xref: /freebsd/sbin/ipfw/ipfw.8 (revision 74ca7bf1d4c7173d5575ba168bc4b5f6d181ff5a)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd March 19, 2018
5.Dt IPFW 8
6.Os
7.Sh NAME
8.Nm ipfw
9.Nd User interface for firewall, traffic shaper, packet scheduler,
10in-kernel NAT.
11.Sh SYNOPSIS
12.Ss FIREWALL CONFIGURATION
13.Nm
14.Op Fl cq
15.Cm add
16.Ar rule
17.Nm
18.Op Fl acdefnNStT
19.Op Cm set Ar N
20.Brq Cm list | show
21.Op Ar rule | first-last ...
22.Nm
23.Op Fl f | q
24.Op Cm set Ar N
25.Cm flush
26.Nm
27.Op Fl q
28.Op Cm set Ar N
29.Brq Cm delete | zero | resetlog
30.Op Ar number ...
31.Pp
32.Nm
33.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
34.Nm
35.Cm set move
36.Op Cm rule
37.Ar number Cm to Ar number
38.Nm
39.Cm set swap Ar number number
40.Nm
41.Cm set show
42.Ss SYSCTL SHORTCUTS
43.Nm
44.Cm enable
45.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
46.Nm
47.Cm disable
48.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
49.Ss LOOKUP TABLES
50.Nm
51.Oo Cm set Ar N Oc Cm table Ar name Cm create Ar create-options
52.Nm
53.Oo Cm set Ar N Oc Cm table
54.Brq Ar name | all
55.Cm destroy
56.Nm
57.Oo Cm set Ar N Oc Cm table Ar name Cm modify Ar modify-options
58.Nm
59.Oo Cm set Ar N Oc Cm table Ar name Cm swap Ar name
60.Nm
61.Oo Cm set Ar N Oc Cm table Ar name Cm add Ar table-key Op Ar value
62.Nm
63.Oo Cm set Ar N Oc Cm table Ar name Cm add Op Ar table-key Ar value ...
64.Nm
65.Oo Cm set Ar N Oc Cm table Ar name Cm atomic add Op Ar table-key Ar value ...
66.Nm
67.Oo Cm set Ar N Oc Cm table Ar name Cm delete Op Ar table-key ...
68.Nm
69.Oo Cm set Ar N Oc Cm table Ar name Cm lookup Ar addr
70.Nm
71.Oo Cm set Ar N Oc Cm table Ar name Cm lock
72.Nm
73.Oo Cm set Ar N Oc Cm table Ar name Cm unlock
74.Nm
75.Oo Cm set Ar N Oc Cm table
76.Brq Ar name | all
77.Cm list
78.Nm
79.Oo Cm set Ar N Oc Cm table
80.Brq Ar name | all
81.Cm info
82.Nm
83.Oo Cm set Ar N Oc Cm table
84.Brq Ar name | all
85.Cm detail
86.Nm
87.Oo Cm set Ar N Oc Cm table
88.Brq Ar name | all
89.Cm flush
90.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
91.Nm
92.Brq Cm pipe | queue | sched
93.Ar number
94.Cm config
95.Ar config-options
96.Nm
97.Op Fl s Op Ar field
98.Brq Cm pipe | queue | sched
99.Brq Cm delete | list | show
100.Op Ar number ...
101.Ss IN-KERNEL NAT
102.Nm
103.Op Fl q
104.Cm nat
105.Ar number
106.Cm config
107.Ar config-options
108.Pp
109.Nm
110.Op Fl cfnNqS
111.Oo
112.Fl p Ar preproc
113.Oo
114.Ar preproc-flags
115.Oc
116.Oc
117.Ar pathname
118.Ss STATEFUL IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
119.Nm
120.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm create Ar create-options
121.Nm
122.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm config Ar config-options
123.Nm
124.Oo Cm set Ar N Oc Cm nat64lsn
125.Brq Ar name | all
126.Brq Cm list | show
127.Op Cm states
128.Nm
129.Oo Cm set Ar N Oc Cm nat64lsn
130.Brq Ar name | all
131.Cm destroy
132.Nm
133.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm stats Op Cm reset
134.Ss STATELESS IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
135.Nm
136.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm create Ar create-options
137.Nm
138.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm config Ar config-options
139.Nm
140.Oo Cm set Ar N Oc Cm nat64stl
141.Brq Ar name | all
142.Brq Cm list | show
143.Nm
144.Oo Cm set Ar N Oc Cm nat64stl
145.Brq Ar name | all
146.Cm destroy
147.Nm
148.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm stats Op Cm reset
149.Ss IPv6-to-IPv6 NETWORK PREFIX TRANSLATION
150.Nm
151.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm create Ar create-options
152.Nm
153.Oo Cm set Ar N Oc Cm nptv6
154.Brq Ar name | all
155.Brq Cm list | show
156.Nm
157.Oo Cm set Ar N Oc Cm nptv6
158.Brq Ar name | all
159.Cm destroy
160.Nm
161.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm stats Op Cm reset
162.Ss INTERNAL DIAGNOSTICS
163.Nm
164.Cm internal iflist
165.Nm
166.Cm internal talist
167.Nm
168.Cm internal vlist
169.Sh DESCRIPTION
170The
171.Nm
172utility is the user interface for controlling the
173.Xr ipfw 4
174firewall, the
175.Xr dummynet 4
176traffic shaper/packet scheduler, and the
177in-kernel NAT services.
178.Pp
179A firewall configuration, or
180.Em ruleset ,
181is made of a list of
182.Em rules
183numbered from 1 to 65535.
184Packets are passed to the firewall
185from a number of different places in the protocol stack
186(depending on the source and destination of the packet,
187it is possible for the firewall to be
188invoked multiple times on the same packet).
189The packet passed to the firewall is compared
190against each of the rules in the
191.Em ruleset ,
192in rule-number order
193(multiple rules with the same number are permitted, in which case
194they are processed in order of insertion).
195When a match is found, the action corresponding to the
196matching rule is performed.
197.Pp
198Depending on the action and certain system settings, packets
199can be reinjected into the firewall at some rule after the
200matching one for further processing.
201.Pp
202A ruleset always includes a
203.Em default
204rule (numbered 65535) which cannot be modified or deleted,
205and matches all packets.
206The action associated with the
207.Em default
208rule can be either
209.Cm deny
210or
211.Cm allow
212depending on how the kernel is configured.
213.Pp
214If the ruleset includes one or more rules with the
215.Cm keep-state
216or
217.Cm limit
218option,
219the firewall will have a
220.Em stateful
221behaviour, i.e., upon a match it will create
222.Em dynamic rules ,
223i.e., rules that match packets with the same 5-tuple
224(protocol, source and destination addresses and ports)
225as the packet which caused their creation.
226Dynamic rules, which have a limited lifetime, are checked
227at the first occurrence of a
228.Cm check-state ,
229.Cm keep-state
230or
231.Cm limit
232rule, and are typically used to open the firewall on-demand to
233legitimate traffic only.
234See the
235.Sx STATEFUL FIREWALL
236and
237.Sx EXAMPLES
238Sections below for more information on the stateful behaviour of
239.Nm .
240.Pp
241All rules (including dynamic ones) have a few associated counters:
242a packet count, a byte count, a log count and a timestamp
243indicating the time of the last match.
244Counters can be displayed or reset with
245.Nm
246commands.
247.Pp
248Each rule belongs to one of 32 different
249.Em sets
250, and there are
251.Nm
252commands to atomically manipulate sets, such as enable,
253disable, swap sets, move all rules in a set to another
254one, delete all rules in a set.
255These can be useful to
256install temporary configurations, or to test them.
257See Section
258.Sx SETS OF RULES
259for more information on
260.Em sets .
261.Pp
262Rules can be added with the
263.Cm add
264command; deleted individually or in groups with the
265.Cm delete
266command, and globally (except those in set 31) with the
267.Cm flush
268command; displayed, optionally with the content of the
269counters, using the
270.Cm show
271and
272.Cm list
273commands.
274Finally, counters can be reset with the
275.Cm zero
276and
277.Cm resetlog
278commands.
279.Pp
280.Ss COMMAND OPTIONS
281The following general options are available when invoking
282.Nm :
283.Bl -tag -width indent
284.It Fl a
285Show counter values when listing rules.
286The
287.Cm show
288command implies this option.
289.It Fl b
290Only show the action and the comment, not the body of a rule.
291Implies
292.Fl c .
293.It Fl c
294When entering or showing rules, print them in compact form,
295i.e., omitting the "ip from any to any" string
296when this does not carry any additional information.
297.It Fl d
298When listing, show dynamic rules in addition to static ones.
299.It Fl e
300When listing and
301.Fl d
302is specified, also show expired dynamic rules.
303.It Fl f
304Do not ask for confirmation for commands that can cause problems
305if misused, i.e.,
306.Cm flush .
307If there is no tty associated with the process, this is implied.
308.It Fl i
309When listing a table (see the
310.Sx LOOKUP TABLES
311section below for more information on lookup tables), format values
312as IP addresses.
313By default, values are shown as integers.
314.It Fl n
315Only check syntax of the command strings, without actually passing
316them to the kernel.
317.It Fl N
318Try to resolve addresses and service names in output.
319.It Fl q
320Be quiet when executing the
321.Cm add ,
322.Cm nat ,
323.Cm zero ,
324.Cm resetlog
325or
326.Cm flush
327commands;
328(implies
329.Fl f ) .
330This is useful when updating rulesets by executing multiple
331.Nm
332commands in a script
333(e.g.,
334.Ql sh\ /etc/rc.firewall ) ,
335or by processing a file with many
336.Nm
337rules across a remote login session.
338It also stops a table add or delete
339from failing if the entry already exists or is not present.
340.Pp
341The reason why this option may be important is that
342for some of these actions,
343.Nm
344may print a message; if the action results in blocking the
345traffic to the remote client,
346the remote login session will be closed
347and the rest of the ruleset will not be processed.
348Access to the console would then be required to recover.
349.It Fl S
350When listing rules, show the
351.Em set
352each rule belongs to.
353If this flag is not specified, disabled rules will not be
354listed.
355.It Fl s Op Ar field
356When listing pipes, sort according to one of the four
357counters (total or current packets or bytes).
358.It Fl t
359When listing, show last match timestamp converted with ctime().
360.It Fl T
361When listing, show last match timestamp as seconds from the epoch.
362This form can be more convenient for postprocessing by scripts.
363.El
364.Ss LIST OF RULES AND PREPROCESSING
365To ease configuration, rules can be put into a file which is
366processed using
367.Nm
368as shown in the last synopsis line.
369An absolute
370.Ar pathname
371must be used.
372The file will be read line by line and applied as arguments to the
373.Nm
374utility.
375.Pp
376Optionally, a preprocessor can be specified using
377.Fl p Ar preproc
378where
379.Ar pathname
380is to be piped through.
381Useful preprocessors include
382.Xr cpp 1
383and
384.Xr m4 1 .
385If
386.Ar preproc
387does not start with a slash
388.Pq Ql /
389as its first character, the usual
390.Ev PATH
391name search is performed.
392Care should be taken with this in environments where not all
393file systems are mounted (yet) by the time
394.Nm
395is being run (e.g.\& when they are mounted over NFS).
396Once
397.Fl p
398has been specified, any additional arguments are passed on to the preprocessor
399for interpretation.
400This allows for flexible configuration files (like conditionalizing
401them on the local hostname) and the use of macros to centralize
402frequently required arguments like IP addresses.
403.Ss TRAFFIC SHAPER CONFIGURATION
404The
405.Nm
406.Cm pipe , queue
407and
408.Cm sched
409commands are used to configure the traffic shaper and packet scheduler.
410See the
411.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
412Section below for details.
413.Pp
414If the world and the kernel get out of sync the
415.Nm
416ABI may break, preventing you from being able to add any rules.
417This can adversely affect the booting process.
418You can use
419.Nm
420.Cm disable
421.Cm firewall
422to temporarily disable the firewall to regain access to the network,
423allowing you to fix the problem.
424.Sh PACKET FLOW
425A packet is checked against the active ruleset in multiple places
426in the protocol stack, under control of several sysctl variables.
427These places and variables are shown below, and it is important to
428have this picture in mind in order to design a correct ruleset.
429.Bd -literal -offset indent
430       ^    to upper layers    V
431       |                       |
432       +----------->-----------+
433       ^                       V
434 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
435       |                       |
436       ^                       V
437 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
438       |                       |
439       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
440       ^                       V
441       |      to devices       |
442.Ed
443.Pp
444The number of
445times the same packet goes through the firewall can
446vary between 0 and 4 depending on packet source and
447destination, and system configuration.
448.Pp
449Note that as packets flow through the stack, headers can be
450stripped or added to it, and so they may or may not be available
451for inspection.
452E.g., incoming packets will include the MAC header when
453.Nm
454is invoked from
455.Cm ether_demux() ,
456but the same packets will have the MAC header stripped off when
457.Nm
458is invoked from
459.Cm ip_input()
460or
461.Cm ip6_input() .
462.Pp
463Also note that each packet is always checked against the complete ruleset,
464irrespective of the place where the check occurs, or the source of the packet.
465If a rule contains some match patterns or actions which are not valid
466for the place of invocation (e.g.\& trying to match a MAC header within
467.Cm ip_input
468or
469.Cm ip6_input ),
470the match pattern will not match, but a
471.Cm not
472operator in front of such patterns
473.Em will
474cause the pattern to
475.Em always
476match on those packets.
477It is thus the responsibility of
478the programmer, if necessary, to write a suitable ruleset to
479differentiate among the possible places.
480.Cm skipto
481rules can be useful here, as an example:
482.Bd -literal -offset indent
483# packets from ether_demux or bdg_forward
484ipfw add 10 skipto 1000 all from any to any layer2 in
485# packets from ip_input
486ipfw add 10 skipto 2000 all from any to any not layer2 in
487# packets from ip_output
488ipfw add 10 skipto 3000 all from any to any not layer2 out
489# packets from ether_output_frame
490ipfw add 10 skipto 4000 all from any to any layer2 out
491.Ed
492.Pp
493(yes, at the moment there is no way to differentiate between
494ether_demux and bdg_forward).
495.Sh SYNTAX
496In general, each keyword or argument must be provided as
497a separate command line argument, with no leading or trailing
498spaces.
499Keywords are case-sensitive, whereas arguments may
500or may not be case-sensitive depending on their nature
501(e.g.\& uid's are, hostnames are not).
502.Pp
503Some arguments (e.g., port or address lists) are comma-separated
504lists of values.
505In this case, spaces after commas ',' are allowed to make
506the line more readable.
507You can also put the entire
508command (including flags) into a single argument.
509E.g., the following forms are equivalent:
510.Bd -literal -offset indent
511ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
512ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
513ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
514.Ed
515.Sh RULE FORMAT
516The format of firewall rules is the following:
517.Bd -ragged -offset indent
518.Bk -words
519.Op Ar rule_number
520.Op Cm set Ar set_number
521.Op Cm prob Ar match_probability
522.Ar action
523.Op Cm log Op Cm logamount Ar number
524.Op Cm altq Ar queue
525.Oo
526.Bro Cm tag | untag
527.Brc Ar number
528.Oc
529.Ar body
530.Ek
531.Ed
532.Pp
533where the body of the rule specifies which information is used
534for filtering packets, among the following:
535.Pp
536.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
537.It Layer-2 header fields
538When available
539.It IPv4 and IPv6 Protocol
540SCTP, TCP, UDP, ICMP, etc.
541.It Source and dest. addresses and ports
542.It Direction
543See Section
544.Sx PACKET FLOW
545.It Transmit and receive interface
546By name or address
547.It Misc. IP header fields
548Version, type of service, datagram length, identification,
549fragment flag (non-zero IP offset),
550Time To Live
551.It IP options
552.It IPv6 Extension headers
553Fragmentation, Hop-by-Hop options,
554Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
555.It IPv6 Flow-ID
556.It Misc. TCP header fields
557TCP flags (SYN, FIN, ACK, RST, etc.),
558sequence number, acknowledgment number,
559window
560.It TCP options
561.It ICMP types
562for ICMP packets
563.It ICMP6 types
564for ICMP6 packets
565.It User/group ID
566When the packet can be associated with a local socket.
567.It Divert status
568Whether a packet came from a divert socket (e.g.,
569.Xr natd 8 ) .
570.It Fib annotation state
571Whether a packet has been tagged for using a specific FIB (routing table)
572in future forwarding decisions.
573.El
574.Pp
575Note that some of the above information, e.g.\& source MAC or IP addresses and
576TCP/UDP ports, can be easily spoofed, so filtering on those fields
577alone might not guarantee the desired results.
578.Bl -tag -width indent
579.It Ar rule_number
580Each rule is associated with a
581.Ar rule_number
582in the range 1..65535, with the latter reserved for the
583.Em default
584rule.
585Rules are checked sequentially by rule number.
