Re: [PATCH v6 2/3] Reapply "drm/xe/xe2: Enable Indirect Ring State support for Xe2"

[Matthew Brost <matthew.brost@intel.com>]

On Tue, Feb 24, 2026 at 09:51:46AM -0800, Julia Filipchuk wrote:
> On 2/20/2026 9:11 AM, Matthew Brost wrote:
> > On Fri, Feb 20, 2026 at 09:01:38AM -0800, Julia Filipchuk wrote:
> >> This reverts commit 03d85ab36bcbcbe9dc962fccd3f8e54d7bb93b35.
> >>
> >> Previously this was reverted due to failures on LRC context switch.
> >> Issues should be root-caused and resolved with GuC WA 14025515070. Now
> >> setting indirect ring state is available by default.
> >>
> >> Note: Will be cleared if WA not available due to older GuC version.
> >>
> > 
> > Is there a reason we need indirect ring state on Xe2? AFAIK, this is
> > only needed for multi-Q or if we did user-space submission. I’d rather
> > not have an officially supported kernel floating around on Xe2 with a
> > mis-match of indirect direct ring state enabled, since it’s a very
> > subtle delta that could surface in odd ways.
> > 
> > IMO, it’s best to leave indirect ring state off on Xe2 unless we have a
> > concrete reason to enable it.
> > 
> > Matt
> 
> Thank you for input. Changing to leave off for Xe2 for now.

I think we should merge the WA though so if we want to experiment on Xe2
for features which require indirect ring state (UMD submission) we don't
trip up on the indirect ring state Xe2 bug.

Matt