Firewalling with OpenBSD's PF packet filter

Peter N. M. Hansteen

Datadokumentasjon A/S

This document is © Copyright 2005 - 2006, Peter N. M. Hansteen. All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

  1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

  2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.


The document is a 'work in progress', based on a manuscript prepared for a lecture at the BLUG (see meeting of January 27th, 2005.

I'm interested in comments of all kinds, and you may if you wish add web or other references to html or pdf versions of the manuscript. If you do, I would like, but can not require, you to send me an email message that you've done it. For communication regarding this document please use the address .

Table of Contents
Before we start
Packet filter? Firewall?
PF today
BSD vs Linux - Configuration
Simplest possible setup (OpenBSD)
Simplest possible setup (FreeBSD)
Simplest possible setup (NetBSD)
First rule set - single machine
Slightly stricter
Statistics from pfctl
Simple gateway with NAT
Gateways and the pitfalls of in, out and on
Setting up
That sad old FTP thing
FTP through NAT: ftp-proxy
FTP through pf with routable addresses: ftpsesame, pftpx and ftp-proxy!
ftp-proxy, new style
Making your network troubleshooting friendly
Then, do we let it all through?
The easy way out: The buck stops here
Letting ping through
Helping traceroute
Path MTU discovery
Hygiene: block-policy, scrub and antispoof
A web server and a mail server on the inside
Taking care of your own - the inside
Tables make your life easier
Taking a peek with tcpdump
But there are limits (an anecdote)
Keeping an eye on things with pftop
Invisible gateway - bridge
Directing traffic with ALTQ
ALTQ - allocation by percentage
ALTQ - prioritizing by traffic type
ALTQ - handling unwanted traffic
CARP and pfsync
Wireless networks made simple
A little IEEE 802.11 background
WEP (Wireless Equivalent Privacy)
WPA (WiFi Protected Access)
Setting up a simple wireless network
An open, yet tightly guarded wireless network with authpf
Turning away the brutes
Giving spammers a hard time
Remember, you are not alone: blacklisting
List of black and grey, and the sticky tarpit
Setting up spamd
Highlights of our spamd experience
Conclusions from our spamd experience
PF - Haiku
Where to find the tutorial on the web
If you enjoyed this: Buy OpenBSD CDs and other items, donate!

Before we start

This lecture[1] will be about firewalls and related functions, starting from a little theory along with a number of examples of filtering and other network traffic directing. As in any number of other endeavors, the things I discuss can be done in more than one way. Under any circumstances I will urge you to interrupt me when you need to. That is, if you will permit me to use what I learn from your comments later, either in revised versions of this lecture or in practice at a later time. But first,

WarningThis is not a HOWTO

This document is not intended as a pre-cooked recipe for cutting and pasting.

Just to hammer this in, please repeat after me

The Pledge of the Network Admin

This is my network. 

It is mine 
or technically my employer's, 
it is my responsibility 
and I care for it with all my heart

there are many other networks a lot like mine,

but none are just like it.

I solemnly swear 

that I will not mindlessly paste from HOWTOs.

The point is, while the rules and configurations I show you do work, I have tested them and they are in some way related to what has been put into production, they may very well be overly simplistic and are almost certain to be at least a little off and possibly quite wrong for your network.

Please keep in mind that this document is intended to show you a few useful things and inspire you to achieve good things.

Please strive to understand your network and what you need to do to make it better.

Please do not paste blindly from this document or any other.

Now, with that out of the way, we can go on to the meat of the matter.


First, a few words about the software we are about to discuss, OpenBSD's PF.

PF was written during the summer and autumn of 2001 by Daniel Hartmeier and a number of OpenBSD developers, and was launched as a default part of the OpenBSD 3.0 base system in December of 2001.

The need for a piece of new firewall software for OpenBSD arose when Darren Reed announced to the world that IPFilter, which at that point had been rather intimately integrated in OpenBSD, was not after all BSD licensed. In fact quite to the contrary. The license itself was almost a word by word copy of the BSD license, omitting only the right to make changes to the code. The OpenBSD version of IPFilter contained quite a number of changes and customizations, which it turned out were not allowed according to the license. IPFilter was removed from the OpenBSD source tree on May 29th, 2001, and for a few weeks OpenBSD-current did not contain any firewalling software.

Fortunately, in Switzerland Daniel Hartmeier was already doing some limited experiments involving kernel hacking in the networking code.

His starting point was hooking a small function of his own into the networking stack, making packets pass through it, and after a while he had started thinking about filtering. Then the license crisis happened.

IPFilter was pruned from the source tree on May 29th. The first commit of the PF code happened Sunday, June 24 2001 at 19:48:58 UTC.

A few months of rather intense activity followed, and the version of PF to be released with OpenBSD 3.0 contained a rather complete implementation of packet filtering, including network address translation.

From the looks of it, Daniel Hartmeier and the other PF developers made good use of their experience with the IPFilter code. Under any circumstances Daniel presented a USENIX 2002 paper with performance tests which show that the OpenBSD 3.1 PF performed equally well or better under stress than IPFilter on the same platform or iptables on Linux.

In addition, some tests were run on the original PF from OpenBSD 3.0. These tests showed mainly that the code had gained in efficiency from version 3.0 to version 3.1. The article which provides the details is available from Daniel Hartmeier's web, see

I have not seen comparable tests performed recently, but in my own experience and that of others, the PF filtering overhead is pretty much negligible. As one data point, the machine which gateways between Datadok's network and the world is a Pentium III 450MHz with 384MB of RAM. When I've remembered to check, I've never seen the machine at less than 96 percent 'idle' according to top.

Packet filter? Firewall?

By now I have already used some terms and concepts before I've bothered to explain them, and I'll correct that oversight shortly. PF is a packet filter, that is, code which inspects network packets at the protocol and port level, and decides what to do with them. In PF's case this code for the most part operates in kernel space, inside the network code.

PF operates in a world which consists of packets, protocols, connections and ports.

Based on where a packet is coming from or where it's going, which protocol, connection of port it is designated for, PF is able to determine where to lead the packet, or decide if it is to be let through at all.

It's equally possible to direct network traffic based on packet contents, usually referred to as application level filtering, but this is not the kind of thing PF does. We will come back later to some cases where PF will hand off these kinds of tasks to other software, but first let us deal with some basics.

We've already mentioned the firewall concept. One important feature of PF and similar software, perhaps the most important feature, is that it is able to identify and block traffic which you do not want to let into your local network or let out to the world outside. At some point the term 'firewall' was coined.


One other thing we will be talking about quit a lot are 'inner' and 'outer' addresses, 'routable' and 'non-routable' adresses. This is, at the heart of things, not directly related to firewalls or packet filtering, but due to the way the world works today, we will need to touch on it. It all comes from the time in the early 1990s when somebody started calculating the number of computers which would be connected to the Internet if the commercialization continued and the great unwashed masses of consumers were to connect at the same time.

At the time the Internet protocols were formulated, computers were usually big, expensive things which would normally serve a large number of simultaneous users, each at their own more or less dumb terminal. Under any circumstances, the only ones allowed to connect were universities and a number of companies with Pentagon contracts. Essentially 32 bit addresses of 4 octets would go an extremely long way. It would accommodate literally millions of machines, even.

Then Internet commercialization happened, and all of a sudden there were actually millions of small, inexpensive machines wanting to connect at the same time. This kind of development showed every sign of continuing and even accelerating. This meant that the smart people who had made the net work, needed to do another few pieces of work. They did a few things more or less at the same time. For one, they started working on a solution based on a larger address space - this has been dubbed IP version 6, or IPv6 for short - which uses 128 bit addresses. This has been designated as the long term solution. I thought I'd mention at this point that IPv6 support is built into OpenBSD by default, and PF has as far as I know always contained IPv6 support.

In addition, some sort of temporary solution was needed. Making the world move to addresses four times the size would take considerable time. The process is as far as we can see still pretty much in an early stage. They found a temporary solution, which consists of two parts. One part was a mechanism to offer the rest of the world 'white lies' by letting the network gateways rewrite packet addresses, the other was offered by designating some address ranges which had not been assigned earlier for use only in networks which would not communicate directly with the Internet at large. This would mean that several different machines at separate locations could have the same local IP address. But this would not matter because the address would be translated before the traffic was let out to the net at large.

If traffic with such "non routable" addresses were to hit the Internet at large, routers seeing the traffic would have a valid reason to refuse the packets to pass any further.

This is what is called "Network Address Translation", sometimes referred to as "IP masquerade" or similar. The two RFCs which define the whats and hows of this are dated 1994 and 1996 respectively [2].

There may be a number of reasons to use the so called "RFC 1918 addresses", but traditionally and historically the main reason has been that official addresses are either not available or practical.

PF today

At this point, we have covered a bit of background. Some years have passed since 2001, and PF in its present OpenBSD 3.8 form is a packet filter which is capable of doing quite a few things, if you want it to.

For one thing, PF classifies packets based on protocol, port, packet type, source or destination address. With a reasonable degree of certainty it is also able to classify packets based on source operating system.

And even if NAT is not a necessary part of a packet filter, for practical reasons it's nice that the address rewriting logic is handled somewhere nearby. Consequently, PF contains NAT logic as well.

PF is able - based on various combinations of protocol, port and other data - to direct traffic to other destinations than those designated by the sender, for example to a different machine or for further processing by a program such as a daemon listening at a port, locally or on a different machine.

Before PF was written, OpenBSD already contained the altq code to handle load balancing and traffic shaping. After a while, altq was integrated with PF. Mainly for practical reasons.

As a result of this, all those features are available to you via one single, essentially human readable configuration file, which is usually called pf.conf, stored in the /etc/ directory.

This is now available as a part of the base system on OpenBSD, on FreeBSD where PF from version 5.3 is one of three firewalling systems to be loaded at will, in NetBSD and DragonFlyBSD. The last two systems I have not had the resources to play much with myself. Something about having both hardware and time available at the same time. Anyway all indications are that only very minor details vary between these systems as far as PF is involved.

BSD vs Linux - Configuration

I assume some in the audience today are more familiar with configuring Linux or other systems than with BSD, so I'll briefly mention a few points about BSD configuration.

BSD network interfaces are not labeled eth0 and so on. The interfaces are assigned names which equal the driver name plus a sequence number, making 3Com cards using the xl driver appear as xl0, xl1, and so on, while Intel cards are likely to end up as em0, em1, SMC cards as sn0, and so on.

In general, the BSDs are organized to read the configuration from /etc/rc.conf, which is read by the /etc/rc script at startup. OpenBSD recommends using /etc/rc.conf.local for local customizations, since rc.conf contains the default values, while FreeBSD uses /etc/defaults/rc.conf to store the default settings, making /etc/rc.conf the correct place to make changes.

PF is configured by editing the /etc/pf.conf file and by using the pfctl command line tool. The pfctl application has a large number of options. We will take a closer look at some of them today.

In case you are wondering, there are web interfaces available for admin tasks, but they are not parts of the base system. The PF developers are not hostile to these things, but rather have not seen any graphical interface to PF configuration which without a doubt is preferable to pf.conf in a text editor, backed up with pfctl invocations and a few unix tricks.

Simplest possible setup (OpenBSD)

This brings us, finally, to the practical point of actually configuring PF in the simplest possible setup. We'll deal with a single machine which will communicate with a network which may very well be the Internet.

In order to start PF, as previously mentioned, you need to tell rc that you want the service to start. On OpenBSD, this is done in /etc/rc.conf.local, with the magical line

pf=YES                # enable PF

quite simply. In addition, you may if you like specify the file where PF will find its rules.

pf_rules=/etc/pf.conf # specify which file contains your rules

At the next startup, PF will be enabled. You can verify this by looking for the message PF enabled on the console. The /etc/pf.conf which comes out of a normal install of OpenBSD, FreeBSD or NetBSD contains a number of useful suggestions, but they're all commented out.

Then you really do not need to restart your machine in order to enable PF. You can do this just as easily by using pfctl. We really do not want to reboot for no good reason, so we type the command

peter@skapet:~$ sudo pfctl -e ; sudo pfctl -f /etc/pf.conf

which enables PF[3][4]. At this point we do not have a rule set, which means that PF does not actually do anything.

It is probably worth noting that if you reboot your machine at this point, the rc script on OpenBSD at least will enable a default rule set, which is in fact loaded before any of the network interfaces are enabled.

This default rule set is designed as a safety measure in case your gateway boots with an invalid configuration. It lets you log in and clean up whichever syntax error caused your rule set not to load. The default rule set allows a basic set of services: ssh from anyhere, basic name resolution and NFS mounts.

Some early versions of PF ports elsewhere appear to have neglected to bring the default rules with them.

Simplest possible setup (FreeBSD)

On FreeBSD you need a little more magic in your /etc/rc.conf, specifically according to the FreeBSD Handbook[5]

pf_enable="YES"                 # Enable PF (load module if required)
pf_rules="/etc/pf.conf"         # rules definition file for PF
pf_flags=""                     # additional flags for pfctl startup
pflog_enable="YES"              # start pflogd(8)
pflog_logfile="/var/log/pflog"  # where pflogd should store the logfile
pflog_flags=""                  # additional flags for pflogd startup

On FreeBSD, PF by default is compiled as a kernel loadable module. This means that you should be able to get started with $ sudo kldload pf, followed by $ sudo pfctl -e to enable PF. Assuming you have put these lines in your /etc/rc.conf, you can use the PF rc script to start PF:

$ sudo /etc/rc.d/pf start

Simplest possible setup (NetBSD)

On NetBSD 2.0 and newer PF is available as a loadable kernel module, installed via packages as pkgsrc/security/pflkm or compiled into a static kernel configuration. In NetBSD 3.0 onwards, PF is part of the base system.

If you want to enable PF in your kernel configuration (rather than loading the kernel module), you add these lines to your kernel configuration:

pseudo-device  pf		# PF packet filter
pseudo-device  pflog		# PF log interface

In /etc/rc.conf you need the lines


to enable PF and the PF log interface, respectively.

If you installed the module, you load it with NetBSD$ sudo modload /usr/lkm/pf.o, followed by NetBSD$ sudo pfctl -e to enable PF. Alternatively, you can run the rc scripts, NetBSD$ sudo /etc/rc.d/pf start to enable PF and NetBSD$ sudo /etc/rc.d/pflogd start to enable the logging.

To load the module automatically at startup, add the following line to /etc/lkm.conf:

/usr/lkm/pf.o - - - - AFTERMOUNT

If it's still all correct at this point, you are ready to create your first PF rule set.

First rule set - single machine

This is the simplest possible setup, for a single machine which will not run any services, and which will talk to one network which may be the Internet. For now, we will use a /etc/pf.conf which looks like this:

block in all
pass out all keep state

that is, deny any incoming traffic, allow traffic we make ourselves, and retain state information on our connections. Keeping state information allows return traffic for all connections we have initiated to pass back to us. This is something you do if this is a machine you know you can trust. If you are ready to use the rule set, you load it with

$ sudo pfctl -e ; sudo pfctl -f /etc/pf.conf

Slightly stricter

For a slightly more structured and complete setup, we start by denying everything and then allowing only those things we know that we need[6]. This gives us the opportunity to introduce two of the features which make PF such a wonderful tool - lists and macros.

We'll make some changes to /etc/pf.conf, starting with

block all

Then we back up a little. Macros need to be defined before use:

tcp_services = "{ ssh, smtp, domain, www, pop3, auth, pop3s }"
udp_services = "{ domain }"

Now we've demonstrated several things at once - what macros look like, we've shown that macros may be lists, and that PF understands rules using port names equally well as it does port numbers. The names are the ones listed in /etc/services. This gives us something to put in our rules, which we edit slightly to look like this:

block all
pass out proto tcp to any port $tcp_services keep state
pass proto udp to any port $udp_services keep state

At this point some of us will point out that UDP is stateless, but PF actually manages to maintain state information despite this. When you ask a name server about a domain name, you probably want to receive its answer.

Since we've made changes to our pf.conf, we load the new rules:

peter@skapet:~$ sudo pfctl -f /etc/pf.conf

and the new rules apply. If there are no syntax errors, pfctl will not output any messages during the rule load. The -v flag will produce more verbose pfctl output.

If you have made extensive changes to your rule set, you may want to check the rules before attempting to load them. The command to do this is, pfctl -nf /etc/pf.conf. The -n option causes the rules to be interpreted only without loading the rules. This gives you an opportunity to correct any errors. Under any circumstances the last valid rule set loaded will be in force until you either disable PF or load a new rule set.

Statistics from pfctl

You may want to check that PF is actually running, and perhaps at the same time look at some statistics. pfctl offers a number of different types of information if you use pfctl -s, adding the type of information you want to display. The following example is taken from my home gateway while I was preparing the this lecture:

peter@skapet:~$ sudo pfctl -s info
Status: Enabled for 17 days 00:24:58          Debug: Urgent

Interface Stats for xl0               IPv4             IPv6
  Bytes In                      9257508558                0
  Bytes Out                      551145119              352
  Packets In
    Passed                         7004355                0
    Blocked                          18975                0
  Packets Out
    Passed                         5222502                3
    Blocked                             65                2

State Table                          Total             Rate
  current entries                       15               
  searches                        19620603           13.3/s
  inserts                           173104            0.1/s
  removals                          173089            0.1/s
  match                             196723            0.1/s
  bad-offset                             0            0.0/s
  fragment                              22            0.0/s
  short                                  0            0.0/s
  normalize                              0            0.0/s
  memory                                 0            0.0/s
  bad-timestamp                          0            0.0/s
  congestion                             0            0.0/s
  ip-option                             28            0.0/s
  proto-cksum                          325            0.0/s
  state-mismatch                       983            0.0/s
  state-insert                           0            0.0/s
  state-limit                            0            0.0/s
  src-limit                             26            0.0/s
  synproxy                               0            0.0/s

The first line here indicates that PF is enabled and has been running for for a little more than two weeks, which is equal to the time since I upgraded to what was then the latest 3.9-beta snapshot. pfctl -s all provides highly detailed information. Try it and have a look, and while there, look into some of the other pfctl options. man 8 pfctl gives you full information.

At this point you have a single machine which should be able to communicate rather well and in a reasonably secure fashion with other internet connected machines.

A few things are still missing. For example, you probably want to let at least some ICMP and UDP traffic through, if nothing else for your own troubleshooting needs.

And even though more modern and more secure options are available, you will probably be required to handle the ftp service.

We will return to these items shortly.

Simple gateway with NAT

At this point we finally move on to the more realistic or at least more common setups, where the machine with a firewall configured also acts as a gateway for at least one other machine. The other machines on the inside may of course also run firewall software, but even if they do, it does not affect what we are interested in here to any significant degree.

Gateways and the pitfalls of in, out and on

In the single machine setup, life is relatively simple. Traffic you create should either pass or not out to the rest of the world, and you decide what you let in from elsewhere. When you set up a gateway, your perspective changes. You go from the "me versus the network out there" setting to "I am the one who decides what to pass to or from all the networks I am connected to". The machine has several, or at least two, network interfaces, each connected to a separate net.

Now it's very reasonable to think that if you want traffic to pass from the network connected to $int_if to hosts on the network connected to $ext_if, you will need a rule like

pass in on $int_if from $int_if:network to $ext_if:network \
     port $ports keep state

which keeps track of states as well. However, one of the most common and most complained-about mistakes in firewall configuration is not realizing that the "to" keyword does not in itself guarantee passage all the way there. The rule we just wrote only lets the traffic pass in to the gateway on the internal interface. To let the packets get a bit further you would need a matching rule which says

pass out on $ext_if from $int_if:network to $ext_if:network \
     port $ports keep state

These rules will work, but they will not necessarily achieve what you want.

If there are good reasons why you need to have rules which are this specific in your rule set, you know you need them and why. For the basic gateway configurations I'll be dealing with here, what you really want to use is probably a rule which says

pass from $int_if:network to any port $ports keep state

to let your local net access the Internet and leave the detective work to the antispoof and scrub code. They are both pretty good these days, and we will get back to them later. For now we just accept the fact that for simple setups, interface bound rules with in/out rules tend to add more clutter than they are worth to your rule sets.

For a busy network admin, a readable rule set is a safer rule set.

For the remainder of this tutorial, with some exceptions, we will keep the rules as simple as possible for readability.

