1 files changed, 147 insertions, 180 deletions

diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
index f42b48b31927..0e5a77c3c2e2 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
@@ -48,6 +48,7 @@
 #include "smu_cmn.h"
 #include "mp/mp_13_0_6_offset.h"
 #include "mp/mp_13_0_6_sh_mask.h"
+#include "umc_v12_0.h"
 
 #undef MP1_Public
 #undef smnMP1_FIRMWARE_FLAGS
@@ -94,22 +95,11 @@ MODULE_FIRMWARE("amdgpu/smu_13_0_6.bin");
 #define PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE__SHIFT 0x5
 #define LINK_SPEED_MAX				4
 
-#define SMU_13_0_6_DSCLK_THRESHOLD 100
+#define SMU_13_0_6_DSCLK_THRESHOLD 140
 
 #define MCA_BANK_IPID(_ip, _hwid, _type) \
 	[AMDGPU_MCA_IP_##_ip] = { .hwid = _hwid, .mcatype = _type, }
 
-enum mca_reg_idx {
-	MCA_REG_IDX_CONTROL		= 0,
-	MCA_REG_IDX_STATUS		= 1,
-	MCA_REG_IDX_ADDR		= 2,
-	MCA_REG_IDX_MISC0		= 3,
-	MCA_REG_IDX_CONFIG		= 4,
-	MCA_REG_IDX_IPID		= 5,
-	MCA_REG_IDX_SYND		= 6,
-	MCA_REG_IDX_COUNT		= 16,
-};
-
 struct mca_bank_ipid {
 	enum amdgpu_mca_ip ip;
 	uint16_t hwid;
@@ -122,7 +112,9 @@ struct mca_ras_info {
 	int *err_code_array;
 	int err_code_count;
 	int (*get_err_count)(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
-			     enum amdgpu_mca_error_type type, int idx, uint32_t *count);
+			     enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count);
+	bool (*bank_is_valid)(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+			      enum amdgpu_mca_error_type type, struct mca_bank_entry *entry);
 };
 
 #define P2S_TABLE_ID_A 0x50325341
@@ -270,7 +262,7 @@ static int smu_v13_0_6_init_microcode(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t p2s_table_id = P2S_TABLE_ID_A;
 	int ret = 0, i, p2stable_count;
-	char ucode_prefix[30];
+	char ucode_prefix[15];
 	char fw_name[30];
 
 	/* No need to load P2S tables in IOV mode */
@@ -1462,7 +1454,7 @@ static int smu_v13_0_6_register_irq_handler(struct smu_context *smu)
 
 static int smu_v13_0_6_notify_unload(struct smu_context *smu)
 {
-	if (smu->smc_fw_version <= 0x553500)
+	if (amdgpu_in_reset(smu->adev))
 		return 0;
 
 	dev_dbg(smu->adev->dev, "Notify PMFW about driver unload");
@@ -2103,6 +2095,14 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 			smu_v13_0_6_get_current_pcie_link_speed(smu);
 		gpu_metrics->pcie_bandwidth_acc =
 				SMUQ10_ROUND(metrics->PcieBandwidthAcc[0]);
+		gpu_metrics->pcie_bandwidth_inst =
+				SMUQ10_ROUND(metrics->PcieBandwidth[0]);
+		gpu_metrics->pcie_l0_to_recov_count_acc =
+				metrics->PCIeL0ToRecoveryCountAcc;
+		gpu_metrics->pcie_replay_count_acc =
+				metrics->PCIenReplayAAcc;
+		gpu_metrics->pcie_replay_rover_count_acc =
+				metrics->PCIenReplayARolloverCountAcc;
 	}
 
 	gpu_metrics->system_clock_counter = ktime_get_boottime_ns();
@@ -2305,7 +2305,7 @@ static int smu_v13_0_6_post_init(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 
 	if (!amdgpu_sriov_vf(adev) && adev->ras_enabled)
-		return smu_v13_0_6_mca_set_debug_mode(smu, true);
+		return smu_v13_0_6_mca_set_debug_mode(smu, false);
 
 	return 0;
 }
@@ -2387,6 +2387,7 @@ static const struct mca_bank_ipid smu_v13_0_6_mca_ipid_table[AMDGPU_MCA_IP_COUNT
 	MCA_BANK_IPID(UMC, 0x96, 0x0),
 	MCA_BANK_IPID(SMU, 0x01, 0x1),
 	MCA_BANK_IPID(MP5, 0x01, 0x2),
+	MCA_BANK_IPID(PCS_XGMI, 0x50, 0x0),
 };
 
 static void mca_bank_entry_info_decode(struct mca_bank_entry *entry, struct mca_bank_info *info)
@@ -2448,53 +2449,60 @@ static int mca_get_mca_entry(struct amdgpu_device *adev, enum amdgpu_mca_error_t
 	return 0;
 }
 
