TCPDUMP(8)                                             TCPDUMP(8)

NAME
       tcpdump - dump traffic on a network

SYNOPSIS
       tcpdump [ -adeflnNOpqStvx ] [ -c count ] [ -F file ]
               [ -i interface ] [ -r file ] [ -s snaplen ]
               [ -T type ] [ -w file ] [ expression ]

DESCRIPTION
       Tcpdump  prints  out  the  headers of packets on a network
       interface that match the boolean expression.

       Under SunOS with nit or bpf: To run tcpdump you must  have
       read  access to /dev/nit or /dev/bpf*.  Under Solaris with
       dlpi: You must have read  access  to  the  network  pseudo
       device, e.g.  /dev/le.  Under HP-UX with dlpi: You must be
       root or it must be installed setuid to root.   Under  IRIX
       with  snoop:  You  must  be  root  or it must be installed
       setuid to root.  Under Linux: You must be root or it  must
       be  installed  setuid  to  root.  Under Ultrix and Digital
       UNIX: Once the  super-user  has  enabled  promiscuous-mode
       operation  using  pfconfig(8),  any  user may run tcpdump.
       Under BSD: You must have read access to /dev/bpf*.

OPTIONS
       -a     Attempt to convert network and broadcast  addresses
              to names.

       -c     Exit after receiving count packets.

       -d     Dump  the  compiled packet-matching code in a human
              readable form to standard output and stop.

       -dd    Dump packet-matching code as a C program  fragment.

       -ddd   Dump  packet-matching code as decimal numbers (pre-
              ceded with a count).

       -e     Print the link-level header on each dump line.

       -f     Print  `foreign'  internet  addresses   numerically
              rather  than  symbolically (this option is intended
              to get around serious  brain  damage  in  Sun's  yp
              server -- usually it hangs forever translating non-
              local internet numbers).

       -F     Use file as input for the  filter  expression.   An
              additional  expression given on the command line is
              ignored.

       -i     Listen  on  interface.   If  unspecified,   tcpdump
              searches  the  system interface list for the lowest
              numbered, configured up interface (excluding  loop-
              back).   Ties  are  broken by choosing the earliest
              match.

       -l     Make stdout line buffered.  Useful if you  want  to
              see the data while capturing it.  E.g.,
              ``tcpdump  -l  |  tee  dat''  or  ``tcpdump  -l   >
              dat  &  tail  -f  dat''.

       -n     Don't convert addresses (i.e., host addresses, port
              numbers, etc.) to names.

       -N     Don't  print  domain  name  qualification  of  host
              names.  E.g., if you give this  flag  then  tcpdump
              will print ``nic'' instead of ``nic.ddn.mil''.

       -O     Do  not  run  the  packet-matching  code optimizer.
              This is useful only if you suspect  a  bug  in  the
              optimizer.

       -p     Don't  put  the  interface  into  promiscuous mode.
              Note that the interface  might  be  in  promiscuous
              mode  for  some other reason; hence, `-p' cannot be
              used as an abbreviation for `ether host  {local-hw-
              addr} or ether broadcast'.

       -q     Quick  (quiet?) output.  Print less protocol infor-
              mation so output lines are shorter.

       -r     Read packets from file (which was created with  the
              -w  option).   Standard  input  is  used if file is
              ``-''.

       -s     Snarf snaplen bytes of data from each packet rather
              than the default of 68 (with SunOS's NIT, the mini-
              mum is actually 96).  68 bytes is adequate for  IP,
              ICMP,  TCP and UDP but may truncate protocol infor-
              mation  from  name  server  and  NFS  packets  (see
              below).   Packets  truncated  because  of a limited
              snapshot  are  indicated   in   the   output   with
              ``[|proto]'', where proto is the name of the proto-
              col level at which  the  truncation  has  occurred.
              Note  that  taking  larger snapshots both increases
              the amount of time it takes to process packets and,
              effectively, decreases the amount of packet buffer-
              ing.  This may  cause  packets  to  be  lost.   You
              should  limit  snaplen  to the smallest number that
              will capture the protocol information you're inter-
              ested in.

       -T     Force packets selected by "expression" to be inter-
              preted the specified type.  Currently  known  types
              are  rpc  (Remote  Procedure  Call), rtp (Real-Time
              Applications protocol),  rtcp  (Real-Time  Applica-
              tions  control  protocol), vat (Visual Audio Tool),
              and wb (distributed White Board).

       -S     Print absolute, rather than relative, TCP  sequence
              numbers.

       -t     Don't print a timestamp on each dump line.

       -tt    Print an unformatted timestamp on each dump line.

       -v     (Slightly  more)  verbose output.  For example, the
              time to live and type of service information in  an
              IP packet is printed.

       -vv    Even  more verbose output.  For example, additional
              fields are printed from NFS reply packets.

       -w     Write the raw packets to file rather  than  parsing
              and  printing  them out.  They can later be printed
              with the -r option.  Standard  output  is  used  if
              file is ``-''.