Setting up

We assume that the machine has acquired another network card or at any rate you have set up a network connection from your local network, via PPP or other means. We will not consider the specific interface configurations. For the discussion and examples below, only the interface names will differ between a PPP setup and an Ethernet one, and we will do our best to get rid of the actual interface names as quickly as possible.

First, we need to turn on gatewaying in order to let the machine forward the network traffic it receives on one interface to other networks via a separate interface. Initially we will do this on the command line with sysctl, for traditional IP version four

# sysctl net.inet.ip.forwarding=1

and if we need to forward IP version six traffic, the command is

# sysctl net.inet6.ip6.forwarding=1

In order for this to work once you reboot the computer at some time in the future, you need to enter these settings into the relevant configuration files.

In OpenBSD and NetBSD, you do this by editing /etc/sysctl.conf, by changing the lines you need, like this


On FreeBSD, you conventionally do the corresponding change by putting these lines in your /etc/rc.conf

gateway_enable="YES" #for ipv4
ipv6_gateway_enable="YES" #for ipv6

The net effect is identical, the FreeBSD rc script sets the two values via the sysctl command. However, a larger part of the FreeBSD configuration is centralized into the rc.conf file.

Are both of the interfaces you intend to use up and running? Use ifconfig -a, or ifconfig interface_name to find out.

If you still intend to allow any traffic initiated by machines on the inside, your /etc/pf.conf could look roughly like this[7]:

ext_if = "xl0" # macro for external interface - use tun0 for PPPoE
int_if = "xl1"  # macro for internal interface
# ext_if IP address could be dynamic, hence ($ext_if)
nat on $ext_if from $int_if:network to any -> ($ext_if) 
block all
pass from { lo0, $int_if:network } to any keep state

Note the use of macros to assign logical names to the network interfaces. Here 3Com cards are used, but this is the last time during this lecture we will find this of any interest whatsoever. In truly simple setups like this one, we may not gain very much by using macros like these, but once the rule sets grow somewhat larger, you will learn to appreciate the readability this adds to the rule sets

Also note the nat rule. This is where we handle the network address translation from the non-routable address inside your local net to the sole official address we assume has been assigned to you.

The parentheses surrounding the last part of the nat rule ($ext_if) serve to compensate for the possibility that the IP address of the external interface may be dynamically assigned. This detail will ensure that your network traffic runs without serious interruptions even if the external IP address changes.

On the other hand, this rule set probably allows more traffic than what you actually want to pass out of your network. Where I work, the equivalent macro is

client_out = "{ ftp-data, ftp, ssh, domain, pop3, auth, nntp, \
                https, 446, cvspserver, 2628, cvsup, 8000, 8080 }"

with the rule

pass inet proto tcp from $int_if:network to any port $client_out \
     flags S/SA keep state

This may be a somewhat peculiar selection of ports, but it's exactly what my colleagues and I need. Some of the numbered ports are needed for systems I am not allowed to discuss any further. Your needs probably differ at least in some specifics, but this should cover at least some of the more useful services.

In addition, we have a few other pass rules. We will be returning to some of the more interesting ones rather soon. One pass rule which is useful to those of us who want the ability to administer our machines from elsewhere is

pass in inet proto tcp from any to any port ssh

or for that matter

pass in inet proto tcp from any to $ext_if port ssh

whichever you like. Lastly we need to make the name service work for our clients

udp_services = "{ domain, ntp }"

supplemented with a rule which passes the traffic we want through our firewall:

pass quick inet proto { tcp, udp } to any port $udp_services keep state

Note the quick keyword in this rule. We have started writing rule sets which consist of several rules, and it is time to take a look at the relationships between the rules in a rule set. The rules are evaluated from top to bottom, in the sequence they are written in the configuration file. For each packet or connection evaluated by PF, the last matching rule in the rule set is the one which is applied. The quick keyword offers an escape from the ordinary sequence. When a packet matches a quick rule, the packet is treated according to the present rule. The rule processing stops without considering any further rules which might have matched the packet. Quite handy when you need a few isolated exceptions to your general rules.

This also takes care of ntp, which is used for time synchronization. One thing common to both protocols is that they may under certain circumstances communicate alternately over TCP and UDP.

That sad old FTP thing

The short list of real life TCP ports we looked at a few moments back contained, among other things, FTP. FTP is a sad old thing and a problem child, emphatically so for anyone trying to combine FTP and firewalls. FTP is an old and weird protocol, with a lot to not like. The most common points against it, are

All of these points make for challenges security-wise, even before considering any potential weaknesses in client or server software which may lead to security issues. These things have tended to happen.

Under any circumstances, other more modern and more secure options for file transfer exist, such as sftp or scp, which feature both authentication and data transfer via encrypted connections. Competent IT professionals should have a preference for some other form of file transfer than FTP.

Regardless of our professionalism and preferences, we are all too aware that at times we will need to handle things we would prefer not to. In the case of FTP through firewalls, the main part of our handling consists of redirecting the traffic to a small program which is written specifically for this purpose.

FTP through NAT: ftp-proxy

NoteOpenBSD 3.8 or earlier only

In November 2005, the old ftp-proxy (/usr/libexec/ftp-proxy) was replaced in OpenBSD-current with the new ftp-proxy, which lives in /usr/sbin. This is the software which is included in OpenBSD 3.9 onwards. See the Section called ftp-proxy, new style for details.

ftp-proxy is a part of the base system on OpenBSD and other systems which offer PF, and is usually called via the inetd "super server" via an appropriate /etc/inetd.conf entry.

The line quoted here specifies that ftp-proxy runs in NAT mode on the loopback interface, lo0: stream tcp nowait root /usr/libexec/ftp-proxy \
  ftp-proxy -n

This line is by default in your inetd.conf, commented out with a # character at the beginning of the line. To enable your change, you restart inetd.

On FreeBSD, NetBSD and other rcng based BSDs you do this with the command

FreeBSD$ sudo /etc/rc.d/inetd restart

or equivalent. Consult man 8 inetd if you are unsure.

The OpenBSD rc system is a tad more traditional, and the command you need is

OpenBSD$ sudo kill -HUP `cat /var/run/` 

At this point inetd is running, with your new settings.

Now for the actual redirection. Redirection rules and NAT rules fall into the same rule class. These rules may be referenced directly by other rules, and filtering rules may depend on these rules. Logically, rdr and nat rules need to be defined before the filtering rules.

We insert our rdr rule immediately after the nat rule in our /etc/pf.conf

rdr on $int_if proto tcp from any to any port ftp -> \
         port 8021

In addition, the redirected traffic must be allowed to pass. We achive this with

pass in on $ext_if inet proto tcp from port ftp-data to ($ext_if) \
    user proxy flags S/SA keep state

Save pf.conf, then load the new rules with

$ sudo pfctl -f /etc/pf.conf

At this point you will probably have users noticing that FTP works before you get around to telling them what you've done.

This example assumes you are using NAT on a gateway with non routable addresses on the inside.

FTP through pf with routable addresses: ftpsesame, pftpx and ftp-proxy!

In cases where the local network uses official, routable address inside the firewall, I must confess I've had trouble making ftp-proxy work properly. When I'd already spent too much time on the problem, I was rather relieved to find a solution to this specific problen offered by a friendly Dutchman called Camiel Dobbelaar in the form of a daemon called ftpsesame.

Local networks using official addresses inside a firewall are apparently rare enough that I'll skip over any further treatment. If you need this and you are running OpenBSD or earlier or one of the other PF enabled operating systems, you could do worse than downloading ftpsesame from Sentia at ftpsesame hooks into your rule set via an anchor, a named sub-ruleset. The documentation consists of a man page with examples which you can more likely than not simply copy and paste.

ftpsesame never made it into the base system, and Camiel went on to write a new solution to the same set of problems.

The new program, at first called pftpx, 0.8 is available from A further developed version, suitably renamed as the new ftp-proxy, will be part of OpenBSD 3.9 as /usr/sbin/ftp-proxy.

ftp-proxy, new style

NoteFor OpenBSD 3.9 and newer

If you are upgrading to OpenBSD 3.9 or working from a fresh 3.9 or newer install, this is the ftp-proxy version to use.

Just like its predecessor, the pftpx successor ftp-proxy configuration is mainly a matter of cut and paste from the man page.

If you are upgrading to the new ftp-proxy from an earlier version, you need remove the ftp-proxy line from your inetd.conf file and restart inetd.

Next, enable ftp-proxy by adding the following line to your /etc/rc.conf.local or /etc/rc.conf


You can start the proxy manually by running /usr/sbin/ftp-proxy if you like.

Moving on to the pf.conf file, you need two anchor definitions in the NAT section:

nat-anchor "ftp-proxy/*"
rdr-anchor "ftp-proxy/*"

Both are needed, even if your setup does not use NAT. If you are migrating from a previous version, your rule set probably contains the appropriate redirection already. If not, you add it:

rdr pass on $int_if proto tcp from any to any port ftp -> \
         port 8021

Moving on down to the filtering rules, you add an anchor for the proxy to fill in,

anchor "ftp-proxy/*"

and finally a pass rule to let the packets pass from the proxy to the rest of the world

pass out proto tcp from $proxy to any port 21 keep state

where $proxy expands to the address the proxy daemon is bound to.

This example covers the simple setup with clients who need to contact FTP servers elsewhere. For other variations and more complicated setups, see the ftp-proxy man page.

Making your network troubleshooting friendly

Making your network troubleshooting friendly is a potentially large subject. At most times, the debugging or troubleshooting friendliness of your TCP/IP network depends on how you treat the Internet protocol which was designed specifically with debugging in mind, the Internet Control Message Protocol, or ICMP as it is usually abbreviated.

ICMP is the protocol for sending and receiving control messages between hosts and gateways, mainly to provide feedback to a sender about any unusual or difficult conditions en route to the target host.

There is a lot of ICMP traffic which usually just happens in the background while you are surfing the web, reading mail or transferring files. Routers (you are aware that you are building one, right?) use ICMP to negotiate packet sizes and other transmission parameters in a process often referred to as path MTU discovery.

You may have heard admins referring to ICMP as either 'just evil', or, if their understanding runs a little deeper, 'a necessary evil'. The reason for this attitude is purely historical. The reason can be found a few years back when it was discovered that several operating systems contained code in their networking stack which could make a machine running one of the affected systems crash and fall over, or in some cases just do really strange things, with a sufficiently large ICMP request.

