]>
gitweb.michael.orlitzky.com - djbdns-logparse.git/blob - bin/djbdns-logparse.py
3 Convert tinydns and dnscache logs to human-readable form
7 from struct
import pack
8 from subprocess
import Popen
, PIPE
9 from time
import strftime
, gmtime
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.
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
20 timestamp_pat
= r
'[\d-]+ [\d:\.]+'
22 # The regex to match dnscache log lines.
23 dnscache_log_re
= re
.compile(fr
'({timestamp_pat}) (\w+)(.*)')
25 # The "hex4" pattern matches a string of four hexadecimal digits. This
26 # is used, for example, by tinydns to encode the query type
28 hex4_pat
= r
'[0-9a-f]{4}'
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}'
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}) (.*)'
40 # A dictionary mapping query type identifiers, in decimal, to their
41 # friendly names for tinydns. Reference:
43 # https://en.wikipedia.org/wiki/List_of_DNS_record_types
45 # Note that mapping here is non-exhaustive, and that tinydns will
46 # log responses for record types that it does not know about.
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.
85 def convert_ip(ip
: str) -> str:
87 Convert a hex string representing an IP address to
94 The hexadecimal representation of either an IPv4 or an IPv6
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
108 >>> convert_ip("7f000001")
110 >>> convert_ip("00000000000000000000ffff7f000001")
111 '0000:0000:0000:0000:0000:ffff:7f00:0001'
114 # IPv4, eg. "7f000001" -> "7f 00 00 01" -> "127.0.0.1"
115 return ".".join(map(str, pack(">L", int(ip
, 16))))
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)])
121 def decode_client(words
: list, i
: int):
123 Helper function to decode the client field in a dnscache log
126 There are two possible formats for the client field,
128 1. clientip:clientport, used by tcpopen/tcpclose entries,
129 2. clientip:clientport:id, used by "query" entries.
135 The ``words`` list (a list of fields) from
136 :func:`handle_dnscache_log`.
139 The index of the client field within ``words``
144 Nothing; the ``i``th entry in the ``words`` list is modified
150 >>> words = ["foo", "bar", "7f000001:9253", "quux"]
151 >>> decode_client(words, 2)
153 ['foo', 'bar', '127.0.0.1:37459', 'quux']
155 >>> words = ["foo", "7f000001:a3db:4fb9", "bar", "quux"]
156 >>> decode_client(words, 1)
158 ['foo', '127.0.0.1:41947 (id 20409)', 'bar', 'quux']
161 chunks
= words
[i
].split(":")
163 ip
= convert_ip(chunks
[0])
164 port
= int(chunks
[1], 16)
165 words
[i
] = f
"{ip}:{port}"
168 # For a "query" entry's clientip:clientport:id field.
169 id = int(chunks
[2], 16)
170 words
[i
] += f
" (id {id})"
172 def decode_ip(words
, i
):
173 words
[i
] = convert_ip(words
[i
])
175 def decode_ttl(words
, i
):
176 words
[i
] = f
"TTL={words[i]}"
178 def decode_serial(words
, i
):
179 serial
= int(words
[i
])
180 words
[i
] = f
"#{serial}"
182 def decode_type(words
, i
):
184 words
[i
] = query_type
.get(int(qt
), qt
)
186 def handle_dnscache_log(line
) -> typing
.Optional
[str]:
188 Handle a single log line if it matches the ``dnscache_log_re`` regex.
194 The log line that might match ``dnscache_log_re``.
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
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.'
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'
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'
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'
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'
226 >>> line = "this line is nonsense"
227 >>> handle_dnscache_log(line)
230 match
= dnscache_log_re
.match(line
)
234 (timestamp
, event
, data
) = match
.groups()
237 if event
== "cached":
238 if words
[0] not in ("cname", "ns", "nxdomain"):
239 decode_type(words
, 0)
241 elif event
== "drop":
242 decode_serial(words
, 0)
244 elif event
== "lame":
247 elif event
== "nodata":
250 decode_type(words
, 2)
252 elif event
== "nxdomain":
256 elif event
== "query":
257 decode_serial(words
, 0)
258 decode_client(words
, 1)
259 decode_type(words
, 2)
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
268 if words
[2] == "txt": # text record
270 if response
.endswith("..."):
272 response
= response
[0:-3]
275 length
= int(response
[0:2], 16)
277 for i
in range(1, len(response
)//2):
278 chars
.append(chr(int(response
[2*i
: (2*i
)+2], 16)))
280 words
[4] = f
"{length}:\"{txt}{ellipsis}\""
282 elif event
== "sent":
283 decode_serial(words
, 0)
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]}"
292 words
[0] = f
"g={words[0]}"
293 decode_type(words
, 1)
295 # words[3] = control (domain for which these servers are believed
296 # to be authoritative)
297 for i
in range(4, len(words
)):
300 elif event
in ("tcpopen", "tcpclose"):
301 decode_client(words
, 0)
303 return f
"{timestamp} {event} " + " ".join(words
)
307 def handle_tinydns_log(line
: str) -> typing
.Optional
[str]:
309 Handle a single log line if it matches the ``tinydns_log_re`` regex.
315 The log line that might match ``tinydns_log_re``.
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
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'
331 >>> line = "this line is nonsense"
332 >>> handle_tinydns_log(line)
335 match
= tinydns_log_re
.match(line
)
339 (timestamp
, ip
, port
, id, code
, type, name
) = match
.groups()
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"
350 line_tpl
= "{timestamp} "
352 reason
= query_drop_reason
[code
]
354 line_tpl
+= "sent response to {ip}:{port} (id {id}): {type} {name}"
356 line_tpl
+= "dropped query ({reason}) from {ip}:{port}"
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}"
362 return line_tpl
.format(timestamp
=timestamp
,
371 def parse_logfile(file : typing
.TextIO
):
373 Process a single log ``file``.
379 An open log file, or stdin.
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')
390 2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
392 >>> from os import remove
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
399 tai
= Popen(["tai64nlocal"], stdin
=PIPE
, stdout
=PIPE
, text
=True, bufsize
=0)
402 tai
.stdin
.write(line
)
403 line
= tai
.stdout
.readline()
405 friendly_line
= handle_tinydns_log(line
)
406 if not friendly_line
:
407 friendly_line
= handle_dnscache_log(line
)
408 if not friendly_line
:
415 The entry point to the program.
417 This function is responsible only for parsing any command-line
418 arguments, and then calling :func`parse_logfile` on them.
420 # Create an argument parser using the file's docsctring as its
422 from argparse
import ArgumentParser
, FileType
423 parser
= ArgumentParser(description
= __doc__
)
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",
433 help="djbdns logfile to process (default: stdin)")
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.
443 # So anyway, don't run this on several million logfiles.
444 args
= parser
.parse_args()
445 for f
in args
.logfiles
:
449 if __name__
== "__main__":