/* * Copyright 2018 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * * */ #include #include #include #include #include #include #include #include "amdgpu.h" #include "amdgpu_ras.h" #include "amdgpu_atomfirmware.h" #include "amdgpu_xgmi.h" #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h" #include "atom.h" static const char *RAS_FS_NAME = "ras"; const char *ras_error_string[] = { "none", "parity", "single_correctable", "multi_uncorrectable", "poison", }; const char *ras_block_string[] = { "umc", "sdma", "gfx", "mmhub", "athub", "pcie_bif", "hdp", "xgmi_wafl", "df", "smn", "sem", "mp0", "mp1", "fuse", }; #define ras_err_str(i) (ras_error_string[ffs(i)]) #define ras_block_str(i) (ras_block_string[i]) #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS) /* inject address is 52 bits */ #define RAS_UMC_INJECT_ADDR_LIMIT (0x1ULL << 52) /* typical ECC bad page rate(1 bad page per 100MB VRAM) */ #define RAS_BAD_PAGE_RATE (100 * 1024 * 1024ULL) enum amdgpu_ras_retire_page_reservation { AMDGPU_RAS_RETIRE_PAGE_RESERVED, AMDGPU_RAS_RETIRE_PAGE_PENDING, AMDGPU_RAS_RETIRE_PAGE_FAULT, }; atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0); 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); void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready) { if (adev && amdgpu_ras_get_context(adev)) amdgpu_ras_get_context(adev)->error_query_ready = ready; } static bool amdgpu_ras_get_error_query_ready(struct amdgpu_device *adev) { if (adev && amdgpu_ras_get_context(adev)) return amdgpu_ras_get_context(adev)->error_query_ready; return false; } static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t address) { struct ras_err_data err_data = {0, 0, 0, NULL}; struct eeprom_table_record err_rec; if ((address >= adev->gmc.mc_vram_size) || (address >= RAS_UMC_INJECT_ADDR_LIMIT)) { dev_warn(adev->dev, "RAS WARN: input address 0x%llx is invalid.\n", address); return -EINVAL; } if (amdgpu_ras_check_bad_page(adev, address)) { dev_warn(adev->dev, "RAS WARN: 0x%llx has already been marked as bad page!\n", address); return 0; } 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; 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); } dev_warn(adev->dev, "WARNING: THIS IS ONLY FOR TEST PURPOSES AND WILL CORRUPT RAS EEPROM\n"); dev_warn(adev->dev, "Clear EEPROM:\n"); dev_warn(adev->dev, " echo 1 > /sys/kernel/debug/dri/0/ras/ras_eeprom_reset\n"); return 0; } static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { struct ras_manager *obj = (struct ras_manager *)file_inode(f)->i_private; struct ras_query_if info = { .head = obj->head, }; ssize_t s; char val[128]; if (amdgpu_ras_query_error_status(obj->adev, &info)) return -EINVAL; s = snprintf(val, sizeof(val), "%s: %lu\n%s: %lu\n", "ue", info.ue_count, "ce", info.ce_count); if (*pos >= s) return 0; s -= *pos; s = min_t(u64, s, size); if (copy_to_user(buf, &val[*pos], s)) return -EINVAL; *pos += s; return s; } static const struct file_operations amdgpu_ras_debugfs_ops = { .owner = THIS_MODULE, .read = amdgpu_ras_debugfs_read, .write = NULL, .llseek = default_llseek }; static int amdgpu_ras_find_block_id_by_name(const char *name, int *block_id) { int i; for (i = 0; i < ARRAY_SIZE(ras_block_string); i++) { *block_id = i; if (strcmp(name, ras_block_str(i)) == 0) return 0; } return -EINVAL; } static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f, const char __user *buf, size_t size, loff_t *pos, struct ras_debug_if *data) { ssize_t s = min_t(u64, 64, size); char str[65]; char block_name[33]; char err[9] = "ue"; int op = -1; int block_id; uint32_t sub_block; u64 address, value; if (*pos) return -EINVAL; *pos = size; memset(str, 0, sizeof(str)); memset(data, 0, sizeof(*data)); if (copy_from_user(str, buf, s)) return -EINVAL; if (sscanf(str, "disable %32s", block_name) == 1) op = 0; else if (sscanf(str, "enable %32s %8s", block_name, err) == 2) op = 1; else if (sscanf(str, "inject %32s %8s", block_name, err) == 2) op = 2; else if (strstr(str, "retire_page") != NULL) op = 3; else if (str[0] && str[1] && str[2] && str[3]) /* ascii string, but commands are not matched. */ return -EINVAL; if (op != -1) { if (op == 3) { if (sscanf(str, "%*s 0x%llx", &address) != 1 && sscanf(str, "%*s %llu", &address) != 1) return -EINVAL; data->op = op; data->inject.address = address; return 0; } if (amdgpu_ras_find_block_id_by_name(block_name, &block_id)) return -EINVAL; data->head.block = block_id; /* only ue and ce errors are supported */ if (!memcmp("ue", err, 2)) data->head.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE; else if (!memcmp("ce", err, 2)) data->head.type = AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE; else return -EINVAL; data->op = op; if (op == 2) { if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx", &sub_block, &address, &value) != 3 && sscanf(str, "%*s %*s %*s %u %llu %llu", &sub_block, &address, &value) != 3) return -EINVAL; data->head.sub_block_index = sub_block; data->inject.address = address; data->inject.value = value; } } else { if (size < sizeof(*data)) return -EINVAL; if (copy_from_user(data, buf, sizeof(*data))) return -EINVAL; } return 0; } /** * DOC: AMDGPU RAS debugfs control interface * * The control interface accepts struct ras_debug_if which has two members. * * First member: ras_debug_if::head or ras_debug_if::inject. * * head is used to indicate which IP block will be under control. * * head has four members, they are block, type, sub_block_index, name. * block: which IP will be under control. * type: what kind of error will be enabled/disabled/injected. * sub_block_index: some IPs have subcomponets. say, GFX, sDMA. * name: the name of IP. * * inject has two more members than head, they are address, value. * As their names indicate, inject operation will write the * value to the address. * * The second member: struct ras_debug_if::op. * It has three kinds of operations. * * - 0: disable RAS on the block. Take ::head as its data. * - 1: enable RAS on the block. Take ::head as its data. * - 2: inject errors on the block. Take ::inject as its data. * * How to use the interface? * * In a program * * Copy the struct ras_debug_if in your code and initialize it. * Write the struct to the control interface. * * From shell * * .. code-block:: bash * * echo "disable " > /sys/kernel/debug/dri//ras/ras_ctrl * echo "enable " > /sys/kernel/debug/dri//ras/ras_ctrl * echo "inject
