xref: /freebsd/sbin/ipfw/ipfw.8 (revision aa77200569e397d6ff1fdb4d255d0fa254d0a128)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd October 25, 2012
5.Dt IPFW 8
6.Os
7.Sh NAME
8.Nm ipfw
9.Nd User interface for firewall, traffic shaper, packet scheduler,
10in-kernel NAT.
11.Sh SYNOPSIS
12.Ss FIREWALL CONFIGURATION
13.Nm
14.Op Fl cq
15.Cm add
16.Ar rule
17.Nm
18.Op Fl acdefnNStT
19.Op Cm set Ar N
20.Brq Cm list | show
21.Op Ar rule | first-last ...
22.Nm
23.Op Fl f | q
24.Op Cm set Ar N
25.Cm flush
26.Nm
27.Op Fl q
28.Op Cm set Ar N
29.Brq Cm delete | zero | resetlog
30.Op Ar number ...
31.Pp
32.Nm
33.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
34.Nm
35.Cm set move
36.Op Cm rule
37.Ar number Cm to Ar number
38.Nm
39.Cm set swap Ar number number
40.Nm
41.Cm set show
42.Ss SYSCTL SHORTCUTS
43.Nm
44.Cm enable
45.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
46.Nm
47.Cm disable
48.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
49.Ss LOOKUP TABLES
50.Nm
51.Cm table Ar number Cm add Ar addr Ns Oo / Ns Ar masklen Oc Op Ar value
52.Nm
53.Cm table Ar number Cm delete Ar addr Ns Op / Ns Ar masklen
54.Nm
55.Cm table
56.Brq Ar number | all
57.Cm flush
58.Nm
59.Cm table
60.Brq Ar number | all
61.Cm list
62.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
63.Nm
64.Brq Cm pipe | queue | sched
65.Ar number
66.Cm config
67.Ar config-options
68.Nm
69.Op Fl s Op Ar field
70.Brq Cm pipe | queue | sched
71.Brq Cm delete | list | show
72.Op Ar number ...
73.Ss IN-KERNEL NAT
74.Nm
75.Op Fl q
76.Cm nat
77.Ar number
78.Cm config
79.Ar config-options
80.Pp
81.Nm
82.Op Fl cfnNqS
83.Oo
84.Fl p Ar preproc
85.Oo
86.Ar preproc-flags
87.Oc
88.Oc
89.Ar pathname
90.Sh DESCRIPTION
91The
92.Nm
93utility is the user interface for controlling the
94.Xr ipfw 4
95firewall, the
96.Xr dummynet 4
97traffic shaper/packet scheduler, and the
98in-kernel NAT services.
99.Pp
100A firewall configuration, or
101.Em ruleset ,
102is made of a list of
103.Em rules
104numbered from 1 to 65535.
105Packets are passed to the firewall
106from a number of different places in the protocol stack
107(depending on the source and destination of the packet,
108it is possible for the firewall to be
109invoked multiple times on the same packet).
110The packet passed to the firewall is compared
111against each of the rules in the
112.Em ruleset ,
113in rule-number order
114(multiple rules with the same number are permitted, in which case
115they are processed in order of insertion).
116When a match is found, the action corresponding to the
117matching rule is performed.
118.Pp
119Depending on the action and certain system settings, packets
120can be reinjected into the firewall at some rule after the
121matching one for further processing.
122.Pp
123A ruleset always includes a
124.Em default
125rule (numbered 65535) which cannot be modified or deleted,
126and matches all packets.
127The action associated with the
128.Em default
129rule can be either
130.Cm deny
131or
132.Cm allow
133depending on how the kernel is configured.
134.Pp
135If the ruleset includes one or more rules with the
136.Cm keep-state
137or
138.Cm limit
139option,
140the firewall will have a
141.Em stateful
142behaviour, i.e., upon a match it will create
143.Em dynamic rules ,
144i.e., rules that match packets with the same 5-tuple
145(protocol, source and destination addresses and ports)
146as the packet which caused their creation.
147Dynamic rules, which have a limited lifetime, are checked
148at the first occurrence of a
149.Cm check-state ,
150.Cm keep-state
151or
152.Cm limit
153rule, and are typically used to open the firewall on-demand to
154legitimate traffic only.
155See the
156.Sx STATEFUL FIREWALL
157and
158.Sx EXAMPLES
159Sections below for more information on the stateful behaviour of
160.Nm .
161.Pp
162All rules (including dynamic ones) have a few associated counters:
163a packet count, a byte count, a log count and a timestamp
164indicating the time of the last match.
165Counters can be displayed or reset with
166.Nm
167commands.
168.Pp
169Each rule belongs to one of 32 different
170.Em sets
171, and there are
172.Nm
173commands to atomically manipulate sets, such as enable,
174disable, swap sets, move all rules in a set to another
175one, delete all rules in a set.
176These can be useful to
177install temporary configurations, or to test them.
178See Section
179.Sx SETS OF RULES
180for more information on
181.Em sets .
182.Pp
183Rules can be added with the
184.Cm add
185command; deleted individually or in groups with the
186.Cm delete
187command, and globally (except those in set 31) with the
188.Cm flush
189command; displayed, optionally with the content of the
190counters, using the
191.Cm show
192and
193.Cm list
194commands.
195Finally, counters can be reset with the
196.Cm zero
197and
198.Cm resetlog
199commands.
200.Pp
201.Ss COMMAND OPTIONS
202The following general options are available when invoking
203.Nm :
204.Bl -tag -width indent
205.It Fl a
206Show counter values when listing rules.
207The
208.Cm show
209command implies this option.
210.It Fl b
211Only show the action and the comment, not the body of a rule.
212Implies
213.Fl c .
214.It Fl c
215When entering or showing rules, print them in compact form,
216i.e., omitting the "ip from any to any" string
217when this does not carry any additional information.
218.It Fl d
219When listing, show dynamic rules in addition to static ones.
220.It Fl e
221When listing and
222.Fl d
223is specified, also show expired dynamic rules.
224.It Fl f
225Do not ask for confirmation for commands that can cause problems
226if misused, i.e.,
227.Cm flush .
228If there is no tty associated with the process, this is implied.
229.It Fl i
230When listing a table (see the
231.Sx LOOKUP TABLES
232section below for more information on lookup tables), format values
233as IP addresses.
234By default, values are shown as integers.
235.It Fl n
236Only check syntax of the command strings, without actually passing
237them to the kernel.
238.It Fl N
239Try to resolve addresses and service names in output.
240.It Fl q
241Be quiet when executing the
242.Cm add ,
243.Cm nat ,
244.Cm zero ,
245.Cm resetlog
246or
247.Cm flush
248commands;
249(implies
250.Fl f ) .
251This is useful when updating rulesets by executing multiple
252.Nm
253commands in a script
254(e.g.,
255.Ql sh\ /etc/rc.firewall ) ,
256or by processing a file with many
257.Nm
258rules across a remote login session.
259It also stops a table add or delete
260from failing if the entry already exists or is not present.
261.Pp
262The reason why this option may be important is that
263for some of these actions,
264.Nm
265may print a message; if the action results in blocking the
266traffic to the remote client,
267the remote login session will be closed
268and the rest of the ruleset will not be processed.
269Access to the console would then be required to recover.
270.It Fl S
271When listing rules, show the
272.Em set
273each rule belongs to.
274If this flag is not specified, disabled rules will not be
275listed.
276.It Fl s Op Ar field
277When listing pipes, sort according to one of the four
278counters (total or current packets or bytes).
279.It Fl t
280When listing, show last match timestamp converted with ctime().
281.It Fl T
282When listing, show last match timestamp as seconds from the epoch.
283This form can be more convenient for postprocessing by scripts.
284.El
285.Ss LIST OF RULES AND PREPROCESSING
286To ease configuration, rules can be put into a file which is
287processed using
288.Nm
289as shown in the last synopsis line.
290An absolute
291.Ar pathname
292must be used.
293The file will be read line by line and applied as arguments to the
294.Nm
295utility.
296.Pp
297Optionally, a preprocessor can be specified using
298.Fl p Ar preproc
299where
300.Ar pathname
301is to be piped through.
302Useful preprocessors include
303.Xr cpp 1
304and
305.Xr m4 1 .
306If
307.Ar preproc
308does not start with a slash
309.Pq Ql /
310as its first character, the usual
311.Ev PATH
312name search is performed.
313Care should be taken with this in environments where not all
314file systems are mounted (yet) by the time
315.Nm
316is being run (e.g.\& when they are mounted over NFS).
317Once
318.Fl p
319has been specified, any additional arguments are passed on to the preprocessor
320for interpretation.
321This allows for flexible configuration files (like conditionalizing
322them on the local hostname) and the use of macros to centralize
323frequently required arguments like IP addresses.
324.Ss TRAFFIC SHAPER CONFIGURATION
325The
326.Nm
327.Cm pipe , queue
328and
329.Cm sched
330commands are used to configure the traffic shaper and packet scheduler.
331See the
332.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
333Section below for details.
334.Pp
335If the world and the kernel get out of sync the
336.Nm
337ABI may break, preventing you from being able to add any rules.
338This can
339adversely effect the booting process.
340You can use
341.Nm
342.Cm disable
343.Cm firewall
344to temporarily disable the firewall to regain access to the network,
345allowing you to fix the problem.
346.Sh PACKET FLOW
347A packet is checked against the active ruleset in multiple places
348in the protocol stack, under control of several sysctl variables.
349These places and variables are shown below, and it is important to
350have this picture in mind in order to design a correct ruleset.
351.Bd -literal -offset indent
352       ^    to upper layers    V
353       |                       |
354       +----------->-----------+
355       ^                       V
356 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
357       |                       |
358       ^                       V
359 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
360       |                       |
361       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
362       ^                       V
363       |      to devices       |
364.Ed
365.Pp
366The number of
367times the same packet goes through the firewall can
368vary between 0 and 4 depending on packet source and
369destination, and system configuration.
370.Pp
371Note that as packets flow through the stack, headers can be
372stripped or added to it, and so they may or may not be available
373for inspection.
374E.g., incoming packets will include the MAC header when
375.Nm
376is invoked from
377.Cm ether_demux() ,
378but the same packets will have the MAC header stripped off when
379.Nm
380is invoked from
381.Cm ip_input()
382or
383.Cm ip6_input() .
384.Pp
385Also note that each packet is always checked against the complete ruleset,
386irrespective of the place where the check occurs, or the source of the packet.
387If a rule contains some match patterns or actions which are not valid
388for the place of invocation (e.g.\& trying to match a MAC header within
389.Cm ip_input
390or
391.Cm ip6_input ),
392the match pattern will not match, but a
393.Cm not
394operator in front of such patterns
395.Em will
396cause the pattern to
397.Em always
398match on those packets.
399It is thus the responsibility of
400the programmer, if necessary, to write a suitable ruleset to
401differentiate among the possible places.
402.Cm skipto
403rules can be useful here, as an example:
404.Bd -literal -offset indent
405# packets from ether_demux or bdg_forward
406ipfw add 10 skipto 1000 all from any to any layer2 in
407# packets from ip_input
408ipfw add 10 skipto 2000 all from any to any not layer2 in
409# packets from ip_output
410ipfw add 10 skipto 3000 all from any to any not layer2 out
411# packets from ether_output_frame
412ipfw add 10 skipto 4000 all from any to any layer2 out
413.Ed
414.Pp
415(yes, at the moment there is no way to differentiate between
416ether_demux and bdg_forward).
417.Sh SYNTAX
418In general, each keyword or argument must be provided as
419a separate command line argument, with no leading or trailing
420spaces.
421Keywords are case-sensitive, whereas arguments may
422or may not be case-sensitive depending on their nature
423(e.g.\& uid's are, hostnames are not).
424.Pp
425Some arguments (e.g., port or address lists) are comma-separated
426lists of values.
427In this case, spaces after commas ',' are allowed to make
428the line more readable.
429You can also put the entire
430command (including flags) into a single argument.
431E.g., the following forms are equivalent:
432.Bd -literal -offset indent
433ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
434ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
435ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
436.Ed
437.Sh RULE FORMAT
438The format of firewall rules is the following:
439.Bd -ragged -offset indent
440.Bk -words
441.Op Ar rule_number
442.Op Cm set Ar set_number
443.Op Cm prob Ar match_probability
444.Ar action
445.Op Cm log Op Cm logamount Ar number
446.Op Cm altq Ar queue
447.Oo
448.Bro Cm tag | untag
449.Brc Ar number
450.Oc
451.Ar body
452.Ek
453.Ed
454.Pp
455where the body of the rule specifies which information is used
456for filtering packets, among the following:
457.Pp
458.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
459.It Layer-2 header fields
460When available
461.It IPv4 and IPv6 Protocol
462TCP, UDP, ICMP, etc.
463.It Source and dest. addresses and ports
464.It Direction
465See Section
466.Sx PACKET FLOW
467.It Transmit and receive interface
468By name or address
469.It Misc. IP header fields
470Version, type of service, datagram length, identification,
471fragment flag (non-zero IP offset),
472Time To Live
473.It IP options
474.It IPv6 Extension headers
475Fragmentation, Hop-by-Hop options,
476Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
477.It IPv6 Flow-ID
478.It Misc. TCP header fields
479TCP flags (SYN, FIN, ACK, RST, etc.),
480sequence number, acknowledgment number,
481window
482.It TCP options
483.It ICMP types
484for ICMP packets
485.It ICMP6 types
486for ICMP6 packets
487.It User/group ID
488When the packet can be associated with a local socket.
489.It Divert status
490Whether a packet came from a divert socket (e.g.,
491.Xr natd 8 ) .
492.It Fib annotation state
493Whether a packet has been tagged for using a specific FIB (routing table)
494in future forwarding decisions.
495.El
496.Pp
497Note that some of the above information, e.g.\& source MAC or IP addresses and
498TCP/UDP ports, can be easily spoofed, so filtering on those fields
499alone might not guarantee the desired results.
