xref: /freebsd/sbin/ipfw/ipfw.8 (revision 4a5216a6dc0c3ce4cf5f2d3ee8af0c3ff3402c4f)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd September 27, 2008
5.Dt IPFW 8
6.Os
7.Sh NAME
8.Nm ipfw
9.Nd IP firewall and traffic shaper control program
10.Sh SYNOPSIS
11.Nm
12.Op Fl cq
13.Cm add
14.Ar rule
15.Nm
16.Op Fl acdefnNStT
17.Op Cm set Ar N
18.Brq Cm list | show
19.Op Ar rule | first-last ...
20.Nm
21.Op Fl f | q
22.Op Cm set Ar N
23.Cm flush
24.Nm
25.Op Fl q
26.Op Cm set Ar N
27.Brq Cm delete | zero | resetlog
28.Op Ar number ...
29.Nm
30.Cm enable
31.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
32.Nm
33.Cm disable
34.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
35.Pp
36.Nm
37.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
38.Nm
39.Cm set move
40.Op Cm rule
41.Ar number Cm to Ar number
42.Nm
43.Cm set swap Ar number number
44.Nm
45.Cm set show
46.Pp
47.Nm
48.Cm table Ar number Cm add Ar addr Ns Oo / Ns Ar masklen Oc Op Ar value
49.Nm
50.Cm table Ar number Cm delete Ar addr Ns Op / Ns Ar masklen
51.Nm
52.Cm table
53.Brq Ar number | all
54.Cm flush
55.Nm
56.Cm table
57.Brq Ar number | all
58.Cm list
59.Pp
60.Nm
61.Brq Cm pipe | queue
62.Ar number
63.Cm config
64.Ar config-options
65.Nm
66.Op Fl s Op Ar field
67.Brq Cm pipe | queue
68.Brq Cm delete | list | show
69.Op Ar number ...
70.Pp
71.Nm
72.Cm nat
73.Ar number
74.Cm config
75.Ar config-options
76.Pp
77.Nm
78.Op Fl cfnNqS
79.Oo
80.Fl p Ar preproc
81.Oo
82.Ar preproc-flags
83.Oc
84.Oc
85.Ar pathname
86.Sh DESCRIPTION
87The
88.Nm
89utility is the user interface for controlling the
90.Xr ipfw 4
91firewall and the
92.Xr dummynet 4
93traffic shaper in
94.Fx .
95.Pp
96An
97.Nm
98configuration, or
99.Em ruleset ,
100is made of a list of
101.Em rules
102numbered from 1 to 65535.
103Packets are passed to
104.Nm
105from a number of different places in the protocol stack
106(depending on the source and destination of the packet,
107it is possible that
108.Nm
109is invoked multiple times on the same packet).
110The packet passed to the firewall is compared
111against each of the rules in the firewall
112.Em ruleset .
113When a match is found, the action corresponding to the
114matching rule is performed.
115.Pp
116Depending on the action and certain system settings, packets
117can be reinjected into the firewall at some rule after the
118matching one for further processing.
119.Pp
120An
121.Nm
122ruleset always includes a
123.Em default
124rule (numbered 65535) which cannot be modified or deleted,
125and matches all packets.
126The action associated with the
127.Em default
128rule can be either
129.Cm deny
130or
131.Cm allow
132depending on how the kernel is configured.
133.Pp
134If the ruleset includes one or more rules with the
135.Cm keep-state
136or
137.Cm limit
138option, then
139.Nm
140assumes a
141.Em stateful
142behaviour, i.e., upon a match it will create dynamic rules matching
143the exact parameters (addresses and ports) of the matching packet.
144.Pp
145These dynamic rules, which have a limited lifetime, are checked
146at the first occurrence of a
147.Cm check-state ,
148.Cm keep-state
149or
150.Cm limit
151rule, and are typically used to open the firewall on-demand to
152legitimate traffic only.
153See the
154.Sx STATEFUL FIREWALL
155and
156.Sx EXAMPLES
157Sections below for more information on the stateful behaviour of
158.Nm .
159.Pp
160All rules (including dynamic ones) have a few associated counters:
161a packet count, a byte count, a log count and a timestamp
162indicating the time of the last match.
163Counters can be displayed or reset with
164.Nm
165commands.
166.Pp
167Rules can be added with the
168.Cm add
169command; deleted individually or in groups with the
170.Cm delete
171command, and globally (except those in set 31) with the
172.Cm flush
173command; displayed, optionally with the content of the
174counters, using the
175.Cm show
176and
177.Cm list
178commands.
179Finally, counters can be reset with the
180.Cm zero
181and
182.Cm resetlog
183commands.
184.Pp
185Also, each rule belongs to one of 32 different
186.Em sets
187, and there are
188.Nm
189commands to atomically manipulate sets, such as enable,
190disable, swap sets, move all rules in a set to another
191one, delete all rules in a set.
192These can be useful to
193install temporary configurations, or to test them.
194See Section
195.Sx SETS OF RULES
196for more information on
197.Em sets .
198.Pp
199The following options are available:
200.Bl -tag -width indent
201.It Fl a
202While listing, show counter values.
203The
204.Cm show
205command just implies this option.
206.It Fl b
207Only show the action and the comment, not the body of a rule.
208Implies
209.Fl c .
210.It Fl c
211When entering or showing rules, print them in compact form,
212i.e., without the optional "ip from any to any" string
213when this does not carry any additional information.
214.It Fl d
215While listing, show dynamic rules in addition to static ones.
216.It Fl e
217While listing, if the
218.Fl d
219option was specified, also show expired dynamic rules.
220.It Fl f
221Do not ask for confirmation for commands that can cause problems
222if misused,
223.No i.e. Cm flush .
224If there is no tty associated with the process, this is implied.
225.It Fl i
226While listing a table (see the
227.Sx LOOKUP TABLES
228section below for more information on lookup tables), format values
229as IP addresses. By default, values are shown as integers.
230.It Fl n
231Only check syntax of the command strings, without actually passing
232them to the kernel.
233.It Fl N
234Try to resolve addresses and service names in output.
235.It Fl q
236While
237.Cm add Ns ing ,
238.Cm zero Ns ing ,
239.Cm resetlog Ns ging
240or
241.Cm flush Ns ing ,
242be quiet about actions
243(implies
244.Fl f ) .
245This is useful for adjusting rules by executing multiple
246.Nm
247commands in a script
248(e.g.,
249.Ql sh\ /etc/rc.firewall ) ,
250or by processing a file of many
251.Nm
252rules across a remote login session.
253It also stops a table add or delete
254from failing if the entry already exists or is not present.
255If a
256.Cm flush
257is performed in normal (verbose) mode (with the default kernel
258configuration), it prints a message.
259Because all rules are flushed, the message might not be delivered
260to the login session, causing the remote login session to be closed
261and the remainder of the ruleset to not be processed.
262Access to the console would then be required to recover.
263.It Fl S
264While listing rules, show the
265.Em set
266each rule belongs to.
267If this flag is not specified, disabled rules will not be
268listed.
269.It Fl s Op Ar field
270While listing pipes, sort according to one of the four
271counters (total or current packets or bytes).
272.It Fl t
273While listing, show last match timestamp (converted with ctime()).
274.It Fl T
275While listing, show last match timestamp (as seconds from the epoch).
276This form can be more convenient for postprocessing by scripts.
277.El
278.Pp
279To ease configuration, rules can be put into a file which is
280processed using
281.Nm
282as shown in the last synopsis line.
283An absolute
284.Ar pathname
285must be used.
286The file will be read line by line and applied as arguments to the
287.Nm
288utility.
289.Pp
290Optionally, a preprocessor can be specified using
291.Fl p Ar preproc
292where
293.Ar pathname
294is to be piped through.
295Useful preprocessors include
296.Xr cpp 1
297and
298.Xr m4 1 .
299If
300.Ar preproc
301does not start with a slash
302.Pq Ql /
303as its first character, the usual
304.Ev PATH
305name search is performed.
306Care should be taken with this in environments where not all
307file systems are mounted (yet) by the time
308.Nm
309is being run (e.g.\& when they are mounted over NFS).
310Once
311.Fl p
312has been specified, any additional arguments as passed on to the preprocessor
313for interpretation.
314This allows for flexible configuration files (like conditionalizing
315them on the local hostname) and the use of macros to centralize
316frequently required arguments like IP addresses.
317.Pp
318The
319.Nm
320.Cm pipe
321and
322.Cm queue
323commands are used to configure the traffic shaper, as shown in the
324.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
325Section below.
326.Pp
327If the world and the kernel get out of sync the
328.Nm
329ABI may break, preventing you from being able to add any rules.
330This can
331adversely effect the booting process.
332You can use
333.Nm
334.Cm disable
335.Cm firewall
336to temporarily disable the firewall to regain access to the network,
337allowing you to fix the problem.
338.Sh PACKET FLOW
339A packet is checked against the active ruleset in multiple places
340in the protocol stack, under control of several sysctl variables.
341These places and variables are shown below, and it is important to
342have this picture in mind in order to design a correct ruleset.
343.Bd -literal -offset indent
344       ^    to upper layers    V
345       |                       |
346       +----------->-----------+
347       ^                       V
348 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
349       |                       |
350       ^                       V
351 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
352       |                       |
353       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
354       ^                       V
355       |      to devices       |
356.Ed
357.Pp
358As can be noted from the above picture, the number of
359times the same packet goes through the firewall can
360vary between 0 and 4 depending on packet source and
361destination, and system configuration.
362.Pp
363Note that as packets flow through the stack, headers can be
364stripped or added to it, and so they may or may not be available
365for inspection.
366E.g., incoming packets will include the MAC header when
367.Nm
368is invoked from
369.Cm ether_demux() ,
370but the same packets will have the MAC header stripped off when
371.Nm
372is invoked from
373.Cm ip_input()
374or
375.Cm ip6_input() .
376.Pp
377Also note that each packet is always checked against the complete ruleset,
378irrespective of the place where the check occurs, or the source of the packet.
379If a rule contains some match patterns or actions which are not valid
380for the place of invocation (e.g.\& trying to match a MAC header within
381.Cm ip_input
382or
383.Cm ip6_input ),
384the match pattern will not match, but a
385.Cm not
386operator in front of such patterns
387.Em will
388cause the pattern to
389.Em always
390match on those packets.
391It is thus the responsibility of
392the programmer, if necessary, to write a suitable ruleset to
393differentiate among the possible places.
394.Cm skipto
395rules can be useful here, as an example:
396.Bd -literal -offset indent
397# packets from ether_demux or bdg_forward
398ipfw add 10 skipto 1000 all from any to any layer2 in
399# packets from ip_input
400ipfw add 10 skipto 2000 all from any to any not layer2 in
401# packets from ip_output
402ipfw add 10 skipto 3000 all from any to any not layer2 out
403# packets from ether_output_frame
404ipfw add 10 skipto 4000 all from any to any layer2 out
405.Ed
406.Pp
407(yes, at the moment there is no way to differentiate between
408ether_demux and bdg_forward).
409.Sh SYNTAX
410In general, each keyword or argument must be provided as
411a separate command line argument, with no leading or trailing
412spaces.