> /sys/kernel/debug/dri//ras/ras_ctrl * * Where N, is the card which you want to affect. * * "disable" requires only the block. * "enable" requires the block and error type. * "inject" requires the block, error type, address, and value. * * The block is one of: umc, sdma, gfx, etc. * see ras_block_string[] for details * * The error type is one of: ue, ce, where, * ue is multi-uncorrectable * ce is single-correctable * * The sub-block is a the sub-block index, pass 0 if there is no sub-block. * The address and value are hexadecimal numbers, leading 0x is optional. * * For instance, * * .. code-block:: bash * * echo inject umc ue 0x0 0x0 0x0 > /sys/kernel/debug/dri/0/ras/ras_ctrl * echo inject umc ce 0 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl * echo disable umc > /sys/kernel/debug/dri/0/ras/ras_ctrl * * How to check the result of the operation? * * To check disable/enable, see "ras" features at, * /sys/class/drm/card[0/1/2...]/device/ras/features * * To check inject, see the corresponding error count at, * /sys/class/drm/card[0/1/2...]/device/ras/[gfx|sdma|umc|...]_err_count * * .. note:: * Operations are only allowed on blocks which are supported. * Check the "ras" mask at /sys/module/amdgpu/parameters/ras_mask * to see which blocks support RAS on a particular asic. * */ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; struct ras_debug_if data; int ret = 0; if (!amdgpu_ras_get_error_query_ready(adev)) { dev_warn(adev->dev, "RAS WARN: error injection " "currently inaccessible\n"); return size; } ret = amdgpu_ras_debugfs_ctrl_parse_data(f, buf, size, pos, &data); if (ret) return -EINVAL; if (data.op == 3) { ret = amdgpu_reserve_page_direct(adev, data.inject.address); if (!ret) return size; else return ret; } if (!amdgpu_ras_is_supported(adev, data.head.block)) return -EINVAL; switch (data.op) { case 0: ret = amdgpu_ras_feature_enable(adev, &data.head, 0); break; case 1: ret = amdgpu_ras_feature_enable(adev, &data.head, 1); break; case 2: if ((data.inject.address >= adev->gmc.mc_vram_size) || (data.inject.address >= RAS_UMC_INJECT_ADDR_LIMIT)) { dev_warn(adev->dev, "RAS WARN: input address " "0x%llx is invalid.", data.inject.address); ret = -EINVAL; break; } /* umc ce/ue error injection for a bad page is not allowed */ if ((data.head.block == AMDGPU_RAS_BLOCK__UMC) && amdgpu_ras_check_bad_page(adev, data.inject.address)) { dev_warn(adev->dev, "RAS WARN: 0x%llx has been marked " "as bad before error injection!\n", data.inject.address); break; } /* data.inject.address is offset instead of absolute gpu address */ ret = amdgpu_ras_error_inject(adev, &data.inject); break; default: ret = -EINVAL; break; } if (ret) return -EINVAL; return size; } /** * DOC: AMDGPU RAS debugfs EEPROM table reset interface * * Some boards contain an EEPROM which is used to persistently store a list of * bad pages which experiences ECC errors in vram. This interface provides * a way to reset the EEPROM, e.g., after testing error injection. * * Usage: * * .. code-block:: bash * * echo 1 > ../ras/ras_eeprom_reset * * will reset EEPROM table to 0 entries. * */ static ssize_t amdgpu_ras_debugfs_eeprom_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private; int ret; ret = amdgpu_ras_eeprom_reset_table( &(amdgpu_ras_get_context(adev)->eeprom_control)); if (ret == 1) { amdgpu_ras_get_context(adev)->flags = RAS_DEFAULT_FLAGS; return size; } else { return -EIO; } } static const struct file_operations amdgpu_ras_debugfs_ctrl_ops = { .owner = THIS_MODULE, .read = NULL, .write = amdgpu_ras_debugfs_ctrl_write, .llseek = default_llseek }; static const struct file_operations amdgpu_ras_debugfs_eeprom_ops = { .owner = THIS_MODULE, .read = NULL, .write = amdgpu_ras_debugfs_eeprom_write, .llseek = default_llseek }; /** * DOC: AMDGPU RAS sysfs Error Count Interface * * It allows the user to read the error count for each IP block on the gpu through * /sys/class/drm/card[0/1/2...]/device/ras/[gfx/sdma/...]_err_count * * It outputs the multiple lines which report the uncorrected (ue) and corrected * (ce) error counts. * * The format of one line is below, * * [ce|ue]: count * * Example: * * .. code-block:: bash * * ue: 0 * ce: 1 * */ static ssize_t amdgpu_ras_sysfs_read(struct device *dev, struct device_attribute *attr, char *buf) { struct ras_manager *obj = container_of(attr, struct ras_manager, sysfs_attr); struct ras_query_if info = { .head = obj->head, }; if (!amdgpu_ras_get_error_query_ready(obj->adev)) return sysfs_emit(buf, "Query currently inaccessible\n"); 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"); } return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count, "ce", info.ce_count); } /* obj begin */ #define get_obj(obj) do { (obj)->use++; } while (0) #define alive_obj(obj) ((obj)->use) 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", obj->head.name); } /* make one obj and return it. */ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev, struct ras_common_if *head) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; if (!adev->ras_enabled || !con) return NULL; if (head->block >= AMDGPU_RAS_BLOCK_COUNT) return NULL; obj = &con->objs[head->block]; /* already exist. return obj? */ if (alive_obj(obj)) return NULL; obj->head = *head; obj->adev = adev; list_add(&obj->node, &con->head); get_obj(obj); return obj; } /* return an obj equal to head, or the first when head is NULL */ struct ras_manager *amdgpu_ras_find_obj(struct amdgpu_device *adev, struct ras_common_if *head) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; int i; if (!adev->ras_enabled || !con) return NULL; if (head) { if (head->block >= AMDGPU_RAS_BLOCK_COUNT) return NULL; obj = &con->objs[head->block]; if (alive_obj(obj)) { WARN_ON(head->block != obj->head.block); return obj; } } else { for (i = 0; i < AMDGPU_RAS_BLOCK_COUNT; i++) { obj = &con->objs[i]; if (alive_obj(obj)) { WARN_ON(i != obj->head.block); return obj; } } } return NULL; } /* obj end */ /* feature ctl begin */ static int amdgpu_ras_is_feature_allowed(struct amdgpu_device *adev, struct ras_common_if *head) { return