Generate CPER records for RAS and commit to CPER ring
Subsystems (3)
v2 cover Generate CPER records for RAS and commit to CPER ring
This patch series generate RAS CPER records for UE/DE/CE/BP threshold exceed event. SMU_TYPE_CE banks are combined into 1 CPER entry, they could be CEs or DEs or both. UEs and BPs are encoded into separate CPER entries.
RAS CPER records for CEs will be generated only after CEs count been queried.
All records are committed to a pure software ring with a limit size, new records will flush older records when overflow happened. User can access the records by reading debugfs node, which is read-only.
Hawking Zhang (5):
drm/amd/include: Add amd cper header drm/amdgpu: Introduce funcs for populating CPER drm/amdgpu: Include ACA error type in aca bank drm/amdgpu: Introduce funcs for generating cper record drm/amdgpu: Generate cper records
Tao Zhou (4):
drm/amdgpu: add RAS CPER ring buffer drm/amdgpu: read CPER ring via debugfs drm/amdgpu: add data write function for CPER ring drm/amdgpu: add mutex lock for cper ring
Xiang Liu (3):
drm/amdgpu: Get timestamp from system time drm/amdgpu: Commit CPER entry
drm/amdgpu: Generate bad page threshold cper records
drivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 46 +- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 16 +- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 559 +++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 104 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 91 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 2 + drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c | 3 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 2 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 2 + drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 + drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 + drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 2 + drivers/gpu/drm/amd/include/amd_cper.h | 269 ++++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 3 + 19 files changed, 1075 insertions(+), 40 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h create mode 100644 drivers/gpu/drm/amd/include/amd_cper.h
Signature · 1 line
2.34.1
PATCH v2 1/12 drm/amd/include: Add amd cper header
From: Hawking Zhang <[email protected]>
AMD is using Common Platform Error Record (CPER) format to report all gpu hardware errors.
v2: add program attribute
--- drivers/gpu/drm/amd/include/amd_cper.h | 269 +++++++++++++++++++++++++ 1 file changed, 269 insertions(+) create mode 100644 drivers/gpu/drm/amd/include/amd_cper.h
drivers/gpu/drm/amd/include/amd_cper.h new file +269 -0
@@ -0,0 +1,269 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* +* Copyright 2025 Advanced Micro Devices, Inc. +* +* Permission is hereby granted, free of charge, to any person obtaining a +* copy of this software and associated documentation files (the "Software"), +* to deal in the Software without restriction, including without limitation +* the rights to use, copy, modify, merge, publish, distribute, sublicense, +* and/or sell copies of the Software, and to permit persons to whom the +* Software is furnished to do so, subject to the following conditions: +* +* The above copyright notice and this permission notice shall be included in +* all copies or substantial portions of the Software. +* +* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL +* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR +* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, +* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR +* OTHER DEALINGS IN THE SOFTWARE. +* +*/ +#ifndef __AMD_CPER_H__ +#define __AMD_CPER_H__ + +#include <linux/uuid.h> + +#define CPER_HDR_REV_1 (0x100) +#define CPER_SEC_MINOR_REV_1 (0x01) +#define CPER_SEC_MAJOR_REV_22 (0x22) +#define CPER_MAX_OAM_COUNT (8) + +#define CPER_CTX_TYPE_CRASH (1) +#define CPER_CTX_TYPE_BOOT (9) + +#define CPER_CREATOR_ID_AMDGPU "amdgpu" + +#define CPER_NOTIFY_MCE \ + GUID_INIT(0xE8F56FFE, 0x919C, 0x4cc5, 0xBA, 0x88, 0x65, 0xAB, \ + 0xE1, 0x49, 0x13, 0xBB) +#define CPER_NOTIFY_CMC \ + GUID_INIT(0x2DCE8BB1, 0xBDD7, 0x450e, 0xB9, 0xAD, 0x9C, 0xF4, \ + 0xEB, 0xD4, 0xF8, 0x90) +#define BOOT_TYPE \ + GUID_INIT(0x3D61A466, 0xAB40, 0x409a, 0xA6, 0x98, 0xF3, 0x62, \ + 0xD4, 0x64, 0xB3, 0x8F) + +#define AMD_CRASHDUMP \ + GUID_INIT(0x32AC0C78, 0x2623, 0x48F6, 0xB0, 0xD0, 0x73, 0x65, \ + 0x72, 0x5F, 0xD6, 0xAE) +#define AMD_GPU_NONSTANDARD_ERROR \ + GUID_INIT(0x32AC0C78, 0x2623, 0x48F6, 0x81, 0xA2, 0xAC, 0x69, \ + 0x17, 0x80, 0x55, 0x1D) +#define PROC_ERR_SECTION_TYPE \ + GUID_INIT(0xDC3EA0B0, 0xA144, 0x4797, 0xB9, 0x5B, 0x53, 0xFA, \ + 0x24, 0x2B, 0x6E, 0x1D) + +enum cper_error_severity { +CPER_SEV_NON_FATAL_UNCORRECTED =0 , +CPER_SEV_FATAL =1 , +CPER_SEV_NON_FATAL_CORRECTED =2 , +CPER_SEV_NUM =3 , + +CPER_SEV_UNUSED =10 , +}; + +enum cper_aca_reg { +CPER_ACA_REG_CTL_LO =0 , +CPER_ACA_REG_CTL_HI =1 , +CPER_ACA_REG_STATUS_LO =2 , +CPER_ACA_REG_STATUS_HI =3 , +CPER_ACA_REG_ADDR_LO =4 , +CPER_ACA_REG_ADDR_HI =5 , +CPER_ACA_REG_MISC0_LO =6 , +CPER_ACA_REG_MISC0_HI =7 , +CPER_ACA_REG_CONFIG_LO =8 , +CPER_ACA_REG_CONFIG_HI =9 , +CPER_ACA_REG_IPID_LO =10 , +CPER_ACA_REG_IPID_HI =11 , +CPER_ACA_REG_SYND_LO =12 , +CPER_ACA_REG_SYND_HI =13 , + +CPER_ACA_REG_COUNT =32 , +}; + +#pragma pack(push, 1) + +struct cper_timestamp { +uint8_t seconds; +uint8_t minutes; +uint8_t hours; +uint8_t flag; +uint8_t day; +uint8_t month; +uint8_t year; +uint8_t century; +}; + +struct cper_hdr { +char signature[4 ];/* "CPER" */ +uint16_t revision; +uint32_t signature_end;/* 0xFFFFFFFF */ +uint16_t sec_cnt; +enum cper_error_severity error_severity; +union { +struct { +uint32_t platform_id :1 ; +uint32_t timestamp :1 ; +uint32_t partition_id :1 ; +uint32_t reserved :29 ; + } valid_bits; +uint32_t valid_mask; + }; +uint32_t record_length;/* Total size of CPER Entry */ +struct cper_timestamp timestamp; +char platform_id[16 ]; +guid_t partition_id;/* Reserved */ +char creator_id[16 ]; +guid_t notify_type;/* CMC, MCE */ +char record_id[8 ];/* Unique CPER Entry ID */ +uint32_t flags;/* Reserved */ +uint64_t persistence_info;/* Reserved */ +uint8_t reserved[12 ];/* Reserved */ +}; + +struct cper_sec_desc { +uint32_t sec_offset;/* Offset from the start of CPER entry */ +uint32_t sec_length; +uint8_t revision_minor;/* CPER_SEC_MINOR_REV_1 */ +uint8_t revision_major;/* CPER_SEC_MAJOR_REV_22 */ +union { +struct { +uint8_t fru_id :1 ; +uint8_t fru_text :1 ; +uint8_t reserved :6 ; + } valid_bits; +uint8_t valid_mask; + }; +uint8_t reserved; +union { +struct { +uint32_t primary :1 ; +uint32_t reserved1 :2 ; +uint32_t exceed_err_threshold :1 ; +uint32_t latent_err :1 ; +uint32_t reserved2 :27 ; + } flag_bits; +uint32_t flag_mask; + }; +guid_t sec_type; +char fru_id[16 ]; +enum cper_error_severity severity; +char fru_text[20 ]; +}; + +struct cper_sec_nonstd_err_hdr { +union { +struct { +uint64_t apic_id :1 ; +uint64_t fw_id :1 ; +uint64_t err_info_cnt :6 ; +uint64_t err_context_cnt :6 ; + } valid_bits; +uint64_t valid_mask; + }; +uint64_t apic_id; +char fw_id[48 ]; +}; + +struct cper_sec_nonstd_err_info { +guid_t error_type; +union { +struct { +uint64_t ms_chk :1 ; +uint64_t target_addr_id :1 ; +uint64_t req_id :1 ; +uint64_t resp_id :1 ; +uint64_t instr_ptr :1 ; +uint64_t reserved :59 ; + } valid_bits; +uint64_t valid_mask; + }; +union { +struct { +uint64_t err_type_valid :1 ; +uint64_t pcc_valid :1 ; +uint64_t uncorr_valid :1 ; +uint64_t precise_ip_valid :1 ; +uint64_t restartable_ip_valid :1 ; +uint64_t overflow_valid :1 ; +uint64_t reserved1 :10 ; +uint64_t err_type :2 ; +uint64_t pcc :1 ; +uint64_t uncorr :1 ; +uint64_t precised_ip :1 ; +uint64_t restartable_ip :1 ; +uint64_t overflow :1 ; +uint64_t reserved2 :41 ; + } ms_chk_bits; +uint64_t ms_chk_mask; + }; +uint64_t target_addr_id; +uint64_t req_id; +uint64_t resp_id; +uint64_t instr_ptr; +}; + +struct cper_sec_nonstd_err_ctx { +uint16_t reg_ctx_type; +uint16_t reg_arr_size; +uint32_t msr_addr; +uint64_t mm_reg_addr; +uint32_t reg_dump[CPER_ACA_REG_COUNT ]; +}; + +struct cper_sec_nonstd_err { +struct cper_sec_nonstd_err_hdr hdr; +struct cper_sec_nonstd_err_info info; +struct cper_sec_nonstd_err_ctx ctx; +}; + +struct cper_sec_crashdump_hdr { +uint64_t reserved1; +uint64_t reserved2; +char fw_id[48 ]; +uint64_t reserved3[8 ]; +}; + +struct cper_sec_crashdump_reg_data { +uint32_t status_lo; +uint32_t status_hi; +uint32_t addr_lo; +uint32_t addr_hi; +uint32_t ipid_lo; +uint32_t ipid_hi; +uint32_t synd_lo; +uint32_t synd_hi; +}; + +struct cper_sec_crashdump_body_fatal { +uint16_t reg_ctx_type; +uint16_t reg_arr_size; +uint32_t reserved1; +uint64_t reserved2; +struct cper_sec_crashdump_reg_data data; +}; + +struct cper_sec_crashdump_body_boot { +uint16_t reg_ctx_type; +uint16_t reg_arr_size; +uint32_t reserved1; +uint64_t reserved2; +uint64_t msg[CPER_MAX_OAM_COUNT ]; +}; + +struct cper_sec_crashdump_fatal { +struct cper_sec_crashdump_hdr hdr; +struct cper_sec_crashdump_body_fatal body; +}; + +struct cper_sec_crashdump_boot { +struct cper_sec_crashdump_hdr hdr; +struct cper_sec_crashdump_body_boot body; +}; + +#pragma pack(pop) + +#endif
Signature · 1 line
2.34.1
PATCH v2 2/12 drm/amdgpu: Introduce funcs for populating CPER
In reply to Xiang LiuFrom: Hawking Zhang <[email protected]>
Introduce utility functions designed to assist
in populating CPER records.
v2: call cper_init/fini in device_ip_init/fini.
--- drivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 281 +++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 91 +++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 + 5 files changed, 381 insertions(+), 1 deletion(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
drivers/gpu/drm/amd/amdgpu/Makefile +1 -1
@@ -66,7 +66,7 @@ amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \ amdgpu_fw_attestation.o amdgpu_securedisplay.o \ amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \ amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \ - amdgpu_userq_fence.o amdgpu_eviction_fence.o + amdgpu_userq_fence.o amdgpu_eviction_fence.o amdgpu_cper.o amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o
drivers/gpu/drm/amd/amdgpu/amdgpu.h +4 -0
@@ -110,6 +110,7 @@#include "amdgpu_mca.h" #include "amdgpu_aca.h" #include "amdgpu_ras.h" +#include "amdgpu_cper.h" #include "amdgpu_xcp.h" #include "amdgpu_seq64.h" #include "amdgpu_reg_state.h"
@@ -1128,6 +1129,9 @@ struct amdgpu_device {/* ACA */ struct amdgpu_aca aca; +/* CPER */ +struct amdgpu_cper cper; +struct amdgpu_ip_block ip_blocks[AMDGPU_MAX_IP_NUM ];uint32_t harvest_ip_mask;int num_ip_blocks;
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c new file +281 -0
281 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h new file +91 -0
91 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +4 -0
@@ -3075,6 +3075,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)amdgpu_fru_get_product_info (adev); + r =amdgpu_cper_init (adev); + init_failed:return r;
@@ -3435,6 +3437,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) {int i, r; +amdgpu_cper_fini (adev); +if (amdgpu_sriov_vf (adev) && adev->virt.ras_init_done)amdgpu_virt_release_ras_err_handler_data (adev);
Signature · 1 line
2.34.1
PATCH v2 3/12 drm/amdgpu: Include ACA error type in aca bank
In reply to Xiang LiuFrom: Hawking Zhang <[email protected]>
ACA error types managed by driver a direct 1:1 correspondence with those managed by firmware.
