[PATCH] drm/xe: Take job list lock in xe_sched_first_pending_job

[Matthew Brost]

On Tue, Nov 05, 2024 at 05:03:27PM +0100, Nirmoy Das wrote:
> Access to the pending_list should always happens under job_list_lock.
> 
> Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")

Is this showing up in any bug reports? The only user of this function is
guc_exec_queue_stop which as stopped the scheduler worker so I think
accessing to the pending_list is in fact safe without this lock. This is
however a micro-optimization in a non-hot path which I think is frowned
upon.

So I think with above the fixes tag is not strickly required.

For the patch though:
Reviewed-by: Matthew Brost <matthew.brost at intel.com>

> Cc: Matthew Brost <matthew.brost at intel.com>
> Signed-off-by: Nirmoy Das <nirmoy.das at intel.com>
> ---
>  drivers/gpu/drm/xe/xe_gpu_scheduler.h | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.h b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> index 64b2ae6839db..c250ea773491 100644
> --- a/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> +++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> @@ -71,8 +71,14 @@ static inline void xe_sched_add_pending_job(struct xe_gpu_scheduler *sched,
>  static inline
>  struct xe_sched_job *xe_sched_first_pending_job(struct xe_gpu_scheduler *sched)
>  {
> -	return list_first_entry_or_null(&sched->base.pending_list,
> -					struct xe_sched_job, drm.list);
> +	struct xe_sched_job *job;
> +
> +	spin_lock(&sched->base.job_list_lock);
> +	job = list_first_entry_or_null(&sched->base.pending_list,
> +				       struct xe_sched_job, drm.list);
> +	spin_unlock(&sched->base.job_list_lock);
> +
> +	return job;
>  }
>  
>  static inline int
> -- 
> 2.46.0
>