zeek/auxil/zeekctl/ZeekControl/ssh_runner.py
Patrick Kelley 8fd444092b initial
2025-05-07 15:35:15 -04:00

371 lines
10 KiB
Python

import ast
import base64
import collections
import json
import logging
import os
import select
import subprocess
import time
import zlib
from queue import Empty, Queue
from threading import Thread
def get_muxer(shell):
# The full path of the Python interpreter. Configured by CMake.
pythonpath = "@Python_EXECUTABLE@"
muxer = r"""
import os,sys,subprocess,signal,select,json
TIMEOUT=120
def w(s):
sys.stdout.write(repr(s) + "\n")
sys.stdout.flush()
def exec_cmds(cmds):
p=[]
for i,cmd in enumerate(cmds):
try:
proc=subprocess.Popen(cmd,stdout=subprocess.PIPE,stderr=subprocess.PIPE __SHELL__)
p.append((i,proc))
except Exception as e:
w((i,(1,'',str(e))))
return p
w("ready")
commands=[]
signal.alarm(TIMEOUT)
for line in iter(sys.stdin.readline,"done\n"):
commands.append(json.loads(line))
procs=exec_cmds(commands)
cmd_map={}
fd_map={}
fds=set()
for i,proc in procs:
o={"idx":i, "proc":proc, "stdout":[], "stderr":[], "waiting":2}
fd_map[proc.stdout]=o["stdout"]
fd_map[proc.stderr]=o["stderr"]
cmd_map[proc.stdout]=o
cmd_map[proc.stderr]=o
fds.update((proc.stdout,proc.stderr))
while fds:
r,_,_=select.select(fds,[],[])
for fd in r:
output=os.read(fd.fileno(),1024)
if output:
fd_map[fd].append(output)
continue
cmd=cmd_map[fd]
fds.remove(fd)
cmd["waiting"]-=1
if cmd["waiting"]:
continue
proc=cmd["proc"]
status=proc.wait()
out=b"".join(cmd["stdout"])
err=b"".join(cmd["stderr"])
w((cmd["idx"],(status,out,err)))
w("done")
"""
if shell:
muxer = muxer.replace("__SHELL__", ",shell=True")
else:
muxer = muxer.replace("__SHELL__", "")
muxer = muxer.encode()
muxer = base64.b64encode(zlib.compress(muxer))
muxer = muxer.decode()
muxer = f"{pythonpath} -c 'import zlib,base64; exec(zlib.decompress(base64.b64decode(b\"{muxer}\")))'\n"
muxer = muxer.encode()
return muxer
CmdResult = collections.namedtuple("CmdResult", "status stdout stderr")
class SSHMaster:
def __init__(self, host, localaddrs):
# The BatchMode=yes disables interactive prompting. The LogLevel=error
# prevents seeing login banners but allows error messages from ssh.
self.base_cmd = [
"ssh",
"-o",
"BatchMode=yes",
"-o",
"LogLevel=error",
host,
]
self.host = host
self.need_connect = True
self.master = None
self.localaddrs = localaddrs
self.run_mux = get_muxer(False)
self.run_mux_shell = get_muxer(True)
def connect(self):
if self.need_connect:
if self.host in self.localaddrs:
cmd = ["sh"]
else:
cmd = self.base_cmd + ["sh"]
self.master = subprocess.Popen(
cmd,
bufsize=0,
stdout=subprocess.PIPE,
stdin=subprocess.PIPE,
close_fds=True,
preexec_fn=os.setsid,
)
self.need_connect = False
def readline_with_timeout(self, timeout):
readable, _, _ = select.select([self.master.stdout], [], [], timeout)
if not readable:
return None
jtxt = self.master.stdout.readline()
jtxt = jtxt.decode()
return jtxt
def exec_command(self, cmd, shell=False, timeout=60):
return self.exec_commands([cmd], shell, timeout)[0]
def exec_commands(self, cmds, shell=False, timeout=60):
self.send_commands(cmds, timeout, shell)
return self.collect_results(timeout)
def send_commands(self, cmds, timeout, shell=False):
self.connect()
if shell:
self.master.stdin.write(self.run_mux_shell)
else:
self.master.stdin.write(self.run_mux)
self.master.stdin.flush()
# Wait until we receive the "ready" message from muxer script
self.readline_with_timeout(timeout)
for cmd in cmds:
jcmd = f"{json.dumps(cmd)}\n"
jcmd = jcmd.encode()
self.master.stdin.write(jcmd)
# Note: the "b" string prefix here for Py3 is ignored by Py2.6-2.7
self.master.stdin.write(b"done\n")
self.master.stdin.flush()
self.sent_commands = len(cmds)
def collect_results(self, timeout):
outputs = [
Exception(f"Command timeout on host {self.host}")
] * self.sent_commands
while True:
line = self.readline_with_timeout(timeout)
if not line:
logging.debug("Command timeout on host %s", self.host)
self.close()
break
resp = ast.literal_eval(line)
if resp == "done":
break
idx, result = resp
status, out, err = result
out = out.decode(errors="replace")
err = err.decode(errors="replace")
outputs[idx] = CmdResult(status, out, err)
return outputs
def close(self):
if not self.master:
return
self.master.stdin.close()
try:
self.master.kill()
except OSError:
pass
self.master.wait()
self.master = None
self.need_connect = True
__del__ = close
STOP_RUNNING = object()
class HostHandler(Thread):
def __init__(self, host, localaddrs, timeout):
self.host = host
self.localaddrs = localaddrs
self.timeout = timeout
self.q = Queue()
self.alive = False
self.master = None
Thread.__init__(self)
def shutdown(self):
self.q.put((STOP_RUNNING, None, None))
def connect(self):
if self.master:
self.master.close()
self.master = SSHMaster(self.host, self.localaddrs)
def ping(self):
