Commit | Line | Data |
---|---|---|
6005ef9b MW |
1 | #! @PYTHON@ |
2 | ### -*-python-*- | |
3 | ### | |
4 | ### Build a CDB file from configuration file | |
5 | ### | |
6 | ### (c) 2007 Straylight/Edgeware | |
7 | ### | |
8 | ||
9 | ###----- Licensing notice --------------------------------------------------- | |
10 | ### | |
11 | ### This file is part of Trivial IP Encryption (TrIPE). | |
12 | ### | |
11ad66c2 MW |
13 | ### TrIPE is free software: you can redistribute it and/or modify it under |
14 | ### the terms of the GNU General Public License as published by the Free | |
15 | ### Software Foundation; either version 3 of the License, or (at your | |
16 | ### option) any later version. | |
6005ef9b | 17 | ### |
11ad66c2 MW |
18 | ### TrIPE is distributed in the hope that it will be useful, but WITHOUT |
19 | ### ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
20 | ### FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
21 | ### for more details. | |
6005ef9b MW |
22 | ### |
23 | ### You should have received a copy of the GNU General Public License | |
11ad66c2 | 24 | ### along with TrIPE. If not, see <https://www.gnu.org/licenses/>. |
6005ef9b MW |
25 | |
26 | VERSION = '@VERSION@' | |
27 | ||
28 | ###-------------------------------------------------------------------------- | |
29 | ### External dependencies. | |
30 | ||
6005ef9b MW |
31 | import mLib as M |
32 | from optparse import OptionParser | |
33 | import cdb as CDB | |
34 | from sys import stdin, stdout, exit, argv | |
35 | import re as RX | |
36 | import os as OS | |
b7e5aa06 | 37 | from cStringIO import StringIO |
6005ef9b MW |
38 | |
39 | ###-------------------------------------------------------------------------- | |
40 | ### Utilities. | |
41 | ||
42 | class CDBFake (object): | |
43 | """Like cdbmake, but just outputs data suitable for cdb-map.""" | |
44 | def __init__(me, file = stdout): | |
45 | me.file = file | |
46 | def add(me, key, value): | |
47 | me.file.write('%s:%s\n' % (key, value)) | |
48 | def finish(me): | |
49 | pass | |
50 | ||
51 | ###-------------------------------------------------------------------------- | |
52 | ### A bulk DNS resolver. | |
53 | ||
54 | class BulkResolver (object): | |
55 | """ | |
56 | Resolve a number of DNS names in parallel. | |
57 | ||
58 | The BulkResovler resolves a number of hostnames in parallel. Using it | |
59 | works in three phases: | |
60 | ||
61 | 1. You call prepare(HOSTNAME) a number of times, to feed in the hostnames | |
62 | you're interested in. | |
63 | ||
64 | 2. You call run() to actually drive the resolver. | |
65 | ||
66 | 3. You call lookup(HOSTNAME) to get the address you wanted. This will | |
67 | fail with KeyError if the resolver couldn't resolve the HOSTNAME. | |
68 | """ | |
69 | ||
70 | def __init__(me): | |
71 | """Initialize the resolver.""" | |
72 | me._resolvers = {} | |
73 | me._namemap = {} | |
74 | ||
75 | def prepare(me, host): | |
76 | """Prime the resolver to resolve the name HOST.""" | |
d8310a3a MW |
77 | if host not in me._resolvers: |
78 | me._resolvers[host] = M.SelResolveByName \ | |
79 | (host, | |
80 | lambda name, alias, addr: | |
81 | me._resolved(host, addr[0]), | |
82 | lambda: me._resolved(host, None)) | |
6005ef9b MW |
83 | |
84 | def run(me): | |
85 | """Run the background DNS resolver until it's finished.""" | |
86 | while me._resolvers: | |
