xref: /netbsd-src/dist/pf/share/man/man5/pf.conf.5 (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1.\"	$NetBSD: pf.conf.5,v 1.16 2009/10/14 17:44:25 joerg Exp $
2.\"	$OpenBSD: pf.conf.5,v 1.383 2007/07/17 16:27:38 jmc Exp $
3.\"
4.\" Copyright (c) 2002, Daniel Hartmeier
5.\" All rights reserved.
6.\"
7.\" Redistribution and use in source and binary forms, with or without
8.\" modification, are permitted provided that the following conditions
9.\" are met:
10.\"
11.\"    - Redistributions of source code must retain the above copyright
12.\"      notice, this list of conditions and the following disclaimer.
13.\"    - Redistributions in binary form must reproduce the above
14.\"      copyright notice, this list of conditions and the following
15.\"      disclaimer in the documentation and/or other materials provided
16.\"      with the distribution.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19.\" "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20.\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21.\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22.\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24.\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25.\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28.\" ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29.\" POSSIBILITY OF SUCH DAMAGE.
30.\"
31.Dd June 26, 2007
32.Dt PF.CONF 5
33.Os
34.Sh NAME
35.Nm pf.conf
36.Nd packet filter configuration file
37.Sh DESCRIPTION
38The
39.Xr pf 4
40packet filter modifies, drops or passes packets according to rules or
41definitions specified in
42.Nm pf.conf .
43.Sh STATEMENT ORDER
44There are seven types of statements in
45.Nm pf.conf :
46.Bl -tag -width xxxx
47.It Cm Macros
48User-defined variables may be defined and used later, simplifying
49the configuration file.
50Macros must be defined before they are referenced in
51.Nm pf.conf .
52.It Cm Tables
53Tables provide a mechanism for increasing the performance and flexibility of
54rules with large numbers of source or destination addresses.
55.It Cm Options
56Options tune the behaviour of the packet filtering engine.
57.It Cm Traffic Normalization Li (e.g. Em scrub )
58Traffic normalization protects internal machines against inconsistencies
59in Internet protocols and implementations.
60.It Cm Queueing
61Queueing provides rule-based bandwidth control.
62.It Cm Translation Li (Various forms of NAT)
63Translation rules specify how addresses are to be mapped or redirected to
64other addresses.
65.It Cm Packet Filtering
66Packet filtering provides rule-based blocking or passing of packets.
67.El
68.Pp
69With the exception of
70.Cm macros
71and
72.Cm tables ,
73the types of statements should be grouped and appear in
74.Nm pf.conf
75in the order shown above, as this matches the operation of the underlying
76packet filtering engine.
77By default
78.Xr pfctl 8
79enforces this order (see
80.Ar set require-order
81below).
82.Sh MACROS
83Macros can be defined that will later be expanded in context.
84Macro names must start with a letter, and may contain letters, digits
85and underscores.
86Macro names may not be reserved words (for example
87.Ar pass ,
88.Ar in ,
89.Ar out ) .
90Macros are not expanded inside quotes.
91.Pp
92For example,
93.Bd -literal -offset indent
94ext_if = \&"kue0\&"
95all_ifs = \&"{\&" $ext_if lo0 \&"}\&"
96pass out on $ext_if from any to any
97pass in  on $ext_if proto tcp from any to any port 25
98.Ed
99.Sh TABLES
100Tables are named structures which can hold a collection of addresses and
101networks.
102Lookups against tables in
103.Xr pf 4
104are relatively fast, making a single rule with tables much more efficient,
105in terms of
106processor usage and memory consumption, than a large number of rules which
107differ only in IP address (either created explicitly or automatically by rule
108expansion).
109.Pp
110Tables can be used as the source or destination of filter rules,
111.Ar scrub
112rules
113or
114translation rules such as
115.Ar nat
116or
117.Ar rdr
118(see below for details on the various rule types).
119Tables can also be used for the redirect address of
120.Ar nat
121and
122.Ar rdr
123rules and in the routing options of filter rules, but only for
124.Ar round-robin
125pools.
126.Pp
127Tables can be defined with any of the following
128.Xr pfctl 8
129mechanisms.
130As with macros, reserved words may not be used as table names.
131.Bl -tag -width "manually"
132.It Ar manually
133Persistent tables can be manually created with the
134.Ar add
135or
136.Ar replace
137option of
138.Xr pfctl 8 ,
139before or after the ruleset has been loaded.
140.It Pa pf.conf
141Table definitions can be placed directly in this file, and loaded at the
142same time as other rules are loaded, atomically.
143Table definitions inside
144.Nm pf.conf
145use the
146.Ar table
147statement, and are especially useful to define non-persistent tables.
148The contents of a pre-existing table defined without a list of addresses
149to initialize it is not altered when
150.Nm pf.conf
151is loaded.
152A table initialized with the empty list,
153.Li { } ,
154will be cleared on load.
155.El
156.Pp
157Tables may be defined with the following two attributes:
158.Bl -tag -width persist
159.It Ar persist
160The
161.Ar persist
162flag forces the kernel to keep the table even when no rules refer to it.
163If the flag is not set, the kernel will automatically remove the table
164when the last rule referring to it is flushed.
165.It Ar const
166The
167.Ar const
168flag prevents the user from altering the contents of the table once it
169has been created.
170Without that flag,
171.Xr pfctl 8
172can be used to add or remove addresses from the table at any time, even
173when running with securelevel = 2.
174.El
175.Pp
176For example,
177.Bd -literal -offset indent
178table \*(Ltprivate\*(Gt const { 10/8, 172.16/12, 192.168/16 }
179table \*(Ltbadhosts\*(Gt persist
180block on fxp0 from { \*(Ltprivate\*(Gt, \*(Ltbadhosts\*(Gt } to any
181.Ed
182.Pp
183creates a table called private, to hold RFC 1918 private network
184blocks, and a table called badhosts, which is initially empty.
185A filter rule is set up to block all traffic coming from addresses listed in
186either table.
187The private table cannot have its contents changed and the badhosts table
188will exist even when no active filter rules reference it.
189Addresses may later be added to the badhosts table, so that traffic from
190these hosts can be blocked by using
191.Bd -literal -offset indent
192# pfctl -t badhosts -Tadd 204.92.77.111
193.Ed
194.Pp
195A table can also be initialized with an address list specified in one or more
196external files, using the following syntax:
197.Bd -literal -offset indent
198table \*(Ltspam\*(Gt persist file \&"/etc/spammers\&" file \&"/etc/openrelays\&"
199block on fxp0 from \*(Ltspam\*(Gt to any
200.Ed
201.Pp
202The files
203.Pa /etc/spammers
204and
205.Pa /etc/openrelays
206list IP addresses, one per line.
207Any lines beginning with a # are treated as comments and ignored.
208In addition to being specified by IP address, hosts may also be
209specified by their hostname.
210When the resolver is called to add a hostname to a table,
211.Em all
212resulting IPv4 and IPv6 addresses are placed into the table.
213IP addresses can also be entered in a table by specifying a valid interface
214name, a valid interface group or the
215.Em self
216keyword, in which case all addresses assigned to the interface(s) will be
217added to the table.
218.Sh OPTIONS
219.Xr pf 4
220may be tuned for various situations using the
221.Ar set
222command.
223.Bl -tag -width xxxx
224.It Ar set timeout
225.Pp
226.Bl -tag -width "src.track" -compact
227.It Ar interval
228Interval between purging expired states and fragments.
229.It Ar frag
230Seconds before an unassembled fragment is expired.
231.It Ar src.track
232Length of time to retain a source tracking entry after the last state
233expires.
234.El
235.Pp
236When a packet matches a stateful connection, the seconds to live for the
237connection will be updated to that of the
238.Ar proto.modifier
239which corresponds to the connection state.
240Each packet which matches this state will reset the TTL.
241Tuning these values may improve the performance of the
242firewall at the risk of dropping valid idle connections.
243.Pp
244.Bl -tag -width xxxx -compact
245.It Ar tcp.first
246The state after the first packet.
247.It Ar tcp.opening
248The state before the destination host ever sends a packet.
249.It Ar tcp.established
250The fully established state.
251.It Ar tcp.closing
252The state after the first FIN has been sent.
253.It Ar tcp.finwait
254The state after both FINs have been exchanged and the connection is closed.
255Some hosts (notably web servers on Solaris) send TCP packets even after closing
256the connection.
257Increasing
258.Ar tcp.finwait
259(and possibly
260.Ar tcp.closing )
261can prevent blocking of such packets.
262.It Ar tcp.closed
263The state after one endpoint sends an RST.
264.El
265.Pp
266ICMP and UDP are handled in a fashion similar to TCP, but with a much more
267limited set of states:
268.Pp
269.Bl -tag -width xxxx -compact
270.It Ar udp.first
271The state after the first packet.
272.It Ar udp.single
273The state if the source host sends more than one packet but the destination
274host has never sent one back.
275.It Ar udp.multiple
276The state if both hosts have sent packets.
277.It Ar icmp.first
278The state after the first packet.
279.It Ar icmp.error
280The state after an ICMP error came back in response to an ICMP packet.
281.El
282.Pp
283Other protocols are handled similarly to UDP:
284.Pp
285.Bl -tag -width xxxx -compact
286.It Ar other.first
287.It Ar other.single
288.It Ar other.multiple
289.El
290.Pp
291Timeout values can be reduced adaptively as the number of state table
292entries grows.
293.Pp
294.Bl -tag -width xxxx -compact
295.It Ar adaptive.start
296When the number of state entries exceeds this value, adaptive scaling
297begins.
298All timeout values are scaled linearly with factor
299(adaptive.end - number of states) / (adaptive.end - adaptive.start).
300.It Ar adaptive.end
301When reaching this number of state entries, all timeout values become
302zero, effectively purging all state entries immediately.
303This value is used to define the scale factor, it should not actually
304be reached (set a lower state limit, see below).
305.El
306.Pp
307Adaptive timeouts are enabled by default, with an adaptive.start value
308equal to 60% of the state limit, and an adaptive.end value equal to
309120% of the state limit.
310They can be disabled by setting both adaptive.start and adaptive.end to 0.
311.Pp
312The adaptive timeout values can be defined both globally and for each rule.
313When used on a per-rule basis, the values relate to the number of
314states created by the rule, otherwise to the total number of
315states.
316.Pp
317For example:
318.Bd -literal -offset indent
319set timeout tcp.first 120
320set timeout tcp.established 86400
321set timeout { adaptive.start 6000, adaptive.end 12000 }
322set limit states 10000
323.Ed
324.Pp
325With 9000 state table entries, the timeout values are scaled to 50%
326(tcp.first 60, tcp.established 43200).
327.Pp
328.It Ar set loginterface
329Enable collection of packet and byte count statistics for the given interface.
330These statistics can be viewed using
331.Bd -literal -offset indent
332# pfctl -s info
333.Ed
334.Pp
335In this example
336.Xr pf 4
337collects statistics on the interface named dc0:
338.Bd -literal -offset indent
339set loginterface dc0
340.Ed
341.Pp
342One can disable the loginterface using:
343.Bd -literal -offset indent
344set loginterface none
345.Ed
346.Pp
347.It Ar set limit
348Sets hard limits on the memory pools used by the packet filter.
349See
350.Xr pool 9
351for an explanation of memory pools.
352.Pp
353For example,
354.Bd -literal -offset indent
355set limit states 20000
356.Ed
357.Pp
358sets the maximum number of entries in the memory pool used by state table
359entries (generated by
360.Ar pass
361rules which do not specify
362.Ar no state )
363to 20000.
364Using
365.Bd -literal -offset indent
366set limit frags 20000
367.Ed
368.Pp
369sets the maximum number of entries in the memory pool used for fragment
370reassembly (generated by
371.Ar scrub
372rules) to 20000.
373Using
374.Bd -literal -offset indent
375set limit src-nodes 2000
376.Ed
377.Pp
378sets the maximum number of entries in the memory pool used for tracking
379source IP addresses (generated by the
380.Ar sticky-address
381and
382.Ar src.track
383options) to 2000.
384Using
385.Bd -literal -offset indent
386set limit tables 1000
387set limit table-entries 100000
388.Ed
389.Pp
390sets limits on the memory pools used by tables.
391The first limits the number of tables that can exist to 1000.
392The second limits the overall number of addresses that can be stored
393in tables to 100000.
394.Pp
395Various limits can be combined on a single line:
396.Bd -literal -offset indent
397set limit { states 20000, frags 20000, src-nodes 2000 }
398.Ed
399.Pp
400.It Ar set ruleset-optimization
401.Bl -tag -width xxxxxxxx -compact
402.It Ar none
403Disable the ruleset optimizer.
404.It Ar basic
405Enable basic ruleset optimization.
406This is the default behaviour.
407Basic ruleset optimization does four things to improve the
408performance of ruleset evaluations:
409.Pp
410.Bl -enum -compact
411.It
412remove duplicate rules
413.It
414remove rules that are a subset of another rule
415.It
416combine multiple rules into a table when advantageous
417.It
418re-order the rules to improve evaluation performance
419.El
420.Pp
421.It Ar profile
422Uses the currently loaded ruleset as a feedback profile to tailor the
423ordering of quick rules to actual network traffic.
424.El
425.Pp
426It is important to note that the ruleset optimizer will modify the ruleset
427to improve performance.
