Merge tag 'amd-drm-next-6.18-2025-08-29' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.18-2025-08-29: amdgpu: - Replay fixes - RAS updates - VCN SRAM load fixes - EDID read fixes - eDP ALPM support - AUX fixes - Documenation updates - Rework how PTE flags are generated - DCE6 fixes - VCN devcoredump cleanup - MMHUB client id fixes - SR-IOV fixes - VRR fixes - VCN 5.0.1 RAS support - Backlight fixes - UserQ fixes - Misc code cleanups - SMU 13.0.12 updates - Expanded PCIe DPC support - Expanded VCN reset support - SMU 13.0.x Updates - VPE per queue reset support - Cusor rotation fix - DSC fixes - GC 12 MES TLB invalidation update - Cursor fixes - Non-DC TMDS clock validation fix amdkfd: - debugfs fixes - Misc code cleanups - Page migration fixes - Partition fixes - SVM fixes radeon: - Misc code cleanups Signed-off-by: Dave Airlie <airlied@redhat.com> From: Alex Deucher <alexander.deucher@amd.com> Link: https://lore.kernel.org/r/20250829190848.1921648-1-alexander.deucher@amd.com
2026-04-18 14:53:58 -04:00 · 2025-09-02 09:35:11 +10:00
parent ddcc2bb288 585b2f685c
commit 14579a6f18
279 changed files with 4889 additions and 2438 deletions
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -122,12 +122,15 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 /* typical ECC bad page rate is 1 bad page per 100MB VRAM */
 #define RAS_BAD_PAGE_COVER              (100 * 1024 * 1024ULL)

-#define MAX_UMC_POISON_POLLING_TIME_ASYNC  300  //ms
+#define MAX_UMC_POISON_POLLING_TIME_ASYNC  10

 #define AMDGPU_RAS_RETIRE_PAGE_INTERVAL 100  //ms

 #define MAX_FLUSH_RETIRE_DWORK_TIMES  100

+#define BYPASS_ALLOCATED_ADDRESS        0x0
+#define BYPASS_INITIALIZATION_ADDRESS   0x1
+
 enum amdgpu_ras_retire_page_reservation {
 	AMDGPU_RAS_RETIRE_PAGE_RESERVED,
 	AMDGPU_RAS_RETIRE_PAGE_PENDING,
@@ -136,10 +139,14 @@ enum amdgpu_ras_retire_page_reservation {

 atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);

-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr);
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr);
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev);
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev);
+
 #ifdef CONFIG_X86_MCE_AMD
 static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev);
 struct mce_notifier_adev_list {
@@ -169,18 +176,16 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre
 	struct eeprom_table_record err_rec;
 	int ret;

-	if ((address >= adev->gmc.mc_vram_size) ||
-	    (address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
+	ret = amdgpu_ras_check_bad_page(adev, address);
+	if (ret == -EINVAL) {
 		dev_warn(adev->dev,
-		         "RAS WARN: input address 0x%llx is invalid.\n",
-		         address);
+			"RAS WARN: input address 0x%llx is invalid.\n",
+			address);
 		return -EINVAL;
-	}
-
-	if (amdgpu_ras_check_bad_page(adev, address)) {
+	} else if (ret == 1) {
 		dev_warn(adev->dev,
-			 "RAS WARN: 0x%llx has already been marked as bad page!\n",
-			 address);
+			"RAS WARN: 0x%llx has already been marked as bad page!\n",
+			address);
 		return 0;
 	}

@@ -207,6 +212,49 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre
 	return 0;
 }

