connect.c is the simple relaying command to make network connection via SOCKS and https proxy. It is mainly intended to be used as proxy command of OpenSSH. You can make SSH session beyond the firewall with this command,
Features of connect.c are:
Download source code from:
http://www.taiyo.co.jp/~gotoh/ssh/connect.c
For windows user, pre-compiled binary is also available:
http://www.taiyo.co.jp/~gotoh/ssh/connect.exe (compiled with MSVC)
OpenSSH development team decides to stop supporting SOCKS and any other tunneling mechanism. It was aimed to separate complexity to support various mechanism of proxying from core code. And they recommends more flexible mechanism: ProxyCommand option instead.
Proxy command mechanism is delegation of network stream communication. If ProxyCommand options is specified, SSH invoke specified external command and talk with standard I/O of thid command. Invoked command undertakes network communication with relaying to/from standard input/output including iniitial communication or negotiation for proxying. Thus, ssh can split out proxying code into external command.
The connect.c program was made for this purpose.
Download source code from here.
If you are MS Windows user, you can get pre-compiled binary from
here.
In most environment, you can compile connect.c simply. On UNIX environment, you can use cc or gcc. On Windows environment, you can use Microsoft Visual C, Borland C or Cygwin gcc.
| Compiler | command line to compile |
|---|---|
| UNIX cc | cc connect.c -o connect |
| UNIX gcc | gcc connect.c -o connect |
| Solaris | gcc connect.c -o connect -lnsl -lsocket -lresolv |
| Microsoft Visual C/C++ | cl connect.c wsock32.lib advapi32.lib |
| Borland C | bcc32 connect.c wsock32.lib advapi32.lib |
| Cygwin gcc | gcc connect.c -o connect |
| Mac OS X | gcc connect.c -o connect -lresolv or gcc connect.c -o connect -DBIND_8_COMPAT=1 |
To install connect command, simply copy compiled binary to directory in your PATH (ex. /usr/local/bin). Like this:
$ cp connect /usr/local/bin
Modify your ~/.ssh/config file to use connect command as
proxy command. For the case of SOCKS server is running on
firewall host socks.local.net with port 1080, you can add
ProxyCommand option in ~/.ssh/config, like this:
Host remote.outside.net ProxyCommand connect -S socks.local.net %h %p
%h and %p will be replaced on invoking proxy command with
target hostname and port specified to SSH command.
If you hate writing many entries of remote hosts, following example may help you.
## Inside of the firewall, use connect command with direct connection. Host *.local.net ProxyCommand connect %h %p ## Outside of the firewall, use connect command with SOCKS conenction. Host * ProxyCommand connect -S socks.local.net %h %p
If you want to use http proxy, use -H option instead of -S option in examle above, like this:
## Inside of the firewall, direct Host *.local.net ProxyCommand connect %h %p ## Outside of the firewall, with HTTP proxy Host * ProxyCommand connect -H proxy.local.net:8080 %h %p
After editing your ~/.ssh/config file, you are ready to use ssh.
You can execute ssh without any special options as if remote host is
IP reachable host. Following is an example to execute hostname
command on host remote.outside.net.
$ ssh remote.outside.net hostname remote.outside.net $
If you have trouble, execute connect command from command line
with -d option to see what is happened. Some debug message may
appear and reports progress. This information may tell you what is
wrong. In this example, error has occurred on authentication stage of
SOCKS5 protocol.
$ connect -d -S socks.local.net unknown.remote.outside.net 110 DEBUG: relay_method = SOCKS (2) DEBUG: relay_host=socks.local.net DEBUG: relay_port=1080 DEBUG: relay_user=gotoh DEBUG: socks_version=5 DEBUG: socks_resolve=REMOTE (2) DEBUG: local_type=stdio DEBUG: dest_host=unknown.remote.outside.net DEBUG: dest_port=110 DEBUG: Program is $Revision: 1.35 $ DEBUG: connecting to xxx.xxx.xxx.xxx:1080 DEBUG: begin_socks_relay() DEBUG: atomic_out() [4 bytes] DEBUG: >>> 05 02 00 02 DEBUG: atomic_in() [2 bytes] DEBUG: <<< 05 02 DEBUG: auth method: USERPASS DEBUG: atomic_out() [some bytes] DEBUG: >>> xx xx xx xx ... DEBUG: atomic_in() [2 bytes] DEBUG: <<< 01 01 ERROR: Authentication faield. FATAL: failed to begin relaying via SOCKS.
Command line usage is here:
usage: connect [-dnhst45] [-R resolve] [-p local-port] [-w sec]
[-H [user@]proxy-server[:port]]
[-S [user@]socks-server[:port]]
[-T socks-server:[port]]
[-c telnet-proxy-command]
host port
host and port is target hostname and port-number to connect.
-H option specify hostname and port number of http proxy server to
relay. If port is omitted, 80 is used. You can specify this value by
environment variable HTTP_PROXY and give -h option to use it.
-S option specify hostname and port number of SOCKS server to
relay. Like -H option, port number can be omit and default is 1080.
You can also specify this value pair by environment variable
SOCKS5_SERVER and give -s option to use it.
-T option specify hostname and port number of telnet proxy to
relay. The port number can be omit and default is 23.
You can also specify this value pair by environment variable
TELNET_PROXY and give -t option to use it.
-4 and -5 is for specifying SOCKS protocol version. It is valid only using with -s or -S. Default is -5 (protocol version 5)
-R is for specifying method to resolve hostname. 3 keywords
(local, remote, both) or dot-notation IP address is
allowed. Keyword both means; "Try local first, then
remote". If dot-notation IP address is specified, use this host as
nameserver (UNIX only). Default is remote for SOCKS5 or local
for others. On SOCKS4 protocol, remote resolving method (remote
and both) use protocol version 4a.
The -p option specifys to wait a local TCP port and make relaying with it instead of standard input and output.
The -w option specifys timeout seconds on making connection with target host.
The -c option specifys request string against telnet proxy server. The special word '%h' and '%p' in this string are replaced as hostname and port number before sending. For telnet proxy by DeleGate, both "telnet %h %p" and "%h:%p" are acceptable. Default is "telnet %h %p".
The -a option specifiys user intended authentication methods
separated by comma. Currently userpass and none are
supported. Default is userpass. You can also specifying this
parameter by the environment variable SOCKS5_AUTH.
The -d option is used for debug. If you fail to connect, use this and check request to and response from server.
You can omit port argument when program name is special format containing port number itself. For example,
$ ln -s connect connect-25 $ ./connect-25 smtphost.outside.net 220 smtphost.outside.net ESMTP Sendmail QUIT 221 2.0.0 smtphost.remote.net closing connection $
This example means that the command name "connect-25" contains port number
25 so you can omit 2nd argument (and used if specified explicitly).
There are 5 environemnt variables to specify user name without command line option. This mechanism is usefull for the user who using another user name different from system account.
Following table describes how user name is determined. Left most number is order to check. If variable is not defined, check next variable, and so on.
| SOCKS v5 | SOCKS v4 | HTTP proxy | |
|---|---|---|---|
| 1 | SOCKS5_USER | SOCKS4_USER | HTTP_PROXY_USER |
| 2 | SOCKS_USER | ||
| 3 | CONNECT_USER | ||
| 4 | (query user name to system) | ||
There are 5 environemnt variables to specify password. If you use this feature, please note that it is not secure way.
Following table describes how password is determined. Left most number is order to check. If variable is not defined, check next variable, and so on. Finally ask to user interactively using external program or tty input.
| SOCKS v5 | HTTP proxy | |
|---|---|---|
| 1 | SOCKS5_PASSWD | HTTP_PROXY_PASSWORD |
| 2 | SOCKS_PASSWORD | |
| 3 | CONNECT_PASSWORD | |
| 4 | (ask to user interactively) | |
Only NO-AUTH and USER/PASSWORD authentications are supported. GSSAPI authentication (RFC 1961) and other draft authentications (CHAP, EAP, MAF, etc.) is not supported.
BASIC authentication is supported but DIGEST authentication is not.
There is no mechanism to switch proxy server regarding to PC environment. This limitation might be bad news for mobile user. Since I do not want to make this program complex, I do not want to support although this feature is already requested. Please advice me if there is good idea of detecting environment to swich and simple way to specify conditioned directive of servers.
One tricky workaround exists. It is replacing ~/.ssh/config file by script on ppp up/down.
There's another example of wrapper script (contributed by Darren Tucker). This script costs executing ifconfig and grep to detect current environment, but it works. (NOTE: you should modify addresses if you use it.)
#!/bin/sh ## ~/bin/myconnect --- Proxy server switching wrapper if ifconfig eth0 |grep "inet addr:192\.168\.1" >/dev/null; then opts="-S 192.168.1.1:1080" elif ifconfig eth0 |grep "inet addr:10\." >/dev/null; then opts="-H 10.1.1.1:80" else opts="-s" fi exec /usr/local/bin/connect $opts $@
At first, note that the telnet proxy support is an partial feature. In this implementation, connect single requestinting and proxy returns some success/error detective in talked back lines including greeting, prompt and connected messages.
The connect simply send request after connection to proxy is established before any response reading, then repeat reading response strings from proxy to decide remote connection request is succeeded or not by checking pre-defined phrase in each lines. There are pre-defined phrases which are good-phrase and bad-phrases. First good-phrase is checked and change state as relaying if exist. connect treat this line as final response from proxy before starting acutal communication with remote host. Or if good-phrase is not matched, bad-phrases will be checked. If one of bad-phrase matched, it cause connection error immediately.
The pre-defined phrases are currently fixed string so you cannot change without modifying and compiling. The good-phrase is: "connected to". The bad-phrases are: " failed", " refused", " rejected", " closed".
In usual, connect.c relays network connection to/from standard input/output. By specifying -p option, however, connect.c relays local network stream instead of standard input/output. With this option, connect command waits connection from other program, then start relaying between both network stream.
