drm/amdgpu: Parse all deferred errors with UMC aca handle

We should only increase the deferred errors in UMC block.

Signed-off-by: Xiang Liu <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

Author: Xiang Liu

drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c

@@ -195,6 +195,10 @@ static bool aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
 {
 	const struct aca_bank_ops *bank_ops = handle->bank_ops;
 
+	/* Parse all deferred errors with UMC aca handle */
+	if (ACA_BANK_ERR_IS_DEFFERED(bank))
+		return handle->hwip == ACA_HWIP_TYPE_UMC;
+
 	if (!aca_bank_hwip_is_matched(bank, handle->hwip))
 		return false;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h

@@ -80,14 +80,6 @@ struct ras_query_context;
 	(ACA_REG__STATUS__POISON((bank)->regs[ACA_REG_IDX_STATUS]) || \
 	 ACA_REG__STATUS__DEFERRED((bank)->regs[ACA_REG_IDX_STATUS]))
 
-#define ACA_BANK_ERR_CE_DE_DECODE(bank)                             \
-	(ACA_BANK_ERR_IS_DEFFERED(bank) ? ACA_ERROR_TYPE_DEFERRED : \
-					  ACA_ERROR_TYPE_CE)
-
-#define ACA_BANK_ERR_UE_DE_DECODE(bank)                             \
-	(ACA_BANK_ERR_IS_DEFFERED(bank) ? ACA_ERROR_TYPE_DEFERRED : \
-					  ACA_ERROR_TYPE_UE)
-
 enum aca_reg_idx {
 	ACA_REG_IDX_CTL			= 0,
 	ACA_REG_IDX_STATUS		= 1,

drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c

@@ -1172,7 +1172,7 @@ static int xgmi_v6_4_0_aca_bank_parser(struct aca_handle *handle, struct aca_ban
 		break;
 	case ACA_SMU_TYPE_CE:
 		count = ext_error_code == 6 ? count : 0ULL;
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, count);
 		break;
 	default:

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c

@@ -867,15 +867,13 @@ static int gfx_v9_4_3_aca_bank_parser(struct aca_handle *handle,
 
 	switch (type) {
 	case ACA_SMU_TYPE_UE:
-		bank->aca_err_type = ACA_BANK_ERR_UE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, 1ULL);
 		break;
 	case ACA_SMU_TYPE_CE:
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
-			(bank->aca_err_type == ACA_ERROR_TYPE_CE) ?
-				ACA_REG__MISC0__ERRCNT(misc0) :
-				1);
+						     ACA_REG__MISC0__ERRCNT(misc0));
 		break;
 	default:
 		return -EINVAL;

drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c

@@ -1328,7 +1328,7 @@ static int jpeg_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_ban
 						     1ULL);
 		break;
 	case ACA_SMU_TYPE_CE:
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
 						     ACA_REG__MISC0__ERRCNT(misc0));
 		break;

drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c

@@ -751,7 +751,7 @@ static int mmhub_v1_8_aca_bank_parser(struct aca_handle *handle, struct aca_bank
 						     1ULL);
 		break;
 	case ACA_SMU_TYPE_CE:
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
 						     ACA_REG__MISC0__ERRCNT(misc0));
 		break;

drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c

@@ -2595,7 +2595,7 @@ static int sdma_v4_4_2_aca_bank_parser(struct aca_handle *handle, struct aca_ban
 						     1ULL);
 		break;
 	case ACA_SMU_TYPE_CE:
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
 						     ACA_REG__MISC0__ERRCNT(misc0));
 		break;

drivers/gpu/drm/amd/amdgpu/umc_v12_0.c

@@ -85,7 +85,8 @@ bool umc_v12_0_is_deferred_error(struct amdgpu_device *adev, uint64_t mc_umc_sta
 
 	return (amdgpu_ras_is_poison_mode_supported(adev) &&
 		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) &&
-		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1));
+		((REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1) ||
+		(REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Poison) == 1)));
 }
 
 bool umc_v12_0_is_uncorrectable_error(struct amdgpu_device *adev, uint64_t mc_umc_status)

drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c

@@ -1965,7 +1965,7 @@ static int vcn_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank
 						     1ULL);
 		break;
 	case ACA_SMU_TYPE_CE:
-		bank->aca_err_type = ACA_BANK_ERR_CE_DE_DECODE(bank);
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
 		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
 						     ACA_REG__MISC0__ERRCNT(misc0));
 		break;