do actually process server-specific [<server> LIMIT] sections
[hippotat] / hippotatlib / __init__.py
1 # -*- python -*-
2 #
3 # Hippotat - Asinine IP Over HTTP program
4 # hippotatlib/__init__.py - common library code
5 #
6 # Copyright 2017 Ian Jackson
7 #
8 # GPLv3+
9 #
10 # This program is free software: you can redistribute it and/or modify
11 # it under the terms of the GNU General Public License as published by
12 # the Free Software Foundation, either version 3 of the License, or
13 # (at your option) any later version.
14 #
15 # This program is distributed in the hope that it will be useful,
16 # but WITHOUT ANY WARRANTY; without even the implied warranty of
17 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 # GNU General Public License for more details.
19 #
20 # You should have received a copy of the GNU General Public License
21 # along with this program, in the file GPLv3. If not,
22 # see <http://www.gnu.org/licenses/>.
23
24
25 import signal
26 signal.signal(signal.SIGINT, signal.SIG_DFL)
27
28 import sys
29 import os
30
31 from zope.interface import implementer
32
33 import twisted
34 from twisted.internet import reactor
35 import twisted.internet.endpoints
36 import twisted.logger
37 from twisted.logger import LogLevel
38 import twisted.python.constants
39 from twisted.python.constants import NamedConstant
40
41 import ipaddress
42 from ipaddress import AddressValueError
43
44 from optparse import OptionParser
45 import configparser
46 from configparser import ConfigParser
47 from configparser import NoOptionError
48
49 from functools import partial
50
51 import collections
52 import time
53 import codecs
54 import traceback
55
56 import re as regexp
57
58 import hippotatlib.slip as slip
59
60 class DBG(twisted.python.constants.Names):
61 INIT = NamedConstant()
62 CONFIG = NamedConstant()
63 ROUTE = NamedConstant()
64 DROP = NamedConstant()
65 OWNSOURCE = NamedConstant()
66 FLOW = NamedConstant()
67 HTTP = NamedConstant()
68 TWISTED = NamedConstant()
69 QUEUE = NamedConstant()
70 HTTP_CTRL = NamedConstant()
71 QUEUE_CTRL = NamedConstant()
72 HTTP_FULL = NamedConstant()
73 CTRL_DUMP = NamedConstant()
74 SLIP_FULL = NamedConstant()
75 DATA_COMPLETE = NamedConstant()
76
77 _hex_codec = codecs.getencoder('hex_codec')
78
79 #---------- logging ----------
80
81 org_stderr = sys.stderr
82
83 log = twisted.logger.Logger()
84
85 debug_set = set()
86 debug_def_detail = DBG.HTTP
87
88 def log_debug(dflag, msg, idof=None, d=None):
89 if dflag not in debug_set: return
90 #print('---------------->',repr((dflag, msg, idof, d)), file=sys.stderr)
91 if idof is not None:
92 msg = '[%#x] %s' % (id(idof), msg)
93 if d is not None:
94 trunc = ''
95 if not DBG.DATA_COMPLETE in debug_set:
96 if len(d) > 64:
97 d = d[0:64]
98 trunc = '...'
