xref: /openbsd-src/share/man/man5/pf.conf.5 (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1.\"	$OpenBSD: pf.conf.5,v 1.508 2011/08/30 00:47:16 mikeb Exp $
2.\"
3.\" Copyright (c) 2002, Daniel Hartmeier
4.\" All rights reserved.
5.\"
6.\" Redistribution and use in source and binary forms, with or without
7.\" modification, are permitted provided that the following conditions
8.\" are met:
9.\"
10.\"    - Redistributions of source code must retain the above copyright
11.\"      notice, this list of conditions and the following disclaimer.
12.\"    - Redistributions in binary form must reproduce the above
13.\"      copyright notice, this list of conditions and the following
14.\"      disclaimer in the documentation and/or other materials provided
15.\"      with the distribution.
16.\"
17.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18.\" "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19.\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
20.\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
21.\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
22.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
23.\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
24.\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
25.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
27.\" ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28.\" POSSIBILITY OF SUCH DAMAGE.
29.\"
30.Dd $Mdocdate: August 30 2011 $
31.Dt PF.CONF 5
32.Os
33.Sh NAME
34.Nm pf.conf
35.Nd packet filter configuration file
36.Sh DESCRIPTION
37The
38.Xr pf 4
39packet filter modifies, drops, or passes packets according to rules or
40definitions specified in
41.Nm .
42.Pp
43This is an overview of the sections in this manual page:
44.Bl -ohang
45.It Sy Packet Filtering
46Packet filtering, including network address translation (NAT).
47.It Sy Options
48Global options tune the behaviour of the packet filtering engine.
49.It Sy Queueing
50Queueing provides rule-based bandwidth control.
51.It Sy Tables
52Tables provide a method for dealing with large numbers of addresses.
53.It Sy Anchors
54Anchors are containers for rules and tables.
55.It Sy Stateful Filtering
56Stateful filtering tracks packets by state.
57.It Sy Traffic Normalisation
58Including scrub, fragment handling, and blocking spoofed traffic.
59.It Sy Operating System Fingerprinting
60A method for detecting a host's operating system.
61.It Sy Examples
62Some example rulesets.
63.El
64.Pp
65The current line can be extended over multiple lines using a backslash
66.Pq Sq \e .
67Comments can be put anywhere in the file using a hash mark
68.Pq Sq # ,
69and extend to the end of the current line.
70Care should be taken when commenting out multi-line text:
71the comment is effective until the end of the entire block.
72.Pp
73Additional configuration files can be included with the
74.Ic include
75keyword, for example:
76.Bd -literal -offset indent
77include "/etc/pf/sub.filter.conf"
78.Ed
79.Pp
80Macros can be defined that will later be expanded in context.
81Macro names must start with a letter, and may contain letters, digits
82and underscores.
83Macro names may not be reserved words (for example
84.Ar pass ,
85.Ar in ,
86.Ar out ) .
87Macros are not expanded inside quotes.
88.Pp
89For example:
90.Bd -literal -offset indent
91ext_if = "kue0"
92all_ifs = "{" $ext_if lo0 "}"
93pass out on $ext_if from any to any
94pass in  on $ext_if proto tcp from any to any port 25
95.Ed
96.Sh PACKET FILTERING
97.Xr pf 4
98has the ability to
99.Ar block ,
100.Ar pass ,
101and
102.Ar match
103packets based on attributes of their layer 3
104and layer 4 headers.
105Filter rules determine which of these actions are taken;
106filter parameters specify the packets to which a rule applies.
107.Pp
108For each packet processed by the packet filter, the filter rules are
109evaluated in sequential order, from first to last.
110For
111.Ar block
112and
113.Ar pass ,
114the last matching rule decides what action is taken;
115if no rule matches the packet, the default action is to pass
116the packet.
117For
118.Ar match ,
119rules are evaluated every time they match;
120the pass/block state of a packet remains unchanged.
121.Pp
122Most parameters are optional.
123If a parameter is specified, the rule only applies to packets with
124matching attributes.
125Certain parameters can be expressed as lists, in which case
126.Xr pfctl 8
127generates all needed rule combinations.
128.Pp
129By default
130.Xr pf 4
131filters packets statefully:
132the first time a packet matches a
133.Ar pass
134rule, a state entry is created.
135The packet filter examines each packet to see if it matches an existing state.
136If it does, the packet is passed without evaluation of any rules.
137After the connection is closed or times out, the state entry is automatically
138removed.
139.Pp
140The following actions can be used in the filter:
141.Bl -tag -width xxxx
142.It Ar block
143The packet is blocked.
144There are a number of ways in which a
145.Ar block
146rule can behave when blocking a packet.
147The default behaviour is to
148.Ar drop
149packets silently, however this can be overridden or made
150explicit either globally, by setting the
151.Ar block-policy
152option, or on a per-rule basis with one of the following options:
153.Pp
154.Bl -tag -width "return-icmp6XXX" -compact
155.It Ar drop
156The packet is silently dropped.
157.It Ar return
158This causes a TCP RST to be returned for TCP packets
159and an ICMP UNREACHABLE for other types of packets.
160.It Ar return-icmp
161.It Ar return-icmp6
162This causes ICMP messages to be returned for packets which match the rule.
163By default this is an ICMP UNREACHABLE message, however this
164can be overridden by specifying a message as a code or number.
165.It Ar return-rst
166This applies only to TCP packets,
167and issues a TCP RST which closes the connection.
168An optional parameter,
169.Ar ttl ,
170may be given with a TTL value.
171.El
172.Pp
173Options returning ICMP packets currently have no effect if
174.Xr pf 4
175operates on a
176.Xr bridge 4 ,
177as the code to support this feature has not yet been implemented.
178.Pp
179The simplest mechanism to block everything by default and only pass
180packets that match explicit rules is specify a first filter rule of:
181.Pp
182.Dl block all
183.It Ar match
184The packet is matched.
185This mechanism is used to provide fine grained filtering
186without altering the block/pass state of a packet.
187.Ar match
188rules differ from block and pass rules in that
189parameters are set every time a packet matches the rule,
190not only on the last matching rule.
191For the following parameters,
192this means that the parameter effectively becomes
193.Dq sticky
194until explicitly overridden:
195.Ar nat-to ,
196.Ar binat-to ,
197.Ar rdr-to ,
198.Ar queue ,
199.Ar rtable ,
200and
201.Ar scrub .
202.Pp
203.Ar log
204is different still,
205in that the action happens every time a rule matches
206i.e. a single packet can get logged more than once.
207.It Ar pass
208The packet is passed;
209state is created unless the
210.Ar no state
211option is specified.
212.El
213.Pp
214The following parameters can be used in the filter:
215.Bl -tag -width Ds
216.It Ar in No or Ar out
217A packet always comes in on, or goes out through, one interface.
218.Ar in
219and
220.Ar out
221apply to incoming and outgoing packets;
222if neither are specified,
223the rule will match packets in both directions.
224.It Ar log
225In addition to the action specified, a log message is generated.
226Only the packet that establishes the state is logged,
227unless the
228.Ar no state
229option is specified.
230The logged packets are sent to a
231.Xr pflog 4
232interface, by default
233.Ar pflog0 .
234This interface is monitored by the
235.Xr pflogd 8
236logging daemon, which dumps the logged packets to the file
237.Pa /var/log/pflog
238in
239.Xr pcap 3
240binary format.
241.It Ar log Pq Ar all
242Used to force logging of all packets for a connection.
243This is not necessary when
244.Ar no state
245is explicitly specified.
246As with
247.Ar log ,
248packets are logged to
249.Xr pflog 4 .
250.It Ar log Pq Ar matches
251Used to force logging of this packet on all subsequent matching rules.
252.It Ar log Pq Ar user
253Logs the UID and PID of the
254socket on the local host used to send or receive a packet,
255in addition to the normal information.
256.It Ar log Pq Ar to Aq Ar interface
257Send logs to the specified
258.Xr pflog 4
259interface instead of
260.Ar pflog0 .
261.It Ar quick
262If a packet matches a rule which has the
263.Ar quick
264option set, this rule
265is considered the last matching rule, and evaluation of subsequent rules
266is skipped.
267.It Ar on Aq Ar interface
268This rule applies only to packets coming in on, or going out through, this
269particular interface or interface group.
270For more information on interface groups,
271see the
272.Ic group
273keyword in
274.Xr ifconfig 8 .
275.It Ar on Ar rdomain Aq Ar number
276This rule applies only to packets coming in on, or going out through, this
277particular routing domain.
278.It Aq Ar af
279This rule applies only to packets of this address family.
280Supported values are
281.Ar inet
282and
283.Ar inet6 .
284.It Ar proto Aq Ar protocol
285This rule applies only to packets of this protocol.
286Common protocols are ICMP, ICMP6, TCP, and UDP.
287For a list of all the protocol name to number mappings used by
288.Xr pfctl 8 ,
289see the file
290.Em /etc/protocols .
291.It Xo
292.Ar from Aq Ar source
293.Ar port Aq Ar source
294.Ar os Aq Ar source
295.Ar to Aq Ar dest
296.Ar port Aq Ar dest
297.Xc
298This rule applies only to packets with the specified source and destination
299addresses and ports.
300.Pp
301Addresses can be specified in CIDR notation (matching netblocks), as
302symbolic host names, interface names or interface group names, or as any
303of the following keywords:
304.Pp
305.Bl -tag -width xxxxxxxxxxxxxx -compact
306.It Ar any
307Any address.
308.It Ar no-route
309Any address which is not currently routable.
310.It Ar route Aq Ar label
311Any address matching the given
312.Xr route 8
313label.
314.It Aq Ar table
315Any address matching the given table.
316.It Ar urpf-failed
317Any source address that fails a unicast reverse path forwarding (URPF)
318check, i.e. packets coming in on an interface other than that which holds
319the route back to the packet's source address.
320.El
321.Pp
322Ranges of addresses are specified using the
323.Sq -
324operator.
325For instance:
326.Dq 10.1.1.10 - 10.1.1.12
327means all addresses from 10.1.1.10 to 10.1.1.12,
328hence addresses 10.1.1.10, 10.1.1.11, and 10.1.1.12.
329.Pp
330Interface names and interface group names can have modifiers appended:
331.Pp
332.Bl -tag -width xxxxxxxxxxxx -compact
333.It Ar :0
334Do not include interface aliases.
335.It Ar :broadcast
336Translates to the interface's broadcast address(es).
337.It Ar :network
338Translates to the network(s) attached to the interface.
339.It Ar :peer
340Translates to the point-to-point interface's peer address(es).
341.El
342.Pp
343Host names may also have the
344.Ar :0
345option appended to restrict the name resolution to the first of each
346v4 and v6 address found.
347.Pp
348Host name resolution and interface to address translation are done at
349ruleset load-time.
350When the address of an interface (or host name) changes (under DHCP or PPP,
351for instance), the ruleset must be reloaded for the change to be reflected
352in the kernel.
353Surrounding the interface name (and optional modifiers) in parentheses
354changes this behaviour.
355When the interface name is surrounded by parentheses, the rule is
356automatically updated whenever the interface changes its address.
357The ruleset does not need to be reloaded.
358This is especially useful with
359.Ar nat .
360.Pp
361Ports can be specified either by number or by name.
362For example, port 80 can be specified as
363.Em www .
364For a list of all port name to number mappings used by
365.Xr pfctl 8 ,
366see the file
367.Pa /etc/services .
368.Pp
369Ports and ranges of ports are specified using these operators:
370.Bd -literal -offset indent
371=	(equal)
372!=	(unequal)
373\*(Lt	(less than)
374\*(Le	(less than or equal)
375\*(Gt	(greater than)
376\*(Ge	(greater than or equal)
377:	(range including boundaries)
378\*(Gt\*(Lt	(range excluding boundaries)
379\*(Lt\*(Gt	(except range)
380.Ed
381.Pp
382.Sq \*(Gt\*(Lt ,
383.Sq \*(Lt\*(Gt
384and
385.Sq \&:
386are binary operators (they take two arguments).
387For instance:
388.Bl -tag -width Ds
389.It port 2000:2004
390means
391.Sq all ports \*(Ge 2000 and \*(Le 2004 ,
392hence ports 2000, 2001, 2002, 2003, and 2004.
393.It port 2000 \*(Gt\*(Lt 2004
394means
395.Sq all ports \*(Gt 2000 and \*(Lt 2004 ,
396hence ports 2001, 2002, and 2003.
397.It port 2000 \*(Lt\*(Gt 2004
398means
399.Sq all ports \*(Lt 2000 or \*(Gt 2004 ,
400hence ports 1\(en1999 and 2005\(en65535.
401.El
402.Pp
403The operating system of the source host can be specified in the case of TCP
404rules with the
405.Ar os
406modifier.
407See the
408.Sx OPERATING SYSTEM FINGERPRINTING
409section for more information.