       -x     Print  each packet (minus its link level header) in
              hex.  The smaller of the entire packet  or  snaplen
              bytes will be printed.

        expression
              selects  which  packets  will  be  dumped.   If  no
              expression is given, all packets on the net will be
              dumped.   Otherwise, only packets for which expres-
              sion is `true' will be dumped.

              The expression consists of one or more  primitives.
              Primitives  usually  consist of an id (name or num-
              ber) preceded by one or more qualifiers.  There are
              three different kinds of qualifier:

              type   qualifiers  say  what  kind  of thing the id
                     name or number refers  to.   Possible  types
                     are  host,  net and port.  E.g., `host foo',
                     `net 128.3', `port 20'.  If there is no type
                     qualifier, host is assumed.

              dir    qualifiers  specify  a  particular  transfer
                     direction  to  and/or  from  id.    Possible
                     directions  are src, dst, src or dst and src
                     and dst.  E.g., `src foo', `dst net  128.3',
                     `src  or dst port ftp-data'.  If there is no
                     dir qualifier, src or dst is  assumed.   For
                     `null' link layers (i.e. point to point pro-
                     tocols such as slip) the  inbound  and  out-
                     bound  qualifiers  can  be used to specify a
                     desired direction.

              proto  qualifiers restrict the match to a  particu-
                     lar  protocol.   Possible protos are: ether,
                     fddi,  ip,  arp,  rarp,  decnet,  lat,  sca,
                     moprc, mopdl, tcp and udp.  E.g., `ether src
                     foo', `arp net 128.3', `tcp  port  21'.   If
                     there  is  no proto qualifier, all protocols
                     consistent with the type are assumed.  E.g.,
                     `src  foo'  means  `(ip  or arp or rarp) src
                     foo' (except the latter is  not  legal  syn-
                     tax),  `net  bar' means `(ip or arp or rarp)
                     net bar' and `port 53' means `(tcp  or  udp)
                     port 53'.

              [`fddi'  is  actually  an  alias  for  `ether'; the
              parser treats them  identically  as  meaning  ``the
              data  link  level  used  on  the  specified network
              interface.''  FDDI  headers  contain  Ethernet-like
              source and destination addresses, and often contain
              Ethernet-like packet types, so you  can  filter  on
              these FDDI fields just as with the analogous Ether-
              net  fields.   FDDI  headers  also  contain   other
              fields,  but  you  cannot name them explicitly in a
              filter expression.]

              In addition to the above, there  are  some  special
              `primitive' keywords that don't follow the pattern:
              gateway, broadcast, less,  greater  and  arithmetic
              expressions.  All of these are described below.

              More  complex  filter  expressions  are built up by
              using the words and, or and not to  combine  primi-
              tives.   E.g.,  `host  foo and not port ftp and not
              port ftp-data'.  To save typing,  identical  quali-
              fier lists can be omitted.  E.g., `tcp dst port ftp
              or ftp-data or domain' is exactly the same as  `tcp
              dst  port  ftp  or tcp dst port ftp-data or tcp dst
              port domain'.

              Allowable primitives are:

              dst host host
                     True if the  IP  destination  field  of  the
                     packet  is  host,  which  may  be  either an
                     address or a name.

              src host host
                     True if the IP source field of the packet is
                     host.

              host host
                     True  if either the IP source or destination
                     of the packet is host.   Any  of  the  above
                     host  expressions  can be prepended with the
                     keywords, ip, arp, or rarp as in:
                          ip host host
                     which is equivalent to:
                          ether proto \ip and host host
                     If  host  is  a  name   with   multiple   IP
                     addresses,  each address will be checked for
                     a match.

              ether dst ehost
                     True if the ethernet destination address  is
                     ehost.   Ehost  may  be  either  a name from
                     /etc/ethers or a number (see ethers(3N)  for
                     numeric format).

              ether src ehost
                     True  if  the  ethernet  source  address  is
                     ehost.

              ether host ehost
                     True if either the ethernet source or desti-
                     nation address is ehost.

              gateway host
                     True  if  the packet used host as a gateway.
                     I.e., the  ethernet  source  or  destination
                     address  was  host but neither the IP source
                     nor the IP destination was host.  Host  must
                     be   a  name  and  must  be  found  in  both
                     /etc/hosts and /etc/ethers.  (An  equivalent
                     expression is
                          ether host ehost and not host host
                     which  can be used with either names or num-
                     bers for host / ehost.)

              dst net net
                     True if the IP destination  address  of  the
                     packet  has a network number of net. Net may
                     be either a name  from  /etc/networks  or  a
                     network    number   (see   networks(4)   for
                     details).

              src net net
                     True if the IP source address of the  packet
                     has a network number of net.

              net net
                     True  if either the IP source or destination
                     address of the packet has a  network  number
                     of net.

              net net mask mask
                     True  if the IP address matches net with the
                     specific netmask.  May be qualified with src
                     or dst.

              net net/len
                     True if the IP address matches net a netmask
                     len bits wide.  May be qualified with src or
                     dst.

