Skip to contentSkip to conversation

drm/amdgpu/userq: lock and validate wptr BOs before reading their GPU offset on restore

PATCH Superseded Jesse Zhang 2 participants 2 messages Aug 17 – Aug 24

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu/userq: lock and validate wptr BOs before reading their GPU offset on restore

On resume, amdgpu_userq_vm_validate_and_restore_queue() updates each queue's wptr GPU address via amdgpu_bo_gpu_offset().

WPTR BOs are VM-mapped, but each BO has its own reservation object and is not implicitly covered by the VM validation path here. This can leave offset reads without proper BO locking/placement state and trigger WARN_ONs.

  ------------[ cut here ]------------
  WARNING: amdgpu_object.c:1486 at amdgpu_bo_gpu_offset+0x75/0xa0 [amdgpu], CPU#3: kworker/3:1/116
  Workqueue: events amdgpu_userq_restore_worker [amdgpu]
  RIP: 0010:amdgpu_bo_gpu_offset+0x75/0xa0 [amdgpu]
  Call Trace:
   <TASK>
   amdgpu_userq_vm_validate_and_restore_queue+0x629/0x960 [amdgpu]
   amdgpu_userq_restore_worker+0xa6/0x180 [amdgpu]
   process_scheduled_works+0xa6/0x460
   worker_thread+0x13c/0x290
   kthread+0xfb/0x140
   ret_from_fork+0x1b6/0x2b0
   ret_from_fork_asm+0x1a/0x30
   </TASK>
  ---[ end trace 0000000000000000 ]---
  ------------[ cut here ]------------
  WARNING: amdgpu_object.c:1485 at amdgpu_bo_gpu_offset+0x9a/0xa0 [amdgpu], CPU#2: kworker/2:1/127
  Workqueue: events amdgpu_userq_restore_worker [amdgpu]
  RIP: 0010:amdgpu_bo_gpu_offset+0x9a/0xa0 [amdgpu]

Add each queue's WPTR BO to the drm_exec ww context and validate it to its allowed placement before the later offset update.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 24 +++++++++++++++++++++++
 1 file changed, 24 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +24 -0
index 17cc48d87c4d..ab8fc14a235b 100644
@@ -1054,6 +1054,30 @@ amdgpu_userq_vm_validate_and_restore_queue(struct amdgpu_userq_mgr *uq_mgr)
 		drm_exec_retry_on_contention(&exec);
 		if (unlikely(ret))
 			goto unlock_all;
+
+		/*
+		 * WPTR BOs are VM-mapped, but each BO has its own reservation
+		 * object. Lock them into this drm_exec ww context so the later
+		 * amdgpu_bo_gpu_offset() reads are done with the BO resv locked.
+		 */
+		xa_for_each(&uq_mgr->userq_xa, tmp_key, queue) {
+			struct ttm_operation_ctx wptr_ctx = { false, false };
+
+			bo = queue->wptr_obj.obj;
+			if (!bo)
+				continue;
+
+			ret = drm_exec_prepare_obj(&exec, &bo->tbo.base,
+						   TTM_NUM_MOVE_FENCES + 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unlock_all;
+
+			amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &wptr_ctx);
+			if (unlikely(ret))
+				goto unlock_all;
+		}
 	}
 
 	if (invalidated) {
Signature · 1 line
2.49.0
maintainer
#1 lore raw
Quoted 56 lines · Jesse Zhang

On resume, amdgpu_userq_vm_validate_and_restore_queue() updates each queue's wptr GPU address via amdgpu_bo_gpu_offset().

WPTR BOs are VM-mapped, but each BO has its own reservation object and is not implicitly covered by the VM validation path here. This can leave offset reads without proper BO locking/placement state and trigger WARN_ONs.

  ------------[ cut here ]------------
  WARNING: amdgpu_object.c:1486 at amdgpu_bo_gpu_offset+0x75/0xa0 [amdgpu], CPU#3: kworker/3:1/116
  Workqueue: events amdgpu_userq_restore_worker [amdgpu]
  RIP: 0010:amdgpu_bo_gpu_offset+0x75/0xa0 [amdgpu]
  Call Trace:
   <TASK>
   amdgpu_userq_vm_validate_and_restore_queue+0x629/0x960 [amdgpu]
   amdgpu_userq_restore_worker+0xa6/0x180 [amdgpu]
   process_scheduled_works+0xa6/0x460
   worker_thread+0x13c/0x290
   kthread+0xfb/0x140
   ret_from_fork+0x1b6/0x2b0
   ret_from_fork_asm+0x1a/0x30
   </TASK>
  ---[ end trace 0000000000000000 ]---
  ------------[ cut here ]------------
  WARNING: amdgpu_object.c:1485 at amdgpu_bo_gpu_offset+0x9a/0xa0 [amdgpu], CPU#2: kworker/2:1/127
  Workqueue: events amdgpu_userq_restore_worker [amdgpu]
  RIP: 0010:amdgpu_bo_gpu_offset+0x9a/0xa0 [amdgpu]

Add each queue's WPTR BO to the drm_exec ww context and validate it to its allowed placement before the later offset update.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c | 24 +++++++++++++++++++++++
 1 file changed, 24 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +14 -0
index 17cc48d87c4d..ab8fc14a235b 100644
@@ -1054,6 +1054,30 @@ amdgpu_userq_vm_validate_and_restore_queue(struct amdgpu_userq_mgr *uq_mgr)
 		drm_exec_retry_on_contention(&exec);
 		if (unlikely(ret))
 			goto unlock_all;
+
+		/*
+		 * WPTR BOs are VM-mapped, but each BO has its own reservation
+		 * object. Lock them into this drm_exec ww context so the later
+		 * amdgpu_bo_gpu_offset() reads are done with the BO resv locked.
+		 */
+		xa_for_each(&uq_mgr->userq_xa, tmp_key, queue) {
+			struct ttm_operation_ctx wptr_ctx = { false, false };
+
+			bo = queue->wptr_obj.obj;
+			if (!bo)
+				continue;
+
+			ret = drm_exec_prepare_obj(&exec, &bo->tbo.base,

drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c amdgpu_userq_vm_validate_and_restore_queue(struct amdgpu_userq_mgr *uq_mgr) · line 1080

Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +10 -0
+						   TTM_NUM_MOVE_FENCES + 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unlock_all;
+
+			amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &wptr_ctx);
+			if (unlikely(ret))
+				goto unlock_all;
+		}

Mhm, something is very wrong here and that is clearly not the right solution.

At this point all BOs which are part of the VM should already be locked and validated and that include the WPTR BOs.

So you are just working around a bug somewhere else.

Regards,
Christian.

Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +0 -0
 	}
 
 	if (invalidated) {