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):
  86     """
  87     Convert a hex string representing an IP address to conventional
  88     human-readable form, ie. dotted-quad decimal for IPv4, and
  89     8 colon-separated hex shorts for IPv6.
  90
  91     Examples
  92     --------
  93
  94         >>> convert_ip("7f000001")
  95         '127.0.0.1'
  96         >>> convert_ip("00000000000000000000ffff7f000001")
  97         '0000:0000:0000:0000:0000:ffff:7f00:0001'
  98
  99     """
 100     if len(ip) == 8:
 101         # IPv4, eg. "7f000001" -> "7f 00 00 01" -> "127.0.0.1"
 102         return "%d.%d.%d.%d" % tuple(pack(">L", int(ip, 16)))
 103     elif len(ip) == 32:
 104         # IPv6 is actually simpler -- it's just a string-slicing operation.
 105         return ":".join([ip[(4*i) : (4*i+4)] for i in range(8)])
 106
 107
 108 def decode_client(words, i):
 109     chunks = words[i].split(":")
 110     if len(chunks) == 2:                # ip:port
 111         words[i] = "%s:%d" % (convert_ip(chunks[0]), int(chunks[1], 16))
 112     elif len(chunks) == 3:
 113         words[i] = "%s:%d (id %d)" % (convert_ip(chunks[0]),
 114                                       int(chunks[1], 16),
 115                                       int(chunks[2], 16))
 116
 117 def decode_ip(words, i):
 118     words[i] = convert_ip(words[i])
 119
 120 def decode_ttl(words, i):
 121     words[i] = "TTL=%s" % words[i]
 122
 123 def decode_serial(words, i):
 124     serial = int(words[i])
 125     words[i] = "#%d" % serial
 126
 127 def decode_type(words, i):
 128     qt = words[i]
 129     words[i] = query_type.get(int(qt), qt)
 130
 131 def handle_dnscache_log(line) -> typing.Optional[str]:
 132     """
 133     Handle a single log line if it matches the ``dnscache_log_re`` regex.
 134
 135     Parameters
 136     ----------
 137
 138     line : string
 139         The log line that might match ``dnscache_log_re``.
 140
 141     Returns
 142     -------
 143
 144     Either the human-readable string if the log line was handled (that
 145     is, if it was really a dnscache log line), or ``None`` if it was
 146     not.
 147
 148     Examples
 149     --------
 150
 151         >>> line = "2022-09-15 18:37:33.863805500 query 1 7f000001:a3db:4fb9 1 www.example.com."
 152         >>> handle_dnscache_log(line)
 153         '2022-09-15 18:37:33.863805500 query #1 127.0.0.1:41947 (id 20409) a www.example.com.'
 154
 155         >>> line = "2022-09-15 18:37:33.863874500 tx 0 1 www.example.com. . c0a80101"
 156         >>> handle_dnscache_log(line)
 157         '2022-09-15 18:37:33.863874500 tx g=0 a www.example.com. . 192.168.1.1'
 158
 159         >>> line = "2022-09-15 18:37:33.878529500 rr c0a80101 20865 1 www.example.com. 5db8d822"
 160         >>> handle_dnscache_log(line)
 161         '2022-09-15 18:37:33.878529500 rr 192.168.1.1 TTL=20865 a www.example.com. 93.184.216.34'
 162
 163         >>> line = "2022-09-15 18:37:33.878532500 stats 1 43 1 0"
 164         >>> handle_dnscache_log(line)
 165         '2022-09-15 18:37:33.878532500 stats count=1 motion=43 udp-active=1 tcp-active=0'
 166
 167         >>> line = "2022-09-15 18:37:33.878602500 sent 1 49"
 168         >>> handle_dnscache_log(line)
 169         '2022-09-15 18:37:33.878602500 sent #1 49'
 170
 171         >>> line = "this line is nonsense"
 172         >>> handle_dnscache_log(line)
 173
 174     """
 175     match = dnscache_log_re.match(line)
 176     if not match:
 177         return None
 178
 179     (timestamp, event, data) = match.groups()
 180
 181     words = data.split()
 182     if event == "cached":
 183         if words[0] not in ("cname", "ns", "nxdomain"):
 184             decode_type(words, 0)
 185
 186     elif event == "drop":
 187         decode_serial(words, 0)
 188
 189     elif event == "lame":
 190         decode_ip(words, 0)
 191
 192     elif event == "nodata":
 193         decode_ip(words, 0)
 194         decode_ttl(words, 1)
 195         decode_type(words, 2)
 196
 197     elif event == "nxdomain":
 198         decode_ip(words, 0)
 199         decode_ttl(words, 1)
 200
 201     elif event == "query":
 202         decode_serial(words, 0)
 203         decode_client(words, 1)
 204         decode_type(words, 2)
 205
 206     elif event == "rr":
 207         decode_ip(words, 0)
 208         decode_ttl(words, 1)
 209         if words[2] not in ("cname", "mx", "ns", "ptr", "soa"):
 210             decode_type(words, 2)
 211             if words[2] == "a":         # decode answer to an A query
 212                 decode_ip(words, 4)
 213             if words[2] == "txt":       # text record
 214                 response = words[4]
 215                 if response.endswith("..."):
 216                     ellipsis = "..."
