[PATCH] lock exclusively in collect_mounts() and drop_collected_mounts()

Taking namespace_sem shared there isn't worth the trouble, especially with
vfsmount ID allocation about to be added.  That way we know that umount_tree(),
copy_tree() and clone_mnt() are _always_ serialized by namespace_sem.
umount_tree() still needs vfsmount_lock (it manipulates hash chains, among
other things), but that's a separate story.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

fs/namespace.c

@@ -1091,20 +1091,20 @@ Enomem:
 struct vfsmount *collect_mounts(struct vfsmount *mnt, struct dentry *dentry)
 {
 	struct vfsmount *tree;
-	down_read(&namespace_sem);
+	down_write(&namespace_sem);
 	tree = copy_tree(mnt, dentry, CL_COPY_ALL | CL_PRIVATE);
-	up_read(&namespace_sem);
+	up_write(&namespace_sem);
 	return tree;
 }
 
 void drop_collected_mounts(struct vfsmount *mnt)
 {
 	LIST_HEAD(umount_list);
-	down_read(&namespace_sem);
+	down_write(&namespace_sem);
 	spin_lock(&vfsmount_lock);
 	umount_tree(mnt, 0, &umount_list);
 	spin_unlock(&vfsmount_lock);
-	up_read(&namespace_sem);
+	up_write(&namespace_sem);
 	release_mounts(&umount_list);
 }