2 # NEPI, a framework to manage network experiments
3 # Copyright (C) 2013 INRIA
5 # This program is free software: you can redistribute it and/or modify
6 # it under the terms of the GNU General Public License as published by
7 # the Free Software Foundation, either version 3 of the License, or
8 # (at your option) any later version.
10 # This program is distributed in the hope that it will be useful,
11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 # GNU General Public License for more details.
15 # You should have received a copy of the GNU General Public License
16 # along with this program. If not, see <http://www.gnu.org/licenses/>.
18 # Author: Alina Quereilhac <alina.quereilhac@inria.fr>
19 # Claudio Freire <claudio-daniel.freire@inria.fr>
21 ## TODO: This code needs reviewing !!!
38 logger = logging.getLogger("sshfuncs")
40 def log(msg, level, out = None, err = None):
42 msg += " - OUT: %s " % out
45 msg += " - ERROR: %s " % err
47 logger.log(level, msg)
50 if hasattr(os, "devnull"):
53 DEV_NULL = "/dev/null"
55 SHELL_SAFE = re.compile('^[-a-zA-Z0-9_=+:.,/]*$')
59 Special value that when given to rspawn in stderr causes stderr to
60 redirect to whatever stdout was redirected to.
65 Codes for status of remote spawned process
67 # Process is still running
73 # Process hasn't started running yet (this should be very rare)
76 hostbyname_cache = dict()
77 hostbyname_cache_lock = threading.Lock()
79 def gethostbyname(host):
80 global hostbyname_cache
81 global hostbyname_cache_lock
83 hostbyname = hostbyname_cache.get(host)
85 with hostbyname_cache_lock:
86 hostbyname = socket.gethostbyname(host)
87 hostbyname_cache[host] = hostbyname
89 msg = " Added hostbyname %s - %s " % (host, hostbyname)
90 log(msg, logging.DEBUG)
94 OPENSSH_HAS_PERSIST = None
96 def openssh_has_persist():
97 """ The ssh_config options ControlMaster and ControlPersist allow to
98 reuse a same network connection for multiple ssh sessions. In this
99 way limitations on number of open ssh connections can be bypassed.
100 However, older versions of openSSH do not support this feature.
101 This function is used to determine if ssh connection persist features
104 global OPENSSH_HAS_PERSIST
105 if OPENSSH_HAS_PERSIST is None:
106 proc = subprocess.Popen(["ssh","-v"],
107 stdout = subprocess.PIPE,
108 stderr = subprocess.STDOUT,
109 stdin = open("/dev/null","r") )
110 out,err = proc.communicate()
113 vre = re.compile(r'OpenSSH_(?:[6-9]|5[.][8-9]|5[.][1-9][0-9]|[1-9][0-9]).*', re.I)
114 OPENSSH_HAS_PERSIST = bool(vre.match(out))
115 return OPENSSH_HAS_PERSIST
117 def make_server_key_args(server_key, host, port):
118 """ Returns a reference to a temporary known_hosts file, to which
119 the server key has been added.
121 Make sure to hold onto the temp file reference until the process is
124 :param server_key: the server public key
125 :type server_key: str
127 :param host: the hostname
130 :param port: the ssh port
135 host = '%s:%s' % (host, str(port))
137 # Create a temporary server key file
138 tmp_known_hosts = tempfile.NamedTemporaryFile()
140 hostbyname = gethostbyname(host)
142 # Add the intended host key
143 tmp_known_hosts.write('%s,%s %s\n' % (host, hostbyname, server_key))
145 # If we're not in strict mode, add user-configured keys
146 if os.environ.get('NEPI_STRICT_AUTH_MODE',"").lower() not in ('1','true','on'):
147 user_hosts_path = '%s/.ssh/known_hosts' % (os.environ.get('HOME',""),)
148 if os.access(user_hosts_path, os.R_OK):
149 f = open(user_hosts_path, "r")
150 tmp_known_hosts.write(f.read())
153 tmp_known_hosts.flush()
155 return tmp_known_hosts
157 def make_control_path(agent, forward_x11):
158 ctrl_path = "/tmp/nepi_ssh"
166 ctrl_path += "-%r@%h:%p"
171 """ Escapes strings so that they are safe to use as command-line
173 if SHELL_SAFE.match(s):
174 # safe string - no escaping needed
177 # unsafe string - escape
179 if (32 <= ord(c) < 127 or c in ('\r','\n','\t')) and c not in ("'",'"'):
182 return "'$'\\x%02x''" % (ord(c),)
183 s = ''.join(map(escp,s))
186 def eintr_retry(func):
187 """Retries a function invocation when a EINTR occurs"""
189 @functools.wraps(func)
191 retry = kw.pop("_retry", False)
192 for i in xrange(0 if retry else 4):
194 return func(*p, **kw)
195 except (select.error, socket.error), args:
196 if args[0] == errno.EINTR:
201 if e.errno == errno.EINTR:
206 return func(*p, **kw)
209 def rexec(command, host, user,
219 connect_timeout = 30,
224 strict_host_checking = True):
226 Executes a remote command, returns ((stdout,stderr),process)
229 tmp_known_hosts = None
231 hostip = gethostbyname(host)
235 # Don't bother with localhost. Makes test easier
236 '-o', 'NoHostAuthenticationForLocalhost=yes',
237 '-o', 'ConnectTimeout=%d' % (int(connect_timeout),),
238 '-o', 'ConnectionAttempts=3',
239 '-o', 'ServerAliveInterval=30',
240 '-o', 'TCPKeepAlive=yes',
241 '-o', 'Batchmode=yes',
242 '-l', user, hostip or host]
244 if persistent and openssh_has_persist():
246 '-o', 'ControlMaster=auto',
247 '-o', 'ControlPath=%s' % (make_control_path(agent, forward_x11),),
248 '-o', 'ControlPersist=60' ])
250 if not strict_host_checking:
251 # Do not check for Host key. Unsafe.
