chroot+overlayfs seems to cause umount mis-behavior

[Serge Hallyn]

If you mount an overlayfs under a loopback fs, chroot into it and mount something, then the loopback device cannot be released. In quantal it could, but in raring it hangs. I will attach a script which triggers the bug (in raring).

It doesn't happen with aufs; it doesn't happen if you mount without chrooting or pivot-rooting; it doesn't happen if you chroot but don't do any mounting.

[Serge Hallyn (serge-hallyn)]

Marking confirmed as it's been seen by >1 people (stgraber and myself).

[Joseph Salisbury (jsalisbury)]

Can you see if this also happens with the latest mainline kernel, which can be downloaded from:
http://kernel.ubuntu.com/~kernel-ppa/mainline/v3.9-rc3-raring/

[Stéphane Graber (stgraber)]

Thanks for the automated comment ;) Overlayfs isn't a mainline feature so we obviously can't test that kernel.

[Joseph Salisbury (jsalisbury)]

Does this issue still exist in Raring with the latest updates? Also, can you see if the bug also exists in Saucy?

[Serge Hallyn (serge-hallyn)]

both raring and saucy still do this.

[Andy Whitcroft (apw)]

@serge -- after a lot of debugging I think we have this one nailed. Could you test out the kernels for raring and saucy below and confirm the resolve the umount issue:

    http://people.canonical.com/~apw/lp1098378-raring/
    http://people.canonical.com/~apw/lp1098378-saucy/

Please report your testing here.

[Serge Hallyn (serge-hallyn)]

Thanks Andy, that fixes it! (Tested the raring kernel).

I haven't explicitly tested in quantal, but suspect it was new in raring.

[Serge Hallyn (serge-hallyn)]

Indeed, quantal is not affected.

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 3.9.0-5.12

---------------
linux (3.9.0-5.12) saucy; urgency=low

  [ Andy Whitcroft ]

  * Revert "HAVE_CPLUS_DEMANGLE=n"
  * SAUCE: intel_pstate -- toggle default to disable
    - LP: #1188647
  * SAUCE: ubuntu: overlayfs -- ovl_path_open should not take path
    reference
    - LP: #1098378
 -- Andy Whitcroft <email address hidden> Wed, 12 Jun 2013 13:28:56 +0100

[Ed Hutchins (eh-s)]

We're seeing something very similar to this bug in 12.04LTS (kernel 3.2.0-49-generic) running LXC containers on overlayfs mounts. In our case a front-end script we're using to run tests in isolation successfully umounts the overlayfs after the contained test completes, but attempting to rmdir the mountpoint fails. There are no processes left keeping the mount busy, and the umount completes without complaint. Interestingly after a few seconds (5 - 20) the mount is no longer busy, and sometimes everything works without incident. I've looked at the overlayfs source in the kernel and it seems to have the same issue as above; any chance this can patch can be back-ported?

[Joseph Salisbury (jsalisbury)]

This bug was nominated against a series that is no longer supported, ie raring. The bug task representing the raring nomination is being closed as Won't Fix.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.