xref: /freebsd/sbin/ipfw/ipfw.8 (revision aa0a1e58f0189b0fde359a8bda032887e72057fa)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd July 27, 2010
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.Pp
44.Nm
45.Cm enable
46.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
47.Nm
48.Cm disable
49.Brq Cm firewall | altq | one_pass | debug | verbose | dyn_keepalive
50.Pp
51.Ss LOOKUP TABLES
52.Nm
53.Cm table Ar number Cm add Ar addr Ns Oo / Ns Ar masklen Oc Op Ar value
54.Nm
55.Cm table Ar number Cm delete Ar addr Ns Op / Ns Ar masklen
56.Nm
57.Cm table
58.Brq Ar number | all
59.Cm flush
60.Nm
61.Cm table
62.Brq Ar number | all
63.Cm list
64.Pp
65.Ss DUMMYNET CONFIGURATION (TRAFFIC SHAPER AND PACKET SCHEDULER)
66.Nm
67.Brq Cm pipe | queue | sched
68.Ar number
69.Cm config
70.Ar config-options
71.Nm
72.Op Fl s Op Ar field
73.Brq Cm pipe | queue | sched
74.Brq Cm delete | list | show
75.Op Ar number ...
76.Pp
77.Ss IN-KERNEL NAT
78.Nm
79.Op Fl q
80.Cm nat
81.Ar number
82.Cm config
83.Ar config-options
84.Pp
85.Nm
86.Op Fl cfnNqS
87.Oo
88.Fl p Ar preproc
89.Oo
90.Ar preproc-flags
91.Oc
92.Oc
93.Ar pathname
94.Sh DESCRIPTION
95The
96.Nm
97utility is the user interface for controlling the
98.Xr ipfw 4
99firewall, the
100.Xr dummynet 4
101traffic shaper/packet scheduler, and the
102in-kernel NAT services.
103.Pp
104A firewall configuration, or
105.Em ruleset ,
106is made of a list of
107.Em rules
108numbered from 1 to 65535.
109Packets are passed to the firewall
110from a number of different places in the protocol stack
111(depending on the source and destination of the packet,
112it is possible for the firewall to be
113invoked multiple times on the same packet).
114The packet passed to the firewall is compared
115against each of the rules in the
116.Em ruleset ,
117in rule-number order
118(multiple rules with the same number are permitted, in which case
119they are processed in order of insertion).
120When a match is found, the action corresponding to the
121matching rule is performed.
122.Pp
123Depending on the action and certain system settings, packets
124can be reinjected into the firewall at some rule after the
125matching one for further processing.
126.Pp
127A ruleset always includes a
128.Em default
129rule (numbered 65535) which cannot be modified or deleted,
130and matches all packets.
131The action associated with the
132.Em default
133rule can be either
134.Cm deny
135or
136.Cm allow
137depending on how the kernel is configured.
138.Pp
139If the ruleset includes one or more rules with the
140.Cm keep-state
141or
142.Cm limit
143option,
144the firewall will have a
145.Em stateful
146behaviour, i.e., upon a match it will create
147.Em dynamic rules ,
148i.e. rules that match packets with the same 5-tuple
149(protocol, source and destination addresses and ports)
150as the packet which caused their creation.
151Dynamic rules, which have a limited lifetime, are checked
152at the first occurrence of a
153.Cm check-state ,
154.Cm keep-state
155or
156.Cm limit
157rule, and are typically used to open the firewall on-demand to
158legitimate traffic only.
159See the
160.Sx STATEFUL FIREWALL
161and
162.Sx EXAMPLES
163Sections below for more information on the stateful behaviour of
164.Nm .
165.Pp
166All rules (including dynamic ones) have a few associated counters:
167a packet count, a byte count, a log count and a timestamp
168indicating the time of the last match.
169Counters can be displayed or reset with
170.Nm
171commands.
172.Pp
173Each rule belongs to one of 32 different
174.Em sets
175, and there are
176.Nm
177commands to atomically manipulate sets, such as enable,
178disable, swap sets, move all rules in a set to another
179one, delete all rules in a set.
180These can be useful to
181install temporary configurations, or to test them.
182See Section
183.Sx SETS OF RULES
184for more information on
185.Em sets .
186.Pp
187Rules can be added with the
188.Cm add
189command; deleted individually or in groups with the
190.Cm delete
191command, and globally (except those in set 31) with the
192.Cm flush
193command; displayed, optionally with the content of the
194counters, using the
195.Cm show
196and
197.Cm list
198commands.
199Finally, counters can be reset with the
200.Cm zero
201and
202.Cm resetlog
203commands.
204.Pp
205.Ss COMMAND OPTIONS
206The following general options are available when invoking
207.Nm :
208.Bl -tag -width indent
209.It Fl a
210Show counter values when listing rules.
211The
212.Cm show
213command implies this option.
214.It Fl b
215Only show the action and the comment, not the body of a rule.
216Implies
217.Fl c .
218.It Fl c
219When entering or showing rules, print them in compact form,
220i.e., omitting the "ip from any to any" string
221when this does not carry any additional information.
222.It Fl d
223When listing, show dynamic rules in addition to static ones.
224.It Fl e
225When listing and
226.Fl d
227is specified, also show expired dynamic rules.
228.It Fl f
229Do not ask for confirmation for commands that can cause problems
230if misused,
231.No i.e. Cm flush .
232If there is no tty associated with the process, this is implied.
233.It Fl i
234When listing a table (see the
235.Sx LOOKUP TABLES
236section below for more information on lookup tables), format values
237as IP addresses. By default, values are shown as integers.
238.It Fl n
239Only check syntax of the command strings, without actually passing
240them to the kernel.
241.It Fl N
242Try to resolve addresses and service names in output.
243.It Fl q
244Be quiet when executing the
245.Cm add ,
246.Cm nat ,
247.Cm zero ,
248.Cm resetlog
249or
250.Cm flush
251commands;
252(implies
253.Fl f ) .
254This is useful when updating rulesets by executing multiple
255.Nm
256commands in a script
257(e.g.,
258.Ql sh\ /etc/rc.firewall ) ,
259or by processing a file with many
260.Nm
261rules across a remote login session.
262It also stops a table add or delete
263from failing if the entry already exists or is not present.
264.Pp
265The reason why this option may be important is that
266for some of these actions,
267.Nm
268may print a message; if the action results in blocking the
269traffic to the remote client,
270the remote login session will be closed
271and the rest of the ruleset will not be processed.
272Access to the console would then be required to recover.
273.It Fl S
274When listing rules, show the
275.Em set
276each rule belongs to.
277If this flag is not specified, disabled rules will not be
278listed.
279.It Fl s Op Ar field
280When listing pipes, sort according to one of the four
281counters (total or current packets or bytes).
282.It Fl t
283When listing, show last match timestamp converted with ctime().
284.It Fl T
285When listing, show last match timestamp as seconds from the epoch.
286This form can be more convenient for postprocessing by scripts.
287.El
288.Pp
289.Ss LIST OF RULES AND PREPROCESSING
290To ease configuration, rules can be put into a file which is
291processed using
292.Nm
293as shown in the last synopsis line.
294An absolute
295.Ar pathname
296must be used.
297The file will be read line by line and applied as arguments to the
298.Nm
299utility.
300.Pp
301Optionally, a preprocessor can be specified using
302.Fl p Ar preproc
303where
304.Ar pathname
305is to be piped through.
306Useful preprocessors include
307.Xr cpp 1
308and
309.Xr m4 1 .
310If
311.Ar preproc
312does not start with a slash
313.Pq Ql /
314as its first character, the usual
315.Ev PATH
316name search is performed.
317Care should be taken with this in environments where not all
318file systems are mounted (yet) by the time
319.Nm
320is being run (e.g.\& when they are mounted over NFS).
321Once
322.Fl p
323has been specified, any additional arguments are passed on to the preprocessor
324for interpretation.
325This allows for flexible configuration files (like conditionalizing
326them on the local hostname) and the use of macros to centralize
327frequently required arguments like IP addresses.
328.Pp
329.Ss TRAFFIC SHAPER CONFIGURATION
330The
331.Nm
332.Cm pipe , queue
333and
334.Cm sched
335commands are used to configure the traffic shaper and packet scheduler.
336See the
337.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
338Section below for details.
339.Pp
340If the world and the kernel get out of sync the
341.Nm
342ABI may break, preventing you from being able to add any rules.
343This can
344adversely effect the booting process.
345You can use
346.Nm
347.Cm disable
348.Cm firewall
349to temporarily disable the firewall to regain access to the network,
350allowing you to fix the problem.
351.Sh PACKET FLOW
352A packet is checked against the active ruleset in multiple places
353in the protocol stack, under control of several sysctl variables.
354These places and variables are shown below, and it is important to
355have this picture in mind in order to design a correct ruleset.
356.Bd -literal -offset indent
357       ^    to upper layers    V
358       |                       |
359       +----------->-----------+
360       ^                       V
361 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
362       |                       |
363       ^                       V
364 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
365       |                       |
366       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
367       ^                       V
368       |      to devices       |
369.Ed
370.Pp
371The number of
372times the same packet goes through the firewall can
373vary between 0 and 4 depending on packet source and
374destination, and system configuration.
375.Pp
376Note that as packets flow through the stack, headers can be
377stripped or added to it, and so they may or may not be available
378for inspection.
379E.g., incoming packets will include the MAC header when
380.Nm
381is invoked from
382.Cm ether_demux() ,
383but the same packets will have the MAC header stripped off when
384.Nm
385is invoked from
386.Cm ip_input()
387or
388.Cm ip6_input() .
389.Pp
390Also note that each packet is always checked against the complete ruleset,
391irrespective of the place where the check occurs, or the source of the packet.
392If a rule contains some match patterns or actions which are not valid
393for the place of invocation (e.g.\& trying to match a MAC header within
394.Cm ip_input
395or
396.Cm ip6_input ),
397the match pattern will not match, but a
398.Cm not
399operator in front of such patterns
400.Em will
401cause the pattern to
402.Em always
403match on those packets.
404It is thus the responsibility of
405the programmer, if necessary, to write a suitable ruleset to
406differentiate among the possible places.
407.Cm skipto
408rules can be useful here, as an example:
409.Bd -literal -offset indent
410# packets from ether_demux or bdg_forward
411ipfw add 10 skipto 1000 all from any to any layer2 in
412# packets from ip_input
413ipfw add 10 skipto 2000 all from any to any not layer2 in
414# packets from ip_output
415ipfw add 10 skipto 3000 all from any to any not layer2 out
416# packets from ether_output_frame
417ipfw add 10 skipto 4000 all from any to any layer2 out
418.Ed
419.Pp
420(yes, at the moment there is no way to differentiate between
421ether_demux and bdg_forward).
422.Sh SYNTAX
423In general, each keyword or argument must be provided as
424a separate command line argument, with no leading or trailing
425spaces.
426Keywords are case-sensitive, whereas arguments may
427or may not be case-sensitive depending on their nature
428(e.g.\& uid's are, hostnames are not).
429.Pp
430Some arguments (e.g. port or address lists) are comma-separated
431lists of values.
432In this case, spaces after commas ',' are allowed to make
433the line more readable.
434You can also put the entire
435command (including flags) into a single argument.
436E.g., the following forms are equivalent:
437.Bd -literal -offset indent
438ipfw -q add deny src-ip 10.0.0.0/24,127.0.0.1/8
439ipfw -q add deny src-ip 10.0.0.0/24, 127.0.0.1/8
440ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
441.Ed
442.Sh RULE FORMAT
443The format of firewall rules is the following:
444.Bd -ragged -offset indent
445.Bk -words
446.Op Ar rule_number
447.Op Cm set Ar set_number
448.Op Cm prob Ar match_probability
449.Ar action
450.Op Cm log Op Cm logamount Ar number
451.Op Cm altq Ar queue
452.Oo
453.Bro Cm tag | untag
454.Brc Ar number
455.Oc
456.Ar body
457.Ek
458.Ed
459.Pp
460where the body of the rule specifies which information is used
461for filtering packets, among the following:
462.Pp
463.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
464.It Layer-2 header fields
465When available
466.It IPv4 and IPv6 Protocol
467TCP, UDP, ICMP, etc.
468.It Source and dest. addresses and ports
469.It Direction
470See Section
471.Sx PACKET FLOW
472.It Transmit and receive interface
473By name or address
474.It Misc. IP header fields
475Version, type of service, datagram length, identification,
476fragment flag (non-zero IP offset),
477Time To Live
478.It IP options
479.It IPv6 Extension headers
480Fragmentation, Hop-by-Hop options,
481Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
482.It IPv6 Flow-ID
483.It Misc. TCP header fields
484TCP flags (SYN, FIN, ACK, RST, etc.),
485sequence number, acknowledgment number,
486window
487.It TCP options
488.It ICMP types
489for ICMP packets
490.It ICMP6 types
491for ICMP6 packets
492.It User/group ID
493When the packet can be associated with a local socket.
494.It Divert status
495Whether a packet came from a divert socket (e.g.,
496.Xr natd 8 ) .
