Oops, broke --noserver mode at some point.

Feature requested by Wayne Scott and Ed Maste.

README.md

@@ -0,0 +1,152 @@
+sshuttle: where transparent proxy meets VPN meets ssh
+=====================================================
+
+I just spent an afternoon working on a new kind of VPN.  You can get
+the first release, <a href="http://github.com/apenwarr/sshuttle">sshuttle
+0.10, on github</a>.
+
+As far as I know, sshuttle is the only program that solves the following
+common case:
+
+ - Your client machine (or router) is Linux.
+
+ - You have access to a remote network via ssh.
+
+ - You don't necessarily have admin access on the remote network.
+
+ - The remote network has no VPN, or only stupid/complex VPN
+    protocols (IPsec, PPTP, etc). Or maybe you <i>are</i> the
+    admin and you just got frustrated with the awful state of
+    VPN tools.
+
+ - You don't want to create an ssh port forward for every
+    single host/port on the remote network.
+
+ - You hate openssh's port forwarding because it's randomly
+    slow and/or stupid.
+ 
+ - You can't use openssh's PermitTunnel feature because
+    it's disabled by default on openssh servers; plus it does
+    TCP-over-TCP, which has terrible performance (see below).
+    
+
+Prerequisites
+-------------
+
+ - sudo, su, or logged in as root on your client machine.
+   (The server doesn't need admin access.)
+   
+ - Linux+iptables on your client machine, including at
+   least the iptables DNAT, REDIRECT, and ttl modules. 
+   This is available by default on most Linux distributions. 
+   (The server doesn't need iptables and doesn't need to be
+   Linux.)
+
+
+This is how you use it:
+-----------------------
+
+ - <tt>git clone git://github.com/apenwarr/sshuttle</tt>
+    on your client and server machines. The server can be
+    any ssh server with python available; the client must
+    be Linux with iptables, and you'll need root or sudo
+    access.
+
+ - <tt>./sshuttle -r username@sshserver 0.0.0.0/0 -vv</tt>
+
+That's it!  Now your local machine can access the remote network as if you
+were right there!  And if your "client" machine is a router, everyone on
+your local network can make connections to your remote network.
+
+This creates a transparent proxy server on your local machine for all IP
+addresses that match 0.0.0.0/0.  (You can use more specific IP addresses if
+you want; use any number of IP addresses or subnets to change which
+addresses get proxied.  Using 0.0.0.0/0 proxies <i>everything</i>, which is
+interesting if you don't trust the people on your local network.)
+
+Any TCP session you initiate to one of the proxied IP addresses will be
+captured by sshuttle and sent over an ssh session to the remote copy of
+sshuttle, which will then regenerate the connection on that end, and funnel
+the data back and forth through ssh.
+
+Fun, right?  A poor man's instant VPN, and you don't even have to have
+admin access on the server.
+
+
+Theory of Operation
+-------------------
+
+sshuttle is not exactly a VPN, and not exactly port forwarding.  It's kind
+of both, and kind of neither.
+
+It's like a VPN, since it can forward every port on an entire network, not
+just ports you specify.  Conveniently, it lets you use the "real" IP
+addresses of each host rather than faking port numbers on localhost.
+
+On the other hand, the way it *works* is more like ssh port forwarding than
+a VPN.  Normally, a VPN forwards your data one packet at a time, and
+doesn't care about individual connections; ie. it's "stateless" with respect
+to the traffic.  sshuttle is the opposite of stateless; it tracks every
+single connection.
+
+You could compare sshuttle to something like the old <a
+href="http://en.wikipedia.org/wiki/Slirp">Slirp</a> program, which was a
+userspace TCP/IP implementation that did something similar.  But it
+operated on a packet-by-packet basis on the client side, reassembling the
+packets on the server side.  That worked okay back in the "real live serial
+port" days, because serial ports had predictable latency and buffering.
+
+But you can't safely just forward TCP packets over a TCP session (like ssh),
+because TCP's performance depends fundamentally on packet loss; it
+<i>must</i> experience packet loss in order to know when to slow down!  At
+the same time, the outer TCP session (ssh, in this case) is a reliable
+transport, which means that what you forward through the tunnel <i>never</i>
+experiences packet loss.  The ssh session itself experiences packet loss, of
+course, but TCP fixes it up and ssh (and thus you) never know the
+difference.  But neither does your inner TCP session, and extremely screwy
+performance ensues.
+
+sshuttle assembles the TCP stream locally, multiplexes it statefully over
+an ssh session, and disassembles it back into packets at the other end.  So
+it never ends up doing TCP-over-TCP.  It's just data-over-TCP, which is
+safe.
+
+
+Useless Trivia
+--------------
+
+Back in 1998 (12 years ago!  Yikes!), I released the first version of <a
+href="http://alumnit.ca/wiki/?TunnelVisionReadMe">Tunnel Vision</a>, a
+semi-intelligent VPN client for Linux.  Unfortunately, I made two big mistakes: 
+I implemented the key exchange myself (oops), and I ended up doing
+TCP-over-TCP (double oops).  The resulting program worked okay - and people
+used it for years - but the performance was always a bit funny.  And nobody
+ever found any security flaws in my key exchange, either, but that doesn't
+mean anything. :)
+
+The same year, dcoombs and I also released Fast Forward, a proxy server
+supporting transparent proxying.  Among other things, we used it for
+automatically splitting traffic across more than one Internet connection (a
+tool we called "Double Vision").
+
+I was still in university at the time.  A couple years after that, one of my
+professors was working with some graduate students on the technology that
+would eventually become <a href="http://www.slipstream.com/">Slipstream
+Internet Acceleration</a>.  He asked me to do a contract for him to build an
+initial prototype of a transparent proxy server for mobile networks.  The
+idea was similar to sshuttle: if you reassemble and then disassemble the TCP
+packets, you can reduce latency and improve performance vs.  just forwarding
+the packets over a plain VPN or mobile network.  (It's unlikely that any of
+my code has persisted in the Slipstream product today, but the concept is
+still pretty cool.  I'm still horrified that people use plain TCP on
+complex mobile networks with crazily variable latency, for which it was
+never really intended.)
+
+That project I did for Slipstream was what first gave me the idea to merge
+the concepts of Fast Forward, Double Vision, and Tunnel Vision into a single
+program that was the best of all worlds.  And here we are, at last, 10 years
+later.  You're welcome.
+
+--
+Avery Pennarun <apenwarr@gmail.com>
+

