On Tue, Oct 08, 2013 at 08:50:45AM +0900, Jorgen Lundman wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> > All in all being able to simplify GFS would help a lot with ZFS
Received: from [91.121.88.72] (helo=mail.dawidek.net)
by node002.open-zfs.net
with esmtp (HybridCluster distributed mail proxy)
(envelope-from <[email protected]>); Tue, 08 Oct 2013 11:32:11 -0000
> > portability. I remember ZFS-on-FUSE implementation didn't have .zfs/
> > support (probably for other reasons too). The same for initial MacOS X
> > port.
> >
>
> Amusingly, we have started the gfs & .zfs port, and it has been quite the
> challenge. The use of nice consistent macros (like that of VTOZ) went
> away, and it uses direct vnode access. The ".." reference seems to be a
> little wrong, often goes to system root.
>
> We too had to cut out the rmdir/mkdir alias to snapshot commands, and
> even trickier is that we can not call "mount" from kernel. Apple do not
> export any of the mount APIs, has the request has to come from userland.
>
> At the moment we may leave .zfs directories out as it is still not quite
> correct.In my opinion automouting snapshots under .zfs/ is a bit of a hack. There are some obvious problems with it: - Snapshots are not unmounted automatically. - Unprivileged users can trigger file system mount which leads to a bit of hacking around in the kernel (as we can't use current context credentials). - NFS exports. Nobody wants to NFS mount every single snapshot on the client and exporting file system and its snapshot as one export can lead to inode collisions. Having snapshots to be handled within ZFS more natively, ie. as integral part of the file system content and not separate mountpoints would make it much more intuitive and user-friendly. Not sure about implementation complexity. -- Pawel Jakub Dawidek http://www.wheelsystems.com FreeBSD committer http://www.FreeBSD.org Am I Evil? Yes, I Am! http://mobter.com
pgpe7LtS7hglC.pgp
Description: PGP signature
_______________________________________________ developer mailing list [email protected] http://lists.open-zfs.org/mailman/listinfo/developer
