xref: /freebsd/sbin/ipfw/ipfw.8 (revision 3b2324c3a800d7599f348c408f01908d0cef05a0)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd April 21, 2019
5.Dt IPFW 8
6.Os
7.Sh NAME
8.Nm ipfw
9.Nd User interface for firewall, traffic shaper, packet scheduler,
10in-kernel NAT.
11.Sh SYNOPSIS
12.Ss FIREWALL CONFIGURATION
13.Nm
14.Op Fl cq
15.Cm add
16.Ar rule
17.Nm
18.Op Fl acdefnNStT
19.Op Cm set Ar N
20.Brq Cm list | show
21.Op Ar rule | first-last ...
22.Nm
23.Op Fl f | q
24.Op Cm set Ar N
25.Cm flush
26.Nm
27.Op Fl q
28.Op Cm set Ar N
29.Brq Cm delete | zero | resetlog
30.Op Ar number ...
31.Pp
32.Nm
33.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
34.Nm
35.Cm set move
36.Op Cm rule
37.Ar number Cm to Ar number
38.Nm
39.Cm set swap Ar number number
40.Nm
41.Cm set show
42.Ss SYSCTL SHORTCUTS
43.Nm
44.Cm enable
45.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
46.Nm
47.Cm disable
48.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
49.Ss LOOKUP TABLES
50.Nm
51.Oo Cm set Ar N Oc Cm table Ar name Cm create Ar create-options
52.Nm
53.Oo Cm set Ar N Oc Cm table
54.Brq Ar name | all
55.Cm destroy
56.Nm
57.Oo Cm set Ar N Oc Cm table Ar name Cm modify Ar modify-options
58.Nm
59.Oo Cm set Ar N Oc Cm table Ar name Cm swap Ar name
60.Nm
61.Oo Cm set Ar N Oc Cm table Ar name Cm add Ar table-key Op Ar value
62.Nm
63.Oo Cm set Ar N Oc Cm table Ar name Cm add Op Ar table-key Ar value ...
64.Nm
65.Oo Cm set Ar N Oc Cm table Ar name Cm atomic add Op Ar table-key Ar value ...
66.Nm
67.Oo Cm set Ar N Oc Cm table Ar name Cm delete Op Ar table-key ...
68.Nm
69.Oo Cm set Ar N Oc Cm table Ar name Cm lookup Ar addr
70.Nm
71.Oo Cm set Ar N Oc Cm table Ar name Cm lock
72.Nm
73.Oo Cm set Ar N Oc Cm table Ar name Cm unlock
74.Nm
75.Oo Cm set Ar N Oc Cm table
76.Brq Ar name | all
77.Cm list
78.Nm
79.Oo Cm set Ar N Oc Cm table
80.Brq Ar name | all
81.Cm info
82.Nm
83.Oo Cm set Ar N Oc Cm table
84.Brq Ar name | all
85.Cm detail
86.Nm
87.Oo Cm set Ar N Oc Cm table
88.Brq Ar name | all
89.Cm flush
90.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
91.Nm
92.Brq Cm pipe | queue | sched
93.Ar number
94.Cm config
95.Ar config-options
96.Nm
97.Op Fl s Op Ar field
98.Brq Cm pipe | queue | sched
99.Brq Cm delete | list | show
100.Op Ar number ...
101.Ss IN-KERNEL NAT
102.Nm
103.Op Fl q
104.Cm nat
105.Ar number
106.Cm config
107.Ar config-options
108.Ss STATEFUL IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
109.Nm
110.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm create Ar create-options
111.Nm
112.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm config Ar config-options
113.Nm
114.Oo Cm set Ar N Oc Cm nat64lsn
115.Brq Ar name | all
116.Brq Cm list | show
117.Op Cm states
118.Nm
119.Oo Cm set Ar N Oc Cm nat64lsn
120.Brq Ar name | all
121.Cm destroy
122.Nm
123.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm stats Op Cm reset
124.Ss STATELESS IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
125.Nm
126.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm create Ar create-options
127.Nm
128.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm config Ar config-options
129.Nm
130.Oo Cm set Ar N Oc Cm nat64stl
131.Brq Ar name | all
132.Brq Cm list | show
133.Nm
134.Oo Cm set Ar N Oc Cm nat64stl
135.Brq Ar name | all
136.Cm destroy
137.Nm
138.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm stats Op Cm reset
139.Ss XLAT464 CLAT IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
140.Nm
141.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm create Ar create-options
142.Nm
143.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm config Ar config-options
144.Nm
145.Oo Cm set Ar N Oc Cm nat64clat
146.Brq Ar name | all
147.Brq Cm list | show
148.Nm
149.Oo Cm set Ar N Oc Cm nat64clat
150.Brq Ar name | all
151.Cm destroy
152.Nm
153.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm stats Op Cm reset
154.Ss IPv6-to-IPv6 NETWORK PREFIX TRANSLATION
155.Nm
156.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm create Ar create-options
157.Nm
158.Oo Cm set Ar N Oc Cm nptv6
159.Brq Ar name | all
160.Brq Cm list | show
161.Nm
162.Oo Cm set Ar N Oc Cm nptv6
163.Brq Ar name | all
164.Cm destroy
165.Nm
166.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm stats Op Cm reset
167.Ss INTERNAL DIAGNOSTICS
168.Nm
169.Cm internal iflist
170.Nm
171.Cm internal talist
172.Nm
173.Cm internal vlist
174.Ss LIST OF RULES AND PREPROCESSING
175.Nm
176.Op Fl cfnNqS
177.Oo
178.Fl p Ar preproc
179.Oo
180.Ar preproc-flags
181.Oc
182.Oc
183.Ar pathname
184.Sh DESCRIPTION
185The
186.Nm
187utility is the user interface for controlling the
188.Xr ipfw 4
189firewall, the
190.Xr dummynet 4
191traffic shaper/packet scheduler, and the
192in-kernel NAT services.
193.Pp
194A firewall configuration, or
195.Em ruleset ,
196is made of a list of
197.Em rules
198numbered from 1 to 65535.
199Packets are passed to the firewall
200from a number of different places in the protocol stack
201(depending on the source and destination of the packet,
202it is possible for the firewall to be
203invoked multiple times on the same packet).
204The packet passed to the firewall is compared
205against each of the rules in the
206.Em ruleset ,
207in rule-number order
208(multiple rules with the same number are permitted, in which case
209they are processed in order of insertion).
210When a match is found, the action corresponding to the
211matching rule is performed.
212.Pp
213Depending on the action and certain system settings, packets
214can be reinjected into the firewall at some rule after the
215matching one for further processing.
216.Pp
217A ruleset always includes a
218.Em default
219rule (numbered 65535) which cannot be modified or deleted,
220and matches all packets.
221The action associated with the
222.Em default
223rule can be either
224.Cm deny
225or
226.Cm allow
227depending on how the kernel is configured.
228.Pp
229If the ruleset includes one or more rules with the
230.Cm keep-state ,
231.Cm record-state ,
232.Cm limit
233or
234.Cm set-limit
235option,
236the firewall will have a
237.Em stateful
238behaviour, i.e., upon a match it will create
239.Em dynamic rules ,
240i.e., rules that match packets with the same 5-tuple
241(protocol, source and destination addresses and ports)
242as the packet which caused their creation.
243Dynamic rules, which have a limited lifetime, are checked
244at the first occurrence of a
245.Cm check-state ,
246.Cm keep-state
247or
248.Cm limit
249rule, and are typically used to open the firewall on-demand to
250legitimate traffic only.
251Please note, that
252.Cm keep-state
253and
254.Cm limit
255imply implicit
256.Cm check-state
257for all packets (not only these matched by the rule) but
258.Cm record-state
259and
260.Cm set-limit
261have no implicit
262.Cm check-state .
263See the
264.Sx STATEFUL FIREWALL
265and
266.Sx EXAMPLES
267Sections below for more information on the stateful behaviour of
268.Nm .
269.Pp
270All rules (including dynamic ones) have a few associated counters:
271a packet count, a byte count, a log count and a timestamp
272indicating the time of the last match.
273Counters can be displayed or reset with
274.Nm
275commands.
276.Pp
277Each rule belongs to one of 32 different
278.Em sets
279, and there are
280.Nm
281commands to atomically manipulate sets, such as enable,
282disable, swap sets, move all rules in a set to another
283one, delete all rules in a set.
284These can be useful to
285install temporary configurations, or to test them.
286See Section
287.Sx SETS OF RULES
288for more information on
289.Em sets .
290.Pp
291Rules can be added with the
292.Cm add
293command; deleted individually or in groups with the
294.Cm delete
295command, and globally (except those in set 31) with the
296.Cm flush
297command; displayed, optionally with the content of the
298counters, using the
299.Cm show
300and
301.Cm list
302commands.
303Finally, counters can be reset with the
304.Cm zero
305and
306.Cm resetlog
307commands.
308.Pp
309.Ss COMMAND OPTIONS
310The following general options are available when invoking
311.Nm :
312.Bl -tag -width indent
313.It Fl a
314Show counter values when listing rules.
315The
316.Cm show
317command implies this option.
318.It Fl b
319Only show the action and the comment, not the body of a rule.
320Implies
321.Fl c .
322.It Fl c
323When entering or showing rules, print them in compact form,
324i.e., omitting the "ip from any to any" string
325when this does not carry any additional information.
326.It Fl d
327When listing, show dynamic rules in addition to static ones.
328.It Fl D
329When listing, show only dynamic states.
330When deleting, delete only dynamic states.
331.It Fl f
332Run without prompting for confirmation for commands that can cause problems if misused,
333i.e.,
334.Cm flush .
335If there is no tty associated with the process, this is implied.
336The
337.Cm delete
338command with this flag ignores possible errors,
339i.e., nonexistent rule number.
340And for batched commands execution continues with the next command.
341.It Fl i
342When listing a table (see the
343.Sx LOOKUP TABLES
344section below for more information on lookup tables), format values
345as IP addresses.
346By default, values are shown as integers.
347.It Fl n
348Only check syntax of the command strings, without actually passing
349them to the kernel.
350.It Fl N
351Try to resolve addresses and service names in output.
352.It Fl q
353Be quiet when executing the
354.Cm add ,
355.Cm nat ,
356.Cm zero ,
357.Cm resetlog
358or
359.Cm flush
360commands;
361(implies
362.Fl f ) .
363This is useful when updating rulesets by executing multiple
364.Nm
365commands in a script
366(e.g.,
367.Ql sh\ /etc/rc.firewall ) ,
368or by processing a file with many
369.Nm
370rules across a remote login session.
371It also stops a table add or delete
372from failing if the entry already exists or is not present.
373.Pp
374The reason why this option may be important is that
375for some of these actions,
376.Nm
377may print a message; if the action results in blocking the
378traffic to the remote client,
379the remote login session will be closed
380and the rest of the ruleset will not be processed.
381Access to the console would then be required to recover.
382.It Fl S
383When listing rules, show the
384.Em set
385each rule belongs to.
386If this flag is not specified, disabled rules will not be
387listed.
388.It Fl s Op Ar field
389When listing pipes, sort according to one of the four
390counters (total or current packets or bytes).
391.It Fl t
392When listing, show last match timestamp converted with ctime().
393.It Fl T
394When listing, show last match timestamp as seconds from the epoch.
395This form can be more convenient for postprocessing by scripts.
396.El
397.Ss LIST OF RULES AND PREPROCESSING
398To ease configuration, rules can be put into a file which is
399processed using
400.Nm
401as shown in the last synopsis line.
402An absolute
403.Ar pathname
404must be used.
405The file will be read line by line and applied as arguments to the
406.Nm
407utility.
408.Pp
409Optionally, a preprocessor can be specified using
410.Fl p Ar preproc
411where
412.Ar pathname
413is to be piped through.
414Useful preprocessors include
415.Xr cpp 1
416and
417.Xr m4 1 .
418If
419.Ar preproc
420does not start with a slash
421.Pq Ql /
422as its first character, the usual
423.Ev PATH
424name search is performed.
425Care should be taken with this in environments where not all
426file systems are mounted (yet) by the time
427.Nm
428is being run (e.g.\& when they are mounted over NFS).
429Once
430.Fl p
431has been specified, any additional arguments are passed on to the preprocessor
432for interpretation.
433This allows for flexible configuration files (like conditionalizing
434them on the local hostname) and the use of macros to centralize
435frequently required arguments like IP addresses.
436.Ss TRAFFIC SHAPER CONFIGURATION
437The
438.Nm
439.Cm pipe , queue
440and
441.Cm sched
442commands are used to configure the traffic shaper and packet scheduler.
443See the
444.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
445Section below for details.
446.Pp
447If the world and the kernel get out of sync the
448.Nm
449ABI may break, preventing you from being able to add any rules.
450This can adversely affect the booting process.
451You can use
452.Nm
453.Cm disable
454.Cm firewall
455to temporarily disable the firewall to regain access to the network,
456allowing you to fix the problem.
457.Sh PACKET FLOW
458A packet is checked against the active ruleset in multiple places
459in the protocol stack, under control of several sysctl variables.
460These places and variables are shown below, and it is important to
461have this picture in mind in order to design a correct ruleset.
462.Bd -literal -offset indent
463       ^    to upper layers    V
464       |                       |
465       +----------->-----------+
466       ^                       V
467 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
468       |                       |
469       ^                       V
470 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
471       |                       |
472       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
473       ^                       V
474       |      to devices       |
475.Ed
476.Pp
477The number of
478times the same packet goes through the firewall can
479vary between 0 and 4 depending on packet source and
480destination, and system configuration.
481.Pp
482Note that as packets flow through the stack, headers can be
483stripped or added to it, and so they may or may not be available
484for inspection.
485E.g., incoming packets will include the MAC header when
486.Nm
487is invoked from
488.Cm ether_demux() ,
489but the same packets will have the MAC header stripped off when
490.Nm
491is invoked from
492.Cm ip_input()
493or
494.Cm ip6_input() .
495.Pp
496Also note that each packet is always checked against the complete ruleset,
497irrespective of the place where the check occurs, or the source of the packet.
498If a rule contains some match patterns or actions which are not valid
499for the place of invocation (e.g.\& trying to match a MAC header within
500.Cm ip_input
501or
502.Cm ip6_input ),
503the match pattern will not match, but a
504.Cm not
505operator in front of such patterns
506.Em will
507cause the pattern to
508.Em always
509match on those packets.
510It is thus the responsibility of
511the programmer, if necessary, to write a suitable ruleset to
512differentiate among the possible places.
513.Cm skipto
514rules can be useful here, as an example:
515.Bd -literal -offset indent
516# packets from ether_demux or bdg_forward
517ipfw add 10 skipto 1000 all from any to any layer2 in
518# packets from ip_input
519ipfw add 10 skipto 2000 all from any to any not layer2 in
520# packets from ip_output
521ipfw add 10 skipto 3000 all from any to any not layer2 out
522# packets from ether_output_frame
523ipfw add 10 skipto 4000 all from any to any layer2 out
524.Ed
525.Pp
526(yes, at the moment there is no way to differentiate between
527ether_demux and bdg_forward).
528.Pp
529Also note that only actions
530.Cm allow,
531.Cm deny,
532.Cm netgraph,
533.Cm ngtee
534and related to
535.Cm dummynet
536are processed for
537.Cm layer2
538frames and all other actions act as if they were
539.Cm allow
540for such frames.
541Full set of actions is supported for IP packets without
542.Cm layer2
543headers only.
544For example,
545.Cm divert
546action does not divert
547.Cm layer2
548frames.
549.Sh SYNTAX
550In general, each keyword or argument must be provided as
551a separate command line argument, with no leading or trailing
552spaces.
553Keywords are case-sensitive, whereas arguments may
554or may not be case-sensitive depending on their nature
555(e.g.\& uid's are, hostnames are not).
556.Pp
557Some arguments (e.g., port or address lists) are comma-separated
558lists of values.
559In this case, spaces after commas ',' are allowed to make
560the line more readable.
561You can also put the entire
562command (including flags) into a single argument.
563E.g., the following forms are equivalent:
564.Bd -literal -offset indent
565ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
566ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
567ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
568.Ed
569.Sh RULE FORMAT
570The format of firewall rules is the following:
571.Bd -ragged -offset indent
572.Bk -words
573.Op Ar rule_number
574.Op Cm set Ar set_number
575.Op Cm prob Ar match_probability
576.Ar action
577.Op Cm log Op Cm logamount Ar number
578.Op Cm altq Ar queue
579.Oo
580.Bro Cm tag | untag
581.Brc Ar number
582.Oc
583.Ar body
584.Ek
585.Ed
586.Pp
587where the body of the rule specifies which information is used
588for filtering packets, among the following:
589.Pp
590.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
591.It Layer-2 header fields
592When available
593.It IPv4 and IPv6 Protocol
594SCTP, TCP, UDP, ICMP, etc.
595.It Source and dest. addresses and ports
596.It Direction
597See Section
598.Sx PACKET FLOW
599.It Transmit and receive interface
600By name or address
601.It Misc. IP header fields
602Version, type of service, datagram length, identification,
603fragment flag (non-zero IP offset),
604Time To Live
605.It IP options
606.It IPv6 Extension headers
607Fragmentation, Hop-by-Hop options,
608Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
609.It IPv6 Flow-ID
610.It Misc. TCP header fields
611TCP flags (SYN, FIN, ACK, RST, etc.),
612sequence number, acknowledgment number,
613window
614.It TCP options
615.It ICMP types
616for ICMP packets
617.It ICMP6 types
618for ICMP6 packets
619.It User/group ID
620When the packet can be associated with a local socket.
621.It Divert status
622Whether a packet came from a divert socket (e.g.,
623.Xr natd 8 ) .
624.It Fib annotation state
625Whether a packet has been tagged for using a specific FIB (routing table)
626in future forwarding decisions.
627.El
628.Pp
629Note that some of the above information, e.g.\& source MAC or IP addresses and
630TCP/UDP ports, can be easily spoofed, so filtering on those fields
631alone might not guarantee the desired results.
632.Bl -tag -width indent
633.It Ar rule_number
634Each rule is associated with a
635.Ar rule_number
636in the range 1..65535, with the latter reserved for the
637.Em default
638rule.
639Rules are checked sequentially by rule number.
640Multiple rules can have the same number, in which case they are
641checked (and listed) according to the order in which they have
642been added.
643If a rule is entered without specifying a number, the kernel will
644assign one in such a way that the rule becomes the last one
645before the
646.Em default
647rule.
648Automatic rule numbers are assigned by incrementing the last
649non-default rule number by the value of the sysctl variable
650.Ar net.inet.ip.fw.autoinc_step
651which defaults to 100.
652If this is not possible (e.g.\& because we would go beyond the
653maximum allowed rule number), the number of the last
654non-default value is used instead.
655.It Cm set Ar set_number
656Each rule is associated with a
657.Ar set_number
658in the range 0..31.
659Sets can be individually disabled and enabled, so this parameter
660is of fundamental importance for atomic ruleset manipulation.
661It can be also used to simplify deletion of groups of rules.
662If a rule is entered without specifying a set number,
663set 0 will be used.
664.br
665Set 31 is special in that it cannot be disabled,
666and rules in set 31 are not deleted by the
667.Nm ipfw flush
668command (but you can delete them with the
669.Nm ipfw delete set 31
670command).
671Set 31 is also used for the
672.Em default
673rule.
674.It Cm prob Ar match_probability
675A match is only declared with the specified probability
676(floating point number between 0 and 1).
677This can be useful for a number of applications such as
678random packet drop or
679(in conjunction with
680.Nm dummynet )
681to simulate the effect of multiple paths leading to out-of-order
682packet delivery.
683.Pp
684Note: this condition is checked before any other condition, including
685ones such as
686.Cm keep-state
687or
688.Cm check-state
689which might have
690side effects.
691.It Cm log Op Cm logamount Ar number
692Packets matching a rule with the
693.Cm log
694keyword will be made available for logging in two ways:
695if the sysctl variable
696.Va net.inet.ip.fw.verbose
697is set to 0 (default), one can use
698.Xr bpf 4
699attached to the
700.Li ipfw0
701pseudo interface.