client.py

@@ -13,17 +13,68 @@ def original_dst(sock):
     return (ip,port)
 
 
-def iptables_setup(port, subnets):
+class IPTables:
+    def __init__(self, port, subnets):
+        self.port = port
+        self.subnets = subnets
         subnets_str = ['%s/%d' % (ip,width) for ip,width in subnets]
-    argv = (['sudo', sys.argv[0]] +
+        argvbase = ([sys.argv[0]] +
                     ['-v'] * (helpers.verbose or 0) +
                     ['--iptables', str(port)] + subnets_str)
-    rv = subprocess.call(argv)
-    if rv != 0:
-        raise Fatal('%r returned %d' % (argv, rv))
+        argv_tries = [
+            ['sudo'] + argvbase,
+            ['su', '-c', ' '.join(argvbase)],
+            argvbase
+        ]
+
+        # we can't use stdin/stdout=subprocess.PIPE here, as we normally would,
+        # because stupid Linux 'su' requires that stdin be attached to a tty.
+        # Instead, attach a *bidirectional* socket to its stdout, and use
+        # that for talking in both directions.
+        (s1,s2) = socket.socketpair()
+        def setup():
+            # run in the child process
+            s2.close()
+        e = None
+        for argv in argv_tries:
+            try:
+                self.p = subprocess.Popen(argv, stdout=s1, preexec_fn=setup)
+                e = None
+                break
+            except OSError, e:
+                pass
+        self.argv = argv
+        s1.close()
+        self.pfile = s2.makefile('wb+')
+        if e:
+            log('Spawning iptables: %r\n' % self.argv)
+            raise Fatal(e)
+        line = self.pfile.readline()
+        self.check()
+        if line != 'READY\n':
+            raise Fatal('%r expected READY, got %r' % (self.argv, line))
+
+    def check(self):
+        rv = self.p.poll()
+        if rv:
+            raise Fatal('%r returned %d' % (self.argv, rv))
+
+    def start(self):
+        self.pfile.write('GO\n')
+        self.pfile.flush()
+        line = self.pfile.readline()
+        self.check()
+        if line != 'STARTED\n':
+            raise Fatal('%r expected STARTED, got %r' % (self.argv, line))
+
+    def done(self):
+        self.pfile.close()
+        rv = self.p.wait()
+        if rv:
+            raise Fatal('cleanup: %r returned %d' % (self.argv, rv))
 
