Re: Error: invalid switch -me200

[Nick Desaulniers]

On Fri, Nov 13, 2020 at 12:04 PM Nathan Chancellor
<natechancellor@xxxxxxxxx> wrote:
>
> On Fri, Nov 13, 2020 at 11:42:03AM -0800, Nick Desaulniers wrote:
> > + MPE, PPC
> >
> > On Fri, Nov 13, 2020 at 11:08 AM Nathan Chancellor
> > <natechancellor@xxxxxxxxx> wrote:
> > >
> > > On Fri, Nov 13, 2020 at 09:28:03AM -0800, Fāng-ruì Sòng wrote:
> > > > On Thu, Nov 12, 2020 at 7:22 PM kernel test robot <lkp@xxxxxxxxx> wrote:
> > > > >
> > > > > Hi Fangrui,
> > > > >
> > > > > FYI, the error/warning still remains.
> > > > >
> > > > > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> > > > > head:   585e5b17b92dead8a3aca4e3c9876fbca5f7e0ba
> > > > > commit: ca9b31f6bb9c6aa9b4e5f0792f39a97bbffb8c51 Makefile: Fix GCC_TOOLCHAIN_DIR prefix for Clang cross compilation
> > > > > date:   4 months ago
> > > > > config: powerpc-randconfig-r031-20201113 (attached as .config)
> >
> > ^ randconfig
> >
> > > > > compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project 9e0c35655b6e8186baef8840b26ba4090503b554)
> > > > > reproduce (this is a W=1 build):
> > > > >         wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
> > > > >         chmod +x ~/bin/make.cross
> > > > >         # install powerpc cross compiling tool for clang build
> > > > >         # apt-get install binutils-powerpc-linux-gnu
> > > > >         # https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ca9b31f6bb9c6aa9b4e5f0792f39a97bbffb8c51
> > > > >         git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> > > > >         git fetch --no-tags linus master
> > > > >         git checkout ca9b31f6bb9c6aa9b4e5f0792f39a97bbffb8c51
> > > > >         # save the attached .config to linux build tree
> > > > >         COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=powerpc
> > > > >
> > > > > If you fix the issue, kindly add following tag as appropriate
> > > > > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > > > >
> > > > > All errors (new ones prefixed by >>):
> > > > >
> > > > >    Assembler messages:
> > > > > >> Error: invalid switch -me200
> > > > > >> Error: unrecognized option -me200
> > > > >    clang-12: error: assembler command failed with exit code 1 (use -v to see invocation)
> > > > >    make[2]: *** [scripts/Makefile.build:281: scripts/mod/empty.o] Error 1
> > > > >    make[2]: Target '__build' not remade because of errors.
> > > > >    make[1]: *** [Makefile:1174: prepare0] Error 2
> > > > >    make[1]: Target 'prepare' not remade because of errors.
> > > > >    make: *** [Makefile:185: __sub-make] Error 2
> > > > >    make: Target 'prepare' not remade because of errors.
> > > > >
> > > > > ---
> > > > > 0-DAY CI Kernel Test Service, Intel Corporation
> > > > > https://lists.01.org/hyperkitty/list/kbuild-all@xxxxxxxxxxxx
> > > >
> > > > This can be ignored. The LLVM integrated assembler does not recognize
> > > > -me200 (-Wa,-me200 in arch/powerpc/Makefile). I guess the GNU as -m
> > > > option is similar to .arch or .machine and controls what instructions
> > > > are recognized. The integrated assembler tends to support all
> > > > instructions (conditional supporting some instructions has some
> > > > challenges; in the end I have patched parsing but ignoring `.arch` for
> > > > x86-64 and ignoring `.machine ppc64` for ppc64)
> > > >
> > > > (In addition, e200 is a 32-bit Power ISA microprocessor. 32-bit
> > > > support may get less attention in LLVM.)
> > >
> > > This is also not a clang specific issue, I see the exact same error
> > > with GCC 10.2.0 and binutils 2.35.
> > >
> > > $ make -skj64 ARCH=powerpc CROSS_COMPILE=powerpc64-linux- olddefconfig vmlinux
> >
> > Does using a non 64b triple produce the same failure?
>
> Yes, CROSS_COMPILE=powerpc-linux- produces the same failure.
>
> > > ...
> > > Error: invalid switch -me200
> > > Error: unrecognized option -me200
> >
> > There's a block in  arch/powerpc/Makefile:
> > 248 cpu-as-$(CONFIG_40x)    += -Wa,-m405
> > 249 cpu-as-$(CONFIG_44x)    += -Wa,-m440
> > 250 cpu-as-$(CONFIG_ALTIVEC)  += $(call
> > as-option,-Wa$(comma)-maltivec)
> > 251 cpu-as-$(CONFIG_E200)   += -Wa,-me200
> > 252 cpu-as-$(CONFIG_E500)   += -Wa,-me500
> >
> > Are those all broken configs, or is Kconfig messed up such that
> > randconfig can select these when it should not?
>
> Hmmm, looks like this flag does not exist in mainline binutils? There is
> a thread in 2010 about this that Segher commented on:
>
> https://lore.kernel.org/linuxppc-dev/9859E645-954D-4D07-8003-FFCD2391AB6E@xxxxxxxxxxxxxxxxxxx/
>
> Guess this config should be eliminated?

If we're going to get pestered by 0day bot randconfigs over code
that's not possible to build, I'm all for deleting it.  I doubt we'll
be seeing patches from anyone to binutils for supporting these.

What has the kernel's policy been for code in tree that other folks
can't build (without proprietary tools)? (ARCH=hexagon is pretty close
to not toeing the line here, not sure ICC actually works either).
-- 
Thanks,
~Nick Desaulniers