87 | M.select() | |
88 | ||
89 | def lookup(me, host): | |
90 | """ | |
91 | Fetch the address corresponding to HOST. | |
92 | """ | |
93 | addr = me._namemap[host] | |
94 | if addr is None: | |
171206b5 | 95 | raise KeyError(host) |
6005ef9b MW |
96 | return addr |
97 | ||
98 | def _resolved(me, host, addr): | |
99 | """Callback function: remember that ADDR is the address for HOST.""" | |
100 | me._namemap[host] = addr | |
101 | del me._resolvers[host] | |
102 | ||
103 | ###-------------------------------------------------------------------------- | |
104 | ### The configuration parser. | |
105 | ||
b7e5aa06 MW |
106 | ## Match a comment or empty line. |
107 | RX_COMMENT = RX.compile(r'(?x) ^ \s* (?: $ | [;#])') | |
108 | ||
109 | ## Match a section group header. | |
110 | RX_GRPHDR = RX.compile(r'(?x) ^ \s* \[ (.*) \] \s* $') | |
111 | ||
112 | ## Match an assignment line. | |
113 | RX_ASSGN = RX.compile(r'''(?x) ^ | |
114 | ([^\s:=] (?: [^:=]* [^\s:=])?) | |
115 | \s* [:=] \s* | |
116 | (| \S | \S.*\S) | |
117 | \s* $''') | |
118 | ||
119 | ## Match a continuation line. | |
120 | RX_CONT = RX.compile(r'''(?x) ^ \s+ | |
121 | (| \S | \S.*\S) | |
122 | \s* $''') | |
123 | ||
6005ef9b | 124 | ## Match a $(VAR) configuration variable reference; group 1 is the VAR. |
2d51bc9f | 125 | RX_REF = RX.compile(r'(?x) \$ \( ([^)]+) \)') |
6005ef9b MW |
126 | |
127 | ## Match a $[HOST] name resolution reference; group 1 is the HOST. | |
2d51bc9f | 128 | RX_RESOLVE = RX.compile(r'(?x) \$ \[ ([^]]+) \]') |
6005ef9b | 129 | |
b7e5aa06 MW |
130 | class ConfigSyntaxError (Exception): |
131 | def __init__(me, fname, lno, msg): | |
132 | me.fname = fname | |
133 | me.lno = lno | |
134 | me.msg = msg | |
135 | def __str__(me): | |
136 | return '%s:%d: %s' % (me.fname, me.lno, me.msg) | |
137 | ||
bd3db76c MW |
138 | def _fmt_path(path): |
139 | return ' -> '.join(["`%s'" % hop for hop in path]) | |
140 | ||
141 | class AmbiguousOptionError (Exception): | |
142 | def __init__(me, key, patha, vala, pathb, valb): | |
143 | me.key = key | |
144 | me.patha, me.vala = patha, vala | |
145 | me.pathb, me.valb = pathb, valb | |
146 | def __str__(me): | |
147 | return "Ambiguous answer resolving key `%s': " \ | |
148 | "path %s yields `%s' but %s yields `%s'" % \ | |
149 | (me.key, _fmt_path(me.patha), me.vala, _fmt_path(me.pathb), me.valb) | |
150 | ||
151 | class InheritanceCycleError (Exception): | |
152 | def __init__(me, key, path): | |
153 | me.key = key | |
154 | me.path = path | |
155 | def __str__(me): | |
156 | return "Found a cycle %s looking up key `%s'" % \ | |
157 | (_fmt_path(me.path), me.key) | |
158 | ||
e3ec3a3a MW |
159 | class MissingSectionException (Exception): |
160 | def __init__(me, sec): | |
161 | me.key = key | |
162 | def __str__(me): | |
163 | return "Section `%s' not found" % (me.sec) | |
164 | ||
bd3db76c MW |
165 | class MissingKeyException (Exception): |
166 | def __init__(me, sec, key): | |
167 | me.sec = sec | |
168 | me.key = key | |
169 | def __str__(me): | |
170 | return "Key `%s' not found in section `%s'" % (me.key, me.sec) | |
171 | ||
e3ec3a3a MW |
172 | class ConfigSection (object): |
173 | """ | |
174 | A section in a configuration parser. | |
175 | ||
176 | This is where a lot of the nitty-gritty stuff actually happens. The | |
177 | `MyConfigParser' knows a lot about the internals of this class, which saves | |
