man tcpdump      Excerpts related to ethereal Capture Filters

       Reading packets from a network interface may require that you have spe-
       cial privileges:

       Under Linux:
              You  must  be  root  or tcpdump must be installed setuid to root.

       Under BSD (this includes Mac OS X):
              You must have read access to /dev/bpf*.  On BSDs  with  a  devfs
              (this includes Mac OS X).

[ OS X - To install UNIX applications, install the Apple Xcode developer
programs, including X windows), then install and use "Darwin Ports"
(http://darwinports.opendarwin.org/) to install "ethereal".

You must enable "su root" by using the Apple "Netinfo Manager" utility. Under
the "Security" menu, do "Authenticate" and then "Enable Root User".  Next,
start X windows and, in the Xterm window, type "su root" and at the next
prompt "ethereal".]

       Reading a saved packet file doesn't require special privileges.

Capture Filters (expressions)

              Set the data  link  type  to  use  while  capturing  packets  to
              datalinktype.

        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 expression is `true' will be dumped.

              The  expression  consists of one or more primitives.  Primitives
              usually consist of an id (name or number)  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 quali-
                     fier, src or dst is assumed.  For some link layers,  such
                     as  SLIP  and  the ``cooked'' Linux capture mode used for
                     the ``any'' device and for some other device  types,  the
                     inbound  and outbound qualifiers can be used to specify a
                     desired direction.

              proto  qualifiers restrict the match to a  particular  protocol.
                     Possible protos are: ether, fddi, tr, wlan, ip, ip6, arp,
                     rarp, decnet, tcp and udp.  E.g., `ether src  foo',  `arp
                     net  128.3',  `tcp port 21'.  If there is no proto quali-
                     fier, 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 syntax), `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 speci-
              fied 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 Ethernet fields.  FDDI headers also con-
              tain other fields, but you cannot name them explicitly in a fil-
              ter expression.

              Similarly, `tr' and `wlan' are aliases for `ether'; the previous
              paragraph's statements about FDDI headers also  apply  to  Token
              Ring  and  802.11 wireless LAN headers.  For 802.11 headers, the
              destination address is the DA field and the  source  address  is
              the SA field; the BSSID, RA, and TA fields aren't tested.]

              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 primitives.  E.g., `host foo and not
              port ftp and not port  ftp-data'.   To  save  typing,  identical
              qualifier 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:  [usually the last item is your value]

              dst host host [the 2nd "host" is a variable, like "www.cnn.com"]
                     True  if  the  IPv4/v6 destination field of the packet is
                     host, which may be either an address or a name.

              src host host
                     True if the IPv4/v6 source field of the packet is host.

              host host
                     True if either the IPv4/v6 source or destination  of  the
                     packet is host.  Any of the above host expressions can be
                     prepended with the keywords, ip, arp, rarp, or ip6 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 destination 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  nei-
                     ther the IP source nor the IP destination was host.  Host
                     must be a name and must be found both  by  the  machine's
                     host-name-to-IP-address  resolution mechanisms (host name
                     file, DNS, NIS, etc.) and by the machine's  host-name-to-
                     Ethernet-address   resolution   mechanism   (/etc/ethers,
                     etc.).  (An equivalent expression is
                          ether host ehost and not host host
                     which can be used with either names or numbers for host /
                     ehost.)   This  syntax does not work in IPv6-enabled con-
                     figuration at this moment.

              dst net net
                     True if the IPv4/v6 destination address of the packet has
                     a  network  number of netNet may be either a name from
                     /etc/networks or a network number  (see  networks(4)  for
                     details).

              src net net
                     True  if  the  IPv4/v6 source address of the packet has a
                     network number of net.

              net net
                     True if either the IPv4/v6 source or destination  address
                     of the packet has a network number of net.

              net net mask netmask     [2nd "net" and "netmask" are variables]

                     True if the IP address matches net with the specific
                     netmask.  May be qualified with src or dst.  Note that  this
                     syntax is not valid for IPv6 net.

              net net/len  [net/len like 130.207.0.0/16]
                     True  if  the  IPv4/v6 address matches net with a netmask
                     len bits wide.  May be qualified with src or dst.

              dst port port
                     True if the packet is ip/tcp, ip/udp, ip6/tcp or  ip6/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  traf-
                     fic,  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       [length should be a number]
                     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 proto-
                     col type protocol.  Protocol can be a number  or  one  of
                     the  names  icmp,  icmp6, igmp, igrp, pim, ah, esp, vrrp,
                     udp, or tcp.

              ip6 proto protocol
                     True  if  the  packet  is an IPv6 packet of protocol type
                     protocol.  Note that this primitive does  not  chase  the
                     protocol header chain.

              ip6 protochain protocol
                     True  if the packet is IPv6 packet, and contains protocol
                     header with type protocol in its protocol  header  chain.
                     For example,
                          ip6 protochain 6
                     matches  any  IPv6 packet with TCP protocol header in the
                     protocol header chain.  The packet may contain, for exam-
                     ple, authentication header, routing header, or hop-by-hop
                     option header, between IPv6 header and TCP  header.   The
                     BPF  code emitted by this primitive is complex and cannot
                     be optimized by BPF optimizer code in  tcpdump,  so  this
                     can be somewhat slow.

              ip protochain protocol
                     Equivalent  to  ip6  protochain protocol, but this is for
                     IPv4.

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

              ip broadcast
                     True  if  the  packet  is  an  IPv4 broadcast packet.  It
                     checks for both the  all-zeroes  and  all-ones  broadcast
                     conventions,  and  looks up the subnet mask on the inter-
                     face on which the capture is being done.

                     If the subnet mask of the interface on which the  capture
                     is being done is not available, either because the inter-
                     face on which capture is being done  has  no  netmask  or
                     because  the  capture  is  being  done on the Linux "any"
                     interface, which can capture on more than one  interface,
                     this check will not work correctly.

              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.

              ip6 multicast
                     True if the packet is an IPv6 multicast packet.

              ether proto protocol
                     True  if  the packet is of ether type protocol.  Protocol
                     can be a number or one of the names ip, ip6,  arp,  rarp,
                     atalk,  aarp,  decnet,  sca, lat, mopdl, moprc, iso, stp,
                     ipx, or netbeui.

                     iso, sap, and netbeui
                            tcpdump checks for an 802.3 frame and then  checks
                            the  LLC  header  as it does for FDDI, Token Ring,
                            and 802.11;

                     atalk  tcpdump checks both for the AppleTalk etype in  an
                            Ethernet  frame and for a SNAP-format packet as it
                            does for FDDI, Token Ring, and 802.11;

                     aarp   tcpdump checks for  the  AppleTalk  ARP  etype  in
                            either  an  Ethernet  frame or an 802.2 SNAP frame
                            with an OUI of 0x000000;

                     ipx    tcpdump checks for the IPX etype  in  an  Ethernet
                            frame,  the  IPX  DSAP  in  the  LLC  header,  the
                            802.3-with-no-LLC-header encapsulation of IPX, and
                            the IPX etype in a SNAP frame.

              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  destination  address
                     is host.

              ifname interface
                     True  if  the packet was logged as coming from the speci-
                     fied  interface  (applies  only  to  packets  logged   by
                     OpenBSD's pf(4)).

              rnr num
                     True  if  the packet was logged as matching the specified
                     PF  rule  number  (applies  only  to  packets  logged  by
                     OpenBSD's pf(4)).

              rulenum num
                     Synonomous with the rnr modifier.

              reason code
                     True  if the packet was logged with the specified PF rea-
                     son code.  The known codes are: match, bad-offset,  frag-
                     ment, short, normalize, and memory (applies only to pack-
                     ets logged by OpenBSD's pf(4)).

              rset name
                     True if the packet was logged as matching  the  specified
                     PF  ruleset  name of an anchored ruleset (applies only to
                     packets logged by pf(4)).

              ruleset name
                     Synonomous with the rset modifier.

              srnr num
                     True if the packet was logged as matching  the  specified
                     PF  rule  number  of an anchored ruleset (applies only to
                     packets logged by pf(4)).