410.Pp
411The host, port, and OS specifications are optional,
412as in the following examples:
413.Bd -literal -offset indent
414pass in all
415pass in from any to any
416pass in proto tcp from any port \*(Le 1024 to any
417pass in proto tcp from any to any port 25
418pass in proto tcp from 10.0.0.0/8 port \*(Gt 1024 \e
419      to ! 10.1.2.3 port != ssh
420pass in proto tcp from any os "OpenBSD"
421pass in proto tcp from route "DTAG"
422.Ed
423.El
424.Pp
425The following additional parameters can be used in the filter:
426.Pp
427.Bl -tag -width Ds -compact
428.It Ar all
429This is equivalent to "from any to any".
430.Pp
431.It Ar allow-opts
432By default, IPv4 packets with IP options or IPv6 packets with routing
433extension headers are blocked.
434When
435.Ar allow-opts
436is specified for a
437.Ar pass
438rule, packets that pass the filter based on that rule (last matching)
439do so even if they contain IP options or routing extension headers.
440For packets that match state, the rule that initially created the
441state is used.
442The implicit
443.Ar pass
444rule that is used when a packet does not match any rules does not
445allow IP options.
446.Pp
447.It Ar divert-packet Ar port Aq Ar port
448Used to send matching packets to
449.Xr divert 4
450sockets bound to port
451.Ar port .
452If the default option of fragment reassembly is enabled, scrubbing with
453.Ar reassemble tcp
454is also enabled for
455.Ar divert-packet
456rules.
457.Pp
458.It Ar divert-reply
459Used to receive replies for sockets that are bound to addresses
460which are not local to the machine.
461See
462.Xr setsockopt 2
463for information on how to bind these sockets.
464.Pp
465.It Xo Ar divert-to Aq Ar host
466.Ar port Aq Ar port
467.Xc
468Used to redirect packets to a local socket bound to
469.Ar host
470and
471.Ar port .
472The packets will not be modified, so
473.Xr getsockname 2
474on the socket will return the original destination address of the packet.
475.Pp
476.It Xo Ar flags Aq Ar a
477.Pf / Ns Aq Ar b
478.No \*(Ba Ar any
479.Xc
480This rule only applies to TCP packets that have the flags
481.Aq Ar a
482set out of set
483.Aq Ar b .
484Flags not specified in
485.Aq Ar b
486are ignored.
487For stateful connections, the default is
488.Ar flags S/SA .
489To indicate that flags should not be checked at all, specify
490.Ar flags any .
491The flags are: (F)IN, (S)YN, (R)ST, (P)USH, (A)CK, (U)RG, (E)CE, and C(W)R.
492.Bl -tag -width Fl
493.It Ar flags S/S
494Flag SYN is set.
495The other flags are ignored.
496.It Ar flags S/SA
497This is the default setting for stateful connections.
498Out of SYN and ACK, exactly SYN may be set.
499SYN, SYN+PSH, and SYN+RST match, but SYN+ACK, ACK, and ACK+RST do not.
500This is more restrictive than the previous example.
501.It Ar flags /SFRA
502If the first set is not specified, it defaults to none.
503All of SYN, FIN, RST, and ACK must be unset.
504.El
505.Pp
506Because
507.Ar flags S/SA
508is applied by default (unless
509.Ar no state
510is specified), only the initial SYN packet of a TCP handshake will create
511a state for a TCP connection.
512It is possible to be less restrictive, and allow state creation from
513intermediate
514.Pq non-SYN
515packets, by specifying
516.Ar flags any .
517This will cause
518.Xr pf 4
519to synchronize to existing connections, for instance
520if one flushes the state table.
521However, states created from such intermediate packets may be missing
522connection details such as the TCP window scaling factor.
523States which modify the packet flow, such as those affected by
524.Ar modulate ,
525.Ar nat-to ,
526.Ar rdr-to ,
527or
528.Ar synproxy state
529options, or scrubbed with
530.Ar reassemble tcp ,
531will also not be recoverable from intermediate packets.
532Such connections will stall and time out.
533.Pp
534.It Ar group Aq Ar group
535Similar to
536.Ar user ,
537this rule only applies to packets of sockets owned by the specified group.
538.Pp
539.It Xo Ar icmp-type Aq Ar type
540.Ar code Aq Ar code
541.Xc
542.It Xo Ar icmp6-type Aq Ar type
543.Ar code Aq Ar code
544.Xc
545This rule only applies to ICMP or ICMP6 packets with the specified type
546and code.
547Text names for ICMP types and codes are listed in
548.Xr icmp 4
549and
550.Xr icmp6 4 .
551The protocol and the ICMP type indicator
552.Po
553.Ar icmp-type
554or
555.Ar icmp6-type
556.Pc
557must match.
558.Pp
559.It Ar label Aq Ar string
560Adds a label to the rule, which can be used to identify the rule.
561For instance,
562.Dq pfctl -s labels
563shows per-rule statistics for rules that have labels.
564.Pp
565The following macros can be used in labels:
566.Pp
567.Bl -tag -width "$srcaddrXXX" -compact -offset indent
568.It Ar $dstaddr
569The destination IP address.
570.It Ar $dstport
571The destination port specification.
572.It Ar $if
573The interface.
574.It Ar $nr
575The rule number.
576.It Ar $proto
577The protocol name.
578.It Ar $srcaddr
579The source IP address.
580.It Ar $srcport
581The source port specification.
582.El
583.Pp
584For example:
585.Bd -literal -offset indent -compact
586ips = "{ 1.2.3.4, 1.2.3.5 }"
587pass in proto tcp from any to $ips \e
588      port \*(Gt 1023 label "$dstaddr:$dstport"
589.Ed
590.Pp
591Expands to:
592.Bd -literal -offset indent -compact
593pass in inet proto tcp from any to 1.2.3.4 \e
594      port \*(Gt 1023 label "1.2.3.4:\*(Gt1023"
595pass in inet proto tcp from any to 1.2.3.5 \e
596      port \*(Gt 1023 label "1.2.3.5:\*(Gt1023"
597.Ed
598.Pp
599The macro expansion for the
600.Ar label
601directive occurs only at configuration file parse time, not during runtime.
602.Pp
603.It Ar once
604Creates a one shot rule that will remove itself from an active ruleset after
605the first match.
606In case this is the only rule in the anchor, the anchor will be destroyed
607automatically after the rule is matched.
608.Pp
609.It Ar probability Aq Ar number
610A probability attribute can be attached to a rule,
611with a value set between 0 and 100%,
612in which case the rule is honoured using the given probability value.
613For example, the following rule will drop 20% of incoming ICMP packets:
614.Pp
615.Dl block in proto icmp probability 20%
616.Pp
617.It Xo Ar prio Aq Ar priority
618.No \*(Ba ( Aq Ar priority ,
619.Aq Ar priority )
620.Xc
621Packets matching this rule will be assigned a specific queueing priority.
622Priorities are assigned as integers 0 through 7,
623with a default priority of 3.
624If two priorities are given, packets which have a TOS of
625.Ar lowdelay
626and TCP ACKs with no data payload will be assigned to the second one.
627.Pp
628For example:
629.Bd -literal -offset indent
630pass in proto tcp to port 25 prio 2
631pass in proto tcp to port 22 prio (2, 5)
632.Ed
633.Pp
634The interface priority queues accessed by the
635.Ar prio
636keyword are always enabled and do not require any additional
637configuration, unlike the queues described below and in the
638.Sx QUEUEING
639section.
640.Pp
641.It Xo Ar queue Aq Ar queue
642.No \*(Ba ( Aq Ar queue ,
643.Aq Ar queue )
644.Xc
645Packets matching this rule will be assigned to the specified queue.
646If two queues are given, packets which have a TOS of
647.Ar lowdelay
648and TCP ACKs with no data payload will be assigned to the second one.
649See
650.Sx QUEUEING
651for setup details.
652.Pp
653For example:
654.Bd -literal -offset indent
655pass in proto tcp to port 25 queue mail
656pass in proto tcp to port 22 queue(ssh_bulk, ssh_prio)
657.Ed
658.Pp
659.It Ar received-on Aq Ar interface
660Only match packets which were received on the specified
661.Ar interface
662(or interface group).
663.Pp
664.It Ar rtable Aq Ar number
665Used to select an alternate routing table for the routing lookup.
666Only effective before the route lookup happened, i.e. when filtering inbound.
667.Pp
668.It Ar tag Aq Ar string
669Packets matching this rule will be tagged with the
670specified string.
671The tag acts as an internal marker that can be used to
672identify these packets later on.
673This can be used, for example, to provide trust between
674interfaces and to determine if packets have been
675processed by translation rules.
676Tags are
677.Qq sticky ,
678meaning that the packet will be tagged even if the rule
679is not the last matching rule.
680Further matching rules can replace the tag with a
681new one but will not remove a previously applied tag.
682A packet is only ever assigned one tag at a time.
683Tags take the same macros as labels (see above).
684.Pp
685.It Ar tagged Aq Ar string
686Used with filter or translation rules
687to specify that packets must already
688be tagged with the given tag in order to match the rule.
689Inverse tag matching can also be done
690by specifying the
691.Cm !\&
692operator before the
693.Ar tagged
694keyword.
695.Pp
696.It Xo Ar tos Aq Ar string
697.No \*(Ba Aq Ar number
698.Xc
699This rule applies to packets with the specified TOS bits set.
700.Ar string
701may be one of
702.Ar critical ,
703.Ar inetcontrol ,
704.Ar lowdelay ,
705.Ar netcontrol ,
706.Ar throughput ,
707.Ar reliability ,
708or one of the DiffServ Code Points:
709.Ar ef ,
710.Ar af11 ... af43 ,
711.Ar cs0 ... cs7 ;
712.Ar number
713may be either a hex or decimal number.
714.Pp
715For example, the following rules are identical:
716.Bd -literal -offset indent
717pass all tos lowdelay
718pass all tos 0x10
719pass all tos 16
720.Ed
721.Pp
722.It Ar user Aq Ar user
723This rule only applies to packets of sockets owned by the specified user.
724For outgoing connections initiated from the firewall, this is the user
725that opened the connection.
726For incoming connections to the firewall itself, this is the user that
727listens on the destination port.
728For forwarded connections, where the firewall is not a connection endpoint,
729the user and group are
730.Em unknown .
731.Pp
732All packets, both outgoing and incoming, of one connection are associated
733with the same user and group.
734Only TCP and UDP packets can be associated with users; for other protocols
735these parameters are ignored.
736.Pp
737User and group refer to the effective (as opposed to the real) IDs, in
738case the socket is created by a setuid/setgid process.
739User and group IDs are stored when a socket is created;
740when a process creates a listening socket as root (for instance, by
741binding to a privileged port) and subsequently changes to another
742user ID (to drop privileges), the credentials will remain root.
743.Pp
744User and group IDs can be specified as either numbers or names.
745The syntax is similar to the one for ports.
746The value
747.Ar unknown
748matches packets of forwarded connections.
749.Ar unknown
750can only be used with the operators
751.Cm =
752and
753.Cm != .
754Other constructs like
755.Cm user \*(Ge unknown
756are invalid.
757Forwarded packets with unknown user and group ID match only rules
758that explicitly compare
759.Ar unknown
760with the operators
761.Cm =
762or
763.Cm != .
764For instance
765.Cm user \*(Ge 0
766does not match forwarded packets.
767The following example allows only selected users to open outgoing
768connections:
769.Bd -literal -offset indent
770block out proto { tcp, udp } all
771pass  out proto { tcp, udp } all user { \*(Lt 1000, dhartmei }
772.Ed
773.El
774.Ss Translation
775Translation options modify either the source or destination address and
776port of the packets associated with a stateful connection.
777.Xr pf 4
778modifies the specified address and/or port in the packet and recalculates
779IP, TCP, and UDP checksums as necessary.
780.Pp
781Subsequent rules will see packets as they look
782after any addresses and ports have been translated.
783These rules will therefore have to filter based on the translated
784address and port number.
785.Pp
786The state entry created permits
787.Xr pf 4
788to keep track of the original address for traffic associated with that state
789and correctly direct return traffic for that connection.
790.Pp
791Different types of translation are possible with pf:
792.Bl -tag -width xxxxxxxx
793.It Ar binat-to
794A
795.Ar binat-to
796rule specifies a bidirectional mapping between an external IP
797netblock and an internal IP netblock.
798It expands to an outbound
799.Ar nat-to
800rule and an inbound
801.Ar rdr-to
802rule.
803.It Ar nat-to
804A
805.Ar nat-to
806option specifies that IP addresses are to be changed as the packet
807traverses the given interface.
808This technique allows one or more IP addresses
809on the translating host to support network traffic for a larger range of
810machines on an "inside" network.
811Although in theory any IP address can be used on the inside, it is strongly
812recommended that one of the address ranges defined by RFC 1918 be used.
813Those netblocks are:
814.Bd -literal -offset indent
81510.0.0.0 \(en 10.255.255.255 (all of net 10, i.e. 10/8)
816172.16.0.0 \(en 172.31.255.255 (i.e. 172.16/12)
817192.168.0.0 \(en 192.168.255.255 (i.e. 192.168/16)
818.Ed
819.Pp
820.Ar nat-to
821is usually applied outbound.
822If applied inbound, nat-to to a local IP address is not supported.
823.It Pa rdr-to
824The packet is redirected to another destination and possibly a
825different port.
826.Ar rdr-to
827can optionally specify port ranges instead of single ports.
828For instance:
829.Bl -tag -width Ds
830.It match in ... port 2000:2999 rdr-to ... port 4000
831redirects ports 2000 to 2999 (inclusive) to port 4000.
832.Pp
833.It match in ... port 2000:2999 rdr-to ... port 4000:*
834redirects port 2000 to 4000, port 2001 to 4001, ..., port 2999 to 4999.
835.El
836.Pp
837.Ar rdr-to
838is usually applied inbound.
839If applied outbound, rdr-to to a local IP address is not supported.
840.El
841.Pp
842In addition to modifying the address, some translation rules may modify
843source or destination ports for TCP or UDP connections;
844implicitly in the case of
845.Ar nat-to
846options and explicitly in the case of
847.Ar rdr-to
848ones.
849Port numbers are never translated with a
850.Ar binat-to
851rule.
852.Pp
853Translation options apply only to packets that pass through the specified
854interface, and if no interface is specified, translation is applied
855to packets on all interfaces.
856For instance, redirecting port 80 on an external interface to an internal
857web server will only work for connections originating from the outside.
858Connections to the address of the external interface from local hosts will
859not be redirected, since such packets do not actually pass through the
860external interface.
861Redirections cannot reflect packets back through the interface they arrive
862on, they can only be redirected to hosts connected to different interfaces
863or to the firewall itself.
864.Pp
865However packets may be redirected to hosts connected to the interface the
866packet arrived on by using redirection with NAT.
867For example:
868.Bd -literal -offset indent
869pass in on $int_if proto tcp from $int_net to $ext_if port 80 \e
870	rdr-to $server
871pass out on $int_if proto tcp to $server port 80 \e
872	received-on $int_if nat-to $int_if
873.Ed
874.Pp
875Note that redirecting external incoming connections to the loopback address
876will effectively allow an external host to connect to daemons
877bound solely to the loopback address, circumventing the traditional
878blocking of such connections on a real interface.
879For example:
880.Bd -literal -offset indent
881pass in on egress proto tcp from any to any port smtp \e
882	rdr-to 127.0.0.1 port spamd
883.Ed
884.Pp
885Unless this effect is desired, any of the local non-loopback addresses
886should be used instead as the redirection target, which allows external
887connections only to daemons bound to this address or not bound to
888any address.
889.Pp
890For
891.Ar nat-to
892and
893.Ar rdr-to
894options for which there is a single redirection address which has a
895subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP
896address), a variety of different methods for assigning this address can be
897used:
898.Bl -tag -width xxxx
899.It Ar bitmask
900The
901.Ar bitmask
902option applies the network portion of the redirection address to the address
903to be modified (source with
904.Ar nat-to ,
905destination with
906.Ar rdr-to ) .
907.It Ar least-states Op Ar sticky-address
908The
909.Ar least-states
910option selects the address with the least active states from
911a given address pool.
912.Pp
913.Ar sticky-address
914can be specified to ensure that multiple connections from the
915same source are mapped to the same redirection address.
916Associations are destroyed as soon as there are
917no longer states which refer to them;
918in order to make the mappings last
919beyond the lifetime of the states,
920increase the global options with
921.Ar set timeout src.track .
922.It Ar random Op Ar sticky-address
923The
924.Ar random
925option selects an address at random within the defined block of addresses.
926.Ar sticky-address
927is as described above.
928.It Ar round-robin Op Ar sticky-address
929The
930.Ar round-robin
931option loops through the redirection address(es) and considers given weights
932associated with address(es).
933Weights can be specified between 1 and 65535.
934Addresses with higher weights are selected more often.
935.Ar sticky-address
936is as described above.
937.It Ar source-hash Op Ar key
938The
939.Ar source-hash
940option uses a hash of the source address to determine the redirection address,
941ensuring that the redirection address is always the same for a given source.
942An optional
943.Ar key
944can be specified after this keyword either in hex or as a string;
945by default
946.Xr pfctl 8
947randomly generates a key for source-hash every time the
948ruleset is reloaded.
949.It Ar static-port
950With
951.Ar nat
952rules, the
953.Ar static-port
954option prevents
955.Xr pf 4
956from modifying the source port on TCP and UDP packets.
957.El
958.Pp
959When more than one redirection address is specified,
960.Ar round-robin
961and
962.Ar least-states
963are the only permitted pool types.
964.Ss Routing
965If a packet matches a rule with one of the following route options set,
966the packet filter will route the packet according to the type of route option.
967When such a rule creates state, the route option is also applied to all
968packets matching the same connection.
969.Bl -tag -width xxxx
970.It Ar dup-to
971The
972.Ar dup-to
973option creates a duplicate of the packet and routes it like
974.Ar route-to .
975The original packet gets routed as it normally would.
976.It Ar reply-to
977The
978.Ar reply-to
979option is similar to
980.Ar route-to ,
981but routes packets that pass in the opposite direction (replies) to the
982specified interface.
983Opposite direction is only defined in the context of a state entry, and
984.Ar reply-to
985is useful only in rules that create state.
986It can be used on systems with multiple external connections to
987route all outgoing packets of a connection through the interface
988the incoming connection arrived through (symmetric routing enforcement).
989.It Ar route-to
990The
991.Ar route-to
992option routes the packet to the specified interface with an optional address
993for the next hop.
994When a
995.Ar route-to
996rule creates state, only packets that pass in the same direction as the
997filter rule specifies will be routed in this way.
998Packets passing in the opposite direction (replies) are not affected
999and are routed normally.
1000.El
1001.Pp
1002For the
1003.Ar dup-to ,
1004.Ar reply-to ,
1005and
1006.Ar route-to
1007route options
1008for which there is a single redirection address which has a
1009subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP
1010address),
1011the methods
1012.Ar least-states ,
1013.Ar random ,
1014.Ar round-robin ,
1015and
1016.Ar source-hash ,
1017as described above,
1018can be used.
1019.Sh OPTIONS
1020.Xr pf 4
1021may be tuned for various situations using the
1022.Ar set
1023command.
1024.Bl -tag -width Ds
1025.It Ar set block-policy
1026The
1027.Ar block-policy
1028option sets the default behaviour for the packet
1029.Ar block
1030action:
1031.Pp
1032.Bl -tag -width xxxxxxxx -compact
1033.It Ar drop
1034Packet is silently dropped.
1035.It Ar return
1036A TCP RST is returned for blocked TCP packets,
1037an ICMP UNREACHABLE is returned for blocked UDP packets,
1038and all other packets are silently dropped.
1039.El
1040.It Ar set debug
1041Set the debug
1042.Ar level ,
1043which limits the severity of log messages printed by
1044.Xr pf 4 .
1045This should be a keyword from the following ordered list
1046(highest to lowest):
1047.Cm emerg ,
1048.Cm alert ,
1049.Cm crit ,
1050.Cm err ,
1051.Cm warning ,
1052.Cm notice ,
1053.Cm info ,
1054and
1055.Cm debug .
1056These keywords correspond to the similar (LOG_) values specified to the
1057.Xr syslog 3
1058library routine.
1059.It Ar set fingerprints
1060Load fingerprints of known operating systems from the given filename.
1061By default fingerprints of known operating systems are automatically
1062loaded from
1063.Xr pf.os 5 ,
1064but can be overridden via this option.
1065Setting this option may leave a small period of time where the fingerprints
1066referenced by the currently active ruleset are inconsistent until the new
1067ruleset finishes loading.
1068.It Ar set hostid
1069The 32-bit
1070.Ar hostid
1071identifies this firewall's state table entries to other firewalls
1072in a
1073.Xr pfsync 4
1074failover cluster.
1075By default the hostid is set to a pseudo-random value, however it may be
1076desirable to manually configure it, for example to more easily identify the
1077source of state table entries.
1078The hostid may be specified in either decimal or hexadecimal.
1079.It Ar set limit
1080Sets hard limits on the memory pools used by the packet filter.
1081See
1082.Xr pool 9
1083for an explanation of memory pools.
1084.Pp
1085For example,
1086to set the maximum number of entries in the memory pool used by state table
1087entries (generated by
1088.Ar pass
1089rules which do not specify
1090.Ar no state )
1091to 20000:
1092.Pp
1093.Dl set limit states 20000
1094.Pp
1095To set the maximum number of entries in the memory pool used for fragment
1096reassembly to 20000:
1097.Pp
1098.Dl set limit frags 20000
1099.Pp
1100To set the maximum number of entries in the memory pool used for tracking
1101source IP addresses (generated by the
1102.Ar sticky-address
1103and
1104.Ar src.track
1105options) to 2000:
1106.Pp
1107.Dl set limit src-nodes 2000
1108.Pp
1109To set limits on the memory pools used by tables:
1110.Bd -literal -offset indent
1111set limit tables 1000
1112set limit table-entries 100000
1113.Ed
1114.Pp
1115The first limits the number of tables that can exist to 1000.
1116The second limits the overall number of addresses that can be stored
1117in tables to 100000.
1118.Pp
1119Various limits can be combined on a single line:
1120.Bd -literal -offset indent
1121set limit { states 20000, frags 20000, src-nodes 2000 }
1122.Ed
1123.It Ar set loginterface
1124Enable collection of packet and byte count statistics for the given
1125interface or interface group.
1126These statistics can be viewed using:
1127.Pp
1128.Dl # pfctl -s info
1129.Pp
1130In this example
1131.Xr pf 4
1132collects statistics on the interface named dc0:
1133.Pp
1134.Dl set loginterface dc0
1135.Pp
1136One can disable the loginterface using:
1137.Pp
1138.Dl set loginterface none
1139.It Ar set optimization
1140Optimize state timeouts for one of the following network environments:
1141.Pp
1142.Bl -tag -width Ds -compact
1143.It Ar aggressive
1144Aggressively expire connections.
1145This can greatly reduce the memory usage of the firewall at the cost of
1146dropping idle connections early.
1147.It Ar conservative
1148Extremely conservative settings.
1149Avoid dropping legitimate connections at the
1150expense of greater memory utilization (possibly much greater on a busy
1151network) and slightly increased processor utilization.
1152.It Ar high-latency
1153A high-latency environment (such as a satellite connection).
1154.It Ar normal
1155A normal network environment.
1156Suitable for almost all networks.
1157.It Ar satellite
1158Alias for
1159.Ar high-latency .
1160.El
1161.It Ar set reassemble
1162The
1163.Ar reassemble
1164option is used to enable or disable the reassembly of fragmented packets,
1165and can be set to
1166.Ar yes
1167(the default) or
1168.Ar no .
1169If
1170.Ar no-df
1171is also specified, fragments with the
1172.Ar dont-fragment
1173bit set are reassembled too,
1174instead of being dropped;
1175the reassembled packet will have the
1176.Ar dont-fragment
1177bit cleared.
1178.It Ar set ruleset-optimization
1179.Bl -tag -width xxxxxxxx -compact
1180.It Ar basic
1181Enable basic ruleset optimization.
1182This is the default behaviour.
1183Basic ruleset optimization does four things to improve the
1184performance of ruleset evaluations:
1185.Pp
1186.Bl -enum -compact
1187.It
1188remove duplicate rules
1189.It
1190remove rules that are a subset of another rule
1191.It
1192combine multiple rules into a table when advantageous
1193.It
1194re-order the rules to improve evaluation performance
1195.El
1196.It Ar none
1197Disable the ruleset optimizer.
1198.It Ar profile
1199Uses the currently loaded ruleset as a feedback profile to tailor the
1200ordering of quick rules to actual network traffic.
1201.El
1202.Pp
1203It is important to note that the ruleset optimizer will modify the ruleset
1204to improve performance.
1205A side effect of the ruleset modification is that per-rule accounting
1206statistics will have different meanings than before.
1207If per-rule accounting is important for billing purposes or whatnot,
1208either the ruleset optimizer should not be used or a label field should
1209be added to all of the accounting rules to act as optimization barriers.
1210.Pp
1211Optimization can also be set as a command-line argument to
1212.Xr pfctl 8 ,
1213overriding the settings in
1214.Nm .
1215.It Ar set skip on Aq Ar ifspec
1216List interfaces for which packets should not be filtered.
1217Packets passing in or out on such interfaces are passed as if pf was
1218disabled, i.e. pf does not process them in any way.
1219This can be useful on loopback and other virtual interfaces, when
1220packet filtering is not desired and can have unexpected effects.
1221.Ar ifspec
1222is only evaluated when the ruleset is loaded; interfaces created
1223later will not be skipped.
1224.It Ar set state-defaults
1225The
1226.Ar state-defaults
1227option sets the state options for states created from rules
1228without an explicit
1229.Ar keep state .
1230For example:
1231.Pp
1232.Dl set state-defaults pflow, no-sync
1233.It Ar set state-policy
1234The
1235.Ar state-policy
1236option sets the default behaviour for states:
1237.Pp
1238.Bl -tag -width if-bound -compact
1239.It Ar if-bound
1240States are bound to an interface.
1241.It Ar floating
1242States can match packets on any interfaces (the default).
1243.El
1244.It Ar set timeout
1245.Bl -tag -width "src.track" -compact
1246.It Ar frag
1247Seconds before an unassembled fragment is expired.
1248.It Ar interval
1249Interval between purging expired states and fragments.
1250.It Ar src.track
1251Length of time to retain a source tracking entry after the last state
1252expires.
1253.El
1254.Pp
1255When a packet matches a stateful connection, the seconds to live for the
1256connection will be updated to that of the
1257protocol and modifier
1258which corresponds to the connection state.
1259Each packet which matches this state will reset the TTL.
1260Tuning these values may improve the performance of the
1261firewall at the risk of dropping valid idle connections.
1262.Pp
1263.Bl -tag -width Ds -compact
1264.It Ar tcp.closed
1265The state after one endpoint sends an RST.
1266.It Ar tcp.closing
1267The state after the first FIN has been sent.
1268.It Ar tcp.established
1269The fully established state.
1270.It Ar tcp.finwait
1271The state after both FINs have been exchanged and the connection is closed.
1272Some hosts (notably web servers on Solaris) send TCP packets even after closing
1273the connection.
1274Increasing
1275.Ar tcp.finwait
1276(and possibly
1277.Ar tcp.closing )
1278can prevent blocking of such packets.
1279.It Ar tcp.first
1280The state after the first packet.
1281.It Ar tcp.opening
1282The state before the destination host ever sends a packet.
1283.El
1284.Pp
1285ICMP and UDP are handled in a fashion similar to TCP, but with a much more
1286limited set of states:
1287.Pp
1288.Bl -tag -width Ds -compact
1289.It Ar icmp.error
1290The state after an ICMP error came back in response to an ICMP packet.
1291.It Ar icmp.first
1292The state after the first packet.
1293.It Ar udp.first
1294The state after the first packet.
1295.It Ar udp.multiple
1296The state if both hosts have sent packets.
1297.It Ar udp.single
1298The state if the source host sends more than one packet but the destination
1299host has never sent one back.
1300.El
1301.Pp
1302Other protocols are handled similarly to UDP:
1303.Pp
1304.Bl -tag -width xxxx -compact
1305.It Ar other.first
1306.It Ar other.multiple
1307.It Ar other.single
1308.El
1309.Pp
1310Timeout values can be reduced adaptively as the number of state table
1311entries grows.
1312.Pp
1313.Bl -tag -width Ds -compact
1314.It Ar adaptive.end
1315When reaching this number of state entries, all timeout values become
1316zero, effectively purging all state entries immediately.
1317This value is used to define the scale factor; it should not actually
1318be reached (set a lower state limit, see below).
1319.It Ar adaptive.start
1320When the number of state entries exceeds this value, adaptive scaling
1321begins.
1322All timeout values are scaled linearly with factor
1323(adaptive.end \- number of states) / (adaptive.end \- adaptive.start).
1324.El
1325.Pp
1326Adaptive timeouts are enabled by default, with an adaptive.start value
1327equal to 60% of the state limit, and an adaptive.end value equal to
1328120% of the state limit.
1329They can be disabled by setting both adaptive.start and adaptive.end to 0.
1330.Pp
1331The adaptive timeout values can be defined both globally and for each rule.
1332When used on a per-rule basis, the values relate to the number of
1333states created by the rule, otherwise to the total number of
1334states.
1335.Pp
1336For example:
1337.Bd -literal -offset indent
1338set timeout tcp.first 120
1339set timeout tcp.established 86400
1340set timeout { adaptive.start 6000, adaptive.end 12000 }
1341set limit states 10000
1342.Ed
1343.Pp
1344With 9000 state table entries, the timeout values are scaled to 50%
1345(tcp.first 60, tcp.established 43200).
1346.El
1347.Sh QUEUEING
1348Packets can be assigned to queues for the purpose of bandwidth
1349control.
1350At least two declarations are required to configure queues, and later
1351any packet filtering rule can reference the defined queues by name.
1352During the filtering component of
1353.Nm ,
1354the last referenced
1355.Ar queue
1356name is where any packets from
1357.Ar pass
1358rules will be queued, while for
1359.Ar block
1360rules it specifies where any resulting ICMP or TCP RST
1361packets should be queued.
1362The
1363.Em scheduler
1364defines the algorithm used to decide which packets get delayed, dropped, or
1365sent out immediately.
1366There are three schedulers currently supported:
1367.Bl -tag -width xxxx
1368.It Ar cbq
1369Class Based Queueing.
1370Queues attached to an interface build a tree,
1371thus each queue can have further child queues.
1372Each queue can have a
1373.Ar priority
1374and a
1375.Ar bandwidth
1376assigned.
1377Priority mainly controls the time packets take to get sent out,
1378while bandwidth has primarily effects on throughput.
1379.Ar cbq
1380achieves both partitioning and sharing of link bandwidth
1381by hierarchically structured classes.
1382Each class has its own queue
1383and is assigned its share of bandwidth.
1384A child class can borrow bandwidth from its parent class
1385as long as excess bandwidth is available
1386(see the option
1387.Ar borrow ,
1388below).
1389.It Ar hfsc
1390Hierarchical Fair Service Curve.
1391Queues attached to an interface build a tree,
1392thus each queue can have further child queues.
1393Each queue can have a
1394.Ar priority
1395and a
1396.Ar bandwidth
1397assigned.
1398Priority mainly controls the time packets take to get sent out,
1399while bandwidth primarily affects throughput.
1400.Ar hfsc
1401supports both link-sharing and guaranteed real-time services.
1402It employs a service curve based QoS model,
1403and its unique feature is an ability to decouple
1404delay and bandwidth allocation.
1405.It Ar priq
1406Priority Queueing.
1407Queues are flat attached to the interface,
1408thus queues cannot have further child queues.
1409Each queue has a unique
1410.Ar priority
1411assigned, ranging from 0 to 15.
1412Packets in the queue with the highest priority are processed first.
1413.El
1414.Pp
1415The interfaces on which queueing should be activated are declared using
1416the
1417.Ar altq on
1418declaration.
1419.Ar altq on
1420has the following keywords:
1421.Bl -tag -width xxxx
1422.It Aq Ar interface
1423Queueing is enabled on the named interface.
1424.It Aq Ar scheduler
1425Specifies which queueing scheduler to use.
1426.It Ar bandwidth Aq Ar bw
1427The maximum bitrate for all queues on an
1428interface may be specified using the
1429.Ar bandwidth
1430keyword.
1431The value can be specified as an absolute value or as a
1432percentage of the interface bandwidth.
1433When using an absolute value, the suffixes
1434.Ar b ,
1435.Ar Kb ,
1436.Ar Mb ,
1437and
1438.Ar Gb
1439are used to represent bits, kilobits, megabits, and
1440gigabits per second, respectively.
1441The value must not exceed the interface bandwidth.
1442If
1443.Ar bandwidth
1444is not specified, the interface bandwidth is used
1445(but take note that some interfaces do not know their bandwidth,
1446or can adapt their bandwidth rates).
1447.It Ar qlimit Aq Ar limit
1448The maximum number of packets held in the queue.
1449The default is 50.
1450.It Ar tbrsize Aq Ar size
1451Adjusts the size, in bytes, of the token bucket regulator.
1452If not specified, heuristics based on the
1453interface bandwidth are used to determine the size.
1454.It Ar queue Aq Ar list
1455Defines a list of subqueues to create on an interface.
1456.El
1457.Pp
1458In the following example, the interface dc0
1459should queue up to 5Mbps in four second-level queues using
1460Class Based Queueing.
1461Those four queues will be shown in a later example.
1462.Bd -literal -offset indent
1463altq on dc0 cbq bandwidth 5Mb queue { std, http, mail, ssh }
1464.Ed
1465.Pp
1466Once interfaces are activated for queueing using the
1467.Ar altq
1468directive, a sequence of
1469.Ar queue
1470directives may be defined.
1471The name associated with a queue
1472must match a queue defined in the altq
1473directive or, except for the
1474.Ar priq
1475scheduler, in a parent queue declaration.
1476The following keywords can be used:
1477.Bl -tag -width xxxx
1478.It Ar on Aq Ar interface
1479Specifies the interface the queue operates on.
1480If not given, it operates on all matching interfaces.
1481.It Ar bandwidth Aq Ar bw
1482Specifies the maximum bitrate to be processed by the queue.
1483This value must not exceed the value of the parent
1484.Ar queue
1485and can be specified as an absolute value or a percentage of the parent
1486queue's bandwidth.
1487If not specified, defaults to 100% of the parent queue's bandwidth.
1488The
1489.Ar priq
1490scheduler does not support bandwidth specification.
1491.It Ar priority Aq Ar level
1492Between queues a priority level can be set.
1493For
1494.Ar cbq
1495and
1496.Ar hfsc ,
1497the range is 0 to 7 and for
1498.Ar priq ,
1499the range is 0 to 15.
1500The default for all is 1.
1501.Ar priq
1502queues with a higher priority are always served first.
1503.Ar cbq
1504and
1505.Ar hfsc
1506queues with a higher priority are preferred in the case of overload.
1507.It Ar qlimit Aq Ar limit
1508The maximum number of packets held in the queue.
1509The default is 50.
1510.El
1511.Pp
1512The scheduler can specify additional parameters using the format
1513.Ar scheduler Ns Pq Ar parameters .
1514The parameters are:
1515.Bl -tag -width Fl
1516.It Ar default
1517Packets not matched by another queue are assigned to this one.
1518Exactly one default queue is required.
1519.It Ar ecn
1520Enables Explicit Congestion Notification (ECN) on this queue.
1521ECN implies RED.
1522.It Ar red
1523Enables Random Early Detection (RED) on this queue.
1524RED drops packets with a probability proportional to the average
1525queue length.
1526.El
1527.Pp
1528The
1529.Ar cbq
1530scheduler supports an additional option:
1531.Pp
1532.Bl -tag -width "upperlimit XXXXXXX" -compact -offset 3n
1533.It Ar borrow
1534The queue can borrow bandwidth from the parent.
1535.El
1536.Pp
1537The
1538.Ar hfsc
1539scheduler supports some additional options:
1540.Pp
1541.Bl -tag -width "upperlimit XXXXXXX" -compact -offset 3n
1542.It Ar linkshare Aq Ar sc
1543The bandwidth share of a backlogged queue.
1544.It Ar realtime Aq Ar sc
1545The minimum required bandwidth for the queue.
1546.It Ar upperlimit Aq Ar sc
1547The maximum allowed bandwidth for the queue.
1548.El
1549.Pp
1550.Aq Ar sc
1551is an abbreviation for
1552.Ar service curve .
1553.Pp
1554The format for service curve specifications is
1555.Ar ( m1 , d , m2 ) .
1556.Ar m2
1557controls the bandwidth assigned to the queue.
1558.Ar m1
1559and
1560.Ar d
1561are optional and can be used to control the initial bandwidth assignment.
1562For the first
1563.Ar d
1564milliseconds the queue gets the bandwidth given as
1565.Ar m1 ,
1566afterwards the value given in
1567.Ar m2 .
1568.Pp
1569Furthermore, with
1570.Ar cbq
1571and
1572.Ar hfsc ,
1573child queues can be specified as in an
1574.Ar altq
1575declaration, thus building a tree of queues using a part of
1576their parent's bandwidth.
1577.Pp
1578Packets can be assigned to queues based on filter rules by using the
1579.Ar queue
1580keyword.
1581Normally only one
1582.Ar queue
1583is specified; when a second one is specified it will instead be used for
1584packets which have a TOS of
1585.Ar lowdelay
1586and for TCP ACKs with no data payload.
1587.Pp
1588To continue the previous example, the examples below would specify the
1589four referenced
1590queues, plus a few child queues.
1591Interactive
1592.Xr ssh 1
1593sessions get priority over bulk transfers like
1594.Xr scp 1
1595and
1596.Xr sftp 1 .
1597The queues may then be referenced by filtering rules (see
1598.Sx PACKET FILTERING ,
1599above).
