xref: /openbsd/share/man/man5/pf.conf.5 (revision 771fbea0)
1.\"	$OpenBSD: pf.conf.5,v 1.586 2021/02/01 00:31:04 dlg Exp $
2.\"
3.\" Copyright (c) 2002, Daniel Hartmeier
4.\" Copyright (c) 2003 - 2013 Henning Brauer <henning@openbsd.org>
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 $Mdocdate: February 1 2021 $
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 .
43.Pp
44This is an overview of the sections in this manual page:
45.Bl -inset
46.It Sx PACKET FILTERING
47including network address translation (NAT).
48.It Sx OPTIONS
49globally tune the behaviour of the packet filtering engine.
50.It Sx QUEUEING
51provides rule-based bandwidth and traffic control.
52.It Sx TABLES
53provide a method for dealing with large numbers of addresses.
54.It Sx ANCHORS
55are containers for rules and tables.
56.It Sx STATEFUL FILTERING
57tracks packets by state.
58.It Sx TRAFFIC NORMALISATION
59includes scrub, fragment handling, and blocking spoofed traffic.
60.It Sx OPERATING SYSTEM FINGERPRINTING
61is a method for detecting a host's operating system.
62.It Sx EXAMPLES
63provides some example rulesets.
64.It Sx GRAMMAR
65provides a complete BNF grammar reference.
66.El
67.Pp
68The current line can be extended over multiple lines using a backslash
69.Pq Sq \e .
70Comments can be put anywhere in the file using a hash mark
71.Pq Sq # ,
72and extend to the end of the current line.
73Care should be taken when commenting out multi-line text:
74the comment is effective until the end of the entire block.
75.Pp
76Argument names not beginning with a letter, digit, or underscore
77must be quoted.
78.Pp
79Additional configuration files can be included with the
80.Ic include
81keyword, for example:
82.Bd -literal -offset indent
83include "/etc/pf/sub.filter.conf"
84.Ed
85.Pp
86Macros can be defined that will later be expanded in context.
87Macro names must start with a letter, digit, or underscore,
88and may contain any of those characters.
89Macro names may not be reserved words (for example
90.Ic pass ,
91.Cm in ,
92.Cm out ) .
93Macros are not expanded inside quotes.
94.Pp
95For example:
96.Bd -literal -offset indent
97ext_if = "kue0"
98all_ifs = "{" $ext_if lo0 "}"
99pass out on $ext_if from any to any
100pass in  on $ext_if proto tcp from any to any port 25
101.Ed
102.Sh PACKET FILTERING
103.Xr pf 4
104has the ability to
105.Ic block ,
106.Ic pass ,
107and
108.Ic match
109packets based on attributes of their layer 3
110and layer 4 headers.
111Filter rules determine which of these actions are taken;
112filter parameters specify the packets to which a rule applies.
113.Pp
114Each time a packet processed by the packet filter comes in on or
115goes out through an interface, the filter rules are evaluated in
116sequential order, from first to last.
117For
118.Ic block
119and
120.Ic pass ,
121the last matching rule decides what action is taken;
122if no rule matches the packet, the default action is to pass
123the packet without creating a state.
124For
125.Ic match ,
126rules are evaluated every time they match;
127the pass/block state of a packet remains unchanged.
128.Pp
129Most parameters are optional.
130If a parameter is specified, the rule only applies to packets with
131matching attributes.
132The matching for some parameters can be inverted with the
133.Cm !\&
134operator.
135Certain parameters can be expressed as lists, in which case
136.Xr pfctl 8
137generates all needed rule combinations.
138.Pp
139By default
140.Xr pf 4
141filters packets statefully:
142the first time a packet matches a
143.Ic pass
144rule, a state entry is created.
145The packet filter examines each packet to see if it matches an existing state.
146If it does, the packet is passed without evaluation of any rules.
147After the connection is closed or times out, the state entry is automatically
148removed.
149.Pp
150The following actions can be used in the filter:
151.Bl -tag -width Ds
152.It Ic block
153The packet is blocked.
154There are a number of ways in which a
155.Ic block
156rule can behave when blocking a packet.
157The default behaviour is to
158.Cm drop
159packets silently, however this can be overridden or made
160explicit either globally, by setting the
161.Cm block-policy
162option, or on a per-rule basis with one of the following options:
163.Pp
164.Bl -tag -width return-icmp6 -compact
165.It Cm drop
166The packet is silently dropped.
167.It Cm return
168This causes a TCP RST to be returned for TCP packets
169and an ICMP UNREACHABLE for other types of packets.
170.It Cm return-icmp
171.It Cm return-icmp6
172This causes ICMP messages to be returned for packets which match the rule.
173By default this is an ICMP UNREACHABLE message, however this
174can be overridden by specifying a message as a code or number.
175.It Cm return-rst
176This applies only to TCP packets,
177and issues a TCP RST which closes the connection.
178An optional parameter,
179.Cm ttl ,
180may be given with a TTL value.
181.El
182.Pp
183Options returning ICMP packets currently have no effect if
184.Xr pf 4
185operates on a
186.Xr bridge 4 ,
187as the code to support this feature has not yet been implemented.
188.Pp
189The simplest mechanism to block everything by default and only pass
190packets that match explicit rules is specify a first filter rule of:
191.Pp
192.Dl block all
193.It Ic match
194The packet is matched.
195This mechanism is used to provide fine grained filtering
196without altering the block/pass state of a packet.
197.Ic match
198rules differ from
199.Ic block
200and
201.Ic pass
202rules in that parameters are set every time a packet matches the
203rule, not only on the last matching rule.
204For the following parameters,
205this means that the parameter effectively becomes
206.Dq sticky
207until explicitly overridden:
208.Cm nat-to ,
209.Cm binat-to ,
210.Cm rdr-to ,
211.Cm queue ,
212.Cm rtable ,
213and
214.Cm scrub .
215.Pp
216.Cm log
217is different still,
218in that the action happens every time a rule matches
219i.e. a single packet can get logged more than once.
220.It Ic pass
221The packet is passed;
222state is created unless the
223.Cm no state
224option is specified.
225.El
226.Pp
227The following parameters can be used in the filter:
228.Bl -tag -width Ds
229.It Cm in No or Cm out
230A packet always comes in on, or goes out through, one interface.
231.Cm in
232and
233.Cm out
234apply to incoming and outgoing packets;
235if neither are specified,
236the rule will match packets in both directions.
237.It Cm log Pq Cm all | matches | to Ar interface | Cm user
238In addition to any action specified,
239log the packet.
240Only the packet that establishes the state is logged,
241unless the
242.Cm no state
243option is specified.
244The logged packets are sent to a
245.Xr pflog 4
246interface, by default
247.Pa pflog0 ;
248pflog0 is monitored by the
249.Xr pflogd 8
250logging daemon which logs to the file
251.Pa /var/log/pflog
252in pcap binary format.
253.Pp
254The keywords
255.Cm all , matches , to ,
256and
257.Cm user
258are optional and can be combined using commas,
259but must be enclosed in parentheses if given.
260.Pp
261Use
262.Cm all
263to force logging of all packets for a connection.
264This is not necessary when
265.Cm no state
266is explicitly specified.
267.Pp
268If
269.Cm matches
270is specified,
271it logs the packet on all subsequent matching rules.
272It is often combined with
273.Cm to Ar interface
274to avoid adding noise to the default log file.
275.Pp
276The keyword
277.Cm user
278logs the UID and PID of the
279socket on the local host used to send or receive a packet,
280in addition to the normal information.
281.Pp
282To specify a logging interface other than
283.Pa pflog0 ,
284use the syntax
285.Cm to Ar interface .
286.It Cm quick
287If a packet matches a rule which has the
288.Cm quick
289option set, this rule
290is considered the last matching rule, and evaluation of subsequent rules
291is skipped.
292.It Cm on Ar interface | Cm any
293This rule applies only to packets coming in on, or going out through, this
294particular interface or interface group.
295For more information on interface groups,
296see the
297.Ic group
298keyword in
299.Xr ifconfig 8 .
300.Cm any
301will match any existing interface except loopback ones.
302.It Cm on rdomain Ar number
303This rule applies only to packets coming in on, or going out through, this
304particular routing domain.
305.It Cm inet | inet6
306This rule applies only to packets of this address family.
307.It Cm proto Ar protocol
308This rule applies only to packets of this protocol.
309Common protocols are ICMP, ICMP6, TCP, and UDP.
310For a list of all the protocol name to number mappings used by
311.Xr pfctl 8 ,
312see the file
313.Pa /etc/protocols .
314.It Xo
315.Cm from Ar source
316.Cm port Ar source
317.Cm os Ar source
318.Cm to Ar dest
319.Cm port Ar dest
320.Xc
321This rule applies only to packets with the specified source and destination
322addresses and ports.
323.Pp
324Addresses can be specified in CIDR notation (matching netblocks), as
325symbolic host names, interface names or interface group names, or as any
326of the following keywords:
327.Pp
328.Bl -tag -width urpf-failed -compact
329.It Cm any
330Any address.
331.It Cm no-route
332Any address which is not currently routable.
333.It Cm route Ar label
334Any address matching the given
335.Xr route 8
336.Ar label .
337.It Cm self
338Expands to all addresses assigned to all interfaces.
339.It Pf < Ar table Ns >
340Any address matching the given
341.Ar table .
342.It Cm urpf-failed
343Any source address that fails a unicast reverse path forwarding (URPF)
344check, i.e. packets coming in on an interface other than that which holds
345the route back to the packet's source address.
346.El
347.Pp
348Ranges of addresses are specified using the
349.Sq -
350operator.
351For instance:
352.Dq 10.1.1.10 - 10.1.1.12
353means all addresses from 10.1.1.10 to 10.1.1.12,
354hence addresses 10.1.1.10, 10.1.1.11, and 10.1.1.12.
355.Pp
356Interface names, interface group names, and
357.Cm self
358can have modifiers appended:
359.Pp
360.Bl -tag -width :broadcast -compact
361.It Cm :0
362Do not include interface aliases.
363.It Cm :broadcast
364Translates to the interface's broadcast address(es).
365.It Cm :network
366Translates to the network(s) attached to the interface.
367.It Cm :peer
368Translates to the point-to-point interface's peer address(es).
369.El
370.Pp
371Host names may also have the
372.Cm :0
373modifier appended to restrict the name resolution to the first of each
374v4 and v6 address found.
375.Pp
376Host name resolution and interface to address translation are done at
377ruleset load-time.
378When the address of an interface (or host name) changes (under DHCP or PPP,
379for instance), the ruleset must be reloaded for the change to be reflected
380in the kernel.
381Surrounding the interface name (and optional modifiers) in parentheses
382changes this behaviour.
383When the interface name is surrounded by parentheses, the rule is
384automatically updated whenever the interface changes its address.
385The ruleset does not need to be reloaded.
386This is especially useful with NAT.
387.Pp
388Ports can be specified either by number or by name.
389For example, port 80 can be specified as
390.Cm www .
391For a list of all port name to number mappings used by
392.Xr pfctl 8 ,
393see the file
394.Pa /etc/services .
395.Pp
396Ports and ranges of ports are specified using these operators:
397.Bd -literal -offset indent
398=	(equal)
399!=	(unequal)
400<	(less than)
401<=	(less than or equal)
402>	(greater than)
403>=	(greater than or equal)
404:	(range including boundaries)
405><	(range excluding boundaries)
406<>	(except range)
407.Ed
408.Pp
409.Sq >< ,
410.Sq <>
411and
412.Sq \&:
413are binary operators (they take two arguments).
414For instance:
415.Bl -tag -width Ds
416.It Li port 2000:2004
417means
418.Sq all ports \(>= 2000 and \(<= 2004 ,
419hence ports 2000, 2001, 2002, 2003, and 2004.
420.It Li port 2000 >< 2004
421means
422.Sq all ports > 2000 and < 2004 ,
423hence ports 2001, 2002, and 2003.
424.It Li port 2000 <> 2004
425means
426.Sq all ports < 2000 or > 2004 ,
427hence ports 1\(en1999 and 2005\(en65535.
428.El
429.Pp
430The operating system of the source host can be specified in the case of TCP
431rules with the
432.Cm os
433modifier.
434See the
435.Sx OPERATING SYSTEM FINGERPRINTING
436section for more information.
437.Pp
438The
439.Cm host ,
440.Cm port ,
441and
442.Cm os
443specifications are optional, as in the following examples:
444.Bd -literal -offset indent
445pass in all
446pass in from any to any
447pass in proto tcp from any port < 1024 to any
448pass in proto tcp from any to any port 25
449pass in proto tcp from 10.0.0.0/8 port >= 1024 \e
450      to ! 10.1.2.3 port != ssh
451pass in proto tcp from any os "OpenBSD"
452pass in proto tcp from route "DTAG"
453.Ed
454.El
455.Pp
456The following additional parameters can be used in the filter:
457.Pp
458.Bl -tag -width Ds -compact
459.It Cm all
460This is equivalent to
461.Ql from any to any .
462.Pp
463.It Cm allow-opts
464By default, packets with IPv4 options or IPv6 hop-by-hop or destination
465options header are blocked.
466When
467.Cm allow-opts
468is specified for a
469.Ic pass
470rule, packets that pass the filter based on that rule (last matching)
471do so even if they contain options.