+static int amdgpu_check_address_validity(struct amdgpu_device *adev,
+			uint64_t address, uint64_t flags)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_block_info blk_info;
+	uint64_t page_pfns[32] = {0};
+	int i, ret, count;
+
+	if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0))
+		return 0;
+
+	if ((address >= adev->gmc.mc_vram_size) ||
+	    (address >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EFAULT;
+
+	count = amdgpu_umc_lookup_bad_pages_in_a_row(adev,
+				address, page_pfns, ARRAY_SIZE(page_pfns));
+	if (count <= 0)
+		return -EPERM;
+
+	for (i = 0; i < count; i++) {
+		memset(&blk_info, 0, sizeof(blk_info));
+		ret = amdgpu_vram_mgr_query_address_block_info(&adev->mman.vram_mgr,
+					page_pfns[i] << AMDGPU_GPU_PAGE_SHIFT, &blk_info);
+		if (!ret) {
+			/* The input address that needs to be checked is allocated by
+			 * current calling process, so it is necessary to exclude
+			 * the calling process.
+			 */
+			if ((flags == BYPASS_ALLOCATED_ADDRESS) &&
+			    ((blk_info.task.pid != task_pid_nr(current)) ||
+				strncmp(blk_info.task.comm, current->comm, TASK_COMM_LEN)))
+				return -EACCES;
+			else if ((flags == BYPASS_INITIALIZATION_ADDRESS) &&
+				(blk_info.task.pid == con->init_task_pid) &&
+				!strncmp(blk_info.task.comm, con->init_task_comm, TASK_COMM_LEN))
+				return -EACCES;
+		}
+	}
+
+	return 0;
+}
+
 static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
 					size_t size, loff_t *pos)
 {
@@ -297,6 +345,8 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 		op = 2;
 	else if (strstr(str, "retire_page") != NULL)
 		op = 3;
+	else if (strstr(str, "check_address") != NULL)
+		op = 4;
 	else if (str[0] && str[1] && str[2] && str[3])
 		/* ascii string, but commands are not matched. */
 		return -EINVAL;
@@ -310,6 +360,15 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 			data->op = op;
 			data->inject.address = address;

+			return 0;
+		} else if (op == 4) {
+			if (sscanf(str, "%*s 0x%llx 0x%llx", &address, &value) != 2 &&
+			    sscanf(str, "%*s %llu %llu", &address, &value) != 2)
+				return -EINVAL;
+
+			data->op = op;
+			data->inject.address = address;
+			data->inject.value = value;
 			return 0;
 		}

@@ -500,6 +559,9 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 			return size;
 		else
 			return ret;
+	} else if (data.op == 4) {
+		ret = amdgpu_check_address_validity(adev, data.inject.address, data.inject.value);
+		return ret ? ret : size;
 	}

 	if (!amdgpu_ras_is_supported(adev, data.head.block))
@@ -513,22 +575,16 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 		ret = amdgpu_ras_feature_enable(adev, &data.head, 1);
 		break;
 	case 2:
-		if ((data.inject.address >= adev->gmc.mc_vram_size &&
-		    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: inject: 0x%llx has "
-				 "already been marked as bad!\n",
-				 data.inject.address);
+		if (data.head.block == AMDGPU_RAS_BLOCK__UMC)
+			ret = amdgpu_ras_check_bad_page(adev, data.inject.address);
+		if (ret == -EINVAL) {
+			dev_warn(adev->dev, "RAS WARN: input address 0x%llx is invalid.",
+					data.inject.address);
+			break;
+		} else if (ret == 1) {
+			dev_warn(adev->dev, "RAS WARN: inject: 0x%llx has already been marked as bad!\n",
+					data.inject.address);
 			break;
 		}

@@ -2566,18 +2622,26 @@ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
 		goto out;
 	}

-	*bps = kmalloc(sizeof(struct ras_badpage) * data->count, GFP_KERNEL);
+	*bps = kmalloc_array(data->count, sizeof(struct ras_badpage), GFP_KERNEL);
 	if (!*bps) {
 		ret = -ENOMEM;
 		goto out;
 	}

 	for (; i < data->count; i++) {
+		if (!data->bps[i].ts)
+			continue;
+
 		(*bps)[i] = (struct ras_badpage){
 			.bp = data->bps[i].retired_page,
 			.size = AMDGPU_GPU_PAGE_SIZE,
 			.flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED,
 		};
+
+		if (amdgpu_ras_check_critical_address(adev,
+			data->bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			continue;
+
 		status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr,
 				data->bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT);
 		if (status == -EBUSY)
@@ -2586,7 +2650,7 @@ static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
 			(*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT;
 	}