586Multiple rules can have the same number, in which case they are
587checked (and listed) according to the order in which they have
588been added.
589If a rule is entered without specifying a number, the kernel will
590assign one in such a way that the rule becomes the last one
591before the
592.Em default
593rule.
594Automatic rule numbers are assigned by incrementing the last
595non-default rule number by the value of the sysctl variable
596.Ar net.inet.ip.fw.autoinc_step
597which defaults to 100.
598If this is not possible (e.g.\& because we would go beyond the
599maximum allowed rule number), the number of the last
600non-default value is used instead.
601.It Cm set Ar set_number
602Each rule is associated with a
603.Ar set_number
604in the range 0..31.
605Sets can be individually disabled and enabled, so this parameter
606is of fundamental importance for atomic ruleset manipulation.
607It can be also used to simplify deletion of groups of rules.
608If a rule is entered without specifying a set number,
609set 0 will be used.
610.br
611Set 31 is special in that it cannot be disabled,
612and rules in set 31 are not deleted by the
613.Nm ipfw flush
614command (but you can delete them with the
615.Nm ipfw delete set 31
616command).
617Set 31 is also used for the
618.Em default
619rule.
620.It Cm prob Ar match_probability
621A match is only declared with the specified probability
622(floating point number between 0 and 1).
623This can be useful for a number of applications such as
624random packet drop or
625(in conjunction with
626.Nm dummynet )
627to simulate the effect of multiple paths leading to out-of-order
628packet delivery.
629.Pp
630Note: this condition is checked before any other condition, including
631ones such as keep-state or check-state which might have side effects.
632.It Cm log Op Cm logamount Ar number
633Packets matching a rule with the
634.Cm log
635keyword will be made available for logging in two ways:
636if the sysctl variable
637.Va net.inet.ip.fw.verbose
638is set to 0 (default), one can use
639.Xr bpf 4
640attached to the
641.Li ipfw0
642pseudo interface.
643This pseudo interface can be created after a boot
644manually by using the following command:
645.Bd -literal -offset indent
646# ifconfig ipfw0 create
647.Ed
648.Pp
649Or, automatically at boot time by adding the following
650line to the
651.Xr rc.conf 5
652file:
653.Bd -literal -offset indent
654firewall_logif="YES"
655.Ed
656.Pp
657There is no overhead if no
658.Xr bpf 4
659is attached to the pseudo interface.
660.Pp
661If
662.Va net.inet.ip.fw.verbose
663is set to 1, packets will be logged to
664.Xr syslogd 8
665with a
666.Dv LOG_SECURITY
667facility up to a maximum of
668.Cm logamount
669packets.
670If no
671.Cm logamount
672is specified, the limit is taken from the sysctl variable
673.Va net.inet.ip.fw.verbose_limit .
674In both cases, a value of 0 means unlimited logging.
675.Pp
676Once the limit is reached, logging can be re-enabled by
677clearing the logging counter or the packet counter for that entry, see the
678.Cm resetlog
679command.
680.Pp
681Note: logging is done after all other packet matching conditions
682have been successfully verified, and before performing the final
683action (accept, deny, etc.) on the packet.
684.It Cm tag Ar number
685When a packet matches a rule with the
686.Cm tag
687keyword, the numeric tag for the given
688.Ar number
689in the range 1..65534 will be attached to the packet.
690The tag acts as an internal marker (it is not sent out over
691the wire) that can be used to identify these packets later on.
692This can be used, for example, to provide trust between interfaces
693and to start doing policy-based filtering.
694A packet can have multiple tags at the same time.
695Tags are "sticky", meaning once a tag is applied to a packet by a
696matching rule it exists until explicit removal.
697Tags are kept with the packet everywhere within the kernel, but are
698lost when packet leaves the kernel, for example, on transmitting
699packet out to the network or sending packet to a
700.Xr divert 4
701socket.
702.Pp
703To check for previously applied tags, use the
704.Cm tagged
705rule option.
706To delete previously applied tag, use the
707.Cm untag
708keyword.
709.Pp
710Note: since tags are kept with the packet everywhere in kernelspace,
711they can be set and unset anywhere in the kernel network subsystem
712(using the
713.Xr mbuf_tags 9
714facility), not only by means of the
715.Xr ipfw 4
716.Cm tag
717and
718.Cm untag
719keywords.
720For example, there can be a specialized
721.Xr netgraph 4
722node doing traffic analyzing and tagging for later inspecting
723in firewall.
724.It Cm untag Ar number
725When a packet matches a rule with the
726.Cm untag
727keyword, the tag with the number
728.Ar number
729is searched among the tags attached to this packet and,
730if found, removed from it.
731Other tags bound to packet, if present, are left untouched.
732.It Cm altq Ar queue
733When a packet matches a rule with the
734.Cm altq
735keyword, the ALTQ identifier for the given
736.Ar queue
737(see
738.Xr altq 4 )
739will be attached.
740Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
741and not being rejected or going to divert sockets.
742Note that if there is insufficient memory at the time the packet is
743processed, it will not be tagged, so it is wise to make your ALTQ
744"default" queue policy account for this.
745If multiple
746.Cm altq
747rules match a single packet, only the first one adds the ALTQ classification
748tag.
749In doing so, traffic may be shaped by using
750.Cm count Cm altq Ar queue
751rules for classification early in the ruleset, then later applying
752the filtering decision.
753For example,
754.Cm check-state
755and
756.Cm keep-state
757rules may come later and provide the actual filtering decisions in
758addition to the fallback ALTQ tag.
759.Pp
760You must run
761.Xr pfctl 8
762to set up the queues before IPFW will be able to look them up by name,
763and if the ALTQ disciplines are rearranged, the rules in containing the
764queue identifiers in the kernel will likely have gone stale and need
765to be reloaded.
766Stale queue identifiers will probably result in misclassification.
767.Pp
768All system ALTQ processing can be turned on or off via
769.Nm
770.Cm enable Ar altq
771and
772.Nm
773.Cm disable Ar altq .
774The usage of
775.Va net.inet.ip.fw.one_pass
776is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
777always after adding an ALTQ tag.
778.El
779.Ss RULE ACTIONS
780A rule can be associated with one of the following actions, which
781will be executed when the packet matches the body of the rule.
782.Bl -tag -width indent
783.It Cm allow | accept | pass | permit
784Allow packets that match rule.
785The search terminates.
786.It Cm check-state Op Ar :flowname | Cm :any
787Checks the packet against the dynamic ruleset.
788If a match is found, execute the action associated with
789the rule which generated this dynamic rule, otherwise
790move to the next rule.
791.br
792.Cm Check-state
793rules do not have a body.
794If no
795.Cm check-state
796rule is found, the dynamic ruleset is checked at the first
797.Cm keep-state
798or
799.Cm limit
800rule.
801The
802.Ar :flowname
803is symbolic name assigned to dynamic rule by
804.Cm keep-state
805opcode.
806The special flowname
807.Cm :any
808can be used to ignore states flowname when matching.
809The
810.Cm :default
811keyword is special name used for compatibility with old rulesets.
812.It Cm count
813Update counters for all packets that match rule.
814The search continues with the next rule.
815.It Cm deny | drop
816Discard packets that match this rule.
817The search terminates.
818.It Cm divert Ar port
819Divert packets that match this rule to the
820.Xr divert 4
821socket bound to port
822.Ar port .
823The search terminates.
824.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
825Change the next-hop on matching packets to
826.Ar ipaddr ,
827which can be an IP address or a host name.
828The next hop can also be supplied by the last table
829looked up for the packet by using the
830.Cm tablearg
831keyword instead of an explicit address.
832The search terminates if this rule matches.
833.Pp
834If
835.Ar ipaddr
836is a local address, then matching packets will be forwarded to
837.Ar port
838(or the port number in the packet if one is not specified in the rule)
839on the local machine.
840.br
841If
842.Ar ipaddr
843is not a local address, then the port number
844(if specified) is ignored, and the packet will be
845forwarded to the remote address, using the route as found in
846the local routing table for that IP.
847.br
848A
849.Ar fwd
850rule will not match layer-2 packets (those received
851on ether_input, ether_output, or bridged).
852.br
853The
854.Cm fwd
855action does not change the contents of the packet at all.
856In particular, the destination address remains unmodified, so
857packets forwarded to another system will usually be rejected by that system
858unless there is a matching rule on that system to capture them.
859For packets forwarded locally,
860the local address of the socket will be
861set to the original destination address of the packet.
862This makes the
863.Xr netstat 1
864entry look rather weird but is intended for
865use with transparent proxy servers.
866.It Cm nat Ar nat_nr | tablearg
867Pass packet to a
868nat instance
869(for network address translation, address redirect, etc.):
870see the
871.Sx NETWORK ADDRESS TRANSLATION (NAT)
872Section for further information.
873.It Cm nat64lsn Ar name
874Pass packet to a stateful NAT64 instance (for IPv6/IPv4 network address and
875protocol translation): see the
876.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
877Section for further information.
878.It Cm nat64stl Ar name
879Pass packet to a stateless NAT64 instance (for IPv6/IPv4 network address and
880protocol translation): see the
881.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
882Section for further information.
883.It Cm nptv6 Ar name
884Pass packet to a NPTv6 instance (for IPv6-to-IPv6 network prefix translation):
885see the
886.Sx IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
887Section for further information.
888.It Cm pipe Ar pipe_nr
889Pass packet to a
890.Nm dummynet
891.Dq pipe
892(for bandwidth limitation, delay, etc.).
893See the
894.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
895Section for further information.
896The search terminates; however, on exit from the pipe and if
897the
898.Xr sysctl 8
899variable
900.Va net.inet.ip.fw.one_pass
901is not set, the packet is passed again to the firewall code
902starting from the next rule.
903.It Cm queue Ar queue_nr
904Pass packet to a
905.Nm dummynet
906.Dq queue
907(for bandwidth limitation using WF2Q+).
908.It Cm reject
909(Deprecated).
910Synonym for
911.Cm unreach host .
912.It Cm reset
913Discard packets that match this rule, and if the
914packet is a TCP packet, try to send a TCP reset (RST) notice.
915The search terminates.
916.It Cm reset6
917Discard packets that match this rule, and if the
918packet is a TCP packet, try to send a TCP reset (RST) notice.
919The search terminates.
920.It Cm skipto Ar number | tablearg
921Skip all subsequent rules numbered less than
922.Ar number .
923The search continues with the first rule numbered
924.Ar number
925or higher.
926It is possible to use the
927.Cm tablearg
928keyword with a skipto for a
929.Em computed
930skipto. Skipto may work either in O(log(N)) or in O(1) depending
931on amount of memory and/or sysctl variables.
932See the
933.Sx SYSCTL VARIABLES
934section for more details.
935.It Cm call Ar number | tablearg
936The current rule number is saved in the internal stack and
937ruleset processing continues with the first rule numbered
938.Ar number
939or higher.
940If later a rule with the
941.Cm return
942action is encountered, the processing returns to the first rule
943with number of this
944.Cm call
945rule plus one or higher
946(the same behaviour as with packets returning from
947.Xr divert 4
948socket after a
949.Cm divert
950action).
951This could be used to make somewhat like an assembly language
952.Dq subroutine
953calls to rules with common checks for different interfaces, etc.
954.Pp
955Rule with any number could be called, not just forward jumps as with
956.Cm skipto .
957So, to prevent endless loops in case of mistakes, both
958.Cm call
959and
960.Cm return
961actions don't do any jumps and simply go to the next rule if memory
962cannot be allocated or stack overflowed/underflowed.
963.Pp
964Internally stack for rule numbers is implemented using
965.Xr mbuf_tags 9
966facility and currently has size of 16 entries.
967As mbuf tags are lost when packet leaves the kernel,
968.Cm divert
969should not be used in subroutines to avoid endless loops
970and other undesired effects.
971.It Cm return
972Takes rule number saved to internal stack by the last
973.Cm call
974action and returns ruleset processing to the first rule
975with number greater than number of corresponding
976.Cm call
977rule.
978See description of the
979.Cm call
980action for more details.
981.Pp
982Note that
983.Cm return
984rules usually end a
985.Dq subroutine
986and thus are unconditional, but
987.Nm
988command-line utility currently requires every action except
989.Cm check-state
990to have body.
991While it is sometimes useful to return only on some packets,
992usually you want to print just
993.Dq return
994for readability.
995A workaround for this is to use new syntax and
996.Fl c
997switch:
998.Bd -literal -offset indent
999# Add a rule without actual body
1000ipfw add 2999 return via any
1001
1002# List rules without "from any to any" part
1003ipfw -c list
1004.Ed
1005.Pp
1006This cosmetic annoyance may be fixed in future releases.
1007.It Cm tee Ar port
1008Send a copy of packets matching this rule to the
1009.Xr divert 4
1010socket bound to port
1011.Ar port .
1012The search continues with the next rule.
1013.It Cm unreach Ar code
1014Discard packets that match this rule, and try to send an ICMP
1015unreachable notice with code
1016.Ar code ,
1017where
1018.Ar code
1019is a number from 0 to 255, or one of these aliases:
1020.Cm net , host , protocol , port ,
1021.Cm needfrag , srcfail , net-unknown , host-unknown ,
1022.Cm isolated , net-prohib , host-prohib , tosnet ,
1023.Cm toshost , filter-prohib , host-precedence
1024or
1025.Cm precedence-cutoff .
1026The search terminates.
1027.It Cm unreach6 Ar code
1028Discard packets that match this rule, and try to send an ICMPv6
1029unreachable notice with code
1030.Ar code ,
1031where
1032.Ar code
1033is a number from 0, 1, 3 or 4, or one of these aliases:
1034.Cm no-route, admin-prohib, address
1035or
1036.Cm port .
1037The search terminates.
1038.It Cm netgraph Ar cookie
1039Divert packet into netgraph with given
1040.Ar cookie .
1041The search terminates.
1042If packet is later returned from netgraph it is either
1043accepted or continues with the next rule, depending on
1044.Va net.inet.ip.fw.one_pass
1045sysctl variable.
1046.It Cm ngtee Ar cookie
1047A copy of packet is diverted into netgraph, original
1048packet continues with the next rule.
1049See
1050.Xr ng_ipfw 4
1051for more information on
1052.Cm netgraph
1053and
1054.Cm ngtee
1055actions.
1056.It Cm setfib Ar fibnum | tablearg
1057The packet is tagged so as to use the FIB (routing table)
1058.Ar fibnum
1059in any subsequent forwarding decisions.
1060In the current implementation, this is limited to the values 0 through 15, see
1061.Xr setfib 2 .
1062Processing continues at the next rule.
1063It is possible to use the
1064.Cm tablearg
1065keyword with setfib.
1066If the tablearg value is not within the compiled range of fibs,
1067the packet's fib is set to 0.
1068.It Cm setdscp Ar DSCP | number | tablearg
1069Set specified DiffServ codepoint for an IPv4/IPv6 packet.
1070Processing continues at the next rule.
1071Supported values are:
1072.Pp
1073.Cm cs0
1074.Pq Dv 000000 ,
1075.Cm cs1
1076.Pq Dv 001000 ,
1077.Cm cs2
1078.Pq Dv 010000 ,
1079.Cm cs3
1080.Pq Dv 011000 ,
1081.Cm cs4
1082.Pq Dv 100000 ,
1083.Cm cs5
1084.Pq Dv 101000 ,
1085.Cm cs6
1086.Pq Dv 110000 ,
1087.Cm cs7
1088.Pq Dv 111000 ,
1089.Cm af11
1090.Pq Dv 001010 ,
1091.Cm af12
1092.Pq Dv 001100 ,
1093.Cm af13
1094.Pq Dv 001110 ,
1095.Cm af21
1096.Pq Dv 010010 ,
1097.Cm af22
1098.Pq Dv 010100 ,
1099.Cm af23
1100.Pq Dv 010110 ,
1101.Cm af31
1102.Pq Dv 011010 ,
1103.Cm af32
1104.Pq Dv 011100 ,
1105.Cm af33
1106.Pq Dv 011110 ,
1107.Cm af41
1108.Pq Dv 100010 ,
1109.Cm af42
1110.Pq Dv 100100 ,
1111.Cm af43
1112.Pq Dv 100110 ,
1113.Cm ef
1114.Pq Dv 101110 ,
1115.Cm be
1116.Pq Dv 000000 .
1117Additionally, DSCP value can be specified by number (0..64).
1118It is also possible to use the
1119.Cm tablearg
1120keyword with setdscp.
1121If the tablearg value is not within the 0..64 range, lower 6 bits of supplied
1122value are used.
1123.It Cm tcp-setmss Ar mss
1124Set the Maximum Segment Size (MSS) in the TCP segment to value
1125.Ar mss .
1126The kernel module
1127.Cm ipfw_pmod
1128should be loaded or kernel should have
1129.Cm options IPFIREWALL_PMOD
1130to be able use this action.
1131This command does not change a packet if original MSS value is lower than
1132specified value.
