFT; - hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn); + if (hash) + hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn); return 0; } @@ -782,7 +783,7 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_alloc_shadow_page( * TODO: guest page type may be different with shadow page type, * when we support PSE page in future. */ - ret = init_shadow_page(vgpu, &spt->shadow_page, type); + ret = init_shadow_page(vgpu, &spt->shadow_page, type, true); if (ret) { gvt_vgpu_err("fail to initialize shadow page for spt\n"); goto err; @@ -1902,11 +1903,11 @@ static int emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, * update the entry in this situation p2m will fail * settting the shadow entry to point to a scratch page */ - ops->set_pfn(&m, gvt->gtt.scratch_ggtt_mfn); + ops->set_pfn(&m, gvt->gtt.scratch_mfn); } } else { m = e; - ops->set_pfn(&m, gvt->gtt.scratch_ggtt_mfn); + ops->set_pfn(&m, gvt->gtt.scratch_mfn); } ggtt_set_shadow_entry(ggtt_mm, &m, g_gtt_index); @@ -2309,16 +2310,16 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) __free_page(virt_to_page(page)); return -ENOMEM; } - gvt->gtt.scratch_ggtt_page = virt_to_page(page); - gvt->gtt.scratch_ggtt_mfn = (unsigned long)(daddr >> - I915_GTT_PAGE_SHIFT); + + gvt->gtt.scratch_page = virt_to_page(page); + gvt->gtt.scratch_mfn = (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT); if (enable_out_of_sync) { ret = setup_spt_oos(gvt); if (ret) { gvt_err("fail to initialize SPT oos\n"); dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); - __free_page(gvt->gtt.scratch_ggtt_page); + __free_page(gvt->gtt.scratch_page); return ret; } } @@ -2337,12 +2338,12 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) void intel_gvt_clean_gtt(struct intel_gvt *gvt) { struct device *dev = &gvt->dev_priv->drm.pdev->dev; - dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_ggtt_mfn << + dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_mfn << I915_GTT_PAGE_SHIFT); dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); - __free_page(gvt->gtt.scratch_ggtt_page); + __free_page(gvt->gtt.scratch_page); if (enable_out_of_sync) clean_spt_oos(gvt); @@ -2368,7 +2369,7 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu) memset(&e, 0, sizeof(struct intel_gvt_gtt_entry)); e.type = GTT_TYPE_GGTT_PTE; - ops->set_pfn(&e, gvt->gtt.scratch_ggtt_mfn); + ops->set_pfn(&e, gvt->gtt.scratch_mfn); e.val64 |= _PAGE_PRESENT; index = vgpu_aperture_gmadr_base(vgpu) >> PAGE_SHIFT; diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index 80b2deb097ad9..ab590278dcf1d 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -85,8 +85,8 @@ struct intel_gvt_gtt { struct list_head oos_page_free_list_head; struct list_head mm_lru_list_head; - struct page *scratch_ggtt_page; - unsigned long scratch_ggtt_mfn; + struct page *scratch_page; + unsigned long scratch_mfn; }; enum { -- GitLab From 7422064883398612b5921bf26b15bd9b0dd6c325 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 10 Oct 2017 17:14:13 +0800 Subject: [PATCH 0296/1471] drm/i915/gvt: Fix a bug of unexpectedly clear scratch page table During a vGPU reset, the scratch page table shouldn't be cleared, what needs to be cleared should be the scratch page. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 912d63b2ff447..d47b8b100ac28 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -2395,8 +2395,6 @@ void intel_vgpu_reset_ggtt(struct intel_vgpu *vgpu) */ void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu) { - int i; - ppgtt_free_all_shadow_page(vgpu); /* Shadow pages are only created when there is no page @@ -2406,11 +2404,4 @@ void intel_vgpu_reset_gtt(struct intel_vgpu *vgpu) intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_PPGTT); intel_vgpu_reset_ggtt(vgpu); - - /* clear scratch page for security */ - for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { - if (vgpu->gtt.scratch_pt[i].page != NULL) - memset(page_address(vgpu->gtt.scratch_pt[i].page), - 0, PAGE_SIZE); - } } -- GitLab From 054f4eba2a2985b1db43353b7b5ce90e96cf9bb9 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 10 Oct 2017 17:19:30 +0800 Subject: [PATCH 0297/1471] drm/i915/gvt: Introduce page table type of current level in GTT type enumerations Need to figure out page table type of current level by GTT entry type during getting a scratch page table entry. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index d47b8b100ac28..74be5e628310a 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -156,13 +156,15 @@ int intel_gvt_ggtt_h2g_index(struct intel_vgpu *vgpu, unsigned long h_index, struct gtt_type_table_entry { int entry_type; + int pt_type; int next_pt_type; int pse_entry_type; }; -#define GTT_TYPE_TABLE_ENTRY(type, e_type, npt_type, pse_type) \ +#define GTT_TYPE_TABLE_ENTRY(type, e_type, cpt_type, npt_type, pse_type) \ [type] = { \ .entry_type = e_type, \ + .pt_type = cpt_type, \ .next_pt_type = npt_type, \ .pse_entry_type = pse_type, \ } @@ -170,55 +172,68 @@ struct gtt_type_table_entry { static struct gtt_type_table_entry gtt_type_table[] = { GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L4_ENTRY, GTT_TYPE_PPGTT_ROOT_L4_ENTRY, + GTT_TYPE_INVALID, GTT_TYPE_PPGTT_PML4_PT, GTT_TYPE_INVALID), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_PT, GTT_TYPE_PPGTT_PML4_ENTRY, + GTT_TYPE_PPGTT_PML4_PT, GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_INVALID), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PML4_ENTRY, GTT_TYPE_PPGTT_PML4_ENTRY, + GTT_TYPE_PPGTT_PML4_PT, GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_INVALID), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_PPGTT_PDP_ENTRY, + GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PTE_1G_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_ROOT_L3_ENTRY, GTT_TYPE_PPGTT_ROOT_L3_ENTRY, + GTT_TYPE_INVALID, GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PTE_1G_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDP_ENTRY, GTT_TYPE_PPGTT_PDP_ENTRY, + GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PTE_1G_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PDE_ENTRY, + GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PTE_PT, GTT_TYPE_PPGTT_PTE_2M_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PDE_ENTRY, GTT_TYPE_PPGTT_PDE_ENTRY, + GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_PPGTT_PTE_PT, GTT_TYPE_PPGTT_PTE_2M_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_PT, GTT_TYPE_PPGTT_PTE_4K_ENTRY, + GTT_TYPE_PPGTT_PTE_PT, GTT_TYPE_INVALID, GTT_TYPE_INVALID), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_4K_ENTRY, GTT_TYPE_PPGTT_PTE_4K_ENTRY, + GTT_TYPE_PPGTT_PTE_PT, GTT_TYPE_INVALID, GTT_TYPE_INVALID), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_2M_ENTRY, GTT_TYPE_PPGTT_PDE_ENTRY, + GTT_TYPE_PPGTT_PDE_PT, GTT_TYPE_INVALID, GTT_TYPE_PPGTT_PTE_2M_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_PPGTT_PTE_1G_ENTRY, GTT_TYPE_PPGTT_PDP_ENTRY, + GTT_TYPE_PPGTT_PDP_PT, GTT_TYPE_INVALID, GTT_TYPE_PPGTT_PTE_1G_ENTRY), GTT_TYPE_TABLE_ENTRY(GTT_TYPE_GGTT_PTE, GTT_TYPE_GGTT_PTE, GTT_TYPE_INVALID, + GTT_TYPE_INVALID, GTT_TYPE_INVALID), }; @@ -227,6 +242,11 @@ static inline int get_next_pt_type(int type) return gtt_type_table[type].next_pt_type; } +static inline int get_pt_type(int type) +{ + return gtt_type_table[type].pt_type; +} + static inline int get_entry_type(int type) { return gtt_type_table[type].entry_type; -- GitLab From 655c64efe36f199bea16f9ba7388c479d5feed5f Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 10 Oct 2017 17:24:26 +0800 Subject: [PATCH 0298/1471] drm/i915/gvt: Introduce ops->set_present() We need ops->set_present() during generating a new scratch page table entry. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 6 ++++++ drivers/gpu/drm/i915/gvt/gtt.h | 1 + 2 files changed, 7 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 74be5e628310a..3d6008b116e57 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -397,6 +397,11 @@ static void gtt_entry_clear_present(struct intel_gvt_gtt_entry *e) e->val64 &= ~BIT(0); } +static void gtt_entry_set_present(struct intel_gvt_gtt_entry *e) +{ + e->val64 |= BIT(0); +} + /* * Per-platform GMA routines. */ @@ -426,6 +431,7 @@ static struct intel_gvt_gtt_pte_ops gen8_gtt_pte_ops = { .get_entry = gtt_get_entry64, .set_entry = gtt_set_entry64, .clear_present = gtt_entry_clear_present, + .set_present = gtt_entry_set_present, .test_present = gen8_gtt_test_present, .test_pse = gen8_gtt_test_pse, .get_pfn = gen8_gtt_get_pfn, diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index ab590278dcf1d..f98c1c19b4cb4 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -62,6 +62,7 @@ struct intel_gvt_gtt_pte_ops { struct intel_vgpu *vgpu); bool (*test_present)(struct intel_gvt_gtt_entry *e); void (*clear_present)(struct intel_gvt_gtt_entry *e); + void (*set_present)(struct intel_gvt_gtt_entry *e); bool (*test_pse)(struct intel_gvt_gtt_entry *e); void (*set_pfn)(struct intel_gvt_gtt_entry *e, unsigned long pfn); unsigned long (*get_pfn)(struct intel_gvt_gtt_entry *e); -- GitLab From c1802534e5a6ec089e2b951116adfc14bb6dae64 Mon Sep 17 00:00:00 2001 From: Zhi Wang Date: Tue, 17 Oct 2017 02:00:27 +0800 Subject: [PATCH 0299/1471] drm/i915/gvt: Refine broken PPGTT scratch Refine previously broken PPGTT scratch. Scratch PTE was no correctly handled and also the handling of scratch entries in page table walk was not well organized, which brings gaps of introducing lazy shadow. Signed-off-by: Zhi Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 196 ++++++++++++++++++--------------- drivers/gpu/drm/i915/gvt/gtt.h | 17 +-- 2 files changed, 112 insertions(+), 101 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 3d6008b116e57..6fa9271e23a56 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -841,20 +841,51 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_find_shadow_page( return NULL; } +static bool ppgtt_is_scratch_entry(struct intel_vgpu *vgpu, + struct intel_gvt_gtt_entry *e) +{ + struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; + int pt_type = get_pt_type(e->type); + + if (pt_type == GTT_TYPE_INVALID) + return false; + + if (ops->get_pfn(e) == vgpu->gtt.ppgtt_scratch_page[pt_type].mfn) + return true; + + return false; +} + +static void ppgtt_get_scratch_entry(struct intel_vgpu *vgpu, int type, + struct intel_gvt_gtt_entry *e) +{ + struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; + struct intel_vgpu_shadow_page *scratch_page; + int pt_type = get_pt_type(type); + + if (WARN_ON(pt_type == GTT_TYPE_INVALID)) + return; + + scratch_page = &vgpu->gtt.ppgtt_scratch_page[pt_type]; + + e->type = get_entry_type(type); + ops->get_entry(scratch_page->vaddr, e, 0, false, 0, vgpu); +} + #define pt_entry_size_shift(spt) \ ((spt)->vgpu->gvt->device_info.gtt_entry_size_shift) #define pt_entries(spt) \ (I915_GTT_PAGE_SIZE >> pt_entry_size_shift(spt)) -#define for_each_present_guest_entry(spt, e, i) \ +#define for_each_guest_entry(spt, e, i) \ for (i = 0; i < pt_entries(spt); i++) \ - if (!ppgtt_get_guest_entry(spt, e, i) && \ - spt->vgpu->gvt->gtt.pte_ops->test_present(e)) + if (!ppgtt_get_guest_entry(spt, e, i)) #define for_each_present_shadow_entry(spt, e, i) \ for (i = 0; i < pt_entries(spt); i++) \ if (!ppgtt_get_shadow_entry(spt, e, i) && \ + !ppgtt_is_scratch_entry(spt->vgpu, e) && \ spt->vgpu->gvt->gtt.pte_ops->test_present(e)) static void ppgtt_get_shadow_page(struct intel_vgpu_ppgtt_spt *spt) @@ -873,18 +904,13 @@ static int ppgtt_invalidate_shadow_page_by_shadow_entry(struct intel_vgpu *vgpu, { struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_vgpu_ppgtt_spt *s; - intel_gvt_gtt_type_t cur_pt_type; if (WARN_ON(!gtt_type_is_pt(get_next_pt_type(e->type)))) return -EINVAL; - if (e->type != GTT_TYPE_PPGTT_ROOT_L3_ENTRY - && e->type != GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { - cur_pt_type = get_next_pt_type(e->type) + 1; - if (ops->get_pfn(e) == - vgpu->gtt.scratch_pt[cur_pt_type].page_mfn) - return 0; - } + if (WARN_ON(ppgtt_is_scratch_entry(vgpu, e))) + return -EINVAL; + s = ppgtt_find_shadow_page(vgpu, ops->get_pfn(e)); if (!s) { gvt_vgpu_err("fail to find shadow page: mfn: 0x%lx\n", @@ -997,6 +1023,7 @@ static inline void ppgtt_generate_shadow_entry(struct intel_gvt_gtt_entry *se, static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) { struct intel_vgpu *vgpu = spt->vgpu; + struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_vgpu_ppgtt_spt *s; struct intel_gvt_gtt_entry se, ge; unsigned long i; @@ -1006,22 +1033,34 @@ static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) spt->guest_page.track.gfn, spt->shadow_page.type); if (gtt_type_is_pte_pt(spt->shadow_page.type)) { - for_each_present_guest_entry(spt, &ge, i) { - ret = gtt_entry_p2m(vgpu, &ge, &se); - if (ret) - goto fail; + for_each_guest_entry(spt, &ge, i) { + if (ops->test_present(&ge)) { + ret = gtt_entry_p2m(vgpu, &ge, &se); + if (ret) + goto fail; + } else { + ppgtt_get_scratch_entry(vgpu, + spt->shadow_page.type, &se); + } ppgtt_set_shadow_entry(spt, &se, i); } return 0; } - for_each_present_guest_entry(spt, &ge, i) { + for_each_guest_entry(spt, &ge, i) { if (!gtt_type_is_pt(get_next_pt_type(ge.type))) { gvt_vgpu_err("GVT doesn't support pse bit now\n"); ret = -EINVAL; goto fail; } + if (!ops->test_present(&ge)) { + ppgtt_get_scratch_entry(vgpu, spt->shadow_page.type, + &se); + ppgtt_set_shadow_entry(spt, &se, i); + continue; + } + s = ppgtt_populate_shadow_page_by_guest_entry(vgpu, &ge); if (IS_ERR(s)) { ret = PTR_ERR(s); @@ -1053,7 +1092,7 @@ static int ppgtt_handle_guest_entry_removal(struct intel_vgpu_guest_page *gpt, if (!ops->test_present(se)) return 0; - if (ops->get_pfn(se) == vgpu->gtt.scratch_pt[sp->type].page_mfn) + if (ppgtt_is_scratch_entry(vgpu, se)) return 0; if (gtt_type_is_pt(get_next_pt_type(se->type))) { @@ -1292,7 +1331,6 @@ static int ppgtt_handle_guest_write_page_table( { struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt); struct intel_vgpu *vgpu = spt->vgpu; - int type = spt->shadow_page.type; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_gvt_gtt_entry se; @@ -1319,7 +1357,7 @@ static int ppgtt_handle_guest_write_page_table( goto fail; if (!new_present) { - ops->set_pfn(&se, vgpu->gtt.scratch_pt[type].page_mfn); + ppgtt_get_scratch_entry(vgpu, spt->shadow_page.type, &se); ppgtt_set_shadow_entry(spt, &se, index); } @@ -1968,106 +2006,85 @@ int intel_vgpu_emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, return ret; } -static int alloc_scratch_pages(struct intel_vgpu *vgpu, +static void ppgtt_destroy_scratch(struct intel_vgpu *vgpu) +{ + struct intel_vgpu_shadow_page *scratch_page; + int i; + + for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { + scratch_page = &vgpu->gtt.ppgtt_scratch_page[i]; + if (scratch_page->page != NULL) { + clean_shadow_page(vgpu, scratch_page); + __free_page(scratch_page->page); + } + } +} + +static int setup_ppgtt_scratch_page(struct intel_vgpu *vgpu, intel_gvt_gtt_type_t type) { + struct intel_gvt *gvt = vgpu->gvt; + struct intel_gvt_device_info *info = &gvt->device_info; + int num_entries = I915_GTT_PAGE_SIZE >> info->gtt_entry_size_shift; struct intel_vgpu_gtt *gtt = &vgpu->gtt; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; - int page_entry_num = I915_GTT_PAGE_SIZE >> - vgpu->gvt->device_info.gtt_entry_size_shift; - void *scratch_pt; - int i; - struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; - dma_addr_t daddr; + struct intel_vgpu_shadow_page *scratch_page; + struct intel_gvt_gtt_entry e; + intel_gvt_gtt_type_t next_pt_type; + int ret, i; if (WARN_ON(type < GTT_TYPE_PPGTT_PTE_PT || type >= GTT_TYPE_MAX)) return -EINVAL; - scratch_pt = (void *)get_zeroed_page(GFP_KERNEL); - if (!scratch_pt) { + scratch_page = >t->ppgtt_scratch_page[type]; + + scratch_page->page = alloc_page(GFP_KERNEL); + if (!scratch_page) { gvt_vgpu_err("fail to allocate scratch page\n"); return -ENOMEM; } - daddr = dma_map_page(dev, virt_to_page(scratch_pt), 0, - 4096, PCI_DMA_BIDIRECTIONAL); - if (dma_mapping_error(dev, daddr)) { - gvt_vgpu_err("fail to dmamap scratch_pt\n"); - __free_page(virt_to_page(scratch_pt)); + ret = init_shadow_page(vgpu, scratch_page, type, false); + if (ret) { + gvt_vgpu_err("fail to allocate scratch page\n"); + __free_page(scratch_page->page); return -ENOMEM; } - gtt->scratch_pt[type].page_mfn = - (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT); - gtt->scratch_pt[type].page = virt_to_page(scratch_pt); - gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n", - vgpu->id, type, gtt->scratch_pt[type].page_mfn); - - /* Build the tree by full filled the scratch pt with the entries which - * point to the next level scratch pt or scratch page. The - * scratch_pt[type] indicate the scratch pt/scratch page used by the - * 'type' pt. - * e.g. scratch_pt[GTT_TYPE_PPGTT_PDE_PT] is used by - * GTT_TYPE_PPGTT_PDE_PT level pt, that means this scratch_pt it self - * is GTT_TYPE_PPGTT_PTE_PT, and full filled by scratch page mfn. - */ - if (type > GTT_TYPE_PPGTT_PTE_PT && type < GTT_TYPE_MAX) { - struct intel_gvt_gtt_entry se; - - memset(&se, 0, sizeof(struct intel_gvt_gtt_entry)); - se.type = get_entry_type(type - 1); - ops->set_pfn(&se, gtt->scratch_pt[type - 1].page_mfn); - - /* The entry parameters like present/writeable/cache type - * set to the same as i915's scratch page tree. - */ - se.val64 |= _PAGE_PRESENT | _PAGE_RW; - if (type == GTT_TYPE_PPGTT_PDE_PT) - se.val64 |= PPAT_CACHED; - - for (i = 0; i < page_entry_num; i++) - ops->set_entry(scratch_pt, &se, i, false, 0, vgpu); - } - return 0; -} - -static int release_scratch_page_tree(struct intel_vgpu *vgpu) -{ - int i; - struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; - dma_addr_t daddr; + memset(&e, 0, sizeof(e)); - for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { - if (vgpu->gtt.scratch_pt[i].page != NULL) { - daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn << - I915_GTT_PAGE_SHIFT); - dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); - __free_page(vgpu->gtt.scratch_pt[i].page); - vgpu->gtt.scratch_pt[i].page = NULL; - vgpu->gtt.scratch_pt[i].page_mfn = 0; - } + if (type == GTT_TYPE_PPGTT_PTE_PT) { + e.type = GTT_TYPE_PPGTT_PTE_4K_ENTRY; + ops->set_pfn(&e, gvt->gtt.scratch_mfn); + } else { + next_pt_type = get_next_pt_type(type); + e.type = get_entry_type(type); + ops->set_pfn(&e, gtt->ppgtt_scratch_page[next_pt_type].mfn); } + ops->set_present(&e); + + for (i = 0; i < num_entries; i++) + ops->set_entry(scratch_page->vaddr, &e, i, false, 0, vgpu); + return 0; } -static int create_scratch_page_tree(struct intel_vgpu *vgpu) +static int ppgtt_create_scratch(struct intel_vgpu *vgpu) { int i, ret; for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { - ret = alloc_scratch_pages(vgpu, i); + ret = setup_ppgtt_scratch_page(vgpu, i); if (ret) goto err; } - return 0; - err: - release_scratch_page_tree(vgpu); + ppgtt_destroy_scratch(vgpu); return ret; } - + /** * intel_vgpu_init_gtt - initialize per-vGPU graphics memory virulization * @vgpu: a vGPU @@ -2100,8 +2117,7 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu) } gtt->ggtt_mm = ggtt_mm; - - return create_scratch_page_tree(vgpu); + return ppgtt_create_scratch(vgpu); } static void intel_vgpu_free_mm(struct intel_vgpu *vgpu, int type) @@ -2133,7 +2149,7 @@ static void intel_vgpu_free_mm(struct intel_vgpu *vgpu, int type) void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu) { ppgtt_free_all_shadow_page(vgpu); - release_scratch_page_tree(vgpu); + ppgtt_destroy_scratch(vgpu); intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_PPGTT); intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_GGTT); diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index f98c1c19b4cb4..416b2f868cf09 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -188,9 +188,12 @@ extern void intel_vgpu_destroy_mm(struct kref *mm_ref); struct intel_vgpu_guest_page; -struct intel_vgpu_scratch_pt { +struct intel_vgpu_shadow_page { + void *vaddr; struct page *page; - unsigned long page_mfn; + int type; + struct hlist_node node; + unsigned long mfn; }; struct intel_vgpu_gtt { @@ -202,7 +205,7 @@ struct intel_vgpu_gtt { atomic_t n_tracked_guest_page; struct list_head oos_page_list_head; struct list_head post_shadow_list_head; - struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX]; + struct intel_vgpu_shadow_page ppgtt_scratch_page[GTT_TYPE_MAX]; }; extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu); @@ -218,14 +221,6 @@ extern struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu, struct intel_vgpu_oos_page; -struct intel_vgpu_shadow_page { - void *vaddr; - struct page *page; - int type; - struct hlist_node node; - unsigned long mfn; -}; - struct intel_vgpu_page_track { struct hlist_node node; bool tracked; -- GitLab From a2ae95af9646316aaf86e2d18f46de1a5f746f1a Mon Sep 17 00:00:00 2001 From: Weinan Li Date: Fri, 20 Oct 2017 15:16:46 +0800 Subject: [PATCH 0300/1471] drm/i915/gvt: update CSB and CSB write pointer in virtual HWSP The engine provides a mirror of the CSB and CSB write pointer in the HWSP. Read these status from virtual HWSP in VM can reduce CPU utilization while applications have much more short GPU workloads. Here we update the corresponding data in virtual HWSP as it in virtual MMIO. Before read these status from HWSP in GVT-g VM, please ensure the host support it by checking the BIT(3) of caps in PVINFO. Virtual HWSP only support GEN8+ platform, since the HWSP MMIO may change follow the platform update, please add the corresponding MMIO emulation when enable new platforms in GVT-g. v3 : Add address audit in HWSP address update. v4 : Separate this patch with enalbe virtual HWSP in VM. Use intel_gvt_render_mmio_to_ring_id() to determine ring_id by offset. v5 : Remove unnessary check about Gen8, GVT-g only support Gen8+. Signed-off-by: Weinan Li Cc: Zhenyu Wang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 16 +++++++++++++++ drivers/gpu/drm/i915/gvt/gvt.h | 1 + drivers/gpu/drm/i915/gvt/handlers.c | 30 ++++++++++++++++++++++++++++- drivers/gpu/drm/i915/gvt/vgpu.c | 3 +++ 4 files changed, 49 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index 5c966edfeefb6..c9fa0fb488d3a 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -133,6 +133,8 @@ static void emulate_csb_update(struct intel_vgpu_execlist *execlist, struct execlist_context_status_pointer_format ctx_status_ptr; u32 write_pointer; u32 ctx_status_ptr_reg, ctx_status_buf_reg, offset; + unsigned long hwsp_gpa; + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id, _EL_OFFSET_STATUS_PTR); @@ -158,6 +160,20 @@ static void emulate_csb_update(struct intel_vgpu_execlist *execlist, ctx_status_ptr.write_ptr = write_pointer; vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw; + /* Update the CSB and CSB write pointer in HWSP */ + hwsp_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm, + vgpu->hws_pga[ring_id]); + if (hwsp_gpa != INTEL_GVT_INVALID_ADDR) { + intel_gvt_hypervisor_write_gpa(vgpu, + hwsp_gpa + I915_HWS_CSB_BUF0_INDEX * 4 + + write_pointer * 8, + status, 8); + intel_gvt_hypervisor_write_gpa(vgpu, + hwsp_gpa + + intel_hws_csb_write_index(dev_priv) * 4, + &write_pointer, 4); + } + gvt_dbg_el("vgpu%d: w pointer %u reg %x csb l %x csb h %x\n", vgpu->id, write_pointer, offset, status->ldw, status->udw); diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index f08d194f639fa..27e8186cbc81a 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -189,6 +189,7 @@ struct intel_vgpu { struct intel_vgpu_opregion opregion; struct intel_vgpu_display display; struct intel_vgpu_submission submission; + u32 hws_pga[I915_NUM_ENGINES]; struct dentry *debugfs; diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index bb16ef5c85104..56ee588377f48 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1380,6 +1380,34 @@ static int mailbox_write(struct intel_vgpu *vgpu, unsigned int offset, return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes); } +static int hws_pga_write(struct intel_vgpu *vgpu, unsigned int offset, + void *p_data, unsigned int bytes) +{ + u32 value = *(u32 *)p_data; + int ring_id = intel_gvt_render_mmio_to_ring_id(vgpu->gvt, offset); + + if (!intel_gvt_ggtt_validate_range(vgpu, value, I915_GTT_PAGE_SIZE)) { + gvt_vgpu_err("VM(%d) write invalid HWSP address, reg:0x%x, value:0x%x\n", + vgpu->id, offset, value); + return -EINVAL; + } + /* + * Need to emulate all the HWSP register write to ensure host can + * update the VM CSB status correctly. Here listed registers can + * support BDW, SKL or other platforms with same HWSP registers. + */ + if (unlikely(ring_id < 0 || ring_id > I915_NUM_ENGINES)) { + gvt_vgpu_err("VM(%d) access unknown hardware status page register:0x%x\n", + vgpu->id, offset); + return -EINVAL; + } + vgpu->hws_pga[ring_id] = value; + gvt_dbg_mmio("VM(%d) write: 0x%x to HWSP: 0x%x\n", + vgpu->id, value, offset); + + return intel_vgpu_default_mmio_write(vgpu, offset, &value, bytes); +} + static int skl_power_well_ctl_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { @@ -2535,7 +2563,7 @@ static int init_broadwell_mmio_info(struct intel_gvt *gvt) MMIO_RING_F(RING_REG, 32, 0, 0, 0, D_BDW_PLUS, NULL, NULL); #undef RING_REG - MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, NULL); + MMIO_RING_GM_RDR(RING_HWS_PGA, D_BDW_PLUS, NULL, hws_pga_write); MMIO_DFH(HDC_CHICKEN0, D_BDW_PLUS, F_MODE_MASK | F_CMD_ACCESS, NULL, NULL); diff --git a/drivers/gpu/drm/i915/gvt/vgpu.c b/drivers/gpu/drm/i915/gvt/vgpu.c index 99dbefadfe919..c6b82d1ba7dee 100644 --- a/drivers/gpu/drm/i915/gvt/vgpu.c +++ b/drivers/gpu/drm/i915/gvt/vgpu.c @@ -43,7 +43,10 @@ void populate_pvinfo_page(struct intel_vgpu *vgpu) vgpu_vreg(vgpu, vgtif_reg(version_minor)) = 0; vgpu_vreg(vgpu, vgtif_reg(display_ready)) = 0; vgpu_vreg(vgpu, vgtif_reg(vgt_id)) = vgpu->id; + vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) = VGT_CAPS_FULL_48BIT_PPGTT; + vgpu_vreg(vgpu, vgtif_reg(vgt_caps)) |= VGT_CAPS_HWSP_EMULATION; + vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.base)) = vgpu_aperture_gmadr_base(vgpu); vgpu_vreg(vgpu, vgtif_reg(avail_rs.mappable_gmadr.size)) = -- GitLab From 4023f301d28fb18aac225fef59592a44b9fd42b3 Mon Sep 17 00:00:00 2001 From: Xiaolin Zhang Date: Fri, 20 Oct 2017 21:52:29 +0800 Subject: [PATCH 0301/1471] drm/i915/gvt: opregion virtualization for win guest this is an enhanced opregion emulation for win guest support by initializing more data members including opregion header size, version and child device propertity for display port. for simplicity, redefined child_device_config structure. Signed-off-by: Xiaolin Zhang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/opregion.c | 82 +++++++++++++++++++++++++---- 1 file changed, 73 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c index db8abac5faad4..2533d1ef1c569 100644 --- a/drivers/gpu/drm/i915/gvt/opregion.c +++ b/drivers/gpu/drm/i915/gvt/opregion.c @@ -63,6 +63,60 @@ struct bdb_data_header { u16 size; /* data size */ } __packed; +struct efp_child_device_config { + u16 handle; + u16 device_type; + u16 device_class; + u8 i2c_speed; + u8 dp_onboard_redriver; /* 158 */ + u8 dp_ondock_redriver; /* 158 */ + u8 hdmi_level_shifter_value:4; /* 169 */ + u8 hdmi_max_data_rate:4; /* 204 */ + u16 dtd_buf_ptr; /* 161 */ + u8 edidless_efp:1; /* 161 */ + u8 compression_enable:1; /* 198 */ + u8 compression_method:1; /* 198 */ + u8 ganged_edp:1; /* 202 */ + u8 skip0:4; + u8 compression_structure_index:4; /* 198 */ + u8 skip1:4; + u8 slave_port; /* 202 */ + u8 skip2; + u8 dvo_port; + u8 i2c_pin; /* for add-in card */ + u8 slave_addr; /* for add-in card */ + u8 ddc_pin; + u16 edid_ptr; + u8 dvo_config; + u8 efp_docked_port:1; /* 158 */ + u8 lane_reversal:1; /* 184 */ + u8 onboard_lspcon:1; /* 192 */ + u8 iboost_enable:1; /* 196 */ + u8 hpd_invert:1; /* BXT 196 */ + u8 slip3:3; + u8 hdmi_compat:1; + u8 dp_compat:1; + u8 tmds_compat:1; + u8 skip4:5; + u8 aux_channel; + u8 dongle_detect; + u8 pipe_cap:2; + u8 sdvo_stall:1; /* 158 */ + u8 hpd_status:2; + u8 integrated_encoder:1; + u8 skip5:2; + u8 dvo_wiring; + u8 mipi_bridge_type; /* 171 */ + u16 device_class_ext; + u8 dvo_function; + u8 dp_usb_type_c:1; /* 195 */ + u8 skip6:7; + u8 dp_usb_type_c_2x_gpio_index; /* 195 */ + u16 dp_usb_type_c_2x_gpio_pin; /* 195 */ + u8 iboost_dp:4; /* 196 */ + u8 iboost_hdmi:4; /* 196 */ +} __packed; + struct vbt { /* header->bdb_offset point to bdb_header offset */ struct vbt_header header; @@ -73,10 +127,11 @@ struct vbt { struct bdb_data_header general_definitions_header; struct bdb_general_definitions general_definitions; - struct child_device_config child0; - struct child_device_config child1; - struct child_device_config child2; - struct child_device_config child3; + + struct efp_child_device_config child0; + struct efp_child_device_config child1; + struct efp_child_device_config child2; + struct efp_child_device_config child3; struct bdb_data_header driver_features_header; struct bdb_driver_features driver_features; @@ -100,7 +155,7 @@ static void virt_vbt_generation(struct vbt *v) v->header.bdb_offset = offsetof(struct vbt, bdb_header); strcpy(&v->bdb_header.signature[0], "BIOS_DATA_BLOCK"); - v->bdb_header.version = 198; /* child_dev_size = 38 */ + v->bdb_header.version = 186; /* child_dev_size = 38 */ v->bdb_header.header_size = sizeof(v->bdb_header); v->bdb_header.bdb_size = sizeof(struct vbt) - sizeof(struct vbt_header) @@ -125,24 +180,32 @@ static void virt_vbt_generation(struct vbt *v) v->child0.device_type = DEVICE_TYPE_DP; v->child0.dvo_port = DVO_PORT_DPA; v->child0.aux_channel = DP_AUX_A; + v->child0.dp_compat = true; + v->child0.integrated_encoder = true; /* portB */ v->child1.handle = DEVICE_TYPE_EFP2; v->child1.device_type = DEVICE_TYPE_DP; v->child1.dvo_port = DVO_PORT_DPB; v->child1.aux_channel = DP_AUX_B; + v->child1.dp_compat = true; + v->child1.integrated_encoder = true; /* portC */ v->child2.handle = DEVICE_TYPE_EFP3; v->child2.device_type = DEVICE_TYPE_DP; v->child2.dvo_port = DVO_PORT_DPC; v->child2.aux_channel = DP_AUX_C; + v->child2.dp_compat = true; + v->child2.integrated_encoder = true; /* portD */ v->child3.handle = DEVICE_TYPE_EFP4; v->child3.device_type = DEVICE_TYPE_DP; v->child3.dvo_port = DVO_PORT_DPD; v->child3.aux_channel = DP_AUX_D; + v->child3.dp_compat = true; + v->child3.integrated_encoder = true; /* driver features */ v->driver_features_header.id = BDB_DRIVER_FEATURES; @@ -278,22 +341,23 @@ int intel_gvt_init_opregion(struct intel_gvt *gvt) } /* emulated opregion with VBT mailbox only */ - header = (struct opregion_header *)gvt->opregion.opregion_va; + buf = (u8 *)gvt->opregion.opregion_va; + header = (struct opregion_header *)buf; memcpy(header->signature, OPREGION_SIGNATURE, sizeof(OPREGION_SIGNATURE)); + header->size = 0x8; + header->opregion_ver = 0x02000000; header->mboxes = MBOX_VBT; /* for unknown reason, the value in LID field is incorrect * which block the windows guest, so workaround it by force * setting it to "OPEN" */ - buf = (u8 *)gvt->opregion.opregion_va; buf[INTEL_GVT_OPREGION_CLID] = 0x3; /* emulated vbt from virt vbt generation */ virt_vbt_generation(&v); - memcpy(gvt->opregion.opregion_va + INTEL_GVT_OPREGION_VBT_OFFSET, - &v, sizeof(struct vbt)); + memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt)); return 0; } -- GitLab From 7cb16018f5b4f04bc58a8752bfd11067bafeb552 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Mon, 23 Oct 2017 11:46:43 +0800 Subject: [PATCH 0302/1471] drm/i915/gvt: Add mmio iterator intel_gvt_for_each_tracked_mmio() This patch add a function intel_gvt_for_each_tracked_mmio() to iterate each tracked mmio. The caller don't be aware of how the tracked mmios are presented internally. v2: remove snapshot_hw_mmio_registers(). Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/firmware.c | 26 ++++++++++------------ drivers/gpu/drm/i915/gvt/handlers.c | 34 +++++++++++++++++++++++++++++ drivers/gpu/drm/i915/gvt/mmio.h | 4 ++++ 3 files changed, 49 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/firmware.c b/drivers/gpu/drm/i915/gvt/firmware.c index a26c1705430eb..a73e1d418c228 100644 --- a/drivers/gpu/drm/i915/gvt/firmware.c +++ b/drivers/gpu/drm/i915/gvt/firmware.c @@ -66,20 +66,23 @@ static struct bin_attribute firmware_attr = { .mmap = NULL, }; -static int expose_firmware_sysfs(struct intel_gvt *gvt) +static int mmio_snapshot_handler(struct intel_gvt *gvt, u32 offset, void *data) { struct drm_i915_private *dev_priv = gvt->dev_priv; + + *(u32 *)(data + offset) = I915_READ_NOTRACE(_MMIO(offset)); + return 0; +} + +static int expose_firmware_sysfs(struct intel_gvt *gvt) +{ struct intel_gvt_device_info *info = &gvt->device_info; struct pci_dev *pdev = gvt->dev_priv->drm.pdev; - struct intel_gvt_mmio_info *e; - struct gvt_mmio_block *block = gvt->mmio.mmio_block; - int num = gvt->mmio.num_mmio_block; struct gvt_firmware_header *h; void *firmware; void *p; unsigned long size, crc32_start; - int i, j; - int ret; + int i, ret; size = sizeof(*h) + info->mmio_size + info->cfg_space_size; firmware = vzalloc(size); @@ -104,15 +107,8 @@ static int expose_firmware_sysfs(struct intel_gvt *gvt) p = firmware + h->mmio_offset; - hash_for_each(gvt->mmio.mmio_info_table, i, e, node) - *(u32 *)(p + e->offset) = I915_READ_NOTRACE(_MMIO(e->offset)); - - for (i = 0; i < num; i++, block++) { - for (j = 0; j < block->size; j += 4) - *(u32 *)(p + INTEL_GVT_MMIO_OFFSET(block->offset) + j) = - I915_READ_NOTRACE(_MMIO(INTEL_GVT_MMIO_OFFSET( - block->offset) + j)); - } + /* Take a snapshot of hw mmio registers. */ + intel_gvt_for_each_tracked_mmio(gvt, mmio_snapshot_handler, p); memcpy(gvt->firmware.mmio, p, info->mmio_size); diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 56ee588377f48..4f8d470dbb3fa 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -2971,6 +2971,40 @@ int intel_gvt_setup_mmio_info(struct intel_gvt *gvt) return ret; } +/** + * intel_gvt_for_each_tracked_mmio - iterate each tracked mmio + * @gvt: a GVT device + * @handler: the handler + * @data: private data given to handler + * + * Returns: + * Zero on success, negative error code if failed. + */ +int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt, + int (*handler)(struct intel_gvt *gvt, u32 offset, void *data), + void *data) +{ + struct gvt_mmio_block *block = gvt->mmio.mmio_block; + struct intel_gvt_mmio_info *e; + int i, j, ret; + + hash_for_each(gvt->mmio.mmio_info_table, i, e, node) { + ret = handler(gvt, e->offset, data); + if (ret) + return ret; + } + + for (i = 0; i < gvt->mmio.num_mmio_block; i++, block++) { + for (j = 0; j < block->size; j += 4) { + ret = handler(gvt, + INTEL_GVT_MMIO_OFFSET(block->offset) + j, + data); + if (ret) + return ret; + } + } + return 0; +} /** * intel_vgpu_default_mmio_read - default MMIO read handler diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h index dbc04ad2c7a1e..62709ac351cdc 100644 --- a/drivers/gpu/drm/i915/gvt/mmio.h +++ b/drivers/gpu/drm/i915/gvt/mmio.h @@ -72,6 +72,10 @@ bool intel_gvt_match_device(struct intel_gvt *gvt, unsigned long device); int intel_gvt_setup_mmio_info(struct intel_gvt *gvt); void intel_gvt_clean_mmio_info(struct intel_gvt *gvt); +int intel_gvt_for_each_tracked_mmio(struct intel_gvt *gvt, + int (*handler)(struct intel_gvt *gvt, u32 offset, void *data), + void *data); + #define INTEL_GVT_MMIO_OFFSET(reg) ({ \ typeof(reg) __reg = reg; \ -- GitLab From cea9083e35b2f771624a18126b10ef1fd7806040 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Mon, 23 Oct 2017 11:46:44 +0800 Subject: [PATCH 0303/1471] drm/i915/gvt: Add new debugfs tool mmio_diff This new debugfs entry is used to figure out which registers of vGPU is different to host. It is a useful tool for new platform enabling and debugging. When read this entry, all the diff mmio are recognized and sorted by mmio offset. Besides, the bit positions of different value are listed in 'Diff' column. Here is a show: $ sudo cat ./mmio_diff Offset HW vGPU Diff 00002030 000025f8 00000000 3-8,10,13 00002034 012025f8 00000000 3-8,10,13,21,24 00002038 027fb000 00000000 12-13,15-22,25 0000203c 00003000 00000000 12-13 00002054 0000000a 00000040 1,3,6 00002074 012025f8 00000000 3-8,10,13,21,24 00002080 fffe6000 00000000 13-14,17-31 000020a8 fffffeff ffffffff 8 000020d4 00000004 00000000 2 .... 00145974 eb42718c 010c11b0 2-5,13-14,17-19,22,25,27,29-31 00145978 0000002f 0000002a 0,2 0014597c 0000002f 0000002a 0,2 00145980 0000002b 00000028 0-1 00145984 a5a87c9e b27d20c0 1-4,6,10-12,14,16,18,20,22-26,28 001459c0 88390000 883c0000 16,18 00146200 88350000 883a0000 16-19 Total: 72432, Diff: 901 v3: fix a typo. v2: add mmio_hw_access_pre/post(). Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/debugfs.c | 115 +++++++++++++++++++++++++++++ 1 file changed, 115 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/debugfs.c b/drivers/gpu/drm/i915/gvt/debugfs.c index 02acd58426055..32a66dfdf1127 100644 --- a/drivers/gpu/drm/i915/gvt/debugfs.c +++ b/drivers/gpu/drm/i915/gvt/debugfs.c @@ -21,9 +21,119 @@ * SOFTWARE. */ #include +#include #include "i915_drv.h" #include "gvt.h" +struct mmio_diff_param { + struct intel_vgpu *vgpu; + int total; + int diff; + struct list_head diff_mmio_list; +}; + +struct diff_mmio { + struct list_head node; + u32 offset; + u32 preg; + u32 vreg; +}; + +/* Compare two diff_mmio items. */ +static int mmio_offset_compare(void *priv, + struct list_head *a, struct list_head *b) +{ + struct diff_mmio *ma; + struct diff_mmio *mb; + + ma = container_of(a, struct diff_mmio, node); + mb = container_of(b, struct diff_mmio, node); + if (ma->offset < mb->offset) + return -1; + else if (ma->offset > mb->offset) + return 1; + return 0; +} + +static inline int mmio_diff_handler(struct intel_gvt *gvt, + u32 offset, void *data) +{ + struct drm_i915_private *dev_priv = gvt->dev_priv; + struct mmio_diff_param *param = data; + struct diff_mmio *node; + u32 preg, vreg; + + preg = I915_READ_NOTRACE(_MMIO(offset)); + vreg = vgpu_vreg(param->vgpu, offset); + + if (preg != vreg) { + node = kmalloc(sizeof(*node), GFP_KERNEL); + if (!node) + return -ENOMEM; + + node->offset = offset; + node->preg = preg; + node->vreg = vreg; + list_add(&node->node, ¶m->diff_mmio_list); + param->diff++; + } + param->total++; + return 0; +} + +/* Show the all the different values of tracked mmio. */ +static int vgpu_mmio_diff_show(struct seq_file *s, void *unused) +{ + struct intel_vgpu *vgpu = s->private; + struct intel_gvt *gvt = vgpu->gvt; + struct mmio_diff_param param = { + .vgpu = vgpu, + .total = 0, + .diff = 0, + }; + struct diff_mmio *node, *next; + + INIT_LIST_HEAD(¶m.diff_mmio_list); + + mutex_lock(&gvt->lock); + spin_lock_bh(&gvt->scheduler.mmio_context_lock); + + mmio_hw_access_pre(gvt->dev_priv); + /* Recognize all the diff mmios to list. */ + intel_gvt_for_each_tracked_mmio(gvt, mmio_diff_handler, ¶m); + mmio_hw_access_post(gvt->dev_priv); + + spin_unlock_bh(&gvt->scheduler.mmio_context_lock); + mutex_unlock(&gvt->lock); + + /* In an ascending order by mmio offset. */ + list_sort(NULL, ¶m.diff_mmio_list, mmio_offset_compare); + + seq_printf(s, "%-8s %-8s %-8s %-8s\n", "Offset", "HW", "vGPU", "Diff"); + list_for_each_entry_safe(node, next, ¶m.diff_mmio_list, node) { + u32 diff = node->preg ^ node->vreg; + + seq_printf(s, "%08x %08x %08x %*pbl\n", + node->offset, node->preg, node->vreg, + 32, &diff); + list_del(&node->node); + kfree(node); + } + seq_printf(s, "Total: %d, Diff: %d\n", param.total, param.diff); + return 0; +} + +static int vgpu_mmio_diff_open(struct inode *inode, struct file *file) +{ + return single_open(file, vgpu_mmio_diff_show, inode->i_private); +} + +static const struct file_operations vgpu_mmio_diff_fops = { + .open = vgpu_mmio_diff_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; /** * intel_gvt_debugfs_add_vgpu - register debugfs entries for a vGPU @@ -47,6 +157,11 @@ int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu) if (!ent) return -ENOMEM; + ent = debugfs_create_file("mmio_diff", 0444, vgpu->debugfs, + vgpu, &vgpu_mmio_diff_fops); + if (!ent) + return -ENOMEM; + return 0; } -- GitLab From 0bde438b7e2200fc961625ac7b828cbc8c6d78e1 Mon Sep 17 00:00:00 2001 From: Zhenyu Wang Date: Thu, 26 Oct 2017 15:29:13 +0800 Subject: [PATCH 0304/1471] MAINTAINERS: Update gvt-linux.git new repo place gvt-linux.git repo is moved for its new place under https://github.com/intel/gvt-linux.git Old https://github.com/01org/gvt-linux.git is set only for redirect now. Signed-off-by: Zhenyu Wang --- MAINTAINERS | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index 6e1f94b4ed263..9e6cf2209e81a 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -6928,7 +6928,7 @@ M: Zhi Wang L: intel-gvt-dev@lists.freedesktop.org L: intel-gfx@lists.freedesktop.org W: https://01.org/igvt-g -T: git https://github.com/01org/gvt-linux.git +T: git https://github.com/intel/gvt-linux.git S: Supported F: drivers/gpu/drm/i915/gvt/ -- GitLab From ffe2a503b06cc300e25c1dfc00c85a2240cf97d7 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Mon, 30 Oct 2017 14:19:15 +0800 Subject: [PATCH 0305/1471] drm/i915/gvt: Reduce rcs mocs switch latency Use I915_WRITE_FW instead of I915_WRITE to reduce overhead. The overall mmio switch latency lowers from ~600us to ~180us. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/render.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c index e16c3551b4a3e..0672178548ef9 100644 --- a/drivers/gpu/drm/i915/gvt/render.c +++ b/drivers/gpu/drm/i915/gvt/render.c @@ -209,7 +209,7 @@ static void load_mocs(struct intel_vgpu *vgpu, int ring_id) offset.reg = regs[ring_id]; for (i = 0; i < 64; i++) { gen9_render_mocs[ring_id][i] = I915_READ_FW(offset); - I915_WRITE(offset, vgpu_vreg(vgpu, offset)); + I915_WRITE_FW(offset, vgpu_vreg(vgpu, offset)); offset.reg += 4; } -- GitLab From e4aeba697943692652246a2d0a040f0859d70e86 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Thu, 2 Nov 2017 13:33:23 +0800 Subject: [PATCH 0306/1471] drm/i915/gvt: Don't dump partial state in cmd parser I have seen the cmd parser dump partial odd info. Stop that and only dump the full verbose info when debug enabled. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 8c8514ee73ac0..18c45734c7a27 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -709,18 +709,13 @@ static void parser_exec_state_dump(struct parser_exec_state *s) print_opcode(cmd_val(s, 0), s->ring_id); - /* print the whole page to trace */ - pr_err(" ip_va=%p: %08x %08x %08x %08x\n", - s->ip_va, cmd_val(s, 0), cmd_val(s, 1), - cmd_val(s, 2), cmd_val(s, 3)); - s->ip_va = (u32 *)((((u64)s->ip_va) >> 12) << 12); while (cnt < 1024) { - pr_err("ip_va=%p: ", s->ip_va); + gvt_dbg_cmd("ip_va=%p: ", s->ip_va); for (i = 0; i < 8; i++) - pr_err("%08x ", cmd_val(s, i)); - pr_err("\n"); + gvt_dbg_cmd("%08x ", cmd_val(s, i)); + gvt_dbg_cmd("\n"); s->ip_va += 8 * sizeof(u32); cnt += 8; -- GitLab From c982c45db6a78e1c23a416785fe920542ae1f33e Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Thu, 2 Nov 2017 13:33:32 +0800 Subject: [PATCH 0307/1471] drm/i915/gvt: Make gvt_vgpu_err use pr_err gvt_vgpu_err means something goes wrong. We need the error propagates to kernel message by default. Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/debug.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/debug.h b/drivers/gpu/drm/i915/gvt/debug.h index d98d5a0b928c7..c6027125c1ec2 100644 --- a/drivers/gpu/drm/i915/gvt/debug.h +++ b/drivers/gpu/drm/i915/gvt/debug.h @@ -30,9 +30,9 @@ #define gvt_vgpu_err(fmt, args...) \ do { \ if (IS_ERR_OR_NULL(vgpu)) \ - pr_debug("gvt: "fmt, ##args); \ + pr_err("gvt: "fmt, ##args); \ else \ - pr_debug("gvt: vgpu %d: "fmt, vgpu->id, ##args);\ + pr_err("gvt: vgpu %d: "fmt, vgpu->id, ##args);\ } while (0) #define gvt_dbg_core(fmt, args...) \ -- GitLab From c4270d122ccff963a021d1beb893d6192336af96 Mon Sep 17 00:00:00 2001 From: Changbin Du Date: Thu, 2 Nov 2017 13:33:42 +0800 Subject: [PATCH 0308/1471] drm/i915/gvt: Emulate PCI expansion ROM base address register Our vGPU doesn't have a device ROM, we need follow the PCI spec to report this info to drivers. Otherwise, we would see below errors. Inspecting possible rom at 0xfe049000 (vd=8086:1912 bdf=00:10.0) qemu-system-x86_64: vfio-pci: Cannot read device rom at 00000000-0000-0000-0000-000000000001 Device option ROM contents are probably invalid (check dmesg). Skip option ROM probe with rombar=0, or load from file with romfile=No option rom signature (got 4860) I will also send a improvement patch to PCI subsystem related to PCI ROM. But no idea to omit below error, since no pattern to detect vbios shadow without touch its content. 0000:00:10.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0x0000 Signed-off-by: Changbin Du Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/cfg_space.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/drivers/gpu/drm/i915/gvt/cfg_space.c b/drivers/gpu/drm/i915/gvt/cfg_space.c index ab19545d59a18..4ce2e6bd06803 100644 --- a/drivers/gpu/drm/i915/gvt/cfg_space.c +++ b/drivers/gpu/drm/i915/gvt/cfg_space.c @@ -208,6 +208,20 @@ static int emulate_pci_command_write(struct intel_vgpu *vgpu, return 0; } +static int emulate_pci_rom_bar_write(struct intel_vgpu *vgpu, + unsigned int offset, void *p_data, unsigned int bytes) +{ + u32 *pval = (u32 *)(vgpu_cfg_space(vgpu) + offset); + u32 new = *(u32 *)(p_data); + + if ((new & PCI_ROM_ADDRESS_MASK) == PCI_ROM_ADDRESS_MASK) + /* We don't have rom, return size of 0. */ + *pval = 0; + else + vgpu_pci_cfg_mem_write(vgpu, offset, p_data, bytes); + return 0; +} + static int emulate_pci_bar_write(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { @@ -300,6 +314,11 @@ int intel_vgpu_emulate_cfg_write(struct intel_vgpu *vgpu, unsigned int offset, } switch (rounddown(offset, 4)) { + case PCI_ROM_ADDRESS: + if (WARN_ON(!IS_ALIGNED(offset, 4))) + return -EINVAL; + return emulate_pci_rom_bar_write(vgpu, offset, p_data, bytes); + case PCI_BASE_ADDRESS_0 ... PCI_BASE_ADDRESS_5: if (WARN_ON(!IS_ALIGNED(offset, 4))) return -EINVAL; @@ -375,6 +394,8 @@ void intel_vgpu_init_cfg_space(struct intel_vgpu *vgpu, pci_resource_len(gvt->dev_priv->drm.pdev, 0); vgpu->cfg_space.bar[INTEL_GVT_PCI_BAR_APERTURE].size = pci_resource_len(gvt->dev_priv->drm.pdev, 2); + + memset(vgpu_cfg_space(vgpu) + PCI_ROM_ADDRESS, 0, 4); } /** -- GitLab From 5c35258de6cfc0829803fe9f9869253a4e600925 Mon Sep 17 00:00:00 2001 From: Zhenyu Wang Date: Thu, 2 Nov 2017 17:44:52 +0800 Subject: [PATCH 0309/1471] Revert "drm/i915/gvt: Refine broken PPGTT scratch" This reverts commit b20d09886fd1b74cd2255d846029a049e524db14. This caused windows driver boot errors for invalid page address. Revert for now. Signed-off-by: Zhenyu Wang Cc: Zhi Wang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 196 +++++++++++++++------------------ drivers/gpu/drm/i915/gvt/gtt.h | 17 ++- 2 files changed, 101 insertions(+), 112 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 6fa9271e23a56..3d6008b116e57 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -841,51 +841,20 @@ static struct intel_vgpu_ppgtt_spt *ppgtt_find_shadow_page( return NULL; } -static bool ppgtt_is_scratch_entry(struct intel_vgpu *vgpu, - struct intel_gvt_gtt_entry *e) -{ - struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; - int pt_type = get_pt_type(e->type); - - if (pt_type == GTT_TYPE_INVALID) - return false; - - if (ops->get_pfn(e) == vgpu->gtt.ppgtt_scratch_page[pt_type].mfn) - return true; - - return false; -} - -static void ppgtt_get_scratch_entry(struct intel_vgpu *vgpu, int type, - struct intel_gvt_gtt_entry *e) -{ - struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; - struct intel_vgpu_shadow_page *scratch_page; - int pt_type = get_pt_type(type); - - if (WARN_ON(pt_type == GTT_TYPE_INVALID)) - return; - - scratch_page = &vgpu->gtt.ppgtt_scratch_page[pt_type]; - - e->type = get_entry_type(type); - ops->get_entry(scratch_page->vaddr, e, 0, false, 0, vgpu); -} - #define pt_entry_size_shift(spt) \ ((spt)->vgpu->gvt->device_info.gtt_entry_size_shift) #define pt_entries(spt) \ (I915_GTT_PAGE_SIZE >> pt_entry_size_shift(spt)) -#define for_each_guest_entry(spt, e, i) \ +#define for_each_present_guest_entry(spt, e, i) \ for (i = 0; i < pt_entries(spt); i++) \ - if (!ppgtt_get_guest_entry(spt, e, i)) + if (!ppgtt_get_guest_entry(spt, e, i) && \ + spt->vgpu->gvt->gtt.pte_ops->test_present(e)) #define for_each_present_shadow_entry(spt, e, i) \ for (i = 0; i < pt_entries(spt); i++) \ if (!ppgtt_get_shadow_entry(spt, e, i) && \ - !ppgtt_is_scratch_entry(spt->vgpu, e) && \ spt->vgpu->gvt->gtt.pte_ops->test_present(e)) static void ppgtt_get_shadow_page(struct intel_vgpu_ppgtt_spt *spt) @@ -904,13 +873,18 @@ static int ppgtt_invalidate_shadow_page_by_shadow_entry(struct intel_vgpu *vgpu, { struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_vgpu_ppgtt_spt *s; + intel_gvt_gtt_type_t cur_pt_type; if (WARN_ON(!gtt_type_is_pt(get_next_pt_type(e->type)))) return -EINVAL; - if (WARN_ON(ppgtt_is_scratch_entry(vgpu, e))) - return -EINVAL; - + if (e->type != GTT_TYPE_PPGTT_ROOT_L3_ENTRY + && e->type != GTT_TYPE_PPGTT_ROOT_L4_ENTRY) { + cur_pt_type = get_next_pt_type(e->type) + 1; + if (ops->get_pfn(e) == + vgpu->gtt.scratch_pt[cur_pt_type].page_mfn) + return 0; + } s = ppgtt_find_shadow_page(vgpu, ops->get_pfn(e)); if (!s) { gvt_vgpu_err("fail to find shadow page: mfn: 0x%lx\n", @@ -1023,7 +997,6 @@ static inline void ppgtt_generate_shadow_entry(struct intel_gvt_gtt_entry *se, static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) { struct intel_vgpu *vgpu = spt->vgpu; - struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_vgpu_ppgtt_spt *s; struct intel_gvt_gtt_entry se, ge; unsigned long i; @@ -1033,34 +1006,22 @@ static int ppgtt_populate_shadow_page(struct intel_vgpu_ppgtt_spt *spt) spt->guest_page.track.gfn, spt->shadow_page.type); if (gtt_type_is_pte_pt(spt->shadow_page.type)) { - for_each_guest_entry(spt, &ge, i) { - if (ops->test_present(&ge)) { - ret = gtt_entry_p2m(vgpu, &ge, &se); - if (ret) - goto fail; - } else { - ppgtt_get_scratch_entry(vgpu, - spt->shadow_page.type, &se); - } + for_each_present_guest_entry(spt, &ge, i) { + ret = gtt_entry_p2m(vgpu, &ge, &se); + if (ret) + goto fail; ppgtt_set_shadow_entry(spt, &se, i); } return 0; } - for_each_guest_entry(spt, &ge, i) { + for_each_present_guest_entry(spt, &ge, i) { if (!gtt_type_is_pt(get_next_pt_type(ge.type))) { gvt_vgpu_err("GVT doesn't support pse bit now\n"); ret = -EINVAL; goto fail; } - if (!ops->test_present(&ge)) { - ppgtt_get_scratch_entry(vgpu, spt->shadow_page.type, - &se); - ppgtt_set_shadow_entry(spt, &se, i); - continue; - } - s = ppgtt_populate_shadow_page_by_guest_entry(vgpu, &ge); if (IS_ERR(s)) { ret = PTR_ERR(s); @@ -1092,7 +1053,7 @@ static int ppgtt_handle_guest_entry_removal(struct intel_vgpu_guest_page *gpt, if (!ops->test_present(se)) return 0; - if (ppgtt_is_scratch_entry(vgpu, se)) + if (ops->get_pfn(se) == vgpu->gtt.scratch_pt[sp->type].page_mfn) return 0; if (gtt_type_is_pt(get_next_pt_type(se->type))) { @@ -1331,6 +1292,7 @@ static int ppgtt_handle_guest_write_page_table( { struct intel_vgpu_ppgtt_spt *spt = guest_page_to_ppgtt_spt(gpt); struct intel_vgpu *vgpu = spt->vgpu; + int type = spt->shadow_page.type; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; struct intel_gvt_gtt_entry se; @@ -1357,7 +1319,7 @@ static int ppgtt_handle_guest_write_page_table( goto fail; if (!new_present) { - ppgtt_get_scratch_entry(vgpu, spt->shadow_page.type, &se); + ops->set_pfn(&se, vgpu->gtt.scratch_pt[type].page_mfn); ppgtt_set_shadow_entry(spt, &se, index); } @@ -2006,85 +1968,106 @@ int intel_vgpu_emulate_gtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off, return ret; } -static void ppgtt_destroy_scratch(struct intel_vgpu *vgpu) -{ - struct intel_vgpu_shadow_page *scratch_page; - int i; - - for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { - scratch_page = &vgpu->gtt.ppgtt_scratch_page[i]; - if (scratch_page->page != NULL) { - clean_shadow_page(vgpu, scratch_page); - __free_page(scratch_page->page); - } - } -} - -static int setup_ppgtt_scratch_page(struct intel_vgpu *vgpu, +static int alloc_scratch_pages(struct intel_vgpu *vgpu, intel_gvt_gtt_type_t type) { - struct intel_gvt *gvt = vgpu->gvt; - struct intel_gvt_device_info *info = &gvt->device_info; - int num_entries = I915_GTT_PAGE_SIZE >> info->gtt_entry_size_shift; struct intel_vgpu_gtt *gtt = &vgpu->gtt; struct intel_gvt_gtt_pte_ops *ops = vgpu->gvt->gtt.pte_ops; - struct intel_vgpu_shadow_page *scratch_page; - struct intel_gvt_gtt_entry e; - intel_gvt_gtt_type_t next_pt_type; - int ret, i; + int page_entry_num = I915_GTT_PAGE_SIZE >> + vgpu->gvt->device_info.gtt_entry_size_shift; + void *scratch_pt; + int i; + struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; if (WARN_ON(type < GTT_TYPE_PPGTT_PTE_PT || type >= GTT_TYPE_MAX)) return -EINVAL; - scratch_page = >t->ppgtt_scratch_page[type]; - - scratch_page->page = alloc_page(GFP_KERNEL); - if (!scratch_page) { + scratch_pt = (void *)get_zeroed_page(GFP_KERNEL); + if (!scratch_pt) { gvt_vgpu_err("fail to allocate scratch page\n"); return -ENOMEM; } - ret = init_shadow_page(vgpu, scratch_page, type, false); - if (ret) { - gvt_vgpu_err("fail to allocate scratch page\n"); - __free_page(scratch_page->page); + daddr = dma_map_page(dev, virt_to_page(scratch_pt), 0, + 4096, PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(dev, daddr)) { + gvt_vgpu_err("fail to dmamap scratch_pt\n"); + __free_page(virt_to_page(scratch_pt)); return -ENOMEM; } - - memset(&e, 0, sizeof(e)); - - if (type == GTT_TYPE_PPGTT_PTE_PT) { - e.type = GTT_TYPE_PPGTT_PTE_4K_ENTRY; - ops->set_pfn(&e, gvt->gtt.scratch_mfn); - } else { - next_pt_type = get_next_pt_type(type); - e.type = get_entry_type(type); - ops->set_pfn(&e, gtt->ppgtt_scratch_page[next_pt_type].mfn); + gtt->scratch_pt[type].page_mfn = + (unsigned long)(daddr >> I915_GTT_PAGE_SHIFT); + gtt->scratch_pt[type].page = virt_to_page(scratch_pt); + gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n", + vgpu->id, type, gtt->scratch_pt[type].page_mfn); + + /* Build the tree by full filled the scratch pt with the entries which + * point to the next level scratch pt or scratch page. The + * scratch_pt[type] indicate the scratch pt/scratch page used by the + * 'type' pt. + * e.g. scratch_pt[GTT_TYPE_PPGTT_PDE_PT] is used by + * GTT_TYPE_PPGTT_PDE_PT level pt, that means this scratch_pt it self + * is GTT_TYPE_PPGTT_PTE_PT, and full filled by scratch page mfn. + */ + if (type > GTT_TYPE_PPGTT_PTE_PT && type < GTT_TYPE_MAX) { + struct intel_gvt_gtt_entry se; + + memset(&se, 0, sizeof(struct intel_gvt_gtt_entry)); + se.type = get_entry_type(type - 1); + ops->set_pfn(&se, gtt->scratch_pt[type - 1].page_mfn); + + /* The entry parameters like present/writeable/cache type + * set to the same as i915's scratch page tree. + */ + se.val64 |= _PAGE_PRESENT | _PAGE_RW; + if (type == GTT_TYPE_PPGTT_PDE_PT) + se.val64 |= PPAT_CACHED; + + for (i = 0; i < page_entry_num; i++) + ops->set_entry(scratch_pt, &se, i, false, 0, vgpu); } - ops->set_present(&e); + return 0; +} + +static int release_scratch_page_tree(struct intel_vgpu *vgpu) +{ + int i; + struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; - for (i = 0; i < num_entries; i++) - ops->set_entry(scratch_page->vaddr, &e, i, false, 0, vgpu); + for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { + if (vgpu->gtt.scratch_pt[i].page != NULL) { + daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn << + I915_GTT_PAGE_SHIFT); + dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); + __free_page(vgpu->gtt.scratch_pt[i].page); + vgpu->gtt.scratch_pt[i].page = NULL; + vgpu->gtt.scratch_pt[i].page_mfn = 0; + } + } return 0; } -static int ppgtt_create_scratch(struct intel_vgpu *vgpu) +static int create_scratch_page_tree(struct intel_vgpu *vgpu) { int i, ret; for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { - ret = setup_ppgtt_scratch_page(vgpu, i); + ret = alloc_scratch_pages(vgpu, i); if (ret) goto err; } + return 0; + err: - ppgtt_destroy_scratch(vgpu); + release_scratch_page_tree(vgpu); return ret; } - + /** * intel_vgpu_init_gtt - initialize per-vGPU graphics memory virulization * @vgpu: a vGPU @@ -2117,7 +2100,8 @@ int intel_vgpu_init_gtt(struct intel_vgpu *vgpu) } gtt->ggtt_mm = ggtt_mm; - return ppgtt_create_scratch(vgpu); + + return create_scratch_page_tree(vgpu); } static void intel_vgpu_free_mm(struct intel_vgpu *vgpu, int type) @@ -2149,7 +2133,7 @@ static void intel_vgpu_free_mm(struct intel_vgpu *vgpu, int type) void intel_vgpu_clean_gtt(struct intel_vgpu *vgpu) { ppgtt_free_all_shadow_page(vgpu); - ppgtt_destroy_scratch(vgpu); + release_scratch_page_tree(vgpu); intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_PPGTT); intel_vgpu_free_mm(vgpu, INTEL_GVT_MM_GGTT); diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h index 416b2f868cf09..f98c1c19b4cb4 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.h +++ b/drivers/gpu/drm/i915/gvt/gtt.h @@ -188,12 +188,9 @@ extern void intel_vgpu_destroy_mm(struct kref *mm_ref); struct intel_vgpu_guest_page; -struct intel_vgpu_shadow_page { - void *vaddr; +struct intel_vgpu_scratch_pt { struct page *page; - int type; - struct hlist_node node; - unsigned long mfn; + unsigned long page_mfn; }; struct intel_vgpu_gtt { @@ -205,7 +202,7 @@ struct intel_vgpu_gtt { atomic_t n_tracked_guest_page; struct list_head oos_page_list_head; struct list_head post_shadow_list_head; - struct intel_vgpu_shadow_page ppgtt_scratch_page[GTT_TYPE_MAX]; + struct intel_vgpu_scratch_pt scratch_pt[GTT_TYPE_MAX]; }; extern int intel_vgpu_init_gtt(struct intel_vgpu *vgpu); @@ -221,6 +218,14 @@ extern struct intel_vgpu_mm *intel_gvt_find_ppgtt_mm(struct intel_vgpu *vgpu, struct intel_vgpu_oos_page; +struct intel_vgpu_shadow_page { + void *vaddr; + struct page *page; + int type; + struct hlist_node node; + unsigned long mfn; +}; + struct intel_vgpu_page_track { struct hlist_node node; bool tracked; -- GitLab From 295764cd2ff41e2c1bc8af4050de77cec5e7a1c0 Mon Sep 17 00:00:00 2001 From: Xiong Zhang Date: Tue, 7 Nov 2017 05:23:02 +0800 Subject: [PATCH 0310/1471] drm/i915/gvt: Limit read hw reg to active vgpu mmio_read_from_hw() let vgpu could read hw reg, if vgpu's workload is running on hw, things is good. Otherwise vgpu will get other vgpu's reg val, it is unsafe. This patch limit such hw access to active vgpu. If vgpu isn't running on hw, the reg read of this vgpu will get the last active val which saved at schedule_out. v2: ring timestamp is walking continuously even if the ring is idle. so read hw directly. (Zhenyu) Signed-off-by: Xiong Zhang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/handlers.c | 26 ++++++++++++++++++++++---- drivers/gpu/drm/i915/gvt/scheduler.c | 15 +++++++++++++++ 2 files changed, 37 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c index 4f8d470dbb3fa..880448d4adc7e 100644 --- a/drivers/gpu/drm/i915/gvt/handlers.c +++ b/drivers/gpu/drm/i915/gvt/handlers.c @@ -1471,11 +1471,29 @@ static int skl_lcpll_write(struct intel_vgpu *vgpu, unsigned int offset, static int mmio_read_from_hw(struct intel_vgpu *vgpu, unsigned int offset, void *p_data, unsigned int bytes) { - struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + struct intel_gvt *gvt = vgpu->gvt; + struct drm_i915_private *dev_priv = gvt->dev_priv; + int ring_id; + u32 ring_base; + + ring_id = intel_gvt_render_mmio_to_ring_id(gvt, offset); + /** + * Read HW reg in following case + * a. the offset isn't a ring mmio + * b. the offset's ring is running on hw. + * c. the offset is ring time stamp mmio + */ + if (ring_id >= 0) + ring_base = dev_priv->engine[ring_id]->mmio_base; + + if (ring_id < 0 || vgpu == gvt->scheduler.engine_owner[ring_id] || + offset == i915_mmio_reg_offset(RING_TIMESTAMP(ring_base)) || + offset == i915_mmio_reg_offset(RING_TIMESTAMP_UDW(ring_base))) { + mmio_hw_access_pre(dev_priv); + vgpu_vreg(vgpu, offset) = I915_READ(_MMIO(offset)); + mmio_hw_access_post(dev_priv); + } - mmio_hw_access_pre(dev_priv); - vgpu_vreg(vgpu, offset) = I915_READ(_MMIO(offset)); - mmio_hw_access_post(dev_priv); return intel_vgpu_default_mmio_read(vgpu, offset, p_data, bytes); } diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 7a1ffaa9ae06d..9749113fccdd5 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -131,6 +131,20 @@ static inline bool is_gvt_request(struct drm_i915_gem_request *req) return i915_gem_context_force_single_submission(req->ctx); } +static void save_ring_hw_state(struct intel_vgpu *vgpu, int ring_id) +{ + struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; + u32 ring_base = dev_priv->engine[ring_id]->mmio_base; + i915_reg_t reg; + + reg = RING_INSTDONE(ring_base); + vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg); + reg = RING_ACTHD(ring_base); + vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg); + reg = RING_ACTHD_UDW(ring_base); + vgpu_vreg(vgpu, i915_mmio_reg_offset(reg)) = I915_READ_FW(reg); +} + static int shadow_context_status_change(struct notifier_block *nb, unsigned long action, void *data) { @@ -175,6 +189,7 @@ static int shadow_context_status_change(struct notifier_block *nb, break; case INTEL_CONTEXT_SCHEDULE_OUT: case INTEL_CONTEXT_SCHEDULE_PREEMPTED: + save_ring_hw_state(workload->vgpu, ring_id); atomic_set(&workload->shadow_ctx_active, 0); break; default: -- GitLab From b2d6ef70614e9e9dacfa9fc7dac49e7dc22dc8b3 Mon Sep 17 00:00:00 2001 From: Xiong Zhang Date: Wed, 8 Nov 2017 00:45:21 +0800 Subject: [PATCH 0311/1471] drm/i915/gvt: Let each vgpu has separate opregion memory Currently every vgpu share a common gvt opregion memory, but it is freed at vgpu destroy, then the later vgpu doesn't have opregion memory once the first vgpu is destroyed. This cause guest function failure like reboot, second or later boot. This patch allocate and init virt opregion memory for each vgpu, so this memory could be freed at vgpu destroy. Signed-off-by: Xiong Zhang Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gvt.c | 9 +-- drivers/gpu/drm/i915/gvt/gvt.h | 10 --- drivers/gpu/drm/i915/gvt/opregion.c | 101 +++++++++++----------------- 3 files changed, 42 insertions(+), 78 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c index cef06d59884e3..3a74a408a9665 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.c +++ b/drivers/gpu/drm/i915/gvt/gvt.c @@ -323,7 +323,6 @@ void intel_gvt_clean_device(struct drm_i915_private *dev_priv) intel_gvt_clean_cmd_parser(gvt); intel_gvt_clean_sched_policy(gvt); intel_gvt_clean_workload_scheduler(gvt); - intel_gvt_clean_opregion(gvt); intel_gvt_clean_gtt(gvt); intel_gvt_clean_irq(gvt); intel_gvt_clean_mmio_info(gvt); @@ -397,13 +396,9 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) if (ret) goto out_clean_irq; - ret = intel_gvt_init_opregion(gvt); - if (ret) - goto out_clean_gtt; - ret = intel_gvt_init_workload_scheduler(gvt); if (ret) - goto out_clean_opregion; + goto out_clean_gtt; ret = intel_gvt_init_sched_policy(gvt); if (ret) @@ -460,8 +455,6 @@ int intel_gvt_init_device(struct drm_i915_private *dev_priv) intel_gvt_clean_sched_policy(gvt); out_clean_workload_scheduler: intel_gvt_clean_workload_scheduler(gvt); -out_clean_opregion: - intel_gvt_clean_opregion(gvt); out_clean_gtt: intel_gvt_clean_gtt(gvt); out_clean_irq: diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h index 27e8186cbc81a..393066726993e 100644 --- a/drivers/gpu/drm/i915/gvt/gvt.h +++ b/drivers/gpu/drm/i915/gvt/gvt.h @@ -125,7 +125,6 @@ struct intel_vgpu_irq { struct intel_vgpu_opregion { void *va; u32 gfn[INTEL_GVT_OPREGION_PAGES]; - struct page *pages[INTEL_GVT_OPREGION_PAGES]; }; #define vgpu_opregion(vgpu) (&(vgpu->opregion)) @@ -265,11 +264,6 @@ struct intel_gvt_firmware { bool firmware_loaded; }; -struct intel_gvt_opregion { - void *opregion_va; - u32 opregion_pa; -}; - #define NR_MAX_INTEL_VGPU_TYPES 20 struct intel_vgpu_type { char name[16]; @@ -293,7 +287,6 @@ struct intel_gvt { struct intel_gvt_firmware firmware; struct intel_gvt_irq irq; struct intel_gvt_gtt gtt; - struct intel_gvt_opregion opregion; struct intel_gvt_workload_scheduler scheduler; struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES]; DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS); @@ -511,9 +504,6 @@ static inline u64 intel_vgpu_get_bar_gpa(struct intel_vgpu *vgpu, int bar) PCI_BASE_ADDRESS_MEM_MASK; } -void intel_gvt_clean_opregion(struct intel_gvt *gvt); -int intel_gvt_init_opregion(struct intel_gvt *gvt); - void intel_vgpu_clean_opregion(struct intel_vgpu *vgpu); int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa); diff --git a/drivers/gpu/drm/i915/gvt/opregion.c b/drivers/gpu/drm/i915/gvt/opregion.c index 2533d1ef1c569..80720e59723ab 100644 --- a/drivers/gpu/drm/i915/gvt/opregion.c +++ b/drivers/gpu/drm/i915/gvt/opregion.c @@ -213,16 +213,55 @@ static void virt_vbt_generation(struct vbt *v) v->driver_features.lvds_config = BDB_DRIVER_FEATURE_NO_LVDS; } +static int alloc_and_init_virt_opregion(struct intel_vgpu *vgpu) +{ + u8 *buf; + struct opregion_header *header; + struct vbt v; + + gvt_dbg_core("init vgpu%d opregion\n", vgpu->id); + vgpu_opregion(vgpu)->va = (void *)__get_free_pages(GFP_KERNEL | + __GFP_ZERO, + get_order(INTEL_GVT_OPREGION_SIZE)); + if (!vgpu_opregion(vgpu)->va) { + gvt_err("fail to get memory for vgpu virt opregion\n"); + return -ENOMEM; + } + + /* emulated opregion with VBT mailbox only */ + buf = (u8 *)vgpu_opregion(vgpu)->va; + header = (struct opregion_header *)buf; + memcpy(header->signature, OPREGION_SIGNATURE, + sizeof(OPREGION_SIGNATURE)); + header->size = 0x8; + header->opregion_ver = 0x02000000; + header->mboxes = MBOX_VBT; + + /* for unknown reason, the value in LID field is incorrect + * which block the windows guest, so workaround it by force + * setting it to "OPEN" + */ + buf[INTEL_GVT_OPREGION_CLID] = 0x3; + + /* emulated vbt from virt vbt generation */ + virt_vbt_generation(&v); + memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt)); + + return 0; +} + static int init_vgpu_opregion(struct intel_vgpu *vgpu, u32 gpa) { - int i; + int i, ret; if (WARN((vgpu_opregion(vgpu)->va), "vgpu%d: opregion has been initialized already.\n", vgpu->id)) return -EINVAL; - vgpu_opregion(vgpu)->va = vgpu->gvt->opregion.opregion_va; + ret = alloc_and_init_virt_opregion(vgpu); + if (ret < 0) + return ret; for (i = 0; i < INTEL_GVT_OPREGION_PAGES; i++) vgpu_opregion(vgpu)->gfn[i] = (gpa >> PAGE_SHIFT) + i; @@ -304,64 +343,6 @@ int intel_vgpu_init_opregion(struct intel_vgpu *vgpu, u32 gpa) return 0; } -/** - * intel_gvt_clean_opregion - clean host opergion related stuffs - * @gvt: a GVT device - * - */ -void intel_gvt_clean_opregion(struct intel_gvt *gvt) -{ - free_pages((unsigned long)gvt->opregion.opregion_va, - get_order(INTEL_GVT_OPREGION_SIZE)); - gvt->opregion.opregion_va = NULL; -} - -/** - * intel_gvt_init_opregion - initialize host opergion related stuffs - * @gvt: a GVT device - * - * Returns: - * Zero on success, negative error code if failed. - */ -int intel_gvt_init_opregion(struct intel_gvt *gvt) -{ - u8 *buf; - struct opregion_header *header; - struct vbt v; - - gvt_dbg_core("init host opregion\n"); - - gvt->opregion.opregion_va = (void *)__get_free_pages(GFP_KERNEL | - __GFP_ZERO, - get_order(INTEL_GVT_OPREGION_SIZE)); - - if (!gvt->opregion.opregion_va) { - gvt_err("fail to get memory for virt opregion\n"); - return -ENOMEM; - } - - /* emulated opregion with VBT mailbox only */ - buf = (u8 *)gvt->opregion.opregion_va; - header = (struct opregion_header *)buf; - memcpy(header->signature, OPREGION_SIGNATURE, - sizeof(OPREGION_SIGNATURE)); - header->size = 0x8; - header->opregion_ver = 0x02000000; - header->mboxes = MBOX_VBT; - - /* for unknown reason, the value in LID field is incorrect - * which block the windows guest, so workaround it by force - * setting it to "OPEN" - */ - buf[INTEL_GVT_OPREGION_CLID] = 0x3; - - /* emulated vbt from virt vbt generation */ - virt_vbt_generation(&v); - memcpy(buf + INTEL_GVT_OPREGION_VBT_OFFSET, &v, sizeof(struct vbt)); - - return 0; -} - #define GVT_OPREGION_FUNC(scic) \ ({ \ u32 __ret; \ -- GitLab From f2880e04f3a5419366926182fc97a3c2e4fd8f2a Mon Sep 17 00:00:00 2001 From: fred gao Date: Tue, 14 Nov 2017 17:09:35 +0800 Subject: [PATCH 0312/1471] drm/i915/gvt: Move request alloc to dispatch_workload path only Previously the performance is improved through the workload auditing and shadowing ahead of vGPU scheduling, however, there is the case that more requests are allocated in submit_context before the previous request is added, the timeline will hold its seqno which is later. This patch is to move the request alloc to dispatch_workload function, where is the same place as request is added. It will fix the issue of kernel BUG for (timeline->seqno != request->fence.seqno) check when add_request. Fixes: 89ea20b930cb ("drm/i915/gvt: Factor out scan and shadow from workload dispatch") Signed-off-by: Chuanxiao Dong Signed-off-by: fred gao Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/scheduler.c | 31 ++++++++++++++++++++++++---- 1 file changed, 27 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index 9749113fccdd5..a742b364c2c3e 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -270,7 +270,6 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; int ring_id = workload->ring_id; struct intel_engine_cs *engine = dev_priv->engine[ring_id]; - struct drm_i915_gem_request *rq; struct intel_ring *ring; int ret; @@ -315,6 +314,27 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) ret = populate_shadow_context(workload); if (ret) goto err_unpin; + workload->shadowed = true; + return 0; + +err_unpin: + engine->context_unpin(engine, shadow_ctx); +err_shadow: + release_shadow_wa_ctx(&workload->wa_ctx); +err_scan: + return ret; +} + +static int intel_gvt_generate_request(struct intel_vgpu_workload *workload) +{ + int ring_id = workload->ring_id; + struct drm_i915_private *dev_priv = workload->vgpu->gvt->dev_priv; + struct intel_engine_cs *engine = dev_priv->engine[ring_id]; + struct drm_i915_gem_request *rq; + struct intel_vgpu *vgpu = workload->vgpu; + struct intel_vgpu_submission *s = &vgpu->submission; + struct i915_gem_context *shadow_ctx = s->shadow_ctx; + int ret; rq = i915_gem_request_alloc(dev_priv->engine[ring_id], shadow_ctx); if (IS_ERR(rq)) { @@ -329,14 +349,11 @@ int intel_gvt_scan_and_shadow_workload(struct intel_vgpu_workload *workload) ret = copy_workload_to_ring_buffer(workload); if (ret) goto err_unpin; - workload->shadowed = true; return 0; err_unpin: engine->context_unpin(engine, shadow_ctx); -err_shadow: release_shadow_wa_ctx(&workload->wa_ctx); -err_scan: return ret; } @@ -496,6 +513,12 @@ static int prepare_workload(struct intel_vgpu_workload *workload) goto err_unpin_mm; } + ret = intel_gvt_generate_request(workload); + if (ret) { + gvt_vgpu_err("fail to generate request\n"); + goto err_unpin_mm; + } + ret = prepare_shadow_batch_buffer(workload); if (ret) { gvt_vgpu_err("fail to prepare_shadow_batch_buffer\n"); -- GitLab From 5926466f543342a098da68b8f771c169dfa8b914 Mon Sep 17 00:00:00 2001 From: Sandy Huang Date: Tue, 14 Nov 2017 19:27:34 +0800 Subject: [PATCH 0313/1471] dt-bindings: add document for rk3126-vop Signed-off-by: Sandy Huang Acked-by: Rob Herring Link: https://patchwork.freedesktop.org/patch/msgid/1510658855-90223-1-git-send-email-hjc@rock-chips.com --- .../devicetree/bindings/display/rockchip/rockchip-vop.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/Documentation/devicetree/bindings/display/rockchip/rockchip-vop.txt b/Documentation/devicetree/bindings/display/rockchip/rockchip-vop.txt index 5d835d9c1ba80..eeda3597011e8 100644 --- a/Documentation/devicetree/bindings/display/rockchip/rockchip-vop.txt +++ b/Documentation/devicetree/bindings/display/rockchip/rockchip-vop.txt @@ -7,6 +7,7 @@ buffer to an external LCD interface. Required properties: - compatible: value should be one of the following "rockchip,rk3036-vop"; + "rockchip,rk3126-vop"; "rockchip,rk3288-vop"; "rockchip,rk3368-vop"; "rockchip,rk3366-vop"; -- GitLab From 460c3b0029923a44a398d4b44d62054c0c249600 Mon Sep 17 00:00:00 2001 From: Sandy Huang Date: Tue, 14 Nov 2017 19:27:37 +0800 Subject: [PATCH 0314/1471] drm/rockchip: vop: add rk3126 vop support RK3126 vop register layout is similar with rk3036, so some feature can reuse with rk3036. RK3126 support two overlay plane and one hwc plane, max output resolution is 1080p. it support IOMMU, and its IOMMU same as rk3288's Signed-off-by: Sandy Huang Reviewed-by: Mark Yao Link: https://patchwork.freedesktop.org/patch/msgid/1510658857-90299-1-git-send-email-hjc@rock-chips.com --- drivers/gpu/drm/rockchip/rockchip_vop_reg.c | 30 +++++++++++++++++++++ drivers/gpu/drm/rockchip/rockchip_vop_reg.h | 6 +++++ 2 files changed, 36 insertions(+) diff --git a/drivers/gpu/drm/rockchip/rockchip_vop_reg.c b/drivers/gpu/drm/rockchip/rockchip_vop_reg.c index 4a39049e901a9..2e4eea3459fe6 100644 --- a/drivers/gpu/drm/rockchip/rockchip_vop_reg.c +++ b/drivers/gpu/drm/rockchip/rockchip_vop_reg.c @@ -149,6 +149,34 @@ static const struct vop_data rk3036_vop = { .win_size = ARRAY_SIZE(rk3036_vop_win_data), }; +static const struct vop_win_phy rk3126_win1_data = { + .data_formats = formats_win_lite, + .nformats = ARRAY_SIZE(formats_win_lite), + .enable = VOP_REG(RK3036_SYS_CTRL, 0x1, 1), + .format = VOP_REG(RK3036_SYS_CTRL, 0x7, 6), + .rb_swap = VOP_REG(RK3036_SYS_CTRL, 0x1, 19), + .dsp_info = VOP_REG(RK3126_WIN1_DSP_INFO, 0x0fff0fff, 0), + .dsp_st = VOP_REG(RK3126_WIN1_DSP_ST, 0x1fff1fff, 0), + .yrgb_mst = VOP_REG(RK3126_WIN1_MST, 0xffffffff, 0), + .yrgb_vir = VOP_REG(RK3036_WIN1_VIR, 0xffff, 0), +}; + +static const struct vop_win_data rk3126_vop_win_data[] = { + { .base = 0x00, .phy = &rk3036_win0_data, + .type = DRM_PLANE_TYPE_PRIMARY }, + { .base = 0x00, .phy = &rk3126_win1_data, + .type = DRM_PLANE_TYPE_CURSOR }, +}; + +static const struct vop_data rk3126_vop = { + .intr = &rk3036_intr, + .common = &rk3036_common, + .modeset = &rk3036_modeset, + .output = &rk3036_output, + .win = rk3126_vop_win_data, + .win_size = ARRAY_SIZE(rk3126_vop_win_data), +}; + static const struct vop_scl_extension rk3288_win_full_scl_ext = { .cbcr_vsd_mode = VOP_REG(RK3288_WIN0_CTRL1, 0x1, 31), .cbcr_vsu_mode = VOP_REG(RK3288_WIN0_CTRL1, 0x1, 30), @@ -510,6 +538,8 @@ static const struct vop_data rk3328_vop = { static const struct of_device_id vop_driver_dt_match[] = { { .compatible = "rockchip,rk3036-vop", .data = &rk3036_vop }, + { .compatible = "rockchip,rk3126-vop", + .data = &rk3126_vop }, { .compatible = "rockchip,rk3288-vop", .data = &rk3288_vop }, { .compatible = "rockchip,rk3368-vop", diff --git a/drivers/gpu/drm/rockchip/rockchip_vop_reg.h b/drivers/gpu/drm/rockchip/rockchip_vop_reg.h index 4a4799ff65dee..f81b510ea99c2 100644 --- a/drivers/gpu/drm/rockchip/rockchip_vop_reg.h +++ b/drivers/gpu/drm/rockchip/rockchip_vop_reg.h @@ -878,4 +878,10 @@ #define RK3036_HWC_LUT_ADDR 0x800 /* rk3036 register definition end */ +/* rk3126 register definition */ +#define RK3126_WIN1_MST 0x4c +#define RK3126_WIN1_DSP_INFO 0x50 +#define RK3126_WIN1_DSP_ST 0x54 +/* rk3126 register definition end */ + #endif /* _ROCKCHIP_VOP_REG_H */ -- GitLab From e7c2e4d368e2efc8789b65d141fb633ebb163a70 Mon Sep 17 00:00:00 2001 From: Marek Szyprowski Date: Thu, 9 Nov 2017 11:28:31 +0100 Subject: [PATCH 0315/1471] drm/bridge/sii8620: filter unsupported modes The maximum pixel clock depends on the version of the connected MHL adapter. Add mode_valid callback to filter out modes with too high pixel clock to avoid failure in mode_fixup later. Signed-off-by: Marek Szyprowski Signed-off-by: Andrzej Hajda Link: https://patchwork.freedesktop.org/patch/msgid/20171109102831.19844-1-m.szyprowski@samsung.com --- drivers/gpu/drm/bridge/sil-sii8620.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/drivers/gpu/drm/bridge/sil-sii8620.c b/drivers/gpu/drm/bridge/sil-sii8620.c index b7eb704d0a8a6..268d66d993374 100644 --- a/drivers/gpu/drm/bridge/sil-sii8620.c +++ b/drivers/gpu/drm/bridge/sil-sii8620.c @@ -2191,6 +2191,19 @@ static void sii8620_detach(struct drm_bridge *bridge) rc_unregister_device(ctx->rc_dev); } +static enum drm_mode_status sii8620_mode_valid(struct drm_bridge *bridge, + const struct drm_display_mode *mode) +{ + struct sii8620 *ctx = bridge_to_sii8620(bridge); + bool can_pack = ctx->devcap[MHL_DCAP_VID_LINK_MODE] & + MHL_DCAP_VID_LINK_PPIXEL; + unsigned int max_pclk = sii8620_is_mhl3(ctx) ? MHL3_MAX_LCLK : + MHL1_MAX_LCLK; + max_pclk /= can_pack ? 2 : 3; + + return (mode->clock > max_pclk) ? MODE_CLOCK_HIGH : MODE_OK; +} + static bool sii8620_mode_fixup(struct drm_bridge *bridge, const struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) @@ -2238,6 +2251,7 @@ static const struct drm_bridge_funcs sii8620_bridge_funcs = { .attach = sii8620_attach, .detach = sii8620_detach, .mode_fixup = sii8620_mode_fixup, + .mode_valid = sii8620_mode_valid, }; static int sii8620_probe(struct i2c_client *client, -- GitLab From 5c632e22c2e9aa45bcf3dc2ee85594dae3781145 Mon Sep 17 00:00:00 2001 From: Maciej Purski Date: Thu, 9 Nov 2017 11:53:42 +0100 Subject: [PATCH 0316/1471] drm/bridge/sii8620: add DVI mode support If the sink device is in HDMI mode, enable infoframe interrupt in scdt irq handle function else call start_video function immediately, because in DVI mode, there is no infoframe interrupt provided. Rename start_hdmi function to start_video and get rid of the old start_video function. In start_video, if the sink is DVI and mode is MHL1 or MHl2, write appropriate values to registers else the path should remain the same as in HDMI mode. Signed-off-by: Maciej Purski Signed-off-by: Andrzej Hajda Link: https://patchwork.freedesktop.org/patch/msgid/1510224822-7732-1-git-send-email-m.purski@samsung.com --- drivers/gpu/drm/bridge/sil-sii8620.c | 36 ++++++++++++++-------------- 1 file changed, 18 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/bridge/sil-sii8620.c b/drivers/gpu/drm/bridge/sil-sii8620.c index 268d66d993374..d58db13502bb6 100644 --- a/drivers/gpu/drm/bridge/sil-sii8620.c +++ b/drivers/gpu/drm/bridge/sil-sii8620.c @@ -1169,8 +1169,18 @@ static void sii8620_set_infoframes(struct sii8620 *ctx) sii8620_write_buf(ctx, REG_TPI_INFO_B0, buf, ret); } -static void sii8620_start_hdmi(struct sii8620 *ctx) +static void sii8620_start_video(struct sii8620 *ctx) { + if (!sii8620_is_mhl3(ctx)) + sii8620_stop_video(ctx); + + if (ctx->sink_type == SINK_DVI && !sii8620_is_mhl3(ctx)) { + sii8620_write(ctx, REG_RX_HDMI_CTRL2, + VAL_RX_HDMI_CTRL2_DEFVAL); + sii8620_write(ctx, REG_TPI_SC, 0); + return; + } + sii8620_write_seq_static(ctx, REG_RX_HDMI_CTRL2, VAL_RX_HDMI_CTRL2_DEFVAL | BIT_RX_HDMI_CTRL2_USE_AV_MUTE, @@ -1229,21 +1239,6 @@ static void sii8620_start_hdmi(struct sii8620 *ctx) sii8620_set_infoframes(ctx); } -static void sii8620_start_video(struct sii8620 *ctx) -{ - if (!sii8620_is_mhl3(ctx)) - sii8620_stop_video(ctx); - - switch (ctx->sink_type) { - case SINK_HDMI: - sii8620_start_hdmi(ctx); - break; - case SINK_DVI: - default: - break; - } -} - static void sii8620_disable_hpd(struct sii8620 *ctx) { sii8620_setbits(ctx, REG_EDID_CTRL, BIT_EDID_CTRL_EDID_PRIME_VALID, 0); @@ -1945,8 +1940,13 @@ static void sii8620_irq_scdt(struct sii8620 *ctx) if (stat & BIT_INTR_SCDT_CHANGE) { u8 cstat = sii8620_readb(ctx, REG_TMDS_CSTAT_P3); - if (cstat & BIT_TMDS_CSTAT_P3_SCDT) - sii8620_scdt_high(ctx); + if (cstat & BIT_TMDS_CSTAT_P3_SCDT) { + if (ctx->sink_type == SINK_HDMI) + /* enable infoframe interrupt */ + sii8620_scdt_high(ctx); + else + sii8620_start_video(ctx); + } } sii8620_write(ctx, REG_INTR5, stat); -- GitLab From d5653ec322ff132c2d6976ca0b909e0a9d1da6af Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Thu, 16 Nov 2017 10:39:54 +0200 Subject: [PATCH 0317/1471] drm/i915: Print the condition causing GEM_BUG_ON It is easier to categorize and debug bugs if the failed condition is in plain sight in the actual dmesg output. Make it so. Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Marta Lofstedt Signed-off-by: Mika Kuoppala Reviewed-by: Joonas Lahtinen Reviewed-by: Marta Lofstedt Link: https://patchwork.freedesktop.org/patch/msgid/20171116083954.3357-1-mika.kuoppala@linux.intel.com --- drivers/gpu/drm/i915/i915_gem.h | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h index ff42b5f0e9812..e920dab7f1b8c 100644 --- a/drivers/gpu/drm/i915/i915_gem.h +++ b/drivers/gpu/drm/i915/i915_gem.h @@ -28,7 +28,11 @@ #include #ifdef CONFIG_DRM_I915_DEBUG_GEM -#define GEM_BUG_ON(expr) BUG_ON(expr) +#define GEM_BUG_ON(condition) do { if (unlikely((condition))) { \ + printk(KERN_ERR "GEM_BUG_ON(%s)\n", __stringify(condition)); \ + BUG(); \ + } \ + } while(0) #define GEM_WARN_ON(expr) WARN_ON(expr) #define GEM_DEBUG_DECL(var) var -- GitLab From c534612e780c4a2c8ef5bfc11583c7d58436baca Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 15 Nov 2017 12:14:58 +0000 Subject: [PATCH 0318/1471] drm/i915: Clear breadcrumb node when cancelling signaling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we call intel_engine_cancel_signaling() to stop reporting when a request is completed via an asynchronous signal, we remove that request from the breadcrumb wait queue. However, we may be concurrently processing that request in the signaler itself, the actual operations on the request's node itself are serialised but we do not actually clear the waiter after removing it from the tree allowing both parties to attempt to do so and corrupting the rbtree. (Previously removing from the breadcrumb wait queue could only be done on behalf of i915_wait_request, so this race could not happen). Reported-by: "He, Bo" Fixes: 9eb143bbec7d ("drm/i915: Allow a request to be cancelled") Signed-off-by: Chris Wilson Cc: "He, Bo" Cc: Tvrtko Ursulin Cc: Michał Winiarski Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171115121458.24655-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_breadcrumbs.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/intel_breadcrumbs.c b/drivers/gpu/drm/i915/intel_breadcrumbs.c index 4c4fbf5f20f9b..5ae2d276f7f3f 100644 --- a/drivers/gpu/drm/i915/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c @@ -549,6 +549,7 @@ static void __intel_engine_remove_wait(struct intel_engine_cs *engine, GEM_BUG_ON(RB_EMPTY_NODE(&wait->node)); rb_erase(&wait->node, &b->waiters); + RB_CLEAR_NODE(&wait->node); out: GEM_BUG_ON(b->irq_wait == wait); -- GitLab From d710fc16ff0749f7a8bd33c69785513fe8b95fb8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 16 Nov 2017 10:50:59 +0000 Subject: [PATCH 0319/1471] drm/i915: Prevent overflow of execbuf.buffer_count and num_cliprects We check whether the multiplies will overflow prior to calling kmalloc_array so that we can respond with -EINVAL for the invalid user arguments rather than treating it as an -ENOMEM that would otherwise occur. However, as Dan Carpenter pointed out, we did an addition on the unsigned int prior to passing to kmalloc_array where it would be promoted to size_t for the calculation, thereby allowing it to overflow and underallocate. v2: buffer_count is currently limited to INT_MAX because we treat it as signaled variable for LUT_HANDLE in eb_lookup_vma v3: Move common checks for eb1/eb2 into the same function v4: Put the check back for nfence*sizeof(user_fence) overflow v5: access_ok uses ULONG_MAX but kvmalloc_array uses SIZE_MAX v6: size_t and unsigned long are not type-equivalent on 32b Reported-by: Dan Carpenter Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171116105059.25142-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 66 ++++++++++++++-------- 1 file changed, 43 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 435ed95df144c..53ccb27bfe917 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -2074,23 +2074,27 @@ static struct drm_syncobj ** get_fence_array(struct drm_i915_gem_execbuffer2 *args, struct drm_file *file) { - const unsigned int nfences = args->num_cliprects; + const unsigned long nfences = args->num_cliprects; struct drm_i915_gem_exec_fence __user *user; struct drm_syncobj **fences; - unsigned int n; + unsigned long n; int err; if (!(args->flags & I915_EXEC_FENCE_ARRAY)) return NULL; - if (nfences > SIZE_MAX / sizeof(*fences)) + /* Check multiplication overflow for access_ok() and kvmalloc_array() */ + BUILD_BUG_ON(sizeof(size_t) > sizeof(unsigned long)); + if (nfences > min_t(unsigned long, + ULONG_MAX / sizeof(*user), + SIZE_MAX / sizeof(*fences))) return ERR_PTR(-EINVAL); user = u64_to_user_ptr(args->cliprects_ptr); - if (!access_ok(VERIFY_READ, user, nfences * 2 * sizeof(u32))) + if (!access_ok(VERIFY_READ, user, nfences * sizeof(*user))) return ERR_PTR(-EFAULT); - fences = kvmalloc_array(args->num_cliprects, sizeof(*fences), + fences = kvmalloc_array(nfences, sizeof(*fences), __GFP_NOWARN | GFP_KERNEL); if (!fences) return ERR_PTR(-ENOMEM); @@ -2447,6 +2451,26 @@ i915_gem_do_execbuffer(struct drm_device *dev, return err; } +static size_t eb_element_size(void) +{ + return (sizeof(struct drm_i915_gem_exec_object2) + + sizeof(struct i915_vma *) + + sizeof(unsigned int)); +} + +static bool check_buffer_count(size_t count) +{ + const size_t sz = eb_element_size(); + + /* + * When using LUT_HANDLE, we impose a limit of INT_MAX for the lookup + * array size (see eb_create()). Otherwise, we can accept an array as + * large as can be addressed (though use large arrays at your peril)! + */ + + return !(count < 1 || count > INT_MAX || count > SIZE_MAX / sz - 1); +} + /* * Legacy execbuffer just creates an exec2 list from the original exec object * list array and passes it to the real function. @@ -2455,18 +2479,16 @@ int i915_gem_execbuffer(struct drm_device *dev, void *data, struct drm_file *file) { - const size_t sz = (sizeof(struct drm_i915_gem_exec_object2) + - sizeof(struct i915_vma *) + - sizeof(unsigned int)); struct drm_i915_gem_execbuffer *args = data; struct drm_i915_gem_execbuffer2 exec2; struct drm_i915_gem_exec_object *exec_list = NULL; struct drm_i915_gem_exec_object2 *exec2_list = NULL; + const size_t count = args->buffer_count; unsigned int i; int err; - if (args->buffer_count < 1 || args->buffer_count > SIZE_MAX / sz - 1) { - DRM_DEBUG("execbuf2 with %d buffers\n", args->buffer_count); + if (!check_buffer_count(count)) { + DRM_DEBUG("execbuf2 with %zd buffers\n", count); return -EINVAL; } @@ -2485,9 +2507,9 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, return -EINVAL; /* Copy in the exec list from userland */ - exec_list = kvmalloc_array(args->buffer_count, sizeof(*exec_list), + exec_list = kvmalloc_array(count, sizeof(*exec_list), __GFP_NOWARN | GFP_KERNEL); - exec2_list = kvmalloc_array(args->buffer_count + 1, sz, + exec2_list = kvmalloc_array(count + 1, eb_element_size(), __GFP_NOWARN | GFP_KERNEL); if (exec_list == NULL || exec2_list == NULL) { DRM_DEBUG("Failed to allocate exec list for %d buffers\n", @@ -2498,7 +2520,7 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, } err = copy_from_user(exec_list, u64_to_user_ptr(args->buffers_ptr), - sizeof(*exec_list) * args->buffer_count); + sizeof(*exec_list) * count); if (err) { DRM_DEBUG("copy %d exec entries failed %d\n", args->buffer_count, err); @@ -2548,16 +2570,14 @@ int i915_gem_execbuffer2(struct drm_device *dev, void *data, struct drm_file *file) { - const size_t sz = (sizeof(struct drm_i915_gem_exec_object2) + - sizeof(struct i915_vma *) + - sizeof(unsigned int)); struct drm_i915_gem_execbuffer2 *args = data; struct drm_i915_gem_exec_object2 *exec2_list; struct drm_syncobj **fences = NULL; + const size_t count = args->buffer_count; int err; - if (args->buffer_count < 1 || args->buffer_count > SIZE_MAX / sz - 1) { - DRM_DEBUG("execbuf2 with %d buffers\n", args->buffer_count); + if (!check_buffer_count(count)) { + DRM_DEBUG("execbuf2 with %zd buffers\n", count); return -EINVAL; } @@ -2565,17 +2585,17 @@ i915_gem_execbuffer2(struct drm_device *dev, void *data, return -EINVAL; /* Allocate an extra slot for use by the command parser */ - exec2_list = kvmalloc_array(args->buffer_count + 1, sz, + exec2_list = kvmalloc_array(count + 1, eb_element_size(), __GFP_NOWARN | GFP_KERNEL); if (exec2_list == NULL) { - DRM_DEBUG("Failed to allocate exec list for %d buffers\n", - args->buffer_count); + DRM_DEBUG("Failed to allocate exec list for %zd buffers\n", + count); return -ENOMEM; } if (copy_from_user(exec2_list, u64_to_user_ptr(args->buffers_ptr), - sizeof(*exec2_list) * args->buffer_count)) { - DRM_DEBUG("copy %d exec entries failed\n", args->buffer_count); + sizeof(*exec2_list) * count)) { + DRM_DEBUG("copy %zd exec entries failed\n", count); kvfree(exec2_list); return -EFAULT; } -- GitLab From c6dce8f140bc19efb04afc8c0d11897ead000946 Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Thu, 16 Nov 2017 19:02:37 +0530 Subject: [PATCH 0320/1471] drm/i915: Update execlists tasklet naming intel_lrc_irq_handler and i915_guc_irq_handler are HW submission related tasklet functions. Name them with "submission_tasklet" suffix and remove intel/i915 prefix as they are static. Also rename irq_tasklet as just tasklet for clarity. v2: s/_bh/_tasklet (Chris) Suggested-by: Michal Wajdeczko Signed-off-by: Sagar Arun Kamble Cc: Michal Wajdeczko Cc: Michal Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Michal Wajdeczko Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/1510839162-25197-2-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_gem.c | 10 +++++----- drivers/gpu/drm/i915/i915_guc_submission.c | 6 +++--- drivers/gpu/drm/i915/i915_irq.c | 2 +- drivers/gpu/drm/i915/intel_engine_cs.c | 2 +- drivers/gpu/drm/i915/intel_lrc.c | 17 +++++++++-------- drivers/gpu/drm/i915/intel_ringbuffer.h | 4 ++-- 6 files changed, 21 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bf8fea7920480..61ba321e9970a 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2933,13 +2933,13 @@ i915_gem_reset_prepare_engine(struct intel_engine_cs *engine) * Prevent request submission to the hardware until we have * completed the reset in i915_gem_reset_finish(). If a request * is completed by one engine, it may then queue a request - * to a second via its engine->irq_tasklet *just* as we are + * to a second via its execlists->tasklet *just* as we are * calling engine->init_hw() and also writing the ELSP. - * Turning off the engine->irq_tasklet until the reset is over + * Turning off the execlists->tasklet until the reset is over * prevents the race. */ - tasklet_kill(&engine->execlists.irq_tasklet); - tasklet_disable(&engine->execlists.irq_tasklet); + tasklet_kill(&engine->execlists.tasklet); + tasklet_disable(&engine->execlists.tasklet); /* * We're using worker to queue preemption requests from the tasklet in @@ -3128,7 +3128,7 @@ void i915_gem_reset(struct drm_i915_private *dev_priv) void i915_gem_reset_finish_engine(struct intel_engine_cs *engine) { - tasklet_enable(&engine->execlists.irq_tasklet); + tasklet_enable(&engine->execlists.tasklet); kthread_unpark(engine->breadcrumbs.signaler); intel_uncore_forcewake_put(engine->i915, FORCEWAKE_ALL); diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 0ba2fc04fe9c1..7f3e63da519b2 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -650,7 +650,7 @@ static void inject_preempt_context(struct work_struct *work) if (WARN_ON(intel_guc_send(guc, data, ARRAY_SIZE(data)))) { execlists_clear_active(&engine->execlists, EXECLISTS_ACTIVE_PREEMPT); - tasklet_schedule(&engine->execlists.irq_tasklet); + tasklet_schedule(&engine->execlists.tasklet); } } @@ -799,7 +799,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) spin_unlock_irq(&engine->timeline->lock); } -static void i915_guc_irq_handler(unsigned long data) +static void guc_submission_tasklet(unsigned long data) { struct intel_engine_cs * const engine = (struct intel_engine_cs *)data; struct intel_engine_execlists * const execlists = &engine->execlists; @@ -1439,7 +1439,7 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) for_each_engine(engine, dev_priv, id) { struct intel_engine_execlists * const execlists = &engine->execlists; - execlists->irq_tasklet.func = i915_guc_irq_handler; + execlists->tasklet.func = guc_submission_tasklet; engine->park = i915_guc_submission_park; engine->unpark = i915_guc_submission_unpark; } diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index ff00e462697ad..4fb183ae7a07e 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1404,7 +1404,7 @@ gen8_cs_irq_handler(struct intel_engine_cs *engine, u32 iir, int test_shift) } if (tasklet) - tasklet_hi_schedule(&execlists->irq_tasklet); + tasklet_hi_schedule(&execlists->tasklet); } static irqreturn_t gen8_gt_irq_ack(struct drm_i915_private *dev_priv, diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index a42b738e79e77..9897c7f78c515 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1585,7 +1585,7 @@ void intel_engines_park(struct drm_i915_private *i915) for_each_engine(engine, i915, id) { /* Flush the residual irq tasklets first. */ intel_engine_disarm_breadcrumbs(engine); - tasklet_kill(&engine->execlists.irq_tasklet); + tasklet_kill(&engine->execlists.tasklet); /* * We are committed now to parking the engines, make sure there diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index ebd9596fe83b7..be6c39adebdf3 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -781,7 +781,7 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) * Check the unread Context Status Buffers and manage the submission of new * contexts to the ELSP accordingly. */ -static void intel_lrc_irq_handler(unsigned long data) +static void execlists_submission_tasklet(unsigned long data) { struct intel_engine_cs * const engine = (struct intel_engine_cs *)data; struct intel_engine_execlists * const execlists = &engine->execlists; @@ -947,7 +947,7 @@ static void insert_request(struct intel_engine_cs *engine, list_add_tail(&pt->link, &ptr_mask_bits(p, 1)->requests); if (ptr_unmask_bits(p, 1)) - tasklet_hi_schedule(&engine->execlists.irq_tasklet); + tasklet_hi_schedule(&engine->execlists.tasklet); } static void execlists_submit_request(struct drm_i915_gem_request *request) @@ -1503,7 +1503,7 @@ static int gen8_init_common_ring(struct intel_engine_cs *engine) /* After a GPU reset, we may have requests to replay */ if (execlists->first) - tasklet_schedule(&execlists->irq_tasklet); + tasklet_schedule(&execlists->tasklet); return 0; } @@ -1881,8 +1881,9 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine) * Tasklet cannot be active at this point due intel_mark_active/idle * so this is just for documentation. */ - if (WARN_ON(test_bit(TASKLET_STATE_SCHED, &engine->execlists.irq_tasklet.state))) - tasklet_kill(&engine->execlists.irq_tasklet); + if (WARN_ON(test_bit(TASKLET_STATE_SCHED, + &engine->execlists.tasklet.state))) + tasklet_kill(&engine->execlists.tasklet); dev_priv = engine->i915; @@ -1906,7 +1907,7 @@ static void execlists_set_default_submission(struct intel_engine_cs *engine) engine->submit_request = execlists_submit_request; engine->cancel_requests = execlists_cancel_requests; engine->schedule = execlists_schedule; - engine->execlists.irq_tasklet.func = intel_lrc_irq_handler; + engine->execlists.tasklet.func = execlists_submission_tasklet; engine->park = NULL; engine->unpark = NULL; @@ -1968,8 +1969,8 @@ logical_ring_setup(struct intel_engine_cs *engine) engine->execlists.fw_domains = fw_domains; - tasklet_init(&engine->execlists.irq_tasklet, - intel_lrc_irq_handler, (unsigned long)engine); + tasklet_init(&engine->execlists.tasklet, + execlists_submission_tasklet, (unsigned long)engine); logical_ring_default_vfuncs(engine); logical_ring_default_irqs(engine); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 15a15cb876a63..c00804ed64c66 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -193,9 +193,9 @@ struct i915_priolist { */ struct intel_engine_execlists { /** - * @irq_tasklet: softirq tasklet for bottom handler + * @tasklet: softirq tasklet for bottom handler */ - struct tasklet_struct irq_tasklet; + struct tasklet_struct tasklet; /** * @default_priolist: priority list for I915_PRIORITY_NORMAL -- GitLab From 6fa1f6fbb18e23dc0f582e296fa83194390a49a6 Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Thu, 16 Nov 2017 19:02:38 +0530 Subject: [PATCH 0321/1471] drm/i915/guc: Update names of submission related static functions i915_guc_submit, i915_guc_dequeue, i915_guc_submission_park and i915_guc_submission_upark are functions internal to GuC submission hence remove "i915_" prefix. Suggested-by: Michal Wajdeczko Signed-off-by: Sagar Arun Kamble Cc: Michal Wajdeczko Cc: Michal Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/1510839162-25197-3-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_guc_submission.