428A side effect of the ruleset modification is that per-rule accounting
429statistics will have different meanings than before.
430If per-rule accounting is important for billing purposes or whatnot,
431either the ruleset optimizer should not be used or a label field should
432be added to all of the accounting rules to act as optimization barriers.
433.Pp
434Optimization can also be set as a command-line argument to
435.Xr pfctl 8 ,
436overriding the settings in
437.Nm .
438.It Ar set optimization
439Optimize state timeouts for one of the following network environments:
440.Pp
441.Bl -tag -width xxxx -compact
442.It Ar normal
443A normal network environment.
444Suitable for almost all networks.
445.It Ar high-latency
446A high-latency environment (such as a satellite connection).
447.It Ar satellite
448Alias for
449.Ar high-latency .
450.It Ar aggressive
451Aggressively expire connections.
452This can greatly reduce the memory usage of the firewall at the cost of
453dropping idle connections early.
454.It Ar conservative
455Extremely conservative settings.
456Avoid dropping legitimate connections at the
457expense of greater memory utilization (possibly much greater on a busy
458network) and slightly increased processor utilization.
459.El
460.Pp
461For example:
462.Bd -literal -offset indent
463set optimization aggressive
464.Ed
465.Pp
466.It Ar set block-policy
467The
468.Ar block-policy
469option sets the default behaviour for the packet
470.Ar block
471action:
472.Pp
473.Bl -tag -width xxxxxxxx -compact
474.It Ar drop
475Packet is silently dropped.
476.It Ar return
477A TCP RST is returned for blocked TCP packets,
478an ICMP UNREACHABLE is returned for blocked UDP packets,
479and all other packets are silently dropped.
480.El
481.Pp
482For example:
483.Bd -literal -offset indent
484set block-policy return
485.Ed
486.It Ar set state-policy
487The
488.Ar state-policy
489option sets the default behaviour for states:
490.Pp
491.Bl -tag -width group-bound -compact
492.It Ar if-bound
493States are bound to interface.
494.It Ar floating
495States can match packets on any interfaces (the default).
496.El
497.Pp
498For example:
499.Bd -literal -offset indent
500set state-policy if-bound
501.Ed
502.It Ar set hostid
503The 32-bit
504.Ar hostid
505identifies this firewall's state table entries to other firewalls
506in a
507.Xr pfsync 4
508failover cluster.
509By default the hostid is set to a pseudo-random value, however it may be
510desirable to manually configure it, for example to more easily identify the
511source of state table entries.
512.Bd -literal -offset indent
513set hostid 1
514.Ed
515.Pp
516The hostid may be specified in either decimal or hexadecimal.
517.It Ar set require-order
518By default
519.Xr pfctl 8
520enforces an ordering of the statement types in the ruleset to:
521.Em options ,
522.Em normalization ,
523.Em queueing ,
524.Em translation ,
525.Em filtering .
526Setting this option to
527.Ar no
528disables this enforcement.
529There may be non-trivial and non-obvious implications to an out of
530order ruleset.
531Consider carefully before disabling the order enforcement.
532.It Ar set fingerprints
533Load fingerprints of known operating systems from the given filename.
534By default fingerprints of known operating systems are automatically
535loaded from
536.Xr pf.os 5
537in
538.Pa /etc
539but can be overridden via this option.
540Setting this option may leave a small period of time where the fingerprints
541referenced by the currently active ruleset are inconsistent until the new
542ruleset finishes loading.
543.Pp
544For example:
545.Pp
546.Dl set fingerprints \&"/etc/pf.os.devel\&"
547.Pp
548.It Ar set skip on Aq Ar ifspec
549List interfaces for which packets should not be filtered.
550Packets passing in or out on such interfaces are passed as if pf was
551disabled, i.e. pf does not process them in any way.
552This can be useful on loopback and other virtual interfaces, when
553packet filtering is not desired and can have unexpected effects.
554For example:
555.Pp
556.Dl set skip on lo0
557.Pp
558.It Ar set debug
559Set the debug
560.Ar level
561to one of the following:
562.Pp
563.Bl -tag -width xxxxxxxxxxxx -compact
564.It Ar none
565Don't generate debug messages.
566.It Ar urgent
567Generate debug messages only for serious errors.
568.It Ar misc
569Generate debug messages for various errors.
570.It Ar loud
571Generate debug messages for common conditions.
572.El
573.El
574.Sh TRAFFIC NORMALIZATION
575Traffic normalization is used to sanitize packet content in such
576a way that there are no ambiguities in packet interpretation on
577the receiving side.
578The normalizer does IP fragment reassembly to prevent attacks
579that confuse intrusion detection systems by sending overlapping
580IP fragments.
581Packet normalization is invoked with the
582.Ar scrub
583directive.
584.Pp
585.Ar scrub
586has the following options:
587.Bl -tag -width xxxx
588.It Ar no-df
589Clears the
590.Ar dont-fragment
591bit from a matching IP packet.
592Some operating systems are known to generate fragmented packets with the
593.Ar dont-fragment
594bit set.
595This is particularly true with NFS.
596.Ar Scrub
597will drop such fragmented
598.Ar dont-fragment
599packets unless
600.Ar no-df
601is specified.
602.Pp
603Unfortunately some operating systems also generate their
604.Ar dont-fragment
605packets with a zero IP identification field.
606Clearing the
607.Ar dont-fragment
608bit on packets with a zero IP ID may cause deleterious results if an
609upstream router later fragments the packet.
610Using the
611.Ar random-id
612modifier (see below) is recommended in combination with the
613.Ar no-df
614modifier to ensure unique IP identifiers.
615.It Ar min-ttl Aq Ar number
616Enforces a minimum TTL for matching IP packets.
617.It Ar max-mss Aq Ar number
618Enforces a maximum MSS for matching TCP packets.
619.It Ar random-id
620Replaces the IP identification field with random values to compensate
621for predictable values generated by many hosts.
622This option only applies to packets that are not fragmented
623after the optional fragment reassembly.
624.It Ar fragment reassemble
625Using
626.Ar scrub
627rules, fragments can be reassembled by normalization.
628In this case, fragments are buffered until they form a complete
629packet, and only the completed packet is passed on to the filter.
630The advantage is that filter rules have to deal only with complete
631packets, and can ignore fragments.
632The drawback of caching fragments is the additional memory cost.
633But the full reassembly method is the only method that currently works
634with NAT.
635This is the default behavior of a
636.Ar scrub
637rule if no fragmentation modifier is supplied.
638.It Ar fragment crop
639The default fragment reassembly method is expensive, hence the option
640to crop is provided.
641In this case,
642.Xr pf 4
643will track the fragments and cache a small range descriptor.
644Duplicate fragments are dropped and overlaps are cropped.
645Thus data will only occur once on the wire with ambiguities resolving to
646the first occurrence.
647Unlike the
648.Ar fragment reassemble
649modifier, fragments are not buffered, they are passed as soon as they
650are received.
651The
652.Ar fragment crop
653reassembly mechanism does not yet work with NAT.
654.Pp
655.It Ar fragment drop-ovl
656This option is similar to the
657.Ar fragment crop
658modifier except that all overlapping or duplicate fragments will be
659dropped, and all further corresponding fragments will be
660dropped as well.
661.It Ar reassemble tcp
662Statefully normalizes TCP connections.
663.Ar scrub reassemble tcp
664rules may not have the direction (in/out) specified.
665.Ar reassemble tcp
666performs the following normalizations:
667.Pp
668.Bl -tag -width timeout -compact
669.It ttl
670Neither side of the connection is allowed to reduce their IP TTL.
671An attacker may send a packet such that it reaches the firewall, affects
672the firewall state, and expires before reaching the destination host.
673.Ar reassemble tcp
674will raise the TTL of all packets back up to the highest value seen on
675the connection.
676.It timestamp modulation
677Modern TCP stacks will send a timestamp on every TCP packet and echo
678the other endpoint's timestamp back to them.
679Many operating systems will merely start the timestamp at zero when
680first booted, and increment it several times a second.
681The uptime of the host can be deduced by reading the timestamp and multiplying
682by a constant.
683Also observing several different timestamps can be used to count hosts
684behind a NAT device.
685And spoofing TCP packets into a connection requires knowing or guessing
686valid timestamps.
687Timestamps merely need to be monotonically increasing and not derived off a
688guessable base time.
689.Ar reassemble tcp
690will cause
691.Ar scrub
692to modulate the TCP timestamps with a random number.
693.It extended PAWS checks
694There is a problem with TCP on long fat pipes, in that a packet might get
695delayed for longer than it takes the connection to wrap its 32-bit sequence
696space.
697In such an occurrence, the old packet would be indistinguishable from a
698new packet and would be accepted as such.
699The solution to this is called PAWS: Protection Against Wrapped Sequence
700numbers.
701It protects against it by making sure the timestamp on each packet does
702not go backwards.
703.Ar reassemble tcp
704also makes sure the timestamp on the packet does not go forward more
705than the RFC allows.
706By doing this,
707.Xr pf 4
708artificially extends the security of TCP sequence numbers by 10 to 18
709bits when the host uses appropriately randomized timestamps, since a
710blind attacker would have to guess the timestamp as well.
711.El
712.El
713.Pp
714For example,
715.Bd -literal -offset indent
716scrub in on $ext_if all fragment reassemble
717.Ed
718.Pp
719The
720.Ar no
721option prefixed to a scrub rule causes matching packets to remain unscrubbed,
722much in the same way as
723.Ar drop quick
724works in the packet filter (see below).
725This mechanism should be used when it is necessary to exclude specific packets
726from broader scrub rules.
727.Sh QUEUEING
728Packets can be assigned to queues for the purpose of bandwidth
729control.
730At least two declarations are required to configure queues, and later
731any packet filtering rule can reference the defined queues by name.
732During the filtering component of
733.Nm pf.conf ,
734the last referenced
735.Ar queue
736name is where any packets from
737.Ar pass
738rules will be queued, while for
739.Ar block
740rules it specifies where any resulting ICMP or TCP RST
741packets should be queued.
742The
743.Ar scheduler
744defines the algorithm used to decide which packets get delayed, dropped, or
745sent out immediately.
746There are three
747.Ar schedulers
748currently supported.
749.Bl -tag -width xxxx
750.It Ar cbq
751Class Based Queueing.
752.Ar Queues
753attached to an interface build a tree, thus each
754.Ar queue
755can have further child
756.Ar queues .
757Each queue can have a
758.Ar priority
759and a
760.Ar bandwidth
761assigned.
762.Ar Priority
763mainly controls the time packets take to get sent out, while
764.Ar bandwidth
765has primarily effects on throughput.
766.Ar cbq
767achieves both partitioning and sharing of link bandwidth
768by hierarchically structured classes.
769Each class has its own
770.Ar queue
771and is assigned its share of
772.Ar bandwidth .
773A child class can borrow bandwidth from its parent class
774as long as excess bandwidth is available
775(see the option
776.Ar borrow ,
777below).
778.It Ar priq
779Priority Queueing.
780.Ar Queues
781are flat attached to the interface, thus,
782.Ar queues
783cannot have further child
784.Ar queues .
785Each
786.Ar queue
787has a unique
788.Ar priority
789assigned, ranging from 0 to 15.
790Packets in the
791.Ar queue
792with the highest
793.Ar priority
794are processed first.
795.It Ar hfsc
796Hierarchical Fair Service Curve.
797.Ar Queues
798attached to an interface build a tree, thus each
799.Ar queue
800can have further child
801.Ar queues .
802Each queue can have a
803.Ar priority
804and a
805.Ar bandwidth
806assigned.
807.Ar Priority
808mainly controls the time packets take to get sent out, while
809.Ar bandwidth
810has primarily effects on throughput.
811.Ar hfsc
812supports both link-sharing and guaranteed real-time services.
813It employs a service curve based QoS model,
814and its unique feature is an ability to decouple
815.Ar delay
816and
817.Ar bandwidth
818allocation.
819.El
820.Pp
821The interfaces on which queueing should be activated are declared using
822the
823.Ar altq on
824declaration.
825.Ar altq on
826has the following keywords:
827.Bl -tag -width xxxx
828.It Aq Ar interface
829Queueing is enabled on the named interface.
830.It Aq Ar scheduler
831Specifies which queueing scheduler to use.
832Currently supported values
833are
834.Ar cbq
835for Class Based Queueing,
836.Ar priq
837for Priority Queueing and
838.Ar hfsc
839for the Hierarchical Fair Service Curve scheduler.
840.It Ar bandwidth Aq Ar bw
841The maximum bitrate for all queues on an
842interface may be specified using the
843.Ar bandwidth
844keyword.
845The value can be specified as an absolute value or as a
846percentage of the interface bandwidth.
847When using an absolute value, the suffixes
848.Ar b ,
849.Ar Kb ,
850.Ar Mb ,
851and
852.Ar Gb
853are used to represent bits, kilobits, megabits, and
854gigabits per second, respectively.
855The value must not exceed the interface bandwidth.
856If
857.Ar bandwidth
858is not specified, the interface bandwidth is used
859(but take note that some interfaces do not know their bandwidth,
860or can adapt their bandwidth rates).
861.It Ar qlimit Aq Ar limit
862The maximum number of packets held in the queue.
863The default is 50.
