1.\" $OpenBSD: pf.conf.5,v 1.402 2008/06/11 07:21:00 jmc 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 June 26, 2018 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.Sh STATEMENT ORDER 43There are seven types of statements in 44.Nm : 45.Bl -tag -width xxxx 46.It Cm Macros 47User-defined variables may be defined and used later, simplifying 48the configuration file. 49Macros must be defined before they are referenced in 50.Nm . 51.It Cm Tables 52Tables provide a mechanism for increasing the performance and flexibility of 53rules with large numbers of source or destination addresses. 54.It Cm Options 55Options tune the behaviour of the packet filtering engine. 56.It Cm Traffic Normalization Li (e.g.\& Em scrub ) 57Traffic normalization protects internal machines against inconsistencies 58in Internet protocols and implementations. 59.It Cm Queueing 60Queueing provides rule-based bandwidth control. 61.It Cm Translation Li (Various forms of NAT) 62Translation rules specify how addresses are to be mapped or redirected to 63other addresses. 64.It Cm Packet Filtering 65Packet filtering provides rule-based blocking or passing of packets. 66.El 67.Pp 68With the exception of 69.Cm macros 70and 71.Cm tables , 72the types of statements should be grouped and appear in 73.Nm 74in the order shown above, as this matches the operation of the underlying 75packet filtering engine. 76By default 77.Xr pfctl 8 78enforces this order (see 79.Ar set require-order 80below). 81.Pp 82Comments can be put anywhere in the file using a hash mark 83.Pq Sq # , 84and extend to the end of the current line. 85.Pp 86Additional configuration files can be included with the 87.Ic include 88keyword, for example: 89.Bd -literal -offset indent 90include "/etc/pf/sub.filter.conf" 91.Ed 92.Sh MACROS 93Macros can be defined that will later be expanded in context. 94Macro names must start with a letter, and may contain letters, digits 95and underscores. 96Macro names may not be reserved words (for example 97.Ar pass , 98.Ar in , 99.Ar out ) . 100Macros are not expanded inside quotes. 101.Pp 102For example, 103.Bd -literal -offset indent 104ext_if = \&"kue0\&" 105all_ifs = \&"{\&" $ext_if lo0 \&"}\&" 106pass out on $ext_if from any to any 107pass in on $ext_if proto tcp from any to any port 25 108.Ed 109.Sh TABLES 110Tables are named structures which can hold a collection of addresses and 111networks. 112Lookups against tables in 113.Xr pf 4 114are relatively fast, making a single rule with tables much more efficient, 115in terms of 116processor usage and memory consumption, than a large number of rules which 117differ only in IP address (either created explicitly or automatically by rule 118expansion). 119.Pp 120Tables can be used as the source or destination of filter rules, 121.Ar scrub 122rules 123or 124translation rules such as 125.Ar nat 126or 127.Ar rdr 128(see below for details on the various rule types). 129Tables can also be used for the redirect address of 130.Ar nat 131and 132.Ar rdr 133rules and in the routing options of filter rules, but only for 134.Ar round-robin 135pools. 136.Pp 137Tables can be defined with any of the following 138.Xr pfctl 8 139mechanisms. 140As with macros, reserved words may not be used as table names. 141.Bl -tag -width "manually" 142.It Ar manually 143Persistent tables can be manually created with the 144.Ar add 145or 146.Ar replace 147option of 148.Xr pfctl 8 , 149before or after the ruleset has been loaded. 150.It Pa pf.conf 151Table definitions can be placed directly in this file, and loaded at the 152same time as other rules are loaded, atomically. 153Table definitions inside 154.Nm 155use the 156.Ar table 157statement, and are especially useful to define non-persistent tables. 158The contents of a pre-existing table defined without a list of addresses 159to initialize it is not altered when 160.Nm 161is loaded. 162A table initialized with the empty list, 163.Li { } , 164will be cleared on load. 165.El 166.Pp 167Tables may be defined with the following two attributes: 168.Bl -tag -width persist 169.It Ar persist 170The 171.Ar persist 172flag forces the kernel to keep the table even when no rules refer to it. 173If the flag is not set, the kernel will automatically remove the table 174when the last rule referring to it is flushed. 175.It Ar const 176The 177.Ar const 178flag prevents the user from altering the contents of the table once it 179has been created. 180Without that flag, 181.Xr pfctl 8 182can be used to add or remove addresses from the table at any time, even 183when running with 184.Xr securelevel 8 185= 2. 186.It Ar counters 187The 188.Ar counters 189flag enables per-address packet and byte counters which can be displayed with 190.Xr pfctl 8 . 191.El 192.Pp 193For example, 194.Bd -literal -offset indent 195table \*(Ltprivate\*(Gt const { 10/8, 172.16/12, 192.168/16 } 196table \*(Ltbadhosts\*(Gt persist 197block on fxp0 from { \*(Ltprivate\*(Gt, \*(Ltbadhosts\*(Gt } to any 198.Ed 199.Pp 200creates a table called private, to hold RFC 1918 private network 201blocks, and a table called badhosts, which is initially empty. 202A filter rule is set up to block all traffic coming from addresses listed in 203either table. 204The private table cannot have its contents changed and the badhosts table 205will exist even when no active filter rules reference it. 206Addresses may later be added to the badhosts table, so that traffic from 207these hosts can be blocked by using 208.Bd -literal -offset indent 209# pfctl -t badhosts -Tadd 204.92.77.111 210.Ed 211.Pp 212A table can also be initialized with an address list specified in one or more 213external files, using the following syntax: 214.Bd -literal -offset indent 215table \*(Ltspam\*(Gt persist file \&"/etc/spammers\&" file \&"/etc/openrelays\&" 216block on fxp0 from \*(Ltspam\*(Gt to any 217.Ed 218.Pp 219The files 220.Pa /etc/spammers 221and 222.Pa /etc/openrelays 223list IP addresses, one per line. 224Any lines beginning with a # are treated as comments and ignored. 225In addition to being specified by IP address, hosts may also be 226specified by their hostname. 227When the resolver is called to add a hostname to a table, 228.Em all 229resulting IPv4 and IPv6 addresses are placed into the table. 230IP addresses can also be entered in a table by specifying a valid interface 231name, a valid interface group or the 232.Em self 233keyword, in which case all addresses assigned to the interface(s) will be 234added to the table. 235.Sh OPTIONS 236.Xr pf 4 237may be tuned for various situations using the 238.Ar set 239command. 240.Bl -tag -width xxxx 241.It Ar set timeout 242.Pp 243.Bl -tag -width "src.track" -compact 244.It Ar interval 245Interval between purging expired states and fragments. 246.It Ar frag 247Seconds before an unassembled fragment is expired. 248.It Ar src.track 249Length of time to retain a source tracking entry after the last state 250expires. 251.El 252.Pp 253When a packet matches a stateful connection, the seconds to live for the 254connection will be updated to that of the 255.Ar proto.modifier 256which corresponds to the connection state. 257Each packet which matches this state will reset the TTL. 258Tuning these values may improve the performance of the 259firewall at the risk of dropping valid idle connections. 260.Pp 261.Bl -tag -width xxxx -compact 262.It Ar tcp.first 263The state after the first packet. 264.It Ar tcp.opening 265The state before the destination host ever sends a packet. 266.It Ar tcp.established 267The fully established state. 268.It Ar tcp.closing 269The state after the first FIN has been sent. 270.It Ar tcp.finwait 271The state after both FINs have been exchanged and the connection is closed. 272Some hosts (notably web servers on Solaris) send TCP packets even after closing 273the connection. 274Increasing 275.Ar tcp.finwait 276(and possibly 277.Ar tcp.closing ) 278can prevent blocking of such packets. 279.It Ar tcp.closed 280The state after one endpoint sends an RST. 281.El 282.Pp 283ICMP and UDP are handled in a fashion similar to TCP, but with a much more 284limited set of states: 285.Pp 286.Bl -tag -width xxxx -compact 287.It Ar udp.first 288The state after the first packet. 289.It Ar udp.single 290The state if the source host sends more than one packet but the destination 291host has never sent one back. 292.It Ar udp.multiple 293The state if both hosts have sent packets. 294.It Ar icmp.first 295The state after the first packet. 296.It Ar icmp.error 297The state after an ICMP error came back in response to an ICMP packet. 298.El 299.Pp 300Other protocols are handled similarly to UDP: 301.Pp 302.Bl -tag -width xxxx -compact 303.It Ar other.first 304.It Ar other.single 305.It Ar other.multiple 306.El 307.Pp 308Timeout values can be reduced adaptively as the number of state table 309entries grows. 310.Pp 311.Bl -tag -width xxxx -compact 312.It Ar adaptive.start 313When the number of state entries exceeds this value, adaptive scaling 314begins. 315All timeout values are scaled linearly with factor 316(adaptive.end - number of states) / (adaptive.end - adaptive.start). 317.It Ar adaptive.end 318When reaching this number of state entries, all timeout values become 319zero, effectively purging all state entries immediately. 320This value is used to define the scale factor, it should not actually 321be reached (set a lower state limit, see below). 322.El 323.Pp 324Adaptive timeouts are enabled by default, with an adaptive.start value 325equal to 60% of the state limit, and an adaptive.end value equal to 326120% of the state limit. 327They can be disabled by setting both adaptive.start and adaptive.end to 0. 328.Pp 329The adaptive timeout values can be defined both globally and for each rule. 330When used on a per-rule basis, the values relate to the number of 331states created by the rule, otherwise to the total number of 332states. 333.Pp 334For example: 335.Bd -literal -offset indent 336set timeout tcp.first 120 337set timeout tcp.established 86400 338set timeout { adaptive.start 6000, adaptive.end 12000 } 339set limit states 10000 340.Ed 341.Pp 342With 9000 state table entries, the timeout values are scaled to 50% 343(tcp.first 60, tcp.established 43200). 344.It Ar set loginterface 345Enable collection of packet and byte count statistics for the given 346interface or interface group. 347These statistics can be viewed using 348.Bd -literal -offset indent 349# pfctl -s info 350.Ed 351.Pp 352In this example 353.Xr pf 4 354collects statistics on the interface named dc0: 355.Bd -literal -offset indent 356set loginterface dc0 357.Ed 358.Pp 359One can disable the loginterface using: 360.Bd -literal -offset indent 361set loginterface none 362.Ed 363.It Ar set limit 364Sets hard limits on the memory pools used by the packet filter. 365See 366.Xr zone 9 367for an explanation of memory pools. 368.Pp 369For example, 370.Bd -literal -offset indent 371set limit states 20000 372.Ed 373.Pp 374sets the maximum number of entries in the memory pool used by state table 375entries (generated by 376.Ar pass 377rules which do not specify 378.Ar no state ) 379to 20000. 380Using 381.Bd -literal -offset indent 382set limit frags 20000 383.Ed 384.Pp 385sets the maximum number of entries in the memory pool used for fragment 386reassembly (generated by 387.Ar scrub 388rules) to 20000. 389Using 390.Bd -literal -offset indent 391set limit src-nodes 2000 392.Ed 393.Pp 394sets the maximum number of entries in the memory pool used for tracking 395source IP addresses (generated by the 396.Ar sticky-address 397and 398.Ar src.track 399options) to 2000. 400Using 401.Bd -literal -offset indent 402set limit tables 1000 403set limit table-entries 100000 404.Ed 405.Pp 406sets limits on the memory pools used by tables. 407The first limits the number of tables that can exist to 1000. 408The second limits the overall number of addresses that can be stored 409in tables to 100000. 410.Pp 411Various limits can be combined on a single line: 412.Bd -literal -offset indent 413set limit { states 20000, frags 20000, src-nodes 2000 } 414.Ed 415.It Ar set ruleset-optimization 416.Bl -tag -width xxxxxxxx -compact 417.It Ar none 418Disable the ruleset optimizer. 419.Pp 420.It Ar basic 421Enable basic ruleset optimization. 422This is the default behaviour. 423Basic ruleset optimization does four things to improve the 424performance of ruleset evaluations: 425.Pp 426.Bl -enum -compact 427.It 428remove duplicate rules 429.It 430remove rules that are a subset of another rule 431.It 432combine multiple rules into a table when advantageous 433.It 434re-order the rules to improve evaluation performance 435.El 436.Pp 437.It Ar profile 438Uses the currently loaded ruleset as a feedback profile to tailor the 439ordering of quick rules to actual network traffic. 440.El 441.Pp 442It is important to note that the ruleset optimizer will modify the ruleset 443to improve performance. 444A side effect of the ruleset modification is that per-rule accounting 445statistics will have different meanings than before. 446If per-rule accounting is important for billing purposes or whatnot, 447either the ruleset optimizer should not be used or a label field should 448be added to all of the accounting rules to act as optimization barriers. 449.Pp 450Optimization can also be set as a command-line argument to 451.Xr pfctl 8 , 452overriding the settings in 453.Nm . 454.It Ar set optimization 455Optimize state timeouts for one of the following network environments: 456.Pp 457.Bl -tag -width xxxx -compact 458.It Ar normal 459A normal network environment. 460Suitable for almost all networks. 461.It Ar high-latency 462A high-latency environment (such as a satellite connection). 