man ipfw(8)

march.jin

ipfw
FreeBSD Hypertext Man Pages


Man Page or Keyword Search:

Man
All Sections1 - General Commands2 - System Calls3 - Subroutines4 - Special Files5 - File Formats6 - Games7 - Macros and Conventions8 - Maintenance Commands9 - Kernel Interfacen - New Commands2.10 BSD2.11 BSD2.8 BSD2.9.1 BSD386BSD 0.0386BSD 0.14.3BSD NET/24.3BSD Reno4.4BSD Lite2Darwin 1.3.1/x86Darwin 1.4.1/x86Darwin 6.0.2/x86Darwin 7.0.1Darwin 8.0.1/ppcFreeBSD 1.0-RELEASEFreeBSD 1.1-RELEASEFreeBSD 1.1.5.1-RELEASEFreeBSD 2.0-RELEASEFreeBSD 2.0.5-RELEASEFreeBSD 2.1.0-RELEASEFreeBSD 2.1.5-RELEASEFreeBSD 2.1.6.1-RELEASEFreeBSD 2.1.7.1-RELEASEFreeBSD 2.2.1-RELEASEFreeBSD 2.2.2-RELEASEFreeBSD 2.2.5-RELEASEFreeBSD 2.2.6-RELEASEFreeBSD 2.2.7-RELEASEFreeBSD 2.2.8-RELEASEFreeBSD 3.0-RELEASEFreeBSD 3.1-RELEASEFreeBSD 3.2-RELEASEFreeBSD 3.3-RELEASEFreeBSD 3.4-RELEASEFreeBSD 3.5.1-RELEASEFreeBSD 4.0-RELEASEFreeBSD 4.1-RELEASEFreeBSD 4.1.1-RELEASEFreeBSD 4.10-RELEASEFreeBSD 4.11-RELEASEFreeBSD 4.2-RELEASEFreeBSD 4.3-RELEASEFreeBSD 4.4-RELEASEFreeBSD 4.5-RELEASEFreeBSD 4.6-RELEASEFreeBSD 4.6.2-RELEASEFreeBSD 4.7-RELEASEFreeBSD 4.8-RELEASEFreeBSD 4.9-RELEASEFreeBSD 5.0-RELEASEFreeBSD 5.1-RELEASEFreeBSD 5.2-RELEASEFreeBSD 5.2.1-RELEASEFreeBSD 5.3-RELEASEFreeBSD 5.4-RELEASEFreeBSD 5.5-RELEASEFreeBSD 5.5-stableFreeBSD 6.0-RELEASEFreeBSD 6.1-RELEASEFreeBSD 6.1-RELEASE and PortsFreeBSD 6.1-stableFreeBSD 6.2-RELEASEFreeBSD 7-currentFreeBSD Ports 4.7-RELEASEFreeBSD Ports 5.0-RELEASEFreeBSD Ports 5.1-RELEASEFreeBSD Ports 6.1-RELEASEHP-UX 10.01HP-UX 10.10HP-UX 10.20HP-UX 11.00HP-UX 11.11HP-UX 11.20HP-UX 11.22Linux Slackware 3.1Minix 2.0NetBSD 1.0NetBSD 1.1NetBSD 1.2NetBSD 1.2.1NetBSD 1.3NetBSD 1.3.1NetBSD 1.3.2NetBSD 1.3.3NetBSD 1.4NetBSD 1.4.1NetBSD 1.4.2NetBSD 1.4.3NetBSD 1.5NetBSD 1.5.1NetBSD 1.5.2NetBSD 1.5.3NetBSD 1.6NetBSD 1.6.1NetBSD 1.6.2NetBSD 2.0NetBSD 2.0.2NetBSD 2.1NetBSD 3.0OSF1 V4.0/alphaOSF1 V5.1/alphaOpenBSD 2.0OpenBSD 2.1OpenBSD 2.2OpenBSD 2.3OpenBSD 2.4OpenBSD 2.5OpenBSD 2.6OpenBSD 2.7OpenBSD 2.8OpenBSD 2.9OpenBSD 3.0OpenBSD 3.1OpenBSD 3.2OpenBSD 3.3OpenBSD 3.4OpenBSD 3.5OpenBSD 3.6OpenBSD 3.7OpenBSD 3.8OpenBSD 3.9OpenDarwin 20030208pre4/ppcOpenDarwin 6.6.1/x86OpenDarwin 6.6.2/x86OpenDarwin 7.2.1Plan 9Red Hat Linux/i386 4.2Red Hat Linux/i386 5.0Red Hat Linux/i386 5.2Red Hat Linux/i386 6.1Red Hat Linux/i386 6.2Red Hat Linux/i386 7.0Red Hat Linux/i386 7.1Red Hat Linux/i386 7.2Red Hat Linux/i386 7.3Red Hat Linux/i386 8.0Red Hat Linux/i386 9SuSE Linux/i386 4.3SuSE Linux/i386 5.0SuSE Linux/i386 5.2SuSE Linux/i386 5.3SuSE Linux/i386 6.0SuSE Linux/i386 6.1SuSE Linux/i386 6.3SuSE Linux/i386 6.4SuSE Linux/i386 7.0SuSE Linux/i386 7.1SuSE Linux/i386 7.2SuSE Linux/i386 7.3SuSE Linux/i386 8.0SuSE Linux/i386 8.1SuSE Linux/i386 8.2SunOS 4.1.3SunOS 5.10SunOS 5.5.1SunOS 5.6SunOS 5.7SunOS 5.8SunOS 5.9ULTRIX 4.2Unix Seventh EditionX11R6.7.0XFree86 3.3XFree86 3.3.6XFree86 4.0XFree86 4.0.1XFree86 4.0.2XFree86 4.1.0XFree86 4.2.0XFree86 4.2.99.3XFree86 4.3.0XFree86 4.4.0deutsch - Linux/GNU

Apropos
Keyword Search (all sections)htmlpspdfasciilatin1Output format
Index Page and Help
|
FAQ
|
Copyright
IPFW(8)                 FreeBSD System Manager's Manual                IPFW(8)
NAME
     ipfw -- IP firewall and traffic shaper control program
SYNOPSIS
     ipfw [-cq] add rule     ipfw [-acdefnNStT] {list | show} [rule | first-last ...]     ipfw [-f | -q] flush     ipfw [-q] {delete | zero | resetlog} [set] [number ...]     ipfw enable          {firewall | altq | one_pass | debug | verbose | dyn_keepalive}     ipfw disable          {firewall | altq | one_pass | debug | verbose | dyn_keepalive}     ipfw set [disable number ...] [enable number ...]     ipfw set move [rule] number to number     ipfw set swap number number     ipfw set show     ipfw table number add addr[/masklen] [value]     ipfw table number delete addr[/masklen]     ipfw table number flush     ipfw table number list     ipfw {pipe | queue} number config config-options     ipfw [-s [field]] {pipe | queue} {delete | list | show} [number ...]     ipfw [-cfnNqS] [-p preproc [preproc-flags]] pathname
DESCRIPTION
     The ipfw utility is the user interface for controlling the
ipfw(4)
fire-     wall and the
dummynet(4)
traffic shaper in FreeBSD.     An ipfw configuration, or ruleset, is made of a list of rules numbered     from 1 to 65535.  Packets are passed to ipfw from a number of different     places in the protocol stack (depending on the source and destination of     the packet, it is possible that ipfw is invoked multiple times on the     same packet).  The packet passed to the firewall is compared against each     of the rules in the firewall ruleset.  When a match is found, the action     corresponding to the matching rule is performed.     Depending on the action and certain system settings, packets can be rein-     jected into the firewall at some rule after the matching one for further     processing.     An ipfw ruleset always includes a default rule (numbered 65535) which     cannot be modified or deleted, and matches all packets.  The action asso-     ciated with the default rule can be either deny or allow depending on how     the kernel is configured.     If the ruleset includes one or more rules with the keep-state or limit     option, then ipfw assumes a stateful behaviour, i.e., upon a match it     will create dynamic rules matching the exact parameters (addresses and     ports) of the matching packet.     These dynamic rules, which have a limited lifetime, are checked at the     first occurrence of a check-state, keep-state or limit rule, and are typ-     ically used to open the firewall on-demand to legitimate traffic only.     See the STATEFUL FIREWALL and EXAMPLES Sections below for more informa-     tion on the stateful behaviour of ipfw.     All rules (including dynamic ones) have a few associated counters: a     packet count, a byte count, a log count and a timestamp indicating the     time of the last match.  Counters can be displayed or reset with ipfw     commands.     Rules can be added with the add command; deleted individually or in     groups with the delete command, and globally (except those in set 31)     with the flush command; displayed, optionally with the content of the     counters, using the show and list commands.  Finally, counters can be     reset with the zero and resetlog commands.     Also, each rule belongs to one of 32 different sets , and there are ipfw     commands to atomically manipulate sets, such as enable, disable, swap     sets, move all rules in a set to another one, delete all rules in a set.     These can be useful to install temporary configurations, or to test them.     See Section SETS OF RULES for more information on sets.     The following options are available:     -a      While listing, show counter values.  The show command just             implies this option.     -b      Only show the action and the comment, not the body of a rule.             Implies -c.     -c      When entering or showing rules, print them in compact form, i.e.,             without the optional "ip from any to any" string when this does             not carry any additional information.     -d      While listing, show dynamic rules in addition to static ones.     -e      While listing, if the -d option was specified, also show expired             dynamic rules.     -f      Do not ask for confirmation for commands that can cause problems             if misused, i.e. flush.  If there is no tty associated with the             process, this is implied.     -n      Only check syntax of the command strings, without actually pass-             ing them to the kernel.     -N      Try to resolve addresses and service names in output.     -q      While adding, zeroing, resetlogging or flushing, be quiet about             actions (implies -f).  This is useful for adjusting rules by exe-             cuting multiple ipfw commands in a script (e.g.,             `sh /etc/rc.firewall'), or by processing a file of many ipfw             rules across a remote login session.  If a flush is performed in             normal (verbose) mode (with the default kernel configuration), it             prints a message.        Because all rules are flushed, the message             might not be delivered to the login session, causing the remote             login session to be closed and the remainder of the ruleset to             not be processed.        Access to the console would then be required             to recover.     -S      While listing rules, show the set each rule belongs to.  If this             flag is not specified, disabled rules will not be listed.     -s [field]             While listing pipes, sort according to one of the four counters             (total or current packets or bytes).     -t      While listing, show last match timestamp (converted with             ctime()).     -T      While listing, show last match timestamp (as seconds from the             epoch).  This form can be more convenient for postprocessing by             scripts.     To ease configuration, rules can be put into a file which is processed     using ipfw as shown in the last synopsis line.  An absolute pathname must     be used.  The file will be read line by line and applied as arguments to     the ipfw utility.     Optionally, a preprocessor can be specified using -p preproc where     pathname is to be piped through.  Useful preprocessors include
cpp(1)
and     
m4(1)
.  If preproc does not start with a slash (`/') as its first charac-     ter, the usual PATH name search is performed.  Care should be taken with     this in environments where not all file systems are mounted (yet) by the     time ipfw is being run (e.g. when they are mounted over NFS).  Once -p     has been specified, any additional arguments as passed on to the pre-     processor for interpretation.  This allows for flexible configuration     files (like conditionalizing them on the local hostname) and the use of     macros to centralize frequently required arguments like IP addresses.     The ipfw pipe and queue commands are used to configure the traffic     shaper, as shown in the TRAFFIC SHAPER (DUMMYNET) CONFIGURATION Section     below.     If the world and the kernel get out of sync the ipfw ABI may break, pre-     venting you from being able to add any rules.  This can adversely effect     the booting process.  You can use ipfw disable firewall to temporarily     disable the firewall to regain access to the network, allowing you to fix     the problem.
PACKET FLOW
     A packet is checked against the active ruleset in multiple places in the     protocol stack, under control of several sysctl variables.  These places     and variables are shown below, and it is important to have this picture     in mind in order to design a correct ruleset.                  ^    to upper layers          V                  |                          |                  +----------->-----------+                  ^                          V            [
ip(6)
_input]            [
ip(6)
_output]     net.inet.ip.fw.enable=1                  |                          |                  ^                          V            [ether_demux]         [ether_output_frame]  net.link.ether.ipfw=1                  |                          |                  +-->--[bdg_forward]-->--+               net.link.ether.bridge_ipfw=1                  ^                          V                  |         to devices          |     As can be noted from the above picture, the number of times the same     packet goes through the firewall can vary between 0 and 4 depending on     packet source and destination, and system configuration.     Note that as packets flow through the stack, headers can be stripped or     added to it, and so they may or may not be available for inspection.     E.g., incoming packets will include the MAC header when ipfw is invoked     from ether_demux(), but the same packets will have the MAC header     stripped off when ipfw is invoked from ip_input() or ip6_input().     Also note that each packet is always checked against the complete rule-     set, irrespective of the place where the check occurs, or the source of     the packet.  If a rule contains some match patterns or actions which are     not valid for the place of invocation (e.g. trying to match a MAC header     within ip_input or ip6_input ), the match pattern will not match, but a     not operator in front of such patterns will cause the pattern to always     match on those packets.  It is thus the responsibility of the programmer,     if necessary, to write a suitable ruleset to differentiate among the pos-     sible places.  skipto rules can be useful here, as an example:           # packets from ether_demux or bdg_forward           ipfw add 10 skipto 1000 all from any to any layer2 in           # packets from ip_input           ipfw add 10 skipto 2000 all from any to any not layer2 in           # packets from ip_output           ipfw add 10 skipto 3000 all from any to any not layer2 out           # packets from ether_output_frame           ipfw add 10 skipto 4000 all from any to any layer2 out     (yes, at the moment there is no way to differentiate between ether_demux     and bdg_forward).