413Keywords are case-sensitive, whereas arguments may
414or may not be case-sensitive depending on their nature
415(e.g.\& uid's are, hostnames are not).
416.Pp
417In
418.Nm ipfw2
419you can introduce spaces after commas ',' to make
420the line more readable.
421You can also put the entire
422command (including flags) into a single argument.
423E.g., the following forms are equivalent:
424.Bd -literal -offset indent
425ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
426ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
427ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
428.Ed
429.Sh RULE FORMAT
430The format of
431.Nm
432rules is the following:
433.Bd -ragged -offset indent
434.Bk -words
435.Op Ar rule_number
436.Op Cm set Ar set_number
437.Op Cm prob Ar match_probability
438.Ar action
439.Op Cm log Op Cm logamount Ar number
440.Op Cm altq Ar queue
441.Oo
442.Bro Cm tag | untag
443.Brc Ar number
444.Oc
445.Ar body
446.Ek
447.Ed
448.Pp
449where the body of the rule specifies which information is used
450for filtering packets, among the following:
451.Pp
452.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
453.It Layer-2 header fields
454When available
455.It IPv4 and IPv6 Protocol
456TCP, UDP, ICMP, etc.
457.It Source and dest. addresses and ports
458.It Direction
459See Section
460.Sx PACKET FLOW
461.It Transmit and receive interface
462By name or address
463.It Misc. IP header fields
464Version, type of service, datagram length, identification,
465fragment flag (non-zero IP offset),
466Time To Live
467.It IP options
468.It IPv6 Extension headers
469Fragmentation, Hop-by-Hop options,
470Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
471.It IPv6 Flow-ID
472.It Misc. TCP header fields
473TCP flags (SYN, FIN, ACK, RST, etc.),
474sequence number, acknowledgment number,
475window
476.It TCP options
477.It ICMP types
478for ICMP packets
479.It ICMP6 types
480for ICMP6 packets
481.It User/group ID
482When the packet can be associated with a local socket.
483.It Divert status
484Whether a packet came from a divert socket (e.g.,
485.Xr natd 8 ) .
486.It Fib annotation state
487Whether a packet has been tagged for using a specific FIB (routing table)
488in future forwarding decisions.
489.El
490.Pp
491Note that some of the above information, e.g.\& source MAC or IP addresses and
492TCP/UDP ports, could easily be spoofed, so filtering on those fields
493alone might not guarantee the desired results.
494.Bl -tag -width indent
495.It Ar rule_number
496Each rule is associated with a
497.Ar rule_number
498in the range 1..65535, with the latter reserved for the
499.Em default
500rule.
501Rules are checked sequentially by rule number.
502Multiple rules can have the same number, in which case they are
503checked (and listed) according to the order in which they have
504been added.
505If a rule is entered without specifying a number, the kernel will
506assign one in such a way that the rule becomes the last one
507before the
508.Em default
509rule.
510Automatic rule numbers are assigned by incrementing the last
511non-default rule number by the value of the sysctl variable
512.Ar net.inet.ip.fw.autoinc_step
513which defaults to 100.
514If this is not possible (e.g.\& because we would go beyond the
515maximum allowed rule number), the number of the last
516non-default value is used instead.
517.It Cm set Ar set_number
518Each rule is associated with a
519.Ar set_number
520in the range 0..31.
521Sets can be individually disabled and enabled, so this parameter
522is of fundamental importance for atomic ruleset manipulation.
523It can be also used to simplify deletion of groups of rules.
524If a rule is entered without specifying a set number,
525set 0 will be used.
526.br
527Set 31 is special in that it cannot be disabled,
528and rules in set 31 are not deleted by the
529.Nm ipfw flush
530command (but you can delete them with the
531.Nm ipfw delete set 31
532command).
533Set 31 is also used for the
534.Em default
535rule.
536.It Cm prob Ar match_probability
537A match is only declared with the specified probability
538(floating point number between 0 and 1).
539This can be useful for a number of applications such as
540random packet drop or
541(in conjunction with
542.Nm dummynet )
543to simulate the effect of multiple paths leading to out-of-order
544packet delivery.
545.Pp
546Note: this condition is checked before any other condition, including
547ones such as keep-state or check-state which might have side effects.
548.It Cm log Op Cm logamount Ar number
549When a packet matches a rule with the
550.Cm log
551keyword, a message will be
552logged to
553.Xr syslogd 8
554with a
555.Dv LOG_SECURITY
556facility.
557The logging only occurs if the sysctl variable
558.Va net.inet.ip.fw.verbose
559is set to 1
560(which is the default when the kernel is compiled with
561.Dv IPFIREWALL_VERBOSE )
562and the number of packets logged so far for that
563particular rule does not exceed the
564.Cm logamount
565parameter.
566If no
567.Cm logamount
568is specified, the limit is taken from the sysctl variable
569.Va net.inet.ip.fw.verbose_limit .
570In both cases, a value of 0 removes the logging limit.
571.Pp
572Once the limit is reached, logging can be re-enabled by
573clearing the logging counter or the packet counter for that entry, see the
574.Cm resetlog
575command.
576.Pp
577Note: logging is done after all other packet matching conditions
578have been successfully verified, and before performing the final
579action (accept, deny, etc.) on the packet.
580.It Cm tag Ar number
581When a packet matches a rule with the
582.Cm tag
583keyword, the numeric tag for the given
584.Ar number
585in the range 1..65534 will be attached to the packet.
586The tag acts as an internal marker (it is not sent out over
587the wire) that can be used to identify these packets later on.
588This can be used, for example, to provide trust between interfaces
589and to start doing policy-based filtering.
590A packet can have mutiple tags at the same time.
591Tags are "sticky", meaning once a tag is applied to a packet by a
592matching rule it exists until explicit removal.
593Tags are kept with the packet everywhere within the kernel, but are
594lost when packet leaves the kernel, for example, on transmitting
595packet out to the network or sending packet to a
596.Xr divert 4
597socket.
598.Pp
599To check for previously applied tags, use the
600.Cm tagged
601rule option.
602To delete previously applied tag, use the
603.Cm untag
604keyword.
605.Pp
606Note: since tags are kept with the packet everywhere in kernelspace,
607they can be set and unset anywhere in kernel network subsystem
608(using
609.Xr mbuf_tags 9
610facility), not only by means of
611.Xr ipfw 4
612.Cm tag
613and
614.Cm untag
615keywords.
616For example, there can be a specialized
617.Xr netgraph 4
618node doing traffic analyzing and tagging for later inspecting
619in firewall.
620.It Cm untag Ar number
621When a packet matches a rule with the
622.Cm untag
623keyword, the tag with the number
624.Ar number
625is searched among the tags attached to this packet and,
626if found, removed from it.
627Other tags bound to packet, if present, are left untouched.
628.It Cm altq Ar queue
629When a packet matches a rule with the
630.Cm altq
631keyword, the ALTQ identifier for the given
632.Ar queue
633(see
634.Xr altq 4 )
635will be attached.
636Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
637and not being rejected or going to divert sockets.
638Note that if there is insufficient memory at the time the packet is
639processed, it will not be tagged, so it is wise to make your ALTQ
640"default" queue policy account for this.
641If multiple
642.Cm altq
643rules match a single packet, only the first one adds the ALTQ classification
644tag.
645In doing so, traffic may be shaped by using
646.Cm count Cm altq Ar queue
647rules for classification early in the ruleset, then later applying
648the filtering decision.
649For example,
650.Cm check-state
651and
652.Cm keep-state
653rules may come later and provide the actual filtering decisions in
654addition to the fallback ALTQ tag.
655.Pp
656You must run
657.Xr pfctl 8
658to set up the queues before IPFW will be able to look them up by name,
659and if the ALTQ disciplines are rearranged, the rules in containing the
660queue identifiers in the kernel will likely have gone stale and need
661to be reloaded.
662Stale queue identifiers will probably result in misclassification.
663.Pp
664All system ALTQ processing can be turned on or off via
665.Nm
666.Cm enable Ar altq
667and
668.Nm
669.Cm disable Ar altq .
670The usage of
671.Va net.inet.ip.fw.one_pass
672is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
673always after adding an ALTQ tag.
674.El
675.Ss RULE ACTIONS
676A rule can be associated with one of the following actions, which
677will be executed when the packet matches the body of the rule.
678.Bl -tag -width indent
679.It Cm allow | accept | pass | permit
680Allow packets that match rule.
681The search terminates.
682.It Cm check-state
683Checks the packet against the dynamic ruleset.
684If a match is found, execute the action associated with
685the rule which generated this dynamic rule, otherwise
686move to the next rule.
687.br
688.Cm Check-state
689rules do not have a body.
690If no
691.Cm check-state
692rule is found, the dynamic ruleset is checked at the first
693.Cm keep-state
694or
695.Cm limit
696rule.
697.It Cm count
698Update counters for all packets that match rule.
699The search continues with the next rule.
700.It Cm deny | drop
701Discard packets that match this rule.
702The search terminates.
703.It Cm divert Ar port
704Divert packets that match this rule to the
705.Xr divert 4
706socket bound to port
707.Ar port .
708The search terminates.
709.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
710Change the next-hop on matching packets to
711.Ar ipaddr ,
712which can be an IP address or a host name.
713The next hop can also be supplied by the last table
714looked up for the packet by using the
715.Cm tablearg
716keyword instead of an explicit address.
717The search terminates if this rule matches.
718.Pp
719If
720.Ar ipaddr
721is a local address, then matching packets will be forwarded to
722.Ar port
723(or the port number in the packet if one is not specified in the rule)
724on the local machine.
725.br
726If
727.Ar ipaddr
728is not a local address, then the port number
729(if specified) is ignored, and the packet will be
730forwarded to the remote address, using the route as found in
731the local routing table for that IP.
732.br
733A
734.Ar fwd
735rule will not match layer-2 packets (those received
736on ether_input, ether_output, or bridged).
737.br
738The
739.Cm fwd
740action does not change the contents of the packet at all.
741In particular, the destination address remains unmodified, so
742packets forwarded to another system will usually be rejected by that system
743unless there is a matching rule on that system to capture them.
744For packets forwarded locally,
745the local address of the socket will be
746set to the original destination address of the packet.
747This makes the
748.Xr netstat 1
749entry look rather weird but is intended for
750use with transparent proxy servers.
751.Pp
752To enable
753.Cm fwd
754a custom kernel needs to be compiled with the option
755.Cd "options IPFIREWALL_FORWARD" .
756.It Cm nat Ar nat_nr
757Pass packet to a
758nat instance
759(for network address translation, address redirect, etc.):
760see the
761.Sx NETWORK ADDRESS TRANSLATION (NAT)
762Section for further information.
763.It Cm pipe Ar pipe_nr
764Pass packet to a
765.Nm dummynet
766.Dq pipe
767(for bandwidth limitation, delay, etc.).
768See the
769.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
770Section for further information.
771The search terminates; however, on exit from the pipe and if
772the
773.Xr sysctl 8
774variable
775.Va net.inet.ip.fw.one_pass
776is not set, the packet is passed again to the firewall code
777starting from the next rule.
778.It Cm queue Ar queue_nr
779Pass packet to a
780.Nm dummynet
781.Dq queue
782(for bandwidth limitation using WF2Q+).
