[git pull] drm fixes for v4.12-rc1

[Dave Airlie]

On 13 May 2017 at 04:56, Linus Torvalds <torvalds at linux-foundation.org> wrote:
> .. and here's the email repeated for the new dri-devel list, since
> apparently Dave sent the pull request to the old no-longer-working one
> that just sends annoying bounces.
>
> Dave, time to update your scripts and address book..

wierd gmail failed me.
>
>               Linus
>
> On Fri, May 12, 2017 at 11:54 AM, Linus Torvalds
> <torvalds at linux-foundation.org> wrote:
>> On Thu, May 11, 2017 at 11:00 PM, Dave Airlie <airlied at gmail.com> wrote:
>>>
>>> It also has an amdgpu fixes pull, with lots of ongoing work on Vega10
>>> which is new in this kernel and is preliminary support so may have a
>>> fair bit of movement.
>>
>> Note: I will *not* be taking these kinds of pull requests after rc1.
>>
>> If Vega10 is in such bad shape that it will need this kind of stuff
>> and isn't worth shipping without them in 4.12, I will take a *oneline*
>> that just disables it.
>>
>> So no "thousands of lines of fixes for a new driver".
>>
>> Being new to 4.12 isn't an excuse for crazy stuff after the merge
>> window. If it will need more of this kind of attention, all it means
>> is that it shouldn't have been sent to me at all in the first place.
>>
>> The drm subsystem is still on my "no more of this shit" list, so I'm
>> going to be very unforgiving of big pull requests when they aren't
>> appropriate.

It's already disabled. It probably won't be enabled in your kernel for 1-2
more releases. You can't even get Vega hardware, I can't even get Vega
hardware at the moment.

For background we've done a lot of work in the graphics community to
get vendors to release code for new hardware early so we can get it all
lined up in the kernel before the hw ships. Some support would be nice.

Vega10 is in bad shape, that's the whole point of merging it early and
turned off, so we don't just dump rebased 100 times code into your tree
on release day and find it doesn't work, instead AMD can develop other
features in their driver for Vega10 and older chips at the same time,
without constant rebasing, and fixup their internal and external driver
ABIs in an upstream kernel tree, instead of dropping a driver based on
Linux 3.10 or something.

In saying that, I did think this pull was probably a bit much, so
Alex, can you please minimise the Vega churn to -next trees unless
changes are needed to fix problems on older GPUs, also I did miss
one pull request from Alex in there somewhere, so that might have
contributed to the overly large pull.

Dave.