Re: [PATCH 6/6] percpu/x86: Enable strict percpu checks via named AS qualifiers

[Uros Bizjak <ubizjak@gmail.com>]

On Fri, Nov 29, 2024 at 4:45 PM Nadav Amit <nadav.amit@gmail.com> wrote:
>
>
> > On 26 Nov 2024, at 19:21, Uros Bizjak <ubizjak@gmail.com> wrote:
> >
> > This patch declares percpu variables in __seg_gs/__seg_fs named AS
> > and keeps them named AS qualified until they are dereferenced with
> > percpu accessor. This approach enables various compiler check
> > for cross-namespace variable assignments.
>
> [snip]
>
> > @@ -95,9 +95,19 @@
> >
> > #endif /* CONFIG_SMP */
> >
> > -#define __my_cpu_type(var) typeof(var) __percpu_seg_override
> > -#define __my_cpu_ptr(ptr) (__my_cpu_type(*(ptr))*)(__force uintptr_t)(ptr)
> > -#define __my_cpu_var(var) (*__my_cpu_ptr(&(var)))
> > +#if defined(CONFIG_USE_X86_SEG_SUPPORT) && \
> > +    defined(CONFIG_CC_HAS_TYPEOF_UNQUAL) && !defined(__CHECKER__)
>
> Is the __CHECKER__ check because of sparse, as in patch 2/6 ?
> If so, do you want to add a similar comment here?

Yes, this is the same check. We can declare _percpu variables in
__seg_gs named address space only when __typeof_unqual__ is used. I
will add a comment in the next revision of the patchset.

> Other than that, I went over the different patches and it looks good as
> much as I can tell.
>
> If it means anything, you have for the series
>
> Acked-by: Nadav Amit <nadav.amit@gmail.com <mailto:nadav.amit@gmail.com>>

Thanks!

Uros.