              subrulenum num
                     Synonomous with the srnr modifier.

              action act
                     True if PF took the specified action when the packet  was
                     logged.   Known actions are: pass and block (applies only
                     to packets logged by OpenBSD's pf(4)).

              ip, ip6, arp, rarp, atalk, aarp, decnet, iso, stp, ipx, netbeui
                     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.

              vlan [vlan_id]
                     True  if  the  packet  is an IEEE 802.1Q VLAN packet.  If
                     [vlan_id] is specified, only true is the packet  has  the
                     specified  vlan_id.   Note  that  the  first vlan keyword
                     encountered in expression changes  the  decoding  offsets
                     for  the  remainder  of expression on the assumption that
                     the packet is a VLAN packet.

              tcp, udp, icmp
                     Abbreviations for:
                          ip proto p or ip6 proto p
                     where p is one of the above protocols.

              iso proto protocol
                     True if the packet is an OSI packet of protocol type pro-
                     tocol.   Protocol  can  be  a  number or one of the names
                     clnp, esis, or isis.

              clnp, esis, isis
                     Abbreviations for:
                          iso proto p
                     where p is one of the above protocols.

              l1, l2, iih, lsp, snp, csnp, psnp
                     Abbreviations for IS-IS PDU types.

              vpi n  True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris, with a virtual path identifier of n.

              vci n  True  if  the  packet  is  an  ATM  packet, for SunATM on
                     Solaris, with a virtual channel identifier of n.

              lane   True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and is an ATM LANE packet.  Note that the first
                     lane keyword encountered in expression changes the  tests
                     done  in  the  remainder  of expression on the assumption
                     that the packet is either a LANE emulated Ethernet packet
                     or  a  LANE  LE Control packet.  If lane isn't specified,
                     the tests are done under the assumption that  the  packet
                     is an LLC-encapsulated packet.

              llc    True  if  the  packet  is  an  ATM  packet, for SunATM on
                     Solaris, and is an LLC-encapsulated packet.

              oamf4s True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and  is  a  segment  OAM  F4 flow cell (VPI=0 &
                     VCI=3).

              oamf4e True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and  is an end-to-end OAM F4 flow cell (VPI=0 &
                     VCI=4).

              oamf4  True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and is a segment or end-to-end OAM F4 flow cell
                     (VPI=0 & (VCI=3 | VCI=4)).

              oam    True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and is a segment or end-to-end OAM F4 flow cell
                     (VPI=0 & (VCI=3 | VCI=4)).

              metac  True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and  is  on  a  meta signaling circuit (VPI=0 &
                     VCI=1).

              bcc    True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris, and is on a broadcast signaling circuit (VPI=0 &
                     VCI=2).

              sc     True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris, and is on a signaling circuit (VPI=0 & VCI=5).

              ilmic  True  if  the  packet  is  an  ATM  packet, for SunATM on
                     Solaris, and is on an ILMI circuit (VPI=0 & VCI=16).

              connectmsg
                     True if the packet  is  an  ATM  packet,  for  SunATM  on
                     Solaris,  and  is  on a signaling circuit and is a Q.2931
                     Setup, Call Proceeding, Connect, Connect Ack, Release, or
                     Release Done message.

              metaconnect
                     True  if  the  packet  is  an  ATM  packet, for SunATM on
                     Solaris, and is on a meta  signaling  circuit  and  is  a
                     Q.2931  Setup,  Call  Proceeding,  Connect,  Release,  or
                     Release Done message.

              expr relop expr
                     True if the relation holds, where relop is one of  >,  <,
                     >=,  <=, =, !=, and expr is an arithmetic expression com-
                     posed of integer constants (expressed in standard C  syn-
                     tax),  the normal binary operators [+, -, *, /, &, |, <<,
                     >>], a length operator, and special  packet  data  acces-
                     sors.   To access data inside the packet, use the follow-
                     ing syntax:
                          proto [ expr : size ]
                     Proto is one of ether, fddi, tr, wlan, ppp,  slip,  link,
                     ip,  arp,  rarp, tcp, udp, icmp or ip6, and indicates the
                     protocol layer for the index  operation.   (ether,  fddi,
                     wlan,  tr,  ppp,  slip  and  link  all  refer to the link
                     layer.)  Note that tcp, udp and other upper-layer  proto-
                     col  types  only  apply  to  IPv4, not IPv6 (this will be
                     fixed in the future).  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.