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 7f3e63da519b2..b7978a243a5c3 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -683,13 +683,13 @@ static void wait_for_guc_preempt_report(struct intel_engine_cs *engine) } /** - * i915_guc_submit() - Submit commands through GuC + * guc_submit() - Submit commands through GuC * @engine: engine associated with the commands * * The only error here arises if the doorbell hardware isn't functioning * as expected, which really shouln't happen. */ -static void i915_guc_submit(struct intel_engine_cs *engine) +static void guc_submit(struct intel_engine_cs *engine) { struct intel_guc *guc = &engine->i915->guc; struct intel_engine_execlists * const execlists = &engine->execlists; @@ -722,7 +722,7 @@ static void port_assign(struct execlist_port *port, port_set(port, port_pack(i915_gem_request_get(rq), port_count(port))); } -static void i915_guc_dequeue(struct intel_engine_cs *engine) +static void guc_dequeue(struct intel_engine_cs *engine) { struct intel_engine_execlists * const execlists = &engine->execlists; struct execlist_port *port = execlists->port; @@ -793,7 +793,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine) if (submit) { port_assign(port, last); execlists_set_active(execlists, EXECLISTS_ACTIVE_USER); - i915_guc_submit(engine); + guc_submit(engine); } unlock: spin_unlock_irq(&engine->timeline->lock); @@ -831,13 +831,13 @@ static void guc_submission_tasklet(unsigned long data) } if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_PREEMPT)) - i915_guc_dequeue(engine); + guc_dequeue(engine); } /* * Everything below here is concerned with setup & teardown, and is * therefore not part of the somewhat time-critical batch-submission - * path of i915_guc_submit() above. + * path of guc_submit() above. */ /* Check that a doorbell register is in the expected state */ @@ -1381,12 +1381,12 @@ static void guc_interrupts_release(struct drm_i915_private *dev_priv) rps->pm_intrmsk_mbz &= ~ARAT_EXPIRED_INTRMSK; } -static void i915_guc_submission_park(struct intel_engine_cs *engine) +static void guc_submission_park(struct intel_engine_cs *engine) { intel_engine_unpin_breadcrumbs_irq(engine); } -static void i915_guc_submission_unpark(struct intel_engine_cs *engine) +static void guc_submission_unpark(struct intel_engine_cs *engine) { intel_engine_pin_breadcrumbs_irq(engine); } @@ -1440,8 +1440,8 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) for_each_engine(engine, dev_priv, id) { struct intel_engine_execlists * const execlists = &engine->execlists; execlists->tasklet.func = guc_submission_tasklet; - engine->park = i915_guc_submission_park; - engine->unpark = i915_guc_submission_unpark; + engine->park = guc_submission_park; + engine->unpark = guc_submission_unpark; } return 0; -- GitLab From db14d0c5ae7c2d8a39b4dcaa584e7503ace6b18f Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Thu, 16 Nov 2017 19:02:39 +0530 Subject: [PATCH 0322/1471] drm/i915/guc: Update name and prototype of GuC submission interface functions i915 GuC submission is hardware interface and GuC APIs that are not user facing should be named intel_guc* hence we change GuC submission related functions name prefix to intel_guc. Also changed the parameter to these functions to intel_guc struct. v2: Using local guc variable in intel_uc_fini_hw. (Michal Wajdeczko) Rebase. Suggested-by: Chris Wilson Signed-off-by: Sagar Arun Kamble Cc: Michal Wajdeczko Cc: Michal Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Acked-by: Chris Wilson Acked-by: Joonas Lahtinen Acked-by: Oscar Mateo Reviewed-by: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/1510839162-25197-4-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_guc_submission.c | 15 ++++++--------- drivers/gpu/drm/i915/i915_guc_submission.h | 8 ++++---- drivers/gpu/drm/i915/intel_uc.c | 16 +++++++++------- 3 files changed, 19 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index b7978a243a5c3..515505c63d074 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -1253,9 +1253,8 @@ static void guc_preempt_work_destroy(struct intel_guc *guc) * Set up the memory resources to be shared with the GuC (via the GGTT) * at firmware loading time. */ -int i915_guc_submission_init(struct drm_i915_private *dev_priv) +int intel_guc_submission_init(struct intel_guc *guc) { - struct intel_guc *guc = &dev_priv->guc; int ret; if (guc->stage_desc_pool) @@ -1302,10 +1301,8 @@ int i915_guc_submission_init(struct drm_i915_private *dev_priv) return ret; } -void i915_guc_submission_fini(struct drm_i915_private *dev_priv) +void intel_guc_submission_fini(struct intel_guc *guc) { - struct intel_guc *guc = &dev_priv->guc; - guc_ads_destroy(guc); guc_preempt_work_destroy(guc); intel_guc_log_destroy(guc); @@ -1391,9 +1388,9 @@ static void guc_submission_unpark(struct intel_engine_cs *engine) intel_engine_pin_breadcrumbs_irq(engine); } -int i915_guc_submission_enable(struct drm_i915_private *dev_priv) +int intel_guc_submission_enable(struct intel_guc *guc) { - struct intel_guc *guc = &dev_priv->guc; + struct drm_i915_private *dev_priv = guc_to_i915(guc); struct intel_engine_cs *engine; enum intel_engine_id id; int err; @@ -1451,9 +1448,9 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv) return err; } -void i915_guc_submission_disable(struct drm_i915_private *dev_priv) +void intel_guc_submission_disable(struct intel_guc *guc) { - struct intel_guc *guc = &dev_priv->guc; + struct drm_i915_private *dev_priv = guc_to_i915(guc); GEM_BUG_ON(dev_priv->gt.awake); /* GT should be parked first */ diff --git a/drivers/gpu/drm/i915/i915_guc_submission.h b/drivers/gpu/drm/i915/i915_guc_submission.h index cb4353b590595..6bdb8fc00a40c 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.h +++ b/drivers/gpu/drm/i915/i915_guc_submission.h @@ -72,9 +72,9 @@ struct i915_guc_client { u64 submissions[I915_NUM_ENGINES]; }; -int i915_guc_submission_init(struct drm_i915_private *dev_priv); -int i915_guc_submission_enable(struct drm_i915_private *dev_priv); -void i915_guc_submission_disable(struct drm_i915_private *dev_priv); -void i915_guc_submission_fini(struct drm_i915_private *dev_priv); +int intel_guc_submission_init(struct intel_guc *guc); +int intel_guc_submission_enable(struct intel_guc *guc); +void intel_guc_submission_disable(struct intel_guc *guc); +void intel_guc_submission_fini(struct intel_guc *guc); #endif diff --git a/drivers/gpu/drm/i915/intel_uc.c b/drivers/gpu/drm/i915/intel_uc.c index aec295470e0d9..2fb65eb7cf748 100644 --- a/drivers/gpu/drm/i915/intel_uc.c +++ b/drivers/gpu/drm/i915/intel_uc.c @@ -168,7 +168,7 @@ int intel_uc_init_hw(struct drm_i915_private *dev_priv) * This is stuff we need to have available at fw load time * if we are planning to enable submission later */ - ret = i915_guc_submission_init(dev_priv); + ret = intel_guc_submission_init(guc); if (ret) goto err_guc; } @@ -217,7 +217,7 @@ int intel_uc_init_hw(struct drm_i915_private *dev_priv) if (i915_modparams.guc_log_level >= 0) gen9_enable_guc_interrupts(dev_priv); - ret = i915_guc_submission_enable(dev_priv); + ret = intel_guc_submission_enable(guc); if (ret) goto err_interrupts; } @@ -246,7 +246,7 @@ int intel_uc_init_hw(struct drm_i915_private *dev_priv) guc_capture_load_err_log(guc); err_submission: if (i915_modparams.enable_guc_submission) - i915_guc_submission_fini(dev_priv); + intel_guc_submission_fini(guc); err_guc: i915_ggtt_disable_guc(dev_priv); @@ -271,19 +271,21 @@ int intel_uc_init_hw(struct drm_i915_private *dev_priv) void intel_uc_fini_hw(struct drm_i915_private *dev_priv) { - guc_free_load_err_log(&dev_priv->guc); + struct intel_guc *guc = &dev_priv->guc; + + guc_free_load_err_log(guc); if (!i915_modparams.enable_guc_loading) return; if (i915_modparams.enable_guc_submission) - i915_guc_submission_disable(dev_priv); + intel_guc_submission_disable(guc); - guc_disable_communication(&dev_priv->guc); + guc_disable_communication(guc); if (i915_modparams.enable_guc_submission) { gen9_disable_guc_interrupts(dev_priv); - i915_guc_submission_fini(dev_priv); + intel_guc_submission_fini(guc); } i915_ggtt_disable_guc(dev_priv); -- GitLab From 5afc8b49e4f06dbd903987df101db5341575a5ba Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Thu, 16 Nov 2017 19:02:40 +0530 Subject: [PATCH 0323/1471] drm/i915/guc: Rename i915_guc_client struct to intel_guc_client GuC submission clients are currently being used in kernel only hence update the structure name to intel_guc_client. Signed-off-by: Sagar Arun Kamble Cc: Michal Wajdeczko Cc: Michal Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Acked-by: Chris Wilson Acked-by: Joonas Lahtinen Acked-by: Oscar Mateo Reviewed-by: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/1510839162-25197-5-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_debugfs.c | 4 +- drivers/gpu/drm/i915/i915_guc_submission.c | 62 +++++++++++----------- drivers/gpu/drm/i915/i915_guc_submission.h | 2 +- drivers/gpu/drm/i915/intel_guc.h | 6 +-- 4 files changed, 37 insertions(+), 37 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 6ba08b0c1c226..ce023587dd30c 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2433,7 +2433,7 @@ static void i915_guc_log_info(struct seq_file *m, static void i915_guc_client_info(struct seq_file *m, struct drm_i915_private *dev_priv, - struct i915_guc_client *client) + struct intel_guc_client *client) { struct intel_engine_cs *engine; enum intel_engine_id id; @@ -2498,7 +2498,7 @@ static int i915_guc_stage_pool(struct seq_file *m, void *data) struct drm_i915_private *dev_priv = node_to_i915(m->private); const struct intel_guc *guc = &dev_priv->guc; struct guc_stage_desc *desc = guc->stage_desc_pool_vaddr; - struct i915_guc_client *client = guc->execbuf_client; + struct intel_guc_client *client = guc->execbuf_client; unsigned int tmp; int index; diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c index 515505c63d074..22dbc254ee85c 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/i915_guc_submission.c @@ -32,7 +32,7 @@ * DOC: GuC-based command submission * * GuC client: - * A i915_guc_client refers to a submission path through GuC. Currently, there + * A intel_guc_client refers to a submission path through GuC. Currently, there * are two clients. One of them (the execbuf_client) is charged with all * submissions to the GuC, the other one (preempt_client) is responsible for * preempting the execbuf_client. This struct is the owner of a doorbell, a @@ -42,7 +42,7 @@ * GuC stage descriptor: * During initialization, the driver allocates a static pool of 1024 such * descriptors, and shares them with the GuC. - * Currently, there exists a 1:1 mapping between a i915_guc_client and a + * Currently, there exists a 1:1 mapping between a intel_guc_client and a * guc_stage_desc (via the client's stage_id), so effectively only one * gets used. This stage descriptor lets the GuC know about the doorbell, * workqueue and process descriptor. Theoretically, it also lets the GuC @@ -82,13 +82,13 @@ * */ -static inline bool is_high_priority(struct i915_guc_client* client) +static inline bool is_high_priority(struct intel_guc_client *client) { return (client->priority == GUC_CLIENT_PRIORITY_KMD_HIGH || client->priority == GUC_CLIENT_PRIORITY_HIGH); } -static int __reserve_doorbell(struct i915_guc_client *client) +static int __reserve_doorbell(struct intel_guc_client *client) { unsigned long offset; unsigned long end; @@ -120,7 +120,7 @@ static int __reserve_doorbell(struct i915_guc_client *client) return 0; } -static void __unreserve_doorbell(struct i915_guc_client *client) +static void __unreserve_doorbell(struct intel_guc_client *client) { GEM_BUG_ON(client->doorbell_id == GUC_DOORBELL_INVALID); @@ -152,7 +152,7 @@ static int __guc_deallocate_doorbell(struct intel_guc *guc, u32 stage_id) return intel_guc_send(guc, action, ARRAY_SIZE(action)); } -static struct guc_stage_desc *__get_stage_desc(struct i915_guc_client *client) +static struct guc_stage_desc *__get_stage_desc(struct intel_guc_client *client) { struct guc_stage_desc *base = client->guc->stage_desc_pool_vaddr; @@ -166,7 +166,7 @@ static struct guc_stage_desc *__get_stage_desc(struct i915_guc_client *client) * client object which contains the page being used for the doorbell */ -static void __update_doorbell_desc(struct i915_guc_client *client, u16 new_id) +static void __update_doorbell_desc(struct intel_guc_client *client, u16 new_id) { struct guc_stage_desc *desc; @@ -175,12 +175,12 @@ static void __update_doorbell_desc(struct i915_guc_client *client, u16 new_id) desc->db_id = new_id; } -static struct guc_doorbell_info *__get_doorbell(struct i915_guc_client *client) +static struct guc_doorbell_info *__get_doorbell(struct intel_guc_client *client) { return client->vaddr + client->doorbell_offset; } -static bool has_doorbell(struct i915_guc_client *client) +static bool has_doorbell(struct intel_guc_client *client) { if (client->doorbell_id == GUC_DOORBELL_INVALID) return false; @@ -188,7 +188,7 @@ static bool has_doorbell(struct i915_guc_client *client) return test_bit(client->doorbell_id, client->guc->doorbell_bitmap); } -static int __create_doorbell(struct i915_guc_client *client) +static int __create_doorbell(struct intel_guc_client *client) { struct guc_doorbell_info *doorbell; int err; @@ -207,7 +207,7 @@ static int __create_doorbell(struct i915_guc_client *client) return err; } -static int __destroy_doorbell(struct i915_guc_client *client) +static int __destroy_doorbell(struct intel_guc_client *client) { struct drm_i915_private *dev_priv = guc_to_i915(client->guc); struct guc_doorbell_info *doorbell; @@ -228,7 +228,7 @@ static int __destroy_doorbell(struct i915_guc_client *client) return __guc_deallocate_doorbell(client->guc, client->stage_id); } -static int create_doorbell(struct i915_guc_client *client) +static int create_doorbell(struct intel_guc_client *client) { int ret; @@ -250,7 +250,7 @@ static int create_doorbell(struct i915_guc_client *client) return ret; } -static int destroy_doorbell(struct i915_guc_client *client) +static int destroy_doorbell(struct intel_guc_client *client) { int err; @@ -286,7 +286,7 @@ static unsigned long __select_cacheline(struct intel_guc* guc) } static inline struct guc_process_desc * -__get_process_desc(struct i915_guc_client *client) +__get_process_desc(struct intel_guc_client *client) { return client->vaddr + client->proc_desc_offset; } @@ -295,7 +295,7 @@ __get_process_desc(struct i915_guc_client *client) * Initialise the process descriptor shared with the GuC firmware. */ static void guc_proc_desc_init(struct intel_guc *guc, - struct i915_guc_client *client) + struct intel_guc_client *client) { struct guc_process_desc *desc; @@ -355,7 +355,7 @@ static void guc_stage_desc_pool_destroy(struct intel_guc *guc) * write queue, etc). */ static void guc_stage_desc_init(struct intel_guc *guc, - struct i915_guc_client *client) + struct intel_guc_client *client) { struct drm_i915_private *dev_priv = guc_to_i915(guc); struct intel_engine_cs *engine; @@ -436,7 +436,7 @@ static void guc_stage_desc_init(struct intel_guc *guc, } static void guc_stage_desc_fini(struct intel_guc *guc, - struct i915_guc_client *client) + struct intel_guc_client *client) { struct guc_stage_desc *desc; @@ -472,7 +472,7 @@ static void guc_shared_data_destroy(struct intel_guc *guc) } /* Construct a Work Item and append it to the GuC's Work Queue */ -static void guc_wq_item_append(struct i915_guc_client *client, +static void guc_wq_item_append(struct intel_guc_client *client, u32 target_engine, u32 context_desc, u32 ring_tail, u32 fence_id) { @@ -517,7 +517,7 @@ static void guc_wq_item_append(struct i915_guc_client *client, WRITE_ONCE(desc->tail, (wq_off + wqi_size) & (GUC_WQ_SIZE - 1)); } -static void guc_reset_wq(struct i915_guc_client *client) +static void guc_reset_wq(struct intel_guc_client *client) { struct guc_process_desc *desc = __get_process_desc(client); @@ -525,7 +525,7 @@ static void guc_reset_wq(struct i915_guc_client *client) desc->tail = 0; } -static void guc_ring_doorbell(struct i915_guc_client *client) +static void guc_ring_doorbell(struct intel_guc_client *client) { struct guc_doorbell_info *db; u32 cookie; @@ -549,7 +549,7 @@ static void guc_ring_doorbell(struct i915_guc_client *client) static void guc_add_request(struct intel_guc *guc, struct drm_i915_gem_request *rq) { - struct i915_guc_client *client = guc->execbuf_client; + struct intel_guc_client *client = guc->execbuf_client; struct intel_engine_cs *engine = rq->engine; u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(rq->ctx, engine)); u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64); @@ -589,7 +589,7 @@ static void inject_preempt_context(struct work_struct *work) struct intel_engine_cs *engine = preempt_work->engine; struct intel_guc *guc = container_of(preempt_work, typeof(*guc), preempt_work[engine->id]); - struct i915_guc_client *client = guc->preempt_client; + struct intel_guc_client *client = guc->preempt_client; struct guc_stage_desc *stage_desc = __get_stage_desc(client); struct intel_ring *ring = client->owner->engine[engine->id].ring; u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(client->owner, @@ -866,7 +866,7 @@ static bool doorbell_ok(struct intel_guc *guc, u16 db_id) * reloaded the GuC FW) we can use this function to tell the GuC to reassign the * doorbell to the rightful owner. */ -static int __reset_doorbell(struct i915_guc_client* client, u16 db_id) +static int __reset_doorbell(struct intel_guc_client *client, u16 db_id) { int err; @@ -887,7 +887,7 @@ static int __reset_doorbell(struct i915_guc_client* client, u16 db_id) */ static int guc_init_doorbell_hw(struct intel_guc *guc) { - struct i915_guc_client *client = guc->execbuf_client; + struct intel_guc_client *client = guc->execbuf_client; bool recreate_first_client = false; u16 db_id; int ret; @@ -937,7 +937,7 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) } /** - * guc_client_alloc() - Allocate an i915_guc_client + * guc_client_alloc() - Allocate an intel_guc_client * @dev_priv: driver private data structure * @engines: The set of engines to enable for this client * @priority: four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW @@ -947,15 +947,15 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) * @ctx: the context that owns the client (we use the default render * context) * - * Return: An i915_guc_client object if success, else NULL. + * Return: An intel_guc_client object if success, else NULL. */ -static struct i915_guc_client * +static struct intel_guc_client * guc_client_alloc(struct drm_i915_private *dev_priv, u32 engines, u32 priority, struct i915_gem_context *ctx) { - struct i915_guc_client *client; + struct intel_guc_client *client; struct intel_guc *guc = &dev_priv->guc; struct i915_vma *vma; void *vaddr; @@ -1033,7 +1033,7 @@ guc_client_alloc(struct drm_i915_private *dev_priv, return ERR_PTR(ret); } -static void guc_client_free(struct i915_guc_client *client) +static void guc_client_free(struct intel_guc_client *client) { /* * XXX: wait for any outstanding submissions before freeing memory. @@ -1054,7 +1054,7 @@ static void guc_client_free(struct i915_guc_client *client) static int guc_clients_create(struct intel_guc *guc) { struct drm_i915_private *dev_priv = guc_to_i915(guc); - struct i915_guc_client *client; + struct intel_guc_client *client; GEM_BUG_ON(guc->execbuf_client); GEM_BUG_ON(guc->preempt_client); @@ -1086,7 +1086,7 @@ static int guc_clients_create(struct intel_guc *guc) static void guc_clients_destroy(struct intel_guc *guc) { - struct i915_guc_client *client; + struct intel_guc_client *client; client = fetch_and_zero(&guc->execbuf_client); guc_client_free(client); diff --git a/drivers/gpu/drm/i915/i915_guc_submission.h b/drivers/gpu/drm/i915/i915_guc_submission.h index 6bdb8fc00a40c..f8ae258c89d12 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.h +++ b/drivers/gpu/drm/i915/i915_guc_submission.h @@ -52,7 +52,7 @@ struct drm_i915_private; * queue (a circular array of work items), again described in the process * descriptor. Work queue pages are mapped momentarily as required. */ -struct i915_guc_client { +struct intel_guc_client { struct i915_vma *vma; void *vaddr; struct i915_gem_context *owner; diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index 607e025002624..75c4cfefdaff1 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -41,7 +41,7 @@ struct guc_preempt_work { /* * Top level structure of GuC. It handles firmware loading and manages client - * pool and doorbells. intel_guc owns a i915_guc_client to replace the legacy + * pool and doorbells. intel_guc owns a intel_guc_client to replace the legacy * ExecList submission. */ struct intel_guc { @@ -62,8 +62,8 @@ struct intel_guc { struct i915_vma *shared_data; void *shared_data_vaddr; - struct i915_guc_client *execbuf_client; - struct i915_guc_client *preempt_client; + struct intel_guc_client *execbuf_client; + struct intel_guc_client *preempt_client; struct guc_preempt_work preempt_work[I915_NUM_ENGINES]; struct workqueue_struct *preempt_wq; -- GitLab From a269574489d4fd4615562090b130cfc79309176e Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Thu, 16 Nov 2017 19:02:41 +0530 Subject: [PATCH 0324/1471] drm/i915/guc: Rename i915_guc_submission.c|h to intel_guc_submission.c|h With all component structures and functions named appropriately, change the names of GuC submission source files. There were bunch of style issues in guc_submission.c that are highlighted now by checkpatch. Fix those. Update name in Documentation/gpu. (Joonas) v2: Rebase. v3: Rebase. Signed-off-by: Sagar Arun Kamble Cc: Michal Wajdeczko Cc: Michal Winiarski Cc: Chris Wilson Cc: Joonas Lahtinen Acked-by: Chris Wilson Acked-by: Joonas Lahtinen Acked-by: Oscar Mateo Reviewed-by: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/1510839162-25197-6-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- Documentation/gpu/i915.rst | 4 +- drivers/gpu/drm/i915/Makefile | 6 +- drivers/gpu/drm/i915/i915_debugfs.c | 2 +- drivers/gpu/drm/i915/intel_guc.c | 2 +- ...uc_submission.c => intel_guc_submission.c} | 58 ++++++++++++------- ...uc_submission.h => intel_guc_submission.h} | 1 + drivers/gpu/drm/i915/intel_uc.c | 2 +- 7 files changed, 45 insertions(+), 30 deletions(-) rename drivers/gpu/drm/i915/{i915_guc_submission.c => intel_guc_submission.c} (97%) rename drivers/gpu/drm/i915/{i915_guc_submission.h => intel_guc_submission.h} (98%) diff --git a/Documentation/gpu/i915.rst b/Documentation/gpu/i915.rst index 2e7ee0313c1cd..21577eabaf785 100644 --- a/Documentation/gpu/i915.rst +++ b/Documentation/gpu/i915.rst @@ -350,10 +350,10 @@ GuC-specific firmware loader GuC-based command submission ---------------------------- -.. kernel-doc:: drivers/gpu/drm/i915/i915_guc_submission.c +.. kernel-doc:: drivers/gpu/drm/i915/intel_guc_submission.c :doc: GuC-based command submission -.. kernel-doc:: drivers/gpu/drm/i915/i915_guc_submission.c +.. kernel-doc:: drivers/gpu/drm/i915/intel_guc_submission.c :internal: GuC Firmware Layout diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index f7afd44214b51..c3649ec5b041c 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -82,10 +82,10 @@ i915-y += intel_uc.o \ intel_uc_fw.o \ intel_guc.o \ intel_guc_ct.o \ - intel_guc_log.o \ intel_guc_fw.o \ - intel_huc.o \ - i915_guc_submission.o + intel_guc_log.o \ + intel_guc_submission.o \ + intel_huc.o # autogenerated null render state i915-y += intel_renderstate_gen6.o \ diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index ce023587dd30c..ff8f508a76614 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -30,7 +30,7 @@ #include #include #include "intel_drv.h" -#include "i915_guc_submission.h" +#include "intel_guc_submission.h" static inline struct drm_i915_private *node_to_i915(struct drm_info_node *node) { diff --git a/drivers/gpu/drm/i915/intel_guc.c b/drivers/gpu/drm/i915/intel_guc.c index 9678630a1c705..823d0c2e9ad2f 100644 --- a/drivers/gpu/drm/i915/intel_guc.c +++ b/drivers/gpu/drm/i915/intel_guc.c @@ -23,8 +23,8 @@ */ #include "intel_guc.h" +#include "intel_guc_submission.h" #include "i915_drv.h" -#include "i915_guc_submission.h" static void gen8_guc_raise_irq(struct intel_guc *guc) { diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c similarity index 97% rename from drivers/gpu/drm/i915/i915_guc_submission.c rename to drivers/gpu/drm/i915/intel_guc_submission.c index 22dbc254ee85c..9c9e6edfd5be5 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -25,7 +25,7 @@ #include #include -#include "i915_guc_submission.h" +#include "intel_guc_submission.h" #include "i915_drv.h" /** @@ -102,7 +102,7 @@ static int __reserve_doorbell(struct intel_guc_client *client) * priority contexts, the second half for high-priority ones. */ offset = 0; - end = GUC_NUM_DOORBELLS/2; + end = GUC_NUM_DOORBELLS / 2; if (is_high_priority(client)) { offset = end; end += offset; @@ -221,7 +221,8 @@ static int __destroy_doorbell(struct intel_guc_client *client) /* Doorbell release flow requires that we wait for GEN8_DRB_VALID bit * to go to zero after updating db_status before we call the GuC to - * release the doorbell */ + * release the doorbell + */ if (wait_for_us(!(I915_READ(GEN8_DRBREGL(db_id)) & GEN8_DRB_VALID), 10)) WARN_ONCE(true, "Doorbell never became invalid after disable\n"); @@ -270,7 +271,7 @@ static int destroy_doorbell(struct intel_guc_client *client) return 0; } -static unsigned long __select_cacheline(struct intel_guc* guc) +static unsigned long __select_cacheline(struct intel_guc *guc) { unsigned long offset; @@ -281,7 +282,7 @@ static unsigned long __select_cacheline(struct intel_guc* guc) guc->db_cacheline += cache_line_size(); DRM_DEBUG_DRIVER("reserved cacheline 0x%lx, next 0x%x, linesize %u\n", - offset, guc->db_cacheline, cache_line_size()); + offset, guc->db_cacheline, cache_line_size()); return offset; } @@ -367,7 +368,8 @@ static void guc_stage_desc_init(struct intel_guc *guc, desc = __get_stage_desc(client); memset(desc, 0, sizeof(*desc)); - desc->attribute = GUC_STAGE_DESC_ATTR_ACTIVE | GUC_STAGE_DESC_ATTR_KERNEL; + desc->attribute = GUC_STAGE_DESC_ATTR_ACTIVE | + GUC_STAGE_DESC_ATTR_KERNEL; if (is_high_priority(client)) desc->attribute |= GUC_STAGE_DESC_ATTR_PREEMPT; desc->stage_id = client->stage_id; @@ -403,7 +405,8 @@ static void guc_stage_desc_init(struct intel_guc *guc, guc_ggtt_offset(ce->state) + LRC_STATE_PN * PAGE_SIZE; /* XXX: In direct submission, the GuC wants the HW context id - * here. In proxy submission, it wants the stage id */ + * here. In proxy submission, it wants the stage id + */ lrc->context_id = (client->stage_id << GUC_ELC_CTXID_OFFSET) | (guc_engine_id << GUC_ELC_ENGINE_OFFSET); @@ -416,7 +419,7 @@ static void guc_stage_desc_init(struct intel_guc *guc, } DRM_DEBUG_DRIVER("Host engines 0x%x => GuC engines used 0x%x\n", - client->engines, desc->engines_used); + client->engines, desc->engines_used); WARN_ON(desc->engines_used == 0); /* @@ -551,7 +554,8 @@ static void guc_add_request(struct intel_guc *guc, { struct intel_guc_client *client = guc->execbuf_client; struct intel_engine_cs *engine = rq->engine; - u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(rq->ctx, engine)); + u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(rq->ctx, + engine)); u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64); spin_lock(&client->wq_lock); @@ -777,7 +781,8 @@ static void guc_dequeue(struct intel_engine_cs *engine) INIT_LIST_HEAD(&rq->priotree.link); __i915_gem_request_submit(rq); - trace_i915_gem_request_in(rq, port_index(port, execlists)); + trace_i915_gem_request_in(rq, + port_index(port, execlists)); last = rq; submit = true; } @@ -910,7 +915,8 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) if (recreate_first_client) { ret = __reserve_doorbell(client); if (unlikely(ret)) { - DRM_ERROR("Couldn't re-reserve first client db: %d\n", ret); + DRM_ERROR("Couldn't re-reserve first client db: %d\n", + ret); return ret; } @@ -918,7 +924,8 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) } /* Now for every client (and not only execbuf_client) make sure their - * doorbells are known by the GuC */ + * doorbells are known by the GuC + */ ret = __create_doorbell(guc->execbuf_client); if (ret) return ret; @@ -941,11 +948,11 @@ static int guc_init_doorbell_hw(struct intel_guc *guc) * @dev_priv: driver private data structure * @engines: The set of engines to enable for this client * @priority: four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW - * The kernel client to replace ExecList submission is created with - * NORMAL priority. Priority of a client for scheduler can be HIGH, - * while a preemption context can use CRITICAL. + * The kernel client to replace ExecList submission is created with + * NORMAL priority. Priority of a client for scheduler can be HIGH, + * while a preemption context can use CRITICAL. * @ctx: the context that owns the client (we use the default render - * context) + * context) * * Return: An intel_guc_client object if success, else NULL. */ @@ -973,7 +980,7 @@ guc_client_alloc(struct drm_i915_private *dev_priv, spin_lock_init(&client->wq_lock); ret = ida_simple_get(&guc->stage_ids, 0, GUC_MAX_STAGE_DESCRIPTORS, - GFP_KERNEL); + GFP_KERNEL); if (ret < 0) goto err_client; @@ -1042,7 +1049,8 @@ static void guc_client_free(struct intel_guc_client *client) /* FIXME: in many cases, by the time we get here the GuC has been * reset, so we cannot destroy the doorbell properly. Ignore the - * error message for now */ + * error message for now + */ destroy_doorbell(client); guc_stage_desc_fini(client->guc, client); i915_gem_object_unpin_map(client->vma->obj); @@ -1178,7 +1186,8 @@ static int guc_ads_create(struct intel_guc *guc) * because our GuC shared data is there. */ blob->ads.golden_context_lrca = - guc_ggtt_offset(dev_priv->kernel_context->engine[RCS].state) + skipped_offset; + guc_ggtt_offset(dev_priv->kernel_context->engine[RCS].state) + + skipped_offset; /* * The GuC expects us to exclude the portion of the context image that @@ -1187,7 +1196,8 @@ static int guc_ads_create(struct intel_guc *guc) * dwords). Weird guc is weird. */ for_each_engine(engine, dev_priv, id) - blob->ads.eng_state_size[engine->guc_id] = engine->context_size - skipped_size; + blob->ads.eng_state_size[engine->guc_id] = + engine->context_size - skipped_size; base = guc_ggtt_offset(vma); blob->ads.scheduler_policies = base + ptr_offset(blob, policies); @@ -1317,7 +1327,9 @@ static void guc_interrupts_capture(struct drm_i915_private *dev_priv) enum intel_engine_id id; int irqs; - /* tell all command streamers to forward interrupts (but not vblank) to GuC */ + /* tell all command streamers to forward interrupts (but not vblank) + * to GuC + */ irqs = _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING); for_each_engine(engine, dev_priv, id) I915_WRITE(RING_MODE_GEN7(engine), irqs); @@ -1435,7 +1447,9 @@ int intel_guc_submission_enable(struct intel_guc *guc) guc_interrupts_capture(dev_priv); for_each_engine(engine, dev_priv, id) { - struct intel_engine_execlists * const execlists = &engine->execlists; + struct intel_engine_execlists * const execlists = + &engine->execlists; + execlists->tasklet.func = guc_submission_tasklet; engine->park = guc_submission_park; engine->unpark = guc_submission_unpark; diff --git a/drivers/gpu/drm/i915/i915_guc_submission.h b/drivers/gpu/drm/i915/intel_guc_submission.h similarity index 98% rename from drivers/gpu/drm/i915/i915_guc_submission.h rename to drivers/gpu/drm/i915/intel_guc_submission.h index f8ae258c89d12..e901192ee4697 100644 --- a/drivers/gpu/drm/i915/i915_guc_submission.h +++ b/drivers/gpu/drm/i915/intel_guc_submission.h @@ -67,6 +67,7 @@ struct intel_guc_client { u16 doorbell_id; unsigned long doorbell_offset; + /* Protects GuC client's WQ access */ spinlock_t wq_lock; /* Per-engine counts of GuC submissions */ u64 submissions[I915_NUM_ENGINES]; diff --git a/drivers/gpu/drm/i915/intel_uc.c b/drivers/gpu/drm/i915/intel_uc.c index 2fb65eb7cf748..1e2a30a40ede3 100644 --- a/drivers/gpu/drm/i915/intel_uc.c +++ b/drivers/gpu/drm/i915/intel_uc.c @@ -23,8 +23,8 @@ */ #include "intel_uc.h" +#include "intel_guc_submission.h" #include "i915_drv.h" -#include "i915_guc_submission.h" /* Reset GuC providing us with fresh state for both GuC and HuC. */ -- GitLab From 738a8143866c4cdbe4c008f1583a0c55338266b2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 15 Nov 2017 22:04:42 +0200 Subject: [PATCH 0325/1471] drm/i915: Don't sanitize frame start delay if the pipe is off MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Avoid touching PIPECONF in intel_sanitize_crtc() unless the pipe is actually on. Should cure some unclaimed register accesses during reset, as we are rather cavalier in our approach to powerdomain management. We don't have to sanitize this if the pipe is off since we will overwrite the frame start delay anyway when turning the pipe on. v2: Amended commit message to implicate the reset path (Chris) Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102249 Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171115200442.15051-1-ville.syrjala@linux.intel.com Reviewed-by: Chris Wilson --- drivers/gpu/drm/i915/intel_display.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 23aa7191024ef..e6fcbc5abc757 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -14812,7 +14812,7 @@ static void intel_sanitize_crtc(struct intel_crtc *crtc, enum transcoder cpu_transcoder = crtc->config->cpu_transcoder; /* Clear any frame start delays used for debugging left by the BIOS */ - if (!transcoder_is_dsi(cpu_transcoder)) { + if (crtc->active && !transcoder_is_dsi(cpu_transcoder)) { i915_reg_t reg = PIPECONF(cpu_transcoder); I915_WRITE(reg, -- GitLab From 4fe95b042d9d36744f6848069a92759c10dd2a0e Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 15 Nov 2017 15:25:58 +0000 Subject: [PATCH 0326/1471] drm/i915/selftests: exercise_ggtt may have nothing to do When operating on the live_ggtt we have to find a usuable hole for our test. It is possible for there to be no hole we can use, so initialise the err to 0 for the early exit. Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171115152558.31252-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index d9560d8a6cc87..3dcf886a2802a 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -958,7 +958,7 @@ static int exercise_ggtt(struct drm_i915_private *i915, u64 hole_start, hole_end, last = 0; struct drm_mm_node *node; IGT_TIMEOUT(end_time); - int err = -ENODEV; + int err = 0; mutex_lock(&i915->drm.struct_mutex); restart: -- GitLab From ec2f343e720eca5ec7015b4d2b5fc2703604835e Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 14 Nov 2017 11:47:53 -0800 Subject: [PATCH 0327/1471] drm/i915/cnl: Remove spurious central_freq. "Display software must leave this field at the default value. It no longer needs to be configured as part of PLL programming." We respect this already and we are setting up the default one line below: "DPLL_CFGCR1_CENTRAL_FREQ". Also we don't touch anywhere else this central_freq for cnl. So let's remove from the final write. No functional change. Only a clean-up patch. Cc: Manasi Navare Cc: Mika Kahola Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Manasi Navare Link: https://patchwork.freedesktop.org/patch/msgid/20171114194759.24541-2-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index be74d4767c8a4..61c684ac47af4 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2265,7 +2265,6 @@ static bool cnl_ddi_hdmi_pll_dividers(struct intel_crtc *crtc, DPLL_CFGCR1_QDIV_MODE(wrpll_params.qdiv_mode) | DPLL_CFGCR1_KDIV(wrpll_params.kdiv) | DPLL_CFGCR1_PDIV(wrpll_params.pdiv) | - wrpll_params.central_freq | DPLL_CFGCR1_CENTRAL_FREQ; memset(&crtc_state->dpll_hw_state, 0, -- GitLab From ecc2069a025e34b90033602a36d4cc171387a950 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 14 Nov 2017 11:47:54 -0800 Subject: [PATCH 0328/1471] drm/i915/cnl: Remove useless conversion. No functional change. Just starting the wrpll fixes with a clean-up to make units a bit more clear. Cc: Mika Kahola Cc: Manasi Navare Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Manasi Navare Link: https://patchwork.freedesktop.org/patch/msgid/20171114194759.24541-3-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index 61c684ac47af4..db7afd3144622 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2198,11 +2198,11 @@ static void cnl_wrpll_params_populate(struct skl_wrpll_params *params, uint32_t } static bool -cnl_ddi_calculate_wrpll(int clock /* in Hz */, +cnl_ddi_calculate_wrpll(int clock, struct drm_i915_private *dev_priv, struct skl_wrpll_params *wrpll_params) { - uint64_t afe_clock = clock * 5 / KHz(1); /* clocks in kHz */ + uint64_t afe_clock = clock * 5; unsigned int dco_min = 7998 * KHz(1); unsigned int dco_max = 10000 * KHz(1); unsigned int dco_mid = (dco_min + dco_max) / 2; @@ -2255,7 +2255,7 @@ static bool cnl_ddi_hdmi_pll_dividers(struct intel_crtc *crtc, cfgcr0 = DPLL_CFGCR0_HDMI_MODE; - if (!cnl_ddi_calculate_wrpll(clock * 1000, dev_priv, &wrpll_params)) + if (!cnl_ddi_calculate_wrpll(clock, dev_priv, &wrpll_params)) return false; cfgcr0 |= DPLL_CFGCR0_DCO_FRACTION(wrpll_params.dco_fraction) | -- GitLab From 5eca81de88927f3e43fdd6e90e8fa5204be996b7 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 14 Nov 2017 11:47:55 -0800 Subject: [PATCH 0329/1471] drm/i915/cnl: Fix, simplify and unify wrpll variable sizes. - 64 bits is not needed for afe_clock now we don't convert that to Hz. - 16 bits is not enough for all dco stuff. - unsigned is not relevant/needed for all divisors values. Cc: Mika Kahola Cc: Manasi Navare Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Manasi Navare Date: Tue, 14 Nov 2017 15:42:23 -0800 Subject: [PATCH 0330/1471] drm/i915/cnl: Fix wrpll math for higher freqs. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Spec describe all values in MHz. We handle our clocks in KHz. This includes the best_dco_centrality that was forgot in the same unity as spec. Consequently we couldn't get a good divider for high frequenies. Hence HDMI 2.0 wasn't working. Spec tells 999999 for initial best_dco_centrality meaning the max value in MHz. Since we convert dco from MHz to KHz we also need to convert this initial best_doc_centrality to 999999000 or 999999999 or even better, to the max that its variable allow. This patch also replaces the use of "* KHz(1)" with the values directly on KHz to avoid future confusion. v2: Use U32_MAX instead of random 99999 as spec tells. (Ville). Cc: Ville Syrjälä Cc: Shashank Sharma Cc: Mika Kahola Cc: Manasi Navare Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171114234223.10600-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index fba969cbda371..6cc12abdd39ed 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2201,8 +2201,8 @@ cnl_ddi_calculate_wrpll(int clock, struct skl_wrpll_params *wrpll_params) { u32 afe_clock = clock * 5; - u32 dco_min = 7998 * KHz(1); - u32 dco_max = 10000 * KHz(1); + u32 dco_min = 7998000; + u32 dco_max = 10000000; u32 dco_mid = (dco_min + dco_max) / 2; static const int dividers[] = { 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30, 32, 36, 40, @@ -2211,7 +2211,7 @@ cnl_ddi_calculate_wrpll(int clock, 84, 88, 90, 92, 96, 98, 100, 102, 3, 5, 7, 9, 15, 21 }; u32 dco, best_dco = 0, dco_centrality = 0; - u32 best_dco_centrality = 999999; + u32 best_dco_centrality = U32_MAX; /* Spec meaning of 999999 MHz */ int d, best_div = 0, pdiv = 0, qdiv = 0, kdiv = 0; for (d = 0; d < ARRAY_SIZE(dividers); d++) { -- GitLab From cacf6fe7c6c4ac9e748f41d0de4293b2aea5ae07 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Tue, 14 Nov 2017 11:47:57 -0800 Subject: [PATCH 0331/1471] drm/i915/cnl: Don't blindly replace qdiv. Accordingly to spec "If Kdiv != 2, then Qdiv must be 1." but we already handle qdiv values properly and this case here should be spurious. But instead of blindly replacing let's warn loudly instead. Because it means something was really wrong on initial setup. Cc: Mika Kahola Cc: Manasi Navare Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Manasi Navare Link: https://patchwork.freedesktop.org/patch/msgid/20171114194759.24541-6-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index 6cc12abdd39ed..361b7102b602b 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2184,8 +2184,7 @@ static void cnl_wrpll_params_populate(struct skl_wrpll_params *params, WARN(1, "Incorrect PDiv\n"); } - if (kdiv != 2) - qdiv = 1; + WARN_ON(kdiv != 2 && qdiv != 1); params->qdiv_ratio = qdiv; params->qdiv_mode = (qdiv == 1) ? 0 : 1; -- GitLab From 8a00678a096445716861d2c0bd2a7007ad373f4d Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 15 Nov 2017 10:42:57 -0800 Subject: [PATCH 0332/1471] drm/i915/cnl: Simplify dco_fraction calculation. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit I confess I never fully understood that previous calculation, so this is not a "fix". But let's simplify this math so poor brains like mine can read and make some sense of it in the future. v2: Don't follow the spec since that gives invalid values and it is also confusing. This Ville's version is much simpler. v3: Use u64 cast instead of declaring a u64 dco. (Ville). Cc: Ville Syrjälä Cc: Mika Kahola Cc: Manasi Navare Cc: James Ausmus Signed-off-by: Rodrigo Vivi Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171115184257.8633-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_dpll_mgr.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dpll_mgr.c b/drivers/gpu/drm/i915/intel_dpll_mgr.c index 361b7102b602b..51c5ae4e91164 100644 --- a/drivers/gpu/drm/i915/intel_dpll_mgr.c +++ b/drivers/gpu/drm/i915/intel_dpll_mgr.c @@ -2153,6 +2153,8 @@ static void cnl_wrpll_params_populate(struct skl_wrpll_params *params, u32 dco_freq, u32 ref_freq, int pdiv, int qdiv, int kdiv) { + u32 dco; + switch (kdiv) { case 1: params->kdiv = 1; @@ -2189,9 +2191,10 @@ static void cnl_wrpll_params_populate(struct skl_wrpll_params *params, params->qdiv_ratio = qdiv; params->qdiv_mode = (qdiv == 1) ? 0 : 1; - params->dco_integer = div_u64(dco_freq, ref_freq); - params->dco_fraction = div_u64((div_u64((uint64_t)dco_freq<<15, (uint64_t)ref_freq) - - ((uint64_t)params->dco_integer<<15)) * 0x8000, 0x8000); + dco = div_u64((u64)dco_freq << 15, ref_freq); + + params->dco_integer = dco >> 15; + params->dco_fraction = dco & 0x7fff; } static bool -- GitLab From 9672a69c761468ec920a8bc2442b8397b20578f8 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Wed, 15 Nov 2017 10:42:05 -0800 Subject: [PATCH 0333/1471] drm/i915/cnl: Extend HDMI 2.0 support to CNL. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Starting on GLK we support HDMI 2.0. So this patch only extend the work Shashank has made to GLK to CNL. v2: The version that compiles :/ v3: Invert order to newer || older platforms check. (Ville). Cc: Ville Syrjälä Cc: Paulo Zanoni Cc: Shashank Sharma Cc: Manasi Navare Signed-off-by: Rodrigo Vivi Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171115184205.8104-1-rodrigo.vivi@intel.com --- drivers/gpu/drm/i915/intel_hdmi.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 2d95db64cdf28..9d5e727284754 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -1235,7 +1235,7 @@ static int intel_hdmi_source_max_tmds_clock(struct intel_encoder *encoder) &dev_priv->vbt.ddi_port_info[encoder->port]; int max_tmds_clock; - if (IS_GEMINILAKE(dev_priv)) + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) max_tmds_clock = 594000; else if (INTEL_GEN(dev_priv) >= 8 || IS_HASWELL(dev_priv)) max_tmds_clock = 300000; @@ -1511,7 +1511,8 @@ bool intel_hdmi_compute_config(struct intel_encoder *encoder, pipe_config->lane_count = 4; - if (scdc->scrambling.supported && IS_GEMINILAKE(dev_priv)) { + if (scdc->scrambling.supported && (INTEL_GEN(dev_priv) >= 10 || + IS_GEMINILAKE(dev_priv))) { if (scdc->scrambling.low_rates) pipe_config->hdmi_scrambling = true; @@ -2033,7 +2034,7 @@ void intel_hdmi_init_connector(struct intel_digital_port *intel_dig_port, connector->doublescan_allowed = 0; connector->stereo_allowed = 1; - if (IS_GEMINILAKE(dev_priv)) + if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) connector->ycbcr_420_allowed = true; intel_hdmi->ddc_bus = intel_hdmi_ddc_pin(dev_priv, port); -- GitLab From 55bd6bd75717ca67b790cc1a457cab92adcce148 Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Thu, 16 Nov 2017 14:06:31 -0800 Subject: [PATCH 0334/1471] drm/i915/selftests: Add a GuC doorbells selftest The first test aims to check guc_init_doorbell_hw, changing the existing guc clients and doorbells state before calling it. The second test tries to create as many clients as it is currently possible (currently limited to max number of doorbells) and exercise the doorbell alloc/dealloc code. Since our usage mode require very few clients/doorbells, this code has been exercised very lightly and it's good to have a simple test for it. As reference, this test already helped identify the bug fixed by commit 7f1ea2ac3017 ("drm/i915/guc: Fix doorbell id selection"). v2: Extend number of clients; check for client allocation failure when number of doorbells is exceeded; validate client properties; reuse guc_init_doorbell_hw (Chris). v3: guc_init_doorbell_hw test added per Chris suggestion. v4: Try to explain why guc_init_doorbell_hw exist and comment some details in the subtest. v5: Remove redundant pr_info at the beginning of each subtest (Chris); rebase (s/i915_guc_client/intel_guc_client/). Signed-off-by: Michel Thierry Cc: Michal Wajdeczko Cc: Daniele Ceraolo Spurio Cc: Chris Wilson Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171116220632.1909-1-michel.thierry@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_submission.c | 4 + .../drm/i915/selftests/i915_live_selftests.h | 1 + drivers/gpu/drm/i915/selftests/intel_guc.c | 367 ++++++++++++++++++ 3 files changed, 372 insertions(+) create mode 100644 drivers/gpu/drm/i915/selftests/intel_guc.c diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index 9c9e6edfd5be5..cbf5a96f58068 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -1475,3 +1475,7 @@ void intel_guc_submission_disable(struct intel_guc *guc) guc_clients_destroy(guc); } + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftests/intel_guc.c" +#endif diff --git a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h index 54a73534b37e8..1b750e92dd08e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h @@ -19,3 +19,4 @@ selftest(evict, i915_gem_evict_live_selftests) selftest(hugepages, i915_gem_huge_page_live_selftests) selftest(contexts, i915_gem_context_live_selftests) selftest(hangcheck, intel_hangcheck_live_selftests) +selftest(guc, intel_guc_live_selftest) diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c new file mode 100644 index 0000000000000..f10029e18820b --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/intel_guc.c @@ -0,0 +1,367 @@ +/* + * Copyright © 2017 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#include "../i915_selftest.h" + +/* max doorbell number + negative test for each client type */ +#define ATTEMPTS (GUC_NUM_DOORBELLS + GUC_CLIENT_PRIORITY_NUM) + +struct intel_guc_client *clients[ATTEMPTS]; + +static bool available_dbs(struct intel_guc *guc, u32 priority) +{ + unsigned long offset; + unsigned long end; + u16 id; + + /* first half is used for normal priority, second half for high */ + offset = 0; + end = GUC_NUM_DOORBELLS / 2; + if (priority <= GUC_CLIENT_PRIORITY_HIGH) { + offset = end; + end += offset; + } + + id = find_next_zero_bit(guc->doorbell_bitmap, end, offset); + if (id < end) + return true; + + return false; +} + +static int check_all_doorbells(struct intel_guc *guc) +{ + u16 db_id; + + pr_info_once("Max number of doorbells: %d", GUC_NUM_DOORBELLS); + for (db_id = 0; db_id < GUC_NUM_DOORBELLS; ++db_id) { + if (!doorbell_ok(guc, db_id)) { + pr_err("doorbell %d, not ok\n", db_id); + return -EIO; + } + } + + return 0; +} + +/* + * Basic client sanity check, handy to validate create_clients. + */ +static int validate_client(struct intel_guc_client *client, + int client_priority, + bool is_preempt_client) +{ + struct drm_i915_private *dev_priv = guc_to_i915(client->guc); + struct i915_gem_context *ctx_owner = is_preempt_client ? + dev_priv->preempt_context : dev_priv->kernel_context; + + if (client->owner != ctx_owner || + client->engines != INTEL_INFO(dev_priv)->ring_mask || + client->priority != client_priority || + client->doorbell_id == GUC_DOORBELL_INVALID) + return -EINVAL; + else + return 0; +} + +/* + * Check that guc_init_doorbell_hw is doing what it should. + * + * During GuC submission enable, we create GuC clients and their doorbells, + * but after resetting the microcontroller (resume & gpu reset), these + * GuC clients are still around, but the status of their doorbells may be + * incorrect. This is the reason behind validating that the doorbells status + * expected by the driver matches what the GuC/HW have. + */ +static int igt_guc_init_doorbell_hw(void *args) +{ + struct drm_i915_private *dev_priv = args; + struct intel_guc *guc; + DECLARE_BITMAP(db_bitmap_bk, GUC_NUM_DOORBELLS); + int i, err = 0; + + GEM_BUG_ON(!HAS_GUC(dev_priv)); + mutex_lock(&dev_priv->drm.struct_mutex); + + guc = &dev_priv->guc; + if (!guc) { + pr_err("No guc object!\n"); + err = -EINVAL; + goto unlock; + } + + err = check_all_doorbells(guc); + if (err) + goto unlock; + + /* + * Get rid of clients created during driver load because the test will + * recreate them. + */ + guc_clients_destroy(guc); + if (guc->execbuf_client || guc->preempt_client) { + pr_err("guc_clients_destroy lied!\n"); + err = -EINVAL; + goto unlock; + } + + err = guc_clients_create(guc); + if (err) { + pr_err("Failed to create clients\n"); + goto unlock; + } + + err = validate_client(guc->execbuf_client, + GUC_CLIENT_PRIORITY_KMD_NORMAL, false); + if (err) { + pr_err("execbug client validation failed\n"); + goto out; + } + + err = validate_client(guc->preempt_client, + GUC_CLIENT_PRIORITY_KMD_HIGH, true); + if (err) { + pr_err("preempt client validation failed\n"); + goto out; + } + + /* each client should have received a doorbell during alloc */ + if (!has_doorbell(guc->execbuf_client) || + !has_doorbell(guc->preempt_client)) { + pr_err("guc_clients_create didn't create doorbells\n"); + err = -EINVAL; + goto out; + } + + /* + * Basic test - an attempt to reallocate a valid doorbell to the + * client it is currently assigned should not cause a failure. + */ + err = guc_init_doorbell_hw(guc); + if (err) + goto out; + + /* + * Negative test - a client with no doorbell (invalid db id). + * Each client gets a doorbell when it is created, after destroying + * the doorbell, the db id is changed to GUC_DOORBELL_INVALID and the + * firmware will reject any attempt to allocate a doorbell with an + * invalid id (db has to be reserved before allocation). + */ + destroy_doorbell(guc->execbuf_client); + if (has_doorbell(guc->execbuf_client)) { + pr_err("destroy db did not work\n"); + err = -EINVAL; + goto out; + } + + err = guc_init_doorbell_hw(guc); + if (err != -EIO) { + pr_err("unexpected (err = %d)", err); + goto out; + } + + if (!available_dbs(guc, guc->execbuf_client->priority)) { + pr_err("doorbell not available when it should\n"); + err = -EIO; + goto out; + } + + /* clean after test */ + err = create_doorbell(guc->execbuf_client); + if (err) { + pr_err("recreate doorbell failed\n"); + goto out; + } + + /* + * Negative test - doorbell_bitmap out of sync, will trigger a few of + * WARN_ON(!doorbell_ok(guc, db_id)) but that's ok as long as the + * doorbells from our clients don't fail. + */ + bitmap_copy(db_bitmap_bk, guc->doorbell_bitmap, GUC_NUM_DOORBELLS); + for (i = 0; i < GUC_NUM_DOORBELLS; i++) + if (i % 2) + test_and_change_bit(i, guc->doorbell_bitmap); + + err = guc_init_doorbell_hw(guc); + if (err) { + pr_err("out of sync doorbell caused an error\n"); + goto out; + } + + /* restore 'correct' db bitmap */ + bitmap_copy(guc->doorbell_bitmap, db_bitmap_bk, GUC_NUM_DOORBELLS); + err = guc_init_doorbell_hw(guc); + if (err) { + pr_err("restored doorbell caused an error\n"); + goto out; + } + +out: + /* + * Leave clean state for other test, plus the driver always destroy the + * clients during unload. + */ + guc_clients_destroy(guc); + guc_clients_create(guc); +unlock: + mutex_unlock(&dev_priv->drm.struct_mutex); + return err; +} + +/* + * Create as many clients as number of doorbells. Note that there's already + * client(s)/doorbell(s) created during driver load, but this test creates + * its own and do not interact with the existing ones. + */ +static int igt_guc_doorbells(void *arg) +{ + struct drm_i915_private *dev_priv = arg; + struct intel_guc *guc; + int i, err = 0; + u16 db_id; + + GEM_BUG_ON(!HAS_GUC(dev_priv)); + mutex_lock(&dev_priv->drm.struct_mutex); + + guc = &dev_priv->guc; + if (!guc) { + pr_err("No guc object!\n"); + err = -EINVAL; + goto unlock; + } + + err = check_all_doorbells(guc); + if (err) + goto unlock; + + for (i = 0; i < ATTEMPTS; i++) { + clients[i] = guc_client_alloc(dev_priv, + INTEL_INFO(dev_priv)->ring_mask, + i % GUC_CLIENT_PRIORITY_NUM, + dev_priv->kernel_context); + + if (!clients[i]) { + pr_err("[%d] No guc client\n", i); + err = -EINVAL; + goto out; + } + + if (IS_ERR(clients[i])) { + if (PTR_ERR(clients[i]) != -ENOSPC) { + pr_err("[%d] unexpected error\n", i); + err = PTR_ERR(clients[i]); + goto out; + } + + if (available_dbs(guc, i % GUC_CLIENT_PRIORITY_NUM)) { + pr_err("[%d] non-db related alloc fail\n", i); + err = -EINVAL; + goto out; + } + + /* expected, ran out of dbs for this client type */ + continue; + } + + /* + * The check below is only valid because we keep a doorbell + * assigned during the whole life of the client. + */ + if (clients[i]->stage_id >= GUC_NUM_DOORBELLS) { + pr_err("[%d] more clients than doorbells (%d >= %d)\n", + i, clients[i]->stage_id, GUC_NUM_DOORBELLS); + err = -EINVAL; + goto out; + } + + err = validate_client(clients[i], + i % GUC_CLIENT_PRIORITY_NUM, false); + if (err) { + pr_err("[%d] client_alloc sanity check failed!\n", i); + err = -EINVAL; + goto out; + } + + db_id = clients[i]->doorbell_id; + + /* + * Client alloc gives us a doorbell, but we want to exercise + * this ourselves (this resembles guc_init_doorbell_hw) + */ + destroy_doorbell(clients[i]); + if (clients[i]->doorbell_id != GUC_DOORBELL_INVALID) { + pr_err("[%d] destroy db did not work!\n", i); + err = -EINVAL; + goto out; + } + + err = __reserve_doorbell(clients[i]); + if (err) { + pr_err("[%d] Failed to reserve a doorbell\n", i); + goto out; + } + + __update_doorbell_desc(clients[i], clients[i]->doorbell_id); + err = __create_doorbell(clients[i]); + if (err) { + pr_err("[%d] Failed to create a doorbell\n", i); + goto out; + } + + /* doorbell id shouldn't change, we are holding the mutex */ + if (db_id != clients[i]->doorbell_id) { + pr_err("[%d] doorbell id changed (%d != %d)\n", + i, db_id, clients[i]->doorbell_id); + err = -EINVAL; + goto out; + } + + err = check_all_doorbells(guc); + if (err) + goto out; + } + +out: + for (i = 0; i < ATTEMPTS; i++) + if (!IS_ERR_OR_NULL(clients[i])) + guc_client_free(clients[i]); +unlock: + mutex_unlock(&dev_priv->drm.struct_mutex); + return err; +} + +int intel_guc_live_selftest(struct drm_i915_private *dev_priv) +{ + static const struct i915_subtest tests[] = { + SUBTEST(igt_guc_init_doorbell_hw), + SUBTEST(igt_guc_doorbells), + }; + + if (!i915_modparams.enable_guc_submission) + return 0; + + return i915_subtests(tests, dev_priv); +} -- GitLab From 290b20a6d34acc188aec01669c210d731cc0ec15 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 21:56:55 +0000 Subject: [PATCH 0335/1471] drm/i915: Add might_sleep() check to wait_for() We should long past the time of trying to use wait_for() from inside atomic contexts, so add a might_sleep() check to prevent misuse. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171114215655.4849-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala Reviewed-by: Tvrtko Ursulin Reviewed-by: Maarten Lankhorst --- drivers/gpu/drm/i915/intel_drv.h | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index e9b66e0cb6477..645c5e8ee9f06 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -47,14 +47,11 @@ * contexts. Note that it's important that we check the condition again after * having timed out, since the timeout could be due to preemption or similar and * we've never had a chance to check the condition before the timeout. - * - * TODO: When modesetting has fully transitioned to atomic, the below - * drm_can_sleep() can be removed and in_atomic()/!in_atomic() asserts - * added. */ #define _wait_for(COND, US, W) ({ \ unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ int ret__; \ + might_sleep(); \ for (;;) { \ bool expired__ = time_after(jiffies, timeout__); \ if (COND) { \ @@ -65,11 +62,7 @@ ret__ = -ETIMEDOUT; \ break; \ } \ - if ((W) && drm_can_sleep()) { \ - usleep_range((W), (W)*2); \ - } else { \ - cpu_relax(); \ - } \ + usleep_range((W), (W) * 2); \ } \ ret__; \ }) -- GitLab From 41729bf2248bc8593e5103d43974079cc269524c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 17:35:20 +0000 Subject: [PATCH 0336/1471] drm/i915: Mark the userptr invalidate workqueue as WQ_MEM_RECLAIM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Commit 21cc6431e0c2 ("drm/i915: Mark the userptr invalidate workqueue as WQ_MEM_RECLAIM") tried to fixup the check_flush_dependency warning for hitting i915_gem_userptr_mn_invalidate_range_start from within the shrinker, but I failed to notice userptr has 2 similarly named workqueues. I marked up i915-userptr-acquire as WQ_MEM_RECLAIM whereas we only wait upon i915-userptr-release from inside the reclaim paths. [62530.869510] workqueue: PF_MEMALLOC task 7983(gem_shrink) is flushing !WQ_MEM_RECLAIM i915-userptr-release: (null) [62530.869515] ------------[ cut here ]------------ [62530.869519] WARNING: CPU: 1 PID: 7983 at kernel/workqueue.c:2434 check_flush_dependency+0x7f/0x110 [62530.869519] Modules linked in: pegasus mii ip6table_filter ip6_tables bnep iptable_filter snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic binfmt_misc nls_iso8859_1 intel_rapl x86_pkg_temp_thermal intel_powerclamp coretemp snd_hda_intel snd_hda_codec kvm_intel snd_hda_core snd_hwdep kvm snd_pcm irqbypass snd_seq_midi snd_seq_midi_event snd_rawmidi crct10dif_pclmul crc32_pclmul 8250_dw ghash_clmulni_intel snd_seq pcbc snd_seq_device snd_timer btusb aesni_intel btrtl btbcm aes_x86_64 iwlwifi btintel crypto_simd glue_helper cryptd bluetooth snd intel_cstate input_leds idma64 intel_rapl_perf ecdh_generic serio_raw soundcore cfg80211 wmi_bmof virt_dma intel_lpss_pci intel_lpss acpi_als kfifo_buf industrialio winbond_cir soc_button_array rc_core spidev tpm_crb intel_hid acpi_pad mac_hid sparse_keymap [62530.869546] parport_pc ppdev lp parport ip_tables x_tables autofs4 hid_generic usbhid i915 i2c_algo_bit prime_numbers drm_kms_helper syscopyarea e1000e sysfillrect sysimgblt fb_sys_fops ahci ptp pps_core libahci drm wmi video i2c_hid hid [62530.869557] CPU: 1 PID: 7983 Comm: gem_shrink Tainted: G U W L 4.14.0-rc8-drm-tip-ww45-commit-1342299+ #1 [62530.869558] Hardware name: Intel Corporation CoffeeLake Client Platform/CoffeeLake H DDR4 RVP, BIOS CNLSFWR1.R00.X098.A00.1707301945 07/30/2017 [62530.869559] task: ffffa1049dbeec80 task.stack: ffffae7d05c44000 [62530.869560] RIP: 0010:check_flush_dependency+0x7f/0x110 [62530.869561] RSP: 0018:ffffae7d05c473a0 EFLAGS: 00010286 [62530.869562] RAX: 000000000000006e RBX: ffffa1049540f400 RCX: ffffffffa3e55788 [62530.869562] RDX: 0000000000000000 RSI: 0000000000000092 RDI: 0000000000000202 [62530.869563] RBP: ffffae7d05c473c0 R08: 000000000000006e R09: 000000000038bb0e [62530.869563] R10: 0000000000000000 R11: 000000000000006e R12: ffffa1049dbeec80 [62530.869564] R13: 0000000000000000 R14: 0000000000000000 R15: ffffae7d05c473e0 [62530.869565] FS: 00007f621b129880(0000) GS:ffffa1050b240000(0000) knlGS:0000000000000000 [62530.869566] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [62530.869566] CR2: 00007f6214400000 CR3: 0000000353a17003 CR4: 00000000003606e0 [62530.869567] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [62530.869567] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [62530.869568] Call Trace: [62530.869570] flush_workqueue+0x115/0x3d0 [62530.869573] ? wake_up_process+0x15/0x20 [62530.869596] i915_gem_userptr_mn_invalidate_range_start+0x12f/0x160 [i915] [62530.869614] ? i915_gem_userptr_mn_invalidate_range_start+0x12f/0x160 [i915] [62530.869616] __mmu_notifier_invalidate_range_start+0x55/0x80 [62530.869618] try_to_unmap_one+0x791/0x8b0 [62530.869620] ? call_rwsem_down_read_failed+0x18/0x30 [62530.869622] rmap_walk_anon+0x10b/0x260 [62530.869624] rmap_walk+0x48/0x60 [62530.869625] try_to_unmap+0x93/0xf0 [62530.869626] ? page_remove_rmap+0x2a0/0x2a0 [62530.869627] ? page_not_mapped+0x20/0x20 [62530.869629] ? page_get_anon_vma+0x90/0x90 [62530.869630] ? invalid_mkclean_vma+0x20/0x20 [62530.869631] migrate_pages+0x946/0xaa0 [62530.869633] ? __ClearPageMovable+0x10/0x10 [62530.869635] ? isolate_freepages_block+0x3c0/0x3c0 [62530.869636] compact_zone+0x22f/0x970 [62530.869638] compact_zone_order+0xa3/0xd0 [62530.869640] try_to_compact_pages+0x1a5/0x2a0 [62530.869641] ? try_to_compact_pages+0x1a5/0x2a0 [62530.869643] __alloc_pages_direct_compact+0x50/0x110 [62530.869644] __alloc_pages_slowpath+0x4da/0xf30 [62530.869646] __alloc_pages_nodemask+0x262/0x280 [62530.869648] alloc_pages_vma+0x165/0x1e0 [62530.869649] shmem_alloc_hugepage+0xd0/0x130 [62530.869651] ? __radix_tree_insert+0x45/0x230 [62530.869652] ? __vm_enough_memory+0x29/0x130 [62530.869654] shmem_alloc_and_acct_page+0x10d/0x1e0 [62530.869655] shmem_getpage_gfp+0x426/0xc00 [62530.869657] shmem_fault+0xa0/0x1e0 [62530.869659] ? file_update_time+0x60/0x110 [62530.869660] __do_fault+0x1e/0xc0 [62530.869661] __handle_mm_fault+0xa35/0x1170 [62530.869662] handle_mm_fault+0xcc/0x1c0 [62530.869664] __do_page_fault+0x262/0x4f0 [62530.869666] do_page_fault+0x2e/0xe0 [62530.869667] page_fault+0x22/0x30 [62530.869668] RIP: 0033:0x404335 [62530.869669] RSP: 002b:00007fff7829e420 EFLAGS: 00010216 [62530.869670] RAX: 00007f6210400000 RBX: 0000000000000004 RCX: 0000000000b80000 [62530.869670] RDX: 0000000000002e01 RSI: 0000000000008000 RDI: 0000000000000004 [62530.869671] RBP: 0000000000000019 R08: 0000000000000002 R09: 0000000000000000 [62530.869671] R10: 0000000000000559 R11: 0000000000000246 R12: 0000000008000000 [62530.869672] R13: 00000000004042f0 R14: 0000000000000004 R15: 000000000000007e [62530.869673] Code: 00 8b b0 18 05 00 00 48 8d 8b b0 00 00 00 48 8d 90 c0 06 00 00 4d 89 f0 48 c7 c7 40 c0 c8 a3 c6 05 68 c5 e8 00 01 e8 c2 68 04 00 <0f> ff 4d 85 ed 74 18 49 8b 45 20 48 8b 70 08 8b 86 00 01 00 00 [62530.869691] ---[ end trace 01e01ad0ff5781f8 ]--- Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103739 Fixes: 21cc6431e0c2 ("drm/i915: Mark the userptr invalidate workqueue as WQ_MEM_RECLAIM") Signed-off-by: Chris Wilson Cc: Michał Winiarski Link: https://patchwork.freedesktop.org/patch/msgid/20171114173520.8829-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/i915_gem_userptr.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c index e26b23171b56f..ccde12d9e5f54 100644 --- a/drivers/gpu/drm/i915/i915_gem_userptr.c +++ b/drivers/gpu/drm/i915/i915_gem_userptr.c @@ -172,7 +172,9 @@ i915_mmu_notifier_create(struct mm_struct *mm) spin_lock_init(&mn->lock); mn->mn.ops = &i915_gem_userptr_notifier; mn->objects = RB_ROOT_CACHED; - mn->wq = alloc_workqueue("i915-userptr-release", WQ_UNBOUND, 0); + mn->wq = alloc_workqueue("i915-userptr-release", + WQ_UNBOUND | WQ_MEM_RECLAIM, + 0); if (mn->wq == NULL) { kfree(mn); return ERR_PTR(-ENOMEM); @@ -827,7 +829,7 @@ int i915_gem_init_userptr(struct drm_i915_private *dev_priv) dev_priv->mm.userptr_wq = alloc_workqueue("i915-userptr-acquire", - WQ_HIGHPRI | WQ_MEM_RECLAIM, + WQ_HIGHPRI | WQ_UNBOUND, 0); if (!dev_priv->mm.userptr_wq) return -ENOMEM; -- GitLab From 33a49868e5d8d98f01b41f902dd487530cd56b5e Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Mon, 13 Nov 2017 15:40:43 +0100 Subject: [PATCH 0337/1471] drm/i915: Enable FIFO underrun reporting after initial fastset, v4. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The firmware may have set up the pipe correctly, but the FIFO underrun and CRC interrupts are likely not enabled. This resulted in debugfs_test.read_all_entries failing on haswell, because of a timeout when reading the crc debugfs entry. Solve this by enabling FIFO underrun reporting after the initial fastset, which lets interrupts be generated as expected. Changes since v1: - Always enable CPU FIFO underrun reporting for >GEN2, and handle GEN2 correctly. Changes since v2: - Remove unneeded HAS_DDI, simplify GEN2 case. Changes since v3: - Use intel_crtc_pch_transcoder to determine pch transcoder for underruns. (Ville) - Remove crtc->config dereference in intel_crtc_pch_transcoder. (Ville) Testcase: debugfs_test.read_all_entries Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171113144043.58658-1-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index e6fcbc5abc757..5196a1683cd35 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1872,8 +1872,6 @@ enum pipe intel_crtc_pch_transcoder(struct intel_crtc *crtc) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - WARN_ON(!crtc->config->has_pch_encoder); - if (HAS_PCH_LPT(dev_priv)) return PIPE_A; else @@ -12940,6 +12938,7 @@ static void intel_begin_crtc_commit(struct drm_crtc *crtc, static void intel_finish_crtc_commit(struct drm_crtc *crtc, struct drm_crtc_state *old_crtc_state) { + struct drm_i915_private *dev_priv = to_i915(crtc->dev); struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct intel_atomic_state *old_intel_state = to_intel_atomic_state(old_crtc_state->state); @@ -12947,6 +12946,20 @@ static void intel_finish_crtc_commit(struct drm_crtc *crtc, intel_atomic_get_new_crtc_state(old_intel_state, intel_crtc); intel_pipe_update_end(new_crtc_state); + + if (new_crtc_state->update_pipe && + !needs_modeset(&new_crtc_state->base) && + old_crtc_state->mode.private_flags & I915_MODE_FLAG_INHERITED) { + if (!IS_GEN2(dev_priv)) + intel_set_cpu_fifo_underrun_reporting(dev_priv, intel_crtc->pipe, true); + + if (new_crtc_state->has_pch_encoder) { + enum pipe pch_transcoder = + intel_crtc_pch_transcoder(intel_crtc); + + intel_set_pch_fifo_underrun_reporting(dev_priv, pch_transcoder, true); + } + } } /** -- GitLab From 93313538c15341f44e8e879aa071663795f07bdf Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:34:59 +0100 Subject: [PATCH 0338/1471] drm/i915: Pass idle crtc_state to intel_dp_sink_crc MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit IPS can only be enabled if the primary plane is visible, so first make sure sw state matches hw state by waiting for hw_done. After this pass crtc_state to intel_dp_sink_crc() so that can be used, instead of using legacy pointers. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-7-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_debugfs.c | 17 +++++++++++++++-- drivers/gpu/drm/i915/intel_dp.c | 23 +++++++++++++---------- drivers/gpu/drm/i915/intel_drv.h | 3 ++- 3 files changed, 30 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index ff8f508a76614..df3852c02a353 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -2746,6 +2746,7 @@ static int i915_sink_crc(struct seq_file *m, void *data) for_each_intel_connector_iter(connector, &conn_iter) { struct drm_crtc *crtc; struct drm_connector_state *state; + struct intel_crtc_state *crtc_state; if (connector->base.connector_type != DRM_MODE_CONNECTOR_eDP) continue; @@ -2764,12 +2765,24 @@ static int i915_sink_crc(struct seq_file *m, void *data) if (ret) goto err; - if (!crtc->state->active) + crtc_state = to_intel_crtc_state(crtc->state); + if (!crtc_state->base.active) continue; + /* + * We need to wait for all crtc updates to complete, to make + * sure any pending modesets and plane updates are completed. + */ + if (crtc_state->base.commit) { + ret = wait_for_completion_interruptible(&crtc_state->base.commit->hw_done); + + if (ret) + goto err; + } + intel_dp = enc_to_intel_dp(state->best_encoder); - ret = intel_dp_sink_crc(intel_dp, crc); + ret = intel_dp_sink_crc(intel_dp, crtc_state, crc); if (ret) goto err; diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index c9c416389d0ef..74c7cd655db48 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3861,11 +3861,12 @@ intel_dp_configure_mst(struct intel_dp *intel_dp) intel_dp->is_mst); } -static int intel_dp_sink_crc_stop(struct intel_dp *intel_dp) +static int intel_dp_sink_crc_stop(struct intel_dp *intel_dp, + struct intel_crtc_state *crtc_state, bool disable_wa) { struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(dig_port->base.base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(crtc_state->base.crtc); u8 buf; int ret = 0; int count = 0; @@ -3901,15 +3902,17 @@ static int intel_dp_sink_crc_stop(struct intel_dp *intel_dp) } out: - hsw_enable_ips(intel_crtc); + if (disable_wa) + hsw_enable_ips(intel_crtc); return ret; } -static int intel_dp_sink_crc_start(struct intel_dp *intel_dp) +static int intel_dp_sink_crc_start(struct intel_dp *intel_dp, + struct intel_crtc_state *crtc_state) { struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(dig_port->base.base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(crtc_state->base.crtc); u8 buf; int ret; @@ -3923,7 +3926,7 @@ static int intel_dp_sink_crc_start(struct intel_dp *intel_dp) return -EIO; if (buf & DP_TEST_SINK_START) { - ret = intel_dp_sink_crc_stop(intel_dp); + ret = intel_dp_sink_crc_stop(intel_dp, crtc_state, false); if (ret) return ret; } @@ -3940,16 +3943,16 @@ static int intel_dp_sink_crc_start(struct intel_dp *intel_dp) return 0; } -int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) +int intel_dp_sink_crc(struct intel_dp *intel_dp, struct intel_crtc_state *crtc_state, u8 *crc) { struct intel_digital_port *dig_port = dp_to_dig_port(intel_dp); struct drm_i915_private *dev_priv = to_i915(dig_port->base.base.dev); - struct intel_crtc *intel_crtc = to_intel_crtc(dig_port->base.base.crtc); + struct intel_crtc *intel_crtc = to_intel_crtc(crtc_state->base.crtc); u8 buf; int count, ret; int attempts = 6; - ret = intel_dp_sink_crc_start(intel_dp); + ret = intel_dp_sink_crc_start(intel_dp, crtc_state); if (ret) return ret; @@ -3977,7 +3980,7 @@ int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc) } stop: - intel_dp_sink_crc_stop(intel_dp); + intel_dp_sink_crc_stop(intel_dp, crtc_state, true); return ret; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 645c5e8ee9f06..225392fe67625 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1531,7 +1531,8 @@ void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode); void intel_dp_encoder_reset(struct drm_encoder *encoder); void intel_dp_encoder_suspend(struct intel_encoder *intel_encoder); void intel_dp_encoder_destroy(struct drm_encoder *encoder); -int intel_dp_sink_crc(struct intel_dp *intel_dp, u8 *crc); +int intel_dp_sink_crc(struct intel_dp *intel_dp, + struct intel_crtc_state *crtc_state, u8 *crc); bool intel_dp_compute_config(struct intel_encoder *encoder, struct intel_crtc_state *pipe_config, struct drm_connector_state *conn_state); -- GitLab From 199ea381d9887bc8b1720176c443e203d9664cd3 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Fri, 10 Nov 2017 12:35:00 +0100 Subject: [PATCH 0339/1471] drm/i915: Pass crtc_state to ips toggle functions, v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Changes since v1: - Only pass crtc_state, not crtc. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171110113503.16253-8-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_color.c | 4 ++-- drivers/gpu/drm/i915/intel_display.c | 24 ++++++++++++++---------- drivers/gpu/drm/i915/intel_dp.c | 6 +++--- drivers/gpu/drm/i915/intel_drv.h | 4 ++-- 4 files changed, 21 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_color.c b/drivers/gpu/drm/i915/intel_color.c index b8315bca852b5..aa66e952a95d3 100644 --- a/drivers/gpu/drm/i915/intel_color.c +++ b/drivers/gpu/drm/i915/intel_color.c @@ -370,7 +370,7 @@ static void haswell_load_luts(struct drm_crtc_state *crtc_state) */ if (IS_HASWELL(dev_priv) && intel_crtc_state->ips_enabled && (intel_crtc_state->gamma_mode == GAMMA_MODE_MODE_SPLIT)) { - hsw_disable_ips(intel_crtc); + hsw_disable_ips(intel_crtc_state); reenable_ips = true; } @@ -380,7 +380,7 @@ static void haswell_load_luts(struct drm_crtc_state *crtc_state) i9xx_load_luts(crtc_state); if (reenable_ips) - hsw_enable_ips(intel_crtc); + hsw_enable_ips(intel_crtc_state); } static void bdw_load_degamma_lut(struct drm_crtc_state *state) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5196a1683cd35..3b3dec1e6640d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4864,8 +4864,9 @@ static void ironlake_pfit_enable(struct intel_crtc *crtc) } } -void hsw_enable_ips(struct intel_crtc *crtc) +void hsw_enable_ips(const struct intel_crtc_state *crtc_state) { + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); @@ -4903,12 +4904,13 @@ void hsw_enable_ips(struct intel_crtc *crtc) } } -void hsw_disable_ips(struct intel_crtc *crtc) +void hsw_disable_ips(const struct intel_crtc_state *crtc_state) { + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - if (!crtc->config->ips_enabled) + if (!crtc_state->ips_enabled) return; assert_plane_enabled(dev_priv, crtc->plane); @@ -4956,7 +4958,8 @@ static void intel_crtc_dpms_overlay_disable(struct intel_crtc *intel_crtc) * completely hide the primary plane. */ static void -intel_post_enable_primary(struct drm_crtc *crtc) +intel_post_enable_primary(struct drm_crtc *crtc, + const struct intel_crtc_state *new_crtc_state) { struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); @@ -4969,7 +4972,7 @@ intel_post_enable_primary(struct drm_crtc *crtc) * when going from primary only to sprite only and vice * versa. */ - hsw_enable_ips(intel_crtc); + hsw_enable_ips(new_crtc_state); /* * Gen2 reports pipe underruns whenever all planes are disabled. @@ -4988,7 +4991,8 @@ intel_post_enable_primary(struct drm_crtc *crtc) /* FIXME move all this to pre_plane_update() with proper state tracking */ static void -intel_pre_disable_primary(struct drm_crtc *crtc) +intel_pre_disable_primary(struct drm_crtc *crtc, + const struct intel_crtc_state *old_crtc_state) { struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); @@ -5010,7 +5014,7 @@ intel_pre_disable_primary(struct drm_crtc *crtc) * when going from primary only to sprite only and vice * versa. */ - hsw_disable_ips(intel_crtc); + hsw_disable_ips(old_crtc_state); } /* FIXME get rid of this and use pre_plane_update */ @@ -5022,7 +5026,7 @@ intel_pre_disable_primary_noatomic(struct drm_crtc *crtc) struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int pipe = intel_crtc->pipe; - intel_pre_disable_primary(crtc); + intel_pre_disable_primary(crtc, to_intel_crtc_state(crtc->state)); /* * Vblank time updates from the shadow to live plane control register @@ -5066,7 +5070,7 @@ static void intel_post_plane_update(struct intel_crtc_state *old_crtc_state) if (primary_state->base.visible && (needs_modeset(&pipe_config->base) || !old_primary_state->base.visible)) - intel_post_enable_primary(&crtc->base); + intel_post_enable_primary(&crtc->base, pipe_config); } } @@ -5095,7 +5099,7 @@ static void intel_pre_plane_update(struct intel_crtc_state *old_crtc_state, if (old_primary_state->base.visible && (modeset || !primary_state->base.visible)) - intel_pre_disable_primary(&crtc->base); + intel_pre_disable_primary(&crtc->base, old_crtc_state); } /* diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index 74c7cd655db48..bc61f38b131df 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -3903,7 +3903,7 @@ static int intel_dp_sink_crc_stop(struct intel_dp *intel_dp, out: if (disable_wa) - hsw_enable_ips(intel_crtc); + hsw_enable_ips(crtc_state); return ret; } @@ -3931,11 +3931,11 @@ static int intel_dp_sink_crc_start(struct intel_dp *intel_dp, return ret; } - hsw_disable_ips(intel_crtc); + hsw_disable_ips(crtc_state); if (drm_dp_dpcd_writeb(&intel_dp->aux, DP_TEST_SINK, buf | DP_TEST_SINK_START) < 0) { - hsw_enable_ips(intel_crtc); + hsw_enable_ips(crtc_state); return -EIO; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 225392fe67625..69aab324aaa1a 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1485,8 +1485,8 @@ bool bxt_find_best_dpll(struct intel_crtc_state *crtc_state, int target_clock, int chv_calc_dpll_params(int refclk, struct dpll *pll_clock); bool intel_crtc_active(struct intel_crtc *crtc); -void hsw_enable_ips(struct intel_crtc *crtc); -void hsw_disable_ips(struct intel_crtc *crtc); +void hsw_enable_ips(const struct intel_crtc_state *crtc_state); +void hsw_disable_ips(const struct intel_crtc_state *crtc_state); enum intel_display_power_domain intel_port_to_power_domain(enum port port); void intel_mode_from_pipe_config(struct drm_display_mode *mode, struct intel_crtc_state *pipe_config); -- GitLab From 870a0b12d0918c8e08d3895e4754a38da2a1850c Mon Sep 17 00:00:00 2001 From: Lukasz Majewski Date: Tue, 7 Nov 2017 16:30:58 +0100 Subject: [PATCH 0340/1471] drm/panel: simple: Add Tianma TM070RVHG71 panel support The Tianma TM070RVHG71 is a 7" WVGA (800x480) TFT LCD panel connected via LVDS and can be supported by the simple-panel binding. Signed-off-by: Lukasz Majewski Acked-by: Rob Herring Signed-off-by: Thierry Reding Link: https://patchwork.freedesktop.org/patch/msgid/20171107153058.5554-1-lukma@denx.de --- .../display/panel/tianma,tm070rvhg71.txt | 29 +++++++++++++++++++ drivers/gpu/drm/panel/panel-simple.c | 27 +++++++++++++++++ 2 files changed, 56 insertions(+) create mode 100644 Documentation/devicetree/bindings/display/panel/tianma,tm070rvhg71.txt diff --git a/Documentation/devicetree/bindings/display/panel/tianma,tm070rvhg71.txt b/Documentation/devicetree/bindings/display/panel/tianma,tm070rvhg71.txt new file mode 100644 index 0000000000000..b25261e63a6d1 --- /dev/null +++ b/Documentation/devicetree/bindings/display/panel/tianma,tm070rvhg71.txt @@ -0,0 +1,29 @@ +Tianma Micro-electronics TM070RVHG71 7.0" WXGA TFT LCD panel + +Required properties: +- compatible: should be "tianma,tm070rvhg71" +- power-supply: single regulator to provide the supply voltage +- backlight: phandle of the backlight device attached to the panel + +Required nodes: +- port: LVDS port mapping to connect this display + +This panel needs single power supply voltage. Its backlight is conntrolled +via PWM signal. + +Example: +-------- + +Example device-tree definition when connected to iMX6Q based board + + panel: panel-lvds0 { + compatible = "tianma,tm070rvhg71"; + backlight = <&backlight_lvds>; + power-supply = <®_lvds>; + + port { + panel_in_lvds0: endpoint { + remote-endpoint = <&lvds0_out>; + }; + }; + }; diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index b7c4709f7b34c..71d8f0e5e664f 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -1837,6 +1837,30 @@ static const struct panel_desc tianma_tm070jdhg30 = { .bus_format = MEDIA_BUS_FMT_RGB888_1X7X4_SPWG, }; +static const struct display_timing tianma_tm070rvhg71_timing = { + .pixelclock = { 27700000, 29200000, 39600000 }, + .hactive = { 800, 800, 800 }, + .hfront_porch = { 12, 40, 212 }, + .hback_porch = { 88, 88, 88 }, + .hsync_len = { 1, 1, 40 }, + .vactive = { 480, 480, 480 }, + .vfront_porch = { 1, 13, 88 }, + .vback_porch = { 32, 32, 32 }, + .vsync_len = { 1, 1, 3 }, + .flags = DISPLAY_FLAGS_DE_HIGH, +}; + +static const struct panel_desc tianma_tm070rvhg71 = { + .timings = &tianma_tm070rvhg71_timing, + .num_timings = 1, + .bpc = 8, + .size = { + .width = 154, + .height = 86, + }, + .bus_format = MEDIA_BUS_FMT_RGB888_1X7X4_SPWG, +}; + static const struct drm_display_mode toshiba_lt089ac29000_mode = { .clock = 79500, .hdisplay = 1280, @@ -2142,6 +2166,9 @@ static const struct of_device_id platform_of_match[] = { }, { .compatible = "tianma,tm070jdhg30", .data = &tianma_tm070jdhg30, + }, { + .compatible = "tianma,tm070rvhg71", + .data = &tianma_tm070rvhg71, }, { .compatible = "toshiba,lt089ac29000", .data = &toshiba_lt089ac29000, -- GitLab From 65c766cad8f81a65790580576a50abd21e3dea44 Mon Sep 17 00:00:00 2001 From: Lukasz Majewski Date: Sat, 21 Oct 2017 00:18:37 +0200 Subject: [PATCH 0341/1471] drm/panel: simple: Add Mitsubishi AA070MC01 panel support The Mitsubishi AA070MC01 is a 7.0" WVGA (800x480) TFT panel working in 8 bit ISP mode (pin 19 "mode" HIGH for 20 pin TFT connector). Signed-off-by: Lukasz Majewski Acked-by: Rob Herring Signed-off-by: Thierry Reding Link: https://patchwork.freedesktop.org/patch/msgid/1508537917-1723-1-git-send-email-lukma@denx.de --- .../display/panel/mitsubishi,aa070mc01.txt | 7 ++++ drivers/gpu/drm/panel/panel-simple.c | 35 +++++++++++++++++++ 2 files changed, 42 insertions(+) create mode 100644 Documentation/devicetree/bindings/display/panel/mitsubishi,aa070mc01.txt diff --git a/Documentation/devicetree/bindings/display/panel/mitsubishi,aa070mc01.txt b/Documentation/devicetree/bindings/display/panel/mitsubishi,aa070mc01.txt new file mode 100644 index 0000000000000..7d8f6eeef6d97 --- /dev/null +++ b/Documentation/devicetree/bindings/display/panel/mitsubishi,aa070mc01.txt @@ -0,0 +1,7 @@ +Mitsubishi "AA070MC01 7.0" WVGA TFT LCD panel + +Required properties: +- compatible: should be "mitsubishi,aa070mc01-ca1" + +This binding is compatible with the simple-panel binding, which is specified +in simple-panel.txt in this directory. diff --git a/drivers/gpu/drm/panel/panel-simple.c b/drivers/gpu/drm/panel/panel-simple.c index 71d8f0e5e664f..5591984a392b7 100644 --- a/drivers/gpu/drm/panel/panel-simple.c +++ b/drivers/gpu/drm/panel/panel-simple.c @@ -1356,6 +1356,38 @@ static const struct panel_desc lg_lp129qe = { }, }; +static const struct drm_display_mode mitsubishi_aa070mc01_mode = { + .clock = 30400, + .hdisplay = 800, + .hsync_start = 800 + 0, + .hsync_end = 800 + 1, + .htotal = 800 + 0 + 1 + 160, + .vdisplay = 480, + .vsync_start = 480 + 0, + .vsync_end = 480 + 48 + 1, + .vtotal = 480 + 48 + 1 + 0, + .vrefresh = 60, + .flags = DRM_MODE_FLAG_NHSYNC | DRM_MODE_FLAG_NVSYNC, +}; + +static const struct panel_desc mitsubishi_aa070mc01 = { + .modes = &mitsubishi_aa070mc01_mode, + .num_modes = 1, + .bpc = 8, + .size = { + .width = 152, + .height = 91, + }, + + .delay = { + .enable = 200, + .unprepare = 200, + .disable = 400, + }, + .bus_format = MEDIA_BUS_FMT_RGB888_1X7X4_SPWG, + .bus_flags = DRM_BUS_FLAG_DE_HIGH, +}; + static const struct display_timing nec_nl12880bc20_05_timing = { .pixelclock = { 67000000, 71000000, 75000000 }, .hactive = { 1280, 1280, 1280 }, @@ -2109,6 +2141,9 @@ static const struct of_device_id platform_of_match[] = { }, { .compatible = "lg,lp129qe", .data = &lg_lp129qe, + }, { + .compatible = "mitsubishi,aa070mc01-ca1", + .data = &mitsubishi_aa070mc01, }, { .compatible = "nec,nl12880bc20-05", .data = &nec_nl12880bc20_05, -- GitLab From 5b9489cb8ee23b41ab5a4cca6652511fa252bad8 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 15 Nov 2017 17:31:56 +0100 Subject: [PATCH 0342/1471] drm/i915: Calculate vlv/chv intermediate watermarks correctly, v3. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The watermarks it should calculate against are the old optimal watermarks. The currently active crtc watermarks are pure fiction, and are invalid in case of a nonblocking modeset, page flip enabling/disabling planes or any other reason. When the crtc is disabled or during a modeset the intermediate watermarks don't need to be programmed separately, and could be directly assigned to the optimal watermarks. CXSR must always be disabled in the intermediate case for modesets, else we get a WARN for vblank wait timeout. Also rename crtc_state to new_crtc_state, to distinguish it from the old state. Changes since v1: - Use intel_atomic_get_old_crtc_state. (ville) Changes since v2: - Always unset cxsr during modeset. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171115163157.14372-1-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_pm.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 8c69ec9eb6eec..f904bf73dbd69 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -2026,16 +2026,27 @@ static void vlv_atomic_update_fifo(struct intel_atomic_state *state, static int vlv_compute_intermediate_wm(struct drm_device *dev, struct intel_crtc *crtc, - struct intel_crtc_state *crtc_state) + struct intel_crtc_state *new_crtc_state) { - struct vlv_wm_state *intermediate = &crtc_state->wm.vlv.intermediate; - const struct vlv_wm_state *optimal = &crtc_state->wm.vlv.optimal; - const struct vlv_wm_state *active = &crtc->wm.active.vlv; + struct vlv_wm_state *intermediate = &new_crtc_state->wm.vlv.intermediate; + const struct vlv_wm_state *optimal = &new_crtc_state->wm.vlv.optimal; + struct intel_atomic_state *intel_state = + to_intel_atomic_state(new_crtc_state->base.state); + const struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(intel_state, crtc); + const struct vlv_wm_state *active = &old_crtc_state->wm.vlv.optimal; int level; + if (!new_crtc_state->base.active || drm_atomic_crtc_needs_modeset(&new_crtc_state->base)) { + *intermediate = *optimal; + + intermediate->cxsr = false; + goto out; + } + intermediate->num_levels = min(optimal->num_levels, active->num_levels); intermediate->cxsr = optimal->cxsr && active->cxsr && - !crtc_state->disable_cxsr; + !new_crtc_state->disable_cxsr; for (level = 0; level < intermediate->num_levels; level++) { enum plane_id plane_id; @@ -2054,12 +2065,13 @@ static int vlv_compute_intermediate_wm(struct drm_device *dev, vlv_invalidate_wms(crtc, intermediate, level); +out: /* * If our intermediate WM are identical to the final WM, then we can * omit the post-vblank programming; only update if it's different. */ if (memcmp(intermediate, optimal, sizeof(*intermediate)) != 0) - crtc_state->wm.need_postvbl_update = true; + new_crtc_state->wm.need_postvbl_update = true; return 0; } -- GitLab From 248c2435cbd75ad098130dda819b2973591d6f6d Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 15 Nov 2017 17:31:57 +0100 Subject: [PATCH 0343/1471] drm/i915: Calculate g4x intermediate watermarks correctly MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The watermarks it should calculate against are the old optimal watermarks. The currently active crtc watermarks are pure fiction, and are invalid in case of a nonblocking modeset, page flip enabling/disabling planes or any other reason. When the crtc is disabled or during a modeset the intermediate watermarks don't need to be programmed separately, and could be directly assigned to the optimal watermarks. CXSR must always be disabled in the intermediate case for modesets, else we get a WARN for vblank wait timeout. Also rename crtc_state to new_crtc_state, to distinguish it from the old state. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171115163157.14372-2-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_pm.c | 27 ++++++++++++++++++++------- 1 file changed, 20 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index f904bf73dbd69..3c55e40263310 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -1406,17 +1406,29 @@ static int g4x_compute_pipe_wm(struct intel_crtc_state *crtc_state) static int g4x_compute_intermediate_wm(struct drm_device *dev, struct intel_crtc *crtc, - struct intel_crtc_state *crtc_state) + struct intel_crtc_state *new_crtc_state) { - struct g4x_wm_state *intermediate = &crtc_state->wm.g4x.intermediate; - const struct g4x_wm_state *optimal = &crtc_state->wm.g4x.optimal; - const struct g4x_wm_state *active = &crtc->wm.active.g4x; + struct g4x_wm_state *intermediate = &new_crtc_state->wm.g4x.intermediate; + const struct g4x_wm_state *optimal = &new_crtc_state->wm.g4x.optimal; + struct intel_atomic_state *intel_state = + to_intel_atomic_state(new_crtc_state->base.state); + const struct intel_crtc_state *old_crtc_state = + intel_atomic_get_old_crtc_state(intel_state, crtc); + const struct g4x_wm_state *active = &old_crtc_state->wm.g4x.optimal; enum plane_id plane_id; + if (!new_crtc_state->base.active || drm_atomic_crtc_needs_modeset(&new_crtc_state->base)) { + *intermediate = *optimal; + + intermediate->cxsr = false; + intermediate->hpll_en = false; + goto out; + } + intermediate->cxsr = optimal->cxsr && active->cxsr && - !crtc_state->disable_cxsr; + !new_crtc_state->disable_cxsr; intermediate->hpll_en = optimal->hpll_en && active->hpll_en && - !crtc_state->disable_cxsr; + !new_crtc_state->disable_cxsr; intermediate->fbc_en = optimal->fbc_en && active->fbc_en; for_each_plane_id_on_crtc(crtc, plane_id) { @@ -1458,12 +1470,13 @@ static int g4x_compute_intermediate_wm(struct drm_device *dev, WARN_ON(intermediate->hpll.fbc > g4x_fbc_fifo_size(2) && intermediate->fbc_en && intermediate->hpll_en); +out: /* * If our intermediate WM are identical to the final WM, then we can * omit the post-vblank programming; only update if it's different. */ if (memcmp(intermediate, optimal, sizeof(*intermediate)) != 0) - crtc_state->wm.need_postvbl_update = true; + new_crtc_state->wm.need_postvbl_update = true; return 0; } -- GitLab From 0cfecb7c4b9b45ed1776162e132b43f92564f3f4 Mon Sep 17 00:00:00 2001 From: Radhakrishna Sripada Date: Thu, 16 Nov 2017 17:08:25 -0800 Subject: [PATCH 0344/1471] Revert "drm/i915: Display WA #1133 WaFbcSkipSegments:cnl, glk" This reverts commit 8f067837c4b713ce2e69be95af7b2a5eb3bd7de8. HSD says "WA withdrawn. It was causing corruption with some images. WA is not strictly necessary since this bug just causes loss of FBC compression with some sizes and images, but doesn't break anything." Fixes: 8f067837c4b7 ("drm/i915: Display WA #1133 WaFbcSkipSegments:cnl, glk") Cc: Rodrigo Vivi Signed-off-by: Radhakrishna Sripada Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171117010825.23118-1-radhakrishna.sripada@intel.com --- drivers/gpu/drm/i915/i915_reg.h | 3 --- drivers/gpu/drm/i915/intel_pm.c | 12 ------------ 2 files changed, 15 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 107e2d7c9fba7..96c80fa0fcacf 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -2985,9 +2985,6 @@ enum i915_power_well_id { #define ILK_DPFC_CHICKEN _MMIO(0x43224) #define ILK_DPFC_DISABLE_DUMMY0 (1<<8) #define ILK_DPFC_NUKE_ON_ANY_MODIFICATION (1<<23) -#define GLK_SKIP_SEG_EN (1<<12) -#define GLK_SKIP_SEG_COUNT_MASK (3<<10) -#define GLK_SKIP_SEG_COUNT(x) ((x)<<10) #define ILK_FBC_RT_BASE _MMIO(0x2128) #define ILK_FBC_RT_VALID (1<<0) #define SNB_FBC_FRONT_BUFFER (1<<1) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 3c55e40263310..4d2cd432f7393 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -121,7 +121,6 @@ static void bxt_init_clock_gating(struct drm_i915_private *dev_priv) static void glk_init_clock_gating(struct drm_i915_private *dev_priv) { - u32 val; gen9_init_clock_gating(dev_priv); /* @@ -141,11 +140,6 @@ static void glk_init_clock_gating(struct drm_i915_private *dev_priv) I915_WRITE(CHICKEN_MISC_2, val); } - /* Display WA #1133: WaFbcSkipSegments:glk */ - val = I915_READ(ILK_DPFC_CHICKEN); - val &= ~GLK_SKIP_SEG_COUNT_MASK; - val |= GLK_SKIP_SEG_EN | GLK_SKIP_SEG_COUNT(1); - I915_WRITE(ILK_DPFC_CHICKEN, val); } static void i915_pineview_get_mem_freq(struct drm_i915_private *dev_priv) @@ -8503,12 +8497,6 @@ static void cnl_init_clock_gating(struct drm_i915_private *dev_priv) if (IS_CNL_REVID(dev_priv, CNL_REVID_A0, CNL_REVID_B0)) val |= SARBUNIT_CLKGATE_DIS; I915_WRITE(SLICE_UNIT_LEVEL_CLKGATE, val); - - /* Display WA #1133: WaFbcSkipSegments:cnl */ - val = I915_READ(ILK_DPFC_CHICKEN); - val &= ~GLK_SKIP_SEG_COUNT_MASK; - val |= GLK_SKIP_SEG_EN | GLK_SKIP_SEG_COUNT(1); - I915_WRITE(ILK_DPFC_CHICKEN, val); } static void cfl_init_clock_gating(struct drm_i915_private *dev_priv) -- GitLab From 223c73a3667a5afe33245d423f867f4c4b1b39ea Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 17 Nov 2017 16:29:45 +0000 Subject: [PATCH 0345/1471] drm/i915/selftests: Report ENOMEM clearly for an allocation failure If we can not run the drunk_hole test because we couldn't allocate the memory for the permutation array (even after we tried trimming the size), report a clear ENOMEM. Similary, if we are asked to operate on a hole too small for ourselves, make it skip quietly. v2: Avoid malloc(0) since that returns ZERO_SIZE_PTR not NULL. v3: Fixup similar construction for lowlevel_hole v4: Use u64 >> 1 to avoid 64b div. Signed-off-by: Chris Wilson Cc: Matthew Auld Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171117101732.4335-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171117162945.16390-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 36 +++++++++++++------ 1 file changed, 26 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c index 3dcf886a2802a..6491cf0a4f464 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c @@ -216,13 +216,21 @@ static int lowlevel_hole(struct drm_i915_private *i915, hole_size = (hole_end - hole_start) >> size; if (hole_size > KMALLOC_MAX_SIZE / sizeof(u32)) hole_size = KMALLOC_MAX_SIZE / sizeof(u32); - count = hole_size; + count = hole_size >> 1; + if (!count) { + pr_debug("%s: hole is too small [%llx - %llx] >> %d: %lld\n", + __func__, hole_start, hole_end, size, hole_size); + break; + } + do { - count >>= 1; order = i915_random_order(count, &prng); - } while (!order && count); - if (!order) - break; + if (order) + break; + } while (count >>= 1); + if (!count) + return -ENOMEM; + GEM_BUG_ON(!order); GEM_BUG_ON(count * BIT_ULL(size) > vm->total); GEM_BUG_ON(hole_start + count * BIT_ULL(size) > hole_end); @@ -704,13 +712,21 @@ static int drunk_hole(struct drm_i915_private *i915, hole_size = (hole_end - hole_start) >> size; if (hole_size > KMALLOC_MAX_SIZE / sizeof(u32)) hole_size = KMALLOC_MAX_SIZE / sizeof(u32); - count = hole_size; + count = hole_size >> 1; + if (!count) { + pr_debug("%s: hole is too small [%llx - %llx] >> %d: %lld\n", + __func__, hole_start, hole_end, size, hole_size); + break; + } + do { - count >>= 1; order = i915_random_order(count, &prng); - } while (!order && count); - if (!order) - break; + if (order) + break; + } while (count >>= 1); + if (!count) + return -ENOMEM; + GEM_BUG_ON(!order); /* Ignore allocation failures (i.e. don't report them as * a test failure) as we are purposefully allocating very -- GitLab From 6a7a6a982a22091b9d67cb3ff33d7782494aa2f8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 17 Nov 2017 10:26:35 +0000 Subject: [PATCH 0346/1471] drm/i915: Add a policy note for removing workarounds Rodrigo gave a persuasive argument for keeping workarounds: that they serve as a good guide for the bring up of the next generation. Not only do workarounds persist into the early revisions, they show where the workarounds were previously added to the code flow and sometimes the old workarounds have an explanation that give insight into their wider implications. Based on his suggestion, document the policy that we want to keep the workarounds from the current generation to guide the next. Older preproduction workarounds we still want to remove to keep the code clean. Signed-off-by: Chris Wilson Cc: Jani Nikula Cc: Rodrigo Vivi Cc: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117102635.8689-1-chris@chris-wilson.co.uk Acked-by: Jani Nikula Reviewed-by: Rodrigo Vivi --- drivers/gpu/drm/i915/i915_drv.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 3423d873123a7..8ea6ce7027d48 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -843,6 +843,11 @@ static void i915_workqueues_cleanup(struct drm_i915_private *dev_priv) * We don't keep the workarounds for pre-production hardware, so we expect our * driver to fail on these machines in one way or another. A little warning on * dmesg may help both the user and the bug triagers. + * + * Our policy for removing pre-production workarounds is to keep the + * current gen workarounds as a guide to the bring-up of the next gen + * (workarounds have a habit of persisting!). Anything older than that + * should be removed along with the complications they introduce. */ static void intel_detect_preproduction_hw(struct drm_i915_private *dev_priv) { -- GitLab From 010d118c20617021025a930bc8e90f371ab99da5 Mon Sep 17 00:00:00 2001 From: Rodrigo Vivi Date: Fri, 17 Nov 2017 14:47:02 -0800 Subject: [PATCH 0347/1471] drm/i915: Update DRIVER_DATE to 20171117 Signed-off-by: Rodrigo Vivi --- drivers/gpu/drm/i915/i915_drv.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2158a758a17d7..36bb4927484a0 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -79,8 +79,8 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20171109" -#define DRIVER_TIMESTAMP 1510270206 +#define DRIVER_DATE "20171117" +#define DRIVER_TIMESTAMP 1510958822 /* Use I915_STATE_WARN(x) and I915_STATE_WARN_ON() (rather than WARN() and * WARN_ON()) for hw state sanity checks to check for unexpected conditions -- GitLab From 17ab7806de0c10d27cdbda8ef57ca680bdd24315 Mon Sep 17 00:00:00 2001 From: Lucas Stach Date: Fri, 13 Jan 2017 18:36:30 +0100 Subject: [PATCH 0348/1471] drm: don't link DP aux i2c adapter to the hardware device node The i2c adapter on DP AUX is purely a software construct. Linking it to the device node of the parent device is wrong, as it leads to 2 devices sharing the same device node, which is bad practice, as well as the i2c trying to populate children of the i2c adapter by looking at the child device nodes of the parent device. Signed-off-by: Lucas Stach Acked-by: Thierry Reding Signed-off-by: Andrzej Hajda Link: https://patchwork.freedesktop.org/patch/msgid/20170113173630.22138-1-l.stach@pengutronix.de --- drivers/gpu/drm/drm_dp_helper.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/drm_dp_helper.c b/drivers/gpu/drm/drm_dp_helper.c index b3d68964b4076..adf79be42c1e2 100644 --- a/drivers/gpu/drm/drm_dp_helper.c +++ b/drivers/gpu/drm/drm_dp_helper.c @@ -1097,7 +1097,6 @@ int drm_dp_aux_register(struct drm_dp_aux *aux) aux->ddc.class = I2C_CLASS_DDC; aux->ddc.owner = THIS_MODULE; aux->ddc.dev.parent = aux->dev; - aux->ddc.dev.of_node = aux->dev->of_node; strlcpy(aux->ddc.name, aux->name ? aux->name : dev_name(aux->dev), sizeof(aux->ddc.name)); -- GitLab From 0df12b3f0297b270960a4ce735607da518feafcb Mon Sep 17 00:00:00 2001 From: Greg Kroah-Hartman Date: Fri, 17 Nov 2017 15:16:32 +0100 Subject: [PATCH 0349/1471] gpu: gma500: remove unneeded DRIVER_LICENSE #define There is no need to #define the license of the driver, just put it in the MODULE_LICENSE() line directly as a text string. This allows tools that check that the module license matches the source code license to work properly, as there is no need to unwind the unneeded dereference, especially when the string is defined in a .h file far away from the .c file it is used in. Cc: Patrik Jakobsson Cc: David Airlie Reported-by: Philippe Ombredanne Signed-off-by: Greg Kroah-Hartman Reviewed-by: Philippe Ombredanne Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117141632.GA17880@kroah.com --- drivers/gpu/drm/gma500/psb_drv.c | 2 +- drivers/gpu/drm/gma500/psb_drv.h | 1 - 2 files changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/gpu/drm/gma500/psb_drv.c b/drivers/gpu/drm/gma500/psb_drv.c index 37a3be71acd90..8f5cc1f471cd2 100644 --- a/drivers/gpu/drm/gma500/psb_drv.c +++ b/drivers/gpu/drm/gma500/psb_drv.c @@ -527,4 +527,4 @@ module_exit(psb_exit); MODULE_AUTHOR(DRIVER_AUTHOR); MODULE_DESCRIPTION(DRIVER_DESC); -MODULE_LICENSE(DRIVER_LICENSE); +MODULE_LICENSE("GPL"); diff --git a/drivers/gpu/drm/gma500/psb_drv.h b/drivers/gpu/drm/gma500/psb_drv.h index 821497dbd3fcb..4918efc57b7ab 100644 --- a/drivers/gpu/drm/gma500/psb_drv.h +++ b/drivers/gpu/drm/gma500/psb_drv.h @@ -36,7 +36,6 @@ #include "mmu.h" #define DRIVER_AUTHOR "Alan Cox and others" -#define DRIVER_LICENSE "GPL" #define DRIVER_NAME "gma500" #define DRIVER_DESC "DRM driver for the Intel GMA500, GMA600, GMA3600, GMA3650" -- GitLab From 675f7ff35bd256e65d3d0f52718d8babf5d1002a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 16 Nov 2017 18:02:15 +0200 Subject: [PATCH 0350/1471] drm/i915: Fix init_clock_gating for resume MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Moving the init_clock_gating() call from intel_modeset_init_hw() to intel_modeset_gem_init() had an unintended effect of not applying some workarounds on resume. This, for example, cause some kind of corruption to appear at the top of my IVB Thinkpad X1 Carbon LVDS screen after hibernation. Fix the problem by explicitly calling init_clock_gating() from the resume path. I really hope this doesn't break something else again. At least the problems reported at https://bugs.freedesktop.org/show_bug.cgi?id=103549 didn't make a comeback, even after a hibernate cycle. v2: Reorder the init_clock_gating vs. modeset_init_hw to match the display reset path (Rodrigo) Cc: stable@vger.kernel.org Cc: Chris Wilson Cc: Rodrigo Vivi Fixes: 6ac43272768c ("drm/i915: Move init_clock_gating() back to where it was") Reviewed-by: Rodrigo Vivi Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171116160215.25715-1-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 8ea6ce7027d48..87452c5558d31 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1714,6 +1714,7 @@ static int i915_drm_resume(struct drm_device *dev) i915_gem_resume(dev_priv); intel_modeset_init_hw(dev); + intel_init_clock_gating(dev_priv); spin_lock_irq(&dev_priv->irq_lock); if (dev_priv->display.hpd_irq_setup) -- GitLab From e40dd226246d2a0fb463a04bd4c6f1b9e0a650ed Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 12:34:55 +0000 Subject: [PATCH 0351/1471] drm/i915/execlists: Listen to COMPLETE context event not ACTIVE_IDLE Since commit e1fee72c2ea2e9c0c6e6743d32a6832f21337d6c Author: Oscar Mateo Date: Thu Jul 24 17:04:40 2014 +0100 drm/i915/bdw: Avoid non-lite-restore preemptions execlists has listened to (ACTIVE_IDLE | ELEMENT_SWITCH) for detecting when one context completed and it either continued onto the next (in port 1) or idled. We would always see COMPLETE | ACTIVE_IDLE on the final context-switch event, but on recent gen it appears that we now get separate ACTIVE_IDLE and COMPLETE events. In particular, the ACTIVE_IDLE events may not be coupled to a context (since it is a general state rather than a specific context completion event). v2: Update the history, execlists did originally start out by listening to the COMPLETE event not ACTIVE_IDLE. v3: Update preempt completion test to also use COMPLETE not ACTIVE_IDLE. References: bspec/12255 References: https://bugs.freedesktop.org/show_bug.cgi?id=103800 Signed-off-by: Chris Wilson Cc: Oscar Mateo Cc: Mika Kuoppala Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Michal Winiarski Cc: Michel Thierry Acked-by: Michel Thierry Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171120123458.23242-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_lrc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index be6c39adebdf3..c2cfdfdc07226 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -154,7 +154,7 @@ #define GEN8_CTX_STATUS_LITE_RESTORE (1 << 15) #define GEN8_CTX_STATUS_COMPLETED_MASK \ - (GEN8_CTX_STATUS_ACTIVE_IDLE | \ + (GEN8_CTX_STATUS_COMPLETE | \ GEN8_CTX_STATUS_PREEMPTED | \ GEN8_CTX_STATUS_ELEMENT_SWITCH) @@ -876,7 +876,7 @@ static void execlists_submission_tasklet(unsigned long data) if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK)) continue; - if (status & GEN8_CTX_STATUS_ACTIVE_IDLE && + if (status & GEN8_CTX_STATUS_COMPLETE && buf[2*head + 1] == PREEMPT_ID) { execlists_cancel_port_requests(execlists); execlists_unwind_incomplete_requests(execlists); -- GitLab From d8747afb115f4967c4e900919d631c84ad58f573 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 12:34:56 +0000 Subject: [PATCH 0352/1471] drm/i915/execlists: Reduce completed event mask to COMPLETE | PREEMPTED Since we get a COMPLETE event when the context switch occurs on RING_HEAD == RING_TAIL and a PREEMPTED event when a switch occurs before that point, COMPLETE | PREEMPTED should cover all possible context switch completion events. We can move the ELEMENT_SWITCH info message from the COMPLETED_MASK into an assertion for when we are performing a switch to port[1]. Suggested-by: Joonas Lahtinen Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Michal Winiarski Cc: Michel Thierry Link: https://patchwork.freedesktop.org/patch/msgid/20171120123458.23242-2-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_lrc.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index c2cfdfdc07226..a2c7fc7bf4a46 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -154,9 +154,7 @@ #define GEN8_CTX_STATUS_LITE_RESTORE (1 << 15) #define GEN8_CTX_STATUS_COMPLETED_MASK \ - (GEN8_CTX_STATUS_COMPLETE | \ - GEN8_CTX_STATUS_PREEMPTED | \ - GEN8_CTX_STATUS_ELEMENT_SWITCH) + (GEN8_CTX_STATUS_COMPLETE | GEN8_CTX_STATUS_PREEMPTED) #define CTX_LRI_HEADER_0 0x01 #define CTX_CONTEXT_CONTROL 0x02 @@ -907,6 +905,8 @@ static void execlists_submission_tasklet(unsigned long data) GEM_BUG_ON(count == 0); if (--count == 0) { GEM_BUG_ON(status & GEN8_CTX_STATUS_PREEMPTED); + GEM_BUG_ON(port_isset(&port[1]) && + !(status & GEN8_CTX_STATUS_ELEMENT_SWITCH)); GEM_BUG_ON(!i915_gem_request_completed(rq)); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); -- GitLab From 1f5f9edb44ca9fd7255cdc60a1880241d2d17e48 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 12:34:57 +0000 Subject: [PATCH 0353/1471] drm/i915/execlists: Assert that we don't get mixed IDLE_ACTIVE | COMPLETE events If IDLE_ACTIVE is set, then all other bits are invalid. For us, we can assert that if we see a COMPLETE | PREEMPTED event, then it should be impossible for it to also contain an IDLE_ACTIVE flag. Suggested-by: Joonas Lahtinen Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Tvrtko Ursulin Cc: Joonas Lahtinen Cc: Michal Winiarski Cc: Michel Thierry Link: https://patchwork.freedesktop.org/patch/msgid/20171120123458.23242-3-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/intel_lrc.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index a2c7fc7bf4a46..4aed003237808 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -874,6 +874,9 @@ static void execlists_submission_tasklet(unsigned long data) if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK)) continue; + /* We should never get a COMPLETED | IDLE_ACTIVE! */ + GEM_BUG_ON(status & GEN8_CTX_STATUS_IDLE_ACTIVE); + if (status & GEN8_CTX_STATUS_COMPLETE && buf[2*head + 1] == PREEMPT_ID) { execlists_cancel_port_requests(execlists); -- GitLab From 7c52a2219da1e551e19af6a4c50619952c6d7bde Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Mon, 13 Nov 2017 23:34:52 +0000 Subject: [PATCH 0354/1471] drm/i915/perf: replace .reg accesses with i915_mmio_reg_offset This replaces accesses to the reg field of the i915_reg_t structure with the i915_mmio_reg_offset() inline function. Suggested-by: Chris Wilson Signed-off-by: Lionel Landwerlin Reviewed-by: Ewelina Musial Link: https://patchwork.freedesktop.org/patch/msgid/20171113233455.12085-2-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_perf.c | 39 ++++++++++++++++++++------------ 1 file changed, 24 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 00be015e01dfc..0f48e666098df 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -3007,7 +3007,7 @@ static bool gen8_is_valid_flex_addr(struct drm_i915_private *dev_priv, u32 addr) int i; for (i = 0; i < ARRAY_SIZE(flex_eu_regs); i++) { - if (flex_eu_regs[i].reg == addr) + if (i915_mmio_reg_offset(flex_eu_regs[i]) == addr) return true; } return false; @@ -3015,38 +3015,47 @@ static bool gen8_is_valid_flex_addr(struct drm_i915_private *dev_priv, u32 addr) static bool gen7_is_valid_b_counter_addr(struct drm_i915_private *dev_priv, u32 addr) { - return (addr >= OASTARTTRIG1.reg && addr <= OASTARTTRIG8.reg) || - (addr >= OAREPORTTRIG1.reg && addr <= OAREPORTTRIG8.reg) || - (addr >= OACEC0_0.reg && addr <= OACEC7_1.reg); + return (addr >= i915_mmio_reg_offset(OASTARTTRIG1) && + addr <= i915_mmio_reg_offset(OASTARTTRIG8)) || + (addr >= i915_mmio_reg_offset(OAREPORTTRIG1) && + addr <= i915_mmio_reg_offset(OAREPORTTRIG8)) || + (addr >= i915_mmio_reg_offset(OACEC0_0) && + addr <= i915_mmio_reg_offset(OACEC7_1)); } static bool gen7_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { - return addr == HALF_SLICE_CHICKEN2.reg || - (addr >= MICRO_BP0_0.reg && addr <= NOA_WRITE.reg) || - (addr >= OA_PERFCNT1_LO.reg && addr <= OA_PERFCNT2_HI.reg) || - (addr >= OA_PERFMATRIX_LO.reg && addr <= OA_PERFMATRIX_HI.reg); + return addr == i915_mmio_reg_offset(HALF_SLICE_CHICKEN2) || + (addr >= i915_mmio_reg_offset(MICRO_BP0_0) && + addr <= i915_mmio_reg_offset(NOA_WRITE)) || + (addr >= i915_mmio_reg_offset(OA_PERFCNT1_LO) && + addr <= i915_mmio_reg_offset(OA_PERFCNT2_HI)) || + (addr >= i915_mmio_reg_offset(OA_PERFMATRIX_LO) && + addr <= i915_mmio_reg_offset(OA_PERFMATRIX_HI)); } static bool gen8_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { return gen7_is_valid_mux_addr(dev_priv, addr) || - addr == WAIT_FOR_RC6_EXIT.reg || - (addr >= RPM_CONFIG0.reg && addr <= NOA_CONFIG(8).reg); + addr == i915_mmio_reg_offset(WAIT_FOR_RC6_EXIT) || + (addr >= i915_mmio_reg_offset(RPM_CONFIG0) && + addr <= i915_mmio_reg_offset(NOA_CONFIG(8))); } static bool gen10_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { return gen8_is_valid_mux_addr(dev_priv, addr) || - (addr >= OA_PERFCNT3_LO.reg && addr <= OA_PERFCNT4_HI.reg); + (addr >= i915_mmio_reg_offset(OA_PERFCNT3_LO) && + addr <= i915_mmio_reg_offset(OA_PERFCNT4_HI)); } static bool hsw_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) { return gen7_is_valid_mux_addr(dev_priv, addr) || (addr >= 0x25100 && addr <= 0x2FF90) || - (addr >= HSW_MBVID2_NOA0.reg && addr <= HSW_MBVID2_NOA9.reg) || - addr == HSW_MBVID2_MISR0.reg; + (addr >= i915_mmio_reg_offset(HSW_MBVID2_NOA0) && + addr <= i915_mmio_reg_offset(HSW_MBVID2_NOA9)) || + addr == i915_mmio_reg_offset(HSW_MBVID2_MISR0); } static bool chv_is_valid_mux_addr(struct drm_i915_private *dev_priv, u32 addr) @@ -3061,14 +3070,14 @@ static uint32_t mask_reg_value(u32 reg, u32 val) * WaDisableSTUnitPowerOptimization workaround. Make sure the value * programmed by userspace doesn't change this. */ - if (HALF_SLICE_CHICKEN2.reg == reg) + if (i915_mmio_reg_offset(HALF_SLICE_CHICKEN2) == reg) val = val & ~_MASKED_BIT_ENABLE(GEN8_ST_PO_DISABLE); /* WAIT_FOR_RC6_EXIT has only one bit fullfilling the function * indicated by its name and a bunch of selection fields used by OA * configs. */ - if (WAIT_FOR_RC6_EXIT.reg == reg) + if (i915_mmio_reg_offset(WAIT_FOR_RC6_EXIT) == reg) val = val & ~_MASKED_BIT_ENABLE(HSW_WAIT_FOR_RC6_EXIT_ENABLE); return val; -- GitLab From 2113184c6f6749f6e4e86a42894f67a50ead6775 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 10:20:01 +0000 Subject: [PATCH 0355/1471] drm/i915: Pull the unconditional GPU cache invalidation into request construction As the request will, in the following patch, implicitly invoke a context-switch on construction, we should precede that with a GPU TLB invalidation. Also, even before using GGTT, we always want to invalidate the TLBs for any updates (as well as the ppgtt invalidates that are unconditionally applied by execbuf). Since we almost always require the TLB invalidate, do it unconditionally on request allocation and so we can remove it from all other paths. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171120102002.22254-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 7 +----- drivers/gpu/drm/i915/i915_gem_render_state.c | 4 ---- drivers/gpu/drm/i915/i915_gem_request.c | 24 +++++++++++++++---- drivers/gpu/drm/i915/selftests/huge_pages.c | 4 ---- .../gpu/drm/i915/selftests/i915_gem_context.c | 4 ---- .../gpu/drm/i915/selftests/i915_gem_request.c | 10 -------- .../gpu/drm/i915/selftests/intel_hangcheck.c | 4 ---- 7 files changed, 20 insertions(+), 37 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 53ccb27bfe917..b7895788bc752 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1111,10 +1111,6 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, if (err) goto err_request; - err = eb->engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - goto err_request; - err = i915_switch_context(rq); if (err) goto err_request; @@ -1818,8 +1814,7 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb) /* Unconditionally flush any chipset caches (for streaming writes). */ i915_gem_chipset_flush(eb->i915); - /* Unconditionally invalidate GPU caches and TLBs. */ - return eb->engine->emit_flush(eb->request, EMIT_INVALIDATE); + return 0; } static bool i915_gem_check_execbuffer(struct drm_i915_gem_execbuffer2 *exec) diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c index c2723a06fbb4e..f7fc0df251acf 100644 --- a/drivers/gpu/drm/i915/i915_gem_render_state.c +++ b/drivers/gpu/drm/i915/i915_gem_render_state.c @@ -208,10 +208,6 @@ int i915_gem_render_state_emit(struct drm_i915_gem_request *rq) if (err) goto err_unpin; - err = engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - goto err_unpin; - err = engine->emit_bb_start(rq, so.batch_offset, so.batch_size, I915_DISPATCH_SECURE); diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index e0d6221022a83..91eae1b20c428 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -703,17 +703,31 @@ i915_gem_request_alloc(struct intel_engine_cs *engine, req->reserved_space = MIN_SPACE_FOR_ADD_REQUEST; GEM_BUG_ON(req->reserved_space < engine->emit_breadcrumb_sz); - ret = engine->request_alloc(req); - if (ret) - goto err_ctx; - - /* Record the position of the start of the request so that + /* + * Record the position of the start of the request so that * should we detect the updated seqno part-way through the * GPU processing the request, we never over-estimate the * position of the head. */ req->head = req->ring->emit; + /* Unconditionally invalidate GPU caches and TLBs. */ + ret = engine->emit_flush(req, EMIT_INVALIDATE); + if (ret) + goto err_ctx; + + ret = engine->request_alloc(req); + if (ret) { + /* + * Past the point-of-no-return. Since we may have updated + * global state after partially completing the request alloc, + * we need to commit any commands so far emitted in the + * request to the HW. + */ + __i915_add_request(req, false); + return ERR_PTR(ret); + } + /* Check that we didn't interrupt ourselves with a new request */ GEM_BUG_ON(req->timeline->seqno != req->fence.seqno); return req; diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index 01af540b6ef9a..159a2cb687658 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -989,10 +989,6 @@ static int gpu_write(struct i915_vma *vma, i915_vma_unpin(batch); i915_vma_close(batch); - err = rq->engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - goto err_request; - err = i915_switch_context(rq); if (err) goto err_request; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c index c82780a9d4556..4ff30b9af1fec 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c @@ -158,10 +158,6 @@ static int gpu_fill(struct drm_i915_gem_object *obj, goto err_batch; } - err = engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - goto err_request; - err = i915_switch_context(rq); if (err) goto err_request; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_request.c b/drivers/gpu/drm/i915/selftests/i915_gem_request.c index 6bce99050e947..d7bf53ff8f84e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_request.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_request.c @@ -459,10 +459,6 @@ empty_request(struct intel_engine_cs *engine, if (IS_ERR(request)) return request; - err = engine->emit_flush(request, EMIT_INVALIDATE); - if (err) - goto out_request; - err = i915_switch_context(request); if (err) goto out_request; @@ -675,9 +671,6 @@ static int live_all_engines(void *arg) goto out_request; } - err = engine->emit_flush(request[id], EMIT_INVALIDATE); - GEM_BUG_ON(err); - err = i915_switch_context(request[id]); GEM_BUG_ON(err); @@ -797,9 +790,6 @@ static int live_sequential_engines(void *arg) } } - err = engine->emit_flush(request[id], EMIT_INVALIDATE); - GEM_BUG_ON(err); - err = i915_switch_context(request[id]); GEM_BUG_ON(err); diff --git a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c index 71ce06680d66a..145bdc26553c0 100644 --- a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c @@ -114,10 +114,6 @@ static int emit_recurse_batch(struct hang *h, if (err) goto unpin_vma; - err = rq->engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - goto unpin_hws; - err = i915_switch_context(rq); if (err) goto unpin_hws; -- GitLab From 3fef5cda970124a15c553c1672d800e40fc08a9e Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 10:20:02 +0000 Subject: [PATCH 0356/1471] drm/i915: Automatic i915_switch_context for legacy During request construction, after pinning the context we know whether or not we have to emit a context switch. So move this common operation from every caller into i915_gem_request_alloc() itself. v2: Always submit the request if we emitted some commands during request construction, as typically it also involves changes in global state. Signed-off-by: Chris Wilson Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171120102002.22254-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 2 +- drivers/gpu/drm/i915/i915_gem_context.c | 7 +------ drivers/gpu/drm/i915/i915_gem_execbuffer.c | 8 -------- drivers/gpu/drm/i915/i915_gem_request.c | 4 ++++ drivers/gpu/drm/i915/i915_perf.c | 3 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 4 ++++ drivers/gpu/drm/i915/selftests/huge_pages.c | 10 +++------- drivers/gpu/drm/i915/selftests/i915_gem_context.c | 4 ---- drivers/gpu/drm/i915/selftests/i915_gem_request.c | 10 ---------- drivers/gpu/drm/i915/selftests/intel_hangcheck.c | 5 ----- 10 files changed, 14 insertions(+), 43 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 61ba321e9970a..e07eb0beef13b 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5045,7 +5045,7 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915) goto out_ctx; } - err = i915_switch_context(rq); + err = 0; if (engine->init_context) err = engine->init_context(rq); diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 2db0406950356..c1efbaf02bf2a 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -842,8 +842,7 @@ int i915_switch_context(struct drm_i915_gem_request *req) struct intel_engine_cs *engine = req->engine; lockdep_assert_held(&req->i915->drm.struct_mutex); - if (i915_modparams.enable_execlists) - return 0; + GEM_BUG_ON(i915_modparams.enable_execlists); if (!req->ctx->engine[engine->id].state) { struct i915_gem_context *to = req->ctx; @@ -899,7 +898,6 @@ int i915_gem_switch_to_kernel_context(struct drm_i915_private *dev_priv) for_each_engine(engine, dev_priv, id) { struct drm_i915_gem_request *req; - int ret; if (engine_has_idle_kernel_context(engine)) continue; @@ -922,10 +920,7 @@ int i915_gem_switch_to_kernel_context(struct drm_i915_private *dev_priv) GFP_KERNEL); } - ret = i915_switch_context(req); i915_add_request(req); - if (ret) - return ret; } return 0; diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index b7895788bc752..14d9e61a1e06d 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -1111,10 +1111,6 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, if (err) goto err_request; - err = i915_switch_context(rq); - if (err) - goto err_request; - err = eb->engine->emit_bb_start(rq, batch->node.start, PAGE_SIZE, cache->gen > 5 ? 0 : I915_DISPATCH_SECURE); @@ -1960,10 +1956,6 @@ static int eb_submit(struct i915_execbuffer *eb) if (err) return err; - err = i915_switch_context(eb->request); - if (err) - return err; - if (eb->args->flags & I915_EXEC_GEN7_SOL_RESET) { err = i915_reset_gen7_sol_offsets(eb->request); if (err) diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index 91eae1b20c428..86e2346357cf9 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -624,6 +624,10 @@ i915_gem_request_alloc(struct intel_engine_cs *engine, if (ret) goto err_unpin; + ret = intel_ring_wait_for_space(ring, MIN_SPACE_FOR_ADD_REQUEST); + if (ret) + goto err_unreserve; + /* Move the oldest request to the slab-cache (if not in use!) */ req = list_first_entry_or_null(&engine->timeline->requests, typeof(*req), link); diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index 0f48e666098df..fd150099978ca 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -1726,10 +1726,9 @@ static int gen8_switch_to_updated_kernel_context(struct drm_i915_private *dev_pr GFP_KERNEL); } - ret = i915_switch_context(req); i915_add_request(req); - return ret; + return 0; } /* diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 12e734b29463d..be98868115bf9 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1592,6 +1592,10 @@ static int ring_request_alloc(struct drm_i915_gem_request *request) if (ret) return ret; + ret = i915_switch_context(request); + if (ret) + return ret; + request->reserved_space -= LEGACY_REQUEST_SIZE; return 0; } diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index 159a2cb687658..db7a0a1f2960e 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -989,13 +989,9 @@ static int gpu_write(struct i915_vma *vma, i915_vma_unpin(batch); i915_vma_close(batch); - err = i915_switch_context(rq); - if (err) - goto err_request; - - err = rq->engine->emit_bb_start(rq, - batch->node.start, batch->node.size, - flags); + err = engine->emit_bb_start(rq, + batch->node.start, batch->node.size, + flags); if (err) goto err_request; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c index 4ff30b9af1fec..09340b3c11562 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c @@ -158,10 +158,6 @@ static int gpu_fill(struct drm_i915_gem_object *obj, goto err_batch; } - err = i915_switch_context(rq); - if (err) - goto err_request; - flags = 0; if (INTEL_GEN(vm->i915) <= 5) flags |= I915_DISPATCH_SECURE; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_request.c b/drivers/gpu/drm/i915/selftests/i915_gem_request.c index d7bf53ff8f84e..647bf2bbd7993 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_request.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_request.c @@ -459,10 +459,6 @@ empty_request(struct intel_engine_cs *engine, if (IS_ERR(request)) return request; - err = i915_switch_context(request); - if (err) - goto out_request; - err = engine->emit_bb_start(request, batch->node.start, batch->node.size, @@ -671,9 +667,6 @@ static int live_all_engines(void *arg) goto out_request; } - err = i915_switch_context(request[id]); - GEM_BUG_ON(err); - err = engine->emit_bb_start(request[id], batch->node.start, batch->node.size, @@ -790,9 +783,6 @@ static int live_sequential_engines(void *arg) } } - err = i915_switch_context(request[id]); - GEM_BUG_ON(err); - err = engine->emit_bb_start(request[id], batch->node.start, batch->node.size, diff --git a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c index 145bdc26553c0..1bbb8c46e2d9c 100644 --- a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c @@ -114,10 +114,6 @@ static int emit_recurse_batch(struct hang *h, if (err) goto unpin_vma; - err = i915_switch_context(rq); - if (err) - goto unpin_hws; - i915_vma_move_to_active(vma, rq, 0); if (!i915_gem_object_has_active_reference(vma->obj)) { i915_gem_object_get(vma->obj); @@ -169,7 +165,6 @@ static int emit_recurse_batch(struct hang *h, err = rq->engine->emit_bb_start(rq, vma->node.start, PAGE_SIZE, flags); -unpin_hws: i915_vma_unpin(hws); unpin_vma: i915_vma_unpin(vma); -- GitLab From 9f9b2792b6d38fbc8028f31b41ebd38a3f28a401 Mon Sep 17 00:00:00 2001 From: Lionel Landwerlin Date: Fri, 27 Oct 2017 15:59:31 +0100 Subject: [PATCH 0357/1471] drm/i915/perf: reuse timestamp frequency from device info Now that we have this stored in the device info, we can drop it from perf part of the driver. Note that this requires to init perf after we've computed the frequency, hence why we move i915_perf_init() from i915_driver_init_early() to after intel_device_info_runtime_init(). v2: Use div_u64 (Chris) v3: Drop u64 divs by switching to kHz (Chris/Ville) Move i915_perf_fini to i915_driver_cleanup_hw (Matthew) Signed-off-by: Lionel Landwerlin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171113181902.12411-2-lionel.g.landwerlin@intel.com --- drivers/gpu/drm/i915/i915_drv.c | 7 ++++--- drivers/gpu/drm/i915/i915_drv.h | 1 - drivers/gpu/drm/i915/i915_perf.c | 36 +++++--------------------------- 3 files changed, 9 insertions(+), 35 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 87452c5558d31..2b4ced93011b7 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -936,8 +936,6 @@ static int i915_driver_init_early(struct drm_i915_private *dev_priv, intel_detect_preproduction_hw(dev_priv); - i915_perf_init(dev_priv); - return 0; err_irq: @@ -954,7 +952,6 @@ static int i915_driver_init_early(struct drm_i915_private *dev_priv, */ static void i915_driver_cleanup_early(struct drm_i915_private *dev_priv) { - i915_perf_fini(dev_priv); i915_gem_load_cleanup(dev_priv); intel_irq_fini(dev_priv); i915_workqueues_cleanup(dev_priv); @@ -1101,6 +1098,8 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv) intel_sanitize_options(dev_priv); + i915_perf_init(dev_priv); + ret = i915_ggtt_probe_hw(dev_priv); if (ret) return ret; @@ -1206,6 +1205,8 @@ static void i915_driver_cleanup_hw(struct drm_i915_private *dev_priv) { struct pci_dev *pdev = dev_priv->drm.pdev; + i915_perf_fini(dev_priv); + if (pdev->msi_enabled) pci_disable_msi(pdev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 36bb4927484a0..2b766253f9706 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2619,7 +2619,6 @@ struct drm_i915_private { bool periodic; int period_exponent; - int timestamp_frequency; struct i915_oa_config test_config; diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index fd150099978ca..adbce839a918b 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -2690,8 +2690,8 @@ i915_perf_open_ioctl_locked(struct drm_i915_private *dev_priv, static u64 oa_exponent_to_ns(struct drm_i915_private *dev_priv, int exponent) { - return div_u64(1000000000ULL * (2ULL << exponent), - dev_priv->perf.oa.timestamp_frequency); + return div64_u64(1000000000ULL * (2ULL << exponent), + 1000ULL * INTEL_INFO(dev_priv)->cs_timestamp_frequency_khz); } /** @@ -3423,8 +3423,6 @@ static struct ctl_table dev_root[] = { */ void i915_perf_init(struct drm_i915_private *dev_priv) { - dev_priv->perf.oa.timestamp_frequency = 0; - if (IS_HASWELL(dev_priv)) { dev_priv->perf.oa.ops.is_valid_b_counter_reg = gen7_is_valid_b_counter_addr; @@ -3440,8 +3438,6 @@ void i915_perf_init(struct drm_i915_private *dev_priv) dev_priv->perf.oa.ops.oa_hw_tail_read = gen7_oa_hw_tail_read; - dev_priv->perf.oa.timestamp_frequency = 12500000; - dev_priv->perf.oa.oa_formats = hsw_oa_formats; } else if (i915_modparams.enable_execlists) { /* Note: that although we could theoretically also support the @@ -3485,23 +3481,6 @@ void i915_perf_init(struct drm_i915_private *dev_priv) dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16); } - - switch (dev_priv->info.platform) { - case INTEL_BROADWELL: - dev_priv->perf.oa.timestamp_frequency = 12500000; - break; - case INTEL_BROXTON: - case INTEL_GEMINILAKE: - dev_priv->perf.oa.timestamp_frequency = 19200000; - break; - case INTEL_SKYLAKE: - case INTEL_KABYLAKE: - case INTEL_COFFEELAKE: - dev_priv->perf.oa.timestamp_frequency = 12000000; - break; - default: - break; - } } else if (IS_GEN10(dev_priv)) { dev_priv->perf.oa.ops.is_valid_b_counter_reg = gen7_is_valid_b_counter_addr; @@ -3517,15 +3496,10 @@ void i915_perf_init(struct drm_i915_private *dev_priv) dev_priv->perf.oa.ctx_flexeu0_offset = 0x3de; dev_priv->perf.oa.gen8_valid_ctx_bit = (1<<16); - - /* Default frequency, although we need to read it from - * the register as it might vary between parts. - */ - dev_priv->perf.oa.timestamp_frequency = 12000000; } } - if (dev_priv->perf.oa.timestamp_frequency) { + if (dev_priv->perf.oa.ops.enable_metric_set) { hrtimer_init(&dev_priv->perf.oa.poll_check_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); dev_priv->perf.oa.poll_check_timer.function = oa_poll_check_timer_cb; @@ -3535,8 +3509,8 @@ void i915_perf_init(struct drm_i915_private *dev_priv) mutex_init(&dev_priv->perf.lock); spin_lock_init(&dev_priv->perf.oa.oa_buffer.ptr_lock); - oa_sample_rate_hard_limit = - dev_priv->perf.oa.timestamp_frequency / 2; + oa_sample_rate_hard_limit = 1000 * + (INTEL_INFO(dev_priv)->cs_timestamp_frequency_khz / 2); dev_priv->perf.sysctl_header = register_sysctl_table(dev_root); mutex_init(&dev_priv->perf.metrics_lock); -- GitLab From 2a6c4241fcc3d942a0a6b09e445561d099fa7884 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 13:26:06 +0000 Subject: [PATCH 0358/1471] drm/i915/selftest: Make guc clients static Make the private array used for stashing test clients static, to silence sparse. References: 55bd6bd75717 ("drm/i915/selftests: Add a GuC doorbells selftest") Signed-off-by: Chris Wilson Cc: Michel Thierry Cc: Michal Wajdeczko Cc: Daniele Ceraolo Spurio Link: https://patchwork.freedesktop.org/patch/msgid/20171120132606.4254-1-chris@chris-wilson.co.uk Reviewed-by: Michel Thierry --- drivers/gpu/drm/i915/selftests/intel_guc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c index f10029e18820b..f1baeda8b4a82 100644 --- a/drivers/gpu/drm/i915/selftests/intel_guc.c +++ b/drivers/gpu/drm/i915/selftests/intel_guc.c @@ -27,7 +27,7 @@ /* max doorbell number + negative test for each client type */ #define ATTEMPTS (GUC_NUM_DOORBELLS + GUC_CLIENT_PRIORITY_NUM) -struct intel_guc_client *clients[ATTEMPTS]; +static struct intel_guc_client *clients[ATTEMPTS]; static bool available_dbs(struct intel_guc *guc, u32 priority) { -- GitLab From ba74cb10c775c839f6e1d0fabd1e772eabd9c43f Mon Sep 17 00:00:00 2001 From: Michel Thierry Date: Mon, 20 Nov 2017 12:34:58 +0000 Subject: [PATCH 0359/1471] drm/i915/execlists: Delay writing to ELSP until HW has processed the previous write The hardware needs some time to process the information received in the ExecList Submission Port, and expects us to not write anything more until it has 'acknowledged' this new submission by sending an IDLE_ACTIVE or PREEMPTED CSB event. If we do not follow this, the driver could write new data into the ELSP before HW had finishing fetching the previous one, putting us in 'undefined behaviour' space. This seems to be the problem causing the spurious PREEMPTED & COMPLETE events after a COMPLETE like the one below: [] vcs0: sw rd pointer = 2, hw wr pointer = 0, current 'head' = 3. [] vcs0: Execlist CSB[0]: 0x00000018 _ 0x00000007 [] vcs0: Execlist CSB[1]: 0x00000001 _ 0x00000000 [] vcs0: Execlist CSB[2]: 0x00000018 _ 0x00000007 <<< COMPLETE [] vcs0: Execlist CSB[3]: 0x00000012 _ 0x00000007 <<< PREEMPTED & COMPLETE [] vcs0: Execlist CSB[4]: 0x00008002 _ 0x00000006 [] vcs0: Execlist CSB[5]: 0x00000014 _ 0x00000006 The ELSP writes that lead to this CSB sequence show that the HW hadn't started executing the previous execlist (the one with only ctx 0x6) by the time the new one was submitted; this is a bit more clear in the data show in the EXECLIST_STATUS register at the time of the ELSP write. [] vcs0: ELSP[0] = 0x0_0 [execlist1] - status_reg = 0x0_302 [] vcs0: ELSP[1] = 0x6_fedb2119 [execlist0] - status_reg = 0x0_8302 [] vcs0: ELSP[2] = 0x7_fedaf119 [execlist1] - status_reg = 0x0_8308 [] vcs0: ELSP[3] = 0x6_fedb2119 [execlist0] - status_reg = 0x7_8308 Note that having to wait for this ack does not disable lite-restores, although it may reduce their numbers. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102035 Signed-off-by: Michel Thierry Link: https://patchwork.freedesktop.org/patch/msgid/<20171118003038.7935-1-michel.thierry@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20171120123458.23242-4-chris@chris-wilson.co.uk Reviewed-by: Chris Wilson Tested-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_lrc.c | 22 ++++++++++++++++++++++ drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + 2 files changed, 23 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 4aed003237808..8d0c493888634 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -477,6 +477,7 @@ static void execlists_submit_ports(struct intel_engine_cs *engine) elsp_write(desc, elsp); } + execlists_clear_active(&engine->execlists, EXECLISTS_ACTIVE_HWACK); } static bool ctx_single_port_submission(const struct i915_gem_context *ctx) @@ -529,6 +530,7 @@ static void inject_preempt_context(struct intel_engine_cs *engine) elsp_write(0, elsp); elsp_write(ce->lrc_desc, elsp); + execlists_clear_active(&engine->execlists, EXECLISTS_ACTIVE_HWACK); } static void execlists_dequeue(struct intel_engine_cs *engine) @@ -575,9 +577,20 @@ static void execlists_dequeue(struct intel_engine_cs *engine) * know the next preemption status we see corresponds * to this ELSP update. */ + GEM_BUG_ON(!port_count(&port[0])); if (port_count(&port[0]) > 1) goto unlock; + /* + * If we write to ELSP a second time before the HW has had + * a chance to respond to the previous write, we can confuse + * the HW and hit "undefined behaviour". After writing to ELSP, + * we must then wait until we see a context-switch event from + * the HW to indicate that it has had a chance to respond. + */ + if (!execlists_is_active(execlists, EXECLISTS_ACTIVE_HWACK)) + goto unlock; + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915) && rb_entry(rb, struct i915_priolist, node)->priority > max(last->priotree.priority, 0)) { @@ -871,6 +884,15 @@ static void execlists_submission_tasklet(unsigned long data) GEM_TRACE("%s csb[%dd]: status=0x%08x:0x%08x\n", engine->name, head, status, buf[2*head + 1]); + + if (status & (GEN8_CTX_STATUS_IDLE_ACTIVE | + GEN8_CTX_STATUS_PREEMPTED)) + execlists_set_active(execlists, + EXECLISTS_ACTIVE_HWACK); + if (status & GEN8_CTX_STATUS_ACTIVE_IDLE) + execlists_clear_active(execlists, + EXECLISTS_ACTIVE_HWACK); + if (!(status & GEN8_CTX_STATUS_COMPLETED_MASK)) continue; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index c00804ed64c66..e437a9556c087 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -251,6 +251,7 @@ struct intel_engine_execlists { unsigned int active; #define EXECLISTS_ACTIVE_USER 0 #define EXECLISTS_ACTIVE_PREEMPT 1 +#define EXECLISTS_ACTIVE_HWACK 2 /** * @port_mask: number of execlist ports - 1 -- GitLab From 3bd6099fd687fdf2874b65a806e85a793496ba10 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 20:29:16 +0200 Subject: [PATCH 0360/1471] drm/vmwgfx: Remove bogus crtc coords vs fb size check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Throw away the bugs crtc coords vs. fb size check. Crtc coords don't define the viewport inside the fb, that's a job for the src coords, which have been checked by the core already. Cc: VMware Graphics Cc: Sinclair Yeh Cc: Thomas Hellstrom Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171101182920.14386-2-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Reviewed-by: Sinclair Yeh --- drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index b850562fbdd65..4f127a3f99e4b 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -476,12 +476,6 @@ int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, vcs = vmw_connector_state_to_vcs(du->connector.state); - if ((dest.x2 > new_fb->width || - dest.y2 > new_fb->height)) { - DRM_ERROR("CRTC area outside of framebuffer\n"); - return -EINVAL; - } - /* Only one active implicit framebuffer at a time. */ mutex_lock(&dev_priv->global_kms_state_mutex); if (vcs->is_implicit && dev_priv->implicit_fb && -- GitLab From 13dd5b62cee2d9f5d300c83b65d81c148d7bff06 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 20:29:17 +0200 Subject: [PATCH 0361/1471] drm/vmwgfx: Use drm_plane_helper_check_state() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Atomic drivers have no reason to use drm_plane_helper_check_update() instead of drm_plane_helper_check_state(). So let's switch over. Cc: VMware Graphics Cc: Sinclair Yeh Cc: Thomas Hellstrom Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171101182920.14386-3-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Reviewed-by: Sinclair Yeh --- drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 17 +++-------------- 1 file changed, 3 insertions(+), 14 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 4f127a3f99e4b..8e5cdf8b2cb95 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -442,29 +442,18 @@ int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, struct drm_plane_state *state) { struct drm_framebuffer *new_fb = state->fb; - bool visible; - - struct drm_rect src = { - .x1 = state->src_x, - .y1 = state->src_y, - .x2 = state->src_x + state->src_w, - .y2 = state->src_y + state->src_h, - }; - struct drm_rect dest = { + struct drm_rect clip = { .x1 = state->crtc_x, .y1 = state->crtc_y, .x2 = state->crtc_x + state->crtc_w, .y2 = state->crtc_y + state->crtc_h, }; - struct drm_rect clip = dest; int ret; - ret = drm_plane_helper_check_update(plane, state->crtc, new_fb, - &src, &dest, &clip, - DRM_MODE_ROTATE_0, + ret = drm_plane_helper_check_state(state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, - false, true, &visible); + false, true); if (!ret && new_fb) { -- GitLab From 58a275aa956498ccf964678767560b0c91f31a23 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 20:29:18 +0200 Subject: [PATCH 0362/1471] drm/vmwgfx: Try to fix plane clipping MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Try to fix the code to actually clip the plane to the crtc bounds instead of the user provided crtc coordinates (which would be a no-op since those are exactly the coordinates before clipping). Cc: VMware Graphics Cc: Sinclair Yeh Cc: Thomas Hellstrom Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171101182920.14386-4-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter Reviewed-by: Sinclair Yeh --- drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 8e5cdf8b2cb95..9eb4d0bc1546b 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -441,20 +441,23 @@ vmw_du_cursor_plane_atomic_update(struct drm_plane *plane, int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, struct drm_plane_state *state) { + struct drm_crtc_state *crtc_state = NULL; struct drm_framebuffer *new_fb = state->fb; - struct drm_rect clip = { - .x1 = state->crtc_x, - .y1 = state->crtc_y, - .x2 = state->crtc_x + state->crtc_w, - .y2 = state->crtc_y + state->crtc_h, - }; + struct drm_rect clip = {}; int ret; - ret = drm_plane_helper_check_state(state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + if (state->crtc) + crtc_state = drm_atomic_get_new_crtc_state(state->state, state->crtc); + if (crtc_state && crtc_state->enable) { + clip.x2 = crtc_state->adjusted_mode.hdisplay; + clip.y2 = crtc_state->adjusted_mode.vdisplay; + } + + ret = drm_plane_helper_check_state(state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); if (!ret && new_fb) { struct drm_crtc *crtc = state->crtc; -- GitLab From 10b47ee02d1ae66160058241cf5b962f64e81b47 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 22:15:58 +0200 Subject: [PATCH 0363/1471] drm: Check crtc_state->enable rather than crtc->enabled in drm_plane_helper_check_state() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_plane_helper_check_state() is supposed to do things the atomic way, so it should not be inspecting crtc->enabled. Rather we should be looking at crtc_state->enable. We have a slight complication due to drm_plane_helper_check_update() reusing drm_plane_helper_check_state() for non-atomic drivers. Thus we'll have to pass the crtc_state in manally and construct a fake crtc_state in drm_plane_helper_check_update(). v2: Fix the WARNs about plane_state->crtc matching crtc_state->crtc Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171101201558.6059-1-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter --- drivers/gpu/drm/arm/hdlcd_crtc.c | 2 +- drivers/gpu/drm/arm/malidp_planes.c | 3 +- drivers/gpu/drm/drm_plane_helper.c | 51 ++++++++++++--------- drivers/gpu/drm/drm_simple_kms_helper.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 2 + drivers/gpu/drm/imx/ipuv3-plane.c | 2 +- drivers/gpu/drm/mediatek/mtk_drm_plane.c | 2 +- drivers/gpu/drm/meson/meson_plane.c | 2 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 4 +- drivers/gpu/drm/nouveau/nv50_display.c | 6 ++- drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 2 +- drivers/gpu/drm/tegra/dc.c | 4 +- drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 2 +- drivers/gpu/drm/zte/zx_plane.c | 4 +- include/drm/drm_plane_helper.h | 3 +- 15 files changed, 52 insertions(+), 39 deletions(-) diff --git a/drivers/gpu/drm/arm/hdlcd_crtc.c b/drivers/gpu/drm/arm/hdlcd_crtc.c index 72b22b805412b..14721723fa8ac 100644 --- a/drivers/gpu/drm/arm/hdlcd_crtc.c +++ b/drivers/gpu/drm/arm/hdlcd_crtc.c @@ -252,7 +252,7 @@ static int hdlcd_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(state, &clip, + return drm_plane_helper_check_state(state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true); diff --git a/drivers/gpu/drm/arm/malidp_planes.c b/drivers/gpu/drm/arm/malidp_planes.c index 94e7e3fa3408c..492d99dd55d46 100644 --- a/drivers/gpu/drm/arm/malidp_planes.c +++ b/drivers/gpu/drm/arm/malidp_planes.c @@ -150,7 +150,8 @@ static int malidp_se_check_scaling(struct malidp_plane *mp, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, &clip, 0, INT_MAX, true, true); + ret = drm_plane_helper_check_state(state, crtc_state, &clip, + 0, INT_MAX, true, true); if (ret) return ret; diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index 759ed93f4ba8f..eef0ffcd7ed52 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -101,7 +101,8 @@ static int get_connectors_for_crtc(struct drm_crtc *crtc, /** * drm_plane_helper_check_state() - Check plane state for validity - * @state: plane state to check + * @plane_state: plane state to check + * @crtc_state: crtc state to check * @clip: integer clipping coordinates * @min_scale: minimum @src:@dest scaling factor in 16.16 fixed point * @max_scale: maximum @src:@dest scaling factor in 16.16 fixed point @@ -120,35 +121,37 @@ static int get_connectors_for_crtc(struct drm_crtc *crtc, * RETURNS: * Zero if update appears valid, error code on failure */ -int drm_plane_helper_check_state(struct drm_plane_state *state, +int drm_plane_helper_check_state(struct drm_plane_state *plane_state, + const struct drm_crtc_state *crtc_state, const struct drm_rect *clip, int min_scale, int max_scale, bool can_position, bool can_update_disabled) { - struct drm_crtc *crtc = state->crtc; - struct drm_framebuffer *fb = state->fb; - struct drm_rect *src = &state->src; - struct drm_rect *dst = &state->dst; - unsigned int rotation = state->rotation; + struct drm_framebuffer *fb = plane_state->fb; + struct drm_rect *src = &plane_state->src; + struct drm_rect *dst = &plane_state->dst; + unsigned int rotation = plane_state->rotation; int hscale, vscale; - *src = drm_plane_state_src(state); - *dst = drm_plane_state_dest(state); + WARN_ON(plane_state->crtc && plane_state->crtc != crtc_state->crtc); + + *src = drm_plane_state_src(plane_state); + *dst = drm_plane_state_dest(plane_state); if (!fb) { - state->visible = false; + plane_state->visible = false; return 0; } /* crtc should only be NULL when disabling (i.e., !fb) */ - if (WARN_ON(!crtc)) { - state->visible = false; + if (WARN_ON(!plane_state->crtc)) { + plane_state->visible = false; return 0; } - if (!crtc->enabled && !can_update_disabled) { + if (!crtc_state->enable && !can_update_disabled) { DRM_DEBUG_KMS("Cannot update plane of a disabled CRTC.\n"); return -EINVAL; } @@ -160,16 +163,16 @@ int drm_plane_helper_check_state(struct drm_plane_state *state, vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale); if (hscale < 0 || vscale < 0) { DRM_DEBUG_KMS("Invalid scaling of plane\n"); - drm_rect_debug_print("src: ", &state->src, true); - drm_rect_debug_print("dst: ", &state->dst, false); + drm_rect_debug_print("src: ", &plane_state->src, true); + drm_rect_debug_print("dst: ", &plane_state->dst, false); return -ERANGE; } - state->visible = drm_rect_clip_scaled(src, dst, clip, hscale, vscale); + plane_state->visible = drm_rect_clip_scaled(src, dst, clip, hscale, vscale); drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, rotation); - if (!state->visible) + if (!plane_state->visible) /* * Plane isn't visible; some drivers can handle this * so we just return success here. Drivers that can't @@ -230,7 +233,7 @@ int drm_plane_helper_check_update(struct drm_plane *plane, bool can_update_disabled, bool *visible) { - struct drm_plane_state state = { + struct drm_plane_state plane_state = { .plane = plane, .crtc = crtc, .fb = fb, @@ -245,18 +248,22 @@ int drm_plane_helper_check_update(struct drm_plane *plane, .rotation = rotation, .visible = *visible, }; + struct drm_crtc_state crtc_state = { + .crtc = crtc, + .enable = crtc->enabled, + }; int ret; - ret = drm_plane_helper_check_state(&state, clip, + ret = drm_plane_helper_check_state(&plane_state, &crtc_state, clip, min_scale, max_scale, can_position, can_update_disabled); if (ret) return ret; - *src = state.src; - *dst = state.dst; - *visible = state.visible; + *src = plane_state.src; + *dst = plane_state.dst; + *visible = plane_state.visible; return 0; } diff --git a/drivers/gpu/drm/drm_simple_kms_helper.c b/drivers/gpu/drm/drm_simple_kms_helper.c index dc9fd109de14f..d428c805025c9 100644 --- a/drivers/gpu/drm/drm_simple_kms_helper.c +++ b/drivers/gpu/drm/drm_simple_kms_helper.c @@ -103,7 +103,7 @@ static int drm_simple_kms_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(plane_state, &clip, + ret = drm_plane_helper_check_state(plane_state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index f780f39e0758c..5e75d762745aa 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9419,6 +9419,7 @@ static int intel_check_cursor(struct intel_crtc_state *crtc_state, int ret; ret = drm_plane_helper_check_state(&plane_state->base, + &crtc_state->base, &plane_state->clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, @@ -12842,6 +12843,7 @@ intel_check_primary_plane(struct intel_plane *plane, } ret = drm_plane_helper_check_state(&state->base, + &crtc_state->base, &state->clip, min_scale, max_scale, can_position, true); diff --git a/drivers/gpu/drm/imx/ipuv3-plane.c b/drivers/gpu/drm/imx/ipuv3-plane.c index 247c60e6bed27..51b23ac175e39 100644 --- a/drivers/gpu/drm/imx/ipuv3-plane.c +++ b/drivers/gpu/drm/imx/ipuv3-plane.c @@ -342,7 +342,7 @@ static int ipu_plane_atomic_check(struct drm_plane *plane, clip.y1 = 0; clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, &clip, + ret = drm_plane_helper_check_state(state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, can_position, true); diff --git a/drivers/gpu/drm/mediatek/mtk_drm_plane.c b/drivers/gpu/drm/mediatek/mtk_drm_plane.c index 6f121891430fa..7ebb336577047 100644 --- a/drivers/gpu/drm/mediatek/mtk_drm_plane.c +++ b/drivers/gpu/drm/mediatek/mtk_drm_plane.c @@ -111,7 +111,7 @@ static int mtk_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->mode.hdisplay; clip.y2 = crtc_state->mode.vdisplay; - return drm_plane_helper_check_state(state, &clip, + return drm_plane_helper_check_state(state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, true, true); diff --git a/drivers/gpu/drm/meson/meson_plane.c b/drivers/gpu/drm/meson/meson_plane.c index 17e96fa478685..2a00b720c3718 100644 --- a/drivers/gpu/drm/meson/meson_plane.c +++ b/drivers/gpu/drm/meson/meson_plane.c @@ -61,7 +61,7 @@ static int meson_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->mode.hdisplay; clip.y2 = crtc_state->mode.vdisplay; - return drm_plane_helper_check_state(state, &clip, + return drm_plane_helper_check_state(state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, true, true); diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 4b22ac3413a10..1c194a9453d97 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -348,8 +348,8 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state, min_scale = FRAC_16_16(1, 8); max_scale = FRAC_16_16(8, 1); - ret = drm_plane_helper_check_state(state, &clip, min_scale, - max_scale, true, true); + ret = drm_plane_helper_check_state(state, crtc_state, &clip, + min_scale, max_scale, true, true); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index d8ddb73967216..0738ae1f6cf4e 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -1143,7 +1143,8 @@ nv50_curs_acquire(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw, { int ret; - ret = drm_plane_helper_check_state(&asyw->state, &asyw->clip, + ret = drm_plane_helper_check_state(&asyw->state, &asyh->state, + &asyw->clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, true, true); @@ -1432,7 +1433,8 @@ nv50_base_acquire(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw, if (!fb->format->depth) return -EINVAL; - ret = drm_plane_helper_check_state(&asyw->state, &asyw->clip, + ret = drm_plane_helper_check_state(&asyw->state, &asyh->state, + &asyw->clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true); diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c index 19128b4dea544..36d0f101e30d0 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c @@ -659,7 +659,7 @@ static int vop_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, &clip, + ret = drm_plane_helper_check_state(state, crtc_state, &clip, min_scale, max_scale, true, true); if (ret) diff --git a/drivers/gpu/drm/tegra/dc.c b/drivers/gpu/drm/tegra/dc.c index 4df39112e38ec..08b899fd74fc4 100644 --- a/drivers/gpu/drm/tegra/dc.c +++ b/drivers/gpu/drm/tegra/dc.c @@ -500,8 +500,8 @@ static int tegra_plane_state_add(struct tegra_plane *plane, clip.y2 = crtc_state->mode.vdisplay; /* Check plane state for visibility and calculate clipping bounds */ - err = drm_plane_helper_check_state(state, &clip, 0, INT_MAX, - true, true); + err = drm_plane_helper_check_state(state, crtc_state, &clip, + 0, INT_MAX, true, true); if (err < 0) return err; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 9eb4d0bc1546b..91555c0370e1d 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -454,7 +454,7 @@ int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, clip.y2 = crtc_state->adjusted_mode.vdisplay; } - ret = drm_plane_helper_check_state(state, &clip, + ret = drm_plane_helper_check_state(state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true); diff --git a/drivers/gpu/drm/zte/zx_plane.c b/drivers/gpu/drm/zte/zx_plane.c index 18e7634932649..ee0002529b8f8 100644 --- a/drivers/gpu/drm/zte/zx_plane.c +++ b/drivers/gpu/drm/zte/zx_plane.c @@ -80,7 +80,7 @@ static int zx_vl_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(plane_state, &clip, + return drm_plane_helper_check_state(plane_state, crtc_state, &clip, min_scale, max_scale, true, true); } @@ -315,7 +315,7 @@ static int zx_gl_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(plane_state, &clip, + return drm_plane_helper_check_state(plane_state, crtc_state, &clip, DRM_PLANE_HELPER_NO_SCALING, DRM_PLANE_HELPER_NO_SCALING, false, true); diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index 7c8a00ceadb74..41b8309b0a57b 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -38,7 +38,8 @@ */ #define DRM_PLANE_HELPER_NO_SCALING (1<<16) -int drm_plane_helper_check_state(struct drm_plane_state *state, +int drm_plane_helper_check_state(struct drm_plane_state *plane_state, + const struct drm_crtc_state *crtc_state, const struct drm_rect *clip, int min_scale, int max_scale, bool can_position, -- GitLab From a01cb8ba3f6282934cff65e89ab36b18b14cbe27 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Wed, 1 Nov 2017 22:16:19 +0200 Subject: [PATCH 0364/1471] drm: Move drm_plane_helper_check_state() into drm_atomic_helper.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drm_plane_helper_check_update() isn't a transitional helper, so let's rename it to drm_atomic_helper_check_plane_state() and move it into drm_atomic_helper.c. v2: Fix the WARNs about plane_state->crtc matching crtc_state->crtc Cc: Daniel Vetter Suggested-by: Daniel Vetter Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171101201619.6175-1-ville.syrjala@linux.intel.com Reviewed-by: Daniel Vetter --- drivers/gpu/drm/arm/hdlcd_crtc.c | 8 +- drivers/gpu/drm/arm/malidp_planes.c | 4 +- drivers/gpu/drm/drm_atomic_helper.c | 94 ++++++++++++++++++ drivers/gpu/drm/drm_plane_helper.c | 102 +------------------- drivers/gpu/drm/drm_simple_kms_helper.c | 9 +- drivers/gpu/drm/i915/intel_display.c | 22 ++--- drivers/gpu/drm/imx/ipuv3-plane.c | 8 +- drivers/gpu/drm/mediatek/mtk_drm_plane.c | 8 +- drivers/gpu/drm/meson/meson_plane.c | 8 +- drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c | 5 +- drivers/gpu/drm/nouveau/nv50_display.c | 20 ++-- drivers/gpu/drm/rockchip/rockchip_drm_vop.c | 6 +- drivers/gpu/drm/tegra/dc.c | 4 +- drivers/gpu/drm/vmwgfx/vmwgfx_kms.c | 8 +- drivers/gpu/drm/zte/zx_plane.c | 15 +-- include/drm/drm_atomic_helper.h | 7 ++ include/drm/drm_plane_helper.h | 6 -- 17 files changed, 169 insertions(+), 165 deletions(-) diff --git a/drivers/gpu/drm/arm/hdlcd_crtc.c b/drivers/gpu/drm/arm/hdlcd_crtc.c index 14721723fa8ac..63511a3bbf6c0 100644 --- a/drivers/gpu/drm/arm/hdlcd_crtc.c +++ b/drivers/gpu/drm/arm/hdlcd_crtc.c @@ -252,10 +252,10 @@ static int hdlcd_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + return drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); } static void hdlcd_plane_atomic_update(struct drm_plane *plane, diff --git a/drivers/gpu/drm/arm/malidp_planes.c b/drivers/gpu/drm/arm/malidp_planes.c index 492d99dd55d46..72a07950167e8 100644 --- a/drivers/gpu/drm/arm/malidp_planes.c +++ b/drivers/gpu/drm/arm/malidp_planes.c @@ -150,8 +150,8 @@ static int malidp_se_check_scaling(struct malidp_plane *mp, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, crtc_state, &clip, - 0, INT_MAX, true, true); + ret = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + 0, INT_MAX, true, true); if (ret) return ret; diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c index 07d4058d97d0d..ce6d11ae9e265 100644 --- a/drivers/gpu/drm/drm_atomic_helper.c +++ b/drivers/gpu/drm/drm_atomic_helper.c @@ -695,6 +695,100 @@ drm_atomic_helper_check_modeset(struct drm_device *dev, } EXPORT_SYMBOL(drm_atomic_helper_check_modeset); +/** + * drm_atomic_helper_check_plane_state() - Check plane state for validity + * @plane_state: plane state to check + * @crtc_state: crtc state to check + * @clip: integer clipping coordinates + * @min_scale: minimum @src:@dest scaling factor in 16.16 fixed point + * @max_scale: maximum @src:@dest scaling factor in 16.16 fixed point + * @can_position: is it legal to position the plane such that it + * doesn't cover the entire crtc? This will generally + * only be false for primary planes. + * @can_update_disabled: can the plane be updated while the crtc + * is disabled? + * + * Checks that a desired plane update is valid, and updates various + * bits of derived state (clipped coordinates etc.). Drivers that provide + * their own plane handling rather than helper-provided implementations may + * still wish to call this function to avoid duplication of error checking + * code. + * + * RETURNS: + * Zero if update appears valid, error code on failure + */ +int drm_atomic_helper_check_plane_state(struct drm_plane_state *plane_state, + const struct drm_crtc_state *crtc_state, + const struct drm_rect *clip, + int min_scale, + int max_scale, + bool can_position, + bool can_update_disabled) +{ + struct drm_framebuffer *fb = plane_state->fb; + struct drm_rect *src = &plane_state->src; + struct drm_rect *dst = &plane_state->dst; + unsigned int rotation = plane_state->rotation; + int hscale, vscale; + + WARN_ON(plane_state->crtc && plane_state->crtc != crtc_state->crtc); + + *src = drm_plane_state_src(plane_state); + *dst = drm_plane_state_dest(plane_state); + + if (!fb) { + plane_state->visible = false; + return 0; + } + + /* crtc should only be NULL when disabling (i.e., !fb) */ + if (WARN_ON(!plane_state->crtc)) { + plane_state->visible = false; + return 0; + } + + if (!crtc_state->enable && !can_update_disabled) { + DRM_DEBUG_KMS("Cannot update plane of a disabled CRTC.\n"); + return -EINVAL; + } + + drm_rect_rotate(src, fb->width << 16, fb->height << 16, rotation); + + /* Check scaling */ + hscale = drm_rect_calc_hscale(src, dst, min_scale, max_scale); + vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale); + if (hscale < 0 || vscale < 0) { + DRM_DEBUG_KMS("Invalid scaling of plane\n"); + drm_rect_debug_print("src: ", &plane_state->src, true); + drm_rect_debug_print("dst: ", &plane_state->dst, false); + return -ERANGE; + } + + plane_state->visible = drm_rect_clip_scaled(src, dst, clip, hscale, vscale); + + drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, rotation); + + if (!plane_state->visible) + /* + * Plane isn't visible; some drivers can handle this + * so we just return success here. Drivers that can't + * (including those that use the primary plane helper's + * update function) will return an error from their + * update_plane handler. + */ + return 0; + + if (!can_position && !drm_rect_equals(dst, clip)) { + DRM_DEBUG_KMS("Plane must cover entire CRTC\n"); + drm_rect_debug_print("dst: ", dst, false); + drm_rect_debug_print("clip: ", clip, false); + return -EINVAL; + } + + return 0; +} +EXPORT_SYMBOL(drm_atomic_helper_check_plane_state); + /** * drm_atomic_helper_check_planes - validate state object for planes changes * @dev: DRM device diff --git a/drivers/gpu/drm/drm_plane_helper.c b/drivers/gpu/drm/drm_plane_helper.c index eef0ffcd7ed52..f1be8cd4e387f 100644 --- a/drivers/gpu/drm/drm_plane_helper.c +++ b/drivers/gpu/drm/drm_plane_helper.c @@ -99,100 +99,6 @@ static int get_connectors_for_crtc(struct drm_crtc *crtc, return count; } -/** - * drm_plane_helper_check_state() - Check plane state for validity - * @plane_state: plane state to check - * @crtc_state: crtc state to check - * @clip: integer clipping coordinates - * @min_scale: minimum @src:@dest scaling factor in 16.16 fixed point - * @max_scale: maximum @src:@dest scaling factor in 16.16 fixed point - * @can_position: is it legal to position the plane such that it - * doesn't cover the entire crtc? This will generally - * only be false for primary planes. - * @can_update_disabled: can the plane be updated while the crtc - * is disabled? - * - * Checks that a desired plane update is valid, and updates various - * bits of derived state (clipped coordinates etc.). Drivers that provide - * their own plane handling rather than helper-provided implementations may - * still wish to call this function to avoid duplication of error checking - * code. - * - * RETURNS: - * Zero if update appears valid, error code on failure - */ -int drm_plane_helper_check_state(struct drm_plane_state *plane_state, - const struct drm_crtc_state *crtc_state, - const struct drm_rect *clip, - int min_scale, - int max_scale, - bool can_position, - bool can_update_disabled) -{ - struct drm_framebuffer *fb = plane_state->fb; - struct drm_rect *src = &plane_state->src; - struct drm_rect *dst = &plane_state->dst; - unsigned int rotation = plane_state->rotation; - int hscale, vscale; - - WARN_ON(plane_state->crtc && plane_state->crtc != crtc_state->crtc); - - *src = drm_plane_state_src(plane_state); - *dst = drm_plane_state_dest(plane_state); - - if (!fb) { - plane_state->visible = false; - return 0; - } - - /* crtc should only be NULL when disabling (i.e., !fb) */ - if (WARN_ON(!plane_state->crtc)) { - plane_state->visible = false; - return 0; - } - - if (!crtc_state->enable && !can_update_disabled) { - DRM_DEBUG_KMS("Cannot update plane of a disabled CRTC.\n"); - return -EINVAL; - } - - drm_rect_rotate(src, fb->width << 16, fb->height << 16, rotation); - - /* Check scaling */ - hscale = drm_rect_calc_hscale(src, dst, min_scale, max_scale); - vscale = drm_rect_calc_vscale(src, dst, min_scale, max_scale); - if (hscale < 0 || vscale < 0) { - DRM_DEBUG_KMS("Invalid scaling of plane\n"); - drm_rect_debug_print("src: ", &plane_state->src, true); - drm_rect_debug_print("dst: ", &plane_state->dst, false); - return -ERANGE; - } - - plane_state->visible = drm_rect_clip_scaled(src, dst, clip, hscale, vscale); - - drm_rect_rotate_inv(src, fb->width << 16, fb->height << 16, rotation); - - if (!plane_state->visible) - /* - * Plane isn't visible; some drivers can handle this - * so we just return success here. Drivers that can't - * (including those that use the primary plane helper's - * update function) will return an error from their - * update_plane handler. - */ - return 0; - - if (!can_position && !drm_rect_equals(dst, clip)) { - DRM_DEBUG_KMS("Plane must cover entire CRTC\n"); - drm_rect_debug_print("dst: ", dst, false); - drm_rect_debug_print("clip: ", clip, false); - return -EINVAL; - } - - return 0; -} -EXPORT_SYMBOL(drm_plane_helper_check_state); - /** * drm_plane_helper_check_update() - Check plane update for validity * @plane: plane object to update @@ -254,10 +160,10 @@ int drm_plane_helper_check_update(struct drm_plane *plane, }; int ret; - ret = drm_plane_helper_check_state(&plane_state, &crtc_state, clip, - min_scale, max_scale, - can_position, - can_update_disabled); + ret = drm_atomic_helper_check_plane_state(&plane_state, &crtc_state, + clip, min_scale, max_scale, + can_position, + can_update_disabled); if (ret) return ret; diff --git a/drivers/gpu/drm/drm_simple_kms_helper.c b/drivers/gpu/drm/drm_simple_kms_helper.c index d428c805025c9..9f3b1c94802b3 100644 --- a/drivers/gpu/drm/drm_simple_kms_helper.c +++ b/drivers/gpu/drm/drm_simple_kms_helper.c @@ -103,10 +103,11 @@ static int drm_simple_kms_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(plane_state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + ret = drm_atomic_helper_check_plane_state(plane_state, crtc_state, + &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); if (ret) return ret; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5e75d762745aa..97406166417dc 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -9418,12 +9418,12 @@ static int intel_check_cursor(struct intel_crtc_state *crtc_state, u32 offset; int ret; - ret = drm_plane_helper_check_state(&plane_state->base, - &crtc_state->base, - &plane_state->clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - true, true); + ret = drm_atomic_helper_check_plane_state(&plane_state->base, + &crtc_state->base, + &plane_state->clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + true, true); if (ret) return ret; @@ -12842,11 +12842,11 @@ intel_check_primary_plane(struct intel_plane *plane, can_position = true; } - ret = drm_plane_helper_check_state(&state->base, - &crtc_state->base, - &state->clip, - min_scale, max_scale, - can_position, true); + ret = drm_atomic_helper_check_plane_state(&state->base, + &crtc_state->base, + &state->clip, + min_scale, max_scale, + can_position, true); if (ret) return ret; diff --git a/drivers/gpu/drm/imx/ipuv3-plane.c b/drivers/gpu/drm/imx/ipuv3-plane.c index 51b23ac175e39..5a67daedcf4d7 100644 --- a/drivers/gpu/drm/imx/ipuv3-plane.c +++ b/drivers/gpu/drm/imx/ipuv3-plane.c @@ -342,10 +342,10 @@ static int ipu_plane_atomic_check(struct drm_plane *plane, clip.y1 = 0; clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - can_position, true); + ret = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + can_position, true); if (ret) return ret; diff --git a/drivers/gpu/drm/mediatek/mtk_drm_plane.c b/drivers/gpu/drm/mediatek/mtk_drm_plane.c index 7ebb336577047..5ef898b93d8d0 100644 --- a/drivers/gpu/drm/mediatek/mtk_drm_plane.c +++ b/drivers/gpu/drm/mediatek/mtk_drm_plane.c @@ -111,10 +111,10 @@ static int mtk_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->mode.hdisplay; clip.y2 = crtc_state->mode.vdisplay; - return drm_plane_helper_check_state(state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - true, true); + return drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + true, true); } static void mtk_plane_atomic_update(struct drm_plane *plane, diff --git a/drivers/gpu/drm/meson/meson_plane.c b/drivers/gpu/drm/meson/meson_plane.c index 2a00b720c3718..d0a6ac8390f39 100644 --- a/drivers/gpu/drm/meson/meson_plane.c +++ b/drivers/gpu/drm/meson/meson_plane.c @@ -61,10 +61,10 @@ static int meson_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->mode.hdisplay; clip.y2 = crtc_state->mode.vdisplay; - return drm_plane_helper_check_state(state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - true, true); + return drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + true, true); } /* Takes a fixed 16.16 number and converts it to integer. */ diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c index 1c194a9453d97..df3360acaccac 100644 --- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c +++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_plane.c @@ -348,8 +348,9 @@ static int mdp5_plane_atomic_check_with_state(struct drm_crtc_state *crtc_state, min_scale = FRAC_16_16(1, 8); max_scale = FRAC_16_16(8, 1); - ret = drm_plane_helper_check_state(state, crtc_state, &clip, - min_scale, max_scale, true, true); + ret = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + min_scale, max_scale, + true, true); if (ret) return ret; diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 0738ae1f6cf4e..9fc8fe0171184 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -1143,11 +1143,11 @@ nv50_curs_acquire(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw, { int ret; - ret = drm_plane_helper_check_state(&asyw->state, &asyh->state, - &asyw->clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - true, true); + ret = drm_atomic_helper_check_plane_state(&asyw->state, &asyh->state, + &asyw->clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + true, true); asyh->curs.visible = asyw->state.visible; if (ret || !asyh->curs.visible) return ret; @@ -1433,11 +1433,11 @@ nv50_base_acquire(struct nv50_wndw *wndw, struct nv50_wndw_atom *asyw, if (!fb->format->depth) return -EINVAL; - ret = drm_plane_helper_check_state(&asyw->state, &asyh->state, - &asyw->clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + ret = drm_atomic_helper_check_plane_state(&asyw->state, &asyh->state, + &asyw->clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); if (ret) return ret; diff --git a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c index 36d0f101e30d0..ba7505292b786 100644 --- a/drivers/gpu/drm/rockchip/rockchip_drm_vop.c +++ b/drivers/gpu/drm/rockchip/rockchip_drm_vop.c @@ -659,9 +659,9 @@ static int vop_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - ret = drm_plane_helper_check_state(state, crtc_state, &clip, - min_scale, max_scale, - true, true); + ret = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + min_scale, max_scale, + true, true); if (ret) return ret; diff --git a/drivers/gpu/drm/tegra/dc.c b/drivers/gpu/drm/tegra/dc.c index 08b899fd74fc4..91919d52b2075 100644 --- a/drivers/gpu/drm/tegra/dc.c +++ b/drivers/gpu/drm/tegra/dc.c @@ -500,8 +500,8 @@ static int tegra_plane_state_add(struct tegra_plane *plane, clip.y2 = crtc_state->mode.vdisplay; /* Check plane state for visibility and calculate clipping bounds */ - err = drm_plane_helper_check_state(state, crtc_state, &clip, - 0, INT_MAX, true, true); + err = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + 0, INT_MAX, true, true); if (err < 0) return err; diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 91555c0370e1d..13ac2c5a04ee1 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -454,10 +454,10 @@ int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, clip.y2 = crtc_state->adjusted_mode.vdisplay; } - ret = drm_plane_helper_check_state(state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + ret = drm_atomic_helper_check_plane_state(state, crtc_state, &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); if (!ret && new_fb) { struct drm_crtc *crtc = state->crtc; diff --git a/drivers/gpu/drm/zte/zx_plane.c b/drivers/gpu/drm/zte/zx_plane.c index ee0002529b8f8..68fd2e2dc78ae 100644 --- a/drivers/gpu/drm/zte/zx_plane.c +++ b/drivers/gpu/drm/zte/zx_plane.c @@ -80,9 +80,9 @@ static int zx_vl_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(plane_state, crtc_state, &clip, - min_scale, max_scale, - true, true); + return drm_atomic_helper_check_plane_state(plane_state, crtc_state, + &clip, min_scale, max_scale, + true, true); } static int zx_vl_get_fmt(uint32_t format) @@ -315,10 +315,11 @@ static int zx_gl_plane_atomic_check(struct drm_plane *plane, clip.x2 = crtc_state->adjusted_mode.hdisplay; clip.y2 = crtc_state->adjusted_mode.vdisplay; - return drm_plane_helper_check_state(plane_state, crtc_state, &clip, - DRM_PLANE_HELPER_NO_SCALING, - DRM_PLANE_HELPER_NO_SCALING, - false, true); + return drm_atomic_helper_check_plane_state(plane_state, crtc_state, + &clip, + DRM_PLANE_HELPER_NO_SCALING, + DRM_PLANE_HELPER_NO_SCALING, + false, true); } static int zx_gl_get_fmt(uint32_t format) diff --git a/include/drm/drm_atomic_helper.h b/include/drm/drm_atomic_helper.h index d2b56cc657e93..4842ee9485cec 100644 --- a/include/drm/drm_atomic_helper.h +++ b/include/drm/drm_atomic_helper.h @@ -38,6 +38,13 @@ struct drm_private_state; int drm_atomic_helper_check_modeset(struct drm_device *dev, struct drm_atomic_state *state); +int drm_atomic_helper_check_plane_state(struct drm_plane_state *plane_state, + const struct drm_crtc_state *crtc_state, + const struct drm_rect *clip, + int min_scale, + int max_scale, + bool can_position, + bool can_update_disabled); int drm_atomic_helper_check_planes(struct drm_device *dev, struct drm_atomic_state *state); int drm_atomic_helper_check(struct drm_device *dev, diff --git a/include/drm/drm_plane_helper.h b/include/drm/drm_plane_helper.h index 41b8309b0a57b..8aa49c0ecd4d6 100644 --- a/include/drm/drm_plane_helper.h +++ b/include/drm/drm_plane_helper.h @@ -38,12 +38,6 @@ */ #define DRM_PLANE_HELPER_NO_SCALING (1<<16) -int drm_plane_helper_check_state(struct drm_plane_state *plane_state, - const struct drm_crtc_state *crtc_state, - const struct drm_rect *clip, - int min_scale, int max_scale, - bool can_position, - bool can_update_disabled); int drm_plane_helper_check_update(struct drm_plane *plane, struct drm_crtc *crtc, struct drm_framebuffer *fb, -- GitLab From fb5c551ad510e4a408c105670f89c725ebbfe6c8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 20:55:00 +0000 Subject: [PATCH 0365/1471] drm/i915: Remove i915.enable_execlists module parameter Execlists and legacy ringbuffer submission are no longer feature comparable (execlists now offer greater functionality that should overcome their performance hit) and obsoletes the unsafe module parameter, i.e. comparing the two modes of execution is no longer useful, so remove the debug tool. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Reviewed-by: Lionel Landwerlin #i915_perf.c Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gvt/render.c | 3 +- drivers/gpu/drm/i915/i915_debugfs.c | 70 ------------------------- drivers/gpu/drm/i915/i915_drv.c | 8 +-- drivers/gpu/drm/i915/i915_drv.h | 3 ++ drivers/gpu/drm/i915/i915_gem.c | 10 ++-- drivers/gpu/drm/i915/i915_gem_context.c | 10 +--- drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +- drivers/gpu/drm/i915/i915_params.c | 4 -- drivers/gpu/drm/i915/i915_params.h | 1 - drivers/gpu/drm/i915/i915_perf.c | 8 +-- drivers/gpu/drm/i915/intel_engine_cs.c | 8 ++- drivers/gpu/drm/i915/intel_gvt.c | 5 -- drivers/gpu/drm/i915/intel_lrc.c | 31 ----------- drivers/gpu/drm/i915/intel_lrc.h | 4 -- 14 files changed, 20 insertions(+), 149 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/render.c b/drivers/gpu/drm/i915/gvt/render.c index 0672178548ef9..dac12c25f349d 100644 --- a/drivers/gpu/drm/i915/gvt/render.c +++ b/drivers/gpu/drm/i915/gvt/render.c @@ -294,8 +294,7 @@ static void switch_mmio_to_vgpu(struct intel_vgpu *vgpu, int ring_id) * write. */ if (mmio->in_context && - ((ctx_ctrl & inhibit_mask) != inhibit_mask) && - i915_modparams.enable_execlists) + (ctx_ctrl & inhibit_mask) != inhibit_mask) continue; if (mmio->mask) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index df3852c02a353..5e2a6e18771f2 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1989,75 +1989,6 @@ static int i915_context_status(struct seq_file *m, void *unused) return 0; } -static void i915_dump_lrc_obj(struct seq_file *m, - struct i915_gem_context *ctx, - struct intel_engine_cs *engine) -{ - struct i915_vma *vma = ctx->engine[engine->id].state; - struct page *page; - int j; - - seq_printf(m, "CONTEXT: %s %u\n", engine->name, ctx->hw_id); - - if (!vma) { - seq_puts(m, "\tFake context\n"); - return; - } - - if (vma->flags & I915_VMA_GLOBAL_BIND) - seq_printf(m, "\tBound in GGTT at 0x%08x\n", - i915_ggtt_offset(vma)); - - if (i915_gem_object_pin_pages(vma->obj)) { - seq_puts(m, "\tFailed to get pages for context object\n\n"); - return; - } - - page = i915_gem_object_get_page(vma->obj, LRC_STATE_PN); - if (page) { - u32 *reg_state = kmap_atomic(page); - - for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) { - seq_printf(m, - "\t[0x%04x] 0x%08x 0x%08x 0x%08x 0x%08x\n", - j * 4, - reg_state[j], reg_state[j + 1], - reg_state[j + 2], reg_state[j + 3]); - } - kunmap_atomic(reg_state); - } - - i915_gem_object_unpin_pages(vma->obj); - seq_putc(m, '\n'); -} - -static int i915_dump_lrc(struct seq_file *m, void *unused) -{ - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct drm_device *dev = &dev_priv->drm; - struct intel_engine_cs *engine; - struct i915_gem_context *ctx; - enum intel_engine_id id; - int ret; - - if (!i915_modparams.enable_execlists) { - seq_printf(m, "Logical Ring Contexts are disabled\n"); - return 0; - } - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - return ret; - - list_for_each_entry(ctx, &dev_priv->contexts.list, link) - for_each_engine(engine, dev_priv, id) - i915_dump_lrc_obj(m, ctx, engine); - - mutex_unlock(&dev->struct_mutex); - - return 0; -} - static const char *swizzle_string(unsigned swizzle) { switch (swizzle) { @@ -4833,7 +4764,6 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_vbt", i915_vbt, 0}, {"i915_gem_framebuffer", i915_gem_framebuffer_info, 0}, {"i915_context_status", i915_context_status, 0}, - {"i915_dump_lrc", i915_dump_lrc, 0}, {"i915_forcewake_domains", i915_forcewake_domains, 0}, {"i915_swizzle_info", i915_swizzle_info, 0}, {"i915_ppgtt_info", i915_ppgtt_info, 0}, diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 2b4ced93011b7..c608fd96771cc 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -371,9 +371,7 @@ static int i915_getparam(struct drm_device *dev, void *data, if (dev_priv->engine[RCS] && dev_priv->engine[RCS]->schedule) { value |= I915_SCHEDULER_CAP_ENABLED; value |= I915_SCHEDULER_CAP_PRIORITY; - - if (HAS_LOGICAL_RING_PREEMPTION(dev_priv) && - i915_modparams.enable_execlists) + if (HAS_LOGICAL_RING_PREEMPTION(dev_priv)) value |= I915_SCHEDULER_CAP_PREEMPTION; } break; @@ -1054,10 +1052,6 @@ static void i915_driver_cleanup_mmio(struct drm_i915_private *dev_priv) static void intel_sanitize_options(struct drm_i915_private *dev_priv) { - i915_modparams.enable_execlists = - intel_sanitize_enable_execlists(dev_priv, - i915_modparams.enable_execlists); - /* * i915.enable_ppgtt is read-only, so do an early pass to validate the * user's requested state against the hardware/driver capabilities. We diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2b766253f9706..d2c6ba7f8acd1 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3153,6 +3153,9 @@ intel_info(const struct drm_i915_private *dev_priv) ((dev_priv)->info.has_logical_ring_contexts) #define HAS_LOGICAL_RING_PREEMPTION(dev_priv) \ ((dev_priv)->info.has_logical_ring_preemption) + +#define HAS_EXECLISTS(dev_priv) HAS_LOGICAL_RING_CONTEXTS(dev_priv) + #define USES_PPGTT(dev_priv) (i915_modparams.enable_ppgtt) #define USES_FULL_PPGTT(dev_priv) (i915_modparams.enable_ppgtt >= 2) #define USES_FULL_48BIT_PPGTT(dev_priv) (i915_modparams.enable_ppgtt == 3) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e07eb0beef13b..494c6be035b27 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5003,7 +5003,7 @@ bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value) return false; /* TODO: make semaphores and Execlists play nicely together */ - if (i915_modparams.enable_execlists) + if (HAS_EXECLISTS(dev_priv)) return false; if (value >= 0) @@ -5147,12 +5147,12 @@ int i915_gem_init(struct drm_i915_private *dev_priv) dev_priv->mm.unordered_timeline = dma_fence_context_alloc(1); - if (!i915_modparams.enable_execlists) { - dev_priv->gt.resume = intel_legacy_submission_resume; - dev_priv->gt.cleanup_engine = intel_engine_cleanup; - } else { + if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { dev_priv->gt.resume = intel_lr_context_resume; dev_priv->gt.cleanup_engine = intel_logical_ring_cleanup; + } else { + dev_priv->gt.resume = intel_legacy_submission_resume; + dev_priv->gt.cleanup_engine = intel_engine_cleanup; } /* This is just a security blanket to placate dragons. diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index c1efbaf02bf2a..c75941d3d7e71 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -460,14 +460,6 @@ int i915_gem_contexts_init(struct drm_i915_private *dev_priv) INIT_WORK(&dev_priv->contexts.free_work, contexts_free_worker); init_llist_head(&dev_priv->contexts.free_list); - if (intel_vgpu_active(dev_priv) && - HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { - if (!i915_modparams.enable_execlists) { - DRM_INFO("Only EXECLIST mode is supported in vgpu.\n"); - return -EINVAL; - } - } - /* Using the simple ida interface, the max is limited by sizeof(int) */ BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX); ida_init(&dev_priv->contexts.hw_ida); @@ -842,7 +834,7 @@ int i915_switch_context(struct drm_i915_gem_request *req) struct intel_engine_cs *engine = req->engine; lockdep_assert_held(&req->i915->drm.struct_mutex); - GEM_BUG_ON(i915_modparams.enable_execlists); + GEM_BUG_ON(HAS_EXECLISTS(req->i915)); if (!req->ctx->engine[engine->id].state) { struct i915_gem_context *to = req->ctx; diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index f92a39fc511cf..e101b9a989573 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -178,7 +178,7 @@ int intel_sanitize_enable_ppgtt(struct drm_i915_private *dev_priv, return 0; } - if (INTEL_GEN(dev_priv) >= 8 && i915_modparams.enable_execlists) { + if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { if (has_full_48bit_ppgtt) return 3; @@ -2162,7 +2162,7 @@ int i915_ppgtt_init_hw(struct drm_i915_private *dev_priv) /* In the case of execlists, PPGTT is enabled by the context descriptor * and the PDPs are contained within the context itself. We don't * need to do anything here. */ - if (i915_modparams.enable_execlists) + if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) return 0; if (!USES_PPGTT(dev_priv)) diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c index b4faeb6aa2bdc..d61c1787c1641 100644 --- a/drivers/gpu/drm/i915/i915_params.c +++ b/drivers/gpu/drm/i915/i915_params.c @@ -99,10 +99,6 @@ i915_param_named_unsafe(enable_ppgtt, int, 0400, "Override PPGTT usage. " "(-1=auto [default], 0=disabled, 1=aliasing, 2=full, 3=full with extended address space)"); -i915_param_named_unsafe(enable_execlists, int, 0400, - "Override execlists usage. " - "(-1=auto [default], 0=disabled, 1=enabled)"); - i915_param_named_unsafe(enable_psr, int, 0600, "Enable PSR " "(0=disabled, 1=enabled - link mode chosen per-platform, 2=force link-standby mode, 3=force link-off mode) " diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h index c7292268ed439..0aef3d7178ebf 100644 --- a/drivers/gpu/drm/i915/i915_params.h +++ b/drivers/gpu/drm/i915/i915_params.h @@ -39,7 +39,6 @@ param(int, enable_dc, -1) \ param(int, enable_fbc, -1) \ param(int, enable_ppgtt, -1) \ - param(int, enable_execlists, -1) \ param(int, enable_psr, -1) \ param(int, disable_power_well, -1) \ param(int, enable_ips, 1) \ diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index adbce839a918b..f0cfdece14ae7 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -1216,9 +1216,9 @@ static int oa_get_render_ctx_id(struct i915_perf_stream *stream) { struct drm_i915_private *dev_priv = stream->dev_priv; - if (i915_modparams.enable_execlists) + if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { dev_priv->perf.oa.specific_ctx_id = stream->ctx->hw_id; - else { + } else { struct intel_engine_cs *engine = dev_priv->engine[RCS]; struct intel_ring *ring; int ret; @@ -1262,7 +1262,7 @@ static void oa_put_render_ctx_id(struct i915_perf_stream *stream) { struct drm_i915_private *dev_priv = stream->dev_priv; - if (i915_modparams.enable_execlists) { + if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { dev_priv->perf.oa.specific_ctx_id = INVALID_CTX_ID; } else { struct intel_engine_cs *engine = dev_priv->engine[RCS]; @@ -3439,7 +3439,7 @@ void i915_perf_init(struct drm_i915_private *dev_priv) gen7_oa_hw_tail_read; dev_priv->perf.oa.oa_formats = hsw_oa_formats; - } else if (i915_modparams.enable_execlists) { + } else if (HAS_LOGICAL_RING_CONTEXTS(dev_priv)) { /* Note: that although we could theoretically also support the * legacy ringbuffer mode on BDW (and earlier iterations of * this driver, before upstreaming did this) it didn't seem diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 9897c7f78c515..cae76f47fec2f 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -164,9 +164,7 @@ __intel_engine_context_size(struct drm_i915_private *dev_priv, u8 class) case 9: return GEN9_LR_CONTEXT_RENDER_SIZE; case 8: - return i915_modparams.enable_execlists ? - GEN8_LR_CONTEXT_RENDER_SIZE : - GEN8_CXT_TOTAL_SIZE; + return GEN8_LR_CONTEXT_RENDER_SIZE; case 7: if (IS_HASWELL(dev_priv)) return HSW_CXT_TOTAL_SIZE; @@ -316,7 +314,7 @@ int intel_engines_init(struct drm_i915_private *dev_priv) &intel_engine_classes[engine->class]; int (*init)(struct intel_engine_cs *engine); - if (i915_modparams.enable_execlists) + if (HAS_EXECLISTS(dev_priv)) init = class_info->init_execlists; else init = class_info->init_legacy; @@ -1739,7 +1737,7 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) drm_printf(m, "\tBBADDR: 0x%08x_%08x\n", upper_32_bits(addr), lower_32_bits(addr)); - if (i915_modparams.enable_execlists) { + if (HAS_EXECLISTS(dev_priv)) { const u32 *hws = &engine->status_page.page_addr[I915_HWS_CSB_BUF0_INDEX]; u32 ptr, read, write; unsigned int idx; diff --git a/drivers/gpu/drm/i915/intel_gvt.c b/drivers/gpu/drm/i915/intel_gvt.c index b4a7f31f02149..126f7c769c694 100644 --- a/drivers/gpu/drm/i915/intel_gvt.c +++ b/drivers/gpu/drm/i915/intel_gvt.c @@ -95,11 +95,6 @@ int intel_gvt_init(struct drm_i915_private *dev_priv) return 0; } - if (!i915_modparams.enable_execlists) { - DRM_ERROR("i915 GVT-g loading failed due to disabled execlists mode\n"); - return -EIO; - } - if (i915_modparams.enable_guc_submission) { DRM_ERROR("i915 GVT-g loading failed due to Graphics virtualization is not yet supported with GuC submission\n"); return -EIO; diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 8d0c493888634..16d4936e2368e 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -218,37 +218,6 @@ static void execlists_init_reg_state(u32 *reg_state, struct intel_engine_cs *engine, struct intel_ring *ring); -/** - * intel_sanitize_enable_execlists() - sanitize i915.enable_execlists - * @dev_priv: i915 device private - * @enable_execlists: value of i915.enable_execlists module parameter. - * - * Only certain platforms support Execlists (the prerequisites being - * support for Logical Ring Contexts and Aliasing PPGTT or better). - * - * Return: 1 if Execlists is supported and has to be enabled. - */ -int intel_sanitize_enable_execlists(struct drm_i915_private *dev_priv, int enable_execlists) -{ - /* On platforms with execlist available, vGPU will only - * support execlist mode, no ring buffer mode. - */ - if (HAS_LOGICAL_RING_CONTEXTS(dev_priv) && intel_vgpu_active(dev_priv)) - return 1; - - if (INTEL_GEN(dev_priv) >= 9) - return 1; - - if (enable_execlists == 0) - return 0; - - if (HAS_LOGICAL_RING_CONTEXTS(dev_priv) && - USES_PPGTT(dev_priv)) - return 1; - - return 0; -} - /** * intel_lr_context_descriptor_update() - calculate & cache the descriptor * descriptor for a pinned context diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h index 17182ce296740..6d4f9b995a110 100644 --- a/drivers/gpu/drm/i915/intel_lrc.h +++ b/drivers/gpu/drm/i915/intel_lrc.h @@ -107,8 +107,4 @@ intel_lr_context_descriptor(struct i915_gem_context *ctx, return ctx->engine[engine->id].lrc_desc; } -/* Execlists */ -int intel_sanitize_enable_execlists(struct drm_i915_private *dev_priv, - int enable_execlists); - #endif /* _INTEL_LRC_H_ */ -- GitLab From 79e6770cb1f5e32eb49590edbb794a97d0113aed Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 20:55:01 +0000 Subject: [PATCH 0366/1471] drm/i915: Remove obsolete ringbuffer emission for gen8+ Since removing the module parameter to force selection of ringbuffer emission for gen8, the code is defunct. Remove it. To put the difference into perspective, a couple of microbenchmarks (bdw i7-5557u, 20170324): ring execlists exec continuous nops on all rings: 1.491us 2.223us exec sequential nops on each ring: 12.508us 53.682us single nop + sync: 9.272us 30.291us vblank_mode=0 glxgears: ~11000fps ~9000fps Since the earlier submission, gen8 ringbuffer submission has fallen further and further behind in features. So while ringbuffer may hold the throughput crown, in terms of interactive latency, execlists is much better. Alas, we have no convenient metrics for such, other than demonstrating things we can do with execlists but can not using legacy ringbuffer submission. We have made a few improvements to lowlevel execlists throughput, and ringbuffer currently panics on boot! (bdw i7-5557u, 20171026): ring execlists exec continuous nops on all rings: n/a 1.921us exec sequential nops on each ring: n/a 44.621us single nop + sync: n/a 21.953us vblank_mode=0 glxgears: n/a ~18500fps References: https://bugs.freedesktop.org/show_bug.cgi?id=87725 Signed-off-by: Chris Wilson Once-upon-a-time-Reviewed-by: Joonas Lahtinen Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_debugfs.c | 44 +-- drivers/gpu/drm/i915/i915_drv.h | 2 - drivers/gpu/drm/i915/i915_gem.c | 2 +- drivers/gpu/drm/i915/i915_gem_context.c | 47 +-- drivers/gpu/drm/i915/i915_gpu_error.c | 36 -- drivers/gpu/drm/i915/intel_engine_cs.c | 14 - drivers/gpu/drm/i915/intel_hangcheck.c | 44 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 431 ++++-------------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 25 +- 9 files changed, 94 insertions(+), 551 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 5e2a6e18771f2..9cef1463d4113 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3254,44 +3254,12 @@ static int i915_semaphore_status(struct seq_file *m, void *unused) return ret; intel_runtime_pm_get(dev_priv); - if (IS_BROADWELL(dev_priv)) { - struct page *page; - uint64_t *seqno; - - page = i915_gem_object_get_page(dev_priv->semaphore->obj, 0); - - seqno = (uint64_t *)kmap_atomic(page); - for_each_engine(engine, dev_priv, id) { - uint64_t offset; - - seq_printf(m, "%s\n", engine->name); - - seq_puts(m, " Last signal:"); - for (j = 0; j < num_rings; j++) { - offset = id * I915_NUM_ENGINES + j; - seq_printf(m, "0x%08llx (0x%02llx) ", - seqno[offset], offset * 8); - } - seq_putc(m, '\n'); - - seq_puts(m, " Last wait: "); - for (j = 0; j < num_rings; j++) { - offset = id + (j * I915_NUM_ENGINES); - seq_printf(m, "0x%08llx (0x%02llx) ", - seqno[offset], offset * 8); - } - seq_putc(m, '\n'); - - } - kunmap_atomic(seqno); - } else { - seq_puts(m, " Last signal:"); - for_each_engine(engine, dev_priv, id) - for (j = 0; j < num_rings; j++) - seq_printf(m, "0x%08x\n", - I915_READ(engine->semaphore.mbox.signal[j])); - seq_putc(m, '\n'); - } + seq_puts(m, " Last signal:"); + for_each_engine(engine, dev_priv, id) + for (j = 0; j < num_rings; j++) + seq_printf(m, "0x%08x\n", + I915_READ(engine->semaphore.mbox.signal[j])); + seq_putc(m, '\n'); intel_runtime_pm_put(dev_priv); mutex_unlock(&dev->struct_mutex); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d2c6ba7f8acd1..5ce778c26bb88 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -942,7 +942,6 @@ struct i915_gpu_state { u64 fence[I915_MAX_NUM_FENCES]; struct intel_overlay_error_state *overlay; struct intel_display_error_state *display; - struct drm_i915_error_object *semaphore; struct drm_i915_error_engine { int engine_id; @@ -2291,7 +2290,6 @@ struct drm_i915_private { struct i915_gem_context *kernel_context; /* Context only to be used for injecting preemption commands */ struct i915_gem_context *preempt_context; - struct i915_vma *semaphore; struct drm_dma_handle *status_page_dmah; struct resource mch_res; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 494c6be035b27..d470e10b2edd0 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4999,7 +4999,7 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value) { - if (INTEL_INFO(dev_priv)->gen < 6) + if (INTEL_GEN(dev_priv) < 6) return false; /* TODO: make semaphores and Execlists play nicely together */ diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index c75941d3d7e71..0704d9af261bc 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -574,21 +574,21 @@ mi_set_context(struct drm_i915_gem_request *req, u32 flags) enum intel_engine_id id; const int num_rings = /* Use an extended w/a on gen7 if signalling from other rings */ - (i915_modparams.semaphores && INTEL_GEN(dev_priv) == 7) ? + (i915_modparams.semaphores && IS_GEN7(dev_priv)) ? INTEL_INFO(dev_priv)->num_rings - 1 : 0; int len; u32 *cs; flags |= MI_MM_SPACE_GTT; - if (IS_HASWELL(dev_priv) || INTEL_GEN(dev_priv) >= 8) + if (IS_HASWELL(dev_priv)) /* These flags are for resource streamer on HSW+ */ flags |= HSW_MI_RS_SAVE_STATE_EN | HSW_MI_RS_RESTORE_STATE_EN; else flags |= MI_SAVE_EXT_STATE_EN | MI_RESTORE_EXT_STATE_EN; len = 4; - if (INTEL_GEN(dev_priv) >= 7) + if (IS_GEN7(dev_priv)) len += 2 + (num_rings ? 