702This pseudo interface can be created manually after a system
703boot by using the following command:
704.Bd -literal -offset indent
705# ifconfig ipfw0 create
706.Ed
707.Pp
708Or, automatically at boot time by adding the following
709line to the
710.Xr rc.conf 5
711file:
712.Bd -literal -offset indent
713firewall_logif="YES"
714.Ed
715.Pp
716There is zero overhead when no
717.Xr bpf 4
718is attached to the pseudo interface.
719.Pp
720If
721.Va net.inet.ip.fw.verbose
722is set to 1, packets will be logged to
723.Xr syslogd 8
724with a
725.Dv LOG_SECURITY
726facility up to a maximum of
727.Cm logamount
728packets.
729If no
730.Cm logamount
731is specified, the limit is taken from the sysctl variable
732.Va net.inet.ip.fw.verbose_limit .
733In both cases, a value of 0 means unlimited logging.
734.Pp
735Once the limit is reached, logging can be re-enabled by
736clearing the logging counter or the packet counter for that entry, see the
737.Cm resetlog
738command.
739.Pp
740Note: logging is done after all other packet matching conditions
741have been successfully verified, and before performing the final
742action (accept, deny, etc.) on the packet.
743.It Cm tag Ar number
744When a packet matches a rule with the
745.Cm tag
746keyword, the numeric tag for the given
747.Ar number
748in the range 1..65534 will be attached to the packet.
749The tag acts as an internal marker (it is not sent out over
750the wire) that can be used to identify these packets later on.
751This can be used, for example, to provide trust between interfaces
752and to start doing policy-based filtering.
753A packet can have multiple tags at the same time.
754Tags are "sticky", meaning once a tag is applied to a packet by a
755matching rule it exists until explicit removal.
756Tags are kept with the packet everywhere within the kernel, but are
757lost when packet leaves the kernel, for example, on transmitting
758packet out to the network or sending packet to a
759.Xr divert 4
760socket.
761.Pp
762To check for previously applied tags, use the
763.Cm tagged
764rule option.
765To delete previously applied tag, use the
766.Cm untag
767keyword.
768.Pp
769Note: since tags are kept with the packet everywhere in kernelspace,
770they can be set and unset anywhere in the kernel network subsystem
771(using the
772.Xr mbuf_tags 9
773facility), not only by means of the
774.Xr ipfw 4
775.Cm tag
776and
777.Cm untag
778keywords.
779For example, there can be a specialized
780.Xr netgraph 4
781node doing traffic analyzing and tagging for later inspecting
782in firewall.
783.It Cm untag Ar number
784When a packet matches a rule with the
785.Cm untag
786keyword, the tag with the number
787.Ar number
788is searched among the tags attached to this packet and,
789if found, removed from it.
790Other tags bound to packet, if present, are left untouched.
791.It Cm altq Ar queue
792When a packet matches a rule with the
793.Cm altq
794keyword, the ALTQ identifier for the given
795.Ar queue
796(see
797.Xr altq 4 )
798will be attached.
799Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
800and not being rejected or going to divert sockets.
801Note that if there is insufficient memory at the time the packet is
802processed, it will not be tagged, so it is wise to make your ALTQ
803"default" queue policy account for this.
804If multiple
805.Cm altq
806rules match a single packet, only the first one adds the ALTQ classification
807tag.
808In doing so, traffic may be shaped by using
809.Cm count Cm altq Ar queue
810rules for classification early in the ruleset, then later applying
811the filtering decision.
812For example,
813.Cm check-state
814and
815.Cm keep-state
816rules may come later and provide the actual filtering decisions in
817addition to the fallback ALTQ tag.
818.Pp
819You must run
820.Xr pfctl 8
821to set up the queues before IPFW will be able to look them up by name,
822and if the ALTQ disciplines are rearranged, the rules in containing the
823queue identifiers in the kernel will likely have gone stale and need
824to be reloaded.
825Stale queue identifiers will probably result in misclassification.
826.Pp
827All system ALTQ processing can be turned on or off via
828.Nm
829.Cm enable Ar altq
830and
831.Nm
832.Cm disable Ar altq .
833The usage of
834.Va net.inet.ip.fw.one_pass
835is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
836always after adding an ALTQ tag.
837.El
838.Ss RULE ACTIONS
839A rule can be associated with one of the following actions, which
840will be executed when the packet matches the body of the rule.
841.Bl -tag -width indent
842.It Cm allow | accept | pass | permit
843Allow packets that match rule.
844The search terminates.
845.It Cm check-state Op Ar :flowname | Cm :any
846Checks the packet against the dynamic ruleset.
847If a match is found, execute the action associated with
848the rule which generated this dynamic rule, otherwise
849move to the next rule.
850.br
851.Cm Check-state
852rules do not have a body.
853If no
854.Cm check-state
855rule is found, the dynamic ruleset is checked at the first
856.Cm keep-state
857or
858.Cm limit
859rule.
860The
861.Ar :flowname
862is symbolic name assigned to dynamic rule by
863.Cm keep-state
864opcode.
865The special flowname
866.Cm :any
867can be used to ignore states flowname when matching.
868The
869.Cm :default
870keyword is special name used for compatibility with old rulesets.
871.It Cm count
872Update counters for all packets that match rule.
873The search continues with the next rule.
874.It Cm deny | drop
875Discard packets that match this rule.
876The search terminates.
877.It Cm divert Ar port
878Divert packets that match this rule to the
879.Xr divert 4
880socket bound to port
881.Ar port .
882The search terminates.
883.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
884Change the next-hop on matching packets to
885.Ar ipaddr ,
886which can be an IP address or a host name.
887The next hop can also be supplied by the last table
888looked up for the packet by using the
889.Cm tablearg
890keyword instead of an explicit address.
891The search terminates if this rule matches.
892.Pp
893If
894.Ar ipaddr
895is a local address, then matching packets will be forwarded to
896.Ar port
897(or the port number in the packet if one is not specified in the rule)
898on the local machine.
899.br
900If
901.Ar ipaddr
902is not a local address, then the port number
903(if specified) is ignored, and the packet will be
904forwarded to the remote address, using the route as found in
905the local routing table for that IP.
906.br
907A
908.Ar fwd
909rule will not match layer-2 packets (those received
910on ether_input, ether_output, or bridged).
911.br
912The
913.Cm fwd
914action does not change the contents of the packet at all.
915In particular, the destination address remains unmodified, so
916packets forwarded to another system will usually be rejected by that system
917unless there is a matching rule on that system to capture them.
918For packets forwarded locally,
919the local address of the socket will be
920set to the original destination address of the packet.
921This makes the
922.Xr netstat 1
923entry look rather weird but is intended for
924use with transparent proxy servers.
925.It Cm nat Ar nat_nr | tablearg
926Pass packet to a
927nat instance
928(for network address translation, address redirect, etc.):
929see the
930.Sx NETWORK ADDRESS TRANSLATION (NAT)
931Section for further information.
932.It Cm nat64lsn Ar name
933Pass packet to a stateful NAT64 instance (for IPv6/IPv4 network address and
934protocol translation): see the
935.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
936Section for further information.
937.It Cm nat64stl Ar name
938Pass packet to a stateless NAT64 instance (for IPv6/IPv4 network address and
939protocol translation): see the
940.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
941Section for further information.
942.It Cm nat64clat Ar name
943Pass packet to a CLAT NAT64 instance (for client-side IPv6/IPv4 network address and
944protocol translation): see the
945.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
946Section for further information.
947.It Cm nptv6 Ar name
948Pass packet to a NPTv6 instance (for IPv6-to-IPv6 network prefix translation):
949see the
950.Sx IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
951Section for further information.
952.It Cm pipe Ar pipe_nr
953Pass packet to a
954.Nm dummynet
955.Dq pipe
956(for bandwidth limitation, delay, etc.).
957See the
958.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
959Section for further information.
960The search terminates; however, on exit from the pipe and if
961the
962.Xr sysctl 8
963variable
964.Va net.inet.ip.fw.one_pass
965is not set, the packet is passed again to the firewall code
966starting from the next rule.
967.It Cm queue Ar queue_nr
968Pass packet to a
969.Nm dummynet
970.Dq queue
971(for bandwidth limitation using WF2Q+).
972.It Cm reject
973(Deprecated).
974Synonym for
975.Cm unreach host .
976.It Cm reset
977Discard packets that match this rule, and if the
978packet is a TCP packet, try to send a TCP reset (RST) notice.
979The search terminates.
980.It Cm reset6
981Discard packets that match this rule, and if the
982packet is a TCP packet, try to send a TCP reset (RST) notice.
983The search terminates.
984.It Cm skipto Ar number | tablearg
985Skip all subsequent rules numbered less than
986.Ar number .
987The search continues with the first rule numbered
988.Ar number
989or higher.
990It is possible to use the
991.Cm tablearg
992keyword with a skipto for a
993.Em computed
994skipto. Skipto may work either in O(log(N)) or in O(1) depending
995on amount of memory and/or sysctl variables.
996See the
997.Sx SYSCTL VARIABLES
998section for more details.
999.It Cm call Ar number | tablearg
1000The current rule number is saved in the internal stack and
1001ruleset processing continues with the first rule numbered
1002.Ar number
1003or higher.
1004If later a rule with the
1005.Cm return
1006action is encountered, the processing returns to the first rule
1007with number of this
1008.Cm call
1009rule plus one or higher
1010(the same behaviour as with packets returning from
1011.Xr divert 4
1012socket after a
1013.Cm divert
1014action).
1015This could be used to make somewhat like an assembly language
1016.Dq subroutine
1017calls to rules with common checks for different interfaces, etc.
1018.Pp
1019Rule with any number could be called, not just forward jumps as with
1020.Cm skipto .
1021So, to prevent endless loops in case of mistakes, both
1022.Cm call
1023and
1024.Cm return
1025actions don't do any jumps and simply go to the next rule if memory
1026cannot be allocated or stack overflowed/underflowed.
1027.Pp
1028Internally stack for rule numbers is implemented using
1029.Xr mbuf_tags 9
1030facility and currently has size of 16 entries.
1031As mbuf tags are lost when packet leaves the kernel,
1032.Cm divert
1033should not be used in subroutines to avoid endless loops
1034and other undesired effects.
1035.It Cm return
1036Takes rule number saved to internal stack by the last
1037.Cm call
1038action and returns ruleset processing to the first rule
1039with number greater than number of corresponding
1040.Cm call
1041rule.
1042See description of the
1043.Cm call
1044action for more details.
1045.Pp
1046Note that
1047.Cm return
1048rules usually end a
1049.Dq subroutine
1050and thus are unconditional, but
1051.Nm
1052command-line utility currently requires every action except
1053.Cm check-state
1054to have body.
1055While it is sometimes useful to return only on some packets,
1056usually you want to print just
1057.Dq return
1058for readability.
1059A workaround for this is to use new syntax and
1060.Fl c
1061switch:
1062.Bd -literal -offset indent
1063# Add a rule without actual body
1064ipfw add 2999 return via any
1065
1066# List rules without "from any to any" part
1067ipfw -c list
1068.Ed
1069.Pp
1070This cosmetic annoyance may be fixed in future releases.
1071.It Cm tee Ar port
1072Send a copy of packets matching this rule to the
1073.Xr divert 4
1074socket bound to port
1075.Ar port .
1076The search continues with the next rule.
1077.It Cm unreach Ar code
1078Discard packets that match this rule, and try to send an ICMP
1079unreachable notice with code
1080.Ar code ,
1081where
1082.Ar code
1083is a number from 0 to 255, or one of these aliases:
1084.Cm net , host , protocol , port ,
1085.Cm needfrag , srcfail , net-unknown , host-unknown ,
1086.Cm isolated , net-prohib , host-prohib , tosnet ,
1087.Cm toshost , filter-prohib , host-precedence
1088or
1089.Cm precedence-cutoff .
1090The search terminates.
1091.It Cm unreach6 Ar code
1092Discard packets that match this rule, and try to send an ICMPv6
1093unreachable notice with code
1094.Ar code ,
1095where
1096.Ar code
1097is a number from 0, 1, 3 or 4, or one of these aliases:
1098.Cm no-route, admin-prohib, address
1099or
1100.Cm port .
1101The search terminates.
1102.It Cm netgraph Ar cookie
1103Divert packet into netgraph with given
1104.Ar cookie .
1105The search terminates.
1106If packet is later returned from netgraph it is either
1107accepted or continues with the next rule, depending on
1108.Va net.inet.ip.fw.one_pass
1109sysctl variable.
1110.It Cm ngtee Ar cookie
1111A copy of packet is diverted into netgraph, original
1112packet continues with the next rule.
1113See
1114.Xr ng_ipfw 4
1115for more information on
1116.Cm netgraph
1117and
1118.Cm ngtee
1119actions.
1120.It Cm setfib Ar fibnum | tablearg
1121The packet is tagged so as to use the FIB (routing table)
1122.Ar fibnum
1123in any subsequent forwarding decisions.
1124In the current implementation, this is limited to the values 0 through 15, see
1125.Xr setfib 2 .
1126Processing continues at the next rule.
1127It is possible to use the
1128.Cm tablearg
1129keyword with setfib.
1130If the tablearg value is not within the compiled range of fibs,
1131the packet's fib is set to 0.
1132.It Cm setdscp Ar DSCP | number | tablearg
1133Set specified DiffServ codepoint for an IPv4/IPv6 packet.
1134Processing continues at the next rule.
1135Supported values are:
1136.Pp
1137.Cm cs0
1138.Pq Dv 000000 ,
1139.Cm cs1
1140.Pq Dv 001000 ,
1141.Cm cs2
1142.Pq Dv 010000 ,
1143.Cm cs3
1144.Pq Dv 011000 ,
1145.Cm cs4
1146.Pq Dv 100000 ,
1147.Cm cs5
1148.Pq Dv 101000 ,
1149.Cm cs6
1150.Pq Dv 110000 ,
1151.Cm cs7
1152.Pq Dv 111000 ,
1153.Cm af11
1154.Pq Dv 001010 ,
1155.Cm af12
1156.Pq Dv 001100 ,
1157.Cm af13
1158.Pq Dv 001110 ,
1159.Cm af21
1160.Pq Dv 010010 ,
1161.Cm af22
1162.Pq Dv 010100 ,
1163.Cm af23
1164.Pq Dv 010110 ,
1165.Cm af31
1166.Pq Dv 011010 ,
1167.Cm af32
1168.Pq Dv 011100 ,
1169.Cm af33
1170.Pq Dv 011110 ,
1171.Cm af41
1172.Pq Dv 100010 ,
1173.Cm af42
1174.Pq Dv 100100 ,
1175.Cm af43
1176.Pq Dv 100110 ,
1177.Cm ef
1178.Pq Dv 101110 ,
1179.Cm be
1180.Pq Dv 000000 .
1181Additionally, DSCP value can be specified by number (0..63).
1182It is also possible to use the
1183.Cm tablearg
1184keyword with setdscp.
1185If the tablearg value is not within the 0..63 range, lower 6 bits of supplied
1186value are used.
1187.It Cm tcp-setmss Ar mss
1188Set the Maximum Segment Size (MSS) in the TCP segment to value
1189.Ar mss .
1190The kernel module
1191.Cm ipfw_pmod
1192should be loaded or kernel should have
1193.Cm options IPFIREWALL_PMOD
1194to be able use this action.
1195This command does not change a packet if original MSS value is lower than
1196specified value.
1197Both TCP over IPv4 and over IPv6 are supported.
1198Regardless of matched a packet or not by the
1199.Cm tcp-setmss
1200rule, the search continues with the next rule.
1201.It Cm reass
1202Queue and reassemble IPv4 fragments.
1203If the packet is not fragmented, counters are updated and
1204processing continues with the next rule.
1205If the packet is the last logical fragment, the packet is reassembled and, if
1206.Va net.inet.ip.fw.one_pass
1207is set to 0, processing continues with the next rule.
1208Otherwise, the packet is allowed to pass and the search terminates.
1209If the packet is a fragment in the middle of a logical group of fragments,
1210it is consumed and
1211processing stops immediately.
1212.Pp
1213Fragment handling can be tuned via
1214.Va net.inet.ip.maxfragpackets
1215and
1216.Va net.inet.ip.maxfragsperpacket
1217which limit, respectively, the maximum number of processable
1218fragments (default: 800) and
1219the maximum number of fragments per packet (default: 16).
1220.Pp
1221NOTA BENE: since fragments do not contain port numbers,
1222they should be avoided with the
1223.Nm reass
1224rule.
1225Alternatively, direction-based (like
1226.Nm in
1227/
1228.Nm out
1229) and source-based (like
1230.Nm via
1231) match patterns can be used to select fragments.
1232.Pp
1233Usually a simple rule like:
1234.Bd -literal -offset indent
1235# reassemble incoming fragments
1236ipfw add reass all from any to any in
1237.Ed
1238.Pp
1239is all you need at the beginning of your ruleset.
1240.It Cm abort
1241Discard packets that match this rule, and if the packet is an SCTP packet,
1242try to send an SCTP packet containing an ABORT chunk.
1243The search terminates.
1244.It Cm abort6
1245Discard packets that match this rule, and if the packet is an SCTP packet,
1246try to send an SCTP packet containing an ABORT chunk.
1247The search terminates.
1248.El
1249.Ss RULE BODY
1250The body of a rule contains zero or more patterns (such as
1251specific source and destination addresses or ports,
1252protocol options, incoming or outgoing interfaces, etc.)
1253that the packet must match in order to be recognised.
1254In general, the patterns are connected by (implicit)
1255.Cm and
1256operators -- i.e., all must match in order for the
1257rule to match.
1258Individual patterns can be prefixed by the
1259.Cm not
1260operator to reverse the result of the match, as in
1261.Pp
1262.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
1263.Pp
1264Additionally, sets of alternative match patterns
1265.Pq Em or-blocks
1266can be constructed by putting the patterns in
1267lists enclosed between parentheses ( ) or braces { }, and
1268using the
1269.Cm or
1270operator as follows:
1271.Pp
1272.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
1273.Pp
1274Only one level of parentheses is allowed.
1275Beware that most shells have special meanings for parentheses
1276or braces, so it is advisable to put a backslash \\ in front of them
1277to prevent such interpretations.
1278.Pp
1279The body of a rule must in general include a source and destination
1280address specifier.
1281The keyword
1282.Ar any
1283can be used in various places to specify that the content of
1284a required field is irrelevant.
1285.Pp
1286The rule body has the following format:
1287.Bd -ragged -offset indent
1288.Op Ar proto Cm from Ar src Cm to Ar dst
1289.Op Ar options
1290.Ed
1291.Pp
1292The first part (proto from src to dst) is for backward
1293compatibility with earlier versions of
1294.Fx .
1295In modern
1296.Fx
1297any match pattern (including MAC headers, IP protocols,
1298addresses and ports) can be specified in the
1299.Ar options
1300section.
1301.Pp
1302Rule fields have the following meaning:
1303.Bl -tag -width indent
1304.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
1305.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
1306An IP protocol specified by number or name
1307(for a complete list see
1308.Pa /etc/protocols ) ,
1309or one of the following keywords:
1310.Bl -tag -width indent
1311.It Cm ip4 | ipv4
1312Matches IPv4 packets.
1313.It Cm ip6 | ipv6
1314Matches IPv6 packets.
1315.It Cm ip | all
1316Matches any packet.
1317.El
1318.Pp
1319The
1320.Cm ipv6
1321in
1322.Cm proto
1323option will be treated as inner protocol.
1324And, the
1325.Cm ipv4
1326is not available in
1327.Cm proto
1328option.
1329.Pp
1330The
1331.Cm { Ar protocol Cm or ... }
1332format (an
1333.Em or-block )
1334is provided for convenience only but its use is deprecated.
1335.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1336An address (or a list, see below)
1337optionally followed by
1338.Ar ports
1339specifiers.
1340.Pp
1341The second format
1342.Em ( or-block
1343with multiple addresses) is provided for convenience only and
1344its use is discouraged.
1345.It Ar addr : Oo Cm not Oc Bro
1346.Cm any | me | me6 |
1347.Cm table Ns Pq Ar name Ns Op , Ns Ar value
1348.Ar | addr-list | addr-set
1349.Brc
1350.Bl -tag -width indent
1351.It Cm any
1352Matches any IP address.
