[PATCH] drm/xe: Do not wedge device on killed exec queues

Rodrigo Vivi rodrigo.vivi at intel.com
Tue Jun 24 20:54:18 UTC 2025


On Tue, Jun 24, 2025 at 10:41:03AM -0700, Matthew Brost wrote:
> When a user closes an exec queue or interrupts an app with Ctrl-C,
> this does not warrant wedging the device in mode 2.
> 
> Avoid this by skipping the wedge check for killed exec queues in
> the TDR and LR exec queue cleanup worker.
> 
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>

Reviewed-by: Rodrigo Vivi <rodrigo.vivi at intel.com>

> ---
>  drivers/gpu/drm/xe/xe_guc_submit.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> index df7a5a4eec74..72477ccc5c5e 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> @@ -908,12 +908,13 @@ static void xe_guc_exec_queue_lr_cleanup(struct work_struct *w)
>  	struct xe_exec_queue *q = ge->q;
>  	struct xe_guc *guc = exec_queue_to_guc(q);
>  	struct xe_gpu_scheduler *sched = &ge->sched;
> -	bool wedged;
> +	bool wedged = false;
>  
>  	xe_gt_assert(guc_to_gt(guc), xe_exec_queue_is_lr(q));
>  	trace_xe_exec_queue_lr_cleanup(q);
>  
> -	wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
> +	if (!exec_queue_killed(q))
> +		wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
>  
>  	/* Kill the run_job / process_msg entry points */
>  	xe_sched_submission_stop(sched);
> @@ -1084,7 +1085,7 @@ guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
>  	int err = -ETIME;
>  	pid_t pid = -1;
>  	int i = 0;
> -	bool wedged, skip_timeout_check;
> +	bool wedged = false, skip_timeout_check;
>  
>  	/*
>  	 * TDR has fired before free job worker. Common if exec queue
> @@ -1130,7 +1131,8 @@ guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
>  	 * doesn't work for SRIOV. For now assuming timeouts in wedged mode are
>  	 * genuine timeouts.
>  	 */
> -	wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
> +	if (!exec_queue_killed(q))
> +		wedged = guc_submit_hint_wedged(exec_queue_to_guc(q));
>  
>  	/* Engine state now stable, disable scheduling to check timestamp */
>  	if (!wedged && exec_queue_registered(q)) {
> -- 
> 2.34.1
> 


More information about the Intel-xe mailing list