Re: [PATCH v3 03/10] drm/xe/guc_submit: Support cancelling submission

["Dong, Zhanjun" <zhanjun.dong@intel.com>]

On 2026-03-08 9:55 a.m., Raag Jadav wrote:
> In preparation of usecases which require cancelling submission before
> PCIe FLR, introduce xe_guc_submit_cancel() helper. This cancels and
> frees any in-flight jobs on the scheduler.

Could you put more info on why add new cancel functions rather than call 
existing xe_sched_submission_stop?
 From commit message, it looks very similar to stop, which also do stop 
+ free action.

Regards,
Zhanjun Dong
> 
> Signed-off-by: Raag Jadav <raag.jadav@intel.com>
> ---
> v3: Cancel in-flight jobs before FLR
> ---
>   drivers/gpu/drm/xe/xe_gpu_scheduler.c | 11 +++++++++++
>   drivers/gpu/drm/xe/xe_gpu_scheduler.h |  1 +
>   drivers/gpu/drm/xe/xe_guc_submit.c    | 24 ++++++++++++++++++++++++
>   drivers/gpu/drm/xe/xe_guc_submit.h    |  1 +
>   4 files changed, 37 insertions(+)
> 
> diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.c b/drivers/gpu/drm/xe/xe_gpu_scheduler.c
> index 9c8004d5dd91..c012dbe84540 100644
> --- a/drivers/gpu/drm/xe/xe_gpu_scheduler.c
> +++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.c
> @@ -90,6 +90,17 @@ void xe_sched_fini(struct xe_gpu_scheduler *sched)
>   	drm_sched_fini(&sched->base);
>   }
>   
> +void xe_sched_submission_cancel(struct xe_gpu_scheduler *sched)
> +{
> +	struct drm_gpu_scheduler *base = &sched->base;
> +	struct drm_sched_job *job, *tmp;
> +
> +	list_for_each_entry_safe_reverse(job, tmp, &base->pending_list, list) {
> +		list_del(&job->list);
> +		base->ops->free_job(job);
> +	}
> +}
> +
>   void xe_sched_submission_start(struct xe_gpu_scheduler *sched)
>   {
>   	drm_sched_wqueue_start(&sched->base);
> diff --git a/drivers/gpu/drm/xe/xe_gpu_scheduler.h b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> index 664c2db56af3..ba7892db8428 100644
> --- a/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> +++ b/drivers/gpu/drm/xe/xe_gpu_scheduler.h
> @@ -19,6 +19,7 @@ int xe_sched_init(struct xe_gpu_scheduler *sched,
>   		  struct device *dev);
>   void xe_sched_fini(struct xe_gpu_scheduler *sched);
>   
> +void xe_sched_submission_cancel(struct xe_gpu_scheduler *sched);
>   void xe_sched_submission_start(struct xe_gpu_scheduler *sched);
>   void xe_sched_submission_stop(struct xe_gpu_scheduler *sched);
>   
> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> index de716c1fb18e..cba544cc185c 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> @@ -2399,6 +2399,30 @@ void xe_guc_submit_stop(struct xe_guc *guc)
>   
>   }
>   
> +/**
> + * xe_guc_submit_cancel - Cancel all runs of submission tasks on given GuC.
> + * @guc: the &xe_guc struct instance whose scheduler is to be cancelled
> + */
> +void xe_guc_submit_cancel(struct xe_guc *guc)
> +{
> +	struct xe_exec_queue *q;
> +	unsigned long index;
> +
> +	mutex_lock(&guc->submission_state.lock);
> +
> +	xa_for_each(&guc->submission_state.exec_queue_lookup, index, q) {
> +		struct xe_gpu_scheduler *sched = &q->guc->sched;
> +
> +		/* Prevent redundant attempts to cancel parallel queues */
> +		if (q->guc->id != index)
> +			continue;
> +
> +		xe_sched_submission_cancel(sched);
> +	}
> +
> +	mutex_unlock(&guc->submission_state.lock);
> +}
> +
>   static void guc_exec_queue_revert_pending_state_change(struct xe_guc *guc,
>   						       struct xe_exec_queue *q)
>   {
> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.h b/drivers/gpu/drm/xe/xe_guc_submit.h
> index b3839a90c142..f361a6d32fd3 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.h
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.h
> @@ -16,6 +16,7 @@ int xe_guc_submit_init(struct xe_guc *guc, unsigned int num_ids);
>   int xe_guc_submit_enable(struct xe_guc *guc);
>   void xe_guc_submit_disable(struct xe_guc *guc);
>   
> +void xe_guc_submit_cancel(struct xe_guc *guc);
>   int xe_guc_submit_reset_prepare(struct xe_guc *guc);
>   void xe_guc_submit_reset_wait(struct xe_guc *guc);
>   void xe_guc_submit_stop(struct xe_guc *guc);