500.Bl -tag -width indent
501.It Ar rule_number
502Each rule is associated with a
503.Ar rule_number
504in the range 1..65535, with the latter reserved for the
505.Em default
506rule.
507Rules are checked sequentially by rule number.
508Multiple rules can have the same number, in which case they are
509checked (and listed) according to the order in which they have
510been added.
511If a rule is entered without specifying a number, the kernel will
512assign one in such a way that the rule becomes the last one
513before the
514.Em default
515rule.
516Automatic rule numbers are assigned by incrementing the last
517non-default rule number by the value of the sysctl variable
518.Ar net.inet.ip.fw.autoinc_step
519which defaults to 100.
520If this is not possible (e.g.\& because we would go beyond the
521maximum allowed rule number), the number of the last
522non-default value is used instead.
523.It Cm set Ar set_number
524Each rule is associated with a
525.Ar set_number
526in the range 0..31.
527Sets can be individually disabled and enabled, so this parameter
528is of fundamental importance for atomic ruleset manipulation.
529It can be also used to simplify deletion of groups of rules.
530If a rule is entered without specifying a set number,
531set 0 will be used.
532.br
533Set 31 is special in that it cannot be disabled,
534and rules in set 31 are not deleted by the
535.Nm ipfw flush
536command (but you can delete them with the
537.Nm ipfw delete set 31
538command).
539Set 31 is also used for the
540.Em default
541rule.
542.It Cm prob Ar match_probability
543A match is only declared with the specified probability
544(floating point number between 0 and 1).
545This can be useful for a number of applications such as
546random packet drop or
547(in conjunction with
548.Nm dummynet )
549to simulate the effect of multiple paths leading to out-of-order
550packet delivery.
551.Pp
552Note: this condition is checked before any other condition, including
553ones such as keep-state or check-state which might have side effects.
554.It Cm log Op Cm logamount Ar number
555Packets matching a rule with the
556.Cm log
557keyword will be made available for logging in two ways:
558if the sysctl variable
559.Va net.inet.ip.fw.verbose
560is set to 0 (default), one can use
561.Xr bpf 4
562attached to the
563.Li ipfw0
564pseudo interface.
565This pseudo interface can be created after a boot
566manually by using the following command:
567.Bd -literal -offset indent
568# ifconfig ipfw0 create
569.Ed
570.Pp
571Or, automatically at boot time by adding the following
572line to the
573.Xr rc.conf 5
574file:
575.Bd -literal -offset indent
576firewall_logif="YES"
577.Ed
578.Pp
579There is no overhead if no
580.Xr bpf 4
581is attached to the pseudo interface.
582.Pp
583If
584.Va net.inet.ip.fw.verbose
585is set to 1, packets will be logged to
586.Xr syslogd 8
587with a
588.Dv LOG_SECURITY
589facility up to a maximum of
590.Cm logamount
591packets.
592If no
593.Cm logamount
594is specified, the limit is taken from the sysctl variable
595.Va net.inet.ip.fw.verbose_limit .
596In both cases, a value of 0 means unlimited logging.
597.Pp
598Once the limit is reached, logging can be re-enabled by
599clearing the logging counter or the packet counter for that entry, see the
600.Cm resetlog
601command.
602.Pp
603Note: logging is done after all other packet matching conditions
604have been successfully verified, and before performing the final
605action (accept, deny, etc.) on the packet.
606.It Cm tag Ar number
607When a packet matches a rule with the
608.Cm tag
609keyword, the numeric tag for the given
610.Ar number
611in the range 1..65534 will be attached to the packet.
612The tag acts as an internal marker (it is not sent out over
613the wire) that can be used to identify these packets later on.
614This can be used, for example, to provide trust between interfaces
615and to start doing policy-based filtering.
616A packet can have multiple tags at the same time.
617Tags are "sticky", meaning once a tag is applied to a packet by a
618matching rule it exists until explicit removal.
619Tags are kept with the packet everywhere within the kernel, but are
620lost when packet leaves the kernel, for example, on transmitting
621packet out to the network or sending packet to a
622.Xr divert 4
623socket.
624.Pp
625To check for previously applied tags, use the
626.Cm tagged
627rule option.
628To delete previously applied tag, use the
629.Cm untag
630keyword.
631.Pp
632Note: since tags are kept with the packet everywhere in kernelspace,
633they can be set and unset anywhere in the kernel network subsystem
634(using the
635.Xr mbuf_tags 9
636facility), not only by means of the
637.Xr ipfw 4
638.Cm tag
639and
640.Cm untag
641keywords.
642For example, there can be a specialized
643.Xr netgraph 4
644node doing traffic analyzing and tagging for later inspecting
645in firewall.
646.It Cm untag Ar number
647When a packet matches a rule with the
648.Cm untag
649keyword, the tag with the number
650.Ar number
651is searched among the tags attached to this packet and,
652if found, removed from it.
653Other tags bound to packet, if present, are left untouched.
654.It Cm altq Ar queue
655When a packet matches a rule with the
656.Cm altq
657keyword, the ALTQ identifier for the given
658.Ar queue
659(see
660.Xr altq 4 )
661will be attached.
662Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
663and not being rejected or going to divert sockets.
664Note that if there is insufficient memory at the time the packet is
665processed, it will not be tagged, so it is wise to make your ALTQ
666"default" queue policy account for this.
667If multiple
668.Cm altq
669rules match a single packet, only the first one adds the ALTQ classification
670tag.
671In doing so, traffic may be shaped by using
672.Cm count Cm altq Ar queue
673rules for classification early in the ruleset, then later applying
674the filtering decision.
675For example,
676.Cm check-state
677and
678.Cm keep-state
679rules may come later and provide the actual filtering decisions in
680addition to the fallback ALTQ tag.
681.Pp
682You must run
683.Xr pfctl 8
684to set up the queues before IPFW will be able to look them up by name,
685and if the ALTQ disciplines are rearranged, the rules in containing the
686queue identifiers in the kernel will likely have gone stale and need
687to be reloaded.
688Stale queue identifiers will probably result in misclassification.
689.Pp
690All system ALTQ processing can be turned on or off via
691.Nm
692.Cm enable Ar altq
693and
694.Nm
695.Cm disable Ar altq .
696The usage of
697.Va net.inet.ip.fw.one_pass
698is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
699always after adding an ALTQ tag.
700.El
701.Ss RULE ACTIONS
702A rule can be associated with one of the following actions, which
703will be executed when the packet matches the body of the rule.
704.Bl -tag -width indent
705.It Cm allow | accept | pass | permit
706Allow packets that match rule.
707The search terminates.
708.It Cm check-state
709Checks the packet against the dynamic ruleset.
710If a match is found, execute the action associated with
711the rule which generated this dynamic rule, otherwise
712move to the next rule.
713.br
714.Cm Check-state
715rules do not have a body.
716If no
717.Cm check-state
718rule is found, the dynamic ruleset is checked at the first
719.Cm keep-state
720or
721.Cm limit
722rule.
723.It Cm count
724Update counters for all packets that match rule.
725The search continues with the next rule.
726.It Cm deny | drop
727Discard packets that match this rule.
728The search terminates.
729.It Cm divert Ar port
730Divert packets that match this rule to the
731.Xr divert 4
732socket bound to port
733.Ar port .
734The search terminates.
735.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
736Change the next-hop on matching packets to
737.Ar ipaddr ,
738which can be an IP address or a host name.
739For IPv4, the next hop can also be supplied by the last table
740looked up for the packet by using the
741.Cm tablearg
742keyword instead of an explicit address.
743The search terminates if this rule matches.
744.Pp
745If
746.Ar ipaddr
747is a local address, then matching packets will be forwarded to
748.Ar port
749(or the port number in the packet if one is not specified in the rule)
750on the local machine.
751.br
752If
753.Ar ipaddr
754is not a local address, then the port number
755(if specified) is ignored, and the packet will be
756forwarded to the remote address, using the route as found in
757the local routing table for that IP.
758.br
759A
760.Ar fwd
761rule will not match layer-2 packets (those received
762on ether_input, ether_output, or bridged).
763.br
764The
765.Cm fwd
766action does not change the contents of the packet at all.
767In particular, the destination address remains unmodified, so
768packets forwarded to another system will usually be rejected by that system
769unless there is a matching rule on that system to capture them.
770For packets forwarded locally,
771the local address of the socket will be
772set to the original destination address of the packet.
773This makes the
774.Xr netstat 1
775entry look rather weird but is intended for
776use with transparent proxy servers.
777.It Cm nat Ar nat_nr | tablearg
778Pass packet to a
779nat instance
780(for network address translation, address redirect, etc.):
781see the
782.Sx NETWORK ADDRESS TRANSLATION (NAT)
783Section for further information.
784.It Cm pipe Ar pipe_nr
785Pass packet to a
786.Nm dummynet
787.Dq pipe
788(for bandwidth limitation, delay, etc.).
789See the
790.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
791Section for further information.
792The search terminates; however, on exit from the pipe and if
793the
794.Xr sysctl 8
795variable
796.Va net.inet.ip.fw.one_pass
797is not set, the packet is passed again to the firewall code
798starting from the next rule.
799.It Cm queue Ar queue_nr
800Pass packet to a
801.Nm dummynet
802.Dq queue
803(for bandwidth limitation using WF2Q+).
804.It Cm reject
805(Deprecated).
806Synonym for
807.Cm unreach host .
808.It Cm reset
809Discard packets that match this rule, and if the
810packet is a TCP packet, try to send a TCP reset (RST) notice.
811The search terminates.
812.It Cm reset6
813Discard packets that match this rule, and if the
814packet is a TCP packet, try to send a TCP reset (RST) notice.
815The search terminates.
816.It Cm skipto Ar number | tablearg
817Skip all subsequent rules numbered less than
818.Ar number .
819The search continues with the first rule numbered
820.Ar number
821or higher.
822It is possible to use the
823.Cm tablearg
824keyword with a skipto for a
825.Em computed
826skipto, but care should be used, as no destination caching
827is possible in this case so the rules are always walked to find it,
828starting from the
829.Cm skipto .
830.It Cm call Ar number | tablearg
831The current rule number is saved in the internal stack and
832ruleset processing continues with the first rule numbered
833.Ar number
834or higher.
835If later a rule with the
836.Cm return
837action is encountered, the processing returns to the first rule
838with number of this
839.Cm call
840rule plus one or higher
841(the same behaviour as with packets returning from
842.Xr divert 4
843socket after a
844.Cm divert
845action).
846This could be used to make somewhat like an assembly language
847.Dq subroutine
848calls to rules with common checks for different interfaces, etc.
849.Pp
850Rule with any number could be called, not just forward jumps as with
851.Cm skipto .
852So, to prevent endless loops in case of mistakes, both
853.Cm call
854and
855.Cm return
856actions don't do any jumps and simply go to the next rule if memory
857can't be allocated or stack overflowed/undeflowed.
858.Pp
859Internally stack for rule numbers is implemented using
860.Xr mbuf_tags 9
861facility and currently has size of 16 entries.
862As mbuf tags are lost when packet leaves the kernel,
863.Cm divert
864should not be used in subroutines to avoid endless loops
865and other undesired effects.
866.It Cm return
867Takes rule number saved to internal stack by the last
868.Cm call
869action and returns ruleset processing to the first rule
870with number greater than number of corresponding
871.Cm call
872rule.
873See description of the
874.Cm call
875action for more details.
876.Pp
877Note that
878.Cm return
879rules usually end a
880.Dq subroutine
881and thus are unconditional, but
882.Nm
883command-line utility currently requires every action except
884.Cm check-state
885to have body.
886While it is sometimes useful to return only on some packets,
887usually you want to print just
888.Dq return
889for readability.
890A workaround for this is to use new syntax and
891.Fl c
892switch:
893.Bd -literal -offset indent
894# Add a rule without actual body
895ipfw add 2999 return via any
896
897# List rules without "from any to any" part
898ipfw -c list
899.Ed
900.Pp
901This cosmetic annoyance may be fixed in future releases.
902.It Cm tee Ar port
903Send a copy of packets matching this rule to the
904.Xr divert 4
905socket bound to port
906.Ar port .
907The search continues with the next rule.
908.It Cm unreach Ar code
909Discard packets that match this rule, and try to send an ICMP
910unreachable notice with code
911.Ar code ,
912where
913.Ar code
914is a number from 0 to 255, or one of these aliases:
915.Cm net , host , protocol , port ,
916.Cm needfrag , srcfail , net-unknown , host-unknown ,
917.Cm isolated , net-prohib , host-prohib , tosnet ,
918.Cm toshost , filter-prohib , host-precedence
919or
920.Cm precedence-cutoff .
921The search terminates.
922.It Cm unreach6 Ar code
923Discard packets that match this rule, and try to send an ICMPv6
924unreachable notice with code
925.Ar code ,
926where
927.Ar code
928is a number from 0, 1, 3 or 4, or one of these aliases:
929.Cm no-route, admin-prohib, address
930or
931.Cm port .
932The search terminates.
933.It Cm netgraph Ar cookie
934Divert packet into netgraph with given
935.Ar cookie .
936The search terminates.
937If packet is later returned from netgraph it is either
938accepted or continues with the next rule, depending on
939.Va net.inet.ip.fw.one_pass
940sysctl variable.
941.It Cm ngtee Ar cookie
942A copy of packet is diverted into netgraph, original
943packet continues with the next rule.
944See
945.Xr ng_ipfw 4
946for more information on
947.Cm netgraph
948and
949.Cm ngtee
950actions.
951.It Cm setfib Ar fibnum | tablearg
952The packet is tagged so as to use the FIB (routing table)
953.Ar fibnum
954in any subsequent forwarding decisions.
955In the current implementation, this is limited to the values 0 through 15, see
956.Xr setfib 2 .
