[PATCH v4 3/3] drm/xe/bmg: Update Wa_14022085890

[Lucas De Marchi]

On Thu, Jun 12, 2025 at 08:23:04PM +0000, Stuart Summers wrote:
>On Thu, 2025-06-12 at 14:31 -0500, Lucas De Marchi wrote:
>> On Thu, Jun 12, 2025 at 07:15:14PM +0000, Stuart Summers wrote:
>> > > diff --git a/drivers/gpu/drm/xe/xe_wa_oob.rules
>> > > b/drivers/gpu/drm/xe/xe_wa_oob.rules
>> > > index 425cb401c2762..38b589fe9198d 100644
>> > > --- a/drivers/gpu/drm/xe/xe_wa_oob.rules
>> > > +++ b/drivers/gpu/drm/xe/xe_wa_oob.rules
>> > > @@ -59,3 +59,7 @@ no_media_l3   MEDIA_VERSION(3000)
>> > >                 MEDIA_VERSION_RANGE(1301, 3000)
>> > >  16026508708    GRAPHICS_VERSION_RANGE(1200, 3001)
>> > >                 MEDIA_VERSION_RANGE(1300, 3000)
>> > > +
>> > > +# SoC workaround - should apply to all platforms using this
>> > > +# graphics version in tile->primary_gt
>> >
>> > If this is an SoC specific workaround that spreads across GTs,
>> > shouldn't we be using the device ID rather than the GMD ID to
>> > determine
>> > the workaround application?
>>
>> devid would be a nightmare to maintain... imagine each time we add a
>> devid we need to add checks in multiple places in the codebase.
>> We should add support for SoC WAs properly and then migrate this WA
>> there. No need to do that now though as that would mean a much harder
>> path to linux-stable and no real benefit.
>>
>> The way it is now, we eventually remove it from here and add to e.g.
>> drivers/gpu/drm/xe/xe_soc_wa_oob.rules. Or even just keep it in the
>> same
>> place and do
>>
>> -14022085890    GRAPHICS_VERSION(2001)
>> +14022085890    SOC_VERSION(1234)
>
>Ok and the 2001 from bspec does seem specific to the skus right now
>that have this soc. It would be nice to have a way to tell that this
>needs to be reviewed if we get something else with the same GT IDs but
>different SoC. I know this says SoC workaround already, but is that
>something we can expand to something like:
>SoC workaround - currently applies to all platforms with the following
>primary GT GMDID.
>
>Otherwise the rest of the changes look good to me:
>Reviewed-by: Stuart Summers <stuart.summers at intel.com>

I reworded it like you mentioned and pushed to drm-xe-next.

[1/3] drm/xe/guc: Ignore GuC CT errors when wedged
       commit: fa424387379650cdc0ce42c2f6e76e020b4c04d1
[2/3] drm/xe/bmg: Update Wa_16023588340
       commit: 6ab42fa03d4c88a0ddf5e56e62794853b198e7bf
[3/3] drm/xe/bmg: Update Wa_14022085890
       commit: bdde16c9ac5cb56ad2ee19792222fa1853577af7

Thanks
Lucas De Marchi

>
>Comment rework isn't a blocker for me though. Feel free to keep it as-
>is, we just need to remember to review this (anything with "soc
>workaround") when we do that rework you mentioned.
>
>Thanks,
>Stuart
>
>>
>> Lucas De Marchi
>