xref: /freebsd/sbin/ipfw/ipfw.8 (revision d8b878873e7aa8df1972cc6a642804b17eb61087)
1.\"
2.\" $FreeBSD$
3.\"
4.Dd March 20, 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
187.Pp
188Rules can be added with the
189.Cm add
190command; deleted individually or in groups with the
191.Cm delete
192command, and globally (except those in set 31) with the
193.Cm flush
194command; displayed, optionally with the content of the
195counters, using the
196.Cm show
197and
198.Cm list
199commands.
200Finally, counters can be reset with the
201.Cm zero
202and
203.Cm resetlog
204commands.
205.Pp
206.Ss COMMAND OPTIONS
207The following general options are available when invoking
208.Nm :
209.Bl -tag -width indent
210.It Fl a
211Show counter values when listing rules.
212The
213.Cm show
214command implies this option.
215.It Fl b
216Only show the action and the comment, not the body of a rule.
217Implies
218.Fl c .
219.It Fl c
220When entering or showing rules, print them in compact form,
221i.e., omitting the "ip from any to any" string
222when this does not carry any additional information.
223.It Fl d
224When listing, show dynamic rules in addition to static ones.
225.It Fl e
226When listing and
227.Fl d
228is specified, also show expired dynamic rules.
229.It Fl f
230Do not ask for confirmation for commands that can cause problems
231if misused,
232.No i.e. Cm flush .
233If there is no tty associated with the process, this is implied.
234.It Fl i
235When listing a table (see the
236.Sx LOOKUP TABLES
237section below for more information on lookup tables), format values
238as IP addresses. By default, values are shown as integers.
239.It Fl n
240Only check syntax of the command strings, without actually passing
241them to the kernel.
242.It Fl N
243Try to resolve addresses and service names in output.
244.It Fl q
245Be quiet when executing the
246.Cm add ,
247.Cm nat ,
248.Cm zero ,
249.Cm resetlog
250or
251.Cm flush
252commands;
253(implies
254.Fl f ) .
255This is useful when updating rulesets by executing multiple
256.Nm
257commands in a script
258(e.g.,
259.Ql sh\ /etc/rc.firewall ) ,
260or by processing a file with many
261.Nm
262rules across a remote login session.
263It also stops a table add or delete
264from failing if the entry already exists or is not present.
265.Pp
266The reason why this option may be important is that
267for some of these actions,
268.Nm
269may print a message; if the action results in blocking the
270traffic to the remote client,
271the remote login session will be closed
272and the rest of the ruleset will not be processed.
273Access to the console would then be required to recover.
274.It Fl S
275When listing rules, show the
276.Em set
277each rule belongs to.
278If this flag is not specified, disabled rules will not be
279listed.
280.It Fl s Op Ar field
281When listing pipes, sort according to one of the four
282counters (total or current packets or bytes).
283.It Fl t
284When listing, show last match timestamp converted with ctime().
285.It Fl T
286When listing, show last match timestamp as seconds from the epoch.
287This form can be more convenient for postprocessing by scripts.
288.El
289.Pp
290.Ss LIST OF RULES AND PREPROCESSING
291To ease configuration, rules can be put into a file which is
292processed using
293.Nm
294as shown in the last synopsis line.
295An absolute
296.Ar pathname
297must be used.
298The file will be read line by line and applied as arguments to the
299.Nm
300utility.
301.Pp
302Optionally, a preprocessor can be specified using
303.Fl p Ar preproc
304where
305.Ar pathname
306is to be piped through.
307Useful preprocessors include
308.Xr cpp 1
309and
310.Xr m4 1 .
311If
312.Ar preproc
313does not start with a slash
314.Pq Ql /
315as its first character, the usual
316.Ev PATH
317name search is performed.
318Care should be taken with this in environments where not all
319file systems are mounted (yet) by the time
320.Nm
321is being run (e.g.\& when they are mounted over NFS).
322Once
323.Fl p
324has been specified, any additional arguments are passed on to the preprocessor
325for interpretation.
326This allows for flexible configuration files (like conditionalizing
327them on the local hostname) and the use of macros to centralize
328frequently required arguments like IP addresses.
329.Pp
330.Ss TRAFFIC SHAPER CONFIGURATION
331The
332.Nm
333.Cm pipe , queue
334and
335.Cm sched
336commands are used to configure the traffic shaper and packet scheduler.
337See the
338.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
339Section below for details.
340.Pp
341If the world and the kernel get out of sync the
342.Nm
343ABI may break, preventing you from being able to add any rules.
344This can
345adversely effect the booting process.
346You can use
347.Nm
348.Cm disable
349.Cm firewall
350to temporarily disable the firewall to regain access to the network,
351allowing you to fix the problem.
352.Sh PACKET FLOW
353A packet is checked against the active ruleset in multiple places
354in the protocol stack, under control of several sysctl variables.
355These places and variables are shown below, and it is important to
356have this picture in mind in order to design a correct ruleset.
357.Bd -literal -offset indent
358       ^    to upper layers    V
359       |                       |
360       +----------->-----------+
361       ^                       V
362 [ip(6)_input]           [ip(6)_output]     net.inet(6).ip(6).fw.enable=1
363       |                       |
364       ^                       V
365 [ether_demux]        [ether_output_frame]  net.link.ether.ipfw=1
366       |                       |
367       +-->--[bdg_forward]-->--+            net.link.bridge.ipfw=1
368       ^                       V
369       |      to devices       |
370.Ed
371.Pp
372The number of
373times the same packet goes through the firewall can
374vary between 0 and 4 depending on packet source and
375destination, and system configuration.
376.Pp
377Note that as packets flow through the stack, headers can be
378stripped or added to it, and so they may or may not be available
379for inspection.
380E.g., incoming packets will include the MAC header when
381.Nm
382is invoked from
383.Cm ether_demux() ,
384but the same packets will have the MAC header stripped off when
385.Nm
386is invoked from
387.Cm ip_input()
388or
389.Cm ip6_input() .
390.Pp
391Also note that each packet is always checked against the complete ruleset,
392irrespective of the place where the check occurs, or the source of the packet.
393If a rule contains some match patterns or actions which are not valid
394for the place of invocation (e.g.\& trying to match a MAC header within
395.Cm ip_input
396or
397.Cm ip6_input ),
398the match pattern will not match, but a
399.Cm not
400operator in front of such patterns
401.Em will
402cause the pattern to
403.Em always
404match on those packets.
405It is thus the responsibility of
406the programmer, if necessary, to write a suitable ruleset to
407differentiate among the possible places.
408.Cm skipto
409rules can be useful here, as an example:
410.Bd -literal -offset indent
411# packets from ether_demux or bdg_forward
412ipfw add 10 skipto 1000 all from any to any layer2 in
413# packets from ip_input
414ipfw add 10 skipto 2000 all from any to any not layer2 in
415# packets from ip_output
416ipfw add 10 skipto 3000 all from any to any not layer2 out
417# packets from ether_output_frame
418ipfw add 10 skipto 4000 all from any to any layer2 out
419.Ed
420.Pp
421(yes, at the moment there is no way to differentiate between
422ether_demux and bdg_forward).
423.Sh SYNTAX
424In general, each keyword or argument must be provided as
425a separate command line argument, with no leading or trailing
426spaces.
427Keywords are case-sensitive, whereas arguments may
428or may not be case-sensitive depending on their nature
429(e.g.\& uid's are, hostnames are not).
430.Pp
431Some arguments (e.g. port or address lists) are comma-separated
432lists of values.
433In this case, spaces after commas ',' are allowed to make
434the line more readable.
435You can also put the entire
436command (including flags) into a single argument.
437E.g., the following forms are equivalent:
438.Bd -literal -offset indent
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
441ipfw "-q add deny src-ip 10.0.0.0/24, 127.0.0.1/8"
442.Ed
443.Sh RULE FORMAT
444The format of firewall rules is the following:
445.Bd -ragged -offset indent
446.Bk -words
447.Op Ar rule_number
448.Op Cm set Ar set_number
449.Op Cm prob Ar match_probability
450.Ar action
451.Op Cm log Op Cm logamount Ar number
452.Op Cm altq Ar queue
453.Oo
454.Bro Cm tag | untag
455.Brc Ar number
456.Oc
457.Ar body
458.Ek
459.Ed
460.Pp
461where the body of the rule specifies which information is used
462for filtering packets, among the following:
463.Pp
464.Bl -tag -width "Source and dest. addresses and ports" -offset XXX -compact
465.It Layer-2 header fields
466When available
467.It IPv4 and IPv6 Protocol
468TCP, UDP, ICMP, etc.
469.It Source and dest. addresses and ports
470.It Direction
471See Section
472.Sx PACKET FLOW
473.It Transmit and receive interface
474By name or address
475.It Misc. IP header fields
476Version, type of service, datagram length, identification,
477fragment flag (non-zero IP offset),
478Time To Live
479.It IP options
480.It IPv6 Extension headers
481Fragmentation, Hop-by-Hop options,
482Routing Headers, Source routing rthdr0, Mobile IPv6 rthdr2, IPSec options.
483.It IPv6 Flow-ID
484.It Misc. TCP header fields
485TCP flags (SYN, FIN, ACK, RST, etc.),
486sequence number, acknowledgment number,
487window
488.It TCP options
489.It ICMP types
490for ICMP packets
491.It ICMP6 types
492for ICMP6 packets
493.It User/group ID
494When the packet can be associated with a local socket.
495.It Divert status
496Whether a packet came from a divert socket (e.g.,
497.Xr natd 8 ) .
498.It Fib annotation state
499Whether a packet has been tagged for using a specific FIB (routing table)
500in future forwarding decisions.
501.El
502.Pp
503Note that some of the above information, e.g.\& source MAC or IP addresses and
504TCP/UDP ports, can be easily spoofed, so filtering on those fields
505alone might not guarantee the desired results.
506.Bl -tag -width indent
507.It Ar rule_number
508Each rule is associated with a
509.Ar rule_number
510in the range 1..65535, with the latter reserved for the
511.Em default
512rule.
513Rules are checked sequentially by rule number.
514Multiple rules can have the same number, in which case they are
515checked (and listed) according to the order in which they have
516been added.
517If a rule is entered without specifying a number, the kernel will
518assign one in such a way that the rule becomes the last one
519before the
520.Em default
521rule.
522Automatic rule numbers are assigned by incrementing the last
523non-default rule number by the value of the sysctl variable
524.Ar net.inet.ip.fw.autoinc_step
525which defaults to 100.
526If this is not possible (e.g.\& because we would go beyond the
527maximum allowed rule number), the number of the last
528non-default value is used instead.
529.It Cm set Ar set_number
530Each rule is associated with a
531.Ar set_number
532in the range 0..31.
533Sets can be individually disabled and enabled, so this parameter
534is of fundamental importance for atomic ruleset manipulation.
535It can be also used to simplify deletion of groups of rules.
536If a rule is entered without specifying a set number,
537set 0 will be used.
538.br
539Set 31 is special in that it cannot be disabled,
540and rules in set 31 are not deleted by the
541.Nm ipfw flush
542command (but you can delete them with the
543.Nm ipfw delete set 31
544command).
545Set 31 is also used for the
546.Em default
547rule.
548.It Cm prob Ar match_probability
549A match is only declared with the specified probability
550(floating point number between 0 and 1).
551This can be useful for a number of applications such as
552random packet drop or
553(in conjunction with
554.Nm dummynet )
555to simulate the effect of multiple paths leading to out-of-order
556packet delivery.
557.Pp
558Note: this condition is checked before any other condition, including
559ones such as keep-state or check-state which might have side effects.
560.It Cm log Op Cm logamount Ar number
561When a packet matches a rule with the
562.Cm log
563keyword, a message will be
564logged to
565.Xr syslogd 8
566with a
567.Dv LOG_SECURITY
568facility.
569The logging only occurs if the sysctl variable
570.Va net.inet.ip.fw.verbose
571is set to 1
572(which is the default when the kernel is compiled with
573.Dv IPFIREWALL_VERBOSE )
574and the number of packets logged so far for that
575particular rule does not exceed the
576.Cm logamount
577parameter.
578If no
579.Cm logamount
580is specified, the limit is taken from the sysctl variable
581.Va net.inet.ip.fw.verbose_limit .
582In both cases, a value of 0 removes the logging limit.
583.Pp
584Once the limit is reached, logging can be re-enabled by
585clearing the logging counter or the packet counter for that entry, see the
586.Cm resetlog
587command.
588.Pp
589Note: logging is done after all other packet matching conditions
590have been successfully verified, and before performing the final
591action (accept, deny, etc.) on the packet.
592.It Cm tag Ar number
593When a packet matches a rule with the
594.Cm tag
595keyword, the numeric tag for the given
596.Ar number
597in the range 1..65534 will be attached to the packet.
598The tag acts as an internal marker (it is not sent out over
599the wire) that can be used to identify these packets later on.
600This can be used, for example, to provide trust between interfaces
601and to start doing policy-based filtering.
602A packet can have multiple tags at the same time.
603Tags are "sticky", meaning once a tag is applied to a packet by a
604matching rule it exists until explicit removal.
605Tags are kept with the packet everywhere within the kernel, but are
606lost when packet leaves the kernel, for example, on transmitting
607packet out to the network or sending packet to a
608.Xr divert 4
609socket.
610.Pp
611To check for previously applied tags, use the
612.Cm tagged
613rule option.
614To delete previously applied tag, use the
615.Cm untag
616keyword.
