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 ".".join(map(str, 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
 111     ip = convert_ip(chunks[0])
 112     port = int(chunks[1], 16)
 113     words[i] = f"{ip}:{port}"
 114
 115     if len(chunks) == 3:
 116         # For a "query" entry's clientip:clientport:id field.
 117         id = int(chunks[2], 16)
 118         words[i] += f" (id {id})"
 119
 120 def decode_ip(words, i):
 121     words[i] = convert_ip(words[i])
 122
 123 def decode_ttl(words, i):
 124     words[i] = f"TTL={words[i]}"
 125
 126 def decode_serial(words, i):
 127     serial = int(words[i])
 128     words[i] = f"#{serial}"
 129
 130 def decode_type(words, i):
 131     qt = words[i]
 132     words[i] = query_type.get(int(qt), qt)
 133
 134 def handle_dnscache_log(line) -> typing.Optional[str]:
 135     """
 136     Handle a single log line if it matches the ``dnscache_log_re`` regex.
 137
 138     Parameters
 139     ----------
 140
 141     line : string
 142         The log line that might match ``dnscache_log_re``.
 143
 144     Returns
 145     -------
 146
 147     Either the human-readable string if the log line was handled (that
 148     is, if it was really a dnscache log line), or ``None`` if it was
 149     not.
 150
 151     Examples
 152     --------
 153
 154         >>> line = "2022-09-15 18:37:33.863805500 query 1 7f000001:a3db:4fb9 1 www.example.com."
 155         >>> handle_dnscache_log(line)
 156         '2022-09-15 18:37:33.863805500 query #1 127.0.0.1:41947 (id 20409) a www.example.com.'
 157
 158         >>> line = "2022-09-15 18:37:33.863874500 tx 0 1 www.example.com. . c0a80101"
 159         >>> handle_dnscache_log(line)
 160         '2022-09-15 18:37:33.863874500 tx g=0 a www.example.com. . 192.168.1.1'
 161
 162         >>> line = "2022-09-15 18:37:33.878529500 rr c0a80101 20865 1 www.example.com. 5db8d822"
 163         >>> handle_dnscache_log(line)
 164         '2022-09-15 18:37:33.878529500 rr 192.168.1.1 TTL=20865 a www.example.com. 93.184.216.34'
 165
 166         >>> line = "2022-09-15 18:37:33.878532500 stats 1 43 1 0"
 167         >>> handle_dnscache_log(line)
 168         '2022-09-15 18:37:33.878532500 stats count=1 motion=43 udp-active=1 tcp-active=0'
 169
 170         >>> line = "2022-09-15 18:37:33.878602500 sent 1 49"
 171         >>> handle_dnscache_log(line)
 172         '2022-09-15 18:37:33.878602500 sent #1 49'
 173
 174         >>> line = "this line is nonsense"
 175         >>> handle_dnscache_log(line)
 176
 177     """
 178     match = dnscache_log_re.match(line)
 179     if not match:
 180         return None
 181
 182     (timestamp, event, data) = match.groups()
 183
 184     words = data.split()
 185     if event == "cached":
 186         if words[0] not in ("cname", "ns", "nxdomain"):
 187             decode_type(words, 0)
 188
 189     elif event == "drop":
 190         decode_serial(words, 0)
 191
 192     elif event == "lame":
 193         decode_ip(words, 0)
 194
 195     elif event == "nodata":
 196         decode_ip(words, 0)
 197         decode_ttl(words, 1)
 198         decode_type(words, 2)
 199
 200     elif event == "nxdomain":
 201         decode_ip(words, 0)
 202         decode_ttl(words, 1)
 203
 204     elif event == "query":
 205         decode_serial(words, 0)
 206         decode_client(words, 1)
 207         decode_type(words, 2)
 208
 209     elif event == "rr":
 210         decode_ip(words, 0)
 211         decode_ttl(words, 1)
 212         if words[2] not in ("cname", "mx", "ns", "ptr", "soa"):
 213             decode_type(words, 2)
 214             if words[2] == "a":         # decode answer to an A query
 215                 decode_ip(words, 4)
 216             if words[2] == "txt":       # text record
 217                 response = words[4]
 218                 if response.endswith("..."):
 219                     ellipsis = "..."
 220                     response = response[0:-3]
 221                 else:
 222                     ellipsis = ""
 223                 length = int(response[0:2], 16)
 224                 chars = []
 225                 for i in range(1, len(response)//2):
 226                     chars.append(chr(int(response[2*i : (2*i)+2], 16)))
 227                 txt = "".join(chars)
 228                 words[4] = f"{length}:\"{txt}{ellipsis}\""
 229
 230     elif event == "sent":
 231         decode_serial(words, 0)
 232
 233     elif event == "stats":
 234         words[0] = f"count={words[0]}"
 235         words[1] = f"motion={words[1]}"
 236         words[2] = f"udp-active={words[2]}"
 237         words[3] = f"tcp-active={words[3]}"
 238
 239     elif event == "tx":
 240         words[0] = f"g={words[0]}"
 241         decode_type(words, 1)
 242         # words[2] = name
 243         # words[3] = control (domain for which these servers are believed
 244         #            to be authoritative)
 245         for i in range(4, len(words)):
 246             decode_ip(words, i)
 247
 248     elif event in ("tcpopen", "tcpclose"):
 249         decode_client(words, 0)
 250
 251     return f"{timestamp} {event} " + " ".join(words)
 252
 253
 254
 255 def handle_tinydns_log(line : str) -> typing.Optional[str]:
 256     """
 257     Handle a single log line if it matches the ``tinydns_log_re`` regex.
 258
 259     Parameters
 260     ----------
 261
 262     line : string
 263         The log line that might match ``tinydns_log_re``.
 264
 265     Returns
 266     -------
 267
 268     Either the human-readable string if the log line was handled (that
 269     is, if it was really a tinydns log line), or ``None`` if it was
 270     not.