SYNTAX
     In general, each keyword or argument must be provided as a separate com-     mand line argument, with no leading or trailing spaces.  Keywords are     case-sensitive, whereas arguments may or may not be case-sensitive     depending on their nature (e.g. uid's are, hostnames are not).     In ipfw2 you can introduce spaces after commas ',' to make the line more     readable.        You can also put the entire command (including flags) into a     single argument.  E.g., the following forms are equivalent:           ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8           ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8           ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
RULE FORMAT
     The format of ipfw rules is the following:           [rule_number] [set set_number] [prob match_probability]               action [log [logamount number]] [altq queue] body     where the body of the rule specifies which information is used for fil-     tering packets, among the following:        Layer-2 header fields                      When available        IPv4 and IPv6 Protocol                      TCP, UDP, ICMP, etc.        Source and dest. addresses and ports        Direction                              See Section PACKET FLOW        Transmit and receive interface              By name or address        Misc. IP header fields                      Version, type of service, data-                                              gram length, identification,                                              fragment flag (non-zero IP off-                                              set), Time To Live        IP options        IPv6 Extension headers                      Fragmentation, Hop-by-Hop                                              options, source routing, IPSec                                              options.        IPv6 Flow-ID        Misc. TCP header fields               TCP flags (SYN, FIN, ACK, RST,                                              etc.), sequence number, acknowl-                                              edgment number, window        TCP options        ICMP types                              for ICMP packets        ICMP6 types                              for ICMP6 packets        User/group ID                              When the packet can be associ-                                              ated with a local socket.        Divert status                              Whether a packet came from a                                              divert socket (e.g.,
natd(8)
).     Note that some of the above information, e.g. source MAC or IP addresses     and TCP/UDP ports, could easily be spoofed, so filtering on those fields     alone might not guarantee the desired results.     rule_number             Each rule is associated with a rule_number in the range 1..65535,             with the latter reserved for the default rule.  Rules are checked             sequentially by rule number.  Multiple rules can have the same             number, in which case they are checked (and listed) according to             the order in which they have been added.  If a rule is entered             without specifying a number, the kernel will assign one in such a             way that the rule becomes the last one before the default rule.             Automatic rule numbers are assigned by incrementing the last non-             default rule number by the value of the sysctl variable             net.inet.ip.fw.autoinc_step which defaults to 100.  If this is             not possible (e.g. because we would go beyond the maximum allowed             rule number), the number of the last non-default value is used             instead.     set set_number             Each rule is associated with a set_number in the range 0..31.             Sets can be individually disabled and enabled, so this parameter             is of fundamental importance for atomic ruleset manipulation.  It             can be also used to simplify deletion of groups of rules.        If a             rule is entered without specifying a set number, set 0 will be             used.             Set 31 is special in that it cannot be disabled, and rules in set             31 are not deleted by the ipfw flush command (but you can delete             them with the ipfw delete set 31 command).  Set 31 is also used             for the default rule.     prob match_probability             A match is only declared with the specified probability (floating             point number between 0 and 1).  This can be useful for a number             of applications such as random packet drop or (in conjunction             with
dummynet(4)
) to simulate the effect of multiple paths lead-             ing to out-of-order packet delivery.             Note: this condition is checked before any other condition,             including ones such as keep-state or check-state which might have             side effects.     log [logamount number]             When a packet matches a rule with the log keyword, a message will             be logged to
syslogd(8)
with a LOG_SECURITY facility.  The log-             ging only occurs if the sysctl variable net.inet.ip.fw.verbose is             set to 1 (which is the default when the kernel is compiled with             IPFIREWALL_VERBOSE) and the number of packets logged so far for             that particular rule does not exceed the logamount parameter.  If             no logamount is specified, the limit is taken from the sysctl             variable net.inet.ip.fw.verbose_limit.  In both cases, a value of             0 removes the logging limit.             Once the limit is reached, logging can be re-enabled by clearing             the logging counter or the packet counter for that entry, see the             resetlog command.             Note: logging is done after all other packet matching conditions             have been successfully verified, and before performing the final             action (accept, deny, etc.) on the packet.     altq queue             When a packet matches a rule with the altq keyword, the ALTQ             identifier for the given queue (see
altq(4)
) will be attached.             Note that this ALTQ tag is only meaningful for packets going             "out" of IPFW, and not being rejected or going to divert sockets.             Note that if there is insufficient memory at the time the packet             is processed, it will not be tagged, so it is wise to make your             ALTQ "default" queue policy account for this.  If multiple altq             rules match a single packet, only the first one adds the ALTQ             classification tag.  In doing so, traffic may be shaped by using             count altq queue rules for classification early in the ruleset,             then later applying the filtering decision.  For example,             check-state and keep-state rules may come later and provide the             actual filtering decisions in addition to the fallback ALTQ tag.             You must run
pfctl(8)
to set up the queues before IPFW will be             able to look them up by name, and if the ALTQ disciplines are             rearranged, the rules in containing the queue identifiers in the             kernel will likely have gone stale and need to be reloaded.             Stale queue identifiers will probably result in misclassifica-             tion.             All system ALTQ processing can be turned on or off via ipfw             enable altq and ipfw disable altq.  The usage of             net.inet.ip.fw.one_pass is irrelevant to ALTQ traffic shaping, as             the actual rule action is followed always after adding an ALTQ             tag.   RULE ACTIONS     A rule can be associated with one of the following actions, which will be     executed when the packet matches the body of the rule.     allow | accept | pass | permit             Allow packets that match rule.  The search terminates.     check-state             Checks the packet against the dynamic ruleset.  If a match is             found, execute the action associated with the rule which gener-             ated this dynamic rule, otherwise move to the next rule.             Check-state rules do not have a body.  If no check-state rule is             found, the dynamic ruleset is checked at the first keep-state or             limit rule.     count   Update counters for all packets that match rule.  The search con-             tinues with the next rule.     deny | drop             Discard packets that match this rule.  The search terminates.     divert port             Divert packets that match this rule to the
divert(4)
socket bound             to port port.  The search terminates.     fwd | forward ipaddr[,port]             Change the next-hop on matching packets to ipaddr, which can be             an IP address or a host name.  The search terminates if this rule             matches.             If ipaddr is a local address, then matching packets will be for-             warded to port (or the port number in the packet if one is not             specified in the rule) on the local machine.             If ipaddr is not a local address, then the port number (if speci-             fied) is ignored, and the packet will be forwarded to the remote             address, using the route as found in the local routing table for             that IP.             A fwd rule will not match layer-2 packets (those received on             ether_input, ether_output, or bridged).             The fwd action does not change the contents of the packet at all.             In particular, the destination address remains unmodified, so             packets forwarded to another system will usually be rejected by             that system unless there is a matching rule on that system to             capture them.  For packets forwarded locally, the local address             of the socket will be set to the original destination address of             the packet.  This makes the
netstat(1)
entry look rather weird             but is intended for use with transparent proxy servers.             To enable fwd a custom kernel needs to be compiled with the             option options IPFIREWALL_FORWARD.  With the additional option             options IPFIREWALL_FORWARD_EXTENDED all safeguards are removed             and it also makes it possible to redirect packets destined to             locally configured IP addresses.  Please note that such rules             apply to locally generated packets as well and great care is             required to ensure proper behaviour for automatically generated             packets like ICMP message size exceeded and others.     pipe pipe_nr             Pass packet to a
dummynet(4)
``pipe'' (for bandwidth limitation,             delay, etc.).  See the TRAFFIC SHAPER (DUMMYNET) CONFIGURATION             Section for further information.  The search terminates; however,             on exit from the pipe and if the
sysctl(8)
variable             net.inet.ip.fw.one_pass is not set, the packet is passed again to             the firewall code starting from the next rule.     queue queue_nr             Pass packet to a
dummynet(4)
``queue'' (for bandwidth limitation             using WF2Q+).     reject  (Deprecated).  Synonym for unreach host.     reset   Discard packets that match this rule, and if the packet is a TCP             packet, try to send a TCP reset (RST) notice.  The search termi-             nates.     reset6  Discard packets that match this rule, and if the packet is a TCP             packet, try to send a TCP reset (RST) notice.  The search termi-             nates.     skipto number             Skip all subsequent rules numbered less than number.  The search             continues with the first rule numbered number or higher.     tee port             Send a copy of packets matching this rule to the
divert(4)
socket             bound to port port.  The search continues with the next rule.     unreach code             Discard packets that match this rule, and try to send an ICMP             unreachable notice with code code, where code is a number from 0             to 255, or one of these aliases: net, host, protocol, port,             needfrag, srcfail, net-unknown, host-unknown, isolated,             net-prohib, host-prohib, tosnet, toshost, filter-prohib,             host-precedence or precedence-cutoff.  The search terminates.     unreach6 code             Discard packets that match this rule, and try to send an ICMPv6             unreachable notice with code code, where code is a number from 0,             1, 3 or 4, or one of these aliases: no-route, admin-prohib,             address or port.  The search terminates.     netgraph cookie             Divert packet into netgraph with given cookie.  The search termi-             nates.  If packet is later returned from netgraph it is either             accepted or continues with the next rule, depending on             net.inet.ip.fw.one_pass sysctl variable.     ngtee cookie             A copy of packet is diverted into netgraph, original packet is             either accepted or continues with the next rule, depending on             net.inet.ip.fw.one_pass sysctl variable.  See
ng_ipfw(4)
for more             information on netgraph and ngtee actions.   RULE BODY     The body of a rule contains zero or more patterns (such as specific     source and destination addresses or ports, protocol options, incoming or     outgoing interfaces, etc.)  that the packet must match in order to be     recognised.  In general, the patterns are connected by (implicit) and     operators -- i.e., all must match in order for the rule to match.        Indi-     vidual patterns can be prefixed by the not operator to reverse the result     of the match, as in           ipfw add 100 allow ip from not 1.2.3.4 to any     Additionally, sets of alternative match patterns (or-blocks) can be con-     structed by putting the patterns in lists enclosed between parentheses (     ) or braces { }, and using the or operator as follows:           ipfw add 100 allow ip from { x or not y or z } to any     Only one level of parentheses is allowed.        Beware that most shells have     special meanings for parentheses or braces, so it is advisable to put a     backslash \ in front of them to prevent such interpretations.     The body of a rule must in general include a source and destination     address specifier.  The keyword any can be used in various places to     specify that the content of a required field is irrelevant.     The rule body has the following format:           [proto from src to dst] [options]     The first part (proto from src to dst) is for backward compatibility with     earlier versions of FreeBSD.  In modern FreeBSD any match pattern     (including MAC headers, IP protocols, addresses and ports) can be speci-     fied in the options section.     Rule fields have the following meaning:     proto: protocol | { protocol or ... }     protocol: [not] protocol-name | protocol-number             An IP protocol specified by number or name (for a complete list             see /etc/protocols), or one of the following keywords:             ip4 | ipv4                     Matches IPv4 packets.             ip6 | ipv6                     Matches IPv6 packets.             ip | all                     Matches any packet.             The ipv6 in proto option will be treated as inner protocol.  And,             the ipv4 is not available in proto option.             The { protocol or ... } format (an or-block) is provided for con-             venience only but its use is deprecated.     src and dst: {addr | { addr or ... }} [[not] ports]             An address (or a list, see below) optionally followed by ports             specifiers.             The second format (or-block with multiple addresses) is provided             for convenience only and its use is discouraged.     addr: [not] {any | me | me6 table(number[,value]) | addr-list | addr-set}     any     matches any IP address.     me      matches any IP address configured on an interface in the system.     me6     matches any IPv6 address configured on an interface in the sys-             tem.  The address list is evaluated at the time the packet is an-             alysed.     table(number[,value])             Matches any IPv4 address for which an entry exists in the lookup             table number.  If an optional 32-bit unsigned value is also spec-             ified, an entry will match only if it has this value.  See the             LOOKUP TABLES section below for more information on lookup             tables.     addr-list: ip-addr[,addr-list]     ip-addr:             A host or subnet address specified in one of the following ways:             numeric-ip | hostname                     Matches a single IPv4 address, specified as dotted-quad                     or a hostname.  Hostnames are resolved at the time the                     rule is added to the firewall list.             addr/masklen                     Matches all addresses with base addr (specified as an IP                     address or a hostname) and mask width of masklen bits.                     As an example, 1.2.3.4/25 will match all IP numbers from                     1.2.3.0 to 1.2.3.127 .             addr:mask                     Matches all addresses with base addr (specified as an IP                     address or a hostname) and the mask of mask, specified as                     a dotted quad.  As an example, 1.2.3.4:255.0.255.0 will                     match 1.*.3.*.  This form is advised only for non-con-                     tiguous masks.  It is better to resort to the                     addr/masklen format for contiguous masks, which is more                     compact and less error-prone.     addr-set: addr[/masklen]{list}     list: {num | num-num}[,list]             Matches all addresses with base address addr (specified as an IP             address or a hostname) and whose last byte is in the list between             braces { } .  Note that there must be no spaces between braces             and numbers (spaces after commas are allowed).  Elements of the             list can be specified as single entries or ranges.  The masklen             field is used to limit the size of the set of addresses, and can             have any value between 24 and 32.        If not specified, it will be             assumed as 24.             This format is particularly useful to handle sparse address sets             within a single rule.  Because the matching occurs using a bit-             mask, it takes constant time and dramatically reduces the com-             plexity of rulesets.             As an example, an address specified as 1.2.3.4/24{128,35-55,89}             will match the following IP addresses:             1.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .     addr6-list: ip6-addr[,addr6-list]     ip6-addr:             A host or subnet specified one of the following ways:             numeric-ip | hostname                     Matches a single IPv6 address as allowed by
inet_pton(3)
                     or a hostname.  Hostnames are resolved at the time the                     rule is added to the firewall list.             addr/masklen                     Matches all IPv6 addresses with base addr (specified as                     allowed by inet_pton or a hostname) and mask width of                     masklen bits.             No support for sets of IPv6 addresses is provided because IPv6             addresses are typically random past the initial prefix.     ports: {port | port-port}[,ports]             For protocols which support port numbers (such as TCP and UDP),             optional ports may be specified as one or more ports or port             ranges, separated by commas but no spaces, and an optional not             operator.        The `-' notation specifies a range of ports (including             boundaries).             Service names (from /etc/services) may be used instead of numeric             port values.  The length of the port list is limited to 30 ports             or ranges, though one can specify larger ranges by using an             or-block in the options section of the rule.             A backslash (`\') can be used to escape the dash (`-') character             in a service name (from a shell, the backslash must be typed             twice to avoid the shell itself interpreting it as an escape             character).                   ipfw add count tcp from any ftp\\-data-ftp to any             Fragmented packets which have a non-zero offset (i.e., not the             first fragment) will never match a rule which has one or more             port specifications.  See the frag option for details on matching             fragmented packets.   RULE OPTIONS (MATCH PATTERNS)     Additional match patterns can be used within rules.  Zero or more of     these so-called options can be present in a rule, optionally prefixed by     the not operand, and possibly grouped into or-blocks.     The following match patterns can be used (listed in alphabetical order):     // this is a comment.             Inserts the specified text as a comment in the rule.  Everything             following // is considered as a comment and stored in the rule.             You can have comment-only rules, which are listed as having a             count action followed by the comment.     bridged             Alias for layer2.     diverted             Matches only packets generated by a divert socket.     diverted-loopback             Matches only packets coming from a divert socket back into the IP             stack input for delivery.     diverted-output             Matches only packets going from a divert socket back outward to             the IP stack output for delivery.     dst-ip ip-address             Matches IPv4 packets whose destination IP is one of the             address(es) specified as argument.     {dst-ip6 | dst-ipv6} ip6-address             Matches IPv6 packets whose destination IP is one of the             address(es) specified as argument.     dst-port ports             Matches IP packets whose destination port is one of the port(s)             specified as argument.     established             Matches TCP packets that have the RST or ACK bits set.     ext6hdr header             Matches IPv6 packets containing the extended header given by             header.  Supported headers are:             Fragment, (frag), Hop-to-hop options (hopopt), Source routing             (route), Destination options (dstopt), IPSec authentication head-             ers (ah), and IPSec encapsulated security payload headers (esp).     flow-id labels             Matches IPv6 packets containing any of the flow labels given in             labels.  labels is a comma seperate list of numeric flow labels.     frag    Matches packets that are fragments and not the first fragment of             an IP datagram.  Note that these packets will not have the next             protocol header (e.g. TCP, UDP) so options that look into these             headers cannot match.     gid group             Matches all TCP or UDP packets sent by or received for a group.             A group may be specified by name or number.  This option should             be used only if debug.mpsafenet=0 to avoid possible deadlocks due             to layering violations in its implementation.     jail prisonID             Matches all TCP or UDP packets sent by or received for the jail             whos prison ID is prisonID.  This option should be used only if             debug.mpsafenet=0 to avoid possible deadlocks due to layering             violations in its implementation.     icmptypes types             Matches ICMP packets whose ICMP type is in the list types.  The             list may be specified as any combination of individual types             (numeric) separated by commas.  Ranges are not allowed. The sup-             ported ICMP types are:             echo reply (0), destination unreachable (3), source quench (4),             redirect (5), echo request (8), router advertisement (9), router             solicitation (10), time-to-live exceeded (11), IP header bad             (12), timestamp request (13), timestamp reply (14), information             request (15), information reply (16), address mask request (17)             and address mask reply (18).     icmp6types types             Matches ICMP6 packets whose ICMP6 type is in the list of types.             The list may be specified as any combination of individual types             (numeric) separated by commas.  Ranges are not allowed.     in | out             Matches incoming or outgoing packets, respectively.  in and out             are mutually exclusive (in fact, out is implemented as not in).     ipid id-list             Matches IPv4 packets whose ip_id field has value included in             id-list, which is either a single value or a list of values or             ranges specified in the same way as ports.     iplen len-list             Matches IP packets whose total length, including header and data,             is in the set len-list, which is either a single value or a list             of values or ranges specified in the same way as ports.     ipoptions spec             Matches packets whose IPv4 header contains the comma separated             list of options specified in spec.  The supported IP options are:             ssrr (strict source route), lsrr (loose source route), rr (record             packet route) and ts (timestamp).        The absence of a particular             option may be denoted with a `!'.     ipprecedence precedence             Matches IPv4 packets whose precedence field is equal to             precedence.     ipsec   Matches packets that have IPSEC history associated with them             (i.e., the packet comes encapsulated in IPSEC, the kernel has             IPSEC support and IPSEC_FILTERGIF option, and can correctly             decapsulate it).             Note that specifying ipsec is different from specifying proto             ipsec as the latter will only look at the specific IP protocol             field, irrespective of IPSEC kernel support and the validity of             the IPSEC data.             Further note that this flag is silently ignored in kernels with-             out IPSEC support.  It does not affect rule processing when given             and the rules are handled as if with no ipsec flag.     iptos spec             Matches IPv4 packets whose tos field contains the comma separated             list of service types specified in spec.  The supported IP types             of service are:             lowdelay (IPTOS_LOWDELAY), throughput (IPTOS_THROUGHPUT),             reliability (IPTOS_RELIABILITY), mincost (IPTOS_MINCOST),             congestion (IPTOS_CE).  The absence of a particular type may be             denoted with a `!'.     ipttl ttl-list             Matches IPv4 packets whose time to live is included in ttl-list,             which is either a single value or a list of values or ranges             specified in the same way as ports.     ipversion ver             Matches IP packets whose IP version field is ver.     keep-state             Upon a match, the firewall will create a dynamic rule, whose             default behaviour is to match bidirectional traffic between             source and destination IP/port using the same protocol.  The rule             has a limited lifetime (controlled by a set of