617.Pp
618Note: since tags are kept with the packet everywhere in kernelspace,
619they can be set and unset anywhere in the kernel network subsystem
620(using the
621.Xr mbuf_tags 9
622facility), not only by means of the
623.Xr ipfw 4
624.Cm tag
625and
626.Cm untag
627keywords.
628For example, there can be a specialized
629.Xr netgraph 4
630node doing traffic analyzing and tagging for later inspecting
631in firewall.
632.It Cm untag Ar number
633When a packet matches a rule with the
634.Cm untag
635keyword, the tag with the number
636.Ar number
637is searched among the tags attached to this packet and,
638if found, removed from it.
639Other tags bound to packet, if present, are left untouched.
640.It Cm altq Ar queue
641When a packet matches a rule with the
642.Cm altq
643keyword, the ALTQ identifier for the given
644.Ar queue
645(see
646.Xr altq 4 )
647will be attached.
648Note that this ALTQ tag is only meaningful for packets going "out" of IPFW,
649and not being rejected or going to divert sockets.
650Note that if there is insufficient memory at the time the packet is
651processed, it will not be tagged, so it is wise to make your ALTQ
652"default" queue policy account for this.
653If multiple
654.Cm altq
655rules match a single packet, only the first one adds the ALTQ classification
656tag.
657In doing so, traffic may be shaped by using
658.Cm count Cm altq Ar queue
659rules for classification early in the ruleset, then later applying
660the filtering decision.
661For example,
662.Cm check-state
663and
664.Cm keep-state
665rules may come later and provide the actual filtering decisions in
666addition to the fallback ALTQ tag.
667.Pp
668You must run
669.Xr pfctl 8
670to set up the queues before IPFW will be able to look them up by name,
671and if the ALTQ disciplines are rearranged, the rules in containing the
672queue identifiers in the kernel will likely have gone stale and need
673to be reloaded.
674Stale queue identifiers will probably result in misclassification.
675.Pp
676All system ALTQ processing can be turned on or off via
677.Nm
678.Cm enable Ar altq
679and
680.Nm
681.Cm disable Ar altq .
682The usage of
683.Va net.inet.ip.fw.one_pass
684is irrelevant to ALTQ traffic shaping, as the actual rule action is followed
685always after adding an ALTQ tag.
686.El
687.Ss RULE ACTIONS
688A rule can be associated with one of the following actions, which
689will be executed when the packet matches the body of the rule.
690.Bl -tag -width indent
691.It Cm allow | accept | pass | permit
692Allow packets that match rule.
693The search terminates.
694.It Cm check-state
695Checks the packet against the dynamic ruleset.
696If a match is found, execute the action associated with
697the rule which generated this dynamic rule, otherwise
698move to the next rule.
699.br
700.Cm Check-state
701rules do not have a body.
702If no
703.Cm check-state
704rule is found, the dynamic ruleset is checked at the first
705.Cm keep-state
706or
707.Cm limit
708rule.
709.It Cm count
710Update counters for all packets that match rule.
711The search continues with the next rule.
712.It Cm deny | drop
713Discard packets that match this rule.
714The search terminates.
715.It Cm divert Ar port
716Divert packets that match this rule to the
717.Xr divert 4
718socket bound to port
719.Ar port .
720The search terminates.
721.It Cm fwd | forward Ar ipaddr | tablearg Ns Op , Ns Ar port
722Change the next-hop on matching packets to
723.Ar ipaddr ,
724which can be an IP address or a host name.
725The next hop can also be supplied by the last table
726looked up for the packet by using the
727.Cm tablearg
728keyword instead of an explicit address.
729The search terminates if this rule matches.
730.Pp
731If
732.Ar ipaddr
733is a local address, then matching packets will be forwarded to
734.Ar port
735(or the port number in the packet if one is not specified in the rule)
736on the local machine.
737.br
738If
739.Ar ipaddr
740is not a local address, then the port number
741(if specified) is ignored, and the packet will be
742forwarded to the remote address, using the route as found in
743the local routing table for that IP.
744.br
745A
746.Ar fwd
747rule will not match layer-2 packets (those received
748on ether_input, ether_output, or bridged).
749.br
750The
751.Cm fwd
752action does not change the contents of the packet at all.
753In particular, the destination address remains unmodified, so
754packets forwarded to another system will usually be rejected by that system
755unless there is a matching rule on that system to capture them.
756For packets forwarded locally,
757the local address of the socket will be
758set to the original destination address of the packet.
759This makes the
760.Xr netstat 1
761entry look rather weird but is intended for
762use with transparent proxy servers.
763.Pp
764To enable
765.Cm fwd
766a custom kernel needs to be compiled with the option
767.Cd "options IPFIREWALL_FORWARD" .
768.It Cm nat Ar nat_nr
769Pass packet to a
770nat instance
771(for network address translation, address redirect, etc.):
772see the
773.Sx NETWORK ADDRESS TRANSLATION (NAT)
774Section for further information.
775.It Cm pipe Ar pipe_nr
776Pass packet to a
777.Nm dummynet
778.Dq pipe
779(for bandwidth limitation, delay, etc.).
780See the
781.Sx TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
782Section for further information.
783The search terminates; however, on exit from the pipe and if
784the
785.Xr sysctl 8
786variable
787.Va net.inet.ip.fw.one_pass
788is not set, the packet is passed again to the firewall code
789starting from the next rule.
790.It Cm queue Ar queue_nr
791Pass packet to a
792.Nm dummynet
793.Dq queue
794(for bandwidth limitation using WF2Q+).
795.It Cm reject
796(Deprecated).
797Synonym for
798.Cm unreach host .
799.It Cm reset
800Discard packets that match this rule, and if the
801packet is a TCP packet, try to send a TCP reset (RST) notice.
802The search terminates.
803.It Cm reset6
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 skipto Ar number | tablearg
808Skip all subsequent rules numbered less than
809.Ar number .
810The search continues with the first rule numbered
811.Ar number
812or higher.
813It is possible to use the
814.Cm tablearg
815keyword with a skipto for a
816.Em computed
817skipto, but care should be used, as no destination caching
818is possible in this case so the rules are always walked to find it,
819starting from the
820.Cm skipto .
821.It Cm tee Ar port
822Send a copy of packets matching this rule to the
823.Xr divert 4
824socket bound to port
825.Ar port .
826The search continues with the next rule.
827.It Cm unreach Ar code
828Discard packets that match this rule, and try to send an ICMP
829unreachable notice with code
830.Ar code ,
831where
832.Ar code
833is a number from 0 to 255, or one of these aliases:
834.Cm net , host , protocol , port ,
835.Cm needfrag , srcfail , net-unknown , host-unknown ,
836.Cm isolated , net-prohib , host-prohib , tosnet ,
837.Cm toshost , filter-prohib , host-precedence
838or
839.Cm precedence-cutoff .
840The search terminates.
841.It Cm unreach6 Ar code
842Discard packets that match this rule, and try to send an ICMPv6
843unreachable notice with code
844.Ar code ,
845where
846.Ar code
847is a number from 0, 1, 3 or 4, or one of these aliases:
848.Cm no-route, admin-prohib, address
849or
850.Cm port .
851The search terminates.
852.It Cm netgraph Ar cookie
853Divert packet into netgraph with given
854.Ar cookie .
855The search terminates.
856If packet is later returned from netgraph it is either
857accepted or continues with the next rule, depending on
858.Va net.inet.ip.fw.one_pass
859sysctl variable.
860.It Cm ngtee Ar cookie
861A copy of packet is diverted into netgraph, original
862packet is either accepted or continues with the next rule, depending on
863.Va net.inet.ip.fw.one_pass
864sysctl variable.
865See
866.Xr ng_ipfw 4
867for more information on
868.Cm netgraph
869and
870.Cm ngtee
871actions.
872.It Cm setfib Ar fibnum
873The packet is tagged so as to use the FIB (routing table)
874.Ar fibnum
875in any subsequent forwarding decisions.
876Initially this is limited to the values 0 through 15, see
877.Xr setfib 1 .
878Processing continues at the next rule.
879.It Cm reass
880Queue and reassemble ip fragments.
881If the packet is not fragmented, counters are updated and processing continues with the next rule.
882If the packet is the last logical fragment, the packet is reassembled and, if
883.Va net.inet.ip.fw.one_pass
884is set to 0, processing continues with the next rule, else packet is allowed to pass and search terminates.
885If the packet is a fragment in the middle, it is consumed and processing stops immediately.
886.Pp
887Fragments handling can be tuned via
888.Va net.inet.ip.maxfragpackets
889and
890.Va net.inet.ip.maxfragsperpacket
891which limit, respectively, the maximum number of processable fragments (default: 800) and
892the maximum number of fragments per packet (default: 16).
893.Pp
894NOTA BENE: since fragments do not contain port numbers, they should be avoided with the
895.Nm reass
896rule.
897Alternatively, direction-based (like
898.Nm in
899/
900.Nm out
901) and source-based (like
902.Nm via
903) match patterns can be used to select fragments.
904.Pp
905Usually a simple rule like:
906.Bd -literal -offset indent
907# reassemble incoming fragments
908ipfw add reass all from any to any in
909.Ed
910.Pp
911is all you need at the beginning of your ruleset.
912.El
913.Ss RULE BODY
914The body of a rule contains zero or more patterns (such as
915specific source and destination addresses or ports,
916protocol options, incoming or outgoing interfaces, etc.)
917that the packet must match in order to be recognised.
918In general, the patterns are connected by (implicit)
919.Cm and
920operators -- i.e., all must match in order for the
921rule to match.
922Individual patterns can be prefixed by the
923.Cm not
924operator to reverse the result of the match, as in
925.Pp
926.Dl "ipfw add 100 allow ip from not 1.2.3.4 to any"
927.Pp
928Additionally, sets of alternative match patterns
929.Pq Em or-blocks
930can be constructed by putting the patterns in
931lists enclosed between parentheses ( ) or braces { }, and
932using the
933.Cm or
934operator as follows:
935.Pp
936.Dl "ipfw add 100 allow ip from { x or not y or z } to any"
937.Pp
938Only one level of parentheses is allowed.
939Beware that most shells have special meanings for parentheses
940or braces, so it is advisable to put a backslash \\ in front of them
941to prevent such interpretations.
942.Pp
943The body of a rule must in general include a source and destination
944address specifier.
945The keyword
946.Ar any
947can be used in various places to specify that the content of
948a required field is irrelevant.
949.Pp
950The rule body has the following format:
951.Bd -ragged -offset indent
952.Op Ar proto Cm from Ar src Cm to Ar dst
953.Op Ar options
954.Ed
955.Pp
956The first part (proto from src to dst) is for backward
957compatibility with earlier versions of
958.Fx .
959In modern
960.Fx
961any match pattern (including MAC headers, IP protocols,
962addresses and ports) can be specified in the
963.Ar options
964section.
965.Pp
966Rule fields have the following meaning:
967.Bl -tag -width indent
968.It Ar proto : protocol | Cm { Ar protocol Cm or ... }
969.It Ar protocol : Oo Cm not Oc Ar protocol-name | protocol-number
970An IP protocol specified by number or name
971(for a complete list see
972.Pa /etc/protocols ) ,
973or one of the following keywords:
974.Bl -tag -width indent
975.It Cm ip4 | ipv4
976Matches IPv4 packets.
977.It Cm ip6 | ipv6
978Matches IPv6 packets.
979.It Cm ip | all
980Matches any packet.
981.El
982.Pp
983The
984.Cm ipv6
985in
986.Cm proto
987option will be treated as inner protocol.
988And, the
989.Cm ipv4
990is not available in
991.Cm proto
992option.
993.Pp
994The
995.Cm { Ar protocol Cm or ... }
996format (an
997.Em or-block )
998is provided for convenience only but its use is deprecated.
999.It Ar src No and Ar dst : Bro Cm addr | Cm { Ar addr Cm or ... } Brc Op Oo Cm not Oc Ar ports
1000An address (or a list, see below)
1001optionally followed by
1002.Ar ports
1003specifiers.
1004.Pp
1005The second format
1006.Em ( or-block
1007with multiple addresses) is provided for convenience only and
1008its use is discouraged.
1009.It Ar addr : Oo Cm not Oc Bro
1010.Bl -tag -width indent
1011.Cm any | me | me6 |
1012.Cm table Ns Pq Ar number Ns Op , Ns Ar value
1013.Ar | addr-list | addr-set
1014.Brc
1015.It Cm any
1016matches any IP address.
1017.It Cm me
1018matches any IP address configured on an interface in the system.
1019.It Cm me6
1020matches any IPv6 address configured on an interface in the system.
1021The address list is evaluated at the time the packet is
1022analysed.
1023.It Cm table Ns Pq Ar number Ns Op , Ns Ar value
1024Matches any IPv4 address for which an entry exists in the lookup table
1025.Ar number .
1026If an optional 32-bit unsigned
1027.Ar value
1028is also specified, an entry will match only if it has this value.
1029See the
1030.Sx LOOKUP TABLES
1031section below for more information on lookup tables.
1032.El
1033.It Ar addr-list : ip-addr Ns Op Ns , Ns Ar addr-list
1034.It Ar ip-addr :
1035A host or subnet address specified in one of the following ways:
1036.Bl -tag -width indent
1037.It Ar numeric-ip | hostname
1038Matches a single IPv4 address, specified as dotted-quad or a hostname.
1039Hostnames are resolved at the time the rule is added to the firewall list.
1040.It Ar addr Ns / Ns Ar masklen
1041Matches all addresses with base
1042.Ar addr
1043(specified as an IP address, a network number, or a hostname)
1044and mask width of
1045.Cm masklen
1046bits.