472For packets that match state, the rule that initially created the
473state is used.
474The implicit pass rule, that is used when a packet does not match
475any rules, does not allow IP options or option headers.
476Note that IPv6 packets with type 0 routing headers are always dropped.
477.Pp
478.It Cm divert-packet port Ar port
479Used to send matching packets to
480.Xr divert 4
481sockets bound to port
482.Ar port .
483If the default option of fragment reassembly is enabled, scrubbing with
484.Cm reassemble tcp
485is also enabled for
486.Cm divert-packet
487rules.
488.Pp
489.It Cm divert-reply
490Used to receive replies for sockets that are bound to addresses
491which are not local to the machine.
492See
493.Xr setsockopt 2
494for information on how to bind these sockets.
495.Pp
496.It Cm divert-to Ar host Cm port Ar port
497Used to redirect packets to a local socket bound to
498.Ar host
499and
500.Ar port .
501The packets will not be modified, so
502.Xr getsockname 2
503on the socket will return the original destination address of the packet.
504.Pp
505.It Cm flags Ar a Ns / Ns Ar b | Cm any
506This rule only applies to TCP packets that have the flags
507.Ar a
508set out of set
509.Ar b .
510Flags not specified in
511.Ar b
512are ignored.
513For stateful connections, the default is
514.Cm flags S/SA .
515To indicate that flags should not be checked at all, specify
516.Cm flags any .
517The flags are: (F)IN, (S)YN, (R)ST, (P)USH, (A)CK, (U)RG, (E)CE, and C(W)R.
518.Bl -tag -width "flags /SFRA"
519.It Cm flags S/S
520Flag SYN is set.
521The other flags are ignored.
522.It Cm flags S/SA
523This is the default setting for stateful connections.
524Out of SYN and ACK, exactly SYN may be set.
525SYN, SYN+PSH, and SYN+RST match, but SYN+ACK, ACK, and ACK+RST do not.
526This is more restrictive than the previous example.
527.It Cm flags /SFRA
528If the first set is not specified, it defaults to none.
529All of SYN, FIN, RST, and ACK must be unset.
530.El
531.Pp
532Because
533.Cm flags S/SA
534is applied by default (unless
535.Cm no state
536is specified), only the initial SYN packet of a TCP handshake will create
537a state for a TCP connection.
538It is possible to be less restrictive, and allow state creation from
539intermediate
540.Pq non-SYN
541packets, by specifying
542.Cm flags any .
543This will cause
544.Xr pf 4
545to synchronize to existing connections, for instance
546if one flushes the state table.
547However, states created from such intermediate packets may be missing
548connection details such as the TCP window scaling factor.
549States which modify the packet flow, such as those affected by
550.Cm af-to ,
551.Cm modulate state ,
552.Cm nat-to ,
553.Cm rdr-to ,
554or
555.Cm synproxy state
556options, or scrubbed with
557.Cm reassemble tcp ,
558will also not be recoverable from intermediate packets.
559Such connections will stall and time out.
560.Pp
561.It Cm group Ar group
562Similar to
563.Cm user ,
564this rule only applies to packets of sockets owned by the specified
565.Ar group .
566.Pp
567.It Cm icmp-type Ar type Cm code Ar code
568.It Cm icmp6-type Ar type Cm code Ar code
569This rule only applies to ICMP or ICMP6 packets with the specified type
570and code.
571Text names for ICMP types and codes are listed in
572.Xr icmp 4
573and
574.Xr icmp6 4 .
575The protocol and the ICMP type indicator
576.Po
577.Cm icmp-type
578or
579.Cm icmp6-type
580.Pc
581must match.
582.Pp
583.It Cm label Ar string
584Adds a label to the rule, which can be used to identify the rule.
585For instance,
586.Ql pfctl -s labels
587shows per-rule statistics for rules that have labels.
588.Pp
589The following macros can be used in labels:
590.Pp
591.Bl -tag -width "$srcaddrXXX" -compact -offset indent
592.It Va $dstaddr
593The destination IP address.
594.It Va $dstport
595The destination port specification.
596.It Va $if
597The interface.
598.It Va $nr
599The rule number.
600.It Va $proto
601The protocol name.
602.It Va $srcaddr
603The source IP address.
604.It Va $srcport
605The source port specification.
606.El
607.Pp
608For example:
609.Bd -literal -offset indent -compact
610ips = "{ 1.2.3.4, 1.2.3.5 }"
611pass in proto tcp from any to $ips \e
612      port > 1023 label "$dstaddr:$dstport"
613.Ed
614.Pp
615Expands to:
616.Bd -literal -offset indent -compact
617pass in inet proto tcp from any to 1.2.3.4 \e
618      port > 1023 label "1.2.3.4:>1023"
619pass in inet proto tcp from any to 1.2.3.5 \e
620      port > 1023 label "1.2.3.5:>1023"
621.Ed
622.Pp
623The macro expansion for the
624.Cm label
625directive occurs only at configuration file parse time, not during runtime.
626.Pp
627.It Cm max-pkt-rate Ar number Ns / Ns Ar seconds
628Measure the rate of packets matching the rule and states created by it.
629When the specified rate is exceeded, the rule stops matching.
630Only packets in the direction in which the state was created are considered,
631so that typically requests are counted and replies are not.
632For example,
633to pass up to 100 ICMP packets per 10 seconds:
634.Bd -literal -offset indent
635block in proto icmp
636pass in proto icmp max-pkt-rate 100/10
637.Ed
638.Pp
639When the rate is exceeded, all ICMP is blocked until the rate falls below
640100 per 10 seconds again.
641.Pp
642.It Cm once
643Creates a one shot rule that will remove itself from an active ruleset after
644the first match.
645In case this is the only rule in the anchor, the anchor will be destroyed
646automatically after the rule is matched.
647.Pp
648.It Cm probability Ar number Ns %
649A probability attribute can be attached to a rule,
650with a value set between 0 and 100%,
651in which case the rule is honoured using the given probability value.
652For example, the following rule will drop 20% of incoming ICMP packets:
653.Pp
654.Dl block in proto icmp probability 20%
655.Pp
656.It Cm prio Ar number
657Only match packets which have the given queueing priority assigned.
658.Pp
659.It Oo Cm \&! Oc Ns Cm received-on Ar interface
660Only match packets which were received on the specified
661.Cm interface
662(or interface group).
663.Cm any
664will match any existing interface except loopback ones.
665.Pp
666.It Cm rtable Ar number
667Used to select an alternate routing table for the routing lookup.
668Only effective before the route lookup happened, i.e. when filtering inbound.
669.Pp
670.It Cm set delay Ar milliseconds
671Packets matching this rule will be delayed at the outbound interface by the
672given number of milliseconds.
673.Pp
674.It Cm set prio Ar priority | Pq Ar priority , priority
675Packets matching this rule will be assigned a specific queueing priority.
676Priorities are assigned as integers 0 through 7,
677with a default priority of 3.
678If the packet is transmitted on a
679.Xr vlan 4
680interface, the queueing priority will also be written as the priority
681code point in the 802.1Q VLAN header.
682If two priorities are given, TCP ACKs with no data payload and packets
683which have a TOS of
684.Cm lowdelay
685will be assigned to the second one.
686Packets with a higher priority number are processed first,
687and packets with the same priority are processed
688in the order in which they are received.
689.Pp
690For example:
691.Bd -literal -offset indent
692pass in proto tcp to port 25 set prio 2
693pass in proto tcp to port 22 set prio (2, 5)
694.Ed
695.Pp
696The interface priority queues accessed by the
697.Cm set prio
698keyword are always enabled and do not require any additional
699configuration, unlike the queues described below and in the
700.Sx QUEUEING
701section.
702.Pp
703.It Cm set queue Ar queue | Pq Ar queue , queue
704Packets matching this rule will be assigned to the specified
705.Ar queue .
706If two queues are given, packets which have a TOS of
707.Cm lowdelay
708and TCP ACKs with no data payload will be assigned to the second one.
709See
710.Sx QUEUEING
711for setup details.
712.Pp
713For example:
714.Bd -literal -offset indent
715pass in proto tcp to port 25 set queue mail
716pass in proto tcp to port 22 set queue(ssh_bulk, ssh_prio)
717.Ed
718.Pp
719.It Cm set tos Ar string | number
720Enforces a TOS for matching packets.
721.Ar string
722may be one of
723.Cm critical ,
724.Cm inetcontrol ,
725.Cm lowdelay ,
726.Cm netcontrol ,
727.Cm throughput ,
728.Cm reliability ,
729or one of the DiffServ Code Points:
730.Cm ef ,
731.Cm af11 No ... Cm af43 ,
732.Cm cs0 No ... Cm cs7 ;
733.Ar number
734may be either a hex or decimal number.
735.Pp
736.It Cm tag Ar string
737Packets matching this rule will be tagged with the specified
738.Ar string .
739The tag acts as an internal marker that can be used to
740identify these packets later on.
741This can be used, for example, to provide trust between
742interfaces and to determine if packets have been
743processed by translation rules.
744Tags are
745.Dq sticky ,
746meaning that the packet will be tagged even if the rule
747is not the last matching rule.
748Further matching rules can replace the tag with a
749new one but will not remove a previously applied tag.
750A packet is only ever assigned one tag at a time.
751Tags take the same macros as labels (see above).
752.Pp
753.It Oo Cm \&! Oc Ns Cm tagged Ar string
754Used with filter or translation rules
755to specify that packets must already
756be tagged with the given
757.Ar string
758in order to match the rule.
759.Pp
760.It Cm tos Ar string | number
761This rule applies to packets with the specified TOS bits set.
762.Ar string
763may be one of
764.Cm critical ,
765.Cm inetcontrol ,
766.Cm lowdelay ,
767.Cm netcontrol ,
768.Cm throughput ,
769.Cm reliability ,
770or one of the DiffServ Code Points:
771.Cm ef ,
772.Cm af11 No ... Cm af43 ,
773.Cm cs0 No ... Cm cs7 ;
774.Ar number
775may be either a hex or decimal number.
776.Pp
777For example, the following rules are identical:
778.Bd -literal -offset indent
779pass all tos lowdelay
780pass all tos 0x10
781pass all tos 16
782.Ed
783.Pp
784.It Cm user Ar user
785This rule only applies to packets of sockets owned by the specified
786.Ar user .
787For outgoing connections initiated from the firewall, this is the user
788that opened the connection.
789For incoming connections to the firewall itself, this is the user that
790listens on the destination port.
791.Pp
792When listening sockets are bound to the wildcard address,
793.Xr pf 4
794cannot determine if a connection is destined for the firewall itself.
795To avoid false matches on just the destination port, combine a
796.Cm user
797rule with source or destination address
798.Cm self .
799.Pp
800All packets, both outgoing and incoming, of one connection are associated
801with the same user and group.
802Only TCP and UDP packets can be associated with users.
803.Pp
804The
805.Ar user
806and
807.Ar group
808arguments refer to the effective (as opposed to the real) IDs, in
809case the socket is created by a setuid/setgid process.
810User and group IDs are stored when a socket is created;
811when a process creates a listening socket as root (for instance, by
812binding to a privileged port) and subsequently changes to another
813user ID (to drop privileges), the credentials will remain root.
814.Pp
815User and group IDs can be specified as either numbers or names.
816The syntax is similar to the one for ports.
817The following example allows only selected users to open outgoing
818connections:
819.Bd -literal -offset indent
820block out proto tcp all
821pass  out proto tcp from self user { < 1000, dhartmei }
822.Ed
823.Pp
824The example below permits users with uid between 1000 and 1500
825to open connections:
826.Bd -literal -offset indent
827block out proto tcp all
828pass  out proto tcp from self user { 999 >< 1501 }
829.Ed
830.Pp
831The
832.Sq \&:
833operator, which works for port number matching, does not work for
834.Cm user
835and
836.Cm group
837match.
838.El
839.Ss Translation
840Translation options modify either the source or destination address and
841port of the packets associated with a stateful connection.
842.Xr pf 4
843modifies the specified address and/or port in the packet and recalculates
844IP, TCP, and UDP checksums as necessary.
845.Pp
846If specified on a
847.Ic match
848rule, subsequent rules will see packets as they look
849after any addresses and ports have been translated.
850These rules will therefore have to filter based on the translated
851address and port number.
852.Pp
853The state entry created permits
854.Xr pf 4
855to keep track of the original address for traffic associated with that state
856and correctly direct return traffic for that connection.
857.Pp
858Different types of translation are possible with pf:
859.Bl -tag -width binat-to
860.It Cm af-to
861Translation between different address families (NAT64) is handled
862using
863.Cm af-to
864rules.
865Because address family translation overrides the routing table, it's
866only possible to use
867.Cm af-to
868on inbound rules, and a source address for the resulting translation
869must always be specified.
870.Pp
871The optional second argument is the host or subnet the original
872addresses are translated into for the destination.
873The lowest bits of the original destination address form the host
874part of the new destination address according to the specified subnet.
875It is possible to embed a complete IPv4 address into an IPv6 address
876using a network prefix of /96 or smaller.
877.Pp
878When a destination address is not specified it is assumed that the host
879part is 32-bit long.
880For IPv6 to IPv4 translation this would mean using only the lower 32
881bits of the original IPv6 destination address.
