Re: [PATCH v3 2/5] mm/mseal: update madvise() logic

[Lorenzo Stoakes]

On Fri, Jul 25, 2025 at 12:10:07PM +0200, David Hildenbrand wrote:
> > > > So - we explicitly disallow FOLL_FORCE write override for CoW file-backed
> > > > mappings.
> > > >
> > > > Obviously if FOLL_FORCE is not set, then we're ALSO not allowed to get past a
> > > > FOLL_WRITE and !VM_WRITE situation.
> > > >
> > > > >
> > > > > >
> > > > > > Hmm maybe I'll soften on this anon_vma idea then. Maybe it is a 'cheap fix' to
> > > > > > rule out the _usual_ cases.
> > > > >
> > > > > Yeah, something to evaluate.
> > > >
> > > > I'm thinking more and more we're probably actually safe with !vma->anon_vma ||
> > > > !(vma->vm_flags & VM_MAYWRITE).
> > >
> > > I think there are possible races, the question is how much you care about
> > > them.
> >
> > Yes I was just wrong. Please just disregard.
> >
> > I mean racing with MADV_POPULATE_WRITE seems a niche thing to worry about, and
> > so what if you did, it's writing a... copy of the underlying file-backed folios
> > no?
>
> MADV_POPULATE_WRITE does not apply. It's only racing with FOLL_FORCE, like
> debugger access.

Yeah right, that too. OK so we might have some breakpoint or changed some data,
then discard, I don't think this is _too_ problematic right? The debugger is
naturally racey anyway.

I think we can accept these.

>
> Again, a race one probably shouldn't worry about in the context of mseal.

Yeah agreed, so I think this is still sensible.

>
> --
> Cheers,
>
> David / dhildenb
>

Cheers, Lorenzo