497.It Fib annotation state
498Whether a packet has been tagged for using a specific FIB (routing table)
499in future forwarding decisions.
500.El
501.Pp
502Note that some of the above information, e.g.\& source MAC or IP addresses and
503TCP/UDP ports, can be easily spoofed, so filtering on those fields
504alone might not guarantee the desired results.
505.Bl -tag -width indent
506.It Ar rule_number
507Each rule is associated with a
508.Ar rule_number
509in the range 1..65535, with the latter reserved for the
510.Em default
511rule.
512Rules are checked sequentially by rule number.
513Multiple rules can have the same number, in which case they are
514checked (and listed) according to the order in which they have
515been added.
516If a rule is entered without specifying a number, the kernel will
517assign one in such a way that the rule becomes the last one
518before the
519.Em default
520rule.
521Automatic rule numbers are assigned by incrementing the last
522non-default rule number by the value of the sysctl variable
523.Ar net.inet.ip.fw.autoinc_step
524which defaults to 100.
525If this is not possible (e.g.\& because we would go beyond the
526maximum allowed rule number), the number of the last
527non-default value is used instead.
528.It Cm set Ar set_number
529Each rule is associated with a
530.Ar set_number
531in the range 0..31.
532Sets can be individually disabled and enabled, so this parameter
533is of fundamental importance for atomic ruleset manipulation.
534It can be also used to simplify deletion of groups of rules.
535If a rule is entered without specifying a set number,
536set 0 will be used.
537.br
538Set 31 is special in that it cannot be disabled,
539and rules in set 31 are not deleted by the
540.Nm ipfw flush
541command (but you can delete them with the
542.Nm ipfw delete set 31
543command).
544Set 31 is also used for the
545.Em default
546rule.
547.It Cm prob Ar match_probability
548A match is only declared with the specified probability
549(floating point number between 0 and 1).
550This can be useful for a number of applications such as
551random packet drop or
552(in conjunction with
553.Nm dummynet )
554to simulate the effect of multiple paths leading to out-of-order
555packet delivery.
556.Pp
557Note: this condition is checked before any other condition, including
558ones such as keep-state or check-state which might have side effects.
559.It Cm log Op Cm logamount Ar number
560Packets matching a rule with the
561.Cm log
562keyword will be made available for logging in two ways:
563if the sysctl variable
564.Va net.inet.ip.fw.verbose
565is set to 0 (default), one can use
566.Xr bpf 4
567attached to the
568.Xr ipfw0
569pseudo interface. There is no overhead if no
570.Xr bpf
571is attached to the pseudo interface.
572.Pp
573If
574.Va net.inet.ip.fw.verbose
575is set to 1, packets will be logged to
576.Xr syslogd 8
577with a
578.Dv LOG_SECURITY
579facility up to a maximum of
580.Cm logamount
581packets.
582If no
583.Cm logamount
584is specified, the limit is taken from the sysctl variable
585.Va net.inet.ip.fw.verbose_limit .
586In both cases, a value of 0 means unlimited logging.
587.Pp
588Once the limit is reached, logging can be re-enabled by
589clearing the logging counter or the packet counter for that entry, see the
590.Cm resetlog
591command.
592.Pp
593Note: logging is done after all other packet matching conditions
594have been successfully verified, and before performing the final
595action (accept, deny, etc.) on the packet.
596.It Cm tag Ar number
597When a packet matches a rule with the
598.Cm tag
599keyword, the numeric tag for the given
600.Ar number
601in the range 1..65534 will be attached to the packet.
602The tag acts as an internal marker (it is not sent out over
603the wire) that can be used to identify these packets later on.
604This can be used, for example, to provide trust between interfaces
605and to start doing policy-based filtering.
606A packet can have multiple tags at the same time.
607Tags are "sticky", meaning once a tag is applied to a packet by a
608matching rule it exists until explicit removal.
609Tags are kept with the packet everywhere within the kernel, but are
610lost when packet leaves the kernel, for example, on transmitting
611packet out to the network or sending packet to a
612.Xr divert 4
613socket.
614.Pp
615To check for previously applied tags, use the
616.Cm tagged
617rule option.
618To delete previously applied tag, use the
619.Cm untag
620keyword.
621.Pp
622Note: since tags are kept with the packet everywhere in kernelspace,
623they can be set and unset anywhere in the kernel network subsystem
624(using the
625.Xr mbuf_tags 9
626facility), not only by means of the
627.Xr ipfw 4
628.Cm tag
629and
630.Cm untag
631keywords.
632For example, there can be a specialized
633.Xr netgraph 4
634node doing traffic analyzing and tagging for later inspecting
635in firewall.
636.It Cm untag Ar number
637When a packet matches a rule with the
638.Cm untag
639keyword, the tag with the number
640.Ar number
641is searched among the tags attached to this packet and,
642if found, removed from it.
643Other tags bound to packet, if present, are left untouched.
644.It Cm altq Ar queue
645When a packet matches a rule with the
646.Cm altq
647keyword, the ALTQ identifier for the given
648.Ar queue
649(see
650.Xr altq 4 )
651will be attached.
652Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
653and not being rejected or going to divert sockets.
654Note that if there is insufficient memory at the time the packet is
655processed, it will not be tagged, so it is wise to make your ALTQ
656"default" queue policy account for this.
657If multiple
658.Cm altq
659rules match a single packet, only the first one adds the ALTQ classification
660tag.
661In doing so, traffic may be shaped by using
662.Cm count Cm altq Ar queue
663rules for classification early in the ruleset, then later applying
664the filtering decision.
665For example,
666.Cm check-state
667and
668.Cm keep-state
669rules may come later and provide the actual filtering decisions in
670addition to the fallback ALTQ tag.
671.Pp
672You must run
673.Xr pfctl 8
674to set up the queues before IPFW will be able to look them up by name,
675and if the ALTQ disciplines are rearranged, the rules in containing the
676queue identifiers in the kernel will likely have gone stale and need
677to be reloaded.
678Stale queue identifiers will probably result in misclassification.
679.Pp
680All system ALTQ processing can be turned on or off via
681.Nm
682.Cm enable Ar altq
683and
684.Nm
685.Cm disable Ar altq .
686The usage of
687.Va net.inet.ip.fw.one_pass
688is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
689always after adding an ALTQ tag.
690.El
691.Ss RULE ACTIONS
692A rule can be associated with one of the following actions, which
693will be executed when the packet matches the body of the rule.
694.Bl -tag -width indent
695.It Cm allow | accept | pass | permit
696Allow packets that match rule.
697The search terminates.
698.It Cm check-state
699Checks the packet against the dynamic ruleset.
700If a match is found, execute the action associated with
701the rule which generated this dynamic rule, otherwise
702move to the next rule.
703.br
704.Cm Check-state
705rules do not have a body.
706If no
707.Cm check-state
708rule is found, the dynamic ruleset is checked at the first
709.Cm keep-state
710or
711.Cm limit
712rule.
713.It Cm count
714Update counters for all packets that match rule.
715The search continues with the next rule.
716.It Cm deny | drop
717Discard packets that match this rule.
718The search terminates.
719.It Cm divert Ar port
720Divert packets that match this rule to the
721.Xr divert 4
722socket bound to port
723.Ar port .
724The search terminates.
725.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
726Change the next-hop on matching packets to
727.Ar ipaddr ,
728which can be an IP address or a host name.
729The next hop can also be supplied by the last table
730looked up for the packet by using the
731.Cm tablearg
732keyword instead of an explicit address.
733The search terminates if this rule matches.
734.Pp
735If
736.Ar ipaddr
737is a local address, then matching packets will be forwarded to
738.Ar port
739(or the port number in the packet if one is not specified in the rule)
740on the local machine.
741.br
742If
743.Ar ipaddr
744is not a local address, then the port number
745(if specified) is ignored, and the packet will be
746forwarded to the remote address, using the route as found in
747the local routing table for that IP.
748.br
749A
750.Ar fwd
751rule will not match layer-2 packets (those received
752on ether_input, ether_output, or bridged).
753.br
754The
755.Cm fwd
756action does not change the contents of the packet at all.
757In particular, the destination address remains unmodified, so
758packets forwarded to another system will usually be rejected by that system
759unless there is a matching rule on that system to capture them.
760For packets forwarded locally,
761the local address of the socket will be
762set to the original destination address of the packet.
763This makes the
764.Xr netstat 1
765entry look rather weird but is intended for
766use with transparent proxy servers.
767.Pp
768To enable
769.Cm fwd
770a custom kernel needs to be compiled with the option
771.Cd "options IPFIREWALL_FORWARD" .
772.It Cm nat Ar nat_nr
773Pass packet to a
774nat instance
775(for network address translation, address redirect, etc.):
776see the
777.Sx NETWORK ADDRESS TRANSLATION (NAT)
778Section for further information.
779.It Cm pipe Ar pipe_nr
780Pass packet to a
781.Nm dummynet
782.Dq pipe
783(for bandwidth limitation, delay, etc.).
784See the
785.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
786Section for further information.
787The search terminates; however, on exit from the pipe and if
788the
789.Xr sysctl 8
790variable
791.Va net.inet.ip.fw.one_pass
792is not set, the packet is passed again to the firewall code
793starting from the next rule.
794.It Cm queue Ar queue_nr
795Pass packet to a
796.Nm dummynet
797.Dq queue
798(for bandwidth limitation using WF2Q+).
799.It Cm reject
800(Deprecated).
801Synonym for
802.Cm unreach host .
803.It Cm reset
804Discard packets that match this rule, and if the
805packet is a TCP packet, try to send a TCP reset (RST) notice.
806The search terminates.
807.It Cm reset6
808Discard packets that match this rule, and if the
809packet is a TCP packet, try to send a TCP reset (RST) notice.
810The search terminates.
811.It Cm skipto Ar number | tablearg
812Skip all subsequent rules numbered less than
813.Ar number .
814The search continues with the first rule numbered
815.Ar number
816or higher.
817It is possible to use the
818.Cm tablearg
819keyword with a skipto for a
820.Em computed
821skipto, but care should be used, as no destination caching
822is possible in this case so the rules are always walked to find it,
823starting from the
824.Cm skipto .
825.It Cm tee Ar port
826Send a copy of packets matching this rule to the
827.Xr divert 4
828socket bound to port
829.Ar port .
830The search continues with the next rule.
831.It Cm unreach Ar code
832Discard packets that match this rule, and try to send an ICMP
833unreachable notice with code
834.Ar code ,
835where
836.Ar code
837is a number from 0 to 255, or one of these aliases:
838.Cm net , host , protocol , port ,
839.Cm needfrag , srcfail , net-unknown , host-unknown ,
840.Cm isolated , net-prohib , host-prohib , tosnet ,
841.Cm toshost , filter-prohib , host-precedence
842or
843.Cm precedence-cutoff .
844The search terminates.
845.It Cm unreach6 Ar code
846Discard packets that match this rule, and try to send an ICMPv6
847unreachable notice with code
848.Ar code ,
849where
850.Ar code
851is a number from 0, 1, 3 or 4, or one of these aliases:
852.Cm no-route, admin-prohib, address
853or
854.Cm port .
855The search terminates.
856.It Cm netgraph Ar cookie
857Divert packet into netgraph with given
858.Ar cookie .
859The search terminates.
860If packet is later returned from netgraph it is either
861accepted or continues with the next rule, depending on
862.Va net.inet.ip.fw.one_pass
863sysctl variable.
864.It Cm ngtee Ar cookie
865A copy of packet is diverted into netgraph, original
866packet continues with the next rule.
867See
868.Xr ng_ipfw 4
869for more information on
870.Cm netgraph
871and
872.Cm ngtee
873actions.
874.It Cm setfib Ar fibnum
875The packet is tagged so as to use the FIB (routing table)
876.Ar fibnum
877in any subsequent forwarding decisions.
878Initially this is limited to the values 0 through 15, see
879.Xr setfib 1 .
880Processing continues at the next rule.
881.It Cm reass
882Queue and reassemble ip fragments.
883If the packet is not fragmented, counters are updated and processing continues with the next rule.
884If the packet is the last logical fragment, the packet is reassembled and, if
885.Va net.inet.ip.fw.one_pass
886is set to 0, processing continues with the next rule, else packet is allowed to pass and search terminates.
887If the packet is a fragment in the middle, it is consumed and processing stops immediately.
888.Pp
889Fragments handling can be tuned via
890.Va net.inet.ip.maxfragpackets
891and
892.Va net.inet.ip.maxfragsperpacket
893which limit, respectively, the maximum number of processable fragments (default: 800) and
894the maximum number of fragments per packet (default: 16).
895.Pp
896NOTA BENE: since fragments do not contain port numbers, they should be avoided with the
897.Nm reass
898rule.
899Alternatively, direction-based (like
900.Nm in
901/
902.Nm out
903) and source-based (like
904.Nm via
905) match patterns can be used to select fragments.
906.Pp
907Usually a simple rule like:
908.Bd -literal -offset indent
909# reassemble incoming fragments
910ipfw add reass all from any to any in
911.Ed
912.Pp
913is all you need at the beginning of your ruleset.