To address this, for each ACA bank, include both the ACA error type and the ACA SMU type.
This addition is useful for creating CPER records.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 4 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 2 ++ drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 2 ++ drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 2 ++ drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 ++ drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 ++ drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 2 ++ 9 files changed, 17 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +1 -1
@@ -168,7 +168,7 @@ static int aca_smu_get_valid_aca_banks(struct amdgpu_device *adev, enum aca_smu_if (ret)return ret; - bank.type = type; + bank.smu_err_type = type;aca_smu_bank_dump (adev, i, count, &bank, qctx);
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h +3 -1
@@ -108,13 +108,15 @@ enum aca_error_type { };enum aca_smu_type { +ACA_SMU_TYPE_INVALID =-1 ,ACA_SMU_TYPE_UE =0 ,ACA_SMU_TYPE_CE ,ACA_SMU_TYPE_COUNT , };struct aca_bank { -enum aca_smu_type type; +enum aca_error_type aca_err_type; +enum aca_smu_type smu_err_type;u64 regs[ACA_MAX_REGS_COUNT ]; };
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +2 -0
@@ -1129,10 +1129,12 @@ static int xgmi_v6_4_0_aca_bank_parser(struct aca_handle *handle, struct aca_banif (ext_error_code !=0 && ext_error_code !=9 ) count =0ULL ; + bank->aca_err_type =ACA_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE , count);break ;case ACA_SMU_TYPE_CE : count = ext_error_code ==6 ? count :0ULL ; + bank->aca_err_type =ACA_SMU_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE , count);break ;default :
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +2 -0
@@ -893,10 +893,12 @@ 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_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE ,1ULL );break ;case ACA_SMU_TYPE_CE : + bank->aca_err_type =ACA_SMU_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE ,ACA_REG__MISC0__ERRCNT (misc0));break ;
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c +2 -0
@@ -1249,10 +1249,12 @@ static int jpeg_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_ban misc0 = bank->regs[ACA_REG_IDX_MISC0 ];switch (type) {case ACA_SMU_TYPE_UE : + bank->aca_err_type =ACA_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE ,1ULL );break ;case ACA_SMU_TYPE_CE : + bank->aca_err_type =ACA_ERROR_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE ,ACA_REG__MISC0__ERRCNT (misc0));break ;
drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c +2 -0
@@ -719,10 +719,12 @@ static int mmhub_v1_8_aca_bank_parser(struct aca_handle *handle, struct aca_bank misc0 = bank->regs[ACA_REG_IDX_MISC0 ];switch (type) {case ACA_SMU_TYPE_UE : + bank->aca_err_type =ACA_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE ,1ULL );break ;case ACA_SMU_TYPE_CE : + bank->aca_err_type =ACA_ERROR_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE ,ACA_REG__MISC0__ERRCNT (misc0));break ;
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +2 -0
@@ -2392,10 +2392,12 @@ static int sdma_v4_4_2_aca_bank_parser(struct aca_handle *handle, struct aca_ban misc0 = bank->regs[ACA_REG_IDX_MISC0 ];switch (type) {case ACA_SMU_TYPE_UE : + bank->aca_err_type =ACA_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE ,1ULL );break ;case ACA_SMU_TYPE_CE : + bank->aca_err_type =ACA_ERROR_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE ,ACA_REG__MISC0__ERRCNT (misc0));break ;
drivers/gpu/drm/amd/amdgpu/umc_v12_0.c +1 -0
@@ -415,6 +415,7 @@ static int umc_v12_0_aca_bank_parser(struct aca_handle *handle, struct aca_bank err_type =ACA_ERROR_TYPE_CE ; else return 0; + bank->aca_err_type = err_type; ret =aca_bank_info_decode (bank, &info);if (ret)
drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c +2 -0
@@ -1935,10 +1935,12 @@ static int vcn_v4_0_3_aca_bank_parser(struct aca_handle *handle, struct aca_bank misc0 = bank->regs[ACA_REG_IDX_MISC0 ];switch (type) {case ACA_SMU_TYPE_UE : + bank->aca_err_type =ACA_ERROR_TYPE_UE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_UE ,1ULL );break ;case ACA_SMU_TYPE_CE : + bank->aca_err_type =ACA_ERROR_TYPE_CE ; ret =aca_error_cache_log_bank_error (handle, &info,ACA_ERROR_TYPE_CE ,ACA_REG__MISC0__ERRCNT (misc0));break ;
Signature · 1 line
2.34.1
PATCH v2 4/12 drm/amdgpu: Introduce funcs for generating cper record
In reply to Xiang LiuFrom: Hawking Zhang <[email protected]>
Introduce new functions that are used to generate cper ue or ce records.