 271
 272     Examples
 273     --------
 274
 275         >>> line = "2022-09-14 21:04:40.206516500 7f000001:9d61:be69 - 0001 www.example.com"
 276         >>> handle_tinydns_log(line)
 277         '2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com'
 278
 279         >>> line = "this line is nonsense"
 280         >>> handle_tinydns_log(line)
 281
 282     """
 283     match = tinydns_log_re.match(line)
 284     if not match:
 285         return None
 286
 287     (timestamp, ip, port, id, code, type, name) = match.groups()
 288     ip = convert_ip(ip)
 289     port = int(port, 16)
 290     id = int(id, 16)
 291
 292     # Convert the "type" field to a human-readable record type name
 293     # using the query_type dictionary. If the right name isn't present
 294     # in the dictionary, we use the (decimal) type id instead.
 295     type = int(type, 16)                # "001c" -> 28
 296     type = query_type.get(type, type)   # 28 -> "aaaa"
 297
 298     line_tpl = "{timestamp} "
 299
 300     reason = query_drop_reason[code]
 301     if code == "+":
 302         line_tpl += "sent response to {ip}:{port} (id {id}): {type} {name}"
 303     else:
 304         line_tpl += "dropped query ({reason}) from {ip}:{port}"
 305         if code != "/":
 306             # If the query can actually be parsed, the log line is a
 307             # bit more informative than it would have been otherwise.
 308             line_tpl += " (id {id}): {type} {name}"
 309
 310     return line_tpl.format(timestamp=timestamp,
 311                            reason=reason,
 312                            ip=ip,
 313                            port=port,
 314                            id=id,
 315                            type=type,
 316                            name=name)
 317
 318
 319 def parse_logfile(file : typing.TextIO):
 320     r"""
 321     Process a single log ``file``.
 322
 323     Parameters
 324     ----------
 325
 326     file : typing.TextIO
 327         An open log file, or stdin.
 328
 329     Examples
 330     --------
 331
 332         >>> line = "@4000000063227a320c4f3114 7f000001:9d61:be69 - 0001 www.example.com\n"
 333         >>> from tempfile import NamedTemporaryFile
 334         >>> with NamedTemporaryFile(mode="w", delete=False) as f:
 335         ...     _ = f.write(line)
 336         >>> f = open(f.name, 'r')
 337         >>> parse_logfile(f)
 338         2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
 339         >>> f.close()
 340         >>> from os import remove
 341         >>> remove(f.name)
 342
 343     """
 344     # Open pipe to tai64nlocal: we will write lines of our input (the
 345     # raw log file) to it, and read log lines with readable timestamps
 346     # from it.
 347     tai = Popen(["tai64nlocal"], stdin=PIPE, stdout=PIPE, text=True, bufsize=0)
 348
 349     for line in file:
 350         tai.stdin.write(line)
 351         line = tai.stdout.readline()
 352
 353         friendly_line = handle_tinydns_log(line)
 354         if not friendly_line:
 355             friendly_line = handle_dnscache_log(line)
 356             if not friendly_line:
 357                 friendly_line = line
 358
 359         print(friendly_line)
 360
 361 def main():
 362     r"""
 363     The entry point to the program.
 364
 365     This function is responsible only for parsing any command-line
 366     arguments, and then calling :func`parse_logfile` on them.
 367     """
 368     # Create an argument parser using the file's docsctring as its
 369     # description.
 370     from argparse import ArgumentParser, FileType
 371     parser = ArgumentParser(description = __doc__)
 372
 373     # Parse zero or more positional arguments into a list of
 374     # "logfiles". If none are given, read from stdin instead.
 375     from sys import stdin
 376     parser.add_argument("logfiles",
 377                         metavar="LOGFILE",
 378                         type=FileType("r"),
 379                         nargs="*",
 380                         default=[stdin],
 381                         help="djbdns logfile to process (default: stdin)")
 382
 383     # Warning: argparse automatically opens its file arguments here,
 384     # and they only get closed when the program terminates. There's no
 385     # real benefit to closing them one-at-a-time after calling
 386     # parse_logfile(), because the "scarce" resource of open file
 387     # descriptors gets consumed immediately, before any processing has
 388     # happened. In other words, if you're going to run out of file
 389     # descriptors, it's going to happen right now.
 390     #
 391     # So anyway, don't run this on several million logfiles.
 392     args = parser.parse_args()
 393     for f in args.logfiles:
 394         parse_logfile(f)
 395
 396
 397 if __name__ == "__main__":
 398     main()