[REGRESSION] AMD GPU regression in 5.16.18..5.17.4 #forregzbot

[Thorsten Leemhuis]

On 29.04.22 13:36, Thorsten Leemhuis wrote:
> TWIMC: this mail is primarily send for documentation purposes and for
> regzbot, my Linux kernel regression tracking bot. These mails usually
> contain '#forregzbot' in the subject, to make them easy to spot and filter.
> 
> #regzbot fixed-by: 78b12008f20

Landed in mainline with a different ID:

#regzbot fixed-by: 19965d8259fdabc68

> On 25.04.22 23:29, Deucher, Alexander wrote:
>> [Public]
>>
>>> -----Original Message-----
>>> From: Demi Marie Obenour <demi at invisiblethingslab.com>
>>> Sent: Sunday, April 24, 2022 7:39 PM
>>> To: Thorsten Leemhuis <regressions at leemhuis.info>; Greg KH 
>>> <gregkh at linuxfoundation.org>
>>> Cc: amd-gfx at lists.freedesktop.org; regressions at lists.linux.dev; 
>>> Deucher, Alexander <Alexander.Deucher at amd.com>; Koenig, Christian 
>>> <Christian.Koenig at amd.com>; Pan, Xinhui <Xinhui.Pan at amd.com>
>>> Subject: Re: [REGRESSION] AMD GPU regression in 5.16.18..5.17.4
>>>
>>> On Sun, Apr 24, 2022 at 11:02:43AM +0200, Thorsten Leemhuis wrote:
>>>> CCing the amdgpu maintainers
>>>>
>>>> On 24.04.22 08:12, Greg KH wrote:
>>>>> On Sat, Apr 23, 2022 at 12:06:33PM -0400, Demi Marie Obenour wrote:
>>>>>> Two Qubes OS users reported that their AMD GPU systems did not 
>>>>>> work on 5.17.4, while 5.16.18 worked fine.  Details can be found 
>>>>>> on https://github.com/QubesOS/qubes-issues/issues/7462.  The 
>>>>>> initial report listed the GPU as “Advanced Micro Devices, Inc. 
>>>>>> [AMD/ATI] Renoir [1002:1636} (rev d3) (prog-if 00 [VGA 
>>>>>> controller])” and the CPU as “AMD Ryzen 5 PRO 4650U with Radeon Graphics”.
>>>>>>
>>>>>> #regzbot introduced: v5.16.18..v5.17.4
>>>>>
>>>>> That's a big range, can you use 'git bisect' to track it down?
>>>>
>>>> FWIW, in another mail in this thread and
>>>> https://github.com/QubesOS/qubes-issues/issues/7462#issuecomment-
>>> 11076
>>>> 81126 it was clarified that the problem was introduced between 
>>>> 5.17.3 and 5.17.4, where a few amdgpu changes were applied. Maybe 
>>>> they are the reason.
>>>>
>>>> Anyway: Yes, as Gregkh said, a bisection really would help. But 
>>>> maybe the amdgfx developers might already have an idea what might be 
>>>> wrong here? The QubesOS ticket linked above has some more details.
>>>
>>> The reporter was able to bisect the regression.  I am not surprised 
>>> that this commit caused problems for Qubes OS, as dom0 in Qubes OS is 
>>> technically a guest of Xen.
>>>
>>> #regzbot ^introduced: b818a5d374542ccec73dcfe578a081574029820e
>>
>> Can you please follow up on the bug ticket:
>> https://gitlab.freedesktop.org/drm/amd/-/issues/1985
>> It will be easier to track everything there.
>>
>> Alex