Message-ID: <ca278e55-1691-45a7-9c67-b810eacb916d@roeck-us.net>
Date: Fri, 5 Dec 2025 10:31:08 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: Nathan Chancellor <nathan@...nel.org>,
 Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc: Naresh Kamboju <naresh.kamboju@...aro.org>, stable@...r.kernel.org,
 patches@...ts.linux.dev, linux-kernel@...r.kernel.org,
 torvalds@...ux-foundation.org, akpm@...ux-foundation.org, shuah@...nel.org,
 patches@...nelci.org, lkft-triage@...ts.linaro.org, pavel@...x.de,
 jonathanh@...dia.com, f.fainelli@...il.com, sudipm.mukherjee@...il.com,
 rwarsow@....de, conor@...nel.org, hargar@...rosoft.com, broonie@...nel.org,
 achill@...ill.org, sr@...dewatkins.com,
 Nick Desaulniers <nick.desaulniers+lkml@...il.com>,
 Arnd Bergmann <arnd@...db.de>, Dan Carpenter <dan.carpenter@...aro.org>,
 Anders Roxell <anders.roxell@...aro.org>,
 Ben Copeland <benjamin.copeland@...aro.org>
Subject: Re: [PATCH 5.15 000/387] 5.15.197-rc2 review

On 12/5/25 09:39, Nathan Chancellor wrote:
> On Fri, Dec 05, 2025 at 05:33:19PM +0100, Greg Kroah-Hartman wrote:
>> On Fri, Dec 05, 2025 at 12:59:37PM +0530, Naresh Kamboju wrote:
>>> On Thu, 4 Dec 2025 at 22:14, Greg Kroah-Hartman
>>> <gregkh@...uxfoundation.org> wrote:
>>>>
>>>> This is the start of the stable review cycle for the 5.15.197 release.
>>>> There are 387 patches in this series, all will be posted as a response
>>>> to this one.  If anyone has any issues with these being applied, please
>>>> let me know.
>>>>
>>>> Responses should be made by Sat, 06 Dec 2025 16:37:24 +0000.
>>>> Anything received after that time might be too late.
>>>>
>>>> The whole patch series can be found in one patch at:
>>>>          https://www.kernel.org/pub/linux/kernel/v5.x/stable-review/patch-5.15.197-rc2.gz
>>>> or in the git tree and branch at:
>>>>          git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git linux-5.15.y
>>>> and the diffstat can be found below.
>>>>
>>>> thanks,
>>>>
>>>> greg k-h
>>>
>>>
>>> The powerpc allnoconfig failed with gcc-8 but passed with gcc-12.
>>>
>>> Build regression: powerpc: allnoconfig: gcc-8: Inconsistent kallsyms data
>>>
>>> Reported-by: Linux Kernel Functional Testing <lkft@...aro.org>
>>>
>>> ### Build error Powerpc
>>> Inconsistent kallsyms data
>>> Try make KALLSYMS_EXTRA_PASS=1 as a workaround
>>> make[1]: *** [Makefile:1244: vmlinux] Error 1
>>>
>>> ### Commit pointing to,
>>>    Makefile.compiler: replace cc-ifversion with compiler-specific macros
>>>    commit 88b61e3bff93f99712718db785b4aa0c1165f35c upstream.
>>>
>>> ### Build
>>>   - https://storage.tuxsuite.com/public/linaro/lkft/builds/36OCnVeYGpKUCXtxVdz6gezHjcQ/
>>>   - https://storage.tuxsuite.com/public/linaro/lkft/builds/36OCnVeYGpKUCXtxVdz6gezHjcQ/config
>>>
>>> ### Steps to reproduce
>>>   - tuxmake --runtime podman --target-arch powerpc --toolchain gcc-8
>>> --kconfig allnoconfig
>>
>> Odd, this works on 5.10 ok?  What is different about 5.15 that keeps
>> this from working?
> 
> This issue does reproduce for me locally but only with gcc-8 from TuxMake (i.e., a
> version from Debian), not with GCC 8.5.0 from kernel.org.
> 
>    $ git show -s --pretty=kernel
>    869807d760ee ("libbpf: Fix invalid return address register in s390")
> 
>    $ tuxmake -r podman -a powerpc -t gcc-8 -k allnoconfig default
>    ...
>    Inconsistent kallsyms data
>    Try make KALLSYMS_EXTRA_PASS=1 as a workaround
>    ...
> 
> However, reverting the backport of 88b61e3bff93 ("Makefile.compiler:
> replace cc-ifversion with compiler-specific macros") does not resolve
> the issue for me:
> 