178 | on building a complicated interface. | |
179 | """ | |
180 | ||
181 | def __init__(me, name, cp): | |
182 | """Initialize a new, empty section with a given NAME and parent CP.""" | |
183 | me.name = name | |
184 | me._itemmap = dict() | |
185 | me._cp = cp | |
186 | ||
187 | def _expand(me, string, resolvep): | |
188 | """ | |
189 | Expands $(...) and (optionally) $[...] placeholders in STRING. | |
190 | ||
191 | RESOLVEP is a boolean switch: do we bother to tax the resolver or not? | |
192 | This is turned off by MyConfigParser's resolve() method while it's | |
193 | collecting hostnames to be resolved. | |
194 | """ | |
195 | string = RX_REF.sub \ | |
196 | (lambda m: me.get(m.group(1), resolvep), string) | |
197 | if resolvep: | |
198 | string = RX_RESOLVE.sub(lambda m: me._cp._resolver.lookup(m.group(1)), | |
199 | string) | |
200 | return string | |
201 | ||
e3ec3a3a MW |
202 | def _get(me, key, map = None, path = None): |
203 | """ | |
204 | Low-level option-fetching method. | |
205 | ||
206 | Fetch the value for the named KEY in this section, or maybe (recursively) | |
207 | a section which it inherits from. | |
208 | ||
209 | Returns a pair VALUE, PATH. The value is not expanded; nor do we check | |
210 | for the special `name' key. The caller is expected to do these things. | |
211 | Returns None if no value could be found. | |
212 | """ | |
213 | ||
214 | ## If we weren't given a memoization map or path, then we'd better make | |
215 | ## one. | |
216 | if map is None: map = {} | |
217 | if path is None: path = [] | |
218 | ||
219 | ## Extend the path to cover us, but remember to remove us again when | |
220 | ## we've finished. If we need to pass the current path back upwards, | |
221 | ## then remember to take a copy. | |
222 | path.append(me.name) | |
223 | try: | |
224 | ||
225 | ## If we've been this way before on another pass through then return | |
226 | ## the value we found then. If we're still thinking about it then | |
227 | ## we've found a cycle. | |
228 | try: threadp, value = map[me.name] | |
229 | except KeyError: pass | |
230 | else: | |
231 | if threadp: raise InheritanceCycleError(key, path[:]) | |
232 | ||
233 | ## See whether the answer is ready waiting for us. | |
234 | try: v = me._itemmap[key] | |
235 | except KeyError: pass | |
236 | else: return v, path[:] | |
237 | ||
238 | ## No, apparently, not. Find out our list of parents. | |
239 | try: | |
240 | parents = [me._cp.section(p) for p in | |
241 | me._itemmap['@inherit'].replace(',', ' ').split()] | |
242 | except KeyError: | |
243 | parents = [] | |
244 | ||
245 | ## Initially we have no idea. | |
246 | value = None | |
247 | winner = None | |
248 | ||
249 | ## Go through our parents and ask them what they think. | |
250 | map[me.name] = True, None | |
251 | for p in parents: | |
252 | ||
253 | ## See whether we get an answer. If not, keep on going. | |
254 | v, pp = p._get(key, map, path) | |
255 | if v is None: continue | |
256 | ||
257 | ## If we got an answer, check that it matches any previous ones. | |
258 | if value is None: | |
259 | value = v | |
260 | winner = pp | |
261 | elif value != v: | |
262 | raise AmbiguousOptionError(key, winner, value, pp, v) | |
263 | ||
264 | ## That's the best we could manage. | |
265 | map[me.name] = False, value | |
266 | return value, winner | |
