Commit af327f75 authored by Jie Wang's avatar Jie Wang Committed by Jiantao Xiao
Browse files

net: hns3: add vf fault detect support 

mainline inclusion
from mainline-v6.7-rc1
commit 8a45c4f9e15902e5c52d5c42cda6239bc473c7c8
category: feature
bugzilla: https://gitee.com/openeuler/kernel/issues/I8E9UJ
CVE: NA

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8a45c4f9e15902e5c52d5c42cda6239bc473c7c8



----------------------------------------------------------------------

Currently hns3 driver supports vf fault detect feature. Several ras caused
by VF resources don't need to do PF function reset for recovery. The driver
only needs to reset the specified VF.

So this patch adds process in ras module. New process will get detailed
information about ras and do the most correct measures based on these
accurate information.

Signed-off-by: default avatarJie Wang <wangjie125@huawei.com>
Signed-off-by: default avatarJijie Shao <shaojijie@huawei.com>
Link: https://lore.kernel.org/r/20231007031215.1067758-3-shaojijie@huawei.com


Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
Signed-off-by: default avatarJiantao Xiao <xiaojiantao1@h-partners.com>
parent 29c78b46

drivers/net/ethernet/hisilicon/hns3/hnae3.h

+1 −0
Original line number Diff line number Diff line
@@ -324,6 +324,7 @@ enum hnae3_reset_type { 
	HNAE3_GLOBAL_RESET,

	HNAE3_IMP_RESET,

	HNAE3_NONE_RESET,

	HNAE3_VF_EXP_RESET,

	HNAE3_MAX_RESET,

};

drivers/net/ethernet/hisilicon/hns3/hns3_common/hclge_comm_cmd.h

+1 −0
Original line number Diff line number Diff line
@@ -94,6 +94,7 @@ enum hclge_opcode_type { 
	HCLGE_OPC_DFX_GEN_REG		= 0x7038,



	HCLGE_OPC_QUERY_DEV_SPECS	= 0x0050,

	HCLGE_OPC_GET_QUEUE_ERR_VF      = 0x0067,



	/* MAC command */

	HCLGE_OPC_CONFIG_MAC_MODE	= 0x0301,

drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_err.c

+111 −5
Original line number Diff line number Diff line
@@ -1737,10 +1737,12 @@ static const struct hclge_hw_type_id hclge_hw_type_id_st[] = { 
		.msg = "tqp_int_ecc_error"

	}, {

		.type_id = PF_ABNORMAL_INT_ERROR,

		.msg = "pf_abnormal_int_error"

		.msg = "pf_abnormal_int_error",

		.cause_by_vf = true

	}, {

		.type_id = MPF_ABNORMAL_INT_ERROR,

		.msg = "mpf_abnormal_int_error"

		.msg = "mpf_abnormal_int_error",

		.cause_by_vf = true

	}, {

		.type_id = COMMON_ERROR,

		.msg = "common_error"
@@ -3200,7 +3202,7 @@ void hclge_handle_occurred_error(struct hclge_dev *hdev) 
		hclge_handle_error_info_log(ae_dev);

}



static void

static bool

hclge_handle_error_type_reg_log(struct hclge_dev *hdev,

				struct hclge_mod_err_info *mod_info,

				struct hclge_type_reg_err_info *type_reg_info)
@@ -3212,6 +3214,7 @@ hclge_handle_error_type_reg_log(struct hclge_dev *hdev, 
	struct device *dev = &hdev->pdev->dev;

	u8 index_module = MODULE_NONE;

	u8 index_type = NONE_ERROR;

	bool cause_by_vf = false;



	mod_id = mod_info->mod_id;

	type_id = type_reg_info->type_id & HCLGE_ERR_TYPE_MASK;
@@ -3230,6 +3233,7 @@ hclge_handle_error_type_reg_log(struct hclge_dev *hdev, 
	for (i = 0; i < total_type; i++) {

		if (type_id == hclge_hw_type_id_st[i].type_id) {

			index_type = i;

			cause_by_vf = hclge_hw_type_id_st[i].cause_by_vf;

			break;

		}

	}
@@ -3250,6 +3254,8 @@ hclge_handle_error_type_reg_log(struct hclge_dev *hdev, 


	if (hclge_hw_module_id_st[index_module].query_reg_info)

		hclge_hw_module_id_st[index_module].query_reg_info(hdev);



	return cause_by_vf;

}



static void hclge_handle_error_module_log(struct hnae3_ae_dev *ae_dev,
@@ -3260,6 +3266,7 @@ static void hclge_handle_error_module_log(struct hnae3_ae_dev *ae_dev, 
	struct device *dev = &hdev->pdev->dev;

	struct hclge_mod_err_info *mod_info;

	struct hclge_sum_err_info *sum_info;

	bool cause_by_vf = false;

	u8 mod_num, err_num, i;

	u32 offset = 0;
@@ -3288,12 +3295,16 @@ static void hclge_handle_error_module_log(struct hnae3_ae_dev *ae_dev, 


			type_reg_info = (struct hclge_type_reg_err_info *)

					    &buf[offset++];

			hclge_handle_error_type_reg_log(dev, mod_info,

							type_reg_info);

			if (hclge_handle_error_type_reg_log(hdev, mod_info,

							    type_reg_info))

				cause_by_vf = true;



			offset += type_reg_info->reg_num;

		}

	}



	if (hnae3_ae_dev_vf_fault_supported(hdev->ae_dev) && cause_by_vf)

		set_bit(HNAE3_VF_EXP_RESET, &ae_dev->hw_err_reset_req);