1353.It Cm me
1354Matches any IP address configured on an interface in the system.
1355.It Cm me6
1356Matches any IPv6 address configured on an interface in the system.
1357The address list is evaluated at the time the packet is
1358analysed.
1359.It Cm table Ns Pq Ar name Ns Op , Ns Ar value
1360Matches any IPv4 or IPv6 address for which an entry exists in the lookup table
1361.Ar number .
1362If an optional 32-bit unsigned
1363.Ar value
1364is also specified, an entry will match only if it has this value.
1365See the
1366.Sx LOOKUP TABLES
1367section below for more information on lookup tables.
1368.El
1369.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1370.It Ar ip-addr :
1371A host or subnet address specified in one of the following ways:
1372.Bl -tag -width indent
1373.It Ar numeric-ip | hostname
1374Matches a single IPv4 address, specified as dotted-quad or a hostname.
1375Hostnames are resolved at the time the rule is added to the firewall list.
1376.It Ar addr Ns / Ns Ar masklen
1377Matches all addresses with base
1378.Ar addr
1379(specified as an IP address, a network number, or a hostname)
1380and mask width of
1381.Cm masklen
1382bits.
1383As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1384all IP numbers from 1.2.3.0 to 1.2.3.127 .
1385.It Ar addr Ns : Ns Ar mask
1386Matches all addresses with base
1387.Ar addr
1388(specified as an IP address, a network number, or a hostname)
1389and the mask of
1390.Ar mask ,
1391specified as a dotted quad.
1392As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
13931.*.3.*.
1394This form is advised only for non-contiguous
1395masks.
1396It is better to resort to the
1397.Ar addr Ns / Ns Ar masklen
1398format for contiguous masks, which is more compact and less
1399error-prone.
1400.El
1401.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1402.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1403Matches all addresses with base address
1404.Ar addr
1405(specified as an IP address, a network number, or a hostname)
1406and whose last byte is in the list between braces { } .
1407Note that there must be no spaces between braces and
1408numbers (spaces after commas are allowed).
1409Elements of the list can be specified as single entries
1410or ranges.
1411The
1412.Ar masklen
1413field is used to limit the size of the set of addresses,
1414and can have any value between 24 and 32.
1415If not specified,
1416it will be assumed as 24.
1417.br
1418This format is particularly useful to handle sparse address sets
1419within a single rule.
1420Because the matching occurs using a
1421bitmask, it takes constant time and dramatically reduces
1422the complexity of rulesets.
1423.br
1424As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1425or 1.2.3.0/24{128,35-55,89}
1426will match the following IP addresses:
1427.br
14281.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1429.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1430.It Ar ip6-addr :
1431A host or subnet specified one of the following ways:
1432.Bl -tag -width indent
1433.It Ar numeric-ip | hostname
1434Matches a single IPv6 address as allowed by
1435.Xr inet_pton 3
1436or a hostname.
1437Hostnames are resolved at the time the rule is added to the firewall
1438list.
1439.It Ar addr Ns / Ns Ar masklen
1440Matches all IPv6 addresses with base
1441.Ar addr
1442(specified as allowed by
1443.Xr inet_pton
1444or a hostname)
1445and mask width of
1446.Cm masklen
1447bits.
1448.It Ar addr Ns / Ns Ar mask
1449Matches all IPv6 addresses with base
1450.Ar addr
1451(specified as allowed by
1452.Xr inet_pton
1453or a hostname)
1454and the mask of
1455.Ar mask ,
1456specified as allowed by
1457.Xr inet_pton.
1458As an example, fe::640:0:0/ffff::ffff:ffff:0:0 will match
1459fe:*:*:*:0:640:*:*.
1460This form is advised only for non-contiguous
1461masks.
1462It is better to resort to the
1463.Ar addr Ns / Ns Ar masklen
1464format for contiguous masks, which is more compact and less
1465error-prone.
1466.El
1467.Pp
1468No support for sets of IPv6 addresses is provided because IPv6 addresses
1469are typically random past the initial prefix.
1470.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1471For protocols which support port numbers (such as SCTP, TCP and UDP), optional
1472.Cm ports
1473may be specified as one or more ports or port ranges, separated
1474by commas but no spaces, and an optional
1475.Cm not
1476operator.
1477The
1478.Ql \&-
1479notation specifies a range of ports (including boundaries).
1480.Pp
1481Service names (from
1482.Pa /etc/services )
1483may be used instead of numeric port values.
1484The length of the port list is limited to 30 ports or ranges,
1485though one can specify larger ranges by using an
1486.Em or-block
1487in the
1488.Cm options
1489section of the rule.
1490.Pp
1491A backslash
1492.Pq Ql \e
1493can be used to escape the dash
1494.Pq Ql -
1495character in a service name (from a shell, the backslash must be
1496typed twice to avoid the shell itself interpreting it as an escape
1497character).
1498.Pp
1499.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1500.Pp
1501Fragmented packets which have a non-zero offset (i.e., not the first
1502fragment) will never match a rule which has one or more port
1503specifications.
1504See the
1505.Cm frag
1506option for details on matching fragmented packets.
1507.El
1508.Ss RULE OPTIONS (MATCH PATTERNS)
1509Additional match patterns can be used within
1510rules.
1511Zero or more of these so-called
1512.Em options
1513can be present in a rule, optionally prefixed by the
1514.Cm not
1515operand, and possibly grouped into
1516.Em or-blocks .
1517.Pp
1518The following match patterns can be used (listed in alphabetical order):
1519.Bl -tag -width indent
1520.It Cm // this is a comment.
1521Inserts the specified text as a comment in the rule.
1522Everything following // is considered as a comment and stored in the rule.
1523You can have comment-only rules, which are listed as having a
1524.Cm count
1525action followed by the comment.
1526.It Cm bridged
1527Alias for
1528.Cm layer2 .
1529.It Cm defer-immediate-action | defer-action
1530A rule with this option will not perform normal action
1531upon a match. This option is intended to be used with
1532.Cm record-state
1533or
1534.Cm keep-state
1535as the dynamic rule, created but ignored on match, will work
1536as intended.
1537Rules with both
1538.Cm record-state
1539and
1540.Cm defer-immediate-action
1541create a dynamic rule and continue with the next rule without actually
1542performing the action part of this rule. When the rule is later activated
1543via the state table, the action is performed as usual.
1544.It Cm diverted
1545Matches only packets generated by a divert socket.
1546.It Cm diverted-loopback
1547Matches only packets coming from a divert socket back into the IP stack
1548input for delivery.
1549.It Cm diverted-output
1550Matches only packets going from a divert socket back outward to the IP
1551stack output for delivery.
1552.It Cm dst-ip Ar ip-address
1553Matches IPv4 packets whose destination IP is one of the address(es)
1554specified as argument.
1555.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1556Matches IPv6 packets whose destination IP is one of the address(es)
1557specified as argument.
1558.It Cm dst-port Ar ports
1559Matches IP packets whose destination port is one of the port(s)
1560specified as argument.
1561.It Cm established
1562Matches TCP packets that have the RST or ACK bits set.
1563.It Cm ext6hdr Ar header
1564Matches IPv6 packets containing the extended header given by
1565.Ar header .
1566Supported headers are:
1567.Pp
1568Fragment,
1569.Pq Cm frag ,
1570Hop-to-hop options
1571.Pq Cm hopopt ,
1572any type of Routing Header
1573.Pq Cm route ,
1574Source routing Routing Header Type 0
1575.Pq Cm rthdr0 ,
1576Mobile IPv6 Routing Header Type 2
1577.Pq Cm rthdr2 ,
1578Destination options
1579.Pq Cm dstopt ,
1580IPSec authentication headers
1581.Pq Cm ah ,
1582and IPsec encapsulated security payload headers
1583.Pq Cm esp .
1584.It Cm fib Ar fibnum
1585Matches a packet that has been tagged to use
1586the given FIB (routing table) number.
1587.It Cm flow Ar table Ns Pq Ar name Ns Op , Ns Ar value
1588Search for the flow entry in lookup table
1589.Ar name .
1590If not found, the match fails.
1591Otherwise, the match succeeds and
1592.Cm tablearg
1593is set to the value extracted from the table.
1594.Pp
1595This option can be useful to quickly dispatch traffic based on
1596certain packet fields.
1597See the
1598.Sx LOOKUP TABLES
1599section below for more information on lookup tables.
1600.It Cm flow-id Ar labels
1601Matches IPv6 packets containing any of the flow labels given in
1602.Ar labels .
1603.Ar labels
1604is a comma separated list of numeric flow labels.
1605.It Cm frag
1606Matches packets that are fragments and not the first
1607fragment of an IP datagram.
1608Note that these packets will not have
1609the next protocol header (e.g.\& TCP, UDP) so options that look into
1610these headers cannot match.
1611.It Cm gid Ar group
1612Matches all TCP or UDP packets sent by or received for a
1613.Ar group .
1614A
1615.Ar group
1616may be specified by name or number.
1617.It Cm jail Ar jail
1618Matches all TCP or UDP packets sent by or received for the
1619jail whose ID or name is
1620.Ar jail .
1621.It Cm icmptypes Ar types
1622Matches ICMP packets whose ICMP type is in the list
1623.Ar types .
1624The list may be specified as any combination of
1625individual types (numeric) separated by commas.
1626.Em Ranges are not allowed .
1627The supported ICMP types are:
1628.Pp
1629echo reply
1630.Pq Cm 0 ,
1631destination unreachable
1632.Pq Cm 3 ,
1633source quench
1634.Pq Cm 4 ,
1635redirect
1636.Pq Cm 5 ,
1637echo request
1638.Pq Cm 8 ,
1639router advertisement
1640.Pq Cm 9 ,
1641router solicitation
1642.Pq Cm 10 ,
1643time-to-live exceeded
1644.Pq Cm 11 ,
1645IP header bad
1646.Pq Cm 12 ,
1647timestamp request
1648.Pq Cm 13 ,
1649timestamp reply
1650.Pq Cm 14 ,
1651information request
1652.Pq Cm 15 ,
1653information reply
1654.Pq Cm 16 ,
1655address mask request
1656.Pq Cm 17
1657and address mask reply
1658.Pq Cm 18 .
1659.It Cm icmp6types Ar types
1660Matches ICMP6 packets whose ICMP6 type is in the list of
1661.Ar types .
1662The list may be specified as any combination of
1663individual types (numeric) separated by commas.
1664.Em Ranges are not allowed .
1665.It Cm in | out
1666Matches incoming or outgoing packets, respectively.
1667.Cm in
1668and
1669.Cm out
1670are mutually exclusive (in fact,
1671.Cm out
1672is implemented as
1673.Cm not in Ns No ).
1674.It Cm ipid Ar id-list
1675Matches IPv4 packets whose
1676.Cm ip_id
1677field has value included in
1678.Ar id-list ,
1679which is either a single value or a list of values or ranges
1680specified in the same way as
1681.Ar ports .
1682.It Cm iplen Ar len-list
1683Matches IP packets whose total length, including header and data, is
1684in the set
1685.Ar len-list ,
1686which is either a single value or a list of values or ranges
1687specified in the same way as
1688.Ar ports .
1689.It Cm ipoptions Ar spec
1690Matches packets whose IPv4 header contains the comma separated list of
1691options specified in
1692.Ar spec .
1693The supported IP options are:
1694.Pp
1695.Cm ssrr
1696(strict source route),
1697.Cm lsrr
1698(loose source route),
1699.Cm rr
1700(record packet route) and
1701.Cm ts
1702(timestamp).
1703The absence of a particular option may be denoted
1704with a
1705.Ql \&! .
1706.It Cm ipprecedence Ar precedence
1707Matches IPv4 packets whose precedence field is equal to
1708.Ar precedence .
1709.It Cm ipsec
1710Matches packets that have IPSEC history associated with them
1711(i.e., the packet comes encapsulated in IPSEC, the kernel
1712has IPSEC support, and can correctly decapsulate it).
1713.Pp
1714Note that specifying
1715.Cm ipsec
1716is different from specifying
1717.Cm proto Ar ipsec
1718as the latter will only look at the specific IP protocol field,
1719irrespective of IPSEC kernel support and the validity of the IPSEC data.
1720.Pp
1721Further note that this flag is silently ignored in kernels without
1722IPSEC support.
1723It does not affect rule processing when given and the
1724rules are handled as if with no
1725.Cm ipsec
1726flag.
1727.It Cm iptos Ar spec
1728Matches IPv4 packets whose
1729.Cm tos
1730field contains the comma separated list of
1731service types specified in
1732.Ar spec .
1733The supported IP types of service are:
1734.Pp
1735.Cm lowdelay
1736.Pq Dv IPTOS_LOWDELAY ,
1737.Cm throughput
1738.Pq Dv IPTOS_THROUGHPUT ,
1739.Cm reliability
1740.Pq Dv IPTOS_RELIABILITY ,
1741.Cm mincost
1742.Pq Dv IPTOS_MINCOST ,
1743.Cm congestion
1744.Pq Dv IPTOS_ECN_CE .
1745The absence of a particular type may be denoted
1746with a
1747.Ql \&! .
1748.It Cm dscp spec Ns Op , Ns Ar spec
1749Matches IPv4/IPv6 packets whose
1750.Cm DS
1751field value is contained in
1752.Ar spec
1753mask.
1754Multiple values can be specified via
1755the comma separated list.
1756Value can be one of keywords used in
1757.Cm setdscp
1758action or exact number.
1759.It Cm ipttl Ar ttl-list
1760Matches IPv4 packets whose time to live is included in
1761.Ar ttl-list ,
1762which is either a single value or a list of values or ranges
1763specified in the same way as
1764.Ar ports .
1765.It Cm ipversion Ar ver
1766Matches IP packets whose IP version field is
1767.Ar ver .
1768.It Cm keep-state Op Ar :flowname
1769Upon a match, the firewall will create a dynamic rule, whose
1770default behaviour is to match bidirectional traffic between
1771source and destination IP/port using the same protocol.
1772The rule has a limited lifetime (controlled by a set of
1773.Xr sysctl 8
1774variables), and the lifetime is refreshed every time a matching
1775packet is found.
1776The
1777.Ar :flowname
1778is used to assign additional to addresses, ports and protocol parameter
1779to dynamic rule. It can be used for more accurate matching by
1780.Cm check-state
1781rule.
1782The
1783.Cm :default
1784keyword is special name used for compatibility with old rulesets.
1785.It Cm layer2
1786Matches only layer2 packets, i.e., those passed to
1787.Nm
1788from ether_demux() and ether_output_frame().
1789.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N Op Ar :flowname
1790The firewall will only allow
1791.Ar N
1792connections with the same
1793set of parameters as specified in the rule.
1794One or more
1795of source and destination addresses and ports can be
1796specified.
1797.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar name
1798Search an entry in lookup table
1799.Ar name
1800that matches the field specified as argument.
1801If not found, the match fails.
1802Otherwise, the match succeeds and
1803.Cm tablearg
1804is set to the value extracted from the table.
1805.Pp
1806This option can be useful to quickly dispatch traffic based on
1807certain packet fields.
1808See the
1809.Sx LOOKUP TABLES
1810section below for more information on lookup tables.
1811.It Cm { MAC | mac } Ar dst-mac src-mac
1812Match packets with a given
1813.Ar dst-mac
1814and
1815.Ar src-mac
1816addresses, specified as the
1817.Cm any
1818keyword (matching any MAC address), or six groups of hex digits
1819separated by colons,
1820and optionally followed by a mask indicating the significant bits.
1821The mask may be specified using either of the following methods:
1822.Bl -enum -width indent
1823.It
1824A slash
1825.Pq /
1826followed by the number of significant bits.
1827For example, an address with 33 significant bits could be specified as:
1828.Pp
1829.Dl "MAC 10:20:30:40:50:60/33 any"
1830.It
1831An ampersand
1832.Pq &
1833followed by a bitmask specified as six groups of hex digits separated
1834by colons.
1835For example, an address in which the last 16 bits are significant could
1836be specified as:
1837.Pp
1838.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1839.Pp
1840Note that the ampersand character has a special meaning in many shells
1841and should generally be escaped.
1842.El
1843Note that the order of MAC addresses (destination first,
1844source second) is
1845the same as on the wire, but the opposite of the one used for
1846IP addresses.
1847.It Cm mac-type Ar mac-type
1848Matches packets whose Ethernet Type field
1849corresponds to one of those specified as argument.
1850.Ar mac-type
1851is specified in the same way as
1852.Cm port numbers
1853(i.e., one or more comma-separated single values or ranges).
1854You can use symbolic names for known values such as
1855.Em vlan , ipv4, ipv6 .
1856Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1857and they are always printed as hexadecimal (unless the
1858.Cm -N
1859option is used, in which case symbolic resolution will be attempted).
1860.It Cm proto Ar protocol
1861Matches packets with the corresponding IP protocol.
1862.It Cm record-state
1863Upon a match, the firewall will create a dynamic rule as if
1864.Cm keep-state
1865was specified.
1866However, this option doesn't imply an implicit
1867.Cm check-state
1868in contrast to
1869.Cm keep-state .
1870.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
1871Matches packets received, transmitted or going through,
1872respectively, the interface specified by exact name
1873.Po Ar ifX Pc ,
1874by device name
1875.Po Ar if* Pc ,
1876by IP address, or through some interface.
1877Table
1878.Ar name
1879may be used to match interface by its kernel ifindex.
1880See the
1881.Sx LOOKUP TABLES
1882section below for more information on lookup tables.
1883.Pp
1884The
1885.Cm via
1886keyword causes the interface to always be checked.
1887If
1888.Cm recv
1889or
1890.Cm xmit
1891is used instead of
1892.Cm via ,
1893then only the receive or transmit interface (respectively)
1894is checked.
1895By specifying both, it is possible to match packets based on
1896both receive and transmit interface, e.g.:
1897.Pp
1898.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1899.Pp
1900The
1901.Cm recv
1902interface can be tested on either incoming or outgoing packets,
1903while the
1904.Cm xmit
1905interface can only be tested on outgoing packets.
1906So
1907.Cm out
1908is required (and
1909.Cm in
1910is invalid) whenever
1911.Cm xmit
1912is used.
1913.Pp
1914A packet might not have a receive or transmit interface: packets
1915originating from the local host have no receive interface,
1916while packets destined for the local host have no transmit
1917interface.
1918.It Cm set-limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
1919Works like
1920.Cm limit
1921but does not have an implicit
1922.Cm check-state
1923attached to it.
1924.It Cm setup
1925Matches TCP packets that have the SYN bit set but no ACK bit.
1926This is the short form of
1927.Dq Li tcpflags\ syn,!ack .
1928.It Cm sockarg
1929Matches packets that are associated to a local socket and
1930for which the SO_USER_COOKIE socket option has been set
1931to a non-zero value.
1932As a side effect, the value of the
1933option is made available as
1934.Cm tablearg
1935value, which in turn can be used as
1936.Cm skipto
1937or
1938.Cm pipe
1939number.
1940.It Cm src-ip Ar ip-address
1941Matches IPv4 packets whose source IP is one of the address(es)
1942specified as an argument.
1943.It Cm src-ip6 Ar ip6-address
1944Matches IPv6 packets whose source IP is one of the address(es)
1945specified as an argument.
1946.It Cm src-port Ar ports
1947Matches IP packets whose source port is one of the port(s)
1948specified as argument.
1949.It Cm tagged Ar tag-list
1950Matches packets whose tags are included in
1951.Ar tag-list ,
1952which is either a single value or a list of values or ranges
1953specified in the same way as
1954.Ar ports .
1955Tags can be applied to the packet using
1956.Cm tag
1957rule action parameter (see it's description for details on tags).
1958.It Cm tcpack Ar ack
1959TCP packets only.
1960Match if the TCP header acknowledgment number field is set to
1961.Ar ack .
1962.It Cm tcpdatalen Ar tcpdatalen-list
1963Matches TCP packets whose length of TCP data is
1964.Ar tcpdatalen-list ,
1965which is either a single value or a list of values or ranges
1966specified in the same way as
1967.Ar ports .