1600.Bd -literal -offset 4n
1601queue std bandwidth 10% cbq(default)
1602queue http bandwidth 60% priority 2 cbq(borrow red) \e
1603      { employees, developers }
1604queue  developers bandwidth 75% cbq(borrow)
1605queue  employees bandwidth 15%
1606queue mail bandwidth 10% priority 0 cbq(borrow ecn)
1607queue ssh bandwidth 20% cbq(borrow) { ssh_interactive, ssh_bulk }
1608queue  ssh_interactive bandwidth 50% priority 7 cbq(borrow)
1609queue  ssh_bulk bandwidth 50% priority 0 cbq(borrow)
1610
1611block return out on dc0 inet all queue std
1612pass out on dc0 inet proto tcp from $developerhosts to any port 80 \e
1613      queue developers
1614pass out on dc0 inet proto tcp from $employeehosts to any port 80 \e
1615      queue employees
1616pass out on dc0 inet proto tcp from any to any port 22 \e
1617      queue(ssh_bulk, ssh_interactive)
1618pass out on dc0 inet proto tcp from any to any port 25 \e
1619      queue mail
1620.Ed
1621.Sh TABLES
1622Tables are named structures which can hold a collection of addresses and
1623networks.
1624Lookups against tables in
1625.Xr pf 4
1626are relatively fast, making a single rule with tables much more efficient,
1627in terms of
1628processor usage and memory consumption, than a large number of rules which
1629differ only in IP address (either created explicitly or automatically by rule
1630expansion).
1631.Pp
1632Tables can be used as the source or destination of filter
1633or translation rules.
1634They can also be used for the redirect address of
1635.Ar nat-to
1636and
1637.Ar rdr-to
1638and in the routing options of filter rules, but only for
1639.Ar least-states
1640and
1641.Ar round-robin
1642pools.
1643.Pp
1644Tables can be defined with any of the following
1645.Xr pfctl 8
1646mechanisms.
1647As with macros, reserved words may not be used as table names.
1648.Bl -tag -width "manually"
1649.It Ar manually
1650Persistent tables can be manually created with the
1651.Ar add
1652or
1653.Ar replace
1654option of
1655.Xr pfctl 8 ,
1656before or after the ruleset has been loaded.
1657.It Pa pf.conf
1658Table definitions can be placed directly in this file and loaded at the
1659same time as other rules are loaded, atomically.
1660Table definitions inside
1661.Nm
1662use the
1663.Ar table
1664statement, and are especially useful to define non-persistent tables.
1665The contents of a pre-existing table defined without a list of addresses
1666to initialize it is not altered when
1667.Nm
1668is loaded.
1669A table initialized with the empty list,
1670.Li { } ,
1671will be cleared on load.
1672.El
1673.Pp
1674Tables may be defined with the following attributes:
1675.Bl -tag -width persist
1676.It Ar const
1677The
1678.Ar const
1679flag prevents the user from altering the contents of the table once it
1680has been created.
1681Without that flag,
1682.Xr pfctl 8
1683can be used to add or remove addresses from the table at any time, even
1684when running with
1685.Xr securelevel 7
1686= 2.
1687.It Ar counters
1688The
1689.Ar counters
1690flag enables per-address packet and byte counters, which can be displayed with
1691.Xr pfctl 8 .
1692.It Ar persist
1693The
1694.Ar persist
1695flag forces the kernel to keep the table even when no rules refer to it.
1696If the flag is not set, the kernel will automatically remove the table
1697when the last rule referring to it is flushed.
1698.El
1699.Pp
1700This example
1701creates a table called private,
1702to hold RFC 1918 private network blocks,
1703and a table called badhosts,
1704which is initially empty.
1705A filter rule is set up to block all traffic coming from addresses listed in
1706either table:
1707.Bd -literal -offset indent
1708table \*(Ltprivate\*(Gt const { 10/8, 172.16/12, 192.168/16 }
1709table \*(Ltbadhosts\*(Gt persist
1710block on fxp0 from { \*(Ltprivate\*(Gt, \*(Ltbadhosts\*(Gt } to any
1711.Ed
1712.Pp
1713The private table cannot have its contents changed and the badhosts table
1714will exist even when no active filter rules reference it.
1715Addresses may later be added to the badhosts table, so that traffic from
1716these hosts can be blocked by using the following:
1717.Pp
1718.Dl # pfctl -t badhosts -Tadd 204.92.77.111
1719.Pp
1720A table can also be initialized with an address list specified in one or more
1721external files, using the following syntax:
1722.Bd -literal -offset indent
1723table \*(Ltspam\*(Gt persist file "/etc/spammers" file "/etc/openrelays"
1724block on fxp0 from \*(Ltspam\*(Gt to any
1725.Ed
1726.Pp
1727The files
1728.Pa /etc/spammers
1729and
1730.Pa /etc/openrelays
1731list IP addresses, one per line.
1732Any lines beginning with a
1733.Sq #
1734are treated as comments and ignored.
1735In addition to being specified by IP address, hosts may also be
1736specified by their hostname.
1737When the resolver is called to add a hostname to a table,
1738.Ar all
1739resulting IPv4 and IPv6 addresses are placed into the table.
1740IP addresses can also be entered in a table by specifying a valid interface
1741name, a valid interface group, or the
1742.Ar self
1743keyword, in which case all addresses assigned to the interface(s) will be
1744added to the table.
1745.Sh ANCHORS
1746Besides the main ruleset,
1747.Nm
1748can specify
1749.Ar anchor
1750attachment points.
1751An anchor is a container that can hold rules,
1752address tables, and other anchors.
1753When evaluation of the main ruleset reaches an
1754.Ar anchor
1755rule,
1756.Xr pf 4
1757will proceed to evaluate all rules specified in that anchor.
1758.Pp
1759The following example blocks all packets on the external interface by default,
1760then evaluates all rules in the anchor named "spam",
1761and finally passes all outgoing connections and
1762incoming connections to port 25:
1763.Bd -literal -offset indent
1764ext_if = "kue0"
1765block on $ext_if all
1766anchor spam
1767pass out on $ext_if all
1768pass in on $ext_if proto tcp from any to $ext_if port smtp
1769.Ed
1770.Pp
1771Anchors can be manipulated through
1772.Xr pfctl 8
1773without reloading the main ruleset or other anchors.
1774This loads a single rule into the anchor,
1775which blocks all packets from a specific address:
1776.Bd -literal -offset indent
1777# echo "block in quick from 1.2.3.4 to any" | pfctl -a spam -f -
1778.Ed
1779.Pp
1780The anchor can also be populated by adding a
1781.Ar load anchor
1782rule after the anchor rule.
1783When
1784.Xr pfctl 8
1785loads
1786.Nm ,
1787it will also load all the rules from the file
1788.Pa /etc/pf-spam.conf
1789into the anchor.
1790.Bd -literal -offset indent
1791anchor spam
1792load anchor spam from "/etc/pf-spam.conf"
1793.Ed
1794.Pp
1795Filter rule anchors can also be loaded inline in the ruleset
1796within a brace-delimited block.
1797Brace delimited blocks may contain rules or other brace-delimited blocks.
1798When anchors are loaded this way the anchor name becomes optional.
1799Since the parser specification for anchor names is a string,
1800double quote characters
1801.Pq Sq \&"
1802should be placed around the anchor name.
1803.Bd -literal -offset indent
1804anchor "external" on egress {
1805	block
1806	anchor out {
1807		pass proto tcp from any to port { 25, 80, 443 }
1808	}
1809	pass in proto tcp to any port 22
1810}
1811.Ed
1812.Pp
1813Anchor rules can also specify packet filtering parameters
1814using the same syntax as filter rules.
1815When parameters are used,
1816the anchor rule is only evaluated for matching packets.
1817This allows conditional evaluation of anchors, like:
1818.Bd -literal -offset indent
1819block on $ext_if all
1820anchor spam proto tcp from any to any port smtp
1821pass out on $ext_if all
1822pass in on $ext_if proto tcp from any to $ext_if port smtp
1823.Ed
1824.Pp
1825The rules inside anchor "spam" are only evaluated
1826for TCP packets with destination port 25.
1827Hence, the following
1828will only block connections from 1.2.3.4 to port 25:
1829.Bd -literal -offset indent
1830# echo "block in quick from 1.2.3.4 to any" | pfctl -a spam -f -
1831.Ed
1832.Pp
1833Matching filter and translation rules marked with the
1834.Ar quick
1835option are final and abort the evaluation of the rules in other
1836anchors and the main ruleset.
1837If the anchor itself is marked with the
1838.Ar quick
1839option,
1840ruleset evaluation will terminate when the anchor is exited if the packet is
1841matched by any rule within the anchor.
1842.Pp
1843An anchor references other anchor attachment points
1844using the following syntax:
1845.Bl -tag -width xxxx
1846.It Ar anchor Aq Ar name
1847Evaluates the filter rules in the specified anchor.
1848.El
1849.Pp
1850An anchor has a name which specifies the path where
1851.Xr pfctl 8
1852can be used to access the anchor to perform operations on it, such as
1853attaching child anchors to it or loading rules into it.
1854Anchors may be nested, with components separated by
1855.Sq /
1856characters, similar to how file system hierarchies are laid out.
1857The main ruleset is actually the default anchor, so filter and
1858translation rules, for example, may also be contained in any anchor.
1859.Pp
1860Anchor rules are evaluated relative to the anchor in which they are contained.
1861For example,
1862all anchor rules specified in the main ruleset will reference
1863anchor attachment points underneath the main ruleset,
1864and anchor rules specified in a file loaded from a
1865.Ar load anchor
1866rule will be attached under that anchor point.
1867.Pp
1868Anchors may end with the asterisk
1869.Pq Sq *
1870character, which signifies that all anchors attached at that point
1871should be evaluated in the alphabetical ordering of their anchor name.
1872For example,
1873the following
1874will evaluate each rule in each anchor attached to the "spam" anchor:
1875.Bd -literal -offset indent
1876anchor "spam/*"
1877.Ed
1878.Pp
1879Note that it will only evaluate anchors that are directly attached to the
1880"spam" anchor, and will not descend to evaluate anchors recursively.
1881.Pp
1882Since anchors are evaluated relative to the anchor in which they are
1883contained, there is a mechanism for accessing the parent and ancestor
1884anchors of a given anchor.
1885Similar to file system path name resolution, if the sequence
1886.Sq ..
1887appears as an anchor path component, the parent anchor of the current
1888anchor in the path evaluation at that point will become the new current
1889anchor.
1890As an example, consider the following:
1891.Bd -literal -offset indent
1892# printf 'anchor "spam/allowed"\en' | pfctl -f -
1893# printf 'anchor "../banned"\enpass\en' | pfctl -a spam/allowed -f -
1894.Ed
1895.Pp
1896Evaluation of the main ruleset will lead into the
1897spam/allowed anchor, which will evaluate the rules in the
1898spam/banned anchor, if any, before finally evaluating the
1899.Ar pass
1900rule.
1901.Sh STATEFUL FILTERING
1902.Xr pf 4
1903filters packets statefully,
1904which has several advantages.
1905For TCP connections, comparing a packet to a state involves checking
1906its sequence numbers, as well as TCP timestamps if a rule using the
1907.Ar reassemble tcp
1908parameter applies to the connection.
1909If these values are outside the narrow windows of expected
1910values, the packet is dropped.
1911This prevents spoofing attacks, such as when an attacker sends packets with
1912a fake source address/port but does not know the connection's sequence
1913numbers.
1914Similarly,
1915.Xr pf 4
1916knows how to match ICMP replies to states.
1917For example,
1918to allow echo requests (such as those created by
1919.Xr ping 8 )
1920out statefully and match incoming echo replies correctly to states:
1921.Pp
1922.Dl pass out inet proto icmp all icmp-type echoreq
1923.Pp
1924Also, looking up states is usually faster than evaluating rules.
1925If there are 50 rules, all of them are evaluated sequentially in O(n).
1926Even with 50000 states, only 16 comparisons are needed to match a
1927state, since states are stored in a binary search tree that allows
1928searches in O(log2 n).
1929.Pp
1930Furthermore, correct handling of ICMP error messages is critical to
1931many protocols, particularly TCP.
1932.Xr pf 4
1933matches ICMP error messages to the correct connection, checks them against
1934connection parameters, and passes them if appropriate.
1935For example if an ICMP source quench message referring to a stateful TCP
1936connection arrives, it will be matched to the state and get passed.
1937.Pp
1938Finally, state tracking is required for
1939.Ar nat-to
1940and
1941.Ar rdr-to
1942options, in order to track address and port translations and reverse the
1943translation on returning packets.
1944.Pp
1945.Xr pf 4
1946will also create state for other protocols which are effectively stateless by
1947nature.
1948UDP packets are matched to states using only host addresses and ports,
1949and other protocols are matched to states using only the host addresses.
1950.Pp
1951If stateless filtering of individual packets is desired,
1952the
1953.Ar no state
1954keyword can be used to specify that state will not be created
1955if this is the last matching rule.
1956Note that packets which match neither block nor pass rules,
1957and thus are passed by default,
1958are effectively passed as if
1959.Ar no state
1960had been specified.
1961.Pp
1962A number of parameters can also be set to affect how
1963.Xr pf 4
1964handles state tracking,
1965as detailed below.
1966.Ss State Modulation
1967Much of the security derived from TCP is attributable to how well the
1968initial sequence numbers (ISNs) are chosen.
