Re: [PATCH v8 21/24] drm/xe: Add ULLS migration job support to GuC submission

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] drm/xe/guc: lost chained ULLS jobs when execution queue is suspended
--

commit 66f017eb1b046b1a95648616e5327a01247f461c
Author: Matthew Brost <matthew.brost@intel.com>
drm/xe: Add ULLS migration job support to GuC submission

This commit adds ULLS migration job support to the GuC submission backend. 
It reduces the max jobs on migration queues to match ULLS semaphores, skips 
writing the saved ring tail for non-first ULLS jobs, sets the ULLS semaphore 
for chained jobs, and suppresses H2G submits for non-first ULLS jobs.

> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> index a2784763e72ba..cd9bb9f9e28c8 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.c

[ ... ]

> @@ -1209,6 +1210,9 @@ static void submit_exec_queue(struct xe_exec_queue *q, struct xe_sched_job *job)
>  	if (exec_queue_suspended(q))
>  		return;
>  
> +	if (xe_sched_job_ulls_is_chained(job))
> +		xe_lrc_set_ulls_semaphore(lrc, xe_sched_job_lrc_seqno(job));
> +

[Severity: High]
Will this result in lost chained ULLS jobs if the execution queue is suspended
concurrently?

If a chained ULLS job is submitted and the queue is suspended (for instance,
due to preemption), submit_exec_queue() returns early at the
exec_queue_suspended() check immediately preceding this addition.

Because the early return completely skips the xe_lrc_set_ulls_semaphore() 
call, when the queue resumes, the hardware will expect the memory semaphore to 
be signaled. Since the signaling was skipped, could this cause the hardware to 
wait indefinitely, triggering a job timeout and queue reset?

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260925201526.1355279-1-matthew.brost@intel.com?part=21