sysctl(8)
vari-             ables), and the lifetime is refreshed every time a matching             packet is found.     layer2  Matches only layer2 packets, i.e., those passed to ipfw from             ether_demux() and ether_output_frame().     limit {src-addr | src-port | dst-addr | dst-port} N             The firewall will only allow N connections with the same set of             parameters as specified in the rule.  One or more of source and             destination addresses and ports can be specified.        Currently,             only IPv4 flows are supported.     { MAC | mac } dst-mac src-mac             Match packets with a given dst-mac and src-mac addresses, speci-             fied as the any keyword (matching any MAC address), or six groups             of hex digits separated by colons, and optionally followed by a             mask indicating the significant bits.  The mask may be specified             using either of the following methods:             1.      A slash (/) followed by the number of significant bits.                     For example, an address with 33 significant bits could be                     specified as:                           MAC 10:20:30:40:50:60/33 any             2.      An ampersand (&) followed by a bitmask specified as six                     groups of hex digits separated by colons.        For example,                     an address in which the last 16 bits are significant                     could be specified as:                           MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any                     Note that the ampersand character has a special meaning                     in many shells and should generally be escaped.             Note that the order of MAC addresses (destination first, source             second) is the same as on the wire, but the opposite of the one             used for IP addresses.     mac-type mac-type             Matches packets whose Ethernet Type field corresponds to one of             those specified as argument.  mac-type is specified in the same             way as port numbers (i.e., one or more comma-separated single             values or ranges).  You can use symbolic names for known values             such as vlan, ipv4, ipv6.        Values can be entered as decimal or             hexadecimal (if prefixed by 0x), and they are always printed as             hexadecimal (unless the -N option is used, in which case symbolic             resolution will be attempted).     proto protocol             Matches packets with the corresponding IP protocol.     recv | xmit | via {ifX | if* | ipno | any}             Matches packets received, transmitted or going through, respec-             tively, the interface specified by exact name (ifX), by device             name (if*), by IP address, or through some interface.             The via keyword causes the interface to always be checked.  If             recv or xmit is used instead of via, then only the receive or             transmit interface (respectively) is checked.  By specifying             both, it is possible to match packets based on both receive and             transmit interface, e.g.:                   ipfw add deny ip from any to any out recv ed0 xmit ed1             The recv interface can be tested on either incoming or outgoing             packets, while the xmit interface can only be tested on outgoing             packets.  So out is required (and in is invalid) whenever xmit is             used.             A packet may not have a receive or transmit interface: packets             originating from the local host have no receive interface, while             packets destined for the local host have no transmit interface.     setup   Matches TCP packets that have the SYN bit set but no ACK bit.             This is the short form of ``tcpflags syn,!ack''.     src-ip ip-address             Matches IPv4 packets whose source IP is one of the address(es)             specified as an argument.     src-ip6 ip6-address             Matches IPv6 packets whose source IP is one of the address(es)             specified as an argument.     src-port ports             Matches IP packets whose source port is one of the port(s) speci-             fied as argument.     tcpack ack             TCP packets only.        Match if the TCP header acknowledgment number             field is set to ack.     tcpdatalen tcpdatalen-list             Matches TCP packets whose length of TCP data is tcpdatalen-list,             which is either a single value or a list of values or ranges             specified in the same way as ports.     tcpflags spec             TCP packets only.        Match if the TCP header contains the comma             separated list of flags specified in spec.  The supported TCP             flags are:             fin, syn, rst, psh, ack and urg.  The absence of a particular             flag may be denoted with a `!'.  A rule which contains a tcpflags             specification can never match a fragmented packet which has a             non-zero offset.  See the frag option for details on matching             fragmented packets.     tcpseq seq             TCP packets only.        Match if the TCP header sequence number field             is set to seq.     tcpwin win             TCP packets only.        Match if the TCP header window field is set to             win.     tcpoptions spec             TCP packets only.        Match if the TCP header contains the comma             separated list of options specified in spec.  The supported TCP             options are:             mss (maximum segment size), window (tcp window advertisement),             sack (selective ack), ts (rfc1323 timestamp) and cc (rfc1644             t/tcp connection count).  The absence of a particular option may             be denoted with a `!'.     uid user             Match all TCP or UDP packets sent by or received for a user.  A             user may be matched by name or identification number.  This             option should be used only if debug.mpsafenet=0 to avoid possible             deadlocks due to layering violations in its implementation.     verrevpath             For incoming packets, a routing table lookup is done on the             packet's source address.  If the interface on which the packet             entered the system matches the outgoing interface for the route,             the packet matches.  If the interfaces do not match up, the             packet does not match.  All outgoing packets or packets with no             incoming interface match.             The name and functionality of the option is intentionally similar             to the Cisco IOS command:                   ip verify unicast reverse-path             This option can be used to make anti-spoofing rules to reject all             packets with source addresses not from this interface.  See also             the option antispoof.     versrcreach             For incoming packets, a routing table lookup is done on the             packet's source address.  If a route to the source address             exists, but not the default route or a blackhole/reject route,             the packet matches.  Otherwise, the packet does not match.  All             outgoing packets match.             The name and functionality of the option is intentionally similar             to the Cisco IOS command:                   ip verify unicast source reachable-via any             This option can be used to make anti-spoofing rules to reject all             packets whose source address is unreachable.     antispoof             For incoming packets, the packet's source address is checked if             it belongs to a directly connected network.  If the network is             directly connected, then the interface the packet came on in is             compared to the interface the network is connected to.  When             incoming interface and directly connected interface are not the             same, the packet does not match.  Otherwise, the packet does             match.  All outgoing packets match.             This option can be used to make anti-spoofing rules to reject all             packets that pretend to be from a directly connected network but             do not come in through that interface.  This option is similar to             but more restricted than verrevpath because it engages only on             packets with source addresses of directly connected networks             instead of all source addresses.
LOOKUP TABLES
     Lookup tables are useful to handle large sparse address sets, typically     from a hundred to several thousands of entries.  There may be up to 128     different lookup tables, numbered 0 to 127.     Each entry is represented by an addr[/masklen] and will match all     addresses with base addr (specified as an IP address or a hostname) and     mask width of masklen bits.  If masklen is not specified, it defaults to     32.  When looking up an IP address in a table, the most specific entry     will match.  Associated with each entry is a 32-bit unsigned value, which     can optionally be checked by a rule matching code.  When adding an entry,     if value is not specified, it defaults to 0.     An entry can be added to a table (add), removed from a table (delete), a     table can be examined (list) or flushed (flush).     Internally, each table is stored in a Radix tree, the same way as the     routing table (see
route(4)
).     Lookup tables currently support IPv4 addresses only.     The tablearg feature provides the ability to use a value, looked up in     the table, as the argument for a rule action.  This can significantly     reduce number of rules in some configurations.  The tablearg argument can     be used with the following actions: pipe, queue, divert, tee, netgraph,     ngtee.  See the EXAMPLES Section for example usage of tables and the     tablearg keyword.
SETS OF
RULES     Each rule belongs to one of 32 different sets , numbered 0 to 31.        Set 31     is reserved for the default rule.     By default, rules are put in set 0, unless you use the set N attribute     when entering a new rule.        Sets can be individually and atomically     enabled or disabled, so this mechanism permits an easy way to store mul-     tiple configurations of the firewall and quickly (and atomically) switch     between them.  The command to enable/disable sets is           ipfw set [disable number ...] [enable number ...]     where multiple enable or disable sections can be specified.  Command exe-     cution is atomic on all the sets specified in the command.  By default,     all sets are enabled.     When you disable a set, its rules behave as if they do not exist in the     firewall configuration, with only one exception:           dynamic rules created from a rule before it had been disabled will           still be active until they expire.  In order to delete dynamic           rules you have to explicitly delete the parent rule which generated           them.     The set number of rules can be changed with the command           ipfw set move {rule rule-number | old-set} to new-set     Also, you can atomically swap two rulesets with the command           ipfw set swap first-set second-set     See the EXAMPLES Section on some possible uses of sets of rules.
STATEFUL FIREWALL
     Stateful operation is a way for the firewall to dynamically create rules     for specific flows when packets that match a given pattern are detected.     Support for stateful operation comes through the check-state, keep-state     and limit options of rules.     Dynamic rules are created when a packet matches a keep-state or limit     rule, causing the creation of a dynamic rule which will match all and     only packets with a given protocol between a src-ip/src-port     dst-ip/dst-port pair of addresses (src and dst are used here only to     denote the initial match addresses, but they are completely equivalent     afterwards).  Dynamic rules will be checked at the first check-state,     keep-state or limit occurrence, and the action performed upon a match     will be the same as in the parent rule.     Note that no additional attributes other than protocol and IP addresses     and ports are checked on dynamic rules.     The typical use of dynamic rules is to keep a closed firewall configura-     tion, but let the first TCP SYN packet from the inside network install a     dynamic rule for the flow so that packets belonging to that session will     be allowed through the firewall:           ipfw add check-state           ipfw add allow tcp from my-subnet to any setup keep-state           ipfw add deny tcp from any to any     A similar approach can be used for UDP, where an UDP packet coming from     the inside will install a dynamic rule to let the response through the     firewall:           ipfw add check-state           ipfw add allow udp from my-subnet to any keep-state           ipfw add deny udp from any to any     Dynamic rules expire after some time, which depends on the status of the     flow and the setting of some sysctl variables.  See Section SYSCTL     VARIABLES for more details.  For TCP sessions, dynamic rules can be     instructed to periodically send keepalive packets to refresh the state of     the rule when it is about to expire.     See Section EXAMPLES for more examples on how to use dynamic rules.
TRAFFIC SHAPER
(DUMMYNET) CONFIGURATION     ipfw is also the user interface for the
dummynet(4)
traffic shaper.     dummynet operates by first using the firewall to classify packets and     divide them into flows, using any match pattern that can be used in ipfw     rules.  Depending on local policies, a flow can contain packets for a     single TCP connection, or from/to a given host, or entire subnet, or a     protocol type, etc.     Packets belonging to the same flow are then passed to either of two dif-     ferent objects, which implement the traffic regulation:         pipe         A pipe emulates a link with given bandwidth, propagation                 delay, queue size and packet loss rate.  Packets are queued                 in front of the pipe as they come out from the classifier,                 and then transferred to the pipe according to the pipe's                 parameters.         queue         A queue is an abstraction used to implement the WF2Q+ (Worst-                 case Fair Weighted Fair Queueing) policy, which is an effi-                 cient variant of the WFQ policy.                 The queue associates a weight and a reference pipe to each                 flow, and then all backlogged (i.e., with packets queued)                 flows linked to the same pipe share the pipe's bandwidth pro-                 portionally to their weights.        Note that weights are not pri-                 orities; a flow with a lower weight is still guaranteed to                 get its fraction of the bandwidth even if a flow with a                 higher weight is permanently backlogged.     In practice, pipes can be used to set hard limits to the bandwidth that a     flow can use, whereas queues can be used to determine how different flow     share the available bandwidth.     The pipe and queue configuration commands are the following:           pipe number config pipe-configuration           queue number config queue-configuration     The following parameters can be configured for a pipe:     bw bandwidth | device             Bandwidth, measured in [K|M]{bit/s|Byte/s}.             A value of 0 (default) means unlimited bandwidth.        The unit must             immediately follow the number, as in                   ipfw pipe 1 config bw 300Kbit/s             If a device name is specified instead of a numeric value, as in                   ipfw pipe 1 config bw tun0             then the transmit clock is supplied by the specified device.  At             the moment only the
tun(4)
device supports this functionality,             for use in conjunction with
ppp(8)
.     delay ms-delay             Propagation delay, measured in milliseconds.  The value is             rounded to the next multiple of the clock tick (typically 10ms,             but it is a good practice to run kernels with ``options HZ=1000''             to reduce the granularity to 1ms or less).  Default value is 0,             meaning no delay.     The following parameters can be configured for a queue:     pipe pipe_nr             Connects a queue to the specified pipe.  Multiple queues (with             the same or different weights) can be connected to the same pipe,             which specifies the aggregate rate for the set of queues.     weight weight             Specifies the weight to be used for flows matching this queue.             The weight must be in the range 1..100, and defaults to 1.     Finally, the following parameters can be configured for both pipes and     queues:     buckets hash-table-size           Specifies the size of the hash table used for storing the various           queues.  Default value is 64 controlled by the
sysctl(8)
variable           net.inet.ip.dummynet.hash_size, allowed range is 16 to 65536.     mask mask-specifier           Packets sent to a given pipe or queue by an ipfw rule can be fur-           ther classified into multiple flows, each of which is then sent to           a different dynamic pipe or queue.  A flow identifier is con-           structed by masking the IP addresses, ports and protocol types as           specified with the mask options in the configuration of the pipe or           queue.  For each different flow identifier, a new pipe or queue is           created with the same parameters as the original object, and match-           ing packets are sent to it.           Thus, when dynamic pipes are used, each flow will get the same           bandwidth as defined by the pipe, whereas when dynamic queues are           used, each flow will share the parent's pipe bandwidth evenly with           other flows generated by the same queue (note that other queues           with different weights might be connected to the same pipe).           Available mask specifiers are a combination of one or more of the           following:           dst-ip mask, dst-ip6 mask, src-ip mask, src-ip6 mask, dst-port           mask, src-port mask, flow-id mask, proto mask or all,           where the latter means all bits in all fields are significant.     noerror           When a packet is dropped by a dummynet queue or pipe, the error is           normally reported to the caller routine in the kernel, in the same           way as it happens when a device queue fills up.  Setting this           option reports the packet as successfully delivered, which can be           needed for some experimental setups where you want to simulate loss           or congestion at a remote router.     plr packet-loss-rate           Packet loss rate.  Argument packet-loss-rate is a floating-point           number between 0 and 1, with 0 meaning no loss, 1 meaning 100%           loss.  The loss rate is internally represented on 31 bits.     queue {slots | sizeKbytes}           Queue size, in slots or KBytes.  Default value is 50 slots, which           is the typical queue size for Ethernet devices.  Note that for slow           speed links you should keep the queue size short or your traffic           might be affected by a significant queueing delay.  E.g., 50 max-           sized ethernet packets (1500 bytes) mean 600Kbit or 20s of queue on           a 30Kbit/s pipe.  Even worse effects can result if you get packets           from an interface with a much larger MTU, e.g. the loopback inter-           face with its 16KB packets.     red | gred w_q/min_th/max_th/max_p           Make use of the RED (Random Early Detection) queue management algo-           rithm.  w_q and max_p are floating point numbers between 0 and 1 (0           not included), while min_th and max_th are integer numbers specify-           ing thresholds for queue management (thresholds are computed in           bytes if the queue has been defined in bytes, in slots otherwise).           The
dummynet(4)
also supports the gentle RED variant (gred).  Three          
sysctl(8)
variables can be used to control the RED behaviour:           net.inet.ip.dummynet.red_lookup_depth                   specifies the accuracy in computing the average queue when                   the link is idle (defaults to 256, must be greater than                   zero)           net.inet.ip.dummynet.red_avg_pkt_size                   specifies the expected average packet size (defaults to                   512, must be greater than zero)           net.inet.ip.dummynet.red_max_pkt_size                   specifies the expected maximum packet size, only used when                   queue thresholds are in bytes (defaults to 1500, must be                   greater than zero).     When used with IPv6 data, dummynet currently has several limitations.     First, debug.mpsafenet=0 must be set.  Second, the information necessi-     cary to route link-local packets to an interface is not avalable after     processing by dummynet so those packets are dropped in the output path.     Care should be taken to insure that link-local packets are not passed to     dummynet.
