Re: [PATCH] mm/rmap: Add anon_vma lifetime debug check

[Jann Horn <jannh@google.com>]

On Thu, Jul 24, 2025 at 11:52 PM Andrew Morton
<akpm@linux-foundation.org> wrote:
> On Thu, 24 Jul 2025 21:13:50 +0200 Jann Horn <jannh@google.com> wrote:
>
> > If an anon page is mapped into userspace, its anon_vma must be alive,
> > otherwise rmap walks can hit UAF.
> >
> > There have been syzkaller reports a few months ago[1][2] of UAF in rmap
> > walks that seems to indicate that there can be pages with elevated mapcount
> > whose anon_vma has already been freed, but I think we never figured out
> > what the cause is; and syzkaller only hit these UAFs when memory pressure
> > randomly caused reclaim to rmap-walk the affected pages, so it of course
> > didn't manage to create a reproducer.
> >
> > Add a VM_WARN_ON_FOLIO() when we add/remove mappings of anonymous pages to
> > hopefully catch such issues more reliably.
> >
> > Implementation note: I'm checking IS_ENABLED(CONFIG_DEBUG_VM) because,
> > unlike the checks above, this one would otherwise be hard to write such
> > that it completely compiles away in non-debug builds by itself, without
> > looking extremely ugly.
> >
> > --- a/include/linux/rmap.h
> > +++ b/include/linux/rmap.h
> > @@ -449,6 +449,19 @@ static inline void __folio_rmap_sanity_checks(const struct folio *folio,
> >       default:
> >               VM_WARN_ON_ONCE(true);
> >       }
> > +
> > +     /*
> > +      * Anon folios must have an associated live anon_vma as long as they're
> > +      * mapped into userspace.
> > +      * Part of the purpose of the atomic_read() is to make KASAN check that
> > +      * the anon_vma is still alive.
> > +      */
> > +     if (IS_ENABLED(CONFIG_DEBUG_VM) && PageAnonNotKsm(page)) {
> > +             unsigned long mapping = (unsigned long)folio->mapping;
> > +             struct anon_vma *anon_vma = (void *)(mapping - PAGE_MAPPING_ANON);
> > +
> > +             VM_WARN_ON_FOLIO(atomic_read(&anon_vma->refcount) == 0, folio);
> > +     }
> >  }
>
> PAGE_MAPPING_ANON is now FOLIO_MAPPING_ANON.

Bleh, sorry about that, I keep forgetting to write MM patches against
the MM tree...

> The subtraction to clear a bitflag works, but my brain would prefer &=
> FOLIO_MAPPING_ANON.  Oh well.

(I'd prefer bitmasking too but the existing code does subtraction, so
I figured I should mirror that.)