1133Both TCP over IPv4 and over IPv6 are supported.
1134Regardless of matched a packet or not by the
1135.Cm tcp-setmss
1136rule, the search continues with the next rule.
1137.It Cm reass
1138Queue and reassemble IPv4 fragments.
1139If the packet is not fragmented, counters are updated and
1140processing continues with the next rule.
1141If the packet is the last logical fragment, the packet is reassembled and, if
1142.Va net.inet.ip.fw.one_pass
1143is set to 0, processing continues with the next rule.
1144Otherwise, the packet is allowed to pass and the search terminates.
1145If the packet is a fragment in the middle of a logical group of fragments,
1146it is consumed and
1147processing stops immediately.
1148.Pp
1149Fragment handling can be tuned via
1150.Va net.inet.ip.maxfragpackets
1151and
1152.Va net.inet.ip.maxfragsperpacket
1153which limit, respectively, the maximum number of processable
1154fragments (default: 800) and
1155the maximum number of fragments per packet (default: 16).
1156.Pp
1157NOTA BENE: since fragments do not contain port numbers,
1158they should be avoided with the
1159.Nm reass
1160rule.
1161Alternatively, direction-based (like
1162.Nm in
1163/
1164.Nm out
1165) and source-based (like
1166.Nm via
1167) match patterns can be used to select fragments.
1168.Pp
1169Usually a simple rule like:
1170.Bd -literal -offset indent
1171# reassemble incoming fragments
1172ipfw add reass all from any to any in
1173.Ed
1174.Pp
1175is all you need at the beginning of your ruleset.
1176.It Cm abort
1177Discard packets that match this rule, and if the packet is an SCTP packet,
1178try to send an SCTP packet containing an ABORT chunk.
1179The search terminates.
1180.It Cm abort6
1181Discard packets that match this rule, and if the packet is an SCTP packet,
1182try to send an SCTP packet containing an ABORT chunk.
1183The search terminates.
1184.El
1185.Ss RULE BODY
1186The body of a rule contains zero or more patterns (such as
1187specific source and destination addresses or ports,
1188protocol options, incoming or outgoing interfaces, etc.)
1189that the packet must match in order to be recognised.
1190In general, the patterns are connected by (implicit)
1191.Cm and
1192operators -- i.e., all must match in order for the
1193rule to match.
1194Individual patterns can be prefixed by the
1195.Cm not
1196operator to reverse the result of the match, as in
1197.Pp
1198.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
1199.Pp
1200Additionally, sets of alternative match patterns
1201.Pq Em or-blocks
1202can be constructed by putting the patterns in
1203lists enclosed between parentheses ( ) or braces { }, and
1204using the
1205.Cm or
1206operator as follows:
1207.Pp
1208.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
1209.Pp
1210Only one level of parentheses is allowed.
1211Beware that most shells have special meanings for parentheses
1212or braces, so it is advisable to put a backslash \\ in front of them
1213to prevent such interpretations.
1214.Pp
1215The body of a rule must in general include a source and destination
1216address specifier.
1217The keyword
1218.Ar any
1219can be used in various places to specify that the content of
1220a required field is irrelevant.
1221.Pp
1222The rule body has the following format:
1223.Bd -ragged -offset indent
1224.Op Ar proto Cm from Ar src Cm to Ar dst
1225.Op Ar options
1226.Ed
1227.Pp
1228The first part (proto from src to dst) is for backward
1229compatibility with earlier versions of
1230.Fx .
1231In modern
1232.Fx
1233any match pattern (including MAC headers, IP protocols,
1234addresses and ports) can be specified in the
1235.Ar options
1236section.
1237.Pp
1238Rule fields have the following meaning:
1239.Bl -tag -width indent
1240.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
1241.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
1242An IP protocol specified by number or name
1243(for a complete list see
1244.Pa /etc/protocols ) ,
1245or one of the following keywords:
1246.Bl -tag -width indent
1247.It Cm ip4 | ipv4
1248Matches IPv4 packets.
1249.It Cm ip6 | ipv6
1250Matches IPv6 packets.
1251.It Cm ip | all
1252Matches any packet.
1253.El
1254.Pp
1255The
1256.Cm ipv6
1257in
1258.Cm proto
1259option will be treated as inner protocol.
1260And, the
1261.Cm ipv4
1262is not available in
1263.Cm proto
1264option.
1265.Pp
1266The
1267.Cm { Ar protocol Cm or ... }
1268format (an
1269.Em or-block )
1270is provided for convenience only but its use is deprecated.
1271.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1272An address (or a list, see below)
1273optionally followed by
1274.Ar ports
1275specifiers.
1276.Pp
1277The second format
1278.Em ( or-block
1279with multiple addresses) is provided for convenience only and
1280its use is discouraged.
1281.It Ar addr : Oo Cm not Oc Bro
1282.Cm any | me | me6 |
1283.Cm table Ns Pq Ar name Ns Op , Ns Ar value
1284.Ar | addr-list | addr-set
1285.Brc
1286.Bl -tag -width indent
1287.It Cm any
1288matches any IP address.
1289.It Cm me
1290matches any IP address configured on an interface in the system.
1291.It Cm me6
1292matches any IPv6 address configured on an interface in the system.
1293The address list is evaluated at the time the packet is
1294analysed.
1295.It Cm table Ns Pq Ar name Ns Op , Ns Ar value
1296Matches any IPv4 or IPv6 address for which an entry exists in the lookup table
1297.Ar number .
1298If an optional 32-bit unsigned
1299.Ar value
1300is also specified, an entry will match only if it has this value.
1301See the
1302.Sx LOOKUP TABLES
1303section below for more information on lookup tables.
1304.El
1305.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1306.It Ar ip-addr :
1307A host or subnet address specified in one of the following ways:
1308.Bl -tag -width indent
1309.It Ar numeric-ip | hostname
1310Matches a single IPv4 address, specified as dotted-quad or a hostname.
1311Hostnames are resolved at the time the rule is added to the firewall list.
1312.It Ar addr Ns / Ns Ar masklen
1313Matches all addresses with base
1314.Ar addr
1315(specified as an IP address, a network number, or a hostname)
1316and mask width of
1317.Cm masklen
1318bits.
1319As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1320all IP numbers from 1.2.3.0 to 1.2.3.127 .
1321.It Ar addr Ns : Ns Ar mask
1322Matches all addresses with base
1323.Ar addr
1324(specified as an IP address, a network number, or a hostname)
1325and the mask of
1326.Ar mask ,
1327specified as a dotted quad.
1328As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
13291.*.3.*.
1330This form is advised only for non-contiguous
1331masks.
1332It is better to resort to the
1333.Ar addr Ns / Ns Ar masklen
1334format for contiguous masks, which is more compact and less
1335error-prone.
1336.El
1337.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1338.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1339Matches all addresses with base address
1340.Ar addr
1341(specified as an IP address, a network number, or a hostname)
1342and whose last byte is in the list between braces { } .
1343Note that there must be no spaces between braces and
1344numbers (spaces after commas are allowed).
1345Elements of the list can be specified as single entries
1346or ranges.
1347The
1348.Ar masklen
1349field is used to limit the size of the set of addresses,
1350and can have any value between 24 and 32.
1351If not specified,
1352it will be assumed as 24.
1353.br
1354This format is particularly useful to handle sparse address sets
1355within a single rule.
1356Because the matching occurs using a
1357bitmask, it takes constant time and dramatically reduces
1358the complexity of rulesets.
1359.br
1360As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1361or 1.2.3.0/24{128,35-55,89}
1362will match the following IP addresses:
1363.br
13641.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1365.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1366.It Ar ip6-addr :
1367A host or subnet specified one of the following ways:
1368.Bl -tag -width indent
1369.It Ar numeric-ip | hostname
1370Matches a single IPv6 address as allowed by
1371.Xr inet_pton 3
1372or a hostname.
1373Hostnames are resolved at the time the rule is added to the firewall
1374list.
1375.It Ar addr Ns / Ns Ar masklen
1376Matches all IPv6 addresses with base
1377.Ar addr
1378(specified as allowed by
1379.Xr inet_pton
1380or a hostname)
1381and mask width of
1382.Cm masklen
1383bits.
1384.It Ar addr Ns / Ns Ar mask
1385Matches all IPv6 addresses with base
1386.Ar addr
1387(specified as allowed by
1388.Xr inet_pton
1389or a hostname)
1390and the mask of
1391.Ar mask ,
1392specified as allowed by
1393.Xr inet_pton.
1394As an example, fe::640:0:0/ffff::ffff:ffff:0:0 will match
1395fe:*:*:*:0:640:*:*.
1396This form is advised only for non-contiguous
1397masks.
1398It is better to resort to the
1399.Ar addr Ns / Ns Ar masklen
1400format for contiguous masks, which is more compact and less
1401error-prone.
1402.El
1403.Pp
1404No support for sets of IPv6 addresses is provided because IPv6 addresses
1405are typically random past the initial prefix.
1406.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1407For protocols which support port numbers (such as SCTP, TCP and UDP), optional
1408.Cm ports
1409may be specified as one or more ports or port ranges, separated
1410by commas but no spaces, and an optional
1411.Cm not
1412operator.
1413The
1414.Ql \&-
1415notation specifies a range of ports (including boundaries).
1416.Pp
1417Service names (from
1418.Pa /etc/services )
1419may be used instead of numeric port values.
1420The length of the port list is limited to 30 ports or ranges,
1421though one can specify larger ranges by using an
1422.Em or-block
1423in the
1424.Cm options
1425section of the rule.
1426.Pp
1427A backslash
1428.Pq Ql \e
1429can be used to escape the dash
1430.Pq Ql -
1431character in a service name (from a shell, the backslash must be
1432typed twice to avoid the shell itself interpreting it as an escape
1433character).
1434.Pp
1435.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1436.Pp
1437Fragmented packets which have a non-zero offset (i.e., not the first
1438fragment) will never match a rule which has one or more port
1439specifications.
1440See the
1441.Cm frag
1442option for details on matching fragmented packets.
1443.El
1444.Ss RULE OPTIONS (MATCH PATTERNS)
1445Additional match patterns can be used within
1446rules.
1447Zero or more of these so-called
1448.Em options
1449can be present in a rule, optionally prefixed by the
1450.Cm not
1451operand, and possibly grouped into
1452.Em or-blocks .
1453.Pp
1454The following match patterns can be used (listed in alphabetical order):
1455.Bl -tag -width indent
1456.It Cm // this is a comment.
1457Inserts the specified text as a comment in the rule.
1458Everything following // is considered as a comment and stored in the rule.
1459You can have comment-only rules, which are listed as having a
1460.Cm count
1461action followed by the comment.
1462.It Cm bridged
1463Alias for
1464.Cm layer2 .
1465.It Cm diverted
1466Matches only packets generated by a divert socket.
1467.It Cm diverted-loopback
1468Matches only packets coming from a divert socket back into the IP stack
1469input for delivery.
1470.It Cm diverted-output
1471Matches only packets going from a divert socket back outward to the IP
1472stack output for delivery.
1473.It Cm dst-ip Ar ip-address
1474Matches IPv4 packets whose destination IP is one of the address(es)
1475specified as argument.
1476.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1477Matches IPv6 packets whose destination IP is one of the address(es)
1478specified as argument.
1479.It Cm dst-port Ar ports
1480Matches IP packets whose destination port is one of the port(s)
1481specified as argument.
1482.It Cm established
1483Matches TCP packets that have the RST or ACK bits set.
1484.It Cm ext6hdr Ar header
1485Matches IPv6 packets containing the extended header given by
1486.Ar header .
1487Supported headers are:
1488.Pp
1489Fragment,
1490.Pq Cm frag ,
1491Hop-to-hop options
1492.Pq Cm hopopt ,
1493any type of Routing Header
1494.Pq Cm route ,
1495Source routing Routing Header Type 0
1496.Pq Cm rthdr0 ,
1497Mobile IPv6 Routing Header Type 2
1498.Pq Cm rthdr2 ,
1499Destination options
1500.Pq Cm dstopt ,
1501IPSec authentication headers
1502.Pq Cm ah ,
1503and IPsec encapsulated security payload headers
1504.Pq Cm esp .
1505.It Cm fib Ar fibnum
1506Matches a packet that has been tagged to use
1507the given FIB (routing table) number.
1508.It Cm flow Ar table Ns Pq Ar name Ns Op , Ns Ar value
1509Search for the flow entry in lookup table
1510.Ar name .
1511If not found, the match fails.
1512Otherwise, the match succeeds and
1513.Cm tablearg
1514is set to the value extracted from the table.
1515.Pp
1516This option can be useful to quickly dispatch traffic based on
1517certain packet fields.
1518See the
1519.Sx LOOKUP TABLES
1520section below for more information on lookup tables.
1521.It Cm flow-id Ar labels
1522Matches IPv6 packets containing any of the flow labels given in
1523.Ar labels .
1524.Ar labels
1525is a comma separated list of numeric flow labels.
1526.It Cm frag
1527Matches packets that are fragments and not the first
1528fragment of an IP datagram.
1529Note that these packets will not have
1530the next protocol header (e.g.\& TCP, UDP) so options that look into
1531these headers cannot match.
1532.It Cm gid Ar group
1533Matches all TCP or UDP packets sent by or received for a
1534.Ar group .
1535A
1536.Ar group
1537may be specified by name or number.
1538.It Cm jail Ar prisonID
1539Matches all TCP or UDP packets sent by or received for the
1540jail whos prison ID is
1541.Ar prisonID .
1542.It Cm icmptypes Ar types
1543Matches ICMP packets whose ICMP type is in the list
1544.Ar types .
1545The list may be specified as any combination of
1546individual types (numeric) separated by commas.
1547.Em Ranges are not allowed .
1548The supported ICMP types are:
1549.Pp
1550echo reply
1551.Pq Cm 0 ,
1552destination unreachable
1553.Pq Cm 3 ,
1554source quench
1555.Pq Cm 4 ,
1556redirect
1557.Pq Cm 5 ,
1558echo request
1559.Pq Cm 8 ,
1560router advertisement
1561.Pq Cm 9 ,
1562router solicitation
1563.Pq Cm 10 ,
1564time-to-live exceeded
1565.Pq Cm 11 ,
1566IP header bad
1567.Pq Cm 12 ,
1568timestamp request
1569.Pq Cm 13 ,
1570timestamp reply
1571.Pq Cm 14 ,
1572information request
1573.Pq Cm 15 ,
1574information reply
1575.Pq Cm 16 ,
1576address mask request
1577.Pq Cm 17
1578and address mask reply
1579.Pq Cm 18 .
1580.It Cm icmp6types Ar types
1581Matches ICMP6 packets whose ICMP6 type is in the list of
1582.Ar types .
1583The list may be specified as any combination of
1584individual types (numeric) separated by commas.
1585.Em Ranges are not allowed .
1586.It Cm in | out
1587Matches incoming or outgoing packets, respectively.
1588.Cm in
1589and
1590.Cm out
1591are mutually exclusive (in fact,
1592.Cm out
1593is implemented as
1594.Cm not in Ns No ).
1595.It Cm ipid Ar id-list
1596Matches IPv4 packets whose
1597.Cm ip_id
1598field has value included in
1599.Ar id-list ,
1600which is either a single value or a list of values or ranges
1601specified in the same way as
1602.Ar ports .
1603.It Cm iplen Ar len-list
1604Matches IP packets whose total length, including header and data, is
1605in the set
1606.Ar len-list ,
1607which is either a single value or a list of values or ranges
1608specified in the same way as
1609.Ar ports .
1610.It Cm ipoptions Ar spec
1611Matches packets whose IPv4 header contains the comma separated list of
1612options specified in
1613.Ar spec .
1614The supported IP options are:
1615.Pp
1616.Cm ssrr
1617(strict source route),
1618.Cm lsrr
1619(loose source route),
1620.Cm rr
1621(record packet route) and
1622.Cm ts
1623(timestamp).
1624The absence of a particular option may be denoted
1625with a
1626.Ql \&! .
1627.It Cm ipprecedence Ar precedence
1628Matches IPv4 packets whose precedence field is equal to
1629.Ar precedence .
1630.It Cm ipsec
1631Matches packets that have IPSEC history associated with them
1632(i.e., the packet comes encapsulated in IPSEC, the kernel
1633has IPSEC support, and can correctly decapsulate it).
1634.Pp
1635Note that specifying
1636.Cm ipsec
1637is different from specifying
1638.Cm proto Ar ipsec
1639as the latter will only look at the specific IP protocol field,
1640irrespective of IPSEC kernel support and the validity of the IPSEC data.
1641.Pp
1642Further note that this flag is silently ignored in kernels without
1643IPSEC support.
1644It does not affect rule processing when given and the
1645rules are handled as if with no
1646.Cm ipsec
1647flag.
1648.It Cm iptos Ar spec
1649Matches IPv4 packets whose
1650.Cm tos
1651field contains the comma separated list of
1652service types specified in
1653.Ar spec .
1654The supported IP types of service are:
1655.Pp
1656.Cm lowdelay
1657.Pq Dv IPTOS_LOWDELAY ,
1658.Cm throughput
1659.Pq Dv IPTOS_THROUGHPUT ,
1660.Cm reliability
1661.Pq Dv IPTOS_RELIABILITY ,
1662.Cm mincost
1663.Pq Dv IPTOS_MINCOST ,
1664.Cm congestion
1665.Pq Dv IPTOS_ECN_CE .
1666The absence of a particular type may be denoted
1667with a
1668.Ql \&! .
1669.It Cm dscp spec Ns Op , Ns Ar spec
1670Matches IPv4/IPv6 packets whose
1671.Cm DS
1672field value is contained in
1673.Ar spec
1674mask.
1675Multiple values can be specified via
1676the comma separated list.
1677Value can be one of keywords used in
1678.Cm setdscp
1679action or exact number.
1680.It Cm ipttl Ar ttl-list
1681Matches IPv4 packets whose time to live is included in
1682.Ar ttl-list ,
1683which is either a single value or a list of values or ranges
1684specified in the same way as
1685.Ar ports .
1686.It Cm ipversion Ar ver
1687Matches IP packets whose IP version field is
1688.Ar ver .
1689.It Cm keep-state Op Ar :flowname
1690Upon a match, the firewall will create a dynamic rule, whose
1691default behaviour is to match bidirectional traffic between
1692source and destination IP/port using the same protocol.
1693The rule has a limited lifetime (controlled by a set of
1694.Xr sysctl 8
1695variables), and the lifetime is refreshed every time a matching
1696packet is found.
1697The
1698.Ar :flowname
1699is used to assign additional to addresses, ports and protocol parameter
1700to dynamic rule. It can be used for more accurate matching by
1701.Cm check-state
1702rule.
1703The
1704.Cm :default
1705keyword is special name used for compatibility with old rulesets.
1706.It Cm layer2
1707Matches only layer2 packets, i.e., those passed to
1708.Nm
1709from ether_demux() and ether_output_frame().
1710.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N Op Ar :flowname
1711The firewall will only allow
1712.Ar N
1713connections with the same
1714set of parameters as specified in the rule.
1715One or more
1716of source and destination addresses and ports can be
1717specified.
1718.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar name
1719Search an entry in lookup table
1720.Ar name
1721that matches the field specified as argument.
1722If not found, the match fails.
1723Otherwise, the match succeeds and
1724.Cm tablearg
1725is set to the value extracted from the table.
1726.Pp
1727This option can be useful to quickly dispatch traffic based on
1728certain packet fields.
1729See the
1730.Sx LOOKUP TABLES
1731section below for more information on lookup tables.
1732.It Cm { MAC | mac } Ar dst-mac src-mac
1733Match packets with a given
1734.Ar dst-mac
1735and
1736.Ar src-mac
1737addresses, specified as the
1738.Cm any
1739keyword (matching any MAC address), or six groups of hex digits
1740separated by colons,
1741and optionally followed by a mask indicating the significant bits.
1742The mask may be specified using either of the following methods:
1743.Bl -enum -width indent
1744.It
1745A slash
1746.Pq /
1747followed by the number of significant bits.
1748For example, an address with 33 significant bits could be specified as:
1749.Pp
1750.Dl "MAC 10:20:30:40:50:60/33 any"
1751.It
1752An ampersand
1753.Pq &
1754followed by a bitmask specified as six groups of hex digits separated
1755by colons.
1756For example, an address in which the last 16 bits are significant could
1757be specified as:
1758.Pp
1759.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1760.Pp
1761Note that the ampersand character has a special meaning in many shells
1762and should generally be escaped.
1763.El
1764Note that the order of MAC addresses (destination first,
1765source second) is
1766the same as on the wire, but the opposite of the one used for
1767IP addresses.
1768.It Cm mac-type Ar mac-type
1769Matches packets whose Ethernet Type field
1770corresponds to one of those specified as argument.
1771.Ar mac-type
1772is specified in the same way as
1773.Cm port numbers
1774(i.e., one or more comma-separated single values or ranges).
1775You can use symbolic names for known values such as
1776.Em vlan , ipv4, ipv6 .
1777Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1778and they are always printed as hexadecimal (unless the
1779.Cm -N
1780option is used, in which case symbolic resolution will be attempted).
1781.It Cm proto Ar protocol
1782Matches packets with the corresponding IP protocol.
1783.It Cm recv | xmit | via Brq Ar ifX | Ar if Ns Cm * | Ar table Ns Po Ar name Ns Oo , Ns Ar value Oc Pc | Ar ipno | Ar any
1784Matches packets received, transmitted or going through,
1785respectively, the interface specified by exact name
1786.Po Ar ifX Pc ,
1787by device name
1788.Po Ar if* Pc ,
1789by IP address, or through some interface.
1790Table
1791.Ar name
1792may be used to match interface by its kernel ifindex.
1793See the
1794.Sx LOOKUP TABLES
1795section below for more information on lookup tables.
1796.Pp
1797The
1798.Cm via
1799keyword causes the interface to always be checked.
1800If
1801.Cm recv
1802or
1803.Cm xmit
1804is used instead of
1805.Cm via ,
1806then only the receive or transmit interface (respectively)
1807is checked.
1808By specifying both, it is possible to match packets based on
1809both receive and transmit interface, e.g.:
1810.Pp
1811.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1812.Pp
1813The
1814.Cm recv
1815interface can be tested on either incoming or outgoing packets,
1816while the
1817.Cm xmit
1818interface can only be tested on outgoing packets.
1819So
1820.Cm out
1821is required (and
1822.Cm in
1823is invalid) whenever
1824.Cm xmit
1825is used.
1826.Pp
1827A packet might not have a receive or transmit interface: packets
1828originating from the local host have no receive interface,
1829while packets destined for the local host have no transmit
1830interface.
1831.It Cm setup
1832Matches TCP packets that have the SYN bit set but no ACK bit.
1833This is the short form of
1834.Dq Li tcpflags\ syn,!ack .
1835.It Cm sockarg
1836Matches packets that are associated to a local socket and
1837for which the SO_USER_COOKIE socket option has been set
1838to a non-zero value.
1839As a side effect, the value of the
1840option is made available as
1841.Cm tablearg
1842value, which in turn can be used as
1843.Cm skipto
1844or
1845.Cm pipe
1846number.
1847.It Cm src-ip Ar ip-address
1848Matches IPv4 packets whose source IP is one of the address(es)
1849specified as an argument.
1850.It Cm src-ip6 Ar ip6-address
1851Matches IPv6 packets whose source IP is one of the address(es)
1852specified as an argument.
1853.It Cm src-port Ar ports
1854Matches IP packets whose source port is one of the port(s)
1855specified as argument.
1856.It Cm tagged Ar tag-list
1857Matches packets whose tags are included in
1858.Ar tag-list ,
1859which is either a single value or a list of values or ranges
1860specified in the same way as
1861.Ar ports .
1862Tags can be applied to the packet using
1863.Cm tag
1864rule action parameter (see it's description for details on tags).
1865.It Cm tcpack Ar ack
1866TCP packets only.
1867Match if the TCP header acknowledgment number field is set to
1868.Ar ack .
1869.It Cm tcpdatalen Ar tcpdatalen-list
1870Matches TCP packets whose length of TCP data is
1871.Ar tcpdatalen-list ,
1872which is either a single value or a list of values or ranges
1873specified in the same way as
1874.Ar ports .
1875.It Cm tcpflags Ar spec
1876TCP packets only.
1877Match if the TCP header contains the comma separated list of
1878flags specified in
1879.Ar spec .
1880The supported TCP flags are:
1881.Pp
1882.Cm fin ,
1883.Cm syn ,
1884.Cm rst ,
1885.Cm psh ,
1886.Cm ack
1887and
1888.Cm urg .
1889The absence of a particular flag may be denoted
1890with a
1891.Ql \&! .
1892A rule which contains a
1893.Cm tcpflags
1894specification can never match a fragmented packet which has
1895a non-zero offset.
1896See the
1897.Cm frag
1898option for details on matching fragmented packets.
1899.It Cm tcpseq Ar seq
1900TCP packets only.
1901Match if the TCP header sequence number field is set to
1902.Ar seq .
1903.It Cm tcpwin Ar tcpwin-list
1904Matches TCP packets whose  header window field is set to
1905.Ar tcpwin-list ,
1906which is either a single value or a list of values or ranges
1907specified in the same way as
1908.Ar ports .
1909.It Cm tcpoptions Ar spec
1910TCP packets only.
1911Match if the TCP header contains the comma separated list of
1912options specified in
1913.Ar spec .
1914The supported TCP options are:
1915.Pp
1916.Cm mss
1917(maximum segment size),
1918.Cm window
1919(tcp window advertisement),
1920.Cm sack
1921(selective ack),
1922.Cm ts
1923(rfc1323 timestamp) and
1924.Cm cc
1925(rfc1644 t/tcp connection count).
1926The absence of a particular option may be denoted
1927with a
1928.Ql \&! .
1929.It Cm uid Ar user
1930Match all TCP or UDP packets sent by or received for a
1931.Ar user .
1932A
1933.Ar user
1934may be matched by name or identification number.
1935.It Cm verrevpath
1936For incoming packets,
1937a routing table lookup is done on the packet's source address.
1938If the interface on which the packet entered the system matches the
1939outgoing interface for the route,
1940the packet matches.
1941If the interfaces do not match up,
1942the packet does not match.
1943All outgoing packets or packets with no incoming interface match.
1944.Pp
1945The name and functionality of the option is intentionally similar to
1946the Cisco IOS command:
1947.Pp
1948.Dl ip verify unicast reverse-path
1949.Pp
1950This option can be used to make anti-spoofing rules to reject all
1951packets with source addresses not from this interface.
1952See also the option
1953.Cm antispoof .
1954.It Cm versrcreach
1955For incoming packets,
1956a routing table lookup is done on the packet's source address.
1957If a route to the source address exists, but not the default route
1958or a blackhole/reject route, the packet matches.
1959Otherwise, the packet does not match.
1960All outgoing packets match.
1961.Pp
1962The name and functionality of the option is intentionally similar to
1963the Cisco IOS command:
1964.Pp
1965.Dl ip verify unicast source reachable-via any
1966.Pp
1967This option can be used to make anti-spoofing rules to reject all
1968packets whose source address is unreachable.
1969.It Cm antispoof
1970For incoming packets, the packet's source address is checked if it
1971belongs to a directly connected network.
1972If the network is directly connected, then the interface the packet
1973came on in is compared to the interface the network is connected to.
1974When incoming interface and directly connected interface are not the
1975same, the packet does not match.
1976Otherwise, the packet does match.
1977All outgoing packets match.
1978.Pp
1979This option can be used to make anti-spoofing rules to reject all
1980packets that pretend to be from a directly connected network but do
1981not come in through that interface.
1982This option is similar to but more restricted than
1983.Cm verrevpath
1984because it engages only on packets with source addresses of directly
1985connected networks instead of all source addresses.
1986.El
1987.Sh LOOKUP TABLES
1988Lookup tables are useful to handle large sparse sets of
1989addresses or other search keys (e.g., ports, jail IDs, interface names).
1990In the rest of this section we will use the term ``key''.
1991Table name needs to match the following spec:
1992.Ar table-name .
1993Tables with the same name can be created in different
1994.Ar sets .
1995However, rule links to the tables in
1996.Ar set 0
1997by default.
1998This behavior can be controlled by
1999.Va net.inet.ip.fw.tables_sets
2000variable.
2001See the
2002.Sx SETS OF RULES
2003section for more information.
2004There may be up to 65535 different lookup tables.
2005.Pp
2006The following table types are supported:
2007.Bl -tag -width indent
2008.It Ar table-type : Ar addr | iface | number | flow
2009.It Ar table-key : Ar addr Ns Oo / Ns Ar masklen Oc | iface-name | number | flow-spec
2010.It Ar flow-spec : Ar flow-field Ns Op , Ns Ar flow-spec
2011.It Ar flow-field : src-ip | proto | src-port | dst-ip | dst-port
2012.It Cm addr
2013matches IPv4 or IPv6 address.
2014Each entry is represented by an
2015.Ar addr Ns Op / Ns Ar masklen
2016and will match all addresses with base
2017.Ar addr
2018(specified as an IPv4/IPv6 address, or a hostname) and mask width of
2019.Ar masklen
2020bits.
2021If
2022.Ar masklen
2023is not specified, it defaults to 32 for IPv4 and 128 for IPv6.
2024When looking up an IP address in a table, the most specific
2025entry will match.
2026.It Cm iface
2027matches interface names.
2028Each entry is represented by string treated as interface name.
2029Wildcards are not supported.
2030.It Cm number
2031maches protocol ports, uids/gids or jail IDs.
2032Each entry is represented by 32-bit unsigned integer.
2033Ranges are not supported.
2034.It Cm flow
2035Matches packet fields specified by
2036.Ar flow
2037type suboptions with table entries.
2038.El
2039.Pp
2040Tables require explicit creation via
2041.Cm create
2042before use.
2043.Pp
2044The following creation options are supported:
2045.Bl -tag -width indent
2046.It Ar create-options : Ar create-option | create-options
2047.It Ar create-option : Cm type Ar table-type | Cm valtype Ar value-mask | Cm algo Ar algo-desc |
2048.Cm limit Ar number | Cm locked
2049.It Cm type
2050Table key type.
2051.It Cm valtype
2052Table value mask.
2053.It Cm algo
2054Table algorithm to use (see below).
2055.It Cm limit
2056Maximum number of items that may be inserted into table.
2057.It Cm locked
2058Restrict any table modifications.
2059.El
2060.Pp
2061Some of these options may be modified later via
2062.Cm modify
2063keyword.
2064The following options can be changed:
2065.Bl -tag -width indent
2066.It Ar modify-options : Ar modify-option | modify-options
2067.It Ar modify-option : Cm limit Ar number
2068.It Cm limit
2069Alter maximum number of items that may be inserted into table.
2070.El
2071.Pp
2072Additionally, table can be locked or unlocked using
2073.Cm lock
2074or
2075.Cm unlock
2076commands.
2077.Pp
2078Tables of the same
2079.Ar type
2080can be swapped with each other using
2081.Cm swap Ar name
2082command.
2083Swap may fail if tables limits are set and data exchange
2084would result in limits hit.
2085Operation is performed atomically.
2086.Pp
2087One or more entries can be added to a table at once using
2088.Cm add
2089command.
2090Addition of all items are performed atomically.
2091By default, error in addition of one entry does not influence
2092addition of other entries. However, non-zero error code is returned
2093in that case.
2094Special
2095.Cm atomic
2096keyword may be specified before
2097.Cm add
2098to indicate all-or-none add request.
2099.Pp
2100One or more entries can be removed from a table at once using
2101.Cm delete
2102command.
2103By default, error in removal of one entry does not influence
2104removing of other entries. However, non-zero error code is returned
2105in that case.
2106.Pp
2107It may be possible to check what entry will be found on particular
2108.Ar table-key
2109using
2110.Cm lookup
2111.Ar table-key
2112command.
2113This functionality is optional and may be unsupported in some algorithms.
2114.Pp
2115The following operations can be performed on
2116.Ar one
2117or
2118.Cm all
2119tables:
2120.Bl -tag -width indent
2121.It Cm list
2122List all entries.
2123.It Cm flush
2124Removes all entries.
2125.It Cm info
2126Shows generic table information.
2127.It Cm detail
2128Shows generic table information and algo-specific data.
2129.El
2130.Pp
2131The following lookup algorithms are supported:
2132.Bl -tag -width indent
2133.It Ar algo-desc : algo-name | "algo-name algo-data"
2134.It Ar algo-name: Ar addr:radix | addr:hash | iface:array | number:array | flow:hash
2135.It Cm addr:radix
2136Separate Radix trees for IPv4 and IPv6, the same way as the routing table (see
2137.Xr route 4 ) .
2138Default choice for
2139.Ar addr
2140type.
2141.It Cm addr:hash
2142Separate auto-growing hashes for IPv4 and IPv6.
2143Accepts entries with the same mask length specified initially via
2144.Cm "addr:hash masks=/v4,/v6"
2145algorithm creation options.
2146Assume /32 and /128 masks by default.
2147Search removes host bits (according to mask) from supplied address and checks
2148resulting key in appropriate hash.
2149Mostly optimized for /64 and byte-ranged IPv6 masks.
2150.It Cm iface:array
2151Array storing sorted indexes for entries which are presented in the system.
2152Optimized for very fast lookup.
2153.It Cm number:array
2154Array storing sorted u32 numbers.
2155.It Cm flow:hash
2156Auto-growing hash storing flow entries.
2157Search calculates hash on required packet fields and searches for matching
2158entries in selected bucket.