CHECKLIST
     Here are some important points to consider when designing your rules:     ·         Remember that you filter both packets going in and out.  Most connec-         tions need packets going in both directions.     ·         Remember to test very carefully.  It is a good idea to be near the         console when doing this.  If you cannot be near the console, use an         auto-recovery script such as the one in         /usr/share/examples/ipfw/change_rules.sh.     ·         Do not forget the loopback interface.
FINE POINTS
     ·         There are circumstances where fragmented datagrams are uncondition-         ally dropped.        TCP packets are dropped if they do not contain at         least 20 bytes of TCP header, UDP packets are dropped if they do not         contain a full 8 byte UDP header, and ICMP packets are dropped if         they do not contain 4 bytes of ICMP header, enough to specify the         ICMP type, code, and checksum.  These packets are simply logged as         ``pullup failed'' since there may not be enough good data in the         packet to produce a meaningful log entry.     ·         Another type of packet is unconditionally dropped, a TCP packet with         a fragment offset of one.  This is a valid packet, but it only has         one use, to try to circumvent firewalls.  When logging is enabled,         these packets are reported as being dropped by rule -1.     ·         If you are logged in over a network, loading the
kld(4)
version of         ipfw is probably not as straightforward as you would think.  I recom-         mend the following command line:               kldload ipfw && \               ipfw add 32000 allow ip from any to any         Along the same lines, doing an               ipfw flush         in similar surroundings is also a bad idea.     ·         The ipfw filter list may not be modified if the system security level         is set to 3 or higher (see
init(8)
for information on system security         levels).
PACKET DIVERSION
     A
divert(4)
socket bound to the specified port will receive all packets     diverted to that port.  If no socket is bound to the destination port, or     if the divert module is not loaded, or if the kernel was not compiled     with divert socket support, the packets are dropped.
SYSCTL VARIABLES
     A set of
sysctl(8)
variables controls the behaviour of the firewall and     associated modules (dummynet, bridge).  These are shown below together     with their default value (but always check with the
sysctl(8)
command     what value is actually in use) and meaning:     net.inet.ip.dummynet.expire: 1             Lazily delete dynamic pipes/queue once they have no pending traf-             fic.  You can disable this by setting the variable to 0, in which             case the pipes/queues will only be deleted when the threshold is             reached.     net.inet.ip.dummynet.hash_size: 64             Default size of the hash table used for dynamic pipes/queues.             This value is used when no buckets option is specified when con-             figuring a pipe/queue.     net.inet.ip.dummynet.max_chain_len: 16             Target value for the maximum number of pipes/queues in a hash             bucket.  The product max_chain_len*hash_size is used to determine             the threshold over which empty pipes/queues will be expired even             when net.inet.ip.dummynet.expire=0.     net.inet.ip.dummynet.red_lookup_depth: 256     net.inet.ip.dummynet.red_avg_pkt_size: 512     net.inet.ip.dummynet.red_max_pkt_size: 1500             Parameters used in the computations of the drop probability for             the RED algorithm.     net.inet.ip.fw.autoinc_step: 100             Delta between rule numbers when auto-generating them.  The value             must be in the range 1..1000.     net.inet.ip.fw.curr_dyn_buckets: net.inet.ip.fw.dyn_buckets             The current number of buckets in the hash table for dynamic rules             (readonly).     net.inet.ip.fw.debug: 1             Controls debugging messages produced by ipfw.     net.inet.ip.fw.dyn_buckets: 256             The number of buckets in the hash table for dynamic rules.  Must             be a power of 2, up to 65536.  It only takes effect when all             dynamic rules have expired, so you are advised to use a flush             command to make sure that the hash table is resized.     net.inet.ip.fw.dyn_count: 3             Current number of dynamic rules (read-only).     net.inet.ip.fw.dyn_keepalive: 1             Enables generation of keepalive packets for keep-state rules on             TCP sessions.  A keepalive is generated to both sides of the con-             nection every 5 seconds for the last 20 seconds of the lifetime             of the rule.     net.inet.ip.fw.dyn_max: 8192             Maximum number of dynamic rules.  When you hit this limit, no             more dynamic rules can be installed until old ones expire.     net.inet.ip.fw.dyn_ack_lifetime: 300     net.inet.ip.fw.dyn_syn_lifetime: 20     net.inet.ip.fw.dyn_fin_lifetime: 1     net.inet.ip.fw.dyn_rst_lifetime: 1     net.inet.ip.fw.dyn_udp_lifetime: 5     net.inet.ip.fw.dyn_short_lifetime: 30             These variables control the lifetime, in seconds, of dynamic             rules.  Upon the initial SYN exchange the lifetime is kept short,             then increased after both SYN have been seen, then decreased             again during the final FIN exchange or when a RST is received.             Both dyn_fin_lifetime and dyn_rst_lifetime must be strictly lower             than 5 seconds, the period of repetition of keepalives.  The             firewall enforces that.     net.inet.ip.fw.enable: 1             Enables the firewall.  Setting this variable to 0 lets you run             your machine without firewall even if compiled in.     net.inet.ip.fw.one_pass: 1             When set, the packet exiting from the
dummynet(4)
pipe or from             
ng_ipfw(4)
node is not passed though the firewall again.  Other-             wise, after an action, the packet is reinjected into the firewall             at the next rule.     net.inet.ip.fw.verbose: 1             Enables verbose messages.     net.inet.ip.fw.verbose_limit: 0             Limits the number of messages produced by a verbose firewall.     net.inet6.ip6.fw.deny_unknown_exthdrs: 1             If enabled packets with unknown IPv6 Extension Headers will be             denied.     net.link.ether.ipfw: 0             Controls whether layer-2 packets are passed to ipfw.  Default is             no.     net.link.ether.bridge_ipfw: 0             Controls whether bridged packets are passed to ipfw.  Default is             no.
EXAMPLES
     There are far too many possible uses of ipfw so this Section will only     give a small set of examples.   BASIC PACKET FILTERING     This command adds an entry which denies all tcp packets from     cracker.evil.org to the telnet port of wolf.tambov.su from being for-     warded by the host:           ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet     This one disallows any connection from the entire cracker's network to my     host:           ipfw add deny ip from 123.45.67.0/24 to my.host.org     A first and efficient way to limit access (not using dynamic rules) is     the use of the following rules:           ipfw add allow tcp from any to any established           ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup           ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup           ...           ipfw add deny tcp from any to any     The first rule will be a quick match for normal TCP packets, but it will     not match the initial SYN packet, which will be matched by the setup     rules only for selected source/destination pairs.        All other SYN packets     will be rejected by the final deny rule.     If you administer one or more subnets, you can take advantage of the     address sets and or-blocks and write extremely compact rulesets which     selectively enable services to blocks of clients, as below:           goodguys="{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }"           badguys="10.1.2.0/24{8,38,60}"           ipfw add allow ip from ${goodguys} to any           ipfw add deny ip from ${badguys} to any           ... normal policies ...     The verrevpath option could be used to do automated anti-spoofing by     adding the following to the top of a ruleset:           ipfw add deny ip from any to any not verrevpath in     This rule drops all incoming packets that appear to be coming to the sys-     tem on the wrong interface.  For example, a packet with a source address     belonging to a host on a protected internal network would be dropped if     it tried to enter the system from an external interface.     The antispoof option could be used to do similar but more restricted     anti-spoofing by adding the following to the top of a ruleset:           ipfw add deny ip from any to any not antispoof in     This rule drops all incoming packets that appear to be coming from     another directly connected system but on the wrong interface.  For exam-     ple, a packet with a source address of 192.168.0.0/24 , configured on     fxp0 , but coming in on fxp1 would be dropped.   DYNAMIC RULES     In order to protect a site from flood attacks involving fake TCP packets,     it is safer to use dynamic rules:           ipfw add check-state           ipfw add deny tcp from any to any established           ipfw add allow tcp from my-net to any setup keep-state     This will let the firewall install dynamic rules only for those connec-     tion which start with a regular SYN packet coming from the inside of our     network.  Dynamic rules are checked when encountering the first     check-state or keep-state rule.  A check-state rule should usually be     placed near the beginning of the ruleset to minimize the amount of work     scanning the ruleset.  Your mileage may vary.     To limit the number of connections a user can open you can use the fol-     lowing type of rules:           ipfw add allow tcp from my-net/24 to any setup limit src-addr 10           ipfw add allow tcp from any to me setup limit src-addr 4     The former (assuming it runs on a gateway) will allow each host on a /24     network to open at most 10 TCP connections.  The latter can be placed on     a server to make sure that a single client does not use more than 4     simultaneous connections.     BEWARE: stateful rules can be subject to denial-of-service attacks by a     SYN-flood which opens a huge number of dynamic rules.  The effects of     such attacks can be partially limited by acting on a set of
sysctl(8)
     variables which control the operation of the firewall.     Here is a good usage of the list command to see accounting records and     timestamp information:           ipfw -at list     or in short form without timestamps:           ipfw -a list     which is equivalent to:           ipfw show     Next rule diverts all incoming packets from 192.168.2.0/24 to divert port     5000:           ipfw divert 5000 ip from 192.168.2.0/24 to any in   TRAFFIC SHAPING     The following rules show some of the applications of ipfw and
dummynet(4)
     for simulations and the like.     This rule drops random incoming packets with a probability of 5%:           ipfw add prob 0.05 deny ip from any to any in     A similar effect can be achieved making use of dummynet pipes:           ipfw add pipe 10 ip from any to any           ipfw pipe 10 config plr 0.05     We can use pipes to artificially limit bandwidth, e.g. on a machine act-     ing as a router, if we want to limit traffic from local clients on     192.168.2.0/24 we do:           ipfw add pipe 1 ip from 192.168.2.0/24 to any out           ipfw pipe 1 config bw 300Kbit/s queue 50KBytes     note that we use the out modifier so that the rule is not used twice.     Remember in fact that ipfw rules are checked both on incoming and outgo-     ing packets.     Should we want to simulate a bidirectional link with bandwidth limita-     tions, the correct way is the following:           ipfw add pipe 1 ip from any to any out           ipfw add pipe 2 ip from any to any in           ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes           ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes     The above can be very useful, e.g. if you want to see how your fancy Web     page will look for a residential user who is connected only through a     slow link.  You should not use only one pipe for both directions, unless     you want to simulate a half-duplex medium (e.g. AppleTalk, Ethernet,     IRDA).  It is not necessary that both pipes have the same configuration,     so we can also simulate asymmetric links.     Should we want to verify network performance with the RED queue manage-     ment algorithm:           ipfw add pipe 1 ip from any to any           ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1     Another typical application of the traffic shaper is to introduce some     delay in the communication.  This can significantly affect applications     which do a lot of Remote Procedure Calls, and where the round-trip-time     of the connection often becomes a limiting factor much more than band-     width:           ipfw add pipe 1 ip from any to any out           ipfw add pipe 2 ip from any to any in           ipfw pipe 1 config delay 250ms bw 1Mbit/s           ipfw pipe 2 config delay 250ms bw 1Mbit/s     Per-flow queueing can be useful for a variety of purposes.  A very simple     one is counting traffic:           ipfw add pipe 1 tcp from any to any           ipfw add pipe 1 udp from any to any           ipfw add pipe 1 ip from any to any           ipfw pipe 1 config mask all     The above set of rules will create queues (and collect statistics) for     all traffic.  Because the pipes have no limitations, the only effect is     collecting statistics.  Note that we need 3 rules, not just the last one,     because when ipfw tries to match IP packets it will not consider ports,     so we would not see connections on separate ports as different ones.     A more sophisticated example is limiting the outbound traffic on a net     with per-host limits, rather than per-network limits:           ipfw add pipe 1 ip from 192.168.2.0/24 to any out           ipfw add pipe 2 ip from any to 192.168.2.0/24 in           ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue           20Kbytes           ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue           20Kbytes   LOOKUP TABLES     In the following example, we need to create several traffic bandwidth     classes and we need different hosts/networks to fall into different     classes.  We create one pipe for each class and configure them accord-     ingly.  Then we create a single table and fill it with IP subnets and     addresses.  For each subnet/host we set the argument equal to the number     of the pipe that it should use.  Then we classify traffic using a single     rule:           ipfw pipe 1 config bw 1000Kbyte/s           ipfw pipe 4 config bw 4000Kbyte/s           ...           ipfw table 1 add 192.168.2.0/24 1           ipfw table 1 add 192.168.0.0/27 4           ipfw table 1 add 192.168.0.2 1           ...           ipfw pipe tablearg ip from
table(1)
to any   SETS OF RULES     To add a set of rules atomically, e.g. set 18:           ipfw set disable 18           ipfw add NN set 18 ...          # repeat as needed           ipfw set enable 18     To delete a set of rules atomically the command is simply:           ipfw delete set 18     To test a ruleset and disable it and regain control if something goes     wrong:           ipfw set disable 18           ipfw add NN set 18 ...          # repeat as needed           ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18     Here if everything goes well, you press control-C before the "sleep" ter-     minates, and your ruleset will be left active.  Otherwise, e.g. if you     cannot access your box, the ruleset will be disabled after the sleep ter-     minates thus restoring the previous situation.
SEE ALSO
     
cpp(1)
,
m4(1)
,
altq(4)
,
bridge(4)
,
divert(4)
,
dummynet(4)
,
ip(4)
,     
ipfirewall(4)
,
ng_ipfw(4)
,
protocols(5)
,
services(5)
,
init(8)
,     
kldload(8)
,
reboot(8)
,
sysctl(8)
,
syslogd(8)
HISTORY
     The ipfw utility first appeared in FreeBSD 2.0.  
dummynet(4)
was intro-     duced in FreeBSD 2.2.8.  Stateful extensions were introduced in     FreeBSD 4.0.  ipfw2 was introduced in Summer 2002.
AUTHORS
     Ugen J. S. Antsilevich,     Poul-Henning Kamp,     Alex Nash,     Archie Cobbs,     Luigi Rizzo.     API based upon code written by Daniel Boulet for BSDI.     Work on
dummynet(4)
traffic shaper supported by Akamba Corp.