783.It Cm reject
784(Deprecated).
785Synonym for
786.Cm unreach host .
787.It Cm reset
788Discard packets that match this rule, and if the
789packet is a TCP packet, try to send a TCP reset (RST) notice.
790The search terminates.
791.It Cm reset6
792Discard packets that match this rule, and if the
793packet is a TCP packet, try to send a TCP reset (RST) notice.
794The search terminates.
795.It Cm skipto Ar number | tablearg
796Skip all subsequent rules numbered less than
797.Ar number .
798The search continues with the first rule numbered
799.Ar number
800or higher.
801It is possible to use the
802.Cm tablearg
803keyword with a skipto for a
804.Em computed
805skipto, but care should be used, as no destination caching
806is possible in this case so the rules are always walked to find it,
807starting from the
808.Cm skipto .
809.It Cm tee Ar port
810Send a copy of packets matching this rule to the
811.Xr divert 4
812socket bound to port
813.Ar port .
814The search continues with the next rule.
815.It Cm unreach Ar code
816Discard packets that match this rule, and try to send an ICMP
817unreachable notice with code
818.Ar code ,
819where
820.Ar code
821is a number from 0 to 255, or one of these aliases:
822.Cm net , host , protocol , port ,
823.Cm needfrag , srcfail , net-unknown , host-unknown ,
824.Cm isolated , net-prohib , host-prohib , tosnet ,
825.Cm toshost , filter-prohib , host-precedence
826or
827.Cm precedence-cutoff .
828The search terminates.
829.It Cm unreach6 Ar code
830Discard packets that match this rule, and try to send an ICMPv6
831unreachable notice with code
832.Ar code ,
833where
834.Ar code
835is a number from 0, 1, 3 or 4, or one of these aliases:
836.Cm no-route, admin-prohib, address
837or
838.Cm port .
839The search terminates.
840.It Cm netgraph Ar cookie
841Divert packet into netgraph with given
842.Ar cookie .
843The search terminates.
844If packet is later returned from netgraph it is either
845accepted or continues with the next rule, depending on
846.Va net.inet.ip.fw.one_pass
847sysctl variable.
848.It Cm ngtee Ar cookie
849A copy of packet is diverted into netgraph, original
850packet is either accepted or continues with the next rule, depending on
851.Va net.inet.ip.fw.one_pass
852sysctl variable.
853See
854.Xr ng_ipfw 4
855for more information on
856.Cm netgraph
857and
858.Cm ngtee
859actions.
860.It Cm setfib Ar fibnum
861The packet is tagged so as to use the FIB (routing table)
862.Ar fibnum
863in any subsequent forwarding decisions. Initially this is
864limited to the values  0 through 15. See
865.Xr setfib 8 .
866Processing continues at the next rule.
867.El
868.Ss RULE BODY
869The body of a rule contains zero or more patterns (such as
870specific source and destination addresses or ports,
871protocol options, incoming or outgoing interfaces, etc.)
872that the packet must match in order to be recognised.
873In general, the patterns are connected by (implicit)
874.Cm and
875operators -- i.e., all must match in order for the
876rule to match.
877Individual patterns can be prefixed by the
878.Cm not
879operator to reverse the result of the match, as in
880.Pp
881.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
882.Pp
883Additionally, sets of alternative match patterns
884.Pq Em or-blocks
885can be constructed by putting the patterns in
886lists enclosed between parentheses ( ) or braces { }, and
887using the
888.Cm or
889operator as follows:
890.Pp
891.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
892.Pp
893Only one level of parentheses is allowed.
894Beware that most shells have special meanings for parentheses
895or braces, so it is advisable to put a backslash \\ in front of them
896to prevent such interpretations.
897.Pp
898The body of a rule must in general include a source and destination
899address specifier.
900The keyword
901.Ar any
902can be used in various places to specify that the content of
903a required field is irrelevant.
904.Pp
905The rule body has the following format:
906.Bd -ragged -offset indent
907.Op Ar proto Cm from Ar src Cm to Ar dst
908.Op Ar options
909.Ed
910.Pp
911The first part (proto from src to dst) is for backward
912compatibility with earlier versions of
913.Fx .
914In modern
915.Fx
916any match pattern (including MAC headers, IP protocols,
917addresses and ports) can be specified in the
918.Ar options
919section.
920.Pp
921Rule fields have the following meaning:
922.Bl -tag -width indent
923.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
924.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
925An IP protocol specified by number or name
926(for a complete list see
927.Pa /etc/protocols ) ,
928or one of the following keywords:
929.Bl -tag -width indent
930.It Cm ip4 | ipv4
931Matches IPv4 packets.
932.It Cm ip6 | ipv6
933Matches IPv6 packets.
934.It Cm ip | all
935Matches any packet.
936.El
937.Pp
938The
939.Cm ipv6
940in
941.Cm proto
942option will be treated as inner protocol.
943And, the
944.Cm ipv4
945is not available in
946.Cm proto
947option.
948.Pp
949The
950.Cm { Ar protocol Cm or ... }
951format (an
952.Em or-block )
953is provided for convenience only but its use is deprecated.
954.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
955An address (or a list, see below)
956optionally followed by
957.Ar ports
958specifiers.
959.Pp
960The second format
961.Em ( or-block
962with multiple addresses) is provided for convenience only and
963its use is discouraged.
964.It Ar addr : Oo Cm not Oc Bro
965.Cm any | me | me6 |
966.Cm table Ns Pq Ar number Ns Op , Ns Ar value
967.Ar | addr-list | addr-set
968.Brc
969.It Cm any
970matches any IP address.
971.It Cm me
972matches any IP address configured on an interface in the system.
973.It Cm me6
974matches any IPv6 address configured on an interface in the system.
975The address list is evaluated at the time the packet is
976analysed.
977.It Cm table Ns Pq Ar number Ns Op , Ns Ar value
978Matches any IPv4 address for which an entry exists in the lookup table
979.Ar number .
980If an optional 32-bit unsigned
981.Ar value
982is also specified, an entry will match only if it has this value.
983See the
984.Sx LOOKUP TABLES
985section below for more information on lookup tables.
986.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
987.It Ar ip-addr :
988A host or subnet address specified in one of the following ways:
989.Bl -tag -width indent
990.It Ar numeric-ip | hostname
991Matches a single IPv4 address, specified as dotted-quad or a hostname.
992Hostnames are resolved at the time the rule is added to the firewall list.
993.It Ar addr Ns / Ns Ar masklen
994Matches all addresses with base
995.Ar addr
996(specified as an IP address, a network number, or a hostname)
997and mask width of
998.Cm masklen
999bits.
1000As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1001all IP numbers from 1.2.3.0 to 1.2.3.127 .
1002.It Ar addr Ns : Ns Ar mask
1003Matches all addresses with base
1004.Ar addr
1005(specified as an IP address, a network number, or a hostname)
1006and the mask of
1007.Ar mask ,
1008specified as a dotted quad.
1009As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
10101.*.3.*.
1011This form is advised only for non-contiguous
1012masks.
1013It is better to resort to the
1014.Ar addr Ns / Ns Ar masklen
1015format for contiguous masks, which is more compact and less
1016error-prone.
1017.El
1018.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1019.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1020Matches all addresses with base address
1021.Ar addr
1022(specified as an IP address, a network number, or a hostname)
1023and whose last byte is in the list between braces { } .
1024Note that there must be no spaces between braces and
1025numbers (spaces after commas are allowed).
1026Elements of the list can be specified as single entries
1027or ranges.
1028The
1029.Ar masklen
1030field is used to limit the size of the set of addresses,
1031and can have any value between 24 and 32.
1032If not specified,
1033it will be assumed as 24.
1034.br
1035This format is particularly useful to handle sparse address sets
1036within a single rule.
1037Because the matching occurs using a
1038bitmask, it takes constant time and dramatically reduces
1039the complexity of rulesets.
1040.br
1041As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1042or 1.2.3.0/24{128,35-55,89}
1043will match the following IP addresses:
1044.br
10451.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1046.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1047.It Ar ip6-addr :
1048A host or subnet specified one of the following ways:
1049.Pp
1050.Bl -tag -width indent
1051.It Ar numeric-ip | hostname
1052Matches a single IPv6 address as allowed by
1053.Xr inet_pton 3
1054or a hostname.
1055Hostnames are resolved at the time the rule is added to the firewall
1056list.
1057.It Ar addr Ns / Ns Ar masklen
1058Matches all IPv6 addresses with base
1059.Ar addr
1060(specified as allowed by
1061.Xr inet_pton
1062or a hostname)
1063and mask width of
1064.Cm masklen
1065bits.
1066.El
1067.Pp
1068No support for sets of IPv6 addresses is provided because IPv6 addresses
1069are typically random past the initial prefix.
1070.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1071For protocols which support port numbers (such as TCP and UDP), optional
1072.Cm ports
1073may be specified as one or more ports or port ranges, separated
1074by commas but no spaces, and an optional
1075.Cm not
1076operator.
1077The
1078.Ql \&-
1079notation specifies a range of ports (including boundaries).
1080.Pp
1081Service names (from
1082.Pa /etc/services )
1083may be used instead of numeric port values.
1084The length of the port list is limited to 30 ports or ranges,
1085though one can specify larger ranges by using an
1086.Em or-block
1087in the
1088.Cm options
1089section of the rule.
1090.Pp
1091A backslash
1092.Pq Ql \e
1093can be used to escape the dash
1094.Pq Ql -
1095character in a service name (from a shell, the backslash must be
1096typed twice to avoid the shell itself interpreting it as an escape
1097character).
1098.Pp
1099.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1100.Pp
1101Fragmented packets which have a non-zero offset (i.e., not the first
1102fragment) will never match a rule which has one or more port
1103specifications.
1104See the
1105.Cm frag
1106option for details on matching fragmented packets.
1107.El
1108.Ss RULE OPTIONS (MATCH PATTERNS)
1109Additional match patterns can be used within
1110rules.
1111Zero or more of these so-called
1112.Em options
1113can be present in a rule, optionally prefixed by the
1114.Cm not
1115operand, and possibly grouped into
1116.Em or-blocks .
1117.Pp
1118The following match patterns can be used (listed in alphabetical order):
1119.Bl -tag -width indent
1120.It Cm // this is a comment.
1121Inserts the specified text as a comment in the rule.
1122Everything following // is considered as a comment and stored in the rule.
1123You can have comment-only rules, which are listed as having a
1124.Cm count
1125action followed by the comment.
1126.It Cm bridged
1127Alias for
1128.Cm layer2 .
1129.It Cm diverted
1130Matches only packets generated by a divert socket.
1131.It Cm diverted-loopback
1132Matches only packets coming from a divert socket back into the IP stack
1133input for delivery.
1134.It Cm diverted-output
1135Matches only packets going from a divert socket back outward to the IP
1136stack output for delivery.
1137.It Cm dst-ip Ar ip-address
1138Matches IPv4 packets whose destination IP is one of the address(es)
1139specified as argument.
1140.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1141Matches IPv6 packets whose destination IP is one of the address(es)
1142specified as argument.
