xref: /freebsd/sbin/ipfw/ipfw.8 (revision 907b59d76938e654f0d040a888e8dfca3de1e222)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd July 19, 2016
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 Ar name Cm destroy
54.Nm
55.Oo Cm set Ar N Oc Cm table Ar name Cm modify Ar modify-options
56.Nm
57.Oo Cm set Ar N Oc Cm table Ar name Cm swap Ar name
58.Nm
59.Oo Cm set Ar N Oc Cm table Ar name Cm add Ar table-key Op Ar value
60.Nm
61.Oo Cm set Ar N Oc Cm table Ar name Cm add Op Ar table-key Ar value ...
62.Nm
63.Oo Cm set Ar N Oc Cm table Ar name Cm atomic add Op Ar table-key Ar value ...
64.Nm
65.Oo Cm set Ar N Oc Cm table Ar name Cm delete Op Ar table-key ...
66.Nm
67.Oo Cm set Ar N Oc Cm table Ar name Cm lookup Ar addr
68.Nm
69.Oo Cm set Ar N Oc Cm table Ar name Cm lock
70.Nm
71.Oo Cm set Ar N Oc Cm table Ar name Cm unlock
72.Nm
73.Oo Cm set Ar N Oc Cm table
74.Brq Ar name | all
75.Cm list
76.Nm
77.Oo Cm set Ar N Oc Cm table
78.Brq Ar name | all
79.Cm info
80.Nm
81.Oo Cm set Ar N Oc Cm table
82.Brq Ar name | all
83.Cm detail
84.Nm
85.Oo Cm set Ar N Oc Cm table
86.Brq Ar name | all
87.Cm flush
88.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
89.Nm
90.Brq Cm pipe | queue | sched
91.Ar number
92.Cm config
93.Ar config-options
94.Nm
95.Op Fl s Op Ar field
96.Brq Cm pipe | queue | sched
97.Brq Cm delete | list | show
98.Op Ar number ...
99.Ss IN-KERNEL NAT
100.Nm
101.Op Fl q
102.Cm nat
103.Ar number
104.Cm config
105.Ar config-options
106.Pp
107.Nm
108.Op Fl cfnNqS
109.Oo
110.Fl p Ar preproc
111.Oo
112.Ar preproc-flags
113.Oc
114.Oc
115.Ar pathname
116.Ss IPv6-to-IPv6 NETWORK PREFIX TRANSLATION
117.Nm
118.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm create Ar create-options
119.Nm
120.Oo Cm set Ar N Oc Cm nptv6
121.Brq Ar name | all
122.Brq Cm list | show
123.Nm
124.Oo Cm set Ar N Oc Cm nptv6
125.Brq Ar name | all
126.Cm destroy
127.Nm
128.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm stats
129.Ss INTERNAL DIAGNOSTICS
130.Nm
131.Cm internal iflist
132.Nm
133.Cm internal talist
134.Nm
135.Cm internal vlist
136.Sh DESCRIPTION
137The
138.Nm
139utility is the user interface for controlling the
140.Xr ipfw 4
141firewall, the
142.Xr dummynet 4
143traffic shaper/packet scheduler, and the
144in-kernel NAT services.
145.Pp
146A firewall configuration, or
147.Em ruleset ,
148is made of a list of
149.Em rules
150numbered from 1 to 65535.
151Packets are passed to the firewall
152from a number of different places in the protocol stack
153(depending on the source and destination of the packet,
154it is possible for the firewall to be
155invoked multiple times on the same packet).
156The packet passed to the firewall is compared
157against each of the rules in the
158.Em ruleset ,
159in rule-number order
160(multiple rules with the same number are permitted, in which case
161they are processed in order of insertion).
162When a match is found, the action corresponding to the
163matching rule is performed.
164.Pp
165Depending on the action and certain system settings, packets
166can be reinjected into the firewall at some rule after the
167matching one for further processing.
168.Pp
169A ruleset always includes a
170.Em default
171rule (numbered 65535) which cannot be modified or deleted,
172and matches all packets.
173The action associated with the
174.Em default
175rule can be either
176.Cm deny
177or
178.Cm allow
179depending on how the kernel is configured.
180.Pp
181If the ruleset includes one or more rules with the
182.Cm keep-state
183or
184.Cm limit
185option,
186the firewall will have a
187.Em stateful
188behaviour, i.e., upon a match it will create
189.Em dynamic rules ,
190i.e., rules that match packets with the same 5-tuple
191(protocol, source and destination addresses and ports)
192as the packet which caused their creation.
193Dynamic rules, which have a limited lifetime, are checked
194at the first occurrence of a
195.Cm check-state ,
196.Cm keep-state
197or
198.Cm limit
199rule, and are typically used to open the firewall on-demand to
200legitimate traffic only.
201See the
202.Sx STATEFUL FIREWALL
203and
204.Sx EXAMPLES
205Sections below for more information on the stateful behaviour of
206.Nm .
207.Pp
208All rules (including dynamic ones) have a few associated counters:
209a packet count, a byte count, a log count and a timestamp
210indicating the time of the last match.
211Counters can be displayed or reset with
212.Nm
213commands.
214.Pp
215Each rule belongs to one of 32 different
216.Em sets
217, and there are
218.Nm
219commands to atomically manipulate sets, such as enable,
220disable, swap sets, move all rules in a set to another
221one, delete all rules in a set.
222These can be useful to
223install temporary configurations, or to test them.
224See Section
225.Sx SETS OF RULES
226for more information on
227.Em sets .
228.Pp
229Rules can be added with the
230.Cm add
231command; deleted individually or in groups with the
232.Cm delete
233command, and globally (except those in set 31) with the
234.Cm flush
235command; displayed, optionally with the content of the
236counters, using the
237.Cm show
238and
239.Cm list
240commands.
241Finally, counters can be reset with the
242.Cm zero
243and
244.Cm resetlog
245commands.
246.Pp
247.Ss COMMAND OPTIONS
248The following general options are available when invoking
249.Nm :
250.Bl -tag -width indent
251.It Fl a
252Show counter values when listing rules.
253The
254.Cm show
255command implies this option.
256.It Fl b
257Only show the action and the comment, not the body of a rule.
258Implies
259.Fl c .
260.It Fl c
261When entering or showing rules, print them in compact form,
262i.e., omitting the "ip from any to any" string
263when this does not carry any additional information.
264.It Fl d
265When listing, show dynamic rules in addition to static ones.
266.It Fl e
267When listing and
268.Fl d
269is specified, also show expired dynamic rules.
270.It Fl f
271Do not ask for confirmation for commands that can cause problems
272if misused, i.e.,
273.Cm flush .
274If there is no tty associated with the process, this is implied.
275.It Fl i
276When listing a table (see the
277.Sx LOOKUP TABLES
278section below for more information on lookup tables), format values
279as IP addresses.
280By default, values are shown as integers.
281.It Fl n
282Only check syntax of the command strings, without actually passing
283them to the kernel.
284.It Fl N
285Try to resolve addresses and service names in output.
286.It Fl q
287Be quiet when executing the
288.Cm add ,
289.Cm nat ,
290.Cm zero ,
291.Cm resetlog
292or
293.Cm flush
294commands;
295(implies
296.Fl f ) .
297This is useful when updating rulesets by executing multiple
298.Nm
299commands in a script
300(e.g.,
301.Ql sh\ /etc/rc.firewall ) ,
302or by processing a file with many
303.Nm
304rules across a remote login session.
305It also stops a table add or delete
306from failing if the entry already exists or is not present.
307.Pp
308The reason why this option may be important is that
309for some of these actions,
310.Nm
311may print a message; if the action results in blocking the
312traffic to the remote client,
313the remote login session will be closed
314and the rest of the ruleset will not be processed.
315Access to the console would then be required to recover.
316.It Fl S
317When listing rules, show the
318.Em set
319each rule belongs to.
320If this flag is not specified, disabled rules will not be
321listed.
322.It Fl s Op Ar field
323When listing pipes, sort according to one of the four
324counters (total or current packets or bytes).
325.It Fl t
326When listing, show last match timestamp converted with ctime().
327.It Fl T
328When listing, show last match timestamp as seconds from the epoch.
329This form can be more convenient for postprocessing by scripts.
330.El
331.Ss LIST OF RULES AND PREPROCESSING
332To ease configuration, rules can be put into a file which is
333processed using
334.Nm
335as shown in the last synopsis line.
336An absolute
337.Ar pathname
338must be used.
339The file will be read line by line and applied as arguments to the
340.Nm
341utility.
342.Pp
343Optionally, a preprocessor can be specified using
344.Fl p Ar preproc
345where
346.Ar pathname
347is to be piped through.
348Useful preprocessors include
349.Xr cpp 1
350and
351.Xr m4 1 .
352If
353.Ar preproc
354does not start with a slash
355.Pq Ql /
356as its first character, the usual
357.Ev PATH
358name search is performed.
359Care should be taken with this in environments where not all
360file systems are mounted (yet) by the time
361.Nm
362is being run (e.g.\& when they are mounted over NFS).
363Once
364.Fl p
365has been specified, any additional arguments are passed on to the preprocessor
366for interpretation.
367This allows for flexible configuration files (like conditionalizing
368them on the local hostname) and the use of macros to centralize
369frequently required arguments like IP addresses.
370.Ss TRAFFIC SHAPER CONFIGURATION
371The
372.Nm
373.Cm pipe , queue
374and
375.Cm sched
376commands are used to configure the traffic shaper and packet scheduler.
377See the
378.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
379Section below for details.
380.Pp
381If the world and the kernel get out of sync the
382.Nm
383ABI may break, preventing you from being able to add any rules.
384This can adversely affect the booting process.
385You can use
386.Nm
387.Cm disable
388.Cm firewall
389to temporarily disable the firewall to regain access to the network,
390allowing you to fix the problem.
391.Sh PACKET FLOW
392A packet is checked against the active ruleset in multiple places
393in the protocol stack, under control of several sysctl variables.
394These places and variables are shown below, and it is important to
395have this picture in mind in order to design a correct ruleset.
396.Bd -literal -offset indent
397       ^    to upper layers    V
398       |                       |
399       +----------->-----------+
400       ^                       V
401 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
402       |                       |
403       ^                       V
404 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
405       |                       |
406       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
407       ^                       V
408       |      to devices       |
409.Ed
410.Pp
411The number of
412times the same packet goes through the firewall can
413vary between 0 and 4 depending on packet source and
414destination, and system configuration.
415.Pp
416Note that as packets flow through the stack, headers can be
417stripped or added to it, and so they may or may not be available
418for inspection.
419E.g., incoming packets will include the MAC header when
420.Nm
421is invoked from
422.Cm ether_demux() ,
423but the same packets will have the MAC header stripped off when
424.Nm
425is invoked from
426.Cm ip_input()
427or
428.Cm ip6_input() .
429.Pp
430Also note that each packet is always checked against the complete ruleset,
431irrespective of the place where the check occurs, or the source of the packet.
432If a rule contains some match patterns or actions which are not valid
433for the place of invocation (e.g.\& trying to match a MAC header within
434.Cm ip_input
435or
436.Cm ip6_input ),
437the match pattern will not match, but a
438.Cm not
439operator in front of such patterns
440.Em will
441cause the pattern to
442.Em always
443match on those packets.
444It is thus the responsibility of
445the programmer, if necessary, to write a suitable ruleset to
446differentiate among the possible places.
447.Cm skipto
448rules can be useful here, as an example:
449.Bd -literal -offset indent
450# packets from ether_demux or bdg_forward
451ipfw add 10 skipto 1000 all from any to any layer2 in
452# packets from ip_input
453ipfw add 10 skipto 2000 all from any to any not layer2 in
454# packets from ip_output
455ipfw add 10 skipto 3000 all from any to any not layer2 out
456# packets from ether_output_frame
457ipfw add 10 skipto 4000 all from any to any layer2 out
458.Ed
459.Pp
460(yes, at the moment there is no way to differentiate between
461ether_demux and bdg_forward).
462.Sh SYNTAX
463In general, each keyword or argument must be provided as
464a separate command line argument, with no leading or trailing
465spaces.
466Keywords are case-sensitive, whereas arguments may
467or may not be case-sensitive depending on their nature
468(e.g.\& uid's are, hostnames are not).
469.Pp
470Some arguments (e.g., port or address lists) are comma-separated
471lists of values.
472In this case, spaces after commas ',' are allowed to make
473the line more readable.
474You can also put the entire
475command (including flags) into a single argument.
476E.g., the following forms are equivalent:
477.Bd -literal -offset indent
478ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
479ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
480ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
481.Ed
482.Sh RULE FORMAT
483The format of firewall rules is the following:
484.Bd -ragged -offset indent
485.Bk -words
486.Op Ar rule_number
487.Op Cm set Ar set_number
488.Op Cm prob Ar match_probability
489.Ar action
490.Op Cm log Op Cm logamount Ar number
491.Op Cm altq Ar queue
492.Oo
493.Bro Cm tag | untag
494.Brc Ar number
495.Oc
496.Ar body
497.Ek
498.Ed
499.Pp
500where the body of the rule specifies which information is used
501for filtering packets, among the following:
502.Pp
503.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
504.It Layer-2 header fields
505When available
506.It IPv4 and IPv6 Protocol
507TCP, UDP, ICMP, etc.
508.It Source and dest. addresses and ports
509.It Direction
510See Section
511.Sx PACKET FLOW
512.It Transmit and receive interface
513By name or address
514.It Misc. IP header fields
515Version, type of service, datagram length, identification,
516fragment flag (non-zero IP offset),
517Time To Live
518.It IP options
519.It IPv6 Extension headers
520Fragmentation, Hop-by-Hop options,
521Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
522.It IPv6 Flow-ID
523.It Misc. TCP header fields
524TCP flags (SYN, FIN, ACK, RST, etc.),
525sequence number, acknowledgment number,
526window
527.It TCP options
528.It ICMP types
529for ICMP packets
530.It ICMP6 types
531for ICMP6 packets
532.It User/group ID
533When the packet can be associated with a local socket.
534.It Divert status
535Whether a packet came from a divert socket (e.g.,
536.Xr natd 8 ) .
537.It Fib annotation state
538Whether a packet has been tagged for using a specific FIB (routing table)
539in future forwarding decisions.
540.El
541.Pp
542Note that some of the above information, e.g.\& source MAC or IP addresses and
543TCP/UDP ports, can be easily spoofed, so filtering on those fields
544alone might not guarantee the desired results.
545.Bl -tag -width indent
546.It Ar rule_number
547Each rule is associated with a
548.Ar rule_number
549in the range 1..65535, with the latter reserved for the
550.Em default
551rule.
552Rules are checked sequentially by rule number.
553Multiple rules can have the same number, in which case they are
554checked (and listed) according to the order in which they have
555been added.
556If a rule is entered without specifying a number, the kernel will
557assign one in such a way that the rule becomes the last one
558before the
559.Em default
560rule.
561Automatic rule numbers are assigned by incrementing the last
562non-default rule number by the value of the sysctl variable
563.Ar net.inet.ip.fw.autoinc_step
564which defaults to 100.
565If this is not possible (e.g.\& because we would go beyond the
566maximum allowed rule number), the number of the last
567non-default value is used instead.
568.It Cm set Ar set_number
569Each rule is associated with a
570.Ar set_number
571in the range 0..31.
572Sets can be individually disabled and enabled, so this parameter
573is of fundamental importance for atomic ruleset manipulation.
574It can be also used to simplify deletion of groups of rules.
