From bc143d8b8387ff0a22e4ef8e2375e63aa24bc311 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Mon, 22 Nov 2021 10:57:20 +0800 Subject: drm/amd/pm: do not expose implementation details to other blocks out of power Those implementation details(whether swsmu supported, some ppt_funcs supported, accessing internal statistics ...)should be kept internally. It's not a good practice and even error prone to expose implementation details. Signed-off-by: Evan Quan Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 586a30ad13e2..4a9970423e7d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -901,7 +901,7 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d * choosing right query method according to * whether smu support query error information */ - ret = smu_get_ecc_info(&adev->smu, (void *)&(ras->umc_ecc)); + ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(ras->umc_ecc)); if (ret == -EOPNOTSUPP) { if (adev->umc.ras_funcs && adev->umc.ras_funcs->query_ras_error_count) @@ -2141,8 +2141,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) if (ret) goto free; - if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->send_hbm_bad_pages_num) - adev->smu.ppt_funcs->send_hbm_bad_pages_num(&adev->smu, con->eeprom_control.ras_num_recs); + amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs); } #ifdef CONFIG_X86_MCE_AMD -- cgit From 6492e1b07c03397f85bd6dc0e230ea6cd9394635 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 13:35:37 +0800 Subject: drm/amdgpu: Unify ras block interface for each ras block 1. Define unified ops interface for each block. 2. Add ras_block_match function pointer in ops interface, each ras block can customize specail match function to identify itself. 3. Add amdgpu_ras_block_match_default new function. If a ras block doesn't define .ras_block_match, default execute amdgpu_ras_block_match_default to identify this ras block. 4. Define unified basic ras block data for each ras block. 5. Create dedicated amdgpu device ras block link list to manage all of the ras blocks. 6. Add amdgpu_ras_register_ras_block new function interface for each ras block to register itself to ras controlling block. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 46 ++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 28 ++++++++++++++++++ 4 files changed, 78 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index f0132a5cc58d..c349337939dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1091,6 +1091,8 @@ struct amdgpu_device { uint32_t ip_versions[MAX_HWIP][HWIP_MAX_INSTANCE]; bool ram_is_direct_mapped; + + struct list_head ras_list; }; static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 596bb2fdb8a2..9230e4476d91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3519,6 +3519,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_LIST_HEAD(&adev->reset_list); + INIT_LIST_HEAD(&adev->ras_list); + INIT_DELAYED_WORK(&adev->delayed_init_work, amdgpu_device_delayed_init_work_handler); INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 4a9970423e7d..bf2983fe5d9d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -866,6 +866,40 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, } /* feature ctl end */ +int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object* block_obj, enum amdgpu_ras_block block) +{ + if(!block_obj) + return -EINVAL; + + if (block_obj->block == block) + return 0; + + return -EINVAL; +} + +static struct amdgpu_ras_block_object* amdgpu_ras_get_ras_block(struct amdgpu_device *adev, + enum amdgpu_ras_block block, uint32_t sub_block_index) +{ + struct amdgpu_ras_block_object *obj, *tmp; + + if (block >= AMDGPU_RAS_BLOCK__LAST) + return NULL; + + if (!amdgpu_ras_is_supported(adev, block)) + return NULL; + + list_for_each_entry_safe(obj, tmp, &adev->ras_list, node) { + if (obj->ras_block_match) { + if (obj->ras_block_match(obj, block, sub_block_index) == 0) + return obj; + } else { + if (amdgpu_ras_block_match_default(obj, block) == 0) + return obj; + } + } + + return NULL; +} static void amdgpu_ras_mca_query_error_status(struct amdgpu_device *adev, struct ras_common_if *ras_block, @@ -2776,3 +2810,15 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev) } } #endif +/* Register each ip ras block into amdgpu ras */ +int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, + struct amdgpu_ras_block_object* ras_block_obj) +{ + if (!adev || !ras_block_obj) + return -EINVAL; + + INIT_LIST_HEAD(&ras_block_obj->node); + list_add_tail(&ras_block_obj->node, &adev->ras_list); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 1c708122d492..f66122fdf477 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -484,6 +484,33 @@ struct ras_debug_if { }; int op; }; + +struct amdgpu_ras_block_object { + /* block name */ + char name[32]; + + enum amdgpu_ras_block block; + + uint32_t sub_block_index; + + /* ras block link */ + struct list_head node; + + int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, enum amdgpu_ras_block block, uint32_t sub_block_index); + int (*ras_late_init)(struct amdgpu_device *adev, void *ras_info); + void (*ras_fini)(struct amdgpu_device *adev); + const struct amdgpu_ras_block_hw_ops *hw_ops; +}; + +struct amdgpu_ras_block_hw_ops { + int (*ras_error_inject)(struct amdgpu_device *adev, void *inject_if); + void (*query_ras_error_count)(struct amdgpu_device *adev,void *ras_error_status); + void (*query_ras_error_status)(struct amdgpu_device *adev); + void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status); + void (*reset_ras_error_count)(struct amdgpu_device *adev); + void (*reset_ras_error_status)(struct amdgpu_device *adev); +}; + /* work flow * vbios * 1: ras feature enable (enabled by default) @@ -667,4 +694,5 @@ const char *get_ras_block_str(struct ras_common_if *ras_block); bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev); +int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object* ras_block_obj); #endif -- cgit From 7cab2124058d2f5f048f435a4631e176dcd1430d Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 13:39:15 +0800 Subject: drm/amdgpu: Modify the compilation failed problem when other ras blocks' .h include amdgpu_ras.h Modify the compilation failed problem when other ras blocks' .h include amdgpu_ras.h. v2: squash in forward declaration warning fix (Alex) Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 39 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 34 +++++++++------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c | 2 +- 4 files changed, 51 insertions(+), 25 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c349337939dd..445a0d077c1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -108,6 +108,7 @@ #include "amdgpu_smuio.h" #include "amdgpu_fdinfo.h" #include "amdgpu_mca.h" +#include "amdgpu_ras.h" #define MAX_GPU_INSTANCE 16 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index bf2983fe5d9d..5b3f4beb2149 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2810,6 +2810,45 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev) } } #endif + +struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev) +{ + if (!adev) + return NULL; + + return adev->psp.ras_context.ras; +} + +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras* ras_con) +{ + if (!adev) + return -EINVAL;; + + adev->psp.ras_context.ras = ras_con; + return 0; +} + +/* check if ras is supported on block, say, sdma, gfx */ +int amdgpu_ras_is_supported(struct amdgpu_device *adev, + unsigned int block) +{ + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + if (block >= AMDGPU_RAS_BLOCK_COUNT) + return 0; + return ras && (adev->ras_enabled & (1 << block)); +} + +int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) +{ + struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); + + if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) + schedule_work(&ras->recovery_work); + return 0; +} + + /* Register each ip ras block into amdgpu ras */ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object* ras_block_obj) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index f66122fdf477..7a4d82378205 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -26,11 +26,11 @@ #include #include -#include "amdgpu.h" -#include "amdgpu_psp.h" #include "ta_ras_if.h" #include "amdgpu_ras_eeprom.h" +struct amdgpu_iv_entry; + #define AMDGPU_RAS_FLAG_INIT_BY_VBIOS (0x1 << 0) enum amdgpu_ras_block { @@ -525,19 +525,6 @@ struct amdgpu_ras_block_hw_ops { * 8: feature disable */ -#define amdgpu_ras_get_context(adev) ((adev)->psp.ras_context.ras) -#define amdgpu_ras_set_context(adev, ras_con) ((adev)->psp.ras_context.ras = (ras_con)) - -/* check if ras is supported on block, say, sdma, gfx */ -static inline int amdgpu_ras_is_supported(struct amdgpu_device *adev, - unsigned int block) -{ - struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - - if (block >= AMDGPU_RAS_BLOCK_COUNT) - return 0; - return ras && (adev->ras_enabled & (1 << block)); -} int amdgpu_ras_recovery_init(struct amdgpu_device *adev); @@ -554,15 +541,6 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev); -static inline int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) -{ - struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - - if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) - schedule_work(&ras->recovery_work); - return 0; -} - static inline enum ta_ras_block amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) { switch (block) { @@ -694,5 +672,13 @@ const char *get_ras_block_str(struct ras_common_if *ras_block); bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev); +int amdgpu_ras_is_supported(struct amdgpu_device *adev, unsigned int block); + +int amdgpu_ras_reset_gpu(struct amdgpu_device *adev); + +struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev); + +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras* ras_con); + int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object* ras_block_obj); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index 25951b2d83b6..e81ce465ff3a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -21,7 +21,7 @@ * */ -#include "amdgpu_ras.h" +#include "amdgpu.h" static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, void *ras_error_status, -- cgit From 8b0fb0e967c1700bd729ae54b6f229501b8587ec Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 13:52:46 +0800 Subject: drm/amdgpu: Modify gfx block to fit for the unified ras block data and ops 1.Modify gfx block to fit for the unified ras block data and ops. 2.Change amdgpu_gfx_ras_funcs to amdgpu_gfx_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of gfx ras variable so that gfx ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register gfx ras block into amdgpu device ras block link list. 5.Remove the redundant code about gfx in amdgpu_ras.c after using the unified ras block. 6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of gfx versions. If .ras_late_init and .ras_fini had been defined by the selected gfx version, the defined functions will take effect; if not defined, default fill with amdgpu_gfx_ras_late_init and amdgpu_gfx_ras_fini. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 8 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 17 +++------ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 63 +++++++++++++++++++++----------- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 65 +++++++++++++++++++++------------ drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c | 24 +++++++----- drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h | 2 +- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c | 25 +++++++------ drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h | 2 +- 8 files changed, 123 insertions(+), 83 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 3d8f82dc8c97..43004822ec6f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -622,7 +622,7 @@ int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value) return r; } -int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev) +int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_fs_if fs_info = { @@ -695,9 +695,9 @@ int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, */ if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) { kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_count) - adev->gfx.ras_funcs->query_ras_error_count(adev, err_data); + if (adev->gfx.ras && adev->gfx.ras->ras_block.hw_ops && + adev->gfx.ras->ras_block.hw_ops->query_ras_error_count) + adev->gfx.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data); amdgpu_ras_reset_gpu(adev); } return AMDGPU_RAS_SUCCESS; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index 776c886fd94a..f99eac544f6d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -31,6 +31,7 @@ #include "amdgpu_ring.h" #include "amdgpu_rlc.h" #include "soc15.h" +#include "amdgpu_ras.h" /* GFX current status */ #define AMDGPU_GFX_NORMAL_MODE 0x00000000L @@ -198,16 +199,8 @@ struct amdgpu_cu_info { uint32_t bitmap[4][4]; }; -struct amdgpu_gfx_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - int (*ras_error_inject)(struct amdgpu_device *adev, - void *inject_if); - int (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); - void (*query_ras_error_status)(struct amdgpu_device *adev); - void (*reset_ras_error_status)(struct amdgpu_device *adev); +struct amdgpu_gfx_ras { + struct amdgpu_ras_block_object ras_block; void (*enable_watchdog_timer)(struct amdgpu_device *adev); }; @@ -331,7 +324,7 @@ struct amdgpu_gfx { /*ras */ struct ras_common_if *ras_if; - const struct amdgpu_gfx_ras_funcs *ras_funcs; + struct amdgpu_gfx_ras *ras; }; #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) @@ -393,7 +386,7 @@ bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me, int pipe, int queue); void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable); int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value); -int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev); +int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, void *ras_info); void amdgpu_gfx_ras_fini(struct amdgpu_device *adev); int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev, void *err_data, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 5b3f4beb2149..a5812c21177e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -89,6 +89,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block) return ras_block_string[ras_block->block]; } +#define ras_block_str(_BLOCK_) (((_BLOCK_) < (sizeof(*ras_block_string)/sizeof(const char*))) ? ras_block_string[_BLOCK_] : "Out Of Range") + #define ras_err_str(i) (ras_error_string[ffs(i)]) #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS) @@ -962,6 +964,7 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) { + struct amdgpu_ras_block_object* block_obj = NULL; struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_err_data err_data = {0, 0, 0, NULL}; int i; @@ -969,6 +972,8 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, if (!obj) return -EINVAL; + block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); + switch (info->head.block) { case AMDGPU_RAS_BLOCK__UMC: amdgpu_ras_get_ecc_info(adev, &err_data); @@ -981,13 +986,16 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } break; case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_count) - adev->gfx.ras_funcs->query_ras_error_count(adev, &err_data); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", + get_ras_block_str(&info->head)); + return -EINVAL; + } + if (block_obj->hw_ops->query_ras_error_count) + block_obj->hw_ops->query_ras_error_count(adev, &err_data); - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_status) - adev->gfx.ras_funcs->query_ras_error_status(adev); + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__MMHUB: if (adev->mmhub.ras_funcs && @@ -1074,18 +1082,23 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, enum amdgpu_ras_block block) { + struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, block, 0); + if (!amdgpu_ras_is_supported(adev, block)) return -EINVAL; switch (block) { case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->reset_ras_error_count) - adev->gfx.ras_funcs->reset_ras_error_count(adev); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); + return -EINVAL; + } + + if (block_obj->hw_ops->reset_ras_error_count) + block_obj->hw_ops->reset_ras_error_count(adev); - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->reset_ras_error_status) - adev->gfx.ras_funcs->reset_ras_error_status(adev); + if (block_obj->hw_ops->reset_ras_error_status) + block_obj->hw_ops->reset_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__MMHUB: if (adev->mmhub.ras_funcs && @@ -1150,7 +1163,8 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, .address = info->address, .value = info->value, }; - int ret = 0; + int ret = -EINVAL; + struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); if (!obj) return -EINVAL; @@ -1164,11 +1178,13 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, switch (info->head.block) { case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_error_inject) - ret = adev->gfx.ras_funcs->ras_error_inject(adev, info); - else - ret = -EINVAL; + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); + return -EINVAL; + } + + if (block_obj->hw_ops->ras_error_inject) + ret = block_obj->hw_ops->ras_error_inject(adev, info); break; case AMDGPU_RAS_BLOCK__UMC: case AMDGPU_RAS_BLOCK__SDMA: @@ -1800,15 +1816,20 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, struct ras_query_if *info) { + struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); /* * Only two block need to query read/write * RspStatus at current state */ switch (info->head.block) { case AMDGPU_RAS_BLOCK__GFX: - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->query_ras_error_status) - adev->gfx.ras_funcs->query_ras_error_status(adev); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); + return ; + } + + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__MMHUB: if (adev->mmhub.ras_funcs && diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 9189fb85a4dd..d36a6bc62560 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -882,7 +882,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev); static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring); static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring); -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status); static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev, void *inject_if); @@ -2197,12 +2197,16 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q, }; -static const struct amdgpu_gfx_ras_funcs gfx_v9_0_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_0_ras_error_inject, - .query_ras_error_count = &gfx_v9_0_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +const struct amdgpu_ras_block_hw_ops gfx_v9_0_ras_ops = { + .ras_error_inject = &gfx_v9_0_ras_error_inject, + .query_ras_error_count = &gfx_v9_0_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_0_reset_ras_error_count, +}; + +static struct amdgpu_gfx_ras gfx_v9_0_ras = { + .ras_block = { + .hw_ops = &gfx_v9_0_ras_ops, + }, }; static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) @@ -2231,7 +2235,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) DRM_INFO("fix gfx.config for vega12\n"); break; case IP_VERSION(9, 4, 0): - adev->gfx.ras_funcs = &gfx_v9_0_ras_funcs; + adev->gfx.ras = &gfx_v9_0_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2258,7 +2262,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; break; case IP_VERSION(9, 4, 1): - adev->gfx.ras_funcs = &gfx_v9_4_ras_funcs; + adev->gfx.ras = &gfx_v9_4_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2279,7 +2283,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config |= 0x22010042; break; case IP_VERSION(9, 4, 2): - adev->gfx.ras_funcs = &gfx_v9_4_2_ras_funcs; + adev->gfx.ras = &gfx_v9_4_2_ras; adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; adev->gfx.config.sc_prim_fifo_size_backend = 0x100; @@ -2298,6 +2302,25 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) break; } + if (adev->gfx.ras) { + err = amdgpu_ras_register_ras_block(adev, &adev->gfx.ras->ras_block); + if (err) { + DRM_ERROR("Failed to register gfx ras block!\n"); + return err; + } + + strcpy(adev->gfx.ras->ras_block.name,"gfx"); + adev->gfx.ras->ras_block.block = AMDGPU_RAS_BLOCK__GFX; + + /* If not define special ras_late_init function, use gfx default ras_late_init */ + if (!adev->gfx.ras->ras_block.ras_late_init) + adev->gfx.ras->ras_block.ras_late_init = amdgpu_gfx_ras_late_init; + + /* If not define special ras_fini function, use gfx default ras_fini */ + if (!adev->gfx.ras->ras_block.ras_fini) + adev->gfx.ras->ras_block.ras_fini = amdgpu_gfx_ras_fini; + } + adev->gfx.config.gb_addr_config = gb_addr_config; adev->gfx.config.gb_addr_config_fields.num_pipes = 1 << @@ -2513,9 +2536,8 @@ static int gfx_v9_0_sw_fini(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_fini) - adev->gfx.ras_funcs->ras_fini(adev); + if (adev->gfx.ras && adev->gfx.ras->ras_block.ras_fini) + adev->gfx.ras->ras_block.ras_fini(adev); for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); @@ -4870,16 +4892,15 @@ static int gfx_v9_0_ecc_late_init(void *handle) if (r) return r; - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->ras_late_init) { - r = adev->gfx.ras_funcs->ras_late_init(adev); + if (adev->gfx.ras && adev->gfx.ras->ras_block.ras_late_init) { + r = adev->gfx.ras->ras_block.ras_late_init(adev, NULL); if (r) return r; } - if (adev->gfx.ras_funcs && - adev->gfx.ras_funcs->enable_watchdog_timer) - adev->gfx.ras_funcs->enable_watchdog_timer(adev); + if (adev->gfx.ras && + adev->gfx.ras->enable_watchdog_timer) + adev->gfx.ras->enable_watchdog_timer(adev); return 0; } @@ -6819,7 +6840,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, mmATC_L2_CACHE_4K_EDC_INDEX, 255); } -static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -6828,7 +6849,7 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, uint32_t reg_value; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -6857,8 +6878,6 @@ static int gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev, mutex_unlock(&adev->grbm_idx_mutex); gfx_v9_0_query_utc_edc_status(adev, err_data); - - return 0; } static void gfx_v9_0_emit_mem_sync(struct amdgpu_ring *ring) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c index b4789dfc2bb9..c67e387a97f5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c @@ -863,7 +863,7 @@ static int gfx_v9_4_ras_error_count(struct amdgpu_device *adev, return 0; } -static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -872,7 +872,7 @@ static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, uint32_t reg_value; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -903,7 +903,6 @@ static int gfx_v9_4_query_ras_error_count(struct amdgpu_device *adev, gfx_v9_4_query_utc_edc_status(adev, err_data); - return 0; } static void gfx_v9_4_reset_ras_error_count(struct amdgpu_device *adev) @@ -1029,11 +1028,16 @@ static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev) mutex_unlock(&adev->grbm_idx_mutex); } -const struct amdgpu_gfx_ras_funcs gfx_v9_4_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_4_ras_error_inject, - .query_ras_error_count = &gfx_v9_4_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_4_reset_ras_error_count, - .query_ras_error_status = &gfx_v9_4_query_ras_error_status, + +const struct amdgpu_ras_block_hw_ops gfx_v9_4_ras_ops = { + .ras_error_inject = &gfx_v9_4_ras_error_inject, + .query_ras_error_count = &gfx_v9_4_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_4_reset_ras_error_count, + .query_ras_error_status = &gfx_v9_4_query_ras_error_status, +}; + +struct amdgpu_gfx_ras gfx_v9_4_ras = { + .ras_block = { + .hw_ops = &gfx_v9_4_ras_ops, + }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h index bdd16b568021..ca520a767267 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.h @@ -24,6 +24,6 @@ #ifndef __GFX_V9_4_H__ #define __GFX_V9_4_H__ -extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_ras_funcs; +extern struct amdgpu_gfx_ras gfx_v9_4_ras; #endif /* __GFX_V9_4_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c index c4f37a161875..7ec6243e015e 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c @@ -1641,14 +1641,14 @@ static int gfx_v9_4_2_query_utc_edc_count(struct amdgpu_device *adev, return 0; } -static int gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, +static void gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; uint32_t sec_count = 0, ded_count = 0; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) - return -EINVAL; + return; err_data->ue_count = 0; err_data->ce_count = 0; @@ -1661,7 +1661,6 @@ static int gfx_v9_4_2_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += sec_count; err_data->ue_count += ded_count; - return 0; } static void gfx_v9_4_2_reset_utc_err_status(struct amdgpu_device *adev) @@ -1931,13 +1930,17 @@ static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev) mutex_unlock(&adev->grbm_idx_mutex); } -const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs = { - .ras_late_init = amdgpu_gfx_ras_late_init, - .ras_fini = amdgpu_gfx_ras_fini, - .ras_error_inject = &gfx_v9_4_2_ras_error_inject, - .query_ras_error_count = &gfx_v9_4_2_query_ras_error_count, - .reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count, - .query_ras_error_status = &gfx_v9_4_2_query_ras_error_status, - .reset_ras_error_status = &gfx_v9_4_2_reset_ras_error_status, +struct amdgpu_ras_block_hw_ops gfx_v9_4_2_ras_ops ={ + .ras_error_inject = &gfx_v9_4_2_ras_error_inject, + .query_ras_error_count = &gfx_v9_4_2_query_ras_error_count, + .reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count, + .query_ras_error_status = &gfx_v9_4_2_query_ras_error_status, + .reset_ras_error_status = &gfx_v9_4_2_reset_ras_error_status, +}; + +struct amdgpu_gfx_ras gfx_v9_4_2_ras = { + .ras_block = { + .hw_ops = &gfx_v9_4_2_ras_ops, + }, .enable_watchdog_timer = &gfx_v9_4_2_enable_watchdog_timer, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h index 6db1f88509af..7584624b641c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h @@ -31,6 +31,6 @@ void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev, void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev); int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev); -extern const struct amdgpu_gfx_ras_funcs gfx_v9_4_2_ras_funcs; +extern struct amdgpu_gfx_ras gfx_v9_4_2_ras; #endif /* __GFX_V9_4_2_H__ */ -- cgit From 6c2453861f48e4e779cafa01c09e78ddc2c23c6b Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 18:56:20 +0800 Subject: drm/amdgpu: Modify xgmi block to fit for the unified ras block data and ops 1.Modify gmc block to fit for the unified ras block data and ops. 2.Change amdgpu_xgmi_ras_funcs to amdgpu_xgmi_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of gmc ras variable so that gmc ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register gmc ras block into amdgpu device ras block link list. 5.Remove the redundant code about gmc in amdgpu_ras.c after using the unified ras block. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 16 ++++++++-------- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 11 ++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 10 +++++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 26 ++++++++++++++++---------- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h | 4 ++-- 5 files changed, 37 insertions(+), 30 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 2430d6223c2d..d86ee530e0a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -454,12 +454,13 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) return r; } - if (!adev->gmc.xgmi.connected_to_cpu) - adev->gmc.xgmi.ras_funcs = &xgmi_ras_funcs; + if (!adev->gmc.xgmi.connected_to_cpu) { + adev->gmc.xgmi.ras = &xgmi_ras; + amdgpu_ras_register_ras_block(adev, &adev->gmc.xgmi.ras->ras_block); + } - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->ras_late_init) { - r = adev->gmc.xgmi.ras_funcs->ras_late_init(adev); + if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_late_init) { + r = adev->gmc.xgmi.ras->ras_block.ras_late_init(adev, NULL); if (r) return r; } @@ -505,9 +506,8 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) adev->mmhub.ras_funcs->ras_fini) adev->mmhub.ras_funcs->ras_fini(adev); - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->ras_fini) - adev->gmc.xgmi.ras_funcs->ras_fini(adev); + if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_fini) + adev->gmc.xgmi.ras->ras_block.ras_fini(adev); if (adev->hdp.ras_funcs && adev->hdp.ras_funcs->ras_fini) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 8458cebc6d5b..0001631cfedb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -29,6 +29,7 @@ #include #include "amdgpu_irq.h" +#include "amdgpu_ras.h" /* VA hole for 48bit addresses on Vega10 */ #define AMDGPU_GMC_HOLE_START 0x0000800000000000ULL @@ -135,12 +136,8 @@ struct amdgpu_gmc_funcs { unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev); }; -struct amdgpu_xgmi_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - int (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_xgmi_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_xgmi { @@ -159,7 +156,7 @@ struct amdgpu_xgmi { struct ras_common_if *ras_if; bool connected_to_cpu; bool pending_reset; - const struct amdgpu_xgmi_ras_funcs *ras_funcs; + struct amdgpu_xgmi_ras *ras; }; struct amdgpu_gmc { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index a5812c21177e..28997b7f7c95 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1012,9 +1012,13 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->nbio.ras_funcs->query_ras_error_count(adev, &err_data); break; case AMDGPU_RAS_BLOCK__XGMI_WAFL: - if (adev->gmc.xgmi.ras_funcs && - adev->gmc.xgmi.ras_funcs->query_ras_error_count) - adev->gmc.xgmi.ras_funcs->query_ras_error_count(adev, &err_data); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", + get_ras_block_str(&info->head)); + return -EINVAL; + } + if (block_obj->hw_ops->query_ras_error_count) + block_obj->hw_ops->query_ras_error_count(adev, &err_data); break; case AMDGPU_RAS_BLOCK__HDP: if (adev->hdp.ras_funcs && diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index e8b8f28c2f72..d29acd33eb11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -732,7 +732,7 @@ int amdgpu_xgmi_remove_device(struct amdgpu_device *adev) return psp_xgmi_terminate(&adev->psp); } -static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev) +static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_ih_if ih_info = { @@ -746,7 +746,7 @@ static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev) adev->gmc.xgmi.num_physical_nodes == 0) return 0; - adev->gmc.xgmi.ras_funcs->reset_ras_error_count(adev); + adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev); if (!adev->gmc.xgmi.ras_if) { adev->gmc.xgmi.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); @@ -865,7 +865,7 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev, return 0; } -static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, +static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -874,7 +874,7 @@ static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, uint32_t ue_cnt = 0, ce_cnt = 0; if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL)) - return -EINVAL; + return ; err_data->ue_count = 0; err_data->ce_count = 0; @@ -940,17 +940,23 @@ static int amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, break; } - adev->gmc.xgmi.ras_funcs->reset_ras_error_count(adev); + adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev); err_data->ue_count += ue_cnt; err_data->ce_count += ce_cnt; - - return 0; } -const struct amdgpu_xgmi_ras_funcs xgmi_ras_funcs = { - .ras_late_init = amdgpu_xgmi_ras_late_init, - .ras_fini = amdgpu_xgmi_ras_fini, +struct amdgpu_ras_block_hw_ops xgmi_ras_hw_ops = { .query_ras_error_count = amdgpu_xgmi_query_ras_error_count, .reset_ras_error_count = amdgpu_xgmi_reset_ras_error_count, }; + +struct amdgpu_xgmi_ras xgmi_ras = { + .ras_block = { + .name = "xgmi", + .block = AMDGPU_RAS_BLOCK__XGMI_WAFL, + .hw_ops = &xgmi_ras_hw_ops, + .ras_late_init = amdgpu_xgmi_ras_late_init, + .ras_fini = amdgpu_xgmi_ras_fini, + }, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h index d2189bf7d428..0afca51c3c0c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h @@ -24,7 +24,7 @@ #include #include "amdgpu_psp.h" - +#include "amdgpu_ras.h" struct amdgpu_hive_info { struct kobject kobj; @@ -50,7 +50,7 @@ struct amdgpu_pcs_ras_field { uint32_t pcs_err_shift; }; -extern const struct amdgpu_xgmi_ras_funcs xgmi_ras_funcs; +extern struct amdgpu_xgmi_ras xgmi_ras; struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev); void amdgpu_put_xgmi_hive(struct amdgpu_hive_info *hive); int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev); -- cgit From 6d76e9049ad92be32704106668c34493c3e4c0d4 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 18:57:58 +0800 Subject: drm/amdgpu: Modify hdp block to fit for the unified ras block data and ops 1.Modify hdp block to fit for the unified ras block data and ops. 2.Change amdgpu_hdp_ras_funcs to amdgpu_hdp_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of hdp ras variable so that hdp ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register hdp ras block into amdgpu device ras block link list. 5.Remove the redundant code about hdp in amdgpu_ras.c after using the unified ras block. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 10 ++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h | 13 +++++-------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 16 ++++++++-------- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 9 +++++---- drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c | 14 +++++++++++--- drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h | 2 +- 7 files changed, 35 insertions(+), 31 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index d86ee530e0a4..58cc4dae1246 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -465,9 +465,8 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) return r; } - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->ras_late_init) { - r = adev->hdp.ras_funcs->ras_late_init(adev); + if (adev->hdp.ras && adev->hdp.ras->ras_block.ras_late_init) { + r = adev->hdp.ras->ras_block.ras_late_init(adev, NULL); if (r) return r; } @@ -509,9 +508,8 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_fini) adev->gmc.xgmi.ras->ras_block.ras_fini(adev); - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->ras_fini) - adev->hdp.ras_funcs->ras_fini(adev); + if (adev->hdp.ras && adev->hdp.ras->ras_block.ras_fini) + adev->hdp.ras->ras_block.ras_fini(adev); } /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c index a766e1aad2b9..518966a26130 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c @@ -24,7 +24,7 @@ #include "amdgpu.h" #include "amdgpu_ras.h" -int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev) +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_ih_if ih_info = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h index 7ec99d591584..c94a4b3c8d6d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h @@ -22,13 +22,10 @@ */ #ifndef __AMDGPU_HDP_H__ #define __AMDGPU_HDP_H__ +#include "amdgpu_ras.h" -struct amdgpu_hdp_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_hdp_ras{ + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_hdp_funcs { @@ -43,9 +40,9 @@ struct amdgpu_hdp_funcs { struct amdgpu_hdp { struct ras_common_if *ras_if; const struct amdgpu_hdp_funcs *funcs; - const struct amdgpu_hdp_ras_funcs *ras_funcs; + struct amdgpu_hdp_ras *ras; }; -int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev); +int amdgpu_hdp_ras_late_init(struct amdgpu_device *adev, void *ras_info); void amdgpu_hdp_ras_fini(struct amdgpu_device *adev); #endif /* __AMDGPU_HDP_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 28997b7f7c95..c47a03252ec8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1012,6 +1012,7 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->nbio.ras_funcs->query_ras_error_count(adev, &err_data); break; case AMDGPU_RAS_BLOCK__XGMI_WAFL: + case AMDGPU_RAS_BLOCK__HDP: if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); @@ -1020,11 +1021,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, if (block_obj->hw_ops->query_ras_error_count) block_obj->hw_ops->query_ras_error_count(adev, &err_data); break; - case AMDGPU_RAS_BLOCK__HDP: - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->query_ras_error_count) - adev->hdp.ras_funcs->query_ras_error_count(adev, &err_data); - break; case AMDGPU_RAS_BLOCK__MCA: amdgpu_ras_mca_query_error_status(adev, &info->head, &err_data); break; @@ -1118,9 +1114,13 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, adev->sdma.funcs->reset_ras_error_count(adev); break; case AMDGPU_RAS_BLOCK__HDP: - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->reset_ras_error_count) - adev->hdp.ras_funcs->reset_ras_error_count(adev); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); + return -EINVAL; + } + + if (block_obj->hw_ops->reset_ras_error_count) + block_obj->hw_ops->reset_ras_error_count(adev); break; default: break; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 88c1eb9ad068..16ab572219ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1269,7 +1269,8 @@ static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) static void gmc_v9_0_set_hdp_ras_funcs(struct amdgpu_device *adev) { - adev->hdp.ras_funcs = &hdp_v4_0_ras_funcs; + adev->hdp.ras = &hdp_v4_0_ras; + amdgpu_ras_register_ras_block(adev, &adev->hdp.ras->ras_block); } static void gmc_v9_0_set_mca_funcs(struct amdgpu_device *adev) @@ -1346,9 +1347,9 @@ static int gmc_v9_0_late_init(void *handle) adev->mmhub.ras_funcs->reset_ras_error_count) adev->mmhub.ras_funcs->reset_ras_error_count(adev); - if (adev->hdp.ras_funcs && - adev->hdp.ras_funcs->reset_ras_error_count) - adev->hdp.ras_funcs->reset_ras_error_count(adev); + if (adev->hdp.ras && adev->hdp.ras->ras_block.hw_ops && + adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count) + adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count(adev); } r = amdgpu_gmc_ras_late_init(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index eecfb1545c1e..6b41fcbf4875 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -150,13 +150,21 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev) WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE_HI, (adev->gmc.vram_start >> 40)); } -const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs = { - .ras_late_init = amdgpu_hdp_ras_late_init, - .ras_fini = amdgpu_hdp_ras_fini, +struct amdgpu_ras_block_hw_ops hdp_v4_0_ras_hw_ops = { .query_ras_error_count = hdp_v4_0_query_ras_error_count, .reset_ras_error_count = hdp_v4_0_reset_ras_error_count, }; +struct amdgpu_hdp_ras hdp_v4_0_ras = { + .ras_block = { + .name = "hdp", + .block = AMDGPU_RAS_BLOCK__HDP, + .hw_ops = &hdp_v4_0_ras_hw_ops, + .ras_late_init = amdgpu_hdp_ras_late_init, + .ras_fini = amdgpu_hdp_ras_fini, + }, +}; + const struct amdgpu_hdp_funcs hdp_v4_0_funcs = { .flush_hdp = hdp_v4_0_flush_hdp, .invalidate_hdp = hdp_v4_0_invalidate_hdp, diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h index dc3a1b81dd62..c44eee9282ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.h @@ -27,6 +27,6 @@ #include "soc15_common.h" extern const struct amdgpu_hdp_funcs hdp_v4_0_funcs; -extern const struct amdgpu_hdp_ras_funcs hdp_v4_0_ras_funcs; +extern struct amdgpu_hdp_ras hdp_v4_0_ras; #endif -- cgit From 5e67bba301156c85251f49df19f5c695875814d1 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 4 Jan 2022 19:04:41 +0800 Subject: drm/amdgpu: Modify mmhub block to fit for the unified ras block data and ops 1.Modify mmhub block to fit for the unified ras block data and ops. 2.Change amdgpu_mmhub_ras_funcs to amdgpu_mmhub_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of mmhub ras variable so that mmhub ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register mmhub ras block into amdgpu device ras block link list. 5.Remove the redundant code about mmhub in amdgpu_ras.c after using the unified ras block. 5.Remove the redundant code about mmhub in amdgpu_ras.c after using the unified ras block. 6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of mmhub versions. If .ras_late_init and .ras_fini had been defined by the selected mmhub version, the defined functions will take effect; if not defined, default fill them with amdgpu_mmhub_ras_late_init and amdgpu_mmhub_ras_fini. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 12 ++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 10 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h | 14 +++------ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 47 +++++++++--------------------- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 27 +++++++++++++---- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 10 +++++-- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h | 2 +- drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c | 10 +++++-- drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h | 2 +- drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 10 +++++-- drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h | 2 +- 12 files changed, 73 insertions(+), 75 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9230e4476d91..71b814fe15f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3307,9 +3307,9 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work) if (adev->asic_reset_res) goto fail; - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops && + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev); } else { task_barrier_full(&hive->tb); @@ -4656,9 +4656,9 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle, if (!r && amdgpu_ras_intr_triggered()) { list_for_each_entry(tmp_adev, device_list_handle, reset_list) { - if (tmp_adev->mmhub.ras_funcs && - tmp_adev->mmhub.ras_funcs->reset_ras_error_count) - tmp_adev->mmhub.ras_funcs->reset_ras_error_count(tmp_adev); + if (tmp_adev->mmhub.ras && tmp_adev->mmhub.ras->ras_block.hw_ops && + tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(tmp_adev); } amdgpu_ras_intr_cleared(); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 58cc4dae1246..acf806c87673 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -447,9 +447,8 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) return r; } - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->ras_late_init) { - r = adev->mmhub.ras_funcs->ras_late_init(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.ras_late_init) { + r = adev->mmhub.ras->ras_block.ras_late_init(adev, NULL); if (r) return r; } @@ -501,9 +500,8 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) adev->umc.ras_funcs->ras_fini) adev->umc.ras_funcs->ras_fini(adev); - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->ras_fini) - adev->mmhub.ras_funcs->ras_fini(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.ras_fini) + adev->mmhub.ras->ras_block.ras_fini(adev); if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_fini) adev->gmc.xgmi.ras->ras_block.ras_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c index 24297dc51434..f9b5472a75d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.c @@ -24,7 +24,7 @@ #include "amdgpu.h" #include "amdgpu_ras.h" -int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev) +int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_ih_if ih_info = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h index b27fcbccce2b..7deda9a3b81e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h @@ -21,14 +21,8 @@ #ifndef __AMDGPU_MMHUB_H__ #define __AMDGPU_MMHUB_H__ -struct amdgpu_mmhub_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*query_ras_error_status)(struct amdgpu_device *adev); - void (*reset_ras_error_count)(struct amdgpu_device *adev); - void (*reset_ras_error_status)(struct amdgpu_device *adev); +struct amdgpu_mmhub_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_mmhub_funcs { @@ -50,10 +44,10 @@ struct amdgpu_mmhub_funcs { struct amdgpu_mmhub { struct ras_common_if *ras_if; const struct amdgpu_mmhub_funcs *funcs; - const struct amdgpu_mmhub_ras_funcs *ras_funcs; + struct amdgpu_mmhub_ras *ras; }; -int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev); +int amdgpu_mmhub_ras_late_init(struct amdgpu_device *adev, void *ras_info); void amdgpu_mmhub_ras_fini(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index c47a03252ec8..0ce26fd6abbd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -986,6 +986,7 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } break; case AMDGPU_RAS_BLOCK__GFX: + case AMDGPU_RAS_BLOCK__MMHUB: if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); @@ -997,15 +998,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, if (block_obj->hw_ops->query_ras_error_status) block_obj->hw_ops->query_ras_error_status(adev); break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_count) - adev->mmhub.ras_funcs->query_ras_error_count(adev, &err_data); - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_status) - adev->mmhub.ras_funcs->query_ras_error_status(adev); - break; case AMDGPU_RAS_BLOCK__PCIE_BIF: if (adev->nbio.ras_funcs && adev->nbio.ras_funcs->query_ras_error_count) @@ -1089,6 +1081,7 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, switch (block) { case AMDGPU_RAS_BLOCK__GFX: + case AMDGPU_RAS_BLOCK__MMHUB: if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); return -EINVAL; @@ -1100,15 +1093,6 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, if (block_obj->hw_ops->reset_ras_error_status) block_obj->hw_ops->reset_ras_error_status(adev); break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); - - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_status) - adev->mmhub.ras_funcs->reset_ras_error_status(adev); - break; case AMDGPU_RAS_BLOCK__SDMA: if (adev->sdma.funcs->reset_ras_error_count) adev->sdma.funcs->reset_ras_error_count(adev); @@ -1825,24 +1809,19 @@ static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, * Only two block need to query read/write * RspStatus at current state */ - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__GFX: - if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); - return ; - } + if ((info->head.block != AMDGPU_RAS_BLOCK__GFX) && + (info->head.block != AMDGPU_RAS_BLOCK__MMHUB)) + return ; - if (block_obj->hw_ops->query_ras_error_status) - block_obj->hw_ops->query_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__MMHUB: - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->query_ras_error_status) - adev->mmhub.ras_funcs->query_ras_error_status(adev); - break; - default: - break; + block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); + return ; } + + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); + } static void amdgpu_ras_query_err_status(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 16ab572219ed..7506e198e6e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1248,18 +1248,33 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev) { switch (adev->ip_versions[MMHUB_HWIP][0]) { case IP_VERSION(9, 4, 0): - adev->mmhub.ras_funcs = &mmhub_v1_0_ras_funcs; + adev->mmhub.ras = &mmhub_v1_0_ras; break; case IP_VERSION(9, 4, 1): - adev->mmhub.ras_funcs = &mmhub_v9_4_ras_funcs; + adev->mmhub.ras = &mmhub_v9_4_ras; break; case IP_VERSION(9, 4, 2): - adev->mmhub.ras_funcs = &mmhub_v1_7_ras_funcs; + adev->mmhub.ras = &mmhub_v1_7_ras; break; default: /* mmhub ras is not available */ break; } + + if (adev->mmhub.ras) { + amdgpu_ras_register_ras_block(adev, &adev->mmhub.ras->ras_block); + + strcpy(adev->mmhub.ras->ras_block.name,"mmhub"); + adev->mmhub.ras->ras_block.block = AMDGPU_RAS_BLOCK__MMHUB; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->mmhub.ras->ras_block.ras_late_init) + adev->mmhub.ras->ras_block.ras_late_init = amdgpu_mmhub_ras_late_init; + + /* If don't define special ras_fini function, use default ras_fini */ + if (!adev->mmhub.ras->ras_block.ras_fini) + adev->mmhub.ras->ras_block.ras_fini = amdgpu_mmhub_ras_fini; + } } static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev) @@ -1343,9 +1358,9 @@ static int gmc_v9_0_late_init(void *handle) } if (!amdgpu_persistent_edc_harvesting_supported(adev)) { - if (adev->mmhub.ras_funcs && - adev->mmhub.ras_funcs->reset_ras_error_count) - adev->mmhub.ras_funcs->reset_ras_error_count(adev); + if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops && + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count) + adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev); if (adev->hdp.ras && adev->hdp.ras->ras_block.hw_ops && adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 1da2ec692057..4c9f0c0f3116 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -774,13 +774,17 @@ static void mmhub_v1_0_reset_ras_error_count(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v1_0_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +struct amdgpu_ras_block_hw_ops mmhub_v1_0_ras_hw_ops = { .query_ras_error_count = mmhub_v1_0_query_ras_error_count, .reset_ras_error_count = mmhub_v1_0_reset_ras_error_count, }; +struct amdgpu_mmhub_ras mmhub_v1_0_ras = { + .ras_block = { + .hw_ops = &mmhub_v1_0_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v1_0_funcs = { .get_fb_location = mmhub_v1_0_get_fb_location, .init = mmhub_v1_0_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h index 4661b094e007..dae7ca48bd8b 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -24,6 +24,6 @@ #define __MMHUB_V1_0_H__ extern const struct amdgpu_mmhub_funcs mmhub_v1_0_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v1_0_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v1_0_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c index f5f7181f9af5..3b901f941627 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c @@ -1321,15 +1321,19 @@ static void mmhub_v1_7_reset_ras_error_status(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +struct amdgpu_ras_block_hw_ops mmhub_v1_7_ras_hw_ops = { .query_ras_error_count = mmhub_v1_7_query_ras_error_count, .reset_ras_error_count = mmhub_v1_7_reset_ras_error_count, .query_ras_error_status = mmhub_v1_7_query_ras_error_status, .reset_ras_error_status = mmhub_v1_7_reset_ras_error_status, }; +struct amdgpu_mmhub_ras mmhub_v1_7_ras = { + .ras_block = { + .hw_ops = &mmhub_v1_7_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs = { .get_fb_location = mmhub_v1_7_get_fb_location, .init = mmhub_v1_7_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h index a7f9dfc24697..629f49052137 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.h @@ -24,6 +24,6 @@ #define __MMHUB_V1_7_H__ extern const struct amdgpu_mmhub_funcs mmhub_v1_7_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v1_7_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v1_7_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c index ff49eeaf7882..619106f7d23d 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c @@ -1655,14 +1655,18 @@ static void mmhub_v9_4_query_ras_error_status(struct amdgpu_device *adev) } } -const struct amdgpu_mmhub_ras_funcs mmhub_v9_4_ras_funcs = { - .ras_late_init = amdgpu_mmhub_ras_late_init, - .ras_fini = amdgpu_mmhub_ras_fini, +const struct amdgpu_ras_block_hw_ops mmhub_v9_4_ras_hw_ops = { .query_ras_error_count = mmhub_v9_4_query_ras_error_count, .reset_ras_error_count = mmhub_v9_4_reset_ras_error_count, .query_ras_error_status = mmhub_v9_4_query_ras_error_status, }; +struct amdgpu_mmhub_ras mmhub_v9_4_ras = { + .ras_block = { + .hw_ops = &mmhub_v9_4_ras_hw_ops, + }, +}; + const struct amdgpu_mmhub_funcs mmhub_v9_4_funcs = { .get_fb_location = mmhub_v9_4_get_fb_location, .init = mmhub_v9_4_init, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h index 90436efa92ef..a48329d95f71 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.h @@ -24,6 +24,6 @@ #define __MMHUB_V9_4_H__ extern const struct amdgpu_mmhub_funcs mmhub_v9_4_funcs; -extern const struct amdgpu_mmhub_ras_funcs mmhub_v9_4_ras_funcs; +extern struct amdgpu_mmhub_ras mmhub_v9_4_ras; #endif -- cgit From 2e54fe5d056e7dc82988ef64ded3dca0ef954f0a Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 5 Jan 2022 14:28:10 +0800 Subject: drm/amdgpu: Modify nbio block to fit for the unified ras block data and ops 1.Modify nbio block to fit for the unified ras block data and ops. 2.Change amdgpu_nbio_ras_funcs to amdgpu_nbio_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of mmhub ras variable so that nbio ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register nbio ras block into amdgpu device ras block link list. 5.Remove the redundant code about nbio in amdgpu_ras.c after using the unified ras block. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 12 ++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h | 11 ++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 22 ++++++++++------------ drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c | 17 +++++++++++++---- drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h | 2 +- drivers/gpu/drm/amd/amdgpu/soc15.c | 18 ++++++++---------- 7 files changed, 43 insertions(+), 41 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index f5cbc2747ac6..ea3e8c66211f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -199,13 +199,13 @@ static irqreturn_t amdgpu_irq_handler(int irq, void *arg) * ack the interrupt if it is there */ if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) { - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->handle_ras_controller_intr_no_bifring) - adev->nbio.ras_funcs->handle_ras_controller_intr_no_bifring(adev); + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_controller_intr_no_bifring) + adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev); - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->handle_ras_err_event_athub_intr_no_bifring) - adev->nbio.ras_funcs->handle_ras_err_event_athub_intr_no_bifring(adev); + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring) + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring(adev); } return ret; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c index 6afb02fef8cf..6ace2e390e77 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c @@ -22,7 +22,7 @@ #include "amdgpu.h" #include "amdgpu_ras.h" -int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev) +int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_ih_if ih_info = { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h index 843052205bd5..4afb76d3cd97 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h @@ -47,15 +47,12 @@ struct nbio_hdp_flush_reg { u32 ref_and_mask_sdma7; }; -struct amdgpu_nbio_ras_funcs { +struct amdgpu_nbio_ras { + struct amdgpu_ras_block_object ras_block; void (*handle_ras_controller_intr_no_bifring)(struct amdgpu_device *adev); void (*handle_ras_err_event_athub_intr_no_bifring)(struct amdgpu_device *adev); int (*init_ras_controller_interrupt)(struct amdgpu_device *adev); int (*init_ras_err_event_athub_interrupt)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); }; struct amdgpu_nbio_funcs { @@ -104,9 +101,9 @@ struct amdgpu_nbio { struct amdgpu_irq_src ras_err_event_athub_irq; struct ras_common_if *ras_if; const struct amdgpu_nbio_funcs *funcs; - const struct amdgpu_nbio_ras_funcs *ras_funcs; + struct amdgpu_nbio_ras *ras; }; -int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev); +int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, void *ras_info); void amdgpu_nbio_ras_fini(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 0ce26fd6abbd..6d1ca9e9795b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -999,10 +999,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, block_obj->hw_ops->query_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__PCIE_BIF: - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->query_ras_error_count) - adev->nbio.ras_funcs->query_ras_error_count(adev, &err_data); - break; case AMDGPU_RAS_BLOCK__XGMI_WAFL: case AMDGPU_RAS_BLOCK__HDP: if (!block_obj || !block_obj->hw_ops) { @@ -2385,24 +2381,26 @@ int amdgpu_ras_init(struct amdgpu_device *adev) case CHIP_VEGA20: case CHIP_ARCTURUS: case CHIP_ALDEBARAN: - if (!adev->gmc.xgmi.connected_to_cpu) - adev->nbio.ras_funcs = &nbio_v7_4_ras_funcs; + if (!adev->gmc.xgmi.connected_to_cpu) { + adev->nbio.ras = &nbio_v7_4_ras; + amdgpu_ras_register_ras_block(adev, &adev->nbio.ras->ras_block); + } break; default: /* nbio ras is not available */ break; } - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_controller_interrupt) { - r = adev->nbio.ras_funcs->init_ras_controller_interrupt(adev); + if (adev->nbio.ras && + adev->nbio.ras->init_ras_controller_interrupt) { + r = adev->nbio.ras->init_ras_controller_interrupt(adev); if (r) goto release_con; } - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt) { - r = adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt(adev); + if (adev->nbio.ras && + adev->nbio.ras->init_ras_err_event_athub_interrupt) { + r = adev->nbio.ras->init_ras_err_event_athub_interrupt(adev); if (r) goto release_con; } diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index dc5e93756fea..39974b449341 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -658,16 +658,25 @@ static void nbio_v7_4_enable_doorbell_interrupt(struct amdgpu_device *adev, DOORBELL_INTERRUPT_DISABLE, enable ? 0 : 1); } -const struct amdgpu_nbio_ras_funcs nbio_v7_4_ras_funcs = { +const struct amdgpu_ras_block_hw_ops nbio_v7_4_ras_hw_ops = { + .query_ras_error_count = nbio_v7_4_query_ras_error_count, +}; + +struct amdgpu_nbio_ras nbio_v7_4_ras = { + .ras_block = { + .name = "pcie_bif", + .block = AMDGPU_RAS_BLOCK__PCIE_BIF, + .hw_ops = &nbio_v7_4_ras_hw_ops, + .ras_late_init = amdgpu_nbio_ras_late_init, + .ras_fini = amdgpu_nbio_ras_fini, + }, .handle_ras_controller_intr_no_bifring = nbio_v7_4_handle_ras_controller_intr_no_bifring, .handle_ras_err_event_athub_intr_no_bifring = nbio_v7_4_handle_ras_err_event_athub_intr_no_bifring, .init_ras_controller_interrupt = nbio_v7_4_init_ras_controller_interrupt, .init_ras_err_event_athub_interrupt = nbio_v7_4_init_ras_err_event_athub_interrupt, - .query_ras_error_count = nbio_v7_4_query_ras_error_count, - .ras_late_init = amdgpu_nbio_ras_late_init, - .ras_fini = amdgpu_nbio_ras_fini, }; + static void nbio_v7_4_program_ltr(struct amdgpu_device *adev) { uint32_t def, data; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h index cc5692db6f98..7490022d79d4 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h @@ -29,6 +29,6 @@ extern const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg; extern const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg_ald; extern const struct amdgpu_nbio_funcs nbio_v7_4_funcs; -extern const struct amdgpu_nbio_ras_funcs nbio_v7_4_ras_funcs; +extern struct amdgpu_nbio_ras nbio_v7_4_ras; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 0fc1747e4a70..6c8fcc4e29f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -1224,9 +1224,8 @@ static int soc15_common_late_init(void *handle) if (amdgpu_sriov_vf(adev)) xgpu_ai_mailbox_get_irq(adev); - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->ras_late_init) - r = adev->nbio.ras_funcs->ras_late_init(adev); + if (adev->nbio.ras && adev->nbio.ras->ras_block.ras_late_init) + r = adev->nbio.ras->ras_block.ras_late_init(adev, NULL); return r; } @@ -1249,9 +1248,8 @@ static int soc15_common_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->ras_fini) - adev->nbio.ras_funcs->ras_fini(adev); + if (adev->nbio.ras && adev->nbio.ras->ras_block.ras_fini) + adev->nbio.ras->ras_block.ras_fini(adev); if (adev->df.funcs && adev->df.funcs->sw_fini) @@ -1318,11 +1316,11 @@ static int soc15_common_hw_fini(void *handle) if (adev->nbio.ras_if && amdgpu_ras_is_supported(adev, adev->nbio.ras_if->block)) { - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_controller_interrupt) + if (adev->nbio.ras && + adev->nbio.ras->init_ras_controller_interrupt) amdgpu_irq_put(adev, &adev->nbio.ras_controller_irq, 0); - if (adev->nbio.ras_funcs && - adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt) + if (adev->nbio.ras && + adev->nbio.ras->init_ras_err_event_athub_interrupt) amdgpu_irq_put(adev, &adev->nbio.ras_err_event_athub_irq, 0); } -- cgit From efe17d5a217e6b7dfd16c80dab522abcf2edf1bc Mon Sep 17 00:00:00 2001 From: yipechai Date: Thu, 6 Jan 2022 14:07:44 +0800 Subject: drm/amdgpu: Modify umc block to fit for the unified ras block data and ops 1.Modify umc block to fit for the unified ras block data and ops. 2.Change amdgpu_umc_ras_funcs to amdgpu_umc_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of umc ras variable so that umc ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register umc ras block into amdgpu device ras block link list. 5.Remove the redundant code about umc in amdgpu_ras.c after using the unified ras block. 6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of umc versions. If .ras_late_init and .ras_fini had been defined by the selected umc version, the defined functions will take effect; if not defined, default fill them with amdgpu_umc_ras_late_init and amdgpu_umc_ras_fini. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 10 ++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 30 +++++++++++++++--------------- drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c | 32 ++++++++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h | 14 +++++--------- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 16 +++++++++++++++- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 21 ++++++++++++++++++--- drivers/gpu/drm/amd/amdgpu/umc_v6_1.c | 12 ++++++++---- drivers/gpu/drm/amd/amdgpu/umc_v6_1.h | 2 +- drivers/gpu/drm/amd/amdgpu/umc_v6_7.c | 10 +++++++--- drivers/gpu/drm/amd/amdgpu/umc_v6_7.h | 2 +- drivers/gpu/drm/amd/amdgpu/umc_v8_7.c | 12 ++++++++---- drivers/gpu/drm/amd/amdgpu/umc_v8_7.h | 2 +- 12 files changed, 99 insertions(+), 64 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index acf806c87673..de0ef21e1501 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -440,9 +440,8 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) { int r; - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ras_late_init) { - r = adev->umc.ras_funcs->ras_late_init(adev); + if (adev->umc.ras && adev->umc.ras->ras_block.ras_late_init) { + r = adev->umc.ras->ras_block.ras_late_init(adev, NULL); if (r) return r; } @@ -496,9 +495,8 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ras_fini) - adev->umc.ras_funcs->ras_fini(adev); + if (adev->umc.ras && adev->umc.ras->ras_block.ras_fini) + adev->umc.ras->ras_block.ras_fini(adev); if (adev->mmhub.ras && adev->mmhub.ras->ras_block.ras_fini) adev->mmhub.ras->ras_block.ras_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 6d1ca9e9795b..fba1c415a2a8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -939,24 +939,24 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d */ ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(ras->umc_ecc)); if (ret == -EOPNOTSUPP) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, err_data); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_count) + adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data); /* umc query_ras_error_address is also responsible for clearing * error status */ - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address) - adev->umc.ras_funcs->query_ras_error_address(adev, err_data); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_address) + adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, err_data); } else if (!ret) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_count) - adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, err_data); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_count) + adev->umc.ras->ecc_info_query_ras_error_count(adev, err_data); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_address) - adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, err_data); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_address) + adev->umc.ras->ecc_info_query_ras_error_address(adev, err_data); } } @@ -2412,12 +2412,12 @@ int amdgpu_ras_init(struct amdgpu_device *adev) } else if (adev->df.funcs && adev->df.funcs->query_ras_poison_mode && - adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_poison_mode) { + adev->umc.ras && + adev->umc.ras->query_ras_poison_mode) { df_poison = adev->df.funcs->query_ras_poison_mode(adev); umc_poison = - adev->umc.ras_funcs->query_ras_poison_mode(adev); + adev->umc.ras->query_ras_poison_mode(adev); /* Only poison is set in both DF and UMC, we can support it */ if (df_poison && umc_poison) con->poison_supported = true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index e81ce465ff3a..b4c68c09e071 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -35,12 +35,12 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc)); if (ret == -EOPNOTSUPP) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_count) - adev->umc.ras_funcs->query_ras_error_count(adev, ras_error_status); + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_count) + adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, ras_error_status); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->query_ras_error_address && + if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops && + adev->umc.ras->ras_block.hw_ops->query_ras_error_address && adev->umc.max_ras_err_cnt_per_query) { err_data->err_addr = kcalloc(adev->umc.max_ras_err_cnt_per_query, @@ -56,15 +56,15 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, /* umc query_ras_error_address is also responsible for clearing * error status */ - adev->umc.ras_funcs->query_ras_error_address(adev, ras_error_status); + adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, ras_error_status); } } else if (!ret) { - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_count) - adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, ras_error_status); + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_count) + adev->umc.ras->ecc_info_query_ras_error_count(adev, ras_error_status); - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->ecc_info_query_ras_error_address && + if (adev->umc.ras && + adev->umc.ras->ecc_info_query_ras_error_address && adev->umc.max_ras_err_cnt_per_query) { err_data->err_addr = kcalloc(adev->umc.max_ras_err_cnt_per_query, @@ -80,7 +80,7 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, /* umc query_ras_error_address is also responsible for clearing * error status */ - adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, ras_error_status); + adev->umc.ras->ecc_info_query_ras_error_address(adev, ras_error_status); } } @@ -136,7 +136,7 @@ static int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev, return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry, true); } -int amdgpu_umc_ras_late_init(struct amdgpu_device *adev) +int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; struct ras_fs_if fs_info = { @@ -172,9 +172,9 @@ int amdgpu_umc_ras_late_init(struct amdgpu_device *adev) } /* ras init of specific umc version */ - if (adev->umc.ras_funcs && - adev->umc.ras_funcs->err_cnt_init) - adev->umc.ras_funcs->err_cnt_init(adev); + if (adev->umc.ras && + adev->umc.ras->err_cnt_init) + adev->umc.ras->err_cnt_init(adev); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index b72194e8bfe5..195740a6d97d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -20,6 +20,7 @@ */ #ifndef __AMDGPU_UMC_H__ #define __AMDGPU_UMC_H__ +#include "amdgpu_ras.h" /* * (addr / 256) * 4096, the higher 26 bits in ErrorAddr @@ -40,14 +41,9 @@ #define LOOP_UMC_CH_INST(ch_inst) for ((ch_inst) = 0; (ch_inst) < adev->umc.channel_inst_num; (ch_inst)++) #define LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) LOOP_UMC_INST((umc_inst)) LOOP_UMC_CH_INST((ch_inst)) -struct amdgpu_umc_ras_funcs { +struct amdgpu_umc_ras { + struct amdgpu_ras_block_object ras_block; void (*err_cnt_init)(struct amdgpu_device *adev); - int (*ras_late_init)(struct amdgpu_device *adev); - void (*ras_fini)(struct amdgpu_device *adev); - void (*query_ras_error_count)(struct amdgpu_device *adev, - void *ras_error_status); - void (*query_ras_error_address)(struct amdgpu_device *adev, - void *ras_error_status); bool (*query_ras_poison_mode)(struct amdgpu_device *adev); void (*ecc_info_query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status); @@ -73,10 +69,10 @@ struct amdgpu_umc { struct ras_common_if *ras_if; const struct amdgpu_umc_funcs *funcs; - const struct amdgpu_umc_ras_funcs *ras_funcs; + struct amdgpu_umc_ras *ras; }; -int amdgpu_umc_ras_late_init(struct amdgpu_device *adev); +int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, void *ras_info); void amdgpu_umc_ras_fini(struct amdgpu_device *adev); int amdgpu_umc_poison_handler(struct amdgpu_device *adev, void *ras_error_status, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index 38bb42727715..5ef4ad28ab26 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -664,11 +664,25 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V8_7_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V8_7_PER_CHANNEL_OFFSET_SIENNA; adev->umc.channel_idx_tbl = &umc_v8_7_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v8_7_ras_funcs; + adev->umc.ras = &umc_v8_7_ras; break; default: break; } + if (adev->umc.ras) { + amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); + + strcpy(adev->umc.ras->ras_block.name,"umc"); + adev->umc.ras->ras_block.block = AMDGPU_RAS_BLOCK__UMC; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->umc.ras->ras_block.ras_late_init) + adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; + + /* If don't define special ras_fini function, use default ras_fini */ + if (!adev->umc.ras->ras_block.ras_fini) + adev->umc.ras->ras_block.ras_fini = amdgpu_umc_ras_fini; + } } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 7506e198e6e1..3965aae435f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1202,7 +1202,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V6_1_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_1_PER_CHANNEL_OFFSET_VG20; adev->umc.channel_idx_tbl = &umc_v6_1_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v6_1_ras_funcs; + adev->umc.ras = &umc_v6_1_ras; break; case IP_VERSION(6, 1, 2): adev->umc.max_ras_err_cnt_per_query = UMC_V6_1_TOTAL_CHANNEL_NUM; @@ -1210,7 +1210,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V6_1_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_1_PER_CHANNEL_OFFSET_ARCT; adev->umc.channel_idx_tbl = &umc_v6_1_channel_idx_tbl[0][0]; - adev->umc.ras_funcs = &umc_v6_1_ras_funcs; + adev->umc.ras = &umc_v6_1_ras; break; case IP_VERSION(6, 7, 0): adev->umc.max_ras_err_cnt_per_query = UMC_V6_7_TOTAL_CHANNEL_NUM; @@ -1218,7 +1218,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) adev->umc.umc_inst_num = UMC_V6_7_UMC_INSTANCE_NUM; adev->umc.channel_offs = UMC_V6_7_PER_CHANNEL_OFFSET; if (!adev->gmc.xgmi.connected_to_cpu) - adev->umc.ras_funcs = &umc_v6_7_ras_funcs; + adev->umc.ras = &umc_v6_7_ras; if (1 & adev->smuio.funcs->get_die_id(adev)) adev->umc.channel_idx_tbl = &umc_v6_7_channel_idx_tbl_first[0][0]; else @@ -1227,6 +1227,21 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) default: break; } + + if (adev->umc.ras) { + amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); + + strcpy(adev->umc.ras->ras_block.name,"umc"); + adev->umc.ras->ras_block.block = AMDGPU_RAS_BLOCK__UMC; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->umc.ras->ras_block.ras_late_init) + adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; + + /* If don't define special ras_fini function, use default ras_fini */ + if (!adev->umc.ras->ras_block.ras_fini) + adev->umc.ras->ras_block.ras_fini = amdgpu_umc_ras_fini; + } } static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c index 20b44983ac94..4776301972d4 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c @@ -465,10 +465,14 @@ static void umc_v6_1_err_cnt_init(struct amdgpu_device *adev) umc_v6_1_enable_umc_index_mode(adev); } -const struct amdgpu_umc_ras_funcs umc_v6_1_ras_funcs = { - .err_cnt_init = umc_v6_1_err_cnt_init, - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v6_1_ras_hw_ops = { .query_ras_error_count = umc_v6_1_query_ras_error_count, .query_ras_error_address = umc_v6_1_query_ras_error_address, }; + +struct amdgpu_umc_ras umc_v6_1_ras = { + .ras_block = { + .hw_ops = &umc_v6_1_ras_hw_ops, + }, + .err_cnt_init = umc_v6_1_err_cnt_init, +}; \ No newline at end of file diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h index 5dc36c730bb2..50c632eb4cc6 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.h @@ -45,7 +45,7 @@ /* umc ce count initial value */ #define UMC_V6_1_CE_CNT_INIT (UMC_V6_1_CE_CNT_MAX - UMC_V6_1_CE_INT_THRESHOLD) -extern const struct amdgpu_umc_ras_funcs umc_v6_1_ras_funcs; +extern struct amdgpu_umc_ras umc_v6_1_ras; extern const uint32_t umc_v6_1_channel_idx_tbl[UMC_V6_1_UMC_INSTANCE_NUM][UMC_V6_1_CHANNEL_INSTANCE_NUM]; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index 6dd1e19e8d43..6953426f0bed 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -480,11 +480,15 @@ static bool umc_v6_7_query_ras_poison_mode(struct amdgpu_device *adev) return true; } -const struct amdgpu_umc_ras_funcs umc_v6_7_ras_funcs = { - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v6_7_ras_hw_ops = { .query_ras_error_count = umc_v6_7_query_ras_error_count, .query_ras_error_address = umc_v6_7_query_ras_error_address, +}; + +struct amdgpu_umc_ras umc_v6_7_ras = { + .ras_block = { + .hw_ops = &umc_v6_7_ras_hw_ops, + }, .query_ras_poison_mode = umc_v6_7_query_ras_poison_mode, .ecc_info_query_ras_error_count = umc_v6_7_ecc_info_query_ras_error_count, .ecc_info_query_ras_error_address = umc_v6_7_ecc_info_query_ras_error_address, diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h index 57f2557e7aca..1f2edf625370 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.h @@ -43,7 +43,7 @@ #define UMC_V6_7_TOTAL_CHANNEL_NUM (UMC_V6_7_CHANNEL_INSTANCE_NUM * UMC_V6_7_UMC_INSTANCE_NUM) /* UMC regiser per channel offset */ #define UMC_V6_7_PER_CHANNEL_OFFSET 0x400 -extern const struct amdgpu_umc_ras_funcs umc_v6_7_ras_funcs; +extern struct amdgpu_umc_ras umc_v6_7_ras; extern const uint32_t umc_v6_7_channel_idx_tbl_second[UMC_V6_7_UMC_INSTANCE_NUM][UMC_V6_7_CHANNEL_INSTANCE_NUM]; extern const uint32_t diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c index af59a35788e3..ff9e1fac616d 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c @@ -324,10 +324,14 @@ static void umc_v8_7_err_cnt_init(struct amdgpu_device *adev) } } -const struct amdgpu_umc_ras_funcs umc_v8_7_ras_funcs = { - .err_cnt_init = umc_v8_7_err_cnt_init, - .ras_late_init = amdgpu_umc_ras_late_init, - .ras_fini = amdgpu_umc_ras_fini, +const struct amdgpu_ras_block_hw_ops umc_v8_7_ras_hw_ops = { .query_ras_error_count = umc_v8_7_query_ras_error_count, .query_ras_error_address = umc_v8_7_query_ras_error_address, }; + +struct amdgpu_umc_ras umc_v8_7_ras = { + .ras_block = { + .hw_ops = &umc_v8_7_ras_hw_ops, + }, + .err_cnt_init = umc_v8_7_err_cnt_init, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h index 37e6dc7c28e0..dd4993f5f78f 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.h @@ -44,7 +44,7 @@ /* umc ce count initial value */ #define UMC_V8_7_CE_CNT_INIT (UMC_V8_7_CE_CNT_MAX - UMC_V8_7_CE_INT_THRESHOLD) -extern const struct amdgpu_umc_ras_funcs umc_v8_7_ras_funcs; +extern struct amdgpu_umc_ras umc_v8_7_ras; extern const uint32_t umc_v8_7_channel_idx_tbl[UMC_V8_7_UMC_INSTANCE_NUM][UMC_V8_7_CHANNEL_INSTANCE_NUM]; -- cgit From bdc4292bd3b4337985f716be789b08eef921f7a6 Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 5 Jan 2022 15:30:37 +0800 Subject: drm/amdgpu: Modify sdma block to fit for the unified ras block data and ops 1.Modify sdma block to fit for the unified ras block data and ops. 2.Change amdgpu_sdma_ras_funcs to amdgpu_sdma_ras, and the corresponding variable name remove _funcs suffix. 3.Remove the const flag of sdma ras variable so that sdma ras block can be able to be inserted into amdgpu device ras block link list. 4.Invoke amdgpu_ras_register_ras_block function to register sdma ras block into amdgpu device ras block link list. 5.Remove the redundant code about sdma in amdgpu_ras.c after using the unified ras block. 