Re: [REGRESSION] PCI/AER: MacBookPro16,1 powers off ~20 s after boot

[Aditya Garg <aditya.garg@linux.dev>]

On 7 October 2026 3:59:17 pm IST, Thorsten Leemhuis <regressions@leemhuis.info> wrote:
>On 10/6/26 23:19, Lukas Wunner wrote:
>> On Tue, Oct 06, 2026 at 12:37:18PM -0400, Alex Deucher wrote:
>>> On Mon, Oct 5, 2026 at 7:19 PM Bjorn Helgaas <helgaas@kernel.org> wrote:
>>>> Alex reported something similar at
>>>> https://bugzilla.kernel.org/show_bug.cgi?id=222095.
>>>
>>> Lukas' patches fixes some boards, but unfortunately, I'm still seeing
>>> failures on others.  Reverting the patch fixes those failures.  I've
>>> updated the ticket.
>> 
>> I've added an analysis of the dmesg output as well as a patch.
>> It looks like the BIOS is raising a false alarm in that it signals
>> a Fatal Error even though the Uncorrectable Error Status register is
>> blank.  The patch causes such events to be reported, but otherwise
>> ignored (i.e. no recovery is attempted for a false alarm).
>> 
>> Newly enabled functionality like Advisory Non-Fatal Errors causes
>> BIOS code paths to be exercised which likely haven't been tested
>> by firmware validation folks, hence the fallout.  However I believe
>> ultimately we *want* to see those errors being reported because they
>> may point to BIOS issues which flew under the radar so far.
>Well, not disagreeing, but given that we have two regression reports
>(one speaking about multiple machines being broken) and we are close to
>-rc7 I wonder:
>
>Is this something we should quickly revert or disable for now and retry
>again later? Maybe multiple mainline devel cycles if needed?

I think the patch needs vigorous testing on a variety of hardware before being mainlined. For now a revert seems to be a good solution to ensure we don't have a broken 7.3 for multiple devices. Also, we are close a new LTS kernel, it's really not a good thing to have a broken one.