v2: return -ENOMEM instead of false
v2: check return value of fill section function
--- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 12 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 12 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 108 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 9 +- 4 files changed, 128 insertions(+), 13 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +1 -11
@@ -30,16 +30,6 @@typedef int bank_handler_t (struct aca_handle *handle,struct aca_bank *bank,enum aca_smu_type type,void *data); -struct aca_banks { -int nr_banks; -struct list_head list; -}; - -struct aca_hwip { -int hwid; -int mcatype; -}; -static struct aca_hwip aca_hwid_mcatypes[ACA_HWIP_TYPE_COUNT ] = {ACA_BANK_HWID (SMU ,0x01 ,0x01 ),ACA_BANK_HWID (PCS_XGMI ,0x50 ,0x00 ),
@@ -111,7 +101,7 @@ static struct aca_regs_dump { {"STATUS", ACA_REG_IDX_STATUS}, {"ADDR", ACA_REG_IDX_ADDR}, {"MISC", ACA_REG_IDX_MISC0}, - {"CONFIG", ACA_REG_IDX_CONFG}, + {"CONFIG", ACA_REG_IDX_CONFIG}, {"IPID", ACA_REG_IDX_IPID}, {"SYND", ACA_REG_IDX_SYND}, {"DESTAT", ACA_REG_IDX_DESTAT},
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h +11 -1
@@ -81,7 +81,7 @@ enum aca_reg_idx {ACA_REG_IDX_STATUS =1 ,ACA_REG_IDX_ADDR =2 ,ACA_REG_IDX_MISC0 =3 , -ACA_REG_IDX_CONFG =4 , +ACA_REG_IDX_CONFIG =4 ,ACA_REG_IDX_IPID =5 ,ACA_REG_IDX_SYND =6 , ACA_REG_IDX_DESTAT = 8,
@@ -114,6 +114,11 @@ enum aca_smu_type { ACA_SMU_TYPE_COUNT, }; +struct aca_hwip { +int hwid; +int mcatype; +}; +struct aca_bank {enum aca_error_type aca_err_type;enum aca_smu_type smu_err_type;
@@ -125,6 +130,11 @@ struct aca_bank_node {struct list_head node; }; +struct aca_banks { +int nr_banks; +struct list_head list; +}; +struct aca_bank_info {int die_id;int socket_id;
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +108 -0
120 lines not on this page · raw message
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h +8 -1
@@ -26,6 +26,7 @@#define __AMDGPU_CPER_H__#include "amd_cper.h" +#include "amdgpu_aca.h" #define CPER_MAX_ALLOWED_COUNT 0x1000#define HDR_LEN (sizeof(struct cper_hdr))
@@ -84,7 +85,13 @@ int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adevstruct cper_hdr *amdgpu_cper_alloc_entry (struct amdgpu_device *adev,enum amdgpu_cper_type type,uint16_t section_count); - +/* UE must be encoded into separated cper entries, 1 UE 1 cper */ +int amdgpu_cper_generate_ue_record (struct amdgpu_device *adev, +struct aca_bank *bank); +/* CEs and DEs are combined into 1 cper entry */ +int amdgpu_cper_generate_ce_records (struct amdgpu_device *adev, +struct aca_banks *banks, +uint16_t bank_count);int amdgpu_cper_init (struct amdgpu_device *adev);int amdgpu_cper_fini (struct amdgpu_device *adev);
Signature · 1 line
2.34.1
PATCH v2 5/12 drm/amdgpu: Generate cper records
In reply to Xiang LiuFrom: Hawking Zhang <[email protected]>
Encode the error information in CPER format and commit to the cper ring
--- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 32 +++++++++++++++++++++++++ 1 file changed, 32 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +32 -0
@@ -384,6 +384,36 @@ static bool aca_bank_should_update(struct amdgpu_device *adev, enum aca_smu_typereturn ret; } +static void aca_banks_generate_cper (struct amdgpu_device *adev, +enum aca_smu_type type, +struct aca_banks *banks, +int count) +{ +struct aca_bank_node *node; +struct aca_bank *bank; + +if (!adev || !banks || !count) { +dev_warn (adev->dev,"fail to generate cper records\n" ); +return ; + } + +/* UEs must be encoded into separate CPER entries */ +if (type ==ACA_SMU_TYPE_UE ) { +list_for_each_entry (node, &banks->list, node) { + bank = &node->bank; +if (amdgpu_cper_generate_ue_record (adev, bank)) +dev_warn (adev->dev,"fail to generate ue cper records\n" ); + } + }else { +/* +* SMU_TYPE_CE banks are combined into 1 CPER entries, +* they could be CEs or DEs or both +*/ +if (amdgpu_cper_generate_ce_records (adev, banks, count)) +dev_warn (adev->dev,"fail to generate ce cper records\n" ); + } +} +static int aca_banks_update (struct amdgpu_device *adev,enum aca_smu_type type,bank_handler_t handler,struct ras_query_context *qctx,void *data) {
@@ -421,6 +451,8 @@ static int aca_banks_update(struct amdgpu_device *adev, enum aca_smu_type type,if (ret) goto err_release_banks; +aca_banks_generate_cper (adev, type, &banks, count); + err_release_banks:aca_banks_release (&banks);
Signature · 1 line
2.34.1
Re: [PATCH v2 05/12] drm/amdgpu: Generate cper records
Op 14-02-2025 om 09:07 schreef Xiang Liu:
Quoted 20 lines
From: Hawking Zhang <[email protected]>
Encode the error information in CPER format and commit to the cper ring
Signed-off-by Hawking Zhang <[email protected]>Reviewed-by Yang Wang <[email protected]>Reviewed-by Tao Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 32 +++++++++++++++++++++++++ 1 file changed, 32 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +0
-0@@ -384,6 +384,36 @@ static bool aca_bank_should_update(struct amdgpu_device *adev, enum aca_smu_typereturn ret; }
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c static bool aca_bank_should_update(struct amdgpu_device *adev, enum aca_smu_type · line 396
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +10
-0+static void aca_banks_generate_cper (struct amdgpu_device *adev, +enum aca_smu_type type, +struct aca_banks *banks, +int count) +{ +struct aca_bank_node *node; +struct aca_bank *bank; + +if (!adev || !banks || !count) { +dev_warn (adev->dev,"fail to generate cper records\n" );
The check for !adev is suspicious in combination with the dev_warn(adev->dev...
Quoted 32 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c +22
-0+return ; + } + +/* UEs must be encoded into separate CPER entries */ +if (type ==ACA_SMU_TYPE_UE ) { +list_for_each_entry (node, &banks->list, node) { + bank = &node->bank; +if (amdgpu_cper_generate_ue_record (adev, bank)) +dev_warn (adev->dev,"fail to generate ue cper records\n" ); + } + }else { +/* +* SMU_TYPE_CE banks are combined into 1 CPER entries, +* they could be CEs or DEs or both +*/ +if (amdgpu_cper_generate_ce_records (adev, banks, count)) +dev_warn (adev->dev,"fail to generate ce cper records\n" ); + } +} +static int aca_banks_update (struct amdgpu_device *adev,enum aca_smu_type type,bank_handler_t handler,struct ras_query_context *qctx,void *data) {@@ -421,6 +451,8 @@ static int aca_banks_update(struct amdgpu_device *adev, enum aca_smu_type type,if (ret) goto err_release_banks; +aca_banks_generate_cper (adev, type, &banks, count); + err_release_banks:aca_banks_release (&banks);
PATCH v2 6/12 drm/amdgpu: add RAS CPER ring buffer
In reply to Xiang LiuFrom: Tao Zhou <[email protected]>
And initialize it, this is a pure software ring to store RAS CPER data.
