Re: [PATCH 1/3] CET: Rename CET_MAX to CET_CONTROL_MASK [BZ #25887]

[Florian Weimer <fw@deneb.enyo.de>]

* H. J. Lu via Libc-alpha:

> diff --git a/sysdeps/x86/cet-tunables.h b/sysdeps/x86/cet-tunables.h
> index 5e1e42df10..0088b89d3e 100644
> --- a/sysdeps/x86/cet-tunables.h
> +++ b/sysdeps/x86/cet-tunables.h
> @@ -16,14 +16,32 @@
>     License along with the GNU C Library; if not, see
>     <https://www.gnu.org/licenses/>.  */
>  
> -/* Valid control values:
> +#ifndef _CET_TUNABLES_H
> +#define _CET_TUNABLES_H
> +
> +/* For each CET feature, IBT and SHSTK, valid control values:
>     0: Enable CET features based on ELF property note.
>     1: Always disable CET features.
>     2: Always enable CET features.
>     3: Enable CET features permissively.
> +
> +   Bits 0-1: IBT
> +   Bits 2-3: SHSTK
>   */
>  #define CET_ELF_PROPERTY	0
>  #define CET_ALWAYS_OFF		1
>  #define CET_ALWAYS_ON		2
>  #define CET_PERMISSIVE		3
> -#define CET_MAX			CET_PERMISSIVE
> +#define CET_CONTROL_MASK	3
> +#define CET_IBT_SHIFT		0
> +#define CET_SHSTK_SHIFT		2
> +
> +/* Get CET control value.  */
> +
> +static inline unsigned int
> +get_cet_control_value (unsigned int shift)
> +{
> +  return (GL(dl_x86_feature_1)[1] >> shift) & CET_CONTROL_MASK;
> +}
> +
> +#endif /* cet-tunables.h */

Is there a reason why this has to be a single bitmask?  Maybe a
bitfield would better document the intent?