From: YiPeng Chai <YiPeng.Chai@amd.com>
Date: Thu, 24 Jul 2025 07:34:29 +0000 (+0800)
Subject: drm/amdgpu: support ras critical address check
X-Git-Url: https://www.infradead.org/git/?a=commitdiff_plain;h=f3486918979030f8982e1af901561dbd6e2cd1bc;p=users%2Fwilly%2Fxarray.git

drm/amdgpu: support ras critical address check

Support ras critical address check.

Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index d49671a2ecd5..b75a0dc22ace 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -143,6 +143,10 @@ static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr);
 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 {
@@ -3728,6 +3732,8 @@ 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 */
@@ -4157,6 +4163,9 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	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);
@@ -4436,6 +4445,9 @@ 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;
@@ -5380,3 +5392,80 @@ 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;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
index aa155cb5d58e..434e23c84962 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
@@ -496,6 +496,13 @@ struct ras_ecc_log_info {
 	uint64_t	prev_de_queried_count;
 };
 
+struct ras_critical_region {
+	struct list_head node;
+	struct amdgpu_bo *bo;
+	uint64_t start;
+	uint64_t size;
+};
+
 struct amdgpu_ras {
 	/* ras infrastructure */
 	/* for ras itself. */
@@ -575,6 +582,10 @@ struct amdgpu_ras {
 	char init_task_comm[TASK_COMM_LEN];
 
 	int bad_page_num;
+
+	struct list_head critical_region_head;
+	struct mutex critical_region_lock;
+
 };
 
 struct ras_fs_data {
@@ -979,6 +990,9 @@ int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_
 
 int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn);
 
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev, struct amdgpu_bo *bo);
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr);
+
 int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
 		enum amdgpu_ras_block block, uint16_t pasid,
 		pasid_notify pasid_fn, void *data, uint32_t reset);