v2: update the initialization of count_dw of cper ring, it's dword variable.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 39 +++++++++++++++++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 29 ++++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c | 3 +- 5 files changed, 57 insertions(+), 16 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +35 -4
@@ -365,6 +365,39 @@ int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,return 0 ; } +static u64 amdgpu_cper_ring_get_rptr (struct amdgpu_ring *ring) +{ +return *(ring->rptr_cpu_addr); +} + +static u64 amdgpu_cper_ring_get_wptr (struct amdgpu_ring *ring) +{ +return ring->wptr; +} + +static const struct amdgpu_ring_funcs cper_ring_funcs = { + .type =AMDGPU_RING_TYPE_CPER , + .align_mask =0xff , + .support_64bit_ptrs = false, + .get_rptr = amdgpu_cper_ring_get_rptr, + .get_wptr = amdgpu_cper_ring_get_wptr, +}; + +static int amdgpu_cper_ring_init (struct amdgpu_device *adev) +{ +struct amdgpu_ring *ring = &(adev->cper.ring_buf); + + ring->adev =NULL ; + ring->ring_obj =NULL ; + ring->use_doorbell = false; + ring->no_scheduler = true; + ring->funcs = &cper_ring_funcs; + +sprintf (ring->name,"cper" ); +return amdgpu_ring_init (adev, ring,PAGE_SIZE ,NULL ,0 , +AMDGPU_RING_PRIO_DEFAULT ,NULL ); +} +int amdgpu_cper_init (struct amdgpu_device *adev) {mutex_init (&adev->cper.cper_lock);
@@ -372,16 +405,14 @@ int amdgpu_cper_init(struct amdgpu_device *adev) adev->cper.enabled = true; adev->cper.max_count =CPER_MAX_ALLOWED_COUNT ; -/*TODO: initialize cper ring*/ - -return 0 ; +return amdgpu_cper_ring_init (adev); }int amdgpu_cper_fini (struct amdgpu_device *adev) { adev->cper.enabled = false; -/*TODO: free cper ring */ +amdgpu_ring_fini (&(adev->cper.ring_buf)); adev->cper.count =0 ; adev->cper.wptr =0 ;
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h +1 -0
@@ -62,6 +62,7 @@ struct amdgpu_cper {uint32_t wptr;void *ring[CPER_MAX_ALLOWED_COUNT ]; +struct amdgpu_ring ring_buf; }; void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +18 -11
@@ -324,20 +324,27 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,/* always set cond_exec_polling to CONTINUE */ *ring->cond_exe_cpu_addr =1 ; - r =amdgpu_fence_driver_start_ring (ring, irq_src, irq_type); -if (r) { -dev_err (adev->dev,"failed initializing fences (%d).\n" , r); -return r; - } +if (ring->funcs->type !=AMDGPU_RING_TYPE_CPER ) { + r =amdgpu_fence_driver_start_ring (ring, irq_src, irq_type); +if (r) { +dev_err (adev->dev,"failed initializing fences (%d).\n" , r); +return r; + } - max_ibs_dw = ring->funcs->emit_frame_size + -amdgpu_ring_max_ibs (ring->funcs->type) * ring->funcs->emit_ib_size; - max_ibs_dw = (max_ibs_dw + ring->funcs->align_mask) & ~ring->funcs->align_mask; + max_ibs_dw = ring->funcs->emit_frame_size + +amdgpu_ring_max_ibs (ring->funcs->type) * ring->funcs->emit_ib_size; + max_ibs_dw = (max_ibs_dw + ring->funcs->align_mask) & ~ring->funcs->align_mask; -if (WARN_ON (max_ibs_dw > max_dw)) - max_dw = max_ibs_dw; +if (WARN_ON (max_ibs_dw > max_dw)) + max_dw = max_ibs_dw; - ring->ring_size =roundup_pow_of_two (max_dw *4 * sched_hw_submission); + ring->ring_size =roundup_pow_of_two (max_dw *4 * sched_hw_submission); + }else { + ring->ring_size =roundup_pow_of_two (max_dw *4 ); + ring->count_dw = (ring->ring_size -4 ) >>2 ; +/* ring buffer is empty now */ + ring->wptr = *ring->rptr_cpu_addr =0 ; + } ring->buf_mask = (ring->ring_size /4 ) -1 ; ring->ptr_mask = ring->funcs->support_64bit_ptrs ?
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +1 -0
@@ -82,6 +82,7 @@ enum amdgpu_ring_type {AMDGPU_RING_TYPE_KIQ ,AMDGPU_RING_TYPE_MES ,AMDGPU_RING_TYPE_UMSCH_MM , + AMDGPU_RING_TYPE_CPER, }; enum amdgpu_ib_pool_type {
drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c +2 -1
@@ -77,7 +77,8 @@ static void aqua_vanjaram_set_xcp_id(struct amdgpu_device *adev, ring->xcp_id =AMDGPU_XCP_NO_PARTITION ;if (ring->funcs->type ==AMDGPU_RING_TYPE_COMPUTE ) adev->gfx.enforce_isolation[0 ].xcp_id = ring->xcp_id; -if (adev->xcp_mgr->mode ==AMDGPU_XCP_MODE_NONE ) +if ((adev->xcp_mgr->mode ==AMDGPU_XCP_MODE_NONE ) || + (ring->funcs->type ==AMDGPU_RING_TYPE_CPER ))return ; inst_mask =1 << inst_idx;
Signature · 1 line
2.34.1
PATCH v2 7/12 drm/amdgpu: read CPER ring via debugfs
In reply to Xiang LiuFrom: Tao Zhou <[email protected]>
We read CPER data from read pointer to write pointer without changing the pointers.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 47 ++++++++++++++++++------ 1 file changed, 36 insertions(+), 11 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +36 -11
@@ -500,6 +500,7 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf, {struct amdgpu_ring *ring =file_inode (f)->i_private;uint32_t value, result, early[3 ]; +uint64_t p;loff_t i;int r;
@@ -523,18 +524,42 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf, } } -while (size) { -if (*pos >= (ring->ring_size +12 )) -return result; +if (ring->funcs->type !=AMDGPU_RING_TYPE_CPER ) { +while (size) { +if (*pos >= (ring->ring_size +12 )) +return result; - value = ring->ring[(*pos -12 )/4 ]; - r =put_user (value, (uint32_t *)buf); -if (r) -return r; - buf +=4 ; - result +=4 ; - size -=4 ; - *pos +=4 ; + value = ring->ring[(*pos -12 )/4 ]; + r =put_user (value, (uint32_t *)buf); +if (r) +return r; + buf +=4 ; + result +=4 ; + size -=4 ; + *pos +=4 ; + } + }else { + p = early[0 ]; +if (early[0 ] <= early[1 ]) + size = (early[1 ] - early[0 ]); +else + size = ring->ring_size - (early[0 ] - early[1 ]); + +while (size) { +if (p == early[1 ]) +return result; + + value = ring->ring[p]; + r =put_user (value, (uint32_t *)buf); +if (r) +return r; + + buf +=4 ; + result +=4 ; + size--; + p++; + p &= ring->ptr_mask; + } }return result;