864.It Ar tbrsize Aq Ar size
865Adjusts the size, in bytes, of the token bucket regulator.
866If not specified, heuristics based on the
867interface bandwidth are used to determine the size.
868.It Ar queue Aq Ar list
869Defines a list of subqueues to create on an interface.
870.El
871.Pp
872In the following example, the interface dc0
873should queue up to 5 Mbit/s in four second-level queues using
874Class Based Queueing.
875Those four queues will be shown in a later example.
876.Bd -literal -offset indent
877altq on dc0 cbq bandwidth 5Mb queue { std, http, mail, ssh }
878.Ed
879.Pp
880Once interfaces are activated for queueing using the
881.Ar altq
882directive, a sequence of
883.Ar queue
884directives may be defined.
885The name associated with a
886.Ar queue
887must match a queue defined in the
888.Ar altq
889directive (e.g. mail), or, except for the
890.Ar priq
891.Ar scheduler ,
892in a parent
893.Ar queue
894declaration.
895The following keywords can be used:
896.Bl -tag -width xxxx
897.It Ar on Aq Ar interface
898Specifies the interface the queue operates on.
899If not given, it operates on all matching interfaces.
900.It Ar bandwidth Aq Ar bw
901Specifies the maximum bitrate to be processed by the queue.
902This value must not exceed the value of the parent
903.Ar queue
904and can be specified as an absolute value or a percentage of the parent
905queue's bandwidth.
906If not specified, defaults to 100% of the parent queue's bandwidth.
907The
908.Ar priq
909scheduler does not support bandwidth specification.
910.It Ar priority Aq Ar level
911Between queues a priority level can be set.
912For
913.Ar cbq
914and
915.Ar hfsc ,
916the range is 0 to 7 and for
917.Ar priq ,
918the range is 0 to 15.
919The default for all is 1.
920.Ar Priq
921queues with a higher priority are always served first.
922.Ar Cbq
923and
924.Ar Hfsc
925queues with a higher priority are preferred in the case of overload.
926.It Ar qlimit Aq Ar limit
927The maximum number of packets held in the queue.
928The default is 50.
929.El
930.Pp
931The
932.Ar scheduler
933can get additional parameters with
934.Xo Aq Ar scheduler
935.Pf ( Aq Ar parameters ) .
936.Xc
937Parameters are as follows:
938.Bl -tag -width Fl
939.It Ar default
940Packets not matched by another queue are assigned to this one.
941Exactly one default queue is required.
942.It Ar red
943Enable RED (Random Early Detection) on this queue.
944RED drops packets with a probability proportional to the average
945queue length.
946.It Ar rio
947Enables RIO on this queue.
948RIO is RED with IN/OUT, thus running
949RED two times more than RIO would achieve the same effect.
950RIO is currently not supported in the GENERIC kernel.
951.It Ar ecn
952Enables ECN (Explicit Congestion Notification) on this queue.
953ECN implies RED.
954.El
955.Pp
956The
957.Ar cbq
958.Ar scheduler
959supports an additional option:
960.Bl -tag -width Fl
961.It Ar borrow
962The queue can borrow bandwidth from the parent.
963.El
964.Pp
965The
966.Ar hfsc
967.Ar scheduler
968supports some additional options:
969.Bl -tag -width Fl
970.It Ar realtime Aq Ar sc
971The minimum required bandwidth for the queue.
972.It Ar upperlimit Aq Ar sc
973The maximum allowed bandwidth for the queue.
974.It Ar linkshare Aq Ar sc
975The bandwidth share of a backlogged queue.
976.El
977.Pp
978.Aq Ar sc
979is an acronym for
980.Ar service curve .
981.Pp
982The format for service curve specifications is
983.Ar ( m1 , d , m2 ) .
984.Ar m2
985controls the bandwidth assigned to the queue.
986.Ar m1
987and
988.Ar d
989are optional and can be used to control the initial bandwidth assignment.
990For the first
991.Ar d
992milliseconds the queue gets the bandwidth given as
993.Ar m1 ,
994afterwards the value given in
995.Ar m2 .
996.Pp
997Furthermore, with
998.Ar cbq
999and
1000.Ar hfsc ,
1001child queues can be specified as in an
1002.Ar altq
1003declaration, thus building a tree of queues using a part of
1004their parent's bandwidth.
1005.Pp
1006Packets can be assigned to queues based on filter rules by using the
1007.Ar queue
1008keyword.
1009Normally only one
1010.Ar queue
1011is specified; when a second one is specified it will instead be used for
1012packets which have a
1013.Em TOS
1014of
1015.Em lowdelay
1016and for TCP ACKs with no data payload.
1017.Pp
1018To continue the previous example, the examples below would specify the
1019four referenced
1020queues, plus a few child queues.
1021Interactive
1022.Xr ssh 1
1023sessions get priority over bulk transfers like
1024.Xr scp 1
1025and
1026.Xr sftp 1 .
1027The queues may then be referenced by filtering rules (see
1028.Sx PACKET FILTERING
1029below).
1030.Bd -literal
1031queue std bandwidth 10% cbq(default)
1032queue http bandwidth 60% priority 2 cbq(borrow red) \e
1033      { employees, developers }
1034queue  developers bandwidth 75% cbq(borrow)
1035queue  employees bandwidth 15%
1036queue mail bandwidth 10% priority 0 cbq(borrow ecn)
1037queue ssh bandwidth 20% cbq(borrow) { ssh_interactive, ssh_bulk }
1038queue  ssh_interactive bandwidth 50% priority 7 cbq(borrow)
1039queue  ssh_bulk bandwidth 50% priority 0 cbq(borrow)
1040
1041block return out on dc0 inet all queue std
1042pass out on dc0 inet proto tcp from $developerhosts to any port 80 \e
1043      queue developers
1044pass out on dc0 inet proto tcp from $employeehosts to any port 80 \e
1045      queue employees
1046pass out on dc0 inet proto tcp from any to any port 22 \e
1047      queue(ssh_bulk, ssh_interactive)
1048pass out on dc0 inet proto tcp from any to any port 25 \e
1049      queue mail
1050.Ed
1051.Sh TRANSLATION
1052Translation rules modify either the source or destination address of the
1053packets associated with a stateful connection.
1054A stateful connection is automatically created to track packets matching
1055such a rule as long as they are not blocked by the filtering section of
1056.Nm pf.conf .
1057The translation engine modifies the specified address and/or port in the
1058packet, recalculates IP, TCP and UDP checksums as necessary, and passes it to
1059the packet filter for evaluation.
1060.Pp
1061Since translation occurs before filtering the filter
1062engine will see packets as they look after any
1063addresses and ports have been translated.
1064Filter rules will therefore have to filter based on the translated
1065address and port number.
1066Packets that match a translation rule are only automatically passed if
1067the
1068.Ar pass
1069modifier is given, otherwise they are
1070still subject to
1071.Ar block
1072and
1073.Ar pass
1074rules.
1075.Pp
1076The state entry created permits
1077.Xr pf 4
1078to keep track of the original address for traffic associated with that state
1079and correctly direct return traffic for that connection.
1080.Pp
1081Various types of translation are possible with pf:
1082.Bl -tag -width xxxx
1083.It Ar binat
1084A
1085.Ar binat
1086rule specifies a bidirectional mapping between an external IP netblock
1087and an internal IP netblock.
1088.It Ar nat
1089A
1090.Ar nat
1091rule specifies that IP addresses are to be changed as the packet
1092traverses the given interface.
1093This technique allows one or more IP addresses
1094on the translating host to support network traffic for a larger range of
1095machines on an "inside" network.
1096Although in theory any IP address can be used on the inside, it is strongly
1097recommended that one of the address ranges defined by RFC 1918 be used.
1098These netblocks are:
1099.Bd -literal
110010.0.0.0 - 10.255.255.255 (all of net 10, i.e., 10/8)
1101172.16.0.0 - 172.31.255.255 (i.e., 172.16/12)
1102192.168.0.0 - 192.168.255.255 (i.e., 192.168/16)
1103.Ed
1104.It Pa rdr
1105The packet is redirected to another destination and possibly a
1106different port.
1107.Ar rdr
1108rules can optionally specify port ranges instead of single ports.
1109rdr ... port 2000:2999 -\*(Gt ... port 4000
1110redirects ports 2000 to 2999 (inclusive) to port 4000.
1111rdr ... port 2000:2999 -\*(Gt ... port 4000:*
1112redirects port 2000 to 4000, 2001 to 4001, ..., 2999 to 4999.
1113.El
1114.Pp
1115In addition to modifying the address, some translation rules may modify
1116source or destination ports for
1117.Xr tcp 4
1118or
1119.Xr udp 4
1120connections; implicitly in the case of
1121.Ar nat
1122rules and explicitly in the case of
1123.Ar rdr
1124rules.
1125Port numbers are never translated with a
1126.Ar binat
1127rule.
1128.Pp
1129Evaluation order of the translation rules is dependent on the type
1130of the translation rules and of the direction of a packet.
1131.Ar binat
1132rules are always evaluated first.
1133Then either the
1134.Ar rdr
1135rules are evaluated on an inbound packet or the
1136.Ar nat
1137rules on an outbound packet.
1138Rules of the same type are evaluated in the same order in which they
1139appear in the ruleset.
1140The first matching rule decides what action is taken.
1141.Pp
1142The
1143.Ar no
1144option prefixed to a translation rule causes packets to remain untranslated,
1145much in the same way as
1146.Ar drop quick
1147works in the packet filter (see below).
1148If no rule matches the packet it is passed to the filter engine unmodified.
1149.Pp
1150Translation rules apply only to packets that pass through
1151the specified interface, and if no interface is specified,
1152translation is applied to packets on all interfaces.
1153For instance, redirecting port 80 on an external interface to an internal
1154web server will only work for connections originating from the outside.
1155Connections to the address of the external interface from local hosts will
1156not be redirected, since such packets do not actually pass through the
1157external interface.
1158Redirections cannot reflect packets back through the interface they arrive
1159on, they can only be redirected to hosts connected to different interfaces
1160or to the firewall itself.
1161.Pp
1162Note that redirecting external incoming connections to the loopback
1163address, as in
1164.Bd -literal -offset indent
1165rdr on ne3 inet proto tcp to port spamd -\*(Gt 127.0.0.1 port smtp
1166.Ed
1167.Pp
1168will effectively allow an external host to connect to daemons
1169bound solely to the loopback address, circumventing the traditional
1170blocking of such connections on a real interface.
1171Unless this effect is desired, any of the local non-loopback addresses
1172should be used as redirection target instead, which allows external
1173connections only to daemons bound to this address or not bound to
1174any address.
1175.Pp
1176See
1177.Sx TRANSLATION EXAMPLES
1178below.
1179.Sh PACKET FILTERING
1180.Xr pf 4
1181has the ability to
1182.Ar block
1183and
1184.Ar pass
1185packets based on attributes of their layer 3 (see
1186.Xr ip 4
1187and
1188.Xr ip6 4 )
1189and layer 4 (see
1190.Xr icmp 4 ,
1191.Xr icmp6 4 ,
1192.Xr tcp 4 ,
1193.Xr udp 4 )
1194headers.
1195In addition, packets may also be
1196assigned to queues for the purpose of bandwidth control.
1197.Pp
1198For each packet processed by the packet filter, the filter rules are
1199evaluated in sequential order, from first to last.
1200The last matching rule decides what action is taken.
1201If no rule matches the packet, the default action is to pass
1202the packet.
1203.Pp
1204The following actions can be used in the filter:
1205.Bl -tag -width xxxx
1206.It Ar block
1207The packet is blocked.
1208There are a number of ways in which a
1209.Ar block
1210rule can behave when blocking a packet.
1211The default behaviour is to
1212.Ar drop
1213packets silently, however this can be overridden or made
1214explicit either globally, by setting the
1215.Ar block-policy
1216option, or on a per-rule basis with one of the following options:
1217.Pp
1218.Bl -tag -width xxxx -compact
1219.It Ar drop
1220The packet is silently dropped.
1221.It Ar return-rst
1222This applies only to
1223.Xr tcp 4
1224packets, and issues a TCP RST which closes the
1225connection.
1226.It Ar return-icmp
1227.It Ar return-icmp6
1228This causes ICMP messages to be returned for packets which match the rule.
1229By default this is an ICMP UNREACHABLE message, however this
1230can be overridden by specifying a message as a code or number.
1231.It Ar return
1232This causes a TCP RST to be returned for
1233.Xr tcp 4
1234packets and an ICMP UNREACHABLE for UDP and other packets.
1235.El
1236.Pp
1237Options returning ICMP packets currently have no effect if
1238.Xr pf 4
1239operates on a
1240.Xr bridge 4 ,
1241as the code to support this feature has not yet been implemented.
1242.Pp
1243The simplest mechanism to block everything by default and only pass
1244packets that match explicit rules is specify a first filter rule of:
1245.Bd -literal -offset indent
1246block all
1247.Ed
1248.It Ar pass
1249The packet is passed;
1250state is created unless the
1251.Ar no state
1252option is specified.
1253.El
1254.Pp
1255By default
1256.Xr pf 4
1257filters packets statefully; the first time a packet matches a
1258.Ar pass
1259rule, a state entry is created; for subsequent packets the filter checks
1260whether the packet matches any state.
