drm/etnaviv: don't block scheduler when GPU is still active

commit 704d3d6 upstream.

Since 45ecaea ("drm/sched: Partial revert of 'drm/sched: Keep
s_fence->parent pointer'") still active jobs aren't put back in the
pending list on drm_sched_start(), as they don't have a active
parent fence anymore, so if the GPU is still working and the timeout
is extended, all currently active jobs will be freed.

To avoid prematurely freeing jobs that are still active on the GPU,
don't block the scheduler until we are fully committed to actually
reset the GPU.

As the current job is already removed from the pending list and
will not be put back when drm_sched_start() isn't called, we must
make sure to put the job back on the pending list when extending
the timeout.

Cc: [email protected] #6.0
Signed-off-by: Lucas Stach <[email protected]>
Reviewed-by: Philipp Zabel <[email protected]>
Reviewed-by: Christian Gmeiner <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

Author: lynxeye-dev

drivers/gpu/drm/etnaviv/etnaviv_sched.c

@@ -38,9 +38,6 @@ static enum drm_gpu_sched_stat etnaviv_sched_timedout_job(struct drm_sched_job
 	u32 dma_addr;
 	int change;
 
-	/* block scheduler */
-	drm_sched_stop(&gpu->sched, sched_job);
-
 	/*
 	 * If the GPU managed to complete this jobs fence, the timout is
 	 * spurious. Bail out.
@@ -62,6 +59,9 @@ static enum drm_gpu_sched_stat etnaviv_sched_timedout_job(struct drm_sched_job
 		goto out_no_timeout;
 	}
 
+	/* block scheduler */
+	drm_sched_stop(&gpu->sched, sched_job);
+
 	if(sched_job)
 		drm_sched_increase_karma(sched_job);
 
@@ -75,8 +75,7 @@ static enum drm_gpu_sched_stat etnaviv_sched_timedout_job(struct drm_sched_job
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 
 out_no_timeout:
-	/* restart scheduler after GPU is usable again */
-	drm_sched_start(&gpu->sched, true);
+	list_add(&sched_job->list, &sched_job->sched->pending_list);
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 }