-	*count = data->count;
+	*count = con->bad_page_num;
 out:
 	mutex_unlock(&con->recovery_lock);
 	return ret;
@@ -2722,7 +2786,7 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	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);
+	void *bps = kmalloc_array(align_space, sizeof(*data->bps), GFP_KERNEL);

 	if (!bps) {
 		return -ENOMEM;
@@ -2814,8 +2878,11 @@ static int __amdgpu_ras_restore_bad_pages(struct amdgpu_device *adev,

 	for (j = 0; j < count; j++) {
 		if (amdgpu_ras_check_bad_page_unlock(con,
-			bps[j].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			bps[j].retired_page << AMDGPU_GPU_PAGE_SHIFT)) {
+			data->count++;
+			data->space_left--;
 			continue;
+		}

 		if (!data->space_left &&
 		    amdgpu_ras_realloc_eh_data_space(adev, data, 256)) {
@@ -2828,6 +2895,7 @@ static int __amdgpu_ras_restore_bad_pages(struct amdgpu_device *adev,
 				sizeof(struct eeprom_table_record));
 		data->count++;
 		data->space_left--;
+		con->bad_page_num++;
 	}

 	return 0;
@@ -2974,7 +3042,7 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 						ret = __amdgpu_ras_convert_rec_array_from_rom(adev,
 										&bps[i], &err_data, nps);
 						if (ret)
-							control->ras_num_bad_pages -= adev->umc.retire_unit;
+							con->bad_page_num -= adev->umc.retire_unit;
 						i += (adev->umc.retire_unit - 1);
 					} else {
 						break;
@@ -2988,8 +3056,10 @@ int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
 			ret = __amdgpu_ras_convert_rec_from_rom(adev,
 				&bps[i], &err_data, nps);
 			if (ret)
-				control->ras_num_bad_pages -= adev->umc.retire_unit;
+				con->bad_page_num -= adev->umc.retire_unit;
 		}
+
+		con->eh_data->count_saved = con->eh_data->count;
 	} else {
 		ret = __amdgpu_ras_restore_bad_pages(adev, bps, pages);
 	}
@@ -3012,7 +3082,7 @@ 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, unit_num, bad_page_num, i;
+	int save_count, unit_num, i;

 	if (!con || !con->eh_data) {
 		if (new_cnt)
@@ -3033,27 +3103,26 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 	mutex_lock(&con->recovery_lock);
 	control = &con->eeprom_control;
 	data = con->eh_data;
-	bad_page_num = control->ras_num_bad_pages;
-	save_count = data->count - bad_page_num;
+	unit_num = data->count / adev->umc.retire_unit - control->ras_num_recs;
+	save_count = con->bad_page_num - control->ras_num_bad_pages;
 	mutex_unlock(&con->recovery_lock);

-	unit_num = save_count / adev->umc.retire_unit;
 	if (new_cnt)
 		*new_cnt = unit_num;

 	/* only new entries are saved */
-	if (save_count > 0) {
+	if (unit_num > 0) {
 		/*old asics only save pa to eeprom like before*/
 		if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12) {
 			if (amdgpu_ras_eeprom_append(control,
-					&data->bps[bad_page_num], save_count)) {
+					&data->bps[data->count_saved], unit_num)) {
 				dev_err(adev->dev, "Failed to save EEPROM table data!");
 				return -EIO;
 			}
 		} else {
 			for (i = 0; i < unit_num; i++) {
 				if (amdgpu_ras_eeprom_append(control,
-						&data->bps[bad_page_num +
+						&data->bps[data->count_saved +
 						i * adev->umc.retire_unit], 1)) {
 					dev_err(adev->dev, "Failed to save EEPROM table data!");
 					return -EIO;
@@ -3062,6 +3131,7 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 		}

 		dev_info(adev->dev, "Saved %d pages to EEPROM table.\n", save_count);
+		data->count_saved = data->count;
 	}

 	return 0;
@@ -3116,17 +3186,17 @@ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev)
 			}
 		}