2159.El
2160.Pp
2161The
2162.Cm tablearg
2163feature provides the ability to use a value, looked up in the table, as
2164the argument for a rule action, action parameter or rule option.
2165This can significantly reduce number of rules in some configurations.
2166If two tables are used in a rule, the result of the second (destination)
2167is used.
2168.Pp
2169Each record may hold one or more values according to
2170.Ar value-mask .
2171This mask is set on table creation via
2172.Cm valtype
2173option.
2174The following value types are supported:
2175.Bl -tag -width indent
2176.It Ar value-mask : Ar value-type Ns Op , Ns Ar value-mask
2177.It Ar value-type : Ar skipto | pipe | fib | nat | dscp | tag | divert |
2178.Ar netgraph | limit | ipv4
2179.It Cm skipto
2180rule number to jump to.
2181.It Cm pipe
2182Pipe number to use.
2183.It Cm fib
2184fib number to match/set.
2185.It Cm nat
2186nat number to jump to.
2187.It Cm dscp
2188dscp value to match/set.
2189.It Cm tag
2190tag number to match/set.
2191.It Cm divert
2192port number to divert traffic to.
2193.It Cm netgraph
2194hook number to move packet to.
2195.It Cm limit
2196maximum number of connections.
2197.It Cm ipv4
2198IPv4 nexthop to fwd packets to.
2199.It Cm ipv6
2200IPv6 nexthop to fwd packets to.
2201.El
2202.Pp
2203The
2204.Cm tablearg
2205argument can be used with the following actions:
2206.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto, setfib,
2207action parameters:
2208.Cm tag, untag,
2209rule options:
2210.Cm limit, tagged.
2211.Pp
2212When used with the
2213.Cm skipto
2214action, the user should be aware that the code will walk the ruleset
2215up to a rule equal to, or past, the given number.
2216.Pp
2217See the
2218.Sx EXAMPLES
2219Section for example usage of tables and the tablearg keyword.
2220.Sh SETS OF RULES
2221Each rule or table belongs to one of 32 different
2222.Em sets
2223, numbered 0 to 31.
2224Set 31 is reserved for the default rule.
2225.Pp
2226By default, rules or tables are put in set 0, unless you use the
2227.Cm set N
2228attribute when adding a new rule or table.
2229Sets can be individually and atomically enabled or disabled,
2230so this mechanism permits an easy way to store multiple configurations
2231of the firewall and quickly (and atomically) switch between them.
2232.Pp
2233By default, tables from set 0 are referenced when adding rule with
2234table opcodes regardless of rule set.
2235This behavior can be changed by setting
2236.Va net.inet.ip.fw.tables_set
2237variable to 1.
2238Rule's set will then be used for table references.
2239.Pp
2240The command to enable/disable sets is
2241.Bd -ragged -offset indent
2242.Nm
2243.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
2244.Ed
2245.Pp
2246where multiple
2247.Cm enable
2248or
2249.Cm disable
2250sections can be specified.
2251Command execution is atomic on all the sets specified in the command.
2252By default, all sets are enabled.
2253.Pp
2254When you disable a set, its rules behave as if they do not exist
2255in the firewall configuration, with only one exception:
2256.Bd -ragged -offset indent
2257dynamic rules created from a rule before it had been disabled
2258will still be active until they expire.
2259In order to delete
2260dynamic rules you have to explicitly delete the parent rule
2261which generated them.
2262.Ed
2263.Pp
2264The set number of rules can be changed with the command
2265.Bd -ragged -offset indent
2266.Nm
2267.Cm set move
2268.Brq Cm rule Ar rule-number | old-set
2269.Cm to Ar new-set
2270.Ed
2271.Pp
2272Also, you can atomically swap two rulesets with the command
2273.Bd -ragged -offset indent
2274.Nm
2275.Cm set swap Ar first-set second-set
2276.Ed
2277.Pp
2278See the
2279.Sx EXAMPLES
2280Section on some possible uses of sets of rules.
2281.Sh STATEFUL FIREWALL
2282Stateful operation is a way for the firewall to dynamically
2283create rules for specific flows when packets that
2284match a given pattern are detected.
2285Support for stateful
2286operation comes through the
2287.Cm check-state , keep-state
2288and
2289.Cm limit
2290options of
2291.Nm rules .
2292.Pp
2293Dynamic rules are created when a packet matches a
2294.Cm keep-state
2295or
2296.Cm limit
2297rule, causing the creation of a
2298.Em dynamic
2299rule which will match all and only packets with
2300a given
2301.Em protocol
2302between a
2303.Em src-ip/src-port dst-ip/dst-port
2304pair of addresses
2305.Em ( src
2306and
2307.Em dst
2308are used here only to denote the initial match addresses, but they
2309are completely equivalent afterwards).
2310Rules created by
2311.Cm keep-state
2312option also have a
2313.Ar :flowname
2314taken from it.
2315This name is used in matching together with addresses, ports and protocol.
2316Dynamic rules will be checked at the first
2317.Cm check-state, keep-state
2318or
2319.Cm limit
2320occurrence, and the action performed upon a match will be the same
2321as in the parent rule.
2322.Pp
2323Note that no additional attributes other than protocol and IP addresses
2324and ports and :flowname are checked on dynamic rules.
2325.Pp
2326The typical use of dynamic rules is to keep a closed firewall configuration,
2327but let the first TCP SYN packet from the inside network install a
2328dynamic rule for the flow so that packets belonging to that session
2329will be allowed through the firewall:
2330.Pp
2331.Dl "ipfw add check-state :OUTBOUND"
2332.Dl "ipfw add allow tcp from my-subnet to any setup keep-state :OUTBOUND"
2333.Dl "ipfw add deny tcp from any to any"
2334.Pp
2335A similar approach can be used for UDP, where an UDP packet coming
2336from the inside will install a dynamic rule to let the response through
2337the firewall:
2338.Pp
2339.Dl "ipfw add check-state :OUTBOUND"
2340.Dl "ipfw add allow udp from my-subnet to any keep-state :OUTBOUND"
2341.Dl "ipfw add deny udp from any to any"
2342.Pp
2343Dynamic rules expire after some time, which depends on the status
2344of the flow and the setting of some
2345.Cm sysctl
2346variables.
2347See Section
2348.Sx SYSCTL VARIABLES
2349for more details.
2350For TCP sessions, dynamic rules can be instructed to periodically
2351send keepalive packets to refresh the state of the rule when it is
2352about to expire.
2353.Pp
2354See Section
2355.Sx EXAMPLES
2356for more examples on how to use dynamic rules.
2357.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
2358.Nm
2359is also the user interface for the
2360.Nm dummynet
2361traffic shaper, packet scheduler and network emulator, a subsystem that
2362can artificially queue, delay or drop packets
2363emulating the behaviour of certain network links
2364or queueing systems.
2365.Pp
2366.Nm dummynet
2367operates by first using the firewall to select packets
2368using any match pattern that can be used in
2369.Nm
2370rules.
2371Matching packets are then passed to either of two
2372different objects, which implement the traffic regulation:
2373.Bl -hang -offset XXXX
2374.It Em pipe
2375A
2376.Em pipe
2377emulates a
2378.Em link
2379with given bandwidth and propagation delay,
2380driven by a FIFO scheduler and a single queue with programmable
2381queue size and packet loss rate.
2382Packets are appended to the queue as they come out from
2383.Nm ipfw ,
2384and then transferred in FIFO order to the link at the desired rate.
2385.It Em queue
2386A
2387.Em queue
2388is an abstraction used to implement packet scheduling
2389using one of several packet scheduling algorithms.
2390Packets sent to a
2391.Em queue
2392are first grouped into flows according to a mask on the 5-tuple.
2393Flows are then passed to the scheduler associated to the
2394.Em queue ,
2395and each flow uses scheduling parameters (weight and others)
2396as configured in the
2397.Em queue
2398itself.
2399A scheduler in turn is connected to an emulated link,
2400and arbitrates the link's bandwidth among backlogged flows according to
2401weights and to the features of the scheduling algorithm in use.
2402.El
2403.Pp
2404In practice,
2405.Em pipes
2406can be used to set hard limits to the bandwidth that a flow can use, whereas
2407.Em queues
2408can be used to determine how different flows share the available bandwidth.
2409.Pp
2410A graphical representation of the binding of queues,
2411flows, schedulers and links is below.
2412.Bd -literal -offset indent
2413                 (flow_mask|sched_mask)  sched_mask
2414         +---------+   weight Wx  +-------------+
2415         |         |->-[flow]-->--|             |-+
2416    -->--| QUEUE x |   ...        |             | |
2417         |         |->-[flow]-->--| SCHEDuler N | |
2418         +---------+              |             | |
2419             ...                  |             +--[LINK N]-->--
2420         +---------+   weight Wy  |             | +--[LINK N]-->--
2421         |         |->-[flow]-->--|             | |
2422    -->--| QUEUE y |   ...        |             | |
2423         |         |->-[flow]-->--|             | |
2424         +---------+              +-------------+ |
2425                                    +-------------+
2426.Ed
2427It is important to understand the role of the SCHED_MASK
2428and FLOW_MASK, which are configured through the commands
2429.Dl "ipfw sched N config mask SCHED_MASK ..."
2430and
2431.Dl "ipfw queue X config mask FLOW_MASK ..." .
2432.Pp
2433The SCHED_MASK is used to assign flows to one or more
2434scheduler instances, one for each
2435value of the packet's 5-tuple after applying SCHED_MASK.
2436As an example, using ``src-ip 0xffffff00'' creates one instance
2437for each /24 destination subnet.
2438.Pp
2439The FLOW_MASK, together with the SCHED_MASK, is used to split
2440packets into flows.
2441As an example, using
2442``src-ip 0x000000ff''
2443together with the previous SCHED_MASK makes a flow for
2444each individual source address.
2445In turn, flows for each /24
2446subnet will be sent to the same scheduler instance.
2447.Pp
2448The above diagram holds even for the
2449.Em pipe
2450case, with the only restriction that a
2451.Em pipe
2452only supports a SCHED_MASK, and forces the use of a FIFO
2453scheduler (these are for backward compatibility reasons;
2454in fact, internally, a
2455.Nm dummynet's
2456pipe is implemented exactly as above).
2457.Pp
2458There are two modes of
2459.Nm dummynet
2460operation:
2461.Dq normal
2462and
2463.Dq fast .
2464The
2465.Dq normal
2466mode tries to emulate a real link: the
2467.Nm dummynet
2468scheduler ensures that the packet will not leave the pipe faster than it
2469would on the real link with a given bandwidth.
2470The
2471.Dq fast
2472mode allows certain packets to bypass the
2473.Nm dummynet
2474scheduler (if packet flow does not exceed pipe's bandwidth).
2475This is the reason why the
2476.Dq fast
2477mode requires less CPU cycles per packet (on average) and packet latency
2478can be significantly lower in comparison to a real link with the same
2479bandwidth.
2480The default mode is
2481.Dq normal .
2482The
2483.Dq fast
2484mode can be enabled by setting the
2485.Va net.inet.ip.dummynet.io_fast
2486.Xr sysctl 8
2487variable to a non-zero value.
2488.Pp
2489.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
2490The
2491.Em pipe ,
2492.Em queue
2493and
2494.Em scheduler
2495configuration commands are the following:
2496.Bd -ragged -offset indent
2497.Cm pipe Ar number Cm config Ar pipe-configuration
2498.Pp
2499.Cm queue Ar number Cm config Ar queue-configuration
2500.Pp
2501.Cm sched Ar number Cm config Ar sched-configuration
2502.Ed
2503.Pp
2504The following parameters can be configured for a pipe:
2505.Pp
2506.Bl -tag -width indent -compact
2507.It Cm bw Ar bandwidth | device
2508Bandwidth, measured in
2509.Sm off
2510.Op Cm K | M | G
2511.Brq Cm bit/s | Byte/s .
2512.Sm on
2513.Pp
2514A value of 0 (default) means unlimited bandwidth.
2515The unit must immediately follow the number, as in
2516.Pp
2517.Dl "ipfw pipe 1 config bw 300Kbit/s"
2518.Pp
2519If a device name is specified instead of a numeric value, as in
2520.Pp
2521.Dl "ipfw pipe 1 config bw tun0"
2522.Pp
2523then the transmit clock is supplied by the specified device.
2524At the moment only the
2525.Xr tun 4
2526device supports this
2527functionality, for use in conjunction with
2528.Xr ppp 8 .
2529.Pp
2530.It Cm delay Ar ms-delay
2531Propagation delay, measured in milliseconds.
2532The value is rounded to the next multiple of the clock tick
2533(typically 10ms, but it is a good practice to run kernels
2534with
2535.Dq "options HZ=1000"
2536to reduce
2537the granularity to 1ms or less).
2538The default value is 0, meaning no delay.
2539.Pp
2540.It Cm burst Ar size
2541If the data to be sent exceeds the pipe's bandwidth limit
2542(and the pipe was previously idle), up to
2543.Ar size
2544bytes of data are allowed to bypass the
2545.Nm dummynet
2546scheduler, and will be sent as fast as the physical link allows.
2547Any additional data will be transmitted at the rate specified
2548by the
2549.Nm pipe
2550bandwidth.
2551The burst size depends on how long the pipe has been idle;
2552the effective burst size is calculated as follows:
2553MAX(
2554.Ar size
2555,
2556.Nm bw
2557* pipe_idle_time).
2558.Pp
2559.It Cm profile Ar filename
2560A file specifying the additional overhead incurred in the transmission
2561of a packet on the link.
2562.Pp
2563Some link types introduce extra delays in the transmission
2564of a packet, e.g., because of MAC level framing, contention on
2565the use of the channel, MAC level retransmissions and so on.
2566From our point of view, the channel is effectively unavailable
2567for this extra time, which is constant or variable depending
2568on the link type.
2569Additionally, packets may be dropped after this
2570time (e.g., on a wireless link after too many retransmissions).
2571We can model the additional delay with an empirical curve
2572that represents its distribution.
2573.Bd -literal -offset indent
2574      cumulative probability
2575      1.0 ^
2576          |
2577      L   +-- loss-level          x
2578          |                 ******
2579          |                *
2580          |           *****
2581          |          *
2582          |        **
2583          |       *
2584          +-------*------------------->
2585                      delay
2586.Ed
2587The empirical curve may have both vertical and horizontal lines.
2588Vertical lines represent constant delay for a range of
2589probabilities.
2590Horizontal lines correspond to a discontinuity in the delay
2591distribution: the pipe will use the largest delay for a
2592given probability.
2593.Pp
2594The file format is the following, with whitespace acting as
2595a separator and '#' indicating the beginning a comment:
2596.Bl -tag -width indent
2597.It Cm name Ar identifier
2598optional name (listed by "ipfw pipe show")
2599to identify the delay distribution;
2600.It Cm bw Ar value
2601the bandwidth used for the pipe.
2602If not specified here, it must be present
2603explicitly as a configuration parameter for the pipe;
2604.It Cm loss-level Ar L
2605the probability above which packets are lost.
2606(0.0 <= L <= 1.0, default 1.0 i.e., no loss);
2607.It Cm samples Ar N
2608the number of samples used in the internal
2609representation of the curve (2..1024; default 100);
2610.It Cm "delay prob" | "prob delay"
2611One of these two lines is mandatory and defines
2612the format of the following lines with data points.
2613.It Ar XXX Ar YYY
26142 or more lines representing points in the curve,
2615with either delay or probability first, according
2616to the chosen format.
2617The unit for delay is milliseconds.
2618Data points do not need to be sorted.
2619Also, the number of actual lines can be different
2620from the value of the "samples" parameter:
2621.Nm
2622utility will sort and interpolate
2623the curve as needed.
2624.El
2625.Pp
2626Example of a profile file:
2627.Bd -literal -offset indent
2628name    bla_bla_bla
2629samples 100
2630loss-level    0.86
2631prob    delay
26320       200	# minimum overhead is 200ms
26330.5     200
26340.5     300
26350.8     1000
26360.9     1300
26371       1300
2638#configuration file end
2639.Ed
2640.El
2641.Pp
2642The following parameters can be configured for a queue:
2643.Pp
2644.Bl -tag -width indent -compact
2645.It Cm pipe Ar pipe_nr
2646Connects a queue to the specified pipe.
2647Multiple queues (with the same or different weights) can be connected to
2648the same pipe, which specifies the aggregate rate for the set of queues.
2649.Pp
2650.It Cm weight Ar weight
2651Specifies the weight to be used for flows matching this queue.
2652The weight must be in the range 1..100, and defaults to 1.
2653.El
2654.Pp
2655The following case-insensitive parameters can be configured for a
2656scheduler:
2657.Pp
2658.Bl -tag -width indent -compact
2659.It Cm type Ar {fifo | wf2q+ | rr | qfq}
2660specifies the scheduling algorithm to use.
2661.Bl -tag -width indent -compact
2662.It Cm fifo
2663is just a FIFO scheduler (which means that all packets
2664are stored in the same queue as they arrive to the scheduler).
