[RFC PATCH RESEND] vfs_ceph: allow mounting a named filesystem

[Jeremy Allison]

On Sat, Mar 28, 2020 at 10:31:35AM -0400, Jeff Layton via samba-technical wrote:
> On Fri, 2020-03-27 at 17:54 +0100, David Disseldorp wrote:
> > Hi Jeff,
> > 
> > On Fri, 27 Mar 2020 12:29:50 -0400, Jeff Layton via samba-technical wrote:
> > 
> > > Signed-off-by: Jeff Layton <jlayton at samba.org>
> > > ---
> > >  docs-xml/manpages/vfs_ceph.8.xml | 16 ++++++++++++++++
> > >  source3/modules/vfs_ceph.c       | 16 ++++++++++++++++
> > >  source3/wscript                  |  6 ++++++
> > >  3 files changed, 38 insertions(+)
> > > 
> > > I had a few spare cycles today and went ahead and plumbed in this
> > > support since Marco had asked about it on the ceph-users list.
> > > 
> > > This builds but is otherwise untested. If it looks reasonable, I can
> > > clean it up and test it out in the near future.
> > 
> > Thanks for the patch!
> > Give that the cmount context may be reused between shares, this would
> > be problematic if the shares are configured with differing fsnames. That
> > said, I think we have the same issue with config_file / user_id, so I
> > don't think it's a showstopper.
> > Otherwise, the change looks reasonable.
> > 
> > Cheers, David
> 
> 
> Oof, yeah...I had forgotten that cmount is global. Ideally we'd have
> some sort of pool of cmounts or create one on a per-service basis.
> 
> Is there some sort of per-share object in the vfs plugin interface off
> which you could hang a cmount?

Yes, you can hang a private pointer off the connection (per share
instance) handle.