+		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs, true);
+		if (ret)
+			goto out;
+
 		ret = amdgpu_ras_eeprom_check(control);
 		if (ret)
 			goto out;

 		/* HW not usable */
-		if (amdgpu_ras_is_rma(adev)) {
+		if (amdgpu_ras_is_rma(adev))
 			ret = -EHWPOISON;
-			goto out;
-		}
-
-		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs, true);
 	}

 out:
@@ -3134,18 +3204,24 @@ out:
 	return ret;
 }

-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr)
 {
 	struct ras_err_handler_data *data = con->eh_data;
+	struct amdgpu_device *adev = con->adev;
 	int i;

+	if ((addr >= adev->gmc.mc_vram_size &&
+	    adev->gmc.mc_vram_size) ||
+	    (addr >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EINVAL;
+
 	addr >>= AMDGPU_GPU_PAGE_SHIFT;
 	for (i = 0; i < data->count; i++)
 		if (addr == data->bps[i].retired_page)
-			return true;
+			return 1;

-	return false;
+	return 0;
 }

 /*
@@ -3153,11 +3229,11 @@ static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 *
 * Note: this check is only for umc block
 */
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	bool ret = false;
+	int ret = 0;

 	if (!con || !con->eh_data)
 		return ret;
@@ -3241,7 +3317,7 @@ static void amdgpu_ras_ecc_log_init(struct ras_ecc_log_info *ecc_log)

 	INIT_RADIX_TREE(&ecc_log->de_page_tree, GFP_KERNEL);
 	ecc_log->de_queried_count = 0;
-	ecc_log->prev_de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
 }

 static void amdgpu_ras_ecc_log_fini(struct ras_ecc_log_info *ecc_log)
@@ -3261,7 +3337,7 @@ static void amdgpu_ras_ecc_log_fini(struct ras_ecc_log_info *ecc_log)

 	mutex_destroy(&ecc_log->lock);
 	ecc_log->de_queried_count = 0;
-	ecc_log->prev_de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
 }

 static bool amdgpu_ras_schedule_retirement_dwork(struct amdgpu_ras *con,
@@ -3287,7 +3363,6 @@ static void amdgpu_ras_do_page_retirement(struct work_struct *work)
 					      page_retirement_dwork.work);
 	struct amdgpu_device *adev = con->adev;
 	struct ras_err_data err_data;
-	unsigned long err_cnt;

 	/* If gpu reset is ongoing, delay retiring the bad pages */
 	if (amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) {
@@ -3299,13 +3374,9 @@ static void amdgpu_ras_do_page_retirement(struct work_struct *work)
 	amdgpu_ras_error_data_init(&err_data);

 	amdgpu_umc_handle_bad_pages(adev, &err_data);
-	err_cnt = err_data.err_addr_cnt;

 	amdgpu_ras_error_data_fini(&err_data);

-	if (err_cnt && amdgpu_ras_is_rma(adev))
-		amdgpu_ras_reset_gpu(adev);
-
 	amdgpu_ras_schedule_retirement_dwork(con,
 			AMDGPU_RAS_RETIRE_PAGE_INTERVAL);
 }
@@ -3316,49 +3387,39 @@ static int amdgpu_ras_poison_creation_handler(struct amdgpu_device *adev,
 	int ret = 0;
 	struct ras_ecc_log_info *ecc_log;
 	struct ras_query_if info;
-	uint32_t timeout = 0;
+	u32 timeout = MAX_UMC_POISON_POLLING_TIME_ASYNC;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
-	uint64_t de_queried_count;
-	uint32_t new_detect_count, total_detect_count;
-	uint32_t need_query_count = poison_creation_count;
+	u64 de_queried_count;
+	u64 consumption_q_count;
 	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CREATION;

 	memset(&info, 0, sizeof(info));
 	info.head.block = AMDGPU_RAS_BLOCK__UMC;

 	ecc_log = &ras->umc_ecc_log;
-	total_detect_count = 0;
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+
 	do {
 		ret = amdgpu_ras_query_error_status_with_event(adev, &info, type);
 		if (ret)
 			return ret;

 		de_queried_count = ecc_log->de_queried_count;
-		if (de_queried_count > ecc_log->prev_de_queried_count) {
-			new_detect_count = de_queried_count - ecc_log->prev_de_queried_count;
-			ecc_log->prev_de_queried_count = de_queried_count;
-			timeout = 0;
-		} else {
-			new_detect_count = 0;
-		}
+		consumption_q_count = ecc_log->consumption_q_count;

