Re: [REGRESSION] XArray commit prevents booting with 6.0-rc1 or later

From: Linux kernel regression tracking (#update)
Date: Thu Jan 05 2023 - 03:26:43 EST


On 12.12.22 08:03, Thorsten Leemhuis wrote:
> On 28.11.22 13:18, Matthew Wilcox wrote:
>> On Sun, Nov 20, 2022 at 12:20:13AM +0100, Jorropo wrote:
>>> Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote :
>>>> On Sat, Nov 19, 2022 at 05:07:45AM +0100, Jorropo wrote:
>>>>>
>>>>> Hi, I recently tried to upgrade to linux v6.0.x but when trying to
>>>>> boot it fails with "error: out of memory" when or after loading
>>>>> initramfs (which then kpanics because the vfs root is missing).
>>>>> The latest releases I tested are v6.0.9 and v6.1-rc5 and it's broken there too.
>>>>>
>>>>> I bisected the error to this patch:
>>>>> 1dd685c414a7b9fdb3d23aca3aedae84f0b998ae "XArray: Add calls to
>>>>> might_alloc()" is the first bad commit.
>>>>> I've confirmed this is not a side effect of a poor bitsect because
>>>>> 1dd685c414a7b9fdb3d23aca3aedae84f0b998ae~1 (v5.19-rc6) works.
>>>>
>>>> That makes no sense. I can't look into this until Wednesday, but I
>>>> suggest that what you have is an intermittent failure to boot, and
>>>> the bisect has led you down the wrong path.
>>>
>>> I rebuilt both 1dd685c414a7b9fdb3d23aca3aedae84f0b998ae and
>>> the parent commit (v5.19-rc6), then tried to start each one 8 times
>>> (shuffled in a Thue morse sequence).
>>> 0 successes for 1dd685c414a7b9fdb3d23aca3aedae84f0b998ae
>>> 8 successes for v5.19-rc6
>>>
>>> This really does not look like an intermittent issue.
>>
>> OK, you convinced me. Can you boot 1dd685c414 with the command line
>> parameters "debug initcall_debug" so we get more information?
>
> Jorropo, did you ever provide the information Matthew asked for? I'm
> asking, as this looks stalled -- and I wonder why. Or was progress made
> somewhere and I just missed it?

For the record: Jorropo sent me a private mail where he states "this is
some installation issue because after some updates the previous release
that used to work does not anymore. You can close this.". Hence:

#regzbot resolve: not a regression according to the reporter

Ciao, Thorsten (wearing his 'the Linux kernel's regression tracker' hat)
--
Everything you wanna know about Linux kernel regression tracking:
https://linux-regtracking.leemhuis.info/about/#tldr
That page also explains what to do if mails like this annoy you.