 
-def _main(listener, listenport, use_server, remotename, subnets):
+def _main(listener, ipt, use_server, remotename):
     handlers = []
     if use_server:
         if helpers.verbose >= 1:
@@ -47,7 +98,7 @@ def _main(listener, listenport, use_server, remotename, subnets):
 
     # we definitely want to do this *after* starting ssh, or we might end
     # up intercepting the ssh connection!
-    iptables_setup(listenport, subnets)
+    ipt.start()
 
     def onaccept():
         sock,srcip = listener.accept()
@@ -87,6 +138,9 @@ def _main(listener, listenport, use_server, remotename, subnets):
         for s in handlers:
             if s.socks & ready:
                 s.callback()
+        if use_server:
+            mux.callback()
+            mux.check_fullness()
 
 
 def main(listenip, use_server, remotename, subnets):
@@ -116,7 +170,9 @@ def main(listenip, use_server, remotename, subnets):
     listenip = listener.getsockname()
     debug1('Listening on %r.\n' % (listenip,))
 
+    ipt = IPTables(listenip[1], subnets)
+    
     try:
-        return _main(listener, listenip[1], use_server, remotename, subnets)
+        return _main(listener, ipt, use_server, remotename)
     finally:
-        iptables_setup(listenip[1], [])
+        ipt.done()

iptables.py

@@ -11,7 +11,7 @@ def chain_exists(name):
             return True
     rv = p.wait()
     if rv:
-        raise Exception('%r returned %d' % (argv, rv))
+        raise Fatal('%r returned %d' % (argv, rv))
 
 
 def ipt(*args):
@@ -19,27 +19,10 @@ def ipt(*args):
     debug1('>> %s\n' % ' '.join(argv))
     rv = subprocess.call(argv)
     if rv:
-        raise Exception('%r returned %d' % (argv, rv))
+        raise Fatal('%r returned %d' % (argv, rv))
 
 
-# This is some iptables voodoo for setting up the Linux kernel's transparent
-# proxying stuff.  If subnets is empty, we just delete our sshuttle chain;
-# otherwise we delete it, then make it from scratch.
-#
-# We name the chain based on the transproxy port number so that it's possible
-# to run multiple copies of sshuttle at the same time.  Of course, the
-# multiple copies shouldn't have overlapping subnets, or only the most-
-# recently-started one will win (because we use "-I OUTPUT 1" instead of
-# "-A OUTPUT").
-#
-# sshuttle is supposed to clean up after itself by deleting extra chains on
-# exit.  In case that fails, it's not the end of the world; future runs will
-# supercede it in the transproxy list, at least, so the leftover iptables
-# chains are mostly harmless.
-def main(port, subnets):
-    assert(port > 0)
-    assert(port <= 65535)
-
+def do_it(port, subnets):
     chain = 'sshuttle-%s' % port
 