1047As an example, 1.2.3.4/25 or 1.2.3.0/25 will match
1048all IP numbers from 1.2.3.0 to 1.2.3.127 .
1049.It Ar addr Ns : Ns Ar mask
1050Matches all addresses with base
1051.Ar addr
1052(specified as an IP address, a network number, or a hostname)
1053and the mask of
1054.Ar mask ,
1055specified as a dotted quad.
1056As an example, 1.2.3.4:255.0.255.0 or 1.0.3.0:255.0.255.0 will match
10571.*.3.*.
1058This form is advised only for non-contiguous
1059masks.
1060It is better to resort to the
1061.Ar addr Ns / Ns Ar masklen
1062format for contiguous masks, which is more compact and less
1063error-prone.
1064.El
1065.It Ar addr-set : addr Ns Oo Ns / Ns Ar masklen Oc Ns Cm { Ns Ar list Ns Cm }
1066.It Ar list : Bro Ar num | num-num Brc Ns Op Ns , Ns Ar list
1067Matches all addresses with base address
1068.Ar addr
1069(specified as an IP address, a network number, or a hostname)
1070and whose last byte is in the list between braces { } .
1071Note that there must be no spaces between braces and
1072numbers (spaces after commas are allowed).
1073Elements of the list can be specified as single entries
1074or ranges.
1075The
1076.Ar masklen
1077field is used to limit the size of the set of addresses,
1078and can have any value between 24 and 32.
1079If not specified,
1080it will be assumed as 24.
1081.br
1082This format is particularly useful to handle sparse address sets
1083within a single rule.
1084Because the matching occurs using a
1085bitmask, it takes constant time and dramatically reduces
1086the complexity of rulesets.
1087.br
1088As an example, an address specified as 1.2.3.4/24{128,35-55,89}
1089or 1.2.3.0/24{128,35-55,89}
1090will match the following IP addresses:
1091.br
10921.2.3.128, 1.2.3.35 to 1.2.3.55, 1.2.3.89 .
1093.It Ar addr6-list : ip6-addr Ns Op Ns , Ns Ar addr6-list
1094.It Ar ip6-addr :
1095A host or subnet specified one of the following ways:
1096.Pp
1097.Bl -tag -width indent
1098.It Ar numeric-ip | hostname
1099Matches a single IPv6 address as allowed by
1100.Xr inet_pton 3
1101or a hostname.
1102Hostnames are resolved at the time the rule is added to the firewall
1103list.
1104.It Ar addr Ns / Ns Ar masklen
1105Matches all IPv6 addresses with base
1106.Ar addr
1107(specified as allowed by
1108.Xr inet_pton
1109or a hostname)
1110and mask width of
1111.Cm masklen
1112bits.
1113.El
1114.Pp
1115No support for sets of IPv6 addresses is provided because IPv6 addresses
1116are typically random past the initial prefix.
1117.It Ar ports : Bro Ar port | port Ns \&- Ns Ar port Ns Brc Ns Op , Ns Ar ports
1118For protocols which support port numbers (such as TCP and UDP), optional
1119.Cm ports
1120may be specified as one or more ports or port ranges, separated
1121by commas but no spaces, and an optional
1122.Cm not
1123operator.
1124The
1125.Ql \&-
1126notation specifies a range of ports (including boundaries).
1127.Pp
1128Service names (from
1129.Pa /etc/services )
1130may be used instead of numeric port values.
1131The length of the port list is limited to 30 ports or ranges,
1132though one can specify larger ranges by using an
1133.Em or-block
1134in the
1135.Cm options
1136section of the rule.
1137.Pp
1138A backslash
1139.Pq Ql \e
1140can be used to escape the dash
1141.Pq Ql -
1142character in a service name (from a shell, the backslash must be
1143typed twice to avoid the shell itself interpreting it as an escape
1144character).
1145.Pp
1146.Dl "ipfw add count tcp from any ftp\e\e-data-ftp to any"
1147.Pp
1148Fragmented packets which have a non-zero offset (i.e., not the first
1149fragment) will never match a rule which has one or more port
1150specifications.
1151See the
1152.Cm frag
1153option for details on matching fragmented packets.
1154.El
1155.Ss RULE OPTIONS (MATCH PATTERNS)
1156Additional match patterns can be used within
1157rules.
1158Zero or more of these so-called
1159.Em options
1160can be present in a rule, optionally prefixed by the
1161.Cm not
1162operand, and possibly grouped into
1163.Em or-blocks .
1164.Pp
1165The following match patterns can be used (listed in alphabetical order):
1166.Bl -tag -width indent
1167.It Cm // this is a comment.
1168Inserts the specified text as a comment in the rule.
1169Everything following // is considered as a comment and stored in the rule.
1170You can have comment-only rules, which are listed as having a
1171.Cm count
1172action followed by the comment.
1173.It Cm bridged
1174Alias for
1175.Cm layer2 .
1176.It Cm diverted
1177Matches only packets generated by a divert socket.
1178.It Cm diverted-loopback
1179Matches only packets coming from a divert socket back into the IP stack
1180input for delivery.
1181.It Cm diverted-output
1182Matches only packets going from a divert socket back outward to the IP
1183stack output for delivery.
1184.It Cm dst-ip Ar ip-address
1185Matches IPv4 packets whose destination IP is one of the address(es)
1186specified as argument.
1187.It Bro Cm dst-ip6 | dst-ipv6 Brc Ar ip6-address
1188Matches IPv6 packets whose destination IP is one of the address(es)
1189specified as argument.
1190.It Cm dst-port Ar ports
1191Matches IP packets whose destination port is one of the port(s)
1192specified as argument.
1193.It Cm established
1194Matches TCP packets that have the RST or ACK bits set.
1195.It Cm ext6hdr Ar header
1196Matches IPv6 packets containing the extended header given by
1197.Ar header .
1198Supported headers are:
1199.Pp
1200Fragment,
1201.Pq Cm frag ,
1202Hop-to-hop options
1203.Pq Cm hopopt ,
1204any type of Routing Header
1205.Pq Cm route ,
1206Source routing Routing Header Type 0
1207.Pq Cm rthdr0 ,
1208Mobile IPv6 Routing Header Type 2
1209.Pq Cm rthdr2 ,
1210Destination options
1211.Pq Cm dstopt ,
1212IPSec authentication headers
1213.Pq Cm ah ,
1214and IPsec encapsulated security payload headers
1215.Pq Cm esp .
1216.It Cm fib Ar fibnum
1217Matches a packet that has been tagged to use
1218the given FIB (routing table) number.
1219.It Cm flow-id Ar labels
1220Matches IPv6 packets containing any of the flow labels given in
1221.Ar labels .
1222.Ar labels
1223is a comma separated list of numeric flow labels.
1224.It Cm frag
1225Matches packets that are fragments and not the first
1226fragment of an IP datagram.
1227Note that these packets will not have
1228the next protocol header (e.g.\& TCP, UDP) so options that look into
1229these headers cannot match.
1230.It Cm gid Ar group
1231Matches all TCP or UDP packets sent by or received for a
1232.Ar group .
1233A
1234.Ar group
1235may be specified by name or number.
1236.It Cm jail Ar prisonID
1237Matches all TCP or UDP packets sent by or received for the
1238jail whos prison ID is
1239.Ar prisonID .
1240.It Cm icmptypes Ar types
1241Matches ICMP packets whose ICMP type is in the list
1242.Ar types .
1243The list may be specified as any combination of
1244individual types (numeric) separated by commas.
1245.Em Ranges are not allowed .
1246The supported ICMP types are:
1247.Pp
1248echo reply
1249.Pq Cm 0 ,
1250destination unreachable
1251.Pq Cm 3 ,
1252source quench
1253.Pq Cm 4 ,
1254redirect
1255.Pq Cm 5 ,
1256echo request
1257.Pq Cm 8 ,
1258router advertisement
1259.Pq Cm 9 ,
1260router solicitation
1261.Pq Cm 10 ,
1262time-to-live exceeded
1263.Pq Cm 11 ,
1264IP header bad
1265.Pq Cm 12 ,
1266timestamp request
1267.Pq Cm 13 ,
1268timestamp reply
1269.Pq Cm 14 ,
1270information request
1271.Pq Cm 15 ,
1272information reply
1273.Pq Cm 16 ,
1274address mask request
1275.Pq Cm 17
1276and address mask reply
1277.Pq Cm 18 .
1278.It Cm icmp6types Ar types
1279Matches ICMP6 packets whose ICMP6 type is in the list of
1280.Ar types .
1281The list may be specified as any combination of
1282individual types (numeric) separated by commas.
1283.Em Ranges are not allowed .
1284.It Cm in | out
1285Matches incoming or outgoing packets, respectively.
1286.Cm in
1287and
1288.Cm out
1289are mutually exclusive (in fact,
1290.Cm out
1291is implemented as
1292.Cm not in Ns No ).
1293.It Cm ipid Ar id-list
1294Matches IPv4 packets whose
1295.Cm ip_id
1296field has value included in
1297.Ar id-list ,
1298which is either a single value or a list of values or ranges
1299specified in the same way as
1300.Ar ports .
1301.It Cm iplen Ar len-list
1302Matches IP packets whose total length, including header and data, is
1303in the set
1304.Ar len-list ,
1305which is either a single value or a list of values or ranges
1306specified in the same way as
1307.Ar ports .
1308.It Cm ipoptions Ar spec
1309Matches packets whose IPv4 header contains the comma separated list of
1310options specified in
1311.Ar spec .
1312The supported IP options are:
1313.Pp
1314.Cm ssrr
1315(strict source route),
1316.Cm lsrr
1317(loose source route),
1318.Cm rr
1319(record packet route) and
1320.Cm ts
1321(timestamp).
1322The absence of a particular option may be denoted
1323with a
1324.Ql \&! .
1325.It Cm ipprecedence Ar precedence
1326Matches IPv4 packets whose precedence field is equal to
1327.Ar precedence .
1328.It Cm ipsec
1329Matches packets that have IPSEC history associated with them
1330(i.e., the packet comes encapsulated in IPSEC, the kernel
1331has IPSEC support and IPSEC_FILTERTUNNEL option, and can correctly
1332decapsulate it).
1333.Pp
1334Note that specifying
1335.Cm ipsec
1336is different from specifying
1337.Cm proto Ar ipsec
1338as the latter will only look at the specific IP protocol field,
1339irrespective of IPSEC kernel support and the validity of the IPSEC data.
1340.Pp
1341Further note that this flag is silently ignored in kernels without
1342IPSEC support.
1343It does not affect rule processing when given and the
1344rules are handled as if with no
1345.Cm ipsec
1346flag.
1347.It Cm iptos Ar spec
1348Matches IPv4 packets whose
1349.Cm tos
1350field contains the comma separated list of
1351service types specified in
1352.Ar spec .
1353The supported IP types of service are:
1354.Pp
1355.Cm lowdelay
1356.Pq Dv IPTOS_LOWDELAY ,
1357.Cm throughput
1358.Pq Dv IPTOS_THROUGHPUT ,
1359.Cm reliability
1360.Pq Dv IPTOS_RELIABILITY ,
1361.Cm mincost
1362.Pq Dv IPTOS_MINCOST ,
1363.Cm congestion
1364.Pq Dv IPTOS_ECN_CE .
1365The absence of a particular type may be denoted
1366with a
1367.Ql \&! .
1368.It Cm ipttl Ar ttl-list
1369Matches IPv4 packets whose time to live is included in
1370.Ar ttl-list ,
1371which is either a single value or a list of values or ranges
1372specified in the same way as
1373.Ar ports .
1374.It Cm ipversion Ar ver
1375Matches IP packets whose IP version field is
1376.Ar ver .
1377.It Cm keep-state
1378Upon a match, the firewall will create a dynamic rule, whose
1379default behaviour is to match bidirectional traffic between
1380source and destination IP/port using the same protocol.
1381The rule has a limited lifetime (controlled by a set of
1382.Xr sysctl 8
1383variables), and the lifetime is refreshed every time a matching
1384packet is found.
1385.It Cm layer2
1386Matches only layer2 packets, i.e., those passed to
1387.Nm
1388from ether_demux() and ether_output_frame().
1389.It Cm limit Bro Cm src-addr | src-port | dst-addr | dst-port Brc Ar N
1390The firewall will only allow
1391.Ar N
1392connections with the same
1393set of parameters as specified in the rule.
1394One or more
1395of source and destination addresses and ports can be
1396specified.
1397Currently,
1398only IPv4 flows are supported.
1399.It Cm lookup Bro Cm dst-ip | dst-port | src-ip | src-port | uid | jail Brc Ar N
1400Search an entry in lookup table
1401.Ar N
1402that matches the field specified as argument.
1403If not found, the match fails.
1404Otherwise, the match succeeds and
1405.Cm tablearg
1406is set to the value extracted from the table.
1407.Pp
1408This option can be useful to quickly dispatch traffic based on
1409certain packet fields.
1410See the
1411.Sx LOOKUP TABLES
1412section below for more information on lookup tables.
1413.It Cm { MAC | mac } Ar dst-mac src-mac
1414Match packets with a given
1415.Ar dst-mac
1416and
1417.Ar src-mac
1418addresses, specified as the
1419.Cm any
1420keyword (matching any MAC address), or six groups of hex digits
1421separated by colons,
1422and optionally followed by a mask indicating the significant bits.
1423The mask may be specified using either of the following methods:
1424.Bl -enum -width indent
1425.It
1426A slash
1427.Pq /
1428followed by the number of significant bits.
1429For example, an address with 33 significant bits could be specified as:
1430.Pp
1431.Dl "MAC 10:20:30:40:50:60/33 any"
1432.Pp
1433.It
1434An ampersand
1435.Pq &
1436followed by a bitmask specified as six groups of hex digits separated
1437by colons.
1438For example, an address in which the last 16 bits are significant could
1439be specified as:
1440.Pp
1441.Dl "MAC 10:20:30:40:50:60&00:00:00:00:ff:ff any"
1442.Pp
1443Note that the ampersand character has a special meaning in many shells
1444and should generally be escaped.
1445.Pp
1446.El
1447Note that the order of MAC addresses (destination first,
1448source second) is
1449the same as on the wire, but the opposite of the one used for
1450IP addresses.
1451.It Cm mac-type Ar mac-type
1452Matches packets whose Ethernet Type field
1453corresponds to one of those specified as argument.
1454.Ar mac-type
1455is specified in the same way as
1456.Cm port numbers
1457(i.e., one or more comma-separated single values or ranges).
1458You can use symbolic names for known values such as
1459.Em vlan , ipv4, ipv6 .
1460Values can be entered as decimal or hexadecimal (if prefixed by 0x),
1461and they are always printed as hexadecimal (unless the
1462.Cm -N
1463option is used, in which case symbolic resolution will be attempted).
1464.It Cm proto Ar protocol
1465Matches packets with the corresponding IP protocol.
1466.It Cm recv | xmit | via Brq Ar ifX | Ar if Ns Cm * | Ar ipno | Ar any
1467Matches packets received, transmitted or going through,
1468respectively, the interface specified by exact name
1469.Ns No ( Ar ifX Ns No ),
1470by device name
1471.Ns No ( Ar if Ns Ar * Ns No ),
1472by IP address, or through some interface.
1473.Pp
1474The
1475.Cm via
1476keyword causes the interface to always be checked.
1477If
1478.Cm recv
1479or
1480.Cm xmit
1481is used instead of
1482.Cm via ,
1483then only the receive or transmit interface (respectively)
1484is checked.
1485By specifying both, it is possible to match packets based on
1486both receive and transmit interface, e.g.:
1487.Pp
1488.Dl "ipfw add deny ip from any to any out recv ed0 xmit ed1"
1489.Pp
1490The
1491.Cm recv
1492interface can be tested on either incoming or outgoing packets,
1493while the
1494.Cm xmit
1495interface can only be tested on outgoing packets.
1496So
1497.Cm out
1498is required (and
1499.Cm in
1500is invalid) whenever
1501.Cm xmit
1502is used.
1503.Pp
1504A packet might not have a receive or transmit interface: packets
1505originating from the local host have no receive interface,
1506while packets destined for the local host have no transmit
1507interface.
1508.It Cm setup
1509Matches TCP packets that have the SYN bit set but no ACK bit.
1510This is the short form of
1511.Dq Li tcpflags\ syn,!ack .
1512.It Cm src-ip Ar ip-address
1513Matches IPv4 packets whose source IP is one of the address(es)
1514specified as an argument.
1515.It Cm src-ip6 Ar ip6-address
1516Matches IPv6 packets whose source IP is one of the address(es)
1517specified as an argument.
1518.It Cm src-port Ar ports
1519Matches IP packets whose source port is one of the port(s)
1520specified as argument.
1521.It Cm tagged Ar tag-list
1522Matches packets whose tags are included in
1523.Ar tag-list ,
1524which is either a single value or a list of values or ranges
1525specified in the same way as
1526.Ar ports .
1527Tags can be applied to the packet using
1528.Cm tag
1529rule action parameter (see it's description for details on tags).
1530.It Cm tcpack Ar ack
1531TCP packets only.
1532Match if the TCP header acknowledgment number field is set to
1533.Ar ack .
1534.It Cm tcpdatalen Ar tcpdatalen-list
1535Matches TCP packets whose length of TCP data is
1536.Ar tcpdatalen-list ,
1537which is either a single value or a list of values or ranges
1538specified in the same way as
1539.Ar ports .
1540.It Cm tcpflags Ar spec
1541TCP packets only.
1542Match if the TCP header contains the comma separated list of
1543flags specified in
1544.Ar spec .
1545The supported TCP flags are:
1546.Pp
1547.Cm fin ,
1548.Cm syn ,
1549.Cm rst ,
1550.Cm psh ,
1551.Cm ack
1552and
1553.Cm urg .
1554The absence of a particular flag may be denoted
1555with a
1556.Ql \&! .
1557A rule which contains a
1558.Cm tcpflags
1559specification can never match a fragmented packet which has
1560a non-zero offset.
1561See the
1562.Cm frag
1563option for details on matching fragmented packets.
1564.It Cm tcpseq Ar seq
1565TCP packets only.
1566Match if the TCP header sequence number field is set to
1567.Ar seq .
1568.It Cm tcpwin Ar win
1569TCP packets only.
1570Match if the TCP header window field is set to
1571.Ar win .
1572.It Cm tcpoptions Ar spec
1573TCP packets only.
1574Match if the TCP header contains the comma separated list of
1575options specified in
1576.Ar spec .
1577The supported TCP options are:
1578.Pp
1579.Cm mss
1580(maximum segment size),
1581.Cm window
1582(tcp window advertisement),
1583.Cm sack
1584(selective ack),
1585.Cm ts
1586(rfc1323 timestamp) and
1587.Cm cc
1588(rfc1644 t/tcp connection count).
1589The absence of a particular option may be denoted
1590with a
1591.Ql \&! .
1592.It Cm uid Ar user
1593Match all TCP or UDP packets sent by or received for a
1594.Ar user .
1595A
1596.Ar user
1597may be matched by name or identification number.
1598.It Cm verrevpath
1599For incoming packets,
1600a routing table lookup is done on the packet's source address.
1601If the interface on which the packet entered the system matches the
1602outgoing interface for the route,
1603the packet matches.
1604If the interfaces do not match up,
1605the packet does not match.
1606All outgoing packets or packets with no incoming interface match.
1607.Pp
1608The name and functionality of the option is intentionally similar to
1609the Cisco IOS command:
1610.Pp
1611.Dl ip verify unicast reverse-path
1612.Pp
1613This option can be used to make anti-spoofing rules to reject all
1614packets with source addresses not from this interface.
1615See also the option
1616.Cm antispoof .
1617.It Cm versrcreach
1618For incoming packets,
1619a routing table lookup is done on the packet's source address.
1620If a route to the source address exists, but not the default route
1621or a blackhole/reject route, the packet matches.
1622Otherwise, the packet does not match.
1623All outgoing packets match.
1624.Pp
1625The name and functionality of the option is intentionally similar to
1626the Cisco IOS command:
1627.Pp
1628.Dl ip verify unicast source reachable-via any
1629.Pp
1630This option can be used to make anti-spoofing rules to reject all
1631packets whose source address is unreachable.
1632.It Cm antispoof
1633For incoming packets, the packet's source address is checked if it
1634belongs to a directly connected network.
1635If the network is directly connected, then the interface the packet
1636came on in is compared to the interface the network is connected to.
1637When incoming interface and directly connected interface are not the
1638same, the packet does not match.
1639Otherwise, the packet does match.
1640All outgoing packets match.
1641.Pp
1642This option can be used to make anti-spoofing rules to reject all
1643packets that pretend to be from a directly connected network but do
1644not come in through that interface.
1645This option is similar to but more restricted than
1646.Cm verrevpath
1647because it engages only on packets with source addresses of directly
1648connected networks instead of all source addresses.
1649.El
1650.Sh LOOKUP TABLES
1651Lookup tables are useful to handle large sparse sets of
1652addresses or other search keys (e.g. ports, jail IDs).
1653In the rest of this section we will use the term ``address''
1654to mean any unsigned value of up to 32-bit.
1655There may be up to 128 different lookup tables, numbered 0 to 127.
1656.Pp
1657Each entry is represented by an
1658.Ar addr Ns Op / Ns Ar masklen
1659and will match all addresses with base
1660.Ar addr
1661(specified as an IP address, a hostname or an unsigned integer)
1662and mask width of
1663.Ar masklen
1664bits.
1665If
1666.Ar masklen
1667is not specified, it defaults to 32.
1668When looking up an IP address in a table, the most specific
1669entry will match.
1670Associated with each entry is a 32-bit unsigned
1671.Ar value ,
1672which can optionally be checked by a rule matching code.
1673When adding an entry, if
1674.Ar value
1675is not specified, it defaults to 0.
1676.Pp
1677An entry can be added to a table
1678.Pq Cm add ,
1679or removed from a table
1680.Pq Cm delete .
1681A table can be examined
1682.Pq Cm list
1683or flushed
1684.Pq Cm flush .
1685.Pp
1686Internally, each table is stored in a Radix tree, the same way as
1687the routing table (see
1688.Xr route 4 ) .
1689.Pp
1690Lookup tables currently support only ports, jail IDs and IPv4 addresses.
1691.Pp
1692The
1693.Cm tablearg
1694feature provides the ability to use a value, looked up in the table, as
1695the argument for a rule action, action parameter or rule option.
1696This can significantly reduce number of rules in some configurations.
1697If two tables are used in a rule, the result of the second (destination)
1698is used.
1699The
1700.Cm tablearg
1701argument can be used with the following actions:
1702.Cm nat, pipe , queue, divert, tee, netgraph, ngtee, fwd, skipto
1703action parameters:
1704.Cm tag, untag,
1705rule options:
1706.Cm limit, tagged.
1707.Pp
1708When used with
1709.Cm fwd
1710it is possible to supply table entries with values
1711that are in the form of IP addresses or hostnames.
1712See the
1713.Sx EXAMPLES
1714Section for example usage of tables and the tablearg keyword.
1715.Pp
1716When used with the
1717.Cm skipto
1718action, the user should be aware that the code will walk the ruleset
1719up to a rule equal to, or past, the given number, and should therefore try keep the
1720ruleset compact between the skipto and the target rules.
1721.Sh SETS OF RULES
1722Each rule belongs to one of 32 different
1723.Em sets
1724, numbered 0 to 31.
1725Set 31 is reserved for the default rule.
1726.Pp
1727By default, rules are put in set 0, unless you use the
1728.Cm set N
1729attribute when entering a new rule.
1730Sets can be individually and atomically enabled or disabled,
1731so this mechanism permits an easy way to store multiple configurations
1732of the firewall and quickly (and atomically) switch between them.
1733The command to enable/disable sets is
1734.Bd -ragged -offset indent
1735.Nm
1736.Cm set Oo Cm disable Ar number ... Oc Op Cm enable Ar number ...
1737.Ed
1738.Pp
1739where multiple
1740.Cm enable
1741or
1742.Cm disable
1743sections can be specified.
1744Command execution is atomic on all the sets specified in the command.
1745By default, all sets are enabled.
1746.Pp
1747When you disable a set, its rules behave as if they do not exist
1748in the firewall configuration, with only one exception:
1749.Bd -ragged -offset indent
1750dynamic rules created from a rule before it had been disabled
1751will still be active until they expire.
1752In order to delete
1753dynamic rules you have to explicitly delete the parent rule
1754which generated them.
1755.Ed
1756.Pp
1757The set number of rules can be changed with the command
1758.Bd -ragged -offset indent
1759.Nm
1760.Cm set move
1761.Brq Cm rule Ar rule-number | old-set
1762.Cm to Ar new-set
1763.Ed
1764.Pp
1765Also, you can atomically swap two rulesets with the command
1766.Bd -ragged -offset indent
1767.Nm
1768.Cm set swap Ar first-set second-set
1769.Ed
1770.Pp
1771See the
1772.Sx EXAMPLES
1773Section on some possible uses of sets of rules.
1774.Sh STATEFUL FIREWALL
1775Stateful operation is a way for the firewall to dynamically
1776create rules for specific flows when packets that
1777match a given pattern are detected.
1778Support for stateful
1779operation comes through the
1780.Cm check-state , keep-state
1781and
1782.Cm limit
1783options of
1784.Nm rules .
1785.Pp
1786Dynamic rules are created when a packet matches a
1787.Cm keep-state
1788or
1789.Cm limit
1790rule, causing the creation of a
1791.Em dynamic
1792rule which will match all and only packets with
1793a given
1794.Em protocol
1795between a
1796.Em src-ip/src-port dst-ip/dst-port
1797pair of addresses
1798.Em ( src
1799and
1800.Em dst
1801are used here only to denote the initial match addresses, but they
1802are completely equivalent afterwards).
1803Dynamic rules will be checked at the first
1804.Cm check-state, keep-state
1805or
1806.Cm limit
1807occurrence, and the action performed upon a match will be the same
1808as in the parent rule.
1809.Pp
1810Note that no additional attributes other than protocol and IP addresses
1811and ports are checked on dynamic rules.
1812.Pp
1813The typical use of dynamic rules is to keep a closed firewall configuration,
1814but let the first TCP SYN packet from the inside network install a
1815dynamic rule for the flow so that packets belonging to that session
1816will be allowed through the firewall:
1817.Pp
1818.Dl "ipfw add check-state"
1819.Dl "ipfw add allow tcp from my-subnet to any setup keep-state"
1820.Dl "ipfw add deny tcp from any to any"
1821.Pp
1822A similar approach can be used for UDP, where an UDP packet coming
1823from the inside will install a dynamic rule to let the response through
1824the firewall:
1825.Pp
1826.Dl "ipfw add check-state"
1827.Dl "ipfw add allow udp from my-subnet to any keep-state"
1828.Dl "ipfw add deny udp from any to any"
1829.Pp
1830Dynamic rules expire after some time, which depends on the status
1831of the flow and the setting of some
1832.Cm sysctl
1833variables.
