On 19/02/2019 17:08, David Howells wrote:

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.

This is a bit ambiguous.  The two cases can be understood by analogy to 
bind / rbind.  But it is also seems natural, to think it could be used 
to implement the exact same thing as current `mount --bind` / 
`--rbind`.  I think it *does* now provide a full equivalence, right?

I was thinking about the case where mount propagation is enabled on the 
source tree, i.e. it is not a private mount.  Suppose a new mount is 
added inside the source tree, between open_tree() and move_mount().

In the previous version of the patch series, Eric suggested there was a 
NULL dereference in this scenario.[1]  This version should be safe.  I 
think the new mount will be propagated to the cloned tree. Furthermore - 
due to the way this version uses a temporary mount namespace - the 
propagated version of the mount will not be locked by 
attach_recursive_mnt().

[1] https://lore.kernel.org/lkml/87bm7n5k1r.fsf@xmission.com/ (local)

It looks very neat now, with the use of the temporary namespaces. 
Congratulations :-).  I have finished looking through these patches 1-3 now.

Thanks also to Alan Jenkins[off-list ref]  for
providing a whole bunch of ways to break things using this interface.

Signed-off-by: Al Viro<viro@zeniv.linux.org.uk>
Signed-off-by: David Howells<dhowells@redhat.com>
Signed-off-by: Al Viro<viro@zeniv.linux.org.uk>

P.S. I guess Al does not need two Signed-off-by lines here.

Thanks
Alan