Coverity has picked up a potential deadlock in the local_flush codepath. The issue is that local_flush acquires the flush_lock and we apparently acquire it again in local_flush_one.

From looking at the Coverity issue, it looked like an entirely different lock (it's complaining that we're locking something in the in-memory file system). But I'd like someone more familiar with this code to take a look at it and verify that it's a false positive.

CC: donald.anderson