vfs: umount_tree() might be called on subtree that had never made it

__mnt_make_shortterm() in there undoes the effect of __mnt_make_longterm()
we'd done back when we set ->mnt_ns non-NULL; it should not be done to
vfsmounts that had never gone through commit_tree() and friends.  Kudos to
lczerner for catching that one...

Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

1 files changed, 2 insertions, 1 deletions

diff --git a/fs/namespace.c b/fs/namespace.c
index 224aff1c0dfd..1e4a5fe3d7b7 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -1074,8 +1074,9 @@ void umount_tree(struct mount *mnt, int propagate, struct list_head *kill)
                 list_del_init(&p->mnt_expire);
                 list_del_init(&p->mnt_list);
                 __touch_mnt_namespace(p->mnt_ns);
+                if (p->mnt_ns)
+                        __mnt_make_shortterm(p);
                 p->mnt_ns = NULL;
-                __mnt_make_shortterm(p);
                 list_del_init(&p->mnt_child);
                 if (mnt_has_parent(p)) {
                         p->mnt_parent->mnt_ghosts++;