OpenBSD manual page server

Manual Page Search Parameters
PFCTL(8) System Manager's Manual PFCTL(8)

pfctl — control the packet filter (PF) device

pfctl [-deghNnPqrvz] [-a anchor] [-D macro=value] [-F modifier] [-f file] [-i interface] [-K key] [-k key] [-L statefile] [-o level] [-p device] [-S statefile] [-s modifier [-R id]] [-t table -T command [address ...]] [-V rdomain] [-x level]

The pfctl utility communicates with the packet filter device using the ioctl interface described in pf(4). It allows ruleset and parameter configuration, and retrieval of status information from the packet filter. Packet filtering restricts the types of packets that pass through network interfaces entering or leaving the host based on filter rules as described in pf.conf(5). The packet filter can also replace addresses and ports of packets.

The packet filter is enabled by default. Should pfctl be unable to load a ruleset, an error occurs and the original ruleset remains in place. If this happens at system startup, the ruleset defined by the RULES variable in rc(8) remains in place.

The packet filter does not itself forward packets between interfaces. Forwarding can be enabled by setting the sysctl(8) variables and/or to 1. Set them permanently in sysctl.conf(5).

At least one option must be specified. The options are as follows:

anchor
Apply flags -f, -F, -s, and -T only to the rules in the specified anchor. In addition to the main ruleset, pfctl can load and manipulate additional rulesets by name, called anchors. The main ruleset is the default anchor.

Anchors are referenced by name and may be nested, with the various components of the anchor path separated by ‘/’ characters, similar to how file system hierarchies are laid out. The last component of the anchor path is where ruleset operations are performed.

Evaluation of anchor rules from the main ruleset is described in pf.conf(5).

For example, the following will show all filter rules (see the -s flag below) inside the anchor “authpf/smith(1234)”, which would have been created for user “smith” by authpf(8), PID 1234:

# pfctl -a "authpf/smith(1234)" -s rules

Private tables can also be put inside anchors, either by having table statements in the pf.conf(5) file that is loaded in the anchor, or by using regular table commands, as in:

# pfctl -a foo/bar -t mytable -T add 1.2.3.4 5.6.7.8

When a rule referring to a table is loaded in an anchor, the rule will use the private table if one is defined, and then fall back to the table defined in the main ruleset, if there is one. This is similar to C rules for variable scope. It is possible to create distinct tables with the same name in the global ruleset and in an anchor, but this is often bad design and a warning will be issued in that case.

By default, recursive inline printing of anchors applies only to unnamed anchors specified inline in the ruleset. If the anchor name is terminated with a ‘*’ character, the -s flag will recursively print all anchors in a brace delimited block. For example the following will print the “authpf” ruleset recursively:

# pfctl -a 'authpf/*' -sr

To print the main ruleset recursively, specify only ‘*’ as the anchor name:

# pfctl -a '*' -sr

To flush all rulesets and tables recursively, specify only ‘*’ as the anchor name:

# pfctl -a '*' -Fa
macro=value
Define macro to be set to value on the command line. Overrides the definition of macro in the ruleset.
Disable the packet filter.
Enable the packet filter.
modifier
Flush the filter parameters specified by modifier (may be abbreviated):

Flush the filter rules.
Flush the state table (NAT and filter).
Flush the source tracking table.
Flush the filter information (statistics that are not bound to rules).
Flush the tables.
Flush the passive operating system fingerprints.
Reset limits, timeouts and other options back to default settings. See the OPTIONS section in pf.conf(5) for details.
Flush all of the above.

If -a is specified as well and anchor is terminated with a ‘*’ character, rules, Tables and all flush the given anchor recursively.

file
Replace the current ruleset with the rules contained in file. This file may contain macros, tables, options, and normalization, queueing, translation, and filtering rules. With the exception of macros and tables, the statements must appear in that order.
Include output helpful for debugging.
Help.
interface
Restrict the operation to the given interface.
key
Kill all of the source tracking entries originating from the host or network specified by key. A second -K option may be specified, which will kill all the source tracking entries from the first host/network to the second.
key
Kill all of the state entries originating from the host or network specified by key. A second -k option may be specified, which will kill all the state entries from the first host/network to the second.

A network prefix length of 0 can be used as a wildcard. To kill all states with the target “host2”:

# pfctl -k 0.0.0.0/0 -k host2

It is also possible to kill states by rule label, state key, or state ID. In this mode the first -k argument is used to specify the type; a second -k gives the actual target.