This feature may be useful for the program which is hard to SOCKSify.
connect.c ask you password when authentication is required. If
you are using on tty/pty terminal, connect can input from terminal
with prompt. But you can also use ssh-askpass program to input
password. If you are graphical environment like X Window or MS
Windows, and program does not have tty/pty, and environment variable
SSH_ASKPASS is specified, then connect.c invoke command
specified by environment variable SSH_ASKPASS to input password.
ssh-askpass program might be installed if you are using OpenSSH on
UNIX environment. On Windows environment, pre-compiled binary is
available from
here.
This feature is limited on window system environment.
And also useful on Emacs on MS Windows (NT Emacs or Meadow). It is hard to send passphrase to connect command (and also ssh) because external command is invoked on hidden terminal and do I/O with this terminal. Using ssh-askpass avoids this problem.
Although connect.c is made for OpenSSH, it is generic and independent from OpenSSH. So we can use this for other purpose. For example, you can use this command in Emacs to open network connection with remote host over the firewall via SOCKS or HTTP proxy without SOCKSifying Emacs itself.
There is sample code: http://www.taiyo.co.jp/~gotoh/lisp/relay.el
With this code, you can use relay-open-network-stream function
instead of open-network-stream to make network connection. See top
comments of source for more detail.
If you are SOCKS4 user on UNIX environment, you might want specify nameserver to resolve remote hostname. You can do it specifying -R option followed by IP address of resolver.
Conbination of ssh and connect command have more interesting usage. Following command makes indirect connection to host2:port from your current host via host1.
ssh host1 connect host2 port
This method is useful for the situations like:
For example, I want to use local NetNews service in my office from home. I cannot make NNTP session directly because NNTP host is barriered by firewall. Fortunately, I have ssh account on internal host and allowed using SOCKS5 on firewall from outside. So I use following command to connect to NNTP service.
$ ssh host1 connect news 119 200 news.my-office.com InterNetNews NNRP server INN 2.3.2 ready (posting ok). quit 205 . $
By combinating hopping connection and relay.el, I can read NetNews using Wanderlust on Emacs at home.
|
External (internet) | Internal (office)
|
+------+ +----------+ +-------+ +-----------+
| HOME | | firewall | | host1 | | NNTP host |
+------+ +----------+ +-------+ +-----------+
emacs <-------------- ssh ---------------> sshd <-- connect --> nntpd
<-- connect --> socksd <-- SOCKS -->
As an advanced example, you can use SSH hopping as fetchmail's plug-in program to access via secure tunnel. This method requires that connect program is insatalled on remote host. There's example of .fetchmailrc bellow. When fetchmail access to mail-server, you will login to remote host using SSH then execute connect program on remote host to relay conversation with pop server. Thus fetchmail can retrieve mails in secure.
poll mail-server protocol pop3 plugin "ssh %h connect localhost %p" username "username" password "password" </exmaple> * <a name="sec23" id="sec23"></a>Break The More Restricted Wall If firewall does not provide SOCKS nor HTTPS other than port 443, you cannot break the wall in usual way. But if you have you own host which is accessible from internet, you can make ssh connection to your own host by configuring sshd as waiting at port 443 instead of standard 22. By this, you can login to your own host via port 443. Once you have logged-in to extenal home machine, you can execute **connect** as second hop to make connection from your own host to final target host, like this: <example> $ cat ~/.ssh/config Host home ProxyCommand connect -H firewall:8080 %h 443 Host server ProxyCommand ssh home connect %h %p ... internal$ ssh home You are logged in to home! home# exit internal$ ssh server You are logged in to server! server# exit internal$
This way is similar to "Hopping connection via SSH" except configuring outer sshd as waiting at port 443 (https). This means that you have a capability to break the strongly restricted wall if you have own host out side of the wall.
|
Internal (office) | External (internet)
|
+--------+ +----------+ +------+ +--------+
| office | | firewall | | home | | server |
+--------+ +----------+ +------+ +--------+
<------------------ ssh --------------------->sshd:443
<-- connect --> http-proxy <-- https:443 --> any
connect <-- tcp --> port
NOTE: If you wanna use this, you should give up hosting https service at port 443 on you external host 'home'.
SOCKS version 4 is first popular implementation which is documented here. Since this protocol provide IP address based requesting, client program should resolve name of outer host by itself. Version 4a (documented here) is enhanced to allow request by hostname instead of IP address.
SOCKS version 5 is re-designed protocol stands on experience of version 4 and 4a. There is no compativility with previous versions. Instead, there's some improvement: IPv6 support, request by hostname, UDP proxying, etc.
Many http proxy servers implementation supports https CONNECT method
(SLL). You might add configuration to allow using https. For the
example of DeleGate (
DeleGate is a multi-purpose application level gateway, or a proxy
server) , you should add https to REMITTABLE parameter to
allow HTTP-Proxy like this:
delegated -Pxxxx ...... REMITTABLE='+,https' ...
For the case of Squid, you should allow target ports via https by ACL, and so on.