by Suren Baghdasaryan

[permalink] [raw]

Subject: Re: [PATCH v4 3/3] mm: fix use-after-free when anon vma name is used after vma is freed

On Wed, Feb 23, 2022 at 7:29 AM Michal Hocko <[email protected]> wrote:
>
> On Wed 23-02-22 06:10:54, Suren Baghdasaryan wrote:
> > On Wed, Feb 23, 2022 at 12:55 AM Michal Hocko <[email protected]> wrote:
> > >
> > > On Tue 22-02-22 07:43:40, Suren Baghdasaryan wrote:
> > > > On Tue, Feb 22, 2022 at 12:06 AM Michal Hocko <[email protected]> wrote:
> > > > >
> > > > > On Mon 21-02-22 21:40:25, Suren Baghdasaryan wrote:
> > > > > > When adjacent vmas are being merged it can result in the vma that was
> > > > > > originally passed to madvise_update_vma being destroyed. In the current
> > > > > > implementation, the name parameter passed to madvise_update_vma points
> > > > > > directly to vma->anon_name->name and it is used after the call to
> > > > > > vma_merge. In the cases when vma_merge merges the original vma and
> > > > > > destroys it, this will result in use-after-free bug as shown below:
> > > > > >
> > > > > > madvise_vma_behavior << passes vma->anon_name->name as name param
> > > > > > madvise_update_vma(name)
> > > > > > vma_merge
> > > > > > __vma_adjust
> > > > > > vm_area_free <-- frees the vma
> > > > > > replace_vma_anon_name(name) <-- UAF
> > > > >
> > > > > This seems to be stale because bare const char pointer is not passed in
> > > > > the call chain. In fact I am not even sure there is any actual UAF here
> > > > > after the rework.
> > > > > Could you be more specific in describing the scenario?
> > > >
> > > > Yes, sorry, I need to update the part of the description talking about
> > > > passing vma->anon_name->name directly.
> > > > I think UAF is still there, it's just harder to reproduce (admittedly
> > > > I could not reproduce it with the previous reproducer). The scenario
> > > > would be when a vma with vma->anon_name->kref == 1 is being merged
> > > > with another one and freed in the process:
> > > >
> > > > madvise_vma_behavior
> > > > anon_name = vma_anon_name(vma) <-- does not increase refcount
> > > > madvise_update_vma(anon_name)
> > > > *prev = vma_merge <-- returns another vma
> > > > __vma_adjust
> > > > vm_area_free(vma)
> > > > free_vma_anon_name
> > > > anon_vma_name_put
> > > > vma_anon_name_free <-- frees the vma->anon_name
> > > > vma = *prev <-- original vma was freed
> > >
> > > How come this is not a UAF in the first place?
> >
> > Sorry, I got you confused. The original vma that was passed as a
> > parameter to vma_merge(vma) was freed and vma_merge() returns the area
> > it was merged with:
> >
> > *prev = vma_merge(vma_a) <-- vma_a is merged with adjacent vma_b,
> > vma_a is freed and vma_b is returned
> > vma = *prev <-- "vma" now points to vma_b
>
> OK, I see the problem now. Essentially if we have two adjacent vmas
> which have the same name but wrapped by two anon_vma_name instances
> and we merge them together then the first one can be freed along with
> its anon_vma_name which is still used by replace_anon_vma_name.
>
> Thanks for the patience.

Thanks for reviewing! I posted the next version of the patchset at
https://lore.kernel.org/linux-mm/[email protected]/

> --
> Michal Hocko
> SUSE Labs

2022-02-24 01:14:19

by Suren Baghdasaryan

[permalink] [raw]

Subject: Re: [PATCH 2/3] mm: prevent vm_area_struct::anon_name refcount saturation

On Tue, Feb 22, 2022 at 1:17 AM Michal Hocko <[email protected]> wrote:
>
> On Mon 21-02-22 21:40:24, Suren Baghdasaryan wrote:
> > A deep process chain with many vmas could grow really high.
>
> This would really benefit from some numbers. With default
> sysctl_max_map_count (64k) and default pid_max (32k) the INT_MAX could
> be theoretically reached but I find it impractical because not all vmas
> can be anonymous same as all available pids can be consumed for a
> theoretical attack (if my counting is proper).
> On the other hand any non-default configuration with any of the values
> increased could hit this theoretically.

re: This would really benefit from some numbers
Should I just add the details you provided above into the description?
Would that suffice?

>
> > kref
> > refcounting interface used in anon_vma_name structure will detect
> > a counter overflow when it reaches REFCOUNT_SATURATED value but will
> > only generate a warning about broken refcounting.
> > To ensure anon_vma_name refcount does not overflow, stop anon_vma_name
> > sharing when the refcount reaches INT_MAX, which still leaves INT_MAX/2
> > values before the counter reaches REFCOUNT_SATURATED. This should provide
> > enough headroom for raising the refcounts temporarily.
> >
> > Suggested-by: Michal Hocko <[email protected]>
> > Signed-off-by: Suren Baghdasaryan <[email protected]>
> > ---
> > include/linux/mm_inline.h | 18 ++++++++++++++----
> > mm/madvise.c | 3 +--
> > 2 files changed, 15 insertions(+), 6 deletions(-)
> >
> > diff --git a/include/linux/mm_inline.h b/include/linux/mm_inline.h
> > index 70b619442d56..b189e2638843 100644
> > --- a/include/linux/mm_inline.h
> > +++ b/include/linux/mm_inline.h
> > @@ -156,15 +156,25 @@ static inline void anon_vma_name_get(struct anon_vma_name *anon_name)
> >
> > extern void anon_vma_name_put(struct anon_vma_name *anon_name);
> >
> > +static inline
> > +struct anon_vma_name *anon_vma_name_reuse(struct anon_vma_name *anon_name)
> > +{
> > + /* Prevent anon_name refcount saturation early on */
> > + if (kref_read(&anon_name->kref) < INT_MAX) {
>
> REFCOUNT_MAX seems to be defined by the kref framework.

Ah, indeed. I missed that. Will change to use it.

>
> Other than that looks good to me.

Thanks for the review!

>
> > + anon_vma_name_get(anon_name);
> > + return anon_name;
> > +
> > + }
> > + return anon_vma_name_alloc(anon_name->name);
> > +}
> > +
> > static inline void dup_vma_anon_name(struct vm_area_struct *orig_vma,
> > struct vm_area_struct *new_vma)
> > {
> > struct anon_vma_name *anon_name = vma_anon_name(orig_vma);
> >
> > - if (anon_name) {
> > - anon_vma_name_get(anon_name);
> > - new_vma->anon_name = anon_name;
> > - }
> > + if (anon_name)
> > + new_vma->anon_name = anon_vma_name_reuse(anon_name);
> > }
> >
> > static inline void free_vma_anon_name(struct vm_area_struct *vma)
> > diff --git a/mm/madvise.c b/mm/madvise.c
> > index f81d62d8ce9b..a395884aeecb 100644
> > --- a/mm/madvise.c
> > +++ b/mm/madvise.c
> > @@ -122,8 +122,7 @@ static int replace_vma_anon_name(struct vm_area_struct *vma,
> > if (anon_vma_name_eq(orig_name, anon_name))
> > return 0;
> >
> > - anon_vma_name_get(anon_name);
> > - vma->anon_name = anon_name;
> > + vma->anon_name = anon_vma_name_reuse(anon_name);
> > anon_vma_name_put(orig_name);
> >
> > return 0;
> > --
> > 2.35.1.473.g83b2b277ed-goog
>
> --
> Michal Hocko
> SUSE Labs