1968.It Cm tcpflags Ar spec
1969TCP packets only.
1970Match if the TCP header contains the comma separated list of
1971flags specified in
1972.Ar spec .
1973The supported TCP flags are:
1974.Pp
1975.Cm fin ,
1976.Cm syn ,
1977.Cm rst ,
1978.Cm psh ,
1979.Cm ack
1980and
1981.Cm urg .
1982The absence of a particular flag may be denoted
1983with a
1984.Ql \&! .
1985A rule which contains a
1986.Cm tcpflags
1987specification can never match a fragmented packet which has
1988a non-zero offset.
1989See the
1990.Cm frag
1991option for details on matching fragmented packets.
1992.It Cm tcpseq Ar seq
1993TCP packets only.
1994Match if the TCP header sequence number field is set to
1995.Ar seq .
1996.It Cm tcpwin Ar tcpwin-list
1997Matches TCP packets whose  header window field is set to
1998.Ar tcpwin-list ,
1999which is either a single value or a list of values or ranges
2000specified in the same way as
2001.Ar ports .
2002.It Cm tcpoptions Ar spec
2003TCP packets only.
2004Match if the TCP header contains the comma separated list of
2005options specified in
2006.Ar spec .
2007The supported TCP options are:
2008.Pp
2009.Cm mss
2010(maximum segment size),
2011.Cm window
2012(tcp window advertisement),
2013.Cm sack
2014(selective ack),
2015.Cm ts
2016(rfc1323 timestamp) and
2017.Cm cc
2018(rfc1644 t/tcp connection count).
2019The absence of a particular option may be denoted
2020with a
2021.Ql \&! .
2022.It Cm uid Ar user
2023Match all TCP or UDP packets sent by or received for a
2024.Ar user .
2025A
2026.Ar user
2027may be matched by name or identification number.
2028.It Cm verrevpath
2029For incoming packets,
2030a routing table lookup is done on the packet's source address.
2031If the interface on which the packet entered the system matches the
2032outgoing interface for the route,
2033the packet matches.
2034If the interfaces do not match up,
2035the packet does not match.
2036All outgoing packets or packets with no incoming interface match.
2037.Pp
2038The name and functionality of the option is intentionally similar to
2039the Cisco IOS command:
2040.Pp
2041.Dl ip verify unicast reverse-path
2042.Pp
2043This option can be used to make anti-spoofing rules to reject all
2044packets with source addresses not from this interface.
2045See also the option
2046.Cm antispoof .
2047.It Cm versrcreach
2048For incoming packets,
2049a routing table lookup is done on the packet's source address.
2050If a route to the source address exists, but not the default route
2051or a blackhole/reject route, the packet matches.
2052Otherwise, the packet does not match.
2053All outgoing packets match.
2054.Pp
2055The name and functionality of the option is intentionally similar to
2056the Cisco IOS command:
2057.Pp
2058.Dl ip verify unicast source reachable-via any
2059.Pp
2060This option can be used to make anti-spoofing rules to reject all
2061packets whose source address is unreachable.
2062.It Cm antispoof
2063For incoming packets, the packet's source address is checked if it
2064belongs to a directly connected network.
2065If the network is directly connected, then the interface the packet
2066came on in is compared to the interface the network is connected to.
2067When incoming interface and directly connected interface are not the
2068same, the packet does not match.
2069Otherwise, the packet does match.
2070All outgoing packets match.
2071.Pp
2072This option can be used to make anti-spoofing rules to reject all
2073packets that pretend to be from a directly connected network but do
2074not come in through that interface.
2075This option is similar to but more restricted than
2076.Cm verrevpath
2077because it engages only on packets with source addresses of directly
2078connected networks instead of all source addresses.
2079.El
2080.Sh LOOKUP TABLES
2081Lookup tables are useful to handle large sparse sets of
2082addresses or other search keys (e.g., ports, jail IDs, interface names).
2083In the rest of this section we will use the term ``key''.
2084Table name needs to match the following spec:
2085.Ar table-name .
2086Tables with the same name can be created in different
2087.Ar sets .
2088However, rule links to the tables in
2089.Ar set 0
2090by default.
2091This behavior can be controlled by
2092.Va net.inet.ip.fw.tables_sets
2093variable.
2094See the
2095.Sx SETS OF RULES
2096section for more information.
2097There may be up to 65535 different lookup tables.
2098.Pp
2099The following table types are supported:
2100.Bl -tag -width indent
2101.It Ar table-type : Ar addr | iface | number | flow
2102.It Ar table-key : Ar addr Ns Oo / Ns Ar masklen Oc | iface-name | number | flow-spec
2103.It Ar flow-spec : Ar flow-field Ns Op , Ns Ar flow-spec
2104.It Ar flow-field : src-ip | proto | src-port | dst-ip | dst-port
2105.It Cm addr
2106Matches IPv4 or IPv6 address.
2107Each entry is represented by an
2108.Ar addr Ns Op / Ns Ar masklen
2109and will match all addresses with base
2110.Ar addr
2111(specified as an IPv4/IPv6 address, or a hostname) and mask width of
2112.Ar masklen
2113bits.
2114If
2115.Ar masklen
2116is not specified, it defaults to 32 for IPv4 and 128 for IPv6.
2117When looking up an IP address in a table, the most specific
2118entry will match.
2119.It Cm iface
2120Matches interface names.
2121Each entry is represented by string treated as interface name.
2122Wildcards are not supported.
2123.It Cm number
2124Matches protocol ports, uids/gids or jail IDs.
2125Each entry is represented by 32-bit unsigned integer.
2126Ranges are not supported.
2127.It Cm flow
2128Matches packet fields specified by
2129.Ar flow
2130type suboptions with table entries.
2131.El
2132.Pp
2133Tables require explicit creation via
2134.Cm create
2135before use.
2136.Pp
2137The following creation options are supported:
2138.Bl -tag -width indent
2139.It Ar create-options : Ar create-option | create-options
2140.It Ar create-option : Cm type Ar table-type | Cm valtype Ar value-mask | Cm algo Ar algo-desc |
2141.Cm limit Ar number | Cm locked
2142.It Cm type
2143Table key type.
2144.It Cm valtype
2145Table value mask.
2146.It Cm algo
2147Table algorithm to use (see below).
2148.It Cm limit
2149Maximum number of items that may be inserted into table.
2150.It Cm locked
2151Restrict any table modifications.
2152.El
2153.Pp
2154Some of these options may be modified later via
2155.Cm modify
2156keyword.
2157The following options can be changed:
2158.Bl -tag -width indent
2159.It Ar modify-options : Ar modify-option | modify-options
2160.It Ar modify-option : Cm limit Ar number
2161.It Cm limit
2162Alter maximum number of items that may be inserted into table.
2163.El
2164.Pp
2165Additionally, table can be locked or unlocked using
2166.Cm lock
2167or
2168.Cm unlock
2169commands.
2170.Pp
2171Tables of the same
2172.Ar type
2173can be swapped with each other using
2174.Cm swap Ar name
2175command.
2176Swap may fail if tables limits are set and data exchange
2177would result in limits hit.
2178Operation is performed atomically.
2179.Pp
2180One or more entries can be added to a table at once using
2181.Cm add
2182command.
2183Addition of all items are performed atomically.
2184By default, error in addition of one entry does not influence
2185addition of other entries. However, non-zero error code is returned
2186in that case.
2187Special
2188.Cm atomic
2189keyword may be specified before
2190.Cm add
2191to indicate all-or-none add request.
2192.Pp
2193One or more entries can be removed from a table at once using
2194.Cm delete
2195command.
2196By default, error in removal of one entry does not influence
2197removing of other entries. However, non-zero error code is returned
2198in that case.
2199.Pp
2200It may be possible to check what entry will be found on particular
2201.Ar table-key
2202using
2203.Cm lookup
2204.Ar table-key
2205command.
2206This functionality is optional and may be unsupported in some algorithms.
2207.Pp
2208The following operations can be performed on
2209.Ar one
2210or
2211.Cm all
2212tables:
2213.Bl -tag -width indent
2214.It Cm list
2215List all entries.
2216.It Cm flush
2217Removes all entries.
2218.It Cm info
2219Shows generic table information.
2220.It Cm detail
2221Shows generic table information and algo-specific data.
2222.El
2223.Pp
2224The following lookup algorithms are supported:
2225.Bl -tag -width indent
2226.It Ar algo-desc : algo-name | "algo-name algo-data"
2227.It Ar algo-name: Ar addr:radix | addr:hash | iface:array | number:array | flow:hash
2228.It Cm addr:radix
2229Separate Radix trees for IPv4 and IPv6, the same way as the routing table (see
2230.Xr route 4 ) .
2231Default choice for
2232.Ar addr
2233type.
2234.It Cm addr:hash
2235Separate auto-growing hashes for IPv4 and IPv6.
2236Accepts entries with the same mask length specified initially via
2237.Cm "addr:hash masks=/v4,/v6"
2238algorithm creation options.
2239Assume /32 and /128 masks by default.
2240Search removes host bits (according to mask) from supplied address and checks
2241resulting key in appropriate hash.
2242Mostly optimized for /64 and byte-ranged IPv6 masks.
2243.It Cm iface:array
2244Array storing sorted indexes for entries which are presented in the system.
2245Optimized for very fast lookup.
2246.It Cm number:array
2247Array storing sorted u32 numbers.
2248.It Cm flow:hash
2249Auto-growing hash storing flow entries.
2250Search calculates hash on required packet fields and searches for matching
2251entries in selected bucket.
2252.El
2253.Pp
2254The
2255.Cm tablearg
2256feature provides the ability to use a value, looked up in the table, as
2257the argument for a rule action, action parameter or rule option.
2258This can significantly reduce number of rules in some configurations.
2259If two tables are used in a rule, the result of the second (destination)
2260is used.
2261.Pp
2262Each record may hold one or more values according to
2263.Ar value-mask .
2264This mask is set on table creation via
2265.Cm valtype
2266option.
2267The following value types are supported:
2268.Bl -tag -width indent
2269.It Ar value-mask : Ar value-type Ns Op , Ns Ar value-mask
2270.It Ar value-type : Ar skipto | pipe | fib | nat | dscp | tag | divert |
2271.Ar netgraph | limit | ipv4
2272.It Cm skipto
2273rule number to jump to.
2274.It Cm pipe
2275Pipe number to use.
2276.It Cm fib
2277fib number to match/set.
2278.It Cm nat
2279nat number to jump to.
2280.It Cm dscp
2281dscp value to match/set.
2282.It Cm tag
2283tag number to match/set.
2284.It Cm divert
2285port number to divert traffic to.
2286.It Cm netgraph
2287hook number to move packet to.
2288.It Cm limit
2289maximum number of connections.
2290.It Cm ipv4
2291IPv4 nexthop to fwd packets to.
2292.It Cm ipv6
2293IPv6 nexthop to fwd packets to.
2294.El
2295.Pp
2296The
2297.Cm tablearg
2298argument can be used with the following actions:
2299.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto, setfib,
2300action parameters:
2301.Cm tag, untag,
2302rule options:
2303.Cm limit, tagged.
2304.Pp
2305When used with the
2306.Cm skipto
2307action, the user should be aware that the code will walk the ruleset
2308up to a rule equal to, or past, the given number.
2309.Pp
2310See the
2311.Sx EXAMPLES
2312Section for example usage of tables and the tablearg keyword.
2313.Sh SETS OF RULES
2314Each rule or table belongs to one of 32 different
2315.Em sets
2316, numbered 0 to 31.
2317Set 31 is reserved for the default rule.
2318.Pp
2319By default, rules or tables are put in set 0, unless you use the
2320.Cm set N
2321attribute when adding a new rule or table.
2322Sets can be individually and atomically enabled or disabled,
2323so this mechanism permits an easy way to store multiple configurations
2324of the firewall and quickly (and atomically) switch between them.
2325.Pp
2326By default, tables from set 0 are referenced when adding rule with
2327table opcodes regardless of rule set.
2328This behavior can be changed by setting
2329.Va net.inet.ip.fw.tables_sets
2330variable to 1.
2331Rule's set will then be used for table references.
2332.Pp
2333The command to enable/disable sets is
2334.Bd -ragged -offset indent
2335.Nm
2336.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
2337.Ed
2338.Pp
2339where multiple
2340.Cm enable
2341or
2342.Cm disable
2343sections can be specified.
2344Command execution is atomic on all the sets specified in the command.
2345By default, all sets are enabled.
2346.Pp
2347When you disable a set, its rules behave as if they do not exist
2348in the firewall configuration, with only one exception:
2349.Bd -ragged -offset indent
2350dynamic rules created from a rule before it had been disabled
2351will still be active until they expire.
2352In order to delete
2353dynamic rules you have to explicitly delete the parent rule
2354which generated them.
2355.Ed
2356.Pp
2357The set number of rules can be changed with the command
2358.Bd -ragged -offset indent
2359.Nm
2360.Cm set move
2361.Brq Cm rule Ar rule-number | old-set
2362.Cm to Ar new-set
2363.Ed
2364.Pp
2365Also, you can atomically swap two rulesets with the command
2366.Bd -ragged -offset indent
2367.Nm
2368.Cm set swap Ar first-set second-set
2369.Ed
2370.Pp
2371See the
2372.Sx EXAMPLES
2373Section on some possible uses of sets of rules.
2374.Sh STATEFUL FIREWALL
2375Stateful operation is a way for the firewall to dynamically
2376create rules for specific flows when packets that
2377match a given pattern are detected.
2378Support for stateful
2379operation comes through the
2380.Cm check-state , keep-state , record-state , limit
2381and
2382.Cm set-limit
2383options of
2384.Nm rules .
2385.Pp
2386Dynamic rules are created when a packet matches a
2387.Cm keep-state ,
2388.Cm record-state ,
2389.Cm limit
2390or
2391.Cm set-limit
2392rule, causing the creation of a
2393.Em dynamic
2394rule which will match all and only packets with
2395a given
2396.Em protocol
2397between a
2398.Em src-ip/src-port dst-ip/dst-port
2399pair of addresses
2400.Em ( src
2401and
2402.Em dst
2403are used here only to denote the initial match addresses, but they
2404are completely equivalent afterwards).
2405Rules created by
2406.Cm keep-state
2407option also have a
2408.Ar :flowname
2409taken from it.
2410This name is used in matching together with addresses, ports and protocol.
2411Dynamic rules will be checked at the first
2412.Cm check-state, keep-state
2413or
2414.Cm limit
2415occurrence, and the action performed upon a match will be the same
2416as in the parent rule.
2417.Pp
2418Note that no additional attributes other than protocol and IP addresses
2419and ports and :flowname are checked on dynamic rules.
2420.Pp
2421The typical use of dynamic rules is to keep a closed firewall configuration,
2422but let the first TCP SYN packet from the inside network install a
2423dynamic rule for the flow so that packets belonging to that session
2424will be allowed through the firewall:
2425.Pp
2426.Dl "ipfw add check-state :OUTBOUND"
2427.Dl "ipfw add allow tcp from my-subnet to any setup keep-state :OUTBOUND"
2428.Dl "ipfw add deny tcp from any to any"
2429.Pp
2430A similar approach can be used for UDP, where an UDP packet coming
2431from the inside will install a dynamic rule to let the response through
2432the firewall:
2433.Pp
2434.Dl "ipfw add check-state :OUTBOUND"
2435.Dl "ipfw add allow udp from my-subnet to any keep-state :OUTBOUND"
2436.Dl "ipfw add deny udp from any to any"
2437.Pp
2438Dynamic rules expire after some time, which depends on the status
2439of the flow and the setting of some
2440.Cm sysctl
2441variables.
2442See Section
2443.Sx SYSCTL VARIABLES
2444for more details.
2445For TCP sessions, dynamic rules can be instructed to periodically
2446send keepalive packets to refresh the state of the rule when it is
2447about to expire.
2448.Pp
2449See Section
2450.Sx EXAMPLES
2451for more examples on how to use dynamic rules.
2452.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
2453.Nm
2454is also the user interface for the
2455.Nm dummynet
2456traffic shaper, packet scheduler and network emulator, a subsystem that
2457can artificially queue, delay or drop packets
2458emulating the behaviour of certain network links
2459or queueing systems.
2460.Pp
2461.Nm dummynet
2462operates by first using the firewall to select packets
2463using any match pattern that can be used in
2464.Nm
2465rules.
2466Matching packets are then passed to either of two
2467different objects, which implement the traffic regulation:
2468.Bl -hang -offset XXXX
2469.It Em pipe
2470A
2471.Em pipe
2472emulates a
2473.Em link
2474with given bandwidth and propagation delay,
2475driven by a FIFO scheduler and a single queue with programmable
2476queue size and packet loss rate.
2477Packets are appended to the queue as they come out from
2478.Nm ipfw ,
2479and then transferred in FIFO order to the link at the desired rate.
2480.It Em queue
2481A
2482.Em queue
2483is an abstraction used to implement packet scheduling
2484using one of several packet scheduling algorithms.
2485Packets sent to a
2486.Em queue
2487are first grouped into flows according to a mask on the 5-tuple.
2488Flows are then passed to the scheduler associated to the
2489.Em queue ,
2490and each flow uses scheduling parameters (weight and others)
2491as configured in the
2492.Em queue
2493itself.
2494A scheduler in turn is connected to an emulated link,
2495and arbitrates the link's bandwidth among backlogged flows according to
2496weights and to the features of the scheduling algorithm in use.
2497.El
2498.Pp
2499In practice,
2500.Em pipes
2501can be used to set hard limits to the bandwidth that a flow can use, whereas
2502.Em queues
2503can be used to determine how different flows share the available bandwidth.
2504.Pp
2505A graphical representation of the binding of queues,
2506flows, schedulers and links is below.
2507.Bd -literal -offset indent
2508                 (flow_mask|sched_mask)  sched_mask
2509         +---------+   weight Wx  +-------------+
2510         |         |->-[flow]-->--|             |-+
2511    -->--| QUEUE x |   ...        |             | |
2512         |         |->-[flow]-->--| SCHEDuler N | |
2513         +---------+              |             | |
2514             ...                  |             +--[LINK N]-->--
2515         +---------+   weight Wy  |             | +--[LINK N]-->--
2516         |         |->-[flow]-->--|             | |
2517    -->--| QUEUE y |   ...        |             | |
2518         |         |->-[flow]-->--|             | |
2519         +---------+              +-------------+ |
2520                                    +-------------+
2521.Ed
2522It is important to understand the role of the SCHED_MASK
2523and FLOW_MASK, which are configured through the commands
2524.Dl "ipfw sched N config mask SCHED_MASK ..."
2525and
2526.Dl "ipfw queue X config mask FLOW_MASK ..." .
2527.Pp
2528The SCHED_MASK is used to assign flows to one or more
2529scheduler instances, one for each
2530value of the packet's 5-tuple after applying SCHED_MASK.
2531As an example, using ``src-ip 0xffffff00'' creates one instance
2532for each /24 destination subnet.
2533.Pp
2534The FLOW_MASK, together with the SCHED_MASK, is used to split
2535packets into flows.
2536As an example, using
2537``src-ip 0x000000ff''
2538together with the previous SCHED_MASK makes a flow for
2539each individual source address.
2540In turn, flows for each /24
2541subnet will be sent to the same scheduler instance.
2542.Pp
2543The above diagram holds even for the
2544.Em pipe
2545case, with the only restriction that a
2546.Em pipe
2547only supports a SCHED_MASK, and forces the use of a FIFO
2548scheduler (these are for backward compatibility reasons;
2549in fact, internally, a
2550.Nm dummynet's
2551pipe is implemented exactly as above).
2552.Pp
2553There are two modes of
2554.Nm dummynet
2555operation:
2556.Dq normal
2557and
2558.Dq fast .
2559The
2560.Dq normal
2561mode tries to emulate a real link: the
2562.Nm dummynet
2563scheduler ensures that the packet will not leave the pipe faster than it
2564would on the real link with a given bandwidth.
2565The
2566.Dq fast
2567mode allows certain packets to bypass the
2568.Nm dummynet
2569scheduler (if packet flow does not exceed pipe's bandwidth).