adev->ras_hw_enabled & BIT(head->block); } static int amdgpu_ras_is_feature_enabled(struct amdgpu_device *adev, struct ras_common_if *head) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); return con->features & BIT(head->block); } /* * if obj is not created, then create one. * set feature enable flag. */ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev, struct ras_common_if *head, int enable) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); /* If hardware does not support ras, then do not create obj. * But if hardware support ras, we can create the obj. * Ras framework checks con->hw_supported to see if it need do * corresponding initialization. * IP checks con->support to see if it need disable ras. */ 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) { obj = amdgpu_ras_create_obj(adev, head); if (!obj) return -EINVAL; } else { /* In case we create obj somewhere else */ get_obj(obj); } con->features |= BIT(head->block); } else { if (obj && amdgpu_ras_is_feature_enabled(adev, head)) { con->features &= ~BIT(head->block); put_obj(obj); } } return 0; } /* wrapper of psp_ras_enable_features */ int amdgpu_ras_feature_enable(struct amdgpu_device *adev, struct ras_common_if *head, bool enable) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); union ta_ras_cmd_input *info; int ret; if (!con) return -EINVAL; info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL); if (!info) return -ENOMEM; if (!enable) { info->disable_features = (struct ta_ras_disable_features_input) { .block_id = amdgpu_ras_block_to_ta(head->block), .error_type = amdgpu_ras_error_to_ta(head->type), }; } else { info->enable_features = (struct ta_ras_enable_features_input) { .block_id = amdgpu_ras_block_to_ta(head->block), .error_type = amdgpu_ras_error_to_ta(head->type), }; } /* 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", enable ? "enable":"disable", ras_block_str(head->block), ret); goto out; } } /* setup the obj */ __amdgpu_ras_feature_enable(adev, head, enable); ret = 0; out: kfree(info); return ret; } /* Only used in device probe stage and called only once. */ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev, struct ras_common_if *head, bool enable) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int ret; if (!con) return -EINVAL; if (con->flags & AMDGPU_RAS_FLAG_INIT_BY_VBIOS) { if (enable) { /* There is no harm to issue a ras TA cmd regardless of * the currecnt ras state. * If current state == target state, it will do nothing * But sometimes it requests driver to reset and repost * with error code -EAGAIN. */ ret = amdgpu_ras_feature_enable(adev, head, 1); /* With old ras TA, we might fail to enable ras. * Log it and just setup the object. * TODO need remove this WA in the future. */ if (ret == -EINVAL) { ret = __amdgpu_ras_feature_enable(adev, head, 1); if (!ret) dev_info(adev->dev, "RAS INFO: %s setup object\n", ras_block_str(head->block)); } } else { /* setup the object then issue a ras TA disable cmd.*/ ret = __amdgpu_ras_feature_enable(adev, head, 1); if (ret) return ret; /* gfx block ras dsiable cmd must send to ras-ta */ if (head->block == AMDGPU_RAS_BLOCK__GFX) con->features |= BIT(head->block); ret = amdgpu_ras_feature_enable(adev, head, 0); /* clean gfx block ras features flag */ if (adev->ras_enabled && head->block == AMDGPU_RAS_BLOCK__GFX) con->features &= ~BIT(head->block); } } else ret = amdgpu_ras_feature_enable(adev, head, enable); return ret; } static int amdgpu_ras_disable_all_features(struct amdgpu_device *adev, bool bypass) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj, *tmp; list_for_each_entry_safe(obj, tmp, &con->head, node) { /* bypass psp. * aka just release the obj and corresponding flags */ if (bypass) { if (__amdgpu_ras_feature_enable(adev, &obj->head, 0)) break; } else { if (amdgpu_ras_feature_enable(adev, &obj->head, 0)) break; } } return con->features; } 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; for (i = 0; i < ras_block_count; i++) { struct ras_common_if head = { .block = i, .type = default_ras_type, .sub_block_index = 0, }; strcpy(head.name, ras_block_str(i)); 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; } } return con->features; } /* feature ctl end */ /* query/inject/cure begin */ int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info) { 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; switch (info->head.block) { case AMDGPU_RAS_BLOCK__UMC: if (adev->umc.ras_funcs && adev->umc.ras_funcs->query_ras_error_count) adev->umc.ras_funcs->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); 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: 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_funcs && adev->gfx.ras_funcs->query_ras_error_status) adev->gfx.ras_funcs->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) 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); 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; default: break; } obj->err_data.ue_count += err_data.ue_count; obj->err_data.ce_count += err_data.ce_count; info->ue_count = obj->err_data.ue_count; info->ce_count = obj->err_data.ce_count; if (err_data.ce_count) { if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id && adev->smuio.funcs->get_die_id) { dev_info(adev->dev, "socket: %d, die: %d " "%ld correctable hardware errors " "detected in %s block, no user " "action is needed.