Fwd: 6.4.4 breaks module-free builds of Debian kernel packages

Fwd: 6.4.4 breaks module-free builds of Debian kernel packages

[Bagas Sanjaya]

Hi,

I notice a regression report on Bugzilla [1]. Quoting from it:

> I'm on AMD64 with Debian testing (trixie), where I build my own kernels (with CONFIG_MODULES unset) using "make bindeb-pkg". The build proceeds through 99% of the process, but fails here:
> 
> Kernel: arch/x86/boot/bzImage is ready  (#2)
> make -f ./Makefile ARCH=x86     KERNELRELEASE=6.4.4-i5 intdeb-pkg
> sh ./scripts/package/builddeb
> ***
> *** The present kernel configuration has modules disabled.
> *** To use the module feature, please run "make menuconfig" etc.
> *** to enable CONFIG_MODULES.
> ***
> make[5]: *** [Makefile:1969: modules_install] Error 1
> make[4]: *** [scripts/Makefile.package:150: intdeb-pkg] Error 2
> make[3]: *** [Makefile:1657: intdeb-pkg] Error 2
> make[2]: *** [debian/rules:16: binary-arch] Error 2
> dpkg-buildpackage: error: debian/rules binary subprocess returned exit status 2
> make[1]: *** [scripts/Makefile.package:139: bindeb-pkg] Error 2
> make: *** [Makefile:1657: bindeb-pkg] Error 2
> 
> 6.3.13 contained the same error, but I "fixed" that by moving to 6.4.3.  But alas, 6.4.4 now has the same issue.
> 
> I worked around the issue by changing "exit 1" to "exit 0" in the main Makefile (at "modules module_install", per the attached patch), but I don't know if this is a true fix or something that simply happens to work for my particular configuration.

See Bugzilla for the full thread and attached patch that ignores the error.

Josh: It looks like this regression is caused by a commit of yours
(and also 1240dabe8d58b4). Would you like to take a look on it?

Anyway, I'm adding this regression to be tracked by regzbot:

#regzbot introduced: 4243afdb932677 https://bugzilla.kernel.org/show_bug.cgi?id=217689
#regzbot title: always doing modules_install breaks CONFIG_MODULES=n builds

Thanks.

[1]: https://bugzilla.kernel.org/show_bug.cgi?id=217689

-- 
An old man doll... just what I always wanted! - Clara

Re: 6.4.4 breaks module-free builds of Debian kernel packages

[Masahiro Yamada]

On Fri, Jul 21, 2023 at 12:19 PM Bagas Sanjaya <bagasdotme@gmail.com> wrote:
>
> Hi,
>
> I notice a regression report on Bugzilla [1]. Quoting from it:
>
> > I'm on AMD64 with Debian testing (trixie), where I build my own kernels (with CONFIG_MODULES unset) using "make bindeb-pkg". The build proceeds through 99% of the process, but fails here:
> >
> > Kernel: arch/x86/boot/bzImage is ready  (#2)
> > make -f ./Makefile ARCH=x86     KERNELRELEASE=6.4.4-i5 intdeb-pkg
> > sh ./scripts/package/builddeb
> > ***
> > *** The present kernel configuration has modules disabled.
> > *** To use the module feature, please run "make menuconfig" etc.
> > *** to enable CONFIG_MODULES.
> > ***
> > make[5]: *** [Makefile:1969: modules_install] Error 1
> > make[4]: *** [scripts/Makefile.package:150: intdeb-pkg] Error 2
> > make[3]: *** [Makefile:1657: intdeb-pkg] Error 2
> > make[2]: *** [debian/rules:16: binary-arch] Error 2
> > dpkg-buildpackage: error: debian/rules binary subprocess returned exit status 2
> > make[1]: *** [scripts/Makefile.package:139: bindeb-pkg] Error 2
> > make: *** [Makefile:1657: bindeb-pkg] Error 2
> >
> > 6.3.13 contained the same error, but I "fixed" that by moving to 6.4.3.  But alas, 6.4.4 now has the same issue.
> >
> > I worked around the issue by changing "exit 1" to "exit 0" in the main Makefile (at "modules module_install", per the attached patch), but I don't know if this is a true fix or something that simply happens to work for my particular configuration.
>
> See Bugzilla for the full thread and attached patch that ignores the error.
>
> Josh: It looks like this regression is caused by a commit of yours
> (and also 1240dabe8d58b4). Would you like to take a look on it?
>
> Anyway, I'm adding this regression to be tracked by regzbot:
>
> #regzbot introduced: 4243afdb932677 https://bugzilla.kernel.org/show_bug.cgi?id=217689
> #regzbot title: always doing modules_install breaks CONFIG_MODULES=n builds
>
> Thanks.
>
> [1]: https://bugzilla.kernel.org/show_bug.cgi?id=217689
>
> --
> An old man doll... just what I always wanted! - Clara