 217                     response = response[0:-3]
 218                 else:
 219                     ellipsis = ""
 220                 length = int(response[0:2], 16)
 221                 chars = []
 222                 for i in range(1, len(response)/2):
 223                     chars.append(chr(int(response[2*i : (2*i)+2], 16)))
 224                 words[4] = "%d:\"%s%s\"" % (length, "".join(chars), ellipsis)
 225
 226     elif event == "sent":
 227         decode_serial(words, 0)
 228
 229     elif event == "stats":
 230         words[0] = "count=%s" % words[0]
 231         words[1] = "motion=%s" % words[1]
 232         words[2] = "udp-active=%s" % words[2]
 233         words[3] = "tcp-active=%s" % words[3]
 234
 235     elif event == "tx":
 236         words[0] = "g=%s" % words[0]
 237         decode_type(words, 1)
 238         # words[2] = name
 239         # words[3] = control (domain for which these servers are believed
 240         #            to be authoritative)
 241         for i in range(4, len(words)):
 242             decode_ip(words, i)
 243
 244     elif event in ("tcpopen", "tcpclose"):
 245         decode_client(words, 0)
 246
 247     return f"{timestamp} {event} " + " ".join(words)
 248
 249
 250
 251 def handle_tinydns_log(line : str) -> typing.Optional[str]:
 252     """
 253     Handle a single log line if it matches the ``tinydns_log_re`` regex.
 254
 255     Parameters
 256     ----------
 257
 258     line : string
 259         The log line that might match ``tinydns_log_re``.
 260
 261     Returns
 262     -------
 263
 264     Either the human-readable string if the log line was handled (that
 265     is, if it was really a tinydns log line), or ``None`` if it was
 266     not.
 267
 268     Examples
 269     --------
 270
 271         >>> line = "2022-09-14 21:04:40.206516500 7f000001:9d61:be69 - 0001 www.example.com"
 272         >>> handle_tinydns_log(line)
 273         '2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com'
 274
 275         >>> line = "this line is nonsense"
 276         >>> handle_tinydns_log(line)
 277
 278     """
 279     match = tinydns_log_re.match(line)
 280     if not match:
 281         return None
 282
 283     (timestamp, ip, port, id, code, type, name) = match.groups()
 284     ip = convert_ip(ip)
 285     port = int(port, 16)
 286     id = int(id, 16)
 287
 288     # Convert the "type" field to a human-readable record type name
 289     # using the query_type dictionary. If the right name isn't present
 290     # in the dictionary, we use the (decimal) type id instead.
 291     type = int(type, 16)                # "001c" -> 28
 292     type = query_type.get(type, type)   # 28 -> "aaaa"
 293
 294     line_tpl = "{timestamp} "
 295
 296     reason = query_drop_reason[code]
 297     if code == "+":
 298         line_tpl += "sent response to {ip}:{port} (id {id}): {type} {name}"
 299     else:
 300         line_tpl += "dropped query ({reason}) from {ip}:{port}"
 301         if code != "/":
 302             # If the query can actually be parsed, the log line is a
 303             # bit more informative than it would have been otherwise.
 304             line_tpl += " (id {id}): {type} {name}"
 305
 306     return line_tpl.format(timestamp=timestamp,
 307                            reason=reason,
 308                            ip=ip,
 309                            port=port,
 310                            id=id,
 311                            type=type,
 312                            name=name)
 313
 314
 315 def parse_logfile(file : typing.TextIO):
 316     r"""
 317     Process a single log ``file``.
 318
 319     Parameters
 320     ----------
 321
 322     file : typing.TextIO
 323         An open log file, or stdin.
 324
 325     Examples
 326     --------
 327
 328         >>> line = "@4000000063227a320c4f3114 7f000001:9d61:be69 - 0001 www.example.com\n"
 329         >>> from tempfile import NamedTemporaryFile
 330         >>> with NamedTemporaryFile(mode="w", delete=False) as f:
 331         ...     _ = f.write(line)
 332         >>> f = open(f.name, 'r')
 333         >>> parse_logfile(f)
 334         2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
 335         >>> f.close()
 336         >>> from os import remove
 337         >>> remove(f.name)
 338
 339     """
 340     # Open pipe to tai64nlocal: we will write lines of our input (the
 341     # raw log file) to it, and read log lines with readable timestamps
 342     # from it.
 343     tai = Popen(["tai64nlocal"], stdin=PIPE, stdout=PIPE, text=True, bufsize=0)
 344
 345     for line in file:
 346         tai.stdin.write(line)
 347         line = tai.stdout.readline()
 348
 349         friendly_line = handle_tinydns_log(line)
 350         if not friendly_line:
 351             friendly_line = handle_dnscache_log(line)
 352             if not friendly_line:
 353                 friendly_line = line
 354
 355         print(friendly_line)
 356
 357 def main():
 358     r"""
 359     The entry point to the program.
 360
 361     This function is responsible only for parsing any command-line
 362     arguments, and then calling :func`parse_logfile` on them.
 363     """
 364     # Create an argument parser using the file's docsctring as its
 365     # description.
 366     from argparse import ArgumentParser, FileType
 367     parser = ArgumentParser(description = __doc__)
 368
 369     # Parse zero or more positional arguments into a list of
 370     # "logfiles". If none are given, read from stdin instead.
 371     from sys import stdin
 372     parser.add_argument("logfiles",
 373                         metavar="LOGFILE",
 374                         type=FileType("r"),
 375                         nargs="*",
 376                         default=[stdin],
 377                         help="djbdns logfile to process (default: stdin)")
 378
 379     # Warning: argparse automatically opens its file arguments here,
 380     # and they only get closed when the program terminates. There's no
 381     # real benefit to closing them one-at-a-time after calling
 382     # parse_logfile(), because the "scarce" resource of open file
 383     # descriptors gets consumed immediately, before any processing has
 384     # happened. In other words, if you're going to run out of file
 385     # descriptors, it's going to happen right now.
 386     #
 387     # So anyway, don't run this on several million logfiles.
 388     args = parser.parse_args()
 389     for f in args.logfiles:
 390         parse_logfile(f)
 391
 392
 393 if __name__ == "__main__":
 394     main()