              dst port port
                     True  if  the packet is ip/tcp or ip/udp and
                     has a destination port value of  port.   The
                     port  can  be  a  number  or  a name used in
                     /etc/services (see tcp(4P) and udp(4P)).  If
                     a  name  is  used,  both the port number and
                     protocol  are  checked.   If  a  number   or
                     ambiguous name is used, only the port number
                     is checked (e.g., dst port  513  will  print
                     both  tcp/login traffic and udp/who traffic,
                     and port domain will print  both  tcp/domain
                     and udp/domain traffic).

              src port port
                     True  if  the packet has a source port value
                     of port.

              port port
                     True if either  the  source  or  destination
                     port  of  the  packet  is  port.  Any of the
                     above port expressions can be prepended with
                     the keywords, tcp or udp, as in:
                          tcp src port port
                     which  matches only tcp packets whose source
                     port is port.

              less length
                     True if the packet has a length less than or
                     equal to length.  This is equivalent to:
                          len <= length.

              greater length
                     True if the packet has a length greater than
                     or equal to length.  This is equivalent to:
                          len >= length.

              ip proto protocol
                     True if the packet  is  an  ip  packet  (see
                     ip(4P)) of protocol type protocol.  Protocol
                     can be a number or one of  the  names  icmp,
                     igrp,  udp, nd, or tcp.  Note that the iden-
                     tifiers tcp, udp, and icmp are also keywords
                     and must be escaped via backslash (\), which
                     is \\ in the C-shell.

              ether broadcast
                     True if the packet is an ethernet  broadcast
                     packet.  The ether keyword is optional.

              ip broadcast
                     True  if  the  packet  is  an  IP  broadcast
                     packet.  It checks for both  the  all-zeroes
                     and   all-ones  broadcast  conventions,  and
                     looks up the local subnet mask.

              ether multicast
                     True if the packet is an ethernet  multicast
                     packet.   The  ether  keyword  is  optional.
                     This is shorthand for `ether[0] & 1 != 0'.

              ip multicast
                     True  if  the  packet  is  an  IP  multicast
                     packet.

              ether proto protocol
                     True  if  the packet is of ether type proto-
                     col.  Protocol can be a  number  or  a  name
                     like  ip,  arp, or rarp.  Note these identi-
                     fiers are also keywords and must be  escaped
                     via  backslash  (\).   [In  the case of FDDI
                     (e.g., `fddi protocol  arp'),  the  protocol
                     identification  comes from the 802.2 Logical
                     Link Control (LLC) header, which is  usually
                     layered  on top of the FDDI header.  Tcpdump
                     assumes,  when  filtering  on  the  protocol
                     identifier, that all FDDI packets include an
                     LLC header, and that the LLC  header  is  in
                     so-called SNAP format.]

              decnet src host
                     True  if  the DECNET source address is host,
                     which  may  be  an  address  of   the   form
                     ``10.123'',  or a DECNET host name.  [DECNET
                     host  name  support  is  only  available  on
                     Ultrix  systems  that  are configured to run
                     DECNET.]

              decnet dst host
                     True if the DECNET  destination  address  is
                     host.

              decnet host host
                     True if either the DECNET source or destina-
                     tion address is host.

              ip, arp, rarp, decnet
                     Abbreviations for:
                          ether proto p
                     where p is one of the above protocols.

              lat, moprc, mopdl
                     Abbreviations for:
                          ether proto p
                     where p is one of the above protocols.  Note
                     that  tcpdump does not currently know how to
                     parse these protocols.

              tcp, udp, icmp
                     Abbreviations for:
                          ip proto p
                     where p is one of the above protocols.

              expr relop expr
                     True if the relation holds, where  relop  is
                     one  of  >, <, >=, <=, =, !=, and expr is an
                     arithmetic expression  composed  of  integer
                     constants  (expressed in standard C syntax),
                     the normal binary operators [+, -, *, /,  &,
                     |],  a  length  operator, and special packet
                     data accessors.  To access data  inside  the
                     packet, use the following syntax:
                          proto [ expr : size ]
                     Proto  is one of ether, fddi, ip, arp, rarp,
                     tcp, udp, or icmp, and indicates the  proto-
                     col layer for the index operation.  The byte
                     offset, relative to the  indicated  protocol
                     layer,  is  given by expr.  Size is optional
                     and indicates the number  of  bytes  in  the
                     field  of  interest;  it  can be either one,
                     two, or four,  and  defaults  to  one.   The
                     length  operator,  indicated  by the keyword
                     len, gives the length of the packet.

                     For example, `ether[0] & 1 != 0' catches all
                     multicast  traffic.  The expression `ip[0] &
                     0xf  !=  5'  catches  all  IP  packets  with
                     options.  The expression `ip[6:2] & 0x1fff =
                     0' catches only unfragmented  datagrams  and
                     frag  zero  of  fragmented  datagrams.  This
                     check is implicitly applied to the  tcp  and
                     udp  index operations.  For instance, tcp[0]
                     always means  the  first  byte  of  the  TCP
                     header, and never means the first byte of an
                     intervening fragment.

              Primitives may be combined using:

                     A  parenthesized  group  of  primitives  and
                     operators  (parentheses  are  special to the
                     Shell and must be escaped).

                     Negation (`!' or `not').

                     Concatenation (`&&' or `and').

                     Alternation (`||' or `or').

              Negation has highest precedence.   Alternation  and
              concatenation  have  equal precedence and associate
              left to right.  Note that explicit and tokens,  not
              juxtaposition,  are now required for concatenation.

              If an identifier is given without  a  keyword,  the
              most recent keyword is assumed.  For example,
                   not host vs and ace
              is short for
                   not host vs and host ace
              which should not be confused with
                   not ( host vs or ace )

              Expression  arguments  can  be passed to tcpdump as
              either a single argument or as multiple  arguments,
              whichever  is  more  convenient.  Generally, if the
              expression contains  Shell  metacharacters,  it  is
              easier  to  pass  it  as a single, quoted argument.
              Multiple arguments  are  concatenated  with  spaces
              before being parsed.

EXAMPLES
       To  print  all  packets arriving at or departing from sun-
       down:
              tcpdump host sundown

       To print traffic between helios and either hot or ace:
              tcpdump host helios and \( hot or ace \)

       To print all IP packets between ace and  any  host  except
       helios:
              tcpdump ip host ace and not helios

       To  print  all  traffic  between  local hosts and hosts at
       Berkeley:
              tcpdump net ucb-ether

       To print all ftp traffic through  internet  gateway  snup:
       (note  that  the expression is quoted to prevent the shell
       from (mis-)interpreting the parentheses):
              tcpdump 'gateway snup and (port ftp or ftp-data)'

       To print traffic neither sourced  from  nor  destined  for
       local  hosts  (if you gateway to one other net, this stuff
       should never make it onto your local net).
              tcpdump ip and not net localnet

       To print the start and end packets (the SYN and FIN  pack-
       ets)  of  each  TCP conversation that involves a non-local
       host.
              tcpdump 'tcp[13] & 3 != 0 and not src and dst net localnet'

       To print IP packets longer than  576  bytes  sent  through
       gateway snup:
              tcpdump 'gateway snup and ip[2:2] > 576'

       To  print  IP broadcast or multicast packets that were not
       sent via ethernet broadcast or multicast:
              tcpdump 'ether[0] & 1 = 0 and ip[16] >= 224'

       To  print   all   ICMP   packets   that   are   not   echo
       requests/replies (i.e., not ping packets):
              tcpdump 'icmp[0] != 8 and icmp[0] != 0"

OUTPUT FORMAT
       The  output of tcpdump is protocol dependent.  The follow-
       ing gives a brief description and examples of most of  the
       formats.

       Link Level Headers

       If  the  '-e'  option  is  given, the link level header is
       printed out.  On ethernets,  the  source  and  destination
       addresses, protocol, and packet length are printed.

       On FDDI networks, the  '-e' option causes tcpdump to print
       the `frame control' field,   the  source  and  destination
       addresses,  and  the  packet length.  (The `frame control'
       field governs  the  interpretation  of  the  rest  of  the
       packet.  Normal packets (such as those containing IP data-
       grams) are `async' packets, with a priority value  between
       0  and 7; for example, `async4'.  Such packets are assumed
       to contain an 802.2 Logical Link Control (LLC) packet; the
       LLC  header  is  printed if it is not an ISO datagram or a
       so-called SNAP packet.

       (N.B.: The following description assumes familiarity  with
       the SLIP compression algorithm described in RFC-1144.)

       On  SLIP  links, a direction indicator (``I'' for inbound,
       ``O'' for outbound), packet type, and compression informa-
       tion  are  printed out.  The packet type is printed first.
       The three types are ip, utcp, and ctcp.  No  further  link
       information  is  printed for ip packets.  For TCP packets,
       the connection identifier is printed following  the  type.
       If the packet is compressed, its encoded header is printed
       out.  The special cases are printed out as *S+n and *SA+n,
       where  n  is  the  amount by which the sequence number (or
       sequence number and ack) has changed.  If it is not a spe-
       cial  case, zero or more changes are printed.  A change is
       indicated by U (urgent pointer), W (window),  A  (ack),  S
       (sequence  number), and I (packet ID), followed by a delta
       (+n or -n), or a new value (=n).  Finally, the  amount  of
       data  in  the  packet  and  compressed  header  length are
       printed.

       For example, the following line  shows  an  outbound  com-
       pressed  TCP  packet,  with an implicit connection identi-
       fier; the ack has changed by 6, the sequence number by 49,
       and  the  packet  ID by 6; there are 3 bytes of data and 6
       bytes of compressed header:
              O ctcp * A+6 S+49 I+6 3 (6)

       ARP/RARP Packets

       Arp/rarp output shows the type of request  and  its  argu-
       ments.   The  format  is  intended to be self explanatory.
       Here is a short sample taken from the start of an `rlogin'
       from host rtsg to host csam:
              arp who-has csam tell rtsg
              arp reply csam is-at CSAM
       The  first  line  says that rtsg sent an arp packet asking
       for the ethernet address  of  internet  host  csam.   Csam
       replies with its ethernet address (in this example, ether-
       net addresses are in caps and internet addresses in  lower
       case).