1143.It Cm dst-port Ar ports
1144Matches IP packets whose destination port is one of the port(s)
1145specified as argument.
1146.It Cm established
1147Matches TCP packets that have the RST or ACK bits set.
1148.It Cm ext6hdr Ar header
1149Matches IPv6 packets containing the extended header given by
1150.Ar header .
1151Supported headers are:
1152.Pp
1153Fragment,
1154.Pq Cm frag ,
1155Hop-to-hop options
1156.Pq Cm hopopt ,
1157any type of Routing Header
1158.Pq Cm route ,
1159Source routing Routing Header Type 0
1160.Pq Cm rthdr0 ,
1161Mobile IPv6 Routing Header Type 2
1162.Pq Cm rthdr2 ,
1163Destination options
1164.Pq Cm dstopt ,
1165IPSec authentication headers
1166.Pq Cm ah ,
1167and IPSec encapsulated security payload headers
1168.Pq Cm esp .
1169.It Cm fib Ar fibnum
1170Matches a packet that has been tagged to use
1171the given FIB (routing table) number.
1172.It Cm flow-id Ar labels
1173Matches IPv6 packets containing any of the flow labels given in
1174.Ar labels .
1175.Ar labels
1176is a comma seperate list of numeric flow labels.
1177.It Cm frag
1178Matches packets that are fragments and not the first
1179fragment of an IP datagram.
1180Note that these packets will not have
1181the next protocol header (e.g.\& TCP, UDP) so options that look into
1182these headers cannot match.
1183.It Cm gid Ar group
1184Matches all TCP or UDP packets sent by or received for a
1185.Ar group .
1186A
1187.Ar group
1188may be specified by name or number.
1189.It Cm jail Ar prisonID
1190Matches all TCP or UDP packets sent by or received for the
1191jail whos prison ID is
1192.Ar prisonID .
1193.It Cm icmptypes Ar types
1194Matches ICMP packets whose ICMP type is in the list
1195.Ar types .
1196The list may be specified as any combination of
1197individual types (numeric) separated by commas.
1198.Em Ranges are not allowed .
1199The supported ICMP types are:
1200.Pp
1201echo reply
1202.Pq Cm 0 ,
1203destination unreachable
1204.Pq Cm 3 ,
1205source quench
1206.Pq Cm 4 ,
1207redirect
1208.Pq Cm 5 ,
1209echo request
1210.Pq Cm 8 ,
1211router advertisement
1212.Pq Cm 9 ,
1213router solicitation
1214.Pq Cm 10 ,
1215time-to-live exceeded
1216.Pq Cm 11 ,
1217IP header bad
1218.Pq Cm 12 ,
1219timestamp request
1220.Pq Cm 13 ,
1221timestamp reply
1222.Pq Cm 14 ,
1223information request
1224.Pq Cm 15 ,
1225information reply
1226.Pq Cm 16 ,
1227address mask request
1228.Pq Cm 17
1229and address mask reply
1230.Pq Cm 18 .
1231.It Cm icmp6types Ar types
1232Matches ICMP6 packets whose ICMP6 type is in the list of
1233.Ar types .
1234The list may be specified as any combination of
1235individual types (numeric) separated by commas.
1236.Em Ranges are not allowed .
1237.It Cm in | out
1238Matches incoming or outgoing packets, respectively.
1239.Cm in
1240and
1241.Cm out
1242are mutually exclusive (in fact,
1243.Cm out
1244is implemented as
1245.Cm not in Ns No ).
1246.It Cm ipid Ar id-list
1247Matches IPv4 packets whose
1248.Cm ip_id
1249field has value included in
1250.Ar id-list ,
1251which is either a single value or a list of values or ranges
1252specified in the same way as
1253.Ar ports .
1254.It Cm iplen Ar len-list
1255Matches IP packets whose total length, including header and data, is
1256in the set
1257.Ar len-list ,
1258which is either a single value or a list of values or ranges
1259specified in the same way as
1260.Ar ports .
1261.It Cm ipoptions Ar spec
1262Matches packets whose IPv4 header contains the comma separated list of
1263options specified in
1264.Ar spec .
1265The supported IP options are:
1266.Pp
1267.Cm ssrr
1268(strict source route),
1269.Cm lsrr
1270(loose source route),
1271.Cm rr
1272(record packet route) and
1273.Cm ts
1274(timestamp).
1275The absence of a particular option may be denoted
1276with a
1277.Ql \&! .
1278.It Cm ipprecedence Ar precedence
1279Matches IPv4 packets whose precedence field is equal to
1280.Ar precedence .
1281.It Cm ipsec
1282Matches packets that have IPSEC history associated with them
1283(i.e., the packet comes encapsulated in IPSEC, the kernel
1284has IPSEC support and IPSEC_FILTERTUNNEL option, and can correctly
1285decapsulate it).
1286.Pp
1287Note that specifying
1288.Cm ipsec
1289is different from specifying
1290.Cm proto Ar ipsec
1291as the latter will only look at the specific IP protocol field,
1292irrespective of IPSEC kernel support and the validity of the IPSEC data.
1293.Pp
1294Further note that this flag is silently ignored in kernels without
1295IPSEC support.
1296It does not affect rule processing when given and the
1297rules are handled as if with no
1298.Cm ipsec
1299flag.
1300.It Cm iptos Ar spec
1301Matches IPv4 packets whose
1302.Cm tos
1303field contains the comma separated list of
1304service types specified in
1305.Ar spec .
1306The supported IP types of service are:
1307.Pp
1308.Cm lowdelay
1309.Pq Dv IPTOS_LOWDELAY ,
1310.Cm throughput
1311.Pq Dv IPTOS_THROUGHPUT ,
1312.Cm reliability
1313.Pq Dv IPTOS_RELIABILITY ,
1314.Cm mincost
1315.Pq Dv IPTOS_MINCOST ,
1316.Cm congestion
1317.Pq Dv IPTOS_ECN_CE .
1318The absence of a particular type may be denoted
1319with a
1320.Ql \&! .
1321.It Cm ipttl Ar ttl-list
1322Matches IPv4 packets whose time to live is included in
1323.Ar ttl-list ,
1324which is either a single value or a list of values or ranges
1325specified in the same way as
1326.Ar ports .
1327.It Cm ipversion Ar ver
1328Matches IP packets whose IP version field is
1329.Ar ver .
1330.It Cm keep-state
1331Upon a match, the firewall will create a dynamic rule, whose
1332default behaviour is to match bidirectional traffic between
1333source and destination IP/port using the same protocol.
1334The rule has a limited lifetime (controlled by a set of
1335.Xr sysctl 8
1336variables), and the lifetime is refreshed every time a matching
1337packet is found.
1338.It Cm layer2
1339Matches only layer2 packets, i.e., those passed to
1340.Nm
1341from ether_demux() and ether_output_frame().
1342.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
1343The firewall will only allow
1344.Ar N
1345connections with the same
1346set of parameters as specified in the rule.
1347One or more
1348of source and destination addresses and ports can be
1349specified.
1350Currently,
1351only IPv4 flows are supported.
1352.It Cm { MAC | mac } Ar dst-mac src-mac
1353Match packets with a given
1354.Ar dst-mac
1355and
1356.Ar src-mac
1357addresses, specified as the
1358.Cm any
1359keyword (matching any MAC address), or six groups of hex digits
1360separated by colons,
1361and optionally followed by a mask indicating the significant bits.
1362The mask may be specified using either of the following methods:
1363.Bl -enum -width indent
1364.It
1365A slash
1366.Pq /
1367followed by the number of significant bits.
1368For example, an address with 33 significant bits could be specified as:
1369.Pp
1370.Dl "MAC 10:20:30:40:50:60/33 any"
1371.Pp
1372.It
1373An ampersand
1374.Pq &
1375followed by a bitmask specified as six groups of hex digits separated
1376by colons.
1377For example, an address in which the last 16 bits are significant could
1378be specified as:
1379.Pp
1380.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1381.Pp
1382Note that the ampersand character has a special meaning in many shells
1383and should generally be escaped.
1384.Pp
1385.El
1386Note that the order of MAC addresses (destination first,
1387source second) is
1388the same as on the wire, but the opposite of the one used for
1389IP addresses.
1390.It Cm mac-type Ar mac-type
1391Matches packets whose Ethernet Type field
1392corresponds to one of those specified as argument.
1393.Ar mac-type
1394is specified in the same way as
1395.Cm port numbers
1396(i.e., one or more comma-separated single values or ranges).
1397You can use symbolic names for known values such as
1398.Em vlan , ipv4, ipv6 .
1399Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1400and they are always printed as hexadecimal (unless the
1401.Cm -N
1402option is used, in which case symbolic resolution will be attempted).
1403.It Cm proto Ar protocol
1404Matches packets with the corresponding IP protocol.
1405.It Cm recv | xmit | via Brq Ar ifX | Ar if Ns Cm * | Ar ipno | Ar any
1406Matches packets received, transmitted or going through,
1407respectively, the interface specified by exact name
1408.Ns No ( Ar ifX Ns No ),
1409by device name
1410.Ns No ( Ar if Ns Ar * Ns No ),
1411by IP address, or through some interface.
1412.Pp
1413The
1414.Cm via
1415keyword causes the interface to always be checked.
1416If
1417.Cm recv
1418or
1419.Cm xmit
1420is used instead of
1421.Cm via ,
1422then only the receive or transmit interface (respectively)
1423is checked.
1424By specifying both, it is possible to match packets based on
1425both receive and transmit interface, e.g.:
1426.Pp
1427.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1428.Pp
1429The
1430.Cm recv
1431interface can be tested on either incoming or outgoing packets,
1432while the
1433.Cm xmit
1434interface can only be tested on outgoing packets.
1435So
1436.Cm out
1437is required (and
1438.Cm in
1439is invalid) whenever
1440.Cm xmit
1441is used.
1442.Pp
1443A packet may not have a receive or transmit interface: packets
1444originating from the local host have no receive interface,
1445while packets destined for the local host have no transmit
1446interface.
1447.It Cm setup
1448Matches TCP packets that have the SYN bit set but no ACK bit.
1449This is the short form of
1450.Dq Li tcpflags\ syn,!ack .
1451.It Cm src-ip Ar ip-address
1452Matches IPv4 packets whose source IP is one of the address(es)
1453specified as an argument.
1454.It Cm src-ip6 Ar ip6-address
1455Matches IPv6 packets whose source IP is one of the address(es)
1456specified as an argument.
1457.It Cm src-port Ar ports
1458Matches IP packets whose source port is one of the port(s)
1459specified as argument.
1460.It Cm tagged Ar tag-list
1461Matches packets whose tags are included in
1462.Ar tag-list ,
1463which is either a single value or a list of values or ranges
1464specified in the same way as
1465.Ar ports .
1466Tags can be applied to the packet using
1467.Cm tag
1468rule action parameter (see it's description for details on tags).
1469.It Cm tcpack Ar ack
1470TCP packets only.
1471Match if the TCP header acknowledgment number field is set to
1472.Ar ack .
1473.It Cm tcpdatalen Ar tcpdatalen-list
1474Matches TCP packets whose length of TCP data is
1475.Ar tcpdatalen-list ,
1476which is either a single value or a list of values or ranges
1477specified in the same way as
1478.Ar ports .
