Re: [FD] [oss-security] Bug in bash <= 4.3 [security feature bypassed]

[Hector Marco <hecmargi () upv es>]

On 04/06/14 11:13, Jose Carlos Luna Duran wrote:
> In my opinion the drop of privs in bash was mostly a "help" measure
> for poorly written setuid programs executing system() calls. I don't
> think is the role of bash to do this as the problem that could be
> exploited by that would really be in the original program that does
> not drop privs before invoking the shell. This has been known for some
> time in some circles at least, but as I said the problem would really
> be in the non-priv-dropping privileged program, that's why most people
> did not really care that much. Last year there was a vuln that is very
> much related to this subject:
> http://blog.cmpxchg8b.com/2013/08/security-debianisms.html

We already knew that this bug was known by the Bash developers.

> Correct me if I'm wrong, but even in that case there is another "help"
> measure that has been implemented at least in linux kernels > 3.1:
> http://lxr.free-electrons.com/source/kernel/sys.c?v=3.1#L628
>
> Therefore setuid calls do not fail anymore even in the case of
> existing resource limits for processes (in linux).

You can still exploit this in the 2.6.x Linux kernel. The 2.6.x versions
are still in widespread use. (Red Hat Enterprise Linux version 6.5, released
a few time ago, is based on version 2.6.32. Possibly Red Hat changed the
RLIMIT_NPROC behavior, but there are other 2.6.x-based Linux distributions also.)

> But in any case, for the sake of correctness I agree that the
> drop_priv code should be fixed (or just completely removed...).

I agree but If finally they decide to remove the code it would seems
as a consequence of the disclosure. Right now it has more sense to fix
the bug. This is because this vulnerability (thanks to "help" measure in
the kernel) is more difficult to exploit. So, the drop privilege code
has more sense nowadays than when was initially coded.

> 2014-06-03 16:16 GMT+02:00 Hector Marco <hecmargi () upv es>:
> > Hi everyone,
> >
> > Recently we discovered a bug in bash. After some time after reporting
> > it to bash developers, it has not been fixed.
> >
> > We think that this is a security issue because in some circumstances
> > the bash security feature could be bypassed allowing the bash to be a
> > valid target shell in an attack.
> >
> > We strongly recommend to patch your bash code.
> >
> > Why don't fix this bug by simple adding mandatory "if" clause ?
> > Any comments about this issue are welcomed.
> >
> >
> > Details at:
> > http://hmarco.org/bugs/bash_4.3-setuid-bug.html
> >
> >
> >
> > Thanks you,
> >
> > Hector Marco
> > http://hmarco.org