1 files changed, 415 insertions, 267 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 8f47c14ecbc7..7e126dff004f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -66,6 +66,8 @@ const char *ras_block_string[] = {
 	"mp1",
 	"fuse",
 	"mca",
+	"vcn",
+	"jpeg",
 };
 
 const char *ras_mca_block_string[] = {
@@ -75,6 +77,13 @@ const char *ras_mca_block_string[] = {
 	"mca_iohc",
 };
 
+struct amdgpu_ras_block_list {
+	/* ras block link */
+	struct list_head node;
+
+	struct amdgpu_ras_block_object *ras_obj;
+};
+
 const char *get_ras_block_str(struct ras_common_if *ras_block)
 {
 	if (!ras_block)
@@ -89,6 +98,9 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 	return ras_block_string[ras_block->block];
 }
 
+#define ras_block_str(_BLOCK_) \
+	(((_BLOCK_) < ARRAY_SIZE(ras_block_string)) ? ras_block_string[_BLOCK_] : "Out Of Range")
+
 #define ras_err_str(i) (ras_error_string[ffs(i)])
 
 #define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS)
@@ -155,14 +167,9 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre
 	}
 
 	memset(&err_rec, 0x0, sizeof(struct eeprom_table_record));
-
-	err_rec.address = address;
-	err_rec.retired_page = address >> AMDGPU_GPU_PAGE_SHIFT;
-	err_rec.ts = (uint64_t)ktime_get_real_seconds();
-	err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE;
-
 	err_data.err_addr = &err_rec;
-	err_data.err_addr_cnt = 1;
+	amdgpu_umc_fill_error_record(&err_data, address,
+			(address >> AMDGPU_GPU_PAGE_SHIFT), 0, 0);
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
@@ -452,7 +459,7 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 	}
 
 	if (ret)
-		return -EINVAL;
+		return ret;
 
 	return size;
 }
@@ -866,30 +873,47 @@ static int amdgpu_ras_enable_all_features(struct amdgpu_device *adev,
 }
 /* feature ctl end */
 
+static int amdgpu_ras_block_match_default(struct amdgpu_ras_block_object *block_obj,
+		enum amdgpu_ras_block block)
+{
+	if (!block_obj)
+		return -EINVAL;
+
+	if (block_obj->ras_comm.block == block)
+		return 0;
+
+	return -EINVAL;
+}
 
-static void amdgpu_ras_mca_query_error_status(struct amdgpu_device *adev,
-					      struct ras_common_if *ras_block,
-					      struct ras_err_data  *err_data)
+static struct amdgpu_ras_block_object *amdgpu_ras_get_ras_block(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block, uint32_t sub_block_index)
 {
-	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;
+	struct amdgpu_ras_block_list *node, *tmp;
+	struct amdgpu_ras_block_object *obj;
+
+	if (block >= AMDGPU_RAS_BLOCK__LAST)
+		return NULL;
+
+	if (!amdgpu_ras_is_supported(adev, block))
+		return NULL;
+
+	list_for_each_entry_safe(node, tmp, &adev->ras_list, node) {
+		if (!node->ras_obj) {
+			dev_warn(adev->dev, "Warning: abnormal ras list node.\n");
+			continue;
+		}
+
+		obj = node->ras_obj;
+		if (obj->ras_block_match) {
+			if (obj->ras_block_match(obj, block, sub_block_index) == 0)
+				return obj;
+		} else {
+			if (amdgpu_ras_block_match_default(obj, block) == 0)
+				return obj;
+		}
 	}
+
+	return NULL;
 }
 
 static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_data *err_data)
@@ -901,26 +925,26 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d
 	 * choosing right query method according to
 	 * whether smu support query error information
 	 */
-	ret = smu_get_ecc_info(&adev->smu, (void *)&(ras->umc_ecc));
+	ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(ras->umc_ecc));
 	if (ret == -EOPNOTSUPP) {
-		if (adev->umc.ras_funcs &&
-			adev->umc.ras_funcs->query_ras_error_count)
-			adev->umc.ras_funcs->query_ras_error_count(adev, err_data);
+		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+			adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
+			adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data);
 
