Re: [PATCH] x86/mm: Use standard C types for sized integers

[Frediano Ziglio <frediano.ziglio@xxxxxxxxx>]

On Wed, Oct 30, 2024 at 11:02 AM Jan Beulich <jbeulich@xxxxxxxx> wrote:
>
> On 30.10.2024 11:44, Frediano Ziglio wrote:
> > The header is already using these types.
> >
> > Signed-off-by: Frediano Ziglio <frediano.ziglio@xxxxxxxxx>
>
> Acked-by: Jan Beulich <jbeulich@xxxxxxxx>
>

Thanks

> Nevertheless I wonder whether ...
>
> > --- a/xen/arch/x86/include/asm/mm.h
> > +++ b/xen/arch/x86/include/asm/mm.h
> > @@ -230,7 +230,7 @@ struct page_info
> >           * Only valid for: a) free pages, and b) pages with zero type count
> >           * (except page table pages when the guest is in shadow mode).
> >           */
> > -        u32 tlbflush_timestamp;
> > +        uint32_t tlbflush_timestamp;
> >
> >          /*
> >           * When PGT_partial is true then the first two fields are valid and
> > @@ -284,8 +284,8 @@ struct page_info
> >           *   in use.
> >           */
> >          struct {
> > -            u16 nr_validated_ptes:PAGETABLE_ORDER + 1;
> > -            u16 :16 - PAGETABLE_ORDER - 1 - 1;
> > +            uint16_t nr_validated_ptes:PAGETABLE_ORDER + 1;
> > +            uint16_t :16 - PAGETABLE_ORDER - 1 - 1;
> >              uint16_t partial_flags:1;
>
> ... fixed width types are really needed here; afaict unsigned int ought to do.
>

Not for gcc/clang. Other compilers (like MS one) differs... in our
case this is not a public header and we only support gcc/clang so
unsigned/int would be the same.

> >              int16_t linear_pt_count;
>
> It's only here where the fixed width type is largely needed (or alternatively
>
>             signed int linear_pt_count:16;
>

That would be different. Compilers do not allow to take addresses of bit-fields.

> ).
>
> Jan

Frediano