       This would look less redundant if we had done tcpdump -n:
              arp who-has 128.3.254.6 tell 128.3.254.68
              arp reply 128.3.254.6 is-at 02:07:01:00:01:c4

       If  we had done tcpdump -e, the fact that the first packet
       is broadcast and the second  is  point-to-point  would  be
       visible:
              RTSG Broadcast 0806  64: arp who-has csam tell rtsg
              CSAM RTSG 0806  64: arp reply csam is-at CSAM
       For the first packet this says the ethernet source address
       is  RTSG,  the  destination  is  the  ethernet   broadcast
       address,   the   type   field  contained  hex  0806  (type
       ETHER_ARP) and the total length was 64 bytes.

       TCP Packets

       (N.B.:The following description assumes  familiarity  with
       the  TCP  protocol  described  in RFC-793.  If you are not
       familiar with the protocol, neither this  description  nor
       tcpdump will be of much use to you.)

       The general format of a tcp protocol line is:
              src > dst: flags data-seqno ack window urgent options
       Src  and  dst  are the source and destination IP addresses
       and ports.  Flags are  some  combination  of  S  (SYN),  F
       (FIN),  P  (PUSH)  or  R (RST) or a single `.' (no flags).
       Data-seqno describes the portion of sequence space covered
       by  the  data  in this packet (see example below).  Ack is
       sequence number of the next data expected the other direc-
       tion on this connection.  Window is the number of bytes of
       receive buffer space available the other direction on this
       connection.   Urg  indicates there is `urgent' data in the
       packet.  Options are tcp options enclosed in angle  brack-
       ets (e.g., <mss 1024>).

       Src,  dst  and flags are always present.  The other fields
       depend on the contents of the packet's tcp protocol header
       and are output only if appropriate.

       Here is the opening portion of an rlogin from host rtsg to
       host csam.
              rtsg.1023 > csam.login: S 768512:768512(0) win 4096 <mss 1024>
              csam.login > rtsg.1023: S 947648:947648(0) ack 768513 win 4096 <ms
s 1024>
              rtsg.1023 > csam.login: . ack 1 win 4096
              rtsg.1023 > csam.login: P 1:2(1) ack 1 win 4096
              csam.login > rtsg.1023: . ack 2 win 4096
              rtsg.1023 > csam.login: P 2:21(19) ack 1 win 4096
              csam.login > rtsg.1023: P 1:2(1) ack 21 win 4077
              csam.login > rtsg.1023: P 2:3(1) ack 21 win 4077 urg 1
              csam.login > rtsg.1023: P 3:4(1) ack 21 win 4077 urg 1
       The first line says that tcp port  1023  on  rtsg  sent  a
       packet  to  port  login on csam.  The S indicates that the
       SYN flag was set.  The packet sequence number  was  768512
       and    it   contained   no   data.    (The   notation   is
       `first:last(nbytes)' which means `sequence  numbers  first
       up to but not including last which is nbytes bytes of user
       data'.)  There was  no  piggy-backed  ack,  the  available
       receive window was 4096 bytes and there was a max-segment-
       size option requesting an mss of 1024 bytes.

       Csam replies with a similar packet except  it  includes  a
       piggy-backed  ack  for  rtsg's SYN.  Rtsg then acks csam's
       SYN.  The `.' means no flags were set.   The  packet  con-
       tained  no data so there is no data sequence number.  Note
       that the ack sequence number is a small integer (1).   The
       first  time  tcpdump  sees a tcp `conversation', it prints
       the sequence number from the packet.  On subsequent  pack-
       ets  of  the conversation, the difference between the cur-
       rent packet's sequence number and  this  initial  sequence
       number is printed.  This means that sequence numbers after
       the first can be interpreted as relative byte positions in
       the  conversation's  data stream (with the first data byte
       each direction being `1').  `-S' will override  this  fea-
       ture,  causing the original sequence numbers to be output.

       On the 6th line, rtsg sends csam 19 bytes of data (bytes 2
       through  20 in the rtsg -> csam side of the conversation).
       The PUSH flag is set in the packet.  On the 7th line, csam
       says it's received data sent by rtsg up to but not includ-
       ing byte 21.  Most of this data is apparently  sitting  in
       the  socket  buffer since csam's receive window has gotten
       19 bytes smaller.  Csam also sends one  byte  of  data  to
       rtsg in this packet.  On the 8th and 9th lines, csam sends
       two bytes of urgent, pushed data to rtsg.