1834See Section
1835.Sx SYSCTL VARIABLES
1836for more details.
1837For TCP sessions, dynamic rules can be instructed to periodically
1838send keepalive packets to refresh the state of the rule when it is
1839about to expire.
1840.Pp
1841See Section
1842.Sx EXAMPLES
1843for more examples on how to use dynamic rules.
1844.Sh TRAFFIC SHAPER (DUMMYNET) CONFIGURATION
1845.Nm
1846is also the user interface for the
1847.Nm dummynet
1848traffic shaper, packet scheduler and network emulator, a subsystem that
1849can artificially queue, delay or drop packets
1850emulating the behaviour of certain network links
1851or queueing systems.
1852.Pp
1853.Nm dummynet
1854operates by first using the firewall to select packets
1855using any match pattern that can be used in
1856.Nm
1857rules.
1858Matching packets are then passed to either of two
1859different objects, which implement the traffic regulation:
1860.Bl -hang -offset XXXX
1861.It Em pipe
1862A
1863.Em pipe
1864emulates a
1865.Em link
1866with given bandwidth and propagation delay,
1867driven by a FIFO scheduler and a single queue with programmable
1868queue size and packet loss rate.
1869Packets are appended to the queue as they come out from
1870.Nm ipfw ,
1871and then transferred in FIFO order to the link at the desired rate.
1872.It Em queue
1873A
1874.Em queue
1875is an abstraction used to implement packet scheduling
1876using one of several packet scheduling algorithms.
1877Packets sent to a
1878.Em queue
1879are first grouped into flows according to a mask on the 5-tuple.
1880Flows are then passed to the scheduler associated to the
1881.Em queue ,
1882and each flow uses scheduling parameters (weight and others)
1883as configured in the
1884.Em queue
1885itself.
1886A scheduler in turn is connected to an emulated link,
1887and arbitrates the link's bandwidth among backlogged flows according to
1888weights and to the features of the scheduling algorithm in use.
1889.El
1890.Pp
1891In practice,
1892.Em pipes
1893can be used to set hard limits to the bandwidth that a flow can use, whereas
1894.Em queues
1895can be used to determine how different flows share the available bandwidth.
1896.Pp
1897A graphical representation of the binding of queues,
1898flows, schedulers and links is below.
1899.Bd -literal -offset indent
1900                 (flow_mask|sched_mask)  sched_mask
1901         +---------+   weight Wx  +-------------+
1902         |         |->-[flow]-->--|             |-+
1903    -->--| QUEUE x |   ...        |             | |
1904         |         |->-[flow]-->--| SCHEDuler N | |
1905         +---------+              |             | |
1906             ...                  |             +--[LINK N]-->--
1907         +---------+   weight Wy  |             | +--[LINK N]-->--
1908         |         |->-[flow]-->--|             | |
1909    -->--| QUEUE y |   ...        |             | |
1910         |         |->-[flow]-->--|             | |
1911         +---------+              +-------------+ |
1912                                    +-------------+
1913.Ed
1914It is important to understand the role of the SCHED_MASK
1915and FLOW_MASK, which are configured through the commands
1916.Dl "ipfw sched N config mask SCHED_MASK ..."
1917and
1918.Dl "ipfw queue X config mask FLOW_MASK ..." .
1919.Pp
1920The SCHED_MASK is used to assign flows to one or more
1921scheduler instances, one for each
1922value of the packet's 5-fuple after applying SCHED_MASK.
1923As an example, using ``src-ip 0xffffff00'' creates one instance
1924for each /24 destination subnet.
1925.Pp
1926The FLOW_MASK, together with the SCHED_MASK, is used to split
1927packets into flows. As an example, using
1928``src-ip 0x000000ff''
1929together with the previous SCHED_MASK makes a flow for
1930each individual source address. In turn, flows for each /24
1931subnet will be sent to the same scheduler instance.
1932.Pp
1933The above diagram holds even for the
1934.Em pipe
1935case, with the only restriction that a
1936.Em pipe
1937only supports a SCHED_MASK, and forces the use of a FIFO
1938scheduler (these are for backward compatibility reasons;
1939in fact, internally, a
1940.Nm dummynet's
1941pipe is implemented exactly as above).
1942.Pp
1943There are two modes of
1944.Nm dummynet
1945operation:
1946.Dq normal
1947and
1948.Dq fast .
1949The
1950.Dq normal
1951mode tries to emulate a real link: the
1952.Nm dummynet
1953scheduler ensures that the packet will not leave the pipe faster than it
1954would on the real link with a given bandwidth.
1955The
1956.Dq fast
1957mode allows certain packets to bypass the
1958.Nm dummynet
1959scheduler (if packet flow does not exceed pipe's bandwidth).
1960This is the reason why the
1961.Dq fast
1962mode requires less CPU cycles per packet (on average) and packet latency
1963can be significantly lower in comparison to a real link with the same
1964bandwidth.
1965The default mode is
1966.Dq normal .
1967The
1968.Dq fast
1969mode can be enabled by setting the
1970.Va net.inet.ip.dummynet.io_fast
1971.Xr sysctl 8
1972variable to a non-zero value.
1973.Pp
1974.Ss PIPE, QUEUE AND SCHEDULER CONFIGURATION
1975The
1976.Em pipe ,
1977.Em queue
1978and
1979.Em scheduler
1980configuration commands are the following:
1981.Bd -ragged -offset indent
1982.Cm pipe Ar number Cm config Ar pipe-configuration
1983.Pp
1984.Cm queue Ar number Cm config Ar queue-configuration
1985.Pp
1986.Cm sched Ar number Cm config Ar sched-configuration
1987.Ed
1988.Pp
1989The following parameters can be configured for a pipe:
1990.Pp
1991.Bl -tag -width indent -compact
1992.It Cm bw Ar bandwidth | device
1993Bandwidth, measured in
1994.Sm off
1995.Op Cm K | M
1996.Brq Cm bit/s | Byte/s .
1997.Sm on
1998.Pp
1999A value of 0 (default) means unlimited bandwidth.
2000The unit must immediately follow the number, as in
2001.Pp
2002.Dl "ipfw pipe 1 config bw 300Kbit/s"
2003.Pp
2004If a device name is specified instead of a numeric value, as in
2005.Pp
2006.Dl "ipfw pipe 1 config bw tun0"
2007.Pp
2008then the transmit clock is supplied by the specified device.
2009At the moment only the
2010.Xr tun 4
2011device supports this
2012functionality, for use in conjunction with
2013.Xr ppp 8 .
2014.Pp
2015.It Cm delay Ar ms-delay
2016Propagation delay, measured in milliseconds.
2017The value is rounded to the next multiple of the clock tick
2018(typically 10ms, but it is a good practice to run kernels
2019with
2020.Dq "options HZ=1000"
2021to reduce
2022the granularity to 1ms or less).
2023The default value is 0, meaning no delay.
2024.Pp
2025.It Cm burst Ar size
2026If the data to be sent exceeds the pipe's bandwidth limit
2027(and the pipe was previously idle), up to
2028.Ar size
2029bytes of data are allowed to bypass the
2030.Nm dummynet
2031scheduler, and will be sent as fast as the physical link allows.
2032Any additional data will be transmitted at the rate specified
2033by the
2034.Nm pipe
2035bandwidth.
2036The burst size depends on how long the pipe has been idle;
2037the effective burst size is calculated as follows:
2038MAX(
2039.Ar size
2040,
2041.Nm bw
2042* pipe_idle_time).
2043.Pp
2044.It Cm profile Ar filename
2045A file specifying the additional overhead incurred in the transmission
2046of a packet on the link.
2047.Pp
2048Some link types introduce extra delays in the transmission
2049of a packet, e.g. because of MAC level framing, contention on
2050the use of the channel, MAC level retransmissions and so on.
2051From our point of view, the channel is effectively unavailable
2052for this extra time, which is constant or variable depending
2053on the link type. Additionally, packets may be dropped after this
2054time (e.g. on a wireless link after too many retransmissions).
2055We can model the additional delay with an empirical curve
2056that represents its distribution.
2057.Bd -literal -offset indent
2058      cumulative probability
2059      1.0 ^
2060          |
2061      L   +-- loss-level          x
2062          |                 ******
2063          |                *
2064          |           *****
2065          |          *
2066          |        **
2067          |       *
2068          +-------*------------------->
2069                      delay
2070.Ed
2071The empirical curve may have both vertical and horizontal lines.
2072Vertical lines represent constant delay for a range of
2073probabilities.
2074Horizontal lines correspond to a discontinuity in the delay
2075distribution: the pipe will use the largest delay for a
2076given probability.
2077.Pp
2078The file format is the following, with whitespace acting as
2079a separator and '#' indicating the beginning a comment:
2080.Bl -tag -width indent
2081.It Cm name Ar identifier
2082optional name (listed by "ipfw pipe show")
2083to identify the delay distribution;
2084.It Cm bw Ar value
2085the bandwidth used for the pipe.
2086If not specified here, it must be present
2087explicitly as a configuration parameter for the pipe;
2088.It Cm loss-level Ar L
2089the probability above which packets are lost.
2090(0.0 <= L <= 1.0, default 1.0 i.e. no loss);
2091.It Cm samples Ar N
2092the number of samples used in the internal
2093representation of the curve (2..1024; default 100);
2094.It Cm "delay prob" | "prob delay"
2095One of these two lines is mandatory and defines
2096the format of the following lines with data points.
2097.It Ar XXX Ar YYY
20982 or more lines representing points in the curve,
2099with either delay or probability first, according
2100to the chosen format.
2101The unit for delay is milliseconds.
2102Data points do not need to be sorted.
2103Also, tne number of actual lines can be different
2104from the value of the "samples" parameter:
2105.Nm
2106utility will sort and interpolate
2107the curve as needed.
2108.El
2109.Pp
2110Example of a profile file:
2111.Bd -literal -offset indent
2112name    bla_bla_bla
2113samples 100
2114loss-level    0.86
2115prob    delay
21160       200	# minimum overhead is 200ms
21170.5     200
21180.5     300
21190.8     1000
21200.9     1300
21211       1300
2122#configuration file end
2123.Ed
2124.El
2125.Pp
2126The following parameters can be configured for a queue:
2127.Pp
2128.Bl -tag -width indent -compact
2129.It Cm pipe Ar pipe_nr
2130Connects a queue to the specified pipe.
2131Multiple queues (with the same or different weights) can be connected to
2132the same pipe, which specifies the aggregate rate for the set of queues.
2133.Pp
2134.It Cm weight Ar weight
2135Specifies the weight to be used for flows matching this queue.
2136The weight must be in the range 1..100, and defaults to 1.
2137.El
2138.Pp
2139The following parameters can be configured for a scheduler:
2140.Pp
2141.Bl -tag -width indent -compact
2142.It Cm type Ar {fifo | wf2qp | rr | qfq}
2143specifies the scheduling algorithm to use.
2144.Bl -tag -width indent -compact
2145.It cm fifo
2146is just a FIFO scheduler (which means that all packets
2147are stored in the same queue as they arrive to the scheduler).
2148FIFO has O(1) per-packet time complexity, with very low
2149constants (estimate 60-80ns on a 2Ghz desktop machine)
2150but gives no service guarantees.
2151.It Cm wf2qp
2152implements the WF2Q+ algorithm, which is a Weighted Fair Queueing
2153algorithm which permits flows to share bandwidth according to
2154their weights. Note that weights are not priorities; even a flow
2155with a minuscule weight will never starve.
2156WF2Q+ has O(log N) per-packet processing cost, where N is the number
2157of flows, and is the default algorithm used by previous versions
2158dummynet's queues.
2159.It Cm rr
2160implements the Deficit Round Robin algorithm, which has O(1) processing
2161costs (roughly, 100-150ns per packet)
2162and permits bandwidth allocation according to weights, but
2163with poor service guarantees.
2164.It Cm qfq
2165implements the QFQ algorithm, which is a very fast variant of
2166WF2Q+, with similar service guarantees and O(1) processing
2167costs (roughly, 200-250ns per packet).
2168.El
2169.El
2170.Pp
2171In addition to the type, all parameters allowed for a pipe can also
2172be specified for a scheduler.
2173.Pp
2174Finally, the following parameters can be configured for both
2175pipes and queues:
2176.Pp
2177.Bl -tag -width XXXX -compact
2178.Pp
2179.It Cm buckets Ar hash-table-size
2180Specifies the size of the hash table used for storing the
2181various queues.
2182Default value is 64 controlled by the
2183.Xr sysctl 8
2184variable
2185.Va net.inet.ip.dummynet.hash_size ,
2186allowed range is 16 to 65536.
2187.Pp
2188.It Cm mask Ar mask-specifier
2189Packets sent to a given pipe or queue by an
2190.Nm
2191rule can be further classified into multiple flows, each of which is then
2192sent to a different
2193.Em dynamic
2194pipe or queue.
2195A flow identifier is constructed by masking the IP addresses,
2196ports and protocol types as specified with the
2197.Cm mask
2198options in the configuration of the pipe or queue.
2199For each different flow identifier, a new pipe or queue is created
2200with the same parameters as the original object, and matching packets
2201are sent to it.
