Re: [PATCH v4 5/5] fuse: Allow user namespace mounts

[Seth Forshee <seth.forshee@canonical.com>]

On Wed, Oct 15, 2014 at 07:58:53AM -0700, Andy Lutomirski wrote:
> On 10/14/2014 07:25 AM, Seth Forshee wrote:
> > Cc: Eric W. Biederman <ebiederm@xmission.com>
> > Cc: Serge H. Hallyn <serge.hallyn@ubuntu.com>
> > Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
> > ---
> >  fs/fuse/inode.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
> > index 5e00a6a76049..6522926b14e4 100644
> > --- a/fs/fuse/inode.c
> > +++ b/fs/fuse/inode.c
> > @@ -1212,7 +1212,7 @@ static void fuse_kill_sb_anon(struct super_block *sb)
> >  static struct file_system_type fuse_fs_type = {
> >  	.owner		= THIS_MODULE,
> >  	.name		= "fuse",
> > -	.fs_flags	= FS_HAS_SUBTYPE,
> > +	.fs_flags	= FS_HAS_SUBTYPE | FS_USERNS_MOUNT,
> >  	.mount		= fuse_mount,
> >  	.kill_sb	= fuse_kill_sb_anon,
> >  };
> > @@ -1244,7 +1244,7 @@ static struct file_system_type fuseblk_fs_type = {
> >  	.name		= "fuseblk",
> >  	.mount		= fuse_mount_blk,
> >  	.kill_sb	= fuse_kill_sb_blk,
> > -	.fs_flags	= FS_REQUIRES_DEV | FS_HAS_SUBTYPE,
> > +	.fs_flags	= FS_REQUIRES_DEV | FS_HAS_SUBTYPE | FS_USERNS_MOUNT,
> 
> I think it's decision time -- if these patches are applied, then FUSE
> will be the first filesystem for which userns nodev behavior matters for
> security, so applying this patch will enshrine an API decision.
> 
> I would very much prefer to make this patch depend on this:
> 
> http://lkml.kernel.org/g/2686c32f00b14148379e8cfee9c028c794d4aa1a.1407974494.git.luto@amacapital.net
> 
> That change will require that anyone who tries to mount one of these
> things explicitly requests MS_NODEV instead of keeping the current
> behavior of implicitly setting MS_NODEV and possibly confusing user code
> that tries to remount.
> 
> If you like my patch, feel free to fold it in to your series, or Eric
> can apply it directly (pretty please).
> 
> For background, with your patches as is, if you mount a FUSE fs and then
> remount it with identical flags, the remount is likely to fail.

(Resending my response since I still don't see it on lkml after 7+
hours)

I discussed this with Eric during LinuxCon NA ... as I recall he was
undecided about whether or not to use your patch at the time. I do
prefer an explicit failure over implicitly adding MS_NODEV, but it's not
up to me. I do agree though that we should make a decision before
merging the fuse patches, I was just assuming that the decision was
already made.

Thanks,
Seth