914.El
915.Ss RULE BODY
916The body of a rule contains zero or more patterns (such as
917specific source and destination addresses or ports,
918protocol options, incoming or outgoing interfaces, etc.)
919that the packet must match in order to be recognised.
920In general, the patterns are connected by (implicit)
921.Cm and
922operators -- i.e., all must match in order for the
923rule to match.
924Individual patterns can be prefixed by the
925.Cm not
926operator to reverse the result of the match, as in
927.Pp
928.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
929.Pp
930Additionally, sets of alternative match patterns
931.Pq Em or-blocks
932can be constructed by putting the patterns in
933lists enclosed between parentheses ( ) or braces { }, and
934using the
935.Cm or
936operator as follows:
937.Pp
938.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
939.Pp
940Only one level of parentheses is allowed.
941Beware that most shells have special meanings for parentheses
942or braces, so it is advisable to put a backslash \\ in front of them
943to prevent such interpretations.
944.Pp
945The body of a rule must in general include a source and destination
946address specifier.
947The keyword
948.Ar any
949can be used in various places to specify that the content of
950a required field is irrelevant.
951.Pp
952The rule body has the following format:
953.Bd -ragged -offset indent
954.Op Ar proto Cm from Ar src Cm to Ar dst
955.Op Ar options
956.Ed
957.Pp
958The first part (proto from src to dst) is for backward
959compatibility with earlier versions of
960.Fx .
961In modern
962.Fx
963any match pattern (including MAC headers, IP protocols,
964addresses and ports) can be specified in the
965.Ar options
966section.
967.Pp
968Rule fields have the following meaning:
969.Bl -tag -width indent
970.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
971.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
972An IP protocol specified by number or name
973(for a complete list see
974.Pa /etc/protocols ) ,
975or one of the following keywords:
976.Bl -tag -width indent
977.It Cm ip4 | ipv4
978Matches IPv4 packets.
979.It Cm ip6 | ipv6
980Matches IPv6 packets.
981.It Cm ip | all
982Matches any packet.
983.El
984.Pp
985The
986.Cm ipv6
987in
988.Cm proto
989option will be treated as inner protocol.
990And, the
991.Cm ipv4
992is not available in
993.Cm proto
994option.
995.Pp
996The
997.Cm { Ar protocol Cm or ... }
998format (an
999.Em or-block )
1000is provided for convenience only but its use is deprecated.
1001.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1002An address (or a list, see below)
1003optionally followed by
1004.Ar ports
1005specifiers.
1006.Pp
1007The second format
1008.Em ( or-block
1009with multiple addresses) is provided for convenience only and
1010its use is discouraged.
1011.It Ar addr : Oo Cm not Oc Bro
1012.Cm any | me | me6 |
1013.Cm table Ns Pq Ar number Ns Op , Ns Ar value
1014.Ar | addr-list | addr-set
1015.Brc
1016.Bl -tag -width indent
1017.It Cm any
1018matches any IP address.
1019.It Cm me
1020matches any IP address configured on an interface in the system.
1021.It Cm me6
1022matches any IPv6 address configured on an interface in the system.
1023The address list is evaluated at the time the packet is
1024analysed.
1025.It Cm table Ns Pq Ar number Ns Op , Ns Ar value
1026Matches any IPv4 address for which an entry exists in the lookup table
1027.Ar number .
1028If an optional 32-bit unsigned
1029.Ar value
1030is also specified, an entry will match only if it has this value.
1031See the
1032.Sx LOOKUP TABLES
1033section below for more information on lookup tables.
1034.El
1035.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1036.It Ar ip-addr :
1037A host or subnet address specified in one of the following ways:
1038.Bl -tag -width indent
1039.It Ar numeric-ip | hostname
1040Matches a single IPv4 address, specified as dotted-quad or a hostname.
1041Hostnames are resolved at the time the rule is added to the firewall list.
1042.It Ar addr Ns / Ns Ar masklen
1043Matches all addresses with base
1044.Ar addr
1045(specified as an IP address, a network number, or a hostname)
1046and mask width of
1047.Cm masklen
1048bits.
1049As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1050all IP numbers from 1.2.3.0 to 1.2.3.127 .
1051.It Ar addr Ns : Ns Ar mask
1052Matches all addresses with base
1053.Ar addr
1054(specified as an IP address, a network number, or a hostname)
1055and the mask of
1056.Ar mask ,
1057specified as a dotted quad.
1058As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
10591.*.3.*.
1060This form is advised only for non-contiguous
1061masks.
1062It is better to resort to the
1063.Ar addr Ns / Ns Ar masklen
1064format for contiguous masks, which is more compact and less
1065error-prone.
1066.El
1067.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1068.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1069Matches all addresses with base address
1070.Ar addr
1071(specified as an IP address, a network number, or a hostname)
1072and whose last byte is in the list between braces { } .
1073Note that there must be no spaces between braces and
1074numbers (spaces after commas are allowed).
1075Elements of the list can be specified as single entries
1076or ranges.
1077The
1078.Ar masklen
1079field is used to limit the size of the set of addresses,
1080and can have any value between 24 and 32.
1081If not specified,
1082it will be assumed as 24.
1083.br
1084This format is particularly useful to handle sparse address sets
1085within a single rule.
1086Because the matching occurs using a
1087bitmask, it takes constant time and dramatically reduces
1088the complexity of rulesets.
1089.br
1090As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1091or 1.2.3.0/24{128,35-55,89}
1092will match the following IP addresses:
1093.br
10941.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1095.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1096.It Ar ip6-addr :
1097A host or subnet specified one of the following ways:
1098.Bl -tag -width indent
1099.It Ar numeric-ip | hostname
1100Matches a single IPv6 address as allowed by
1101.Xr inet_pton 3
1102or a hostname.
1103Hostnames are resolved at the time the rule is added to the firewall
1104list.
1105.It Ar addr Ns / Ns Ar masklen
1106Matches all IPv6 addresses with base
1107.Ar addr
1108(specified as allowed by
1109.Xr inet_pton
1110or a hostname)
1111and mask width of
1112.Cm masklen
1113bits.
1114.El
1115.Pp
1116No support for sets of IPv6 addresses is provided because IPv6 addresses
1117are typically random past the initial prefix.
1118.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1119For protocols which support port numbers (such as TCP and UDP), optional
1120.Cm ports
1121may be specified as one or more ports or port ranges, separated
1122by commas but no spaces, and an optional
1123.Cm not
1124operator.
1125The
1126.Ql \&-
1127notation specifies a range of ports (including boundaries).
1128.Pp
1129Service names (from
1130.Pa /etc/services )
1131may be used instead of numeric port values.
1132The length of the port list is limited to 30 ports or ranges,
1133though one can specify larger ranges by using an
1134.Em or-block
1135in the
1136.Cm options
1137section of the rule.
1138.Pp
1139A backslash
1140.Pq Ql \e
1141can be used to escape the dash
1142.Pq Ql -
1143character in a service name (from a shell, the backslash must be
1144typed twice to avoid the shell itself interpreting it as an escape
1145character).
1146.Pp
1147.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1148.Pp
1149Fragmented packets which have a non-zero offset (i.e., not the first
1150fragment) will never match a rule which has one or more port
1151specifications.
1152See the
1153.Cm frag
1154option for details on matching fragmented packets.
1155.El
1156.Ss RULE OPTIONS (MATCH PATTERNS)
1157Additional match patterns can be used within
1158rules.
1159Zero or more of these so-called
1160.Em options
1161can be present in a rule, optionally prefixed by the
1162.Cm not
1163operand, and possibly grouped into
1164.Em or-blocks .
1165.Pp
1166The following match patterns can be used (listed in alphabetical order):
1167.Bl -tag -width indent
1168.It Cm // this is a comment.
1169Inserts the specified text as a comment in the rule.
1170Everything following // is considered as a comment and stored in the rule.
1171You can have comment-only rules, which are listed as having a
1172.Cm count
1173action followed by the comment.
1174.It Cm bridged
1175Alias for
1176.Cm layer2 .
1177.It Cm diverted
1178Matches only packets generated by a divert socket.
1179.It Cm diverted-loopback
1180Matches only packets coming from a divert socket back into the IP stack
1181input for delivery.
1182.It Cm diverted-output
1183Matches only packets going from a divert socket back outward to the IP
1184stack output for delivery.
1185.It Cm dst-ip Ar ip-address
1186Matches IPv4 packets whose destination IP is one of the address(es)
1187specified as argument.
1188.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1189Matches IPv6 packets whose destination IP is one of the address(es)
1190specified as argument.
1191.It Cm dst-port Ar ports
1192Matches IP packets whose destination port is one of the port(s)
1193specified as argument.
1194.It Cm established
1195Matches TCP packets that have the RST or ACK bits set.
1196.It Cm ext6hdr Ar header
1197Matches IPv6 packets containing the extended header given by
1198.Ar header .
1199Supported headers are:
1200.Pp
1201Fragment,
1202.Pq Cm frag ,
1203Hop-to-hop options
1204.Pq Cm hopopt ,
1205any type of Routing Header
1206.Pq Cm route ,
1207Source routing Routing Header Type 0
1208.Pq Cm rthdr0 ,
1209Mobile IPv6 Routing Header Type 2
1210.Pq Cm rthdr2 ,
1211Destination options
1212.Pq Cm dstopt ,
1213IPSec authentication headers
1214.Pq Cm ah ,
1215and IPsec encapsulated security payload headers
1216.Pq Cm esp .
1217.It Cm fib Ar fibnum
1218Matches a packet that has been tagged to use
1219the given FIB (routing table) number.
1220.It Cm flow-id Ar labels
1221Matches IPv6 packets containing any of the flow labels given in
1222.Ar labels .
1223.Ar labels
1224is a comma separated list of numeric flow labels.
1225.It Cm frag
1226Matches packets that are fragments and not the first
1227fragment of an IP datagram.
1228Note that these packets will not have
1229the next protocol header (e.g.\& TCP, UDP) so options that look into
1230these headers cannot match.
1231.It Cm gid Ar group
1232Matches all TCP or UDP packets sent by or received for a
1233.Ar group .
1234A
1235.Ar group
1236may be specified by name or number.
1237.It Cm jail Ar prisonID
1238Matches all TCP or UDP packets sent by or received for the
1239jail whos prison ID is
1240.Ar prisonID .
1241.It Cm icmptypes Ar types
1242Matches ICMP packets whose ICMP type is in the list
1243.Ar types .
1244The list may be specified as any combination of
1245individual types (numeric) separated by commas.
1246.Em Ranges are not allowed .
1247The supported ICMP types are:
1248.Pp
1249echo reply
1250.Pq Cm 0 ,
1251destination unreachable
1252.Pq Cm 3 ,
1253source quench
1254.Pq Cm 4 ,
1255redirect
1256.Pq Cm 5 ,
1257echo request
1258.Pq Cm 8 ,
1259router advertisement
1260.Pq Cm 9 ,
1261router solicitation
1262.Pq Cm 10 ,
1263time-to-live exceeded
1264.Pq Cm 11 ,
1265IP header bad
1266.Pq Cm 12 ,
1267timestamp request
1268.Pq Cm 13 ,
1269timestamp reply
1270.Pq Cm 14 ,
1271information request
1272.Pq Cm 15 ,
1273information reply
1274.Pq Cm 16 ,
1275address mask request
1276.Pq Cm 17
1277and address mask reply
1278.Pq Cm 18 .
1279.It Cm icmp6types Ar types
1280Matches ICMP6 packets whose ICMP6 type is in the list of
1281.Ar types .
1282The list may be specified as any combination of
1283individual types (numeric) separated by commas.
1284.Em Ranges are not allowed .
1285.It Cm in | out
1286Matches incoming or outgoing packets, respectively.
1287.Cm in
1288and
1289.Cm out
1290are mutually exclusive (in fact,
1291.Cm out
1292is implemented as
1293.Cm not in Ns No ).
1294.It Cm ipid Ar id-list
1295Matches IPv4 packets whose
1296.Cm ip_id
1297field has value included in
1298.Ar id-list ,
1299which is either a single value or a list of values or ranges
1300specified in the same way as
1301.Ar ports .
1302.It Cm iplen Ar len-list
1303Matches IP packets whose total length, including header and data, is
1304in the set
1305.Ar len-list ,
1306which is either a single value or a list of values or ranges
1307specified in the same way as
1308.Ar ports .
1309.It Cm ipoptions Ar spec
1310Matches packets whose IPv4 header contains the comma separated list of
1311options specified in
1312.Ar spec .
1313The supported IP options are:
1314.Pp
1315.Cm ssrr
1316(strict source route),
1317.Cm lsrr
1318(loose source route),
1319.Cm rr
1320(record packet route) and
1321.Cm ts
1322(timestamp).
1323The absence of a particular option may be denoted
1324with a
1325.Ql \&! .
1326.It Cm ipprecedence Ar precedence
1327Matches IPv4 packets whose precedence field is equal to
1328.Ar precedence .
1329.It Cm ipsec
1330Matches packets that have IPSEC history associated with them
1331(i.e., the packet comes encapsulated in IPSEC, the kernel
1332has IPSEC support and IPSEC_FILTERTUNNEL option, and can correctly
1333decapsulate it).
