Re: [RFC 1/1] shiftfs: uid/gid shifting bind mount

From: James Bottomley <James.Bottomley@HansenPartnership.com>
To: Amir Goldstein <amir73il@gmail.com>,
	"J. R. Okajima" <hooanon05g@gmail.com>
Cc: Djalal Harouni <tixxdz@gmail.com>, Chris Mason <clm@fb.com>,
	Theodore Tso <tytso@mit.edu>,
	Josh Triplett <josh@joshtriplett.org>,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	Andy Lutomirski <luto@kernel.org>,
	Seth Forshee <seth.forshee@canonical.com>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	LSM List <linux-security-module@vger.kernel.org>,
	Dongsu Park <dongsu@endocode.com>,
	David Herrmann <dh.herrmann@googlemail.com>,
	Miklos Szeredi <mszeredi@redhat.com>,
	Alban Crequy <alban.crequy@gmail.com>,
	Al Viro <viro@zeniv.linux.org.uk>,
	"Serge E. Hallyn" <serge@hallyn.com>,
	Phil Estes <estesp@gmail.com>
Subject: Re: [RFC 1/1] shiftfs: uid/gid shifting bind mount
Date: Mon, 06 Feb 2017 08:29:43 -0800	[thread overview]
Message-ID: <1486398583.2474.61.camel@HansenPartnership.com> (raw)
In-Reply-To: <CAOQ4uxiyPrzvH-JJaqfqeZZ14=QKyQckW=tvaT6PxmoNJTtqRw@mail.gmail.com>

On Mon, 2017-02-06 at 08:38 +0200, Amir Goldstein wrote:
> On Mon, Feb 6, 2017 at 5:25 AM, J. R. Okajima <hooanon05g@gmail.com>
> wrote:
> > James Bottomley:
> > > This allows any subtree to be uid/gid shifted and bound
> > > elsewhere.  It
> >         :::
> > 
> > Interesting.
> > But I am afraid that the inconsistency problem of the inode numbers 
> > will happen.
> > 
> 
> Yet another example that overlayfs already is in the process of 
> solving (it is fixed for stat of merged directory inode).
> In fact, fir the case of single layer overlay (as well as shiftfs) 
> the solution is trivial - preserve underlying inode st_ino/d_ino and 
> use the overlayed fs st_dev.

not sure I follow what st_ino is, do you mean  s_root->d_inode->i_ino?
or did you mean s_dev (which is more traditional)?

The problem with this is there's no way to ensure global uniqueness in
a mapping that goes (ino, ino) -> (ino) (or (s_dev, ino) -> (ino)) and
I believe global uniqueness is more important because the i_ino is used
in the hashed lookups.  Secondly you're not guaranteed that s_root
->d_inode->i_ino is unique ... historically a lot of filesystems use a
well known inode number as the root, that's why filehandles
traditionally used something representing the device and the inode
number (we also have s_dev uniqueness problems for tmpfs which is used
in some overlays).

We can certainly construct a filehandle using an export operations
override that is unique and can be used to lookup the underlying object
(based on the underlying device and inode).

James