Re: compiler_types.h in UAPI?

[Arnd Bergmann]

On Mon, May 6, 2024, at 10:39, Andy Shevchenko wrote:
> On Fri, May 03, 2024 at 09:34:13PM +0200, Arnd Bergmann wrote:
>> On Fri, May 3, 2024, at 20:57, Andy Shevchenko wrote:
>
>> > Today I have stumbled over use of __force and other compiler_types.h related
>> > things in UAPI headers. Can anybody explain to me how do they suppose to work
>> > outside of the kernel? Or did I miss something obvious? Or it was a mistake
>> > during UAPI split to move swab.h and byteorder/ (most of the users of those)
>> > to UAPI in the first place?
>> 
>> These get stripped out by scripts/headers_install.sh during
>> the 'make headers_install' stage:
>> 
>> sed -E -e '
>>         s/([[:space:](])(__user|__force|__iomem)[[:space:]]/\1/g
>>         s/__attribute_const__([[:space:]]|$)/\1/g
>>         s@^#include <linux/compiler(|_types).h>@@
>>         s/(^|[^a-zA-Z0-9])__packed([^a-zA-Z0-9_]|$)/\1__attribute__((packed))\2/g
>>         s/(^|[[:space:](])(inline|asm|volatile)([[:space:](]|$)/\1__\2__\3/g
>>         s@#(ifndef|define|endif[[:space:]]*/[*])[[:space:]]*_UAPI@#\1 @
>> ' $INFILE > $TMPFILE || exit 1
>
> Thanks, Arnd, TIL!
>
> But do we discourage using these in UAPI in general? I mean do we have
> this somewhere being documented?

I don't think they are discouraged in uapi headers, since the
annotations tend to be required for clean kernel builds with
sparse.

I could not find any documentation about it though, so it might
be good to mention them in Documentation/dev-tools/sparse.rst.

      Arnd