1334.Pp
1335Note that specifying
1336.Cm ipsec
1337is different from specifying
1338.Cm proto Ar ipsec
1339as the latter will only look at the specific IP protocol field,
1340irrespective of IPSEC kernel support and the validity of the IPSEC data.
1341.Pp
1342Further note that this flag is silently ignored in kernels without
1343IPSEC support.
1344It does not affect rule processing when given and the
1345rules are handled as if with no
1346.Cm ipsec
1347flag.
1348.It Cm iptos Ar spec
1349Matches IPv4 packets whose
1350.Cm tos
1351field contains the comma separated list of
1352service types specified in
1353.Ar spec .
1354The supported IP types of service are:
1355.Pp
1356.Cm lowdelay
1357.Pq Dv IPTOS_LOWDELAY ,
1358.Cm throughput
1359.Pq Dv IPTOS_THROUGHPUT ,
1360.Cm reliability
1361.Pq Dv IPTOS_RELIABILITY ,
1362.Cm mincost
1363.Pq Dv IPTOS_MINCOST ,
1364.Cm congestion
1365.Pq Dv IPTOS_ECN_CE .
1366The absence of a particular type may be denoted
1367with a
1368.Ql \&! .
1369.It Cm ipttl Ar ttl-list
1370Matches IPv4 packets whose time to live is included in
1371.Ar ttl-list ,
1372which is either a single value or a list of values or ranges
1373specified in the same way as
1374.Ar ports .
1375.It Cm ipversion Ar ver
1376Matches IP packets whose IP version field is
1377.Ar ver .
1378.It Cm keep-state
1379Upon a match, the firewall will create a dynamic rule, whose
1380default behaviour is to match bidirectional traffic between
1381source and destination IP/port using the same protocol.
1382The rule has a limited lifetime (controlled by a set of
1383.Xr sysctl 8
1384variables), and the lifetime is refreshed every time a matching
1385packet is found.
1386.It Cm layer2
1387Matches only layer2 packets, i.e., those passed to
1388.Nm
1389from ether_demux() and ether_output_frame().
1390.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
1391The firewall will only allow
1392.Ar N
1393connections with the same
1394set of parameters as specified in the rule.
1395One or more
1396of source and destination addresses and ports can be
1397specified.
1398Currently,
1399only IPv4 flows are supported.
1400.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar N
1401Search an entry in lookup table
1402.Ar N
1403that matches the field specified as argument.
1404If not found, the match fails.
1405Otherwise, the match succeeds and
1406.Cm tablearg
1407is set to the value extracted from the table.
1408.Pp
1409This option can be useful to quickly dispatch traffic based on
1410certain packet fields.
1411See the
1412.Sx LOOKUP TABLES
1413section below for more information on lookup tables.
1414.It Cm { MAC | mac } Ar dst-mac src-mac
1415Match packets with a given
1416.Ar dst-mac
1417and
1418.Ar src-mac
1419addresses, specified as the
1420.Cm any
1421keyword (matching any MAC address), or six groups of hex digits
1422separated by colons,
1423and optionally followed by a mask indicating the significant bits.
1424The mask may be specified using either of the following methods:
1425.Bl -enum -width indent
1426.It
1427A slash
1428.Pq /
1429followed by the number of significant bits.
1430For example, an address with 33 significant bits could be specified as:
1431.Pp
1432.Dl "MAC 10:20:30:40:50:60/33 any"
1433.Pp
1434.It
1435An ampersand
1436.Pq &
1437followed by a bitmask specified as six groups of hex digits separated
1438by colons.
1439For example, an address in which the last 16 bits are significant could
1440be specified as:
1441.Pp
1442.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1443.Pp
1444Note that the ampersand character has a special meaning in many shells
1445and should generally be escaped.
1446.Pp
1447.El
1448Note that the order of MAC addresses (destination first,
1449source second) is
1450the same as on the wire, but the opposite of the one used for
1451IP addresses.
1452.It Cm mac-type Ar mac-type
1453Matches packets whose Ethernet Type field
1454corresponds to one of those specified as argument.
1455.Ar mac-type
1456is specified in the same way as
1457.Cm port numbers
1458(i.e., one or more comma-separated single values or ranges).
1459You can use symbolic names for known values such as
1460.Em vlan , ipv4, ipv6 .
1461Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1462and they are always printed as hexadecimal (unless the
1463.Cm -N
1464option is used, in which case symbolic resolution will be attempted).
1465.It Cm proto Ar protocol
1466Matches packets with the corresponding IP protocol.
1467.It Cm recv | xmit | via Brq Ar ifX | Ar if Ns Cm * | Ar ipno | Ar any
1468Matches packets received, transmitted or going through,
1469respectively, the interface specified by exact name
1470.Ns No ( Ar ifX Ns No ),
1471by device name
1472.Ns No ( Ar if Ns Ar * Ns No ),
1473by IP address, or through some interface.
1474.Pp
1475The
1476.Cm via
1477keyword causes the interface to always be checked.
1478If
1479.Cm recv
1480or
1481.Cm xmit
1482is used instead of
1483.Cm via ,
1484then only the receive or transmit interface (respectively)
1485is checked.
1486By specifying both, it is possible to match packets based on
1487both receive and transmit interface, e.g.:
1488.Pp
1489.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1490.Pp
1491The
1492.Cm recv
1493interface can be tested on either incoming or outgoing packets,
1494while the
1495.Cm xmit
1496interface can only be tested on outgoing packets.
1497So
1498.Cm out
1499is required (and
1500.Cm in
1501is invalid) whenever
1502.Cm xmit
1503is used.
1504.Pp
1505A packet might not have a receive or transmit interface: packets
1506originating from the local host have no receive interface,
1507while packets destined for the local host have no transmit
1508interface.
1509.It Cm setup
1510Matches TCP packets that have the SYN bit set but no ACK bit.
1511This is the short form of
1512.Dq Li tcpflags\ syn,!ack .
1513.It Cm sockarg
1514Matches packets that are associated to a local socket and
1515for which the SO_USER_COOKIE socket option has been set
1516to a non-zero value. As a side effect, the value of the
1517option is made available as
1518.Cm tablearg
1519value, which in turn can be used as
1520.Cm skipto
1521or
1522.Cm pipe
1523number.
1524.It Cm src-ip Ar ip-address
1525Matches IPv4 packets whose source IP is one of the address(es)
1526specified as an argument.
1527.It Cm src-ip6 Ar ip6-address
1528Matches IPv6 packets whose source IP is one of the address(es)
1529specified as an argument.
1530.It Cm src-port Ar ports
1531Matches IP packets whose source port is one of the port(s)
1532specified as argument.
1533.It Cm tagged Ar tag-list
1534Matches packets whose tags are included in
1535.Ar tag-list ,
1536which is either a single value or a list of values or ranges
1537specified in the same way as
1538.Ar ports .
1539Tags can be applied to the packet using
1540.Cm tag
1541rule action parameter (see it's description for details on tags).
1542.It Cm tcpack Ar ack
1543TCP packets only.
1544Match if the TCP header acknowledgment number field is set to
1545.Ar ack .
1546.It Cm tcpdatalen Ar tcpdatalen-list
1547Matches TCP packets whose length of TCP data is
1548.Ar tcpdatalen-list ,
1549which is either a single value or a list of values or ranges
1550specified in the same way as
1551.Ar ports .
1552.It Cm tcpflags Ar spec
1553TCP packets only.
1554Match if the TCP header contains the comma separated list of
1555flags specified in
1556.Ar spec .
1557The supported TCP flags are:
1558.Pp
1559.Cm fin ,
1560.Cm syn ,
1561.Cm rst ,
1562.Cm psh ,
1563.Cm ack
1564and
1565.Cm urg .
1566The absence of a particular flag may be denoted
1567with a
1568.Ql \&! .
1569A rule which contains a
1570.Cm tcpflags
1571specification can never match a fragmented packet which has
1572a non-zero offset.
1573See the
1574.Cm frag
1575option for details on matching fragmented packets.
1576.It Cm tcpseq Ar seq
1577TCP packets only.
1578Match if the TCP header sequence number field is set to
1579.Ar seq .
1580.It Cm tcpwin Ar win
1581TCP packets only.
1582Match if the TCP header window field is set to
1583.Ar win .
1584.It Cm tcpoptions Ar spec
1585TCP packets only.
1586Match if the TCP header contains the comma separated list of
1587options specified in
1588.Ar spec .
1589The supported TCP options are:
1590.Pp
1591.Cm mss
1592(maximum segment size),
1593.Cm window
1594(tcp window advertisement),
1595.Cm sack
1596(selective ack),
1597.Cm ts
1598(rfc1323 timestamp) and
1599.Cm cc
1600(rfc1644 t/tcp connection count).
1601The absence of a particular option may be denoted
1602with a
1603.Ql \&! .
1604.It Cm uid Ar user
1605Match all TCP or UDP packets sent by or received for a
1606.Ar user .
1607A
1608.Ar user
1609may be matched by name or identification number.
1610.It Cm verrevpath
1611For incoming packets,
1612a routing table lookup is done on the packet's source address.
1613If the interface on which the packet entered the system matches the
1614outgoing interface for the route,
1615the packet matches.
1616If the interfaces do not match up,
1617the packet does not match.
1618All outgoing packets or packets with no incoming interface match.
1619.Pp
1620The name and functionality of the option is intentionally similar to
1621the Cisco IOS command:
1622.Pp
1623.Dl ip verify unicast reverse-path
1624.Pp
1625This option can be used to make anti-spoofing rules to reject all
1626packets with source addresses not from this interface.
1627See also the option
1628.Cm antispoof .
1629.It Cm versrcreach
1630For incoming packets,
1631a routing table lookup is done on the packet's source address.
1632If a route to the source address exists, but not the default route
1633or a blackhole/reject route, the packet matches.
1634Otherwise, the packet does not match.
1635All outgoing packets match.
1636.Pp
1637The name and functionality of the option is intentionally similar to
1638the Cisco IOS command:
1639.Pp
1640.Dl ip verify unicast source reachable-via any
1641.Pp
1642This option can be used to make anti-spoofing rules to reject all
1643packets whose source address is unreachable.
1644.It Cm antispoof
1645For incoming packets, the packet's source address is checked if it
1646belongs to a directly connected network.
1647If the network is directly connected, then the interface the packet
1648came on in is compared to the interface the network is connected to.
1649When incoming interface and directly connected interface are not the
1650same, the packet does not match.
1651Otherwise, the packet does match.
1652All outgoing packets match.
1653.Pp
1654This option can be used to make anti-spoofing rules to reject all
1655packets that pretend to be from a directly connected network but do
1656not come in through that interface.
1657This option is similar to but more restricted than
1658.Cm verrevpath
1659because it engages only on packets with source addresses of directly
1660connected networks instead of all source addresses.
1661.El
1662.Sh LOOKUP TABLES
1663Lookup tables are useful to handle large sparse sets of
1664addresses or other search keys (e.g. ports, jail IDs).
1665In the rest of this section we will use the term ``address''
1666to mean any unsigned value of up to 32-bit.
1667There may be up to 128 different lookup tables, numbered 0 to 127.
1668.Pp
1669Each entry is represented by an
1670.Ar addr Ns Op / Ns Ar masklen
1671and will match all addresses with base
1672.Ar addr
1673(specified as an IP address, a hostname or an unsigned integer)
1674and mask width of
1675.Ar masklen
1676bits.
1677If
1678.Ar masklen
1679is not specified, it defaults to 32.
1680When looking up an IP address in a table, the most specific
1681entry will match.
1682Associated with each entry is a 32-bit unsigned
1683.Ar value ,
1684which can optionally be checked by a rule matching code.
1685When adding an entry, if
1686.Ar value
1687is not specified, it defaults to 0.
1688.Pp
1689An entry can be added to a table
1690.Pq Cm add ,
1691or removed from a table
1692.Pq Cm delete .
1693A table can be examined
1694.Pq Cm list
1695or flushed
1696.Pq Cm flush .
1697.Pp
1698Internally, each table is stored in a Radix tree, the same way as
1699the routing table (see
1700.Xr route 4 ) .
1701.Pp
1702Lookup tables currently support only ports, jail IDs and IPv4 addresses.
1703.Pp
1704The
1705.Cm tablearg
1706feature provides the ability to use a value, looked up in the table, as
1707the argument for a rule action, action parameter or rule option.
1708This can significantly reduce number of rules in some configurations.
1709If two tables are used in a rule, the result of the second (destination)
1710is used.
1711The
1712.Cm tablearg
1713argument can be used with the following actions:
1714.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto
1715action parameters:
1716.Cm tag, untag,
1717rule options:
1718.Cm limit, tagged.
1719.Pp
1720When used with
1721.Cm fwd
1722it is possible to supply table entries with values
1723that are in the form of IP addresses or hostnames.
1724See the
1725.Sx EXAMPLES
1726Section for example usage of tables and the tablearg keyword.
