From owner-freebsd-hackers  Wed Jan 29 14:47:28 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id OAA27254
          for hackers-outgoing; Wed, 29 Jan 1997 14:47:28 -0800 (PST)
Received: from whistle.com (s205m131.whistle.com [207.76.205.131])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id OAA27246
          for <hackers@freebsd.org>; Wed, 29 Jan 1997 14:47:26 -0800 (PST)
Received: (from smap@localhost) by whistle.com (8.7.5/8.6.12) id OAA19160; Wed, 29 Jan 1997 14:46:43 -0800 (PST)
Received: from bubba.whistle.com(207.76.205.7) by whistle.com via smap (V1.3)
	id sma019158; Wed Jan 29 14:46:19 1997
Received: (from archie@localhost) by bubba.whistle.com (8.7.5/8.6.12) id OAA24842; Wed, 29 Jan 1997 14:46:19 -0800 (PST)
From: Archie Cobbs <archie@whistle.com>
Message-Id: <199701292246.OAA24842@bubba.whistle.com>
Subject: Re: ipdivert & masqd
In-Reply-To: <199701292130.NAA19395@gatekeeper> from Darren Reed at "Jan 30, 97 08:29:09 am"
To: avalon@coombs.anu.edu.au (Darren Reed)
Date: Wed, 29 Jan 1997 14:46:19 -0800 (PST)
Cc: archie@whistle.com, terry@lambert.org, ari.suutari@ps.carel.fi,
        brian@awfulhak.demon.co.uk, hackers@freebsd.org, cmott@srv.net
X-Mailer: ELM [version 2.4ME+ PL25 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk


> > The theory was that this loop avoidance was working too well, and
> > seemed to be applying to packets other than the one that it was
> > supposed to. What I'm trying to prove to myself is that this can't
> > be happening.
> 
> Does ip_divert_flag get set/reset inside or outside the loop in
> ip_input() which dequeues packets ?  (src isn't handy)

Ah.. well, it doesn't get reset until ip_input() returns. Perhaps
this is the problem... certainly if calling ip_input() with one
packet can trigger the ipfw processing of other packets, that would
be bad.

[ checking source .. ]

>From my reading it doesn't seem like this can happen. Packet fragments
are queued up and then merged when the last packet arrives, but sending
ip_input() a whole separate packet shouldn't trigger this.

-Archie

___________________________________________________________________________
Archie Cobbs   *   Whistle Communications, Inc.  *   http://www.whistle.com