GFS2: Instruct DLM to avoid queue convert slowdown