                     Some  offsets  and field values may be expressed as names
                     rather than as numeric values.   The  following  protocol
                     header  field  offsets are available: icmptype (ICMP type
                     field), icmpcode (ICMP code  field),  and  tcpflags  (TCP
                     flags field).

                     The following ICMP type field values are available: icmp-
                     echoreply,  icmp-unreach,  icmp-sourcequench,  icmp-redi-
                     rect,  icmp-echo,  icmp-routeradvert, icmp-routersolicit,
                     icmp-timxceed, icmp-paramprob,  icmp-tstamp,  icmp-tstam-
                     preply,  icmp-ireq,  icmp-ireqreply,  icmp-maskreq, icmp-
                     maskreply.

                     The following TCP flags field values are available:  tcp-
                     fin, tcp-syn, tcp-rst, tcp-push, tcp-ack, tcp-urg.

              Primitives may be combined using:

                     A parenthesized group of primitives and operators (paren-
                     theses 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 sundown:
              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  packets)  of  each
       TCP conversation that involves a non-local host.
              tcpdump 'tcp[tcpflags] & (tcp-syn|tcp-fin) != 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 eth-
       ernet 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[icmptype] != icmp-echo and icmp[icmptype] != icmp-echoreply'

  = = =
       Capturing  TCP packets with particular flag combinations (SYN-ACK, URG-
       ACK, etc.)

       There are 8 bits in the control bits section of the TCP header:

              CWR | ECE | URG | ACK | PSH | RST | SYN | FIN

       Let's assume that we want to watch packets used in establishing  a  TCP
       connection.   Recall  that  TCP uses a 3-way handshake protocol when it
       initializes a new connection; the connection sequence  with  regard  to
       the TCP control bits is

              1) Caller sends SYN
              2) Recipient responds with SYN, ACK
              3) Caller sends ACK

       Now  we're  interested  in capturing packets that have only the SYN bit
       set (Step 1).  Note that we don't want packets from step  2  (SYN-ACK),
       just  a plain initial SYN.  What we need is a correct filter expression
       for tcpdump.

       Recall the structure of a TCP header without options:

        0                            15                              31
       -----------------------------------------------------------------
       |          source port          |       destination port        |
       -----------------------------------------------------------------
       |                        sequence number                        |
       -----------------------------------------------------------------
       |                     acknowledgment number                     |
       -----------------------------------------------------------------
       |  HL   | rsvd  |C|E|U|A|P|R|S|F|        window size            |
       -----------------------------------------------------------------
       |         TCP checksum          |       urgent pointer          |
       -----------------------------------------------------------------

       A TCP header usually holds  20  octets  of  data,  unless  options  are
       present.  The first line of the graph contains octets 0 - 3, the second
       line shows octets 4 - 7 etc.

       Starting to count with 0, the relevant TCP control bits  are  contained
       in octet 13:

        0             7|             15|             23|             31
       ----------------|---------------|---------------|----------------
       |  HL   | rsvd  |C|E|U|A|P|R|S|F|        window size            |
       ----------------|---------------|---------------|----------------
       |               |  13th octet   |               |               |

       Let's have a closer look at octet no. 13:

                       |               |
                       |---------------|
                       |C|E|U|A|P|R|S|F|
                       |---------------|
                       |7   5   3     0|

       These  are the TCP control bits we are interested in.  We have numbered
       the bits in this octet from 0 to 7, right to left, so the  PSH  bit  is
       bit number 3, while the URG bit is number 5.