99 d = _hex_codec(d)[0].decode('ascii')
100 msg += ' ' + d + trunc
101 log.info('{dflag} {msgcore}', dflag=dflag, msgcore=msg)
102
103 def logevent_is_boringtwisted(event):
104 try:
105 if event.get('log_level') != LogLevel.info:
106 return False
107 dflag = event.get('dflag')
108 if dflag is False : return False
109 if dflag in debug_set: return False
110 if dflag is None and DBG.TWISTED in debug_set: return False
111 return True
112 except Exception:
113 print('EXCEPTION (IN BORINGTWISTED CHECK)',
114 traceback.format_exc(), file=org_stderr)
115 return False
116
117 @implementer(twisted.logger.ILogFilterPredicate)
118 class LogNotBoringTwisted:
119 def __call__(self, event):
120 return (
121 twisted.logger.PredicateResult.no
122 if logevent_is_boringtwisted(event) else
123 twisted.logger.PredicateResult.yes
124 )
125
126 #---------- default config ----------
127
128 defcfg = '''
129 [COMMON]
130 max_batch_down = 65536
131 max_queue_time = 10
132 target_requests_outstanding = 3
133 http_timeout = 30
134 http_timeout_grace = 5
135 max_requests_outstanding = 6
136 max_batch_up = 4000
137 http_retry = 5
138 port = 80
139 vroutes = ''
140 ifname_client = hippo%%d
141 ifname_server = shippo%%d
142
143 #[server] or [<client>] overrides
144 ipif = userv root ipif %(local)s,%(peer)s,%(mtu)s,slip,%(ifname)s %(rnets)s
145
146 # relating to virtual network
147 mtu = 1500
148
149 # addrs = 127.0.0.1 ::1
150 # url
151
152 # relating to virtual network
153 vvnetwork = 172.24.230.192
154 # vnetwork = <prefix>/<len>
155 # vaddr = <ipaddr>
156 # vrelay = <ipaddr>
157
158
159 # [<client-ip4-or-ipv6-address>]
160 # password = <password> # used by both, must match
161
162 [LIMIT]
163 max_batch_down = 262144
164 max_queue_time = 121
165 http_timeout = 121
166 target_requests_outstanding = 10
167 '''
168
169 # these need to be defined here so that they can be imported by import *
170 cfg = ConfigParser(strict=False)
171 optparser = OptionParser()
172
173 _mimetrans = bytes.maketrans(b'-'+slip.esc, slip.esc+b'-')
174 def mime_translate(s):
175 # SLIP-encoded packets cannot contain ESC ESC.
176 # Swap `-' and ESC. The result cannot contain `--'
177 return s.translate(_mimetrans)
178
179 class ConfigResults:
180 def __init__(self):
181 pass
182 def __repr__(self):
183 return 'ConfigResults('+repr(self.__dict__)+')'
184
185 def log_discard(packet, iface, saddr, daddr, why):
186 log_debug(DBG.DROP,
187 'discarded packet [%s] %s -> %s: %s' % (iface, saddr, daddr, why),
188 d=packet)
189
190 #---------- packet parsing ----------
191
192 def packet_addrs(packet):
193 version = packet[0] >> 4
194 if version == 4:
195 addrlen = 4
196 saddroff = 3*4
197 factory = ipaddress.IPv4Address
198 elif version == 6:
199 addrlen = 16
200 saddroff = 2*4
201 factory = ipaddress.IPv6Address
202 else:
203 raise ValueError('unsupported IP version %d' % version)
204 saddr = factory(packet[ saddroff : saddroff + addrlen ])
205 daddr = factory(packet[ saddroff + addrlen : saddroff + addrlen*2 ])
206 return (saddr, daddr)
207
208 #---------- address handling ----------
209
210 def ipaddr(input):
211 try:
212 r = ipaddress.IPv4Address(input)
213 except AddressValueError:
214 r = ipaddress.IPv6Address(input)
215 return r
216
217 def ipnetwork(input):
218 try:
219 r = ipaddress.IPv4Network(input)
220 except NetworkValueError:
221 r = ipaddress.IPv6Network(input)
222 return r
223
224 #---------- ipif (SLIP) subprocess ----------
225
226 class SlipStreamDecoder():
227 def __init__(self, desc, on_packet):
228 self._buffer = b''
229 self._on_packet = on_packet
230 self._desc = desc
231 self._log('__init__')
232
233 def _log(self, msg, **kwargs):
234 log_debug(DBG.SLIP_FULL, 'slip %s: %s' % (self._desc, msg), **kwargs)
235
236 def inputdata(self, data):
237 self._log('inputdata', d=data)
238 data = self._buffer + data
239 self._buffer = b''
240 packets = slip.decode(data, True)
241 self._buffer = packets.pop()
242 for packet in packets:
243 self._maybe_packet(packet)
244 self._log('bufremain', d=self._buffer)
245
246 def _maybe_packet(self, packet):
247 self._log('maybepacket', d=packet)
248 if len(packet):
249 self._on_packet(packet)
250
251 def flush(self):
252 self._log('flush')
253 data = self._buffer
254 self._buffer = b''
255 packets = slip.decode(data)
256 assert(len(packets) == 1)
257 self._maybe_packet(packets[0])
258
259 class _IpifProcessProtocol(twisted.internet.protocol.ProcessProtocol):
260 def __init__(self, router):
261 self._router = router
262 self._decoder = SlipStreamDecoder('ipif', self.slip_on_packet)
263 def connectionMade(self): pass
264 def outReceived(self, data):
265 self._decoder.inputdata(data)
266 def slip_on_packet(self, packet):
267 (saddr, daddr) = packet_addrs(packet)
268 if saddr.is_link_local or daddr.is_link_local:
269 log_discard(packet, 'ipif', saddr, daddr, 'link-local')
270 return
271 self._router(packet, saddr, daddr)
272 def processEnded(self, status):
273 status.raiseException()
274
275 def start_ipif(command, router):
276 ipif = _IpifProcessProtocol(router)
277 reactor.spawnProcess(ipif,
278 '/bin/sh',['sh','-xc', command],
279 childFDs={0:'w', 1:'r', 2:2},
280 env=None)
281 return ipif
282
283 def queue_inbound(ipif, packet):
284 log_debug(DBG.FLOW, "queue_inbound", d=packet)
285 ipif.transport.write(slip.delimiter)
286 ipif.transport.write(slip.encode(packet))
287 ipif.transport.write(slip.delimiter)
288
289 #---------- packet queue ----------
290
291 class PacketQueue():
292 def __init__(self, desc, max_queue_time):
293 self._desc = desc
294 assert(desc + '')
295 self._max_queue_time = max_queue_time
296 self._pq = collections.deque() # packets
297
298 def _log(self, dflag, msg, **kwargs):
299 log_debug(dflag, self._desc+' pq: '+msg, **kwargs)
300
301 def append(self, packet):
302 self._log(DBG.QUEUE, 'append', d=packet)
303 self._pq.append((time.monotonic(), packet))
304
305 def nonempty(self):
306 self._log(DBG.QUEUE, 'nonempty ?')
307 while True:
308 try: (queuetime, packet) = self._pq[0]
309 except IndexError:
310 self._log(DBG.QUEUE, 'nonempty ? empty.')
311 return False
312
313 age = time.monotonic() - queuetime
314 if age > self._max_queue_time:
315 # strip old packets off the front
316 self._log(DBG.QUEUE, 'dropping (old)', d=packet)
317 self._pq.popleft()
318 continue
319
320 self._log(DBG.QUEUE, 'nonempty ? nonempty.')
321 return True
322
323 def process(self, sizequery, moredata, max_batch):
324 # sizequery() should return size of batch so far
325 # moredata(s) should add s to batch
326 self._log(DBG.QUEUE, 'process...')