267 | ||
268 | finally: | |
269 | ## Remove us from the path again. | |
270 | path.pop() | |
271 | ||
272 | def get(me, key, resolvep = True): | |
273 | """ | |
274 | Retrieve the value of KEY from this section. | |
275 | """ | |
276 | ||
277 | ## Special handling for the `name' key. | |
278 | if key == 'name': | |
279 | value = me._itemmap.get('name', me.name) | |
280 | else: | |
281 | value, _ = me._get(key) | |
282 | if value is None: | |
283 | raise MissingKeyException(me.name, key) | |
284 | ||
285 | ## Expand the value and return it. | |
286 | return me._expand(value, resolvep) | |
287 | ||
288 | def items(me, resolvep = True): | |
289 | """ | |
290 | Return a list of (NAME, VALUE) items in this section. | |
291 | ||
292 | This extends the default method by handling the inheritance chain. | |
293 | """ | |
294 | ||
295 | ## Initialize for a depth-first walk of the inheritance graph. | |
296 | d = {} | |
297 | visited = {} | |
298 | stack = [me.name] | |
299 | ||
300 | ## Visit nodes, collecting their keys. Don't believe the values: | |
301 | ## resolving inheritance is too hard to do like this. | |
302 | while stack: | |
303 | sec = me._cp.section(stack.pop()) | |
304 | if sec.name in visited: continue | |
305 | visited[sec.name] = True | |
306 | ||
307 | for key, value in sec._itemmap.iteritems(): | |
308 | if key == '@inherit': stack += value.replace(',', ' ').split() | |
309 | else: d[key] = None | |
310 | ||
311 | ## Now collect the values for the known keys, one by one. | |
312 | items = [] | |
313 | for key in d: items.append((key, me.get(key, resolvep))) | |
314 | ||
315 | ## And we're done. | |
316 | return items | |
317 | ||
b7e5aa06 | 318 | class MyConfigParser (object): |
6005ef9b MW |
319 | """ |
320 | A more advanced configuration parser. | |
321 | ||
b7e5aa06 | 322 | This has four major enhancements over the standard ConfigParser which are |
6005ef9b MW |
323 | relevant to us. |
324 | ||
325 | * It recognizes `@inherits' keys and follows them when expanding a | |
326 | value. | |
327 | ||
328 | * It recognizes `$(VAR)' references to configuration variables during | |
329 | expansion and processes them correctly. | |
330 | ||
331 | * It recognizes `$[HOST]' name-resolver requests and handles them | |
332 | correctly. | |
333 | ||
b7e5aa06 MW |
334 | * Its parsing behaviour is well-defined. |
335 | ||
6005ef9b MW |
336 | Use: |
337 | ||
b7e5aa06 | 338 | 1. Call parse(FILENAME) to slurp in the configuration data. |
6005ef9b MW |
339 | |
340 | 2. Call resolve() to collect the hostnames which need to be resolved and | |
341 | actually do the name resolution. | |
342 | ||
e3ec3a3a MW |
343 | 3. Call sections() to get a list of the configuration sections, or |
344 | section(NAME) to find a named section. | |
345 | ||
346 | 4. Call get(ITEM) on a section to collect the results, or items() to | |
6005ef9b MW |
347 | iterate over them. |
348 | """ | |
349 | ||
350 | def __init__(me): | |
351 | """ | |
352 | Initialize a new, empty configuration parser. | |
353 | """ | |
b7e5aa06 | 354 | me._sectmap = dict() |
6005ef9b MW |
355 | me._resolver = BulkResolver() |
356 | ||
b7e5aa06 MW |
357 | def parse(me, f): |
358 | """ | |
359 | Parse configuration from a file F. | |
360 | """ | |
361 | ||
362 | ## Initial parser state. | |
363 | sect = None | |
364 | key = None | |
365 | val = None | |