     # basic cleanup/setup of chains
@@ -63,4 +46,72 @@ def main(port, subnets):
                 '--to-ports', str(port),
                 '-m', 'ttl', '!', '--ttl', '42'  # to prevent infinite loops
                 )
-    subnets_str = ['%s/%d' % (ip,width) for ip,width in subnets]
+
+
+# This is some iptables voodoo for setting up the Linux kernel's transparent
+# proxying stuff.  If subnets is empty, we just delete our sshuttle chain;
+# otherwise we delete it, then make it from scratch.
+#
+# We name the chain based on the transproxy port number so that it's possible
+# to run multiple copies of sshuttle at the same time.  Of course, the
+# multiple copies shouldn't have overlapping subnets, or only the most-
+# recently-started one will win (because we use "-I OUTPUT 1" instead of
+# "-A OUTPUT").
+#
+# This code is supposed to clean up after itself by deleting extra chains on
+# exit.  In case that fails, it's not the end of the world; future runs will
+# supercede it in the transproxy list, at least, so the leftover iptables
+# chains are mostly harmless.
+def main(port, subnets):
+    assert(port > 0)
+    assert(port <= 65535)
+
+    if os.getuid() != 0:
+        raise Fatal('you must be root (or enable su/sudo) to set up iptables')
+
+    # because of limitations of the 'su' command, the *real* stdin/stdout
+    # are both attached to stdout initially.  Clone stdout into stdin so we
+    # can read from it.
+    os.dup2(1, 0)
+
+    debug1('iptables manager ready.\n')
+    sys.stdout.write('READY\n')
+    sys.stdout.flush()
+
+    # ctrl-c shouldn't be passed along to me.  When the main sshuttle dies,
+    # I'll die automatically.
+    os.setsid()
+
+    # we wait until we get some input before creating the rules.  That way,
+    # sshuttle can launch us as early as possible (and get sudo password
+    # authentication as early in the startup process as possible).
+    line = sys.stdin.readline(128)
+    if not line:
+        return  # parent died; nothing to do
+    if line != 'GO\n':
+        raise Fatal('iptables: expected GO but got %r' % line)
+    try:
+        if line:
+            debug1('iptables manager: starting transproxy.\n')
+            do_it(port, subnets)
+            sys.stdout.write('STARTED\n')
+        
+        try:
+            sys.stdout.flush()
+            
+            # Now we wait until EOF or any other kind of exception.  We need
+            # to stay running so that we don't need a *second* password
+            # authentication at shutdown time - that cleanup is important!
+            while sys.stdin.readline(128):
+                pass
+        except IOError:
+            # the parent process died for some reason; he's surely been loud
+            # enough, so no reason to report another error
+            return
+
+    finally:
+        try:
+            debug1('iptables manager: undoing changes.\n')
+        except:
+            pass
+        do_it(port, [])

main.py

@@ -45,7 +45,7 @@ def parse_ipport(s):
 
 
 optspec = """
-sshuttle [-l [ip:]port] [-r [username@]sshserver] <subnets...>
+sshuttle [-l [ip:]port] [-r [username@]sshserver[:port]] <subnets...>
 sshuttle --iptables <port> <subnets...>
 sshuttle --server
 --
@@ -83,5 +83,6 @@ except Fatal, e:
     log('fatal: %s\n' % e)
     sys.exit(99)
 except KeyboardInterrupt:
-    log('\nKeyboard interrupt: exiting.\n')
+    log('\n')
+    log('Keyboard interrupt: exiting.\n')
     sys.exit(1)

server.py

@@ -35,11 +35,15 @@ def main():
         handlers = filter(lambda s: s.ok, handlers)
         for s in handlers:
             s.pre_select(r,w,x)
-        debug2('Waiting: %d[%d,%d,%d]...\n' 
-               % (len(handlers), len(r), len(w), len(x)))
+        debug2('Waiting: %d[%d,%d,%d] (fullness=%d/%d)...\n' 
+               % (len(handlers), len(r), len(w), len(x),
+                  mux.fullness, mux.too_full))
         (r,w,x) = select.select(r,w,x)
         #log('r=%r w=%r x=%r\n' % (r,w,x))
         ready = set(r) | set(w) | set(x)
         for s in handlers:
+            #debug2('check: %r: %r\n' % (s, s.socks & ready))
             if s.socks & ready:
                 s.callback()
+        mux.check_fullness()
+        mux.callback()

ssh.py

@@ -1,10 +1,21 @@
 import sys, os, re, subprocess, socket
 import helpers
+from helpers import *
 