 		/* umc query_ras_error_address is also responsible for clearing
 		 * error status
 		 */
-		if (adev->umc.ras_funcs &&
-		    adev->umc.ras_funcs->query_ras_error_address)
-			adev->umc.ras_funcs->query_ras_error_address(adev, err_data);
+		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+		    adev->umc.ras->ras_block.hw_ops->query_ras_error_address)
+			adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, err_data);
 	} else if (!ret) {
-		if (adev->umc.ras_funcs &&
-			adev->umc.ras_funcs->ecc_info_query_ras_error_count)
-			adev->umc.ras_funcs->ecc_info_query_ras_error_count(adev, err_data);
+		if (adev->umc.ras &&
+			adev->umc.ras->ecc_info_query_ras_error_count)
+			adev->umc.ras->ecc_info_query_ras_error_count(adev, err_data);
 
-		if (adev->umc.ras_funcs &&
-			adev->umc.ras_funcs->ecc_info_query_ras_error_address)
-			adev->umc.ras_funcs->ecc_info_query_ras_error_address(adev, err_data);
+		if (adev->umc.ras &&
+			adev->umc.ras->ecc_info_query_ras_error_address)
+			adev->umc.ras->ecc_info_query_ras_error_address(adev, err_data);
 	}
 }
 
@@ -928,62 +952,32 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d
 int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 				  struct ras_query_if *info)
 {
+	struct amdgpu_ras_block_object *block_obj = NULL;
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
 	struct ras_err_data err_data = {0, 0, 0, NULL};
-	int i;
 
 	if (!obj)
 		return -EINVAL;
 
-	switch (info->head.block) {
-	case AMDGPU_RAS_BLOCK__UMC:
+	if (info->head.block == AMDGPU_RAS_BLOCK__UMC) {
 		amdgpu_ras_get_ecc_info(adev, &err_data);
-		break;
-	case AMDGPU_RAS_BLOCK__SDMA:
-		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);
+	} else {
+		block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0);
+		if (!block_obj || !block_obj->hw_ops)   {
+			dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+				     get_ras_block_str(&info->head));
+			return -EINVAL;
 		}
-		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;
-	case AMDGPU_RAS_BLOCK__MCA:
-		amdgpu_ras_mca_query_error_status(adev, &info->head, &err_data);
-		break;
-	default:
-		break;
+
+		if (block_obj->hw_ops->query_ras_error_count)
+			block_obj->hw_ops->query_ras_error_count(adev, &err_data);
+
+		if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) ||
+		    (info->head.block == AMDGPU_RAS_BLOCK__GFX) ||
+		    (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) {
+				if (block_obj->hw_ops->query_ras_error_status)
+					block_obj->hw_ops->query_ras_error_status(adev);
+			}
 	}
 
 	obj->err_data.ue_count += err_data.ue_count;
@@ -1040,68 +1034,27 @@ int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
 		enum amdgpu_ras_block block)
 {
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+
 	if (!amdgpu_ras_is_supported(adev, block))
 		return -EINVAL;
 
-	switch (block) {
-	case AMDGPU_RAS_BLOCK__GFX:
-		if (adev->gfx.ras_funcs &&
-		    adev->gfx.ras_funcs->reset_ras_error_count)
-			adev->gfx.ras_funcs->reset_ras_error_count(adev);
-
-		if (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;
+	if (!block_obj || !block_obj->hw_ops)   {
+		dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+			     ras_block_str(block));
+		return -EINVAL;
 	}
 
-	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 (block_obj->hw_ops->reset_ras_error_count)
+		block_obj->hw_ops->reset_ras_error_count(adev);
 
