Re: Feature Request: disallow world-writable files in chmod

[Joseph D. Wagner]

On 06/27/2013 1:03 pm, Ben Lentz wrote:

> > And the kernel devs would never allow it. You may still want to patch
> > your local systems, either chmod or the kernel. However, this will 
> > not
> > be accepted upstream.
>
> My suggestion was merely meant to insight thought in the user
> attempting to set files world-writable, perhaps triggering a
> discussion with their system administrator about proper usage of
>
> I certainly don't aim to remove S_WOTH from the kernel; I certainly
> don't think chmod providing some resistance or objection to being used
> to set this bit would be a bad thing.
>
> Maybe the world isn't ready yet. Oh well, thanks anyway I guess.

Sorry, I wasn't trying to be dismissive.  I was trying to point you
to a better direction to take this discussion.  To the contrary, I
think your idea has some merits in certain use-cases, especially on
locked-down systems.

If I could waive a magic wand and implement this any way I wanted,
I would create it as a optional (bool) selinux policy which prevents
members of users_u from adding o+w.  The advantages of doing it this
way:

 1. Truly secure.  Can't be bypassed by another program.
 2. You would have fine-grained control of who's locked out by
    managing their groups -- users_u, staff_u, root_u, or
    unconfined_u.
 3. You wouldn't need to be root in order to do it.  You could be
    a non-root member of staff_u.

However, I don't know if the kernel has selinux checks at that
particular point in the codepath to support such a policy. Perhaps
the selinux guys would already know the answer to this.

Joseph D. Wagner