diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 385 |
1 files changed, 304 insertions, 81 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index 96a8fd0ca1df..e8875351967e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -35,7 +35,11 @@ #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" #include "atom.h" +#ifdef CONFIG_X86_MCE_AMD +#include <asm/mce.h> +static bool notifier_registered; +#endif static const char *RAS_FS_NAME = "ras"; const char *ras_error_string[] = { @@ -61,8 +65,30 @@ const char *ras_block_string[] = { "mp0", "mp1", "fuse", + "mca", }; +const char *ras_mca_block_string[] = { + "mca_mp0", + "mca_mp1", + "mca_mpio", + "mca_iohc", +}; + +const char *get_ras_block_str(struct ras_common_if *ras_block) +{ + if (!ras_block) + return "NULL"; + + if (ras_block->block >= AMDGPU_RAS_BLOCK_COUNT) + return "OUT OF RANGE"; + + if (ras_block->block == AMDGPU_RAS_BLOCK__MCA) + return ras_mca_block_string[ras_block->sub_block_index]; + + return ras_block_string[ras_block->block]; +} + #define ras_err_str(i) (ras_error_string[ffs(i)]) #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS) @@ -85,6 +111,9 @@ static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con, uint64_t addr); static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev, uint64_t addr); +#ifdef CONFIG_X86_MCE_AMD +static void amdgpu_register_bad_pages_mca_notifier(void); +#endif void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready) { @@ -187,7 +216,7 @@ static int amdgpu_ras_find_block_id_by_name(const char *name, int *block_id) for (i = 0; i < ARRAY_SIZE(ras_block_string); i++) { *block_id = i; - if (strcmp(name, ras_block_str(i)) == 0) + if (strcmp(name, ras_block_string[i]) == 0) return 0; } return -EINVAL; @@ -509,7 +538,6 @@ 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 (amdgpu_ras_reset_error_status(obj->adev, info.head.block)) DRM_WARN("Failed to reset error counter and error status"); @@ -529,7 +557,7 @@ static inline void put_obj(struct ras_manager *obj) if (obj && (--obj->use == 0)) list_del(&obj->node); if (obj && (obj->use < 0)) - DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", ras_block_str(obj->head.block)); + DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", get_ras_block_str(&obj->head)); } /* make one obj and return it. */ @@ -545,7 +573,14 @@ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev, if (head->block >= AMDGPU_RAS_BLOCK_COUNT) return NULL; - obj = &con->objs[head->block]; + if (head->block == AMDGPU_RAS_BLOCK__MCA) { + if (head->sub_block_index >= AMDGPU_RAS_MCA_BLOCK__LAST) + return NULL; + + obj = &con->objs[AMDGPU_RAS_BLOCK__LAST + head->sub_block_index]; + } else + obj = &con->objs[head->block]; + /* already exist. return obj? */ if (alive_obj(obj)) return NULL; @@ -573,19 +608,21 @@ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, if (head->block >= AMDGPU_RAS_BLOCK_COUNT) return NULL; - obj = &con->objs[head->block]; + if (head->block == AMDGPU_RAS_BLOCK__MCA) { + if (head->sub_block_index >= AMDGPU_RAS_MCA_BLOCK__LAST) + return NULL; + + obj = &con->objs[AMDGPU_RAS_BLOCK__LAST + head->sub_block_index]; + } else + obj = &con->objs[head->block]; - if (alive_obj(obj)) { - WARN_ON(head->block != obj->head.block); + if (alive_obj(obj)) return obj; - } } else { - for (i = 0; i < AMDGPU_RAS_BLOCK_COUNT; i++) { + for (i = 0; i < AMDGPU_RAS_BLOCK_COUNT + AMDGPU_RAS_MCA_BLOCK_COUNT; i++) { obj = &con->objs[i]; - if (alive_obj(obj)) { - WARN_ON(i != obj->head.block); + if (alive_obj(obj)) return obj; - } } } @@ -626,8 +663,6 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev, */ if (!amdgpu_ras_is_feature_allowed(adev, head)) return 0; - if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head))) - return 0; if (enable) { if (!obj) { @@ -678,19 +713,14 @@ 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)); - /* Are we alerady in that state we are going to set? */ - if (!