On Mon, Sep 29, 2014 at 11:31:48PM +0100, Steven Hartland via illumos-zfs wrote: > > ----- Original Message ----- > From: "Matthew Ahrens" <[email protected]> > To: "illumos-zfs" <[email protected]>; "Steven Hartland" > <[email protected]> > Cc: "developer" <[email protected]> > Sent: Monday, September 29, 2014 9:52 PM > Subject: Re: [zfs] Possible miss-merge of 4369 implement zfs bookmarks > > > > On Mon, Sep 29, 2014 at 12:49 PM, Steven Hartland via illumos-zfs < > > [email protected]> wrote: > > > >> I've just spotted what looks like a possible miss-merge of: > >> 4369 implement zfs bookmarks > >> > > > > This is not a mis-merge, at least not from DelphixOS, where the patch is > > the same. > > > > > >> > >> It seems we've lost the snap must be in the specified pool > >> check as well as the check for error on unmount which was > >> specifically added by: > >> 3744 zfs shouldn't ignore errors unmounting snapshots > >> > > > > I believe that the semantics should be the same as if that code had been > > left in place. dsl_destroy_snapshot_check() will fail if a snapshot is in > > a different pool. It will also fail if a snapshot is mounted. > > > > Do you have a test case which shows that the current behavior is > > undesirable? > > What actually drew my attention was the loss of the FreeBSD:- > (void) zvol_remove_minors(name);
Jumping in to explain what zvol_remove_minors() really does. In ZFS on Solaris in the past ZVOLs were created immediately. At some point it changed so that now (also in illumos) ZVOLs are created on-demand when a lookup in /dev/ happens. We cannot implement it this way in FreeBSD, so that in various places we create/destroy/rename GEOM providers so they are always available. -- Pawel Jakub Dawidek http://www.wheelsystems.com FreeBSD committer http://www.FreeBSD.org Am I Evil? Yes, I Am! http://mobter.com _______________________________________________ developer mailing list [email protected] http://lists.open-zfs.org/mailman/listinfo/developer
