This has to rank as the most confusing topic in using the Linux command line. It is even confusing using Fugu because there are three user, host and ports that can be used leading to a lot of possibilities to get it wrong. To the best of my understanding, SSH tunnelling is TCP forwarding - at least as far as the SSH man page goes.
The canonical problem that one would want to solve using tunnelling is to burrow through an intermediate public-facing machine to get to a private machine that shares an network with the public-facing machine. I'll label the three machines
TERMINAL. The goal is to get to
The syntax is as follows:
[email protected]:~$ ssh -f -L [CLIENT_NAME:]NEW_CLIENT_PORT:TERMINAL_NAME:TERMINAL_PORT \ [email protected]_NAME [-p TUNNEL_PORT] sleep 10
NEW_CLIENT_PORT is a local port with value over 1023, say 2000.
) indicate optional parameters:
CLIENT_NAMEif it is 'localhost'
-p TUNNEL_PORTif you use SSH (
-p 22) unless it is on a non-standard port
CLIENT_NAMEcould refer to the hostname or the IP address.
-fargument indicates that this command should be run in the background. You want SSH to return the console so that you can connect to
sleep 10command will be run in
TUNNELupon successful connection to
TUNNEL. In this case, you have 10 seconds to try to connect to
TERMINAL. If 10 seconds elapses before a successful connection then the tunnel will be closed.
Once you are connected to
TUNNEL and get back the command prompt type the following:
[email protected]:~$ ssh [email protected]_NAME -p NEW_CLIENT_PORT
CLIENT_NAME = localhost NEW_CLIENT_PORT = 2000 TERMINAL_NAME = terminal.server.ten TERMINAL_PORT = 22 TUNNEL_USER = tunneler TUNNEL_NAME = public.server.ten TUNNEL_PORT = 22 TERMINAL_USER = terminator [email protected]:~$ ssh -f -L 2000:terminal.server.ten:22 [email protected] sleep 10 [email protected]:~$ ssh [email protected] -p 2000
SSH man page