One of the companies which was hit hard by this was Microsoft, and you can find rather a lot of material on the 'ping of death' bug by using your favourite search engine. This all happened in the second half of the 1990s, and all modern operating systems, at least the ones we can read, have thoroughly sanitized their network code since then. At least that's what we are lead to believe.

One of the early workarounds was to simply block either all ICMP traffic or at least ICMP ECHO, which is what ping uses. Now these rule sets have been around for roughly ten years, and the people who put them there are still scared.

Then, do we let it all through?

The obvious question then becomes, if ICMP is such a good and useful thing, should we not let it all through, all the time? The answer is, 'It depends'.

Letting diagnostic traffic pass unconditionally of course makes debugging easier, but it also makes it relatively easy for others to extract information about your network. That means that a rule like

pass inet proto icmp from any to any

might not be optimal if you want to cloak the internal workings of your network in a bit of mystery. In all fairness it should also be said that you might find some ICMP traffic quite harmlessly riding piggyback on your keep state rules.

The easy way out: The buck stops here

The easiest solution could very well be to let all ICMP traffic from your local net through and let probes from elsewhere stop at your gateway:

pass inet proto icmp icmp-type $icmp_types from $int_if:network \
          to any keep state
pass inet proto icmp icmp-type $icmp_types from any to $ext_if \
          keep state

Stopping probes at the gateway might be an attractive option anyway, but let us have a look at a few other options which will show you some of PF's flexibility.

Letting ping through

The rule set we have developed so far has one clear disadvantage: common troubleshooting commands such as ping and traceroute will not work. That may not matter too much to your users, and since it was the ping command which scared people into filtering or blocking ICMP traffic in the first place, there are apparently some people who feel we are better off without it. If you are in my perceived target audience, you will be rather fond of having those troubleshooting tools avalable. With a couple of small additions to the rule set, they will be. ping uses ICMP, and in order to keep our rule set tidy, we start by defining another macro:

icmp_types = "echoreq"

and a rule which uses the definition,

pass inet proto icmp all icmp-type $icmp_types keep state

You may need more or other types of ICMP packets to go through, and you can then expand icmp_types to a list of those packet types you want to allow.

Helping traceroute

traceroute is another command which is quite useful when your users claim that the Internet isn't working. By default, Unix traceroute uses UDP connections according to a set formula based on destination. The rule below works with the traceroute command on all unixes I've had access to, including GNU/Linux:

# allow out the default range for traceroute(8):
# "base+nhops*nqueries-1" (33434+64*3-1)
pass out on $ext_if inet proto udp from any to any \
             port 33433 >< 33626 keep state

Experience so far indicates that traceroute implementations on other operating systems work roughly the same. Except, of course, Microsoft Windows. On that platform, TRACERT.EXE uses ICMP ECHO for this purpose. So if you want to let Windows traceroutes through, you only need the first rule. Unix traceroutes can be instructed to use other protocols as well, and will behave remarkably like its Microsoft counterpart if you use its -I command line option. You can check the traceroute man page (or its source code, for that matter) for all the details.

Under any circumstances, this solution was lifted from an openbsd-misc post. I've found that list, and the searchable list archives (accessible among other places from, to be a very valuable resource whenever you need OpenBSD or PF related information.

Path MTU discovery

The last bit I will remind you about when it comes to troubleshooting is the 'path MTU discovery'. Internet protocols are designed to be device independent, and one consequence of device independence is that you can not always predict reliably what the optimal packet size is for a given connection. The main constraint on your packet size is called the Maximum Transmission Unit, or MTU, which sets the upper limit on the packet size for an interface. The ifconfig command will show you the MTU for your network interfaces.

The way modern TCP/IP implementations work, they expect to be able to determine the right packet size for a connection through a process which simply puts involves sending packets of varying sizes with the 'Do not fragment' flag set, expecting an ICMP return packet indicating "type 3, code 4", when the upper limit has been reached. Now you don't need to dive for the RFCs right away. Type 3 means "destination unreachable", while code 4 is short for "fragmentation needed, but the do not fragment flag is set". So if your connections to networks which may have other MTUs than your own seem sub-optimal, and you do not need to be that specific, you could try changing your list of ICMP types slightly to let the Destination unreachable packets through, too:

icmp_types = "{ echoreq, unreach }"

as we can see, this means we do not need to change the pass rule itself:

pass inet proto icmp all icmp-type $icmp_types keep state

PF lets you filter on all variations of ICMP types and codes, and if you want to delve into what to pass and not of ICMP traffic, the list of possible types and codes are documented in the icmp(4) and icmp6(4) man pages. The background information is available in the RFCs[8] .

Hygiene: block-policy, scrub and antispoof

Our gateway does not feel quite complete without a few more items in the configuration which will make it behave a bit more sanely towards hosts on the wide net and our local network.

block-policy is an option which can be set in the options part of the ruleset, which precedes the redirection and filtering rules. This option determines which feedback, if any, PF will give to hosts which try to create connections which are subsequently blocked. The option has two possible values, drop which drops blocked packets with no feedback, and return which returns with status codes such as Connection refused or similar.

The correct strategy for block policies has been the subject of rather a lot of discussion. We choose to play nicely and instruct our firewall to issue returns:

set block-policy return

scrub is a keyword which enables network packet normalization, causing fragmented packets to be assembled and removing ambiguity. Enabling scrub provides a measure of protection against certain kinds of attacks based on incorrect handling of packet fragments. A number of supplementing options are available, but we choose the simplest form which is suitable for most configurations.

scrub in all

Some services, such as NFS, require some specific fragment handling options. This is extensively documented in the PF user guide and man pages provide all the information you could need.

antispoof is a common special case of filtering and blocking. This mechanism protects against activity from spoofed or forged IP addresses, mainly by blocking packets appearing on interfaces and in directions which are logically not possible. We specify that we want to weed out spoofed traffic coming in from the rest of the world and any spoofed packets which, however unlikely, were to originate in our own network:

antispoof for $ext_if
antispoof for $int_if

This completes our simple NATing firewall for a small local network.

A web server and a mail server on the inside

Time passes, and needs change. Rather frequently, a need to run externally accessible services develops. This quite frequently becomes just a little harder because externally visible addresses are either not available or too expensive, and running several other services on a machine which is primarily a firewall is not a desirable option.

The redirection mechanisms in PF makes it relatively easy to keep servers on the inside. If we assume that we need to run a web server which serves up data in clear text (http) and encrypted (https) and in addition we want a mail server which sends and receives e-mail while letting clients inside and outside the local network use a number of well known submission and retrieval protocols, the following lines may be all that's needed in addition to the rule set we developed earlier:

webserver = ""
webports = "{ http, https }"
emailserver = ""
email = "{ smtp, pop3, imap, imap3, imaps, pop3s }"

rdr on $ext_if proto tcp from any to $ext_if port \
       $webports -> $webserver
rdr on $ext_if proto tcp from any to $ext_if port \
       $email -> $emailserver

pass proto tcp from any to $webserver port $webports \
   flags S/SA synproxy state
pass proto tcp from any to $emailserver port $email \
   flags S/SA synproxy state
pass proto tcp from $emailserver to any port smtp \
   flags S/SA synproxy state

Notice the flag 'synproxy' in the new rules. This means that PF will handle the connection setup (three way handshake) on behalf of your server or client before handing the connection over to the application. This provides a certain amount of protection against certain types of attacks.

Rule sets for configurations with DMZ networks isolated behind separate network interfaces and in some cases services running on alternative ports will not necessarily be much different from this one.

Taking care of your own - the inside

Everything I've said so far is excellent and correct as long as all you are interested in is getting traffic from hosts outside your local net to reach your servers.

If you want the hosts in your local net to be able to use the services on these machines, you will soon see that the traffic originating in your local network most likely never reaches the external interface. The external interface is where all the redirection and translation happens, and consequently the redirections do not quite work from the inside. The problem is common enough that the PF documentation lists four different solutions to the problem.[9]

The options listed in the PF user guide are

  • 'Split horizon' DNS, which means configuring your name service to provide one set of replies for requests originating in the local net and a different one for requests from elsewhere

  • proxying using software such as nc (NetCat)

  • treating the local net as a special case for redirection and NAT.

    We will be looking into this option below.

  • Or simply moving your servers to a separate network, aka a 'DMZ', with only minor changes to your PF rules.

We need to intercept the network packets originating in the local network and handle those connections correctly, making sure any returning traffic is directed to the communication partner who actually originated the connection.

Returning to our previous example, we achieve this by adding these special case rules::

rdr on $int_if proto tcp from $int_if:network to $ext_if \
       port $webports -> $webserver
rdr on $int_if proto tcp from $int_if:network to $ext_if \
       port $email -> $emailserver
no nat on $int_if proto tcp from $int_if:network to $int_if:network
nat on $int_if proto tcp from $int_if:network to $webserver \
       port $webports -> $int_if 
nat on $int_if proto tcp from $int_if:network to $emailserver \
       port $email -> $int_if 

Tables make your life easier

By this time you may be thinking that this gets awfully static and rigid. There will after all be some kinds of data which are relevant to filtering and redirection at a given time, but do not deserve to be put into a configuration file! Quite right, and PF offers mechanisms for handling these situations as well. Tables are one such feature, mainly useful as lists which can be manipulated without needing to reload the entire rule set, and where fast lookups are desirable. Table names are always enclosed in < >, like this:

table <clients> {, ! }

here, the network is part of the table, except the address, which is excluded using the ! operator (logical NOT). It is also possible to load tables from files where each item is on a separate line, such as the file /etc/clients

which in turn is used to initialize the table in /etc/pf.conf:

table <clients> persist file /etc/clients

Then, for example, you can change one of our earlier rules to read

pass inet proto tcp from <clients> to any port $client_out \
         flags S/SA keep state

to manage outgoing traffic from your client computers. With this in hand, you can manipulate the table's contents live, such as

$ sudo pfctl -t clients -T add 192.168.1/16

Note that this changes the in-memory copy of the table only, meaning that the change will not survive a power failure or other reboot unless you arrange to store your changes.

