Re: [PATCH for-4.14 3/3] xen/x86: atomic: Don't allow to write atomically in a pointer to const

[Julien Grall <julien@xxxxxxx>]

Hi Jan,

On 04/05/2020 11:47, Jan Beulich wrote:
> On 02.05.2020 18:07, Julien Grall wrote:
> > From: Julien Grall <jgrall@xxxxxxxxxx>
> >
> > At the moment, write_atomic() will happily write to a pointer to const.
> > While there are no use in Xen, it would be best to catch them at
> > compilation time.
> >
> > Signed-off-by: Julien Grall <jgrall@xxxxxxxxxx>
> Acked-by: Jan Beulich <jbeulich@xxxxxxxx>
> albeit ...
>
> > --- a/xen/include/asm-x86/atomic.h
> > +++ b/xen/include/asm-x86/atomic.h
> > @@ -63,6 +63,8 @@ void __bad_atomic_size(void);
> >   
> >   #define write_atomic(p, x) ({                             \
> >       typeof(*(p)) __x = (x);                               \
> > +    /* Check that the pointer is not const */             \
> > +    void *__maybe_unused p_ = &__x;                       \
> ... along the lines of the similar case with guest handles I'd
> like to suggest for the comment to be more precise: It's not
> the pointer's const-ness you're after, but the pointed to
> object's. Maybe "Check that the pointer is not to a const
> type" or even just "Check that the pointer is not to const"?
I am happy with "Check that the pointer is not to a const type".

Cheers,

--
Julien Grall