2665FIFO has O(1) per-packet time complexity, with very low
2666constants (estimate 60-80ns on a 2GHz desktop machine)
2667but gives no service guarantees.
2668.It Cm wf2q+
2669implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2670algorithm which permits flows to share bandwidth according to
2671their weights.
2672Note that weights are not priorities; even a flow
2673with a minuscule weight will never starve.
2674WF2Q+ has O(log N) per-packet processing cost, where N is the number
2675of flows, and is the default algorithm used by previous versions
2676dummynet's queues.
2677.It Cm rr
2678implements the Deficit Round Robin algorithm, which has O(1) processing
2679costs (roughly, 100-150ns per packet)
2680and permits bandwidth allocation according to weights, but
2681with poor service guarantees.
2682.It Cm qfq
2683implements the QFQ algorithm, which is a very fast variant of
2684WF2Q+, with similar service guarantees and O(1) processing
2685costs (roughly, 200-250ns per packet).
2686.El
2687.El
2688.Pp
2689In addition to the type, all parameters allowed for a pipe can also
2690be specified for a scheduler.
2691.Pp
2692Finally, the following parameters can be configured for both
2693pipes and queues:
2694.Pp
2695.Bl -tag -width XXXX -compact
2696.It Cm buckets Ar hash-table-size
2697Specifies the size of the hash table used for storing the
2698various queues.
2699Default value is 64 controlled by the
2700.Xr sysctl 8
2701variable
2702.Va net.inet.ip.dummynet.hash_size ,
2703allowed range is 16 to 65536.
2704.Pp
2705.It Cm mask Ar mask-specifier
2706Packets sent to a given pipe or queue by an
2707.Nm
2708rule can be further classified into multiple flows, each of which is then
2709sent to a different
2710.Em dynamic
2711pipe or queue.
2712A flow identifier is constructed by masking the IP addresses,
2713ports and protocol types as specified with the
2714.Cm mask
2715options in the configuration of the pipe or queue.
2716For each different flow identifier, a new pipe or queue is created
2717with the same parameters as the original object, and matching packets
2718are sent to it.
2719.Pp
2720Thus, when
2721.Em dynamic pipes
2722are used, each flow will get the same bandwidth as defined by the pipe,
2723whereas when
2724.Em dynamic queues
2725are used, each flow will share the parent's pipe bandwidth evenly
2726with other flows generated by the same queue (note that other queues
2727with different weights might be connected to the same pipe).
2728.br
2729Available mask specifiers are a combination of one or more of the following:
2730.Pp
2731.Cm dst-ip Ar mask ,
2732.Cm dst-ip6 Ar mask ,
2733.Cm src-ip Ar mask ,
2734.Cm src-ip6 Ar mask ,
2735.Cm dst-port Ar mask ,
2736.Cm src-port Ar mask ,
2737.Cm flow-id Ar mask ,
2738.Cm proto Ar mask
2739or
2740.Cm all ,
2741.Pp
2742where the latter means all bits in all fields are significant.
2743.Pp
2744.It Cm noerror
2745When a packet is dropped by a
2746.Nm dummynet
2747queue or pipe, the error
2748is normally reported to the caller routine in the kernel, in the
2749same way as it happens when a device queue fills up.
2750Setting this
2751option reports the packet as successfully delivered, which can be
2752needed for some experimental setups where you want to simulate
2753loss or congestion at a remote router.
2754.Pp
2755.It Cm plr Ar packet-loss-rate
2756Packet loss rate.
2757Argument
2758.Ar packet-loss-rate
2759is a floating-point number between 0 and 1, with 0 meaning no
2760loss, 1 meaning 100% loss.
2761The loss rate is internally represented on 31 bits.
2762.Pp
2763.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2764Queue size, in
2765.Ar slots
2766or
2767.Cm KBytes .
2768Default value is 50 slots, which
2769is the typical queue size for Ethernet devices.
2770Note that for slow speed links you should keep the queue
2771size short or your traffic might be affected by a significant
2772queueing delay.
2773E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2774or 20s of queue on a 30Kbit/s pipe.
2775Even worse effects can result if you get packets from an
2776interface with a much larger MTU, e.g.\& the loopback interface
2777with its 16KB packets.
2778The
2779.Xr sysctl 8
2780variables
2781.Em net.inet.ip.dummynet.pipe_byte_limit
2782and
2783.Em net.inet.ip.dummynet.pipe_slot_limit
2784control the maximum lengths that can be specified.
2785.Pp
2786.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2787[ecn]
2788Make use of the RED (Random Early Detection) queue management algorithm.
2789.Ar w_q
2790and
2791.Ar max_p
2792are floating
2793point numbers between 0 and 1 (inclusive), while
2794.Ar min_th
2795and
2796.Ar max_th
2797are integer numbers specifying thresholds for queue management
2798(thresholds are computed in bytes if the queue has been defined
2799in bytes, in slots otherwise).
2800The two parameters can also be of the same value if needed. The
2801.Nm dummynet
2802also supports the gentle RED variant (gred) and ECN (Explicit Congestion
2803Notification) as optional. Three
2804.Xr sysctl 8
2805variables can be used to control the RED behaviour:
2806.Bl -tag -width indent
2807.It Va net.inet.ip.dummynet.red_lookup_depth
2808specifies the accuracy in computing the average queue
2809when the link is idle (defaults to 256, must be greater than zero)
2810.It Va net.inet.ip.dummynet.red_avg_pkt_size
2811specifies the expected average packet size (defaults to 512, must be
2812greater than zero)
2813.It Va net.inet.ip.dummynet.red_max_pkt_size
2814specifies the expected maximum packet size, only used when queue
2815thresholds are in bytes (defaults to 1500, must be greater than zero).
2816.El
2817.El
2818.Pp
2819When used with IPv6 data,
2820.Nm dummynet
2821currently has several limitations.
2822Information necessary to route link-local packets to an
2823interface is not available after processing by
2824.Nm dummynet
2825so those packets are dropped in the output path.
2826Care should be taken to ensure that link-local packets are not passed to
2827.Nm dummynet .
2828.Sh CHECKLIST
2829Here are some important points to consider when designing your
2830rules:
2831.Bl -bullet
2832.It
2833Remember that you filter both packets going
2834.Cm in
2835and
2836.Cm out .
2837Most connections need packets going in both directions.
2838.It
2839Remember to test very carefully.
2840It is a good idea to be near the console when doing this.
2841If you cannot be near the console,
2842use an auto-recovery script such as the one in
2843.Pa /usr/share/examples/ipfw/change_rules.sh .
2844.It
2845Do not forget the loopback interface.
2846.El
2847.Sh FINE POINTS
2848.Bl -bullet
2849.It
2850There are circumstances where fragmented datagrams are unconditionally
2851dropped.
2852TCP packets are dropped if they do not contain at least 20 bytes of
2853TCP header, UDP packets are dropped if they do not contain a full 8
2854byte UDP header, and ICMP packets are dropped if they do not contain
28554 bytes of ICMP header, enough to specify the ICMP type, code, and
2856checksum.
2857These packets are simply logged as
2858.Dq pullup failed
2859since there may not be enough good data in the packet to produce a
2860meaningful log entry.
2861.It
2862Another type of packet is unconditionally dropped, a TCP packet with a
2863fragment offset of one.
2864This is a valid packet, but it only has one use, to try
2865to circumvent firewalls.
2866When logging is enabled, these packets are
2867reported as being dropped by rule -1.
2868.It
2869If you are logged in over a network, loading the
2870.Xr kld 4
2871version of
2872.Nm
2873is probably not as straightforward as you would think.
2874The following command line is recommended:
2875.Bd -literal -offset indent
2876kldload ipfw && \e
2877ipfw add 32000 allow ip from any to any
2878.Ed
2879.Pp
2880Along the same lines, doing an
2881.Bd -literal -offset indent
2882ipfw flush
2883.Ed
2884.Pp
2885in similar surroundings is also a bad idea.
2886.It
2887The
2888.Nm
2889filter list may not be modified if the system security level
2890is set to 3 or higher
2891(see
2892.Xr init 8
2893for information on system security levels).
2894.El
2895.Sh PACKET DIVERSION
2896A
2897.Xr divert 4
2898socket bound to the specified port will receive all packets
2899diverted to that port.
2900If no socket is bound to the destination port, or if the divert module is
2901not loaded, or if the kernel was not compiled with divert socket support,
2902the packets are dropped.
2903.Sh NETWORK ADDRESS TRANSLATION (NAT)
2904.Nm
2905support in-kernel NAT using the kernel version of
2906.Xr libalias 3 .
2907The kernel module
2908.Cm ipfw_nat
2909should be loaded or kernel should have
2910.Cm options IPFIREWALL_NAT
2911to be able use NAT.
2912.Pp
2913The nat configuration command is the following:
2914.Bd -ragged -offset indent
2915.Bk -words
2916.Cm nat
2917.Ar nat_number
2918.Cm config
2919.Ar nat-configuration
2920.Ek
2921.Ed
2922.Pp
2923The following parameters can be configured:
2924.Bl -tag -width indent
2925.It Cm ip Ar ip_address
2926Define an ip address to use for aliasing.
2927.It Cm if Ar nic
2928Use ip address of NIC for aliasing, dynamically changing
2929it if NIC's ip address changes.
2930.It Cm log
2931Enable logging on this nat instance.
2932.It Cm deny_in
2933Deny any incoming connection from outside world.
2934.It Cm same_ports
2935Try to leave the alias port numbers unchanged from
2936the actual local port numbers.
2937.It Cm unreg_only
2938Traffic on the local network not originating from an
2939unregistered address spaces will be ignored.
2940.It Cm reset
2941Reset table of the packet aliasing engine on address change.
2942.It Cm reverse
2943Reverse the way libalias handles aliasing.
2944.It Cm proxy_only
2945Obey transparent proxy rules only, packet aliasing is not performed.
2946.It Cm skip_global
2947Skip instance in case of global state lookup (see below).
2948.El
2949.Pp
2950Some specials value can be supplied instead of
2951.Va nat_number:
2952.Bl -tag -width indent
2953.It Cm global
2954Looks up translation state in all configured nat instances.
2955If an entry is found, packet is aliased according to that entry.
2956If no entry was found in any of the instances, packet is passed unchanged,
2957and no new entry will be created.
2958See section
2959.Sx MULTIPLE INSTANCES
2960in
2961.Xr natd 8
2962for more information.
2963.It Cm tablearg
2964Uses argument supplied in lookup table.
2965See
2966.Sx LOOKUP TABLES
2967section below for more information on lookup tables.
2968.El
2969.Pp
2970To let the packet continue after being (de)aliased, set the sysctl variable
2971.Va net.inet.ip.fw.one_pass
2972to 0.
2973For more information about aliasing modes, refer to
2974.Xr libalias 3 .
2975See Section
2976.Sx EXAMPLES
2977for some examples about nat usage.
2978.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
2979Redirect and LSNAT support follow closely the syntax used in
2980.Xr natd 8 .
2981See Section
2982.Sx EXAMPLES
2983for some examples on how to do redirect and lsnat.
2984.Ss SCTP NAT SUPPORT
2985SCTP nat can be configured in a similar manner to TCP through the
2986.Nm
2987command line tool.
2988The main difference is that
2989.Nm sctp nat
2990does not do port translation.
2991Since the local and global side ports will be the same,
2992there is no need to specify both.
2993Ports are redirected as follows:
2994.Bd -ragged -offset indent
2995.Bk -words
2996.Cm nat
2997.Ar nat_number
2998.Cm config if
2999.Ar nic
3000.Cm redirect_port sctp
3001.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
3002.Ek
3003.Ed
3004.Pp
3005Most
3006.Nm sctp nat
3007configuration can be done in real-time through the
3008.Xr sysctl 8
3009interface.
3010All may be changed dynamically, though the hash_table size will only
3011change for new
3012.Nm nat
3013instances.
3014See
3015.Sx SYSCTL VARIABLES
3016for more info.
3017.Sh IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
3018.Nm
3019supports in-kernel IPv6/IPv4 network address and protocol translation.
3020Stateful NAT64 translation allows IPv6-only clients to contact IPv4 servers
3021using unicast TCP, UDP or ICMP protocols.
3022One or more IPv4 addresses assigned to a stateful NAT64 translator are shared
3023among several IPv6-only clients.
3024When stateful NAT64 is used in conjunction with DNS64, no changes are usually
3025required in the IPv6 client or the IPv4 server.
3026The kernel module
3027.Cm ipfw_nat64
3028should be loaded or kernel should have
3029.Cm options IPFIREWALL_NAT64
3030to be able use stateful NAT64 translator.
3031.Pp
3032Stateful NAT64 uses a bunch of memory for several types of objects.
3033When IPv6 client initiates connection, NAT64 translator creates a host entry
3034in the states table.
3035Each host entry has a number of ports group entries allocated on demand.
3036Ports group entries contains connection state entries.
3037There are several options to control limits and lifetime for these objects.
3038.Pp
3039NAT64 translator follows RFC7915 when does ICMPv6/ICMP translation,
3040unsupported message types will be silently dropped.
3041IPv6 needs several ICMPv6 message types to be explicitly allowed for correct
3042operation.
3043Make sure that ND6 neighbor solicitation (ICMPv6 type 135) and neighbor
3044advertisement (ICMPv6 type 136) messages will not be handled by translation
3045rules.
3046.Pp
3047After translation NAT64 translator sends packets through corresponding netisr
3048queue.
3049Thus translator host should be configured as IPv4 and IPv6 router.
3050.Pp
3051Currently both stateful and stateless NAT64 translators use Well-Known IPv6
3052Prefix
3053.Ar 64:ff9b::/96
3054to represent IPv4 addresses in the IPv6 address.
3055Thus DNS64 service and routing should be configured to use Well-Known IPv6
3056Prefix.
3057.Pp
3058The stateful NAT64 configuration command is the following:
3059.Bd -ragged -offset indent
3060.Bk -words
3061.Cm nat64lsn
3062.Ar name
3063.Cm create
3064.Ar create-options
3065.Ek
3066.Ed
3067.Pp
3068The following parameters can be configured:
3069.Bl -tag -width indent
3070.It Cm prefix4 Ar ipv4_prefix/mask
3071The IPv4 prefix with mask defines the pool of IPv4 addresses used as
3072source address after translation.
3073Stateful NAT64 module translates IPv6 source address of client to one
3074IPv4 address from this pool.
3075Note that incoming IPv4 packets that don't have corresponding state entry
3076in the states table will be dropped by translator.
3077Make sure that translation rules handle packets, destined to configured prefix.
3078.It Cm max_ports Ar number
3079Maximum number of ports reserved for upper level protocols to one IPv6 client.
3080All reserved ports are divided into chunks between supported protocols.
3081The number of connections from one IPv6 client is limited by this option.
3082Note that closed TCP connections still remain in the list of connections until
3083.Cm tcp_close_age
3084interval will not expire.
3085Default value is
3086.Ar 2048 .
3087.It Cm host_del_age Ar seconds
3088The number of seconds until the host entry for a IPv6 client will be deleted
3089and all its resources will be released due to inactivity.
3090Default value is
3091.Ar 3600 .
3092.It Cm pg_del_age Ar seconds
3093The number of seconds until a ports group with unused state entries will
3094be released.
3095Default value is
3096.Ar 900 .
3097.It Cm tcp_syn_age Ar seconds
3098The number of seconds while a state entry for TCP connection with only SYN
3099sent will be kept.
3100If TCP connection establishing will not be finished,
3101state entry will be deleted.
3102Default value is
3103.Ar 10 .
3104.It Cm tcp_est_age Ar seconds
3105The number of seconds while a state entry for established TCP connection
3106will be kept.
3107Default value is
3108.Ar 7200 .
3109.It Cm tcp_close_age Ar seconds
3110The number of seconds while a state entry for closed TCP connection
3111will be kept.
3112Keeping state entries for closed connections is needed, because IPv4 servers
3113typically keep closed connections in a TIME_WAIT state for a several minutes.
3114Since translator's IPv4 addresses are shared among all IPv6 clients,
3115new connections from the same addresses and ports may be rejected by server,
3116because these connections are still in a TIME_WAIT state.
3117Keeping them in translator's state table protects from such rejects.
3118Default value is
3119.Ar 180 .
3120.It Cm udp_age Ar seconds
3121The number of seconds while translator keeps state entry in a waiting for
3122reply to the sent UDP datagram.
3123Default value is
3124.Ar 120 .
3125.It Cm icmp_age Ar seconds
3126The number of seconds while translator keeps state entry in a waiting for
3127reply to the sent ICMP message.
3128Default value is
3129.Ar 60 .
3130.It Cm log
3131Turn on logging of all handled packets via BPF through
3132.Ar ipfwlog0
3133interface.