You might opt to maintain the on-disk copy of the table using a cron job which dumps the table content to disk at regular intervals, using a command such as pfctl -t clients -T show >/etc/clients. Alternatively, you could edit the /etc/clients file and replace the in-memory table contents with the file data:

$ sudo pfctl -t clients -T replace -f /etc/clients

For operations you will be performing frequently, you will sooner or later end up writing shell scripts for tasks such as inserting or removing items or replacing table contents. The only real limitations lie in your own needs and your creativity.

We will be returning to other handy uses of tables shortly.


Up to now we have not mentioned much about logging. PF provides the opportunity to log exactly what you want by adding the 'log' keyword to the rules you want logged. You may want to limit the amount of data a bit by specifying one interface where the logging is to be done. You do this by adding

set loginterface $ext_if

and then editing the rules you want to log, such as

pass out log from <client> to any port $email \
         label client-email keep state

This causes the traffic to be logged in a binary format which is really only intended to be used as tcpdump input.

The label part creates a new set of counters for various statistics for the rule. This can be quite convenient if you are invoicing others for bandwidth use, for example.

Taking a peek with tcpdump

Once you have enabled logging in one or more rules, PF logs via the pflog0 interface, and stores binary log data in the log file /var/log/pflog. The log file is useful for a permanent record and for those cases where you want to periodically convert some of the data to other formats. However, if you want to look at your traffic in real time, you can tell tcpdump to look at the pflog0 log interface instead.

Here is what the output from a couple of log rules can look like on a lazy Thursday afternoon:

peter@skapet:~$ sudo tcpdump -n -e -ttt -i pflog0
tcpdump: WARNING: pflog0: no IPv4 address assigned
tcpdump: listening on pflog0, link-type PFLOG
Feb 16 16:43:20.152187 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 16:48:26.073244 rule 27/(match) pass in on xl0: > icmp: echo request
Feb 16 16:49:09.563448 rule 0/(match) block in on xl0: > [|tcp]
Feb 16 16:49:14.601022 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 16:53:10.110110 rule 0/(match) block in on xl0: > [|icmp]
Feb 16 16:55:54.818549 rule 27/(match) pass in on xl0: > icmp: echo request
Feb 16 16:57:55.577782 rule 27/(match) pass in on xl0: > icmp: echo request
Feb 16 17:01:27.108404 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 17:11:02.137310 rule 0/(match) block in on xl0: > [|tcp]
Feb 16 17:14:05.739403 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 17:14:08.715163 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 17:14:09.308355 rule 0/(match) block in on xl0: > [|tcp] (DF)
Feb 16 17:19:01.853730 rule 27/(match) pass in on xl0: > icmp: echo request

The PF User Guide has a section devoted to logging which contains a number of very useful suggestions. Combined with among other things the tcpdump man pages, you should be able to extract any log data you will find useful.

But there are limits (an anecdote)

It might feel tempting at first to put something like this in

block log all

- just to make sure you don't miss anything.

The PF user guide contains a detailed description of how to make PF log to a human readable text format via syslog, and this does sound rather attractive. I went through the procedure described there when I set up my first PF configuration at work, and the experience sums up rather neatly: Logging is useful, but by all means, be selective. After a little more than an hour the PF text log file had grown to more than a gigabyte, on a machine with less than ten gigabytes of disk space total.

The explanation is simply that even in a rather unexciting Internet backwater, at the far end of an unexceptional ADSL line there's still an incredible amount of uncontrolled Windows traffic such as file sharing and various types of searches trying to get to you. The Windows boxes on the inside probably weren't totally quiet either. At any rate: put some sensible limit on what you log, or make arrangements for sufficient disk space, somewhere.

Keeping an eye on things with pftop

If you are interested in keeping an eye on what passes in to and out of your network, Can Erkin Acar's pftop is a very useful tool. The name is a strong hint at what it does - pftop shows a running snapshot of your traffic in a format which is strongly inspired by top(1):

pfTop: Up State 1-21/67, View: default, Order: none, Cache: 10000       19:52:28

PR   DIR SRC                  DEST                 STATE   AGE   EXP  PKTS BYTES
tcp  Out      9:9     28    67    29  3608
tcp  In        4:4     15 86400    30  1594
tcp  In       10:10   418    75   810 44675
tcp  In      4:4    146 86395   158 37326
tcp  In      4:4    193 86243   131 21186
tcp  In      9:9    154    59    11  1570
tcp  In      4:4      4 86397     6  1370
tcp  In      9:9     28    67    29  3608
tcp  Out      9:9    214     0     9  1490
tcp  Out      4:4     64 86337     7  1410
udp  Out        2:1     36     0     2   235
udp  In      1:2     36     0     2   219
udp  In      1:2     36     0     2   255
udp  In      1:2     36     0     2   271

Your connections can be shown sorted by a number of different criteria, among others by PF rule, volume, age and so on.

This program is not in the base system itself, but is in ports on OpenBSD and FreeBSD both as /usr/ports/sysutils/pftop, on NetBSD via pkgsrc as sysutils/pftop.

Invisible gateway - bridge

A bridge in our context is a machine with two or more network interfaces, located in between the Internet and one or more internal networks, and the network interfaces are not assigned IP addresses. If the machine in question runs OpenBSD or a similarly capable operating system, it is still able to filter and redirect traffic. The advantage of such a setup is that attacking the firewall itself is more difficult. The disadvantage is that all admin tasks must be performed at the firewall's console, unless you configure a network interface which is reachable via a secured network of some kind, or even a serial console.

The exact method for configuring bridges differs in some details between the operating systems. Below is a short recipe for use on OpenBSD, which for good measure blocks all non-Internet protocol traffic. Setting up a bridge with two interfaces:






  add xl0 add xl1 blocknonip xl0 blocknonip xl1 up


ext_if = xl0
int_if  = xl1
interesting-traffic = { ... }
block all
pass quick on $ext_if all
pass log on $int_if from $int_if to any port $interesting-traffic \
     keep state

Significantly more complicated setups are possible. Experienced bridgers recommend picking one of the interfaces to perform all filtering and redirection. All packets pass through PF's view twice, making for potentially extremely complicated rules.

In addition, the OpenBSD brconfig command offers its own set of filtering options in addition to other configuration options. The bridge(4) and brconfig(8) man pages offer further information.

FreeBSD uses a slightly different set of commands to configure bridges, while the NetBSD PF implementation does not support bridging.

Directing traffic with ALTQ

ALTQ - short for ALTernate Queueing - is a very flexible mechanism for directing network traffic which lived a life of its own before getting integrated into PF. Altq was another one of those things which were integrated into PF because of the additional convenience it offered when integrated.

Altq uses the term queue about the main traffic control mechanisms. Queues are defined with a defined amount of bandwidth or a specific part of available bandwidth, where a queue can be assigned subqueues of various types.

To complete the picture, you write filtering rules which assign packets to specified queues or a selection of subqueues where packets pass according to specified criteria.

Queues are created with one of several queue disciplines. The default queue discipline without ALTQ is FIFO (first, in first out). A slightly more interesting discipline is the class based discipline (CBQ), which in practical terms means you define the queue's bandwidth as a set amount of data per second, as a percentage or in units of kilobits, megabits and so on, with an additional priority as an option, or priority based (priq), where you assign priority only. Priorities can be set at 0 to 7 for cbq queues, 0 to 15 for priq queues, with a higher value assigning a higher priority and preferential treatment. In addtion, the hierarcical queue algoritm "Hierarchical Fair Service Curve" or HFSC is available. Briefly, a simplified syntax is

altq on interface type [options ... ] main_queue { sub_q1, sub_q2 ..}
  queue sub_q1 [ options ... ]
  queue sub_q2 [ options ... ]
pass [ ... ] queue sub_q1
pass [ ... ] queue sub_q2

If you will be using these features in you own rule sets, you should under any circumstances read the pf.conf man page and the PF user guide. These documents offer a very detailed and reasonably well laid out explanation of the syntax and options.[10] [11]

ALTQ - allocation by percentage

Now for a real example, which I for all practical purposes swiped from The queues are set up on the external interface. This is probably the more common approach, since the limitations on bandwidth are usually more severe on the external interface. In principle, however, allocating queues and running traffic shaping can be done on any network interface. Here, the setup includes a cbq queue for a total bandwidth of 640KB with six sub queues.

altq on $ext_if cbq bandwidth 640Kb queue { def, ftp, udp, http, \
         ssh, icmp }
queue def bandwidth 18% cbq(default borrow red)
queue ftp bandwidth 10% cbq(borrow red)
queue udp bandwidth 30% cbq(borrow red)
queue http bandwidth 20% cbq(borrow red)
queue ssh bandwidth 20% cbq(borrow red) { ssh_interactive, ssh_bulk }
queue ssh_interactive priority 7
queue ssh_bulk priority 0
queue icmp bandwidth 2% cbq

We see the subqueue def with 18 percent of the bandwidth is designated as the default queue, that is any traffic not explicitly assigned to some other queue ends up here. The borrow and red keywords mean that the queue may 'borrow' bandwidth from its parent queue, while the system attempts to avoid congestion by applying the RED (Random Early Detection) algorithm. The other queues follow more or less the same pattern, up to the subqueue ssh, which itself has two subqueues with individual priorities.

Finally, the pass rules which show which traffic gets assigned to the queues, and their criteria:

pass log quick on $ext_if proto tcp from any to any port 22 flags S/SA \ 
    keep state queue (ssh_bulk, ssh_interactive)
pass in quick on $ext_if proto tcp from any to any port 20 flags S/SA  \
    keep state queue ftp
pass in quick on $ext_if proto tcp from any to any port 80 flags S/SA \
    keep state queue http
pass out on $ext_if proto udp all keep state queue udp
pass out on $ext_if proto icmp all keep state queue icmp

We can reasonably assume that this allocation meets the site's needs.

ALTQ - prioritizing by traffic type

We move on to another example, lifted from Daniel Hartmeier's web. Like quite a few of us, Daniel is on an asymmetric connection, and naturally he wanted to get better bandwidth utilization.

One symptom in particular seemed to indicate that there was room for improvement. Incoming traffic (downloads) apparently slowed down outgoing traffic.

Analyzing the data indicated that the ACK packets for each data packet transferred caused a disproportionately large slowdown, possibly due to the FIFO (First In, First Out) queue discipline in effect on the outgoing traffic.

