README revision 190207
1190207Srpaulo@(#) $Header: /tcpdump/master/tcpdump/README,v 1.66.2.1 2008-05-30 01:38:20 guy Exp $ (LBL) 217680Spst 3146773SsamTCPDUMP 3.9 475115SfennerNow maintained by "The Tcpdump Group" 575115SfennerSee www.tcpdump.org 617680Spst 7190207SrpauloPlease send inquiries/comments/reports to: 8190207Srpaulo tcpdump-workers@lists.tcpdump.org 975115Sfenner 1075115SfennerAnonymous CVS is available via: 11111726Sfenner cvs -d :pserver:cvs.tcpdump.org:/tcpdump/master login 1275115Sfenner (password "anoncvs") 13111726Sfenner cvs -d :pserver:cvs.tcpdump.org:/tcpdump/master checkout tcpdump 1475115Sfenner 15172683SmlaierVersion 3.9 of TCPDUMP can be retrieved with the CVS tag "tcpdump_3_9rel1": 16146773Ssam cvs -d :pserver:cvs.tcpdump.org:/tcpdump/master checkout -r tcpdump_3_9rel1 tcpdump 1775115Sfenner 18172683SmlaierPlease submit patches against the master copy to the tcpdump project on 19172683Smlaiersourceforge.net. 2075115Sfenner 2175115Sfennerformerly from Lawrence Berkeley National Laboratory 2275115Sfenner Network Research Group <tcpdump@ee.lbl.gov> 2375115Sfenner ftp://ftp.ee.lbl.gov/tcpdump.tar.Z (3.4) 2475115Sfenner 2517680SpstThis directory contains source code for tcpdump, a tool for network 2675115Sfennermonitoring and data acquisition. This software was originally 2775115Sfennerdeveloped by the Network Research Group at the Lawrence Berkeley 2875115SfennerNational Laboratory. The original distribution is available via 2975115Sfenneranonymous ftp to ftp.ee.lbl.gov, in tcpdump.tar.Z. More recent 3075115Sfennerdevelopment is performed at tcpdump.org, http://www.tcpdump.org/ 3117680Spst 3275115SfennerTcpdump uses libpcap, a system-independent interface for user-level 3317680Spstpacket capture. Before building tcpdump, you must first retrieve and 3475115Sfennerbuild libpcap, also originally from LBL and now being maintained by 3575115Sfennertcpdump.org; see http://www.tcpdump.org/ . 3617680Spst 3717680SpstOnce libpcap is built (either install it or make sure it's in 3817680Spst../libpcap), you can build tcpdump using the procedure in the INSTALL 3917680Spstfile. 4017680Spst 4139297SfennerThe program is loosely based on SMI's "etherfind" although none of the 4239297Sfenneretherfind code remains. It was originally written by Van Jacobson as 4339297Sfennerpart of an ongoing research project to investigate and improve tcp and 4439297Sfennerinternet gateway performance. The parts of the program originally 4539297Sfennertaken from Sun's etherfind were later re-written by Steven McCanne of 4639297SfennerLBL. To insure that there would be no vestige of proprietary code in 4739297Sfennertcpdump, Steve wrote these pieces from the specification given by the 4839297Sfennermanual entry, with no access to the source of tcpdump or etherfind. 4917680Spst 5039297SfennerOver the past few years, tcpdump has been steadily improved by the 5139297Sfennerexcellent contributions from the Internet community (just browse 5239297Sfennerthrough the CHANGES file). We are grateful for all the input. 5317680Spst 5439297SfennerRichard Stevens gives an excellent treatment of the Internet protocols 5539297Sfennerin his book ``TCP/IP Illustrated, Volume 1''. If you want to learn more 5639297Sfennerabout tcpdump and how to interpret its output, pick up this book. 5717680Spst 5817680SpstSome tools for viewing and analyzing tcpdump trace files are available 5917680Spstfrom the Internet Traffic Archive: 6017680Spst 6139297Sfenner http://www.acm.org/sigcomm/ITA/ 6217680Spst 6339297SfennerAnother tool that tcpdump users might find useful is tcpslice: 6439297Sfenner 6539297Sfenner ftp://ftp.ee.lbl.gov/tcpslice.tar.Z 6639297Sfenner 6739297SfennerIt is a program that can be used to extract portions of tcpdump binary 6839297Sfennertrace files. See the above distribution for further details and 6939297Sfennerdocumentation. 