4*num_rings + 6 : 0); cs = intel_ring_begin(req, len); @@ -596,7 +596,7 @@ mi_set_context(struct drm_i915_gem_request *req, u32 flags) return PTR_ERR(cs); /* WaProgramMiArbOnOffAroundMiSetContext:ivb,vlv,hsw,bdw,chv */ - if (INTEL_GEN(dev_priv) >= 7) { + if (IS_GEN7(dev_priv)) { *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE; if (num_rings) { struct intel_engine_cs *signaller; @@ -623,7 +623,7 @@ mi_set_context(struct drm_i915_gem_request *req, u32 flags) */ *cs++ = MI_NOOP; - if (INTEL_GEN(dev_priv) >= 7) { + if (IS_GEN7(dev_priv)) { if (num_rings) { struct intel_engine_cs *signaller; i915_reg_t last_reg = {}; /* keep gcc quiet */ @@ -714,27 +714,7 @@ needs_pd_load_pre(struct i915_hw_ppgtt *ppgtt, struct intel_engine_cs *engine) if (engine->id != RCS) return true; - if (INTEL_GEN(engine->i915) < 8) - return true; - - return false; -} - -static bool -needs_pd_load_post(struct i915_hw_ppgtt *ppgtt, - struct i915_gem_context *to, - u32 hw_flags) -{ - if (!ppgtt) - return false; - - if (!IS_GEN8(to->i915)) - return false; - - if (hw_flags & MI_RESTORE_INHIBIT) - return true; - - return false; + return true; } static int do_rcs_switch(struct drm_i915_gem_request *req) @@ -784,21 +764,6 @@ static int do_rcs_switch(struct drm_i915_gem_request *req) engine->legacy_active_context = to; } - /* GEN8 does *not* require an explicit reload if the PDPs have been - * setup, and we do not wish to move them. - */ - if (needs_pd_load_post(ppgtt, to, hw_flags)) { - trace_switch_mm(engine, to); - ret = ppgtt->switch_mm(ppgtt, req); - /* The hardware context switch is emitted, but we haven't - * actually changed the state - so it's probably safe to bail - * here. Still, let the user know something dangerous has - * happened. - */ - if (ret) - return ret; - } - if (ppgtt) ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine); diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 7481c8e1b5a8f..3d18d67d065f1 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -793,8 +793,6 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m, "WA batchbuffer", ee->wa_batchbuffer); } - print_error_obj(m, NULL, "Semaphores", error->semaphore); - if (error->overlay) intel_overlay_print_error_state(m, error->overlay); @@ -903,8 +901,6 @@ void __i915_gpu_state_free(struct kref *error_ref) kfree(ee->waiters); } - i915_error_object_free(error->semaphore); - for (i = 0; i < ARRAY_SIZE(error->active_bo); i++) kfree(error->active_bo[i]); kfree(error->pinned_bo); @@ -1116,34 +1112,6 @@ gen8_engine_sync_index(struct intel_engine_cs *engine, return idx; } -static void gen8_record_semaphore_state(struct i915_gpu_state *error, - struct intel_engine_cs *engine, - struct drm_i915_error_engine *ee) -{ - struct drm_i915_private *dev_priv = engine->i915; - struct intel_engine_cs *to; - enum intel_engine_id id; - - if (!error->semaphore) - return; - - for_each_engine(to, dev_priv, id) { - int idx; - u16 signal_offset; - u32 *tmp; - - if (engine == to) - continue; - - signal_offset = - (GEN8_SIGNAL_OFFSET(engine, id) & (PAGE_SIZE - 1)) / 4; - tmp = error->semaphore->pages[0]; - idx = gen8_engine_sync_index(engine, to); - - ee->semaphore_mboxes[idx] = tmp[signal_offset]; - } -} - static void gen6_record_semaphore_state(struct intel_engine_cs *engine, struct drm_i915_error_engine *ee) { @@ -1218,7 +1186,6 @@ static void error_record_engine_registers(struct i915_gpu_state *error, if (INTEL_GEN(dev_priv) >= 6) { ee->rc_psmi = I915_READ(RING_PSMI_CTL(engine->mmio_base)); if (INTEL_GEN(dev_priv) >= 8) { - gen8_record_semaphore_state(error, engine, ee); ee->fault_reg = I915_READ(GEN8_RING_FAULT_REG); } else { gen6_record_semaphore_state(engine, ee); @@ -1453,9 +1420,6 @@ static void i915_gem_record_rings(struct drm_i915_private *dev_priv, struct i915_ggtt *ggtt = &dev_priv->ggtt; int i; - error->semaphore = - i915_error_object_create(dev_priv, dev_priv->semaphore); - for (i = 0; i < I915_NUM_ENGINES; i++) { struct intel_engine_cs *engine = dev_priv->engine[i]; struct drm_i915_error_engine *ee = &error->engine[i]; diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index cae76f47fec2f..1fca7ac3b059d 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -37,8 +37,6 @@ * Resource Streamer, is 66944 bytes, which rounds to 17 pages. */ #define HSW_CXT_TOTAL_SIZE (17 * PAGE_SIZE) -/* Same as Haswell, but 72064 bytes now. */ -#define GEN8_CXT_TOTAL_SIZE (18 * PAGE_SIZE) #define GEN8_LR_CONTEXT_RENDER_SIZE (20 * PAGE_SIZE) #define GEN9_LR_CONTEXT_RENDER_SIZE (22 * PAGE_SIZE) @@ -364,18 +362,6 @@ void intel_engine_init_global_seqno(struct intel_engine_cs *engine, u32 seqno) if (HAS_VEBOX(dev_priv)) I915_WRITE(RING_SYNC_2(engine->mmio_base), 0); } - if (dev_priv->semaphore) { - struct page *page = i915_vma_first_page(dev_priv->semaphore); - void *semaphores; - - /* Semaphores are in noncoherent memory, flush to be safe */ - semaphores = kmap_atomic(page); - memset(semaphores + GEN8_SEMAPHORE_OFFSET(engine->id, 0), - 0, I915_NUM_ENGINES * gen8_semaphore_seqno_size); - drm_clflush_virt_range(semaphores + GEN8_SEMAPHORE_OFFSET(engine->id, 0), - I915_NUM_ENGINES * gen8_semaphore_seqno_size); - kunmap_atomic(semaphores); - } intel_write_status_page(engine, I915_GEM_HWS_INDEX, seqno); clear_bit(ENGINE_IRQ_BREADCRUMB, &engine->irq_posted); diff --git a/drivers/gpu/drm/i915/intel_hangcheck.c b/drivers/gpu/drm/i915/intel_hangcheck.c index 12ac270a5f93e..95bbb5a79c4f8 100644 --- a/drivers/gpu/drm/i915/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/intel_hangcheck.c @@ -27,13 +27,9 @@ static bool ipehr_is_semaphore_wait(struct intel_engine_cs *engine, u32 ipehr) { - if (INTEL_GEN(engine->i915) >= 8) { - return (ipehr >> 23) == 0x1c; - } else { - ipehr &= ~MI_SEMAPHORE_SYNC_MASK; - return ipehr == (MI_SEMAPHORE_MBOX | MI_SEMAPHORE_COMPARE | - MI_SEMAPHORE_REGISTER); - } + ipehr &= ~MI_SEMAPHORE_SYNC_MASK; + return ipehr == (MI_SEMAPHORE_MBOX | MI_SEMAPHORE_COMPARE | + MI_SEMAPHORE_REGISTER); } static struct intel_engine_cs * @@ -41,31 +37,20 @@ semaphore_wait_to_signaller_ring(struct intel_engine_cs *engine, u32 ipehr, u64 offset) { struct drm_i915_private *dev_priv = engine->i915; + u32 sync_bits = ipehr & MI_SEMAPHORE_SYNC_MASK; struct intel_engine_cs *signaller; enum intel_engine_id id; - if (INTEL_GEN(dev_priv) >= 8) { - for_each_engine(signaller, dev_priv, id) { - if (engine == signaller) - continue; - - if (offset == signaller->semaphore.signal_ggtt[engine->hw_id]) - return signaller; - } - } else { - u32 sync_bits = ipehr & MI_SEMAPHORE_SYNC_MASK; - - for_each_engine(signaller, dev_priv, id) { - if(engine == signaller) - continue; + for_each_engine(signaller, dev_priv, id) { + if (engine == signaller) + continue; - if (sync_bits == signaller->semaphore.mbox.wait[engine->hw_id]) - return signaller; - } + if (sync_bits == signaller->semaphore.mbox.wait[engine->hw_id]) + return signaller; } - DRM_DEBUG_DRIVER("No signaller ring found for %s, ipehr 0x%08x, offset 0x%016llx\n", - engine->name, ipehr, offset); + DRM_DEBUG_DRIVER("No signaller ring found for %s, ipehr 0x%08x\n", + engine->name, ipehr); return ERR_PTR(-ENODEV); } @@ -135,11 +120,6 @@ semaphore_waits_for(struct intel_engine_cs *engine, u32 *seqno) return NULL; *seqno = ioread32(vaddr + head + 4) + 1; - if (INTEL_GEN(dev_priv) >= 8) { - offset = ioread32(vaddr + head + 12); - offset <<= 32; - offset |= ioread32(vaddr + head + 8); - } return semaphore_wait_to_signaller_ring(engine, ipehr, offset); } @@ -273,7 +253,7 @@ engine_stuck(struct intel_engine_cs *engine, u64 acthd) return ENGINE_WAIT_KICK; } - if (INTEL_GEN(dev_priv) >= 6 && tmp & RING_WAIT_SEMAPHORE) { + if (IS_GEN(dev_priv, 6, 7) && tmp & RING_WAIT_SEMAPHORE) { switch (semaphore_passed(engine)) { default: return ENGINE_DEAD; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index be98868115bf9..5465e708545ff 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -340,50 +340,6 @@ gen7_render_ring_flush(struct drm_i915_gem_request *req, u32 mode) return 0; } -static int -gen8_render_ring_flush(struct drm_i915_gem_request *req, u32 mode) -{ - u32 flags; - u32 *cs; - - cs = intel_ring_begin(req, mode & EMIT_INVALIDATE ? 12 : 6); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - flags = PIPE_CONTROL_CS_STALL; - - if (mode & EMIT_FLUSH) { - flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; - flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; - flags |= PIPE_CONTROL_DC_FLUSH_ENABLE; - flags |= PIPE_CONTROL_FLUSH_ENABLE; - } - if (mode & EMIT_INVALIDATE) { - flags |= PIPE_CONTROL_TLB_INVALIDATE; - flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; - flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; - flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; - flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; - flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; - flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; - - /* WaCsStallBeforeStateCacheInvalidate:bdw,chv */ - cs = gen8_emit_pipe_control(cs, - PIPE_CONTROL_CS_STALL | - PIPE_CONTROL_STALL_AT_SCOREBOARD, - 0); - } - - cs = gen8_emit_pipe_control(cs, flags, - i915_ggtt_offset(req->engine->scratch) + - 2 * CACHELINE_BYTES); - - intel_ring_advance(req, cs); - - return 0; -} - static void ring_setup_phys_status_page(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; @@ -427,7 +383,6 @@ static void intel_ring_setup_status_page(struct intel_engine_cs *engine) } else if (IS_GEN6(dev_priv)) { mmio = RING_HWS_PGA_GEN6(engine->mmio_base); } else { - /* XXX: gen8 returns to sanity */ mmio = RING_HWS_PGA(engine->mmio_base); } @@ -437,13 +392,7 @@ static void intel_ring_setup_status_page(struct intel_engine_cs *engine) I915_WRITE(mmio, engine->status_page.ggtt_offset); POSTING_READ(mmio); - /* - * Flush the TLB for this page - * - * FIXME: These two bits have disappeared on gen8, so a question - * arises: do we still need this and if so how should we go about - * invalidating the TLB? - */ + /* Flush the TLB for this page */ if (IS_GEN(dev_priv, 6, 7)) { i915_reg_t reg = RING_INSTPM(engine->mmio_base); @@ -611,8 +560,6 @@ static void reset_ring_common(struct intel_engine_cs *engine, struct intel_context *ce = &request->ctx->engine[engine->id]; struct i915_hw_ppgtt *ppgtt; - /* FIXME consider gen8 reset */ - if (ce->state) { I915_WRITE(CCID, i915_ggtt_offset(ce->state) | @@ -713,62 +660,6 @@ static int init_render_ring(struct intel_engine_cs *engine) return init_workarounds_ring(engine); } -static void render_ring_cleanup(struct intel_engine_cs *engine) -{ - struct drm_i915_private *dev_priv = engine->i915; - - i915_vma_unpin_and_release(&dev_priv->semaphore); -} - -static u32 *gen8_rcs_signal(struct drm_i915_gem_request *req, u32 *cs) -{ - struct drm_i915_private *dev_priv = req->i915; - struct intel_engine_cs *waiter; - enum intel_engine_id id; - - for_each_engine(waiter, dev_priv, id) { - u64 gtt_offset = req->engine->semaphore.signal_ggtt[id]; - if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID) - continue; - - *cs++ = GFX_OP_PIPE_CONTROL(6); - *cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_QW_WRITE | - PIPE_CONTROL_CS_STALL; - *cs++ = lower_32_bits(gtt_offset); - *cs++ = upper_32_bits(gtt_offset); - *cs++ = req->global_seqno; - *cs++ = 0; - *cs++ = MI_SEMAPHORE_SIGNAL | - MI_SEMAPHORE_TARGET(waiter->hw_id); - *cs++ = 0; - } - - return cs; -} - -static u32 *gen8_xcs_signal(struct drm_i915_gem_request *req, u32 *cs) -{ - struct drm_i915_private *dev_priv = req->i915; - struct intel_engine_cs *waiter; - enum intel_engine_id id; - - for_each_engine(waiter, dev_priv, id) { - u64 gtt_offset = req->engine->semaphore.signal_ggtt[id]; - if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID) - continue; - - *cs++ = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW; - *cs++ = lower_32_bits(gtt_offset) | MI_FLUSH_DW_USE_GTT; - *cs++ = upper_32_bits(gtt_offset); - *cs++ = req->global_seqno; - *cs++ = MI_SEMAPHORE_SIGNAL | - MI_SEMAPHORE_TARGET(waiter->hw_id); - *cs++ = 0; - } - - return cs; -} - static u32 *gen6_signal(struct drm_i915_gem_request *req, u32 *cs) { struct drm_i915_private *dev_priv = req->i915; @@ -851,70 +742,6 @@ static void gen6_sema_emit_breadcrumb(struct drm_i915_gem_request *req, u32 *cs) req->engine->semaphore.signal(req, cs)); } -static void gen8_render_emit_breadcrumb(struct drm_i915_gem_request *req, - u32 *cs) -{ - struct intel_engine_cs *engine = req->engine; - - if (engine->semaphore.signal) - cs = engine->semaphore.signal(req, cs); - - *cs++ = GFX_OP_PIPE_CONTROL(6); - *cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL | - PIPE_CONTROL_QW_WRITE; - *cs++ = intel_hws_seqno_address(engine); - *cs++ = 0; - *cs++ = req->global_seqno; - /* We're thrashing one dword of HWS. */ - *cs++ = 0; - *cs++ = MI_USER_INTERRUPT; - *cs++ = MI_NOOP; - - req->tail = intel_ring_offset(req, cs); - assert_ring_tail_valid(req->ring, req->tail); -} - -static const int gen8_render_emit_breadcrumb_sz = 8; - -/** - * intel_ring_sync - sync the waiter to the signaller on seqno - * - * @waiter - ring that is waiting - * @signaller - ring which has, or will signal - * @seqno - seqno which the waiter will block on - */ - -static int -gen8_ring_sync_to(struct drm_i915_gem_request *req, - struct drm_i915_gem_request *signal) -{ - struct drm_i915_private *dev_priv = req->i915; - u64 offset = GEN8_WAIT_OFFSET(req->engine, signal->engine->id); - struct i915_hw_ppgtt *ppgtt; - u32 *cs; - - cs = intel_ring_begin(req, 4); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - *cs++ = MI_SEMAPHORE_WAIT | MI_SEMAPHORE_GLOBAL_GTT | - MI_SEMAPHORE_SAD_GTE_SDD; - *cs++ = signal->global_seqno; - *cs++ = lower_32_bits(offset); - *cs++ = upper_32_bits(offset); - intel_ring_advance(req, cs); - - /* When the !RCS engines idle waiting upon a semaphore, they lose their - * pagetables and we must reload them before executing the batch. - * We do this on the i915_switch_context() following the wait and - * before the dispatch. - */ - ppgtt = req->ctx->ppgtt; - if (ppgtt && req->engine->id != RCS) - ppgtt->pd_dirty_rings |= intel_engine_flag(req->engine); - return 0; -} - static int gen6_ring_sync_to(struct drm_i915_gem_request *req, struct drm_i915_gem_request *signal) @@ -1090,25 +917,6 @@ hsw_vebox_irq_disable(struct intel_engine_cs *engine) gen6_mask_pm_irq(dev_priv, engine->irq_enable_mask); } -static void -gen8_irq_enable(struct intel_engine_cs *engine) -{ - struct drm_i915_private *dev_priv = engine->i915; - - I915_WRITE_IMR(engine, - ~(engine->irq_enable_mask | - engine->irq_keep_mask)); - POSTING_READ_FW(RING_IMR(engine->mmio_base)); -} - -static void -gen8_irq_disable(struct intel_engine_cs *engine) -{ - struct drm_i915_private *dev_priv = engine->i915; - - I915_WRITE_IMR(engine, ~engine->irq_keep_mask); -} - static int i965_emit_bb_start(struct drm_i915_gem_request *req, u64 offset, u32 length, @@ -1796,8 +1604,6 @@ static int gen6_bsd_ring_flush(struct drm_i915_gem_request *req, u32 mode) return PTR_ERR(cs); cmd = MI_FLUSH_DW; - if (INTEL_GEN(req->i915) >= 8) - cmd += 1; /* We always require a command barrier so that subsequent * commands, such as breadcrumb interrupts, are strictly ordered @@ -1817,38 +1623,9 @@ static int gen6_bsd_ring_flush(struct drm_i915_gem_request *req, u32 mode) *cs++ = cmd; *cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT; - if (INTEL_GEN(req->i915) >= 8) { - *cs++ = 0; /* upper addr */ - *cs++ = 0; /* value */ - } else { - *cs++ = 0; - *cs++ = MI_NOOP; - } - intel_ring_advance(req, cs); - return 0; -} - -static int -gen8_emit_bb_start(struct drm_i915_gem_request *req, - u64 offset, u32 len, - unsigned int dispatch_flags) -{ - bool ppgtt = USES_PPGTT(req->i915) && - !(dispatch_flags & I915_DISPATCH_SECURE); - u32 *cs; - - cs = intel_ring_begin(req, 4); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - /* FIXME(BDW): Address space and security selectors. */ - *cs++ = MI_BATCH_BUFFER_START_GEN8 | (ppgtt << 8) | (dispatch_flags & - I915_DISPATCH_RS ? MI_BATCH_RESOURCE_STREAMER : 0); - *cs++ = lower_32_bits(offset); - *cs++ = upper_32_bits(offset); + *cs++ = 0; *cs++ = MI_NOOP; intel_ring_advance(req, cs); - return 0; } @@ -1905,8 +1682,6 @@ static int gen6_ring_flush(struct drm_i915_gem_request *req, u32 mode) return PTR_ERR(cs); cmd = MI_FLUSH_DW; - if (INTEL_GEN(req->i915) >= 8) - cmd += 1; /* We always require a command barrier so that subsequent * commands, such as breadcrumb interrupts, are strictly ordered @@ -1925,13 +1700,8 @@ static int gen6_ring_flush(struct drm_i915_gem_request *req, u32 mode) cmd |= MI_INVALIDATE_TLB; *cs++ = cmd; *cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT; - if (INTEL_GEN(req->i915) >= 8) { - *cs++ = 0; /* upper addr */ - *cs++ = 0; /* value */ - } else { - *cs++ = 0; - *cs++ = MI_NOOP; - } + *cs++ = 0; + *cs++ = MI_NOOP; intel_ring_advance(req, cs); return 0; @@ -1940,110 +1710,61 @@ static int gen6_ring_flush(struct drm_i915_gem_request *req, u32 mode) static void intel_ring_init_semaphores(struct drm_i915_private *dev_priv, struct intel_engine_cs *engine) { - struct drm_i915_gem_object *obj; - int ret, i; + int i; if (!i915_modparams.semaphores) return; - if (INTEL_GEN(dev_priv) >= 8 && !dev_priv->semaphore) { - struct i915_vma *vma; - - obj = i915_gem_object_create(dev_priv, PAGE_SIZE); - if (IS_ERR(obj)) - goto err; - - vma = i915_vma_instance(obj, &dev_priv->ggtt.base, NULL); - if (IS_ERR(vma)) - goto err_obj; - - ret = i915_gem_object_set_to_gtt_domain(obj, false); - if (ret) - goto err_obj; - - ret = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH); - if (ret) - goto err_obj; + GEM_BUG_ON(INTEL_GEN(dev_priv) < 6); + engine->semaphore.sync_to = gen6_ring_sync_to; + engine->semaphore.signal = gen6_signal; - dev_priv->semaphore = vma; - } - - if (INTEL_GEN(dev_priv) >= 8) { - u32 offset = i915_ggtt_offset(dev_priv->semaphore); - - engine->semaphore.sync_to = gen8_ring_sync_to; - engine->semaphore.signal = gen8_xcs_signal; - - for (i = 0; i < I915_NUM_ENGINES; i++) { - u32 ring_offset; - - if (i != engine->id) - ring_offset = offset + GEN8_SEMAPHORE_OFFSET(engine->id, i); - else - ring_offset = MI_SEMAPHORE_SYNC_INVALID; - - engine->semaphore.signal_ggtt[i] = ring_offset; - } - } else if (INTEL_GEN(dev_priv) >= 6) { - engine->semaphore.sync_to = gen6_ring_sync_to; - engine->semaphore.signal = gen6_signal; - - /* - * The current semaphore is only applied on pre-gen8 - * platform. And there is no VCS2 ring on the pre-gen8 - * platform. So the semaphore between RCS and VCS2 is - * initialized as INVALID. Gen8 will initialize the - * sema between VCS2 and RCS later. - */ - for (i = 0; i < GEN6_NUM_SEMAPHORES; i++) { - static const struct { - u32 wait_mbox; - i915_reg_t mbox_reg; - } sem_data[GEN6_NUM_SEMAPHORES][GEN6_NUM_SEMAPHORES] = { - [RCS_HW] = { - [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RV, .mbox_reg = GEN6_VRSYNC }, - [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RB, .mbox_reg = GEN6_BRSYNC }, - [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RVE, .mbox_reg = GEN6_VERSYNC }, - }, - [VCS_HW] = { - [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VR, .mbox_reg = GEN6_RVSYNC }, - [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VB, .mbox_reg = GEN6_BVSYNC }, - [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VVE, .mbox_reg = GEN6_VEVSYNC }, - }, - [BCS_HW] = { - [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BR, .mbox_reg = GEN6_RBSYNC }, - [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BV, .mbox_reg = GEN6_VBSYNC }, - [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BVE, .mbox_reg = GEN6_VEBSYNC }, - }, - [VECS_HW] = { - [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VER, .mbox_reg = GEN6_RVESYNC }, - [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEV, .mbox_reg = GEN6_VVESYNC }, - [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEB, .mbox_reg = GEN6_BVESYNC }, - }, - }; + /* + * The current semaphore is only applied on pre-gen8 + * platform. And there is no VCS2 ring on the pre-gen8 + * platform. So the semaphore between RCS and VCS2 is + * initialized as INVALID. + */ + for (i = 0; i < GEN6_NUM_SEMAPHORES; i++) { + static const struct { u32 wait_mbox; i915_reg_t mbox_reg; + } sem_data[GEN6_NUM_SEMAPHORES][GEN6_NUM_SEMAPHORES] = { + [RCS_HW] = { + [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RV, .mbox_reg = GEN6_VRSYNC }, + [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RB, .mbox_reg = GEN6_BRSYNC }, + [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RVE, .mbox_reg = GEN6_VERSYNC }, + }, + [VCS_HW] = { + [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VR, .mbox_reg = GEN6_RVSYNC }, + [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VB, .mbox_reg = GEN6_BVSYNC }, + [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VVE, .mbox_reg = GEN6_VEVSYNC }, + }, + [BCS_HW] = { + [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BR, .mbox_reg = GEN6_RBSYNC }, + [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BV, .mbox_reg = GEN6_VBSYNC }, + [VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BVE, .mbox_reg = GEN6_VEBSYNC }, + }, + [VECS_HW] = { + [RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VER, .mbox_reg = GEN6_RVESYNC }, + [VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEV, .mbox_reg = GEN6_VVESYNC }, + [BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEB, .mbox_reg = GEN6_BVESYNC }, + }, + }; + u32 wait_mbox; + i915_reg_t mbox_reg; - if (i == engine->hw_id) { - wait_mbox = MI_SEMAPHORE_SYNC_INVALID; - mbox_reg = GEN6_NOSYNC; - } else { - wait_mbox = sem_data[engine->hw_id][i].wait_mbox; - mbox_reg = sem_data[engine->hw_id][i].mbox_reg; - } - - engine->semaphore.mbox.wait[i] = wait_mbox; - engine->semaphore.mbox.signal[i] = mbox_reg; + if (i == engine->hw_id) { + wait_mbox = MI_SEMAPHORE_SYNC_INVALID; + mbox_reg = GEN6_NOSYNC; + } else { + wait_mbox = sem_data[engine->hw_id][i].wait_mbox; + mbox_reg = sem_data[engine->hw_id][i].mbox_reg; } - } - - return; -err_obj: - i915_gem_object_put(obj); -err: - DRM_DEBUG_DRIVER("Failed to allocate space for semaphores, disabling\n"); - i915_modparams.semaphores = 0; + engine->semaphore.mbox.wait[i] = wait_mbox; + engine->semaphore.mbox.signal[i] = mbox_reg; + } } static void intel_ring_init_irq(struct drm_i915_private *dev_priv, @@ -2051,11 +1772,7 @@ static void intel_ring_init_irq(struct drm_i915_private *dev_priv, { engine->irq_enable_mask = GT_RENDER_USER_INTERRUPT << engine->irq_shift; - if (INTEL_GEN(dev_priv) >= 8) { - engine->irq_enable = gen8_irq_enable; - engine->irq_disable = gen8_irq_disable; - engine->irq_seqno_barrier = gen6_seqno_barrier; - } else if (INTEL_GEN(dev_priv) >= 6) { + if (INTEL_GEN(dev_priv) >= 6) { engine->irq_enable = gen6_irq_enable; engine->irq_disable = gen6_irq_disable; engine->irq_seqno_barrier = gen6_seqno_barrier; @@ -2090,6 +1807,9 @@ static void gen6_bsd_set_default_submission(struct intel_engine_cs *engine) static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv, struct intel_engine_cs *engine) { + /* gen8+ are only supported with execlists */ + GEM_BUG_ON(INTEL_GEN(dev_priv) >= 8); + intel_ring_init_irq(dev_priv, engine); intel_ring_init_semaphores(dev_priv, engine); @@ -2109,20 +1829,14 @@ static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv, engine->emit_breadcrumb = gen6_sema_emit_breadcrumb; num_rings = INTEL_INFO(dev_priv)->num_rings - 1; - if (INTEL_GEN(dev_priv) >= 8) { - engine->emit_breadcrumb_sz += num_rings * 6; - } else { - engine->emit_breadcrumb_sz += num_rings * 3; - if (num_rings & 1) - engine->emit_breadcrumb_sz++; - } + engine->emit_breadcrumb_sz += num_rings * 3; + if (num_rings & 1) + engine->emit_breadcrumb_sz++; } engine->set_default_submission = i9xx_set_default_submission; - if (INTEL_GEN(dev_priv) >= 8) - engine->emit_bb_start = gen8_emit_bb_start; - else if (INTEL_GEN(dev_priv) >= 6) + if (INTEL_GEN(dev_priv) >= 6) engine->emit_bb_start = gen6_emit_bb_start; else if (INTEL_GEN(dev_priv) >= 4) engine->emit_bb_start = i965_emit_bb_start; @@ -2142,20 +1856,7 @@ int intel_init_render_ring_buffer(struct intel_engine_cs *engine) if (HAS_L3_DPF(dev_priv)) engine->irq_keep_mask = GT_RENDER_L3_PARITY_ERROR_INTERRUPT; - if (INTEL_GEN(dev_priv) >= 8) { - engine->init_context = intel_rcs_ctx_init; - engine->emit_breadcrumb = gen8_render_emit_breadcrumb; - engine->emit_breadcrumb_sz = gen8_render_emit_breadcrumb_sz; - engine->emit_flush = gen8_render_ring_flush; - if (i915_modparams.semaphores) { - int num_rings; - - engine->semaphore.signal = gen8_rcs_signal; - - num_rings = INTEL_INFO(dev_priv)->num_rings - 1; - engine->emit_breadcrumb_sz += num_rings * 8; - } - } else if (INTEL_GEN(dev_priv) >= 6) { + if (INTEL_GEN(dev_priv) >= 6) { engine->init_context = intel_rcs_ctx_init; engine->emit_flush = gen7_render_ring_flush; if (IS_GEN6(dev_priv)) @@ -2174,7 +1875,6 @@ int intel_init_render_ring_buffer(struct intel_engine_cs *engine) engine->emit_bb_start = hsw_emit_bb_start; engine->init_hw = init_render_ring; - engine->cleanup = render_ring_cleanup; ret = intel_init_ring_buffer(engine); if (ret) @@ -2204,8 +1904,7 @@ int intel_init_bsd_ring_buffer(struct intel_engine_cs *engine) if (IS_GEN6(dev_priv)) engine->set_default_submission = gen6_bsd_set_default_submission; engine->emit_flush = gen6_bsd_ring_flush; - if (INTEL_GEN(dev_priv) < 8) - engine->irq_enable_mask = GT_BSD_USER_INTERRUPT; + engine->irq_enable_mask = GT_BSD_USER_INTERRUPT; } else { engine->mmio_base = BSD_RING_BASE; engine->emit_flush = bsd_ring_flush; @@ -2225,8 +1924,7 @@ int intel_init_blt_ring_buffer(struct intel_engine_cs *engine) intel_ring_default_vfuncs(dev_priv, engine); engine->emit_flush = gen6_ring_flush; - if (INTEL_GEN(dev_priv) < 8) - engine->irq_enable_mask = GT_BLT_USER_INTERRUPT; + engine->irq_enable_mask = GT_BLT_USER_INTERRUPT; return intel_init_ring_buffer(engine); } @@ -2238,12 +1936,9 @@ int intel_init_vebox_ring_buffer(struct intel_engine_cs *engine) intel_ring_default_vfuncs(dev_priv, engine); engine->emit_flush = gen6_ring_flush; - - if (INTEL_GEN(dev_priv) < 8) { - engine->irq_enable_mask = PM_VEBOX_USER_INTERRUPT; - engine->irq_enable = hsw_vebox_irq_enable; - engine->irq_disable = hsw_vebox_irq_disable; - } + engine->irq_enable_mask = PM_VEBOX_USER_INTERRUPT; + engine->irq_enable = hsw_vebox_irq_enable; + engine->irq_disable = hsw_vebox_irq_disable; return intel_init_ring_buffer(engine); } diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index e437a9556c087..80cd7812ce023 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -46,16 +46,6 @@ struct intel_hw_status_page { /* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to * do the writes, and that must have qw aligned offsets, simply pretend it's 8b. */ -#define gen8_semaphore_seqno_size sizeof(uint64_t) -#define GEN8_SEMAPHORE_OFFSET(__from, __to) \ - (((__from) * I915_NUM_ENGINES + (__to)) * gen8_semaphore_seqno_size) -#define GEN8_SIGNAL_OFFSET(__ring, to) \ - (dev_priv->semaphore->node.start + \ - GEN8_SEMAPHORE_OFFSET((__ring)->id, (to))) -#define GEN8_WAIT_OFFSET(__ring, from) \ - (dev_priv->semaphore->node.start + \ - GEN8_SEMAPHORE_OFFSET(from, (__ring)->id)) - enum intel_engine_hangcheck_action { ENGINE_IDLE = 0, ENGINE_WAIT, @@ -467,18 +457,15 @@ struct intel_engine_cs { * ie. transpose of f(x, y) */ struct { - union { #define GEN6_SEMAPHORE_LAST VECS_HW #define GEN6_NUM_SEMAPHORES (GEN6_SEMAPHORE_LAST + 1) #define GEN6_SEMAPHORES_MASK GENMASK(GEN6_SEMAPHORE_LAST, 0) - struct { - /* our mbox written by others */ - u32 wait[GEN6_NUM_SEMAPHORES]; - /* mboxes this ring signals to */ - i915_reg_t signal[GEN6_NUM_SEMAPHORES]; - } mbox; - u64 signal_ggtt[I915_NUM_ENGINES]; - }; + struct { + /* our mbox written by others */ + u32 wait[GEN6_NUM_SEMAPHORES]; + /* mboxes this ring signals to */ + i915_reg_t signal[GEN6_NUM_SEMAPHORES]; + } mbox; /* AKA wait() */ int (*sync_to)(struct drm_i915_gem_request *req, -- GitLab From 0da715ee60774401bea00dc71fca6fd1096c734a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 20:55:02 +0000 Subject: [PATCH 0367/1471] drm/i915: Disable semaphores on Sandybridge I should have admitted defeat long ago as there has been a rare but persistent error on Sandybridge where semaphore signaling did not propagate to the waiter, leading to a GPU hang. With the work on fence signaling for v4.9, the impact of using CPU driven signaling was greatly reduced wrt to the latency of GPU semaphores, though without logical rings support, the benefit of reordering work to avoid bubbles is not realised (i.e. as it stands fence signaling is just a slower, more costly version of HW semaphores; but works more consistently). As a rough indicator of the difference, with semaphores: Sequential (3 engines, 1 processes): average 5.470us per cycle [expected 4.988us] w/o semaphores: Sequential (3 engines, 1 processes): average 15.771us per cycle [expected 4.923us] In comparison, v3.4: with semaphores: Sequential (3 engines, 1 processes): average 16.066us per cycle [expected 11.842us] w/o semaphores: Sequential (3 engines, 1 processes): average 23.460us per cycle [expected 11.839us] Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54226 #and 100+ dupes Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Acked-by: Mika Kuoppala Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-3-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d470e10b2edd0..d53bb8e872ba6 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4999,20 +4999,12 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value) { - if (INTEL_GEN(dev_priv) < 6) - return false; - - /* TODO: make semaphores and Execlists play nicely together */ - if (HAS_EXECLISTS(dev_priv)) + if (!IS_GEN7(dev_priv)) return false; if (value >= 0) return value; - /* Enable semaphores on SNB when IO remapping is off */ - if (IS_GEN6(dev_priv) && intel_vtd_active()) - return false; - return true; } -- GitLab From af9ff6c70df452b42779aac1bb787ec72d5f257b Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 20:55:03 +0000 Subject: [PATCH 0368/1471] drm/i915: Move debugfs/i915_semaphore_status to i915_engine_info As the semaphores is just part of the engine, include it with the general pretty printer universally used for debugging. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-4-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_debugfs.c | 32 -------------------------- drivers/gpu/drm/i915/intel_engine_cs.c | 9 ++++++++ 2 files changed, 9 insertions(+), 32 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 9cef1463d4113..41d49a4d25d35 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -3235,37 +3235,6 @@ static int i915_shrinker_info(struct seq_file *m, void *unused) return 0; } -static int i915_semaphore_status(struct seq_file *m, void *unused) -{ - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct drm_device *dev = &dev_priv->drm; - struct intel_engine_cs *engine; - int num_rings = INTEL_INFO(dev_priv)->num_rings; - enum intel_engine_id id; - int j, ret; - - if (!i915_modparams.semaphores) { - seq_puts(m, "Semaphores are disabled\n"); - return 0; - } - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - return ret; - intel_runtime_pm_get(dev_priv); - - seq_puts(m, " Last signal:"); - for_each_engine(engine, dev_priv, id) - for (j = 0; j < num_rings; j++) - seq_printf(m, "0x%08x\n", - I915_READ(engine->semaphore.mbox.signal[j])); - seq_putc(m, '\n'); - - intel_runtime_pm_put(dev_priv); - mutex_unlock(&dev->struct_mutex); - return 0; -} - static int i915_shared_dplls_info(struct seq_file *m, void *unused) { struct drm_i915_private *dev_priv = node_to_i915(m->private); @@ -4745,7 +4714,6 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_display_info", i915_display_info, 0}, {"i915_engine_info", i915_engine_info, 0}, {"i915_shrinker_info", i915_shrinker_info, 0}, - {"i915_semaphore_status", i915_semaphore_status, 0}, {"i915_shared_dplls_info", i915_shared_dplls_info, 0}, {"i915_dp_mst_info", i915_dp_mst_info, 0}, {"i915_wa_registers", i915_wa_registers, 0}, diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 1fca7ac3b059d..ef8e101ebd987 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1713,6 +1713,15 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) I915_READ(RING_MI_MODE(engine->mmio_base)), I915_READ(RING_MI_MODE(engine->mmio_base)) & (MODE_IDLE) ? " [idle]" : ""); } + if (i915_modparams.semaphores) { + drm_printf(m, "\tSYNC_0: 0x%08x\n", + I915_READ(RING_SYNC_0(engine->mmio_base))); + drm_printf(m, "\tSYNC_1: 0x%08x\n", + I915_READ(RING_SYNC_1(engine->mmio_base))); + if (HAS_VEBOX(dev_priv)) + drm_printf(m, "\tSYNC_2: 0x%08x\n", + I915_READ(RING_SYNC_2(engine->mmio_base))); + } rcu_read_unlock(); -- GitLab From 93c6e966b4cd1c99d02b488e149df606f811587b Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 20:55:04 +0000 Subject: [PATCH 0369/1471] drm/i915: Remove i915.semaphores modparam Having disabled the broken semaphores on Sandybridge, there is no need for a modparam any more, so remove it in favour of a simple HAS_LEGACY_SEMAPHORES() guard. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Maarten Lankhorst Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-5-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_drv.c | 7 +------ drivers/gpu/drm/i915/i915_drv.h | 4 ++-- drivers/gpu/drm/i915/i915_gem.c | 11 ----------- drivers/gpu/drm/i915/i915_gem_context.c | 2 +- drivers/gpu/drm/i915/i915_params.c | 4 ---- drivers/gpu/drm/i915/i915_params.h | 1 - drivers/gpu/drm/i915/intel_engine_cs.c | 2 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 4 ++-- 8 files changed, 7 insertions(+), 28 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index c608fd96771cc..8dbcb03b5f54b 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -321,7 +321,7 @@ static int i915_getparam(struct drm_device *dev, void *data, value = USES_PPGTT(dev_priv); break; case I915_PARAM_HAS_SEMAPHORES: - value = i915_modparams.semaphores; + value = HAS_LEGACY_SEMAPHORES(dev_priv); break; case I915_PARAM_HAS_SECURE_BATCHES: value = capable(CAP_SYS_ADMIN); @@ -1063,11 +1063,6 @@ static void intel_sanitize_options(struct drm_i915_private *dev_priv) i915_modparams.enable_ppgtt); DRM_DEBUG_DRIVER("ppgtt mode: %i\n", i915_modparams.enable_ppgtt); - i915_modparams.semaphores = - intel_sanitize_semaphores(dev_priv, i915_modparams.semaphores); - DRM_DEBUG_DRIVER("use GPU semaphores? %s\n", - yesno(i915_modparams.semaphores)); - intel_uc_sanitize_options(dev_priv); intel_gvt_sanitize_options(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5ce778c26bb88..9a0654e10a554 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3139,6 +3139,8 @@ intel_info(const struct drm_i915_private *dev_priv) #define HAS_BLT(dev_priv) HAS_ENGINE(dev_priv, BCS) #define HAS_VEBOX(dev_priv) HAS_ENGINE(dev_priv, VECS) +#define HAS_LEGACY_SEMAPHORES(dev_priv) IS_GEN7(dev_priv) + #define HAS_LLC(dev_priv) ((dev_priv)->info.has_llc) #define HAS_SNOOP(dev_priv) ((dev_priv)->info.has_snoop) #define HAS_EDRAM(dev_priv) (!!((dev_priv)->edram_cap & EDRAM_ENABLED)) @@ -3302,8 +3304,6 @@ intel_ggtt_update_needs_vtd_wa(struct drm_i915_private *dev_priv) int intel_sanitize_enable_ppgtt(struct drm_i915_private *dev_priv, int enable_ppgtt); -bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value); - /* i915_drv.c */ void __printf(3, 4) __i915_printk(struct drm_i915_private *dev_priv, const char *level, diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index d53bb8e872ba6..792e6dc7e19b1 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4997,17 +4997,6 @@ int i915_gem_init_hw(struct drm_i915_private *dev_priv) return ret; } -bool intel_sanitize_semaphores(struct drm_i915_private *dev_priv, int value) -{ - if (!IS_GEN7(dev_priv)) - return false; - - if (value >= 0) - return value; - - return true; -} - static int __intel_engines_record_defaults(struct drm_i915_private *i915) { struct i915_gem_context *ctx; diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 0704d9af261bc..6ca56e482d793 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -574,7 +574,7 @@ mi_set_context(struct drm_i915_gem_request *req, u32 flags) enum intel_engine_id id; const int num_rings = /* Use an extended w/a on gen7 if signalling from other rings */ - (i915_modparams.semaphores && IS_GEN7(dev_priv)) ? + (HAS_LEGACY_SEMAPHORES(dev_priv) && IS_GEN7(dev_priv)) ? INTEL_INFO(dev_priv)->num_rings - 1 : 0; int len; diff --git a/drivers/gpu/drm/i915/i915_params.c b/drivers/gpu/drm/i915/i915_params.c index d61c1787c1641..3328147b48630 100644 --- a/drivers/gpu/drm/i915/i915_params.c +++ b/drivers/gpu/drm/i915/i915_params.c @@ -46,10 +46,6 @@ i915_param_named_unsafe(panel_ignore_lid, int, 0600, "Override lid status (0=autodetect, 1=autodetect disabled [default], " "-1=force lid closed, -2=force lid open)"); -i915_param_named_unsafe(semaphores, int, 0400, - "Use semaphores for inter-ring sync " - "(default: -1 (use per-chip defaults))"); - i915_param_named_unsafe(enable_rc6, int, 0400, "Enable power-saving render C-state 6. " "Different stages can be selected via bitmask values " diff --git a/drivers/gpu/drm/i915/i915_params.h b/drivers/gpu/drm/i915/i915_params.h index 0aef3d7178ebf..8321bd86cba54 100644 --- a/drivers/gpu/drm/i915/i915_params.h +++ b/drivers/gpu/drm/i915/i915_params.h @@ -31,7 +31,6 @@ param(char *, vbt_firmware, NULL) \ param(int, modeset, -1) \ param(int, panel_ignore_lid, 1) \ - param(int, semaphores, -1) \ param(int, lvds_channel_mode, 0) \ param(int, panel_use_ssc, -1) \ param(int, vbt_sdvo_panel_type, -1) \ diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index ef8e101ebd987..22c095035539c 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1713,7 +1713,7 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) I915_READ(RING_MI_MODE(engine->mmio_base)), I915_READ(RING_MI_MODE(engine->mmio_base)) & (MODE_IDLE) ? " [idle]" : ""); } - if (i915_modparams.semaphores) { + if (HAS_LEGACY_SEMAPHORES(dev_priv)) { drm_printf(m, "\tSYNC_0: 0x%08x\n", I915_READ(RING_SYNC_0(engine->mmio_base))); drm_printf(m, "\tSYNC_1: 0x%08x\n", diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 5465e708545ff..bfa11a84e476e 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1712,7 +1712,7 @@ static void intel_ring_init_semaphores(struct drm_i915_private *dev_priv, { int i; - if (!i915_modparams.semaphores) + if (!HAS_LEGACY_SEMAPHORES(dev_priv)) return; GEM_BUG_ON(INTEL_GEN(dev_priv) < 6); @@ -1823,7 +1823,7 @@ static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv, engine->emit_breadcrumb = i9xx_emit_breadcrumb; engine->emit_breadcrumb_sz = i9xx_emit_breadcrumb_sz; - if (i915_modparams.semaphores) { + if (HAS_LEGACY_SEMAPHORES(dev_priv)) { int num_rings; engine->emit_breadcrumb = gen6_sema_emit_breadcrumb; -- GitLab From 0ab42a78712acbdcfefe5aa3205a15393ebbc6d7 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 21:19:07 +0000 Subject: [PATCH 0370/1471] drm/i915/selftests: Declare we allocated the guc clients Silence smatch over drivers/gpu/drm/i915/selftests/intel_guc.c:135 igt_guc_init_doorbell_hw() error: we previously assumed 'guc->execbuf_client' could be null (see line 123) drivers/gpu/drm/i915/selftests/intel_guc.c:142 igt_guc_init_doorbell_hw() error: we previously assumed 'guc->preempt_client' could be null (see line 123) by asserting that we did succeed in creating the pair of clients for testing. References: 55bd6bd75717 ("drm/i915/selftests: Add a GuC doorbells selftest") Signed-off-by: Chris Wilson Cc: Michel Thierry Cc: Michal Wajdeczko Cc: Daniele Ceraolo Spurio Link: https://patchwork.freedesktop.org/patch/msgid/20171120211907.1649-1-chris@chris-wilson.co.uk Reviewed-by: Michel Thierry --- drivers/gpu/drm/i915/selftests/intel_guc.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c index f1baeda8b4a82..7b23597858bbf 100644 --- a/drivers/gpu/drm/i915/selftests/intel_guc.c +++ b/drivers/gpu/drm/i915/selftests/intel_guc.c @@ -131,6 +131,8 @@ static int igt_guc_init_doorbell_hw(void *args) pr_err("Failed to create clients\n"); goto unlock; } + GEM_BUG_ON(!guc->execbuf_client); + GEM_BUG_ON(!guc->preempt_client); err = validate_client(guc->execbuf_client, GUC_CLIENT_PRIORITY_KMD_NORMAL, false); -- GitLab From 36fe778a489e0461126fc657ff333522c24d5391 Mon Sep 17 00:00:00 2001 From: David Weinehall Date: Fri, 17 Nov 2017 10:01:46 +0200 Subject: [PATCH 0371/1471] drm/i915: Don't use GEN6_RC_VIDEO_FREQ on gen10+ GEN6_RC_VIDEO_FREQ is deprecated for >= gen10; don't try to program it. v2: Use IS_GEN9() instead of INTEL_GEN() and remove comment (Rodrigo) Signed-off-by: David Weinehall Reviewed-by: Rodrigo Vivi Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171117080146.20150-1-david.weinehall@linux.intel.com --- drivers/gpu/drm/i915/intel_pm.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 4d2cd432f7393..ce5cd75b81301 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6602,9 +6602,10 @@ static void gen9_enable_rps(struct drm_i915_private *dev_priv) { intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL); - /* Program defaults and thresholds for RPS*/ - I915_WRITE(GEN6_RC_VIDEO_FREQ, - GEN9_FREQUENCY(dev_priv->gt_pm.rps.rp1_freq)); + /* Program defaults and thresholds for RPS */ + if (IS_GEN9(dev_priv)) + I915_WRITE(GEN6_RC_VIDEO_FREQ, + GEN9_FREQUENCY(dev_priv->gt_pm.rps.rp1_freq)); /* 1 second timeout*/ I915_WRITE(GEN6_RP_DOWN_TIMEOUT, -- GitLab From 88f8156fba43d040dc5af42f88db2c53d6c69443 Mon Sep 17 00:00:00 2001 From: Dave Stevenson Date: Thu, 16 Nov 2017 14:22:29 +0000 Subject: [PATCH 0372/1471] drm/vc4: Add support for DRM_FORMAT_RGB888 and DRM_FORMAT_BGR888 Filling out the list of supported formats based on those the hardware can support. Signed-off-by: Dave Stevenson Signed-off-by: Eric Anholt Reviewed-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/b551205d1c33fa49eef2c33ed2d60c5339b2f299.1510841336.git.dave.stevenson@raspberrypi.org --- drivers/gpu/drm/vc4/vc4_plane.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/drivers/gpu/drm/vc4/vc4_plane.c b/drivers/gpu/drm/vc4/vc4_plane.c index 423a23ed8fc2f..513b4b1175bb7 100644 --- a/drivers/gpu/drm/vc4/vc4_plane.c +++ b/drivers/gpu/drm/vc4/vc4_plane.c @@ -120,6 +120,14 @@ static const struct hvs_format { .drm = DRM_FORMAT_XRGB1555, .hvs = HVS_PIXEL_FORMAT_RGBA5551, .pixel_order = HVS_PIXEL_ORDER_ABGR, .has_alpha = false, }, + { + .drm = DRM_FORMAT_RGB888, .hvs = HVS_PIXEL_FORMAT_RGB888, + .pixel_order = HVS_PIXEL_ORDER_XRGB, .has_alpha = false, + }, + { + .drm = DRM_FORMAT_BGR888, .hvs = HVS_PIXEL_FORMAT_RGB888, + .pixel_order = HVS_PIXEL_ORDER_XBGR, .has_alpha = false, + }, { .drm = DRM_FORMAT_YUV422, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_3PLANE, -- GitLab From 090cb0c690183be849e2bfa0427220f1e435fa30 Mon Sep 17 00:00:00 2001 From: Dave Stevenson Date: Thu, 16 Nov 2017 14:22:30 +0000 Subject: [PATCH 0373/1471] drm/vc4: Use .pixel_order instead of custom .flip_cbcr The hardware has enums for altering the Cr and Cb order, so use this instead of having a flag which swaps the order the pointers are presented to the hardware (that only worked for 3 plane formats anyway). Explicitly sets .pixel_order in each case, rather than relying on then default XYCBCR order being a value 0. Signed-off-by: Dave Stevenson Signed-off-by: Eric Anholt Reviewed-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/563872b69c1e5df142cb15ebfca7f20056b8a64c.1510841336.git.dave.stevenson@raspberrypi.org --- drivers/gpu/drm/vc4/vc4_plane.c | 20 ++++++++------------ 1 file changed, 8 insertions(+), 12 deletions(-) diff --git a/drivers/gpu/drm/vc4/vc4_plane.c b/drivers/gpu/drm/vc4/vc4_plane.c index 513b4b1175bb7..6373fd5686ade 100644 --- a/drivers/gpu/drm/vc4/vc4_plane.c +++ b/drivers/gpu/drm/vc4/vc4_plane.c @@ -86,7 +86,6 @@ static const struct hvs_format { u32 hvs; /* HVS_FORMAT_* */ u32 pixel_order; bool has_alpha; - bool flip_cbcr; } hvs_formats[] = { { .drm = DRM_FORMAT_XRGB8888, .hvs = HVS_PIXEL_FORMAT_RGBA8888, @@ -131,28 +130,32 @@ static const struct hvs_format { { .drm = DRM_FORMAT_YUV422, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_3PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, { .drm = DRM_FORMAT_YVU422, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_3PLANE, - .flip_cbcr = true, + .pixel_order = HVS_PIXEL_ORDER_XYCRCB, }, { .drm = DRM_FORMAT_YUV420, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV420_3PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, { .drm = DRM_FORMAT_YVU420, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV420_3PLANE, - .flip_cbcr = true, + .pixel_order = HVS_PIXEL_ORDER_XYCRCB, }, { .drm = DRM_FORMAT_NV12, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV420_2PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, { .drm = DRM_FORMAT_NV16, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_2PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, }; @@ -625,15 +628,8 @@ static int vc4_plane_mode_set(struct drm_plane *plane, * The pointers may be any byte address. */ vc4_state->ptr0_offset = vc4_state->dlist_count; - if (!format->flip_cbcr) { - for (i = 0; i < num_planes; i++) - vc4_dlist_write(vc4_state, vc4_state->offsets[i]); - } else { - WARN_ON_ONCE(num_planes != 3); - vc4_dlist_write(vc4_state, vc4_state->offsets[0]); - vc4_dlist_write(vc4_state, vc4_state->offsets[2]); - vc4_dlist_write(vc4_state, vc4_state->offsets[1]); - } + for (i = 0; i < num_planes; i++) + vc4_dlist_write(vc4_state, vc4_state->offsets[i]); /* Pointer Context Word 0/1/2: Written by the HVS */ for (i = 0; i < num_planes; i++) -- GitLab From cb20dd170d6a7d41e0f347998771b0e0db183438 Mon Sep 17 00:00:00 2001 From: Dave Stevenson Date: Thu, 16 Nov 2017 14:22:31 +0000 Subject: [PATCH 0374/1471] drm/vc4: Add support for NV21 and NV61. NV12 (YUV420 2 plane) and NV16 (YUV422 2 plane) were supported, but NV21 and NV61 (same but with Cb and Cr swapped) weren't. Add them. Signed-off-by: Dave Stevenson Signed-off-by: Eric Anholt Reviewed-by: Eric Anholt Link: https://patchwork.freedesktop.org/patch/msgid/1f50799525e3401551dff2b0b2828b9ab892f75f.1510841336.git.dave.stevenson@raspberrypi.org --- drivers/gpu/drm/vc4/vc4_plane.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/gpu/drm/vc4/vc4_plane.c b/drivers/gpu/drm/vc4/vc4_plane.c index 6373fd5686ade..515f979976240 100644 --- a/drivers/gpu/drm/vc4/vc4_plane.c +++ b/drivers/gpu/drm/vc4/vc4_plane.c @@ -152,11 +152,21 @@ static const struct hvs_format { .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV420_2PLANE, .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, + { + .drm = DRM_FORMAT_NV21, + .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV420_2PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCRCB, + }, { .drm = DRM_FORMAT_NV16, .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_2PLANE, .pixel_order = HVS_PIXEL_ORDER_XYCBCR, }, + { + .drm = DRM_FORMAT_NV61, + .hvs = HVS_PIXEL_FORMAT_YCBCR_YUV422_2PLANE, + .pixel_order = HVS_PIXEL_ORDER_XYCRCB, + }, }; static const struct hvs_format *vc4_get_hvs_format(u32 drm_format) -- GitLab From 7b30255698edc91f9235faf586d21625ca7bbbac Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 20 Nov 2017 13:29:58 +0000 Subject: [PATCH 0375/1471] drm/i915/gvt: Cleanup unwanted public symbols drivers/gpu/drm/i915/gvt/execlist.c:531:6: warning: symbol 'clean_execlist' was not declared. Should it be static? drivers/gpu/drm/i915/gvt/execlist.c:545:6: warning: symbol 'reset_execlist' was not declared. Should it be static? drivers/gpu/drm/i915/gvt/execlist.c:556:5: warning: symbol 'init_execlist' was not declared. Should it be static? drivers/gpu/drm/i915/gvt/scheduler.c:248:6: warning: symbol 'release_shadow_wa_ctx' was not declared. Should it be static? References: 06bb372f9ace ("drm/i915/gvt: Introduce intel_vgpu_reset_submission") Signed-off-by: Chris Wilson Cc: Zhi Wang Cc: Zhenyu Wang Cc: intel-gvt-dev@lists.freedesktop.org Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/execlist.c | 6 +++--- drivers/gpu/drm/i915/gvt/scheduler.c | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c index c9fa0fb488d3a..fa4929584744a 100644 --- a/drivers/gpu/drm/i915/gvt/execlist.c +++ b/drivers/gpu/drm/i915/gvt/execlist.c @@ -528,7 +528,7 @@ static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id) vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw; } -void clean_execlist(struct intel_vgpu *vgpu) +static void clean_execlist(struct intel_vgpu *vgpu) { enum intel_engine_id i; struct intel_engine_cs *engine; @@ -542,7 +542,7 @@ void clean_execlist(struct intel_vgpu *vgpu) } } -void reset_execlist(struct intel_vgpu *vgpu, +static void reset_execlist(struct intel_vgpu *vgpu, unsigned long engine_mask) { struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv; @@ -553,7 +553,7 @@ void reset_execlist(struct intel_vgpu *vgpu, init_vgpu_execlist(vgpu, engine->id); } -int init_execlist(struct intel_vgpu *vgpu) +static int init_execlist(struct intel_vgpu *vgpu) { reset_execlist(vgpu, ALL_ENGINES); return 0; diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c index a742b364c2c3e..13ccc00f0d408 100644 --- a/drivers/gpu/drm/i915/gvt/scheduler.c +++ b/drivers/gpu/drm/i915/gvt/scheduler.c @@ -245,7 +245,7 @@ static int copy_workload_to_ring_buffer(struct intel_vgpu_workload *workload) return 0; } -void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) +static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx) { if (!wa_ctx->indirect_ctx.obj) return; -- GitLab From c83ecfa5851f4d35be88f32dabb3a53f51cf5c32 Mon Sep 17 00:00:00 2001 From: Arvind Yadav Date: Fri, 17 Nov 2017 16:06:06 +0530 Subject: [PATCH 0376/1471] drm/sti: Handle return value of platform_get_irq_byname platform_get_irq_byname() can fail here and we must check its return value. Signed-off-by: Arvind Yadav Signed-off-by: Benjamin Gaignard Link: https://patchwork.freedesktop.org/patch/msgid/f4b4f1b2cfa5e302ef7ffad4e3efb0d3147709d3.1510914877.git.arvind.yadav.cs@gmail.com --- drivers/gpu/drm/sti/sti_hdmi.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index d1902750a85da..1c7423f4d423e 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -1413,6 +1413,11 @@ static int sti_hdmi_probe(struct platform_device *pdev) init_waitqueue_head(&hdmi->wait_event); hdmi->irq = platform_get_irq_byname(pdev, "irq"); + if (hdmi->irq < 0) { + DRM_ERROR("Cannot get HDMI irq\n"); + ret = hdmi->irq; + goto release_adapter; + } ret = devm_request_threaded_irq(dev, hdmi->irq, hdmi_irq, hdmi_irq_thread, IRQF_ONESHOT, dev_name(dev), hdmi); -- GitLab From 03e4e0a9e02cf703da331ff6cfd57d0be9bf5692 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 14 Nov 2017 16:27:19 +0000 Subject: [PATCH 0377/1471] dma-buf/fence: Fix lock inversion within dma-fence-array MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Ages ago Rob Clark noted, "Currently with fence-array, we have a potential deadlock situation. If we fence_add_callback() on an array-fence, the array-fence's lock is acquired first, and in it's ->enable_signaling() callback, it will install cbs on it's array-member fences, so the array-member's lock is acquired second. But in the signal path, the array-member's lock is acquired first, and the array-fence's lock acquired second." Rob proposed either extensive changes to dma-fence to unnest the fence-array signaling, or to defer the signaling onto a workqueue. This is a more refined version of the later, that should keep the latency of the fence signaling to a minimum by using an irq-work, which is executed asap. Reported-by: Rob Clark Suggested-by: Rob Clark References: 1476635975-21981-1-git-send-email-robdclark@gmail.com Signed-off-by: Chris Wilson Cc: Rob Clark Cc: Gustavo Padovan Cc: Sumit Semwal Cc: Christian König Reviewed-by: Christian König Signed-off-by: Sumit Semwal Link: https://patchwork.freedesktop.org/patch/msgid/20171114162719.30958-1-chris@chris-wilson.co.uk --- drivers/base/Kconfig | 1 + drivers/dma-buf/dma-fence-array.c | 14 ++++++++++++-- include/linux/dma-fence-array.h | 3 +++ 3 files changed, 16 insertions(+), 2 deletions(-) diff --git a/drivers/base/Kconfig b/drivers/base/Kconfig index 1a5f6a157a57d..62b0de06836e1 100644 --- a/drivers/base/Kconfig +++ b/drivers/base/Kconfig @@ -244,6 +244,7 @@ config DMA_SHARED_BUFFER bool default n select ANON_INODES + select IRQ_WORK help This option enables the framework for buffer-sharing between multiple drivers. A buffer is associated with a file using driver diff --git a/drivers/dma-buf/dma-fence-array.c b/drivers/dma-buf/dma-fence-array.c index 0350829ba62e7..dd1edfb27b61a 100644 --- a/drivers/dma-buf/dma-fence-array.c +++ b/drivers/dma-buf/dma-fence-array.c @@ -31,6 +31,14 @@ static const char *dma_fence_array_get_timeline_name(struct dma_fence *fence) return "unbound"; } +static void irq_dma_fence_array_work(struct irq_work *wrk) +{ + struct dma_fence_array *array = container_of(wrk, typeof(*array), work); + + dma_fence_signal(&array->base); + dma_fence_put(&array->base); +} + static void dma_fence_array_cb_func(struct dma_fence *f, struct dma_fence_cb *cb) { @@ -39,8 +47,9 @@ static void dma_fence_array_cb_func(struct dma_fence *f, struct dma_fence_array *array = array_cb->array; if (atomic_dec_and_test(&array->num_pending)) - dma_fence_signal(&array->base); - dma_fence_put(&array->base); + irq_work_queue(&array->work); + else + dma_fence_put(&array->base); } static bool dma_fence_array_enable_signaling(struct dma_fence *fence) @@ -136,6 +145,7 @@ struct dma_fence_array *dma_fence_array_create(int num_fences, spin_lock_init(&array->lock); dma_fence_init(&array->base, &dma_fence_array_ops, &array->lock, context, seqno); + init_irq_work(&array->work, irq_dma_fence_array_work); array->num_fences = num_fences; atomic_set(&array->num_pending, signal_on_any ? 1 : num_fences); diff --git a/include/linux/dma-fence-array.h b/include/linux/dma-fence-array.h index 332a5420243c4..bc8940ca280dc 100644 --- a/include/linux/dma-fence-array.h +++ b/include/linux/dma-fence-array.h @@ -21,6 +21,7 @@ #define __LINUX_DMA_FENCE_ARRAY_H #include +#include /** * struct dma_fence_array_cb - callback helper for fence array @@ -47,6 +48,8 @@ struct dma_fence_array { unsigned num_fences; atomic_t num_pending; struct dma_fence **fences; + + struct irq_work work; }; extern const struct dma_fence_ops dma_fence_array_ops; -- GitLab From 51f5a096398433a881e845d3685a2c1dac756019 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:08 +0200 Subject: [PATCH 0378/1471] drm/i915: Add .get_hw_state() method for planes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add a .get_hw_state() method for planes, returning true or false depending on whether the plane is enabled. Use it to rewrite the plane enabled/disabled asserts in platform agnostic fashion. We do lose the pre-gen4 plane<->pipe mapping checks, but since we're supposed sanitize that anyway it doesn't really matter. v2: Reoder patches to not depend on enum old_plane_id Just call assert_plane_disabled() from assert_planes_disabled() v3: Deal with disabled power wells in .get_hw_state() v4: Rebase due skl primary plane code removal Cc: Thierry Reding Cc: Alex Villacís Lasso Reviewed-by: Daniel Vetter #v2 Tested-by: Thierry Reding #v2 Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-2-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 188 +++++++++++++-------------- drivers/gpu/drm/i915/intel_drv.h | 2 + drivers/gpu/drm/i915/intel_sprite.c | 83 ++++++++++++ 3 files changed, 175 insertions(+), 98 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 3b3dec1e6640d..7bf8290f0343d 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1190,23 +1190,6 @@ void assert_panel_unlocked(struct drm_i915_private *dev_priv, enum pipe pipe) pipe_name(pipe)); } -static void assert_cursor(struct drm_i915_private *dev_priv, - enum pipe pipe, bool state) -{ - bool cur_state; - - if (IS_I845G(dev_priv) || IS_I865G(dev_priv)) - cur_state = I915_READ(CURCNTR(PIPE_A)) & CURSOR_ENABLE; - else - cur_state = I915_READ(CURCNTR(pipe)) & CURSOR_MODE; - - I915_STATE_WARN(cur_state != state, - "cursor on pipe %c assertion failure (expected %s, current %s)\n", - pipe_name(pipe), onoff(state), onoff(cur_state)); -} -#define assert_cursor_enabled(d, p) assert_cursor(d, p, true) -#define assert_cursor_disabled(d, p) assert_cursor(d, p, false) - void assert_pipe(struct drm_i915_private *dev_priv, enum pipe pipe, bool state) { @@ -1234,77 +1217,25 @@ void assert_pipe(struct drm_i915_private *dev_priv, pipe_name(pipe), onoff(state), onoff(cur_state)); } -static void assert_plane(struct drm_i915_private *dev_priv, - enum plane plane, bool state) +static void assert_plane(struct intel_plane *plane, bool state) { - u32 val; - bool cur_state; + bool cur_state = plane->get_hw_state(plane); - val = I915_READ(DSPCNTR(plane)); - cur_state = !!(val & DISPLAY_PLANE_ENABLE); I915_STATE_WARN(cur_state != state, - "plane %c assertion failure (expected %s, current %s)\n", - plane_name(plane), onoff(state), onoff(cur_state)); + "%s assertion failure (expected %s, current %s)\n", + plane->base.name, onoff(state), onoff(cur_state)); } -#define assert_plane_enabled(d, p) assert_plane(d, p, true) -#define assert_plane_disabled(d, p) assert_plane(d, p, false) +#define assert_plane_enabled(p) assert_plane(p, true) +#define assert_plane_disabled(p) assert_plane(p, false) -static void assert_planes_disabled(struct drm_i915_private *dev_priv, - enum pipe pipe) +static void assert_planes_disabled(struct intel_crtc *crtc) { - int i; - - /* Primary planes are fixed to pipes on gen4+ */ - if (INTEL_GEN(dev_priv) >= 4) { - u32 val = I915_READ(DSPCNTR(pipe)); - I915_STATE_WARN(val & DISPLAY_PLANE_ENABLE, - "plane %c assertion failure, should be disabled but not\n", - plane_name(pipe)); - return; - } - - /* Need to check both planes against the pipe */ - for_each_pipe(dev_priv, i) { - u32 val = I915_READ(DSPCNTR(i)); - enum pipe cur_pipe = (val & DISPPLANE_SEL_PIPE_MASK) >> - DISPPLANE_SEL_PIPE_SHIFT; - I915_STATE_WARN((val & DISPLAY_PLANE_ENABLE) && pipe == cur_pipe, - "plane %c assertion failure, should be off on pipe %c but is still active\n", - plane_name(i), pipe_name(pipe)); - } -} - -static void assert_sprites_disabled(struct drm_i915_private *dev_priv, - enum pipe pipe) -{ - int sprite; + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_plane *plane; - if (INTEL_GEN(dev_priv) >= 9) { - for_each_sprite(dev_priv, pipe, sprite) { - u32 val = I915_READ(PLANE_CTL(pipe, sprite)); - I915_STATE_WARN(val & PLANE_CTL_ENABLE, - "plane %d assertion failure, should be off on pipe %c but is still active\n", - sprite, pipe_name(pipe)); - } - } else if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) { - for_each_sprite(dev_priv, pipe, sprite) { - u32 val = I915_READ(SPCNTR(pipe, PLANE_SPRITE0 + sprite)); - I915_STATE_WARN(val & SP_ENABLE, - "sprite %c assertion failure, should be off on pipe %c but is still active\n", - sprite_name(pipe, sprite), pipe_name(pipe)); - } - } else if (INTEL_GEN(dev_priv) >= 7) { - u32 val = I915_READ(SPRCTL(pipe)); - I915_STATE_WARN(val & SPRITE_ENABLE, - "sprite %c assertion failure, should be off on pipe %c but is still active\n", - plane_name(pipe), pipe_name(pipe)); - } else if (INTEL_GEN(dev_priv) >= 5 || IS_G4X(dev_priv)) { - u32 val = I915_READ(DVSCNTR(pipe)); - I915_STATE_WARN(val & DVS_ENABLE, - "sprite %c assertion failure, should be off on pipe %c but is still active\n", - plane_name(pipe), pipe_name(pipe)); - } + for_each_intel_plane_on_crtc(&dev_priv->drm, crtc, plane) + assert_plane_disabled(plane); } static void assert_vblank_disabled(struct drm_crtc *crtc) @@ -1896,9 +1827,7 @@ static void intel_enable_pipe(struct intel_crtc *crtc) DRM_DEBUG_KMS("enabling pipe %c\n", pipe_name(pipe)); - assert_planes_disabled(dev_priv, pipe); - assert_cursor_disabled(dev_priv, pipe); - assert_sprites_disabled(dev_priv, pipe); + assert_planes_disabled(crtc); /* * A pipe without a PLL won't actually be able to drive bits from @@ -1968,9 +1897,7 @@ static void intel_disable_pipe(struct intel_crtc *crtc) * Make sure planes won't keep trying to pump pixels to us, * or we might hang the display. */ - assert_planes_disabled(dev_priv, pipe); - assert_cursor_disabled(dev_priv, pipe); - assert_sprites_disabled(dev_priv, pipe); + assert_planes_disabled(crtc); reg = PIPECONF(cpu_transcoder); val = I915_READ(reg); @@ -3364,6 +3291,31 @@ static void i9xx_disable_primary_plane(struct intel_plane *primary, spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +static bool i9xx_plane_get_hw_state(struct intel_plane *primary) +{ + + struct drm_i915_private *dev_priv = to_i915(primary->base.dev); + enum intel_display_power_domain power_domain; + enum plane plane = primary->plane; + enum pipe pipe = primary->pipe; + bool ret; + + /* + * Not 100% correct for planes that can move between pipes, + * but that's only the case for gen2-4 which don't have any + * display power wells. + */ + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(DSPCNTR(plane)) & DISPLAY_PLANE_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static u32 intel_fb_stride_alignment(const struct drm_framebuffer *fb, int plane) { @@ -4879,7 +4831,8 @@ void hsw_enable_ips(const struct intel_crtc_state *crtc_state) * a vblank wait. */ - assert_plane_enabled(dev_priv, crtc->plane); + assert_plane_enabled(to_intel_plane(crtc->base.primary)); + if (IS_BROADWELL(dev_priv)) { mutex_lock(&dev_priv->pcu_lock); WARN_ON(sandybridge_pcode_write(dev_priv, DISPLAY_IPS_CONTROL, @@ -4913,7 +4866,8 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state) if (!crtc_state->ips_enabled) return; - assert_plane_enabled(dev_priv, crtc->plane); + assert_plane_enabled(to_intel_plane(crtc->base.primary)); + if (IS_BROADWELL(dev_priv)) { mutex_lock(&dev_priv->pcu_lock); WARN_ON(sandybridge_pcode_write(dev_priv, DISPLAY_IPS_CONTROL, 0)); @@ -9499,6 +9453,23 @@ static void i845_disable_cursor(struct intel_plane *plane, i845_update_cursor(plane, NULL, NULL); } +static bool i845_cursor_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + bool ret; + + power_domain = POWER_DOMAIN_PIPE(PIPE_A); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(CURCNTR(PIPE_A)) & CURSOR_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static u32 i9xx_cursor_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state) { @@ -9692,6 +9663,28 @@ static void i9xx_disable_cursor(struct intel_plane *plane, i9xx_update_cursor(plane, NULL, NULL); } +static bool i9xx_cursor_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + enum pipe pipe = plane->pipe; + bool ret; + + /* + * Not 100% correct for planes that can move between pipes, + * but that's only the case for gen2-3 which don't have any + * display power wells. + */ + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(CURCNTR(pipe)) & CURSOR_MODE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} /* VESA 640x480x72Hz mode to set on the pipe */ static const struct drm_display_mode load_detect_mode = { @@ -13279,6 +13272,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->update_plane = skl_update_plane; primary->disable_plane = skl_disable_plane; + primary->get_hw_state = skl_plane_get_hw_state; } else if (INTEL_GEN(dev_priv) >= 9) { intel_primary_formats = skl_primary_formats; num_formats = ARRAY_SIZE(skl_primary_formats); @@ -13289,6 +13283,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->update_plane = skl_update_plane; primary->disable_plane = skl_disable_plane; + primary->get_hw_state = skl_plane_get_hw_state; } else if (INTEL_GEN(dev_priv) >= 4) { intel_primary_formats = i965_primary_formats; num_formats = ARRAY_SIZE(i965_primary_formats); @@ -13296,6 +13291,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->update_plane = i9xx_update_primary_plane; primary->disable_plane = i9xx_disable_primary_plane; + primary->get_hw_state = i9xx_plane_get_hw_state; } else { intel_primary_formats = i8xx_primary_formats; num_formats = ARRAY_SIZE(i8xx_primary_formats); @@ -13303,6 +13299,7 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) primary->update_plane = i9xx_update_primary_plane; primary->disable_plane = i9xx_disable_primary_plane; + primary->get_hw_state = i9xx_plane_get_hw_state; } if (INTEL_GEN(dev_priv) >= 9) @@ -13392,10 +13389,12 @@ intel_cursor_plane_create(struct drm_i915_private *dev_priv, if (IS_I845G(dev_priv) || IS_I865G(dev_priv)) { cursor->update_plane = i845_update_cursor; cursor->disable_plane = i845_disable_cursor; + cursor->get_hw_state = i845_cursor_get_hw_state; cursor->check_plane = i845_check_cursor; } else { cursor->update_plane = i9xx_update_cursor; cursor->disable_plane = i9xx_disable_cursor; + cursor->get_hw_state = i9xx_cursor_get_hw_state; cursor->check_plane = i9xx_check_cursor; } @@ -14761,8 +14760,8 @@ void i830_disable_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) DRM_DEBUG_KMS("disabling pipe %c due to force quirk\n", pipe_name(pipe)); - assert_plane_disabled(dev_priv, PLANE_A); - assert_plane_disabled(dev_priv, PLANE_B); + assert_planes_disabled(intel_get_crtc_for_pipe(dev_priv, PIPE_A)); + assert_planes_disabled(intel_get_crtc_for_pipe(dev_priv, PIPE_B)); I915_WRITE(PIPECONF(pipe), 0); POSTING_READ(PIPECONF(pipe)); @@ -14976,20 +14975,13 @@ void i915_redisable_vga(struct drm_i915_private *dev_priv) intel_display_power_put(dev_priv, POWER_DOMAIN_VGA); } -static bool primary_get_hw_state(struct intel_plane *plane) -{ - struct drm_i915_private *dev_priv = to_i915(plane->base.dev); - - return I915_READ(DSPCNTR(plane->plane)) & DISPLAY_PLANE_ENABLE; -} - /* FIXME read out full plane state for all planes */ static void readout_plane_state(struct intel_crtc *crtc) { struct intel_plane *primary = to_intel_plane(crtc->base.primary); bool visible; - visible = crtc->active && primary_get_hw_state(primary); + visible = crtc->active && primary->get_hw_state(primary); intel_set_plane_visible(to_intel_crtc_state(crtc->base.state), to_intel_plane_state(primary->base.state), diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 69aab324aaa1a..99840f3940c7e 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -866,6 +866,7 @@ struct intel_plane { const struct intel_plane_state *plane_state); void (*disable_plane)(struct intel_plane *plane, struct intel_crtc *crtc); + bool (*get_hw_state)(struct intel_plane *plane); int (*check_plane)(struct intel_plane *plane, struct intel_crtc_state *crtc_state, struct intel_plane_state *state); @@ -1934,6 +1935,7 @@ void skl_update_plane(struct intel_plane *plane, const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state); void skl_disable_plane(struct intel_plane *plane, struct intel_crtc *crtc); +bool skl_plane_get_hw_state(struct intel_plane *plane); /* intel_tv.c */ void intel_tv_init(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index ce615704982ae..5baa0023964ef 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -325,6 +325,26 @@ skl_disable_plane(struct intel_plane *plane, struct intel_crtc *crtc) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +bool +skl_plane_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + enum plane_id plane_id = plane->id; + enum pipe pipe = plane->pipe; + bool ret; + + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(PLANE_CTL(pipe, plane_id)) & PLANE_CTL_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static void chv_update_csc(struct intel_plane *plane, uint32_t format) { @@ -502,6 +522,26 @@ vlv_disable_plane(struct intel_plane *plane, struct intel_crtc *crtc) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +static bool +vlv_plane_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + enum plane_id plane_id = plane->id; + enum pipe pipe = plane->pipe; + bool ret; + + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(SPCNTR(pipe, plane_id)) & SP_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static u32 ivb_sprite_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state) { @@ -642,6 +682,25 @@ ivb_disable_plane(struct intel_plane *plane, struct intel_crtc *crtc) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +static bool +ivb_plane_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + enum pipe pipe = plane->pipe; + bool ret; + + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(SPRCTL(pipe)) & SPRITE_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static u32 g4x_sprite_ctl(const struct intel_crtc_state *crtc_state, const struct intel_plane_state *plane_state) { @@ -773,6 +832,25 @@ g4x_disable_plane(struct intel_plane *plane, struct intel_crtc *crtc) spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } +static bool +g4x_plane_get_hw_state(struct intel_plane *plane) +{ + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum intel_display_power_domain power_domain; + enum pipe pipe = plane->pipe; + bool ret; + + power_domain = POWER_DOMAIN_PIPE(pipe); + if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) + return false; + + ret = I915_READ(DVSCNTR(pipe)) & DVS_ENABLE; + + intel_display_power_put(dev_priv, power_domain); + + return ret; +} + static int intel_check_sprite_plane(struct intel_plane *plane, struct intel_crtc_state *crtc_state, @@ -1231,6 +1309,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, intel_plane->update_plane = skl_update_plane; intel_plane->disable_plane = skl_disable_plane; + intel_plane->get_hw_state = skl_plane_get_hw_state; plane_formats = skl_plane_formats; num_plane_formats = ARRAY_SIZE(skl_plane_formats); @@ -1241,6 +1320,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, intel_plane->update_plane = skl_update_plane; intel_plane->disable_plane = skl_disable_plane; + intel_plane->get_hw_state = skl_plane_get_hw_state; plane_formats = skl_plane_formats; num_plane_formats = ARRAY_SIZE(skl_plane_formats); @@ -1251,6 +1331,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, intel_plane->update_plane = vlv_update_plane; intel_plane->disable_plane = vlv_disable_plane; + intel_plane->get_hw_state = vlv_plane_get_hw_state; plane_formats = vlv_plane_formats; num_plane_formats = ARRAY_SIZE(vlv_plane_formats); @@ -1266,6 +1347,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, intel_plane->update_plane = ivb_update_plane; intel_plane->disable_plane = ivb_disable_plane; + intel_plane->get_hw_state = ivb_plane_get_hw_state; plane_formats = snb_plane_formats; num_plane_formats = ARRAY_SIZE(snb_plane_formats); @@ -1276,6 +1358,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, intel_plane->update_plane = g4x_update_plane; intel_plane->disable_plane = g4x_disable_plane; + intel_plane->get_hw_state = g4x_plane_get_hw_state; modifiers = i9xx_plane_format_modifiers; if (IS_GEN6(dev_priv)) { -- GitLab From b1e01595a66dc206a2c75401ec4c285740537f3f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:09 +0200 Subject: [PATCH 0379/1471] drm/i915: Redo plane sanitation during readout MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Unify the plane disabling during state readout by pulling the code into a new helper intel_plane_disable_noatomic(). We'll also read out the state of all planes, so that we know which planes really need to be diabled. Additonally we change the plane<->pipe mapping sanitation to work by simply disabling the offending planes instead of entire pipes. And we do it before we otherwise sanitize the crtcs, which means we don't have to worry about misassigned planes during crtc sanitation anymore. v2: Reoder patches to not depend on enum old_plane_id v3: s/for_each_pipe/for_each_intel_crtc/ Cc: Thierry Reding Cc: Alex Villacís Lasso Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103223 Reviewed-by: Daniel Vetter Tested-by: Thierry Reding Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-3-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 114 +++++++++++++++------------ 1 file changed, 65 insertions(+), 49 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 7bf8290f0343d..91f74c5373b3b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2726,6 +2726,23 @@ intel_set_plane_visible(struct intel_crtc_state *crtc_state, crtc_state->active_planes); } +static void intel_plane_disable_noatomic(struct intel_crtc *crtc, + struct intel_plane *plane) +{ + struct intel_crtc_state *crtc_state = + to_intel_crtc_state(crtc->base.state); + struct intel_plane_state *plane_state = + to_intel_plane_state(plane->base.state); + + intel_set_plane_visible(crtc_state, plane_state, false); + + if (plane->id == PLANE_PRIMARY) + intel_pre_disable_primary_noatomic(&crtc->base); + + trace_intel_disable_plane(&plane->base, crtc); + plane->disable_plane(plane, crtc); +} + static void intel_find_initial_plane_obj(struct intel_crtc *intel_crtc, struct intel_initial_plane_config *plane_config) @@ -2783,12 +2800,7 @@ intel_find_initial_plane_obj(struct intel_crtc *intel_crtc, * simplest solution is to just disable the primary plane now and * pretend the BIOS never had it enabled. */ - intel_set_plane_visible(to_intel_crtc_state(crtc_state), - to_intel_plane_state(plane_state), - false); - intel_pre_disable_primary_noatomic(&intel_crtc->base); - trace_intel_disable_plane(primary, intel_crtc); - intel_plane->disable_plane(intel_plane, intel_crtc); + intel_plane_disable_noatomic(intel_crtc, intel_plane); return; @@ -5869,6 +5881,7 @@ static void intel_crtc_disable_noatomic(struct drm_crtc *crtc, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); struct drm_i915_private *dev_priv = to_i915(crtc->dev); enum intel_display_power_domain domain; + struct intel_plane *plane; u64 domains; struct drm_atomic_state *state; struct intel_crtc_state *crtc_state; @@ -5877,11 +5890,12 @@ static void intel_crtc_disable_noatomic(struct drm_crtc *crtc, if (!intel_crtc->active) return; - if (crtc->primary->state->visible) { - intel_pre_disable_primary_noatomic(crtc); + for_each_intel_plane_on_crtc(&dev_priv->drm, intel_crtc, plane) { + const struct intel_plane_state *plane_state = + to_intel_plane_state(plane->base.state); - intel_crtc_disable_planes(crtc, 1 << drm_plane_index(crtc->primary)); - crtc->primary->state->visible = false; + if (plane_state->base.visible) + intel_plane_disable_noatomic(intel_crtc, plane); } state = drm_atomic_state_alloc(crtc->dev); @@ -14773,22 +14787,36 @@ void i830_disable_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) POSTING_READ(DPLL(pipe)); } -static bool -intel_check_plane_mapping(struct intel_crtc *crtc) +static bool intel_plane_mapping_ok(struct intel_crtc *crtc, + struct intel_plane *primary) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - u32 val; + enum plane plane = primary->plane; + u32 val = I915_READ(DSPCNTR(plane)); - if (INTEL_INFO(dev_priv)->num_pipes == 1) - return true; + return (val & DISPLAY_PLANE_ENABLE) == 0 || + (val & DISPPLANE_SEL_PIPE_MASK) == DISPPLANE_SEL_PIPE(crtc->pipe); +} - val = I915_READ(DSPCNTR(!crtc->plane)); +static void +intel_sanitize_plane_mapping(struct drm_i915_private *dev_priv) +{ + struct intel_crtc *crtc; - if ((val & DISPLAY_PLANE_ENABLE) && - (!!(val & DISPPLANE_SEL_PIPE_MASK) == crtc->pipe)) - return false; + if (INTEL_GEN(dev_priv) >= 4) + return; - return true; + for_each_intel_crtc(&dev_priv->drm, crtc) { + struct intel_plane *plane = + to_intel_plane(crtc->base.primary); + + if (intel_plane_mapping_ok(crtc, plane)) + continue; + + DRM_DEBUG_KMS("%s attached to the wrong pipe, disabling plane\n", + plane->base.name); + intel_plane_disable_noatomic(crtc, plane); + } } static bool intel_crtc_has_encoders(struct intel_crtc *crtc) @@ -14844,33 +14872,15 @@ static void intel_sanitize_crtc(struct intel_crtc *crtc, /* Disable everything but the primary plane */ for_each_intel_plane_on_crtc(dev, crtc, plane) { - if (plane->base.type == DRM_PLANE_TYPE_PRIMARY) - continue; + const struct intel_plane_state *plane_state = + to_intel_plane_state(plane->base.state); - trace_intel_disable_plane(&plane->base, crtc); - plane->disable_plane(plane, crtc); + if (plane_state->base.visible && + plane->base.type != DRM_PLANE_TYPE_PRIMARY) + intel_plane_disable_noatomic(crtc, plane); } } - /* We need to sanitize the plane -> pipe mapping first because this will - * disable the crtc (and hence change the state) if it is wrong. Note - * that gen4+ has a fixed plane -> pipe mapping. */ - if (INTEL_GEN(dev_priv) < 4 && !intel_check_plane_mapping(crtc)) { - bool plane; - - DRM_DEBUG_KMS("[CRTC:%d:%s] wrong plane connection detected!\n", - crtc->base.base.id, crtc->base.name); - - /* Pipe has the wrong plane attached and the plane is active. - * Temporarily change the plane mapping and disable everything - * ... */ - plane = crtc->plane; - crtc->base.primary->state->visible = true; - crtc->plane = !plane; - intel_crtc_disable_noatomic(&crtc->base, ctx); - crtc->plane = plane; - } - /* Adjust the state of the output pipe according to whether we * have active connectors/encoders. */ if (crtc->active && !intel_crtc_has_encoders(crtc)) @@ -14978,14 +14988,18 @@ void i915_redisable_vga(struct drm_i915_private *dev_priv) /* FIXME read out full plane state for all planes */ static void readout_plane_state(struct intel_crtc *crtc) { - struct intel_plane *primary = to_intel_plane(crtc->base.primary); - bool visible; + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + struct intel_crtc_state *crtc_state = + to_intel_crtc_state(crtc->base.state); + struct intel_plane *plane; - visible = crtc->active && primary->get_hw_state(primary); + for_each_intel_plane_on_crtc(&dev_priv->drm, crtc, plane) { + struct intel_plane_state *plane_state = + to_intel_plane_state(plane->base.state); + bool visible = plane->get_hw_state(plane); - intel_set_plane_visible(to_intel_crtc_state(crtc->base.state), - to_intel_plane_state(primary->base.state), - visible); + intel_set_plane_visible(crtc_state, plane_state, visible); + } } static void intel_modeset_readout_hw_state(struct drm_device *dev) @@ -15193,6 +15207,8 @@ intel_modeset_setup_hw_state(struct drm_device *dev, /* HW state is read out, now we need to sanitize this mess. */ get_encoder_power_domains(dev_priv); + intel_sanitize_plane_mapping(dev_priv); + for_each_intel_encoder(dev, encoder) { intel_sanitize_encoder(encoder); } -- GitLab From ed15030d7ab0cfe395a494f03e5ee2e66ccb8be9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:10 +0200 Subject: [PATCH 0380/1471] drm/i915: s/enum plane/enum i9xx_plane_id/ MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Rename enum plane to enum i9xx_plane_id to make it clear that it only applies to pre-SKL platforms. enum i9xx_plane_id is a global identifier, whereas enum plane_id is per-pipe. We need the old global identifier to index the primary plane (and the pre-g4x sprite C if we ever expose it) registers on pre-SKL platforms. v2: Reorder patches v3: s/old_plane_id/i9xx_plane_id/ (Daniel) Pimp the commit message a bit Note that i9xx_plane_id doesn't apply to SKL+ v4: Rebase due to power domain handling in plane readout v5: Rebase due to crtc->dspaddr_offset removal v6: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-4-ville.syrjala@linux.intel.com Reviewed-by: James Ausmus Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.h | 6 +- drivers/gpu/drm/i915/intel_display.c | 98 ++++++++++++++-------------- drivers/gpu/drm/i915/intel_drv.h | 6 +- drivers/gpu/drm/i915/intel_fbc.c | 12 ++-- drivers/gpu/drm/i915/intel_sprite.c | 2 +- 5 files changed, 62 insertions(+), 62 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 9a0654e10a554..d63a731ef60bf 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -304,9 +304,9 @@ static inline bool transcoder_is_dsi(enum transcoder transcoder) /* * Global legacy plane identifier. Valid only for primary/sprite - * planes on pre-g4x, and only for primary planes on g4x+. + * planes on pre-g4x, and only for primary planes on g4x-bdw. */ -enum plane { +enum i9xx_plane_id { PLANE_A, PLANE_B, PLANE_C, @@ -1144,7 +1144,7 @@ struct intel_fbc { struct { enum pipe pipe; - enum plane plane; + enum i9xx_plane_id i9xx_plane; unsigned int fence_y_offset; } crtc; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 91f74c5373b3b..16ac86816f288 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -3220,16 +3220,16 @@ int i9xx_check_plane_surface(struct intel_plane_state *plane_state) return 0; } -static void i9xx_update_primary_plane(struct intel_plane *primary, - const struct intel_crtc_state *crtc_state, - const struct intel_plane_state *plane_state) +static void i9xx_update_plane(struct intel_plane *plane, + const struct intel_crtc_state *crtc_state, + const struct intel_plane_state *plane_state) { - struct drm_i915_private *dev_priv = to_i915(primary->base.dev); + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); const struct drm_framebuffer *fb = plane_state->base.fb; - enum plane plane = primary->plane; + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; u32 linear_offset; u32 dspcntr = plane_state->ctl; - i915_reg_t reg = DSPCNTR(plane); + i915_reg_t reg = DSPCNTR(i9xx_plane); int x = plane_state->main.x; int y = plane_state->main.y; unsigned long irqflags; @@ -3248,34 +3248,34 @@ static void i9xx_update_primary_plane(struct intel_plane *primary, /* pipesrc and dspsize control the size that is scaled from, * which should always be the user's requested size. */ - I915_WRITE_FW(DSPSIZE(plane), + I915_WRITE_FW(DSPSIZE(i9xx_plane), ((crtc_state->pipe_src_h - 1) << 16) | (crtc_state->pipe_src_w - 1)); - I915_WRITE_FW(DSPPOS(plane), 0); - } else if (IS_CHERRYVIEW(dev_priv) && plane == PLANE_B) { - I915_WRITE_FW(PRIMSIZE(plane), + I915_WRITE_FW(DSPPOS(i9xx_plane), 0); + } else if (IS_CHERRYVIEW(dev_priv) && i9xx_plane == PLANE_B) { + I915_WRITE_FW(PRIMSIZE(i9xx_plane), ((crtc_state->pipe_src_h - 1) << 16) | (crtc_state->pipe_src_w - 1)); - I915_WRITE_FW(PRIMPOS(plane), 0); - I915_WRITE_FW(PRIMCNSTALPHA(plane), 0); + I915_WRITE_FW(PRIMPOS(i9xx_plane), 0); + I915_WRITE_FW(PRIMCNSTALPHA(i9xx_plane), 0); } I915_WRITE_FW(reg, dspcntr); - I915_WRITE_FW(DSPSTRIDE(plane), fb->pitches[0]); + I915_WRITE_FW(DSPSTRIDE(i9xx_plane), fb->pitches[0]); if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) { - I915_WRITE_FW(DSPSURF(plane), + I915_WRITE_FW(DSPSURF(i9xx_plane), intel_plane_ggtt_offset(plane_state) + dspaddr_offset); - I915_WRITE_FW(DSPOFFSET(plane), (y << 16) | x); + I915_WRITE_FW(DSPOFFSET(i9xx_plane), (y << 16) | x); } else if (INTEL_GEN(dev_priv) >= 4) { - I915_WRITE_FW(DSPSURF(plane), + I915_WRITE_FW(DSPSURF(i9xx_plane), intel_plane_ggtt_offset(plane_state) + dspaddr_offset); - I915_WRITE_FW(DSPTILEOFF(plane), (y << 16) | x); - I915_WRITE_FW(DSPLINOFF(plane), linear_offset); + I915_WRITE_FW(DSPTILEOFF(i9xx_plane), (y << 16) | x); + I915_WRITE_FW(DSPLINOFF(i9xx_plane), linear_offset); } else { - I915_WRITE_FW(DSPADDR(plane), + I915_WRITE_FW(DSPADDR(i9xx_plane), intel_plane_ggtt_offset(plane_state) + dspaddr_offset); } @@ -3284,32 +3284,31 @@ static void i9xx_update_primary_plane(struct intel_plane *primary, spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } -static void i9xx_disable_primary_plane(struct intel_plane *primary, - struct intel_crtc *crtc) +static void i9xx_disable_plane(struct intel_plane *plane, + struct intel_crtc *crtc) { - struct drm_i915_private *dev_priv = to_i915(primary->base.dev); - enum plane plane = primary->plane; + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; unsigned long irqflags; spin_lock_irqsave(&dev_priv->uncore.lock, irqflags); - I915_WRITE_FW(DSPCNTR(plane), 0); - if (INTEL_INFO(dev_priv)->gen >= 4) - I915_WRITE_FW(DSPSURF(plane), 0); + I915_WRITE_FW(DSPCNTR(i9xx_plane), 0); + if (INTEL_GEN(dev_priv) >= 4) + I915_WRITE_FW(DSPSURF(i9xx_plane), 0); else - I915_WRITE_FW(DSPADDR(plane), 0); - POSTING_READ_FW(DSPCNTR(plane)); + I915_WRITE_FW(DSPADDR(i9xx_plane), 0); + POSTING_READ_FW(DSPCNTR(i9xx_plane)); spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); } -static bool i9xx_plane_get_hw_state(struct intel_plane *primary) +static bool i9xx_plane_get_hw_state(struct intel_plane *plane) { - - struct drm_i915_private *dev_priv = to_i915(primary->base.dev); + struct drm_i915_private *dev_priv = to_i915(plane->base.dev); enum intel_display_power_domain power_domain; - enum plane plane = primary->plane; - enum pipe pipe = primary->pipe; + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; + enum pipe pipe = plane->pipe; bool ret; /* @@ -3321,7 +3320,7 @@ static bool i9xx_plane_get_hw_state(struct intel_plane *primary) if (!intel_display_power_get_if_enabled(dev_priv, power_domain)) return false; - ret = I915_READ(DSPCNTR(plane)) & DISPLAY_PLANE_ENABLE; + ret = I915_READ(DSPCNTR(i9xx_plane)) & DISPLAY_PLANE_ENABLE; intel_display_power_put(dev_priv, power_domain); @@ -7406,7 +7405,7 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); u32 val, base, offset; - int pipe = crtc->pipe, plane = crtc->plane; + int pipe = crtc->pipe, plane = crtc->i9xx_plane; int fourcc, pixel_format; unsigned int aligned_height; struct drm_framebuffer *fb; @@ -13272,9 +13271,9 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) * port is hooked to pipe B. Hence we want plane A feeding pipe B. */ if (HAS_FBC(dev_priv) && INTEL_GEN(dev_priv) < 4) - primary->plane = (enum plane) !pipe; + primary->i9xx_plane = (enum i9xx_plane_id) !pipe; else - primary->plane = (enum plane) pipe; + primary->i9xx_plane = (enum i9xx_plane_id) pipe; primary->id = PLANE_PRIMARY; primary->frontbuffer_bit = INTEL_FRONTBUFFER_PRIMARY(pipe); primary->check_plane = intel_check_primary_plane; @@ -13303,16 +13302,16 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) num_formats = ARRAY_SIZE(i965_primary_formats); modifiers = i9xx_format_modifiers; - primary->update_plane = i9xx_update_primary_plane; - primary->disable_plane = i9xx_disable_primary_plane; + primary->update_plane = i9xx_update_plane; + primary->disable_plane = i9xx_disable_plane; primary->get_hw_state = i9xx_plane_get_hw_state; } else { intel_primary_formats = i8xx_primary_formats; num_formats = ARRAY_SIZE(i8xx_primary_formats); modifiers = i9xx_format_modifiers; - primary->update_plane = i9xx_update_primary_plane; - primary->disable_plane = i9xx_disable_primary_plane; + primary->update_plane = i9xx_update_plane; + primary->disable_plane = i9xx_disable_plane; primary->get_hw_state = i9xx_plane_get_hw_state; } @@ -13336,7 +13335,8 @@ intel_primary_plane_create(struct drm_i915_private *dev_priv, enum pipe pipe) intel_primary_formats, num_formats, modifiers, DRM_PLANE_TYPE_PRIMARY, - "plane %c", plane_name(primary->plane)); + "plane %c", + plane_name(primary->i9xx_plane)); if (ret) goto fail; @@ -13396,7 +13396,7 @@ intel_cursor_plane_create(struct drm_i915_private *dev_priv, cursor->can_scale = false; cursor->max_downscale = 1; cursor->pipe = pipe; - cursor->plane = pipe; + cursor->i9xx_plane = (enum i9xx_plane_id) pipe; cursor->id = PLANE_CURSOR; cursor->frontbuffer_bit = INTEL_FRONTBUFFER_CURSOR(pipe); @@ -13524,14 +13524,14 @@ static int intel_crtc_init(struct drm_i915_private *dev_priv, enum pipe pipe) goto fail; intel_crtc->pipe = pipe; - intel_crtc->plane = primary->plane; + intel_crtc->i9xx_plane = primary->i9xx_plane; /* initialize shared scalers */ intel_crtc_init_scalers(intel_crtc, crtc_state); BUG_ON(pipe >= ARRAY_SIZE(dev_priv->plane_to_crtc_mapping) || - dev_priv->plane_to_crtc_mapping[intel_crtc->plane] != NULL); - dev_priv->plane_to_crtc_mapping[intel_crtc->plane] = intel_crtc; + dev_priv->plane_to_crtc_mapping[intel_crtc->i9xx_plane] != NULL); + dev_priv->plane_to_crtc_mapping[intel_crtc->i9xx_plane] = intel_crtc; dev_priv->pipe_to_crtc_mapping[intel_crtc->pipe] = intel_crtc; drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs); @@ -14788,11 +14788,11 @@ void i830_disable_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) } static bool intel_plane_mapping_ok(struct intel_crtc *crtc, - struct intel_plane *primary) + struct intel_plane *plane) { struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); - enum plane plane = primary->plane; - u32 val = I915_READ(DSPCNTR(plane)); + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; + u32 val = I915_READ(DSPCNTR(i9xx_plane)); return (val & DISPLAY_PLANE_ENABLE) == 0 || (val & DISPPLANE_SEL_PIPE_MASK) == DISPPLANE_SEL_PIPE(crtc->pipe); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 99840f3940c7e..d1fe7be94b62f 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -799,7 +799,7 @@ struct intel_crtc_state { struct intel_crtc { struct drm_crtc base; enum pipe pipe; - enum plane plane; + enum i9xx_plane_id i9xx_plane; /* * Whether the crtc and the connected output pipeline is active. Implies * that crtc->enabled is set, i.e. the current mode configuration has @@ -844,7 +844,7 @@ struct intel_crtc { struct intel_plane { struct drm_plane base; - u8 plane; + enum i9xx_plane_id i9xx_plane; enum plane_id id; enum pipe pipe; bool can_scale; @@ -1130,7 +1130,7 @@ intel_get_crtc_for_pipe(struct drm_i915_private *dev_priv, enum pipe pipe) } static inline struct intel_crtc * -intel_get_crtc_for_plane(struct drm_i915_private *dev_priv, enum plane plane) +intel_get_crtc_for_plane(struct drm_i915_private *dev_priv, enum i9xx_plane_id plane) { return dev_priv->plane_to_crtc_mapping[plane]; } diff --git a/drivers/gpu/drm/i915/intel_fbc.c b/drivers/gpu/drm/i915/intel_fbc.c index 1a0f5e0c8d105..3133131306a9c 100644 --- a/drivers/gpu/drm/i915/intel_fbc.c +++ b/drivers/gpu/drm/i915/intel_fbc.c @@ -151,7 +151,7 @@ static void i8xx_fbc_activate(struct drm_i915_private *dev_priv) /* Set it up... */ fbc_ctl2 = FBC_CTL_FENCE_DBL | FBC_CTL_IDLE_IMM | FBC_CTL_CPU_FENCE; - fbc_ctl2 |= FBC_CTL_PLANE(params->crtc.plane); + fbc_ctl2 |= FBC_CTL_PLANE(params->crtc.i9xx_plane); I915_WRITE(FBC_CONTROL2, fbc_ctl2); I915_WRITE(FBC_FENCE_OFF, params->crtc.fence_y_offset); } @@ -177,7 +177,7 @@ static void g4x_fbc_activate(struct drm_i915_private *dev_priv) struct intel_fbc_reg_params *params = &dev_priv->fbc.params; u32 dpfc_ctl; - dpfc_ctl = DPFC_CTL_PLANE(params->crtc.plane) | DPFC_SR_EN; + dpfc_ctl = DPFC_CTL_PLANE(params->crtc.i9xx_plane) | DPFC_SR_EN; if (params->fb.format->cpp[0] == 2) dpfc_ctl |= DPFC_CTL_LIMIT_2X; else @@ -224,7 +224,7 @@ static void ilk_fbc_activate(struct drm_i915_private *dev_priv) u32 dpfc_ctl; int threshold = dev_priv->fbc.threshold; - dpfc_ctl = DPFC_CTL_PLANE(params->crtc.plane); + dpfc_ctl = DPFC_CTL_PLANE(params->crtc.i9xx_plane); if (params->fb.format->cpp[0] == 2) threshold++; @@ -306,7 +306,7 @@ static void gen7_fbc_activate(struct drm_i915_private *dev_priv) dpfc_ctl = 0; if (IS_IVYBRIDGE(dev_priv)) - dpfc_ctl |= IVB_DPFC_CTL_PLANE(params->crtc.plane); + dpfc_ctl |= IVB_DPFC_CTL_PLANE(params->crtc.i9xx_plane); if (params->fb.format->cpp[0] == 2) threshold++; @@ -890,7 +890,7 @@ static void intel_fbc_get_reg_params(struct intel_crtc *crtc, params->vma = cache->vma; params->crtc.pipe = crtc->pipe; - params->crtc.plane = crtc->plane; + params->crtc.i9xx_plane = crtc->i9xx_plane; params->crtc.fence_y_offset = get_crtc_fence_y_offset(fbc); params->fb.format = cache->fb.format; @@ -1088,7 +1088,7 @@ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, if (fbc_on_pipe_a_only(dev_priv) && crtc->pipe != PIPE_A) continue; - if (fbc_on_plane_a_only(dev_priv) && crtc->plane != PLANE_A) + if (fbc_on_plane_a_only(dev_priv) && crtc->i9xx_plane != PLANE_A) continue; intel_crtc_state = to_intel_crtc_state( diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 5baa0023964ef..dd485f59eb1dc 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -1384,7 +1384,7 @@ intel_sprite_plane_create(struct drm_i915_private *dev_priv, } intel_plane->pipe = pipe; - intel_plane->plane = plane; + intel_plane->i9xx_plane = plane; intel_plane->id = PLANE_SPRITE0 + plane; intel_plane->frontbuffer_bit = INTEL_FRONTBUFFER_SPRITE(pipe, plane); intel_plane->check_plane = intel_check_sprite_plane; -- GitLab From bdaf8439ba8176908b1b33e7db724f3de35ee0f3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:11 +0200 Subject: [PATCH 0381/1471] drm/i915: Use enum i9xx_plane_id for the .get_fifo_size() hooks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace the 0 and 1 with PLANE_A and PLANE_B in the pre-g4x wm code. v2: s/old_plane_id/i9xx_plane_id/ (Daniel) v3: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Daniel Vetter Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-5-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.h | 3 ++- drivers/gpu/drm/i915/intel_pm.c | 36 +++++++++++++++++---------------- 2 files changed, 21 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d63a731ef60bf..a1268d706c5e1 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -698,7 +698,8 @@ struct drm_i915_display_funcs { struct intel_cdclk_state *cdclk_state); void (*set_cdclk)(struct drm_i915_private *dev_priv, const struct intel_cdclk_state *cdclk_state); - int (*get_fifo_size)(struct drm_i915_private *dev_priv, int plane); + int (*get_fifo_size)(struct drm_i915_private *dev_priv, + enum i9xx_plane_id i9xx_plane); int (*compute_pipe_wm)(struct intel_crtc_state *cstate); int (*compute_intermediate_wm)(struct drm_device *dev, struct intel_crtc *intel_crtc, diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index ce5cd75b81301..e445ec1748317 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -512,38 +512,41 @@ static void vlv_get_fifo_size(struct intel_crtc_state *crtc_state) fifo_state->plane[PLANE_CURSOR] = 63; } -static int i9xx_get_fifo_size(struct drm_i915_private *dev_priv, int plane) +static int i9xx_get_fifo_size(struct drm_i915_private *dev_priv, + enum i9xx_plane_id i9xx_plane) { uint32_t dsparb = I915_READ(DSPARB); int size; size = dsparb & 0x7f; - if (plane) + if (i9xx_plane == PLANE_B) size = ((dsparb >> DSPARB_CSTART_SHIFT) & 0x7f) - size; - DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb, - plane ? "B" : "A", size); + DRM_DEBUG_KMS("FIFO size - (0x%08x) %c: %d\n", + dsparb, plane_name(i9xx_plane), size); return size; } -static int i830_get_fifo_size(struct drm_i915_private *dev_priv, int plane) +static int i830_get_fifo_size(struct drm_i915_private *dev_priv, + enum i9xx_plane_id i9xx_plane) { uint32_t dsparb = I915_READ(DSPARB); int size; size = dsparb & 0x1ff; - if (plane) + if (i9xx_plane == PLANE_B) size = ((dsparb >> DSPARB_BEND_SHIFT) & 0x1ff) - size; size >>= 1; /* Convert to cachelines */ - DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb, - plane ? "B" : "A", size); + DRM_DEBUG_KMS("FIFO size - (0x%08x) %c: %d\n", + dsparb, plane_name(i9xx_plane), size); return size; } -static int i845_get_fifo_size(struct drm_i915_private *dev_priv, int plane) +static int i845_get_fifo_size(struct drm_i915_private *dev_priv, + enum i9xx_plane_id i9xx_plane) { uint32_t dsparb = I915_READ(DSPARB); int size; @@ -551,9 +554,8 @@ static int i845_get_fifo_size(struct drm_i915_private *dev_priv, int plane) size = dsparb & 0x7f; size >>= 2; /* Convert to cachelines */ - DRM_DEBUG_KMS("FIFO size - (0x%08x) %s: %d\n", dsparb, - plane ? "B" : "A", - size); + DRM_DEBUG_KMS("FIFO size - (0x%08x) %c: %d\n", + dsparb, plane_name(i9xx_plane), size); return size; } @@ -2277,8 +2279,8 @@ static void i9xx_update_wm(struct intel_crtc *unused_crtc) else wm_info = &i830_a_wm_info; - fifo_size = dev_priv->display.get_fifo_size(dev_priv, 0); - crtc = intel_get_crtc_for_plane(dev_priv, 0); + fifo_size = dev_priv->display.get_fifo_size(dev_priv, PLANE_A); + crtc = intel_get_crtc_for_plane(dev_priv, PLANE_A); if (intel_crtc_active(crtc)) { const struct drm_display_mode *adjusted_mode = &crtc->config->base.adjusted_mode; @@ -2304,8 +2306,8 @@ static void i9xx_update_wm(struct intel_crtc *unused_crtc) if (IS_GEN2(dev_priv)) wm_info = &i830_bc_wm_info; - fifo_size = dev_priv->display.get_fifo_size(dev_priv, 1); - crtc = intel_get_crtc_for_plane(dev_priv, 1); + fifo_size = dev_priv->display.get_fifo_size(dev_priv, PLANE_B); + crtc = intel_get_crtc_for_plane(dev_priv, PLANE_B); if (intel_crtc_active(crtc)) { const struct drm_display_mode *adjusted_mode = &crtc->config->base.adjusted_mode; @@ -2417,7 +2419,7 @@ static void i845_update_wm(struct intel_crtc *unused_crtc) adjusted_mode = &crtc->config->base.adjusted_mode; planea_wm = intel_calculate_wm(adjusted_mode->crtc_clock, &i845_wm_info, - dev_priv->display.get_fifo_size(dev_priv, 0), + dev_priv->display.get_fifo_size(dev_priv, PLANE_A), 4, pessimal_latency_ns); fwater_lo = I915_READ(FW_BLC) & ~0xfff; fwater_lo |= (3<<8) | planea_wm; -- GitLab From 282e83ef6289542914697eae633444c7089d4479 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:12 +0200 Subject: [PATCH 0382/1471] drm/i915: Cleanup enum pipe/enum plane_id/enum i9xx_plane_id in initial fb readout MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use enum pipe, enum plane_id, and enum i9xx_plane_id consistently in the initial framebuffe readout. v2: Use old_plane_id in the ilk code v3: s/old_plane_id/i9xx_plane_id/ (Daniel) v4: Rebase due to GLK/CNL PLANE_COLOR_CTL alpha stuff v5: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Daniel Vetter Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-6-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 60 +++++++++++++++------------- 1 file changed, 33 insertions(+), 27 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 16ac86816f288..0c407cb0e6aa4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7404,14 +7404,16 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); + struct intel_plane *plane = to_intel_plane(crtc->base.primary); + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; + enum pipe pipe = crtc->pipe; u32 val, base, offset; - int pipe = crtc->pipe, plane = crtc->i9xx_plane; int fourcc, pixel_format; unsigned int aligned_height; struct drm_framebuffer *fb; struct intel_framebuffer *intel_fb; - val = I915_READ(DSPCNTR(plane)); + val = I915_READ(DSPCNTR(i9xx_plane)); if (!