[PATCH] drm/sched: Always trace the dependencies we wait on, to fix a race.

[Koenig, Christian]

Am 07.12.18 um 20:16 schrieb Eric Anholt:
> The entity->dependency can go away completely once we've called
> drm_sched_entity_add_dependency_cb() (if the cb is called before we
> get around to tracing).  The tracepoint is more useful if we trace
> every dependency instead of just ones that get callbacks installed,
> anyway, so just do that.
>
> Fixes any easy-to-produce OOPS when tracing the scheduler on V3D with
> "perf record -a -e gpu_scheduler:.\* glxgears" and DEBUG_SLAB enabled.
>
> Signed-off-by: Eric Anholt <eric at anholt.net>

Reviewed-by: Christian König <christian.koenig at amd.com>

Going to pick that up for upstream and will add with a CC: stable.

Thanks,
Christian.

> ---
>   drivers/gpu/drm/scheduler/sched_entity.c | 7 ++-----
>   1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index 4463d3826ecb..e2942c9a11a7 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -440,13 +440,10 @@ struct drm_sched_job *drm_sched_entity_pop_job(struct drm_sched_entity *entity)
>   
>   	while ((entity->dependency =
>   			sched->ops->dependency(sched_job, entity))) {
> +		trace_drm_sched_job_wait_dep(sched_job, entity->dependency);
>   
> -		if (drm_sched_entity_add_dependency_cb(entity)) {
> -
> -			trace_drm_sched_job_wait_dep(sched_job,
> -						     entity->dependency);
> +		if (drm_sched_entity_add_dependency_cb(entity))
>   			return NULL;
> -		}
>   	}
>   
>   	/* skip jobs from entity that marked guilty */