\n", 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)); } 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)); } } if (err_data.ue_count) { if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id && adev->smuio.funcs->get_die_id) { dev_info(adev->dev, "socket: %d, die: %d " "%ld uncorrectable hardware errors " "detected in %s block\n", 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)); } 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)); } } return 0; } int amdgpu_ras_reset_error_status(struct amdgpu_device *adev, enum amdgpu_ras_block block) { 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 (adev->gfx.ras_funcs && adev->gfx.ras_funcs->reset_ras_error_status) adev->gfx.ras_funcs->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); 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); break; default: break; } 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) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ta_ras_trigger_error_input block_info = { .block_id = amdgpu_ras_block_to_ta(info->head.block), .inject_error_type = amdgpu_ras_error_to_ta(info->head.type), .sub_block_index = info->head.sub_block_index, .address = info->address, .value = info->value, }; int ret = 0; if (!obj) return -EINVAL; /* Calculate XGMI relative offset */ if (adev->gmc.xgmi.num_physical_nodes > 1) { block_info.address = amdgpu_xgmi_get_relative_phy_addr(adev, block_info.address); } 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; break; case AMDGPU_RAS_BLOCK__UMC: case AMDGPU_RAS_BLOCK__SDMA: case AMDGPU_RAS_BLOCK__MMHUB: case AMDGPU_RAS_BLOCK__PCIE_BIF: 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", ras_block_str(info->head.block)); ret = -EINVAL; } if (ret) dev_err(adev->dev, "ras inject %s failed %d\n", ras_block_str(info->head.block), ret); return ret; } /** * amdgpu_ras_query_error_count -- Get error counts of all IPs * adev: pointer to AMD GPU device * ce_count: pointer to an integer to be set to the count of correctible errors. * ue_count: pointer to an integer to be set to the count of uncorrectible * errors. * * If set, @ce_count or @ue_count, count and return the corresponding * error counts in those integer pointers. Return 0 if the device * supports RAS. Return -EOPNOTSUPP if the device doesn't support RAS. */ int amdgpu_ras_query_error_count(struct amdgpu_device *adev, unsigned long *ce_count, unsigned long *ue_count) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; unsigned long ce, ue; if (!adev->ras_enabled || !con) return -EOPNOTSUPP; /* Don't count since no reporting. */ if (!ce_count && !ue_count) return 0; ce = 0; ue = 0; list_for_each_entry(obj, &con->head, node) { struct ras_query_if info = { .head = obj->head, }; int res; res = amdgpu_ras_query_error_status(adev, &info); if (res) return res; ce += info.ce_count; ue += info.ue_count; } if (ce_count) *ce_count = ce; if (ue_count) *ue_count = ue; return 0; } /* query/inject/cure end */ /* sysfs begin */ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev, struct ras_badpage **bps, unsigned int *count); static char *amdgpu_ras_badpage_flags_str(unsigned int flags) { switch (flags) { case AMDGPU_RAS_RETIRE_PAGE_RESERVED: return "R"; case AMDGPU_RAS_RETIRE_PAGE_PENDING: return "P"; case AMDGPU_RAS_RETIRE_PAGE_FAULT: default: return "F"; } } /** * DOC: AMDGPU RAS sysfs gpu_vram_bad_pages Interface * * It allows user to read the bad pages of vram on the gpu through * /sys/class/drm/card[0/1/2...]/device/ras/gpu_vram_bad_pages * * It outputs multiple lines, and each line stands for one gpu page. * * The format of one line is below, * gpu pfn : gpu page size : flags * * gpu pfn and gpu page size are printed in hex format. * flags can be one of below character, * * R: reserved, this gpu page is reserved and not able to use. * * P: pending for reserve, this gpu page is marked as bad, will be reserved * in next window of page_reserve. * * F: unable to reserve. this gpu page can't be reserved due to some reasons. * * Examples: * * .. code-block:: bash * * 0x00000001 : 0x00001000 : R * 0x00000002 : 0x00001000 : P * */ static ssize_t amdgpu_ras_sysfs_badpages_read(struct file *f, struct kobject *kobj, struct bin_attribute *attr, char *buf, loff_t ppos, size_t count) { struct amdgpu_ras *con = container_of(attr, struct amdgpu_ras, badpages_attr); struct amdgpu_device *adev = con->adev; const unsigned int element_size = sizeof("0xabcdabcd : 0x12345678 : R\n") - 1; unsigned int start = div64_ul(ppos + element_size - 1, element_size); unsigned int end = div64_ul(ppos + count - 1, element_size); ssize_t s = 0; struct ras_badpage *bps = NULL; unsigned int bps_count = 0; memset(buf, 0, count); if (amdgpu_ras_badpages_read(adev, &bps, &bps_count)) return 0; for (; start < end && start < bps_count; start++) s += scnprintf(&buf[s], element_size + 1, "0x%08x : 0x%08x : %1s\n", bps[start].bp, bps[start].size, amdgpu_ras_badpage_flags_str(bps[start].flags)); kfree(bps); return s; } static ssize_t amdgpu_ras_sysfs_features_read(struct device *dev, struct device_attribute *attr, char *buf) { struct amdgpu_ras *con = container_of(attr, struct amdgpu_ras, features_attr); return scnprintf(buf, PAGE_SIZE, "feature mask: 0x%x\n", con->features); } static void amdgpu_ras_sysfs_remove_bad_page_node(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); sysfs_remove_file_from_group(&adev->dev->kobj, &con->badpages_attr.attr, RAS_FS_NAME); } static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct attribute *attrs[] = { &con->features_attr.attr, NULL }; struct attribute_group group = { .name = RAS_FS_NAME, .attrs = attrs, }; sysfs_remove_group(&adev->dev->kobj, &group); return 0; } int amdgpu_ras_sysfs_create(struct amdgpu_device *adev, struct ras_fs_if *head) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->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)); obj->sysfs_attr = (struct device_attribute){ .attr = { .name = obj->fs_data.sysfs_name, .mode = S_IRUGO, }, .show = amdgpu_ras_sysfs_read, }; sysfs_attr_init(&obj->sysfs_attr.attr); if (sysfs_add_file_to_group(&adev->dev->kobj, &obj->sysfs_attr.attr, RAS_FS_NAME)) { put_obj(obj); return -EINVAL; } obj->attr_inuse = 1; return 0; } int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev, struct ras_common_if *head) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, head); if (!obj || !obj->attr_inuse) return -EINVAL; sysfs_remove_file_from_group(&adev->dev->kobj, &obj->sysfs_attr.attr, RAS_FS_NAME); obj->attr_inuse = 0; put_obj(obj); return 0; } static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj, *tmp; list_for_each_entry_safe(obj, tmp, &con->head, node) { amdgpu_ras_sysfs_remove(adev, &obj->head); } if (amdgpu_bad_page_threshold != 0) amdgpu_ras_sysfs_remove_bad_page_node(adev); amdgpu_ras_sysfs_remove_feature_node(adev); return 0; } /* sysfs end */ /** * DOC: AMDGPU RAS Reboot Behavior for Unrecoverable Errors * * Normally when there is an uncorrectable error, the driver will reset * the GPU to recover. However, in the event of an unrecoverable error, * the driver provides an interface to reboot the system automatically * in that event. * * The following file in debugfs provides that interface: * /sys/kernel/debug/dri/[0/1/2...]