Re: [PATCH v7 2/7] drm/xe/pm: introduce PM PME support

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Unbalanced GT suspend in error path
--

commit 5ef22270a5fbbcf3b5f920591e425278fe8a87c6
Author: Vinod Govindapillai <vinod.govindapillai@intel.com>

drm/xe/pm: introduce PM PME support

Introduce PME support for PME capable devices. Whether device is
PME capable is assessed during PCI probe routine. And the whether
PME is enabled for a specific context is assessed during the PM
runtime suspend call if the device is PME capable.

> diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
> index f517bf453b54c..ca5665d015e98 100644
> --- a/drivers/gpu/drm/xe/xe_pm.c
> +++ b/drivers/gpu/drm/xe/xe_pm.c
[ ... ]
> @@ -650,6 +663,7 @@ int xe_pm_runtime_suspend(struct xe_device *xe)
>  	return 0;
>  
>  out_resume:
> +	xe_pm_update_pme_enabled(xe, false);
>  	xe_display_pm_runtime_resume(xe);
>  	xe_pxp_pm_resume(xe->pxp);
>  out:

[Severity: High]
This isn't a bug introduced by this patch, but does xe_pm_runtime_suspend()
in drivers/gpu/drm/xe/xe_pm.c leave GTs in a suspended state if a failure
occurs during the suspension loop?

When xe_pm_runtime_suspend() loops over the GTs, if a suspend fails on any
GT after the first one, it jumps to out_resume:

	for_each_gt(gt, xe, id) {
		err = xe->d3cold.allowed ? xe_gt_suspend(gt) : xe_gt_runtime_suspend(gt);
		if (err)
			goto out_resume;
	}

The out_resume error path handles resuming the display and PXP, but does
not unwind the GT suspends that already succeeded before the failure.

Since the suspend function returns an error, the runtime PM core would
consider the device active, leaving it in a partially suspended state.

Could this cause hardware hangs or crashes on subsequent accesses to the
successfully suspended GTs?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260914204034.309566-1-vinod.govindapillai@intel.com?part=2