xref: /freebsd/contrib/tcpdump/README.md (revision d0b2dbfa0ecf2bbc9709efc5e20baf8e4b44bbbf)
1# TCPDUMP 4.x.y by [The Tcpdump Group](https://www.tcpdump.org/)
2
3**To report a security issue please send an e-mail to security@tcpdump.org.**
4
5To report bugs and other problems, contribute patches, request a
6feature, provide generic feedback etc please see the
7[guidelines for contributing](CONTRIBUTING.md) in the tcpdump source tree root.
8
9Anonymous Git is available via
10
11	https://github.com/the-tcpdump-group/tcpdump.git
12
13This directory contains source code for tcpdump, a tool for network
14monitoring and data acquisition.
15
16Over the past few years, tcpdump has been steadily improved by the
17excellent contributions from the Internet community (just browse
18through the [change log](CHANGES)).  We are grateful for all the input.
19
20### Supported platforms
21In many operating systems tcpdump is available as a native package or port,
22which simplifies installation of updates and long-term maintenance. However,
23the native packages are sometimes a few versions behind and to try a more
24recent snapshot it will take to compile tcpdump from the source code.
25
26tcpdump compiles and works on at least the following platforms:
27
28* AIX
29* DragonFly BSD
30* FreeBSD
31* Haiku
32* HP-UX 11i
33* illumos (OmniOS, OpenIndiana)
34* GNU/Linux
35* {Mac} OS X / macOS
36* NetBSD
37* OpenBSD
38* OpenWrt
39* Solaris
40* Windows (requires WinPcap or Npcap, and Visual Studio with CMake)
41
42### Dependency on libpcap
43Tcpdump uses libpcap, a system-independent interface for user-level
44packet capture.  Before building tcpdump, you must first retrieve and
45build libpcap.
46
47Once libpcap is built (either install it or make sure it's in
48`../libpcap`), you can build tcpdump using the procedure in the
49[installation notes](INSTALL.md).
50
51### Origins of tcpdump
52The program is loosely based on SMI's "etherfind" although none of the
53etherfind code remains.  It was originally written by Van Jacobson as
54part of an ongoing research project to investigate and improve TCP and
55Internet gateway performance.  The parts of the program originally
56taken from Sun's etherfind were later re-written by Steven McCanne of
57LBL.  To insure that there would be no vestige of proprietary code in
58tcpdump, Steve wrote these pieces from the specification given by the
59manual entry, with no access to the source of tcpdump or etherfind.
60```text
61formerly from	Lawrence Berkeley National Laboratory
62		Network Research Group <tcpdump@ee.lbl.gov>
63		ftp://ftp.ee.lbl.gov/old/tcpdump.tar.Z (3.4)
64```
65
66### See also
67Richard Stevens gives an excellent treatment of the Internet protocols
68in his book *"TCP/IP Illustrated, Volume 1"*. If you want to learn more
69about tcpdump and how to interpret its output, pick up this book.
70
71Another tool that tcpdump users might find useful is
72[tcpslice](https://github.com/the-tcpdump-group/tcpslice).
73It is a program that can be used to extract portions of tcpdump binary
74trace files.
75
76### The original LBL README by Steve McCanne, Craig Leres and Van Jacobson
77```
78This directory also contains some short awk programs intended as
79examples of ways to reduce tcpdump data when you're tracking
80particular network problems:
81
82send-ack.awk
83	Simplifies the tcpdump trace for an ftp (or other unidirectional
84	tcp transfer).  Since we assume that one host only sends and
85	the other only acks, all address information is left off and
86	we just note if the packet is a "send" or an "ack".
87
88	There is one output line per line of the original trace.
89	Field 1 is the packet time in decimal seconds, relative
90	to the start of the conversation.  Field 2 is delta-time
91	from last packet.  Field 3 is packet type/direction.
92	"Send" means data going from sender to receiver, "ack"
93	means an ack going from the receiver to the sender.  A
94	preceding "*" indicates that the data is a retransmission.
95	A preceding "-" indicates a hole in the sequence space
96	(i.e., missing packet(s)), a "#" means an odd-size (not max
97	seg size) packet.  Field 4 has the packet flags
98	(same format as raw trace).  Field 5 is the sequence
99	number (start seq. num for sender, next expected seq number
100	for acks).  The number in parens following an ack is
101	the delta-time from the first send of the packet to the
102	ack.  A number in parens following a send is the
103	delta-time from the first send of the packet to the
104	current send (on duplicate packets only).  Duplicate
105	sends or acks have a number in square brackets showing
106	the number of duplicates so far.