(!!enable ^ !!amdgpu_ras_is_feature_enabled(adev, head))) { - ret = 0; - goto out; - } if (!amdgpu_ras_intr_triggered()) { ret = psp_ras_enable_features(&adev->psp, info, enable); if (ret) { - dev_err(adev->dev, "ras %s %s failed %d\n", + dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n", enable ? "enable":"disable", - ras_block_str(head->block), - ret); + get_ras_block_str(head), + amdgpu_ras_is_poison_mode_supported(adev), ret); goto out; } } @@ -731,7 +761,7 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, if (!ret) dev_info(adev->dev, "RAS INFO: %s setup object\n", - ras_block_str(head->block)); + get_ras_block_str(head)); } } else { /* setup the object then issue a ras TA disable cmd.*/ @@ -781,17 +811,39 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, bool bypass) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); - int ras_block_count = AMDGPU_RAS_BLOCK_COUNT; int i; - const enum amdgpu_ras_error_type default_ras_type = - AMDGPU_RAS_ERROR__NONE; + const enum amdgpu_ras_error_type default_ras_type = AMDGPU_RAS_ERROR__NONE; - for (i = 0; i < ras_block_count; i++) { + for (i = 0; i < AMDGPU_RAS_BLOCK_COUNT; i++) { struct ras_common_if head = { .block = i, .type = default_ras_type, .sub_block_index = 0, }; + + if (i == AMDGPU_RAS_BLOCK__MCA) + continue; + + if (bypass) { + /* + * bypass psp. vbios enable ras for us. + * so just create the obj + */ + if (__amdgpu_ras_feature_enable(adev, &head, 1)) + break; + } else { + if (amdgpu_ras_feature_enable(adev, &head, 1)) + break; + } + } + + for (i = 0; i < AMDGPU_RAS_MCA_BLOCK_COUNT; i++) { + struct ras_common_if head = { + .block = AMDGPU_RAS_BLOCK__MCA, + .type = default_ras_type, + .sub_block_index = i, + }; + if (bypass) { /* * bypass psp. vbios enable ras for us. @@ -809,6 +861,32 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev, } /* feature ctl end */ + +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; + } +} + /* query/inject/cure begin */ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) @@ -872,6 +950,9 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, 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; default: break; } @@ -893,13 +974,13 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->smuio.funcs->get_socket_id(adev), adev->smuio.funcs->get_die_id(adev), obj->err_data.ce_count, - ras_block_str(info->head.block)); + get_ras_block_str(&info->head)); } else { dev_info(adev->dev, "%ld correctable hardware errors " "detected in %s block, no user " "action is needed.\n", obj->err_data.ce_count, - ras_block_str(info->head.block)); + get_ras_block_str(&info->head)); } } if (err_data.ue_count) { @@ -912,15 +993,18 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, adev->smuio.funcs->get_socket_id(adev), adev->smuio.funcs->get_die_id(adev), obj->err_data.ue_count, - ras_block_str(info->head.block)); + get_ras_block_str(&info->head)); } else { dev_info(adev->dev, "%ld uncorrectable hardware errors " "detected in %s block\n", obj->err_data.ue_count, - ras_block_str(info->head.block)); + get_ras_block_str(&info->head)); } } + if (!amdgpu_persistent_edc_harvesting_supported(adev)) + amdgpu_ras_reset_error_status(adev, info->head.block); + return 0; } @@ -1027,6 +1111,7 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, 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: @@ -1034,13 +1119,13 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev, break; default: dev_info(adev->dev, "%s error injection is not supported yet\n", - ras_block_str(info->head.block)); + get_ras_block_str(&info->head)); ret = -EINVAL; } if (ret) dev_err(adev->dev, "ras inject %s failed %d\n", - ras_block_str(info->head.block), ret); + get_ras_block_str(&info->head), ret); return ret; } @@ -1383,7 +1468,7 @@ void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev) if (amdgpu_ras_is_supported(adev, obj->head.block) && (obj->attr_inuse == 1)) { sprintf(fs_info.debugfs_name, "%s_err_inject", - ras_block_str(obj->head.block)); + get_ras_block_str(&obj->head)); fs_info.head = obj->head; amdgpu_ras_debugfs_create(adev, &fs_info, dir); } @@ -1469,22 +1554,28 @@ static void amdgpu_ras_interrupt_handler(struct ras_manager *obj) data->rptr = (data->aligned_element_size + data->rptr) % data->ring_size; - /* Let IP handle its data, maybe we need get the output - * from the callback to udpate the error type/count, etc - */ if (data->cb) { - 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 + 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 + */ + 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. */ - obj->err_data.ue_count += err_data.ue_count; - obj->err_data.ce_count += err_data.ce_count; + 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; + } } } } @@ -2014,6 +2105,11 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev) adev->smu.ppt_funcs->send_hbm_bad_pages_num(&adev->smu, con->eeprom_control.ras_num_recs); } +#ifdef CONFIG_X86_MCE_AMD + if ((adev->asic_type == CHIP_ALDEBARAN) && + (adev->gmc.xgmi.connected_to_cpu)) + amdgpu_register_bad_pages_mca_notifier(); +#endif return 0; free: @@ -2056,19 +2152,6 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev) } /* recovery end */ -/* return 0 if ras will reset gpu and repost.*/ -int amdgpu_ras_request_reset_on_boot(struct amdgpu_device *adev, - unsigned int block) -{ - struct amdgpu_ras *ras = amdgpu_ras_get_context(adev); - - if (!ras) - return -EINVAL; - - ras->flags |= AMDGPU_RAS_FLAG_INIT_NEED_RESET; - return 0; -} - static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev) { return adev->asic_type == CHIP_VEGA10 || @@ -2176,12 +2259,14 @@ int amdgpu_ras_init(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int r; + bool df_poison, umc_poison; if (con) return 0; con = kmalloc(sizeof(struct amdgpu_ras) + - sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT, + sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT + + sizeof(struct ras_manager) * AMDGPU_RAS_MCA_BLOCK_COUNT, GFP_KERNEL|__GFP_ZERO); if (!con) return -ENOMEM; @@ -2245,6 +2330,23 @@ int amdgpu_ras_init(struct amdgpu_device *adev) goto release_con; } + /* Init poison supported flag, the default value is false */ + if (adev->df.funcs && + adev->df.funcs->query_ras_poison_mode && + adev->umc.ras_funcs && + adev->umc.ras_funcs->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); + /* Only poison is set in both DF and UMC, we can support it */ + if (df_poison && umc_poison) + con->poison_supported = true; + else if (df_poison != umc_poison) + dev_warn(adev->dev, "Poison setting is inconsistent in DF/UMC(%d:%d)!\n", + df_poison, umc_poison); + } + if (amdgpu_ras_fs_init(adev)) { r = -EINVAL; goto release_con; @@ -2288,6 +2390,16 @@ static int amdgpu_persistent_edc_harvesting(struct amdgpu_device *adev, return 0; } +bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev) +{ + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); + + if (!con) + return false; + + return con->poison_supported; +} + /* 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, @@ -2306,12 +2418,7 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev, r = amdgpu_ras_feature_enable_on_boot(adev, ras_block, 1); if (r) { - if (r == -EAGAIN) { - /* request gpu reset. will run again */ - amdgpu_ras_request_reset_on_boot(adev, - ras_block->block); - return 