Skip to contentSkip to conversation

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

PATCH In review Eric Anholt 3 participants 4 messages Dec 7, 2018 – Feb 8, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/sched: Always trace the dependencies we wait on, to fix a race.

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.

---
 drivers/gpu/drm/scheduler/sched_entity.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
drivers/gpu/drm/scheduler/sched_entity.c +2 -5
index 4463d3826ecb..e2942c9a11a7 100644
@@ -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 */
Signature · 1 line
2.20.0.rc1
Reviewed-by on v1 1/1
#1 lore raw

Am 07.12.18 um 20:16 schrieb Eric Anholt:

Quoted 10 lines

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.

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

Thanks,
Christian.

Quoted 24 lines
---
  drivers/gpu/drm/scheduler/sched_entity.c | 7 ++-----
  1 file changed, 2 insertions(+), 5 deletions(-)
drivers/gpu/drm/scheduler/sched_entity.c +2 -5
index 4463d3826ecb..e2942c9a11a7 100644
@@ -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 */
#2 lore raw
Quoted 7 lines · Christian

Am 07.12.18 um 20:16 schrieb Eric Anholt:

Quoted 6 lines

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.

Quoted 8 lines
Quoted 4 lines

Fixes any easy-to-produce OOPS when tracing the scheduler on V3D with "perf record -a -e gpu_scheduler:.\* glxgears" and DEBUG_SLAB enabled.

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

Looks like this got misplaced.

#3 lore raw

Am 08.02.19 um 00:10 schrieb Eric Anholt:

Quoted 9 lines
Quoted 7 lines · Christian

Am 07.12.18 um 20:16 schrieb Eric Anholt:

Quoted 6 lines

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.

Quoted 8 lines
Quoted 7 lines
Quoted 4 lines

Fixes any easy-to-produce OOPS when tracing the scheduler on V3D with "perf record -a -e gpu_scheduler:.\* glxgears" and DEBUG_SLAB enabled.

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

Looks like this got misplaced.

My fault, pushed to our internal branch now with a CC: stable tag on it.

Christian.

Quoted 5 lines

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel