On 17 January 2014 13:38, Chris Murphy <lists(a)colorremedies.com> wrote:
On Jan 17, 2014, at 6:06 AM, James Hogarth <james.hogarth(a)gmail.com>
wrote:
So far as I understand it when in the same pool cross-subvolume COW should
be valid (as shown in /data on the server) - even if technically under
separate mounts.
I haven't yet found anything newer than this:
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id...
The clone can be in different subvolumes but the source and destination
must be the same mount point.
There's a similar effect with btrfs send that might be related, and it has
patches upstream.
http://www.spinics.net/lists/linux-btrfs/msg30852.html
Thanks that commit does indeed match my behaviour ... even if it seems a
bit messed up all things considered ...
I might use a Fedora Live instance and remove the /home mount point whilst
maintaining as a subvolume within root ...
In fact given this behaviour I'm thinking that a bug against anaconda might
be warranted to make home a subvolume (for easy snapshotting etc) but to
not explicitly mount it to get the maximum flexibility...
James