A testable hypothesis formed - if the tiny, practically dataless ACK packets were able to slip inbetween the larger data packets, this would lead to a more efficient use of available bandwidth. The means were two queues with different priorities. The relevant parts of the rule set follows::


altq on $ext_if priq bandwidth 100Kb queue { q_pri, q_def }
queue q_pri priority 7
queue q_def priority 1 priq(default)

pass out on $ext_if proto tcp from $ext_if to any flags S/SA \
        keep state queue (q_def, q_pri)

pass in  on $ext_if proto tcp from any to $ext_if flags S/SA \
        keep state queue (q_def, q_pri)

The result was indeed better performance. Daniel's article is available from his web site at

ALTQ - handling unwanted traffic

Our last altq example is one which surfaced around the time of one of the many spam or virus storms we've seen during the last few years. It's fairly common knowledge that the machines causing these bursts of email are practically all Windows machines. PF has a fairly reliable operating system fingerprinting mechanism which detects the operating system at the other end of a network connection. One OpenBSD user got sufficiently tired of all this meaningless traffic, and posted some selected bits of his pf.conf on his blog:

altq on $ext_if cbq queue { q_default q_web q_mail }

   queue q_default cbq(default)
   queue q_web (...) 

    ## all mail limited to 1Mb/sec
    queue q_mail bandwidth 1Mb { q_mail_windows }
    ## windows mail limited to 56Kb/sec
    queue q_mail_windows bandwidth 56Kb

    pass in quick proto tcp from any os "Windows" to $ext_if port 25 \
            keep state queue q_mail_windows
    pass in quick proto tcp from any to $ext_if port 25 label "smtp" \
            keep state queue q_mail
" I can't believe I didn't see this earlier. Oh, how sweet. ... 
  Already a huge difference in my load. Bwa ha ha. "

Randal L. Schwartz, 29. january 2004,

Here all email traffic is assigned one megabit worth of bandwidth, while email traffic originating at Windows hosts get to share a subqueue of 56 kbit total. No wonder the total load went down and the blog post ends with what must be an evil chuckle.

I must confess this is something I've wanted very much to do myself, but I've never dared. A few too many of our customers have for their own reasons chosen to run their mail service on Windows, and we do like to receive most of their mail. In a little while, we'll have a look at a different PF approach which may have achieved much of the same effect.

CARP and pfsync

CARP and pfsync were two of the main new items in OpenBSD 3.5. CARP is short for Common Address Redundancy Protocol. The protocol was developed as a non patent encumbered alternative to VRRP (Virtual Router Redundancy Protocol, RFC 2281, RFC 3768), which was quite far along the track to becoming an IETF sanctioned standard, even though possible patent issued has not been resolved. The patents involved are held by Cisco, IBM and Nokia.

Both protocols are intended to ensure redundancy for essential network features, with automatic failover.

CARP is based on setting up a group of machines as one 'master' and one of more redundant 'slaves', all of which are equipped to handle a common IP address. If the master goes down, one of the slaves will inherit the IP address, and if the synchronization has been properly handled, active connections will be handed over. The handover may be authenticated using cryptographic keys.

One of the main purposes of CARP is to ensure that the network will keep functioning as usual even when a firewall or other service goes down due to errors or planned maintenance activities such as upgrades.

In the case of PF firewalls, pfsync can handle the synchronization. pfsync is a type of virtual network interface specially designed to synchronize state information between PF firewalls. pfsync interfaces are assigned to physical interfaces with ifconfig. On networks where uptime requirements are strict enough to dictate automatic failover, the number of simultaneous network connections is likely to be large enough that it will make sense to assign the pfsync network its own physical network.

I hope to explore these exciting and advanced features in a future advanced section of the tutorial. At this point, the best pfsync and CARP references are the OpenBSD FAQ, the man pages and Ryan McBride's overview article at

Wireless networks made simple

It rather tempting to say that on BSD, and on OpenBSD in particular, there's no need to 'make wireless networking simple', because it already is. Getting a wireless network running is basically not very different from getting a wired one up and running, but then of course there are some issues which turn up simply because we are dealing with radio waves and not wires. We will look briefly at some of the issues before moving on to the practical setup issues.

A little IEEE 802.11 background

Moving to wireless networks provides an opportunity to view security at various level in the networking stack from a new perspective. We look briefly at two of the basic IEEE 802.11 security mechanisms below.[12]

It goes almost without saying that you will need further security measures, such as SSH or SSL encryption, to maintain any significant level of confidentiality for your data stream.

WEP (Wireless Equivalent Privacy)

One consequence of using radio waves instead of wires is that it is comparatively easy for outsiders to capture your data in transit. The designers of the 802.11 family of wireless network standards seem to have been aware of this fact, and came up with a solution which they marketed under the name Wired Equivalent Privacy, or WEP.

WEP is a link level encryption scheme which is considered a pretty primitive homebrew among cryptography professionals. It was no great surprise that WEP encryption was reverse-engineered and cracked within a few months after the first products were released. Even though you can download tools for free to descramble WEP encoded traffic in a matter of minutes, for a variety of reasons it is still widely supported and used. You should consider network traffic protected only by WEP to be only marginally more secure than data broadcast in the clear. Then again, the token effort needed to crack into a WEP network may be sufficient to deter lazy and unsophisticated attackers.

WPA (WiFi Protected Access)

It dawned fairly quickly on the 802.11 designers that their Wired Equivalent Privacy system was not quite what it was cracked up to be, and they came up with a revised and slightly more comprehensive solution which was dubbed WiFi Protected Access, or WPA.

WPA looks better than WEP, at least on paper, but the specification is complicated enough that it is not nearly as widely supported or implemented as its designers intended. In addition WPA has also attracted its share of criticism over design issues and bugs. Combined with the familiar issues of access to documentation and hardware, free software support varies. If your project specification includes WPA, look carefully at your operating system and driver documentation.

Setting up a simple wireless network

The first part is to make sure you have a supported card and check your dmesg output to see that the driver loads and initializes the card properly[13] With a successfully configured card you should see someting like

ath0 at pci1 dev 4 function 0 "Atheros AR5212" rev 0x01: irq 11
ath0: AR5212 5.6 phy 4.1 rf5111 1.7 rf2111 2.3, ETSI1W, address 

Next, you need to configure the interface for TCP/IP. On OpenBSD, this means an /etc/hostname.ath0 roughly like this:

up media autoselect mediaopt hostap mode 11b chan 6 nwid unwiredbsd \
 nwkey 0x1deadbeef9

Note that the configuration is divided over two lines. The first line generates an ifconfig command which sets up the interface with the correct parameters for the physical wireless network, the second command, which gets executed only after the first one completes, sets the IP address. Note that we set the channel explicitly, and we enable a weak WEP encryption by setting the nwkey parameter.

The FreeBSD and NetBSD version would be to put the following lines in your /etc/rc.conf:

ifconfig_ath0="up media autoselect mediaopt hostap mode 11b chan 6 \
  nwid unwiredbsd nwkey 0x1deadbeef9"

Then you most likely want to set up dhcpd to serve addresses and other relevant network information to clients. Your clients would need an /etc/hostname.ath0 configuration of

up media autoselect mode 11b chan 6 nwid unwiredbsd nwkey 0x1deadbeef9

or on FreeBSD and NetBSD,

ifconfig_ath0="up media autoselect mode 11b chan 6 nwid unwiredbsd \
  nwkey 0x1deadbeef9"

in /etc/rc.conf.

Assuming your gateway does NAT, you will want to set up NAT for the wireless network as well, by making some small changes to your /etc/pf.conf:

air_if = "ath0"


nat on $ext_if from $air_if:network to any -> ($ext_if) static-port

You will need a similar near duplicate line for your ftp-proxy config, and include $air_if in your pass rules.

That's all there is to it. This configuration gives you a functional BSD access point, with at least token security via WEP encryption.

An open, yet tightly guarded wireless network with authpf

As always, there are other ways to configure the security of your wireless network than the one we have just seen. What little protection WEP encryption offers, security professionals tend to agree is barely enough to signal to an attacker that you do not intend to let all and sundry use your network resources.

A different approach appeared one day in my mail as a message from my friend Vegard Engen, who told me he had been setting up authpf. authpf is a user shell which lets you load PF rules on a per user basis, effectively deciding which user gets to do what. With a reasonable setup, only traffic originated by authenticated users will be let through. Vegard's annotated config follows below. His wireless network is configured without WEP encryption, preferring to handle the security side of things via PF and authpf:

Start with creating an empty /etc/authpf/authpf.conf. It needs to be there for authpf to work, but doesn't actually need any content.

The other relevant bits of /etc/pf.conf follow. First, interface macros:

wi_if = "wi0"

The use of this address will become apparent later:


The traditional authpf table

table <authpf_users> persist 

We could put the NAT part in authpf.rules, but keeping it in the main pf.conf doesn't hurt:

nat on $ext_if from $wi_if:network to any -> ($ext_if)

Redirects to let traffic reach servers on the internal net. These could be put in authpf.rules too, but since they do not actually provide access without pass rules, keeping them here won't hurt anything.

rdr on $wi_if proto tcp from any to $myaddr port $tcp_in -> $server
rdr on $wi_if proto udp from any to $myaddr port $udp_in -> $server

The next redirect sends all web traffic from non authenticated users to port 80 on $auth_web. In Vegard's setup, this is a web server which displays contact info for people who stumble onto the wireless net. In a commercial setting, this would be where you would put something which could handle credit cards and create users.

rdr on $wi_if proto tcp from ! <authpf_users> to any \
 port 80 -> $auth_web

To activate nat, binat or redirects in authpf

nat-anchor "authpf/*"
binat-anchor "authpf/*"
rdr-anchor "authpf/*"

On to the filtering rules, we start with a sensible default

block all

Other global, user independent rules would go here. Next for the authpf anchor, we make sure non-authenticated users connecting to the wireless interface get redirected to $auth_web

anchor "authpf/*" in on wi0

pass in on $wi_if inet proto tcp from any to $auth_web \
 port 80 keep state

There are three things we want anyway on the wireless interface: Name service (DNS), DHCP and SSH in to the gateway. Three rules do the trick

pass in on $wi_if inet proto udp from any port 53 keep state

pass in on $wi_if inet proto udp from any to $wi_if port 67

pass in on $wi_if inet proto tcp from any to $wi_if \
 port 22 keep state

Next up, the we define the rules which get loaded for all users who log in with their shell set to /usr/sbin/authpf. These rules go in /etc/authpf/authpf.rules,

int_if = "sis1"
ext_if = "sis0"
wi_if = "wi0"
server = ""
myaddr = "213.187.n.m"