-static int mca_decode_mca_ipid(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, int idx, int *ip)
+static int mca_decode_ipid_to_hwip(uint64_t val)
 {
 	const struct mca_bank_ipid *ipid;
-	uint64_t val;
 	uint16_t hwid, mcatype;
-	int i, ret;
-
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_IPID, &val);
-	if (ret)
-		return ret;
+	int i;
 
 	hwid = REG_GET_FIELD(val, MCMP1_IPIDT0, HardwareID);
 	mcatype = REG_GET_FIELD(val, MCMP1_IPIDT0, McaType);
 
-	if (hwid) {
-		for (i = 0; i < ARRAY_SIZE(smu_v13_0_6_mca_ipid_table); i++) {
-			ipid = &smu_v13_0_6_mca_ipid_table[i];
+	for (i = 0; i < ARRAY_SIZE(smu_v13_0_6_mca_ipid_table); i++) {
+		ipid = &smu_v13_0_6_mca_ipid_table[i];
 
-			if (!ipid->hwid)
-				continue;
+		if (!ipid->hwid)
+			continue;
 
-			if (ipid->hwid == hwid && ipid->mcatype == mcatype) {
-				*ip = i;
-				return 0;
-			}
-		}
+		if (ipid->hwid == hwid && ipid->mcatype == mcatype)
+			return i;
 	}
 
-	*ip = AMDGPU_MCA_IP_UNKNOW;
+	return AMDGPU_MCA_IP_UNKNOW;
+}
+
+static int mca_umc_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+				     enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count)
+{
+	uint64_t status0;
+
+	status0 = entry->regs[MCA_REG_IDX_STATUS];
+
+	if (!REG_GET_FIELD(status0, MCMP1_STATUST0, Val)) {
+		*count = 0;
+		return 0;
+	}
+
+	if (type == AMDGPU_MCA_ERROR_TYPE_UE && umc_v12_0_is_uncorrectable_error(status0))
+		*count = 1;
+	else if (type == AMDGPU_MCA_ERROR_TYPE_CE && umc_v12_0_is_correctable_error(status0))
+		*count = 1;
 
 	return 0;
 }
 
-static int mca_normal_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
-					enum amdgpu_mca_error_type type, int idx, uint32_t *count)
+static int mca_pcs_xgmi_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+					  enum amdgpu_mca_error_type type, struct mca_bank_entry *entry,
+					  uint32_t *count)
 {
-	uint64_t status0;
-	int ret;
+	u32 ext_error_code;
 
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_STATUS, &status0);
-	if (ret)
-		return ret;
+	ext_error_code = MCA_REG__STATUS__ERRORCODEEXT(entry->regs[MCA_REG_IDX_STATUS]);
 
-	if (REG_GET_FIELD(status0, MCMP1_STATUST0, Val))
+	if (type == AMDGPU_MCA_ERROR_TYPE_UE && ext_error_code == 0)
+		*count = 1;
+	else if (type == AMDGPU_MCA_ERROR_TYPE_CE && ext_error_code == 6)
 		*count = 1;
-	else
-		*count = 0;
 
 	return 0;
 }
@@ -2515,70 +2523,41 @@ static bool mca_smu_check_error_code(struct amdgpu_device *adev, const struct mc
 	return false;
 }
 
-static int mca_mp5_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
-				     enum amdgpu_mca_error_type type, int idx, uint32_t *count)
+static int mca_gfx_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+				     enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count)
 {
-	uint64_t status0 = 0, misc0 = 0;
-	uint32_t errcode;
-	int ret;
-
-	if (mca_ras->ip != AMDGPU_MCA_IP_MP5)
-		return -EINVAL;
-
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_STATUS, &status0);
-	if (ret)
-		return ret;
+	uint64_t status0, misc0;
 
+	status0 = entry->regs[MCA_REG_IDX_STATUS];
 	if (!REG_GET_FIELD(status0, MCMP1_STATUST0, Val)) {
 		*count = 0;
 		return 0;
 	}
 