327 while True:
328 try: (dummy, packet) = self._pq[0]
329 except IndexError:
330 self._log(DBG.QUEUE, 'process... empty')
331 break
332
333 self._log(DBG.QUEUE_CTRL, 'process... packet', d=packet)
334
335 encoded = slip.encode(packet)
336 sofar = sizequery()
337
338 self._log(DBG.QUEUE_CTRL,
339 'process... (sofar=%d, max=%d) encoded' % (sofar, max_batch),
340 d=encoded)
341
342 if sofar > 0:
343 if sofar + len(slip.delimiter) + len(encoded) > max_batch:
344 self._log(DBG.QUEUE_CTRL, 'process... overflow')
345 break
346 moredata(slip.delimiter)
347
348 moredata(encoded)
349 self._pq.popleft()
350
351 #---------- error handling ----------
352
353 _crashing = False
354
355 def crash(err):
356 global _crashing
357 _crashing = True
358 print('========== CRASH ==========', err,
359 '===========================', file=sys.stderr)
360 try: reactor.stop()
361 except twisted.internet.error.ReactorNotRunning: pass
362
363 def crash_on_defer(defer):
364 defer.addErrback(lambda err: crash(err))
365
366 def crash_on_critical(event):
367 if event.get('log_level') >= LogLevel.critical:
368 crash(twisted.logger.formatEvent(event))
369
370 #---------- config processing ----------
371
372 def _cfg_process_putatives():
373 servers = { }
374 clients = { }
375 # maps from abstract object to canonical name for cs's
376
377 def putative(cmap, abstract, canoncs):
378 try:
379 current_canoncs = cmap[abstract]
380 except KeyError:
381 pass
382 else:
383 assert(current_canoncs == canoncs)
384 cmap[abstract] = canoncs
385
386 server_pat = r'[-.0-9A-Za-z]+'
387 client_pat = r'[.:0-9a-f]+'
388 server_re = regexp.compile(server_pat)
389 serverclient_re = regexp.compile(
390 server_pat + r' ' + '(?:' + client_pat + '|LIMIT)')
391
392 for cs in cfg.sections():
393 if cs == 'LIMIT' or cs == 'COMMON':
394 # plan A "[LIMIT]" or "[COMMON]"
395 continue
396
397 try:
398 # plan B "[<client>]" part 1
399 ci = ipaddr(cs)
400 except AddressValueError:
401
402 if server_re.fullmatch(cs):
403 # plan C "[<servername>]"
404 putative(servers, cs, cs)
405 continue
406
407 if serverclient_re.fullmatch(cs):
408 # plan D "[<servername> <client>]" part 1
409 (pss,pcs) = cs.split(' ')
410
411 if pcs == 'LIMIT':
412 # plan E "[<servername> LIMIT]"
413 continue
414
415 try:
416 # plan D "[<servername> <client>]" part 2
417 ci = ipaddr(pc)
418 except AddressValueError:
419 # plan F "[<some thing we do not understand>]"
420 # well, we ignore this
421 print('warning: ignoring config section %s' % cs, file=sys.stderr)
422 continue
423
424 else: # no AddressValueError
425 # plan D "[<servername> <client]" part 3
426 putative(clients, ci, pcs)
427 putative(servers, pss, pss)
428 continue
429
430 else: # no AddressValueError
431 # plan B "[<client>" part 2
432 putative(clients, ci, cs)
433 continue
434
435 return (servers, clients)
436
437 def cfg_process_general(c, ss):
438 c.mtu = cfg1getint(ss, 'mtu')
439
440 def cfg_process_saddrs(c, ss):
441 class ServerAddr():
442 def __init__(self, port, addrspec):
443 self.port = port
444 # also self.addr
445 try:
446 self.addr = ipaddress.IPv4Address(addrspec)
447 self._endpointfactory = twisted.internet.endpoints.TCP4ServerEndpoint
448 self._inurl = b'%s'
449 except AddressValueError:
450 self.addr = ipaddress.IPv6Address(addrspec)
451 self._endpointfactory = twisted.internet.endpoints.TCP6ServerEndpoint
452 self._inurl = b'[%s]'
453 def make_endpoint(self):
454 return self._endpointfactory(reactor, self.port,
455 interface= '%s' % self.addr)
456 def url(self):
457 url = b'http://' + (self._inurl % str(self.addr).encode('ascii'))
