[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <9cd36c00-b2e0-4d57-94b1-840b084d0a3b@gmx.com>
Date: Mon, 17 Nov 2025 15:40:35 +1030
From: Qu Wenruo <quwenruo.btrfs@....com>
To: LKML <linux-kernel@...r.kernel.org>,
linux-btrfs <linux-btrfs@...r.kernel.org>
Subject: Re: Objtool segfault inside an VM, based on commit 24172e0d7990
在 2025/11/17 13:33, Qu Wenruo 写道:
>
>
> 在 2025/11/17 13:10, Qu Wenruo 写道:
>> Hi,
>>
>> Recently I'm hitting pretty frequent objtool crashes, sometimes during
>> module linking (btrfs in my case) sometimes during the vmlinux linking.
>>
>> Unfortunately I don't have a coredump for it.
>>
>> The only info so far is from dmesg (and obvious the compile failure):
>>
>> [ 625.066787] traps: objtool[46220] general protection fault
>> ip:563ab54c6eb0 sp:7ffd9c2ba7c8 error:0 in
>> objtool[19eb0,563ab54b2000+1f000]
>>
>> The involved VM has the following spec:
>>
>> - Kernel branch
>> Btrfs' development branch.
>>
>> The base commit is 24172e0d79900908cf5ebf366600616d29c9b417, around
>> v6.18-rc6.
Furthermore, if just compiling on old kernels (in my case, 6.17.8), I
haven't yet hit a linking time segfault.
So it's something in the v6.18 release cycle.
Not sure if it's the recent Zen5 related hassles.
Thanks,
Qu
>>
>> - 10 vCPU, host CPU features pass-through, x86_64
>> Host is using AMD HX370.
>>
>> Haven't yet hit the segfault on aarch64.
>>
>> - 8G vRAM
>> Thus it should be enough for most cases.
>>
>> - Toolchain versions:
>> gcc 15.2.1+r301+gf24307422d1d-1
>> binutils 2.45.1-1
>>
>> All from the latest Archlinux.
>>
>> Older gcc (15.2.1+r22+gc4e96a094636-1) and binutils
>> (2.45+r29+g2b2e51a31ec7-1) are causing the same segfault.
>>
>> Haven't yet tried LLVM.
>
> Unfortunately LLVM is hitting the same crash.
> I'm using llvm/clang/lld 21.1.5.
>
> So it doesn't look like a toolchain specific bug, but something specific
> to objtool of the kernel.
>
>
> Thanks,
> Qu
>
>>
>> - Kernel config
>> Attached.
>>
>> Sometimes tweaking the config can workaround the bug a little,
>> but not reliably.
>>
>> Is this a known problem and can I work around the segfault?
>>
>> Thanks,
>> Qu
>
>
Powered by blists - more mailing lists