Re: Spurious RST, segment rejected - possible pattern?

Hugo, good day.

Tue, Jul 24, 2007 at 05:50:17PM +0100, Hugo Silva wrote:
>While conducting some benchmarks against various webserver software, (7-CURRENT
>client vs 7-CURRENT server), I couldn't help notice that the flood of Spurious
>RST (which only happens on the server side) seems to follow a specific pattern.
<...>
> Here's how it goes:
> * webserver restarted _at_ #a
> * benchmark runs and FAILS _at_ #b
> Jul 24 17:31:46 bluebunny kernel: TCP: [192.168.0.1]:60033 to 
> [192.168.0.100]:80 tcpflags 0x4<RST>; tcp_input: Listen socket: Spurious RST, 
> segment rejected
<...more RSTs...>

This topic was raised in the freebsd-net,
    http://lists.freebsd.org/pipermail/freebsd-net/2007-July/014804.html
    http://lists.freebsd.org/pipermail/freebsd-net/2007-July/014808.html
and in the freebsd-current,
    http://lists.freebsd.org/pipermail/freebsd-current/2007-July/074837.html

The issue is understood, but it is currently not publicly known
how to deal with it.  But Andre Oppermann and Robert Watson are
aware of it and Andre told us that he will think how to fix it.

So, stay tuned for the fix ;))

Thank you.
-- 
Eygene