On Fri, Nov 19, 2021 at 5:12 PM Alejandro Colomar (man-pages)
<alx.manpages@xxxxxxxxx> wrote:
>
> On 11/19/21 16:57, Arnd Bergmann wrote:
> >
> > From what I can tell, linux/stddef.h is tiny, I don't think it's really
> > worth optimizing this part. I have spent some time last year
> > trying to untangle some of the more interesting headers, but ended
> > up not completing this as there are some really hard problems
> > once you start getting to the interesting bits.
>
> In this case it was not about being worth it or not,
> but that the fact that adding memberof() would break,
> unless I use 0 instead of NULL for the implementation of memberof(),
> which I'm against, or I split stddef.
>
> If I don't do either of those,
> I'm creating a circular dependency,
> and it doesn't compile.
Sorry for missing the background here, but I don't see what that
dependency is. If memberof() is a macro, then including the definition
should not require having the NULL definition first, you just need to
have both at the time you use it.
> > The main issue here is that user space code should not
> > include anything outside of include/uapi/ and arch/*/include/uapi/
>
> Okay. That's good to know.
>
> So everything can use uapi code,
> and uapi code can only use uapi code,
> right?
Correct.
> > offsetof() is defined in include/linux/stddef.h, so this is by
> > definition not accessible here. It appears that there is also
> > an include/uapi/linux/stddef.h that is really strange because
> > it includes linux/compiler_types.h, which in turn is outside
> > of uapi/. This should probably be fixed.
>
> I see.
> Then,
> perhaps it would be better to define offsetof() _only_ inside uapi/,
> and use that definition from everywhere else,
> and therefore remove the non-uapi version,
> right?
No, because the user-space <stddef.h> provided by the compiler
also includes an offsetof() definition. In the uapi/ namespace, the
kernel must only provide definitions that do not clash with anything
in user space.
Arnd