The following commit must be back-ported.



commit 8ae071fc216a25f4f797f33c56857f4dd6b4408e
Author: Masahiro Yamada <masahiroy@kernel.org>
Date:   Thu Jun 15 20:17:43 2023 +0900

    kbuild: make modules_install copy modules.builtin(.modinfo)









-- 
Best Regards
Masahiro Yamada

Re: 6.4.4 breaks module-free builds of Debian kernel packages

[Masahiro Yamada]

On Fri, Jul 21, 2023 at 12:19 PM Bagas Sanjaya <bagasdotme@gmail.com> wrote:
>
> Hi,
>
> I notice a regression report on Bugzilla [1]. Quoting from it:
>
> > I'm on AMD64 with Debian testing (trixie), where I build my own kernels (with CONFIG_MODULES unset) using "make bindeb-pkg". The build proceeds through 99% of the process, but fails here:
> >
> > Kernel: arch/x86/boot/bzImage is ready  (#2)
> > make -f ./Makefile ARCH=x86     KERNELRELEASE=6.4.4-i5 intdeb-pkg
> > sh ./scripts/package/builddeb
> > ***
> > *** The present kernel configuration has modules disabled.
> > *** To use the module feature, please run "make menuconfig" etc.
> > *** to enable CONFIG_MODULES.
> > ***
> > make[5]: *** [Makefile:1969: modules_install] Error 1
> > make[4]: *** [scripts/Makefile.package:150: intdeb-pkg] Error 2
> > make[3]: *** [Makefile:1657: intdeb-pkg] Error 2
> > make[2]: *** [debian/rules:16: binary-arch] Error 2
> > dpkg-buildpackage: error: debian/rules binary subprocess returned exit status 2
> > make[1]: *** [scripts/Makefile.package:139: bindeb-pkg] Error 2
> > make: *** [Makefile:1657: bindeb-pkg] Error 2
> >
> > 6.3.13 contained the same error, but I "fixed" that by moving to 6.4.3.  But alas, 6.4.4 now has the same issue.



I sent a back-port request.
https://lore.kernel.org/stable/CAK7LNAQNwjRYQDCD3=VoddnFmhxruzGpyppHr+2ZF3SgqDme-w@mail.gmail.com/T/#u

Hopefully, 6.4.x will be fixed in the next release.



The 6.3.x series is EOL.
So, we cannot fix it.






--
Best Regards
Masahiro Yamada

Re: Fwd: 6.4.4 breaks module-free builds of Debian kernel packages

[Randy Dunlap]

