xref: /freebsd/sbin/ipfw/ipfw.8 (revision f126890ac5386406dadf7c4cfa9566cbb56537c5)
1.\"
2.Dd December 17, 2023
3.Dt IPFW 8
4.Os
5.Sh NAME
6.Nm ipfw , dnctl
7.Nd User interface for firewall, traffic shaper, packet scheduler,
8in-kernel NAT.
9.Sh SYNOPSIS
10.Ss FIREWALL CONFIGURATION
11.Nm
12.Op Fl cq
13.Cm add
14.Ar rule
15.Nm
16.Op Fl acdefnNStT
17.Op Cm set Ar N
18.Brq Cm list | show
19.Op Ar rule | first-last ...
20.Nm
21.Op Fl f | q
22.Op Cm set Ar N
23.Cm flush
24.Nm
25.Op Fl q
26.Op Cm set Ar N
27.Brq Cm delete | zero | resetlog
28.Op Ar number ...
29.Pp
30.Nm
31.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
32.Nm
33.Cm set move
34.Op Cm rule
35.Ar number Cm to Ar number
36.Nm
37.Cm set swap Ar number number
38.Nm
39.Cm set show
40.Ss SYSCTL SHORTCUTS
41.Nm
42.Cm enable
43.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
44.Nm
45.Cm disable
46.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
47.Ss LOOKUP TABLES
48.Nm
49.Oo Cm set Ar N Oc Cm table Ar name Cm create Ar create-options
50.Nm
51.Oo Cm set Ar N Oc Cm table
52.Brq Ar name | all
53.Cm destroy
54.Nm
55.Oo Cm set Ar N Oc Cm table Ar name Cm modify Ar modify-options
56.Nm
57.Oo Cm set Ar N Oc Cm table Ar name Cm swap Ar name
58.Nm
59.Oo Cm set Ar N Oc Cm table Ar name Cm add Ar table-key Op Ar value
60.Nm
61.Oo Cm set Ar N Oc Cm table Ar name Cm add Op Ar table-key Ar value ...
62.Nm
63.Oo Cm set Ar N Oc Cm table Ar name Cm atomic add Op Ar table-key Ar value ...
64.Nm
65.Oo Cm set Ar N Oc Cm table Ar name Cm delete Op Ar table-key ...
66.Nm
67.Oo Cm set Ar N Oc Cm table Ar name Cm lookup Ar addr
68.Nm
69.Oo Cm set Ar N Oc Cm table Ar name Cm lock
70.Nm
71.Oo Cm set Ar N Oc Cm table Ar name Cm unlock
72.Nm
73.Oo Cm set Ar N Oc Cm table
74.Brq Ar name | all
75.Cm list
76.Nm
77.Oo Cm set Ar N Oc Cm table
78.Brq Ar name | all
79.Cm info
80.Nm
81.Oo Cm set Ar N Oc Cm table
82.Brq Ar name | all
83.Cm detail
84.Nm
85.Oo Cm set Ar N Oc Cm table
86.Brq Ar name | all
87.Cm flush
88.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
89.Nm dnctl
90.Brq Cm pipe | queue | sched
91.Ar number
92.Cm config
93.Ar config-options
94.Nm dnctl
95.Op Fl s Op Ar field
96.Brq Cm pipe | queue | sched
97.Brq Cm delete | list | show
98.Op Ar number ...
99.Ss IN-KERNEL NAT
100.Nm
101.Op Fl q
102.Cm nat
103.Ar number
104.Cm config
105.Ar config-options
106.Nm
107.Cm nat
108.Ar number
109.Cm show
110.Brq Cm config | log
111.Ss STATEFUL IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
112.Nm
113.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm create Ar create-options
114.Nm
115.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm config Ar config-options
116.Nm
117.Oo Cm set Ar N Oc Cm nat64lsn
118.Brq Ar name | all
119.Brq Cm list | show
120.Op Cm states
121.Nm
122.Oo Cm set Ar N Oc Cm nat64lsn
123.Brq Ar name | all
124.Cm destroy
125.Nm
126.Oo Cm set Ar N Oc Cm nat64lsn Ar name Cm stats Op Cm reset
127.Ss STATELESS IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
128.Nm
129.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm create Ar create-options
130.Nm
131.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm config Ar config-options
132.Nm
133.Oo Cm set Ar N Oc Cm nat64stl
134.Brq Ar name | all
135.Brq Cm list | show
136.Nm
137.Oo Cm set Ar N Oc Cm nat64stl
138.Brq Ar name | all
139.Cm destroy
140.Nm
141.Oo Cm set Ar N Oc Cm nat64stl Ar name Cm stats Op Cm reset
142.Ss XLAT464 CLAT IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
143.Nm
144.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm create Ar create-options
145.Nm
146.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm config Ar config-options
147.Nm
148.Oo Cm set Ar N Oc Cm nat64clat
149.Brq Ar name | all
150.Brq Cm list | show
151.Nm
152.Oo Cm set Ar N Oc Cm nat64clat
153.Brq Ar name | all
154.Cm destroy
155.Nm
156.Oo Cm set Ar N Oc Cm nat64clat Ar name Cm stats Op Cm reset
157.Ss IPv6-to-IPv6 NETWORK PREFIX TRANSLATION
158.Nm
159.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm create Ar create-options
160.Nm
161.Oo Cm set Ar N Oc Cm nptv6
162.Brq Ar name | all
163.Brq Cm list | show
164.Nm
165.Oo Cm set Ar N Oc Cm nptv6
166.Brq Ar name | all
167.Cm destroy
168.Nm
169.Oo Cm set Ar N Oc Cm nptv6 Ar name Cm stats Op Cm reset
170.Ss INTERNAL DIAGNOSTICS
171.Nm
172.Cm internal iflist
173.Nm
174.Cm internal talist
175.Nm
176.Cm internal vlist
177.Ss LIST OF RULES AND PREPROCESSING
178.Nm
179.Op Fl cfnNqS
180.Oo
181.Fl p Ar preproc
182.Oo
183.Ar preproc-flags
184.Oc
185.Oc
186.Ar pathname
187.Sh DESCRIPTION
188The
189.Nm
190utility is the user interface for controlling the
191.Xr ipfw 4
192firewall, the
193.Xr dummynet 4
194traffic shaper/packet scheduler, and the
195in-kernel NAT services.
196.Pp
197A firewall configuration, or
198.Em ruleset ,
199is made of a list of
200.Em rules
201numbered from 1 to 65535.
202Packets are passed to the firewall
203from a number of different places in the protocol stack
204(depending on the source and destination of the packet,
205it is possible for the firewall to be
206invoked multiple times on the same packet).
207The packet passed to the firewall is compared
208against each of the rules in the
209.Em ruleset ,
210in rule-number order
211(multiple rules with the same number are permitted, in which case
212they are processed in order of insertion).
213When a match is found, the action corresponding to the
214matching rule is performed.
215.Pp
216Depending on the action and certain system settings, packets
217can be reinjected into the firewall at some rule after the
218matching one for further processing.
219.Pp
220A ruleset always includes a
221.Em default
222rule (numbered 65535) which cannot be modified or deleted,
223and matches all packets.
224The action associated with the
225.Em default
226rule can be either
227.Cm deny
228or
229.Cm allow
230depending on how the kernel is configured.
231.Pp
232If the ruleset includes one or more rules with the
233.Cm keep-state ,
234.Cm record-state ,
235.Cm limit
236or
237.Cm set-limit
238option,
239the firewall will have a
240.Em stateful
241behaviour, i.e., upon a match it will create
242.Em dynamic rules ,
243i.e., rules that match packets with the same 5-tuple
244(protocol, source and destination addresses and ports)
245as the packet which caused their creation.
246Dynamic rules, which have a limited lifetime, are checked
247at the first occurrence of a
248.Cm check-state ,
249.Cm keep-state
250or
251.Cm limit
252rule, and are typically used to open the firewall on-demand to
253legitimate traffic only.
254Please note, that
255.Cm keep-state
256and
257.Cm limit
258imply implicit
259.Cm check-state
260for all packets (not only these matched by the rule) but
261.Cm record-state
262and
263.Cm set-limit
264have no implicit
265.Cm check-state .
266See the
267.Sx STATEFUL FIREWALL
268and
269.Sx EXAMPLES
270Sections below for more information on the stateful behaviour of
271.Nm .
272.Pp
273All rules (including dynamic ones) have a few associated counters:
274a packet count, a byte count, a log count and a timestamp
275indicating the time of the last match.
276Counters can be displayed or reset with
277.Nm
278commands.
279.Pp
280Each rule belongs to one of 32 different
281.Em sets
282, and there are
283.Nm
284commands to atomically manipulate sets, such as enable,
285disable, swap sets, move all rules in a set to another
286one, delete all rules in a set.
287These can be useful to
288install temporary configurations, or to test them.
289See Section
290.Sx SETS OF RULES
291for more information on
292.Em sets .
293.Pp
294Rules can be added with the
295.Cm add
296command; deleted individually or in groups with the
297.Cm delete
298command, and globally (except those in set 31) with the
299.Cm flush
300command; displayed, optionally with the content of the
301counters, using the
302.Cm show
303and
304.Cm list
305commands.
306Finally, counters can be reset with the
307.Cm zero
308and
309.Cm resetlog
310commands.
311.Ss COMMAND OPTIONS
312The following general options are available when invoking
313.Nm :
314.Bl -tag -width indent
315.It Fl a
316Show counter values when listing rules.
317The
318.Cm show
319command implies this option.
320.It Fl b
321Only show the action and the comment, not the body of a rule.
322Implies
323.Fl c .
324.It Fl c
325When entering or showing rules, print them in compact form,
326i.e., omitting the "ip from any to any" string
327when this does not carry any additional information.
328.It Fl d
329When listing, show dynamic rules in addition to static ones.
330.It Fl D
331When listing, show only dynamic states.
332When deleting, delete only dynamic states.
333.It Fl f
334Run without prompting for confirmation for commands that can cause problems if misused,
335i.e.,
336.Cm flush .
337If there is no tty associated with the process, this is implied.
338The
339.Cm delete
340command with this flag ignores possible errors,
341i.e., nonexistent rule number.
342And for batched commands execution continues with the next command.
343.It Fl i
344When listing a table (see the
345.Sx LOOKUP TABLES
346section below for more information on lookup tables), format values
347as IP addresses.
348By default, values are shown as integers.
349.It Fl n
350Only check syntax of the command strings, without actually passing
351them to the kernel.
352.It Fl N
353Try to resolve addresses and service names in output.
354.It Fl q
355Be quiet when executing the
356.Cm add ,
357.Cm nat ,
358.Cm zero ,
359.Cm resetlog
360or
361.Cm flush
362commands;
363(implies
364.Fl f ) .
365This is useful when updating rulesets by executing multiple
366.Nm
367commands in a script
368(e.g.,
369.Ql sh\ /etc/rc.firewall ) ,
370or by processing a file with many
371.Nm
372rules across a remote login session.
373It also stops a table add or delete
374from failing if the entry already exists or is not present.
375.Pp
376The reason why this option may be important is that
377for some of these actions,
378.Nm
379may print a message; if the action results in blocking the
380traffic to the remote client,
381the remote login session will be closed
382and the rest of the ruleset will not be processed.
383Access to the console would then be required to recover.
384.It Fl S
385When listing rules, show the
386.Em set
387each rule belongs to.
388If this flag is not specified, disabled rules will not be
389listed.
390.It Fl s Op Ar field
391When listing pipes, sort according to one of the four
392counters (total or current packets or bytes).
393.It Fl t
394When listing, show last match timestamp converted with
395.Fn ctime .
396.It Fl T
397When listing, show last match timestamp as seconds from the epoch.
398This form can be more convenient for postprocessing by scripts.
399.El
400.Ss LIST OF RULES AND PREPROCESSING
401To ease configuration, rules can be put into a file which is
402processed using
403.Nm
404as shown in the last synopsis line.
405An absolute
406.Ar pathname
407must be used.
408The file will be read line by line and applied as arguments to the
409.Nm
410utility.
411.Pp
412Optionally, a preprocessor can be specified using
413.Fl p Ar preproc
414where
415.Ar pathname
416is to be piped through.
417Useful preprocessors include
418.Xr cpp 1
419and
420.Xr m4 1 .
421If
422.Ar preproc
423does not start with a slash
424.Pq Ql /
425as its first character, the usual
426.Ev PATH
427name search is performed.
428Care should be taken with this in environments where not all
429file systems are mounted (yet) by the time
430.Nm
431is being run (e.g.\& when they are mounted over NFS).
432Once
433.Fl p
434has been specified, any additional arguments are passed on to the preprocessor
435for interpretation.
436This allows for flexible configuration files (like conditionalizing
437them on the local hostname) and the use of macros to centralize
438frequently required arguments like IP addresses.
439.Ss TRAFFIC SHAPER CONFIGURATION
440The
441.Nm dnctl
442.Cm pipe , queue
443and
444.Cm sched
445commands are used to configure the traffic shaper and packet scheduler.
446See the
447.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
448Section below for details.
449.Pp
450If the world and the kernel get out of sync the
451.Nm
452ABI may break, preventing you from being able to add any rules.
453This can adversely affect the booting process.
454You can use
455.Nm
456.Cm disable
457.Cm firewall
458to temporarily disable the firewall to regain access to the network,
459allowing you to fix the problem.
460.Sh PACKET FLOW
461A packet is checked against the active ruleset in multiple places
462in the protocol stack, under control of several sysctl variables.
463These places and variables are shown below, and it is important to
464have this picture in mind in order to design a correct ruleset.
465.Bd -literal -offset indent
466       ^    to upper layers    V
467       |                       |
468       +----------->-----------+
469       ^                       V
470 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
471       |                       |
472       ^                       V
473 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
474       |                       |
475       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
476       ^                       V
477       |      to devices       |
478.Ed
479.Pp
480The number of
481times the same packet goes through the firewall can
482vary between 0 and 4 depending on packet source and
483destination, and system configuration.
484.Pp
485Note that as packets flow through the stack, headers can be
486stripped or added to it, and so they may or may not be available
487for inspection.
488E.g., incoming packets will include the MAC header when
489.Nm
490is invoked from
491.Cm ether_demux() ,
492but the same packets will have the MAC header stripped off when
493.Nm
494is invoked from
495.Cm ip_input()
496or
497.Cm ip6_input() .
498.Pp
499Also note that each packet is always checked against the complete ruleset,
500irrespective of the place where the check occurs, or the source of the packet.
501If a rule contains some match patterns or actions which are not valid
502for the place of invocation (e.g.\& trying to match a MAC header within
503.Cm ip_input
504or
505.Cm ip6_input ),
506the match pattern will not match, but a
507.Cm not
508operator in front of such patterns
509.Em will
510cause the pattern to
511.Em always
512match on those packets.
513It is thus the responsibility of
514the programmer, if necessary, to write a suitable ruleset to
515differentiate among the possible places.
516.Cm skipto
517rules can be useful here, as an example:
518.Bd -literal -offset indent
519# packets from ether_demux or bdg_forward
520ipfw add 10 skipto 1000 all from any to any layer2 in
521# packets from ip_input
522ipfw add 10 skipto 2000 all from any to any not layer2 in
523# packets from ip_output
524ipfw add 10 skipto 3000 all from any to any not layer2 out
525# packets from ether_output_frame
526ipfw add 10 skipto 4000 all from any to any layer2 out
527.Ed
528.Pp
529(yes, at the moment there is no way to differentiate between
530ether_demux and bdg_forward).
531.Pp
532Also note that only actions
533.Cm allow ,
534.Cm deny ,
535.Cm netgraph ,
536.Cm ngtee
537and related to
538.Cm dummynet
539are processed for
540.Cm layer2
541frames and all other actions act as if they were
542.Cm allow
543for such frames.
544Full set of actions is supported for IP packets without
545.Cm layer2
546headers only.
547For example,
548.Cm divert
549action does not divert
550.Cm layer2
551frames.
552.Sh SYNTAX
553In general, each keyword or argument must be provided as
554a separate command line argument, with no leading or trailing
555spaces.
556Keywords are case-sensitive, whereas arguments may
557or may not be case-sensitive depending on their nature
558(e.g.\& uid's are, hostnames are not).
559.Pp
560Some arguments (e.g., port or address lists) are comma-separated
561lists of values.
562In this case, spaces after commas ',' are allowed to make
563the line more readable.
564You can also put the entire
565command (including flags) into a single argument.
566E.g., the following forms are equivalent:
567.Bd -literal -offset indent
568ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
569ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
570ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
571.Ed
572.Sh RULE FORMAT
573The format of firewall rules is the following:
574.Bd -ragged -offset indent
575.Bk -words
576.Op Ar rule_number
577.Op Cm set Ar set_number
578.Op Cm prob Ar match_probability
579.Ar action
580.Op Cm log Op Cm logamount Ar number
581.Op Cm altq Ar queue
582.Oo
583.Bro Cm tag | untag
584.Brc Ar number
585.Oc
586.Ar body
587.Ek
588.Ed
589.Pp
590where the body of the rule specifies which information is used
591for filtering packets, among the following:
592.Pp
593.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
594.It Layer2 header fields
595When available
596.It IPv4 and IPv6 Protocol
597SCTP, TCP, UDP, ICMP, etc.
598.It Source and dest. addresses and ports
599.It Direction
600See Section
601.Sx PACKET FLOW
602.It Transmit and receive interface
603By name or address
604.It Misc. IP header fields
605Version, type of service, datagram length, identification,
606fragmentation flags,
607Time To Live
608.It IP options
609.It IPv6 Extension headers
610Fragmentation, Hop-by-Hop options,
611Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
612.It IPv6 Flow-ID
613.It Misc. TCP header fields
614TCP flags (SYN, FIN, ACK, RST, etc.),
615sequence number, acknowledgment number,
616window
617.It TCP options
618.It ICMP types
619for ICMP packets
620.It ICMP6 types
621for ICMP6 packets
622.It User/group ID
623When the packet can be associated with a local socket.
624.It Divert status
625Whether a packet came from a divert socket (e.g.,
626.Xr natd 8 ) .
627.It Fib annotation state
628Whether a packet has been tagged for using a specific FIB (routing table)
629in future forwarding decisions.
630.El
631.Pp
632Note that some of the above information, e.g.\& source MAC or IP addresses and
633TCP/UDP ports, can be easily spoofed, so filtering on those fields
634alone might not guarantee the desired results.
635.Bl -tag -width indent
636.It Ar rule_number
637Each rule is associated with a
638.Ar rule_number
639in the range 1..65535, with the latter reserved for the
640.Em default
641rule.
642Rules are checked sequentially by rule number.
643Multiple rules can have the same number, in which case they are
644checked (and listed) according to the order in which they have
645been added.
646If a rule is entered without specifying a number, the kernel will
647assign one in such a way that the rule becomes the last one
648before the
649.Em default
650rule.
651Automatic rule numbers are assigned by incrementing the last
652non-default rule number by the value of the sysctl variable
653.Ar net.inet.ip.fw.autoinc_step
654which defaults to 100.
655If this is not possible (e.g.\& because we would go beyond the
656maximum allowed rule number), the number of the last
657non-default value is used instead.
658.It Cm set Ar set_number
659Each rule is associated with a
660.Ar set_number
661in the range 0..31.
662Sets can be individually disabled and enabled, so this parameter
663is of fundamental importance for atomic ruleset manipulation.
664It can be also used to simplify deletion of groups of rules.
665If a rule is entered without specifying a set number,
666set 0 will be used.
667.br
668Set 31 is special in that it cannot be disabled,
669and rules in set 31 are not deleted by the
670.Nm ipfw flush
671command (but you can delete them with the
672.Nm ipfw delete set 31
673command).
674Set 31 is also used for the
675.Em default
676rule.
677.It Cm prob Ar match_probability
678A match is only declared with the specified probability
679(floating point number between 0 and 1).
680This can be useful for a number of applications such as
681random packet drop or
682(in conjunction with
683.Nm dummynet )
684to simulate the effect of multiple paths leading to out-of-order
685packet delivery.
686.Pp
687Note: this condition is checked before any other condition, including
688ones such as
689.Cm keep-state
690or
691.Cm check-state
692which might have
693side effects.
694.It Cm log Op Cm logamount Ar number
695Packets matching a rule with the
696.Cm log
697keyword will be made available for logging in two ways:
698if the sysctl variable
699.Va net.inet.ip.fw.verbose
700is set to 0 (default), one can use
701.Xr bpf 4
702attached to the
703.Li ipfw0
704pseudo interface.
705This pseudo interface can be created manually after a system
706boot by using the following command:
707.Bd -literal -offset indent
708# ifconfig ipfw0 create
709.Ed
710.Pp
711Or, automatically at boot time by adding the following
712line to the
713.Xr rc.conf 5
714file:
715.Bd -literal -offset indent
716firewall_logif="YES"
717.Ed
718.Pp
719There is zero overhead when no
720.Xr bpf 4
721is attached to the pseudo interface.
722.Pp
723If
724.Va net.inet.ip.fw.verbose
725is set to 1, packets will be logged to
726.Xr syslogd 8
727with a
728.Dv LOG_SECURITY
729facility up to a maximum of
730.Cm logamount
731packets.
