ensure libguestfs has completed before proceeding

This avoids a FUSE issue referenced here:
http://code.google.com/p/s3ql/issues/detail?id=159
which in turn references this thread on the subject from 2006:
http://thread.gmane.org/gmane.comp.file-systems.fuse.devel/3903
And another case from 2008:
http://permalink.gmane.org/gmane.comp.file-systems.fuse.devel/6502

The vast majority of this work was reproducing and classifying
done by David Naori and Richard W.M. Jones at:
https://bugzilla.redhat.com/show_bug.cgi?id=835466

Fixes bug: 1013689
Change-Id: I36fd11d5f01562f65a6b6f07e759ea066490b067

nova/virt/disk/guestfs.py

@@ -91,4 +91,16 @@ def unmnt_dev(self):
         # root users don't need a specific unmnt_dev()
         # but ordinary users do
         utils.execute('fusermount', '-u', self.mount_dir, run_as_root=True)
+
+        # Unfortunately FUSE has an issue where it doesn't wait
+        # for processes associated with the mount to terminate.
+        # Therefore we do this manually here.  Note later versions
+        # of guestmount have the --pid-file option to help with this.
+        # Here we check every .2 seconds whether guestmount is finished
+        # but do this for at most 10 seconds.
+        wait_cmd = 'until ! ps -C guestmount -o args= | grep -qF "%s"; '
+        wait_cmd += 'do sleep .2; done'
+        wait_cmd %= self.mount_dir
+        utils.execute('timeout', '10s', 'sh', '-c', wait_cmd)
+
         self.mounted = False