["ronnie sahlberg" <ronniesahlberg@xxxxxxxxx>]

can you put the tcp and nfs dissector s back one by one and see which
one of them is slowing it down?


On 3/10/06, ronnie sahlberg <ronniesahlberg@xxxxxxxxx> wrote:
> ok   can you test with  the latest SVN    there are two defines in
> emem.c   you should remove    WANT_GUARD_PAGES   and
> DEBUG_USE_CANARIES.
>
>
> comment both of them out   and pelase test how   slow it becomes.
>
>
>
> On 3/10/06, didier <dgautheron@xxxxxxxx> wrote:
> > Hi,
> > ronnie sahlberg wrote:
> > > that is very odd.    can you verify by reversing the nfs patch to see
> > > that it was indeed changing to a tree that caused this slowdown?
> > Both times are with svn 17537 (I think, version.pl is broken if you
> > don't build in the source tree) .
> > 50s with packet-tcp.c and packet-nfs.c back to hash.
> > >
> > > in my initial test there was no slowdown.
> > >
> > >
> > > ill test with some more large nfs traces.
> > >
> > > (can you make that nfs trace available to me for testing?)
> > It's a make clean; make
> > with the source tree (not the build dir) nfs mounted on loopback UDP,
> > NFS 3, linux 2.4.something, captured with tcpdump -i any -s 0 -w
> foo.pcap.
> > all nfs prefs set.
> > It's big :)
> >
> > Didier
> >
> > _______________________________________________
> > Ethereal-dev mailing list
> > Ethereal-dev@xxxxxxxxxxxx
> > http://www.ethereal.com/mailman/listinfo/ethereal-dev
> >
>