Signature · 1 line
2.34.1
PATCH v2 8/12 drm/amdgpu: add data write function for CPER ring
In reply to Xiang LiuFrom: Tao Zhou <[email protected]>
Old CPER data will be overwritten if ring buffer is full, and read pointer always points to CPER header.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 93 ++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 2 + 2 files changed, 95 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +93 -0
@@ -365,6 +365,99 @@ int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,return 0 ; } +static bool amdgpu_cper_is_hdr (struct amdgpu_ring *ring,u64 pos) +{ +struct cper_hdr *chdr; + + chdr = (struct cper_hdr *)&(ring->ring[pos]); +return strcmp (chdr->signature,"CPER" ) ? false : true; +} + +static u32 amdgpu_cper_ring_get_ent_sz (struct amdgpu_ring *ring,u64 pos) +{ +struct cper_hdr *chdr; +u64 p; +u32 chunk, rec_len =0 ; + + chdr = (struct cper_hdr *)&(ring->ring[pos]); + chunk = ring->ring_size - (pos <<2 ); + +if (!strcmp (chdr->signature,"CPER" )) { + rec_len = chdr->record_length; + goto calc; + } + +/* ring buffer is not full, no cper data after ring->wptr */ +if (ring->count_dw) + goto calc; + +for (p = pos +1 ; p <= ring->buf_mask; p++) { + chdr = (struct cper_hdr *)&(ring->ring[p]); +if (!strcmp (chdr->signature,"CPER" )) { + rec_len = (p - pos) <<2 ; + goto calc; + } + } + +calc: +if (!rec_len) +return chunk; +else +return min (rec_len, chunk); +} + +void amdgpu_cper_ring_write (struct amdgpu_ring *ring, +void *src,int count) +{ +u64 pos, wptr_old, rptr = *ring->rptr_cpu_addr & ring->ptr_mask; +u32 chunk, ent_sz; +u8 *s = (u8 *)src; + +if (count >= ring->ring_size -4 ) { +dev_err (ring->adev->dev, +"CPER data size(%d) is larger than ring size(%d)\n" , + count, ring->ring_size -4 ); + +return ; + } + + wptr_old = ring->wptr; + +while (count) { + ent_sz =amdgpu_cper_ring_get_ent_sz (ring, ring->wptr); + chunk =min (ent_sz, count); + +memcpy (&ring->ring[ring->wptr], s, chunk); + + ring->wptr += (chunk >>2 ); + ring->wptr &= ring->ptr_mask; + count -= chunk; + s += chunk; + } + +/* the buffer is overflow, adjust rptr */ +if (((wptr_old < rptr) && (rptr <= ring->wptr)) || + ((ring->wptr < wptr_old) && (wptr_old < rptr)) || + ((rptr <= ring->wptr) && (ring->wptr < wptr_old))) { + pos = (ring->wptr +1 ) & ring->ptr_mask; + +do { + ent_sz =amdgpu_cper_ring_get_ent_sz (ring, pos); + + rptr += (ent_sz >>2 ); + rptr &= ring->ptr_mask; + *ring->rptr_cpu_addr = rptr; + + pos = rptr; + }while (!amdgpu_cper_is_hdr (ring, rptr)); + } + +if (ring->count_dw >= (count >>2 )) + ring->count_dw - (count >>2 ); +else + ring->count_dw =0 ; +} +static u64 amdgpu_cper_ring_get_rptr (struct amdgpu_ring *ring) {return *(ring->rptr_cpu_addr);
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h +2 -0
@@ -93,6 +93,8 @@ int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,int amdgpu_cper_generate_ce_records (struct amdgpu_device *adev,struct aca_banks *banks,uint16_t bank_count); +void amdgpu_cper_ring_write (struct amdgpu_ring *ring, +void *src,int count);int amdgpu_cper_init (struct amdgpu_device *adev);int amdgpu_cper_fini (struct amdgpu_device *adev);
Signature · 1 line
2.34.1
PATCH v2 9/12 drm/amdgpu: add mutex lock for cper ring
In reply to Xiang LiuFrom: Tao Zhou <[email protected]>
Avoid the confliction between read and write of ring buffer.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 4 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 21 ++++++++++++++++----- 3 files changed, 21 insertions(+), 5 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +4 -0
@@ -423,6 +423,7 @@ void amdgpu_cper_ring_write(struct amdgpu_ring *ring, wptr_old = ring->wptr; +mutex_lock (&ring->adev->cper.ring_lock);while (count) { ent_sz =amdgpu_cper_ring_get_ent_sz (ring, ring->wptr); chunk =min (ent_sz, count);
@@ -451,6 +452,7 @@ void amdgpu_cper_ring_write(struct amdgpu_ring *ring, pos = rptr; }while (!amdgpu_cper_is_hdr (ring, rptr)); } +mutex_unlock (&ring->adev->cper.ring_lock);if (ring->count_dw >= (count >>2 )) ring->count_dw - (count >>2 );
@@ -480,6 +482,8 @@ static int amdgpu_cper_ring_init(struct amdgpu_device *adev) {struct amdgpu_ring *ring = &(adev->cper.ring_buf); +mutex_init (&adev->cper.ring_lock); + ring->adev =NULL ; ring->ring_obj =NULL ; ring->use_doorbell = false;
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h +1 -0
@@ -63,6 +63,7 @@ struct amdgpu_cper {void *ring[CPER_MAX_ALLOWED_COUNT ];struct amdgpu_ring ring_buf; +struct mutex ring_lock; }; void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +16 -5
@@ -510,13 +510,18 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf, result =0 ;if (*pos <12 ) { +if (ring->funcs->type ==AMDGPU_RING_TYPE_CPER ) +mutex_lock (&ring->adev->cper.ring_lock); + early[0 ] =amdgpu_ring_get_rptr (ring) & ring->buf_mask; early[1 ] =amdgpu_ring_get_wptr (ring) & ring->buf_mask; early[2 ] = ring->wptr & ring->buf_mask;for (i = *pos /4 ; i <3 && size; i++) { r =put_user (early[i], (uint32_t *)buf); -if (r) -return r; +if (r) { + result = r; + goto out; + } buf +=4 ; result +=4 ; size -=4 ;
@@ -547,12 +552,14 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,while (size) {if (p == early[1 ]) -return result; + goto out; value = ring->ring[p]; r =put_user (value, (uint32_t *)buf); -if (r) -return r; +if (r) { + result = r; + goto out; + } buf +=4 ; result +=4 ;
@@ -562,6 +569,10 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf, } } +out: +if (ring->funcs->type ==AMDGPU_RING_TYPE_CPER ) +mutex_unlock (&ring->adev->cper.ring_lock); +return result; }
Signature · 1 line
2.34.1
PATCH v2 10/12 drm/amdgpu: Get timestamp from system time