957Processing continues at the next rule.
958It is possible to use the
959.Cm tablearg
960keyword with setfib.
961If the tablearg value is not within the compiled range of fibs,
962the packet's fib is set to 0.
963.It Cm reass
964Queue and reassemble IP fragments.
965If the packet is not fragmented, counters are updated and
966processing continues with the next rule.
967If the packet is the last logical fragment, the packet is reassembled and, if
968.Va net.inet.ip.fw.one_pass
969is set to 0, processing continues with the next rule.
970Otherwise, the packet is allowed to pass and the search terminates.
971If the packet is a fragment in the middle of a logical group of fragments,
972it is consumed and
973processing stops immediately.
974.Pp
975Fragment handling can be tuned via
976.Va net.inet.ip.maxfragpackets
977and
978.Va net.inet.ip.maxfragsperpacket
979which limit, respectively, the maximum number of processable
980fragments (default: 800) and
981the maximum number of fragments per packet (default: 16).
982.Pp
983NOTA BENE: since fragments do not contain port numbers,
984they should be avoided with the
985.Nm reass
986rule.
987Alternatively, direction-based (like
988.Nm in
989/
990.Nm out
991) and source-based (like
992.Nm via
993) match patterns can be used to select fragments.
994.Pp
995Usually a simple rule like:
996.Bd -literal -offset indent
997# reassemble incoming fragments
998ipfw add reass all from any to any in
999.Ed
1000.Pp
1001is all you need at the beginning of your ruleset.
1002.El
1003.Ss RULE BODY
1004The body of a rule contains zero or more patterns (such as
1005specific source and destination addresses or ports,
1006protocol options, incoming or outgoing interfaces, etc.)
1007that the packet must match in order to be recognised.
1008In general, the patterns are connected by (implicit)
1009.Cm and
1010operators -- i.e., all must match in order for the
1011rule to match.
1012Individual patterns can be prefixed by the
1013.Cm not
1014operator to reverse the result of the match, as in
1015.Pp
1016.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
1017.Pp
1018Additionally, sets of alternative match patterns
1019.Pq Em or-blocks
1020can be constructed by putting the patterns in
1021lists enclosed between parentheses ( ) or braces { }, and
1022using the
1023.Cm or
1024operator as follows:
1025.Pp
1026.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
1027.Pp
1028Only one level of parentheses is allowed.
1029Beware that most shells have special meanings for parentheses
1030or braces, so it is advisable to put a backslash \\ in front of them
1031to prevent such interpretations.
1032.Pp
1033The body of a rule must in general include a source and destination
1034address specifier.
1035The keyword
1036.Ar any
1037can be used in various places to specify that the content of
1038a required field is irrelevant.
1039.Pp
1040The rule body has the following format:
1041.Bd -ragged -offset indent
1042.Op Ar proto Cm from Ar src Cm to Ar dst
1043.Op Ar options
1044.Ed
1045.Pp
1046The first part (proto from src to dst) is for backward
1047compatibility with earlier versions of
1048.Fx .
1049In modern
1050.Fx
1051any match pattern (including MAC headers, IP protocols,
1052addresses and ports) can be specified in the
1053.Ar options
1054section.
1055.Pp
1056Rule fields have the following meaning:
1057.Bl -tag -width indent
1058.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
1059.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
1060An IP protocol specified by number or name
1061(for a complete list see
1062.Pa /etc/protocols ) ,
1063or one of the following keywords:
1064.Bl -tag -width indent
1065.It Cm ip4 | ipv4
1066Matches IPv4 packets.
1067.It Cm ip6 | ipv6
1068Matches IPv6 packets.
1069.It Cm ip | all
1070Matches any packet.
1071.El
1072.Pp
1073The
1074.Cm ipv6
1075in
1076.Cm proto
1077option will be treated as inner protocol.
1078And, the
1079.Cm ipv4
1080is not available in
1081.Cm proto
1082option.
1083.Pp
1084The
1085.Cm { Ar protocol Cm or ... }
1086format (an
1087.Em or-block )
1088is provided for convenience only but its use is deprecated.
1089.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1090An address (or a list, see below)
1091optionally followed by
1092.Ar ports
1093specifiers.
1094.Pp
1095The second format
1096.Em ( or-block
1097with multiple addresses) is provided for convenience only and
1098its use is discouraged.
1099.It Ar addr : Oo Cm not Oc Bro
1100.Cm any | me | me6 |
1101.Cm table Ns Pq Ar number Ns Op , Ns Ar value
1102.Ar | addr-list | addr-set
1103.Brc
1104.Bl -tag -width indent
1105.It Cm any
1106matches any IP address.
1107.It Cm me
1108matches any IP address configured on an interface in the system.
1109.It Cm me6
1110matches any IPv6 address configured on an interface in the system.
1111The address list is evaluated at the time the packet is
1112analysed.
1113.It Cm table Ns Pq Ar number Ns Op , Ns Ar value
1114Matches any IPv4 address for which an entry exists in the lookup table
1115.Ar number .
1116If an optional 32-bit unsigned
1117.Ar value
1118is also specified, an entry will match only if it has this value.
1119See the
1120.Sx LOOKUP TABLES
1121section below for more information on lookup tables.
1122.El
1123.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1124.It Ar ip-addr :
1125A host or subnet address specified in one of the following ways:
1126.Bl -tag -width indent
1127.It Ar numeric-ip | hostname
1128Matches a single IPv4 address, specified as dotted-quad or a hostname.
1129Hostnames are resolved at the time the rule is added to the firewall list.
1130.It Ar addr Ns / Ns Ar masklen
1131Matches all addresses with base
1132.Ar addr
1133(specified as an IP address, a network number, or a hostname)
1134and mask width of
1135.Cm masklen
1136bits.
1137As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1138all IP numbers from 1.2.3.0 to 1.2.3.127 .
1139.It Ar addr Ns : Ns Ar mask
1140Matches all addresses with base
1141.Ar addr
1142(specified as an IP address, a network number, or a hostname)
1143and the mask of
1144.Ar mask ,
1145specified as a dotted quad.
1146As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
11471.*.3.*.
1148This form is advised only for non-contiguous
1149masks.
1150It is better to resort to the
1151.Ar addr Ns / Ns Ar masklen
1152format for contiguous masks, which is more compact and less
1153error-prone.
1154.El
1155.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1156.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1157Matches all addresses with base address
1158.Ar addr
1159(specified as an IP address, a network number, or a hostname)
1160and whose last byte is in the list between braces { } .
1161Note that there must be no spaces between braces and
1162numbers (spaces after commas are allowed).
1163Elements of the list can be specified as single entries
1164or ranges.
1165The
1166.Ar masklen
1167field is used to limit the size of the set of addresses,
1168and can have any value between 24 and 32.
1169If not specified,
1170it will be assumed as 24.
1171.br
1172This format is particularly useful to handle sparse address sets
1173within a single rule.
1174Because the matching occurs using a
1175bitmask, it takes constant time and dramatically reduces
1176the complexity of rulesets.
1177.br
1178As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1179or 1.2.3.0/24{128,35-55,89}
1180will match the following IP addresses:
1181.br
11821.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1183.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1184.It Ar ip6-addr :
1185A host or subnet specified one of the following ways:
1186.Bl -tag -width indent
1187.It Ar numeric-ip | hostname
1188Matches a single IPv6 address as allowed by
1189.Xr inet_pton 3
1190or a hostname.
1191Hostnames are resolved at the time the rule is added to the firewall
1192list.
1193.It Ar addr Ns / Ns Ar masklen
1194Matches all IPv6 addresses with base
1195.Ar addr
1196(specified as allowed by
1197.Xr inet_pton
1198or a hostname)
1199and mask width of
1200.Cm masklen
1201bits.
1202.El
1203.Pp
1204No support for sets of IPv6 addresses is provided because IPv6 addresses
1205are typically random past the initial prefix.
1206.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1207For protocols which support port numbers (such as TCP and UDP), optional
1208.Cm ports
1209may be specified as one or more ports or port ranges, separated
1210by commas but no spaces, and an optional
1211.Cm not
1212operator.
1213The
1214.Ql \&-
1215notation specifies a range of ports (including boundaries).
1216.Pp
1217Service names (from
1218.Pa /etc/services )
1219may be used instead of numeric port values.
1220The length of the port list is limited to 30 ports or ranges,
1221though one can specify larger ranges by using an
1222.Em or-block
1223in the
1224.Cm options
1225section of the rule.
1226.Pp
1227A backslash
1228.Pq Ql \e
1229can be used to escape the dash
1230.Pq Ql -
1231character in a service name (from a shell, the backslash must be
1232typed twice to avoid the shell itself interpreting it as an escape
1233character).
1234.Pp
1235.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1236.Pp
1237Fragmented packets which have a non-zero offset (i.e., not the first
1238fragment) will never match a rule which has one or more port
1239specifications.
1240See the
1241.Cm frag
1242option for details on matching fragmented packets.
1243.El
1244.Ss RULE OPTIONS (MATCH PATTERNS)
1245Additional match patterns can be used within
1246rules.
1247Zero or more of these so-called
1248.Em options
1249can be present in a rule, optionally prefixed by the
1250.Cm not
1251operand, and possibly grouped into
1252.Em or-blocks .
1253.Pp
1254The following match patterns can be used (listed in alphabetical order):
1255.Bl -tag -width indent
1256.It Cm // this is a comment.
1257Inserts the specified text as a comment in the rule.
1258Everything following // is considered as a comment and stored in the rule.
1259You can have comment-only rules, which are listed as having a
1260.Cm count
1261action followed by the comment.
1262.It Cm bridged
1263Alias for
1264.Cm layer2 .
1265.It Cm diverted
1266Matches only packets generated by a divert socket.
1267.It Cm diverted-loopback
1268Matches only packets coming from a divert socket back into the IP stack
1269input for delivery.
1270.It Cm diverted-output
1271Matches only packets going from a divert socket back outward to the IP
1272stack output for delivery.
1273.It Cm dst-ip Ar ip-address
1274Matches IPv4 packets whose destination IP is one of the address(es)
1275specified as argument.
1276.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1277Matches IPv6 packets whose destination IP is one of the address(es)
1278specified as argument.
1279.It Cm dst-port Ar ports
1280Matches IP packets whose destination port is one of the port(s)
1281specified as argument.
1282.It Cm established
1283Matches TCP packets that have the RST or ACK bits set.
1284.It Cm ext6hdr Ar header
1285Matches IPv6 packets containing the extended header given by
1286.Ar header .
1287Supported headers are:
1288.Pp
1289Fragment,
1290.Pq Cm frag ,
1291Hop-to-hop options
1292.Pq Cm hopopt ,
1293any type of Routing Header
1294.Pq Cm route ,
1295Source routing Routing Header Type 0
1296.Pq Cm rthdr0 ,
1297Mobile IPv6 Routing Header Type 2
1298.Pq Cm rthdr2 ,
1299Destination options
1300.Pq Cm dstopt ,
1301IPSec authentication headers
1302.Pq Cm ah ,
1303and IPsec encapsulated security payload headers
1304.Pq Cm esp .
1305.It Cm fib Ar fibnum
1306Matches a packet that has been tagged to use
1307the given FIB (routing table) number.
1308.It Cm flow-id Ar labels
1309Matches IPv6 packets containing any of the flow labels given in
1310.Ar labels .
1311.Ar labels
1312is a comma separated list of numeric flow labels.
1313.It Cm frag
1314Matches packets that are fragments and not the first
1315fragment of an IP datagram.
1316Note that these packets will not have
1317the next protocol header (e.g.\& TCP, UDP) so options that look into
1318these headers cannot match.
1319.It Cm gid Ar group
1320Matches all TCP or UDP packets sent by or received for a
1321.Ar group .
1322A
1323.Ar group
1324may be specified by name or number.
1325.It Cm jail Ar prisonID
1326Matches all TCP or UDP packets sent by or received for the
1327jail whos prison ID is
1328.Ar prisonID .
1329.It Cm icmptypes Ar types
1330Matches ICMP packets whose ICMP type is in the list
1331.Ar types .
1332The list may be specified as any combination of
1333individual types (numeric) separated by commas.
1334.Em Ranges are not allowed .
1335The supported ICMP types are:
1336.Pp
1337echo reply
1338.Pq Cm 0 ,
1339destination unreachable
1340.Pq Cm 3 ,
1341source quench
1342.Pq Cm 4 ,
1343redirect
1344.Pq Cm 5 ,
1345echo request
1346.Pq Cm 8 ,
1347router advertisement
1348.Pq Cm 9 ,
1349router solicitation
1350.Pq Cm 10 ,
1351time-to-live exceeded
1352.Pq Cm 11 ,
1353IP header bad
1354.Pq Cm 12 ,
1355timestamp request
1356.Pq Cm 13 ,
1357timestamp reply
1358.Pq Cm 14 ,
1359information request
1360.Pq Cm 15 ,
1361information reply
1362.Pq Cm 16 ,
1363address mask request
1364.Pq Cm 17
1365and address mask reply
1366.Pq Cm 18 .
1367.It Cm icmp6types Ar types
1368Matches ICMP6 packets whose ICMP6 type is in the list of
1369.Ar types .
1370The list may be specified as any combination of
1371individual types (numeric) separated by commas.
1372.Em Ranges are not allowed .
1373.It Cm in | out
1374Matches incoming or outgoing packets, respectively.
1375.Cm in
1376and
1377.Cm out
1378are mutually exclusive (in fact,
1379.Cm out
1380is implemented as
1381.Cm not in Ns No ).
1382.It Cm ipid Ar id-list
1383Matches IPv4 packets whose
1384.Cm ip_id
1385field has value included in
1386.Ar id-list ,
1387which is either a single value or a list of values or ranges
1388specified in the same way as
1389.Ar ports .