252 args.extend(['-o', 'StrictHostKeyChecking=no'])
256 proxycommand = 'ProxyCommand=ssh %s@%s -W %%h:%%p' % (gwuser, gw)
258 proxycommand = 'ProxyCommand=ssh %%r@%s -W %%h:%%p' % gw
259 args.extend(['-o', proxycommand])
265 args.append('-p%d' % port)
268 identity = os.path.expanduser(identity)
269 args.extend(('-i', identity))
279 # Create a temporary server key file
280 tmp_known_hosts = make_server_key_args(server_key, host, port)
281 args.extend(['-o', 'UserKnownHostsFile=%s' % (tmp_known_hosts.name,)])
284 command = "sudo " + command
288 log_msg = " rexec - host %s - command %s " % (host, " ".join(args))
290 stdout = stderr = stdin = subprocess.PIPE
292 stdout = stderr = stdin = None
294 return _retry_rexec(args, log_msg,
300 tmp_known_hosts = tmp_known_hosts,
303 def rcopy(source, dest,
312 strict_host_checking = True):
314 Copies from/to remote sites.
316 Source and destination should have the user and host encoded
319 Source can be a list of files to copy to a single destination,
320 (in which case it is advised that the destination be a folder),
321 a single file in a string or a semi-colon separated list of files
325 # Parse destination as <user>@<server>:<path>
326 if isinstance(dest, str) and ':' in dest:
327 remspec, path = dest.split(':',1)
328 elif isinstance(source, str) and ':' in source:
329 remspec, path = source.split(':',1)
331 raise ValueError, "Both endpoints cannot be local"
332 user,host = remspec.rsplit('@',1)
335 tmp_known_hosts = None
337 args = ['scp', '-q', '-p', '-C',
338 # Speed up transfer using blowfish cypher specification which is
339 # faster than the default one (3des)
341 # Don't bother with localhost. Makes test easier
342 '-o', 'NoHostAuthenticationForLocalhost=yes',
343 '-o', 'ConnectTimeout=60',
344 '-o', 'ConnectionAttempts=3',
345 '-o', 'ServerAliveInterval=30',
346 '-o', 'TCPKeepAlive=yes' ]
349 args.append('-P%d' % port)
353 proxycommand = 'ProxyCommand=ssh %s@%s -W %%h:%%p' % (gwuser, gw)
355 proxycommand = 'ProxyCommand=ssh %%r@%s -W %%h:%%p' % gw
356 args.extend(['-o', proxycommand])
362 identity = os.path.expanduser(identity)
363 args.extend(('-i', identity))
366 # Create a temporary server key file
367 tmp_known_hosts = make_server_key_args(server_key, host, port)
368 args.extend(['-o', 'UserKnownHostsFile=%s' % (tmp_known_hosts.name,)])
370 if not strict_host_checking:
371 # Do not check for Host key. Unsafe.
372 args.extend(['-o', 'StrictHostKeyChecking=no'])
374 if isinstance(source, list):
377 if openssh_has_persist():
379 '-o', 'ControlMaster=auto',
380 '-o', 'ControlPath=%s' % (make_control_path(agent, False),)
386 log_msg = " rcopy - host %s - command %s " % (host, " ".join(args))
388 return _retry_rexec(args, log_msg, env = None, retry = retry,
389 tmp_known_hosts = tmp_known_hosts,
392 def rspawn(command, pidfile,
393 stdout = '/dev/null',
409 Spawn a remote command such that it will continue working asynchronously in
412 :param command: The command to run, it should be a single line.
415 :param pidfile: Path to a file where to store the pid and ppid of the
419 :param stdout: Path to file to redirect standard output.
420 The default value is /dev/null
423 :param stderr: Path to file to redirect standard error.
424 If the special STDOUT value is used, stderr will
425 be redirected to the same file as stdout
428 :param stdin: Path to a file with input to be piped into the command's standard input
431 :param home: Path to working directory folder.