2570This is the reason why the
2571.Dq fast
2572mode requires less CPU cycles per packet (on average) and packet latency
2573can be significantly lower in comparison to a real link with the same
2574bandwidth.
2575The default mode is
2576.Dq normal .
2577The
2578.Dq fast
2579mode can be enabled by setting the
2580.Va net.inet.ip.dummynet.io_fast
2581.Xr sysctl 8
2582variable to a non-zero value.
2583.Pp
2584.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
2585The
2586.Em pipe ,
2587.Em queue
2588and
2589.Em scheduler
2590configuration commands are the following:
2591.Bd -ragged -offset indent
2592.Cm pipe Ar number Cm config Ar pipe-configuration
2593.Pp
2594.Cm queue Ar number Cm config Ar queue-configuration
2595.Pp
2596.Cm sched Ar number Cm config Ar sched-configuration
2597.Ed
2598.Pp
2599The following parameters can be configured for a pipe:
2600.Pp
2601.Bl -tag -width indent -compact
2602.It Cm bw Ar bandwidth | device
2603Bandwidth, measured in
2604.Sm off
2605.Op Cm K | M | G
2606.Brq Cm bit/s | Byte/s .
2607.Sm on
2608.Pp
2609A value of 0 (default) means unlimited bandwidth.
2610The unit must immediately follow the number, as in
2611.Pp
2612.Dl "ipfw pipe 1 config bw 300Kbit/s"
2613.Pp
2614If a device name is specified instead of a numeric value, as in
2615.Pp
2616.Dl "ipfw pipe 1 config bw tun0"
2617.Pp
2618then the transmit clock is supplied by the specified device.
2619At the moment only the
2620.Xr tun 4
2621device supports this
2622functionality, for use in conjunction with
2623.Xr ppp 8 .
2624.Pp
2625.It Cm delay Ar ms-delay
2626Propagation delay, measured in milliseconds.
2627The value is rounded to the next multiple of the clock tick
2628(typically 10ms, but it is a good practice to run kernels
2629with
2630.Dq "options HZ=1000"
2631to reduce
2632the granularity to 1ms or less).
2633The default value is 0, meaning no delay.
2634.Pp
2635.It Cm burst Ar size
2636If the data to be sent exceeds the pipe's bandwidth limit
2637(and the pipe was previously idle), up to
2638.Ar size
2639bytes of data are allowed to bypass the
2640.Nm dummynet
2641scheduler, and will be sent as fast as the physical link allows.
2642Any additional data will be transmitted at the rate specified
2643by the
2644.Nm pipe
2645bandwidth.
2646The burst size depends on how long the pipe has been idle;
2647the effective burst size is calculated as follows:
2648MAX(
2649.Ar size
2650,
2651.Nm bw
2652* pipe_idle_time).
2653.Pp
2654.It Cm profile Ar filename
2655A file specifying the additional overhead incurred in the transmission
2656of a packet on the link.
2657.Pp
2658Some link types introduce extra delays in the transmission
2659of a packet, e.g., because of MAC level framing, contention on
2660the use of the channel, MAC level retransmissions and so on.
2661From our point of view, the channel is effectively unavailable
2662for this extra time, which is constant or variable depending
2663on the link type.
2664Additionally, packets may be dropped after this
2665time (e.g., on a wireless link after too many retransmissions).
2666We can model the additional delay with an empirical curve
2667that represents its distribution.
2668.Bd -literal -offset indent
2669      cumulative probability
2670      1.0 ^
2671          |
2672      L   +-- loss-level          x
2673          |                 ******
2674          |                *
2675          |           *****
2676          |          *
2677          |        **
2678          |       *
2679          +-------*------------------->
2680                      delay
2681.Ed
2682The empirical curve may have both vertical and horizontal lines.
2683Vertical lines represent constant delay for a range of
2684probabilities.
2685Horizontal lines correspond to a discontinuity in the delay
2686distribution: the pipe will use the largest delay for a
2687given probability.
2688.Pp
2689The file format is the following, with whitespace acting as
2690a separator and '#' indicating the beginning a comment:
2691.Bl -tag -width indent
2692.It Cm name Ar identifier
2693optional name (listed by "ipfw pipe show")
2694to identify the delay distribution;
2695.It Cm bw Ar value
2696the bandwidth used for the pipe.
2697If not specified here, it must be present
2698explicitly as a configuration parameter for the pipe;
2699.It Cm loss-level Ar L
2700the probability above which packets are lost.
2701(0.0 <= L <= 1.0, default 1.0 i.e., no loss);
2702.It Cm samples Ar N
2703the number of samples used in the internal
2704representation of the curve (2..1024; default 100);
2705.It Cm "delay prob" | "prob delay"
2706One of these two lines is mandatory and defines
2707the format of the following lines with data points.
2708.It Ar XXX Ar YYY
27092 or more lines representing points in the curve,
2710with either delay or probability first, according
2711to the chosen format.
2712The unit for delay is milliseconds.
2713Data points do not need to be sorted.
2714Also, the number of actual lines can be different
2715from the value of the "samples" parameter:
2716.Nm
2717utility will sort and interpolate
2718the curve as needed.
2719.El
2720.Pp
2721Example of a profile file:
2722.Bd -literal -offset indent
2723name    bla_bla_bla
2724samples 100
2725loss-level    0.86
2726prob    delay
27270       200	# minimum overhead is 200ms
27280.5     200
27290.5     300
27300.8     1000
27310.9     1300
27321       1300
2733#configuration file end
2734.Ed
2735.El
2736.Pp
2737The following parameters can be configured for a queue:
2738.Pp
2739.Bl -tag -width indent -compact
2740.It Cm pipe Ar pipe_nr
2741Connects a queue to the specified pipe.
2742Multiple queues (with the same or different weights) can be connected to
2743the same pipe, which specifies the aggregate rate for the set of queues.
2744.Pp
2745.It Cm weight Ar weight
2746Specifies the weight to be used for flows matching this queue.
2747The weight must be in the range 1..100, and defaults to 1.
2748.El
2749.Pp
2750The following case-insensitive parameters can be configured for a
2751scheduler:
2752.Pp
2753.Bl -tag -width indent -compact
2754.It Cm type Ar {fifo | wf2q+ | rr | qfq | fq_codel | fq_pie}
2755specifies the scheduling algorithm to use.
2756.Bl -tag -width indent -compact
2757.It Cm fifo
2758is just a FIFO scheduler (which means that all packets
2759are stored in the same queue as they arrive to the scheduler).
2760FIFO has O(1) per-packet time complexity, with very low
2761constants (estimate 60-80ns on a 2GHz desktop machine)
2762but gives no service guarantees.
2763.It Cm wf2q+
2764implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2765algorithm which permits flows to share bandwidth according to
2766their weights.
2767Note that weights are not priorities; even a flow
2768with a minuscule weight will never starve.
2769WF2Q+ has O(log N) per-packet processing cost, where N is the number
2770of flows, and is the default algorithm used by previous versions
2771dummynet's queues.
2772.It Cm rr
2773implements the Deficit Round Robin algorithm, which has O(1) processing
2774costs (roughly, 100-150ns per packet)
2775and permits bandwidth allocation according to weights, but
2776with poor service guarantees.
2777.It Cm qfq
2778implements the QFQ algorithm, which is a very fast variant of
2779WF2Q+, with similar service guarantees and O(1) processing
2780costs (roughly, 200-250ns per packet).
2781.It Cm fq_codel
2782implements the FQ-CoDel (FlowQueue-CoDel) scheduler/AQM algorithm, which
2783uses a modified Deficit Round Robin scheduler to manage two lists of sub-queues
2784(old sub-queues and new sub-queues) for providing brief periods of priority to
2785lightweight or short burst flows.
2786By default, the total number of sub-queues is 1024.
2787FQ-CoDel's internal, dynamically
2788created sub-queues are controlled by separate instances of CoDel AQM.
2789.It Cm fq_pie
2790implements the FQ-PIE (FlowQueue-PIE) scheduler/AQM algorithm, which similar to
2791.Cm fq_codel
2792but uses per sub-queue PIE AQM instance to control the queue delay.
2793.El
2794.Pp
2795.Cm fq_codel
2796inherits AQM parameters and options from
2797.Cm codel
2798(see below), and
2799.Cm fq_pie
2800inherits AQM parameters and options from
2801.Cm pie
2802(see below).
2803Additionally, both of
2804.Cm fq_codel
2805and
2806.Cm fq_pie
2807have shared scheduler parameters which are:
2808.Bl -tag -width indent
2809.It Cm quantum
2810.Ar m
2811specifies the quantum (credit) of the scheduler.
2812.Ar m
2813is the number of bytes a queue can serve before being moved to the tail
2814of old queues list.
2815The default is 1514 bytes, and the maximum acceptable value
2816is 9000 bytes.
2817.It Cm limit
2818.Ar m
2819specifies the hard size limit (in unit of packets) of all queues managed by an
2820instance of the scheduler.
2821The default value of
2822.Ar m
2823is 10240 packets, and the maximum acceptable value is 20480 packets.
2824.It Cm flows
2825.Ar m
2826specifies the total number of flow queues (sub-queues) that fq_*
2827creates and manages.
2828By default, 1024 sub-queues are created when an instance
2829of the fq_{codel/pie} scheduler is created.
2830The maximum acceptable value is
283165536.
2832.El
2833.Pp
2834Note that any token after
2835.Cm fq_codel
2836or
2837.Cm fq_pie
2838is considered a parameter for fq_{codel/pie}.
2839So, ensure all scheduler
2840configuration options not related to fq_{codel/pie} are written before
2841.Cm fq_codel/fq_pie
2842tokens.
2843.El
2844.Pp
2845In addition to the type, all parameters allowed for a pipe can also
2846be specified for a scheduler.
2847.Pp
2848Finally, the following parameters can be configured for both
2849pipes and queues:
2850.Pp
2851.Bl -tag -width XXXX -compact
2852.It Cm buckets Ar hash-table-size
2853Specifies the size of the hash table used for storing the
2854various queues.
2855Default value is 64 controlled by the
2856.Xr sysctl 8
2857variable
2858.Va net.inet.ip.dummynet.hash_size ,
2859allowed range is 16 to 65536.
2860.Pp
2861.It Cm mask Ar mask-specifier
2862Packets sent to a given pipe or queue by an
2863.Nm
2864rule can be further classified into multiple flows, each of which is then
2865sent to a different
2866.Em dynamic
2867pipe or queue.
2868A flow identifier is constructed by masking the IP addresses,
2869ports and protocol types as specified with the
2870.Cm mask
2871options in the configuration of the pipe or queue.
2872For each different flow identifier, a new pipe or queue is created
2873with the same parameters as the original object, and matching packets
2874are sent to it.
2875.Pp
2876Thus, when
2877.Em dynamic pipes
2878are used, each flow will get the same bandwidth as defined by the pipe,
2879whereas when
2880.Em dynamic queues
2881are used, each flow will share the parent's pipe bandwidth evenly
2882with other flows generated by the same queue (note that other queues
2883with different weights might be connected to the same pipe).
2884.br
2885Available mask specifiers are a combination of one or more of the following:
2886.Pp
2887.Cm dst-ip Ar mask ,
2888.Cm dst-ip6 Ar mask ,
2889.Cm src-ip Ar mask ,
2890.Cm src-ip6 Ar mask ,
2891.Cm dst-port Ar mask ,
2892.Cm src-port Ar mask ,
2893.Cm flow-id Ar mask ,
2894.Cm proto Ar mask
2895or
2896.Cm all ,
2897.Pp
2898where the latter means all bits in all fields are significant.
2899.Pp
2900.It Cm noerror
2901When a packet is dropped by a
2902.Nm dummynet
2903queue or pipe, the error
2904is normally reported to the caller routine in the kernel, in the
2905same way as it happens when a device queue fills up.
2906Setting this
2907option reports the packet as successfully delivered, which can be
2908needed for some experimental setups where you want to simulate
2909loss or congestion at a remote router.
2910.Pp
2911.It Cm plr Ar packet-loss-rate
2912Packet loss rate.
2913Argument
2914.Ar packet-loss-rate
2915is a floating-point number between 0 and 1, with 0 meaning no
2916loss, 1 meaning 100% loss.
2917The loss rate is internally represented on 31 bits.
2918.Pp
2919.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2920Queue size, in
2921.Ar slots
2922or
2923.Cm KBytes .
2924Default value is 50 slots, which
2925is the typical queue size for Ethernet devices.
2926Note that for slow speed links you should keep the queue
2927size short or your traffic might be affected by a significant
2928queueing delay.
2929E.g., 50 max-sized Ethernet packets (1500 bytes) mean 600Kbit
2930or 20s of queue on a 30Kbit/s pipe.
2931Even worse effects can result if you get packets from an
2932interface with a much larger MTU, e.g.\& the loopback interface
2933with its 16KB packets.
2934The
2935.Xr sysctl 8
2936variables
2937.Em net.inet.ip.dummynet.pipe_byte_limit
2938and
2939.Em net.inet.ip.dummynet.pipe_slot_limit
2940control the maximum lengths that can be specified.
2941.Pp
2942.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2943[ecn]
2944Make use of the RED (Random Early Detection) queue management algorithm.
2945.Ar w_q
2946and
2947.Ar max_p
2948are floating
2949point numbers between 0 and 1 (inclusive), while
2950.Ar min_th
2951and
2952.Ar max_th
2953are integer numbers specifying thresholds for queue management
2954(thresholds are computed in bytes if the queue has been defined
2955in bytes, in slots otherwise).
2956The two parameters can also be of the same value if needed. The
2957.Nm dummynet
2958also supports the gentle RED variant (gred) and ECN (Explicit Congestion
2959Notification) as optional. Three
2960.Xr sysctl 8
2961variables can be used to control the RED behaviour:
2962.Bl -tag -width indent
2963.It Va net.inet.ip.dummynet.red_lookup_depth
2964specifies the accuracy in computing the average queue
2965when the link is idle (defaults to 256, must be greater than zero)
2966.It Va net.inet.ip.dummynet.red_avg_pkt_size
2967specifies the expected average packet size (defaults to 512, must be
2968greater than zero)
2969.It Va net.inet.ip.dummynet.red_max_pkt_size
2970specifies the expected maximum packet size, only used when queue
2971thresholds are in bytes (defaults to 1500, must be greater than zero).
2972.El
2973.Pp
2974.It Cm codel Oo Cm target Ar time Oc Oo Cm interval Ar time Oc Oo Cm ecn |
2975.Cm noecn Oc
2976Make use of the CoDel (Controlled-Delay) queue management algorithm.
2977.Ar time
2978is interpreted as milliseconds by default but seconds (s), milliseconds (ms) or
2979microseconds (us) can be specified instead.
2980CoDel drops or marks (ECN) packets
2981depending on packet sojourn time in the queue.
2982.Cm target
2983.Ar time
2984(5ms by default) is the minimum acceptable persistent queue delay that CoDel
2985allows.
2986CoDel does not drop packets directly after packets sojourn time becomes
2987higher than
2988.Cm target
2989.Ar time
2990but waits for
2991.Cm interval
2992.Ar time
2993(100ms default) before dropping.
2994.Cm interval
2995.Ar time
2996should be set to maximum RTT for all expected connections.
2997.Cm ecn
2998enables (disabled by default) packet marking (instead of dropping) for
2999ECN-enabled TCP flows when queue delay becomes high.
3000.Pp
3001Note that any token after
3002.Cm codel
3003is considered a parameter for CoDel.
3004So, ensure all pipe/queue
3005configuration options are written before
3006.Cm codel
3007token.
3008.Pp
3009The
3010.Xr sysctl 8
3011variables
3012.Va net.inet.ip.dummynet.codel.target
3013and
3014.Va net.inet.ip.dummynet.codel.interval
3015can be used to set CoDel default parameters.
3016.Pp
3017.It Cm pie Oo Cm target Ar time Oc Oo Cm tupdate Ar time Oc Oo
3018.Cm alpha Ar n Oc Oo Cm beta Ar n Oc Oo Cm max_burst Ar time Oc Oo
3019.Cm max_ecnth Ar n Oc Oo Cm ecn | Cm noecn Oc Oo Cm capdrop |
3020.Cm nocapdrop Oc Oo Cm drand | Cm nodrand Oc Oo Cm onoff
3021.Oc Oo Cm dre | Cm ts Oc
3022Make use of the PIE (Proportional Integral controller Enhanced) queue management
3023algorithm.
3024PIE drops or marks packets depending on a calculated drop probability during
3025en-queue process, with the aim of achieving high throughput while keeping queue
3026delay low.
3027At regular time intervals of
3028.Cm tupdate
3029.Ar time
3030(15ms by default) a background process (re)calculates the probability based on queue delay
3031deviations from
3032.Cm target
3033.Ar time
3034(15ms by default) and queue delay trends.
3035PIE approximates current queue
3036delay by using a departure rate estimation method, or (optionally) by using a
3037packet timestamp method similar to CoDel.
3038.Ar time
3039is interpreted as milliseconds by default but seconds (s), milliseconds (ms) or
3040microseconds (us) can be specified instead.
3041The other PIE parameters and options are as follows:
3042.Bl -tag -width indent
3043.It Cm alpha Ar n
3044.Ar n
3045is a floating point number between 0 and 7 which specifies the weight of queue
3046delay deviations that is used in drop probability calculation.
30470.125 is the default.
3048.It Cm beta Ar n
3049.Ar n
3050is a floating point number between 0 and 7 which specifies is the weight of queue
3051delay trend that is used in drop probability calculation.
30521.25 is the default.
3053.It Cm max_burst Ar time
3054The maximum period of time that PIE does not drop/mark packets.
3055150ms is the
3056default and 10s is the maximum value.
3057.It Cm max_ecnth Ar n
3058Even when ECN is enabled, PIE drops packets instead of marking them when drop
3059probability becomes higher than ECN probability threshold
3060.Cm max_ecnth Ar n
3061, the default is 0.1 (i.e 10%) and 1 is the maximum value.
3062.It Cm ecn | noecn
3063enable or disable ECN marking for ECN-enabled TCP flows.
3064Disabled by default.
3065.It Cm capdrop | nocapdrop
3066enable or disable cap drop adjustment.
3067Cap drop adjustment is enabled by default.
3068.It Cm drand | nodrand
3069enable or disable drop probability de-randomisation.
3070De-randomisation eliminates
3071the problem of dropping packets too close or too far.
3072De-randomisation is enabled by default.
3073.It Cm onoff
3074enable turning PIE on and off depending on queue load.
3075If this option is enabled,
3076PIE turns on when over 1/3 of queue becomes full.
3077This option is disabled by
3078default.
3079.It Cm dre | ts
3080Calculate queue delay using departure rate estimation
3081.Cm dre
3082or timestamps
3083.Cm ts .
3084.Cm dre
3085is used by default.
3086.El
3087.Pp
3088Note that any token after
3089.Cm pie
3090is considered a parameter for PIE.
3091So ensure all pipe/queue
3092the configuration options are written before
3093.Cm pie
3094token.
3095.Xr sysctl 8
3096variables can be used to control the
3097.Cm pie
3098default parameters.
3099See the
3100.Sx SYSCTL VARIABLES
3101section for more details.
3102.El
3103.Pp
3104When used with IPv6 data,
3105.Nm dummynet
3106currently has several limitations.
3107Information necessary to route link-local packets to an
3108interface is not available after processing by
3109.Nm dummynet
3110so those packets are dropped in the output path.
3111Care should be taken to ensure that link-local packets are not passed to
3112.Nm dummynet .
3113.Sh CHECKLIST
3114Here are some important points to consider when designing your
3115rules:
3116.Bl -bullet
3117.It
3118Remember that you filter both packets going
3119.Cm in
3120and
3121.Cm out .
3122Most connections need packets going in both directions.
3123.It
3124Remember to test very carefully.
3125It is a good idea to be near the console when doing this.
3126If you cannot be near the console,
3127use an auto-recovery script such as the one in
3128.Pa /usr/share/examples/ipfw/change_rules.sh .
3129.It
3130Do not forget the loopback interface.