}



static int hclge_query_all_err_bd_num(struct hclge_dev *hdev, u32 *bd_num)
@@ -3385,3 +3396,98 @@ int hclge_handle_error_info_log(struct hnae3_ae_dev *ae_dev) 
out:

	return ret;

}



static bool hclge_reset_vf_in_bitmap(struct hclge_dev *hdev,

				     unsigned long *bitmap)

{

	struct hclge_vport *vport;

	bool exist_set = false;

	int func_id;

	int ret;



	func_id = find_first_bit(bitmap, HCLGE_VPORT_NUM);

	if (func_id == PF_VPORT_ID)

		return false;



	while (func_id != HCLGE_VPORT_NUM) {

		vport = hclge_get_vf_vport(hdev,

					   func_id - HCLGE_VF_VPORT_START_NUM);

		if (!vport) {

			dev_err(&hdev->pdev->dev, "invalid func id(%d)\n",

				func_id);

			return false;

		}



		dev_info(&hdev->pdev->dev, "do function %d recovery.", func_id);



		ret = hclge_reset_tqp(&vport->nic);

		if (ret) {

			dev_err(&hdev->pdev->dev,

				"failed to reset tqp, ret = %d.", ret);

			return false;

		}



		ret = hclge_inform_vf_reset(vport, HNAE3_VF_FUNC_RESET);

		if (ret) {

			dev_err(&hdev->pdev->dev,

				"failed to reset func %d, ret = %d.",

				func_id, ret);

			return false;

		}



		exist_set = true;

		clear_bit(func_id, bitmap);

		func_id = find_first_bit(bitmap, HCLGE_VPORT_NUM);

	}



	return exist_set;

}



static void hclge_get_vf_fault_bitmap(struct hclge_desc *desc,

				      unsigned long *bitmap)

{

#define HCLGE_FIR_FAULT_BYTES	24

#define HCLGE_SEC_FAULT_BYTES	8



	u8 *buff;



	BUILD_BUG_ON(HCLGE_FIR_FAULT_BYTES + HCLGE_SEC_FAULT_BYTES !=

		     BITS_TO_BYTES(HCLGE_VPORT_NUM));



	memcpy(bitmap, desc[0].data, HCLGE_FIR_FAULT_BYTES);

	buff = (u8 *)bitmap + HCLGE_FIR_FAULT_BYTES;

	memcpy(buff, desc[1].data, HCLGE_SEC_FAULT_BYTES);

}



int hclge_handle_vf_queue_err_ras(struct hclge_dev *hdev)

{

	unsigned long vf_fault_bitmap[BITS_TO_LONGS(HCLGE_VPORT_NUM)];

	struct hclge_desc desc[2];

	bool cause_by_vf = false;

	int ret;



	if (!test_and_clear_bit(HNAE3_VF_EXP_RESET,

				&hdev->ae_dev->hw_err_reset_req) ||

	    !hnae3_ae_dev_vf_fault_supported(hdev->ae_dev))

		return 0;



	hclge_comm_cmd_setup_basic_desc(&desc[0], HCLGE_OPC_GET_QUEUE_ERR_VF,

					true);

	desc[0].flag |= cpu_to_le16(HCLGE_COMM_CMD_FLAG_NEXT);

	hclge_comm_cmd_setup_basic_desc(&desc[1], HCLGE_OPC_GET_QUEUE_ERR_VF,

					true);



	ret = hclge_comm_cmd_send(&hdev->hw.hw, desc, 2);

	if (ret) {

		dev_err(&hdev->pdev->dev,

			"failed to get vf bitmap, ret = %d.\n", ret);

		return ret;

	}

	hclge_get_vf_fault_bitmap(desc, vf_fault_bitmap);



	cause_by_vf = hclge_reset_vf_in_bitmap(hdev, vf_fault_bitmap);

	if (cause_by_vf)

		hdev->ae_dev->hw_err_reset_req = 0;



	return 0;

}

drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_err.h

+2 −0
Original line number Diff line number Diff line
@@ -220,6 +220,7 @@ struct hclge_hw_module_id { 
struct hclge_hw_type_id {

	enum hclge_err_type_list type_id;

	const char *msg;

	bool cause_by_vf; /* indicate the error may from vf exception */

};



struct hclge_sum_err_info {
@@ -274,4 +275,5 @@ int hclge_handle_hw_msix_error(struct hclge_dev *hdev, 
			       unsigned long *reset_requests);

int hclge_handle_error_info_log(struct hnae3_ae_dev *ae_dev);

int hclge_handle_mac_tnl(struct hclge_dev *hdev);

int hclge_handle_vf_queue_err_ras(struct hclge_dev *hdev);

#endif

drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c

+2 −1
Original line number Diff line number Diff line
@@ -3391,7 +3391,7 @@ static int hclge_get_status(struct hnae3_handle *handle) 
	return hdev->hw.mac.link;

}



static struct hclge_vport *hclge_get_vf_vport(struct hclge_dev *hdev, int vf)

struct hclge_vport *hclge_get_vf_vport(struct hclge_dev *hdev, int vf)

{

	if (!pci_num_vf(hdev->pdev)) {

		dev_err(&hdev->pdev->dev,
@@ -4515,6 +4515,7 @@ static void hclge_handle_err_recovery(struct hclge_dev *hdev) 
	if (hclge_find_error_source(hdev)) {

		hclge_handle_error_info_log(ae_dev);

		hclge_handle_mac_tnl(hdev);

		hclge_handle_vf_queue_err_ras(hdev);

	}



	hclge_handle_err_reset_request(hdev);