/ras/auto_reboot * * Usage: * * .. code-block:: bash * * echo true > .../ras/auto_reboot * */ /* debugfs begin */ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct drm_minor *minor = adev_to_drm(adev)->primary; struct dentry *dir; dir = debugfs_create_dir(RAS_FS_NAME, minor->debugfs_root); debugfs_create_file("ras_ctrl", S_IWUGO | S_IRUGO, dir, adev, &amdgpu_ras_debugfs_ctrl_ops); debugfs_create_file("ras_eeprom_reset", S_IWUGO | S_IRUGO, dir, adev, &amdgpu_ras_debugfs_eeprom_ops); debugfs_create_u32("bad_page_cnt_threshold", 0444, dir, &con->bad_page_cnt_threshold); debugfs_create_x32("ras_hw_enabled", 0444, dir, &adev->ras_hw_enabled); debugfs_create_x32("ras_enabled", 0444, dir, &adev->ras_enabled); /* * After one uncorrectable error happens, usually GPU recovery will * be scheduled. But due to the known problem in GPU recovery failing * to bring GPU back, below interface provides one direct way to * user to reboot system automatically in such case within * ERREVENT_ATHUB_INTERRUPT generated. Normal GPU recovery routine * will never be called. */ debugfs_create_bool("auto_reboot", S_IWUGO | S_IRUGO, dir, &con->reboot); /* * User could set this not to clean up hardware's error count register * of RAS IPs during ras recovery. */ debugfs_create_bool("disable_ras_err_cnt_harvest", 0644, dir, &con->disable_ras_err_cnt_harvest); return dir; } static void amdgpu_ras_debugfs_create(struct amdgpu_device *adev, struct ras_fs_if *head, struct dentry *dir) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->head); if (!obj || !dir) return; get_obj(obj); memcpy(obj->fs_data.debugfs_name, head->debugfs_name, sizeof(obj->fs_data.debugfs_name)); debugfs_create_file(obj->fs_data.debugfs_name, S_IWUGO | S_IRUGO, dir, obj, &amdgpu_ras_debugfs_ops); } void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct dentry *dir; struct ras_manager *obj; struct ras_fs_if fs_info; /* * it won't be called in resume path, no need to check * suspend and gpu reset status */ if (!IS_ENABLED(CONFIG_DEBUG_FS) || !con) return; dir = amdgpu_ras_debugfs_create_ctrl_node(adev); list_for_each_entry(obj, &con->head, node) { 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)); fs_info.head = obj->head; amdgpu_ras_debugfs_create(adev, &fs_info, dir); } } } /* debugfs end */ /* ras fs */ static BIN_ATTR(gpu_vram_bad_pages, S_IRUGO, amdgpu_ras_sysfs_badpages_read, NULL, 0); static DEVICE_ATTR(features, S_IRUGO, amdgpu_ras_sysfs_features_read, NULL); static int amdgpu_ras_fs_init(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct attribute_group group = { .name = RAS_FS_NAME, }; struct attribute *attrs[] = { &con->features_attr.attr, NULL }; struct bin_attribute *bin_attrs[] = { NULL, NULL, }; int r; /* add features entry */ con->features_attr = dev_attr_features; group.attrs = attrs; sysfs_attr_init(attrs[0]); if (amdgpu_bad_page_threshold != 0) { /* add bad_page_features entry */ bin_attr_gpu_vram_bad_pages.private = NULL; con->badpages_attr = bin_attr_gpu_vram_bad_pages; bin_attrs[0] = &con->badpages_attr; group.bin_attrs = bin_attrs; sysfs_bin_attr_init(bin_attrs[0]); } r = sysfs_create_group(&adev->dev->kobj, &group); if (r) dev_err(adev->dev, "Failed to create RAS sysfs group!"); return 0; } static int amdgpu_ras_fs_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *con_obj, *ip_obj, *tmp; if (IS_ENABLED(CONFIG_DEBUG_FS)) { list_for_each_entry_safe(con_obj, tmp, &con->head, node) { ip_obj = amdgpu_ras_find_obj(adev, &con_obj->head); if (ip_obj) put_obj(ip_obj); } } amdgpu_ras_sysfs_remove_all(adev); return 0; } /* ras fs end */ /* ih begin */ 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(); memcpy(&entry, &data->ring[data->rptr], data->element_size); wmb(); 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 */ obj->err_data.ue_count += err_data.ue_count; obj->err_data.ce_count += err_data.ce_count; } } } } static void amdgpu_ras_interrupt_process_handler(struct work_struct *work) { struct ras_ih_data *data = container_of(work, struct ras_ih_data, ih_work); struct ras_manager *obj = container_of(data, struct ras_manager, ih_data); amdgpu_ras_interrupt_handler(obj); } int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev, struct ras_dispatch_if *info) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_ih_data *data = &obj->ih_data; if (!obj) return -EINVAL; if (data->inuse == 0) return 0; /* Might be overflow... */ memcpy(&data->ring[data->wptr], info->entry, data->element_size); wmb(); data->wptr = (data->aligned_element_size + data->wptr) % data->ring_size; schedule_work(&data->ih_work); return 0; } int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev, struct ras_ih_if *info) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_ih_data *data; if (!obj) return -EINVAL; data = &obj->ih_data; if (data->inuse == 0) return 0; cancel_work_sync(&data->ih_work); kfree(data->ring); memset(data, 0, sizeof(*data)); put_obj(obj); return 0; } int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev, struct ras_ih_if *info) { struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head); struct ras_ih_data *data; if (!obj) { /* in case we registe the IH before enable ras feature */ obj = amdgpu_ras_create_obj(adev, &info->head); if (!obj) return -EINVAL; } else get_obj(obj); data = &obj->ih_data; /* add the callback.etc */ *data = (struct ras_ih_data) { .inuse = 0, .cb = info->cb, .element_size = sizeof(struct amdgpu_iv_entry), .rptr = 0, .wptr = 0, }; INIT_WORK(&data->ih_work, amdgpu_ras_interrupt_process_handler); data->aligned_element_size = ALIGN(data->element_size, 8); /* the ring can store 64 iv entries. */ data->ring_size = 64 * data->aligned_element_size; data->ring = kmalloc(data->ring_size, GFP_KERNEL); if (!data->ring) { put_obj(obj); return -ENOMEM; } /* IH is ready */ data->inuse = 1; return 0; } static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(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); } return 0; } /* ih end */ /* traversal all IPs except NBIO to query error counter */ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { struct ras_query_if info = { .head = obj->head, }; /* * PCIE_BIF IP has one different isr by ras controller * interrupt, the specific ras counter query will be * done in that isr. So skip such block from common * sync flood interrupt isr calling. */ if (info.head.block == AMDGPU_RAS_BLOCK__PCIE_BIF) continue; amdgpu_ras_query_error_status(adev, &info); } } /* Parse RdRspStatus and WrRspStatus */ static void amdgpu_ras_error_status_query(struct amdgpu_device *adev, struct ras_query_if *info) { /* * 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); 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; } } static void amdgpu_ras_query_err_status(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj; if (!adev->ras_enabled || !con) return; list_for_each_entry(obj, &con->head, node) { struct ras_query_if info = { .head = obj->head, }; amdgpu_ras_error_status_query(adev, &info); } } /* recovery begin */ /* return 0 on success. * caller need free bps. */ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev, struct ras_badpage **bps, unsigned int *count) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_err_handler_data *data; int i = 0; int ret = 0, status; if (!con || !con->eh_data || !bps || !count) return -EINVAL; mutex_lock(&con->recovery_lock); data = con->eh_data; if (!data || data->count == 0) { *bps = NULL; ret = -EINVAL; goto out; } *bps = kmalloc(sizeof(struct ras_badpage) * data->count, GFP_KERNEL); if (!*bps) { ret = -ENOMEM; goto out; } for (; i < data->count; i++) { (*bps)[i] = (struct ras_badpage){ .bp = data->bps[i].retired_page, .size = AMDGPU_GPU_PAGE_SIZE, .flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED, }; status = amdgpu_vram_mgr_query_page_status( ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM), data->bps[i].retired_page); if (status == -EBUSY) (*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING; else if (status == -ENOENT) (*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT; } *count = data->count; out: mutex_unlock(&con->recovery_lock); return ret; } static void amdgpu_ras_do_recovery(struct work_struct *work) { struct amdgpu_ras *ras = container_of(work, struct amdgpu_ras, recovery_work); struct amdgpu_device *remote_adev = NULL; struct amdgpu_device *adev = ras->adev; struct list_head device_list, *device_list_handle = NULL; if (!ras->disable_ras_err_cnt_harvest) { struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev); /* Build list of devices to query RAS related errors */ if (hive && adev->gmc.xgmi.num_physical_nodes > 1) { device_list_handle = &hive->device_list; } else { INIT_LIST_HEAD(&device_list); list_add_tail(&adev->gmc.xgmi.head, &device_list); device_list_handle = &device_list; } list_for_each_entry(remote_adev, device_list_handle, gmc.xgmi.head) { amdgpu_ras_query_err_status(remote_adev); amdgpu_ras_log_on_err_counter(remote_adev); } amdgpu_put_xgmi_hive(hive); } if (amdgpu_device_should_recover_gpu(ras->adev)) amdgpu_device_gpu_recover(ras->adev, NULL); atomic_set(&ras->in_recovery, 0); } /* alloc/realloc bps array */ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev, struct ras_err_handler_data *data, int pages) { unsigned int old_space = data->count + data->space_left; unsigned int new_space = old_space + pages; unsigned int align_space = ALIGN(new_space, 512); void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL); if (!bps) { kfree(bps); return -ENOMEM; } if (data->bps) { memcpy(bps, data->bps, data->count * sizeof(*data->bps)); kfree(data->bps); } data->bps = bps; data->space_left += align_space - old_space; return 0; } /* it deal with vram only. */ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev, struct eeprom_table_record *bps, int pages) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_err_handler_data *data; int ret = 0; uint32_t i; if (!con || !con->eh_data || !bps || pages <= 0) return 0; mutex_lock(&con->recovery_lock); data = con->eh_data; if (!data) goto out; for (i = 0; i < pages; i++) { if (amdgpu_ras_check_bad_page_unlock(con, bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT)) continue; if (!data->space_left && amdgpu_ras_realloc_eh_data_space(adev, data, 256)) { ret = -ENOMEM; goto out; } amdgpu_vram_mgr_reserve_range( ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM), bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT, AMDGPU_GPU_PAGE_SIZE); memcpy(&data->bps[data->count], &bps[i], sizeof(*data->bps)); data->count++; data->space_left--; } out: mutex_unlock(&con->recovery_lock); return ret; } /* * write error record array to eeprom, the function should be * protected by recovery_lock */ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_err_handler_data *data; struct amdgpu_ras_eeprom_control *control; int save_count; if (!con || !con->eh_data) return 0; control = &con->eeprom_control; data = con->eh_data; save_count = data->count - control->num_recs; /* only new entries are saved */ if (save_count > 0) { if (amdgpu_ras_eeprom_process_recods(control, &data->bps[control->num_recs], true, save_count)) { dev_err(adev->dev, "Failed to save EEPROM table data!"); return -EIO; } dev_info(adev->dev, "Saved %d pages to EEPROM table.\n", save_count); } return 0; } /* * read error record array in eeprom and reserve enough space for * storing new bad pages */ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev) { struct amdgpu_ras_eeprom_control *control = &adev->psp.ras.ras->eeprom_control; struct eeprom_table_record *bps = NULL; int ret = 0; /* no bad page record, skip eeprom access */ if (!control->num_recs || (amdgpu_bad_page_threshold == 0)) return ret; bps = kcalloc(control->num_recs, sizeof(*bps), GFP_KERNEL); if (!bps) return -ENOMEM; if (amdgpu_ras_eeprom_process_recods(control, bps, false, control->num_recs)) { dev_err(adev->dev, "Failed to load EEPROM table records!"); ret = -EIO; goto out; } ret = amdgpu_ras_add_bad_pages(adev, bps, control->num_recs); out: kfree(bps); return ret; } static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con, uint64_t addr) { struct ras_err_handler_data *data = con->eh_data; int i; addr >>= AMDGPU_GPU_PAGE_SHIFT; for (i = 0; i < data->count; i++) if (addr == data->bps[i].retired_page) return true; return false; } /* * check if an address belongs to bad page * * Note: this