458 if self.port != 80: url += b':%d' % self.port
459 url += b'/'
460 return url
461 def __repr__(self):
462 return 'ServerAddr'+repr((self.port,self.addr))
463
464 c.port = cfg1getint(ss,'port')
465 c.saddrs = [ ]
466 for addrspec in cfg1get(ss, 'addrs').split():
467 sa = ServerAddr(c.port, addrspec)
468 c.saddrs.append(sa)
469
470 def cfg_process_vnetwork(c, ss):
471 c.vnetwork = ipnetwork(cfg1get(ss,'vnetwork'))
472 if c.vnetwork.num_addresses < 3 + 2:
473 raise ValueError('vnetwork needs at least 2^3 addresses')
474
475 def cfg_process_vaddr(c, ss):
476 try:
477 c.vaddr = cfg1get(ss,'vaddr')
478 except NoOptionError:
479 cfg_process_vnetwork(c, ss)
480 c.vaddr = next(c.vnetwork.hosts())
481
482 def cfg_search_section(key,sections):
483 for section in sections:
484 if cfg.has_option(section, key):
485 return section
486 raise NoOptionError(key, repr(sections))
487
488 def cfg_get_raw(*args, **kwargs):
489 # for passing to cfg_search
490 return cfg.get(*args, raw=True, **kwargs)
491
492 def cfg_search(getter,key,sections):
493 section = cfg_search_section(key,sections)
494 return getter(section, key)
495
496 def cfg1get(section,key, getter=cfg.get,**kwargs):
497 section = cfg_search_section(key,[section,'COMMON'])
498 return getter(section,key,**kwargs)
499
500 def cfg1getint(section,key, **kwargs):
501 return cfg1get(section,key, getter=cfg.getint,**kwargs);
502
503 def cfg_process_client_limited(cc,ss,sections,key):
504 val = cfg_search(cfg1getint, key, sections)
505 lim = cfg_search(cfg1getint, key, ['%s LIMIT' % ss, 'LIMIT'])
506 cc.__dict__[key] = min(val,lim)
507
508 def cfg_process_client_common(cc,ss,cs,ci):
509 # returns sections to search in, iff password is defined, otherwise None
510 cc.ci = ci
511
512 sections = ['%s %s' % (ss,cs),
513 cs,
514 ss,
515 'COMMON']
516
517 try: pwsection = cfg_search_section('password', sections)
518 except NoOptionError: return None
519
520 pw = cfg1get(pwsection, 'password')
521 cc.password = pw.encode('utf-8')
522
523 cfg_process_client_limited(cc,ss,sections,'target_requests_outstanding')
524 cfg_process_client_limited(cc,ss,sections,'http_timeout')
525
526 return sections
527
528 def cfg_process_ipif(c, sections, varmap):
529 for d, s in varmap:
530 try: v = getattr(c, s)
531 except AttributeError: continue
532 setattr(c, d, v)
533
534 #print('CFGIPIF',repr((varmap, sections, c.__dict__)),file=sys.stderr)
535
536 section = cfg_search_section('ipif', sections)
537 c.ipif_command = cfg1get(section,'ipif', vars=c.__dict__)
538
539 #---------- startup ----------
540
541 def log_debug_config(m):
542 if not DBG.CONFIG in debug_set: return
543 print('DBG.CONFIG:', m)
544
545 def common_startup(process_cfg):
546 # calls process_cfg(putative_clients, putative_servers)
547
548 # ConfigParser hates #-comments after values
549 trailingcomments_re = regexp.compile(r'#.*')
550 cfg.read_string(trailingcomments_re.sub('', defcfg))
551 need_defcfg = True
552
553 def readconfig(pathname, mandatory=True):
554 def log(m, p=pathname):
555 if not DBG.CONFIG in debug_set: return
556 log_debug_config('%s: %s' % (m, p))
557
558 try:
559 files = os.listdir(pathname)
560
561 except FileNotFoundError:
562 if mandatory: raise
563 log('skipped')
564 return
565
566 except NotADirectoryError:
567 cfg.read(pathname)
568 log('read file')
569 return
570
571 # is a directory
572 log('directory')
573 re = regexp.compile('[^-A-Za-z0-9_]')
574 for f in os.listdir(pathname):
575 if re.search(f): continue
576 subpath = pathname + '/' + f
577 try:
578 os.stat(subpath)
579 except FileNotFoundError:
580 log('entry skipped', subpath)
581 continue