1261If it does, the packet is passed without evaluation of any rules.
1262After the connection is closed or times out, the state entry is automatically
1263removed.
1264.Pp
1265This has several advantages.
1266For TCP connections, comparing a packet to a state involves checking
1267its sequence numbers, as well as TCP timestamps if a
1268.Ar scrub reassemble tcp
1269rule applies to the connection.
1270If these values are outside the narrow windows of expected
1271values, the packet is dropped.
1272This prevents spoofing attacks, such as when an attacker sends packets with
1273a fake source address/port but does not know the connection's sequence
1274numbers.
1275Similarly,
1276.Xr pf 4
1277knows how to match ICMP replies to states.
1278For example,
1279.Bd -literal -offset indent
1280pass out inet proto icmp all icmp-type echoreq
1281.Ed
1282.Pp
1283allows echo requests (such as those created by
1284.Xr ping 8 )
1285out statefully, and matches incoming echo replies correctly to states.
1286.Pp
1287Also, looking up states is usually faster than evaluating rules.
1288If there are 50 rules, all of them are evaluated sequentially in O(n).
1289Even with 50000 states, only 16 comparisons are needed to match a
1290state, since states are stored in a binary search tree that allows
1291searches in O(log2 n).
1292.Pp
1293Furthermore, correct handling of ICMP error messages is critical to
1294many protocols, particularly TCP.
1295.Xr pf 4
1296matches ICMP error messages to the correct connection, checks them against
1297connection parameters, and passes them if appropriate.
1298For example if an ICMP source quench message referring to a stateful TCP
1299connection arrives, it will be matched to the state and get passed.
1300.Pp
1301Finally, state tracking is required for
1302.Ar nat , binat No and Ar rdr
1303rules, in order to track address and port translations and reverse the
1304translation on returning packets.
1305.Pp
1306.Xr pf 4
1307will also create state for other protocols which are effectively stateless by
1308nature.
1309UDP packets are matched to states using only host addresses and ports,
1310and other protocols are matched to states using only the host addresses.
1311.Pp
1312If stateless filtering of individual packets is desired,
1313the
1314.Ar no state
1315keyword can be used to specify that state will not be created
1316if this is the last matching rule.
1317A number of parameters can also be set to affect how
1318.Xr pf 4
1319handles state tracking.
1320See
1321.Sx STATEFUL TRACKING OPTIONS
1322below for further details.
1323.Sh PARAMETERS
1324The rule parameters specify the packets to which a rule applies.
1325A packet always comes in on, or goes out through, one interface.
1326Most parameters are optional.
1327If a parameter is specified, the rule only applies to packets with
1328matching attributes.
1329Certain parameters can be expressed as lists, in which case
1330.Xr pfctl 8
1331generates all needed rule combinations.
1332.Bl -tag -width xxxx
1333.It Ar in No or Ar out
1334This rule applies to incoming or outgoing packets.
1335If neither
1336.Ar in
1337nor
1338.Ar out
1339are specified, the rule will match packets in both directions.
1340.It Ar log
1341In addition to the action specified, a log message is generated.
1342Only the packet that establishes the state is logged,
1343unless the
1344.Ar no state
1345option is specified.
1346The logged packets are sent to a
1347.Xr pflog 4
1348interface, by default
1349.Ar pflog0 .
1350This interface is monitored by the
1351.Xr pflogd 8
1352logging daemon, which dumps the logged packets to the file
1353.Pa /var/log/pflog
1354in
1355.Xr pcap 3
1356binary format.
1357.It Ar log (all)
1358Used to force logging of all packets for a connection.
1359This is not necessary when
1360.Ar no state
1361is explicitly specified.
1362As with
1363.Ar log ,
1364packets are logged to
1365.Xr pflog 4 .
1366.It Ar log (user)
1367Logs the
1368.Ux
1369user ID of the user that owns the socket and the PID of the process that
1370has the socket open where the packet is sourced from or destined to
1371(depending on which socket is local).
1372This is in addition to the normal information logged.
1373.It Ar log (to Aq Ar interface )
1374Send logs to the specified
1375.Xr pflog 4
1376interface instead of
1377.Ar pflog0 .
1378.It Ar quick
1379If a packet matches a rule which has the
1380.Ar quick
1381option set, this rule
1382is considered the last matching rule, and evaluation of subsequent rules
1383is skipped.
1384.It Ar on Aq Ar interface
1385This rule applies only to packets coming in on, or going out through, this
1386particular interface or interface group.
1387For more information on interface groups,
1388see the
1389.Ic group
1390keyword in
1391.Xr ifconfig 8 .
1392.It Aq Ar af
1393This rule applies only to packets of this address family.
1394Supported values are
1395.Ar inet
1396and
1397.Ar inet6 .
1398.It Ar proto Aq Ar protocol
1399This rule applies only to packets of this protocol.
1400Common protocols are
1401.Xr icmp 4 ,
1402.Xr icmp6 4 ,
1403.Xr tcp 4 ,
1404and
1405.Xr udp 4 .
1406For a list of all the protocol name to number mappings used by
1407.Xr pfctl 8 ,
1408see the file
1409.Em /etc/protocols .
1410.It Ar from Ao Ar source Ac Ar port Ao Ar source Ac Ar os Ao Ar source Ac \
1411Ar to Ao Ar dest Ac Ar port Aq Ar dest
1412This rule applies only to packets with the specified source and destination
1413addresses and ports.
1414.Pp
1415Addresses can be specified in CIDR notation (matching netblocks), as
1416symbolic host names, interface names or interface group names, or as any
1417of the following keywords:
1418.Pp
1419.Bl -tag -width xxxxxxxxxxxxxx -compact
1420.It Ar any
1421Any address.
1422.It Ar route Aq Ar label
1423Any address whose associated route has label
1424.Aq Ar label .
1425See
1426.Xr route 4
1427and
1428.Xr route 8 .
1429.It Ar no-route
1430Any address which is not currently routable.
1431.It Ar urpf-failed
1432Any source address that fails a unicast reverse path forwarding (URPF)
1433check, i.e. packets coming in on an interface other than that which holds
1434the route back to the packet's source address.
1435.It Aq Ar table
1436Any address that matches the given table.
1437.El
1438.Pp
1439Interface names and interface group names can have modifiers appended:
1440.Pp
1441.Bl -tag -width xxxxxxxxxxxx -compact
1442.It Ar :network
1443Translates to the network(s) attached to the interface.
1444.It Ar :broadcast
1445Translates to the interface's broadcast address(es).
1446.It Ar :peer
1447Translates to the point to point interface's peer address(es).
1448.It Ar :0
1449Do not include interface aliases.
1450.El
1451.Pp
1452Host names may also have the
1453.Ar :0
1454option appended to restrict the name resolution to the first of each
1455v4 and v6 address found.
1456.Pp
1457Host name resolution and interface to address translation are done at
1458ruleset load-time.
1459When the address of an interface (or host name) changes (under DHCP or PPP,
1460for instance), the ruleset must be reloaded for the change to be reflected
1461in the kernel.
1462Surrounding the interface name (and optional modifiers) in parentheses
1463changes this behaviour.
1464When the interface name is surrounded by parentheses, the rule is
1465automatically updated whenever the interface changes its address.
1466The ruleset does not need to be reloaded.
1467This is especially useful with
1468.Ar nat .
1469.Pp
1470Ports can be specified either by number or by name.
1471For example, port 80 can be specified as
1472.Em www .
1473For a list of all port name to number mappings used by
1474.Xr pfctl 8 ,
1475see the file
1476.Pa /etc/services .
1477.Pp
1478Ports and ranges of ports are specified by using these operators:
1479.Bd -literal -offset indent
1480=	(equal)
1481!=	(unequal)
1482\*(Lt	(less than)
1483\*(Le	(less than or equal)
1484\*(Gt	(greater than)
1485\*(Ge	(greater than or equal)
1486:	(range including boundaries)
1487\*(Gt\*(Lt	(range excluding boundaries)
1488\*(Lt\*(Gt	(except range)
1489.Ed
1490.Pp
1491.Sq \*(Gt\*(Lt ,
1492.Sq \*(Lt\*(Gt
1493and
1494.Sq \&:
1495are binary operators (they take two arguments).
1496For instance:
1497.Bl -tag -width Fl
1498.It Ar port 2000:2004
1499means
1500.Sq all ports \*(Ge 2000 and \*(Le 2004 ,
1501hence ports 2000, 2001, 2002, 2003 and 2004.
1502.It Ar port 2000 \*(Gt\*(Lt 2004
1503means
1504.Sq all ports \*(Gt 2000 and \*(Lt 2004 ,
1505hence ports 2001, 2002 and 2003.
1506.It Ar port 2000 \*(Lt\*(Gt 2004
1507means
1508.Sq all ports \*(Lt 2000 or \*(Gt 2004 ,
1509hence ports 1-1999 and 2005-65535.
1510.El
1511.Pp
1512The operating system of the source host can be specified in the case of TCP
1513rules with the
1514.Ar OS
1515modifier.
1516See the
1517.Sx OPERATING SYSTEM FINGERPRINTING
1518section for more information.
1519.Pp
1520The host, port and OS specifications are optional, as in the following examples:
1521.Bd -literal -offset indent
1522pass in all
1523pass in from any to any
1524pass in proto tcp from any port \*(Le 1024 to any
1525pass in proto tcp from any to any port 25
1526pass in proto tcp from 10.0.0.0/8 port \*(Gt 1024 \e
1527      to ! 10.1.2.3 port != ssh
1528pass in proto tcp from any os "OpenBSD"
1529pass in proto tcp from route "DTAG"
1530.Ed
1531.It Ar all
1532This is equivalent to "from any to any".
1533.It Ar group Aq Ar group
1534This functionality is not supported in this version of
1535.Nx .
1536.\" Similar to
1537.\" .Ar user ,
1538.\" this rule only applies to packets of sockets owned by the specified group.
1539.It Ar user Aq Ar user
1540This rule only applies to packets of sockets owned by the specified user.
1541For outgoing connections initiated from the firewall, this is the user
1542that opened the connection.
1543For incoming connections to the firewall itself, this is the user that
1544listens on the destination port.
1545For forwarded connections, where the firewall is not a connection endpoint,
1546the user and group are
1547.Em unknown .
1548.Pp
1549All packets, both outgoing and incoming, of one connection are associated
1550with the same user and group.
1551Only TCP and UDP packets can be associated with users; for other protocols
1552these parameters are ignored.
1553.Pp
1554User and group refer to the effective (as opposed to the real) IDs, in
1555case the socket is created by a setuid/setgid process.
1556User and group IDs are stored when a socket is created;
1557when a process creates a listening socket as root (for instance, by
1558binding to a privileged port) and subsequently changes to another
1559user ID (to drop privileges), the credentials will remain root.
1560.Pp
1561User and group IDs can be specified as either numbers or names.
1562The syntax is similar to the one for ports.
1563The value
1564.Em unknown
1565matches packets of forwarded connections.
1566.Em unknown
1567can only be used with the operators
1568.Cm =
1569and
1570.Cm != .
1571Other constructs like
1572.Cm user \*(Ge unknown
1573are invalid.
1574Forwarded packets with unknown user and group ID match only rules
1575that explicitly compare against
1576.Em unknown
1577with the operators
1578.Cm =
1579or
1580.Cm != .
1581For instance
1582.Cm user \*(Ge 0
1583does not match forwarded packets.
1584The following example allows only selected users to open outgoing
1585connections:
1586.Bd -literal -offset indent
1587block out proto { tcp, udp } all
1588pass  out proto { tcp, udp } all user { \*(Lt 1000, dhartmei }
1589.Ed
1590.It Ar flags Ao Ar a Ac Pf / Ns Ao Ar b Ac No \*(Ba / Ns \
1591Ao Ar b Ac No \*(Ba any
1592This rule only applies to TCP packets that have the flags
1593.Aq Ar a
1594set out of set
1595.Aq Ar b .
1596Flags not specified in
1597.Aq Ar b
1598are ignored.
1599For stateful connections, the default is
1600.Ar flags S/SA .
1601To indicate that flags should not be checked at all, specify
1602.Ar flags any .
1603The flags are: (F)IN, (S)YN, (R)ST, (P)USH, (A)CK, (U)RG, (E)CE, and C(W)R.
1604.Bl -tag -width Fl
1605.It Ar flags S/S
1606Flag SYN is set.
1607The other flags are ignored.
1608.It Ar flags S/SA
1609This is the default setting for stateful connections.
1610Out of SYN and ACK, exactly SYN may be set.
1611SYN, SYN+PSH and SYN+RST match, but SYN+ACK, ACK and ACK+RST do not.
1612This is more restrictive than the previous example.
1613.It Ar flags /SFRA
1614If the first set is not specified, it defaults to none.
1615All of SYN, FIN, RST and ACK must be unset.
1616.El
1617.Pp
1618Because
1619.Ar flags S/SA
1620is applied by default (unless
1621.Ar no state
1622is specified), only the initial SYN packet of a TCP handshake will create
1623a state for a TCP connection.
1624It is possible to be less restrictive, and allow state creation from
1625intermediate
1626.Pq non-SYN
1627packets, by specifying
1628.Ar flags any .
1629This will cause
1630.Xr pf 4
1631to synchronize to existing connections, for instance
1632if one flushes the state table.