6.Fill unified ras block .name .block .ras_late_init and .ras_fini for all of sdma versions. If .ras_late_init and .ras_fini had been defined by the selected sdma version, the defined functions will take effect; if not defined, default fill them with amdgpu_sdma_ras_late_init and amdgpu_sdma_ras_fini. v2: squash in warning fix (Alex) Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 10 ------ drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h | 12 +++---- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 58 +++++++++++++++++++++++++------- drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c | 25 +++++++++++--- drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h | 2 +- 5 files changed, 71 insertions(+), 36 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index fba1c415a2a8..7c21eab95fc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -967,7 +967,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct amdgpu_ras_block_object* block_obj = NULL; struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_err_data err_data = {0, 0, 0, NULL}; - int i; if (!obj) return -EINVAL; @@ -979,12 +978,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, amdgpu_ras_get_ecc_info(adev, &err_data); break; case AMDGPU_RAS_BLOCK__SDMA: - if (adev->sdma.funcs->query_ras_error_count) { - for (i = 0; i < adev->sdma.num_instances; i++) - adev->sdma.funcs->query_ras_error_count(adev, i, - &err_data); - } - break; case AMDGPU_RAS_BLOCK__GFX: case AMDGPU_RAS_BLOCK__MMHUB: if (!block_obj || !block_obj->hw_ops) { @@ -1090,9 +1083,6 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, block_obj->hw_ops->reset_ras_error_status(adev); break; case AMDGPU_RAS_BLOCK__SDMA: - if (adev->sdma.funcs->reset_ras_error_count) - adev->sdma.funcs->reset_ras_error_count(adev); - break; case AMDGPU_RAS_BLOCK__HDP: if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h index f8fb755e3aa6..eaee12ab6518 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -23,6 +23,7 @@ #ifndef __AMDGPU_SDMA_H__ #define __AMDGPU_SDMA_H__ +#include "amdgpu_ras.h" /* max number of IP instances */ #define AMDGPU_MAX_SDMA_INSTANCES 8 @@ -50,13 +51,8 @@ struct amdgpu_sdma_instance { bool burst_nop; }; -struct amdgpu_sdma_ras_funcs { - int (*ras_late_init)(struct amdgpu_device *adev, - void *ras_ih_info); - void (*ras_fini)(struct amdgpu_device *adev); - int (*query_ras_error_count)(struct amdgpu_device *adev, - uint32_t instance, void *ras_error_status); - void (*reset_ras_error_count)(struct amdgpu_device *adev); +struct amdgpu_sdma_ras { + struct amdgpu_ras_block_object ras_block; }; struct amdgpu_sdma { @@ -73,7 +69,7 @@ struct amdgpu_sdma { uint32_t srbm_soft_reset; bool has_page_queue; struct ras_common_if *ras_if; - const struct amdgpu_sdma_ras_funcs *funcs; + struct amdgpu_sdma_ras *ras; }; /* diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index e8e4749e9c79..3c1483dc113e 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1892,13 +1892,13 @@ static int sdma_v4_0_late_init(void *handle) sdma_v4_0_setup_ulv(adev); if (!amdgpu_persistent_edc_harvesting_supported(adev)) { - if (adev->sdma.funcs && - adev->sdma.funcs->reset_ras_error_count) - adev->sdma.funcs->reset_ras_error_count(adev); + if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops && + adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count) + adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count(adev); } - if (adev->sdma.funcs && adev->sdma.funcs->ras_late_init) - return adev->sdma.funcs->ras_late_init(adev, &ih_info); + if (adev->sdma.ras && adev->sdma.ras->ras_block.ras_late_init) + return adev->sdma.ras->ras_block.ras_late_init(adev, &ih_info); else return 0; } @@ -2001,8 +2001,9 @@ static int sdma_v4_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - if (adev->sdma.funcs && adev->sdma.funcs->ras_fini) - adev->sdma.funcs->ras_fini(adev); + if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops && + adev->sdma.ras->ras_block.ras_fini) + adev->sdma.ras->ras_block.ras_fini(adev); for (i = 0; i < adev->sdma.num_instances; i++) { amdgpu_ring_fini(&adev->sdma.instance[i].ring); @@ -2740,7 +2741,7 @@ static void sdma_v4_0_get_ras_error_count(uint32_t value, } } -static int sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, +static int sdma_v4_0_query_ras_error_count_by_instance(struct amdgpu_device *adev, uint32_t instance, void *ras_error_status) { struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status; @@ -2762,6 +2763,18 @@ static int sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, return 0; }; +static void sdma_v4_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) +{ + int i = 0; + for (i = 0; i < adev->sdma.num_instances; i++) { + if (sdma_v4_0_query_ras_error_count_by_instance(adev, i, ras_error_status)) + { + dev_err(adev->dev, "Query ras error count failed in SDMA%d \n", i); + return; + } + } +} + static void sdma_v4_0_reset_ras_error_count(struct amdgpu_device *adev) { int i; @@ -2773,26 +2786,45 @@ static void sdma_v4_0_reset_ras_error_count(struct amdgpu_device *adev) } } -static const struct amdgpu_sdma_ras_funcs sdma_v4_0_ras_funcs = { - .ras_late_init = amdgpu_sdma_ras_late_init, - .ras_fini = amdgpu_sdma_ras_fini, +const struct amdgpu_ras_block_hw_ops sdma_v4_0_ras_hw_ops = { .query_ras_error_count = sdma_v4_0_query_ras_error_count, .reset_ras_error_count = sdma_v4_0_reset_ras_error_count, }; +static struct amdgpu_sdma_ras sdma_v4_0_ras = { + .ras_block = { + .hw_ops = &sdma_v4_0_ras_hw_ops, + }, +}; + static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev) { switch (adev->ip_versions[SDMA0_HWIP][0]) { case IP_VERSION(4, 2, 0): case IP_VERSION(4, 2, 2): - adev->sdma.funcs = &sdma_v4_0_ras_funcs; + adev->sdma.ras = &sdma_v4_0_ras; break; case IP_VERSION(4, 4, 0): - adev->sdma.funcs = &sdma_v4_4_ras_funcs; + adev->sdma.ras = &sdma_v4_4_ras; break; default: break; } + + if (adev->sdma.ras) { + amdgpu_ras_register_ras_block(adev, &adev->sdma.ras->ras_block); + + strcpy(adev->sdma.ras->ras_block.name,"sdma"); + adev->sdma.ras->ras_block.block = AMDGPU_RAS_BLOCK__SDMA; + + /* If don't define special ras_late_init function, use default ras_late_init */ + if (!adev->sdma.ras->ras_block.ras_late_init) + adev->sdma.ras->ras_block.ras_late_init = amdgpu_sdma_ras_late_init; + + /* If don't define special ras_fini function, use default ras_fini */ + if (!adev->sdma.ras->ras_block.ras_fini) + adev->sdma.ras->ras_block.ras_fini = amdgpu_sdma_ras_fini; + } } const struct amdgpu_ip_block_version sdma_v4_0_ip_block = { diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c index bf95007f0843..5c1ba1116e5c 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.c @@ -188,7 +188,7 @@ static void sdma_v4_4_get_ras_error_count(struct amdgpu_device *adev, } } -static int sdma_v4_4_query_ras_error_count(struct amdgpu_device *adev, +static int sdma_v4_4_query_ras_error_count_by_instance(struct amdgpu_device *adev, uint32_t instance, void *ras_error_status) { @@ -245,9 +245,26 @@ static void sdma_v4_4_reset_ras_error_count(struct amdgpu_device *adev) } } -const struct amdgpu_sdma_ras_funcs sdma_v4_4_ras_funcs = { - .ras_late_init = amdgpu_sdma_ras_late_init, - .ras_fini = amdgpu_sdma_ras_fini, +static void sdma_v4_4_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status) +{ + int i = 0; + for (i = 0; i < adev->sdma.num_instances; i++) { + if (sdma_v4_4_query_ras_error_count_by_instance(adev, i, ras_error_status)) + { + dev_err(adev->dev, "Query ras error count failed in SDMA%d \n", i); + return; + } + } + +} + +const struct amdgpu_ras_block_hw_ops sdma_v4_4_ras_hw_ops = { .query_ras_error_count = sdma_v4_4_query_ras_error_count, .reset_ras_error_count = sdma_v4_4_reset_ras_error_count, }; + +struct amdgpu_sdma_ras sdma_v4_4_ras = { + .ras_block = { + .hw_ops = &sdma_v4_4_ras_hw_ops, + }, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h index 74a6e5b5e949..a9f0c68359e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4.h @@ -23,6 +23,6 @@ #ifndef __SDMA_V4_4_H__ #define __SDMA_V4_4_H__ -extern const struct amdgpu_sdma_ras_funcs sdma_v4_4_ras_funcs; +extern struct amdgpu_sdma_ras sdma_v4_4_ras; #endif -- cgit From 22d4ba53b1c10de6832e588f01d916e24306f6a1 Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 5 Jan 2022 15:40:26 +0800 Subject: drm/amdgpu: Adjust error inject function code style in amdgpu_ras.c 1. Move xgmi special error inject function from amdgpu_ras.c to xgmi block. 2. Support to use psp_ras_trigger_error as default error inject function in amdgpu_ras.c. If .ras_error_inject isn't defined in ras block, default error inject function will take effect. v2: squash in warning fix (Alex) Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 89 +++++--------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 27 ++++++++++ 2 files changed, 40 insertions(+), 76 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 7c21eab95fc8..bcd33e5350f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -903,31 +903,6 @@ static struct amdgpu_ras_block_object* amdgpu_ras_get_ras_block(struct amdgpu_de return NULL; } -static void amdgpu_ras_mca_query_error_status(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_err_data *err_data) -{ - switch (ras_block->sub_block_index) { - case AMDGPU_RAS_MCA_BLOCK__MP0: - if (adev->mca.mp0.ras_funcs && - adev->mca.mp0.ras_funcs->query_ras_error_count) - adev->mca.mp0.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_MCA_BLOCK__MP1: - if (adev->mca.mp1.ras_funcs && - adev->mca.mp1.ras_funcs->query_ras_error_count) - adev->mca.mp1.ras_funcs->query_ras_error_count(adev, &err_data); - break; - case AMDGPU_RAS_MCA_BLOCK__MPIO: - if (adev->mca.mpio.ras_funcs && - adev->mca.mpio.ras_funcs->query_ras_error_count) - adev->mca.mpio.ras_funcs->query_ras_error_count(adev, &err_data); - break; - default: - break; - } -} - static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_data *err_data) { struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); @@ -994,6 +969,7 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, case AMDGPU_RAS_BLOCK__PCIE_BIF: case AMDGPU_RAS_BLOCK__XGMI_WAFL: case AMDGPU_RAS_BLOCK__HDP: + case AMDGPU_RAS_BLOCK__MCA: if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); @@ -1002,9 +978,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, if (block_obj->hw_ops->query_ras_error_count) block_obj->hw_ops->query_ras_error_count(adev, &err_data); break; - case AMDGPU_RAS_BLOCK__MCA: - amdgpu_ras_mca_query_error_status(adev, &info->head, &err_data); - break; default: break; } @@ -1099,32 +1072,6 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, return 0; } -/* Trigger XGMI/WAFL error */ -static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev, - struct ta_ras_trigger_error_input *block_info) -{ - int ret; - - if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW)) - dev_warn(adev->dev, "Failed to disallow df cstate"); - - if (amdgpu_dpm_allow_xgmi_power_down(adev, false)) - dev_warn(adev->dev, "Failed to disallow XGMI power down"); - - ret = psp_ras_trigger_error(&adev->psp, block_info); - - if (amdgpu_ras_intr_triggered()) - return ret; - - if (amdgpu_dpm_allow_xgmi_power_down(adev, true)) - dev_warn(adev->dev, "Failed to allow XGMI power down"); - - if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW)) - dev_warn(adev->dev, "Failed to allow df cstate"); - - return ret; -} - /* wrapper of psp_ras_trigger_error */ int amdgpu_ras_error_inject(struct amdgpu_device *adev, struct ras_inject_if *info) @@ -1143,6 +1090,11 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, if (!obj) return -EINVAL; + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); + return -EINVAL; + } + /* Calculate XGMI relative offset */ if (adev->gmc.xgmi.num_physical_nodes > 1) { block_info.address = @@ -1150,30 +1102,15 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, block_info.address); } - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__GFX: - if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); - return -EINVAL; - } - + if (info->head.block == AMDGPU_RAS_BLOCK__GFX) { if (block_obj->hw_ops->ras_error_inject) ret = block_obj->hw_ops->ras_error_inject(adev, info); - break; - case AMDGPU_RAS_BLOCK__UMC: - case AMDGPU_RAS_BLOCK__SDMA: - case AMDGPU_RAS_BLOCK__MMHUB: - case AMDGPU_RAS_BLOCK__PCIE_BIF: - case AMDGPU_RAS_BLOCK__MCA: - ret = psp_ras_trigger_error(&adev->psp, &block_info); - break; - case AMDGPU_RAS_BLOCK__XGMI_WAFL: - ret = amdgpu_ras_error_inject_xgmi(adev, &block_info); - break; - default: - dev_info(adev->dev, "%s error injection is not supported yet\n", - get_ras_block_str(&info->head)); - ret = -EINVAL; + } else { + /* If defined special ras_error_inject(e.g: xgmi), implement special ras_error_inject */ + if (block_obj->hw_ops->ras_error_inject) + ret = block_obj->hw_ops->ras_error_inject(adev, &block_info); + else /*If not defined .ras_error_inject, use default ras_error_inject*/ + ret = psp_ras_trigger_error(&adev->psp, &block_info); } if (ret) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index d29acd33eb11..478457637d29 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -946,9 +946,36 @@ static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev, err_data->ce_count += ce_cnt; } +/* Trigger XGMI/WAFL error */ +static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev, void *inject_if) +{ + int ret = 0; + struct ta_ras_trigger_error_input *block_info = (struct ta_ras_trigger_error_input *)inject_if; + + if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW)) + dev_warn(adev->dev, "Failed to disallow df cstate"); + + if (amdgpu_dpm_allow_xgmi_power_down(adev, false)) + dev_warn(adev->dev, "Failed to disallow XGMI power down"); + + ret = psp_ras_trigger_error(&adev->psp, block_info); + + if (amdgpu_ras_intr_triggered()) + return ret; + + if (amdgpu_dpm_allow_xgmi_power_down(adev, true)) + dev_warn(adev->dev, "Failed to allow XGMI power down"); + + if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW)) + dev_warn(adev->dev, "Failed to allow df cstate"); + + return ret; +} + struct amdgpu_ras_block_hw_ops xgmi_ras_hw_ops = { .query_ras_error_count = amdgpu_xgmi_query_ras_error_count, .reset_ras_error_count = amdgpu_xgmi_reset_ras_error_count, + .ras_error_inject = amdgpu_ras_error_inject_xgmi, }; struct amdgpu_xgmi_ras xgmi_ras = { -- cgit From 7389a5b837cde5e5cd771619e9f006ae102f5d7d Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 5 Jan 2022 15:48:21 +0800 Subject: drm/amdgpu: Removed redundant ras code Removed redundant ras code. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: John Clements Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 67 ++++++++++----------------------- 1 file changed, 19 insertions(+), 48 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index bcd33e5350f8..517650d286a7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -946,40 +946,25 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, if (!obj) return -EINVAL; - block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); - - switch (info->head.block) { - case AMDGPU_RAS_BLOCK__UMC: + if (info->head.block == AMDGPU_RAS_BLOCK__UMC) { amdgpu_ras_get_ecc_info(adev, &err_data); - break; - case AMDGPU_RAS_BLOCK__SDMA: - case AMDGPU_RAS_BLOCK__GFX: - case AMDGPU_RAS_BLOCK__MMHUB: + } else { + block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); if (!block_obj || !block_obj->hw_ops) { dev_info(adev->dev, "%s doesn't config ras function \n", - get_ras_block_str(&info->head)); + get_ras_block_str(&info->head)); return -EINVAL; } - if (block_obj->hw_ops->query_ras_error_count) - block_obj->hw_ops->query_ras_error_count(adev, &err_data); - if (block_obj->hw_ops->query_ras_error_status) - block_obj->hw_ops->query_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__PCIE_BIF: - case AMDGPU_RAS_BLOCK__XGMI_WAFL: - case AMDGPU_RAS_BLOCK__HDP: - case AMDGPU_RAS_BLOCK__MCA: - if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", - get_ras_block_str(&info->head)); - return -EINVAL; - } if (block_obj->hw_ops->query_ras_error_count) block_obj->hw_ops->query_ras_error_count(adev, &err_data); - break; - default: - break; + + if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) || + (info->head.block == AMDGPU_RAS_BLOCK__GFX) || + (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) { + if (block_obj->hw_ops->query_ras_error_status) + block_obj->hw_ops->query_ras_error_status(adev); + } } obj->err_data.ue_count += err_data.ue_count; @@ -1041,32 +1026,18 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, if (!amdgpu_ras_is_supported(adev, block)) return -EINVAL; - switch (block) { - case AMDGPU_RAS_BLOCK__GFX: - case AMDGPU_RAS_BLOCK__MMHUB: - if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); - return -EINVAL; - } + if (!block_obj || !block_obj->hw_ops) { + dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); + return -EINVAL; + } - if (block_obj->hw_ops->reset_ras_error_count) - block_obj->hw_ops->reset_ras_error_count(adev); + if (block_obj->hw_ops->reset_ras_error_count) + block_obj->hw_ops->reset_ras_error_count(adev); + if ((block == AMDGPU_RAS_BLOCK__GFX) || + (block == AMDGPU_RAS_BLOCK__MMHUB)) { if (block_obj->hw_ops->reset_ras_error_status) block_obj->hw_ops->reset_ras_error_status(adev); - break; - case AMDGPU_RAS_BLOCK__SDMA: - case AMDGPU_RAS_BLOCK__HDP: - if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); - return -EINVAL; - } - - if (block_obj->hw_ops->reset_ras_error_count) - block_obj->hw_ops->reset_ras_error_count(adev); - break; - default: - break; } return 0; -- cgit From df01fe73ee98daf00c94189967550bd2d2098912 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 11 Jan 2022 16:45:19 +0800 Subject: drm/amdgpu: Add ras supported check for register_ras_block Add ras supported check for register_ras_block. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 517650d286a7..a1b4623ab1dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2756,6 +2756,9 @@ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, if (!adev || !ras_block_obj) return -EINVAL; + if (!amdgpu_ras_asic_supported(adev)) + return 0; + INIT_LIST_HEAD(&ras_block_obj->node); list_add_tail(&ras_block_obj->node, &adev->ras_list); -- cgit From df4f0041c6ef497e598a67e367db835489162754 Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 12 Jan 2022 10:19:07 +0800 Subject: drm/amdgpu: No longer insert ras blocks into ras_list if it already exists in ras_list No longer insert ras blocks into ras_list if it already exists in ras_list. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index a1b4623ab1dd..23f4290b2fde 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2753,12 +2753,20 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object* ras_block_obj) { + struct amdgpu_ras_block_object *obj, *tmp; if (!adev || !ras_block_obj) return -EINVAL; if (!amdgpu_ras_asic_supported(adev)) return 0; + /* If the ras object is in ras_list, don't add it again */ + list_for_each_entry_safe(obj, tmp, &adev->ras_list, node) { + if (obj == ras_block_obj) { + return 0; + } + } + INIT_LIST_HEAD(&ras_block_obj->node); list_add_tail(&ras_block_obj->node, &adev->ras_list); -- cgit From 69f91d32c6632e09f0954e690d61ac4921dacbd3 Mon Sep 17 00:00:00 2001 From: Yang Li Date: Thu, 13 Jan 2022 15:11:31 +0800 Subject: drm/amdgpu: remove unneeded semicolon Eliminate the following coccicheck warning: ./drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2725:16-17: Unneeded semicolon Reported-by: Abaci Robot Reviewed-by: Guchun Chen Signed-off-by: Yang Li Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 23f4290b2fde..64eed01d740c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2722,7 +2722,7 @@ struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev) int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras* ras_con) { if (!adev) - return -EINVAL;; + return -EINVAL; adev->psp.ras_context.ras = ras_con; return 0; -- cgit From ab3b9de65bfeded1d4646c9f66897c163e89abd8 Mon Sep 17 00:00:00 2001 From: Yang Li Date: Thu, 13 Jan 2022 15:11:32 +0800 Subject: drm/amdgpu: clean up some inconsistent indenting Eliminate the follow smatch warnings: drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:3504 amdgpu_device_init() warn: inconsistent indenting drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:1716 amdgpu_ras_error_status_query() warn: if statement not indented drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:1058 amdgpu_ras_error_inject() warn: inconsistent indenting Reported-by: Abaci Robot Reviewed-by: Guchun Chen Signed-off-by: Yang Li Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 8 +++++--- 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 71b814fe15f9..7dd7ef76c720 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3498,7 +3498,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->psp.mutex); mutex_init(&adev->notifier_lock); - amdgpu_device_init_apu_flags(adev); + amdgpu_device_init_apu_flags(adev); r = amdgpu_device_check_arguments(adev); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 64eed01d740c..d37b09b2d33a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1055,8 +1055,10 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, .address = info->address, .value = info->value, }; - int ret = -EINVAL; - struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); + int ret = -EINVAL; + struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, + info->head.block, + info->head.sub_block_index); if (!obj) return -EINVAL; @@ -1714,7 +1716,7 @@ static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, } if (block_obj->hw_ops->query_ras_error_status) - block_obj->hw_ops->query_ras_error_status(adev); + block_obj->hw_ops->query_ras_error_status(adev); } -- cgit From 954ea6aa1545a13036851327b4ed251fa7ab2f22 Mon Sep 17 00:00:00 2001 From: yipechai Date: Thu, 13 Jan 2022 11:01:17 +0800 Subject: drm/amdgpu: Use ARRAY_SIZE to get array length Use ARRAY_SIZE to get array length. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index d37b09b2d33a..9864f730b958 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -89,7 +89,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block) return ras_block_string[ras_block->block]; } -#define ras_block_str(_BLOCK_) (((_BLOCK_) < (sizeof(*ras_block_string)/sizeof(const char*))) ? ras_block_string[_BLOCK_] : "Out Of Range") +#define ras_block_str(_BLOCK_) \ + (((_BLOCK_) < ARRAY_SIZE(ras_block_string)) ? ras_block_string[_BLOCK_] : "Out Of Range") #define ras_err_str(i) (ras_error_string[ffs(i)]) -- cgit From e3d833f41c46b8c59c4af53897a6619bf667ebe5 Mon Sep 17 00:00:00 2001 From: yipechai Date: Thu, 13 Jan 2022 11:29:55 +0800 Subject: drm/amdgpu: fix compile warning for ras_block_match_default fix compile warning for ras_block_match_default Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 9864f730b958..0bb6b5354802 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -869,7 +869,8 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, } /* feature ctl end */ -int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object* block_obj, enum amdgpu_ras_block block) +static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_obj, + enum amdgpu_ras_block block) { if(!block_obj) return -EINVAL; -- cgit From 79c0462159a1fa3810ae1869a5fc9fd7782b6b70 Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Tue, 11 Jan 2022 14:14:50 +0800 Subject: drm/amdgpu: handle denied inject error into critical regions v2 Changed from v1: remove unused brace Signed-off-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 9 ++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 +- drivers/gpu/drm/amd/amdgpu/ta_ras_if.h | 3 ++- 3 files changed, 11 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index c984b5a34679..5c9b67ab168f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -1335,6 +1335,11 @@ static void psp_ras_ta_check_status(struct psp_context *psp) break; case TA_RAS_STATUS__SUCCESS: break; + case TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED: + if (ras_cmd->cmd_id == TA_RAS_COMMAND__TRIGGER_ERROR) + dev_warn(psp->adev->dev, + "RAS WARNING: Inject error to critical region is not allowed\n"); + break; default: dev_warn(psp->adev->dev, "RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status); @@ -1547,7 +1552,9 @@ int psp_ras_trigger_error(struct psp_context *psp, if (amdgpu_ras_intr_triggered()) return 0; - if (ras_cmd->ras_status) + if (ras_cmd->ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED) + return -EACCES; + else if (ras_cmd->ras_status) return -EINVAL; return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 0bb6b5354802..3538032e40d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -455,7 +455,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, } if (ret) - return -EINVAL; + return ret; return size; } diff --git a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h index 5093826a43d1..509d8a1945eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h +++ b/drivers/gpu/drm/amd/amdgpu/ta_ras_if.h @@ -64,7 +64,8 @@ enum ta_ras_status { TA_RAS_STATUS__ERROR_PCS_STATE_ERROR = 0xA016, TA_RAS_STATUS__ERROR_PCS_STATE_HANG = 0xA017, TA_RAS_STATUS__ERROR_PCS_STATE_UNKNOWN = 0xA018, - TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ = 0xA019 + TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ = 0xA019, + TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED = 0xA01A }; enum ta_ras_block { -- cgit From b6efdb02d23ef615464cd0077c211b40a1faca26 Mon Sep 17 00:00:00 2001 From: yipechai Date: Fri, 14 Jan 2022 10:24:30 +0800 Subject: drm/amdgpu: Fix the code style warnings in amdgpu_ras MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix the code style warnings in amdgpu_ras: 1. ERROR: space required before the open parenthesis '('. 2. WARNING: line length of xxx exceeds 100 columns. 3. ERROR: "foo* bar" should be "foo *bar". 4. WARNING: unnecessary whitespace before a quoted newline. 5. WARNING: space prohibited before semicolon. 6. WARNING: suspect code indent for conditional statements. 7. WARNING: braces {} are not necessary for single statement blocks. Signed-off-by: yipechai Reviewed-by: Tao Zhou Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 41 +++++++++++++++++++-------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 10 ++++---- 2 files changed, 30 insertions(+), 21 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 3538032e40d5..7a1d2bac698e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -872,7 +872,7 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_obj, enum amdgpu_ras_block block) { - if(!block_obj) + if (!block_obj) return -EINVAL; if (block_obj->block == block) @@ -881,7 +881,7 @@ static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_ return -EINVAL; } -static struct amdgpu_ras_block_object* amdgpu_ras_get_ras_block(struct amdgpu_device *adev, +static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_device *adev, enum amdgpu_ras_block block, uint32_t sub_block_index) { struct amdgpu_ras_block_object *obj, *tmp; @@ -941,7 +941,7 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) { - struct amdgpu_ras_block_object* block_obj = NULL; + struct amdgpu_ras_block_object *block_obj = NULL; struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_err_data err_data = {0, 0, 0, NULL}; @@ -953,7 +953,7 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } else { block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", + dev_info(adev->dev, "%s doesn't config ras function.\n", get_ras_block_str(&info->head)); return -EINVAL; } @@ -1023,13 +1023,14 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, enum amdgpu_ras_block block) { - struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, block, 0); + struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0); if (!amdgpu_ras_is_supported(adev, block)) return -EINVAL; if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", ras_block_str(block)); + dev_info(adev->dev, "%s doesn't config ras function.