Required to protect the escalate/de-escalate read lock process on GFS (and potentially other) mounted filesystems.

Why we need this: While GFS supports cluster-wide locking via flock(), it's implementation isn't quite what it should be. The locking semantics that work consistently on a local filesystem tend to fail on GFS mounted filesystems. This appears to be a design defect in the implementation of GFS. How this manifests itself is that conditional promotion of a shared lock to exclusive will always fail, lock release requests are honored but not immediately processed (causing erratic failures of subsequent conditional requests) and the releasing of the exclusive lock before the shared lock is set when a lock is demoted (which can open a window of opportunity for another process to gain an exclusive lock when it shoudln't be allowed to).