575If a rule is entered without specifying a set number,
576set 0 will be used.
577.br
578Set 31 is special in that it cannot be disabled,
579and rules in set 31 are not deleted by the
580.Nm ipfw flush
581command (but you can delete them with the
582.Nm ipfw delete set 31
583command).
584Set 31 is also used for the
585.Em default
586rule.
587.It Cm prob Ar match_probability
588A match is only declared with the specified probability
589(floating point number between 0 and 1).
590This can be useful for a number of applications such as
591random packet drop or
592(in conjunction with
593.Nm dummynet )
594to simulate the effect of multiple paths leading to out-of-order
595packet delivery.
596.Pp
597Note: this condition is checked before any other condition, including
598ones such as keep-state or check-state which might have side effects.
599.It Cm log Op Cm logamount Ar number
600Packets matching a rule with the
601.Cm log
602keyword will be made available for logging in two ways:
603if the sysctl variable
604.Va net.inet.ip.fw.verbose
605is set to 0 (default), one can use
606.Xr bpf 4
607attached to the
608.Li ipfw0
609pseudo interface.
610This pseudo interface can be created after a boot
611manually by using the following command:
612.Bd -literal -offset indent
613# ifconfig ipfw0 create
614.Ed
615.Pp
616Or, automatically at boot time by adding the following
617line to the
618.Xr rc.conf 5
619file:
620.Bd -literal -offset indent
621firewall_logif="YES"
622.Ed
623.Pp
624There is no overhead if no
625.Xr bpf 4
626is attached to the pseudo interface.
627.Pp
628If
629.Va net.inet.ip.fw.verbose
630is set to 1, packets will be logged to
631.Xr syslogd 8
632with a
633.Dv LOG_SECURITY
634facility up to a maximum of
635.Cm logamount
636packets.
637If no
638.Cm logamount
639is specified, the limit is taken from the sysctl variable
640.Va net.inet.ip.fw.verbose_limit .
641In both cases, a value of 0 means unlimited logging.
642.Pp
643Once the limit is reached, logging can be re-enabled by
644clearing the logging counter or the packet counter for that entry, see the
645.Cm resetlog
646command.
647.Pp
648Note: logging is done after all other packet matching conditions
649have been successfully verified, and before performing the final
650action (accept, deny, etc.) on the packet.
651.It Cm tag Ar number
652When a packet matches a rule with the
653.Cm tag
654keyword, the numeric tag for the given
655.Ar number
656in the range 1..65534 will be attached to the packet.
657The tag acts as an internal marker (it is not sent out over
658the wire) that can be used to identify these packets later on.
659This can be used, for example, to provide trust between interfaces
660and to start doing policy-based filtering.
661A packet can have multiple tags at the same time.
662Tags are "sticky", meaning once a tag is applied to a packet by a
663matching rule it exists until explicit removal.
664Tags are kept with the packet everywhere within the kernel, but are
665lost when packet leaves the kernel, for example, on transmitting
666packet out to the network or sending packet to a
667.Xr divert 4
668socket.
669.Pp
670To check for previously applied tags, use the
671.Cm tagged
672rule option.
673To delete previously applied tag, use the
674.Cm untag
675keyword.
676.Pp
677Note: since tags are kept with the packet everywhere in kernelspace,
678they can be set and unset anywhere in the kernel network subsystem
679(using the
680.Xr mbuf_tags 9
681facility), not only by means of the
682.Xr ipfw 4
683.Cm tag
684and
685.Cm untag
686keywords.
687For example, there can be a specialized
688.Xr netgraph 4
689node doing traffic analyzing and tagging for later inspecting
690in firewall.
691.It Cm untag Ar number
692When a packet matches a rule with the
693.Cm untag
694keyword, the tag with the number
695.Ar number
696is searched among the tags attached to this packet and,
697if found, removed from it.
698Other tags bound to packet, if present, are left untouched.
699.It Cm altq Ar queue
700When a packet matches a rule with the
701.Cm altq
702keyword, the ALTQ identifier for the given
703.Ar queue
704(see
705.Xr altq 4 )
706will be attached.
707Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
708and not being rejected or going to divert sockets.
709Note that if there is insufficient memory at the time the packet is
710processed, it will not be tagged, so it is wise to make your ALTQ
711"default" queue policy account for this.
712If multiple
713.Cm altq
714rules match a single packet, only the first one adds the ALTQ classification
715tag.
716In doing so, traffic may be shaped by using
717.Cm count Cm altq Ar queue
718rules for classification early in the ruleset, then later applying
719the filtering decision.
720For example,
721.Cm check-state
722and
723.Cm keep-state
724rules may come later and provide the actual filtering decisions in
725addition to the fallback ALTQ tag.
726.Pp
727You must run
728.Xr pfctl 8
729to set up the queues before IPFW will be able to look them up by name,
730and if the ALTQ disciplines are rearranged, the rules in containing the
731queue identifiers in the kernel will likely have gone stale and need
732to be reloaded.
733Stale queue identifiers will probably result in misclassification.
734.Pp
735All system ALTQ processing can be turned on or off via
736.Nm
737.Cm enable Ar altq
738and
739.Nm
740.Cm disable Ar altq .
741The usage of
742.Va net.inet.ip.fw.one_pass
743is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
744always after adding an ALTQ tag.
745.El
746.Ss RULE ACTIONS
747A rule can be associated with one of the following actions, which
748will be executed when the packet matches the body of the rule.
749.Bl -tag -width indent
750.It Cm allow | accept | pass | permit
751Allow packets that match rule.
752The search terminates.
753.It Cm check-state Op Ar flowname | Cm any
754Checks the packet against the dynamic ruleset.
755If a match is found, execute the action associated with
756the rule which generated this dynamic rule, otherwise
757move to the next rule.
758.br
759.Cm Check-state
760rules do not have a body.
761If no
762.Cm check-state
763rule is found, the dynamic ruleset is checked at the first
764.Cm keep-state
765or
766.Cm limit
767rule.
768The
769.Ar flowname
770is symbolic name assigned to dynamic rule by
771.Cm keep-state
772opcode.
773The special flowname
774.Cm any
775can be used to ignore states flowname when matching.
776The
777.Cm default
778keyword is special name used for compatibility with old rulesets.
779.It Cm count
780Update counters for all packets that match rule.
781The search continues with the next rule.
782.It Cm deny | drop
783Discard packets that match this rule.
784The search terminates.
785.It Cm divert Ar port
786Divert packets that match this rule to the
787.Xr divert 4
788socket bound to port
789.Ar port .
790The search terminates.
791.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
792Change the next-hop on matching packets to
793.Ar ipaddr ,
794which can be an IP address or a host name.
795For IPv4, the next hop can also be supplied by the last table
796looked up for the packet by using the
797.Cm tablearg
798keyword instead of an explicit address.
799The search terminates if this rule matches.
800.Pp
801If
802.Ar ipaddr
803is a local address, then matching packets will be forwarded to
804.Ar port
805(or the port number in the packet if one is not specified in the rule)
806on the local machine.
807.br
808If
809.Ar ipaddr
810is not a local address, then the port number
811(if specified) is ignored, and the packet will be
812forwarded to the remote address, using the route as found in
813the local routing table for that IP.
814.br
815A
816.Ar fwd
817rule will not match layer-2 packets (those received
818on ether_input, ether_output, or bridged).
819.br
820The
821.Cm fwd
822action does not change the contents of the packet at all.
823In particular, the destination address remains unmodified, so
824packets forwarded to another system will usually be rejected by that system
825unless there is a matching rule on that system to capture them.
826For packets forwarded locally,
827the local address of the socket will be
828set to the original destination address of the packet.
829This makes the
830.Xr netstat 1
831entry look rather weird but is intended for
832use with transparent proxy servers.
833.It Cm nat Ar nat_nr | tablearg
834Pass packet to a
835nat instance
836(for network address translation, address redirect, etc.):
837see the
838.Sx NETWORK ADDRESS TRANSLATION (NAT)
839Section for further information.
840.It Cm nptv6 Ar name
841Pass packet to a NPTv6 instance (for IPv6-to-IPv6 network prefix translation):
842see the
843.Sx IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
844Section for further information.
845.It Cm pipe Ar pipe_nr
846Pass packet to a
847.Nm dummynet
848.Dq pipe
849(for bandwidth limitation, delay, etc.).
850See the
851.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
852Section for further information.
853The search terminates; however, on exit from the pipe and if
854the
855.Xr sysctl 8
856variable
857.Va net.inet.ip.fw.one_pass
858is not set, the packet is passed again to the firewall code
859starting from the next rule.
860.It Cm queue Ar queue_nr
861Pass packet to a
862.Nm dummynet
863.Dq queue
864(for bandwidth limitation using WF2Q+).
865.It Cm reject
866(Deprecated).
867Synonym for
868.Cm unreach host .
869.It Cm reset
870Discard packets that match this rule, and if the
871packet is a TCP packet, try to send a TCP reset (RST) notice.
872The search terminates.
873.It Cm reset6
874Discard packets that match this rule, and if the
875packet is a TCP packet, try to send a TCP reset (RST) notice.
876The search terminates.
877.It Cm skipto Ar number | tablearg
878Skip all subsequent rules numbered less than
879.Ar number .
880The search continues with the first rule numbered
881.Ar number
882or higher.
883It is possible to use the
884.Cm tablearg
885keyword with a skipto for a
886.Em computed
887skipto. Skipto may work either in O(log(N)) or in O(1) depending
888on amount of memory and/or sysctl variables.
889See the
890.Sx SYSCTL VARIABLES
891section for more details.
892.It Cm call Ar number | tablearg
893The current rule number is saved in the internal stack and
894ruleset processing continues with the first rule numbered
895.Ar number
896or higher.
897If later a rule with the
898.Cm return
899action is encountered, the processing returns to the first rule
900with number of this
901.Cm call
902rule plus one or higher
903(the same behaviour as with packets returning from
904.Xr divert 4
905socket after a
906.Cm divert
907action).
908This could be used to make somewhat like an assembly language
909.Dq subroutine
910calls to rules with common checks for different interfaces, etc.
911.Pp
912Rule with any number could be called, not just forward jumps as with
913.Cm skipto .
914So, to prevent endless loops in case of mistakes, both
915.Cm call
916and
917.Cm return
918actions don't do any jumps and simply go to the next rule if memory
919cannot be allocated or stack overflowed/underflowed.
920.Pp
921Internally stack for rule numbers is implemented using
922.Xr mbuf_tags 9
923facility and currently has size of 16 entries.
924As mbuf tags are lost when packet leaves the kernel,
925.Cm divert
926should not be used in subroutines to avoid endless loops
927and other undesired effects.
928.It Cm return
929Takes rule number saved to internal stack by the last
930.Cm call
931action and returns ruleset processing to the first rule
932with number greater than number of corresponding
933.Cm call
934rule.
935See description of the
936.Cm call
937action for more details.
938.Pp
939Note that
940.Cm return
941rules usually end a
942.Dq subroutine
943and thus are unconditional, but
944.Nm
945command-line utility currently requires every action except
946.Cm check-state
947to have body.
948While it is sometimes useful to return only on some packets,
949usually you want to print just
950.Dq return
951for readability.
952A workaround for this is to use new syntax and
953.Fl c
954switch:
955.Bd -literal -offset indent
956# Add a rule without actual body
957ipfw add 2999 return via any
958
959# List rules without "from any to any" part
960ipfw -c list
961.Ed
962.Pp
963This cosmetic annoyance may be fixed in future releases.
964.It Cm tee Ar port
965Send a copy of packets matching this rule to the
966.Xr divert 4
967socket bound to port
968.Ar port .
969The search continues with the next rule.
970.It Cm unreach Ar code
971Discard packets that match this rule, and try to send an ICMP
972unreachable notice with code
973.Ar code ,
974where
975.Ar code
976is a number from 0 to 255, or one of these aliases:
977.Cm net , host , protocol , port ,
978.Cm needfrag , srcfail , net-unknown , host-unknown ,
979.Cm isolated , net-prohib , host-prohib , tosnet ,
980.Cm toshost , filter-prohib , host-precedence
981or
982.Cm precedence-cutoff .
983The search terminates.
984.It Cm unreach6 Ar code
985Discard packets that match this rule, and try to send an ICMPv6
986unreachable notice with code
987.Ar code ,
988where
989.Ar code
990is a number from 0, 1, 3 or 4, or one of these aliases:
991.Cm no-route, admin-prohib, address
992or
993.Cm port .
994The search terminates.
995.It Cm netgraph Ar cookie
996Divert packet into netgraph with given
997.Ar cookie .
998The search terminates.
999If packet is later returned from netgraph it is either
1000accepted or continues with the next rule, depending on
1001.Va net.inet.ip.fw.one_pass
1002sysctl variable.
1003.It Cm ngtee Ar cookie
1004A copy of packet is diverted into netgraph, original
1005packet continues with the next rule.
1006See
1007.Xr ng_ipfw 4
1008for more information on
1009.Cm netgraph
1010and
1011.Cm ngtee
1012actions.
1013.It Cm setfib Ar fibnum | tablearg
1014The packet is tagged so as to use the FIB (routing table)
1015.Ar fibnum
1016in any subsequent forwarding decisions.
1017In the current implementation, this is limited to the values 0 through 15, see
1018.Xr setfib 2 .
1019Processing continues at the next rule.
1020It is possible to use the
1021.Cm tablearg
1022keyword with setfib.
1023If the tablearg value is not within the compiled range of fibs,
1024the packet's fib is set to 0.
1025.It Cm setdscp Ar DSCP | number | tablearg
1026Set specified DiffServ codepoint for an IPv4/IPv6 packet.
1027Processing continues at the next rule.
1028Supported values are:
1029.Pp
1030.Cm CS0
1031.Pq Dv 000000 ,
1032.Cm CS1
1033.Pq Dv 001000 ,
1034.Cm CS2
1035.Pq Dv 010000 ,
1036.Cm CS3
1037.Pq Dv 011000 ,
1038.Cm CS4
1039.Pq Dv 100000 ,
1040.Cm CS5
1041.Pq Dv 101000 ,
1042.Cm CS6
1043.Pq Dv 110000 ,
1044.Cm CS7
1045.Pq Dv 111000 ,
1046.Cm AF11
1047.Pq Dv 001010 ,
1048.Cm AF12
1049.Pq Dv 001100 ,
1050.Cm AF13
1051.Pq Dv 001110 ,
1052.Cm AF21
1053.Pq Dv 010010 ,
1054.Cm AF22
1055.Pq Dv 010100 ,
1056.Cm AF23
1057.Pq Dv 010110 ,
1058.Cm AF31
1059.Pq Dv 011010 ,
1060.Cm AF32
1061.Pq Dv 011100 ,
1062.Cm AF33
1063.Pq Dv 011110 ,
1064.Cm AF41
1065.Pq Dv 100010 ,
1066.Cm AF42
1067.Pq Dv 100100 ,
1068.Cm AF43
1069.Pq Dv 100110 ,
1070.Cm EF
1071.Pq Dv 101110 ,
1072.Cm BE
1073.Pq Dv 000000 .
1074Additionally, DSCP value can be specified by number (0..64).
1075It is also possible to use the
1076.Cm tablearg
1077keyword with setdscp.
1078If the tablearg value is not within the 0..64 range, lower 6 bits of supplied
1079value are used.
1080.It Cm reass
1081Queue and reassemble IP fragments.
1082If the packet is not fragmented, counters are updated and
1083processing continues with the next rule.