1390.It Cm iplen Ar len-list
1391Matches IP packets whose total length, including header and data, is
1392in the set
1393.Ar len-list ,
1394which is either a single value or a list of values or ranges
1395specified in the same way as
1396.Ar ports .
1397.It Cm ipoptions Ar spec
1398Matches packets whose IPv4 header contains the comma separated list of
1399options specified in
1400.Ar spec .
1401The supported IP options are:
1402.Pp
1403.Cm ssrr
1404(strict source route),
1405.Cm lsrr
1406(loose source route),
1407.Cm rr
1408(record packet route) and
1409.Cm ts
1410(timestamp).
1411The absence of a particular option may be denoted
1412with a
1413.Ql \&! .
1414.It Cm ipprecedence Ar precedence
1415Matches IPv4 packets whose precedence field is equal to
1416.Ar precedence .
1417.It Cm ipsec
1418Matches packets that have IPSEC history associated with them
1419(i.e., the packet comes encapsulated in IPSEC, the kernel
1420has IPSEC support and IPSEC_FILTERTUNNEL option, and can correctly
1421decapsulate it).
1422.Pp
1423Note that specifying
1424.Cm ipsec
1425is different from specifying
1426.Cm proto Ar ipsec
1427as the latter will only look at the specific IP protocol field,
1428irrespective of IPSEC kernel support and the validity of the IPSEC data.
1429.Pp
1430Further note that this flag is silently ignored in kernels without
1431IPSEC support.
1432It does not affect rule processing when given and the
1433rules are handled as if with no
1434.Cm ipsec
1435flag.
1436.It Cm iptos Ar spec
1437Matches IPv4 packets whose
1438.Cm tos
1439field contains the comma separated list of
1440service types specified in
1441.Ar spec .
1442The supported IP types of service are:
1443.Pp
1444.Cm lowdelay
1445.Pq Dv IPTOS_LOWDELAY ,
1446.Cm throughput
1447.Pq Dv IPTOS_THROUGHPUT ,
1448.Cm reliability
1449.Pq Dv IPTOS_RELIABILITY ,
1450.Cm mincost
1451.Pq Dv IPTOS_MINCOST ,
1452.Cm congestion
1453.Pq Dv IPTOS_ECN_CE .
1454The absence of a particular type may be denoted
1455with a
1456.Ql \&! .
1457.It Cm ipttl Ar ttl-list
1458Matches IPv4 packets whose time to live is included in
1459.Ar ttl-list ,
1460which is either a single value or a list of values or ranges
1461specified in the same way as
1462.Ar ports .
1463.It Cm ipversion Ar ver
1464Matches IP packets whose IP version field is
1465.Ar ver .
1466.It Cm keep-state
1467Upon a match, the firewall will create a dynamic rule, whose
1468default behaviour is to match bidirectional traffic between
1469source and destination IP/port using the same protocol.
1470The rule has a limited lifetime (controlled by a set of
1471.Xr sysctl 8
1472variables), and the lifetime is refreshed every time a matching
1473packet is found.
1474.It Cm layer2
1475Matches only layer2 packets, i.e., those passed to
1476.Nm
1477from ether_demux() and ether_output_frame().
1478.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
1479The firewall will only allow
1480.Ar N
1481connections with the same
1482set of parameters as specified in the rule.
1483One or more
1484of source and destination addresses and ports can be
1485specified.
1486Currently,
1487only IPv4 flows are supported.
1488.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar N
1489Search an entry in lookup table
1490.Ar N
1491that matches the field specified as argument.
1492If not found, the match fails.
1493Otherwise, the match succeeds and
1494.Cm tablearg
1495is set to the value extracted from the table.
1496.Pp
1497This option can be useful to quickly dispatch traffic based on
1498certain packet fields.
1499See the
1500.Sx LOOKUP TABLES
1501section below for more information on lookup tables.
1502.It Cm { MAC | mac } Ar dst-mac src-mac
1503Match packets with a given
1504.Ar dst-mac
1505and
1506.Ar src-mac
1507addresses, specified as the
1508.Cm any
1509keyword (matching any MAC address), or six groups of hex digits
1510separated by colons,
1511and optionally followed by a mask indicating the significant bits.
1512The mask may be specified using either of the following methods:
1513.Bl -enum -width indent
1514.It
1515A slash
1516.Pq /
1517followed by the number of significant bits.
1518For example, an address with 33 significant bits could be specified as:
1519.Pp
1520.Dl "MAC 10:20:30:40:50:60/33 any"
1521.Pp
1522.It
1523An ampersand
1524.Pq &
1525followed by a bitmask specified as six groups of hex digits separated
1526by colons.
1527For example, an address in which the last 16 bits are significant could
1528be specified as:
1529.Pp
1530.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1531.Pp
1532Note that the ampersand character has a special meaning in many shells
1533and should generally be escaped.
1534.Pp
1535.El
1536Note that the order of MAC addresses (destination first,
1537source second) is
1538the same as on the wire, but the opposite of the one used for
1539IP addresses.
1540.It Cm mac-type Ar mac-type
1541Matches packets whose Ethernet Type field
1542corresponds to one of those specified as argument.
1543.Ar mac-type
1544is specified in the same way as
1545.Cm port numbers
1546(i.e., one or more comma-separated single values or ranges).
1547You can use symbolic names for known values such as
1548.Em vlan , ipv4, ipv6 .
1549Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1550and they are always printed as hexadecimal (unless the
1551.Cm -N
1552option is used, in which case symbolic resolution will be attempted).
1553.It Cm proto Ar protocol
1554Matches packets with the corresponding IP protocol.
1555.It Cm recv | xmit | via Brq Ar ifX | Ar if Ns Cm * | Ar table Ns Pq Ar number Ns Op , Ns Ar value | Ar ipno | Ar any
1556Matches packets received, transmitted or going through,
1557respectively, the interface specified by exact name
1558.Po Ar ifX Pc ,
1559by device name
1560.Po Ar if* Pc ,
1561by IP address, or through some interface.
1562.Pp
1563The
1564.Cm via
1565keyword causes the interface to always be checked.
1566If
1567.Cm recv
1568or
1569.Cm xmit
1570is used instead of
1571.Cm via ,
1572then only the receive or transmit interface (respectively)
1573is checked.
1574By specifying both, it is possible to match packets based on
1575both receive and transmit interface, e.g.:
1576.Pp
1577.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1578.Pp
1579The
1580.Cm recv
1581interface can be tested on either incoming or outgoing packets,
1582while the
1583.Cm xmit
1584interface can only be tested on outgoing packets.
1585So
1586.Cm out
1587is required (and
1588.Cm in
1589is invalid) whenever
1590.Cm xmit
1591is used.
1592.Pp
1593A packet might not have a receive or transmit interface: packets
1594originating from the local host have no receive interface,
1595while packets destined for the local host have no transmit
1596interface.
1597.It Cm setup
1598Matches TCP packets that have the SYN bit set but no ACK bit.
1599This is the short form of
1600.Dq Li tcpflags\ syn,!ack .
1601.It Cm sockarg
1602Matches packets that are associated to a local socket and
1603for which the SO_USER_COOKIE socket option has been set
1604to a non-zero value.
1605As a side effect, the value of the
1606option is made available as
1607.Cm tablearg
1608value, which in turn can be used as
1609.Cm skipto
1610or
1611.Cm pipe
1612number.
1613.It Cm src-ip Ar ip-address
1614Matches IPv4 packets whose source IP is one of the address(es)
1615specified as an argument.
1616.It Cm src-ip6 Ar ip6-address
1617Matches IPv6 packets whose source IP is one of the address(es)
1618specified as an argument.
1619.It Cm src-port Ar ports
1620Matches IP packets whose source port is one of the port(s)
1621specified as argument.
1622.It Cm tagged Ar tag-list
1623Matches packets whose tags are included in
1624.Ar tag-list ,
1625which is either a single value or a list of values or ranges
1626specified in the same way as
1627.Ar ports .
1628Tags can be applied to the packet using
1629.Cm tag
1630rule action parameter (see it's description for details on tags).
1631.It Cm tcpack Ar ack
1632TCP packets only.
1633Match if the TCP header acknowledgment number field is set to
1634.Ar ack .
1635.It Cm tcpdatalen Ar tcpdatalen-list
1636Matches TCP packets whose length of TCP data is
1637.Ar tcpdatalen-list ,
1638which is either a single value or a list of values or ranges
1639specified in the same way as
1640.Ar ports .
1641.It Cm tcpflags Ar spec
1642TCP packets only.
1643Match if the TCP header contains the comma separated list of
1644flags specified in
1645.Ar spec .
1646The supported TCP flags are:
1647.Pp
1648.Cm fin ,
1649.Cm syn ,
1650.Cm rst ,
1651.Cm psh ,
1652.Cm ack
1653and
1654.Cm urg .
1655The absence of a particular flag may be denoted
1656with a
1657.Ql \&! .
1658A rule which contains a
1659.Cm tcpflags
1660specification can never match a fragmented packet which has
1661a non-zero offset.
1662See the
1663.Cm frag
1664option for details on matching fragmented packets.
1665.It Cm tcpseq Ar seq
1666TCP packets only.
1667Match if the TCP header sequence number field is set to
1668.Ar seq .
1669.It Cm tcpwin Ar tcpwin-list
1670Matches TCP packets whose  header window field is set to
1671.Ar tcpwin-list ,
1672which is either a single value or a list of values or ranges
1673specified in the same way as
1674.Ar ports .
1675.It Cm tcpoptions Ar spec
1676TCP packets only.
1677Match if the TCP header contains the comma separated list of
1678options specified in
1679.Ar spec .
1680The supported TCP options are:
1681.Pp
1682.Cm mss
1683(maximum segment size),
1684.Cm window
1685(tcp window advertisement),
1686.Cm sack
1687(selective ack),
1688.Cm ts
1689(rfc1323 timestamp) and
1690.Cm cc
1691(rfc1644 t/tcp connection count).
1692The absence of a particular option may be denoted
1693with a
1694.Ql \&! .
1695.It Cm uid Ar user
1696Match all TCP or UDP packets sent by or received for a
1697.Ar user .
1698A
1699.Ar user
1700may be matched by name or identification number.
1701.It Cm verrevpath
1702For incoming packets,
1703a routing table lookup is done on the packet's source address.
1704If the interface on which the packet entered the system matches the
1705outgoing interface for the route,
1706the packet matches.
1707If the interfaces do not match up,
1708the packet does not match.
1709All outgoing packets or packets with no incoming interface match.
1710.Pp
1711The name and functionality of the option is intentionally similar to
1712the Cisco IOS command:
1713.Pp
1714.Dl ip verify unicast reverse-path
1715.Pp
1716This option can be used to make anti-spoofing rules to reject all
1717packets with source addresses not from this interface.
1718See also the option
1719.Cm antispoof .
1720.It Cm versrcreach
1721For incoming packets,
1722a routing table lookup is done on the packet's source address.
1723If a route to the source address exists, but not the default route
1724or a blackhole/reject route, the packet matches.
1725Otherwise, the packet does not match.
1726All outgoing packets match.
1727.Pp
1728The name and functionality of the option is intentionally similar to
1729the Cisco IOS command:
1730.Pp
1731.Dl ip verify unicast source reachable-via any
1732.Pp
1733This option can be used to make anti-spoofing rules to reject all
1734packets whose source address is unreachable.
1735.It Cm antispoof
1736For incoming packets, the packet's source address is checked if it
1737belongs to a directly connected network.
1738If the network is directly connected, then the interface the packet
1739came on in is compared to the interface the network is connected to.
1740When incoming interface and directly connected interface are not the
1741same, the packet does not match.
1742Otherwise, the packet does match.
1743All outgoing packets match.
1744.Pp
1745This option can be used to make anti-spoofing rules to reject all
1746packets that pretend to be from a directly connected network but do
1747not come in through that interface.
1748This option is similar to but more restricted than
1749.Cm verrevpath
1750because it engages only on packets with source addresses of directly
1751connected networks instead of all source addresses.
1752.El
1753.Sh LOOKUP TABLES
1754Lookup tables are useful to handle large sparse sets of
1755addresses or other search keys (e.g., ports, jail IDs, interface names).
1756In the rest of this section we will use the term ``address''.
1757There may be up to 65535 different lookup tables, numbered 0 to 65534.
1758.Pp
1759Each entry is represented by an
1760.Ar addr Ns Op / Ns Ar masklen
1761and will match all addresses with base
1762.Ar addr
1763(specified as an IPv4/IPv6 address, a hostname or an unsigned integer)
1764and mask width of
1765.Ar masklen
1766bits.
1767If
1768.Ar masklen
1769is not specified, it defaults to 32 for IPv4 and 128 for IPv6.
1770When looking up an IP address in a table, the most specific
1771entry will match.
1772Associated with each entry is a 32-bit unsigned
1773.Ar value ,
1774which can optionally be checked by a rule matching code.
1775When adding an entry, if
1776.Ar value
1777is not specified, it defaults to 0.
1778.Pp
1779An entry can be added to a table
1780.Pq Cm add ,
1781or removed from a table
1782.Pq Cm delete .
1783A table can be examined
1784.Pq Cm list
1785or flushed
1786.Pq Cm flush .
1787.Pp
1788Internally, each table is stored in a Radix tree, the same way as
1789the routing table (see
1790.Xr route 4 ) .
1791.Pp
1792Lookup tables currently support only ports, jail IDs, IPv4/IPv6  addresses
1793and interface names.
1794Wildcards is not supported for interface names.
1795.Pp
1796The
1797.Cm tablearg
1798feature provides the ability to use a value, looked up in the table, as
1799the argument for a rule action, action parameter or rule option.
1800This can significantly reduce number of rules in some configurations.
1801If two tables are used in a rule, the result of the second (destination)
1802is used.
