Re: [PATCH 03/17] um: don't generate asm/bpf_perf_event.h

[Arnd Bergmann]

On Fri, Jul 12, 2024, at 10:37, Masahiro Yamada wrote:
> On Thu, Jul 4, 2024 at 11:37 PM Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
>> --- a/arch/um/include/asm/Kbuild
>> +++ b/arch/um/include/asm/Kbuild
>> @@ -1,5 +1,4 @@
>>  # SPDX-License-Identifier: GPL-2.0
>> -generic-y += bpf_perf_event.h
>>  generic-y += bug.h
>>  generic-y += compat.h
>>  generic-y += current.h
>> diff --git a/arch/um/include/asm/bpf_perf_event.h b/arch/um/include/asm/bpf_perf_event.h
>> new file mode 100644
>> index 000000000000..0a30420c83de
>> --- /dev/null
>> +++ b/arch/um/include/asm/bpf_perf_event.h
>> @@ -0,0 +1,3 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +
>> +#include <asm-generic/bpf_perf_event.h>
>
> I guess this is a step backward.
>
> Technically, kernel-space asm/*.h files are allowed to
> wrap UAPI <asm-generic/*.h>.
> There is no reason why we ban generic-y for doing this,
> whereas check-in source is allowed.

I think in this case, the simplicity is the more important
argument: this is the only case we have of wrapping a
uapi header from a non-uapi header, and no other architecture
would need to do this.

The way the syscall-y rule works relies on enforcing some
checking for existing asm-generic headers, and I could extend
it to allow this special case, but that would make the
Makefile rule less readable  in exchange for avoiding
effectively a one-line file.

    Arnd