Re: [PATCH 1/3] drm/xe: fix job timeout recovery for unstarted jobs and kernel queues

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Thu, Jun 04, 2026 at 12:16:38PM +0530, Yadav, Sanjay Kumar wrote:
> 
> On 03-06-2026 20:31, Rodrigo Vivi wrote:
> > Jobs that GuC never scheduled were silently errored out instead of
> > triggering a GT reset. Kernel jobs that exhaust all recovery attempts
> > should wedge the device rather than silently fail, and userspace VM bind
> > queues should stay permanently banned rather than being reset and retried.
> > 
> > The queue is banned early in the timeout handler to signal the G2H
> > scheduling-done handler so it wakes the disable-scheduling waiter; without
> > it the waiter sleeps the full 5s timeout. For kernel queues the ban is
> > cleared before rearming so that guc_exec_queue_start() can resubmit jobs
> > after the GT reset — a banned queue would block resubmission and cause an
> > infinite TDR loop.
> > 
> > Cc: Matthew Auld <matthew.auld@intel.com>
> > Cc: Sanjay Yadav <sanjay.kumar.yadav@intel.com>
> > Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
> > Assisted-by: GitHub-Copilot:claude-sonnet-4.6
> > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Tested-by: Sanjay Yadav <sanjay.kumar.yadav@intel.com>
> LGTM
> Tested with Reproducer IGT and xe_evict - No IGT stuck recovery works
> correctly and subsequent migration tests pass.

Thank you. Since I changed the code a bit on the v2, I didn't keep your
tested by.
Could you please check the v2 as well?
https://lore.kernel.org/intel-xe/20260609144412.244678-3-rodrigo.vivi@intel.com/T/#t

> > ---
> >   drivers/gpu/drm/xe/xe_guc_submit.c | 31 +++++++++++++++++++++---------
> >   1 file changed, 22 insertions(+), 9 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> > index ab501513d806..bbccba367626 100644
> > --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> > @@ -158,6 +158,11 @@ static void set_exec_queue_banned(struct xe_exec_queue *q)
> >   	atomic_or(EXEC_QUEUE_STATE_BANNED, &q->guc->state);
> >   }
> > +static void clear_exec_queue_banned(struct xe_exec_queue *q)
> > +{
> > +	atomic_andnot(EXEC_QUEUE_STATE_BANNED, &q->guc->state);
> > +}
> > +
> >   static bool exec_queue_suspended(struct xe_exec_queue *q)
> >   {
> >   	return atomic_read(&q->guc->state) & EXEC_QUEUE_STATE_SUSPENDED;
> > @@ -1376,7 +1381,8 @@ static bool check_timeout(struct xe_exec_queue *q, struct xe_sched_job *job)
> >   			   xe_sched_job_seqno(job), xe_sched_job_lrc_seqno(job),
> >   			   q->guc->id);
> > -		return xe_sched_invalidate_job(job, 2);
> > +		/* GuC never scheduled this job - let the caller trigger a GT reset. */
> > +		return true;
> >   	}
> >   	ctx_timestamp = lower_32_bits(xe_lrc_timestamp(q->lrc[0]));
> > @@ -1622,19 +1628,26 @@ guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
> >   			       q->guc->id, q->flags);
> >   	/*
> > -	 * Kernel jobs should never fail, nor should VM jobs if they do
> > -	 * somethings has gone wrong and the GT needs a reset
> > +	 * Kernel jobs should never fail permanently. Attempt GT reset and
> > +	 * resubmit; if karma is exhausted the hardware is unrecoverable so
> > +	 * wedge the device.
> > +	 *
> > +	 * Userspace VM bind queues are banned permanently on timeout
> > +.	 * No reset is attempted, the ban already
> > +	 * signals the G2H handler, and the queue stays banned so the job
> > +	 * errors out cleanly.
> >   	 */
> > -	xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_KERNEL,
> > -		   "Kernel-submitted job timed out\n");
> > -	xe_gt_WARN(q->gt, q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q),
> > -		   "VM job timed out on non-killed execqueue\n");
> > -	if (!wedged && (q->flags & EXEC_QUEUE_FLAG_KERNEL ||
> > -			(q->flags & EXEC_QUEUE_FLAG_VM && !exec_queue_killed(q)))) {
> > +	if (!wedged && q->flags & EXEC_QUEUE_FLAG_KERNEL) {
> >   		if (!xe_sched_invalidate_job(job, 2)) {
> > +			clear_exec_queue_banned(q);
> >   			xe_gt_reset_async(q->gt);
> >   			goto rearm;
> >   		}
> > +		xe_gt_WARN(q->gt, true, "Kernel-submitted job timed out\n");
> > +		xe_device_declare_wedged(gt_to_xe(q->gt));
> > +	} else if (!wedged && q->flags & EXEC_QUEUE_FLAG_VM &&
> > +		   !exec_queue_killed(q)) {
> > +		xe_gt_WARN(q->gt, true, "VM job timed out on non-killed execqueue\n");
> >   	}
> >   	/* Mark all outstanding jobs as bad, thus completing them */