1727.Pp
1728When used with the
1729.Cm skipto
1730action, the user should be aware that the code will walk the ruleset
1731up to a rule equal to, or past, the given number, and should therefore try keep the
1732ruleset compact between the skipto and the target rules.
1733.Sh SETS OF RULES
1734Each rule belongs to one of 32 different
1735.Em sets
1736, numbered 0 to 31.
1737Set 31 is reserved for the default rule.
1738.Pp
1739By default, rules are put in set 0, unless you use the
1740.Cm set N
1741attribute when entering a new rule.
1742Sets can be individually and atomically enabled or disabled,
1743so this mechanism permits an easy way to store multiple configurations
1744of the firewall and quickly (and atomically) switch between them.
1745The command to enable/disable sets is
1746.Bd -ragged -offset indent
1747.Nm
1748.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
1749.Ed
1750.Pp
1751where multiple
1752.Cm enable
1753or
1754.Cm disable
1755sections can be specified.
1756Command execution is atomic on all the sets specified in the command.
1757By default, all sets are enabled.
1758.Pp
1759When you disable a set, its rules behave as if they do not exist
1760in the firewall configuration, with only one exception:
1761.Bd -ragged -offset indent
1762dynamic rules created from a rule before it had been disabled
1763will still be active until they expire.
1764In order to delete
1765dynamic rules you have to explicitly delete the parent rule
1766which generated them.
1767.Ed
1768.Pp
1769The set number of rules can be changed with the command
1770.Bd -ragged -offset indent
1771.Nm
1772.Cm set move
1773.Brq Cm rule Ar rule-number | old-set
1774.Cm to Ar new-set
1775.Ed
1776.Pp
1777Also, you can atomically swap two rulesets with the command
1778.Bd -ragged -offset indent
1779.Nm
1780.Cm set swap Ar first-set second-set
1781.Ed
1782.Pp
1783See the
1784.Sx EXAMPLES
1785Section on some possible uses of sets of rules.
1786.Sh STATEFUL FIREWALL
1787Stateful operation is a way for the firewall to dynamically
1788create rules for specific flows when packets that
1789match a given pattern are detected.
1790Support for stateful
1791operation comes through the
1792.Cm check-state , keep-state
1793and
1794.Cm limit
1795options of
1796.Nm rules .
1797.Pp
1798Dynamic rules are created when a packet matches a
1799.Cm keep-state
1800or
1801.Cm limit
1802rule, causing the creation of a
1803.Em dynamic
1804rule which will match all and only packets with
1805a given
1806.Em protocol
1807between a
1808.Em src-ip/src-port dst-ip/dst-port
1809pair of addresses
1810.Em ( src
1811and
1812.Em dst
1813are used here only to denote the initial match addresses, but they
1814are completely equivalent afterwards).
1815Dynamic rules will be checked at the first
1816.Cm check-state, keep-state
1817or
1818.Cm limit
1819occurrence, and the action performed upon a match will be the same
1820as in the parent rule.
1821.Pp
1822Note that no additional attributes other than protocol and IP addresses
1823and ports are checked on dynamic rules.
1824.Pp
1825The typical use of dynamic rules is to keep a closed firewall configuration,
1826but let the first TCP SYN packet from the inside network install a
1827dynamic rule for the flow so that packets belonging to that session
1828will be allowed through the firewall:
1829.Pp
1830.Dl "ipfw add check-state"
1831.Dl "ipfw add allow tcp from my-subnet to any setup keep-state"
1832.Dl "ipfw add deny tcp from any to any"
1833.Pp
1834A similar approach can be used for UDP, where an UDP packet coming
1835from the inside will install a dynamic rule to let the response through
1836the firewall:
1837.Pp
1838.Dl "ipfw add check-state"
1839.Dl "ipfw add allow udp from my-subnet to any keep-state"
1840.Dl "ipfw add deny udp from any to any"
1841.Pp
1842Dynamic rules expire after some time, which depends on the status
1843of the flow and the setting of some
1844.Cm sysctl
1845variables.
1846See Section
1847.Sx SYSCTL VARIABLES
1848for more details.
1849For TCP sessions, dynamic rules can be instructed to periodically
1850send keepalive packets to refresh the state of the rule when it is
1851about to expire.
1852.Pp
1853See Section
1854.Sx EXAMPLES
1855for more examples on how to use dynamic rules.
1856.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
1857.Nm
1858is also the user interface for the
1859.Nm dummynet
1860traffic shaper, packet scheduler and network emulator, a subsystem that
1861can artificially queue, delay or drop packets
1862emulating the behaviour of certain network links
1863or queueing systems.
1864.Pp
1865.Nm dummynet
1866operates by first using the firewall to select packets
1867using any match pattern that can be used in
1868.Nm
1869rules.
1870Matching packets are then passed to either of two
1871different objects, which implement the traffic regulation:
1872.Bl -hang -offset XXXX
1873.It Em pipe
1874A
1875.Em pipe
1876emulates a
1877.Em link
1878with given bandwidth and propagation delay,
1879driven by a FIFO scheduler and a single queue with programmable
1880queue size and packet loss rate.
1881Packets are appended to the queue as they come out from
1882.Nm ipfw ,
1883and then transferred in FIFO order to the link at the desired rate.
1884.It Em queue
1885A
1886.Em queue
1887is an abstraction used to implement packet scheduling
1888using one of several packet scheduling algorithms.
1889Packets sent to a
1890.Em queue
1891are first grouped into flows according to a mask on the 5-tuple.
1892Flows are then passed to the scheduler associated to the
1893.Em queue ,
1894and each flow uses scheduling parameters (weight and others)
1895as configured in the
1896.Em queue
1897itself.
1898A scheduler in turn is connected to an emulated link,
1899and arbitrates the link's bandwidth among backlogged flows according to
1900weights and to the features of the scheduling algorithm in use.
1901.El
1902.Pp
1903In practice,
1904.Em pipes
1905can be used to set hard limits to the bandwidth that a flow can use, whereas
1906.Em queues
1907can be used to determine how different flows share the available bandwidth.
1908.Pp
1909A graphical representation of the binding of queues,
1910flows, schedulers and links is below.
1911.Bd -literal -offset indent
1912                 (flow_mask|sched_mask)  sched_mask
1913         +---------+   weight Wx  +-------------+
1914         |         |->-[flow]-->--|             |-+
1915    -->--| QUEUE x |   ...        |             | |
1916         |         |->-[flow]-->--| SCHEDuler N | |
1917         +---------+              |             | |
1918             ...                  |             +--[LINK N]-->--
1919         +---------+   weight Wy  |             | +--[LINK N]-->--
1920         |         |->-[flow]-->--|             | |
1921    -->--| QUEUE y |   ...        |             | |
1922         |         |->-[flow]-->--|             | |
1923         +---------+              +-------------+ |
1924                                    +-------------+
1925.Ed
1926It is important to understand the role of the SCHED_MASK
1927and FLOW_MASK, which are configured through the commands
1928.Dl "ipfw sched N config mask SCHED_MASK ..."
1929and
1930.Dl "ipfw queue X config mask FLOW_MASK ..." .
1931.Pp
1932The SCHED_MASK is used to assign flows to one or more
1933scheduler instances, one for each
1934value of the packet's 5-tuple after applying SCHED_MASK.
1935As an example, using ``src-ip 0xffffff00'' creates one instance
1936for each /24 destination subnet.
1937.Pp
1938The FLOW_MASK, together with the SCHED_MASK, is used to split
1939packets into flows. As an example, using
1940``src-ip 0x000000ff''
1941together with the previous SCHED_MASK makes a flow for
1942each individual source address. In turn, flows for each /24
1943subnet will be sent to the same scheduler instance.
1944.Pp
1945The above diagram holds even for the
1946.Em pipe
1947case, with the only restriction that a
1948.Em pipe
1949only supports a SCHED_MASK, and forces the use of a FIFO
1950scheduler (these are for backward compatibility reasons;
1951in fact, internally, a
1952.Nm dummynet's
1953pipe is implemented exactly as above).
1954.Pp
1955There are two modes of
1956.Nm dummynet
1957operation:
1958.Dq normal
1959and
1960.Dq fast .
1961The
1962.Dq normal
1963mode tries to emulate a real link: the
1964.Nm dummynet
1965scheduler ensures that the packet will not leave the pipe faster than it
1966would on the real link with a given bandwidth.
1967The
1968.Dq fast
1969mode allows certain packets to bypass the
1970.Nm dummynet
1971scheduler (if packet flow does not exceed pipe's bandwidth).
1972This is the reason why the
1973.Dq fast
1974mode requires less CPU cycles per packet (on average) and packet latency
1975can be significantly lower in comparison to a real link with the same
1976bandwidth.
1977The default mode is
1978.Dq normal .
1979The
1980.Dq fast
1981mode can be enabled by setting the
1982.Va net.inet.ip.dummynet.io_fast
1983.Xr sysctl 8
1984variable to a non-zero value.
1985.Pp
1986.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
1987The
1988.Em pipe ,
1989.Em queue
1990and
1991.Em scheduler
1992configuration commands are the following:
1993.Bd -ragged -offset indent
1994.Cm pipe Ar number Cm config Ar pipe-configuration
1995.Pp
1996.Cm queue Ar number Cm config Ar queue-configuration
1997.Pp
1998.Cm sched Ar number Cm config Ar sched-configuration
1999.Ed
2000.Pp
2001The following parameters can be configured for a pipe:
2002.Pp
2003.Bl -tag -width indent -compact
2004.It Cm bw Ar bandwidth | device
2005Bandwidth, measured in
2006.Sm off
2007.Op Cm K | M
2008.Brq Cm bit/s | Byte/s .
2009.Sm on
2010.Pp
2011A value of 0 (default) means unlimited bandwidth.
2012The unit must immediately follow the number, as in
2013.Pp
2014.Dl "ipfw pipe 1 config bw 300Kbit/s"
2015.Pp
2016If a device name is specified instead of a numeric value, as in
2017.Pp
2018.Dl "ipfw pipe 1 config bw tun0"
2019.Pp
2020then the transmit clock is supplied by the specified device.
2021At the moment only the
2022.Xr tun 4
2023device supports this
2024functionality, for use in conjunction with
2025.Xr ppp 8 .
2026.Pp
2027.It Cm delay Ar ms-delay
2028Propagation delay, measured in milliseconds.
2029The value is rounded to the next multiple of the clock tick
2030(typically 10ms, but it is a good practice to run kernels
2031with
2032.Dq "options HZ=1000"
2033to reduce
2034the granularity to 1ms or less).
2035The default value is 0, meaning no delay.
2036.Pp
2037.It Cm burst Ar size
2038If the data to be sent exceeds the pipe's bandwidth limit
2039(and the pipe was previously idle), up to
2040.Ar size
2041bytes of data are allowed to bypass the
2042.Nm dummynet
2043scheduler, and will be sent as fast as the physical link allows.
2044Any additional data will be transmitted at the rate specified
2045by the
2046.Nm pipe
2047bandwidth.
2048The burst size depends on how long the pipe has been idle;
2049the effective burst size is calculated as follows:
2050MAX(
2051.Ar size
2052,
2053.Nm bw
2054* pipe_idle_time).
2055.Pp
2056.It Cm profile Ar filename
2057A file specifying the additional overhead incurred in the transmission
2058of a packet on the link.
2059.Pp
2060Some link types introduce extra delays in the transmission
2061of a packet, e.g. because of MAC level framing, contention on
2062the use of the channel, MAC level retransmissions and so on.
2063From our point of view, the channel is effectively unavailable
2064for this extra time, which is constant or variable depending
2065on the link type. Additionally, packets may be dropped after this
2066time (e.g. on a wireless link after too many retransmissions).
2067We can model the additional delay with an empirical curve
2068that represents its distribution.
2069.Bd -literal -offset indent
2070      cumulative probability
2071      1.0 ^
2072          |
2073      L   +-- loss-level          x
2074          |                 ******
2075          |                *
2076          |           *****
2077          |          *
2078          |        **
2079          |       *
2080          +-------*------------------->
2081                      delay
2082.Ed
2083The empirical curve may have both vertical and horizontal lines.
2084Vertical lines represent constant delay for a range of
2085probabilities.
2086Horizontal lines correspond to a discontinuity in the delay
2087distribution: the pipe will use the largest delay for a
2088given probability.
2089.Pp
2090The file format is the following, with whitespace acting as
2091a separator and '#' indicating the beginning a comment:
2092.Bl -tag -width indent
2093.It Cm name Ar identifier
2094optional name (listed by "ipfw pipe show")
2095to identify the delay distribution;
2096.It Cm bw Ar value
2097the bandwidth used for the pipe.
2098If not specified here, it must be present
2099explicitly as a configuration parameter for the pipe;
2100.It Cm loss-level Ar L
2101the probability above which packets are lost.
2102(0.0 <= L <= 1.0, default 1.0 i.e. no loss);
2103.It Cm samples Ar N
2104the number of samples used in the internal
2105representation of the curve (2..1024; default 100);
2106.It Cm "delay prob" | "prob delay"
2107One of these two lines is mandatory and defines
2108the format of the following lines with data points.
2109.It Ar XXX Ar YYY
21102 or more lines representing points in the curve,
2111with either delay or probability first, according
2112to the chosen format.