732If no
733.Cm logamount
734is specified, the limit is taken from the sysctl variable
735.Va net.inet.ip.fw.verbose_limit .
736In both cases, a value of 0 means unlimited logging.
737.Pp
738Once the limit is reached, logging can be re-enabled by
739clearing the logging counter or the packet counter for that entry, see the
740.Cm resetlog
741command.
742.Pp
743Note: logging is done after all other packet matching conditions
744have been successfully verified, and before performing the final
745action (accept, deny, etc.) on the packet.
746.It Cm tag Ar number
747When a packet matches a rule with the
748.Cm tag
749keyword, the numeric tag for the given
750.Ar number
751in the range 1..65534 will be attached to the packet.
752The tag acts as an internal marker (it is not sent out over
753the wire) that can be used to identify these packets later on.
754This can be used, for example, to provide trust between interfaces
755and to start doing policy-based filtering.
756A packet can have multiple tags at the same time.
757Tags are "sticky", meaning once a tag is applied to a packet by a
758matching rule it exists until explicit removal.
759Tags are kept with the packet everywhere within the kernel, but are
760lost when the packet leaves the kernel, for example, on transmitting
761packet out to the network or sending packet to a
762.Xr divert 4
763socket.
764.Pp
765To check for previously applied tags, use the
766.Cm tagged
767rule option.
768To delete previously applied tag, use the
769.Cm untag
770keyword.
771.Pp
772Note: since tags are kept with the packet everywhere in kernelspace,
773they can be set and unset anywhere in the kernel network subsystem
774(using the
775.Xr mbuf_tags 9
776facility), not only by means of the
777.Xr ipfw 4
778.Cm tag
779and
780.Cm untag
781keywords.
782For example, there can be a specialized
783.Xr netgraph 4
784node doing traffic analyzing and tagging for later inspecting
785in firewall.
786.It Cm untag Ar number
787When a packet matches a rule with the
788.Cm untag
789keyword, the tag with the number
790.Ar number
791is searched among the tags attached to this packet and,
792if found, removed from it.
793Other tags bound to packet, if present, are left untouched.
794.It Cm setmark Ar value | tablearg
795When a packet matches a rule with the
796.Cm setmark
797keyword, a 32-bit numeric mark is assigned to the packet.
798The mark is an extension to the tags.
799As tags, mark is "sticky" so the value is kept the same within the kernel and
800is lost when the packet leaves the kernel.
801Unlike tags, mark can be matched as a lookup table key or compared with bitwise
802mask applied against another value.
803Each packet can have only one mark, so
804.Cm setmark
805always overwrites the previous mark value.
806.Pp
807The initial mark value is 0.
808To check the current mark value, use the
809.Cm mark
810rule option.
811Mark
812.Ar value
813can be entered as decimal or hexadecimal (if prefixed by 0x), and they
814are always printed as hexadecimal.
815.It Cm altq Ar queue
816When a packet matches a rule with the
817.Cm altq
818keyword, the ALTQ identifier for the given
819.Ar queue
820(see
821.Xr altq 4 )
822will be attached.
823Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
824and not being rejected or going to divert sockets.
825Note that if there is insufficient memory at the time the packet is
826processed, it will not be tagged, so it is wise to make your ALTQ
827"default" queue policy account for this.
828If multiple
829.Cm altq
830rules match a single packet, only the first one adds the ALTQ classification
831tag.
832In doing so, traffic may be shaped by using
833.Cm count Cm altq Ar queue
834rules for classification early in the ruleset, then later applying
835the filtering decision.
836For example,
837.Cm check-state
838and
839.Cm keep-state
840rules may come later and provide the actual filtering decisions in
841addition to the fallback ALTQ tag.
842.Pp
843You must run
844.Xr pfctl 8
845to set up the queues before IPFW will be able to look them up by name,
846and if the ALTQ disciplines are rearranged, the rules in containing the
847queue identifiers in the kernel will likely have gone stale and need
848to be reloaded.
849Stale queue identifiers will probably result in misclassification.
850.Pp
851All system ALTQ processing can be turned on or off via
852.Nm
853.Cm enable Ar altq
854and
855.Nm
856.Cm disable Ar altq .
857The usage of
858.Va net.inet.ip.fw.one_pass
859is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
860always after adding an ALTQ tag.
861.El
862.Ss RULE ACTIONS
863A rule can be associated with one of the following actions, which
864will be executed when the packet matches the body of the rule.
865.Bl -tag -width indent
866.It Cm allow | accept | pass | permit
867Allow packets that match rule.
868The search terminates.
869.It Cm check-state Op Ar :flowname | Cm :any
870Checks the packet against the dynamic ruleset.
871If a match is found, execute the action associated with
872the rule which generated this dynamic rule, otherwise
873move to the next rule.
874.br
875.Cm Check-state
876rules do not have a body.
877If no
878.Cm check-state
879rule is found, the dynamic ruleset is checked at the first
880.Cm keep-state
881or
882.Cm limit
883rule.
884The
885.Ar :flowname
886is symbolic name assigned to dynamic rule by
887.Cm keep-state
888opcode.
889The special flowname
890.Cm :any
891can be used to ignore states flowname when matching.
892The
893.Cm :default
894keyword is special name used for compatibility with old rulesets.
895.It Cm count
896Update counters for all packets that match rule.
897The search continues with the next rule.
898.It Cm deny | drop
899Discard packets that match this rule.
900The search terminates.
901.It Cm divert Ar port
902Divert packets that match this rule to the
903.Xr divert 4
904socket bound to port
905.Ar port .
906The search terminates.
907.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
908Change the next-hop on matching packets to
909.Ar ipaddr ,
910which can be an IP address or a host name.
911The next hop can also be supplied by the last table
912looked up for the packet by using the
913.Cm tablearg
914keyword instead of an explicit address.
915The search terminates if this rule matches.
916.Pp
917If
918.Ar ipaddr
919is a local address, then matching packets will be forwarded to
920.Ar port
921(or the port number in the packet if one is not specified in the rule)
922on the local machine.
923.br
924If
925.Ar ipaddr
926is not a local address, then the port number
927(if specified) is ignored, and the packet will be
928forwarded to the remote address, using the route as found in
929the local routing table for that IP.
930.br
931A
932.Ar fwd
933rule will not match layer2 packets (those received
934on ether_input, ether_output, or bridged).
935.br
936The
937.Cm fwd
938action does not change the contents of the packet at all.
939In particular, the destination address remains unmodified, so
940packets forwarded to another system will usually be rejected by that system
941unless there is a matching rule on that system to capture them.
942For packets forwarded locally,
943the local address of the socket will be
944set to the original destination address of the packet.
945This makes the
946.Xr netstat 1
947entry look rather weird but is intended for
948use with transparent proxy servers.
949.It Cm nat Ar nat_nr | global | tablearg
950Pass packet to a
951nat instance
952(for network address translation, address redirect, etc.):
953see the
954.Sx NETWORK ADDRESS TRANSLATION (NAT)
955Section for further information.
956.It Cm nat64lsn Ar name
957Pass packet to a stateful NAT64 instance (for IPv6/IPv4 network address and
958protocol translation): see the
959.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
960Section for further information.
961.It Cm nat64stl Ar name
962Pass packet to a stateless NAT64 instance (for IPv6/IPv4 network address and
963protocol translation): see the
964.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
965Section for further information.
966.It Cm nat64clat Ar name
967Pass packet to a CLAT NAT64 instance (for client-side IPv6/IPv4 network address and
968protocol translation): see the
969.Sx IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
970Section for further information.
971.It Cm nptv6 Ar name
972Pass packet to a NPTv6 instance (for IPv6-to-IPv6 network prefix translation):
973see the
974.Sx IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
975Section for further information.
976.It Cm pipe Ar pipe_nr
977Pass packet to a
978.Nm dummynet
979.Dq pipe
980(for bandwidth limitation, delay, etc.).
981See the
982.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
983Section for further information.
984The search terminates; however, on exit from the pipe and if
985the
986.Xr sysctl 8
987variable
988.Va net.inet.ip.fw.one_pass
989is not set, the packet is passed again to the firewall code
990starting from the next rule.
991.It Cm queue Ar queue_nr
992Pass packet to a
993.Nm dummynet
994.Dq queue
995(for bandwidth limitation using WF2Q+).
996.It Cm reject
997(Deprecated).
998Synonym for
999.Cm unreach host .
1000.It Cm reset
1001Discard packets that match this rule, and if the
1002packet is a TCP packet, try to send a TCP reset (RST) notice.
1003The search terminates.
1004.It Cm reset6
1005Discard packets that match this rule, and if the
1006packet is a TCP packet, try to send a TCP reset (RST) notice.
1007The search terminates.
1008.It Cm skipto Ar number | tablearg
1009Skip all subsequent rules numbered less than
1010.Ar number .
1011The search continues with the first rule numbered
1012.Ar number
1013or higher.
1014It is possible to use the
1015.Cm tablearg
1016keyword with a skipto for a
1017.Em computed
1018skipto.
1019Skipto may work either in O(log(N)) or in O(1) depending
1020on amount of memory and/or sysctl variables.
1021See the
1022.Sx SYSCTL VARIABLES
1023section for more details.
1024.It Cm call Ar number | tablearg
1025The current rule number is saved in the internal stack and
1026ruleset processing continues with the first rule numbered
1027.Ar number
1028or higher.
1029If later a rule with the
1030.Cm return
1031action is encountered, the processing returns to the first rule
1032with number of this
1033.Cm call
1034rule plus one or higher
1035(the same behaviour as with packets returning from
1036.Xr divert 4
1037socket after a
1038.Cm divert
1039action).
1040This could be used to make somewhat like an assembly language
1041.Dq subroutine
1042calls to rules with common checks for different interfaces, etc.
1043.Pp
1044Rule with any number could be called, not just forward jumps as with
1045.Cm skipto .
1046So, to prevent endless loops in case of mistakes, both
1047.Cm call
1048and
1049.Cm return
1050actions don't do any jumps and simply go to the next rule if memory
1051cannot be allocated or stack overflowed/underflowed.
1052.Pp
1053Internally stack for rule numbers is implemented using
1054.Xr mbuf_tags 9
1055facility and currently has size of 16 entries.
1056As mbuf tags are lost when packet leaves the kernel,
1057.Cm divert
1058should not be used in subroutines to avoid endless loops
1059and other undesired effects.
1060.It Cm return
1061Takes rule number saved to internal stack by the last
1062.Cm call
1063action and returns ruleset processing to the first rule
1064with number greater than number of corresponding
1065.Cm call
1066rule.
1067See description of the
1068.Cm call
1069action for more details.
1070.Pp
1071Note that
1072.Cm return
1073rules usually end a
1074.Dq subroutine
1075and thus are unconditional, but
1076.Nm
1077command-line utility currently requires every action except
1078.Cm check-state
1079to have body.
1080While it is sometimes useful to return only on some packets,
1081usually you want to print just
1082.Dq return
1083for readability.
1084A workaround for this is to use new syntax and
1085.Fl c
1086switch:
1087.Bd -literal -offset indent
1088# Add a rule without actual body
1089ipfw add 2999 return via any
1090
1091# List rules without "from any to any" part
1092ipfw -c list
1093.Ed
1094.Pp
1095This cosmetic annoyance may be fixed in future releases.
1096.It Cm tee Ar port
1097Send a copy of packets matching this rule to the
1098.Xr divert 4
1099socket bound to port
1100.Ar port .
1101The search continues with the next rule.
1102.It Cm unreach Ar code Op mtu
1103Discard packets that match this rule, and try to send an ICMP
1104unreachable notice with code
1105.Ar code ,
1106where
1107.Ar code
1108is a number from 0 to 255, or one of these aliases:
1109.Cm net , host , protocol , port ,
1110.Cm needfrag , srcfail , net-unknown , host-unknown ,
1111.Cm isolated , net-prohib , host-prohib , tosnet ,
1112.Cm toshost , filter-prohib , host-precedence
1113or
1114.Cm precedence-cutoff .
1115The
1116.Cm needfrag
1117code may have an optional
1118.Ar mtu
1119parameter.
1120If specified, the MTU value will be put into generated ICMP packet.
1121The search terminates.
1122.It Cm unreach6 Ar code
1123Discard packets that match this rule, and try to send an ICMPv6
1124unreachable notice with code
1125.Ar code ,
1126where
1127.Ar code
1128is a number from 0, 1, 3 or 4, or one of these aliases:
1129.Cm no-route, admin-prohib, address
1130or
1131.Cm port .
1132The search terminates.
1133.It Cm netgraph Ar cookie
1134Divert packet into netgraph with given
1135.Ar cookie .
1136The search terminates.
1137If packet is later returned from netgraph it is either
1138accepted or continues with the next rule, depending on
1139.Va net.inet.ip.fw.one_pass
1140sysctl variable.
1141.It Cm ngtee Ar cookie
1142A copy of packet is diverted into netgraph, original
1143packet continues with the next rule.
1144See
1145.Xr ng_ipfw 4
1146for more information on
1147.Cm netgraph
1148and
1149.Cm ngtee
1150actions.
1151.It Cm setfib Ar fibnum | tablearg
1152The packet is tagged so as to use the FIB (routing table)
1153.Ar fibnum
1154in any subsequent forwarding decisions.
1155In the current implementation, this is limited to the values 0 through 15, see
1156.Xr setfib 2 .
1157Processing continues at the next rule.
1158It is possible to use the
1159.Cm tablearg
1160keyword with setfib.
1161If the tablearg value is not within the compiled range of fibs,
1162the packet's fib is set to 0.
1163.It Cm setdscp Ar DSCP | number | tablearg
1164Set specified DiffServ codepoint for an IPv4/IPv6 packet.
1165Processing continues at the next rule.
1166Supported values are:
1167.Pp
1168.Cm cs0
1169.Pq Dv 000000 ,
1170.Cm cs1
1171.Pq Dv 001000 ,
1172.Cm cs2
1173.Pq Dv 010000 ,
1174.Cm cs3
1175.Pq Dv 011000 ,
1176.Cm cs4
1177.Pq Dv 100000 ,
1178.Cm cs5
1179.Pq Dv 101000 ,
1180.Cm cs6
1181.Pq Dv 110000 ,
1182.Cm cs7
1183.Pq Dv 111000 ,
1184.Cm af11
1185.Pq Dv 001010 ,
1186.Cm af12
1187.Pq Dv 001100 ,
1188.Cm af13
1189.Pq Dv 001110 ,
1190.Cm af21
1191.Pq Dv 010010 ,
1192.Cm af22
1193.Pq Dv 010100 ,
1194.Cm af23
1195.Pq Dv 010110 ,
1196.Cm af31
1197.Pq Dv 011010 ,
1198.Cm af32
1199.Pq Dv 011100 ,
1200.Cm af33
1201.Pq Dv 011110 ,
1202.Cm af41
1203.Pq Dv 100010 ,
1204.Cm af42
1205.Pq Dv 100100 ,
1206.Cm af43
1207.Pq Dv 100110 ,
1208.Cm va
1209.Pq Dv 101100 ,
1210.Cm ef
1211.Pq Dv 101110 ,
1212.Cm be
1213.Pq Dv 000000 .
1214Additionally, DSCP value can be specified by number (0..63).
1215It is also possible to use the
1216.Cm tablearg
1217keyword with setdscp.
1218If the tablearg value is not within the 0..63 range, lower 6 bits of supplied
1219value are used.
1220.It Cm tcp-setmss Ar mss
1221Set the Maximum Segment Size (MSS) in the TCP segment to value
1222.Ar mss .
1223The kernel module
1224.Cm ipfw_pmod
1225should be loaded or kernel should have
1226.Cm options IPFIREWALL_PMOD
1227to be able use this action.
1228This command does not change a packet if original MSS value is lower than
1229specified value.
1230Both TCP over IPv4 and over IPv6 are supported.
1231Regardless of matched a packet or not by the
1232.Cm tcp-setmss
1233rule, the search continues with the next rule.
1234.It Cm reass
1235Queue and reassemble IPv4 fragments.
1236If the packet is not fragmented, counters are updated and
1237processing continues with the next rule.
1238If the packet is the last logical fragment, the packet is reassembled and, if
1239.Va net.inet.ip.fw.one_pass
1240is set to 0, processing continues with the next rule.
1241Otherwise, the packet is allowed to pass and the search terminates.
1242If the packet is a fragment in the middle of a logical group of fragments,
1243it is consumed and
1244processing stops immediately.
1245.Pp
1246Fragment handling can be tuned via
1247.Va net.inet.ip.maxfragpackets
1248and
1249.Va net.inet.ip.maxfragsperpacket
1250which limit, respectively, the maximum number of processable
1251fragments (default: 800) and
1252the maximum number of fragments per packet (default: 16).
1253.Pp
1254NOTA BENE: since fragments do not contain port numbers,
1255they should be avoided with the
1256.Nm reass
1257rule.
1258Alternatively, direction-based (like
1259.Nm in
1260/
1261.Nm out
1262) and source-based (like
1263.Nm via
1264) match patterns can be used to select fragments.
1265.Pp
1266Usually a simple rule like:
1267.Bd -literal -offset indent
1268# reassemble incoming fragments
1269ipfw add reass all from any to any in
1270.Ed
1271.Pp
1272is all you need at the beginning of your ruleset.
1273.It Cm abort
1274Discard packets that match this rule, and if the packet is an SCTP packet,
1275try to send an SCTP packet containing an ABORT chunk.
1276The search terminates.
1277.It Cm abort6
1278Discard packets that match this rule, and if the packet is an SCTP packet,
1279try to send an SCTP packet containing an ABORT chunk.
1280The search terminates.
1281.El
1282.Ss RULE BODY
1283The body of a rule contains zero or more patterns (such as
1284specific source and destination addresses or ports,
1285protocol options, incoming or outgoing interfaces, etc.)
1286that the packet must match in order to be recognised.
1287In general, the patterns are connected by (implicit)
1288.Cm and
1289operators -- i.e., all must match in order for the
1290rule to match.
1291Individual patterns can be prefixed by the
1292.Cm not
1293operator to reverse the result of the match, as in
1294.Pp
1295.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
1296.Pp
1297Additionally, sets of alternative match patterns
1298.Pq Em or-blocks
1299can be constructed by putting the patterns in
1300lists enclosed between parentheses ( ) or braces { }, and
1301using the
1302.Cm or
1303operator as follows:
1304.Pp
1305.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
1306.Pp
1307Only one level of parentheses is allowed.
1308Beware that most shells have special meanings for parentheses
1309or braces, so it is advisable to put a backslash \\ in front of them
1310to prevent such interpretations.
1311.Pp
1312The body of a rule must in general include a source and destination
1313address specifier.
1314The keyword
1315.Ar any
1316can be used in various places to specify that the content of
1317a required field is irrelevant.
1318.Pp
1319The rule body has the following format:
1320.Bd -ragged -offset indent
1321.Op Ar proto Cm from Ar src Cm to Ar dst
1322.Op Ar options
1323.Ed
1324.Pp
1325The first part (proto from src to dst) is for backward
1326compatibility with earlier versions of
1327.Fx .
1328In modern
1329.Fx
1330any match pattern (including MAC headers, IP protocols,
1331addresses and ports) can be specified in the
1332.Ar options
1333section.
1334.Pp
1335Rule fields have the following meaning:
1336.Bl -tag -width indent
1337.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
1338.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
1339An IP protocol specified by number or name
1340(for a complete list see
1341.Pa /etc/protocols ) ,
1342or one of the following keywords:
1343.Bl -tag -width indent
1344.It Cm ip4 | ipv4
1345Matches IPv4 packets.
1346.It Cm ip6 | ipv6
1347Matches IPv6 packets.
1348.It Cm ip | all
1349Matches any packet.
1350.El
1351.Pp
1352The
1353.Cm ipv6
1354in
1355.Cm proto
1356option will be treated as inner protocol.
1357And, the
1358.Cm ipv4
1359is not available in
1360.Cm proto
1361option.
1362.Pp
1363The
1364.Cm { Ar protocol Cm or ... }
1365format (an
1366.Em or-block )
1367is provided for convenience only but its use is deprecated.
1368.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1369An address (or a list, see below)
1370optionally followed by
1371.Ar ports
1372specifiers.
1373.Pp
1374The second format
1375.Em ( or-block
1376with multiple addresses) is provided for convenience only and
1377its use is discouraged.
1378.It Ar addr : Oo Cm not Oc Bro
1379.Cm any | me | me6 |
1380.Cm table Ns Pq Ar name Ns Op , Ns Ar value
1381.Ar | addr-list | addr-set
1382.Brc
1383.Bl -tag -width indent
1384.It Cm any
1385Matches any IP address.
1386.It Cm me
1387Matches any IP address configured on an interface in the system.
1388.It Cm me6
1389Matches any IPv6 address configured on an interface in the system.
1390The address list is evaluated at the time the packet is
1391analysed.
1392.It Cm table Ns Pq Ar name Ns Op , Ns Ar value
1393Matches any IPv4 or IPv6 address for which an entry exists in the lookup table
1394.Ar number .
1395If an optional 32-bit unsigned
1396.Ar value
1397is also specified, an entry will match only if it has this value.