To kill states by rule label, use the label modifier. To kill all states created from rules carrying the label “foobar”:

# pfctl -k label -k foobar

To kill one specific state by its state key (as shown by pfctl -s state), use the key modifier. To kill a state originating from 10.0.0.101:32123 to 10.0.0.1:80, protocol TCP, use:

# pfctl -k key -k 'tcp 10.0.0.1:80 <- 10.0.0.101:32123'

To kill one specific state by its unique state ID (as shown by pfctl -s state -vv), use the id modifier. To kill a state with ID 4823e84500000003 use:

# pfctl -k id -k 4823e84500000003

To kill a state with ID 4823e84500000018 created from a backup firewall with hostid 00000002 use:

# pfctl -k id -k 4823e84500000018/2
statefile
Load pf states from the file specified by statefile.
Do not perform domain name resolution. If a name cannot be resolved without DNS, an error will be reported.
Do not actually load rules, just parse them.
level
Control the ruleset optimizer, overriding any rule file settings.

Disable the ruleset optimizer.
Enable basic ruleset optimizations. This is the default behaviour.
Enable basic ruleset optimizations with profiling.

For further information on the ruleset optimizer, see pf.conf(5).

Print ports using their names in /etc/services if available.
device
Use the device file device instead of the default /dev/pf.
Only print errors and warnings.
Perform reverse DNS lookups on states and tables when displaying them. -N and -r are mutually exclusive.
statefile
Store the pf state table in the file specified by statefile.
modifier [-R id]
Show the filter parameters specified by modifier (may be abbreviated):

Show the currently loaded queue definitions. When used together with -v, per-queue statistics are also shown. When used together with -v -v, pfctl will loop and show updated queue statistics every five seconds, including measured bandwidth and packets per second.
Show the currently loaded filter rules. If -R id is specified as well, only the rule with the specified numeric ID is shown. When used together with -v, the per-rule statistics (number of evaluations, packets and bytes) are also shown. When used together with -g or -vv, expired rules (marked as “# expired”) are also shown. Note that the “skip step” optimization done automatically by the kernel will skip evaluation of rules where possible. Packets passed statefully are counted in the rule that created the state (even though the rule isn't evaluated more than once for the entire connection).
Show the currently loaded anchors directly attached to the main ruleset. If -a anchor is specified as well, the anchors loaded directly below the given anchor are shown instead. If -v is specified, all anchors attached under the target anchor will be displayed recursively.
Show the contents of the state table. If -R id is specified as well, only states created by the rule with the specified numeric ID are shown.
Show the contents of the source tracking table.
Show filter information (statistics and counters). When used together with -v, source tracking statistics, the firewall's 32-bit hostid number and the main ruleset's MD5 checksum for use with pfsync(4) are also shown.
Show per-rule statistics (label, evaluations, packets total, bytes total, packets in, bytes in, packets out, bytes out, state creations) of filter rules with labels, useful for accounting. If -R id is specified as well, only the statistics for the rule with the specified numeric ID are shown.
Show the current global timeouts.
Show the current pool memory hard limits.
Show the list of tables.
Show the list of operating system fingerprints.
Show the list of interfaces and interface groups available to PF. When used together with -v, it additionally lists which interfaces have skip rules activated. When used together with -vv, interface statistics are also shown. -i can be used to select an interface or a group of interfaces.
Show all of the above, except for the lists of interfaces and operating system fingerprints.

Counters shown with -s info are:

match
explicit rule match
bad-offset
currently unused
fragment
invalid fragments dropped
short
short packets dropped
normalize
dropped by normalizer: illegal packets
memory
memory could not be allocated
bad-timestamp
bad TCP timestamp; RFC 1323
congestion
network interface queue congested
ip-option
bad IP/IPv6 options
proto-cksum
invalid protocol checksum
state-mismatch
packet was associated with a state entry, but sequence numbers did not match
state-insert
state insertion failure
state-limit
configured state limit was reached
src-limit
source node/connection limit
synproxy
dropped by synproxy
translate
no free ports in translation port range
no-route
dropped by no-route
table -T command [address ...]
Specify the command (may be abbreviated) to apply to table. Commands include:

Add one or more addresses to a table. Automatically create a persistent table if it does not exist.
Delete one or more addresses from a table.
number
Delete addresses which had their statistics cleared more than number seconds ago. For entries which have never had their statistics cleared, number refers to the time they were added to the table.
Flush all addresses in a table.
Kill a table.
Replace the addresses of the table. Automatically create a persistent table if it does not exist.
Show the content (addresses) of a table.
Test if the given addresses match a table.
Clear all the statistics of a table, or only for specified addresses.