1969Some popular stack implementations choose
1970.Em very
1971poor ISNs and thus are normally susceptible to ISN prediction exploits.
1972By applying a
1973.Ar modulate state
1974rule to a TCP connection,
1975.Xr pf 4
1976will create a high quality random sequence number for each connection
1977endpoint.
1978.Pp
1979The
1980.Ar modulate state
1981directive implicitly keeps state on the rule and is
1982only applicable to TCP connections.
1983.Pp
1984For instance:
1985.Bd -literal -offset indent
1986block all
1987pass out proto tcp from any to any modulate state
1988pass in  proto tcp from any to any port 25 flags S/SFRA \e
1989      modulate state
1990.Ed
1991.Pp
1992Note that modulated connections will not recover when the state table
1993is lost (firewall reboot, flushing the state table, etc.).
1994.Xr pf 4
1995will not be able to infer a connection again after the state table flushes
1996the connection's modulator.
1997When the state is lost, the connection may be left dangling until the
1998respective endpoints time out the connection.
1999It is possible on a fast local network for the endpoints to start an ACK
2000storm while trying to resynchronize after the loss of the modulator.
2001The default
2002.Ar flags
2003settings (or a more strict equivalent) should be used on
2004.Ar modulate state
2005rules to prevent ACK storms.
2006.Pp
2007Note that alternative methods are available
2008to prevent loss of the state table
2009and allow for firewall failover.
2010See
2011.Xr carp 4
2012and
2013.Xr pfsync 4
2014for further information.
2015.Ss SYN Proxy
2016By default,
2017.Xr pf 4
2018passes packets that are part of a
2019TCP handshake between the endpoints.
2020The
2021.Ar synproxy state
2022option can be used to cause
2023.Xr pf 4
2024itself to complete the handshake with the active endpoint, perform a handshake
2025with the passive endpoint, and then forward packets between the endpoints.
2026.Pp
2027No packets are sent to the passive endpoint before the active endpoint has
2028completed the handshake, hence so-called SYN floods with spoofed source
2029addresses will not reach the passive endpoint, as the sender can't complete the
2030handshake.
2031.Pp
2032The proxy is transparent to both endpoints; they each see a single
2033connection from/to the other endpoint.
2034.Xr pf 4
2035chooses random initial sequence numbers for both handshakes.
2036Once the handshakes are completed, the sequence number modulators
2037(see previous section) are used to translate further packets of the
2038connection.
2039.Ar synproxy state
2040includes
2041.Ar modulate state .
2042.Pp
2043Rules with
2044.Ar synproxy
2045will not work if
2046.Xr pf 4
2047operates on a
2048.Xr bridge 4 .
2049.Pp
2050Example:
2051.Bd -literal -offset indent
2052pass in proto tcp from any to any port www synproxy state
2053.Ed
2054.Ss Stateful Tracking Options
2055A number of options related to stateful tracking can be applied on a
2056per-rule basis.
2057One of
2058.Ar keep state ,
2059.Ar modulate state ,
2060or
2061.Ar synproxy state
2062must be specified explicitly to apply these options to a rule.
2063.Pp
2064.Bl -tag -width xxxx -compact
2065.It Ar floating
2066States can match packets on any interfaces
2067(the opposite of
2068.Ar if-bound ) .
2069This is the default.
2070.It Ar if-bound
2071States are bound to an interface
2072(the opposite of
2073.Ar floating ) .
2074.It Ar max Aq Ar number
2075Limits the number of concurrent states the rule may create.
2076When this limit is reached, further packets that would create
2077state are dropped until existing states time out.
2078.It Ar no-sync
2079Prevent state changes for states created by this rule from appearing on the
2080.Xr pfsync 4
2081interface.
2082.It Ar pflow
2083States created by this rule are exported on the
2084.Xr pflow 4
2085interface.
2086.It Ar sloppy
2087Uses a sloppy TCP connection tracker that does not check sequence
2088numbers at all, which makes insertion and ICMP teardown attacks way
2089easier.
2090This is intended to be used in situations where one does not see all
2091packets of a connection, e.g. in asymmetric routing situations.
2092It cannot be used with
2093.Ar modulate
2094or
2095.Ar synproxy state .
2096.It Xo Aq Ar timeout
2097.Aq Ar seconds
2098.Xc
2099Changes the timeout values used for states created by this rule.
2100For a list of all valid timeout names, see
2101.Sx OPTIONS
2102above.
2103.El
2104.Pp
2105Multiple options can be specified, separated by commas:
2106.Bd -literal -offset indent
2107pass in proto tcp from any to any \e
2108      port www keep state \e
2109      (max 100, source-track rule, max-src-nodes 75, \e
2110      max-src-states 3, tcp.established 60, tcp.closing 5)
2111.Ed
2112.Pp
2113When the
2114.Ar source-track
2115keyword is specified, the number of states per source IP is tracked.
2116.Pp
2117.Bl -tag -width xxxx -compact
2118.It Ar source-track global
2119The number of states created by all rules that use this option is limited.
2120Each rule can specify different
2121.Ar max-src-nodes
2122and
2123.Ar max-src-states
2124options, however state entries created by any participating rule count towards
2125each individual rule's limits.
2126.It Ar source-track rule
2127The maximum number of states created by this rule is limited by the rule's
2128.Ar max-src-nodes
2129and
2130.Ar max-src-states
2131options.
2132Only state entries created by this particular rule count toward the rule's
2133limits.
2134.El
2135.Pp
2136The following limits can be set:
2137.Pp
2138.Bl -tag -width xxxx -compact
2139.It Ar max-src-nodes Aq Ar number
2140Limits the maximum number of source addresses which can simultaneously
2141have state table entries.
2142.It Ar max-src-states Aq Ar number
2143Limits the maximum number of simultaneous state entries that a single
2144source address can create with this rule.
2145.El
2146.Pp
2147For stateful TCP connections, limits on established connections (connections
2148which have completed the TCP 3-way handshake) can also be enforced
2149per source IP.
2150.Pp
2151.Bl -tag -width xxxx -compact
2152.It Ar max-src-conn Aq Ar number
2153Limits the maximum number of simultaneous TCP connections which have
2154completed the 3-way handshake that a single host can make.
2155.It Xo Ar max-src-conn-rate Aq Ar number
2156.No / Aq Ar seconds
2157.Xc
2158Limit the rate of new connections over a time interval.
2159The connection rate is an approximation calculated as a moving average.
2160.El
2161.Pp
2162When one of these limits is reached, further packets that would create
2163state are dropped until existing states time out.
2164.Pp
2165Because the 3-way handshake ensures that the source address is not being
2166spoofed, more aggressive action can be taken based on these limits.
2167With the
2168.Ar overload Aq Ar table
2169state option, source IP addresses which hit either of the limits on
2170established connections will be added to the named table.
2171This table can be used in the ruleset to block further activity from
2172the offending host, redirect it to a tarpit process, or restrict its
2173bandwidth.
2174.Pp
2175The optional
2176.Ar flush
2177keyword kills all states created by the matching rule which originate
2178from the host which exceeds these limits.
2179The
2180.Ar global
2181modifier to the flush command kills all states originating from the
2182offending host, regardless of which rule created the state.
2183.Pp
2184For example, the following rules will protect the webserver against
2185hosts making more than 100 connections in 10 seconds.
2186Any host which connects faster than this rate will have its address added
2187to the
2188.Aq bad_hosts
2189table and have all states originating from it flushed.
2190Any new packets arriving from this host will be dropped unconditionally
2191by the block rule.
2192.Bd -literal -offset indent
2193block quick from \*(Ltbad_hosts\*(Gt
2194pass in on $ext_if proto tcp to $webserver port www keep state \e
2195      (max-src-conn-rate 100/10, overload \*(Ltbad_hosts\*(Gt flush global)
2196.Ed
2197.Sh TRAFFIC NORMALISATION
2198Traffic normalisation is a broad umbrella term
2199for aspects of the packet filter which deal with
2200verifying packets, packet fragments, spoof traffic,
2201and other irregularities.
2202.Ss Scrub
2203Scrub involves sanitising packet content in such a way
2204that there are no ambiguities in packet interpretation on the receiving side.
2205It is invoked with the
2206.Ar scrub
2207option, added to regular rules.
2208.Pp
2209Parameters are specified enclosed in parentheses.
2210At least one of the following parameters must be specified:
2211.Bl -tag -width xxxx
2212.It Ar max-mss Aq Ar number
2213Enforces a maximum segment size (MSS) for matching TCP packets.
2214.It Ar min-ttl Aq Ar number
2215Enforces a minimum TTL for matching IP packets.
2216.It Ar no-df
2217Clears the
2218.Ar dont-fragment
2219bit from a matching IPv4 packet.
2220Some operating systems have NFS implementations
2221which are known to generate fragmented packets with the
2222.Ar dont-fragment
2223bit set.
2224.Xr pf 4
2225will drop such fragmented
2226.Ar dont-fragment
2227packets unless
2228.Ar no-df
2229is specified.
2230.Pp
2231Unfortunately some operating systems also generate their
2232.Ar dont-fragment
2233packets with a zero IP identification field.
2234Clearing the
2235.Ar dont-fragment
2236bit on packets with a zero IP ID may cause deleterious results if an
2237upstream router later fragments the packet.
2238Using
2239.Ar random-id
2240is recommended in combination with
2241.Ar no-df
2242to ensure unique IP identifiers.
2243.It Ar random-id
2244Replaces the IPv4 identification field with random values to compensate
2245for predictable values generated by many hosts.
2246This option only applies to packets that are not fragmented
2247after the optional fragment reassembly.
2248.It Ar reassemble tcp
2249Statefully normalises TCP connections.
2250.Ar reassemble tcp
2251performs the following normalisations:
2252.Bl -ohang
2253.It TTL
2254Neither side of the connection is allowed to reduce their IP TTL.
2255An attacker may send a packet such that it reaches the firewall, affects
2256the firewall state, and expires before reaching the destination host.
2257.Ar reassemble tcp
2258will raise the TTL of all packets back up to the highest value seen on
2259the connection.
2260.It Timestamp Modulation
2261Modern TCP stacks will send a timestamp on every TCP packet and echo
2262the other endpoint's timestamp back to them.
2263Many operating systems will merely start the timestamp at zero when
2264first booted, and increment it several times a second.
2265The uptime of the host can be deduced by reading the timestamp and multiplying
2266by a constant.
2267Also observing several different timestamps can be used to count hosts
2268behind a NAT device.
2269And spoofing TCP packets into a connection requires knowing or guessing
2270valid timestamps.
2271Timestamps merely need to be monotonically increasing and not derived off a
2272guessable base time.
2273.Ar reassemble tcp
2274will cause
2275.Ar scrub
2276to modulate the TCP timestamps with a random number.
2277.It Extended PAWS Checks
2278There is a problem with TCP on long fat pipes, in that a packet might get
2279delayed for longer than it takes the connection to wrap its 32-bit sequence
2280space.
2281In such an occurrence, the old packet would be indistinguishable from a
2282new packet and would be accepted as such.
2283The solution to this is called PAWS: Protection Against Wrapped Sequence
2284numbers.
2285It protects against it by making sure the timestamp on each packet does
2286not go backwards.
2287.Ar reassemble tcp
2288also makes sure the timestamp on the packet does not go forward more
2289than the RFC allows.
2290By doing this,
2291.Xr pf 4
2292artificially extends the security of TCP sequence numbers by 10 to 18
2293bits when the host uses appropriately randomized timestamps, since a
2294blind attacker would have to guess the timestamp as well.
2295.El
2296.It Xo Ar set-tos Aq Ar string
2297.No \*(Ba Aq Ar number
2298.Xc
2299Enforces a TOS for matching IPv4 packets.
2300.Ar string
2301may be one of
2302.Ar critical ,
2303.Ar inetcontrol ,
2304.Ar lowdelay ,
2305.Ar netcontrol ,
2306.Ar throughput ,
2307.Ar reliability ,
2308or one of the DiffServ Code Points:
2309.Ar ef ,
2310.Ar af11 ... af43 ,
2311.Ar cs0 ... cs7 ;
2312.Ar number
2313may be either a hex or decimal number.
2314.El
2315.Pp
2316For example:
2317.Pp
2318.Dl match in all scrub (no-df max-mss 1440)
2319.Ss Fragment Handling
2320The size of IP datagrams (packets) can be significantly larger than the
2321maximum transmission unit (MTU) of the network.
2322In cases when it is necessary or more efficient to send such large packets,
2323the large packet will be fragmented into many smaller packets that will each
2324fit onto the wire.
2325Unfortunately for a firewalling device, only the first logical fragment will
2326contain the necessary header information for the subprotocol that allows
2327.Xr pf 4
2328to filter on things such as TCP ports or to perform NAT.
2329.Pp
2330One alternative is to filter individual fragments with filter rules.
2331If packet reassembly is turned off, it is passed to the filter.
2332Filter rules with matching IP header parameters decide whether the
2333fragment is passed or blocked, in the same way as complete packets
2334are filtered.