1398See the
1399.Sx LOOKUP TABLES
1400section below for more information on lookup tables.
1401.El
1402.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1403.It Ar ip-addr :
1404A host or subnet address specified in one of the following ways:
1405.Bl -tag -width indent
1406.It Ar numeric-ip | hostname
1407Matches a single IPv4 address, specified as dotted-quad or a hostname.
1408Hostnames are resolved at the time the rule is added to the firewall list.
1409.It Ar addr Ns / Ns Ar masklen
1410Matches all addresses with base
1411.Ar addr
1412(specified as an IP address, a network number, or a hostname)
1413and mask width of
1414.Cm masklen
1415bits.
1416As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1417all IP numbers from 1.2.3.0 to 1.2.3.127 .
1418.It Ar addr Ns : Ns Ar mask
1419Matches all addresses with base
1420.Ar addr
1421(specified as an IP address, a network number, or a hostname)
1422and the mask of
1423.Ar mask ,
1424specified as a dotted quad.
1425As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
14261.*.3.*.
1427This form is advised only for non-contiguous
1428masks.
1429It is better to resort to the
1430.Ar addr Ns / Ns Ar masklen
1431format for contiguous masks, which is more compact and less
1432error-prone.
1433.El
1434.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1435.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1436Matches all addresses with base address
1437.Ar addr
1438(specified as an IP address, a network number, or a hostname)
1439and whose last byte is in the list between braces { } .
1440Note that there must be no spaces between braces and
1441numbers (spaces after commas are allowed).
1442Elements of the list can be specified as single entries
1443or ranges.
1444The
1445.Ar masklen
1446field is used to limit the size of the set of addresses,
1447and can have any value between 24 and 32.
1448If not specified,
1449it will be assumed as 24.
1450.br
1451This format is particularly useful to handle sparse address sets
1452within a single rule.
1453Because the matching occurs using a
1454bitmask, it takes constant time and dramatically reduces
1455the complexity of rulesets.
1456.br
1457As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1458or 1.2.3.0/24{128,35-55,89}
1459will match the following IP addresses:
1460.br
14611.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1462.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1463.It Ar ip6-addr :
1464A host or subnet specified one of the following ways:
1465.Bl -tag -width indent
1466.It Ar numeric-ip | hostname
1467Matches a single IPv6 address as allowed by
1468.Xr inet_pton 3
1469or a hostname.
1470Hostnames are resolved at the time the rule is added to the firewall
1471list.
1472.It Ar addr Ns / Ns Ar masklen
1473Matches all IPv6 addresses with base
1474.Ar addr
1475(specified as allowed by
1476.Xr inet_pton 3
1477or a hostname)
1478and mask width of
1479.Cm masklen
1480bits.
1481.It Ar addr Ns / Ns Ar mask
1482Matches all IPv6 addresses with base
1483.Ar addr
1484(specified as allowed by
1485.Xr inet_pton 3
1486or a hostname)
1487and the mask of
1488.Ar mask ,
1489specified as allowed by
1490.Xr inet_pton 3 .
1491As an example, fe::640:0:0/ffff::ffff:ffff:0:0 will match
1492fe:*:*:*:0:640:*:*.
1493This form is advised only for non-contiguous
1494masks.
1495It is better to resort to the
1496.Ar addr Ns / Ns Ar masklen
1497format for contiguous masks, which is more compact and less
1498error-prone.
1499.El
1500.Pp
1501No support for sets of IPv6 addresses is provided because IPv6 addresses
1502are typically random past the initial prefix.
1503.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1504For protocols which support port numbers (such as SCTP, TCP and UDP), optional
1505.Cm ports
1506may be specified as one or more ports or port ranges, separated
1507by commas but no spaces, and an optional
1508.Cm not
1509operator.
1510The
1511.Ql \&-
1512notation specifies a range of ports (including boundaries).
1513.Pp
1514Service names (from
1515.Pa /etc/services )
1516may be used instead of numeric port values.
1517The length of the port list is limited to 30 ports or ranges,
1518though one can specify larger ranges by using an
1519.Em or-block
1520in the
1521.Cm options
1522section of the rule.
1523.Pp
1524A backslash
1525.Pq Ql \e
1526can be used to escape the dash
1527.Pq Ql -
1528character in a service name (from a shell, the backslash must be
1529typed twice to avoid the shell itself interpreting it as an escape
1530character).
1531.Pp
1532.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1533.Pp
1534Fragmented packets which have a non-zero offset (i.e., not the first
1535fragment) will never match a rule which has one or more port
1536specifications.
1537See the
1538.Cm frag
1539option for details on matching fragmented packets.
1540.El
1541.Ss RULE OPTIONS (MATCH PATTERNS)
1542Additional match patterns can be used within
1543rules.
1544Zero or more of these so-called
1545.Em options
1546can be present in a rule, optionally prefixed by the
1547.Cm not
1548operand, and possibly grouped into
1549.Em or-blocks .
1550.Pp
1551The following match patterns can be used (listed in alphabetical order):
1552.Bl -tag -width indent
1553.It Cm // this is a comment .
1554Inserts the specified text as a comment in the rule.
1555Everything following // is considered as a comment and stored in the rule.
1556You can have comment-only rules, which are listed as having a
1557.Cm count
1558action followed by the comment.
1559.It Cm bridged
1560Alias for
1561.Cm layer2 .
1562.It Cm defer-immediate-action | defer-action
1563A rule with this option will not perform normal action
1564upon a match.
1565This option is intended to be used with
1566.Cm record-state
1567or
1568.Cm keep-state
1569as the dynamic rule, created but ignored on match, will work
1570as intended.
1571Rules with both
1572.Cm record-state
1573and
1574.Cm defer-immediate-action
1575create a dynamic rule and continue with the next rule without actually
1576performing the action part of this rule.
1577When the rule is later activated via the state table, the action is
1578performed as usual.
1579.It Cm diverted
1580Matches only packets generated by a divert socket.
1581.It Cm diverted-loopback
1582Matches only packets coming from a divert socket back into the IP stack
1583input for delivery.
1584.It Cm diverted-output
1585Matches only packets going from a divert socket back outward to the IP
1586stack output for delivery.
1587.It Cm dst-ip Ar ip-address
1588Matches IPv4 packets whose destination IP is one of the address(es)
1589specified as argument.
1590.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1591Matches IPv6 packets whose destination IP is one of the address(es)
1592specified as argument.
1593.It Cm dst-port Ar ports
1594Matches IP packets whose destination port is one of the port(s)
1595specified as argument.
1596.It Cm established
1597Matches TCP packets that have the RST or ACK bits set.
1598.It Cm ext6hdr Ar header
1599Matches IPv6 packets containing the extended header given by
1600.Ar header .
1601Supported headers are:
1602.Pp
1603Fragment,
1604.Pq Cm frag ,
1605Hop-to-hop options
1606.Pq Cm hopopt ,
1607any type of Routing Header
1608.Pq Cm route ,
1609Source routing Routing Header Type 0
1610.Pq Cm rthdr0 ,
1611Mobile IPv6 Routing Header Type 2
1612.Pq Cm rthdr2 ,
1613Destination options
1614.Pq Cm dstopt ,
1615IPSec authentication headers
1616.Pq Cm ah ,
1617and IPsec encapsulated security payload headers
1618.Pq Cm esp .
1619.It Cm fib Ar fibnum
1620Matches a packet that has been tagged to use
1621the given FIB (routing table) number.
1622.It Cm flow Ar table Ns Pq Ar name Ns Op , Ns Ar value
1623Search for the flow entry in lookup table
1624.Ar name .
1625If not found, the match fails.
1626Otherwise, the match succeeds and
1627.Cm tablearg
1628is set to the value extracted from the table.
1629.Pp
1630This option can be useful to quickly dispatch traffic based on
1631certain packet fields.
1632See the
1633.Sx LOOKUP TABLES
1634section below for more information on lookup tables.
1635.It Cm flow-id Ar labels
1636Matches IPv6 packets containing any of the flow labels given in
1637.Ar labels .
1638.Ar labels
1639is a comma separated list of numeric flow labels.
1640.It Cm dst-mac Ar table Ns Pq Ar name Ns Op , Ns Ar value
1641Search for the destination MAC address entry in lookup table
1642.Ar name .
1643If not found, the match fails.
1644Otherwise, the match succeeds and
1645.Cm tablearg
1646is set to the value extracted from the table.
1647.It Cm src-mac Ar table Ns Pq Ar name Ns Op , Ns Ar value
1648Search for the source MAC address entry in lookup table
1649.Ar name .
1650If not found, the match fails.
1651Otherwise, the match succeeds and
1652.Cm tablearg
1653is set to the value extracted from the table.
1654.It Cm frag Ar spec
1655Matches IPv4 packets whose
1656.Cm ip_off
1657field contains the comma separated list of IPv4 fragmentation
1658options specified in
1659.Ar spec .
1660The recognized options are:
1661.Cm df
1662.Pq Dv don't fragment ,
1663.Cm mf
1664.Pq Dv more fragments ,
1665.Cm rf
1666.Pq Dv reserved fragment bit
1667.Cm offset
1668.Pq Dv non-zero fragment offset .
1669The absence of a particular options may be denoted
1670with a
1671.Ql \&! .
1672.Pp
1673Empty list of options defaults to matching on non-zero fragment offset.
1674Such rule would match all not the first fragment datagrams,
1675both IPv4 and IPv6.
1676This is a backward compatibility with older rulesets.
1677.It Cm gid Ar group
1678Matches all TCP or UDP packets sent by or received for a
1679.Ar group .
1680A
1681.Ar group
1682may be specified by name or number.
1683.It Cm jail Ar jail
1684Matches all TCP or UDP packets sent by or received for the
1685jail whose ID or name is
1686.Ar jail .
1687.It Cm icmptypes Ar types
1688Matches ICMP packets whose ICMP type is in the list
1689.Ar types .
1690The list may be specified as any combination of
1691individual types (numeric) separated by commas.
1692.Em Ranges are not allowed .
1693The supported ICMP types are:
1694.Pp
1695echo reply
1696.Pq Cm 0 ,
1697destination unreachable
1698.Pq Cm 3 ,
1699source quench
1700.Pq Cm 4 ,
1701redirect
1702.Pq Cm 5 ,
1703echo request
1704.Pq Cm 8 ,
1705router advertisement
1706.Pq Cm 9 ,
1707router solicitation
1708.Pq Cm 10 ,
1709time-to-live exceeded
1710.Pq Cm 11 ,
1711IP header bad
1712.Pq Cm 12 ,
1713timestamp request
1714.Pq Cm 13 ,
1715timestamp reply
1716.Pq Cm 14 ,
1717information request
1718.Pq Cm 15 ,
1719information reply
1720.Pq Cm 16 ,
1721address mask request
1722.Pq Cm 17
1723and address mask reply
1724.Pq Cm 18 .
1725.It Cm icmp6types Ar types
1726Matches ICMP6 packets whose ICMP6 type is in the list of
1727.Ar types .
1728The list may be specified as any combination of
1729individual types (numeric) separated by commas.
1730.Em Ranges are not allowed .
1731.It Cm in | out
1732Matches incoming or outgoing packets, respectively.
1733.Cm in
1734and
1735.Cm out
1736are mutually exclusive (in fact,
1737.Cm out
1738is implemented as
1739.Cm not in Ns No ).
1740.It Cm ipid Ar id-list
1741Matches IPv4 packets whose
1742.Cm ip_id
1743field has value included in
1744.Ar id-list ,
1745which is either a single value or a list of values or ranges
1746specified in the same way as
1747.Ar ports .
1748.It Cm iplen Ar len-list
1749Matches IP packets whose total length, including header and data, is
1750in the set
1751.Ar len-list ,
1752which is either a single value or a list of values or ranges
1753specified in the same way as
1754.Ar ports .
1755.It Cm ipoptions Ar spec
1756Matches packets whose IPv4 header contains the comma separated list of
1757options specified in
1758.Ar spec .
1759The supported IP options are:
1760.Pp
1761.Cm ssrr
1762(strict source route),
1763.Cm lsrr
1764(loose source route),
1765.Cm rr
1766(record packet route) and
1767.Cm ts
1768(timestamp).
1769The absence of a particular option may be denoted
1770with a
1771.Ql \&! .
1772.It Cm ipprecedence Ar precedence
1773Matches IPv4 packets whose precedence field is equal to
1774.Ar precedence .
1775.It Cm ipsec
1776Matches packets that have IPSEC history associated with them
1777(i.e., the packet comes encapsulated in IPSEC, the kernel
1778has IPSEC support, and can correctly decapsulate it).
1779.Pp
1780Note that specifying
1781.Cm ipsec
1782is different from specifying
1783.Cm proto Ar ipsec
1784as the latter will only look at the specific IP protocol field,
1785irrespective of IPSEC kernel support and the validity of the IPSEC data.
1786.Pp
1787Further note that this flag is silently ignored in kernels without
1788IPSEC support.
1789It does not affect rule processing when given and the
1790rules are handled as if with no
1791.Cm ipsec
1792flag.
1793.It Cm iptos Ar spec
1794Matches IPv4 packets whose
1795.Cm tos
1796field contains the comma separated list of
1797service types specified in
1798.Ar spec .
1799The supported IP types of service are:
1800.Pp
1801.Cm lowdelay
1802.Pq Dv IPTOS_LOWDELAY ,
1803.Cm throughput
1804.Pq Dv IPTOS_THROUGHPUT ,
1805.Cm reliability
1806.Pq Dv IPTOS_RELIABILITY ,
1807.Cm mincost
1808.Pq Dv IPTOS_MINCOST ,
1809.Cm congestion
1810.Pq Dv IPTOS_ECN_CE .
1811The absence of a particular type may be denoted
1812with a
1813.Ql \&! .
1814.It Cm dscp spec Ns Op , Ns Ar spec
1815Matches IPv4/IPv6 packets whose
1816.Cm DS
1817field value is contained in
1818.Ar spec
1819mask.
1820Multiple values can be specified via
1821the comma separated list.
1822Value can be one of keywords used in
1823.Cm setdscp
1824action or exact number.
1825.It Cm ipttl Ar ttl-list
1826Matches IPv4 packets whose time to live is included in
1827.Ar ttl-list ,
1828which is either a single value or a list of values or ranges
1829specified in the same way as
1830.Ar ports .
1831.It Cm ipversion Ar ver
1832Matches IP packets whose IP version field is
1833.Ar ver .
1834.It Cm keep-state Op Ar :flowname
1835Upon a match, the firewall will create a dynamic rule, whose
1836default behaviour is to match bidirectional traffic between
1837source and destination IP/port using the same protocol.
1838The rule has a limited lifetime (controlled by a set of
1839.Xr sysctl 8
1840variables), and the lifetime is refreshed every time a matching
1841packet is found.
1842The
1843.Ar :flowname
1844is used to assign additional to addresses, ports and protocol parameter
1845to dynamic rule.
1846It can be used for more accurate matching by
1847.Cm check-state
1848rule.
1849The
1850.Cm :default
1851keyword is special name used for compatibility with old rulesets.
1852.It Cm layer2
1853Matches only layer2 packets, i.e., those passed to
1854.Nm
1855from
1856.Fn ether_demux
1857and
1858.Fn ether_output_frame .
1859.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N Op Ar :flowname
1860The firewall will only allow
1861.Ar N
1862connections with the same
1863set of parameters as specified in the rule.
1864One or more
1865of source and destination addresses and ports can be
1866specified.
1867.It Cm lookup Bro Cm dst-ip | dst-port | dst-mac | src-ip | src-port | src-mac | uid |
1868.Cm jail | dscp | mark Brc Ar name
1869Search an entry in lookup table
1870.Ar name
1871that matches the field specified as argument.
1872If not found, the match fails.
1873Otherwise, the match succeeds and
1874.Cm tablearg
1875is set to the value extracted from the table.
1876.Pp
1877This option can be useful to quickly dispatch traffic based on
1878certain packet fields.
1879See the
1880.Sx LOOKUP TABLES
1881section below for more information on lookup tables.
1882.It Cm { MAC | mac } Ar dst-mac src-mac
1883Match packets with a given
1884.Ar dst-mac
1885and
1886.Ar src-mac
1887addresses, specified as the
1888.Cm any
1889keyword (matching any MAC address), or six groups of hex digits
1890separated by colons,
1891and optionally followed by a mask indicating the significant bits.
1892The mask may be specified using either of the following methods:
1893.Bl -enum -width indent
1894.It
1895A slash
1896.Pq /
1897followed by the number of significant bits.
1898For example, an address with 33 significant bits could be specified as:
1899.Pp
1900.Dl "MAC 10:20:30:40:50:60/33 any"
1901.It
1902An ampersand
1903.Pq &
1904followed by a bitmask specified as six groups of hex digits separated
1905by colons.
1906For example, an address in which the last 16 bits are significant could
1907be specified as:
1908.Pp
1909.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1910.Pp
1911Note that the ampersand character has a special meaning in many shells
1912and should generally be escaped.
1913.El
1914Note that the order of MAC addresses (destination first,
1915source second) is
1916the same as on the wire, but the opposite of the one used for
1917IP addresses.
1918.It Cm mac-type Ar mac-type
1919Matches packets whose Ethernet Type field
1920corresponds to one of those specified as argument.
1921.Ar mac-type
1922is specified in the same way as
1923.Cm port numbers
1924(i.e., one or more comma-separated single values or ranges).
1925You can use symbolic names for known values such as
1926.Em vlan , ipv4, ipv6 .
1927Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1928and they are always printed as hexadecimal (unless the
1929.Cm -N
1930option is used, in which case symbolic resolution will be attempted).
1931.It Cm proto Ar protocol
1932Matches packets with the corresponding IP protocol.
1933.It Cm record-state
1934Upon a match, the firewall will create a dynamic rule as if
1935.Cm keep-state
1936was specified.
1937However, this option doesn't imply an implicit
1938.Cm check-state
1939in contrast to
1940.Cm keep-state .
1941.It Cm recv | xmit | via Brq Ar ifX | Ar ifmask | Ar table Ns Po Ar name Ns Oo , Ns Ar value Oc Pc | Ar ipno | Ar any
1942Matches packets received, transmitted or going through,
1943respectively, the interface specified by exact name
1944.Po Ar ifX Pc ,
1945by device mask
1946.Po Ar ifmask Pc ,
1947by IP address, or through some interface.
1948.Pp
1949Interface
1950name may be matched against
1951.Ar ifmask
1952with
1953.Xr fnmatch 3
1954according to the rules used by the shell (f.e. tun*).
1955See also the
1956.Sx EXAMPLES
1957section.
1958.Pp
1959Table
1960.Ar name
1961may be used to match interface by its kernel ifindex.
1962See the
1963.Sx LOOKUP TABLES
1964section below for more information on lookup tables.
1965.Pp
1966The
1967.Cm via
1968keyword causes the interface to always be checked.
1969If
1970.Cm recv
1971or
1972.Cm xmit
1973is used instead of
1974.Cm via ,
1975then only the receive or transmit interface (respectively)
1976is checked.
1977By specifying both, it is possible to match packets based on
1978both receive and transmit interface, e.g.:
1979.Pp
1980.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1981.Pp
1982The
1983.Cm recv
1984interface can be tested on either incoming or outgoing packets,
1985while the
1986.Cm xmit
1987interface can only be tested on outgoing packets.
1988So
1989.Cm out
1990is required (and
1991.Cm in
1992is invalid) whenever
1993.Cm xmit
1994is used.
1995.Pp
1996A packet might not have a receive or transmit interface: packets
1997originating from the local host have no receive interface,
1998while packets destined for the local host have no transmit
1999interface.
2000.It Cm set-limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
2001Works like
2002.Cm limit
2003but does not have an implicit
2004.Cm check-state
2005attached to it.
2006.It Cm setup
2007Matches TCP packets that have the SYN bit set but no ACK bit.
2008This is the short form of
2009.Dq Li tcpflags\ syn,!ack .
2010.It Cm sockarg
2011Matches packets that are associated to a local socket and
2012for which the SO_USER_COOKIE socket option has been set
2013to a non-zero value.
2014As a side effect, the value of the
2015option is made available as
2016.Cm tablearg
2017value, which in turn can be used as
2018.Cm skipto
2019or
2020.Cm pipe
2021number.
2022.It Cm src-ip Ar ip-address
2023Matches IPv4 packets whose source IP is one of the address(es)
2024specified as an argument.
2025.It Cm src-ip6 Ar ip6-address
2026Matches IPv6 packets whose source IP is one of the address(es)
2027specified as an argument.
2028.It Cm src-port Ar ports
2029Matches IP packets whose source port is one of the port(s)
2030specified as argument.
2031.It Cm tagged Ar tag-list
2032Matches packets whose tags are included in
2033.Ar tag-list ,
2034which is either a single value or a list of values or ranges
2035specified in the same way as
2036.Ar ports .
2037Tags can be applied to the packet using
2038.Cm tag
2039rule action parameter (see it's description for details on tags).
2040.It Cm mark Ar value[:bitmask] | tablearg[:bitmask]
2041Matches packets whose mark is equal to
2042.Ar value
2043with optional
2044.Ar bitmask
2045applied to it.