For the add, delete, replace, and test commands, the list of addresses can be specified either directly on the command line and/or in an unformatted text file, using the -f flag. Comments starting with a ‘#’ are allowed in the text file. With these commands, the -v flag can also be used once or twice, in which case pfctl will print the detailed result of the operation for each individual address, prefixed by one of the following letters:

A
The address/network has been added.
C
The address/network has been changed (negated).
D
The address/network has been deleted.
M
The address matches (test operation only).
X
The address/network is duplicated and therefore ignored.
Y
The address/network cannot be added/deleted due to conflicting ‘!’ attributes.
Z
The address/network has been cleared (statistics).

Each table can maintain a set of counters that can be retrieved using the -v flag of pfctl. For example, the following commands define a wide open firewall which will keep track of packets going to or coming from the OpenBSD FTP server. The following commands configure the firewall and send 10 pings to the FTP server:

# printf "table <test> counters { ftp.openbsd.org }\n \
    pass out to <test>\n" | pfctl -f-
# ping -qc10 ftp.openbsd.org

We can now use the table show command to output, for each address and packet direction, the number of packets and bytes that are being passed, matched or blocked by rules referencing the table. Note that the match counters are incremented for every match rule in which they are referenced, meaning that a single packet may be counted multiple times. The time at which the current accounting started is also shown with the “Cleared” line.

# pfctl -t test -vTshow
   198.51.100.81
        Cleared:        Fri Jun 28 11:17:37 2013
        In/Block:       [ Packets: 0	Bytes: 0	]
        In/Match        [ Packets: 54	Bytes: 10028	]
        In/Pass:        [ Packets: 5	Bytes: 1949	]
        Out/Block:      [ Packets: 0	Bytes: 0	]
        Out/Match       [ Packets: 65	Bytes: 12684	]
        Out/Pass:       [ Packets: 6	Bytes: 389	]

Similarly, it is possible to view global information about the tables by using the -v modifier twice and the -s Tables command. This will display the number of addresses on each table, the number of rules which reference the table, and the global packet statistics for the whole table:

# pfctl -vvsTables
--a-r-C test
        Addresses:   1
        Cleared:     Fri Jun 28 11:17:37 2013
        References:  [ Anchors: 0	Rules: 4	]
        Evaluations: [ NoMatch: 35	Match: 8	]
        In/Block:    [ Packets: 0	Bytes: 0	]
        In/Match:    [ Packets: 54	Bytes: 10028	]
        In/Pass:     [ Packets: 5	Bytes: 1949	]
        In/XPass:    [ Packets: 0	Bytes: 0	]
        Out/Block:   [ Packets: 0	Bytes: 0	]
        Out/Match:   [ Packets: 65	Bytes: 12684	]
        Out/Pass:    [ Packets: 6	Bytes: 389	]
        Out/XPass:   [ Packets: 0	Bytes: 0	]

Only packets creating state are matched in the Evaluations line, but all packets passing as a result of the state are correctly accounted for. Reloading the table(s) or ruleset will not affect packet accounting in any way. The two “XPass” counters are incremented instead of the “Pass” counters when a “stateful” packet is passed but doesn't match the table anymore. This will happen in our example if someone flushes the table while the ping(8) command is running.

When used with a single -v, pfctl will only display the first line containing the table flags and name. The flags are defined as follows:

c
For constant tables, which cannot be altered outside pf.conf(5).
p
For persistent tables, which don't get automatically killed when no rules refer to them.
a
For tables which are part of the tableset. Tables without this flag do not really exist, cannot contain addresses, and are only listed if the -g flag is given.
i
For tables which are part of the tableset. This flag can only be witnessed briefly during the loading of pf.conf(5).
r
For tables which are referenced (used) by rules.
h
This flag is set when a table in the main ruleset is hidden by one or more tables of the same name from anchors attached below it.
C
This flag is set when per-address counters are enabled on the table.
rdomain
Select the routing domain to be used to kill states by host or by label. The rdomain of a state is displayed in parentheses before the host by -s states.
Produce more verbose output. A second use of -v will produce even more verbose output including ruleset warnings. See the previous section for its effect on table commands.
level
Set the debug level, which limits the severity of log messages printed by pf(4). This should be a keyword from the following ordered list (highest to lowest): emerg, alert, crit, err, warning, notice, info, and debug. These keywords correspond to the similar (LOG_) values specified to the syslog(3) library routine, and may be abbreviated on the command line.
Clear per-rule statistics.

/etc/pf.conf
Packet filter rules file.
/etc/pf.os
Passive operating system fingerprint database.

pf(4), pf.conf(5), pf.os(5), sysctl.conf(5), authpf(8), ftp-proxy(8), rc(8), rc.conf(8), sysctl(8)

The pfctl program and the pf(4) filter mechanism first appeared in OpenBSD 3.0.

November 21, 2024 OpenBSD-current