From owner-cvs-sys  Mon Oct 27 23:51:22 1997
Return-Path: <owner-cvs-sys>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id XAA05158
          for cvs-sys-outgoing; Mon, 27 Oct 1997 23:51:22 -0800 (PST)
          (envelope-from owner-cvs-sys)
Received: from gvr.gvr.org (root@gvr.gvr.org [194.151.74.97])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id XAA05138;
          Mon, 27 Oct 1997 23:51:10 -0800 (PST)
          (envelope-from guido@gvr.org)
Received: (from guido@localhost)
	by gvr.gvr.org (8.8.6/8.8.5) id IAA14324;
	Tue, 28 Oct 1997 08:51:03 +0100 (MET)
From: Guido van Rooij <guido@gvr.org>
Message-Id: <199710280751.IAA14324@gvr.gvr.org>
Subject: Re: cvs commit: src/sys/netinet ip_input.c
In-Reply-To: <97Oct27.204111pst.177487@crevenia.parc.xerox.com> from Bill Fenner at "Oct 27, 97 08:41:04 pm"
To: fenner@parc.xerox.com (Bill Fenner)
Date: Tue, 28 Oct 1997 08:51:03 +0100 (MET)
Cc: guido@freebsd.org, cvs-committers@freebsd.org, cvs-all@freebsd.org,
        cvs-sys@freebsd.org
X-Mailer: ELM [version 2.4ME+ PL32 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-cvs-sys@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

Bill Fenner wrote:
> Guido van Rooij <guido@FreeBSD.ORG> wrote:
> >  Modified files:
> >    sys/netinet          ip_input.c 
> >  Log:
> >  When dosourcerouting is set do not sourceoute....
> 
> Well, let's be clear about what the semantics are.  Previously, they were:
> 
> "Do not allow yourself to be named as a router in a source-route, but
> forward LSRR packets if you're not one of the routers in the
> source-route, and accept source-routed packets if you're the ultimate
> destination."
> 
> Now, they are:
> 
> "Do not forward any source-routed packets, but still accept
> source-routed packets if you're the ultimate destination".
> 
> Are these the semantics we want?  (My personal feeling is that there
> should be 2 variables, one for forwarding and one for accepting).
> 

Yes. If dosourceroute is NOT set, the machine should never forward
source routes. 

The second variable (accept_sourceroute) will come later; it's
already implemented over here after a deabte with Paul Traina,
Garrett Wollman an David Greenmen.


> However, there is a problem with the actual change: the "goto" bypasses
> the strcpy() which gives log() something to print.  In addition, the whole
> log(LOG_WARNING, ...) for each packet provides a wonderful denial of service
> attack (I understand this isn't new with this commit, but it's still an
> issue).

I will correct the strcpy thing tonight. The denail os service attack is indeed
present, but we do want the logging.
A kludge would be to add yet another sysctl variable, that controls
the logging that is cleared after a number of logs, together with a message
that logging has stopped and can be restarted by resetting the flag.
Another solution is to not allow more then XXX logging entries per minut.

-Guido