2046.Cm tablearg
2047can also be used instead of an explicit
2048.Ar value
2049to match a value supplied by the last table lookup.
2050.Pp
2051Both
2052.Ar value
2053and
2054.Ar bitmask
2055can be entered as decimal or hexadecimal (if prefixed by 0x), and they
2056are always printed as hexadecimal.
2057.It Cm tcpack Ar ack
2058TCP packets only.
2059Match if the TCP header acknowledgment number field is set to
2060.Ar ack .
2061.It Cm tcpdatalen Ar tcpdatalen-list
2062Matches TCP packets whose length of TCP data is
2063.Ar tcpdatalen-list ,
2064which is either a single value or a list of values or ranges
2065specified in the same way as
2066.Ar ports .
2067.It Cm tcpflags Ar spec
2068TCP packets only.
2069Match if the TCP header contains the comma separated list of
2070flags specified in
2071.Ar spec .
2072The supported TCP flags are:
2073.Pp
2074.Cm fin ,
2075.Cm syn ,
2076.Cm rst ,
2077.Cm psh ,
2078.Cm ack
2079and
2080.Cm urg .
2081The absence of a particular flag may be denoted
2082with a
2083.Ql \&! .
2084A rule which contains a
2085.Cm tcpflags
2086specification can never match a fragmented packet which has
2087a non-zero offset.
2088See the
2089.Cm frag
2090option for details on matching fragmented packets.
2091.It Cm tcpmss Ar tcpmss-list
2092Matches TCP packets whose MSS (maximum segment size) value is set to
2093.Ar tcpmss-list ,
2094which is either a single value or a list of values or ranges
2095specified in the same way as
2096.Ar ports .
2097.It Cm tcpseq Ar seq
2098TCP packets only.
2099Match if the TCP header sequence number field is set to
2100.Ar seq .
2101.It Cm tcpwin Ar tcpwin-list
2102Matches TCP packets whose  header window field is set to
2103.Ar tcpwin-list ,
2104which is either a single value or a list of values or ranges
2105specified in the same way as
2106.Ar ports .
2107.It Cm tcpoptions Ar spec
2108TCP packets only.
2109Match if the TCP header contains the comma separated list of
2110options specified in
2111.Ar spec .
2112The supported TCP options are:
2113.Pp
2114.Cm mss
2115(maximum segment size),
2116.Cm window
2117(tcp window advertisement),
2118.Cm sack
2119(selective ack),
2120.Cm ts
2121(rfc1323 timestamp) and
2122.Cm cc
2123(rfc1644 t/tcp connection count).
2124The absence of a particular option may be denoted
2125with a
2126.Ql \&! .
2127.It Cm uid Ar user
2128Match all TCP or UDP packets sent by or received for a
2129.Ar user .
2130A
2131.Ar user
2132may be matched by name or identification number.
2133.It Cm verrevpath
2134For incoming packets,
2135a routing table lookup is done on the packet's source address.
2136If the interface on which the packet entered the system matches the
2137outgoing interface for the route,
2138the packet matches.
2139If the interfaces do not match up,
2140the packet does not match.
2141All outgoing packets or packets with no incoming interface match.
2142.Pp
2143The name and functionality of the option is intentionally similar to
2144the Cisco IOS command:
2145.Pp
2146.Dl ip verify unicast reverse-path
2147.Pp
2148This option can be used to make anti-spoofing rules to reject all
2149packets with source addresses not from this interface.
2150See also the option
2151.Cm antispoof .
2152.It Cm versrcreach
2153For incoming packets,
2154a routing table lookup is done on the packet's source address.
2155If a route to the source address exists, but not the default route
2156or a blackhole/reject route, the packet matches.
2157Otherwise, the packet does not match.
2158All outgoing packets match.
2159.Pp
2160The name and functionality of the option is intentionally similar to
2161the Cisco IOS command:
2162.Pp
2163.Dl ip verify unicast source reachable-via any
2164.Pp
2165This option can be used to make anti-spoofing rules to reject all
2166packets whose source address is unreachable.
2167.It Cm antispoof
2168For incoming packets, the packet's source address is checked if it
2169belongs to a directly connected network.
2170If the network is directly connected, then the interface the packet
2171came on in is compared to the interface the network is connected to.
2172When incoming interface and directly connected interface are not the
2173same, the packet does not match.
2174Otherwise, the packet does match.
2175All outgoing packets match.
2176.Pp
2177This option can be used to make anti-spoofing rules to reject all
2178packets that pretend to be from a directly connected network but do
2179not come in through that interface.
2180This option is similar to but more restricted than
2181.Cm verrevpath
2182because it engages only on packets with source addresses of directly
2183connected networks instead of all source addresses.
2184.El
2185.Sh LOOKUP TABLES
2186Lookup tables are useful to handle large sparse sets of
2187addresses or other search keys (e.g., ports, jail IDs, interface names).
2188In the rest of this section we will use the term ``key''.
2189Table name needs to match the following spec:
2190.Ar table-name .
2191Tables with the same name can be created in different
2192.Ar sets .
2193However, rule links to the tables in
2194.Ar set 0
2195by default.
2196This behavior can be controlled by
2197.Va net.inet.ip.fw.tables_sets
2198variable.
2199See the
2200.Sx SETS OF RULES
2201section for more information.
2202There may be up to 65535 different lookup tables.
2203.Pp
2204The following table types are supported:
2205.Bl -tag -width indent
2206.It Ar table-type : Ar addr | iface | number | flow | mac
2207.It Ar table-key : Ar addr Ns Oo / Ns Ar masklen Oc | iface-name | number | flow-spec
2208.It Ar flow-spec : Ar flow-field Ns Op , Ns Ar flow-spec
2209.It Ar flow-field : src-ip | proto | src-port | dst-ip | dst-port
2210.It Cm addr
2211Matches IPv4 or IPv6 address.
2212Each entry is represented by an
2213.Ar addr Ns Op / Ns Ar masklen
2214and will match all addresses with base
2215.Ar addr
2216(specified as an IPv4/IPv6 address, or a hostname) and mask width of
2217.Ar masklen
2218bits.
2219If
2220.Ar masklen
2221is not specified, it defaults to 32 for IPv4 and 128 for IPv6.
2222When looking up an IP address in a table, the most specific
2223entry will match.
2224.It Cm iface
2225Matches interface names.
2226Each entry is represented by string treated as interface name.
2227Wildcards are not supported.
2228.It Cm number
2229Matches protocol ports, uids/gids or jail IDs.
2230Each entry is represented by 32-bit unsigned integer.
2231Ranges are not supported.
2232.It Cm flow
2233Matches packet fields specified by
2234.Ar flow
2235type suboptions with table entries.
2236.It Cm mac
2237Matches MAC address.
2238Each entry is represented by an
2239.Ar addr Ns Op / Ns Ar masklen
2240and will match all addresses with base
2241.Ar addr
2242and mask width of
2243.Ar masklen
2244bits.
2245If
2246.Ar masklen
2247is not specified, it defaults to 48.
2248When looking up an MAC address in a table, the most specific
2249entry will match.
2250.El
2251.Pp
2252Tables require explicit creation via
2253.Cm create
2254before use.
2255.Pp
2256The following creation options are supported:
2257.Bl -tag -width indent
2258.It Ar create-options : Ar create-option | create-options
2259.It Ar create-option : Cm type Ar table-type | Cm valtype Ar value-mask | Cm algo Ar algo-desc |
2260.Cm limit Ar number | Cm locked | Cm missing | Cm or-flush
2261.It Cm type
2262Table key type.
2263.It Cm valtype
2264Table value mask.
2265.It Cm algo
2266Table algorithm to use (see below).
2267.It Cm limit
2268Maximum number of items that may be inserted into table.
2269.It Cm locked
2270Restrict any table modifications.
2271.It Cm missing
2272Do not fail if table already exists and has exactly same options as new one.
2273.It Cm or-flush
2274Flush existing table with same name instead of returning error.
2275Implies
2276.Cm missing
2277so existing table must be compatible with new one.
2278.El
2279.Pp
2280Some of these options may be modified later via
2281.Cm modify
2282keyword.
2283The following options can be changed:
2284.Bl -tag -width indent
2285.It Ar modify-options : Ar modify-option | modify-options
2286.It Ar modify-option : Cm limit Ar number
2287.It Cm limit
2288Alter maximum number of items that may be inserted into table.
2289.El
2290.Pp
2291Additionally, table can be locked or unlocked using
2292.Cm lock
2293or
2294.Cm unlock
2295commands.
2296.Pp
2297Tables of the same
2298.Ar type
2299can be swapped with each other using
2300.Cm swap Ar name
2301command.
2302Swap may fail if tables limits are set and data exchange
2303would result in limits hit.
2304Operation is performed atomically.
2305.Pp
2306One or more entries can be added to a table at once using
2307.Cm add
2308command.
2309Addition of all items are performed atomically.
2310By default, error in addition of one entry does not influence
2311addition of other entries.
2312However, non-zero error code is returned in that case.
2313Special
2314.Cm atomic
2315keyword may be specified before
2316.Cm add
2317to indicate all-or-none add request.
2318.Pp
2319One or more entries can be removed from a table at once using
2320.Cm delete
2321command.
2322By default, error in removal of one entry does not influence
2323removing of other entries.
2324However, non-zero error code is returned in that case.
2325.Pp
2326It may be possible to check what entry will be found on particular
2327.Ar table-key
2328using
2329.Cm lookup
2330.Ar table-key
2331command.
2332This functionality is optional and may be unsupported in some algorithms.
2333.Pp
2334The following operations can be performed on
2335.Ar one
2336or
2337.Cm all
2338tables:
2339.Bl -tag -width indent
2340.It Cm list
2341List all entries.
2342.It Cm flush
2343Removes all entries.
2344.It Cm info
2345Shows generic table information.
2346.It Cm detail
2347Shows generic table information and algo-specific data.
2348.El
2349.Pp
2350The following lookup algorithms are supported:
2351.Bl -tag -width indent
2352.It Ar algo-desc : algo-name | "algo-name algo-data"
2353.It Ar algo-name : Ar addr: radix | addr: hash | iface: array | number: array | flow: hash | mac: radix
2354.It Cm addr: radix
2355Separate Radix trees for IPv4 and IPv6, the same way as the routing table (see
2356.Xr route 4 ) .
2357Default choice for
2358.Ar addr
2359type.
2360.It Cm addr:hash
2361Separate auto-growing hashes for IPv4 and IPv6.
2362Accepts entries with the same mask length specified initially via
2363.Cm "addr:hash masks=/v4,/v6"
2364algorithm creation options.
2365Assume /32 and /128 masks by default.
2366Search removes host bits (according to mask) from supplied address and checks
2367resulting key in appropriate hash.
2368Mostly optimized for /64 and byte-ranged IPv6 masks.
2369.It Cm iface:array
2370Array storing sorted indexes for entries which are presented in the system.
2371Optimized for very fast lookup.
2372.It Cm number:array
2373Array storing sorted u32 numbers.
2374.It Cm flow:hash
2375Auto-growing hash storing flow entries.
2376Search calculates hash on required packet fields and searches for matching
2377entries in selected bucket.
2378.It Cm mac: radix
2379Radix tree for MAC address
2380.El
2381.Pp
2382The
2383.Cm tablearg
2384feature provides the ability to use a value, looked up in the table, as
2385the argument for a rule action, action parameter or rule option.
2386This can significantly reduce number of rules in some configurations.
2387If two tables are used in a rule, the result of the second (destination)
2388is used.
2389.Pp
2390Each record may hold one or more values according to
2391.Ar value-mask .
2392This mask is set on table creation via
2393.Cm valtype
2394option.
2395The following value types are supported:
2396.Bl -tag -width indent
2397.It Ar value-mask : Ar value-type Ns Op , Ns Ar value-mask
2398.It Ar value-type : Ar skipto | pipe | fib | nat | dscp | tag | divert |
2399.Ar netgraph | limit | ipv4 | ipv6 | mark
2400.It Cm skipto
2401rule number to jump to.
2402.It Cm pipe
2403Pipe number to use.
2404.It Cm fib
2405fib number to match/set.
2406.It Cm nat
2407nat number to jump to.
2408.It Cm dscp
2409dscp value to match/set.
2410.It Cm tag
2411tag number to match/set.
2412.It Cm divert
2413port number to divert traffic to.
2414.It Cm netgraph
2415hook number to move packet to.
2416.It Cm limit
2417maximum number of connections.
2418.It Cm ipv4
2419IPv4 nexthop to fwd packets to.
2420.It Cm ipv6
2421IPv6 nexthop to fwd packets to.
2422.It Cm mark
2423mark value to match/set.
2424.El
2425.Pp
2426The
2427.Cm tablearg
2428argument can be used with the following actions:
2429.Cm nat, pipe, queue, divert, tee, netgraph, ngtee, fwd, skipto, setfib ,
2430.Cm setmark ,
2431action parameters:
2432.Cm tag, untag ,
2433rule options:
2434.Cm limit, tagged, mark .
2435.Pp
2436When used with the
2437.Cm skipto
2438action, the user should be aware that the code will walk the ruleset
2439up to a rule equal to, or past, the given number.
2440.Pp
2441See the
2442.Sx EXAMPLES
2443Section for example usage of tables and the tablearg keyword.
2444.Sh SETS OF RULES
2445Each rule or table belongs to one of 32 different
2446.Em sets
2447, numbered 0 to 31.
2448Set 31 is reserved for the default rule.
2449.Pp
2450By default, rules or tables are put in set 0, unless you use the
2451.Cm set N
2452attribute when adding a new rule or table.
2453Sets can be individually and atomically enabled or disabled,
2454so this mechanism permits an easy way to store multiple configurations
2455of the firewall and quickly (and atomically) switch between them.
2456.Pp
2457By default, tables from set 0 are referenced when adding rule with
2458table opcodes regardless of rule set.
2459This behavior can be changed by setting
2460.Va net.inet.ip.fw.tables_sets
2461variable to 1.
2462Rule's set will then be used for table references.
2463.Pp
2464The command to enable/disable sets is
2465.Bd -ragged -offset indent
2466.Nm
2467.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
2468.Ed
2469.Pp
2470where multiple
2471.Cm enable
2472or
2473.Cm disable
2474sections can be specified.
2475Command execution is atomic on all the sets specified in the command.
2476By default, all sets are enabled.
2477.Pp
2478When you disable a set, its rules behave as if they do not exist
2479in the firewall configuration, with only one exception:
2480.Bd -ragged -offset indent
2481dynamic rules created from a rule before it had been disabled
2482will still be active until they expire.
2483In order to delete
2484dynamic rules you have to explicitly delete the parent rule
2485which generated them.
2486.Ed
2487.Pp
2488The set number of rules can be changed with the command
2489.Bd -ragged -offset indent
2490.Nm
2491.Cm set move
2492.Brq Cm rule Ar rule-number | old-set
2493.Cm to Ar new-set
2494.Ed
2495.Pp
2496Also, you can atomically swap two rulesets with the command
2497.Bd -ragged -offset indent
2498.Nm
2499.Cm set swap Ar first-set second-set
2500.Ed
2501.Pp
2502See the
2503.Sx EXAMPLES
2504Section on some possible uses of sets of rules.
2505.Sh STATEFUL FIREWALL
2506Stateful operation is a way for the firewall to dynamically
2507create rules for specific flows when packets that
2508match a given pattern are detected.
2509Support for stateful
2510operation comes through the
2511.Cm check-state , keep-state , record-state , limit
2512and
2513.Cm set-limit
2514options of
2515.Nm rules .
2516.Pp
2517Dynamic rules are created when a packet matches a
2518.Cm keep-state ,
2519.Cm record-state ,
2520.Cm limit
2521or
2522.Cm set-limit
2523rule, causing the creation of a
2524.Em dynamic
2525rule which will match all and only packets with
2526a given
2527.Em protocol
2528between a
2529.Em src-ip/src-port dst-ip/dst-port
2530pair of addresses
2531.Em ( src
2532and
2533.Em dst
2534are used here only to denote the initial match addresses, but they
2535are completely equivalent afterwards).
2536Rules created by
2537.Cm keep-state
2538option also have a
2539.Ar :flowname
2540taken from it.
2541This name is used in matching together with addresses, ports and protocol.
2542Dynamic rules will be checked at the first
2543.Cm check-state, keep-state
2544or
2545.Cm limit
2546occurrence, and the action performed upon a match will be the same
2547as in the parent rule.
2548.Pp
2549Note that no additional attributes other than protocol and IP addresses
2550and ports and :flowname are checked on dynamic rules.
2551.Pp
2552The typical use of dynamic rules is to keep a closed firewall configuration,
2553but let the first TCP SYN packet from the inside network install a
2554dynamic rule for the flow so that packets belonging to that session
2555will be allowed through the firewall:
2556.Pp
2557.Dl "ipfw add check-state :OUTBOUND"
2558.Dl "ipfw add allow tcp from my-subnet to any setup keep-state :OUTBOUND"
2559.Dl "ipfw add deny tcp from any to any"
2560.Pp
2561A similar approach can be used for UDP, where an UDP packet coming
2562from the inside will install a dynamic rule to let the response through
2563the firewall:
2564.Pp
2565.Dl "ipfw add check-state :OUTBOUND"
2566.Dl "ipfw add allow udp from my-subnet to any keep-state :OUTBOUND"
2567.Dl "ipfw add deny udp from any to any"
2568.Pp
2569Dynamic rules expire after some time, which depends on the status
2570of the flow and the setting of some
2571.Cm sysctl
2572variables.
2573See Section
2574.Sx SYSCTL VARIABLES
2575for more details.
2576For TCP sessions, dynamic rules can be instructed to periodically
2577send keepalive packets to refresh the state of the rule when it is
2578about to expire.
2579.Pp
2580See Section
2581.Sx EXAMPLES
2582for more examples on how to use dynamic rules.
2583.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
2584.Nm
2585is also the user interface for the
2586.Nm dummynet
2587traffic shaper, packet scheduler and network emulator, a subsystem that
2588can artificially queue, delay or drop packets
2589emulating the behaviour of certain network links
2590or queueing systems.
2591.Pp
2592.Nm dummynet
2593operates by first using the firewall to select packets
2594using any match pattern that can be used in
2595.Nm
2596rules.
2597Matching packets are then passed to either of two
2598different objects, which implement the traffic regulation:
2599.Bl -hang -offset XXXX
2600.It Em pipe
2601A
2602.Em pipe
2603emulates a
2604.Em link
2605with given bandwidth and propagation delay,
2606driven by a FIFO scheduler and a single queue with programmable
2607queue size and packet loss rate.
2608Packets are appended to the queue as they come out from
2609.Nm ipfw ,
2610and then transferred in FIFO order to the link at the desired rate.
2611.It Em queue
2612A
2613.Em queue
2614is an abstraction used to implement packet scheduling
2615using one of several packet scheduling algorithms.
2616Packets sent to a
2617.Em queue
2618are first grouped into flows according to a mask on the 5-tuple.
2619Flows are then passed to the scheduler associated to the
2620.Em queue ,
2621and each flow uses scheduling parameters (weight and others)
2622as configured in the
2623.Em queue
2624itself.
2625A scheduler in turn is connected to an emulated link,
2626and arbitrates the link's bandwidth among backlogged flows according to
2627weights and to the features of the scheduling algorithm in use.
2628.El
2629.Pp
2630In practice,
2631.Em pipes
2632can be used to set hard limits to the bandwidth that a flow can use, whereas
2633.Em queues
2634can be used to determine how different flows share the available bandwidth.
2635.Pp
2636A graphical representation of the binding of queues,
2637flows, schedulers and links is below.
2638.Bd -literal -offset indent
2639                 (flow_mask|sched_mask)  sched_mask
2640         +---------+   weight Wx  +-------------+
2641         |         |->-[flow]-->--|             |-+
2642    -->--| QUEUE x |   ...        |             | |
2643         |         |->-[flow]-->--| SCHEDuler N | |
2644         +---------+              |             | |
2645             ...                  |             +--[LINK N]-->--
2646         +---------+   weight Wy  |             | +--[LINK N]-->--
2647         |         |->-[flow]-->--|             | |
2648    -->--| QUEUE y |   ...        |             | |
2649         |         |->-[flow]-->--|             | |
2650         +---------+              +-------------+ |
2651                                    +-------------+
2652.Ed
2653It is important to understand the role of the SCHED_MASK
2654and FLOW_MASK, which are configured through the commands
2655.Dl "ipfw sched N config mask SCHED_MASK ..."
2656and
2657.Dl "ipfw queue X config mask FLOW_MASK ..." .
2658.Pp
2659The SCHED_MASK is used to assign flows to one or more
2660scheduler instances, one for each
2661value of the packet's 5-tuple after applying SCHED_MASK.
2662As an example, using ``src-ip 0xffffff00'' creates one instance
2663for each /24 destination subnet.
2664.Pp
2665The FLOW_MASK, together with the SCHED_MASK, is used to split
2666packets into flows.
2667As an example, using
2668``src-ip 0x000000ff''
2669together with the previous SCHED_MASK makes a flow for
2670each individual source address.
2671In turn, flows for each /24
2672subnet will be sent to the same scheduler instance.
