Re: clang: Powerpc: clang-nightly-maple_defconfig — FAIL

[Nathan Chancellor <nathan@kernel.org>]

Hi Naresh,

On Tue, Jun 13, 2023 at 12:10:30AM +0530, Naresh Kamboju wrote:
> [Please ignore if it is already reported]
> 
> Following two builds failed on stable-rc 6.1.34-rc1.
> 
>   - Powerpc: clang-nightly-maple_defconfig — FAIL
>   - Powerpc: clang-nightly-cell_defconfig — FAIL
> 
> Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
> 
> make --silent --keep-going --jobs=8
> O=/home/tuxbuild/.cache/tuxmake/builds/1/build ARCH=powerpc
> CROSS_COMPILE=powerpc64le-linux-gnu- HOSTCC=clang CC=clang LLVM=1
> LLVM_IAS=0 LD=powerpc64le-linux-gnu-ld
> 
> arch/powerpc/lib/copypage_power7.S: Assembler messages:
> arch/powerpc/lib/copypage_power7.S:34: Error: junk at end of line: `0b01000'
> arch/powerpc/lib/copypage_power7.S:35: Error: junk at end of line: `0b01010'
> arch/powerpc/lib/copypage_power7.S:37: Error: junk at end of line: `0b01000'
> arch/powerpc/lib/copypage_power7.S:38: Error: junk at end of line: `0b01010'
> arch/powerpc/lib/copypage_power7.S:40: Error: junk at end of line: `0b01010'
> clang: error: assembler command failed with exit code 1 (use -v to see
> invocation)
> make[4]: *** [scripts/Makefile.build:382:
> arch/powerpc/lib/copypage_power7.o] Error 1
> make[4]: Target 'arch/powerpc/lib/' not remade because of errors.
> make[3]: *** [scripts/Makefile.build:500: arch/powerpc/lib] Error 2
> arch/powerpc/kernel/exceptions-64s.S: Assembler messages:
> arch/powerpc/kernel/exceptions-64s.S:2959: Error: junk at end of line: `0b01010'
> arch/powerpc/kernel/exceptions-64s.S:2979: Error: junk at end of line: `0b01010'
> arch/powerpc/kernel/exceptions-64s.S:2994: Error: junk at end of line: `0b01010'
> arch/powerpc/kernel/exceptions-64s.S:3012: Error: junk at end of line: `0b01010'
> arch/powerpc/kernel/exceptions-64s.S:3032: Error: junk at end of line: `0b01010'
> arch/powerpc/kernel/exceptions-64s.S:3079: Error: junk at end of line: `0b01010'
> clang: error: assembler command failed with exit code 1 (use -v to see
> invocation)
> make[4]: *** [scripts/Makefile.build:382: arch/powerpc/kernel/head_64.o] Error 1
> arch/powerpc/kernel/entry_64.S: Assembler messages:
> arch/powerpc/kernel/entry_64.S:172: Error: junk at end of line: `0b01010'
> clang: error: assembler command failed with exit code 1 (use -v to see
> invocation)
> make[4]: *** [scripts/Makefile.build:382:
> arch/powerpc/kernel/entry_64.o] Error 1
> make[4]: Target 'arch/powerpc/kernel/' not remade because of errors.
> make[3]: *** [scripts/Makefile.build:500: arch/powerpc/kernel] Error 2
> make[3]: Target 'arch/powerpc/' not remade because of errors.
> make[2]: *** [scripts/Makefile.build:500: arch/powerpc] Error 2
> make[2]: Target './' not remade because of errors.
> make[1]: *** [Makefile:2012: .] Error 2
> make[1]: Target '__all' not remade because of errors.
> make: *** [Makefile:238: __sub-make] Error 2
> make: Target '__all' not remade because of errors.

As always, thanks for the report. This is an LLVM regression/change in
behavior caused by [1], which can break as-option and as-instr on
releases prior to commit d5c8d6e0fa61 ("kbuild: Update assembler calls
to use proper flags and language target"), as unsupported flags for the
current target ('-x') may be present (KBUILD_CFLAGS is used for these
tests instead of KBUILD_AFLAGS). Inside try-run, the macro behind
as-instr and as-option, I see

  clang-17: error: unsupported option '-mno-prefixed' for target 'powerpc64le-linux-gnu'
  clang-17: error: unsupported option '-mno-pcrel' for target 'powerpc64le-linux-gnu'
  clang-17: error: unsupported option '-mno-altivec' for target 'powerpc64le-linux-gnu'
  clang-17: error: unsupported option '-mno-vsx' for target 'powerpc64le-linux-gnu'
  clang-17: error: unsupported option '-mno-mma' for target 'powerpc64le-linux-gnu'
  clang-17: error: unsupported option '-mno-spe' for target 'powerpc64le-linux-gnu'

This has come up recently elsewhere in PowerPC, see
commit 2b694fc96fe3 ("powerpc/boot: Disable power10 features after
BOOTAFLAGS assignment"). While I think it is dubious that clang errors
on these flags for the assembler target, this is already fixed on the
Linux side by using KBUILD_AFLAGS for these make macros.

I am preparing a series of d5c8d6e0fa61 and its dependencies for 6.1 but
I want to do sufficient build testing first, which is currently running
for me. Would you be able to point your matrix to [2] to make sure
everything works properly with both GCC and LLVM? It is a work in
progress as the second patch in the stack needs a proper commit message
but it is the diff I expect to ship so that it all that matters.

[1]: https://github.com/llvm/llvm-project/commit/5548843d692a92a7840f14002debc3cebcb3cdc3
[2]: https://git.kernel.org/pub/scm/linux/kernel/git/nathan/linux.git/log/?h=wip/b4/6-1-asssembler-target-llvm-17

Cheers,
Nathan