1633However, states created from such intermediate packets may be missing
1634connection details such as the TCP window scaling factor.
1635States which modify the packet flow, such as those affected by
1636.Ar nat , binat No or Ar rdr
1637rules,
1638.Ar modulate No or Ar synproxy state
1639options, or scrubbed with
1640.Ar reassemble tcp
1641will also not be recoverable from intermediate packets.
1642Such connections will stall and time out.
1643.It Ar icmp-type Ao Ar type Ac Ar code Ao Ar code Ac
1644.It Ar icmp6-type Ao Ar type Ac Ar code Ao Ar code Ac
1645This rule only applies to ICMP or ICMPv6 packets with the specified type
1646and code.
1647Text names for ICMP types and codes are listed in
1648.Xr icmp 4
1649and
1650.Xr icmp6 4 .
1651This parameter is only valid for rules that cover protocols ICMP or
1652ICMP6.
1653The protocol and the ICMP type indicator
1654.Po
1655.Ar icmp-type
1656or
1657.Ar icmp6-type
1658.Pc
1659must match.
1660.It Ar tos Ao Ar string Ac No \*(Ba Aq Ar number
1661This rule applies to packets with the specified
1662.Em TOS
1663bits set.
1664.Em TOS
1665may be
1666given as one of
1667.Ar lowdelay ,
1668.Ar throughput ,
1669.Ar reliability ,
1670or as either hex or decimal.
1671.Pp
1672For example, the following rules are identical:
1673.Bd -literal -offset indent
1674pass all tos lowdelay
1675pass all tos 0x10
1676pass all tos 16
1677.Ed
1678.It Ar allow-opts
1679By default, IPv4 packets with IP options or IPv6 packets with routing
1680extension headers are blocked.
1681When
1682.Ar allow-opts
1683is specified for a
1684.Ar pass
1685rule, packets that pass the filter based on that rule (last matching)
1686do so even if they contain IP options or routing extension headers.
1687For packets that match state, the rule that initially created the
1688state is used.
1689The implicit
1690.Ar pass
1691rule that is used when a packet does not match any rules does not
1692allow IP options.
1693.It Ar label Aq Ar string
1694Adds a label (name) to the rule, which can be used to identify the rule.
1695For instance,
1696pfctl -s labels
1697shows per-rule statistics for rules that have labels.
1698.Pp
1699The following macros can be used in labels:
1700.Pp
1701.Bl -tag -width $srcaddr -compact -offset indent
1702.It Ar $if
1703The interface.
1704.It Ar $srcaddr
1705The source IP address.
1706.It Ar $dstaddr
1707The destination IP address.
1708.It Ar $srcport
1709The source port specification.
1710.It Ar $dstport
1711The destination port specification.
1712.It Ar $proto
1713The protocol name.
1714.It Ar $nr
1715The rule number.
1716.El
1717.Pp
1718For example:
1719.Bd -literal -offset indent
1720ips = \&"{ 1.2.3.4, 1.2.3.5 }\&"
1721pass in proto tcp from any to $ips \e
1722      port \*(Gt 1023 label \&"$dstaddr:$dstport\&"
1723.Ed
1724.Pp
1725expands to
1726.Bd -literal -offset indent
1727pass in inet proto tcp from any to 1.2.3.4 \e
1728      port \*(Gt 1023 label \&"1.2.3.4:\*(Gt1023\&"
1729pass in inet proto tcp from any to 1.2.3.5 \e
1730      port \*(Gt 1023 label \&"1.2.3.5:\*(Gt1023\&"
1731.Ed
1732.Pp
1733The macro expansion for the
1734.Ar label
1735directive occurs only at configuration file parse time, not during runtime.
1736.It Ar queue Ao Ar queue Ac No \*(Ba ( Ao Ar queue Ac , Aq Ar queue )
1737Packets matching this rule will be assigned to the specified queue.
1738If two queues are given, packets which have a
1739.Em TOS
1740of
1741.Em lowdelay
1742and TCP ACKs with no data payload will be assigned to the second one.
1743See
1744.Sx QUEUEING
1745for setup details.
1746.Pp
1747For example:
1748.Bd -literal -offset indent
1749pass in proto tcp to port 25 queue mail
1750pass in proto tcp to port 22 queue(ssh_bulk, ssh_prio)
1751.Ed
1752.It Ar tag Aq Ar string
1753Packets matching this rule will be tagged with the
1754specified string.
1755The tag acts as an internal marker that can be used to
1756identify these packets later on.
1757This can be used, for example, to provide trust between
1758interfaces and to determine if packets have been
1759processed by translation rules.
1760Tags are
1761.Qq sticky ,
1762meaning that the packet will be tagged even if the rule
1763is not the last matching rule.
1764Further matching rules can replace the tag with a
1765new one but will not remove a previously applied tag.
1766A packet is only ever assigned one tag at a time.
1767Packet tagging can be done during
1768.Ar nat ,
1769.Ar rdr ,
1770or
1771.Ar binat
1772rules in addition to filter rules.
1773Tags take the same macros as labels (see above).
1774.It Ar tagged Aq Ar string
1775Used with filter or translation rules to specify that packets must already
1776be tagged with the given tag in order to match the rule.
1777Inverse tag matching can also be done
1778by specifying the
1779.Cm !\&
1780operator before the
1781.Ar tagged
1782keyword.
1783.It Ar rtable Aq Ar number
1784Used to select an alternate routing table for the routing lookup.
1785Only effective before the route lookup happened, i.e. when filtering inbound.
1786.It Ar probability Aq Ar number
1787A probability attribute can be attached to a rule, with a value set between
17880 and 1, bounds not included.
1789In that case, the rule will be honoured using the given probability value
1790only.
1791For example, the following rule will drop 20% of incoming ICMP packets:
1792.Bd -literal -offset indent
1793block in proto icmp probability 20%
1794.Ed
1795.El
1796.Sh ROUTING
1797If a packet matches a rule with a route option set, the packet filter will
1798route the packet according to the type of route option.
1799When such a rule creates state, the route option is also applied to all
1800packets matching the same connection.
1801.Bl -tag -width xxxx
1802.It Ar fastroute
1803The
1804.Ar fastroute
1805option does a normal route lookup to find the next hop for the packet.
1806.It Ar route-to
1807The
1808.Ar route-to
1809option routes the packet to the specified interface with an optional address
1810for the next hop.
1811When a
1812.Ar route-to
1813rule creates state, only packets that pass in the same direction as the
1814filter rule specifies will be routed in this way.
1815Packets passing in the opposite direction (replies) are not affected
1816and are routed normally.
1817.It Ar reply-to
1818The
1819.Ar reply-to
1820option is similar to
1821.Ar route-to ,
1822but routes packets that pass in the opposite direction (replies) to the
1823specified interface.
1824Opposite direction is only defined in the context of a state entry, and
1825.Ar reply-to
1826is useful only in rules that create state.
1827It can be used on systems with multiple external connections to
1828route all outgoing packets of a connection through the interface
1829the incoming connection arrived through (symmetric routing enforcement).
1830.It Ar dup-to
1831The
1832.Ar dup-to
1833option creates a duplicate of the packet and routes it like
1834.Ar route-to .
1835The original packet gets routed as it normally would.
1836.El
1837.Sh POOL OPTIONS
1838For
1839.Ar nat
1840and
1841.Ar rdr
1842rules, (as well as for the
1843.Ar route-to ,
1844.Ar reply-to
1845and
1846.Ar dup-to
1847rule options) for which there is a single redirection address which has a
1848subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP
1849address), a variety of different methods for assigning this address can be
1850used:
1851.Bl -tag -width xxxx
1852.It Ar bitmask
1853The
1854.Ar bitmask
1855option applies the network portion of the redirection address to the address
1856to be modified (source with
1857.Ar nat ,
1858destination with
1859.Ar rdr ) .
1860.It Ar random
1861The
1862.Ar random
1863option selects an address at random within the defined block of addresses.
1864.It Ar source-hash
1865The
1866.Ar source-hash
1867option uses a hash of the source address to determine the redirection address,
1868ensuring that the redirection address is always the same for a given source.
1869An optional key can be specified after this keyword either in hex or as a
1870string; by default
1871.Xr pfctl 8
1872randomly generates a key for source-hash every time the
1873ruleset is reloaded.
1874.It Ar round-robin
1875The
1876.Ar round-robin
1877option loops through the redirection address(es).
1878.Pp
1879When more than one redirection address is specified,
1880.Ar round-robin
1881is the only permitted pool type.
1882.It Ar static-port
1883With
1884.Ar nat
1885rules, the
1886.Ar static-port
1887option prevents
1888.Xr pf 4
1889from modifying the source port on TCP and UDP packets.
1890.El
1891.Pp
1892Additionally, the
1893.Ar sticky-address
1894option can be specified to help ensure that multiple connections from the
1895same source are mapped to the same redirection address.
1896This option can be used with the
1897.Ar random
1898and
1899.Ar round-robin
1900pool options.
1901Note that by default these associations are destroyed as soon as there are
1902no longer states which refer to them; in order to make the mappings last
1903beyond the lifetime of the states, increase the global options with
1904.Ar set timeout src.track .
1905See
1906.Sx STATEFUL TRACKING OPTIONS
1907for more ways to control the source tracking.
1908.Sh STATE MODULATION
1909Much of the security derived from TCP is attributable to how well the
1910initial sequence numbers (ISNs) are chosen.
1911Some popular stack implementations choose
1912.Em very
1913poor ISNs and thus are normally susceptible to ISN prediction exploits.
1914By applying a
1915.Ar modulate state
1916rule to a TCP connection,
1917.Xr pf 4
1918will create a high quality random sequence number for each connection
1919endpoint.
1920.Pp
1921The
1922.Ar modulate state
1923directive implicitly keeps state on the rule and is
1924only applicable to TCP connections.
1925.Pp
1926For instance:
1927.Bd -literal -offset indent
1928block all
1929pass out proto tcp from any to any modulate state
1930pass in  proto tcp from any to any port 25 flags S/SFRA modulate state
1931.Ed
1932.Pp
1933Note that modulated connections will not recover when the state table
1934is lost (firewall reboot, flushing the state table, etc...).
1935.Xr pf 4
1936will not be able to infer a connection again after the state table flushes
1937the connection's modulator.
1938When the state is lost, the connection may be left dangling until the
1939respective endpoints time out the connection.
1940It is possible on a fast local network for the endpoints to start an ACK
1941storm while trying to resynchronize after the loss of the modulator.
1942The default
1943.Ar flags
1944settings (or a more strict equivalent) should be used on
1945.Ar modulate state
1946rules to prevent ACK storms.
1947.Pp
1948Note that alternative methods are available
1949to prevent loss of the state table
1950and allow for firewall failover.
1951See
1952.Xr carp 4
1953and
1954.Xr pfsync 4
1955for further information.
1956.Sh SYN PROXY
1957By default,
1958.Xr pf 4
1959passes packets that are part of a
1960.Xr tcp 4
1961handshake between the endpoints.
1962The
1963.Ar synproxy state
1964option can be used to cause
1965.Xr pf 4
1966itself to complete the handshake with the active endpoint, perform a handshake
1967with the passive endpoint, and then forward packets between the endpoints.
1968.Pp
1969No packets are sent to the passive endpoint before the active endpoint has
1970completed the handshake, hence so-called SYN floods with spoofed source
1971addresses will not reach the passive endpoint, as the sender can't complete the
1972handshake.
1973.Pp
1974The proxy is transparent to both endpoints, they each see a single
1975connection from/to the other endpoint.
1976.Xr pf 4
1977chooses random initial sequence numbers for both handshakes.
1978Once the handshakes are completed, the sequence number modulators
1979(see previous section) are used to translate further packets of the
1980connection.
1981.Ar synproxy state
1982includes
1983.Ar modulate state .
1984.Pp
1985Rules with
1986.Ar synproxy
1987will not work if
1988.Xr pf 4
1989operates on a
1990.Xr bridge 4 .
1991.Pp
1992Example:
1993.Bd -literal -offset indent
1994pass in proto tcp from any to any port www synproxy state
1995.Ed
1996.Sh STATEFUL TRACKING OPTIONS
1997A number of options related to stateful tracking can be applied on a
1998per-rule basis.
1999.Ar keep state ,
2000.Ar modulate state
2001and
2002.Ar synproxy state
2003support these options, and
2004.Ar keep state
2005must be specified explicitly to apply options to a rule.
2006.Pp
2007.Bl -tag -width xxxx -compact
2008.It Ar max Aq Ar number
2009Limits the number of concurrent states the rule may create.
2010When this limit is reached, further packets matching the rule that would
2011create state are dropped, until existing states time out.
2012.\" .It Ar no-sync
2013.\" Prevent state changes for states created by this rule from appearing on the
2014.\" .Xr pfsync 4
2015.\" interface.
2016.It Ao Ar timeout Ac Aq Ar seconds
2017Changes the timeout values used for states created by this rule.
2018For a list of all valid timeout names, see
2019.Sx OPTIONS
2020above.
2021.El
2022.Pp
2023Multiple options can be specified, separated by commas:
2024.Bd -literal -offset indent
2025pass in proto tcp from any to any \e
2026      port www keep state \e
2027      (max 100, source-track rule, max-src-nodes 75, \e
2028      max-src-states 3, tcp.established 60, tcp.closing 5)
2029.Ed
2030.Pp
2031When the
2032.Ar source-track
2033keyword is specified, the number of states per source IP is tracked.