3131.El
3132.Sh FINE POINTS
3133.Bl -bullet
3134.It
3135There are circumstances where fragmented datagrams are unconditionally
3136dropped.
3137TCP packets are dropped if they do not contain at least 20 bytes of
3138TCP header, UDP packets are dropped if they do not contain a full 8
3139byte UDP header, and ICMP packets are dropped if they do not contain
31404 bytes of ICMP header, enough to specify the ICMP type, code, and
3141checksum.
3142These packets are simply logged as
3143.Dq pullup failed
3144since there may not be enough good data in the packet to produce a
3145meaningful log entry.
3146.It
3147Another type of packet is unconditionally dropped, a TCP packet with a
3148fragment offset of one.
3149This is a valid packet, but it only has one use, to try
3150to circumvent firewalls.
3151When logging is enabled, these packets are
3152reported as being dropped by rule -1.
3153.It
3154If you are logged in over a network, loading the
3155.Xr kld 4
3156version of
3157.Nm
3158is probably not as straightforward as you would think.
3159The following command line is recommended:
3160.Bd -literal -offset indent
3161kldload ipfw && \e
3162ipfw add 32000 allow ip from any to any
3163.Ed
3164.Pp
3165Along the same lines, doing an
3166.Bd -literal -offset indent
3167ipfw flush
3168.Ed
3169.Pp
3170in similar surroundings is also a bad idea.
3171.It
3172The
3173.Nm
3174filter list may not be modified if the system security level
3175is set to 3 or higher
3176(see
3177.Xr init 8
3178for information on system security levels).
3179.El
3180.Sh PACKET DIVERSION
3181A
3182.Xr divert 4
3183socket bound to the specified port will receive all packets
3184diverted to that port.
3185If no socket is bound to the destination port, or if the divert module is
3186not loaded, or if the kernel was not compiled with divert socket support,
3187the packets are dropped.
3188.Sh NETWORK ADDRESS TRANSLATION (NAT)
3189.Nm
3190support in-kernel NAT using the kernel version of
3191.Xr libalias 3 .
3192The kernel module
3193.Cm ipfw_nat
3194should be loaded or kernel should have
3195.Cm options IPFIREWALL_NAT
3196to be able use NAT.
3197.Pp
3198The nat configuration command is the following:
3199.Bd -ragged -offset indent
3200.Bk -words
3201.Cm nat
3202.Ar nat_number
3203.Cm config
3204.Ar nat-configuration
3205.Ek
3206.Ed
3207.Pp
3208The following parameters can be configured:
3209.Bl -tag -width indent
3210.It Cm ip Ar ip_address
3211Define an ip address to use for aliasing.
3212.It Cm if Ar nic
3213Use ip address of NIC for aliasing, dynamically changing
3214it if NIC's ip address changes.
3215.It Cm log
3216Enable logging on this nat instance.
3217.It Cm deny_in
3218Deny any incoming connection from outside world.
3219.It Cm same_ports
3220Try to leave the alias port numbers unchanged from
3221the actual local port numbers.
3222.It Cm unreg_only
3223Traffic on the local network not originating from an
3224unregistered address spaces will be ignored.
3225.It Cm reset
3226Reset table of the packet aliasing engine on address change.
3227.It Cm reverse
3228Reverse the way libalias handles aliasing.
3229.It Cm proxy_only
3230Obey transparent proxy rules only, packet aliasing is not performed.
3231.It Cm skip_global
3232Skip instance in case of global state lookup (see below).
3233.El
3234.Pp
3235Some specials value can be supplied instead of
3236.Va nat_number:
3237.Bl -tag -width indent
3238.It Cm global
3239Looks up translation state in all configured nat instances.
3240If an entry is found, packet is aliased according to that entry.
3241If no entry was found in any of the instances, packet is passed unchanged,
3242and no new entry will be created.
3243See section
3244.Sx MULTIPLE INSTANCES
3245in
3246.Xr natd 8
3247for more information.
3248.It Cm tablearg
3249Uses argument supplied in lookup table.
3250See
3251.Sx LOOKUP TABLES
3252section below for more information on lookup tables.
3253.El
3254.Pp
3255To let the packet continue after being (de)aliased, set the sysctl variable
3256.Va net.inet.ip.fw.one_pass
3257to 0.
3258For more information about aliasing modes, refer to
3259.Xr libalias 3 .
3260See Section
3261.Sx EXAMPLES
3262for some examples about nat usage.
3263.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
3264Redirect and LSNAT support follow closely the syntax used in
3265.Xr natd 8 .
3266See Section
3267.Sx EXAMPLES
3268for some examples on how to do redirect and lsnat.
3269.Ss SCTP NAT SUPPORT
3270SCTP nat can be configured in a similar manner to TCP through the
3271.Nm
3272command line tool.
3273The main difference is that
3274.Nm sctp nat
3275does not do port translation.
3276Since the local and global side ports will be the same,
3277there is no need to specify both.
3278Ports are redirected as follows:
3279.Bd -ragged -offset indent
3280.Bk -words
3281.Cm nat
3282.Ar nat_number
3283.Cm config if
3284.Ar nic
3285.Cm redirect_port sctp
3286.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
3287.Ek
3288.Ed
3289.Pp
3290Most
3291.Nm sctp nat
3292configuration can be done in real-time through the
3293.Xr sysctl 8
3294interface.
3295All may be changed dynamically, though the hash_table size will only
3296change for new
3297.Nm nat
3298instances.
3299See
3300.Sx SYSCTL VARIABLES
3301for more info.
3302.Sh IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
3303.Ss Stateful translation
3304.Nm
3305supports in-kernel IPv6/IPv4 network address and protocol translation.
3306Stateful NAT64 translation allows IPv6-only clients to contact IPv4 servers
3307using unicast TCP, UDP or ICMP protocols.
3308One or more IPv4 addresses assigned to a stateful NAT64 translator are shared
3309among several IPv6-only clients.
3310When stateful NAT64 is used in conjunction with DNS64, no changes are usually
3311required in the IPv6 client or the IPv4 server.
3312The kernel module
3313.Cm ipfw_nat64
3314should be loaded or kernel should have
3315.Cm options IPFIREWALL_NAT64
3316to be able use stateful NAT64 translator.
3317.Pp
3318Stateful NAT64 uses a bunch of memory for several types of objects.
3319When IPv6 client initiates connection, NAT64 translator creates a host entry
3320in the states table.
3321Each host entry uses preallocated IPv4 alias entry.
3322Each alias entry has a number of ports group entries allocated on demand.
3323Ports group entries contains connection state entries.
3324There are several options to control limits and lifetime for these objects.
3325.Pp
3326NAT64 translator follows RFC7915 when does ICMPv6/ICMP translation,
3327unsupported message types will be silently dropped.
3328IPv6 needs several ICMPv6 message types to be explicitly allowed for correct
3329operation.
3330Make sure that ND6 neighbor solicitation (ICMPv6 type 135) and neighbor
3331advertisement (ICMPv6 type 136) messages will not be handled by translation
3332rules.
3333.Pp
3334After translation NAT64 translator by default sends packets through
3335corresponding netisr queue.
3336Thus translator host should be configured as IPv4 and IPv6 router.
3337Also this means, that a packet is handled by firewall twice.
3338First time an original packet is handled and consumed by translator,
3339and then it is handled again as translated packet.
3340This behavior can be changed by sysctl variable
3341.Va net.inet.ip.fw.nat64_direct_output .
3342Also translated packet can be tagged using
3343.Cm tag
3344rule action, and then matched by
3345.Cm tagged
3346opcode to avoid loops and extra overhead.
3347.Pp
3348The stateful NAT64 configuration command is the following:
3349.Bd -ragged -offset indent
3350.Bk -words
3351.Cm nat64lsn
3352.Ar name
3353.Cm create
3354.Ar create-options
3355.Ek
3356.Ed
3357.Pp
3358The following parameters can be configured:
3359.Bl -tag -width indent
3360.It Cm prefix4 Ar ipv4_prefix/plen
3361The IPv4 prefix with mask defines the pool of IPv4 addresses used as
3362source address after translation.
3363Stateful NAT64 module translates IPv6 source address of client to one
3364IPv4 address from this pool.
3365Note that incoming IPv4 packets that don't have corresponding state entry
3366in the states table will be dropped by translator.
3367Make sure that translation rules handle packets, destined to configured prefix.
3368.It Cm prefix6 Ar ipv6_prefix/length
3369The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3370to represent IPv4 addresses. This IPv6 prefix should be configured in DNS64.
3371The translator implementation follows RFC6052, that restricts the length of
3372prefixes to one of following: 32, 40, 48, 56, 64, or 96.
3373The Well-Known IPv6 Prefix 64:ff9b:: must be 96 bits long.
3374The special
3375.Ar ::/length
3376prefix can be used to handle several IPv6 prefixes with one NAT64 instance.
3377The NAT64 instance will determine a destination IPv4 address from prefix
3378.Ar length .
3379.It Cm states_chunks Ar number
3380The number of states chunks in single ports group.
3381Each ports group by default can keep 64 state entries in single chunk.
3382The above value affects the maximum number of states that can be associated with single IPv4 alias address and port.
3383The value must be power of 2, and up to 128.
3384.It Cm host_del_age Ar seconds
3385The number of seconds until the host entry for a IPv6 client will be deleted
3386and all its resources will be released due to inactivity.
3387Default value is
3388.Ar 3600 .
3389.It Cm pg_del_age Ar seconds
3390The number of seconds until a ports group with unused state entries will
3391be released.
3392Default value is
3393.Ar 900 .
3394.It Cm tcp_syn_age Ar seconds
3395The number of seconds while a state entry for TCP connection with only SYN
3396sent will be kept.
3397If TCP connection establishing will not be finished,
3398state entry will be deleted.
3399Default value is
3400.Ar 10 .
3401.It Cm tcp_est_age Ar seconds
3402The number of seconds while a state entry for established TCP connection
3403will be kept.
3404Default value is
3405.Ar 7200 .
3406.It Cm tcp_close_age Ar seconds
3407The number of seconds while a state entry for closed TCP connection
3408will be kept.
3409Keeping state entries for closed connections is needed, because IPv4 servers
3410typically keep closed connections in a TIME_WAIT state for a several minutes.
3411Since translator's IPv4 addresses are shared among all IPv6 clients,
3412new connections from the same addresses and ports may be rejected by server,
3413because these connections are still in a TIME_WAIT state.
3414Keeping them in translator's state table protects from such rejects.
3415Default value is
3416.Ar 180 .
3417.It Cm udp_age Ar seconds
3418The number of seconds while translator keeps state entry in a waiting for
3419reply to the sent UDP datagram.
3420Default value is
3421.Ar 120 .
3422.It Cm icmp_age Ar seconds
3423The number of seconds while translator keeps state entry in a waiting for
3424reply to the sent ICMP message.
3425Default value is
3426.Ar 60 .
3427.It Cm log
3428Turn on logging of all handled packets via BPF through
3429.Ar ipfwlog0
3430interface.
3431.Ar ipfwlog0
3432is a pseudo interface and can be created after a boot manually with
3433.Cm ifconfig
3434command.
3435Note that it has different purpose than
3436.Ar ipfw0
3437interface.
3438Translators sends to BPF an additional information with each packet.
3439With
3440.Cm tcpdump
3441you are able to see each handled packet before and after translation.
3442.It Cm -log
3443Turn off logging of all handled packets via BPF.
3444.It Cm allow_private
3445Turn on processing private IPv4 addresses. By default IPv6 packets with
3446destinations mapped to private address ranges defined by RFC1918 are not
3447processed.
3448.It Cm -allow_private
3449Turn off private address handling in
3450.Nm nat64
3451instance.
3452.El
3453.Pp
3454To inspect a states table of stateful NAT64 the following command can be used:
3455.Bd -ragged -offset indent
3456.Bk -words
3457.Cm nat64lsn
3458.Ar name
3459.Cm show Cm states
3460.Ek
3461.Ed
3462.Pp
3463.Pp
3464Stateless NAT64 translator doesn't use a states table for translation
3465and converts IPv4 addresses to IPv6 and vice versa solely based on the
3466mappings taken from configured lookup tables.
3467Since a states table doesn't used by stateless translator,
3468it can be configured to pass IPv4 clients to IPv6-only servers.
3469.Pp
3470The stateless NAT64 configuration command is the following:
3471.Bd -ragged -offset indent
3472.Bk -words
3473.Cm nat64stl
3474.Ar name
3475.Cm create
3476.Ar create-options
3477.Ek
3478.Ed
3479.Pp
3480The following parameters can be configured:
3481.Bl -tag -width indent
3482.It Cm prefix6 Ar ipv6_prefix/length
3483The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3484to represent IPv4 addresses. This IPv6 prefix should be configured in DNS64.
3485.It Cm table4 Ar table46
3486The lookup table
3487.Ar table46
3488contains mapping how IPv4 addresses should be translated to IPv6 addresses.
3489.It Cm table6 Ar table64
3490The lookup table
3491.Ar table64
3492contains mapping how IPv6 addresses should be translated to IPv4 addresses.
3493.It Cm log
3494Turn on logging of all handled packets via BPF through
3495.Ar ipfwlog0
3496interface.
3497.It Cm -log
3498Turn off logging of all handled packets via BPF.
3499.It Cm allow_private
3500Turn on processing private IPv4 addresses. By default IPv6 packets with
3501destinations mapped to private address ranges defined by RFC1918 are not
3502processed.
3503.It Cm -allow_private
3504Turn off private address handling in
3505.Nm nat64
3506instance.
3507.El
3508.Pp
3509Note that the behavior of stateless translator with respect to not matched
3510packets differs from stateful translator.
3511If corresponding addresses was not found in the lookup tables, the packet
3512will not be dropped and the search continues.
3513.Pp
3514.Pp
3515.Ss XLAT464 CLAT translation
3516XLAT464 CLAT NAT64 translator implements client-side stateless translation as
3517defined in RFC6877 and is very similar to statless NAT64 translator
3518explained above. Instead of lookup tables it uses one-to-one mapping
3519between IPv4 and IPv6 addresses using configured prefixes.
3520This mode can be used as a replacement of DNS64 service for applications
3521that are not using it (e.g. VoIP) allowing them to access IPv4-only Internet
3522over IPv6-only networks with help of remote NAT64 translator.
3523.Pp
3524The CLAT NAT64 configuration command is the following:
3525.Bd -ragged -offset indent
3526.Bk -words
3527.Cm nat64clat
3528.Ar name
3529.Cm create
3530.Ar create-options
3531.Ek
3532.Ed
3533.Pp
3534The following parameters can be configured:
3535.Bl -tag -width indent
3536.It Cm clat_prefix Ar ipv6_prefix/length
3537The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3538to represent source IPv4 addresses.
3539.It Cm plat_prefix Ar ipv6_prefix/length
3540The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3541to represent destination IPv4 addresses. This IPv6 prefix should be configured
3542on a remote NAT64 translator.
3543.It Cm log
3544Turn on logging of all handled packets via BPF through
3545.Ar ipfwlog0
3546interface.
3547.It Cm -log
3548Turn off logging of all handled packets via BPF.
3549.It Cm allow_private
3550Turn on processing private IPv4 addresses. By default
3551.Nm nat64clat
3552instance will not process IPv4 packets with destination address from private
3553ranges as defined in RFC1918.
3554.It Cm -allow_private
3555Turn off private address handling in
3556.Nm nat64clat
3557instance.
3558.El
3559.Pp
3560Note that the behavior of CLAT translator with respect to not matched
3561packets differs from stateful translator.
3562If corresponding addresses were not matched against prefixes configured,
3563the packet will not be dropped and the search continues.
3564.Sh IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
3565.Nm
3566supports in-kernel IPv6-to-IPv6 network prefix translation as described
3567in RFC6296.
3568The kernel module
3569.Cm ipfw_nptv6
3570should be loaded or kernel should has
3571.Cm options IPFIREWALL_NPTV6
3572to be able use NPTv6 translator.
3573.Pp
3574The NPTv6 configuration command is the following:
3575.Bd -ragged -offset indent
3576.Bk -words
3577.Cm nptv6
3578.Ar name
3579.Cm create
3580.Ar create-options
3581.Ek
3582.Ed
3583.Pp
3584The following parameters can be configured:
3585.Bl -tag -width indent
3586.It Cm int_prefix Ar ipv6_prefix
3587IPv6 prefix used in internal network.
3588NPTv6 module translates source address when it matches this prefix.
3589.It Cm ext_prefix Ar ipv6_prefix
3590IPv6 prefix used in external network.
3591NPTv6 module translates destination address when it matches this prefix.
3592.It Cm ext_if Ar nic
3593The NPTv6 module will use first global IPv6 address from interface
3594.Ar nic
3595as external prefix.
3596It can be useful when IPv6 prefix of external network is dynamically obtained.
3597.Cm ext_prefix
3598and
3599.Cm ext_if
3600options are mutually exclusive.
3601.It Cm prefixlen Ar length
3602The length of specified IPv6 prefixes. It must be in range from 8 to 64.
3603.El
3604.Pp
3605Note that the prefix translation rules are silently ignored when IPv6 packet
3606forwarding is disabled.
3607To enable the packet forwarding, set the sysctl variable
3608.Va net.inet6.ip6.forwarding
3609to 1.
3610.Pp
3611To let the packet continue after being translated, set the sysctl variable
3612.Va net.inet.ip.fw.one_pass
3613to 0.
3614.Sh LOADER TUNABLES
3615Tunables can be set in
3616.Xr loader 8
3617prompt,
3618.Xr loader.conf 5
3619or
3620.Xr kenv 1
3621before ipfw module gets loaded.
3622.Bl -tag -width indent
3623.It Va net.inet.ip.fw.default_to_accept: No 0
3624Defines ipfw last rule behavior.
3625This value overrides
3626.Cd "options IPFW_DEFAULT_TO_(ACCEPT|DENY)"
3627from kernel configuration file.
3628.It Va net.inet.ip.fw.tables_max: No 128
3629Defines number of tables available in ipfw.
3630Number cannot exceed 65534.
3631.El
3632.Sh SYSCTL VARIABLES
3633A set of
3634.Xr sysctl 8
3635variables controls the behaviour of the firewall and
3636associated modules
3637.Pq Nm dummynet , bridge , sctp nat .
3638These are shown below together with their default value
3639(but always check with the
3640.Xr sysctl 8
3641command what value is actually in use) and meaning:
3642.Bl -tag -width indent
3643.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
3644Defines how the
3645.Nm nat
3646responds to receipt of global OOTB ASCONF-AddIP:
3647.Bl -tag -width indent
3648.It Cm 0
3649No response (unless a partially matching association exists -
3650ports and vtags match but global address does not)
3651.It Cm 1
3652.Nm nat
3653will accept and process all OOTB global AddIP messages.
3654.El
3655.Pp
3656Option 1 should never be selected as this forms a security risk.
3657An attacker can
3658establish multiple fake associations by sending AddIP messages.
3659.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
3660Defines the maximum number of chunks in an SCTP packet that will be
3661parsed for a
3662packet that matches an existing association.
3663This value is enforced to be greater or equal than
3664.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
3665A high value is
3666a DoS risk yet setting too low a value may result in
3667important control chunks in
3668the packet not being located and parsed.
3669.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
3670Defines when the
3671.Nm nat
3672responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
3673An OOTB packet is a packet that arrives with no existing association
3674registered in the
3675.Nm nat
3676and is not an INIT or ASCONF-AddIP packet:
3677.Bl -tag -width indent
3678.It Cm 0
3679ErrorM is never sent in response to OOTB packets.
3680.It Cm 1
3681ErrorM is only sent to OOTB packets received on the local side.
3682.It Cm 2
3683ErrorM is sent to the local side and on the global side ONLY if there is a
3684partial match (ports and vtags match but the source global IP does not).
3685This value is only useful if the
3686.Nm nat
3687is tracking global IP addresses.
3688.It Cm 3
3689ErrorM is sent in response to all OOTB packets on both
3690the local and global side
3691(DoS risk).
3692.El
3693.Pp
3694At the moment the default is 0, since the ErrorM packet is not yet
3695supported by most SCTP stacks.