882For IPv4 to IPv6 translation the destination subnet defaults to the
883subnet of the new IPv6 source address with a prefix length of /96.
884See RFC 6052 Section 2.2 for details on how the prefix determines the
885destination address encoding.
886.Pp
887For example, the following rules are identical:
888.Bd -literal -offset indent
889pass in inet af-to inet6 from 2001:db8::1 to 2001:db8::/96
890pass in inet af-to inet6 from 2001:db8::1
891.Ed
892.Pp
893In the above example the matching IPv4 packets will be modified to
894have a source address of 2001:db8::1 and a destination address will
895get prefixed with 2001:db8::/96, e.g. 198.51.100.100 will be
896translated to 2001:db8::c633:6464.
897.Pp
898In the reverse case the following rules are identical:
899.Bd -literal -offset indent
900pass in inet6 from any to 64:ff9b::/96 af-to inet \e
901	from 198.51.100.1 to 0.0.0.0/0
902pass in inet6 from any to 64:ff9b::/96 af-to inet \e
903	from 198.51.100.1
904.Ed
905.Pp
906The destination IPv4 address is assumed to be embedded inside the
907original IPv6 destination address, e.g. 64:ff9b::c633:6464 will be
908translated to 198.51.100.100.
909.Pp
910The current implementation will only extract IPv4 addresses from the
911IPv6 addresses with a prefix length of /96 and greater.
912.It Cm binat-to
913A
914.Cm binat-to
915rule specifies a bidirectional mapping between an external IP
916netblock and an internal IP netblock.
917It expands to an outbound
918.Cm nat-to
919rule and an inbound
920.Cm rdr-to
921rule.
922.It Cm nat-to
923A
924.Cm nat-to
925option specifies that IP addresses are to be changed as the packet
926traverses the given interface.
927This technique allows one or more IP addresses
928on the translating host to support network traffic for a larger range of
929machines on an
930.Dq inside
931network.
932Although in theory any IP address can be used on the inside, it is strongly
933recommended that one of the address ranges defined by RFC 1918 be used.
934Those netblocks are:
935.Bd -literal -offset indent
93610.0.0.0 \(en 10.255.255.255 (all of net 10, i.e. 10/8)
937172.16.0.0 \(en 172.31.255.255 (i.e. 172.16/12)
938192.168.0.0 \(en 192.168.255.255 (i.e. 192.168/16)
939.Ed
940.Pp
941.Cm nat-to
942is usually applied outbound.
943If applied inbound, nat-to to a local IP address is not supported.
944.It Cm rdr-to
945The packet is redirected to another destination and possibly a
946different port.
947.Cm rdr-to
948can optionally specify port ranges instead of single ports.
949For instance:
950.Bl -tag -width Ds
951.It match in ... port 2000:2999 rdr-to ... port 4000
952redirects ports 2000 to 2999 (inclusive) to port 4000.
953.It match in ... port 2000:2999 rdr-to ... port 4000:*
954redirects port 2000 to 4000, port 2001 to 4001, ..., port 2999 to 4999.
955.El
956.Pp
957.Cm rdr-to
958is usually applied inbound.
959If applied outbound, rdr-to to a local IP address is not supported.
960.El
961.Pp
962In addition to modifying the address, some translation rules may modify
963source or destination ports for TCP or UDP connections;
964implicitly in the case of
965.Cm nat-to
966options and explicitly in the case of
967.Cm rdr-to
968ones.
969Port numbers are never translated with a
970.Cm binat-to
971rule.
972.Pp
973Translation options apply only to packets that pass through the specified
974interface, and if no interface is specified, translation is applied
975to packets on all interfaces.
976For instance, redirecting port 80 on an external interface to an internal
977web server will only work for connections originating from the outside.
978Connections to the address of the external interface from local hosts will
979not be redirected, since such packets do not actually pass through the
980external interface.
981Redirections cannot reflect packets back through the interface they arrive
982on, they can only be redirected to hosts connected to different interfaces
983or to the firewall itself.
984.Pp
985However packets may be redirected to hosts connected to the interface the
986packet arrived on by using redirection with NAT.
987For example:
988.Bd -literal -offset indent
989pass in on $int_if proto tcp from $int_net to $ext_if port 80 \e
990	rdr-to $server
991pass out on $int_if proto tcp to $server port 80 \e
992	received-on $int_if nat-to $int_if
993.Ed
994.Pp
995Note that redirecting external incoming connections to the loopback address
996will effectively allow an external host to connect to daemons
997bound solely to the loopback address, circumventing the traditional
998blocking of such connections on a real interface.
999For example:
1000.Bd -literal -offset indent
1001pass in on egress proto tcp from any to any port smtp \e
1002	rdr-to 127.0.0.1 port spamd
1003.Ed
1004.Pp
1005Unless this effect is desired, any of the local non-loopback addresses
1006should be used instead as the redirection target, which allows external
1007connections only to daemons bound to this address or not bound to
1008any address.
1009.Pp
1010For
1011.Cm af-to ,
1012.Cm nat-to
1013and
1014.Cm rdr-to
1015options for which there is a single redirection address which has a
1016subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP
1017address), a variety of different methods for assigning this address can be
1018used:
1019.Bl -tag -width xxxx
1020.It Cm bitmask
1021The
1022.Cm bitmask
1023option applies the network portion of the redirection address to the address
1024to be modified (source with
1025.Cm nat-to ,
1026destination with
1027.Cm rdr-to ) .
1028.It Cm least-states Op Cm sticky-address
1029The
1030.Cm least-states
1031option selects the address with the least active states from
1032a given address pool and considers given weights
1033associated with address(es).
1034Weights can be specified between 1 and 65535.
1035Addresses with higher weights are selected more often.
1036.Pp
1037.Cm sticky-address
1038can be specified to ensure that multiple connections from the
1039same source are mapped to the same redirection address.
1040Associations are destroyed as soon as there are
1041no longer states which refer to them;
1042in order to make the mappings last
1043beyond the lifetime of the states,
1044increase the global options with
1045.Ic set Cm timeout src.track .
1046.It Cm random Op Cm sticky-address
1047The
1048.Cm random
1049option selects an address at random within the defined block of addresses.
1050.Cm sticky-address
1051is as described above.
1052.It Cm round-robin Op Cm sticky-address
1053The
1054.Cm round-robin
1055option loops through the redirection address(es) and considers given weights
1056associated with address(es).
1057Weights can be specified between 1 and 65535.
1058Addresses with higher weights are selected more often.
1059.Cm sticky-address
1060is as described above.
1061.It Cm source-hash Oo Ar key Oc Op Cm sticky-address
1062The
1063.Cm source-hash
1064option uses a hash of the source address to determine the redirection address,
1065ensuring that the redirection address is always the same for a given source.
1066An optional
1067.Ar key
1068can be specified after this keyword either in hex or as a string;
1069by default
1070.Xr pfctl 8
1071randomly generates a key for source-hash every time the
1072ruleset is reloaded.
1073.Cm sticky-address
1074is as described above.
1075.It Cm static-port
1076With
1077.Cm nat-to
1078rules, the
1079.Cm static-port
1080option prevents
1081.Xr pf 4
1082from modifying the source port on TCP and UDP packets.
1083.El
1084.Pp
1085When more than one redirection address or a table is specified,
1086.Cm bitmask
1087is not permitted as a pool type.
1088.Ss Routing
1089If a packet matches a rule with one of the following route options set,
1090the packet filter will route the packet according to the type of route option.
1091When such a rule creates state, the route option is also applied to all
1092packets matching the same connection.
1093.Bl -tag -width route-to
1094.It Cm dup-to
1095The
1096.Cm dup-to
1097option creates a duplicate of the packet and routes it like
1098.Cm route-to .
1099The original packet gets routed as it normally would.
1100.It Cm reply-to
1101The
1102.Cm reply-to
1103option is similar to
1104.Cm route-to ,
1105but routes packets that pass in the opposite direction (replies) to the
1106specified interface.
1107Opposite direction is only defined in the context of a state entry, and
1108.Cm reply-to
1109is useful only in rules that create state.
1110It can be used on systems with multiple external connections to
1111route all outgoing packets of a connection through the interface
1112the incoming connection arrived through (symmetric routing enforcement).
1113.It Cm route-to
1114The
1115.Cm route-to
1116option routes the packet to the specified destination address instead
1117of the destination address in the packet header.
1118When a
1119.Cm route-to
1120rule creates state, only packets that pass in the same direction as the
1121filter rule specifies will be routed in this way.
1122Packets passing in the opposite direction (replies) are not affected
1123and are routed normally.
1124.El
1125.Pp
1126For the
1127.Cm dup-to ,
1128.Cm reply-to ,
1129and
1130.Cm route-to
1131route options
1132for which there is a single redirection address which has a
1133subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP
1134address),
1135the methods
1136.Cm least-states ,
1137.Cm random ,
1138.Cm round-robin ,
1139and
1140.Cm source-hash ,
1141as described above,
1142can be used.
1143.Sh OPTIONS
1144.Xr pf 4
1145may be tuned for various situations using the
1146.Ic set
1147command.
1148.Bl -tag -width Ds
1149.It Ic set Cm block-policy drop | return
1150The
1151.Cm block-policy
1152option sets the default behaviour for the packet
1153.Ic block
1154action:
1155.Pp
1156.Bl -tag -width return -compact
1157.It Cm drop
1158Packet is silently dropped.
1159.It Cm return
1160A TCP RST is returned for blocked TCP packets,
1161an ICMP UNREACHABLE is returned for blocked UDP packets,
1162and all other packets are silently dropped.
1163.El
1164.Pp
1165The default value is
1166.Cm drop .
1167.It Ic set Cm debug Ar level
1168Set the debug
1169.Ar level ,
1170which limits the severity of log messages printed by
1171.Xr pf 4 .
1172This should be a keyword from the following ordered list
1173(highest to lowest):
1174.Cm emerg ,
1175.Cm alert ,
1176.Cm crit ,
1177.Cm err ,
1178.Cm warning ,
1179.Cm notice ,
1180.Cm info ,
1181and
1182.Cm debug .
1183These keywords correspond to the similar (LOG_) values specified to the
1184.Xr syslog 3
1185library routine.
1186The default value is
1187.Cm err .
1188.It Cm set Cm fingerprints Ar filename
1189Load fingerprints of known operating systems from the given
1190.Ar filename .
1191By default fingerprints of known operating systems are automatically
1192loaded from
1193.Xr pf.os 5 ,
1194but can be overridden via this option.
1195Setting this option may leave a small period of time where the fingerprints
1196referenced by the currently active ruleset are inconsistent until the new
1197ruleset finishes loading.
1198The default location for fingerprints is
1199.Pa /etc/pf.os .
1200.It Ic set Cm hostid Ar number
1201The 32-bit hostid
1202.Ar number
1203identifies this firewall's state table entries to other firewalls
1204in a
1205.Xr pfsync 4
1206failover cluster.
1207By default the hostid is set to a pseudo-random value, however it may be
1208desirable to manually configure it, for example to more easily identify the
1209source of state table entries.
1210The hostid may be specified in either decimal or hexadecimal.
1211.It Ic set Cm limit Ar limit-item number
1212Sets hard limits on the memory pools used by the packet filter.
1213See
1214.Xr pool 9
1215for an explanation of memory pools.
1216.Pp
1217For example,
1218to set the maximum number of entries in the memory pool used by state table
1219entries (generated by
1220.Ic pass
1221rules which do not specify
1222.Cm no state )
1223to 20000:
1224.Pp
1225.Dl set limit states 20000
1226.Pp
1227To set the maximum number of entries in the memory pool used for fragment
1228reassembly to 2000:
1229.Pp
1230.Dl set limit frags 2000
1231.Pp
1232This maximum may not exceed, and should be well below, the maximum number
1233of mbuf clusters
1234.Pq sysctl kern.maxclusters
1235in the system.
1236.Pp
1237To set the maximum number of entries in the memory pool used for tracking
1238source IP addresses (generated by the
1239.Cm sticky-address
1240and
1241.Cm src.track
1242options) to 2000:
1243.Pp
1244.Dl set limit src-nodes 2000
1245.Pp
1246To set limits on the memory pools used by tables:
1247.Bd -literal -offset indent
1248set limit tables 1000
1249set limit table-entries 100000
1250.Ed
1251.Pp
1252The first limits the number of tables that can exist to 1000.
1253The second limits the overall number of addresses that can be stored
1254in tables to 100000.
1255.Pp
1256Various limits can be combined on a single line:
1257.Bd -literal -offset indent
1258set limit { states 20000, frags 2000, src-nodes 2000 }
1259.Ed
1260.Pp
1261.Xr pf 4
1262has the following defaults:
1263.Bl -column table-entries PFR_KENTRY_HIWAT_SMALL platform_dependent
1264.It states Ta Dv PFSTATE_HIWAT Ta Pq 100000
1265.It tables Ta Dv PFR_KTABLE_HIWAT Ta Pq 1000
1266.It table-entries Ta Dv PFR_KENTRY_HIWAT Ta Pq 200000
1267.It table-entries Ta Dv PFR_KENTRY_HIWAT_SMALL Ta Pq 100000
1268.It frags Ta Dv NMBCLUSTERS Ns /32 Ta Pq platform dependent
1269.El
1270.Pp
1271.Dv NMBCLUSTERS
1272defines the total number of packets which can exist in-system at any one time.
