ports/159102: fusefs daemonizes before mount is available

[Martin Pärtel]

>Number:         159102
>Category:       ports
>Synopsis:       fusefs daemonizes before mount is available
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    freebsd-ports-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Fri Jul 22 08:00:20 UTC 2011
>Closed-Date:
>Last-Modified:
>Originator:     Martin Pärtel
>Release:        Debian GNU/kFreeBSD
>Organization:
>Environment:
GNU/kFreeBSD debian 8.2-1-amd64 #0 Sun Jul 10 02:32:34 CEST 2011 x86_64 amd64 Intel(R) Core(TM)2 Duo CPU     E8500  @ 3.16GHz GNU/kFreeBSD
>Description:
When you start a FUSE filesystem, such as bindfs, with fuse4bsd, then fuse_main() in libfuse daemonizes the process before the mount is usable. This becomes a problem in scripts that first mount a filesystem and then immediately use it. libfuse on Linux does not seem to have this problem.

bindfs sources are available here: http://code.google.com/p/bindfs/downloads/list
and here: https://github.com/mpartel/bindfs
Other FUSE filesystems, such as the example filesystem shipped with it, probably have the same behavior.

>How-To-Repeat:
#!/bin/bash
# Test case to show that when mounting fuse4bsd filesystem like bindfs,
# libfuse daemonizes before the mount becomes usable.
# Since this test case relies on being fast enough, it may be unreliable.

mkdir -p a
mkdir -p b
rm -f a/file
touch a/file

echo "Testing..."
for i in `seq 1 1000`; do
  echo "$i/1000"
  bindfs a b
  #sleep 0.5  # Uncomment this and there should be no fail
  if ! test -e b/file; then
    echo "FAIL. The mount appeared too late."
    sleep 1
    umount b
    exit 1
  fi
  umount b
done

>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted: