On 12/5/19 4:43 PM, Christian König wrote:
Am 05.12.19 um 15:48 schrieb Nirmoy Das:
EXPORT_SYMBOL(drm_sched_entity_set_priority);
@@ -490,18 +474,20 @@ void drm_sched_entity_select_rq(struct
drm_sched_entity *entity)
struct dma_fence *fence;
struct drm_sched_rq *rq;
- if (spsc_queue_count(&entity->job_queue) ||
entity->num_rq_list <= 1)
+ if (spsc_queue_count(&entity->job_queue) ||
entity->num_sched_list <= 1)
return;
fence = READ_ONCE(entity->last_scheduled);
if (fence && !dma_fence_is_signaled(fence))
return;
+ spin_lock(&entity->rq_lock);
rq = drm_sched_entity_get_free_sched(entity);
- if (rq == entity->rq)
+ if (rq == entity->rq) {
+ spin_unlock(&entity->rq_lock);
return;
+ }
Should we redesign that into:
if (rq != entity->rq) {
- spin_lock(&entity->rq_lock);
drm_sched_rq_remove_entity(entity->rq, entity);
entity->rq = rq;
}
Would probably be easier to understand.
Yes, looks much cleaner with that. Sent a v3 with that change.
Apart from that the patch is Reviewed-by: Christian König
<christian.koenig@xxxxxxx>.
Do you want to work on a follow up patch first or should I push this
to drm-misc-next?
Please push this to drm-misc-next for now. I am thinking of doing rest
of cleanup and on-demad entity creation for each HW IP together.
Regards,
Christian.
_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
