[kernel-hardening] Re: [PATCH 1/2] procfs: restore 0400 permissions on /proc/*/{syscall,stack,personality}

[Ryan Mallon <rmallon@gmail.com>]

On 04/10/13 10:41, Kees Cook wrote:
> On Wed, Aug 28, 2013 at 1:49 PM, Kees Cook <keescook@chromium.org> wrote:

<snip>

> 
> BTW, this just came to my attention:
> http://marc.info/?l=linux-kernel&m=138049414321387&w=2
> 
> Same problem, just for /proc/kallsyms. This would benefit from the
> open vs read cred check as well, I think.

I was actually just about to put together a repost of this. Sorry I
missed you off the original Cc list, get_maintainer didn't list you.

I wanted to at least change the comment mentioning "badly written"
setuid binaries. That isn't really true, as George Spelvin pointed out,
even a setuid binary which opens the file with dropped priviledges, but
reads it after re-elevating privileges will be susceptible to this.

Setuid apps could be more precautious by doing the open + read into
memory of user files with the privileges dropped, so that once
privileges are re-elevated only the in-memory copy is used.

I still think in-kernel fixing is a good idea too though, since it
hardens against user-space setuid apps that don't do this. This was just
the simplest approach to fixing the problem that I could think of. I'm
open to suggestions for a better solution.

~Ryan