Re: [PATCH v2] mm: align larger anonymous mappings on THP boundaries

[Yang Shi <shy828301@gmail.com>]

On Wed, Jan 17, 2024 at 11:04 PM Jiri Slaby <jirislaby@kernel.org> wrote:
>
> On 18. 01. 24, 1:07, Yang Shi wrote:
> >> This works around the problem, of course (but is a band-aid, not a fix):
> >>
> >> --- a/mm/mmap.c
> >> +++ b/mm/mmap.c
> >> @@ -1829,7 +1829,7 @@ get_unmapped_area(struct file *file, unsigned long
> >> addr, unsigned long len,
> >>                    */
> >>                   pgoff = 0;
> >>                   get_area = shmem_get_unmapped_area;
> >> -       } else if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE)) {
> >> +       } else if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE) &&
> >> !in_32bit_syscall()) {
> >>                   /* Ensures that larger anonymous mappings are THP
> >> aligned. */
> >>                   get_area = thp_get_unmapped_area;
> >>           }
> >>
> >>
> >> thp_get_unmapped_area() does not take care of the legacy stuff...
> >
> > Could you please help test the below patch? It is compiled, but I
> > don't have 32 bit userspace or machine to test it.
>
> Yeah, for x86_64, it's semantically the same as the above, so this works
> too:
>
> Tested-by: Jiri Slaby <jirislaby@kernel.org>

Thanks!

>
> > --- a/mm/huge_memory.c
> > +++ b/mm/huge_memory.c
> > @@ -811,6 +811,9 @@ static unsigned long
> > __thp_get_unmapped_area(struct file *filp,
> >          loff_t off_align = round_up(off, size);
> >          unsigned long len_pad, ret;
> >
> > +       if (IS_ENABLED(CONFIG_32BIT) || in_compat_syscall())
> > +               return 0;
> > +
> >          if (off_end <= off_align || (off_end - off_align) < size)
> >                  return 0;
>
> thanks,
> --
> js
> suse labs
>