Patch: Implement FSCTL_LMR_REQUEST_RESILIENCY

Jeremy Allison jra at samba.org
Fri Oct 10 09:55:40 MDT 2014


On Fri, Oct 10, 2014 at 08:02:44AM +0200, Michael Adam wrote:
> On 2014-10-09 at 16:50 -0700, Jeremy Allison wrote:

> > Unless you talk about them on the list,
> > no one knows they exist, and no one knows
> > what is being done there.
> > 
> > They might as well be vaporware :-).
> 
> Er, I'm talking about them. On the list. Ain't I?

Well *NOW* you are :-). But lots of work
goes on in there that unless you know about
them and check them out you don't know about.

> > Actually, you need to make sure everything
> > is in master, so others can play with it
> > (sorry if it already is, been working on
> > other stuff at the moment :-).
> 
> No, this is humbug. We only need to make sure
> code gets into master once it is ready. And
> as I said before, we have been and are putting those
> parts into master that were ready. Putting
> WIP/POC code into master prematurely is
> the wrong move. This is what personal repos
> are for: letting code ripen until it is ready.

Ok. It's a fair cop... :-). I take that back.
I don't really want code in master that isn't
ready (yes Simo, I agree with you too :-),
I just hate not knowing what's going on in
all branches.

I'm just a control freak at heart, with
nostalgia for the old days where I used
to know (and understand) everything that
was going on in Samba :-).

But master isn't the place for experiments,
I agree :-).

> > We can work up to this slowly I think.
> > We certainly need to be able to cope
> > with a closed fd being reopened.
> 
> We do that already with durable handle reconnect since 4.0.

Yeah, I knew about that. As we
already have the close->reopen model
working it might be easier to build
the more advanced handle support on
top of that rather than fd passing,
for now.

I understand the "no guarantees from the underlying
OS once we close the fd" problem, and
ultimately having the fd remain
open is best - but the guarantees
we need are so above and beyond what
even Linux can provide at the moment
(w.r.t. leases etc.) that making these
features "Samba share accessible only"
isn't such a big deal IMHO.

Understand I'm just brainstorming here,
and if it turns out to be easier to
build on top of the existing fd passing code,
then just go for it and ignore me :-).

Jeremy.


More information about the samba-technical mailing list