check is only for umc block */ static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev, uint64_t addr) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); bool ret = false; if (!con || !con->eh_data) return ret; mutex_lock(&con->recovery_lock); ret = amdgpu_ras_check_bad_page_unlock(con, addr); mutex_unlock(&con->recovery_lock); return ret; } static void amdgpu_ras_validate_threshold(struct amdgpu_device *adev, uint32_t max_length) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int tmp_threshold = amdgpu_bad_page_threshold; u64 val; /* * Justification of value bad_page_cnt_threshold in ras structure * * Generally, -1 <= amdgpu_bad_page_threshold <= max record length * in eeprom, and introduce two scenarios accordingly. * * Bad page retirement enablement: * - If amdgpu_bad_page_threshold = -1, * bad_page_cnt_threshold = typical value by formula. * * - When the value from user is 0 < amdgpu_bad_page_threshold < * max record length in eeprom, use it directly. * * Bad page retirement disablement: * - If amdgpu_bad_page_threshold = 0, bad page retirement * functionality is disabled, and bad_page_cnt_threshold will * take no effect. */ if (tmp_threshold < -1) tmp_threshold = -1; else if (tmp_threshold > max_length) tmp_threshold = max_length; if (tmp_threshold == -1) { val = adev->gmc.mc_vram_size; do_div(val, RAS_BAD_PAGE_RATE); con->bad_page_cnt_threshold = min(lower_32_bits(val), max_length); } else { con->bad_page_cnt_threshold = tmp_threshold; } } int amdgpu_ras_recovery_init(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_err_handler_data **data; uint32_t max_eeprom_records_len = 0; bool exc_err_limit = false; int ret; if (adev->ras_enabled && con) data = &con->eh_data; else return 0; *data = kmalloc(sizeof(**data), GFP_KERNEL | __GFP_ZERO); if (!*data) { ret = -ENOMEM; goto out; } mutex_init(&con->recovery_lock); INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery); atomic_set(&con->in_recovery, 0); con->adev = adev; max_eeprom_records_len = amdgpu_ras_eeprom_get_record_max_length(); amdgpu_ras_validate_threshold(adev, max_eeprom_records_len); /* Todo: During test the SMU might fail to read the eeprom through I2C * when the GPU is pending on XGMI reset during probe time * (Mostly after second bus reset), skip it now */ if (adev->gmc.xgmi.pending_reset) return 0; ret = amdgpu_ras_eeprom_init(&con->eeprom_control, &exc_err_limit); /* * This calling fails when exc_err_limit is true or * ret != 0. */ if (exc_err_limit || ret) goto free; if (con->eeprom_control.num_recs) { ret = amdgpu_ras_load_bad_pages(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.num_recs); } return 0; free: kfree((*data)->bps); kfree(*data); con->eh_data = NULL; out: dev_warn(adev->dev, "Failed to initialize ras recovery!\n"); /* * Except error threshold exceeding case, other failure cases in this * function would not fail amdgpu driver init. */ if (!exc_err_limit) ret = 0; else ret = -EINVAL; return ret; } static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_err_handler_data *data = con->eh_data; /* recovery_init failed to init it, fini is useless */ if (!data) return 0; cancel_work_sync(&con->recovery_work); mutex_lock(&con->recovery_lock); con->eh_data = NULL; kfree(data->bps); kfree(data); mutex_unlock(&con->recovery_lock); return 0; } /* 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 || adev->asic_type == CHIP_VEGA20 || adev->asic_type == CHIP_ARCTURUS || adev->asic_type == CHIP_ALDEBARAN || adev->asic_type == CHIP_SIENNA_CICHLID; } /* * this is workaround for vega20 workstation sku, * force enable gfx ras, ignore vbios gfx ras flag * due to GC EDC can not write */ static void amdgpu_ras_get_quirks(struct amdgpu_device *adev) { struct atom_context *ctx = adev->mode_info.atom_context; if (!ctx) return; if (strnstr(ctx->vbios_version, "D16406", sizeof(ctx->vbios_version)) || strnstr(ctx->vbios_version, "D36002", sizeof(ctx->vbios_version))) adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX); } /* * check hardware's ras ability which will be saved in hw_supported. * if hardware does not support ras, we can skip some ras initializtion and * forbid some ras operations from IP. * if software itself, say boot parameter, limit the ras ability. We still * need allow IP do some limited operations, like disable. In such case, * we have to initialize ras as normal. but need check if operation is * allowed or not in each function. */ 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 || !amdgpu_ras_asic_supported(adev)) return; if (!adev->gmc.xgmi.connected_to_cpu) { if (amdgpu_atomfirmware_mem_ecc_supported(adev)) { dev_info(adev->dev, "MEM ECC is active.\n"); adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC | 1 << AMDGPU_RAS_BLOCK__DF); } else { dev_info(adev->dev, "MEM ECC is not presented.\n"); } 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); } else { dev_info(adev->dev, "SRAM ECC is not presented.\n"); } } else { /* driver only manages a few IP blocks RAS feature * when GPU is connected cpu through XGMI */ adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX | 1 << AMDGPU_RAS_BLOCK__SDMA | 1 << AMDGPU_RAS_BLOCK__MMHUB); } amdgpu_ras_get_quirks(adev); /* hw_supported needs to be aligned with RAS block mask. */ adev->ras_hw_enabled &= AMDGPU_RAS_BLOCK_MASK; adev->ras_enabled = amdgpu_ras_enable == 0 ? 