7039297Sfenner 71111726SfennerProblems, bugs, questions, desirable enhancements, etc. should be sent 72190207Srpauloto the address "tcpdump-workers@lists.tcpdump.org". Bugs, support 73190207Srpaulorequests, and feature requests may also be submitted on the SourceForge 74190207Srpaulosite for tcpdump at 7517680Spst 76111726Sfenner http://sourceforge.net/projects/tcpdump/ 7775115Sfenner 78111726SfennerSource code contributions, etc. should be sent to the email address 79190207Srpaulosubmitted as patches on the SourceForge site for tcpdump. 8075115Sfenner 81111726SfennerCurrent versions can be found at www.tcpdump.org, or the SourceForge 82111726Sfennersite for tcpdump. 83111726Sfenner 8475115Sfenner - The TCPdump team 8575115Sfenner 8675115Sfenneroriginal text by: Steve McCanne, Craig Leres, Van Jacobson 8775115Sfenner 8817680Spst------------------------------------- 8917680SpstThis directory also contains some short awk programs intended as 9017680Spstexamples of ways to reduce tcpdump data when you're tracking 9117680Spstparticular network problems: 9217680Spst 9317680Spstsend-ack.awk 9417680Spst Simplifies the tcpdump trace for an ftp (or other unidirectional 9517680Spst tcp transfer). Since we assume that one host only sends and 9617680Spst the other only acks, all address information is left off and 9717680Spst we just note if the packet is a "send" or an "ack". 9817680Spst 9917680Spst There is one output line per line of the original trace. 10017680Spst Field 1 is the packet time in decimal seconds, relative 10117680Spst to the start of the conversation. Field 2 is delta-time 10217680Spst from last packet. Field 3 is packet type/direction. 10317680Spst "Send" means data going from sender to receiver, "ack" 10417680Spst means an ack going from the receiver to the sender. A 10517680Spst preceding "*" indicates that the data is a retransmission. 10617680Spst A preceding "-" indicates a hole in the sequence space 10717680Spst (i.e., missing packet(s)), a "#" means an odd-size (not max 10817680Spst seg size) packet. Field 4 has the packet flags 10917680Spst (same format as raw trace). Field 5 is the sequence 11017680Spst number (start seq. num for sender, next expected seq number 11117680Spst for acks). The number in parens following an ack is 11217680Spst the delta-time from the first send of the packet to the 11317680Spst ack. A number in parens following a send is the 11417680Spst delta-time from the first send of the packet to the 11517680Spst current send (on duplicate packets only). Duplicate 11617680Spst sends or acks have a number in square brackets showing 11717680Spst the number of duplicates so far. 11817680Spst 11917680Spst Here is a short sample from near the start of an ftp: 12017680Spst 3.00 0.20 send . 512 12117680Spst 3.20 0.20 ack . 1024 (0.20) 12217680Spst 3.20 0.00 send P 1024 12317680Spst 3.40 0.20 ack . 1536 (0.20) 12417680Spst 3.80 0.40 * send . 0 (3.80) [2] 12517680Spst 3.82 0.02 * ack . 1536 (0.62) [2] 12617680Spst Three seconds into the conversation, bytes 512 through 1023 12717680Spst were sent. 200ms later they were acked. Shortly thereafter 12817680Spst bytes 1024-1535 were sent and again acked after 200ms. 12917680Spst Then, for no apparent reason, 0-511 is retransmitted, 3.8 13017680Spst seconds after its initial send (the round trip time for this 13117680Spst ftp was 1sec, +-500ms). Since the receiver is expecting 13217680Spst 1536, 1536 is re-acked when 0 arrives. 13317680Spst 13417680Spstpacketdat.awk 13517680Spst Computes chunk summary data for an ftp (or similar 13617680Spst unidirectional tcp transfer). [A "chunk" refers to 13717680Spst a chunk of the sequence space -- essentially the packet 13817680Spst sequence number divided by the max segment size.] 13917680Spst 14017680Spst A summary line is printed showing the number of chunks, 14117680Spst the number of packets it took to send that many chunks 14217680Spst (if there are no lost or duplicated packets, the number 14317680Spst of packets should equal the number of chunks) and the 14417680Spst number of acks. 14517680Spst 14617680Spst Following the summary line is one line of information 14717680Spst per chunk. The line contains eight fields: 14817680Spst 1 - the chunk number 14917680Spst 2 - the start sequence number for this chunk 15017680Spst 3 - time of first send 15117680Spst 4 - time of last send 15217680Spst 5 - time of first ack 15317680Spst 6 - time of last ack 15417680Spst 7 - number of times chunk was sent 15517680Spst 8 - number of times chunk was acked 15617680Spst (all times are in decimal seconds, relative to the start 15717680Spst of the conversation.) 