463.It Ar satellite 464Alias for 465.Ar high-latency . 466.It Ar aggressive 467Aggressively expire connections. 468This can greatly reduce the memory usage of the firewall at the cost of 469dropping idle connections early. 470.It Ar conservative 471Extremely conservative settings. 472Avoid dropping legitimate connections at the 473expense of greater memory utilization (possibly much greater on a busy 474network) and slightly increased processor utilization. 475.El 476.Pp 477For example: 478.Bd -literal -offset indent 479set optimization aggressive 480.Ed 481.It Ar set keep-policy keep_rule 482The 483.Ar keep-policy 484option sets the default state retention policy for all 485.Ar pass 486rules. 487See 488.Sx STATEFUL TRACKING OPTIONS 489or 490.Sx GRAMMAR 491(keep) for format of 492.Ar keep_rule . 493Any 494.Ar no Ns / Ns Ar keep Ns / Ns Ar modulate Ns / Ns Ar synproxy state 495directives in a 496.Ar pass 497rule will override the default. 498For example: 499.Bd -literal -offset indent 500set keep-policy keep state (pickups) 501.Ed 502.It Ar set block-policy 503The 504.Ar block-policy 505option sets the default behaviour for the packet 506.Ar block 507action: 508.Pp 509.Bl -tag -width xxxxxxxx -compact 510.It Ar drop 511Packet is silently dropped. 512.It Ar return 513A TCP RST is returned for blocked TCP packets, 514an ICMP UNREACHABLE is returned for blocked UDP packets, 515and all other packets are silently dropped. 516.El 517.Pp 518For example: 519.Bd -literal -offset indent 520set block-policy return 521.Ed 522.It Ar set state-policy 523The 524.Ar state-policy 525option sets the default behaviour for states: 526.Pp 527.Bl -tag -width if-bound -compact 528.It Ar if-bound 529States are bound to interface. 530.It Ar floating 531States can match packets on any interfaces (the default). 532.El 533.Pp 534For example: 535.Bd -literal -offset indent 536set state-policy if-bound 537.Ed 538.It Ar set hostid 539The 32-bit 540.Ar hostid 541identifies this firewall's state table entries to other firewalls 542in a 543.Xr pfsync 4 544failover cluster. 545By default the hostid is set to a pseudo-random value, however it may be 546desirable to manually configure it, for example to more easily identify the 547source of state table entries. 548.Bd -literal -offset indent 549set hostid 1 550.Ed 551.Pp 552The hostid may be specified in either decimal or hexadecimal. 553.It Ar set require-order 554By default 555.Xr pfctl 8 556enforces an ordering of the statement types in the ruleset to: 557.Em options , 558.Em normalization , 559.Em queueing , 560.Em translation , 561.Em filtering . 562Setting this option to 563.Ar no 564disables this enforcement. 565There may be non-trivial and non-obvious implications to an out of 566order ruleset. 567Consider carefully before disabling the order enforcement. 568.It Ar set fingerprints 569Load fingerprints of known operating systems from the given filename. 570By default fingerprints of known operating systems are automatically 571loaded from 572.Xr pf.os 5 573in 574.Pa /etc 575but can be overridden via this option. 576Setting this option may leave a small period of time where the fingerprints 577referenced by the currently active ruleset are inconsistent until the new 578ruleset finishes loading. 579.Pp 580For example: 581.Pp 582.Dl set fingerprints \&"/etc/pf.os.devel\&" 583.It Ar set skip on Aq Ar ifspec 584List interfaces for which packets should not be filtered. 585Packets passing in or out on such interfaces are passed as if pf was 586disabled, i.e. pf does not process them in any way. 587This can be useful on loopback and other virtual interfaces, when 588packet filtering is not desired and can have unexpected effects. 589For example: 590.Pp 591.Dl set skip on lo0 592.It Ar set debug 593Set the debug 594.Ar level 595to one of the following: 596.Pp 597.Bl -tag -width xxxxxxxxxxxx -compact 598.It Ar none 599Don't generate debug messages. 600.It Ar urgent 601Generate debug messages only for serious errors. 602.It Ar misc 603Generate debug messages for various errors. 604.It Ar loud 605Generate debug messages for common conditions. 606.El 607.El 608.Sh TRAFFIC NORMALIZATION 609Traffic normalization is used to sanitize packet content in such 610a way that there are no ambiguities in packet interpretation on 611the receiving side. 612The normalizer does IP fragment reassembly to prevent attacks 613that confuse intrusion detection systems by sending overlapping 614IP fragments. 615Packet normalization is invoked with the 616.Ar scrub 617directive. 618.Pp 619.Ar scrub 620has the following options: 621.Bl -tag -width xxxx 622.It Ar no-df 623Clears the 624.Ar dont-fragment 625bit from a matching IP packet. 626Some operating systems are known to generate fragmented packets with the 627.Ar dont-fragment 628bit set. 629This is particularly true with NFS. 630.Ar Scrub 631will drop such fragmented 632.Ar dont-fragment 633packets unless 634.Ar no-df 635is specified. 636.Pp 637Unfortunately some operating systems also generate their 638.Ar dont-fragment 639packets with a zero IP identification field. 640Clearing the 641.Ar dont-fragment 642bit on packets with a zero IP ID may cause deleterious results if an 643upstream router later fragments the packet. 644Using the 645.Ar random-id 646modifier (see below) is recommended in combination with the 647.Ar no-df 648modifier to ensure unique IP identifiers. 649.It Ar min-ttl Aq Ar number 650Enforces a minimum TTL for matching IP packets. 651.It Ar max-mss Aq Ar number 652Enforces a maximum MSS for matching TCP packets. 653.It Xo Ar set-tos Aq Ar string 654.No \*(Ba Aq Ar number 655.Xc 656Enforces a 657.Em TOS 658for matching IP packets. 659.Em TOS 660may be 661given as one of 662.Ar lowdelay , 663.Ar throughput , 664.Ar reliability , 665or as either hex or decimal. 666.It Ar random-id 667Replaces the IP identification field with random values to compensate 668for predictable values generated by many hosts. 669This option only applies to packets that are not fragmented 670after the optional fragment reassembly. 671.It Ar fragment reassemble 672Using 673.Ar scrub 674rules, fragments can be reassembled by normalization. 675In this case, fragments are buffered until they form a complete 676packet, and only the completed packet is passed on to the filter. 677The advantage is that filter rules have to deal only with complete 678packets, and can ignore fragments. 679The drawback of caching fragments is the additional memory cost. 680But the full reassembly method is the only method that currently works 681with NAT. 682This is the default behavior of a 683.Ar scrub 684rule if no fragmentation modifier is supplied. 685.It Ar fragment crop 686The default fragment reassembly method is expensive, hence the option 687to crop is provided. 688In this case, 689.Xr pf 4 690will track the fragments and cache a small range descriptor. 691Duplicate fragments are dropped and overlaps are cropped. 692Thus data will only occur once on the wire with ambiguities resolving to 693the first occurrence. 694Unlike the 695.Ar fragment reassemble 696modifier, fragments are not buffered, they are passed as soon as they 697are received. 698The 699.Ar fragment crop 700reassembly mechanism does not yet work with NAT. 701.It Ar fragment drop-ovl 702This option is similar to the 703.Ar fragment crop 704modifier except that all overlapping or duplicate fragments will be 705dropped, and all further corresponding fragments will be 706dropped as well. 707.It Ar reassemble tcp 708Statefully normalizes TCP connections. 709.Ar scrub reassemble tcp 710rules may not have the direction (in/out) specified. 711.Ar reassemble tcp 712performs the following normalizations: 713.Pp 714.Bl -tag -width timeout -compact 715.It ttl 716Neither side of the connection is allowed to reduce their IP TTL. 717An attacker may send a packet such that it reaches the firewall, affects 718the firewall state, and expires before reaching the destination host. 719.Ar reassemble tcp 720will raise the TTL of all packets back up to the highest value seen on 721the connection. 722.It timestamp modulation 723Modern TCP stacks will send a timestamp on every TCP packet and echo 724the other endpoint's timestamp back to them. 725Many operating systems will merely start the timestamp at zero when 726first booted, and increment it several times a second. 727The uptime of the host can be deduced by reading the timestamp and multiplying 728by a constant. 729Also observing several different timestamps can be used to count hosts 730behind a NAT device. 731And spoofing TCP packets into a connection requires knowing or guessing 732valid timestamps. 733Timestamps merely need to be monotonically increasing and not derived off a 734guessable base time. 735.Ar reassemble tcp 736will cause 737.Ar scrub 738to modulate the TCP timestamps with a random number. 739.It extended PAWS checks 740There is a problem with TCP on long fat pipes, in that a packet might get 741delayed for longer than it takes the connection to wrap its 32-bit sequence 742space. 743In such an occurrence, the old packet would be indistinguishable from a 744new packet and would be accepted as such. 745The solution to this is called PAWS: Protection Against Wrapped Sequence 746numbers. 747It protects against it by making sure the timestamp on each packet does 748not go backwards. 749.Ar reassemble tcp 750also makes sure the timestamp on the packet does not go forward more 751than the RFC allows. 752By doing this, 753.Xr pf 4 754artificially extends the security of TCP sequence numbers by 10 to 18 755bits when the host uses appropriately randomized timestamps, since a 756blind attacker would have to guess the timestamp as well. 757.El 758.El 759.Pp 760For example, 761.Bd -literal -offset indent 762scrub in on $ext_if all fragment reassemble 763.Ed 764.Pp 765The 766.Ar no 767option prefixed to a scrub rule causes matching packets to remain unscrubbed, 768much in the same way as 769.Ar drop quick 770works in the packet filter (see below). 771This mechanism should be used when it is necessary to exclude specific packets 772from broader scrub rules. 773.Sh QUEUEING 774Packets can be assigned to queues for the purpose of bandwidth 775control. 776At least two declarations are required to configure queues, and later 777any packet filtering rule can reference the defined queues by name. 778During the filtering component of 779.Nm , 780the last referenced 781.Ar queue 782name is where any packets from 783.Ar pass 784rules will be queued, while for 785.Ar block 786rules it specifies where any resulting ICMP or TCP RST 787packets should be queued. 788The 789.Ar scheduler 790defines the algorithm used to decide which packets get delayed, dropped, or 791sent out immediately. 792There are four 793.Ar schedulers 794currently supported. 795.Bl -tag -width ".Ar fairq" 796.It Ar cbq 797Class Based Queueing. 798.Ar Queues 799attached to an interface build a tree, thus each 800.Ar queue 801can have further child 802.Ar queues . 803Each queue can have a 804.Ar priority 805and a 806.Ar bandwidth 807assigned. 808.Ar Priority 809mainly controls the time packets take to get sent out, while 810.Ar bandwidth 811has primarily effects on throughput. 812.Ar cbq 813achieves both partitioning and sharing of link bandwidth 814by hierarchically structured classes. 815Each class has its own 816.Ar queue 817and is assigned its share of 818.Ar bandwidth . 819A child class can borrow bandwidth from its parent class 820as long as excess bandwidth is available 821(see the option 822.Ar borrow , 823below). 824.It Ar priq 825Priority Queueing. 826.Ar Queues 827are flat attached to the interface, thus, 828.Ar queues 829cannot have further child 830.Ar queues . 831Each 832.Ar queue 833has a unique 834.Ar priority 835assigned, ranging from 0 to 15. 836Packets in the 837.Ar queue 838with the highest 839.Ar priority 840are processed first. 841.It Ar hfsc 842Hierarchical Fair Service Curve. 843.Ar Queues 844attached to an interface build a tree, thus each 845.Ar queue 846can have further child 847.Ar queues . 848Each queue can have a 849.Ar priority 850and a 851.Ar bandwidth 852assigned. 853.Ar Priority 854mainly controls the time packets take to get sent out, while 855.Ar bandwidth 856primarily affects throughput. 857.Ar hfsc 858supports both link-sharing and guaranteed real-time services. 859It employs a service curve based QoS model, 860and its unique feature is an ability to decouple 861.Ar delay 862and 863.Ar bandwidth 864allocation. 865.It Ar fairq 866Fair Queue. 867.Ar Queues 868are flat attached to the interface, thus, 869.Ar queues 870cannot have further child 871.Ar queues . 872Each queue must be given a unique 873.Ar priority 874and one must be marked 875as the default queue. 876Each queue implements a number of 877.Ar buckets 878(default 256) which sorts the 879traffic based on a hash key generated by the 880.Ar keep state 881facility in your 882.Ar pass 883rules. 884Each bucket contains a list of packets controlled by 885.Ar qlimit . 886In order for 887.Ar fairq 888to function properly, 889.Ar keep state 890must be enabled on most of the rule sets that route packets to the queue. 891Any rules for which keep state is not enabled are added to the end of the 892queue. 893If you do not wish keep state to do TCP sequence space checks use 894.Ar "keep state (no-pickups)" 895or 896.Ar "keep state (hash-only)" . 897.Pp 898Packet selection operates as follows: 899The queues are scanned from highest priority to lowest priority. 900If a queue has pending packets and is under its bandwidth minimum the 901scan stops and a packet is selected from that queue. 902If all queues have reached their bandwidth minimum a scale factor based 903on each queue's bandwidth minimum versus that queue's current bandwidth 904usage is calculated and the queue with the lowest scale factor is selected. 905This effectively uses the minimum bandwidth specification as a relative 906weighting for apportioning any remaining bandwidth on the link. 907.Pp 908The priority mechanic is only applicable in cases where the aggregate 909minimum bandwidth guarantees exceed the link bandwidth, and also has 910a small effect on queue selection when prioritizing between equal scale 911calculations. 912.Pp 913A 914.Ar fairq 915round robins between its 916.Ar buckets , 917extracting one packet from each bucket. 918This essentially prevents large backlogs of packets from high volume 919connections from destroying the interactive response of other connections. 920.Pp 921The 922.Ar bandwidth 923parameter for a 924.Ar fairq 925is guaranteed minimum and more will be used if no higher priority traffic is 926present. 927Creating a queue with one bucket as a catch-all for 928.Ar pass 929rules not characterized by 930.Ar keep state 931is supported. 932Such a queue serves as a basic priority queue with a bandwidth specification. 933.Pp 934Also note that when specifying rules it is always a good idea to specify 935a secondary queue for any tcp rules. 936The secondary queue is selected for pure ACKs without payloads and should 937generally be dedicated to that purpose with a minimum bandwidth specification 938sufficient to max-out the bandwidth for your incoming traffic. 939.El 940.Pp 941The interfaces on which queueing should be activated are declared using 942the 943.Ar altq on 944declaration. 945.Ar altq on 946has the following keywords: 947.Bl -tag -width xxxx 948.It Aq Ar interface 949Queueing is enabled on the named interface. 950.It Aq Ar scheduler 951Specifies which queueing scheduler to use. 952Currently supported values 953are 954.Ar cbq 955for Class Based Queueing, 956.Ar priq 957for Priority Queueing, 958.Ar hfsc 959for the Hierarchical Fair Service Curve scheduler, and 960.Ar fairq 961for the Fair Queueing. 962.It Ar bandwidth Aq Ar bw 963The maximum bitrate for all queues on an 964interface may be specified using the 965.Ar bandwidth 966keyword. 967The value can be specified as an absolute value or as a 968percentage of the interface bandwidth. 969When using an absolute value, the suffixes 970.Ar b , 971.Ar Kb , 972.Ar Mb , 973and 974.Ar Gb 975are used to represent bits, kilobits, megabits, and 976gigabits per second, respectively. 977The value must not exceed the interface bandwidth. 978If 979.Ar bandwidth 980is not specified, the interface bandwidth is used 981(but take note that some interfaces do not know their bandwidth, 982or can adapt their bandwidth rates). 983.Pp 984When used with 985.Ar fairq , 986.Ar bandwidth 987specifies a guaranteed minimum but the fairq is allowed to exceed it. 988.It Ar qlimit Aq Ar limit 989The maximum number of packets held in the queue. 990The default is 50. 991.It Ar tbrsize Aq Ar size 992Adjusts the size, in bytes, of the token bucket regulator. 993If not specified, heuristics based on the 994interface bandwidth are used to determine the size. 995.It Ar queue Aq Ar list 996Defines a list of subqueues to create on an interface. 997.El 998.Pp 999In the following example, the interface dc0 1000should queue up to 5 Mbit/s in four second-level queues using 1001Class Based Queueing. 1002Those four queues will be shown in a later example. 1003.Bd -literal -offset indent 1004altq on dc0 cbq bandwidth 5Mb queue { std, http, mail, ssh } 1005.Ed 1006.Pp 1007Once interfaces are activated for queueing using the 1008.Ar altq 1009directive, a sequence of 1010.Ar queue 1011directives may be defined. 1012The name associated with a 1013.Ar queue 1014must match a queue defined in the 1015.Ar altq 1016directive (e.g.\& mail), or, except for the 1017.Ar priq 1018and 1019.Ar fairq 1020.Ar schedulers , 1021in a parent 1022.Ar queue 1023declaration. 1024The following keywords can be used: 1025.Bl -tag -width xxxx 1026.It Ar on Aq Ar interface 1027Specifies the interface the queue operates on. 1028If not given, it operates on all matching interfaces. 1029.It Ar bandwidth Aq Ar bw 1030Specifies the maximum bitrate to be processed by the queue. 1031This value must not exceed the value of the parent 1032.Ar queue 1033and can be specified as an absolute value or a percentage of the parent 1034queue's bandwidth. 1035If not specified, defaults to 100% of the parent queue's bandwidth. 1036The 1037.Ar priq 1038scheduler does not support bandwidth specification. 1039The 1040.Ar fairq 1041scheduler uses the bandwidth specification as a guaranteed minimum and 1042may exceed it. 1043.It Ar priority Aq Ar level 1044Between queues a priority level can be set. 1045For 1046.Ar cbq , 1047.Ar hfsc , 1048and 1049.Ar fairq 1050the range is 0 to 7 and for 1051.Ar priq , 1052the range is 0 to 15. 1053The default for all is 1. 1054.Ar Priq 1055queues with a higher priority are always served first. 1056.Ar Fairq 1057queues with a higher priority are served first unless they exceed their 1058bandwidth specification. 1059.Ar Cbq 1060and 1061.Ar hfsc 1062queues with a higher priority are preferred in the case of overload. 1063.It Ar qlimit Aq Ar limit 1064The maximum number of packets held in the queue. 1065The default is 50. 1066When used with a 1067.Ar fairq 1068this specified the maximum number of packets held per bucket. 1069.El 1070.Pp 1071The 1072.Ar scheduler 1073can get additional parameters with 1074.Xo Aq Ar scheduler 1075.Pf ( Aq Ar parameters ) . 1076.Xc 1077Parameters are as follows: 1078.Bl -tag -width Fl 1079.It Ar default 1080Packets not matched by another queue are assigned to this one. 1081Exactly one default queue is required. 1082.It Ar red 1083Enable RED (Random Early Detection) on this queue. 1084RED drops packets with a probability proportional to the average 1085queue length. 1086.It Ar rio 1087Enables RIO on this queue. 1088RIO is RED with IN/OUT, thus running 1089RED two times more than RIO would achieve the same effect. 1090.It Ar ecn 1091Enables ECN (Explicit Congestion Notification) on this queue. 1092ECN implies RED. 1093.El 1094.Pp 1095The 1096.Ar fairq 1097.Ar scheduler 1098supports the following additional options: 1099.Bl -tag -width Fl 1100.It Ar buckets Aq Ar number 1101Specify the number of buckets, from 1 to 2048 in powers of 2. 1102A bucket size of 1 causes a 1103.Ar fairq 1104to essentially degenerate into a priority queue. 1105.It Ar linkshare Aq Ar sc 1106The bandwidth share of a backlogged queue. 1107This option is parsed but not yet supported. 1108.It Ar hogs Aq Ar bandwidth 1109This option allows low bandwidth connections to burst up to the specified 1110bandwidth by not advancing the round robin when taking packets out of 1111the related queue. 1112When using this option a small value no greater than 1/20 available interface 1113bandwidth is recommended. 1114.El 1115.Pp 1116The 1117.Ar cbq 1118.Ar scheduler 1119supports an additional option: 1120.Bl -tag -width Fl 1121.It Ar borrow 1122The queue can borrow bandwidth from the parent. 1123.El 1124.Pp 1125The 1126.Ar hfsc 1127.Ar scheduler 1128supports some additional options: 1129.Bl -tag -width Fl 1130.It Ar realtime Aq Ar sc 1131The minimum required bandwidth for the queue. 1132.It Ar upperlimit Aq Ar sc 1133The maximum allowed bandwidth for the queue. 1134.It Ar linkshare Aq Ar sc 1135The bandwidth share of a backlogged queue. 1136.El 1137.Pp 1138.Aq Ar sc 1139is an acronym for 1140.Ar service curve . 1141.Pp 1142The format for service curve specifications is 1143.Ar m2 1144or 1145.Ar "(m1 d m2)" . 1146.Ar m2 1147controls the bandwidth assigned to the queue. 1148.Ar m1 1149and 1150.Ar d 1151are optional and can be used to control the initial bandwidth assignment. 1152For the first 1153.Ar d 1154milliseconds the queue gets the bandwidth given as 1155.Ar m1 , 1156afterwards the value given in 1157.Ar m2 . 1158.Pp 1159Furthermore, with 1160.Ar cbq 1161and 1162.Ar hfsc , 1163child queues can be specified as in an 1164.Ar altq 1165declaration, thus building a tree of queues using a part of 1166their parent's bandwidth. 1167.Pp 1168Packets can be assigned to queues based on filter rules by using the 1169.Ar queue 1170keyword. 1171Normally only one 1172.Ar queue 1173is specified; when a second one is specified it will instead be used for 1174packets which have a 1175.Em TOS 1176of 1177.Em lowdelay 1178and for TCP ACKs with no data payload. 1179.Pp 1180To continue the previous example, the examples below would specify the 1181four referenced 1182queues, plus a few child queues. 1183Interactive 1184.Xr ssh 1 1185sessions get priority over bulk transfers like 1186.Xr scp 1 1187and 1188.Xr sftp 1 . 1189The queues may then be referenced by filtering rules (see 1190.Sx PACKET FILTERING 1191below). 1192.Bd -literal 1193queue std bandwidth 10% cbq(default) 1194queue http bandwidth 60% priority 2 cbq(borrow red) \e 1195 { employees, developers } 1196queue developers bandwidth 75% cbq(borrow) 1197queue employees bandwidth 15% 1198queue mail bandwidth 10% priority 0 cbq(borrow ecn) 1199queue ssh bandwidth 20% cbq(borrow) { ssh_interactive, ssh_bulk } 1200queue ssh_interactive bandwidth 50% priority 7 cbq(borrow) 1201queue ssh_bulk bandwidth 50% priority 0 cbq(borrow) 1202 1203block return out on dc0 inet all queue std 1204pass out on dc0 inet proto tcp from $developerhosts to any port 80 \e 1205 queue developers 1206pass out on dc0 inet proto tcp from $employeehosts to any port 80 \e 1207 queue employees 1208pass out on dc0 inet proto tcp from any to any port 22 \e 1209 queue(ssh_bulk, ssh_interactive) 1210pass out on dc0 inet proto tcp from any to any port 25 \e 1211 queue mail 1212.Ed 1213.Sh TRANSLATION 1214Translation rules modify either the source or destination address of the 1215packets associated with a stateful connection. 1216A stateful connection is automatically created to track packets matching 1217such a rule as long as they are not blocked by the filtering section of 1218.Nm . 1219The translation engine modifies the specified address and/or port in the 1220packet, recalculates IP, TCP and UDP checksums as necessary, and passes it to 1221the packet filter for evaluation. 1222.Pp 1223Since translation occurs before filtering the filter 1224engine will see packets as they look after any 1225addresses and ports have been translated. 1226Filter rules will therefore have to filter based on the translated 1227address and port number. 1228Packets that match a translation rule are only automatically passed if 1229the 1230.Ar pass 1231modifier is given, otherwise they are 1232still subject to 1233.Ar block 1234and 1235.Ar pass 1236rules. 1237.Pp 1238The state entry created permits 1239.Xr pf 4 1240to keep track of the original address for traffic associated with that state 1241and correctly direct return traffic for that connection. 1242.Pp 1243Various types of translation are possible with pf: 1244.Bl -tag -width xxxx 1245.It Ar binat 1246A 1247.Ar binat 1248rule specifies a bidirectional mapping between an external IP netblock 1249and an internal IP netblock. 1250.It Ar nat 1251A 1252.Ar nat 1253rule specifies that IP addresses are to be changed as the packet 1254traverses the given interface. 1255This technique allows one or more IP addresses 1256on the translating host to support network traffic for a larger range of 1257machines on an "inside" network. 1258Although in theory any IP address can be used on the inside, it is strongly 1259recommended that one of the address ranges defined by RFC 1918 be used. 1260These netblocks are: 1261.Bd -literal 126210.0.0.0 - 10.255.255.255 (all of net 10, i.e., 10/8) 1263172.16.0.0 - 172.31.255.255 (i.e., 172.16/12) 1264192.168.0.0 - 192.168.255.255 (i.e., 192.168/16) 1265.Ed 1266.It Pa rdr 1267The packet is redirected to another destination and possibly a 1268different port. 1269.Ar rdr 1270rules can optionally specify port ranges instead of single ports. 1271rdr ... port 2000:2999 -\*(Gt ... port 4000 1272redirects ports 2000 to 2999 (inclusive) to port 4000. 1273rdr ... port 2000:2999 -\*(Gt ... port 4000:* 1274redirects port 2000 to 4000, 2001 to 4001, ..., 2999 to 4999. 1275.El 1276.Pp 1277In addition to modifying the address, some translation rules may modify 1278source or destination ports for 1279.Xr tcp 4 1280or 1281.Xr udp 4 1282connections; implicitly in the case of 1283.Ar nat 1284rules and explicitly in the case of 1285.Ar rdr 1286rules. 1287Port numbers are never translated with a 1288.Ar binat 1289rule. 1290.Pp 1291Evaluation order of the translation rules is dependent on the type 1292of the translation rules and of the direction of a packet. 1293.Ar binat 1294rules are always evaluated first. 1295Then either the 1296.Ar rdr 1297rules are evaluated on an inbound packet or the 1298.Ar nat 1299rules on an outbound packet. 1300Rules of the same type are evaluated in the same order in which they 1301appear in the ruleset. 1302The first matching rule decides what action is taken. 1303.Pp 1304The 1305.Ar no 1306option prefixed to a translation rule causes packets to remain untranslated, 1307much in the same way as 1308.Ar drop quick 1309works in the packet filter (see below). 1310If no rule matches the packet it is passed to the filter engine unmodified. 1311.Pp 1312Translation rules apply only to packets that pass through 1313the specified interface, and if no interface is specified, 1314translation is applied to packets on all interfaces. 