0 : adev->ras_hw_enabled & amdgpu_ras_mask; } static void amdgpu_ras_counte_dw(struct work_struct *work) { struct amdgpu_ras *con = container_of(work, struct amdgpu_ras, ras_counte_delay_work.work); struct amdgpu_device *adev = con->adev; struct drm_device *dev = adev_to_drm(adev); unsigned long ce_count, ue_count; int res; res = pm_runtime_get_sync(dev->dev); if (res < 0) goto Out; /* Cache new values. */ if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count) == 0) { atomic_set(&con->ras_ce_count, ce_count); atomic_set(&con->ras_ue_count, ue_count); } pm_runtime_mark_last_busy(dev->dev); Out: pm_runtime_put_autosuspend(dev->dev); } int amdgpu_ras_init(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int r; if (con) return 0; con = kmalloc(sizeof(struct amdgpu_ras) + sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT, GFP_KERNEL|__GFP_ZERO); if (!con) return -ENOMEM; con->adev = adev; INIT_DELAYED_WORK(&con->ras_counte_delay_work, amdgpu_ras_counte_dw); atomic_set(&con->ras_ce_count, 0); atomic_set(&con->ras_ue_count, 0); con->objs = (struct ras_manager *)(con + 1); amdgpu_ras_set_context(adev, con); amdgpu_ras_check_supported(adev); if (!adev->ras_enabled || adev->asic_type == CHIP_VEGA10) { /* set gfx block ras context feature for VEGA20 Gaming * send ras disable cmd to ras ta during ras late init. */ if (!adev->ras_enabled && adev->asic_type == CHIP_VEGA20) { con->features |= BIT(AMDGPU_RAS_BLOCK__GFX); return 0; } r = 0; goto release_con; } con->features = 0; INIT_LIST_HEAD(&con->head); /* Might need get this flag from vbios. */ con->flags = RAS_DEFAULT_FLAGS; /* initialize nbio ras function ahead of any other * ras functions so hardware fatal error interrupt * can be enabled as early as possible */ switch (adev->asic_type) { 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; 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 (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 (r) goto release_con; } if (amdgpu_ras_fs_init(adev)) { r = -EINVAL; goto release_con; } dev_info(adev->dev, "RAS INFO: ras initialized successfully, " "hardware ability[%x] ras_mask[%x]\n", adev->ras_hw_enabled, adev->ras_enabled); return 0; release_con: amdgpu_ras_set_context(adev, NULL); kfree(con); return r; } int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev) { if (adev->gmc.xgmi.connected_to_cpu) return 1; return 0; } static int amdgpu_persistent_edc_harvesting(struct amdgpu_device *adev, struct ras_common_if *ras_block) { struct ras_query_if info = { .head = *ras_block, }; if (!amdgpu_persistent_edc_harvesting_supported(adev)) return 0; if (amdgpu_ras_query_error_status(adev, &info) != 0) DRM_WARN("RAS init harvest failure"); if (amdgpu_ras_reset_error_status(adev, ras_block->block) != 0) DRM_WARN("RAS init harvest reset failure"); return 0; } /* 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) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); unsigned long ue_count, ce_count; int r; /* disable RAS feature per IP block if it is not supported */ if (!amdgpu_ras_is_supported(adev, ras_block->block)) { amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0); return 0; } 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)) { /* in resume phase, if fail to enable ras, * clean up all ras fs nodes, and disable ras */ goto cleanup; } else return r; } /* check for errors on warm reset edc persisant supported ASIC */ amdgpu_persistent_edc_harvesting(adev, ras_block); /* in resume phase, no need to create ras fs node */ if (adev->in_suspend || amdgpu_in_reset(adev)) return 0; if (ih_info->cb) { r = amdgpu_ras_interrupt_add_handler(adev, ih_info); if (r) goto interrupt; } r = amdgpu_ras_sysfs_create(adev, fs_info); if (r) goto sysfs; /* Those are the cached values at init. */ if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count) == 0) { atomic_set(&con->ras_ce_count, ce_count); atomic_set(&con->ras_ue_count, ue_count); } return 0; cleanup: amdgpu_ras_sysfs_remove(adev, ras_block); sysfs: if (ih_info->cb) amdgpu_ras_interrupt_remove_handler(adev, ih_info); interrupt: amdgpu_ras_feature_enable(adev, ras_block, 0); return r; } /* 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); amdgpu_ras_feature_enable(adev, ras_block, 0); } /* do some init work after IP late init as dependence. * and it runs in resume/gpu reset/booting up cases. */ void amdgpu_ras_resume(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); struct ras_manager *obj, *tmp; if (!adev->ras_enabled || !con) { /* clean ras context for VEGA20 Gaming after send ras disable cmd */ amdgpu_release_ras_context(adev); return; } if (con->flags & AMDGPU_RAS_FLAG_INIT_BY_VBIOS) { /* Set up all other IPs which are not implemented. There is a * tricky thing that IP's actual ras error type should be * MULTI_UNCORRECTABLE, but as driver does not handle it, so * ERROR_NONE make sense anyway. */ amdgpu_ras_enable_all_features(adev, 1); /* We enable ras on all hw_supported block, but as boot * parameter might disable some of them and one or more IP has * not implemented yet. So we disable them on behalf. */ list_for_each_entry_safe(obj, tmp, &con->head, node) { if (!amdgpu_ras_is_supported(adev, obj->head.block)) { amdgpu_ras_feature_enable(adev, &obj->head, 0); /* there should be no any reference. */ WARN_ON(alive_obj(obj)); } } } 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) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) return; amdgpu_ras_disable_all_features(adev, 0); /* Make sure all ras objects are disabled. */ if (con->features) amdgpu_ras_disable_all_features(adev, 1); } /* do some fini work before IP fini as dependence */ int amdgpu_ras_pre_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) return 0; /* Need disable ras on all IPs here before ip [hw/sw]fini */ amdgpu_ras_disable_all_features(adev, 0); amdgpu_ras_recovery_fini(adev); return 0; } int amdgpu_ras_fini(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!adev->ras_enabled || !con) return 0; amdgpu_ras_fs_fini(adev); amdgpu_ras_interrupt_remove_all(adev); WARN(con->features, "Feature mask is not cleared"); if (con->features) amdgpu_ras_disable_all_features(adev, 1); cancel_delayed_work_sync(&con->ras_counte_delay_work); amdgpu_ras_set_context(adev, NULL); kfree(con); return 0; } void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev) { amdgpu_ras_check_supported(adev); if (!adev->ras_hw_enabled) return; if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) { dev_info(adev->dev, "uncorrectable hardware error" "(ERREVENT_ATHUB_INTERRUPT) detected!\n"); amdgpu_ras_reset_gpu(adev); } } bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) { if (adev->asic_type == CHIP_VEGA20 && adev->pm.fw_version <= 0x283400) { return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) && amdgpu_ras_intr_triggered(); } return false; } void amdgpu_release_ras_context(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); if (!con) return; if (!adev->ras_enabled && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) { con->features &= ~BIT(AMDGPU_RAS_BLOCK__GFX); amdgpu_ras_set_context(adev, NULL); kfree(con); } }