15817680Spst 15917680Spst As an example, here is the first part of the output for 16017680Spst an ftp trace: 16117680Spst 16217680Spst # 134 chunks. 536 packets sent. 508 acks. 16317680Spst 1 1 0.00 5.80 0.20 0.20 4 1 16417680Spst 2 513 0.28 6.20 0.40 0.40 4 1 16517680Spst 3 1025 1.16 6.32 1.20 1.20 4 1 16617680Spst 4 1561 1.86 15.00 2.00 2.00 6 1 16717680Spst 5 2049 2.16 15.44 2.20 2.20 5 1 16817680Spst 6 2585 2.64 16.44 2.80 2.80 5 1 16917680Spst 7 3073 3.00 16.66 3.20 3.20 4 1 17017680Spst 8 3609 3.20 17.24 3.40 5.82 4 11 17117680Spst 9 4097 6.02 6.58 6.20 6.80 2 5 17217680Spst 17317680Spst This says that 134 chunks were transferred (about 70K 17417680Spst since the average packet size was 512 bytes). It took 17517680Spst 536 packets to transfer the data (i.e., on the average 17617680Spst each chunk was transmitted four times). Looking at, 17717680Spst say, chunk 4, we see it represents the 512 bytes of 17817680Spst sequence space from 1561 to 2048. It was first sent 17917680Spst 1.86 seconds into the conversation. It was last 18017680Spst sent 15 seconds into the conversation and was sent 18117680Spst a total of 6 times (i.e., it was retransmitted every 18217680Spst 2 seconds on the average). It was acked once, 140ms 18317680Spst after it first arrived. 18417680Spst 18517680Spststime.awk 18617680Spstatime.awk 18717680Spst Output one line per send or ack, respectively, in the form 18817680Spst <time> <seq. number> 18917680Spst where <time> is the time in seconds since the start of the 19017680Spst transfer and <seq. number> is the sequence number being sent 19117680Spst or acked. I typically plot this data looking for suspicious 19217680Spst patterns. 19317680Spst 19417680Spst 19517680SpstThe problem I was looking at was the bulk-data-transfer 19617680Spstthroughput of medium delay network paths (1-6 sec. round trip 19717680Spsttime) under typical DARPA Internet conditions. The trace of the 19817680Spstftp transfer of a large file was used as the raw data source. 19917680SpstThe method was: 20017680Spst 20117680Spst - On a local host (but not the Sun running tcpdump), connect to 20217680Spst the remote ftp. 20317680Spst 20417680Spst - On the monitor Sun, start the trace going. E.g., 20517680Spst tcpdump host local-host and remote-host and port ftp-data >tracefile 20617680Spst 20717680Spst - On local, do either a get or put of a large file (~500KB), 20817680Spst preferably to the null device (to minimize effects like 20917680Spst closing the receive window while waiting for a disk write). 21017680Spst 21117680Spst - When transfer is finished, stop tcpdump. Use awk to make up 21217680Spst two files of summary data (maxsize is the maximum packet size, 21317680Spst tracedata is the file of tcpdump tracedata): 21417680Spst awk -f send-ack.awk packetsize=avgsize tracedata >sa 21517680Spst awk -f packetdat.awk packetsize=avgsize tracedata >pd 21617680Spst 21717680Spst - While the summary data files are printing, take a look at 21817680Spst how the transfer behaved: 21917680Spst awk -f stime.awk tracedata | xgraph 22017680Spst (90% of what you learn seems to happen in this step). 22117680Spst 22217680Spst - Do all of the above steps several times, both directions, 22317680Spst at different times of day, with different protocol 22417680Spst implementations on the other end. 22517680Spst 22617680Spst - Using one of the Unix data analysis packages (in my case, 22717680Spst S and Gary Perlman's Unix|Stat), spend a few months staring 22817680Spst at the data. 22917680Spst 23017680Spst - Change something in the local protocol implementation and 23117680Spst redo the steps above. 23217680Spst 23317680Spst - Once a week, tell your funding agent that you're discovering 23417680Spst wonderful things and you'll write up that research report 23517680Spst "real soon now". 236