2034.Pp
2035.Bl -tag -width xxxx -compact
2036.It Ar source-track rule
2037The maximum number of states created by this rule is limited by the rule's
2038.Ar max-src-nodes
2039and
2040.Ar max-src-states
2041options.
2042Only state entries created by this particular rule count toward the rule's
2043limits.
2044.It Ar source-track global
2045The number of states created by all rules that use this option is limited.
2046Each rule can specify different
2047.Ar max-src-nodes
2048and
2049.Ar max-src-states
2050options, however state entries created by any participating rule count towards
2051each individual rule's limits.
2052.El
2053.Pp
2054The following limits can be set:
2055.Pp
2056.Bl -tag -width xxxx -compact
2057.It Ar max-src-nodes Aq Ar number
2058Limits the maximum number of source addresses which can simultaneously
2059have state table entries.
2060.It Ar max-src-states Aq Ar number
2061Limits the maximum number of simultaneous state entries that a single
2062source address can create with this rule.
2063.El
2064.Pp
2065For stateful TCP connections, limits on established connections (connections
2066which have completed the TCP 3-way handshake) can also be enforced
2067per source IP.
2068.Pp
2069.Bl -tag -width xxxx -compact
2070.It Ar max-src-conn Aq Ar number
2071Limits the maximum number of simultaneous TCP connections which have
2072completed the 3-way handshake that a single host can make.
2073.It Ar max-src-conn-rate Ao Ar number Ac No / Aq Ar seconds
2074Limit the rate of new connections over a time interval.
2075The connection rate is an approximation calculated as a moving average.
2076.El
2077.Pp
2078Because the 3-way handshake ensures that the source address is not being
2079spoofed, more aggressive action can be taken based on these limits.
2080With the
2081.Ar overload Aq Ar table
2082state option, source IP addresses which hit either of the limits on
2083established connections will be added to the named table.
2084This table can be used in the ruleset to block further activity from
2085the offending host, redirect it to a tarpit process, or restrict its
2086bandwidth.
2087.Pp
2088The optional
2089.Ar flush
2090keyword kills all states created by the matching rule which originate
2091from the host which exceeds these limits.
2092The
2093.Ar global
2094modifier to the flush command kills all states originating from the
2095offending host, regardless of which rule created the state.
2096.Pp
2097For example, the following rules will protect the webserver against
2098hosts making more than 100 connections in 10 seconds.
2099Any host which connects faster than this rate will have its address added
2100to the
2101.Aq bad_hosts
2102table and have all states originating from it flushed.
2103Any new packets arriving from this host will be dropped unconditionally
2104by the block rule.
2105.Bd -literal -offset indent
2106block quick from \*(Ltbad_hosts\*(Gt
2107pass in on $ext_if proto tcp to $webserver port www keep state \e
2108	(max-src-conn-rate 100/10, overload \*(Ltbad_hosts\*(Gt flush global)
2109.Ed
2110.Pp
2111You can adjust the state policy on individual
2112.Ar nat
2113and
2114.Ar rdr
2115translation rules by adding a keyword
2116.Ar if-bound ,
2117.Ar group-bound
2118or
2119.Ar floating
2120at the end of the rule.
2121For example, a rule such as this,
2122.Bd -literal -offset indent
2123nat on sip0 from 10/8 to ! 10/8 -> 192.168.1.4/32 if-bound
2124.Ed
2125.Pp
2126will create states that only match packets on sip0.
2127.Sh OPERATING SYSTEM FINGERPRINTING
2128Passive OS Fingerprinting is a mechanism to inspect nuances of a TCP
2129connection's initial SYN packet and guess at the host's operating system.
2130Unfortunately these nuances are easily spoofed by an attacker so the
2131fingerprint is not useful in making security decisions.
2132But the fingerprint is typically accurate enough to make policy decisions
2133upon.
2134.Pp
2135The fingerprints may be specified by operating system class, by
2136version, or by subtype/patchlevel.
2137The class of an operating system is typically the vendor or genre
2138and would be
2139.Ox
2140for the
2141.Xr pf 4
2142firewall itself.
2143The version of the oldest available
2144.Ox
2145release on the main FTP site
2146would be 2.6 and the fingerprint would be written
2147.Pp
2148.Dl \&"OpenBSD 2.6\&"
2149.Pp
2150The subtype of an operating system is typically used to describe the
2151patchlevel if that patch led to changes in the TCP stack behavior.
2152In the case of
2153.Ox ,
2154the only subtype is for a fingerprint that was
2155normalized by the
2156.Ar no-df
2157scrub option and would be specified as
2158.Pp
2159.Dl \&"OpenBSD 3.3 no-df\&"
2160.Pp
2161Fingerprints for most popular operating systems are provided by
2162.Xr pf.os 5 .
2163Once
2164.Xr pf 4
2165is running, a complete list of known operating system fingerprints may
2166be listed by running:
2167.Pp
2168.Dl # pfctl -so
2169.Pp
2170Filter rules can enforce policy at any level of operating system specification
2171assuming a fingerprint is present.
2172Policy could limit traffic to approved operating systems or even ban traffic
2173from hosts that aren't at the latest service pack.
2174.Pp
2175The
2176.Ar unknown
2177class can also be used as the fingerprint which will match packets for
2178which no operating system fingerprint is known.
2179.Pp
2180Examples:
2181.Bd -literal -offset indent
2182pass  out proto tcp from any os OpenBSD
2183block out proto tcp from any os Doors
2184block out proto tcp from any os "Doors PT"
2185block out proto tcp from any os "Doors PT SP3"
2186block out from any os "unknown"
2187pass on lo0 proto tcp from any os "OpenBSD 3.3 lo0"
2188.Ed
2189.Pp
2190Operating system fingerprinting is limited only to the TCP SYN packet.
2191This means that it will not work on other protocols and will not match
2192a currently established connection.
2193.Pp
2194Caveat: operating system fingerprints are occasionally wrong.
2195There are three problems: an attacker can trivially craft his packets to
2196appear as any operating system he chooses;
2197an operating system patch could change the stack behavior and no fingerprints
2198will match it until the database is updated;
2199and multiple operating systems may have the same fingerprint.
2200.Sh BLOCKING SPOOFED TRAFFIC
2201"Spoofing" is the faking of IP addresses, typically for malicious
2202purposes.
2203The
2204.Ar antispoof
2205directive expands to a set of filter rules which will block all
2206traffic with a source IP from the network(s) directly connected
2207to the specified interface(s) from entering the system through
2208any other interface.
2209.Pp
2210For example, the line
2211.Bd -literal -offset indent
2212antispoof for lo0
2213.Ed
2214.Pp
2215expands to
2216.Bd -literal -offset indent
2217block drop in on ! lo0 inet from 127.0.0.1/8 to any
2218block drop in on ! lo0 inet6 from ::1 to any
2219.Ed
2220.Pp
2221For non-loopback interfaces, there are additional rules to block incoming
2222packets with a source IP address identical to the interface's IP(s).
2223For example, assuming the interface wi0 had an IP address of 10.0.0.1 and a
2224netmask of 255.255.255.0,
2225the line
2226.Bd -literal -offset indent
2227antispoof for wi0 inet
2228.Ed
2229.Pp
2230expands to
2231.Bd -literal -offset indent
2232block drop in on ! wi0 inet from 10.0.0.0/24 to any
2233block drop in inet from 10.0.0.1 to any
2234.Ed
2235.Pp
2236Caveat: Rules created by the
2237.Ar antispoof
2238directive interfere with packets sent over loopback interfaces
2239to local addresses.
2240One should pass these explicitly.
2241.Sh FRAGMENT HANDLING
2242The size of IP datagrams (packets) can be significantly larger than the
2243maximum transmission unit (MTU) of the network.
2244In cases when it is necessary or more efficient to send such large packets,
2245the large packet will be fragmented into many smaller packets that will each
2246fit onto the wire.
2247Unfortunately for a firewalling device, only the first logical fragment will
2248contain the necessary header information for the subprotocol that allows
2249.Xr pf 4
2250to filter on things such as TCP ports or to perform NAT.
2251.Pp
2252Besides the use of
2253.Ar scrub
2254rules as described in
2255.Sx TRAFFIC NORMALIZATION
2256above, there are three options for handling fragments in the packet filter.
2257.Pp
2258One alternative is to filter individual fragments with filter rules.
2259If no
2260.Ar scrub
2261rule applies to a fragment, it is passed to the filter.
2262Filter rules with matching IP header parameters decide whether the
2263fragment is passed or blocked, in the same way as complete packets
2264are filtered.
2265Without reassembly, fragments can only be filtered based on IP header
2266fields (source/destination address, protocol), since subprotocol header
2267fields are not available (TCP/UDP port numbers, ICMP code/type).
2268The
2269.Ar fragment
2270option can be used to restrict filter rules to apply only to
2271fragments, but not complete packets.
2272Filter rules without the
2273.Ar fragment
2274option still apply to fragments, if they only specify IP header fields.
2275For instance, the rule
2276.Bd -literal -offset indent
2277pass in proto tcp from any to any port 80
2278.Ed
2279.Pp
2280never applies to a fragment, even if the fragment is part of a TCP
2281packet with destination port 80, because without reassembly this information
2282is not available for each fragment.
2283This also means that fragments cannot create new or match existing
2284state table entries, which makes stateful filtering and address
2285translation (NAT, redirection) for fragments impossible.
2286.Pp
2287It's also possible to reassemble only certain fragments by specifying
2288source or destination addresses or protocols as parameters in
2289.Ar scrub
2290rules.
2291.Pp
2292In most cases, the benefits of reassembly outweigh the additional
2293memory cost, and it's recommended to use
2294.Ar scrub
2295rules to reassemble
2296all fragments via the
2297.Ar fragment reassemble
2298modifier.
2299.Pp
2300The memory allocated for fragment caching can be limited using
2301.Xr pfctl 8 .
2302Once this limit is reached, fragments that would have to be cached
2303are dropped until other entries time out.
2304The timeout value can also be adjusted.
2305.Pp
2306Currently, only IPv4 fragments are supported and IPv6 fragments
2307are blocked unconditionally.
2308.Sh ANCHORS
2309Besides the main ruleset,
2310.Xr pfctl 8
2311can load rulesets into
2312.Ar anchor
2313attachment points.
2314An
2315.Ar anchor
2316is a container that can hold rules, address tables, and other anchors.
2317.Pp
2318An
2319.Ar anchor
2320has a name which specifies the path where
2321.Xr pfctl 8
2322can be used to access the anchor to perform operations on it, such as
2323attaching child anchors to it or loading rules into it.
2324Anchors may be nested, with components separated by
2325.Sq /
2326characters, similar to how file system hierarchies are laid out.
2327The main ruleset is actually the default anchor, so filter and
2328translation rules, for example, may also be contained in any anchor.
2329.Pp
2330An anchor can reference another
2331.Ar anchor
2332attachment point
2333using the following kinds
2334of rules:
2335.Bl -tag -width xxxx
2336.It Ar nat-anchor Aq Ar name
2337Evaluates the
2338.Ar nat
2339rules in the specified
2340.Ar anchor .
2341.It Ar rdr-anchor Aq Ar name
2342Evaluates the
2343.Ar rdr
2344rules in the specified
2345.Ar anchor .
2346.It Ar binat-anchor Aq Ar name
2347Evaluates the
2348.Ar binat
2349rules in the specified
2350.Ar anchor .
2351.It Ar anchor Aq Ar name
2352Evaluates the filter rules in the specified
2353.Ar anchor .
2354.It Ar load anchor Ao Ar name Ac Ar from Aq Ar file
2355Loads the rules from the specified file into the
2356anchor
2357.Ar name .
2358.El
2359.Pp
2360When evaluation of the main ruleset reaches an
2361.Ar anchor
2362rule,
2363.Xr pf 4
2364will proceed to evaluate all rules specified in that anchor.
2365.Pp
2366Matching filter and translation rules marked with the
2367.Ar quick
2368option are final and abort the evaluation of the rules in other
2369anchors and the main ruleset.
2370If the
2371.Ar anchor
2372itself is marked with the
2373.Ar quick
2374option,
2375ruleset evaluation will terminate when the anchor is exited if the packet is
2376matched by any rule within the anchor.
2377.Pp
2378.Ar anchor
2379rules are evaluated relative to the anchor in which they are contained.
2380For example, all
2381.Ar anchor
2382rules specified in the main ruleset will reference anchor
2383attachment points underneath the main ruleset, and
2384.Ar anchor
2385rules specified in a file loaded from a
2386.Ar load anchor
2387rule will be attached under that anchor point.
2388.Pp
2389Rules may be contained in
2390.Ar anchor
2391attachment points which do not contain any rules when the main ruleset
2392is loaded, and later such anchors can be manipulated through
2393.Xr pfctl 8
2394without reloading the main ruleset or other anchors.
2395For example,
2396.Bd -literal -offset indent
2397ext_if = \&"kue0\&"
2398block on $ext_if all
2399anchor spam
2400pass out on $ext_if all
2401pass in on $ext_if proto tcp from any \e
2402      to $ext_if port smtp
2403.Ed
2404.Pp
2405blocks all packets on the external interface by default, then evaluates
2406all rules in the
2407.Ar anchor
2408named "spam", and finally passes all outgoing connections and
2409incoming connections to port 25.
