[PATCH 08/34] drm/amdgpu: add per ring reset support (v2)

[Alex Deucher]

If a specific job is hung, try and reset just the
ring associated with the job.

v2: move to amdgpu_job.c

Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index e238f2832f65..5d4883df64d8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -72,6 +72,23 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 
 	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 
+	/* attempt a per ring reset */
+	if (amdgpu_gpu_recovery &&
+	    ring->funcs->reset) {
+		if (amdgpu_ring_sched_ready(ring))
+			drm_sched_stop(&ring->sched, s_job);
+		r = amdgpu_ring_reset(ring, job->vmid);
+		if (!r) {
+			/* XXX: these are required for subsequent jobs to work */
+			amdgpu_fence_driver_clear_job_fences(ring);
+			amdgpu_fence_driver_force_completion(ring);
+			drm_sched_increase_karma(s_job);
+			if (amdgpu_ring_sched_ready(ring))
+				drm_sched_start(&ring->sched, true);
+			goto exit;
+		}
+	}
+
 	if (amdgpu_device_should_recover_gpu(ring->adev)) {
 		struct amdgpu_reset_context reset_context;
 		memset(&reset_context, 0, sizeof(reset_context));
-- 
2.45.2