1803The
1804.Cm tablearg
1805argument can be used with the following actions:
1806.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto, setfib,
1807action parameters:
1808.Cm tag, untag,
1809rule options:
1810.Cm limit, tagged.
1811.Pp
1812When used with
1813.Cm fwd
1814it is possible to supply table entries with values
1815that are in the form of IP addresses or hostnames.
1816See the
1817.Sx EXAMPLES
1818Section for example usage of tables and the tablearg keyword.
1819.Pp
1820When used with the
1821.Cm skipto
1822action, the user should be aware that the code will walk the ruleset
1823up to a rule equal to, or past, the given number,
1824and should therefore try keep the
1825ruleset compact between the skipto and the target rules.
1826.Sh SETS OF RULES
1827Each rule belongs to one of 32 different
1828.Em sets
1829, numbered 0 to 31.
1830Set 31 is reserved for the default rule.
1831.Pp
1832By default, rules are put in set 0, unless you use the
1833.Cm set N
1834attribute when entering a new rule.
1835Sets can be individually and atomically enabled or disabled,
1836so this mechanism permits an easy way to store multiple configurations
1837of the firewall and quickly (and atomically) switch between them.
1838The command to enable/disable sets is
1839.Bd -ragged -offset indent
1840.Nm
1841.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
1842.Ed
1843.Pp
1844where multiple
1845.Cm enable
1846or
1847.Cm disable
1848sections can be specified.
1849Command execution is atomic on all the sets specified in the command.
1850By default, all sets are enabled.
1851.Pp
1852When you disable a set, its rules behave as if they do not exist
1853in the firewall configuration, with only one exception:
1854.Bd -ragged -offset indent
1855dynamic rules created from a rule before it had been disabled
1856will still be active until they expire.
1857In order to delete
1858dynamic rules you have to explicitly delete the parent rule
1859which generated them.
1860.Ed
1861.Pp
1862The set number of rules can be changed with the command
1863.Bd -ragged -offset indent
1864.Nm
1865.Cm set move
1866.Brq Cm rule Ar rule-number | old-set
1867.Cm to Ar new-set
1868.Ed
1869.Pp
1870Also, you can atomically swap two rulesets with the command
1871.Bd -ragged -offset indent
1872.Nm
1873.Cm set swap Ar first-set second-set
1874.Ed
1875.Pp
1876See the
1877.Sx EXAMPLES
1878Section on some possible uses of sets of rules.
1879.Sh STATEFUL FIREWALL
1880Stateful operation is a way for the firewall to dynamically
1881create rules for specific flows when packets that
1882match a given pattern are detected.
1883Support for stateful
1884operation comes through the
1885.Cm check-state , keep-state
1886and
1887.Cm limit
1888options of
1889.Nm rules .
1890.Pp
1891Dynamic rules are created when a packet matches a
1892.Cm keep-state
1893or
1894.Cm limit
1895rule, causing the creation of a
1896.Em dynamic
1897rule which will match all and only packets with
1898a given
1899.Em protocol
1900between a
1901.Em src-ip/src-port dst-ip/dst-port
1902pair of addresses
1903.Em ( src
1904and
1905.Em dst
1906are used here only to denote the initial match addresses, but they
1907are completely equivalent afterwards).
1908Dynamic rules will be checked at the first
1909.Cm check-state, keep-state
1910or
1911.Cm limit
1912occurrence, and the action performed upon a match will be the same
1913as in the parent rule.
1914.Pp
1915Note that no additional attributes other than protocol and IP addresses
1916and ports are checked on dynamic rules.
1917.Pp
1918The typical use of dynamic rules is to keep a closed firewall configuration,
1919but let the first TCP SYN packet from the inside network install a
1920dynamic rule for the flow so that packets belonging to that session
1921will be allowed through the firewall:
1922.Pp
1923.Dl "ipfw add check-state"
1924.Dl "ipfw add allow tcp from my-subnet to any setup keep-state"
1925.Dl "ipfw add deny tcp from any to any"
1926.Pp
1927A similar approach can be used for UDP, where an UDP packet coming
1928from the inside will install a dynamic rule to let the response through
1929the firewall:
1930.Pp
1931.Dl "ipfw add check-state"
1932.Dl "ipfw add allow udp from my-subnet to any keep-state"
1933.Dl "ipfw add deny udp from any to any"
1934.Pp
1935Dynamic rules expire after some time, which depends on the status
1936of the flow and the setting of some
1937.Cm sysctl
1938variables.
1939See Section
1940.Sx SYSCTL VARIABLES
1941for more details.
1942For TCP sessions, dynamic rules can be instructed to periodically
1943send keepalive packets to refresh the state of the rule when it is
1944about to expire.
1945.Pp
1946See Section
1947.Sx EXAMPLES
1948for more examples on how to use dynamic rules.
1949.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
1950.Nm
1951is also the user interface for the
1952.Nm dummynet
1953traffic shaper, packet scheduler and network emulator, a subsystem that
1954can artificially queue, delay or drop packets
1955emulating the behaviour of certain network links
1956or queueing systems.
1957.Pp
1958.Nm dummynet
1959operates by first using the firewall to select packets
1960using any match pattern that can be used in
1961.Nm
1962rules.
1963Matching packets are then passed to either of two
1964different objects, which implement the traffic regulation:
1965.Bl -hang -offset XXXX
1966.It Em pipe
1967A
1968.Em pipe
1969emulates a
1970.Em link
1971with given bandwidth and propagation delay,
1972driven by a FIFO scheduler and a single queue with programmable
1973queue size and packet loss rate.
1974Packets are appended to the queue as they come out from
1975.Nm ipfw ,
1976and then transferred in FIFO order to the link at the desired rate.
1977.It Em queue
1978A
1979.Em queue
1980is an abstraction used to implement packet scheduling
1981using one of several packet scheduling algorithms.
1982Packets sent to a
1983.Em queue
1984are first grouped into flows according to a mask on the 5-tuple.
1985Flows are then passed to the scheduler associated to the
1986.Em queue ,
1987and each flow uses scheduling parameters (weight and others)
1988as configured in the
1989.Em queue
1990itself.
1991A scheduler in turn is connected to an emulated link,
1992and arbitrates the link's bandwidth among backlogged flows according to
1993weights and to the features of the scheduling algorithm in use.
1994.El
1995.Pp
1996In practice,
1997.Em pipes
1998can be used to set hard limits to the bandwidth that a flow can use, whereas
1999.Em queues
2000can be used to determine how different flows share the available bandwidth.
2001.Pp
2002A graphical representation of the binding of queues,
2003flows, schedulers and links is below.
2004.Bd -literal -offset indent
2005                 (flow_mask|sched_mask)  sched_mask
2006         +---------+   weight Wx  +-------------+
2007         |         |->-[flow]-->--|             |-+
2008    -->--| QUEUE x |   ...        |             | |
2009         |         |->-[flow]-->--| SCHEDuler N | |
2010         +---------+              |             | |
2011             ...                  |             +--[LINK N]-->--
2012         +---------+   weight Wy  |             | +--[LINK N]-->--
2013         |         |->-[flow]-->--|             | |
2014    -->--| QUEUE y |   ...        |             | |
2015         |         |->-[flow]-->--|             | |
2016         +---------+              +-------------+ |
2017                                    +-------------+
2018.Ed
2019It is important to understand the role of the SCHED_MASK
2020and FLOW_MASK, which are configured through the commands
2021.Dl "ipfw sched N config mask SCHED_MASK ..."
2022and
2023.Dl "ipfw queue X config mask FLOW_MASK ..." .
2024.Pp
2025The SCHED_MASK is used to assign flows to one or more
2026scheduler instances, one for each
2027value of the packet's 5-tuple after applying SCHED_MASK.
2028As an example, using ``src-ip 0xffffff00'' creates one instance
2029for each /24 destination subnet.
2030.Pp
2031The FLOW_MASK, together with the SCHED_MASK, is used to split
2032packets into flows.
2033As an example, using
2034``src-ip 0x000000ff''
2035together with the previous SCHED_MASK makes a flow for
2036each individual source address.
2037In turn, flows for each /24
2038subnet will be sent to the same scheduler instance.
2039.Pp
2040The above diagram holds even for the
2041.Em pipe
2042case, with the only restriction that a
2043.Em pipe
2044only supports a SCHED_MASK, and forces the use of a FIFO
2045scheduler (these are for backward compatibility reasons;
2046in fact, internally, a
2047.Nm dummynet's
2048pipe is implemented exactly as above).
2049.Pp
2050There are two modes of
2051.Nm dummynet
2052operation:
2053.Dq normal
2054and
2055.Dq fast .
2056The
2057.Dq normal
2058mode tries to emulate a real link: the
2059.Nm dummynet
2060scheduler ensures that the packet will not leave the pipe faster than it
2061would on the real link with a given bandwidth.
2062The
2063.Dq fast
2064mode allows certain packets to bypass the
2065.Nm dummynet
2066scheduler (if packet flow does not exceed pipe's bandwidth).
2067This is the reason why the
2068.Dq fast
2069mode requires less CPU cycles per packet (on average) and packet latency
2070can be significantly lower in comparison to a real link with the same
2071bandwidth.
2072The default mode is
2073.Dq normal .
2074The
2075.Dq fast
2076mode can be enabled by setting the
2077.Va net.inet.ip.dummynet.io_fast
2078.Xr sysctl 8
2079variable to a non-zero value.
2080.Pp
2081.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
2082The
2083.Em pipe ,
2084.Em queue
2085and
2086.Em scheduler
2087configuration commands are the following:
2088.Bd -ragged -offset indent
2089.Cm pipe Ar number Cm config Ar pipe-configuration
2090.Pp
2091.Cm queue Ar number Cm config Ar queue-configuration
2092.Pp
2093.Cm sched Ar number Cm config Ar sched-configuration
2094.Ed
2095.Pp
2096The following parameters can be configured for a pipe:
2097.Pp
2098.Bl -tag -width indent -compact
2099.It Cm bw Ar bandwidth | device
2100Bandwidth, measured in
2101.Sm off
2102.Op Cm K | M
2103.Brq Cm bit/s | Byte/s .
2104.Sm on
2105.Pp
2106A value of 0 (default) means unlimited bandwidth.
2107The unit must immediately follow the number, as in
2108.Pp
2109.Dl "ipfw pipe 1 config bw 300Kbit/s"
2110.Pp
2111If a device name is specified instead of a numeric value, as in
2112.Pp
2113.Dl "ipfw pipe 1 config bw tun0"
2114.Pp
2115then the transmit clock is supplied by the specified device.
2116At the moment only the
2117.Xr tun 4
2118device supports this
2119functionality, for use in conjunction with
2120.Xr ppp 8 .
2121.Pp
2122.It Cm delay Ar ms-delay
2123Propagation delay, measured in milliseconds.
2124The value is rounded to the next multiple of the clock tick
2125(typically 10ms, but it is a good practice to run kernels
2126with
2127.Dq "options HZ=1000"
2128to reduce
2129the granularity to 1ms or less).
2130The default value is 0, meaning no delay.
2131.Pp
2132.It Cm burst Ar size
2133If the data to be sent exceeds the pipe's bandwidth limit
2134(and the pipe was previously idle), up to
2135.Ar size
2136bytes of data are allowed to bypass the
2137.Nm dummynet
2138scheduler, and will be sent as fast as the physical link allows.
2139Any additional data will be transmitted at the rate specified
2140by the
2141.Nm pipe
2142bandwidth.
2143The burst size depends on how long the pipe has been idle;
2144the effective burst size is calculated as follows:
2145MAX(
2146.Ar size
2147,
2148.Nm bw
2149* pipe_idle_time).
2150.Pp
2151.It Cm profile Ar filename
2152A file specifying the additional overhead incurred in the transmission
2153of a packet on the link.
2154.Pp
2155Some link types introduce extra delays in the transmission
2156of a packet, e.g., because of MAC level framing, contention on
2157the use of the channel, MAC level retransmissions and so on.
2158From our point of view, the channel is effectively unavailable
2159for this extra time, which is constant or variable depending
2160on the link type.
2161Additionally, packets may be dropped after this
2162time (e.g., on a wireless link after too many retransmissions).
2163We can model the additional delay with an empirical curve
2164that represents its distribution.
2165.Bd -literal -offset indent
2166      cumulative probability
2167      1.0 ^
2168          |
2169      L   +-- loss-level          x
2170          |                 ******
2171          |                *
2172          |           *****
2173          |          *
2174          |        **
2175          |       *
2176          +-------*------------------->
2177                      delay
2178.Ed
2179The empirical curve may have both vertical and horizontal lines.
2180Vertical lines represent constant delay for a range of
2181probabilities.
2182Horizontal lines correspond to a discontinuity in the delay
2183distribution: the pipe will use the largest delay for a
2184given probability.
2185.Pp
2186The file format is the following, with whitespace acting as
2187a separator and '#' indicating the beginning a comment:
2188.Bl -tag -width indent
2189.It Cm name Ar identifier
2190optional name (listed by "ipfw pipe show")
2191to identify the delay distribution;
2192.It Cm bw Ar value
2193the bandwidth used for the pipe.
2194If not specified here, it must be present
2195explicitly as a configuration parameter for the pipe;
2196.It Cm loss-level Ar L
2197the probability above which packets are lost.
2198(0.0 <= L <= 1.0, default 1.0 i.e., no loss);
2199.It Cm samples Ar N
2200the number of samples used in the internal
2201representation of the curve (2..1024; default 100);
2202.It Cm "delay prob" | "prob delay"
2203One of these two lines is mandatory and defines
2204the format of the following lines with data points.
2205.It Ar XXX Ar YYY
22062 or more lines representing points in the curve,
2207with either delay or probability first, according
2208to the chosen format.
2209The unit for delay is milliseconds.
2210Data points do not need to be sorted.
2211Also, the number of actual lines can be different
2212from the value of the "samples" parameter:
2213.Nm
2214utility will sort and interpolate
2215the curve as needed.
