nat-traverse(1) Use of UDP to traverse NAT gateways

SYNOPSIS

To create a simple text-only tunnel, use the commands


[email protected] $ nat-traverse 40000:natgw-of-right:40001
[email protected] $ nat-traverse 40001:natgw-of-left:40000

where 40000 is an unused UDP port on "left" and 40001 is an unused port on "right". See ``EXAMPLES'' for more.

VERSION

This document describes nat-traverse v0.5.

DESCRIPTION

nat-traverse establishes connections between nodes which are behind NAT gateways, i.e. hosts which do not have public IP addresses. Additionally, you can setup a small VPN by using pppd on top of nat-traverse (see ``EXAMPLES''). nat-traverse does not need an external server on the Internet, and it isn't necessary to reconfigure the involved NAT gateways, either. nat-traverse works out-of-the-box.

See ``TECHNIQUE'' for how this is achieved.

Limitation: nat-traverse does not work with gateways which change the port numbers. This is a fundamental problem of nat-traverse's design, as the changed port numbers are (in general) not predictable.

OPTIONS

"local_port:peer:remote_port" (required)
Sets the local port to use and the remote address to connect to.

Note that you have to give the

IP address or hostname of the NAT gateway of the host you want to connect to, as the target host doesn't have a public IP address.
"--cmd="pppd...""
Runs the specified command after establishing the connection.

The command will be run with its

STDIN and STDOUT bound to the socket, i.e. everything the command writes to STDOUT will be forwarded to the peer.

If no command is specified, nat-traverse will relay input from STDIN to the peer and vice versa, i.e. nat-traverse degrades to netcat.

"--window=
10"
Sets the number of initial garbage packets to send. The default, 10, should work with most firewalls.
"--timeout=
10"
Sets the maximum number of seconds to wait for an acknowledgement by the peer.
"--quit-after-connect"
Quits nat-traverse after the tunnel has been established successfully.

nat-traverse returns a non-0

statuscode to indicate that it wasn't able to establish the tunnel.

"--quit-after-connect" is useful if you want another program to use the tunnel. For example, you could configure OpenVPN to use the the same ports as nat-traverse --- thus OpenVPN would be able to cross NAT gateways.

"--version", "--help"

TECHNIQUE

nat-traverse establishes connections between hosts behind NAT gateways without need for reconfiguration of the involved NAT gateways.
1.
Firstly, nat-traverse on host "left" sends garbage UDP packets to the NAT gateway of "right". These packets are, of course, discarded by the firewall.
2.
Then "right"'s nat-traverse sends garbage UDP packets to the NAT gateway of "left". These packets are not discarded, as "left"'s NAT gateway thinks these packets are replies to the packets sent in step 1!
3.
"left"'s nat-traverse continues to send garbage packets to "right"'s NAT gateway. These packets are now not dropped either, as the NAT gateway thinks the packets are replies to the packets sent in step 2.
4.
Finally, both hosts send an acknowledgement packet to signal readiness. When these packets are received, the connection is established and nat-traverse can either relay STDIN/STDOUT to the socket or execute a program.

EXAMPLES

Setup of a small VPN with PPP

It's easy to setup a VPN (Virtual Private Network) by using the Point-to-Point Protocol Daemon, "pppd":

  [email protected] # nat-traverse \
      --cmd="pppd updetach noauth passive notty \
             ipparam vpn 10.0.0.1:10.0.0.2"
      40000:natgw-of-right:40001
  [email protected] # nat-traverse \
      --cmd="pppd nodetach notty noauth"
      40001:natgw-of-left:40000

"pppd" creates a new interface, typically "ppp0". Using this interface, you can ping 10.0.0.1 or 10.0.0.2. As you can see, "pppd" upgrades the data-only tunnel nat-traverse provides to a full IP tunnel. Thus you can establish reliable TCP connections over the tunnel, even though the tunnel uses UDP! Furthermore, you could even add IPv6 addresses to "ppp0" by running "ip -6 addr add..."!

Note though that although this VPN is arguably a private network, it is not secured in any way. You may want to use SSH to encrypt the connection.

Port Forwarding with netcat

You can use "netcat" to forward one of your local UDP or TCP ports to an arbitrary UDP or TCP port of the remote host, similar to "ssh -L" or "ssh -R":

  [email protected]  $ nat-traverse 10001:natgw-of-right:10002 \
        --cmd="nc -vl 20000"
  [email protected] $ nat-traverse 10002:natgw-of-left:10001 \
        --cmd="nc -v localhost 22"

As soon as the tunnel is established (using UDP ports 10001 and 10002), "left"'s TCP port 20000 is forwarded to "right"'s SSH Daemon (TCP port 22):

  [email protected] $ ssh -p 20000 [email protected]
  # Will connect to right's SSH daemon!

But do note that you lose the reliability of TCP in this example, as the actual data is transported via UDP; so this is only a toy example. If you want reliable streams, use PPP on top of nat-traverse, as described above.

Setup of a VPN with OpenVPN

You can use <OpenVPN> over nat-traverse if you want to have a secure VPN.

Using OpenVPN over nat-traverse requires only one change to OpenVPN's configuration file, presuming that you don't want to use OpenVPN's multi-client mode: You have to adjust the "code" and "lport" options accordingly, for example:

  # Options to add to left's and right's OpenVPN config:
  port  60001
  lport 60001
  # Command to execute on left resp. right:
  [email protected]  # until \
                 nat-traverse --quit-after-connect 60001:right:60001 \
               do \
                 sleep 5 \
               done; \
               openvpn [...]
  [email protected] # until \
                 nat-traverse --quit-after-connect 60001:left:60001 \
               do \
                 sleep 5 \
               done; \
               openvpn [...]<!--

The "until" loop ensures that OpenVPN will not be started before nat-traverse was able to establish the connection. Michael Kugele ("michael (at) kugele.net") also reported a way to still be able to use OpenVPN's multi-client mode with nat-traverse: As all instances of nat-traverse have to use unique ports (because a connection is identified by the source/destination port combination), you've to use redirection rules to redirect the ports used by nat-traverse to the port the OpenVPN daemon listens on:

  iptables -t nat -A PREROUTING -p udp \
    --dport $LPORT -j DNAT --to $HOST:$PORT
  iptables -t nat -A PREROUTING -p udp \
    --dport $PORT -j REDIRECT --to-port $LPORT

$LPORT specifies the source port nat-traverse uses on the server side, and "$HOST:$PORT" is the address of the OpenVPN server.)

LIMITATIONS

Only IPv4 is supported, nat-traverse won't work with IPv6 addresses. Drop me a note if you do need IPv6 support.

nat-traverse does not work with gateways which change the port numbers. This is a fundamental problem of nat-traverse's design, as the changed port numbers are (in general) not predictable.

AUTHOR

Copyright (C) 2005, 2012 Ingo Blechschmidt, <[email protected]>.

You may want to visit nat-traverse's Freecode project page, <http://freecode.com/projects/nat-traverse/>.

LICENSE

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.