Volume deletion can write lots of data if device node is missing

[Eric Harney]

If something happens to the LVM stack so that the /dev/ node for a volume is missing, and a user deletes the volume from Cinder, the volume wiping code will run dd from /dev/zero to a file that does not exist on /dev/, putting GBs of data onto the file system.

There should be a check that the device node exists before copying /dev/zero to it.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/14631

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/14631
Committed: http://github.com/openstack/cinder/commit/8b58a4605f83de6a6286ac9de2634dd3269c22b1
Submitter: Jenkins
Branch: master

commit 8b58a4605f83de6a6286ac9de2634dd3269c22b1
Author: Eric Harney <email address hidden>
Date: Mon Oct 22 18:02:17 2012 -0400

    Ensure device node exists before wiping during volume deletion

    If the LVM device node is missing when a Cinder volume is deleted,
    a large amount of data will be written to /dev/. Check that the
    device node exists before performing a _copy_volume('/dev/zero'...)
    to wipe it. Fixes bug 1070023.

    Change-Id: I77a0a5545a54a4adc905a781ed8120db7adccc27