2216.El
2217.Pp
2218Example of a profile file:
2219.Bd -literal -offset indent
2220name    bla_bla_bla
2221samples 100
2222loss-level    0.86
2223prob    delay
22240       200	# minimum overhead is 200ms
22250.5     200
22260.5     300
22270.8     1000
22280.9     1300
22291       1300
2230#configuration file end
2231.Ed
2232.El
2233.Pp
2234The following parameters can be configured for a queue:
2235.Pp
2236.Bl -tag -width indent -compact
2237.It Cm pipe Ar pipe_nr
2238Connects a queue to the specified pipe.
2239Multiple queues (with the same or different weights) can be connected to
2240the same pipe, which specifies the aggregate rate for the set of queues.
2241.Pp
2242.It Cm weight Ar weight
2243Specifies the weight to be used for flows matching this queue.
2244The weight must be in the range 1..100, and defaults to 1.
2245.El
2246.Pp
2247The following case-insensitive parameters can be configured for a
2248scheduler:
2249.Pp
2250.Bl -tag -width indent -compact
2251.It Cm type Ar {fifo | wf2q+ | rr | qfq}
2252specifies the scheduling algorithm to use.
2253.Bl -tag -width indent -compact
2254.It Cm fifo
2255is just a FIFO scheduler (which means that all packets
2256are stored in the same queue as they arrive to the scheduler).
2257FIFO has O(1) per-packet time complexity, with very low
2258constants (estimate 60-80ns on a 2GHz desktop machine)
2259but gives no service guarantees.
2260.It Cm wf2q+
2261implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2262algorithm which permits flows to share bandwidth according to
2263their weights.
2264Note that weights are not priorities; even a flow
2265with a minuscule weight will never starve.
2266WF2Q+ has O(log N) per-packet processing cost, where N is the number
2267of flows, and is the default algorithm used by previous versions
2268dummynet's queues.
2269.It Cm rr
2270implements the Deficit Round Robin algorithm, which has O(1) processing
2271costs (roughly, 100-150ns per packet)
2272and permits bandwidth allocation according to weights, but
2273with poor service guarantees.
2274.It Cm qfq
2275implements the QFQ algorithm, which is a very fast variant of
2276WF2Q+, with similar service guarantees and O(1) processing
2277costs (roughly, 200-250ns per packet).
2278.El
2279.El
2280.Pp
2281In addition to the type, all parameters allowed for a pipe can also
2282be specified for a scheduler.
2283.Pp
2284Finally, the following parameters can be configured for both
2285pipes and queues:
2286.Pp
2287.Bl -tag -width XXXX -compact
2288.It Cm buckets Ar hash-table-size
2289Specifies the size of the hash table used for storing the
2290various queues.
2291Default value is 64 controlled by the
2292.Xr sysctl 8
2293variable
2294.Va net.inet.ip.dummynet.hash_size ,
2295allowed range is 16 to 65536.
2296.Pp
2297.It Cm mask Ar mask-specifier
2298Packets sent to a given pipe or queue by an
2299.Nm
2300rule can be further classified into multiple flows, each of which is then
2301sent to a different
2302.Em dynamic
2303pipe or queue.
2304A flow identifier is constructed by masking the IP addresses,
2305ports and protocol types as specified with the
2306.Cm mask
2307options in the configuration of the pipe or queue.
2308For each different flow identifier, a new pipe or queue is created
2309with the same parameters as the original object, and matching packets
2310are sent to it.
2311.Pp
2312Thus, when
2313.Em dynamic pipes
2314are used, each flow will get the same bandwidth as defined by the pipe,
2315whereas when
2316.Em dynamic queues
2317are used, each flow will share the parent's pipe bandwidth evenly
2318with other flows generated by the same queue (note that other queues
2319with different weights might be connected to the same pipe).
2320.br
2321Available mask specifiers are a combination of one or more of the following:
2322.Pp
2323.Cm dst-ip Ar mask ,
2324.Cm dst-ip6 Ar mask ,
2325.Cm src-ip Ar mask ,
2326.Cm src-ip6 Ar mask ,
2327.Cm dst-port Ar mask ,
2328.Cm src-port Ar mask ,
2329.Cm flow-id Ar mask ,
2330.Cm proto Ar mask
2331or
2332.Cm all ,
2333.Pp
2334where the latter means all bits in all fields are significant.
2335.Pp
2336.It Cm noerror
2337When a packet is dropped by a
2338.Nm dummynet
2339queue or pipe, the error
2340is normally reported to the caller routine in the kernel, in the
2341same way as it happens when a device queue fills up.
2342Setting this
2343option reports the packet as successfully delivered, which can be
2344needed for some experimental setups where you want to simulate
2345loss or congestion at a remote router.
2346.Pp
2347.It Cm plr Ar packet-loss-rate
2348Packet loss rate.
2349Argument
2350.Ar packet-loss-rate
2351is a floating-point number between 0 and 1, with 0 meaning no
2352loss, 1 meaning 100% loss.
2353The loss rate is internally represented on 31 bits.
2354.Pp
2355.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2356Queue size, in
2357.Ar slots
2358or
2359.Cm KBytes .
2360Default value is 50 slots, which
2361is the typical queue size for Ethernet devices.
2362Note that for slow speed links you should keep the queue
2363size short or your traffic might be affected by a significant
2364queueing delay.
2365E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2366or 20s of queue on a 30Kbit/s pipe.
2367Even worse effects can result if you get packets from an
2368interface with a much larger MTU, e.g.\& the loopback interface
2369with its 16KB packets.
2370The
2371.Xr sysctl 8
2372variables
2373.Em net.inet.ip.dummynet.pipe_byte_limit
2374and
2375.Em net.inet.ip.dummynet.pipe_slot_limit
2376control the maximum lengths that can be specified.
2377.Pp
2378.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2379Make use of the RED (Random Early Detection) queue management algorithm.
2380.Ar w_q
2381and
2382.Ar max_p
2383are floating
2384point numbers between 0 and 1 (0 not included), while
2385.Ar min_th
2386and
2387.Ar max_th
2388are integer numbers specifying thresholds for queue management
2389(thresholds are computed in bytes if the queue has been defined
2390in bytes, in slots otherwise).
2391The
2392.Nm dummynet
2393also supports the gentle RED variant (gred).
2394Three
2395.Xr sysctl 8
2396variables can be used to control the RED behaviour:
2397.Bl -tag -width indent
2398.It Va net.inet.ip.dummynet.red_lookup_depth
2399specifies the accuracy in computing the average queue
2400when the link is idle (defaults to 256, must be greater than zero)
2401.It Va net.inet.ip.dummynet.red_avg_pkt_size
2402specifies the expected average packet size (defaults to 512, must be
2403greater than zero)
2404.It Va net.inet.ip.dummynet.red_max_pkt_size
2405specifies the expected maximum packet size, only used when queue
2406thresholds are in bytes (defaults to 1500, must be greater than zero).
2407.El
2408.El
2409.Pp
2410When used with IPv6 data,
2411.Nm dummynet
2412currently has several limitations.
2413Information necessary to route link-local packets to an
2414interface is not available after processing by
2415.Nm dummynet
2416so those packets are dropped in the output path.
2417Care should be taken to ensure that link-local packets are not passed to
2418.Nm dummynet .
2419.Sh CHECKLIST
2420Here are some important points to consider when designing your
2421rules:
2422.Bl -bullet
2423.It
2424Remember that you filter both packets going
2425.Cm in
2426and
2427.Cm out .
2428Most connections need packets going in both directions.
2429.It
2430Remember to test very carefully.
2431It is a good idea to be near the console when doing this.
2432If you cannot be near the console,
2433use an auto-recovery script such as the one in
2434.Pa /usr/share/examples/ipfw/change_rules.sh .
2435.It
2436Do not forget the loopback interface.
2437.El
2438.Sh FINE POINTS
2439.Bl -bullet
2440.It
2441There are circumstances where fragmented datagrams are unconditionally
2442dropped.
2443TCP packets are dropped if they do not contain at least 20 bytes of
2444TCP header, UDP packets are dropped if they do not contain a full 8
2445byte UDP header, and ICMP packets are dropped if they do not contain
24464 bytes of ICMP header, enough to specify the ICMP type, code, and
2447checksum.
2448These packets are simply logged as
2449.Dq pullup failed
2450since there may not be enough good data in the packet to produce a
2451meaningful log entry.
2452.It
2453Another type of packet is unconditionally dropped, a TCP packet with a
2454fragment offset of one.
2455This is a valid packet, but it only has one use, to try
2456to circumvent firewalls.
2457When logging is enabled, these packets are
2458reported as being dropped by rule -1.
2459.It
2460If you are logged in over a network, loading the
2461.Xr kld 4
2462version of
2463.Nm
2464is probably not as straightforward as you would think.
2465The following command line is recommended:
2466.Bd -literal -offset indent
2467kldload ipfw && \e
2468ipfw add 32000 allow ip from any to any
2469.Ed
2470.Pp
2471Along the same lines, doing an
2472.Bd -literal -offset indent
2473ipfw flush
2474.Ed
2475.Pp
2476in similar surroundings is also a bad idea.
2477.It
2478The
2479.Nm
2480filter list may not be modified if the system security level
2481is set to 3 or higher
2482(see
2483.Xr init 8
2484for information on system security levels).
2485.El
2486.Sh PACKET DIVERSION
2487A
2488.Xr divert 4
2489socket bound to the specified port will receive all packets
2490diverted to that port.
2491If no socket is bound to the destination port, or if the divert module is
2492not loaded, or if the kernel was not compiled with divert socket support,
2493the packets are dropped.
2494.Sh NETWORK ADDRESS TRANSLATION (NAT)
2495.Nm
2496support in-kernel NAT using the kernel version of
2497.Xr libalias 3 .
2498.Pp
2499The nat configuration command is the following:
2500.Bd -ragged -offset indent
2501.Bk -words
2502.Cm nat
2503.Ar nat_number
2504.Cm config
2505.Ar nat-configuration
2506.Ek
2507.Ed
2508.Pp
2509The following parameters can be configured:
2510.Bl -tag -width indent
2511.It Cm ip Ar ip_address
2512Define an ip address to use for aliasing.
2513.It Cm if Ar nic
2514Use ip address of NIC for aliasing, dynamically changing
2515it if NIC's ip address changes.
2516.It Cm log
2517Enable logging on this nat instance.
2518.It Cm deny_in
2519Deny any incoming connection from outside world.
2520.It Cm same_ports
2521Try to leave the alias port numbers unchanged from
2522the actual local port numbers.
2523.It Cm unreg_only
2524Traffic on the local network not originating from an
2525unregistered address spaces will be ignored.
2526.It Cm reset
2527Reset table of the packet aliasing engine on address change.
2528.It Cm reverse
2529Reverse the way libalias handles aliasing.
2530.It Cm proxy_only
2531Obey transparent proxy rules only, packet aliasing is not performed.
2532.It Cm skip_global
2533Skip instance in case of global state lookup (see below).
2534.El
2535.Pp
2536Some specials value can be supplied instead of
2537.Va nat_number:
2538.Bl -tag -width indent
2539.It Cm global
2540Looks up translation state in all configured nat instances.
2541If an entry is found, packet is aliased according to that entry.
2542If no entry was found in any of the instances, packet is passed unchanged,
2543and no new entry will be created.
2544See section
2545.Sx MULTIPLE INSTANCES
2546in
2547.Xr natd 8
2548for more information.
2549.It Cm tablearg
2550Uses argument supplied in lookup table.
2551See
2552.Sx LOOKUP TABLES
2553section below for more information on lookup tables.
2554.El
2555.Pp
2556To let the packet continue after being (de)aliased, set the sysctl variable
2557.Va net.inet.ip.fw.one_pass
2558to 0.
2559For more information about aliasing modes, refer to
2560.Xr libalias 3 .
2561See Section
2562.Sx EXAMPLES
2563for some examples about nat usage.
2564.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
2565Redirect and LSNAT support follow closely the syntax used in
2566.Xr natd 8 .
2567See Section
2568.Sx EXAMPLES
2569for some examples on how to do redirect and lsnat.
2570.Ss SCTP NAT SUPPORT
2571SCTP nat can be configured in a similar manner to TCP through the
2572.Nm
2573command line tool.
2574The main difference is that
2575.Nm sctp nat
2576does not do port translation.
2577Since the local and global side ports will be the same,
2578there is no need to specify both.
2579Ports are redirected as follows:
2580.Bd -ragged -offset indent
2581.Bk -words
2582.Cm nat
2583.Ar nat_number
2584.Cm config if
2585.Ar nic
2586.Cm redirect_port sctp
2587.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
2588.Ek
2589.Ed
2590.Pp
2591Most
2592.Nm sctp nat
2593configuration can be done in real-time through the
2594.Xr sysctl 8
2595interface.
2596All may be changed dynamically, though the hash_table size will only
2597change for new
2598.Nm nat
2599instances.
2600See
2601.Sx SYSCTL VARIABLES
2602for more info.
2603.Sh LOADER TUNABLES
2604Tunables can be set in
2605.Xr loader 8
2606prompt,
2607.Xr loader.conf 5
2608or
2609.Xr kenv 1
2610before ipfw module gets loaded.
2611.Bl -tag -width indent
2612.It Va net.inet.ip.fw.default_to_accept: No 0
2613Defines ipfw last rule behavior.
2614This value overrides
2615.Cd "options IPFW_DEFAULT_TO_(ACCEPT|DENY)"
2616from kernel configuration file.
2617.It Va net.inet.ip.fw.tables_max: No 128
2618Defines number of tables available in ipfw.
2619Number cannot exceed 65534.
2620.El
2621.Sh SYSCTL VARIABLES
2622A set of
2623.Xr sysctl 8
2624variables controls the behaviour of the firewall and
2625associated modules
2626.Pq Nm dummynet , bridge , sctp nat .