3134.Ar ipfwlog0
3135is a pseudo interface and can be created after a boot manually with
3136.Cm ifconfig
3137command.
3138Note that it has different purpose than
3139.Ar ipfw0
3140interface.
3141Translators sends to BPF an additional information with each packet.
3142With
3143.Cm tcpdump
3144you are able to see each handled packet before and after translation.
3145.It Cm -log
3146Turn off logging of all handled packets via BPF.
3147.El
3148.Pp
3149To inspect a states table of stateful NAT64 the following command can be used:
3150.Bd -ragged -offset indent
3151.Bk -words
3152.Cm nat64lsn
3153.Ar name
3154.Cm show Cm states
3155.Ek
3156.Ed
3157.Pp
3158.Pp
3159Stateless NAT64 translator doesn't use a states table for translation
3160and converts IPv4 addresses to IPv6 and vice versa solely based on the
3161mappings taken from configured lookup tables.
3162Since a states table doesn't used by stateless translator,
3163it can be configured to pass IPv4 clients to IPv6-only servers.
3164.Pp
3165The stateless NAT64 configuration command is the following:
3166.Bd -ragged -offset indent
3167.Bk -words
3168.Cm nat64stl
3169.Ar name
3170.Cm create
3171.Ar create-options
3172.Ek
3173.Ed
3174.Pp
3175The following parameters can be configured:
3176.Bl -tag -width indent
3177.It Cm table4 Ar table46
3178The lookup table
3179.Ar table46
3180contains mapping how IPv4 addresses should be translated to IPv6 addresses.
3181.It Cm table6 Ar table64
3182The lookup table
3183.Ar table64
3184contains mapping how IPv6 addresses should be translated to IPv4 addresses.
3185.It Cm log
3186Turn on logging of all handled packets via BPF through
3187.Ar ipfwlog0
3188interface.
3189.It Cm -log
3190Turn off logging of all handled packets via BPF.
3191.El
3192.Pp
3193Note that the behavior of stateless translator with respect to not matched
3194packets differs from stateful translator.
3195If corresponding addresses was not found in the lookup tables, the packet
3196will not be dropped and the search continues.
3197.Sh IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
3198.Nm
3199supports in-kernel IPv6-to-IPv6 network prefix translation as described
3200in RFC6296.
3201The kernel module
3202.Cm ipfw_nptv6
3203should be loaded or kernel should has
3204.Cm options IPFIREWALL_NPTV6
3205to be able use NPTv6 translator.
3206.Pp
3207The NPTv6 configuration command is the following:
3208.Bd -ragged -offset indent
3209.Bk -words
3210.Cm nptv6
3211.Ar name
3212.Cm create
3213.Ar create-options
3214.Ek
3215.Ed
3216.Pp
3217The following parameters can be configured:
3218.Bl -tag -width indent
3219.It Cm int_prefix Ar ipv6_prefix
3220IPv6 prefix used in internal network.
3221NPTv6 module translates source address when it matches this prefix.
3222.It Cm ext_prefix Ar ipv6_prefix
3223IPv6 prefix used in external network.
3224NPTv6 module translates destination address when it matches this prefix.
3225.It Cm prefixlen Ar length
3226The length of specified IPv6 prefixes. It must be in range from 8 to 64.
3227.El
3228.Pp
3229Note that the prefix translation rules are silently ignored when IPv6 packet
3230forwarding is disabled.
3231To enable the packet forwarding, set the sysctl variable
3232.Va net.inet6.ip6.forwarding
3233to 1.
3234.Pp
3235To let the packet continue after being translated, set the sysctl variable
3236.Va net.inet.ip.fw.one_pass
3237to 0.
3238.Sh LOADER TUNABLES
3239Tunables can be set in
3240.Xr loader 8
3241prompt,
3242.Xr loader.conf 5
3243or
3244.Xr kenv 1
3245before ipfw module gets loaded.
3246.Bl -tag -width indent
3247.It Va net.inet.ip.fw.default_to_accept: No 0
3248Defines ipfw last rule behavior.
3249This value overrides
3250.Cd "options IPFW_DEFAULT_TO_(ACCEPT|DENY)"
3251from kernel configuration file.
3252.It Va net.inet.ip.fw.tables_max: No 128
3253Defines number of tables available in ipfw.
3254Number cannot exceed 65534.
3255.El
3256.Sh SYSCTL VARIABLES
3257A set of
3258.Xr sysctl 8
3259variables controls the behaviour of the firewall and
3260associated modules
3261.Pq Nm dummynet , bridge , sctp nat .
3262These are shown below together with their default value
3263(but always check with the
3264.Xr sysctl 8
3265command what value is actually in use) and meaning:
3266.Bl -tag -width indent
3267.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
3268Defines how the
3269.Nm nat
3270responds to receipt of global OOTB ASCONF-AddIP:
3271.Bl -tag -width indent
3272.It Cm 0
3273No response (unless a partially matching association exists -
3274ports and vtags match but global address does not)
3275.It Cm 1
3276.Nm nat
3277will accept and process all OOTB global AddIP messages.
3278.El
3279.Pp
3280Option 1 should never be selected as this forms a security risk.
3281An attacker can
3282establish multiple fake associations by sending AddIP messages.
3283.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
3284Defines the maximum number of chunks in an SCTP packet that will be
3285parsed for a
3286packet that matches an existing association.
3287This value is enforced to be greater or equal than
3288.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
3289A high value is
3290a DoS risk yet setting too low a value may result in
3291important control chunks in
3292the packet not being located and parsed.
3293.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
3294Defines when the
3295.Nm nat
3296responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
3297An OOTB packet is a packet that arrives with no existing association
3298registered in the
3299.Nm nat
3300and is not an INIT or ASCONF-AddIP packet:
3301.Bl -tag -width indent
3302.It Cm 0
3303ErrorM is never sent in response to OOTB packets.
3304.It Cm 1
3305ErrorM is only sent to OOTB packets received on the local side.
3306.It Cm 2
3307ErrorM is sent to the local side and on the global side ONLY if there is a
3308partial match (ports and vtags match but the source global IP does not).
3309This value is only useful if the
3310.Nm nat
3311is tracking global IP addresses.
3312.It Cm 3
3313ErrorM is sent in response to all OOTB packets on both
3314the local and global side
3315(DoS risk).
3316.El
3317.Pp
3318At the moment the default is 0, since the ErrorM packet is not yet
3319supported by most SCTP stacks.
3320When it is supported, and if not tracking
3321global addresses, we recommend setting this value to 1 to allow
3322multi-homed local hosts to function with the
3323.Nm nat .
3324To track global addresses, we recommend setting this value to 2 to
3325allow global hosts to be informed when they need to (re)send an
3326ASCONF-AddIP.
3327Value 3 should never be chosen (except for debugging) as the
3328.Nm nat
3329will respond to all OOTB global packets (a DoS risk).
3330.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
3331Size of hash tables used for
3332.Nm nat
3333lookups (100 < prime_number > 1000001).
3334This value sets the
3335.Nm hash table
3336size for any future created
3337.Nm nat
3338instance and therefore must be set prior to creating a
3339.Nm nat
3340instance.
3341The table sizes may be changed to suit specific needs.
3342If there will be few
3343concurrent associations, and memory is scarce, you may make these smaller.
3344If there will be many thousands (or millions) of concurrent associations, you
3345should make these larger.
3346A prime number is best for the table size.
3347The sysctl
3348update function will adjust your input value to the next highest prime number.
3349.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
3350Hold association in table for this many seconds after receiving a
3351SHUTDOWN-COMPLETE.
3352This allows endpoints to correct shutdown gracefully if a
3353shutdown_complete is lost and retransmissions are required.
3354.It Va net.inet.ip.alias.sctp.init_timer: No 15
3355Timeout value while waiting for (INIT-ACK|AddIP-ACK).
3356This value cannot be 0.
3357.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
3358Defines the maximum number of chunks in an SCTP packet that will be parsed when
3359no existing association exists that matches that packet.
3360Ideally this packet
3361will only be an INIT or ASCONF-AddIP packet.
3362A higher value may become a DoS
3363risk as malformed packets can consume processing resources.
3364.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
3365Defines the maximum number of parameters within a chunk that will be
3366parsed in a
3367packet.
3368As for other similar sysctl variables, larger values pose a DoS risk.
3369.It Va net.inet.ip.alias.sctp.log_level: No 0
3370Level of detail in the system log messages (0 \- minimal, 1 \- event,
33712 \- info, 3 \- detail, 4 \- debug, 5 \- max debug).
3372May be a good
3373option in high loss environments.
3374.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
3375Timeout value while waiting for SHUTDOWN-COMPLETE.
3376This value cannot be 0.
3377.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
3378Enables/disables global IP address tracking within the
3379.Nm nat
3380and places an
3381upper limit on the number of addresses tracked for each association:
3382.Bl -tag -width indent
3383.It Cm 0
3384Global tracking is disabled
3385.It Cm >1
3386Enables tracking, the maximum number of addresses tracked for each
3387association is limited to this value
3388.El
3389.Pp
3390This variable is fully dynamic, the new value will be adopted for all newly
3391arriving associations, existing associations are treated
3392as they were previously.
3393Global tracking will decrease the number of collisions within the
3394.Nm nat
3395at a cost
3396of increased processing load, memory usage, complexity, and possible
3397.Nm nat
3398state
3399problems in complex networks with multiple
3400.Nm nats .
3401We recommend not tracking
3402global IP addresses, this will still result in a fully functional
3403.Nm nat .
3404.It Va net.inet.ip.alias.sctp.up_timer: No 300
3405Timeout value to keep an association up with no traffic.
3406This value cannot be 0.
3407.It Va net.inet.ip.dummynet.expire : No 1
3408Lazily delete dynamic pipes/queue once they have no pending traffic.
3409You can disable this by setting the variable to 0, in which case
3410the pipes/queues will only be deleted when the threshold is reached.
3411.It Va net.inet.ip.dummynet.hash_size : No 64
3412Default size of the hash table used for dynamic pipes/queues.
3413This value is used when no
3414.Cm buckets
3415option is specified when configuring a pipe/queue.
3416.It Va net.inet.ip.dummynet.io_fast : No 0
3417If set to a non-zero value,
3418the
3419.Dq fast
3420mode of
3421.Nm dummynet
3422operation (see above) is enabled.
3423.It Va net.inet.ip.dummynet.io_pkt
3424Number of packets passed to
3425.Nm dummynet .
3426.It Va net.inet.ip.dummynet.io_pkt_drop
3427Number of packets dropped by
3428.Nm dummynet .
3429.It Va net.inet.ip.dummynet.io_pkt_fast
3430Number of packets bypassed by the
3431.Nm dummynet
3432scheduler.
3433.It Va net.inet.ip.dummynet.max_chain_len : No 16
3434Target value for the maximum number of pipes/queues in a hash bucket.
3435The product
3436.Cm max_chain_len*hash_size
3437is used to determine the threshold over which empty pipes/queues
3438will be expired even when
3439.Cm net.inet.ip.dummynet.expire=0 .
3440.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
3441.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
3442.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
3443Parameters used in the computations of the drop probability
3444for the RED algorithm.
3445.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
3446.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
3447The maximum queue size that can be specified in bytes or packets.
3448These limits prevent accidental exhaustion of resources such as mbufs.
3449If you raise these limits,
3450you should make sure the system is configured so that sufficient resources
3451are available.
3452.It Va net.inet.ip.fw.autoinc_step : No 100
3453Delta between rule numbers when auto-generating them.
3454The value must be in the range 1..1000.
3455.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
3456The current number of buckets in the hash table for dynamic rules
3457(readonly).
3458.It Va net.inet.ip.fw.debug : No 1
3459Controls debugging messages produced by
3460.Nm .
3461.It Va net.inet.ip.fw.default_rule : No 65535
3462The default rule number (read-only).
3463By the design of
3464.Nm , the default rule is the last one, so its number
3465can also serve as the highest number allowed for a rule.
3466.It Va net.inet.ip.fw.dyn_buckets : No 256
3467The number of buckets in the hash table for dynamic rules.
3468Must be a power of 2, up to 65536.
3469It only takes effect when all dynamic rules have expired, so you
3470are advised to use a
3471.Cm flush
3472command to make sure that the hash table is resized.
3473.It Va net.inet.ip.fw.dyn_count : No 3
3474Current number of dynamic rules
3475(read-only).
3476.It Va net.inet.ip.fw.dyn_keepalive : No 1
3477Enables generation of keepalive packets for
3478.Cm keep-state
3479rules on TCP sessions.
3480A keepalive is generated to both
3481sides of the connection every 5 seconds for the last 20
3482seconds of the lifetime of the rule.
3483.It Va net.inet.ip.fw.dyn_max : No 8192
3484Maximum number of dynamic rules.
3485When you hit this limit, no more dynamic rules can be
3486installed until old ones expire.
3487.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
3488.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
3489.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
3490.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
3491.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
3492.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
3493These variables control the lifetime, in seconds, of dynamic
3494rules.
3495Upon the initial SYN exchange the lifetime is kept short,
3496then increased after both SYN have been seen, then decreased
3497again during the final FIN exchange or when a RST is received.
3498Both
3499.Em dyn_fin_lifetime
3500and
3501.Em dyn_rst_lifetime
3502must be strictly lower than 5 seconds, the period of
3503repetition of keepalives.
3504The firewall enforces that.
3505.It Va net.inet.ip.fw.dyn_keep_states: No 0
3506Keep dynamic states on rule/set deletion.
3507States are relinked to default rule (65535).
3508This can be handly for ruleset reload.
3509Turned off by default.
3510.It Va net.inet.ip.fw.enable : No 1
3511Enables the firewall.
3512Setting this variable to 0 lets you run your machine without
3513firewall even if compiled in.
3514.It Va net.inet6.ip6.fw.enable : No 1
3515provides the same functionality as above for the IPv6 case.
3516.It Va net.inet.ip.fw.one_pass : No 1
3517When set, the packet exiting from the
3518.Nm dummynet
3519pipe or from
3520.Xr ng_ipfw 4
3521node is not passed though the firewall again.
3522Otherwise, after an action, the packet is
3523reinjected into the firewall at the next rule.
3524.It Va net.inet.ip.fw.tables_max : No 128
3525Maximum number of tables.
3526.It Va net.inet.ip.fw.verbose : No 1
3527Enables verbose messages.
3528.It Va net.inet.ip.fw.verbose_limit : No 0
3529Limits the number of messages produced by a verbose firewall.
3530.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
3531If enabled packets with unknown IPv6 Extension Headers will be denied.
3532.It Va net.link.ether.ipfw : No 0
3533Controls whether layer-2 packets are passed to
3534.Nm .
3535Default is no.
3536.It Va net.link.bridge.ipfw : No 0
3537Controls whether bridged packets are passed to
3538.Nm .
3539Default is no.
3540.El
3541.Sh INTERNAL DIAGNOSTICS
3542There are some commands that may be useful to understand current state
3543of certain subsystems inside kernel module.
3544These commands provide debugging output which may change without notice.
3545.Pp
3546Currently the following commands are available as
3547.Cm internal
3548sub-options:
3549.Bl -tag -width indent
3550.It Cm iflist
3551Lists all interface which are currently tracked by
3552.Nm
3553with their in-kernel status.
3554.It Cm talist
3555List all table lookup algorithms currently available.
3556.El
3557.Sh EXAMPLES
3558There are far too many possible uses of
3559.Nm
3560so this Section will only give a small set of examples.
3561.Pp
3562.Ss BASIC PACKET FILTERING
3563This command adds an entry which denies all tcp packets from
3564.Em cracker.evil.org
3565to the telnet port of
3566.Em wolf.tambov.su
3567from being forwarded by the host:
3568.Pp
3569.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
3570.Pp
3571This one disallows any connection from the entire cracker's
3572network to my host:
3573.Pp
3574.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
3575.Pp
3576A first and efficient way to limit access (not using dynamic rules)
3577is the use of the following rules:
3578.Pp
3579.Dl "ipfw add allow tcp from any to any established"
3580.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
3581.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
3582.Dl "..."
3583.Dl "ipfw add deny tcp from any to any"
3584.Pp
3585The first rule will be a quick match for normal TCP packets,
3586but it will not match the initial SYN packet, which will be
3587matched by the
3588.Cm setup
3589rules only for selected source/destination pairs.
3590All other SYN packets will be rejected by the final
3591.Cm deny
3592rule.
3593.Pp
3594If you administer one or more subnets, you can take advantage
3595of the address sets and or-blocks and write extremely
3596compact rulesets which selectively enable services to blocks
3597of clients, as below:
3598.Pp
3599.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
3600.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
3601.Dl ""
3602.Dl "ipfw add allow ip from ${goodguys} to any"
3603.Dl "ipfw add deny ip from ${badguys} to any"
3604.Dl "... normal policies ..."
