From mboxrd@z Thu Jan  1 00:00:00 1970
From: Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: Re: ip_nonlocal_bind and sendto (fwd)
Date: Fri, 27 Feb 2009 11:50:15 -0300
Message-ID: <20090227145015.GG22152@ghostprotocols.net>
References: <alpine.DEB.1.10.0902270119060.7158@asgard.lang.hm>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: netdev@vger.kernel.org
To: david@lang.hm
Return-path: <netdev-owner@vger.kernel.org>
Received: from mx2.redhat.com ([66.187.237.31]:33705 "EHLO mx2.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1754753AbZB0OuU (ORCPT <rfc822;netdev@vger.kernel.org>);
	Fri, 27 Feb 2009 09:50:20 -0500
Content-Disposition: inline
In-Reply-To: <alpine.DEB.1.10.0902270119060.7158@asgard.lang.hm>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

Em Fri, Feb 27, 2009 at 01:20:07AM -0800, david@lang.hm escreveu:
> I've got a need to forge the source IP of UDP packets (a stupid syslog 
> receiver app that I need to deal with)
>
> I know that this can be done with raw sockets, but with the 
> IP_nonlocal_bind option I thought that it may be possible to do this 
> without dealing with raw sockets
>
> so I have an app that does the socket call, followed by the bind call 
> without any errors, but when I issue the sendto call it generates error 
> 22 (invalid parameter) if the source IP doesn't exist on the local box 
> somewhere.
>
> is this the kernel that is doing whatever check is failing? or is glibc  
> wrapping the kernel syscall and doing some additional checking?
>
> if it is the kernel that's throwing the error, is there some way of 
> disabling this check? or do I have to go to raw sockets?

Follow:

udp_sendmsg
    ip_route_output_flow
        __ip_route_output_key

And then look at this line:

	rth->fl.fl4_src == flp->fl4_src &&

It'll fail and so another function will be called:

	    ip_route_output_slow

Many, many details later, this comment:

                           Because we are allowed to send to iface
                           even if it has NO routes and NO assigned
                           addresses. When oif is specified, routing
                           tables are looked up with only one purpose:
                           to catch if destination is gatewayed, rather than
                           direct. Moreover, if MSG_DONTROUTE is set,
                           we send packet, ignoring both routing tables
                           and ifaddr state. --ANK

So try using:

setsockopt(udp_sock, SOL_SOCKET, SO_BINDTODEVICE, "eth0", 5);

Replacing "eth0" with the interface you want your forged packets to fly by.

The device name will be transformed into an index and stored in
sk->sk_bound_dev_if, that later, at udp_sendmsg time, will get stashed into
that oif field and then that comment will hold.

Caveat emptor: I never tried this, this is just after 5 minutes code inspection 8)

- Arnaldo