Date: Mon, 23 Nov 2009 18:15:45 -0500
From: Jeff Layton <jlayton@redhat.com>
To: Jamie Lokier <jamie@shareable.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>, linux-fsdevel@vger.kernel.org,
       linux-kernel@vger.kernel.org, pavel@ucw.cz, miklos@szeredi.hu,
       viro@ZenIV.linux.org.uk
Subject: Re: [PATCH 0/3] vfs: plug some holes involving LAST_BIND symlinks
 and file bind mounts (try #5)
Message-ID: <20091123181545.05ad004d@tlielax.poochiereds.net>
In-Reply-To: <20091123224948.GB5598@shareable.org>
References: <1258998084-26797-1-git-send-email-jlayton@redhat.com>
	<m1my2cdhi3.fsf@fess.ebiederm.org>
	<20091123173616.75c3f600@tlielax.poochiereds.net>
	<20091123224948.GB5598@shareable.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2136
Lines: 61

On Mon, 23 Nov 2009 22:49:48 +0000
Jamie Lokier <jamie@shareable.org> wrote:

> Jeff Layton wrote:
> > > check_path_accessible seems pretty horrible.   If a process is running
> > > inside of a subdirectory it doesn't have permissions to access, say
> > > a chroot, /proc/self/fd/XXX becomes completely unusable.
> > > 
> > 
> > Hmm...I have this in there:
> > 
> > +		/* are we at global root or root of namespace? */
> > +		if ((tdentry == root.dentry && vfsmnt == root.mnt) ||
> > +		    vfsmnt->mnt_parent == vfsmnt)
> > +			break;
> > 
> > ...In the case of a chroot, wouldn't "current->fs->root" point to the
> > root of the process' namespace? Or am I misunderstanding what
> > current->fs actually represents?
> 
> A process can run inside a subdirectory it doesn't have permissions to
> access without that being a chroot.
> 

Certainly.

> It can also run inside a subdirectory that isn't accessible from it's
> root, if that's how it was started - as well as having other
> descriptors pointing to things outside its root.
> 

Yes.

> It can also be passed file descriptors from outside it's root while
> it's running.
> 

Yep.

> Really, I think the /proc/PID/fd/N check should restrict the open to
> the O_* limitations that were used to open fd N before, and not have
> any connection to actual paths at the time of this check.
> 

The big question with all of this is: Should a task have the ability
to follow a /proc/pid symlink to a path that it wouldn't ordinarily be
able to resolve with a path lookup. The concensus that I got from the
bugtraq discussion was that it should not, and this patch is an attempt
to prevent that.

I take it from you and Eric's comments that you disagree? If so, what's
your rationale for allowing a task to resolve this symlink when it
wouldn't ordinarily be able to do so if it were a "normal" symlink?

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