vfs: Fix vfsmount_lock imbalance in path_init()

When backporting commit 4023bfc9 ("be careful with nd->inode in
path_init() and follow_dotdot_rcu()"), I failed to account for the
vfsmount_lock that is used in 3.2 but not upstream.  path_init() takes
the lock if performing RCU lookup, but must drop it if (and only if)
it subsequently fails.

Reported-by: nuxi@vault24.org
References: https://bugzilla.kernel.org/show_bug.cgi?id=92531Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Tested-by: nuxi@vault24.org

fs/namei.c

@@ -1567,6 +1567,7 @@ static int path_init(int dfd, const char *name, unsigned int flags,
 	if (!(nd->flags & LOOKUP_ROOT))
 		nd->root.mnt = NULL;
 	rcu_read_unlock();
+	br_read_unlock(vfsmount_lock);
 	return -ECHILD;
 
 fput_fail: