[Intel-gfx] -stable regression in Intel graphics, introduced in Linux 5.10.9

[Hans de Goede]

Hi,

On 2/28/21 4:14 PM, Greg KH wrote:
> On Sun, Feb 28, 2021 at 03:29:07PM +0100, Diego Calleja wrote:
>> Hi,
>>
>> There is a regression in Linux 5.10.9 that does not happen in 5.10.8. It is still there as
>> of 5.11.1
> 
> Is this the same issue reported here:
> 	https://lore.kernel.org/r/f1070486-891a-8ec0-0390-b9aeb03178ce@redhat.com
> ?
> 
> If so, is this a problem in 5.11 as well?

I see in the original email:
https://lore.kernel.org/stable/3423617.kz1aARBMGD@archlinux/

That Diego is using the iGPU part of a Haswell CPU, so yes this is almost
certainly the same issue.

Diego as I already mentioned to another arch user, it would be good if the
arch kernel-maintainers can pick-up these 3 commits from the drm-intel tree
as downstream patches for now:

e627d5923cae ("drm/i915/gt: One more flush for Baytrail clear residuals")
d30bbd62b1bf ("drm/i915/gt: Flush before changing register state")
1914911f4aa0 ("drm/i915/gt: Correct surface base address for renderclear")

We (Fedora) have added these as downstream patches for now and we have
multiple reports that these resolve the problem.

Chris, can you please send the 2nd and 3th commit of the above list on
their way to Linus ASAP, so that Greg can add them to the stable series?

ATM only the 1st commit is in Linus tree (unless the others have landed
with different hashes?)

Regards,

Hans