-		if (new_detect_count) {
-			total_detect_count += new_detect_count;
-		} else {
-			if (!timeout && need_query_count)
-				timeout = MAX_UMC_POISON_POLLING_TIME_ASYNC;
+		if (de_queried_count && consumption_q_count)
+			break;

-			if (timeout) {
-				if (!--timeout)
-					break;
-				msleep(1);
-			}
-		}
-	} while (total_detect_count < need_query_count);
+		msleep(100);
+	} while (--timeout);

-	if (total_detect_count)
+	if (de_queried_count)
 		schedule_delayed_work(&ras->page_retirement_dwork, 0);

+	if (amdgpu_ras_is_rma(adev) && atomic_cmpxchg(&ras->rma_in_recovery, 0, 1) == 0)
+		amdgpu_ras_reset_gpu(adev);
+
 	return 0;
 }

@@ -3394,6 +3455,12 @@ static int amdgpu_ras_poison_consumption_handler(struct amdgpu_device *adev,
 		reset_flags |= msg.reset;
 	}

+	/*
+	 * Try to ensure poison creation handler is completed first
+	 * to set rma if bad page exceed threshold.
+	 */
+	flush_delayed_work(&con->page_retirement_dwork);
+
 	/* for RMA, amdgpu_ras_poison_creation_handler will trigger gpu reset */
 	if (reset_flags && !amdgpu_ras_is_rma(adev)) {
 		if (reset_flags & AMDGPU_RAS_GPU_RESET_MODE1_RESET)
@@ -3403,8 +3470,6 @@ static int amdgpu_ras_poison_consumption_handler(struct amdgpu_device *adev,
 		else
 			reset = reset_flags;

-		flush_delayed_work(&con->page_retirement_dwork);
-
 		con->gpu_reset_flags |= reset;
 		amdgpu_ras_reset_gpu(adev);

@@ -3434,6 +3499,7 @@ static int amdgpu_ras_page_retirement_thread(void *param)
 		if (kthread_should_stop())
 			break;

+		mutex_lock(&con->poison_lock);
 		gpu_reset = 0;

 		do {
@@ -3446,7 +3512,8 @@ static int amdgpu_ras_page_retirement_thread(void *param)
 				atomic_sub(poison_creation_count, &con->poison_creation_count);
 				atomic_sub(poison_creation_count, &con->page_retirement_req_cnt);
 			}
-		} while (atomic_read(&con->poison_creation_count));
+		} while (atomic_read(&con->poison_creation_count) &&
+			!atomic_read(&con->poison_consumption_count));

 		if (ret != -EIO) {
 			msg_count = kfifo_len(&con->poison_fifo);
@@ -3463,6 +3530,7 @@ static int amdgpu_ras_page_retirement_thread(void *param)
 			/* gpu mode-1 reset is ongoing or just completed ras mode-1 reset */
 			/* Clear poison creation request */
 			atomic_set(&con->poison_creation_count, 0);
+			atomic_set(&con->poison_consumption_count, 0);

 			/* Clear poison fifo */
 			amdgpu_ras_clear_poison_fifo(adev);
@@ -3487,9 +3555,12 @@ static int amdgpu_ras_page_retirement_thread(void *param)
 				atomic_sub(msg_count, &con->page_retirement_req_cnt);
 			}

+			atomic_set(&con->poison_consumption_count, 0);
+
 			/* Wake up work to save bad pages to eeprom */
 			schedule_delayed_work(&con->page_retirement_dwork, 0);
 		}
+		mutex_unlock(&con->poison_lock);
 	}

 	return 0;
@@ -3570,8 +3641,10 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info)
 	}

 	mutex_init(&con->recovery_lock);
+	mutex_init(&con->poison_lock);
 	INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery);
 	atomic_set(&con->in_recovery, 0);
+	atomic_set(&con->rma_in_recovery, 0);
 	con->eeprom_control.bad_channel_bitmap = 0;