2202.Pp
2203Thus, when
2204.Em dynamic pipes
2205are used, each flow will get the same bandwidth as defined by the pipe,
2206whereas when
2207.Em dynamic queues
2208are used, each flow will share the parent's pipe bandwidth evenly
2209with other flows generated by the same queue (note that other queues
2210with different weights might be connected to the same pipe).
2211.br
2212Available mask specifiers are a combination of one or more of the following:
2213.Pp
2214.Cm dst-ip Ar mask ,
2215.Cm dst-ip6 Ar mask ,
2216.Cm src-ip Ar mask ,
2217.Cm src-ip6 Ar mask ,
2218.Cm dst-port Ar mask ,
2219.Cm src-port Ar mask ,
2220.Cm flow-id Ar mask ,
2221.Cm proto Ar mask
2222or
2223.Cm all ,
2224.Pp
2225where the latter means all bits in all fields are significant.
2226.Pp
2227.It Cm noerror
2228When a packet is dropped by a
2229.Nm dummynet
2230queue or pipe, the error
2231is normally reported to the caller routine in the kernel, in the
2232same way as it happens when a device queue fills up.
2233Setting this
2234option reports the packet as successfully delivered, which can be
2235needed for some experimental setups where you want to simulate
2236loss or congestion at a remote router.
2237.Pp
2238.It Cm plr Ar packet-loss-rate
2239Packet loss rate.
2240Argument
2241.Ar packet-loss-rate
2242is a floating-point number between 0 and 1, with 0 meaning no
2243loss, 1 meaning 100% loss.
2244The loss rate is internally represented on 31 bits.
2245.Pp
2246.It Cm queue Brq Ar slots | size Ns Cm Kbytes
2247Queue size, in
2248.Ar slots
2249or
2250.Cm KBytes .
2251Default value is 50 slots, which
2252is the typical queue size for Ethernet devices.
2253Note that for slow speed links you should keep the queue
2254size short or your traffic might be affected by a significant
2255queueing delay.
2256E.g., 50 max-sized ethernet packets (1500 bytes) mean 600Kbit
2257or 20s of queue on a 30Kbit/s pipe.
2258Even worse effects can result if you get packets from an
2259interface with a much larger MTU, e.g.\& the loopback interface
2260with its 16KB packets.
2261The
2262.Xr sysctl 8
2263variables
2264.Em net.inet.ip.dummynet.pipe_byte_limit
2265and
2266.Em net.inet.ip.dummynet.pipe_slot_limit
2267control the maximum lengths that can be specified.
2268.Pp
2269.It Cm red | gred Ar w_q Ns / Ns Ar min_th Ns / Ns Ar max_th Ns / Ns Ar max_p
2270Make use of the RED (Random Early Detection) queue management algorithm.
2271.Ar w_q
2272and
2273.Ar max_p
2274are floating
2275point numbers between 0 and 1 (0 not included), while
2276.Ar min_th
2277and
2278.Ar max_th
2279are integer numbers specifying thresholds for queue management
2280(thresholds are computed in bytes if the queue has been defined
2281in bytes, in slots otherwise).
2282The
2283.Nm dummynet
2284also supports the gentle RED variant (gred).
2285Three
2286.Xr sysctl 8
2287variables can be used to control the RED behaviour:
2288.Bl -tag -width indent
2289.It Va net.inet.ip.dummynet.red_lookup_depth
2290specifies the accuracy in computing the average queue
2291when the link is idle (defaults to 256, must be greater than zero)
2292.It Va net.inet.ip.dummynet.red_avg_pkt_size
2293specifies the expected average packet size (defaults to 512, must be
2294greater than zero)
2295.It Va net.inet.ip.dummynet.red_max_pkt_size
2296specifies the expected maximum packet size, only used when queue
2297thresholds are in bytes (defaults to 1500, must be greater than zero).
2298.El
2299.El
2300.Pp
2301When used with IPv6 data,
2302.Nm dummynet
2303currently has several limitations.
2304Information necessary to route link-local packets to an
2305interface is not available after processing by
2306.Nm dummynet
2307so those packets are dropped in the output path.
2308Care should be taken to insure that link-local packets are not passed to
2309.Nm dummynet .
2310.Sh CHECKLIST
2311Here are some important points to consider when designing your
2312rules:
2313.Bl -bullet
2314.It
2315Remember that you filter both packets going
2316.Cm in
2317and
2318.Cm out .
2319Most connections need packets going in both directions.
2320.It
2321Remember to test very carefully.
2322It is a good idea to be near the console when doing this.
2323If you cannot be near the console,
2324use an auto-recovery script such as the one in
2325.Pa /usr/share/examples/ipfw/change_rules.sh .
2326.It
2327Do not forget the loopback interface.
2328.El
2329.Sh FINE POINTS
2330.Bl -bullet
2331.It
2332There are circumstances where fragmented datagrams are unconditionally
2333dropped.
2334TCP packets are dropped if they do not contain at least 20 bytes of
2335TCP header, UDP packets are dropped if they do not contain a full 8
2336byte UDP header, and ICMP packets are dropped if they do not contain
23374 bytes of ICMP header, enough to specify the ICMP type, code, and
2338checksum.
2339These packets are simply logged as
2340.Dq pullup failed
2341since there may not be enough good data in the packet to produce a
2342meaningful log entry.
2343.It
2344Another type of packet is unconditionally dropped, a TCP packet with a
2345fragment offset of one.
2346This is a valid packet, but it only has one use, to try
2347to circumvent firewalls.
2348When logging is enabled, these packets are
2349reported as being dropped by rule -1.
2350.It
2351If you are logged in over a network, loading the
2352.Xr kld 4
2353version of
2354.Nm
2355is probably not as straightforward as you would think.
2356The following command line is recommended:
2357.Bd -literal -offset indent
2358kldload ipfw && \e
2359ipfw add 32000 allow ip from any to any
2360.Ed
2361.Pp
2362Along the same lines, doing an
2363.Bd -literal -offset indent
2364ipfw flush
2365.Ed
2366.Pp
2367in similar surroundings is also a bad idea.
2368.It
2369The
2370.Nm
2371filter list may not be modified if the system security level
2372is set to 3 or higher
2373(see
2374.Xr init 8
2375for information on system security levels).
2376.El
2377.Sh PACKET DIVERSION
2378A
2379.Xr divert 4
2380socket bound to the specified port will receive all packets
2381diverted to that port.
2382If no socket is bound to the destination port, or if the divert module is
2383not loaded, or if the kernel was not compiled with divert socket support,
2384the packets are dropped.
2385.Sh NETWORK ADDRESS TRANSLATION (NAT)
2386.Pp
2387.Nm
2388support in-kernel NAT using the kernel version of
2389.Xr libalias 3 .
2390.Pp
2391The nat configuration command is the following:
2392.Bd -ragged -offset indent
2393.Bk -words
2394.Cm nat
2395.Ar nat_number
2396.Cm config
2397.Ar nat-configuration
2398.Ek
2399.Ed
2400.Pp
2401The following parameters can be configured:
2402.Bl -tag -width indent
2403.It Cm ip Ar ip_address
2404Define an ip address to use for aliasing.
2405.It Cm if Ar nic
2406Use ip address of NIC for aliasing, dynamically changing
2407it if NIC's ip address changes.
2408.It Cm log
2409Enable logging on this nat instance.
2410.It Cm deny_in
2411Deny any incoming connection from outside world.
2412.It Cm same_ports
2413Try to leave the alias port numbers unchanged from
2414the actual local port numbers.
2415.It Cm unreg_only
2416Traffic on the local network not originating from an
2417unregistered address spaces will be ignored.
2418.It Cm reset
2419Reset table of the packet aliasing engine on address change.
2420.It Cm reverse
2421Reverse the way libalias handles aliasing.
2422.It Cm proxy_only
2423Obey transparent proxy rules only, packet aliasing is not performed.
2424.El
2425.Pp
2426To let the packet continue after being (de)aliased, set the sysctl variable
2427.Va net.inet.ip.fw.one_pass
2428to 0.
2429For more information about aliasing modes, refer to
2430.Xr libalias 3 .
2431See Section
2432.Sx EXAMPLES
2433for some examples about nat usage.
2434.Ss REDIRECT AND LSNAT SUPPORT IN IPFW
2435Redirect and LSNAT support follow closely the syntax used in
2436.Xr natd 8 .
2437See Section
2438.Sx EXAMPLES
2439for some examples on how to do redirect and lsnat.
2440.Ss SCTP NAT SUPPORT
2441SCTP nat can be configured in a similar manner to TCP through the
2442.Nm
2443command line tool.
2444The main difference is that
2445.Nm sctp nat
2446does not do port translation.
2447Since the local and global side ports will be the same,
2448there is no need to specify both.
2449Ports are redirected as follows:
2450.Bd -ragged -offset indent
2451.Bk -words
2452.Cm nat
2453.Ar nat_number
2454.Cm config if
2455.Ar nic
2456.Cm redirect_port sctp
2457.Ar ip_address [,addr_list] {[port | port-port] [,ports]}
2458.Ek
2459.Ed
2460.Pp
2461Most
2462.Nm sctp nat
2463configuration can be done in real-time through the
2464.Xr sysctl 8
2465interface.
2466All may be changed dynamically, though the hash_table size will only
2467change for new
2468.Nm nat
2469instances.
2470See
2471.Sx SYSCTL VARIABLES
2472for more info.
2473.Sh SYSCTL VARIABLES
2474A set of
2475.Xr sysctl 8
2476variables controls the behaviour of the firewall and
2477associated modules
2478.Pq Nm dummynet , bridge , sctp nat .
2479These are shown below together with their default value
2480(but always check with the
2481.Xr sysctl 8
2482command what value is actually in use) and meaning:
2483.Bl -tag -width indent
2484.It Va net.inet.ip.alias.sctp.accept_global_ootb_addip: No 0
2485Defines how the
2486.Nm nat
2487responds to receipt of global OOTB ASCONF-AddIP:
2488.Bl -tag -width indent
2489.It Cm 0
2490No response (unless a partially matching association exists -
2491ports and vtags match but global address does not)
2492.It Cm 1
2493.Nm nat
2494will accept and process all OOTB global AddIP messages.
2495.El
2496.Pp
2497Option 1 should never be selected as this forms a security risk.
2498An attacker can
2499establish multiple fake associations by sending AddIP messages.
2500.It Va net.inet.ip.alias.sctp.chunk_proc_limit: No 5
2501Defines the maximum number of chunks in an SCTP packet that will be parsed for a
2502packet that matches an existing association.
2503This value is enforced to be greater or equal than
2504.Cm net.inet.ip.alias.sctp.initialising_chunk_proc_limit .
2505A high value is
2506a DoS risk yet setting too low a value may result in important control chunks in
2507the packet not being located and parsed.
2508.It Va net.inet.ip.alias.sctp.error_on_ootb: No 1
2509Defines when the
2510.Nm nat
2511responds to any Out-of-the-Blue (OOTB) packets with ErrorM packets.
2512An OOTB packet is a packet that arrives with no existing association
2513registered in the
2514.Nm nat
2515and is not an INIT or ASCONF-AddIP packet:
2516.Bl -tag -width indent
2517.It Cm 0
2518ErrorM is never sent in response to OOTB packets.
2519.It Cm 1
2520ErrorM is only sent to OOTB packets received on the local side.
2521.It Cm 2
2522ErrorM is sent to the local side and on the global side ONLY if there is a
2523partial match (ports and vtags match but the source global IP does not).
2524This value is only useful if the
2525.Nm nat
2526is tracking global IP addresses.
2527.It Cm 3
2528ErrorM is sent in response to all OOTB packets on both the local and global side
2529(DoS risk).
2530.El
2531.Pp
2532At the moment the default is 0, since the ErrorM packet is not yet
2533supported by most SCTP stacks.
2534When it is supported, and if not tracking
2535global addresses, we recommend setting this value to 1 to allow
2536multi-homed local hosts to function with the
2537.Nm nat .
2538To track global addresses, we recommend setting this value to 2 to
2539allow global hosts to be informed when they need to (re)send an
2540ASCONF-AddIP.
2541Value 3 should never be chosen (except for debugging) as the
2542.Nm nat
2543will respond to all OOTB global packets (a DoS risk).
2544.It Va net.inet.ip.alias.sctp.hashtable_size: No 2003
2545Size of hash tables used for
2546.Nm nat
2547lookups (100 < prime_number > 1000001).
2548This value sets the
2549.Nm hash table
2550size for any future created
2551.Nm nat
2552instance and therefore must be set prior to creating a
2553.Nm nat
2554instance.
2555The table sizes may be changed to suit specific needs.
2556If there will be few
2557concurrent associations, and memory is scarce, you may make these smaller.
2558If there will be many thousands (or millions) of concurrent associations, you
2559should make these larger.
2560A prime number is best for the table size.
2561The sysctl
2562update function will adjust your input value to the next highest prime number.
2563.It Va net.inet.ip.alias.sctp.holddown_time:  No 0
2564Hold association in table for this many seconds after receiving a
2565SHUTDOWN-COMPLETE.
2566This allows endpoints to correct shutdown gracefully if a
2567shutdown_complete is lost and retransmissions are required.
2568.It Va net.inet.ip.alias.sctp.init_timer: No 15
2569Timeout value while waiting for (INIT-ACK|AddIP-ACK).
2570This value cannot be 0.
2571.It Va net.inet.ip.alias.sctp.initialising_chunk_proc_limit: No 2
2572Defines the maximum number of chunks in an SCTP packet that will be parsed when
2573no existing association exists that matches that packet.