2673.Pp
2674The above diagram holds even for the
2675.Em pipe
2676case, with the only restriction that a
2677.Em pipe
2678only supports a SCHED_MASK, and forces the use of a FIFO
2679scheduler (these are for backward compatibility reasons;
2680in fact, internally, a
2681.Nm dummynet's
2682pipe is implemented exactly as above).
2683.Pp
2684There are two modes of
2685.Nm dummynet
2686operation:
2687.Dq normal
2688and
2689.Dq fast .
2690The
2691.Dq normal
2692mode tries to emulate a real link: the
2693.Nm dummynet
2694scheduler ensures that the packet will not leave the pipe faster than it
2695would on the real link with a given bandwidth.
2696The
2697.Dq fast
2698mode allows certain packets to bypass the
2699.Nm dummynet
2700scheduler (if packet flow does not exceed pipe's bandwidth).
2701This is the reason why the
2702.Dq fast
2703mode requires less CPU cycles per packet (on average) and packet latency
2704can be significantly lower in comparison to a real link with the same
2705bandwidth.
2706The default mode is
2707.Dq normal .
2708The
2709.Dq fast
2710mode can be enabled by setting the
2711.Va net.inet.ip.dummynet.io_fast
2712.Xr sysctl 8
2713variable to a non-zero value.
2714.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
2715The
2716.Em pipe ,
2717.Em queue
2718and
2719.Em scheduler
2720configuration commands are the following:
2721.Bd -ragged -offset indent
2722.Cm pipe Ar number Cm config Ar pipe-configuration
2723.Pp
2724.Cm queue Ar number Cm config Ar queue-configuration
2725.Pp
2726.Cm sched Ar number Cm config Ar sched-configuration
2727.Ed
2728.Pp
2729The following parameters can be configured for a pipe:
2730.Pp
2731.Bl -tag -width indent -compact
2732.It Cm bw Ar bandwidth | device
2733Bandwidth, measured in
2734.Sm off
2735.Op Cm K | M | G
2736.Brq Cm bit/s | Byte/s .
2737.Sm on
2738.Pp
2739A value of 0 (default) means unlimited bandwidth.
2740The unit must immediately follow the number, as in
2741.Pp
2742.Dl "dnctl pipe 1 config bw 300Kbit/s"
2743.Pp
2744If a device name is specified instead of a numeric value, as in
2745.Pp
2746.Dl "dnctl pipe 1 config bw tun0"
2747.Pp
2748then the transmit clock is supplied by the specified device.
2749At the moment only the
2750.Xr tun 4
2751device supports this
2752functionality, for use in conjunction with
2753.Xr ppp 8 .
2754.Pp
2755.It Cm delay Ar ms-delay
2756Propagation delay, measured in milliseconds.
2757The value is rounded to the next multiple of the clock tick
2758(typically 10ms, but it is a good practice to run kernels
2759with
2760.Dq "options HZ=1000"
2761to reduce
2762the granularity to 1ms or less).
2763The default value is 0, meaning no delay.
2764.Pp
2765.It Cm burst Ar size
2766If the data to be sent exceeds the pipe's bandwidth limit
2767(and the pipe was previously idle), up to
2768.Ar size
2769bytes of data are allowed to bypass the
2770.Nm dummynet
2771scheduler, and will be sent as fast as the physical link allows.
2772Any additional data will be transmitted at the rate specified
2773by the
2774.Nm pipe
2775bandwidth.
2776The burst size depends on how long the pipe has been idle;
2777the effective burst size is calculated as follows:
2778MAX(
2779.Ar size
2780,
2781.Nm bw
2782* pipe_idle_time).
2783.Pp
2784.It Cm profile Ar filename
2785A file specifying the additional overhead incurred in the transmission
2786of a packet on the link.
2787.Pp
2788Some link types introduce extra delays in the transmission
2789of a packet, e.g., because of MAC level framing, contention on
2790the use of the channel, MAC level retransmissions and so on.
2791From our point of view, the channel is effectively unavailable
2792for this extra time, which is constant or variable depending
2793on the link type.
2794Additionally, packets may be dropped after this
2795time (e.g., on a wireless link after too many retransmissions).
2796We can model the additional delay with an empirical curve
2797that represents its distribution.
2798.Bd -literal -offset indent
2799      cumulative probability
2800      1.0 ^
2801          |
2802      L   +-- loss-level          x
2803          |                 ******
2804          |                *
2805          |           *****
2806          |          *
2807          |        **
2808          |       *
2809          +-------*------------------->
2810                      delay
2811.Ed
2812The empirical curve may have both vertical and horizontal lines.
2813Vertical lines represent constant delay for a range of
2814probabilities.
2815Horizontal lines correspond to a discontinuity in the delay
2816distribution: the pipe will use the largest delay for a
2817given probability.
2818.Pp
2819The file format is the following, with whitespace acting as
2820a separator and '#' indicating the beginning a comment:
2821.Bl -tag -width indent
2822.It Cm name Ar identifier
2823optional name (listed by "dnctl pipe show")
2824to identify the delay distribution;
2825.It Cm bw Ar value
2826the bandwidth used for the pipe.
2827If not specified here, it must be present
2828explicitly as a configuration parameter for the pipe;
2829.It Cm loss-level Ar L
2830the probability above which packets are lost.
2831(0.0 <= L <= 1.0, default 1.0 i.e., no loss);
2832.It Cm samples Ar N
2833the number of samples used in the internal
2834representation of the curve (2..1024; default 100);
2835.It Cm "delay prob" | "prob delay"
2836One of these two lines is mandatory and defines
2837the format of the following lines with data points.
2838.It Ar XXX Ar YYY
28392 or more lines representing points in the curve,
2840with either delay or probability first, according
2841to the chosen format.
2842The unit for delay is milliseconds.
2843Data points do not need to be sorted.
2844Also, the number of actual lines can be different
2845from the value of the "samples" parameter:
2846.Nm
2847utility will sort and interpolate
2848the curve as needed.
2849.El
2850.Pp
2851Example of a profile file:
2852.Bd -literal -offset indent
2853name    bla_bla_bla
2854samples 100
2855loss-level    0.86
2856prob    delay
28570       200	# minimum overhead is 200ms
28580.5     200
28590.5     300
28600.8     1000
28610.9     1300
28621       1300
2863#configuration file end
2864.Ed
2865.El
2866.Pp
2867The following parameters can be configured for a queue:
2868.Pp
2869.Bl -tag -width indent -compact
2870.It Cm pipe Ar pipe_nr
2871Connects a queue to the specified pipe.
2872Multiple queues (with the same or different weights) can be connected to
2873the same pipe, which specifies the aggregate rate for the set of queues.
2874.Pp
2875.It Cm weight Ar weight
2876Specifies the weight to be used for flows matching this queue.
2877The weight must be in the range 1..100, and defaults to 1.
2878.El
2879.Pp
2880The following case-insensitive parameters can be configured for a
2881scheduler:
2882.Pp
2883.Bl -tag -width indent -compact
2884.It Cm type Ar {fifo | wf2q+ | rr | qfq | fq_codel | fq_pie}
2885specifies the scheduling algorithm to use.
2886.Bl -tag -width indent -compact
2887.It Cm fifo
2888is just a FIFO scheduler (which means that all packets
2889are stored in the same queue as they arrive to the scheduler).
2890FIFO has O(1) per-packet time complexity, with very low
2891constants (estimate 60-80ns on a 2GHz desktop machine)
2892but gives no service guarantees.
2893.It Cm wf2q+
2894implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2895algorithm which permits flows to share bandwidth according to
2896their weights.
2897Note that weights are not priorities; even a flow
2898with a minuscule weight will never starve.
2899WF2Q+ has O(log N) per-packet processing cost, where N is the number
2900of flows, and is the default algorithm used by previous versions
2901dummynet's queues.
2902.It Cm rr
2903implements the Deficit Round Robin algorithm, which has O(1) processing
2904costs (roughly, 100-150ns per packet)
2905and permits bandwidth allocation according to weights, but
2906with poor service guarantees.
2907.It Cm qfq
2908implements the QFQ algorithm, which is a very fast variant of
2909WF2Q+, with similar service guarantees and O(1) processing
2910costs (roughly, 200-250ns per packet).
2911.It Cm fq_codel
2912implements the FQ-CoDel (FlowQueue-CoDel) scheduler/AQM algorithm, which
2913uses a modified Deficit Round Robin scheduler to manage two lists of sub-queues
2914(old sub-queues and new sub-queues) for providing brief periods of priority to
2915lightweight or short burst flows.
2916By default, the total number of sub-queues is 1024.
2917FQ-CoDel's internal, dynamically
2918created sub-queues are controlled by separate instances of CoDel AQM.
2919.It Cm fq_pie
2920implements the FQ-PIE (FlowQueue-PIE) scheduler/AQM algorithm, which similar to
2921.Cm fq_codel
2922but uses per sub-queue PIE AQM instance to control the queue delay.
2923.El
2924.Pp
2925.Cm fq_codel
2926inherits AQM parameters and options from
2927.Cm codel
2928(see below), and
2929.Cm fq_pie
2930inherits AQM parameters and options from
2931.Cm pie
2932(see below).
2933Additionally, both of
2934.Cm fq_codel
2935and
2936.Cm fq_pie
2937have shared scheduler parameters which are:
2938.Bl -tag -width indent
2939.It Cm quantum
2940.Ar m
2941specifies the quantum (credit) of the scheduler.
2942.Ar m
2943is the number of bytes a queue can serve before being moved to the tail
2944of old queues list.
2945The default is 1514 bytes, and the maximum acceptable value
2946is 9000 bytes.
2947.It Cm limit
2948.Ar m
2949specifies the hard size limit (in unit of packets) of all queues managed by an
2950instance of the scheduler.
2951The default value of
2952.Ar m
2953is 10240 packets, and the maximum acceptable value is 20480 packets.
2954.It Cm flows
2955.Ar m
2956specifies the total number of flow queues (sub-queues) that fq_*
2957creates and manages.
2958By default, 1024 sub-queues are created when an instance
2959of the fq_{codel/pie} scheduler is created.
2960The maximum acceptable value is
296165536.
2962.El
2963.Pp
2964Note that any token after
2965.Cm fq_codel
2966or
2967.Cm fq_pie
2968is considered a parameter for fq_{codel/pie}.
2969So, ensure all scheduler
2970configuration options not related to fq_{codel/pie} are written before
2971.Cm fq_codel/fq_pie
2972tokens.
2973.El
2974.Pp
2975In addition to the type, all parameters allowed for a pipe can also
2976be specified for a scheduler.
2977.Pp
2978Finally, the following parameters can be configured for both
2979pipes and queues:
2980.Pp
2981.Bl -tag -width XXXX -compact
2982.It Cm buckets Ar hash-table-size
2983Specifies the size of the hash table used for storing the
2984various queues.
2985Default value is 64 controlled by the
2986.Xr sysctl 8
2987variable
2988.Va net.inet.ip.dummynet.hash_size ,
2989allowed range is 16 to 65536.
2990.Pp
2991.It Cm mask Ar mask-specifier
2992Packets sent to a given pipe or queue by an
2993.Nm
2994rule can be further classified into multiple flows, each of which is then
2995sent to a different
2996.Em dynamic
2997pipe or queue.
2998A flow identifier is constructed by masking the IP addresses,
2999ports and protocol types as specified with the
3000.Cm mask
3001options in the configuration of the pipe or queue.
3002For each different flow identifier, a new pipe or queue is created
3003with the same parameters as the original object, and matching packets
3004are sent to it.
3005.Pp
3006Thus, when
3007.Em dynamic pipes
3008are used, each flow will get the same bandwidth as defined by the pipe,
3009whereas when
3010.Em dynamic queues
3011are used, each flow will share the parent's pipe bandwidth evenly
3012with other flows generated by the same queue (note that other queues
3013with different weights might be connected to the same pipe).
3014.br
3015Available mask specifiers are a combination of one or more of the following:
3016.Pp
3017.Cm dst-ip Ar mask ,
3018.Cm dst-ip6 Ar mask ,
3019.Cm src-ip Ar mask ,
3020.Cm src-ip6 Ar mask ,
3021.Cm dst-port Ar mask ,
3022.Cm src-port Ar mask ,
3023.Cm flow-id Ar mask ,
3024.Cm proto Ar mask
3025or
3026.Cm all ,
3027.Pp
3028where the latter means all bits in all fields are significant.
3029.Pp
3030.It Cm noerror
3031When a packet is dropped by a
3032.Nm dummynet
3033queue or pipe, the error
3034is normally reported to the caller routine in the kernel, in the
3035same way as it happens when a device queue fills up.
3036Setting this
3037option reports the packet as successfully delivered, which can be
3038needed for some experimental setups where you want to simulate
3039loss or congestion at a remote router.
3040.Pp
3041.It Cm plr Ar packet-loss-rate
3042.It Cm plr Ar K,p,H,r
3043Packet loss rate.
3044Argument
3045.Ar packet-loss-rate
3046is a floating-point number between 0 and 1, with 0 meaning no
3047loss, 1 meaning 100% loss.
3048.Pp
3049When invoked with four arguments, the simple Gilbert-Elliott
3050channel model with two states (Good and Bad) is used.
3051.Bd -literal -offset indent
3052                        r
3053               .----------------.
3054               v                |
3055         .------------.   .------------.
3056         |     G      |   |     B      |
3057         |  drop (K)  |   |  drop (H)  |
3058         '------------'   '------------'
3059               |                ^
3060               '----------------'
3061                        p
3062
3063.Ed
3064This has the associated probabilities
3065.Po Ar K
3066and
3067.Ar H Pc
3068for the loss probability. This is different from the literature,
3069where this model is described with probabilities of successful
3070transmission k and h. However, converting from literature is
3071easy:
3072.Pp
3073K = 1 - k ; H = 1 - h
3074.Pp
3075This is to retain consistency within the interface and allow the
3076quick re-use of loss probability when giving only a single argument.
3077In addition the state change probabilities
3078.Po Ar p
3079and
3080.Ar r Pc
3081are given.
3082All of the above probabilities are internally represented on 31 bits.
3083.Pp
3084.It Cm queue Brq Ar slots | size Ns Cm Kbytes
3085Queue size, in
3086.Ar slots
3087or
3088.Cm KBytes .
3089Default value is 50 slots, which
3090is the typical queue size for Ethernet devices.
3091Note that for slow speed links you should keep the queue
3092size short or your traffic might be affected by a significant
3093queueing delay.
3094E.g., 50 max-sized Ethernet packets (1500 bytes) mean 600Kbit
3095or 20s of queue on a 30Kbit/s pipe.
3096Even worse effects can result if you get packets from an
3097interface with a much larger MTU, e.g.\& the loopback interface
3098with its 16KB packets.
3099The
3100.Xr sysctl 8
3101variables
3102.Em net.inet.ip.dummynet.pipe_byte_limit
3103and
3104.Em net.inet.ip.dummynet.pipe_slot_limit
3105control the maximum lengths that can be specified.
3106.Pp
3107.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
3108[ecn]
3109Make use of the RED (Random Early Detection) queue management algorithm.
3110.Ar w_q
3111and
3112.Ar max_p
3113are floating
3114point numbers between 0 and 1 (inclusive), while
3115.Ar min_th
3116and
3117.Ar max_th
3118are integer numbers specifying thresholds for queue management
3119(thresholds are computed in bytes if the queue has been defined
3120in bytes, in slots otherwise).
3121The two parameters can also be of the same value if needed.
3122The
3123.Nm dummynet
3124also supports the gentle RED variant (gred) and ECN (Explicit Congestion
3125Notification) as optional.
3126Three
3127.Xr sysctl 8
3128variables can be used to control the RED behaviour:
3129.Bl -tag -width indent
3130.It Va net.inet.ip.dummynet.red_lookup_depth
3131specifies the accuracy in computing the average queue
3132when the link is idle (defaults to 256, must be greater than zero)
3133.It Va net.inet.ip.dummynet.red_avg_pkt_size
3134specifies the expected average packet size (defaults to 512, must be
3135greater than zero)
3136.It Va net.inet.ip.dummynet.red_max_pkt_size
3137specifies the expected maximum packet size, only used when queue
3138thresholds are in bytes (defaults to 1500, must be greater than zero).
3139.El
3140.Pp
3141.It Cm codel Oo Cm target Ar time Oc Oo Cm interval Ar time Oc Oo Cm ecn |
3142.Cm noecn Oc
3143Make use of the CoDel (Controlled-Delay) queue management algorithm.
3144.Ar time
3145is interpreted as milliseconds by default but seconds (s), milliseconds (ms) or
3146microseconds (us) can be specified instead.
3147CoDel drops or marks (ECN) packets
3148depending on packet sojourn time in the queue.
3149.Cm target
3150.Ar time
3151(5ms by default) is the minimum acceptable persistent queue delay that CoDel
3152allows.
3153CoDel does not drop packets directly after packets sojourn time becomes
3154higher than
3155.Cm target
3156.Ar time
3157but waits for
3158.Cm interval
3159.Ar time
3160(100ms default) before dropping.
3161.Cm interval
3162.Ar time
3163should be set to maximum RTT for all expected connections.
3164.Cm ecn
3165enables (disabled by default) packet marking (instead of dropping) for
3166ECN-enabled TCP flows when queue delay becomes high.
3167.Pp
3168Note that any token after
3169.Cm codel
3170is considered a parameter for CoDel.
3171So, ensure all pipe/queue
3172configuration options are written before
3173.Cm codel
3174token.
3175.Pp
3176The
3177.Xr sysctl 8
3178variables
3179.Va net.inet.ip.dummynet.codel.target
3180and
3181.Va net.inet.ip.dummynet.codel.interval
3182can be used to set CoDel default parameters.
3183.Pp
3184.It Cm pie Oo Cm target Ar time Oc Oo Cm tupdate Ar time Oc Oo
3185.Cm alpha Ar n Oc Oo Cm beta Ar n Oc Oo Cm max_burst Ar time Oc Oo
3186.Cm max_ecnth Ar n Oc Oo Cm ecn | Cm noecn Oc Oo Cm capdrop |
3187.Cm nocapdrop Oc Oo Cm drand | Cm nodrand Oc Oo Cm onoff
3188.Oc Oo Cm dre | Cm ts Oc
3189Make use of the PIE (Proportional Integral controller Enhanced) queue management
3190algorithm.
3191PIE drops or marks packets depending on a calculated drop probability during
3192en-queue process, with the aim of achieving high throughput while keeping queue
3193delay low.
3194At regular time intervals of
3195.Cm tupdate
3196.Ar time
3197(15ms by default) a background process (re)calculates the probability based on queue delay
3198deviations from
3199.Cm target
3200.Ar time
3201(15ms by default) and queue delay trends.
3202PIE approximates current queue
3203delay by using a departure rate estimation method, or (optionally) by using a
3204packet timestamp method similar to CoDel.
3205.Ar time
3206is interpreted as milliseconds by default but seconds (s), milliseconds (ms) or
3207microseconds (us) can be specified instead.
3208The other PIE parameters and options are as follows:
3209.Bl -tag -width indent
3210.It Cm alpha Ar n
3211.Ar n
3212is a floating point number between 0 and 7 which specifies the weight of queue
3213delay deviations that is used in drop probability calculation.
32140.125 is the default.
3215.It Cm beta Ar n
3216.Ar n
3217is a floating point number between 0 and 7 which specifies is the weight of queue
3218delay trend that is used in drop probability calculation.
32191.25 is the default.
3220.It Cm max_burst Ar time
3221The maximum period of time that PIE does not drop/mark packets.
3222150ms is the
3223default and 10s is the maximum value.
3224.It Cm max_ecnth Ar n
3225Even when ECN is enabled, PIE drops packets instead of marking them when drop
3226probability becomes higher than ECN probability threshold
3227.Cm max_ecnth Ar n
3228, the default is 0.1 (i.e 10%) and 1 is the maximum value.
3229.It Cm ecn | noecn
3230enable or disable ECN marking for ECN-enabled TCP flows.
3231Disabled by default.
3232.It Cm capdrop | nocapdrop
3233enable or disable cap drop adjustment.
3234Cap drop adjustment is enabled by default.
3235.It Cm drand | nodrand
3236enable or disable drop probability de-randomisation.
3237De-randomisation eliminates
3238the problem of dropping packets too close or too far.
3239De-randomisation is enabled by default.
3240.It Cm onoff
3241enable turning PIE on and off depending on queue load.
3242If this option is enabled,
3243PIE turns on when over 1/3 of queue becomes full.
3244This option is disabled by
3245default.
3246.It Cm dre | ts
3247Calculate queue delay using departure rate estimation
3248.Cm dre
3249or timestamps
3250.Cm ts .
3251.Cm dre
3252is used by default.
3253.El
3254.Pp
3255Note that any token after
3256.Cm pie
3257is considered a parameter for PIE.
3258So ensure all pipe/queue
3259the configuration options are written before
3260.Cm pie
3261token.
3262.Xr sysctl 8
3263variables can be used to control the
3264.Cm pie
3265default parameters.
3266See the
3267.Sx SYSCTL VARIABLES
3268section for more details.
3269.El
3270.Pp
3271When used with IPv6 data,
3272.Nm dummynet
3273currently has several limitations.
