Re: [RFC] Implement ambient capability set.

[Andy Lutomirski]

On Wed, Feb 4, 2015 at 1:57 PM, Christoph Lameter <cl@xxxxxxxxx> wrote:
> On Wed, 4 Feb 2015, Andy Lutomirski wrote:
>
>> But someone will want to run *bash* as an untrusted user with, say,
>> CAP_NET_BIND permitted and ambient.  Then that user has a non-empty
>> ambient set, and they can run a setuid-root program, and who knows
>> what will go wrong?  Requiring no_new_privs would prevent this type of
>> failure entirely.
>>
>> If we need to relax that later, it's easy, I think.  The rule's not
>> that convoluted, and there's precedent for having new fancy features
>> require setting no_new_privs first.
>
> It would make the patch pointless. The case of having to run a setuid root
> prpgrams from a shell that has the caps enabled is a routine thing for
> testing etc.
>

That's unfortunate.  In that case, we need to figure out what happens
when you run such a setuid root program.  I think the answer should be
that pA gets cleared, and that pA also gets cleared if you run a
program that has file caps.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/