Re: linux-next: build failure after merge of the amdgpu tree

[Alex Deucher]

On Mon, May 22, 2023 at 12:47 PM Nick Desaulniers
<ndesaulniers@xxxxxxxxxx> wrote:
>
> On Mon, May 22, 2023 at 9:36 AM Alex Deucher <alexdeucher@xxxxxxxxx> wrote:
> >
> > On Mon, May 22, 2023 at 12:29 PM Nick Desaulniers
> > <ndesaulniers@xxxxxxxxxx> wrote:
> > >
> > > On Mon, May 22, 2023 at 05:15:57PM +1000, Stephen Rothwell wrote:
> > > > Hi all,
> > > >
> > > > After merging the amdgpu tree, today's linux-next build (arm allmodconfig
> > > > clang-17) failed like this:
> > > >
> > > > drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c:146:54: error: format specifies type 'unsigned long long' but the argument has type 'dma_addr_t' (aka 'unsigned int') [-Werror,-Wformat]
> > > >
> > > > Caused by commit
> > > >
> > > >   d020a29b6b58 ("drm/amdgpu: Allocate GART table in RAM for AMD APU")
> > > >
> > > > Reported by the kernelci.org bot.
> > >
> > > Alex,
> > > This is the third report of linux-next being broken for clang due to the
> > > AMDGPU tree.
> > > 1. https://lore.kernel.org/lkml/20230522171557.32027acf@xxxxxxxxxxxxxxxx/
> > > 2. https://lore.kernel.org/lkml/20230522171145.38a8bd4d@xxxxxxxxxxxxxxxx/
> > > 3. https://lore.kernel.org/lkml/20230522170031.5fb87a64@xxxxxxxxxxxxxxxx/
> > >
> > > Our CI is red as a result.
> > > https://github.com/ClangBuiltLinux/continuous-integration2/actions/runs/5045716034/jobs/9053211936
> > >
> > > When will AMD start testing their kernels with Clang?
> >
> > We have clang as part of our CI system and have had it for a while.
> > I'm not sure why it didn't catch these.  Our CI clang builds are
> > currently passing.
>
> Can you verify that the driver configs are enabled for those builds?

They are definitely enabled.  It's the whole point of the builds.

>
> Looking through my CI reports, it looks like allmodconfig/allyesconfig
> is red from this tree for ARCH=x86_64 and ARCH=arm64.
> Examples:
> x86_64: https://github.com/ClangBuiltLinux/continuous-integration2/actions/runs/5045594636/jobs/9052932014
> arm64: https://github.com/ClangBuiltLinux/continuous-integration2/actions/runs/5045594636/jobs/9052930995

I looked into this quickly and I think in this particular case, it's
the result of a manual merge last week where the fixes from the
fallout didn't end up getting mirrored to my drm-next tree, and I
don't think CONFIG_WERROR is set in the config.  IIRC, we had a lot of
problems with that config option and older compilers when it was added
and I think we ultimately turned it off at the time.  We should
revisit that.  I'll talk to the CI team tomorrow.

Alex


>
> >
> > Alex
> >
> > >
> > > >
> > > > --
> > > > Cheers,
> > > > Stephen Rothwell
> > >
> > > Thanks for reporting these, Stephen.
>
>
>
> --
> Thanks,
> ~Nick Desaulniers