-	errcode = REG_GET_FIELD(status0, MCMP1_STATUST0, ErrorCode);
-	if (!mca_smu_check_error_code(adev, mca_ras, errcode))
-		return 0;
-
 	if (type == AMDGPU_MCA_ERROR_TYPE_UE &&
 	    REG_GET_FIELD(status0, MCMP1_STATUST0, UC) == 1 &&
 	    REG_GET_FIELD(status0, MCMP1_STATUST0, PCC) == 1) {
-		if (count)
-			*count = 1;
+		*count = 1;
 		return 0;
-	}
-
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_MISC0, &misc0);
-	if (ret)
-		return ret;
-
-	if (count)
+	} else {
+		misc0 = entry->regs[MCA_REG_IDX_MISC0];
 		*count = REG_GET_FIELD(misc0, MCMP1_MISC0T0, ErrCnt);
+	}
 
 	return 0;
 }
 
 static int mca_smu_mca_get_err_count(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
-				     enum amdgpu_mca_error_type type, int idx, uint32_t *count)
+				     enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count)
 {
-	uint64_t status0 = 0, misc0 = 0;
-	uint32_t errcode;
-	int ret;
-
-	if (mca_ras->ip != AMDGPU_MCA_IP_SMU)
-		return -EINVAL;
-
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_STATUS, &status0);
-	if (ret)
-		return ret;
+	uint64_t status0, misc0;
 
+	status0 = entry->regs[MCA_REG_IDX_STATUS];
 	if (!REG_GET_FIELD(status0, MCMP1_STATUST0, Val)) {
 		*count = 0;
 		return 0;
 	}
 
-	errcode = REG_GET_FIELD(status0, MCMP1_STATUST0, ErrorCode);
-	if (!mca_smu_check_error_code(adev, mca_ras, errcode))
-		return 0;
-
 	if (type == AMDGPU_MCA_ERROR_TYPE_UE &&
 	    REG_GET_FIELD(status0, MCMP1_STATUST0, UC) == 1 &&
 	    REG_GET_FIELD(status0, MCMP1_STATUST0, PCC) == 1) {
@@ -2587,16 +2566,43 @@ static int mca_smu_mca_get_err_count(const struct mca_ras_info *mca_ras, struct
 		return 0;
 	}
 
-	ret = mca_bank_read_reg(adev, type, idx, MCA_REG_IDX_MISC0, &misc0);
-	if (ret)
-		return ret;
-
-	if (count)
-		*count = REG_GET_FIELD(misc0, MCMP1_MISC0T0, ErrCnt);
+	misc0 = entry->regs[MCA_REG_IDX_MISC0];
+	*count = REG_GET_FIELD(misc0, MCMP1_MISC0T0, ErrCnt);
 
 	return 0;
 }
 
+static bool mca_gfx_smu_bank_is_valid(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+				      enum amdgpu_mca_error_type type, struct mca_bank_entry *entry)
+{
+	uint32_t instlo;
+
+	instlo = REG_GET_FIELD(entry->regs[MCA_REG_IDX_IPID], MCMP1_IPIDT0, InstanceIdLo);
+	switch (instlo) {
+	case 0x36430400: /* SMNAID XCD 0 */
+	case 0x38430400: /* SMNAID XCD 1 */
+	case 0x40430400: /* SMNXCD XCD 0, NOTE: FIXME: fix this error later */
+		return true;
+	default:
+		return false;
+	}
+
+	return false;
+};
+
+static bool mca_smu_bank_is_valid(const struct mca_ras_info *mca_ras, struct amdgpu_device *adev,
+				  enum amdgpu_mca_error_type type, struct mca_bank_entry *entry)
+{
+	uint32_t errcode, instlo;
+
+	instlo = REG_GET_FIELD(entry->regs[MCA_REG_IDX_IPID], MCMP1_IPIDT0, InstanceIdLo);
+	if (instlo != 0x03b30400)
+		return false;
+
+	errcode = REG_GET_FIELD(entry->regs[MCA_REG_IDX_STATUS], MCMP1_STATUST0, ErrorCode);
+	return mca_smu_check_error_code(adev, mca_ras, errcode);
+}
+
 static int sdma_err_codes[] = { CODE_SDMA0, CODE_SDMA1, CODE_SDMA2, CODE_SDMA3 };
 static int mmhub_err_codes[] = {
 	CODE_DAGB0, CODE_DAGB0 + 1, CODE_DAGB0 + 2, CODE_DAGB0 + 3, CODE_DAGB0 + 4, /* DAGB0-4 */
@@ -2608,23 +2614,30 @@ static const struct mca_ras_info mca_ras_table[] = {
 	{
 		.blkid = AMDGPU_RAS_BLOCK__UMC,
 		.ip = AMDGPU_MCA_IP_UMC,
-		.get_err_count = mca_normal_mca_get_err_count,
+		.get_err_count = mca_umc_mca_get_err_count,
 	}, {
 		.blkid = AMDGPU_RAS_BLOCK__GFX,
-		.ip = AMDGPU_MCA_IP_MP5,
-		.get_err_count = mca_mp5_mca_get_err_count,
+		.ip = AMDGPU_MCA_IP_SMU,
+		.get_err_count = mca_gfx_mca_get_err_count,
+		.bank_is_valid = mca_gfx_smu_bank_is_valid,
 	}, {
 		.blkid = AMDGPU_RAS_BLOCK__SDMA,
 		.ip = AMDGPU_MCA_IP_SMU,
 		.err_code_array = sdma_err_codes,
 		.err_code_count = ARRAY_SIZE(sdma_err_codes),
 		.get_err_count = mca_smu_mca_get_err_count,
+		.bank_is_valid = mca_smu_bank_is_valid,
 	}, {
 		.blkid = AMDGPU_RAS_BLOCK__MMHUB,
 		.ip = AMDGPU_MCA_IP_SMU,
 		.err_code_array = mmhub_err_codes,
 		.err_code_count = ARRAY_SIZE(mmhub_err_codes),
 		.get_err_count = mca_smu_mca_get_err_count,
+		.bank_is_valid = mca_smu_bank_is_valid,
+	}, {
+		.blkid = AMDGPU_RAS_BLOCK__XGMI_WAFL,
+		.ip = AMDGPU_MCA_IP_PCS_XGMI,
+		.get_err_count = mca_pcs_xgmi_mca_get_err_count,
 	},
 };
 
