Support ras critical address check.
Signed-off-by: YiPeng Chai <yipeng.c...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 89 +++++++++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 14 ++++
2 files changed, 103 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 234f0de31917..0ad3a9eedfd2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -143,6 +143,10 @@ 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);
+
+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 {
@@ -3709,6 +3713,8 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev,
bool init_bp_info)
amdgpu_register_bad_pages_mca_notifier(adev);
#endif
+ amdgpu_ras_critical_region_init(adev);
+
return 0;
free:
@@ -4198,6 +4204,9 @@ 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);
@@ -4477,6 +4486,9 @@ 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;
@@ -5433,3 +5445,80 @@ 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(®ion->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(®ion->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 9faedfc494af..9614cec7c3c3 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 {
void *ras_mgr;
/* ras infrastructure */
@@ -574,6 +581,10 @@ struct amdgpu_ras {
pid_t init_task_pid;
char init_task_comm[TASK_COMM_LEN];
+
+ struct list_head critical_region_head;
+ struct mutex critical_region_lock;
+
};
struct ras_fs_data {
@@ -977,6 +988,9 @@ 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);
--
2.34.1
