]> gitweb.michael.orlitzky.com - djbdns-logparse.git/blob - bin/djbdns-logparse.py
8f30e1430cbfa1e8c6f9af11e398f5159d3ebd75
[djbdns-logparse.git] / bin / djbdns-logparse.py
1 #!/usr/bin/python3
2 """
3 Convert tinydns and dnscache logs to human-readable form
4 """
5
6 import re, typing
7 from struct import pack
8
9 ## Regular expressions for matching tinydns/dnscache log lines. We
10 ## compile these once here rather than within the corresponding
11 ## matching functions, because the latter get executed repeatedly.
12
13 # This first pattern is used to match the timestamp format that the
14 # tai64nlocal program produces. It appears in both dnscache and
15 # tinydns lines, after they've been piped through tai64nlocal, of
16 # course.
17 timestamp_pat = r'[\d-]+ [\d:\.]+'
18
19 # The regex to match dnscache log lines.
20 dnscache_log_re = re.compile(fr'({timestamp_pat}) (\w+)(.*)')
21
22 # The "hex4" pattern matches a string of four hexadecimal digits. This
23 # is used, for example, by tinydns to encode the query type
24 # identifier.
25 hex4_pat = r'[0-9a-f]{4}'
26
27 # The IP pattern matches a string of either 8 or 32 hexadecimal
28 # characters, which correspond to IPv4 and IPv6 addresses,
29 # respectively, in tinydns logs.
30 ip_pat = r'[0-9a-f]{8,32}'
31
32 # The regex to match tinydns log lines.
33 tinydns_log_re = re.compile(
34 rf'({timestamp_pat}) ({ip_pat}):({hex4_pat}):({hex4_pat}) ([\+\-IC/]) ({hex4_pat}) (.*)'
35 )
36
37 # A dictionary mapping query type identifiers, in decimal, to their
38 # friendly names for tinydns. Reference:
39 #
40 # https://en.wikipedia.org/wiki/List_of_DNS_record_types
41 #
42 # Note that mapping here is non-exhaustive, and that tinydns will
43 # log responses for record types that it does not know about.
44 query_type = {
45 1: "a",
46 2: "ns",
47 5: "cname",
48 6: "soa",
49 12: "ptr",
50 13: "hinfo",
51 15: "mx",
52 16: "txt",
53 17: "rp",
54 24: "sig",
55 25: "key",
56 28: "aaaa",
57 33: "srv",
58 35: "naptr",
59 38: "a6",
60 48: "dnskey",
61 52: "tlsa",
62 65: "https",
63 252: "axfr",
64 255: "any",
65 257: "caa"
66 }
67
68 # tinydns can drop a query for one of three reasons; this dictionary
69 # maps the symbol that gets logged in each case to a human-readable
70 # reason. We include the "+" case here, indicating that the query was
71 # NOT dropped, to avoid a special case later on when we're formatting
72 # the human-readable output.
73 query_drop_reason = {
74 "+": None,
75 "-": "no authority",
76 "I": "invalid query",
77 "C": "invalid class",
78 "/": "couldn't parse"
79 }
80
81
82 def convert_ip(ip : str) -> str:
83 """
84 Convert a hex string representing an IP address to
85 human-readable form.
86
87 Parameters
88 ----------
89
90 ip : str
91 The hexadecimal representation of either an IPv4 or an IPv6
92 address.
93
94 Returns
95 -------
96
97 The usual decimal dotted-quad representation is returned for an
98 IPv4 address. IPv6 addresses are returned almost as-is, but with
99 colons inserted in the appropriate places, between every four
100 characters.
101
102 Examples
103 --------
104
105 >>> convert_ip("7f000001")
106 '127.0.0.1'
107 >>> convert_ip("00000000000000000000ffff7f000001")
108 '0000:0000:0000:0000:0000:ffff:7f00:0001'
109 """
110 if len(ip) == 8:
111 # IPv4, eg. "7f000001" -> "7f 00 00 01" -> "127.0.0.1"
112 return ".".join(map(str, pack(">L", int(ip, 16))))
113 elif len(ip) == 32:
114 # IPv6 is actually simpler -- it's just a string-slicing operation.
115 return ":".join([ip[(4*i) : (4*i+4)] for i in range(8)])
116
117
118 def decode_client(words : list, i : int):
119 r"""
120 Helper function to decode the client field in a dnscache log
121 entry.
122
123 There are two possible formats for the client field,
124
125 1. clientip:clientport, used by tcpopen/tcpclose entries,
126 2. clientip:clientport:id, used by "query" entries.
127
128 Parameters
129 ----------
130
131 words : list
132 The ``words`` list (a list of fields) from
133 :func:`handle_dnscache_log`.
134
135 i : int
136 The index of the client field within ``words``
137
138 Returns
139 -------
140
141 Nothing; the ``i``th entry in the ``words`` list is modified
142 in-place.