# Services which live on the internal network 
# and need to be accessible
tcp_services = "{ 22, 25, 53, 80, 110, 113, 995 }"
udp_services = "{ 53 }"
tcp_in = " { 22, 25, 53, 80, 993, 2317, pop3}"
udp_in = "{ 53 }"

# Pass traffic to elsewhere, that is the outside world
pass in on $wi_if inet from <authpf_users> to ! $int_if:network \
   keep state

# Let authenticated users use services on 
# the internal network.

pass in on $wi_if inet proto tcp from <authpf_users> to $server \
  port $tcp_in keep state
pass in on $wi_if inet proto udp from <authpf_users> to $server \
  port $udp_in keep state

# Also pass to external address. This means you can access 
# internal services on external addesses.

pass in on $wi_if inet proto tcp from <authpf_users> to $myaddr \
    port $tcp_in keep state
pass in on $wi_if inet proto udp from <authpf_users> to $myaddr \
    port $udp_in keep state

At this point we have an open net where anybody can connect and get an IP address from DHCP. All HTTP requests get redirected to port 80 on, which is a server on the internal net where all requests are answered with the same page, which displays contact info in case you want to be registered and be allowed to use the net.

You are allowed to ssh in to the gateway. Users with valid user IDs and passwords get rule sets with appropriate pass rules loaded for their assigned IP address.

We can fine tune this even more by making user specific rules in /etc/authpf/users/$user/authpf.rules. Per user rules can use the $user_ip macro for the user's IP address. For example, if I want to give myself unlimited access, create the following /etc/authpf/users/vegard/authpf.rules:

pass in on $wi_if from $user_ip to any keep state

Turning away the brutes

If you run a Secure Shell login service anywhere which is accessible from the Internet, I'm sure you've seen things like these in your authentication logs:

Sep 26 03:12:34 skapet sshd[25771]: Failed password for root from port 40992 ssh2
Sep 26 03:12:34 skapet sshd[5279]: Failed password for root from port 40992 ssh2
Sep 26 03:12:35 skapet sshd[5279]: Received disconnect from 11: Bye Bye
Sep 26 03:12:44 skapet sshd[29635]: Invalid user admin from
Sep 26 03:12:44 skapet sshd[24703]: input_userauth_request: 
invalid user admin
Sep 26 03:12:44 skapet sshd[24703]: Failed password for invalid user 
admin from port 41484 ssh2
Sep 26 03:12:44 skapet sshd[29635]: Failed password for invalid user 
admin from port 41484 ssh2
Sep 26 03:12:45 skapet sshd[24703]: Connection closed by
Sep 26 03:13:10 skapet sshd[11459]: Failed password for root from port 43344 ssh2
Sep 26 03:13:10 skapet sshd[7635]: Failed password for root from port 43344 ssh2
Sep 26 03:13:10 skapet sshd[11459]: Received disconnect from 11: Bye Bye
Sep 26 03:13:15 skapet sshd[31357]: Invalid user admin from
Sep 26 03:13:15 skapet sshd[10543]: input_userauth_request: invalid 
user admin
Sep 26 03:13:15 skapet sshd[10543]: Failed password for invalid user 
admin from port 43811 ssh2
Sep 26 03:13:15 skapet sshd[31357]: Failed password for invalid user 
admin from port 43811 ssh2
Sep 26 03:13:15 skapet sshd[10543]: Received disconnect from 11: Bye Bye
Sep 26 03:13:25 skapet sshd[6526]: Connection closed by

It gets repetetive after that. This is what a brute force attack looks like. Essentially somebody, or more likely, a cracked computer somewhere, is trying by brute force to find a combination of user name and password which will let them into your system.

The simplest response would be to write a pf.conf rule which blocks all access. This leads to another class of problems, including what you do in order to let people with legitimate business on your system access it anyway. You might consider moving the service to some other port, but then again, the ones flooding you on port 22 would probably be able to scan their way to port 22222 for a repeat performance.

Since OpenBSD 3.7, PF has offered a slightly more elegant solution. You can write your pass rules so they maintain certain limits on what connecting hosts can do. For good measure, you can banish violators to a table of addresses which which you deny some or all access. You can even choose to drop all existing connections from machines which overreach your limits, if you like. Here's how it's done:

Now first set up the table. In your tables section, add

table <bruteforce> persist

Then somewhere fairly early in your rule set you set up to block from the bruteforcers

block quick from <bruteforce>

And finally, your pass rule.

pass inet proto tcp from any to $int_if:network port $tcp_services \
        flags S/SA keep state \
	(max-src-conn 100, max-src-conn-rate 15/5, \
         overload <bruteforce> flush global)

This is rather similar to what we've seen before, isn't it? In fact, the first part is identical to the one we constructed earlier. The part in brackets is the new stuff which will ease your network load even further.

max-src-conn is the number of simultaneous connections you allow from one host. In this example, I've set it at 100, in your setup you may want a slightly higher or lower value.

max-src-conn-rate is the rate of new connections allowed from any single host, here 15 connections per 5 seconds. Again, you are the one to judge what suits your setup.

overload <bruteforce> means that any host which exceeds these limits gets its address added to the table bruteforce. Our rule set blocks all traffic from addresses in the bruteforce table.

finally, flush global says that when a host reaches the limit, that host's connections will be terminated (flushed). The global part says that for good measure, this applies to connections which match other pass rules too.

The effect is dramatic. My bruteforcers more often than not end up with "Fatal: timeout before authentication" messages, which is exactly what we want.

Once again, please keep in mind that this example rule is intended mainly as an illustration. It is not unlikely that your network's needs are better served by rather different rules or combinations of rules.

If, for example, you want to allow a generous number of connections in general, but would like to be a little more tight fisted when it comes to ssh, you could supplement the rule above with something like one early on in your rule set:

pass quick proto { tcp, udp } from any to any port ssh \
        flags S/SA keep state \
        (max-src-conn 15, max-src-conn-rate 5/3, \
        overload <bruteforce> flush global)

You should be able to find the set of parameters which is just right for your situation by reading the relevant man pages and the PF User Guide, and perhaps a bit of experimentation.

Giving spammers a hard time

At this point we've covered quite some ground, and I'm more than happy to present something really useful: PF as a means to make spammers' lives harder. Based on our recent exposure to PF rulesets, understanding the following /etc/pf.conf parts should be straighforward:

table <spamd> persist
table <spamd-white> persist file "/var/mail/whitelist.txt"
rdr pass on $ext_if inet proto tcp from <spamd> to \
         { $ext_if, $int_if:network } port smtp -> port 8025
rdr pass on $ext_if inet proto tcp from !<spamd-white> to \
         { $ext_if, $int_if:network } port smtp -> port 8025

We have two tables, and one of them gets filled up from a file somewhere in the file system. SMTP traffic from the addresses in the first table plus the ones which are not in the other table are redirected to a daemon listening at port 8025.

Remember, you are not alone: blacklisting

The main point underlying the spamd design is the fact that spammers send a large number of messages, and the probability that you are the first person receiving a particular message is incredibly small. In addition, spam is mainly sent via a few spammer friendly networks and a large number of hijacked machines. Both the individual messages and the machines will be reported to blacklists fairly quickly, and this is the data which eventually ends up in the first table in our example.

List of black and grey, and the sticky tarpit

What spamd does to SMTP connections from addresses in the blacklist is to present its banner and immediately switch to a mode where it answers SMTP traffic 1 byte at the time. This technique,which is intended to waste as much time as possible on the sending end while costing the receiver pretty much nothing, is called tarpitting. The specific implementation with 1 byte SMTP replies is often referred to as stuttering.

spamd also supports greylisting, which works by rejecting messages from unknown hosts temporarily with 45n codes, letting messages from hosts which try again within a reasonable time through. Traffic from well behaved hosts, that is, senders which are set up to behave within the limits set up in the relevant RFCs[14], will be let through.

Greylisting as a technique was presented in a 2003 paper by Evan Harris[15], and a number of implementations followed over the next few months. OpenBSD's spamd aquired its ability to greylist in version OpenBSD 3.5, which was released in May 2004.

The most amazing thing about greylisting, apart from its simplicity, is that it still works. Spammers and malware writers have been very slow to adapt. We will see a few examples later.

Setting up spamd

Configuring spamd is fairly straightforward[16]. You simply edit /etc/spamd.conf according to your own needs. The file itself offers quite a bit of explanation, and the man page offers additional information. The suggested default blacklists quite a bit, including Korean addresses. I work in a company which actually does the odd bit of business with Koreans, and consequently I needed to edit out that particular entry from our configuration. You are the judge of which data sources to use, and using other lists than the default ones is possible.

Put the lines for spamd and the startup parameters you want in your /etc/rc.conf or /etc/rc.conf.local. Note for example that you can fine tune several of the greylisting related parameters via spamd command line parameters.

When you are done with editing the setup, you start spamd with the options you want, and complete the configuration using spamd-setup. Finally, you create a cron job which updates the tables at reasonable intervals.

Once the tables are filled, you can list and manipulate their contents using pfctl, just like any other tables.

Note that the example above uses rdr rules which are also pass rules. If your rdr rules do not include a 'pass' part, you need to set up pass rules to let traffic through to your redirection. You also need to set up rules to let legitimate email through. If you are already running an email service on your network, you can probably go on using your old SMTP pass rules.

Highlights of our spamd experience

What is spamd like in practical use? We started using spamd in earnest in early December of 2004, after running spamassassin and clamav as parts of the exim delivery process for a while. Our exim is configured to tag and deliver messages with a spamassassin score in the interval from 5 to 9.99 points, while discarding messages with 10 points or more along with malware carrying messages. As the autumn progressed, spamassassin's success rate had been steadily declining, letting ever more spam through.

