[Intel-gfx] [PATCH] x86/gpu: add CFL to early quirks

[Lucas De Marchi]

On Wed, Dec 13, 2017 at 2:11 AM, Jani Nikula
<jani.nikula at linux.intel.com> wrote:
> On Tue, 12 Dec 2017, Lucas De Marchi <lucas.de.marchi at gmail.com> wrote:
>> On Tue, Dec 12, 2017 at 1:53 AM, Joonas Lahtinen
>> <joonas.lahtinen at linux.intel.com> wrote:
>>> + Jani, who'll continue with -fixes
>>>
>>> On Mon, 2017-12-11 at 13:50 -0800, Lucas De Marchi wrote:
>>>> On Mon, Dec 11, 2017 at 2:26 AM, Joonas Lahtinen
>>>> <joonas.lahtinen at linux.intel.com> wrote:
>>>> > On Fri, 2017-12-08 at 10:47 -0800, Lucas De Marchi wrote:
>>>> > > CFL was missing from intel_early_ids[].
>>>> > >
>>>> > > Cc: Ingo Molnar <mingo at kernel.org>
>>>> > > Cc: H. Peter Anvin <hpa at zytor.com>
>>>> > > Cc: Thomas Gleixner <tglx at linutronix.de>
>>>> > > Cc: x86 at kernel.org
>>>> > > Cc: Rodrigo Vivi <rodrigo.vivi at intel.com>
>>>> > > Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
>>>> >
>>>> > This should come with a Fixes: line to be picked up to -fixes. The IDs
>>>>
>>>> I thought this didn't deserve CC to stable since alpha support was
>>>> removed for CFL only for 4.15.
>>>
>>> I don't think system memory corruption is really acceptable even for
>>> alpha quality support :P
>>>
>>>> > have been added in smaller chunks and reworked after, so backporting
>>>> > will be required. For this level of fix, my recommendation would be to
>>>> > actively provide a cleanly applying backports to affected stable
>>>> > versions.
>>>>
>>>> Are you saying this should be proactive rather than reactive? I don't
>>>> see this mentioned on
>>>> Documentation/process/stable-kernel-rules.rst... the only thing I see
>>>> there regarding patches that don't apply
>>>> cleanly is that I may bring more patches through a tag for each version.
>>>>
>>>> If we are indeed going to cc stable I can submit a v2 with added tags.
>>>> If a patch that can be cc'ed to stable
>>>> needs to be provided we may need to improve our docs, too.
>>>
>>> That's correct. But once Cc:d stable, we can see from the GIT history
>>> that it'll bounce back because it won't apply. For this specific case
>>> that might cause system memory corruption, I'd make an exception and be
>>> proactive.
>>
>> Another option would be to cherry-pick
>> 0890540e21cf1156b4cf960a4c1c734db4e816f9 and
>> 41693fd5237397d3c61b311af0fda1f6f39297c2 so then all commits apply cleanly.
>
> There's a cc: stable annotation for dependencies like that, see stable

Yep, that's why I suggested the alternative. I think bring those
commits to stable will be
better because they will always cause conflicts on future backports.
If we have them applied
it will make future backports to apply cleanly rather than needing to
diverge more from
master.

> kernel rules. I think Greg has stated a preference for picking up
> dependencies rather than having manually backported patches.

Great. I will follow that approach for v2.

thanks
Lucas De Marchi