Re: [PATCH] drm/sched: Clarify docu concerning drm_sched_job_arm()

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

 



Am 13.03.25 um 10:30 schrieb Philipp Stanner:
> The documentation for drm_sched_job_arm() and especially
> drm_sched_job_cleanup() does not make it very clear why
> drm_sched_job_arm() is a point of no return, which it indeed is.
>
> Make the nature of drm_sched_job_arm() in the docu as clear as possible.
>
> Suggested-by: Christian König <christian.koenig@xxxxxxx>
> Signed-off-by: Philipp Stanner <phasta@xxxxxxxxxx>

Reviewed-by: Christian König <christian.koenig@xxxxxxx>

I'm currently looking into how to fix the amdgpu CS path for gang submission regarding this.

Any objections that I add a preload function to allocate the memory for the XA outside of the critical section?

Regards,
Christian.

> ---
>  drivers/gpu/drm/scheduler/sched_main.c | 24 ++++++++++++++++++------
>  1 file changed, 18 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
> index 4d4219fbe49d..829579c41c6b 100644
> --- a/drivers/gpu/drm/scheduler/sched_main.c
> +++ b/drivers/gpu/drm/scheduler/sched_main.c
> @@ -828,11 +828,15 @@ EXPORT_SYMBOL(drm_sched_job_init);
>   *
>   * This arms a scheduler job for execution. Specifically it initializes the
>   * &drm_sched_job.s_fence of @job, so that it can be attached to struct dma_resv
> - * or other places that need to track the completion of this job.
> + * or other places that need to track the completion of this job. It also
> + * initializes sequence numbers, which are fundamental for fence ordering.
>   *
>   * Refer to drm_sched_entity_push_job() documentation for locking
>   * considerations.
>   *
> + * Once this function was called, you *must* submit @job with
> + * drm_sched_entity_push_job().
> + *
>   * This can only be called if drm_sched_job_init() succeeded.
>   */
>  void drm_sched_job_arm(struct drm_sched_job *job)
> @@ -1017,9 +1021,12 @@ EXPORT_SYMBOL(drm_sched_job_has_dependency);
>   * Drivers should call this from their error unwind code if @job is aborted
>   * before drm_sched_job_arm() is called.
>   *
> - * After that point of no return @job is committed to be executed by the
> - * scheduler, and this function should be called from the
> - * &drm_sched_backend_ops.free_job callback.
> + * drm_sched_job_arm() is a point of no return since it initializes the fences
> + * and their sequence number etc. Once that function has been called, you *must*
> + * submit it with drm_sched_entity_push_job() and cannot simply abort it by
> + * calling drm_sched_job_cleanup().
> + *
> + * This function should be called in the &drm_sched_backend_ops.free_job callback.
>   */
>  void drm_sched_job_cleanup(struct drm_sched_job *job)
>  {
> @@ -1027,10 +1034,15 @@ void drm_sched_job_cleanup(struct drm_sched_job *job)
>  	unsigned long index;
>  
>  	if (kref_read(&job->s_fence->finished.refcount)) {
> -		/* drm_sched_job_arm() has been called */
> +		/* The job has been processed by the scheduler, i.e.,
> +		 * drm_sched_job_arm() and drm_sched_entity_push_job() have
> +		 * been called.
> +		 */
>  		dma_fence_put(&job->s_fence->finished);
>  	} else {
> -		/* aborted job before committing to run it */
> +		/* The job was aborted before it has been committed to be run;
> +		 * notably, drm_sched_job_arm() has not been called.
> +		 */
>  		drm_sched_fence_free(job->s_fence);
>  	}
>  




[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