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/write 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 (preceded with a count).
-e
Print the link-level header on each dump line.
-E
Use algo:secret for decrypting IPsec ESP packets. Algorithms may be
des-cbc,
3des-cbc,
blowfish-cbc,
rc3-cbc,
cast128-cbc, or
none.
The default is des-cbc.
The ability to decrypt packets is only present if tcpdump was compiled
with cryptography enabled.
secret the ascii text for ESP secret key.
We cannot take arbitrary binary value at this moment.
The option assumes RFC2406 ESP, not RFC1827 ESP.
The option is only for debugging purposes, and
the use of this option with truly `secret' key is discouraged.
By presenting IPsec secret key onto command line
you make it visible to others, via
ps(1)
and other occasions.
-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 loopback).
Ties are broken by choosing the earliest match.
On Linux systems with 2.2 or later kernels, an
interface
argument of ``any'' can be used to capture packets from all interfaces.
Note that captures on the ``any'' device will not be done in promiscuous
mode.
-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''.
-m
Load SMI MIB module definitions from file module. This option
can be used several times to load several MIB modules into tcpdump.
-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 information 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 minimum is actually 96).
68 bytes is adequate for IP, ICMP, TCP
and UDP but may truncate protocol information 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 protocol 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 buffering. This may cause packets to be
lost. You should limit snaplen to the smallest number that will
capture the protocol information you're interested in. Setting
snaplen to 0 means use the required length to catch whole packets.
-T
Force packets selected by "expression" to be interpreted the
specified type. Currently known types are
cnfp (Cisco NetFlow protocol),
rpc (Remote Procedure Call),
rtp (Real-Time Applications protocol),
rtcp (Real-Time Applications control protocol),
snmp (Simple Network Management Protocol),
vat (Visual Audio Tool),
and
wb (distributed White Board).
-R
Assume ESP/AH packets to be based on old specification (RFC1825 to RFC1829).
If specified, tcpdump will not print replay prevention field.
Since there is no protocol version field in ESP/AH specification,
tcpdump cannot deduce the version of ESP/AH protocol.
-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,
identification, total length and options in an IP packet are printed.
Also enables additional packet integrity checks such as verifying the
IP and ICMP header checksum.
-vv
Even more verbose output. For example, additional fields are
printed from NFS reply packets.
-vvv
Even more verbose output. For example,
telnet SB ... SE options
are printed in full. With
-X
telnet options are printed in hex as well.
-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.
-X
When printing hex, print ascii too. Thus if
-x
is also set, the packet is printed in hex/ascii.
This is very handy for analysing new protocols.
Even if
-x
is not also set, some parts of some packets may be printed
in hex/ascii.
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 anddst.
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 protocols such as slip) 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,
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 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 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 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 Ethernet fields. FDDI headers also contain other fields,
but you cannot name them explicitly in a filter expression.
Similarly, `tr' is an alias for `ether'; the previous paragraph's
statements about FDDI headers also apply to Token Ring headers.]
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:
dst host host
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.
gatewayhost
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 numbers for host / ehost.)
This syntax does not work in IPv6-enabled configuration at this moment.
dst net net
True if the IPv4/v6 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 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 netmask mask
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
True if the IPv4/v6 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, 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 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, icmp6, igmp, igrp, pim, ah,
esp, udp, or tcp.
Note that the identifiers tcp, udp, and icmp are also
keywords and must be escaped via backslash (\), which is \\ in the C-shell.
Note that this primitive does not chase protocol header chain.
ip6 proto protocol
True if the packet is an IPv6 packet of protocol type protocol.
Note that this primitive does not chase 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 example,
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 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.
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, or
iso.
Note these identifiers 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. The same applies to Token Ring.]
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.
ip, ip6, arp, rarp, atalk, aarp, decnet, iso
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 protocol.
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.
Note that tcpdump does an incomplete job of parsing these 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, tr,
ip, arp, rarp, tcp, udp, icmp or ip6, and
indicates the protocol layer for the index operation.
Note that tcp, udp and other upper-layer protocol 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.
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 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[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 following
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 datagrams) 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.
On Token Ring networks, the '-e' option causes tcpdump to print
the `access control' and `frame control' fields, the source and
destination addresses, and the packet length. As on FDDI networks,
packets are assumed to contain an LLC packet. Regardless of whether
the '-e' option is specified or not, the source routing information is
printed for source-routed packets.
(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 information 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 special 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 compressed TCP packet,
with an implicit connection identifier; 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 arguments. 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:
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, ethernet addresses
are in caps and internet addresses in lower case).
This would look less redundant if we had done tcpdump -n:
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.)
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
direction 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 brackets (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.
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 contained 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 packets of the conversation, the difference between
the current 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
feature, 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 including 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 capture
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 present but the IP datagram length is not
long enough for the options to actually be there, tcpdump reports
it as ``[bad hdr length]''.
Capturing TCP packets with particular flag combinations (SYN-ACK, URG-ACK, etc.)
There are 6 bits in the control bits section of the TCP header:
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 | reserved |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 fist 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:
We see that this octet contains 2 bytes from the reserved field.