3274Information necessary to route link-local packets to an
3275interface is not available after processing by
3276.Nm dummynet
3277so those packets are dropped in the output path.
3278Care should be taken to ensure that link-local packets are not passed to
3279.Nm dummynet .
3280.Sh CHECKLIST
3281Here are some important points to consider when designing your
3282rules:
3283.Bl -bullet
3284.It
3285Remember that you filter both packets going
3286.Cm in
3287and
3288.Cm out .
3289Most connections need packets going in both directions.
3290.It
3291Remember to test very carefully.
3292It is a good idea to be near the console when doing this.
3293If you cannot be near the console,
3294use an auto-recovery script such as the one in
3295.Pa /usr/share/examples/ipfw/change_rules.sh .
3296.It
3297Do not forget the loopback interface.
3298.El
3299.Sh FINE POINTS
3300.Bl -bullet
3301.It
3302There are circumstances where fragmented datagrams are unconditionally
3303dropped.
3304TCP packets are dropped if they do not contain at least 20 bytes of
3305TCP header, UDP packets are dropped if they do not contain a full 8
3306byte UDP header, and ICMP packets are dropped if they do not contain
33074 bytes of ICMP header, enough to specify the ICMP type, code, and
3308checksum.
3309These packets are simply logged as
3310.Dq pullup failed
3311since there may not be enough good data in the packet to produce a
3312meaningful log entry.
3313.It
3314Another type of packet is unconditionally dropped, a TCP packet with a
3315fragment offset of one.
3316This is a valid packet, but it only has one use, to try
3317to circumvent firewalls.
3318When logging is enabled, these packets are
3319reported as being dropped by rule -1.
3320.It
3321If you are logged in over a network, loading the
3322.Xr kld 4
3323version of
3324.Nm
3325is probably not as straightforward as you would think.
3326The following command line is recommended:
3327.Bd -literal -offset indent
3328kldload ipfw && \e
3329ipfw add 32000 allow ip from any to any
3330.Ed
3331.Pp
3332Along the same lines, doing an
3333.Bd -literal -offset indent
3334ipfw flush
3335.Ed
3336.Pp
3337in similar surroundings is also a bad idea.
3338.It
3339The
3340.Nm
3341filter list may not be modified if the system security level
3342is set to 3 or higher
3343(see
3344.Xr init 8
3345for information on system security levels).
3346.El
3347.Sh PACKET DIVERSION
3348A
3349.Xr divert 4
3350socket bound to the specified port will receive all packets
3351diverted to that port.
3352If no socket is bound to the destination port, or if the divert module is
3353not loaded, or if the kernel was not compiled with divert socket support,
3354the packets are dropped.
3355.Sh NETWORK ADDRESS TRANSLATION (NAT)
3356.Nm
3357support in-kernel NAT using the kernel version of
3358.Xr libalias 3 .
3359The kernel module
3360.Cm ipfw_nat
3361should be loaded or kernel should have
3362.Cm options IPFIREWALL_NAT
3363to be able use NAT.
3364.Pp
3365The nat configuration command is the following:
3366.Bd -ragged -offset indent
3367.Bk -words
3368.Cm nat
3369.Ar nat_number
3370.Cm config
3371.Ar nat-configuration
3372.Ek
3373.Ed
3374.Pp
3375The following parameters can be configured:
3376.Bl -tag -width indent
3377.It Cm ip Ar ip_address
3378Define an ip address to use for aliasing.
3379.It Cm if Ar nic
3380Use ip address of NIC for aliasing, dynamically changing
3381it if NIC's ip address changes.
3382.It Cm log
3383Enable logging on this nat instance.
3384.It Cm deny_in
3385Deny any incoming connection from outside world.
3386.It Cm same_ports
3387Try to leave the alias port numbers unchanged from
3388the actual local port numbers.
3389.It Cm unreg_only
3390Traffic on the local network not originating from a RFC 1918
3391unregistered address spaces will be ignored.
3392.It Cm unreg_cgn
3393Like unreg_only, but includes the RFC 6598 (Carrier Grade NAT)
3394address range.
3395.It Cm reset
3396Reset table of the packet aliasing engine on address change.
3397.It Cm reverse
3398Reverse the way libalias handles aliasing.
3399.It Cm proxy_only
3400Obey transparent proxy rules only, packet aliasing is not performed.
3401.It Cm skip_global
3402Skip instance in case of global state lookup (see below).
3403.It Cm port_range Ar lower-upper
3404Set the aliasing ports between the ranges given.
3405Upper port has to be greater than lower.
3406.El
3407.Pp
3408Some special values can be supplied instead of
3409.Va nat_number
3410in nat rule actions:
3411.Bl -tag -width indent
3412.It Cm global
3413Looks up translation state in all configured nat instances.
3414If an entry is found, packet is aliased according to that entry.
3415If no entry was found in any of the instances, packet is passed unchanged,
3416and no new entry will be created.
3417See section
3418.Sx MULTIPLE INSTANCES
3419in
3420.Xr natd 8
3421for more information.
3422.It Cm tablearg
3423Uses argument supplied in lookup table.
3424See
3425.Sx LOOKUP TABLES
3426section below for more information on lookup tables.
3427.El
3428.Pp
3429To let the packet continue after being (de)aliased, set the sysctl variable
3430.Va net.inet.ip.fw.one_pass
3431to 0.
3432For more information about aliasing modes, refer to
3433.Xr libalias 3 .
3434See Section
3435.Sx EXAMPLES
3436for some examples of nat usage.
3437.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
3438Redirect and LSNAT support follow closely the syntax used in
3439.Xr natd 8 .
3440See Section
3441.Sx EXAMPLES
3442for some examples on how to do redirect and lsnat.
3443.Ss SCTP NAT SUPPORT
3444SCTP nat can be configured in a similar manner to TCP through the
3445.Nm
3446command line tool.
3447The main difference is that
3448.Nm sctp nat
3449does not do port translation.
3450Since the local and global side ports will be the same,
3451there is no need to specify both.
3452Ports are redirected as follows:
3453.Bd -ragged -offset indent
3454.Bk -words
3455.Cm nat
3456.Ar nat_number
3457.Cm config if
3458.Ar nic
3459.Cm redirect_port sctp
3460.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
3461.Ek
3462.Ed
3463.Pp
3464Most
3465.Nm sctp nat
3466configuration can be done in real-time through the
3467.Xr sysctl 8
3468interface.
3469All may be changed dynamically, though the hash_table size will only
3470change for new
3471.Nm nat
3472instances.
3473See
3474.Sx SYSCTL VARIABLES
3475for more info.
3476.Sh IPv6/IPv4 NETWORK ADDRESS AND PROTOCOL TRANSLATION
3477.Ss Stateful translation
3478.Nm
3479supports in-kernel IPv6/IPv4 network address and protocol translation.
3480Stateful NAT64 translation allows IPv6-only clients to contact IPv4 servers
3481using unicast TCP, UDP or ICMP protocols.
3482One or more IPv4 addresses assigned to a stateful NAT64 translator are shared
3483among several IPv6-only clients.
3484When stateful NAT64 is used in conjunction with DNS64, no changes are usually
3485required in the IPv6 client or the IPv4 server.
3486The kernel module
3487.Cm ipfw_nat64
3488should be loaded or kernel should have
3489.Cm options IPFIREWALL_NAT64
3490to be able use stateful NAT64 translator.
3491.Pp
3492Stateful NAT64 uses a bunch of memory for several types of objects.
3493When IPv6 client initiates connection, NAT64 translator creates a host entry
3494in the states table.
3495Each host entry uses preallocated IPv4 alias entry.
3496Each alias entry has a number of ports group entries allocated on demand.
3497Ports group entries contains connection state entries.
3498There are several options to control limits and lifetime for these objects.
3499.Pp
3500NAT64 translator follows RFC7915 when does ICMPv6/ICMP translation,
3501unsupported message types will be silently dropped.
3502IPv6 needs several ICMPv6 message types to be explicitly allowed for correct
3503operation.
3504Make sure that ND6 neighbor solicitation (ICMPv6 type 135) and neighbor
3505advertisement (ICMPv6 type 136) messages will not be handled by translation
3506rules.
3507.Pp
3508After translation NAT64 translator by default sends packets through
3509corresponding netisr queue.
3510Thus translator host should be configured as IPv4 and IPv6 router.
3511Also this means, that a packet is handled by firewall twice.
3512First time an original packet is handled and consumed by translator,
3513and then it is handled again as translated packet.
3514This behavior can be changed by sysctl variable
3515.Va net.inet.ip.fw.nat64_direct_output .
3516Also translated packet can be tagged using
3517.Cm tag
3518rule action, and then matched by
3519.Cm tagged
3520opcode to avoid loops and extra overhead.
3521.Pp
3522The stateful NAT64 configuration command is the following:
3523.Bd -ragged -offset indent
3524.Bk -words
3525.Cm nat64lsn
3526.Ar name
3527.Cm create
3528.Ar create-options
3529.Ek
3530.Ed
3531.Pp
3532The following parameters can be configured:
3533.Bl -tag -width indent
3534.It Cm prefix4 Ar ipv4_prefix/plen
3535The IPv4 prefix with mask defines the pool of IPv4 addresses used as
3536source address after translation.
3537Stateful NAT64 module translates IPv6 source address of client to one
3538IPv4 address from this pool.
3539Note that incoming IPv4 packets that don't have corresponding state entry
3540in the states table will be dropped by translator.
3541Make sure that translation rules handle packets, destined to configured prefix.
3542.It Cm prefix6 Ar ipv6_prefix/length
3543The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3544to represent IPv4 addresses.
3545This IPv6 prefix should be configured in DNS64.
3546The translator implementation follows RFC6052, that restricts the length of
3547prefixes to one of following: 32, 40, 48, 56, 64, or 96.
3548The Well-Known IPv6 Prefix 64:ff9b:: must be 96 bits long.
3549The special
3550.Ar ::/length
3551prefix can be used to handle several IPv6 prefixes with one NAT64 instance.
3552The NAT64 instance will determine a destination IPv4 address from prefix
3553.Ar length .
3554.It Cm states_chunks Ar number
3555The number of states chunks in single ports group.
3556Each ports group by default can keep 64 state entries in single chunk.
3557The above value affects the maximum number of states that can be associated with single IPv4 alias address and port.
3558The value must be power of 2, and up to 128.
3559.It Cm host_del_age Ar seconds
3560The number of seconds until the host entry for a IPv6 client will be deleted
3561and all its resources will be released due to inactivity.
3562Default value is
3563.Ar 3600 .
3564.It Cm pg_del_age Ar seconds
3565The number of seconds until a ports group with unused state entries will
3566be released.
3567Default value is
3568.Ar 900 .
3569.It Cm tcp_syn_age Ar seconds
3570The number of seconds while a state entry for TCP connection with only SYN
3571sent will be kept.
3572If TCP connection establishing will not be finished,
3573state entry will be deleted.
3574Default value is
3575.Ar 10 .
3576.It Cm tcp_est_age Ar seconds
3577The number of seconds while a state entry for established TCP connection
3578will be kept.
3579Default value is
3580.Ar 7200 .
3581.It Cm tcp_close_age Ar seconds
3582The number of seconds while a state entry for closed TCP connection
3583will be kept.
3584Keeping state entries for closed connections is needed, because IPv4 servers
3585typically keep closed connections in a TIME_WAIT state for a several minutes.
3586Since translator's IPv4 addresses are shared among all IPv6 clients,
3587new connections from the same addresses and ports may be rejected by server,
3588because these connections are still in a TIME_WAIT state.
3589Keeping them in translator's state table protects from such rejects.
3590Default value is
3591.Ar 180 .
3592.It Cm udp_age Ar seconds
3593The number of seconds while translator keeps state entry in a waiting for
3594reply to the sent UDP datagram.
3595Default value is
3596.Ar 120 .
3597.It Cm icmp_age Ar seconds
3598The number of seconds while translator keeps state entry in a waiting for
3599reply to the sent ICMP message.
3600Default value is
3601.Ar 60 .
3602.It Cm log
3603Turn on logging of all handled packets via BPF through
3604.Ar ipfwlog0
3605interface.
3606.Ar ipfwlog0
3607is a pseudo interface and can be created after a boot manually with
3608.Cm ifconfig
3609command.
3610Note that it has different purpose than
3611.Ar ipfw0
3612interface.
3613Translators sends to BPF an additional information with each packet.
3614With
3615.Cm tcpdump
3616you are able to see each handled packet before and after translation.
3617.It Cm -log
3618Turn off logging of all handled packets via BPF.
3619.It Cm allow_private
3620Turn on processing private IPv4 addresses.
3621By default IPv6 packets with destinations mapped to private address ranges
3622defined by RFC1918 are not processed.
3623.It Cm -allow_private
3624Turn off private address handling in
3625.Nm nat64
3626instance.
3627.El
3628.Pp
3629To inspect a states table of stateful NAT64 the following command can be used:
3630.Bd -ragged -offset indent
3631.Bk -words
3632.Cm nat64lsn
3633.Ar name
3634.Cm show Cm states
3635.Ek
3636.Ed
3637.Pp
3638Stateless NAT64 translator doesn't use a states table for translation
3639and converts IPv4 addresses to IPv6 and vice versa solely based on the
3640mappings taken from configured lookup tables.
3641Since a states table doesn't used by stateless translator,
3642it can be configured to pass IPv4 clients to IPv6-only servers.
3643.Pp
3644The stateless NAT64 configuration command is the following:
3645.Bd -ragged -offset indent
3646.Bk -words
3647.Cm nat64stl
3648.Ar name
3649.Cm create
3650.Ar create-options
3651.Ek
3652.Ed
3653.Pp
3654The following parameters can be configured:
3655.Bl -tag -width indent
3656.It Cm prefix6 Ar ipv6_prefix/length
3657The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3658to represent IPv4 addresses.
3659This IPv6 prefix should be configured in DNS64.
3660.It Cm table4 Ar table46
3661The lookup table
3662.Ar table46
3663contains mapping how IPv4 addresses should be translated to IPv6 addresses.
3664.It Cm table6 Ar table64
3665The lookup table
3666.Ar table64
3667contains mapping how IPv6 addresses should be translated to IPv4 addresses.
3668.It Cm log
3669Turn on logging of all handled packets via BPF through
3670.Ar ipfwlog0
3671interface.
3672.It Cm -log
3673Turn off logging of all handled packets via BPF.
3674.It Cm allow_private
3675Turn on processing private IPv4 addresses.
3676By default IPv6 packets with destinations mapped to private address ranges
3677defined by RFC1918 are not processed.
3678.It Cm -allow_private
3679Turn off private address handling in
3680.Nm nat64
3681instance.
3682.El
3683.Pp
3684Note that the behavior of stateless translator with respect to not matched
3685packets differs from stateful translator.
3686If corresponding addresses was not found in the lookup tables, the packet
3687will not be dropped and the search continues.
3688.Ss XLAT464 CLAT translation
3689XLAT464 CLAT NAT64 translator implements client-side stateless translation as
3690defined in RFC6877 and is very similar to statless NAT64 translator
3691explained above.
3692Instead of lookup tables it uses one-to-one mapping between IPv4 and IPv6
3693addresses using configured prefixes.
3694This mode can be used as a replacement of DNS64 service for applications
3695that are not using it (e.g. VoIP) allowing them to access IPv4-only Internet
3696over IPv6-only networks with help of remote NAT64 translator.
3697.Pp
3698The CLAT NAT64 configuration command is the following:
3699.Bd -ragged -offset indent
3700.Bk -words
3701.Cm nat64clat
3702.Ar name
3703.Cm create
3704.Ar create-options
3705.Ek
3706.Ed
3707.Pp
3708The following parameters can be configured:
3709.Bl -tag -width indent
3710.It Cm clat_prefix Ar ipv6_prefix/length
3711The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3712to represent source IPv4 addresses.
3713.It Cm plat_prefix Ar ipv6_prefix/length
3714The IPv6 prefix defines IPv4-embedded IPv6 addresses used by translator
3715to represent destination IPv4 addresses.
3716This IPv6 prefix should be configured on a remote NAT64 translator.
3717.It Cm log
3718Turn on logging of all handled packets via BPF through
3719.Ar ipfwlog0
3720interface.
3721.It Cm -log
3722Turn off logging of all handled packets via BPF.
3723.It Cm allow_private
3724Turn on processing private IPv4 addresses.
3725By default
3726.Nm nat64clat
3727instance will not process IPv4 packets with destination address from private
3728ranges as defined in RFC1918.
3729.It Cm -allow_private
3730Turn off private address handling in
3731.Nm nat64clat
3732instance.
3733.El
3734.Pp
3735Note that the behavior of CLAT translator with respect to not matched
3736packets differs from stateful translator.
3737If corresponding addresses were not matched against prefixes configured,
3738the packet will not be dropped and the search continues.
3739.Sh IPv6-to-IPv6 NETWORK PREFIX TRANSLATION (NPTv6)
3740.Nm
3741supports in-kernel IPv6-to-IPv6 network prefix translation as described
3742in RFC6296.
3743The kernel module
3744.Cm ipfw_nptv6
3745should be loaded or kernel should has
3746.Cm options IPFIREWALL_NPTV6
3747to be able use NPTv6 translator.
3748.Pp
3749The NPTv6 configuration command is the following:
3750.Bd -ragged -offset indent
3751.Bk -words
3752.Cm nptv6
3753.Ar name
3754.Cm create
3755.Ar create-options
3756.Ek
3757.Ed
3758.Pp
3759The following parameters can be configured:
3760.Bl -tag -width indent
3761.It Cm int_prefix Ar ipv6_prefix
3762IPv6 prefix used in internal network.
3763NPTv6 module translates source address when it matches this prefix.
3764.It Cm ext_prefix Ar ipv6_prefix
3765IPv6 prefix used in external network.
3766NPTv6 module translates destination address when it matches this prefix.
3767.It Cm ext_if Ar nic
3768The NPTv6 module will use first global IPv6 address from interface
3769.Ar nic
3770as external prefix.
3771It can be useful when IPv6 prefix of external network is dynamically obtained.
3772.Cm ext_prefix
3773and
3774.Cm ext_if
3775options are mutually exclusive.
3776.It Cm prefixlen Ar length
3777The length of specified IPv6 prefixes.
3778It must be in range from 8 to 64.
3779.El
3780.Pp
3781Note that the prefix translation rules are silently ignored when IPv6 packet
3782forwarding is disabled.
3783To enable the packet forwarding, set the sysctl variable
3784.Va net.inet6.ip6.forwarding
3785to 1.
3786.Pp
3787To let the packet continue after being translated, set the sysctl variable
3788.Va net.inet.ip.fw.one_pass
3789to 0.
3790.Sh LOADER TUNABLES
3791Tunables can be set in
3792.Xr loader 8
3793prompt,
3794.Xr loader.conf 5
3795or
3796.Xr kenv 1
3797before ipfw module gets loaded.
3798.Bl -tag -width indent
3799.It Va net.inet.ip.fw.enable : No 1
3800Enables the firewall.
3801Setting this variable to 0 lets you run your machine without
3802firewall even if compiled in.
3803.It Va net.inet6.ip6.fw.enable : No 1
3804provides the same functionality as above for the IPv6 case.
3805.It Va net.link.ether.ipfw : No 0
3806Controls whether layer2 packets are passed to
3807.Nm .
3808Default is no.
3809.It Va net.inet.ip.fw.default_to_accept : No 0
3810Defines ipfw last rule behavior.
3811This value overrides
3812.Cd "options IPFW_DEFAULT_TO_(ACCEPT|DENY)"
3813from kernel configuration file.
3814.It Va net.inet.ip.fw.tables_max : No 128
3815Defines number of tables available in ipfw.
3816Number cannot exceed 65534.
3817.El
3818.Sh SYSCTL VARIABLES
3819A set of
3820.Xr sysctl 8
3821variables controls the behaviour of the firewall and
3822associated modules
3823.Pq Nm dummynet , bridge , sctp nat .
3824These are shown below together with their default value
3825(but always check with the
3826.Xr sysctl 8
3827command what value is actually in use) and meaning:
3828.Bl -tag -width indent
3829.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip : No 0
3830Defines how the
3831.Nm nat
3832responds to receipt of global OOTB ASCONF-AddIP:
3833.Bl -tag -width indent
3834.It Cm 0
3835No response (unless a partially matching association exists -
3836ports and vtags match but global address does not)
3837.It Cm 1
3838.Nm nat
3839will accept and process all OOTB global AddIP messages.
3840.El
3841.Pp
3842Option 1 should never be selected as this forms a security risk.
3843An attacker can
3844establish multiple fake associations by sending AddIP messages.
3845.It Va net.inet.ip.alias.sctp.chunk_proc_limit : No 5
3846Defines the maximum number of chunks in an SCTP packet that will be
3847parsed for a
3848packet that matches an existing association.
3849This value is enforced to be greater or equal than
3850.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
3851A high value is
3852a DoS risk yet setting too low a value may result in
3853important control chunks in
3854the packet not being located and parsed.
