Commit a259cc14 authored by Thomas Hellström's avatar Thomas Hellström
Browse files

drm/i915: Reduce the number of objects subject to memcpy recover 



We really only need memcpy restore for objects that affect the
operability of the migrate context. That is, primarily the page-table
objects of the migrate VM.

Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early
restores using memcpy and a way to assign LMEM page-table object flags
to be used by the vms.

Restore objects without this flag with the gpu blitter and only objects
carrying the flag using TTM memcpy.

Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and
defer for a later audit which vms actually need it. Most importantly, user-
allocated vms with pinned page-table objects can be restored using the
blitter.

Performance-wise memcpy restore is probably as fast as gpu restore if not
faster, but using gpu restore will help tackling future restrictions in
mappable LMEM size.

v4:
- Don't mark the aliasing ppgtt page table flags for early resume, but
  rather the ggtt page table flags as intended. (Matthew Auld)
- The check for user buffer objects during early resume is pointless, since
  they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld)
v5:
- Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored
  before we fire up the migrate context.

Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: default avatarThomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: default avatarMatthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
parent 0d8ee5ba

drivers/gpu/drm/i915/gem/i915_gem_context.c

+2 −2
Original line number Diff line number Diff line
@@ -1287,7 +1287,7 @@ i915_gem_create_context(struct drm_i915_private *i915, 
	} else if (HAS_FULL_PPGTT(i915)) {

		struct i915_ppgtt *ppgtt;



		ppgtt = i915_ppgtt_create(&i915->gt);

		ppgtt = i915_ppgtt_create(&i915->gt, 0);

		if (IS_ERR(ppgtt)) {

			drm_dbg(&i915->drm, "PPGTT setup failed (%ld)\n",

				PTR_ERR(ppgtt));
@@ -1465,7 +1465,7 @@ int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data, 
	if (args->flags)

		return -EINVAL;



	ppgtt = i915_ppgtt_create(&i915->gt);

	ppgtt = i915_ppgtt_create(&i915->gt, 0);

	if (IS_ERR(ppgtt))

		return PTR_ERR(ppgtt);

drivers/gpu/drm/i915/gem/i915_gem_object_types.h

+6 −3
Original line number Diff line number Diff line
@@ -294,13 +294,16 @@ struct drm_i915_gem_object { 
#define I915_BO_ALLOC_USER        BIT(3)

/* Object is allowed to lose its contents on suspend / resume, even if pinned */

#define I915_BO_ALLOC_PM_VOLATILE BIT(4)

/* Object needs to be restored early using memcpy during resume */

#define I915_BO_ALLOC_PM_EARLY    BIT(5)

#define I915_BO_ALLOC_FLAGS (I915_BO_ALLOC_CONTIGUOUS | \

			     I915_BO_ALLOC_VOLATILE | \

			     I915_BO_ALLOC_CPU_CLEAR | \

			     I915_BO_ALLOC_USER | \

			     I915_BO_ALLOC_PM_VOLATILE)

#define I915_BO_READONLY          BIT(5)

#define I915_TILING_QUIRK_BIT     6 /* unknown swizzling; do not release! */

			     I915_BO_ALLOC_PM_VOLATILE | \

			     I915_BO_ALLOC_PM_EARLY)

#define I915_BO_READONLY          BIT(6)

#define I915_TILING_QUIRK_BIT     7 /* unknown swizzling; do not release! */



	/**

	 * @mem_flags - Mutable placement-related flags

drivers/gpu/drm/i915/gem/i915_gem_pm.c

+5 −1
Original line number Diff line number Diff line
@@ -97,8 +97,12 @@ int i915_gem_backup_suspend(struct drm_i915_private *i915) 
	 * More objects may have become unpinned as requests were

	 * retired. Now try to evict again. The gt may be wedged here

	 * in which case we automatically fall back to memcpy.

	 * We allow also backing up pinned objects that have not been

	 * marked for early recover, and that may contain, for example,

	 * page-tables for the migrate context.

	 */

	ret = lmem_suspend(i915, I915_TTM_BACKUP_ALLOW_GPU);

	ret = lmem_suspend(i915, I915_TTM_BACKUP_ALLOW_GPU |

			   I915_TTM_BACKUP_PINNED);

	if (ret)

		goto out_recover;

drivers/gpu/drm/i915/gem/i915_gem_ttm_pm.c

+3 −2
Original line number Diff line number Diff line
@@ -57,7 +57,8 @@ static int i915_ttm_backup(struct i915_gem_apply_to_region *apply, 
	if (pm_apply->allow_gpu && i915_gem_object_evictable(obj))

		return ttm_bo_validate(bo, i915_ttm_sys_placement(), &ctx);



	if (!pm_apply->backup_pinned)

	if (!pm_apply->backup_pinned ||

	    (pm_apply->allow_gpu && (obj->flags & I915_BO_ALLOC_PM_EARLY)))

		return 0;



	if (obj->flags & I915_BO_ALLOC_PM_VOLATILE)
@@ -155,7 +156,7 @@ static int i915_ttm_restore(struct i915_gem_apply_to_region *apply, 
	if (!backup)

		return 0;



	if (!pm_apply->allow_gpu && (obj->flags & I915_BO_ALLOC_USER))

	if (!pm_apply->allow_gpu && !(obj->flags & I915_BO_ALLOC_PM_EARLY))

		return 0;



	err = i915_gem_object_lock(backup, apply->ww);

drivers/gpu/drm/i915/gem/selftests/huge_pages.c

+1 −1
Original line number Diff line number Diff line
@@ -1645,7 +1645,7 @@ int i915_gem_huge_page_mock_selftests(void) 
	mkwrite_device_info(dev_priv)->ppgtt_type = INTEL_PPGTT_FULL;

	mkwrite_device_info(dev_priv)->ppgtt_size = 48;



	ppgtt = i915_ppgtt_create(&dev_priv->gt);

	ppgtt = i915_ppgtt_create(&dev_priv->gt, 0);

	if (IS_ERR(ppgtt)) {

		err = PTR_ERR(ppgtt);

		goto out_unlock;