Re: [arnd-asm-generic:asm-generic 15/15] /usr/aarch64-linux-gnu/include/linux/types.h:12: Error: unknown mnemonic `typedef' -- `typedef __signed__ __int128 __s128 __attribute__((aligned(16)))'

[Thomas Huth <thuth@redhat.com>]

On 29/10/2026 23.13, Arnd Bergmann wrote:
> On Thu, Oct 29, 2026, at 19:09, kernel test robot wrote:
>> tree:
>> https://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic.git
>> asm-generic
>> head:   dd7cc8c26b6a37a54adc5ce38d286521b7201044
>> commit: dd7cc8c26b6a37a54adc5ce38d286521b7201044 [15/15] treewide: Stop
>> defining __ASSEMBLY__ for assembler files
> 
> Hi Thomas,
> 
>> config: arm64-allnoconfig-bpf
>> (https://download.01.org/0day-ci/archive/20261029/202610291918.kHO0GTDB-lkp@intel.com/config)
>> compiler: aarch64-linux-gnu-gcc (Debian 14.2.0-19) 14.2.0
>> reproduce (this is a W=1 build):
>> (https://download.01.org/0day-ci/archive/20261029/202610291918.kHO0GTDB-lkp@intel.com/reproduce)
>>
>> If you fix the issue in a separate patch/commit (i.e. not just a new
>> version of
>> the same patch/commit), kindly add following tags
>> | Reported-by: kernel test robot <lkp@intel.com>
>> | Closes:
>> https://lore.kernel.org/oe-kbuild-all/202610291918.kHO0GTDB-lkp@intel.com/
>>
>> Note: the arnd-asm-generic/asm-generic HEAD
>> dd7cc8c26b6a37a54adc5ce38d286521b7201044 builds fine.
>>        It only hurts bisectability.
>>
>> All errors (new ones prefixed by >>):
>>
>>     /usr/aarch64-linux-gnu/include/asm-generic/int-ll64.h: Assembler messages:
>>>> /usr/aarch64-linux-gnu/include/asm-generic/int-ll64.h:20: Error: unknown mnemonic `typedef' -- `typedef __signed__ char __s8'
>>>> /usr/aarch64-linux-gnu/include/asm-generic/int-ll64.h:21: Error: unknown mnemonic `typedef' -- `typedef unsigned char __u8'
> 
> 
>  From what I can tell, this happens after your last patch when cross-compiling
> kselftest against system headers that still check for __ASSEMBLY__ instead of
> __ASSEMBLER__.
> 
> The command from the build log is:
>          COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-14.2.0 ~/lkp-tests/kbuild/make.cross W=1 -C tools/testing/selftests
> 
> The command that causes this is
> 
> gcc -std=gnu99 -O2 -Wall -Wstrict-prototypes -D_GNU_SOURCE= -I/home/arnd/arm-soc/asm-generic/tools/testing/selftests/../../../tools/testing/selftests -idirafter /home/arnd/arm-soc/asm-generic/tools/testing/selftests/../../../tools/include -idirafter /home/arnd/arm-soc/asm-generic/tools/testing/selftests/../../../tools/include/generated -idirafter /home/arnd/arm-soc/asm-generic/tools/testing/selftests/../../../arch/arm64/include -idirafter /home/arnd/arm-soc/asm-generic/tools/testing/selftests/../../../include -Wa,--noexecstack    vdso_test_chacha.c vgetrandom-chacha.S  -o /home/arnd/arm-soc/build/tmp/kselftest/vDSO/vdso_test_chacha
> 
> and looking at the preprocessed file, I see a wild mix of tools/include,
> ${srctree}/arch/arm64/include/ and /usr/include. I also see that at least
> part of the problem is that arch/arm64/kernel/vdso/vgetrandom-chacha.S
> includes arch/arm64/include/asm/assembler.h, which then goes on to include
> the world.
After getting access to an aarch64 host, I was finally able to reproduce the 
problem. I think it's likely best if we force the selftest to use the 
headers from the current kernel sources instead of using the headers from
/usr/include ... i.e. this patch fixes the issue for me:

  https://lore.kernel.org/lkml/20261030174323.370687-1-thuth@redhat.com/

WDYT?

  Thomas