3855.It Va net.inet.ip.alias.sctp.error_on_ootb : No 1
3856Defines when the
3857.Nm nat
3858responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
3859An OOTB packet is a packet that arrives with no existing association
3860registered in the
3861.Nm nat
3862and is not an INIT or ASCONF-AddIP packet:
3863.Bl -tag -width indent
3864.It Cm 0
3865ErrorM is never sent in response to OOTB packets.
3866.It Cm 1
3867ErrorM is only sent to OOTB packets received on the local side.
3868.It Cm 2
3869ErrorM is sent to the local side and on the global side ONLY if there is a
3870partial match (ports and vtags match but the source global IP does not).
3871This value is only useful if the
3872.Nm nat
3873is tracking global IP addresses.
3874.It Cm 3
3875ErrorM is sent in response to all OOTB packets on both
3876the local and global side
3877(DoS risk).
3878.El
3879.Pp
3880At the moment the default is 0, since the ErrorM packet is not yet
3881supported by most SCTP stacks.
3882When it is supported, and if not tracking
3883global addresses, we recommend setting this value to 1 to allow
3884multi-homed local hosts to function with the
3885.Nm nat .
3886To track global addresses, we recommend setting this value to 2 to
3887allow global hosts to be informed when they need to (re)send an
3888ASCONF-AddIP.
3889Value 3 should never be chosen (except for debugging) as the
3890.Nm nat
3891will respond to all OOTB global packets (a DoS risk).
3892.It Va net.inet.ip.alias.sctp.hashtable_size : No 2003
3893Size of hash tables used for
3894.Nm nat
3895lookups (100 < prime_number > 1000001).
3896This value sets the
3897.Nm hash table
3898size for any future created
3899.Nm nat
3900instance and therefore must be set prior to creating a
3901.Nm nat
3902instance.
3903The table sizes may be changed to suit specific needs.
3904If there will be few
3905concurrent associations, and memory is scarce, you may make these smaller.
3906If there will be many thousands (or millions) of concurrent associations, you
3907should make these larger.
3908A prime number is best for the table size.
3909The sysctl
3910update function will adjust your input value to the next highest prime number.
3911.It Va net.inet.ip.alias.sctp.holddown_time : No 0
3912Hold association in table for this many seconds after receiving a
3913SHUTDOWN-COMPLETE.
3914This allows endpoints to correct shutdown gracefully if a
3915shutdown_complete is lost and retransmissions are required.
3916.It Va net.inet.ip.alias.sctp.init_timer : No 15
3917Timeout value while waiting for (INIT-ACK|AddIP-ACK).
3918This value cannot be 0.
3919.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit : No 2
3920Defines the maximum number of chunks in an SCTP packet that will be parsed when
3921no existing association exists that matches that packet.
3922Ideally this packet
3923will only be an INIT or ASCONF-AddIP packet.
3924A higher value may become a DoS
3925risk as malformed packets can consume processing resources.
3926.It Va net.inet.ip.alias.sctp.param_proc_limit : No 25
3927Defines the maximum number of parameters within a chunk that will be
3928parsed in a
3929packet.
3930As for other similar sysctl variables, larger values pose a DoS risk.
3931.It Va net.inet.ip.alias.sctp.log_level : No 0
3932Level of detail in the system log messages (0 \- minimal, 1 \- event,
39332 \- info, 3 \- detail, 4 \- debug, 5 \- max debug).
3934May be a good
3935option in high loss environments.
3936.It Va net.inet.ip.alias.sctp.shutdown_time : No 15
3937Timeout value while waiting for SHUTDOWN-COMPLETE.
3938This value cannot be 0.
3939.It Va net.inet.ip.alias.sctp.track_global_addresses : No 0
3940Enables/disables global IP address tracking within the
3941.Nm nat
3942and places an
3943upper limit on the number of addresses tracked for each association:
3944.Bl -tag -width indent
3945.It Cm 0
3946Global tracking is disabled
3947.It Cm >1
3948Enables tracking, the maximum number of addresses tracked for each
3949association is limited to this value
3950.El
3951.Pp
3952This variable is fully dynamic, the new value will be adopted for all newly
3953arriving associations, existing associations are treated
3954as they were previously.
3955Global tracking will decrease the number of collisions within the
3956.Nm nat
3957at a cost
3958of increased processing load, memory usage, complexity, and possible
3959.Nm nat
3960state
3961problems in complex networks with multiple
3962.Nm nats .
3963We recommend not tracking
3964global IP addresses, this will still result in a fully functional
3965.Nm nat .
3966.It Va net.inet.ip.alias.sctp.up_timer : No 300
3967Timeout value to keep an association up with no traffic.
3968This value cannot be 0.
3969.It Va net.inet.ip.dummynet.codel.interval : No 100000
3970Default
3971.Cm codel
3972AQM interval in microseconds.
3973The value must be in the range 1..5000000.
3974.It Va net.inet.ip.dummynet.codel.target : No 5000
3975Default
3976.Cm codel
3977AQM target delay time in microseconds (the minimum acceptable persistent queue
3978delay).
3979The value must be in the range 1..5000000.
3980.It Va net.inet.ip.dummynet.expire : No 1
3981Lazily delete dynamic pipes/queue once they have no pending traffic.
3982You can disable this by setting the variable to 0, in which case
3983the pipes/queues will only be deleted when the threshold is reached.
3984.It Va net.inet.ip.dummynet.fqcodel.flows : No 1024
3985Defines the default total number of flow queues (sub-queues) that
3986.Cm fq_codel
3987creates and manages.
3988The value must be in the range 1..65536.
3989.It Va net.inet.ip.dummynet.fqcodel.interval : No 100000
3990Default
3991.Cm fq_codel
3992scheduler/AQM interval in microseconds.
3993The value must be in the range 1..5000000.
3994.It Va net.inet.ip.dummynet.fqcodel.limit : No 10240
3995The default hard size limit (in unit of packet) of all queues managed by an
3996instance of the
3997.Cm fq_codel
3998scheduler.
3999The value must be in the range 1..20480.
4000.It Va net.inet.ip.dummynet.fqcodel.quantum : No 1514
4001The default quantum (credit) of the
4002.Cm fq_codel
4003in unit of byte.
4004The value must be in the range 1..9000.
4005.It Va net.inet.ip.dummynet.fqcodel.target : No 5000
4006Default
4007.Cm fq_codel
4008scheduler/AQM target delay time in microseconds (the minimum acceptable
4009persistent queue delay).
4010The value must be in the range 1..5000000.
4011.It Va net.inet.ip.dummynet.fqpie.alpha : No 125
4012The default
4013.Ar alpha
4014parameter (scaled by 1000) for
4015.Cm fq_pie
4016scheduler/AQM.
4017The value must be in the range 1..7000.
4018.It Va net.inet.ip.dummynet.fqpie.beta : No 1250
4019The default
4020.Ar beta
4021parameter (scaled by 1000) for
4022.Cm fq_pie
4023scheduler/AQM.
4024The value must be in the range 1..7000.
4025.It Va net.inet.ip.dummynet.fqpie.flows : No 1024
4026Defines the default total number of flow queues (sub-queues) that
4027.Cm fq_pie
4028creates and manages.
4029The value must be in the range 1..65536.
4030.It Va net.inet.ip.dummynet.fqpie.limit : No 10240
4031The default hard size limit (in unit of packet) of all queues managed by an
4032instance of the
4033.Cm fq_pie
4034scheduler.
4035The value must be in the range 1..20480.
4036.It Va net.inet.ip.dummynet.fqpie.max_burst : No 150000
4037The default maximum period of microseconds that
4038.Cm fq_pie
4039scheduler/AQM does not drop/mark packets.
4040The value must be in the range 1..10000000.
4041.It Va net.inet.ip.dummynet.fqpie.max_ecnth : No 99
4042The default maximum ECN probability threshold (scaled by 1000) for
4043.Cm fq_pie
4044scheduler/AQM.
4045The value must be in the range 1..7000.
4046.It Va net.inet.ip.dummynet.fqpie.quantum : No 1514
4047The default quantum (credit) of the
4048.Cm fq_pie
4049in unit of byte.
4050The value must be in the range 1..9000.
4051.It Va net.inet.ip.dummynet.fqpie.target : No 15000
4052The default
4053.Cm target
4054delay of the
4055.Cm fq_pie
4056in unit of microsecond.
4057The value must be in the range 1..5000000.
4058.It Va net.inet.ip.dummynet.fqpie.tupdate : No 15000
4059The default
4060.Cm tupdate
4061of the
4062.Cm fq_pie
4063in unit of microsecond.
4064The value must be in the range 1..5000000.
4065.It Va net.inet.ip.dummynet.hash_size : No 64
4066Default size of the hash table used for dynamic pipes/queues.
4067This value is used when no
4068.Cm buckets
4069option is specified when configuring a pipe/queue.
4070.It Va net.inet.ip.dummynet.io_fast : No 0
4071If set to a non-zero value,
4072the
4073.Dq fast
4074mode of
4075.Nm dummynet
4076operation (see above) is enabled.
4077.It Va net.inet.ip.dummynet.io_pkt
4078Number of packets passed to
4079.Nm dummynet .
4080.It Va net.inet.ip.dummynet.io_pkt_drop
4081Number of packets dropped by
4082.Nm dummynet .
4083.It Va net.inet.ip.dummynet.io_pkt_fast
4084Number of packets bypassed by the
4085.Nm dummynet
4086scheduler.
4087.It Va net.inet.ip.dummynet.max_chain_len : No 16
4088Target value for the maximum number of pipes/queues in a hash bucket.
4089The product
4090.Cm max_chain_len*hash_size
4091is used to determine the threshold over which empty pipes/queues
4092will be expired even when
4093.Cm net.inet.ip.dummynet.expire=0 .
4094.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
4095.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
4096.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
4097Parameters used in the computations of the drop probability
4098for the RED algorithm.
4099.It Va net.inet.ip.dummynet.pie.alpha : No 125
4100The default
4101.Ar alpha
4102parameter (scaled by 1000) for
4103.Cm pie
4104AQM.
4105The value must be in the range 1..7000.
4106.It Va net.inet.ip.dummynet.pie.beta : No 1250
4107The default
4108.Ar beta
4109parameter (scaled by 1000) for
4110.Cm pie
4111AQM.
4112The value must be in the range 1..7000.
4113.It Va net.inet.ip.dummynet.pie.max_burst : No 150000
4114The default maximum period of microseconds that
4115.Cm pie
4116AQM does not drop/mark packets.
4117The value must be in the range 1..10000000.
4118.It Va net.inet.ip.dummynet.pie.max_ecnth : No 99
4119The default maximum ECN probability threshold (scaled by 1000) for
4120.Cm pie
4121AQM.
4122The value must be in the range 1..7000.
4123.It Va net.inet.ip.dummynet.pie.target : No 15000
4124The default
4125.Cm target
4126delay of
4127.Cm pie
4128AQM in unit of microsecond.
4129The value must be in the range 1..5000000.
4130.It Va net.inet.ip.dummynet.pie.tupdate : No 15000
4131The default
4132.Cm tupdate
4133of
4134.Cm pie
4135AQM in unit of microsecond.
4136The value must be in the range 1..5000000.
4137.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
4138.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
4139The maximum queue size that can be specified in bytes or packets.
4140These limits prevent accidental exhaustion of resources such as mbufs.
4141If you raise these limits,
4142you should make sure the system is configured so that sufficient resources
4143are available.
4144.It Va net.inet.ip.fw.autoinc_step : No 100
4145Delta between rule numbers when auto-generating them.
4146The value must be in the range 1..1000.
4147.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
4148The current number of buckets in the hash table for dynamic rules
4149(readonly).
4150.It Va net.inet.ip.fw.debug : No 1
4151Controls debugging messages produced by
4152.Nm .
4153.It Va net.inet.ip.fw.default_rule : No 65535
4154The default rule number (read-only).
4155By the design of
4156.Nm , the default rule is the last one, so its number
4157can also serve as the highest number allowed for a rule.
4158.It Va net.inet.ip.fw.dyn_buckets : No 256
4159The number of buckets in the hash table for dynamic rules.
4160Must be a power of 2, up to 65536.
4161It only takes effect when all dynamic rules have expired, so you
4162are advised to use a
4163.Cm flush
4164command to make sure that the hash table is resized.
4165.It Va net.inet.ip.fw.dyn_count : No 3
4166Current number of dynamic rules
4167(read-only).
4168.It Va net.inet.ip.fw.dyn_keepalive : No 1
4169Enables generation of keepalive packets for
4170.Cm keep-state
4171rules on TCP sessions.
4172A keepalive is generated to both
4173sides of the connection every 5 seconds for the last 20
4174seconds of the lifetime of the rule.
4175.It Va net.inet.ip.fw.dyn_max : No 8192
4176Maximum number of dynamic rules.
4177When you hit this limit, no more dynamic rules can be
4178installed until old ones expire.
4179.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
4180.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
4181.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
4182.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
4183.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
4184.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
4185These variables control the lifetime, in seconds, of dynamic
4186rules.
4187Upon the initial SYN exchange the lifetime is kept short,
4188then increased after both SYN have been seen, then decreased
4189again during the final FIN exchange or when a RST is received.
4190Both
4191.Em dyn_fin_lifetime
4192and
4193.Em dyn_rst_lifetime
4194must be strictly lower than 5 seconds, the period of
4195repetition of keepalives.
4196The firewall enforces that.
4197.It Va net.inet.ip.fw.dyn_keep_states : No 0
4198Keep dynamic states on rule/set deletion.
4199States are relinked to default rule (65535).
4200This can be handly for ruleset reload.
4201Turned off by default.
4202.It Va net.inet.ip.fw.one_pass : No 1
4203When set, the packet exiting from the
4204.Nm dummynet
4205pipe or from
4206.Xr ng_ipfw 4
4207node is not passed though the firewall again.
4208Otherwise, after an action, the packet is
4209reinjected into the firewall at the next rule.
4210.It Va net.inet.ip.fw.tables_max : No 128
4211Maximum number of tables.
4212.It Va net.inet.ip.fw.verbose : No 1
4213Enables verbose messages.
4214.It Va net.inet.ip.fw.verbose_limit : No 0
4215Limits the number of messages produced by a verbose firewall.
4216.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
4217If enabled packets with unknown IPv6 Extension Headers will be denied.
4218.It Va net.link.bridge.ipfw : No 0
4219Controls whether bridged packets are passed to
4220.Nm .
4221Default is no.
4222.It Va net.inet.ip.fw.nat64_debug : No 0
4223Controls debugging messages produced by
4224.Nm ipfw_nat64
4225module.
4226.It Va net.inet.ip.fw.nat64_direct_output : No 0
4227Controls the output method used by
4228.Nm ipfw_nat64
4229module:
4230.Bl -tag -width indent
4231.It Cm 0
4232A packet is handled by
4233.Nm ipfw
4234twice.
4235First time an original packet is handled by
4236.Nm ipfw
4237and consumed by
4238.Nm ipfw_nat64
4239translator.
4240Then translated packet is queued via netisr to input processing again.
4241.It Cm 1
4242A packet is handled by
4243.Nm ipfw
4244only once, and after translation it will be pushed directly to outgoing
4245interface.
4246.El
4247.El
4248.Sh INTERNAL DIAGNOSTICS
4249There are some commands that may be useful to understand current state
4250of certain subsystems inside kernel module.
4251These commands provide debugging output which may change without notice.
4252.Pp
4253Currently the following commands are available as
4254.Cm internal
4255sub-options:
4256.Bl -tag -width indent
4257.It Cm iflist
4258Lists all interface which are currently tracked by
4259.Nm
4260with their in-kernel status.
4261.It Cm talist
4262List all table lookup algorithms currently available.
4263.El
4264.Sh EXAMPLES
4265There are far too many possible uses of
4266.Nm
4267so this Section will only give a small set of examples.
4268.Ss BASIC PACKET FILTERING
4269This command adds an entry which denies all tcp packets from
4270.Em cracker.evil.org
4271to the telnet port of
4272.Em wolf.tambov.su
4273from being forwarded by the host:
4274.Pp
4275.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
4276.Pp
4277This one disallows any connection from the entire cracker's
4278network to my host:
4279.Pp
4280.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
4281.Pp
4282A first and efficient way to limit access (not using dynamic rules)
4283is the use of the following rules:
4284.Pp
4285.Dl "ipfw add allow tcp from any to any established"
4286.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
4287.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
4288.Dl "..."
4289.Dl "ipfw add deny tcp from any to any"
4290.Pp
4291The first rule will be a quick match for normal TCP packets,
4292but it will not match the initial SYN packet, which will be
4293matched by the
4294.Cm setup
4295rules only for selected source/destination pairs.
4296All other SYN packets will be rejected by the final
4297.Cm deny
4298rule.
4299.Pp
4300If you administer one or more subnets, you can take advantage
4301of the address sets and or-blocks and write extremely
4302compact rulesets which selectively enable services to blocks
4303of clients, as below:
4304.Pp
4305.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
4306.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
4307.Dl ""
4308.Dl "ipfw add allow ip from ${goodguys} to any"
4309.Dl "ipfw add deny ip from ${badguys} to any"
4310.Dl "... normal policies ..."
4311.Pp
4312Allow any transit packets coming from single vlan 10 and
4313going out to vlans 100-1000:
4314.Pp
4315.Dl "ipfw add 10 allow out recv vlan10 \e"
4316.Dl "{ xmit vlan1000 or xmit \*qvlan[1-9]??\*q }"
4317.Pp
4318The
4319.Cm verrevpath
4320option could be used to do automated anti-spoofing by adding the
4321following to the top of a ruleset:
4322.Pp
4323.Dl "ipfw add deny ip from any to any not verrevpath in"
4324.Pp
4325This rule drops all incoming packets that appear to be coming to the
4326system on the wrong interface.
4327For example, a packet with a source
4328address belonging to a host on a protected internal network would be
4329dropped if it tried to enter the system from an external interface.
4330.Pp
4331The
4332.Cm antispoof
4333option could be used to do similar but more restricted anti-spoofing
4334by adding the following to the top of a ruleset:
4335.Pp
4336.Dl "ipfw add deny ip from any to any not antispoof in"
4337.Pp
4338This rule drops all incoming packets that appear to be coming from another
4339directly connected system but on the wrong interface.
4340For example, a packet with a source address of
4341.Li 192.168.0.0/24 ,
4342configured on
4343.Li fxp0 ,
4344but coming in on
4345.Li fxp1
4346would be dropped.
4347.Pp
4348The
4349.Cm setdscp
4350option could be used to (re)mark user traffic,
4351by adding the following to the appropriate place in ruleset:
4352.Pp
4353.Dl "ipfw add setdscp be ip from any to any dscp af11,af21"
4354.Ss SELECTIVE MIRRORING
4355If your network has network traffic analyzer
4356connected to your host directly via dedicated interface
4357or remotely via RSPAN vlan, you can selectively mirror
4358some Ethernet layer2 frames to the analyzer.
4359.Pp
4360First, make sure your firewall is already configured and runs.
4361Then, enable layer2 processing if not already enabled:
4362.Pp
4363.Dl "sysctl net.link.ether.ipfw=1"
4364.Pp
4365Next, load needed additional kernel modules:
4366.Pp
4367.Dl "kldload ng_ether ng_ipfw"
4368.Pp
4369Optionally, make system load these modules automatically
4370at startup:
4371.Pp
4372.Dl sysrc kld_list+="ng_ether ng_ipfw"
4373.Pp
4374Next, configure
4375.Xr ng_ipfw 4
4376kernel module to transmit mirrored copies of layer2 frames
4377out via vlan900 interface:
4378.Pp
4379.Dl "ngctl connect ipfw: vlan900: 1 lower"
4380.Pp
4381Think of "1" here as of "mirroring instance index" and vlan900 is its
4382destination.
4383You can have arbitrary number of instances.
4384Refer to
4385.Xr ng_ipfw 4
4386for details.
4387.Pp
4388At last, actually start mirroring of selected frames using "instance 1".
4389For frames incoming from em0 interface:
4390.Pp
4391.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 in recv em0"
4392.Pp
4393For frames outgoing to em0 interface:
4394.Pp
4395.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 out xmit em0"
4396.Pp
4397For both incoming and outgoing frames while flowing through em0:
4398.Pp
4399.Dl "ipfw add ngtee 1 ip from any to 192.168.0.1 layer2 via em0"
4400.Pp
4401Make sure you do not perform mirroring for already duplicated frames
4402or kernel may hang as there is no safety net.
4403.Ss DYNAMIC RULES
4404In order to protect a site from flood attacks involving fake
4405TCP packets, it is safer to use dynamic rules:
4406.Pp
4407.Dl "ipfw add check-state"
4408.Dl "ipfw add deny tcp from any to any established"
4409.Dl "ipfw add allow tcp from my-net to any setup keep-state"
4410.Pp
4411This will let the firewall install dynamic rules only for
4412those connection which start with a regular SYN packet coming
4413from the inside of our network.
4414Dynamic rules are checked when encountering the first
4415occurrence of a
4416.Cm check-state ,
4417.Cm keep-state
4418or
4419.Cm limit
4420rule.
4421A
4422.Cm check-state
4423rule should usually be placed near the beginning of the
4424ruleset to minimize the amount of work scanning the ruleset.
4425Your mileage may vary.