1273Refer to
1274.In machine/param.h
1275for the platform-specific value.
1276.It Ic set Cm loginterface Ar interface | Cm none
1277Enable collection of packet and byte count statistics for the given
1278interface or interface group.
1279These statistics can be viewed using:
1280.Pp
1281.Dl # pfctl -s info
1282.Pp
1283In this example
1284.Xr pf 4
1285collects statistics on the interface named dc0:
1286.Pp
1287.Dl set loginterface dc0
1288.Pp
1289One can disable the loginterface using:
1290.Pp
1291.Dl set loginterface none
1292.Pp
1293The default value is
1294.Cm none .
1295.It Ic set Cm optimization Ar environment
1296Optimize state timeouts for one of the following network environments:
1297.Pp
1298.Bl -tag -width Ds -compact
1299.It Cm aggressive
1300Aggressively expire connections.
1301This can greatly reduce the memory usage of the firewall at the cost of
1302dropping idle connections early.
1303.It Cm conservative
1304Extremely conservative settings.
1305Avoid dropping legitimate connections at the
1306expense of greater memory utilization (possibly much greater on a busy
1307network) and slightly increased processor utilization.
1308.It Cm high-latency
1309A high-latency environment (such as a satellite connection).
1310.It Cm normal
1311A normal network environment.
1312Suitable for almost all networks.
1313.It Cm satellite
1314Alias for
1315.Cm high-latency .
1316.El
1317.Pp
1318The default value is
1319.Cm normal .
1320.It Ic set Cm reassemble yes | no Op Cm no-df
1321The
1322.Cm reassemble
1323option is used to enable or disable the reassembly of fragmented packets,
1324and can be set to
1325.Cm yes
1326(the default) or
1327.Cm no .
1328If
1329.Cm no-df
1330is also specified, fragments with the
1331.Dq dont-fragment
1332bit set are reassembled too,
1333instead of being dropped;
1334the reassembled packet will have the
1335.Dq dont-fragment
1336bit cleared.
1337The default value is
1338.Cm yes .
1339.It Ic set Cm ruleset-optimization Ar level
1340.Bl -tag -width profile -compact
1341.It Cm basic
1342Enable basic ruleset optimization.
1343This is the default behaviour.
1344Basic ruleset optimization does four things to improve the
1345performance of ruleset evaluations:
1346.Pp
1347.Bl -enum -compact
1348.It
1349remove duplicate rules
1350.It
1351remove rules that are a subset of another rule
1352.It
1353combine multiple rules into a table when advantageous
1354.It
1355reorder the rules to improve evaluation performance
1356.El
1357.Pp
1358.It Cm none
1359Disable the ruleset optimizer.
1360.It Cm profile
1361Uses the currently loaded ruleset as a feedback profile to tailor the
1362ordering of
1363.Cm quick
1364rules to actual network traffic.
1365.El
1366.Pp
1367It is important to note that the ruleset optimizer will modify the ruleset
1368to improve performance.
1369A side effect of the ruleset modification is that per-rule accounting
1370statistics will have different meanings than before.
1371If per-rule accounting is important for billing purposes or whatnot,
1372either the ruleset optimizer should not be used or a label field should
1373be added to all of the accounting rules to act as optimization barriers.
1374.Pp
1375Optimization can also be set as a command-line argument to
1376.Xr pfctl 8 ,
1377overriding the settings in
1378.Nm .
1379.It Ic set Cm skip on Ar ifspec
1380List interfaces for which packets should not be filtered.
1381Packets passing in or out on such interfaces are passed as if pf was
1382disabled, i.e. pf does not process them in any way.
1383This can be useful on loopback and other virtual interfaces, when
1384packet filtering is not desired and can have unexpected effects.
1385.Ar ifspec
1386is only evaluated when the ruleset is loaded; interfaces created
1387later will not be skipped.
1388PF filters traffic on all interfaces by default.
1389.It Ic set Cm state-defaults Ar state-option , ...
1390The
1391.Cm state-defaults
1392option sets the state options for states created from rules
1393without an explicit
1394.Cm keep state .
1395For example:
1396.Pp
1397.Dl set state-defaults pflow, no-sync
1398.It Ic set Cm state-policy if-bound | floating
1399The
1400.Cm state-policy
1401option sets the default behaviour for states:
1402.Pp
1403.Bl -tag -width if-bound -compact
1404.It Cm if-bound
1405States are bound to an interface.
1406.It Cm floating
1407States can match packets on any interfaces (the default).
1408.El
1409.It Ic set Cm syncookies never | always | adaptive
1410When
1411.Cm syncookies
1412are active, pf will answer each and every incoming TCP SYN with a
1413syncookie SYNACK, without allocating any resources.
1414Upon reception of the client's ACK in response to the syncookie
1415SYNACK, pf will evaluate the ruleset and create state if the ruleset
1416permits it, complete the three way handshake with the target host,
1417and continue the connection with synproxy in place.
1418This allows pf to be resilient against large synflood attacks,
1419which could otherwise exhaust the state table.
1420Due to the blind answers to each and every SYN,
1421syncookies share the caveats of synproxy:
1422seemingly accepting connections that will be dropped later on.
1423.Pp
1424.Bl -tag -width adaptive -compact
1425.It Cm never
1426pf will never send syncookie SYNACKs (the default).
1427.It Cm always
1428pf will always send syncookie SYNACKs.
1429.It Cm adaptive
1430pf will enable syncookie mode when a given percentage of the state table
1431is used up by half-open TCP connections, such as those that saw the initial
1432SYN but didn't finish the three way handshake.
1433The thresholds for entering and leaving syncookie mode can be specified using:
1434.Bd -literal -offset indent
1435set syncookies adaptive (start 25%, end 12%)
1436.Ed
1437.El
1438.It Ic set Cm timeout Ar variable value
1439.Bl -tag -width "src.track" -compact
1440.It Cm frag
1441Seconds before an unassembled fragment is expired (60 by default).
1442.It Cm interval
1443Interval between purging expired states and fragments (10 seconds by default).
1444.It Cm src.track
1445Length of time to retain a source tracking entry after the last state
1446expires (0 by default, which means there is no global limit.
1447The value is defined by the rule which creates the state.).
1448.El
1449.Pp
1450When a packet matches a stateful connection, the seconds to live for the
1451connection will be updated to that of the
1452protocol and modifier
1453which corresponds to the connection state.
1454Each packet which matches this state will reset the TTL.
1455Tuning these values may improve the performance of the
1456firewall at the risk of dropping valid idle connections.
1457.Pp
1458.Bl -tag -width Ds -compact
1459.It Cm tcp.closed Pq 90 seconds by default
1460The state after one endpoint sends an RST.
1461.It Cm tcp.closing Pq 900 seconds by default
1462The state after the first FIN has been sent.
1463.It Cm tcp.established Pq 24 hours by default
1464The fully established state.
1465.It Cm tcp.finwait Pq 45 seconds by default
1466The state after both FINs have been exchanged and the connection is closed.
1467Some hosts (notably web servers on Solaris) send TCP packets even after closing
1468the connection.
1469Increasing
1470.Cm tcp.finwait
1471(and possibly
1472.Cm tcp.closing )
1473can prevent blocking of such packets.
1474.It Cm tcp.first Pq 120 seconds by default
1475The state after the first packet.
1476.It Cm tcp.opening Pq 30 seconds by default
1477The state after the second packet but before both endpoints have
1478acknowledged the connection.
1479.El
1480.Pp
1481ICMP and UDP are handled in a fashion similar to TCP, but with a much more
1482limited set of states:
1483.Pp
1484.Bl -tag -width Ds -compact
1485.It Cm icmp.error Pq 10 seconds by default
1486The state after an ICMP error came back in response to an ICMP packet.
1487.It Cm icmp.first Pq 20 seconds by default
1488The state after the first packet.
1489.It Cm udp.first Pq 60 seconds by default
1490The state after the first packet.
1491.It Cm udp.multiple Pq 60 seconds by default
1492The state if both hosts have sent packets.
1493.It Cm udp.single Pq 30 seconds by default
1494The state if the source host sends more than one packet but the destination
1495host has never sent one back.
1496.El
1497.Pp
1498Other protocols are handled similarly to UDP:
1499.Pp
1500.Bl -tag -width xxxx -compact
1501.It Cm other.first Pq 60 seconds by default
1502.It Cm other.multiple Pq 60 seconds by default
1503.It Cm other.single Pq 30 seconds by default
1504.El
1505.Pp
1506Timeout values can be reduced adaptively as the number of state table
1507entries grows.
1508.Pp
1509.Bl -tag -width Ds -compact
1510.It Cm adaptive.start Pq 60000 states by default
1511When the number of state entries exceeds this value, adaptive scaling
1512begins.
1513All timeout values are scaled linearly with factor
1514(adaptive.end \- number of states) / (adaptive.end \- adaptive.start).
1515.It Cm adaptive.end Pq 120000 states by default
1516When reaching this number of state entries, all timeout values become
1517zero, effectively purging all state entries immediately.
1518This value is used to define the scale factor; it should not actually
1519be reached (set a lower state limit, see below).
1520.El
1521.Pp
1522Adaptive timeouts are enabled by default, with an adaptive.start value
1523equal to 60% of the state limit, and an adaptive.end value equal to
1524120% of the state limit.
1525They can be disabled by setting both adaptive.start and adaptive.end to 0.
1526.Pp
1527The adaptive timeout values can be defined both globally and for each rule.
1528When used on a per-rule basis, the values relate to the number of
1529states created by the rule, otherwise to the total number of
1530states.
1531.Pp
1532For example:
1533.Bd -literal -offset indent
1534set timeout tcp.first 120
1535set timeout tcp.established 86400
1536set timeout { adaptive.start 60000, adaptive.end 120000 }
1537set limit states 100000
1538.Ed
1539.Pp
1540With 9000 state table entries, the timeout values are scaled to 50%
1541(tcp.first 60, tcp.established 43200).
1542.El
1543.Pp
1544.Dq pfctl -F Reset
1545restores default values for the following options: debug, all limit options,
1546loginterface, reassemble, skip, syncookies, all timeouts.
1547.Sh QUEUEING
1548Packets can be assigned to queues for the purpose of bandwidth
1549control.
1550At least one declaration is required to configure queues, and later
1551any packet filtering rule can reference the defined queues by name.
1552When filtering, the last referenced
1553.Ar queue
1554name is where any passed packets will be queued, while for
1555blocked packets it specifies where any resulting ICMP or TCP RST
1556packets should be queued.
1557If the referenced queue does not exist on the outgoing interface,
1558the default queue for that interface is used.
1559Queues attached to an interface build a tree,
1560thus each queue can have further child queues.
1561Only leaf queues, i.e. queues without children, can be used to assign
1562packets to.
1563The root queue must specifically reference an interface, all other queues
1564pick up the interfaces they should be created on from their parent queues.
1565.Pp
1566In the following example, a queue named std is created on the interface em0,
1567with 3 child queues ssh, mail and http:
1568.Bd -literal -offset indent
1569queue std on em0 bandwidth 100M
1570queue ssh parent std bandwidth 10M
1571queue mail parent std bandwidth 10M
1572queue http parent std bandwidth 80M default
1573.Ed
1574.Pp
1575The specified bandwidth is the target bandwidth, every queue can receive
1576more bandwidth as long as the parent still has some available.
1577The maximum bandwidth that should be assigned to a given queue can be limited
1578using the
1579.Cm max
1580keyword.
1581If a limitation isn't imposed on the root queue, borrowing can result in
1582saturating the bandwidth of the outgoing interface.
1583Similarly, a minimum (reserved) bandwidth can be specified:
1584.Pp
1585.Dl queue ssh parent std bandwidth 10M min 5M max 25M
1586.Pp
1587For each of these 3 bandwidth specifications an additional burst bandwidth and
1588time can be specified:
1589.Pp
1590.Dl queue ssh parent std bandwidth 10M burst 90M for 100ms
1591.Pp
1592All
1593.Cm bandwidth
1594values are specified as bits per second or using the suffixes
1595.Cm K ,
1596.Cm M ,
1597and
1598.Cm G
1599to represent kilobits, megabits, and gigabits per second, respectively.
1600The value must not exceed the interface bandwidth.
1601.Pp
1602If multiple connections are assigned the same queue, they're not guaranteed
1603to share the queue bandwidth fairly.
1604An alternative flow queue manager can be used to achieve fair sharing by
1605indicating how many simultaneous states are expected with a
1606.Cm flows
1607option, unless a minimum bandwidth has been specified as well.
1608.Pp
1609When packets are classified by the stateful inspection engine, a flow
1610identifier is assigned to all packets belonging to the state,
1611thus limiting the number of individual flows that can be recognized
1612by the resolution of a flow identifier.
1613The current implementation is able to classify traffic into 32767 distinct
1614flows.
1615However, efficient fair sharing is observed even with a much smaller number
1616of flows.
1617For example on a 10Mbit/s DSL or a cable modem uplink, the following simple
1618configuration can be used:
1619.Bd -literal -offset 4n
1620queue outq on em0 bandwidth 9M max 9M flows 1024 qlimit 1024 \e
1621      default
1622.Ed
1623.Pp
1624It's important to specify the upper bound within 90-95% of the expected
1625bandwidth and raise the default queue limit.
1626.Pp
1627If a
1628.Cm flows
1629option appears without a
1630.Cm bandwidth
1631specification, the flow queue manager is selected as the queueing discipline
1632for the corresponding interface acting as a default queue for all outgoing
1633packets.
1634In such a scenario, a queueing hierarchy is not supported.
1635.Pp
1636In addition to the bandwidth and flow specifications, queues support the
1637following options:
1638.Bl -tag -width xxxx
1639.It Cm default
1640Packets not matched by another queue are assigned to this queue.
1641Exactly one default queue per interface is required.
1642.It Cm on Ar interface
1643Specifies the interface the queue operates on.
1644If not given, it operates on all matching interfaces.
1645.It Cm parent Ar name
1646Defines which parent queue the queue should be attached to.
1647Mandatory for all queues except root queues.
1648The parent queue must exist.
1649.It Cm quantum Ar size
1650Specifies the quantum of service for the flow queue manager.
1651The lower the quantum size the more advantage is given to streams of smaller
1652packets at the expense of bulk transfers.
1653The default value is set to the configured Maximum Transmission Unit (MTU)
1654of the specified interface.
1655.It Cm qlimit Ar limit
1656The maximum number of packets held in the queue.
1657The default is 50.
1658.El
1659.Pp
1660Packets can be assigned to queues based on filter rules by using the
1661.Cm queue
1662keyword.
1663Normally only one
1664.Ar queue
1665is specified; when a second one is specified it will instead be used for
1666packets which have a TOS of
1667.Cm lowdelay
1668and for TCP ACKs with no data payload.
1669.Pp
1670To continue the previous example, the examples below would specify the
1671four referenced
1672queues, plus a few child queues.
1673Interactive
1674.Xr ssh 1
1675sessions get a queue with a minimum bandwidth;
1676.Xr scp 1
1677and
1678.Xr sftp 1
1679bulk transfers go to a separate queue.
1680The queues are then referenced by filtering rules.
1681.Bd -literal -offset 4n
1682queue rootq on em0 bandwidth 100M max 100M
1683queue http parent rootq bandwidth 60M burst 90M for 100ms
1684queue  developers parent http bandwidth 45M
1685queue  employees parent http bandwidth 15M
1686queue mail parent rootq bandwidth 10M
1687queue ssh parent rootq bandwidth 20M
1688queue  ssh_interactive parent ssh bandwidth 10M min 5M
1689queue  ssh_bulk parent ssh bandwidth 10M
1690queue std parent rootq bandwidth 20M default
1691
1692block return out on em0 inet all set queue std
1693pass out on em0 inet proto tcp from $developerhosts to any port 80 \e
1694      set queue developers
1695pass out on em0 inet proto tcp from $employeehosts to any port 80 \e
1696      set queue employees
1697pass out on em0 inet proto tcp from any to any port 22 \e
1698      set queue(ssh_bulk, ssh_interactive)
1699pass out on em0 inet proto tcp from any to any port 25 \e
1700      set queue mail
1701.Ed
1702.Sh TABLES
1703Tables are named structures which can hold a collection of addresses and
1704networks.
1705Lookups against tables in
1706.Xr pf 4
1707are relatively fast, making a single rule with tables much more efficient,
1708in terms of
1709processor usage and memory consumption, than a large number of rules which
1710differ only in IP address (either created explicitly or automatically by rule
1711expansion).
1712.Pp
1713Tables can be used as the source or destination of filter
1714or translation rules.
1715They can also be used for the redirect address of
1716.Cm nat-to
1717and
1718.Cm rdr-to
1719and in the routing options of filter rules, but not for
1720.Cm bitmask
1721pools.
1722.Pp
1723Tables can be defined with any of the following
1724.Xr pfctl 8
1725mechanisms.
1726As with macros, reserved words may not be used as table names.
1727.Bl -tag -width "manually"
1728.It manually
1729Persistent tables can be manually created with the
1730.Cm add
1731or
1732.Cm replace
1733option of
1734.Xr pfctl 8 ,
1735before or after the ruleset has been loaded.
1736.It Nm
1737Table definitions can be placed directly in this file and loaded at the
1738same time as other rules are loaded, atomically.
1739Table definitions inside
1740.Nm
1741use the
1742.Ic table
1743statement, and are especially useful to define non-persistent tables.
1744The contents of a pre-existing table defined without a list of addresses
1745to initialize it is not altered when
1746.Nm
1747is loaded.
1748A table initialized with the empty list,
1749.Li { } ,
1750will be cleared on load.
1751.El
1752.Pp
1753Tables may be defined with the following attributes:
1754.Bl -tag -width counters
1755.It Cm const
1756The
1757.Cm const
1758flag prevents the user from altering the contents of the table once it
1759has been created.
1760Without that flag,
1761.Xr pfctl 8
1762can be used to add or remove addresses from the table at any time, even
1763when running with
1764.Xr securelevel 7
1765= 2.
1766.It Cm counters
1767The
1768.Cm counters
1769flag enables per-address packet and byte counters, which can be displayed with
1770.Xr pfctl 8 .
1771.It Cm persist
1772The
1773.Cm persist
1774flag forces the kernel to keep the table even when no rules refer to it.
1775If the flag is not set, the kernel will automatically remove the table
1776when the last rule referring to it is flushed.
1777.El
1778.Pp
1779This example creates a table called
1780.Dq private ,
1781to hold RFC 1918 private network blocks,
1782and a table called
1783.Dq badhosts ,
1784which is initially empty.
1785A filter rule is set up to block all traffic coming from addresses listed in
1786either table:
1787.Bd -literal -offset indent
1788table <private> const { 10/8, 172.16/12, 192.168/16 }
1789table <badhosts> persist
1790block on fxp0 from { <private>, <badhosts> } to any
1791.Ed
1792.Pp
1793The private table cannot have its contents changed and the badhosts table
1794will exist even when no active filter rules reference it.
1795Addresses may later be added to the badhosts table, so that traffic from
1796these hosts can be blocked by using the following:
1797.Pp
1798.Dl # pfctl -t badhosts -Tadd 204.92.77.111
1799.Pp
1800A table can also be initialized with an address list specified in one or more
1801external files, using the following syntax:
1802.Bd -literal -offset indent
1803table <spam> persist file "/etc/spammers" file "/etc/openrelays"
1804block on fxp0 from <spam> to any
1805.Ed
1806.Pp
1807The files
1808.Pa /etc/spammers
1809and
1810.Pa /etc/openrelays
1811list IP addresses, one per line.
1812Any lines beginning with a
1813.Sq #
1814are treated as comments and ignored.
1815In addition to being specified by IP address, hosts may also be
1816specified by their hostname.
1817When the resolver is called to add a hostname to a table,
1818.Em all
1819resulting IPv4 and IPv6 addresses are placed into the table.
1820IP addresses can also be entered in a table by specifying a valid interface
1821name, a valid interface group, or the
1822.Cm self
1823keyword, in which case all addresses assigned to the interface(s) will be
1824added to the table.
1825.Sh ANCHORS
1826Besides the main ruleset,
1827.Nm
1828can specify anchor attachment points.
1829An anchor is a container that can hold rules,
1830address tables, and other anchors.
1831When evaluation of the main ruleset reaches an
1832.Ic anchor
1833rule,
1834.Xr pf 4
1835will proceed to evaluate all rules specified in that anchor.
1836.Pp
1837The following example blocks all packets on the external interface by default,
1838then evaluates all rules in the anchor named "spam",
1839and finally passes all outgoing connections and
1840incoming connections to port 25:
1841.Bd -literal -offset indent
1842ext_if = "kue0"
1843block on $ext_if all
1844anchor spam
1845pass out on $ext_if all
1846pass in on $ext_if proto tcp from any to $ext_if port smtp
1847.Ed
1848.Pp
1849Anchors can be manipulated through
1850.Xr pfctl 8
1851without reloading the main ruleset or other anchors.
1852This loads a single rule into the anchor,
1853which blocks all packets from a specific address:
1854.Bd -literal -offset indent
1855# echo "block in quick from 1.2.3.4 to any" | pfctl -a spam -f -
1856.Ed
1857.Pp
1858The anchor can also be populated by adding a
1859.Ic load anchor
1860rule after the anchor rule.
1861When
1862.Xr pfctl 8
1863loads
1864.Nm ,
1865it will also load all the rules from the file
1866.Pa /etc/pf-spam.conf
1867into the anchor.
1868.Bd -literal -offset indent
1869anchor spam
1870load anchor spam from "/etc/pf-spam.conf"
1871.Ed
1872.Pp
1873An anchor rule can also contain a filter ruleset
1874in a brace-delimited block.
1875In that case, no separate loading of rules into the anchor
1876is required.
1877Brace delimited blocks may contain rules or other brace-delimited blocks.
1878When an anchor is populated this way the anchor name becomes optional.
1879Since the parser specification for anchor names is a string,
1880double quote characters
1881.Pq Sq \&"
1882should be placed around the anchor name.
1883.Bd -literal -offset indent
1884anchor "external" on egress {
1885	block
1886	anchor out {
1887		pass proto tcp from any to port { 25, 80, 443 }
1888	}
1889	pass in proto tcp to any port 22
1890}
1891.Ed
1892.Pp
1893Anchor rules can also specify packet filtering parameters
1894using the same syntax as filter rules.
1895When parameters are used,
1896the anchor rule is only evaluated for matching packets.
1897This allows conditional evaluation of anchors, like:
1898.Bd -literal -offset indent
1899block on $ext_if all
1900anchor spam proto tcp from any to any port smtp
1901pass out on $ext_if all
1902pass in on $ext_if proto tcp from any to $ext_if port smtp
1903.Ed
1904.Pp
1905The rules inside anchor "spam" are only evaluated
1906for TCP packets with destination port 25.
1907Hence, the following
1908will only block connections from 1.2.3.4 to port 25:
1909.Bd -literal -offset indent
1910# echo "block in quick from 1.2.3.4 to any" | pfctl -a spam -f -
1911.Ed
1912.Pp
1913Matching filter and translation rules marked with the
1914.Cm quick
1915option are final and abort the evaluation of the rules in other
1916anchors and the main ruleset.
1917If the anchor itself is marked with the
1918.Cm quick
1919option,
1920ruleset evaluation will terminate when the anchor is exited if the packet is
1921matched by any rule within the anchor.
1922.Pp
1923An anchor references other anchor attachment points
1924using the following syntax:
1925.Bl -tag -width xxxx
1926.It Ic anchor Ar name
1927Evaluates the filter rules in the specified anchor.
1928.El
1929.Pp
1930An anchor has a name which specifies the path where
1931.Xr pfctl 8
1932can be used to access the anchor to perform operations on it, such as
1933attaching child anchors to it or loading rules into it.
1934Anchors may be nested, with components separated by
1935.Sq /
1936characters, similar to how file system hierarchies are laid out.
1937The main ruleset is actually the default anchor, so filter and
1938translation rules, for example, may also be contained in any anchor.
1939.Pp
1940Anchor rules are evaluated relative to the anchor in which they are contained.
1941For example,
1942all anchor rules specified in the main ruleset will reference
1943anchor attachment points underneath the main ruleset,
1944and anchor rules specified in a file loaded from a
1945.Ic load anchor
1946rule will be attached under that anchor point.
1947.Pp
1948Anchors may end with the asterisk
1949.Pq Sq *
1950character, which signifies that all anchors attached at that point
1951should be evaluated in the alphabetical ordering of their anchor name.
1952For example,
1953the following
1954will evaluate each rule in each anchor attached to the "spam" anchor:
1955.Bd -literal -offset indent
1956anchor "spam/*"
1957.Ed
1958.Pp
1959Note that it will only evaluate anchors that are directly attached to the
1960"spam" anchor, and will not descend to evaluate anchors recursively.
1961.Pp
1962Since anchors are evaluated relative to the anchor in which they are
1963contained, there is a mechanism for accessing the parent and ancestor
1964anchors of a given anchor.
1965Similar to file system path name resolution, if the sequence
1966.Sq ..
1967appears as an anchor path component, the parent anchor of the current
1968anchor in the path evaluation at that point will become the new current
1969anchor.
1970As an example, consider the following:
1971.Bd -literal -offset indent
1972# printf 'anchor "spam/allowed"\en' | pfctl -f -
1973# printf 'anchor "../banned"\enpass\en' | pfctl -a spam/allowed -f -
1974.Ed
1975.Pp
1976Evaluation of the main ruleset will lead into the
1977spam/allowed anchor, which will evaluate the rules in the
1978spam/banned anchor, if any, before finally evaluating the
1979.Ic pass
1980rule.
1981.Sh STATEFUL FILTERING
1982.Xr pf 4
1983filters packets statefully,
1984which has several advantages.
1985For TCP connections, comparing a packet to a state involves checking
1986its sequence numbers, as well as TCP timestamps if a rule using the
1987.Cm reassemble tcp
1988parameter applies to the connection.
1989If these values are outside the narrow windows of expected
1990values, the packet is dropped.
1991This prevents spoofing attacks, such as when an attacker sends packets with
1992a fake source address/port but does not know the connection's sequence
1993numbers.
1994Similarly,
1995.Xr pf 4
1996knows how to match ICMP replies to states.