In reply to Xiang LiuGet system local time and encode it to timestamp for CPER.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +18 -1
@@ -36,6 +36,22 @@ static void __inc_entry_length(struct cper_hdr *hdr, uint32_t size) hdr->record_length += size; } +static void amdgpu_cper_get_timestamp (struct cper_timestamp *timestamp) +{ +struct tm tm; +time64_t now =ktime_get_real_seconds (); + +time64_to_tm (now,0 , &tm); + timestamp->seconds = tm.tm_sec; + timestamp->minutes = tm.tm_min; + timestamp->hours = tm.tm_hour; + timestamp->flag =0 ; + timestamp->day = tm.tm_mday; + timestamp->month =1 + tm.tm_mon; + timestamp->year = (1900 + tm.tm_year) %100 ; + timestamp->century = (1900 + tm.tm_year) /100 ; +} + void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev, struct cper_hdr *hdr, enum amdgpu_cper_type type,
@@ -52,7 +68,8 @@ void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev, hdr->valid_bits.platform_id =1 ; hdr->valid_bits.partition_id =1 ; hdr->valid_bits.timestamp =1 ; -/*TODO need to initialize hdr->timestamp */ + +amdgpu_cper_get_timestamp (&hdr->timestamp);snprintf (hdr->record_id,8 ,"%d" ,atomic_inc_return (&adev->cper.unique_id)); snprintf(hdr->platform_id, 16, "0x%04X:0x%04X",
Signature · 1 line
2.34.1
PATCH v2 11/12 drm/amdgpu: Commit CPER entry
In reply to Xiang LiuCommit the CPER entry to the ring buffer.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +4 -2
@@ -280,6 +280,7 @@ int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev, {struct cper_hdr *fatal =NULL ;struct cper_sec_crashdump_reg_data reg_data = {0 }; +struct amdgpu_ring *ring = &adev->cper.ring_buf;int ret; fatal =amdgpu_cper_alloc_entry (adev,AMDGPU_CPER_TYPE_FATAL ,1 );
@@ -302,7 +303,7 @@ int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,if (ret)return ret; -/*TODO: commit the cper entry to cper ring */ +amdgpu_cper_ring_write (ring, fatal, fatal->record_length);return 0 ; }
@@ -329,6 +330,7 @@ int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev, {struct cper_hdr *corrected =NULL ;enum cper_error_severity sev =CPER_SEV_NON_FATAL_CORRECTED ; +struct amdgpu_ring *ring = &adev->cper.ring_buf;uint32_t reg_data[CPER_ACA_REG_COUNT ] = {0 };struct aca_bank_node *node;struct aca_bank *bank;
@@ -377,7 +379,7 @@ int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,return ret; } -/*TODO: commit the cper entry to cper ring */ +amdgpu_cper_ring_write (ring, corrected, corrected->record_length);return 0 ; }
Signature · 1 line
2.34.1
PATCH v2 12/12 drm/amdgpu: Generate bad page threshold cper records
In reply to Xiang LiuGenerate CPER record when bad page threshold exceed and commit to CPER ring.
v2: return -ENOMEM instead of false
v2: check return value of fill section function
--- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 23 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 2 ++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 3 +++ 3 files changed, 28 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c +23 -0
@@ -308,6 +308,29 @@ int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,return 0 ; } +int amdgpu_cper_generate_bp_threshold_record (struct amdgpu_device *adev) +{ +struct cper_hdr *bp_threshold =NULL ; +struct amdgpu_ring *ring = &adev->cper.ring_buf; +int ret; + + bp_threshold =amdgpu_cper_alloc_entry (adev,AMDGPU_CPER_TYPE_FATAL ,1 ); +if (!bp_threshold) { +dev_err (adev->dev,"fail to alloc cper entry for bad page threshold record\n" ); +return -ENOMEM ; + } + +amdgpu_cper_entry_fill_hdr (adev, bp_threshold,AMDGPU_CPER_TYPE_BP_THRESHOLD , +CPER_SEV_FATAL ); + ret =amdgpu_cper_entry_fill_bad_page_threshold_section (adev, bp_threshold,0 ); +if (ret) +return ret; + +amdgpu_cper_ring_write (ring, bp_threshold, bp_threshold->record_length); + +return 0 ; +} +static enum cper_error_severity amdgpu_aca_err_type_to_cper_sev (struct amdgpu_device *adev,enum aca_error_type aca_err_type) {
drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h +2 -0
@@ -94,6 +94,8 @@ int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,int amdgpu_cper_generate_ce_records (struct amdgpu_device *adev,struct aca_banks *banks,uint16_t bank_count); +/* Bad page threshold is encoded into separated cper entry */ +int amdgpu_cper_generate_bp_threshold_record (struct amdgpu_device *adev);void amdgpu_cper_ring_write (struct amdgpu_ring *ring,void *src,int count);int amdgpu_cper_init (struct amdgpu_device *adev);
drivers/gpu/drm/amd/pm/amdgpu_dpm.c +3 -0
@@ -716,6 +716,9 @@ int amdgpu_dpm_send_rma_reason(struct amdgpu_device *adev) ret =smu_send_rma_reason (smu);mutex_unlock (&adev->pm.mutex); +if (amdgpu_cper_generate_bp_threshold_record (adev)) +dev_warn (adev->dev,"fail to generate bad page threshold cper records\n" ); +return ret; }
Signature · 1 line
2.34.1
Hi,
Based on the discussions around using Netlink for RAS purposes, as summarized in this blog post [1] by Dave Airlie. I had proposed a series regarding RAS infrastructure in DRM [2].
I came across your work, which appears to address related areas and I'm particularly interested in understanding how it aligns with or could be adapted to the ongoing discussions around leveraging Netlink for RAS.
Could you share your perspective on the potential integration of your efforts with Netlink? Do you foresee any challenges or opportunities in aligning with the approach discussed in the above-mentioned blog post and series?
Looking forward to your insights and any additional thoughts you may have on this topic.
[1]
https://airlied.blogspot.com/2022/09/accelerators-bof-outcomes-summary.html
[2]
https://lore.kernel.org/all/[email protected]/
Thanks,
Aravind.
Quoted 52 lines · Xiang Liu
This patch series generate RAS CPER records for UE/DE/CE/BP threshold exceed event.