1479.It Cm tcpflags Ar spec
1480TCP packets only.
1481Match if the TCP header contains the comma separated list of
1482flags specified in
1483.Ar spec .
1484The supported TCP flags are:
1485.Pp
1486.Cm fin ,
1487.Cm syn ,
1488.Cm rst ,
1489.Cm psh ,
1490.Cm ack
1491and
1492.Cm urg .
1493The absence of a particular flag may be denoted
1494with a
1495.Ql \&! .
1496A rule which contains a
1497.Cm tcpflags
1498specification can never match a fragmented packet which has
1499a non-zero offset.
1500See the
1501.Cm frag
1502option for details on matching fragmented packets.
1503.It Cm tcpseq Ar seq
1504TCP packets only.
1505Match if the TCP header sequence number field is set to
1506.Ar seq .
1507.It Cm tcpwin Ar win
1508TCP packets only.
1509Match if the TCP header window field is set to
1510.Ar win .
1511.It Cm tcpoptions Ar spec
1512TCP packets only.
1513Match if the TCP header contains the comma separated list of
1514options specified in
1515.Ar spec .
1516The supported TCP options are:
1517.Pp
1518.Cm mss
1519(maximum segment size),
1520.Cm window
1521(tcp window advertisement),
1522.Cm sack
1523(selective ack),
1524.Cm ts
1525(rfc1323 timestamp) and
1526.Cm cc
1527(rfc1644 t/tcp connection count).
1528The absence of a particular option may be denoted
1529with a
1530.Ql \&! .
1531.It Cm uid Ar user
1532Match all TCP or UDP packets sent by or received for a
1533.Ar user .
1534A
1535.Ar user
1536may be matched by name or identification number.
1537.It Cm verrevpath
1538For incoming packets,
1539a routing table lookup is done on the packet's source address.
1540If the interface on which the packet entered the system matches the
1541outgoing interface for the route,
1542the packet matches.
1543If the interfaces do not match up,
1544the packet does not match.
1545All outgoing packets or packets with no incoming interface match.
1546.Pp
1547The name and functionality of the option is intentionally similar to
1548the Cisco IOS command:
1549.Pp
1550.Dl ip verify unicast reverse-path
1551.Pp
1552This option can be used to make anti-spoofing rules to reject all
1553packets with source addresses not from this interface.
1554See also the option
1555.Cm antispoof .
1556.It Cm versrcreach
1557For incoming packets,
1558a routing table lookup is done on the packet's source address.
1559If a route to the source address exists, but not the default route
1560or a blackhole/reject route, the packet matches.
1561Otherwise, the packet does not match.
1562All outgoing packets match.
1563.Pp
1564The name and functionality of the option is intentionally similar to
1565the Cisco IOS command:
1566.Pp
1567.Dl ip verify unicast source reachable-via any
1568.Pp
1569This option can be used to make anti-spoofing rules to reject all
1570packets whose source address is unreachable.
1571.It Cm antispoof
1572For incoming packets, the packet's source address is checked if it
1573belongs to a directly connected network.
1574If the network is directly connected, then the interface the packet
1575came on in is compared to the interface the network is connected to.
1576When incoming interface and directly connected interface are not the
1577same, the packet does not match.
1578Otherwise, the packet does match.
1579All outgoing packets match.
1580.Pp
1581This option can be used to make anti-spoofing rules to reject all
1582packets that pretend to be from a directly connected network but do
1583not come in through that interface.
1584This option is similar to but more restricted than
1585.Cm verrevpath
1586because it engages only on packets with source addresses of directly
1587connected networks instead of all source addresses.
1588.El
1589.Sh LOOKUP TABLES
1590Lookup tables are useful to handle large sparse address sets,
1591typically from a hundred to several thousands of entries.
1592There may be up to 128 different lookup tables, numbered 0 to 127.
1593.Pp
1594Each entry is represented by an
1595.Ar addr Ns Op / Ns Ar masklen
1596and will match all addresses with base
1597.Ar addr
1598(specified as an IP address or a hostname)
1599and mask width of
1600.Ar masklen
1601bits.
1602If
1603.Ar masklen
1604is not specified, it defaults to 32.
1605When looking up an IP address in a table, the most specific
1606entry will match.
1607Associated with each entry is a 32-bit unsigned
1608.Ar value ,
1609which can optionally be checked by a rule matching code.
1610When adding an entry, if
1611.Ar value
1612is not specified, it defaults to 0.
1613.Pp
1614An entry can be added to a table
1615.Pq Cm add ,
1616removed from a table
1617.Pq Cm delete ,
1618a table can be examined
1619.Pq Cm list
1620or flushed
1621.Pq Cm flush .
1622.Pp
1623Internally, each table is stored in a Radix tree, the same way as
1624the routing table (see
1625.Xr route 4 ) .
1626.Pp
1627Lookup tables currently support IPv4 addresses only.
1628.Pp
1629The
1630.Cm tablearg
1631feature provides the ability to use a value, looked up in the table, as
1632the argument for a rule action, action parameter or rule option.
1633This can significantly reduce number of rules in some configurations.
1634If two tables are used in a rule, the result of the second (destination)
1635is used.
1636The
1637.Cm tablearg
1638argument can be used with the following actions:
1639.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto
1640action parameters:
1641.Cm tag, untag,
1642rule options:
1643.Cm limit, tagged.
1644.Pp
1645When used with
1646.Cm fwd
1647it is possible to supply table entries with values
1648that are in the form of IP addresses or hostnames.
1649See the
1650.Sx EXAMPLES
1651Section for example usage of tables and the tablearg keyword.
1652.Pp
1653When used with the
1654.Cm skipto
1655action, the user should be aware that the code will walk the ruleset
1656up to a rule equal to, or past, the given number, and should therefore try keep the
1657ruleset compact between the skipto and the target rules.
1658.Sh SETS OF RULES
1659Each rule belongs to one of 32 different
1660.Em sets
1661, numbered 0 to 31.
1662Set 31 is reserved for the default rule.
1663.Pp
1664By default, rules are put in set 0, unless you use the
1665.Cm set N
1666attribute when entering a new rule.
1667Sets can be individually and atomically enabled or disabled,
1668so this mechanism permits an easy way to store multiple configurations
1669of the firewall and quickly (and atomically) switch between them.
1670The command to enable/disable sets is
1671.Bd -ragged -offset indent
1672.Nm
1673.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
1674.Ed
1675.Pp
1676where multiple
1677.Cm enable
1678or
1679.Cm disable
1680sections can be specified.
1681Command execution is atomic on all the sets specified in the command.
1682By default, all sets are enabled.
1683.Pp
1684When you disable a set, its rules behave as if they do not exist
1685in the firewall configuration, with only one exception:
1686.Bd -ragged -offset indent
1687dynamic rules created from a rule before it had been disabled
1688will still be active until they expire.
1689In order to delete
1690dynamic rules you have to explicitly delete the parent rule
1691which generated them.
1692.Ed
1693.Pp
1694The set number of rules can be changed with the command
1695.Bd -ragged -offset indent
1696.Nm
1697.Cm set move
1698.Brq Cm rule Ar rule-number | old-set
1699.Cm to Ar new-set
1700.Ed
1701.Pp
1702Also, you can atomically swap two rulesets with the command
1703.Bd -ragged -offset indent
1704.Nm
1705.Cm set swap Ar first-set second-set
1706.Ed
1707.Pp
1708See the
1709.Sx EXAMPLES
1710Section on some possible uses of sets of rules.
1711.Sh STATEFUL FIREWALL
1712Stateful operation is a way for the firewall to dynamically
1713create rules for specific flows when packets that
1714match a given pattern are detected.
1715Support for stateful
1716operation comes through the
1717.Cm check-state , keep-state
1718and
1719.Cm limit
1720options of
1721.Nm rules .
1722.Pp
1723Dynamic rules are created when a packet matches a
1724.Cm keep-state
1725or
1726.Cm limit
1727rule, causing the creation of a
1728.Em dynamic
1729rule which will match all and only packets with
1730a given
1731.Em protocol
1732between a
1733.Em src-ip/src-port dst-ip/dst-port
1734pair of addresses
1735.Em ( src
1736and
1737.Em dst
1738are used here only to denote the initial match addresses, but they
1739are completely equivalent afterwards).
1740Dynamic rules will be checked at the first
1741.Cm check-state, keep-state
1742or
1743.Cm limit
1744occurrence, and the action performed upon a match will be the same
1745as in the parent rule.
1746.Pp
1747Note that no additional attributes other than protocol and IP addresses
1748and ports are checked on dynamic rules.
1749.Pp
1750The typical use of dynamic rules is to keep a closed firewall configuration,
1751but let the first TCP SYN packet from the inside network install a
1752dynamic rule for the flow so that packets belonging to that session
1753will be allowed through the firewall:
1754.Pp
1755.Dl "ipfw add check-state"
1756.Dl "ipfw add allow tcp from my-subnet to any setup keep-state"
1757.Dl "ipfw add deny tcp from any to any"
1758.Pp
1759A similar approach can be used for UDP, where an UDP packet coming
1760from the inside will install a dynamic rule to let the response through
1761the firewall:
1762.Pp
1763.Dl "ipfw add check-state"
1764.Dl "ipfw add allow udp from my-subnet to any keep-state"
1765.Dl "ipfw add deny udp from any to any"
1766.Pp
1767Dynamic rules expire after some time, which depends on the status
1768of the flow and the setting of some
1769.Cm sysctl
1770variables.
1771See Section
1772.Sx SYSCTL VARIABLES
1773for more details.
1774For TCP sessions, dynamic rules can be instructed to periodically
1775send keepalive packets to refresh the state of the rule when it is
1776about to expire.
1777.Pp
1778See Section
1779.Sx EXAMPLES
1780for more examples on how to use dynamic rules.
1781.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
1782.Nm
1783is also the user interface for the
1784.Nm dummynet
1785traffic shaper.
1786.Pp
1787.Nm dummynet
1788operates by first using the firewall to classify packets and divide them into
1789.Em flows ,
1790using any match pattern that can be used in
1791.Nm
1792rules.
1793Depending on local policies, a flow can contain packets for a single
1794TCP connection, or from/to a given host, or entire subnet, or a
1795protocol type, etc.
1796.Pp
1797There are two modes of
1798.Nm dummynet
1799operation:
1800.Dq normal
1801and
1802.Dq fast .
1803The
1804.Dq normal
1805mode tries to emulate a real link: the
1806.Nm dummynet
1807scheduler ensures that the packet will not leave the pipe faster than it
1808would on the real link with a given bandwidth.
1809The
1810.Dq fast
1811mode allows certain packets to bypass the
1812.Nm dummynet
1813scheduler (if packet flow does not exceed pipe's bandwidth).
1814This is the reason why the
1815.Dq fast
1816mode requires less CPU cycles per packet (on average) and packet latency
1817can be significantly lower in comparison to a real link with the same
1818bandwidth.
1819The default mode is
1820.Dq normal .
1821The
1822.Dq fast
1823mode can be enabled by setting the
1824.Va net.inet.ip.dummynet.io_fast
1825.Xr sysctl 8
1826variable to a non-zero value.