2113The unit for delay is milliseconds.
2114Data points do not need to be sorted.
2115Also, the number of actual lines can be different
2116from the value of the "samples" parameter:
2117.Nm
2118utility will sort and interpolate
2119the curve as needed.
2120.El
2121.Pp
2122Example of a profile file:
2123.Bd -literal -offset indent
2124name    bla_bla_bla
2125samples 100
2126loss-level    0.86
2127prob    delay
21280       200	# minimum overhead is 200ms
21290.5     200
21300.5     300
21310.8     1000
21320.9     1300
21331       1300
2134#configuration file end
2135.Ed
2136.El
2137.Pp
2138The following parameters can be configured for a queue:
2139.Pp
2140.Bl -tag -width indent -compact
2141.It Cm pipe Ar pipe_nr
2142Connects a queue to the specified pipe.
2143Multiple queues (with the same or different weights) can be connected to
2144the same pipe, which specifies the aggregate rate for the set of queues.
2145.Pp
2146.It Cm weight Ar weight
2147Specifies the weight to be used for flows matching this queue.
2148The weight must be in the range 1..100, and defaults to 1.
2149.El
2150.Pp
2151The following parameters can be configured for a scheduler:
2152.Pp
2153.Bl -tag -width indent -compact
2154.It Cm type Ar {fifo | wf2qp | rr | qfq}
2155specifies the scheduling algorithm to use.
2156.Bl -tag -width indent -compact
2157.It cm fifo
2158is just a FIFO scheduler (which means that all packets
2159are stored in the same queue as they arrive to the scheduler).
2160FIFO has O(1) per-packet time complexity, with very low
2161constants (estimate 60-80ns on a 2Ghz desktop machine)
2162but gives no service guarantees.
2163.It Cm wf2qp
2164implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2165algorithm which permits flows to share bandwidth according to
2166their weights. Note that weights are not priorities; even a flow
2167with a minuscule weight will never starve.
2168WF2Q+ has O(log N) per-packet processing cost, where N is the number
2169of flows, and is the default algorithm used by previous versions
2170dummynet's queues.
2171.It Cm rr
2172implements the Deficit Round Robin algorithm, which has O(1) processing
2173costs (roughly, 100-150ns per packet)
2174and permits bandwidth allocation according to weights, but
2175with poor service guarantees.
2176.It Cm qfq
2177implements the QFQ algorithm, which is a very fast variant of
2178WF2Q+, with similar service guarantees and O(1) processing
2179costs (roughly, 200-250ns per packet).
2180.El
2181.El
2182.Pp
2183In addition to the type, all parameters allowed for a pipe can also
2184be specified for a scheduler.
2185.Pp
2186Finally, the following parameters can be configured for both
2187pipes and queues:
2188.Pp
2189.Bl -tag -width XXXX -compact
2190.It Cm buckets Ar hash-table-size
2191Specifies the size of the hash table used for storing the
2192various queues.
2193Default value is 64 controlled by the
2194.Xr sysctl 8
2195variable
2196.Va net.inet.ip.dummynet.hash_size ,
2197allowed range is 16 to 65536.
2198.Pp
2199.It Cm mask Ar mask-specifier
2200Packets sent to a given pipe or queue by an
2201.Nm
2202rule can be further classified into multiple flows, each of which is then
2203sent to a different
2204.Em dynamic
2205pipe or queue.
2206A flow identifier is constructed by masking the IP addresses,
2207ports and protocol types as specified with the
2208.Cm mask
2209options in the configuration of the pipe or queue.
2210For each different flow identifier, a new pipe or queue is created
2211with the same parameters as the original object, and matching packets
2212are sent to it.
2213.Pp
2214Thus, when
2215.Em dynamic pipes
2216are used, each flow will get the same bandwidth as defined by the pipe,
2217whereas when
2218.Em dynamic queues
2219are used, each flow will share the parent's pipe bandwidth evenly
2220with other flows generated by the same queue (note that other queues
2221with different weights might be connected to the same pipe).
2222.br
2223Available mask specifiers are a combination of one or more of the following:
2224.Pp
2225.Cm dst-ip Ar mask ,
2226.Cm dst-ip6 Ar mask ,
2227.Cm src-ip Ar mask ,
2228.Cm src-ip6 Ar mask ,
2229.Cm dst-port Ar mask ,
2230.Cm src-port Ar mask ,
2231.Cm flow-id Ar mask ,
2232.Cm proto Ar mask
2233or
2234.Cm all ,
2235.Pp
2236where the latter means all bits in all fields are significant.
2237.Pp
2238.It Cm noerror
2239When a packet is dropped by a
2240.Nm dummynet
2241queue or pipe, the error
2242is normally reported to the caller routine in the kernel, in the
2243same way as it happens when a device queue fills up.
2244Setting this
2245option reports the packet as successfully delivered, which can be
2246needed for some experimental setups where you want to simulate
2247loss or congestion at a remote router.
2248.Pp
2249.It Cm plr Ar packet-loss-rate
2250Packet loss rate.
2251Argument
2252.Ar packet-loss-rate
2253is a floating-point number between 0 and 1, with 0 meaning no
2254loss, 1 meaning 100% loss.
2255The loss rate is internally represented on 31 bits.
2256.Pp
2257.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2258Queue size, in
2259.Ar slots
2260or
2261.Cm KBytes .
2262Default value is 50 slots, which
2263is the typical queue size for Ethernet devices.
2264Note that for slow speed links you should keep the queue
2265size short or your traffic might be affected by a significant
2266queueing delay.
2267E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2268or 20s of queue on a 30Kbit/s pipe.
2269Even worse effects can result if you get packets from an
2270interface with a much larger MTU, e.g.\& the loopback interface
2271with its 16KB packets.
2272The
2273.Xr sysctl 8
2274variables
2275.Em net.inet.ip.dummynet.pipe_byte_limit
2276and
2277.Em net.inet.ip.dummynet.pipe_slot_limit
2278control the maximum lengths that can be specified.
2279.Pp
2280.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2281Make use of the RED (Random Early Detection) queue management algorithm.
2282.Ar w_q
2283and
2284.Ar max_p
2285are floating
2286point numbers between 0 and 1 (0 not included), while
2287.Ar min_th
2288and
2289.Ar max_th
2290are integer numbers specifying thresholds for queue management
2291(thresholds are computed in bytes if the queue has been defined
2292in bytes, in slots otherwise).
2293The
2294.Nm dummynet
2295also supports the gentle RED variant (gred).
2296Three
2297.Xr sysctl 8
2298variables can be used to control the RED behaviour:
2299.Bl -tag -width indent
2300.It Va net.inet.ip.dummynet.red_lookup_depth
2301specifies the accuracy in computing the average queue
2302when the link is idle (defaults to 256, must be greater than zero)
2303.It Va net.inet.ip.dummynet.red_avg_pkt_size
2304specifies the expected average packet size (defaults to 512, must be
2305greater than zero)
2306.It Va net.inet.ip.dummynet.red_max_pkt_size
2307specifies the expected maximum packet size, only used when queue
2308thresholds are in bytes (defaults to 1500, must be greater than zero).
2309.El
2310.El
2311.Pp
2312When used with IPv6 data,
2313.Nm dummynet
2314currently has several limitations.
2315Information necessary to route link-local packets to an
2316interface is not available after processing by
2317.Nm dummynet
2318so those packets are dropped in the output path.
2319Care should be taken to ensure that link-local packets are not passed to
2320.Nm dummynet .
2321.Sh CHECKLIST
2322Here are some important points to consider when designing your
2323rules:
2324.Bl -bullet
2325.It
2326Remember that you filter both packets going
2327.Cm in
2328and
2329.Cm out .
2330Most connections need packets going in both directions.
2331.It
2332Remember to test very carefully.
2333It is a good idea to be near the console when doing this.
2334If you cannot be near the console,
2335use an auto-recovery script such as the one in
2336.Pa /usr/share/examples/ipfw/change_rules.sh .
2337.It
2338Do not forget the loopback interface.
2339.El
2340.Sh FINE POINTS
2341.Bl -bullet
2342.It
2343There are circumstances where fragmented datagrams are unconditionally
2344dropped.
2345TCP packets are dropped if they do not contain at least 20 bytes of
2346TCP header, UDP packets are dropped if they do not contain a full 8
2347byte UDP header, and ICMP packets are dropped if they do not contain
23484 bytes of ICMP header, enough to specify the ICMP type, code, and
2349checksum.
2350These packets are simply logged as
2351.Dq pullup failed
2352since there may not be enough good data in the packet to produce a
2353meaningful log entry.
2354.It
2355Another type of packet is unconditionally dropped, a TCP packet with a
2356fragment offset of one.
2357This is a valid packet, but it only has one use, to try
2358to circumvent firewalls.
2359When logging is enabled, these packets are
2360reported as being dropped by rule -1.
2361.It
2362If you are logged in over a network, loading the
2363.Xr kld 4
2364version of
2365.Nm
2366is probably not as straightforward as you would think.
2367The following command line is recommended:
2368.Bd -literal -offset indent
2369kldload ipfw && \e
2370ipfw add 32000 allow ip from any to any
2371.Ed
2372.Pp
2373Along the same lines, doing an
2374.Bd -literal -offset indent
2375ipfw flush
2376.Ed
2377.Pp
2378in similar surroundings is also a bad idea.
2379.It
2380The
2381.Nm
2382filter list may not be modified if the system security level
2383is set to 3 or higher
2384(see
2385.Xr init 8
2386for information on system security levels).
2387.El
2388.Sh PACKET DIVERSION
2389A
2390.Xr divert 4
2391socket bound to the specified port will receive all packets
2392diverted to that port.
2393If no socket is bound to the destination port, or if the divert module is
2394not loaded, or if the kernel was not compiled with divert socket support,
2395the packets are dropped.
2396.Sh NETWORK ADDRESS TRANSLATION (NAT)
2397.Pp
2398.Nm
2399support in-kernel NAT using the kernel version of
2400.Xr libalias 3 .
2401.Pp
2402The nat configuration command is the following:
2403.Bd -ragged -offset indent
2404.Bk -words
2405.Cm nat
2406.Ar nat_number
2407.Cm config
2408.Ar nat-configuration
2409.Ek
2410.Ed
2411.Pp
2412The following parameters can be configured:
2413.Bl -tag -width indent
2414.It Cm ip Ar ip_address
2415Define an ip address to use for aliasing.
2416.It Cm if Ar nic
2417Use ip address of NIC for aliasing, dynamically changing
2418it if NIC's ip address changes.
2419.It Cm log
2420Enable logging on this nat instance.
2421.It Cm deny_in
2422Deny any incoming connection from outside world.
2423.It Cm same_ports
2424Try to leave the alias port numbers unchanged from
2425the actual local port numbers.
2426.It Cm unreg_only
2427Traffic on the local network not originating from an
2428unregistered address spaces will be ignored.
2429.It Cm reset
2430Reset table of the packet aliasing engine on address change.
2431.It Cm reverse
2432Reverse the way libalias handles aliasing.
2433.It Cm proxy_only
2434Obey transparent proxy rules only, packet aliasing is not performed.
2435.El
2436.Pp
2437To let the packet continue after being (de)aliased, set the sysctl variable
2438.Va net.inet.ip.fw.one_pass
2439to 0.
2440For more information about aliasing modes, refer to
2441.Xr libalias 3 .
2442See Section
2443.Sx EXAMPLES
2444for some examples about nat usage.
2445.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
2446Redirect and LSNAT support follow closely the syntax used in
2447.Xr natd 8 .
2448See Section
2449.Sx EXAMPLES
2450for some examples on how to do redirect and lsnat.
2451.Ss SCTP NAT SUPPORT
2452SCTP nat can be configured in a similar manner to TCP through the
2453.Nm
2454command line tool.
2455The main difference is that
2456.Nm sctp nat
2457does not do port translation.
2458Since the local and global side ports will be the same,
2459there is no need to specify both.
2460Ports are redirected as follows:
2461.Bd -ragged -offset indent
2462.Bk -words
2463.Cm nat
2464.Ar nat_number
2465.Cm config if
2466.Ar nic
2467.Cm redirect_port sctp
2468.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
2469.Ek
2470.Ed
2471.Pp
2472Most
2473.Nm sctp nat
2474configuration can be done in real-time through the
2475.Xr sysctl 8
2476interface.
2477All may be changed dynamically, though the hash_table size will only
2478change for new
2479.Nm nat
2480instances.
2481See
2482.Sx SYSCTL VARIABLES
2483for more info.
2484.Sh SYSCTL VARIABLES
2485A set of
2486.Xr sysctl 8
2487variables controls the behaviour of the firewall and
2488associated modules
2489.Pq Nm dummynet , bridge , sctp nat .
2490These are shown below together with their default value
2491(but always check with the
2492.Xr sysctl 8
2493command what value is actually in use) and meaning:
2494.Bl -tag -width indent
2495.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
2496Defines how the
2497.Nm nat
2498responds to receipt of global OOTB ASCONF-AddIP:
2499.Bl -tag -width indent
2500.It Cm 0
2501No response (unless a partially matching association exists -
2502ports and vtags match but global address does not)
2503.It Cm 1
2504.Nm nat
2505will accept and process all OOTB global AddIP messages.
