Description of problem:

fuse-encfs does not seem to implement flocks properly.  It allows multiple
processes to acquire LOCK_EX flocks on the same file.  Additionally, multiple
locks can cause encfs mount to "freeze" - all processes attempting an IO on any
file in encfs filesystem are blocked and uninterruptible.


Version-Release number of selected component (if applicable):

$ rpm -q fuse fuse-encfs
fuse-2.7.3-2.fc8
fuse-encfs-1.4.1.1-1.fc8


Steps to Reproduce:
1. Create encfs filesystem / mount.
2. Compile attached program.
3. Run it inside encfs mount as: ./encfs-flock & ./encfs-flock & ./encfs-flock
4. Try e.g. ls encfsdir
5. Do: killall encfs-flock ; killall ulockmgr_server  to bring encfs mount back
to life.


Actual results:
Multiple locks acquired, encfs mount IO blocked.


Expected results:
No blocked IO, no multiple locks.


Additional info:

Previous F8 fuse-encfs version was not affected by this problem.

Tested also with fuse-sshfs-1.9-2.fc8, problem does not occur there, which may
imply this is encfs issue rather than fuse issue.

This problem was spotted when running autotools inside encfs mount.  You should
be able to reproduce with rpmbuild rebuild of any SRPM that calls autoconf /
autoreconf in %prep (e.g. openssh, cups).