0; - } else if (adev->in_suspend || amdgpu_in_reset(adev)) { + if (adev->in_suspend || amdgpu_in_reset(adev)) { /* in resume phase, if fail to enable ras, * clean up all ras fs nodes, and disable ras */ goto cleanup; @@ -2403,19 +2510,6 @@ void amdgpu_ras_resume(struct amdgpu_device *adev) } } } - - if (con->flags & AMDGPU_RAS_FLAG_INIT_NEED_RESET) { - con->flags &= ~AMDGPU_RAS_FLAG_INIT_NEED_RESET; - /* setup ras obj state as disabled. - * for init_by_vbios case. - * if we want to enable ras, just enable it in a normal way. - * If we want do disable it, need setup ras obj as enabled, - * then issue another TA disable cmd. - * See feature_enable_on_boot - */ - amdgpu_ras_disable_all_features(adev, 1); - amdgpu_ras_reset_gpu(adev); - } } void amdgpu_ras_suspend(struct amdgpu_device *adev) @@ -2507,3 +2601,132 @@ void amdgpu_release_ras_context(struct amdgpu_device *adev) kfree(con); } } + +#ifdef CONFIG_X86_MCE_AMD +static struct amdgpu_device *find_adev(uint32_t node_id) +{ + struct amdgpu_gpu_instance *gpu_instance; + int i; + struct amdgpu_device *adev = NULL; + + mutex_lock(&mgpu_info.mutex); + + for (i = 0; i < mgpu_info.num_gpu; i++) { + gpu_instance = &(mgpu_info.gpu_ins[i]); + adev = gpu_instance->adev; + + if (adev->gmc.xgmi.connected_to_cpu && + adev->gmc.xgmi.physical_node_id == node_id) + break; + adev = NULL; + } + + mutex_unlock(&mgpu_info.mutex); + + return adev; +} + +#define GET_MCA_IPID_GPUID(m) (((m) >> 44) & 0xF) +#define GET_UMC_INST(m) (((m) >> 21) & 0x7) +#define GET_CHAN_INDEX(m) ((((m) >> 12) & 0x3) | (((m) >> 18) & 0x4)) +#define GPU_ID_OFFSET 8 + +static int amdgpu_bad_page_notifier(struct notifier_block *nb, + unsigned long val, void *data) +{ + struct mce *m = (struct mce *)data; + struct amdgpu_device *adev = NULL; + uint32_t gpu_id = 0; + uint32_t umc_inst = 0; + uint32_t ch_inst, channel_index = 0; + struct ras_err_data err_data = {0, 0, 0, NULL}; + struct eeprom_table_record err_rec; + uint64_t retired_page; + + /* + * If the error was generated in UMC_V2, which belongs to GPU UMCs, + * and error occurred in DramECC (Extended error code = 0) then only + * process the error, else bail out. + */ + if (!m || !((smca_get_bank_type(m->bank) == SMCA_UMC_V2) && + (XEC(m->status, 0x3f) == 0x0))) + return NOTIFY_DONE; + + /* + * If it is correctable error, return. + */ + if (mce_is_correctable(m)) + return NOTIFY_OK; + + /* + * GPU Id is offset by GPU_ID_OFFSET in MCA_IPID_UMC register. + */ + gpu_id = GET_MCA_IPID_GPUID(m->ipid) - GPU_ID_OFFSET; + + adev = find_adev(gpu_id); + if (!adev) { + DRM_WARN("%s: Unable to find adev for gpu_id: %d\n", __func__, + gpu_id); + return NOTIFY_DONE; + } + + /* + * If it is uncorrectable error, then find out UMC instance and + * channel index. + */ + umc_inst = GET_UMC_INST(m->ipid); + ch_inst = GET_CHAN_INDEX(m->ipid); + + 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 + */ + channel_index = + adev->umc.channel_idx_tbl[umc_inst * adev->umc.channel_inst_num + + ch_inst]; + + retired_page = ADDR_OF_8KB_BLOCK(m->addr) | + 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; + + err_data.err_addr = &err_rec; + err_data.err_addr_cnt = 1; + + if (amdgpu_bad_page_threshold != 0) { + amdgpu_ras_add_bad_pages(adev, err_data.err_addr, + err_data.err_addr_cnt); + amdgpu_ras_save_bad_pages(adev); + } + + return NOTIFY_OK; +} + +static struct notifier_block amdgpu_bad_page_nb = { + .notifier_call = amdgpu_bad_page_notifier, + .priority = MCE_PRIO_UC, +}; + +static void amdgpu_register_bad_pages_mca_notifier(void) +{ + /* + * Register the x86 notifier only once + * with MCE subsystem. + */ + if (notifier_registered == false) { + mce_register_decode_chain(&amdgpu_bad_page_nb); + notifier_registered = true; + } +} +#endif |