-	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");
+	if ((block == AMDGPU_RAS_BLOCK__GFX) ||
+	    (block == AMDGPU_RAS_BLOCK__MMHUB)) {
+		if (block_obj->hw_ops->reset_ras_error_status)
+			block_obj->hw_ops->reset_ras_error_status(adev);
+	}
 
-	return ret;
+	return 0;
 }
 
 /* wrapper of psp_ras_trigger_error */
@@ -1116,11 +1069,20 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 		.address = info->address,
 		.value = info->value,
 	};
-	int ret = 0;
+	int ret = -EINVAL;
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev,
+							info->head.block,
+							info->head.sub_block_index);
 
 	if (!obj)
 		return -EINVAL;
 
+	if (!block_obj || !block_obj->hw_ops)	{
+		dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+			     get_ras_block_str(&info->head));
+		return -EINVAL;
+	}
+
 	/* Calculate XGMI relative offset */
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
 		block_info.address =
@@ -1128,28 +1090,15 @@ int amdgpu_ras_error_inject(struct amdgpu_device *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:
-	case AMDGPU_RAS_BLOCK__MCA:
-		ret = psp_ras_trigger_error(&adev->psp, &block_info);
-		break;
-	case AMDGPU_RAS_BLOCK__XGMI_WAFL:
-		ret = amdgpu_ras_error_inject_xgmi(adev, &block_info);
-		break;
-	default:
-		dev_info(adev->dev, "%s error injection is not supported yet\n",
-			 get_ras_block_str(&info->head));
-		ret = -EINVAL;
+	if (info->head.block == AMDGPU_RAS_BLOCK__GFX) {
+		if (block_obj->hw_ops->ras_error_inject)
+			ret = block_obj->hw_ops->ras_error_inject(adev, info);
+	} else {
+		/* If defined special ras_error_inject(e.g: xgmi), implement special ras_error_inject */
+		if (block_obj->hw_ops->ras_error_inject)
+			ret = block_obj->hw_ops->ras_error_inject(adev, &block_info);
+		else  /*If not defined .ras_error_inject, use default ras_error_inject*/
+			ret = psp_ras_trigger_error(&adev->psp, &block_info);
 	}
 
 	if (ret)
@@ -1329,18 +1278,17 @@ static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev)
 }
 
 int amdgpu_ras_sysfs_create(struct amdgpu_device *adev,
-		struct ras_fs_if *head)
+		struct ras_common_if *head)
 {
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->head);
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 
 	if (!obj || obj->attr_inuse)
 		return -EINVAL;
 
 	get_obj(obj);
 
-	memcpy(obj->fs_data.sysfs_name,
-			head->sysfs_name,
-			sizeof(obj->fs_data.sysfs_name));
+	snprintf(obj->fs_data.sysfs_name, sizeof(obj->fs_data.sysfs_name),
+		"%s_err_count", head->name);
 
 	obj->sysfs_attr = (struct device_attribute){
 		.attr = {
@@ -1567,12 +1515,97 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev)
 /* ras fs end */
 