       Recall  that  we  want to capture packets with only SYN set.  Let's see
       what happens to octet 13 if a TCP datagram arrives with the SYN bit set
       in its header:

                       |C|E|U|A|P|R|S|F|
                       |---------------|
                       |0 0 0 0 0 0 1 0|
                       |---------------|
                       |7 6 5 4 3 2 1 0|

       Looking at the control bits section we see that only bit number 1 (SYN)
       is set.

       Assuming that octet number 13 is an 8-bit unsigned integer  in  network
       byte order, the binary value of this octet is

              00000010

       and its decimal representation is

          7     6     5     4     3     2     1     0
       0*2 + 0*2 + 0*2 + 0*2 + 0*2 + 0*2 + 1*2 + 0*2  =  2

       We're  almost  done,  because  now we know that if only SYN is set, the
       value of the 13th octet in the TCP header, when interpreted as a  8-bit
       unsigned integer in network byte order, must be exactly 2.

       This relationship can be expressed as
              tcp[13] == 2

       We  can use this expression as the filter for tcpdump in order to watch
       packets which have only SYN set:
              tcpdump -i xl0 tcp[13] == 2

       The expression says "let the 13th octet of a TCP datagram have the dec-
       imal value 2", which is exactly what we want.

       Now,  let's  assume  that  we need to capture SYN packets, but we don't
       care if ACK or any other TCP control bit  is  set  at  the  same  time.
       Let's see what happens to octet 13 when a TCP datagram with SYN-ACK set
       arrives:

            |C|E|U|A|P|R|S|F|
            |---------------|
            |0 0 0 1 0 0 1 0|
            |---------------|
            |7 6 5 4 3 2 1 0|

       Now bits 1 and 4 are set in the 13th octet.  The binary value of  octet
       13 is

                   00010010

       which translates to decimal

          7     6     5     4     3     2     1     0
       0*2 + 0*2 + 0*2 + 1*2 + 0*2 + 0*2 + 1*2 + 0*2   = 18

       Now we can't just use 'tcp[13] == 18' in the tcpdump filter expression,
       because that would select only those packets that have SYN-ACK set, but
       not those with only SYN set.  Remember that we don't care if ACK or any
       other control bit is set as long as SYN is set.

       In order to achieve our goal, we need to logically AND the binary value
       of  octet  13  with  some other value to preserve the SYN bit.  We know
       that we want SYN to be set in any case,  so  we'll  logically  AND  the
       value in the 13th octet with the binary value of a SYN:


              00010010 SYN-ACK              00000010 SYN
         AND  00000010 (we want SYN)   AND  00000010 (we want SYN)
              --------                      --------
         =    00000010                 =    00000010

       We  see  that  this  AND  operation delivers the same result regardless
       whether ACK or another TCP control bit is set.  The decimal representa-
       tion  of  the  AND  value  as well as the result of this operation is 2
       (binary 00000010), so we know that for packets with SYN set the follow-
       ing relation must hold true:

              ( ( value of octet 13 ) AND ( 2 ) ) == ( 2 )

       This points us to the tcpdump filter expression

                   tcpdump -i xl0 'tcp[13] & 2 == 2'

       Note that you should use single quotes or a backslash in the expression
       to hide the AND ('&') special character from the shell. [not used in ethereal]

       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  datagram  to  port
       who on host broadcast, the Internet broadcast address.  The packet con-
       tained 84 bytes of user data.

       Some UDP services are recognized (from the source or  destination  port
       number) and the higher level protocol information printed.  In particu-
       lar, 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 description 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 ucbvax.berkeley.edu.  The query id
       was `3'.  The `+' indicates the recursion desired flag  was  set.   The
       query  length was 37 bytes, not including the UDP and IP protocol head-
       ers.  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 immediately after the `A'.

       A few anomalies are checked and may result in extra fields enclosed  in
       square  brackets:   If a query contains an answer, authority records or
       additional records 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 additional 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  (NoEr-
       ror) 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 seri-
       ously investigate name server traffic.  `-s 128' has  worked  well  for
       me.


       
       Timestamps

       By default, all output lines are preceded by a  timestamp.   The  time-
       stamp 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.
===
       tcpdump is currently being maintained by tcpdump.org.

       The current version is available via http:

              http://www.tcpdump.org/