BUGS
     Use of dummynet with IPv6 requires that debug.mpsafenet be set to 0.     The syntax has grown over the years and sometimes it might be confusing.     Unfortunately, backward compatibility prevents cleaning up mistakes made     in the definition of the syntax.     !!! WARNING !!!     Misconfiguring the firewall can put your computer in an unusable state,     possibly shutting down network services and requiring console access to     regain control of it.     Incoming packet fragments diverted by divert are reassembled before     delivery to the socket.  The action used on those packet is the one from     the rule which matches the first fragment of the packet.     Packets diverted to userland, and then reinserted by a userland process     may lose various packet attributes.  The packet source interface name     will be preserved if it is shorter than 8 bytes and the userland process     saves and reuses the sockaddr_in (as does
natd(8)
); otherwise, it may be     lost.  If a packet is reinserted in this manner, later rules may be     incorrectly applied, making the order of divert rules in the rule     sequence very important.     Dummynet drops all packets with IPv6 link-local addresses.     Rules using uid or gid may not behave as expected.  In particular, incom-     ing SYN packets may have no uid or gid associated with them since they do     not yet belong to a TCP connection, and the uid/gid associated with a     packet may not be as expected if the associated process calls
setuid(2)
     or similar system calls.     Rules which use uid, gid or jail based matching should be used only if     debug.mpsafenet=0 to avoid possible deadlocks due to layering violations     in its implementation.FreeBSD 4.11                       January 16, 2006                   FreeBSD 4.11
NAME
|
SYNOPSIS
|
DESCRIPTION
|
PACKET FLOW
|
SYNTAX
|
RULE FORMAT
|
LOOKUP TABLES
|
SETS OF
|
STATEFUL FIREWALL
|
TRAFFIC SHAPER
|
CHECKLIST
|
FINE POINTS
|
PACKET DIVERSION
|
SYSCTL VARIABLES
|
EXAMPLES
|
SEE ALSO
|
HISTORY
|
AUTHORS
|
BUGS
IPFW(8)                 FreeBSD System Manager's Manual                IPFW(8)
NAME
     ipfw -- IP firewall and traffic shaper control program
SYNOPSIS
     ipfw [-cq] add rule
     ipfw [-acdefnNStT] {list | show} [rule | first-last ...]
     ipfw [-f | -q] flush
     ipfw [-q] {delete | zero | resetlog} [set] [number ...]
     ipfw enable
          {firewall | altq | one_pass | debug | verbose | dyn_keepalive}
     ipfw disable
          {firewall | altq | one_pass | debug | verbose | dyn_keepalive}
     ipfw set [disable number ...] [enable number ...]
     ipfw set move [rule] number to number
     ipfw set swap number number
     ipfw set show
     ipfw table number add addr[/masklen] [value]
     ipfw table number delete addr[/masklen]
     ipfw table number flush
     ipfw table number list
     ipfw {pipe | queue} number config config-options
     ipfw [-s [field]] {pipe | queue} {delete | list | show} [number ...]
     ipfw [-cfnNqS] [-p preproc [preproc-flags]] pathname
DESCRIPTION
     The ipfw utility is the user interface for controlling the
ipfw(4)
fire-
     wall and the
dummynet(4)
traffic shaper in FreeBSD.
     An ipfw configuration, or ruleset, is made of a list of rules numbered
     from 1 to 65535.  Packets are passed to ipfw from a number of different
     places in the protocol stack (depending on the source and destination of
     the packet, it is possible that ipfw is invoked multiple times on the
     same packet).  The packet passed to the firewall is compared against each
     of the rules in the firewall ruleset.  When a match is found, the action
     corresponding to the matching rule is performed.
     Depending on the action and certain system settings, packets can be rein-
     jected into the firewall at some rule after the matching one for further
     processing.
     An ipfw ruleset always includes a default rule (numbered 65535) which
     cannot be modified or deleted, and matches all packets.  The action asso-
     ciated with the default rule can be either deny or allow depending on how
     the kernel is configured.
     If the ruleset includes one or more rules with the keep-state or limit
     option, then ipfw assumes a stateful behaviour, i.e., upon a match it
     will create dynamic rules matching the exact parameters (addresses and
     ports) of the matching packet.
     These dynamic rules, which have a limited lifetime, are checked at the
     first occurrence of a check-state, keep-state or limit rule, and are typ-
     ically used to open the firewall on-demand to legitimate traffic only.
     See the STATEFUL FIREWALL and EXAMPLES Sections below for more informa-
     tion on the stateful behaviour of ipfw.
     All rules (including dynamic ones) have a few associated counters: a
     packet count, a byte count, a log count and a timestamp indicating the
     time of the last match.  Counters can be displayed or reset with ipfw
     commands.
     Rules can be added with the add command; deleted individually or in
     groups with the delete command, and globally (except those in set 31)
     with the flush command; displayed, optionally with the content of the
     counters, using the show and list commands.  Finally, counters can be
     reset with the zero and resetlog commands.
     Also, each rule belongs to one of 32 different sets , and there are ipfw
     commands to atomically manipulate sets, such as enable, disable, swap
     sets, move all rules in a set to another one, delete all rules in a set.
     These can be useful to install temporary configurations, or to test them.
     See Section SETS OF RULES for more information on sets.
     The following options are available:
     -a      While listing, show counter values.  The show command just
             implies this option.
     -b      Only show the action and the comment, not the body of a rule.
             Implies -c.
     -c      When entering or showing rules, print them in compact form, i.e.,
             without the optional "ip from any to any" string when this does
             not carry any additional information.
     -d      While listing, show dynamic rules in addition to static ones.
     -e      While listing, if the -d option was specified, also show expired
             dynamic rules.
     -f      Do not ask for confirmation for commands that can cause problems
             if misused, i.e. flush.  If there is no tty associated with the
             process, this is implied.
     -n      Only check syntax of the command strings, without actually pass-
             ing them to the kernel.
     -N      Try to resolve addresses and service names in output.
     -q      While adding, zeroing, resetlogging or flushing, be quiet about
             actions (implies -f).  This is useful for adjusting rules by exe-
             cuting multiple ipfw commands in a script (e.g.,
             `sh /etc/rc.firewall'), or by processing a file of many ipfw
             rules across a remote login session.  If a flush is performed in
             normal (verbose) mode (with the default kernel configuration), it
             prints a message.        Because all rules are flushed, the message
             might not be delivered to the login session, causing the remote
             login session to be closed and the remainder of the ruleset to
             not be processed.        Access to the console would then be required
             to recover.
     -S      While listing rules, show the set each rule belongs to.  If this
             flag is not specified, disabled rules will not be listed.
     -s [field]
             While listing pipes, sort according to one of the four counters
             (total or current packets or bytes).
     -t      While listing, show last match timestamp (converted with
             ctime()).
     -T      While listing, show last match timestamp (as seconds from the
             epoch).  This form can be more convenient for postprocessing by
             scripts.
     To ease configuration, rules can be put into a file which is processed
     using ipfw as shown in the last synopsis line.  An absolute pathname must
     be used.  The file will be read line by line and applied as arguments to
     the ipfw utility.
     Optionally, a preprocessor can be specified using -p preproc where
     pathname is to be piped through.  Useful preprocessors include
cpp(1)
and
     
m4(1)
.  If preproc does not start with a slash (`/') as its first charac-
     ter, the usual PATH name search is performed.  Care should be taken with
     this in environments where not all file systems are mounted (yet) by the
     time ipfw is being run (e.g. when they are mounted over NFS).  Once -p
     has been specified, any additional arguments as passed on to the pre-
     processor for interpretation.  This allows for flexible configuration
     files (like conditionalizing them on the local hostname) and the use of
     macros to centralize frequently required arguments like IP addresses.
     The ipfw pipe and queue commands are used to configure the traffic
     shaper, as shown in the TRAFFIC SHAPER (DUMMYNET) CONFIGURATION Section
     below.
     If the world and the kernel get out of sync the ipfw ABI may break, pre-
     venting you from being able to add any rules.  This can adversely effect
     the booting process.  You can use ipfw disable firewall to temporarily
     disable the firewall to regain access to the network, allowing you to fix
     the problem.
PACKET FLOW
     A packet is checked against the active ruleset in multiple places in the
     protocol stack, under control of several sysctl variables.  These places
     and variables are shown below, and it is important to have this picture
     in mind in order to design a correct ruleset.
                  ^    to upper layers          V
                  |                          |
                  +----------->-----------+
                  ^                          V
            [
ip(6)
_input]            [
ip(6)
_output]     net.inet.ip.fw.enable=1
                  |                          |
                  ^                          V
            [ether_demux]         [ether_output_frame]  net.link.ether.ipfw=1
                  |                          |
                  +-->--[bdg_forward]-->--+               net.link.ether.bridge_ipfw=1
                  ^                          V
                  |         to devices          |
     As can be noted from the above picture, the number of times the same
     packet goes through the firewall can vary between 0 and 4 depending on
     packet source and destination, and system configuration.
     Note that as packets flow through the stack, headers can be stripped or
     added to it, and so they may or may not be available for inspection.
     E.g., incoming packets will include the MAC header when ipfw is invoked
     from ether_demux(), but the same packets will have the MAC header
     stripped off when ipfw is invoked from ip_input() or ip6_input().
     Also note that each packet is always checked against the complete rule-
     set, irrespective of the place where the check occurs, or the source of
     the packet.  If a rule contains some match patterns or actions which are
     not valid for the place of invocation (e.g. trying to match a MAC header
     within ip_input or ip6_input ), the match pattern will not match, but a
     not operator in front of such patterns will cause the pattern to always
     match on those packets.  It is thus the responsibility of the programmer,
     if necessary, to write a suitable ruleset to differentiate among the pos-
     sible places.  skipto rules can be useful here, as an example:
           # packets from ether_demux or bdg_forward
           ipfw add 10 skipto 1000 all from any to any layer2 in
           # packets from ip_input
           ipfw add 10 skipto 2000 all from any to any not layer2 in
           # packets from ip_output
           ipfw add 10 skipto 3000 all from any to any not layer2 out
           # packets from ether_output_frame
           ipfw add 10 skipto 4000 all from any to any layer2 out
     (yes, at the moment there is no way to differentiate between ether_demux
     and bdg_forward).
SYNTAX
     In general, each keyword or argument must be provided as a separate com-
     mand line argument, with no leading or trailing spaces.  Keywords are
     case-sensitive, whereas arguments may or may not be case-sensitive
     depending on their nature (e.g. uid's are, hostnames are not).
     In ipfw2 you can introduce spaces after commas ',' to make the line more
     readable.        You can also put the entire command (including flags) into a
     single argument.  E.g., the following forms are equivalent:
           ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
           ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
           ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
RULE FORMAT
     The format of ipfw rules is the following:
           [rule_number] [set set_number] [prob match_probability]
               action [log [logamount number]] [altq queue] body
     where the body of the rule specifies which information is used for fil-
     tering packets, among the following:
        Layer-2 header fields                      When available
        IPv4 and IPv6 Protocol                      TCP, UDP, ICMP, etc.
        Source and dest. addresses and ports
        Direction                              See Section PACKET FLOW
        Transmit and receive interface              By name or address
        Misc. IP header fields                      Version, type of service, data-
                                              gram length, identification,
                                              fragment flag (non-zero IP off-
                                              set), Time To Live
        IP options
        IPv6 Extension headers                      Fragmentation, Hop-by-Hop
                                              options, source routing, IPSec
                                              options.
        IPv6 Flow-ID
        Misc. TCP header fields               TCP flags (SYN, FIN, ACK, RST,
                                              etc.), sequence number, acknowl-
                                              edgment number, window
        TCP options
        ICMP types                              for ICMP packets
        ICMP6 types                              for ICMP6 packets
        User/group ID                              When the packet can be associ-
                                              ated with a local socket.
        Divert status                              Whether a packet came from a
                                              divert socket (e.g.,
natd(8)
).
     Note that some of the above information, e.g. source MAC or IP addresses
     and TCP/UDP ports, could easily be spoofed, so filtering on those fields
     alone might not guarantee the desired results.
     rule_number
             Each rule is associated with a rule_number in the range 1..65535,
             with the latter reserved for the default rule.  Rules are checked
             sequentially by rule number.  Multiple rules can have the same
             number, in which case they are checked (and listed) according to
             the order in which they have been added.  If a rule is entered
             without specifying a number, the kernel will assign one in such a
             way that the rule becomes the last one before the default rule.
             Automatic rule numbers are assigned by incrementing the last non-
             default rule number by the value of the sysctl variable
             net.inet.ip.fw.autoinc_step which defaults to 100.  If this is
             not possible (e.g. because we would go beyond the maximum allowed
             rule number), the number of the last non-default value is used
             instead.
     set set_number
             Each rule is associated with a set_number in the range 0..31.
             Sets can be individually disabled and enabled, so this parameter
             is of fundamental importance for atomic ruleset manipulation.  It
             can be also used to simplify deletion of groups of rules.        If a
             rule is entered without specifying a set number, set 0 will be
             used.
             Set 31 is special in that it cannot be disabled, and rules in set
             31 are not deleted by the ipfw flush command (but you can delete
             them with the ipfw delete set 31 command).  Set 31 is also used
             for the default rule.
     prob match_probability
             A match is only declared with the specified probability (floating
             point number between 0 and 1).  This can be useful for a number
             of applications such as random packet drop or (in conjunction
             with
dummynet(4)
) to simulate the effect of multiple paths lead-
             ing to out-of-order packet delivery.
             Note: this condition is checked before any other condition,
             including ones such as keep-state or check-state which might have
             side effects.
     log [logamount number]
             When a packet matches a rule with the log keyword, a message will
             be logged to
syslogd(8)
with a LOG_SECURITY facility.  The log-
             ging only occurs if the sysctl variable net.inet.ip.fw.verbose is
             set to 1 (which is the default when the kernel is compiled with
             IPFIREWALL_VERBOSE) and the number of packets logged so far for
             that particular rule does not exceed the logamount parameter.  If
             no logamount is specified, the limit is taken from the sysctl
             variable net.inet.ip.fw.verbose_limit.  In both cases, a value of
             0 removes the logging limit.
             Once the limit is reached, logging can be re-enabled by clearing
             the logging counter or the packet counter for that entry, see the
             resetlog command.
             Note: logging is done after all other packet matching conditions
             have been successfully verified, and before performing the final
             action (accept, deny, etc.) on the packet.
     altq queue
             When a packet matches a rule with the altq keyword, the ALTQ
             identifier for the given queue (see
altq(4)
) will be attached.
             Note that this ALTQ tag is only meaningful for packets going
             "out" of IPFW, and not being rejected or going to divert sockets.
             Note that if there is insufficient memory at the time the packet
             is processed, it will not be tagged, so it is wise to make your
             ALTQ "default" queue policy account for this.  If multiple altq
             rules match a single packet, only the first one adds the ALTQ
             classification tag.  In doing so, traffic may be shaped by using
             count altq queue rules for classification early in the ruleset,
             then later applying the filtering decision.  For example,
             check-state and keep-state rules may come later and provide the
             actual filtering decisions in addition to the fallback ALTQ tag.
             You must run
pfctl(8)
to set up the queues before IPFW will be
             able to look them up by name, and if the ALTQ disciplines are
             rearranged, the rules in containing the queue identifiers in the
             kernel will likely have gone stale and need to be reloaded.
             Stale queue identifiers will probably result in misclassifica-
             tion.
             All system ALTQ processing can be turned on or off via ipfw
             enable altq and ipfw disable altq.  The usage of
             net.inet.ip.fw.one_pass is irrelevant to ALTQ traffic shaping, as
             the actual rule action is followed always after adding an ALTQ
             tag.
   RULE ACTIONS
     A rule can be associated with one of the following actions, which will be
     executed when the packet matches the body of the rule.
     allow | accept | pass | permit
             Allow packets that match rule.  The search terminates.
     check-state
             Checks the packet against the dynamic ruleset.  If a match is
             found, execute the action associated with the rule which gener-
             ated this dynamic rule, otherwise move to the next rule.
