]>
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 time
import strftime
, gmtime
9 from subprocess
import Popen
, PIPE
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):
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.
94 >>> convert_ip("7f000001")
96 >>> convert_ip("00000000000000000000ffff7f000001")
97 '0000:0000:0000:0000:0000:ffff:7f00:0001'
101 # IPv4, eg. "7f000001" -> "7f 00 00 01" -> "127.0.0.1"
102 return "%d.%d.%d.%d" % tuple(pack(">L", int(ip
, 16)))
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)])
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]),
117 def decode_ip(words
, i
):
118 words
[i
] = convert_ip(words
[i
])
120 def decode_ttl(words
, i
):
121 words
[i
] = "TTL=%s" % words
[i
]
123 def decode_serial(words
, i
):
124 serial
= int(words
[i
])
125 words
[i
] = "#%d" % serial
127 def decode_type(words
, i
):
129 words
[i
] = query_type
.get(int(qt
), qt
)
131 def handle_dnscache_log(line
, match
):
132 (timestamp
, event
, data
) = match
.groups()
135 if event
== "cached":
136 if words
[0] not in ("cname", "ns", "nxdomain"):
137 decode_type(words
, 0)
139 elif event
== "drop":
140 decode_serial(words
, 0)
142 elif event
== "lame":
145 elif event
== "nodata":
148 decode_type(words
, 2)
150 elif event
== "nxdomain":
154 elif event
== "query":
155 decode_serial(words
, 0)
156 decode_client(words
, 1)
157 decode_type(words
, 2)
162 if words
[2] not in ("cname", "mx", "ns", "ptr", "soa"):
163 decode_type(words
, 2)
164 if words
[2] == "a": # decode answer to an A query
166 if words
[2] == "txt": # text record
168 if response
.endswith("..."):
170 response
= response
[0:-3]
173 length
= int(response
[0:2], 16)
175 for i
in range(1, len(response
)/2):
176 chars
.append(chr(int(response
[2*i
: (2*i
)+2], 16)))
177 words
[4] = "%d:\"%s%s\"" % (length
, "".join(chars
), ellipsis
)
179 elif event
== "sent":
180 decode_serial(words
, 0)
182 elif event
== "stats":
183 words
[0] = "count=%s" % words
[0]
184 words
[1] = "motion=%s" % words
[1]
185 words
[2] = "udp-active=%s" % words
[2]
186 words
[3] = "tcp-active=%s" % words
[3]
189 words
[0] = "g=%s" % words
[0]
190 decode_type(words
, 1)
192 # words[3] = control (domain for which these servers are believed
193 # to be authoritative)
194 for i
in range(4, len(words
)):
197 elif event
in ("tcpopen", "tcpclose"):
198 decode_client(words
, 0)
200 print(timestamp
, event
, " ".join(words
))
203 def handle_tinydns_log(line
: str, match
: re
.Match
):
205 Handle a line that matched the ``tinydns_log_re`` regex.
211 The tinydns log line that matched ``tinydns_log_re``.
214 The match object that was returned when ``line`` was
215 tested against ``tinydns_log_re``.
220 >>> line = "2022-09-14 21:04:40.206516500 7f000001:9d61:be69 - 0001 www.example.com"
221 >>> match = tinydns_log_re.match(line)
222 >>> handle_tinydns_log(line, match)
223 2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
226 (timestamp
, ip
, port
, id, code
, type, name
) = match
.groups()
231 # Convert the "type" field to a human-readable record type name
232 # using the query_type dictionary. If the right name isn't present
233 # in the dictionary, we use the (decimal) type id instead.
234 type = int(type, 16) # "001c" -> 28
235 type = query_type
.get(type, type) # 28 -> "aaaa"
237 print(timestamp
, end
=' ')
239 reason
= query_drop_reason
[code
]
241 line_tpl
= "sent response to {ip}:{port} (id {id}): {type} {name}"
243 line_tpl
= "dropped query ({reason}) from {ip}:{port}"
245 # If the query can actually be parsed, the log line is a
246 # bit more informative than it would have been otherwise.
247 line_tpl
+= " (id {id}): {type} {name}"
249 print(line_tpl
.format(reason
=reason
,
257 def parse_logfile(file : typing
.TextIO
):
259 Process a single log ``file``.
265 An open log file, or stdin.
270 >>> line = "@4000000063227a320c4f3114 7f000001:9d61:be69 - 0001 www.example.com\n"
271 >>> from tempfile import NamedTemporaryFile
272 >>> with NamedTemporaryFile(mode="w", delete=False) as f:
273 ... _ = f.write(line)
274 >>> f = open(f.name, 'r')
276 2022-09-14 21:04:40.206516500 dropped query (no authority) from 127.0.0.1:40289 (id 48745): a www.example.com
278 >>> from os import remove
282 # Open pipe to tai64nlocal: we will write lines of our input (the
283 # raw log file) to it, and read log lines with readable timestamps
285 tai
= Popen(["tai64nlocal"], stdin
=PIPE
, stdout
=PIPE
, text
=True, bufsize
=0)
288 tai
.stdin
.write(line
)
289 line
= tai
.stdout
.readline()
291 match
= tinydns_log_re
.match(line
)
293 handle_tinydns_log(line
, match
)
296 match
= dnscache_log_re
.match(line
)
298 handle_dnscache_log(line
, match
)
304 # Create an argument parser using the file's docsctring as its
306 from argparse
import ArgumentParser
, FileType
307 parser
= ArgumentParser(description
= __doc__
)
309 # Parse zero or more positional arguments into a list of
310 # "logfiles". If none are given, read from stdin instead.
311 from sys
import stdin
312 parser
.add_argument("logfiles",
317 help="djbdns logfile to process (default: stdin)")
319 # Warning: argparse automatically opens its file arguments here,
320 # and they only get closed when the program terminates. There's no
321 # real benefit to closing them one-at-a-time after calling
322 # parse_logfile(), because the "scarce" resource of open file
323 # descriptors gets consumed immediately, before any processing has
324 # happened. In other words, if you're going to run out of file
325 # descriptors, it's going to happen right now.
327 # So anyway, don't run this on several million logfiles.
328 args
= parser
.parse_args()
329 for f
in args
.logfiles
:
333 if __name__
== "__main__":