drm/amdgpu: set an error on all fences from a bad context

When we backup ring contents to reemit after a queue reset,
we don't backup ring contents from the bad context.  When
we signal the fences, we should set an error on those
fences as well.

v2: misc cleanups
v3: add locking for fence error, fix comment (Christian)
v4: fix wrap around, locking (Christian)

Fixes: 77cc0da ("drm/amdgpu: track ring state associated with a fence")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: alexdeucher

drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c

@@ -758,11 +758,42 @@ void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring)
  * @fence: fence of the ring to signal
  *
  */
-void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *fence)
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af)
 {
-	dma_fence_set_error(&fence->base, -ETIME);
-	amdgpu_fence_write(fence->ring, fence->seq);
-	amdgpu_fence_process(fence->ring);
+	struct dma_fence *unprocessed;
+	struct dma_fence __rcu **ptr;
+	struct amdgpu_fence *fence;
+	struct amdgpu_ring *ring = af->ring;
+	unsigned long flags;
+	u32 seq, last_seq;
+
+	last_seq = amdgpu_fence_read(ring) & ring->fence_drv.num_fences_mask;
+	seq = ring->fence_drv.sync_seq & ring->fence_drv.num_fences_mask;
+
+	/* mark all fences from the guilty context with an error */
+	spin_lock_irqsave(&ring->fence_drv.lock, flags);
+	do {
+		last_seq++;
+		last_seq &= ring->fence_drv.num_fences_mask;
+
+		ptr = &ring->fence_drv.fences[last_seq];
+		rcu_read_lock();
+		unprocessed = rcu_dereference(*ptr);
+
+		if (unprocessed && !dma_fence_is_signaled_locked(unprocessed)) {
+			fence = container_of(unprocessed, struct amdgpu_fence, base);
+
+			if (fence == af)
+				dma_fence_set_error(&fence->base, -ETIME);
+			else if (fence->context == af->context)
+				dma_fence_set_error(&fence->base, -ECANCELED);
+		}
+		rcu_read_unlock();
+	} while (last_seq != seq);
+	spin_unlock_irqrestore(&ring->fence_drv.lock, flags);
+	/* signal the guilty fence */
+	amdgpu_fence_write(ring, af->seq);
+	amdgpu_fence_process(ring);
 }
 
 void amdgpu_fence_save_wptr(struct dma_fence *fence)

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -811,7 +811,7 @@ int amdgpu_ring_reset_helper_end(struct amdgpu_ring *ring,
 	if (r)
 		return r;
 
-	/* signal the fence of the bad job */
+	/* signal the guilty fence and set an error on all fences from the context */
 	if (guilty_fence)
 		amdgpu_fence_driver_guilty_force_completion(guilty_fence);
 	/* Re-emit the non-guilty commands */

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h

@@ -155,7 +155,7 @@ extern const struct drm_sched_backend_ops amdgpu_sched_ops;
 void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring);
 void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error);
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
-void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *fence);
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af);
 void amdgpu_fence_save_wptr(struct dma_fence *fence);
 
 int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);