1084If the packet is the last logical fragment, the packet is reassembled and, if
1085.Va net.inet.ip.fw.one_pass
1086is set to 0, processing continues with the next rule.
1087Otherwise, the packet is allowed to pass and the search terminates.
1088If the packet is a fragment in the middle of a logical group of fragments,
1089it is consumed and
1090processing stops immediately.
1091.Pp
1092Fragment handling can be tuned via
1093.Va net.inet.ip.maxfragpackets
1094and
1095.Va net.inet.ip.maxfragsperpacket
1096which limit, respectively, the maximum number of processable
1097fragments (default: 800) and
1098the maximum number of fragments per packet (default: 16).
1099.Pp
1100NOTA BENE: since fragments do not contain port numbers,
1101they should be avoided with the
1102.Nm reass
1103rule.
1104Alternatively, direction-based (like
1105.Nm in
1106/
1107.Nm out
1108) and source-based (like
1109.Nm via
1110) match patterns can be used to select fragments.
1111.Pp
1112Usually a simple rule like:
1113.Bd -literal -offset indent
1114# reassemble incoming fragments
1115ipfw add reass all from any to any in
1116.Ed
1117.Pp
1118is all you need at the beginning of your ruleset.
1119.El
1120.Ss RULE BODY
1121The body of a rule contains zero or more patterns (such as
1122specific source and destination addresses or ports,
1123protocol options, incoming or outgoing interfaces, etc.)
1124that the packet must match in order to be recognised.
1125In general, the patterns are connected by (implicit)
1126.Cm and
1127operators -- i.e., all must match in order for the
1128rule to match.
1129Individual patterns can be prefixed by the
1130.Cm not
1131operator to reverse the result of the match, as in
1132.Pp
1133.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
1134.Pp
1135Additionally, sets of alternative match patterns
1136.Pq Em or-blocks
1137can be constructed by putting the patterns in
1138lists enclosed between parentheses ( ) or braces { }, and
1139using the
1140.Cm or
1141operator as follows:
1142.Pp
1143.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
1144.Pp
1145Only one level of parentheses is allowed.
1146Beware that most shells have special meanings for parentheses
1147or braces, so it is advisable to put a backslash \\ in front of them
1148to prevent such interpretations.
1149.Pp
1150The body of a rule must in general include a source and destination
1151address specifier.
1152The keyword
1153.Ar any
1154can be used in various places to specify that the content of
1155a required field is irrelevant.
1156.Pp
1157The rule body has the following format:
1158.Bd -ragged -offset indent
1159.Op Ar proto Cm from Ar src Cm to Ar dst
1160.Op Ar options
1161.Ed
1162.Pp
1163The first part (proto from src to dst) is for backward
1164compatibility with earlier versions of
1165.Fx .
1166In modern
1167.Fx
1168any match pattern (including MAC headers, IP protocols,
1169addresses and ports) can be specified in the
1170.Ar options
1171section.
1172.Pp
1173Rule fields have the following meaning:
1174.Bl -tag -width indent
1175.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
1176.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
1177An IP protocol specified by number or name
1178(for a complete list see
1179.Pa /etc/protocols ) ,
1180or one of the following keywords:
1181.Bl -tag -width indent
1182.It Cm ip4 | ipv4
1183Matches IPv4 packets.
1184.It Cm ip6 | ipv6
1185Matches IPv6 packets.
1186.It Cm ip | all
1187Matches any packet.
1188.El
1189.Pp
1190The
1191.Cm ipv6
1192in
1193.Cm proto
1194option will be treated as inner protocol.
1195And, the
1196.Cm ipv4
1197is not available in
1198.Cm proto
1199option.
1200.Pp
1201The
1202.Cm { Ar protocol Cm or ... }
1203format (an
1204.Em or-block )
1205is provided for convenience only but its use is deprecated.
1206.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1207An address (or a list, see below)
1208optionally followed by
1209.Ar ports
1210specifiers.
1211.Pp
1212The second format
1213.Em ( or-block
1214with multiple addresses) is provided for convenience only and
1215its use is discouraged.
1216.It Ar addr : Oo Cm not Oc Bro
1217.Cm any | me | me6 |
1218.Cm table Ns Pq Ar name Ns Op , Ns Ar value
1219.Ar | addr-list | addr-set
1220.Brc
1221.Bl -tag -width indent
1222.It Cm any
1223matches any IP address.
1224.It Cm me
1225matches any IP address configured on an interface in the system.
1226.It Cm me6
1227matches any IPv6 address configured on an interface in the system.
1228The address list is evaluated at the time the packet is
1229analysed.
1230.It Cm table Ns Pq Ar name Ns Op , Ns Ar value
1231Matches any IPv4 or IPv6 address for which an entry exists in the lookup table
1232.Ar number .
1233If an optional 32-bit unsigned
1234.Ar value
1235is also specified, an entry will match only if it has this value.
1236See the
1237.Sx LOOKUP TABLES
1238section below for more information on lookup tables.
1239.El
1240.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1241.It Ar ip-addr :
1242A host or subnet address specified in one of the following ways:
1243.Bl -tag -width indent
1244.It Ar numeric-ip | hostname
1245Matches a single IPv4 address, specified as dotted-quad or a hostname.
1246Hostnames are resolved at the time the rule is added to the firewall list.
1247.It Ar addr Ns / Ns Ar masklen
1248Matches all addresses with base
1249.Ar addr
1250(specified as an IP address, a network number, or a hostname)
1251and mask width of
1252.Cm masklen
1253bits.
1254As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1255all IP numbers from 1.2.3.0 to 1.2.3.127 .
1256.It Ar addr Ns : Ns Ar mask
1257Matches all addresses with base
1258.Ar addr
1259(specified as an IP address, a network number, or a hostname)
1260and the mask of
1261.Ar mask ,
1262specified as a dotted quad.
1263As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
12641.*.3.*.
1265This form is advised only for non-contiguous
1266masks.
1267It is better to resort to the
1268.Ar addr Ns / Ns Ar masklen
1269format for contiguous masks, which is more compact and less
1270error-prone.
1271.El
1272.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1273.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1274Matches all addresses with base address
1275.Ar addr
1276(specified as an IP address, a network number, or a hostname)
1277and whose last byte is in the list between braces { } .
1278Note that there must be no spaces between braces and
1279numbers (spaces after commas are allowed).
1280Elements of the list can be specified as single entries
1281or ranges.
1282The
1283.Ar masklen
1284field is used to limit the size of the set of addresses,
1285and can have any value between 24 and 32.
1286If not specified,
1287it will be assumed as 24.
1288.br
1289This format is particularly useful to handle sparse address sets
1290within a single rule.
1291Because the matching occurs using a
1292bitmask, it takes constant time and dramatically reduces
1293the complexity of rulesets.
1294.br
1295As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1296or 1.2.3.0/24{128,35-55,89}
1297will match the following IP addresses:
1298.br
12991.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1300.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1301.It Ar ip6-addr :
1302A host or subnet specified one of the following ways:
1303.Bl -tag -width indent
1304.It Ar numeric-ip | hostname
1305Matches a single IPv6 address as allowed by
1306.Xr inet_pton 3
1307or a hostname.
1308Hostnames are resolved at the time the rule is added to the firewall
1309list.
1310.It Ar addr Ns / Ns Ar masklen
1311Matches all IPv6 addresses with base
1312.Ar addr
1313(specified as allowed by
1314.Xr inet_pton
1315or a hostname)
1316and mask width of
1317.Cm masklen
1318bits.
1319.El
1320.Pp
1321No support for sets of IPv6 addresses is provided because IPv6 addresses
1322are typically random past the initial prefix.
1323.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1324For protocols which support port numbers (such as TCP and UDP), optional
1325.Cm ports
1326may be specified as one or more ports or port ranges, separated
1327by commas but no spaces, and an optional
1328.Cm not
1329operator.
1330The
1331.Ql \&-
1332notation specifies a range of ports (including boundaries).
1333.Pp
1334Service names (from
1335.Pa /etc/services )
1336may be used instead of numeric port values.
1337The length of the port list is limited to 30 ports or ranges,
1338though one can specify larger ranges by using an
1339.Em or-block
1340in the
1341.Cm options
1342section of the rule.
1343.Pp
1344A backslash
1345.Pq Ql \e
1346can be used to escape the dash
1347.Pq Ql -
1348character in a service name (from a shell, the backslash must be
1349typed twice to avoid the shell itself interpreting it as an escape
1350character).
1351.Pp
1352.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1353.Pp
1354Fragmented packets which have a non-zero offset (i.e., not the first
1355fragment) will never match a rule which has one or more port
1356specifications.
1357See the
1358.Cm frag
1359option for details on matching fragmented packets.
1360.El
1361.Ss RULE OPTIONS (MATCH PATTERNS)
1362Additional match patterns can be used within
1363rules.
1364Zero or more of these so-called
1365.Em options
1366can be present in a rule, optionally prefixed by the
1367.Cm not
1368operand, and possibly grouped into
1369.Em or-blocks .
1370.Pp
1371The following match patterns can be used (listed in alphabetical order):
1372.Bl -tag -width indent
1373.It Cm // this is a comment.
1374Inserts the specified text as a comment in the rule.
1375Everything following // is considered as a comment and stored in the rule.
1376You can have comment-only rules, which are listed as having a
1377.Cm count
1378action followed by the comment.
1379.It Cm bridged
1380Alias for
1381.Cm layer2 .
1382.It Cm diverted
1383Matches only packets generated by a divert socket.
1384.It Cm diverted-loopback
1385Matches only packets coming from a divert socket back into the IP stack
1386input for delivery.
1387.It Cm diverted-output
1388Matches only packets going from a divert socket back outward to the IP
1389stack output for delivery.
1390.It Cm dst-ip Ar ip-address
1391Matches IPv4 packets whose destination IP is one of the address(es)
1392specified as argument.
1393.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1394Matches IPv6 packets whose destination IP is one of the address(es)
1395specified as argument.
1396.It Cm dst-port Ar ports
1397Matches IP packets whose destination port is one of the port(s)
1398specified as argument.
1399.It Cm established
1400Matches TCP packets that have the RST or ACK bits set.
1401.It Cm ext6hdr Ar header
1402Matches IPv6 packets containing the extended header given by
1403.Ar header .
1404Supported headers are:
1405.Pp
1406Fragment,
1407.Pq Cm frag ,
1408Hop-to-hop options
1409.Pq Cm hopopt ,
1410any type of Routing Header
1411.Pq Cm route ,
1412Source routing Routing Header Type 0
1413.Pq Cm rthdr0 ,
1414Mobile IPv6 Routing Header Type 2
1415.Pq Cm rthdr2 ,
1416Destination options
1417.Pq Cm dstopt ,
1418IPSec authentication headers
1419.Pq Cm ah ,
1420and IPsec encapsulated security payload headers
1421.Pq Cm esp .
1422.It Cm fib Ar fibnum
1423Matches a packet that has been tagged to use
1424the given FIB (routing table) number.
1425.It Cm flow Ar table Ns Pq Ar name Ns Op , Ns Ar value
1426Search for the flow entry in lookup table
1427.Ar name .
1428If not found, the match fails.
1429Otherwise, the match succeeds and
1430.Cm tablearg
1431is set to the value extracted from the table.
1432.Pp
1433This option can be useful to quickly dispatch traffic based on
1434certain packet fields.
1435See the
1436.Sx LOOKUP TABLES
1437section below for more information on lookup tables.
1438.It Cm flow-id Ar labels
1439Matches IPv6 packets containing any of the flow labels given in
1440.Ar labels .
1441.Ar labels
1442is a comma separated list of numeric flow labels.
1443.It Cm frag
1444Matches packets that are fragments and not the first
1445fragment of an IP datagram.
1446Note that these packets will not have
1447the next protocol header (e.g.\& TCP, UDP) so options that look into
1448these headers cannot match.
1449.It Cm gid Ar group
1450Matches all TCP or UDP packets sent by or received for a
1451.Ar group .
1452A
1453.Ar group
1454may be specified by name or number.
1455.It Cm jail Ar prisonID
1456Matches all TCP or UDP packets sent by or received for the
1457jail whos prison ID is
1458.Ar prisonID .
1459.It Cm icmptypes Ar types
1460Matches ICMP packets whose ICMP type is in the list
1461.Ar types .
1462The list may be specified as any combination of
1463individual types (numeric) separated by commas.
1464.Em Ranges are not allowed .
1465The supported ICMP types are:
1466.Pp
1467echo reply
1468.Pq Cm 0 ,
1469destination unreachable
1470.Pq Cm 3 ,
1471source quench
1472.Pq Cm 4 ,
1473redirect
1474.Pq Cm 5 ,
1475echo request
1476.Pq Cm 8 ,
1477router advertisement
1478.Pq Cm 9 ,
1479router solicitation
1480.Pq Cm 10 ,
1481time-to-live exceeded
1482.Pq Cm 11 ,
1483IP header bad
1484.Pq Cm 12 ,
1485timestamp request
1486.Pq Cm 13 ,
1487timestamp reply
1488.Pq Cm 14 ,
1489information request
1490.Pq Cm 15 ,
1491information reply
1492.Pq Cm 16 ,
1493address mask request
1494.Pq Cm 17
1495and address mask reply
1496.Pq Cm 18 .
1497.It Cm icmp6types Ar types
1498Matches ICMP6 packets whose ICMP6 type is in the list of
1499.Ar types .
1500The list may be specified as any combination of
1501individual types (numeric) separated by commas.
1502.Em Ranges are not allowed .
1503.It Cm in | out
1504Matches incoming or outgoing packets, respectively.
1505.Cm in
1506and
1507.Cm out
1508are mutually exclusive (in fact,
1509.Cm out
1510is implemented as
1511.Cm not in Ns No ).
1512.It Cm ipid Ar id-list
1513Matches IPv4 packets whose
1514.Cm ip_id
1515field has value included in
1516.Ar id-list ,
1517which is either a single value or a list of values or ranges
1518specified in the same way as
1519.Ar ports .
1520.It Cm iplen Ar len-list
1521Matches IP packets whose total length, including header and data, is
1522in the set
1523.Ar len-list ,
1524which is either a single value or a list of values or ranges
1525specified in the same way as
1526.Ar ports .
1527.It Cm ipoptions Ar spec
1528Matches packets whose IPv4 header contains the comma separated list of
1529options specified in
1530.Ar spec .
1531The supported IP options are:
1532.Pp
1533.Cm ssrr
1534(strict source route),
1535.Cm lsrr
1536(loose source route),
1537.Cm rr
1538(record packet route) and
1539.Cm ts
1540(timestamp).
1541The absence of a particular option may be denoted
1542with a
1543.Ql \&! .
1544.It Cm ipprecedence Ar precedence
1545Matches IPv4 packets whose precedence field is equal to
1546.Ar precedence .
1547.It Cm ipsec
1548Matches packets that have IPSEC history associated with them
1549(i.e., the packet comes encapsulated in IPSEC, the kernel
1550has IPSEC support and IPSEC_FILTERTUNNEL option, and can correctly
1551decapsulate it).
1552.Pp
1553Note that specifying
1554.Cm ipsec
1555is different from specifying
1556.Cm proto Ar ipsec
1557as the latter will only look at the specific IP protocol field,
1558irrespective of IPSEC kernel support and the validity of the IPSEC data.
1559.Pp
1560Further note that this flag is silently ignored in kernels without
1561IPSEC support.
1562It does not affect rule processing when given and the
1563rules are handled as if with no
1564.Cm ipsec
1565flag.
