Re: [PATCH v3 12/18] drm/xe/multi_queue: Reset GT upon CGP_SYNC failure

[Matthew Brost <matthew.brost@intel.com>]

On Thu, Nov 20, 2025 at 07:51:46PM -0800, Niranjana Vishwanathapura wrote:
> If GuC doesn't response to CGP_SYNC message, trigger
> GT reset and cleanup of all the queues of the multi
> queue group.
> 
> Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>

Reviewed-by: Matthew Brost <matthew.brost@intel.com>

> ---
>  drivers/gpu/drm/xe/xe_guc_submit.c | 38 ++++++++++++++++++++++++++++++
>  1 file changed, 38 insertions(+)
> 
> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> index 87a33c06fe26..932be1e5001e 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> @@ -593,6 +593,23 @@ static void xe_guc_exec_queue_trigger_cleanup(struct xe_exec_queue *q)
>  		xe_sched_tdr_queue_imm(&q->guc->sched);
>  }
>  
> +static void xe_guc_exec_queue_group_trigger_cleanup(struct xe_exec_queue *q)
> +{
> +	struct xe_exec_queue *primary = xe_exec_queue_multi_queue_primary(q);
> +	struct xe_exec_queue_group *group = q->multi_queue.group;
> +	struct xe_exec_queue *eq;
> +
> +	xe_gt_assert(guc_to_gt(exec_queue_to_guc(q)),
> +		     xe_exec_queue_is_multi_queue(q));
> +
> +	xe_guc_exec_queue_trigger_cleanup(primary);
> +
> +	mutex_lock(&group->list_lock);
> +	list_for_each_entry(eq, &group->list, multi_queue.link)
> +		xe_guc_exec_queue_trigger_cleanup(eq);
> +	mutex_unlock(&group->list_lock);
> +}
> +
>  static void xe_guc_exec_queue_reset_trigger_cleanup(struct xe_exec_queue *q)
>  {
>  	if (xe_exec_queue_is_multi_queue(q)) {
> @@ -618,6 +635,23 @@ static void xe_guc_exec_queue_reset_trigger_cleanup(struct xe_exec_queue *q)
>  	}
>  }
>  
> +static void set_exec_queue_group_banned(struct xe_exec_queue *q)
> +{
> +	struct xe_exec_queue *primary = xe_exec_queue_multi_queue_primary(q);
> +	struct xe_exec_queue_group *group = q->multi_queue.group;
> +	struct xe_exec_queue *eq;
> +
> +	/* Ban all queues of the multi-queue group */
> +	xe_gt_assert(guc_to_gt(exec_queue_to_guc(q)),
> +		     xe_exec_queue_is_multi_queue(q));
> +	set_exec_queue_banned(primary);
> +
> +	mutex_lock(&group->list_lock);
> +	list_for_each_entry(eq, &group->list, multi_queue.link)
> +		set_exec_queue_banned(eq);
> +	mutex_unlock(&group->list_lock);
> +}
> +
>  #define parallel_read(xe_, map_, field_) \
>  	xe_map_rd_field(xe_, &map_, 0, struct guc_submit_parallel_scratch, \
>  			field_)
> @@ -675,7 +709,11 @@ static void xe_guc_exec_queue_group_cgp_sync(struct xe_guc *guc,
>  				 !READ_ONCE(group->sync_pending) ||
>  				 xe_guc_read_stopped(guc), HZ);
>  	if ((!ret && !vf_recovery(guc)) || xe_guc_read_stopped(guc)) {
> +		/* CGP_SYNC failed. Reset gt, cleanup the group */
>  		xe_gt_warn(guc_to_gt(guc), "Wait for CGP_SYNC_DONE response failed!\n");
> +		set_exec_queue_group_banned(q);
> +		xe_gt_reset_async(q->gt);
> +		xe_guc_exec_queue_group_trigger_cleanup(q);
>  		return;
>  	}
>  
> -- 
> 2.43.0
>