Re: [PATCH v2] drm/xe/pm: Handle GT resume failure

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

On Thu, Jan 08, 2026 at 12:54:17PM +0100, Raag Jadav wrote:
> On Sun, Dec 21, 2025 at 08:38:24AM +0530, Vivi, Rodrigo wrote:
> > On Sat, 2025-12-20 at 13:06 +0530, Raag Jadav wrote:
> > > We've been historically ignoring GT resume failure. Since the
> > > function
> > > can return error, handle it properly.
> > > 
> > > v2: Bring up display before bailing (Matt Roper, Rodrigo)
> > 
> > I'm afraid this is not exactly what we meant.
> > 
> > 1. it is not just about display.
> > 2. display cases might depend on gt (the chicken and egg case I had
> > mentioned).
> 
> Yes, this is based on commit cb8f81c17531.
> 
> > It won't be a simple solution. You probably will need to experiment
> > with error injection cases and manual experiments before deciding
> > the proper approach...
> 
> I did two experiments.
> 
> Without GUI: I injected errors in GT resume path at mulitple places and
> still managed to land a working console after S3 resume.
> 
> With GUI: Crash and burn, but able to see resume failure on tty with
> this patch.
> 
> Hope this helps.

Yes, it does. thank you for the diligence and for the patch.

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

pushing soon to drm-xe-next...

> 
> Raag
> 
> > > Signed-off-by: Raag Jadav <raag.jadav@intel.com>
> > > ---
> > >  drivers/gpu/drm/xe/xe_pm.c | 26 ++++++++++++++++++++++----
> > >  1 file changed, 22 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
> > > index 4390ba69610d..559cf5490ac0 100644
> > > --- a/drivers/gpu/drm/xe/xe_pm.c
> > > +++ b/drivers/gpu/drm/xe/xe_pm.c
> > > @@ -260,10 +260,19 @@ int xe_pm_resume(struct xe_device *xe)
> > >  
> > >  	xe_irq_resume(xe);
> > >  
> > > -	for_each_gt(gt, xe, id)
> > > -		xe_gt_resume(gt);
> > > +	for_each_gt(gt, xe, id) {
> > > +		err = xe_gt_resume(gt);
> > > +		if (err)
> > > +			break;
> > > +	}
> > >  
> > > +	/*
> > > +	 * Try to bring up display before bailing from GT resume
> > > failure,
> > > +	 * so we don't leave the user clueless with a blank screen.
> > > +	 */
> > >  	xe_display_pm_resume(xe);
> > > +	if (err)
> > > +		goto err;
> > >  
> > >  	err = xe_bo_restore_late(xe);
> > >  	if (err)
> > > @@ -656,10 +665,19 @@ int xe_pm_runtime_resume(struct xe_device *xe)
> > >  
> > >  	xe_irq_resume(xe);
> > >  
> > > -	for_each_gt(gt, xe, id)
> > > -		xe->d3cold.allowed ? xe_gt_resume(gt) :
> > > xe_gt_runtime_resume(gt);
> > > +	for_each_gt(gt, xe, id) {
> > > +		err = xe->d3cold.allowed ? xe_gt_resume(gt) :
> > > xe_gt_runtime_resume(gt);
> > > +		if (err)
> > > +			break;
> > > +	}
> > >  
> > > +	/*
> > > +	 * Try to bring up display before bailing from GT resume
> > > failure,
> > > +	 * so we don't leave the user clueless with a blank screen.
> > > +	 */
> > >  	xe_display_pm_runtime_resume(xe);
> > > +	if (err)
> > > +		goto out;
> > >  
> > >  	if (xe->d3cold.allowed) {
> > >  		err = xe_bo_restore_late(xe);