Re: [RFC] capabilities: Ambient capabilities

From: Andy Lutomirski
Date: Fri Mar 13 2015 - 13:58:01 EST

On Mar 13, 2015 6:24 AM, "Andrew G. Morgan" <morgan@xxxxxxxxxx> wrote:
> > It's to preserve the invariant that pA is always a subset of pI.
> But since a user can always raise a bit in pI if it is present in pP,
> what does this invariant add to your model other than inconvenience?

The useful part is that dropping a bit from pI also drops it from pA.
To keep the model consistent, I also require that you add the bit to
pI before adding it to pA.

> >> I'm also unclear how you can turn off this new 'feature' for a process
> >> tree? As it is, the code creates an exploit path for a capable (pP !=
> >> 0) program with an exploitable flaw to create a privilege escalation
> >> for an arbitrary child program.
> >
> > Huh? If you exploit the parent, you already win. Yes, if a kiddie
> > injects shellcode that does system("/bin/bash") into some pP != 0
> > program, they don't actually elevate their privileges. On the other
> > hand, by the time an attacker injected shellcode for:
> >
> > system("/bin/bash");
> Let's call the above two lines [a] and [b]. With this patch, you are
> encouraging folk to write programs that contain a line like [a]
> already. So, yes, I am saying that you are creating an exploitable
> path in these programs that says if someone can inject
> system("/bin/bash") into these programs they can get a new (because of
> this patch) privilege escalation.
> In the prevailing model, this kind of privilege escalation (resulting
> from naive inheritance) is designed out. I recognize that you want to
> add it back in, but I am concerned that you are not allowing for the
> possibility that some folk might still want to still be able to
> prevent it.

If you have a program that deliberately uses PR_CAP_AMBIENT, then
setting such a securebit will break the program, so it still doesn't
buy you anything.

> > into a target, they can already do whatever they want.
> >
> >> While I understand that everyone
> >> 'knows what they are doing' in implementing this change, I'm convinced
> >> that folk that are up to no good also do... Why not provide a lockable
> >> secure bit to selectively disable this support?
> >
> > Show me a legitimate use case and I'll gladly implement a secure bit.
> Thanks. I was kind of hoping that you would add a lockable secure bit
> that defaults this support to off, but can be used to turn it on with
> or without a lock. That way, you can avoid disturbing the legacy
> defaults (no surprises).

I think this thing needs to default on to be useful.

> > In the mean time, I don't even believe that there's a legitimate use
> > for any of the other secure bits (except keepcaps, and I don't know
> > why that's a securebit in the first place).
> Those bits currently make it possible to run a subsystem with no
> [set]uid-0 support in its process tree.

Not usefully, because even with all the securebits set to their
non-legacy modes, caps don't inherit, so it doesn't work. I've tried.

> > In the mean time, see CVE-2014-3215 for an example of why securebits
> > are probably more trouble than they're worth.
> I think it is safe to say that naive privilege inheritance has a fair
> track record of being exploited orders of magnitude more frequently
> than this. After all, these are the reasons LD_PRELOAD and shell
> script setuid bits are suppressed.

I don't know what you mean here by naive privilege inheritance. The
examples you're taking about aren't inheritance at all; they're
exploring privilege *grants* during execve. My patch deliberately
leaves grants like that alone.


> Cheers
> Andrew
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at