 /* ih begin */
+
+/* For the hardware that cannot enable bif ring for both ras_controller_irq
+ * and ras_err_evnet_athub_irq ih cookies, the driver has to poll status
+ * register to check whether the interrupt is triggered or not, and properly
+ * ack the interrupt if it is there
+ */
+void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev)
+{
+	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF))
+		return;
+
+	if (adev->nbio.ras &&
+	    adev->nbio.ras->handle_ras_controller_intr_no_bifring)
+		adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev);
+
+	if (adev->nbio.ras &&
+	    adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring)
+		adev->nbio.ras->handle_ras_err_event_athub_intr_no_bifring(adev);
+}
+
+static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *obj,
+				struct amdgpu_iv_entry *entry)
+{
+	bool poison_stat = true, need_reset = true;
+	struct amdgpu_device *adev = obj->adev;
+	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct amdgpu_ras_block_object *block_obj =
+		amdgpu_ras_get_ras_block(adev, obj->head.block, 0);
+
+	if (!adev->gmc.xgmi.connected_to_cpu)
+		amdgpu_umc_poison_handler(adev, &err_data, false);
+
+	/* both query_poison_status and handle_poison_consumption are optional */
+	if (block_obj && block_obj->hw_ops) {
+		if (block_obj->hw_ops->query_poison_status) {
+			poison_stat = block_obj->hw_ops->query_poison_status(adev);
+			if (!poison_stat)
+				dev_info(adev->dev, "No RAS poison status in %s poison IH.\n",
+						block_obj->ras_comm.name);
+		}
+
+		if (poison_stat && block_obj->hw_ops->handle_poison_consumption) {
+			poison_stat = block_obj->hw_ops->handle_poison_consumption(adev);
+			need_reset = poison_stat;
+		}
+	}
+
+	/* gpu reset is fallback for all failed cases */
+	if (need_reset)
+		amdgpu_ras_reset_gpu(adev);
+}
+
+static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj,
+				struct amdgpu_iv_entry *entry)
+{
+	dev_info(obj->adev->dev,
+		"Poison is created, no user action is needed.\n");
+}
+
+static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj,
+				struct amdgpu_iv_entry *entry)
+{
+	struct ras_ih_data *data = &obj->ih_data;
+	struct ras_err_data err_data = {0, 0, 0, NULL};
+	int ret;
+
+	if (!data->cb)
+		return;
+
+	/* Let IP handle its data, maybe we need get the output
+	 * from the callback to update the error type/count, etc
+	 */
+	ret = data->cb(obj->adev, &err_data, entry);
+	/* ue will trigger an interrupt, and in that case
+	 * we need do a reset to recovery the whole system.
+	 * But leave IP do that recovery, here we just dispatch
+	 * the error.
+	 */
+	if (ret == AMDGPU_RAS_SUCCESS) {
+		/* these counts could be left as 0 if
+		 * some blocks do not count error number
+		 */
+		obj->err_data.ue_count += err_data.ue_count;
+		obj->err_data.ce_count += err_data.ce_count;
+	}
+}
+
 static void amdgpu_ras_interrupt_handler(struct ras_manager *obj)
 {
 	struct ras_ih_data *data = &obj->ih_data;
 	struct amdgpu_iv_entry entry;
-	int ret;
-	struct ras_err_data err_data = {0, 0, 0, NULL};
 
 	while (data->rptr != data->wptr) {
 		rmb();
@@ -1583,30 +1616,17 @@ static void amdgpu_ras_interrupt_handler(struct ras_manager *obj)
 		data->rptr = (data->aligned_element_size +
 				data->rptr) % data->ring_size;
 
-		if (data->cb) {
-			if (amdgpu_ras_is_poison_mode_supported(obj->adev) &&
-			    obj->head.block == AMDGPU_RAS_BLOCK__UMC)
-				dev_info(obj->adev->dev,
-						"Poison is created, no user action is needed.\n");
-			else {
-				/* Let IP handle its data, maybe we need get the output
-				 * from the callback to udpate the error type/count, etc
-				 */
-				memset(&err_data, 0, sizeof(err_data));
-				ret = data->cb(obj->adev, &err_data, &entry);
-				/* ue will trigger an interrupt, and in that case
-				 * we need do a reset to recovery the whole system.
-				 * But leave IP do that recovery, here we just dispatch
-				 * the error.
-				 */
-				if (ret == AMDGPU_RAS_SUCCESS) {
-					/* these counts could be left as 0 if
-					 * some blocks do not count error number
-					 */
-					obj->err_data.ue_count += err_data.ue_count;
-					obj->err_data.ce_count += err_data.ce_count;
-				}
-			}
+		if (amdgpu_ras_is_poison_mode_supported(obj->adev)) {
+			if (obj->head.block == AMDGPU_RAS_BLOCK__UMC)
+				amdgpu_ras_interrupt_poison_creation_handler(obj, &entry);
+			else
+				amdgpu_ras_interrupt_poison_consumption_handler(obj, &entry);
+		} else {
+			if (obj->head.block == AMDGPU_RAS_BLOCK__UMC)
+				amdgpu_ras_interrupt_umc_handler(obj, &entry);
+			else
+				dev_warn(obj->adev->dev,
+					"No RAS interrupt handler for non-UMC block with poison disabled.\n");
 		}
 	}
 }
@@ -1647,9 +1667,9 @@ int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev,
 }
 
