Re: [lxc-devel] Kernel bug? Setuid apps and user namespaces

["Serge E. Hallyn" <serge@hallyn.com>]

Quoting Andy Lutomirski (luto@amacapital.net):
> On Mon, Apr 7, 2014 at 11:13 AM, Serge E. Hallyn <serge@hallyn.com> wrote:
> > Quoting Andy Lutomirski (luto@amacapital.net):
> >> I'm starting to think that we need to extend dumpable to something
> >> much more general like a list of struct creds that someone needs to be
> >> able to ptrace, *in addition to current creds* in order to access
> >> sensitive /proc files, coredumps, etc.  If you get started as setuid,
> >
> > Hm, yeah, this sort of makes sense.
> >
> >> then you start with two struct creds in the list (or maybe just your
> >> euid and uid).  If you get started !setuid, then your initial creds
> >> are in the list.  It's possible that few or no things will need to
> >> change that list after execve.
> >>
> >> If all of the entries and current->cred are in the same user_ns, then
> >> we can dump as userns root.  If they're in different usernses, then we
> >> dump as global root or maybe the common ancestor root.
> >> setuid(getuid()) and other such nastiness may have to empty the list,
> >> or maybe we can just use a prctl for that.
> >
> > A few questions,
> >
> > 1. is there any other action which would trigger adding a new cred to
> > the ist?
> 
> I don't think so.  Anyone who can ptrace you from the start can
> corrupt you such that you leak rights even if some future action
> prevents new ptracers from attaching.
> 
> OTOH, it might be nice for something like an HTTPS server to be able
> to fork and shove its private key into the child, while preventing
> anyone from ptracing the child.  But doing this securely without help
> from someone with a different uid might be impossible anyway.
> 
> >
> > 2. would execve clear (and re-init) the list of creds?
> 
> Probably.  Thoughts?

Yeah it seems to me it should be re-initialized, with a cred added
to the list for every open fd.

> We could have a way to ask execve not to reinit the list.  Such a
> mechanism would have to require no_new_privs to prevent a
> straightforward attack on any setuid binary.

If we don't add a cred for every open fd, then I think we need this.  If
we do, then I'm not sure this makes sense.

> We's also want PR_SET_DUMPABLE or a new prctl to be able reset the
> list to contain just current-.cred, I think.

Yeah PR_SET_DUMPABLE should reset it, agreed.

-serge