1315For instance, redirecting port 80 on an external interface to an internal 1316web server will only work for connections originating from the outside. 1317Connections to the address of the external interface from local hosts will 1318not be redirected, since such packets do not actually pass through the 1319external interface. 1320Redirections cannot reflect packets back through the interface they arrive 1321on, they can only be redirected to hosts connected to different interfaces 1322or to the firewall itself. 1323.Pp 1324Note that redirecting external incoming connections to the loopback 1325address, as in 1326.Bd -literal -offset indent 1327rdr on ne3 inet proto tcp to port smtp -\*(Gt 127.0.0.1 port spamd 1328.Ed 1329.Pp 1330will effectively allow an external host to connect to daemons 1331bound solely to the loopback address, circumventing the traditional 1332blocking of such connections on a real interface. 1333Unless this effect is desired, any of the local non-loopback addresses 1334should be used as redirection target instead, which allows external 1335connections only to daemons bound to this address or not bound to 1336any address. 1337.Pp 1338See 1339.Sx TRANSLATION EXAMPLES 1340below. 1341.Sh PACKET FILTERING 1342.Xr pf 4 1343has the ability to 1344.Ar block 1345and 1346.Ar pass 1347packets based on attributes of their layer 3 (see 1348.Xr ip 4 1349and 1350.Xr ip6 4 ) 1351and layer 4 (see 1352.Xr icmp 4 , 1353.Xr icmp6 4 , 1354.Xr tcp 4 , 1355.Xr udp 4 ) 1356headers. 1357In addition, packets may also be 1358assigned to queues for the purpose of bandwidth control. 1359.Pp 1360For each packet processed by the packet filter, the filter rules are 1361evaluated in sequential order, from first to last. 1362The last matching rule decides what action is taken. 1363If no rule matches the packet, the default action is to pass 1364the packet. 1365.Pp 1366The following actions can be used in the filter: 1367.Bl -tag -width xxxx 1368.It Ar block 1369The packet is blocked. 1370There are a number of ways in which a 1371.Ar block 1372rule can behave when blocking a packet. 1373The default behaviour is to 1374.Ar drop 1375packets silently, however this can be overridden or made 1376explicit either globally, by setting the 1377.Ar block-policy 1378option, or on a per-rule basis with one of the following options: 1379.Pp 1380.Bl -tag -width xxxx -compact 1381.It Ar drop 1382The packet is silently dropped. 1383.It Ar return-rst 1384This applies only to 1385.Xr tcp 4 1386packets, and issues a TCP RST which closes the 1387connection. 1388.It Ar return-icmp 1389.It Ar return-icmp6 1390This causes ICMP messages to be returned for packets which match the rule. 1391By default this is an ICMP UNREACHABLE message, however this 1392can be overridden by specifying a message as a code or number. 1393.It Ar return 1394This causes a TCP RST to be returned for 1395.Xr tcp 4 1396packets and an ICMP UNREACHABLE for UDP and other packets. 1397.El 1398.Pp 1399Options returning ICMP packets currently have no effect if 1400.Xr pf 4 1401operates on a 1402.Xr bridge 4 , 1403as the code to support this feature has not yet been implemented. 1404.Pp 1405The simplest mechanism to block everything by default and only pass 1406packets that match explicit rules is specify a first filter rule of: 1407.Bd -literal -offset indent 1408block all 1409.Ed 1410.It Ar pass 1411The packet is passed; 1412state is created unless the 1413.Ar no state 1414option is specified. 1415.El 1416.Pp 1417By default 1418.Xr pf 4 1419filters packets statefully; the first time a packet matches a 1420.Ar pass 1421rule, a state entry is created; for subsequent packets the filter checks 1422whether the packet matches any state. 1423If it does, the packet is passed without evaluation of any rules. 1424After the connection is closed or times out, the state entry is automatically 1425removed. 1426.Pp 1427This has several advantages. 1428For TCP connections, comparing a packet to a state involves checking 1429its sequence numbers, as well as TCP timestamps if a 1430.Ar scrub reassemble tcp 1431rule applies to the connection. 1432If these values are outside the narrow windows of expected 1433values, the packet is dropped. 1434This prevents spoofing attacks, such as when an attacker sends packets with 1435a fake source address/port but does not know the connection's sequence 1436numbers. 1437Similarly, 1438.Xr pf 4 1439knows how to match ICMP replies to states. 1440For example, 1441.Bd -literal -offset indent 1442pass out inet proto icmp all icmp-type echoreq 1443.Ed 1444.Pp 1445allows echo requests (such as those created by 1446.Xr ping 8 ) 1447out statefully, and matches incoming echo replies correctly to states. 1448.Pp 1449Also, looking up states is usually faster than evaluating rules. 1450If there are 50 rules, all of them are evaluated sequentially in O(n). 1451Even with 50000 states, only 16 comparisons are needed to match a 1452state, since states are stored in a binary search tree that allows 1453searches in O(log2 n). 1454.Pp 1455Furthermore, correct handling of ICMP error messages is critical to 1456many protocols, particularly TCP. 1457.Xr pf 4 1458matches ICMP error messages to the correct connection, checks them against 1459connection parameters, and passes them if appropriate. 1460For example if an ICMP source quench message referring to a stateful TCP 1461connection arrives, it will be matched to the state and get passed. 1462.Pp 1463Finally, state tracking is required for 1464.Ar nat , binat No and Ar rdr 1465rules, in order to track address and port translations and reverse the 1466translation on returning packets. 1467.Pp 1468.Xr pf 4 1469will also create state for other protocols which are effectively stateless by 1470nature. 1471UDP packets are matched to states using only host addresses and ports, 1472and other protocols are matched to states using only the host addresses. 1473.Pp 1474If stateless filtering of individual packets is desired, 1475the 1476.Ar no state 1477keyword can be used to specify that state will not be created 1478if this is the last matching rule. 1479A number of parameters can also be set to affect how 1480.Xr pf 4 1481handles state tracking. 1482See 1483.Sx STATEFUL TRACKING OPTIONS 1484below for further details. 1485.Sh PARAMETERS 1486The rule parameters specify the packets to which a rule applies. 1487A packet always comes in on, or goes out through, one interface. 1488Most parameters are optional. 1489If a parameter is specified, the rule only applies to packets with 1490matching attributes. 1491Certain parameters can be expressed as lists, in which case 1492.Xr pfctl 8 1493generates all needed rule combinations. 1494.Bl -tag -width xxxx 1495.It Ar in No or Ar out 1496This rule applies to incoming or outgoing packets. 1497If neither 1498.Ar in 1499nor 1500.Ar out 1501are specified, the rule will match packets in both directions. 1502.It Ar log 1503In addition to the action specified, a log message is generated. 1504Only the packet that establishes the state is logged, 1505unless the 1506.Ar no state 1507option is specified. 1508The logged packets are sent to a 1509.Xr pflog 4 1510interface, by default 1511.Ar pflog0 . 1512This interface is monitored by the 1513.Xr pflogd 8 1514logging daemon, which dumps the logged packets to the file 1515.Pa /var/log/pflog 1516in 1517.Xr pcap 3 1518binary format. 1519.It Ar log (all) 1520Used to force logging of all packets for a connection. 1521This is not necessary when 1522.Ar no state 1523is explicitly specified. 1524As with 1525.Ar log , 1526packets are logged to 1527.Xr pflog 4 . 1528.It Ar log (user) 1529Logs the 1530.Ux 1531user ID of the user that owns the socket and the PID of the process that 1532has the socket open where the packet is sourced from or destined to 1533(depending on which socket is local). 1534This is in addition to the normal information logged. 1535.It Ar log (to Aq Ar interface ) 1536Send logs to the specified 1537.Xr pflog 4 1538interface instead of 1539.Ar pflog0 . 1540.It Ar quick 1541If a packet matches a rule which has the 1542.Ar quick 1543option set, this rule 1544is considered the last matching rule, and evaluation of subsequent rules 1545is skipped. 1546.It Ar on Aq Ar interface 1547This rule applies only to packets coming in on, or going out through, this 1548particular interface or interface group. 1549For more information on interface groups, 1550see the 1551.Ic group 1552keyword in 1553.Xr ifconfig 8 . 1554.It Aq Ar af 1555This rule applies only to packets of this address family. 1556Supported values are 1557.Ar inet 1558and 1559.Ar inet6 . 1560.It Ar proto Aq Ar protocol 1561This rule applies only to packets of this protocol. 1562Common protocols are 1563.Xr icmp 4 , 1564.Xr icmp6 4 , 1565.Xr tcp 4 , 1566and 1567.Xr udp 4 . 1568For a list of all the protocol name to number mappings used by 1569.Xr pfctl 8 , 1570see the file 1571.Pa /etc/protocols . 1572.It Xo 1573.Ar from Aq Ar source 1574.Ar port Aq Ar source 1575.Ar os Aq Ar source 1576.Ar to Aq Ar dest 1577.Ar port Aq Ar dest 1578.Xc 1579This rule applies only to packets with the specified source and destination 1580addresses and ports. 1581.Pp 1582Addresses can be specified in CIDR notation (matching netblocks), as 1583symbolic host names, interface names or interface group names, or as any 1584of the following keywords: 1585.Pp 1586.Bl -tag -width xxxxxxxxxxxxxx -compact 1587.It Ar any 1588Any address. 1589.It Ar route Aq Ar label 1590Any address whose associated route has label 1591.Aq Ar label . 1592See 1593.Xr route 4 1594and 1595.Xr route 8 . 1596.It Ar no-route 1597Any address which is not currently routable. 1598.It Ar urpf-failed 1599Any source address that fails a unicast reverse path forwarding (URPF) 1600check, i.e. packets coming in on an interface other than that which holds 1601the route back to the packet's source address. 1602.It Aq Ar table 1603Any address that matches the given table. 1604.El 1605.Pp 1606Ranges of addresses are specified by using the 1607.Sq - 1608operator. 1609For instance: 1610.Dq 10.1.1.10 - 10.1.1.12 1611means all addresses from 10.1.1.10 to 10.1.1.12, 1612hence addresses 10.1.1.10, 10.1.1.11, and 10.1.1.12. 1613.Pp 1614Interface names and interface group names can have modifiers appended: 1615.Pp 1616.Bl -tag -width xxxxxxxxxxxx -compact 1617.It Ar :network 1618Translates to the network(s) attached to the interface. 1619.It Ar :broadcast 1620Translates to the interface's broadcast address(es). 1621.It Ar :peer 1622Translates to the point-to-point interface's peer address(es). 1623.It Ar :0 1624Do not include interface aliases. 1625.El 1626.Pp 1627Host names may also have the 1628.Ar :0 1629option appended to restrict the name resolution to the first of each 1630v4 and v6 address found. 1631.Pp 1632Host name resolution and interface to address translation are done at 1633ruleset load-time. 1634When the address of an interface (or host name) changes (under DHCP or PPP, 1635for instance), the ruleset must be reloaded for the change to be reflected 1636in the kernel. 1637Surrounding the interface name (and optional modifiers) in parentheses 1638changes this behaviour. 1639When the interface name is surrounded by parentheses, the rule is 1640automatically updated whenever the interface changes its address. 1641The ruleset does not need to be reloaded. 1642This is especially useful with 1643.Ar nat . 1644.Pp 1645Ports can be specified either by number or by name. 1646For example, port 80 can be specified as 1647.Em www . 1648For a list of all port name to number mappings used by 1649.Xr pfctl 8 , 1650see the file 1651.Pa /etc/services . 1652.Pp 1653Ports and ranges of ports are specified by using these operators: 1654.Bd -literal -offset indent 1655= (equal) 1656!= (unequal) 1657\*(Lt (less than) 1658\*(Le (less than or equal) 1659\*(Gt (greater than) 1660\*(Ge (greater than or equal) 1661: (range including boundaries) 1662\*(Gt\*(Lt (range excluding boundaries) 1663\*(Lt\*(Gt (except range) 1664.Ed 1665.Pp 1666.Sq \*(Gt\*(Lt , 1667.Sq \*(Lt\*(Gt 1668and 1669.Sq \&: 1670are binary operators (they take two arguments). 1671For instance: 1672.Bl -tag -width Fl 1673.It Ar port 2000:2004 1674means 1675.Sq all ports \*(Ge 2000 and \*(Le 2004 , 1676hence ports 2000, 2001, 2002, 2003 and 2004. 1677.It Ar port 2000 \*(Gt\*(Lt 2004 1678means 1679.Sq all ports \*(Gt 2000 and \*(Lt 2004 , 1680hence ports 2001, 2002 and 2003. 1681.It Ar port 2000 \*(Lt\*(Gt 2004 1682means 1683.Sq all ports \*(Lt 2000 or \*(Gt 2004 , 1684hence ports 1-1999 and 2005-65535. 1685.El 1686.Pp 1687The operating system of the source host can be specified in the case of TCP 1688rules with the 1689.Ar OS 1690modifier. 1691See the 1692.Sx OPERATING SYSTEM FINGERPRINTING 1693section for more information. 1694.Pp 1695The host, port and OS specifications are optional, as in the following examples: 1696.Bd -literal -offset indent 1697pass in all 1698pass in from any to any 1699pass in proto tcp from any port \*(Le 1024 to any 1700pass in proto tcp from any to any port 25 1701pass in proto tcp from 10.0.0.0/8 port \*(Gt 1024 \e 1702 to ! 10.1.2.3 port != ssh 1703pass in proto tcp from any os "OpenBSD" 1704pass in proto tcp from route "DTAG" 1705.Ed 1706.It Ar all 1707This is equivalent to "from any to any". 1708.It Ar group Aq Ar group 1709Similar to 1710.Ar user , 1711this rule only applies to packets of sockets owned by the specified group. 1712.It Ar user Aq Ar user 1713This rule only applies to packets of sockets owned by the specified user. 1714For outgoing connections initiated from the firewall, this is the user 1715that opened the connection. 