1566.It Cm iptos Ar spec
1567Matches IPv4 packets whose
1568.Cm tos
1569field contains the comma separated list of
1570service types specified in
1571.Ar spec .
1572The supported IP types of service are:
1573.Pp
1574.Cm lowdelay
1575.Pq Dv IPTOS_LOWDELAY ,
1576.Cm throughput
1577.Pq Dv IPTOS_THROUGHPUT ,
1578.Cm reliability
1579.Pq Dv IPTOS_RELIABILITY ,
1580.Cm mincost
1581.Pq Dv IPTOS_MINCOST ,
1582.Cm congestion
1583.Pq Dv IPTOS_ECN_CE .
1584The absence of a particular type may be denoted
1585with a
1586.Ql \&! .
1587.It Cm dscp spec Ns Op , Ns Ar spec
1588Matches IPv4/IPv6 packets whose
1589.Cm DS
1590field value is contained in
1591.Ar spec
1592mask.
1593Multiple values can be specified via
1594the comma separated list.
1595Value can be one of keywords used in
1596.Cm setdscp
1597action or exact number.
1598.It Cm ipttl Ar ttl-list
1599Matches IPv4 packets whose time to live is included in
1600.Ar ttl-list ,
1601which is either a single value or a list of values or ranges
1602specified in the same way as
1603.Ar ports .
1604.It Cm ipversion Ar ver
1605Matches IP packets whose IP version field is
1606.Ar ver .
1607.It Cm keep-state Op Ar flowname
1608Upon a match, the firewall will create a dynamic rule, whose
1609default behaviour is to match bidirectional traffic between
1610source and destination IP/port using the same protocol.
1611The rule has a limited lifetime (controlled by a set of
1612.Xr sysctl 8
1613variables), and the lifetime is refreshed every time a matching
1614packet is found.
1615The
1616.Ar flowname
1617is used to assign additional to addresses, ports and protocol parameter
1618to dynamic rule. It can be used for more accurate matching by
1619.Cm check-state
1620rule.
1621The
1622.Cm default
1623keyword is special name used for compatibility with old rulesets.
1624.It Cm layer2
1625Matches only layer2 packets, i.e., those passed to
1626.Nm
1627from ether_demux() and ether_output_frame().
1628.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N Op Ar flowname
1629The firewall will only allow
1630.Ar N
1631connections with the same
1632set of parameters as specified in the rule.
1633One or more
1634of source and destination addresses and ports can be
1635specified.
1636.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar name
1637Search an entry in lookup table
1638.Ar name
1639that matches the field specified as argument.
1640If not found, the match fails.
1641Otherwise, the match succeeds and
1642.Cm tablearg
1643is set to the value extracted from the table.
1644.Pp
1645This option can be useful to quickly dispatch traffic based on
1646certain packet fields.
1647See the
1648.Sx LOOKUP TABLES
1649section below for more information on lookup tables.
1650.It Cm { MAC | mac } Ar dst-mac src-mac
1651Match packets with a given
1652.Ar dst-mac
1653and
1654.Ar src-mac
1655addresses, specified as the
1656.Cm any
1657keyword (matching any MAC address), or six groups of hex digits
1658separated by colons,
1659and optionally followed by a mask indicating the significant bits.
1660The mask may be specified using either of the following methods:
1661.Bl -enum -width indent
1662.It
1663A slash
1664.Pq /
1665followed by the number of significant bits.
1666For example, an address with 33 significant bits could be specified as:
1667.Pp
1668.Dl "MAC 10:20:30:40:50:60/33 any"
1669.It
1670An ampersand
1671.Pq &
1672followed by a bitmask specified as six groups of hex digits separated
1673by colons.
1674For example, an address in which the last 16 bits are significant could
1675be specified as:
1676.Pp
1677.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1678.Pp
1679Note that the ampersand character has a special meaning in many shells
1680and should generally be escaped.
1681.El
1682Note that the order of MAC addresses (destination first,
1683source second) is
1684the same as on the wire, but the opposite of the one used for
1685IP addresses.
1686.It Cm mac-type Ar mac-type
1687Matches packets whose Ethernet Type field
1688corresponds to one of those specified as argument.
1689.Ar mac-type
1690is specified in the same way as
1691.Cm port numbers
1692(i.e., one or more comma-separated single values or ranges).
1693You can use symbolic names for known values such as
1694.Em vlan , ipv4, ipv6 .
1695Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1696and they are always printed as hexadecimal (unless the
1697.Cm -N
1698option is used, in which case symbolic resolution will be attempted).
1699.It Cm proto Ar protocol
1700Matches packets with the corresponding IP protocol.
1701.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
1702Matches packets received, transmitted or going through,
1703respectively, the interface specified by exact name
1704.Po Ar ifX Pc ,
1705by device name
1706.Po Ar if* Pc ,
1707by IP address, or through some interface.
1708Table
1709.Ar name
1710may be used to match interface by its kernel ifindex.
1711See the
1712.Sx LOOKUP TABLES
1713section below for more information on lookup tables.
1714.Pp
1715The
1716.Cm via
1717keyword causes the interface to always be checked.
1718If
1719.Cm recv
1720or
1721.Cm xmit
1722is used instead of
1723.Cm via ,
1724then only the receive or transmit interface (respectively)
1725is checked.
1726By specifying both, it is possible to match packets based on
1727both receive and transmit interface, e.g.:
1728.Pp
1729.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1730.Pp
1731The
1732.Cm recv
1733interface can be tested on either incoming or outgoing packets,
1734while the
1735.Cm xmit
1736interface can only be tested on outgoing packets.
1737So
1738.Cm out
1739is required (and
1740.Cm in
1741is invalid) whenever
1742.Cm xmit
1743is used.
1744.Pp
1745A packet might not have a receive or transmit interface: packets
1746originating from the local host have no receive interface,
1747while packets destined for the local host have no transmit
1748interface.
1749.It Cm setup
1750Matches TCP packets that have the SYN bit set but no ACK bit.
1751This is the short form of
1752.Dq Li tcpflags\ syn,!ack .
1753.It Cm sockarg
1754Matches packets that are associated to a local socket and
1755for which the SO_USER_COOKIE socket option has been set
1756to a non-zero value.
1757As a side effect, the value of the
1758option is made available as
1759.Cm tablearg
1760value, which in turn can be used as
1761.Cm skipto
1762or
1763.Cm pipe
1764number.
1765.It Cm src-ip Ar ip-address
1766Matches IPv4 packets whose source IP is one of the address(es)
1767specified as an argument.
1768.It Cm src-ip6 Ar ip6-address
1769Matches IPv6 packets whose source IP is one of the address(es)
1770specified as an argument.
1771.It Cm src-port Ar ports
1772Matches IP packets whose source port is one of the port(s)
1773specified as argument.
1774.It Cm tagged Ar tag-list
1775Matches packets whose tags are included in
1776.Ar tag-list ,
1777which is either a single value or a list of values or ranges
1778specified in the same way as
1779.Ar ports .
1780Tags can be applied to the packet using
1781.Cm tag
1782rule action parameter (see it's description for details on tags).
1783.It Cm tcpack Ar ack
1784TCP packets only.
1785Match if the TCP header acknowledgment number field is set to
1786.Ar ack .
1787.It Cm tcpdatalen Ar tcpdatalen-list
1788Matches TCP packets whose length of TCP data is
1789.Ar tcpdatalen-list ,
1790which is either a single value or a list of values or ranges
1791specified in the same way as
1792.Ar ports .
1793.It Cm tcpflags Ar spec
1794TCP packets only.
1795Match if the TCP header contains the comma separated list of
1796flags specified in
1797.Ar spec .
1798The supported TCP flags are:
1799.Pp
1800.Cm fin ,
1801.Cm syn ,
1802.Cm rst ,
1803.Cm psh ,
1804.Cm ack
1805and
1806.Cm urg .
1807The absence of a particular flag may be denoted
1808with a
1809.Ql \&! .
1810A rule which contains a
1811.Cm tcpflags
1812specification can never match a fragmented packet which has
1813a non-zero offset.
1814See the
1815.Cm frag
1816option for details on matching fragmented packets.
1817.It Cm tcpseq Ar seq
1818TCP packets only.
1819Match if the TCP header sequence number field is set to
1820.Ar seq .
1821.It Cm tcpwin Ar tcpwin-list
1822Matches TCP packets whose  header window field is set to
1823.Ar tcpwin-list ,
1824which is either a single value or a list of values or ranges
1825specified in the same way as
1826.Ar ports .
1827.It Cm tcpoptions Ar spec
1828TCP packets only.
1829Match if the TCP header contains the comma separated list of
1830options specified in
1831.Ar spec .
1832The supported TCP options are:
1833.Pp
1834.Cm mss
1835(maximum segment size),
1836.Cm window
1837(tcp window advertisement),
1838.Cm sack
1839(selective ack),
1840.Cm ts
1841(rfc1323 timestamp) and
1842.Cm cc
1843(rfc1644 t/tcp connection count).
1844The absence of a particular option may be denoted
1845with a
1846.Ql \&! .
1847.It Cm uid Ar user
1848Match all TCP or UDP packets sent by or received for a
1849.Ar user .
1850A
1851.Ar user
1852may be matched by name or identification number.
1853.It Cm verrevpath
1854For incoming packets,
1855a routing table lookup is done on the packet's source address.
1856If the interface on which the packet entered the system matches the
1857outgoing interface for the route,
1858the packet matches.
1859If the interfaces do not match up,
1860the packet does not match.
1861All outgoing packets or packets with no incoming interface match.
1862.Pp
1863The name and functionality of the option is intentionally similar to
1864the Cisco IOS command:
1865.Pp
1866.Dl ip verify unicast reverse-path
1867.Pp
1868This option can be used to make anti-spoofing rules to reject all
1869packets with source addresses not from this interface.
1870See also the option
1871.Cm antispoof .
1872.It Cm versrcreach
1873For incoming packets,
1874a routing table lookup is done on the packet's source address.
1875If a route to the source address exists, but not the default route
1876or a blackhole/reject route, the packet matches.
1877Otherwise, the packet does not match.
1878All outgoing packets match.
1879.Pp
1880The name and functionality of the option is intentionally similar to
1881the Cisco IOS command:
1882.Pp
1883.Dl ip verify unicast source reachable-via any
1884.Pp
1885This option can be used to make anti-spoofing rules to reject all
1886packets whose source address is unreachable.
1887.It Cm antispoof
1888For incoming packets, the packet's source address is checked if it
1889belongs to a directly connected network.
1890If the network is directly connected, then the interface the packet
1891came on in is compared to the interface the network is connected to.
1892When incoming interface and directly connected interface are not the
1893same, the packet does not match.
1894Otherwise, the packet does match.
1895All outgoing packets match.
1896.Pp
1897This option can be used to make anti-spoofing rules to reject all
1898packets that pretend to be from a directly connected network but do
1899not come in through that interface.
1900This option is similar to but more restricted than
1901.Cm verrevpath
1902because it engages only on packets with source addresses of directly
1903connected networks instead of all source addresses.
1904.El
1905.Sh LOOKUP TABLES
1906Lookup tables are useful to handle large sparse sets of
1907addresses or other search keys (e.g., ports, jail IDs, interface names).
1908In the rest of this section we will use the term ``key''.
1909Table name needs to match the following spec:
1910.Ar table-name .
1911Tables with the same name can be created in different
1912.Ar sets .
1913However, rule links to the tables in
1914.Ar set 0
1915by default.
1916This behavior can be controlled by
1917.Va net.inet.ip.fw.tables_sets
1918variable.
1919See the
1920.Sx SETS OF RULES
1921section for more information.
1922There may be up to 65535 different lookup tables.
1923.Pp
1924The following table types are supported:
1925.Bl -tag -width indent
1926.It Ar table-type : Ar addr | iface | number | flow
1927.It Ar table-key : Ar addr Ns Oo / Ns Ar masklen Oc | iface-name | number | flow-spec
1928.It Ar flow-spec : Ar flow-field Ns Op , Ns Ar flow-spec
1929.It Ar flow-field : src-ip | proto | src-port | dst-ip | dst-port
1930.It Cm addr
1931matches IPv4 or IPv6 address.
1932Each entry is represented by an
1933.Ar addr Ns Op / Ns Ar masklen
1934and will match all addresses with base
1935.Ar addr
1936(specified as an IPv4/IPv6 address, or a hostname) and mask width of
1937.Ar masklen
1938bits.
1939If
1940.Ar masklen
1941is not specified, it defaults to 32 for IPv4 and 128 for IPv6.
1942When looking up an IP address in a table, the most specific
1943entry will match.
1944.It Cm iface
1945matches interface names.
1946Each entry is represented by string treated as interface name.
1947Wildcards are not supported.
1948.It Cm number
1949maches protocol ports, uids/gids or jail IDs.
1950Each entry is represented by 32-bit unsigned integer.
1951Ranges are not supported.
1952.It Cm flow
1953Matches packet fields specified by
1954.Ar flow
1955type suboptions with table entries.
1956.El
1957.Pp
1958Tables require explicit creation via
1959.Cm create
1960before use.
1961.Pp
1962The following creation options are supported:
1963.Bl -tag -width indent
1964.It Ar create-options : Ar create-option | create-options
1965.It Ar create-option : Cm type Ar table-type | Cm valtype Ar value-mask | Cm algo Ar algo-desc |
1966.Cm limit Ar number | Cm locked
1967.It Cm type
1968Table key type.
1969.It Cm valtype
1970Table value mask.
1971.It Cm algo
1972Table algorithm to use (see below).
1973.It Cm limit
1974Maximum number of items that may be inserted into table.
1975.It Cm locked
1976Restrict any table modifications.
1977.El
1978.Pp
1979Some of these options may be modified later via
1980.Cm modify
1981keyword.
1982The following options can be changed:
1983.Bl -tag -width indent
1984.It Ar modify-options : Ar modify-option | modify-options
1985.It Ar modify-option : Cm limit Ar number
1986.It Cm limit
1987Alter maximum number of items that may be inserted into table.
1988.El
1989.Pp
1990Additionally, table can be locked or unlocked using
1991.Cm lock
1992or
1993.Cm unlock
1994commands.
1995.Pp
1996Tables of the same
1997.Ar type
1998can be swapped with each other using
1999.Cm swap Ar name
2000command.
2001Swap may fail if tables limits are set and data exchange
2002would result in limits hit.
2003Operation is performed atomically.
2004.Pp
2005One or more entries can be added to a table at once using
2006.Cm add
2007command.
2008Addition of all items are performed atomically.
2009By default, error in addition of one entry does not influence
2010addition of other entries. However, non-zero error code is returned
2011in that case.
2012Special
2013.Cm atomic
2014keyword may be specified before
2015.Cm add
2016to indicate all-or-none add request.
2017.Pp
2018One or more entries can be removed from a table at once using
2019.Cm delete
2020command.
2021By default, error in removal of one entry does not influence
2022removing of other entries. However, non-zero error code is returned
2023in that case.
2024.Pp
2025It may be possible to check what entry will be found on particular
2026.Ar table-key
2027using
2028.Cm lookup
2029.Ar table-key
2030command.
2031This functionality is optional and may be unsupported in some algorithms.
2032.Pp
2033The following operations can be performed on
2034.Ar one
2035or
2036.Cm all
2037tables:
2038.Bl -tag -width indent
2039.It Cm list
2040List all entries.
2041.It Cm flush
2042Removes all entries.
2043.It Cm info
2044Shows generic table information.
