[PATCH 1/2] aarch64: Convert GCS policy states to an enum

[Yury Khrustalev]

On Tue, Mar 31, 2026 at 08:52:26AM -0300, Adhemerval Zanella Netto wrote:
> 
> 
> On 31/03/26 05:38, Yury Khrustalev wrote:
> > On Mon, Mar 30, 2026 at 03:26:23PM -0300, Adhemerval Zanella Netto wrote:
> >>
> >> ...
> >>
> >>>> diff --git a/sysdeps/unix/sysv/linux/aarch64/dl-procruntime.c b/sysdeps/unix/sysv/linux/aarch64/dl-procruntime.c
> >>>> index 1f3b58d0fc..d49bb6cf5d 100644
> >>>> --- a/sysdeps/unix/sysv/linux/aarch64/dl-procruntime.c
> >>>> +++ b/sysdeps/unix/sysv/linux/aarch64/dl-procruntime.c
> >>>> @@ -24,7 +24,7 @@
> >>>>  # if !defined PROCINFO_DECL && defined SHARED
> >>>>    ._dl_aarch64_gcs
> >>>>  # else
> >>>> -PROCINFO_CLASS unsigned long _dl_aarch64_gcs
> >>>> +PROCINFO_CLASS aarch64_gcs_mode _dl_aarch64_gcs
> >>>
> >>> Is this change not back-portable to previous releases? If so, I'd like
> >>> to avoid making it.
> >>
> >> Do we intend do backport this fix?
> > 
> > Maybe not this fix but some other fix that depends on this code.
> 
> Which is not the case now, so there is no blocker either.

I don't want to have to fix it back later. Is there a reason why this has to
become 32-bit instead of 64-bit?

> 
> > 
> >> This does not seems a good candidate: it does
> >> not fix any inherent and I don't think this would be a requisite for other
> >> fixes.
> > 
> > Exactly, so I don't really see any reason to make this change at all.
> > 
> 
> This is an internal type improvement, and I am not understanding why you are
> blocking this change.

Like I said previously, this patch is a good improvement. I just
requested several changes to be made in v2. Regarding this change:

  -PROCINFO_CLASS unsigned long _dl_aarch64_gcs
  +PROCINFO_CLASS aarch64_gcs_mode _dl_aarch64_gcs

(and the relevant x-reg --> w-reg change in assembly), I think it's not
necessary and we shouldn't do it (unless I'm missing something here).

Otherwise, this patch seems OK, pending the requested changes, I'm not
blocking anything.

Thanks,
Yury