2506.El
2507.Pp
2508Option 1 should never be selected as this forms a security risk.
2509An attacker can
2510establish multiple fake associations by sending AddIP messages.
2511.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
2512Defines the maximum number of chunks in an SCTP packet that will be parsed for a
2513packet that matches an existing association.
2514This value is enforced to be greater or equal than
2515.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
2516A high value is
2517a DoS risk yet setting too low a value may result in important control chunks in
2518the packet not being located and parsed.
2519.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
2520Defines when the
2521.Nm nat
2522responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
2523An OOTB packet is a packet that arrives with no existing association
2524registered in the
2525.Nm nat
2526and is not an INIT or ASCONF-AddIP packet:
2527.Bl -tag -width indent
2528.It Cm 0
2529ErrorM is never sent in response to OOTB packets.
2530.It Cm 1
2531ErrorM is only sent to OOTB packets received on the local side.
2532.It Cm 2
2533ErrorM is sent to the local side and on the global side ONLY if there is a
2534partial match (ports and vtags match but the source global IP does not).
2535This value is only useful if the
2536.Nm nat
2537is tracking global IP addresses.
2538.It Cm 3
2539ErrorM is sent in response to all OOTB packets on both the local and global side
2540(DoS risk).
2541.El
2542.Pp
2543At the moment the default is 0, since the ErrorM packet is not yet
2544supported by most SCTP stacks.
2545When it is supported, and if not tracking
2546global addresses, we recommend setting this value to 1 to allow
2547multi-homed local hosts to function with the
2548.Nm nat .
2549To track global addresses, we recommend setting this value to 2 to
2550allow global hosts to be informed when they need to (re)send an
2551ASCONF-AddIP.
2552Value 3 should never be chosen (except for debugging) as the
2553.Nm nat
2554will respond to all OOTB global packets (a DoS risk).
2555.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
2556Size of hash tables used for
2557.Nm nat
2558lookups (100 < prime_number > 1000001).
2559This value sets the
2560.Nm hash table
2561size for any future created
2562.Nm nat
2563instance and therefore must be set prior to creating a
2564.Nm nat
2565instance.
2566The table sizes may be changed to suit specific needs.
2567If there will be few
2568concurrent associations, and memory is scarce, you may make these smaller.
2569If there will be many thousands (or millions) of concurrent associations, you
2570should make these larger.
2571A prime number is best for the table size.
2572The sysctl
2573update function will adjust your input value to the next highest prime number.
2574.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
2575Hold association in table for this many seconds after receiving a
2576SHUTDOWN-COMPLETE.
2577This allows endpoints to correct shutdown gracefully if a
2578shutdown_complete is lost and retransmissions are required.
2579.It Va net.inet.ip.alias.sctp.init_timer: No 15
2580Timeout value while waiting for (INIT-ACK|AddIP-ACK).
2581This value cannot be 0.
2582.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
2583Defines the maximum number of chunks in an SCTP packet that will be parsed when
2584no existing association exists that matches that packet.
2585Ideally this packet
2586will only be an INIT or ASCONF-AddIP packet.
2587A higher value may become a DoS
2588risk as malformed packets can consume processing resources.
2589.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
2590Defines the maximum number of parameters within a chunk that will be parsed in a
2591packet.
2592As for other similar sysctl variables, larger values pose a DoS risk.
2593.It Va net.inet.ip.alias.sctp.log_level: No 0
2594Level of detail in the system log messages (0 \- minimal, 1 \- event,
25952 \- info, 3 \- detail, 4 \- debug, 5 \- max debug). May be a good
2596option in high loss environments.
2597.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
2598Timeout value while waiting for SHUTDOWN-COMPLETE.
2599This value cannot be 0.
2600.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
2601Enables/disables global IP address tracking within the
2602.Nm nat
2603and places an
2604upper limit on the number of addresses tracked for each association:
2605.Bl -tag -width indent
2606.It Cm 0
2607Global tracking is disabled
2608.It Cm >1
2609Enables tracking, the maximum number of addresses tracked for each
2610association is limited to this value
2611.El
2612.Pp
2613This variable is fully dynamic, the new value will be adopted for all newly
2614arriving associations, existing associations are treated as they were previously.
2615Global tracking will decrease the number of collisions within the
2616.Nm nat
2617at a cost
2618of increased processing load, memory usage, complexity, and possible
2619.Nm nat
2620state
2621problems in complex networks with multiple
2622.Nm nats .
2623We recommend not tracking
2624global IP addresses, this will still result in a fully functional
2625.Nm nat .
2626.It Va net.inet.ip.alias.sctp.up_timer: No 300
2627Timeout value to keep an association up with no traffic.
2628This value cannot be 0.
2629.It Va net.inet.ip.dummynet.expire : No 1
2630Lazily delete dynamic pipes/queue once they have no pending traffic.
2631You can disable this by setting the variable to 0, in which case
2632the pipes/queues will only be deleted when the threshold is reached.
2633.It Va net.inet.ip.dummynet.hash_size : No 64
2634Default size of the hash table used for dynamic pipes/queues.
2635This value is used when no
2636.Cm buckets
2637option is specified when configuring a pipe/queue.
2638.It Va net.inet.ip.dummynet.io_fast : No 0
2639If set to a non-zero value,
2640the
2641.Dq fast
2642mode of
2643.Nm dummynet
2644operation (see above) is enabled.
2645.It Va net.inet.ip.dummynet.io_pkt
2646Number of packets passed to
2647.Nm dummynet .
2648.It Va net.inet.ip.dummynet.io_pkt_drop
2649Number of packets dropped by
2650.Nm dummynet .
2651.It Va net.inet.ip.dummynet.io_pkt_fast
2652Number of packets bypassed by the
2653.Nm dummynet
2654scheduler.
2655.It Va net.inet.ip.dummynet.max_chain_len : No 16
2656Target value for the maximum number of pipes/queues in a hash bucket.
2657The product
2658.Cm max_chain_len*hash_size
2659is used to determine the threshold over which empty pipes/queues
2660will be expired even when
2661.Cm net.inet.ip.dummynet.expire=0 .
2662.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
2663.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
2664.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
2665Parameters used in the computations of the drop probability
2666for the RED algorithm.
2667.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
2668.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
2669The maximum queue size that can be specified in bytes or packets.
2670These limits prevent accidental exhaustion of resources such as mbufs.
2671If you raise these limits,
2672you should make sure the system is configured so that sufficient resources
2673are available.
2674.It Va net.inet.ip.fw.autoinc_step : No 100
2675Delta between rule numbers when auto-generating them.
2676The value must be in the range 1..1000.
2677.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
2678The current number of buckets in the hash table for dynamic rules
2679(readonly).
2680.It Va net.inet.ip.fw.debug : No 1
2681Controls debugging messages produced by
2682.Nm .
2683.It Va net.inet.ip.fw.default_rule : No 65535
2684The default rule number (read-only).
2685By the design of
2686.Nm , the default rule is the last one, so its number
2687can also serve as the highest number allowed for a rule.
2688.It Va net.inet.ip.fw.dyn_buckets : No 256
2689The number of buckets in the hash table for dynamic rules.
2690Must be a power of 2, up to 65536.
2691It only takes effect when all dynamic rules have expired, so you
2692are advised to use a
2693.Cm flush
2694command to make sure that the hash table is resized.
2695.It Va net.inet.ip.fw.dyn_count : No 3
2696Current number of dynamic rules
2697(read-only).
2698.It Va net.inet.ip.fw.dyn_keepalive : No 1
2699Enables generation of keepalive packets for
2700.Cm keep-state
2701rules on TCP sessions.
2702A keepalive is generated to both
2703sides of the connection every 5 seconds for the last 20
2704seconds of the lifetime of the rule.
2705.It Va net.inet.ip.fw.dyn_max : No 8192
2706Maximum number of dynamic rules.
2707When you hit this limit, no more dynamic rules can be
2708installed until old ones expire.
2709.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
2710.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
2711.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
2712.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
2713.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
2714.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
2715These variables control the lifetime, in seconds, of dynamic
2716rules.
2717Upon the initial SYN exchange the lifetime is kept short,
2718then increased after both SYN have been seen, then decreased
2719again during the final FIN exchange or when a RST is received.
2720Both
2721.Em dyn_fin_lifetime
2722and
2723.Em dyn_rst_lifetime
2724must be strictly lower than 5 seconds, the period of
2725repetition of keepalives.
2726The firewall enforces that.
2727.It Va net.inet.ip.fw.enable : No 1
2728Enables the firewall.
2729Setting this variable to 0 lets you run your machine without
2730firewall even if compiled in.
2731.It Va net.inet6.ip6.fw.enable : No 1
2732provides the same functionality as above for the IPv6 case.
2733.It Va net.inet.ip.fw.one_pass : No 1
2734When set, the packet exiting from the
2735.Nm dummynet
2736pipe or from
2737.Xr ng_ipfw 4
2738node is not passed though the firewall again.
2739Otherwise, after an action, the packet is
2740reinjected into the firewall at the next rule.
2741.It Va net.inet.ip.fw.tables_max : No 128
2742Maximum number of tables (read-only).
2743.It Va net.inet.ip.fw.verbose : No 1
2744Enables verbose messages.
2745.It Va net.inet.ip.fw.verbose_limit : No 0
2746Limits the number of messages produced by a verbose firewall.
2747.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
2748If enabled packets with unknown IPv6 Extension Headers will be denied.
2749.It Va net.link.ether.ipfw : No 0
2750Controls whether layer-2 packets are passed to
2751.Nm .
2752Default is no.
2753.It Va net.link.bridge.ipfw : No 0
2754Controls whether bridged packets are passed to
2755.Nm .
2756Default is no.
2757.El
2758.Pp
2759.Sh EXAMPLES
2760There are far too many possible uses of
2761.Nm
2762so this Section will only give a small set of examples.
2763.Pp
2764.Ss BASIC PACKET FILTERING
2765This command adds an entry which denies all tcp packets from
2766.Em cracker.evil.org
2767to the telnet port of
2768.Em wolf.tambov.su
2769from being forwarded by the host:
2770.Pp
2771.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
2772.Pp
2773This one disallows any connection from the entire cracker's
2774network to my host:
2775.Pp
2776.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
2777.Pp
2778A first and efficient way to limit access (not using dynamic rules)
2779is the use of the following rules:
2780.Pp
2781.Dl "ipfw add allow tcp from any to any established"
2782.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
2783.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
2784.Dl "..."
2785.Dl "ipfw add deny tcp from any to any"
2786.Pp
2787The first rule will be a quick match for normal TCP packets,
2788but it will not match the initial SYN packet, which will be
2789matched by the
2790.Cm setup
2791rules only for selected source/destination pairs.
2792All other SYN packets will be rejected by the final
2793.Cm deny
2794rule.
2795.Pp
2796If you administer one or more subnets, you can take advantage
2797of the address sets and or-blocks and write extremely
2798compact rulesets which selectively enable services to blocks
2799of clients, as below:
2800.Pp
2801.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
2802.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
2803.Dl ""
2804.Dl "ipfw add allow ip from ${goodguys} to any"
2805.Dl "ipfw add deny ip from ${badguys} to any"
2806.Dl "... normal policies ..."
2807.Pp
2808The
2809.Cm verrevpath
2810option could be used to do automated anti-spoofing by adding the
2811following to the top of a ruleset:
2812.Pp
2813.Dl "ipfw add deny ip from any to any not verrevpath in"
2814.Pp
2815This rule drops all incoming packets that appear to be coming to the
2816system on the wrong interface.
2817For example, a packet with a source
2818address belonging to a host on a protected internal network would be
2819dropped if it tried to enter the system from an external interface.
2820.Pp
2821The
2822.Cm antispoof
2823option could be used to do similar but more restricted anti-spoofing
2824by adding the following to the top of a ruleset:
2825.Pp
2826.Dl "ipfw add deny ip from any to any not antispoof in"
2827.Pp
2828This rule drops all incoming packets that appear to be coming from another
2829directly connected system but on the wrong interface.
2830For example, a packet with a source address of
2831.Li 192.168.0.0/24 ,
2832configured on
2833.Li fxp0 ,
2834but coming in on
2835.Li fxp1
2836would be dropped.
2837.Ss DYNAMIC RULES
2838In order to protect a site from flood attacks involving fake
2839TCP packets, it is safer to use dynamic rules:
2840.Pp
2841.Dl "ipfw add check-state"
2842.Dl "ipfw add deny tcp from any to any established"
2843.Dl "ipfw add allow tcp from my-net to any setup keep-state"
2844.Pp
2845This will let the firewall install dynamic rules only for
2846those connection which start with a regular SYN packet coming
2847from the inside of our network.
2848Dynamic rules are checked when encountering the first
2849.Cm check-state
2850or
2851.Cm keep-state
2852rule.
2853A
2854.Cm check-state
2855rule should usually be placed near the beginning of the
2856ruleset to minimize the amount of work scanning the ruleset.
