From 94ac2ab3fff96814d7460a27a0e9d004abbd4128 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 19 Jun 2024 23:00:37 +0200 Subject: Merging upstream version 6.9.2. Signed-off-by: Daniel Baumann --- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 81 +++++++++++++++++++++++++++----- 1 file changed, 70 insertions(+), 11 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index a6c88f2fe6..20d51f6c9b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -1035,15 +1035,74 @@ int amdgpu_xgmi_remove_device(struct amdgpu_device *adev) return 0; } +static int xgmi_v6_4_0_aca_bank_generate_report(struct aca_handle *handle, struct aca_bank *bank, enum aca_error_type type, + struct aca_bank_report *report, void *data) +{ + struct amdgpu_device *adev = handle->adev; + const char *error_str; + u64 status; + int ret, ext_error_code; + + ret = aca_bank_info_decode(bank, &report->info); + if (ret) + return ret; + + status = bank->regs[ACA_REG_IDX_STATUS]; + ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status); + + error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ? + xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL; + if (error_str) + dev_info(adev->dev, "%s detected\n", error_str); + + if ((type == ACA_ERROR_TYPE_UE && ext_error_code == 0) || + (type == ACA_ERROR_TYPE_CE && ext_error_code == 6)) + report->count[type] = ACA_REG__MISC0__ERRCNT(bank->regs[ACA_REG_IDX_MISC0]); + + return 0; +} + +static const struct aca_bank_ops xgmi_v6_4_0_aca_bank_ops = { + .aca_bank_generate_report = xgmi_v6_4_0_aca_bank_generate_report, +}; + +static const struct aca_info xgmi_v6_4_0_aca_info = { + .hwip = ACA_HWIP_TYPE_PCS_XGMI, + .mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK, + .bank_ops = &xgmi_v6_4_0_aca_bank_ops, +}; + static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { + int r; + if (!adev->gmc.xgmi.supported || adev->gmc.xgmi.num_physical_nodes == 0) return 0; amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL); - return amdgpu_ras_block_late_init(adev, ras_block); + r = amdgpu_ras_block_late_init(adev, ras_block); + if (r) + return r; + + switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) { + case IP_VERSION(6, 4, 0): + r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL, + &xgmi_v6_4_0_aca_info, NULL); + if (r) + goto late_fini; + break; + default: + break; + } + + return 0; + +late_fini: + amdgpu_ras_block_late_fini(adev, ras_block); + + return r; } uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev, @@ -1099,7 +1158,7 @@ static void amdgpu_xgmi_legacy_reset_ras_error_count(struct amdgpu_device *adev) static void __xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst, u64 mca_base) { - WREG64_MCA(xgmi_inst, mca_base, MCA_REG_IDX_STATUS, 0ULL); + WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL); } static void xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst) @@ -1277,12 +1336,12 @@ static void amdgpu_xgmi_legacy_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += ce_cnt; } -static enum amdgpu_mca_error_type xgmi_v6_4_0_pcs_mca_get_error_type(struct amdgpu_device *adev, u64 status) +static enum aca_error_type xgmi_v6_4_0_pcs_mca_get_error_type(struct amdgpu_device *adev, u64 status) { const char *error_str; int ext_error_code; - ext_error_code = MCA_REG__STATUS__ERRORCODEEXT(status); + ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status); error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ? xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL; @@ -1291,9 +1350,9 @@ static enum amdgpu_mca_error_type xgmi_v6_4_0_pcs_mca_get_error_type(struct amdg switch (ext_error_code) { case 0: - return AMDGPU_MCA_ERROR_TYPE_UE; + return ACA_ERROR_TYPE_UE; case 6: - return AMDGPU_MCA_ERROR_TYPE_CE; + return ACA_ERROR_TYPE_CE; default: return -EINVAL; } @@ -1307,22 +1366,22 @@ static void __xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, struct a int xgmi_inst = mcm_info->die_id; u64 status = 0; - status = RREG64_MCA(xgmi_inst, mca_base, MCA_REG_IDX_STATUS); - if (!MCA_REG__STATUS__VAL(status)) + status = RREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS); + if (!ACA_REG__STATUS__VAL(status)) return; switch (xgmi_v6_4_0_pcs_mca_get_error_type(adev, status)) { - case AMDGPU_MCA_ERROR_TYPE_UE: + case ACA_ERROR_TYPE_UE: amdgpu_ras_error_statistic_ue_count(err_data, mcm_info, NULL, 1ULL); break; - case AMDGPU_MCA_ERROR_TYPE_CE: + case ACA_ERROR_TYPE_CE: amdgpu_ras_error_statistic_ce_count(err_data, mcm_info, NULL, 1ULL); break; default: break; } - WREG64_MCA(xgmi_inst, mca_base, MCA_REG_IDX_STATUS, 0ULL); + WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL); } static void xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, int xgmi_inst, struct ras_err_data *err_data) -- cgit v1.2.3