366 | lno = 0 | |
367 | ||
368 | ## An unpleasant hack. Python makes it hard to capture a value in a | |
369 | ## variable and examine it in a single action, and this is the best that | |
370 | ## I came up with. | |
371 | m = [None] | |
372 | def match(rx): m[0] = rx.match(line); return m[0] | |
373 | ||
374 | ## Commit a key's value when we've determined that there are no further | |
375 | ## continuation lines. | |
376 | def flush(): | |
e3ec3a3a | 377 | if key is not None: sect._itemmap[key] = val.getvalue() |
b7e5aa06 MW |
378 | |
379 | ## Work through all of the input lines. | |
380 | for line in f: | |
381 | lno += 1 | |
382 | ||
383 | if match(RX_COMMENT): | |
384 | ## A comment or a blank line. Nothing doing. (This means that we | |
385 | ## leave out blank lines which look like they might be continuation | |
386 | ## lines.) | |
387 | ||
388 | pass | |
389 | ||
390 | elif match(RX_GRPHDR): | |
391 | ## A section header. Flush out any previous value and set up the new | |
392 | ## group. | |
393 | ||
394 | flush() | |
395 | name = m[0].group(1) | |
396 | try: sect = me._sectmap[name] | |
e3ec3a3a | 397 | except KeyError: sect = me._sectmap[name] = ConfigSection(name, me) |
b7e5aa06 MW |
398 | key = None |
399 | ||
400 | elif match(RX_ASSGN): | |
401 | ## A new assignment. Flush out the old one, and set up to store this | |
402 | ## one. | |
403 | ||
404 | if sect is None: | |
405 | raise ConfigSyntaxError(f.name, lno, 'no active section to update') | |
406 | flush() | |
407 | key = m[0].group(1) | |
408 | val = StringIO(); val.write(m[0].group(2)) | |
409 | ||
410 | elif match(RX_CONT): | |
411 | ## A continuation line. Accumulate the value. | |
412 | ||
413 | if key is None: | |
414 | raise ConfigSyntaxError(f.name, lno, 'no config value to continue') | |
415 | val.write('\n'); val.write(m[0].group(1)) | |
416 | ||
417 | else: | |
418 | ## Something else. | |
419 | ||
420 | raise ConfigSyntaxError(f.name, lno, 'incomprehensible line') | |
421 | ||
422 | ## Don't forget to commit any final value material. | |
423 | flush() | |
424 | ||
e3ec3a3a MW |
425 | def section(me, name): |
426 | """Return a ConfigSection with the given NAME.""" | |
427 | try: return me._sectmap[name] | |
428 | except KeyError: raise MissingSectionException(name) | |
429 | ||
b7e5aa06 | 430 | def sections(me): |
e3ec3a3a MW |
431 | """Yield the known sections.""" |
432 | return me._sectmap.itervalues() | |
b7e5aa06 | 433 | |
6005ef9b MW |
434 | def resolve(me): |
435 | """ | |
436 | Works out all of the hostnames which need resolving and resolves them. | |
437 | ||
438 | Until you call this, attempts to fetch configuration items which need to | |
439 | resolve hostnames will fail! | |
440 | """ | |
e3ec3a3a MW |
441 | for sec in me.sections(): |
442 | for key, value in sec.items(resolvep = False): | |
2d51bc9f | 443 | for match in RX_RESOLVE.finditer(value): |
6005ef9b MW |
444 | me._resolver.prepare(match.group(1)) |
445 | me._resolver.run() | |
446 | ||
6005ef9b MW |
447 | ###-------------------------------------------------------------------------- |
448 | ### Command-line handling. | |
449 | ||
450 | def inputiter(things): | |
451 | """ | |
452 | Iterate over command-line arguments, returning corresponding open files. | |
453 | ||
454 | If none were given, or one is `-', assume standard input; if one is a | |