2045.It Cm detail
2046Shows generic table information and algo-specific data.
2047.El
2048.Pp
2049The following lookup algorithms are supported:
2050.Bl -tag -width indent
2051.It Ar algo-desc : algo-name | "algo-name algo-data"
2052.It Ar algo-name: Ar addr:radix | addr:hash | iface:array | number:array | flow:hash
2053.It Cm addr:radix
2054Separate Radix trees for IPv4 and IPv6, the same way as the routing table (see
2055.Xr route 4 ) .
2056Default choice for
2057.Ar addr
2058type.
2059.It Cm addr:hash
2060Separate auto-growing hashes for IPv4 and IPv6.
2061Accepts entries with the same mask length specified initially via
2062.Cm "addr:hash masks=/v4,/v6"
2063algorithm creation options.
2064Assume /32 and /128 masks by default.
2065Search removes host bits (according to mask) from supplied address and checks
2066resulting key in appropriate hash.
2067Mostly optimized for /64 and byte-ranged IPv6 masks.
2068.It Cm iface:array
2069Array storing sorted indexes for entries which are presented in the system.
2070Optimized for very fast lookup.
2071.It Cm number:array
2072Array storing sorted u32 numbers.
2073.It Cm flow:hash
2074Auto-growing hash storing flow entries.
2075Search calculates hash on required packet fields and searches for matching
2076entries in selected bucket.
2077.El
2078.Pp
2079The
2080.Cm tablearg
2081feature provides the ability to use a value, looked up in the table, as
2082the argument for a rule action, action parameter or rule option.
2083This can significantly reduce number of rules in some configurations.
2084If two tables are used in a rule, the result of the second (destination)
2085is used.
2086.Pp
2087Each record may hold one or more values according to
2088.Ar value-mask .
2089This mask is set on table creation via
2090.Cm valtype
2091option.
2092The following value types are supported:
2093.Bl -tag -width indent
2094.It Ar value-mask : Ar value-type Ns Op , Ns Ar value-mask
2095.It Ar value-type : Ar skipto | pipe | fib | nat | dscp | tag | divert |
2096.Ar netgraph | limit | ipv4
2097.It Cm skipto
2098rule number to jump to.
2099.It Cm pipe
2100Pipe number to use.
2101.It Cm fib
2102fib number to match/set.
2103.It Cm nat
2104nat number to jump to.
2105.It Cm dscp
2106dscp value to match/set.
2107.It Cm tag
2108tag number to match/set.
2109.It Cm divert
2110port number to divert traffic to.
2111.It Cm netgraph
2112hook number to move packet to.
2113.It Cm limit
2114maximum number of connections.
2115.It Cm ipv4
2116IPv4 nexthop to fwd packets to.
2117.It Cm ipv6
2118IPv6 nexthop to fwd packets to.
2119.El
2120.Pp
2121The
2122.Cm tablearg
2123argument can be used with the following actions:
2124.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto, setfib,
2125action parameters:
2126.Cm tag, untag,
2127rule options:
2128.Cm limit, tagged.
2129.Pp
2130When used with the
2131.Cm skipto
2132action, the user should be aware that the code will walk the ruleset
2133up to a rule equal to, or past, the given number.
2134.Pp
2135See the
2136.Sx EXAMPLES
2137Section for example usage of tables and the tablearg keyword.
2138.Sh SETS OF RULES
2139Each rule or table belongs to one of 32 different
2140.Em sets
2141, numbered 0 to 31.
2142Set 31 is reserved for the default rule.
2143.Pp
2144By default, rules or tables are put in set 0, unless you use the
2145.Cm set N
2146attribute when adding a new rule or table.
2147Sets can be individually and atomically enabled or disabled,
2148so this mechanism permits an easy way to store multiple configurations
2149of the firewall and quickly (and atomically) switch between them.
2150.Pp
2151By default, tables from set 0 are referenced when adding rule with
2152table opcodes regardless of rule set.
2153This behavior can be changed by setting
2154.Va net.inet.ip.fw.tables_set
2155variable to 1.
2156Rule's set will then be used for table references.
2157.Pp
2158The command to enable/disable sets is
2159.Bd -ragged -offset indent
2160.Nm
2161.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
2162.Ed
2163.Pp
2164where multiple
2165.Cm enable
2166or
2167.Cm disable
2168sections can be specified.
2169Command execution is atomic on all the sets specified in the command.
2170By default, all sets are enabled.
2171.Pp
2172When you disable a set, its rules behave as if they do not exist
2173in the firewall configuration, with only one exception:
2174.Bd -ragged -offset indent
2175dynamic rules created from a rule before it had been disabled
2176will still be active until they expire.
2177In order to delete
2178dynamic rules you have to explicitly delete the parent rule
2179which generated them.
2180.Ed
2181.Pp
2182The set number of rules can be changed with the command
2183.Bd -ragged -offset indent
2184.Nm
2185.Cm set move
2186.Brq Cm rule Ar rule-number | old-set
2187.Cm to Ar new-set
2188.Ed
2189.Pp
2190Also, you can atomically swap two rulesets with the command
2191.Bd -ragged -offset indent
2192.Nm
2193.Cm set swap Ar first-set second-set
2194.Ed
2195.Pp
2196See the
2197.Sx EXAMPLES
2198Section on some possible uses of sets of rules.
2199.Sh STATEFUL FIREWALL
2200Stateful operation is a way for the firewall to dynamically
2201create rules for specific flows when packets that
2202match a given pattern are detected.
2203Support for stateful
2204operation comes through the
2205.Cm check-state , keep-state
2206and
2207.Cm limit
2208options of
2209.Nm rules .
2210.Pp
2211Dynamic rules are created when a packet matches a
2212.Cm keep-state
2213or
2214.Cm limit
2215rule, causing the creation of a
2216.Em dynamic
2217rule which will match all and only packets with
2218a given
2219.Em protocol
2220between a
2221.Em src-ip/src-port dst-ip/dst-port
2222pair of addresses
2223.Em ( src
2224and
2225.Em dst
2226are used here only to denote the initial match addresses, but they
2227are completely equivalent afterwards).
2228Rules created by
2229.Cm keep-state
2230option also have a
2231.Ar flowname
2232taken from it.
2233This name is used in matching together with addresses, ports and protocol.
2234Dynamic rules will be checked at the first
2235.Cm check-state, keep-state
2236or
2237.Cm limit
2238occurrence, and the action performed upon a match will be the same
2239as in the parent rule.
2240.Pp
2241Note that no additional attributes other than protocol and IP addresses
2242and ports and flowname are checked on dynamic rules.
2243.Pp
2244The typical use of dynamic rules is to keep a closed firewall configuration,
2245but let the first TCP SYN packet from the inside network install a
2246dynamic rule for the flow so that packets belonging to that session
2247will be allowed through the firewall:
2248.Pp
2249.Dl "ipfw add check-state OUTBOUND"
2250.Dl "ipfw add allow tcp from my-subnet to any setup keep-state OUTBOUND"
2251.Dl "ipfw add deny tcp from any to any"
2252.Pp
2253A similar approach can be used for UDP, where an UDP packet coming
2254from the inside will install a dynamic rule to let the response through
2255the firewall:
2256.Pp
2257.Dl "ipfw add check-state OUTBOUND"
2258.Dl "ipfw add allow udp from my-subnet to any keep-state OUTBOUND"
2259.Dl "ipfw add deny udp from any to any"
2260.Pp
2261Dynamic rules expire after some time, which depends on the status
2262of the flow and the setting of some
2263.Cm sysctl
2264variables.
2265See Section
2266.Sx SYSCTL VARIABLES
2267for more details.
2268For TCP sessions, dynamic rules can be instructed to periodically
2269send keepalive packets to refresh the state of the rule when it is
2270about to expire.
2271.Pp
2272See Section
2273.Sx EXAMPLES
2274for more examples on how to use dynamic rules.
2275.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
2276.Nm
2277is also the user interface for the
2278.Nm dummynet
2279traffic shaper, packet scheduler and network emulator, a subsystem that
2280can artificially queue, delay or drop packets
2281emulating the behaviour of certain network links
2282or queueing systems.
2283.Pp
2284.Nm dummynet
2285operates by first using the firewall to select packets
2286using any match pattern that can be used in
2287.Nm
2288rules.
2289Matching packets are then passed to either of two
2290different objects, which implement the traffic regulation:
2291.Bl -hang -offset XXXX
2292.It Em pipe
2293A
2294.Em pipe
2295emulates a
2296.Em link
2297with given bandwidth and propagation delay,
2298driven by a FIFO scheduler and a single queue with programmable
2299queue size and packet loss rate.
2300Packets are appended to the queue as they come out from
2301.Nm ipfw ,
2302and then transferred in FIFO order to the link at the desired rate.
2303.It Em queue
2304A
2305.Em queue
2306is an abstraction used to implement packet scheduling
2307using one of several packet scheduling algorithms.
2308Packets sent to a
2309.Em queue
2310are first grouped into flows according to a mask on the 5-tuple.
2311Flows are then passed to the scheduler associated to the
2312.Em queue ,
2313and each flow uses scheduling parameters (weight and others)
2314as configured in the
2315.Em queue
2316itself.
2317A scheduler in turn is connected to an emulated link,
2318and arbitrates the link's bandwidth among backlogged flows according to
2319weights and to the features of the scheduling algorithm in use.
2320.El
2321.Pp
2322In practice,
2323.Em pipes
2324can be used to set hard limits to the bandwidth that a flow can use, whereas
2325.Em queues
2326can be used to determine how different flows share the available bandwidth.
2327.Pp
2328A graphical representation of the binding of queues,
2329flows, schedulers and links is below.
2330.Bd -literal -offset indent
2331                 (flow_mask|sched_mask)  sched_mask
2332         +---------+   weight Wx  +-------------+
2333         |         |->-[flow]-->--|             |-+
2334    -->--| QUEUE x |   ...        |             | |
2335         |         |->-[flow]-->--| SCHEDuler N | |
2336         +---------+              |             | |
2337             ...                  |             +--[LINK N]-->--
2338         +---------+   weight Wy  |             | +--[LINK N]-->--
2339         |         |->-[flow]-->--|             | |
2340    -->--| QUEUE y |   ...        |             | |
2341         |         |->-[flow]-->--|             | |
2342         +---------+              +-------------+ |
2343                                    +-------------+
2344.Ed
2345It is important to understand the role of the SCHED_MASK
2346and FLOW_MASK, which are configured through the commands
2347.Dl "ipfw sched N config mask SCHED_MASK ..."
2348and
2349.Dl "ipfw queue X config mask FLOW_MASK ..." .
2350.Pp
2351The SCHED_MASK is used to assign flows to one or more
2352scheduler instances, one for each
2353value of the packet's 5-tuple after applying SCHED_MASK.
2354As an example, using ``src-ip 0xffffff00'' creates one instance
2355for each /24 destination subnet.
2356.Pp
2357The FLOW_MASK, together with the SCHED_MASK, is used to split
2358packets into flows.
2359As an example, using
2360``src-ip 0x000000ff''
2361together with the previous SCHED_MASK makes a flow for
2362each individual source address.
2363In turn, flows for each /24
2364subnet will be sent to the same scheduler instance.
2365.Pp
2366The above diagram holds even for the
2367.Em pipe
2368case, with the only restriction that a
2369.Em pipe
2370only supports a SCHED_MASK, and forces the use of a FIFO
2371scheduler (these are for backward compatibility reasons;
2372in fact, internally, a
2373.Nm dummynet's
2374pipe is implemented exactly as above).
2375.Pp
2376There are two modes of
2377.Nm dummynet
2378operation:
2379.Dq normal
2380and
2381.Dq fast .
2382The
2383.Dq normal
2384mode tries to emulate a real link: the
2385.Nm dummynet
2386scheduler ensures that the packet will not leave the pipe faster than it
2387would on the real link with a given bandwidth.
2388The
2389.Dq fast
2390mode allows certain packets to bypass the
2391.Nm dummynet
2392scheduler (if packet flow does not exceed pipe's bandwidth).
2393This is the reason why the
2394.Dq fast
2395mode requires less CPU cycles per packet (on average) and packet latency
2396can be significantly lower in comparison to a real link with the same
2397bandwidth.
2398The default mode is
2399.Dq normal .
2400The
2401.Dq fast
2402mode can be enabled by setting the
2403.Va net.inet.ip.dummynet.io_fast
2404.Xr sysctl 8
2405variable to a non-zero value.
2406.Pp
2407.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
2408The
2409.Em pipe ,
2410.Em queue
2411and
2412.Em scheduler
2413configuration commands are the following:
2414.Bd -ragged -offset indent
2415.Cm pipe Ar number Cm config Ar pipe-configuration
2416.Pp
2417.Cm queue Ar number Cm config Ar queue-configuration
2418.Pp
2419.Cm sched Ar number Cm config Ar sched-configuration
2420.Ed
2421.Pp
2422The following parameters can be configured for a pipe:
2423.Pp
2424.Bl -tag -width indent -compact
2425.It Cm bw Ar bandwidth | device
2426Bandwidth, measured in
2427.Sm off
2428.Op Cm K | M
2429.Brq Cm bit/s | Byte/s .
2430.Sm on
2431.Pp
2432A value of 0 (default) means unlimited bandwidth.
2433The unit must immediately follow the number, as in
2434.Pp
2435.Dl "ipfw pipe 1 config bw 300Kbit/s"
2436.Pp
2437If a device name is specified instead of a numeric value, as in
2438.Pp
2439.Dl "ipfw pipe 1 config bw tun0"
2440.Pp
2441then the transmit clock is supplied by the specified device.
2442At the moment only the
2443.Xr tun 4
2444device supports this
2445functionality, for use in conjunction with
2446.Xr ppp 8 .
2447.Pp
2448.It Cm delay Ar ms-delay
2449Propagation delay, measured in milliseconds.
2450The value is rounded to the next multiple of the clock tick
2451(typically 10ms, but it is a good practice to run kernels
2452with
2453.Dq "options HZ=1000"
2454to reduce
2455the granularity to 1ms or less).
2456The default value is 0, meaning no delay.
2457.Pp
2458.It Cm burst Ar size
2459If the data to be sent exceeds the pipe's bandwidth limit
2460(and the pipe was previously idle), up to
2461.Ar size
2462bytes of data are allowed to bypass the
2463.Nm dummynet
2464scheduler, and will be sent as fast as the physical link allows.
2465Any additional data will be transmitted at the rate specified
2466by the
2467.Nm pipe
2468bandwidth.
2469The burst size depends on how long the pipe has been idle;
2470the effective burst size is calculated as follows:
2471MAX(
2472.Ar size
2473,
2474.Nm bw
2475* pipe_idle_time).
2476.Pp
2477.It Cm profile Ar filename
2478A file specifying the additional overhead incurred in the transmission
2479of a packet on the link.
2480.Pp
2481Some link types introduce extra delays in the transmission
2482of a packet, e.g., because of MAC level framing, contention on
2483the use of the channel, MAC level retransmissions and so on.
2484From our point of view, the channel is effectively unavailable
2485for this extra time, which is constant or variable depending
2486on the link type.
2487Additionally, packets may be dropped after this
2488time (e.g., on a wireless link after too many retransmissions).
2489We can model the additional delay with an empirical curve
2490that represents its distribution.
2491.Bd -literal -offset indent
2492      cumulative probability
2493      1.0 ^
2494          |
2495      L   +-- loss-level          x
2496          |                 ******
2497          |                *
2498          |           *****
2499          |          *
2500          |        **
2501          |       *
2502          +-------*------------------->
2503                      delay
2504.Ed
2505The empirical curve may have both vertical and horizontal lines.