3605.Pp
3606The
3607.Cm verrevpath
3608option could be used to do automated anti-spoofing by adding the
3609following to the top of a ruleset:
3610.Pp
3611.Dl "ipfw add deny ip from any to any not verrevpath in"
3612.Pp
3613This rule drops all incoming packets that appear to be coming to the
3614system on the wrong interface.
3615For example, a packet with a source
3616address belonging to a host on a protected internal network would be
3617dropped if it tried to enter the system from an external interface.
3618.Pp
3619The
3620.Cm antispoof
3621option could be used to do similar but more restricted anti-spoofing
3622by adding the following to the top of a ruleset:
3623.Pp
3624.Dl "ipfw add deny ip from any to any not antispoof in"
3625.Pp
3626This rule drops all incoming packets that appear to be coming from another
3627directly connected system but on the wrong interface.
3628For example, a packet with a source address of
3629.Li 192.168.0.0/24 ,
3630configured on
3631.Li fxp0 ,
3632but coming in on
3633.Li fxp1
3634would be dropped.
3635.Pp
3636The
3637.Cm setdscp
3638option could be used to (re)mark user traffic,
3639by adding the following to the appropriate place in ruleset:
3640.Pp
3641.Dl "ipfw add setdscp be ip from any to any dscp af11,af21"
3642.Ss DYNAMIC RULES
3643In order to protect a site from flood attacks involving fake
3644TCP packets, it is safer to use dynamic rules:
3645.Pp
3646.Dl "ipfw add check-state"
3647.Dl "ipfw add deny tcp from any to any established"
3648.Dl "ipfw add allow tcp from my-net to any setup keep-state"
3649.Pp
3650This will let the firewall install dynamic rules only for
3651those connection which start with a regular SYN packet coming
3652from the inside of our network.
3653Dynamic rules are checked when encountering the first
3654occurrence of a
3655.Cm check-state ,
3656.Cm keep-state
3657or
3658.Cm limit
3659rule.
3660A
3661.Cm check-state
3662rule should usually be placed near the beginning of the
3663ruleset to minimize the amount of work scanning the ruleset.
3664Your mileage may vary.
3665.Pp
3666To limit the number of connections a user can open
3667you can use the following type of rules:
3668.Pp
3669.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
3670.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
3671.Pp
3672The former (assuming it runs on a gateway) will allow each host
3673on a /24 network to open at most 10 TCP connections.
3674The latter can be placed on a server to make sure that a single
3675client does not use more than 4 simultaneous connections.
3676.Pp
3677.Em BEWARE :
3678stateful rules can be subject to denial-of-service attacks
3679by a SYN-flood which opens a huge number of dynamic rules.
3680The effects of such attacks can be partially limited by
3681acting on a set of
3682.Xr sysctl 8
3683variables which control the operation of the firewall.
3684.Pp
3685Here is a good usage of the
3686.Cm list
3687command to see accounting records and timestamp information:
3688.Pp
3689.Dl ipfw -at list
3690.Pp
3691or in short form without timestamps:
3692.Pp
3693.Dl ipfw -a list
3694.Pp
3695which is equivalent to:
3696.Pp
3697.Dl ipfw show
3698.Pp
3699Next rule diverts all incoming packets from 192.168.2.0/24
3700to divert port 5000:
3701.Pp
3702.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
3703.Ss TRAFFIC SHAPING
3704The following rules show some of the applications of
3705.Nm
3706and
3707.Nm dummynet
3708for simulations and the like.
3709.Pp
3710This rule drops random incoming packets with a probability
3711of 5%:
3712.Pp
3713.Dl "ipfw add prob 0.05 deny ip from any to any in"
3714.Pp
3715A similar effect can be achieved making use of
3716.Nm dummynet
3717pipes:
3718.Pp
3719.Dl "ipfw add pipe 10 ip from any to any"
3720.Dl "ipfw pipe 10 config plr 0.05"
3721.Pp
3722We can use pipes to artificially limit bandwidth, e.g.\& on a
3723machine acting as a router, if we want to limit traffic from
3724local clients on 192.168.2.0/24 we do:
3725.Pp
3726.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3727.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
3728.Pp
3729note that we use the
3730.Cm out
3731modifier so that the rule is not used twice.
3732Remember in fact that
3733.Nm
3734rules are checked both on incoming and outgoing packets.
3735.Pp
3736Should we want to simulate a bidirectional link with bandwidth
3737limitations, the correct way is the following:
3738.Pp
3739.Dl "ipfw add pipe 1 ip from any to any out"
3740.Dl "ipfw add pipe 2 ip from any to any in"
3741.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
3742.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
3743.Pp
3744The above can be very useful, e.g.\& if you want to see how
3745your fancy Web page will look for a residential user who
3746is connected only through a slow link.
3747You should not use only one pipe for both directions, unless
3748you want to simulate a half-duplex medium (e.g.\& AppleTalk,
3749Ethernet, IRDA).
3750It is not necessary that both pipes have the same configuration,
3751so we can also simulate asymmetric links.
3752.Pp
3753Should we want to verify network performance with the RED queue
3754management algorithm:
3755.Pp
3756.Dl "ipfw add pipe 1 ip from any to any"
3757.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
3758.Pp
3759Another typical application of the traffic shaper is to
3760introduce some delay in the communication.
3761This can significantly affect applications which do a lot of Remote
3762Procedure Calls, and where the round-trip-time of the
3763connection often becomes a limiting factor much more than
3764bandwidth:
3765.Pp
3766.Dl "ipfw add pipe 1 ip from any to any out"
3767.Dl "ipfw add pipe 2 ip from any to any in"
3768.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
3769.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
3770.Pp
3771Per-flow queueing can be useful for a variety of purposes.
3772A very simple one is counting traffic:
3773.Pp
3774.Dl "ipfw add pipe 1 tcp from any to any"
3775.Dl "ipfw add pipe 1 udp from any to any"
3776.Dl "ipfw add pipe 1 ip from any to any"
3777.Dl "ipfw pipe 1 config mask all"
3778.Pp
3779The above set of rules will create queues (and collect
3780statistics) for all traffic.
3781Because the pipes have no limitations, the only effect is
3782collecting statistics.
3783Note that we need 3 rules, not just the last one, because
3784when
3785.Nm
3786tries to match IP packets it will not consider ports, so we
3787would not see connections on separate ports as different
3788ones.
3789.Pp
3790A more sophisticated example is limiting the outbound traffic
3791on a net with per-host limits, rather than per-network limits:
3792.Pp
3793.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3794.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
3795.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3796.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3797.Ss LOOKUP TABLES
3798In the following example, we need to create several traffic bandwidth
3799classes and we need different hosts/networks to fall into different classes.
3800We create one pipe for each class and configure them accordingly.
3801Then we create a single table and fill it with IP subnets and addresses.
3802For each subnet/host we set the argument equal to the number of the pipe
3803that it should use.
3804Then we classify traffic using a single rule:
3805.Pp
3806.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
3807.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
3808.Dl "..."
3809.Dl "ipfw table T1 create type addr"
3810.Dl "ipfw table T1 add 192.168.2.0/24 1"
3811.Dl "ipfw table T1 add 192.168.0.0/27 4"
3812.Dl "ipfw table T1 add 192.168.0.2 1"
3813.Dl "..."
3814.Dl "ipfw add pipe tablearg ip from 'table(T1)' to any"
3815.Pp
3816Using the
3817.Cm fwd
3818action, the table entries may include hostnames and IP addresses.
3819.Pp
3820.Dl "ipfw table T2 create type addr ftype ip"
3821.Dl "ipfw table T2 add 192.168.2.0/24 10.23.2.1"
3822.Dl "ipfw table T21 add 192.168.0.0/27 router1.dmz"
3823.Dl "..."
3824.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
3825.Pp
3826In the following example per-interface firewall is created:
3827.Pp
3828.Dl "ipfw table IN create type iface valtype skipto,fib"
3829.Dl "ipfw table IN add vlan20 12000,12"
3830.Dl "ipfw table IN add vlan30 13000,13"
3831.Dl "ipfw table OUT create type iface valtype skipto"
3832.Dl "ipfw table OUT add vlan20 22000"
3833.Dl "ipfw table OUT add vlan30 23000"
3834.Dl ".."
3835.Dl "ipfw add 100 ipfw setfib tablearg ip from any to any recv 'table(IN)' in"
3836.Dl "ipfw add 200 ipfw skipto tablearg ip from any to any recv 'table(IN)' in"
3837.Dl "ipfw add 300 ipfw skipto tablearg ip from any to any xmit 'table(OUT)' out"
3838.Pp
3839The following example illustrate usage of flow tables:
3840.Pp
3841.Dl "ipfw table fl create type flow:flow:src-ip,proto,dst-ip,dst-port"
3842.Dl "ipfw table fl add 2a02:6b8:77::88,tcp,2a02:6b8:77::99,80 11"
3843.Dl "ipfw table fl add 10.0.0.1,udp,10.0.0.2,53 12"
3844.Dl ".."
3845.Dl "ipfw add 100 allow ip from any to any flow 'table(fl,11)' recv ix0"
3846.Ss SETS OF RULES
3847To add a set of rules atomically, e.g.\& set 18:
3848.Pp
3849.Dl "ipfw set disable 18"
3850.Dl "ipfw add NN set 18 ...         # repeat as needed"
3851.Dl "ipfw set enable 18"
3852.Pp
3853To delete a set of rules atomically the command is simply:
3854.Pp
3855.Dl "ipfw delete set 18"
3856.Pp
3857To test a ruleset and disable it and regain control if something goes wrong:
3858.Pp
3859.Dl "ipfw set disable 18"
3860.Dl "ipfw add NN set 18 ...         # repeat as needed"
3861.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
3862.Pp
3863Here if everything goes well, you press control-C before the "sleep"
3864terminates, and your ruleset will be left active.
3865Otherwise, e.g.\& if
3866you cannot access your box, the ruleset will be disabled after
3867the sleep terminates thus restoring the previous situation.
3868.Pp
3869To show rules of the specific set:
3870.Pp
3871.Dl "ipfw set 18 show"
3872.Pp
3873To show rules of the disabled set:
3874.Pp
3875.Dl "ipfw -S set 18 show"
3876.Pp
3877To clear a specific rule counters of the specific set:
3878.Pp
3879.Dl "ipfw set 18 zero NN"
3880.Pp
3881To delete a specific rule of the specific set:
3882.Pp
3883.Dl "ipfw set 18 delete NN"
3884.Ss NAT, REDIRECT AND LSNAT
3885First redirect all the traffic to nat instance 123:
3886.Pp
3887.Dl "ipfw add nat 123 all from any to any"
3888.Pp
3889Then to configure nat instance 123 to alias all the outgoing traffic with ip
3890192.168.0.123, blocking all incoming connections, trying to keep
3891same ports on both sides, clearing aliasing table on address change
3892and keeping a log of traffic/link statistics:
3893.Pp
3894.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
3895.Pp
3896Or to change address of instance 123, aliasing table will be cleared (see
3897reset option):
3898.Pp
3899.Dl "ipfw nat 123 config ip 10.0.0.1"
3900.Pp
3901To see configuration of nat instance 123:
3902.Pp
3903.Dl "ipfw nat 123 show config"
3904.Pp
3905To show logs of all the instances in range 111-999:
3906.Pp
3907.Dl "ipfw nat 111-999 show"
3908.Pp
3909To see configurations of all instances:
3910.Pp
3911.Dl "ipfw nat show config"
3912.Pp
3913Or a redirect rule with mixed modes could looks like:
3914.Pp
3915.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
3916.Dl "			 redirect_port tcp 192.168.0.1:80 500"
3917.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
3918.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
3919.Dl "			 	    10.0.0.100	# LSNAT"
3920.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
3921.Dl "			 	    500		# LSNAT"
3922.Pp
3923or it could be split in:
3924.Pp
3925.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
3926.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
3927.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
3928.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
3929.Dl "				         10.0.0.100"
3930.Dl "ipfw nat 5 config redirect_port tcp"
3931.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
3932.Sh SEE ALSO
3933.Xr cpp 1 ,
3934.Xr m4 1 ,
3935.Xr altq 4 ,
3936.Xr divert 4 ,
3937.Xr dummynet 4 ,
3938.Xr if_bridge 4 ,
3939.Xr ip 4 ,
3940.Xr ipfirewall 4 ,
3941.Xr ng_ipfw 4 ,
3942.Xr protocols 5 ,
3943.Xr services 5 ,
3944.Xr init 8 ,
3945.Xr kldload 8 ,
3946.Xr reboot 8 ,
3947.Xr sysctl 8 ,
3948.Xr syslogd 8
3949.Sh HISTORY
3950The
3951.Nm
3952utility first appeared in
3953.Fx 2.0 .
3954.Nm dummynet
3955was introduced in
3956.Fx 2.2.8 .
3957Stateful extensions were introduced in
3958.Fx 4.0 .
3959.Nm ipfw2
3960was introduced in Summer 2002.
3961.Sh AUTHORS
3962.An Ugen J. S. Antsilevich ,
3963.An Poul-Henning Kamp ,
3964.An Alex Nash ,
3965.An Archie Cobbs ,
3966.An Luigi Rizzo .
3967.Pp
3968.An -nosplit
3969API based upon code written by
3970.An Daniel Boulet
3971for BSDI.
3972.Pp
3973Dummynet has been introduced by Luigi Rizzo in 1997-1998.
3974.Pp
3975Some early work (1999-2000) on the
3976.Nm dummynet
3977traffic shaper supported by Akamba Corp.
3978.Pp
3979The ipfw core (ipfw2) has been completely redesigned and
3980reimplemented by Luigi Rizzo in summer 2002.
3981Further
3982actions and
3983options have been added by various developer over the years.
3984.Pp
3985.An -nosplit
3986In-kernel NAT support written by
3987.An Paolo Pisati Aq Mt piso@FreeBSD.org
3988as part of a Summer of Code 2005 project.
3989.Pp
3990SCTP
3991.Nm nat
3992support has been developed by
3993.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
3994The primary developers and maintainers are David Hayes and Jason But.
3995For further information visit:
3996.Aq http://www.caia.swin.edu.au/urp/SONATA
3997.Pp
3998Delay profiles have been developed by Alessandro Cerri and
3999Luigi Rizzo, supported by the
4000European Commission within Projects Onelab and Onelab2.
4001.Sh BUGS
4002The syntax has grown over the years and sometimes it might be confusing.
4003Unfortunately, backward compatibility prevents cleaning up mistakes
4004made in the definition of the syntax.
4005.Pp
4006.Em !!! WARNING !!!
4007.Pp
4008Misconfiguring the firewall can put your computer in an unusable state,
4009possibly shutting down network services and requiring console access to
4010regain control of it.
4011.Pp
4012Incoming packet fragments diverted by
4013.Cm divert
4014are reassembled before delivery to the socket.
4015The action used on those packet is the one from the
4016rule which matches the first fragment of the packet.
4017.Pp
4018Packets diverted to userland, and then reinserted by a userland process
4019may lose various packet attributes.
4020The packet source interface name
4021will be preserved if it is shorter than 8 bytes and the userland process
4022saves and reuses the sockaddr_in
4023(as does
4024.Xr natd 8 ) ;
4025otherwise, it may be lost.
4026If a packet is reinserted in this manner, later rules may be incorrectly
4027applied, making the order of
4028.Cm divert
4029rules in the rule sequence very important.
4030.Pp
4031Dummynet drops all packets with IPv6 link-local addresses.
4032.Pp
4033Rules using
4034.Cm uid
4035or
4036.Cm gid
4037may not behave as expected.
4038In particular, incoming SYN packets may
4039have no uid or gid associated with them since they do not yet belong
4040to a TCP connection, and the uid/gid associated with a packet may not
4041be as expected if the associated process calls
4042.Xr setuid 2
4043or similar system calls.
4044.Pp
4045Rule syntax is subject to the command line environment and some patterns
4046may need to be escaped with the backslash character
4047or quoted appropriately.
4048.Pp
4049Due to the architecture of
4050.Xr libalias 3 ,
4051ipfw nat is not compatible with the TCP segmentation offloading (TSO).
4052Thus, to reliably nat your network traffic, please disable TSO
4053on your NICs using
4054.Xr ifconfig 8 .
4055.Pp
4056ICMP error messages are not implicitly matched by dynamic rules
4057for the respective conversations.
4058To avoid failures of network error detection and path MTU discovery,
4059ICMP error messages may need to be allowed explicitly through static
4060rules.
4061.Pp
4062Rules using
4063.Cm call
4064and
4065.Cm return
4066actions may lead to confusing behaviour if ruleset has mistakes,
4067and/or interaction with other subsystems (netgraph, dummynet, etc.) is used.
4068One possible case for this is packet leaving
4069.Nm
4070in subroutine on the input pass, while later on output encountering unpaired
4071.Cm return
4072first.
4073As the call stack is kept intact after input pass, packet will suddenly
4074return to the rule number used on input pass, not on output one.
4075Order of processing should be checked carefully to avoid such mistakes.
4076