On Fri, 25 Mar 2022 16:07:47 +0100
Janosch Frank <frankja@xxxxxxxxxxxxx> wrote:
> On 3/25/22 15:30, Claudio Imbrenda wrote:
> > On Fri, 25 Mar 2022 08:29:11 +0100
> > Janosch Frank <frankja@xxxxxxxxxxxxx> wrote:
> >
> >> On 3/24/22 18:40, Heiko Carstens wrote:
> >>> On Wed, Mar 23, 2022 at 06:03:19PM +0100, Nico Boehr wrote:
> >>> ...
> >>>> +static inline unsigned long load_guarded(unsigned long *p)
> >>>> +{
> >>>> + unsigned long v;
> >>>> +
> >>>> + asm(".insn rxy,0xe3000000004c, %0,%1"
> >>>> + : "=d" (v)
> >>>> + : "m" (*p)
> >>>> + : "r14", "memory");
> >>>> + return v;
> >>>> +}
> >>>
> >>> It was like that before, but why is r14 within the clobber list?
> >>> That doesn't make sense.
> >>
> >> r14 is changed in the gs handler of the gs test which is executed if the
> >> "guarded" part of the load takes place.
> >
> > I will add a comment explaining that when picking the patch
>
> Do we need load_guarded() in this new header?
> The load/store_gscb() functions have potential to be shared across tests
> but the lg doesn't need to be executed, no?
>
> We could opt to leave it in gs.c instead
yes, probably a better idea. I'd still add the comment, though :)
shall I just fix this up when picking?