>    $ git revert --no-edit c8dad7eb1e6221e0363ee468dc46700bfbad6dd2
>    [detached HEAD 1669da6455e4] Revert "Makefile.compiler: replace cc-ifversion with compiler-specific macros"
>     Date: Fri Dec 5 10:30:10 2025 -0700
>     13 files changed, 30 insertions(+), 37 deletions(-)
> 
>    $ tuxmake -r podman -a powerpc -t gcc-8 -k allnoconfig default
>    ...
>    Inconsistent kallsyms data
>    Try make KALLSYMS_EXTRA_PASS=1 as a workaround
>    ...
> 
> My bisect landed on the backport of 19de03b312d6 ("block: make
> REQ_OP_ZONE_OPEN a write operation") and reverting that actually
> resolves the error.
> 
>    $ git bisect log
>    # bad: [869807d760eeef0e0132eed3f1be6be16d084401] libbpf: Fix invalid return address register in s390
>    # good: [cc5ec87693063acebb60f587e8a019ba9b94ae0e] Linux 5.15.196
>    git bisect start '869807d760eeef0e0132eed3f1be6be16d084401' 'cc5ec87693063acebb60f587e8a019ba9b94ae0e'
>    # bad: [ffd15ced026694355243064968a3b84269e0ee09] ARM: at91: pm: save and restore ACR during PLL disable/enable
>    git bisect bad ffd15ced026694355243064968a3b84269e0ee09
>    # bad: [1ba52b54e5f45e982d61b4eeabbf28d78a5f9e75] drm/amdkfd: return -ENOTTY for unsupported IOCTLs
>    git bisect bad 1ba52b54e5f45e982d61b4eeabbf28d78a5f9e75
>    # good: [8966a057d07281a5e65eaa3225cb072713921b25] Revert "docs/process/howto: Replace C89 with C11"
>    git bisect good 8966a057d07281a5e65eaa3225cb072713921b25
>    # bad: [27df52e05f7dd606e18e869cb4f52ce1fbfe699f] tee: allow a driver to allocate a tee_device without a pool
>    git bisect bad 27df52e05f7dd606e18e869cb4f52ce1fbfe699f
>    # bad: [2b426fa49ef76bbe04671b3a861352b924a01b96] memstick: Add timeout to prevent indefinite waiting
>    git bisect bad 2b426fa49ef76bbe04671b3a861352b924a01b96
>    # bad: [c126c39dc662661531c96acbbf5fc129ed7f535a] soc: qcom: smem: Fix endian-unaware access of num_entries
>    git bisect bad c126c39dc662661531c96acbbf5fc129ed7f535a
>    # good: [c142f0d16766ab7189af4b9128a2c81c56f7a01f] drm/sysfb: Do not dereference NULL pointer in plane reset
>    git bisect good c142f0d16766ab7189af4b9128a2c81c56f7a01f
>    # bad: [02dc541fc61c3e2dabc3574fe46a19f554ea5d8c] soc: aspeed: socinfo: Add AST27xx silicon IDs
>    git bisect bad 02dc541fc61c3e2dabc3574fe46a19f554ea5d8c
>    # bad: [23e0fecb7be5010e96b2948490799ef59ac4bea6] block: make REQ_OP_ZONE_OPEN a write operation
>    git bisect bad 23e0fecb7be5010e96b2948490799ef59ac4bea6
>    # first bad commit: [23e0fecb7be5010e96b2948490799ef59ac4bea6] block: make REQ_OP_ZONE_OPEN a write operation
> 
>    $ git revert --no-edit 23e0fecb7be5010e96b2948490799ef59ac4bea6
>    [detached HEAD ddf3a34910b0] Revert "block: make REQ_OP_ZONE_OPEN a write operation"
>     Date: Fri Dec 5 10:26:26 2025 -0700
>      1 file changed, 5 insertions(+), 5 deletions(-)
> 
>    $ tuxmake -r podman -a powerpc -t gcc-8 -k allnoconfig default
>    ...
>    I: config: PASS in 0:00:01.410526
>    I: default: PASS in 0:00:04.933923
>    ...
> 
> No idea why that commit would cause such an issue... but it does not
> appear to be 88b61e3bff93 so I guess I am off the hook ;)
> 

It doesn't. Problem is that code changes causes symbol addresses to shift,
which causes the output of the kallsyms calculations to change. And each
round of kallsyms data calculation changes it again. Dropping a commit
will solve the problem for one compiler version but create it for others.
There is nothing that can be done about it but to make kallsyms calculation
more robust, but as I said in my other reply my attempt to do that was rejected.

Guenter

Powered by blists - more mailing lists