SMU_TYPE_CEbanks are combined into 1 CPER entry, they could be CEs or DEs or both. UEs and BPs are encoded into separate CPER entries.RAS CPER records for CEs will be generated only after CEs count been queried.
All records are committed to a pure software ring with a limit size, new records will flush older records when overflow happened. User can access the records by reading debugfs node, which is read-only.
Hawking Zhang (5):
drm/amd/include: Add amd cper header drm/amdgpu: Introduce funcs for populating CPER drm/amdgpu: Include ACA error type in aca bank drm/amdgpu: Introduce funcs for generating cper record drm/amdgpu: Generate cper recordsTao Zhou (4):
drm/amdgpu: add RAS CPER ring buffer drm/amdgpu: read CPER ring via debugfs drm/amdgpu: add data write function for CPER ring drm/amdgpu: add mutex lock for cper ringXiang Liu (3):
drm/amdgpu: Get timestamp from system time drm/amdgpu: Commit CPER entry
drm/amdgpu: Generate bad page threshold cper recordsdrivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 46 +- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 16 +- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 559 +++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 104 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 91 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 2 + drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c | 3 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 2 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 2 + drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 + drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 + drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 2 + drivers/gpu/drm/amd/include/amd_cper.h | 269 ++++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 3 + 19 files changed, 1075 insertions(+), 40 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h create mode 100644 drivers/gpu/drm/amd/include/amd_cper.h
++ dri-devel
Quoted 81 lines · Aravind Iddamsetty
Hi,
Based on the discussions around using Netlink for RAS purposes, as summarized in this blog post [1] by Dave Airlie. I had proposed a series regarding RAS infrastructure in DRM [2].
I came across your work, which appears to address related areas and I'm particularly interested in understanding how it aligns with or could be adapted to the ongoing discussions around leveraging Netlink for RAS.
Could you share your perspective on the potential integration of your efforts with Netlink? Do you foresee any challenges or opportunities in aligning with the approach discussed in the above-mentioned blog post and series?
Looking forward to your insights and any additional thoughts you may have on this topic.
[1]
https://airlied.blogspot.com/2022/09/accelerators-bof-outcomes-summary.html[2]
https://lore.kernel.org/all/[email protected]/Thanks,
Aravind.Quoted 52 lines · Xiang Liu
This patch series generate RAS CPER records for UE/DE/CE/BP threshold exceed event.
SMU_TYPE_CEbanks are combined into 1 CPER entry, they could be CEs or DEs or both. UEs and BPs are encoded into separate CPER entries.RAS CPER records for CEs will be generated only after CEs count been queried.
All records are committed to a pure software ring with a limit size, new records will flush older records when overflow happened. User can access the records by reading debugfs node, which is read-only.
Hawking Zhang (5):
drm/amd/include: Add amd cper header drm/amdgpu: Introduce funcs for populating CPER drm/amdgpu: Include ACA error type in aca bank drm/amdgpu: Introduce funcs for generating cper record drm/amdgpu: Generate cper recordsTao Zhou (4):
drm/amdgpu: add RAS CPER ring buffer drm/amdgpu: read CPER ring via debugfs drm/amdgpu: add data write function for CPER ring drm/amdgpu: add mutex lock for cper ringXiang Liu (3):
drm/amdgpu: Get timestamp from system time drm/amdgpu: Commit CPER entry
drm/amdgpu: Generate bad page threshold cper recordsdrivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 46 +- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 16 +- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 559 +++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 104 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 91 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 2 + drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c | 3 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 2 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 2 + drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 + drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 + drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 2 + drivers/gpu/drm/amd/include/amd_cper.h | 269 ++++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 3 + 19 files changed, 1075 insertions(+), 40 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h create mode 100644 drivers/gpu/drm/amd/include/amd_cper.h
Hi,
I would appreciate it if you could kindly let me know your thoughts.
Thanks,
Aravind.
Quoted 84 lines · Aravind Iddamsetty
++ dri-devel
Quoted 81 lines · Aravind Iddamsetty
Hi,
Based on the discussions around using Netlink for RAS purposes, as summarized in this blog post [1] by Dave Airlie. I had proposed a series regarding RAS infrastructure in DRM [2].
I came across your work, which appears to address related areas and I'm particularly interested in understanding how it aligns with or could be adapted to the ongoing discussions around leveraging Netlink for RAS.
Could you share your perspective on the potential integration of your efforts with Netlink? Do you foresee any challenges or opportunities in aligning with the approach discussed in the above-mentioned blog post and series?
Looking forward to your insights and any additional thoughts you may have on this topic.
[1]
https://airlied.blogspot.com/2022/09/accelerators-bof-outcomes-summary.html[2]
https://lore.kernel.org/all/[email protected]/Thanks,
Aravind.Quoted 52 lines · Xiang Liu
This patch series generate RAS CPER records for UE/DE/CE/BP threshold exceed event.
SMU_TYPE_CEbanks are combined into 1 CPER entry, they could be CEs or DEs or both. UEs and BPs are encoded into separate CPER entries.RAS CPER records for CEs will be generated only after CEs count been queried.
All records are committed to a pure software ring with a limit size, new records will flush older records when overflow happened. User can access the records by reading debugfs node, which is read-only.
Hawking Zhang (5):
drm/amd/include: Add amd cper header drm/amdgpu: Introduce funcs for populating CPER drm/amdgpu: Include ACA error type in aca bank drm/amdgpu: Introduce funcs for generating cper record drm/amdgpu: Generate cper recordsTao Zhou (4):
drm/amdgpu: add RAS CPER ring buffer drm/amdgpu: read CPER ring via debugfs drm/amdgpu: add data write function for CPER ring drm/amdgpu: add mutex lock for cper ringXiang Liu (3):
drm/amdgpu: Get timestamp from system time drm/amdgpu: Commit CPER entry
drm/amdgpu: Generate bad page threshold cper recordsdrivers/gpu/drm/amd/amdgpu/Makefile | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c | 46 +- drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h | 16 +- drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c | 559 +++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h | 104 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 91 +++- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 2 + drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c | 3 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 2 + drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c | 2 + drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 + drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 + drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 1 + drivers/gpu/drm/amd/amdgpu/vcn_v4_0_3.c | 2 + drivers/gpu/drm/amd/include/amd_cper.h | 269 ++++++++++ drivers/gpu/drm/amd/pm/amdgpu_dpm.c | 3 + 19 files changed, 1075 insertions(+), 40 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h create mode 100644 drivers/gpu/drm/amd/include/amd_cper.h