[5683] in bugtraq
Re: 44BSD port of land.c
daemon@ATHENA.MIT.EDU (Charles M. Hannum)
Sat Nov 22 20:38:03 1997
Date: Sat, 22 Nov 1997 13:54:57 -0500
Reply-To: "Charles M. Hannum" <mycroft@MIT.EDU>
From: "Charles M. Hannum" <mycroft@MIT.EDU>
X-To: deviant@unixnet.org
To: BUGTRAQ@NETSPACE.ORG
In-Reply-To: Peter's message of Sat, 22 Nov 1997 08:00:52 +0000
Peter <deviant@UNIXNET.ORG> writes:
>
> This seems to work, as long as you've got NetCat:
>
> -----
> #!/bin/bash
> nc -s $1 -p $2 $1 $2
> -----
>
> where $1 is the host, and $2 is the port (139,23, 25, whatever)
This is actually a separate bug. It used to be that in the 4.4BSD
stack (and probably earlier versions) a TCP socket connecting to
itself would cause a SYN war, via a different code path than the
`land' sttack. We fixed this a few years ago in NetBSD, and our fix
for the `land' attack (which I'll post about in a moment) still allows
a socket to connect to itself -- although truthfully I'm not sure how
useful this behaviour really is.
