Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 28 Oct 2003 05:01:03 +0000
From:      Bruce M Simpson <bms@spc.org>
To:        Mike Silbersack <silby@silby.com>
Cc:        freebsd-net@freebsd.org
Subject:   Re: Changes to PCBPORTHASH wrt TCP, review needed
Message-ID:  <20031028050103.GA7279@saboteur.dek.spc.org>
In-Reply-To: <20031027014854.K2023@odysseus.silby.com>
References:  <20031027014854.K2023@odysseus.silby.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Oct 27, 2003 at 02:16:13AM -0600, Mike Silbersack wrote:
> One easy way to test this patch is to install http_load, set your
> ephemeral port range to something in the range of 30, and have it start
> testing a host.  It will quickly create TIME_WAIT sockets filling all
> ephemeral ports.  Without this patch, you will be unable to create
> outgoing connections; with this patch, other outgoing connections will be
> fine.

I can confirm I can replicate this behaviour with 5.1-CURRENT and 5.1-RELEASE
with http_load.

We discussed on IRC that this problem of ephemeral port hash mapping may also
affect udp PCBs, and that it may be having undesirable effects with multiple
concurrent media streams, as RTP/RTCP is a heavy udp socket consumer in a
large installation, and has specific requirements for binding consecutive
odd/even port pairs (more details in RFC 3550 for those who care).

I will test the patch shortly. I have looked over it and can't find any
immediate problems with it, but further digestion on my part is required.

BMS



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031028050103.GA7279>