       If the snapshot was small enough that tcpdump didn't  cap-
       ture  the  full  TCP  header, it interprets as much of the
       header as it can and then reports ``[|tcp]''  to  indicate
       the  remainder  could  not  be interpreted.  If the header
       contains a bogus option (one with a length  that's  either
       too  small  or  beyond  the  end  of  the header), tcpdump
       reports it as ``[bad opt]'' and  does  not  interpret  any
       further  options (since it's impossible to tell where they
       start).  If the header length indicates options  are  pre-
       sent but the IP datagram length is not long enough for the
       options to actually be there, tcpdump reports it as ``[bad
       hdr length]''.

       UDP Packets

       UDP format is illustrated by this rwho packet:
              actinide.who > broadcast.who: udp 84
       This  says that port who on host actinide sent a udp data-
       gram to port who on host broadcast, the Internet broadcast
       address.  The packet contained 84 bytes of user data.

       Some  UDP services are recognized (from the source or des-
       tination port number) and the higher level protocol infor-
       mation   printed.   In  particular,  Domain  Name  service
       requests (RFC-1034/1035) and Sun RPC calls  (RFC-1050)  to
       NFS.

       UDP Name Server Requests

       (N.B.:The  following  description assumes familiarity with
       the Domain Service protocol described in RFC-1035.  If you
       are not familiar with the protocol, the following descrip-
       tion will appear to be written in greek.)

       Name server requests are formatted as
              src > dst: id op? flags qtype qclass name (len)
              h2opolo.1538 > helios.domain: 3+ A? ucbvax.berkeley.edu. (37)
       Host h2opolo asked the domain  server  on  helios  for  an
       address  record  (qtype=A)  associated  with the name ucb-
       vax.berkeley.edu.  The query id was `3'.   The  `+'  indi-
       cates  the  recursion  desired  flag  was  set.  The query
       length was 37 bytes, not including the UDP and IP protocol
       headers.   The  query operation was the normal one, Query,
       so the op field was omitted.  If the op had been  anything
       else,  it  would have been printed between the `3' and the
       `+'.  Similarly, the qclass was the normal one, C_IN,  and
       omitted.  Any other qclass would have been printed immedi-
       ately after the `A'.

       A few anomalies are checked and may result in extra fields
       enclosed  in  square  brackets:   If  a  query contains an
       answer,  name  server  or  authority   section,   ancount,
       nscount,  or  arcount  are  printed  as  `[na]', `[nn]' or
       `[nau]' where n is the appropriate count.  If any  of  the
       response  bits  are  set  (AA,  RA or rcode) or any of the
       `must be zero' bits  are  set  in  bytes  two  and  three,
       `[b2&3=x]'  is printed, where x is the hex value of header
       bytes two and three.

       UDP Name Server Responses

       Name server responses are formatted as
              src > dst:  id op rcode flags a/n/au type class data (len)
              helios.domain > h2opolo.1538: 3 3/3/7 A 128.32.137.3 (273)
              helios.domain > h2opolo.1537: 2 NXDomain* 0/1/0 (97)
       In the first example, helios responds to query id  3  from
       h2opolo with 3 answer records, 3 name server records and 7
       authority records.  The first  answer  record  is  type  A
       (address)  and  its data is internet address 128.32.137.3.
       The total size of the response was  273  bytes,  excluding
       UDP  and  IP  headers.   The  op (Query) and response code
       (NoError) were omitted, as was the class (C_IN) of  the  A
       record.

       In  the  second example, helios responds to query 2 with a
       response code of non-existent domain  (NXDomain)  with  no
       answers,  one  name  server and no authority records.  The
       `*' indicates that the authoritative answer bit  was  set.
       Since  there  were no answers, no type, class or data were
       printed.

       Other flag characters that might appear are `-' (recursion
       available,  RA,  not  set) and `|' (truncated message, TC,
       set).  If the `question' section doesn't  contain  exactly
       one entry, `[nq]' is printed.

       Note  that  name  server requests and responses tend to be
       large and the default snaplen of 68 bytes may not  capture
       enough  of  the  packet  to  print.   Use  the  -s flag to
       increase the snaplen if you need to seriously  investigate
       name server traffic.  `-s 128' has worked well for me.

       NFS Requests and Replies

       Sun  NFS  (Network  File  System) requests and replies are
       printed as:
              src.xid > dst.nfs: len op args
              src.nfs > dst.xid: reply stat len op results

              sushi.6709 > wrl.nfs: 112 readlink fh 21,24/10.73165
              wrl.nfs > sushi.6709: reply ok 40 readlink "../var"
              sushi.201b > wrl.nfs:
                   144 lookup fh 9,74/4096.6878 "xcolors"
              wrl.nfs > sushi.201b:
                   reply ok 128 lookup fh 9,74/4134.3150

       In the first line, host sushi sends a transaction with  id
       6709  to  wrl (note that the number following the src host
       is a transaction id, not the source  port).   The  request
       was  112  bytes,  excluding  the  UDP and IP headers.  The
       operation was a readlink (read symbolic link) on file han-
       dle (fh) 21,24/10.731657119.  (If one is lucky, as in this
       case, the file handle can be interpreted as a  major,minor
       device  number pair, followed by the inode number and gen-
       eration number.)  Wrl replies `ok' with  the  contents  of
       the link.