\n", + ras_block_str(block)); return -EINVAL; } @@ -1066,7 +1067,8 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, return -EINVAL; if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); + dev_info(adev->dev, "%s doesn't config ras function.\n", + get_ras_block_str(&info->head)); return -EINVAL; } @@ -1702,19 +1704,25 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, struct ras_query_if *info) { - struct amdgpu_ras_block_object* block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); + struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, + info->head.block, + info->head.sub_block_index); /* * Only two block need to query read/write * RspStatus at current state */ if ((info->head.block != AMDGPU_RAS_BLOCK__GFX) && (info->head.block != AMDGPU_RAS_BLOCK__MMHUB)) - return ; + return; + + block_obj = amdgpu_ras_get_ras_block(adev, + info->head.block, + info->head.sub_block_index); - block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, info->head.sub_block_index); if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function \n", get_ras_block_str(&info->head)); - return ; + dev_info(adev->dev, "%s doesn't config ras function.\n", + get_ras_block_str(&info->head)); + return; } if (block_obj->hw_ops->query_ras_error_status) @@ -2715,7 +2723,7 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev) } #endif -struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev) +struct amdgpu_ras *amdgpu_ras_get_context(struct amdgpu_device *adev) { if (!adev) return NULL; @@ -2723,7 +2731,7 @@ struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev) return adev->psp.ras_context.ras; } -int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras* ras_con) +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con) { if (!adev) return -EINVAL; @@ -2755,7 +2763,7 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) /* Register each ip ras block into amdgpu ras */ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, - struct amdgpu_ras_block_object* ras_block_obj) + struct amdgpu_ras_block_object *ras_block_obj) { struct amdgpu_ras_block_object *obj, *tmp; if (!adev || !ras_block_obj) @@ -2766,9 +2774,8 @@ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, /* If the ras object is in ras_list, don't add it again */ list_for_each_entry_safe(obj, tmp, &adev->ras_list, node) { - if (obj == ras_block_obj) { + if (obj == ras_block_obj) return 0; - } } INIT_LIST_HEAD(&ras_block_obj->node); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 7a4d82378205..a51a281bd91a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -496,7 +496,8 @@ struct amdgpu_ras_block_object { /* ras block link */ struct list_head node; - int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, enum amdgpu_ras_block block, uint32_t sub_block_index); + int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, + enum amdgpu_ras_block block, uint32_t sub_block_index); int (*ras_late_init)(struct amdgpu_device *adev, void *ras_info); void (*ras_fini)(struct amdgpu_device *adev); const struct amdgpu_ras_block_hw_ops *hw_ops; @@ -504,7 +505,7 @@ struct amdgpu_ras_block_object { struct amdgpu_ras_block_hw_ops { int (*ras_error_inject)(struct amdgpu_device *adev, void *inject_if); - void (*query_ras_error_count)(struct amdgpu_device *adev,void *ras_error_status); + void (*query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status); void (*query_ras_error_status)(struct amdgpu_device *adev); void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status); void (*reset_ras_error_count)(struct amdgpu_device *adev); @@ -678,7 +679,8 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev); struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev); -int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras* ras_con); +int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con); -int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object* ras_block_obj); +int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, + struct amdgpu_ras_block_object *ras_block_obj); #endif -- cgit From 8eb53bb2aa8afa170ba40f9460f2de4d4d138764 Mon Sep 17 00:00:00 2001 From: yipechai Date: Mon, 17 Jan 2022 11:23:59 +0800 Subject: drm/amdgpu: Remove repeated calls Remove repeated calls. Signed-off-by: yipechai Reviewed-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 7a1d2bac698e..4992bc554c0c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1704,9 +1704,7 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, struct ras_query_if *info) { - struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, - info->head.block, - info->head.sub_block_index); + struct amdgpu_ras_block_object *block_obj; /* * Only two block need to query read/write * RspStatus at current state -- cgit From e9287ef8d46cee538c9a71bb8978b2f7e975c452 Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 19 Jan 2022 16:49:42 +0800 Subject: Revert "drm/amdgpu: No longer insert ras blocks into ras_list if it already exists in ras_list" This reverts commit df4f0041c6ef497e598a67e367db835489162754. Xgmi ras initialization had been moved from .late_init to early_init, the defect of repeated calling amdgpu_ras_register_ras_block had been fixed, so revert this patch. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 4992bc554c0c..d4e07d0acb66 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2763,19 +2763,12 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object *ras_block_obj) { - struct amdgpu_ras_block_object *obj, *tmp; if (!adev || !ras_block_obj) return -EINVAL; if (!amdgpu_ras_asic_supported(adev)) return 0; - /* If the ras object is in ras_list, don't add it again */ - list_for_each_entry_safe(obj, tmp, &adev->ras_list, node) { - if (obj == ras_block_obj) - return 0; - } - INIT_LIST_HEAD(&ras_block_obj->node); list_add_tail(&ras_block_obj->node, &adev->ras_list); -- cgit From 400013b268cb666a44c0827b136bfd4bb741b13d Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Wed, 19 Jan 2022 15:42:55 +0800 Subject: drm/amdgpu: add umc_fill_error_record to make code more simple Create common amdgpu_umc_fill_error_record function for all versions of UMC and clean up related codes. Signed-off-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 23 +++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c | 21 +++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h | 5 +++++ drivers/gpu/drm/amd/amdgpu/umc_v6_1.c | 18 +++-------------- drivers/gpu/drm/amd/amdgpu/umc_v6_7.c | 36 ++++++--------------------------- drivers/gpu/drm/amd/amdgpu/umc_v8_7.c | 36 ++++++--------------------------- 6 files changed, 46 insertions(+), 93 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index d4e07d0acb66..e6324995fc54 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -158,14 +158,9 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre } memset(&err_rec, 0x0, sizeof(struct eeprom_table_record)); - - err_rec.address = address; - err_rec.retired_page = address >> AMDGPU_GPU_PAGE_SHIFT; - err_rec.ts = (uint64_t)ktime_get_real_seconds(); - err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_data.err_addr = &err_rec; - err_data.err_addr_cnt = 1; + amdgpu_umc_fill_error_record(&err_data, address, + (address >> AMDGPU_GPU_PAGE_SHIFT), 0, 0); if (amdgpu_bad_page_threshold != 0) { amdgpu_ras_add_bad_pages(adev, err_data.err_addr, @@ -2660,8 +2655,6 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, dev_info(adev->dev, "Uncorrectable error detected in UMC inst: %d, chan_idx: %d", umc_inst, ch_inst); - memset(&err_rec, 0x0, sizeof(struct eeprom_table_record)); - /* * Translate UMC channel address to Physical address */ @@ -2673,16 +2666,10 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb, ADDR_OF_256B_BLOCK(channel_index) | OFFSET_IN_256B_BLOCK(m->addr); - err_rec.address = m->addr; - err_rec.retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec.ts = (uint64_t)ktime_get_real_seconds(); - err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec.cu = 0; - err_rec.mem_channel = channel_index; - err_rec.mcumc_id = umc_inst; - + memset(&err_rec, 0x0, sizeof(struct eeprom_table_record)); err_data.err_addr = &err_rec; - err_data.err_addr_cnt = 1; + amdgpu_umc_fill_error_record(&err_data, m->addr, + retired_page, channel_index, umc_inst); if (amdgpu_bad_page_threshold != 0) { amdgpu_ras_add_bad_pages(adev, err_data.err_addr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index b4c68c09e071..ff7805beda38 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -218,3 +218,24 @@ int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, amdgpu_ras_interrupt_dispatch(adev, &ih_data); return 0; } + +void amdgpu_umc_fill_error_record(struct ras_err_data *err_data, + uint64_t err_addr, + uint64_t retired_page, + uint32_t channel_index, + uint32_t umc_inst) +{ + struct eeprom_table_record *err_rec = + &err_data->err_addr[err_data->err_addr_cnt]; + + err_rec->address = err_addr; + /* page frame address is saved */ + err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; + err_rec->ts = (uint64_t)ktime_get_real_seconds(); + err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; + err_rec->cu = 0; + err_rec->mem_channel = channel_index; + err_rec->mcumc_id = umc_inst; + + err_data->err_addr_cnt++; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h index 195740a6d97d..4db0526d0be4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h @@ -80,4 +80,9 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev, int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry); +void amdgpu_umc_fill_error_record(struct ras_err_data *err_data, + uint64_t err_addr, + uint64_t retired_page, + uint32_t channel_index, + uint32_t umc_inst); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c index 4776301972d4..939cb203f7ad 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c @@ -300,7 +300,6 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, { uint32_t lsb, mc_umc_status_addr; uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; uint32_t channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; if (adev->asic_type == CHIP_ARCTURUS) { @@ -328,8 +327,6 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - /* calculate error address if ue/ce error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -348,18 +345,9 @@ static void umc_v6_1_query_error_address(struct amdgpu_device *adev, /* we only save ue error information currently, ce is skipped */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + == 1) + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } /* clear umc status */ diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c index f5a1ba7db75a..300dee9ec6b4 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c @@ -120,7 +120,6 @@ static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, uint32_t umc_inst) { uint64_t mc_umc_status, err_addr, retired_page; - struct eeprom_table_record *err_rec; uint32_t channel_index; uint32_t eccinfo_table_idx; struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); @@ -137,8 +136,6 @@ static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, if (!err_data->err_addr) return; - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - /* calculate error address if ue/ce error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -154,18 +151,9 @@ static void umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev, /* we only save ue error information currently, ce is skipped */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + == 1) + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } } @@ -345,7 +333,6 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev, { uint32_t mc_umc_status_addr; uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; uint32_t channel_index; mc_umc_status_addr = @@ -364,8 +351,6 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; @@ -384,18 +369,9 @@ static void umc_v6_7_query_error_address(struct amdgpu_device *adev, /* we only save ue error information currently, ce is skipped */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + == 1) + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } /* clear umc status */ diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c index d70417196662..de85a998ef99 100644 --- a/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c +++ b/drivers/gpu/drm/amd/amdgpu/umc_v8_7.c @@ -121,7 +121,6 @@ static void umc_v8_7_ecc_info_query_error_address(struct amdgpu_device *adev, uint32_t umc_inst) { uint64_t mc_umc_status, err_addr, retired_page; - struct eeprom_table_record *err_rec; uint32_t channel_index; uint32_t eccinfo_table_idx; struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); @@ -138,8 +137,6 @@ static void umc_v8_7_ecc_info_query_error_address(struct amdgpu_device *adev, if (!err_data->err_addr) return; - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - /* calculate error address if ue/ce error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -155,18 +152,9 @@ static void umc_v8_7_ecc_info_query_error_address(struct amdgpu_device *adev, /* we only save ue error information currently, ce is skipped */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + == 1) + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } } @@ -344,7 +332,6 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, { uint32_t lsb, mc_umc_status_addr; uint64_t mc_umc_status, err_addr, retired_page, mc_umc_addrt0; - struct eeprom_table_record *err_rec; uint32_t channel_index = adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + ch_inst]; mc_umc_status_addr = @@ -363,8 +350,6 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, return; } - err_rec = &err_data->err_addr[err_data->err_addr_cnt]; - /* calculate error address if ue/ce error is detected */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 && (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -383,18 +368,9 @@ static void umc_v8_7_query_error_address(struct amdgpu_device *adev, /* we only save ue error information currently, ce is skipped */ if (REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, UECC) - == 1) { - err_rec->address = err_addr; - /* page frame address is saved */ - err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT; - err_rec->ts = (uint64_t)ktime_get_real_seconds(); - err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE; - err_rec->cu = 0; - err_rec->mem_channel = channel_index; - err_rec->mcumc_id = umc_inst; - - err_data->err_addr_cnt++; - } + == 1) + amdgpu_umc_fill_error_record(err_data, err_addr, + retired_page, channel_index, umc_inst); } /* clear umc status */ -- cgit From a2170b4af62fd347c699ae3cd9ec0a8096466f9d Mon Sep 17 00:00:00 2001 From: yipechai Date: Sat, 29 Jan 2022 11:20:50 +0800 Subject: drm/amdgpu: Add judgement to avoid infinite loop 1. The infinite loop causing soft lock occurs on multiple amdgpu cards supporting ras feature. 2. This a workaround patch to fix 6492e1b07c03397f85bd6dc0e230ea6cd9394635. It is valid for multiple amdgpu cards of the same type. 3. The root cause is that each GPU card device has a separate .ras_list link header, but the instance and linked list node of each ras block are unique. When each device is initialized, each ras instance will repeatedly add link node to the device every time. In this way, only the .ras_list of the last initialized device is completely correct. the .ras_list->prev and .ras_list->next of the device initialzied before can still point to the correct ras instance, but the prev pointer and next pointer of the pointed ras instance both point to the last initialized device's .ras_ list instead of the beginning .ras_ list. When using list_for_each_entry_safe searches for non-existent Ras nodes on devices other than the last device, the last ras instance next pointer cannot always be equal to the beginning .ras_list, so that the loop cannot be terminated, the program enters a infinite loop. BTW: Since the data and initialization process of each card are the same, the link list between ras instances will not be destroyed every time the device is initialized. 4. The soft locked logs are as follows: [ 262.165690] CPU: 93 PID: 758 Comm: kworker/93:1 Tainted: G OE 5.13.0-27-generic #29~20.04.1-Ubuntu [ 262.165695] Hardware name: Supermicro AS -4124GS-TNR/H12DSG-O-CPU, BIOS T20200717143848 07/17/2020 [ 262.165698] Workqueue: events amdgpu_ras_do_recovery [amdgpu] [ 262.165980] RIP: 0010:amdgpu_ras_get_ras_block+0x86/0xd0 [amdgpu] [ 262.166239] Code: 68 d8 4c 8d 71 d8 48 39 c3 74 54 49 8b 45 38 48 85 c0 74 32 44 89 fa 44 89 e6 4c 89 ef e8 82 e4 9b dc 85 c0 74 3c 49 8b 46 28 <49> 8d 56 28 4d 89 f5 48 83 e8 28 48 39 d3 74 25 49 89 c6 49 8b 45 [ 262.166243] RSP: 0018:ffffac908fa87d80 EFLAGS: 00000202 [ 262.166247] RAX: ffffffffc1394248 RBX: ffff91e4ab8d6e20 RCX: ffffffffc1394248 [ 262.166249] RDX: ffff91e4aa356e20 RSI: 000000000000000e RDI: ffff91e4ab8c0000 [ 262.166252] RBP: ffffac908fa87da8 R08: 0000000000000007 R09: 0000000000000001 [ 262.166254] R10: ffff91e4930b64ec R11: 0000000000000000 R12: 000000000000000e [ 262.166256] R13: ffff91e4aa356df8 R14: ffffffffc1394320 R15: 0000000000000003 [ 262.166258] FS: 0000000000000000(0000) GS:ffff92238fb40000(0000) knlGS:0000000000000000 [ 262.166261] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 262.166264] CR2: 00000001004865d0 CR3: 000000406d796000 CR4: 0000000000350ee0 [ 262.166267] Call Trace: [ 262.166272] amdgpu_ras_do_recovery+0x130/0x290 [amdgpu] [ 262.166529] ? psi_task_switch+0xd2/0x250 [ 262.166537] ? __switch_to+0x11d/0x460 [ 262.166542] ? __switch_to_asm+0x36/0x70 [ 262.166549] process_one_work+0x220/0x3c0 [ 262.166556] worker_thread+0x4d/0x3f0 [ 262.166560] ? process_one_work+0x3c0/0x3c0 [ 262.166563] kthread+0x12b/0x150 [ 262.166568] ? set_kthread_struct+0x40/0x40 [ 262.166571] ret_from_fork+0x22/0x30 Fixes: 6492e1b07c0339 ("drm/amdgpu: Unify ras block interface for each ras block") Signed-off-by: yipechai Reviewed-by: John Clements Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index e6324995fc54..9d7c778c1a2d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -879,6 +879,7 @@ static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_device *adev, enum amdgpu_ras_block block, uint32_t sub_block_index) { + int loop_cnt = 0; struct amdgpu_ras_block_object *obj, *tmp; if (block >= AMDGPU_RAS_BLOCK__LAST) @@ -895,6 +896,9 @@ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_de if (amdgpu_ras_block_match_default(obj, block) == 0) return obj; } + + if (++loop_cnt >= AMDGPU_RAS_BLOCK__LAST) + break; } return NULL; -- cgit From afa37315917bbc9e71a4359f921eb887470c008c Mon Sep 17 00:00:00 2001 From: Luben Tuikov Date: Thu, 3 Feb 2022 17:05:12 -0500 Subject: drm/amdgpu: Print once if RAS unsupported MESA polls for errors every 2-3 seconds. Printing with dev_info() causes the dmesg log to fill up with the same message, e.g, [18028.206676] amdgpu 0000:0b:00.0: amdgpu: df doesn't config ras function. Make it dev_dbg_once(), as it isn't something correctible during boot or thereafter, so printing just once is sufficient. Also sanitize the message. Cc: Alex Deucher Cc: Hawking Zhang Cc: John Clements Cc: Tao Zhou Cc: yipechai Fixes: 8b0fb0e967c1 ("drm/amdgpu: Modify gfx block to fit for the unified ras block data and ops") Signed-off-by: Luben Tuikov Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 9d7c778c1a2d..e440a5268ace 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -952,8 +952,8 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } else { block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0); if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function.\n", - get_ras_block_str(&info->head)); + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); return -EINVAL; } @@ -1028,8 +1028,8 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, return -EINVAL; if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function.\n", - ras_block_str(block)); + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + ras_block_str(block)); return -EINVAL; } @@ -1066,8 +1066,8 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, return -EINVAL; if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function.\n", - get_ras_block_str(&info->head)); + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); return -EINVAL; } @@ -1717,8 +1717,8 @@ static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, info->head.sub_block_index); if (!block_obj || !block_obj->hw_ops) { - dev_info(adev->dev, "%s doesn't config ras function.\n", - get_ras_block_str(&info->head)); + dev_dbg_once(adev->dev, "%s doesn't config RAS function\n", + get_ras_block_str(&info->head)); return; } -- cgit From d5e8ff5f7b2a41d503914d4896ed3c6b3befe933 Mon Sep 17 00:00:00 2001 From: yipechai Date: Sat, 29 Jan 2022 17:09:08 +0800 Subject: drm/amdgpu: Fixed the defect of soft lock caused by infinite loop 1. The infinite loop case only occurs on multiple cards support ras functions. 2. The explanation of root cause refer to commit 76641cbbf196 ("drm/amdgpu: Add judgement to avoid infinite loop"). 3. Create new node to manage each unique ras instance to guarantee each device .ras_list is completely independent. 4. Fixes: commit 7a6b8ab3231b51 ("drm/amdgpu: Unify ras block interface for each ras block"). 5. The soft locked logs are as follows: [ 262.165690] CPU: 93 PID: 758 Comm: kworker/93:1 Tainted: G OE 5.13.0-27-generic #29~20.04.1-Ubuntu [ 262.165695] Hardware name: Supermicro AS -4124GS-TNR/H12DSG-O-CPU, BIOS T20200717143848 07/17/2020 [ 262.165698] Workqueue: events amdgpu_ras_do_recovery [amdgpu] [ 262.165980] RIP: 0010:amdgpu_ras_get_ras_block+0x86/0xd0 [amdgpu] [ 262.166239] Code: 68 d8 4c 8d 71 d8 48 39 c3 74 54 49 8b 45 38 48 85 c0 74 32 44 89 fa 44 89 e6 4c 89 ef e8 82 e4 9b dc 85 c0 74 3c 49 8b 46 28 <49> 8d 56 28 4d 89 f5 48 83 e8 28 48 39 d3 74 25 49 89 c6 49 8b 45 [ 262.166243] RSP: 0018:ffffac908fa87d80 EFLAGS: 00000202 [ 262.166247] RAX: ffffffffc1394248 RBX: ffff91e4ab8d6e20 RCX: ffffffffc1394248 [ 262.166249] RDX: ffff91e4aa356e20 RSI: 000000000000000e RDI: ffff91e4ab8c0000 [ 262.166252] RBP: ffffac908fa87da8 R08: 0000000000000007 R09: 0000000000000001 [ 262.166254] R10: ffff91e4930b64ec R11: 0000000000000000 R12: 000000000000000e [ 262.166256] R13: ffff91e4aa356df8 R14: ffffffffc1394320 R15: 0000000000000003 [ 262.166258] FS: 0000000000000000(0000) GS:ffff92238fb40000(0000) knlGS:0000000000000000 [ 262.166261] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 262.166264] CR2: 00000001004865d0 CR3: 000000406d796000 CR4: 0000000000350ee0 [ 262.166267] Call Trace: [ 262.166272] amdgpu_ras_do_recovery+0x130/0x290 [amdgpu] [ 262.166529] ? psi_task_switch+0xd2/0x250 [ 262.166537] ? __switch_to+0x11d/0x460 [ 262.166542] ? __switch_to_asm+0x36/0x70 [ 262.166549] process_one_work+0x220/0x3c0 [ 262.166556] worker_thread+0x4d/0x3f0 [ 262.166560] ? process_one_work+0x3c0/0x3c0 [ 262.166563] kthread+0x12b/0x150 [ 262.166568] ? set_kthread_struct+0x40/0x40 [ 262.166571] ret_from_fork+0x22/0x30 Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 35 +++++++++++++++++++++++++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 3 --- 2 files changed, 31 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index e440a5268ace..d6608ae3c127 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -75,6 +75,13 @@ const char *ras_mca_block_string[] = { "mca_iohc", }; +struct amdgpu_ras_block_list { + /* ras block link */ + struct list_head node; + + struct amdgpu_ras_block_object *ras_obj; +}; + const char *get_ras_block_str(struct ras_common_if *ras_block) { if (!ras_block) @@ -880,7 +887,8 @@ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_de enum amdgpu_ras_block block, uint32_t sub_block_index) { int loop_cnt = 0; - struct amdgpu_ras_block_object *obj, *tmp; + struct amdgpu_ras_block_list *node, *tmp; + struct amdgpu_ras_block_object *obj; if (block >= AMDGPU_RAS_BLOCK__LAST) return NULL; @@ -888,7 +896,13 @@ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_de if (!amdgpu_ras_is_supported(adev, block)) return NULL; - list_for_each_entry_safe(obj, tmp, &adev->ras_list, node) { + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { + if (!node->ras_obj) { + dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); + continue; + } + + obj = node->ras_obj; if (obj->ras_block_match) { if (obj->ras_block_match(obj, block, sub_block_index) == 0) return obj; @@ -2527,6 +2541,7 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) int amdgpu_ras_fini(struct amdgpu_device *adev) { + struct amdgpu_ras_block_list *ras_node, *tmp; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) @@ -2545,6 +2560,12 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) amdgpu_ras_set_context(adev, NULL); kfree(con); + /* Clear ras blocks from ras_list and free ras block list node */ + list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) { + list_del(&ras_node->node); + kfree(ras_node); + } + return 0; } @@ -2754,14 +2775,20 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev) int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object *ras_block_obj) { + struct amdgpu_ras_block_list *ras_node; if (!adev || !ras_block_obj) return -EINVAL; if (!amdgpu_ras_asic_supported(adev)) return 0; - INIT_LIST_HEAD(&ras_block_obj->node); - list_add_tail(&ras_block_obj->node, &adev->ras_list); + ras_node = kzalloc(sizeof(*ras_node), GFP_KERNEL); + if (!ras_node) + return -ENOMEM; + + INIT_LIST_HEAD(&ras_node->node); + ras_node->ras_obj = ras_block_obj; + list_add_tail(&ras_node->node, &adev->ras_list); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index a51a281bd91a..a55743b12d57 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -493,9 +493,6 @@ struct amdgpu_ras_block_object { uint32_t sub_block_index; - /* ras block link */ - struct list_head node; - int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, enum amdgpu_ras_block block, uint32_t sub_block_index); int (*ras_late_init)(struct amdgpu_device *adev, void *ras_info); -- cgit From a50b048276c4e1bc6f7e869c99b6bdc91b4e237f Mon Sep 17 00:00:00 2001 From: yipechai Date: Mon, 7 Feb 2022 11:28:28 +0800 Subject: Revert "drm/amdgpu: Add judgement to avoid infinite loop" The commit d5e8ff5f7b2a ("drm/amdgpu: Fixed the defect of soft lock caused by infinite loop") had fixed this defect. Revert workaround commit a2170b4af62f ("drm/amdgpu: Add judgement to avoid infinite loop"). Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index d6608ae3c127..5934326b9db3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -886,7 +886,6 @@ static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_device *adev, enum amdgpu_ras_block block, uint32_t sub_block_index) { - int loop_cnt = 0; struct amdgpu_ras_block_list *node, *tmp; struct amdgpu_ras_block_object *obj; @@ -910,9 +909,6 @@ static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_de if (amdgpu_ras_block_match_default(obj, block) == 0) return obj; } - - if (++loop_cnt >= AMDGPU_RAS_BLOCK__LAST) - break; } return NULL; -- cgit From bdb3489cfca16815e9a737359e9e90a4af5d0ff3 Mon Sep 17 00:00:00 2001 From: yipechai Date: Sun, 30 Jan 2022 17:03:32 +0800 Subject: drm/amdgpu: Optimize xxx_ras_late_init/xxx_ras_late_fini for each ras block 1. Define amdgpu_ras_block_late_init to create sysfs nodes and interrupt handles. 2. Define amdgpu_ras_block_late_fini to remove sysfs nodes and interrupt handles. 3. Replace ras block variable members in struct amdgpu_ras_block_object with struct ras_common_if, which can make it easy to associate each ras block instance with each ras block functional interface. 4. Add .ras_cb to struct amdgpu_ras_block_object. 5. Change each ras block to fit for the changement of struct amdgpu_ras_block_object. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c | 7 ++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 35 +++++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 15 ++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 8 ++++---- drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/mca_v3_0.c | 28 +++++++++++++++---------- drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c | 6 ++++-- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 4 ++-- 11 files changed, 86 insertions(+), 37 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c index 52a60c2316a2..ad057d6b2c77 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c @@ -83,14 +83,15 @@ int amdgpu_mca_ras_late_init(struct amdgpu_device *adev, .sysfs_name = sysfs_name, }; - snprintf(sysfs_name, sizeof(sysfs_name), "%s_err_count", mca_dev->ras->ras_block.name); + snprintf(sysfs_name, sizeof(sysfs_name), "%s_err_count", + mca_dev->ras->ras_block.ras_comm.name); if (!mca_dev->ras_if) { mca_dev->ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); if (!mca_dev->ras_if) return -ENOMEM; - mca_dev->ras_if->block = mca_dev->ras->ras_block.block; - mca_dev->ras_if->sub_block_index = mca_dev->ras->ras_block.sub_block_index; + mca_dev->ras_if->block = mca_dev->ras->ras_block.ras_comm.block; + mca_dev->ras_if->sub_block_index = mca_dev->ras->ras_block.ras_comm.sub_block_index; mca_dev->ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; } ih_info.head = fs_info.head = *mca_dev->ras_if; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index a9c133a09be5..877b0e023648 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -877,7 +877,7 @@ static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_ if (!block_obj) return -EINVAL; - if (block_obj->block == block) + if (block_obj->ras_comm.block == block) return 0; return -EINVAL; @@ -2457,6 +2457,23 @@ interrupt: return r; } +int amdgpu_ras_block_late_init(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + char sysfs_name[32]; + struct ras_ih_if ih_info; + struct ras_fs_if fs_info; + struct amdgpu_ras_block_object *obj; + + obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + ih_info.cb = obj->ras_cb; + ih_info.head = *ras_block; + snprintf(sysfs_name, sizeof(sysfs_name), "%s_err_count", ras_block->name); + fs_info.sysfs_name = (const char *)sysfs_name; + fs_info.head = *ras_block; + return amdgpu_ras_late_init(adev, ras_block, &fs_info, &ih_info); +} + /* helper function to remove ras fs node and interrupt handler */ void amdgpu_ras_late_fini(struct amdgpu_device *adev, struct ras_common_if *ras_block, @@ -2470,6 +2487,22 @@ void amdgpu_ras_late_fini(struct amdgpu_device *adev, amdgpu_ras_interrupt_remove_handler(adev, ih_info); } +void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + struct ras_ih_if ih_info; + struct amdgpu_ras_block_object *obj; + + if (!ras_block) + return; + + obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + ih_info.head = *ras_block; + ih_info.cb = obj->ras_cb; + + amdgpu_ras_late_fini(adev, ras_block, &ih_info); +} + /* do some init work after IP late init as dependence. * and it runs in resume/gpu reset/booting up cases. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index a55743b12d57..8b94b556baf6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -486,17 +486,13 @@ struct ras_debug_if { }; struct amdgpu_ras_block_object { - /* block name */ - char name[32]; - - enum amdgpu_ras_block block; - - uint32_t sub_block_index; + struct ras_common_if ras_comm; int (*ras_block_match)(struct amdgpu_ras_block_object *block_obj, enum amdgpu_ras_block block, uint32_t sub_block_index); int (*ras_late_init)(struct amdgpu_device *adev, void *ras_info); void (*ras_fini)(struct amdgpu_device *adev); + ras_ih_cb ras_cb; const struct amdgpu_ras_block_hw_ops *hw_ops; }; @@ -605,10 +601,17 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block, struct ras_fs_if *fs_info, struct ras_ih_if *ih_info); + +int amdgpu_ras_block_late_init(struct amdgpu_device *adev, + struct ras_common_if *ras_block); + void amdgpu_ras_late_fini(struct amdgpu_device *adev, struct ras_common_if *ras_block, struct ras_ih_if *ih_info); +void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, + struct ras_common_if *ras_block); + int amdgpu_ras_feature_enable(struct amdgpu_device *adev, struct ras_common_if *head, bool enable); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 5929d6f528c9..15707af89212 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -981,8 +981,10 @@ struct amdgpu_ras_block_hw_ops xgmi_ras_hw_ops = { struct amdgpu_xgmi_ras xgmi_ras = { .ras_block = { - .name = "xgmi", - .block = AMDGPU_RAS_BLOCK__XGMI_WAFL, + .ras_comm = { + .name = "xgmi", + .block = AMDGPU_RAS_BLOCK__XGMI_WAFL, + }, .hw_ops = &xgmi_ras_hw_ops, .ras_late_init = amdgpu_xgmi_ras_late_init, .ras_fini = amdgpu_xgmi_ras_fini, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 744253be5142..1405f8145789 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2195,8 +2195,8 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) return err; } - strcpy(adev->gfx.ras->ras_block.name,"gfx"); - adev->gfx.ras->ras_block.block = AMDGPU_RAS_BLOCK__GFX; + strcpy(adev->gfx.ras->ras_block.ras_comm.name, "gfx"); + adev->gfx.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__GFX; /* If not define special ras_late_init function, use gfx default ras_late_init */ if (!adev->gfx.ras->ras_block.ras_late_init) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index c64e3a391c99..80f6a29ed30c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -672,8 +672,8 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev) if (adev->umc.ras) { amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); - strcpy(adev->umc.ras->ras_block.name, "umc"); - adev->umc.ras->ras_block.block = AMDGPU_RAS_BLOCK__UMC; + strcpy(adev->umc.ras->ras_block.ras_comm.name, "umc"); + adev->umc.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__UMC; /* If don't define special ras_late_init function, use default ras_late_init */ if (!adev->umc.ras->ras_block.ras_late_init) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 4595027a8c63..af873c99d5e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1232,8 +1232,8 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) if (adev->umc.ras) { amdgpu_ras_register_ras_block(adev, &adev->umc.ras->ras_block); - strcpy(adev->umc.ras->ras_block.name, "umc"); - adev->umc.ras->ras_block.block = AMDGPU_RAS_BLOCK__UMC; + strcpy(adev->umc.ras->ras_block.ras_comm.name, "umc"); + adev->umc.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__UMC; /* If don't define special ras_late_init function, use default ras_late_init */ if (!adev->umc.ras->ras_block.ras_late_init) @@ -1280,8 +1280,8 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev) if (adev->mmhub.ras) { amdgpu_ras_register_ras_block(adev, &adev->mmhub.ras->ras_block); - strcpy(adev->mmhub.ras->ras_block.name,"mmhub"); - adev->mmhub.ras->ras_block.block = AMDGPU_RAS_BLOCK__MMHUB; + strcpy(adev->mmhub.ras->ras_block.ras_comm.name, "mmhub"); + adev->mmhub.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__MMHUB; /* If don't define special ras_late_init function, use default ras_late_init */ if (!adev->mmhub.ras->ras_block.ras_late_init) diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index 6b41fcbf4875..503c292b321e 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -157,8 +157,10 @@ struct amdgpu_ras_block_hw_ops hdp_v4_0_ras_hw_ops = { struct amdgpu_hdp_ras hdp_v4_0_ras = { .ras_block = { - .name = "hdp", - .block = AMDGPU_RAS_BLOCK__HDP, + .ras_comm = { + .name = "hdp", + .block = AMDGPU_RAS_BLOCK__HDP, + }, .hw_ops = &hdp_v4_0_ras_hw_ops, .ras_late_init = amdgpu_hdp_ras_late_init, .ras_fini = amdgpu_hdp_ras_fini, diff --git a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c index 68565262af9c..386416378a82 100644 --- a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c @@ -53,8 +53,8 @@ static int mca_v3_0_ras_block_match(struct amdgpu_ras_block_object *block_obj, if (!block_obj) return -EINVAL; - if ((block_obj->block == block) && - (block_obj->sub_block_index == sub_block_index)) { + if ((block_obj->ras_comm.block == block) && + (block_obj->ras_comm.sub_block_index == sub_block_index)) { return 0; } @@ -68,9 +68,11 @@ const struct amdgpu_ras_block_hw_ops mca_v3_0_mp0_hw_ops = { struct amdgpu_mca_ras_block mca_v3_0_mp0_ras = { .ras_block = { - .block = AMDGPU_RAS_BLOCK__MCA, - .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP0, - .name = "mp0", + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP0, + .name = "mp0", + }, .hw_ops = &mca_v3_0_mp0_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, .ras_late_init = mca_v3_0_mp0_ras_late_init, @@ -103,9 +105,11 @@ const struct amdgpu_ras_block_hw_ops mca_v3_0_mp1_hw_ops = { struct amdgpu_mca_ras_block mca_v3_0_mp1_ras = { .ras_block = { - .block = AMDGPU_RAS_BLOCK__MCA, - .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP1, - .name = "mp1", + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MP1, + .name = "mp1", + }, .hw_ops = &mca_v3_0_mp1_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, .ras_late_init = mca_v3_0_mp1_ras_late_init, @@ -138,9 +142,11 @@ const struct amdgpu_ras_block_hw_ops mca_v3_0_mpio_hw_ops = { struct amdgpu_mca_ras_block mca_v3_0_mpio_ras = { .ras_block = { - .block = AMDGPU_RAS_BLOCK__MCA, - .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MPIO, - .name = "mpio", + .ras_comm = { + .block = AMDGPU_RAS_BLOCK__MCA, + .sub_block_index = AMDGPU_RAS_MCA_BLOCK__MPIO, + .name = "mpio", + }, .hw_ops = &mca_v3_0_mpio_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, .ras_late_init = mca_v3_0_mpio_ras_late_init, diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index 39974b449341..c7cca87f1647 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -664,8 +664,10 @@ const struct amdgpu_ras_block_hw_ops nbio_v7_4_ras_hw_ops = { struct amdgpu_nbio_ras nbio_v7_4_ras = { .ras_block = { - .name = "pcie_bif", - .block = AMDGPU_RAS_BLOCK__PCIE_BIF, + .ras_comm = { + .name = "pcie_bif", + .block = AMDGPU_RAS_BLOCK__PCIE_BIF, + }, .hw_ops = &nbio_v7_4_ras_hw_ops, .ras_late_init = amdgpu_nbio_ras_late_init, .ras_fini = amdgpu_nbio_ras_fini, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 02115d63b071..f07cb6a256c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -2822,8 +2822,8 @@ static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev) if (adev->sdma.ras) { amdgpu_ras_register_ras_block(adev, &adev->sdma.ras->ras_block); - strcpy(adev->sdma.ras->ras_block.name, "sdma"); - adev->sdma.ras->ras_block.block = AMDGPU_RAS_BLOCK__SDMA; + strcpy(adev->sdma.ras->ras_block.ras_comm.name, "sdma"); + adev->sdma.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__SDMA; /* If don't define special ras_late_init function, use default ras_late_init */ if (!adev->sdma.ras->ras_block.ras_late_init) -- cgit From 80ed77f971c3911d6a6f7c537540b4ee62859455 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 8 Feb 2022 10:48:56 +0800 Subject: drm/amdgpu: Optimize amdgpu_nbio_ras_late_init/amdgpu_nbio_ras_fini function code Optimize amdgpu_nbio_ras_late_init/amdgpu_nbio_ras_fini function code. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c | 40 ++++---------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 1 + drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c | 1 + 3 files changed, 7 insertions(+), 35 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c index 6ace2e390e77..89e61fdd3580 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c @@ -25,26 +25,9 @@ int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, void *ras_info) { int r; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - struct ras_fs_if fs_info = { - .sysfs_name = "pcie_bif_err_count", - }; - - if (!adev->nbio.ras_if) { - adev->nbio.ras_if = kmalloc(sizeof(struct ras_common_if), GFP_KERNEL); - if (!adev->nbio.ras_if) - return -ENOMEM; - adev->nbio.ras_if->block = AMDGPU_RAS_BLOCK__PCIE_BIF; - adev->nbio.ras_if->type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; - adev->nbio.ras_if->sub_block_index = 0; - } - ih_info.head = fs_info.head = *adev->nbio.ras_if; - r = amdgpu_ras_late_init(adev, adev->nbio.ras_if, - &fs_info, &ih_info); + r = amdgpu_ras_block_late_init(adev, adev->nbio.ras_if); if (r) - goto free; + return r; if (amdgpu_ras_is_supported(adev, adev->nbio.ras_if->block)) { r = amdgpu_irq_get(adev, &adev->nbio.ras_controller_irq, 0); @@ -53,30 +36,17 @@ int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, void *ras_info) r = amdgpu_irq_get(adev, &adev->nbio.ras_err_event_athub_irq, 0); if (r) goto late_fini; - } else { - r = 0; - goto free; } return 0; late_fini: - amdgpu_ras_late_fini(adev, adev->nbio.ras_if, &ih_info); -free: - kfree(adev->nbio.ras_if); - adev->nbio.ras_if = NULL; + amdgpu_ras_block_late_fini(adev, adev->nbio.ras_if); return r; } void amdgpu_nbio_ras_fini(struct amdgpu_device *adev) { if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF) && - adev->nbio.ras_if) { - struct ras_common_if *ras_if = adev->nbio.ras_if; - struct ras_ih_if ih_info = { - .cb = NULL, - }; - - amdgpu_ras_late_fini(adev, ras_if, &ih_info); - kfree(ras_if); - } + adev->nbio.ras_if) + amdgpu_ras_block_late_fini(adev, adev->nbio.ras_if); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 877b0e023648..75f5fda357ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2301,6 +2301,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev) if (!adev->gmc.xgmi.connected_to_cpu) { adev->nbio.ras = &nbio_v7_4_ras; amdgpu_ras_register_ras_block(adev, &adev->nbio.ras->ras_block); + adev->nbio.ras_if = &adev->nbio.ras->ras_block.ras_comm; } break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index c7cca87f1647..14768570c298 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -667,6 +667,7 @@ struct amdgpu_nbio_ras nbio_v7_4_ras = { .ras_comm = { .name = "pcie_bif", .block = AMDGPU_RAS_BLOCK__PCIE_BIF, + .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, }, .hw_ops = &nbio_v7_4_ras_hw_ops, .ras_late_init = amdgpu_nbio_ras_late_init, -- cgit From 9252d33df597a60416f3718b9b41457657c8540c Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 9 Feb 2022 11:05:27 +0800 Subject: drm/amdgpu: Optimize operating sysfs and interrupt function interface in amdgpu_ras.c In order to reduce redundant struct conversion, modify operating sysfs and interrupt function interface parameters. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 37 ++++++++++++++++----------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 6 +++--- 2 files changed, 21 insertions(+), 22 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 75f5fda357ac..9fb944cbc755 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1276,18 +1276,17 @@ static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev) } int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, - struct ras_fs_if *head) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); if (!obj || obj->attr_inuse) return -EINVAL; get_obj(obj); - memcpy(obj->fs_data.sysfs_name, - head->sysfs_name, - sizeof(obj->fs_data.sysfs_name)); + snprintf(obj->fs_data.sysfs_name, sizeof(obj->fs_data.sysfs_name), + "%s_err_count", head->name); obj->sysfs_attr = (struct device_attribute){ .attr = { @@ -1594,9 +1593,9 @@ int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, } int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, - struct ras_ih_if *info) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); struct ras_ih_data *data; if (!obj) @@ -1616,24 +1615,27 @@ int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, } int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, - struct ras_ih_if *info) + struct ras_common_if *head) { - struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); + struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); struct ras_ih_data *data; + struct amdgpu_ras_block_object *ras_obj; if (!obj) { /* in case we registe the IH before enable ras feature */ - obj = amdgpu_ras_create_obj(adev, &info->head); + obj = amdgpu_ras_create_obj(adev, head); if (!obj) return -EINVAL; } else get_obj(obj); + ras_obj = container_of(head, struct amdgpu_ras_block_object, ras_comm); + data = &obj->ih_data; /* add the callback.etc */ *data = (struct ras_ih_data) { .inuse = 0, - .cb = info->cb, + .cb = ras_obj->ras_cb, .element_size = sizeof(struct amdgpu_iv_entry), .rptr = 0, .wptr = 0, @@ -1662,10 +1664,7 @@ static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev) struct ras_manager *obj, *tmp; list_for_each_entry_safe(obj, tmp, &con->head, node) { - struct ras_ih_if info = { - .head = obj->head, - }; - amdgpu_ras_interrupt_remove_handler(adev, &info); + amdgpu_ras_interrupt_remove_handler(adev, &obj->head); } return 0; @@ -2431,12 +2430,12 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, return 0; if (ih_info->cb) { - r = amdgpu_ras_interrupt_add_handler(adev, ih_info); + r = amdgpu_ras_interrupt_add_handler(adev, ras_block); if (r) goto interrupt; } - r = amdgpu_ras_sysfs_create(adev, fs_info); + r = amdgpu_ras_sysfs_create(adev, ras_block); if (r) goto sysfs; @@ -2452,7 +2451,7 @@ cleanup: amdgpu_ras_sysfs_remove(adev, ras_block); sysfs: if (ih_info->cb) - amdgpu_ras_interrupt_remove_handler(adev, ih_info); + amdgpu_ras_interrupt_remove_handler(adev, ras_block); interrupt: amdgpu_ras_feature_enable(adev, ras_block, 0); return r; @@ -2485,7 +2484,7 @@ void amdgpu_ras_late_fini(struct amdgpu_device *adev, amdgpu_ras_sysfs_remove(adev, ras_block); if (ih_info->cb) - amdgpu_ras_interrupt_remove_handler(adev, ih_info); + amdgpu_ras_interrupt_remove_handler(adev, &ih_info->head); } void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 8b94b556baf6..ae8741ac526f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -619,7 +619,7 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, struct ras_common_if *head, bool enable); int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, - struct ras_fs_if *head); + struct ras_common_if *head); int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev, struct ras_common_if *head); @@ -636,10 +636,10 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, struct ras_inject_if *info); int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, - struct ras_ih_if *info); + struct ras_common_if *head); int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, - struct ras_ih_if *info); + struct ras_common_if *head); int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, struct ras_dispatch_if *info); -- cgit From 563285c85ecaa1fcecf304dabf87cbeee1ddbc3f Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 9 Feb 2022 11:15:29 +0800 Subject: drm/amdgpu: Merge amdgpu_ras_late_init/amdgpu_ras_late_fini to amdgpu_ras_block_late_init/amdgpu_ras_block_late_fini 1. Merge amdgpu_ras_late_init to amdgpu_ras_block_late_init. 2. Remove amdgpu_ras_late_init since no ras block calls amdgpu_ras_late_init. 3. Merge amdgpu_ras_late_fini to amdgpu_ras_block_late_fini. 4. Remove amdgpu_ras_late_fini since no ras block calls amdgpu_ras_late_fini. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 53 +++++++-------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 8 ----- 2 files changed, 11 insertions(+), 50 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 9fb944cbc755..b5cd21cb6e58 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2397,11 +2397,10 @@ bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev) } /* helper function to handle common stuff in ip late init phase */ -int amdgpu_ras_late_init(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_fs_if *fs_info, - struct ras_ih_if *ih_info) +int amdgpu_ras_block_late_init(struct amdgpu_device *adev, + struct ras_common_if *ras_block) { + struct amdgpu_ras_block_object *ras_obj; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); unsigned long ue_count, ce_count; int r; @@ -2429,7 +2428,8 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, if (adev->in_suspend || amdgpu_in_reset(adev)) return 0; - if (ih_info->cb) { + ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + if (ras_obj->ras_cb) { r = amdgpu_ras_interrupt_add_handler(adev, ras_block); if (r) goto interrupt; @@ -2450,57 +2450,26 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, cleanup: amdgpu_ras_sysfs_remove(adev, ras_block); sysfs: - if (ih_info->cb) + if (ras_obj->ras_cb) amdgpu_ras_interrupt_remove_handler(adev, ras_block); interrupt: amdgpu_ras_feature_enable(adev, ras_block, 0); return r; } -int amdgpu_ras_block_late_init(struct amdgpu_device *adev, - struct ras_common_if *ras_block) -{ - char sysfs_name[32]; - struct ras_ih_if ih_info; - struct ras_fs_if fs_info; - struct amdgpu_ras_block_object *obj; - - obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); - ih_info.cb = obj->ras_cb; - ih_info.head = *ras_block; - snprintf(sysfs_name, sizeof(sysfs_name), "%s_err_count", ras_block->name); - fs_info.sysfs_name = (const char *)sysfs_name; - fs_info.head = *ras_block; - return amdgpu_ras_late_init(adev, ras_block, &fs_info, &ih_info); -} - /* helper function to remove ras fs node and interrupt handler */ -void amdgpu_ras_late_fini(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_ih_if *ih_info) -{ - if (!ras_block || !ih_info) - return; - - amdgpu_ras_sysfs_remove(adev, ras_block); - if (ih_info->cb) - amdgpu_ras_interrupt_remove_handler(adev, &ih_info->head); -} - void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, struct ras_common_if *ras_block) { - struct ras_ih_if ih_info; - struct amdgpu_ras_block_object *obj; - + struct amdgpu_ras_block_object *ras_obj; if (!ras_block) return; - obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); - ih_info.head = *ras_block; - ih_info.cb = obj->ras_cb; + amdgpu_ras_sysfs_remove(adev, ras_block); - amdgpu_ras_late_fini(adev, ras_block, &ih_info); + ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); + if (ras_obj->ras_cb) + amdgpu_ras_interrupt_remove_handler(adev, ras_block); } /* do some init work after IP late init as dependence. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index ae8741ac526f..5de567c6a8f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -597,18 +597,10 @@ amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) { int amdgpu_ras_init(struct amdgpu_device *adev); int amdgpu_ras_fini(struct amdgpu_device *adev); int amdgpu_ras_pre_fini(struct amdgpu_device *adev); -int amdgpu_ras_late_init(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_fs_if *fs_info, - struct ras_ih_if *ih_info); int amdgpu_ras_block_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block); -void amdgpu_ras_late_fini(struct amdgpu_device *adev, - struct ras_common_if *ras_block, - struct ras_ih_if *ih_info); - void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, struct ras_common_if *ras_block); -- cgit From 867e24ca4945249baf34ea07ae6b27ca927210a1 Mon Sep 17 00:00:00 2001 From: yipechai Date: Mon, 14 Feb 2022 14:53:37 +0800 Subject: drm/amdgpu: define amdgpu_ras_late_init to call all ras blocks' .ras_late_init Define amdgpu_ras_late_init to call all ras blocks' .ras_late_init. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 44 ------------------------------ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 25 +++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 6 ---- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 5 +--- drivers/gpu/drm/amd/amdgpu/soc15.c | 6 +--- 7 files changed, 34 insertions(+), 59 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6d4d59458cff..5b772f2af6eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2629,6 +2629,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) adev->ip_blocks[i].status.late_initialized = true; } + r = amdgpu_ras_late_init(adev); + if (r) { + DRM_ERROR("amdgpu_ras_late_init failed %d", r); + return r; + } + amdgpu_ras_set_error_query_ready(adev, true); amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 7bc68c94e0b8..a71688d1c1b1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -449,50 +449,6 @@ int amdgpu_gmc_ras_early_init(struct amdgpu_device *adev) int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) { - int r; - - if (adev->umc.ras && adev->umc.ras->ras_block.ras_late_init) { - r = adev->umc.ras->ras_block.ras_late_init(adev, adev->umc.ras_if); - if (r) - return r; - } - - if (adev->mmhub.ras && adev->mmhub.ras->ras_block.ras_late_init) { - r = adev->mmhub.ras->ras_block.ras_late_init(adev, adev->mmhub.ras_if); - if (r) - return r; - } - - if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_late_init) { - r = adev->gmc.xgmi.ras->ras_block.ras_late_init(adev, adev->gmc.xgmi.ras_if); - if (r) - return r; - } - - if (adev->hdp.ras && adev->hdp.ras->ras_block.ras_late_init) { - r = adev->hdp.ras->ras_block.ras_late_init(adev, adev->hdp.ras_if); - if (r) - return r; - } - - if (adev->mca.mp0.ras && adev->mca.mp0.ras->ras_block.ras_late_init) { - r = adev->mca.mp0.ras->ras_block.ras_late_init(adev, adev->mca.mp0.ras_if); - if (r) - return r; - } - - if (adev->mca.mp1.ras && adev->mca.mp1.ras->ras_block.ras_late_init) { - r = adev->mca.mp1.ras->ras_block.ras_late_init(adev, adev->mca.mp1.ras_if); - if (r) - return r; - } - - if (adev->mca.mpio.ras && adev->mca.mpio.ras->ras_block.ras_late_init) { - r = adev->mca.mpio.ras->ras_block.ras_late_init(adev, adev->mca.mpio.ras_if); - if (r) - return r; - } - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b5cd21cb6e58..73e8d6f70541 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2522,6 +2522,31 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev) amdgpu_ras_disable_all_features(adev, 1); } +int amdgpu_ras_late_init(struct amdgpu_device *adev) +{ + struct amdgpu_ras_block_list *node, *tmp; + struct amdgpu_ras_block_object *obj; + int r; + + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { + if (!node->ras_obj) { + dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); + continue; + } + obj = node->ras_obj; + if (obj->ras_late_init) { + r = obj->ras_late_init(adev, &obj->ras_comm); + if (r) { + dev_err(adev->dev, "%s failed to execute ras_late_init! ret:%d\n", + obj->ras_comm.name, r); + return r; + } + } + } + + return 0; +} + /* do some fini work before IP fini as dependence */ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 837d1b79a9cb..143a83043d7c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -595,6 +595,7 @@ amdgpu_ras_error_to_ta(enum amdgpu_ras_error_type error) { /* called in ip_init and ip_fini */ int amdgpu_ras_init(struct amdgpu_device *adev); +int amdgpu_ras_late_init(struct amdgpu_device *adev); int amdgpu_ras_fini(struct amdgpu_device *adev); int amdgpu_ras_pre_fini(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index bb40ab83fc22..1997f129db9c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -4791,12 +4791,6 @@ static int gfx_v9_0_ecc_late_init(void *handle) if (r) return r; - if (adev->gfx.ras && adev->gfx.ras->ras_block.ras_late_init) { - r = adev->gfx.ras->ras_block.ras_late_init(adev, adev->gfx.ras_if); - if (r) - return r; - } - if (adev->gfx.ras && adev->gfx.ras->enable_watchdog_timer) adev->gfx.ras->enable_watchdog_timer(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index af5a1c93861b..e26c39fcd336 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1894,10 +1894,7 @@ static int sdma_v4_0_late_init(void *handle) adev->sdma.ras->ras_block.hw_ops->reset_ras_error_count(adev); } - if (adev->sdma.ras && adev->sdma.ras->ras_block.ras_late_init) - return adev->sdma.ras->ras_block.ras_late_init(adev, adev->sdma.ras_if); - else - return 0; + return 0; } static int sdma_v4_0_sw_init(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 99a88f4d8050..25e1ee2bd2f7 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -1189,15 +1189,11 @@ static int soc15_common_early_init(void *handle) static int soc15_common_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int r = 0; if (amdgpu_sriov_vf(adev)) xgpu_ai_mailbox_get_irq(adev); - if (adev->nbio.ras && adev->nbio.ras->ras_block.ras_late_init) - r = adev->nbio.ras->ras_block.ras_late_init(adev, adev->nbio.ras_if); - - return r; + return 0; } static int soc15_common_sw_init(void *handle) -- cgit From 418abce203fc2e936bf8c7632a9a429e861f6283 Mon Sep 17 00:00:00 2001 From: yipechai Date: Tue, 15 Feb 2022 14:38:13 +0800 Subject: drm/amdgpu: Remove redundant .ras_late_init initialization in some ras blocks 1. Define amdgpu_ras_block_late_init_default in amdgpu_ras.c as .ras_late_init common function, which is called when .ras_late_init of ras block isn't initialized. 2. Remove the code of using amdgpu_ras_block_late_init to initialize .ras_late_init in ras blocks. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 10 +++++++++- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 4 ---- drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c | 1 - drivers/gpu/drm/amd/amdgpu/mca_v3_0.c | 3 --- 4 files changed, 9 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 73e8d6f70541..d60123e9d045 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2457,6 +2457,12 @@ interrupt: return r; } +int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + return amdgpu_ras_block_late_init(adev, ras_block); +} + /* helper function to remove ras fs node and interrupt handler */ void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, struct ras_common_if *ras_block) @@ -2533,6 +2539,7 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev) dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); continue; } + obj = node->ras_obj; if (obj->ras_late_init) { r = obj->ras_late_init(adev, &obj->ras_comm); @@ -2541,7 +2548,8 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev) obj->ras_comm.name, r); return r; } - } + } else + amdgpu_ras_block_late_init_default(adev, &obj->ras_comm); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index b719d2c3003b..412e44af1608 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1291,10 +1291,6 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev) adev->mmhub.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; adev->mmhub.ras_if = &adev->mmhub.ras->ras_block.ras_comm; - /* If don't define special ras_late_init function, use default ras_late_init */ - if (!adev->mmhub.ras->ras_block.ras_late_init) - adev->mmhub.ras->ras_block.ras_late_init = amdgpu_ras_block_late_init; - /* If don't define special ras_fini function, use default ras_fini */ if (!adev->mmhub.ras->ras_block.ras_fini) adev->mmhub.ras->ras_block.ras_fini = amdgpu_mmhub_ras_fini; diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index c9e931f046f7..d7811e0327cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -163,7 +163,6 @@ struct amdgpu_hdp_ras hdp_v4_0_ras = { .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, }, .hw_ops = &hdp_v4_0_ras_hw_ops, - .ras_late_init = amdgpu_ras_block_late_init, .ras_fini = amdgpu_hdp_ras_fini, }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c index 12d09a58b644..b4b36899f5c6 100644 --- a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c @@ -71,7 +71,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mp0_ras = { }, .hw_ops = &mca_v3_0_mp0_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_late_init = amdgpu_ras_block_late_init, .ras_fini = mca_v3_0_mp0_ras_fini, }, }; @@ -104,7 +103,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mp1_ras = { }, .hw_ops = &mca_v3_0_mp1_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_late_init = amdgpu_ras_block_late_init, .ras_fini = mca_v3_0_mp1_ras_fini, }, }; @@ -137,7 +135,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mpio_ras = { }, .hw_ops = &mca_v3_0_mpio_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_late_init = amdgpu_ras_block_late_init, .ras_fini = mca_v3_0_mpio_ras_fini, }, }; -- cgit From 779596ce6a79e187995f04f143fc5ea44a565ea9 Mon Sep 17 00:00:00 2001 From: Tom Rix Date: Thu, 17 Feb 2022 07:38:42 -0800 Subject: drm/amdgpu: fix amdgpu_ras_block_late_init error handler Clang build fails with amdgpu_ras.c:2416:7: error: variable 'ras_obj' is used uninitialized whenever 'if' condition is true if (adev->in_suspend || amdgpu_in_reset(adev)) { ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ amdgpu_ras.c:2453:6: note: uninitialized use occurs here if (ras_obj->ras_cb) ^~~~~~~ There is a logic error in the error handler's labels. ex/ The sysfs: is the last goto label in the normal code but is the middle of error handler. Rework the error handler. cleanup: is the first error, so it's handler should be last. interrupt: is the second error, it's handler is next. interrupt: handles the failure of amdgpu_ras_interrupt_add_hander() by calling amdgpu_ras_interrupt_remove_handler(). This is wrong, remove() assumes the interrupt has been setup, not torn down by add(). Change the goto label to cleanup. sysfs is the last error, it's handler should be first. sysfs: handles the failure of amdgpu_ras_sysfs_create() by calling amdgpu_ras_sysfs_remove(). But when the create() fails there is nothing added so there is nothing to remove. This error handler is not needed. Remove the error handler and change goto label to interrupt. Fixes: b293e891b057 ("drm/amdgpu: add helper function to do common ras_late_init/fini (v3)") Reviewed-by: Luben Tuikov Signed-off-by: Tom Rix Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index d60123e9d045..61f589b6fe5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2432,12 +2432,12 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev, if (ras_obj->ras_cb) { r = amdgpu_ras_interrupt_add_handler(adev, ras_block); if (r) - goto interrupt; + goto cleanup; } r = amdgpu_ras_sysfs_create(adev, ras_block); if (r) - goto sysfs; + goto interrupt; /* Those are the cached values at init. */ @@ -2447,12 +2447,11 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev, } return 0; -cleanup: - amdgpu_ras_sysfs_remove(adev, ras_block); -sysfs: + +interrupt: if (ras_obj->ras_cb) amdgpu_ras_interrupt_remove_handler(adev, ras_block); -interrupt: +cleanup: amdgpu_ras_feature_enable(adev, ras_block, 0); return r; } -- cgit From d41ff22a4eefe2b933a82c0d3ce8ff47fa86e212 Mon Sep 17 00:00:00 2001 From: Maíra Canal Date: Tue, 22 Feb 2022 10:16:52 -0300 Subject: drm/amdgpu: Change amdgpu_ras_block_late_init_default function scope MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Turn previously global function into a static function to avoid the following Clang warning: drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2459:5: warning: no previous prototype for function 'amdgpu_ras_block_late_init_default' [-Wmissing-prototypes] int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, ^ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c:2459:1: note: declare 'static' if the function is not intended to be used outside of this translation unit int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, ^ static Signed-off-by: Maíra Canal Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 61f589b6fe5e..0e89175ee0a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2456,7 +2456,7 @@ cleanup: return r; } -int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, +static int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev, struct ras_common_if *ras_block) { return amdgpu_ras_block_late_init(adev, ras_block); -- cgit From 29c9b6cd581de79ef848a50d1c9f06cf567bfd78 Mon Sep 17 00:00:00 2001 From: yipechai Date: Wed, 23 Feb 2022 11:05:43 +0800 Subject: drm/amdgpu: Fixed warning reported by kernel test robot Fixed warning reported by kernel test robot: 1.warning: variable 'ras_obj' is used uninitialized whenever '||' condition is true. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 0e89175ee0a9..b53abc15a830 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2400,7 +2400,7 @@ bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev) int amdgpu_ras_block_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { - struct amdgpu_ras_block_object *ras_obj; + struct amdgpu_ras_block_object *ras_obj = NULL; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); unsigned long ue_count, ce_count; int r; -- cgit From 1f211a827cbda87bd0ec145a1f44f2615c3f56d3 Mon Sep 17 00:00:00 2001 From: yipechai Date: Mon, 14 Feb 2022 15:46:32 +0800 Subject: drm/amdgpu: centrally calls the .ras_fini function of all ras blocks centrally calls the .ras_fini function of all ras blocks. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 10 ---------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 20 ++++++++++++++------ drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 3 --- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 4 ---- drivers/gpu/drm/amd/amdgpu/soc15.c | 3 --- 5 files changed, 14 insertions(+), 26 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 62083d3aa9b8..65f1d3f35c60 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -454,17 +454,7 @@ int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev) void amdgpu_gmc_ras_fini(struct amdgpu_device *adev) { - if (adev->umc.ras && adev->umc.ras->ras_block.ras_fini) - adev->umc.ras->ras_block.ras_fini(adev, adev->umc.ras_if); - if (adev->mmhub.ras && adev->mmhub.ras->ras_block.ras_fini) - adev->mmhub.ras->ras_block.ras_fini(adev, adev->mmhub.ras_if); - - if (adev->gmc.xgmi.ras && adev->gmc.xgmi.ras->ras_block.ras_fini) - adev->gmc.xgmi.ras->ras_block.ras_fini(adev, adev->gmc.xgmi.ras_if); - - if (adev->hdp.ras && adev->hdp.ras->ras_block.ras_fini) - adev->hdp.ras->ras_block.ras_fini(adev, adev->hdp.ras_if); } /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b53abc15a830..bbbf22860c41 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2572,11 +2572,25 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) int amdgpu_ras_fini(struct amdgpu_device *adev) { struct amdgpu_ras_block_list *ras_node, *tmp; + struct amdgpu_ras_block_object *obj = NULL; struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) return 0; + list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) { + if (ras_node->ras_obj) { + obj = ras_node->ras_obj; + if (amdgpu_ras_is_supported(adev, obj->ras_comm.block) && + obj->ras_fini) + obj->ras_fini(adev, &obj->ras_comm); + } + + /* Clear ras blocks from ras_list and free ras block list node */ + list_del(&ras_node->node); + kfree(ras_node); + } + amdgpu_ras_fs_fini(adev); amdgpu_ras_interrupt_remove_all(adev); @@ -2590,12 +2604,6 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) amdgpu_ras_set_context(adev, NULL); kfree(con); - /* Clear ras blocks from ras_list and free ras block list node */ - list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) { - list_del(&ras_node->node); - kfree(ras_node); - } - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index e8446967a4d4..dc6e6fe6c978 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2432,9 +2432,6 @@ static int gfx_v9_0_sw_fini(void *handle) int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->gfx.ras && adev->gfx.ras->ras_block.ras_fini) - adev->gfx.ras->ras_block.ras_fini(adev, adev->gfx.ras_if); - for (i = 0; i < adev->gfx.num_gfx_rings; i++) amdgpu_ring_fini(&adev->gfx.gfx_ring[i]); for (i = 0; i < adev->gfx.num_compute_rings; i++) diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 222d25a0413a..d1d40bbb2892 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1995,10 +1995,6 @@ static int sdma_v4_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; - if (adev->sdma.ras && adev->sdma.ras->ras_block.hw_ops && - adev->sdma.ras->ras_block.ras_fini) - adev->sdma.ras->ras_block.ras_fini(adev, adev->sdma.ras_if); - for (i = 0; i < adev->sdma.num_instances; i++) { amdgpu_ring_fini(&adev->sdma.instance[i].ring); if (adev->sdma.has_page_queue) diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 0631ebd39db1..496c4a6e23ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -1214,9 +1214,6 @@ static int soc15_common_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->nbio.ras && adev->nbio.ras->ras_block.ras_fini) - adev->nbio.ras->ras_block.ras_fini(adev, adev->nbio.ras_if); - if (adev->df.funcs && adev->df.funcs->sw_fini) adev->df.funcs->sw_fini(adev); -- cgit From 80e0c2cb37b6e00ec0b41c7670e24f72b2d54ceb Mon Sep 17 00:00:00 2001 From: yipechai Date: Thu, 17 Feb 2022 15:40:12 +0800 Subject: drm/amdgpu: Remove redundant .ras_fini initialization in some ras blocks 1. Define amdgpu_ras_block_late_fini_default in amdgpu_ras.c as .ras_fini common function, which is called when .ras_fini of ras block isn't initialized. 2. Remove the code of using amdgpu_ras_block_late_fini to initialize .ras_fini in ras blocks. Signed-off-by: yipechai Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 8 ++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c | 1 - drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 ---- drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 4 ---- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 8 -------- drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c | 1 - drivers/gpu/drm/amd/amdgpu/mca_v3_0.c | 3 --- drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c | 1 - drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 4 ---- 9 files changed, 8 insertions(+), 26 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index bbbf22860c41..d78c2970e558 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2477,6 +2477,12 @@ void amdgpu_ras_block_late_fini(struct amdgpu_device *adev, amdgpu_ras_interrupt_remove_handler(adev, ras_block); } +static void amdgpu_ras_block_late_fini_default(struct amdgpu_device *adev, + struct ras_common_if *ras_block) +{ + return amdgpu_ras_block_late_fini(adev, ras_block); +} + /* do some init work after IP late init as dependence. * and it runs in resume/gpu reset/booting up cases. */ @@ -2584,6 +2590,8 @@ int amdgpu_ras_fini(struct amdgpu_device *adev) if (amdgpu_ras_is_supported(adev, obj->ras_comm.block) && obj->ras_fini) obj->ras_fini(adev, &obj->ras_comm); + else + amdgpu_ras_block_late_fini_default(adev, &obj->ras_comm); } /* Clear ras blocks from ras_list and free ras block list node */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c index 05dc76f6fdc6..05c0d6e2c75c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -975,6 +975,5 @@ struct amdgpu_xgmi_ras xgmi_ras = { }, .hw_ops = &xgmi_ras_hw_ops, .ras_late_init = amdgpu_xgmi_ras_late_init, - .ras_fini = amdgpu_ras_block_late_fini, }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index f0cc073e6bb0..8def7f630d4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2204,10 +2204,6 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) if (!adev->gfx.ras->ras_block.ras_late_init) adev->gfx.ras->ras_block.ras_late_init = amdgpu_gfx_ras_late_init; - /* If not define special ras_fini function, use gfx default ras_fini */ - if (!adev->gfx.ras->ras_block.ras_fini) - adev->gfx.ras->ras_block.ras_fini = amdgpu_ras_block_late_fini; - /* If not defined special ras_cb function, use default ras_cb */ if (!adev->gfx.ras->ras_block.ras_cb) adev->gfx.ras->ras_block.ras_cb = amdgpu_gfx_process_ras_data_cb; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c index d4ffe08f9fb5..f60b7bd4dbf5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c @@ -683,10 +683,6 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev) if (!adev->umc.ras->ras_block.ras_late_init) adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; - /* If don't define special ras_fini function, use default ras_fini */ - if (!adev->umc.ras->ras_block.ras_fini) - adev->umc.ras->ras_block.ras_fini = amdgpu_ras_block_late_fini; - /* If not defined special ras_cb function, use default ras_cb */ if (!adev->umc.ras->ras_block.ras_cb) adev->umc.ras->ras_block.ras_cb = amdgpu_umc_process_ras_data_cb; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 1b7a59e427e5..e3b78b4cbd64 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -1243,10 +1243,6 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev) if (!adev->umc.ras->ras_block.ras_late_init) adev->umc.ras->ras_block.ras_late_init = amdgpu_umc_ras_late_init; - /* If don't define special ras_fini function, use default ras_fini */ - if (!adev->umc.ras->ras_block.ras_fini) - adev->umc.ras->ras_block.ras_fini = amdgpu_ras_block_late_fini; - /* If not defined special ras_cb function, use default ras_cb */ if (!adev->umc.ras->ras_block.ras_cb) adev->umc.ras->ras_block.ras_cb = amdgpu_umc_process_ras_data_cb; @@ -1292,10 +1288,6 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev) adev->mmhub.ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__MMHUB; adev->mmhub.ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; adev->mmhub.ras_if = &adev->mmhub.ras->ras_block.ras_comm; - - /* If don't define special ras_fini function, use default ras_fini */ - if (!adev->mmhub.ras->ras_block.ras_fini) - adev->mmhub.ras->ras_block.ras_fini = amdgpu_ras_block_late_fini; } } diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c index d1e27133b2ca..046216635262 100644 --- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c @@ -166,7 +166,6 @@ struct amdgpu_hdp_ras hdp_v4_0_ras = { .type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, }, .hw_ops = &hdp_v4_0_ras_hw_ops, - .ras_fini = amdgpu_ras_block_late_fini, }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c index 5ce6778a821d..d4bd7d1d2649 100644 --- a/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mca_v3_0.c @@ -66,7 +66,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mp0_ras = { }, .hw_ops = &mca_v3_0_mp0_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_fini = amdgpu_ras_block_late_fini, }, }; @@ -93,7 +92,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mp1_ras = { }, .hw_ops = &mca_v3_0_mp1_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_fini = amdgpu_ras_block_late_fini, }, }; @@ -120,7 +118,6 @@ struct amdgpu_mca_ras_block mca_v3_0_mpio_ras = { }, .hw_ops = &mca_v3_0_mpio_hw_ops, .ras_block_match = mca_v3_0_ras_block_match, - .ras_fini = amdgpu_ras_block_late_fini, }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c index 485fd9239cd5..b31df4db01fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -671,7 +671,6 @@ struct amdgpu_nbio_ras nbio_v7_4_ras = { }, .hw_ops = &nbio_v7_4_ras_hw_ops, .ras_late_init = amdgpu_nbio_ras_late_init, - .ras_fini = amdgpu_ras_block_late_fini, }, .handle_ras_controller_intr_no_bifring = nbio_v7_4_handle_ras_controller_intr_no_bifring, .handle_ras_err_event_athub_intr_no_bifring = nbio_v7_4_handle_ras_err_event_athub_intr_no_bifring, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 206acb6a5b20..01b385568c14 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -2822,10 +2822,6 @@ static void sdma_v4_0_set_ras_funcs(struct amdgpu_device *adev) if (!adev->sdma.ras->ras_block.ras_late_init) adev->sdma.ras->ras_block.ras_late_init = amdgpu_sdma_ras_late_init; - /* If don't define special ras_fini function, use default ras_fini */ - if (!adev->sdma.ras->ras_block.ras_fini) - adev->sdma.ras->ras_block.ras_fini = amdgpu_ras_block_late_fini; - /* If not defined special ras_cb function, use default ras_cb */ if (!adev->sdma.ras->ras_block.ras_cb) adev->sdma.ras->ras_block.ras_cb = amdgpu_sdma_process_ras_data_cb; -- cgit From 69691c823531c36c7283ecaa040e99e9c12ece07 Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Thu, 3 Mar 2022 17:56:33 +0800 Subject: drm/amdgpu: message smu to update bad channel info It should notice SMU to update bad channel info when detected uncorrectable error in UMC block Signed-off-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 7 +++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 3 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c | 25 +++++++++++++++++++++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h | 4 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c | 5 +++++ 5 files changed, 42 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index d78c2970e558..424c22a841f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2068,6 +2068,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) mutex_init(&con->recovery_lock); INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery); atomic_set(&con->in_recovery, 0); + con->eeprom_control.bad_channel_bitmap = 0; max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count(); amdgpu_ras_validate_threshold(adev, max_eeprom_records_count); @@ -2092,6 +2093,11 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) goto free; amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs); + + if (con->update_channel_flag == true) { + amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap); + con->update_channel_flag = false; + } } #ifdef CONFIG_X86_MCE_AMD @@ -2285,6 +2291,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev) goto release_con; } + con->update_channel_flag = false; con->features = 0; INIT_LIST_HEAD(&con->head); /* Might need get this flag from vbios. */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 7cddaad90d6d..9314fde81e68 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -374,6 +374,9 @@ struct amdgpu_ras { /* record umc error info queried from smu */ struct umc_ecc_info umc_ecc; + + /* Indicates smu whether need update bad channel info */ + bool update_channel_flag; }; struct ras_fs_data { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c index a44f2eeed6ef..c4283987bb1e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c @@ -267,6 +267,7 @@ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control) { struct amdgpu_device *adev = to_amdgpu_device(control); struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr; + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); u8 csum; int res; @@ -287,6 +288,10 @@ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control) amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_recs); + control->bad_channel_bitmap = 0; + amdgpu_dpm_send_hbm_bad_channel_flag(adev, control->bad_channel_bitmap); + con->update_channel_flag = false; + amdgpu_ras_debugfs_set_ret_size(control); mutex_unlock(&control->ras_tbl_mutex); @@ -420,6 +425,7 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control, struct eeprom_table_record *record, const u32 num) { + struct amdgpu_ras *con = amdgpu_ras_get_context(to_amdgpu_device(control)); u32 a, b, i; u8 *buf, *pp; int res; @@ -431,9 +437,16 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control, /* Encode all of them in one go. */ pp = buf; - for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) + for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) { __encode_table_record_to_buf(control, &record[i], pp); + /* update bad channel bitmap */ + if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) { + control->bad_channel_bitmap |= 1 << record[i].mem_channel; + con->update_channel_flag = true; + } + } + /* a, first record index to write into. * b, last record index to write into. * a = first index to read (fri) + number of records in the table, @@ -686,6 +699,7 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control, const u32 num) { struct amdgpu_device *adev = to_amdgpu_device(control); + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int i, res; u8 *buf, *pp; u32 g0, g1; @@ -753,8 +767,15 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control, /* Read up everything? Then transform. */ pp = buf; - for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) + for (i = 0; i < num; i++, pp += RAS_TABLE_RECORD_SIZE) { __decode_table_record_from_buf(control, &record[i], pp); + + /* update bad channel bitmap */ + if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) { + control->bad_channel_bitmap |= 1 << record[i].mem_channel; + con->update_channel_flag = true; + } + } Out: kfree(buf); mutex_unlock(&control->ras_tbl_mutex); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h index 6bb00578bfbb..54d9bfe0881d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h @@ -80,6 +80,10 @@ struct amdgpu_ras_eeprom_control { /* Protect table access via this mutex. */ struct mutex ras_tbl_mutex; + + /* Record channel info which occurred bad pages + */ + u32 bad_channel_bitmap; }; /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c index 85da6cbaf3b7..aad3c8b4c810 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c @@ -97,6 +97,11 @@ static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev, amdgpu_ras_save_bad_pages(adev); amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs); + + if (con->update_channel_flag == true) { + amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap); + con->update_channel_flag = false; + } } if (reset) -- cgit From a3d63c62bdf06c5a3f8a71c207f13b26fc6030f5 Mon Sep 17 00:00:00 2001 From: Mohammad Zafar Ziya Date: Thu, 17 Mar 2022 09:45:42 +0800 Subject: drm/amdgpu: Add vcn and jpeg ras support flag Add vcn and jpeg ras support options V2: vcn and jpeg ras flag enabled for aldebaran asic only V3: vcn and jpeg ras flag disabled for error counter query Generic poison query interface added VCN and JPEG ras enabled based on IP version check V4: vcn and jpeg ras flag moved under ecc flag for dGPU Signed-off-by: Mohammad Zafar Ziya Reviewed-by: Hawking Zhang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 9 +++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 2 ++ 2 files changed, 11 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 424c22a841f4..ec709997c9c7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -66,6 +66,8 @@ const char *ras_block_string[] = { "mp1", "fuse", "mca", + "vcn", + "jpeg", }; const char *ras_mca_block_string[] = { @@ -2205,6 +2207,13 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) dev_info(adev->dev, "SRAM ECC is active.