 int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev,
-		struct ras_ih_if *info)
+		struct ras_common_if *head)
 {
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 	struct ras_ih_data *data;
 
 	if (!obj)
@@ -1669,24 +1689,27 @@ int amdgpu_ras_interrupt_remove_handler(struct amdgpu_device *adev,
 }
 
 int amdgpu_ras_interrupt_add_handler(struct amdgpu_device *adev,
-		struct ras_ih_if *info)
+		struct ras_common_if *head)
 {
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 	struct ras_ih_data *data;
+	struct amdgpu_ras_block_object *ras_obj;
 
 	if (!obj) {
 		/* in case we registe the IH before enable ras feature */
-		obj = amdgpu_ras_create_obj(adev, &info->head);
+		obj = amdgpu_ras_create_obj(adev, head);
 		if (!obj)
 			return -EINVAL;
 	} else
 		get_obj(obj);
 
+	ras_obj = container_of(head, struct amdgpu_ras_block_object, ras_comm);
+
 	data = &obj->ih_data;
 	/* add the callback.etc */
 	*data = (struct ras_ih_data) {
 		.inuse = 0,
-		.cb = info->cb,
+		.cb = ras_obj->ras_cb,
 		.element_size = sizeof(struct amdgpu_iv_entry),
 		.rptr = 0,
 		.wptr = 0,
@@ -1715,10 +1738,7 @@ static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev)
 	struct ras_manager *obj, *tmp;
 
 	list_for_each_entry_safe(obj, tmp, &con->head, node) {
-		struct ras_ih_if info = {
-			.head = obj->head,
-		};
-		amdgpu_ras_interrupt_remove_handler(adev, &info);
+		amdgpu_ras_interrupt_remove_handler(adev, &obj->head);
 	}
 
 	return 0;
@@ -1766,24 +1786,28 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev)
 static void amdgpu_ras_error_status_query(struct amdgpu_device *adev,
 					  struct ras_query_if *info)
 {
+	struct amdgpu_ras_block_object *block_obj;
 	/*
 	 * 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;
+	if ((info->head.block != AMDGPU_RAS_BLOCK__GFX) &&
+		(info->head.block != AMDGPU_RAS_BLOCK__MMHUB))
+		return;
+
+	block_obj = amdgpu_ras_get_ras_block(adev,
+					info->head.block,
+					info->head.sub_block_index);
+
+	if (!block_obj || !block_obj->hw_ops) {
+		dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+			     get_ras_block_str(&info->head));
+		return;
 	}
+
+	if (block_obj->hw_ops->query_ras_error_status)
+		block_obj->hw_ops->query_ras_error_status(adev);
+
 }
 
 static void amdgpu_ras_query_err_status(struct amdgpu_device *adev)
@@ -1897,7 +1921,6 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL);
 
 	if (!bps) {
-		kfree(bps);
 		return -ENOMEM;
 	}
 
@@ -2118,6 +2141,7 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
 	mutex_init(&con->recovery_lock);
 	INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery);
 	atomic_set(&con->in_recovery, 0);
+	con->eeprom_control.bad_channel_bitmap = 0;
 