1716For incoming connections to the firewall itself, this is the user that 1717listens on the destination port. 1718For forwarded connections, where the firewall is not a connection endpoint, 1719the user and group are 1720.Em unknown . 1721.Pp 1722All packets, both outgoing and incoming, of one connection are associated 1723with the same user and group. 1724Only TCP and UDP packets can be associated with users; for other protocols 1725these parameters are ignored. 1726.Pp 1727User and group refer to the effective (as opposed to the real) IDs, in 1728case the socket is created by a setuid/setgid process. 1729User and group IDs are stored when a socket is created; 1730when a process creates a listening socket as root (for instance, by 1731binding to a privileged port) and subsequently changes to another 1732user ID (to drop privileges), the credentials will remain root. 1733.Pp 1734User and group IDs can be specified as either numbers or names. 1735The syntax is similar to the one for ports. 1736The value 1737.Em unknown 1738matches packets of forwarded connections. 1739.Em unknown 1740can only be used with the operators 1741.Cm = 1742and 1743.Cm != . 1744Other constructs like 1745.Cm user \*(Ge unknown 1746are invalid. 1747Forwarded packets with unknown user and group ID match only rules 1748that explicitly compare against 1749.Em unknown 1750with the operators 1751.Cm = 1752or 1753.Cm != . 1754For instance 1755.Cm user \*(Ge 0 1756does not match forwarded packets. 1757The following example allows only selected users to open outgoing 1758connections: 1759.Bd -literal -offset indent 1760block out proto { tcp, udp } all 1761pass out proto { tcp, udp } all user { \*(Lt 1000, dhartmei } 1762.Ed 1763.It Xo Ar flags Aq Ar a 1764.Pf / Ns Aq Ar b 1765.No \*(Ba / Ns Aq Ar b 1766.No \*(Ba any 1767.Xc 1768This rule only applies to TCP packets that have the flags 1769.Aq Ar a 1770set out of set 1771.Aq Ar b . 1772Flags not specified in 1773.Aq Ar b 1774are ignored. 1775For stateful connections, the default is 1776.Ar flags S/SA . 1777To indicate that flags should not be checked at all, specify 1778.Ar flags any . 1779The flags are: (F)IN, (S)YN, (R)ST, (P)USH, (A)CK, (U)RG, (E)CE, and C(W)R. 1780.Bl -tag -width Fl 1781.It Ar flags S/S 1782Flag SYN is set. 1783The other flags are ignored. 1784.It Ar flags S/SA 1785This is the default setting for stateful connections. 1786Out of SYN and ACK, exactly SYN may be set. 1787SYN, SYN+PSH and SYN+RST match, but SYN+ACK, ACK and ACK+RST do not. 1788This is more restrictive than the previous example. 1789.It Ar flags /SFRA 1790If the first set is not specified, it defaults to none. 1791All of SYN, FIN, RST and ACK must be unset. 1792.El 1793.Pp 1794Because 1795.Ar flags S/SA 1796is applied by default (unless 1797.Ar no state 1798is specified), only the initial SYN packet of a TCP handshake will create 1799a state for a TCP connection. 1800It is possible to be less restrictive, and allow state creation from 1801intermediate 1802.Pq non-SYN 1803packets, by specifying 1804.Ar flags any . 1805This will cause 1806.Xr pf 4 1807to synchronize to existing connections, for instance 1808if one flushes the state table. 1809However, states created from such intermediate packets may be missing 1810connection details such as the TCP window scaling factor. 1811States which modify the packet flow, such as those affected by 1812.Ar nat , binat No or Ar rdr 1813rules, 1814.Ar modulate No or Ar synproxy state 1815options, or scrubbed with 1816.Ar reassemble tcp 1817will also not be recoverable from intermediate packets. 1818Such connections will stall and time out. 1819.It Xo Ar icmp-type Aq Ar type 1820.Ar code Aq Ar code 1821.Xc 1822.It Xo Ar icmp6-type Aq Ar type 1823.Ar code Aq Ar code 1824.Xc 1825This rule only applies to ICMP or ICMPv6 packets with the specified type 1826and code. 1827Text names for ICMP types and codes are listed in 1828.Xr icmp 4 1829and 1830.Xr icmp6 4 . 1831This parameter is only valid for rules that cover protocols ICMP or 1832ICMP6. 1833The protocol and the ICMP type indicator 1834.Po 1835.Ar icmp-type 1836or 1837.Ar icmp6-type 1838.Pc 1839must match. 1840.It Xo Ar tos Aq Ar string 1841.No \*(Ba Aq Ar number 1842.Xc 1843This rule applies to packets with the specified 1844.Em TOS 1845bits set. 1846.Em TOS 1847may be 1848given as one of 1849.Ar lowdelay , 1850.Ar throughput , 1851.Ar reliability , 1852or as either hex or decimal. 1853.Pp 1854For example, the following rules are identical: 1855.Bd -literal -offset indent 1856pass all tos lowdelay 1857pass all tos 0x10 1858pass all tos 16 1859.Ed 1860.It Ar allow-opts 1861By default, IPv4 packets with IP options or IPv6 packets with routing 1862extension headers are blocked. 1863When 1864.Ar allow-opts 1865is specified for a 1866.Ar pass 1867rule, packets that pass the filter based on that rule (last matching) 1868do so even if they contain IP options or routing extension headers. 1869For packets that match state, the rule that initially created the 1870state is used. 1871The implicit 1872.Ar pass 1873rule that is used when a packet does not match any rules does not 1874allow IP options. 1875.It Ar label Aq Ar string 1876Adds a label (name) to the rule, which can be used to identify the rule. 1877For instance, 1878pfctl -s labels 1879shows per-rule statistics for rules that have labels. 1880.Pp 1881The following macros can be used in labels: 1882.Pp 1883.Bl -tag -width $srcaddr -compact -offset indent 1884.It Ar $if 1885The interface. 1886.It Ar $srcaddr 1887The source IP address. 1888.It Ar $dstaddr 1889The destination IP address. 1890.It Ar $srcport 1891The source port specification. 1892.It Ar $dstport 1893The destination port specification. 1894.It Ar $proto 1895The protocol name. 1896.It Ar $nr 1897The rule number. 1898.El 1899.Pp 1900For example: 1901.Bd -literal -offset indent 1902ips = \&"{ 1.2.3.4, 1.2.3.5 }\&" 1903pass in proto tcp from any to $ips \e 1904 port \*(Gt 1023 label \&"$dstaddr:$dstport\&" 1905.Ed 1906.Pp 1907expands to 1908.Bd -literal -offset indent 1909pass in inet proto tcp from any to 1.2.3.4 \e 1910 port \*(Gt 1023 label \&"1.2.3.4:\*(Gt1023\&" 1911pass in inet proto tcp from any to 1.2.3.5 \e 1912 port \*(Gt 1023 label \&"1.2.3.5:\*(Gt1023\&" 1913.Ed 1914.Pp 1915The macro expansion for the 1916.Ar label 1917directive occurs only at configuration file parse time, not during runtime. 1918.It Xo Ar queue Aq Ar queue 1919.No \*(Ba ( Aq Ar queue , 1920.Aq Ar queue ) 1921.Xc 1922Packets matching this rule will be assigned to the specified queue. 1923If two queues are given, packets which have a 1924.Em TOS 1925of 1926.Em lowdelay 1927and TCP ACKs with no data payload will be assigned to the second one. 1928See 1929.Sx QUEUEING 1930for setup details. 1931.Pp 1932For example: 1933.Bd -literal -offset indent 1934pass in proto tcp to port 25 queue mail 1935pass in proto tcp to port 22 queue(ssh_bulk, ssh_prio) 1936.Ed 1937.It Ar tag Aq Ar string 1938Packets matching this rule will be tagged with the 1939specified string. 1940The tag acts as an internal marker that can be used to 1941identify these packets later on. 1942This can be used, for example, to provide trust between 1943interfaces and to determine if packets have been 1944processed by translation rules. 1945Tags are 1946.Qq sticky , 1947meaning that the packet will be tagged even if the rule 1948is not the last matching rule. 1949Further matching rules can replace the tag with a 1950new one but will not remove a previously applied tag. 1951A packet is only ever assigned one tag at a time. 1952Packet tagging can be done during 1953.Ar nat , 1954.Ar rdr , 1955or 1956.Ar binat 1957rules in addition to filter rules. 1958Tags take the same macros as labels (see above). 1959.It Ar tagged Aq Ar string 1960Used with filter, translation or scrub rules 1961to specify that packets must already 1962be tagged with the given tag in order to match the rule. 1963Inverse tag matching can also be done 1964by specifying the 1965.Cm !\& 1966operator before the 1967.Ar tagged 1968keyword. 1969.It Ar rtable Aq Ar number 1970Used to select an alternate routing table for the routing lookup. 1971Only effective before the route lookup happened, i.e. when filtering inbound. 1972.It Xo Ar divert-to Aq Ar host 1973.Ar port Aq Ar port 1974.Xc 1975Used to redirect packets to a local socket bound to 1976.Ar host 1977and 1978.Ar port . 1979The packets will not be modified, so 1980.Xr getsockname 2 1981on the socket will return the original destination address of the packet. 1982.It Ar divert-reply 1983Used to receive replies for sockets that are bound to addresses 1984which are not local to the machine. 1985See 1986.Xr setsockopt 2 1987for information on how to bind these sockets. 1988.It Ar probability Aq Ar number 1989A probability attribute can be attached to a rule, with a value set between 19900 and 1, bounds not included. 1991In that case, the rule will be honoured using the given probability value 1992only. 1993For example, the following rule will drop 20% of incoming ICMP packets: 1994.Bd -literal -offset indent 1995block in proto icmp probability 20% 1996.Ed 1997.El 1998.Sh ROUTING 1999If a packet matches a rule with a route option set, the packet filter will 2000route the packet according to the type of route option. 2001When such a rule creates state, the route option is also applied to all 2002packets matching the same connection. 2003.Bl -tag -width xxxx 2004.It Ar fastroute 2005The 2006.Ar fastroute 2007option does a normal route lookup to find the next hop for the packet. 2008.It Ar route-to 2009The 2010.Ar route-to 2011option routes the packet to the specified interface with an optional address 2012for the next hop. 2013When a 2014.Ar route-to 2015rule creates state, only packets that pass in the same direction as the 2016filter rule specifies will be routed in this way. 2017Packets passing in the opposite direction (replies) are not affected 2018and are routed normally. 2019.It Ar reply-to 2020The 2021.Ar reply-to 2022option is similar to 2023.Ar route-to , 2024but routes packets that pass in the opposite direction (replies) to the 2025specified interface. 2026Opposite direction is only defined in the context of a state entry, and 2027.Ar reply-to 2028is useful only in rules that create state. 2029It can be used on systems with multiple external connections to 2030route all outgoing packets of a connection through the interface 2031the incoming connection arrived through (symmetric routing enforcement). 2032.It Ar dup-to 2033The 2034.Ar dup-to 2035option creates a duplicate of the packet and routes it like 2036.Ar route-to . 2037The original packet gets routed as it normally would. 2038.El 2039.Sh POOL OPTIONS 2040For 2041.Ar nat 2042and 2043.Ar rdr 2044rules, (as well as for the 2045.Ar route-to , 2046.Ar reply-to 2047and 2048.Ar dup-to 2049rule options) for which there is a single redirection address which has a 2050subnet mask smaller than 32 for IPv4 or 128 for IPv6 (more than one IP 2051address), a variety of different methods for assigning this address can be 2052used: 2053.Bl -tag -width xxxx 2054.It Ar bitmask 2055The 2056.Ar bitmask 2057option applies the network portion of the redirection address to the address 2058to be modified (source with 2059.Ar nat , 2060destination with 2061.Ar rdr ) . 2062.It Ar random 2063The 2064.Ar random 2065option selects an address at random within the defined block of addresses. 2066.It Ar source-hash 2067The 2068.Ar source-hash 2069option uses a hash of the source address to determine the redirection address, 2070ensuring that the redirection address is always the same for a given source. 2071An optional key can be specified after this keyword either in hex or as a 2072string; by default 2073.Xr pfctl 8 2074randomly generates a key for source-hash every time the 2075ruleset is reloaded. 2076.It Ar round-robin 2077The 2078.Ar round-robin 2079option loops through the redirection address(es). 2080.Pp 2081When more than one redirection address is specified, 2082.Ar round-robin 2083is the only permitted pool type. 2084.It Ar static-port 2085With 2086.Ar nat 2087rules, the 2088.Ar static-port 2089option prevents 2090.Xr pf 4 2091from modifying the source port on TCP and UDP packets. 2092.El 2093.Pp 2094Additionally, the 2095.Ar sticky-address 2096option can be specified to help ensure that multiple connections from the 2097same source are mapped to the same redirection address. 2098This option can be used with the 2099.Ar random 2100and 2101.Ar round-robin 2102pool options. 2103Note that by default these associations are destroyed as soon as there are 2104no longer states which refer to them; in order to make the mappings last 2105beyond the lifetime of the states, increase the global options with 2106.Ar set timeout src.track . 2107See 2108.Sx STATEFUL TRACKING OPTIONS 2109for more ways to control the source tracking. 2110.Sh STATE MODULATION 2111Much of the security derived from TCP is attributable to how well the 2112initial sequence numbers (ISNs) are chosen. 2113Some popular stack implementations choose 2114.Em very 2115poor ISNs and thus are normally susceptible to ISN prediction exploits. 2116By applying a 2117.Ar modulate state 2118rule to a TCP connection, 2119.Xr pf 4 2120will create a high quality random sequence number for each connection 2121endpoint. 2122.Pp 2123The 2124.Ar modulate state 2125directive implicitly keeps state on the rule and is 2126only applicable to TCP connections. 2127.Pp 2128For instance: 2129.Bd -literal -offset indent 2130block all 2131pass out proto tcp from any to any modulate state 2132pass in proto tcp from any to any port 25 flags S/SFRA modulate state 2133.Ed 2134.Pp 2135Note that modulated connections will not recover when the state table 2136is lost (firewall reboot, flushing the state table, etc...). 