Latest leases patchset - getting there !

[Jeremy Allison]

On Thu, Nov 20, 2014 at 04:48:23PM -0800, Jeremy Allison wrote:
> On Tue, Nov 18, 2014 at 04:15:02PM -0800, Jeremy Allison wrote:
> > 
> > I tidied up this code. Here is the complete patchset
> > I'm working with that goes on top of master (for
> > OEMs who might be following along :-).
> > 
> > It includes extra tests for dynamic shares,
> > leases+writes and leases+byte range locks.
> > 
> > I'll do full make tests's tomorrow, and I
> > thought you had some more work you wanted
> > to do on smb2.lease.breaking2 and
> > smb2.lease.breaking4, but other than that
> > I think we need to start merging this into
> > a patchset we can apply on master soon.
> > 
> > After all we don't want to slip 4.2rc3
> > again do we :-) :-).

OK, I'm testing breaking4 by removing
the commented out code, and testing
directly against W2K12.

W2K12 doesn't do what the test expects
it to do so I need to understand what
you're trying to test here. Your code
does:

Create LEASE1(R)		->
				<-	H1,LEASE1(R)
Create Lease2(RH)		->
				<-	H2,LEASE2(RH)
Create-Overwrite (NO OPLOCK)	->

What W2K12 does here is:

				<- H3 (NO OPLOCK)
immediately followed by:
				<- Break to none (LEASE2), ack required.
				<- Break to none (LEASE1), No ack required.

The test code seems to expect that the reply
to the Create-Overwrite (NO OPLOCK) would
be deferred until the reply to the

Break to none (LEASE2), ack required

packet is received (which to be honest
I'd expect too :-), but that's not
how Windows behaves.

I'm pretty comfortable with how our
lease code is behaving now w.r.t. conflicting
non-lease opens, so I'm going to start
squashing some of the commits into a more
coherent patchset, in the hope we can
get this merged asap.

Cheers,

Jeremy.