2410.Bd -literal -offset indent
2411# echo \&"block in quick from 1.2.3.4 to any\&" \&| \e
2412      pfctl -a spam -f -
2413.Ed
2414.Pp
2415This loads a single rule into the
2416.Ar anchor ,
2417which blocks all packets from a specific address.
2418.Pp
2419The anchor can also be populated by adding a
2420.Ar load anchor
2421rule after the
2422.Ar anchor
2423rule:
2424.Bd -literal -offset indent
2425anchor spam
2426load anchor spam from "/etc/pf-spam.conf"
2427.Ed
2428.Pp
2429When
2430.Xr pfctl 8
2431loads
2432.Nm pf.conf ,
2433it will also load all the rules from the file
2434.Pa /etc/pf-spam.conf
2435into the anchor.
2436.Pp
2437Optionally,
2438.Ar anchor
2439rules can specify the parameter's
2440direction, interface, address family, protocol and source/destination
2441address/port
2442using the same syntax as filter rules.
2443When parameters are used, the
2444.Ar anchor
2445rule is only evaluated for matching packets.
2446This allows conditional evaluation of anchors, like:
2447.Bd -literal -offset indent
2448block on $ext_if all
2449anchor spam proto tcp from any to any port smtp
2450pass out on $ext_if all
2451pass in on $ext_if proto tcp from any to $ext_if port smtp
2452.Ed
2453.Pp
2454The rules inside
2455.Ar anchor
2456spam are only evaluated for
2457.Ar tcp
2458packets with destination port 25.
2459Hence,
2460.Bd -literal -offset indent
2461# echo \&"block in quick from 1.2.3.4 to any" \&| \e
2462      pfctl -a spam -f -
2463.Ed
2464.Pp
2465will only block connections from 1.2.3.4 to port 25.
2466.Pp
2467Anchors may end with the asterisk
2468.Pq Sq *
2469character, which signifies that all anchors attached at that point
2470should be evaluated in the alphabetical ordering of their anchor name.
2471For example,
2472.Bd -literal -offset indent
2473anchor "spam/*"
2474.Ed
2475.Pp
2476will evaluate each rule in each anchor attached to the
2477.Li spam
2478anchor.
2479Note that it will only evaluate anchors that are directly attached to the
2480.Li spam
2481anchor, and will not descend to evaluate anchors recursively.
2482.Pp
2483Since anchors are evaluated relative to the anchor in which they are
2484contained, there is a mechanism for accessing the parent and ancestor
2485anchors of a given anchor.
2486Similar to file system path name resolution, if the sequence
2487.Dq ..
2488appears as an anchor path component, the parent anchor of the current
2489anchor in the path evaluation at that point will become the new current
2490anchor.
2491As an example, consider the following:
2492.Bd -literal -offset indent
2493# echo ' anchor "spam/allowed" ' | pfctl -f -
2494# echo -e ' anchor "../banned" \en pass' | \e
2495      pfctl -a spam/allowed -f -
2496.Ed
2497.Pp
2498Evaluation of the main ruleset will lead into the
2499.Li spam/allowed
2500anchor, which will evaluate the rules in the
2501.Li spam/banned
2502anchor, if any, before finally evaluating the
2503.Ar pass
2504rule.
2505.Pp
2506Filter rule
2507.Ar anchors
2508can also be loaded inline in the ruleset within a brace ('{' '}') delimited
2509block.
2510Brace delimited blocks may contain rules or other brace-delimited blocks.
2511When anchors are loaded this way the anchor name becomes optional.
2512.Bd -literal -offset indent
2513anchor "external" on egress {
2514	block
2515	anchor out {
2516		pass proto tcp from any to port { 25, 80, 443 }
2517	}
2518	pass in proto tcp to any port 22
2519}
2520.Ed
2521.Pp
2522Since the parser specification for anchor names is a string, any
2523reference to an anchor name containing solidus
2524.Pq Sq /
2525characters will require double quote
2526.Pq Sq \&"
2527characters around the anchor name.
2528.Sh TRANSLATION EXAMPLES
2529This example maps incoming requests on port 80 to port 8080, on
2530which a daemon is running (because, for example, it is not run as root,
2531and therefore lacks permission to bind to port 80).
2532.Bd -literal
2533# use a macro for the interface name, so it can be changed easily
2534ext_if = \&"ne3\&"
2535
2536# map daemon on 8080 to appear to be on 80
2537rdr on $ext_if proto tcp from any to any port 80 -\*(Gt 127.0.0.1 port 8080
2538.Ed
2539.Pp
2540If the
2541.Ar pass
2542modifier is given, packets matching the translation rule are passed without
2543inspecting the filter rules:
2544.Bd -literal
2545rdr pass on $ext_if proto tcp from any to any port 80 -\*(Gt 127.0.0.1 \e
2546      port 8080
2547.Ed
2548.Pp
2549In the example below, vlan12 is configured as 192.168.168.1;
2550the machine translates all packets coming from 192.168.168.0/24 to 204.92.77.111
2551when they are going out any interface except vlan12.
2552This has the net effect of making traffic from the 192.168.168.0/24
2553network appear as though it is the Internet routable address
2554204.92.77.111 to nodes behind any interface on the router except
2555for the nodes on vlan12.
2556(Thus, 192.168.168.1 can talk to the 192.168.168.0/24 nodes.)
2557.Bd -literal
2558nat on ! vlan12 from 192.168.168.0/24 to any -\*(Gt 204.92.77.111
2559.Ed
2560.Pp
2561In the example below, the machine sits between a fake internal 144.19.74.*
2562network, and a routable external IP of 204.92.77.100.
2563The
2564.Ar no nat
2565rule excludes protocol AH from being translated.
2566.Bd -literal
2567# NO NAT
2568no nat on $ext_if proto ah from 144.19.74.0/24 to any
2569nat on $ext_if from 144.19.74.0/24 to any -\*(Gt 204.92.77.100
2570.Ed
2571.Pp
2572In the example below, packets bound for one specific server, as well as those
2573generated by the sysadmins are not proxied; all other connections are.
2574.Bd -literal
2575# NO RDR
2576no rdr on $int_if proto { tcp, udp } from any to $server port 80
2577no rdr on $int_if proto { tcp, udp } from $sysadmins to any port 80
2578rdr on $int_if proto { tcp, udp } from any to any port 80 -\*(Gt 127.0.0.1 \e
2579      port 80
2580.Ed
2581.Pp
2582This longer example uses both a NAT and a redirection.
2583The external interface has the address 157.161.48.183.
2584On localhost, we are running
2585.Xr ftp-proxy 8 ,
2586waiting for FTP sessions to be redirected to it.
2587The three mandatory anchors for
2588.Xr ftp-proxy 8
2589are omitted from this example; see the
2590.Xr ftp-proxy 8
2591manpage.
2592.Bd -literal
2593# NAT
2594# Translate outgoing packets' source addresses (any protocol).
2595# In this case, any address but the gateway's external address is mapped.
2596nat on $ext_if inet from ! ($ext_if) to any -\*(Gt ($ext_if)
2597
2598# NAT PROXYING
2599# Map outgoing packets' source port to an assigned proxy port instead of
2600# an arbitrary port.
2601# In this case, proxy outgoing isakmp with port 500 on the gateway.
2602nat on $ext_if inet proto udp from any port = isakmp to any -\*(Gt ($ext_if) \e
2603      port 500
2604
2605# BINAT
2606# Translate outgoing packets' source address (any protocol).
2607# Translate incoming packets' destination address to an internal machine
2608# (bidirectional).
2609binat on $ext_if from 10.1.2.150 to any -\*(Gt $ext_if
2610
2611# RDR
2612# Translate incoming packets' destination addresses.
2613# As an example, redirect a TCP and UDP port to an internal machine.
2614rdr on $ext_if inet proto tcp from any to ($ext_if) port 8080 \e
2615      -\*(Gt 10.1.2.151 port 22
2616rdr on $ext_if inet proto udp from any to ($ext_if) port 8080 \e
2617      -\*(Gt 10.1.2.151 port 53
2618
2619# RDR
2620# Translate outgoing ftp control connections to send them to localhost
2621# for proxying with ftp-proxy(8) running on port 8021.
2622rdr on $int_if proto tcp from any to any port 21 -\*(Gt 127.0.0.1 port 8021
2623.Ed
2624.Pp
2625In this example, a NAT gateway is set up to translate internal addresses
2626using a pool of public addresses (192.0.2.16/28) and to redirect
2627incoming web server connections to a group of web servers on the internal
2628network.
2629.Bd -literal
2630# NAT LOAD BALANCE
2631# Translate outgoing packets' source addresses using an address pool.
2632# A given source address is always translated to the same pool address by
2633# using the source-hash keyword.
2634nat on $ext_if inet from any to any -\*(Gt 192.0.2.16/28 source-hash
2635
2636# RDR ROUND ROBIN
2637# Translate incoming web server connections to a group of web servers on
2638# the internal network.
2639rdr on $ext_if proto tcp from any to any port 80 \e
2640      -\*(Gt { 10.1.2.155, 10.1.2.160, 10.1.2.161 } round-robin
2641.Ed
2642.Sh FILTER EXAMPLES
2643.Bd -literal
2644# The external interface is kue0
2645# (157.161.48.183, the only routable address)
2646# and the private network is 10.0.0.0/8, for which we are doing NAT.
2647
2648# use a macro for the interface name, so it can be changed easily
2649ext_if = \&"kue0\&"
2650
2651# normalize all incoming traffic
2652scrub in on $ext_if all fragment reassemble
2653
2654# block and log everything by default
2655block return log on $ext_if all
2656
2657# block anything coming from source we have no back routes for
2658block in from no-route to any
2659
2660# block packets whose ingress interface does not match the one in
2661# the route back to their source address
2662block in from urpf-failed to any
2663
2664# block and log outgoing packets that do not have our address as source,
2665# they are either spoofed or something is misconfigured (NAT disabled,
2666# for instance), we want to be nice and do not send out garbage.
2667block out log quick on $ext_if from ! 157.161.48.183 to any
2668
2669# silently drop broadcasts (cable modem noise)
2670block in quick on $ext_if from any to 255.255.255.255
2671
2672# block and log incoming packets from reserved address space and invalid
2673# addresses, they are either spoofed or misconfigured, we cannot reply to
2674# them anyway (hence, no return-rst).
2675block in log quick on $ext_if from { 10.0.0.0/8, 172.16.0.0/12, \e
2676      192.168.0.0/16, 255.255.255.255/32 } to any
2677
2678# ICMP
2679
2680# pass out/in certain ICMP queries and keep state (ping)
2681# state matching is done on host addresses and ICMP id (not type/code),
2682# so replies (like 0/0 for 8/0) will match queries
2683# ICMP error messages (which always refer to a TCP/UDP packet) are
2684# handled by the TCP/UDP states
2685pass on $ext_if inet proto icmp all icmp-type 8 code 0
2686
2687# UDP
2688
2689# pass out all UDP connections and keep state
2690pass out on $ext_if proto udp all
2691
2692# pass in certain UDP connections and keep state (DNS)
2693pass in on $ext_if proto udp from any to any port domain
2694
2695# TCP
2696
2697# pass out all TCP connections and modulate state
2698pass out on $ext_if proto tcp all modulate state
2699
2700# pass in certain TCP connections and keep state (SSH, SMTP, DNS, IDENT)
2701pass in on $ext_if proto tcp from any to any port { ssh, smtp, domain, \e
2702      auth }
2703
2704# Do not allow Windows 9x SMTP connections since they are typically
2705# a viral worm. Alternately we could limit these OSes to 1 connection each.
2706block in on $ext_if proto tcp from any os {"Windows 95", "Windows 98"} \e
2707      to any port smtp
2708
2709# IPv6
2710# pass in/out all IPv6 traffic: note that we have to enable this in two
2711# different ways, on both our physical interface and our tunnel
2712pass quick on gif0 inet6
2713pass quick on $ext_if proto ipv6
2714
2715# Packet Tagging
2716
2717# three interfaces: $int_if, $ext_if, and $wifi_if (wireless). NAT is
2718# being done on $ext_if for all outgoing packets. tag packets in on
2719# $int_if and pass those tagged packets out on $ext_if.  all other
2720# outgoing packets (i.e., packets from the wireless network) are only
2721# permitted to access port 80.
2722
2723pass in on $int_if from any to any tag INTNET
2724pass in on $wifi_if from any to any
2725
2726block out on $ext_if from any to any
2727pass out quick on $ext_if tagged INTNET
2728pass out on $ext_if proto tcp from any to any port 80
2729
2730# tag incoming packets as they are redirected to spamd(8). use the tag
2731# to pass those packets through the packet filter.
