[lxc-users] pre-mount hook namespace

[Serge Hallyn]

Quoting Wolfgang Bumiller (w.bumiller at proxmox.com):
> > On November 11, 2015 at 6:04 PM Serge Hallyn <serge.hallyn at ubuntu.com> wrote:
> > > > 2.
> > > > If you are just using unpriv containers to use user namespaces, you can
> > > > actually have the container be owned/started by root.  That's what I do
> > > > for some containers where their rootfs is a dmcrypt device which I
> > > > couldn't mount as an unpriv user.
> > > 
> > > They are started as root, which means I can prepare the mounts as you
> > > suggested above, but I'd again be clobbering the host's namespace.
> > 
> > Oh, right.  I forget that even when starting as root, this only works
> > for the rootfs itself, not other mounts.  (Lxd actually does handle this,
> > but at the cost of having a MS_SLAVE mount per container)
> 
> So we ended up doing just that, but now with the latest lxcfs
> upgrades (I suspect cgmanager/cgfs changes) AppArmor suddenly
> denies lxc-start to bind mount something. Here's what happens
> with raw lxc-start commands:
> 
> # lxc-start -n 406
> 
> works, but (simplified to just unshare -m):
> 
> # unshare -m -- lxc-start -n 406
> 
> audit: type=1400 audit(1447670720.554:74): apparmor="DENIED" operation="mount"
> profile="/usr/bin/lxc-start"
> name="/usr/lib/x86_64-linux-gnu/lxc/rootfs/sys/fs/cgroup/hugetlb/lxc/406/"
> pid=21536 comm="lxc-start" flags="rw, bind"
> 
> This doesn't make sense to me, I don't see how the namespace
> change would affect this? (Using unshare -m and then running
> `mount --make-r{slave,private,shared} /` doesn't change the
> outcome.)

Can you make sure that your apparmor profile has the
attach_disconnected flag?