LR compute WA

LR compute WA

[Mario Limonciello (AMD) (kernel.org)]

Hi,

We have some reports of long compute jobs on APUs hanging the system. 
This has been root caused and a workaround has been introduced in the 
mainline kernel.  I didn't CC stable on the original W/A because I 
wanted to make sure we've had enough time to test it didn't have 
unintended side effects.

I feel comfortable with the testing at this point and I think it's worth 
bringing back to any stable kernels it will apply to 6.12.y and newer. 
The commit is:

1fb710793ce2619223adffaf981b1ff13cd48f17

Thanks,

Re: LR compute WA

[Greg KH]

On Sat, Oct 04, 2025 at 01:41:29PM -0500, Mario Limonciello (AMD) (kernel.org) wrote:
> Hi,
> 
> We have some reports of long compute jobs on APUs hanging the system. This
> has been root caused and a workaround has been introduced in the mainline
> kernel.  I didn't CC stable on the original W/A because I wanted to make
> sure we've had enough time to test it didn't have unintended side effects.
> 
> I feel comfortable with the testing at this point and I think it's worth
> bringing back to any stable kernels it will apply to 6.12.y and newer. The
> commit is:
> 
> 1fb710793ce2619223adffaf981b1ff13cd48f17
 

It did not apply to 6.12.y, so if you want it there, can you provide a
working backport?

thanks,

greg k-h

Re: LR compute WA

[Mario Limonciello (AMD) (kernel.org)]

On 10/6/2025 5:04 AM, Greg KH wrote:
> On Sat, Oct 04, 2025 at 01:41:29PM -0500, Mario Limonciello (AMD) (kernel.org) wrote:
>> Hi,
>>
>> We have some reports of long compute jobs on APUs hanging the system. This
>> has been root caused and a workaround has been introduced in the mainline
>> kernel.  I didn't CC stable on the original W/A because I wanted to make
>> sure we've had enough time to test it didn't have unintended side effects.
>>
>> I feel comfortable with the testing at this point and I think it's worth
>> bringing back to any stable kernels it will apply to 6.12.y and newer. The
>> commit is:
>>
>> 1fb710793ce2619223adffaf981b1ff13cd48f17
>   
> 
> It did not apply to 6.12.y, so if you want it there, can you provide a
> working backport?
> 
> thanks,
> 
> greg k-h

Thanks, I see 6.16 and 6.17 had no problem.  I'll find the contextually 
missing patches and send out 6.12.y separately.

Re: LR compute WA

[Mario Limonciello (AMD) (kernel.org)]

On 10/6/2025 8:56 AM, Mario Limonciello (AMD) (kernel.org) wrote:
> 
> 
> On 10/6/2025 5:04 AM, Greg KH wrote:
>> On Sat, Oct 04, 2025 at 01:41:29PM -0500, Mario Limonciello (AMD) 
>> (kernel.org) wrote:
>>> Hi,
>>>
>>> We have some reports of long compute jobs on APUs hanging the system. 
>>> This
>>> has been root caused and a workaround has been introduced in the 
>>> mainline
>>> kernel.  I didn't CC stable on the original W/A because I wanted to make
>>> sure we've had enough time to test it didn't have unintended side 
>>> effects.
>>>
>>> I feel comfortable with the testing at this point and I think it's worth
>>> bringing back to any stable kernels it will apply to 6.12.y and 
>>> newer. The
>>> commit is:
>>>
>>> 1fb710793ce2619223adffaf981b1ff13cd48f17
>>
>> It did not apply to 6.12.y, so if you want it there, can you provide a
>> working backport?
>>
>> thanks,
>>
>> greg k-h
> 
> Thanks, I see 6.16 and 6.17 had no problem.  I'll find the contextually 
> missing patches and send out 6.12.y separately.

OK - here's the 3 patches needed for 6.12.y to cleanly cherry-pick:

ce4971388c79d36b3f50f607c3278dbfae6c789b
1c687c0da9efb7c627793483a8927554764e7a55
15d8c92f107c17c2e585cb4888c67873538f9722

Thanks!

Re: LR compute WA

[Greg KH]

On Mon, Oct 06, 2025 at 12:37:30PM -0500, Mario Limonciello (AMD) (kernel.org) wrote:
> 
> 
> On 10/6/2025 8:56 AM, Mario Limonciello (AMD) (kernel.org) wrote:
> > 
> > 
> > On 10/6/2025 5:04 AM, Greg KH wrote:
> > > On Sat, Oct 04, 2025 at 01:41:29PM -0500, Mario Limonciello (AMD)
> > > (kernel.org) wrote:
> > > > Hi,
> > > > 
> > > > We have some reports of long compute jobs on APUs hanging the
> > > > system. This
> > > > has been root caused and a workaround has been introduced in the
> > > > mainline
> > > > kernel.  I didn't CC stable on the original W/A because I wanted to make
> > > > sure we've had enough time to test it didn't have unintended
> > > > side effects.
> > > > 
> > > > I feel comfortable with the testing at this point and I think it's worth
> > > > bringing back to any stable kernels it will apply to 6.12.y and
> > > > newer. The
> > > > commit is:
> > > > 
> > > > 1fb710793ce2619223adffaf981b1ff13cd48f17
> > > 
> > > It did not apply to 6.12.y, so if you want it there, can you provide a
> > > working backport?
> > > 
> > > thanks,
> > > 
> > > greg k-h
> > 
> > Thanks, I see 6.16 and 6.17 had no problem.  I'll find the contextually
> > missing patches and send out 6.12.y separately.
> 
> OK - here's the 3 patches needed for 6.12.y to cleanly cherry-pick:
> 
> ce4971388c79d36b3f50f607c3278dbfae6c789b
> 1c687c0da9efb7c627793483a8927554764e7a55
> 15d8c92f107c17c2e585cb4888c67873538f9722

All now queued up, thanks.

greg k-h