Hi Bagas,

On 7/20/23 20:18, Bagas Sanjaya wrote:
> Hi,
> 
> I notice a regression report on Bugzilla [1]. Quoting from it:
> 
>> I'm on AMD64 with Debian testing (trixie), where I build my own kernels (with CONFIG_MODULES unset) using "make bindeb-pkg". The build proceeds through 99% of the process, but fails here:
>>
>> Kernel: arch/x86/boot/bzImage is ready  (#2)
>> make -f ./Makefile ARCH=x86     KERNELRELEASE=6.4.4-i5 intdeb-pkg
>> sh ./scripts/package/builddeb
>> ***
>> *** The present kernel configuration has modules disabled.
>> *** To use the module feature, please run "make menuconfig" etc.
>> *** to enable CONFIG_MODULES.
>> ***
>> make[5]: *** [Makefile:1969: modules_install] Error 1
>> make[4]: *** [scripts/Makefile.package:150: intdeb-pkg] Error 2
>> make[3]: *** [Makefile:1657: intdeb-pkg] Error 2
>> make[2]: *** [debian/rules:16: binary-arch] Error 2
>> dpkg-buildpackage: error: debian/rules binary subprocess returned exit status 2
>> make[1]: *** [scripts/Makefile.package:139: bindeb-pkg] Error 2
>> make: *** [Makefile:1657: bindeb-pkg] Error 2
>>
>> 6.3.13 contained the same error, but I "fixed" that by moving to 6.4.3.  But alas, 6.4.4 now has the same issue.
>>
>> I worked around the issue by changing "exit 1" to "exit 0" in the main Makefile (at "modules module_install", per the attached patch), but I don't know if this is a true fix or something that simply happens to work for my particular configuration.
> 
> See Bugzilla for the full thread and attached patch that ignores the error.
> 
> Josh: It looks like this regression is caused by a commit of yours
> (and also 1240dabe8d58b4). Would you like to take a look on it?
> 
> Anyway, I'm adding this regression to be tracked by regzbot:
> 
> #regzbot introduced: 4243afdb932677 https://bugzilla.kernel.org/show_bug.cgi?id=217689
> #regzbot title: always doing modules_install breaks CONFIG_MODULES=n builds
> 
> Thanks.
> 
> [1]: https://bugzilla.kernel.org/show_bug.cgi?id=217689


Do you also take care of marking the bugzilla entries as completed/fixed/solved
etc.?

thanks.
-- 
~Randy

Re: Fwd: 6.4.4 breaks module-free builds of Debian kernel packages

[Linux regression tracking (Thorsten Leemhuis)]

On 21.07.23 06:13, Randy Dunlap wrote:
> On 7/20/23 20:18, Bagas Sanjaya wrote:
>>
>>> I'm on AMD64 with Debian testing (trixie), where I build my own kernels (with CONFIG_MODULES unset) using "make bindeb-pkg". The build proceeds through 99% of the process, but fails here:
>>>
> [...]
>>>
>>> 6.3.13 contained the same error, but I "fixed" that by moving to 6.4.3.  But alas, 6.4.4 now has the same issue.
>>>
>>> I worked around the issue by changing "exit 1" to "exit 0" in the main Makefile (at "modules module_install", per the attached patch), but I don't know if this is a true fix or something that simply happens to work for my particular configuration.
>>
>> See Bugzilla for the full thread and attached patch that ignores the error.
>>
>> Josh: It looks like this regression is caused by a commit of yours
>> (and also 1240dabe8d58b4). Would you like to take a look on it?
>>
>> Anyway, I'm adding this regression to be tracked by regzbot:
>>
>> #regzbot introduced: 4243afdb932677 https://bugzilla.kernel.org/show_bug.cgi?id=217689
>> #regzbot title: always doing modules_install breaks CONFIG_MODULES=n builds

Masahiro Yamada: thx for taking care of this and Greg for picking the
fix up.

BTW, Bagas, this apparently is a regression that only affected stable.
In that case please tell rezbot about the stable commit-id of the
change, as otherwise it will consider the problem a regression in
mainline (and there it's was never a problem or already solved [didn't
look]).

Let me fix this up and tell regzbot about the incoming fix while at it:

#regzbot introduced: 6061ac50f1e04
#regzbot fix: kbuild: make modules_install copy modules.builtin(.modinfo)

>> [1]: https://bugzilla.kernel.org/show_bug.cgi?id=217689
> 
> Do you also take care of marking the bugzilla entries as completed/fixed/solved
> etc.?

I don't known if Bagas does, but I do not and have no plans to do so.

I consider my work on looking at bugzilla for regressions a courtesy I
perform in the interest of the "no regressions" rule, as some or many of
those report otherwise will be ignored. And that's bad for our reputation.

But that is where my courtesy stops. Those that think having a bugzilla
around should take care of maintaining the state. Bugbot might soon
solve part of the problem. But I guess it won't handle this case.

Ciao, Thorsten