2506Vertical lines represent constant delay for a range of
2507probabilities.
2508Horizontal lines correspond to a discontinuity in the delay
2509distribution: the pipe will use the largest delay for a
2510given probability.
2511.Pp
2512The file format is the following, with whitespace acting as
2513a separator and '#' indicating the beginning a comment:
2514.Bl -tag -width indent
2515.It Cm name Ar identifier
2516optional name (listed by "ipfw pipe show")
2517to identify the delay distribution;
2518.It Cm bw Ar value
2519the bandwidth used for the pipe.
2520If not specified here, it must be present
2521explicitly as a configuration parameter for the pipe;
2522.It Cm loss-level Ar L
2523the probability above which packets are lost.
2524(0.0 <= L <= 1.0, default 1.0 i.e., no loss);
2525.It Cm samples Ar N
2526the number of samples used in the internal
2527representation of the curve (2..1024; default 100);
2528.It Cm "delay prob" | "prob delay"
2529One of these two lines is mandatory and defines
2530the format of the following lines with data points.
2531.It Ar XXX Ar YYY
25322 or more lines representing points in the curve,
2533with either delay or probability first, according
2534to the chosen format.
2535The unit for delay is milliseconds.
2536Data points do not need to be sorted.
2537Also, the number of actual lines can be different
2538from the value of the "samples" parameter:
2539.Nm
2540utility will sort and interpolate
2541the curve as needed.
2542.El
2543.Pp
2544Example of a profile file:
2545.Bd -literal -offset indent
2546name    bla_bla_bla
2547samples 100
2548loss-level    0.86
2549prob    delay
25500       200	# minimum overhead is 200ms
25510.5     200
25520.5     300
25530.8     1000
25540.9     1300
25551       1300
2556#configuration file end
2557.Ed
2558.El
2559.Pp
2560The following parameters can be configured for a queue:
2561.Pp
2562.Bl -tag -width indent -compact
2563.It Cm pipe Ar pipe_nr
2564Connects a queue to the specified pipe.
2565Multiple queues (with the same or different weights) can be connected to
2566the same pipe, which specifies the aggregate rate for the set of queues.
2567.Pp
2568.It Cm weight Ar weight
2569Specifies the weight to be used for flows matching this queue.
2570The weight must be in the range 1..100, and defaults to 1.
2571.El
2572.Pp
2573The following case-insensitive parameters can be configured for a
2574scheduler:
2575.Pp
2576.Bl -tag -width indent -compact
2577.It Cm type Ar {fifo | wf2q+ | rr | qfq}
2578specifies the scheduling algorithm to use.
2579.Bl -tag -width indent -compact
2580.It Cm fifo
2581is just a FIFO scheduler (which means that all packets
2582are stored in the same queue as they arrive to the scheduler).
2583FIFO has O(1) per-packet time complexity, with very low
2584constants (estimate 60-80ns on a 2GHz desktop machine)
2585but gives no service guarantees.
2586.It Cm wf2q+
2587implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2588algorithm which permits flows to share bandwidth according to
2589their weights.
2590Note that weights are not priorities; even a flow
2591with a minuscule weight will never starve.
2592WF2Q+ has O(log N) per-packet processing cost, where N is the number
2593of flows, and is the default algorithm used by previous versions
2594dummynet's queues.
2595.It Cm rr
2596implements the Deficit Round Robin algorithm, which has O(1) processing
2597costs (roughly, 100-150ns per packet)
2598and permits bandwidth allocation according to weights, but
2599with poor service guarantees.
2600.It Cm qfq
2601implements the QFQ algorithm, which is a very fast variant of
2602WF2Q+, with similar service guarantees and O(1) processing
2603costs (roughly, 200-250ns per packet).
2604.El
2605.El
2606.Pp
2607In addition to the type, all parameters allowed for a pipe can also
2608be specified for a scheduler.
2609.Pp
2610Finally, the following parameters can be configured for both
2611pipes and queues:
2612.Pp
2613.Bl -tag -width XXXX -compact
2614.It Cm buckets Ar hash-table-size
2615Specifies the size of the hash table used for storing the
2616various queues.
2617Default value is 64 controlled by the
2618.Xr sysctl 8
2619variable
2620.Va net.inet.ip.dummynet.hash_size ,
2621allowed range is 16 to 65536.
2622.Pp
2623.It Cm mask Ar mask-specifier
2624Packets sent to a given pipe or queue by an
2625.Nm
2626rule can be further classified into multiple flows, each of which is then
2627sent to a different
2628.Em dynamic
2629pipe or queue.
2630A flow identifier is constructed by masking the IP addresses,
2631ports and protocol types as specified with the
2632.Cm mask
2633options in the configuration of the pipe or queue.
2634For each different flow identifier, a new pipe or queue is created
2635with the same parameters as the original object, and matching packets
2636are sent to it.
2637.Pp
2638Thus, when
2639.Em dynamic pipes
2640are used, each flow will get the same bandwidth as defined by the pipe,
2641whereas when
2642.Em dynamic queues
2643are used, each flow will share the parent's pipe bandwidth evenly
2644with other flows generated by the same queue (note that other queues
2645with different weights might be connected to the same pipe).
2646.br
2647Available mask specifiers are a combination of one or more of the following:
2648.Pp
2649.Cm dst-ip Ar mask ,
2650.Cm dst-ip6 Ar mask ,
2651.Cm src-ip Ar mask ,
2652.Cm src-ip6 Ar mask ,
2653.Cm dst-port Ar mask ,
2654.Cm src-port Ar mask ,
2655.Cm flow-id Ar mask ,
2656.Cm proto Ar mask
2657or
2658.Cm all ,
2659.Pp
2660where the latter means all bits in all fields are significant.
2661.Pp
2662.It Cm noerror
2663When a packet is dropped by a
2664.Nm dummynet
2665queue or pipe, the error
2666is normally reported to the caller routine in the kernel, in the
2667same way as it happens when a device queue fills up.
2668Setting this
2669option reports the packet as successfully delivered, which can be
2670needed for some experimental setups where you want to simulate
2671loss or congestion at a remote router.
2672.Pp
2673.It Cm plr Ar packet-loss-rate
2674Packet loss rate.
2675Argument
2676.Ar packet-loss-rate
2677is a floating-point number between 0 and 1, with 0 meaning no
2678loss, 1 meaning 100% loss.
2679The loss rate is internally represented on 31 bits.
2680.Pp
2681.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2682Queue size, in
2683.Ar slots
2684or
2685.Cm KBytes .
2686Default value is 50 slots, which
2687is the typical queue size for Ethernet devices.
2688Note that for slow speed links you should keep the queue
2689size short or your traffic might be affected by a significant
2690queueing delay.
2691E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2692or 20s of queue on a 30Kbit/s pipe.
2693Even worse effects can result if you get packets from an
2694interface with a much larger MTU, e.g.\& the loopback interface
2695with its 16KB packets.
2696The
2697.Xr sysctl 8
2698variables
2699.Em net.inet.ip.dummynet.pipe_byte_limit
2700and
2701.Em net.inet.ip.dummynet.pipe_slot_limit
2702control the maximum lengths that can be specified.
2703.Pp
2704.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2705[ecn]
2706Make use of the RED (Random Early Detection) queue management algorithm.
2707.Ar w_q
2708and
2709.Ar max_p
2710are floating
2711point numbers between 0 and 1 (inclusive), while
2712.Ar min_th
2713and
2714.Ar max_th
2715are integer numbers specifying thresholds for queue management
2716(thresholds are computed in bytes if the queue has been defined
2717in bytes, in slots otherwise).
2718The two parameters can also be of the same value if needed. The
2719.Nm dummynet
2720also supports the gentle RED variant (gred) and ECN (Explicit Congestion
2721Notification) as optional. Three
2722.Xr sysctl 8
2723variables can be used to control the RED behaviour:
2724.Bl -tag -width indent
2725.It Va net.inet.ip.dummynet.red_lookup_depth
2726specifies the accuracy in computing the average queue
2727when the link is idle (defaults to 256, must be greater than zero)
2728.It Va net.inet.ip.dummynet.red_avg_pkt_size
2729specifies the expected average packet size (defaults to 512, must be
2730greater than zero)
2731.It Va net.inet.ip.dummynet.red_max_pkt_size
2732specifies the expected maximum packet size, only used when queue
2733thresholds are in bytes (defaults to 1500, must be greater than zero).
2734.El
2735.El
2736.Pp
2737When used with IPv6 data,
2738.Nm dummynet
2739currently has several limitations.
2740Information necessary to route link-local packets to an
2741interface is not available after processing by
2742.Nm dummynet
2743so those packets are dropped in the output path.
2744Care should be taken to ensure that link-local packets are not passed to
2745.Nm dummynet .
2746.Sh CHECKLIST
2747Here are some important points to consider when designing your
2748rules:
2749.Bl -bullet
2750.It
2751Remember that you filter both packets going
2752.Cm in
2753and
2754.Cm out .
2755Most connections need packets going in both directions.
2756.It
2757Remember to test very carefully.
2758It is a good idea to be near the console when doing this.
2759If you cannot be near the console,
2760use an auto-recovery script such as the one in
2761.Pa /usr/share/examples/ipfw/change_rules.sh .
2762.It
2763Do not forget the loopback interface.
2764.El
2765.Sh FINE POINTS
2766.Bl -bullet
2767.It
2768There are circumstances where fragmented datagrams are unconditionally
2769dropped.
2770TCP packets are dropped if they do not contain at least 20 bytes of
2771TCP header, UDP packets are dropped if they do not contain a full 8
2772byte UDP header, and ICMP packets are dropped if they do not contain
27734 bytes of ICMP header, enough to specify the ICMP type, code, and
2774checksum.
2775These packets are simply logged as
2776.Dq pullup failed
2777since there may not be enough good data in the packet to produce a
2778meaningful log entry.
2779.It
2780Another type of packet is unconditionally dropped, a TCP packet with a
2781fragment offset of one.
2782This is a valid packet, but it only has one use, to try
2783to circumvent firewalls.
2784When logging is enabled, these packets are
2785reported as being dropped by rule -1.
2786.It
2787If you are logged in over a network, loading the
2788.Xr kld 4
2789version of
2790.Nm
2791is probably not as straightforward as you would think.
2792The following command line is recommended:
2793.Bd -literal -offset indent
2794kldload ipfw && \e
2795ipfw add 32000 allow ip from any to any
2796.Ed
2797.Pp
2798Along the same lines, doing an
2799.Bd -literal -offset indent
2800ipfw flush
2801.Ed
2802.Pp
2803in similar surroundings is also a bad idea.
2804.It
2805The
2806.Nm
2807filter list may not be modified if the system security level
2808is set to 3 or higher
2809(see
2810.Xr init 8
2811for information on system security levels).
2812.El
2813.Sh PACKET DIVERSION
2814A
2815.Xr divert 4
2816socket bound to the specified port will receive all packets
2817diverted to that port.
2818If no socket is bound to the destination port, or if the divert module is
2819not loaded, or if the kernel was not compiled with divert socket support,
2820the packets are dropped.
2821.Sh NETWORK ADDRESS TRANSLATION (NAT)
2822.Nm
2823support in-kernel NAT using the kernel version of
2824.Xr libalias 3 .
2825.Pp
2826The nat configuration command is the following:
2827.Bd -ragged -offset indent
2828.Bk -words
2829.Cm nat
2830.Ar nat_number
2831.Cm config
2832.Ar nat-configuration
2833.Ek
2834.Ed
2835.Pp
2836The following parameters can be configured:
2837.Bl -tag -width indent
2838.It Cm ip Ar ip_address
2839Define an ip address to use for aliasing.
2840.It Cm if Ar nic
2841Use ip address of NIC for aliasing, dynamically changing
2842it if NIC's ip address changes.
2843.It Cm log
2844Enable logging on this nat instance.
2845.It Cm deny_in
2846Deny any incoming connection from outside world.
2847.It Cm same_ports
2848Try to leave the alias port numbers unchanged from
2849the actual local port numbers.
2850.It Cm unreg_only
2851Traffic on the local network not originating from an
2852unregistered address spaces will be ignored.
2853.It Cm reset
2854Reset table of the packet aliasing engine on address change.
2855.It Cm reverse
2856Reverse the way libalias handles aliasing.
2857.It Cm proxy_only
2858Obey transparent proxy rules only, packet aliasing is not performed.
2859.It Cm skip_global
2860Skip instance in case of global state lookup (see below).
2861.El
2862.Pp
2863Some specials value can be supplied instead of
2864.Va nat_number:
2865.Bl -tag -width indent
2866.It Cm global
2867Looks up translation state in all configured nat instances.
2868If an entry is found, packet is aliased according to that entry.
2869If no entry was found in any of the instances, packet is passed unchanged,
2870and no new entry will be created.
2871See section
2872.Sx MULTIPLE INSTANCES
2873in
2874.Xr natd 8
2875for more information.
2876.It Cm tablearg
2877Uses argument supplied in lookup table.
2878See
2879.Sx LOOKUP TABLES
2880section below for more information on lookup tables.
2881.El
2882.Pp
2883To let the packet continue after being (de)aliased, set the sysctl variable
2884.Va net.inet.ip.fw.one_pass
2885to 0.
2886For more information about aliasing modes, refer to
2887.Xr libalias 3 .
2888See Section
2889.Sx EXAMPLES
2890for some examples about nat usage.
2891.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
2892Redirect and LSNAT support follow closely the syntax used in
2893.Xr natd 8 .
2894See Section
2895.Sx EXAMPLES
2896for some examples on how to do redirect and lsnat.
2897.Ss SCTP NAT SUPPORT
2898SCTP nat can be configured in a similar manner to TCP through the
2899.Nm
2900command line tool.
2901The main difference is that
2902.Nm sctp nat
2903does not do port translation.
2904Since the local and global side ports will be the same,
2905there is no need to specify both.
2906Ports are redirected as follows:
2907.Bd -ragged -offset indent
2908.Bk -words
2909.Cm nat
2910.Ar nat_number
2911.Cm config if
2912.Ar nic
2913.Cm redirect_port sctp
2914.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
2915.Ek
2916.Ed
2917.Pp
2918Most
2919.Nm sctp nat
2920configuration can be done in real-time through the
2921.Xr sysctl 8
2922interface.
2923All may be changed dynamically, though the hash_table size will only
2924change for new
2925.Nm nat
2926instances.
2927See
2928.Sx SYSCTL VARIABLES
2929for more info.
2930.Sh IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
2931.Nm
2932support in-kernel IPv6-to-IPv6 network prefix translation as described
2933in RFC6296.
2934The kernel module
2935.Cm ipfw_nptv6
2936should be loaded or kernel should has
2937.Cm options IPFIREWALL_NPTV6
2938to be able use NPTv6 translator.
2939.Pp
2940The NPTv6 configuration command is the following:
2941.Bd -ragged -offset indent
2942.Bk -words
2943.Cm nptv6
2944.Ar name
2945.Cm create
2946.Ar create-options
2947.Ek
2948.Ed
2949.Pp
2950The following parameters can be configured:
2951.Bl -tag -width indent
2952.It Cm int_prefix Ar ipv6_prefix
2953IPv6 prefix used in internal network.
2954NPTv6 module translates source address when it matches this prefix.
2955.It Cm ext_prefix Ar ipv6_prefix
2956IPv6 prefix used in external network.