1827.Pp
1828Packets belonging to the same flow are then passed to either of two
1829different objects, which implement the traffic regulation:
1830.Bl -hang -offset XXXX
1831.It Em pipe
1832A pipe emulates a link with given bandwidth, propagation delay,
1833queue size and packet loss rate.
1834Packets are queued in front of the pipe as they come out from the classifier,
1835and then transferred to the pipe according to the pipe's parameters.
1836.Pp
1837.It Em queue
1838A queue
1839is an abstraction used to implement the WF2Q+
1840(Worst-case Fair Weighted Fair Queueing) policy, which is
1841an efficient variant of the WFQ policy.
1842.br
1843The queue associates a
1844.Em weight
1845and a reference pipe to each flow, and then all backlogged (i.e.,
1846with packets queued) flows linked to the same pipe share the pipe's
1847bandwidth proportionally to their weights.
1848Note that weights are not priorities; a flow with a lower weight
1849is still guaranteed to get its fraction of the bandwidth even if a
1850flow with a higher weight is permanently backlogged.
1851.Pp
1852.El
1853In practice,
1854.Em pipes
1855can be used to set hard limits to the bandwidth that a flow can use, whereas
1856.Em queues
1857can be used to determine how different flows share the available bandwidth.
1858.Pp
1859The
1860.Em pipe
1861and
1862.Em queue
1863configuration commands are the following:
1864.Bd -ragged -offset indent
1865.Cm pipe Ar number Cm config Ar pipe-configuration
1866.Pp
1867.Cm queue Ar number Cm config Ar queue-configuration
1868.Ed
1869.Pp
1870The following parameters can be configured for a pipe:
1871.Pp
1872.Bl -tag -width indent -compact
1873.It Cm bw Ar bandwidth | device
1874Bandwidth, measured in
1875.Sm off
1876.Op Cm K | M
1877.Brq Cm bit/s | Byte/s .
1878.Sm on
1879.Pp
1880A value of 0 (default) means unlimited bandwidth.
1881The unit must immediately follow the number, as in
1882.Pp
1883.Dl "ipfw pipe 1 config bw 300Kbit/s"
1884.Pp
1885If a device name is specified instead of a numeric value, as in
1886.Pp
1887.Dl "ipfw pipe 1 config bw tun0"
1888.Pp
1889then the transmit clock is supplied by the specified device.
1890At the moment only the
1891.Xr tun 4
1892device supports this
1893functionality, for use in conjunction with
1894.Xr ppp 8 .
1895.Pp
1896.It Cm delay Ar ms-delay
1897Propagation delay, measured in milliseconds.
1898The value is rounded to the next multiple of the clock tick
1899(typically 10ms, but it is a good practice to run kernels
1900with
1901.Dq "options HZ=1000"
1902to reduce
1903the granularity to 1ms or less).
1904Default value is 0, meaning no delay.
1905.El
1906.Pp
1907The following parameters can be configured for a queue:
1908.Pp
1909.Bl -tag -width indent -compact
1910.It Cm pipe Ar pipe_nr
1911Connects a queue to the specified pipe.
1912Multiple queues (with the same or different weights) can be connected to
1913the same pipe, which specifies the aggregate rate for the set of queues.
1914.Pp
1915.It Cm weight Ar weight
1916Specifies the weight to be used for flows matching this queue.
1917The weight must be in the range 1..100, and defaults to 1.
1918.El
1919.Pp
1920Finally, the following parameters can be configured for both
1921pipes and queues:
1922.Pp
1923.Bl -tag -width XXXX -compact
1924.Pp
1925.It Cm buckets Ar hash-table-size
1926Specifies the size of the hash table used for storing the
1927various queues.
1928Default value is 64 controlled by the
1929.Xr sysctl 8
1930variable
1931.Va net.inet.ip.dummynet.hash_size ,
1932allowed range is 16 to 65536.
1933.Pp
1934.It Cm mask Ar mask-specifier
1935Packets sent to a given pipe or queue by an
1936.Nm
1937rule can be further classified into multiple flows, each of which is then
1938sent to a different
1939.Em dynamic
1940pipe or queue.
1941A flow identifier is constructed by masking the IP addresses,
1942ports and protocol types as specified with the
1943.Cm mask
1944options in the configuration of the pipe or queue.
1945For each different flow identifier, a new pipe or queue is created
1946with the same parameters as the original object, and matching packets
1947are sent to it.
1948.Pp
1949Thus, when
1950.Em dynamic pipes
1951are used, each flow will get the same bandwidth as defined by the pipe,
1952whereas when
1953.Em dynamic queues
1954are used, each flow will share the parent's pipe bandwidth evenly
1955with other flows generated by the same queue (note that other queues
1956with different weights might be connected to the same pipe).
1957.br
1958Available mask specifiers are a combination of one or more of the following:
1959.Pp
1960.Cm dst-ip Ar mask ,
1961.Cm dst-ip6 Ar mask ,
1962.Cm src-ip Ar mask ,
1963.Cm src-ip6 Ar mask ,
1964.Cm dst-port Ar mask ,
1965.Cm src-port Ar mask ,
1966.Cm flow-id Ar mask ,
1967.Cm proto Ar mask
1968or
1969.Cm all ,
1970.Pp
1971where the latter means all bits in all fields are significant.
1972.Pp
1973.It Cm noerror
1974When a packet is dropped by a
1975.Nm dummynet
1976queue or pipe, the error
1977is normally reported to the caller routine in the kernel, in the
1978same way as it happens when a device queue fills up.
1979Setting this
1980option reports the packet as successfully delivered, which can be
1981needed for some experimental setups where you want to simulate
1982loss or congestion at a remote router.
1983.Pp
1984.It Cm plr Ar packet-loss-rate
1985Packet loss rate.
1986Argument
1987.Ar packet-loss-rate
1988is a floating-point number between 0 and 1, with 0 meaning no
1989loss, 1 meaning 100% loss.
1990The loss rate is internally represented on 31 bits.
1991.Pp
1992.It Cm queue Brq Ar slots | size Ns Cm Kbytes
1993Queue size, in
1994.Ar slots
1995or
1996.Cm KBytes .
1997Default value is 50 slots, which
1998is the typical queue size for Ethernet devices.
1999Note that for slow speed links you should keep the queue
2000size short or your traffic might be affected by a significant
2001queueing delay.
2002E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2003or 20s of queue on a 30Kbit/s pipe.
2004Even worse effects can result if you get packets from an
2005interface with a much larger MTU, e.g.\& the loopback interface
2006with its 16KB packets.
2007The
2008.Xr sysctl 8
2009variables
2010.Em net.inet.ip.dummynet.pipe_byte_limit
2011and
2012.Em net.inet.ip.dummynet.pipe_slot_limit
2013control the maximum lengths that can be specified.
2014.Pp
2015.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2016Make use of the RED (Random Early Detection) queue management algorithm.
2017.Ar w_q
2018and
2019.Ar max_p
2020are floating
2021point numbers between 0 and 1 (0 not included), while
2022.Ar min_th
2023and
2024.Ar max_th
2025are integer numbers specifying thresholds for queue management
2026(thresholds are computed in bytes if the queue has been defined
2027in bytes, in slots otherwise).
2028The
2029.Nm dummynet
2030also supports the gentle RED variant (gred).
2031Three
2032.Xr sysctl 8
2033variables can be used to control the RED behaviour:
2034.Bl -tag -width indent
2035.It Va net.inet.ip.dummynet.red_lookup_depth
2036specifies the accuracy in computing the average queue
2037when the link is idle (defaults to 256, must be greater than zero)
2038.It Va net.inet.ip.dummynet.red_avg_pkt_size
2039specifies the expected average packet size (defaults to 512, must be
2040greater than zero)
2041.It Va net.inet.ip.dummynet.red_max_pkt_size
2042specifies the expected maximum packet size, only used when queue
2043thresholds are in bytes (defaults to 1500, must be greater than zero).
2044.El
2045.El
2046.Pp
2047When used with IPv6 data,
2048.Nm dummynet
2049currently has several limitations.
2050Information necessary to route link-local packets to an
2051interface is not available after processing by
2052.Nm dummynet
2053so those packets are dropped in the output path.
2054Care should be taken to insure that link-local packets are not passed to
2055.Nm dummynet .
2056.Sh CHECKLIST
2057Here are some important points to consider when designing your
2058rules:
2059.Bl -bullet
2060.It
2061Remember that you filter both packets going
2062.Cm in
2063and
2064.Cm out .
2065Most connections need packets going in both directions.
2066.It
2067Remember to test very carefully.
2068It is a good idea to be near the console when doing this.
2069If you cannot be near the console,
2070use an auto-recovery script such as the one in
2071.Pa /usr/share/examples/ipfw/change_rules.sh .
2072.It
2073Do not forget the loopback interface.
2074.El
2075.Sh FINE POINTS
2076.Bl -bullet
2077.It
2078There are circumstances where fragmented datagrams are unconditionally
2079dropped.
2080TCP packets are dropped if they do not contain at least 20 bytes of
2081TCP header, UDP packets are dropped if they do not contain a full 8
2082byte UDP header, and ICMP packets are dropped if they do not contain
20834 bytes of ICMP header, enough to specify the ICMP type, code, and
2084checksum.
2085These packets are simply logged as
2086.Dq pullup failed
2087since there may not be enough good data in the packet to produce a
2088meaningful log entry.
2089.It
2090Another type of packet is unconditionally dropped, a TCP packet with a
2091fragment offset of one.
2092This is a valid packet, but it only has one use, to try
2093to circumvent firewalls.
2094When logging is enabled, these packets are
2095reported as being dropped by rule -1.
2096.It
2097If you are logged in over a network, loading the
2098.Xr kld 4
2099version of
2100.Nm
2101is probably not as straightforward as you would think.
2102I recommend the following command line:
2103.Bd -literal -offset indent
2104kldload ipfw && \e
2105ipfw add 32000 allow ip from any to any
2106.Ed
2107.Pp
2108Along the same lines, doing an
2109.Bd -literal -offset indent
2110ipfw flush
2111.Ed
2112.Pp
2113in similar surroundings is also a bad idea.
2114.It
2115The
2116.Nm
2117filter list may not be modified if the system security level
2118is set to 3 or higher
2119(see
2120.Xr init 8
2121for information on system security levels).
2122.El
2123.Sh PACKET DIVERSION
2124A
2125.Xr divert 4
2126socket bound to the specified port will receive all packets
2127diverted to that port.
2128If no socket is bound to the destination port, or if the divert module is
2129not loaded, or if the kernel was not compiled with divert socket support,
2130the packets are dropped.
2131.Sh NETWORK ADDRESS TRANSLATION (NAT)
2132The nat configuration command is the following:
2133.Bd -ragged -offset indent
2134.Bk -words
2135.Cm nat
2136.Ar nat_number
2137.Cm config
2138.Ar nat-configuration
2139.Ek
2140.Ed
2141.Pp
2142.
2143The following parameters can be configured:
2144.Bl -tag -width indent
2145.It Cm ip Ar ip_address
2146Define an ip address to use for aliasing.