 	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count();
 	amdgpu_ras_validate_threshold(adev, max_eeprom_records_count);
@@ -2141,8 +2165,12 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
 		if (ret)
 			goto free;
 
-		if (adev->smu.ppt_funcs && adev->smu.ppt_funcs->send_hbm_bad_pages_num)
-			adev->smu.ppt_funcs->send_hbm_bad_pages_num(&adev->smu, con->eeprom_control.ras_num_recs);
+		amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
+
+		if (con->update_channel_flag == true) {
+			amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
+			con->update_channel_flag = false;
+		}
 	}
 
 #ifdef CONFIG_X86_MCE_AMD
@@ -2250,6 +2278,13 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 			dev_info(adev->dev, "SRAM ECC is active.\n");
 			adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
 						    1 << AMDGPU_RAS_BLOCK__DF);
+
+			if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0))
+				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
+						1 << AMDGPU_RAS_BLOCK__JPEG);
+			else
+				adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
+						1 << AMDGPU_RAS_BLOCK__JPEG);
 		} else {
 			dev_info(adev->dev, "SRAM ECC is not presented.\n");
 		}
@@ -2336,6 +2371,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 		goto release_con;
 	}
 
+	con->update_channel_flag = false;
 	con->features = 0;
 	INIT_LIST_HEAD(&con->head);
 	/* Might need get this flag from vbios. */
@@ -2348,24 +2384,27 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	case CHIP_VEGA20:
 	case CHIP_ARCTURUS:
 	case CHIP_ALDEBARAN:
-		if (!adev->gmc.xgmi.connected_to_cpu)
-			adev->nbio.ras_funcs = &nbio_v7_4_ras_funcs;
+		if (!adev->gmc.xgmi.connected_to_cpu) {
+			adev->nbio.ras = &nbio_v7_4_ras;
+			amdgpu_ras_register_ras_block(adev, &adev->nbio.ras->ras_block);
+			adev->nbio.ras_if = &adev->nbio.ras->ras_block.ras_comm;
+		}
 		break;
 	default:
 		/* nbio ras is not available */
 		break;
 	}
 
-	if (adev->nbio.ras_funcs &&
-	    adev->nbio.ras_funcs->init_ras_controller_interrupt) {
-		r = adev->nbio.ras_funcs->init_ras_controller_interrupt(adev);
+	if (adev->nbio.ras &&
+	    adev->nbio.ras->init_ras_controller_interrupt) {
+		r = adev->nbio.ras->init_ras_controller_interrupt(adev);
 		if (r)
 			goto release_con;
 	}
 
-	if (adev->nbio.ras_funcs &&
-	    adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt) {
-		r = adev->nbio.ras_funcs->init_ras_err_event_athub_interrupt(adev);
+	if (adev->nbio.ras &&
+	    adev->nbio.ras->init_ras_err_event_athub_interrupt) {
+		r = adev->nbio.ras->init_ras_err_event_athub_interrupt(adev);
 		if (r)
 			goto release_con;
 	}
@@ -2377,12 +2416,12 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	}
 	else if (adev->df.funcs &&
 	    adev->df.funcs->query_ras_poison_mode &&
-	    adev->umc.ras_funcs &&
-	    adev->umc.ras_funcs->query_ras_poison_mode) {
+	    adev->umc.ras &&
+	    adev->umc.ras->query_ras_poison_mode) {
 		df_poison =
 			adev->df.funcs->query_ras_poison_mode(adev);
 		umc_poison =
-			adev->umc.ras_funcs->query_ras_poison_mode(adev);
+			adev->umc.ras->query_ras_poison_mode(adev);
 		/* Only poison is set in both DF and UMC, we can support it */
 		if (df_poison && umc_poison)
 			con->poison_supported = true;
@@ -2445,11 +2484,10 @@ bool amdgpu_ras_is_poison_mode_supported(struct amdgpu_device *adev)
 }
 