2335Without reassembly, fragments can only be filtered based on IP header
2336fields (source/destination address, protocol), since subprotocol header
2337fields are not available (TCP/UDP port numbers, ICMP code/type).
2338The
2339.Ar fragment
2340option can be used to restrict filter rules to apply only to
2341fragments, but not complete packets.
2342Filter rules without the
2343.Ar fragment
2344option still apply to fragments, if they only specify IP header fields.
2345For instance:
2346.Bd -literal -offset indent
2347pass in proto tcp from any to any port 80
2348.Ed
2349.Pp
2350The rule above never applies to a fragment,
2351even if the fragment is part of a TCP packet with destination port 80,
2352because without reassembly this information
2353is not available for each fragment.
2354This also means that fragments cannot create new or match existing
2355state table entries, which makes stateful filtering and address
2356translation (NAT, redirection) for fragments impossible.
2357.Pp
2358In most cases, the benefits of reassembly outweigh the additional
2359memory cost,
2360so reassembly is on by default.
2361.Pp
2362The memory allocated for fragment caching can be limited using
2363.Xr pfctl 8 .
2364Once this limit is reached, fragments that would have to be cached
2365are dropped until other entries time out.
2366The timeout value can also be adjusted.
2367.Pp
2368When forwarding reassembled IPv6 packets, pf refragments them with
2369the original maximum fragment size.
2370This allows the sender to determine the optimal fragment size by
2371path MTU discovery.
2372.Ss Blocking Spoofed Traffic
2373Spoofing is the faking of IP addresses,
2374typically for malicious purposes.
2375The
2376.Ar antispoof
2377directive expands to a set of filter rules which will block all
2378traffic with a source IP from the network(s) directly connected
2379to the specified interface(s) from entering the system through
2380any other interface.
2381.Pp
2382For example:
2383.Dl antispoof for lo0
2384.Pp
2385Expands to:
2386.Bd -literal -offset indent -compact
2387block drop in on ! lo0 inet from 127.0.0.1/8 to any
2388block drop in on ! lo0 inet6 from ::1 to any
2389.Ed
2390.Pp
2391For non-loopback interfaces, there are additional rules to block incoming
2392packets with a source IP address identical to the interface's IP(s).
2393For example, assuming the interface wi0 had an IP address of 10.0.0.1 and a
2394netmask of 255.255.255.0:
2395.Pp
2396.Dl antispoof for wi0 inet
2397.Pp
2398Expands to:
2399.Bd -literal -offset indent -compact
2400block drop in on ! wi0 inet from 10.0.0.0/24 to any
2401block drop in inet from 10.0.0.1 to any
2402.Ed
2403.Pp
2404Caveat: Rules created by the
2405.Ar antispoof
2406directive interfere with packets sent over loopback interfaces
2407to local addresses.
2408One should pass these explicitly.
2409.Sh OPERATING SYSTEM FINGERPRINTING
2410Passive OS fingerprinting is a mechanism to inspect nuances of a TCP
2411connection's initial SYN packet and guess at the host's operating system.
2412Unfortunately these nuances are easily spoofed by an attacker so the
2413fingerprint is not useful in making security decisions.
2414But the fingerprint is typically accurate enough to make policy decisions
2415upon.
2416.Pp
2417The fingerprints may be specified by operating system class, by
2418version, or by subtype/patchlevel.
2419The class of an operating system is typically the vendor or genre
2420and would be
2421.Ox
2422for the
2423.Xr pf 4
2424firewall itself.
2425The version of the oldest available
2426.Ox
2427release on the main FTP site
2428would be 2.6 and the fingerprint would be written as:
2429.Pp
2430.Dl \&"OpenBSD 2.6\&"
2431.Pp
2432The subtype of an operating system is typically used to describe the
2433patchlevel if that patch led to changes in the TCP stack behavior.
2434In the case of
2435.Ox ,
2436the only subtype is for a fingerprint that was
2437normalised by the
2438.Ar no-df
2439scrub option and would be specified as:
2440.Pp
2441.Dl \&"OpenBSD 3.3 no-df\&"
2442.Pp
2443Fingerprints for most popular operating systems are provided by
2444.Xr pf.os 5 .
2445Once
2446.Xr pf 4
2447is running, a complete list of known operating system fingerprints may
2448be listed by running:
2449.Pp
2450.Dl # pfctl -so
2451.Pp
2452Filter rules can enforce policy at any level of operating system specification
2453assuming a fingerprint is present.
2454Policy could limit traffic to approved operating systems or even ban traffic
2455from hosts that aren't at the latest service pack.
2456.Pp
2457The
2458.Ar unknown
2459class can also be used as the fingerprint which will match packets for
2460which no operating system fingerprint is known.
2461.Pp
2462Examples:
2463.Bd -literal -offset indent
2464pass  out proto tcp from any os OpenBSD
2465block out proto tcp from any os Doors
2466block out proto tcp from any os "Doors PT"
2467block out proto tcp from any os "Doors PT SP3"
2468block out from any os "unknown"
2469pass on lo0 proto tcp from any os "OpenBSD 3.3 lo0"
2470.Ed
2471.Pp
2472Operating system fingerprinting is limited only to the TCP SYN packet.
2473This means that it will not work on other protocols and will not match
2474a currently established connection.
2475.Pp
2476Caveat: operating system fingerprints are occasionally wrong.
2477There are three problems: an attacker can trivially craft his packets to
2478appear as any operating system he chooses;
2479an operating system patch could change the stack behavior and no fingerprints
2480will match it until the database is updated;
2481and multiple operating systems may have the same fingerprint.
2482.Sh EXAMPLES
2483In this example,
2484the external interface is kue0.
2485We use a macro for the interface name, so it can be changed easily.
2486All incoming traffic is "normalised",
2487and everything is blocked and logged by default.
2488.Bd -literal -offset 4n
2489ext_if = "kue0"
2490match in all scrub (no-df max-mss 1440)
2491block return log on $ext_if all
2492.Ed
2493.Pp
2494Here we specifically block packets we don't want:
2495anything coming from source we have no back routes for;
2496packets whose ingress interface does not match the one in
2497the route back to their source address;
2498anything that does not have our address (157.161.48.183) as source;
2499broadcasts (cable modem noise);
2500and anything from reserved address space or invalid addresses.
2501.Bd -literal -offset 4n
2502block in from no-route to any
2503block in from urpf-failed to any
2504block out log quick on $ext_if from ! 157.161.48.183 to any
2505block in quick on $ext_if from any to 255.255.255.255
2506block in log quick on $ext_if from { 10.0.0.0/8, 172.16.0.0/12, \e
2507    192.168.0.0/16, 255.255.255.255/32 } to any
2508.Ed
2509.Pp
2510For ICMP,
2511pass out/in ping queries.
2512State matching is done on host addresses and ICMP ID (not type/code),
2513so replies (like 0/0 for 8/0) will match queries.
2514ICMP error messages (which always refer to a TCP/UDP packet)
2515are handled by the TCP/UDP states.
2516.Bd -literal -offset 4n
2517pass on $ext_if inet proto icmp all icmp-type 8 code 0
2518.Ed
2519.Pp
2520For UDP,
2521pass out all UDP connections.
2522DNS connections are passed in.
2523.Bd -literal -offset 4n
2524pass out on $ext_if proto udp all
2525pass in on $ext_if proto udp from any to any port domain
2526.Ed
2527.Pp
2528For TCP,
2529pass out all TCP connections and modulate state.
2530SSH, SMTP, DNS, and IDENT connections are passed in.
2531We do not allow Windows 9x SMTP connections since they are typically
2532a viral worm.
2533.Bd -literal -offset 4n
2534pass out on $ext_if proto tcp all modulate state
2535pass in on $ext_if proto tcp from any to any \e
2536    port { ssh, smtp, domain, auth }
2537block in on $ext_if proto tcp from any \e
2538    os { "Windows 95", "Windows 98" } to any port smtp
2539.Ed
2540.Pp
2541Here we pass in/out all IPv6 traffic:
2542note that we have to enable this in two different ways,
2543on both our physical interface and our tunnel.
2544.Bd -literal -offset 4n
2545pass quick on gif0 inet6
2546pass quick on $ext_if proto ipv6
2547.Ed
2548.Pp
2549This example illustrates packet tagging.
2550There are three interfaces: $int_if, $ext_if, and $wifi_if (wireless).
2551NAT is being done on $ext_if for all outgoing packets.
2552Packets in on $int_if are tagged and passed out on $ext_if.
2553All other outgoing packets
2554(i.e. packets from the wireless network)
2555are only permitted to access port 80.
2556.Bd -literal -offset 4n
2557pass in on $int_if from any to any tag INTNET
2558pass in on $wifi_if from any to any
2559
2560block out on $ext_if from any to any
2561pass out quick on $ext_if tagged INTNET
2562pass out on $ext_if proto tcp from any to any port 80
2563.Ed
2564.Pp
2565In this example,
2566we tag incoming packets as they are redirected to spamd(8).
2567The tag is used to pass those packets through the packet filter.
2568.Bd -literal -offset 4n
2569match in on $ext_if inet proto tcp from \*(Ltspammers\*(Gt to port smtp \e
2570     tag SPAMD rdr-to 127.0.0.1 port spamd
2571
2572block in on $ext_if
2573pass in on $ext_if inet proto tcp tagged SPAMD
2574.Ed
2575.Pp
2576This example maps incoming requests on port 80 to port 8080, on
2577which a daemon is running (because, for example, it is not run as root,
2578and therefore lacks permission to bind to port 80).
2579.Bd -literal -offset 4n
2580match in on $ext_if proto tcp from any to any port 80 \e
2581      rdr-to 127.0.0.1 port 8080
2582.Ed
2583.Pp
2584If a
2585.Ar pass
2586rule is used with the
2587.Ar quick
2588modifier, packets matching the translation rule are passed without
2589inspecting subsequent filter rules.
2590.Bd -literal -offset 4n
2591pass in quick on $ext_if proto tcp from any to any port 80 \e
2592      rdr-to 127.0.0.1 port 8080
2593.Ed
2594.Pp
2595In the example below, vlan12 is configured as 192.168.168.1;
2596the machine translates all packets coming from 192.168.168.0/24 to 204.92.77.111
2597when they are going out any interface except vlan12.
2598This has the net effect of making traffic from the 192.168.168.0/24
2599network appear as though it is the Internet routable address
2600204.92.77.111 to nodes behind any interface on the router except
2601for the nodes on vlan12.
2602Thus, 192.168.168.1 can talk to the 192.168.168.0/24 nodes.
2603.Bd -literal -offset 4n
2604match out on ! vlan12 from 192.168.168.0/24 to any nat-to 204.92.77.111
2605.Ed
2606.Pp
2607In the example below, the machine sits between a fake internal
2608144.19.74.* network, and a routable external IP of 204.92.77.100.
2609The last rule excludes protocol AH from being translated.
2610.Bd -literal -offset 4n
2611pass out on $ext_if from 144.19.74.0/24 nat-to 204.92.77.100
2612pass out on $ext_if proto ah from 144.19.74.0/24
2613.Ed
2614.Pp
2615In the example below, packets bound for one specific server, as well as those
2616generated by the sysadmins are not proxied; all other connections are.
2617.Bd -literal -offset 4n
2618pass in on $int_if proto { tcp, udp } from any to any port 80 \e
2619      rdr-to 127.0.0.1 port 80
2620pass in on $int_if proto { tcp, udp } from any to $server port 80
2621pass in on $int_if proto { tcp, udp } from $sysadmins to any port 80
2622.Ed
2623.Pp
2624This example maps outgoing packets' source port
2625to an assigned proxy port instead of an arbitrary port.
2626In this case, proxy outgoing isakmp with port 500 on the gateway.
2627.Bd -literal -offset 4n
2628match out on $ext_if inet proto udp from any port isakmp to any \e
2629    nat-to ($ext_if) port 500
2630.Ed
2631.Pp
2632One more example uses
2633.Ar rdr-to
2634to redirect a TCP and UDP port to an internal machine.
2635.Bd -literal -offset 4n
2636match in on $ext_if inet proto tcp from any to ($ext_if) port 8080 \e
2637      rdr-to 10.1.2.151 port 22
2638match in on $ext_if inet proto udp from any to ($ext_if) port 8080 \e
2639      rdr-to 10.1.2.151 port 53
2640.Ed
2641.Pp
2642In this example, a NAT gateway is set up to translate internal addresses
2643using a pool of public addresses (192.0.2.16/28).
2644A given source address is always translated to the same pool address by
2645using the source-hash keyword.
2646The gateway also translates incoming web server connections
2647to a group of web servers on the internal network.
2648.Bd -literal -offset 4n
2649match out on $ext_if inet from any to any nat-to 192.0.2.16/28 \e
2650    source-hash
2651match in  on $ext_if proto tcp from any to any port 80 \e
2652    rdr-to { 10.1.2.155 weight 2, 10.1.2.160 weight 1, \e
2653             10.1.2.161 weight 8 } round-robin
2654.Ed
2655.Pp
2656The bidirectional address translation example uses a single
2657.Ar binat-to
2658rule that expands to a
2659.Ar nat-to
2660and an
2661.Ar rdr-to
2662rule.