2957NPTv6 module translates destination address when it matches this prefix.
2958.It Cm prefixlen Ar length
2959The length of specified IPv6 prefixes. It must be in range from 8 to 64.
2960.El
2961.Pp
2962Note that the prefix translation rules are silently ignored when IPv6 packet
2963forwarding is disabled.
2964To enable the packet forwarding, set the sysctl variable
2965.Va net.inet6.ip6.forwarding
2966to 1.
2967.Pp
2968To let the packet continue after being translated, set the sysctl variable
2969.Va net.inet.ip.fw.one_pass
2970to 0.
2971.Sh LOADER TUNABLES
2972Tunables can be set in
2973.Xr loader 8
2974prompt,
2975.Xr loader.conf 5
2976or
2977.Xr kenv 1
2978before ipfw module gets loaded.
2979.Bl -tag -width indent
2980.It Va net.inet.ip.fw.default_to_accept: No 0
2981Defines ipfw last rule behavior.
2982This value overrides
2983.Cd "options IPFW_DEFAULT_TO_(ACCEPT|DENY)"
2984from kernel configuration file.
2985.It Va net.inet.ip.fw.tables_max: No 128
2986Defines number of tables available in ipfw.
2987Number cannot exceed 65534.
2988.El
2989.Sh SYSCTL VARIABLES
2990A set of
2991.Xr sysctl 8
2992variables controls the behaviour of the firewall and
2993associated modules
2994.Pq Nm dummynet , bridge , sctp nat .
2995These are shown below together with their default value
2996(but always check with the
2997.Xr sysctl 8
2998command what value is actually in use) and meaning:
2999.Bl -tag -width indent
3000.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
3001Defines how the
3002.Nm nat
3003responds to receipt of global OOTB ASCONF-AddIP:
3004.Bl -tag -width indent
3005.It Cm 0
3006No response (unless a partially matching association exists -
3007ports and vtags match but global address does not)
3008.It Cm 1
3009.Nm nat
3010will accept and process all OOTB global AddIP messages.
3011.El
3012.Pp
3013Option 1 should never be selected as this forms a security risk.
3014An attacker can
3015establish multiple fake associations by sending AddIP messages.
3016.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
3017Defines the maximum number of chunks in an SCTP packet that will be
3018parsed for a
3019packet that matches an existing association.
3020This value is enforced to be greater or equal than
3021.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
3022A high value is
3023a DoS risk yet setting too low a value may result in
3024important control chunks in
3025the packet not being located and parsed.
3026.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
3027Defines when the
3028.Nm nat
3029responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
3030An OOTB packet is a packet that arrives with no existing association
3031registered in the
3032.Nm nat
3033and is not an INIT or ASCONF-AddIP packet:
3034.Bl -tag -width indent
3035.It Cm 0
3036ErrorM is never sent in response to OOTB packets.
3037.It Cm 1
3038ErrorM is only sent to OOTB packets received on the local side.
3039.It Cm 2
3040ErrorM is sent to the local side and on the global side ONLY if there is a
3041partial match (ports and vtags match but the source global IP does not).
3042This value is only useful if the
3043.Nm nat
3044is tracking global IP addresses.
3045.It Cm 3
3046ErrorM is sent in response to all OOTB packets on both
3047the local and global side
3048(DoS risk).
3049.El
3050.Pp
3051At the moment the default is 0, since the ErrorM packet is not yet
3052supported by most SCTP stacks.
3053When it is supported, and if not tracking
3054global addresses, we recommend setting this value to 1 to allow
3055multi-homed local hosts to function with the
3056.Nm nat .
3057To track global addresses, we recommend setting this value to 2 to
3058allow global hosts to be informed when they need to (re)send an
3059ASCONF-AddIP.
3060Value 3 should never be chosen (except for debugging) as the
3061.Nm nat
3062will respond to all OOTB global packets (a DoS risk).
3063.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
3064Size of hash tables used for
3065.Nm nat
3066lookups (100 < prime_number > 1000001).
3067This value sets the
3068.Nm hash table
3069size for any future created
3070.Nm nat
3071instance and therefore must be set prior to creating a
3072.Nm nat
3073instance.
3074The table sizes may be changed to suit specific needs.
3075If there will be few
3076concurrent associations, and memory is scarce, you may make these smaller.
3077If there will be many thousands (or millions) of concurrent associations, you
3078should make these larger.
3079A prime number is best for the table size.
3080The sysctl
3081update function will adjust your input value to the next highest prime number.
3082.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
3083Hold association in table for this many seconds after receiving a
3084SHUTDOWN-COMPLETE.
3085This allows endpoints to correct shutdown gracefully if a
3086shutdown_complete is lost and retransmissions are required.
3087.It Va net.inet.ip.alias.sctp.init_timer: No 15
3088Timeout value while waiting for (INIT-ACK|AddIP-ACK).
3089This value cannot be 0.
3090.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
3091Defines the maximum number of chunks in an SCTP packet that will be parsed when
3092no existing association exists that matches that packet.
3093Ideally this packet
3094will only be an INIT or ASCONF-AddIP packet.
3095A higher value may become a DoS
3096risk as malformed packets can consume processing resources.
3097.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
3098Defines the maximum number of parameters within a chunk that will be
3099parsed in a
3100packet.
3101As for other similar sysctl variables, larger values pose a DoS risk.
3102.It Va net.inet.ip.alias.sctp.log_level: No 0
3103Level of detail in the system log messages (0 \- minimal, 1 \- event,
31042 \- info, 3 \- detail, 4 \- debug, 5 \- max debug).
3105May be a good
3106option in high loss environments.
3107.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
3108Timeout value while waiting for SHUTDOWN-COMPLETE.
3109This value cannot be 0.
3110.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
3111Enables/disables global IP address tracking within the
3112.Nm nat
3113and places an
3114upper limit on the number of addresses tracked for each association:
3115.Bl -tag -width indent
3116.It Cm 0
3117Global tracking is disabled
3118.It Cm >1
3119Enables tracking, the maximum number of addresses tracked for each
3120association is limited to this value
3121.El
3122.Pp
3123This variable is fully dynamic, the new value will be adopted for all newly
3124arriving associations, existing associations are treated
3125as they were previously.
3126Global tracking will decrease the number of collisions within the
3127.Nm nat
3128at a cost
3129of increased processing load, memory usage, complexity, and possible
3130.Nm nat
3131state
3132problems in complex networks with multiple
3133.Nm nats .
3134We recommend not tracking
3135global IP addresses, this will still result in a fully functional
3136.Nm nat .
3137.It Va net.inet.ip.alias.sctp.up_timer: No 300
3138Timeout value to keep an association up with no traffic.
3139This value cannot be 0.
3140.It Va net.inet.ip.dummynet.expire : No 1
3141Lazily delete dynamic pipes/queue once they have no pending traffic.
3142You can disable this by setting the variable to 0, in which case
3143the pipes/queues will only be deleted when the threshold is reached.
3144.It Va net.inet.ip.dummynet.hash_size : No 64
3145Default size of the hash table used for dynamic pipes/queues.
3146This value is used when no
3147.Cm buckets
3148option is specified when configuring a pipe/queue.
3149.It Va net.inet.ip.dummynet.io_fast : No 0
3150If set to a non-zero value,
3151the
3152.Dq fast
3153mode of
3154.Nm dummynet
3155operation (see above) is enabled.
3156.It Va net.inet.ip.dummynet.io_pkt
3157Number of packets passed to
3158.Nm dummynet .
3159.It Va net.inet.ip.dummynet.io_pkt_drop
3160Number of packets dropped by
3161.Nm dummynet .
3162.It Va net.inet.ip.dummynet.io_pkt_fast
3163Number of packets bypassed by the
3164.Nm dummynet
3165scheduler.
3166.It Va net.inet.ip.dummynet.max_chain_len : No 16
3167Target value for the maximum number of pipes/queues in a hash bucket.
3168The product
3169.Cm max_chain_len*hash_size
3170is used to determine the threshold over which empty pipes/queues
3171will be expired even when
3172.Cm net.inet.ip.dummynet.expire=0 .
3173.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
3174.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
3175.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
3176Parameters used in the computations of the drop probability
3177for the RED algorithm.
3178.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
3179.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
3180The maximum queue size that can be specified in bytes or packets.
3181These limits prevent accidental exhaustion of resources such as mbufs.
3182If you raise these limits,
3183you should make sure the system is configured so that sufficient resources
3184are available.
3185.It Va net.inet.ip.fw.autoinc_step : No 100
3186Delta between rule numbers when auto-generating them.
3187The value must be in the range 1..1000.
3188.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
3189The current number of buckets in the hash table for dynamic rules
3190(readonly).
3191.It Va net.inet.ip.fw.debug : No 1
3192Controls debugging messages produced by
3193.Nm .
3194.It Va net.inet.ip.fw.default_rule : No 65535
3195The default rule number (read-only).
3196By the design of
3197.Nm , the default rule is the last one, so its number
3198can also serve as the highest number allowed for a rule.
3199.It Va net.inet.ip.fw.dyn_buckets : No 256
3200The number of buckets in the hash table for dynamic rules.
3201Must be a power of 2, up to 65536.
3202It only takes effect when all dynamic rules have expired, so you
3203are advised to use a
3204.Cm flush
3205command to make sure that the hash table is resized.
3206.It Va net.inet.ip.fw.dyn_count : No 3
3207Current number of dynamic rules
3208(read-only).
3209.It Va net.inet.ip.fw.dyn_keepalive : No 1
3210Enables generation of keepalive packets for
3211.Cm keep-state
3212rules on TCP sessions.
3213A keepalive is generated to both
3214sides of the connection every 5 seconds for the last 20
3215seconds of the lifetime of the rule.
3216.It Va net.inet.ip.fw.dyn_max : No 8192
3217Maximum number of dynamic rules.
3218When you hit this limit, no more dynamic rules can be
3219installed until old ones expire.
3220.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
3221.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
3222.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
3223.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
3224.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
3225.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
3226These variables control the lifetime, in seconds, of dynamic
3227rules.
3228Upon the initial SYN exchange the lifetime is kept short,
3229then increased after both SYN have been seen, then decreased
3230again during the final FIN exchange or when a RST is received.
3231Both
3232.Em dyn_fin_lifetime
3233and
3234.Em dyn_rst_lifetime
3235must be strictly lower than 5 seconds, the period of
3236repetition of keepalives.
3237The firewall enforces that.
3238.It Va net.inet.ip.fw.dyn_keep_states: No 0
3239Keep dynamic states on rule/set deletion.
3240States are relinked to default rule (65535).
3241This can be handly for ruleset reload.
3242Turned off by default.
3243.It Va net.inet.ip.fw.enable : No 1
3244Enables the firewall.
3245Setting this variable to 0 lets you run your machine without
3246firewall even if compiled in.
3247.It Va net.inet6.ip6.fw.enable : No 1
3248provides the same functionality as above for the IPv6 case.
3249.It Va net.inet.ip.fw.one_pass : No 1
3250When set, the packet exiting from the
3251.Nm dummynet
3252pipe or from
3253.Xr ng_ipfw 4
3254node is not passed though the firewall again.
3255Otherwise, after an action, the packet is
3256reinjected into the firewall at the next rule.
3257.It Va net.inet.ip.fw.tables_max : No 128
3258Maximum number of tables.
3259.It Va net.inet.ip.fw.verbose : No 1
3260Enables verbose messages.
3261.It Va net.inet.ip.fw.verbose_limit : No 0
3262Limits the number of messages produced by a verbose firewall.
3263.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
3264If enabled packets with unknown IPv6 Extension Headers will be denied.
3265.It Va net.link.ether.ipfw : No 0
3266Controls whether layer-2 packets are passed to
3267.Nm .
3268Default is no.
3269.It Va net.link.bridge.ipfw : No 0
3270Controls whether bridged packets are passed to
3271.Nm .
3272Default is no.
3273.El
3274.Sh INTERNAL DIAGNOSTICS
3275There are some commands that may be useful to understand current state
3276of certain subsystems inside kernel module.
3277These commands provide debugging output which may change without notice.
3278.Pp
3279Currently the following commands are available as
3280.Cm internal
3281sub-options:
3282.Bl -tag -width indent
3283.It Cm iflist
3284Lists all interface which are currently tracked by
3285.Nm
3286with their in-kernel status.
3287.It Cm talist
3288List all table lookup algorithms currently available.
3289.El
3290.Sh EXAMPLES
3291There are far too many possible uses of
3292.Nm
3293so this Section will only give a small set of examples.
3294.Pp
3295.Ss BASIC PACKET FILTERING
3296This command adds an entry which denies all tcp packets from
3297.Em cracker.evil.org
3298to the telnet port of
3299.Em wolf.tambov.su
3300from being forwarded by the host:
3301.Pp
3302.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
3303.Pp
3304This one disallows any connection from the entire cracker's
3305network to my host:
3306.Pp
3307.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
3308.Pp
3309A first and efficient way to limit access (not using dynamic rules)
3310is the use of the following rules:
3311.Pp
3312.Dl "ipfw add allow tcp from any to any established"
3313.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
3314.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
3315.Dl "..."
3316.Dl "ipfw add deny tcp from any to any"
3317.Pp
3318The first rule will be a quick match for normal TCP packets,
3319but it will not match the initial SYN packet, which will be
3320matched by the
3321.Cm setup
3322rules only for selected source/destination pairs.
3323All other SYN packets will be rejected by the final
3324.Cm deny
3325rule.
3326.Pp
3327If you administer one or more subnets, you can take advantage
3328of the address sets and or-blocks and write extremely
3329compact rulesets which selectively enable services to blocks
3330of clients, as below:
3331.Pp
3332.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
3333.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
3334.Dl ""
3335.Dl "ipfw add allow ip from ${goodguys} to any"
3336.Dl "ipfw add deny ip from ${badguys} to any"
3337.Dl "... normal policies ..."
3338.Pp
3339The
3340.Cm verrevpath
3341option could be used to do automated anti-spoofing by adding the
3342following to the top of a ruleset:
3343.Pp
3344.Dl "ipfw add deny ip from any to any not verrevpath in"
3345.Pp
3346This rule drops all incoming packets that appear to be coming to the
3347system on the wrong interface.
3348For example, a packet with a source
3349address belonging to a host on a protected internal network would be
3350dropped if it tried to enter the system from an external interface.
3351.Pp
3352The
3353.Cm antispoof
3354option could be used to do similar but more restricted anti-spoofing
3355by adding the following to the top of a ruleset:
3356.Pp
3357.Dl "ipfw add deny ip from any to any not antispoof in"
3358.Pp
3359This rule drops all incoming packets that appear to be coming from another
3360directly connected system but on the wrong interface.
3361For example, a packet with a source address of
3362.Li 192.168.0.0/24 ,
3363configured on
3364.Li fxp0 ,
3365but coming in on
3366.Li fxp1
3367would be dropped.
3368.Pp
3369The
3370.Cm setdscp
3371option could be used to (re)mark user traffic,
3372by adding the following to the appropriate place in ruleset:
3373.Pp
3374.Dl "ipfw add setdscp be ip from any to any dscp af11,af21"
3375.Ss DYNAMIC RULES
3376In order to protect a site from flood attacks involving fake
3377TCP packets, it is safer to use dynamic rules:
3378.Pp
3379.Dl "ipfw add check-state"
3380.Dl "ipfw add deny tcp from any to any established"
3381.Dl "ipfw add allow tcp from my-net to any setup keep-state"
3382.Pp
3383This will let the firewall install dynamic rules only for
3384those connection which start with a regular SYN packet coming
3385from the inside of our network.