432 It is assumed to exist unless the create_home flag is set.
435 :param create_home: Flag to force creation of the home folder before
437 :type create_home: bool
439 :param sudo: Flag forcing execution with sudo user
444 (stdout, stderr), process
446 Of the spawning process, which only captures errors at spawning time.
447 Usually only useful for diagnostics.
449 # Start process in a "daemonized" way, using nohup and heavy
450 # stdin/out redirection to avoid connection issues
454 stderr = ' ' + stderr
456 daemon_command = '{ { %(command)s > %(stdout)s 2>%(stderr)s < %(stdin)s & } ; echo $! 1 > %(pidfile)s ; }' % {
458 'pidfile' : shell_escape(pidfile),
464 cmd = "%(create)s%(gohome)s rm -f %(pidfile)s ; %(sudo)s nohup bash -c %(command)s " % {
465 'command' : shell_escape(daemon_command),
466 'sudo' : 'sudo -S' if sudo else '',
467 'pidfile' : shell_escape(pidfile),
468 'gohome' : 'cd %s ; ' % (shell_escape(home),) if home else '',
469 'create' : 'mkdir -p %s ; ' % (shell_escape(home),) if create_home and home else '',
472 (out,err),proc = rexec(
481 server_key = server_key,
486 raise RuntimeError, "Failed to set up application on host %s: %s %s" % (host, out,err,)
488 return ((out, err), proc)
501 Returns the pid and ppid of a process from a remote file where the
502 information was stored.
504 :param home: Path to directory where the pidfile is located
507 :param pidfile: Name of file containing the pid information
512 A (pid, ppid) tuple useful for calling rstatus and rkill,
513 or None if the pidfile isn't valid yet (can happen when process is staring up)
516 (out,err),proc = rexec(
517 "cat %(pidfile)s" % {
527 server_key = server_key
535 return map(int,out.strip().split(' ',1))
537 # Ignore, many ways to fail that don't matter that much
541 def rstatus(pid, ppid,
551 Returns a code representing the the status of a remote process
553 :param pid: Process id of the process
556 :param ppid: Parent process id of process
559 :rtype: int (One of NOT_STARTED, RUNNING, FINISHED)
562 (out,err),proc = rexec(
563 # Check only by pid. pid+ppid does not always work (especially with sudo)
564 " (( ps --pid %(pid)d -o pid | grep -c %(pid)d && echo 'wait') || echo 'done' ) | tail -n 1" % {
575 server_key = server_key
579 return ProcStatus.NOT_STARTED
583 if err.strip().find("Error, do this: mount -t proc none /proc") >= 0:
586 status = (out.strip() == 'wait')
588 return ProcStatus.NOT_STARTED
589 return ProcStatus.RUNNING if status else ProcStatus.FINISHED
604 Sends a kill signal to a remote process.
606 First tries a SIGTERM, and if the process does not end in 10 seconds,
609 :param pid: Process id of process to be killed
612 :param ppid: Parent process id of process to be killed
615 :param sudo: Flag indicating if sudo should be used to kill the process
619 subkill = "$(ps --ppid %(pid)d -o pid h)" % { 'pid' : pid }
621 SUBKILL="%(subkill)s" ;
622 %(sudo)s kill -- -%(pid)d $SUBKILL || /bin/true
623 %(sudo)s kill %(pid)d $SUBKILL || /bin/true
624 for x in 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 ; do
626 if [ `ps --pid %(pid)d -o pid | grep -c %(pid)d` == '0' ]; then
629 %(sudo)s kill -- -%(pid)d $SUBKILL || /bin/true
630 %(sudo)s kill %(pid)d $SUBKILL || /bin/true
634 if [ `ps --pid %(pid)d -o pid | grep -c %(pid)d` != '0' ]; then
635 %(sudo)s kill -9 -- -%(pid)d $SUBKILL || /bin/true
636 %(sudo)s kill -9 %(pid)d $SUBKILL || /bin/true
640 cmd = "( %s ) >/dev/null 2>/dev/null </dev/null &" % (cmd,)
642 (out,err),proc = rexec(
646 'sudo' : 'sudo -S' if sudo else '',
656 server_key = server_key
659 # wait, don't leave zombies around
662 return (out, err), proc
664 def _retry_rexec(args,
666 stdout = subprocess.PIPE,
667 stdin = subprocess.PIPE,
668 stderr = subprocess.PIPE,
671 tmp_known_hosts = None,
674 for x in xrange(retry):
675 # connects to the remote host and starts a remote connection
676 proc = subprocess.Popen(args,
682 # attach tempfile object to the process, to make sure the file stays
683 # alive until the process is finished with it
684 proc._known_hosts = tmp_known_hosts
686 # The argument block == False forces to rexec to return immediately,
691 (out, err) = proc.communicate()
693 out = proc.stdout.read()
694 if proc.poll() and stderr:
695 err = proc.stderr.read()
697 log(log_msg, logging.DEBUG, out, err)
702 if err.strip().startswith('ssh: ') or err.strip().startswith('mux_client_hello_exchange: '):
703 # SSH error, can safely retry
706 # Probably timed out or plain failed but can retry
711 msg = "SLEEPING %d ... ATEMPT %d - command %s " % (
712 t, x, " ".join(args))
713 log(msg, logging.DEBUG)
718 except RuntimeError, e:
719 msg = " rexec EXCEPTION - TIMEOUT -> %s \n %s" % ( e.args, log_msg )
720 log(msg, logging.DEBUG, out, err)
726 return ((out, err), proc)