Re: [PATCH 2/2] drm/scheduler: stop setting rq to NULL

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On Thu, Aug 2, 2018, 8:09 PM Andrey Grodzovsky <Andrey.Grodzovsky@xxxxxxx> wrote:



On 08/02/2018 02:43 AM, Nayan Deshmukh wrote:
Hi Andrey,

Good catch. I guess since both Christian and I were working on it parallelly we didn't catch this problem.

If it is only causing a problem in push_job then we can handle it something like this:

----------------------------------------------------------------------------------------------------------------------------
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c
index 05dc6ecd4003..f344ce32f128 100644
--- a/drivers/gpu/drm/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c
@@ -563,6 +563,11 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job,
        first = spsc_queue_count(&entity->job_queue) == 0;
        reschedule = idle && first && (entity->num_rq_list > 1);
 
+       if (first && list_empty(&entity->list)) {

first might be false if the other process interrupted by SIGKILL  in middle of  wait_event_killable in drm_sched_entity_flush when there were still item in queue.
I don't see a good way besides adding a 'stopped' flag to drm_sched_enitity.
Sorry I missed this mail. This case might happen but this was also not being handled previously. 

Nayan 

Andrey

+               DRM_ERROR("Trying to push to a killed entity\n");
+               return;
+       }
+
        if (reschedule) {
                rq = drm_sched_entity_get_free_sched(entity);
                spin_lock(&entity->rq_lock);
@@ -580,11 +585,6 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job,
        if (first) {
                /* Add the entity to the run queue */
                spin_lock(&entity->rq_lock);
-               if (!entity->rq) {
-                       DRM_ERROR("Trying to push to a killed entity\n");
-                       spin_unlock(&entity->rq_lock);
-                       return;
-               }
                drm_sched_rq_add_entity(entity->rq, entity);
                spin_unlock(&entity->rq_lock);
                drm_sched_wakeup(entity->rq->sched);
-----------------------------------------------------------------------------------------------------------------------------


On Thu, Aug 2, 2018 at 3:55 AM Andrey Grodzovsky <Andrey.Grodzovsky@xxxxxxx> wrote:
Thinking again about this change and 53d5f1e drm/scheduler: move idle
entities to scheduler with less load v2

Looks to me like use case for which fc9a539 drm/scheduler: add NULL
pointer check for run queue (v2) was done

will not work anymore.

First of all in drm_sched_entity_push_job, 'if (!entity->rq)' will never
be true any more since we stopped entity->rq to NULL in

drm_sched_entity_flush.

Second of all, even after we removed the entity from rq in
drm_sched_entity_flush to terminate any subsequent submissions

to the queue the other thread doing push job can just acquire again a
run queue

from drm_sched_entity_get_free_sched and continue submission so you
can't substitute ' if (!entity->rq)' to 'if (list_empty(&entity->list))'.

What about adding a 'stopped' flag to drm_sched_entity to be set in
drm_sched_entity_flush and in

But it might be worth adding a stopped flag field if it is required at other places as well. 

Thanks,
Nayan
drm_sched_entity_push_job check for  'if (!entity->stopped)' instead of 
' if (!entity->rq)' ?

Andrey


On 07/30/2018 07:03 AM, Christian König wrote:
> We removed the redundancy of having an extra scheduler field, so we
> can't set the rq to NULL any more or otherwise won't know which
> scheduler to use for the cleanup.
>
> Just remove the entity from the scheduling list instead.
>
> Signed-off-by: Christian König <christian.koenig@xxxxxxx>
> ---
>   drivers/gpu/drm/scheduler/gpu_scheduler.c | 35 +++++++------------------------
>   1 file changed, 8 insertions(+), 27 deletions(-)
>
> diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c
> index f563e4fbb4b6..1b733229201e 100644
> --- a/drivers/gpu/drm/scheduler/gpu_scheduler.c
> +++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c
> @@ -198,21 +198,6 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
>   }
>   EXPORT_SYMBOL(drm_sched_entity_init);
>   
> -/**
> - * drm_sched_entity_is_initialized - Query if entity is initialized
> - *
> - * @sched: Pointer to scheduler instance
> - * @entity: The pointer to a valid scheduler entity
> - *
> - * return true if entity is initialized, false otherwise
> -*/
> -static bool drm_sched_entity_is_initialized(struct drm_gpu_scheduler *sched,
> -                                         struct drm_sched_entity *entity)
> -{
> -     return entity->rq != NULL &&
> -             entity->rq->sched == sched;
> -}
> -
>   /**
>    * drm_sched_entity_is_idle - Check if entity is idle
>    *
> @@ -224,7 +209,8 @@ static bool drm_sched_entity_is_idle(struct drm_sched_entity *entity)
>   {
>       rmb();
>   
> -     if (!entity->rq || spsc_queue_peek(&entity->job_queue) == NULL)
> +     if (list_empty(&entity->list) ||
> +         spsc_queue_peek(&entity->job_queue) == NULL)
>               return true;
>   
>       return false;
> @@ -279,8 +265,6 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>       long ret = timeout;
>   
>       sched = entity->rq->sched;
> -     if (!drm_sched_entity_is_initialized(sched, entity))
> -             return ret;
>       /**
>        * The client will not queue more IBs during this fini, consume existing
>        * queued IBs or discard them on SIGKILL
> @@ -299,7 +283,7 @@ long drm_sched_entity_flush(struct drm_sched_entity *entity, long timeout)
>       last_user = cmpxchg(&entity->last_user, current->group_leader, NULL);
>       if ((!last_user || last_user == current->group_leader) &&
>           (current->flags & PF_EXITING) && (current->exit_code == SIGKILL))
> -             drm_sched_entity_set_rq(entity, NULL);
> +             drm_sched_rq_remove_entity(entity->rq, entity);
>   
>       return ret;
>   }
> @@ -320,7 +304,7 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity)
>       struct drm_gpu_scheduler *sched;
>   
>       sched = entity->rq->sched;
> -     drm_sched_entity_set_rq(entity, NULL);
> +     drm_sched_rq_remove_entity(entity->rq, entity);
>   
>       /* Consumption of existing IBs wasn't completed. Forcefully
>        * remove them here.
> @@ -416,15 +400,12 @@ void drm_sched_entity_set_rq(struct drm_sched_entity *entity,
>       if (entity->rq == rq)
>               return;
>   
> -     spin_lock(&entity->rq_lock);
> -
> -     if (entity->rq)
> -             drm_sched_rq_remove_entity(entity->rq, entity);
> +     BUG_ON(!rq);
>   
> +     spin_lock(&entity->rq_lock);
> +     drm_sched_rq_remove_entity(entity->rq, entity);
>       entity->rq = rq;
> -     if (rq)
> -             drm_sched_rq_add_entity(rq, entity);
> -
> +     drm_sched_rq_add_entity(rq, entity);
>       spin_unlock(&entity->rq_lock);
>   }
>   EXPORT_SYMBOL(drm_sched_entity_set_rq);


_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux