IPv4

TRACEROUTE(1)                Traceroute For Linux                TRACEROUTE(1)

NAME
       traceroute - print the route packets trace to network host

SYNOPSIS
       traceroute [-46dFITUnreAV] [-f first_ttl] [-g gate,...]
               [-i device] [-m max_ttl] [-p port] [-s src_addr]
               [-q nqueries] [-N squeries] [-t tos]
               [-l flow_label] [-w waittimes] [-z sendwait] [-UL] [-D]
               [-P proto] [--sport=port] [-M method] [-O mod_options]
               [--mtu] [--back]
               host [packet_len]
       traceroute6  [options]
       tcptraceroute  [options]
       lft  [options]

DESCRIPTION
       traceroute  tracks  the route packets taken from an IP network on their
       way to a given host. It utilizes the IP protocol's time to  live  (TTL)
       field  and  attempts to elicit an ICMP TIME_EXCEEDED response from each
       gateway along the path to the host.

       traceroute6 is equivalent to traceroute -6

       tcptraceroute is equivalent to traceroute -T

       lft , the Layer  Four  Traceroute,  performs  a  TCP  traceroute,  like
       traceroute -T , but attempts to provide compatibility with the original
       such implementation, also called "lft".

       The only required parameter is the name or IP address of  the  destina‐
       tion host .  The optional packet_len`gth is the total size of the prob‐
       ing packet (default 60 bytes for IPv4 and 80 for IPv6).  The  specified
       size  can  be  ignored  in some situations or increased up to a minimal
       value.

       This program attempts to trace the route an IP packet would  follow  to
       some internet host by launching probe packets with a small ttl (time to
       live) then listening for an ICMP "time exceeded" reply from a  gateway.
       We  start our probes with a ttl of one and increase by one until we get
       an ICMP "port unreachable" (or TCP reset), which means we  got  to  the
       "host",  or hit a max (which defaults to 30 hops). Three probes (by de‐
       fault) are sent at each ttl setting and a line is printed  showing  the
       ttl,  address of the gateway and round trip time of each probe. The ad‐
       dress can be followed by additional information when requested. If  the
       probe  answers  come  from  different gateways, the address of each re‐
       sponding system will be printed.  If there is no response within a cer‐
       tain timeout, an "*" (asterisk) is printed for that probe.

       After the trip time, some additional annotation can be printed: !H, !N,
       or !P  (host,  network  or  protocol  unreachable),  !S  (source  route
       failed),  !F (fragmentation needed), !X (communication administratively
       prohibited), !V (host precedence violation), !C (precedence  cutoff  in
       effect),  or  !<num>  (ICMP unreachable code <num>).  If almost all the
       probes result in some kind of unreachable, traceroute will give up  and
       exit.

       We don't want the destination host to process the UDP probe packets, so
       the destination port is set to an unlikely value  (you  can  change  it
       with  the  -p flag). There is no such a problem for ICMP or TCP tracer‐
       outing (for TCP we use half-open technique, which prevents  our  probes
       to be seen by applications on the destination host).

       In  the  modern  network environment the traditional traceroute methods
       can not be always applicable, because of widespread use  of  firewalls.
       Such  firewalls  filter  the "unlikely" UDP ports, or even ICMP echoes.
       To solve this, some additional  tracerouting  methods  are  implemented
       (including  tcp), see LIST OF AVAILABLE METHODS below. Such methods try
       to use particular protocol and source/destination port, in order to by‐
       pass firewalls (to be seen by firewalls just as a start of allowed type
       of a network session).

OPTIONS
       --help Print help info and exit.

       -4, -6 Explicitly force IPv4 or IPv6 tracerouting. By default, the pro‐
              gram  will  try to resolve the name given, and choose the appro‐
              priate protocol automatically. If resolving a host name  returns
              both IPv4 and IPv6 addresses, traceroute will use IPv4.

       -I, --icmp
              Use ICMP ECHO for probes

       -T, --tcp
              Use TCP SYN for probes

       -d, --debug
              Enable  socket  level  debugging (when the Linux kernel supports
              it)

       -F, --dont-fragment
              Do not fragment probe packets. (For IPv4 it also  sets  DF  bit,
              which  tells  intermediate  routers  not to fragment remotely as
              well).

              Varying the size of the probing packet by the packet_len command
              line  parameter,  you  can manually obtain information about the
              MTU of individual network hops. The  --mtu  option  (see  below)
              tries to do this automatically.

              Note, that non-fragmented features (like -F or --mtu) work prop‐
              erly since the Linux kernel 2.6.22 only.  Before  that  version,
              IPv6  was always fragmented, IPv4 could use the once the discov‐
              ered final mtu only (from the route cache), which  can  be  less
              than the actual mtu of a device.

       -f first_ttl, --first=first_ttl
              Specifies with what TTL to start. Defaults to 1.

       -g gateway, --gateway=gateway
              Tells  traceroute to add an IP source routing option to the out‐
              going packet that tells the network to route the packet  through
              the specified gateway (most routers have disabled source routing
              for security reasons).  In general, several gateway's is allowed
              (comma  separated).  For  IPv6, the form of num,addr,addr...  is
              allowed, where num is a route header type (default is  type  2).
              Note the type 0 route header is now deprecated (rfc5095).

       -i interface, --interface=interface
              Specifies  the  interface  through  which traceroute should send
              packets. By default, the interface is selected according to  the
              routing table.

       -m max_ttl, --max-hops=max_ttl
              Specifies  the  maximum  number of hops (max time-to-live value)
              traceroute will probe. The default is 30.

       -N squeries, --sim-queries=squeries
              Specifies the number of probe packets sent  out  simultaneously.
              Sending several probes concurrently can speed up traceroute con‐
              siderably. The default value is 16.
              Note that some routers and hosts can use ICMP  rate  throttling.
              In such a situation specifying too large number can lead to loss
              of some responses.

       -n     Do not try to map IP addresses to  host  names  when  displaying
              them.

       -p port, --port=port
              For  UDP tracing, specifies the destination port base traceroute
              will use (the destination port number  will  be  incremented  by
              each probe).
              For ICMP tracing, specifies the initial ICMP sequence value (in‐
              cremented by each probe too).
              For TCP and others specifies  just  the  (constant)  destination
              port to connect. When using the tcptraceroute wrapper, -p speci‐
              fies the source port.

       -t tos, --tos=tos
              For IPv4, set the Type of Service (TOS)  and  Precedence  value.
              Useful  values  are 16 (low delay) and 8 (high throughput). Note
              that in order to use some TOS precedence values, you have to  be
              super user.
              For IPv6, set the Traffic Control value.

       -l flow_label, --flowlabel=flow_label
              Use specified flow_label for IPv6 packets.

       -w max[,here,near], --wait=max[,here,near]
              Determines how long to wait for a response to a probe.

              There  are  three (in general) float values separated by a comma
              (or a slash).  Max specifies the maximum time (in  seconds,  de‐
              fault 5.0) to wait, in any case.

              Traditional  traceroute  implementation  always waited whole max
              seconds for any probe. But if we already have some replies  from
              the  same  hop, or even from some next hop, we can use the round
              trip time of such a reply as a hint to determine the actual rea‐
              sonable amount of time to wait.

              The  optional  here (default 3.0) specifies a factor to multiply
              the round trip time of an already  received  response  from  the
              same  hop.  The  resulting  value  is  used as a timeout for the
              probe, instead of (but no more than)  max.   The  optional  near
              (default  10.0)  specifies  a similar factor for a response from
              some next hop.  (The time of the first found result is  used  in
              both cases).

              First,  we  look  for  the  same hop (of the probe which will be
              printed first from now).  If nothing found, then look  for  some
              next  hop.  If nothing found, use max.  If here and/or near have
              zero values, the corresponding computation is skipped.
              Here and near are always set to zero if only  max  is  specified
              (for compatibility with previous versions).

       -q nqueries, --queries=nqueries
              Sets the number of probe packets per hop. The default is 3.

       -r     Bypass  the normal routing tables and send directly to a host on
              an attached network.  If the host is not on a  directly-attached
              network,  an error is returned.  This option can be used to ping
              a local host through an interface that has no route through it.

       -s source_addr, --source=source_addr
              Chooses an alternative source address. Note that you must select
              the  address  of one of the interfaces.  By default, the address
              of the outgoing interface is used.

       -z sendwait, --sendwait=sendwait
              Minimal time interval between probes (default 0).  If the  value
              is  more  than  10,  then it specifies a number in milliseconds,
              else it is a number of seconds (float point values allowed too).
              Useful when some routers use rate-limit for ICMP messages.

       -e, --extensions
              Show  ICMP extensions (rfc4884). The general form is CLASS/TYPE:
              followed by a hexadecimal dump.  The  MPLS  (rfc4950)  is  shown
              parsed,  in  a form: MPLS:L=label,E=exp_use,S=stack_bottom,T=TTL
              (more objects separated by / ).

       -A, --as-path-lookups
              Perform AS path lookups in routing registries and print  results
              directly after the corresponding addresses.

       -V, --version
              Print the version and exit.

       There  are  additional options intended for advanced usage (such as al‐
       ternate trace methods etc.):

       --sport=port
              Chooses the source port to use. Implies  -N 1 -w 5  .   Normally
              source ports (if applicable) are chosen by the system.

       --fwmark=mark
              Set the firewall mark for outgoing packets (since the Linux ker‐
              nel 2.6.25).

       -M method, --module=name
              Use specified method for traceroute operations.  Default  tradi‐
              tional  udp method has name default, icmp (-I) and tcp (-T) have
              names icmp and tcp respectively.
              Method-specific options can be passed by -O .  Most methods have
              their simple shortcuts, (-I means -M icmp, etc).

       -O option, --options=options
              Specifies some method-specific option. Several options are sepa‐
              rated by comma (or use several -O on cmdline).  Each method  may
              have its own specific options, or many not have them at all.  To
              print information about available options, use -O help.

       -U, --udp
              Use UDP to particular destination port for tracerouting (instead
              of  increasing  the  port  per  each  probe). Default port is 53
              (dns).

       -UL    Use UDPLITE for tracerouting (default port is 53).

       -D, --dccp
              Use DCCP Requests for probes.

       -P protocol, --protocol=protocol
              Use raw packet of specified protocol for  tracerouting.  Default
              protocol is 253 (rfc3692).

       --mtu  Discover  MTU along the path being traced. Implies -F -N 1.  New
              mtu is printed once in a form of F=NUM at the first probe  of  a
              hop which requires such mtu to be reached. (Actually, the corre‐
              spond "frag needed" icmp message normally is sent by the  previ‐
              ous hop).

              Note, that some routers might cache once the seen information on
              a fragmentation. Thus you can  receive  the  final  mtu  from  a
              closer hop.  Try to specify an unusual tos by -t , this can help
              for one attempt (then it can be cached there as well).
              See -F option for more info.

       --back Print the number of backward hops when it seems  different  with
              the forward direction. This number is guessed in assumption that
              remote hops send reply packets with initial ttl  set  to  either
              64, or 128 or 255 (which seems a common practice). It is printed
              as a negate value in a form of '-NUM' .

LIST OF AVAILABLE METHODS
       In general, a particular traceroute method may have  to  be  chosen  by
       -M name,  but  most  of  the methods have their simple cmdline switches
       (you can see them after the method name, if present).

   default
       The traditional, ancient method of tracerouting. Used by default.

       Probe packets are udp datagrams with so-called  "unlikely"  destination
       ports.   The "unlikely" port of the first probe is 33434, then for each
       next probe it is incremented by one. Since the ports are expected to be
       unused,  the destination host normally returns "icmp unreach port" as a
       final response.  (Nobody knows what happens when some application  lis‐
       tens for such ports, though).

       This method is allowed for unprivileged users.

   icmp       -I
       Most usual method for now, which uses icmp echo packets for probes.
       If  you can ping(8) the destination host, icmp tracerouting is applica‐
       ble as well.

       This method may be allowed for unprivileged users since the kernel  3.0
       (IPv4,  for IPv6 since 3.11), which supports new dgram icmp (or "ping")
       sockets.   To   allow   such   sockets,   sysadmin    should    provide
       net/ipv4/ping_group_range sysctl range to match any group of the user.
       Options:

       raw    Use only raw sockets (the traditional way).
              This  way is tried first by default (for compatibility reasons),
              then new dgram icmp sockets as fallback.

       dgram  Use only dgram icmp sockets.

   tcp        -T
       Well-known modern method, intended to bypass firewalls.
       Uses the constant destination port (default is 80, http).