2732
2733rdr on $ext_if inet proto tcp from \*(Ltspammers\*(Gt to port smtp \e
2734	tag SPAMD -\*(Gt 127.0.0.1 port spamd
2735
2736block in on $ext_if
2737pass in on $ext_if inet proto tcp tagged SPAMD
2738.Ed
2739.Sh GRAMMAR
2740Syntax for
2741.Nm
2742in BNF:
2743.Bd -literal
2744line           = ( option | pf-rule | nat-rule | binat-rule | rdr-rule |
2745                 antispoof-rule | altq-rule | queue-rule | trans-anchors |
2746		 anchor-rule | anchor-close | load-anchor | table-rule | )
2747
2748option         = "set" ( [ "timeout" ( timeout | "{" timeout-list "}" ) ] |
2749		 [ "ruleset-optimization" [ "none" | "basic" | "profile" ]] |
2750                 [ "optimization" [ "default" | "normal" |
2751                 "high-latency" | "satellite" |
2752                 "aggressive" | "conservative" ] ]
2753                 [ "limit" ( limit-item | "{" limit-list "}" ) ] |
2754                 [ "loginterface" ( interface-name | "none" ) ] |
2755                 [ "block-policy" ( "drop" | "return" ) ] |
2756                 [ "state-policy" ( "if-bound" | "floating" ) ]
2757                 [ "require-order" ( "yes" | "no" ) ]
2758                 [ "fingerprints" filename ] |
2759                 [ "skip on" ifspec ] |
2760                 [ "debug" ( "none" | "urgent" | "misc" | "loud" ) ] )
2761
2762pf-rule        = action [ ( "in" | "out" ) ]
2763                 [ "log" [ "(" logopts ")"] ] [ "quick" ]
2764                 [ "on" ifspec ] [ "fastroute" | route ] [ af ] [ protospec ]
2765                 hosts [ filteropt-list ]
2766
2767logopts        = logopt [ "," logopts ]
2768logopt         = "all" | "user" | "to" interface-name
2769
2770filteropt-list = filteropt-list filteropt | filteropt
2771filteropt      = user | group | flags | icmp-type | icmp6-type | tos |
2772                 ( "no" | "keep" | "modulate" | "synproxy" ) "state"
2773                 [ "(" state-opts ")" ] |
2774                 "fragment" | "no-df" | "min-ttl" number |
2775                 "max-mss" number | "random-id" | "reassemble tcp" |
2776                 fragmentation | "allow-opts" |
2777                 "label" string | "tag" string | [ ! ] "tagged" string |
2778                 "queue" ( string | "(" string [ [ "," ] string ] ")" ) |
2779                 "rtable" number | "probability" number"%"
2780
2781nat-rule       = [ "no" ] "nat" [ "pass" [ "log" [ "(" logopts ")" ] ] ]
2782                 [ "on" ifspec ] [ af ]
2783                 [ protospec ] hosts [ "tag" string ] [ "tagged" string ]
2784                 [ "-\*(Gt" ( redirhost | "{" redirhost-list "}" )
2785                 [ portspec ] [ pooltype ] [ "static-port" ] ]
2786                 [ ( "if-bound" | "group-bound" | "floating" ) ]
2787
2788binat-rule     = [ "no" ] "binat" [ "pass" [ "log" [ "(" logopts ")" ] ] ]
2789                 [ "on" interface-name ] [ af ]
2790                 [ "proto" ( proto-name | proto-number ) ]
2791                 "from" address [ "/" mask-bits ] "to" ipspec
2792                 [ "tag" string ] [ "tagged" string ]
2793                 [ "-\*(Gt" address [ "/" mask-bits ] ]
2794
2795rdr-rule       = [ "no" ] "rdr" [ "pass" [ "log" [ "(" logopts ")" ] ] ]
2796                 [ "on" ifspec ] [ af ]
2797                 [ protospec ] hosts [ "tag" string ] [ "tagged" string ]
2798                 [ "-\*(Gt" ( redirhost | "{" redirhost-list "}" )
2799                 [ portspec ] [ pooltype ] ]
2800                 [ ( "if-bound" | "group-bound" | "floating" ) ]
2801
2802antispoof-rule = "antispoof" [ "log" ] [ "quick" ]
2803                 "for" ifspec [ af ] [ "label" string ]
2804
2805table-rule     = "table" "\*(Lt" string "\*(Gt" [ tableopts-list ]
2806tableopts-list = tableopts-list tableopts | tableopts
2807tableopts      = "persist" | "const" | "file" string |
2808                 "{" [ tableaddr-list ] "}"
2809tableaddr-list = tableaddr-list [ "," ] tableaddr-spec | tableaddr-spec
2810tableaddr-spec = [ "!" ] tableaddr [ "/" mask-bits ]
2811tableaddr      = hostname | ifspec | "self" |
2812                 ipv4-dotted-quad | ipv6-coloned-hex
2813
2814altq-rule      = "altq on" interface-name queueopts-list
2815                 "queue" subqueue
2816queue-rule     = "queue" string [ "on" interface-name ] queueopts-list
2817                 subqueue
2818
2819anchor-rule    = "anchor" [ string ] [ ( "in" | "out" ) ] [ "on" ifspec ]
2820                 [ af ] [ protospec ] [ hosts ] [ "{" ]
2821
2822anchor-close   = "}"
2823
2824trans-anchors  = ( "nat-anchor" | "rdr-anchor" | "binat-anchor" ) string
2825                 [ "on" ifspec ] [ af ] [ "proto" ] [ protospec ] [ hosts ]
2826
2827load-anchor    = "load anchor" string "from" filename
2828
2829queueopts-list = queueopts-list queueopts | queueopts
2830queueopts      = [ "bandwidth" bandwidth-spec ] |
2831                 [ "qlimit" number ] | [ "tbrsize" number ] |
2832                 [ "priority" number ] | [ schedulers ]
2833schedulers     = ( cbq-def | priq-def | hfsc-def )
2834bandwidth-spec = "number" ( "b" | "Kb" | "Mb" | "Gb" | "%" )
2835
2836action         = "pass" | "block" [ return ] | [ "no" ] "scrub"
2837return         = "drop" | "return" | "return-rst" [ "( ttl" number ")" ] |
2838                 "return-icmp" [ "(" icmpcode [ [ "," ] icmp6code ] ")" ] |
2839                 "return-icmp6" [ "(" icmp6code ")" ]
2840icmpcode       = ( icmp-code-name | icmp-code-number )
2841icmp6code      = ( icmp6-code-name | icmp6-code-number )
2842
2843ifspec         = ( [ "!" ] ( interface-name | interface-group ) ) |
2844                 "{" interface-list "}"
2845interface-list = [ "!" ] ( interface-name | interface-group )
2846                 [ [ "," ] interface-list ]
2847route          = ( "route-to" | "reply-to" | "dup-to" )
2848                 ( routehost | "{" routehost-list "}" )
2849                 [ pooltype ]
2850af             = "inet" | "inet6"
2851
2852protospec      = "proto" ( proto-name | proto-number |
2853                 "{" proto-list "}" )
2854proto-list     = ( proto-name | proto-number ) [ [ "," ] proto-list ]
2855
2856hosts          = "all" |
2857                 "from" ( "any" | "no-route" | "urpf-failed" | "self" | host |
2858                 "{" host-list "}" | "route" string ) [ port ] [ os ]
2859                 "to"   ( "any" | "no-route" | "self" | host |
2860                 "{" host-list "}" | "route" string ) [ port ]
2861
2862ipspec         = "any" | host | "{" host-list "}"
2863host           = [ "!" ] ( address [ "/" mask-bits ] | "\*(Lt" string "\*(Gt" )
2864redirhost      = address [ "/" mask-bits ]
2865routehost      = "(" interface-name [ address [ "/" mask-bits ] ] ")"
2866address        = ( interface-name | interface-group |
2867                 "(" ( interface-name | interface-group ) ")" |
2868                 hostname | ipv4-dotted-quad | ipv6-coloned-hex )
2869host-list      = host [ [ "," ] host-list ]
2870redirhost-list = redirhost [ [ "," ] redirhost-list ]
2871routehost-list = routehost [ [ "," ] routehost-list ]
2872
2873port           = "port" ( unary-op | binary-op | "{" op-list "}" )
2874portspec       = "port" ( number | name ) [ ":" ( "*" | number | name ) ]
2875os             = "os"  ( os-name | "{" os-list "}" )
2876user           = "user" ( unary-op | binary-op | "{" op-list "}" )
2877.\" group          = "group" ( unary-op | binary-op | "{" op-list "}" )
2878
2879unary-op       = [ "=" | "!=" | "\*(Lt" | "\*(Le" | "\*(Gt" | "\*(Ge" ]
2880                 ( name | number )
2881binary-op      = number ( "\*(Lt\*(Gt" | "\*(Gt\*(Lt" | ":" ) number
2882op-list        = ( unary-op | binary-op ) [ [ "," ] op-list ]
2883
2884os-name        = operating-system-name
2885os-list        = os-name [ [ "," ] os-list ]
2886
2887flags          = "flags" ( [ flag-set ] "/"  flag-set | "any" )
2888flag-set       = [ "F" ] [ "S" ] [ "R" ] [ "P" ] [ "A" ] [ "U" ] [ "E" ]
2889                 [ "W" ]
2890
2891icmp-type      = "icmp-type" ( icmp-type-code | "{" icmp-list "}" )
2892icmp6-type     = "icmp6-type" ( icmp-type-code | "{" icmp-list "}" )
2893icmp-type-code = ( icmp-type-name | icmp-type-number )
2894                 [ "code" ( icmp-code-name | icmp-code-number ) ]
2895icmp-list      = icmp-type-code [ [ "," ] icmp-list ]
2896
2897tos            = "tos" ( "lowdelay" | "throughput" | "reliability" |
2898                 [ "0x" ] number )
2899
2900state-opts     = state-opt [ [ "," ] state-opts ]
2901state-opt      = ( "max" number | timeout |
2902                 "source-track" [ ( "rule" | "global" ) ] |
2903                 "max-src-nodes" number | "max-src-states" number |
2904                 "max-src-conn" number |
2905                 "max-src-conn-rate" number "/" number |
2906                 "overload" "\*(Lt" string "\*(Gt" [ "flush" ] |
2907                 "if-bound" | "floating" )
2908
2909fragmentation  = [ "fragment reassemble" | "fragment crop" |
2910                 "fragment drop-ovl" ]
2911
2912timeout-list   = timeout [ [ "," ] timeout-list ]
2913timeout        = ( "tcp.first" | "tcp.opening" | "tcp.established" |
2914                 "tcp.closing" | "tcp.finwait" | "tcp.closed" |
2915                 "udp.first" | "udp.single" | "udp.multiple" |
2916                 "icmp.first" | "icmp.error" |
2917                 "other.first" | "other.single" | "other.multiple" |
2918                 "frag" | "interval" | "src.track" |
2919                 "adaptive.start" | "adaptive.end" ) number
2920
2921limit-list     = limit-item [ [ "," ] limit-list ]
2922limit-item     = ( "states" | "frags" | "src-nodes" ) number
2923
2924pooltype       = ( "bitmask" | "random" |
2925                 "source-hash" [ ( hex-key | string-key ) ] |
2926                 "round-robin" ) [ sticky-address ]
2927
2928subqueue       = string | "{" queue-list "}"
2929queue-list     = string [ [ "," ] string ]
2930cbq-def        = "cbq" [ "(" cbq-opt [ [ "," ] cbq-opt ] ")" ]
2931priq-def       = "priq" [ "(" priq-opt [ [ "," ] priq-opt ] ")" ]
2932hfsc-def       = "hfsc" [ "(" hfsc-opt [ [ "," ] hfsc-opt ] ")" ]
2933cbq-opt        = ( "default" | "borrow" | "red" | "ecn" | "rio" )
2934priq-opt       = ( "default" | "red" | "ecn" | "rio" )
2935hfsc-opt       = ( "default" | "red" | "ecn" | "rio" |
2936                 linkshare-sc | realtime-sc | upperlimit-sc )
2937linkshare-sc   = "linkshare" sc-spec
2938realtime-sc    = "realtime" sc-spec
2939upperlimit-sc  = "upperlimit" sc-spec
2940sc-spec        = ( bandwidth-spec |
2941                 "(" bandwidth-spec number bandwidth-spec ")" )
2942.Ed
2943.Sh FILES
2944.Bl -tag -width "/etc/protocols" -compact
2945.It Pa /etc/hosts
2946Host name database.
2947.It Pa /etc/pf.conf
2948Default location of the ruleset file.
2949.It Pa /etc/pf.os
2950Default location of OS fingerprints.
2951.It Pa /etc/protocols
2952Protocol name database.
2953.It Pa /etc/services
2954Service name database.
2955.It Pa /usr/share/examples/pf
2956Example rulesets.
2957.El
2958.Sh SEE ALSO
2959.Xr carp 4 ,
2960.Xr icmp 4 ,
2961.Xr icmp6 4 ,
2962.Xr ip 4 ,
2963.Xr ip6 4 ,
2964.Xr pf 4 ,
2965.\" .Xr pfsync 4 ,
2966.Xr route 4 ,
2967.Xr tcp 4 ,
2968.Xr udp 4 ,
2969.Xr hosts 5 ,
2970.Xr pf.os 5 ,
2971.Xr protocols 5 ,
2972.Xr services 5 ,
2973.Xr ftp-proxy 8 ,
2974.Xr pfctl 8 ,
2975.Xr pflogd 8 ,
2976.Xr route 8
2977.Sh HISTORY
2978The
2979.Nm
2980file format first appeared in
2981.Ox 3.0 .
2982