[patch 0/8] unprivileged mount syscall

[Serge E. Hallyn]

Quoting Miklos Szeredi (miklos at szeredi.hu):
> > > > Agreed on desired behavior, but not on chroot sufficing.  It actually
> > > > sounds like you want exactly what was outlined in the OLS paper.
> > > > 
> > > > Users still need to be in a different mounts namespace from the admin
> > > > user so long as we consider the deluser and backup problems
> > > 
> > > I don't think it matters, because /share/$USER duplicates a part or
> > > the whole of the user's namespace.
> > > 
> > > So backup would have to be taught about /share anyway, and deluser
> > > operates on /home/$USER and not on /share/*, so there shouldn't be any
> > > problem.
> > 
> > In what I was thinking of, /share/$USER is bind mounted to
> > ~$USER/share, so it would have to be done in a private namespace in
> > order for deluser to not be tricked.
> 
> But /share/$USER is surely not bind mounted to ~$USER/share in the
> _global_ namespace, is it?  I can't see any sense in that.

No it's not, only in the private namespace.

> > > There's actually very little difference between rbind+chroot, and
> > > CLONE_NEWNS.  In a private namespace:
> > > 
> > >   1) when no more processes reference the namespace, the tree will be
> > >     disbanded
> > > 
> > >   2) the mount tree won't be accessible from outside the namespace
> > 
> > But it *can* be, if properly set up.  That's part of the point of the
> > example in the OLS paper.  When a user logs in, sshd clones a new
> > namespace, then bind-mounts /share/$USER into ~$USER/share.  So assuming
> > that /share/$USER was --make-shared'd, it and ~$USER are now in the
> > same peer group, and any changes made by the user under ~$USER will
> > be reflected back into /share/$USER.
> 
> I acknowledge, that it can be done.  My point was that it can be done
> more simply _without_ using CLONE_NS.

Seems like a matter of preference, but I see what you're saying.

> > > Wanting a persistent namespace contradicts 1).
> > 
> > Not necessarily, see above.
> > 
> > > Wanting a per-user (as opposed to per-session) namespace contradicts
> > > 2).  The namespace _has_ to be accessible from outside, so that a new
> > > session can access/copy it.
> > 
> > Again, I *think* you are wrong that private namespace contradicts this
> > requirement.
> 
> I'm not saying there's any contradiction, I'm saying rbind+chroot is a
> better fit.

Ok, I see.

> I haven't yet heard a single reason why a per-session namespace with
> parts shared per-user is better than just a per-user namespace.

In fact I suspect we could show that they are functionally equivalent
(for your purposes) by drawing the fs tree and peer groups from
current->fs->root on up for both methods.

And not using private namespaces leaves the admin (at least for now)
better able to diagnose the state of the system.

-serge