2574Ideally this packet
2575will only be an INIT or ASCONF-AddIP packet.
2576A higher value may become a DoS
2577risk as malformed packets can consume processing resources.
2578.It Va net.inet.ip.alias.sctp.param_proc_limit: No 25
2579Defines the maximum number of parameters within a chunk that will be parsed in a
2580packet.
2581As for other similar sysctl variables, larger values pose a DoS risk.
2582.It Va net.inet.ip.alias.sctp.log_level: No 0
2583Level of detail in the system log messages (0 \- minimal, 1 \- event,
25842 \- info, 3 \- detail, 4 \- debug, 5 \- max debug). May be a good
2585option in high loss environments.
2586.It Va net.inet.ip.alias.sctp.shutdown_time: No 15
2587Timeout value while waiting for SHUTDOWN-COMPLETE.
2588This value cannot be 0.
2589.It Va net.inet.ip.alias.sctp.track_global_addresses: No 0
2590Enables/disables global IP address tracking within the
2591.Nm nat
2592and places an
2593upper limit on the number of addresses tracked for each association:
2594.Bl -tag -width indent
2595.It Cm 0
2596Global tracking is disabled
2597.It Cm >1
2598Enables tracking, the maximum number of addresses tracked for each
2599association is limited to this value
2600.El
2601.Pp
2602This variable is fully dynamic, the new value will be adopted for all newly
2603arriving associations, existing associations are treated as they were previously.
2604Global tracking will decrease the number of collisions within the
2605.Nm nat
2606at a cost
2607of increased processing load, memory usage, complexity, and possible
2608.Nm nat
2609state
2610problems in complex networks with multiple
2611.Nm nats .
2612We recommend not tracking
2613global IP addresses, this will still result in a fully functional
2614.Nm nat .
2615.It Va net.inet.ip.alias.sctp.up_timer: No 300
2616Timeout value to keep an association up with no traffic.
2617This value cannot be 0.
2618.It Va net.inet.ip.dummynet.expire : No 1
2619Lazily delete dynamic pipes/queue once they have no pending traffic.
2620You can disable this by setting the variable to 0, in which case
2621the pipes/queues will only be deleted when the threshold is reached.
2622.It Va net.inet.ip.dummynet.hash_size : No 64
2623Default size of the hash table used for dynamic pipes/queues.
2624This value is used when no
2625.Cm buckets
2626option is specified when configuring a pipe/queue.
2627.It Va net.inet.ip.dummynet.io_fast : No 0
2628If set to a non-zero value,
2629the
2630.Dq fast
2631mode of
2632.Nm dummynet
2633operation (see above) is enabled.
2634.It Va net.inet.ip.dummynet.io_pkt
2635Number of packets passed to
2636.Nm dummynet .
2637.It Va net.inet.ip.dummynet.io_pkt_drop
2638Number of packets dropped by
2639.Nm dummynet .
2640.It Va net.inet.ip.dummynet.io_pkt_fast
2641Number of packets bypassed by the
2642.Nm dummynet
2643scheduler.
2644.It Va net.inet.ip.dummynet.max_chain_len : No 16
2645Target value for the maximum number of pipes/queues in a hash bucket.
2646The product
2647.Cm max_chain_len*hash_size
2648is used to determine the threshold over which empty pipes/queues
2649will be expired even when
2650.Cm net.inet.ip.dummynet.expire=0 .
2651.It Va net.inet.ip.dummynet.red_lookup_depth : No 256
2652.It Va net.inet.ip.dummynet.red_avg_pkt_size : No 512
2653.It Va net.inet.ip.dummynet.red_max_pkt_size : No 1500
2654Parameters used in the computations of the drop probability
2655for the RED algorithm.
2656.It Va net.inet.ip.dummynet.pipe_byte_limit : No 1048576
2657.It Va net.inet.ip.dummynet.pipe_slot_limit : No 100
2658The maximum queue size that can be specified in bytes or packets.
2659These limits prevent accidental exhaustion of resources such as mbufs.
2660If you raise these limits,
2661you should make sure the system is configured so that sufficient resources
2662are available.
2663.It Va net.inet.ip.fw.autoinc_step : No 100
2664Delta between rule numbers when auto-generating them.
2665The value must be in the range 1..1000.
2666.It Va net.inet.ip.fw.curr_dyn_buckets : Va net.inet.ip.fw.dyn_buckets
2667The current number of buckets in the hash table for dynamic rules
2668(readonly).
2669.It Va net.inet.ip.fw.debug : No 1
2670Controls debugging messages produced by
2671.Nm .
2672.It Va net.inet.ip.fw.default_rule : No 65535
2673The default rule number (read-only).
2674By the design of
2675.Nm , the default rule is the last one, so its number
2676can also serve as the highest number allowed for a rule.
2677.It Va net.inet.ip.fw.dyn_buckets : No 256
2678The number of buckets in the hash table for dynamic rules.
2679Must be a power of 2, up to 65536.
2680It only takes effect when all dynamic rules have expired, so you
2681are advised to use a
2682.Cm flush
2683command to make sure that the hash table is resized.
2684.It Va net.inet.ip.fw.dyn_count : No 3
2685Current number of dynamic rules
2686(read-only).
2687.It Va net.inet.ip.fw.dyn_keepalive : No 1
2688Enables generation of keepalive packets for
2689.Cm keep-state
2690rules on TCP sessions.
2691A keepalive is generated to both
2692sides of the connection every 5 seconds for the last 20
2693seconds of the lifetime of the rule.
2694.It Va net.inet.ip.fw.dyn_max : No 8192
2695Maximum number of dynamic rules.
2696When you hit this limit, no more dynamic rules can be
2697installed until old ones expire.
2698.It Va net.inet.ip.fw.dyn_ack_lifetime : No 300
2699.It Va net.inet.ip.fw.dyn_syn_lifetime : No 20
2700.It Va net.inet.ip.fw.dyn_fin_lifetime : No 1
2701.It Va net.inet.ip.fw.dyn_rst_lifetime : No 1
2702.It Va net.inet.ip.fw.dyn_udp_lifetime : No 5
2703.It Va net.inet.ip.fw.dyn_short_lifetime : No 30
2704These variables control the lifetime, in seconds, of dynamic
2705rules.
2706Upon the initial SYN exchange the lifetime is kept short,
2707then increased after both SYN have been seen, then decreased
2708again during the final FIN exchange or when a RST is received.
2709Both
2710.Em dyn_fin_lifetime
2711and
2712.Em dyn_rst_lifetime
2713must be strictly lower than 5 seconds, the period of
2714repetition of keepalives.
2715The firewall enforces that.
2716.It Va net.inet.ip.fw.enable : No 1
2717Enables the firewall.
2718Setting this variable to 0 lets you run your machine without
2719firewall even if compiled in.
2720.It Va net.inet6.ip6.fw.enable : No 1
2721provides the same functionality as above for the IPv6 case.
2722.It Va net.inet.ip.fw.one_pass : No 1
2723When set, the packet exiting from the
2724.Nm dummynet
2725pipe or from
2726.Xr ng_ipfw 4
2727node is not passed though the firewall again.
2728Otherwise, after an action, the packet is
2729reinjected into the firewall at the next rule.
2730.It Va net.inet.ip.fw.tables_max : No 128
2731Maximum number of tables (read-only).
2732.It Va net.inet.ip.fw.verbose : No 1
2733Enables verbose messages.
2734.It Va net.inet.ip.fw.verbose_limit : No 0
2735Limits the number of messages produced by a verbose firewall.
2736.It Va net.inet6.ip6.fw.deny_unknown_exthdrs : No 1
2737If enabled packets with unknown IPv6 Extension Headers will be denied.
2738.It Va net.link.ether.ipfw : No 0
2739Controls whether layer-2 packets are passed to
2740.Nm .
2741Default is no.
2742.It Va net.link.bridge.ipfw : No 0
2743Controls whether bridged packets are passed to
2744.Nm .
2745Default is no.
2746.El
2747.Pp
2748.Sh EXAMPLES
2749There are far too many possible uses of
2750.Nm
2751so this Section will only give a small set of examples.
2752.Pp
2753.Ss BASIC PACKET FILTERING
2754This command adds an entry which denies all tcp packets from
2755.Em cracker.evil.org
2756to the telnet port of
2757.Em wolf.tambov.su
2758from being forwarded by the host:
2759.Pp
2760.Dl "ipfw add deny tcp from cracker.evil.org to wolf.tambov.su telnet"
2761.Pp
2762This one disallows any connection from the entire cracker's
2763network to my host:
2764.Pp
2765.Dl "ipfw add deny ip from 123.45.67.0/24 to my.host.org"
2766.Pp
2767A first and efficient way to limit access (not using dynamic rules)
2768is the use of the following rules:
2769.Pp
2770.Dl "ipfw add allow tcp from any to any established"
2771.Dl "ipfw add allow tcp from net1 portlist1 to net2 portlist2 setup"
2772.Dl "ipfw add allow tcp from net3 portlist3 to net3 portlist3 setup"
2773.Dl "..."
2774.Dl "ipfw add deny tcp from any to any"
2775.Pp
2776The first rule will be a quick match for normal TCP packets,
2777but it will not match the initial SYN packet, which will be
2778matched by the
2779.Cm setup
2780rules only for selected source/destination pairs.
2781All other SYN packets will be rejected by the final
2782.Cm deny
2783rule.
2784.Pp
2785If you administer one or more subnets, you can take advantage
2786of the address sets and or-blocks and write extremely
2787compact rulesets which selectively enable services to blocks
2788of clients, as below:
2789.Pp
2790.Dl "goodguys=\*q{ 10.1.2.0/24{20,35,66,18} or 10.2.3.0/28{6,3,11} }\*q"
2791.Dl "badguys=\*q10.1.2.0/24{8,38,60}\*q"
2792.Dl ""
2793.Dl "ipfw add allow ip from ${goodguys} to any"
2794.Dl "ipfw add deny ip from ${badguys} to any"
2795.Dl "... normal policies ..."
2796.Pp
2797The
2798.Cm verrevpath
2799option could be used to do automated anti-spoofing by adding the
2800following to the top of a ruleset:
2801.Pp
2802.Dl "ipfw add deny ip from any to any not verrevpath in"
2803.Pp
2804This rule drops all incoming packets that appear to be coming to the
2805system on the wrong interface.
2806For example, a packet with a source
2807address belonging to a host on a protected internal network would be
2808dropped if it tried to enter the system from an external interface.
2809.Pp
2810The
2811.Cm antispoof
2812option could be used to do similar but more restricted anti-spoofing
2813by adding the following to the top of a ruleset:
2814.Pp
2815.Dl "ipfw add deny ip from any to any not antispoof in"
2816.Pp
2817This rule drops all incoming packets that appear to be coming from another
2818directly connected system but on the wrong interface.
2819For example, a packet with a source address of
2820.Li 192.168.0.0/24 ,
2821configured on
2822.Li fxp0 ,
2823but coming in on
2824.Li fxp1
2825would be dropped.
2826.Ss DYNAMIC RULES
2827In order to protect a site from flood attacks involving fake
2828TCP packets, it is safer to use dynamic rules:
2829.Pp
2830.Dl "ipfw add check-state"
2831.Dl "ipfw add deny tcp from any to any established"
2832.Dl "ipfw add allow tcp from my-net to any setup keep-state"
2833.Pp
2834This will let the firewall install dynamic rules only for
2835those connection which start with a regular SYN packet coming
2836from the inside of our network.
2837Dynamic rules are checked when encountering the first
2838.Cm check-state
2839or
2840.Cm keep-state
2841rule.
2842A
2843.Cm check-state
2844rule should usually be placed near the beginning of the
2845ruleset to minimize the amount of work scanning the ruleset.
2846Your mileage may vary.
2847.Pp
2848To limit the number of connections a user can open
2849you can use the following type of rules:
2850.Pp
2851.Dl "ipfw add allow tcp from my-net/24 to any setup limit src-addr 10"
2852.Dl "ipfw add allow tcp from any to me setup limit src-addr 4"
2853.Pp
2854The former (assuming it runs on a gateway) will allow each host
2855on a /24 network to open at most 10 TCP connections.
2856The latter can be placed on a server to make sure that a single
2857client does not use more than 4 simultaneous connections.
2858.Pp
2859.Em BEWARE :
2860stateful rules can be subject to denial-of-service attacks
2861by a SYN-flood which opens a huge number of dynamic rules.
2862The effects of such attacks can be partially limited by
2863acting on a set of
2864.Xr sysctl 8
2865variables which control the operation of the firewall.
2866.Pp
2867Here is a good usage of the
2868.Cm list
2869command to see accounting records and timestamp information:
2870.Pp
2871.Dl ipfw -at list
2872.Pp
2873or in short form without timestamps:
2874.Pp
2875.Dl ipfw -a list
2876.Pp
2877which is equivalent to:
2878.Pp
2879.Dl ipfw show
2880.Pp
2881Next rule diverts all incoming packets from 192.168.2.0/24
2882to divert port 5000:
2883.Pp
2884.Dl ipfw divert 5000 ip from 192.168.2.0/24 to any in
2885.Pp
2886.Ss TRAFFIC SHAPING
2887The following rules show some of the applications of
2888.Nm
2889and
2890.Nm dummynet
2891for simulations and the like.