3696When it is supported, and if not tracking
3697global addresses, we recommend setting this value to 1 to allow
3698multi-homed local hosts to function with the
3699.Nm nat .
3700To track global addresses, we recommend setting this value to 2 to
3701allow global hosts to be informed when they need to (re)send an
3702ASCONF-AddIP.
3703Value 3 should never be chosen (except for debugging) as the
3704.Nm nat
3705will respond to all OOTB global packets (a DoS risk).
3706.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
3707Size of hash tables used for
3708.Nm nat
3709lookups (100 < prime_number > 1000001).
3710This value sets the
3711.Nm hash table
3712size for any future created
3713.Nm nat
3714instance and therefore must be set prior to creating a
3715.Nm nat
3716instance.
3717The table sizes may be changed to suit specific needs.
3718If there will be few
3719concurrent associations, and memory is scarce, you may make these smaller.
3720If there will be many thousands (or millions) of concurrent associations, you
3721should make these larger.
3722A prime number is best for the table size.
3723The sysctl
3724update function will adjust your input value to the next highest prime number.
3725.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
3726Hold association in table for this many seconds after receiving a
3727SHUTDOWN-COMPLETE.
3728This allows endpoints to correct shutdown gracefully if a
3729shutdown_complete is lost and retransmissions are required.
3730.It Va net.inet.ip.alias.sctp.init_timer: No 15
3731Timeout value while waiting for (INIT-ACK|AddIP-ACK).
3732This value cannot be 0.
3733.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
3734Defines the maximum number of chunks in an SCTP packet that will be parsed when
3735no existing association exists that matches that packet.
3736Ideally this packet
3737will only be an INIT or ASCONF-AddIP packet.
3738A higher value may become a DoS
3739risk as malformed packets can consume processing resources.
3740.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
3741Defines the maximum number of parameters within a chunk that will be
3742parsed in a
3743packet.
3744As for other similar sysctl variables, larger values pose a DoS risk.
3745.It Va net.inet.ip.alias.sctp.log_level: No 0
3746Level of detail in the system log messages (0 \- minimal, 1 \- event,
37472 \- info, 3 \- detail, 4 \- debug, 5 \- max debug).
3748May be a good
3749option in high loss environments.
3750.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
3751Timeout value while waiting for SHUTDOWN-COMPLETE.
3752This value cannot be 0.
3753.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
3754Enables/disables global IP address tracking within the
3755.Nm nat
3756and places an
3757upper limit on the number of addresses tracked for each association:
3758.Bl -tag -width indent
3759.It Cm 0
3760Global tracking is disabled
3761.It Cm >1
3762Enables tracking, the maximum number of addresses tracked for each
3763association is limited to this value
3764.El
3765.Pp
3766This variable is fully dynamic, the new value will be adopted for all newly
3767arriving associations, existing associations are treated
3768as they were previously.
3769Global tracking will decrease the number of collisions within the
3770.Nm nat
3771at a cost
3772of increased processing load, memory usage, complexity, and possible
3773.Nm nat
3774state
3775problems in complex networks with multiple
3776.Nm nats .
3777We recommend not tracking
3778global IP addresses, this will still result in a fully functional
3779.Nm nat .
3780.It Va net.inet.ip.alias.sctp.up_timer: No 300
3781Timeout value to keep an association up with no traffic.
3782This value cannot be 0.
3783.It Va net.inet.ip.dummynet.codel.interval : No 100000
3784Default
3785.Cm codel
3786AQM interval in microseconds.
3787The value must be in the range 1..5000000.
3788.It Va net.inet.ip.dummynet.codel.target : No 5000
3789Default
3790.Cm codel
3791AQM target delay time in microseconds (the minimum acceptable persistent queue
3792delay).
3793The value must be in the range 1..5000000.
3794.It Va net.inet.ip.dummynet.expire : No 1
3795Lazily delete dynamic pipes/queue once they have no pending traffic.
3796You can disable this by setting the variable to 0, in which case
3797the pipes/queues will only be deleted when the threshold is reached.
3798.It Va net.inet.ip.dummynet.fqcodel.flows : No 1024
3799Defines the default total number of flow queues (sub-queues) that
3800.Cm fq_codel
3801creates and manages.
3802The value must be in the range 1..65536.
3803.It Va net.inet.ip.dummynet.fqcodel.interval : No 100000
3804Default
3805.Cm fq_codel
3806scheduler/AQM interval in microseconds.
3807The value must be in the range 1..5000000.
3808.It Va net.inet.ip.dummynet.fqcodel.limit : No 10240
3809The default hard size limit (in unit of packet) of all queues managed by an
3810instance of the
3811.Cm fq_codel
3812scheduler.
3813The value must be in the range 1..20480.
3814.It Va net.inet.ip.dummynet.fqcodel.quantum : No 1514
3815The default quantum (credit) of the
3816.Cm fq_codel
3817in unit of byte.
3818The value must be in the range 1..9000.
3819.It Va net.inet.ip.dummynet.fqcodel.target : No 5000
3820Default
3821.Cm fq_codel
3822scheduler/AQM target delay time in microseconds (the minimum acceptable
3823persistent queue delay).
3824The value must be in the range 1..5000000.
3825.It Va net.inet.ip.dummynet.fqpie.alpha : No 125
3826The default
3827.Ar alpha
3828parameter (scaled by 1000) for
3829.Cm fq_pie
3830scheduler/AQM.
3831The value must be in the range 1..7000.
3832.It Va net.inet.ip.dummynet.fqpie.beta : No 1250
3833The default
3834.Ar beta
3835parameter (scaled by 1000) for
3836.Cm fq_pie
3837scheduler/AQM.
3838The value must be in the range 1..7000.
3839.It Va net.inet.ip.dummynet.fqpie.flows : No 1024
3840Defines the default total number of flow queues (sub-queues) that
3841.Cm fq_pie
3842creates and manages.
3843The value must be in the range 1..65536.
3844.It Va net.inet.ip.dummynet.fqpie.limit : No 10240
3845The default hard size limit (in unit of packet) of all queues managed by an
3846instance of the
3847.Cm fq_pie
3848scheduler.
3849The value must be in the range 1..20480.
3850.It Va net.inet.ip.dummynet.fqpie.max_burst : No 150000
3851The default maximum period of microseconds that
3852.Cm fq_pie
3853scheduler/AQM does not drop/mark packets.
3854The value must be in the range 1..10000000.
3855.It Va net.inet.ip.dummynet.fqpie.max_ecnth : No 99
3856The default maximum ECN probability threshold (scaled by 1000) for
3857.Cm fq_pie
3858scheduler/AQM.
3859The value must be in the range 1..7000.
3860.It Va net.inet.ip.dummynet.fqpie.quantum : No 1514
3861The default quantum (credit) of the
3862.Cm fq_pie
3863in unit of byte.
3864The value must be in the range 1..9000.
3865.It Va net.inet.ip.dummynet.fqpie.target : No 15000
3866The default
3867.Cm target
3868delay of the
3869.Cm fq_pie
3870in unit of microsecond.
3871The value must be in the range 1..5000000.
3872.It Va net.inet.ip.dummynet.fqpie.tupdate : No 15000
3873The default
3874.Cm tupdate
3875of the
3876.Cm fq_pie
3877in unit of microsecond.
3878The value must be in the range 1..5000000.
3879.It Va net.inet.ip.dummynet.hash_size : No 64
3880Default size of the hash table used for dynamic pipes/queues.
3881This value is used when no
3882.Cm buckets
3883option is specified when configuring a pipe/queue.
3884.It Va net.inet.ip.dummynet.io_fast : No 0
3885If set to a non-zero value,
3886the
3887.Dq fast
3888mode of
3889.Nm dummynet
3890operation (see above) is enabled.
3891.It Va net.inet.ip.dummynet.io_pkt
3892Number of packets passed to
3893.Nm dummynet .
3894.It Va net.inet.ip.dummynet.io_pkt_drop
3895Number of packets dropped by
3896.Nm dummynet .
3897.It Va net.inet.ip.dummynet.io_pkt_fast
3898Number of packets bypassed by the
3899.Nm dummynet
3900scheduler.
3901.It Va net.inet.ip.dummynet.max_chain_len : No 16
3902Target value for the maximum number of pipes/queues in a hash bucket.
3903The product
3904.Cm max_chain_len*hash_size
3905is used to determine the threshold over which empty pipes/queues
3906will be expired even when
3907.Cm net.inet.ip.dummynet.expire=0 .
3908.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
3909.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
3910.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
3911Parameters used in the computations of the drop probability
3912for the RED algorithm.
3913.It Va net.inet.ip.dummynet.pie.alpha : No 125
3914The default
3915.Ar alpha
3916parameter (scaled by 1000) for
3917.Cm pie
3918AQM.
3919The value must be in the range 1..7000.
3920.It Va net.inet.ip.dummynet.pie.beta : No 1250
3921The default
3922.Ar beta
3923parameter (scaled by 1000) for
3924.Cm pie
3925AQM.
3926The value must be in the range 1..7000.
3927.It Va net.inet.ip.dummynet.pie.max_burst : No 150000
3928The default maximum period of microseconds that
3929.Cm pie
3930AQM does not drop/mark packets.
3931The value must be in the range 1..10000000.
3932.It Va net.inet.ip.dummynet.pie.max_ecnth : No 99
3933The default maximum ECN probability threshold (scaled by 1000) for
3934.Cm pie
3935AQM.
3936The value must be in the range 1..7000.
3937.It Va net.inet.ip.dummynet.pie.target : No 15000
3938The default
3939.Cm target
3940delay of
3941.Cm pie
3942AQM in unit of microsecond.
3943The value must be in the range 1..5000000.
3944.It Va net.inet.ip.dummynet.pie.tupdate : No 15000
3945The default
3946.Cm tupdate
3947of
3948.Cm pie
3949AQM in unit of microsecond.
3950The value must be in the range 1..5000000.
3951.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
3952.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
3953The maximum queue size that can be specified in bytes or packets.
3954These limits prevent accidental exhaustion of resources such as mbufs.
3955If you raise these limits,
3956you should make sure the system is configured so that sufficient resources
3957are available.
3958.It Va net.inet.ip.fw.autoinc_step : No 100
3959Delta between rule numbers when auto-generating them.
3960The value must be in the range 1..1000.
3961.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
3962The current number of buckets in the hash table for dynamic rules
3963(readonly).
3964.It Va net.inet.ip.fw.debug : No 1
3965Controls debugging messages produced by
3966.Nm .
3967.It Va net.inet.ip.fw.default_rule : No 65535
3968The default rule number (read-only).
3969By the design of
3970.Nm , the default rule is the last one, so its number
3971can also serve as the highest number allowed for a rule.
3972.It Va net.inet.ip.fw.dyn_buckets : No 256
3973The number of buckets in the hash table for dynamic rules.
3974Must be a power of 2, up to 65536.
3975It only takes effect when all dynamic rules have expired, so you
3976are advised to use a
3977.Cm flush
3978command to make sure that the hash table is resized.
3979.It Va net.inet.ip.fw.dyn_count : No 3
3980Current number of dynamic rules
3981(read-only).
3982.It Va net.inet.ip.fw.dyn_keepalive : No 1
3983Enables generation of keepalive packets for
3984.Cm keep-state
3985rules on TCP sessions.
3986A keepalive is generated to both
3987sides of the connection every 5 seconds for the last 20
3988seconds of the lifetime of the rule.
3989.It Va net.inet.ip.fw.dyn_max : No 8192
3990Maximum number of dynamic rules.
3991When you hit this limit, no more dynamic rules can be
3992installed until old ones expire.
3993.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
3994.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
3995.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
3996.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
3997.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
3998.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
3999These variables control the lifetime, in seconds, of dynamic
4000rules.
4001Upon the initial SYN exchange the lifetime is kept short,
4002then increased after both SYN have been seen, then decreased
4003again during the final FIN exchange or when a RST is received.
4004Both
4005.Em dyn_fin_lifetime
4006and
4007.Em dyn_rst_lifetime
4008must be strictly lower than 5 seconds, the period of
4009repetition of keepalives.
4010The firewall enforces that.
4011.It Va net.inet.ip.fw.dyn_keep_states: No 0
4012Keep dynamic states on rule/set deletion.
4013States are relinked to default rule (65535).
4014This can be handly for ruleset reload.
4015Turned off by default.
4016.It Va net.inet.ip.fw.enable : No 1
4017Enables the firewall.
4018Setting this variable to 0 lets you run your machine without
4019firewall even if compiled in.
4020.It Va net.inet6.ip6.fw.enable : No 1
4021provides the same functionality as above for the IPv6 case.
4022.It Va net.inet.ip.fw.one_pass : No 1
4023When set, the packet exiting from the
4024.Nm dummynet
4025pipe or from
4026.Xr ng_ipfw 4
4027node is not passed though the firewall again.
4028Otherwise, after an action, the packet is
4029reinjected into the firewall at the next rule.
4030.It Va net.inet.ip.fw.tables_max : No 128
4031Maximum number of tables.
4032.It Va net.inet.ip.fw.verbose : No 1
4033Enables verbose messages.
4034.It Va net.inet.ip.fw.verbose_limit : No 0
4035Limits the number of messages produced by a verbose firewall.
4036.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
4037If enabled packets with unknown IPv6 Extension Headers will be denied.
4038.It Va net.link.ether.ipfw : No 0
4039Controls whether layer-2 packets are passed to
4040.Nm .
4041Default is no.
4042.It Va net.link.bridge.ipfw : No 0
4043Controls whether bridged packets are passed to
4044.Nm .
4045Default is no.
4046.It Va net.inet.ip.fw.nat64_debug : No 0
4047Controls debugging messages produced by
4048.Nm ipfw_nat64
4049module.
4050.It Va net.inet.ip.fw.nat64_direct_output : No 0
4051Controls the output method used by
4052.Nm ipfw_nat64
4053module:
4054.Bl -tag -width indent
4055.It Cm 0
4056A packet is handled by
4057.Nm ipfw
4058twice.
4059First time an original packet is handled by
4060.Nm ipfw
4061and consumed by
4062.Nm ipfw_nat64
4063translator.
4064Then translated packet is queued via netisr to input processing again.
4065.It Cm 1
4066A packet is handled by
4067.Nm ipfw
4068only once, and after translation it will be pushed directly to outgoing
4069interface.
4070.El
4071.El
4072.Sh INTERNAL DIAGNOSTICS
4073There are some commands that may be useful to understand current state
4074of certain subsystems inside kernel module.
4075These commands provide debugging output which may change without notice.
4076.Pp
4077Currently the following commands are available as
4078.Cm internal
4079sub-options:
4080.Bl -tag -width indent
4081.It Cm iflist
4082Lists all interface which are currently tracked by
4083.Nm
4084with their in-kernel status.
4085.It Cm talist
4086List all table lookup algorithms currently available.
4087.El
4088.Sh EXAMPLES
4089There are far too many possible uses of
4090.Nm
4091so this Section will only give a small set of examples.
4092.Pp
4093.Ss BASIC PACKET FILTERING
4094This command adds an entry which denies all tcp packets from
4095.Em cracker.evil.org
4096to the telnet port of
4097.Em wolf.tambov.su
4098from being forwarded by the host:
4099.Pp
4100.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
4101.Pp
4102This one disallows any connection from the entire cracker's
4103network to my host:
4104.Pp
4105.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
4106.Pp
4107A first and efficient way to limit access (not using dynamic rules)
4108is the use of the following rules:
4109.Pp
4110.Dl "ipfw add allow tcp from any to any established"
4111.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
4112.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
4113.Dl "..."
4114.Dl "ipfw add deny tcp from any to any"
4115.Pp
4116The first rule will be a quick match for normal TCP packets,
4117but it will not match the initial SYN packet, which will be
4118matched by the
4119.Cm setup
4120rules only for selected source/destination pairs.
4121All other SYN packets will be rejected by the final
4122.Cm deny
4123rule.
4124.Pp
4125If you administer one or more subnets, you can take advantage
4126of the address sets and or-blocks and write extremely
4127compact rulesets which selectively enable services to blocks
4128of clients, as below:
4129.Pp
4130.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
4131.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
4132.Dl ""
4133.Dl "ipfw add allow ip from ${goodguys} to any"
4134.Dl "ipfw add deny ip from ${badguys} to any"
4135.Dl "... normal policies ..."
4136.Pp
4137The
4138.Cm verrevpath
4139option could be used to do automated anti-spoofing by adding the
4140following to the top of a ruleset:
4141.Pp
4142.Dl "ipfw add deny ip from any to any not verrevpath in"
4143.Pp
4144This rule drops all incoming packets that appear to be coming to the
4145system on the wrong interface.
4146For example, a packet with a source
4147address belonging to a host on a protected internal network would be
4148dropped if it tried to enter the system from an external interface.
4149.Pp
4150The
4151.Cm antispoof
4152option could be used to do similar but more restricted anti-spoofing
4153by adding the following to the top of a ruleset:
4154.Pp
4155.Dl "ipfw add deny ip from any to any not antispoof in"
4156.Pp
4157This rule drops all incoming packets that appear to be coming from another
4158directly connected system but on the wrong interface.
4159For example, a packet with a source address of
4160.Li 192.168.0.0/24 ,
4161configured on
4162.Li fxp0 ,
4163but coming in on
4164.Li fxp1
4165would be dropped.
4166.Pp
4167The
4168.Cm setdscp
4169option could be used to (re)mark user traffic,
4170by adding the following to the appropriate place in ruleset:
4171.Pp
4172.Dl "ipfw add setdscp be ip from any to any dscp af11,af21"
4173.Ss SELECTIVE MIRRORING
4174If your network has network traffic analyzer
4175connected to your host directly via dedicated interface
4176or remotely via RSPAN vlan, you can selectively mirror
4177some Ethernet layer2 frames to the analyzer.
4178.Pp
4179First, make sure your firewall is already configured and runs.
4180Then, enable layer2 processing if not already enabled:
4181.Pp
4182.Dl "sysctl net.link.ether.ipfw=1"
4183.Pp
4184Next, load needed additional kernel modules:
4185.Pp
4186.Dl "kldload ng_ether ng_ipfw"
4187.Pp
4188Optionally, make system load these modules automatically
4189at startup:
4190.Pp
4191.Dl sysrc kld_list+="ng_ether ng_ipfw"
4192.Pp
4193Next, configure
4194.Xr ng_ipfw 4
4195kernel module to transmit mirrored copies of layer2 frames
4196out via vlan900 interface:
4197.Pp
4198.Dl "ngctl connect ipfw: vlan900: 1 lower"
4199.Pp
4200Think of "1" here as of "mirroring instance index" and vlan900 is its
4201destination.
4202You can have arbitrary number of instances.
4203Refer to
4204.Xr ng_ipfw 4
4205for details.
4206.Pp
4207At last, actually start mirroring of selected frames using "instance 1".
4208For frames incoming from em0 interface:
4209.Pp
4210.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 in recv em0"
4211.Pp
4212For frames outgoing to em0 interface:
4213.Pp
4214.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 out xmit em0"
4215.Pp
4216For both incoming and outgoing frames while flowing through em0:
4217.Pp
4218.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 via em0"
4219.Pp
4220Make sure you do not perform mirroring for already duplicated frames
4221or kernel may hang as there is no safety net.
4222.Ss DYNAMIC RULES
4223In order to protect a site from flood attacks involving fake
4224TCP packets, it is safer to use dynamic rules:
4225.Pp
4226.Dl "ipfw add check-state"
4227.Dl "ipfw add deny tcp from any to any established"
4228.Dl "ipfw add allow tcp from my-net to any setup keep-state"
4229.Pp
4230This will let the firewall install dynamic rules only for
4231those connection which start with a regular SYN packet coming
4232from the inside of our network.
4233Dynamic rules are checked when encountering the first
4234occurrence of a
4235.Cm check-state ,
4236.Cm keep-state
4237or
4238.Cm limit
4239rule.
4240A
4241.Cm check-state
4242rule should usually be placed near the beginning of the
4243ruleset to minimize the amount of work scanning the ruleset.
4244Your mileage may vary.
4245.Pp
4246For more complex scenarios with dynamic rules
4247.Cm record-state
4248and
4249.Cm defer-action
4250can be used to precisely control creation and checking of dynamic rules.
