[Previous: Issues with FTP] [Contents] [Next: Firewall Redundancy with CARP and pfsync]
Authpf loads a user's filter/NAT rules into a unique anchor point. The anchor is named by combining the user's UNIX username and the authpf process-id into the format "username(PID)". Each users' anchor is stored within the authpf anchor which is in turn anchored to the main ruleset. The "fully qualified anchor path" then becomes:
main_ruleset/authpf/username(PID)
The rules that authpf loads can be configured on a per-user or global basis.
Example uses of authpf include:
Authpf logs the username and IP address of each user who authenticates successfully as well as the start and end times of their login session via syslogd(8). By using this information, an administrator can determine who was logged in when and also make users accountable for their network traffic.
Authpf will not run if the config file /etc/authpf/authpf.conf is not present. Even if the file is empty (zero size), it must still be present or authpf will exit immediately after a user authenticates successfully.
The following configuration directives can be placed in authpf.conf:
nat-anchor "authpf/*"
rdr-anchor "authpf/*"
binat-anchor "authpf/*"
anchor "authpf/*"
Wherever the anchor rules are placed within the ruleset is where PF will branch off from the main ruleset to evaluate the authpf rules. It's not necessary for all four anchor rules to be present; for example, if authpf hasn't been setup to load any nat rules, the nat-anchor rule can be omitted.
The first file contains rules that are only loaded when the user $USER (which is replaced with the user's username) logs in. The per-user rule configuration is used when a specific user -- such as an administrator -- requires a set of rules that is different than the default set. The second file contains the default rules which are loaded for any user that doesn't have their own authpf.rules file. If the user-specific file exists, it will override the default file. At least one of the files must exist or authpf will not run.
Filter and translation rules have the same syntax as in any other PF ruleset with one exception: Authpf allows for the use of two predefined macros:
It's recommended practice to use the $user_ip macro to only permit traffic through the gateway from the authenticated user's computer.
In addition to the $user_ip macro, authpf will make use of the authpf_users table (if it exists) for storing the IP addresses of all authenticated users. Be sure to define the table before using it:
table <authpf_users> persist
pass in on $ext_if proto tcp from <authpf_users> \
to port smtp flags S/SA keep state
This table should only be used in rules that are meant to apply to all authenticated users.
Conversely, it's also possible to allow only specific users access by placing usernames in the /etc/authpf/authpf.allow file. If the /etc/authpf/authpf.allow file does not exist or "*" is entered into the file, then authpf will permit access to any user who successfully logs in via SSH as long as they are not explicitly banned.
If authpf is unable to determine if a username is allowed or denied, it will print a brief message and then disconnect the user. An entry in /etc/authpf/banned/ always overrides an entry in /etc/authpf/authpf.allow.
Whenever a user successfully authenticates to authpf, a greeting is printed that indicates that the user is authenticated.
Hello charlie. You are authenticated from host "64.59.56.140"
This message can be supplemented by putting a custom message in /etc/authpf/authpf.message. The contents of this file will be displayed after the default welcome message.
There are a couple ways of assigning authpf as a user's shell:
Login classes are created in the login.conf(5) file. OpenBSD comes with an authpf login class defined as:
authpf:\
:welcome=/etc/motd.authpf:\
:shell=/usr/sbin/authpf:\
:tc=default:
Users are assigned to a login class by editing the class field of the user's passwd(5) database entry. One way to do this is with the chsh(1) command.
# ps -ax | grep authpf
23664 p0 Is+ 0:00.11 -authpf: charlie@192.168.1.3 (authpf)
Here the user charlie is logged in from the machine
192.168.1.3. By sending a SIGTERM signal to the authpf process, the
user can be forcefully logged out. Authpf will also remove any rules
loaded for the user and kill any stateful connections the user has open.
# kill -TERM 23664
The /etc/authpf/authpf.rules file contains the following rules:
wifi_if = "wi0"
pass in quick on $wifi_if proto tcp from $user_ip to port { ssh, http, \
https } flags S/SA keep state
|
The administrative user charlie needs to be able to access the
campus SMTP and POP3 servers in addition to surfing the web and using
SSH. The following rules are setup in
/etc/authpf/users/charlie/authpf.rules:
wifi_if = "wi0"
smtp_server = "10.0.1.50"
pop3_server = "10.0.1.51"
pass in quick on $wifi_if proto tcp from $user_ip to $smtp_server \
port smtp flags S/SA keep state
pass in quick on $wifi_if proto tcp from $user_ip to $pop3_server \
port pop3 flags S/SA keep state
pass in quick on $wifi_if proto tcp from $user_ip to port { ssh, http, \
https } flags S/SA keep state
|
The main ruleset -- located in /etc/pf.conf -- is setup
as follows:
# macros
wifi_if = "wi0"
ext_if = "fxp0"
dns_servers = "{ 10.0.1.56, 10.0.2.56 }"
table <authpf_users> persist
scrub in all
# filter
block drop all
pass out quick on $ext_if inet proto tcp from \
{ $wifi_if:network, $ext_if } flags S/SA modulate state
pass out quick on $ext_if inet proto { udp, icmp } from \
{ $wifi_if:network, $ext_if } keep state
pass in quick on $wifi_if inet proto tcp from $wifi_if:network to $wifi_if \
port ssh flags S/SA keep state
pass in quick on $wifi_if inet proto { tcp, udp } from <authpf_users> \
to $dns_servers port domain keep state
anchor "authpf/*" in on $wifi_if
|
The ruleset is very simple and does the following:
The idea behind the main ruleset is to block everything and allow the least amount of traffic through as possible. Traffic is free to flow out on the external interface but is blocked from entering the wireless interface by the default deny policy. Once a user authenticates, their traffic is permitted to pass in on the wireless interface and to then flow through the gateway into the rest of the network. The quick keyword is used throughout so that PF doesn't have to evaluate each named ruleset when a new connection passes through the gateway.
[Previous: Issues with FTP] [Contents] [Next: Firewall Redundancy with CARP and pfsync]