SMBFS driver has locking issues

[Alessandro Pilotti]

The SMBFS uses a single lock for every volume and snapshot operations.
This leads to bottlenecks and starvation during tempest runs.

[Eric Harney (eharney)]

Patch: https://review.openstack.org/#/c/169858/

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/169858
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=acf19befe59614d15b89ea4be1217de2d1712f3b
Submitter: Jenkins
Branch: master

commit acf19befe59614d15b89ea4be1217de2d1712f3b
Author: Alessandro Pilotti <email address hidden>
Date: Wed Apr 1 20:57:14 2015 +0300

    SMBFS: Lock on a per-volume basis

    This allows operations that do not conflict with each
    other (i.e. are on different volumes) to run concurrently.
    The prior locking scheme was too coarse and essentially
    made the driver single-threaded.

    This patch moves the implementation of the GlusterFS driver locking
    scheme to the RemoteFS base driver so that other similar volume
    drivers can use it.

    Closes-Bug: #1439352
    Change-Id: I37be14e7525406e0ad568f6228c1303998c5235f