-def connect(rhost):
+def connect(rhostport):
     main_exe = sys.argv[0]
+    l = (rhostport or '').split(':', 1)
+    rhost = l[0]
+    portl = []
+    if len(l) > 1:
+        portl = ['-p', str(int(l[1]))]
     nicedir = os.path.split(os.path.abspath(main_exe))[0]
     nicedir = re.sub(r':', "_", nicedir)
+    myhome = os.path.expanduser('~') + '/'
+    if nicedir.startswith(myhome):
+        nicedir2 = nicedir[len(myhome):]
+    else:
+        nicedir2 = nicedir
     if rhost == '-':
         rhost = None
     if not rhost:
@@ -18,10 +29,13 @@ def connect(rhost):
         # allow shellquoting.  So we end up having to double-shellquote
         # stuff here.
         escapedir  = re.sub(r'([^\w/])', r'\\\\\\\1', nicedir)
+        escapedir2 = re.sub(r'([^\w/])', r'\\\\\\\1', nicedir2)
         cmd = r"""
-                   sh -c PATH=%s:'$PATH exec sshuttle --server%s'
-               """ % (escapedir, ' -v' * (helpers.verbose or 0))
-        argv = ['ssh', rhost, '--', cmd.strip()]
+                   sh -c PATH=%s:'$HOME'/%s:'$PATH exec sshuttle --server%s'
+               """ % (escapedir, escapedir2,
+                      ' -v' * (helpers.verbose or 0))
+        argv = ['ssh'] + portl + [rhost, '--', cmd.strip()]
+        debug2('executing: %r\n' % argv)
     (s1,s2) = socket.socketpair()
     def setup():
         # runs in the child process
@@ -32,7 +46,8 @@ def connect(rhost):
         os.setsid()
     s1a,s1b = os.dup(s1.fileno()), os.dup(s1.fileno())
     s1.close()
-    p = subprocess.Popen(argv, stdin=s1a, stdout=s1b, preexec_fn=setup)
+    p = subprocess.Popen(argv, stdin=s1a, stdout=s1b, preexec_fn=setup,
+                         close_fds=True)
     os.close(s1a)
     os.close(s1b)
     return p, s2

ssnet.py

@@ -81,6 +81,7 @@ class SockWrapper:
                 pass  # not connected yet
             elif e.args[0] in [errno.ECONNREFUSED, errno.ETIMEDOUT]:
                 # a "normal" kind of error
+                self.connect_to = None
                 self.seterr(e)
             else:
                 raise  # error we've never heard of?!  barf completely.
@@ -100,6 +101,9 @@ class SockWrapper:
             except socket.error, e:
                 self.seterr(e)
 
+    def too_full(self):
+        return False  # fullness is determined by the socket's select() state
+
     def uwrite(self, buf):
         if self.connect_to:
             return 0  # still connecting
@@ -180,6 +184,7 @@ class Proxy(Handler):
         if self.wrap1.connect_to:
             w.add(self.wrap1.rsock)
         elif self.wrap1.buf:
+            if not self.wrap2.too_full():
                 w.add(self.wrap2.wsock)
         elif not self.wrap1.shut_read:
             r.add(self.wrap1.rsock)
@@ -187,6 +192,7 @@ class Proxy(Handler):
         if self.wrap2.connect_to:
             w.add(self.wrap2.rsock)
         elif self.wrap2.buf:
+            if not self.wrap1.too_full():
                 w.add(self.wrap1.wsock)
         elif not self.wrap2.shut_read:
             r.add(self.wrap2.rsock)
@@ -214,6 +220,8 @@ class Mux(Handler):
         self.want = 0
         self.inbuf = ''
         self.outbuf = []
+        self.fullness = 0
+        self.too_full = False
         self.send(0, CMD_PING, 'chicken')
 
     def next_channel(self):
@@ -225,15 +233,28 @@ class Mux(Handler):
             if not self.channels.get(self.chani):
                 return self.chani
 
