i386: tinyconfig: perf_event.h:838:21: error: invalid output size for constraint '=q'

i386: tinyconfig: perf_event.h:838:21: error: invalid output size for constraint '=q'

[Naresh Kamboju]

Following i386 tinyconfig  clang-13 and clang-nightly failed on
stable-rc queue/5.4.

Fail (119 errors) with status message
'failure while building tuxmake target(s): default': ea3681525113
 ("net: enetc: fix ethtool counter name for PM0_TERR")
 i386 (tinyconfig) with clang-nightly
@ https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/


make --silent --keep-going --jobs=8
O=/home/tuxbuild/.cache/tuxmake/builds/current ARCH=i386
CROSS_COMPILE=i686-linux-gnu- HOSTCC=clang CC=clang
In file included from /builds/linux/arch/x86/events/amd/core.c:12:
/builds/linux/arch/x86/events/amd/../perf_event.h:838:21: error:
invalid output size for constraint '=q'
        u64 disable_mask = __this_cpu_read(cpu_hw_events.perf_ctr_virt_mask);

build link,
https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/

Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>


--
Linaro LKFT
https://lkft.linaro.org                           ^

Re: i386: tinyconfig: perf_event.h:838:21: error: invalid output size for constraint '=q'

[Nathan Chancellor]

Hi Naresh,

On Sun, Oct 24, 2021 at 08:05:01AM +0530, Naresh Kamboju wrote:
> Following i386 tinyconfig  clang-13 and clang-nightly failed on
> stable-rc queue/5.4.
> 
> Fail (119 errors) with status message
> 'failure while building tuxmake target(s): default': ea3681525113
>  ("net: enetc: fix ethtool counter name for PM0_TERR")
>  i386 (tinyconfig) with clang-nightly
> @ https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/
> 
> 
> make --silent --keep-going --jobs=8
> O=/home/tuxbuild/.cache/tuxmake/builds/current ARCH=i386
> CROSS_COMPILE=i686-linux-gnu- HOSTCC=clang CC=clang
> In file included from /builds/linux/arch/x86/events/amd/core.c:12:
> /builds/linux/arch/x86/events/amd/../perf_event.h:838:21: error:
> invalid output size for constraint '=q'
>         u64 disable_mask = __this_cpu_read(cpu_hw_events.perf_ctr_virt_mask);
> 
> build link,
> https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/
> 
> Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>

I am surprised this is being reported as a new issue for 5.4, as it
should have always had this error [1]. We did not fix this until 5.9
[2], meaning that 5.10 and beyond is fine. The series that did fix it
was rather long so I am not sure it is suitable for stable. My
recommendation is to disable i386 testing for 5.4 and earlier with
clang.

[1]: https://github.com/ClangBuiltLinux/linux/issues/194
[2]: https://git.kernel.org/linus/ba77c568f3160657a5f7905289c07d18c2dfde78

Cheers,
Nathan

Re: i386: tinyconfig: perf_event.h:838:21: error: invalid output size for constraint '=q'

[Naresh Kamboju]

On Mon, 25 Oct 2021 at 20:11, Nathan Chancellor <nathan@kernel.org> wrote:
>
> Hi Naresh,
>
> On Sun, Oct 24, 2021 at 08:05:01AM +0530, Naresh Kamboju wrote:
> > Following i386 tinyconfig  clang-13 and clang-nightly failed on
> > stable-rc queue/5.4.
> >
> > Fail (119 errors) with status message
> > 'failure while building tuxmake target(s): default': ea3681525113
> >  ("net: enetc: fix ethtool counter name for PM0_TERR")
> >  i386 (tinyconfig) with clang-nightly
> > @ https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/
> >
> >
> > make --silent --keep-going --jobs=8
> > O=/home/tuxbuild/.cache/tuxmake/builds/current ARCH=i386
> > CROSS_COMPILE=i686-linux-gnu- HOSTCC=clang CC=clang
> > In file included from /builds/linux/arch/x86/events/amd/core.c:12:
> > /builds/linux/arch/x86/events/amd/../perf_event.h:838:21: error:
> > invalid output size for constraint '=q'
> >         u64 disable_mask = __this_cpu_read(cpu_hw_events.perf_ctr_virt_mask);
> >
> > build link,
> > https://builds.tuxbuild.com/1zvtvNS4eyYkOMiXtFgR7n1k0Yc/
> >
> > Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
>
> I am surprised this is being reported as a new issue for 5.4, as it
> should have always had this error [1]. We did not fix this until 5.9
> [2], meaning that 5.10 and beyond is fine. The series that did fix it
> was rather long so I am not sure it is suitable for stable. My
> recommendation is to disable i386 testing for 5.4 and earlier with
> clang.

We will disable i386 clang builds on LTS 5.4.
Thank you.

- Naresh