Re: [PATCH 03/33] teach move_mount(2) to work with OPEN_TREE_CLONE [ver #11]

From: Alan Jenkins
Date: Fri Oct 12 2018 - 10:25:49 EST


On 01/08/2018 16:24, David Howells wrote:
From: Al Viro<viro@xxxxxxxxxxxxxxxxxx>

Allow a detached tree created by open_tree(..., OPEN_TREE_CLONE) to be
attached by move_mount(2).

If by the time of final fput() of OPEN_TREE_CLONE-opened file its tree is
not detached anymore, it won't be dissolved. move_mount(2) is adjusted
to handle detached source.

That gives us equivalents of mount --bind and mount --rbind.

Signed-off-by: Al Viro<viro@xxxxxxxxxxxxxxxxxx>
Signed-off-by: David Howells<dhowells@xxxxxxxxxx>
---

fs/namespace.c | 26 ++++++++++++++++++++------
1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/fs/namespace.c b/fs/namespace.c
index e2934a4f342b..3981fd7b13f5 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c

@@ -2464,10 +2467,19 @@ static int do_move_mount(struct path *old_path, struct path *new_path)
p = real_mount(new_path->mnt);
err = -EINVAL;
- if (!check_mnt(p) || !check_mnt(old))
+ /* The mountpoint must be in our namespace. */
+ if (!check_mnt(p))
+ goto out1;
+ /* The thing moved should be either ours or completely unattached. */
+ if (old->mnt_ns && !check_mnt(old))
goto out1;
- if (!mnt_has_parent(old))
+ attached = mnt_has_parent(old);
+ /*
+ * We need to allow open_tree(OPEN_TREE_CLONE) followed by
+ * move_mount(), but mustn't allow "/" to be moved.
+ */
+ if (old->mnt_ns && !attached)
goto out1;


Technically, I think the comment should say "mustn't allow rootfs to be moved". "rootfs", as in Documentation/filesystems/ramfs-rootfs-initramfs.txt. Moving "/" is allowed.

# unshare -m
# cd /mnt
# mount -ttmpfs none /
# mount --move --no-canonicalize /.. /mnt
#

(Or if you want to quibble about "/.." v.s. "/" -

# unshare -m
# cd /mnt
# mount --rbind / /
# chroot --skip-chdir /..
# mount --move --no-canonicalize / .