bug#72278: 29.4; Proced reports incorrect results with 16KB page size

[Paul Eggert]

On 2024-07-26 15:33, Stefan Kangas wrote:
> Paul Eggert<eggert@cs.ucla.edu> writes:
>
> > On 2024-07-25 02:55, Stefan Kangas wrote:
> > > On Linux, page sizes can vary per process ("multiple page size
> > > support").  Should we bother with that?
> > I thought that support didn't change what getpagesize returns or what
> > /proc/PID/stat reports, at least not in mainline Linux kernels. Am I
> > missing something?
> I don't know to be honest, which is why I'm asking.

For recent Asahi Linux user-mode pages are always 16 KiB; see 
<https://github.com/AsahiLinux/docs/wiki/Broken-Software>.

I'd be surprised if any Linux distro lets getpagesize calls differ among 
processes running atop the same kernel. Although Linux does support huge 
pages via (for example) mmap with MAP_HUGETLB, I would expect this to 
not affect getpagesize or /proc/PID/stat. I haven't verified this by 
running code or by scouting through kernel source, though.