smbtorture/netbench against w2k

Simo Sorce simo.sorce at
Mon Oct 14 21:42:00 GMT 2002

Not tried but the problem is easy to find and we encountered it in our
auth code.
It seem that w2k has some sort of very dumb option active by default
that drops connection if it finds more than one from the same ip under
some cicumstances (mostly when auth is not yet finished and a new
connection starts from the same IP), I do not need to say this stupid
behaviour broke many natted networks.

I'm sure that if you dig into this list you will also find a patch from
microsoft about this problem.

Needles to say that netbench from a single interface use a single ip ...


On Mon, 2002-10-14 at 23:03, Peter Siering wrote:
> Hi,
> I used smbtorture's netbench runs against Samba on Linux and Mac OS X to
> get a clue wich system performs better than the other. Now I've tried the
> same against a Windows 2000 Server, but had no success.
> What happens: w2k seems to have some optimizations regarding client
> connections. If I raise the (emulated) client count in smbtorture's
> netbench w2k starts to refuse connections - that even happens with just a
> few clients (10 or so). It's even dropping an advice in its event log,
> that it could not handle every connection request. I've tried w2k server
> plain and with sp2 with no differences.
> But: If I use a couple of instances of smbtorture on virtual interfaces
> (just tried up to ten both on test client and server) that does not happen
> (as long as the client count per interface stays low). The w2k server
> starts to serve dozens of (emulated) clients that way.
> Did anybody else tried that?
> Peter
Simo Sorce - simo.sorce at
Xsec s.r.l.
via Durando 10 Ed. G - 20158 - Milano
tel. +39 02 2399 7130 - fax: +39 02 700 442 399
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: This is a digitally signed message part
Url :

More information about the samba-technical mailing list