       If some filters are present in the network path, then most probably any
       "unlikely"  udp  ports  (as for default method) or even icmp echoes (as
       for icmp) are filtered, and whole tracerouting will just stop at such a
       firewall.  To bypass a network filter, we have to use only allowed pro‐
       tocol/port combinations. If we trace for some,  say,  mailserver,  then
       more likely -T -p 25 can reach it, even when -I can not.

       This  method  uses well-known "half-open technique", which prevents ap‐
       plications on the destination host from seeing our probes at all.  Nor‐
       mally,  a tcp syn is sent. For non-listened ports we receive tcp reset,
       and all is done. For active listening ports we receive tcp syn+ack, but
       answer  by tcp reset (instead of expected tcp ack), this way the remote
       tcp session is dropped even without the application ever taking notice.

       There is a couple of options for tcp method:

       syn,ack,fin,rst,psh,urg,ece,cwr
              Sets specified tcp flags for probe packet, in any combination.

       flags=num
              Sets the flags field in the tcp header exactly to num.

       ecn    Send syn packet with tcp flags ECE and CWR (for Explicit Conges‐
              tion Notification, rfc3168).

       sack,timestamps,window_scaling
              Use  the  corresponding  tcp header option in the outgoing probe
              packet.

       sysctl Use current sysctl (/proc/sys/net/*) setting for the tcp  header
              options  above  and ecn.  Always set by default, if nothing else
              specified.

       mss=num
              Use value of num for maxseg tcp header option (when syn).

       info   Print tcp flags of final tcp replies when  the  target  host  is
              reached.  Allows to determine whether an application listens the
              port and other useful things.

       Default options is syn,sysctl.

   tcpconn
       An initial implementation of tcp method, simple using connect(2)  call,
       which  does  full  tcp session opening. Not recommended for normal use,
       because a destination application is always affected (and can  be  con‐
       fused).

   udp        -U
       Use udp datagram with constant destination port (default 53, dns).
       Intended to bypass firewall as well.

       Note, that unlike in tcp method, the correspond application on the des‐
       tination host always receive our probes (with random  data),  and  most
       can  easily  be confused by them. Most cases it will not respond to our
       packets though, so we will never see the final hop in the trace.  (For‐
       tunately, it seems that at least dns servers replies with something an‐
       gry).

       This method is allowed for unprivileged users.

   udplite    -UL
       Use udplite datagram for probes (with constant  destination  port,  de‐
       fault 53).

       This method is allowed for unprivileged users.
       Options:

       coverage=num
              Set udplite send coverage to num.

   dccp    -D
       Use DCCP Request packets for probes (rfc4340).

       This  method  uses the same "half-open technique" as used for TCP.  The
       default destination port is 33434.

       Options:

       service=num
              Set DCCP service code to num (default is 1885957735).

   raw        -P proto
       Send raw packet of protocol proto.
       No protocol-specific headers are used, just IP header only.
       Implies -N 1 -w 5 .
       Options:

       protocol=proto
              Use IP protocol proto (default 253).

NOTES
       To speed up work, normally several probes are sent simultaneously.   On
       the other hand, it creates a "storm of packages", especially in the re‐
       ply direction. Routers can throttle the rate  of  icmp  responses,  and
       some  of replies can be lost. To avoid this, decrease the number of si‐
       multaneous probes, or even set it to 1 (like in initial traceroute  im‐
       plementation), i.e.  -N 1

       The  final  (target) host can drop some of the simultaneous probes, and
       might even answer only the latest ones. It can  lead  to  extra  "looks
       like  expired"  hops  near  the  final hop. We use a smart algorithm to
       auto-detect such a situation, but if it cannot help in your case,  just
       use -N 1 too.

       For  even  greater stability you can slow down the program's work by -z
       option, for example use -z 0.5 for half-second pause between probes.

       To avoid an extra waiting, we use adaptive algorithm for timeouts  (see
       -w  option  for more info). It can lead to premature expiry (especially
       when response times differ at times) and  printing  "*"  instead  of  a
       time.  In such a case, switch this algorithm off, by specifying -w with
       the desired timeout only (for example, -w 5).

       If some hops report nothing for every method, the last chance to obtain
       something  is  to  use  ping  -R  command (IPv4, and for nearest 8 hops
       only).

SEE ALSO
       ping(8), ping6(8), tcpdump(8), netstat(8)

Traceroute                      11 October 2006                  TRACEROUTE(1)