Re: [PATCH v6 7/8] fs/proc/task_mmu: read proc/pid/maps under per-vma lock

[Suren Baghdasaryan <surenb@google.com>]

On Wed, Jul 9, 2025 at 3:03 AM Vlastimil Babka <vbabka@suse.cz> wrote:
>
> On 7/4/25 08:07, Suren Baghdasaryan wrote:
> > --- a/mm/mmap_lock.c
> > +++ b/mm/mmap_lock.c
> > @@ -178,6 +178,94 @@ struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
> >       count_vm_vma_lock_event(VMA_LOCK_ABORT);
> >       return NULL;
> >  }
> > +
> > +static struct vm_area_struct *lock_vma_under_mmap_lock(struct mm_struct *mm,
> > +                                                    struct vma_iterator *iter,
> > +                                                    unsigned long address)
> > +{
> > +     struct vm_area_struct *vma;
> > +     int ret;
> > +
> > +     ret = mmap_read_lock_killable(mm);
> > +     if (ret)
> > +             return ERR_PTR(ret);
> > +
> > +     /* Lookup the vma at the last position again under mmap_read_lock */
> > +     vma_iter_init(iter, mm, address);
> > +     vma = vma_next(iter);
> > +     if (vma)
> > +             vma_start_read_locked(vma);
>
> This can in theory return false (refcount overflow?) so it should be handled?

Yes, I should handle it by falling back to mmap_lock. Thanks!

>
> > +
> > +     mmap_read_unlock(mm);
> > +
> > +     return vma;
> > +}
> > +