# Error message should indicate whether or not ssh is being used.
msgstr = "" if self.host in self.localaddrs else "ssh "
# Error message shows if a connection was previously established.
if self.alive:
msg = f"Lost {msgstr}connection to host {self.host}"
else:
msg = f"Failed to establish {msgstr}connection to host {self.host}"
# This will be set to True below only if the "ping" is received.
self.alive = False
try:
resp = self.master.exec_command(["/bin/echo", "ping"], timeout=10)
except Exception as e:
# This happens most likely due to broken pipe (i.e., ssh
# terminates, usually because it couldn't connect, or its own
# timeout occurred).
return f"{msg}: {e}"
try:
ping_recvd = resp.stdout.strip() == "ping"
except Exception:
# This happens when there was a timeout in SSHMaster (in this
# situation, that almost always means unable to establish
# connection or a loss of connection).
return msg
if ping_recvd:
self.alive = True
return ""
# This should probably never happen.
return f"Communication failure with host {self.host} when checking connection"
def connect_and_ping(self):
if not self.alive:
self.connect()
return self.ping()
def run(self):
while True:
if self.iteration():
return
def iteration(self):
try:
item, shell, rq = self.q.get(timeout=30)
except Empty:
self.connect_and_ping()
return False
if item is STOP_RUNNING:
return True
msg = self.connect_and_ping()
if not self.alive:
logging.debug(msg)
resp = [Exception(msg)] * len(item)
rq.put(resp)
return False
try:
resp = self.master.exec_commands(item, shell, self.timeout)
except Exception as e:
self.alive = False
msgstr = "" if self.host in self.localaddrs else "ssh "
msg = f"Lost {msgstr}connection while running command on host {self.host}: {e}"
logging.debug(msg)
resp = [Exception(msg)] * len(item)
time.sleep(2)
rq.put(resp)
return False
def send_commands(self, commands, shell, rq):
self.q.put((commands, shell, rq))
class MultiMasterManager:
def __init__(self, localaddrs=[]):
self.masters = {}
self.response_queues = {}
self.localaddrs = localaddrs
def setup(self, host, timeout):
if host not in self.masters:
self.masters[host] = HostHandler(host, self.localaddrs, timeout)
self.masters[host].start()
def send_commands(self, host, commands, timeout, shell=False):
self.setup(host, timeout)
rq = Queue()
self.response_queues[host] = rq
self.masters[host].send_commands(commands, shell, rq)
def get_result(self, host, hosttimeout):
# Add a few seconds to the host timeout in order to let the
# command timeout happen first.
hosttimeout += 5
rq = self.response_queues[host]
try:
return rq.get(timeout=hosttimeout)
except Empty:
self.shutdown(host)
# This can happen due to commands that take a while to run, a
# loss of connectivity to remote host, or both.
return [
Exception(f"Timeout waiting for commands to finish on host {host}")
] # FIXME: needs to be the right length
def exec_command(self, host, command, timeout=30):
return self.exec_commands(host, [command], timeout)[0]
def exec_commands(self, host, commands, timeout=60):
self.send_commands(host, commands, timeout)
return self.get_result(host, timeout)
def exec_multihost_commands(self, cmds, shell=False, timeout=60):
hosts = collections.defaultdict(list)
for host, cmd in cmds:
hosts[host].append(cmd)
for host, cmds in hosts.items():
self.send_commands(host, cmds, timeout, shell)
for host in hosts:
for res in self.get_result(host, timeout):
yield host, res
def host_status(self):
for h, o in self.masters.items():
if h not in self.localaddrs:
yield h, o.alive
def shutdown(self, host):
self.masters[host].shutdown()
del self.masters[host]
def shutdown_all(self):
for handler in self.masters.values():
handler.shutdown()
self.masters = {}
__del__ = shutdown_all