143
144 Examples
145 --------
146
147 >>> words = ["foo", "bar", "7f000001:9253", "quux"]
148 >>> decode_client(words, 2)
149 >>> words
150 ['foo', 'bar', '127.0.0.1:37459', 'quux']
151
152 >>> words = ["foo", "7f000001:a3db:4fb9", "bar", "quux"]
153 >>> decode_client(words, 1)
154 >>> words
155 ['foo', '127.0.0.1:41947 (id 20409)', 'bar', 'quux']
156
157 """
158 chunks = words[i].split(":")
159
160 ip = convert_ip(chunks[0])
161 port = int(chunks[1], 16)
162 words[i] = f"{ip}:{port}"
163
164 if len(chunks) == 3:
165 # For a "query" entry's clientip:clientport:id field.
166 id = int(chunks[2], 16)
167 words[i] += f" (id {id})"
168
169 def decode_ip(words, i):
170 words[i] = convert_ip(words[i])
171
172 def decode_ttl(words, i):
173 words[i] = f"TTL={words[i]}"
174
175 def decode_serial(words, i):
176 serial = int(words[i])
177 words[i] = f"#{serial}"
178
179 def decode_type(words, i):
180 qt = words[i]
181 words[i] = query_type.get(int(qt), qt)
182
183 def handle_dnscache_log(line) -> typing.Optional[str]:
184 """
185 Handle a single log line if it matches the ``dnscache_log_re`` regex.
186
187 Parameters
188 ----------
189
190 line : string
191 The log line that might match ``dnscache_log_re``.
192
193 Returns
194 -------
195
196 Either the human-readable string if the log line was handled (that
197 is, if it was really a dnscache log line), or ``None`` if it was
198 not.
199
200 Examples
201 --------
202
203 >>> line = "2022-09-15 18:37:33.863805500 query 1 7f000001:a3db:4fb9 1 www.example.com."
204 >>> handle_dnscache_log(line)
205 '2022-09-15 18:37:33.863805500 query #1 127.0.0.1:41947 (id 20409) a www.example.com.'
206
207 >>> line = "2022-09-15 18:37:33.863874500 tx 0 1 www.example.com. . c0a80101"
208 >>> handle_dnscache_log(line)
209 '2022-09-15 18:37:33.863874500 tx g=0 a www.example.com. . 192.168.1.1'
210
211 >>> line = "2022-09-15 18:37:33.878529500 rr c0a80101 20865 1 www.example.com. 5db8d822"
212 >>> handle_dnscache_log(line)
213 '2022-09-15 18:37:33.878529500 rr 192.168.1.1 TTL=20865 a www.example.com. 93.184.216.34'
214
215 >>> line = "2022-09-15 18:37:33.878532500 stats 1 43 1 0"
216 >>> handle_dnscache_log(line)
217 '2022-09-15 18:37:33.878532500 stats count=1 motion=43 udp-active=1 tcp-active=0'
218
219 >>> line = "2022-09-15 18:37:33.878602500 sent 1 49"
220 >>> handle_dnscache_log(line)
221 '2022-09-15 18:37:33.878602500 sent #1 49'
222
223 >>> line = "this line is nonsense"
224 >>> handle_dnscache_log(line)
225
226 """
227 match = dnscache_log_re.match(line)
228 if not match:
229 return None
230
231 (timestamp, event, data) = match.groups()
232
233 words = data.split()
234 if event == "cached":
235 if words[0] not in ("cname", "ns", "nxdomain"):
236 decode_type(words, 0)
237
238 elif event == "drop":
239 decode_serial(words, 0)
240
241 elif event == "lame":
242 decode_ip(words, 0)
243
244 elif event == "nodata":
245 decode_ip(words, 0)
246 decode_ttl(words, 1)
247 decode_type(words, 2)
248
249 elif event == "nxdomain":
250 decode_ip(words, 0)
251 decode_ttl(words, 1)
252
253 elif event == "query":
254 decode_serial(words, 0)
255 decode_client(words, 1)
256 decode_type(words, 2)
257
258 elif event == "rr":
259 decode_ip(words, 0)
260 decode_ttl(words, 1)
261 if words[2] not in ("cname", "mx", "ns", "ptr", "soa"):
262 decode_type(words, 2)
263 if words[2] == "a": # decode answer to an A query
264 decode_ip(words, 4)
265 if words[2] == "txt": # text record
266 response = words[4]
267 if response.endswith("..."):
268 ellipsis = "..."
269 response = response[0:-3]
270 else:
271 ellipsis = ""
272 length = int(response[0:2], 16)
273 chars = []
274 for i in range(1, len(response)//2):
275 chars.append(chr(int(response[2*i : (2*i)+2], 16)))
276 txt = "".join(chars)
277 words[4] = f"{length}:\"{txt}{ellipsis}\""
278
279 elif event == "sent":
280 decode_serial(words, 0)
281
282 elif event == "stats":
283 words[0] = f"count={words[0]}"
284 words[1] = f"motion={words[1]}"
285 words[2] = f"udp-active={words[2]}"
286 words[3] = f"tcp-active={words[3]}"
287
288 elif event == "tx":
289 words[0] = f"g={words[0]}"
290 decode_type(words, 1)
291 # words[2] = name
292 # words[3] = control (domain for which these servers are believed
293 # to be authoritative)
294 for i in range(4, len(words)):
295 decode_ip(words, i)
296
297 elif event in ("tcpopen", "tcpclose"):
298 decode_client(words, 0)
299
300 # Reconstitute "data" (i.e. everything after the timestamp and the
301 # event) from "words", which was originally obtained by splitting
302 # "data".