4251Example of usage of these options are provided in
4252.Sx NETWORK ADDRESS TRANSLATION (NAT)
4253Section.
4254.Pp
4255To limit the number of connections a user can open
4256you can use the following type of rules:
4257.Pp
4258.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
4259.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
4260.Pp
4261The former (assuming it runs on a gateway) will allow each host
4262on a /24 network to open at most 10 TCP connections.
4263The latter can be placed on a server to make sure that a single
4264client does not use more than 4 simultaneous connections.
4265.Pp
4266.Em BEWARE :
4267stateful rules can be subject to denial-of-service attacks
4268by a SYN-flood which opens a huge number of dynamic rules.
4269The effects of such attacks can be partially limited by
4270acting on a set of
4271.Xr sysctl 8
4272variables which control the operation of the firewall.
4273.Pp
4274Here is a good usage of the
4275.Cm list
4276command to see accounting records and timestamp information:
4277.Pp
4278.Dl ipfw -at list
4279.Pp
4280or in short form without timestamps:
4281.Pp
4282.Dl ipfw -a list
4283.Pp
4284which is equivalent to:
4285.Pp
4286.Dl ipfw show
4287.Pp
4288Next rule diverts all incoming packets from 192.168.2.0/24
4289to divert port 5000:
4290.Pp
4291.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
4292.Ss TRAFFIC SHAPING
4293The following rules show some of the applications of
4294.Nm
4295and
4296.Nm dummynet
4297for simulations and the like.
4298.Pp
4299This rule drops random incoming packets with a probability
4300of 5%:
4301.Pp
4302.Dl "ipfw add prob 0.05 deny ip from any to any in"
4303.Pp
4304A similar effect can be achieved making use of
4305.Nm dummynet
4306pipes:
4307.Pp
4308.Dl "ipfw add pipe 10 ip from any to any"
4309.Dl "ipfw pipe 10 config plr 0.05"
4310.Pp
4311We can use pipes to artificially limit bandwidth, e.g.\& on a
4312machine acting as a router, if we want to limit traffic from
4313local clients on 192.168.2.0/24 we do:
4314.Pp
4315.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
4316.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
4317.Pp
4318note that we use the
4319.Cm out
4320modifier so that the rule is not used twice.
4321Remember in fact that
4322.Nm
4323rules are checked both on incoming and outgoing packets.
4324.Pp
4325Should we want to simulate a bidirectional link with bandwidth
4326limitations, the correct way is the following:
4327.Pp
4328.Dl "ipfw add pipe 1 ip from any to any out"
4329.Dl "ipfw add pipe 2 ip from any to any in"
4330.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
4331.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
4332.Pp
4333The above can be very useful, e.g.\& if you want to see how
4334your fancy Web page will look for a residential user who
4335is connected only through a slow link.
4336You should not use only one pipe for both directions, unless
4337you want to simulate a half-duplex medium (e.g.\& AppleTalk,
4338Ethernet, IRDA).
4339It is not necessary that both pipes have the same configuration,
4340so we can also simulate asymmetric links.
4341.Pp
4342Should we want to verify network performance with the RED queue
4343management algorithm:
4344.Pp
4345.Dl "ipfw add pipe 1 ip from any to any"
4346.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
4347.Pp
4348Another typical application of the traffic shaper is to
4349introduce some delay in the communication.
4350This can significantly affect applications which do a lot of Remote
4351Procedure Calls, and where the round-trip-time of the
4352connection often becomes a limiting factor much more than
4353bandwidth:
4354.Pp
4355.Dl "ipfw add pipe 1 ip from any to any out"
4356.Dl "ipfw add pipe 2 ip from any to any in"
4357.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
4358.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
4359.Pp
4360Per-flow queueing can be useful for a variety of purposes.
4361A very simple one is counting traffic:
4362.Pp
4363.Dl "ipfw add pipe 1 tcp from any to any"
4364.Dl "ipfw add pipe 1 udp from any to any"
4365.Dl "ipfw add pipe 1 ip from any to any"
4366.Dl "ipfw pipe 1 config mask all"
4367.Pp
4368The above set of rules will create queues (and collect
4369statistics) for all traffic.
4370Because the pipes have no limitations, the only effect is
4371collecting statistics.
4372Note that we need 3 rules, not just the last one, because
4373when
4374.Nm
4375tries to match IP packets it will not consider ports, so we
4376would not see connections on separate ports as different
4377ones.
4378.Pp
4379A more sophisticated example is limiting the outbound traffic
4380on a net with per-host limits, rather than per-network limits:
4381.Pp
4382.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
4383.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
4384.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
4385.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
4386.Ss LOOKUP TABLES
4387In the following example, we need to create several traffic bandwidth
4388classes and we need different hosts/networks to fall into different classes.
4389We create one pipe for each class and configure them accordingly.
4390Then we create a single table and fill it with IP subnets and addresses.
4391For each subnet/host we set the argument equal to the number of the pipe
4392that it should use.
4393Then we classify traffic using a single rule:
4394.Pp
4395.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
4396.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
4397.Dl "..."
4398.Dl "ipfw table T1 create type addr"
4399.Dl "ipfw table T1 add 192.168.2.0/24 1"
4400.Dl "ipfw table T1 add 192.168.0.0/27 4"
4401.Dl "ipfw table T1 add 192.168.0.2 1"
4402.Dl "..."
4403.Dl "ipfw add pipe tablearg ip from 'table(T1)' to any"
4404.Pp
4405Using the
4406.Cm fwd
4407action, the table entries may include hostnames and IP addresses.
4408.Pp
4409.Dl "ipfw table T2 create type addr ftype ip"
4410.Dl "ipfw table T2 add 192.168.2.0/24 10.23.2.1"
4411.Dl "ipfw table T21 add 192.168.0.0/27 router1.dmz"
4412.Dl "..."
4413.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
4414.Pp
4415In the following example per-interface firewall is created:
4416.Pp
4417.Dl "ipfw table IN create type iface valtype skipto,fib"
4418.Dl "ipfw table IN add vlan20 12000,12"
4419.Dl "ipfw table IN add vlan30 13000,13"
4420.Dl "ipfw table OUT create type iface valtype skipto"
4421.Dl "ipfw table OUT add vlan20 22000"
4422.Dl "ipfw table OUT add vlan30 23000"
4423.Dl ".."
4424.Dl "ipfw add 100 setfib tablearg ip from any to any recv 'table(IN)' in"
4425.Dl "ipfw add 200 skipto tablearg ip from any to any recv 'table(IN)' in"
4426.Dl "ipfw add 300 skipto tablearg ip from any to any xmit 'table(OUT)' out"
4427.Pp
4428The following example illustrate usage of flow tables:
4429.Pp
4430.Dl "ipfw table fl create type flow:src-ip,proto,dst-ip,dst-port"
4431.Dl "ipfw table fl add 2a02:6b8:77::88,tcp,2a02:6b8:77::99,80 11"
4432.Dl "ipfw table fl add 10.0.0.1,udp,10.0.0.2,53 12"
4433.Dl ".."
4434.Dl "ipfw add 100 allow ip from any to any flow 'table(fl,11)' recv ix0"
4435.Ss SETS OF RULES
4436To add a set of rules atomically, e.g.\& set 18:
4437.Pp
4438.Dl "ipfw set disable 18"
4439.Dl "ipfw add NN set 18 ...         # repeat as needed"
4440.Dl "ipfw set enable 18"
4441.Pp
4442To delete a set of rules atomically the command is simply:
4443.Pp
4444.Dl "ipfw delete set 18"
4445.Pp
4446To test a ruleset and disable it and regain control if something goes wrong:
4447.Pp
4448.Dl "ipfw set disable 18"
4449.Dl "ipfw add NN set 18 ...         # repeat as needed"
4450.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
4451.Pp
4452Here if everything goes well, you press control-C before the "sleep"
4453terminates, and your ruleset will be left active.
4454Otherwise, e.g.\& if
4455you cannot access your box, the ruleset will be disabled after
4456the sleep terminates thus restoring the previous situation.
4457.Pp
4458To show rules of the specific set:
4459.Pp
4460.Dl "ipfw set 18 show"
4461.Pp
4462To show rules of the disabled set:
4463.Pp
4464.Dl "ipfw -S set 18 show"
4465.Pp
4466To clear a specific rule counters of the specific set:
4467.Pp
4468.Dl "ipfw set 18 zero NN"
4469.Pp
4470To delete a specific rule of the specific set:
4471.Pp
4472.Dl "ipfw set 18 delete NN"
4473.Ss NAT, REDIRECT AND LSNAT
4474First redirect all the traffic to nat instance 123:
4475.Pp
4476.Dl "ipfw add nat 123 all from any to any"
4477.Pp
4478Then to configure nat instance 123 to alias all the outgoing traffic with ip
4479192.168.0.123, blocking all incoming connections, trying to keep
4480same ports on both sides, clearing aliasing table on address change
4481and keeping a log of traffic/link statistics:
4482.Pp
4483.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
4484.Pp
4485Or to change address of instance 123, aliasing table will be cleared (see
4486reset option):
4487.Pp
4488.Dl "ipfw nat 123 config ip 10.0.0.1"
4489.Pp
4490To see configuration of nat instance 123:
4491.Pp
4492.Dl "ipfw nat 123 show config"
4493.Pp
4494To show logs of all the instances in range 111-999:
4495.Pp
4496.Dl "ipfw nat 111-999 show"
4497.Pp
4498To see configurations of all instances:
4499.Pp
4500.Dl "ipfw nat show config"
4501.Pp
4502Or a redirect rule with mixed modes could looks like:
4503.Pp
4504.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
4505.Dl "			 redirect_port tcp 192.168.0.1:80 500"
4506.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
4507.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
4508.Dl "			 	    10.0.0.100	# LSNAT"
4509.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
4510.Dl "			 	    500		# LSNAT"
4511.Pp
4512or it could be split in:
4513.Pp
4514.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
4515.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
4516.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
4517.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
4518.Dl "				         10.0.0.100"
4519.Dl "ipfw nat 5 config redirect_port tcp"
4520.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
4521.Pp
4522Sometimes you may want to mix NAT and dynamic rules. It could be achieved with
4523.Cm record-state
4524and
4525.Cm defer-action
4526options. Problem is, you need to create dynamic rule before NAT and check it
4527after NAT actions (or vice versa) to have consistent addresses and ports.
4528Rule with
4529.Cm keep-state
4530option will trigger activation of existing dynamic state, and action of such
4531rule will be performed as soon as rule is matched. In case of NAT and
4532.Cm allow
4533rule packet need to be passed to NAT, not allowed as soon is possible.
4534.Pp
4535There is example of set of rules to achieve this. Bear in mind that this
4536is exmaple only and it is not very useful by itself.
4537.Pp
4538On way out, after all checks place this rules:
4539.Pp
4540.Dl "ipfw add allow record-state skip-action"
4541.Dl "ipfw add nat 1"
4542.Pp
4543And on way in there should be something like this:
4544.Pp
4545.Dl "ipfw add nat 1"
4546.Dl "ipfw add check-state"
4547.Pp
4548Please note, that first rule on way out doesn't allow packet and doesn't
4549execute existing dynamic rules. All it does, create new dynamic rule with
4550.Cm allow
4551action, if it is not created yet. Later, this dynamic rule is used on way
4552in by
4553.Cm check-state
4554rule.
4555.Ss CONFIGURING CODEL, PIE, FQ-CODEL and FQ-PIE AQM
4556.Cm codel
4557and
4558.Cm pie
4559AQM can be configured for
4560.Nm dummynet
4561.Cm pipe
4562or
4563.Cm queue.
4564.Pp
4565To configure a
4566.Cm pipe
4567with
4568.Cm codel
4569AQM using default configuration for traffic from 192.168.0.0/24 and 1Mbits/s
4570rate limit, we do:
4571.Pp
4572.Dl "ipfw pipe 1 config bw 1mbits/s codel"
4573.Dl "ipfw add 100 pipe 1 ip from 192.168.0.0/24 to any"
4574.Pp
4575To configure a
4576.Cm queue
4577with
4578.Cm codel
4579AQM using different configurations parameters for traffic from
4580192.168.0.0/24 and 1Mbits/s rate limit, we do:
4581.Pp
4582.Dl "ipfw pipe 1 config bw 1mbits/s"
4583.Dl "ipfw queue 1 config pipe 1 codel target 8ms interval 160ms ecn"
4584.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4585.Pp
4586To configure a
4587.Cm pipe
4588with
4589.Cm pie
4590AQM using default configuration for traffic from 192.168.0.0/24 and 1Mbits/s
4591rate limit, we do:
4592.Pp
4593.Dl "ipfw pipe 1 config bw 1mbits/s pie"
4594.Dl "ipfw add 100 pipe 1 ip from 192.168.0.0/24 to any"
4595.Pp
4596To configure a
4597.Cm queue
4598with
4599.Cm pie
4600AQM using different configuration parameters for traffic from
4601192.168.0.0/24 and 1Mbits/s rate limit, we do:
4602.Pp
4603.Dl "ipfw pipe 1 config bw 1mbits/s"
4604.Dl "ipfw queue 1 config pipe 1 pie target 20ms tupdate 30ms ecn"
4605.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4606.Pp
4607.Cm fq_codel
4608and
4609.Cm fq_pie
4610AQM can be configured for
4611.Nm dummynet
4612schedulers.
4613.Pp
4614To configure
4615.Cm fq_codel
4616scheduler using different configurations parameters for traffic from
4617192.168.0.0/24 and 1Mbits/s rate limit, we do:
4618.Pp
4619.Dl "ipfw pipe 1 config bw 1mbits/s"
4620.Dl "ipfw sched 1 config pipe 1 type fq_codel"
4621.Dl "ipfw queue 1 config sched 1"
4622.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4623.Pp
4624To change
4625.Cm fq_codel
4626default configuration for a
4627.Cm sched
4628such as disable ECN and change the
4629.Ar target
4630to 10ms, we do:
4631.Pp
4632.Dl "ipfw sched 1 config pipe 1 type fq_codel target 10ms noecn"
4633.Pp
4634Similar to
4635.Cm fq_codel,
4636to configure
4637.Cm fq_pie
4638scheduler using different configurations parameters for traffic from
4639192.168.0.0/24 and 1Mbits/s rate limit, we do:
4640.Pp
4641.Dl "ipfw pipe 1 config bw 1mbits/s"
4642.Dl "ipfw sched 1 config pipe 1 type fq_pie"
4643.Dl "ipfw queue 1 config sched 1"
4644.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4645.Pp
4646The configurations of
4647.Cm fq_pie
4648.Cm sched
4649can be changed in a similar way as for
4650.Cm fq_codel
4651.Sh SEE ALSO
4652.Xr cpp 1 ,
4653.Xr m4 1 ,
4654.Xr altq 4 ,
4655.Xr divert 4 ,
4656.Xr dummynet 4 ,
4657.Xr if_bridge 4 ,
4658.Xr ip 4 ,
4659.Xr ipfirewall 4 ,
4660.Xr ng_ether 4 ,
4661.Xr ng_ipfw 4 ,
4662.Xr protocols 5 ,
4663.Xr services 5 ,
4664.Xr init 8 ,
4665.Xr kldload 8 ,
4666.Xr reboot 8 ,
4667.Xr sysctl 8 ,
4668.Xr sysrc 8 ,
4669.Xr syslogd 8
4670.Sh HISTORY
4671The
4672.Nm
4673utility first appeared in
4674.Fx 2.0 .
4675.Nm dummynet
4676was introduced in
4677.Fx 2.2.8 .
4678Stateful extensions were introduced in
4679.Fx 4.0 .
4680.Nm ipfw2
4681was introduced in Summer 2002.
4682.Sh AUTHORS
4683.An Ugen J. S. Antsilevich ,
4684.An Poul-Henning Kamp ,
4685.An Alex Nash ,
4686.An Archie Cobbs ,
4687.An Luigi Rizzo ,
4688.An Rasool Al-Saadi .
4689.Pp
4690.An -nosplit
4691API based upon code written by
4692.An Daniel Boulet
4693for BSDI.
4694.Pp
4695Dummynet has been introduced by Luigi Rizzo in 1997-1998.
4696.Pp
4697Some early work (1999-2000) on the
4698.Nm dummynet
4699traffic shaper supported by Akamba Corp.
4700.Pp
4701The ipfw core (ipfw2) has been completely redesigned and
4702reimplemented by Luigi Rizzo in summer 2002.
4703Further
4704actions and
4705options have been added by various developers over the years.
4706.Pp
4707.An -nosplit
4708In-kernel NAT support written by
4709.An Paolo Pisati Aq Mt piso@FreeBSD.org
4710as part of a Summer of Code 2005 project.
4711.Pp
4712SCTP
4713.Nm nat
4714support has been developed by
4715.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
4716The primary developers and maintainers are David Hayes and Jason But.
4717For further information visit:
4718.Aq http://www.caia.swin.edu.au/urp/SONATA
4719.Pp
4720Delay profiles have been developed by Alessandro Cerri and
4721Luigi Rizzo, supported by the
4722European Commission within Projects Onelab and Onelab2.
4723.Pp
4724CoDel, PIE, FQ-CoDel and FQ-PIE AQM for Dummynet have been implemented by
4725.An The Centre for Advanced Internet Architectures (CAIA)
4726in 2016, supported by The Comcast Innovation Fund.
4727The primary developer is
4728Rasool Al-Saadi.
4729.Sh BUGS
4730The syntax has grown over the years and sometimes it might be confusing.
4731Unfortunately, backward compatibility prevents cleaning up mistakes
4732made in the definition of the syntax.
4733.Pp
4734.Em !!! WARNING !!!
4735.Pp
4736Misconfiguring the firewall can put your computer in an unusable state,
4737possibly shutting down network services and requiring console access to
4738regain control of it.
4739.Pp
4740Incoming packet fragments diverted by
4741.Cm divert
4742are reassembled before delivery to the socket.
4743The action used on those packet is the one from the
4744rule which matches the first fragment of the packet.
4745.Pp
4746Packets diverted to userland, and then reinserted by a userland process
4747may lose various packet attributes.
4748The packet source interface name
4749will be preserved if it is shorter than 8 bytes and the userland process
4750saves and reuses the sockaddr_in
4751(as does
4752.Xr natd 8 ) ;
4753otherwise, it may be lost.
4754If a packet is reinserted in this manner, later rules may be incorrectly
4755applied, making the order of
4756.Cm divert
4757rules in the rule sequence very important.
4758.Pp
4759Dummynet drops all packets with IPv6 link-local addresses.
4760.Pp
4761Rules using
4762.Cm uid
4763or
4764.Cm gid
4765may not behave as expected.
4766In particular, incoming SYN packets may
4767have no uid or gid associated with them since they do not yet belong
4768to a TCP connection, and the uid/gid associated with a packet may not
4769be as expected if the associated process calls
4770.Xr setuid 2
4771or similar system calls.
4772.Pp
4773Rule syntax is subject to the command line environment and some patterns
4774may need to be escaped with the backslash character
4775or quoted appropriately.
4776.Pp
4777Due to the architecture of
4778.Xr libalias 3 ,
4779ipfw nat is not compatible with the TCP segmentation offloading (TSO).
4780Thus, to reliably nat your network traffic, please disable TSO
4781on your NICs using
4782.Xr ifconfig 8 .
4783.Pp
4784ICMP error messages are not implicitly matched by dynamic rules
4785for the respective conversations.
4786To avoid failures of network error detection and path MTU discovery,
4787ICMP error messages may need to be allowed explicitly through static
4788rules.
4789.Pp
4790Rules using
4791.Cm call
4792and
4793.Cm return
4794actions may lead to confusing behaviour if ruleset has mistakes,
4795and/or interaction with other subsystems (netgraph, dummynet, etc.) is used.
4796One possible case for this is packet leaving
4797.Nm
4798in subroutine on the input pass, while later on output encountering unpaired
4799.Cm return
4800first.
4801As the call stack is kept intact after input pass, packet will suddenly
4802return to the rule number used on input pass, not on output one.
4803Order of processing should be checked carefully to avoid such mistakes.
4804