             Check-state rules do not have a body.  If no check-state rule is
             found, the dynamic ruleset is checked at the first keep-state or
             limit rule.
     count   Update counters for all packets that match rule.  The search con-
             tinues with the next rule.
     deny | drop
             Discard packets that match this rule.  The search terminates.
     divert port
             Divert packets that match this rule to the
divert(4)
socket bound
             to port port.  The search terminates.
     fwd | forward ipaddr[,port]
             Change the next-hop on matching packets to ipaddr, which can be
             an IP address or a host name.  The search terminates if this rule
             matches.
             If ipaddr is a local address, then matching packets will be for-
             warded to port (or the port number in the packet if one is not
             specified in the rule) on the local machine.
             If ipaddr is not a local address, then the port number (if speci-
             fied) is ignored, and the packet will be forwarded to the remote
             address, using the route as found in the local routing table for
             that IP.
             A fwd rule will not match layer-2 packets (those received on
             ether_input, ether_output, or bridged).
             The fwd action does not change the contents of the packet at all.
             In particular, the destination address remains unmodified, so
             packets forwarded to another system will usually be rejected by
             that system unless there is a matching rule on that system to
             capture them.  For packets forwarded locally, the local address
             of the socket will be set to the original destination address of
             the packet.  This makes the
netstat(1)
entry look rather weird
             but is intended for use with transparent proxy servers.
             To enable fwd a custom kernel needs to be compiled with the
             option options IPFIREWALL_FORWARD.  With the additional option
             options IPFIREWALL_FORWARD_EXTENDED all safeguards are removed
             and it also makes it possible to redirect packets destined to
             locally configured IP addresses.  Please note that such rules
             apply to locally generated packets as well and great care is
             required to ensure proper behaviour for automatically generated
             packets like ICMP message size exceeded and others.
     pipe pipe_nr
             Pass packet to a
dummynet(4)
``pipe'' (for bandwidth limitation,
             delay, etc.).  See the TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
             Section for further information.  The search terminates; however,
             on exit from the pipe and if the
sysctl(8)
variable
             net.inet.ip.fw.one_pass is not set, the packet is passed again to
             the firewall code starting from the next rule.
     queue queue_nr
             Pass packet to a
dummynet(4)
``queue'' (for bandwidth limitation
             using WF2Q+).
     reject  (Deprecated).  Synonym for unreach host.
     reset   Discard packets that match this rule, and if the packet is a TCP
             packet, try to send a TCP reset (RST) notice.  The search termi-
             nates.
     reset6  Discard packets that match this rule, and if the packet is a TCP
             packet, try to send a TCP reset (RST) notice.  The search termi-
             nates.
     skipto number
             Skip all subsequent rules numbered less than number.  The search
             continues with the first rule numbered number or higher.
     tee port
             Send a copy of packets matching this rule to the
divert(4)
socket
             bound to port port.  The search continues with the next rule.
     unreach code
             Discard packets that match this rule, and try to send an ICMP
             unreachable notice with code code, where code is a number from 0
             to 255, or one of these aliases: net, host, protocol, port,
             needfrag, srcfail, net-unknown, host-unknown, isolated,
             net-prohib, host-prohib, tosnet, toshost, filter-prohib,
             host-precedence or precedence-cutoff.  The search terminates.
     unreach6 code
             Discard packets that match this rule, and try to send an ICMPv6
             unreachable notice with code code, where code is a number from 0,
             1, 3 or 4, or one of these aliases: no-route, admin-prohib,
             address or port.  The search terminates.
     netgraph cookie
             Divert packet into netgraph with given cookie.  The search termi-
             nates.  If packet is later returned from netgraph it is either
             accepted or continues with the next rule, depending on
             net.inet.ip.fw.one_pass sysctl variable.
     ngtee cookie
             A copy of packet is diverted into netgraph, original packet is
             either accepted or continues with the next rule, depending on
             net.inet.ip.fw.one_pass sysctl variable.  See
ng_ipfw(4)
for more
             information on netgraph and ngtee actions.
   RULE BODY
     The body of a rule contains zero or more patterns (such as specific
     source and destination addresses or ports, protocol options, incoming or
     outgoing interfaces, etc.)  that the packet must match in order to be
     recognised.  In general, the patterns are connected by (implicit) and
     operators -- i.e., all must match in order for the rule to match.        Indi-
     vidual patterns can be prefixed by the not operator to reverse the result
     of the match, as in
           ipfw add 100 allow ip from not 1.2.3.4 to any
     Additionally, sets of alternative match patterns (or-blocks) can be con-
     structed by putting the patterns in lists enclosed between parentheses (
     ) or braces { }, and using the or operator as follows:
           ipfw add 100 allow ip from { x or not y or z } to any
     Only one level of parentheses is allowed.        Beware that most shells have
     special meanings for parentheses or braces, so it is advisable to put a
     backslash \ in front of them to prevent such interpretations.
     The body of a rule must in general include a source and destination
     address specifier.  The keyword any can be used in various places to
     specify that the content of a required field is irrelevant.
     The rule body has the following format:
           [proto from src to dst] [options]
     The first part (proto from src to dst) is for backward compatibility with
     earlier versions of FreeBSD.  In modern FreeBSD any match pattern
     (including MAC headers, IP protocols, addresses and ports) can be speci-
     fied in the options section.
     Rule fields have the following meaning:
     proto: protocol | { protocol or ... }
     protocol: [not] protocol-name | protocol-number
             An IP protocol specified by number or name (for a complete list
             see /etc/protocols), or one of the following keywords:
             ip4 | ipv4
                     Matches IPv4 packets.
             ip6 | ipv6
                     Matches IPv6 packets.
             ip | all
                     Matches any packet.
             The ipv6 in proto option will be treated as inner protocol.  And,
             the ipv4 is not available in proto option.
             The { protocol or ... } format (an or-block) is provided for con-
             venience only but its use is deprecated.
     src and dst: {addr | { addr or ... }} [[not] ports]
             An address (or a list, see below) optionally followed by ports
             specifiers.
             The second format (or-block with multiple addresses) is provided
             for convenience only and its use is discouraged.
     addr: [not] {any | me | me6 table(number[,value]) | addr-list | addr-set}
     any     matches any IP address.
     me      matches any IP address configured on an interface in the system.
     me6     matches any IPv6 address configured on an interface in the sys-
             tem.  The address list is evaluated at the time the packet is an-
             alysed.
     table(number[,value])
             Matches any IPv4 address for which an entry exists in the lookup
             table number.  If an optional 32-bit unsigned value is also spec-
             ified, an entry will match only if it has this value.  See the
             LOOKUP TABLES section below for more information on lookup
             tables.
     addr-list: ip-addr[,addr-list]
     ip-addr:
             A host or subnet address specified in one of the following ways:
             numeric-ip | hostname
                     Matches a single IPv4 address, specified as dotted-quad
                     or a hostname.  Hostnames are resolved at the time the
                     rule is added to the firewall list.
             addr/masklen
                     Matches all addresses with base addr (specified as an IP
                     address or a hostname) and mask width of masklen bits.
                     As an example, 1.2.3.4/25 will match all IP numbers from
                     1.2.3.0 to 1.2.3.127 .
             addr:mask
                     Matches all addresses with base addr (specified as an IP
                     address or a hostname) and the mask of mask, specified as
                     a dotted quad.  As an example, 1.2.3.4:255.0.255.0 will
                     match 1.*.3.*.  This form is advised only for non-con-
                     tiguous masks.  It is better to resort to the
                     addr/masklen format for contiguous masks, which is more
                     compact and less error-prone.
     addr-set: addr[/masklen]{list}
     list: {num | num-num}[,list]
             Matches all addresses with base address addr (specified as an IP
             address or a hostname) and whose last byte is in the list between
             braces { } .  Note that there must be no spaces between braces
             and numbers (spaces after commas are allowed).  Elements of the
             list can be specified as single entries or ranges.  The masklen
             field is used to limit the size of the set of addresses, and can
             have any value between 24 and 32.        If not specified, it will be
             assumed as 24.
             This format is particularly useful to handle sparse address sets
             within a single rule.  Because the matching occurs using a bit-
             mask, it takes constant time and dramatically reduces the com-
             plexity of rulesets.
             As an example, an address specified as 1.2.3.4/24{128,35-55,89}
             will match the following IP addresses:
             1.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
     addr6-list: ip6-addr[,addr6-list]
     ip6-addr:
             A host or subnet specified one of the following ways:
             numeric-ip | hostname
                     Matches a single IPv6 address as allowed by
inet_pton(3)
                     or a hostname.  Hostnames are resolved at the time the
                     rule is added to the firewall list.
             addr/masklen
                     Matches all IPv6 addresses with base addr (specified as
                     allowed by inet_pton or a hostname) and mask width of
                     masklen bits.
             No support for sets of IPv6 addresses is provided because IPv6
             addresses are typically random past the initial prefix.
     ports: {port | port-port}[,ports]
             For protocols which support port numbers (such as TCP and UDP),
             optional ports may be specified as one or more ports or port
             ranges, separated by commas but no spaces, and an optional not
             operator.        The `-' notation specifies a range of ports (including
             boundaries).
             Service names (from /etc/services) may be used instead of numeric
             port values.  The length of the port list is limited to 30 ports
             or ranges, though one can specify larger ranges by using an
             or-block in the options section of the rule.
             A backslash (`\') can be used to escape the dash (`-') character
             in a service name (from a shell, the backslash must be typed
             twice to avoid the shell itself interpreting it as an escape
             character).
                   ipfw add count tcp from any ftp\\-data-ftp to any
             Fragmented packets which have a non-zero offset (i.e., not the
             first fragment) will never match a rule which has one or more
             port specifications.  See the frag option for details on matching
             fragmented packets.
   RULE OPTIONS (MATCH PATTERNS)
     Additional match patterns can be used within rules.  Zero or more of
     these so-called options can be present in a rule, optionally prefixed by
     the not operand, and possibly grouped into or-blocks.
     The following match patterns can be used (listed in alphabetical order):
     // this is a comment.
             Inserts the specified text as a comment in the rule.  Everything
             following // is considered as a comment and stored in the rule.
             You can have comment-only rules, which are listed as having a
             count action followed by the comment.
     bridged
             Alias for layer2.
     diverted
             Matches only packets generated by a divert socket.
     diverted-loopback
             Matches only packets coming from a divert socket back into the IP
             stack input for delivery.
     diverted-output
             Matches only packets going from a divert socket back outward to
             the IP stack output for delivery.
     dst-ip ip-address
             Matches IPv4 packets whose destination IP is one of the
             address(es) specified as argument.
     {dst-ip6 | dst-ipv6} ip6-address
             Matches IPv6 packets whose destination IP is one of the
             address(es) specified as argument.
     dst-port ports
             Matches IP packets whose destination port is one of the port(s)
             specified as argument.
     established
             Matches TCP packets that have the RST or ACK bits set.
     ext6hdr header
             Matches IPv6 packets containing the extended header given by
             header.  Supported headers are:
             Fragment, (frag), Hop-to-hop options (hopopt), Source routing
             (route), Destination options (dstopt), IPSec authentication head-
             ers (ah), and IPSec encapsulated security payload headers (esp).
     flow-id labels
             Matches IPv6 packets containing any of the flow labels given in
             labels.  labels is a comma seperate list of numeric flow labels.
     frag    Matches packets that are fragments and not the first fragment of
             an IP datagram.  Note that these packets will not have the next
             protocol header (e.g. TCP, UDP) so options that look into these
             headers cannot match.
     gid group
             Matches all TCP or UDP packets sent by or received for a group.
             A group may be specified by name or number.  This option should
             be used only if debug.mpsafenet=0 to avoid possible deadlocks due
             to layering violations in its implementation.
     jail prisonID
             Matches all TCP or UDP packets sent by or received for the jail
             whos prison ID is prisonID.  This option should be used only if
             debug.mpsafenet=0 to avoid possible deadlocks due to layering
             violations in its implementation.
     icmptypes types
             Matches ICMP packets whose ICMP type is in the list types.  The
             list may be specified as any combination of individual types
             (numeric) separated by commas.  Ranges are not allowed. The sup-
             ported ICMP types are:
             echo reply (0), destination unreachable (3), source quench (4),
             redirect (5), echo request (8), router advertisement (9), router
             solicitation (10), time-to-live exceeded (11), IP header bad
             (12), timestamp request (13), timestamp reply (14), information
             request (15), information reply (16), address mask request (17)
             and address mask reply (18).
     icmp6types types
             Matches ICMP6 packets whose ICMP6 type is in the list of types.
             The list may be specified as any combination of individual types
             (numeric) separated by commas.  Ranges are not allowed.
     in | out
             Matches incoming or outgoing packets, respectively.  in and out
             are mutually exclusive (in fact, out is implemented as not in).
     ipid id-list
             Matches IPv4 packets whose ip_id field has value included in
             id-list, which is either a single value or a list of values or
             ranges specified in the same way as ports.
     iplen len-list
             Matches IP packets whose total length, including header and data,
             is in the set len-list, which is either a single value or a list
             of values or ranges specified in the same way as ports.
     ipoptions spec
             Matches packets whose IPv4 header contains the comma separated
             list of options specified in spec.  The supported IP options are:
             ssrr (strict source route), lsrr (loose source route), rr (record
             packet route) and ts (timestamp).        The absence of a particular
             option may be denoted with a `!'.
     ipprecedence precedence
             Matches IPv4 packets whose precedence field is equal to
             precedence.
     ipsec   Matches packets that have IPSEC history associated with them
             (i.e., the packet comes encapsulated in IPSEC, the kernel has
             IPSEC support and IPSEC_FILTERGIF option, and can correctly
             decapsulate it).
             Note that specifying ipsec is different from specifying proto
             ipsec as the latter will only look at the specific IP protocol
             field, irrespective of IPSEC kernel support and the validity of
             the IPSEC data.
             Further note that this flag is silently ignored in kernels with-
             out IPSEC support.  It does not affect rule processing when given
             and the rules are handled as if with no ipsec flag.
     iptos spec
             Matches IPv4 packets whose tos field contains the comma separated
             list of service types specified in spec.  The supported IP types
             of service are:
             lowdelay (IPTOS_LOWDELAY), throughput (IPTOS_THROUGHPUT),
             reliability (IPTOS_RELIABILITY), mincost (IPTOS_MINCOST),
             congestion (IPTOS_CE).  The absence of a particular type may be
             denoted with a `!'.
     ipttl ttl-list
             Matches IPv4 packets whose time to live is included in ttl-list,
             which is either a single value or a list of values or ranges
             specified in the same way as ports.
     ipversion ver
             Matches IP packets whose IP version field is ver.
     keep-state
             Upon a match, the firewall will create a dynamic rule, whose
             default behaviour is to match bidirectional traffic between
             source and destination IP/port using the same protocol.  The rule
             has a limited lifetime (controlled by a set of
sysctl(8)
vari-
             ables), and the lifetime is refreshed every time a matching
             packet is found.
     layer2  Matches only layer2 packets, i.e., those passed to ipfw from
             ether_demux() and ether_output_frame().
     limit {src-addr | src-port | dst-addr | dst-port} N
             The firewall will only allow N connections with the same set of
             parameters as specified in the rule.  One or more of source and
             destination addresses and ports can be specified.        Currently,
             only IPv4 flows are supported.
     { MAC | mac } dst-mac src-mac
             Match packets with a given dst-mac and src-mac addresses, speci-
             fied as the any keyword (matching any MAC address), or six groups
             of hex digits separated by colons, and optionally followed by a
             mask indicating the significant bits.  The mask may be specified
             using either of the following methods:
             1.      A slash (/) followed by the number of significant bits.
                     For example, an address with 33 significant bits could be
                     specified as:
                           MAC 10:20:30:40:50:60/33 any
             2.      An ampersand (&) followed by a bitmask specified as six
                     groups of hex digits separated by colons.        For example,
                     an address in which the last 16 bits are significant
                     could be specified as:
                           MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any
                     Note that the ampersand character has a special meaning
                     in many shells and should generally be escaped.