1997For example,
1998to allow echo requests (such as those created by
1999.Xr ping 8 )
2000out statefully and match incoming echo replies correctly to states:
2001.Pp
2002.Dl pass out inet proto icmp all icmp-type echoreq
2003.Pp
2004Also, looking up states is usually faster than evaluating rules.
2005If there are 50 rules, all of them are evaluated sequentially in O(n).
2006Even with 50000 states, only 16 comparisons are needed to match a
2007state, since states are stored in a binary search tree that allows
2008searches in O(log2\~n).
2009.Pp
2010Furthermore, correct handling of ICMP error messages is critical to
2011many protocols, particularly TCP.
2012.Xr pf 4
2013matches ICMP error messages to the correct connection, checks them against
2014connection parameters, and passes them if appropriate.
2015For example if an ICMP source quench message referring to a stateful TCP
2016connection arrives, it will be matched to the state and get passed.
2017.Pp
2018Finally, state tracking is required for
2019.Cm nat-to
2020and
2021.Cm rdr-to
2022options, in order to track address and port translations and reverse the
2023translation on returning packets.
2024.Pp
2025.Xr pf 4
2026will also create state for other protocols which are effectively stateless by
2027nature.
2028UDP packets are matched to states using only host addresses and ports,
2029and other protocols are matched to states using only the host addresses.
2030.Pp
2031If stateless filtering of individual packets is desired,
2032the
2033.Cm no state
2034keyword can be used to specify that state will not be created
2035if this is the last matching rule.
2036Note that packets which match neither block nor pass rules,
2037and thus are passed by default,
2038are effectively passed as if
2039.Cm no state
2040had been specified.
2041.Pp
2042A number of parameters can also be set to affect how
2043.Xr pf 4
2044handles state tracking,
2045as detailed below.
2046.Ss State Modulation
2047Much of the security derived from TCP is attributable to how well the
2048initial sequence numbers (ISNs) are chosen.
2049Some popular stack implementations choose
2050.Em very
2051poor ISNs and thus are normally susceptible to ISN prediction exploits.
2052By applying a
2053.Cm modulate state
2054rule to a TCP connection,
2055.Xr pf 4
2056will create a high quality random sequence number for each connection
2057endpoint.
2058.Pp
2059The
2060.Cm modulate state
2061directive implicitly keeps state on the rule and is
2062only applicable to TCP connections.
2063.Pp
2064For instance:
2065.Bd -literal -offset indent
2066block all
2067pass out proto tcp from any to any modulate state
2068pass in  proto tcp from any to any port 25 flags S/SFRA \e
2069      modulate state
2070.Ed
2071.Pp
2072Note that modulated connections will not recover when the state table
2073is lost (firewall reboot, flushing the state table, etc.).
2074.Xr pf 4
2075will not be able to infer a connection again after the state table flushes
2076the connection's modulator.
2077When the state is lost, the connection may be left dangling until the
2078respective endpoints time out the connection.
2079It is possible on a fast local network for the endpoints to start an ACK
2080storm while trying to resynchronize after the loss of the modulator.
2081The default
2082.Cm flags
2083settings (or a more strict equivalent) should be used on
2084.Cm modulate state
2085rules to prevent ACK storms.
2086.Pp
2087Note that alternative methods are available
2088to prevent loss of the state table
2089and allow for firewall failover.
2090See
2091.Xr carp 4
2092and
2093.Xr pfsync 4
2094for further information.
2095.Ss SYN Proxy
2096By default,
2097.Xr pf 4
2098passes packets that are part of a
2099TCP handshake between the endpoints.
2100The
2101.Cm synproxy state
2102option can be used to cause
2103.Xr pf 4
2104itself to complete the handshake with the active endpoint, perform a handshake
2105with the passive endpoint, and then forward packets between the endpoints.
2106.Pp
2107No packets are sent to the passive endpoint before the active endpoint has
2108completed the handshake, hence so-called SYN floods with spoofed source
2109addresses will not reach the passive endpoint, as the sender can't complete the
2110handshake.
2111.Pp
2112The proxy is transparent to both endpoints; they each see a single
2113connection from/to the other endpoint.
2114.Xr pf 4
2115chooses random initial sequence numbers for both handshakes.
2116Once the handshakes are completed, the sequence number modulators
2117(see previous section) are used to translate further packets of the
2118connection.
2119.Cm synproxy state
2120includes
2121.Cm modulate state .
2122.Pp
2123Rules with
2124.Cm synproxy state
2125will not work if
2126.Xr pf 4
2127operates on a
2128.Xr bridge 4 .
2129Also they act on incoming SYN packets only.
2130.Pp
2131Example:
2132.Bd -literal -offset indent
2133pass in proto tcp from any to any port www synproxy state
2134.Ed
2135.Ss Stateful Tracking Options
2136A number of options related to stateful tracking can be applied on a
2137per-rule basis.
2138One of
2139.Cm keep state ,
2140.Cm modulate state ,
2141or
2142.Cm synproxy state
2143must be specified explicitly to apply these options to a rule.
2144.Pp
2145.Bl -tag -width xxxx -compact
2146.It Cm floating
2147States can match packets on any interfaces
2148(the opposite of
2149.Cm if-bound ) .
2150This is the default.
2151.It Cm if-bound
2152States are bound to an interface
2153(the opposite of
2154.Cm floating ) .
2155.It Cm max Ar number
2156Limits the number of concurrent states the rule may create.
2157When this limit is reached, further packets that would create
2158state are dropped until existing states time out.
2159.It Cm no-sync
2160Prevent state changes for states created by this rule from appearing on the
2161.Xr pfsync 4
2162interface.
2163.It Cm pflow
2164States created by this rule are exported on the
2165.Xr pflow 4
2166interface.
2167.It Cm sloppy
2168Uses a sloppy TCP connection tracker that does not check sequence
2169numbers at all, which makes insertion and ICMP teardown attacks way
2170easier.
2171This is intended to be used in situations where one does not see all
2172packets of a connection, e.g. in asymmetric routing situations.
2173It cannot be used with
2174.Cm modulate state
2175or
2176.Cm synproxy state .
2177.It Ar timeout seconds
2178Changes the
2179.Ar timeout
2180values used for states created by this rule.
2181For a list of all valid
2182.Ar timeout
2183names, see
2184.Sx OPTIONS
2185above.
2186.El
2187.Pp
2188Multiple options can be specified, separated by commas:
2189.Bd -literal -offset indent
2190pass in proto tcp from any to any \e
2191      port www keep state \e
2192      (max 100, source-track rule, max-src-nodes 75, \e
2193      max-src-states 3, tcp.established 60, tcp.closing 5)
2194.Ed
2195.Pp
2196When the
2197.Cm source-track
2198keyword is specified, the number of states per source IP is tracked.
2199.Pp
2200.Bl -tag -width xxxx -compact
2201.It Cm source-track global
2202The number of states created by all rules that use this option is limited.
2203Each rule can specify different
2204.Cm max-src-nodes
2205and
2206.Cm max-src-states
2207options, however state entries created by any participating rule count towards
2208each individual rule's limits.
2209.It Cm source-track rule
2210The maximum number of states created by this rule is limited by the rule's
2211.Cm max-src-nodes
2212and
2213.Cm max-src-states
2214options.
2215Only state entries created by this particular rule count toward the rule's
2216limits.
2217.El
2218.Pp
2219The following limits can be set:
2220.Pp
2221.Bl -tag -width xxxx -compact
2222.It Cm max-src-nodes Ar number
2223Limits the maximum number of source addresses which can simultaneously
2224have state table entries.
2225.It Cm max-src-states Ar number
2226Limits the maximum number of simultaneous state entries that a single
2227source address can create with this rule.
2228.El
2229.Pp
2230For stateful TCP connections, limits on established connections (connections
2231which have completed the TCP 3-way handshake) can also be enforced
2232per source IP.
2233.Pp
2234.Bl -tag -width xxxx -compact
2235.It Cm max-src-conn Ar number
2236Limits the maximum number of simultaneous TCP connections which have
2237completed the 3-way handshake that a single host can make.
2238.It Cm max-src-conn-rate Ar number Ns / Ns Ar seconds
2239Limit the rate of new connections over a time interval.
2240The connection rate is an approximation calculated as a moving average.
2241.El
2242.Pp
2243When one of these limits is reached, further packets that would create
2244state are dropped until existing states time out.
2245.Pp
2246Because the 3-way handshake ensures that the source address is not being
2247spoofed, more aggressive action can be taken based on these limits.
2248With the
2249.Cm overload Pf < Ar table Ns >
2250state option, source IP addresses which hit either of the limits on
2251established connections will be added to the named
2252.Ar table .
2253This table can be used in the ruleset to block further activity from
2254the offending host, redirect it to a tarpit process, or restrict its
2255bandwidth.
2256.Pp
2257The optional
2258.Cm flush
2259keyword kills all states created by the matching rule which originate
2260from the host which exceeds these limits.
2261The
2262.Cm global
2263modifier to the
2264.Cm flush
2265command kills all states originating from the
2266offending host, regardless of which rule created the state.
2267.Pp
2268For example, the following rules will protect the webserver against
2269hosts making more than 100 connections in 10 seconds.
2270Any host which connects faster than this rate will have its address added
2271to the <bad_hosts> table and have all states originating from it flushed.
2272Any new packets arriving from this host will be dropped unconditionally
2273by the block rule.
2274.Bd -literal -offset indent
2275block quick from <bad_hosts>
2276pass in on $ext_if proto tcp to $webserver port www keep state \e
2277      (max-src-conn-rate 100/10, overload <bad_hosts> flush global)
2278.Ed
2279.Sh TRAFFIC NORMALISATION
2280Traffic normalisation is a broad umbrella term
2281for aspects of the packet filter which deal with
2282verifying packets, packet fragments, spoof traffic,
2283and other irregularities.
2284.Ss Scrub
2285Scrub involves sanitising packet content in such a way
2286that there are no ambiguities in packet interpretation on the receiving side.
2287It is invoked with the
2288.Cm scrub
2289option, added to regular rules.
2290.Pp
2291Parameters are specified enclosed in parentheses.
2292At least one of the following parameters must be specified:
2293.Bl -tag -width xxxx
2294.It Cm max-mss Ar number
2295Enforces a maximum segment size (MSS) for matching TCP packets.
2296.It Cm min-ttl Ar number
2297Enforces a minimum TTL for matching IP packets.
2298.It Cm no-df
2299Clears the
2300.Dq dont-fragment
2301bit from a matching IPv4 packet.
2302Some operating systems have NFS implementations
2303which are known to generate fragmented packets with the
2304.Dq dont-fragment
2305bit set.
2306.Xr pf 4
2307will drop such fragmented
2308.Dq dont-fragment
2309packets unless
2310.Cm no-df
2311is specified.
2312.Pp
2313Unfortunately some operating systems also generate their
2314.Dq dont-fragment
2315packets with a zero IP identification field.
2316Clearing the
2317.Dq dont-fragment
2318bit on packets with a zero IP ID may cause deleterious results if an
2319upstream router later fragments the packet.
2320Using
2321.Cm random-id
2322is recommended in combination with
2323.Cm no-df
2324to ensure unique IP identifiers.
2325.It Cm random-id
2326Replaces the IPv4 identification field with random values to compensate
2327for predictable values generated by many hosts.
2328This option only applies to packets that are not fragmented
2329after the optional fragment reassembly.
2330.It Cm reassemble tcp
2331Statefully normalises TCP connections.
2332.Cm reassemble tcp
2333performs the following normalisations:
2334.Bl -ohang
2335.It TTL
2336Neither side of the connection is allowed to reduce their IP TTL.
2337An attacker may send a packet such that it reaches the firewall, affects
2338the firewall state, and expires before reaching the destination host.
2339.Cm reassemble tcp
2340will raise the TTL of all packets back up to the highest value seen on
2341the connection.
2342.It Timestamp Modulation
2343Modern TCP stacks will send a timestamp on every TCP packet and echo
2344the other endpoint's timestamp back to them.
2345Many operating systems will merely start the timestamp at zero when
2346first booted, and increment it several times a second.
2347The uptime of the host can be deduced by reading the timestamp and multiplying
2348by a constant.
2349Also observing several different timestamps can be used to count hosts
2350behind a NAT device.
2351And spoofing TCP packets into a connection requires knowing or guessing
2352valid timestamps.
2353Timestamps merely need to be monotonically increasing and not derived off a
2354guessable base time.
2355.Cm reassemble tcp
2356will cause
2357.Cm scrub
2358to modulate the TCP timestamps with a random number.
2359.It Extended PAWS Checks
2360There is a problem with TCP on long fat pipes, in that a packet might get
2361delayed for longer than it takes the connection to wrap its 32-bit sequence
2362space.
2363In such an occurrence, the old packet would be indistinguishable from a
2364new packet and would be accepted as such.
2365The solution to this is called PAWS: Protection Against Wrapped Sequence
2366numbers.
2367It protects against it by making sure the timestamp on each packet does
2368not go backwards.
2369.Cm reassemble tcp
2370also makes sure the timestamp on the packet does not go forward more
2371than the RFC allows.