According to RFC 793 this field is reserved for future use and must
be 0. The remaining 6 bits 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:
We already mentioned that bits number 7 and 6 belong to the
reserved field, so they must must be 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
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 decimal 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:
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 representation 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 following
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.
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 contained 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 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 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 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
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, 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.
SMB/CIFS decoding
tcpdump now includes fairly extensive SMB/CIFS/NBT decoding for data
on UDP/137, UDP/138 and TCP/139. Some primitive decoding of IPX and
NetBEUI SMB data is also done.
By default a fairly minimal decode is done, with a much more detailed
decode done if -v is used. Be warned that with -v a single SMB packet
may take up a page or more, so only use -v if you really want all the
gory details.
If you are decoding SMB sessions containing unicode strings then you
may wish to set the environment variable USE_UNICODE to 1. A patch to
auto-detect unicode srings would be welcome.
For information on SMB packet formats and what all te fields mean see
www.cifs.org or the pub/samba/specs/ directory on your favourite
samba.org mirror site. The SMB patches were written by Andrew Tridgell
(tridge@samba.org).
NFS Requests and Replies
Sun NFS (Network File System) requests and replies are printed as:
src.xid > dst.nfs: len op argssrc.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 handle (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
generation 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:
(-v also prints the IP header TTL, ID, length, 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 operation. Instead,
tcpdump keeps track of ``recent'' requests, and matches them to the
replies using the transaction ID. If a reply does not closely follow the
corresponding request, it might not be parsable.
AFS Requests and Replies
Transarc AFS (Andrew File System) requests and replies are printed
as:
src.sport > dst.dport: rx packet-typesrc.sport > dst.dport: rx packet-type service call call-name argssrc.sport > dst.dport: rx packet-type service reply call-name args
elvis.7001 > pike.afsfs:
rx data fs call rename old fid 536876964/1/1 ".newsrc.new"
new fid 536876964/1/1 ".newsrc"
pike.afsfs > elvis.7001: rx data fs reply rename
In the first line, host elvis sends a RX packet to pike. This was
a RX data packet to the fs (fileserver) service, and is the start of
an RPC call. The RPC call was a rename, with the old directory file id
of 536876964/1/1 and an old filename of `.newsrc.new', and a new directory
file id of 536876964/1/1 and a new filename of `.newsrc'. The host pike
responds with a RPC reply to the rename call (which was successful, because
it was a data packet and not an abort packet).
In general, all AFS RPCs are decoded at least by RPC call name. Most
AFS RPCs have at least some of the arguments decoded (generally only
the `interesting' arguments, for some definition of interesting).
The format is intended to be self-describing, but it will probably
not be useful to people who are not familiar with the workings of
AFS and RX.
If the -v (verbose) flag is given twice, acknowledgement packets and
additional header information is printed, such as the the RX call ID,
call number, sequence number, serial number, and the RX packet flags.
If the -v flag is given twice, additional information is printed,
such as the the RX call ID, serial number, and the RX packet flags.
The MTU negotiation information is also printed from RX ack packets.
If the -v flag is given three times, the security index and service id
are printed.
Error codes are printed for abort packets, with the exception of Ubik
beacon packets (because abort packets are used to signify a yes vote
for the Ubik protocol).
Note that AFS requests are very large and many of the arguments won't
be printed unless snaplen is increased. Try using `-s 256'
to watch AFS traffic.
AFS reply packets do not explicitly identify the RPC operation. Instead,
tcpdump keeps track of ``recent'' requests, and matches them to the
replies using the call number and service ID. If a reply does not closely
follow the
corresponding 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 name1.254 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 separated by
whitespace (blanks or tabs).
The
/etc/atalk.names
file may contain blank lines or comment lines (lines starting with
a `#').
(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 indicated 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:
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 initiates transaction id 12266 with host helios by requesting
up to 8 packets (the `<0-7>'). The hex number 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 retransmitted. 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 connection that doesn't appear to handle 576 byte datagrams:
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.
NIT doesn't let you watch your own outbound traffic, BPF will.
We recommend that you use the latter.
On Linux systems with 2.0[.x] kernels:
packets on the loopback device will be seen twice;
packet filtering cannot be done in the kernel, so that all packets must
be copied from the kernel in order to be filtered in user mode;
all of a packet, not just the part that's within the snapshot length,
will be copied from the kernel (the 2.0[.x] packet capture mechanism, if
asked to copy only part of a packet to userland, will not report the
true length of the packet; this would cause most IP packets to get an
error from
tcpdump).
We recommend that you upgrade to a 2.2 or later kernel.
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 generating them rather than tcpdump.
A packet trace that crosses a daylight savings time change will give
skewed time stamps (the time change is ignored).
Filter expressions that manipulate FDDI or Token Ring headers assume
that all FDDI and Token Ring packets are SNAP-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 filter may inadvertently
accept certain packets that do not properly match the filter expression.
Filter expressions on fields other than those that manipulate Token Ring
headers will not correctly handle source-routed Token Ring packets.
ip6 proto
should chase header chain, but at this moment it does not.
ip6 protochain
is supplied for this behavior.
Arithmetic expression against transport layer headers, like tcp[0],
does not work against IPv6 packets.
It only looks at IPv4 packets.