582 cfg.read(subpath)
583 log('entry read', subpath)
584
585 def oc_config(od,os, value, op):
586 nonlocal need_defcfg
587 need_defcfg = False
588 readconfig(value)
589
590 def oc_extra_config(od,os, value, op):
591 readconfig(value)
592
593 def read_defconfig():
594 readconfig('/etc/hippotat/config.d', False)
595 readconfig('/etc/hippotat/passwords.d', False)
596 readconfig('/etc/hippotat/master.cfg', False)
597
598 def oc_defconfig(od,os, value, op):
599 nonlocal need_defcfg
600 need_defcfg = False
601 read_defconfig(value)
602
603 def dfs_less_detailed(dl):
604 return [df for df in DBG.iterconstants() if df <= dl]
605
606 def ds_default(od,os,dl,op):
607 global debug_set
608 debug_set.clear
609 debug_set |= set(dfs_less_detailed(debug_def_detail))
610
611 def ds_select(od,os, spec, op):
612 for it in spec.split(','):
613
614 if it.startswith('-'):
615 mutator = debug_set.discard
616 it = it[1:]
617 else:
618 mutator = debug_set.add
619
620 if it == '+':
621 dfs = DBG.iterconstants()
622
623 else:
624 if it.endswith('+'):
625 mapper = dfs_less_detailed
626 it = it[0:len(it)-1]
627 else:
628 mapper = lambda x: [x]
629
630 try:
631 dfspec = DBG.lookupByName(it)
632 except ValueError:
633 optparser.error('unknown debug flag %s in --debug-select' % it)
634
635 dfs = mapper(dfspec)
636
637 for df in dfs:
638 mutator(df)
639
640 optparser.add_option('-D', '--debug',
641 nargs=0,
642 action='callback',
643 help='enable default debug (to stdout)',
644 callback= ds_default)
645
646 optparser.add_option('--debug-select',
647 nargs=1,
648 type='string',
649 metavar='[-]DFLAG[+]|[-]+,...',
650 help=
651 '''enable (`-': disable) each specified DFLAG;
652 `+': do same for all "more interesting" DFLAGSs;
653 just `+': all DFLAGs.
654 DFLAGS: ''' + ' '.join([df.name for df in DBG.iterconstants()]),
655 action='callback',
656 callback= ds_select)
657
658 optparser.add_option('-c', '--config',
659 nargs=1,
660 type='string',
661 metavar='CONFIGFILE',
662 dest='configfile',
663 action='callback',
664 callback= oc_config)
665
666 optparser.add_option('--extra-config',
667 nargs=1,
668 type='string',
669 metavar='CONFIGFILE',
670 dest='configfile',
671 action='callback',
672 callback= oc_extra_config)
673
674 optparser.add_option('--default-config',
675 action='callback',
676 callback= oc_defconfig)
677
678 (opts, args) = optparser.parse_args()
679 if len(args): optparser.error('no non-option arguments please')
680
681 if need_defcfg:
682 read_defconfig()
683
684 try:
685 (pss, pcs) = _cfg_process_putatives()
686 process_cfg(opts, pss, pcs)
687 except (configparser.Error, ValueError):
688 traceback.print_exc(file=sys.stderr)
689 print('\nInvalid configuration, giving up.', file=sys.stderr)
690 sys.exit(12)
691
692
693 #print('X', debug_set, file=sys.stderr)
694
695 log_formatter = twisted.logger.formatEventAsClassicLogText
696 stdout_obs = twisted.logger.FileLogObserver(sys.stdout, log_formatter)
697 stderr_obs = twisted.logger.FileLogObserver(sys.stderr, log_formatter)
698 pred = twisted.logger.LogLevelFilterPredicate(LogLevel.error)
699 stdsomething_obs = twisted.logger.FilteringLogObserver(
700 stderr_obs, [pred], stdout_obs
701 )
702 global file_log_observer
703 file_log_observer = twisted.logger.FilteringLogObserver(
704 stdsomething_obs, [LogNotBoringTwisted()]
705 )
706 #log_observer = stdsomething_obs
707 twisted.logger.globalLogBeginner.beginLoggingTo(
708 [ file_log_observer, crash_on_critical ]
709 )
710
711 def common_run():
712 log_debug(DBG.INIT, 'entering reactor')
713 if not _crashing: reactor.run()
714 print('ENDED', file=sys.stderr)
715 sys.exit(16)