2372By doing this,
2373.Xr pf 4
2374artificially extends the security of TCP sequence numbers by 10 to 18
2375bits when the host uses appropriately randomized timestamps, since a
2376blind attacker would have to guess the timestamp as well.
2377.El
2378.El
2379.Pp
2380For example:
2381.Pp
2382.Dl match in all scrub (no-df random-id max-mss 1440)
2383.Ss Fragment Handling
2384The size of IP datagrams (packets) can be significantly larger than the
2385maximum transmission unit (MTU) of the network.
2386In cases when it is necessary or more efficient to send such large packets,
2387the large packet will be fragmented into many smaller packets that will each
2388fit onto the wire.
2389Unfortunately for a firewalling device, only the first logical fragment will
2390contain the necessary header information for the subprotocol that allows
2391.Xr pf 4
2392to filter on things such as TCP ports or to perform NAT.
2393.Pp
2394One alternative is to filter individual fragments with filter rules.
2395If packet reassembly is turned off, it is passed to the filter.
2396Filter rules with matching IP header parameters decide whether the
2397fragment is passed or blocked, in the same way as complete packets
2398are filtered.
2399Without reassembly, fragments can only be filtered based on IP header
2400fields (source/destination address, protocol), since subprotocol header
2401fields are not available (TCP/UDP port numbers, ICMP code/type).
2402The
2403.Cm fragment
2404option can be used to restrict filter rules to apply only to
2405fragments, but not complete packets.
2406Filter rules without the
2407.Cm fragment
2408option still apply to fragments, if they only specify IP header fields.
2409For instance:
2410.Bd -literal -offset indent
2411pass in proto tcp from any to any port 80
2412.Ed
2413.Pp
2414The rule above never applies to a fragment,
2415even if the fragment is part of a TCP packet with destination port 80,
2416because without reassembly this information
2417is not available for each fragment.
2418This also means that fragments cannot create new or match existing
2419state table entries, which makes stateful filtering and address
2420translation (NAT, redirection) for fragments impossible.
2421.Pp
2422In most cases, the benefits of reassembly outweigh the additional
2423memory cost,
2424so reassembly is on by default.
2425.Pp
2426The memory allocated for fragment caching can be limited using
2427.Xr pfctl 8 .
2428Once this limit is reached, fragments that would have to be cached
2429are dropped until other entries time out.
2430The timeout value can also be adjusted.
2431.Pp
2432When forwarding reassembled IPv6 packets, pf refragments them with
2433the original maximum fragment size.
2434This allows the sender to determine the optimal fragment size by
2435path MTU discovery.
2436.Ss Blocking Spoofed Traffic
2437Spoofing is the faking of IP addresses,
2438typically for malicious purposes.
2439The
2440.Ic antispoof
2441directive expands to a set of filter rules which will block all
2442traffic with a source IP from the network(s) directly connected
2443to the specified interface(s) from entering the system through
2444any other interface.
2445.Pp
2446For example:
2447.Dl antispoof for lo0
2448.Pp
2449Expands to:
2450.Bd -literal -offset indent -compact
2451block drop in on ! lo0 inet from 127.0.0.1/8 to any
2452block drop in on ! lo0 inet6 from ::1 to any
2453.Ed
2454.Pp
2455For non-loopback interfaces, there are additional rules to block incoming
2456packets with a source IP address identical to the interface's IP(s).
2457For example, assuming the interface wi0 had an IP address of 10.0.0.1 and a
2458netmask of 255.255.255.0:
2459.Pp
2460.Dl antispoof for wi0 inet
2461.Pp
2462Expands to:
2463.Bd -literal -offset indent -compact
2464block drop in on ! wi0 inet from 10.0.0.0/24 to any
2465block drop in inet from 10.0.0.1 to any
2466.Ed
2467.Pp
2468Caveat: Rules created by the
2469.Ic antispoof
2470directive interfere with packets sent over loopback interfaces
2471to local addresses.
2472One should pass these explicitly.
2473.Sh OPERATING SYSTEM FINGERPRINTING
2474Passive OS fingerprinting is a mechanism to inspect nuances of a TCP
2475connection's initial SYN packet and guess at the host's operating system.
2476Unfortunately these nuances are easily spoofed by an attacker so the
2477fingerprint is not useful in making security decisions.
2478But the fingerprint is typically accurate enough to make policy decisions
2479upon.
2480.Pp
2481The fingerprints may be specified by operating system class, by
2482version, or by subtype/patchlevel.
2483The class of an operating system is typically the vendor or genre
2484and would be
2485.Ox
2486for the
2487.Xr pf 4
2488firewall itself.
2489The version of the oldest available
2490.Ox
2491release on the main FTP site
2492would be 2.6 and the fingerprint would be written as:
2493.Pp
2494.Dl \&"OpenBSD 2.6\&"
2495.Pp
2496The subtype of an operating system is typically used to describe the
2497patchlevel if that patch led to changes in the TCP stack behavior.
2498In the case of
2499.Ox ,
2500the only subtype is for a fingerprint that was
2501normalised by the
2502.Cm no-df
2503scrub option and would be specified as:
2504.Pp
2505.Dl \&"OpenBSD 3.3 no-df\&"
2506.Pp
2507Fingerprints for most popular operating systems are provided by
2508.Xr pf.os 5 .
2509Once
2510.Xr pf 4
2511is running, a complete list of known operating system fingerprints may
2512be listed by running:
2513.Pp
2514.Dl # pfctl -so
2515.Pp
2516Filter rules can enforce policy at any level of operating system specification
2517assuming a fingerprint is present.
2518Policy could limit traffic to approved operating systems or even ban traffic
2519from hosts that aren't at the latest service pack.
2520.Pp
2521The
2522.Cm unknown
2523class can also be used as the fingerprint which will match packets for
2524which no operating system fingerprint is known.
2525.Pp
2526Examples:
2527.Bd -literal -offset indent
2528pass  out proto tcp from any os OpenBSD
2529block out proto tcp from any os Doors
2530block out proto tcp from any os "Doors PT"
2531block out proto tcp from any os "Doors PT SP3"
2532block out from any os "unknown"
2533pass on lo0 proto tcp from any os "OpenBSD 3.3 lo0"
2534.Ed
2535.Pp
2536Operating system fingerprinting is limited only to the TCP SYN packet.
2537This means that it will not work on other protocols and will not match
2538a currently established connection.
2539.Pp
2540Caveat: operating system fingerprints are occasionally wrong.
2541There are three problems: an attacker can trivially craft his packets to
2542appear as any operating system he chooses;
2543an operating system patch could change the stack behavior and no fingerprints
2544will match it until the database is updated;
2545and multiple operating systems may have the same fingerprint.
2546.Sh EXAMPLES
2547In this example,
2548the external interface is
2549.Pa kue0 .
2550We use a macro for the interface name, so it can be changed easily.
2551All incoming traffic is "normalised",
2552and everything is blocked and logged by default.
2553.Bd -literal -offset 4n
2554ext_if = "kue0"
2555match in all scrub (no-df max-mss 1440)
2556block return log on $ext_if all
2557.Ed
2558.Pp
2559Here we specifically block packets we don't want:
2560anything coming from source we have no back routes for;
2561packets whose ingress interface does not match the one in
2562the route back to their source address;
2563anything that does not have our address (157.161.48.183) as source;
2564broadcasts (cable modem noise);
2565and anything from reserved address space or invalid addresses.
2566.Bd -literal -offset 4n
2567block in from no-route to any
2568block in from urpf-failed to any
2569block out log quick on $ext_if from ! 157.161.48.183 to any
2570block in quick on $ext_if from any to 255.255.255.255
2571block in log quick on $ext_if from { 10.0.0.0/8, 172.16.0.0/12, \e
2572    192.168.0.0/16, 255.255.255.255/32 } to any
2573.Ed
2574.Pp
2575For ICMP,
2576pass out/in ping queries.
2577State matching is done on host addresses and ICMP ID (not type/code),
2578so replies (like 0/0 for 8/0) will match queries.
2579ICMP error messages (which always refer to a TCP/UDP packet)
2580are handled by the TCP/UDP states.
2581.Bd -literal -offset 4n
2582pass on $ext_if inet proto icmp all icmp-type 8 code 0
2583.Ed
2584.Pp
2585For UDP,
2586pass out all UDP connections.
2587DNS connections are passed in.
2588.Bd -literal -offset 4n
2589pass out on $ext_if proto udp all
2590pass in on $ext_if proto udp from any to any port domain
2591.Ed
2592.Pp
2593For TCP,
2594pass out all TCP connections and modulate state.
2595SSH, SMTP, DNS, and IDENT connections are passed in.
2596We do not allow Windows 9x SMTP connections since they are typically
2597a viral worm.
2598.Bd -literal -offset 4n
2599pass out on $ext_if proto tcp all modulate state
2600pass in on $ext_if proto tcp from any to any \e
2601    port { ssh, smtp, domain, auth }
2602block in on $ext_if proto tcp from any \e
2603    os { "Windows 95", "Windows 98" } to any port smtp
2604.Ed
2605.Pp
2606Here we pass in/out all IPv6 traffic:
2607note that we have to enable this in two different ways,
2608on both our physical interface and our tunnel.
2609.Bd -literal -offset 4n
2610pass quick on gif0 inet6
2611pass quick on $ext_if proto ipv6
2612.Ed
2613.Pp
2614This example illustrates packet tagging.
2615There are three interfaces: $int_if, $ext_if, and $wifi_if (wireless).
2616NAT is being done on $ext_if for all outgoing packets.
2617Packets in on $int_if are tagged and passed out on $ext_if.
2618All other outgoing packets
2619(i.e. packets from the wireless network)
2620are only permitted to access port 80.
2621.Bd -literal -offset 4n
2622pass in on $int_if from any to any tag INTNET
2623pass in on $wifi_if from any to any
2624
2625block out on $ext_if from any to any
2626pass out quick on $ext_if tagged INTNET
2627pass out on $ext_if proto tcp from any to any port 80
2628.Ed
2629.Pp
2630In this example,
2631we tag incoming packets as they are redirected to
2632.Xr spamd 8 .
2633The tag is used to pass those packets through the packet filter.
2634.Bd -literal -offset 4n
2635match in on $ext_if inet proto tcp from <spammers> to port smtp \e
2636     tag SPAMD rdr-to 127.0.0.1 port spamd
2637
2638block in on $ext_if
2639pass in on $ext_if inet proto tcp tagged SPAMD
2640.Ed
2641.Pp
2642This example maps incoming requests on port 80 to port 8080, on
2643which a daemon is running (because, for example, it is not run as root,
2644and therefore lacks permission to bind to port 80).
2645.Bd -literal -offset 4n
2646match in on $ext_if proto tcp from any to any port 80 \e
2647      rdr-to 127.0.0.1 port 8080
2648.Ed
2649.Pp
2650If a
2651.Ic pass
2652rule is used with the
2653.Cm quick
2654modifier, packets matching the translation rule are passed without
2655inspecting subsequent filter rules.
2656.Bd -literal -offset 4n
2657pass in quick on $ext_if proto tcp from any to any port 80 \e
2658      rdr-to 127.0.0.1 port 8080
2659.Ed
2660.Pp
2661In the example below, vlan12 is configured as 192.168.168.1;
2662the machine translates all packets coming from 192.168.168.0/24 to 204.92.77.111
2663when they are going out any interface except vlan12.
2664This has the net effect of making traffic from the 192.168.168.0/24
2665network appear as though it is the Internet routable address
2666204.92.77.111 to nodes behind any interface on the router except
2667for the nodes on vlan12.
2668Thus, 192.168.168.1 can talk to the 192.168.168.0/24 nodes.
2669.Bd -literal -offset 4n
2670match out on ! vlan12 from 192.168.168.0/24 to any nat-to 204.92.77.111
2671.Ed
2672.Pp
2673In the example below, the machine sits between a fake internal
2674144.19.74.* network, and a routable external IP of 204.92.77.100.
2675The last rule excludes protocol AH from being translated.
2676.Bd -literal -offset 4n
2677pass out on $ext_if from 144.19.74.0/24 nat-to 204.92.77.100
2678pass out on $ext_if proto ah from 144.19.74.0/24
2679.Ed
2680.Pp
2681In the example below, packets bound for one specific server, as well as those
2682generated by the sysadmins are not proxied; all other connections are.
2683.Bd -literal -offset 4n
2684pass in on $int_if proto { tcp, udp } from any to any port 80 \e
2685      rdr-to 127.0.0.1 port 80
2686pass in on $int_if proto { tcp, udp } from any to $server port 80
2687pass in on $int_if proto { tcp, udp } from $sysadmins to any port 80
2688.Ed
2689.Pp
2690This example maps outgoing packets' source port
2691to an assigned proxy port instead of an arbitrary port.
2692In this case, proxy outgoing isakmp with port 500 on the gateway.
2693.Bd -literal -offset 4n
2694match out on $ext_if inet proto udp from any port isakmp to any \e
2695    nat-to ($ext_if) port 500
2696.Ed
2697.Pp
2698One more example uses
2699.Cm rdr-to
2700to redirect a TCP and UDP port to an internal machine.
2701.Bd -literal -offset 4n
2702match in on $ext_if inet proto tcp from any to ($ext_if) port 8080 \e
2703      rdr-to 10.1.2.151 port 22
2704match in on $ext_if inet proto udp from any to ($ext_if) port 8080 \e
2705      rdr-to 10.1.2.151 port 53
2706.Ed
2707.Pp
2708In this example, a NAT gateway is set up to translate internal addresses
2709using a pool of public addresses (192.0.2.16/28).