303 data = " ".join(words)
304 return f"{timestamp} {event} {data}"
305
306
307
308 def handle_tinydns_log(line : str) -> typing.Optional[str]:
309 """
310 Handle a single log line if it matches the ``tinydns_log_re`` regex.
311
312 Parameters
313 ----------
314
315 line : string
316 The log line that might match ``tinydns_log_re``.
317
318 Returns
319 -------
320
321 Either the human-readable string if the log line was handled (that
322 is, if it was really a tinydns log line), or ``None`` if it was
323 not.
324
325 Examples
326 --------
327
328 >>> line = "2022-09-14 21:04:40.206516500 7f000001:9d61:be69 - 0001 www.example.com"
329 >>> handle_tinydns_log(line)
330 '2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com'
331
332 >>> line = "this line is nonsense"
333 >>> handle_tinydns_log(line)
334
335 """
336 match = tinydns_log_re.match(line)
337 if not match:
338 return None
339
340 (timestamp, ip, port, id, code, type, name) = match.groups()
341 ip = convert_ip(ip)
342 port = int(port, 16)
343 id = int(id, 16)
344
345 # Convert the "type" field to a human-readable record type name
346 # using the query_type dictionary. If the right name isn't present
347 # in the dictionary, we use the (decimal) type id instead.
348 type = int(type, 16) # "001c" -> 28
349 type = query_type.get(type, type) # 28 -> "aaaa"
350
351 line_tpl = "{timestamp} "
352
353 reason = query_drop_reason[code]
354 if code == "+":
355 line_tpl += "sent response to {ip}:{port} (id {id}): {type} {name}"
356 else:
357 line_tpl += "dropped query ({reason}) from {ip}:{port}"
358 if code != "/":
359 # If the query can actually be parsed, the log line is a
360 # bit more informative than it would have been otherwise.
361 line_tpl += " (id {id}): {type} {name}"
362
363 return line_tpl.format(timestamp=timestamp,
364 reason=reason,
365 ip=ip,
366 port=port,
367 id=id,
368 type=type,
369 name=name)
370
371
372 def parse_logfile(file : typing.TextIO):
373 r"""
374 Process a single log ``file``.
375
376 Parameters
377 ----------
378
379 file : typing.TextIO
380 An open log file, or stdin.
381
382 Examples
383 --------
384
385 >>> line = "@4000000063227a320c4f3114 7f000001:9d61:be69 - 0001 www.example.com\n"
386 >>> from tempfile import NamedTemporaryFile
387 >>> with NamedTemporaryFile(mode="w", delete=False) as f:
388 ... _ = f.write(line)
389 >>> f = open(f.name, 'r')
390 >>> parse_logfile(f)
391 2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
392 >>> f.close()
393 >>> from os import remove
394 >>> remove(f.name)
395
396 """
397 # Open a pipe to tai64nlocal. We'll write lines of our input file
398 # (the log file) to it, and read back the same lines but with
399 # friendly timestamps in them.
400 from subprocess import Popen, PIPE
401 tai = Popen(["tai64nlocal"], stdin=PIPE, stdout=PIPE, text=True, bufsize=0)
402
403 for line in file:
404 tai.stdin.write(line)
405 line = tai.stdout.readline()
406
407 friendly_line = handle_tinydns_log(line)
408 if not friendly_line:
409 friendly_line = handle_dnscache_log(line)
410 if not friendly_line:
411 friendly_line = line
412
413 print(friendly_line)
414
415 def main():
416 r"""
417 The entry point to the program.
418
419 This function is responsible only for parsing any command-line
420 arguments, and then calling :func`parse_logfile` on them.
421 """
422 # Create an argument parser using the file's docsctring as its
423 # description.
424 from argparse import ArgumentParser, FileType
425 parser = ArgumentParser(description = __doc__)
426
427 # Parse zero or more positional arguments into a list of
428 # "logfiles". If none are given, read from stdin instead.
429 from sys import stdin
430 parser.add_argument("logfiles",
431 metavar="LOGFILE",
432 type=FileType("r"),
433 nargs="*",
434 default=[stdin],
435 help="djbdns logfile to process (default: stdin)")
436
437 # Warning: argparse automatically opens its file arguments here,
438 # and they only get closed when the program terminates. There's no
439 # real benefit to closing them one-at-a-time after calling
440 # parse_logfile(), because the "scarce" resource of open file
441 # descriptors gets consumed immediately, before any processing has
442 # happened. In other words, if you're going to run out of file
443 # descriptors, it's going to happen right now.
444 #
445 # So anyway, don't run this on several million logfiles.
446 args = parser.parse_args()
447 for f in args.logfiles:
448 parse_logfile(f)
449
450
451 if __name__ == "__main__":
452 main()