 /* helper function to handle common stuff in ip late init phase */
-int amdgpu_ras_late_init(struct amdgpu_device *adev,
-			 struct ras_common_if *ras_block,
-			 struct ras_fs_if *fs_info,
-			 struct ras_ih_if *ih_info)
+int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
+			 struct ras_common_if *ras_block)
 {
+	struct amdgpu_ras_block_object *ras_obj = NULL;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	unsigned long ue_count, ce_count;
 	int r;
@@ -2477,15 +2515,16 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev,
 	if (adev->in_suspend || amdgpu_in_reset(adev))
 		return 0;
 
-	if (ih_info->cb) {
-		r = amdgpu_ras_interrupt_add_handler(adev, ih_info);
+	ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm);
+	if (ras_obj->ras_cb) {
+		r = amdgpu_ras_interrupt_add_handler(adev, ras_block);
 		if (r)
-			goto interrupt;
+			goto cleanup;
 	}
 
-	r = amdgpu_ras_sysfs_create(adev, fs_info);
+	r = amdgpu_ras_sysfs_create(adev, ras_block);
 	if (r)
-		goto sysfs;
+		goto interrupt;
 
 	/* Those are the cached values at init.
 	 */
@@ -2495,27 +2534,40 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev,
 	}
 
 	return 0;
-cleanup:
-	amdgpu_ras_sysfs_remove(adev, ras_block);
-sysfs:
-	if (ih_info->cb)
-		amdgpu_ras_interrupt_remove_handler(adev, ih_info);
+
 interrupt:
+	if (ras_obj->ras_cb)
+		amdgpu_ras_interrupt_remove_handler(adev, ras_block);
+cleanup:
 	amdgpu_ras_feature_enable(adev, ras_block, 0);
 	return r;
 }
 
+static int amdgpu_ras_block_late_init_default(struct amdgpu_device *adev,
+			 struct ras_common_if *ras_block)
+{
+	return amdgpu_ras_block_late_init(adev, ras_block);
+}
+
 /* 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)
+void amdgpu_ras_block_late_fini(struct amdgpu_device *adev,
+			  struct ras_common_if *ras_block)
 {
-	if (!ras_block || !ih_info)
+	struct amdgpu_ras_block_object *ras_obj;
+	if (!ras_block)
 		return;
 
 	amdgpu_ras_sysfs_remove(adev, ras_block);
-	if (ih_info->cb)
-		amdgpu_ras_interrupt_remove_handler(adev, ih_info);
+
+	ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm);
+	if (ras_obj->ras_cb)
+		amdgpu_ras_interrupt_remove_handler(adev, ras_block);
+}
+
+static void amdgpu_ras_block_late_fini_default(struct amdgpu_device *adev,
+			  struct ras_common_if *ras_block)
+{
+	return amdgpu_ras_block_late_fini(adev, ras_block);
 }
 
 /* do some init work after IP late init as dependence.
@@ -2568,6 +2620,33 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev)
 		amdgpu_ras_disable_all_features(adev, 1);
 }
 
+int amdgpu_ras_late_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras_block_list *node, *tmp;
+	struct amdgpu_ras_block_object *obj;
+	int r;
+
+	list_for_each_entry_safe(node, tmp, &adev->ras_list, node) {
+		if (!node->ras_obj) {
+			dev_warn(adev->dev, "Warning: abnormal ras list node.\n");
+			continue;
+		}
+
+		obj = node->ras_obj;
+		if (obj->ras_late_init) {
+			r = obj->ras_late_init(adev, &obj->ras_comm);
+			if (r) {
+				dev_err(adev->dev, "%s failed to execute ras_late_init! ret:%d\n",
+					obj->ras_comm.name, r);
+				return r;
+			}
+		} else
+			amdgpu_ras_block_late_init_default(adev, &obj->ras_comm);
+	}
+
+	return 0;
+}
+
 /* do some fini work before IP fini as dependence */
 int amdgpu_ras_pre_fini(struct amdgpu_device *adev)
 {
@@ -2585,11 +2664,28 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev)
 
 int amdgpu_ras_fini(struct amdgpu_device *adev)
 {
+	struct amdgpu_ras_block_list *ras_node, *tmp;
+	struct amdgpu_ras_block_object *obj = NULL;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
 	if (!adev->ras_enabled || !con)
 		return 0;
 