2857Your mileage may vary.
2858.Pp
2859To limit the number of connections a user can open
2860you can use the following type of rules:
2861.Pp
2862.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
2863.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
2864.Pp
2865The former (assuming it runs on a gateway) will allow each host
2866on a /24 network to open at most 10 TCP connections.
2867The latter can be placed on a server to make sure that a single
2868client does not use more than 4 simultaneous connections.
2869.Pp
2870.Em BEWARE :
2871stateful rules can be subject to denial-of-service attacks
2872by a SYN-flood which opens a huge number of dynamic rules.
2873The effects of such attacks can be partially limited by
2874acting on a set of
2875.Xr sysctl 8
2876variables which control the operation of the firewall.
2877.Pp
2878Here is a good usage of the
2879.Cm list
2880command to see accounting records and timestamp information:
2881.Pp
2882.Dl ipfw -at list
2883.Pp
2884or in short form without timestamps:
2885.Pp
2886.Dl ipfw -a list
2887.Pp
2888which is equivalent to:
2889.Pp
2890.Dl ipfw show
2891.Pp
2892Next rule diverts all incoming packets from 192.168.2.0/24
2893to divert port 5000:
2894.Pp
2895.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
2896.Pp
2897.Ss TRAFFIC SHAPING
2898The following rules show some of the applications of
2899.Nm
2900and
2901.Nm dummynet
2902for simulations and the like.
2903.Pp
2904This rule drops random incoming packets with a probability
2905of 5%:
2906.Pp
2907.Dl "ipfw add prob 0.05 deny ip from any to any in"
2908.Pp
2909A similar effect can be achieved making use of
2910.Nm dummynet
2911pipes:
2912.Pp
2913.Dl "ipfw add pipe 10 ip from any to any"
2914.Dl "ipfw pipe 10 config plr 0.05"
2915.Pp
2916We can use pipes to artificially limit bandwidth, e.g.\& on a
2917machine acting as a router, if we want to limit traffic from
2918local clients on 192.168.2.0/24 we do:
2919.Pp
2920.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2921.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
2922.Pp
2923note that we use the
2924.Cm out
2925modifier so that the rule is not used twice.
2926Remember in fact that
2927.Nm
2928rules are checked both on incoming and outgoing packets.
2929.Pp
2930Should we want to simulate a bidirectional link with bandwidth
2931limitations, the correct way is the following:
2932.Pp
2933.Dl "ipfw add pipe 1 ip from any to any out"
2934.Dl "ipfw add pipe 2 ip from any to any in"
2935.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
2936.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
2937.Pp
2938The above can be very useful, e.g.\& if you want to see how
2939your fancy Web page will look for a residential user who
2940is connected only through a slow link.
2941You should not use only one pipe for both directions, unless
2942you want to simulate a half-duplex medium (e.g.\& AppleTalk,
2943Ethernet, IRDA).
2944It is not necessary that both pipes have the same configuration,
2945so we can also simulate asymmetric links.
2946.Pp
2947Should we want to verify network performance with the RED queue
2948management algorithm:
2949.Pp
2950.Dl "ipfw add pipe 1 ip from any to any"
2951.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
2952.Pp
2953Another typical application of the traffic shaper is to
2954introduce some delay in the communication.
2955This can significantly affect applications which do a lot of Remote
2956Procedure Calls, and where the round-trip-time of the
2957connection often becomes a limiting factor much more than
2958bandwidth:
2959.Pp
2960.Dl "ipfw add pipe 1 ip from any to any out"
2961.Dl "ipfw add pipe 2 ip from any to any in"
2962.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
2963.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
2964.Pp
2965Per-flow queueing can be useful for a variety of purposes.
2966A very simple one is counting traffic:
2967.Pp
2968.Dl "ipfw add pipe 1 tcp from any to any"
2969.Dl "ipfw add pipe 1 udp from any to any"
2970.Dl "ipfw add pipe 1 ip from any to any"
2971.Dl "ipfw pipe 1 config mask all"
2972.Pp
2973The above set of rules will create queues (and collect
2974statistics) for all traffic.
2975Because the pipes have no limitations, the only effect is
2976collecting statistics.
2977Note that we need 3 rules, not just the last one, because
2978when
2979.Nm
2980tries to match IP packets it will not consider ports, so we
2981would not see connections on separate ports as different
2982ones.
2983.Pp
2984A more sophisticated example is limiting the outbound traffic
2985on a net with per-host limits, rather than per-network limits:
2986.Pp
2987.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2988.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
2989.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2990.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2991.Ss LOOKUP TABLES
2992In the following example, we need to create several traffic bandwidth
2993classes and we need different hosts/networks to fall into different classes.
2994We create one pipe for each class and configure them accordingly.
2995Then we create a single table and fill it with IP subnets and addresses.
2996For each subnet/host we set the argument equal to the number of the pipe
2997that it should use.
2998Then we classify traffic using a single rule:
2999.Pp
3000.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
3001.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
3002.Dl "..."
3003.Dl "ipfw table 1 add 192.168.2.0/24 1"
3004.Dl "ipfw table 1 add 192.168.0.0/27 4"
3005.Dl "ipfw table 1 add 192.168.0.2 1"
3006.Dl "..."
3007.Dl "ipfw add pipe tablearg ip from table(1) to any"
3008.Pp
3009Using the
3010.Cm fwd
3011action, the table entries may include hostnames and IP addresses.
3012.Pp
3013.Dl "ipfw table 1 add 192.168.2.0/24 10.23.2.1"
3014.Dl "ipfw table 1 add 192.168.0.0/27 router1.dmz"
3015.Dl "..."
3016.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
3017.Ss SETS OF RULES
3018To add a set of rules atomically, e.g.\& set 18:
3019.Pp
3020.Dl "ipfw set disable 18"
3021.Dl "ipfw add NN set 18 ...         # repeat as needed"
3022.Dl "ipfw set enable 18"
3023.Pp
3024To delete a set of rules atomically the command is simply:
3025.Pp
3026.Dl "ipfw delete set 18"
3027.Pp
3028To test a ruleset and disable it and regain control if something goes wrong:
3029.Pp
3030.Dl "ipfw set disable 18"
3031.Dl "ipfw add NN set 18 ...         # repeat as needed"
3032.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
3033.Pp
3034Here if everything goes well, you press control-C before the "sleep"
3035terminates, and your ruleset will be left active.
3036Otherwise, e.g.\& if
3037you cannot access your box, the ruleset will be disabled after
3038the sleep terminates thus restoring the previous situation.
3039.Pp
3040To show rules of the specific set:
3041.Pp
3042.Dl "ipfw set 18 show"
3043.Pp
3044To show rules of the disabled set:
3045.Pp
3046.Dl "ipfw -S set 18 show"
3047.Pp
3048To clear a specific rule counters of the specific set:
3049.Pp
3050.Dl "ipfw set 18 zero NN"
3051.Pp
3052To delete a specific rule of the specific set:
3053.Pp
3054.Dl "ipfw set 18 delete NN"
3055.Ss NAT, REDIRECT AND LSNAT
3056First redirect all the traffic to nat instance 123:
3057.Pp
3058.Dl "ipfw add nat 123 all from any to any"
3059.Pp
3060Then to configure nat instance 123 to alias all the outgoing traffic with ip
3061192.168.0.123, blocking all incoming connections, trying to keep
3062same ports on both sides, clearing aliasing table on address change
3063and keeping a log of traffic/link statistics:
3064.Pp
3065.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
3066.Pp
3067Or to change address of instance 123, aliasing table will be cleared (see
3068reset option):
3069.Pp
3070.Dl "ipfw nat 123 config ip 10.0.0.1"
3071.Pp
3072To see configuration of nat instance 123:
3073.Pp
3074.Dl "ipfw nat 123 show config"
3075.Pp
3076To show logs of all the instances in range 111-999:
3077.Pp
3078.Dl "ipfw nat 111-999 show"
3079.Pp
3080To see configurations of all instances:
3081.Pp
3082.Dl "ipfw nat show config"
3083.Pp
3084Or a redirect rule with mixed modes could looks like:
3085.Pp
3086.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
3087.Dl "			 redirect_port tcp 192.168.0.1:80 500"
3088.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
3089.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
3090.Dl "			 	    10.0.0.100	# LSNAT"
3091.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
3092.Dl "			 	    500		# LSNAT"
3093.Pp
3094or it could be split in:
3095.Pp
3096.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
3097.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
3098.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
3099.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
3100.Dl "				         10.0.0.100"
3101.Dl "ipfw nat 5 config redirect_port tcp"
3102.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
3103.Pp
3104.Sh SEE ALSO
3105.Xr cpp 1 ,
3106.Xr m4 1 ,
3107.Xr altq 4 ,
3108.Xr divert 4 ,
3109.Xr dummynet 4 ,
3110.Xr if_bridge 4 ,
3111.Xr ip 4 ,
3112.Xr ipfirewall 4 ,
3113.Xr ng_ipfw 4 ,
3114.Xr protocols 5 ,
3115.Xr services 5 ,
3116.Xr init 8 ,
3117.Xr kldload 8 ,
3118.Xr reboot 8 ,
3119.Xr sysctl 8 ,
3120.Xr syslogd 8
3121.Sh HISTORY
3122The
3123.Nm
3124utility first appeared in
3125.Fx 2.0 .
3126.Nm dummynet
3127was introduced in
3128.Fx 2.2.8 .
3129Stateful extensions were introduced in
3130.Fx 4.0 .
3131.Nm ipfw2
3132was introduced in Summer 2002.
3133.Sh AUTHORS
3134.An Ugen J. S. Antsilevich ,
3135.An Poul-Henning Kamp ,
3136.An Alex Nash ,
3137.An Archie Cobbs ,
3138.An Luigi Rizzo .
3139.Pp
3140.An -nosplit
3141API based upon code written by
3142.An Daniel Boulet
3143for BSDI.
3144.Pp
3145Dummynet has been introduced by Luigi Rizzo in 1997-1998.
3146.Pp
3147Some early work (1999-2000) on the
3148.Nm dummynet
3149traffic shaper supported by Akamba Corp.
3150.Pp
3151The ipfw core (ipfw2) has been completely redesigned and
3152reimplemented by Luigi Rizzo in summer 2002. Further
3153actions and
3154options have been added by various developer over the years.
3155.Pp
3156.An -nosplit
3157In-kernel NAT support written by
3158.An Paolo Pisati Aq piso@FreeBSD.org
3159as part of a Summer of Code 2005 project.
3160.Pp
3161SCTP
3162.Nm nat
3163support has been developed by
3164.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
3165The primary developers and maintainers are David Hayes and Jason But.
3166For further information visit:
3167.Aq http://www.caia.swin.edu.au/urp/SONATA
3168.Pp
3169Delay profiles have been developed by Alessandro Cerri and
3170Luigi Rizzo, supported by the
3171European Commission within Projects Onelab and Onelab2.
3172.Sh BUGS
3173The syntax has grown over the years and sometimes it might be confusing.
3174Unfortunately, backward compatibility prevents cleaning up mistakes
3175made in the definition of the syntax.
3176.Pp
3177.Em !!! WARNING !!!
3178.Pp
3179Misconfiguring the firewall can put your computer in an unusable state,
3180possibly shutting down network services and requiring console access to
3181regain control of it.
3182.Pp
3183Incoming packet fragments diverted by
3184.Cm divert
3185are reassembled before delivery to the socket.
3186The action used on those packet is the one from the
3187rule which matches the first fragment of the packet.
3188.Pp
3189Packets diverted to userland, and then reinserted by a userland process
3190may lose various packet attributes.
3191The packet source interface name
3192will be preserved if it is shorter than 8 bytes and the userland process
3193saves and reuses the sockaddr_in
3194(as does
3195.Xr natd 8 ) ;
3196otherwise, it may be lost.
3197If a packet is reinserted in this manner, later rules may be incorrectly
3198applied, making the order of
3199.Cm divert
3200rules in the rule sequence very important.
3201.Pp
3202Dummynet drops all packets with IPv6 link-local addresses.
3203.Pp
3204Rules using
3205.Cm uid
3206or
3207.Cm gid
3208may not behave as expected.
3209In particular, incoming SYN packets may
3210have no uid or gid associated with them since they do not yet belong
3211to a TCP connection, and the uid/gid associated with a packet may not
3212be as expected if the associated process calls
3213.Xr setuid 2
3214or similar system calls.
3215.Pp
3216Rule syntax is subject to the command line environment and some patterns
3217may need to be escaped with the backslash character
3218or quoted appropriately.
3219.Pp
3220Due to the architecture of
3221.Xr libalias 3 ,
3222ipfw nat is not compatible with the TCP segmentation offloading (TSO).
3223Thus, to reliably nat your network traffic, please disable TSO
3224on your NICs using
3225.Xr ifconfig 8 .
3226.Pp
3227ICMP error messages are not implicitly matched by dynamic rules
3228for the respective conversations.
3229To avoid failures of network error detection and path MTU discovery,
3230ICMP error messages may need to be allowed explicitly through static
3231rules.
3232