@@ -2659,130 +2672,84 @@ static int mca_get_valid_mca_count(struct amdgpu_device *adev, enum amdgpu_mca_e
 }
 
 static bool mca_bank_is_valid(struct amdgpu_device *adev, const struct mca_ras_info *mca_ras,
-			      enum amdgpu_mca_error_type type, int idx)
+			      enum amdgpu_mca_error_type type, struct mca_bank_entry *entry)
 {
-	int ret, ip = AMDGPU_MCA_IP_UNKNOW;
-
-	ret = mca_decode_mca_ipid(adev, type, idx, &ip);
-	if (ret)
-		return false;
-
-	if (ip == AMDGPU_MCA_IP_UNKNOW)
+	if (mca_decode_ipid_to_hwip(entry->regs[MCA_REG_IDX_IPID]) != mca_ras->ip)
 		return false;
 
-	return ip == mca_ras->ip;
-}
+	if (mca_ras->bank_is_valid)
+		return mca_ras->bank_is_valid(mca_ras, adev, type, entry);
 
-static int mca_get_valid_mca_idx(struct amdgpu_device *adev, const struct mca_ras_info *mca_ras,
-				 enum amdgpu_mca_error_type type,
-				 uint32_t mca_cnt, int *idx_array, int idx_array_size)
-{
-	int i, idx_cnt = 0;
-
-	for (i = 0; i < mca_cnt; i++) {
-		if (!mca_bank_is_valid(adev, mca_ras, type, i))
-			continue;
-
-		if (idx_array) {
-			if (idx_cnt < idx_array_size)
-				idx_array[idx_cnt] = i;
-			else
-				return -EINVAL;
-		}
-
-		idx_cnt++;
-	}
-
-	return idx_cnt;
+	return true;
 }
 
-static int __mca_smu_get_error_count(struct amdgpu_device *adev, const struct mca_ras_info *mca_ras, enum amdgpu_mca_error_type type, uint32_t *count)
+static int __mca_smu_get_ras_mca_set(struct amdgpu_device *adev, const struct mca_ras_info *mca_ras,
+				     enum amdgpu_mca_error_type type, struct mca_bank_set *mca_set)
 {
-	uint32_t result, mca_cnt, total = 0;
-	int idx_array[16];
-	int i, ret, idx_cnt = 0;
+	struct mca_bank_entry entry;
+	uint32_t mca_cnt;
+	int i, ret;
 
 	ret = mca_get_valid_mca_count(adev, type, &mca_cnt);
 	if (ret)
 		return ret;
 
 	/* if valid mca bank count is 0, the driver can return 0 directly */
-	if (!mca_cnt) {
-		*count = 0;
+	if (!mca_cnt)
 		return 0;
-	}
 
-	if (!mca_ras->get_err_count)
-		return -EINVAL;
+	for (i = 0; i < mca_cnt; i++) {
+		memset(&entry, 0, sizeof(entry));
+		ret = mca_get_mca_entry(adev, type, i, &entry);
+		if (ret)
+			return ret;
 