2892.Pp
2893This rule drops random incoming packets with a probability
2894of 5%:
2895.Pp
2896.Dl "ipfw add prob 0.05 deny ip from any to any in"
2897.Pp
2898A similar effect can be achieved making use of
2899.Nm dummynet
2900pipes:
2901.Pp
2902.Dl "ipfw add pipe 10 ip from any to any"
2903.Dl "ipfw pipe 10 config plr 0.05"
2904.Pp
2905We can use pipes to artificially limit bandwidth, e.g.\& on a
2906machine acting as a router, if we want to limit traffic from
2907local clients on 192.168.2.0/24 we do:
2908.Pp
2909.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2910.Dl "ipfw pipe 1 config bw 300Kbit/s queue 50KBytes"
2911.Pp
2912note that we use the
2913.Cm out
2914modifier so that the rule is not used twice.
2915Remember in fact that
2916.Nm
2917rules are checked both on incoming and outgoing packets.
2918.Pp
2919Should we want to simulate a bidirectional link with bandwidth
2920limitations, the correct way is the following:
2921.Pp
2922.Dl "ipfw add pipe 1 ip from any to any out"
2923.Dl "ipfw add pipe 2 ip from any to any in"
2924.Dl "ipfw pipe 1 config bw 64Kbit/s queue 10Kbytes"
2925.Dl "ipfw pipe 2 config bw 64Kbit/s queue 10Kbytes"
2926.Pp
2927The above can be very useful, e.g.\& if you want to see how
2928your fancy Web page will look for a residential user who
2929is connected only through a slow link.
2930You should not use only one pipe for both directions, unless
2931you want to simulate a half-duplex medium (e.g.\& AppleTalk,
2932Ethernet, IRDA).
2933It is not necessary that both pipes have the same configuration,
2934so we can also simulate asymmetric links.
2935.Pp
2936Should we want to verify network performance with the RED queue
2937management algorithm:
2938.Pp
2939.Dl "ipfw add pipe 1 ip from any to any"
2940.Dl "ipfw pipe 1 config bw 500Kbit/s queue 100 red 0.002/30/80/0.1"
2941.Pp
2942Another typical application of the traffic shaper is to
2943introduce some delay in the communication.
2944This can significantly affect applications which do a lot of Remote
2945Procedure Calls, and where the round-trip-time of the
2946connection often becomes a limiting factor much more than
2947bandwidth:
2948.Pp
2949.Dl "ipfw add pipe 1 ip from any to any out"
2950.Dl "ipfw add pipe 2 ip from any to any in"
2951.Dl "ipfw pipe 1 config delay 250ms bw 1Mbit/s"
2952.Dl "ipfw pipe 2 config delay 250ms bw 1Mbit/s"
2953.Pp
2954Per-flow queueing can be useful for a variety of purposes.
2955A very simple one is counting traffic:
2956.Pp
2957.Dl "ipfw add pipe 1 tcp from any to any"
2958.Dl "ipfw add pipe 1 udp from any to any"
2959.Dl "ipfw add pipe 1 ip from any to any"
2960.Dl "ipfw pipe 1 config mask all"
2961.Pp
2962The above set of rules will create queues (and collect
2963statistics) for all traffic.
2964Because the pipes have no limitations, the only effect is
2965collecting statistics.
2966Note that we need 3 rules, not just the last one, because
2967when
2968.Nm
2969tries to match IP packets it will not consider ports, so we
2970would not see connections on separate ports as different
2971ones.
2972.Pp
2973A more sophisticated example is limiting the outbound traffic
2974on a net with per-host limits, rather than per-network limits:
2975.Pp
2976.Dl "ipfw add pipe 1 ip from 192.168.2.0/24 to any out"
2977.Dl "ipfw add pipe 2 ip from any to 192.168.2.0/24 in"
2978.Dl "ipfw pipe 1 config mask src-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2979.Dl "ipfw pipe 2 config mask dst-ip 0x000000ff bw 200Kbit/s queue 20Kbytes"
2980.Ss LOOKUP TABLES
2981In the following example, we need to create several traffic bandwidth
2982classes and we need different hosts/networks to fall into different classes.
2983We create one pipe for each class and configure them accordingly.
2984Then we create a single table and fill it with IP subnets and addresses.
2985For each subnet/host we set the argument equal to the number of the pipe
2986that it should use.
2987Then we classify traffic using a single rule:
2988.Pp
2989.Dl "ipfw pipe 1 config bw 1000Kbyte/s"
2990.Dl "ipfw pipe 4 config bw 4000Kbyte/s"
2991.Dl "..."
2992.Dl "ipfw table 1 add 192.168.2.0/24 1"
2993.Dl "ipfw table 1 add 192.168.0.0/27 4"
2994.Dl "ipfw table 1 add 192.168.0.2 1"
2995.Dl "..."
2996.Dl "ipfw add pipe tablearg ip from table(1) to any"
2997.Pp
2998Using the
2999.Cm fwd
3000action, the table entries may include hostnames and IP addresses.
3001.Pp
3002.Dl "ipfw table 1 add 192.168.2.0/24 10.23.2.1"
3003.Dl "ipfw table 1 add 192.168.0.0/27 router1.dmz"
3004.Dl "..."
3005.Dl "ipfw add 100 fwd tablearg ip from any to table(1)"
3006.Ss SETS OF RULES
3007To add a set of rules atomically, e.g.\& set 18:
3008.Pp
3009.Dl "ipfw set disable 18"
3010.Dl "ipfw add NN set 18 ...         # repeat as needed"
3011.Dl "ipfw set enable 18"
3012.Pp
3013To delete a set of rules atomically the command is simply:
3014.Pp
3015.Dl "ipfw delete set 18"
3016.Pp
3017To test a ruleset and disable it and regain control if something goes wrong:
3018.Pp
3019.Dl "ipfw set disable 18"
3020.Dl "ipfw add NN set 18 ...         # repeat as needed"
3021.Dl "ipfw set enable 18; echo done; sleep 30 && ipfw set disable 18"
3022.Pp
3023Here if everything goes well, you press control-C before the "sleep"
3024terminates, and your ruleset will be left active.
3025Otherwise, e.g.\& if
3026you cannot access your box, the ruleset will be disabled after
3027the sleep terminates thus restoring the previous situation.
3028.Pp
3029To show rules of the specific set:
3030.Pp
3031.Dl "ipfw set 18 show"
3032.Pp
3033To show rules of the disabled set:
3034.Pp
3035.Dl "ipfw -S set 18 show"
3036.Pp
3037To clear a specific rule counters of the specific set:
3038.Pp
3039.Dl "ipfw set 18 zero NN"
3040.Pp
3041To delete a specific rule of the specific set:
3042.Pp
3043.Dl "ipfw set 18 delete NN"
3044.Ss NAT, REDIRECT AND LSNAT
3045First redirect all the traffic to nat instance 123:
3046.Pp
3047.Dl "ipfw add nat 123 all from any to any"
3048.Pp
3049Then to configure nat instance 123 to alias all the outgoing traffic with ip
3050192.168.0.123, blocking all incoming connections, trying to keep
3051same ports on both sides, clearing aliasing table on address change
3052and keeping a log of traffic/link statistics:
3053.Pp
3054.Dl "ipfw nat 123 config ip 192.168.0.123 log deny_in reset same_ports"
3055.Pp
3056Or to change address of instance 123, aliasing table will be cleared (see
3057reset option):
3058.Pp
3059.Dl "ipfw nat 123 config ip 10.0.0.1"
3060.Pp
3061To see configuration of nat instance 123:
3062.Pp
3063.Dl "ipfw nat 123 show config"
3064.Pp
3065To show logs of all the instances in range 111-999:
3066.Pp
3067.Dl "ipfw nat 111-999 show"
3068.Pp
3069To see configurations of all instances:
3070.Pp
3071.Dl "ipfw nat show config"
3072.Pp
3073Or a redirect rule with mixed modes could looks like:
3074.Pp
3075.Dl "ipfw nat 123 config redirect_addr 10.0.0.1 10.0.0.66"
3076.Dl "			 redirect_port tcp 192.168.0.1:80 500"
3077.Dl "			 redirect_proto udp 192.168.1.43 192.168.1.1"
3078.Dl "			 redirect_addr 192.168.0.10,192.168.0.11"
3079.Dl "			 	    10.0.0.100	# LSNAT"
3080.Dl "			 redirect_port tcp 192.168.0.1:80,192.168.0.10:22"
3081.Dl "			 	    500		# LSNAT"
3082.Pp
3083or it could be split in:
3084.Pp
3085.Dl "ipfw nat 1 config redirect_addr 10.0.0.1 10.0.0.66"
3086.Dl "ipfw nat 2 config redirect_port tcp 192.168.0.1:80 500"
3087.Dl "ipfw nat 3 config redirect_proto udp 192.168.1.43 192.168.1.1"
3088.Dl "ipfw nat 4 config redirect_addr 192.168.0.10,192.168.0.11,192.168.0.12"
3089.Dl "				         10.0.0.100"
3090.Dl "ipfw nat 5 config redirect_port tcp"
3091.Dl "			192.168.0.1:80,192.168.0.10:22,192.168.0.20:25 500"
3092.Pp
3093.Sh SEE ALSO
3094.Xr cpp 1 ,
3095.Xr m4 1 ,
3096.Xr altq 4 ,
3097.Xr divert 4 ,
3098.Xr dummynet 4 ,
3099.Xr if_bridge 4 ,
3100.Xr ip 4 ,
3101.Xr ipfirewall 4 ,
3102.Xr ng_ipfw 4 ,
3103.Xr protocols 5 ,
3104.Xr services 5 ,
3105.Xr init 8 ,
3106.Xr kldload 8 ,
3107.Xr reboot 8 ,
3108.Xr sysctl 8 ,
3109.Xr syslogd 8
3110.Sh HISTORY
3111The
3112.Nm
3113utility first appeared in
3114.Fx 2.0 .
3115.Nm dummynet
3116was introduced in
3117.Fx 2.2.8 .
3118Stateful extensions were introduced in
3119.Fx 4.0 .
3120.Nm ipfw2
3121was introduced in Summer 2002.
3122.Sh AUTHORS
3123.An Ugen J. S. Antsilevich ,
3124.An Poul-Henning Kamp ,
3125.An Alex Nash ,
3126.An Archie Cobbs ,
3127.An Luigi Rizzo .
3128.Pp
3129.An -nosplit
3130API based upon code written by
3131.An Daniel Boulet
3132for BSDI.
3133.Pp
3134Dummynet has been introduced by Luigi Rizzo in 1997-1998.
3135.Pp
3136Some early work (1999-2000) on the
3137.Nm dummynet
3138traffic shaper supported by Akamba Corp.
3139.Pp
3140The ipfw core (ipfw2) has been completely redesigned and
3141reimplemented by Luigi Rizzo in summer 2002. Further
3142actions and
3143options have been added by various developer over the years.
3144.Pp
3145.An -nosplit
3146In-kernel NAT support written by
3147.An Paolo Pisati Aq piso@FreeBSD.org
3148as part of a Summer of Code 2005 project.
3149.Pp
3150SCTP
3151.Nm nat
3152support has been developed by
3153.An The Centre for Advanced Internet Architectures (CAIA) Aq http://www.caia.swin.edu.au .
3154The primary developers and maintainers are David Hayes and Jason But.
3155For further information visit:
3156.Aq http://www.caia.swin.edu.au/urp/SONATA
3157.Pp
3158Delay profiles have been developed by Alessandro Cerri and
3159Luigi Rizzo, supported by the
3160European Commission within Projects Onelab and Onelab2.
3161.Sh BUGS
3162The syntax has grown over the years and sometimes it might be confusing.
3163Unfortunately, backward compatibility prevents cleaning up mistakes
3164made in the definition of the syntax.
3165.Pp
3166.Em !!! WARNING !!!
3167.Pp
3168Misconfiguring the firewall can put your computer in an unusable state,
3169possibly shutting down network services and requiring console access to
3170regain control of it.
3171.Pp
3172Incoming packet fragments diverted by
3173.Cm divert
3174are reassembled before delivery to the socket.
3175The action used on those packet is the one from the
3176rule which matches the first fragment of the packet.
3177.Pp
3178Packets diverted to userland, and then reinserted by a userland process
3179may lose various packet attributes.
3180The packet source interface name
3181will be preserved if it is shorter than 8 bytes and the userland process
3182saves and reuses the sockaddr_in
3183(as does
3184.Xr natd 8 ) ;
3185otherwise, it may be lost.
3186If a packet is reinserted in this manner, later rules may be incorrectly
3187applied, making the order of
3188.Cm divert
3189rules in the rule sequence very important.
3190.Pp
3191Dummynet drops all packets with IPv6 link-local addresses.
3192.Pp
3193Rules using
3194.Cm uid
3195or
3196.Cm gid
3197may not behave as expected.
3198In particular, incoming SYN packets may
3199have no uid or gid associated with them since they do not yet belong
3200to a TCP connection, and the uid/gid associated with a packet may not
3201be as expected if the associated process calls
3202.Xr setuid 2
3203or similar system calls.
3204.Pp
3205Rule syntax is subject to the command line environment and some patterns
3206may need to be escaped with the backslash character
3207or quoted appropriately.
3208.Pp
3209Due to the architecture of
3210.Xr libalias 3 ,
3211ipfw nat is not compatible with the TCP segmentation offloading (TSO).
3212Thus, to reliably nat your network traffic, please disable TSO
3213on your NICs using
3214.Xr ifconfig 8 .
3215.Pp
3216ICMP error messages are not implicitly matched by dynamic rules
3217for the respective conversations.
3218To avoid failures of network error detection and path MTU discovery,
3219ICMP error messages may need to be allowed explicitly through static
3220rules.
3221