[PATCH] SMB2_FIND improvement for clustered environments

[Jeremy Allison]

On Thu, Mar 23, 2017 at 06:53:09PM +0100, Ralph Böhme wrote:
> On Thu, Mar 23, 2017 at 06:36:58AM -0700, Jeremy Allison wrote:
> > On Thu, Mar 23, 2017 at 01:02:02PM +0100, Stefan Metzmacher via samba-technical wrote:
> > > Am 23.03.2017 um 12:43 schrieb Volker Lendecke via samba-technical:
> > > > On Thu, Mar 23, 2017 at 11:32:23AM +0100, Ralph Böhme wrote:
> > > >> On Thu, Mar 23, 2017 at 11:09:16AM +0100, Volker Lendecke wrote:
> > > >>> On Thu, Mar 23, 2017 at 10:54:09AM +0100, Ralph Böhme via samba-technical wrote:
> > > >>>> This new feature is still off by default as it doesn't
> > > >>>> give any improvement in the non-clustered usecase.
> > > >>>> "smbd:async search ask sharemode" can be used to activate it,
> > > >>>> which makes only sense with "clustering = yes" (execept for testing).
> > > >>>
> > > >>> Can we always enable this in clustered mode, i.e. not introduce the parameter?
> > > >>
> > > >> sure. I was a bit hesitant, given the size of the change, to enable this by
> > > >> default. But if you and others think we should do this, I'm won't object. :)
> > > > 
> > > > It is a big change, but we won't find the bugs if people don't use it.
> > > > And they won't use it if it's hidden behind a parametric option. I
> > > > know this a difficult decision, but because this does not affect
> > > > functionality at all, only performance, I'm not sure a parameter is
> > > > the right thing to do.
> > > 
> > > Either that or we default it to yes (and may remove it later).
> > 
> > I just went through this code. It's *great*. Very
> > impressive change !
> 
> Thanks!
> 
> > Please turn this on by default. Volker is correct
> > that we will not find all the bugs in this until it is
> > turned on by default.
> 
> What about folding the existing and the new parametric options
> 
>   smbd:search ask sharemode (existing)
>   smbd:async search ask sharemode (new)
> 
> into a full fledged option with extended syntax:
> 
>   smbd:search ask sharemode = yes | sync | async | no (default async)
> 
> yes | sync = same behaviour as current setting of "yes"
> async      = fetch write time, async version
> no         = same behaviour as current setting of "no"
> 
> ? Maybe the "sync" alias is not needed.

I think that's horribly complicated :-). Can we just turn
it on by default, and have an (undocumented) parametric
option to turn if off that we can remove once we're
confident it works correctly ?

Don't want to drop this on the floor :-).