Fix #494 sshuttle caught in infinite select() loop.

Improve detection of when the ssh process exits in both daemon and
foreground modes. Previously, sshuttle could infinite loop with 100%
cpu usage if the ssh process died. On machines that use suspend, the
ssh connection might not resume after wakeup. Now, this situation is
detected and sshuttle exits. The fix involves changing the return
value we check for when we call poll() and using a psutil function to
detect when the process exits if we are running sshuttle as a daemon.

requirements.txt

@@ -1 +1,2 @@
 setuptools-scm==4.1.2
+psutil

sshuttle/client.py

@@ -6,6 +6,7 @@ import subprocess as ssubprocess
 import os
 import sys
 import platform
+import psutil
 
 import sshuttle.helpers as helpers
 import sshuttle.ssnet as ssnet
@@ -541,11 +542,22 @@ def _main(tcp_listener, udp_listener, fw, ssh_cmd, remotename,
         debug1('seed_hosts: %r\n' % seed_hosts)
         mux.send(0, ssnet.CMD_HOST_REQ, str.encode('\n'.join(seed_hosts)))
 
-    while 1:
-        rv = serverproc.poll()
-        if rv:
-            raise Fatal('server died with error code %d' % rv)
 
+    def check_ssh_alive():
+        if daemon:
+            # poll() won't tell us when process exited since the
+            # process is no longer our child (it returns 0 all the
+            # time).
+            if not psutil.pid_exists(serverproc.pid):
+                raise Fatal('ssh connection to server (pid %d) exited.' % serverproc.pid)
+        else:
+            rv = serverproc.poll()
+            # poll returns None if process hasn't exited.
+            if rv is not None:
+                raise Fatal('ssh connection to server (pid %d) exited with returncode %d' % (serverproc.pid, rv))
+        
+    while 1:
+        check_ssh_alive()
         ssnet.runonce(handlers, mux)
         if latency_control:
             mux.check_fullness()