107
108	Here is a short sample from near the start of an ftp:
109		3.00    0.20   send . 512
110		3.20    0.20    ack . 1024  (0.20)
111		3.20    0.00   send P 1024
112		3.40    0.20    ack . 1536  (0.20)
113		3.80    0.40 * send . 0  (3.80) [2]
114		3.82    0.02 *  ack . 1536  (0.62) [2]
115	Three seconds into the conversation, bytes 512 through 1023
116	were sent.  200ms later they were acked.  Shortly thereafter
117	bytes 1024-1535 were sent and again acked after 200ms.
118	Then, for no apparent reason, 0-511 is retransmitted, 3.8
119	seconds after its initial send (the round trip time for this
120	ftp was 1sec, +-500ms).  Since the receiver is expecting
121	1536, 1536 is re-acked when 0 arrives.
122
123packetdat.awk
124	Computes chunk summary data for an ftp (or similar
125	unidirectional tcp transfer). [A "chunk" refers to
126	a chunk of the sequence space -- essentially the packet
127	sequence number divided by the max segment size.]
128
129	A summary line is printed showing the number of chunks,
130	the number of packets it took to send that many chunks
131	(if there are no lost or duplicated packets, the number
132	of packets should equal the number of chunks) and the
133	number of acks.
134
135	Following the summary line is one line of information
136	per chunk.  The line contains eight fields:
137	   1 - the chunk number
138	   2 - the start sequence number for this chunk
139	   3 - time of first send
140	   4 - time of last send
141	   5 - time of first ack
142	   6 - time of last ack
143	   7 - number of times chunk was sent
144	   8 - number of times chunk was acked
145	(all times are in decimal seconds, relative to the start
146	of the conversation.)
147
148	As an example, here is the first part of the output for
149	an ftp trace:
150
151	# 134 chunks.  536 packets sent.  508 acks.
152	1       1       0.00    5.80    0.20    0.20    4       1
153	2       513     0.28    6.20    0.40    0.40    4       1
154	3       1025    1.16    6.32    1.20    1.20    4       1
155	4       1561    1.86    15.00   2.00    2.00    6       1
156	5       2049    2.16    15.44   2.20    2.20    5       1
157	6       2585    2.64    16.44   2.80    2.80    5       1
158	7       3073    3.00    16.66   3.20    3.20    4       1
159	8       3609    3.20    17.24   3.40    5.82    4       11
160	9       4097    6.02    6.58    6.20    6.80    2       5
161
162	This says that 134 chunks were transferred (about 70K
163	since the average packet size was 512 bytes).  It took
164	536 packets to transfer the data (i.e., on the average
165	each chunk was transmitted four times).  Looking at,
166	say, chunk 4, we see it represents the 512 bytes of
167	sequence space from 1561 to 2048.  It was first sent
168	1.86 seconds into the conversation.  It was last
169	sent 15 seconds into the conversation and was sent
170	a total of 6 times (i.e., it was retransmitted every
171	2 seconds on the average).  It was acked once, 140ms
172	after it first arrived.
173
174stime.awk
175atime.awk
176	Output one line per send or ack, respectively, in the form
177		<time> <seq. number>
178	where <time> is the time in seconds since the start of the
179	transfer and <seq. number> is the sequence number being sent
180	or acked.  I typically plot this data looking for suspicious
181	patterns.
182
183
184The problem I was looking at was the bulk-data-transfer
185throughput of medium delay network paths (1-6 sec.  round trip
186time) under typical DARPA Internet conditions.  The trace of the
187ftp transfer of a large file was used as the raw data source.
188The method was:
189
190  - On a local host (but not the Sun running tcpdump), connect to
191    the remote ftp.
192
193  - On the monitor Sun, start the trace going.  E.g.,
194      tcpdump host local-host and remote-host and port ftp-data >tracefile
195
196  - On local, do either a get or put of a large file (~500KB),
197    preferably to the null device (to minimize effects like
198    closing the receive window while waiting for a disk write).
199
200  - When transfer is finished, stop tcpdump.  Use awk to make up
201    two files of summary data (maxsize is the maximum packet size,
202    tracedata is the file of tcpdump tracedata):
203      awk -f send-ack.awk packetsize=avgsize tracedata >sa
204      awk -f packetdat.awk packetsize=avgsize tracedata >pd
205
206  - While the summary data files are printing, take a look at
207    how the transfer behaved:
208      awk -f stime.awk tracedata | xgraph
209    (90% of what you learn seems to happen in this step).
210
211  - Do all of the above steps several times, both directions,
212    at different times of day, with different protocol
213    implementations on the other end.
214
215  - Using one of the Unix data analysis packages (in my case,
216    S and Gary Perlman's Unix|Stat), spend a few months staring
217    at the data.
218
219  - Change something in the local protocol implementation and
220    redo the steps above.
221
222  - Once a week, tell your funding agent that you're discovering
223    wonderful things and you'll write up that research report
224    "real soon now".
225```
226