drm/sched: Add locking to drm_sched_entity_modify_sched

Without the locking amdgpu currently can race between
amdgpu_ctx_set_entity_priority() (via drm_sched_entity_modify_sched()) and
drm_sched_job_arm(), leading to the latter accesing potentially
inconsitent entity->sched_list and entity->num_sched_list pair.

v2:
 * Improve commit message. (Philipp)

Signed-off-by: Tvrtko Ursulin <[email protected]>
Fixes: b37aced31eb0 ("drm/scheduler: implement a function to modify sched list")
Cc: Christian König <[email protected]>
Cc: Alex Deucher <[email protected]>
Cc: Luben Tuikov <[email protected]>
Cc: Matthew Brost <[email protected]>
Cc: David Airlie <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: [email protected]
Cc: Philipp Stanner <[email protected]>
Cc: <[email protected]> # v5.7+
Reviewed-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]
Signed-off-by: Christian König <[email protected]>

1 files changed, 2 insertions, 0 deletions

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index ffa3e765f5db..b06c0696f970 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -133,8 +133,10 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
 {
 	WARN_ON(!num_sched_list || !sched_list);
 
+	spin_lock(&entity->rq_lock);
 	entity->sched_list = sched_list;
 	entity->num_sched_list = num_sched_list;
+	spin_unlock(&entity->rq_lock);
 }
 EXPORT_SYMBOL(drm_sched_entity_modify_sched);