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