2147.It Cm if Ar nic
2148Use ip addres of NIC for aliasing, dynamically changing
2149it if NIC's ip address change.
2150.It Cm log
2151Enable logging on this nat instance.
2152.It Cm deny_in
2153Deny any incoming connection from outside world.
2154.It Cm same_ports
2155Try to leave the alias port numbers unchanged from
2156the actual local port numbers.
2157.It Cm unreg_only
2158Traffic on the local network not originating from an
2159unregistered address spaces will be ignored.
2160.It Cm reset
2161Reset table of the packet aliasing engine on address change.
2162.It Cm reverse
2163Reverse the way libalias handles aliasing.
2164.It Cm proxy_only
2165Obey transparent proxy rules only, packet aliasing is not performed.
2166.El
2167.Pp
2168To let the packet continue after being (de)aliased, set the sysctl variable
2169.Va net.inet.ip.fw.one_pass
2170to 0.
2171For more information about aliasing modes, refer to
2172.Xr libalias 3
2173.
2174See Section
2175.Sx EXAMPLES
2176for some examples about nat usage.
2177.Sh REDIRECT AND LSNAT SUPPORT IN IPFW
2178Redirect and LSNAT support follow closely the syntax used in
2179.Xr natd 8
2180.
2181See Section
2182.Sx EXAMPLES
2183for some examples on how to do redirect and lsnat.
2184.Sh SYSCTL VARIABLES
2185A set of
2186.Xr sysctl 8
2187variables controls the behaviour of the firewall and
2188associated modules
2189.Pq Nm dummynet , bridge .
2190These are shown below together with their default value
2191(but always check with the
2192.Xr sysctl 8
2193command what value is actually in use) and meaning:
2194.Bl -tag -width indent
2195.It Va net.inet.ip.dummynet.expire : No 1
2196Lazily delete dynamic pipes/queue once they have no pending traffic.
2197You can disable this by setting the variable to 0, in which case
2198the pipes/queues will only be deleted when the threshold is reached.
2199.It Va net.inet.ip.dummynet.hash_size : No 64
2200Default size of the hash table used for dynamic pipes/queues.
2201This value is used when no
2202.Cm buckets
2203option is specified when configuring a pipe/queue.
2204.It Va net.inet.ip.dummynet.io_fast : No 0
2205If set to a non-zero value,
2206the
2207.Dq fast
2208mode of
2209.Nm dummynet
2210operation (see above) is enabled.
2211.It Va net.inet.ip.dummynet.io_pkt
2212Number of packets passed to
2213.Nm dummynet .
2214.It Va net.inet.ip.dummynet.io_pkt_drop
2215Number of packets dropped by
2216.Nm dummynet .
2217.It Va net.inet.ip.dummynet.io_pkt_fast
2218Number of packets bypassed by the
2219.Nm dummynet
2220scheduler.
2221.It Va net.inet.ip.dummynet.max_chain_len : No 16
2222Target value for the maximum number of pipes/queues in a hash bucket.
2223The product
2224.Cm max_chain_len*hash_size
2225is used to determine the threshold over which empty pipes/queues
2226will be expired even when
2227.Cm net.inet.ip.dummynet.expire=0 .
2228.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
2229.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
2230.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
2231Parameters used in the computations of the drop probability
2232for the RED algorithm.
2233.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
2234.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
2235The maximum queue size that can be specified in bytes or packets.
2236These limits prevent accidental exhaustion of resources such as mbufs.
2237If you raise these limits,
2238you should make sure the system is configured so that sufficient resources
2239are available.
2240.It Va net.inet.ip.fw.autoinc_step : No 100
2241Delta between rule numbers when auto-generating them.
2242The value must be in the range 1..1000.
2243.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
2244The current number of buckets in the hash table for dynamic rules
2245(readonly).
2246.It Va net.inet.ip.fw.debug : No 1
2247Controls debugging messages produced by
2248.Nm .
2249.It Va net.inet.ip.fw.default_rule : No 65535
2250The default rule number (read-only).
2251By the design of
2252.Nm , the default rule is the last one, so its number
2253can also serve as the highest number allowed for a rule.
2254.It Va net.inet.ip.fw.dyn_buckets : No 256
2255The number of buckets in the hash table for dynamic rules.
2256Must be a power of 2, up to 65536.
2257It only takes effect when all dynamic rules have expired, so you
2258are advised to use a
2259.Cm flush
2260command to make sure that the hash table is resized.
2261.It Va net.inet.ip.fw.dyn_count : No 3
2262Current number of dynamic rules
2263(read-only).
2264.It Va net.inet.ip.fw.dyn_keepalive : No 1
2265Enables generation of keepalive packets for
2266.Cm keep-state
2267rules on TCP sessions.
2268A keepalive is generated to both
2269sides of the connection every 5 seconds for the last 20
2270seconds of the lifetime of the rule.
2271.It Va net.inet.ip.fw.dyn_max : No 8192
2272Maximum number of dynamic rules.
2273When you hit this limit, no more dynamic rules can be
2274installed until old ones expire.
2275.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
2276.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
2277.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
2278.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
2279.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
2280.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
2281These variables control the lifetime, in seconds, of dynamic
2282rules.
2283Upon the initial SYN exchange the lifetime is kept short,
2284then increased after both SYN have been seen, then decreased
2285again during the final FIN exchange or when a RST is received.
2286Both
2287.Em dyn_fin_lifetime
2288and
2289.Em dyn_rst_lifetime
2290must be strictly lower than 5 seconds, the period of
2291repetition of keepalives.
2292The firewall enforces that.
2293.It Va net.inet.ip.fw.enable : No 1
2294Enables the firewall.
2295Setting this variable to 0 lets you run your machine without
2296firewall even if compiled in.
2297.It Va net.inet6.ip6.fw.enable : No 1
2298provides the same functionality as above for the IPv6 case.
2299.It Va net.inet.ip.fw.one_pass : No 1
2300When set, the packet exiting from the
2301.Nm dummynet
2302pipe or from
2303.Xr ng_ipfw 4
2304node is not passed though the firewall again.
2305Otherwise, after an action, the packet is
2306reinjected into the firewall at the next rule.
2307.It Va net.inet.ip.fw.tables_max : No 128
2308Maximum number of tables (read-only).
2309.It Va net.inet.ip.fw.verbose : No 1
2310Enables verbose messages.
2311.It Va net.inet.ip.fw.verbose_limit : No 0
2312Limits the number of messages produced by a verbose firewall.
2313.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
2314If enabled packets with unknown IPv6 Extension Headers will be denied.
2315.It Va net.link.ether.ipfw : No 0
2316Controls whether layer-2 packets are passed to
2317.Nm .
2318Default is no.
2319.It Va net.link.bridge.ipfw : No 0
2320Controls whether bridged packets are passed to
2321.Nm .
2322Default is no.
2323.El
2324.Pp
2325.Sh EXAMPLES
2326There are far too many possible uses of
2327.Nm
2328so this Section will only give a small set of examples.
2329.Pp
2330.Ss BASIC PACKET FILTERING
2331This command adds an entry which denies all tcp packets from
2332.Em cracker.evil.org
2333to the telnet port of
2334.Em wolf.tambov.su
2335from being forwarded by the host:
2336.Pp
2337.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
2338.Pp
2339This one disallows any connection from the entire cracker's
2340network to my host:
2341.Pp
2342.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
2343.Pp
2344A first and efficient way to limit access (not using dynamic rules)
2345is the use of the following rules:
2346.Pp
2347.Dl "ipfw add allow tcp from any to any established"
2348.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
2349.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
2350.Dl "..."
2351.Dl "ipfw add deny tcp from any to any"
2352.Pp
2353The first rule will be a quick match for normal TCP packets,
2354but it will not match the initial SYN packet, which will be
2355matched by the
2356.Cm setup
2357rules only for selected source/destination pairs.
2358All other SYN packets will be rejected by the final
2359.Cm deny
2360rule.
2361.Pp
2362If you administer one or more subnets, you can take advantage
2363of the address sets and or-blocks and write extremely
2364compact rulesets which selectively enable services to blocks
2365of clients, as below:
2366.Pp
2367.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
2368.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
2369.Dl ""
2370.Dl "ipfw add allow ip from ${goodguys} to any"
2371.Dl "ipfw add deny ip from ${badguys} to any"
2372.Dl "... normal policies ..."
2373.Pp
2374The
2375.Cm verrevpath
2376option could be used to do automated anti-spoofing by adding the
2377following to the top of a ruleset:
2378.Pp
2379.Dl "ipfw add deny ip from any to any not verrevpath in"
2380.Pp
2381This rule drops all incoming packets that appear to be coming to the
2382system on the wrong interface.
2383For example, a packet with a source
2384address belonging to a host on a protected internal network would be
2385dropped if it tried to enter the system from an external interface.
2386.Pp
2387The
2388.Cm antispoof
2389option could be used to do similar but more restricted anti-spoofing
2390by adding the following to the top of a ruleset:
2391.Pp
2392.Dl "ipfw add deny ip from any to any not antispoof in"
2393.Pp
2394This rule drops all incoming packets that appear to be coming from another
2395directly connected system but on the wrong interface.
2396For example, a packet with a source address of
2397.Li 192.168.0.0/24
2398, configured on
2399.Li fxp0
2400, but coming in on
2401.Li fxp1
2402would be dropped.
2403.Ss DYNAMIC RULES
2404In order to protect a site from flood attacks involving fake
2405TCP packets, it is safer to use dynamic rules:
2406.Pp
2407.Dl "ipfw add check-state"
2408.Dl "ipfw add deny tcp from any to any established"
2409.Dl "ipfw add allow tcp from my-net to any setup keep-state"
2410.Pp
2411This will let the firewall install dynamic rules only for
2412those connection which start with a regular SYN packet coming
2413from the inside of our network.
2414Dynamic rules are checked when encountering the first
2415.Cm check-state
2416or
2417.Cm keep-state
2418rule.
2419A
2420.Cm check-state
2421rule should usually be placed near the beginning of the
2422ruleset to minimize the amount of work scanning the ruleset.
2423Your mileage may vary.
2424.Pp
2425To limit the number of connections a user can open
2426you can use the following type of rules:
2427.Pp
2428.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
2429.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
2430.Pp
2431The former (assuming it runs on a gateway) will allow each host
2432on a /24 network to open at most 10 TCP connections.
2433The latter can be placed on a server to make sure that a single
2434client does not use more than 4 simultaneous connections.
2435.Pp
2436.Em BEWARE :
2437stateful rules can be subject to denial-of-service attacks
2438by a SYN-flood which opens a huge number of dynamic rules.
2439The effects of such attacks can be partially limited by
2440acting on a set of
2441.Xr sysctl 8
2442variables which control the operation of the firewall.
2443.Pp
2444Here is a good usage of the
2445.Cm list
2446command to see accounting records and timestamp information:
2447.Pp
2448.Dl ipfw -at list
2449.Pp
2450or in short form without timestamps:
2451.Pp
2452.Dl ipfw -a list
2453.Pp
2454which is equivalent to:
2455.Pp
2456.Dl ipfw show
2457.Pp
2458Next rule diverts all incoming packets from 192.168.2.0/24
2459to divert port 5000:
2460.Pp
2461.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
2462.Pp
2463.Ss TRAFFIC SHAPING
2464The following rules show some of the applications of
2465.Nm
2466and
2467.Nm dummynet
2468for simulations and the like.