(val & DISPLAY_PLANE_ENABLE)) return; @@ -7438,12 +7440,12 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, if (INTEL_GEN(dev_priv) >= 4) { if (plane_config->tiling) - offset = I915_READ(DSPTILEOFF(plane)); + offset = I915_READ(DSPTILEOFF(i9xx_plane)); else - offset = I915_READ(DSPLINOFF(plane)); - base = I915_READ(DSPSURF(plane)) & 0xfffff000; + offset = I915_READ(DSPLINOFF(i9xx_plane)); + base = I915_READ(DSPSURF(i9xx_plane)) & 0xfffff000; } else { - base = I915_READ(DSPADDR(plane)); + base = I915_READ(DSPADDR(i9xx_plane)); } plane_config->base = base; @@ -7451,15 +7453,15 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, fb->width = ((val >> 16) & 0xfff) + 1; fb->height = ((val >> 0) & 0xfff) + 1; - val = I915_READ(DSPSTRIDE(pipe)); + val = I915_READ(DSPSTRIDE(i9xx_plane)); fb->pitches[0] = val & 0xffffffc0; aligned_height = intel_fb_align_height(fb, 0, fb->height); plane_config->size = fb->pitches[0] * aligned_height; - DRM_DEBUG_KMS("pipe/plane %c/%d with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", - pipe_name(pipe), plane, fb->width, fb->height, + DRM_DEBUG_KMS("%s/%s with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", + crtc->base.name, plane->base.name, fb->width, fb->height, fb->format->cpp[0] * 8, base, fb->pitches[0], plane_config->size); @@ -8428,8 +8430,10 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); + struct intel_plane *plane = to_intel_plane(crtc->base.primary); + enum plane_id plane_id = plane->id; + enum pipe pipe = crtc->pipe; u32 val, base, offset, stride_mult, tiling, alpha; - int pipe = crtc->pipe; int fourcc, pixel_format; unsigned int aligned_height; struct drm_framebuffer *fb; @@ -8445,14 +8449,14 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, fb->dev = dev; - val = I915_READ(PLANE_CTL(pipe, 0)); + val = I915_READ(PLANE_CTL(pipe, plane_id)); if (!(val & PLANE_CTL_ENABLE)) goto error; pixel_format = val & PLANE_CTL_FORMAT_MASK; if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) { - alpha = I915_READ(PLANE_COLOR_CTL(pipe, 0)); + alpha = I915_READ(PLANE_COLOR_CTL(pipe, plane_id)); alpha &= PLANE_COLOR_ALPHA_MASK; } else { alpha = val & PLANE_CTL_ALPHA_MASK; @@ -8488,16 +8492,16 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, goto error; } - base = I915_READ(PLANE_SURF(pipe, 0)) & 0xfffff000; + base = I915_READ(PLANE_SURF(pipe, plane_id)) & 0xfffff000; plane_config->base = base; - offset = I915_READ(PLANE_OFFSET(pipe, 0)); + offset = I915_READ(PLANE_OFFSET(pipe, plane_id)); - val = I915_READ(PLANE_SIZE(pipe, 0)); + val = I915_READ(PLANE_SIZE(pipe, plane_id)); fb->height = ((val >> 16) & 0xfff) + 1; fb->width = ((val >> 0) & 0x1fff) + 1; - val = I915_READ(PLANE_STRIDE(pipe, 0)); + val = I915_READ(PLANE_STRIDE(pipe, plane_id)); stride_mult = intel_fb_stride_alignment(fb, 0); fb->pitches[0] = (val & 0x3ff) * stride_mult; @@ -8505,8 +8509,8 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, plane_config->size = fb->pitches[0] * aligned_height; - DRM_DEBUG_KMS("pipe %c with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", - pipe_name(pipe), fb->width, fb->height, + DRM_DEBUG_KMS("%s/%s with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", + crtc->base.name, plane->base.name, fb->width, fb->height, fb->format->cpp[0] * 8, base, fb->pitches[0], plane_config->size); @@ -8547,14 +8551,16 @@ ironlake_get_initial_plane_config(struct intel_crtc *crtc, { struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); + struct intel_plane *plane = to_intel_plane(crtc->base.primary); + enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; + enum pipe pipe = crtc->pipe; u32 val, base, offset; - int pipe = crtc->pipe; int fourcc, pixel_format; unsigned int aligned_height; struct drm_framebuffer *fb; struct intel_framebuffer *intel_fb; - val = I915_READ(DSPCNTR(pipe)); + val = I915_READ(DSPCNTR(i9xx_plane)); if (!(val & DISPLAY_PLANE_ENABLE)) return; @@ -8579,14 +8585,14 @@ ironlake_get_initial_plane_config(struct intel_crtc *crtc, fourcc = i9xx_format_to_fourcc(pixel_format); fb->format = drm_format_info(fourcc); - base = I915_READ(DSPSURF(pipe)) & 0xfffff000; + base = I915_READ(DSPSURF(i9xx_plane)) & 0xfffff000; if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) { - offset = I915_READ(DSPOFFSET(pipe)); + offset = I915_READ(DSPOFFSET(i9xx_plane)); } else { if (plane_config->tiling) - offset = I915_READ(DSPTILEOFF(pipe)); + offset = I915_READ(DSPTILEOFF(i9xx_plane)); else - offset = I915_READ(DSPLINOFF(pipe)); + offset = I915_READ(DSPLINOFF(i9xx_plane)); } plane_config->base = base; @@ -8594,15 +8600,15 @@ ironlake_get_initial_plane_config(struct intel_crtc *crtc, fb->width = ((val >> 16) & 0xfff) + 1; fb->height = ((val >> 0) & 0xfff) + 1; - val = I915_READ(DSPSTRIDE(pipe)); + val = I915_READ(DSPSTRIDE(i9xx_plane)); fb->pitches[0] = val & 0xffffffc0; aligned_height = intel_fb_align_height(fb, 0, fb->height); plane_config->size = fb->pitches[0] * aligned_height; - DRM_DEBUG_KMS("pipe %c with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", - pipe_name(pipe), fb->width, fb->height, + DRM_DEBUG_KMS("%s/%s with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", + crtc->base.name, plane->base.name, fb->width, fb->height, fb->format->cpp[0] * 8, base, fb->pitches[0], plane_config->size); -- GitLab From 81894b2fb9285f21dc2ca5f4c1ab0c9197d48607 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:13 +0200 Subject: [PATCH 0383/1471] drm/i915: Nuke ironlake_get_initial_plane_config() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The only relevant difference between i9xx_get_initial_plane_config() and ironlake_get_initial_plane_config() is the HSW/BDW TILEOFF handling. Add that to i9xx_get_initial_plane_config() and nuke ironlake_get_initial_plane_config(). v2: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-7-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 79 +++------------------------- 1 file changed, 6 insertions(+), 73 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0c407cb0e6aa4..c1d7547c14577 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7438,7 +7438,10 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, fourcc = i9xx_format_to_fourcc(pixel_format); fb->format = drm_format_info(fourcc); - if (INTEL_GEN(dev_priv) >= 4) { + if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) { + offset = I915_READ(DSPOFFSET(i9xx_plane)); + base = I915_READ(DSPSURF(i9xx_plane)) & 0xfffff000; + } else if (INTEL_GEN(dev_priv) >= 4) { if (plane_config->tiling) offset = I915_READ(DSPTILEOFF(i9xx_plane)); else @@ -8545,76 +8548,6 @@ static void ironlake_get_pfit_config(struct intel_crtc *crtc, } } -static void -ironlake_get_initial_plane_config(struct intel_crtc *crtc, - struct intel_initial_plane_config *plane_config) -{ - struct drm_device *dev = crtc->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_plane *plane = to_intel_plane(crtc->base.primary); - enum i9xx_plane_id i9xx_plane = plane->i9xx_plane; - enum pipe pipe = crtc->pipe; - u32 val, base, offset; - int fourcc, pixel_format; - unsigned int aligned_height; - struct drm_framebuffer *fb; - struct intel_framebuffer *intel_fb; - - val = I915_READ(DSPCNTR(i9xx_plane)); - if (!(val & DISPLAY_PLANE_ENABLE)) - return; - - intel_fb = kzalloc(sizeof(*intel_fb), GFP_KERNEL); - if (!intel_fb) { - DRM_DEBUG_KMS("failed to alloc fb\n"); - return; - } - - fb = &intel_fb->base; - - fb->dev = dev; - - if (INTEL_GEN(dev_priv) >= 4) { - if (val & DISPPLANE_TILED) { - plane_config->tiling = I915_TILING_X; - fb->modifier = I915_FORMAT_MOD_X_TILED; - } - } - - pixel_format = val & DISPPLANE_PIXFORMAT_MASK; - fourcc = i9xx_format_to_fourcc(pixel_format); - fb->format = drm_format_info(fourcc); - - base = I915_READ(DSPSURF(i9xx_plane)) & 0xfffff000; - if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) { - offset = I915_READ(DSPOFFSET(i9xx_plane)); - } else { - if (plane_config->tiling) - offset = I915_READ(DSPTILEOFF(i9xx_plane)); - else - offset = I915_READ(DSPLINOFF(i9xx_plane)); - } - plane_config->base = base; - - val = I915_READ(PIPESRC(pipe)); - fb->width = ((val >> 16) & 0xfff) + 1; - fb->height = ((val >> 0) & 0xfff) + 1; - - val = I915_READ(DSPSTRIDE(i9xx_plane)); - fb->pitches[0] = val & 0xffffffc0; - - aligned_height = intel_fb_align_height(fb, 0, fb->height); - - plane_config->size = fb->pitches[0] * aligned_height; - - DRM_DEBUG_KMS("%s/%s with fb: size=%dx%d@%d, offset=%x, pitch %d, size 0x%x\n", - crtc->base.name, plane->base.name, fb->width, fb->height, - fb->format->cpp[0] * 8, base, fb->pitches[0], - plane_config->size); - - plane_config->fb = intel_fb; -} - static bool ironlake_get_pipe_config(struct intel_crtc *crtc, struct intel_crtc_state *pipe_config) { @@ -14217,7 +14150,7 @@ void intel_init_display_hooks(struct drm_i915_private *dev_priv) } else if (HAS_DDI(dev_priv)) { dev_priv->display.get_pipe_config = haswell_get_pipe_config; dev_priv->display.get_initial_plane_config = - ironlake_get_initial_plane_config; + i9xx_get_initial_plane_config; dev_priv->display.crtc_compute_clock = haswell_crtc_compute_clock; dev_priv->display.crtc_enable = haswell_crtc_enable; @@ -14225,7 +14158,7 @@ void intel_init_display_hooks(struct drm_i915_private *dev_priv) } else if (HAS_PCH_SPLIT(dev_priv)) { dev_priv->display.get_pipe_config = ironlake_get_pipe_config; dev_priv->display.get_initial_plane_config = - ironlake_get_initial_plane_config; + i9xx_get_initial_plane_config; dev_priv->display.crtc_compute_clock = ironlake_crtc_compute_clock; dev_priv->display.crtc_enable = ironlake_crtc_enable; -- GitLab From dd57602efbce0d4f39770ec253fe6f95a6f0791f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:14 +0200 Subject: [PATCH 0384/1471] drm/i915: Switch fbc over to for_each_new_intel_plane_in_state() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Stop using the old for_each_intel_plane_in_state() type iteration macro and replace it with for_each_new_intel_plane_in_state(). And similarly replace drm_atomic_get_existing_crtc_state() with intel_atomic_get_new_crtc_state(). Switch over to intel_ types as well to make the code less cluttered. v2: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Maarten Lankhorst Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-8-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/i915_drv.h | 7 +++---- drivers/gpu/drm/i915/intel_display.c | 2 +- drivers/gpu/drm/i915/intel_drv.h | 2 +- drivers/gpu/drm/i915/intel_fbc.c | 23 ++++++++++------------- 4 files changed, 15 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a1268d706c5e1..9a890f6ad2df5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -560,13 +560,13 @@ struct i915_hotplug { for_each_power_well_rev(__dev_priv, __power_well) \ for_each_if ((__power_well)->domains & (__domain_mask)) -#define for_each_intel_plane_in_state(__state, plane, plane_state, __i) \ +#define for_each_new_intel_plane_in_state(__state, plane, new_plane_state, __i) \ for ((__i) = 0; \ (__i) < (__state)->base.dev->mode_config.num_total_plane && \ ((plane) = to_intel_plane((__state)->base.planes[__i].ptr), \ - (plane_state) = to_intel_plane_state((__state)->base.planes[__i].state), 1); \ + (new_plane_state) = to_intel_plane_state((__state)->base.planes[__i].new_state), 1); \ (__i)++) \ - for_each_if (plane_state) + for_each_if (plane) #define for_each_new_intel_crtc_in_state(__state, crtc, new_crtc_state, __i) \ for ((__i) = 0; \ @@ -576,7 +576,6 @@ struct i915_hotplug { (__i)++) \ for_each_if (crtc) - #define for_each_oldnew_intel_plane_in_state(__state, plane, old_plane_state, new_plane_state, __i) \ for ((__i) = 0; \ (__i) < (__state)->base.dev->mode_config.num_total_plane && \ diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index c1d7547c14577..b1ead3f95cde6 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -12080,7 +12080,7 @@ static int intel_atomic_check(struct drm_device *dev, if (ret) return ret; - intel_fbc_choose_crtc(dev_priv, state); + intel_fbc_choose_crtc(dev_priv, intel_state); return calc_watermark_data(state); } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index d1fe7be94b62f..3ebe626661083 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1652,7 +1652,7 @@ static inline void intel_fbdev_restore_mode(struct drm_device *dev) /* intel_fbc.c */ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, - struct drm_atomic_state *state); + struct intel_atomic_state *state); bool intel_fbc_is_active(struct drm_i915_private *dev_priv); void intel_fbc_pre_update(struct intel_crtc *crtc, struct intel_crtc_state *crtc_state, diff --git a/drivers/gpu/drm/i915/intel_fbc.c b/drivers/gpu/drm/i915/intel_fbc.c index 3133131306a9c..474234322b8bb 100644 --- a/drivers/gpu/drm/i915/intel_fbc.c +++ b/drivers/gpu/drm/i915/intel_fbc.c @@ -1054,11 +1054,11 @@ void intel_fbc_flush(struct drm_i915_private *dev_priv, * enable FBC for the chosen CRTC. If it does, it will set dev_priv->fbc.crtc. */ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, - struct drm_atomic_state *state) + struct intel_atomic_state *state) { struct intel_fbc *fbc = &dev_priv->fbc; - struct drm_plane *plane; - struct drm_plane_state *plane_state; + struct intel_plane *plane; + struct intel_plane_state *plane_state; bool crtc_chosen = false; int i; @@ -1066,7 +1066,7 @@ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, /* Does this atomic commit involve the CRTC currently tied to FBC? */ if (fbc->crtc && - !drm_atomic_get_existing_crtc_state(state, &fbc->crtc->base)) + !intel_atomic_get_new_crtc_state(state, fbc->crtc)) goto out; if (!intel_fbc_can_enable(dev_priv)) @@ -1076,13 +1076,11 @@ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, * plane. We could go for fancier schemes such as checking the plane * size, but this would just affect the few platforms that don't tie FBC * to pipe or plane A. */ - for_each_new_plane_in_state(state, plane, plane_state, i) { - struct intel_plane_state *intel_plane_state = - to_intel_plane_state(plane_state); - struct intel_crtc_state *intel_crtc_state; - struct intel_crtc *crtc = to_intel_crtc(plane_state->crtc); + for_each_new_intel_plane_in_state(state, plane, plane_state, i) { + struct intel_crtc_state *crtc_state; + struct intel_crtc *crtc = to_intel_crtc(plane_state->base.crtc); - if (!intel_plane_state->base.visible) + if (!plane_state->base.visible) continue; if (fbc_on_pipe_a_only(dev_priv) && crtc->pipe != PIPE_A) @@ -1091,10 +1089,9 @@ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, if (fbc_on_plane_a_only(dev_priv) && crtc->i9xx_plane != PLANE_A) continue; - intel_crtc_state = to_intel_crtc_state( - drm_atomic_get_existing_crtc_state(state, &crtc->base)); + crtc_state = intel_atomic_get_new_crtc_state(state, crtc); - intel_crtc_state->enable_fbc = true; + crtc_state->enable_fbc = true; crtc_chosen = true; break; } -- GitLab From b1558c7ea16038f3c3dcc40dd5c1cf7ad6004ada Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:15 +0200 Subject: [PATCH 0385/1471] drm/i915: Nuke crtc->plane MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Eliminate crtc->plane since it's pretty much a layering violation. We can always get the plane via crtc->primary if we actually need it. The only ugly thing left is plane_to_crtc_mapping[], but that's still needed by the pre-g4x watermark code. v2: Removed a misplaced comment change (Daniel) v3: Rebase due to fbc crtc->y usage removal v4: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Daniel Vetter Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-9-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 5 ++--- drivers/gpu/drm/i915/intel_drv.h | 1 - drivers/gpu/drm/i915/intel_fbc.c | 4 ++-- 3 files changed, 4 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index b1ead3f95cde6..62cc2a600205a 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -13463,14 +13463,13 @@ static int intel_crtc_init(struct drm_i915_private *dev_priv, enum pipe pipe) goto fail; intel_crtc->pipe = pipe; - intel_crtc->i9xx_plane = primary->i9xx_plane; /* initialize shared scalers */ intel_crtc_init_scalers(intel_crtc, crtc_state); BUG_ON(pipe >= ARRAY_SIZE(dev_priv->plane_to_crtc_mapping) || - dev_priv->plane_to_crtc_mapping[intel_crtc->i9xx_plane] != NULL); - dev_priv->plane_to_crtc_mapping[intel_crtc->i9xx_plane] = intel_crtc; + dev_priv->plane_to_crtc_mapping[primary->i9xx_plane] != NULL); + dev_priv->plane_to_crtc_mapping[primary->i9xx_plane] = intel_crtc; dev_priv->pipe_to_crtc_mapping[intel_crtc->pipe] = intel_crtc; drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3ebe626661083..635a96fcd7888 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -799,7 +799,6 @@ struct intel_crtc_state { struct intel_crtc { struct drm_crtc base; enum pipe pipe; - enum i9xx_plane_id i9xx_plane; /* * Whether the crtc and the connected output pipeline is active. Implies * that crtc->enabled is set, i.e. the current mode configuration has diff --git a/drivers/gpu/drm/i915/intel_fbc.c b/drivers/gpu/drm/i915/intel_fbc.c index 474234322b8bb..4aefc658a5cf6 100644 --- a/drivers/gpu/drm/i915/intel_fbc.c +++ b/drivers/gpu/drm/i915/intel_fbc.c @@ -890,7 +890,7 @@ static void intel_fbc_get_reg_params(struct intel_crtc *crtc, params->vma = cache->vma; params->crtc.pipe = crtc->pipe; - params->crtc.i9xx_plane = crtc->i9xx_plane; + params->crtc.i9xx_plane = to_intel_plane(crtc->base.primary)->i9xx_plane; params->crtc.fence_y_offset = get_crtc_fence_y_offset(fbc); params->fb.format = cache->fb.format; @@ -1086,7 +1086,7 @@ void intel_fbc_choose_crtc(struct drm_i915_private *dev_priv, if (fbc_on_pipe_a_only(dev_priv) && crtc->pipe != PIPE_A) continue; - if (fbc_on_plane_a_only(dev_priv) && crtc->i9xx_plane != PLANE_A) + if (fbc_on_plane_a_only(dev_priv) && plane->i9xx_plane != PLANE_A) continue; crtc_state = intel_atomic_get_new_crtc_state(state, crtc); -- GitLab From 2924b8cc415a950687a890cb6dea8dbc0e930447 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:16 +0200 Subject: [PATCH 0386/1471] drm/i915: Use plane->get_hw_state() for initial plane fb readout MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since we now have a ->get_hw_state() method for planes, let's use that during the initial plane fb readout. v2: s/plane/i9xx_plane/ etc. (James) Cc: James Ausmus Cc: Daniel Vetter Suggested-by: Daniel Vetter Reviewed-by: James Ausmus Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-10-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 62cc2a600205a..ed58311e8da06 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7413,8 +7413,7 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, struct drm_framebuffer *fb; struct intel_framebuffer *intel_fb; - val = I915_READ(DSPCNTR(i9xx_plane)); - if (!(val & DISPLAY_PLANE_ENABLE)) + if (!plane->get_hw_state(plane)) return; intel_fb = kzalloc(sizeof(*intel_fb), GFP_KERNEL); @@ -7427,6 +7426,8 @@ i9xx_get_initial_plane_config(struct intel_crtc *crtc, fb->dev = dev; + val = I915_READ(DSPCNTR(i9xx_plane)); + if (INTEL_GEN(dev_priv) >= 4) { if (val & DISPPLANE_TILED) { plane_config->tiling = I915_TILING_X; @@ -8442,6 +8443,9 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, struct drm_framebuffer *fb; struct intel_framebuffer *intel_fb; + if (!plane->get_hw_state(plane)) + return; + intel_fb = kzalloc(sizeof(*intel_fb), GFP_KERNEL); if (!intel_fb) { DRM_DEBUG_KMS("failed to alloc fb\n"); @@ -8453,8 +8457,6 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc, fb->dev = dev; val = I915_READ(PLANE_CTL(pipe, plane_id)); - if (!(val & PLANE_CTL_ENABLE)) - goto error; pixel_format = val & PLANE_CTL_FORMAT_MASK; -- GitLab From cff109f06dfebc379a5a36d49b541212caaec607 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Fri, 17 Nov 2017 21:19:17 +0200 Subject: [PATCH 0387/1471] drm/i915: Add rudimentary plane state verification MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Check that the planes are in the state we expect them to be. For now we can only check whether each plane is correctly enabled or disabled. In the future we may want to expand the plane state readout to support a more thorough verification. v2: Verify all planes part of the state as long as at least one crtc is doing a modeset (Daniel) v3: Fix typoes (James) Cc: James Ausmus Cc: Daniel Vetter Suggested-by: Daniel Vetter Reviewed-by: James Ausmus Link: https://patchwork.freedesktop.org/patch/msgid/20171117191917.11506-11-ville.syrjala@linux.intel.com Signed-off-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ed58311e8da06..5ca7b33554ef4 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -11600,6 +11600,18 @@ verify_crtc_state(struct drm_crtc *crtc, } } +static void +intel_verify_planes(struct intel_atomic_state *state) +{ + struct intel_plane *plane; + const struct intel_plane_state *plane_state; + int i; + + for_each_new_intel_plane_in_state(state, plane, + plane_state, i) + assert_plane(plane, plane_state->base.visible); +} + static void verify_single_dpll_state(struct drm_i915_private *dev_priv, struct intel_shared_dpll *pll, @@ -12393,6 +12405,9 @@ static void intel_atomic_commit_tail(struct drm_atomic_state *state) intel_modeset_verify_crtc(crtc, state, old_crtc_state, new_crtc_state); } + if (intel_state->modeset) + intel_verify_planes(intel_state); + if (intel_state->modeset && intel_can_enable_sagv(state)) intel_enable_sagv(dev_priv); -- GitLab From f9eb63b98c91f4cfaddf54b769f971c77da10917 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 21 Nov 2017 11:06:52 +0000 Subject: [PATCH 0388/1471] drm/i915/selftests: Avoid drm_gem_handle_create under struct_mutex Despite us reloading the module around every selftest, the lockclasses persist and the chains used in selftesting may then dictate how we are allowed to nest locks during runtime testing. As such we have to be just as careful, and in particular it turns out we are not allowed to nest dev->object_name_lock (drm_gem_handle_create) inside dev->struct_mutex. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103830 Signed-off-by: Chris Wilson Cc: Matthew Auld Cc: Joonas Lahtinen Cc: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171121110652.1107-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- .../gpu/drm/i915/selftests/i915_gem_context.c | 21 ++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c index 09340b3c11562..ec1eff739e012 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c @@ -264,6 +264,23 @@ static int cpu_check(struct drm_i915_gem_object *obj, unsigned int max) return err; } +static int file_add_object(struct drm_file *file, + struct drm_i915_gem_object *obj) +{ + int err; + + GEM_BUG_ON(obj->base.handle_count); + + /* tie the object to the drm_file for easy reaping */ + err = idr_alloc(&file->object_idr, &obj->base, 1, 0, GFP_KERNEL); + if (err < 0) + return err; + + i915_gem_object_get(obj); + obj->base.handle_count++; + return 0; +} + static struct drm_i915_gem_object * create_test_object(struct i915_gem_context *ctx, struct drm_file *file, @@ -273,7 +290,6 @@ create_test_object(struct i915_gem_context *ctx, struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &ctx->i915->ggtt.base; u64 size; - u32 handle; int err; size = min(vm->total / 2, 1024ull * DW_PER_PAGE * PAGE_SIZE); @@ -283,8 +299,7 @@ create_test_object(struct i915_gem_context *ctx, if (IS_ERR(obj)) return obj; - /* tie the handle to the drm_file for easy reaping */ - err = drm_gem_handle_create(file, &obj->base, &handle); + err = file_add_object(file, obj); i915_gem_object_put(obj); if (err) return ERR_PTR(err); -- GitLab From c84b27054643db73e56a5e933191bda435b6c46e Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:44 +0000 Subject: [PATCH 0389/1471] drm/i915: Extract intel_get_cagf Code to be shared between debugfs and the PMU implementation. v2: Checkpatch cleanup. v3: Also consolidate i915_sysfs.c/gt_act_freq_mhz_show. v4: Rebase. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_debugfs.c | 9 ++------- drivers/gpu/drm/i915/i915_drv.h | 2 ++ drivers/gpu/drm/i915/i915_sysfs.c | 11 +++-------- drivers/gpu/drm/i915/intel_pm.c | 14 ++++++++++++++ 4 files changed, 21 insertions(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 41d49a4d25d35..28294470ae31f 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1151,13 +1151,8 @@ static int i915_frequency_info(struct seq_file *m, void *unused) rpdownei = I915_READ(GEN6_RP_CUR_DOWN_EI) & GEN6_CURIAVG_MASK; rpcurdown = I915_READ(GEN6_RP_CUR_DOWN) & GEN6_CURBSYTAVG_MASK; rpprevdown = I915_READ(GEN6_RP_PREV_DOWN) & GEN6_CURBSYTAVG_MASK; - if (INTEL_GEN(dev_priv) >= 9) - cagf = (rpstat & GEN9_CAGF_MASK) >> GEN9_CAGF_SHIFT; - else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) - cagf = (rpstat & HSW_CAGF_MASK) >> HSW_CAGF_SHIFT; - else - cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT; - cagf = intel_gpu_freq(dev_priv, cagf); + cagf = intel_gpu_freq(dev_priv, + intel_get_cagf(dev_priv, rpstat)); intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 9a890f6ad2df5..019117144b3bc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -4226,6 +4226,8 @@ int intel_freq_opcode(struct drm_i915_private *dev_priv, int val); u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, const i915_reg_t reg); +u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat1); + #define I915_READ8(reg) dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true) #define I915_WRITE8(reg, val) dev_priv->uncore.funcs.mmio_writeb(dev_priv, (reg), (val), true) diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c index 791759f632e1d..450ac7d343ad8 100644 --- a/drivers/gpu/drm/i915/i915_sysfs.c +++ b/drivers/gpu/drm/i915/i915_sysfs.c @@ -252,14 +252,9 @@ static ssize_t gt_act_freq_mhz_show(struct device *kdev, freq = vlv_punit_read(dev_priv, PUNIT_REG_GPU_FREQ_STS); ret = intel_gpu_freq(dev_priv, (freq >> 8) & 0xff); } else { - u32 rpstat = I915_READ(GEN6_RPSTAT1); - if (INTEL_GEN(dev_priv) >= 9) - ret = (rpstat & GEN9_CAGF_MASK) >> GEN9_CAGF_SHIFT; - else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) - ret = (rpstat & HSW_CAGF_MASK) >> HSW_CAGF_SHIFT; - else - ret = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT; - ret = intel_gpu_freq(dev_priv, ret); + ret = intel_gpu_freq(dev_priv, + intel_get_cagf(dev_priv, + I915_READ(GEN6_RPSTAT1))); } mutex_unlock(&dev_priv->pcu_lock); diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index e445ec1748317..f1dc36c1fd697 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -9468,3 +9468,17 @@ u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, intel_runtime_pm_put(dev_priv); return DIV_ROUND_UP_ULL(time_hw * units, div); } + +u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat) +{ + u32 cagf; + + if (INTEL_GEN(dev_priv) >= 9) + cagf = (rpstat & GEN9_CAGF_MASK) >> GEN9_CAGF_SHIFT; + else if (IS_HASWELL(dev_priv) || IS_BROADWELL(dev_priv)) + cagf = (rpstat & HSW_CAGF_MASK) >> HSW_CAGF_SHIFT; + else + cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT; + + return cagf; +} -- GitLab From b46a33e271ed81bd765c632b972c49d5b44729c7 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:45 +0000 Subject: [PATCH 0390/1471] drm/i915/pmu: Expose a PMU interface for perf queries MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Chris Wilson From: Tvrtko Ursulin From: Dmitry Rogozhkin The first goal is to be able to measure GPU (and invidual ring) busyness without having to poll registers from userspace. (Which not only incurs holding the forcewake lock indefinitely, perturbing the system, but also runs the risk of hanging the machine.) As an alternative we can use the perf event counter interface to sample the ring registers periodically and send those results to userspace. Functionality we are exporting to userspace is via the existing perf PMU API and can be exercised via the existing tools. For example: perf stat -a -e i915/rcs0-busy/ -I 1000 Will print the render engine busynnes once per second. All the performance counters can be enumerated (perf list) and have their unit of measure correctly reported in sysfs. v1-v2 (Chris Wilson): v2: Use a common timer for the ring sampling. v3: (Tvrtko Ursulin) * Decouple uAPI from i915 engine ids. * Complete uAPI defines. * Refactor some code to helpers for clarity. * Skip sampling disabled engines. * Expose counters in sysfs. * Pass in fake regs to avoid null ptr deref in perf core. * Convert to class/instance uAPI. * Use shared driver code for rc6 residency, power and frequency. v4: (Dmitry Rogozhkin) * Register PMU with .task_ctx_nr=perf_invalid_context * Expose cpumask for the PMU with the single CPU in the mask * Properly support pmu->stop(): it should call pmu->read() * Properly support pmu->del(): it should call stop(event, PERF_EF_UPDATE) * Introduce refcounting of event subscriptions. * Make pmu.busy_stats a refcounter to avoid busy stats going away with some deleted event. * Expose cpumask for i915 PMU to avoid multiple events creation of the same type followed by counter aggregation by perf-stat. * Track CPUs getting online/offline to migrate perf context. If (likely) cpumask will initially set CPU0, CONFIG_BOOTPARAM_HOTPLUG_CPU0 will be needed to see effect of CPU status tracking. * End result is that only global events are supported and perf stat works correctly. * Deny perf driver level sampling - it is prohibited for uncore PMU. v5: (Tvrtko Ursulin) * Don't hardcode number of engine samplers. * Rewrite event ref-counting for correctness and simplicity. * Store initial counter value when starting already enabled events to correctly report values to all listeners. * Fix RC6 residency readout. * Comments, GPL header. v6: * Add missing entry to v4 changelog. * Fix accounting in CPU hotplug case by copying the approach from arch/x86/events/intel/cstate.c. (Dmitry Rogozhkin) v7: * Log failure message only on failure. * Remove CPU hotplug notification state on unregister. v8: * Fix error unwind on failed registration. * Checkpatch cleanup. v9: * Drop the energy metric, it is available via intel_rapl_perf. (Ville Syrjälä) * Use HAS_RC6(p). (Chris Wilson) * Handle unsupported non-engine events. (Dmitry Rogozhkin) * Rebase for intel_rc6_residency_ns needing caller managed runtime pm. * Drop HAS_RC6 checks from the read callback since creating those events will be rejected at init time already. * Add counter units to sysfs so perf stat output is nicer. * Cleanup the attribute tables for brevity and readability. v10: * Fixed queued accounting. v11: * Move intel_engine_lookup_user to intel_engine_cs.c * Commit update. (Joonas Lahtinen) v12: * More accurate sampling. (Chris Wilson) * Store and report frequency in MHz for better usability from perf stat. * Removed metrics: queued, interrupts, rc6 counters. * Sample engine busyness based on seqno difference only for less MMIO (and forcewake) on all platforms. (Chris Wilson) v13: * Comment spelling, use mul_u32_u32 to work around potential GCC issue and somne code alignment changes. (Chris Wilson) v14: * Rebase. v15: * Rebase for RPS refactoring. v16: * Use the dynamic slot in the CPU hotplug state machine so that we are free to setup our state as multi-instance. Previously we were re-using the CPUHP_AP_PERF_X86_UNCORE_ONLINE slot which is neither used as multi-instance, nor owned by our driver to start with. * Register the CPU hotplug handlers after the PMU, otherwise the callback will get called before the PMU is initialized which can end up in perf_pmu_migrate_context with an un-initialized base. * Added workaround for a probable bug in cpuhp core. v17: * Remove workaround for the cpuhp bug. v18: * Rebase for drm_i915_gem_engine_class getting upstream before us. v19: * Rebase. (trivial) Signed-off-by: Chris Wilson Signed-off-by: Tvrtko Ursulin Signed-off-by: Dmitry Rogozhkin Cc: Tvrtko Ursulin Cc: Chris Wilson Cc: Dmitry Rogozhkin Cc: Peter Zijlstra Reviewed-by: Chris Wilson Signed-off-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-2-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/Makefile | 1 + drivers/gpu/drm/i915/i915_drv.c | 3 + drivers/gpu/drm/i915/i915_drv.h | 5 + drivers/gpu/drm/i915/i915_pmu.c | 688 ++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_pmu.h | 104 ++++ drivers/gpu/drm/i915/i915_reg.h | 3 + drivers/gpu/drm/i915/intel_engine_cs.c | 33 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 26 + include/uapi/drm/i915_drm.h | 39 ++ 9 files changed, 902 insertions(+) create mode 100644 drivers/gpu/drm/i915/i915_pmu.c create mode 100644 drivers/gpu/drm/i915/i915_pmu.h diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index c3649ec5b041c..42bc8bd4ff061 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -45,6 +45,7 @@ i915-y := i915_drv.o \ i915-$(CONFIG_COMPAT) += i915_ioc32.o i915-$(CONFIG_DEBUG_FS) += i915_debugfs.o intel_pipe_crc.o +i915-$(CONFIG_PERF_EVENTS) += i915_pmu.o # GEM code i915-y += i915_cmd_parser.o \ diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 8dbcb03b5f54b..0793a27e2b95b 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -48,6 +48,7 @@ #include "i915_drv.h" #include "i915_trace.h" +#include "i915_pmu.h" #include "i915_vgpu.h" #include "intel_drv.h" #include "intel_uc.h" @@ -1215,6 +1216,7 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) struct drm_device *dev = &dev_priv->drm; i915_gem_shrinker_init(dev_priv); + i915_pmu_register(dev_priv); /* * Notify a valid surface after modesetting, @@ -1269,6 +1271,7 @@ static void i915_driver_unregister(struct drm_i915_private *dev_priv) intel_opregion_unregister(dev_priv); i915_perf_unregister(dev_priv); + i915_pmu_unregister(dev_priv); i915_teardown_sysfs(dev_priv); i915_guc_log_unregister(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 019117144b3bc..5bd5ac4cd03e9 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -40,6 +40,7 @@ #include #include #include +#include #include #include #include @@ -2290,6 +2291,8 @@ struct drm_i915_private { struct i915_gem_context *kernel_context; /* Context only to be used for injecting preemption commands */ struct i915_gem_context *preempt_context; + struct intel_engine_cs *engine_class[MAX_ENGINE_CLASS + 1] + [MAX_ENGINE_INSTANCE + 1]; struct drm_dma_handle *status_page_dmah; struct resource mch_res; @@ -2761,6 +2764,8 @@ struct drm_i915_private { int irq; } lpe_audio; + struct i915_pmu pmu; + /* * NOTE: This is the dri1/ums dungeon, don't add stuff here. Your patch * will be rejected. Instead look for a better place. diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c new file mode 100644 index 0000000000000..01b5ee67c1bf8 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -0,0 +1,688 @@ +/* + * Copyright © 2017 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#include +#include + +#include "i915_drv.h" +#include "i915_pmu.h" +#include "intel_ringbuffer.h" + +/* Frequency for the sampling timer for events which need it. */ +#define FREQUENCY 200 +#define PERIOD max_t(u64, 10000, NSEC_PER_SEC / FREQUENCY) + +#define ENGINE_SAMPLE_MASK \ + (BIT(I915_SAMPLE_BUSY) | \ + BIT(I915_SAMPLE_WAIT) | \ + BIT(I915_SAMPLE_SEMA)) + +#define ENGINE_SAMPLE_BITS (1 << I915_PMU_SAMPLE_BITS) + +static cpumask_t i915_pmu_cpumask = CPU_MASK_NONE; + +static u8 engine_config_sample(u64 config) +{ + return config & I915_PMU_SAMPLE_MASK; +} + +static u8 engine_event_sample(struct perf_event *event) +{ + return engine_config_sample(event->attr.config); +} + +static u8 engine_event_class(struct perf_event *event) +{ + return (event->attr.config >> I915_PMU_CLASS_SHIFT) & 0xff; +} + +static u8 engine_event_instance(struct perf_event *event) +{ + return (event->attr.config >> I915_PMU_SAMPLE_BITS) & 0xff; +} + +static bool is_engine_config(u64 config) +{ + return config < __I915_PMU_OTHER(0); +} + +static unsigned int config_enabled_bit(u64 config) +{ + if (is_engine_config(config)) + return engine_config_sample(config); + else + return ENGINE_SAMPLE_BITS + (config - __I915_PMU_OTHER(0)); +} + +static u64 config_enabled_mask(u64 config) +{ + return BIT_ULL(config_enabled_bit(config)); +} + +static bool is_engine_event(struct perf_event *event) +{ + return is_engine_config(event->attr.config); +} + +static unsigned int event_enabled_bit(struct perf_event *event) +{ + return config_enabled_bit(event->attr.config); +} + +static bool grab_forcewake(struct drm_i915_private *i915, bool fw) +{ + if (!fw) + intel_uncore_forcewake_get(i915, FORCEWAKE_ALL); + + return true; +} + +static void +update_sample(struct i915_pmu_sample *sample, u32 unit, u32 val) +{ + /* + * Since we are doing stochastic sampling for these counters, + * average the delta with the previous value for better accuracy. + */ + sample->cur += div_u64(mul_u32_u32(sample->prev + val, unit), 2); + sample->prev = val; +} + +static void engines_sample(struct drm_i915_private *dev_priv) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + bool fw = false; + + if ((dev_priv->pmu.enable & ENGINE_SAMPLE_MASK) == 0) + return; + + if (!dev_priv->gt.awake) + return; + + if (!intel_runtime_pm_get_if_in_use(dev_priv)) + return; + + for_each_engine(engine, dev_priv, id) { + u32 current_seqno = intel_engine_get_seqno(engine); + u32 last_seqno = intel_engine_last_submit(engine); + u32 val; + + val = !i915_seqno_passed(current_seqno, last_seqno); + + update_sample(&engine->pmu.sample[I915_SAMPLE_BUSY], + PERIOD, val); + + if (val && (engine->pmu.enable & + (BIT(I915_SAMPLE_WAIT) | BIT(I915_SAMPLE_SEMA)))) { + fw = grab_forcewake(dev_priv, fw); + + val = I915_READ_FW(RING_CTL(engine->mmio_base)); + } else { + val = 0; + } + + update_sample(&engine->pmu.sample[I915_SAMPLE_WAIT], + PERIOD, !!(val & RING_WAIT)); + + update_sample(&engine->pmu.sample[I915_SAMPLE_SEMA], + PERIOD, !!(val & RING_WAIT_SEMAPHORE)); + } + + if (fw) + intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL); + + intel_runtime_pm_put(dev_priv); +} + +static void frequency_sample(struct drm_i915_private *dev_priv) +{ + if (dev_priv->pmu.enable & + config_enabled_mask(I915_PMU_ACTUAL_FREQUENCY)) { + u32 val; + + val = dev_priv->gt_pm.rps.cur_freq; + if (dev_priv->gt.awake && + intel_runtime_pm_get_if_in_use(dev_priv)) { + val = intel_get_cagf(dev_priv, + I915_READ_NOTRACE(GEN6_RPSTAT1)); + intel_runtime_pm_put(dev_priv); + } + + update_sample(&dev_priv->pmu.sample[__I915_SAMPLE_FREQ_ACT], + 1, intel_gpu_freq(dev_priv, val)); + } + + if (dev_priv->pmu.enable & + config_enabled_mask(I915_PMU_REQUESTED_FREQUENCY)) { + update_sample(&dev_priv->pmu.sample[__I915_SAMPLE_FREQ_REQ], 1, + intel_gpu_freq(dev_priv, + dev_priv->gt_pm.rps.cur_freq)); + } +} + +static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer) +{ + struct drm_i915_private *i915 = + container_of(hrtimer, struct drm_i915_private, pmu.timer); + + if (i915->pmu.enable == 0) + return HRTIMER_NORESTART; + + engines_sample(i915); + frequency_sample(i915); + + hrtimer_forward_now(hrtimer, ns_to_ktime(PERIOD)); + return HRTIMER_RESTART; +} + +static void i915_pmu_event_destroy(struct perf_event *event) +{ + WARN_ON(event->parent); +} + +static int engine_event_init(struct perf_event *event) +{ + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + + if (!intel_engine_lookup_user(i915, engine_event_class(event), + engine_event_instance(event))) + return -ENODEV; + + switch (engine_event_sample(event)) { + case I915_SAMPLE_BUSY: + case I915_SAMPLE_WAIT: + break; + case I915_SAMPLE_SEMA: + if (INTEL_GEN(i915) < 6) + return -ENODEV; + break; + default: + return -ENOENT; + } + + return 0; +} + +static int i915_pmu_event_init(struct perf_event *event) +{ + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + int cpu, ret; + + if (event->attr.type != event->pmu->type) + return -ENOENT; + + /* unsupported modes and filters */ + if (event->attr.sample_period) /* no sampling */ + return -EINVAL; + + if (has_branch_stack(event)) + return -EOPNOTSUPP; + + if (event->cpu < 0) + return -EINVAL; + + cpu = cpumask_any_and(&i915_pmu_cpumask, + topology_sibling_cpumask(event->cpu)); + if (cpu >= nr_cpu_ids) + return -ENODEV; + + if (is_engine_event(event)) { + ret = engine_event_init(event); + } else { + ret = 0; + switch (event->attr.config) { + case I915_PMU_ACTUAL_FREQUENCY: + if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) + /* Requires a mutex for sampling! */ + ret = -ENODEV; + case I915_PMU_REQUESTED_FREQUENCY: + if (INTEL_GEN(i915) < 6) + ret = -ENODEV; + break; + default: + ret = -ENOENT; + break; + } + } + if (ret) + return ret; + + event->cpu = cpu; + if (!event->parent) + event->destroy = i915_pmu_event_destroy; + + return 0; +} + +static u64 __i915_pmu_event_read(struct perf_event *event) +{ + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + u64 val = 0; + + if (is_engine_event(event)) { + u8 sample = engine_event_sample(event); + struct intel_engine_cs *engine; + + engine = intel_engine_lookup_user(i915, + engine_event_class(event), + engine_event_instance(event)); + + if (WARN_ON_ONCE(!engine)) { + /* Do nothing */ + } else { + val = engine->pmu.sample[sample].cur; + } + } else { + switch (event->attr.config) { + case I915_PMU_ACTUAL_FREQUENCY: + val = + div_u64(i915->pmu.sample[__I915_SAMPLE_FREQ_ACT].cur, + FREQUENCY); + break; + case I915_PMU_REQUESTED_FREQUENCY: + val = + div_u64(i915->pmu.sample[__I915_SAMPLE_FREQ_REQ].cur, + FREQUENCY); + break; + } + } + + return val; +} + +static void i915_pmu_event_read(struct perf_event *event) +{ + struct hw_perf_event *hwc = &event->hw; + u64 prev, new; + +again: + prev = local64_read(&hwc->prev_count); + new = __i915_pmu_event_read(event); + + if (local64_cmpxchg(&hwc->prev_count, prev, new) != prev) + goto again; + + local64_add(new - prev, &event->count); +} + +static void i915_pmu_enable(struct perf_event *event) +{ + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + unsigned int bit = event_enabled_bit(event); + unsigned long flags; + + spin_lock_irqsave(&i915->pmu.lock, flags); + + /* + * Start the sampling timer when enabling the first event. + */ + if (i915->pmu.enable == 0) + hrtimer_start_range_ns(&i915->pmu.timer, + ns_to_ktime(PERIOD), 0, + HRTIMER_MODE_REL_PINNED); + + /* + * Update the bitmask of enabled events and increment + * the event reference counter. + */ + GEM_BUG_ON(bit >= I915_PMU_MASK_BITS); + GEM_BUG_ON(i915->pmu.enable_count[bit] == ~0); + i915->pmu.enable |= BIT_ULL(bit); + i915->pmu.enable_count[bit]++; + + /* + * For per-engine events the bitmask and reference counting + * is stored per engine. + */ + if (is_engine_event(event)) { + u8 sample = engine_event_sample(event); + struct intel_engine_cs *engine; + + engine = intel_engine_lookup_user(i915, + engine_event_class(event), + engine_event_instance(event)); + GEM_BUG_ON(!engine); + engine->pmu.enable |= BIT(sample); + + GEM_BUG_ON(sample >= I915_PMU_SAMPLE_BITS); + GEM_BUG_ON(engine->pmu.enable_count[sample] == ~0); + engine->pmu.enable_count[sample]++; + } + + /* + * Store the current counter value so we can report the correct delta + * for all listeners. Even when the event was already enabled and has + * an existing non-zero value. + */ + local64_set(&event->hw.prev_count, __i915_pmu_event_read(event)); + + spin_unlock_irqrestore(&i915->pmu.lock, flags); +} + +static void i915_pmu_disable(struct perf_event *event) +{ + struct drm_i915_private *i915 = + container_of(event->pmu, typeof(*i915), pmu.base); + unsigned int bit = event_enabled_bit(event); + unsigned long flags; + + spin_lock_irqsave(&i915->pmu.lock, flags); + + if (is_engine_event(event)) { + u8 sample = engine_event_sample(event); + struct intel_engine_cs *engine; + + engine = intel_engine_lookup_user(i915, + engine_event_class(event), + engine_event_instance(event)); + GEM_BUG_ON(!engine); + GEM_BUG_ON(sample >= I915_PMU_SAMPLE_BITS); + GEM_BUG_ON(engine->pmu.enable_count[sample] == 0); + /* + * Decrement the reference count and clear the enabled + * bitmask when the last listener on an event goes away. + */ + if (--engine->pmu.enable_count[sample] == 0) + engine->pmu.enable &= ~BIT(sample); + } + + GEM_BUG_ON(bit >= I915_PMU_MASK_BITS); + GEM_BUG_ON(i915->pmu.enable_count[bit] == 0); + /* + * Decrement the reference count and clear the enabled + * bitmask when the last listener on an event goes away. + */ + if (--i915->pmu.enable_count[bit] == 0) + i915->pmu.enable &= ~BIT_ULL(bit); + + spin_unlock_irqrestore(&i915->pmu.lock, flags); +} + +static void i915_pmu_event_start(struct perf_event *event, int flags) +{ + i915_pmu_enable(event); + event->hw.state = 0; +} + +static void i915_pmu_event_stop(struct perf_event *event, int flags) +{ + if (flags & PERF_EF_UPDATE) + i915_pmu_event_read(event); + i915_pmu_disable(event); + event->hw.state = PERF_HES_STOPPED; +} + +static int i915_pmu_event_add(struct perf_event *event, int flags) +{ + if (flags & PERF_EF_START) + i915_pmu_event_start(event, flags); + + return 0; +} + +static void i915_pmu_event_del(struct perf_event *event, int flags) +{ + i915_pmu_event_stop(event, PERF_EF_UPDATE); +} + +static int i915_pmu_event_event_idx(struct perf_event *event) +{ + return 0; +} + +static ssize_t i915_pmu_format_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct dev_ext_attribute *eattr; + + eattr = container_of(attr, struct dev_ext_attribute, attr); + return sprintf(buf, "%s\n", (char *)eattr->var); +} + +#define I915_PMU_FORMAT_ATTR(_name, _config) \ + (&((struct dev_ext_attribute[]) { \ + { .attr = __ATTR(_name, 0444, i915_pmu_format_show, NULL), \ + .var = (void *)_config, } \ + })[0].attr.attr) + +static struct attribute *i915_pmu_format_attrs[] = { + I915_PMU_FORMAT_ATTR(i915_eventid, "config:0-20"), + NULL, +}; + +static const struct attribute_group i915_pmu_format_attr_group = { + .name = "format", + .attrs = i915_pmu_format_attrs, +}; + +static ssize_t i915_pmu_event_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct dev_ext_attribute *eattr; + + eattr = container_of(attr, struct dev_ext_attribute, attr); + return sprintf(buf, "config=0x%lx\n", (unsigned long)eattr->var); +} + +#define I915_EVENT_ATTR(_name, _config) \ + (&((struct dev_ext_attribute[]) { \ + { .attr = __ATTR(_name, 0444, i915_pmu_event_show, NULL), \ + .var = (void *)_config, } \ + })[0].attr.attr) + +#define I915_EVENT_STR(_name, _str) \ + (&((struct perf_pmu_events_attr[]) { \ + { .attr = __ATTR(_name, 0444, perf_event_sysfs_show, NULL), \ + .id = 0, \ + .event_str = _str, } \ + })[0].attr.attr) + +#define I915_EVENT(_name, _config, _unit) \ + I915_EVENT_ATTR(_name, _config), \ + I915_EVENT_STR(_name.unit, _unit) + +#define I915_ENGINE_EVENT(_name, _class, _instance, _sample) \ + I915_EVENT_ATTR(_name, __I915_PMU_ENGINE(_class, _instance, _sample)), \ + I915_EVENT_STR(_name.unit, "ns") + +#define I915_ENGINE_EVENTS(_name, _class, _instance) \ + I915_ENGINE_EVENT(_name##_instance-busy, _class, _instance, I915_SAMPLE_BUSY), \ + I915_ENGINE_EVENT(_name##_instance-sema, _class, _instance, I915_SAMPLE_SEMA), \ + I915_ENGINE_EVENT(_name##_instance-wait, _class, _instance, I915_SAMPLE_WAIT) + +static struct attribute *i915_pmu_events_attrs[] = { + I915_ENGINE_EVENTS(rcs, I915_ENGINE_CLASS_RENDER, 0), + I915_ENGINE_EVENTS(bcs, I915_ENGINE_CLASS_COPY, 0), + I915_ENGINE_EVENTS(vcs, I915_ENGINE_CLASS_VIDEO, 0), + I915_ENGINE_EVENTS(vcs, I915_ENGINE_CLASS_VIDEO, 1), + I915_ENGINE_EVENTS(vecs, I915_ENGINE_CLASS_VIDEO_ENHANCE, 0), + + I915_EVENT(actual-frequency, I915_PMU_ACTUAL_FREQUENCY, "MHz"), + I915_EVENT(requested-frequency, I915_PMU_REQUESTED_FREQUENCY, "MHz"), + + NULL, +}; + +static const struct attribute_group i915_pmu_events_attr_group = { + .name = "events", + .attrs = i915_pmu_events_attrs, +}; + +static ssize_t +i915_pmu_get_attr_cpumask(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + return cpumap_print_to_pagebuf(true, buf, &i915_pmu_cpumask); +} + +static DEVICE_ATTR(cpumask, 0444, i915_pmu_get_attr_cpumask, NULL); + +static struct attribute *i915_cpumask_attrs[] = { + &dev_attr_cpumask.attr, + NULL, +}; + +static struct attribute_group i915_pmu_cpumask_attr_group = { + .attrs = i915_cpumask_attrs, +}; + +static const struct attribute_group *i915_pmu_attr_groups[] = { + &i915_pmu_format_attr_group, + &i915_pmu_events_attr_group, + &i915_pmu_cpumask_attr_group, + NULL +}; + +#ifdef CONFIG_HOTPLUG_CPU +static int i915_pmu_cpu_online(unsigned int cpu, struct hlist_node *node) +{ + struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), node); + unsigned int target; + + GEM_BUG_ON(!pmu->base.event_init); + + target = cpumask_any_and(&i915_pmu_cpumask, &i915_pmu_cpumask); + /* Select the first online CPU as a designated reader. */ + if (target >= nr_cpu_ids) + cpumask_set_cpu(cpu, &i915_pmu_cpumask); + + return 0; +} + +static int i915_pmu_cpu_offline(unsigned int cpu, struct hlist_node *node) +{ + struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), node); + unsigned int target; + + GEM_BUG_ON(!pmu->base.event_init); + + if (cpumask_test_and_clear_cpu(cpu, &i915_pmu_cpumask)) { + target = cpumask_any_but(topology_sibling_cpumask(cpu), cpu); + /* Migrate events if there is a valid target */ + if (target < nr_cpu_ids) { + cpumask_set_cpu(target, &i915_pmu_cpumask); + perf_pmu_migrate_context(&pmu->base, cpu, target); + } + } + + return 0; +} + +static enum cpuhp_state cpuhp_slot = CPUHP_INVALID; +#endif + +static int i915_pmu_register_cpuhp_state(struct drm_i915_private *i915) +{ +#ifdef CONFIG_HOTPLUG_CPU + enum cpuhp_state slot; + int ret; + + ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN, + "perf/x86/intel/i915:online", + i915_pmu_cpu_online, + i915_pmu_cpu_offline); + if (ret < 0) + return ret; + + slot = ret; + ret = cpuhp_state_add_instance(slot, &i915->pmu.node); + if (ret) { + cpuhp_remove_multi_state(slot); + return ret; + } + + cpuhp_slot = slot; +#endif + return 0; +} + +static void i915_pmu_unregister_cpuhp_state(struct drm_i915_private *i915) +{ +#ifdef CONFIG_HOTPLUG_CPU + WARN_ON(cpuhp_slot == CPUHP_INVALID); + WARN_ON(cpuhp_state_remove_instance(cpuhp_slot, &i915->pmu.node)); + cpuhp_remove_multi_state(cpuhp_slot); +#endif +} + +void i915_pmu_register(struct drm_i915_private *i915) +{ + int ret; + + if (INTEL_GEN(i915) <= 2) { + DRM_INFO("PMU not supported for this GPU."); + return; + } + + i915->pmu.base.attr_groups = i915_pmu_attr_groups; + i915->pmu.base.task_ctx_nr = perf_invalid_context; + i915->pmu.base.event_init = i915_pmu_event_init; + i915->pmu.base.add = i915_pmu_event_add; + i915->pmu.base.del = i915_pmu_event_del; + i915->pmu.base.start = i915_pmu_event_start; + i915->pmu.base.stop = i915_pmu_event_stop; + i915->pmu.base.read = i915_pmu_event_read; + i915->pmu.base.event_idx = i915_pmu_event_event_idx; + + spin_lock_init(&i915->pmu.lock); + hrtimer_init(&i915->pmu.timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + i915->pmu.timer.function = i915_sample; + + ret = perf_pmu_register(&i915->pmu.base, "i915", -1); + if (ret) + goto err; + + ret = i915_pmu_register_cpuhp_state(i915); + if (ret) + goto err_unreg; + + return; + +err_unreg: + perf_pmu_unregister(&i915->pmu.base); +err: + i915->pmu.base.event_init = NULL; + DRM_NOTE("Failed to register PMU! (err=%d)\n", ret); +} + +void i915_pmu_unregister(struct drm_i915_private *i915) +{ + if (!i915->pmu.base.event_init) + return; + + WARN_ON(i915->pmu.enable); + + hrtimer_cancel(&i915->pmu.timer); + + i915_pmu_unregister_cpuhp_state(i915); + + perf_pmu_unregister(&i915->pmu.base); + i915->pmu.base.event_init = NULL; +} diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h new file mode 100644 index 0000000000000..1ac8b2e346075 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_pmu.h @@ -0,0 +1,104 @@ +/* + * Copyright © 2017 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ +#ifndef __I915_PMU_H__ +#define __I915_PMU_H__ + +enum { + __I915_SAMPLE_FREQ_ACT = 0, + __I915_SAMPLE_FREQ_REQ, + __I915_NUM_PMU_SAMPLERS +}; + +/** + * How many different events we track in the global PMU mask. + * + * It is also used to know to needed number of event reference counters. + */ +#define I915_PMU_MASK_BITS \ + ((1 << I915_PMU_SAMPLE_BITS) + \ + (I915_PMU_LAST + 1 - __I915_PMU_OTHER(0))) + +struct i915_pmu_sample { + u64 cur; + u32 prev; +}; + +struct i915_pmu { + /** + * @node: List node for CPU hotplug handling. + */ + struct hlist_node node; + /** + * @base: PMU base. + */ + struct pmu base; + /** + * @lock: Lock protecting enable mask and ref count handling. + */ + spinlock_t lock; + /** + * @timer: Timer for internal i915 PMU sampling. + */ + struct hrtimer timer; + /** + * @enable: Bitmask of all currently enabled events. + * + * Bits are derived from uAPI event numbers in a way that low 16 bits + * correspond to engine event _sample_ _type_ (I915_SAMPLE_QUEUED is + * bit 0), and higher bits correspond to other events (for instance + * I915_PMU_ACTUAL_FREQUENCY is bit 16 etc). + * + * In other words, low 16 bits are not per engine but per engine + * sampler type, while the upper bits are directly mapped to other + * event types. + */ + u64 enable; + /** + * @enable_count: Reference counts for the enabled events. + * + * Array indices are mapped in the same way as bits in the @enable field + * and they are used to control sampling on/off when multiple clients + * are using the PMU API. + */ + unsigned int enable_count[I915_PMU_MASK_BITS]; + /** + * @sample: Current and previous (raw) counters for sampling events. + * + * These counters are updated from the i915 PMU sampling timer. + * + * Only global counters are held here, while the per-engine ones are in + * struct intel_engine_cs. + */ + struct i915_pmu_sample sample[__I915_NUM_PMU_SAMPLERS]; +}; + +#ifdef CONFIG_PERF_EVENTS +void i915_pmu_register(struct drm_i915_private *i915); +void i915_pmu_unregister(struct drm_i915_private *i915); +#else +static inline void i915_pmu_register(struct drm_i915_private *i915) {} +static inline void i915_pmu_unregister(struct drm_i915_private *i915) {} +#endif + +#endif diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 96c80fa0fcacf..09bf043c1c2e7 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -186,6 +186,9 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg) #define VIDEO_ENHANCEMENT_CLASS 2 #define COPY_ENGINE_CLASS 3 #define OTHER_CLASS 4 +#define MAX_ENGINE_CLASS 4 + +#define MAX_ENGINE_INSTANCE 1 /* PCI config space */ diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 22c095035539c..a5a494210b9ef 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -205,6 +205,15 @@ intel_engine_setup(struct drm_i915_private *dev_priv, GEM_BUG_ON(info->class >= ARRAY_SIZE(intel_engine_classes)); class_info = &intel_engine_classes[info->class]; + if (GEM_WARN_ON(info->class > MAX_ENGINE_CLASS)) + return -EINVAL; + + if (GEM_WARN_ON(info->instance > MAX_ENGINE_INSTANCE)) + return -EINVAL; + + if (GEM_WARN_ON(dev_priv->engine_class[info->class][info->instance])) + return -EINVAL; + GEM_BUG_ON(dev_priv->engine[id]); engine = kzalloc(sizeof(*engine), GFP_KERNEL); if (!engine) @@ -234,6 +243,7 @@ intel_engine_setup(struct drm_i915_private *dev_priv, ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier); + dev_priv->engine_class[info->class][info->instance] = engine; dev_priv->engine[id] = engine; return 0; } @@ -1816,6 +1826,29 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m) drm_printf(m, "\n"); } +static u8 user_class_map[] = { + [I915_ENGINE_CLASS_RENDER] = RENDER_CLASS, + [I915_ENGINE_CLASS_COPY] = COPY_ENGINE_CLASS, + [I915_ENGINE_CLASS_VIDEO] = VIDEO_DECODE_CLASS, + [I915_ENGINE_CLASS_VIDEO_ENHANCE] = VIDEO_ENHANCEMENT_CLASS, +}; + +struct intel_engine_cs * +intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance) +{ + if (class >= ARRAY_SIZE(user_class_map)) + return NULL; + + class = user_class_map[class]; + + GEM_BUG_ON(class > MAX_ENGINE_CLASS); + + if (instance > MAX_ENGINE_INSTANCE) + return NULL; + + return i915->engine_class[class][instance]; +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftests/mock_engine.c" #endif diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 80cd7812ce023..7ee0f18d4179f 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -5,6 +5,7 @@ #include "i915_gem_batch_pool.h" #include "i915_gem_request.h" #include "i915_gem_timeline.h" +#include "i915_pmu.h" #include "i915_selftest.h" struct drm_printer; @@ -338,6 +339,28 @@ struct intel_engine_cs { I915_SELFTEST_DECLARE(bool mock : 1); } breadcrumbs; + struct { + /** + * @enable: Bitmask of enable sample events on this engine. + * + * Bits correspond to sample event types, for instance + * I915_SAMPLE_QUEUED is bit 0 etc. + */ + u32 enable; + /** + * @enable_count: Reference count for the enabled samplers. + * + * Index number corresponds to the bit number from @enable. + */ + unsigned int enable_count[I915_PMU_SAMPLE_BITS]; + /** + * @sample: Counter values for sampling events. + * + * Our internal timer stores the current counters in this field. + */ + struct i915_pmu_sample sample[I915_ENGINE_SAMPLE_MAX]; + } pmu; + /* * A pool of objects to use as shadow copies of client batch buffers * when the command parser is enabled. Prevents the client from @@ -926,4 +949,7 @@ bool intel_engine_can_store_dword(struct intel_engine_cs *engine); void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *p); +struct intel_engine_cs * +intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance); + #endif /* _INTEL_RINGBUFFER_H_ */ diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index b579859295536..40e7b438bdaa1 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -102,6 +102,45 @@ enum drm_i915_gem_engine_class { I915_ENGINE_CLASS_INVALID = -1 }; +/** + * DOC: perf_events exposed by i915 through /sys/bus/event_sources/drivers/i915 + * + */ + +enum drm_i915_pmu_engine_sample { + I915_SAMPLE_BUSY = 0, + I915_SAMPLE_WAIT = 1, + I915_SAMPLE_SEMA = 2, + I915_ENGINE_SAMPLE_MAX /* non-ABI */ +}; + +#define I915_PMU_SAMPLE_BITS (4) +#define I915_PMU_SAMPLE_MASK (0xf) +#define I915_PMU_SAMPLE_INSTANCE_BITS (8) +#define I915_PMU_CLASS_SHIFT \ + (I915_PMU_SAMPLE_BITS + I915_PMU_SAMPLE_INSTANCE_BITS) + +#define __I915_PMU_ENGINE(class, instance, sample) \ + ((class) << I915_PMU_CLASS_SHIFT | \ + (instance) << I915_PMU_SAMPLE_BITS | \ + (sample)) + +#define I915_PMU_ENGINE_BUSY(class, instance) \ + __I915_PMU_ENGINE(class, instance, I915_SAMPLE_BUSY) + +#define I915_PMU_ENGINE_WAIT(class, instance) \ + __I915_PMU_ENGINE(class, instance, I915_SAMPLE_WAIT) + +#define I915_PMU_ENGINE_SEMA(class, instance) \ + __I915_PMU_ENGINE(class, instance, I915_SAMPLE_SEMA) + +#define __I915_PMU_OTHER(x) (__I915_PMU_ENGINE(0xff, 0xff, 0xf) + 1 + (x)) + +#define I915_PMU_ACTUAL_FREQUENCY __I915_PMU_OTHER(0) +#define I915_PMU_REQUESTED_FREQUENCY __I915_PMU_OTHER(1) + +#define I915_PMU_LAST I915_PMU_REQUESTED_FREQUENCY + /* Each region is a minimum of 16k, and there are at most 255 of them. */ #define I915_NR_TEX_REGIONS 255 /* table size 2k - maximum due to use -- GitLab From feff0dc6cdc527f8a6fd170d8de7966f56890fc0 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:46 +0000 Subject: [PATCH 0391/1471] drm/i915/pmu: Suspend sampling when GPU is idle If only a subset of events is enabled we can afford to suspend the sampling timer when the GPU is idle and so save some cycles and power. v2: Rebase and limit timer even more. v3: Rebase. v4: Rebase. v5: Skip action if perf PMU failed to register. v6: Checkpatch cleanup. v7: * Add a common helper to start the timer if needed. (Chris Wilson) * Add comment explaining bitwise logic in pmu_needs_timer. v8: Fix some comments styles. (Chris Wilson) v9: Rebase. v10: Move function declarations to i915_pmu.h. v11: Rename functions to i915_pmu_gt_(un)parked. (Chris Wilson) Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-3-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_gem.c | 1 + drivers/gpu/drm/i915/i915_gem_request.c | 1 + drivers/gpu/drm/i915/i915_pmu.c | 88 ++++++++++++++++++++++--- drivers/gpu/drm/i915/i915_pmu.h | 8 +++ 4 files changed, 88 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 792e6dc7e19b1..21ca680e9e633 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3372,6 +3372,7 @@ i915_gem_idle_work_handler(struct work_struct *work) intel_engines_park(dev_priv); i915_gem_timelines_mark_idle(dev_priv); + i915_pmu_gt_parked(dev_priv); GEM_BUG_ON(!dev_priv->gt.awake); dev_priv->gt.awake = false; diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index 86e2346357cf9..7325469ce754a 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -258,6 +258,7 @@ static void mark_busy(struct drm_i915_private *i915) i915_update_gfx_val(i915); if (INTEL_GEN(i915) >= 6) gen6_rps_busy(i915); + i915_pmu_gt_unparked(i915); intel_engines_unpark(i915); diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 01b5ee67c1bf8..fb95f0ac30eac 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -90,6 +90,75 @@ static unsigned int event_enabled_bit(struct perf_event *event) return config_enabled_bit(event->attr.config); } +static bool pmu_needs_timer(struct drm_i915_private *i915, bool gpu_active) +{ + u64 enable; + + /* + * Only some counters need the sampling timer. + * + * We start with a bitmask of all currently enabled events. + */ + enable = i915->pmu.enable; + + /* + * Mask out all the ones which do not need the timer, or in + * other words keep all the ones that could need the timer. + */ + enable &= config_enabled_mask(I915_PMU_ACTUAL_FREQUENCY) | + config_enabled_mask(I915_PMU_REQUESTED_FREQUENCY) | + ENGINE_SAMPLE_MASK; + + /* + * When the GPU is idle per-engine counters do not need to be + * running so clear those bits out. + */ + if (!gpu_active) + enable &= ~ENGINE_SAMPLE_MASK; + + /* + * If some bits remain it means we need the sampling timer running. + */ + return enable; +} + +void i915_pmu_gt_parked(struct drm_i915_private *i915) +{ + if (!i915->pmu.base.event_init) + return; + + spin_lock_irq(&i915->pmu.lock); + /* + * Signal sampling timer to stop if only engine events are enabled and + * GPU went idle. + */ + i915->pmu.timer_enabled = pmu_needs_timer(i915, false); + spin_unlock_irq(&i915->pmu.lock); +} + +static void __i915_pmu_maybe_start_timer(struct drm_i915_private *i915) +{ + if (!i915->pmu.timer_enabled && pmu_needs_timer(i915, true)) { + i915->pmu.timer_enabled = true; + hrtimer_start_range_ns(&i915->pmu.timer, + ns_to_ktime(PERIOD), 0, + HRTIMER_MODE_REL_PINNED); + } +} + +void i915_pmu_gt_unparked(struct drm_i915_private *i915) +{ + if (!i915->pmu.base.event_init) + return; + + spin_lock_irq(&i915->pmu.lock); + /* + * Re-enable sampling timer when GPU goes active. + */ + __i915_pmu_maybe_start_timer(i915); + spin_unlock_irq(&i915->pmu.lock); +} + static bool grab_forcewake(struct drm_i915_private *i915, bool fw) { if (!fw) @@ -187,7 +256,7 @@ static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer) struct drm_i915_private *i915 = container_of(hrtimer, struct drm_i915_private, pmu.timer); - if (i915->pmu.enable == 0) + if (!READ_ONCE(i915->pmu.timer_enabled)) return HRTIMER_NORESTART; engines_sample(i915); @@ -339,14 +408,6 @@ static void i915_pmu_enable(struct perf_event *event) spin_lock_irqsave(&i915->pmu.lock, flags); - /* - * Start the sampling timer when enabling the first event. - */ - if (i915->pmu.enable == 0) - hrtimer_start_range_ns(&i915->pmu.timer, - ns_to_ktime(PERIOD), 0, - HRTIMER_MODE_REL_PINNED); - /* * Update the bitmask of enabled events and increment * the event reference counter. @@ -356,6 +417,11 @@ static void i915_pmu_enable(struct perf_event *event) i915->pmu.enable |= BIT_ULL(bit); i915->pmu.enable_count[bit]++; + /* + * Start the sampling timer if needed and not already enabled. + */ + __i915_pmu_maybe_start_timer(i915); + /* * For per-engine events the bitmask and reference counting * is stored per engine. @@ -418,8 +484,10 @@ static void i915_pmu_disable(struct perf_event *event) * Decrement the reference count and clear the enabled * bitmask when the last listener on an event goes away. */ - if (--i915->pmu.enable_count[bit] == 0) + if (--i915->pmu.enable_count[bit] == 0) { i915->pmu.enable &= ~BIT_ULL(bit); + i915->pmu.timer_enabled &= pmu_needs_timer(i915, true); + } spin_unlock_irqrestore(&i915->pmu.lock, flags); } diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h index 1ac8b2e346075..e209783a4c5fa 100644 --- a/drivers/gpu/drm/i915/i915_pmu.h +++ b/drivers/gpu/drm/i915/i915_pmu.h @@ -82,6 +82,10 @@ struct i915_pmu { * are using the PMU API. */ unsigned int enable_count[I915_PMU_MASK_BITS]; + /** + * @timer_enabled: Should the internal sampling timer be running. + */ + bool timer_enabled; /** * @sample: Current and previous (raw) counters for sampling events. * @@ -96,9 +100,13 @@ struct i915_pmu { #ifdef CONFIG_PERF_EVENTS void i915_pmu_register(struct drm_i915_private *i915); void i915_pmu_unregister(struct drm_i915_private *i915); +void i915_pmu_gt_parked(struct drm_i915_private *i915); +void i915_pmu_gt_unparked(struct drm_i915_private *i915); #else static inline void i915_pmu_register(struct drm_i915_private *i915) {} static inline void i915_pmu_unregister(struct drm_i915_private *i915) {} +static inline void i915_pmu_gt_parked(struct drm_i915_private *i915) {} +static inline void i915_pmu_gt_unparked(struct drm_i915_private *i915) {} #endif #endif -- GitLab From 73fd9d381693dee6f28bb24de4b81b643d4d6d63 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:47 +0000 Subject: [PATCH 0392/1471] drm/i915: Wrap context schedule notification No functional change just something which will be handy in the following patch. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-4-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/intel_lrc.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 16d4936e2368e..8e3990d5191c4 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -379,6 +379,18 @@ execlists_context_status_change(struct drm_i915_gem_request *rq, status, rq); } +static inline void +execlists_context_schedule_in(struct drm_i915_gem_request *rq) +{ + execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN); +} + +static inline void +execlists_context_schedule_out(struct drm_i915_gem_request *rq) +{ + execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); +} + static void execlists_update_context_pdps(struct i915_hw_ppgtt *ppgtt, u32 *reg_state) { @@ -430,7 +442,7 @@ static void execlists_submit_ports(struct intel_engine_cs *engine) if (rq) { GEM_BUG_ON(count > !n); if (!count++) - execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN); + execlists_context_schedule_in(rq); port_set(&port[n], port_pack(rq, count)); desc = execlists_update_context(rq); GEM_DEBUG_EXEC(port[n].context_id = upper_32_bits(desc)); @@ -902,8 +914,7 @@ static void execlists_submission_tasklet(unsigned long data) GEM_BUG_ON(port_isset(&port[1]) && !(status & GEN8_CTX_STATUS_ELEMENT_SWITCH)); GEM_BUG_ON(!i915_gem_request_completed(rq)); - execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); - + execlists_context_schedule_out(rq); trace_i915_gem_request_out(rq); i915_gem_request_put(rq); -- GitLab From 30e17b7847f54732a91e5804144d445e9f29493b Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:48 +0000 Subject: [PATCH 0393/1471] drm/i915: Engine busy time tracking Track total time requests have been executing on the hardware. We add new kernel API to allow software tracking of time GPU engines are spending executing requests. Both per-engine and global API is added with the latter also being exported for use by external users. v2: * Squashed with the internal API. * Dropped static key. * Made per-engine. * Store time in monotonic ktime. v3: Moved stats clearing to disable. v4: * Comments. * Don't export the API just yet. v5: Whitespace cleanup. v6: * Rename ref to active. * Drop engine aggregate stats for now. * Account initial busy period after enabling stats. v7: * Rebase. v8: * Move context in notification after the notifier. (Chris Wilson) v9: In cases where stats tracking is getting disabled while there is an active context on an engine, add up the current value to the total. This also implies we don't clear the total when tracking is disabled any longer. There is no real need to do so because we define the stats as relative while enabled, meaning comparison between two samples while tracking is enabled is the valid usage. However, when busy stats will later be plugged into the perf PMU API, it is beneficial to not reset the total, since the PMU core likes to do some counter disable/enable cycles on startup, and while doing so during a single long context executing on an engine we would lose some accuracy and so make unit testing more difficult than needs to be. v10: * Fix accounting for preemption. v11: * Rebase for i915_modparams.enable_execlists removal. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-5-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/intel_engine_cs.c | 87 +++++++++++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 3 + drivers/gpu/drm/i915/intel_ringbuffer.h | 92 +++++++++++++++++++++++++ 3 files changed, 182 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index a5a494210b9ef..fede62daf3e15 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -241,6 +241,8 @@ intel_engine_setup(struct drm_i915_private *dev_priv, /* Nothing to do here, execute in order of dependencies */ engine->schedule = NULL; + spin_lock_init(&engine->stats.lock); + ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier); dev_priv->engine_class[info->class][info->instance] = engine; @@ -1849,6 +1851,91 @@ intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance) return i915->engine_class[class][instance]; } +/** + * intel_enable_engine_stats() - Enable engine busy tracking on engine + * @engine: engine to enable stats collection + * + * Start collecting the engine busyness data for @engine. + * + * Returns 0 on success or a negative error code. + */ +int intel_enable_engine_stats(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (INTEL_GEN(engine->i915) < 8) + return -ENODEV; + + spin_lock_irqsave(&engine->stats.lock, flags); + if (engine->stats.enabled == ~0) + goto busy; + if (engine->stats.enabled++ == 0) + engine->stats.enabled_at = ktime_get(); + spin_unlock_irqrestore(&engine->stats.lock, flags); + + return 0; + +busy: + spin_unlock_irqrestore(&engine->stats.lock, flags); + + return -EBUSY; +} + +static ktime_t __intel_engine_get_busy_time(struct intel_engine_cs *engine) +{ + ktime_t total = engine->stats.total; + + /* + * If the engine is executing something at the moment + * add it to the total. + */ + if (engine->stats.active) + total = ktime_add(total, + ktime_sub(ktime_get(), engine->stats.start)); + + return total; +} + +/** + * intel_engine_get_busy_time() - Return current accumulated engine busyness + * @engine: engine to report on + * + * Returns accumulated time @engine was busy since engine stats were enabled. + */ +ktime_t intel_engine_get_busy_time(struct intel_engine_cs *engine) +{ + ktime_t total; + unsigned long flags; + + spin_lock_irqsave(&engine->stats.lock, flags); + total = __intel_engine_get_busy_time(engine); + spin_unlock_irqrestore(&engine->stats.lock, flags); + + return total; +} + +/** + * intel_disable_engine_stats() - Disable engine busy tracking on engine + * @engine: engine to disable stats collection + * + * Stops collecting the engine busyness data for @engine. + */ +void intel_disable_engine_stats(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (INTEL_GEN(engine->i915) < 8) + return; + + spin_lock_irqsave(&engine->stats.lock, flags); + WARN_ON_ONCE(engine->stats.enabled == 0); + if (--engine->stats.enabled == 0) { + engine->stats.total = __intel_engine_get_busy_time(engine); + engine->stats.active = 0; + } + spin_unlock_irqrestore(&engine->stats.lock, flags); +} + #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) #include "selftests/mock_engine.c" #endif diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 8e3990d5191c4..570864583e283 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -383,11 +383,13 @@ static inline void execlists_context_schedule_in(struct drm_i915_gem_request *rq) { execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN); + intel_engine_context_in(rq->engine); } static inline void execlists_context_schedule_out(struct drm_i915_gem_request *rq) { + intel_engine_context_out(rq->engine); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT); } @@ -705,6 +707,7 @@ execlists_cancel_port_requests(struct intel_engine_execlists * const execlists) struct drm_i915_gem_request *rq = port_request(port); GEM_BUG_ON(!execlists->active); + intel_engine_context_out(rq->engine); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_PREEMPTED); i915_gem_request_put(rq); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 7ee0f18d4179f..215edfa424d5f 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -547,6 +547,38 @@ struct intel_engine_cs { * certain bits to encode the command length in the header). */ u32 (*get_cmd_length_mask)(u32 cmd_header); + + struct { + /** + * @lock: Lock protecting the below fields. + */ + spinlock_t lock; + /** + * @enabled: Reference count indicating number of listeners. + */ + unsigned int enabled; + /** + * @active: Number of contexts currently scheduled in. + */ + unsigned int active; + /** + * @enabled_at: Timestamp when busy stats were enabled. + */ + ktime_t enabled_at; + /** + * @start: Timestamp of the last idle to active transition. + * + * Idle is defined as active == 0, active is active > 0. + */ + ktime_t start; + /** + * @total: Total time this engine was busy. + * + * Accumulated time not counting the most recent block in cases + * where engine is currently busy (active > 0). + */ + ktime_t total; + } stats; }; static inline void @@ -952,4 +984,64 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *p); struct intel_engine_cs * intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance); +static inline void intel_engine_context_in(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (READ_ONCE(engine->stats.enabled) == 0) + return; + + spin_lock_irqsave(&engine->stats.lock, flags); + + if (engine->stats.enabled > 0) { + if (engine->stats.active++ == 0) + engine->stats.start = ktime_get(); + GEM_BUG_ON(engine->stats.active == 0); + } + + spin_unlock_irqrestore(&engine->stats.lock, flags); +} + +static inline void intel_engine_context_out(struct intel_engine_cs *engine) +{ + unsigned long flags; + + if (READ_ONCE(engine->stats.enabled) == 0) + return; + + spin_lock_irqsave(&engine->stats.lock, flags); + + if (engine->stats.enabled > 0) { + ktime_t last; + + if (engine->stats.active && --engine->stats.active == 0) { + /* + * Decrement the active context count and in case GPU + * is now idle add up to the running total. + */ + last = ktime_sub(ktime_get(), engine->stats.start); + + engine->stats.total = ktime_add(engine->stats.total, + last); + } else if (engine->stats.active == 0) { + /* + * After turning on engine stats, context out might be + * the first event in which case we account from the + * time stats gathering was turned on. + */ + last = ktime_sub(ktime_get(), engine->stats.enabled_at); + + engine->stats.total = ktime_add(engine->stats.total, + last); + } + } + + spin_unlock_irqrestore(&engine->stats.lock, flags); +} + +int intel_enable_engine_stats(struct intel_engine_cs *engine); +void intel_disable_engine_stats(struct intel_engine_cs *engine); + +ktime_t intel_engine_get_busy_time(struct intel_engine_cs *engine); + #endif /* _INTEL_RINGBUFFER_H_ */ -- GitLab From b3add01ee217db8e0da8a6596f03c487ed33b06e Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:49 +0000 Subject: [PATCH 0394/1471] drm/i915/pmu: Wire up engine busy stats to PMU We can use engine busy stats instead of the sampling timer for better accuracy. By doing this we replace the stohastic sampling with busyness metric derived directly from engine activity. This is context switch interrupt driven, so as accurate as we can get from software tracking. As a secondary benefit, we can also not run the sampling timer in cases only busyness metric is enabled. v2: Rebase. v3: * Rebase, comments. * Leave engine busyness controls out of workers. v4: Checkpatch cleanup. v5: Added comment to pmu_needs_timer change. v6: * Rebase. * Fix style of some comments. (Chris Wilson) v7: Rebase and commit message update. (Chris Wilson) v8: Add delayed stats disabling to improve accuracy in face of CPU hotplug events. v9: Rebase. v10: Rebase - i915_modparams.enable_execlists removal. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-6-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 78 ++++++++++++++++++++++++- drivers/gpu/drm/i915/intel_ringbuffer.h | 14 +++++ 2 files changed, 90 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index fb95f0ac30eac..6a428e7218d2b 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -90,6 +90,11 @@ static unsigned int event_enabled_bit(struct perf_event *event) return config_enabled_bit(event->attr.config); } +static bool supports_busy_stats(struct drm_i915_private *i915) +{ + return INTEL_GEN(i915) >= 8; +} + static bool pmu_needs_timer(struct drm_i915_private *i915, bool gpu_active) { u64 enable; @@ -115,6 +120,12 @@ static bool pmu_needs_timer(struct drm_i915_private *i915, bool gpu_active) */ if (!gpu_active) enable &= ~ENGINE_SAMPLE_MASK; + /* + * Also there is software busyness tracking available we do not + * need the timer for I915_SAMPLE_BUSY counter. + */ + else if (supports_busy_stats(i915)) + enable &= ~BIT(I915_SAMPLE_BUSY); /* * If some bits remain it means we need the sampling timer running. @@ -363,6 +374,9 @@ static u64 __i915_pmu_event_read(struct perf_event *event) if (WARN_ON_ONCE(!engine)) { /* Do nothing */ + } else if (sample == I915_SAMPLE_BUSY && + engine->pmu.busy_stats) { + val = ktime_to_ns(intel_engine_get_busy_time(engine)); } else { val = engine->pmu.sample[sample].cur; } @@ -399,6 +413,12 @@ static void i915_pmu_event_read(struct perf_event *event) local64_add(new - prev, &event->count); } +static bool engine_needs_busy_stats(struct intel_engine_cs *engine) +{ + return supports_busy_stats(engine->i915) && + (engine->pmu.enable & BIT(I915_SAMPLE_BUSY)); +} + static void i915_pmu_enable(struct perf_event *event) { struct drm_i915_private *i915 = @@ -438,7 +458,21 @@ static void i915_pmu_enable(struct perf_event *event) GEM_BUG_ON(sample >= I915_PMU_SAMPLE_BITS); GEM_BUG_ON(engine->pmu.enable_count[sample] == ~0); - engine->pmu.enable_count[sample]++; + if (engine->pmu.enable_count[sample]++ == 0) { + /* + * Enable engine busy stats tracking if needed or + * alternatively cancel the scheduled disable. + * + * If the delayed disable was pending, cancel it and + * in this case do not enable since it already is. + */ + if (engine_needs_busy_stats(engine) && + !engine->pmu.busy_stats) { + engine->pmu.busy_stats = true; + if (!cancel_delayed_work(&engine->pmu.disable_busy_stats)) + intel_enable_engine_stats(engine); + } + } } /* @@ -451,6 +485,14 @@ static void i915_pmu_enable(struct perf_event *event) spin_unlock_irqrestore(&i915->pmu.lock, flags); } +static void __disable_busy_stats(struct work_struct *work) +{ + struct intel_engine_cs *engine = + container_of(work, typeof(*engine), pmu.disable_busy_stats.work); + + intel_disable_engine_stats(engine); +} + static void i915_pmu_disable(struct perf_event *event) { struct drm_i915_private *i915 = @@ -474,8 +516,26 @@ static void i915_pmu_disable(struct perf_event *event) * Decrement the reference count and clear the enabled * bitmask when the last listener on an event goes away. */ - if (--engine->pmu.enable_count[sample] == 0) + if (--engine->pmu.enable_count[sample] == 0) { engine->pmu.enable &= ~BIT(sample); + if (!engine_needs_busy_stats(engine) && + engine->pmu.busy_stats) { + engine->pmu.busy_stats = false; + /* + * We request a delayed disable to handle the + * rapid on/off cycles on events, which can + * happen when tools like perf stat start, in a + * nicer way. + * + * In addition, this also helps with busy stats + * accuracy with background CPU offline/online + * migration events. + */ + queue_delayed_work(system_wq, + &engine->pmu.disable_busy_stats, + round_jiffies_up_relative(HZ)); + } + } } GEM_BUG_ON(bit >= I915_PMU_MASK_BITS); @@ -702,6 +762,8 @@ static void i915_pmu_unregister_cpuhp_state(struct drm_i915_private *i915) void i915_pmu_register(struct drm_i915_private *i915) { + struct intel_engine_cs *engine; + enum intel_engine_id id; int ret; if (INTEL_GEN(i915) <= 2) { @@ -723,6 +785,10 @@ void i915_pmu_register(struct drm_i915_private *i915) hrtimer_init(&i915->pmu.timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); i915->pmu.timer.function = i915_sample; + for_each_engine(engine, i915, id) + INIT_DELAYED_WORK(&engine->pmu.disable_busy_stats, + __disable_busy_stats); + ret = perf_pmu_register(&i915->pmu.base, "i915", -1); if (ret) goto err; @@ -742,6 +808,9 @@ void i915_pmu_register(struct drm_i915_private *i915) void i915_pmu_unregister(struct drm_i915_private *i915) { + struct intel_engine_cs *engine; + enum intel_engine_id id; + if (!i915->pmu.base.event_init) return; @@ -749,6 +818,11 @@ void i915_pmu_unregister(struct drm_i915_private *i915) hrtimer_cancel(&i915->pmu.timer); + for_each_engine(engine, i915, id) { + GEM_BUG_ON(engine->pmu.busy_stats); + flush_delayed_work(&engine->pmu.disable_busy_stats); + } + i915_pmu_unregister_cpuhp_state(i915); perf_pmu_unregister(&i915->pmu.base); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 215edfa424d5f..43473e6d1a4f8 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -359,6 +359,20 @@ struct intel_engine_cs { * Our internal timer stores the current counters in this field. */ struct i915_pmu_sample sample[I915_ENGINE_SAMPLE_MAX]; + /** + * @busy_stats: Has enablement of engine stats tracking been + * requested. + */ + bool busy_stats; + /** + * @disable_busy_stats: Work item for busy stats disabling. + * + * Same as with @enable_busy_stats action, with the difference + * that we delay it in case there are rapid enable-disable + * actions, which can happen during tool startup (like perf + * stat). + */ + struct delayed_work disable_busy_stats; } pmu; /* -- GitLab From 0cd4684d6ea9a4ffec33fc19de4dd667bb90d0a5 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:50 +0000 Subject: [PATCH 0395/1471] drm/i915/pmu: Add interrupt count metric For clients like intel-gpu-overlay it is easier to read the count via the perf API than having to parse /proc. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-7-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 23 +++++++++++++++++++++++ include/uapi/drm/i915_drm.h | 4 +++- 2 files changed, 26 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 6a428e7218d2b..fef389ebf92ca 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -277,6 +277,22 @@ static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer) return HRTIMER_RESTART; } +static u64 count_interrupts(struct drm_i915_private *i915) +{ + /* open-coded kstat_irqs() */ + struct irq_desc *desc = irq_to_desc(i915->drm.pdev->irq); + u64 sum = 0; + int cpu; + + if (!desc || !desc->kstat_irqs) + return 0; + + for_each_possible_cpu(cpu) + sum += *per_cpu_ptr(desc->kstat_irqs, cpu); + + return sum; +} + static void i915_pmu_event_destroy(struct perf_event *event) { WARN_ON(event->parent); @@ -343,6 +359,8 @@ static int i915_pmu_event_init(struct perf_event *event) if (INTEL_GEN(i915) < 6) ret = -ENODEV; break; + case I915_PMU_INTERRUPTS: + break; default: ret = -ENOENT; break; @@ -392,6 +410,9 @@ static u64 __i915_pmu_event_read(struct perf_event *event) div_u64(i915->pmu.sample[__I915_SAMPLE_FREQ_REQ].cur, FREQUENCY); break; + case I915_PMU_INTERRUPTS: + val = count_interrupts(i915); + break; } } @@ -654,6 +675,8 @@ static struct attribute *i915_pmu_events_attrs[] = { I915_EVENT(actual-frequency, I915_PMU_ACTUAL_FREQUENCY, "MHz"), I915_EVENT(requested-frequency, I915_PMU_REQUESTED_FREQUENCY, "MHz"), + I915_EVENT_ATTR(interrupts, I915_PMU_INTERRUPTS), + NULL, }; diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 40e7b438bdaa1..d840ff0835205 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -139,7 +139,9 @@ enum drm_i915_pmu_engine_sample { #define I915_PMU_ACTUAL_FREQUENCY __I915_PMU_OTHER(0) #define I915_PMU_REQUESTED_FREQUENCY __I915_PMU_OTHER(1) -#define I915_PMU_LAST I915_PMU_REQUESTED_FREQUENCY +#define I915_PMU_INTERRUPTS __I915_PMU_OTHER(2) + +#define I915_PMU_LAST I915_PMU_INTERRUPTS /* Each region is a minimum of 16k, and there are at most 255 of them. */ -- GitLab From 36cc8b963f5eb46352dc79d79e4bd5e8dd2ac2d6 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:51 +0000 Subject: [PATCH 0396/1471] drm/i915: Convert intel_rc6_residency_us to ns Will be used for exposing the PMU counters. v2: * Move intel_runtime_pm_get/put to the callers. (Chris Wilson) * Restore full unit conversion precision. