Re: x86: Declarations for compat_set_{cx,px}_pminfo()

[Nicola Vetrini <nicola.vetrini@xxxxxxxxxxx>]

On 2024-11-18 22:41, Andrew Cooper wrote:
> Hello,
>
> Eclair reports Rule 8.4 violations for these two.  They're real
> violations, but fixing them is proving challenging.
>
> Because of how x86_64/platform_hypercall.c sets up some defines and
> includes ../platform_hypercall.c, the declarations of
> compat_set_{cx,px}_pminfo() of pmstat.h become compat ones.
>
> Therefore the real hypercall handler does see a correct declaration.
>
> However, the implementation in cpufreq.c does not get the compat-ified
> version of pmstat.h, so misses the declaration and thus the violation.
>
> Worse however is the fact that cpufreq.c depends on not having the:
>
> #define xen_processor_performance compat_processor_performance
>
> in scope while it's transforming between the two formats, meaning we
> can't simply reuse pmstat.h with suitable defines.
>
> Any clever ideas for how to untangle this mess, before I go for the
> brute force approach of an #if COMPAT section in the main hypercall 
> handler?
Yeah, that is why these were among the last ones. I'm afraid I didn't 
come up with a reasonable to deal with this either.
You can find attached a tentative patch I made (not build tested, and 
also based on an older tree). A last resort may be to special-case those 
declarations, but that's undesirable.
--
Nicola Vetrini, BSc
Software Engineer, BUGSENG srl (https://bugseng.com)

Attachment: R8.4.patch
Description: Text Data