2137.Xr pf 4 2138will not be able to infer a connection again after the state table flushes 2139the connection's modulator. 2140When the state is lost, the connection may be left dangling until the 2141respective endpoints time out the connection. 2142It is possible on a fast local network for the endpoints to start an ACK 2143storm while trying to resynchronize after the loss of the modulator. 2144The default 2145.Ar flags 2146settings (or a more strict equivalent) should be used on 2147.Ar modulate state 2148rules to prevent ACK storms. 2149.Pp 2150Note that alternative methods are available 2151to prevent loss of the state table 2152and allow for firewall failover. 2153See 2154.Xr carp 4 2155and 2156.Xr pfsync 4 2157for further information. 2158.Sh SYN PROXY 2159By default, 2160.Xr pf 4 2161passes packets that are part of a 2162.Xr tcp 4 2163handshake between the endpoints. 2164The 2165.Ar synproxy state 2166option can be used to cause 2167.Xr pf 4 2168itself to complete the handshake with the active endpoint, perform a handshake 2169with the passive endpoint, and then forward packets between the endpoints. 2170.Pp 2171No packets are sent to the passive endpoint before the active endpoint has 2172completed the handshake, hence so-called SYN floods with spoofed source 2173addresses will not reach the passive endpoint, as the sender can't complete the 2174handshake. 2175.Pp 2176The proxy is transparent to both endpoints, they each see a single 2177connection from/to the other endpoint. 2178.Xr pf 4 2179chooses random initial sequence numbers for both handshakes. 2180Once the handshakes are completed, the sequence number modulators 2181(see previous section) are used to translate further packets of the 2182connection. 2183.Ar synproxy state 2184includes 2185.Ar modulate state . 2186.Pp 2187Rules with 2188.Ar synproxy 2189will not work if 2190.Xr pf 4 2191operates on a 2192.Xr bridge 4 . 2193.Pp 2194Example: 2195.Bd -literal -offset indent 2196pass in proto tcp from any to any port www synproxy state 2197.Ed 2198.Sh STATEFUL TRACKING OPTIONS 2199A number of options related to stateful tracking can be applied on a 2200per-rule basis. 2201.Ar keep state , 2202.Ar modulate state 2203and 2204.Ar synproxy state 2205support these options, and 2206.Ar keep state 2207must be specified explicitly to apply options to a rule. 2208.Pp 2209.Bl -tag -width xxxx -compact 2210.It Ar max Aq Ar number 2211Limits the number of concurrent states the rule may create. 2212When this limit is reached, further packets that would create 2213state will not match this rule until existing states time out. 2214.It Ar no-sync 2215Prevent state changes for states created by this rule from appearing on the 2216.Xr pfsync 4 2217interface. 2218.It Xo Aq Ar timeout 2219.Aq Ar seconds 2220.Xc 2221Changes the timeout values used for states created by this rule. 2222For a list of all valid timeout names, see 2223.Sx OPTIONS 2224above. 2225.It Ar sloppy 2226Uses a sloppy TCP connection tracker that does not check sequence 2227numbers at all, which makes insertion and ICMP teardown attacks way 2228easier. 2229This is intended to be used in situations where one does not see all 2230packets of a connection, e.g. in asymmetric routing situations. 2231Cannot be used with modulate or synproxy state. 2232.El 2233.Pp 2234Multiple options can be specified, separated by commas: 2235.Bd -literal -offset indent 2236pass in proto tcp from any to any \e 2237 port www keep state \e 2238 (max 100, source-track rule, max-src-nodes 75, \e 2239 max-src-states 3, tcp.established 60, tcp.closing 5) 2240.Ed 2241.Pp 2242When the 2243.Ar source-track 2244keyword is specified, the number of states per source IP is tracked. 2245.Pp 2246.Bl -tag -width xxxx -compact 2247.It Ar source-track rule 2248The maximum number of states created by this rule is limited by the rule's 2249.Ar max-src-nodes 2250and 2251.Ar max-src-states 2252options. 2253Only state entries created by this particular rule count toward the rule's 2254limits. 2255.It Ar source-track global 2256The number of states created by all rules that use this option is limited. 2257Each rule can specify different 2258.Ar max-src-nodes 2259and 2260.Ar max-src-states 2261options, however state entries created by any participating rule count towards 2262each individual rule's limits. 2263.El 2264.Pp 2265The following limits can be set: 2266.Pp 2267.Bl -tag -width xxxx -compact 2268.It Ar max-src-nodes Aq Ar number 2269Limits the maximum number of source addresses which can simultaneously 2270have state table entries. 2271.It Ar max-src-states Aq Ar number 2272Limits the maximum number of simultaneous state entries that a single 2273source address can create with this rule. 2274.It Ar pickups 2275Specify that mid-stream pickups are to be allowed. 2276The default is to NOT allow mid-stream pickups and implies flags 2277S/SA for TCP connections. 2278If pickups are enabled, flags S/SA are not implied 2279for TCP connections and state can be created for any packet. 2280.Pp 2281The implied flags parameters need not be specified in either case 2282unless you explicitly wish to override them, which also allows 2283you to roll-up several protocols into a single rule. 2284.Pp 2285Certain validations are disabled when mid-stream pickups occur. 2286For example, the window scaling options are not known for 2287TCP pickups and sequence space comparisons must be disabled. 2288.Pp 2289This does not effect state representing fully quantified 2290connections (for which the SYN/SYN-ACK passed through the routing 2291engine). 2292Those connections continue to be fully validated. 2293.It Ar hash-only 2294Specify that mid-stream pickups are to be allowed, but unconditionally 2295disables sequence space checks even if full state is available. 2296.It Ar no-pickups 2297Specify that mid-stream pickups are not to be allowed. 2298This is the 2299default and this keyword does not normally need to be specified. 2300However, if you are concerned about rule set portability then 2301specifying this keyword will at least result in an error from 2302.Xr pfctl 8 2303if it doesn't understand the feature. 2304TCP flags of S/SA are implied 2305and do not need to explicitly specified. 2306.El 2307.Pp 2308For stateful TCP connections, limits on established connections (connections 2309which have completed the TCP 3-way handshake) can also be enforced 2310per source IP. 2311.Pp 2312.Bl -tag -width xxxx -compact 2313.It Ar max-src-conn Aq Ar number 2314Limits the maximum number of simultaneous TCP connections which have 2315completed the 3-way handshake that a single host can make. 2316.It Xo Ar max-src-conn-rate Aq Ar number 2317.No / Aq Ar seconds 2318.Xc 2319Limit the rate of new connections over a time interval. 2320The connection rate is an approximation calculated as a moving average. 2321.El 2322.Pp 2323Because the 3-way handshake ensures that the source address is not being 2324spoofed, more aggressive action can be taken based on these limits. 2325With the 2326.Ar overload Aq Ar table 2327state option, source IP addresses which hit either of the limits on 2328established connections will be added to the named table. 2329This table can be used in the ruleset to block further activity from 2330the offending host, redirect it to a tarpit process, or restrict its 2331bandwidth. 2332.Pp 2333The optional 2334.Ar flush 2335keyword kills all states created by the matching rule which originate 2336from the host which exceeds these limits. 2337The 2338.Ar global 2339modifier to the flush command kills all states originating from the 2340offending host, regardless of which rule created the state. 2341.Pp 2342For example, the following rules will protect the webserver against 2343hosts making more than 100 connections in 10 seconds. 2344Any host which connects faster than this rate will have its address added 2345to the 2346.Aq bad_hosts 2347table and have all states originating from it flushed. 2348Any new packets arriving from this host will be dropped unconditionally 2349by the block rule. 2350.Bd -literal -offset indent 2351block quick from \*(Ltbad_hosts\*(Gt 2352pass in on $ext_if proto tcp to $webserver port www keep state \e 2353 (max-src-conn-rate 100/10, overload \*(Ltbad_hosts\*(Gt flush global) 2354.Ed 2355.Sh OPERATING SYSTEM FINGERPRINTING 2356Passive OS Fingerprinting is a mechanism to inspect nuances of a TCP 2357connection's initial SYN packet and guess at the host's operating system. 2358Unfortunately these nuances are easily spoofed by an attacker so the 2359fingerprint is not useful in making security decisions. 2360But the fingerprint is typically accurate enough to make policy decisions 2361upon. 2362.Pp 2363The fingerprints may be specified by operating system class, by 2364version, or by subtype/patchlevel. 2365The class of an operating system is typically the vendor or genre 2366and would be 2367.Ox 2368for the 2369.Xr pf 4 2370firewall itself. 2371The version of the oldest available 2372.Ox 2373release on the main FTP site 2374would be 2.6 and the fingerprint would be written 2375.Pp 2376.Dl \&"OpenBSD 2.6\&" 2377.Pp 2378The subtype of an operating system is typically used to describe the 2379patchlevel if that patch led to changes in the TCP stack behavior. 2380In the case of 2381.Ox , 2382the only subtype is for a fingerprint that was 2383normalized by the 2384.Ar no-df 2385scrub option and would be specified as 2386.Pp 2387.Dl \&"OpenBSD 3.3 no-df\&" 2388.Pp 2389Fingerprints for most popular operating systems are provided by 2390.Xr pf.os 5 . 2391Once 2392.Xr pf 4 2393is running, a complete list of known operating system fingerprints may 2394be listed by running: 2395.Pp 2396.Dl # pfctl -so 2397.Pp 2398Filter rules can enforce policy at any level of operating system specification 2399assuming a fingerprint is present. 2400Policy could limit traffic to approved operating systems or even ban traffic 2401from hosts that aren't at the latest service pack. 2402.Pp 2403The 2404.Ar unknown 2405class can also be used as the fingerprint which will match packets for 2406which no operating system fingerprint is known. 2407.Pp 2408Examples: 2409.Bd -literal -offset indent 2410pass out proto tcp from any os OpenBSD 2411block out proto tcp from any os Doors 2412block out proto tcp from any os "Doors PT" 2413block out proto tcp from any os "Doors PT SP3" 2414block out from any os "unknown" 2415pass on lo0 proto tcp from any os "OpenBSD 3.3 lo0" 2416.Ed 2417.Pp 2418Operating system fingerprinting is limited only to the TCP SYN packet. 2419This means that it will not work on other protocols and will not match 2420a currently established connection. 2421.Pp 2422Caveat: operating system fingerprints are occasionally wrong. 2423There are three problems: an attacker can trivially craft his packets to 2424appear as any operating system he chooses; 2425an operating system patch could change the stack behavior and no fingerprints 2426will match it until the database is updated; 2427and multiple operating systems may have the same fingerprint. 2428.Sh BLOCKING SPOOFED TRAFFIC 2429"Spoofing" is the faking of IP addresses, typically for malicious 2430purposes. 2431The 2432.Ar antispoof 2433directive expands to a set of filter rules which will block all 2434traffic with a source IP from the network(s) directly connected 2435to the specified interface(s) from entering the system through 2436any other interface. 2437.Pp 2438For example, the line 2439.Bd -literal -offset indent 2440antispoof for lo0 2441.Ed 2442.Pp 2443expands to 2444.Bd -literal -offset indent 2445block drop in on ! lo0 inet from 127.0.0.1/8 to any 2446block drop in on ! lo0 inet6 from ::1 to any 2447.Ed 2448.Pp 2449For non-loopback interfaces, there are additional rules to block incoming 2450packets with a source IP address identical to the interface's IP(s). 2451For example, assuming the interface wi0 had an IP address of 10.0.0.1 and a 2452netmask of 255.255.255.0, 2453the line 2454.Bd -literal -offset indent 2455antispoof for wi0 inet 2456.Ed 2457.Pp 2458expands to 2459.Bd -literal -offset indent 2460block drop in on ! wi0 inet from 10.0.0.0/24 to any 2461block drop in inet from 10.0.0.1 to any 2462.Ed 2463.Pp 2464Caveat: Rules created by the 2465.Ar antispoof 2466directive interfere with packets sent over loopback interfaces 2467to local addresses. 2468One should pass these explicitly. 2469.Sh FRAGMENT HANDLING 2470The size of IP datagrams (packets) can be significantly larger than the 2471maximum transmission unit (MTU) of the network. 2472In cases when it is necessary or more efficient to send such large packets, 2473the large packet will be fragmented into many smaller packets that will each 2474fit onto the wire. 2475Unfortunately for a firewalling device, only the first logical fragment will 2476contain the necessary header information for the subprotocol that allows 2477.Xr pf 4 2478to filter on things such as TCP ports or to perform NAT. 2479.Pp 2480Besides the use of 2481.Ar scrub 2482rules as described in 2483.Sx TRAFFIC NORMALIZATION 2484above, there are three options for handling fragments in the packet filter. 2485.Pp 2486One alternative is to filter individual fragments with filter rules. 2487If no 2488.Ar scrub 2489rule applies to a fragment, it is passed to the filter. 2490Filter rules with matching IP header parameters decide whether the 2491fragment is passed or blocked, in the same way as complete packets 2492are filtered. 