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-8-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_drv.h | 8 +++++++- drivers/gpu/drm/i915/i915_sysfs.c | 9 +++++++-- drivers/gpu/drm/i915/intel_pm.c | 27 +++++++++++++-------------- 3 files changed, 27 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 5bd5ac4cd03e9..d575a56fc1000 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -4228,11 +4228,17 @@ void vlv_phy_reset_lanes(struct intel_encoder *encoder, int intel_gpu_freq(struct drm_i915_private *dev_priv, int val); int intel_freq_opcode(struct drm_i915_private *dev_priv, int val); -u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, +u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, const i915_reg_t reg); u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat1); +static inline u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, + const i915_reg_t reg) +{ + return DIV_ROUND_UP_ULL(intel_rc6_residency_ns(dev_priv, reg), 1000); +} + #define I915_READ8(reg) dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true) #define I915_WRITE8(reg, val) dev_priv->uncore.funcs.mmio_writeb(dev_priv, (reg), (val), true) diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c index 450ac7d343ad8..c290cb600eea5 100644 --- a/drivers/gpu/drm/i915/i915_sysfs.c +++ b/drivers/gpu/drm/i915/i915_sysfs.c @@ -42,8 +42,13 @@ static inline struct drm_i915_private *kdev_minor_to_i915(struct device *kdev) static u32 calc_residency(struct drm_i915_private *dev_priv, i915_reg_t reg) { - return DIV_ROUND_CLOSEST_ULL(intel_rc6_residency_us(dev_priv, reg), - 1000); + u64 res; + + intel_runtime_pm_get(dev_priv); + res = intel_rc6_residency_us(dev_priv, reg); + intel_runtime_pm_put(dev_priv); + + return DIV_ROUND_CLOSEST_ULL(res, 1000); } static ssize_t diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index f1dc36c1fd697..7d41aad791665 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -9437,36 +9437,35 @@ static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, return lower | (u64)upper << 8; } -u64 intel_rc6_residency_us(struct drm_i915_private *dev_priv, +u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, const i915_reg_t reg) { - u64 time_hw, units, div; + u64 time_hw; + u32 mul, div; if (!intel_rc6_enabled()) return 0; - intel_runtime_pm_get(dev_priv); - /* On VLV and CHV, residency time is in CZ units rather than 1.28us */ if (IS_VALLEYVIEW(dev_priv) || IS_CHERRYVIEW(dev_priv)) { - units = 1000; + mul = 1000000; div = dev_priv->czclk_freq; - time_hw = vlv_residency_raw(dev_priv, reg); - } else if (IS_GEN9_LP(dev_priv)) { - units = 1000; - div = 1200; /* 833.33ns */ - time_hw = I915_READ(reg); } else { - units = 128000; /* 1.28us */ - div = 100000; + /* 833.33ns units on Gen9LP, 1.28us elsewhere. */ + if (IS_GEN9_LP(dev_priv)) { + mul = 10000; + div = 12; + } else { + mul = 1280; + div = 1; + } time_hw = I915_READ(reg); } - intel_runtime_pm_put(dev_priv); - return DIV_ROUND_UP_ULL(time_hw * units, div); + return DIV_ROUND_UP_ULL(time_hw * mul, div); } u32 intel_get_cagf(struct drm_i915_private *dev_priv, u32 rpstat) -- GitLab From 6060b6aec03c76f9ce0977b70c27429d39d2956e Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 21 Nov 2017 18:18:52 +0000 Subject: [PATCH 0397/1471] drm/i915/pmu: Add RC6 residency metrics For clients like intel-gpu-overlay it is easier to read the counters via the perf API than having to parse sysfs. Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171121181852.16128-9-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 31 +++++++++++++++++++++++++++++++ include/uapi/drm/i915_drm.h | 6 +++++- 2 files changed, 36 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index fef389ebf92ca..1071935bfa67c 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -361,6 +361,15 @@ static int i915_pmu_event_init(struct perf_event *event) break; case I915_PMU_INTERRUPTS: break; + case I915_PMU_RC6_RESIDENCY: + if (!HAS_RC6(i915)) + ret = -ENODEV; + break; + case I915_PMU_RC6p_RESIDENCY: + case I915_PMU_RC6pp_RESIDENCY: + if (!HAS_RC6p(i915)) + ret = -ENODEV; + break; default: ret = -ENOENT; break; @@ -413,6 +422,24 @@ static u64 __i915_pmu_event_read(struct perf_event *event) case I915_PMU_INTERRUPTS: val = count_interrupts(i915); break; + case I915_PMU_RC6_RESIDENCY: + intel_runtime_pm_get(i915); + val = intel_rc6_residency_ns(i915, + IS_VALLEYVIEW(i915) ? + VLV_GT_RENDER_RC6 : + GEN6_GT_GFX_RC6); + intel_runtime_pm_put(i915); + break; + case I915_PMU_RC6p_RESIDENCY: + intel_runtime_pm_get(i915); + val = intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6p); + intel_runtime_pm_put(i915); + break; + case I915_PMU_RC6pp_RESIDENCY: + intel_runtime_pm_get(i915); + val = intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6pp); + intel_runtime_pm_put(i915); + break; } } @@ -677,6 +704,10 @@ static struct attribute *i915_pmu_events_attrs[] = { I915_EVENT_ATTR(interrupts, I915_PMU_INTERRUPTS), + I915_EVENT(rc6-residency, I915_PMU_RC6_RESIDENCY, "ns"), + I915_EVENT(rc6p-residency, I915_PMU_RC6p_RESIDENCY, "ns"), + I915_EVENT(rc6pp-residency, I915_PMU_RC6pp_RESIDENCY, "ns"), + NULL, }; diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index d840ff0835205..915a6e85a8550 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -141,7 +141,11 @@ enum drm_i915_pmu_engine_sample { #define I915_PMU_INTERRUPTS __I915_PMU_OTHER(2) -#define I915_PMU_LAST I915_PMU_INTERRUPTS +#define I915_PMU_RC6_RESIDENCY __I915_PMU_OTHER(3) +#define I915_PMU_RC6p_RESIDENCY __I915_PMU_OTHER(4) +#define I915_PMU_RC6pp_RESIDENCY __I915_PMU_OTHER(5) + +#define I915_PMU_LAST I915_PMU_RC6pp_RESIDENCY /* Each region is a minimum of 16k, and there are at most 255 of them. */ -- GitLab From c65c8b0f7a1f90e84c273e111fc9391c50c3c482 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 22 Nov 2017 12:06:00 +0000 Subject: [PATCH 0398/1471] drm/i915/selftests: Use NOWARN for large allocations We may try to do a large kmalloc for the permutation array, falling back to a smaller array/test if the first allocation fails. Since we are intentionally trying a large allocation which may fail, pass __GFP_NOWARN. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103842 Signed-off-by: Chris Wilson Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171122120600.27025-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/selftests/i915_random.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/selftests/i915_random.c b/drivers/gpu/drm/i915/selftests/i915_random.c index b85872cc7fbe0..2088ae57aa899 100644 --- a/drivers/gpu/drm/i915/selftests/i915_random.c +++ b/drivers/gpu/drm/i915/selftests/i915_random.c @@ -57,7 +57,7 @@ unsigned int *i915_random_order(unsigned int count, struct rnd_state *state) { unsigned int *order, i; - order = kmalloc_array(count, sizeof(*order), GFP_KERNEL); + order = kmalloc_array(count, sizeof(*order), GFP_KERNEL | __GFP_NOWARN); if (!order) return order; -- GitLab From 593f4b19a094c4426bd1e1e3cbab87a48bd13c71 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 13 Nov 2017 19:04:18 +0200 Subject: [PATCH 0399/1471] video/hdmi: Allow "empty" HDMI infoframes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit HDMI 2.0 Appendix F suggest that we should keep sending the infoframe when switching from 3D to 2D mode, even if the infoframe isn't strictly necessary (ie. not needed to transmit the VIC or stereo information). This is a workaround against some sinks that fail to realize that they should switch from 3D to 2D mode when the source stop transmitting the infoframe. v2: Handle unpack() as well Pull the length calculation into a helper Cc: Shashank Sharma Cc: Andrzej Hajda Cc: Thierry Reding Cc: Hans Verkuil Cc: linux-media@vger.kernel.org Reviewed-by: Andrzej Hajda #v1 Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171113170427.4150-2-ville.syrjala@linux.intel.com Reviewed-by: Shashank Sharma --- drivers/video/hdmi.c | 51 +++++++++++++++++++++++++++----------------- 1 file changed, 31 insertions(+), 20 deletions(-) diff --git a/drivers/video/hdmi.c b/drivers/video/hdmi.c index 1cf907ecded4e..111a0ab6280a0 100644 --- a/drivers/video/hdmi.c +++ b/drivers/video/hdmi.c @@ -321,6 +321,17 @@ int hdmi_vendor_infoframe_init(struct hdmi_vendor_infoframe *frame) } EXPORT_SYMBOL(hdmi_vendor_infoframe_init); +static int hdmi_vendor_infoframe_length(const struct hdmi_vendor_infoframe *frame) +{ + /* for side by side (half) we also need to provide 3D_Ext_Data */ + if (frame->s3d_struct >= HDMI_3D_STRUCTURE_SIDE_BY_SIDE_HALF) + return 6; + else if (frame->vic != 0 || frame->s3d_struct != HDMI_3D_STRUCTURE_INVALID) + return 5; + else + return 4; +} + /** * hdmi_vendor_infoframe_pack() - write a HDMI vendor infoframe to binary buffer * @frame: HDMI infoframe @@ -341,19 +352,11 @@ ssize_t hdmi_vendor_infoframe_pack(struct hdmi_vendor_infoframe *frame, u8 *ptr = buffer; size_t length; - /* empty info frame */ - if (frame->vic == 0 && frame->s3d_struct == HDMI_3D_STRUCTURE_INVALID) - return -EINVAL; - /* only one of those can be supplied */ if (frame->vic != 0 && frame->s3d_struct != HDMI_3D_STRUCTURE_INVALID) return -EINVAL; - /* for side by side (half) we also need to provide 3D_Ext_Data */ - if (frame->s3d_struct >= HDMI_3D_STRUCTURE_SIDE_BY_SIDE_HALF) - frame->length = 6; - else - frame->length = 5; + frame->length = hdmi_vendor_infoframe_length(frame); length = HDMI_INFOFRAME_HEADER_SIZE + frame->length; @@ -372,14 +375,16 @@ ssize_t hdmi_vendor_infoframe_pack(struct hdmi_vendor_infoframe *frame, ptr[5] = 0x0c; ptr[6] = 0x00; - if (frame->vic) { - ptr[7] = 0x1 << 5; /* video format */ - ptr[8] = frame->vic; - } else { + if (frame->s3d_struct != HDMI_3D_STRUCTURE_INVALID) { ptr[7] = 0x2 << 5; /* video format */ ptr[8] = (frame->s3d_struct & 0xf) << 4; if (frame->s3d_struct >= HDMI_3D_STRUCTURE_SIDE_BY_SIDE_HALF) ptr[9] = (frame->s3d_ext_data & 0xf) << 4; + } else if (frame->vic) { + ptr[7] = 0x1 << 5; /* video format */ + ptr[8] = frame->vic; + } else { + ptr[7] = 0x0 << 5; /* video format */ } hdmi_infoframe_set_checksum(buffer, length); @@ -1165,7 +1170,7 @@ hdmi_vendor_any_infoframe_unpack(union hdmi_vendor_any_infoframe *frame, if (ptr[0] != HDMI_INFOFRAME_TYPE_VENDOR || ptr[1] != 1 || - (ptr[2] != 5 && ptr[2] != 6)) + (ptr[2] != 4 && ptr[2] != 5 && ptr[2] != 6)) return -EINVAL; length = ptr[2]; @@ -1193,16 +1198,22 @@ hdmi_vendor_any_infoframe_unpack(union hdmi_vendor_any_infoframe *frame, hvf->length = length; - if (hdmi_video_format == 0x1) { - hvf->vic = ptr[4]; - } else if (hdmi_video_format == 0x2) { + if (hdmi_video_format == 0x2) { + if (length != 5 && length != 6) + return -EINVAL; hvf->s3d_struct = ptr[4] >> 4; if (hvf->s3d_struct >= HDMI_3D_STRUCTURE_SIDE_BY_SIDE_HALF) { - if (length == 6) - hvf->s3d_ext_data = ptr[5] >> 4; - else + if (length != 6) return -EINVAL; + hvf->s3d_ext_data = ptr[5] >> 4; } + } else if (hdmi_video_format == 0x1) { + if (length != 5) + return -EINVAL; + hvf->vic = ptr[4]; + } else { + if (length != 4) + return -EINVAL; } return 0; -- GitLab From f1781e9bb2dd2305d8d7ffbede1888ae22119557 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 13 Nov 2017 19:04:19 +0200 Subject: [PATCH 0400/1471] drm/edid: Allow HDMI infoframe without VIC or S3D MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Appedix F of HDMI 2.0 says that some HDMI sink may fail to switch from 3D to 2D mode in a timely fashion if the source simply stops sending the HDMI infoframe. The suggested workaround is to keep sending the infoframe even when strictly not necessary (ie. no VIC and no S3D). HDMI 1.4 does allow for this behaviour, stating that sending the infoframe is optional in this case. The infoframe was first specified in HDMI 1.4, so in theory sinks predating that may not appreciate us sending an uknown infoframe their way. To avoid regressions let's try to determine if the sink supports the infoframe or not. Unfortunately there's no direct way to do that, so instead we'll just check if we managed to parse any HDMI 1.4 4k or stereo modes from the EDID, and if so we assume the sink will accept the infoframe. Also if the EDID contains the HDMI 2.0 HDMI Forum VSDB we can assume the sink is prepared to receive the infoframe. v2: Fix getting has_hdmi_infoframe from display_info Always fail constructing the infoframe if the display possibly can't handle it Cc: Shashank Sharma Cc: Andrzej Hajda Cc: Laurent Pinchart Signed-off-by: Ville Syrjälä Reviewed-by: Shashank Sharma Link: https://patchwork.freedesktop.org/patch/msgid/20171113170427.4150-3-ville.syrjala@linux.intel.com --- drivers/gpu/drm/bridge/sil-sii8620.c | 3 +- drivers/gpu/drm/bridge/synopsys/dw-hdmi.c | 4 ++- drivers/gpu/drm/drm_edid.c | 34 +++++++++++++++++++---- drivers/gpu/drm/exynos/exynos_hdmi.c | 3 +- drivers/gpu/drm/i915/intel_hdmi.c | 14 ++++++---- drivers/gpu/drm/mediatek/mtk_hdmi.c | 3 +- drivers/gpu/drm/nouveau/nv50_display.c | 3 +- drivers/gpu/drm/rockchip/inno_hdmi.c | 1 + drivers/gpu/drm/sti/sti_hdmi.c | 4 ++- drivers/gpu/drm/zte/zx_hdmi.c | 1 + include/drm/drm_connector.h | 5 ++++ include/drm/drm_edid.h | 1 + 12 files changed, 58 insertions(+), 18 deletions(-) diff --git a/drivers/gpu/drm/bridge/sil-sii8620.c b/drivers/gpu/drm/bridge/sil-sii8620.c index d58db13502bb6..86789f8918a49 100644 --- a/drivers/gpu/drm/bridge/sil-sii8620.c +++ b/drivers/gpu/drm/bridge/sil-sii8620.c @@ -2233,8 +2233,9 @@ static bool sii8620_mode_fixup(struct drm_bridge *bridge, union hdmi_infoframe frm; u8 mhl_vic[] = { 0, 95, 94, 93, 98 }; + /* FIXME: We need the connector here */ drm_hdmi_vendor_infoframe_from_display_mode( - &frm.vendor.hdmi, adjusted_mode); + &frm.vendor.hdmi, NULL, adjusted_mode); vic = frm.vendor.hdmi.vic; if (vic >= ARRAY_SIZE(mhl_vic)) vic = 0; diff --git a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c index a64ce71122881..b172139502d6d 100644 --- a/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c +++ b/drivers/gpu/drm/bridge/synopsys/dw-hdmi.c @@ -1437,7 +1437,9 @@ static void hdmi_config_vendor_specific_infoframe(struct dw_hdmi *hdmi, u8 buffer[10]; ssize_t err; - err = drm_hdmi_vendor_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_vendor_infoframe_from_display_mode(&frame, + &hdmi->connector, + mode); if (err < 0) /* * Going into that statement does not means vendor infoframe diff --git a/drivers/gpu/drm/drm_edid.c b/drivers/gpu/drm/drm_edid.c index 749d07a017729..9ada0ccf50dff 100644 --- a/drivers/gpu/drm/drm_edid.c +++ b/drivers/gpu/drm/drm_edid.c @@ -3393,6 +3393,7 @@ static int do_hdmi_vsdb_modes(struct drm_connector *connector, const u8 *db, u8 len, const u8 *video_db, u8 video_len) { + struct drm_display_info *info = &connector->display_info; int modes = 0, offset = 0, i, multi_present = 0, multi_len; u8 vic_len, hdmi_3d_len = 0; u16 mask; @@ -3520,6 +3521,8 @@ do_hdmi_vsdb_modes(struct drm_connector *connector, const u8 *db, u8 len, } out: + if (modes > 0) + info->has_hdmi_infoframe = true; return modes; } @@ -4243,6 +4246,8 @@ static void drm_parse_hdmi_forum_vsdb(struct drm_connector *connector, struct drm_display_info *display = &connector->display_info; struct drm_hdmi_info *hdmi = &display->hdmi; + display->has_hdmi_infoframe = true; + if (hf_vsdb[6] & 0x80) { hdmi->scdc.supported = true; if (hf_vsdb[6] & 0x40) @@ -4416,6 +4421,7 @@ static void drm_add_display_info(struct drm_connector *connector, info->cea_rev = 0; info->max_tmds_clock = 0; info->dvi_dual = false; + info->has_hdmi_infoframe = false; if (edid->revision < 3) return; @@ -4903,6 +4909,7 @@ s3d_structure_from_display_mode(const struct drm_display_mode *mode) * drm_hdmi_vendor_infoframe_from_display_mode() - fill an HDMI infoframe with * data from a DRM display mode * @frame: HDMI vendor infoframe + * @connector: the connector * @mode: DRM display mode * * Note that there's is a need to send HDMI vendor infoframes only when using a @@ -4913,8 +4920,15 @@ s3d_structure_from_display_mode(const struct drm_display_mode *mode) */ int drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, + struct drm_connector *connector, const struct drm_display_mode *mode) { + /* + * FIXME: sil-sii8620 doesn't have a connector around when + * we need one, so we have to be prepared for a NULL connector. + */ + bool has_hdmi_infoframe = connector ? + connector->display_info.has_hdmi_infoframe : false; int err; u32 s3d_flags; u8 vic; @@ -4922,11 +4936,21 @@ drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, if (!frame || !mode) return -EINVAL; + if (!has_hdmi_infoframe) + return -EINVAL; + vic = drm_match_hdmi_mode(mode); s3d_flags = mode->flags & DRM_MODE_FLAG_3D_MASK; - if (!vic && !s3d_flags) - return -EINVAL; + /* + * Even if it's not absolutely necessary to send the infoframe + * (ie.vic==0 and s3d_struct==0) we will still send it if we + * know that the sink can handle it. This is based on a + * suggestion in HDMI 2.0 Appendix F. Apparently some sinks + * have trouble realizing that they shuld switch from 3D to 2D + * mode if the source simply stops sending the infoframe when + * it wants to switch from 3D to 2D. + */ if (vic && s3d_flags) return -EINVAL; @@ -4935,10 +4959,8 @@ drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, if (err < 0) return err; - if (vic) - frame->vic = vic; - else - frame->s3d_struct = s3d_structure_from_display_mode(mode); + frame->vic = vic; + frame->s3d_struct = s3d_structure_from_display_mode(mode); return 0; } diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c index 82d1b7e2febea..a4b75a46f9463 100644 --- a/drivers/gpu/drm/exynos/exynos_hdmi.c +++ b/drivers/gpu/drm/exynos/exynos_hdmi.c @@ -829,7 +829,8 @@ static void hdmi_reg_infoframes(struct hdmi_context *hdata) DRM_INFO("%s: invalid AVI infoframe (%d)\n", __func__, ret); } - ret = drm_hdmi_vendor_infoframe_from_display_mode(&frm.vendor.hdmi, m); + ret = drm_hdmi_vendor_infoframe_from_display_mode(&frm.vendor.hdmi, + &hdata->connector, m); if (!ret) ret = hdmi_vendor_infoframe_pack(&frm.vendor.hdmi, buf, sizeof(buf)); diff --git a/drivers/gpu/drm/i915/intel_hdmi.c b/drivers/gpu/drm/i915/intel_hdmi.c index 5132dc8147884..59247a49a0773 100644 --- a/drivers/gpu/drm/i915/intel_hdmi.c +++ b/drivers/gpu/drm/i915/intel_hdmi.c @@ -512,12 +512,14 @@ static void intel_hdmi_set_spd_infoframe(struct drm_encoder *encoder, static void intel_hdmi_set_hdmi_infoframe(struct drm_encoder *encoder, - const struct intel_crtc_state *crtc_state) + const struct intel_crtc_state *crtc_state, + const struct drm_connector_state *conn_state) { union hdmi_infoframe frame; int ret; ret = drm_hdmi_vendor_infoframe_from_display_mode(&frame.vendor.hdmi, + conn_state->connector, &crtc_state->base.adjusted_mode); if (ret < 0) return; @@ -584,7 +586,7 @@ static void g4x_set_infoframes(struct drm_encoder *encoder, intel_hdmi_set_avi_infoframe(encoder, crtc_state); intel_hdmi_set_spd_infoframe(encoder, crtc_state); - intel_hdmi_set_hdmi_infoframe(encoder, crtc_state); + intel_hdmi_set_hdmi_infoframe(encoder, crtc_state, conn_state); } static bool hdmi_sink_is_deep_color(const struct drm_connector_state *conn_state) @@ -725,7 +727,7 @@ static void ibx_set_infoframes(struct drm_encoder *encoder, intel_hdmi_set_avi_infoframe(encoder, crtc_state); intel_hdmi_set_spd_infoframe(encoder, crtc_state); - intel_hdmi_set_hdmi_infoframe(encoder, crtc_state); + intel_hdmi_set_hdmi_infoframe(encoder, crtc_state, conn_state); } static void cpt_set_infoframes(struct drm_encoder *encoder, @@ -768,7 +770,7 @@ static void cpt_set_infoframes(struct drm_encoder *encoder, intel_hdmi_set_avi_infoframe(encoder, crtc_state); intel_hdmi_set_spd_infoframe(encoder, crtc_state); - intel_hdmi_set_hdmi_infoframe(encoder, crtc_state); + intel_hdmi_set_hdmi_infoframe(encoder, crtc_state, conn_state); } static void vlv_set_infoframes(struct drm_encoder *encoder, @@ -821,7 +823,7 @@ static void vlv_set_infoframes(struct drm_encoder *encoder, intel_hdmi_set_avi_infoframe(encoder, crtc_state); intel_hdmi_set_spd_infoframe(encoder, crtc_state); - intel_hdmi_set_hdmi_infoframe(encoder, crtc_state); + intel_hdmi_set_hdmi_infoframe(encoder, crtc_state, conn_state); } static void hsw_set_infoframes(struct drm_encoder *encoder, @@ -854,7 +856,7 @@ static void hsw_set_infoframes(struct drm_encoder *encoder, intel_hdmi_set_avi_infoframe(encoder, crtc_state); intel_hdmi_set_spd_infoframe(encoder, crtc_state); - intel_hdmi_set_hdmi_infoframe(encoder, crtc_state); + intel_hdmi_set_hdmi_infoframe(encoder, crtc_state, conn_state); } void intel_dp_dual_mode_set_tmds_output(struct intel_hdmi *hdmi, bool enable) diff --git a/drivers/gpu/drm/mediatek/mtk_hdmi.c b/drivers/gpu/drm/mediatek/mtk_hdmi.c index b787910619839..59a11026dceb4 100644 --- a/drivers/gpu/drm/mediatek/mtk_hdmi.c +++ b/drivers/gpu/drm/mediatek/mtk_hdmi.c @@ -1054,7 +1054,8 @@ static int mtk_hdmi_setup_vendor_specific_infoframe(struct mtk_hdmi *hdmi, u8 buffer[10]; ssize_t err; - err = drm_hdmi_vendor_infoframe_from_display_mode(&frame, mode); + err = drm_hdmi_vendor_infoframe_from_display_mode(&frame, + &hdmi->conn, mode); if (err) { dev_err(hdmi->dev, "Failed to get vendor infoframe from mode: %zd\n", err); diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c index 9c61e767950d6..fdfeae12ef622 100644 --- a/drivers/gpu/drm/nouveau/nv50_display.c +++ b/drivers/gpu/drm/nouveau/nv50_display.c @@ -2756,7 +2756,8 @@ nv50_hdmi_enable(struct drm_encoder *encoder, struct drm_display_mode *mode) = hdmi_infoframe_pack(&avi_frame, args.infoframes, 17); } - ret = drm_hdmi_vendor_infoframe_from_display_mode(&vendor_frame.vendor.hdmi, mode); + ret = drm_hdmi_vendor_infoframe_from_display_mode(&vendor_frame.vendor.hdmi, + &nv_connector->base, mode); if (!ret) { /* We have a Vendor InfoFrame, populate it to the display */ args.pwr.vendor_infoframe_length diff --git a/drivers/gpu/drm/rockchip/inno_hdmi.c b/drivers/gpu/drm/rockchip/inno_hdmi.c index ee584d87111fb..fab30927a8894 100644 --- a/drivers/gpu/drm/rockchip/inno_hdmi.c +++ b/drivers/gpu/drm/rockchip/inno_hdmi.c @@ -282,6 +282,7 @@ static int inno_hdmi_config_video_vsi(struct inno_hdmi *hdmi, int rc; rc = drm_hdmi_vendor_infoframe_from_display_mode(&frame.vendor.hdmi, + &hdmi->connector, mode); return inno_hdmi_upload_frame(hdmi, rc, &frame, INFOFRAME_VSI, diff --git a/drivers/gpu/drm/sti/sti_hdmi.c b/drivers/gpu/drm/sti/sti_hdmi.c index 1c7423f4d423e..4ea1cc1c032e2 100644 --- a/drivers/gpu/drm/sti/sti_hdmi.c +++ b/drivers/gpu/drm/sti/sti_hdmi.c @@ -515,7 +515,9 @@ static int hdmi_vendor_infoframe_config(struct sti_hdmi *hdmi) DRM_DEBUG_DRIVER("\n"); - ret = drm_hdmi_vendor_infoframe_from_display_mode(&infoframe, mode); + ret = drm_hdmi_vendor_infoframe_from_display_mode(&infoframe, + hdmi->drm_connector, + mode); if (ret < 0) { /* * Going into that statement does not means vendor infoframe diff --git a/drivers/gpu/drm/zte/zx_hdmi.c b/drivers/gpu/drm/zte/zx_hdmi.c index b8abb1b496ff5..13ea90f7a185a 100644 --- a/drivers/gpu/drm/zte/zx_hdmi.c +++ b/drivers/gpu/drm/zte/zx_hdmi.c @@ -108,6 +108,7 @@ static int zx_hdmi_config_video_vsi(struct zx_hdmi *hdmi, int ret; ret = drm_hdmi_vendor_infoframe_from_display_mode(&frame.vendor.hdmi, + &hdmi->connector, mode); if (ret) { DRM_DEV_ERROR(hdmi->dev, "failed to get vendor infoframe: %d\n", diff --git a/include/drm/drm_connector.h b/include/drm/drm_connector.h index 2b97d1e28f602..1543212b04496 100644 --- a/include/drm/drm_connector.h +++ b/include/drm/drm_connector.h @@ -269,6 +269,11 @@ struct drm_display_info { */ bool dvi_dual; + /** + * @has_hdmi_infoframe: Does the sink support the HDMI infoframe? + */ + bool has_hdmi_infoframe; + /** * @edid_hdmi_dc_modes: Mask of supported hdmi deep color modes. Even * more stuff redundant with @bus_formats. diff --git a/include/drm/drm_edid.h b/include/drm/drm_edid.h index 9e4e23524840e..3c8740ad1db68 100644 --- a/include/drm/drm_edid.h +++ b/include/drm/drm_edid.h @@ -356,6 +356,7 @@ drm_hdmi_avi_infoframe_from_display_mode(struct hdmi_avi_infoframe *frame, bool is_hdmi2_sink); int drm_hdmi_vendor_infoframe_from_display_mode(struct hdmi_vendor_infoframe *frame, + struct drm_connector *connector, const struct drm_display_mode *mode); void drm_hdmi_avi_infoframe_quant_range(struct hdmi_avi_infoframe *frame, -- GitLab From 62d0fe45297d213ba923e252dcdbcee6f41a3bc6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 22 Nov 2017 14:56:46 +0000 Subject: [PATCH 0401/1471] drm/i915: Remove success dmesg noise for intel_rotate_pages() During selftesting intel_rotate_pages() is very, very verbose without giving us any information. Suppress the noise. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171122145646.1859-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_gem_gtt.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index e101b9a989573..41a203e0c1604 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -3737,9 +3737,6 @@ intel_rotate_pages(struct intel_rotation_info *rot_info, rot_info->plane[i].stride, st, sg); } - DRM_DEBUG_KMS("Created rotated page mapping for object size %zu (%ux%u tiles, %u pages)\n", - obj->base.size, rot_info->plane[0].width, rot_info->plane[0].height, size); - kvfree(page_addr_list); return st; @@ -3749,8 +3746,8 @@ intel_rotate_pages(struct intel_rotation_info *rot_info, err_st_alloc: kvfree(page_addr_list); - DRM_DEBUG_KMS("Failed to create rotated mapping for object size %zu! (%ux%u tiles, %u pages)\n", - obj->base.size, rot_info->plane[0].width, rot_info->plane[0].height, size); + DRM_DEBUG_DRIVER("Failed to create rotated mapping for object size %zu! (%ux%u tiles, %u pages)\n", + obj->base.size, rot_info->plane[0].width, rot_info->plane[0].height, size); return ERR_PTR(ret); } -- GitLab From ee48700dd57d9ce783ec40f035b324d0b75632e4 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 22 Nov 2017 17:26:21 +0000 Subject: [PATCH 0402/1471] drm/i915: Call i915_gem_init_userptr() before taking struct_mutex We don't need struct_mutex to initialise userptr (it just allocates a workqueue for itself etc), but we do need struct_mutex later on in i915_gem_init() in order to feed requests onto the HW. This should break the chain [ 385.697902] ====================================================== [ 385.697907] WARNING: possible circular locking dependency detected [ 385.697913] 4.14.0-CI-Patchwork_7234+ #1 Tainted: G U [ 385.697917] ------------------------------------------------------ [ 385.697922] perf_pmu/2631 is trying to acquire lock: [ 385.697927] (&mm->mmap_sem){++++}, at: [] __might_fault+0x3e/0x90 [ 385.697941] but task is already holding lock: [ 385.697946] (&cpuctx_mutex){+.+.}, at: [] perf_event_ctx_lock_nested+0xbc/0x1d0 [ 385.697957] which lock already depends on the new lock. [ 385.697963] the existing dependency chain (in reverse order) is: [ 385.697970] -> #4 (&cpuctx_mutex){+.+.}: [ 385.697980] __mutex_lock+0x86/0x9b0 [ 385.697985] perf_event_init_cpu+0x5a/0x90 [ 385.697991] perf_event_init+0x178/0x1a4 [ 385.697997] start_kernel+0x27f/0x3f1 [ 385.698003] verify_cpu+0x0/0xfb [ 385.698006] -> #3 (pmus_lock){+.+.}: [ 385.698015] __mutex_lock+0x86/0x9b0 [ 385.698020] perf_event_init_cpu+0x21/0x90 [ 385.698025] cpuhp_invoke_callback+0xca/0xc00 [ 385.698030] _cpu_up+0xa7/0x170 [ 385.698035] do_cpu_up+0x57/0x70 [ 385.698039] smp_init+0x62/0xa6 [ 385.698044] kernel_init_freeable+0x97/0x193 [ 385.698050] kernel_init+0xa/0x100 [ 385.698055] ret_from_fork+0x27/0x40 [ 385.698058] -> #2 (cpu_hotplug_lock.rw_sem){++++}: [ 385.698068] cpus_read_lock+0x39/0xa0 [ 385.698073] apply_workqueue_attrs+0x12/0x50 [ 385.698078] __alloc_workqueue_key+0x1d8/0x4d8 [ 385.698134] i915_gem_init_userptr+0x5f/0x80 [i915] [ 385.698176] i915_gem_init+0x7c/0x390 [i915] [ 385.698213] i915_driver_load+0x99e/0x15c0 [i915] [ 385.698250] i915_pci_probe+0x33/0x90 [i915] [ 385.698256] pci_device_probe+0xa1/0x130 [ 385.698262] driver_probe_device+0x293/0x440 [ 385.698267] __driver_attach+0xde/0xe0 [ 385.698272] bus_for_each_dev+0x5c/0x90 [ 385.698277] bus_add_driver+0x16d/0x260 [ 385.698282] driver_register+0x57/0xc0 [ 385.698287] do_one_initcall+0x3e/0x160 [ 385.698292] do_init_module+0x5b/0x1fa [ 385.698297] load_module+0x2374/0x2dc0 [ 385.698302] SyS_finit_module+0xaa/0xe0 [ 385.698307] entry_SYSCALL_64_fastpath+0x1c/0xb1 [ 385.698311] -> #1 (&dev->struct_mutex){+.+.}: [ 385.698320] __mutex_lock+0x86/0x9b0 [ 385.698361] i915_mutex_lock_interruptible+0x4c/0x130 [i915] [ 385.698403] i915_gem_fault+0x206/0x760 [i915] [ 385.698409] __do_fault+0x1a/0x70 [ 385.698413] __handle_mm_fault+0x7c4/0xdb0 [ 385.698417] handle_mm_fault+0x154/0x300 [ 385.698440] __do_page_fault+0x2d6/0x570 [ 385.698445] page_fault+0x22/0x30 [ 385.698449] -> #0 (&mm->mmap_sem){++++}: [ 385.698459] lock_acquire+0xaf/0x200 [ 385.698464] __might_fault+0x68/0x90 [ 385.698470] _copy_to_user+0x1e/0x70 [ 385.698475] perf_read+0x1aa/0x290 [ 385.698480] __vfs_read+0x23/0x120 [ 385.698484] vfs_read+0xa3/0x150 [ 385.698488] SyS_read+0x45/0xb0 [ 385.698493] entry_SYSCALL_64_fastpath+0x1c/0xb1 [ 385.698497] other info that might help us debug this: [ 385.698505] Chain exists of: &mm->mmap_sem --> pmus_lock --> &cpuctx_mutex [ 385.698517] Possible unsafe locking scenario: [ 385.698522] CPU0 CPU1 [ 385.698526] ---- ---- [ 385.698529] lock(&cpuctx_mutex); [ 385.698553] lock(pmus_lock); [ 385.698558] lock(&cpuctx_mutex); [ 385.698564] lock(&mm->mmap_sem); [ 385.698568] *** DEADLOCK *** [ 385.698574] 1 lock held by perf_pmu/2631: [ 385.698578] #0: (&cpuctx_mutex){+.+.}, at: [] perf_event_ctx_lock_nested+0xbc/0x1d0 [ 385.698589] stack backtrace: [ 385.698595] CPU: 3 PID: 2631 Comm: perf_pmu Tainted: G U 4.14.0-CI-Patchwork_7234+ #1 [ 385.698602] Hardware name: /NUC6CAYB, BIOS AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017 [ 385.698609] Call Trace: [ 385.698615] dump_stack+0x5f/0x86 [ 385.698621] print_circular_bug.isra.18+0x1d0/0x2c0 [ 385.698627] __lock_acquire+0x19c3/0x1b60 [ 385.698634] ? generic_exec_single+0x77/0xe0 [ 385.698640] ? lock_acquire+0xaf/0x200 [ 385.698644] lock_acquire+0xaf/0x200 [ 385.698650] ? __might_fault+0x3e/0x90 [ 385.698655] __might_fault+0x68/0x90 [ 385.698660] ? __might_fault+0x3e/0x90 [ 385.698665] _copy_to_user+0x1e/0x70 [ 385.698670] perf_read+0x1aa/0x290 [ 385.698675] __vfs_read+0x23/0x120 [ 385.698682] ? __fget+0x101/0x1f0 [ 385.698686] vfs_read+0xa3/0x150 [ 385.698691] SyS_read+0x45/0xb0 [ 385.698696] entry_SYSCALL_64_fastpath+0x1c/0xb1 [ 385.698701] RIP: 0033:0x7ff1c46876ed [ 385.698705] RSP: 002b:00007fff13552f90 EFLAGS: 00000293 ORIG_RAX: 0000000000000000 [ 385.698712] RAX: ffffffffffffffda RBX: ffffc90000647ff0 RCX: 00007ff1c46876ed [ 385.698718] RDX: 0000000000000010 RSI: 00007fff13552fa0 RDI: 0000000000000005 [ 385.698723] RBP: 000056063d300580 R08: 0000000000000000 R09: 0000000000000060 [ 385.698729] R10: 0000000000000000 R11: 0000000000000293 R12: 0000000000000046 [ 385.698734] R13: 00007fff13552c6f R14: 00007ff1c6279d00 R15: 00007ff1c6279a40 Testcase: igt/perf_pmu Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171122172621.16158-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_gem.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 21ca680e9e633..e03d6c2554e2b 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -5116,8 +5116,6 @@ int i915_gem_init(struct drm_i915_private *dev_priv) { int ret; - mutex_lock(&dev_priv->drm.struct_mutex); - /* * We need to fallback to 4K pages since gvt gtt handling doesn't * support huge page entries - we will need to check either hypervisor @@ -5137,18 +5135,19 @@ int i915_gem_init(struct drm_i915_private *dev_priv) dev_priv->gt.cleanup_engine = intel_engine_cleanup; } + ret = i915_gem_init_userptr(dev_priv); + if (ret) + return ret; + /* This is just a security blanket to placate dragons. * On some systems, we very sporadically observe that the first TLBs * used by the CS may be stale, despite us poking the TLB reset. If * we hold the forcewake during initialisation these problems * just magically go away. */ + mutex_lock(&dev_priv->drm.struct_mutex); intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL); - ret = i915_gem_init_userptr(dev_priv); - if (ret) - goto out_unlock; - ret = i915_gem_init_ggtt(dev_priv); if (ret) goto out_unlock; -- GitLab From b4e3c935b2313b3e557066b630efd0367c7fb34c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 22 Nov 2017 22:25:10 +0000 Subject: [PATCH 0403/1471] drm/i915: Save/restore irq state for vlv_residency_raw() Since commit 6060b6aec03c ("drm/i915/pmu: Add RC6 residency metrics"), vlv_residency_raw() may be called from an irq-disabled context (via perf event sampling on remote cpu). As such, we can no longer assume that we are called from process context and must save/restore the irq state for the spinlock. Fixes: 6060b6aec03c ("drm/i915/pmu: Add RC6 residency metrics") Testcase: igt/perf_pmu/other-init-3 Signed-off-by: Chris Wilson Cc: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171122222510.22627-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/intel_pm.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 7d41aad791665..03d67d8ab647d 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -9396,12 +9396,13 @@ static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, const i915_reg_t reg) { u32 lower, upper, tmp; + unsigned long flags; int loop = 2; /* The register accessed do not need forcewake. We borrow * uncore lock to prevent concurrent access to range reg. */ - spin_lock_irq(&dev_priv->uncore.lock); + spin_lock_irqsave(&dev_priv->uncore.lock, flags); /* vlv and chv residency counters are 40 bits in width. * With a control bit, we can choose between upper or lower @@ -9432,7 +9433,7 @@ static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, * now. */ - spin_unlock_irq(&dev_priv->uncore.lock); + spin_unlock_irqrestore(&dev_priv->uncore.lock, flags); return lower | (u64)upper << 8; } @@ -9451,7 +9452,6 @@ u64 intel_rc6_residency_ns(struct drm_i915_private *dev_priv, mul = 1000000; div = dev_priv->czclk_freq; time_hw = vlv_residency_raw(dev_priv, reg); - } else { /* 833.33ns units on Gen9LP, 1.28us elsewhere. */ if (IS_GEN9_LP(dev_priv)) { -- GitLab From 4f0aa1fa3e3849caee450ee5d14fcc289cf16703 Mon Sep 17 00:00:00 2001 From: Anusha Srivatsa Date: Thu, 9 Nov 2017 10:51:43 -0800 Subject: [PATCH 0404/1471] drm/i915/dmc: DMC 1.04 for Kabylake There is a new version of DMC available for KBL. The release notes mentions: 1. Fix for the issue where DC_STATE was getting enabled even when disabled by driver causing data corruption. v2: Remove pull request from commit message (Rodrigo). Cc: Rodrigo Vivi Signed-off-by: Anusha Srivatsa Reviewed-by: Rodrigo Vivi Signed-off-by: Jani Nikula Link: https://patchwork.freedesktop.org/patch/msgid/1510253503-12634-1-git-send-email-anusha.srivatsa@intel.com --- drivers/gpu/drm/i915/intel_csr.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c index 77d8b3d483ca2..07e4f7bc44127 100644 --- a/drivers/gpu/drm/i915/intel_csr.c +++ b/drivers/gpu/drm/i915/intel_csr.c @@ -40,9 +40,9 @@ #define I915_CSR_CNL "i915/cnl_dmc_ver1_06.bin" #define CNL_CSR_VERSION_REQUIRED CSR_VERSION(1, 6) -#define I915_CSR_KBL "i915/kbl_dmc_ver1_01.bin" +#define I915_CSR_KBL "i915/kbl_dmc_ver1_04.bin" MODULE_FIRMWARE(I915_CSR_KBL); -#define KBL_CSR_VERSION_REQUIRED CSR_VERSION(1, 1) +#define KBL_CSR_VERSION_REQUIRED CSR_VERSION(1, 4) #define I915_CSR_SKL "i915/skl_dmc_ver1_26.bin" MODULE_FIRMWARE(I915_CSR_SKL); -- GitLab From e2b155e9924c39c074339cf77e22a0a4cead66e0 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 08:40:46 +0000 Subject: [PATCH 0405/1471] drm/printer: Add drm_vprintf() Simple va_args equivalent to the existing drm_printf() for use with the drm_printer. v2: Fixup kerneldoc to match final parameter names. v3: Turn it into a kerneldoc comment Signed-off-by: Chris Wilson Cc: Rob Clark Reviewed-by: Daniel Vetter Cc: Dave Airlie Acked-by: Dave Airlie Signed-off-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171123084051.30203-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/drm_print.c | 5 +---- include/drm/drm_print.h | 15 +++++++++++++++ 2 files changed, 16 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/drm_print.c b/drivers/gpu/drm/drm_print.c index 82ff327eb2df8..781518fd88e36 100644 --- a/drivers/gpu/drm/drm_print.c +++ b/drivers/gpu/drm/drm_print.c @@ -55,13 +55,10 @@ EXPORT_SYMBOL(__drm_printfn_debug); */ void drm_printf(struct drm_printer *p, const char *f, ...) { - struct va_format vaf; va_list args; va_start(args, f); - vaf.fmt = f; - vaf.va = &args; - p->printfn(p, &vaf); + drm_vprintf(p, f, &args); va_end(args); } EXPORT_SYMBOL(drm_printf); diff --git a/include/drm/drm_print.h b/include/drm/drm_print.h index 0968e411f5627..5f9932e2246e1 100644 --- a/include/drm/drm_print.h +++ b/include/drm/drm_print.h @@ -80,6 +80,21 @@ void __drm_printfn_debug(struct drm_printer *p, struct va_format *vaf); __printf(2, 3) void drm_printf(struct drm_printer *p, const char *f, ...); +/** + * drm_vprintf - print to a &drm_printer stream + * @p: the &drm_printer + * @fmt: format string + * @va: the va_list + */ +__printf(2, 0) +static inline void +drm_vprintf(struct drm_printer *p, const char *fmt, va_list *va) +{ + struct va_format vaf = { .fmt = fmt, .va = va }; + + p->printfn(p, &vaf); +} + /** * drm_printf_indent - Print to a &drm_printer stream with indentation * @printer: DRM printer -- GitLab From 298b6a815bba69c12336114c546104ba9d25bc8f Mon Sep 17 00:00:00 2001 From: Vasyl Gomonovych Date: Wed, 22 Nov 2017 16:22:41 +0100 Subject: [PATCH 0406/1471] dma-buf: Fix ifnullfree.cocci warnings NULL check before some freeing functions is not needed. drivers/dma-buf/dma-buf.c:1183:2-26: WARNING: NULL check before freeing debugfs_remove_recursive Generated by: scripts/coccinelle/free/ifnullfree.cocci Signed-off-by: Vasyl Gomonovych Signed-off-by: Sumit Semwal Link: https://patchwork.freedesktop.org/patch/msgid/1511364161-6074-1-git-send-email-gomonovych@gmail.com --- drivers/dma-buf/dma-buf.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c index 1792385405f0e..058805b6d1640 100644 --- a/drivers/dma-buf/dma-buf.c +++ b/drivers/dma-buf/dma-buf.c @@ -1179,8 +1179,7 @@ static int dma_buf_init_debugfs(void) static void dma_buf_uninit_debugfs(void) { - if (dma_buf_debugfs_dir) - debugfs_remove_recursive(dma_buf_debugfs_dir); + debugfs_remove_recursive(dma_buf_debugfs_dir); } #else static inline int dma_buf_init_debugfs(void) -- GitLab From b552ae444e454eb3254c958e05b69820c0ef346d Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Thu, 23 Nov 2017 10:07:01 +0000 Subject: [PATCH 0407/1471] drm/i915/pmu: Drop I915_ENGINE_SAMPLE_MAX from uapi headers We have agreed during the engine classes discussion that fields marked as non-ABI are better left out altogether from uapi headers. v2: Use a local define for maintanability. (Chris Wilson) Signed-off-by: Tvrtko Ursulin Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171123100701.18430-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + include/uapi/drm/i915_drm.h | 3 +-- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 43473e6d1a4f8..d38d059285dca 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -358,6 +358,7 @@ struct intel_engine_cs { * * Our internal timer stores the current counters in this field. */ +#define I915_ENGINE_SAMPLE_MAX (I915_SAMPLE_SEMA + 1) struct i915_pmu_sample sample[I915_ENGINE_SAMPLE_MAX]; /** * @busy_stats: Has enablement of engine stats tracking been diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 915a6e85a8550..239e8633edc91 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -110,8 +110,7 @@ enum drm_i915_gem_engine_class { enum drm_i915_pmu_engine_sample { I915_SAMPLE_BUSY = 0, I915_SAMPLE_WAIT = 1, - I915_SAMPLE_SEMA = 2, - I915_ENGINE_SAMPLE_MAX /* non-ABI */ + I915_SAMPLE_SEMA = 2 }; #define I915_PMU_SAMPLE_BITS (4) -- GitLab From fbba5559d91e812db0fd446fb383a37fde48b1cb Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Thu, 23 Nov 2017 10:26:54 +0000 Subject: [PATCH 0408/1471] drm/i915/pmu: Clear the previous sample value when parking When turning off the engines, and the pmu sampling, clear the previous value as the current measurement should be 0. v2: Use a for-loop v3: * Move clearing to timer self-dis-arm to avoid race with parking. * Clear frequency samples as well. v4: * Init frequency to idle_freq. (Chris Wilson) Signed-off-by: Tvrtko Ursulin Suggested-by: Chris Wilson Reviewed-by: Chris Wilson (v3) Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171123102654.29296-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 1071935bfa67c..6a42e7f7967dd 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -262,13 +262,31 @@ static void frequency_sample(struct drm_i915_private *dev_priv) } } +static void pmu_init_previous_samples(struct drm_i915_private *i915) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + unsigned int i; + + for_each_engine(engine, i915, id) { + for (i = 0; i < ARRAY_SIZE(engine->pmu.sample); i++) + engine->pmu.sample[i].prev = 0; + } + + for (i = 0; i < ARRAY_SIZE(i915->pmu.sample); i++) + i915->pmu.sample[i].prev = i915->gt_pm.rps.idle_freq; +} + static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer) { struct drm_i915_private *i915 = container_of(hrtimer, struct drm_i915_private, pmu.timer); - if (!READ_ONCE(i915->pmu.timer_enabled)) + if (!READ_ONCE(i915->pmu.timer_enabled)) { + pmu_init_previous_samples(i915); + return HRTIMER_NORESTART; + } engines_sample(i915); frequency_sample(i915); @@ -839,6 +857,8 @@ void i915_pmu_register(struct drm_i915_private *i915) hrtimer_init(&i915->pmu.timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); i915->pmu.timer.function = i915_sample; + pmu_init_previous_samples(i915); + for_each_engine(engine, i915, id) INIT_DELAYED_WORK(&engine->pmu.disable_busy_stats, __disable_busy_stats); -- GitLab From 42e5e65765265485ecf2a480c244d76c2c624449 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 13 Nov 2017 17:01:40 +0100 Subject: [PATCH 0409/1471] drm/i915: sync dp link status checks against atomic commmits MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Two bits: - check actual atomic state, the legacy stuff can only be looked at from within the atomic_commit_tail function, since it's only protected by ordering and not by any locks. - Make sure we don't wreak the work an ongoing nonblocking commit is doing. v2: We need the crtc lock too, because a plane update might change it without having to acquire the connection_mutex (Maarten). Use Maarten's changes for this locking, while keeping the logic that uses the connection->commit->hw_done signal for syncing with nonblocking commits. v3: The initial state objects from the hw state readout do not have a commit object. Check for that (spotted by CI). v4: Fix deadlock from jumping to put_power with locks still held. (mlankhorst) Cc: Manasi Navare Cc: Maarten Lankhorst Cc: Ville Syrjälä References: https://bugs.freedesktop.org/show_bug.cgi?id=103336 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99272 Signed-off-by: Daniel Vetter Signed-off-by: Maarten Lankhorst Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171113160140.22679-1-maarten.lankhorst@linux.intel.com --- drivers/gpu/drm/i915/intel_dp.c | 61 +++++++++++++++++++++++++++++---- 1 file changed, 55 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index bc61f38b131df..dd0b3a0187926 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -4277,6 +4277,8 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) { struct drm_i915_private *dev_priv = to_i915(intel_dp_to_dev(intel_dp)); struct intel_encoder *intel_encoder = &dp_to_dig_port(intel_dp)->base; + struct drm_connector_state *conn_state = + intel_dp->attached_connector->base.state; u8 link_status[DP_LINK_STATUS_SIZE]; WARN_ON(!drm_modeset_is_locked(&dev_priv->drm.mode_config.connection_mutex)); @@ -4286,10 +4288,16 @@ intel_dp_check_link_status(struct intel_dp *intel_dp) return; } - if (!intel_encoder->base.crtc) + if (!conn_state->crtc) return; - if (!to_intel_crtc(intel_encoder->base.crtc)->active) + WARN_ON(!drm_modeset_is_locked(&conn_state->crtc->mutex)); + + if (!conn_state->crtc->state->active) + return; + + if (conn_state->commit && + !try_wait_for_completion(&conn_state->commit->hw_done)) return; /* @@ -4364,9 +4372,8 @@ intel_dp_short_pulse(struct intel_dp *intel_dp) DRM_DEBUG_DRIVER("CP or sink specific irq unhandled\n"); } - drm_modeset_lock(&dev_priv->drm.mode_config.connection_mutex, NULL); intel_dp_check_link_status(intel_dp); - drm_modeset_unlock(&dev_priv->drm.mode_config.connection_mutex); + if (intel_dp->compliance.test_type == DP_TEST_LINK_TRAINING) { DRM_DEBUG_KMS("Link Training Compliance Test requested\n"); /* Send a Hotplug Uevent to userspace to start modeset */ @@ -4814,8 +4821,19 @@ intel_dp_detect(struct drm_connector *connector, connector->base.id, connector->name); /* If full detect is not performed yet, do a full detect */ - if (!intel_dp->detect_done) + if (!intel_dp->detect_done) { + struct drm_crtc *crtc; + int ret; + + crtc = connector->state->crtc; + if (crtc) { + ret = drm_modeset_lock(&crtc->mutex, ctx); + if (ret) + return ret; + } + status = intel_dp_long_pulse(intel_dp->attached_connector); + } intel_dp->detect_done = false; @@ -5097,7 +5115,38 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd) } if (!intel_dp->is_mst) { - if (!intel_dp_short_pulse(intel_dp)) { + struct drm_modeset_acquire_ctx ctx; + struct drm_connector *connector = &intel_dp->attached_connector->base; + struct drm_crtc *crtc; + int iret; + bool handled = false; + + drm_modeset_acquire_init(&ctx, 0); +retry: + iret = drm_modeset_lock(&dev_priv->drm.mode_config.connection_mutex, &ctx); + if (iret) + goto err; + + crtc = connector->state->crtc; + if (crtc) { + iret = drm_modeset_lock(&crtc->mutex, &ctx); + if (iret) + goto err; + } + + handled = intel_dp_short_pulse(intel_dp); + +err: + if (iret == -EDEADLK) { + drm_modeset_backoff(&ctx); + goto retry; + } + + drm_modeset_drop_locks(&ctx); + drm_modeset_acquire_fini(&ctx); + WARN(iret, "Acquiring modeset locks failed with %i\n", iret); + + if (!handled) { intel_dp->detect_done = false; goto put_power; } -- GitLab From 6a44e1772177ed9f987447bb6c601ee7451eeb52 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 21 Nov 2017 10:42:41 +0100 Subject: [PATCH 0410/1471] drm/i915: remove stale comment from sanitize_encoder MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This goes back to pre-atomic, where due to intermediate dpms states connectors and encoder states might indeed not have matched. With atomic that's all smashed together (and hopefully no bios ever enables a vga output in dpms standby/suspedn state or we're toast). In commit 873ffe69a9097fb241fff2967ea6f0bf2c179195 Author: Maarten Lankhorst Date: Wed Aug 5 12:37:07 2015 +0200 drm/i915: Remove connectors_active from sanitization, v2. sanitize_encoders was changed to disable the encoder in all cases, which made the comment obsolete. Remove the misleading comment. Reported-by: Ville Syrjälä Cc: Ville Syrjälä Cc: Maarten Lankhorst Reviewed-by: Maarten Lankhorst Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171121094241.9129-1-daniel.vetter@ffwll.ch --- drivers/gpu/drm/i915/intel_display.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5ca7b33554ef4..fec267dfc0623 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -14909,8 +14909,6 @@ static void intel_sanitize_encoder(struct intel_encoder *encoder) connector->base.dpms = DRM_MODE_DPMS_OFF; connector->base.encoder = NULL; } - /* Enabled encoders without active connectors will be fixed in - * the crtc fixup. */ } void i915_redisable_vga_power_on(struct drm_i915_private *dev_priv) -- GitLab From 621d07b20eb8e7c47381178fc0f5724caa49afcd Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Thu, 23 Nov 2017 13:54:20 +0000 Subject: [PATCH 0411/1471] drm/i915/selftests: rein in igt_write_huge Rather than repeat the test for each engine, which takes a long time, let's try alternating between the engines in some randomized order. v2: fix gen2 blunder fix !order blunder more cunning permutation construction! Suggested-by: Chris Wilson Signed-off-by: Matthew Auld Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171123135421.17967-1-matthew.auld@intel.com --- drivers/gpu/drm/i915/selftests/huge_pages.c | 111 ++++++++++++-------- 1 file changed, 66 insertions(+), 45 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index db7a0a1f2960e..83b3a27370a47 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -27,6 +27,7 @@ #include #include "mock_drm.h" +#include "i915_random.h" static const unsigned int page_sizes[] = { I915_GTT_PAGE_SIZE_2M, @@ -1044,7 +1045,10 @@ static int igt_write_huge(struct i915_gem_context *ctx, { struct drm_i915_private *i915 = to_i915(obj->base.dev); struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base; + static struct intel_engine_cs *engines[I915_NUM_ENGINES]; struct intel_engine_cs *engine; + I915_RND_STATE(prng); + IGT_TIMEOUT(end_time); struct i915_vma *vma; unsigned int flags = PIN_USER | PIN_OFFSET_FIXED; unsigned int max_page_size; @@ -1052,6 +1056,8 @@ static int igt_write_huge(struct i915_gem_context *ctx, u64 max; u64 num; u64 size; + int *order; + int i, n; int err = 0; GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); @@ -1067,67 +1073,81 @@ static int igt_write_huge(struct i915_gem_context *ctx, if (IS_ERR(vma)) return PTR_ERR(vma); + n = 0; for_each_engine(engine, i915, id) { - IGT_TIMEOUT(end_time); - if (!intel_engine_can_store_dword(engine)) { - pr_info("store-dword-imm not supported on engine=%u\n", - id); + pr_info("store-dword-imm not supported on engine=%u\n", id); continue; } + engines[n++] = engine; + } - /* - * Try various offsets until we timeout -- we want to avoid - * issues hidden by effectively always using offset = 0. - */ - for_each_prime_number_from(num, 0, max) { - u64 offset = num * max_page_size; - u32 dword; + if (!n) + return 0; - err = i915_vma_unbind(vma); - if (err) - goto out_vma_close; + /* + * To keep things interesting when alternating between engines in our + * randomized order, lets also make feeding to the same engine a few + * times in succession a possibility by enlarging the permutation array. + */ + order = i915_random_order(n * I915_NUM_ENGINES, &prng); + if (!order) + return -ENOMEM; - err = i915_vma_pin(vma, size, max_page_size, flags | offset); - if (err) { - /* - * The ggtt may have some pages reserved so - * refrain from erroring out. - */ - if (err == -ENOSPC && i915_is_ggtt(vm)) { - err = 0; - continue; - } + /* + * Try various offsets until we timeout -- we want to avoid + * issues hidden by effectively always using offset = 0. + */ + i = 0; + for_each_prime_number_from(num, 0, max) { + u64 offset = num * max_page_size; + u32 dword; - goto out_vma_close; + err = i915_vma_unbind(vma); + if (err) + goto out_vma_close; + + err = i915_vma_pin(vma, size, max_page_size, flags | offset); + if (err) { + /* + * The ggtt may have some pages reserved so + * refrain from erroring out. + */ + if (err == -ENOSPC && i915_is_ggtt(vm)) { + err = 0; + continue; } - err = igt_check_page_sizes(vma); - if (err) - goto out_vma_unpin; + goto out_vma_close; + } - dword = offset_in_page(num) / 4; + err = igt_check_page_sizes(vma); + if (err) + goto out_vma_unpin; - err = gpu_write(vma, ctx, engine, dword, num + 1); - if (err) { - pr_err("gpu-write failed at offset=%llx", offset); - goto out_vma_unpin; - } + dword = offset_in_page(num) / 4; - err = cpu_check(obj, dword, num + 1); - if (err) { - pr_err("cpu-check failed at offset=%llx", offset); - goto out_vma_unpin; - } + engine = engines[order[i] % n]; + i = (i + 1) % (n * I915_NUM_ENGINES); - i915_vma_unpin(vma); + err = gpu_write(vma, ctx, engine, dword, num + 1); + if (err) { + pr_err("gpu-write failed at offset=%llx", offset); + goto out_vma_unpin; + } - if (num > 0 && - igt_timeout(end_time, - "%s timed out on engine=%u at offset=%llx, max_page_size=%x\n", - __func__, id, offset, max_page_size)) - break; + err = cpu_check(obj, dword, num + 1); + if (err) { + pr_err("cpu-check failed at offset=%llx", offset); + goto out_vma_unpin; } + + i915_vma_unpin(vma); + + if (igt_timeout(end_time, + "%s timed out on engine=%u at offset=%llx, max_page_size=%x\n", + __func__, engine->id, offset, max_page_size)) + break; } out_vma_unpin: @@ -1135,6 +1155,7 @@ static int igt_write_huge(struct i915_gem_context *ctx, i915_vma_unpin(vma); out_vma_close: i915_vma_close(vma); + kfree(order); return err; } -- GitLab From c83a8d4a2ec93c53f78e109dee5e21953485cbfd Mon Sep 17 00:00:00 2001 From: Matthew Auld Date: Thu, 23 Nov 2017 13:54:21 +0000 Subject: [PATCH 0412/1471] drm/i915/selftests: test descending addresses For igt_write_huge make sure the higher gtt offsets don't feel left out, which is especially true when dealing with the 48b PPGTT, where we timeout long before we are able exhaust the address space. v2: just use IGT_TIMEOUT Suggested-by: Chris Wilson Signed-off-by: Matthew Auld Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171123135421.17967-2-matthew.auld@intel.com --- drivers/gpu/drm/i915/selftests/huge_pages.c | 123 +++++++++++--------- 1 file changed, 70 insertions(+), 53 deletions(-) diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index 83b3a27370a47..df1befe2c7938 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -1040,6 +1040,62 @@ static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val) return err; } +static int __igt_write_huge(struct i915_gem_context *ctx, + struct intel_engine_cs *engine, + struct drm_i915_gem_object *obj, + u64 size, u64 offset, + u32 dword, u32 val) +{ + struct drm_i915_private *i915 = to_i915(obj->base.dev); + struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base; + unsigned int flags = PIN_USER | PIN_OFFSET_FIXED; + struct i915_vma *vma; + int err; + + vma = i915_vma_instance(obj, vm, NULL); + if (IS_ERR(vma)) + return PTR_ERR(vma); + + err = i915_vma_unbind(vma); + if (err) + goto out_vma_close; + + err = i915_vma_pin(vma, size, 0, flags | offset); + if (err) { + /* + * The ggtt may have some pages reserved so + * refrain from erroring out. + */ + if (err == -ENOSPC && i915_is_ggtt(vm)) + err = 0; + + goto out_vma_close; + } + + err = igt_check_page_sizes(vma); + if (err) + goto out_vma_unpin; + + err = gpu_write(vma, ctx, engine, dword, val); + if (err) { + pr_err("gpu-write failed at offset=%llx\n", offset); + goto out_vma_unpin; + } + + err = cpu_check(obj, dword, val); + if (err) { + pr_err("cpu-check failed at offset=%llx\n", offset); + goto out_vma_unpin; + } + +out_vma_unpin: + i915_vma_unpin(vma); +out_vma_close: + i915_vma_close(vma); + + return err; +} + static int igt_write_huge(struct i915_gem_context *ctx, struct drm_i915_gem_object *obj) { @@ -1049,8 +1105,6 @@ static int igt_write_huge(struct i915_gem_context *ctx, struct intel_engine_cs *engine; I915_RND_STATE(prng); IGT_TIMEOUT(end_time); - struct i915_vma *vma; - unsigned int flags = PIN_USER | PIN_OFFSET_FIXED; unsigned int max_page_size; unsigned int id; u64 max; @@ -1069,10 +1123,6 @@ static int igt_write_huge(struct i915_gem_context *ctx, max_page_size = rounddown_pow_of_two(obj->mm.page_sizes.sg); max = div_u64((vm->total - size), max_page_size); - vma = i915_vma_instance(obj, vm, NULL); - if (IS_ERR(vma)) - return PTR_ERR(vma); - n = 0; for_each_engine(engine, i915, id) { if (!intel_engine_can_store_dword(engine)) { @@ -1095,66 +1145,33 @@ static int igt_write_huge(struct i915_gem_context *ctx, return -ENOMEM; /* - * Try various offsets until we timeout -- we want to avoid - * issues hidden by effectively always using offset = 0. + * Try various offsets in an ascending/descending fashion until we + * timeout -- we want to avoid issues hidden by effectively always using + * offset = 0. */ i = 0; for_each_prime_number_from(num, 0, max) { - u64 offset = num * max_page_size; - u32 dword; - - err = i915_vma_unbind(vma); - if (err) - goto out_vma_close; - - err = i915_vma_pin(vma, size, max_page_size, flags | offset); - if (err) { - /* - * The ggtt may have some pages reserved so - * refrain from erroring out. - */ - if (err == -ENOSPC && i915_is_ggtt(vm)) { - err = 0; - continue; - } - - goto out_vma_close; - } - - err = igt_check_page_sizes(vma); - if (err) - goto out_vma_unpin; - - dword = offset_in_page(num) / 4; + u64 offset_low = num * max_page_size; + u64 offset_high = (max - num) * max_page_size; + u32 dword = offset_in_page(num) / 4; engine = engines[order[i] % n]; i = (i + 1) % (n * I915_NUM_ENGINES); - err = gpu_write(vma, ctx, engine, dword, num + 1); - if (err) { - pr_err("gpu-write failed at offset=%llx", offset); - goto out_vma_unpin; - } - - err = cpu_check(obj, dword, num + 1); - if (err) { - pr_err("cpu-check failed at offset=%llx", offset); - goto out_vma_unpin; - } + err = __igt_write_huge(ctx, engine, obj, size, offset_low, dword, num + 1); + if (err) + break; - i915_vma_unpin(vma); + err = __igt_write_huge(ctx, engine, obj, size, offset_high, dword, num + 1); + if (err) + break; if (igt_timeout(end_time, - "%s timed out on engine=%u at offset=%llx, max_page_size=%x\n", - __func__, engine->id, offset, max_page_size)) + "%s timed out on engine=%u, offset_low=%llx offset_high=%llx, max_page_size=%x\n", + __func__, engine->id, offset_low, offset_high, max_page_size)) break; } -out_vma_unpin: - if (i915_vma_is_pinned(vma)) - i915_vma_unpin(vma); -out_vma_close: - i915_vma_close(vma); kfree(order); return err; -- GitLab From b1c24a6137af11fca49192a42face03cacbd7fc5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 15:26:30 +0000 Subject: [PATCH 0413/1471] drm/i915: Unwind incomplete legacy context switches The legacy context switch for ringbuffer submission is multistaged, where each of those stages may fail. However, we were updating global state after some stages, and so we had to force the incomplete request to be submitted because we could not unwind. Save the global state before performing the switches, and so enable us to unwind back to the previous global state should any phase fail. We then must cancel the request instead of submitting it should the construction fail. v2: s/saved_ctx/from_ctx/; s/ctx/to_ctx/ etc. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171123152631.31385-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem_context.c | 173 ++++++++---------------- drivers/gpu/drm/i915/i915_gem_request.c | 18 +-- drivers/gpu/drm/i915/intel_ringbuffer.c | 1 + drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + 4 files changed, 65 insertions(+), 128 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index 6ca56e482d793..da4a64b3484ca 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -507,6 +507,7 @@ void i915_gem_contexts_lost(struct drm_i915_private *dev_priv) for_each_engine(engine, dev_priv, id) { engine->legacy_active_context = NULL; + engine->legacy_active_ppgtt = NULL; if (!engine->last_retired_context) continue; @@ -681,68 +682,49 @@ static int remap_l3(struct drm_i915_gem_request *req, int slice) return 0; } -static inline bool skip_rcs_switch(struct i915_hw_ppgtt *ppgtt, - struct intel_engine_cs *engine, - struct i915_gem_context *to) -{ - if (to->remap_slice) - return false; - - if (ppgtt && (intel_engine_flag(engine) & ppgtt->pd_dirty_rings)) - return false; - - return to == engine->legacy_active_context; -} - -static bool -needs_pd_load_pre(struct i915_hw_ppgtt *ppgtt, struct intel_engine_cs *engine) -{ - struct i915_gem_context *from = engine->legacy_active_context; - - if (!ppgtt) - return false; - - /* Always load the ppgtt on first use */ - if (!from) - return true; - - /* Same context without new entries, skip */ - if ((!from->ppgtt || from->ppgtt == ppgtt) && - !(intel_engine_flag(engine) & ppgtt->pd_dirty_rings)) - return false; - - if (engine->id != RCS) - return true; - - return true; -} - -static int do_rcs_switch(struct drm_i915_gem_request *req) +/** + * i915_switch_context() - perform a GPU context switch. + * @rq: request for which we'll execute the context switch + * + * The context life cycle is simple. The context refcount is incremented and + * decremented by 1 on create and destroy. If the context is in use by the GPU, + * it will have a refcount > 1. This allows us to destroy the context abstract + * object while letting the normal object tracking destroy the backing BO. + * + * This function should not be used in execlists mode. Instead the context is + * switched by writing to the ELSP and requests keep a reference to their + * context. + */ +int i915_switch_context(struct drm_i915_gem_request *rq) { - struct i915_gem_context *to = req->ctx; - struct intel_engine_cs *engine = req->engine; - struct i915_hw_ppgtt *ppgtt = to->ppgtt ?: req->i915->mm.aliasing_ppgtt; - struct i915_gem_context *from = engine->legacy_active_context; - u32 hw_flags; + struct intel_engine_cs *engine = rq->engine; + struct i915_gem_context *to_ctx = rq->ctx; + struct i915_hw_ppgtt *to_mm = + to_ctx->ppgtt ?: rq->i915->mm.aliasing_ppgtt; + struct i915_gem_context *from_ctx = engine->legacy_active_context; + struct i915_hw_ppgtt *from_mm = engine->legacy_active_ppgtt; + u32 hw_flags = 0; int ret, i; - GEM_BUG_ON(engine->id != RCS); + lockdep_assert_held(&rq->i915->drm.struct_mutex); + GEM_BUG_ON(HAS_EXECLISTS(rq->i915)); - if (skip_rcs_switch(ppgtt, engine, to)) - return 0; - - if (needs_pd_load_pre(ppgtt, engine)) { - /* Older GENs and non render rings still want the load first, - * "PP_DCLV followed by PP_DIR_BASE register through Load - * Register Immediate commands in Ring Buffer before submitting - * a context."*/ - trace_switch_mm(engine, to); - ret = ppgtt->switch_mm(ppgtt, req); + if (to_mm != from_mm || + (to_mm && intel_engine_flag(engine) & to_mm->pd_dirty_rings)) { + trace_switch_mm(engine, to_ctx); + ret = to_mm->switch_mm(to_mm, rq); if (ret) - return ret; + goto err; + + to_mm->pd_dirty_rings &= ~intel_engine_flag(engine); + engine->legacy_active_ppgtt = to_mm; + hw_flags = MI_FORCE_RESTORE; } - if (i915_gem_context_is_kernel(to)) + if (to_ctx->engine[engine->id].state && + (to_ctx != from_ctx || hw_flags & MI_FORCE_RESTORE)) { + GEM_BUG_ON(engine->id != RCS); + /* * The kernel context(s) is treated as pure scratch and is not * expected to retain any state (as we sacrifice it during @@ -750,78 +732,37 @@ static int do_rcs_switch(struct drm_i915_gem_request *req) * as nothing actually executes using the kernel context; it * is purely used for flushing user contexts. */ - hw_flags = MI_RESTORE_INHIBIT; - else if (ppgtt && intel_engine_flag(engine) & ppgtt->pd_dirty_rings) - hw_flags = MI_FORCE_RESTORE; - else - hw_flags = 0; + if (i915_gem_context_is_kernel(to_ctx)) + hw_flags = MI_RESTORE_INHIBIT; - if (to != from || (hw_flags & MI_FORCE_RESTORE)) { - ret = mi_set_context(req, hw_flags); + ret = mi_set_context(rq, hw_flags); if (ret) - return ret; + goto err_mm; - engine->legacy_active_context = to; + engine->legacy_active_context = to_ctx; } - if (ppgtt) - ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine); + if (to_ctx->remap_slice) { + for (i = 0; i < MAX_L3_SLICES; i++) { + if (!(to_ctx->remap_slice & BIT(i))) + continue; - for (i = 0; i < MAX_L3_SLICES; i++) { - if (!(to->remap_slice & (1<remap_slice &= ~(1< 1. This allows us to destroy the context abstract - * object while letting the normal object tracking destroy the backing BO. - * - * This function should not be used in execlists mode. Instead the context is - * switched by writing to the ELSP and requests keep a reference to their - * context. - */ -int i915_switch_context(struct drm_i915_gem_request *req) -{ - struct intel_engine_cs *engine = req->engine; - - lockdep_assert_held(&req->i915->drm.struct_mutex); - GEM_BUG_ON(HAS_EXECLISTS(req->i915)); - - if (!req->ctx->engine[engine->id].state) { - struct i915_gem_context *to = req->ctx; - struct i915_hw_ppgtt *ppgtt = - to->ppgtt ?: req->i915->mm.aliasing_ppgtt; - - if (needs_pd_load_pre(ppgtt, engine)) { - int ret; - - trace_switch_mm(engine, to); - ret = ppgtt->switch_mm(ppgtt, req); + ret = remap_l3(rq, i); if (ret) - return ret; - - ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine); + goto err_ctx; } - engine->legacy_active_context = to; - return 0; + to_ctx->remap_slice = 0; } - return do_rcs_switch(req); + return 0; + +err_ctx: + engine->legacy_active_context = from_ctx; +err_mm: + engine->legacy_active_ppgtt = from_mm; +err: + return ret; } static bool engine_has_idle_kernel_context(struct intel_engine_cs *engine) diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c index 7325469ce754a..a90bdd26571f5 100644 --- a/drivers/gpu/drm/i915/i915_gem_request.c +++ b/drivers/gpu/drm/i915/i915_gem_request.c @@ -719,25 +719,19 @@ i915_gem_request_alloc(struct intel_engine_cs *engine, /* Unconditionally invalidate GPU caches and TLBs. */ ret = engine->emit_flush(req, EMIT_INVALIDATE); if (ret) - goto err_ctx; + goto err_unwind; ret = engine->request_alloc(req); - if (ret) { - /* - * Past the point-of-no-return. Since we may have updated - * global state after partially completing the request alloc, - * we need to commit any commands so far emitted in the - * request to the HW. - */ - __i915_add_request(req, false); - return ERR_PTR(ret); - } + if (ret) + goto err_unwind; /* Check that we didn't interrupt ourselves with a new request */ GEM_BUG_ON(req->timeline->seqno != req->fence.seqno); return req; -err_ctx: +err_unwind: + req->ring->emit = req->head; + /* Make sure we didn't add ourselves to external state before freeing */ GEM_BUG_ON(!list_empty(&req->active_list)); GEM_BUG_ON(!list_empty(&req->priotree.signalers_list)); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index bfa11a84e476e..a904b0353bec7 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -591,6 +591,7 @@ static void reset_ring_common(struct intel_engine_cs *engine, request->ring->head = request->postfix; } else { engine->legacy_active_context = NULL; + engine->legacy_active_ppgtt = NULL; } } diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index d38d059285dca..ef304c99954c8 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -531,6 +531,7 @@ struct intel_engine_cs { * stream (ring). */ struct i915_gem_context *legacy_active_context; + struct i915_hw_ppgtt *legacy_active_ppgtt; /* status_notifier: list of callbacks for context-switch changes */ struct atomic_notifier_head context_status_notifier; -- GitLab From 8911a31c813275882fdc15554235a914e678759e Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 15:26:31 +0000 Subject: [PATCH 0414/1471] drm/i915: Move mi_set_context() into the legacy ringbuffer submission The legacy i915_switch_context() is only applicable to the legacy ringbuffer submission method, so move it from the general i915_gem_context.c to intel_ringbuffer.c (rename pending!). Signed-off-by: Chris Wilson Cc: Mika Kuoppala Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171123152631.31385-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem_context.c | 198 ------------------------ drivers/gpu/drm/i915/intel_ringbuffer.c | 186 +++++++++++++++++++++- 2 files changed, 185 insertions(+), 199 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index da4a64b3484ca..aee0f6d72d337 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -567,204 +567,6 @@ void i915_gem_context_close(struct drm_file *file) idr_destroy(&file_priv->context_idr); } -static inline int -mi_set_context(struct drm_i915_gem_request *req, u32 flags) -{ - struct drm_i915_private *dev_priv = req->i915; - struct intel_engine_cs *engine = req->engine; - enum intel_engine_id id; - const int num_rings = - /* Use an extended w/a on gen7 if signalling from other rings */ - (HAS_LEGACY_SEMAPHORES(dev_priv) && IS_GEN7(dev_priv)) ? - INTEL_INFO(dev_priv)->num_rings - 1 : - 0; - int len; - u32 *cs; - - flags |= MI_MM_SPACE_GTT; - if (IS_HASWELL(dev_priv)) - /* These flags are for resource streamer on HSW+ */ - flags |= HSW_MI_RS_SAVE_STATE_EN | HSW_MI_RS_RESTORE_STATE_EN; - else - flags |= MI_SAVE_EXT_STATE_EN | MI_RESTORE_EXT_STATE_EN; - - len = 4; - if (IS_GEN7(dev_priv)) - len += 2 + (num_rings ? 4*num_rings + 6 : 0); - - cs = intel_ring_begin(req, len); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - /* WaProgramMiArbOnOffAroundMiSetContext:ivb,vlv,hsw,bdw,chv */ - if (IS_GEN7(dev_priv)) { - *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE; - if (num_rings) { - struct intel_engine_cs *signaller; - - *cs++ = MI_LOAD_REGISTER_IMM(num_rings); - for_each_engine(signaller, dev_priv, id) { - if (signaller == engine) - continue; - - *cs++ = i915_mmio_reg_offset( - RING_PSMI_CTL(signaller->mmio_base)); - *cs++ = _MASKED_BIT_ENABLE( - GEN6_PSMI_SLEEP_MSG_DISABLE); - } - } - } - - *cs++ = MI_NOOP; - *cs++ = MI_SET_CONTEXT; - *cs++ = i915_ggtt_offset(req->ctx->engine[RCS].state) | flags; - /* - * w/a: MI_SET_CONTEXT must always be followed by MI_NOOP - * WaMiSetContext_Hang:snb,ivb,vlv - */ - *cs++ = MI_NOOP; - - if (IS_GEN7(dev_priv)) { - if (num_rings) { - struct intel_engine_cs *signaller; - i915_reg_t last_reg = {}; /* keep gcc quiet */ - - *cs++ = MI_LOAD_REGISTER_IMM(num_rings); - for_each_engine(signaller, dev_priv, id) { - if (signaller == engine) - continue; - - last_reg = RING_PSMI_CTL(signaller->mmio_base); - *cs++ = i915_mmio_reg_offset(last_reg); - *cs++ = _MASKED_BIT_DISABLE( - GEN6_PSMI_SLEEP_MSG_DISABLE); - } - - /* Insert a delay before the next switch! */ - *cs++ = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT; - *cs++ = i915_mmio_reg_offset(last_reg); - *cs++ = i915_ggtt_offset(engine->scratch); - *cs++ = MI_NOOP; - } - *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; - } - - intel_ring_advance(req, cs); - - return 0; -} - -static int remap_l3(struct drm_i915_gem_request *req, int slice) -{ - u32 *cs, *remap_info = req->i915->l3_parity.remap_info[slice]; - int i; - - if (!remap_info) - return 0; - - cs = intel_ring_begin(req, GEN7_L3LOG_SIZE/4 * 2 + 2); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - /* - * Note: We do not worry about the concurrent register cacheline hang - * here because no other code should access these registers other than - * at initialization time. - */ - *cs++ = MI_LOAD_REGISTER_IMM(GEN7_L3LOG_SIZE/4); - for (i = 0; i < GEN7_L3LOG_SIZE/4; i++) { - *cs++ = i915_mmio_reg_offset(GEN7_L3LOG(slice, i)); - *cs++ = remap_info[i]; - } - *cs++ = MI_NOOP; - intel_ring_advance(req, cs); - - return 0; -} - -/** - * i915_switch_context() - perform a GPU context switch. - * @rq: request for which we'll execute the context switch - * - * The context life cycle is simple. The context refcount is incremented and - * decremented by 1 on create and destroy. If the context is in use by the GPU, - * it will have a refcount > 1. This allows us to destroy the context abstract - * object while letting the normal object tracking destroy the backing BO. - * - * This function should not be used in execlists mode. Instead the context is - * switched by writing to the ELSP and requests keep a reference to their - * context. - */ -int i915_switch_context(struct drm_i915_gem_request *rq) -{ - struct intel_engine_cs *engine = rq->engine; - struct i915_gem_context *to_ctx = rq->ctx; - struct i915_hw_ppgtt *to_mm = - to_ctx->ppgtt ?: rq->i915->mm.aliasing_ppgtt; - struct i915_gem_context *from_ctx = engine->legacy_active_context; - struct i915_hw_ppgtt *from_mm = engine->legacy_active_ppgtt; - u32 hw_flags = 0; - int ret, i; - - lockdep_assert_held(&rq->i915->drm.struct_mutex); - GEM_BUG_ON(HAS_EXECLISTS(rq->i915)); - - if (to_mm != from_mm || - (to_mm && intel_engine_flag(engine) & to_mm->pd_dirty_rings)) { - trace_switch_mm(engine, to_ctx); - ret = to_mm->switch_mm(to_mm, rq); - if (ret) - goto err; - - to_mm->pd_dirty_rings &= ~intel_engine_flag(engine); - engine->legacy_active_ppgtt = to_mm; - hw_flags = MI_FORCE_RESTORE; - } - - if (to_ctx->engine[engine->id].state && - (to_ctx != from_ctx || hw_flags & MI_FORCE_RESTORE)) { - GEM_BUG_ON(engine->id != RCS); - - /* - * The kernel context(s) is treated as pure scratch and is not - * expected to retain any state (as we sacrifice it during - * suspend and on resume it may be corrupted). This is ok, - * as nothing actually executes using the kernel context; it - * is purely used for flushing user contexts. - */ - if (i915_gem_context_is_kernel(to_ctx)) - hw_flags = MI_RESTORE_INHIBIT; - - ret = mi_set_context(rq, hw_flags); - if (ret) - goto err_mm; - - engine->legacy_active_context = to_ctx; - } - - if (to_ctx->remap_slice) { - for (i = 0; i < MAX_L3_SLICES; i++) { - if (!(to_ctx->remap_slice & BIT(i))) - continue; - - ret = remap_l3(rq, i); - if (ret) - goto err_ctx; - } - - to_ctx->remap_slice = 0; - } - - return 0; - -err_ctx: - engine->legacy_active_context = from_ctx; -err_mm: - engine->legacy_active_ppgtt = from_mm; -err: - return ret; -} - static bool engine_has_idle_kernel_context(struct intel_engine_cs *engine) { struct i915_gem_timeline *timeline; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index a904b0353bec7..e2085820b586b 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1385,6 +1385,190 @@ void intel_legacy_submission_resume(struct drm_i915_private *dev_priv) intel_ring_reset(engine->buffer, 0); } +static inline int mi_set_context(struct drm_i915_gem_request *rq, u32 flags) +{ + struct drm_i915_private *i915 = rq->i915; + struct intel_engine_cs *engine = rq->engine; + enum intel_engine_id id; + const int num_rings = + /* Use an extended w/a on gen7 if signalling from other rings */ + (HAS_LEGACY_SEMAPHORES(i915) && IS_GEN7(i915)) ? + INTEL_INFO(i915)->num_rings - 1 : + 0; + int len; + u32 *cs; + + flags |= MI_MM_SPACE_GTT; + if (IS_HASWELL(i915)) + /* These flags are for resource streamer on HSW+ */ + flags |= HSW_MI_RS_SAVE_STATE_EN | HSW_MI_RS_RESTORE_STATE_EN; + else + flags |= MI_SAVE_EXT_STATE_EN | MI_RESTORE_EXT_STATE_EN; + + len = 4; + if (IS_GEN7(i915)) + len += 2 + (num_rings ? 4*num_rings + 6 : 0); + + cs = intel_ring_begin(rq, len); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + /* WaProgramMiArbOnOffAroundMiSetContext:ivb,vlv,hsw,bdw,chv */ + if (IS_GEN7(i915)) { + *cs++ = MI_ARB_ON_OFF | MI_ARB_DISABLE; + if (num_rings) { + struct intel_engine_cs *signaller; + + *cs++ = MI_LOAD_REGISTER_IMM(num_rings); + for_each_engine(signaller, i915, id) { + if (signaller == engine) + continue; + + *cs++ = i915_mmio_reg_offset( + RING_PSMI_CTL(signaller->mmio_base)); + *cs++ = _MASKED_BIT_ENABLE( + GEN6_PSMI_SLEEP_MSG_DISABLE); + } + } + } + + *cs++ = MI_NOOP; + *cs++ = MI_SET_CONTEXT; + *cs++ = i915_ggtt_offset(rq->ctx->engine[RCS].state) | flags; + /* + * w/a: MI_SET_CONTEXT must always be followed by MI_NOOP + * WaMiSetContext_Hang:snb,ivb,vlv + */ + *cs++ = MI_NOOP; + + if (IS_GEN7(i915)) { + if (num_rings) { + struct intel_engine_cs *signaller; + i915_reg_t last_reg = {}; /* keep gcc quiet */ + + *cs++ = MI_LOAD_REGISTER_IMM(num_rings); + for_each_engine(signaller, i915, id) { + if (signaller == engine) + continue; + + last_reg = RING_PSMI_CTL(signaller->mmio_base); + *cs++ = i915_mmio_reg_offset(last_reg); + *cs++ = _MASKED_BIT_DISABLE( + GEN6_PSMI_SLEEP_MSG_DISABLE); + } + + /* Insert a delay before the next switch! */ + *cs++ = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT; + *cs++ = i915_mmio_reg_offset(last_reg); + *cs++ = i915_ggtt_offset(engine->scratch); + *cs++ = MI_NOOP; + } + *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; + } + + intel_ring_advance(rq, cs); + + return 0; +} + +static int remap_l3(struct drm_i915_gem_request *rq, int slice) +{ + u32 *cs, *remap_info = rq->i915->l3_parity.remap_info[slice]; + int i; + + if (!remap_info) + return 0; + + cs = intel_ring_begin(rq, GEN7_L3LOG_SIZE/4 * 2 + 2); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + /* + * Note: We do not worry about the concurrent register cacheline hang + * here because no other code should access these registers other than + * at initialization time. + */ + *cs++ = MI_LOAD_REGISTER_IMM(GEN7_L3LOG_SIZE/4); + for (i = 0; i < GEN7_L3LOG_SIZE/4; i++) { + *cs++ = i915_mmio_reg_offset(GEN7_L3LOG(slice, i)); + *cs++ = remap_info[i]; + } + *cs++ = MI_NOOP; + intel_ring_advance(rq, cs); + + return 0; +} + +static int switch_context(struct drm_i915_gem_request *rq) +{ + struct intel_engine_cs *engine = rq->engine; + struct i915_gem_context *to_ctx = rq->ctx; + struct i915_hw_ppgtt *to_mm = + to_ctx->ppgtt ?: rq->i915->mm.aliasing_ppgtt; + struct i915_gem_context *from_ctx = engine->legacy_active_context; + struct i915_hw_ppgtt *from_mm = engine->legacy_active_ppgtt; + u32 hw_flags = 0; + int ret, i; + + lockdep_assert_held(&rq->i915->drm.struct_mutex); + GEM_BUG_ON(HAS_EXECLISTS(rq->i915)); + + if (to_mm != from_mm || + (to_mm && intel_engine_flag(engine) & to_mm->pd_dirty_rings)) { + trace_switch_mm(engine, to_ctx); + ret = to_mm->switch_mm(to_mm, rq); + if (ret) + goto err; + + to_mm->pd_dirty_rings &= ~intel_engine_flag(engine); + engine->legacy_active_ppgtt = to_mm; + hw_flags = MI_FORCE_RESTORE; + } + + if (to_ctx->engine[engine->id].state && + (to_ctx != from_ctx || hw_flags & MI_FORCE_RESTORE)) { + GEM_BUG_ON(engine->id != RCS); + + /* + * The kernel context(s) is treated as pure scratch and is not + * expected to retain any state (as we sacrifice it during + * suspend and on resume it may be corrupted). This is ok, + * as nothing actually executes using the kernel context; it + * is purely used for flushing user contexts. + */ + if (i915_gem_context_is_kernel(to_ctx)) + hw_flags = MI_RESTORE_INHIBIT; + + ret = mi_set_context(rq, hw_flags); + if (ret) + goto err_mm; + + engine->legacy_active_context = to_ctx; + } + + if (to_ctx->remap_slice) { + for (i = 0; i < MAX_L3_SLICES; i++) { + if (!(to_ctx->remap_slice & BIT(i))) + continue; + + ret = remap_l3(rq, i); + if (ret) + goto err_ctx; + } + + to_ctx->remap_slice = 0; + } + + return 0; + +err_ctx: + engine->legacy_active_context = from_ctx; +err_mm: + engine->legacy_active_ppgtt = from_mm; +err: + return ret; +} + static int ring_request_alloc(struct drm_i915_gem_request *request) { int ret; @@ -1401,7 +1585,7 @@ static int ring_request_alloc(struct drm_i915_gem_request *request) if (ret) return ret; - ret = i915_switch_context(request); + ret = switch_context(request); if (ret) return ret; -- GitLab From b7d3aabf900728b6b13b860157026fd26099208a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 21:17:51 +0000 Subject: [PATCH 0415/1471] drm/i915/pmu: Hide the (unsigned long)ptr cast We pretend the PMU config id is a pointer value when encoding it into the device parameters for presentation via sysfs. This requires casting of an unsigned long into and out of the pointer member, which annoys smatch: drivers/gpu/drm/i915/i915_pmu.c:684 i915_pmu_event_show() warn: argument 3 to %lx specifier is cast from pointer Instead of abusing a generic dev_ext_attribute, define our own typesafe attributes. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/20171123211751.2885-1-chris@chris-wilson.co.uk Reviewed-by: Tvrtko Ursulin --- drivers/gpu/drm/i915/i915_pmu.c | 30 ++++++++++++++++++++---------- 1 file changed, 20 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 6a42e7f7967dd..5170a46893d79 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -650,19 +650,24 @@ static int i915_pmu_event_event_idx(struct perf_event *event) return 0; } +struct i915_str_attribute { + struct device_attribute attr; + const char *str; +}; + static ssize_t i915_pmu_format_show(struct device *dev, struct device_attribute *attr, char *buf) { - struct dev_ext_attribute *eattr; + struct i915_str_attribute *eattr; - eattr = container_of(attr, struct dev_ext_attribute, attr); - return sprintf(buf, "%s\n", (char *)eattr->var); + eattr = container_of(attr, struct i915_str_attribute, attr); + return sprintf(buf, "%s\n", eattr->str); } #define I915_PMU_FORMAT_ATTR(_name, _config) \ - (&((struct dev_ext_attribute[]) { \ + (&((struct i915_str_attribute[]) { \ { .attr = __ATTR(_name, 0444, i915_pmu_format_show, NULL), \ - .var = (void *)_config, } \ + .str = _config, } \ })[0].attr.attr) static struct attribute *i915_pmu_format_attrs[] = { @@ -675,19 +680,24 @@ static const struct attribute_group i915_pmu_format_attr_group = { .attrs = i915_pmu_format_attrs, }; +struct i915_ext_attribute { + struct device_attribute attr; + unsigned long val; +}; + static ssize_t i915_pmu_event_show(struct device *dev, struct device_attribute *attr, char *buf) { - struct dev_ext_attribute *eattr; + struct i915_ext_attribute *eattr; - eattr = container_of(attr, struct dev_ext_attribute, attr); - return sprintf(buf, "config=0x%lx\n", (unsigned long)eattr->var); + eattr = container_of(attr, struct i915_ext_attribute, attr); + return sprintf(buf, "config=0x%lx\n", eattr->val); } #define I915_EVENT_ATTR(_name, _config) \ - (&((struct dev_ext_attribute[]) { \ + (&((struct i915_ext_attribute[]) { \ { .attr = __ATTR(_name, 0444, i915_pmu_event_show, NULL), \ - .var = (void *)_config, } \ + .val = _config, } \ })[0].attr.attr) #define I915_EVENT_STR(_name, _str) \ -- GitLab From e7e5da7127d5c9c076f05d6ed71bdf6826fc536c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 23:37:12 +0000 Subject: [PATCH 0416/1471] drm/i915/selftests: Hold rpm wakeref for request + ggtt usage Since the removal of the delayed rc6 enabling, we now setup and drop the early rpm wakeref during modules initialisation before we start the live selftests. As such, we are now detecting errors in the tests where we were not holding the required wakeref for various actions. As rpm is not the primary goal of the tests involved, take a coarse and convenient rpm wakeref around the tests. Signed-off-by: Chris Wilson Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171123233712.21836-1-chris@chris-wilson.co.uk Reviewed-by: Matthew Auld --- drivers/gpu/drm/i915/selftests/huge_pages.c | 2 ++ drivers/gpu/drm/i915/selftests/i915_gem_evict.c | 2 ++ 2 files changed, 4 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index df1befe2c7938..e6b31041cc881 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -1749,6 +1749,7 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv) return PTR_ERR(file); mutex_lock(&dev_priv->drm.struct_mutex); + intel_runtime_pm_get(dev_priv); ctx = live_context(dev_priv, file); if (IS_ERR(ctx)) { @@ -1759,6 +1760,7 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv) err = i915_subtests(tests, ctx); out_unlock: + intel_runtime_pm_put(dev_priv); mutex_unlock(&dev_priv->drm.struct_mutex); mock_file_free(dev_priv, file); diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c index f463105ff48dc..e1ddad635d73e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c @@ -355,6 +355,7 @@ static int igt_evict_contexts(void *arg) return 0; mutex_lock(&i915->drm.struct_mutex); + intel_runtime_pm_get(i915); /* Reserve a block so that we know we have enough to fit a few rq */ memset(&hole, 0, sizeof(hole)); @@ -463,6 +464,7 @@ static int igt_evict_contexts(void *arg) } if (drm_mm_node_allocated(&hole)) drm_mm_remove_node(&hole); + intel_runtime_pm_put(i915); mutex_unlock(&i915->drm.struct_mutex); return err; -- GitLab From 141a0895d5ad3d2540c3d08651f2b63443fb8dfd Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 12:34:31 +0000 Subject: [PATCH 0417/1471] drm/i915/pmu: Remove conditional HOTPLUG_CPU registration Even for static CPU configurations, the hotplug CPU framework is still used to determine the CPU topology, and is still being used by the perf event register to check for valid CPUs. Fixes: b46a33e271ed ("drm/i915/pmu: Expose a PMU interface for perf queries") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Dmitry Rogozhkin Reviewed-by: Tvrtko Ursulin Signed-off-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20171123123432.25035-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 5170a46893d79..21006e8065a03 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -40,7 +40,7 @@ #define ENGINE_SAMPLE_BITS (1 << I915_PMU_SAMPLE_BITS) -static cpumask_t i915_pmu_cpumask = CPU_MASK_NONE; +static cpumask_t i915_pmu_cpumask; static u8 engine_config_sample(u64 config) { @@ -770,7 +770,6 @@ static const struct attribute_group *i915_pmu_attr_groups[] = { NULL }; -#ifdef CONFIG_HOTPLUG_CPU static int i915_pmu_cpu_online(unsigned int cpu, struct hlist_node *node) { struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), node); @@ -806,11 +805,9 @@ static int i915_pmu_cpu_offline(unsigned int cpu, struct hlist_node *node) } static enum cpuhp_state cpuhp_slot = CPUHP_INVALID; -#endif static int i915_pmu_register_cpuhp_state(struct drm_i915_private *i915) { -#ifdef CONFIG_HOTPLUG_CPU enum cpuhp_state slot; int ret; @@ -829,17 +826,14 @@ static int i915_pmu_register_cpuhp_state(struct drm_i915_private *i915) } cpuhp_slot = slot; -#endif return 0; } static void i915_pmu_unregister_cpuhp_state(struct drm_i915_private *i915) { -#ifdef CONFIG_HOTPLUG_CPU WARN_ON(cpuhp_slot == CPUHP_INVALID); WARN_ON(cpuhp_state_remove_instance(cpuhp_slot, &i915->pmu.node)); cpuhp_remove_multi_state(cpuhp_slot); -#endif } void i915_pmu_register(struct drm_i915_private *i915) -- GitLab From 0426c046546140fc42db5addeb7e2298dd0b6fb1 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Thu, 23 Nov 2017 12:34:32 +0000 Subject: [PATCH 0418/1471] drm/i915/pmu: Only allow running on a single CPU We do two things, both of which are purely to simplify and clarify the implementation: 1. Simplify the CPU online callback so it is more obvious that the purpose there is to set a single CPU mask bit for the first CPU which comes online. Using cpumask_weight for this reads more obvious than the trick with cpumask_and_any. 2. Modify the event init so that events can be created only on a single CPU. This removes looking at the requested CPU thread siblings, and only allows creating on the current active CPU. Signed-off-by: Tvrtko Ursulin Cc: Chris Wilson Cc: Dmitry Rogozhkin Tested-by: Chris Wilson Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171123123432.25035-2-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 21006e8065a03..2b7a4779f1847 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -344,7 +344,7 @@ static int i915_pmu_event_init(struct perf_event *event) { struct drm_i915_private *i915 = container_of(event->pmu, typeof(*i915), pmu.base); - int cpu, ret; + int ret; if (event->attr.type != event->pmu->type) return -ENOENT; @@ -359,9 +359,8 @@ static int i915_pmu_event_init(struct perf_event *event) if (event->cpu < 0) return -EINVAL; - cpu = cpumask_any_and(&i915_pmu_cpumask, - topology_sibling_cpumask(event->cpu)); - if (cpu >= nr_cpu_ids) + /* only allow running on one cpu at a time */ + if (!cpumask_test_cpu(event->cpu, &i915_pmu_cpumask)) return -ENODEV; if (is_engine_event(event)) { @@ -396,7 +395,6 @@ static int i915_pmu_event_init(struct perf_event *event) if (ret) return ret; - event->cpu = cpu; if (!event->parent) event->destroy = i915_pmu_event_destroy; @@ -773,13 +771,11 @@ static const struct attribute_group *i915_pmu_attr_groups[] = { static int i915_pmu_cpu_online(unsigned int cpu, struct hlist_node *node) { struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), node); - unsigned int target; GEM_BUG_ON(!pmu->base.event_init); - target = cpumask_any_and(&i915_pmu_cpumask, &i915_pmu_cpumask); /* Select the first online CPU as a designated reader. */ - if (target >= nr_cpu_ids) + if (!cpumask_weight(&i915_pmu_cpumask)) cpumask_set_cpu(cpu, &i915_pmu_cpumask); return 0; -- GitLab From 95c71769811cb3ca55ecbb9f6a7fc126a35c1500 Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Fri, 24 Nov 2017 13:50:50 +0530 Subject: [PATCH 0419/1471] drm/i915/guc: Update header guard name of intel_guc_submission.h While renaming i915_guc_submission.h to intel_guc_submission.h, update of header guard name to _INTEL_GUC_SUBMISSION_H_ was missed. References: a269574489d4 (drm/i915/guc: Rename i915_guc_submission.c|h to intel_guc_submission.c|h) Signed-off-by: Sagar Arun Kamble Cc: Joonas Lahtinen Reviewed-by: Michal Wajdeczko Link: https://patchwork.freedesktop.org/patch/msgid/1511511650-13756-1-git-send-email-sagar.a.kamble@intel.com Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_submission.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_submission.h b/drivers/gpu/drm/i915/intel_guc_submission.h index e901192ee4697..021fe85c8f718 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.h +++ b/drivers/gpu/drm/i915/intel_guc_submission.h @@ -22,8 +22,8 @@ * */ -#ifndef _I915_GUC_SUBMISSION_H_ -#define _I915_GUC_SUBMISSION_H_ +#ifndef _INTEL_GUC_SUBMISSION_H_ +#define _INTEL_GUC_SUBMISSION_H_ #include -- GitLab From 3e8f5b0805e5f93770e420e3f455b7d14d08173f Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 24 Nov 2017 09:53:40 +0000 Subject: [PATCH 0420/1471] drm/i915/guc: Rename i915_guc_reg.h to intel_guc_reg.h We are using intel_ prefix for all file names with hardware related definitions. GuC registers also fall into this category. Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Reviewed-by: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171124095340.1500-1-michal.wajdeczko@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc.h | 2 +- drivers/gpu/drm/i915/{i915_guc_reg.h => intel_guc_reg.h} | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) rename drivers/gpu/drm/i915/{i915_guc_reg.h => intel_guc_reg.h} (98%) diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h index 75c4cfefdaff1..59856726d2bca 100644 --- a/drivers/gpu/drm/i915/intel_guc.h +++ b/drivers/gpu/drm/i915/intel_guc.h @@ -30,8 +30,8 @@ #include "intel_guc_fwif.h" #include "intel_guc_ct.h" #include "intel_guc_log.h" +#include "intel_guc_reg.h" #include "intel_uc_fw.h" -#include "i915_guc_reg.h" #include "i915_vma.h" struct guc_preempt_work { diff --git a/drivers/gpu/drm/i915/i915_guc_reg.h b/drivers/gpu/drm/i915/intel_guc_reg.h similarity index 98% rename from drivers/gpu/drm/i915/i915_guc_reg.h rename to drivers/gpu/drm/i915/intel_guc_reg.h index bc1ae7d8f424d..0a8ff03d026f6 100644 --- a/drivers/gpu/drm/i915/i915_guc_reg.h +++ b/drivers/gpu/drm/i915/intel_guc_reg.h @@ -21,8 +21,8 @@ * IN THE SOFTWARE. * */ -#ifndef _I915_GUC_REG_H_ -#define _I915_GUC_REG_H_ +#ifndef _INTEL_GUC_REG_H_ +#define _INTEL_GUC_REG_H_ /* Definitions of GuC H/W registers, bits, etc */ -- GitLab From 1b2b659141cfa7c96c62c79516662a34ad112284 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 12 Oct 2017 10:21:47 +0100 Subject: [PATCH 0421/1471] drm/i915: Select DEBUG_FS for our test suite Ensure that we build our test kernels with DEBUG_FS enabled as many of our igt tests require poking around debugfs/dri. This should also fixup the kbuild complaint that we tried to select SW_SYNC without meeting its dependencies. Reported-by: kbuild-all@01.org Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171012092147.28986-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/Kconfig.debug | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug index 9e53edbc713b9..fa36491495b11 100644 --- a/drivers/gpu/drm/i915/Kconfig.debug +++ b/drivers/gpu/drm/i915/Kconfig.debug @@ -18,6 +18,7 @@ config DRM_I915_WERROR config DRM_I915_DEBUG bool "Enable additional driver debugging" depends on DRM_I915 + select DEBUG_FS select PREEMPT_COUNT select I2C_CHARDEV select DRM_DP_AUX_CHARDEV -- GitLab From 8ee4f19c47031f23340055da4d9f2af537de23f4 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 24 Nov 2017 09:49:59 +0000 Subject: [PATCH 0422/1471] drm/i915/pmu: Stop averaging with the previous sample Averaging with the previous sample brings a small statistical improvement to sampling counters, but can leek a little bit of state from a current client to the next which mulls the border between past and present for observing clients. This is because on event enable clients record the current counter value and use it as reference, but with rapid off-on event cycles, and due the delayed nature of sampling timer self-disarm, previous sample value does not get cleared under these circumstances. Solution is to stop averaging with the previous sample. This has a small downside of losing some precision with short and spiky signals, but the alternatives look too complicated for the benefit. Signed-off-by: Tvrtko Ursulin Cc: Chris Wilson Cc: Sagar Arun Kamble Reviewed-by: Chris Wilson Tested-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171124094959.10725-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 29 ++--------------------------- drivers/gpu/drm/i915/i915_pmu.h | 1 - 2 files changed, 2 insertions(+), 28 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 2b7a4779f1847..39310cf13c3af 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -181,12 +181,7 @@ static bool grab_forcewake(struct drm_i915_private *i915, bool fw) static void update_sample(struct i915_pmu_sample *sample, u32 unit, u32 val) { - /* - * Since we are doing stochastic sampling for these counters, - * average the delta with the previous value for better accuracy. - */ - sample->cur += div_u64(mul_u32_u32(sample->prev + val, unit), 2); - sample->prev = val; + sample->cur += mul_u32_u32(val, unit); } static void engines_sample(struct drm_i915_private *dev_priv) @@ -262,31 +257,13 @@ static void frequency_sample(struct drm_i915_private *dev_priv) } } -static void pmu_init_previous_samples(struct drm_i915_private *i915) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - unsigned int i; - - for_each_engine(engine, i915, id) { - for (i = 0; i < ARRAY_SIZE(engine->pmu.sample); i++) - engine->pmu.sample[i].prev = 0; - } - - for (i = 0; i < ARRAY_SIZE(i915->pmu.sample); i++) - i915->pmu.sample[i].prev = i915->gt_pm.rps.idle_freq; -} - static enum hrtimer_restart i915_sample(struct hrtimer *hrtimer) { struct drm_i915_private *i915 = container_of(hrtimer, struct drm_i915_private, pmu.timer); - if (!READ_ONCE(i915->pmu.timer_enabled)) { - pmu_init_previous_samples(i915); - + if (!READ_ONCE(i915->pmu.timer_enabled)) return HRTIMER_NORESTART; - } engines_sample(i915); frequency_sample(i915); @@ -857,8 +834,6 @@ void i915_pmu_register(struct drm_i915_private *i915) hrtimer_init(&i915->pmu.timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); i915->pmu.timer.function = i915_sample; - pmu_init_previous_samples(i915); - for_each_engine(engine, i915, id) INIT_DELAYED_WORK(&engine->pmu.disable_busy_stats, __disable_busy_stats); diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h index e209783a4c5fa..40c154d13565a 100644 --- a/drivers/gpu/drm/i915/i915_pmu.h +++ b/drivers/gpu/drm/i915/i915_pmu.h @@ -41,7 +41,6 @@ enum { struct i915_pmu_sample { u64 cur; - u32 prev; }; struct i915_pmu { -- GitLab From c4deb62d7821672265b87952bcd1c808f3bf3e8f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 23 Nov 2017 21:41:56 +0200 Subject: [PATCH 0423/1471] drm/i915: Don't try indexed reads to alternate slave addresses MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We can only specify the one slave address to indexed reads/writes. Make sure the messages we check are destined to the same slave address before deciding to do an indexed transfer. Cc: stable@vger.kernel.org Cc: Daniel Kurtz Cc: Chris Wilson Cc: Daniel Vetter Cc: Sean Paul Fixes: 56f9eac05489 ("drm/i915/intel_i2c: use INDEX cycles for i2c read transactions") Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171123194157.25367-2-ville.syrjala@linux.intel.com Reviewed-by: Chris Wilson --- drivers/gpu/drm/i915/intel_i2c.