-	idx_cnt = mca_get_valid_mca_idx(adev, mca_ras, type, mca_cnt, idx_array, ARRAY_SIZE(idx_array));
-	if (idx_cnt < 0)
-		return -EINVAL;
+		if (mca_ras && !mca_bank_is_valid(adev, mca_ras, type, &entry))
+			continue;
 
-	for (i = 0; i < idx_cnt; i++) {
-		result = 0;
-		ret = mca_ras->get_err_count(mca_ras, adev, type, idx_array[i], &result);
+		ret = amdgpu_mca_bank_set_add_entry(mca_set, &entry);
 		if (ret)
 			return ret;
-
-		total += result;
 	}
 
-	*count = total;
-
 	return 0;
 }
 
-static int mca_smu_get_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
-				   enum amdgpu_mca_error_type type, uint32_t *count)
+static int mca_smu_get_ras_mca_set(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+				   enum amdgpu_mca_error_type type, struct mca_bank_set *mca_set)
 {
-	const struct mca_ras_info *mca_ras;
+	const struct mca_ras_info *mca_ras = NULL;
 
-	if (!count)
+	if (!mca_set)
 		return -EINVAL;
 
-	mca_ras = mca_get_mca_ras_info(adev, blk);
-	if (!mca_ras)
-		return -EOPNOTSUPP;
-
-	return __mca_smu_get_error_count(adev, mca_ras, type, count);
-}
-
-static int __mca_smu_get_ras_mca_idx_array(struct amdgpu_device *adev, const struct mca_ras_info *mca_ras,
-					   enum amdgpu_mca_error_type type, int *idx_array, int *idx_array_size)
-{
-	uint32_t mca_cnt = 0;
-	int ret, idx_cnt = 0;
-
-	ret = mca_get_valid_mca_count(adev, type, &mca_cnt);
-	if (ret)
-		return ret;
-
-	/* if valid mca bank count is 0, the driver can return 0 directly */
-	if (!mca_cnt) {
-		*idx_array_size = 0;
-		return 0;
+	if (blk != AMDGPU_RAS_BLOCK_COUNT) {
+		mca_ras = mca_get_mca_ras_info(adev, blk);
+		if (!mca_ras)
+			return -EOPNOTSUPP;
 	}
 
-	idx_cnt = mca_get_valid_mca_idx(adev, mca_ras, type, mca_cnt, idx_array, *idx_array_size);
-	if (idx_cnt < 0)
-		return -EINVAL;
-
-	*idx_array_size = idx_cnt;
-
-	return 0;
+	return __mca_smu_get_ras_mca_set(adev, mca_ras, type, mca_set);
 }
 
-static int mca_smu_get_ras_mca_idx_array(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
-					 enum amdgpu_mca_error_type type, int *idx_array, int *idx_array_size)
+static int mca_smu_parse_mca_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+					 struct mca_bank_entry *entry, uint32_t *count)
 {
 	const struct mca_ras_info *mca_ras;
 
+	if (!entry || !count)
+		return -EINVAL;
+
 	mca_ras = mca_get_mca_ras_info(adev, blk);
 	if (!mca_ras)
 		return -EOPNOTSUPP;
 
-	return __mca_smu_get_ras_mca_idx_array(adev, mca_ras, type, idx_array, idx_array_size);
+	if (!mca_bank_is_valid(adev, mca_ras, type, entry)) {
+		*count = 0;
+		return 0;
+	}
+
+	return mca_ras->get_err_count(mca_ras, adev, type, entry, count);
 }
 
 static int mca_smu_get_mca_entry(struct amdgpu_device *adev,
@@ -2801,10 +2768,10 @@ static const struct amdgpu_mca_smu_funcs smu_v13_0_6_mca_smu_funcs = {
 	.max_ue_count = 12,
 	.max_ce_count = 12,
 	.mca_set_debug_mode = mca_smu_set_debug_mode,
-	.mca_get_error_count = mca_smu_get_error_count,
+	.mca_get_ras_mca_set = mca_smu_get_ras_mca_set,
+	.mca_parse_mca_error_count = mca_smu_parse_mca_error_count,
 	.mca_get_mca_entry = mca_smu_get_mca_entry,
 	.mca_get_valid_mca_count = mca_smu_get_valid_mca_count,
-	.mca_get_ras_mca_idx_array = mca_smu_get_ras_mca_idx_array,
 };
 
 static int smu_v13_0_6_select_xgmi_plpd_policy(struct smu_context *smu,