2663.Bd -literal -offset 4n
2664pass on $ext_if from 10.1.2.120 to any binat-to 192.0.2.17
2665.Ed
2666.Pp
2667The previous example is identical to the following set of rules:
2668.Bd -literal -offset 4n
2669pass out on $ext_if inet from 10.1.2.120 to any \e
2670      nat-to 192.0.2.17 static-port
2671pass in on $ext_if inet from any to 192.0.2.17 rdr-to 10.1.2.120
2672.Ed
2673.Sh GRAMMAR
2674Syntax for
2675.Nm
2676in BNF:
2677.Bd -literal
2678line           = ( option | pf-rule |
2679                 antispoof-rule | altq-rule | queue-rule | anchor-rule |
2680                 anchor-close | load-anchor | table-rule | include )
2681
2682option         = "set" ( [ "timeout" ( timeout | "{" timeout-list "}" ) ] |
2683                 [ "ruleset-optimization" [ "none" | "basic" |
2684                 "profile" ] ] |
2685                 [ "optimization" [ "default" | "normal" | "high-latency" |
2686                 "satellite" | "aggressive" | "conservative" ] ]
2687                 [ "limit" ( limit-item | "{" limit-list "}" ) ] |
2688                 [ "loginterface" ( interface-name | "none" ) ] |
2689                 [ "block-policy" ( "drop" | "return" ) ] |
2690                 [ "state-policy" ( "if-bound" | "floating" ) ]
2691                 [ "state-defaults" state-opts ]
2692                 [ "fingerprints" filename ] |
2693                 [ "skip on" ifspec ] |
2694                 [ "debug" ( "none" | "urgent" | "misc" | "loud" ) ] |
2695		 [ "reassemble" ( "yes" | "no" ) [ "no-df" ] ] )
2696
2697pf-rule        = action [ ( "in" | "out" ) ]
2698                 [ "log" [ "(" logopts ")"] ] [ "quick" ]
2699                 [ "on" ( ifspec | "rdomain" number ) ] [ af ]
2700                 [ protospec ] hosts [ filteropts ]
2701
2702logopts        = logopt [ [ "," ] logopts ]
2703logopt         = "all" | "matches" | "user" | "to" interface-name
2704
2705filteropts     = filteropt [ [ "," ] filteropts ]
2706filteropt      = user | group | flags | icmp-type | icmp6-type |
2707                 "tos" tos |
2708                 ( "no" | "keep" | "modulate" | "synproxy" ) "state"
2709                 [ "(" state-opts ")" ] | "scrub" "(" scrubopts ")" |
2710                 "fragment" | "allow-opts" | "once" |
2711		 "divert-packet" "port" port | "divert-reply" |
2712		 "divert-to" host "port" port |
2713                 "label" string | "tag" string | [ ! ] "tagged" string |
2714                 "prio" ( number | "(" number [ [ "," ] number ] ")" ) |
2715                 "queue" ( string | "(" string [ [ "," ] string ] ")" ) |
2716                 "rtable" number | "probability" number"%" |
2717		 "binat-to" ( redirhost | "{" redirhost-list "}" )
2718		 [ portspec ] [ pooltype ] |
2719		 "rdr-to" ( redirhost | "{" redirhost-list "}" )
2720		 [ portspec ] [ pooltype ] |
2721		 "nat-to" ( redirhost | "{" redirhost-list "}" )
2722		 [ portspec ] [ pooltype ] [ "static-port" ] |
2723		 [ route ] |
2724		 [ "received-on" ( interface-name | interface-group ) ]
2725
2726scrubopts      = scrubopt [ [ "," ] scrubopts ]
2727scrubopt       = "no-df" | "min-ttl" number | "max-mss" number |
2728                 "set-tos" tos | "reassemble tcp" | "random-id"
2729
2730antispoof-rule = "antispoof" [ "log" ] [ "quick" ]
2731                 "for" ifspec [ af ] [ "label" string ]
2732
2733table-rule     = "table" "\*(Lt" string "\*(Gt" [ tableopts ]
2734tableopts      = tableopt [ tableopts ]
2735tableopt       = "persist" | "const" | "counters" |
2736                 "file" string | "{" [ tableaddrs ] "}"
2737tableaddrs     = tableaddr-spec [ [ "," ] tableaddrs ]
2738tableaddr-spec = [ "!" ] tableaddr [ "/" mask-bits ]
2739tableaddr      = hostname | ifspec | "self" |
2740                 ipv4-dotted-quad | ipv6-coloned-hex
2741
2742altq-rule      = "altq on" interface-name queueopts-list
2743                 "queue" subqueue
2744queue-rule     = "queue" string [ "on" interface-name ] queueopts-list
2745                 subqueue
2746
2747anchor-rule    = "anchor" [ string ] [ ( "in" | "out" ) ] [ "on" ifspec ]
2748                 [ af ] [ protospec ] [ hosts ] [ filteropt-list ] [ "{" ]
2749
2750anchor-close   = "}"
2751
2752load-anchor    = "load anchor" string "from" filename
2753
2754queueopts-list = queueopts-list queueopts | queueopts
2755queueopts      = [ "bandwidth" bandwidth-spec ] |
2756                 [ "qlimit" number ] | [ "tbrsize" number ] |
2757                 [ "priority" number ] | [ schedulers ]
2758schedulers     = ( cbq-def | priq-def | hfsc-def )
2759bandwidth-spec = "number" ( "b" | "Kb" | "Mb" | "Gb" | "%" )
2760
2761action         = "pass" | "match" | "block" [ return ]
2762return         = "drop" | "return" |
2763                 "return-rst" [ "(" "ttl" number ")" ] |
2764                 "return-icmp" [ "(" icmpcode [ [ "," ] icmp6code ] ")" ] |
2765                 "return-icmp6" [ "(" icmp6code ")" ]
2766icmpcode       = ( icmp-code-name | icmp-code-number )
2767icmp6code      = ( icmp6-code-name | icmp6-code-number )
2768
2769ifspec         = ( [ "!" ] ( interface-name | interface-group ) ) |
2770                 "{" interface-list "}"
2771interface-list = [ "!" ] ( interface-name | interface-group )
2772                 [ [ "," ] interface-list ]
2773route          = ( "route-to" | "reply-to" | "dup-to" )
2774                 ( routehost | "{" routehost-list "}" )
2775                 [ pooltype ]
2776af             = "inet" | "inet6"
2777
2778protospec      = "proto" ( proto-name | proto-number |
2779                 "{" proto-list "}" )
2780proto-list     = ( proto-name | proto-number ) [ [ "," ] proto-list ]
2781
2782hosts          = "all" |
2783                 "from" ( "any" | "no-route" | "urpf-failed" | "self" |
2784                 host | "{" host-list "}" | "route" string ) [ port ]
2785                 [ os ]
2786                 "to"   ( "any" | "no-route" | "self" | host |
2787                 "{" host-list "}" | "route" string ) [ port ]
2788
2789ipspec         = "any" | host | "{" host-list "}"
2790host           = [ "!" ] ( address [ "weight" number ] |
2791                 address [ "/" mask-bits ] [ "weight" number ] |
2792                 "\*(Lt" string "\*(Gt" )
2793redirhost      = address [ "/" mask-bits ]
2794routehost      = host | host "@" interface-name |
2795                 "(" interface-name [ address [ "/" mask-bits ] ] ")"
2796address        = ( interface-name | interface-group |
2797                 "(" ( interface-name | interface-group ) ")" |
2798                 hostname | ipv4-dotted-quad | ipv6-coloned-hex )
2799host-list      = host [ [ "," ] host-list ]
2800redirhost-list = redirhost [ [ "," ] redirhost-list ]
2801routehost-list = routehost [ [ "," ] routehost-list ]
2802
2803port           = "port" ( unary-op | binary-op | "{" op-list "}" )
2804portspec       = "port" ( number | name ) [ ":" ( "*" | number | name ) ]
2805os             = "os"  ( os-name | "{" os-list "}" )
2806user           = "user" ( unary-op | binary-op | "{" op-list "}" )
2807group          = "group" ( unary-op | binary-op | "{" op-list "}" )
2808
2809unary-op       = [ "=" | "!=" | "\*(Lt" | "\*(Le" | "\*(Gt" | "\*(Ge" ]
2810                 ( name | number )
2811binary-op      = number ( "\*(Lt\*(Gt" | "\*(Gt\*(Lt" | ":" ) number
2812op-list        = ( unary-op | binary-op ) [ [ "," ] op-list ]
2813
2814os-name        = operating-system-name
2815os-list        = os-name [ [ "," ] os-list ]
2816
2817flags          = "flags" ( [ flag-set ] "/"  flag-set | "any" )
2818flag-set       = [ "F" ] [ "S" ] [ "R" ] [ "P" ] [ "A" ] [ "U" ] [ "E" ]
2819                 [ "W" ]
2820
2821icmp-type      = "icmp-type" ( icmp-type-code | "{" icmp-list "}" )
2822icmp6-type     = "icmp6-type" ( icmp-type-code | "{" icmp-list "}" )
2823icmp-type-code = ( icmp-type-name | icmp-type-number )
2824                 [ "code" ( icmp-code-name | icmp-code-number ) ]
2825icmp-list      = icmp-type-code [ [ "," ] icmp-list ]
2826
2827tos            = ( "lowdelay" | "throughput" | "reliability" |
2828                 [ "0x" ] number )
2829
2830state-opts     = state-opt [ [ "," ] state-opts ]
2831state-opt      = ( "max" number | "no-sync" | timeout | "sloppy" |
2832                 "pflow" | "source-track" [ ( "rule" | "global" ) ] |
2833                 "max-src-nodes" number | "max-src-states" number |
2834                 "max-src-conn" number |
2835                 "max-src-conn-rate" number "/" number |
2836                 "overload" "\*(Lt" string "\*(Gt" [ "flush" [ "global" ] ] |
2837                 "if-bound" | "floating" )
2838
2839timeout-list   = timeout [ [ "," ] timeout-list ]
2840timeout        = ( "tcp.first" | "tcp.opening" | "tcp.established" |
2841                 "tcp.closing" | "tcp.finwait" | "tcp.closed" |
2842                 "udp.first" | "udp.single" | "udp.multiple" |
2843                 "icmp.first" | "icmp.error" |
2844                 "other.first" | "other.single" | "other.multiple" |
2845                 "frag" | "interval" | "src.track" |
2846                 "adaptive.start" | "adaptive.end" ) number
2847
2848limit-list     = limit-item [ [ "," ] limit-list ]
2849limit-item     = ( "states" | "frags" | "src-nodes" | "tables" |
2850                 "table-entries" ) number
2851
2852pooltype       = ( "bitmask" | "least-states" |
2853                 "random" | "round-robin" |
2854                 "source-hash" [ ( hex-key | string-key ) ] )
2855                 [ sticky-address ]
2856
2857subqueue       = string | "{" queue-list "}"
2858queue-list     = string [ [ "," ] string ]
2859cbq-def        = "cbq" [ "(" cbq-opt [ [ "," ] cbq-opt ] ")" ]
2860priq-def       = "priq" [ "(" priq-opt [ [ "," ] priq-opt ] ")" ]
2861hfsc-def       = "hfsc" [ "(" hfsc-opt [ [ "," ] hfsc-opt ] ")" ]
2862cbq-opt        = ( "default" | "borrow" | "red" | "ecn" )
2863priq-opt       = ( "default" | "red" | "ecn" )
2864hfsc-opt       = ( "default" | "red" | "ecn" |
2865                 linkshare-sc | realtime-sc | upperlimit-sc )
2866linkshare-sc   = "linkshare" sc-spec
2867realtime-sc    = "realtime" sc-spec
2868upperlimit-sc  = "upperlimit" sc-spec
2869sc-spec        = ( bandwidth-spec |
2870                 "(" bandwidth-spec number bandwidth-spec ")" )
2871include        = "include" filename
2872.Ed
2873.Sh FILES
2874.Bl -tag -width "/etc/protocolsXXX" -compact
2875.It Pa /etc/hosts
2876Host name database.
2877.It Pa /etc/pf.conf
2878Default location of the ruleset file.
2879.It Pa /etc/pf.os
2880Default location of OS fingerprints.
2881.It Pa /etc/protocols
2882Protocol name database.
2883.It Pa /etc/services
2884Service name database.
2885.El
2886.Sh SEE ALSO
2887.Xr pf 4 ,
2888.Xr pflow 4 ,
2889.Xr pfsync 4 ,
2890.Xr pf.os 5 ,
2891.Xr pfctl 8 ,
2892.Xr pflogd 8
2893.Sh HISTORY
2894The
2895.Nm
2896file format first appeared in
2897.Ox 3.0 .
2898