3386Dynamic rules are checked when encountering the first
3387occurrence of a
3388.Cm check-state ,
3389.Cm keep-state
3390or
3391.Cm limit
3392rule.
3393A
3394.Cm check-state
3395rule should usually be placed near the beginning of the
3396ruleset to minimize the amount of work scanning the ruleset.
3397Your mileage may vary.
3398.Pp
3399To limit the number of connections a user can open
3400you can use the following type of rules:
3401.Pp
3402.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
3403.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
3404.Pp
3405The former (assuming it runs on a gateway) will allow each host
3406on a /24 network to open at most 10 TCP connections.
3407The latter can be placed on a server to make sure that a single
3408client does not use more than 4 simultaneous connections.
3409.Pp
3410.Em BEWARE :
3411stateful rules can be subject to denial-of-service attacks
3412by a SYN-flood which opens a huge number of dynamic rules.
3413The effects of such attacks can be partially limited by
3414acting on a set of
3415.Xr sysctl 8
3416variables which control the operation of the firewall.
3417.Pp
3418Here is a good usage of the
3419.Cm list
3420command to see accounting records and timestamp information:
3421.Pp
3422.Dl ipfw -at list
3423.Pp
3424or in short form without timestamps:
3425.Pp
3426.Dl ipfw -a list
3427.Pp
3428which is equivalent to:
3429.Pp
3430.Dl ipfw show
3431.Pp
3432Next rule diverts all incoming packets from 192.168.2.0/24
3433to divert port 5000:
3434.Pp
3435.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
3436.Ss TRAFFIC SHAPING
3437The following rules show some of the applications of
3438.Nm
3439and
3440.Nm dummynet
3441for simulations and the like.
3442.Pp
3443This rule drops random incoming packets with a probability
3444of 5%:
3445.Pp
3446.Dl "ipfw add prob 0.05 deny ip from any to any in"
3447.Pp
3448A similar effect can be achieved making use of
3449.Nm dummynet
3450pipes:
3451.Pp
3452.Dl "ipfw add pipe 10 ip from any to any"
3453.Dl "ipfw pipe 10 config plr 0.05"
3454.Pp
3455We can use pipes to artificially limit bandwidth, e.g.\& on a
3456machine acting as a router, if we want to limit traffic from
3457local clients on 192.168.2.0/24 we do:
3458.Pp
3459.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3460.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
3461.Pp
3462note that we use the
3463.Cm out
3464modifier so that the rule is not used twice.
3465Remember in fact that
3466.Nm
3467rules are checked both on incoming and outgoing packets.
3468.Pp
3469Should we want to simulate a bidirectional link with bandwidth
3470limitations, the correct way is the following:
3471.Pp
3472.Dl "ipfw add pipe 1 ip from any to any out"
3473.Dl "ipfw add pipe 2 ip from any to any in"
3474.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
3475.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
3476.Pp
3477The above can be very useful, e.g.\& if you want to see how
3478your fancy Web page will look for a residential user who
3479is connected only through a slow link.
3480You should not use only one pipe for both directions, unless
3481you want to simulate a half-duplex medium (e.g.\& AppleTalk,
3482Ethernet, IRDA).
3483It is not necessary that both pipes have the same configuration,
3484so we can also simulate asymmetric links.
3485.Pp
3486Should we want to verify network performance with the RED queue
3487management algorithm:
3488.Pp
3489.Dl "ipfw add pipe 1 ip from any to any"
3490.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
3491.Pp
3492Another typical application of the traffic shaper is to
3493introduce some delay in the communication.
3494This can significantly affect applications which do a lot of Remote
3495Procedure Calls, and where the round-trip-time of the
3496connection often becomes a limiting factor much more than
3497bandwidth:
3498.Pp
3499.Dl "ipfw add pipe 1 ip from any to any out"
3500.Dl "ipfw add pipe 2 ip from any to any in"
3501.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
3502.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
3503.Pp
3504Per-flow queueing can be useful for a variety of purposes.
3505A very simple one is counting traffic:
3506.Pp
3507.Dl "ipfw add pipe 1 tcp from any to any"
3508.Dl "ipfw add pipe 1 udp from any to any"
3509.Dl "ipfw add pipe 1 ip from any to any"
3510.Dl "ipfw pipe 1 config mask all"
3511.Pp
3512The above set of rules will create queues (and collect
3513statistics) for all traffic.
3514Because the pipes have no limitations, the only effect is
3515collecting statistics.
3516Note that we need 3 rules, not just the last one, because
3517when
3518.Nm
3519tries to match IP packets it will not consider ports, so we
3520would not see connections on separate ports as different
3521ones.
3522.Pp
3523A more sophisticated example is limiting the outbound traffic
3524on a net with per-host limits, rather than per-network limits:
3525.Pp
3526.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3527.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
3528.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3529.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3530.Ss LOOKUP TABLES
3531In the following example, we need to create several traffic bandwidth
3532classes and we need different hosts/networks to fall into different classes.
3533We create one pipe for each class and configure them accordingly.
3534Then we create a single table and fill it with IP subnets and addresses.
3535For each subnet/host we set the argument equal to the number of the pipe
3536that it should use.
3537Then we classify traffic using a single rule:
3538.Pp
3539.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
3540.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
3541.Dl "..."
3542.Dl "ipfw table T1 create type addr"
3543.Dl "ipfw table T1 add 192.168.2.0/24 1"
3544.Dl "ipfw table T1 add 192.168.0.0/27 4"
3545.Dl "ipfw table T1 add 192.168.0.2 1"
3546.Dl "..."
3547.Dl "ipfw add pipe tablearg ip from 'table(T1)' to any"
3548.Pp
3549Using the
3550.Cm fwd
3551action, the table entries may include hostnames and IP addresses.
3552.Pp
3553.Dl "ipfw table T2 create type addr ftype ip"
3554.Dl "ipfw table T2 add 192.168.2.0/24 10.23.2.1"
3555.Dl "ipfw table T21 add 192.168.0.0/27 router1.dmz"
3556.Dl "..."
3557.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
3558.Pp
3559In the following example per-interface firewall is created:
3560.Pp
3561.Dl "ipfw table IN create type iface valtype skipto,fib"
3562.Dl "ipfw table IN add vlan20 12000,12"
3563.Dl "ipfw table IN add vlan30 13000,13"
3564.Dl "ipfw table OUT create type iface valtype skipto"
3565.Dl "ipfw table OUT add vlan20 22000"
3566.Dl "ipfw table OUT add vlan30 23000"
3567.Dl ".."
3568.Dl "ipfw add 100 ipfw setfib tablearg ip from any to any recv 'table(IN)' in"
3569.Dl "ipfw add 200 ipfw skipto tablearg ip from any to any recv 'table(IN)' in"
3570.Dl "ipfw add 300 ipfw skipto tablearg ip from any to any xmit 'table(OUT)' out"
3571.Pp
3572The following example illustrate usage of flow tables:
3573.Pp
3574.Dl "ipfw table fl create type flow:flow:src-ip,proto,dst-ip,dst-port"
3575.Dl "ipfw table fl add 2a02:6b8:77::88,tcp,2a02:6b8:77::99,80 11"
3576.Dl "ipfw table fl add 10.0.0.1,udp,10.0.0.2,53 12"
3577.Dl ".."
3578.Dl "ipfw add 100 allow ip from any to any flow 'table(fl,11)' recv ix0"
3579.Ss SETS OF RULES
3580To add a set of rules atomically, e.g.\& set 18:
3581.Pp
3582.Dl "ipfw set disable 18"
3583.Dl "ipfw add NN set 18 ...         # repeat as needed"
3584.Dl "ipfw set enable 18"
3585.Pp
3586To delete a set of rules atomically the command is simply:
3587.Pp
3588.Dl "ipfw delete set 18"
3589.Pp
3590To test a ruleset and disable it and regain control if something goes wrong:
3591.Pp
3592.Dl "ipfw set disable 18"
3593.Dl "ipfw add NN set 18 ...         # repeat as needed"
3594.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
3595.Pp
3596Here if everything goes well, you press control-C before the "sleep"
3597terminates, and your ruleset will be left active.
3598Otherwise, e.g.\& if
3599you cannot access your box, the ruleset will be disabled after
3600the sleep terminates thus restoring the previous situation.
3601.Pp
3602To show rules of the specific set:
3603.Pp
3604.Dl "ipfw set 18 show"
3605.Pp
3606To show rules of the disabled set:
3607.Pp
3608.Dl "ipfw -S set 18 show"
3609.Pp
3610To clear a specific rule counters of the specific set:
3611.Pp
3612.Dl "ipfw set 18 zero NN"
3613.Pp
3614To delete a specific rule of the specific set:
3615.Pp
3616.Dl "ipfw set 18 delete NN"
3617.Ss NAT, REDIRECT AND LSNAT
3618First redirect all the traffic to nat instance 123:
3619.Pp
3620.Dl "ipfw add nat 123 all from any to any"
3621.Pp
3622Then to configure nat instance 123 to alias all the outgoing traffic with ip
3623192.168.0.123, blocking all incoming connections, trying to keep
3624same ports on both sides, clearing aliasing table on address change
3625and keeping a log of traffic/link statistics:
3626.Pp
3627.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
3628.Pp
3629Or to change address of instance 123, aliasing table will be cleared (see
3630reset option):
3631.Pp
3632.Dl "ipfw nat 123 config ip 10.0.0.1"
3633.Pp
3634To see configuration of nat instance 123:
3635.Pp
3636.Dl "ipfw nat 123 show config"
3637.Pp
3638To show logs of all the instances in range 111-999:
3639.Pp
3640.Dl "ipfw nat 111-999 show"
3641.Pp
3642To see configurations of all instances:
3643.Pp
3644.Dl "ipfw nat show config"
3645.Pp
3646Or a redirect rule with mixed modes could looks like:
3647.Pp
3648.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
3649.Dl "			 redirect_port tcp 192.168.0.1:80 500"
3650.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
3651.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
3652.Dl "			 	    10.0.0.100	# LSNAT"
3653.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
3654.Dl "			 	    500		# LSNAT"
3655.Pp
3656or it could be split in:
3657.Pp
3658.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
3659.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
3660.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
3661.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
3662.Dl "				         10.0.0.100"
3663.Dl "ipfw nat 5 config redirect_port tcp"
3664.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
3665.Sh SEE ALSO
3666.Xr cpp 1 ,
3667.Xr m4 1 ,
3668.Xr altq 4 ,
3669.Xr divert 4 ,
3670.Xr dummynet 4 ,
3671.Xr if_bridge 4 ,
3672.Xr ip 4 ,
3673.Xr ipfirewall 4 ,
3674.Xr ng_ipfw 4 ,
3675.Xr protocols 5 ,
3676.Xr services 5 ,
3677.Xr init 8 ,
3678.Xr kldload 8 ,
3679.Xr reboot 8 ,
3680.Xr sysctl 8 ,
3681.Xr syslogd 8
3682.Sh HISTORY
3683The
3684.Nm
3685utility first appeared in
3686.Fx 2.0 .
3687.Nm dummynet
3688was introduced in
3689.Fx 2.2.8 .
3690Stateful extensions were introduced in
3691.Fx 4.0 .
3692.Nm ipfw2
3693was introduced in Summer 2002.
3694.Sh AUTHORS
3695.An Ugen J. S. Antsilevich ,
3696.An Poul-Henning Kamp ,
3697.An Alex Nash ,
3698.An Archie Cobbs ,
3699.An Luigi Rizzo .
3700.Pp
3701.An -nosplit
3702API based upon code written by
3703.An Daniel Boulet
3704for BSDI.
3705.Pp
3706Dummynet has been introduced by Luigi Rizzo in 1997-1998.
3707.Pp
3708Some early work (1999-2000) on the
3709.Nm dummynet
3710traffic shaper supported by Akamba Corp.
3711.Pp
3712The ipfw core (ipfw2) has been completely redesigned and
3713reimplemented by Luigi Rizzo in summer 2002.
3714Further
3715actions and
3716options have been added by various developer over the years.
3717.Pp
3718.An -nosplit
3719In-kernel NAT support written by
3720.An Paolo Pisati Aq Mt piso@FreeBSD.org
3721as part of a Summer of Code 2005 project.
3722.Pp
3723SCTP
3724.Nm nat
3725support has been developed by
3726.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
3727The primary developers and maintainers are David Hayes and Jason But.
3728For further information visit:
3729.Aq http://www.caia.swin.edu.au/urp/SONATA
3730.Pp
3731Delay profiles have been developed by Alessandro Cerri and
3732Luigi Rizzo, supported by the
3733European Commission within Projects Onelab and Onelab2.
3734.Sh BUGS
3735The syntax has grown over the years and sometimes it might be confusing.
3736Unfortunately, backward compatibility prevents cleaning up mistakes
3737made in the definition of the syntax.
3738.Pp
3739.Em !!! WARNING !!!
3740.Pp
3741Misconfiguring the firewall can put your computer in an unusable state,
3742possibly shutting down network services and requiring console access to
3743regain control of it.
3744.Pp
3745Incoming packet fragments diverted by
3746.Cm divert
3747are reassembled before delivery to the socket.
3748The action used on those packet is the one from the
3749rule which matches the first fragment of the packet.
3750.Pp
3751Packets diverted to userland, and then reinserted by a userland process
3752may lose various packet attributes.
3753The packet source interface name
3754will be preserved if it is shorter than 8 bytes and the userland process
3755saves and reuses the sockaddr_in
3756(as does
3757.Xr natd 8 ) ;
3758otherwise, it may be lost.
3759If a packet is reinserted in this manner, later rules may be incorrectly
3760applied, making the order of
3761.Cm divert
3762rules in the rule sequence very important.
3763.Pp
3764Dummynet drops all packets with IPv6 link-local addresses.
3765.Pp
3766Rules using
3767.Cm uid
3768or
3769.Cm gid
3770may not behave as expected.
3771In particular, incoming SYN packets may
3772have no uid or gid associated with them since they do not yet belong
3773to a TCP connection, and the uid/gid associated with a packet may not
3774be as expected if the associated process calls
3775.Xr setuid 2
3776or similar system calls.
3777.Pp
3778Rule syntax is subject to the command line environment and some patterns
3779may need to be escaped with the backslash character
3780or quoted appropriately.
3781.Pp
3782Due to the architecture of
3783.Xr libalias 3 ,
3784ipfw nat is not compatible with the TCP segmentation offloading (TSO).
3785Thus, to reliably nat your network traffic, please disable TSO
3786on your NICs using
3787.Xr ifconfig 8 .
3788.Pp
3789ICMP error messages are not implicitly matched by dynamic rules
3790for the respective conversations.
3791To avoid failures of network error detection and path MTU discovery,
3792ICMP error messages may need to be allowed explicitly through static
3793rules.
3794.Pp
3795Rules using
3796.Cm call
3797and
3798.Cm return
3799actions may lead to confusing behaviour if ruleset has mistakes,
3800and/or interaction with other subsystems (netgraph, dummynet, etc.) is used.
3801One possible case for this is packet leaving
3802.Nm
3803in subroutine on the input pass, while later on output encountering unpaired
3804.Cm return
3805first.
3806As the call stack is kept intact after input pass, packet will suddenly
3807return to the rule number used on input pass, not on output one.
3808Order of processing should be checked carefully to avoid such mistakes.
3809