2493Without reassembly, fragments can only be filtered based on IP header 2494fields (source/destination address, protocol), since subprotocol header 2495fields are not available (TCP/UDP port numbers, ICMP code/type). 2496The 2497.Ar fragment 2498option can be used to restrict filter rules to apply only to 2499fragments, but not complete packets. 2500Filter rules without the 2501.Ar fragment 2502option still apply to fragments, if they only specify IP header fields. 2503For instance, the rule 2504.Bd -literal -offset indent 2505pass in proto tcp from any to any port 80 2506.Ed 2507.Pp 2508never applies to a fragment, even if the fragment is part of a TCP 2509packet with destination port 80, because without reassembly this information 2510is not available for each fragment. 2511This also means that fragments cannot create new or match existing 2512state table entries, which makes stateful filtering and address 2513translation (NAT, redirection) for fragments impossible. 2514.Pp 2515It's also possible to reassemble only certain fragments by specifying 2516source or destination addresses or protocols as parameters in 2517.Ar scrub 2518rules. 2519.Pp 2520In most cases, the benefits of reassembly outweigh the additional 2521memory cost, and it's recommended to use 2522.Ar scrub 2523rules to reassemble 2524all fragments via the 2525.Ar fragment reassemble 2526modifier. 2527.Pp 2528The memory allocated for fragment caching can be limited using 2529.Xr pfctl 8 . 2530Once this limit is reached, fragments that would have to be cached 2531are dropped until other entries time out. 2532The timeout value can also be adjusted. 2533.Pp 2534Currently, only IPv4 fragments are supported and IPv6 fragments 2535are blocked unconditionally. 2536.Sh ANCHORS 2537Besides the main ruleset, 2538.Xr pfctl 8 2539can load rulesets into 2540.Ar anchor 2541attachment points. 2542An 2543.Ar anchor 2544is a container that can hold rules, address tables, and other anchors. 2545.Pp 2546An 2547.Ar anchor 2548has a name which specifies the path where 2549.Xr pfctl 8 2550can be used to access the anchor to perform operations on it, such as 2551attaching child anchors to it or loading rules into it. 2552Anchors may be nested, with components separated by 2553.Sq / 2554characters, similar to how file system hierarchies are laid out. 2555The main ruleset is actually the default anchor, so filter and 2556translation rules, for example, may also be contained in any anchor. 2557.Pp 2558An anchor can reference another 2559.Ar anchor 2560attachment point 2561using the following kinds 2562of rules: 2563.Bl -tag -width xxxx 2564.It Ar nat-anchor Aq Ar name 2565Evaluates the 2566.Ar nat 2567rules in the specified 2568.Ar anchor . 2569.It Ar rdr-anchor Aq Ar name 2570Evaluates the 2571.Ar rdr 2572rules in the specified 2573.Ar anchor . 2574.It Ar binat-anchor Aq Ar name 2575Evaluates the 2576.Ar binat 2577rules in the specified 2578.Ar anchor . 2579.It Ar anchor Aq Ar name 2580Evaluates the filter rules in the specified 2581.Ar anchor . 2582.It Xo Ar load anchor 2583.Aq Ar name 2584.Ar from Aq Ar file 2585.Xc 2586Loads the rules from the specified file into the 2587anchor 2588.Ar name . 2589.El 2590.Pp 2591When evaluation of the main ruleset reaches an 2592.Ar anchor 2593rule, 2594.Xr pf 4 2595will proceed to evaluate all rules specified in that anchor. 2596.Pp 2597Matching filter and translation rules marked with the 2598.Ar quick 2599option are final and abort the evaluation of the rules in other 2600anchors and the main ruleset. 2601If the 2602.Ar anchor 2603itself is marked with the 2604.Ar quick 2605option, 2606ruleset evaluation will terminate when the anchor is exited if the packet is 2607matched by any rule within the anchor. 2608.Pp 2609.Ar anchor 2610rules are evaluated relative to the anchor in which they are contained. 2611For example, all 2612.Ar anchor 2613rules specified in the main ruleset will reference anchor 2614attachment points underneath the main ruleset, and 2615.Ar anchor 2616rules specified in a file loaded from a 2617.Ar load anchor 2618rule will be attached under that anchor point. 2619.Pp 2620Rules may be contained in 2621.Ar anchor 2622attachment points which do not contain any rules when the main ruleset 2623is loaded, and later such anchors can be manipulated through 2624.Xr pfctl 8 2625without reloading the main ruleset or other anchors. 2626For example, 2627.Bd -literal -offset indent 2628ext_if = \&"kue0\&" 2629block on $ext_if all 2630anchor spam 2631pass out on $ext_if all 2632pass in on $ext_if proto tcp from any \e 2633 to $ext_if port smtp 2634.Ed 2635.Pp 2636blocks all packets on the external interface by default, then evaluates 2637all rules in the 2638.Ar anchor 2639named "spam", and finally passes all outgoing connections and 2640incoming connections to port 25. 2641.Bd -literal -offset indent 2642# echo \&"block in quick from 1.2.3.4 to any\&" \&| \e 2643 pfctl -a spam -f - 2644.Ed 2645.Pp 2646This loads a single rule into the 2647.Ar anchor , 2648which blocks all packets from a specific address. 2649.Pp 2650The anchor can also be populated by adding a 2651.Ar load anchor 2652rule after the 2653.Ar anchor 2654rule: 2655.Bd -literal -offset indent 2656anchor spam 2657load anchor spam from "/etc/pf-spam.conf" 2658.Ed 2659.Pp 2660When 2661.Xr pfctl 8 2662loads 2663.Nm , 2664it will also load all the rules from the file 2665.Pa /etc/pf-spam.conf 2666into the anchor. 2667.Pp 2668Optionally, 2669.Ar anchor 2670rules can specify packet filtering parameters using the same syntax as 2671filter rules. 2672When parameters are used, the 2673.Ar anchor 2674rule is only evaluated for matching packets. 2675This allows conditional evaluation of anchors, like: 2676.Bd -literal -offset indent 2677block on $ext_if all 2678anchor spam proto tcp from any to any port smtp 2679pass out on $ext_if all 2680pass in on $ext_if proto tcp from any to $ext_if port smtp 2681.Ed 2682.Pp 2683The rules inside 2684.Ar anchor 2685spam are only evaluated for 2686.Ar tcp 2687packets with destination port 25. 2688Hence, 2689.Bd -literal -offset indent 2690# echo \&"block in quick from 1.2.3.4 to any" \&| \e 2691 pfctl -a spam -f - 2692.Ed 2693.Pp 2694will only block connections from 1.2.3.4 to port 25. 2695.Pp 2696Anchors may end with the asterisk 2697.Pq Sq * 2698character, which signifies that all anchors attached at that point 2699should be evaluated in the alphabetical ordering of their anchor name. 2700For example, 2701.Bd -literal -offset indent 2702anchor "spam/*" 2703.Ed 2704.Pp 2705will evaluate each rule in each anchor attached to the 2706.Li spam 2707anchor. 2708Note that it will only evaluate anchors that are directly attached to the 2709.Li spam 2710anchor, and will not descend to evaluate anchors recursively. 2711.Pp 2712Since anchors are evaluated relative to the anchor in which they are 2713contained, there is a mechanism for accessing the parent and ancestor 2714anchors of a given anchor. 2715Similar to file system path name resolution, if the sequence 2716.Dq .. 2717appears as an anchor path component, the parent anchor of the current 2718anchor in the path evaluation at that point will become the new current 2719anchor. 2720As an example, consider the following: 2721.Bd -literal -offset indent 2722# echo ' anchor "spam/allowed" ' | pfctl -f - 2723# echo -e ' anchor "../banned" \en pass' | \e 2724 pfctl -a spam/allowed -f - 2725.Ed 2726.Pp 2727Evaluation of the main ruleset will lead into the 2728.Li spam/allowed 2729anchor, which will evaluate the rules in the 2730.Li spam/banned 2731anchor, if any, before finally evaluating the 2732.Ar pass 2733rule. 2734.Pp 2735Filter rule 2736.Ar anchors 2737can also be loaded inline in the ruleset within a brace ('{' '}') delimited 2738block. 2739Brace delimited blocks may contain rules or other brace-delimited blocks. 2740When anchors are loaded this way the anchor name becomes optional. 2741.Bd -literal -offset indent 2742anchor "external" on egress { 2743 block 2744 anchor out { 2745 pass proto tcp from any to port { 25, 80, 443 } 2746 } 2747 pass in proto tcp to any port 22 2748} 2749.Ed 2750.Pp 2751Since the parser specification for anchor names is a string, any 2752reference to an anchor name containing 2753.Sq / 2754characters will require double quote 2755.Pq Sq \&" 2756characters around the anchor name. 2757.Sh TRANSLATION EXAMPLES 2758This example maps incoming requests on port 80 to port 8080, on 2759which a daemon is running (because, for example, it is not run as root, 2760and therefore lacks permission to bind to port 80). 2761.Bd -literal 2762# use a macro for the interface name, so it can be changed easily 2763ext_if = \&"ne3\&" 2764 2765# map daemon on 8080 to appear to be on 80 2766rdr on $ext_if proto tcp from any to any port 80 -\*(Gt 127.0.0.1 port 8080 2767.Ed 2768.Pp 2769If the 2770.Ar pass 2771modifier is given, packets matching the translation rule are passed without 2772inspecting the filter rules: 2773.Bd -literal 2774rdr pass on $ext_if proto tcp from any to any port 80 -\*(Gt 127.0.0.1 \e 2775 port 8080 2776.Ed 2777.Pp 2778In the example below, vlan12 is configured as 192.168.168.1; 2779the machine translates all packets coming from 192.168.168.0/24 to 204.92.77.111 2780when they are going out any interface except vlan12. 2781This has the net effect of making traffic from the 192.168.168.0/24 2782network appear as though it is the Internet routable address 2783204.92.77.111 to nodes behind any interface on the router except 2784for the nodes on vlan12. 2785(Thus, 192.168.168.1 can talk to the 192.168.168.0/24 nodes.) 2786.Bd -literal 2787nat on ! vlan12 from 192.168.168.0/24 to any -\*(Gt 204.92.77.111 2788.Ed 2789.Pp 2790In the example below, the machine sits between a fake internal 144.19.74.* 2791network, and a routable external IP of 204.92.77.100. 2792The 2793.Ar no nat 2794rule excludes protocol AH from being translated. 2795.Bd -literal 2796# NO NAT 2797no nat on $ext_if proto ah from 144.19.74.0/24 to any 2798nat on $ext_if from 144.19.74.0/24 to any -\*(Gt 204.92.77.100 2799.Ed 2800.Pp 2801In the example below, packets bound for one specific server, as well as those 2802generated by the sysadmins are not proxied; all other connections are. 2803.Bd -literal 2804# NO RDR 2805no rdr on $int_if proto { tcp, udp } from any to $server port 80 2806no rdr on $int_if proto { tcp, udp } from $sysadmins to any port 80 2807rdr on $int_if proto { tcp, udp } from any to any port 80 -\*(Gt 127.0.0.1 \e 2808 port 80 2809.Ed 2810.Pp 2811This longer example uses both a NAT and a redirection. 2812The external interface has the address 157.161.48.183. 2813On localhost, we are running 2814.Xr ftp-proxy 8 , 2815waiting for FTP sessions to be redirected to it. 2816The three mandatory anchors for 2817.Xr ftp-proxy 8 2818are omitted from this example; see the 2819.Xr ftp-proxy 8 2820manpage. 2821.Bd -literal 2822# NAT 2823# Translate outgoing packets' source addresses (any protocol). 2824# In this case, any address but the gateway's external address is mapped. 2825nat on $ext_if inet from ! ($ext_if) to any -\*(Gt ($ext_if) 2826 2827# NAT PROXYING 2828# Map outgoing packets' source port to an assigned proxy port instead of 2829# an arbitrary port. 2830# In this case, proxy outgoing isakmp with port 500 on the gateway. 2831nat on $ext_if inet proto udp from any port = isakmp to any -\*(Gt ($ext_if) \e 2832 port 500 2833 2834# BINAT 2835# Translate outgoing packets' source address (any protocol). 2836# Translate incoming packets' destination address to an internal machine 2837# (bidirectional). 2838binat on $ext_if from 10.1.2.150 to any -\*(Gt $ext_if 2839 2840# RDR 2841# Translate incoming packets' destination addresses. 2842# As an example, redirect a TCP and UDP port to an internal machine. 2843rdr on $ext_if inet proto tcp from any to ($ext_if) port 8080 \e 2844 -\*(Gt 10.1.2.151 port 22 2845rdr on $ext_if inet proto udp from any to ($ext_if) port 8080 \e 2846 -\*(Gt 10.1.2.151 port 53 2847 2848# RDR 2849# Translate outgoing ftp control connections to send them to localhost 2850# for proxying with ftp-proxy(8) running on port 8021. 2851rdr on $int_if proto tcp from any to any port 21 -\*(Gt 127.0.0.1 port 8021 2852.Ed 2853.Pp 2854In this example, a NAT gateway is set up to translate internal addresses 2855using a pool of public addresses (192.0.2.16/28) and to redirect 2856incoming web server connections to a group of web servers on the internal 2857network. 2858.Bd -literal 2859# NAT LOAD BALANCE 2860# Translate outgoing packets' source addresses using an address pool. 2861# A given source address is always translated to the same pool address by 2862# using the source-hash keyword. 2863nat on $ext_if inet from any to any -\*(Gt 192.0.2.16/28 source-hash 2864 2865# RDR ROUND ROBIN 2866# Translate incoming web server connections to a group of web servers on 2867# the internal network. 2868rdr on $ext_if proto tcp from any to any port 80 \e 2869 -\*(Gt { 10.1.2.155, 10.1.2.160, 10.1.2.161 } round-robin 2870.Ed 2871.Sh FILTER EXAMPLES 2872.Bd -literal 2873# The external interface is kue0 2874# (157.161.48.183, the only routable address) 2875# and the private network is 10.0.0.0/8, for which we are doing NAT. 2876 2877# use a macro for the interface name, so it can be changed easily 2878ext_if = \&"kue0\&" 2879 2880# normalize all incoming traffic 2881scrub in on $ext_if all fragment reassemble 2882 2883# block and log everything by default 2884block return log on $ext_if all 2885 2886# block anything coming from source we have no back routes for 2887block in from no-route to any 2888 2889# block packets whose ingress interface does not match the one in 2890# the route back to their source address 2891block in from urpf-failed to any 2892 2893# block and log outgoing packets that do not have our address as source, 2894# they are either spoofed or something is misconfigured (NAT disabled, 2895# for instance), we want to be nice and do not send out garbage. 