2710A given source address is always translated to the same pool address by
2711using the
2712.Cm source-hash
2713keyword.
2714The gateway also translates incoming web server connections
2715to a group of web servers on the internal network.
2716.Bd -literal -offset 4n
2717match out on $ext_if inet from any to any nat-to 192.0.2.16/28 \e
2718    source-hash
2719match in  on $ext_if proto tcp from any to any port 80 \e
2720    rdr-to { 10.1.2.155 weight 2, 10.1.2.160 weight 1, \e
2721             10.1.2.161 weight 8 } round-robin
2722.Ed
2723.Pp
2724The bidirectional address translation example uses a single
2725.Cm binat-to
2726rule that expands to a
2727.Cm nat-to
2728and an
2729.Cm rdr-to
2730rule.
2731.Bd -literal -offset 4n
2732pass on $ext_if from 10.1.2.120 to any binat-to 192.0.2.17
2733.Ed
2734.Pp
2735The previous example is identical to the following set of rules:
2736.Bd -literal -offset 4n
2737pass out on $ext_if inet from 10.1.2.120 to any \e
2738      nat-to 192.0.2.17 static-port
2739pass in on $ext_if inet from any to 192.0.2.17 rdr-to 10.1.2.120
2740.Ed
2741.Pp
2742In the example below, a router handling both address families
2743translates an internal IPv4 subnet to IPv6 using the well-known
274464:ff9b::/96 prefix:
2745.Bd -literal -offset 4n
2746pass in on $v4_if inet af-to inet6 from ($v6_if) to 64:ff9b::/96
2747.Ed
2748.Pp
2749Paired with the example above, the example below can be used on
2750another router handling both address families to translate back
2751to IPv4:
2752.Bd -literal -offset 4n
2753pass in on $v6_if inet6 to 64:ff9b::/96 af-to inet from ($v4_if)
2754.Ed
2755.Sh GRAMMAR
2756Syntax for
2757.Nm
2758in BNF:
2759.Bd -literal
2760line           = ( option | pf-rule |
2761                 antispoof-rule | queue-rule | anchor-rule |
2762                 anchor-close | load-anchor | table-rule | include )
2763
2764option         = "set" ( [ "timeout" ( timeout | "{" timeout-list "}" ) ] |
2765                 [ "ruleset-optimization" [ "none" | "basic" |
2766                 "profile" ] ] |
2767                 [ "optimization" [ "default" | "normal" | "high-latency" |
2768                 "satellite" | "aggressive" | "conservative" ] ]
2769                 [ "limit" ( limit-item | "{" limit-list "}" ) ] |
2770                 [ "loginterface" ( interface-name | "none" ) ] |
2771                 [ "block-policy" ( "drop" | "return" ) ] |
2772                 [ "state-policy" ( "if-bound" | "floating" ) ]
2773                 [ "state-defaults" state-opts ]
2774                 [ "fingerprints" filename ] |
2775                 [ "skip on" ifspec ] |
2776                 [ "debug" ( "emerg" | "alert" | "crit" | "err" |
2777                 "warning" | "notice" | "info" | "debug" ) ] |
2778                 [ "reassemble" ( "yes" | "no" ) [ "no-df" ] ] )
2779
2780pf-rule        = action [ ( "in" | "out" ) ]
2781                 [ "log" [ "(" logopts ")"] ] [ "quick" ]
2782                 [ "on" ( ifspec | "rdomain" number ) ] [ af ]
2783                 [ protospec ] [ hosts ] [ filteropts ]
2784
2785logopts        = logopt [ [ "," ] logopts ]
2786logopt         = "all" | "matches" | "user" | "to" interface-name
2787
2788filteropts     = filteropt [ [ "," ] filteropts ]
2789filteropt      = user | group | flags | icmp-type | icmp6-type |
2790                 "tos" tos |
2791                 ( "no" | "keep" | "modulate" | "synproxy" ) "state"
2792                 [ "(" state-opts ")" ] | "scrub" "(" scrubopts ")" |
2793                 "fragment" | "allow-opts" | "once" |
2794                 "divert-packet" "port" port | "divert-reply" |
2795                 "divert-to" host "port" port |
2796                 "label" string | "tag" string | [ "!" ] "tagged" string |
2797                 "max-pkt-rate" number "/" seconds |
2798                 "set delay" number |
2799                 "set prio" ( number | "(" number [ [ "," ] number ] ")" ) |
2800                 "set queue" ( string | "(" string [ [ "," ] string ] ")" ) |
2801                 "rtable" number | "probability" number"%" | "prio" number |
2802                 "af-to" af "from" ( redirhost | "{" redirhost-list "}" )
2803                 [ "to" ( redirhost | "{" redirhost-list "}" ) ] |
2804                 "binat-to" ( redirhost | "{" redirhost-list "}" )
2805                 [ portspec ] [ pooltype ] |
2806                 "rdr-to" ( redirhost | "{" redirhost-list "}" )
2807                 [ portspec ] [ pooltype ] |
2808                 "nat-to" ( redirhost | "{" redirhost-list "}" )
2809                 [ portspec ] [ pooltype ] [ "static-port" ] |
2810                 [ route ] | [ "set tos" tos ] |
2811                 [ [ "!" ] "received-on" ( interface-name | interface-group ) ]
2812
2813scrubopts      = scrubopt [ [ "," ] scrubopts ]
2814scrubopt       = "no-df" | "min-ttl" number | "max-mss" number |
2815                 "reassemble tcp" | "random-id"
2816
2817antispoof-rule = "antispoof" [ "log" ] [ "quick" ]
2818                 "for" ifspec [ af ] [ "label" string ]
2819
2820table-rule     = "table" "<" string ">" [ tableopts ]
2821tableopts      = tableopt [ tableopts ]
2822tableopt       = "persist" | "const" | "counters" |
2823                 "file" string | "{" [ tableaddrs ] "}"
2824tableaddrs     = tableaddr-spec [ [ "," ] tableaddrs ]
2825tableaddr-spec = [ "!" ] tableaddr [ "/" mask-bits ]
2826tableaddr      = hostname | ifspec | "self" |
2827                 ipv4-dotted-quad | ipv6-coloned-hex
2828
2829queue-rule     = "queue" string [ "on" interface-name ] queueopts-list
2830
2831anchor-rule    = "anchor" [ string ] [ ( "in" | "out" ) ] [ "on" ifspec ]
2832                 [ af ] [ protospec ] [ hosts ] [ filteropt-list ] [ "{" ]
2833
2834anchor-close   = "}"
2835
2836load-anchor    = "load anchor" string "from" filename
2837
2838queueopts-list = queueopts-list queueopts | queueopts
2839queueopts      = ([ "bandwidth" bandwidth ] | [ "min" bandwidth ] |
2840                 [ "max" bandwidth ] | [ "parent" string ] |
2841                 [ "default" ]) |
2842                 ([ "flows" number ] | [ "quantum" number ]) |
2843                 [ "qlimit" number ]
2844
2845bandwidth      = bandwidth-spec [ "burst" bandwidth-spec "for" number "ms" ]
2846bandwidth-spec = number ( "" | "K" | "M" | "G" )
2847
2848action         = "pass" | "match" | "block" [ return ]
2849return         = "drop" | "return" |
2850                 "return-rst" [ "(" "ttl" number ")" ] |
2851                 "return-icmp" [ "(" icmpcode [ [ "," ] icmp6code ] ")" ] |
2852                 "return-icmp6" [ "(" icmp6code ")" ]
2853icmpcode       = ( icmp-code-name | icmp-code-number )
2854icmp6code      = ( icmp6-code-name | icmp6-code-number )
2855
2856ifspec         = ( [ "!" ] ( interface-name | interface-group ) ) |
2857                 "{" interface-list "}"
2858interface-list = [ "!" ] ( interface-name | interface-group )
2859                 [ [ "," ] interface-list ]
2860route          = ( "route-to" | "reply-to" | "dup-to" )
2861                 ( redirhost | "{" redirhost-list "}" )
2862af             = "inet" | "inet6"
2863
2864protospec      = "proto" ( proto-name | proto-number |
2865                 "{" proto-list "}" )
2866proto-list     = ( proto-name | proto-number ) [ [ "," ] proto-list ]
2867
2868hosts          = "all" |
2869                 "from" ( "any" | "no-route" | "urpf-failed" | "self" |
2870                 host | "{" host-list "}" | "route" string ) [ port ]
2871                 [ os ]
2872                 "to"   ( "any" | "no-route" | "self" | host |
2873                 "{" host-list "}" | "route" string ) [ port ]
2874
2875ipspec         = "any" | host | "{" host-list "}"
2876host           = [ "!" ] ( address [ "weight" number ] |
2877                 address [ "/" mask-bits ] [ "weight" number ] |
2878                 "<" string ">" )
2879redirhost      = address [ "/" mask-bits ]
2880address        = ( interface-name | interface-group |
2881                 "(" ( interface-name | interface-group ) ")" |
2882                 hostname | ipv4-dotted-quad | ipv6-coloned-hex )
2883host-list      = host [ [ "," ] host-list ]
2884redirhost-list = redirhost [ [ "," ] redirhost-list ]
2885
2886port           = "port" ( unary-op | binary-op | "{" op-list "}" )
2887portspec       = "port" ( number | name ) [ ":" ( "*" | number | name ) ]
2888os             = "os"  ( os-name | "{" os-list "}" )
2889user           = "user" ( unary-op | binary-op | "{" op-list "}" )
2890group          = "group" ( unary-op | binary-op | "{" op-list "}" )
2891
2892unary-op       = [ "=" | "!=" | "<" | "<=" | ">" | ">=" ]
2893                 ( name | number )
2894binary-op      = number ( "<>" | "><" | ":" ) number
2895op-list        = ( unary-op | binary-op ) [ [ "," ] op-list ]
2896
2897os-name        = operating-system-name
2898os-list        = os-name [ [ "," ] os-list ]
2899
2900flags          = "flags" ( [ flag-set ] "/"  flag-set | "any" )
2901flag-set       = [ "F" ] [ "S" ] [ "R" ] [ "P" ] [ "A" ] [ "U" ] [ "E" ]
2902                 [ "W" ]
2903
2904icmp-type      = "icmp-type" ( icmp-type-code | "{" icmp-list "}" )
2905icmp6-type     = "icmp6-type" ( icmp-type-code | "{" icmp-list "}" )
2906icmp-type-code = ( icmp-type-name | icmp-type-number )
2907                 [ "code" ( icmp-code-name | icmp-code-number ) ]
2908icmp-list      = icmp-type-code [ [ "," ] icmp-list ]
2909
2910tos            = ( "lowdelay" | "throughput" | "reliability" |
2911                 [ "0x" ] number )
2912
2913state-opts     = state-opt [ [ "," ] state-opts ]
2914state-opt      = ( "max" number | "no-sync" | timeout | "sloppy" |
2915                 "pflow" | "source-track" [ ( "rule" | "global" ) ] |
2916                 "max-src-nodes" number | "max-src-states" number |
2917                 "max-src-conn" number |
2918                 "max-src-conn-rate" number "/" number |
2919                 "overload" "<" string ">" [ "flush" [ "global" ] ] |
2920                 "if-bound" | "floating" )
2921
2922timeout-list   = timeout [ [ "," ] timeout-list ]
2923timeout        = ( "tcp.first" | "tcp.opening" | "tcp.established" |
2924                 "tcp.closing" | "tcp.finwait" | "tcp.closed" |
2925                 "udp.first" | "udp.single" | "udp.multiple" |
2926                 "icmp.first" | "icmp.error" |
2927                 "other.first" | "other.single" | "other.multiple" |
2928                 "frag" | "interval" | "src.track" |
2929                 "adaptive.start" | "adaptive.end" ) number
2930
2931limit-list     = limit-item [ [ "," ] limit-list ]
2932limit-item     = ( "states" | "frags" | "src-nodes" | "tables" |
2933                 "table-entries" ) number
2934
2935pooltype       = ( "bitmask" | "least-states" |
2936                 "random" | "round-robin" |
2937                 "source-hash" [ ( hex-key | string-key ) ] )
2938                 [ "sticky-address" ]
2939
2940include        = "include" filename
2941.Ed
2942.Sh FILES
2943.Bl -tag -width /etc/examples/pf.conf -compact
2944.It Pa /etc/hosts
2945Host name database.
2946.It Pa /etc/pf.conf
2947Default location of the ruleset file.
2948.It Pa /etc/examples/pf.conf
2949Example ruleset file.
2950.It Pa /etc/pf.os
2951Default location of OS fingerprints.
2952.It Pa /etc/protocols
2953Protocol name database.
2954.It Pa /etc/services
2955Service name database.
2956.El
2957.Sh SEE ALSO
2958.Xr pf 4 ,
2959.Xr pflow 4 ,
2960.Xr pfsync 4 ,
2961.Xr pf.os 5 ,
2962.Xr pfctl 8 ,
2963.Xr pflogd 8
2964.Sh HISTORY
2965The
2966.Nm
2967file format first appeared in
2968.Ox 3.0 .
2969