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/intel_i2c.c b/drivers/gpu/drm/i915/intel_i2c.c index eb5827110d8ff..8affd47b98b81 100644 --- a/drivers/gpu/drm/i915/intel_i2c.c +++ b/drivers/gpu/drm/i915/intel_i2c.c @@ -438,6 +438,7 @@ static bool gmbus_is_index_read(struct i2c_msg *msgs, int i, int num) { return (i + 1 < num && + msgs[i].addr == msgs[i + 1].addr && !(msgs[i].flags & I2C_M_RD) && msgs[i].len <= 2 && (msgs[i + 1].flags & I2C_M_RD)); } -- GitLab From bb9e0d4bca50f429152e74a459160b41f3d60fb2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 23 Nov 2017 21:41:57 +0200 Subject: [PATCH 0424/1471] drm/i915: Prevent zero length "index" write MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The hardware always writes one or two bytes in the index portion of an indexed transfer. Make sure the message we send as the index doesn't have a zero length. Cc: stable@vger.kernel.org Cc: Daniel Kurtz Cc: Chris Wilson Cc: Daniel Vetter Cc: Sean Paul Fixes: 56f9eac05489 ("drm/i915/intel_i2c: use INDEX cycles for i2c read transactions") Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171123194157.25367-3-ville.syrjala@linux.intel.com Reviewed-by: Chris Wilson --- drivers/gpu/drm/i915/intel_i2c.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_i2c.c b/drivers/gpu/drm/i915/intel_i2c.c index 8affd47b98b81..49fdf09f9919c 100644 --- a/drivers/gpu/drm/i915/intel_i2c.c +++ b/drivers/gpu/drm/i915/intel_i2c.c @@ -439,7 +439,8 @@ gmbus_is_index_read(struct i2c_msg *msgs, int i, int num) { return (i + 1 < num && msgs[i].addr == msgs[i + 1].addr && - !(msgs[i].flags & I2C_M_RD) && msgs[i].len <= 2 && + !(msgs[i].flags & I2C_M_RD) && + (msgs[i].len == 1 || msgs[i].len == 2) && (msgs[i + 1].flags & I2C_M_RD)); } -- GitLab From aa34a6fc2722f042e372741661c0fae8640245ac Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 24 Nov 2017 13:37:44 +0000 Subject: [PATCH 0425/1471] drm/i915/guc: Advance over port[0] if set and not preempting MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Our execlist emulation is intended to only use a maximum of 2 ports per engine, so as to not overflow the wq. (By knowing the limits, we can avoid having to handle the wq exhaustion.) However, upon adding preemption, we lost the skip over the first port if set for the non-preemption path. Restore it. Reported-by: Mika Kuoppala Fixes: c41937fd994a ("drm/i915/guc: Preemption! With GuC") Signed-off-by: Chris Wilson Cc: Michał Winiarski Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171124133745.5173-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala Reviewed-by: Michał Winiarski --- drivers/gpu/drm/i915/intel_guc_submission.c | 29 ++++++++++++--------- 1 file changed, 16 insertions(+), 13 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index cbf5a96f58068..70e64bdb73dd2 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -743,23 +743,26 @@ static void guc_dequeue(struct intel_engine_cs *engine) if (!rb) goto unlock; - if (HAS_LOGICAL_RING_PREEMPTION(engine->i915) && port_isset(port)) { - struct guc_preempt_work *preempt_work = - &engine->i915->guc.preempt_work[engine->id]; - - if (rb_entry(rb, struct i915_priolist, node)->priority > - max(port_request(port)->priotree.priority, 0)) { - execlists_set_active(execlists, - EXECLISTS_ACTIVE_PREEMPT); - queue_work(engine->i915->guc.preempt_wq, - &preempt_work->work); - goto unlock; - } else if (port_isset(last_port)) { - goto unlock; + if (port_isset(port)) { + if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) { + struct guc_preempt_work *preempt_work = + &engine->i915->guc.preempt_work[engine->id]; + + if (rb_entry(rb, struct i915_priolist, node)->priority > + max(port_request(port)->priotree.priority, 0)) { + execlists_set_active(execlists, + EXECLISTS_ACTIVE_PREEMPT); + queue_work(engine->i915->guc.preempt_wq, + &preempt_work->work); + goto unlock; + } } port++; + if (port_isset(port)) + goto unlock; } + GEM_BUG_ON(port_isset(port)); do { struct i915_priolist *p = rb_entry(rb, typeof(*p), node); -- GitLab From 8c5da3bbcfd9cf763623f2d9515ace8eb3f01307 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 24 Nov 2017 13:00:29 +0000 Subject: [PATCH 0426/1471] drm/i915/guc: Tidy ELSP port assignment MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since we know that the port is empty, we do not need to extract the count from the old request it and copy it over to the new request, or attempt to unref the NULL old request pointer. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Michał Winiarski Reviewed-by: Michał Winiarski Link: https://patchwork.freedesktop.org/patch/msgid/20171124130031.20761-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_guc_submission.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index 70e64bdb73dd2..cf1cc2cb67225 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -718,12 +718,9 @@ static void guc_submit(struct intel_engine_cs *engine) static void port_assign(struct execlist_port *port, struct drm_i915_gem_request *rq) { - GEM_BUG_ON(rq == port_request(port)); - - if (port_isset(port)) - i915_gem_request_put(port_request(port)); + GEM_BUG_ON(port_isset(port)); - port_set(port, port_pack(i915_gem_request_get(rq), port_count(port))); + port_set(port, i915_gem_request_get(rq)); } static void guc_dequeue(struct intel_engine_cs *engine) -- GitLab From a54b1873c0e3ffec3041cf72cb10d986d6bc4fe5 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 24 Nov 2017 13:00:30 +0000 Subject: [PATCH 0427/1471] drm/i915: Use exponential backoff for wait_for() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead of sleeping for a fixed 1ms (roughly, depending on timer slack), start with a small sleep and exponentially increase the sleep on each cycle. A good example of a beneficiary is the guc mmio communication channel. Typically we expect (and so spin) for 10us for a quick response, but this doesn't cover everything and so sometimes we fallback to the millisecond+ sleep. This incurs a significant delay in time-critical operations like preemption (igt/gem_exec_latency), which can be improved significantly by using a small sleep after the spin fails. We've made this suggestion many times, but had little experimental data to support adding the complexity. v2: Bump the minimum usleep to 10us on advice of Documentation/timers/timers-howto.txt (Tvrko) v3: Specify min, max range for usleep intervals -- some code may crucially depend upon and so want to specify the sleep pattern. References: 1758b90e38f5 ("drm/i915: Use a hybrid scheme for fast register waits") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: John Harrison Cc: Michał Winiarski Cc: Ville Syrjala Reviewed-by: Michał Winiarski Link: https://patchwork.freedesktop.org/patch/msgid/20171124130031.20761-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_drv.h | 11 +++++++---- drivers/gpu/drm/i915/intel_pm.c | 2 +- 2 files changed, 8 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 635a96fcd7888..c00441a3d6495 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -48,8 +48,9 @@ * having timed out, since the timeout could be due to preemption or similar and * we've never had a chance to check the condition before the timeout. */ -#define _wait_for(COND, US, W) ({ \ +#define _wait_for(COND, US, Wmin, Wmax) ({ \ unsigned long timeout__ = jiffies + usecs_to_jiffies(US) + 1; \ + long wait__ = (Wmin); /* recommended min for usleep is 10 us */ \ int ret__; \ might_sleep(); \ for (;;) { \ @@ -62,12 +63,14 @@ ret__ = -ETIMEDOUT; \ break; \ } \ - usleep_range((W), (W) * 2); \ + usleep_range(wait__, wait__ * 2); \ + if (wait__ < (Wmax)) \ + wait__ <<= 1; \ } \ ret__; \ }) -#define wait_for(COND, MS) _wait_for((COND), (MS) * 1000, 1000) +#define wait_for(COND, MS) _wait_for((COND), (MS) * 1000, 10, 1000) /* If CONFIG_PREEMPT_COUNT is disabled, in_atomic() always reports false. */ #if defined(CONFIG_DRM_I915_DEBUG) && defined(CONFIG_PREEMPT_COUNT) @@ -116,7 +119,7 @@ int ret__; \ BUILD_BUG_ON(!__builtin_constant_p(US)); \ if ((US) > 10) \ - ret__ = _wait_for((COND), (US), 10); \ + ret__ = _wait_for((COND), (US), 10, 10); \ else \ ret__ = _wait_for_atomic((COND), (US), 0); \ ret__; \ diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 03d67d8ab647d..a80c322c5b43e 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -9294,7 +9294,7 @@ int skl_pcode_request(struct drm_i915_private *dev_priv, u32 mbox, u32 request, ret = 0; goto out; } - ret = _wait_for(COND, timeout_base_ms * 1000, 10); + ret = _wait_for(COND, timeout_base_ms * 1000, 10, 10); if (!ret) goto out; -- GitLab From 3452fa3095e91acbcb1f6290e0d70fa7d3695a3a Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 24 Nov 2017 17:13:31 +0000 Subject: [PATCH 0428/1471] drm/i915/pmu: Aggregate all RC6 states into one counter Chris has discovered that RC6, RC6p and RC6pp counters are mutually exclusive, and even that on some SNB SKUs you get RC6p increasing, and on the others RC6. Furthermore RC6p and RC6pp were only present starting from GEN6 until, GEN7, not including Haswell. All this combined makes it questionable whether we need to reserve new ABI for these counters. One idea was to just combine them all under the RC6 counter to simplify things for userspace. So that is what this patch does. Signed-off-by: Tvrtko Ursulin Suggested-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171124171331.17981-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 23 ++++++----------------- include/uapi/drm/i915_drm.h | 6 +----- 2 files changed, 7 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 39310cf13c3af..3357b690ce909 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -359,11 +359,6 @@ static int i915_pmu_event_init(struct perf_event *event) if (!HAS_RC6(i915)) ret = -ENODEV; break; - case I915_PMU_RC6p_RESIDENCY: - case I915_PMU_RC6pp_RESIDENCY: - if (!HAS_RC6p(i915)) - ret = -ENODEV; - break; default: ret = -ENOENT; break; @@ -421,16 +416,12 @@ static u64 __i915_pmu_event_read(struct perf_event *event) IS_VALLEYVIEW(i915) ? VLV_GT_RENDER_RC6 : GEN6_GT_GFX_RC6); - intel_runtime_pm_put(i915); - break; - case I915_PMU_RC6p_RESIDENCY: - intel_runtime_pm_get(i915); - val = intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6p); - intel_runtime_pm_put(i915); - break; - case I915_PMU_RC6pp_RESIDENCY: - intel_runtime_pm_get(i915); - val = intel_rc6_residency_ns(i915, GEN6_GT_GFX_RC6pp); + if (HAS_RC6p(i915)) { + val += intel_rc6_residency_ns(i915, + GEN6_GT_GFX_RC6p); + val += intel_rc6_residency_ns(i915, + GEN6_GT_GFX_RC6pp); + } intel_runtime_pm_put(i915); break; } @@ -708,8 +699,6 @@ static struct attribute *i915_pmu_events_attrs[] = { I915_EVENT_ATTR(interrupts, I915_PMU_INTERRUPTS), I915_EVENT(rc6-residency, I915_PMU_RC6_RESIDENCY, "ns"), - I915_EVENT(rc6p-residency, I915_PMU_RC6p_RESIDENCY, "ns"), - I915_EVENT(rc6pp-residency, I915_PMU_RC6pp_RESIDENCY, "ns"), NULL, }; diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 239e8633edc91..536ee4febd746 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -137,14 +137,10 @@ enum drm_i915_pmu_engine_sample { #define I915_PMU_ACTUAL_FREQUENCY __I915_PMU_OTHER(0) #define I915_PMU_REQUESTED_FREQUENCY __I915_PMU_OTHER(1) - #define I915_PMU_INTERRUPTS __I915_PMU_OTHER(2) - #define I915_PMU_RC6_RESIDENCY __I915_PMU_OTHER(3) -#define I915_PMU_RC6p_RESIDENCY __I915_PMU_OTHER(4) -#define I915_PMU_RC6pp_RESIDENCY __I915_PMU_OTHER(5) -#define I915_PMU_LAST I915_PMU_RC6pp_RESIDENCY +#define I915_PMU_LAST I915_PMU_RC6_RESIDENCY /* Each region is a minimum of 16k, and there are at most 255 of them. */ -- GitLab From 53fa54a6ef4fa5a261468d13dc0f54739adaa456 Mon Sep 17 00:00:00 2001 From: Michal Wajdeczko Date: Fri, 24 Nov 2017 17:02:39 +0000 Subject: [PATCH 0429/1471] drm/i915/guc: Use consistent name for scratch register count We should be consistent on naming of similar definitions. Signed-off-by: Michal Wajdeczko Cc: Chris Wilson Cc: Joonas Lahtinen Cc: Sagar Arun Kamble Link: https://patchwork.freedesktop.org/patch/msgid/20171124170239.29360-1-michal.wajdeczko@intel.com Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/intel_guc_fw.c | 4 ++-- drivers/gpu/drm/i915/intel_guc_reg.h | 3 ++- drivers/gpu/drm/i915/intel_uc_fw.c | 2 +- 3 files changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index 69ba015995757..bbab4e1106f33 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -130,14 +130,14 @@ static int guc_xfer_rsa(struct intel_guc *guc, struct i915_vma *vma) struct drm_i915_private *dev_priv = guc_to_i915(guc); struct intel_uc_fw *guc_fw = &guc->fw; struct sg_table *sg = vma->pages; - u32 rsa[UOS_RSA_SCRATCH_MAX_COUNT]; + u32 rsa[UOS_RSA_SCRATCH_COUNT]; int i; if (sg_pcopy_to_buffer(sg->sgl, sg->nents, rsa, sizeof(rsa), guc_fw->rsa_offset) != sizeof(rsa)) return -EINVAL; - for (i = 0; i < UOS_RSA_SCRATCH_MAX_COUNT; i++) + for (i = 0; i < UOS_RSA_SCRATCH_COUNT; i++) I915_WRITE(UOS_RSA_SCRATCH(i), rsa[i]); return 0; diff --git a/drivers/gpu/drm/i915/intel_guc_reg.h b/drivers/gpu/drm/i915/intel_guc_reg.h index 0a8ff03d026f6..19a9247c56646 100644 --- a/drivers/gpu/drm/i915/intel_guc_reg.h +++ b/drivers/gpu/drm/i915/intel_guc_reg.h @@ -52,7 +52,8 @@ #define SOFT_SCRATCH_COUNT 16 #define UOS_RSA_SCRATCH(i) _MMIO(0xc200 + (i) * 4) -#define UOS_RSA_SCRATCH_MAX_COUNT 64 +#define UOS_RSA_SCRATCH_COUNT 64 + #define DMA_ADDR_0_LOW _MMIO(0xc300) #define DMA_ADDR_0_HIGH _MMIO(0xc304) #define DMA_ADDR_1_LOW _MMIO(0xc308) diff --git a/drivers/gpu/drm/i915/intel_uc_fw.c b/drivers/gpu/drm/i915/intel_uc_fw.c index 4bc82d3005ff3..b376dd3b28cc8 100644 --- a/drivers/gpu/drm/i915/intel_uc_fw.c +++ b/drivers/gpu/drm/i915/intel_uc_fw.c @@ -105,7 +105,7 @@ void intel_uc_fw_fetch(struct drm_i915_private *dev_priv, } /* now RSA */ - if (css->key_size_dw != UOS_RSA_SCRATCH_MAX_COUNT) { + if (css->key_size_dw != UOS_RSA_SCRATCH_COUNT) { DRM_WARN("%s: Mismatched firmware RSA key size (%u)\n", intel_uc_fw_type_repr(uc_fw->type), css->key_size_dw); err = -ENOEXEC; -- GitLab From b46cb7d70e77ef5639bf2dcbef27be6e21e07a61 Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Fri, 24 Nov 2017 18:10:54 -0500 Subject: [PATCH 0430/1471] drm/amdkfd: Delete a useless parameter from create_queue function pointer Signed-off-by: Yong Zhao Reviewed-by: Felix Kuehling Signed-off-by: Oded Gabbay --- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 13 +++---------- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.h | 3 +-- .../gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 6 ++---- 3 files changed, 6 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 8447810c9a1e8..81ec7bb89b04c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -149,8 +149,7 @@ static void deallocate_vmid(struct device_queue_manager *dqm, static int create_queue_nocpsch(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd, - int *allocated_vmid) + struct qcm_process_device *qpd) { int retval; @@ -170,7 +169,6 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm, if (retval) goto out_unlock; } - *allocated_vmid = qpd->vmid; q->properties.vmid = qpd->vmid; q->properties.tba_addr = qpd->tba_addr; @@ -184,10 +182,8 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm, retval = -EINVAL; if (retval) { - if (list_empty(&qpd->queues_list)) { + if (list_empty(&qpd->queues_list)) deallocate_vmid(dqm, qpd, q); - *allocated_vmid = 0; - } goto out_unlock; } @@ -812,16 +808,13 @@ static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm, } static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd, int *allocate_vmid) + struct qcm_process_device *qpd) { int retval; struct mqd_manager *mqd; retval = 0; - if (allocate_vmid) - *allocate_vmid = 0; - mutex_lock(&dqm->lock); if (dqm->total_queue_count >= max_num_of_queues_per_device) { diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index 8752edf9cd9b3..c61b693bfa8c3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -84,8 +84,7 @@ struct device_process_node { struct device_queue_manager_ops { int (*create_queue)(struct device_queue_manager *dqm, struct queue *q, - struct qcm_process_device *qpd, - int *allocate_vmid); + struct qcm_process_device *qpd); int (*destroy_queue)(struct device_queue_manager *dqm, struct qcm_process_device *qpd, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index eeb7726b62991..fbfa2741e2a78 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -201,8 +201,7 @@ int pqm_create_queue(struct process_queue_manager *pqm, goto err_create_queue; pqn->q = q; pqn->kq = NULL; - retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd, - &q->properties.vmid); + retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd); pr_debug("DQM returned %d for create_queue\n", retval); print_queue(q); break; @@ -222,8 +221,7 @@ int pqm_create_queue(struct process_queue_manager *pqm, goto err_create_queue; pqn->q = q; pqn->kq = NULL; - retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd, - &q->properties.vmid); + retval = dev->dqm->ops.create_queue(dev->dqm, q, &pdd->qpd); pr_debug("DQM returned %d for create_queue\n", retval); print_queue(q); break; -- GitLab From ad88d7fc6c032ddfb32b8d496a070ab71de3a64f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sat, 25 Nov 2017 19:41:55 +0000 Subject: [PATCH 0431/1471] drm/i915/fbdev: Serialise early hotplug events with async fbdev config As both the hotplug event and fbdev configuration run asynchronously, it is possible for them to run concurrently. If configuration fails, we were freeing the fbdev causing a use-after-free in the hotplug event. <7>[ 3069.935211] [drm:intel_fb_initial_config [i915]] Not using firmware configuration <7>[ 3069.935225] [drm:drm_setup_crtcs] looking for cmdline mode on connector 77 <7>[ 3069.935229] [drm:drm_setup_crtcs] looking for preferred mode on connector 77 0 <7>[ 3069.935233] [drm:drm_setup_crtcs] found mode 3200x1800 <7>[ 3069.935236] [drm:drm_setup_crtcs] picking CRTCs for 8192x8192 config <7>[ 3069.935253] [drm:drm_setup_crtcs] desired mode 3200x1800 set on crtc 43 (0,0) <7>[ 3069.935323] [drm:intelfb_create [i915]] no BIOS fb, allocating a new one <4>[ 3069.967737] general protection fault: 0000 [#1] PREEMPT SMP <0>[ 3069.977453] --------------------------------- <4>[ 3069.977457] Modules linked in: i915(+) vgem snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm r8169 mei_me mii prime_numbers mei i2c_hid pinctrl_geminilake pinctrl_intel [last unloaded: i915] <4>[ 3069.977492] CPU: 1 PID: 15414 Comm: kworker/1:0 Tainted: G U 4.14.0-CI-CI_DRM_3388+ #1 <4>[ 3069.977497] Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017 <4>[ 3069.977508] Workqueue: events output_poll_execute <4>[ 3069.977512] task: ffff880177734e40 task.stack: ffffc90001fe4000 <4>[ 3069.977519] RIP: 0010:__lock_acquire+0x109/0x1b60 <4>[ 3069.977523] RSP: 0018:ffffc90001fe7bb0 EFLAGS: 00010002 <4>[ 3069.977526] RAX: 6b6b6b6b6b6b6b6b RBX: 0000000000000282 RCX: 0000000000000000 <4>[ 3069.977530] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff880170d4efd0 <4>[ 3069.977534] RBP: ffffc90001fe7c70 R08: 0000000000000001 R09: 0000000000000000 <4>[ 3069.977538] R10: 0000000000000000 R11: ffffffff81899609 R12: ffff880170d4efd0 <4>[ 3069.977542] R13: ffff880177734e40 R14: 0000000000000001 R15: 0000000000000000 <4>[ 3069.977547] FS: 0000000000000000(0000) GS:ffff88017fc80000(0000) knlGS:0000000000000000 <4>[ 3069.977551] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 <4>[ 3069.977555] CR2: 00007f7e8b7bcf04 CR3: 0000000003e0f000 CR4: 00000000003406e0 <4>[ 3069.977559] Call Trace: <4>[ 3069.977565] ? mark_held_locks+0x64/0x90 <4>[ 3069.977571] ? _raw_spin_unlock_irq+0x24/0x50 <4>[ 3069.977575] ? _raw_spin_unlock_irq+0x24/0x50 <4>[ 3069.977579] ? trace_hardirqs_on_caller+0xde/0x1c0 <4>[ 3069.977583] ? _raw_spin_unlock_irq+0x2f/0x50 <4>[ 3069.977588] ? finish_task_switch+0xa5/0x210 <4>[ 3069.977592] ? lock_acquire+0xaf/0x200 <4>[ 3069.977596] lock_acquire+0xaf/0x200 <4>[ 3069.977600] ? __mutex_lock+0x5e9/0x9b0 <4>[ 3069.977604] _raw_spin_lock+0x2a/0x40 <4>[ 3069.977608] ? __mutex_lock+0x5e9/0x9b0 <4>[ 3069.977612] __mutex_lock+0x5e9/0x9b0 <4>[ 3069.977616] ? drm_fb_helper_hotplug_event.part.19+0x16/0xa0 <4>[ 3069.977621] ? drm_fb_helper_hotplug_event.part.19+0x16/0xa0 <4>[ 3069.977625] drm_fb_helper_hotplug_event.part.19+0x16/0xa0 <4>[ 3069.977630] output_poll_execute+0x8d/0x180 <4>[ 3069.977635] process_one_work+0x22e/0x660 <4>[ 3069.977640] worker_thread+0x48/0x3a0 <4>[ 3069.977644] ? _raw_spin_unlock_irqrestore+0x4c/0x60 <4>[ 3069.977649] kthread+0x102/0x140 <4>[ 3069.977653] ? process_one_work+0x660/0x660 <4>[ 3069.977657] ? kthread_create_on_node+0x40/0x40 <4>[ 3069.977662] ret_from_fork+0x27/0x40 <4>[ 3069.977666] Code: 8d 62 f8 c3 49 81 3c 24 e0 fa 3c 82 41 be 00 00 00 00 45 0f 45 f0 83 fe 01 77 86 89 f0 49 8b 44 c4 08 48 85 c0 0f 84 76 ff ff ff ff 80 38 01 00 00 8b 1d 62 f9 e8 01 45 8b 85 b8 08 00 00 85 <1>[ 3069.977707] RIP: __lock_acquire+0x109/0x1b60 RSP: ffffc90001fe7bb0 <4>[ 3069.977712] ---[ end trace 4ad012eb3af62df7 ]--- In order to keep the dev_priv->ifbdev alive after failure, we have to avoid the free and leave it empty until we unload the module (which is less than ideal, but a necessary evil for simplicity). Then we can use intel_fbdev_sync() to serialise the hotplug event with the configuration. The serialisation between the two was removed in commit 934458c2c95d ("Revert "drm/i915: Fix races on fbdev""), but the use after free is much older, commit 366e39b4d2c5 ("drm/i915: Tear down fbdev if initialization fails") Fixes: 366e39b4d2c5 ("drm/i915: Tear down fbdev if initialization fails") Fixes: 934458c2c95d ("Revert "drm/i915: Fix races on fbdev"") Signed-off-by: Chris Wilson Cc: Lukas Wunner Cc: Joonas Lahtinen Cc: Daniel Vetter Cc: stable@vger.kernel.org Reviewed-by: Lukas Wunner Link: https://patchwork.freedesktop.org/patch/msgid/20171125194155.355-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/intel_fbdev.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index b8af35187d226..ea96682568e88 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -697,10 +697,8 @@ static void intel_fbdev_initial_config(void *data, async_cookie_t cookie) /* Due to peculiar init order wrt to hpd handling this is separate. */ if (drm_fb_helper_initial_config(&ifbdev->helper, - ifbdev->preferred_bpp)) { + ifbdev->preferred_bpp)) intel_fbdev_unregister(to_i915(ifbdev->helper.dev)); - intel_fbdev_fini(to_i915(ifbdev->helper.dev)); - } } void intel_fbdev_initial_config_async(struct drm_device *dev) @@ -800,7 +798,11 @@ void intel_fbdev_output_poll_changed(struct drm_device *dev) { struct intel_fbdev *ifbdev = to_i915(dev)->fbdev; - if (ifbdev) + if (!ifbdev) + return; + + intel_fbdev_sync(ifbdev); + if (ifbdev->vma) drm_fb_helper_hotplug_event(&ifbdev->helper); } -- GitLab From 44a2d56429bc9d1f9edd56761bfb504ed68fef01 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Mon, 27 Nov 2017 12:17:03 +0100 Subject: [PATCH 0432/1471] drm/via: use monotonic time for VIA_WAIT_IRQ The normal DRM vblank interrupt events started using monotonic times in commit c61eef726a78 ("drm: add support for monotonic vblank timestamps"), which is useful for a number of reasons, including the possible y2038 overflow. The VIA_WAIT_IRQ ioctl uses the same drm_wait_vblank_reply as DRM_IOCTL_WAIT_VBLANK, but still uses wall-clock time. This converts it to using monotonic time as well, which is more consistent, and avoids problems with the y2038 overflow as well as synchronization issues when the real time skips. I could not find the matching user space that calls the VIA_WAIT_IRQ ioctl to verify that this is safe, but it very likely is. Please either test or review the user space side before applying this. Signed-off-by: Arnd Bergmann Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171127111728.1710742-1-arnd@arndb.de --- drivers/gpu/drm/via/via_irq.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/via/via_irq.c b/drivers/gpu/drm/via/via_irq.c index 24e71578af4df..c96830ccc0ec4 100644 --- a/drivers/gpu/drm/via/via_irq.c +++ b/drivers/gpu/drm/via/via_irq.c @@ -343,7 +343,7 @@ void via_driver_irq_uninstall(struct drm_device *dev) int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_via_irqwait_t *irqwait = data; - struct timeval now; + struct timespec64 now; int ret = 0; drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; drm_via_irq_t *cur_irq = dev_priv->via_irqs; @@ -377,9 +377,9 @@ int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv) ret = via_driver_irq_wait(dev, irqwait->request.irq, force_sequence, &irqwait->request.sequence); - do_gettimeofday(&now); + ktime_get_ts64(&now); irqwait->reply.tval_sec = now.tv_sec; - irqwait->reply.tval_usec = now.tv_usec; + irqwait->reply.tval_usec = now.tv_nsec / NSEC_PER_USEC; return ret; } -- GitLab From 2de2d0b063b08becb2c67a2c338c44e37bdcffee Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Mon, 27 Nov 2017 16:10:27 +0100 Subject: [PATCH 0433/1471] drm/i915: fix intel_backlight_device_register declaration The alternative intel_backlight_device_register() definition apparently never got used, but I have now run into a case of i915 being compiled without CONFIG_BACKLIGHT_CLASS_DEVICE, resulting in a number of identical warnings: drivers/gpu/drm/i915/intel_drv.h:1739:12: error: 'intel_backlight_device_register' defined but not used [-Werror=unused-function] This marks the function as 'inline', which was surely the original intention here. Fixes: 1ebaa0b9c2d4 ("drm/i915: Move backlight registration to connector registration") Signed-off-by: Arnd Bergmann Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171127151239.1813673-1-arnd@arndb.de --- drivers/gpu/drm/i915/intel_drv.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index c00441a3d6495..c679df99d5300 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1750,7 +1750,7 @@ extern struct drm_display_mode *intel_find_panel_downclock( int intel_backlight_device_register(struct intel_connector *connector); void intel_backlight_device_unregister(struct intel_connector *connector); #else /* CONFIG_BACKLIGHT_CLASS_DEVICE */ -static int intel_backlight_device_register(struct intel_connector *connector) +static inline int intel_backlight_device_register(struct intel_connector *connector) { return 0; } -- GitLab From fd94d53e55bd487368dfee9f1af24da78b2bb582 Mon Sep 17 00:00:00 2001 From: Arnd Bergmann Date: Mon, 27 Nov 2017 16:10:28 +0100 Subject: [PATCH 0434/1471] drm/i915: hide unused intel_panel_set_backlight function Building i915 without backlight support results in a harmless warning for intel_panel_set_backlight: drivers/gpu/drm/i915/intel_panel.c:653:13: error: 'intel_panel_set_backlight' defined but not used [-Werror=unused-function] This moves it into the CONFIG_BACKLIGHT_CLASS_DEVICE section that its caller is in. Signed-off-by: Arnd Bergmann Signed-off-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171127151239.1813673-2-arnd@arndb.de --- drivers/gpu/drm/i915/intel_panel.c | 88 +++++++++++++++--------------- 1 file changed, 44 insertions(+), 44 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index adc51e452e3ee..fa6831f8c004e 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -543,25 +543,6 @@ static u32 pwm_get_backlight(struct intel_connector *connector) return DIV_ROUND_UP(duty_ns * 100, CRC_PMIC_PWM_PERIOD_NS); } -static u32 intel_panel_get_backlight(struct intel_connector *connector) -{ - struct drm_i915_private *dev_priv = to_i915(connector->base.dev); - struct intel_panel *panel = &connector->panel; - u32 val = 0; - - mutex_lock(&dev_priv->backlight_lock); - - if (panel->backlight.enabled) { - val = panel->backlight.get(connector); - val = intel_panel_compute_brightness(connector, val); - } - - mutex_unlock(&dev_priv->backlight_lock); - - DRM_DEBUG_DRIVER("get backlight PWM = %d\n", val); - return val; -} - static void lpt_set_backlight(const struct drm_connector_state *conn_state, u32 level) { struct intel_connector *connector = to_intel_connector(conn_state->connector); @@ -649,31 +630,6 @@ intel_panel_actually_set_backlight(const struct drm_connector_state *conn_state, panel->backlight.set(conn_state, level); } -/* set backlight brightness to level in range [0..max], scaling wrt hw min */ -static void intel_panel_set_backlight(const struct drm_connector_state *conn_state, - u32 user_level, u32 user_max) -{ - struct intel_connector *connector = to_intel_connector(conn_state->connector); - struct drm_i915_private *dev_priv = to_i915(connector->base.dev); - struct intel_panel *panel = &connector->panel; - u32 hw_level; - - if (!panel->backlight.present) - return; - - mutex_lock(&dev_priv->backlight_lock); - - WARN_ON(panel->backlight.max == 0); - - hw_level = scale_user_to_hw(connector, user_level, user_max); - panel->backlight.level = hw_level; - - if (panel->backlight.enabled) - intel_panel_actually_set_backlight(conn_state, hw_level); - - mutex_unlock(&dev_priv->backlight_lock); -} - /* set backlight brightness to level in range [0..max], assuming hw min is * respected. */ @@ -1182,6 +1138,50 @@ void intel_panel_enable_backlight(const struct intel_crtc_state *crtc_state, } #if IS_ENABLED(CONFIG_BACKLIGHT_CLASS_DEVICE) +static u32 intel_panel_get_backlight(struct intel_connector *connector) +{ + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); + struct intel_panel *panel = &connector->panel; + u32 val = 0; + + mutex_lock(&dev_priv->backlight_lock); + + if (panel->backlight.enabled) { + val = panel->backlight.get(connector); + val = intel_panel_compute_brightness(connector, val); + } + + mutex_unlock(&dev_priv->backlight_lock); + + DRM_DEBUG_DRIVER("get backlight PWM = %d\n", val); + return val; +} + +/* set backlight brightness to level in range [0..max], scaling wrt hw min */ +static void intel_panel_set_backlight(const struct drm_connector_state *conn_state, + u32 user_level, u32 user_max) +{ + struct intel_connector *connector = to_intel_connector(conn_state->connector); + struct drm_i915_private *dev_priv = to_i915(connector->base.dev); + struct intel_panel *panel = &connector->panel; + u32 hw_level; + + if (!panel->backlight.present) + return; + + mutex_lock(&dev_priv->backlight_lock); + + WARN_ON(panel->backlight.max == 0); + + hw_level = scale_user_to_hw(connector, user_level, user_max); + panel->backlight.level = hw_level; + + if (panel->backlight.enabled) + intel_panel_actually_set_backlight(conn_state, hw_level); + + mutex_unlock(&dev_priv->backlight_lock); +} + static int intel_backlight_device_update_status(struct backlight_device *bd) { struct intel_connector *connector = bl_get_data(bd); -- GitLab From c82297fcfaff2862686094061772070e6763e00c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 26 Nov 2017 21:48:56 +0000 Subject: [PATCH 0435/1471] drm/i915: Flush everything on switching to the kernel_context Even though all rendering should have been flushed at the end of the previous requests, add an extra flush after switching to the kernel_context. As the switch to the kernel_context is used when idling the gpu (e.g. suspend), having an extra layer of paranoia to ensure everything is flushed to memory seems sensible. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171126214856.23702-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem_context.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c index aee0f6d72d337..ce3139e5ec4c4 100644 --- a/drivers/gpu/drm/i915/i915_gem_context.c +++ b/drivers/gpu/drm/i915/i915_gem_context.c @@ -620,7 +620,14 @@ int i915_gem_switch_to_kernel_context(struct drm_i915_private *dev_priv) GFP_KERNEL); } - i915_add_request(req); + /* + * Force a flush after the switch to ensure that all rendering + * and operations prior to switching to the kernel context hits + * memory. This should be guaranteed by the previous request, + * but an extra layer of paranoia before we declare the system + * idle (on suspend etc) is advisable! + */ + __i915_add_request(req, true); } return 0; -- GitLab From 4e90a6e222720dd0ec529f87eca990c736ba8ede Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 26 Nov 2017 22:09:01 +0000 Subject: [PATCH 0436/1471] drm/i915: Record default HW state in the GPU error state It may be of interest to both compare the active HW context against the default (aka NULL) context, to see what has been changed and if either are corrupt. v2: Rename the fake vma as fake. Signed-off-by: Chris Wilson Cc: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20171126220901.14735-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gpu_error.c | 23 +++++++++++++++++++++++ 2 files changed, 24 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d575a56fc1000..50a945d3c11de 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1009,6 +1009,7 @@ struct i915_gpu_state { long user_bo_count; struct drm_i915_error_object *wa_ctx; + struct drm_i915_error_object *default_state; struct drm_i915_error_request { long jiffies; diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 3d18d67d065f1..876be8f1d9306 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -791,6 +791,9 @@ int i915_error_state_to_str(struct drm_i915_error_state_buf *m, print_error_obj(m, dev_priv->engine[i], "WA batchbuffer", ee->wa_batchbuffer); + + print_error_obj(m, dev_priv->engine[i], + "NULL context", ee->default_state); } if (error->overlay) @@ -1414,6 +1417,23 @@ static void request_record_user_bo(struct drm_i915_gem_request *request, ee->user_bo_count = count; } +static struct drm_i915_error_object * +capture_object(struct drm_i915_private *dev_priv, + struct drm_i915_gem_object *obj) +{ + if (obj && i915_gem_object_has_pages(obj)) { + struct i915_vma fake = { + .node = { .start = U64_MAX, .size = obj->base.size }, + .pages = obj->mm.pages, + .obj = obj, + }; + + return i915_error_object_create(dev_priv, &fake); + } else { + return NULL; + } +} + static void i915_gem_record_rings(struct drm_i915_private *dev_priv, struct i915_gpu_state *error) { @@ -1485,6 +1505,9 @@ static void i915_gem_record_rings(struct drm_i915_private *dev_priv, ee->wa_ctx = i915_error_object_create(dev_priv, engine->wa_ctx.vma); + + ee->default_state = + capture_object(dev_priv, engine->default_state); } } -- GitLab From 848b365d5d1f0d41d07773c6a69149b2561f29de Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 11:53:37 +0000 Subject: [PATCH 0437/1471] drm/i915: Rename shrinker init/cleanup to match driver initialisation phase Since the shrinker is registered and unregistered during i915_driver_register and i915_driver_unregister, respectively, rename the init/cleanup functions to match. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171123115338.10270-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_drv.c | 4 ++-- drivers/gpu/drm/i915/i915_drv.h | 4 ++-- drivers/gpu/drm/i915/i915_gem_shrinker.c | 8 ++++---- 3 files changed, 8 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 0793a27e2b95b..9c519aa4f550d 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1215,7 +1215,7 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) { struct drm_device *dev = &dev_priv->drm; - i915_gem_shrinker_init(dev_priv); + i915_gem_shrinker_register(dev_priv); i915_pmu_register(dev_priv); /* @@ -1277,7 +1277,7 @@ static void i915_driver_unregister(struct drm_i915_private *dev_priv) i915_guc_log_unregister(dev_priv); drm_dev_unregister(&dev_priv->drm); - i915_gem_shrinker_cleanup(dev_priv); + i915_gem_shrinker_unregister(dev_priv); } /** diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 50a945d3c11de..2dd32de47f9a2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3921,8 +3921,8 @@ unsigned long i915_gem_shrink(struct drm_i915_private *dev_priv, #define I915_SHRINK_ACTIVE 0x8 #define I915_SHRINK_VMAPS 0x10 unsigned long i915_gem_shrink_all(struct drm_i915_private *dev_priv); -void i915_gem_shrinker_init(struct drm_i915_private *dev_priv); -void i915_gem_shrinker_cleanup(struct drm_i915_private *dev_priv); +void i915_gem_shrinker_register(struct drm_i915_private *dev_priv); +void i915_gem_shrinker_unregister(struct drm_i915_private *dev_priv); /* i915_gem_tiling.c */ diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c index 3770e3323fc8e..7205dc123d92e 100644 --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c @@ -494,12 +494,12 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr } /** - * i915_gem_shrinker_init - Initialize i915 shrinker + * i915_gem_shrinker_register - Register the i915 shrinker * @dev_priv: i915 device * * This function registers and sets up the i915 shrinker and OOM handler. */ -void i915_gem_shrinker_init(struct drm_i915_private *dev_priv) +void i915_gem_shrinker_register(struct drm_i915_private *dev_priv) { dev_priv->mm.shrinker.scan_objects = i915_gem_shrinker_scan; dev_priv->mm.shrinker.count_objects = i915_gem_shrinker_count; @@ -515,12 +515,12 @@ void i915_gem_shrinker_init(struct drm_i915_private *dev_priv) } /** - * i915_gem_shrinker_cleanup - Clean up i915 shrinker + * i915_gem_shrinker_unregister - Unregisters the i915 shrinker * @dev_priv: i915 device * * This function unregisters the i915 shrinker and OOM handler. */ -void i915_gem_shrinker_cleanup(struct drm_i915_private *dev_priv) +void i915_gem_shrinker_unregister(struct drm_i915_private *dev_priv) { WARN_ON(unregister_vmap_purge_notifier(&dev_priv->mm.vmap_notifier)); WARN_ON(unregister_oom_notifier(&dev_priv->mm.oom_notifier)); -- GitLab From 56fa4bf2b2f08405739a9166f306ceaea421447f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 23 Nov 2017 11:53:38 +0000 Subject: [PATCH 0438/1471] drm/i915: Update shrinker drm_i915_private naming convention Switch over from the non-descript dev_priv locals to i915. Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171123115338.10270-2-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_drv.h | 8 +- drivers/gpu/drm/i915/i915_gem_shrinker.c | 134 +++++++++++------------ 2 files changed, 71 insertions(+), 71 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2dd32de47f9a2..bddd65839f604 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -3911,7 +3911,7 @@ i915_gem_object_create_internal(struct drm_i915_private *dev_priv, phys_addr_t size); /* i915_gem_shrinker.c */ -unsigned long i915_gem_shrink(struct drm_i915_private *dev_priv, +unsigned long i915_gem_shrink(struct drm_i915_private *i915, unsigned long target, unsigned long *nr_scanned, unsigned flags); @@ -3920,9 +3920,9 @@ unsigned long i915_gem_shrink(struct drm_i915_private *dev_priv, #define I915_SHRINK_BOUND 0x4 #define I915_SHRINK_ACTIVE 0x8 #define I915_SHRINK_VMAPS 0x10 -unsigned long i915_gem_shrink_all(struct drm_i915_private *dev_priv); -void i915_gem_shrinker_register(struct drm_i915_private *dev_priv); -void i915_gem_shrinker_unregister(struct drm_i915_private *dev_priv); +unsigned long i915_gem_shrink_all(struct drm_i915_private *i915); +void i915_gem_shrinker_register(struct drm_i915_private *i915); +void i915_gem_shrinker_unregister(struct drm_i915_private *i915); /* i915_gem_tiling.c */ diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c index 7205dc123d92e..9029ed04879c3 100644 --- a/drivers/gpu/drm/i915/i915_gem_shrinker.c +++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c @@ -35,9 +35,9 @@ #include "i915_drv.h" #include "i915_trace.h" -static bool shrinker_lock(struct drm_i915_private *dev_priv, bool *unlock) +static bool shrinker_lock(struct drm_i915_private *i915, bool *unlock) { - switch (mutex_trylock_recursive(&dev_priv->drm.struct_mutex)) { + switch (mutex_trylock_recursive(&i915->drm.struct_mutex)) { case MUTEX_TRYLOCK_RECURSIVE: *unlock = false; return true; @@ -47,7 +47,7 @@ static bool shrinker_lock(struct drm_i915_private *dev_priv, bool *unlock) preempt_disable(); do { cpu_relax(); - if (mutex_trylock(&dev_priv->drm.struct_mutex)) { + if (mutex_trylock(&i915->drm.struct_mutex)) { *unlock = true; break; } @@ -63,12 +63,12 @@ static bool shrinker_lock(struct drm_i915_private *dev_priv, bool *unlock) BUG(); } -static void shrinker_unlock(struct drm_i915_private *dev_priv, bool unlock) +static void shrinker_unlock(struct drm_i915_private *i915, bool unlock) { if (!unlock) return; - mutex_unlock(&dev_priv->drm.struct_mutex); + mutex_unlock(&i915->drm.struct_mutex); } static bool swap_available(void) @@ -118,7 +118,7 @@ static bool unsafe_drop_pages(struct drm_i915_gem_object *obj) /** * i915_gem_shrink - Shrink buffer object caches - * @dev_priv: i915 device + * @i915: i915 device * @target: amount of memory to make available, in pages * @nr_scanned: optional output for number of pages scanned (incremental) * @flags: control flags for selecting cache types @@ -142,7 +142,7 @@ static bool unsafe_drop_pages(struct drm_i915_gem_object *obj) * The number of pages of backing storage actually released. */ unsigned long -i915_gem_shrink(struct drm_i915_private *dev_priv, +i915_gem_shrink(struct drm_i915_private *i915, unsigned long target, unsigned long *nr_scanned, unsigned flags) @@ -151,15 +151,15 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, struct list_head *list; unsigned int bit; } phases[] = { - { &dev_priv->mm.unbound_list, I915_SHRINK_UNBOUND }, - { &dev_priv->mm.bound_list, I915_SHRINK_BOUND }, + { &i915->mm.unbound_list, I915_SHRINK_UNBOUND }, + { &i915->mm.bound_list, I915_SHRINK_BOUND }, { NULL, 0 }, }, *phase; unsigned long count = 0; unsigned long scanned = 0; bool unlock; - if (!shrinker_lock(dev_priv, &unlock)) + if (!shrinker_lock(i915, &unlock)) return 0; /* @@ -172,10 +172,10 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, * we will free as much as we can and hope to get a second chance. */ if (flags & I915_SHRINK_ACTIVE) - i915_gem_wait_for_idle(dev_priv, I915_WAIT_LOCKED); + i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED); - trace_i915_gem_shrink(dev_priv, target, flags); - i915_gem_retire_requests(dev_priv); + trace_i915_gem_shrink(i915, target, flags); + i915_gem_retire_requests(i915); /* * Unbinding of objects will require HW access; Let us not wake the @@ -183,7 +183,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, * we will force the wake during oom-notifier. */ if ((flags & I915_SHRINK_BOUND) && - !intel_runtime_pm_get_if_in_use(dev_priv)) + !intel_runtime_pm_get_if_in_use(i915)) flags &= ~I915_SHRINK_BOUND; /* @@ -221,7 +221,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, * to be able to shrink their pages, so they remain on * the unbound/bound list until actually freed. */ - spin_lock(&dev_priv->mm.obj_lock); + spin_lock(&i915->mm.obj_lock); while (count < target && (obj = list_first_entry_or_null(phase->list, typeof(*obj), @@ -244,7 +244,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, if (!can_release_pages(obj)) continue; - spin_unlock(&dev_priv->mm.obj_lock); + spin_unlock(&i915->mm.obj_lock); if (unsafe_drop_pages(obj)) { /* May arrive from get_pages on another bo */ @@ -258,18 +258,18 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, } scanned += obj->base.size >> PAGE_SHIFT; - spin_lock(&dev_priv->mm.obj_lock); + spin_lock(&i915->mm.obj_lock); } list_splice_tail(&still_in_list, phase->list); - spin_unlock(&dev_priv->mm.obj_lock); + spin_unlock(&i915->mm.obj_lock); } if (flags & I915_SHRINK_BOUND) - intel_runtime_pm_put(dev_priv); + intel_runtime_pm_put(i915); - i915_gem_retire_requests(dev_priv); + i915_gem_retire_requests(i915); - shrinker_unlock(dev_priv, unlock); + shrinker_unlock(i915, unlock); if (nr_scanned) *nr_scanned += scanned; @@ -278,7 +278,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, /** * i915_gem_shrink_all - Shrink buffer object caches completely - * @dev_priv: i915 device + * @i915: i915 device * * This is a simple wraper around i915_gem_shrink() to aggressively shrink all * caches completely. It also first waits for and retires all outstanding @@ -290,16 +290,16 @@ i915_gem_shrink(struct drm_i915_private *dev_priv, * Returns: * The number of pages of backing storage actually released. */ -unsigned long i915_gem_shrink_all(struct drm_i915_private *dev_priv) +unsigned long i915_gem_shrink_all(struct drm_i915_private *i915) { unsigned long freed; - intel_runtime_pm_get(dev_priv); - freed = i915_gem_shrink(dev_priv, -1UL, NULL, + intel_runtime_pm_get(i915); + freed = i915_gem_shrink(i915, -1UL, NULL, I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_ACTIVE); - intel_runtime_pm_put(dev_priv); + intel_runtime_pm_put(i915); return freed; } @@ -347,53 +347,53 @@ i915_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc) static unsigned long i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc) { - struct drm_i915_private *dev_priv = + struct drm_i915_private *i915 = container_of(shrinker, struct drm_i915_private, mm.shrinker); unsigned long freed; bool unlock; sc->nr_scanned = 0; - if (!shrinker_lock(dev_priv, &unlock)) + if (!shrinker_lock(i915, &unlock)) return SHRINK_STOP; - freed = i915_gem_shrink(dev_priv, + freed = i915_gem_shrink(i915, sc->nr_to_scan, &sc->nr_scanned, I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_PURGEABLE); if (freed < sc->nr_to_scan) - freed += i915_gem_shrink(dev_priv, + freed += i915_gem_shrink(i915, sc->nr_to_scan - sc->nr_scanned, &sc->nr_scanned, I915_SHRINK_BOUND | I915_SHRINK_UNBOUND); if (freed < sc->nr_to_scan && current_is_kswapd()) { - intel_runtime_pm_get(dev_priv); - freed += i915_gem_shrink(dev_priv, + intel_runtime_pm_get(i915); + freed += i915_gem_shrink(i915, sc->nr_to_scan - sc->nr_scanned, &sc->nr_scanned, I915_SHRINK_ACTIVE | I915_SHRINK_BOUND | I915_SHRINK_UNBOUND); - intel_runtime_pm_put(dev_priv); + intel_runtime_pm_put(i915); } - shrinker_unlock(dev_priv, unlock); + shrinker_unlock(i915, unlock); return sc->nr_scanned ? freed : SHRINK_STOP; } static bool -shrinker_lock_uninterruptible(struct drm_i915_private *dev_priv, bool *unlock, +shrinker_lock_uninterruptible(struct drm_i915_private *i915, bool *unlock, int timeout_ms) { unsigned long timeout = jiffies + msecs_to_jiffies_timeout(timeout_ms); do { - if (i915_gem_wait_for_idle(dev_priv, 0) == 0 && - shrinker_lock(dev_priv, unlock)) + if (i915_gem_wait_for_idle(i915, 0) == 0 && + shrinker_lock(i915, unlock)) break; schedule_timeout_killable(1); @@ -412,32 +412,32 @@ shrinker_lock_uninterruptible(struct drm_i915_private *dev_priv, bool *unlock, static int i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) { - struct drm_i915_private *dev_priv = + struct drm_i915_private *i915 = container_of(nb, struct drm_i915_private, mm.oom_notifier); struct drm_i915_gem_object *obj; unsigned long unevictable, bound, unbound, freed_pages; - freed_pages = i915_gem_shrink_all(dev_priv); + freed_pages = i915_gem_shrink_all(i915); /* Because we may be allocating inside our own driver, we cannot * assert that there are no objects with pinned pages that are not * being pointed to by hardware. */ unbound = bound = unevictable = 0; - spin_lock(&dev_priv->mm.obj_lock); - list_for_each_entry(obj, &dev_priv->mm.unbound_list, mm.link) { + spin_lock(&i915->mm.obj_lock); + list_for_each_entry(obj, &i915->mm.unbound_list, mm.link) { if (!can_release_pages(obj)) unevictable += obj->base.size >> PAGE_SHIFT; else unbound += obj->base.size >> PAGE_SHIFT; } - list_for_each_entry(obj, &dev_priv->mm.bound_list, mm.link) { + list_for_each_entry(obj, &i915->mm.bound_list, mm.link) { if (!can_release_pages(obj)) unevictable += obj->base.size >> PAGE_SHIFT; else bound += obj->base.size >> PAGE_SHIFT; } - spin_unlock(&dev_priv->mm.obj_lock); + spin_unlock(&i915->mm.obj_lock); if (freed_pages || unbound || bound) pr_info("Purging GPU memory, %lu pages freed, " @@ -455,39 +455,39 @@ i915_gem_shrinker_oom(struct notifier_block *nb, unsigned long event, void *ptr) static int i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr) { - struct drm_i915_private *dev_priv = + struct drm_i915_private *i915 = container_of(nb, struct drm_i915_private, mm.vmap_notifier); struct i915_vma *vma, *next; unsigned long freed_pages = 0; bool unlock; int ret; - if (!shrinker_lock_uninterruptible(dev_priv, &unlock, 5000)) + if (!shrinker_lock_uninterruptible(i915, &unlock, 5000)) return NOTIFY_DONE; /* Force everything onto the inactive lists */ - ret = i915_gem_wait_for_idle(dev_priv, I915_WAIT_LOCKED); + ret = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED); if (ret) goto out; - intel_runtime_pm_get(dev_priv); - freed_pages += i915_gem_shrink(dev_priv, -1UL, NULL, + intel_runtime_pm_get(i915); + freed_pages += i915_gem_shrink(i915, -1UL, NULL, I915_SHRINK_BOUND | I915_SHRINK_UNBOUND | I915_SHRINK_ACTIVE | I915_SHRINK_VMAPS); - intel_runtime_pm_put(dev_priv); + intel_runtime_pm_put(i915); /* We also want to clear any cached iomaps as they wrap vmap */ list_for_each_entry_safe(vma, next, - &dev_priv->ggtt.base.inactive_list, vm_link) { + &i915->ggtt.base.inactive_list, vm_link) { unsigned long count = vma->node.size >> PAGE_SHIFT; if (vma->iomap && i915_vma_unbind(vma) == 0) freed_pages += count; } out: - shrinker_unlock(dev_priv, unlock); + shrinker_unlock(i915, unlock); *(unsigned long *)ptr += freed_pages; return NOTIFY_DONE; @@ -495,34 +495,34 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr /** * i915_gem_shrinker_register - Register the i915 shrinker - * @dev_priv: i915 device + * @i915: i915 device * * This function registers and sets up the i915 shrinker and OOM handler. */ -void i915_gem_shrinker_register(struct drm_i915_private *dev_priv) +void i915_gem_shrinker_register(struct drm_i915_private *i915) { - dev_priv->mm.shrinker.scan_objects = i915_gem_shrinker_scan; - dev_priv->mm.shrinker.count_objects = i915_gem_shrinker_count; - dev_priv->mm.shrinker.seeks = DEFAULT_SEEKS; - dev_priv->mm.shrinker.batch = 4096; - WARN_ON(register_shrinker(&dev_priv->mm.shrinker)); + i915->mm.shrinker.scan_objects = i915_gem_shrinker_scan; + i915->mm.shrinker.count_objects = i915_gem_shrinker_count; + i915->mm.shrinker.seeks = DEFAULT_SEEKS; + i915->mm.shrinker.batch = 4096; + WARN_ON(register_shrinker(&i915->mm.shrinker)); - dev_priv->mm.oom_notifier.notifier_call = i915_gem_shrinker_oom; - WARN_ON(register_oom_notifier(&dev_priv->mm.oom_notifier)); + i915->mm.oom_notifier.notifier_call = i915_gem_shrinker_oom; + WARN_ON(register_oom_notifier(&i915->mm.oom_notifier)); - dev_priv->mm.vmap_notifier.notifier_call = i915_gem_shrinker_vmap; - WARN_ON(register_vmap_purge_notifier(&dev_priv->mm.vmap_notifier)); + i915->mm.vmap_notifier.notifier_call = i915_gem_shrinker_vmap; + WARN_ON(register_vmap_purge_notifier(&i915->mm.vmap_notifier)); } /** * i915_gem_shrinker_unregister - Unregisters the i915 shrinker - * @dev_priv: i915 device + * @i915: i915 device * * This function unregisters the i915 shrinker and OOM handler. */ -void i915_gem_shrinker_unregister(struct drm_i915_private *dev_priv) +void i915_gem_shrinker_unregister(struct drm_i915_private *i915) { - WARN_ON(unregister_vmap_purge_notifier(&dev_priv->mm.vmap_notifier)); - WARN_ON(unregister_oom_notifier(&dev_priv->mm.oom_notifier)); - unregister_shrinker(&dev_priv->mm.shrinker); + WARN_ON(unregister_vmap_purge_notifier(&i915->mm.vmap_notifier)); + WARN_ON(unregister_oom_notifier(&i915->mm.oom_notifier)); + unregister_shrinker(&i915->mm.shrinker); } -- GitLab From d02a1d830880b611000aaa0df8a8d7aabf1d8718 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 27 Nov 2017 12:30:54 +0000 Subject: [PATCH 0439/1471] drm/i915: Rename i915_gem_timelines_mark_idle The kerneldoc markup for i915_gem_timelines_mark_idle() was incorrect, so take the opportunity to also convert it from the "mark_idle" to "park" naming scheme. drivers/gpu/drm/i915/i915_gem_timeline.c:120: warning: No description found for parameter 'i915' Signed-off-by: Chris Wilson Cc: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171127123054.20966-1-chris@chris-wilson.co.uk Reviewed-by: Joonas Lahtinen --- drivers/gpu/drm/i915/i915_gem.c | 3 ++- drivers/gpu/drm/i915/i915_gem_timeline.c | 6 +++--- drivers/gpu/drm/i915/i915_gem_timeline.h | 2 +- 3 files changed, 6 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index e03d6c2554e2b..354b0546a1915 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3371,7 +3371,8 @@ i915_gem_idle_work_handler(struct work_struct *work) synchronize_irq(dev_priv->drm.irq); intel_engines_park(dev_priv); - i915_gem_timelines_mark_idle(dev_priv); + i915_gem_timelines_park(dev_priv); + i915_pmu_gt_parked(dev_priv); GEM_BUG_ON(!dev_priv->gt.awake); diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.c b/drivers/gpu/drm/i915/i915_gem_timeline.c index c597ce277a043..c01905d6450c8 100644 --- a/drivers/gpu/drm/i915/i915_gem_timeline.c +++ b/drivers/gpu/drm/i915/i915_gem_timeline.c @@ -107,8 +107,8 @@ int i915_gem_timeline_init__global(struct drm_i915_private *i915) } /** - * i915_gem_timelines_mark_idle -- called when the driver idles - * @i915 - the drm_i915_private device + * i915_gem_timelines_park - called when the driver idles + * @i915: the drm_i915_private device * * When the driver is completely idle, we know that all of our sync points * have been signaled and our tracking is then entirely redundant. Any request @@ -116,7 +116,7 @@ int i915_gem_timeline_init__global(struct drm_i915_private *i915) * the fence is signaled and therefore we will not even look them up in the * sync point map. */ -void i915_gem_timelines_mark_idle(struct drm_i915_private *i915) +void i915_gem_timelines_park(struct drm_i915_private *i915) { struct i915_gem_timeline *timeline; int i; diff --git a/drivers/gpu/drm/i915/i915_gem_timeline.h b/drivers/gpu/drm/i915/i915_gem_timeline.h index bfb5eb94c64d0..b5a22400a01f8 100644 --- a/drivers/gpu/drm/i915/i915_gem_timeline.h +++ b/drivers/gpu/drm/i915/i915_gem_timeline.h @@ -93,7 +93,7 @@ int i915_gem_timeline_init(struct drm_i915_private *i915, struct i915_gem_timeline *tl, const char *name); int i915_gem_timeline_init__global(struct drm_i915_private *i915); -void i915_gem_timelines_mark_idle(struct drm_i915_private *i915); +void i915_gem_timelines_park(struct drm_i915_private *i915); void i915_gem_timeline_fini(struct i915_gem_timeline *tl); static inline int __intel_timeline_sync_set(struct intel_timeline *tl, -- GitLab From a9efcc19161e20623c285fac967a32842972cebe Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:43 -0500 Subject: [PATCH 0440/1471] drm/amdgpu: fix get_max_engine_clock_in_mhz Use proper powerplay function. This fixes OpenCL initialization problems. Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 896b16db58aac..01aa851938f81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -298,6 +298,9 @@ uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; - /* The sclk is in quantas of 10kHz */ - return adev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk / 100; + /* the sclk is in quantas of 10kHz */ + if (amdgpu_sriov_vf(adev)) + return adev->clock.default_sclk / 100; + + return amdgpu_dpm_get_sclk(adev, false) / 100; } -- GitLab From 3c0b42809071cb6d779b9c4a573edd5d12654f43 Mon Sep 17 00:00:00 2001 From: Philip Yang Date: Mon, 27 Nov 2017 18:29:44 -0500 Subject: [PATCH 0441/1471] drm/amdkfd: Add crash protection in debugger register path After debugger is registered, the pqm_destroy_queue fails because is_debug is true, the queue should not be removed from process_queue_list since the count is not reduced. Test application calls debugger unregister without register debugger, add null pointer check protection to avoid crash for this case Signed-off-by: Philip Yang Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 5 +++++ 2 files changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index cc61ec2898802..62c3d9cd6ef18 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -526,7 +526,7 @@ static int kfd_ioctl_dbg_unregister(struct file *filep, long status; dev = kfd_device_by_id(args->gpu_id); - if (!dev) + if (!dev || !dev->dbgmgr) return -EINVAL; if (dev->device_info->asic_family == CHIP_CARRIZO) { diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index fbfa2741e2a78..991b3bf1e8bcd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -311,6 +311,10 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) if (pqn->q) { dqm = pqn->q->device->dqm; retval = dqm->ops.destroy_queue(dqm, &pdd->qpd, pqn->q); + if (retval) { + pr_debug("Destroy queue failed, returned %d\n", retval); + goto err_destroy_queue; + } uninit_queue(pqn->q); } @@ -322,6 +326,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) list_empty(&pdd->qpd.priv_queue_list)) dqm->ops.unregister_process(dqm, &pdd->qpd); +err_destroy_queue: return retval; } -- GitLab From 8f8fb9b9d05c44e452ebefac5f01f8f32ffe306c Mon Sep 17 00:00:00 2001 From: Kent Russell Date: Mon, 4 Dec 2017 06:50:17 -0500 Subject: [PATCH 0442/1471] drm/amdkfd: Fix printing pointer cast Just print a pointer instead of casting v2: Remove the 0x prefix, since %p prints that automatically, and remove it from one other spot as well Signed-off-by: Kent Russell Reviewed-by: Felix Kuehling Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c index feb76c235b1a6..ebb4da14e3df7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c @@ -116,8 +116,7 @@ int kfd_doorbell_init(struct kfd_dev *kfd) pr_debug("doorbell aperture size == 0x%08lX\n", kfd->shared_resources.doorbell_aperture_size); - pr_debug("doorbell kernel address == 0x%08lX\n", - (uintptr_t)kfd->doorbell_kernel_ptr); + pr_debug("doorbell kernel address == %p\n", kfd->doorbell_kernel_ptr); return 0; } @@ -194,8 +193,8 @@ u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, pr_debug("Get kernel queue doorbell\n" " doorbell offset == 0x%08X\n" - " kernel address == 0x%08lX\n", - *doorbell_off, (uintptr_t)(kfd->doorbell_kernel_ptr + inx)); + " kernel address == %p\n", + *doorbell_off, (kfd->doorbell_kernel_ptr + inx)); return kfd->doorbell_kernel_ptr + inx; } @@ -215,7 +214,7 @@ inline void write_kernel_doorbell(u32 __iomem *db, u32 value) { if (db) { writel(value, db); - pr_debug("Writing %d to doorbell address 0x%p\n", value, db); + pr_debug("Writing %d to doorbell address %p\n", value, db); } } -- GitLab From a99c6d4fdc0470c3137d834d578441d057a5ebd0 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:45 -0500 Subject: [PATCH 0443/1471] drm/amdkfd: map multiple processes to HW scheduler Allow HWS to to execute multiple processes on the hardware concurrently. The number of concurrent processes is limited by the number of VMIDs allocated to the HWS. A module parameter can be used for limiting this further or turn it off altogether (mainly for debugging purposes). Signed-off-by: Yong Zhao Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_device.c | 11 +++++++ drivers/gpu/drm/amd/amdkfd/kfd_module.c | 5 ++++ .../gpu/drm/amd/amdkfd/kfd_packet_manager.c | 30 +++++++++++++++++-- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 9 ++++++ 4 files changed, 53 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 4f05eacca786a..a8fa33a08de30 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -238,6 +238,17 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd, kfd->vm_info.vmid_num_kfd = kfd->vm_info.last_vmid_kfd - kfd->vm_info.first_vmid_kfd + 1; + /* Verify module parameters regarding mapped process number*/ + if ((hws_max_conc_proc < 0) + || (hws_max_conc_proc > kfd->vm_info.vmid_num_kfd)) { + dev_err(kfd_device, + "hws_max_conc_proc %d must be between 0 and %d, use %d instead\n", + hws_max_conc_proc, kfd->vm_info.vmid_num_kfd, + kfd->vm_info.vmid_num_kfd); + kfd->max_proc_per_quantum = kfd->vm_info.vmid_num_kfd; + } else + kfd->max_proc_per_quantum = hws_max_conc_proc; + /* calculate max size of mqds needed for queues */ size = max_num_of_queues_per_device * kfd->device_info->mqd_size_aligned; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index ee8adf654cd00..4e060c864c214 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -50,6 +50,11 @@ module_param(sched_policy, int, 0444); MODULE_PARM_DESC(sched_policy, "Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)"); +int hws_max_conc_proc = 8; +module_param(hws_max_conc_proc, int, 0444); +MODULE_PARM_DESC(hws_max_conc_proc, + "Max # processes HWS can execute concurrently when sched_policy=0 (0 = no concurrency, #VMIDs for KFD = Maximum(default))"); + int cwsr_enable = 1; module_param(cwsr_enable, int, 0444); MODULE_PARM_DESC(cwsr_enable, "CWSR enable (0 = Off, 1 = On (Default))"); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index 69c147a835918..0b7092e59bf94 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -57,13 +57,24 @@ static void pm_calc_rlib_size(struct packet_manager *pm, { unsigned int process_count, queue_count; unsigned int map_queue_size; + unsigned int max_proc_per_quantum = 1; + struct kfd_dev *dev = pm->dqm->dev; process_count = pm->dqm->processes_count; queue_count = pm->dqm->queue_count; - /* check if there is over subscription*/ + /* check if there is over subscription + * Note: the arbitration between the number of VMIDs and + * hws_max_conc_proc has been done in + * kgd2kfd_device_init(). + */ *over_subscription = false; - if ((process_count > 1) || queue_count > get_queues_num(pm->dqm)) { + + if (dev->max_proc_per_quantum > 1) + max_proc_per_quantum = dev->max_proc_per_quantum; + + if ((process_count > max_proc_per_quantum) || + queue_count > get_queues_num(pm->dqm)) { *over_subscription = true; pr_debug("Over subscribed runlist\n"); } @@ -116,10 +127,24 @@ static int pm_create_runlist(struct packet_manager *pm, uint32_t *buffer, uint64_t ib, size_t ib_size_in_dwords, bool chain) { struct pm4_mes_runlist *packet; + int concurrent_proc_cnt = 0; + struct kfd_dev *kfd = pm->dqm->dev; if (WARN_ON(!ib)) return -EFAULT; + /* Determine the number of processes to map together to HW: + * it can not exceed the number of VMIDs available to the + * scheduler, and it is determined by the smaller of the number + * of processes in the runlist and kfd module parameter + * hws_max_conc_proc. + * Note: the arbitration between the number of VMIDs and + * hws_max_conc_proc has been done in + * kgd2kfd_device_init(). + */ + concurrent_proc_cnt = min(pm->dqm->processes_count, + kfd->max_proc_per_quantum); + packet = (struct pm4_mes_runlist *)buffer; memset(buffer, 0, sizeof(struct pm4_mes_runlist)); @@ -130,6 +155,7 @@ static int pm_create_runlist(struct packet_manager *pm, uint32_t *buffer, packet->bitfields4.chain = chain ? 1 : 0; packet->bitfields4.offload_polling = 0; packet->bitfields4.valid = 1; + packet->bitfields4.process_cnt = concurrent_proc_cnt; packet->ordinal2 = lower_32_bits(ib); packet->bitfields3.ib_base_hi = upper_32_bits(ib); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index a668764679958..1edab2199f0b8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -88,6 +88,12 @@ extern int max_num_of_queues_per_device; /* Kernel module parameter to specify the scheduling policy */ extern int sched_policy; +/* + * Kernel module parameter to specify the maximum process + * number per HW scheduler + */ +extern int hws_max_conc_proc; + extern int cwsr_enable; /* @@ -214,6 +220,9 @@ struct kfd_dev { /* Debug manager */ struct kfd_dbgmgr *dbgmgr; + /* Maximum process number mapped to HW scheduler */ + unsigned int max_proc_per_quantum; + /* CWSR */ bool cwsr_enabled; const void *cwsr_isa; -- GitLab From 36582fa5165bca6cfab33bbe1d95f1598ab5fb7d Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:46 -0500 Subject: [PATCH 0444/1471] drm/amdkfd: Fix oversubscription accounting Don't count SDMA queues towards compute HQD oversubscription when deciding whether to create a chained runlist. Signed-off-by: Jay Cornwall Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index 0b7092e59bf94..c3230b96241e9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -55,13 +55,14 @@ static void pm_calc_rlib_size(struct packet_manager *pm, unsigned int *rlib_size, bool *over_subscription) { - unsigned int process_count, queue_count; + unsigned int process_count, queue_count, compute_queue_count; unsigned int map_queue_size; unsigned int max_proc_per_quantum = 1; struct kfd_dev *dev = pm->dqm->dev; process_count = pm->dqm->processes_count; queue_count = pm->dqm->queue_count; + compute_queue_count = queue_count - pm->dqm->sdma_queue_count; /* check if there is over subscription * Note: the arbitration between the number of VMIDs and @@ -74,7 +75,7 @@ static void pm_calc_rlib_size(struct packet_manager *pm, max_proc_per_quantum = dev->max_proc_per_quantum; if ((process_count > max_proc_per_quantum) || - queue_count > get_queues_num(pm->dqm)) { + compute_queue_count > get_queues_num(pm->dqm)) { *over_subscription = true; pr_debug("Over subscribed runlist\n"); } -- GitLab From fdcba29c5d9cfc67ba3ff6135d392ea91b6bd61a Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:47 -0500 Subject: [PATCH 0445/1471] drm/amdgpu: Fix definition of KFD_CIK_SDMA_QUEUE_OFFSET This counts the queue offset in register index, not register address. Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdgpu/cikd.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/cikd.h b/drivers/gpu/drm/amd/amdgpu/cikd.h index 6a9e38a3d2a0b..cee6e8a3ad9c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/cikd.h +++ b/drivers/gpu/drm/amd/amdgpu/cikd.h @@ -562,7 +562,7 @@ #define PRIVATE_BASE(x) ((x) << 0) /* scratch */ #define SHARED_BASE(x) ((x) << 16) /* LDS */ -#define KFD_CIK_SDMA_QUEUE_OFFSET 0x200 +#define KFD_CIK_SDMA_QUEUE_OFFSET (mmSDMA0_RLC1_RB_CNTL - mmSDMA0_RLC0_RB_CNTL) /* valid for both DEFAULT_MTYPE and APE1_MTYPE */ enum { -- GitLab From 80c195f5c2bd4da139f913b4debc5b75173f6fca Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:48 -0500 Subject: [PATCH 0446/1471] drm/amdgpu: Add kfd2kgd APIs for dumping HQDs This can be used by KFD for debugging features, such as dumping HQDs in debugfs. Signed-off-by: Felix Kuehling Acked-by: Oded Gabbay Signed-off-by: Oded Gabbay --- .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 71 ++++++++++++++++ .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 80 +++++++++++++++++++ .../gpu/drm/amd/include/kgd_kfd_interface.h | 14 ++++ 3 files changed, 165 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 14333af92ed9d..12feba8091df0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -105,8 +105,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm); +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); @@ -178,6 +184,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .init_interrupts = kgd_init_interrupts, .hqd_load = kgd_hqd_load, .hqd_sdma_load = kgd_hqd_sdma_load, + .hqd_dump = kgd_hqd_dump, + .hqd_sdma_dump = kgd_hqd_sdma_dump, .hqd_is_occupied = kgd_hqd_is_occupied, .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_hqd_destroy, @@ -376,6 +384,42 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, return 0; } +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t i = 0, reg; +#define HQD_N_REGS (35+4) +#define DUMP_REG(addr) do { \ + if (WARN_ON_ONCE(i >= HQD_N_REGS)) \ + break; \ + (*dump)[i][0] = (addr) << 2; \ + (*dump)[i++][1] = RREG32(addr); \ + } while (0) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + acquire_queue(kgd, pipe_id, queue_id); + + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE2); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE3); + + for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++) + DUMP_REG(reg); + + release_queue(kgd); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { @@ -440,6 +484,33 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET + + queue_id * KFD_CIK_SDMA_QUEUE_OFFSET; + uint32_t i = 0, reg; +#undef HQD_N_REGS +#define HQD_N_REGS (19+4) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++) + DUMP_REG(sdma_offset + reg); + for (reg = mmSDMA0_RLC0_VIRTUAL_ADDR; reg <= mmSDMA0_RLC0_WATERMARK; + reg++) + DUMP_REG(sdma_offset + reg); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 1d989e4915b11..b38049569264f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -64,8 +64,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, uint32_t wptr_shift, uint32_t wptr_mask, struct mm_struct *mm); +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm); +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); @@ -137,6 +143,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .init_interrupts = kgd_init_interrupts, .hqd_load = kgd_hqd_load, .hqd_sdma_load = kgd_hqd_sdma_load, + .hqd_dump = kgd_hqd_dump, + .hqd_sdma_dump = kgd_hqd_sdma_dump, .hqd_is_occupied = kgd_hqd_is_occupied, .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, .hqd_destroy = kgd_hqd_destroy, @@ -365,6 +373,42 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, return 0; } +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t i = 0, reg; +#define HQD_N_REGS (54+4) +#define DUMP_REG(addr) do { \ + if (WARN_ON_ONCE(i >= HQD_N_REGS)) \ + break; \ + (*dump)[i][0] = (addr) << 2; \ + (*dump)[i++][1] = RREG32(addr); \ + } while (0) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + acquire_queue(kgd, pipe_id, queue_id); + + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE2); + DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE3); + + for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_DONES; reg++) + DUMP_REG(reg); + + release_queue(kgd); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm) { @@ -428,6 +472,42 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, return 0; } +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET + + queue_id * KFD_VI_SDMA_QUEUE_OFFSET; + uint32_t i = 0, reg; +#undef HQD_N_REGS +#define HQD_N_REGS (19+4+2+3+7) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++) + DUMP_REG(sdma_offset + reg); + for (reg = mmSDMA0_RLC0_VIRTUAL_ADDR; reg <= mmSDMA0_RLC0_WATERMARK; + reg++) + DUMP_REG(sdma_offset + reg); + for (reg = mmSDMA0_RLC0_CSA_ADDR_LO; reg <= mmSDMA0_RLC0_CSA_ADDR_HI; + reg++) + DUMP_REG(sdma_offset + reg); + for (reg = mmSDMA0_RLC0_IB_SUB_REMAIN; reg <= mmSDMA0_RLC0_DUMMY_REG; + reg++) + DUMP_REG(sdma_offset + reg); + for (reg = mmSDMA0_RLC0_MIDCMD_DATA0; reg <= mmSDMA0_RLC0_MIDCMD_CNTL; + reg++) + DUMP_REG(sdma_offset + reg); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id) { diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index c6d4e64531c71..fe3079a33a3d4 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -131,6 +131,12 @@ struct tile_config { * @hqd_sdma_load: Loads the SDMA mqd structure to a H/W SDMA hqd slot. * used only for no HWS mode. * + * @hqd_dump: Dumps CPC HQD registers to an array of address-value pairs. + * Array is allocated with kmalloc, needs to be freed with kfree by caller. + * + * @hqd_sdma_dump: Dumps SDMA HQD registers to an array of address-value pairs. + * Array is allocated with kmalloc, needs to be freed with kfree by caller. + * * @hqd_is_occupies: Checks if a hqd slot is occupied. * * @hqd_destroy: Destructs and preempts the queue assigned to that hqd slot. @@ -187,6 +193,14 @@ struct kfd2kgd_calls { int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd, uint32_t __user *wptr, struct mm_struct *mm); + int (*hqd_dump)(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); + + int (*hqd_sdma_dump)(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); + bool (*hqd_is_occupied)(struct kgd_dev *kgd, uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); -- GitLab From 851a645efd0fee7f27819ab30f3b353bace3cd98 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:49 -0500 Subject: [PATCH 0447/1471] drm/amdkfd: Add debugfs support to KFD This commit adds several debugfs entries for kfd: kfd/hqds: dumps all HQDs on all GPUs for KFD-controlled compute and SDMA RLC queues kfd/mqds: dumps all MQDs of all KFD processes on all GPUs kfd/rls: dumps HWS runlists on all GPUs Signed-off-by: Yong Zhao Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/Makefile | 2 + drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c | 75 +++++++++++++++++++ .../drm/amd/amdkfd/kfd_device_queue_manager.c | 71 ++++++++++++++++++ drivers/gpu/drm/amd/amdkfd/kfd_module.c | 3 + drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h | 4 + .../gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c | 27 +++++++ .../gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c | 25 +++++++ .../gpu/drm/amd/amdkfd/kfd_packet_manager.c | 24 ++++++ drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 21 ++++++ drivers/gpu/drm/amd/amdkfd/kfd_process.c | 29 +++++++ .../amd/amdkfd/kfd_process_queue_manager.c | 63 ++++++++++++++++ drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 55 ++++++++++++++ 12 files changed, 399 insertions(+) create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 342c2d937b17b..67e2c42a5e653 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -37,4 +37,6 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_interrupt.o kfd_events.o cik_event_interrupt.o \ kfd_dbgdev.o kfd_dbgmgr.o +amdkfd-$(CONFIG_DEBUG_FS) += kfd_debugfs.o + obj-$(CONFIG_HSA_AMD) += amdkfd.o diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c b/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c new file mode 100644 index 0000000000000..4bd6ebfaf425b --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_debugfs.c @@ -0,0 +1,75 @@ +/* + * Copyright 2016-2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include "kfd_priv.h" + +static struct dentry *debugfs_root; + +static int kfd_debugfs_open(struct inode *inode, struct file *file) +{ + int (*show)(struct seq_file *, void *) = inode->i_private; + + return single_open(file, show, NULL); +} + +static const struct file_operations kfd_debugfs_fops = { + .owner = THIS_MODULE, + .open = kfd_debugfs_open, + .read = seq_read, + .llseek = seq_lseek, + .release = single_release, +}; + +void kfd_debugfs_init(void) +{ + struct dentry *ent; + + debugfs_root = debugfs_create_dir("kfd", NULL); + if (!debugfs_root || debugfs_root == ERR_PTR(-ENODEV)) { + pr_warn("Failed to create kfd debugfs dir\n"); + return; + } + + ent = debugfs_create_file("mqds", S_IFREG | 0444, debugfs_root, + kfd_debugfs_mqds_by_process, + &kfd_debugfs_fops); + if (!ent) + pr_warn("Failed to create mqds in kfd debugfs\n"); + + ent = debugfs_create_file("hqds", S_IFREG | 0444, debugfs_root, + kfd_debugfs_hqds_by_device, + &kfd_debugfs_fops); + if (!ent) + pr_warn("Failed to create hqds in kfd debugfs\n"); + + ent = debugfs_create_file("rls", S_IFREG | 0444, debugfs_root, + kfd_debugfs_rls_by_device, + &kfd_debugfs_fops); + if (!ent) + pr_warn("Failed to create rls in kfd debugfs\n"); +} + +void kfd_debugfs_fini(void) +{ + debugfs_remove_recursive(debugfs_root); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 81ec7bb89b04c..d0693fd8cbf87 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -1311,3 +1311,74 @@ void device_queue_manager_uninit(struct device_queue_manager *dqm) dqm->ops.uninitialize(dqm); kfree(dqm); } + +#if defined(CONFIG_DEBUG_FS) + +static void seq_reg_dump(struct seq_file *m, + uint32_t (*dump)[2], uint32_t n_regs) +{ + uint32_t i, count; + + for (i = 0, count = 0; i < n_regs; i++) { + if (count == 0 || + dump[i-1][0] + sizeof(uint32_t) != dump[i][0]) { + seq_printf(m, "%s %08x: %08x", + i ? "\n" : "", + dump[i][0], dump[i][1]); + count = 7; + } else { + seq_printf(m, " %08x", dump[i][1]); + count--; + } + } + + seq_puts(m, "\n"); +} + +int dqm_debugfs_hqds(struct seq_file *m, void *data) +{ + struct device_queue_manager *dqm = data; + uint32_t (*dump)[2], n_regs; + int pipe, queue; + int r = 0; + + for (pipe = 0; pipe < get_pipes_per_mec(dqm); pipe++) { + int pipe_offset = pipe * get_queues_per_pipe(dqm); + + for (queue = 0; queue < get_queues_per_pipe(dqm); queue++) { + if (!test_bit(pipe_offset + queue, + dqm->dev->shared_resources.queue_bitmap)) + continue; + + r = dqm->dev->kfd2kgd->hqd_dump( + dqm->dev->kgd, pipe, queue, &dump, &n_regs); + if (r) + break; + + seq_printf(m, " CP Pipe %d, Queue %d\n", + pipe, queue); + seq_reg_dump(m, dump, n_regs); + + kfree(dump); + } + } + + for (pipe = 0; pipe < CIK_SDMA_ENGINE_NUM; pipe++) { + for (queue = 0; queue < CIK_SDMA_QUEUES_PER_ENGINE; queue++) { + r = dqm->dev->kfd2kgd->hqd_sdma_dump( + dqm->dev->kgd, pipe, queue, &dump, &n_regs); + if (r) + break; + + seq_printf(m, " SDMA Engine %d, RLC %d\n", + pipe, queue); + seq_reg_dump(m, dump, n_regs); + + kfree(dump); + } + } + + return r; +} + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index 4e060c864c214..f50e4949a2ed2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -123,6 +123,8 @@ static int __init kfd_module_init(void) kfd_process_create_wq(); + kfd_debugfs_init(); + amdkfd_init_completed = 1; dev_info(kfd_device, "Initialized module\n"); @@ -139,6 +141,7 @@ static void __exit kfd_module_exit(void) { amdkfd_init_completed = 0; + kfd_debugfs_fini(); kfd_process_destroy_wq(); kfd_topology_shutdown(); kfd_chardev_exit(); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h index 1f3a6ba7eed21..8972bcfbf701c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h @@ -85,6 +85,10 @@ struct mqd_manager { uint64_t queue_address, uint32_t pipe_id, uint32_t queue_id); +#if defined(CONFIG_DEBUG_FS) + int (*debugfs_show_mqd)(struct seq_file *m, void *data); +#endif + struct mutex mqd_mutex; struct kfd_dev *dev; }; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index 7aa57abebbcd0..f8ef4a051e082 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -365,6 +365,24 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, return 0; } +#if defined(CONFIG_DEBUG_FS) + +static int debugfs_show_mqd(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct cik_mqd), false); + return 0; +} + +static int debugfs_show_mqd_sdma(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct cik_sdma_rlc_registers), false); + return 0; +} + +#endif + struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, struct kfd_dev *dev) @@ -389,6 +407,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd; mqd->destroy_mqd = destroy_mqd; mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif break; case KFD_MQD_TYPE_HIQ: mqd->init_mqd = init_mqd_hiq; @@ -397,6 +418,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd_hiq; mqd->destroy_mqd = destroy_mqd; mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif break; case KFD_MQD_TYPE_SDMA: mqd->init_mqd = init_mqd_sdma; @@ -405,6 +429,9 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd_sdma; mqd->destroy_mqd = destroy_mqd_sdma; mqd->is_occupied = is_occupied_sdma; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd_sdma; +#endif break; default: kfree(mqd); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index 00e1f1a9728b2..971aec0637dc7 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -358,7 +358,23 @@ static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); } +#if defined(CONFIG_DEBUG_FS) +static int debugfs_show_mqd(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct vi_mqd), false); + return 0; +} + +static int debugfs_show_mqd_sdma(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct vi_sdma_mqd), false); + return 0; +} + +#endif struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, struct kfd_dev *dev) @@ -383,6 +399,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd; mqd->destroy_mqd = destroy_mqd; mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif break; case KFD_MQD_TYPE_HIQ: mqd->init_mqd = init_mqd_hiq; @@ -391,6 +410,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd_hiq; mqd->destroy_mqd = destroy_mqd; mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif break; case KFD_MQD_TYPE_SDMA: mqd->init_mqd = init_mqd_sdma; @@ -399,6 +421,9 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, mqd->update_mqd = update_mqd_sdma; mqd->destroy_mqd = destroy_mqd_sdma; mqd->is_occupied = is_occupied_sdma; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd_sdma; +#endif break; default: kfree(mqd); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index c3230b96241e9..0ecbd1f9b606b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -278,6 +278,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, return retval; *rl_size_bytes = alloc_size_bytes; + pm->ib_size_bytes = alloc_size_bytes; pr_debug("Building runlist ib process count: %d queues count %d\n", pm->dqm->processes_count, pm->dqm->queue_count); @@ -591,3 +592,26 @@ void pm_release_ib(struct packet_manager *pm) } mutex_unlock(&pm->lock); } + +#if defined(CONFIG_DEBUG_FS) + +int pm_debugfs_runlist(struct seq_file *m, void *data) +{ + struct packet_manager *pm = data; + + mutex_lock(&pm->lock); + + if (!pm->allocated) { + seq_puts(m, " No active runlist\n"); + goto out; + } + + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + pm->ib_buffer_obj->cpu_ptr, pm->ib_size_bytes, false); + +out: + mutex_unlock(&pm->lock); + return 0; +} + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 1edab2199f0b8..dca493b4e17d5 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -33,6 +33,7 @@ #include #include #include +#include #include #include "amd_shared.h" @@ -735,6 +736,7 @@ struct packet_manager { struct mutex lock; bool allocated; struct kfd_mem_obj *ib_buffer_obj; + unsigned int ib_size_bytes; }; int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm); @@ -781,4 +783,23 @@ int kfd_event_destroy(struct kfd_process *p, uint32_t event_id); int dbgdev_wave_reset_wavefronts(struct kfd_dev *dev, struct kfd_process *p); +/* Debugfs */ +#if defined(CONFIG_DEBUG_FS) + +void kfd_debugfs_init(void); +void kfd_debugfs_fini(void); +int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data); +int pqm_debugfs_mqds(struct seq_file *m, void *data); +int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data); +int dqm_debugfs_hqds(struct seq_file *m, void *data); +int kfd_debugfs_rls_by_device(struct seq_file *m, void *data); +int pm_debugfs_runlist(struct seq_file *m, void *data); + +#else + +static inline void kfd_debugfs_init(void) {} +static inline void kfd_debugfs_fini(void) {} + +#endif + #endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 39f4c19aaf617..99c18ee65a764 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -620,3 +620,32 @@ int kfd_reserved_mem_mmap(struct kfd_process *process, PFN_DOWN(__pa(qpd->cwsr_kaddr)), KFD_CWSR_TBA_TMA_SIZE, vma->vm_page_prot); } + +#if defined(CONFIG_DEBUG_FS) + +int kfd_debugfs_mqds_by_process(struct seq_file *m, void *data) +{ + struct kfd_process *p; + unsigned int temp; + int r = 0; + + int idx = srcu_read_lock(&kfd_processes_srcu); + + hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { + seq_printf(m, "Process %d PASID %d:\n", + p->lead_thread->tgid, p->pasid); + + mutex_lock(&p->mutex); + r = pqm_debugfs_mqds(m, &p->pqm); + mutex_unlock(&p->mutex); + + if (r) + break; + } + + srcu_read_unlock(&kfd_processes_srcu, idx); + + return r; +} + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 991b3bf1e8bcd..8763806326682 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -368,4 +368,67 @@ struct kernel_queue *pqm_get_kernel_queue( return NULL; } +#if defined(CONFIG_DEBUG_FS) +int pqm_debugfs_mqds(struct seq_file *m, void *data) +{ + struct process_queue_manager *pqm = data; + struct process_queue_node *pqn; + struct queue *q; + enum KFD_MQD_TYPE mqd_type; + struct mqd_manager *mqd_manager; + int r = 0; + + list_for_each_entry(pqn, &pqm->queues, process_queue_list) { + if (pqn->q) { + q = pqn->q; + switch (q->properties.type) { + case KFD_QUEUE_TYPE_SDMA: + seq_printf(m, " SDMA queue on device %x\n", + q->device->id); + mqd_type = KFD_MQD_TYPE_SDMA; + break; + case KFD_QUEUE_TYPE_COMPUTE: + seq_printf(m, " Compute queue on device %x\n", + q->device->id); + mqd_type = KFD_MQD_TYPE_CP; + break; + default: + seq_printf(m, + " Bad user queue type %d on device %x\n", + q->properties.type, q->device->id); + continue; + } + mqd_manager = q->device->dqm->ops.get_mqd_manager( + q->device->dqm, mqd_type); + } else if (pqn->kq) { + q = pqn->kq->queue; + mqd_manager = pqn->kq->mqd; + switch (q->properties.type) { + case KFD_QUEUE_TYPE_DIQ: + seq_printf(m, " DIQ on device %x\n", + pqn->kq->dev->id); + mqd_type = KFD_MQD_TYPE_HIQ; + break; + default: + seq_printf(m, + " Bad kernel queue type %d on device %x\n", + q->properties.type, + pqn->kq->dev->id); + continue; + } + } else { + seq_printf(m, + " Weird: Queue node with neither kernel nor user queue\n"); + continue; + } + + r = mqd_manager->debugfs_show_mqd(m, q->mqd); + if (r != 0) + break; + } + + return r; +} + +#endif diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index 19ce59028d6bd..9d03a56fd9c75 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -32,6 +32,7 @@ #include "kfd_priv.h" #include "kfd_crat.h" #include "kfd_topology.h" +#include "kfd_device_queue_manager.h" static struct list_head topology_device_list; static int topology_crat_parsed; @@ -1233,3 +1234,57 @@ struct kfd_dev *kfd_topology_enum_kfd_devices(uint8_t idx) return device; } + +#if defined(CONFIG_DEBUG_FS) + +int kfd_debugfs_hqds_by_device(struct seq_file *m, void *data) +{ + struct kfd_topology_device *dev; + unsigned int i = 0; + int r = 0; + + down_read(&topology_lock); + + list_for_each_entry(dev, &topology_device_list, list) { + if (!dev->gpu) { + i++; + continue; + } + + seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id); + r = dqm_debugfs_hqds(m, dev->gpu->dqm); + if (r) + break; + } + + up_read(&topology_lock); + + return r; +} + +int kfd_debugfs_rls_by_device(struct seq_file *m, void *data) +{ + struct kfd_topology_device *dev; + unsigned int i = 0; + int r = 0; + + down_read(&topology_lock); + + list_for_each_entry(dev, &topology_device_list, list) { + if (!dev->gpu) { + i++; + continue; + } + + seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id); + r = pm_debugfs_runlist(m, &dev->gpu->dqm->packets); + if (r) + break; + } + + up_read(&topology_lock); + + return r; +} + +#endif -- GitLab From c7b1243eef128c49aabdfab71cd4f5ecb6602d97 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:50 -0500 Subject: [PATCH 0448/1471] drm/amdkfd: Get reference to lead_thread task struct Increment the kfd_process.lead_thread's reference counter to make it safe to dereference. This is needed for getting a safe reference to the process' mm_struct. Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 99c18ee65a764..660d8bcb93e05 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -24,6 +24,7 @@ #include #include #include +#include #include #include #include @@ -191,6 +192,8 @@ static void kfd_process_wq_release(struct work_struct *work) mutex_destroy(&p->mutex); + put_task_struct(p->lead_thread); + kfree(p); kfree(work); @@ -342,6 +345,7 @@ static struct kfd_process *create_process(const struct task_struct *thread) (uintptr_t)process->mm); process->lead_thread = thread->group_leader; + get_task_struct(process->lead_thread); INIT_LIST_HEAD(&process->per_device_data); -- GitLab From 5ce10687ae5ba2aa06aeaaf3fad9ce64dff45a85 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:51 -0500 Subject: [PATCH 0449/1471] drm/amdkfd: Make kfd_process reference counted This will be used to elliminate the use of the process lock for preventing concurrent process destruction. This will simplify lock dependencies between KFD and KGD. This also simplifies the process destruction in a few ways: * Don't allocate work struct dynamically * Remove unnecessary hack that increments mm reference counter * Remove unnecessary process locking during destruction Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 4 ++ drivers/gpu/drm/amd/amdkfd/kfd_process.c | 58 +++++++++--------------- 2 files changed, 26 insertions(+), 36 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index dca493b4e17d5..248e4f5108247 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -34,6 +34,7 @@ #include #include #include +#include #include #include "amd_shared.h" @@ -537,6 +538,9 @@ struct kfd_process { */ void *mm; + struct kref ref; + struct work_struct release_work; + struct mutex mutex; /* diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 660d8bcb93e05..e02e8a2e7b289 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -48,11 +48,6 @@ DEFINE_STATIC_SRCU(kfd_processes_srcu); static struct workqueue_struct *kfd_process_wq; -struct kfd_process_release_work { - struct work_struct kfd_work; - struct kfd_process *p; -}; - static struct kfd_process *find_process(const struct task_struct *thread); static struct kfd_process *create_process(const struct task_struct *thread); static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep); @@ -151,21 +146,20 @@ static struct kfd_process *find_process(const struct task_struct *thread) return p; } +/* No process locking is needed in this function, because the process + * is not findable any more. We must assume that no other thread is + * using it any more, otherwise we couldn't safely free the process + * structure in the end. + */ static void kfd_process_wq_release(struct work_struct *work) { - struct kfd_process_release_work *my_work; + struct kfd_process *p = container_of(work, struct kfd_process, + release_work); struct kfd_process_device *pdd, *temp; - struct kfd_process *p; - - my_work = (struct kfd_process_release_work *) work; - - p = my_work->p; pr_debug("Releasing process (pasid %d) in workqueue\n", p->pasid); - mutex_lock(&p->mutex); - list_for_each_entry_safe(pdd, temp, &p->per_device_data, per_device_list) { pr_debug("Releasing pdd (topology id %d) for process (pasid %d) in workqueue\n", @@ -188,33 +182,26 @@ static void kfd_process_wq_release(struct work_struct *work) kfd_pasid_free(p->pasid); kfd_free_process_doorbells(p); - mutex_unlock(&p->mutex); - mutex_destroy(&p->mutex); put_task_struct(p->lead_thread); kfree(p); - - kfree(work); } -static void kfd_process_destroy_delayed(struct rcu_head *rcu) +static void kfd_process_ref_release(struct kref *ref) { - struct kfd_process_release_work *work; - struct kfd_process *p; - - p = container_of(rcu, struct kfd_process, rcu); + struct kfd_process *p = container_of(ref, struct kfd_process, ref); - mmdrop(p->mm); + INIT_WORK(&p->release_work, kfd_process_wq_release); + queue_work(kfd_process_wq, &p->release_work); +} - work = kmalloc(sizeof(struct kfd_process_release_work), GFP_ATOMIC); +static void kfd_process_destroy_delayed(struct rcu_head *rcu) +{ + struct kfd_process *p = container_of(rcu, struct kfd_process, rcu); - if (work) { - INIT_WORK((struct work_struct *) work, kfd_process_wq_release); - work->p = p; - queue_work(kfd_process_wq, (struct work_struct *) work); - } + kref_put(&p->ref, kfd_process_ref_release); } static void kfd_process_notifier_release(struct mmu_notifier *mn, @@ -258,15 +245,12 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn, kfd_process_dequeue_from_all_devices(p); pqm_uninit(&p->pqm); + /* Indicate to other users that MM is no longer valid */ + p->mm = NULL; + mutex_unlock(&p->mutex); - /* - * Because we drop mm_count inside kfd_process_destroy_delayed - * and because the mmu_notifier_unregister function also drop - * mm_count we need to take an extra count here. - */ - mmgrab(p->mm); - mmu_notifier_unregister_no_release(&p->mmu_notifier, p->mm); + mmu_notifier_unregister_no_release(&p->mmu_notifier, mm); mmu_notifier_call_srcu(&p->rcu, &kfd_process_destroy_delayed); } @@ -331,6 +315,8 @@ static struct kfd_process *create_process(const struct task_struct *thread) if (kfd_alloc_process_doorbells(process) < 0) goto err_alloc_doorbells; + kref_init(&process->ref); + mutex_init(&process->mutex); process->mm = thread->mm; -- GitLab From abb208a8d490bac9d765c45f20cf306757733778 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:52 -0500 Subject: [PATCH 0450/1471] drm/amdkfd: Use ref count to prevent kfd_process destruction MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use a reference counter instead of a lock to prevent process destruction while functions running out of process context are using the kfd_process structure. In many cases these functions don't need the structure to be locked. In the few cases that really do need the process lock, take it explicitly. This helps simplify lock dependencies between the process lock and other locks, particularly amdgpu and mm_struct locks. This will be important when amdgpu calls back to amdkfd for memory evictions. Signed-off-by: Felix Kuehling Acked-by: Christian König Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_events.c | 14 +++++++------- drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 1 + drivers/gpu/drm/amd/amdkfd/kfd_process.c | 16 +++++++++++++--- 3 files changed, 21 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c index cb92d4b724009..93aae5c1e78bf 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -441,7 +441,7 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id, /* * Because we are called from arbitrary context (workqueue) as opposed * to process context, kfd_process could attempt to exit while we are - * running so the lookup function returns a locked process. + * running so the lookup function increments the process ref count. */ struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); @@ -493,7 +493,7 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id, } mutex_unlock(&p->event_mutex); - mutex_unlock(&p->mutex); + kfd_unref_process(p); } static struct kfd_event_waiter *alloc_event_waiters(uint32_t num_events) @@ -847,7 +847,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, /* * Because we are called from arbitrary context (workqueue) as opposed * to process context, kfd_process could attempt to exit while we are - * running so the lookup function returns a locked process. + * running so the lookup function increments the process ref count. */ struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); struct mm_struct *mm; @@ -860,7 +860,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, */ mm = get_task_mm(p->lead_thread); if (!mm) { - mutex_unlock(&p->mutex); + kfd_unref_process(p); return; /* Process is exiting */ } @@ -903,7 +903,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, &memory_exception_data); mutex_unlock(&p->event_mutex); - mutex_unlock(&p->mutex); + kfd_unref_process(p); } void kfd_signal_hw_exception_event(unsigned int pasid) @@ -911,7 +911,7 @@ void kfd_signal_hw_exception_event(unsigned int pasid) /* * Because we are called from arbitrary context (workqueue) as opposed * to process context, kfd_process could attempt to exit while we are - * running so the lookup function returns a locked process. + * running so the lookup function increments the process ref count. */ struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); @@ -924,5 +924,5 @@ void kfd_signal_hw_exception_event(unsigned int pasid) lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_HW_EXCEPTION, NULL); mutex_unlock(&p->event_mutex); - mutex_unlock(&p->mutex); + kfd_unref_process(p); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 248e4f5108247..0c96a6b3a173a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -606,6 +606,7 @@ void kfd_process_destroy_wq(void); struct kfd_process *kfd_create_process(struct file *filep); struct kfd_process *kfd_get_process(const struct task_struct *); struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid); +void kfd_unref_process(struct kfd_process *p); struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, struct kfd_process *p); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index e02e8a2e7b289..509f987f4994c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -49,6 +49,7 @@ DEFINE_STATIC_SRCU(kfd_processes_srcu); static struct workqueue_struct *kfd_process_wq; static struct kfd_process *find_process(const struct task_struct *thread); +static void kfd_process_ref_release(struct kref *ref); static struct kfd_process *create_process(const struct task_struct *thread); static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep); @@ -146,6 +147,11 @@ static struct kfd_process *find_process(const struct task_struct *thread) return p; } +void kfd_unref_process(struct kfd_process *p) +{ + kref_put(&p->ref, kfd_process_ref_release); +} + /* No process locking is needed in this function, because the process * is not findable any more. We must assume that no other thread is * using it any more, otherwise we couldn't safely free the process @@ -201,7 +207,7 @@ static void kfd_process_destroy_delayed(struct rcu_head *rcu) { struct kfd_process *p = container_of(rcu, struct kfd_process, rcu); - kref_put(&p->ref, kfd_process_ref_release); + kfd_unref_process(p); } static void kfd_process_notifier_release(struct mmu_notifier *mn, @@ -525,6 +531,8 @@ void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned int pasid) mutex_unlock(kfd_get_dbgmgr_mutex()); + mutex_lock(&p->mutex); + pdd = kfd_get_process_device_data(dev, p); if (pdd) /* For GPU relying on IOMMU, we need to dequeue here @@ -533,6 +541,8 @@ void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned int pasid) kfd_process_dequeue_from_device(pdd); mutex_unlock(&p->mutex); + + kfd_unref_process(p); } struct kfd_process_device *kfd_get_first_process_device_data( @@ -557,7 +567,7 @@ bool kfd_has_process_device_data(struct kfd_process *p) return !(list_empty(&p->per_device_data)); } -/* This returns with process->mutex locked. */ +/* This increments the process->ref counter. */ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) { struct kfd_process *p; @@ -567,7 +577,7 @@ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { if (p->pasid == pasid) { - mutex_lock(&p->mutex); + kref_get(&p->ref); break; } } -- GitLab From 82c16b4280bc42bc035af566e1480d88666126b1 Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Mon, 27 Nov 2017 18:29:53 -0500 Subject: [PATCH 0451/1471] drm/amdkfd: Return NULL if kfd_lookup_process_by_pasid fails If no matching process is found, return NULL instead of a pointer to the last process in the kfd_processes_table. Signed-off-by: Yong Zhao Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 509f987f4994c..93f90191509fe 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -570,7 +570,7 @@ bool kfd_has_process_device_data(struct kfd_process *p) /* This increments the process->ref counter. */ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) { - struct kfd_process *p; + struct kfd_process *p, *ret_p = NULL; unsigned int temp; int idx = srcu_read_lock(&kfd_processes_srcu); @@ -578,13 +578,14 @@ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { if (p->pasid == pasid) { kref_get(&p->ref); + ret_p = p; break; } } srcu_read_unlock(&kfd_processes_srcu, idx); - return p; + return ret_p; } int kfd_reserved_mem_mmap(struct kfd_process *process, -- GitLab From 2d9b36f98343624282f81c1886e95bc552d1a9b7 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:54 -0500 Subject: [PATCH 0452/1471] drm/amdkfd: Reduce nesting in kfd_create_process_device_data Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 93f90191509fe..88fc822c9e5a4 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -390,17 +390,18 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, struct kfd_process_device *pdd = NULL; pdd = kzalloc(sizeof(*pdd), GFP_KERNEL); - if (pdd != NULL) { - pdd->dev = dev; - INIT_LIST_HEAD(&pdd->qpd.queues_list); - INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); - pdd->qpd.dqm = dev->dqm; - pdd->qpd.pqm = &p->pqm; - pdd->process = p; - pdd->bound = PDD_UNBOUND; - pdd->already_dequeued = false; - list_add(&pdd->per_device_list, &p->per_device_data); - } + if (!pdd) + return NULL; + + pdd->dev = dev; + INIT_LIST_HEAD(&pdd->qpd.queues_list); + INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); + pdd->qpd.dqm = dev->dqm; + pdd->qpd.pqm = &p->pqm; + pdd->process = p; + pdd->bound = PDD_UNBOUND; + pdd->already_dequeued = false; + list_add(&pdd->per_device_list, &p->per_device_data); return pdd; } -- GitLab From de1450a5597a642bdc4987aa31873d8475810aae Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Mon, 27 Nov 2017 18:29:55 -0500 Subject: [PATCH 0453/1471] drm/amdkfd: Factor PDD destruction out of kfd_process_wq_release Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 40 ++++++++++++++---------- 1 file changed, 24 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 88fc822c9e5a4..096710c3766d3 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -152,28 +152,15 @@ void kfd_unref_process(struct kfd_process *p) kref_put(&p->ref, kfd_process_ref_release); } -/* No process locking is needed in this function, because the process - * is not findable any more. We must assume that no other thread is - * using it any more, otherwise we couldn't safely free the process - * structure in the end. - */ -static void kfd_process_wq_release(struct work_struct *work) +static void kfd_process_destroy_pdds(struct kfd_process *p) { - struct kfd_process *p = container_of(work, struct kfd_process, - release_work); struct kfd_process_device *pdd, *temp; - pr_debug("Releasing process (pasid %d) in workqueue\n", - p->pasid); - list_for_each_entry_safe(pdd, temp, &p->per_device_data, - per_device_list) { - pr_debug("Releasing pdd (topology id %d) for process (pasid %d) in workqueue\n", + per_device_list) { + pr_debug("Releasing pdd (topology id %d) for process (pasid %d)\n", pdd->dev->id, p->pasid); - if (pdd->bound == PDD_BOUND) - amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid); - list_del(&pdd->per_device_list); if (pdd->qpd.cwsr_kaddr) @@ -182,6 +169,27 @@ static void kfd_process_wq_release(struct work_struct *work) kfree(pdd); } +} + +/* No process locking is needed in this function, because the process + * is not findable any more. We must assume that no other thread is + * using it any more, otherwise we couldn't safely free the process + * structure in the end. + */ +static void kfd_process_wq_release(struct work_struct *work) +{ + struct kfd_process *p = container_of(work, struct kfd_process, + release_work); + struct kfd_process_device *pdd; + + pr_debug("Releasing process (pasid %d) in workqueue\n", p->pasid); + + list_for_each_entry(pdd, &p->per_device_data, per_device_list) { + if (pdd->bound == PDD_BOUND) + amd_iommu_unbind_pasid(pdd->dev->pdev, p->pasid); + } + + kfd_process_destroy_pdds(p); kfd_event_free_process(p); -- GitLab From c0ede1f8dc3a22464b126f4d54ad7999ebb789a0 Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Mon, 27 Nov 2017 18:29:56 -0500 Subject: [PATCH 0454/1471] drm/amdkfd: Simplify locking during process creation Also fixes error handling if kfd_process_init_cwsr fails. Signed-off-by: Yong Zhao Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 46 +++++++++++------------- 1 file changed, 21 insertions(+), 25 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 096710c3766d3..a22fb0710f153 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -50,7 +50,8 @@ static struct workqueue_struct *kfd_process_wq; static struct kfd_process *find_process(const struct task_struct *thread); static void kfd_process_ref_release(struct kref *ref); -static struct kfd_process *create_process(const struct task_struct *thread); +static struct kfd_process *create_process(const struct task_struct *thread, + struct file *filep); static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep); @@ -80,9 +81,6 @@ struct kfd_process *kfd_create_process(struct file *filep) if (thread->group_leader->mm != thread->mm) return ERR_PTR(-EINVAL); - /* Take mmap_sem because we call __mmu_notifier_register inside */ - down_write(&thread->mm->mmap_sem); - /* * take kfd processes mutex before starting of process creation * so there won't be a case where two threads of the same process @@ -94,16 +92,11 @@ struct kfd_process *kfd_create_process(struct file *filep) process = find_process(thread); if (process) pr_debug("Process already found\n"); - - if (!process) - process = create_process(thread); + else + process = create_process(thread, filep); mutex_unlock(&kfd_processes_mutex); - up_write(&thread->mm->mmap_sem); - - kfd_process_init_cwsr(process, filep); - return process; } @@ -274,15 +267,12 @@ static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = { static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep) { - int err = 0; unsigned long offset; - struct kfd_process_device *temp, *pdd = NULL; + struct kfd_process_device *pdd = NULL; struct kfd_dev *dev = NULL; struct qcm_process_device *qpd = NULL; - mutex_lock(&p->mutex); - list_for_each_entry_safe(pdd, temp, &p->per_device_data, - per_device_list) { + list_for_each_entry(pdd, &p->per_device_data, per_device_list) { dev = pdd->dev; qpd = &pdd->qpd; if (!dev->cwsr_enabled || qpd->cwsr_kaddr) @@ -293,12 +283,12 @@ static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep) MAP_SHARED, offset); if (IS_ERR_VALUE(qpd->tba_addr)) { - pr_err("Failure to set tba address. error -%d.\n", - (int)qpd->tba_addr); - err = qpd->tba_addr; + int err = qpd->tba_addr; + + pr_err("Failure to set tba address. error %d.\n", err); qpd->tba_addr = 0; qpd->cwsr_kaddr = NULL; - goto out; + return err; } memcpy(qpd->cwsr_kaddr, dev->cwsr_isa, dev->cwsr_isa_size); @@ -307,12 +297,12 @@ static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep) pr_debug("set tba :0x%llx, tma:0x%llx, cwsr_kaddr:%p for pqm.\n", qpd->tba_addr, qpd->tma_addr, qpd->cwsr_kaddr); } -out: - mutex_unlock(&p->mutex); - return err; + + return 0; } -static struct kfd_process *create_process(const struct task_struct *thread) +static struct kfd_process *create_process(const struct task_struct *thread, + struct file *filep) { struct kfd_process *process; int err = -ENOMEM; @@ -337,7 +327,7 @@ static struct kfd_process *create_process(const struct task_struct *thread) /* register notifier */ process->mmu_notifier.ops = &kfd_process_mmu_notifier_ops; - err = __mmu_notifier_register(&process->mmu_notifier, process->mm); + err = mmu_notifier_register(&process->mmu_notifier, process->mm); if (err) goto err_mmu_notifier; @@ -361,8 +351,14 @@ static struct kfd_process *create_process(const struct task_struct *thread) if (err != 0) goto err_init_apertures; + err = kfd_process_init_cwsr(process, filep); + if (err) + goto err_init_cwsr; + return process; +err_init_cwsr: + kfd_process_destroy_pdds(process); err_init_apertures: pqm_uninit(&process->pqm); err_process_pqm_init: -- GitLab From 448aa9117cfa6591c4e18cb32a0329a66b9cd3b7 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 28 Nov 2017 11:01:47 +0000 Subject: [PATCH 0455/1471] drm/i915: Enable hotplug polling after registering the outputs Previously we would enable hotplug polling on the outputs immediately upon construction. This would allow a very early hotplug event to trigger before we had finishing setting up the driver to handle it. Instead, move the output polling to the last step of registration, after we have set up all handlers, including the fbdev configuration. v2: Symmetrically turnoff the hotplug helper in unregister after the fbdev is first synchronised then finalized. This stops a late hotplug event being processed after the interrupts are disabled. Signed-off-by: Chris Wilson Cc: Daniel Vetter Cc: Ville Syrjala Cc: Maarten Lankhorst Reviewed-by: Maarten Lankhorst #v1 Link: https://patchwork.freedesktop.org/patch/msgid/20171128110147.28654-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_drv.c | 16 ++++++++++++++-- drivers/gpu/drm/i915/intel_display.c | 5 +---- 2 files changed, 15 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 9c519aa4f550d..1c419893e3a9d 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -693,8 +693,6 @@ static int i915_load_modeset_init(struct drm_device *dev) /* Only enable hotplug handling once the fbdev is fully set up. */ intel_hpd_init(dev_priv); - drm_kms_helper_poll_init(dev); - return 0; cleanup_gem: @@ -1255,6 +1253,13 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) * cannot run before the connectors are registered. */ intel_fbdev_initial_config_async(dev); + + /* + * We need to coordinate the hotplugs with the asynchronous fbdev + * configuration, for which we use the fbdev->async_cookie. + */ + if (INTEL_INFO(dev_priv)->num_pipes) + drm_kms_helper_poll_init(dev); } /** @@ -1266,6 +1271,13 @@ static void i915_driver_unregister(struct drm_i915_private *dev_priv) intel_fbdev_unregister(dev_priv); intel_audio_deinit(dev_priv); + /* + * After flushing the fbdev (incl. a late async config which will + * have delayed queuing of a hotplug event), then flush the hotplug + * events. + */ + drm_kms_helper_poll_fini(&dev_priv->drm); + intel_gpu_ips_teardown(); acpi_video_unregister(); intel_opregion_unregister(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index fec267dfc0623..76c75d34e799b 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15276,10 +15276,7 @@ static void intel_hpd_poll_fini(struct drm_device *dev) struct intel_connector *connector; struct drm_connector_list_iter conn_iter; - /* First disable polling... */ - drm_kms_helper_poll_fini(dev); - - /* Then kill the work that may have been queued by hpd. */ + /* Kill all the work that may have been queued by hpd. */ drm_connector_list_iter_begin(dev, &conn_iter); for_each_intel_connector_iter(connector, &conn_iter) { if (connector->modeset_retry_work.func) -- GitLab From 00a797225e53a4f488ab725e0b438c2970aa41d7 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Tue, 28 Nov 2017 10:55:15 +0000 Subject: [PATCH 0456/1471] drm/i915/pmu: Return -EINVAL when selecting the inactive CPU In commit 0426c0465461 ("drm/i915/pmu: Only allow running on a single CPU") I attempted to clarify the CPU hotplug logic in our PMU implementation, but missed that a more logical error to return, when attempting to initialize an event on a currently inactive CPU, is -EINVAL rather than -ENODEV. This is because i915 PMU explicitly disallows running counters on more than one CPU at a time, and is not reporting that the requested CPU does not exist, or is off-line. Signed-off-by: Tvrtko Ursulin Reported-by: Chris Wilson Cc: Chris Wilson Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20171128105515.21998-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 3357b690ce909..1c0ee9d68b046 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -338,7 +338,7 @@ static int i915_pmu_event_init(struct perf_event *event) /* only allow running on one cpu at a time */ if (!cpumask_test_cpu(event->cpu, &i915_pmu_cpumask)) - return -ENODEV; + return -EINVAL; if (is_engine_event(event)) { ret = engine_event_init(event); -- GitLab From 9987da4b5dcfc8b94b702d4bb94b30955eb73c75 Mon Sep 17 00:00:00 2001 From: Joonas Lahtinen Date: Mon, 27 Nov 2017 11:12:33 +0200 Subject: [PATCH 0457/1471] drm/i915: Disable THP until we have a GPU read BW W/A We seem to be missing some W/A for 2M pages and are getting a hit on raw GPU read bandwidths (even 30%) even though the GPU write bandwidths improve (even 10%). For now, disable THP, which is our only practical source of 2M pages until we have a W/A for the issue. v2: - Be explicit that we talk about GPU bandwidths (Eero) - s/deny/never/ because that's why (Chris) Reported-by: Valtteri Rantala Fixes: b901bb89324a ("drm/i915/gemfs: enable THP") Signed-off-by: Joonas Lahtinen Cc: Matthew Auld Cc: Chris Wilson Cc: Jani Nikula Cc: Rodrigo Vivi Cc: Valtteri Rantala Cc: Eero Tamminen Reviewed-by: Chris Wilson Reviewed-by: Matthew Auld Tested-by: Valtteri Rantala Link: https://patchwork.freedesktop.org/patch/msgid/20171127091233.7001-1-joonas.lahtinen@linux.intel.com --- drivers/gpu/drm/i915/i915_gemfs.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gemfs.c b/drivers/gpu/drm/i915/i915_gemfs.c index e2993857df37b..888b7d3f04c30 100644 --- a/drivers/gpu/drm/i915/i915_gemfs.c +++ b/drivers/gpu/drm/i915/i915_gemfs.c @@ -52,7 +52,8 @@ int i915_gemfs_init(struct drm_i915_private *i915) if (has_transparent_hugepage()) { struct super_block *sb = gemfs->mnt_sb; - char options[] = "huge=within_size"; + /* FIXME: Disabled until we get W/A for read BW issue. */ + char options[] = "huge=never"; int flags = 0; int err; -- GitLab From 439e2ee4ca520e72870e4fa44aa0076060ad6857 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Wed, 29 Nov 2017 08:24:09 +0000 Subject: [PATCH 0458/1471] drm/i915: Move engine->needs_cmd_parser to engine->flags Will be adding a new per-engine flags shortly so it makes sense to consolidate. v2: Keep the original code flow in intel_engine_cleanup_cmd_parser. (Joonas Lahtinen) Signed-off-by: Tvrtko Ursulin Suggested-by: Chris Wilson Reviewed-by: Chris Wilson Reviewed-by: Sagar Arun Kamble Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171129082409.18189-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_cmd_parser.c | 7 ++++--- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/intel_ringbuffer.h | 8 +++++++- 3 files changed, 12 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index b11629beeb634..ccb5ba043b63a 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -26,6 +26,7 @@ */ #include "i915_drv.h" +#include "intel_ringbuffer.h" /** * DOC: batch buffer command parser @@ -940,7 +941,7 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) return; } - engine->needs_cmd_parser = true; + engine->flags |= I915_ENGINE_NEEDS_CMD_PARSER; } /** @@ -952,7 +953,7 @@ void intel_engine_init_cmd_parser(struct intel_engine_cs *engine) */ void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine) { - if (!engine->needs_cmd_parser) + if (!intel_engine_needs_cmd_parser(engine)) return; fini_hash_table(engine); @@ -1350,7 +1351,7 @@ int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv) /* If the command parser is not enabled, report 0 - unsupported */ for_each_engine(engine, dev_priv, id) { - if (engine->needs_cmd_parser) { + if (intel_engine_needs_cmd_parser(engine)) { active = true; break; } diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 14d9e61a1e06d..70ccd63cbf8ee 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -271,7 +271,7 @@ static inline u64 gen8_noncanonical_addr(u64 address) static inline bool eb_use_cmdparser(const struct i915_execbuffer *eb) { - return eb->engine->needs_cmd_parser && eb->batch_len; + return intel_engine_needs_cmd_parser(eb->engine) && eb->batch_len; } static int eb_create(struct i915_execbuffer *eb) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index ef304c99954c8..2b978b363c0b7 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -538,7 +538,8 @@ struct intel_engine_cs { struct intel_engine_hangcheck hangcheck; - bool needs_cmd_parser; +#define I915_ENGINE_NEEDS_CMD_PARSER BIT(0) + unsigned int flags; /* * Table of commands the command parser needs to know about @@ -597,6 +598,11 @@ struct intel_engine_cs { } stats; }; +static inline bool intel_engine_needs_cmd_parser(struct intel_engine_cs *engine) +{ + return engine->flags & I915_ENGINE_NEEDS_CMD_PARSER; +} + static inline void execlists_set_active(struct intel_engine_execlists *execlists, unsigned int bit) -- GitLab From cf669b4e9fdce4cf1a7bef4e7ab8b3db0da021a2 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Wed, 29 Nov 2017 10:28:05 +0000 Subject: [PATCH 0459/1471] drm/i915: Consolidate checks for engine stats availability Sagar noticed the check can be consolidated between the engine stats implementation and the PMU. My first choice was a static inline helper but that got into include ordering mess quickly fast so I went with a macro instead. At some point we should perhaps looking into taking out the non-ringubffer bits from intel_ringbuffer.h into a new intel_engine.h or something. v2: Use engine->flags. (Chris Wilson) v3: Rebase and mark GuC as not yet supported. (Chris Wilson) v4: Move flag setting to intel_engines_reset_default_submission. (Chris Wilson) v5: Move flag setting to logical_ring_setup. v6: intel_engines_reset_default_submission is the wrong place to set the flag - it needs to be in execlists_set_default_submission. (Sagar) v7: Flag setting in logical_ring_setup is not required. (Chris) Signed-off-by: Tvrtko Ursulin Suggested-by: Sagar Arun Kamble Cc: Sagar Arun Kamble Reviewed-by: Chris Wilson Reviewed-by: Sagar Arun Kamble (v6) Link: https://patchwork.freedesktop.org/patch/msgid/20171129102805.22690-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/i915_pmu.c | 11 ++++------- drivers/gpu/drm/i915/intel_engine_cs.c | 4 ++-- drivers/gpu/drm/i915/intel_guc_submission.c | 2 ++ drivers/gpu/drm/i915/intel_lrc.c | 2 ++ drivers/gpu/drm/i915/intel_ringbuffer.h | 6 ++++++ 5 files changed, 16 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index 1c0ee9d68b046..e8e2faf4982f1 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -90,11 +90,6 @@ static unsigned int event_enabled_bit(struct perf_event *event) return config_enabled_bit(event->attr.config); } -static bool supports_busy_stats(struct drm_i915_private *i915) -{ - return INTEL_GEN(i915) >= 8; -} - static bool pmu_needs_timer(struct drm_i915_private *i915, bool gpu_active) { u64 enable; @@ -123,8 +118,10 @@ static bool pmu_needs_timer(struct drm_i915_private *i915, bool gpu_active) /* * Also there is software busyness tracking available we do not * need the timer for I915_SAMPLE_BUSY counter. + * + * Use RCS as proxy for all engines. */ - else if (supports_busy_stats(i915)) + else if (intel_engine_supports_stats(i915->engine[RCS])) enable &= ~BIT(I915_SAMPLE_BUSY); /* @@ -447,7 +444,7 @@ static void i915_pmu_event_read(struct perf_event *event) static bool engine_needs_busy_stats(struct intel_engine_cs *engine) { - return supports_busy_stats(engine->i915) && + return intel_engine_supports_stats(engine) && (engine->pmu.enable & BIT(I915_SAMPLE_BUSY)); } diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index fede62daf3e15..cffd0c812b7e3 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1863,7 +1863,7 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine) { unsigned long flags; - if (INTEL_GEN(engine->i915) < 8) + if (!intel_engine_supports_stats(engine)) return -ENODEV; spin_lock_irqsave(&engine->stats.lock, flags); @@ -1924,7 +1924,7 @@ void intel_disable_engine_stats(struct intel_engine_cs *engine) { unsigned long flags; - if (INTEL_GEN(engine->i915) < 8) + if (!intel_engine_supports_stats(engine)) return; spin_lock_irqsave(&engine->stats.lock, flags); diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index cf1cc2cb67225..912ff143d5312 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -1453,6 +1453,8 @@ int intel_guc_submission_enable(struct intel_guc *guc) execlists->tasklet.func = guc_submission_tasklet; engine->park = guc_submission_park; engine->unpark = guc_submission_unpark; + + engine->flags &= ~I915_ENGINE_SUPPORTS_STATS; } return 0; diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index 570864583e283..2a8160f603ab0 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1919,6 +1919,8 @@ static void execlists_set_default_submission(struct intel_engine_cs *engine) engine->park = NULL; engine->unpark = NULL; + + engine->flags |= I915_ENGINE_SUPPORTS_STATS; } static void diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 2b978b363c0b7..3d097bbeb2d9c 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -539,6 +539,7 @@ struct intel_engine_cs { struct intel_engine_hangcheck hangcheck; #define I915_ENGINE_NEEDS_CMD_PARSER BIT(0) +#define I915_ENGINE_SUPPORTS_STATS BIT(1) unsigned int flags; /* @@ -603,6 +604,11 @@ static inline bool intel_engine_needs_cmd_parser(struct intel_engine_cs *engine) return engine->flags & I915_ENGINE_NEEDS_CMD_PARSER; } +static inline bool intel_engine_supports_stats(struct intel_engine_cs *engine) +{ + return engine->flags & I915_ENGINE_SUPPORTS_STATS; +} + static inline void execlists_set_active(struct intel_engine_execlists *execlists, unsigned int bit) -- GitLab From bcbec31ce500fe036f75a19bca5c73bfa6dd420b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 13 Nov 2017 17:02:10 +0200 Subject: [PATCH 0460/1471] drm/vblank: Fix vblank timestamp debugs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We're currently calling ktime_to_timespec64() on stack garbage hence the debug output for vblank timestamps also contains garbage. Let's assing something to the ktime_t first before we go converting it to a timespec. While at it micro-optimize the ktime_to_timespec64() calls away when vblank debugging isn't enabled. Fixes: 67680d3c0464 ("drm: vblank: use ktime_t instead of timeval") Cc: Arnd Bergmann Cc: Keith Packard Cc: Sean Paul Cc: Dave Airlie Signed-off-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171113150210.11311-1-ville.syrjala@linux.intel.com Acked-by: Arnd Bergmann --- drivers/gpu/drm/drm_vblank.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c index 09c1c4ff93ca4..2c68909637b53 100644 --- a/drivers/gpu/drm/drm_vblank.c +++ b/drivers/gpu/drm/drm_vblank.c @@ -664,14 +664,16 @@ bool drm_calc_vbltimestamp_from_scanoutpos(struct drm_device *dev, delta_ns = div_s64(1000000LL * (vpos * mode->crtc_htotal + hpos), mode->crtc_clock); - /* save this only for debugging purposes */ - ts_etime = ktime_to_timespec64(etime); - ts_vblank_time = ktime_to_timespec64(*vblank_time); /* Subtract time delta from raw timestamp to get final * vblank_time timestamp for end of vblank. */ - etime = ktime_sub_ns(etime, delta_ns); - *vblank_time = etime; + *vblank_time = ktime_sub_ns(etime, delta_ns); + + if ((drm_debug & DRM_UT_VBL) == 0) + return true; + + ts_etime = ktime_to_timespec64(etime); + ts_vblank_time = ktime_to_timespec64(*vblank_time); DRM_DEBUG_VBL("crtc %u : v p(%d,%d)@ %lld.%06ld -> %lld.%06ld [e %d us, %d rep]\n", pipe, hpos, vpos, -- GitLab From b415b79529391abd29b6a30b93697821f01e082f Mon Sep 17 00:00:00 2001 From: Mark Yao Date: Wed, 29 Nov 2017 09:24:42 -0500 Subject: [PATCH 0461/1471] MAINTAINERS: change maintainer for Rockchip drm drivers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit For personal reasons, Mark Yao will leave rockchip, can not continue maintain drm/rockchip, Sandy Huang and Heiko Stübner will take over drm/rockchip. Cc: Sandy Huang Cc: Heiko Stübner Signed-off-by: Mark Yao Acked-by: Heiko Stuebner [seanpaul added Heiko] Signed-off-by: Sean Paul Link: https://patchwork.freedesktop.org/patch/msgid/20171129142459.32351-1-seanpaul@chromium.org --- MAINTAINERS | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index d4b63c7ffb0e0..4f0756a4ff1aa 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4625,7 +4625,8 @@ F: Documentation/devicetree/bindings/display/bridge/renesas,dw-hdmi.txt F: Documentation/devicetree/bindings/display/renesas,du.txt DRM DRIVERS FOR ROCKCHIP -M: Mark Yao +M: Sandy Huang +M: Heiko Stübner L: dri-devel@lists.freedesktop.org S: Maintained F: drivers/gpu/drm/rockchip/ -- GitLab From 580b9d05a34ef81ecab6a55e068ab269b148b32e Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Wed, 29 Nov 2017 21:49:09 +0530 Subject: [PATCH 0462/1471] drm/i915/guc: Change default GuC FW for SKL to v9.33 This patch makes v9.33 firmware as default firmware for SKL. Note: GuC logging control is changed with this firmware. GuC is expecting i915 to set control bit to enable "default logging" while using GuC action UK_LOG_ENABLE_LOGGING. However i915 is currently not doing this because it is version specific change and can be handled entirely in GuC. It will need to be fixed in future firmwares. This update includes (since v6.1): - HuC RSA Keys updated. - Adding per engine preemption support in GuC scheduler - Minor bug fixes. - Added support to log media reset count for host to read it - Sub-feature level control for power management features. - Minor clean-up for power management interface. - Unified power management interface and scheduler interface into 1 file using same version. - Bug Fix for multi context scheduler flag. - DCC spec changes for BXT + DCT enabling - SB based Pre-ETM/ETM flow enabling for debug signed GuC/HuC - Moving GuC non_critical r/w data to lower SRAM 64KB - Media engine Reset fix. Correctly marking context for resubmission in Media Reset case. - ABT Disable bug fix. Disabled Evaluation mode on context change. - Async FW in Engine Schedule feature (not enabled from KMD) - GuC clean up to align developer build in line to production build. - DCC consistency fix for SKL - Disable ARAT interrupt before programming ARAT delta. - Memory range check in Parse to avoid failure due to overflow. - Enabled WA for MSGCH hang issue - Clear forcewake in CSB when SQ is empty. - Move UkGuckmdInterface.h file from 2016 folders to common 2016 folder. - This is file location change.No functional change done as part of this check in. - Enable decoupled freq for SKL GT4 - 3 tries of wake request needed from GuC2CSME for ME to wake up. Request has come from ME spec - During reset one parameter was not getting accounted - Enabling Guc Log changes for ultra low logging for OCA - Enabling build failure check to catch critical section overflow. - Disable build.bat redundant prints. - Move few least used functions to non-critical section. - Rearrange GuC documentation folder structure. - Synchronize SLPC internal debug interface with other branches. - Fixing Issue with Default Guc Log changes for OCA using special Control Bit v2: Rebase. Updated commit message. Signed-off-by: Jeff McGee Signed-off-by: Sagar Arun Kamble Cc: Spotswood John A Cc: Anusha Srivatsa Cc: Michal Wajdeczko Cc: Rodrigo Vivi Cc: Joonas Lahtinen Reviewed-by: Anusha Srivatsa Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/1511972351-574-2-git-send-email-sagar.a.kamble@intel.com --- drivers/gpu/drm/i915/intel_guc_fw.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index bbab4e1106f33..631e932c99537 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -30,8 +30,8 @@ #include "intel_guc_fw.h" #include "i915_drv.h" -#define SKL_FW_MAJOR 6 -#define SKL_FW_MINOR 1 +#define SKL_FW_MAJOR 9 +#define SKL_FW_MINOR 33 #define BXT_FW_MAJOR 8 #define BXT_FW_MINOR 7 -- GitLab From d416ac780eecc76511fcf15974f54f5e9520cea6 Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Wed, 29 Nov 2017 21:49:10 +0530 Subject: [PATCH 0463/1471] drm/i915/guc: Change default GuC FW for BXT to v9.29 This patch makes v9.29 firmware as default firmware for BXT. Note: GuC logging control is changed with this firmware. GuC is expecting i915 to set control bit to enable "default logging" while using GuC action UK_LOG_ENABLE_LOGGING. However i915 is currently not doing this because it is version specific change and can be handled entirely in GuC. It will need to be fixed in future firmwares. This update includes (since v8.7): - Added support to log media reset count for host to read it - BXT WA for fixing MTP hangs. WaDisableDOPRenderClkGatingAtSubmit - Sub-feature level control for power management features. - Minor clean-up for power management interface. - Unified power management interface and scheduler interface into 1 file using same version. - Bug Fix for multi context scheduler flag. - DCC spec changes for BXT + DCT enabling - Springboard based Pre-ETM/ETM flow enabling for debug signed GuC/HuC - Moving GuC non_critical r/w data to lower SRAM 64KB - Enabled IBC for BXT - Media engine Reset fix. Correctly marking context for resubmission in Media Reset case. - SLPC Dynamic RPe fix to resolve issues where incorrect frequency was set. - ABT Disable bug fix. Disabled Evaluation mode on context change. - GuC clean up to align developer build in line to production build. - Disable ARAT interrupt before programming ARAT delta. - Memory range check in Parse to avoid failure due to overflow. - Clear forcewake in CSB when SQ is empty. - SLPC IBC 1.6 for APL to ensure multiplier does not cap IA below Pe. - Move UkGuckmdInterface.h file from 2016 folders to common 2016 folder. - This is file location change. No functional change done as part of this check in. - 3 tries of wake request needed from GuC2CSME for ME to wake up. Request has come from ME spec - During reset one parameter was not getting accounted - Enabling Guc Log changes for ultra low logging for OCA - Disable build.bat redundant prints. - Move few least used functions to non-critical section. - Rearrange GuC documentation folder structure. - Fixing Issue with Default Guc Log changes for OCA using special Control Bit v2: Rebase. Updated commit message. Signed-off-by: Jeff McGee Signed-off-by: Sagar Arun Kamble Cc: Spotswood John A Cc: Anusha Srivatsa Cc: Michal Wajdeczko Cc: Rodrigo Vivi Cc: Joonas Lahtinen Reviewed-by: Anusha Srivatsa Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/1511972351-574-3-git-send-email-sagar.a.kamble@intel.com --- drivers/gpu/drm/i915/intel_guc_fw.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index 631e932c99537..df2ff965655a8 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -33,8 +33,8 @@ #define SKL_FW_MAJOR 9 #define SKL_FW_MINOR 33 -#define BXT_FW_MAJOR 8 -#define BXT_FW_MINOR 7 +#define BXT_FW_MAJOR 9 +#define BXT_FW_MINOR 29 #define KBL_FW_MAJOR 9 #define KBL_FW_MINOR 14 -- GitLab From cc44085616a94a998dc8f708e37ec9df2df97a87 Mon Sep 17 00:00:00 2001 From: Sagar Arun Kamble Date: Wed, 29 Nov 2017 21:49:11 +0530 Subject: [PATCH 0464/1471] drm/i915/guc: Change default GuC FW for KBL to v9.39 This patch makes v9.39 firmware as default firmware for KBL. Note: GuC logging control is changed with this firmware. GuC is expecting i915 to set control bit to enable "default logging" while using GuC action UK_LOG_ENABLE_LOGGING. However i915 is currently not doing this because it is version specific change and can be handled entirely in GuC. It will need to be fixed in future firmwares. This update includes (since v9.14): - DCC spec changes for BXT + DCT enabling - Bug Fix for power conservation feature SLPC_DCC - Scheduler 1-element submission during DCC cycles. - SB based Pre-ETM/ETM flow enabling for debug signed GuC/HuC - Moving GuC non_critical r/w data to lower SRAM 64KB - Media engine Reset fix. Correctly marking context for resubmission in Media Reset case. - ABT Disable bug fix. Disabled Evaluation mode on context change. - Async FW in Engine Schedule feature (not enabled from KMD) - GuC clean up to align developer build in line to production build. - Disable ARAT interrupt before programming ARAT delta. - Memory range check in Parse to avoid failure due to overflow. - GuC Msg Channel Hang WA - Stall GUC for mmio access when IDI is low during CPD flow. - Fix for submit queue over flow issue - Enabling IBC on KBL GT3 15W, GT4 45W - Disabling wrong device ID WA in production signed kernel - Enabling WA for MSGCH hang issue upto required KBL stepping - Clear forcewake in CSB when SQ is empty. - 3Tries of GuC2CSME wake request - During reset one parameter was not getting accounted - Disable DCC 1-elem mode submission - Move UkGuckmdInterface.h file from 2016 folders to common 2016 folder. - This is file location change.No functional change done as part of this check in. - Enabling Guc Log changes for ultra low logging for OCA - Enabling Dynamic Render Power Well Hysteresis Programming for Compute Worklaods - Enabling build failure check to catch critical section overflow. - Disable build.bat redundant prints. - Move few least used functions to non-critical section. - Rearrange GuC documentation folder structure. - Synchronize SLPC internal debug interface with other branches. - Fixing Issue with Default Guc Log changes for OCA using special Control Bit - Aggressive DCC implementation for supported platforms. v2: Rebase. Updated commit message. Signed-off-by: Jeff McGee Signed-off-by: Sagar Arun Kamble Cc: Spotswood John A Cc: Anusha Srivatsa Cc: Michal Wajdeczko Cc: Rodrigo Vivi Cc: Joonas Lahtinen Reviewed-by: Anusha Srivatsa Signed-off-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/1511972351-574-4-git-send-email-sagar.a.kamble@intel.com --- drivers/gpu/drm/i915/intel_guc_fw.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_guc_fw.c b/drivers/gpu/drm/i915/intel_guc_fw.c index df2ff965655a8..89862fa8ab425 100644 --- a/drivers/gpu/drm/i915/intel_guc_fw.c +++ b/drivers/gpu/drm/i915/intel_guc_fw.c @@ -37,7 +37,7 @@ #define BXT_FW_MINOR 29 #define KBL_FW_MAJOR 9 -#define KBL_FW_MINOR 14 +#define KBL_FW_MINOR 39 #define GLK_FW_MAJOR 10 #define GLK_FW_MINOR 56 -- GitLab From 2f51be0945b9e0b9baafe47289c802be6d4c7980 Mon Sep 17 00:00:00 2001 From: Andrey Gusakov Date: Tue, 7 Nov 2017 19:56:25 +0300 Subject: [PATCH 0465/1471] drm/bridge: tc358767: add copyright lines Add copyright lines for Zodiac who paid for driver development. Signed-off-by: Andrey Gusakov Acked-by: Chris Healy Signed-off-by: Andrzej Hajda Link: https://patchwork.freedesktop.org/patch/msgid/1510073785-16108-8-git-send-email-andrey.gusakov@cogentembedded.com --- drivers/gpu/drm/bridge/tc358767.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/bridge/tc358767.c b/drivers/gpu/drm/bridge/tc358767.c index 8571cfd877c52..e79fbc1b77044 100644 --- a/drivers/gpu/drm/bridge/tc358767.c +++ b/drivers/gpu/drm/bridge/tc358767.c @@ -6,6 +6,8 @@ * * Copyright (C) 2016 Pengutronix, Philipp Zabel * + * Copyright (C) 2016 Zodiac Inflight Innovations + * * Initially based on: drivers/gpu/drm/i2c/tda998x_drv.c * * Copyright (C) 2012 Texas Instruments -- GitLab From 7436830c8d1a3aab61d131d5e0494e2dbb9e55bd Mon Sep 17 00:00:00 2001 From: Valtteri Rantala Date: Tue, 28 Nov 2017 16:45:05 +0200 Subject: [PATCH 0466/1471] drm/i915/glk: Apply WaProgramL3SqcReg1DefaultForPerf for GLK too Testing the texture read performance shows that the same tuning for the SQ credits is needed on GLK as on BXT/APL. This has been also confirmed by Altug from the HW team. V4: Rebase + fix Signed-off-by: Valtteri Rantala Reviewed-by: David Weinehall (v1) Reviewed-by: Rodrigo Vivi Signed-off-by: Imre Deak Link: https://patchwork.freedesktop.org/patch/msgid/1511880305-12166-1-git-send-email-valtteri.rantala@intel.com --- drivers/gpu/drm/i915/intel_engine_cs.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index cffd0c812b7e3..86d4c85c8725f 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1067,6 +1067,15 @@ static int gen9_init_workarounds(struct intel_engine_cs *engine) /* WaDisableSTUnitPowerOptimization:skl,bxt,kbl,glk,cfl */ WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN2, GEN8_ST_PO_DISABLE); + /* WaProgramL3SqcReg1DefaultForPerf:bxt,glk */ + if (IS_GEN9_LP(dev_priv)) { + u32 val = I915_READ(GEN8_L3SQCREG1); + + val &= ~L3_PRIO_CREDITS_MASK; + val |= L3_GENERAL_PRIO_CREDITS(62) | L3_HIGH_PRIO_CREDITS(2); + I915_WRITE(GEN8_L3SQCREG1, val); + } + /* WaOCLCoherentLineFlush:skl,bxt,kbl,cfl */ I915_WRITE(GEN8_L3SQCREG4, (I915_READ(GEN8_L3SQCREG4) | GEN8_LQSC_FLUSH_COHERENT_LINES)); @@ -1184,7 +1193,6 @@ static int skl_init_workarounds(struct intel_engine_cs *engine) static int bxt_init_workarounds(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; - u32 val; int ret; ret = gen9_init_workarounds(engine); @@ -1199,12 +1207,6 @@ static int bxt_init_workarounds(struct intel_engine_cs *engine) I915_WRITE(FF_SLICE_CS_CHICKEN2, _MASKED_BIT_ENABLE(GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE)); - /* WaProgramL3SqcReg1DefaultForPerf:bxt */ - val = I915_READ(GEN8_L3SQCREG1); - val &= ~L3_PRIO_CREDITS_MASK; - val |= L3_GENERAL_PRIO_CREDITS(62) | L3_HIGH_PRIO_CREDITS(2); - I915_WRITE(GEN8_L3SQCREG1, val); - /* WaToEnableHwFixForPushConstHWBug:bxt */ WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2, GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION); -- GitLab From ecf73eb2d27d43b2153bb80671768a06d35521f1 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 30 Nov 2017 10:29:51 +0000 Subject: [PATCH 0467/1471] drm/i915: Skip switch-to-kernel-context on suspend when wedged If the HW is already wedged, attempting to submit a request will generate an -EIO. If we tried this during suspend, we would abort whereas all we want to do is to go sleep and throw away the corrupt state. Fixes: 5ab57c702069 ("drm/i915: Flush logical context image out to memory upon suspend") Testcase: igt/gem_eio/suspend Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Joonas Lahtinen Reviewed-by: Joonas Lahtinen Link: https://patchwork.freedesktop.org/patch/msgid/20171130102951.14965-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/i915_gem.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 354b0546a1915..e083f242b8dc6 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4774,17 +4774,19 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) * state. Fortunately, the kernel_context is disposable and we do * not rely on its state. */ - ret = i915_gem_switch_to_kernel_context(dev_priv); - if (ret) - goto err_unlock; + if (!i915_terminally_wedged(&dev_priv->gpu_error)) { + ret = i915_gem_switch_to_kernel_context(dev_priv); + if (ret) + goto err_unlock; - ret = i915_gem_wait_for_idle(dev_priv, - I915_WAIT_INTERRUPTIBLE | - I915_WAIT_LOCKED); - if (ret && ret != -EIO) - goto err_unlock; + ret = i915_gem_wait_for_idle(dev_priv, + I915_WAIT_INTERRUPTIBLE | + I915_WAIT_LOCKED); + if (ret && ret != -EIO) + goto err_unlock; - assert_kernel_context_is_current(dev_priv); + assert_kernel_context_is_current(dev_priv); + } i915_gem_contexts_lost(dev_priv); mutex_unlock(&dev->struct_mutex); -- GitLab From 5643205c6340b565a3be0fe0e7305dc4aa551c74 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Wed, 29 Nov 2017 19:51:37 +0200 Subject: [PATCH 0468/1471] drm/i915: Avoid PPS HW/SW state mismatch due to rounding MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We store a SW state of the t11_t12 timing in 100usec units but have to program it in 100msec as required by HW. The rounding used during programming means there will be a mismatch between the SW and HW states of this value triggering a "PPS state mismatch" error. Avoid this by storing the already rounded-up value in the SW state. Note that we still calculate panel_power_cycle_delay with the finer 100usec granularity to avoid any needless waits using that version of the delay. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103903 Cc: joks Signed-off-by: Imre Deak Reviewed-by: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171129175137.2889-1-imre.deak@intel.com --- drivers/gpu/drm/i915/intel_dp.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c index dd0b3a0187926..c603d4c903e13 100644 --- a/drivers/gpu/drm/i915/intel_dp.c +++ b/drivers/gpu/drm/i915/intel_dp.c @@ -5355,6 +5355,12 @@ intel_dp_init_panel_power_sequencer(struct intel_dp *intel_dp) */ final->t8 = 1; final->t9 = 1; + + /* + * HW has only a 100msec granularity for t11_t12 so round it up + * accordingly. + */ + final->t11_t12 = roundup(final->t11_t12, 100 * 10); } static void -- GitLab From ff71c2cf5eb255b7dd03da5de4dd414394d9e346 Mon Sep 17 00:00:00 2001 From: Chen-Yu Tsai Date: Mon, 27 Nov 2017 16:46:32 +0800 Subject: [PATCH 0469/1471] drm/sun4i: use sun4i_tcon_of_table to check if a device node is a TCON The sun4i DRM driver maintains a list of compatible strings it uses to check if a device node within the display component graph is a TCON. The TCON driver also has this list, used to bind the TCON driver to the device. These two lists are identical. Instead of maintaining two identical lists, export the list from the TCON driver for the DRM driver to use. Suggested-by: Rob Herring Signed-off-by: Chen-Yu Tsai Signed-off-by: Maxime Ripard Link: https://patchwork.freedesktop.org/patch/msgid/20171127084632.25511-1-wens@csie.org --- drivers/gpu/drm/sun4i/sun4i_drv.c | 8 +------- drivers/gpu/drm/sun4i/sun4i_tcon.c | 4 +++- drivers/gpu/drm/sun4i/sun4i_tcon.h | 2 ++ 3 files changed, 6 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/sun4i/sun4i_drv.c b/drivers/gpu/drm/sun4i/sun4i_drv.c index 75c76cdd82bc6..49215d91c8538 100644 --- a/drivers/gpu/drm/sun4i/sun4i_drv.c +++ b/drivers/gpu/drm/sun4i/sun4i_drv.c @@ -187,13 +187,7 @@ static bool sun4i_drv_node_is_frontend(struct device_node *node) static bool sun4i_drv_node_is_tcon(struct device_node *node) { - return of_device_is_compatible(node, "allwinner,sun4i-a10-tcon") || - of_device_is_compatible(node, "allwinner,sun5i-a13-tcon") || - of_device_is_compatible(node, "allwinner,sun6i-a31-tcon") || - of_device_is_compatible(node, "allwinner,sun6i-a31s-tcon") || - of_device_is_compatible(node, "allwinner,sun7i-a20-tcon") || - of_device_is_compatible(node, "allwinner,sun8i-a33-tcon") || - of_device_is_compatible(node, "allwinner,sun8i-v3s-tcon"); + return !!of_match_node(sun4i_tcon_of_table, node); } static int compare_of(struct device *dev, void *data) diff --git a/drivers/gpu/drm/sun4i/sun4i_tcon.c b/drivers/gpu/drm/sun4i/sun4i_tcon.c index e122f5b2a3955..a1ed462c24307 100644 --- a/drivers/gpu/drm/sun4i/sun4i_tcon.c +++ b/drivers/gpu/drm/sun4i/sun4i_tcon.c @@ -900,7 +900,8 @@ static const struct sun4i_tcon_quirks sun8i_v3s_quirks = { /* nothing is supported */ }; -static const struct of_device_id sun4i_tcon_of_table[] = { +/* sun4i_drv uses this list to check if a device node is a TCON */ +const struct of_device_id sun4i_tcon_of_table[] = { { .compatible = "allwinner,sun4i-a10-tcon", .data = &sun4i_a10_quirks }, { .compatible = "allwinner,sun5i-a13-tcon", .data = &sun5i_a13_quirks }, { .compatible = "allwinner,sun6i-a31-tcon", .data = &sun6i_a31_quirks }, @@ -911,6 +912,7 @@ static const struct of_device_id sun4i_tcon_of_table[] = { { } }; MODULE_DEVICE_TABLE(of, sun4i_tcon_of_table); +EXPORT_SYMBOL(sun4i_tcon_of_table); static struct platform_driver sun4i_tcon_platform_driver = { .probe = sun4i_tcon_probe, diff --git a/drivers/gpu/drm/sun4i/sun4i_tcon.h b/drivers/gpu/drm/sun4i/sun4i_tcon.h index f61bf6d83b4a0..839266a385059 100644 --- a/drivers/gpu/drm/sun4i/sun4i_tcon.h +++ b/drivers/gpu/drm/sun4i/sun4i_tcon.h @@ -197,4 +197,6 @@ void sun4i_tcon_mode_set(struct sun4i_tcon *tcon, void sun4i_tcon_set_status(struct sun4i_tcon *crtc, const struct drm_encoder *encoder, bool enable); +extern const struct of_device_id sun4i_tcon_of_table[]; + #endif /* __SUN4I_TCON_H__ */ -- GitLab From 24f2845056892b297fa95edbddceefd38acc0e62 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 22 Nov 2017 19:39:01 +0100 Subject: [PATCH 0470/1471] drm/i915: Make ips_enabled a property depending on whether IPS is enabled, v3. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ips_enabled was used as a variable of whether IPS can be enabled or not, but should be used to test whether IPS is actually enabled. Changes since v1: - Call needs_modeset on new crtc state. (Ville) - IPS can be enabled with sprite plane enabled too. (Ville) - Fix CDCLK vs IPS workaround. (Ville) Changes since v2: - Only re-enable fastset when inheriting mode. (Ville) - Put the conditions for enabling and disabling IPS in a helper. Changes since v3: - Keep the max_cdclk workaround working. (Ville) - Also check logical cdclk out of paranoia. - Remove planes check from IPS disable function for initial disable. - Remove assert_plane_enabled/disabled checks and use crtc_state->active_planes for hsw_enable_ips only, always allow calling hsw_disable_ips to disable it initially in hw. Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171122183901.47720-1-maarten.lankhorst@linux.intel.com [mlankhorst: pipe_config -> crtc_state (Ville)] Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_cdclk.c | 2 +- drivers/gpu/drm/i915/intel_display.c | 168 ++++++++++++++++---------- drivers/gpu/drm/i915/intel_drv.h | 1 + drivers/gpu/drm/i915/intel_pipe_crc.c | 2 - 4 files changed, 109 insertions(+), 64 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c index e8884c2ade98b..9c5ceb98d48f4 100644 --- a/drivers/gpu/drm/i915/intel_cdclk.c +++ b/drivers/gpu/drm/i915/intel_cdclk.c @@ -1896,7 +1896,7 @@ int intel_crtc_compute_min_cdclk(const struct intel_crtc_state *crtc_state) min_cdclk = intel_pixel_rate_to_cdclk(dev_priv, crtc_state->pixel_rate); /* pixel rate mustn't exceed 95% of cdclk with IPS on BDW */ - if (IS_BROADWELL(dev_priv) && crtc_state->ips_enabled) + if (IS_BROADWELL(dev_priv) && hsw_crtc_state_ips_capable(crtc_state)) min_cdclk = DIV_ROUND_UP(min_cdclk * 100, 95); /* BSpec says "Do not use DisplayPort with CDCLK less than 432 MHz, diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 76c75d34e799b..68f101c0e9091 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -489,7 +489,7 @@ static const struct intel_limit intel_limits_bxt = { }; static bool -needs_modeset(struct drm_crtc_state *state) +needs_modeset(const struct drm_crtc_state *state) { return drm_atomic_crtc_needs_modeset(state); } @@ -4833,7 +4833,7 @@ void hsw_enable_ips(const struct intel_crtc_state *crtc_state) struct drm_device *dev = crtc->base.dev; struct drm_i915_private *dev_priv = to_i915(dev); - if (!crtc->config->ips_enabled) + if (!crtc_state->ips_enabled) return; /* @@ -4841,8 +4841,7 @@ void hsw_enable_ips(const struct intel_crtc_state *crtc_state) * This function is called from post_plane_update, which is run after * a vblank wait. */ - - assert_plane_enabled(to_intel_plane(crtc->base.primary)); + WARN_ON(!(crtc_state->active_planes & ~BIT(PLANE_CURSOR))); if (IS_BROADWELL(dev_priv)) { mutex_lock(&dev_priv->pcu_lock); @@ -4877,8 +4876,6 @@ void hsw_disable_ips(const struct intel_crtc_state *crtc_state) if (!crtc_state->ips_enabled) return; - assert_plane_enabled(to_intel_plane(crtc->base.primary)); - if (IS_BROADWELL(dev_priv)) { mutex_lock(&dev_priv->pcu_lock); WARN_ON(sandybridge_pcode_write(dev_priv, DISPLAY_IPS_CONTROL, 0)); @@ -4931,14 +4928,6 @@ intel_post_enable_primary(struct drm_crtc *crtc, struct intel_crtc *intel_crtc = to_intel_crtc(crtc); int pipe = intel_crtc->pipe; - /* - * FIXME IPS should be fine as long as one plane is - * enabled, but in practice it seems to have problems - * when going from primary only to sprite only and vice - * versa. - */ - hsw_enable_ips(new_crtc_state); - /* * Gen2 reports pipe underruns whenever all planes are disabled. * So don't enable underrun reporting before at least some planes @@ -4954,10 +4943,9 @@ intel_post_enable_primary(struct drm_crtc *crtc, intel_check_pch_fifo_underruns(dev_priv); } -/* FIXME move all this to pre_plane_update() with proper state tracking */ +/* FIXME get rid of this and use pre_plane_update */ static void -intel_pre_disable_primary(struct drm_crtc *crtc, - const struct intel_crtc_state *old_crtc_state) +intel_pre_disable_primary_noatomic(struct drm_crtc *crtc) { struct drm_device *dev = crtc->dev; struct drm_i915_private *dev_priv = to_i915(dev); @@ -4966,32 +4954,12 @@ intel_pre_disable_primary(struct drm_crtc *crtc, /* * Gen2 reports pipe underruns whenever all planes are disabled. - * So diasble underrun reporting before all the planes get disabled. - * FIXME: Need to fix the logic to work when we turn off all planes - * but leave the pipe running. + * So disable underrun reporting before all the planes get disabled. */ if (IS_GEN2(dev_priv)) intel_set_cpu_fifo_underrun_reporting(dev_priv, pipe, false); - /* - * FIXME IPS should be fine as long as one plane is - * enabled, but in practice it seems to have problems - * when going from primary only to sprite only and vice - * versa. - */ - hsw_disable_ips(old_crtc_state); -} - -/* FIXME get rid of this and use pre_plane_update */ -static void -intel_pre_disable_primary_noatomic(struct drm_crtc *crtc) -{ - struct drm_device *dev = crtc->dev; - struct drm_i915_private *dev_priv = to_i915(dev); - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - int pipe = intel_crtc->pipe; - - intel_pre_disable_primary(crtc, to_intel_crtc_state(crtc->state)); + hsw_disable_ips(to_intel_crtc_state(crtc->state)); /* * Vblank time updates from the shadow to live plane control register @@ -5007,6 +4975,38 @@ intel_pre_disable_primary_noatomic(struct drm_crtc *crtc) intel_wait_for_vblank(dev_priv, pipe); } +static bool hsw_pre_update_disable_ips(const struct intel_crtc_state *old_crtc_state, + const struct intel_crtc_state *new_crtc_state) +{ + if (!old_crtc_state->ips_enabled) + return false; + + if (needs_modeset(&new_crtc_state->base)) + return true; + + return !new_crtc_state->ips_enabled; +} + +static bool hsw_post_update_enable_ips(const struct intel_crtc_state *old_crtc_state, + const struct intel_crtc_state *new_crtc_state) +{ + if (!new_crtc_state->ips_enabled) + return false; + + if (needs_modeset(&new_crtc_state->base)) + return true; + + /* + * We can't read out IPS on broadwell, assume the worst and + * forcibly enable IPS on the first fastset. + */ + if (new_crtc_state->update_pipe && + old_crtc_state->base.adjusted_mode.private_flags & I915_MODE_FLAG_INHERITED) + return true; + + return !old_crtc_state->ips_enabled; +} + static void intel_post_plane_update(struct intel_crtc_state *old_crtc_state) { struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->base.crtc); @@ -5023,6 +5023,9 @@ static void intel_post_plane_update(struct intel_crtc_state *old_crtc_state) if (pipe_config->update_wm_post && pipe_config->base.active) intel_update_watermarks(crtc); + if (hsw_post_update_enable_ips(old_crtc_state, pipe_config)) + hsw_enable_ips(pipe_config); + if (old_pri_state) { struct intel_plane_state *primary_state = intel_atomic_get_new_plane_state(to_intel_atomic_state(old_state), @@ -5053,6 +5056,9 @@ static void intel_pre_plane_update(struct intel_crtc_state *old_crtc_state, struct intel_atomic_state *old_intel_state = to_intel_atomic_state(old_state); + if (hsw_pre_update_disable_ips(old_crtc_state, pipe_config)) + hsw_disable_ips(old_crtc_state); + if (old_pri_state) { struct intel_plane_state *primary_state = intel_atomic_get_new_plane_state(old_intel_state, @@ -5061,10 +5067,13 @@ static void intel_pre_plane_update(struct intel_crtc_state *old_crtc_state, to_intel_plane_state(old_pri_state); intel_fbc_pre_update(crtc, pipe_config, primary_state); - - if (old_primary_state->base.visible && + /* + * Gen2 reports pipe underruns whenever all planes are disabled. + * So disable underrun reporting before all the planes get disabled. + */ + if (IS_GEN2(dev_priv) && old_primary_state->base.visible && (modeset || !primary_state->base.visible)) - intel_pre_disable_primary(&crtc->base, old_crtc_state); + intel_set_cpu_fifo_underrun_reporting(dev_priv, crtc->pipe, false); } /* @@ -6195,18 +6204,20 @@ static int ironlake_fdi_compute_config(struct intel_crtc *intel_crtc, return ret; } -static bool pipe_config_supports_ips(struct drm_i915_private *dev_priv, - struct intel_crtc_state *pipe_config) +bool hsw_crtc_state_ips_capable(const struct intel_crtc_state *crtc_state) { - if (pipe_config->ips_force_disable) + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc); + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev); + + /* IPS only exists on ULT machines and is tied to pipe A. */ + if (!hsw_crtc_supports_ips(crtc)) return false; - if (pipe_config->pipe_bpp > 24) + if (!i915_modparams.enable_ips) return false; - /* HSW can handle pixel rate up to cdclk? */ - if (IS_HASWELL(dev_priv)) - return true; + if (crtc_state->pipe_bpp > 24) + return false; /* * We compare against max which means we must take @@ -6215,19 +6226,41 @@ static bool pipe_config_supports_ips(struct drm_i915_private *dev_priv, * * Should measure whether using a lower cdclk w/o IPS */ - return pipe_config->pixel_rate <= - dev_priv->max_cdclk_freq * 95 / 100; + if (IS_BROADWELL(dev_priv) && + crtc_state->pixel_rate > dev_priv->max_cdclk_freq * 95 / 100) + return false; + + return true; } -static void hsw_compute_ips_config(struct intel_crtc *crtc, - struct intel_crtc_state *pipe_config) +static bool hsw_compute_ips_config(struct intel_crtc_state *crtc_state) { - struct drm_device *dev = crtc->base.dev; - struct drm_i915_private *dev_priv = to_i915(dev); + struct drm_i915_private *dev_priv = + to_i915(crtc_state->base.crtc->dev); + struct intel_atomic_state *intel_state = + to_intel_atomic_state(crtc_state->base.state); - pipe_config->ips_enabled = i915_modparams.enable_ips && - hsw_crtc_supports_ips(crtc) && - pipe_config_supports_ips(dev_priv, pipe_config); + if (!hsw_crtc_state_ips_capable(crtc_state)) + return false; + + if (crtc_state->ips_force_disable) + return false; + + /* + * FIXME IPS should be fine as long as one plane is + * enabled, but in practice it seems to have problems + * when going from primary only to sprite only and vice + * versa. + */ + if (!(crtc_state->active_planes & BIT(PLANE_PRIMARY))) + return false; + + /* pixel rate mustn't exceed 95% of cdclk with IPS on BDW */ + if (IS_BROADWELL(dev_priv) && + crtc_state->pixel_rate > intel_state->cdclk.logical.cdclk * 95 / 100) + return false; + + return true; } static bool intel_crtc_supports_double_wide(const struct intel_crtc *crtc) @@ -6345,9 +6378,6 @@ static int intel_crtc_compute_config(struct intel_crtc *crtc, intel_crtc_compute_pixel_rate(pipe_config); - if (HAS_IPS(dev_priv)) - hsw_compute_ips_config(crtc, pipe_config); - if (pipe_config->has_pch_encoder) return ironlake_fdi_compute_config(crtc, pipe_config); @@ -9183,6 +9213,19 @@ static bool haswell_get_pipe_config(struct intel_crtc *crtc, ironlake_get_pfit_config(crtc, pipe_config); } + if (hsw_crtc_supports_ips(crtc)) { + if (IS_HASWELL(dev_priv)) + pipe_config->ips_enabled = I915_READ(IPS_CTL) & IPS_ENABLE; + else { + /* + * We cannot readout IPS state on broadwell, set to + * true so we can set it to a defined state on first + * commit. + */ + pipe_config->ips_enabled = true; + } + } + if (pipe_config->cpu_transcoder != TRANSCODER_EDP && !transcoder_is_dsi(pipe_config->cpu_transcoder)) { pipe_config->pixel_multiplier = @@ -10435,6 +10478,9 @@ static int intel_crtc_atomic_check(struct drm_crtc *crtc, pipe_config); } + if (HAS_IPS(dev_priv)) + pipe_config->ips_enabled = hsw_compute_ips_config(pipe_config); + return ret; } diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index c679df99d5300..64426d3e078eb 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1488,6 +1488,7 @@ bool bxt_find_best_dpll(struct intel_crtc_state *crtc_state, int target_clock, int chv_calc_dpll_params(int refclk, struct dpll *pll_clock); bool intel_crtc_active(struct intel_crtc *crtc); +bool hsw_crtc_state_ips_capable(const struct intel_crtc_state *crtc_state); void hsw_enable_ips(const struct intel_crtc_state *crtc_state); void hsw_disable_ips(const struct intel_crtc_state *crtc_state); enum intel_display_power_domain intel_port_to_power_domain(enum port port); diff --git a/drivers/gpu/drm/i915/intel_pipe_crc.c b/drivers/gpu/drm/i915/intel_pipe_crc.c index 61641d479b938..1f5cd572a7ffe 100644 --- a/drivers/gpu/drm/i915/intel_pipe_crc.c +++ b/drivers/gpu/drm/i915/intel_pipe_crc.c @@ -541,8 +541,6 @@ static void hsw_pipe_A_crc_wa(struct drm_i915_private *dev_priv, * completely disable it. */ pipe_config->ips_force_disable = enable; - if (pipe_config->ips_enabled == enable) - pipe_config->base.connectors_changed = true; } if (IS_HASWELL(dev_priv)) { -- GitLab From adbe5c5cd32c732a4ea1d4c5f41714fa836d69b6 Mon Sep 17 00:00:00 2001 From: Maarten Lankhorst Date: Wed, 22 Nov 2017 19:39:06 +0100 Subject: [PATCH 0471/1471] drm/i915: Enable IPS with only sprite plane visible too, v4. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This comment predates atomic, and I think with the way we currently track IPS, it's safe to enable this for the case we switch too. Changes since v1: - Keep IPS enabled when switching planes. Changes since v2: - Enable IPS when at least one plane is enabled. (Ville) Changes since v3: - Actually do what was advertised in v3, sigh! (Ville, CI) Signed-off-by: Maarten Lankhorst Link: https://patchwork.freedesktop.org/patch/msgid/20171122183906.47767-1-maarten.lankhorst@linux.intel.com Reviewed-by: Ville Syrjälä --- drivers/gpu/drm/i915/intel_display.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 68f101c0e9091..601c23be82641 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -6246,13 +6246,8 @@ static bool hsw_compute_ips_config(struct intel_crtc_state *crtc_state) if (crtc_state->ips_force_disable) return false; - /* - * FIXME IPS should be fine as long as one plane is - * enabled, but in practice it seems to have problems - * when going from primary only to sprite only and vice - * versa. - */ - if (!(crtc_state->active_planes & BIT(PLANE_PRIMARY))) + /* IPS should be fine as long as at least one plane is enabled. */ + if (!(crtc_state->active_planes & ~BIT(PLANE_CURSOR))) return false; /* pixel rate mustn't exceed 95% of cdclk with IPS on BDW */ -- GitLab From b66d0f348624101203dbec22e45c74e1c895b881 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 6 Nov 2017 20:18:07 +0100 Subject: [PATCH 0472/1471] drm/probe-helper: Fix drm_kms_helper_poll_enable() docs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix docs to reflect code and drm_kms_helper_poll_disable() docs by saying that calling drm_kms_helper_poll_enable() is fine even if output polling is not enabled. Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171106191812.38927-2-noralf@tronnes.org --- drivers/gpu/drm/drm_probe_helper.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/gpu/drm/drm_probe_helper.c b/drivers/gpu/drm/drm_probe_helper.c index 6dc2dde5b6720..555fbe54d6e22 100644 --- a/drivers/gpu/drm/drm_probe_helper.c +++ b/drivers/gpu/drm/drm_probe_helper.c @@ -216,8 +216,7 @@ enum drm_mode_status drm_connector_mode_valid(struct drm_connector *connector, * suspend/resume. * * Drivers can call this helper from their device resume implementation. It is - * an error to call this when the output polling support has not yet been set - * up. + * not an error to call this even when output polling isn't enabled. * * Note that calls to enable and disable polling must be strictly ordered, which * is automatically the case when they're only call from suspend/resume -- GitLab From ca038cfb5cfa8d48b427b916efe97e711f5a18c7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 6 Nov 2017 20:18:08 +0100 Subject: [PATCH 0473/1471] drm/modeset-helper: Add simple modeset suspend/resume helpers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add drm_mode_config_helper_suspend/resume() which takes care of atomic modeset suspend/resume for simple use cases. The suspend state is stored in struct drm_mode_config. Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171106191812.38927-3-noralf@tronnes.org --- drivers/gpu/drm/drm_modeset_helper.c | 76 ++++++++++++++++++++++++++++ include/drm/drm_mode_config.h | 9 ++++ include/drm/drm_modeset_helper.h | 3 ++ 3 files changed, 88 insertions(+) diff --git a/drivers/gpu/drm/drm_modeset_helper.c b/drivers/gpu/drm/drm_modeset_helper.c index 9cb1eede0b4d9..f1c24ab0ef09f 100644 --- a/drivers/gpu/drm/drm_modeset_helper.c +++ b/drivers/gpu/drm/drm_modeset_helper.c @@ -20,6 +20,9 @@ * OF THIS SOFTWARE. */ +#include +#include +#include #include #include @@ -156,3 +159,76 @@ int drm_crtc_init(struct drm_device *dev, struct drm_crtc *crtc, NULL); } EXPORT_SYMBOL(drm_crtc_init); + +/** + * drm_mode_config_helper_suspend - Modeset suspend helper + * @dev: DRM device + * + * This helper function takes care of suspending the modeset side. It disables + * output polling if initialized, suspends fbdev if used and finally calls + * drm_atomic_helper_suspend(). + * If suspending fails, fbdev and polling is re-enabled. + * + * Returns: + * Zero on success, negative error code on error. + * + * See also: + * drm_kms_helper_poll_disable() and drm_fb_helper_set_suspend_unlocked(). + */ +int drm_mode_config_helper_suspend(struct drm_device *dev) +{ + struct drm_atomic_state *state; + + if (!dev) + return 0; + + drm_kms_helper_poll_disable(dev); + drm_fb_helper_set_suspend_unlocked(dev->fb_helper, 1); + state = drm_atomic_helper_suspend(dev); + if (IS_ERR(state)) { + drm_fb_helper_set_suspend_unlocked(dev->fb_helper, 0); + drm_kms_helper_poll_enable(dev); + return PTR_ERR(state); + } + + dev->mode_config.suspend_state = state; + + return 0; +} +EXPORT_SYMBOL(drm_mode_config_helper_suspend); + +/** + * drm_mode_config_helper_resume - Modeset resume helper + * @dev: DRM device + * + * This helper function takes care of resuming the modeset side. It calls + * drm_atomic_helper_resume(), resumes fbdev if used and enables output polling + * if initiaized. + * + * Returns: + * Zero on success, negative error code on error. + * + * See also: + * drm_fb_helper_set_suspend_unlocked() and drm_kms_helper_poll_enable(). + */ +int drm_mode_config_helper_resume(struct drm_device *dev) +{ + int ret; + + if (!dev) + return 0; + + if (WARN_ON(!dev->mode_config.suspend_state)) + return -EINVAL; + + ret = drm_atomic_helper_resume(dev, dev->mode_config.suspend_state); + if (ret) + DRM_ERROR("Failed to resume (%d)\n", ret); + dev->mode_config.suspend_state = NULL; + + drm_fb_helper_set_suspend_unlocked(dev->fb_helper, 0); + drm_kms_helper_poll_enable(dev); + + return ret; +} +EXPORT_SYMBOL(drm_mode_config_helper_resume); diff --git a/include/drm/drm_mode_config.h b/include/drm/drm_mode_config.h index cf73c5eaab980..eb3b8dba3e2c7 100644 --- a/include/drm/drm_mode_config.h +++ b/include/drm/drm_mode_config.h @@ -753,6 +753,15 @@ struct drm_mode_config { /* cursor size */ uint32_t cursor_width, cursor_height; + /** + * @suspend_state: + * + * Atomic state when suspended. + * Set by drm_mode_config_helper_suspend() and cleared by + * drm_mode_config_helper_resume(). + */ + struct drm_atomic_state *suspend_state; + const struct drm_mode_config_helper_funcs *helper_private; }; diff --git a/include/drm/drm_modeset_helper.h b/include/drm/drm_modeset_helper.h index cb0ec92e11e60..efa337f031299 100644 --- a/include/drm/drm_modeset_helper.h +++ b/include/drm/drm_modeset_helper.h @@ -34,4 +34,7 @@ void drm_helper_mode_fill_fb_struct(struct drm_device *dev, int drm_crtc_init(struct drm_device *dev, struct drm_crtc *crtc, const struct drm_crtc_funcs *funcs); +int drm_mode_config_helper_suspend(struct drm_device *dev); +int drm_mode_config_helper_resume(struct drm_device *dev); + #endif -- GitLab From 194b8799d27cd61bbc6da4e3ed333de261c07407 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 6 Nov 2017 20:18:09 +0100 Subject: [PATCH 0474/1471] drm/arm/mali: Use drm_mode_config_helper_suspend/resume() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace driver's code with the generic helpers that do the same thing. Cc: Liviu Dudau Cc: Brian Starkey Signed-off-by: Noralf Trønnes Reviewed-by: Liviu Dudau Link: https://patchwork.freedesktop.org/patch/msgid/20171106191812.38927-4-noralf@tronnes.org --- drivers/gpu/drm/arm/malidp_drv.c | 24 +++--------------------- drivers/gpu/drm/arm/malidp_drv.h | 1 - 2 files changed, 3 insertions(+), 22 deletions(-) diff --git a/drivers/gpu/drm/arm/malidp_drv.c b/drivers/gpu/drm/arm/malidp_drv.c index b8944666a18f0..75f0bce339418 100644 --- a/drivers/gpu/drm/arm/malidp_drv.c +++ b/drivers/gpu/drm/arm/malidp_drv.c @@ -27,6 +27,7 @@ #include #include #include +#include #include #include "malidp_drv.h" @@ -749,34 +750,15 @@ static int malidp_platform_remove(struct platform_device *pdev) static int __maybe_unused malidp_pm_suspend(struct device *dev) { struct drm_device *drm = dev_get_drvdata(dev); - struct malidp_drm *malidp = drm->dev_private; - - drm_kms_helper_poll_disable(drm); - console_lock(); - drm_fbdev_cma_set_suspend(malidp->fbdev, 1); - console_unlock(); - malidp->pm_state = drm_atomic_helper_suspend(drm); - if (IS_ERR(malidp->pm_state)) { - console_lock(); - drm_fbdev_cma_set_suspend(malidp->fbdev, 0); - console_unlock(); - drm_kms_helper_poll_enable(drm); - return PTR_ERR(malidp->pm_state); - } - return 0; + return drm_mode_config_helper_suspend(drm); } static int __maybe_unused malidp_pm_resume(struct device *dev) { struct drm_device *drm = dev_get_drvdata(dev); - struct malidp_drm *malidp = drm->dev_private; - drm_atomic_helper_resume(drm, malidp->pm_state); - console_lock(); - drm_fbdev_cma_set_suspend(malidp->fbdev, 0); - console_unlock(); - drm_kms_helper_poll_enable(drm); + drm_mode_config_helper_resume(drm); return 0; } diff --git a/drivers/gpu/drm/arm/malidp_drv.h b/drivers/gpu/drm/arm/malidp_drv.h index 2e2033140efc0..70ed6aeccf052 100644 --- a/drivers/gpu/drm/arm/malidp_drv.h +++ b/drivers/gpu/drm/arm/malidp_drv.h @@ -24,7 +24,6 @@ struct malidp_drm { struct drm_crtc crtc; wait_queue_head_t wq; atomic_t config_valid; - struct drm_atomic_state *pm_state; u32 core_id; }; -- GitLab From 6e8e9a01ec2a35dcb502a216d5c84452ebbed90c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 6 Nov 2017 20:18:11 +0100 Subject: [PATCH 0475/1471] drm/tinydrm: Use drm_mode_config_helper_suspend/resume() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Replace driver's code with the generic helpers that do the same thing. Remove todo entry. Signed-off-by: Noralf Trønnes Reviewed-by: Stefan Agner Link: https://patchwork.freedesktop.org/patch/msgid/20171106191812.38927-6-noralf@tronnes.org --- Documentation/gpu/todo.rst | 5 -- drivers/gpu/drm/tinydrm/core/tinydrm-core.c | 67 --------------------- drivers/gpu/drm/tinydrm/mi0283qt.c | 7 ++- include/drm/tinydrm/tinydrm.h | 4 -- 4 files changed, 5 insertions(+), 78 deletions(-) diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst index e9840d693a863..a44f379d2b253 100644 --- a/Documentation/gpu/todo.rst +++ b/Documentation/gpu/todo.rst @@ -404,11 +404,6 @@ those drivers as simple as possible, so lots of room for refactoring: a drm_device wrong. Doesn't matter, since everyone else gets it wrong too :-) -- With the fbdev pointer in dev->mode_config we could also make - suspend/resume helpers entirely generic, at least if we add a - dev->mode_config.suspend_state. We could even provide a generic pm_ops - structure with those. - - also rework the drm_framebuffer_funcs->dirty hook wire-up, see above. Contact: Noralf Trønnes, Daniel Vetter diff --git a/drivers/gpu/drm/tinydrm/core/tinydrm-core.c b/drivers/gpu/drm/tinydrm/core/tinydrm-core.c index 1a8a57cad4312..bd7b82824a34c 100644 --- a/drivers/gpu/drm/tinydrm/core/tinydrm-core.c +++ b/drivers/gpu/drm/tinydrm/core/tinydrm-core.c @@ -292,71 +292,4 @@ void tinydrm_shutdown(struct tinydrm_device *tdev) } EXPORT_SYMBOL(tinydrm_shutdown); -/** - * tinydrm_suspend - Suspend tinydrm - * @tdev: tinydrm device - * - * Used in driver PM operations to suspend tinydrm. - * Suspends fbdev and DRM. - * Resume with tinydrm_resume(). - * - * Returns: - * Zero on success, negative error code on failure. - */ -int tinydrm_suspend(struct tinydrm_device *tdev) -{ - struct drm_atomic_state *state; - - if (tdev->suspend_state) { - DRM_ERROR("Failed to suspend: state already set\n"); - return -EINVAL; - } - - drm_fbdev_cma_set_suspend_unlocked(tdev->fbdev_cma, 1); - state = drm_atomic_helper_suspend(tdev->drm); - if (IS_ERR(state)) { - drm_fbdev_cma_set_suspend_unlocked(tdev->fbdev_cma, 0); - return PTR_ERR(state); - } - - tdev->suspend_state = state; - - return 0; -} -EXPORT_SYMBOL(tinydrm_suspend); - -/** - * tinydrm_resume - Resume tinydrm - * @tdev: tinydrm device - * - * Used in driver PM operations to resume tinydrm. - * Suspend with tinydrm_suspend(). - * - * Returns: - * Zero on success, negative error code on failure. - */ -int tinydrm_resume(struct tinydrm_device *tdev) -{ - struct drm_atomic_state *state = tdev->suspend_state; - int ret; - - if (!state) { - DRM_ERROR("Failed to resume: state is not set\n"); - return -EINVAL; - } - - tdev->suspend_state = NULL; - - ret = drm_atomic_helper_resume(tdev->drm, state); - if (ret) { - DRM_ERROR("Error resuming state: %d\n", ret); - return ret; - } - - drm_fbdev_cma_set_suspend_unlocked(tdev->fbdev_cma, 0); - - return 0; -} -EXPORT_SYMBOL(tinydrm_resume); - MODULE_LICENSE("GPL"); diff --git a/drivers/gpu/drm/tinydrm/mi0283qt.c b/drivers/gpu/drm/tinydrm/mi0283qt.c index 6a83b30932543..70ae4f76f4551 100644 --- a/drivers/gpu/drm/tinydrm/mi0283qt.c +++ b/drivers/gpu/drm/tinydrm/mi0283qt.c @@ -9,6 +9,7 @@ * (at your option) any later version. */ +#include #include #include #include @@ -231,7 +232,7 @@ static int __maybe_unused mi0283qt_pm_suspend(struct device *dev) struct mipi_dbi *mipi = dev_get_drvdata(dev); int ret; - ret = tinydrm_suspend(&mipi->tinydrm); + ret = drm_mode_config_helper_suspend(mipi->tinydrm.drm); if (ret) return ret; @@ -249,7 +250,9 @@ static int __maybe_unused mi0283qt_pm_resume(struct device *dev) if (ret) return ret; - return tinydrm_resume(&mipi->tinydrm); + drm_mode_config_helper_resume(mipi->tinydrm.drm); + + return 0; } static const struct dev_pm_ops mi0283qt_pm_ops = { diff --git a/include/drm/tinydrm/tinydrm.h b/include/drm/tinydrm/tinydrm.h index 423828922e5ab..03cd9d72308c1 100644 --- a/include/drm/tinydrm/tinydrm.h +++ b/include/drm/tinydrm/tinydrm.h @@ -20,7 +20,6 @@ * @pipe: Display pipe structure * @dirty_lock: Serializes framebuffer flushing * @fbdev_cma: CMA fbdev structure - * @suspend_state: Atomic state when suspended * @fb_funcs: Framebuffer functions used when creating framebuffers */ struct tinydrm_device { @@ -28,7 +27,6 @@ struct tinydrm_device { struct drm_simple_display_pipe pipe; struct mutex dirty_lock; struct drm_fbdev_cma *fbdev_cma; - struct drm_atomic_state *suspend_state; const struct drm_framebuffer_funcs *fb_funcs; }; @@ -93,8 +91,6 @@ int devm_tinydrm_init(struct device *parent, struct tinydrm_device *tdev, struct drm_driver *driver); int devm_tinydrm_register(struct tinydrm_device *tdev); void tinydrm_shutdown(struct tinydrm_device *tdev); -int tinydrm_suspend(struct tinydrm_device *tdev); -int tinydrm_resume(struct tinydrm_device *tdev); void tinydrm_display_pipe_update(struct drm_simple_display_pipe *pipe, struct drm_plane_state *old_state); -- GitLab From 3233fc0a691a351b6c1f6ff080b2a7d893337d20 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Noralf=20Tr=C3=B8nnes?= Date: Mon, 6 Nov 2017 20:18:12 +0100 Subject: [PATCH 0476/1471] drm/docs: Add todo entry for simple modeset suspend/resume MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add entry for conversion of drivers to new helpers. Signed-off-by: Noralf Trønnes Reviewed-by: Daniel Vetter Link: https://patchwork.freedesktop.org/patch/msgid/20171106191812.38927-7-noralf@tronnes.org --- Documentation/gpu/todo.rst | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/Documentation/gpu/todo.rst b/Documentation/gpu/todo.rst index a44f379d2b253..6bce1beafabe1 100644 --- a/Documentation/gpu/todo.rst +++ b/Documentation/gpu/todo.rst @@ -185,6 +185,15 @@ are better. Contact: Sean Paul, Maintainer of the driver you plan to convert +Convert drivers to use simple modeset suspend/resume +---------------------------------------------------- + +Most drivers (except i915 and nouveau) that use +drm_atomic_helper_suspend/resume() can probably be converted to use +drm_mode_config_helper_suspend/resume(). + +Contact: Maintainer of the driver you plan to convert + Core refactorings ================= -- GitLab From 8bd099a71db660f3bd000dd2256032dbf49a6e84 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 30 Nov 2017 12:52:53 +0000 Subject: [PATCH 0477/1471] drm/i915: Tidy up signed/unsigned comparison MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit drivers/gpu/drm/i915/i915_irq.c|3096 col 29| error: comparison between signed and unsigned integer expressions [-Werror=sign-compare] Signed-off-by: Chris Wilson Cc: Ville Syrjälä Link: https://patchwork.freedesktop.org/patch/msgid/20171130125253.21405-1-chris@chris-wilson.co.uk Reviewed-by: Michal Wajdeczko --- drivers/gpu/drm/i915/i915_irq.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 4fb183ae7a07e..7cac07db89b99 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3068,7 +3068,7 @@ static void vlv_display_irq_reset(struct drm_i915_private *dev_priv) i9xx_pipestat_irq_reset(dev_priv); GEN3_IRQ_RESET(VLV_); - dev_priv->irq_mask = ~0; + dev_priv->irq_mask = ~0u; } static void vlv_display_irq_postinstall(struct drm_i915_private *dev_priv) @@ -3093,7 +3093,7 @@ static void vlv_display_irq_postinstall(struct drm_i915_private *dev_priv) enable_mask |= I915_DISPLAY_PIPE_C_EVENT_INTERRUPT | I915_LPE_PIPE_C_INTERRUPT; - WARN_ON(dev_priv->irq_mask != ~0); + WARN_ON(dev_priv->irq_mask != ~0u); dev_priv->irq_mask = ~enable_mask; -- GitLab From b5e0a9418e09a7b6df1728a26832c7c34aa1adf8 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 1 Dec 2017 00:15:36 +0000 Subject: [PATCH 0478/1471] drm/i915: Set fake_vma.size as well as fake_vma.node.size for capture When capturing the bo, we allocate an error object with an array of min(vma->size, vma->node.size) pages, plus a bit for compression overhead. However, when creating the fake vma to describe the bo, only one of the sizes was filled in, resulting in a too small array. Through my and CI testing, this was sufficient for the mostly empty NULL context as it compressed well (or the out-of-bounds access simply didn't cause an issue). However, in real workloads on Cannonlake, we were overflowing that array and causing havoc with the random memory corruption. Reported-by: Rafael Antognolli Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103964 Fixes: 4e90a6e22272 ("drm/i915: Record default HW state in the GPU error state") Signed-off-by: Chris Wilson Cc: Chris Wilson Cc: Mika Kuoppala Cc: Joonas Lahtinen Tested-by: Rodrigo Vivi Link: https://patchwork.freedesktop.org/patch/msgid/20171201001536.13941-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/i915_gpu_error.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 876be8f1d9306..48418fb810660 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -1424,6 +1424,7 @@ capture_object(struct drm_i915_private *dev_priv, if (obj && i915_gem_object_has_pages(obj)) { struct i915_vma fake = { .node = { .start = U64_MAX, .size = obj->base.size }, + .size = obj->base.size, .pages = obj->mm.pages, .obj = obj, }; -- GitLab From b37aef5f1b5a4788eca662de6cb33e1521a7e404 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 30 Nov 2017 09:42:31 +0000 Subject: [PATCH 0479/1471] drm/i915/selftests: Wake the device before executing requests on the GPU To execute a requests requires us to have first woken the device, using the rpm wakeref (as the request needs to write to hardware to setup the context/ppGTT and execute on the GPU). So call intel_runtime_pm_get() around queuing the request; the request itself will then carry a wakeref until completion. References: https://bugs.freedesktop.org/show_bug.cgi?id=103994 Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20171130094231.10405-1-chris@chris-wilson.co.uk Reviewed-by: Mika Kuoppala --- drivers/gpu/drm/i915/selftests/i915_gem_context.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/selftests/i915_gem_context.c index ec1eff739e012..56a803d11916e 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_context.c @@ -376,7 +376,9 @@ static int igt_ctx_exec(void *arg) } } + intel_runtime_pm_get(i915); err = gpu_fill(obj, ctx, engine, dw); + intel_runtime_pm_put(i915); if (err) { pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) in ctx %u [full-ppgtt? %s], err=%d\n", ndwords, dw, max_dwords(obj), -- GitLab From 2bade36cae87070e66261f4ec96740773e80d95d Mon Sep 17 00:00:00 2001 From: David Lechner Date: Sun, 19 Nov 2017 14:12:05 -0600 Subject: [PATCH 0480/1471] dt-bindings: Add vendor prefix for ilitek MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This adds the vendor prefix ilitek for ILI Technology Corporation (ILITEK). This prefix is already used several places and I will be adding more. Signed-off-by: David Lechner Acked-by: Rob Herring Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/1511122328-31133-2-git-send-email-david@lechnology.com --- Documentation/devicetree/bindings/vendor-prefixes.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/Documentation/devicetree/bindings/vendor-prefixes.txt b/Documentation/devicetree/bindings/vendor-prefixes.txt index 6cf1dc5bc77ed..cf41a33800711 100644 --- a/Documentation/devicetree/bindings/vendor-prefixes.txt +++ b/Documentation/devicetree/bindings/vendor-prefixes.txt @@ -150,6 +150,7 @@ i2se I2SE GmbH ibm International Business Machines (IBM) idt Integrated Device Technologies, Inc. ifi Ingenieurburo Fur Ic-Technologie (I/F/I) +ilitek ILI Technology Corporation (ILITEK) img Imagination Technologies Ltd. infineon Infineon Technologies inforce Inforce Computing -- GitLab From 3fd8d9e00f44b7057aa02a50b06e8320cc822526 Mon Sep 17 00:00:00 2001 From: David Lechner Date: Sun, 19 Nov 2017 14:12:06 -0600 Subject: [PATCH 0481/1471] dt-bindings: Add binding for Ilitek ILI9225 display panels MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This adds a new binding for display panels that use an Ilitek ILI9225 controller. The "ilitek,ili9225-2.2in-176x220" device listed has no identification markings whatsoever and an hour of googling turned up nothing, hence the use of the size and resolution in the name instead of a model name. An example of this nameless device can be found at: https://github.com/Nkawu/TFT_22_ILI9225 Signed-off-by: David Lechner Acked-by: Rob Herring Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/1511122328-31133-3-git-send-email-david@lechnology.com --- .../bindings/display/ilitek,ili9225.txt | 25 +++++++++++++++++++ 1 file changed, 25 insertions(+) create mode 100644 Documentation/devicetree/bindings/display/ilitek,ili9225.txt diff --git a/Documentation/devicetree/bindings/display/ilitek,ili9225.txt b/Documentation/devicetree/bindings/display/ilitek,ili9225.txt new file mode 100644 index 0000000000000..21607a541c332 --- /dev/null +++ b/Documentation/devicetree/bindings/display/ilitek,ili9225.txt @@ -0,0 +1,25 @@ +Ilitek ILI9225 display panels + +This binding is for display panels using an Ilitek ILI9225 controller in SPI +mode. + +Required properties: +- compatible: "ilitek,ili9225-2.2in-176x220" +- rs-gpios: Register select signal +- reset-gpios: Reset pin + +The node for this driver must be a child node of a SPI controller, hence +all mandatory properties described in ../spi/spi-bus.txt must be specified. + +Optional properties: +- rotation: panel rotation in degrees counter clockwise (0,90,180,270) + +Example: + display@0{ + compatible = "ilitek,ili9225-2.2in-176x220"; + reg = <0>; + spi-max-frequency = <12000000>; + rs-gpios = <&gpio0 9 GPIO_ACTIVE_HIGH>; + reset-gpios = <&gpio0 8 GPIO_ACTIVE_HIGH>; + rotation = <270>; + }; -- GitLab From 13deee8111ed600ecd1809b0bfbab232c82159d9 Mon Sep 17 00:00:00 2001 From: David Lechner Date: Sun, 19 Nov 2017 14:12:07 -0600 Subject: [PATCH 0482/1471] drm/tinydrm: export mipi_dbi_buf_copy and mipi_dbi_spi_cmd_max_speed MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This exports the mipi_dbi_buf_copy() and mipi_dbi_spi_cmd_max_speed() functions so that they can be shared with other drivers. Signed-off-by: David Lechner Reviewed-by: Noralf Trønnes Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/1511122328-31133-4-git-send-email-david@lechnology.com --- drivers/gpu/drm/tinydrm/mipi-dbi.c | 24 ++++++++++++++++++++---- include/drm/tinydrm/mipi-dbi.h | 4 +++- 2 files changed, 23 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/tinydrm/mipi-dbi.c b/drivers/gpu/drm/tinydrm/mipi-dbi.c index 347f9b226f267..aa6b6ce568912 100644 --- a/drivers/gpu/drm/tinydrm/mipi-dbi.c +++ b/drivers/gpu/drm/tinydrm/mipi-dbi.c @@ -154,8 +154,18 @@ int mipi_dbi_command_buf(struct mipi_dbi *mipi, u8 cmd, u8 *data, size_t len) } EXPORT_SYMBOL(mipi_dbi_command_buf); -static int mipi_dbi_buf_copy(void *dst, struct drm_framebuffer *fb, - struct drm_clip_rect *clip, bool swap) +/** + * mipi_dbi_buf_copy - Copy a framebuffer, transforming it if necessary + * @dst: The destination buffer + * @fb: The source framebuffer + * @clip: Clipping rectangle of the area to be copied + * @swap: When true, swap MSB/LSB of 16-bit values + * + * Returns: + * Zero on success, negative error code on failure. + */ +int mipi_dbi_buf_copy(void *dst, struct drm_framebuffer *fb, + struct drm_clip_rect *clip, bool swap) { struct drm_gem_cma_object *cma_obj = drm_fb_cma_get_gem_obj(fb, 0); struct dma_buf_attachment *import_attach = cma_obj->base.import_attach; @@ -192,6 +202,7 @@ static int mipi_dbi_buf_copy(void *dst, struct drm_framebuffer *fb, DMA_FROM_DEVICE); return ret; } +EXPORT_SYMBOL(mipi_dbi_buf_copy); static int mipi_dbi_fb_dirty(struct drm_framebuffer *fb, struct drm_file *file_priv, @@ -444,18 +455,23 @@ EXPORT_SYMBOL(mipi_dbi_display_is_on); #if IS_ENABLED(CONFIG_SPI) -/* +/** + * mipi_dbi_spi_cmd_max_speed - get the maximum SPI bus speed + * @spi: SPI device + * @len: The transfer buffer length. + * * Many controllers have a max speed of 10MHz, but can be pushed way beyond * that. Increase reliability by running pixel data at max speed and the rest * at 10MHz, preventing transfer glitches from messing up the init settings. */ -static u32 mipi_dbi_spi_cmd_max_speed(struct spi_device *spi, size_t len) +u32 mipi_dbi_spi_cmd_max_speed(struct spi_device *spi, size_t len) { if (len > 64) return 0; /* use default */ return min_t(u32, 10000000, spi->max_speed_hz); } +EXPORT_SYMBOL(mipi_dbi_spi_cmd_max_speed); /* * MIPI DBI Type C Option 1 diff --git a/include/drm/tinydrm/mipi-dbi.h b/include/drm/tinydrm/mipi-dbi.h index 83346ddb9dba5..5d0e82b36eaf1 100644 --- a/include/drm/tinydrm/mipi-dbi.h +++ b/include/drm/tinydrm/mipi-dbi.h @@ -72,10 +72,12 @@ void mipi_dbi_pipe_enable(struct drm_simple_display_pipe *pipe, void mipi_dbi_pipe_disable(struct drm_simple_display_pipe *pipe); void mipi_dbi_hw_reset(struct mipi_dbi *mipi); bool mipi_dbi_display_is_on(struct mipi_dbi *mipi); +u32 mipi_dbi_spi_cmd_max_speed(struct spi_device *spi, size_t len); int mipi_dbi_command_read(struct mipi_dbi *mipi, u8 cmd, u8 *val); int mipi_dbi_command_buf(struct mipi_dbi *mipi, u8 cmd, u8 *data, size_t len); - +int mipi_dbi_buf_copy(void *dst, struct drm_framebuffer *fb, + struct drm_clip_rect *clip, bool swap); /** * mipi_dbi_command - MIPI DCS command with optional parameter(s) * @mipi: MIPI structure -- GitLab From b57e8b7661e04690643031af276c7bfc5c969dc9 Mon Sep 17 00:00:00 2001 From: David Lechner Date: Sun, 19 Nov 2017 14:12:08 -0600 Subject: [PATCH 0483/1471] drm/tinydrm: add driver for ILI9225 panels MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This adds a new driver for display panels based on the Ilitek ILI9225 controller. This was developed for a no-name panel with a red PCB that is commonly marketed for Arduino. See . Signed-off-by: David Lechner Reviewed-by: Noralf Trønnes Signed-off-by: Noralf Trønnes Link: https://patchwork.freedesktop.org/patch/msgid/1511122328-31133-5-git-send-email-david@lechnology.com --- MAINTAINERS | 6 + drivers/gpu/drm/tinydrm/Kconfig | 10 + drivers/gpu/drm/tinydrm/Makefile | 1 + drivers/gpu/drm/tinydrm/ili9225.c | 468 ++++++++++++++++++++++++++++++ 4 files changed, 485 insertions(+) create mode 100644 drivers/gpu/drm/tinydrm/ili9225.c diff --git a/MAINTAINERS b/MAINTAINERS index 4f0756a4ff1aa..712fcf3c47934 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -4370,6 +4370,12 @@ T: git git://anongit.freedesktop.org/drm/drm-misc S: Maintained F: drivers/gpu/drm/tve200/ +DRM DRIVER FOR ILITEK ILI9225 PANELS +M: David Lechner +S: Maintained +F: drivers/gpu/drm/tinydrm/ili9225.c +F: Documentation/devicetree/bindings/display/ili9225.txt + DRM DRIVER FOR INTEL I810 VIDEO CARDS S: Orphan / Obsolete F: drivers/gpu/drm/i810/ diff --git a/drivers/gpu/drm/tinydrm/Kconfig b/drivers/gpu/drm/tinydrm/Kconfig index 2e790e7dced5d..90c5bd5ef81b2 100644 --- a/drivers/gpu/drm/tinydrm/Kconfig +++ b/drivers/gpu/drm/tinydrm/Kconfig @@ -12,6 +12,16 @@ menuconfig DRM_TINYDRM config TINYDRM_MIPI_DBI tristate +config TINYDRM_ILI9225 + tristate "DRM support for ILI9225 display panels" + depends on DRM_TINYDRM && SPI + select TINYDRM_MIPI_DBI + help + DRM driver for the following Ilitek ILI9225 panels: + * No-name 2.2" color screen module + + If M is selected the module will be called ili9225. + config TINYDRM_MI0283QT tristate "DRM support for MI0283QT" depends on DRM_TINYDRM && SPI diff --git a/drivers/gpu/drm/tinydrm/Makefile b/drivers/gpu/drm/tinydrm/Makefile index 0c184bd1bb595..8aeee532474f5 100644 --- a/drivers/gpu/drm/tinydrm/Makefile +++ b/drivers/gpu/drm/tinydrm/Makefile @@ -4,6 +4,7 @@ obj-$(CONFIG_DRM_TINYDRM) += core/ obj-$(CONFIG_TINYDRM_MIPI_DBI) += mipi-dbi.o # Displays +obj-$(CONFIG_TINYDRM_ILI9225) += ili9225.o obj-$(CONFIG_TINYDRM_MI0283QT) += mi0283qt.o obj-$(CONFIG_TINYDRM_REPAPER) += repaper.o obj-$(CONFIG_TINYDRM_ST7586) += st7586.o diff --git a/drivers/gpu/drm/tinydrm/ili9225.c b/drivers/gpu/drm/tinydrm/ili9225.c new file mode 100644 index 0000000000000..3b766a26aa612 --- /dev/null +++ b/drivers/gpu/drm/tinydrm/ili9225.c @@ -0,0 +1,468 @@ +/* + * DRM driver for Ilitek ILI9225 panels + * + * Copyright 2017 David Lechner + * + * Some code copied from mipi-dbi.c + * Copyright 2016 Noralf Trønnes + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + */ + +#include +#include +#include +#include +#include +#include +#include