\n"); adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | 1 << AMDGPU_RAS_BLOCK__DF); + + if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0)) + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); + else + adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 9314fde81e68..1e1a3b736859 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -49,6 +49,8 @@ enum amdgpu_ras_block { AMDGPU_RAS_BLOCK__MP1, AMDGPU_RAS_BLOCK__FUSE, AMDGPU_RAS_BLOCK__MCA, + AMDGPU_RAS_BLOCK__VCN, + AMDGPU_RAS_BLOCK__JPEG, AMDGPU_RAS_BLOCK__LAST }; -- cgit From 50a7d025cacdb518a413590a5f2028ace77289f5 Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Fri, 8 Apr 2022 19:51:20 +0800 Subject: drm/amdgpu: add RAS poison creation handler (v2) Prepare for the implementation of poison consumption handler. v2: separate umc handler from poison creation. Signed-off-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 70 +++++++++++++++++++++------------ 1 file changed, 44 insertions(+), 26 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index ec709997c9c7..f10afc1ea017 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1515,12 +1515,45 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) /* ras fs end */ /* ih begin */ +static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + dev_info(obj->adev->dev, + "Poison is created, no user action is needed.\n"); +} + +static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + struct ras_ih_data *data = &obj->ih_data; + struct ras_err_data err_data = {0, 0, 0, NULL}; + int ret; + + if (!data->cb) + return; + + /* Let IP handle its data, maybe we need get the output + * from the callback to update the error type/count, etc + */ + ret = data->cb(obj->adev, &err_data, entry); + /* ue will trigger an interrupt, and in that case + * we need do a reset to recovery the whole system. + * But leave IP do that recovery, here we just dispatch + * the error. + */ + if (ret == AMDGPU_RAS_SUCCESS) { + /* these counts could be left as 0 if + * some blocks do not count error number + */ + obj->err_data.ue_count += err_data.ue_count; + obj->err_data.ce_count += err_data.ce_count; + } +} + static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) { struct ras_ih_data *data = &obj->ih_data; struct amdgpu_iv_entry entry; - int ret; - struct ras_err_data err_data = {0, 0, 0, NULL}; while (data->rptr != data->wptr) { rmb(); @@ -1531,30 +1564,15 @@ static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) data->rptr = (data->aligned_element_size + data->rptr) % data->ring_size; - if (data->cb) { - if (amdgpu_ras_is_poison_mode_supported(obj->adev) && - obj->head.block == AMDGPU_RAS_BLOCK__UMC) - dev_info(obj->adev->dev, - "Poison is created, no user action is needed.\n"); - else { - /* Let IP handle its data, maybe we need get the output - * from the callback to udpate the error type/count, etc - */ - memset(&err_data, 0, sizeof(err_data)); - ret = data->cb(obj->adev, &err_data, &entry); - /* ue will trigger an interrupt, and in that case - * we need do a reset to recovery the whole system. - * But leave IP do that recovery, here we just dispatch - * the error. - */ - if (ret == AMDGPU_RAS_SUCCESS) { - /* these counts could be left as 0 if - * some blocks do not count error number - */ - obj->err_data.ue_count += err_data.ue_count; - obj->err_data.ce_count += err_data.ce_count; - } - } + if (amdgpu_ras_is_poison_mode_supported(obj->adev)) { + if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) + amdgpu_ras_interrupt_poison_creation_handler(obj, &entry); + } else { + if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) + amdgpu_ras_interrupt_umc_handler(obj, &entry); + else + dev_warn(obj->adev->dev, + "No RAS interrupt handler for non-UMC block with poison disabled.\n"); } } } -- cgit From 66f879496121c18c541125d04a444c62f3ca82ad Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Tue, 19 Apr 2022 11:04:19 +0800 Subject: drm/amdgpu: add RAS poison consumption handler (v2) Add support for general RAS poison consumption handler. v2: remove callback function for poison consumption. Signed-off-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 34 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + 2 files changed, 35 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index f10afc1ea017..1c86ec9ab139 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1515,6 +1515,38 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) /* ras fs end */ /* ih begin */ +static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *obj, + struct amdgpu_iv_entry *entry) +{ + bool poison_stat = true, need_reset = true; + struct amdgpu_device *adev = obj->adev; + struct ras_err_data err_data = {0, 0, 0, NULL}; + struct amdgpu_ras_block_object *block_obj = + amdgpu_ras_get_ras_block(adev, obj->head.block, 0); + + if (!adev->gmc.xgmi.connected_to_cpu) + amdgpu_umc_poison_handler(adev, &err_data, false); + + /* both query_poison_status and handle_poison_consumption are optional */ + if (block_obj && block_obj->hw_ops) { + if (block_obj->hw_ops->query_poison_status) { + poison_stat = block_obj->hw_ops->query_poison_status(adev); + if (!poison_stat) + dev_info(adev->dev, "No RAS poison status in %s poison IH.\n", + block_obj->ras_comm.name); + } + + if (poison_stat && block_obj->hw_ops->handle_poison_consumption) { + poison_stat = block_obj->hw_ops->handle_poison_consumption(adev); + need_reset = poison_stat; + } + } + + /* gpu reset is fallback for all failed cases */ + if (need_reset) + amdgpu_ras_reset_gpu(adev); +} + static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj, struct amdgpu_iv_entry *entry) { @@ -1567,6 +1599,8 @@ static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) if (amdgpu_ras_is_poison_mode_supported(obj->adev)) { if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) amdgpu_ras_interrupt_poison_creation_handler(obj, &entry); + else + amdgpu_ras_interrupt_poison_consumption_handler(obj, &entry); } else { if (obj->head.block == AMDGPU_RAS_BLOCK__UMC) amdgpu_ras_interrupt_umc_handler(obj, &entry); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 606df8869b89..c4b61785ab5c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -509,6 +509,7 @@ struct amdgpu_ras_block_hw_ops { void (*reset_ras_error_count)(struct amdgpu_device *adev); void (*reset_ras_error_status)(struct amdgpu_device *adev); bool (*query_poison_status)(struct amdgpu_device *adev); + bool (*handle_poison_consumption)(struct amdgpu_device *adev); }; /* work flow -- cgit From b3c76814ce5b043faa2f07108f1c87ed1cbc8cd1 Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Tue, 19 Apr 2022 14:45:09 +0800 Subject: drm/amdgpu: add RAS fatal error interrupt handler The fatal error handler is independent from general ras interrupt handler since there is no related IH ring. Signed-off-by: Tao Zhou Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 15 +-------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 20 ++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 1 + 3 files changed, 22 insertions(+), 14 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index ea3e8c66211f..b4cf8717f554 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -193,20 +193,7 @@ static irqreturn_t amdgpu_irq_handler(int irq, void *arg) if (ret == IRQ_HANDLED) pm_runtime_mark_last_busy(dev->dev); - /* For the hardware that cannot enable bif ring for both ras_controller_irq - * and ras_err_evnet_athub_irq ih cookies, the driver has to poll status - * register to check whether the interrupt is triggered or not, and properly - * ack the interrupt if it is there - */ - if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) { - if (adev->nbio.ras && - adev->nbio.ras->handle_ras_controller_intr_no_bifring) - adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev); - - if (adev->nbio.ras && - adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring) - adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring(adev); - } + amdgpu_ras_interrupt_fatal_error_handler(adev); return ret; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 1c86ec9ab139..03ce3ce913e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1515,6 +1515,26 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) /* ras fs end */ /* ih begin */ + +/* For the hardware that cannot enable bif ring for both ras_controller_irq + * and ras_err_evnet_athub_irq ih cookies, the driver has to poll status + * register to check whether the interrupt is triggered or not, and properly + * ack the interrupt if it is there + */ +void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev) +{ + if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) + return; + + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_controller_intr_no_bifring) + adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev); + + if (adev->nbio.ras && + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring) + adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring(adev); +} + static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *obj, struct amdgpu_iv_entry *entry) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index c4b61785ab5c..b9a6fac2b8b2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -683,4 +683,5 @@ int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_co int amdgpu_ras_register_ras_block(struct amdgpu_device *adev, struct amdgpu_ras_block_object *ras_block_obj); +void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev); #endif -- cgit From b3ef3205bc5e5d5a91f7ef53306686967c76492b Mon Sep 17 00:00:00 2001 From: Haowen Bai Date: Fri, 22 Apr 2022 14:03:57 +0800 Subject: drm/amdgpu: Remove useless kfree After alloc fail, we do not need to kfree. Signed-off-by: Haowen Bai Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 03ce3ce913e4..7e126dff004f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1921,7 +1921,6 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev, void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL); if (!bps) { - kfree(bps); return -ENOMEM; } -- cgit From 367806068283725d91ffcf49149db1cda5c1fd23 Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Mon, 9 May 2022 14:49:16 +0800 Subject: drm/amdgpu: enable RAS IH for poison consumption Enable RAS IH if poison consumption handler is implemented. Signed-off-by: Tao Zhou Reviewed-by: Mohammad Zafar Ziya Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 7e126dff004f..defc6a53c7dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2516,7 +2516,9 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev, return 0; ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm); - if (ras_obj->ras_cb) { + if (ras_obj->ras_cb || (ras_obj->hw_ops && + (ras_obj->hw_ops->query_poison_status || + ras_obj->hw_ops->handle_poison_consumption))) { r = amdgpu_ras_interrupt_add_handler(adev, ras_block); if (r) goto cleanup; -- cgit From b63ac5d3033976301f296d048c54d584dfb3ac30 Mon Sep 17 00:00:00 2001 From: Tao Zhou Date: Mon, 9 May 2022 17:52:15 +0800 Subject: drm/amdgpu: refine RAS poison consumption handler Qeury ras status before ras poison consumption handling, add more comment and log. Signed-off-by: Tao Zhou Reviewed-and-tested-by: Mohammad Zafar Ziya Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 43 ++++++++++++++++++++------------- 1 file changed, 26 insertions(+), 17 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index defc6a53c7dc..035891ec59d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -1538,33 +1538,42 @@ void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev) static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *obj, struct amdgpu_iv_entry *entry) { - bool poison_stat = true, need_reset = true; + bool poison_stat = false; struct amdgpu_device *adev = obj->adev; struct ras_err_data err_data = {0, 0, 0, NULL}; struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, obj->head.block, 0); - if (!adev->gmc.xgmi.connected_to_cpu) - amdgpu_umc_poison_handler(adev, &err_data, false); - - /* both query_poison_status and handle_poison_consumption are optional */ - if (block_obj && block_obj->hw_ops) { - if (block_obj->hw_ops->query_poison_status) { - poison_stat = block_obj->hw_ops->query_poison_status(adev); - if (!poison_stat) - dev_info(adev->dev, "No RAS poison status in %s poison IH.\n", - block_obj->ras_comm.name); - } + if (!block_obj || !block_obj->hw_ops) + return; - if (poison_stat && block_obj->hw_ops->handle_poison_consumption) { - poison_stat = block_obj->hw_ops->handle_poison_consumption(adev); - need_reset = poison_stat; + /* both query_poison_status and handle_poison_consumption are optional, + * but at least one of them should be implemented if we need poison + * consumption handler + */ + if (block_obj->hw_ops->query_poison_status) { + poison_stat = block_obj->hw_ops->query_poison_status(adev); + if (!poison_stat) { + /* Not poison consumption interrupt, no need to handle it */ + dev_info(adev->dev, "No RAS poison status in %s poison IH.\n", + block_obj->ras_comm.name); + + return; } } - /* gpu reset is fallback for all failed cases */ - if (need_reset) + if (!adev->gmc.xgmi.connected_to_cpu) + amdgpu_umc_poison_handler(adev, &err_data, false); + + if (block_obj->hw_ops->handle_poison_consumption) + poison_stat = block_obj->hw_ops->handle_poison_consumption(adev); + + /* gpu reset is fallback for failed and default cases */ + if (poison_stat) { + dev_info(adev->dev, "GPU reset for %s RAS poison consumption is issued!\n", + block_obj->ras_comm.name); amdgpu_ras_reset_gpu(adev); + } } static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj, -- cgit From 950d64250fba5fbb8c290f692de8ffdee380726c Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Wed, 27 Apr 2022 12:16:51 +0800 Subject: drm/amdgpu: support ras on SRIOV support umc/gfx/sdma ras on guest side Changed from V1: move sriov judgment in amdgpu_ras_interrupt_fatal_error_handler Signed-off-by: Stanley.Yang Reviewed-by: Tao Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 42 +++++++++++++++++++++--------- drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c | 4 +++ drivers/gpu/drm/amd/amdgpu/psp_v13_0.c | 9 ++++--- 4 files changed, 44 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9af8d7a1d011..c45736a902fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -5219,6 +5219,10 @@ retry: /* Rest of adevs pre asic reset from XGMI hive. */ r = amdgpu_device_reset_sriov(adev, job ? false : true); if (r) adev->asic_reset_res = r; + + /* Aldebaran supports ras in SRIOV, so need resume ras during reset */ + if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2)) + amdgpu_ras_resume(adev); } else { r = amdgpu_do_asic_reset(device_list_handle, &reset_context); if (r && r == -EAGAIN) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 035891ec59d5..2de9309a4193 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -726,7 +726,9 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, /* Do not enable if it is not allowed. */ WARN_ON(enable && !amdgpu_ras_is_feature_allowed(adev, head)); - if (!amdgpu_ras_intr_triggered()) { + /* Only enable ras feature operation handle on host side */ + if (!amdgpu_sriov_vf(adev) && + !amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n", @@ -1523,7 +1525,9 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) */ void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev) { - if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) + /* Fatal error events are handled on host side */ + if (amdgpu_sriov_vf(adev) || + !amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF)) return; if (adev->nbio.ras && @@ -2270,10 +2274,14 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) { adev->ras_hw_enabled = adev->ras_enabled = 0; - if (amdgpu_sriov_vf(adev) || !adev->is_atom_fw || + if (!adev->is_atom_fw || !amdgpu_ras_asic_supported(adev)) return; + if (!(amdgpu_sriov_vf(adev) && + (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2)))) + return; + if (!adev->gmc.xgmi.connected_to_cpu) { if (amdgpu_atomfirmware_mem_ecc_supported(adev)) { dev_info(adev->dev, "MEM ECC is active.\n"); @@ -2285,15 +2293,21 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) if (amdgpu_atomfirmware_sram_ecc_supported(adev)) { dev_info(adev->dev, "SRAM ECC is active.\n"); - adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | - 1 << AMDGPU_RAS_BLOCK__DF); - - if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0)) - adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | - 1 << AMDGPU_RAS_BLOCK__JPEG); - else - adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN | - 1 << AMDGPU_RAS_BLOCK__JPEG); + if (!amdgpu_sriov_vf(adev)) { + adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC | + 1 << AMDGPU_RAS_BLOCK__DF); + + if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0)) + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); + else + adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN | + 1 << AMDGPU_RAS_BLOCK__JPEG); + } else { + adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF | + 1 << AMDGPU_RAS_BLOCK__SDMA | + 1 << AMDGPU_RAS_BLOCK__GFX); + } } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } @@ -2637,6 +2651,10 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev) struct amdgpu_ras_block_object *obj; int r; + /* Guest side doesn't need init ras feature */ + if (amdgpu_sriov_vf(adev)) + return 0; + list_for_each_entry_safe(node, tmp, &adev->ras_list, node) { if (!node->ras_obj) { dev_warn(adev->dev, "Warning: abnormal ras list node.\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c index 8e221a1ba937..42c1f050542f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -124,6 +124,10 @@ int amdgpu_sdma_process_ras_data_cb(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { kgd2kfd_set_sram_ecc_flag(adev->kfd.dev); + + if (amdgpu_sriov_vf(adev)) + return AMDGPU_RAS_SUCCESS; + amdgpu_ras_reset_gpu(adev); return AMDGPU_RAS_SUCCESS; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c index d6d79e97def9..18014ed0e853 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v13_0.c @@ -85,9 +85,12 @@ static int psp_v13_0_init_microcode(struct psp_context *psp) err = psp_init_sos_microcode(psp, chip_name); if (err) return err; - err = psp_init_ta_microcode(&adev->psp, chip_name); - if (err) - return err; + /* It's not necessary to load ras ta on Guest side */ + if (!amdgpu_sriov_vf(adev)) { + err = psp_init_ta_microcode(&adev->psp, chip_name); + if (err) + return err; + } break; case IP_VERSION(13, 0, 1): case IP_VERSION(13, 0, 3): -- cgit From 28caf8c467e2cb98b90a760b65767fa9aaf52ffb Mon Sep 17 00:00:00 2001 From: "Stanley.Yang" Date: Tue, 31 May 2022 18:57:06 +0800 Subject: drm/amdgpu: fix ras supported check Fix aldebaran ras supported check on SRIOV guest side, the previous check conditicon block all ras feature on baremetal Signed-off-by: Stanley.Yang Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 2de9309a4193..1b1b502897ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2278,8 +2278,9 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev) !amdgpu_ras_asic_supported(adev)) return; - if (!(amdgpu_sriov_vf(adev) && - (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2)))) + /* If driver run on sriov guest side, only enable ras for aldebaran */ + if (amdgpu_sriov_vf(adev) && + adev->ip_versions[MP1_HWIP][0] != IP_VERSION(13, 0, 2)) return; if (!adev->gmc.xgmi.connected_to_cpu) { -- cgit From 2a460963350ec6b1534d28d7f943b5f84815aff2 Mon Sep 17 00:00:00 2001 From: Candice Li Date: Wed, 1 Jun 2022 17:10:44 +0800 Subject: drm/amdgpu: Resolve RAS GFX error count issue after cold boot on Arcturus Adjust the sequence for ras late init and separate ras reset error status from query status. v2: squash in fix from Candice Signed-off-by: Candice Li Reviewed-by: Hawking Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 9 ++++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 27 ++++++++++++++++++++++----- 2 files changed, 28 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index ede2fa56f6c9..16699158e00d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -594,17 +594,20 @@ int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value) int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block) { int r; - r = amdgpu_ras_block_late_init(adev, ras_block); - if (r) - return r; if (amdgpu_ras_is_supported(adev, ras_block->block)) { if (!amdgpu_persistent_edc_harvesting_supported(adev)) amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX); + r = amdgpu_ras_block_late_init(adev, ras_block); + if (r) + return r; + r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0); if (r) goto late_fini; + } else { + amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 1b1b502897ef..dac202ae864d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -197,6 +197,13 @@ static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf, if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; + /* Hardware counter will be reset automatically after the query on Vega20 and Arcturus */ + if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { + if (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) + dev_warn(obj->adev->dev, "Failed to reset error counter and error status"); + } + s = snprintf(val, sizeof(val), "%s: %lu\n%s: %lu\n", "ue", info.ue_count, "ce", info.ce_count); @@ -550,9 +557,10 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev, if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; - if (obj->adev->asic_type == CHIP_ALDEBARAN) { + if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { if (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) - DRM_WARN("Failed to reset error counter and error status"); + dev_warn(obj->adev->dev, "Failed to reset error counter and error status"); } return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count, @@ -1027,9 +1035,6 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, } } - if (!amdgpu_persistent_edc_harvesting_supported(adev)) - amdgpu_ras_reset_error_status(adev, info->head.block); - return 0; } @@ -1149,6 +1154,12 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev, if (res) return res; + if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { + if (amdgpu_ras_reset_error_status(adev, info.head.block)) + dev_warn(adev->dev, "Failed to reset error counter and error status"); + } + ce += info.ce_count; ue += info.ue_count; } @@ -1792,6 +1803,12 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) continue; amdgpu_ras_query_error_status(adev, &info); + + if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) && + adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) { + if (amdgpu_ras_reset_error_status(adev, info.head.block)) + dev_warn(adev->dev, "Failed to reset error counter and error status"); + } } } -- cgit