Re: [PATCH v2 00/39] Shadowstacks for userspace

From: Jann Horn
Date: Mon Oct 03 2022 - 13:26:22 EST

On Mon, Oct 3, 2022 at 7:04 PM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> On Thu, Sep 29, 2022 at 03:28:57PM -0700, Rick Edgecombe wrote:
> > This is an overdue followup to the “Shadow stacks for userspace” CET series.
> > Thanks for all the comments on the first version [0]. They drove a decent
> > amount of changes for v2. Since it has been awhile, I’ll try to summarize the
> > areas that got major changes since last time. Smaller changes are listed in
> > each patch.
> Thanks for the write-up!
> > [...]
> > GUP
> > ---
> > Shadow stack memory is generally treated as writable by the kernel, but
> > it behaves differently then other writable memory with respect to GUP.
> > FOLL_WRITE will not GUP shadow stack memory unless FOLL_FORCE is also
> > set. Shadow stack memory is writable from the perspective of being
> > changeable by userspace, but it is also protected memory from
> > userspace’s perspective. So preventing it from being writable via
> > FOLL_WRITE help’s make it harder for userspace to arbitrarily write to
> > it. However, like read-only memory, FOLL_FORCE can still write through
> > it. This means shadow stacks can be written to via things like
> > “/proc/self/mem”. Apps that want extra security will have to prevent
> > access to kernel features that can write with FOLL_FORCE.
> This seems like a problem to me -- the point of SS is that there cannot be
> a way to write to them without specific instruction sequences. The fact
> that /proc/self/mem bypasses memory protections was an old design mistake
> that keeps leading to surprising behaviors. It would be much nicer to
> draw the line somewhere and just say that FOLL_FORCE doesn't work on
> VM_SHADOW_STACK. Why must FOLL_FORCE be allowed to write to SS?

But once you have FOLL_FORCE, you can also just write over stuff like
executable code instead of writing over the stack. I don't think
allowing FOLL_FORCE writes over shadow stacks from /proc/$pid/mem is
making things worse in any way, and it's probably helpful for stuff
like debuggers.

If you don't want /proc/$pid/mem to be able to do stuff like that,
then IMO the way to go is to change when /proc/$pid/mem uses
FOLL_FORCE, or to limit overall write access to /proc/$pid/mem.