       In  the  third  line,  sushi  asks  wrl to lookup the name
       `xcolors' in directory file 9,74/4096.6878.  Note that the
       data printed depends on the operation type.  The format is
       intended to be self explanatory  if  read  in  conjunction
       with an NFS protocol spec.

       If  the -v (verbose) flag is given, additional information
       is printed.  For example:

              sushi.1372a > wrl.nfs:
                   148 read fh 21,11/12.195 8192 bytes @ 24576
              wrl.nfs > sushi.1372a:
                   reply ok 1472 read REG 100664 ids 417/0 sz 29388

       (-v also prints the IP header TTL, ID,  and  fragmentation
       fields,  which  have  been omitted from this example.)  In
       the first line, sushi asks wrl to  read  8192  bytes  from
       file  21,11/12.195,  at  byte  offset  24576.  Wrl replies
       `ok'; the packet shown on the second  line  is  the  first
       fragment  of  the reply, and hence is only 1472 bytes long
       (the other bytes will follow in subsequent fragments,  but
       these fragments do not have NFS or even UDP headers and so
       might not be printed, depending on the  filter  expression
       used).   Because  the  -v  flag is given, some of the file
       attributes (which are returned in  addition  to  the  file
       data)  are  printed:  the  file type (``REG'', for regular
       file), the file mode (in octal), the uid and gid, and  the
       file size.

       If  the -v flag is given more than once, even more details
       are printed.

       Note that NFS requests are very  large  and  much  of  the
       detail  won't be printed unless snaplen is increased.  Try
       using `-s 192' to watch NFS traffic.

       NFS reply packets do not explicitly identify the RPC oper-
       ation.    Instead,   tcpdump  keeps  track  of  ``recent''
       requests, and matches them to the replies using the trans-
       action  ID.  If a reply does not closely follow the corre-
       sponding request, it might not be parsable.

       KIP Appletalk (DDP in UDP)

       Appletalk DDP packets encapsulated in  UDP  datagrams  are
       de-encapsulated  and  dumped as DDP packets (i.e., all the
       UDP  header   information   is   discarded).    The   file
       /etc/atalk.names  is  used  to translate appletalk net and
       node numbers to names.  Lines in this file have the form
              number    name

              1.254          ether
              16.1      icsd-net
              1.254.110 ace
       The first two lines give the names of appletalk  networks.
       The third line gives the name of a particular host (a host
       is distinguished from a net by the 3rd octet in the number
       - a net number must have two octets and a host number must
       have three octets.)  The number and name should  be  sepa-
       rated    by    whitespace    (blanks    or   tabs).    The
       /etc/atalk.names file may contain blank lines  or  comment
       lines (lines starting with a `#').

       Appletalk addresses are printed in the form
              net.host.port

              144.1.209.2 > icsd-net.112.220
              office.2 > icsd-net.112.220
              jssmag.149.235 > icsd-net.2
       (If  the /etc/atalk.names doesn't exist or doesn't contain
       an entry for some appletalk host/net number, addresses are
       printed  in numeric form.)  In the first example, NBP (DDP
       port 2) on net 144.1 node 209 is sending  to  whatever  is
       listening  on  port  220 of net icsd node 112.  The second
       line is the same except the full name of the  source  node
       is  known  (`office').  The third line is a send from port
       235 on net jssmag node 149 to broadcast  on  the  icsd-net
       NBP  port  (note that the broadcast address (255) is indi-
       cated by a net name with no host number - for this  reason
       it's a good idea to keep node names and net names distinct
       in /etc/atalk.names).

       NBP (name binding protocol) and ATP (Appletalk transaction
       protocol)  packets have their contents interpreted.  Other
       protocols just dump the protocol name  (or  number  if  no
       name is registered for the protocol) and packet size.

       NBP packets are formatted like the following examples:
              icsd-net.112.220 > jssmag.2: nbp-lkup 190: "=:LaserWriter@*"
              jssmag.209.2 > icsd-net.112.220: nbp-reply 190: "RM1140:LaserWrite
r@*" 250
              techpit.2 > icsd-net.112.220: nbp-reply 190: "techpit:LaserWriter@
*" 186
       The  first  line is a name lookup request for laserwriters
       sent by net icsd host 112 and  broadcast  on  net  jssmag.
       The nbp id for the lookup is 190.  The second line shows a
       reply for this request (note that it has the same id) from
       host  jssmag.209 saying that it has a laserwriter resource
       named "RM1140" registered on port 250.  The third line  is
       another  reply to the same request saying host techpit has
       laserwriter "techpit" registered on port 186.