2469.Pp
2470This rule drops random incoming packets with a probability
2471of 5%:
2472.Pp
2473.Dl "ipfw add prob 0.05 deny ip from any to any in"
2474.Pp
2475A similar effect can be achieved making use of
2476.Nm dummynet
2477pipes:
2478.Pp
2479.Dl "ipfw add pipe 10 ip from any to any"
2480.Dl "ipfw pipe 10 config plr 0.05"
2481.Pp
2482We can use pipes to artificially limit bandwidth, e.g.\& on a
2483machine acting as a router, if we want to limit traffic from
2484local clients on 192.168.2.0/24 we do:
2485.Pp
2486.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2487.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
2488.Pp
2489note that we use the
2490.Cm out
2491modifier so that the rule is not used twice.
2492Remember in fact that
2493.Nm
2494rules are checked both on incoming and outgoing packets.
2495.Pp
2496Should we want to simulate a bidirectional link with bandwidth
2497limitations, the correct way is the following:
2498.Pp
2499.Dl "ipfw add pipe 1 ip from any to any out"
2500.Dl "ipfw add pipe 2 ip from any to any in"
2501.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
2502.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
2503.Pp
2504The above can be very useful, e.g.\& if you want to see how
2505your fancy Web page will look for a residential user who
2506is connected only through a slow link.
2507You should not use only one pipe for both directions, unless
2508you want to simulate a half-duplex medium (e.g.\& AppleTalk,
2509Ethernet, IRDA).
2510It is not necessary that both pipes have the same configuration,
2511so we can also simulate asymmetric links.
2512.Pp
2513Should we want to verify network performance with the RED queue
2514management algorithm:
2515.Pp
2516.Dl "ipfw add pipe 1 ip from any to any"
2517.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
2518.Pp
2519Another typical application of the traffic shaper is to
2520introduce some delay in the communication.
2521This can significantly affect applications which do a lot of Remote
2522Procedure Calls, and where the round-trip-time of the
2523connection often becomes a limiting factor much more than
2524bandwidth:
2525.Pp
2526.Dl "ipfw add pipe 1 ip from any to any out"
2527.Dl "ipfw add pipe 2 ip from any to any in"
2528.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
2529.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
2530.Pp
2531Per-flow queueing can be useful for a variety of purposes.
2532A very simple one is counting traffic:
2533.Pp
2534.Dl "ipfw add pipe 1 tcp from any to any"
2535.Dl "ipfw add pipe 1 udp from any to any"
2536.Dl "ipfw add pipe 1 ip from any to any"
2537.Dl "ipfw pipe 1 config mask all"
2538.Pp
2539The above set of rules will create queues (and collect
2540statistics) for all traffic.
2541Because the pipes have no limitations, the only effect is
2542collecting statistics.
2543Note that we need 3 rules, not just the last one, because
2544when
2545.Nm
2546tries to match IP packets it will not consider ports, so we
2547would not see connections on separate ports as different
2548ones.
2549.Pp
2550A more sophisticated example is limiting the outbound traffic
2551on a net with per-host limits, rather than per-network limits:
2552.Pp
2553.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2554.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
2555.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2556.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2557.Ss LOOKUP TABLES
2558In the following example, we need to create several traffic bandwidth
2559classes and we need different hosts/networks to fall into different classes.
2560We create one pipe for each class and configure them accordingly.
2561Then we create a single table and fill it with IP subnets and addresses.
2562For each subnet/host we set the argument equal to the number of the pipe
2563that it should use.
2564Then we classify traffic using a single rule:
2565.Pp
2566.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
2567.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
2568.Dl "..."
2569.Dl "ipfw table 1 add 192.168.2.0/24 1"
2570.Dl "ipfw table 1 add 192.168.0.0/27 4"
2571.Dl "ipfw table 1 add 192.168.0.2 1"
2572.Dl "..."
2573.Dl "ipfw add pipe tablearg ip from table(1) to any"
2574.Pp
2575Using the
2576.Cm fwd
2577action, the table entries may include hostnames and IP addresses.
2578.Pp
2579.Dl "ipfw table 1 add 192.168.2.0/24 10.23.2.1"
2580.Dl "ipfw table 1 add 192.168.0.0/27 router1.dmz"
2581.Dl "..."
2582.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
2583.Ss SETS OF RULES
2584To add a set of rules atomically, e.g.\& set 18:
2585.Pp
2586.Dl "ipfw set disable 18"
2587.Dl "ipfw add NN set 18 ...         # repeat as needed"
2588.Dl "ipfw set enable 18"
2589.Pp
2590To delete a set of rules atomically the command is simply:
2591.Pp
2592.Dl "ipfw delete set 18"
2593.Pp
2594To test a ruleset and disable it and regain control if something goes wrong:
2595.Pp
2596.Dl "ipfw set disable 18"
2597.Dl "ipfw add NN set 18 ...         # repeat as needed"
2598.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
2599.Pp
2600Here if everything goes well, you press control-C before the "sleep"
2601terminates, and your ruleset will be left active.
2602Otherwise, e.g.\& if
2603you cannot access your box, the ruleset will be disabled after
2604the sleep terminates thus restoring the previous situation.
2605.Pp
2606To show rules of the specific set:
2607.Pp
2608.Dl "ipfw set 18 show"
2609.Pp
2610To show rules of the disabled set:
2611.Pp
2612.Dl "ipfw -S set 18 show"
2613.Pp
2614To clear a specific rule counters of the specific set:
2615.Pp
2616.Dl "ipfw set 18 zero NN"
2617.Pp
2618To delete a specific rule of the specific set:
2619.Pp
2620.Dl "ipfw set 18 delete NN"
2621.Ss NAT, REDIRECT AND LSNAT
2622First redirect all the traffic to nat instance 123:
2623.Pp
2624.Dl "ipfw add nat 123 all from any to any"
2625.Pp
2626Then to configure nat instance 123 to alias all the outgoing traffic with ip
2627192.168.0.123, blocking all incoming connections, trying to keep
2628same ports on both sides, clearing aliasing table on address change
2629and keeping a log of traffic/link statistics:
2630.Pp
2631.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
2632.Pp
2633Or to change address of instance 123, aliasing table will be cleared (see
2634reset option):
2635.Pp
2636.Dl "ipfw nat 123 config ip 10.0.0.1"
2637.Pp
2638To see configuration of nat instance 123:
2639.Pp
2640.Dl "ipfw nat 123 show config"
2641.Pp
2642To show logs of all the instances in range 111-999:
2643.Pp
2644.Dl "ipfw nat 111-999 show"
2645.Pp
2646To see configurations of all instances:
2647.Pp
2648.Dl "ipfw nat show config"
2649.Pp
2650Or a redirect rule with mixed modes could looks like:
2651.Pp
2652.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
2653.Dl "			 redirect_port tcp 192.168.0.1:80 500"
2654.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
2655.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
2656.Dl "			 	    10.0.0.100	# LSNAT"
2657.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
2658.Dl "			 	    500		# LSNAT"
2659.Pp
2660or it could be splitted in:
2661.Pp
2662.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
2663.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
2664.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
2665.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
2666.Dl "				         10.0.0.100"
2667.Dl "ipfw nat 5 config redirect_port tcp"
2668.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
2669.Pp
2670.Sh SEE ALSO
2671.Xr cpp 1 ,
2672.Xr m4 1 ,
2673.Xr altq 4 ,
2674.Xr divert 4 ,
2675.Xr dummynet 4 ,
2676.Xr if_bridge 4 ,
2677.Xr ip 4 ,
2678.Xr ipfirewall 4 ,
2679.Xr ng_ipfw 4 ,
2680.Xr protocols 5 ,
2681.Xr services 5 ,
2682.Xr init 8 ,
2683.Xr kldload 8 ,
2684.Xr reboot 8 ,
2685.Xr sysctl 8 ,
2686.Xr syslogd 8
2687.Sh HISTORY
2688The
2689.Nm
2690utility first appeared in
2691.Fx 2.0 .
2692.Nm dummynet
2693was introduced in
2694.Fx 2.2.8 .
2695Stateful extensions were introduced in
2696.Fx 4.0 .
2697.Nm ipfw2
2698was introduced in Summer 2002.
2699.Sh AUTHORS
2700.An Ugen J. S. Antsilevich ,
2701.An Poul-Henning Kamp ,
2702.An Alex Nash ,
2703.An Archie Cobbs ,
2704.An Luigi Rizzo .
2705.Pp
2706.An -nosplit
2707API based upon code written by
2708.An Daniel Boulet
2709for BSDI.
2710.Pp
2711.An -nosplit
2712In-kernel NAT support written by
2713.An Paolo Pisati Aq piso@FreeBSD.org
2714as part of a Summer of Code 2005 project.
2715.Pp
2716Work on
2717.Nm dummynet
2718traffic shaper supported by Akamba Corp.
2719.Sh BUGS
2720The syntax has grown over the years and sometimes it might be confusing.
2721Unfortunately, backward compatibility prevents cleaning up mistakes
2722made in the definition of the syntax.
2723.Pp
2724.Em !!! WARNING !!!
2725.Pp
2726Misconfiguring the firewall can put your computer in an unusable state,
2727possibly shutting down network services and requiring console access to
2728regain control of it.
2729.Pp
2730Incoming packet fragments diverted by
2731.Cm divert
2732are reassembled before delivery to the socket.
2733The action used on those packet is the one from the
2734rule which matches the first fragment of the packet.
2735.Pp
2736Packets diverted to userland, and then reinserted by a userland process
2737may lose various packet attributes.
2738The packet source interface name
2739will be preserved if it is shorter than 8 bytes and the userland process
2740saves and reuses the sockaddr_in
2741(as does
2742.Xr natd 8 ) ;
2743otherwise, it may be lost.
2744If a packet is reinserted in this manner, later rules may be incorrectly
2745applied, making the order of
2746.Cm divert
2747rules in the rule sequence very important.
2748.Pp
2749Dummynet drops all packets with IPv6 link-local addresses.
2750.Pp
2751Rules using
2752.Cm uid
2753or
2754.Cm gid
2755may not behave as expected.
2756In particular, incoming SYN packets may
2757have no uid or gid associated with them since they do not yet belong
2758to a TCP connection, and the uid/gid associated with a packet may not
2759be as expected if the associated process calls
2760.Xr setuid 2
2761or similar system calls.
2762.Pp
2763Rule syntax is subject to the command line environment and some patterns
2764may need to be escaped with the backslash character
2765or quoted appropriately.
2766.Pp
2767Due to the architecture of
2768.Xr libalias 3 ,
2769ipfw nat is not compatible with the tcp segmentation offloading
2770(TSO). Thus, to reliably nat your network traffic, please disable TSO
2771on your NICs using
2772.Xr ifconfig 8 .
2773.Pp
2774ICMP error messages are not implicitly matched by dynamic rules
2775for the respective conversations.
2776To avoid failures of network error detection and path MTU discovery,
2777ICMP error messages may need to be allowed explicitly through static
2778rules.
2779