NAK: [PATCH 0/1] [F linux-aws-5.11] c4.large boot regression in upstream stable patchset 2022-01-19

Krzysztof Kozlowski krzysztof.kozlowski at canonical.com
Thu Feb 10 14:02:58 UTC 2022


On 10/02/2022 14:57, Tim Gardner wrote:
> 
> 
> On 2/10/22 6:43 AM, Krzysztof Kozlowski wrote:
>> On 05/02/2022 19:00, Tim Gardner wrote:
>>> SRU Justification
>>>
>>> [Impact]
>>>
>>> Upstream stable patchset 2022-01-19 as applied to linux-aws-5.11 5.11.0-1029.32_20.04.1 causes a boot regression.
>>>
>>> [Fix]
>>>
>>> A bisect narrowed down the offending commit. Reverting on top of Ubuntu-aws-5.11-5.11.0-1029.32_20.04.1
>>> solves the issue.
>>>
>>> Revert commit caefebc4b66f841612504c9bb3b37ce6a326d994 ("PCI/MSI: Mask MSI-X vectors only on success")
>>>
>>> [Test Case]
>>>
>>> Install a kernel in a Focal/aws c4.large instance. Check for boot success.
>>>
>>> [Where things could go wrong]
>>>
>>> Other instance types could fail without this patch, though it is unlikely since
>>> they worked before this stable update.
>>
>> Initial report from Andrea went unanswered:
>> https://lore.kernel.org/all/4264ac63-6072-5a43-2f5e-79d899cf0882@denx.de/#t
>>
>> but we should not revert-it and leave it like this.
>>
>> First, the revert should be UBUNTU SAUCE.
>>
>> Second, how do we follow up on this? Do we keep the revert even though
>> upstream doesn't revert?
>>
>>
>> Best regards,
>> Krzysztof
> 
> I didn't realize Andrea had issues with this commit in a 5.15 kernel. I 
> guess I should have asked, though I did do some googling to see if 
> anyone else had encountered this problem. Knowing would have saved a day 
> of bisecting. Bisecting a derived kernel is kinda of a PITA.
> 
> I'm fine with making this a SAUCE patch. I don't think its that 
> important wrt 5.11-hwe how we track upstream progress because all of the 
> derived kernels are EOL (or very close).

Hm, but it might bite us also on v5.13 and v5.15 AWS kernels. I
literally don't know how we usually handle such cases - something buggy,
but upstream is lazy to investigate. The reverted commit might not be
the real cause, but it just brings other issue to the light.

To be clear - my NAK was more about the commit subject (sauce), not the
approach. Approach I would be happy to discuss. :)


Best regards,
Krzysztof



More information about the kernel-team mailing list