       ATP packet formatting is  demonstrated  by  the  following
       example:
              jssmag.209.165 > helios.132: atp-req  12266<0-7> 0xae030001
              helios.132 > jssmag.209.165: atp-resp 12266:0 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:1 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:2 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:3 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:4 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:5 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:6 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp*12266:7 (512) 0xae040000
              jssmag.209.165 > helios.132: atp-req  12266<3,5> 0xae030001
              helios.132 > jssmag.209.165: atp-resp 12266:3 (512) 0xae040000
              helios.132 > jssmag.209.165: atp-resp 12266:5 (512) 0xae040000
              jssmag.209.165 > helios.132: atp-rel  12266<0-7> 0xae030001
              jssmag.209.133 > helios.132: atp-req* 12267<0-7> 0xae030002
       Jssmag.209 initiates transaction id 12266 with host helios
       by requesting up to 8 packets (the `<0-7>').  The hex num-
       ber  at the end of the line is the value of the `userdata'
       field in the request.

       Helios responds with 8  512-byte  packets.   The  `:digit'
       following  the  transaction  id  gives the packet sequence
       number in the transaction and the number in parens is  the
       amount  of  data  in the packet, excluding the atp header.
       The `*' on packet 7 indicates that the EOM bit was set.

       Jssmag.209 then requests that packets 3 & 5 be retransmit-
       ted.   Helios  resends  them  then jssmag.209 releases the
       transaction.   Finally,  jssmag.209  initiates  the   next
       request.   The  `*'  on  the  request  indicates  that  XO
       (`exactly once') was not set.

       IP Fragmentation

       Fragmented Internet datagrams are printed as
              (frag id:size@offset+)
              (frag id:size@offset)
       (The first form indicates there are more  fragments.   The
       second indicates this is the last fragment.)

       Id  is  the  fragment  id.   Size is the fragment size (in
       bytes) excluding the IP header.  Offset is this fragment's
       offset (in bytes) in the original datagram.

       The fragment information is output for each fragment.  The
       first fragment contains the higher level  protocol  header
       and  the  frag  info  is  printed after the protocol info.
       Fragments after the first contain no higher level protocol
       header  and  the frag info is printed after the source and
       destination addresses.  For example, here is  part  of  an
       ftp from arizona.edu to lbl-rtsg.arpa over a CSNET connec-
       tion that doesn't appear to handle 576 byte datagrams:
              arizona.ftp-data > rtsg.1170: . 1024:1332(308) ack 1 win 4096 (fra
g 595a:328@0+)
              arizona > rtsg: (frag 595a:204@328)
              rtsg.1170 > arizona.ftp-data: . ack 1536 win 2560
       There are  a  couple  of  things  to  note  here:   First,
       addresses  in  the  2nd  line  don't include port numbers.
       This is because the TCP protocol information is all in the
       first  fragment  and  we  have  no  idea  what the port or
       sequence numbers are when we print  the  later  fragments.
       Second,  the tcp sequence information in the first line is
       printed as if there were 308 bytes of user data  when,  in
       fact,  there  are 512 bytes (308 in the first frag and 204
       in the second).  If you  are  looking  for  holes  in  the
       sequence  space  or  trying to match up acks with packets,
       this can fool you.

       A packet with the IP don't fragment flag is marked with  a
       trailing (DF).

       Timestamps

       By  default, all output lines are preceded by a timestamp.
       The timestamp is the current clock time in the form
              hh:mm:ss.frac
       and is as accurate as the kernel's clock.   The  timestamp
       reflects  the  time  the  kernel first saw the packet.  No
       attempt is made to account for the time lag  between  when
       the  ethernet  interface  removed the packet from the wire
       and when the kernel serviced the `new packet' interrupt.

SEE ALSO
       traffic(1C), nit(4P), bpf(4), pcap(3)

AUTHORS
       Van Jacobson, Craig Leres and Steven McCanne, all  of  the
       Lawrence Berkeley National Laboratory, University of Cali-
       fornia, Berkeley, CA.

       The current version is available via anonymous ftp:

              ftp://ftp.ee.lbl.gov/tcpdump.tar.Z

BUGS
       Please send bug reports to tcpdump@ee.lbl.gov.

       NIT doesn't let you watch your own outbound  traffic,  BPF
       will.  We recommend that you use the latter.

       Some attempt should be made to reassemble IP fragments or,
       at least to compute the right length for the higher  level
       protocol.

       Name  server inverse queries are not dumped correctly: The
       (empty) question section is printed rather than real query
       in  the answer section.  Some believe that inverse queries
       are themselves a bug and prefer to fix the program  gener-
       ating them rather than tcpdump.

       Apple  Ethertalk  DDP packets could be dumped as easily as
       KIP DDP packets but aren't.  Even if we were  inclined  to
       do  anything  to promote the use of Ethertalk (we aren't),
       LBL doesn't allow Ethertalk on any of its networks so we'd
       would have no way of testing this code.

       A packet trace that crosses a daylight savings time change
       will give skewed time stamps (the time change is ignored).

       Filters  expressions  that  manipulate FDDI headers assume
       that all FDDI packets are encapsulated  Ethernet  packets.
       This  is true for IP, ARP, and DECNET Phase IV, but is not
       true for protocols such as ISO CLNS.  Therefore, the  fil-
       ter  may  inadvertently accept certain packets that do not
       properly match the filter expression.

                           30 June 1997                         1