2627These are shown below together with their default value
2628(but always check with the
2629.Xr sysctl 8
2630command what value is actually in use) and meaning:
2631.Bl -tag -width indent
2632.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
2633Defines how the
2634.Nm nat
2635responds to receipt of global OOTB ASCONF-AddIP:
2636.Bl -tag -width indent
2637.It Cm 0
2638No response (unless a partially matching association exists -
2639ports and vtags match but global address does not)
2640.It Cm 1
2641.Nm nat
2642will accept and process all OOTB global AddIP messages.
2643.El
2644.Pp
2645Option 1 should never be selected as this forms a security risk.
2646An attacker can
2647establish multiple fake associations by sending AddIP messages.
2648.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
2649Defines the maximum number of chunks in an SCTP packet that will be
2650parsed for a
2651packet that matches an existing association.
2652This value is enforced to be greater or equal than
2653.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
2654A high value is
2655a DoS risk yet setting too low a value may result in
2656important control chunks in
2657the packet not being located and parsed.
2658.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
2659Defines when the
2660.Nm nat
2661responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
2662An OOTB packet is a packet that arrives with no existing association
2663registered in the
2664.Nm nat
2665and is not an INIT or ASCONF-AddIP packet:
2666.Bl -tag -width indent
2667.It Cm 0
2668ErrorM is never sent in response to OOTB packets.
2669.It Cm 1
2670ErrorM is only sent to OOTB packets received on the local side.
2671.It Cm 2
2672ErrorM is sent to the local side and on the global side ONLY if there is a
2673partial match (ports and vtags match but the source global IP does not).
2674This value is only useful if the
2675.Nm nat
2676is tracking global IP addresses.
2677.It Cm 3
2678ErrorM is sent in response to all OOTB packets on both
2679the local and global side
2680(DoS risk).
2681.El
2682.Pp
2683At the moment the default is 0, since the ErrorM packet is not yet
2684supported by most SCTP stacks.
2685When it is supported, and if not tracking
2686global addresses, we recommend setting this value to 1 to allow
2687multi-homed local hosts to function with the
2688.Nm nat .
2689To track global addresses, we recommend setting this value to 2 to
2690allow global hosts to be informed when they need to (re)send an
2691ASCONF-AddIP.
2692Value 3 should never be chosen (except for debugging) as the
2693.Nm nat
2694will respond to all OOTB global packets (a DoS risk).
2695.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
2696Size of hash tables used for
2697.Nm nat
2698lookups (100 < prime_number > 1000001).
2699This value sets the
2700.Nm hash table
2701size for any future created
2702.Nm nat
2703instance and therefore must be set prior to creating a
2704.Nm nat
2705instance.
2706The table sizes may be changed to suit specific needs.
2707If there will be few
2708concurrent associations, and memory is scarce, you may make these smaller.
2709If there will be many thousands (or millions) of concurrent associations, you
2710should make these larger.
2711A prime number is best for the table size.
2712The sysctl
2713update function will adjust your input value to the next highest prime number.
2714.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
2715Hold association in table for this many seconds after receiving a
2716SHUTDOWN-COMPLETE.
2717This allows endpoints to correct shutdown gracefully if a
2718shutdown_complete is lost and retransmissions are required.
2719.It Va net.inet.ip.alias.sctp.init_timer: No 15
2720Timeout value while waiting for (INIT-ACK|AddIP-ACK).
2721This value cannot be 0.
2722.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
2723Defines the maximum number of chunks in an SCTP packet that will be parsed when
2724no existing association exists that matches that packet.
2725Ideally this packet
2726will only be an INIT or ASCONF-AddIP packet.
2727A higher value may become a DoS
2728risk as malformed packets can consume processing resources.
2729.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
2730Defines the maximum number of parameters within a chunk that will be
2731parsed in a
2732packet.
2733As for other similar sysctl variables, larger values pose a DoS risk.
2734.It Va net.inet.ip.alias.sctp.log_level: No 0
2735Level of detail in the system log messages (0 \- minimal, 1 \- event,
27362 \- info, 3 \- detail, 4 \- debug, 5 \- max debug).
2737May be a good
2738option in high loss environments.
2739.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
2740Timeout value while waiting for SHUTDOWN-COMPLETE.
2741This value cannot be 0.
2742.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
2743Enables/disables global IP address tracking within the
2744.Nm nat
2745and places an
2746upper limit on the number of addresses tracked for each association:
2747.Bl -tag -width indent
2748.It Cm 0
2749Global tracking is disabled
2750.It Cm >1
2751Enables tracking, the maximum number of addresses tracked for each
2752association is limited to this value
2753.El
2754.Pp
2755This variable is fully dynamic, the new value will be adopted for all newly
2756arriving associations, existing associations are treated
2757as they were previously.
2758Global tracking will decrease the number of collisions within the
2759.Nm nat
2760at a cost
2761of increased processing load, memory usage, complexity, and possible
2762.Nm nat
2763state
2764problems in complex networks with multiple
2765.Nm nats .
2766We recommend not tracking
2767global IP addresses, this will still result in a fully functional
2768.Nm nat .
2769.It Va net.inet.ip.alias.sctp.up_timer: No 300
2770Timeout value to keep an association up with no traffic.
2771This value cannot be 0.
2772.It Va net.inet.ip.dummynet.expire : No 1
2773Lazily delete dynamic pipes/queue once they have no pending traffic.
2774You can disable this by setting the variable to 0, in which case
2775the pipes/queues will only be deleted when the threshold is reached.
2776.It Va net.inet.ip.dummynet.hash_size : No 64
2777Default size of the hash table used for dynamic pipes/queues.
2778This value is used when no
2779.Cm buckets
2780option is specified when configuring a pipe/queue.
2781.It Va net.inet.ip.dummynet.io_fast : No 0
2782If set to a non-zero value,
2783the
2784.Dq fast
2785mode of
2786.Nm dummynet
2787operation (see above) is enabled.
2788.It Va net.inet.ip.dummynet.io_pkt
2789Number of packets passed to
2790.Nm dummynet .
2791.It Va net.inet.ip.dummynet.io_pkt_drop
2792Number of packets dropped by
2793.Nm dummynet .
2794.It Va net.inet.ip.dummynet.io_pkt_fast
2795Number of packets bypassed by the
2796.Nm dummynet
2797scheduler.
2798.It Va net.inet.ip.dummynet.max_chain_len : No 16
2799Target value for the maximum number of pipes/queues in a hash bucket.
2800The product
2801.Cm max_chain_len*hash_size
2802is used to determine the threshold over which empty pipes/queues
2803will be expired even when
2804.Cm net.inet.ip.dummynet.expire=0 .
2805.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
2806.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
2807.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
2808Parameters used in the computations of the drop probability
2809for the RED algorithm.
2810.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
2811.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
2812The maximum queue size that can be specified in bytes or packets.
2813These limits prevent accidental exhaustion of resources such as mbufs.
2814If you raise these limits,
2815you should make sure the system is configured so that sufficient resources
2816are available.
2817.It Va net.inet.ip.fw.autoinc_step : No 100
2818Delta between rule numbers when auto-generating them.
2819The value must be in the range 1..1000.
2820.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
2821The current number of buckets in the hash table for dynamic rules
2822(readonly).
2823.It Va net.inet.ip.fw.debug : No 1
2824Controls debugging messages produced by
2825.Nm .
2826.It Va net.inet.ip.fw.default_rule : No 65535
2827The default rule number (read-only).
2828By the design of
2829.Nm , the default rule is the last one, so its number
2830can also serve as the highest number allowed for a rule.
2831.It Va net.inet.ip.fw.dyn_buckets : No 256
2832The number of buckets in the hash table for dynamic rules.
2833Must be a power of 2, up to 65536.
2834It only takes effect when all dynamic rules have expired, so you
2835are advised to use a
2836.Cm flush
2837command to make sure that the hash table is resized.
2838.It Va net.inet.ip.fw.dyn_count : No 3
2839Current number of dynamic rules
2840(read-only).
2841.It Va net.inet.ip.fw.dyn_keepalive : No 1
2842Enables generation of keepalive packets for
2843.Cm keep-state
2844rules on TCP sessions.
2845A keepalive is generated to both
2846sides of the connection every 5 seconds for the last 20
2847seconds of the lifetime of the rule.
2848.It Va net.inet.ip.fw.dyn_max : No 8192
2849Maximum number of dynamic rules.
2850When you hit this limit, no more dynamic rules can be
2851installed until old ones expire.
2852.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
2853.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
2854.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
2855.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
2856.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
2857.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
2858These variables control the lifetime, in seconds, of dynamic
2859rules.
2860Upon the initial SYN exchange the lifetime is kept short,
2861then increased after both SYN have been seen, then decreased
2862again during the final FIN exchange or when a RST is received.
2863Both
2864.Em dyn_fin_lifetime
2865and
2866.Em dyn_rst_lifetime
2867must be strictly lower than 5 seconds, the period of
2868repetition of keepalives.
2869The firewall enforces that.
2870.It Va net.inet.ip.fw.enable : No 1
2871Enables the firewall.
2872Setting this variable to 0 lets you run your machine without
2873firewall even if compiled in.
2874.It Va net.inet6.ip6.fw.enable : No 1
2875provides the same functionality as above for the IPv6 case.
2876.It Va net.inet.ip.fw.one_pass : No 1
2877When set, the packet exiting from the
2878.Nm dummynet
2879pipe or from
2880.Xr ng_ipfw 4
2881node is not passed though the firewall again.
2882Otherwise, after an action, the packet is
2883reinjected into the firewall at the next rule.
2884.It Va net.inet.ip.fw.tables_max : No 128
2885Maximum number of tables.
2886.It Va net.inet.ip.fw.verbose : No 1
2887Enables verbose messages.
2888.It Va net.inet.ip.fw.verbose_limit : No 0
2889Limits the number of messages produced by a verbose firewall.
2890.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
2891If enabled packets with unknown IPv6 Extension Headers will be denied.
2892.It Va net.link.ether.ipfw : No 0
2893Controls whether layer-2 packets are passed to
2894.Nm .
2895Default is no.
2896.It Va net.link.bridge.ipfw : No 0
2897Controls whether bridged packets are passed to
2898.Nm .
2899Default is no.
2900.El
2901.Sh EXAMPLES
2902There are far too many possible uses of
2903.Nm
2904so this Section will only give a small set of examples.
2905.Pp
2906.Ss BASIC PACKET FILTERING
2907This command adds an entry which denies all tcp packets from
2908.Em cracker.evil.org
2909to the telnet port of
2910.Em wolf.tambov.su
2911from being forwarded by the host:
2912.Pp
2913.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
2914.Pp
2915This one disallows any connection from the entire cracker's
2916network to my host:
2917.Pp
2918.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
2919.Pp
2920A first and efficient way to limit access (not using dynamic rules)
2921is the use of the following rules:
2922.Pp
2923.Dl "ipfw add allow tcp from any to any established"
2924.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
2925.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
2926.Dl "..."
2927.Dl "ipfw add deny tcp from any to any"
2928.Pp
2929The first rule will be a quick match for normal TCP packets,
2930but it will not match the initial SYN packet, which will be
2931matched by the
2932.Cm setup
2933rules only for selected source/destination pairs.
2934All other SYN packets will be rejected by the final
2935.Cm deny
2936rule.
2937.Pp
2938If you administer one or more subnets, you can take advantage
2939of the address sets and or-blocks and write extremely
2940compact rulesets which selectively enable services to blocks
2941of clients, as below:
2942.Pp
2943.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
2944.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
2945.Dl ""
2946.Dl "ipfw add allow ip from ${goodguys} to any"
2947.Dl "ipfw add deny ip from ${badguys} to any"
2948.Dl "... normal policies ..."
2949.Pp
2950The
2951.Cm verrevpath
2952option could be used to do automated anti-spoofing by adding the
2953following to the top of a ruleset:
2954.Pp
2955.Dl "ipfw add deny ip from any to any not verrevpath in"
2956.Pp
2957This rule drops all incoming packets that appear to be coming to the
2958system on the wrong interface.
2959For example, a packet with a source
2960address belonging to a host on a protected internal network would be
2961dropped if it tried to enter the system from an external interface.
2962.Pp
2963The
2964.Cm antispoof
2965option could be used to do similar but more restricted anti-spoofing
2966by adding the following to the top of a ruleset:
2967.Pp
2968.Dl "ipfw add deny ip from any to any not antispoof in"
2969.Pp
2970This rule drops all incoming packets that appear to be coming from another
2971directly connected system but on the wrong interface.
2972For example, a packet with a source address of
2973.Li 192.168.0.0/24 ,
2974configured on
2975.Li fxp0 ,
2976but coming in on
2977.Li fxp1
2978would be dropped.
2979.Ss DYNAMIC RULES
2980In order to protect a site from flood attacks involving fake
2981TCP packets, it is safer to use dynamic rules:
2982.Pp
2983.Dl "ipfw add check-state"
2984.Dl "ipfw add deny tcp from any to any established"
2985.Dl "ipfw add allow tcp from my-net to any setup keep-state"
2986.Pp
2987This will let the firewall install dynamic rules only for
2988those connection which start with a regular SYN packet coming
2989from the inside of our network.
2990Dynamic rules are checked when encountering the first
2991occurrence of a
2992.Cm check-state ,
2993.Cm keep-state
2994or
2995.Cm limit
2996rule.
2997A
2998.Cm check-state
2999rule should usually be placed near the beginning of the
3000ruleset to minimize the amount of work scanning the ruleset.
3001Your mileage may vary.
3002.Pp
3003To limit the number of connections a user can open
3004you can use the following type of rules:
3005.Pp
3006.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
3007.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
3008.Pp
3009The former (assuming it runs on a gateway) will allow each host
3010on a /24 network to open at most 10 TCP connections.