2896block out log quick on $ext_if from ! 157.161.48.183 to any 2897 2898# silently drop broadcasts (cable modem noise) 2899block in quick on $ext_if from any to 255.255.255.255 2900 2901# block and log incoming packets from reserved address space and invalid 2902# addresses, they are either spoofed or misconfigured, we cannot reply to 2903# them anyway (hence, no return-rst). 2904block in log quick on $ext_if from { 10.0.0.0/8, 172.16.0.0/12, \e 2905 192.168.0.0/16, 255.255.255.255/32 } to any 2906 2907# ICMP 2908 2909# pass out/in certain ICMP queries and keep state (ping) 2910# state matching is done on host addresses and ICMP id (not type/code), 2911# so replies (like 0/0 for 8/0) will match queries 2912# ICMP error messages (which always refer to a TCP/UDP packet) are 2913# handled by the TCP/UDP states 2914pass on $ext_if inet proto icmp all icmp-type 8 code 0 2915 2916# UDP 2917 2918# pass out all UDP connections and keep state 2919pass out on $ext_if proto udp all 2920 2921# pass in certain UDP connections and keep state (DNS) 2922pass in on $ext_if proto udp from any to any port domain 2923 2924# TCP 2925 2926# pass out all TCP connections and modulate state 2927pass out on $ext_if proto tcp all modulate state 2928 2929# pass in certain TCP connections and keep state (SSH, SMTP, DNS, IDENT) 2930pass in on $ext_if proto tcp from any to any port { ssh, smtp, domain, \e 2931 auth } 2932 2933# Do not allow Windows 9x SMTP connections since they are typically 2934# a viral worm. Alternately we could limit these OSes to 1 connection each. 2935block in on $ext_if proto tcp from any os {"Windows 95", "Windows 98"} \e 2936 to any port smtp 2937 2938# IPv6 2939# pass in/out all IPv6 traffic: note that we have to enable this in two 2940# different ways, on both our physical interface and our tunnel 2941pass quick on gif0 inet6 2942pass quick on $ext_if proto ipv6 2943 2944# Using the pickup options to keep/modulate/synproxy state 2945# 2946# no-pickups (default) Do not allow connections to be picked up in the 2947# middle. Implies flags S/SA (the 'no-pickups' option need 2948# not be specified, it is the default). 2949# 2950# pickups Allow connections to be picked up in the middle, even if 2951# no window scaling information is known. Such connections 2952# will disable sequence space checks. Implies no flag 2953# restrictions. 2954# 2955# hash-only Do not fail packets on sequence space checks. Implies no 2956# flag restrictions. 2957 2958pass in on $ext_if proto tcp ... keep state (no-pickups) 2959pass in on $ext_if proto tcp ... keep state (pickups) 2960pass in on $ext_if proto tcp ... keep state (hash-only) 2961 2962# Packet Tagging 2963 2964# three interfaces: $int_if, $ext_if, and $wifi_if (wireless). NAT is 2965# being done on $ext_if for all outgoing packets. tag packets in on 2966# $int_if and pass those tagged packets out on $ext_if. all other 2967# outgoing packets (i.e., packets from the wireless network) are only 2968# permitted to access port 80. 2969 2970pass in on $int_if from any to any tag INTNET 2971pass in on $wifi_if from any to any 2972 2973block out on $ext_if from any to any 2974pass out quick on $ext_if tagged INTNET 2975pass out on $ext_if proto tcp from any to any port 80 2976 2977# tag incoming packets as they are redirected to spamd(8). use the tag 2978# to pass those packets through the packet filter. 2979 2980rdr on $ext_if inet proto tcp from \*(Ltspammers\*(Gt to port smtp \e 2981 tag SPAMD -\*(Gt 127.0.0.1 port spamd 2982 2983block in on $ext_if 2984pass in on $ext_if inet proto tcp tagged SPAMD 2985.Ed 2986.Sh GRAMMAR 2987Syntax for 2988.Nm 2989in BNF: 2990.Bd -literal 2991line = ( option | pf-rule | nat-rule | binat-rule | rdr-rule | 2992 antispoof-rule | altq-rule | queue-rule | trans-anchors | 2993 anchor-rule | anchor-close | load-anchor | table-rule | 2994 include ) 2995 2996option = "set" ( [ "timeout" ( timeout | "{" timeout-list "}" ) ] | 2997 [ "ruleset-optimization" [ "none" | "basic" | "profile" ] ] | 2998 [ "optimization" [ "default" | "normal" | 2999 "high-latency" | "satellite" | 3000 "aggressive" | "conservative" ] ] 3001 [ "limit" ( limit-item | "{" limit-list "}" ) ] | 3002 [ "loginterface" ( interface-name | "none" ) ] | 3003 [ "block-policy" ( "drop" | "return" ) ] | 3004 [ "keep-policy" keep ] | 3005 [ "state-policy" ( "if-bound" | "floating" ) ] 3006 [ "require-order" ( "yes" | "no" ) ] 3007 [ "fingerprints" filename ] | 3008 [ "skip on" ifspec ] | 3009 [ "debug" ( "none" | "urgent" | "misc" | "loud" ) ] ) 3010 3011pf-rule = action [ ( "in" | "out" ) ] 3012 [ "log" [ "(" logopts ")"] ] [ "quick" ] 3013 [ "on" ifspec ] [ "fastroute" | route ] [ af ] [ protospec ] 3014 hosts [ filteropt-list ] 3015 3016logopts = logopt [ "," logopts ] 3017logopt = "all" | "user" | "to" interface-name 3018 3019filteropt-list = filteropt-list filteropt | filteropt 3020filteropt = user | group | flags | icmp-type | icmp6-type | tos | 3021 keep | "fragment" | "no-df" | "min-ttl" number | 3022 "max-mss" number | "random-id" | "reassemble tcp" | 3023 fragmentation | "allow-opts" | 3024 "label" string | "tag" string | [ ! ] "tagged" string | 3025 "queue" ( string | "(" string [ [ "," ] string ] ")" ) | 3026 "probability" number"%" 3027 3028keep = "no" "state" | 3029 ( "keep" | "modulate" | "synproxy" ) "state" 3030 [ "(" state-opts ")" ] 3031 3032nat-rule = [ "no" ] "nat" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3033 [ "on" ifspec ] [ af ] 3034 [ protospec ] hosts [ "tag" string ] [ "tagged" string ] 3035 [ "-\*(Gt" ( redirhost | "{" redirhost-list "}" ) 3036 [ portspec ] [ pooltype ] [ "static-port" ] ] 3037 3038binat-rule = [ "no" ] "binat" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3039 [ "on" interface-name ] [ af ] 3040 [ "proto" ( proto-name | proto-number ) ] 3041 "from" address [ "/" mask-bits ] "to" ipspec 3042 [ "tag" string ] [ "tagged" string ] 3043 [ "-\*(Gt" address [ "/" mask-bits ] ] 3044 3045rdr-rule = [ "no" ] "rdr" [ "pass" [ "log" [ "(" logopts ")" ] ] ] 3046 [ "on" ifspec ] [ af ] 3047 [ protospec ] hosts [ "tag" string ] [ "tagged" string ] 3048 [ "-\*(Gt" ( redirhost | "{" redirhost-list "}" ) 3049 [ portspec ] [ pooltype ] ] 3050 3051antispoof-rule = "antispoof" [ "log" ] [ "quick" ] 3052 "for" ifspec [ af ] [ "label" string ] 3053 3054table-rule = "table" "\*(Lt" string "\*(Gt" [ tableopts-list ] 3055tableopts-list = tableopts-list tableopts | tableopts 3056tableopts = "persist" | "const" | "counters" | "file" string | 3057 "{" [ tableaddr-list ] "}" 3058tableaddr-list = tableaddr-list [ "," ] tableaddr-spec | tableaddr-spec 3059tableaddr-spec = [ "!" ] tableaddr [ "/" mask-bits ] 3060tableaddr = hostname | ifspec | "self" | 3061 ipv4-dotted-quad | ipv6-coloned-hex 3062 3063altq-rule = "altq on" interface-name queueopts-list 3064 "queue" subqueue 3065queue-rule = "queue" string [ "on" interface-name ] queueopts-list 3066 subqueue 3067 3068anchor-rule = "anchor" [ string ] [ ( "in" | "out" ) ] [ "on" ifspec ] 3069 [ af ] [ "proto" ] [ protospec ] [ hosts ] 3070 3071trans-anchors = ( "nat-anchor" | "rdr-anchor" | "binat-anchor" ) string 3072 [ "on" ifspec ] [ af ] [ "proto" ] [ protospec ] [ hosts ] 3073 3074load-anchor = "load anchor" string "from" filename 3075 3076queueopts-list = queueopts-list queueopts | queueopts 3077queueopts = [ "bandwidth" bandwidth-spec ] | 3078 [ "qlimit" number ] | [ "tbrsize" number ] | 3079 [ "priority" number ] | [ schedulers ] 3080schedulers = ( cbq-def | hfsc-def | priq-def | fairq-def ) 3081bandwidth-spec = "number" ( "b" | "Kb" | "Mb" | "Gb" | "%" ) 3082 3083action = "pass" | "block" [ return ] | [ "no" ] "scrub" 3084return = "drop" | "return" | "return-rst" [ "( ttl" number ")" ] | 3085 "return-icmp" [ "(" icmpcode [ [ "," ] icmp6code ] ")" ] | 3086 "return-icmp6" [ "(" icmp6code ")" ] 3087icmpcode = ( icmp-code-name | icmp-code-number ) 3088icmp6code = ( icmp6-code-name | icmp6-code-number ) 3089 3090ifspec = ( [ "!" ] ( interface-name | interface-group ) ) | 3091 "{" interface-list "}" 3092interface-list = [ "!" ] ( interface-name | interface-group ) 3093 [ [ "," ] interface-list ] 3094route = ( "route-to" | "reply-to" | "dup-to" ) 3095 ( routehost | "{" routehost-list "}" ) 3096 [ pooltype ] 3097af = "inet" | "inet6" 3098 3099protospec = "proto" ( proto-name | proto-number | 3100 "{" proto-list "}" ) 3101proto-list = ( proto-name | proto-number ) [ [ "," ] proto-list ] 3102 3103hosts = "all" | 3104 "from" ( "any" | "no-route" | "urpf-failed" | "self" | host | 3105 "{" host-list "}" | "route" string ) [ port ] [ os ] 3106 "to" ( "any" | "no-route" | "self" | host | 3107 "{" host-list "}" | "route" string ) [ port ] 3108 3109ipspec = "any" | host | "{" host-list "}" 3110host = [ "!" ] ( address [ "/" mask-bits ] | "\*(Lt" string "\*(Gt" ) 3111redirhost = address [ "/" mask-bits ] 3112routehost = "(" interface-name [ address [ "/" mask-bits ] ] ")" 3113address = ( interface-name | interface-group | 3114 "(" ( interface-name | interface-group ) ")" | 3115 hostname | ipv4-dotted-quad | ipv6-coloned-hex ) 3116host-list = host [ [ "," ] host-list ] 3117redirhost-list = redirhost [ [ "," ] redirhost-list ] 3118routehost-list = routehost [ [ "," ] routehost-list ] 3119 3120port = "port" ( unary-op | binary-op | "{" op-list "}" ) 3121portspec = "port" ( number | name ) [ ":" ( "*" | number | name ) ] 3122os = "os" ( os-name | "{" os-list "}" ) 3123user = "user" ( unary-op | binary-op | "{" op-list "}" ) 3124group = "group" ( unary-op | binary-op | "{" op-list "}" ) 3125 3126unary-op = [ "=" | "!=" | "\*(Lt" | "\*(Le" | "\*(Gt" | "\*(Ge" ] 3127 ( name | number ) 3128binary-op = number ( "\*(Lt\*(Gt" | "\*(Gt\*(Lt" | ":" ) number 3129op-list = ( unary-op | binary-op ) [ [ "," ] op-list ] 3130 3131os-name = operating-system-name 3132os-list = os-name [ [ "," ] os-list ] 3133 3134flags = "flags" ( [ flag-set ] "/" flag-set | "any" ) 3135flag-set = [ "F" ] [ "S" ] [ "R" ] [ "P" ] [ "A" ] [ "U" ] [ "E" ] 3136 [ "W" ] 3137 3138icmp-type = "icmp-type" ( icmp-type-code | "{" icmp-list "}" ) 3139icmp6-type = "icmp6-type" ( icmp-type-code | "{" icmp-list "}" ) 3140icmp-type-code = ( icmp-type-name | icmp-type-number ) 3141 [ "code" ( icmp-code-name | icmp-code-number ) ] 3142icmp-list = icmp-type-code [ [ "," ] icmp-list ] 3143 3144tos = ( "lowdelay" | "throughput" | "reliability" | 3145 [ "0x" ] number ) 3146 3147state-opts = state-opt [ [ "," ] state-opts ] 3148state-opt = "max" number | "no-sync" | timeout | 3149 "source-track" [ "rule" | "global" ] | 3150 "max-src-nodes" number | "max-src-states" number | 3151 "max-src-conn" number | 3152 "max-src-conn-rate" number "/" number | 3153 "overload" "\*(Lt" string "\*(Gt" [ "flush" ] | 3154 "if-bound" | "floating" | 3155 "pickups" | "no-pickups" | "hash-only" 3156 3157fragmentation = [ "fragment reassemble" | "fragment crop" | 3158 "fragment drop-ovl" ] 3159 3160timeout-list = timeout [ [ "," ] timeout-list ] 3161timeout = ( "tcp.first" | "tcp.opening" | "tcp.established" | 3162 "tcp.closing" | "tcp.finwait" | "tcp.closed" | 3163 "udp.first" | "udp.single" | "udp.multiple" | 3164 "icmp.first" | "icmp.error" | 3165 "other.first" | "other.single" | "other.multiple" | 3166 "frag" | "interval" | "src.track" | 3167 "adaptive.start" | "adaptive.end" ) number 3168 3169limit-list = limit-item [ [ "," ] limit-list ] 3170limit-item = ( "states" | "frags" | "src-nodes" ) number 3171 3172pooltype = ( "bitmask" | "random" | 3173 "source-hash" [ hex-key | string-key ] | 3174 "round-robin" ) [ sticky-address ] 3175 3176subqueue = string | "{" queue-list "}" 3177queue-list = string [ [ "," ] string ] 3178 3179cbq-def = "cbq" [ "(" cbq-opts ")" ] 3180priq-def = "priq" [ "(" priq-opts ")" ] 3181hfsc-def = "hfsc" [ "(" hfsc-opts ")" ] 3182fairq-def = "fairq" [ "(" fairq-opts ")" ] 3183 3184cbq-opts = cbq-opt [ [ "," ] cbq-opts ] 3185priq-opts = priq-opt [ [ "," ] priq-opts ] 3186hfsc-opts = hfsc-opt [ [ "," ] hfsc-opts ] 3187fairq-opts = fairq-opt [ [ "," ] fairq-opts ] 3188 3189cbq-opt = "default" | "borrow" | "red" | "ecn" | "rio" 3190priq-opt = "default" | "red" | "ecn" | "rio" 3191hfsc-opt = "default" | "red" | "ecn" | "rio" | 3192 linkshare-sc | realtime-sc | upperlimit-sc 3193fairq-opt = "default" | "red" | "ecn" | "rio" | 3194 "buckets" number | "hogs" number | linkshare-sc 3195 3196linkshare-sc = "linkshare" sc-spec 3197realtime-sc = "realtime" sc-spec 3198upperlimit-sc = "upperlimit" sc-spec 3199sc-spec = ( bandwidth-spec | 3200 "(" bandwidth-spec number bandwidth-spec ")" ) 3201include = "include" filename 3202.Ed 3203.Sh FILES 3204.Bl -tag -width ".Pa /usr/share/examples/pf" -compact 3205.It Pa /etc/hosts 3206Host name database. 3207.It Pa /etc/pf.conf 3208Default location of the ruleset file. 3209.It Pa /etc/pf.os 3210Default location of OS fingerprints. 3211.It Pa /etc/protocols 3212Protocol name database. 3213.It Pa /etc/services 3214Service name database. 3215.It Pa /usr/share/examples/pf 3216Example rulesets. 3217.El 3218.Sh SEE ALSO 3219.Xr altq 4 , 3220.Xr carp 4 , 3221.Xr icmp 4 , 3222.Xr icmp6 4 , 3223.Xr ip 4 , 3224.Xr ip6 4 , 3225.Xr pf 4 , 3226.Xr pfsync 4 , 3227.Xr route 4 , 3228.Xr tcp 4 , 3229.Xr udp 4 , 3230.Xr hosts 5 , 3231.Xr pf.os 5 , 3232.Xr protocols 5 , 3233.Xr services 5 , 3234.Xr ftp-proxy 8 , 3235.Xr pfctl 8 , 3236.Xr pflogd 8 , 3237.Xr route 8 3238.Sh HISTORY 3239The 3240.Nm 3241file format first appeared in 3242.Ox 3.0 . 3243