             Note that the order of MAC addresses (destination first, source
             second) is the same as on the wire, but the opposite of the one
             used for IP addresses.
     mac-type mac-type
             Matches packets whose Ethernet Type field corresponds to one of
             those specified as argument.  mac-type is specified in the same
             way as port numbers (i.e., one or more comma-separated single
             values or ranges).  You can use symbolic names for known values
             such as vlan, ipv4, ipv6.        Values can be entered as decimal or
             hexadecimal (if prefixed by 0x), and they are always printed as
             hexadecimal (unless the -N option is used, in which case symbolic
             resolution will be attempted).
     proto protocol
             Matches packets with the corresponding IP protocol.
     recv | xmit | via {ifX | if* | ipno | any}
             Matches packets received, transmitted or going through, respec-
             tively, the interface specified by exact name (ifX), by device
             name (if*), by IP address, or through some interface.
             The via keyword causes the interface to always be checked.  If
             recv or xmit is used instead of via, then only the receive or
             transmit interface (respectively) is checked.  By specifying
             both, it is possible to match packets based on both receive and
             transmit interface, e.g.:
                   ipfw add deny ip from any to any out recv ed0 xmit ed1
             The recv interface can be tested on either incoming or outgoing
             packets, while the xmit interface can only be tested on outgoing
             packets.  So out is required (and in is invalid) whenever xmit is
             used.
             A packet may not have a receive or transmit interface: packets
             originating from the local host have no receive interface, while
             packets destined for the local host have no transmit interface.
     setup   Matches TCP packets that have the SYN bit set but no ACK bit.
             This is the short form of ``tcpflags syn,!ack''.
     src-ip ip-address
             Matches IPv4 packets whose source IP is one of the address(es)
             specified as an argument.
     src-ip6 ip6-address
             Matches IPv6 packets whose source IP is one of the address(es)
             specified as an argument.
     src-port ports
             Matches IP packets whose source port is one of the port(s) speci-
             fied as argument.
     tcpack ack
             TCP packets only.        Match if the TCP header acknowledgment number
             field is set to ack.
     tcpdatalen tcpdatalen-list
             Matches TCP packets whose length of TCP data is tcpdatalen-list,
             which is either a single value or a list of values or ranges
             specified in the same way as ports.
     tcpflags spec
             TCP packets only.        Match if the TCP header contains the comma
             separated list of flags specified in spec.  The supported TCP
             flags are:
             fin, syn, rst, psh, ack and urg.  The absence of a particular
             flag may be denoted with a `!'.  A rule which contains a tcpflags
             specification can never match a fragmented packet which has a
             non-zero offset.  See the frag option for details on matching
             fragmented packets.
     tcpseq seq
             TCP packets only.        Match if the TCP header sequence number field
             is set to seq.
     tcpwin win
             TCP packets only.        Match if the TCP header window field is set to
             win.
     tcpoptions spec
             TCP packets only.        Match if the TCP header contains the comma
             separated list of options specified in spec.  The supported TCP
             options are:
             mss (maximum segment size), window (tcp window advertisement),
             sack (selective ack), ts (rfc1323 timestamp) and cc (rfc1644
             t/tcp connection count).  The absence of a particular option may
             be denoted with a `!'.
     uid user
             Match all TCP or UDP packets sent by or received for a user.  A
             user may be matched by name or identification number.  This
             option should be used only if debug.mpsafenet=0 to avoid possible
             deadlocks due to layering violations in its implementation.
     verrevpath
             For incoming packets, a routing table lookup is done on the
             packet's source address.  If the interface on which the packet
             entered the system matches the outgoing interface for the route,
             the packet matches.  If the interfaces do not match up, the
             packet does not match.  All outgoing packets or packets with no
             incoming interface match.
             The name and functionality of the option is intentionally similar
             to the Cisco IOS command:
                   ip verify unicast reverse-path
             This option can be used to make anti-spoofing rules to reject all
             packets with source addresses not from this interface.  See also
             the option antispoof.
     versrcreach
             For incoming packets, a routing table lookup is done on the
             packet's source address.  If a route to the source address
             exists, but not the default route or a blackhole/reject route,
             the packet matches.  Otherwise, the packet does not match.  All
             outgoing packets match.
             The name and functionality of the option is intentionally similar
             to the Cisco IOS command:
                   ip verify unicast source reachable-via any
             This option can be used to make anti-spoofing rules to reject all
             packets whose source address is unreachable.
     antispoof
             For incoming packets, the packet's source address is checked if
             it belongs to a directly connected network.  If the network is
             directly connected, then the interface the packet came on in is
             compared to the interface the network is connected to.  When
             incoming interface and directly connected interface are not the
             same, the packet does not match.  Otherwise, the packet does
             match.  All outgoing packets match.
             This option can be used to make anti-spoofing rules to reject all
             packets that pretend to be from a directly connected network but
             do not come in through that interface.  This option is similar to
             but more restricted than verrevpath because it engages only on
             packets with source addresses of directly connected networks
             instead of all source addresses.
LOOKUP TABLES
     Lookup tables are useful to handle large sparse address sets, typically
     from a hundred to several thousands of entries.  There may be up to 128
     different lookup tables, numbered 0 to 127.
     Each entry is represented by an addr[/masklen] and will match all
     addresses with base addr (specified as an IP address or a hostname) and
     mask width of masklen bits.  If masklen is not specified, it defaults to
     32.  When looking up an IP address in a table, the most specific entry
     will match.  Associated with each entry is a 32-bit unsigned value, which
     can optionally be checked by a rule matching code.  When adding an entry,
     if value is not specified, it defaults to 0.
     An entry can be added to a table (add), removed from a table (delete), a
     table can be examined (list) or flushed (flush).
     Internally, each table is stored in a Radix tree, the same way as the
     routing table (see
route(4)
).
     Lookup tables currently support IPv4 addresses only.
     The tablearg feature provides the ability to use a value, looked up in
     the table, as the argument for a rule action.  This can significantly
     reduce number of rules in some configurations.  The tablearg argument can
     be used with the following actions: pipe, queue, divert, tee, netgraph,
     ngtee.  See the EXAMPLES Section for example usage of tables and the
     tablearg keyword.
SETS OF
RULES
     Each rule belongs to one of 32 different sets , numbered 0 to 31.        Set 31
     is reserved for the default rule.
     By default, rules are put in set 0, unless you use the set N attribute
     when entering a new rule.        Sets can be individually and atomically
     enabled or disabled, so this mechanism permits an easy way to store mul-
     tiple configurations of the firewall and quickly (and atomically) switch
     between them.  The command to enable/disable sets is
           ipfw set [disable number ...] [enable number ...]
     where multiple enable or disable sections can be specified.  Command exe-
     cution is atomic on all the sets specified in the command.  By default,
     all sets are enabled.
     When you disable a set, its rules behave as if they do not exist in the
     firewall configuration, with only one exception:
           dynamic rules created from a rule before it had been disabled will
           still be active until they expire.  In order to delete dynamic
           rules you have to explicitly delete the parent rule which generated
           them.
     The set number of rules can be changed with the command
           ipfw set move {rule rule-number | old-set} to new-set
     Also, you can atomically swap two rulesets with the command
           ipfw set swap first-set second-set
     See the EXAMPLES Section on some possible uses of sets of rules.
STATEFUL FIREWALL
     Stateful operation is a way for the firewall to dynamically create rules
     for specific flows when packets that match a given pattern are detected.
     Support for stateful operation comes through the check-state, keep-state
     and limit options of rules.
     Dynamic rules are created when a packet matches a keep-state or limit
     rule, causing the creation of a dynamic rule which will match all and
     only packets with a given protocol between a src-ip/src-port
     dst-ip/dst-port pair of addresses (src and dst are used here only to
     denote the initial match addresses, but they are completely equivalent
     afterwards).  Dynamic rules will be checked at the first check-state,
     keep-state or limit occurrence, and the action performed upon a match
     will be the same as in the parent rule.
     Note that no additional attributes other than protocol and IP addresses
     and ports are checked on dynamic rules.
     The typical use of dynamic rules is to keep a closed firewall configura-
     tion, but let the first TCP SYN packet from the inside network install a
     dynamic rule for the flow so that packets belonging to that session will
     be allowed through the firewall:
           ipfw add check-state
           ipfw add allow tcp from my-subnet to any setup keep-state
           ipfw add deny tcp from any to any
     A similar approach can be used for UDP, where an UDP packet coming from
     the inside will install a dynamic rule to let the response through the
     firewall:
           ipfw add check-state
           ipfw add allow udp from my-subnet to any keep-state
           ipfw add deny udp from any to any
     Dynamic rules expire after some time, which depends on the status of the
     flow and the setting of some sysctl variables.  See Section SYSCTL
     VARIABLES for more details.  For TCP sessions, dynamic rules can be
     instructed to periodically send keepalive packets to refresh the state of
     the rule when it is about to expire.
     See Section EXAMPLES for more examples on how to use dynamic rules.
TRAFFIC SHAPER
(DUMMYNET) CONFIGURATION
     ipfw is also the user interface for the
dummynet(4)
traffic shaper.
     dummynet operates by first using the firewall to classify packets and
     divide them into flows, using any match pattern that can be used in ipfw
     rules.  Depending on local policies, a flow can contain packets for a
     single TCP connection, or from/to a given host, or entire subnet, or a
     protocol type, etc.
     Packets belonging to the same flow are then passed to either of two dif-
     ferent objects, which implement the traffic regulation:
         pipe         A pipe emulates a link with given bandwidth, propagation
                 delay, queue size and packet loss rate.  Packets are queued
                 in front of the pipe as they come out from the classifier,
                 and then transferred to the pipe according to the pipe's
                 parameters.
         queue         A queue is an abstraction used to implement the WF2Q+ (Worst-
                 case Fair Weighted Fair Queueing) policy, which is an effi-
                 cient variant of the WFQ policy.
                 The queue associates a weight and a reference pipe to each
                 flow, and then all backlogged (i.e., with packets queued)
                 flows linked to the same pipe share the pipe's bandwidth pro-
                 portionally to their weights.        Note that weights are not pri-
                 orities; a flow with a lower weight is still guaranteed to
                 get its fraction of the bandwidth even if a flow with a
                 higher weight is permanently backlogged.
     In practice, pipes can be used to set hard limits to the bandwidth that a
     flow can use, whereas queues can be used to determine how different flow
     share the available bandwidth.
     The pipe and queue configuration commands are the following:
           pipe number config pipe-configuration
           queue number config queue-configuration
     The following parameters can be configured for a pipe:
     bw bandwidth | device
             Bandwidth, measured in [K|M]{bit/s|Byte/s}.
             A value of 0 (default) means unlimited bandwidth.        The unit must
             immediately follow the number, as in
                   ipfw pipe 1 config bw 300Kbit/s
             If a device name is specified instead of a numeric value, as in
                   ipfw pipe 1 config bw tun0
             then the transmit clock is supplied by the specified device.  At
             the moment only the
tun(4)
device supports this functionality,
             for use in conjunction with
ppp(8)
.
     delay ms-delay
             Propagation delay, measured in milliseconds.  The value is
             rounded to the next multiple of the clock tick (typically 10ms,
             but it is a good practice to run kernels with ``options HZ=1000''
             to reduce the granularity to 1ms or less).  Default value is 0,
             meaning no delay.
     The following parameters can be configured for a queue:
     pipe pipe_nr
             Connects a queue to the specified pipe.  Multiple queues (with
             the same or different weights) can be connected to the same pipe,
             which specifies the aggregate rate for the set of queues.
     weight weight
             Specifies the weight to be used for flows matching this queue.
             The weight must be in the range 1..100, and defaults to 1.
     Finally, the following parameters can be configured for both pipes and
     queues:
     buckets hash-table-size
           Specifies the size of the hash table used for storing the various
           queues.  Default value is 64 controlled by the
sysctl(8)
variable
           net.inet.ip.dummynet.hash_size, allowed range is 16 to 65536.
     mask mask-specifier
           Packets sent to a given pipe or queue by an ipfw rule can be fur-
           ther classified into multiple flows, each of which is then sent to
           a different dynamic pipe or queue.  A flow identifier is con-
           structed by masking the IP addresses, ports and protocol types as
           specified with the mask options in the configuration of the pipe or
           queue.  For each different flow identifier, a new pipe or queue is
           created with the same parameters as the original object, and match-
           ing packets are sent to it.
           Thus, when dynamic pipes are used, each flow will get the same
           bandwidth as defined by the pipe, whereas when dynamic queues are
           used, each flow will share the parent's pipe bandwidth evenly with
           other flows generated by the same queue (note that other queues
           with different weights might be connected to the same pipe).
           Available mask specifiers are a combination of one or more of the
           following:
           dst-ip mask, dst-ip6 mask, src-ip mask, src-ip6 mask, dst-port
           mask, src-port mask, flow-id mask, proto mask or all,
           where the latter means all bits in all fields are significant.
     noerror
           When a packet is dropped by a dummynet queue or pipe, the error is
           normally reported to the caller routine in the kernel, in the same
           way as it happens when a device queue fills up.  Setting this
           option reports the packet as successfully delivered, which can be
           needed for some experimental setups where you want to simulate loss
           or congestion at a remote router.
     plr packet-loss-rate
           Packet loss rate.  Argument packet-loss-rate is a floating-point
           number between 0 and 1, with 0 meaning no loss, 1 meaning 100%
           loss.  The loss rate is internally represented on 31 bits.
     queue {slots | sizeKbytes}
           Queue size, in slots or KBytes.  Default value is 50 slots, which
           is the typical queue size for Ethernet devices.  Note that for slow
           speed links you should keep the queue size short or your traffic
           might be affected by a significant queueing delay.  E.g., 50 max-
           sized ethernet packets (1500 bytes) mean 600Kbit or 20s of queue on
           a 30Kbit/s pipe.  Even worse effects can result if you get packets
           from an interface with a much larger MTU, e.g. the loopback inter-
           face with its 16KB packets.
     red | gred w_q/min_th/max_th/max_p
           Make use of the RED (Random Early Detection) queue management algo-
           rithm.  w_q and max_p are floating point numbers between 0 and 1 (0
           not included), while min_th and max_th are integer numbers specify-
           ing thresholds for queue management (thresholds are computed in
           bytes if the queue has been defined in bytes, in slots otherwise).
           The
dummynet(4)
also supports the gentle RED variant (gred).  Three
          
sysctl(8)
variables can be used to control the RED behaviour:
           net.inet.ip.dummynet.red_lookup_depth
                   specifies the accuracy in computing the average queue when
                   the link is idle (defaults to 256, must be greater than
                   zero)
           net.inet.ip.dummynet.red_avg_pkt_size
                   specifies the expected average packet size (defaults to
                   512, must be greater than zero)
           net.inet.ip.dummynet.red_max_pkt_size
                   specifies the expected maximum packet size, only used when
                   queue thresholds are in bytes (defaults to 1500, must be
                   greater than zero).
     When used with IPv6 data, dummynet currently has several limitations.
     First, debug.mpsafenet=0 must be set.  Second, the information necessi-
     cary to route link-local packets to an interface is not avalable after
     processing by dummynet so those packets are dropped in the output path.
     Care should be taken to insure that link-local packets are not passed to
     dummynet.
CHECKLIST
     Here are some important points to consider when designing your rules:
     ·         Remember that you filter both packets going in and out.  Most connec-
         tions need packets going in both directions.
     ·         Remember to test very carefully.  It is a good idea to be near the
         console when doing this.  If you cannot be near the console, use an
         auto-recovery script such as the one in
         /usr/share/examples/ipfw/change_rules.sh.
     ·         Do not forget the loopback interface.
