drm/amdgpu: fix lockup when clean pending fences

The first lockup fence will lock the fence list of scheduler.
Then cancel the delayed workqueues for all clean pending fences
without waiting the workqueues to finish.

Change-Id: I9bec826de1aa49d587b0662f3fb4a95333979429
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
index a413dee7cd190d4c6ee43c59bf3342bb7a2544a2..7fa1d7a438e9c71ba76477f76d3f2d5379ca5955 100644 (file)
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
@@ -352,14 +352,14 @@ static void amd_sched_fence_work_func(struct work_struct *work)
        DRM_ERROR("[%s] scheduler is timeout!\n", sched->name);
 
        /* Clean all pending fences */
+       spin_lock_irqsave(&sched->fence_list_lock, flags);
        list_for_each_entry_safe(entity, tmp, &sched->fence_list, list) {
                DRM_ERROR("  fence no %d\n", entity->base.seqno);
-               cancel_delayed_work_sync(&entity->dwork);
-               spin_lock_irqsave(&sched->fence_list_lock, flags);
+               cancel_delayed_work(&entity->dwork);
                list_del_init(&entity->list);
-               spin_unlock_irqrestore(&sched->fence_list_lock, flags);
                fence_put(&entity->base);
        }
+       spin_unlock_irqrestore(&sched->fence_list_lock, flags);
 }
 
 static int amd_sched_main(void *param)