+	list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) {
+		if (ras_node->ras_obj) {
+			obj = ras_node->ras_obj;
+			if (amdgpu_ras_is_supported(adev, obj->ras_comm.block) &&
+			    obj->ras_fini)
+				obj->ras_fini(adev, &obj->ras_comm);
+			else
+				amdgpu_ras_block_late_fini_default(adev, &obj->ras_comm);
+		}
+
+		/* Clear ras blocks from ras_list and free ras block list node */
+		list_del(&ras_node->node);
+		kfree(ras_node);
+	}
+
 	amdgpu_ras_fs_fini(adev);
 	amdgpu_ras_interrupt_remove_all(adev);
 
@@ -2717,8 +2813,6 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb,
 	dev_info(adev->dev, "Uncorrectable error detected in UMC inst: %d, chan_idx: %d",
 			     umc_inst, ch_inst);
 
-	memset(&err_rec, 0x0, sizeof(struct eeprom_table_record));
-
 	/*
 	 * Translate UMC channel address to Physical address
 	 */
@@ -2730,16 +2824,10 @@ static int amdgpu_bad_page_notifier(struct notifier_block *nb,
 			ADDR_OF_256B_BLOCK(channel_index) |
 			OFFSET_IN_256B_BLOCK(m->addr);
 
-	err_rec.address = m->addr;
-	err_rec.retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT;
-	err_rec.ts = (uint64_t)ktime_get_real_seconds();
-	err_rec.err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE;
-	err_rec.cu = 0;
-	err_rec.mem_channel = channel_index;
-	err_rec.mcumc_id = umc_inst;
-
+	memset(&err_rec, 0x0, sizeof(struct eeprom_table_record));
 	err_data.err_addr = &err_rec;
-	err_data.err_addr_cnt = 1;
+	amdgpu_umc_fill_error_record(&err_data, m->addr,
+			retired_page, channel_index, umc_inst);
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
@@ -2777,3 +2865,63 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev)
 	}
 }
 #endif
+
+struct amdgpu_ras *amdgpu_ras_get_context(struct amdgpu_device *adev)
+{
+	if (!adev)
+		return NULL;
+
+	return adev->psp.ras_context.ras;
+}
+
+int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con)
+{
+	if (!adev)
+		return -EINVAL;
+
+	adev->psp.ras_context.ras = ras_con;
+	return 0;
+}
+
+/* check if ras is supported on block, say, sdma, gfx */
+int amdgpu_ras_is_supported(struct amdgpu_device *adev,
+		unsigned int block)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (block >= AMDGPU_RAS_BLOCK_COUNT)
+		return 0;
+	return ras && (adev->ras_enabled & (1 << block));
+}
+
+int amdgpu_ras_reset_gpu(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0)
+		schedule_work(&ras->recovery_work);
+	return 0;
+}
+
+
+/* Register each ip ras block into amdgpu ras */
+int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
+		struct amdgpu_ras_block_object *ras_block_obj)
+{
+	struct amdgpu_ras_block_list *ras_node;
+	if (!adev || !ras_block_obj)
+		return -EINVAL;
+
+	if (!amdgpu_ras_asic_supported(adev))
+		return 0;
+
+	ras_node = kzalloc(sizeof(*ras_node), GFP_KERNEL);
+	if (!ras_node)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&ras_node->node);
+	ras_node->ras_obj = ras_block_obj;
+	list_add_tail(&ras_node->node, &adev->ras_list);
+
+	return 0;
+}