4426.Pp
4427For more complex scenarios with dynamic rules
4428.Cm record-state
4429and
4430.Cm defer-action
4431can be used to precisely control creation and checking of dynamic rules.
4432Example of usage of these options are provided in
4433.Sx NETWORK ADDRESS TRANSLATION (NAT)
4434Section.
4435.Pp
4436To limit the number of connections a user can open
4437you can use the following type of rules:
4438.Pp
4439.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
4440.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
4441.Pp
4442The former (assuming it runs on a gateway) will allow each host
4443on a /24 network to open at most 10 TCP connections.
4444The latter can be placed on a server to make sure that a single
4445client does not use more than 4 simultaneous connections.
4446.Pp
4447.Em BEWARE :
4448stateful rules can be subject to denial-of-service attacks
4449by a SYN-flood which opens a huge number of dynamic rules.
4450The effects of such attacks can be partially limited by
4451acting on a set of
4452.Xr sysctl 8
4453variables which control the operation of the firewall.
4454.Pp
4455Here is a good usage of the
4456.Cm list
4457command to see accounting records and timestamp information:
4458.Pp
4459.Dl ipfw -at list
4460.Pp
4461or in short form without timestamps:
4462.Pp
4463.Dl ipfw -a list
4464.Pp
4465which is equivalent to:
4466.Pp
4467.Dl ipfw show
4468.Pp
4469Next rule diverts all incoming packets from 192.168.2.0/24
4470to divert port 5000:
4471.Pp
4472.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
4473.Ss TRAFFIC SHAPING
4474The following rules show some of the applications of
4475.Nm
4476and
4477.Nm dummynet
4478for simulations and the like.
4479.Pp
4480This rule drops random incoming packets with a probability
4481of 5%:
4482.Pp
4483.Dl "ipfw add prob 0.05 deny ip from any to any in"
4484.Pp
4485A similar effect can be achieved making use of
4486.Nm dummynet
4487pipes:
4488.Pp
4489.Dl "dnctl add pipe 10 ip from any to any"
4490.Dl "dnctl pipe 10 config plr 0.05"
4491.Pp
4492We can use pipes to artificially limit bandwidth, e.g.\& on a
4493machine acting as a router, if we want to limit traffic from
4494local clients on 192.168.2.0/24 we do:
4495.Pp
4496.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
4497.Dl "dnctl pipe 1 config bw 300Kbit/s queue 50KBytes"
4498.Pp
4499note that we use the
4500.Cm out
4501modifier so that the rule is not used twice.
4502Remember in fact that
4503.Nm
4504rules are checked both on incoming and outgoing packets.
4505.Pp
4506Should we want to simulate a bidirectional link with bandwidth
4507limitations, the correct way is the following:
4508.Pp
4509.Dl "ipfw add pipe 1 ip from any to any out"
4510.Dl "ipfw add pipe 2 ip from any to any in"
4511.Dl "dnctl pipe 1 config bw 64Kbit/s queue 10Kbytes"
4512.Dl "dnctl pipe 2 config bw 64Kbit/s queue 10Kbytes"
4513.Pp
4514The above can be very useful, e.g.\& if you want to see how
4515your fancy Web page will look for a residential user who
4516is connected only through a slow link.
4517You should not use only one pipe for both directions, unless
4518you want to simulate a half-duplex medium (e.g.\& AppleTalk,
4519Ethernet, IRDA).
4520It is not necessary that both pipes have the same configuration,
4521so we can also simulate asymmetric links.
4522.Pp
4523Should we want to verify network performance with the RED queue
4524management algorithm:
4525.Pp
4526.Dl "ipfw add pipe 1 ip from any to any"
4527.Dl "dnctl pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
4528.Pp
4529Another typical application of the traffic shaper is to
4530introduce some delay in the communication.
4531This can significantly affect applications which do a lot of Remote
4532Procedure Calls, and where the round-trip-time of the
4533connection often becomes a limiting factor much more than
4534bandwidth:
4535.Pp
4536.Dl "ipfw add pipe 1 ip from any to any out"
4537.Dl "ipfw add pipe 2 ip from any to any in"
4538.Dl "dnctl pipe 1 config delay 250ms bw 1Mbit/s"
4539.Dl "dnctl pipe 2 config delay 250ms bw 1Mbit/s"
4540.Pp
4541Per-flow queueing can be useful for a variety of purposes.
4542A very simple one is counting traffic:
4543.Pp
4544.Dl "ipfw add pipe 1 tcp from any to any"
4545.Dl "ipfw add pipe 1 udp from any to any"
4546.Dl "ipfw add pipe 1 ip from any to any"
4547.Dl "dnctl pipe 1 config mask all"
4548.Pp
4549The above set of rules will create queues (and collect
4550statistics) for all traffic.
4551Because the pipes have no limitations, the only effect is
4552collecting statistics.
4553Note that we need 3 rules, not just the last one, because
4554when
4555.Nm
4556tries to match IP packets it will not consider ports, so we
4557would not see connections on separate ports as different
4558ones.
4559.Pp
4560A more sophisticated example is limiting the outbound traffic
4561on a net with per-host limits, rather than per-network limits:
4562.Pp
4563.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
4564.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
4565.Dl "dnctl pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
4566.Dl "dnctl pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
4567.Ss LOOKUP TABLES
4568In the following example, we need to create several traffic bandwidth
4569classes and we need different hosts/networks to fall into different classes.
4570We create one pipe for each class and configure them accordingly.
4571Then we create a single table and fill it with IP subnets and addresses.
4572For each subnet/host we set the argument equal to the number of the pipe
4573that it should use.
4574Then we classify traffic using a single rule:
4575.Pp
4576.Dl "dnctl pipe 1 config bw 1000Kbyte/s"
4577.Dl "dnctl pipe 4 config bw 4000Kbyte/s"
4578.Dl "..."
4579.Dl "ipfw table T1 create type addr"
4580.Dl "ipfw table T1 add 192.168.2.0/24 1"
4581.Dl "ipfw table T1 add 192.168.0.0/27 4"
4582.Dl "ipfw table T1 add 192.168.0.2 1"
4583.Dl "..."
4584.Dl "ipfw add pipe tablearg ip from 'table(T1)' to any"
4585.Pp
4586Using the
4587.Cm fwd
4588action, the table entries may include hostnames and IP addresses.
4589.Pp
4590.Dl "ipfw table T2 create type addr valtype ipv4"
4591.Dl "ipfw table T2 add 192.168.2.0/24 10.23.2.1"
4592.Dl "ipfw table T2 add 192.168.0.0/27 router1.dmz"
4593.Dl "..."
4594.Dl "ipfw add 100 fwd tablearg ip from any to 'table(T2)'"
4595.Pp
4596In the following example per-interface firewall is created:
4597.Pp
4598.Dl "ipfw table IN create type iface valtype skipto,fib"
4599.Dl "ipfw table IN add vlan20 12000,12"
4600.Dl "ipfw table IN add vlan30 13000,13"
4601.Dl "ipfw table OUT create type iface valtype skipto"
4602.Dl "ipfw table OUT add vlan20 22000"
4603.Dl "ipfw table OUT add vlan30 23000"
4604.Dl ".."
4605.Dl "ipfw add 100 setfib tablearg ip from any to any recv 'table(IN)' in"
4606.Dl "ipfw add 200 skipto tablearg ip from any to any recv 'table(IN)' in"
4607.Dl "ipfw add 300 skipto tablearg ip from any to any xmit 'table(OUT)' out"
4608.Pp
4609The following example illustrate usage of flow tables:
4610.Pp
4611.Dl "ipfw table fl create type flow:src-ip,proto,dst-ip,dst-port"
4612.Dl "ipfw table fl add 2a02:6b8:77::88,tcp,2a02:6b8:77::99,80 11"
4613.Dl "ipfw table fl add 10.0.0.1,udp,10.0.0.2,53 12"
4614.Dl ".."
4615.Dl "ipfw add 100 allow ip from any to any flow 'table(fl,11)' recv ix0"
4616.Ss SETS OF RULES
4617To add a set of rules atomically, e.g.\& set 18:
4618.Pp
4619.Dl "ipfw set disable 18"
4620.Dl "ipfw add NN set 18 ...         # repeat as needed"
4621.Dl "ipfw set enable 18"
4622.Pp
4623To delete a set of rules atomically the command is simply:
4624.Pp
4625.Dl "ipfw delete set 18"
4626.Pp
4627To test a ruleset and disable it and regain control if something goes wrong:
4628.Pp
4629.Dl "ipfw set disable 18"
4630.Dl "ipfw add NN set 18 ...         # repeat as needed"
4631.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
4632.Pp
4633Here if everything goes well, you press control-C before the "sleep"
4634terminates, and your ruleset will be left active.
4635Otherwise, e.g.\& if
4636you cannot access your box, the ruleset will be disabled after
4637the sleep terminates thus restoring the previous situation.
4638.Pp
4639To show rules of the specific set:
4640.Pp
4641.Dl "ipfw set 18 show"
4642.Pp
4643To show rules of the disabled set:
4644.Pp
4645.Dl "ipfw -S set 18 show"
4646.Pp
4647To clear a specific rule counters of the specific set:
4648.Pp
4649.Dl "ipfw set 18 zero NN"
4650.Pp
4651To delete a specific rule of the specific set:
4652.Pp
4653.Dl "ipfw set 18 delete NN"
4654.Ss NAT, REDIRECT AND LSNAT
4655First redirect all the traffic to nat instance 123:
4656.Pp
4657.Dl "ipfw add nat 123 all from any to any"
4658.Pp
4659Then to configure nat instance 123 to alias all the outgoing traffic with ip
4660192.168.0.123, blocking all incoming connections, trying to keep
4661same ports on both sides, clearing aliasing table on address change
4662and keeping a log of traffic/link statistics:
4663.Pp
4664.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
4665.Pp
4666Or to change address of instance 123, aliasing table will be cleared (see
4667reset option):
4668.Pp
4669.Dl "ipfw nat 123 config ip 10.0.0.1"
4670.Pp
4671To see configuration of nat instance 123:
4672.Pp
4673.Dl "ipfw nat 123 show config"
4674.Pp
4675To show logs of all instances:
4676.Pp
4677.Dl "ipfw nat show log"
4678.Pp
4679To see configurations of all instances:
4680.Pp
4681.Dl "ipfw nat show config"
4682.Pp
4683Or a redirect rule with mixed modes could looks like:
4684.Bd -literal -offset 2n
4685ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66
4686			 redirect_port tcp 192.168.0.1:80 500
4687			 redirect_proto udp 192.168.1.43 192.168.1.1
4688			 redirect_addr 192.168.0.10,192.168.0.11
4689			 	    10.0.0.100	# LSNAT
4690			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22
4691			 	    500		# LSNAT
4692.Ed
4693.Pp
4694or it could be split in:
4695.Bd -literal -offset 2n
4696ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66
4697ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500
4698ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1
4699ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12
4700				         10.0.0.100
4701ipfw nat 5 config redirect_port tcp
4702			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500
4703.Ed
4704.Pp
4705Sometimes you may want to mix NAT and dynamic rules.
4706It could be achieved with
4707.Cm record-state
4708and
4709.Cm defer-action
4710options.
4711Problem is, you need to create dynamic rule before NAT and check it
4712after NAT actions (or vice versa) to have consistent addresses and ports.
4713Rule with
4714.Cm keep-state
4715option will trigger activation of existing dynamic state, and action of such
4716rule will be performed as soon as rule is matched.
4717In case of NAT and
4718.Cm allow
4719rule packet need to be passed to NAT, not allowed as soon is possible.
4720.Pp
4721There is example of set of rules to achieve this.
4722Bear in mind that this is example only and it is not very useful by itself.
4723.Pp
4724On way out, after all checks place this rules:
4725.Pp
4726.Dl "ipfw add allow record-state defer-action"
4727.Dl "ipfw add nat 1"
4728.Pp
4729And on way in there should be something like this:
4730.Pp
4731.Dl "ipfw add nat 1"
4732.Dl "ipfw add check-state"
4733.Pp
4734Please note, that first rule on way out doesn't allow packet and doesn't
4735execute existing dynamic rules.
4736All it does, create new dynamic rule with
4737.Cm allow
4738action, if it is not created yet.
4739Later, this dynamic rule is used on way in by
4740.Cm check-state
4741rule.
4742.Ss CONFIGURING CODEL, PIE, FQ-CODEL and FQ-PIE AQM
4743.Cm codel
4744and
4745.Cm pie
4746AQM can be configured for
4747.Nm dummynet
4748.Cm pipe
4749or
4750.Cm queue .
4751.Pp
4752To configure a
4753.Cm pipe
4754with
4755.Cm codel
4756AQM using default configuration for traffic from 192.168.0.0/24 and 1Mbits/s
4757rate limit, we do:
4758.Pp
4759.Dl "dnctl pipe 1 config bw 1mbits/s codel"
4760.Dl "ipfw add 100 pipe 1 ip from 192.168.0.0/24 to any"
4761.Pp
4762To configure a
4763.Cm queue
4764with
4765.Cm codel
4766AQM using different configurations parameters for traffic from
4767192.168.0.0/24 and 1Mbits/s rate limit, we do:
4768.Pp
4769.Dl "dnctl pipe 1 config bw 1mbits/s"
4770.Dl "dnctl queue 1 config pipe 1 codel target 8ms interval 160ms ecn"
4771.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4772.Pp
4773To configure a
4774.Cm pipe
4775with
4776.Cm pie
4777AQM using default configuration for traffic from 192.168.0.0/24 and 1Mbits/s
4778rate limit, we do:
4779.Pp
4780.Dl "dnctl pipe 1 config bw 1mbits/s pie"
4781.Dl "ipfw add 100 pipe 1 ip from 192.168.0.0/24 to any"
4782.Pp
4783To configure a
4784.Cm queue
4785with
4786.Cm pie
4787AQM using different configuration parameters for traffic from
4788192.168.0.0/24 and 1Mbits/s rate limit, we do:
4789.Pp
4790.Dl "dnctl pipe 1 config bw 1mbits/s"
4791.Dl "dnctl queue 1 config pipe 1 pie target 20ms tupdate 30ms ecn"
4792.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4793.Pp
4794.Cm fq_codel
4795and
4796.Cm fq_pie
4797AQM can be configured for
4798.Nm dummynet
4799schedulers.
4800.Pp
4801To configure
4802.Cm fq_codel
4803scheduler using different configurations parameters for traffic from
4804192.168.0.0/24 and 1Mbits/s rate limit, we do:
4805.Pp
4806.Dl "dnctl pipe 1 config bw 1mbits/s"
4807.Dl "dnctl sched 1 config pipe 1 type fq_codel"
4808.Dl "dnctl queue 1 config sched 1"
4809.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4810.Pp
4811To change
4812.Cm fq_codel
4813default configuration for a
4814.Cm sched
4815such as disable ECN and change the
4816.Ar target
4817to 10ms, we do:
4818.Pp
4819.Dl "dnctl sched 1 config pipe 1 type fq_codel target 10ms noecn"
4820.Pp
4821Similar to
4822.Cm fq_codel ,
4823to configure
4824.Cm fq_pie
4825scheduler using different configurations parameters for traffic from
4826192.168.0.0/24 and 1Mbits/s rate limit, we do:
4827.Pp
4828.Dl "dnctl pipe 1 config bw 1mbits/s"
4829.Dl "dnctl sched 1 config pipe 1 type fq_pie"
4830.Dl "dnctl queue 1 config sched 1"
4831.Dl "ipfw add 100 queue 1 ip from 192.168.0.0/24 to any"
4832.Pp
4833The configurations of
4834.Cm fq_pie
4835.Cm sched
4836can be changed in a similar way as for
4837.Cm fq_codel
4838.Sh SEE ALSO
4839.Xr cpp 1 ,
4840.Xr m4 1 ,
4841.Xr fnmatch 3 ,
4842.Xr altq 4 ,
4843.Xr divert 4 ,
4844.Xr dummynet 4 ,
4845.Xr if_bridge 4 ,
4846.Xr ip 4 ,
4847.Xr ipfirewall 4 ,
4848.Xr ng_ether 4 ,
4849.Xr ng_ipfw 4 ,
4850.Xr protocols 5 ,
4851.Xr services 5 ,
4852.Xr init 8 ,
4853.Xr kldload 8 ,
4854.Xr reboot 8 ,
4855.Xr sysctl 8 ,
4856.Xr syslogd 8 ,
4857.Xr sysrc 8
4858.Sh HISTORY
4859The
4860.Nm
4861utility first appeared in
4862.Fx 2.0 .
4863.Nm dummynet
4864was introduced in
4865.Fx 2.2.8 .
4866Stateful extensions were introduced in
4867.Fx 4.0 .
4868.Nm ipfw2
4869was introduced in Summer 2002.
4870.Sh AUTHORS
4871.An Ugen J. S. Antsilevich ,
4872.An Poul-Henning Kamp ,
4873.An Alex Nash ,
4874.An Archie Cobbs ,
4875.An Luigi Rizzo ,
4876.An Rasool Al-Saadi .
4877.Pp
4878.An -nosplit
4879API based upon code written by
4880.An Daniel Boulet
4881for BSDI.
4882.Pp
4883Dummynet has been introduced by Luigi Rizzo in 1997-1998.
4884.Pp
4885Some early work (1999-2000) on the
4886.Nm dummynet
4887traffic shaper supported by Akamba Corp.
4888.Pp
4889The ipfw core (ipfw2) has been completely redesigned and
4890reimplemented by Luigi Rizzo in summer 2002.
4891Further
4892actions and
4893options have been added by various developers over the years.
4894.Pp
4895.An -nosplit
4896In-kernel NAT support written by
4897.An Paolo Pisati Aq Mt piso@FreeBSD.org
4898as part of a Summer of Code 2005 project.
4899.Pp
4900SCTP
4901.Nm nat
4902support has been developed by
4903.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
4904The primary developers and maintainers are David Hayes and Jason But.
4905For further information visit:
4906.Aq http://www.caia.swin.edu.au/urp/SONATA
4907.Pp
4908Delay profiles have been developed by Alessandro Cerri and
4909Luigi Rizzo, supported by the
4910European Commission within Projects Onelab and Onelab2.
4911.Pp
4912CoDel, PIE, FQ-CoDel and FQ-PIE AQM for Dummynet have been implemented by
4913.An The Centre for Advanced Internet Architectures (CAIA)
4914in 2016, supported by The Comcast Innovation Fund.
4915The primary developer is
4916Rasool Al-Saadi.
4917.Sh BUGS
4918The syntax has grown over the years and sometimes it might be confusing.
4919Unfortunately, backward compatibility prevents cleaning up mistakes
4920made in the definition of the syntax.
4921.Pp
4922.Em !!! WARNING !!!
4923.Pp
4924Misconfiguring the firewall can put your computer in an unusable state,
4925possibly shutting down network services and requiring console access to
4926regain control of it.
4927.Pp
4928Incoming packet fragments diverted by
4929.Cm divert
4930are reassembled before delivery to the socket.
4931The action used on those packet is the one from the
4932rule which matches the first fragment of the packet.
4933.Pp
4934Packets diverted to userland, and then reinserted by a userland process
4935may lose various packet attributes.
4936The packet source interface name
4937will be preserved if it is shorter than 8 bytes and the userland process
4938saves and reuses the sockaddr_in
4939(as does
4940.Xr natd 8 ) ;
4941otherwise, it may be lost.
4942If a packet is reinserted in this manner, later rules may be incorrectly
4943applied, making the order of
4944.Cm divert
4945rules in the rule sequence very important.
4946.Pp
4947Dummynet drops all packets with IPv6 link-local addresses.
4948.Pp
4949Rules using
4950.Cm uid
4951or
4952.Cm gid
4953may not behave as expected.
4954In particular, incoming SYN packets may
4955have no uid or gid associated with them since they do not yet belong
4956to a TCP connection, and the uid/gid associated with a packet may not
4957be as expected if the associated process calls
4958.Xr setuid 2
4959or similar system calls.
4960.Pp
4961Rule syntax is subject to the command line environment and some patterns
4962may need to be escaped with the backslash character
4963or quoted appropriately.
4964.Pp
4965Due to the architecture of
4966.Xr libalias 3 ,
4967ipfw nat is not compatible with the TCP segmentation offloading (TSO).
4968Thus, to reliably nat your network traffic, please disable TSO
4969on your NICs using
4970.Xr ifconfig 8 .
4971.Pp
4972ICMP error messages are not implicitly matched by dynamic rules
4973for the respective conversations.
4974To avoid failures of network error detection and path MTU discovery,
4975ICMP error messages may need to be allowed explicitly through static
4976rules.
4977.Pp
4978Rules using
4979.Cm call
4980and
4981.Cm return
4982actions may lead to confusing behaviour if ruleset has mistakes,
4983and/or interaction with other subsystems (netgraph, dummynet, etc.) is used.
4984One possible case for this is packet leaving
4985.Nm
4986in subroutine on the input pass, while later on output encountering unpaired
4987.Cm return
4988first.
4989As the call stack is kept intact after input pass, packet will suddenly
4990return to the rule number used on input pass, not on output one.
4991Order of processing should be checked carefully to avoid such mistakes.
4992