+    def amount_queued(self):
+        return sum(len(b) for b in self.outbuf)
+            
+    def check_fullness(self):
+        if self.fullness > 32768:
+            if not self.too_full:
+                self.send(0, CMD_PING, 'rttest')
+            self.too_full = True
+        #ob = []
+        #for b in self.outbuf:
+        #    (s1,s2,c) = struct.unpack('!ccH', b[:4])
+        #    ob.append(c)
+        #log('outbuf: %d %r\n' % (self.amount_queued(), ob))
+        
     def send(self, channel, cmd, data):
         data = str(data)
         assert(len(data) <= 65535)
         p = struct.pack('!ccHHH', 'S', 'S', channel, cmd, len(data)) + data
         self.outbuf.append(p)
-        debug2(' > channel=%d cmd=%s len=%d\n' 
-               % (channel, cmd_to_name[cmd], len(data)))
-        #log('Mux: send queue is %d/%d\n' 
-        #    % (len(self.outbuf), sum(len(b) for b in self.outbuf)))
+        debug2(' > channel=%d cmd=%s len=%d (fullness=%d)\n'
+               % (channel, cmd_to_name[cmd], len(data), self.fullness))
+        self.fullness += len(data)
 
     def got_packet(self, channel, cmd, data):
         debug2('<  channel=%d cmd=%s len=%d\n' 
@@ -242,6 +263,8 @@ class Mux(Handler):
             self.send(0, CMD_PONG, data)
         elif cmd == CMD_PONG:
             debug2('received PING response\n')
+            self.too_full = False
+            self.fullness = 0
         elif cmd == CMD_EXIT:
             self.ok = False
         elif cmd == CMD_CONNECT:
@@ -256,6 +279,7 @@ class Mux(Handler):
         self.wsock.setblocking(False)
         if self.outbuf and self.outbuf[0]:
             wrote = _nb_clean(os.write, self.wsock.fileno(), self.outbuf[0])
+            debug2('mux wrote: %d/%d\n' % (wrote, len(self.outbuf[0])))
             if wrote:
                 self.outbuf[0] = self.outbuf[0][wrote:]
         while self.outbuf and not self.outbuf[0]:
@@ -263,7 +287,10 @@ class Mux(Handler):
 
     def fill(self):
         self.rsock.setblocking(False)
+        try:
             b = _nb_clean(os.read, self.rsock.fileno(), 32768)
+        except OSError, e:
+            raise Fatal('other end: %r' % e)
         #log('<<< %r\n' % b)
         if b == '': # EOF
             self.ok = False
@@ -308,6 +335,7 @@ class MuxWrapper(SockWrapper):
         self.mux = mux
         self.channel = channel
         self.mux.channels[channel] = self.got_packet
+        self.socks = []
         debug2('new channel: %d\n' % channel)
 
     def __del__(self):
@@ -326,9 +354,14 @@ class MuxWrapper(SockWrapper):
             self.shut_write = True
             self.mux.send(self.channel, CMD_EOF, '')
 
+    def too_full(self):
+        return self.mux.too_full
+
     def uwrite(self, buf):
-        if len(buf) > 65535:
-            buf = buf[:32768]
+        if self.mux.too_full:
+            return 0  # too much already enqueued
+        if len(buf) > 2048:
+            buf = buf[:2048]
         self.mux.send(self.channel, CMD_DATA, buf)
         return len(buf)