3011The latter can be placed on a server to make sure that a single
3012client does not use more than 4 simultaneous connections.
3013.Pp
3014.Em BEWARE :
3015stateful rules can be subject to denial-of-service attacks
3016by a SYN-flood which opens a huge number of dynamic rules.
3017The effects of such attacks can be partially limited by
3018acting on a set of
3019.Xr sysctl 8
3020variables which control the operation of the firewall.
3021.Pp
3022Here is a good usage of the
3023.Cm list
3024command to see accounting records and timestamp information:
3025.Pp
3026.Dl ipfw -at list
3027.Pp
3028or in short form without timestamps:
3029.Pp
3030.Dl ipfw -a list
3031.Pp
3032which is equivalent to:
3033.Pp
3034.Dl ipfw show
3035.Pp
3036Next rule diverts all incoming packets from 192.168.2.0/24
3037to divert port 5000:
3038.Pp
3039.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
3040.Ss TRAFFIC SHAPING
3041The following rules show some of the applications of
3042.Nm
3043and
3044.Nm dummynet
3045for simulations and the like.
3046.Pp
3047This rule drops random incoming packets with a probability
3048of 5%:
3049.Pp
3050.Dl "ipfw add prob 0.05 deny ip from any to any in"
3051.Pp
3052A similar effect can be achieved making use of
3053.Nm dummynet
3054pipes:
3055.Pp
3056.Dl "ipfw add pipe 10 ip from any to any"
3057.Dl "ipfw pipe 10 config plr 0.05"
3058.Pp
3059We can use pipes to artificially limit bandwidth, e.g.\& on a
3060machine acting as a router, if we want to limit traffic from
3061local clients on 192.168.2.0/24 we do:
3062.Pp
3063.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3064.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
3065.Pp
3066note that we use the
3067.Cm out
3068modifier so that the rule is not used twice.
3069Remember in fact that
3070.Nm
3071rules are checked both on incoming and outgoing packets.
3072.Pp
3073Should we want to simulate a bidirectional link with bandwidth
3074limitations, the correct way is the following:
3075.Pp
3076.Dl "ipfw add pipe 1 ip from any to any out"
3077.Dl "ipfw add pipe 2 ip from any to any in"
3078.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
3079.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
3080.Pp
3081The above can be very useful, e.g.\& if you want to see how
3082your fancy Web page will look for a residential user who
3083is connected only through a slow link.
3084You should not use only one pipe for both directions, unless
3085you want to simulate a half-duplex medium (e.g.\& AppleTalk,
3086Ethernet, IRDA).
3087It is not necessary that both pipes have the same configuration,
3088so we can also simulate asymmetric links.
3089.Pp
3090Should we want to verify network performance with the RED queue
3091management algorithm:
3092.Pp
3093.Dl "ipfw add pipe 1 ip from any to any"
3094.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
3095.Pp
3096Another typical application of the traffic shaper is to
3097introduce some delay in the communication.
3098This can significantly affect applications which do a lot of Remote
3099Procedure Calls, and where the round-trip-time of the
3100connection often becomes a limiting factor much more than
3101bandwidth:
3102.Pp
3103.Dl "ipfw add pipe 1 ip from any to any out"
3104.Dl "ipfw add pipe 2 ip from any to any in"
3105.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
3106.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
3107.Pp
3108Per-flow queueing can be useful for a variety of purposes.
3109A very simple one is counting traffic:
3110.Pp
3111.Dl "ipfw add pipe 1 tcp from any to any"
3112.Dl "ipfw add pipe 1 udp from any to any"
3113.Dl "ipfw add pipe 1 ip from any to any"
3114.Dl "ipfw pipe 1 config mask all"
3115.Pp
3116The above set of rules will create queues (and collect
3117statistics) for all traffic.
3118Because the pipes have no limitations, the only effect is
3119collecting statistics.
3120Note that we need 3 rules, not just the last one, because
3121when
3122.Nm
3123tries to match IP packets it will not consider ports, so we
3124would not see connections on separate ports as different
3125ones.
3126.Pp
3127A more sophisticated example is limiting the outbound traffic
3128on a net with per-host limits, rather than per-network limits:
3129.Pp
3130.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
3131.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
3132.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3133.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
3134.Ss LOOKUP TABLES
3135In the following example, we need to create several traffic bandwidth
3136classes and we need different hosts/networks to fall into different classes.
3137We create one pipe for each class and configure them accordingly.
3138Then we create a single table and fill it with IP subnets and addresses.
3139For each subnet/host we set the argument equal to the number of the pipe
3140that it should use.
3141Then we classify traffic using a single rule:
3142.Pp
3143.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
3144.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
3145.Dl "..."
3146.Dl "ipfw table 1 add 192.168.2.0/24 1"
3147.Dl "ipfw table 1 add 192.168.0.0/27 4"
3148.Dl "ipfw table 1 add 192.168.0.2 1"
3149.Dl "..."
3150.Dl "ipfw add pipe tablearg ip from table(1) to any"
3151.Pp
3152Using the
3153.Cm fwd
3154action, the table entries may include hostnames and IP addresses.
3155.Pp
3156.Dl "ipfw table 1 add 192.168.2.0/24 10.23.2.1"
3157.Dl "ipfw table 1 add 192.168.0.0/27 router1.dmz"
3158.Dl "..."
3159.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
3160.Pp
3161In the following example per-interface firewall is created:
3162.Pp
3163.Dl "ipfw table 10 add vlan20 12000"
3164.Dl "ipfw table 10 add vlan30 13000"
3165.Dl "ipfw table 20 add vlan20 22000"
3166.Dl "ipfw table 20 add vlan30 23000"
3167.Dl ".."
3168.Dl "ipfw add 100 ipfw skipto tablearg ip from any to any recv 'table(10)' in"
3169.Dl "ipfw add 200 ipfw skipto tablearg ip from any to any xmit 'table(10)' out"
3170.Ss SETS OF RULES
3171To add a set of rules atomically, e.g.\& set 18:
3172.Pp
3173.Dl "ipfw set disable 18"
3174.Dl "ipfw add NN set 18 ...         # repeat as needed"
3175.Dl "ipfw set enable 18"
3176.Pp
3177To delete a set of rules atomically the command is simply:
3178.Pp
3179.Dl "ipfw delete set 18"
3180.Pp
3181To test a ruleset and disable it and regain control if something goes wrong:
3182.Pp
3183.Dl "ipfw set disable 18"
3184.Dl "ipfw add NN set 18 ...         # repeat as needed"
3185.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
3186.Pp
3187Here if everything goes well, you press control-C before the "sleep"
3188terminates, and your ruleset will be left active.
3189Otherwise, e.g.\& if
3190you cannot access your box, the ruleset will be disabled after
3191the sleep terminates thus restoring the previous situation.
3192.Pp
3193To show rules of the specific set:
3194.Pp
3195.Dl "ipfw set 18 show"
3196.Pp
3197To show rules of the disabled set:
3198.Pp
3199.Dl "ipfw -S set 18 show"
3200.Pp
3201To clear a specific rule counters of the specific set:
3202.Pp
3203.Dl "ipfw set 18 zero NN"
3204.Pp
3205To delete a specific rule of the specific set:
3206.Pp
3207.Dl "ipfw set 18 delete NN"
3208.Ss NAT, REDIRECT AND LSNAT
3209First redirect all the traffic to nat instance 123:
3210.Pp
3211.Dl "ipfw add nat 123 all from any to any"
3212.Pp
3213Then to configure nat instance 123 to alias all the outgoing traffic with ip
3214192.168.0.123, blocking all incoming connections, trying to keep
3215same ports on both sides, clearing aliasing table on address change
3216and keeping a log of traffic/link statistics:
3217.Pp
3218.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
3219.Pp
3220Or to change address of instance 123, aliasing table will be cleared (see
3221reset option):
3222.Pp
3223.Dl "ipfw nat 123 config ip 10.0.0.1"
3224.Pp
3225To see configuration of nat instance 123:
3226.Pp
3227.Dl "ipfw nat 123 show config"
3228.Pp
3229To show logs of all the instances in range 111-999:
3230.Pp
3231.Dl "ipfw nat 111-999 show"
3232.Pp
3233To see configurations of all instances:
3234.Pp
3235.Dl "ipfw nat show config"
3236.Pp
3237Or a redirect rule with mixed modes could looks like:
3238.Pp
3239.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
3240.Dl "			 redirect_port tcp 192.168.0.1:80 500"
3241.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
3242.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
3243.Dl "			 	    10.0.0.100	# LSNAT"
3244.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
3245.Dl "			 	    500		# LSNAT"
3246.Pp
3247or it could be split in:
3248.Pp
3249.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
3250.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
3251.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
3252.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
3253.Dl "				         10.0.0.100"
3254.Dl "ipfw nat 5 config redirect_port tcp"
3255.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
3256.Sh SEE ALSO
3257.Xr cpp 1 ,
3258.Xr m4 1 ,
3259.Xr altq 4 ,
3260.Xr divert 4 ,
3261.Xr dummynet 4 ,
3262.Xr if_bridge 4 ,
3263.Xr ip 4 ,
3264.Xr ipfirewall 4 ,
3265.Xr ng_ipfw 4 ,
3266.Xr protocols 5 ,
3267.Xr services 5 ,
3268.Xr init 8 ,
3269.Xr kldload 8 ,
3270.Xr reboot 8 ,
3271.Xr sysctl 8 ,
3272.Xr syslogd 8
3273.Sh HISTORY
3274The
3275.Nm
3276utility first appeared in
3277.Fx 2.0 .
3278.Nm dummynet
3279was introduced in
3280.Fx 2.2.8 .
3281Stateful extensions were introduced in
3282.Fx 4.0 .
3283.Nm ipfw2
3284was introduced in Summer 2002.
3285.Sh AUTHORS
3286.An Ugen J. S. Antsilevich ,
3287.An Poul-Henning Kamp ,
3288.An Alex Nash ,
3289.An Archie Cobbs ,
3290.An Luigi Rizzo .
3291.Pp
3292.An -nosplit
3293API based upon code written by
3294.An Daniel Boulet
3295for BSDI.
3296.Pp
3297Dummynet has been introduced by Luigi Rizzo in 1997-1998.
3298.Pp
3299Some early work (1999-2000) on the
3300.Nm dummynet
3301traffic shaper supported by Akamba Corp.
3302.Pp
3303The ipfw core (ipfw2) has been completely redesigned and
3304reimplemented by Luigi Rizzo in summer 2002.
3305Further
3306actions and
3307options have been added by various developer over the years.
3308.Pp
3309.An -nosplit
3310In-kernel NAT support written by
3311.An Paolo Pisati Aq piso@FreeBSD.org
3312as part of a Summer of Code 2005 project.
3313.Pp
3314SCTP
3315.Nm nat
3316support has been developed by
3317.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
3318The primary developers and maintainers are David Hayes and Jason But.
3319For further information visit:
3320.Aq http://www.caia.swin.edu.au/urp/SONATA
3321.Pp
3322Delay profiles have been developed by Alessandro Cerri and
3323Luigi Rizzo, supported by the
3324European Commission within Projects Onelab and Onelab2.
3325.Sh BUGS
3326The syntax has grown over the years and sometimes it might be confusing.
3327Unfortunately, backward compatibility prevents cleaning up mistakes
3328made in the definition of the syntax.
3329.Pp
3330.Em !!! WARNING !!!
3331.Pp
3332Misconfiguring the firewall can put your computer in an unusable state,
3333possibly shutting down network services and requiring console access to
3334regain control of it.
3335.Pp
3336Incoming packet fragments diverted by
3337.Cm divert
3338are reassembled before delivery to the socket.
3339The action used on those packet is the one from the
3340rule which matches the first fragment of the packet.
3341.Pp
3342Packets diverted to userland, and then reinserted by a userland process
3343may lose various packet attributes.
3344The packet source interface name
3345will be preserved if it is shorter than 8 bytes and the userland process
3346saves and reuses the sockaddr_in
3347(as does
3348.Xr natd 8 ) ;
3349otherwise, it may be lost.
3350If a packet is reinserted in this manner, later rules may be incorrectly
3351applied, making the order of
3352.Cm divert
3353rules in the rule sequence very important.
3354.Pp
3355Dummynet drops all packets with IPv6 link-local addresses.
3356.Pp
3357Rules using
3358.Cm uid
3359or
3360.Cm gid
3361may not behave as expected.
3362In particular, incoming SYN packets may
3363have no uid or gid associated with them since they do not yet belong
3364to a TCP connection, and the uid/gid associated with a packet may not
3365be as expected if the associated process calls
3366.Xr setuid 2
3367or similar system calls.
3368.Pp
3369Rule syntax is subject to the command line environment and some patterns
3370may need to be escaped with the backslash character
3371or quoted appropriately.
3372.Pp
3373Due to the architecture of
3374.Xr libalias 3 ,
3375ipfw nat is not compatible with the TCP segmentation offloading (TSO).
3376Thus, to reliably nat your network traffic, please disable TSO
3377on your NICs using
3378.Xr ifconfig 8 .
3379.Pp
3380ICMP error messages are not implicitly matched by dynamic rules
3381for the respective conversations.
3382To avoid failures of network error detection and path MTU discovery,
3383ICMP error messages may need to be allowed explicitly through static
3384rules.
3385.Pp
3386Rules using
3387.Cm call
3388and
3389.Cm return
3390actions may lead to confusing behaviour if ruleset has mistakes,
3391and/or interaction with other subsystems (netgraph, dummynet, etc.) is used.
3392One possible case for this is packet leaving
3393.Nm
3394in subroutine on the input pass, while later on output encountering unpaired
3395.Cm return
3396first.
3397As the call stack is kept intact after input pass, packet will suddenly
3398return to the rule number used on input pass, not on output one.
3399Order of processing should be checked carefully to avoid such mistakes.
3400