455 | directory, scan it for files other than backups; otherwise return the | |
456 | opened files. | |
457 | """ | |
458 | ||
459 | if not things: | |
460 | if OS.isatty(stdin.fileno()): | |
461 | M.die('no input given, and stdin is a terminal') | |
462 | yield stdin | |
463 | else: | |
464 | for thing in things: | |
465 | if thing == '-': | |
466 | yield stdin | |
467 | elif OS.path.isdir(thing): | |
468 | for item in OS.listdir(thing): | |
469 | if item.endswith('~') or item.endswith('#'): | |
470 | continue | |
471 | name = OS.path.join(thing, item) | |
472 | if not OS.path.isfile(name): | |
473 | continue | |
474 | yield file(name) | |
475 | else: | |
476 | yield file(thing) | |
477 | ||
478 | def parse_options(argv = argv): | |
479 | """ | |
480 | Parse command-line options, returning a pair (OPTS, ARGS). | |
481 | """ | |
482 | M.ego(argv[0]) | |
483 | op = OptionParser(usage = '%prog [-c CDB] INPUT...', | |
484 | version = '%%prog (tripe, version %s)' % VERSION) | |
485 | op.add_option('-c', '--cdb', metavar = 'CDB', | |
486 | dest = 'cdbfile', default = None, | |
487 | help = 'Compile output into a CDB file.') | |
488 | opts, args = op.parse_args(argv) | |
489 | return opts, args | |
490 | ||
491 | ###-------------------------------------------------------------------------- | |
492 | ### Main code. | |
493 | ||
494 | def getconf(args): | |
495 | """ | |
496 | Read the configuration files and return the accumulated result. | |
497 | ||
498 | We make sure that all hostnames have been properly resolved. | |
499 | """ | |
500 | conf = MyConfigParser() | |
501 | for f in inputiter(args): | |
b7e5aa06 | 502 | conf.parse(f) |
6005ef9b MW |
503 | conf.resolve() |
504 | return conf | |
505 | ||
506 | def output(conf, cdb): | |
507 | """ | |
508 | Output the configuration information CONF to the database CDB. | |
509 | ||
510 | This is where the special `user' and `auto' database entries get set. | |
511 | """ | |
512 | auto = [] | |
e3ec3a3a MW |
513 | for sec in sorted(conf.sections(), key = lambda sec: sec.name): |
514 | if sec.name.startswith('@'): | |
6005ef9b | 515 | continue |
e3ec3a3a MW |
516 | elif sec.name.startswith('$'): |
517 | label = sec.name | |
6005ef9b | 518 | else: |
e3ec3a3a | 519 | label = 'P%s' % sec.name |
fd1ba90c MW |
520 | try: a = sec.get('auto') |
521 | except MissingKeyException: pass | |
522 | else: | |
523 | if a in ('y', 'yes', 't', 'true', '1', 'on'): auto.append(sec.name) | |
524 | try: u = sec.get('user') | |
525 | except MissingKeyException: pass | |
526 | else: cdb.add('U%s' % u) | |
6005ef9b | 527 | url = M.URLEncode(laxp = True, semip = True) |
e3ec3a3a | 528 | for key, value in sorted(sec.items(), key = lambda (k, v): k): |
6005ef9b MW |
529 | if not key.startswith('@'): |
530 | url.encode(key, ' '.join(M.split(value)[0])) | |
531 | cdb.add(label, url.result) | |
532 | cdb.add('%AUTO', ' '.join(auto)) | |
533 | cdb.finish() | |
534 | ||
535 | def main(): | |
536 | """Main program.""" | |
537 | opts, args = parse_options() | |
538 | if opts.cdbfile: | |
539 | cdb = CDB.cdbmake(opts.cdbfile, opts.cdbfile + '.new') | |
540 | else: | |
541 | cdb = CDBFake() | |
542 | conf = getconf(args[1:]) | |
543 | output(conf, cdb) | |
544 | ||
545 | if __name__ == '__main__': | |
546 | main() | |
547 | ||
548 | ###----- That's all, folks -------------------------------------------------- |