 	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count(&con->eeprom_control);
@@ -3589,6 +3662,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info)
 	init_waitqueue_head(&con->page_retirement_wq);
 	atomic_set(&con->page_retirement_req_cnt, 0);
 	atomic_set(&con->poison_creation_count, 0);
+	atomic_set(&con->poison_consumption_count, 0);
 	con->page_retirement_thread =
 		kthread_run(amdgpu_ras_page_retirement_thread, adev, "umc_page_retirement");
 	if (IS_ERR(con->page_retirement_thread)) {
@@ -3661,6 +3735,8 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 	kfree(data);
 	mutex_unlock(&con->recovery_lock);

+	amdgpu_ras_critical_region_init(adev);
+
 	return 0;
 }
 /* recovery end */
@@ -4087,6 +4163,12 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 			goto release_con;
 	}

+	con->init_task_pid = task_pid_nr(current);
+	get_task_comm(con->init_task_comm, current);
+
+	mutex_init(&con->critical_region_lock);
+	INIT_LIST_HEAD(&con->critical_region_head);
+
 	dev_info(adev->dev, "RAS INFO: ras initialized successfully, "
 		 "hardware ability[%x] ras_mask[%x]\n",
 		 adev->ras_hw_enabled, adev->ras_enabled);
@@ -4366,6 +4448,9 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	if (!adev->ras_enabled || !con)
 		return 0;

+	amdgpu_ras_critical_region_fini(adev);
+	mutex_destroy(&con->critical_region_lock);
+
 	list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) {
 		if (ras_node->ras_obj) {
 			obj = ras_node->ras_obj;
@@ -5274,6 +5359,9 @@ int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn)
 	uint64_t start = pfn << AMDGPU_GPU_PAGE_SHIFT;
 	int ret = 0;

+	if (amdgpu_ras_check_critical_address(adev, start))
+		return 0;
+
 	mutex_lock(&con->page_rsv_lock);
 	ret = amdgpu_vram_mgr_query_page_status(mgr, start);
 	if (ret == -ENOENT)
@@ -5310,3 +5398,80 @@ bool amdgpu_ras_is_rma(struct amdgpu_device *adev)

 	return con->is_rma;
 }
+
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev,
+			struct amdgpu_bo *bo)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_mgr_resource *vres;
+	struct ras_critical_region *region;
+	struct drm_buddy_block *block;
+	int ret = 0;
+
+	if (!bo || !bo->tbo.resource)
+		return -EINVAL;
+
+	vres = to_amdgpu_vram_mgr_resource(bo->tbo.resource);
+
+	mutex_lock(&con->critical_region_lock);
+
+	/* Check if the bo had been recorded */
+	list_for_each_entry(region, &con->critical_region_head, node)
+		if (region->bo == bo)
+			goto out;
+
+	/* Record new critical amdgpu bo */
+	list_for_each_entry(block, &vres->blocks, link) {
+		region = kzalloc(sizeof(*region), GFP_KERNEL);
+		if (!region) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		region->bo = bo;
+		region->start = amdgpu_vram_mgr_block_start(block);
+		region->size = amdgpu_vram_mgr_block_size(block);
+		list_add_tail(&region->node, &con->critical_region_head);
+	}
+
+out:
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev)
+{
+	amdgpu_ras_add_critical_region(adev, adev->mman.fw_reserved_memory);
+}
+
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region, *tmp;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry_safe(region, tmp, &con->critical_region_head, node) {
+		list_del(&region->node);
+		kfree(region);
+	}
+	mutex_unlock(&con->critical_region_lock);
+}
+
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region;
+	bool ret = false;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry(region, &con->critical_region_head, node) {
+		if ((region->start <= addr) &&
+		    (addr < (region->start + region->size))) {
+			ret = true;
+			break;
+		}
+	}
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}