FINE POINTS
     ·         There are circumstances where fragmented datagrams are uncondition-
         ally dropped.        TCP packets are dropped if they do not contain at
         least 20 bytes of TCP header, UDP packets are dropped if they do not
         contain a full 8 byte UDP header, and ICMP packets are dropped if
         they do not contain 4 bytes of ICMP header, enough to specify the
         ICMP type, code, and checksum.  These packets are simply logged as
         ``pullup failed'' since there may not be enough good data in the
         packet to produce a meaningful log entry.
     ·         Another type of packet is unconditionally dropped, a TCP packet with
         a fragment offset of one.  This is a valid packet, but it only has
         one use, to try to circumvent firewalls.  When logging is enabled,
         these packets are reported as being dropped by rule -1.
     ·         If you are logged in over a network, loading the
kld(4)
version of
         ipfw is probably not as straightforward as you would think.  I recom-
         mend the following command line:
               kldload ipfw && \
               ipfw add 32000 allow ip from any to any
         Along the same lines, doing an
               ipfw flush
         in similar surroundings is also a bad idea.
     ·         The ipfw filter list may not be modified if the system security level
         is set to 3 or higher (see
init(8)
for information on system security
         levels).
PACKET DIVERSION
     A
divert(4)
socket bound to the specified port will receive all packets
     diverted to that port.  If no socket is bound to the destination port, or
     if the divert module is not loaded, or if the kernel was not compiled
     with divert socket support, the packets are dropped.
SYSCTL VARIABLES
     A set of
sysctl(8)
variables controls the behaviour of the firewall and
     associated modules (dummynet, bridge).  These are shown below together
     with their default value (but always check with the
sysctl(8)
command
     what value is actually in use) and meaning:
     net.inet.ip.dummynet.expire: 1
             Lazily delete dynamic pipes/queue once they have no pending traf-
             fic.  You can disable this by setting the variable to 0, in which
             case the pipes/queues will only be deleted when the threshold is
             reached.
     net.inet.ip.dummynet.hash_size: 64
             Default size of the hash table used for dynamic pipes/queues.
             This value is used when no buckets option is specified when con-
             figuring a pipe/queue.
     net.inet.ip.dummynet.max_chain_len: 16
             Target value for the maximum number of pipes/queues in a hash
             bucket.  The product max_chain_len*hash_size is used to determine
             the threshold over which empty pipes/queues will be expired even
             when net.inet.ip.dummynet.expire=0.
     net.inet.ip.dummynet.red_lookup_depth: 256
     net.inet.ip.dummynet.red_avg_pkt_size: 512
     net.inet.ip.dummynet.red_max_pkt_size: 1500
             Parameters used in the computations of the drop probability for
             the RED algorithm.
     net.inet.ip.fw.autoinc_step: 100
             Delta between rule numbers when auto-generating them.  The value
             must be in the range 1..1000.
     net.inet.ip.fw.curr_dyn_buckets: net.inet.ip.fw.dyn_buckets
             The current number of buckets in the hash table for dynamic rules
             (readonly).
     net.inet.ip.fw.debug: 1
             Controls debugging messages produced by ipfw.
     net.inet.ip.fw.dyn_buckets: 256
             The number of buckets in the hash table for dynamic rules.  Must
             be a power of 2, up to 65536.  It only takes effect when all
             dynamic rules have expired, so you are advised to use a flush
             command to make sure that the hash table is resized.
     net.inet.ip.fw.dyn_count: 3
             Current number of dynamic rules (read-only).
     net.inet.ip.fw.dyn_keepalive: 1
             Enables generation of keepalive packets for keep-state rules on
             TCP sessions.  A keepalive is generated to both sides of the con-
             nection every 5 seconds for the last 20 seconds of the lifetime
             of the rule.
     net.inet.ip.fw.dyn_max: 8192
             Maximum number of dynamic rules.  When you hit this limit, no
             more dynamic rules can be installed until old ones expire.
     net.inet.ip.fw.dyn_ack_lifetime: 300
     net.inet.ip.fw.dyn_syn_lifetime: 20
     net.inet.ip.fw.dyn_fin_lifetime: 1
     net.inet.ip.fw.dyn_rst_lifetime: 1
     net.inet.ip.fw.dyn_udp_lifetime: 5
     net.inet.ip.fw.dyn_short_lifetime: 30
             These variables control the lifetime, in seconds, of dynamic
             rules.  Upon the initial SYN exchange the lifetime is kept short,
             then increased after both SYN have been seen, then decreased
             again during the final FIN exchange or when a RST is received.
             Both dyn_fin_lifetime and dyn_rst_lifetime must be strictly lower
             than 5 seconds, the period of repetition of keepalives.  The
             firewall enforces that.
     net.inet.ip.fw.enable: 1
             Enables the firewall.  Setting this variable to 0 lets you run
             your machine without firewall even if compiled in.
     net.inet.ip.fw.one_pass: 1
             When set, the packet exiting from the
dummynet(4)
pipe or from
             
ng_ipfw(4)
node is not passed though the firewall again.  Other-
             wise, after an action, the packet is reinjected into the firewall
             at the next rule.
     net.inet.ip.fw.verbose: 1
             Enables verbose messages.
     net.inet.ip.fw.verbose_limit: 0
             Limits the number of messages produced by a verbose firewall.
     net.inet6.ip6.fw.deny_unknown_exthdrs: 1
             If enabled packets with unknown IPv6 Extension Headers will be
             denied.
     net.link.ether.ipfw: 0
             Controls whether layer-2 packets are passed to ipfw.  Default is
             no.
     net.link.ether.bridge_ipfw: 0
             Controls whether bridged packets are passed to ipfw.  Default is
             no.
EXAMPLES
     There are far too many possible uses of ipfw so this Section will only
     give a small set of examples.
   BASIC PACKET FILTERING
     This command adds an entry which denies all tcp packets from
     cracker.evil.org to the telnet port of wolf.tambov.su from being for-
     warded by the host:
           ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet
     This one disallows any connection from the entire cracker's network to my
     host:
           ipfw add deny ip from 123.45.67.0/24 to my.host.org
     A first and efficient way to limit access (not using dynamic rules) is
     the use of the following rules:
           ipfw add allow tcp from any to any established
           ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup
           ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup
           ...
           ipfw add deny tcp from any to any
     The first rule will be a quick match for normal TCP packets, but it will
     not match the initial SYN packet, which will be matched by the setup
     rules only for selected source/destination pairs.        All other SYN packets
     will be rejected by the final deny rule.
     If you administer one or more subnets, you can take advantage of the
     address sets and or-blocks and write extremely compact rulesets which
     selectively enable services to blocks of clients, as below:
           goodguys="{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }"
           badguys="10.1.2.0/24{8,38,60}"
           ipfw add allow ip from ${goodguys} to any
           ipfw add deny ip from ${badguys} to any
           ... normal policies ...
     The verrevpath option could be used to do automated anti-spoofing by
     adding the following to the top of a ruleset:
           ipfw add deny ip from any to any not verrevpath in
     This rule drops all incoming packets that appear to be coming to the sys-
     tem on the wrong interface.  For example, a packet with a source address
     belonging to a host on a protected internal network would be dropped if
     it tried to enter the system from an external interface.
     The antispoof option could be used to do similar but more restricted
     anti-spoofing by adding the following to the top of a ruleset:
           ipfw add deny ip from any to any not antispoof in
     This rule drops all incoming packets that appear to be coming from
     another directly connected system but on the wrong interface.  For exam-
     ple, a packet with a source address of 192.168.0.0/24 , configured on
     fxp0 , but coming in on fxp1 would be dropped.
   DYNAMIC RULES
     In order to protect a site from flood attacks involving fake TCP packets,
     it is safer to use dynamic rules:
           ipfw add check-state
           ipfw add deny tcp from any to any established
           ipfw add allow tcp from my-net to any setup keep-state
     This will let the firewall install dynamic rules only for those connec-
     tion which start with a regular SYN packet coming from the inside of our
     network.  Dynamic rules are checked when encountering the first
     check-state or keep-state rule.  A check-state rule should usually be
     placed near the beginning of the ruleset to minimize the amount of work
     scanning the ruleset.  Your mileage may vary.
     To limit the number of connections a user can open you can use the fol-
     lowing type of rules:
           ipfw add allow tcp from my-net/24 to any setup limit src-addr 10
           ipfw add allow tcp from any to me setup limit src-addr 4
     The former (assuming it runs on a gateway) will allow each host on a /24
     network to open at most 10 TCP connections.  The latter can be placed on
     a server to make sure that a single client does not use more than 4
     simultaneous connections.
     BEWARE: stateful rules can be subject to denial-of-service attacks by a
     SYN-flood which opens a huge number of dynamic rules.  The effects of
     such attacks can be partially limited by acting on a set of
sysctl(8)
     variables which control the operation of the firewall.
     Here is a good usage of the list command to see accounting records and
     timestamp information:
           ipfw -at list
     or in short form without timestamps:
           ipfw -a list
     which is equivalent to:
           ipfw show
     Next rule diverts all incoming packets from 192.168.2.0/24 to divert port
     5000:
           ipfw divert 5000 ip from 192.168.2.0/24 to any in
   TRAFFIC SHAPING
     The following rules show some of the applications of ipfw and
dummynet(4)
     for simulations and the like.
     This rule drops random incoming packets with a probability of 5%:
           ipfw add prob 0.05 deny ip from any to any in
     A similar effect can be achieved making use of dummynet pipes:
           ipfw add pipe 10 ip from any to any
           ipfw pipe 10 config plr 0.05
     We can use pipes to artificially limit bandwidth, e.g. on a machine act-
     ing as a router, if we want to limit traffic from local clients on
     192.168.2.0/24 we do:
           ipfw add pipe 1 ip from 192.168.2.0/24 to any out
           ipfw pipe 1 config bw 300Kbit/s queue 50KBytes
     note that we use the out modifier so that the rule is not used twice.
     Remember in fact that ipfw rules are checked both on incoming and outgo-
     ing packets.
     Should we want to simulate a bidirectional link with bandwidth limita-
     tions, the correct way is the following:
           ipfw add pipe 1 ip from any to any out
           ipfw add pipe 2 ip from any to any in
           ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes
           ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes
     The above can be very useful, e.g. if you want to see how your fancy Web
     page will look for a residential user who is connected only through a
     slow link.  You should not use only one pipe for both directions, unless
     you want to simulate a half-duplex medium (e.g. AppleTalk, Ethernet,
     IRDA).  It is not necessary that both pipes have the same configuration,
     so we can also simulate asymmetric links.
     Should we want to verify network performance with the RED queue manage-
     ment algorithm:
           ipfw add pipe 1 ip from any to any
           ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1
     Another typical application of the traffic shaper is to introduce some
     delay in the communication.  This can significantly affect applications
     which do a lot of Remote Procedure Calls, and where the round-trip-time
     of the connection often becomes a limiting factor much more than band-
     width:
           ipfw add pipe 1 ip from any to any out
           ipfw add pipe 2 ip from any to any in
           ipfw pipe 1 config delay 250ms bw 1Mbit/s
           ipfw pipe 2 config delay 250ms bw 1Mbit/s
     Per-flow queueing can be useful for a variety of purposes.  A very simple
     one is counting traffic:
           ipfw add pipe 1 tcp from any to any
           ipfw add pipe 1 udp from any to any
           ipfw add pipe 1 ip from any to any
           ipfw pipe 1 config mask all
     The above set of rules will create queues (and collect statistics) for
     all traffic.  Because the pipes have no limitations, the only effect is
     collecting statistics.  Note that we need 3 rules, not just the last one,
     because when ipfw tries to match IP packets it will not consider ports,
     so we would not see connections on separate ports as different ones.
     A more sophisticated example is limiting the outbound traffic on a net
     with per-host limits, rather than per-network limits:
           ipfw add pipe 1 ip from 192.168.2.0/24 to any out
           ipfw add pipe 2 ip from any to 192.168.2.0/24 in
           ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue
           20Kbytes
           ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue
           20Kbytes
   LOOKUP TABLES
     In the following example, we need to create several traffic bandwidth
     classes and we need different hosts/networks to fall into different
     classes.  We create one pipe for each class and configure them accord-
     ingly.  Then we create a single table and fill it with IP subnets and
     addresses.  For each subnet/host we set the argument equal to the number
     of the pipe that it should use.  Then we classify traffic using a single
     rule:
           ipfw pipe 1 config bw 1000Kbyte/s
           ipfw pipe 4 config bw 4000Kbyte/s
           ...
           ipfw table 1 add 192.168.2.0/24 1
           ipfw table 1 add 192.168.0.0/27 4
           ipfw table 1 add 192.168.0.2 1
           ...
           ipfw pipe tablearg ip from
table(1)
to any
   SETS OF RULES
     To add a set of rules atomically, e.g. set 18:
           ipfw set disable 18
           ipfw add NN set 18 ...          # repeat as needed
           ipfw set enable 18
     To delete a set of rules atomically the command is simply:
           ipfw delete set 18
     To test a ruleset and disable it and regain control if something goes
     wrong:
           ipfw set disable 18
           ipfw add NN set 18 ...          # repeat as needed
           ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18
     Here if everything goes well, you press control-C before the "sleep" ter-
     minates, and your ruleset will be left active.  Otherwise, e.g. if you
     cannot access your box, the ruleset will be disabled after the sleep ter-
     minates thus restoring the previous situation.
SEE ALSO
     
cpp(1)
,
m4(1)
,
altq(4)
,
bridge(4)
,
divert(4)
,
dummynet(4)
,
ip(4)
,
     
ipfirewall(4)
,
ng_ipfw(4)
,
protocols(5)
,
services(5)
,
init(8)
,
     
kldload(8)
,
reboot(8)
,
sysctl(8)
,
syslogd(8)
HISTORY
     The ipfw utility first appeared in FreeBSD 2.0.  
dummynet(4)
was intro-
     duced in FreeBSD 2.2.8.  Stateful extensions were introduced in
     FreeBSD 4.0.  ipfw2 was introduced in Summer 2002.
AUTHORS
     Ugen J. S. Antsilevich,
     Poul-Henning Kamp,
     Alex Nash,
     Archie Cobbs,
     Luigi Rizzo.
     API based upon code written by Daniel Boulet for BSDI.
     Work on
dummynet(4)
traffic shaper supported by Akamba Corp.
BUGS
     Use of dummynet with IPv6 requires that debug.mpsafenet be set to 0.
     The syntax has grown over the years and sometimes it might be confusing.
     Unfortunately, backward compatibility prevents cleaning up mistakes made
     in the definition of the syntax.
     !!! WARNING !!!
     Misconfiguring the firewall can put your computer in an unusable state,
     possibly shutting down network services and requiring console access to
     regain control of it.
     Incoming packet fragments diverted by divert are reassembled before
     delivery to the socket.  The action used on those packet is the one from
     the rule which matches the first fragment of the packet.
     Packets diverted to userland, and then reinserted by a userland process
     may lose various packet attributes.  The packet source interface name
     will be preserved if it is shorter than 8 bytes and the userland process
     saves and reuses the sockaddr_in (as does
natd(8)
); otherwise, it may be
     lost.  If a packet is reinserted in this manner, later rules may be
     incorrectly applied, making the order of divert rules in the rule
     sequence very important.
     Dummynet drops all packets with IPv6 link-local addresses.
     Rules using uid or gid may not behave as expected.  In particular, incom-
     ing SYN packets may have no uid or gid associated with them since they do
     not yet belong to a TCP connection, and the uid/gid associated with a
     packet may not be as expected if the associated process calls
setuid(2)
     or similar system calls.
     Rules which use uid, gid or jail based matching should be used only if
     debug.mpsafenet=0 to avoid possible deadlocks due to layering violations
     in its implementation.
FreeBSD 4.11                       January 16, 2006                   FreeBSD 4.11


本文来自ChinaUnix博客，如果查看原文请点：http://blog.chinaunix.net/u/27478/showart_209744.html