Re: [PATCH v9 08/10] drm/xe: Improve wedged state management
Raag Jadav <[email protected]>
| Newsgroups | org.freedesktop.lists.intel-xe |
|---|---|
| Message-ID | <[email protected]> |
On Fri, Aug 21, 2026 at 11:29:35AM +0200, Raag Jadav wrote:
> On Fri, Aug 21, 2026 at 10:26:58AM +0200, Laguna, Lukasz wrote:
> > On 7/10/2026 07:01, Raag Jadav wrote:
>
> ...
>
> > > +bool xe_device_wedged(struct xe_device *xe)
> > > +{
> > > + return atomic_read(&xe->wedged.ref);
> > > +}
> >
> > __guc_ct_send_locked() and g2h_read() return -ENOTRECOVERABLE when
> > xe_device_wedged() is true, but with the new refcount approach device may
> > only be temporarily wedged.
>
> Good catch. I was expecting this path to not be hit during FLR as we're
> tearing down exec queues and disabling interrupts, but fully agree that
> there's a whole lot of TODOs we need to revisit as a follow up. I'll add
> this one too.
I've sent a standalone fix[1] for this. However, I'm unsure about it's
correctness, so please weigh in.
Raag
[1] https://lore.kernel.org/intel-xe/[email protected]