[SCM] Samba Shared Repository - branch master updated

[Jeremy Allison]

On Tue, Dec 17, 2013 at 09:45:24AM +0100, Volker Lendecke wrote:
> On Mon, Dec 16, 2013 at 10:34:46AM -0800, Jeremy Allison wrote:
> > On Sat, Dec 14, 2013 at 01:20:04PM +0100, Stefan Metzmacher wrote:
> > > 
> > > commit 5baa7402ba8eebd9b2ddc6b259ae9bb2852b4bb1
> > > Author: Volker Lendecke <vl at samba.org>
> > > Date:   Fri Dec 6 14:34:05 2013 +0000
> > > 
> > >     smbd: Implement and use full_path_tos
> > >     
> > >     Yes, this looks like a hack, but talloc_asprintf does show up high in
> > >     profiles called from these routines
> > >     
> > >     Signed-off-by: Volker Lendecke <vl at samba.org>
> > >     Reviewed-by: Stefan Metzmacher <metze at samba.org>
> > 
> > Hahaha ! Coming this summer to a cinema near you :
> > 
> > "Samba 4 - Revenge of the pstrings" :-) :-).
> 
> Sorry for that :-)

I don't mind. I just think it's funny :-).

> > Great work though. I take it this was a cachegrind
> > find ?
> 
> Yes, and the sum of all those changes is significantly
> measurable using netbench even on a reasonably powerful
> server maching, not only on the Raspberry Pi. It was not the
> talloc mainly here, it was the double run of snprintf that
> bit here I think. So talloc with explicit strlen + memcpy
> would have gained us quite a bit as well. But while there I
> thought why not remove the talloc also.

Very nice work. Have you tried using Intel NASPT as
a load generator as well ? I might try and get this
going over the Christmas holidays and take a look at
some bottlenecks myself :-).

Cheers,

	Jeremy.