When we put spamd into production, the total number of messages handled and the number of messages handled by spamassassin decreased drastically. The number of spam messages which make it through untagged is now stabilized at roughly five a day, based on a reporting population of a handful of users.

If you start spamd with the -v command line option for verbose logging, the logs start including a few more items of information in addition to the IP addresses. With verbose logging, a typical log excerpt looks like this:

Oct  2 19:55:05 delilah spamd[26905]: (GREY) 
<> -> <>
Oct  2 19:55:05 delilah spamd[26905]: disconnected after 
0 seconds.
Oct  2 19:55:05 delilah spamd[26905]: connected (2/1)
Oct  2 19:55:06 delilah spamd[26905]: (GREY) 
<> -> <>
Oct  2 19:55:06 delilah spamd[26905]: disconnected after 
1 seconds.
Oct  2 19:57:07 delilah spamd[26905]: (BLACK)
 <> -> <>
Oct  2 19:58:50 delilah spamd[26905]: From: Auto 
lnsurance Savings <>
Oct  2 19:58:50 delilah spamd[26905]: Subject: Start 
SAVlNG M0NEY on Auto lnsurance
Oct  2 19:58:50 delilah spamd[26905]: To:
Oct  2 20:00:05 delilah spamd[26905]: disconnected after 
404 seconds. lists: spews1
Oct  2 20:03:48 delilah spamd[26905]: connected (1/0)
Oct  2 20:03:48 delilah spamd[26905]: disconnected after 
0 seconds.
Oct  2 20:06:51 delilah spamd[26905]: connected (1/1), 
lists: spews1
Oct  2 20:07:00 delilah spamd[26905]: connected (2/1)
Oct  2 20:07:00 delilah spamd[26905]: disconnected after 
0 seconds.
Oct  2 20:07:12 delilah spamd[26905]: disconnected after 
21 seconds. lists: spews1

The first three lines say that a machine connects, as the second active connection, with one connection from a blacklisted host. The (GREY) and (BLACK) before the addresses indicate greylisting or blacklisting status, respectively. After 404 seconds (or 6 minutes, 44 seconds), the blacklisted host gives up without completing the delivery. The following lines may be the first ever contact from a machine, which is then greylisted.

Our preliminary conclusion is that spamd stops quite a bit of spam. Unfortunately, we have also seen some false positives. As far as I can tell, these have all been connected with entries in the spews2 (spews level 2) list. For now we have stopped using this list as a blacklist, without a noticeable increase in the spam volume.

Now for what used to be the the climax of my spamd experience. One log entry stood out for a long time:

Dec 11 23:57:24 delilah spamd[32048]: connected (1/1), 
lists: spamhaus spews1 spews2
Dec 12 00:30:08 delilah spamd[32048]: disconnected 
after 1964 seconds. lists: spamhaus spews1 spews2

This entry concerns a sender at This particular machine made 13 attempts at delivery during the period from December 9th to December 12th, 2004. The last attempt lasted 32 minutes, 44 seconds, without completing the delivery.

I update the tutorial now and again, and recently I found a few more entries which exceeded this:

peter@delilah:~$ grep disconnected /var/log/spamd | awk '{print $9}' \
| sort -rn | uniq -c | head
   1 36099
   1 14714
   1 5495
   1 2193
   1 1964
   1 1872
   1 1718
   1 1677
   1 1617
   1 1594

The first, at 36099 seconds, which is more than ten hours,

Aug 26 10:10:17 delilah spamd[26905]: connected (1/0)
Aug 26 20:11:56 delilah spamd[26905]: disconnected 
after 36099 seconds.

concerns a machine apparently at the University of Wisconsin, Madison campus, which was probably infected by an extremely naive spam sending worm, which just took a long time waiting for the rest of the SMTP dialogue. The next entries, 4 hours, 5 minutes and 14 seconds and 1 hour, 31 minutes and 35 seconds respectively, seem to have behaved in much the same way.

Conclusions from our spamd experience

Summing up, selectively used, blacklists combined with spamd are powerful, precise and efficient spam fighting tools. The load on the spamd machine is minimal. On the other hand, spamd will never perform better than its weakest data source, which means you will need to monitor your logs and use whitelisting when neccessary.

PF - Haiku

Finally, an indication of the level of feeling inspired by PF in its users is in order. On the PF mailing list, a message with the subject of "Things pf can't do?" appeared in May 2004. The message had been written by someone who did not have a lot of firewalls experience, and who consequently found it hard to get the setup he or she wanted.

This, of course, lead to some discussion, with several participants saying that if PF was hard on a newbie, the alternatives were certainly not a bit better. The thread ended in the following haiku of praise from Jason Dixon, which is given intact as it came, along with Jason's comments:

Compared to working with iptables, PF is like this haiku:

A breath of fresh air,
floating on white rose petals,
eating strawberries.

Now I'm getting carried away:

Hartmeier codes now,
Henning knows not why it fails,
fails only for n00b.

Tables load my lists,
tarpit for the asshole spammer,
death to his mail store.

CARP due to Cisco,
redundant blessed packets,
licensed free for me.

Jason Dixon, on the PF email list, May 20th, 2004 (


OpenBSDs web


PF User Guide

Daniel Hartmeier's PF pages,

Daniel Hartmeier: Design and Performance of the OpenBSD Stateful Packet Filter (pf), (presented at Usenix 2002)

Nate Underwood: HOWTO: Transparent Packet Filtering with OpenBSD,

Randal L. Schwartz: Monitoring Net Traffic with OpenBSD's Packet Filter, Brandvägg med OpenBSD,

Randal L. Schwartz: Blog for Thu, Jan 29, 2004,

RFC 1631, "The IP Network Address Translator (NAT)", May 1994

RFC 1918, "Address Allocation for Private Internets", February 1996

The FreeBSD PF home page,

Peter Postma's PF on NetBSD pages,

Marcus Ranum: The Six Dumbest Ideas in Computer Security, September 1, 2005

Kjell Jørgen Hole WiFi courseware,, also see; also The Unofficial 802.11 Security Web Page comes higly recommended.

Evan Harris: The Next Step in the Spam Control War: Greylisting (the original greylisting paper)

Where to find the tutorial on the web

Work in progress edition:

Several formats

This is where updated versions will appear more or less in the natural course of tinkering, in between events. Please let me know if you want to be told of future updates.

If you enjoyed this: Buy OpenBSD CDs and other items, donate!

If you have enjoyed this tutorial or found it useful, please go to the Orders page and buy CD sets, or for that matter, support further development work by the OpenBSD project via a donation.

If you've taken this in at a conference, there might even be an OpenBSD booth nearby where you can buy CDs, T-shirts or other items.

Remember, even free software takes real work and real money to develop and maintain.



This manuscript is a slightly further developed version of a manuscript prepared for a lecture which was announced as (translated from Norwegian): "This lecture is about firewalls and related functions, with examples from real life with the OpenBSD project's PF (Packet Filter). PF offers firewalling, NAT, traffic control and bandwidth management in a single, flexible and sysadmin friendly system. Peter hopes that the lecture will give you some ideas about how to control your network traffic the way you want - keeping some things outside your network, directing traffic to specified hosts or services, and of course, giving spammers a hard time."

People who have offered significant and useful input regarding this manuscript include Eystein Roll Aarseth, David Snyder, Peter Postma, Henrik Kramshøj, Vegard Engen, Greg Lehey, Ian Darwin, Daniel Hartmeier, Mark Uemura, Hallvor Engen and probably a few who will remain lost in my mail archive until I can grep them out of there.


The two documents are RFC 1631, "The IP Network Address Translator (NAT)", dated May 1994 and RFC 1918, "Address Allocation for Private Internets", dated February 1996. See the Chapter called References


As a footnoted aside, I tend to use sudo when I need to do something which requires privileges. Sudo is in the base system on OpenBSD, and is within easy reach as a port or package elsewhere. If you have not started using sudo yet, you should. Then you'll avoid shooting your own foot simply because you forgot you were root in that terminal window.


For convenience if you want it - pfctl is able to handle several operations on a single command line. You can, for example, enable PF and load the rule set with the command sudo pfctl -e -f /etc/pf.conf


There are some differences between the FreeBSD 4.n and 5.n and newer releases with respect to PF. Refer to the FreeBSD Handbook, specifically the PF chapter to see which information applies in your case.


You may ask why do I write the rule set to default deny? The short answer is, it gives you better control at the expense of some thinking. The point of packet filtering is to take control, not to run catch-up with what the bad guys do. Marcus Ranum has written a very entertaining and informative article about this, The Six Dumbest Ideas in Computer Security, which is a good read to boot.


For dialup users and a significant subset of ADSL users, specifically those using PPP over Ethernet (PPPoE), the external interface will be tun0, instead of an Ethernet interface.


The main internet RFCs describing ICMP and some related techhiques are RFC792, RFC950, RFC1191, RFC1256, RFC2521, rfc2765, while necessary updates for ICMP for IPv6 are found in RFC1885, RFC2463, RFC2466. These documents are available in a number of places on the net, such as the and web sites, and probably also via your package system.

It is quite possible that I will return to ICMP filtering in a future advanced section of the tutorial.


See Redirection and Reflection in the PF user guide.


On FreeBSD, ALTQ requires the ALTQ and queue discipline options for the disciplines you want to use to be compiled into the running kernel. Refer to the PF chapter of the FreeBSD Handbook for further information.


At the time of writing, ALTQ has not been integrated in the NetBSD PF implementation. Peter Postma maintains a NetBSD patch to enable PF/ALTQ functions. Up to date information on this is available from Peter Postmas PF on NetBSD pages,


For a more complete overview of issues surrounding security in wireless networks, see eg Professor Kjell Jørgen Hole's articles and slides at For fresh developments in the WiFI field, the sites and The Unofficial 802.11 Security Web Page come higly recommended.


Wireless network support in OpenBSD and BSDs in general is getting better all the time, but this does not mean that getting all the bits you need is necessarily easy. A brief history of my home network goes like this: I started out buying two CNet CWP-854 cards, which should be supported in OpenBSD 3.7 via the new ral driver. The one I put in the brand new Dell machine running a non-free operating system worked right out of the box. My gateway, which had been running without incident since the 3.3 days however, was a little more problematic. The card did get recognized and configured, but once the Dell tried to get an IP address, the gateway went down with a kernel panic. The gory details are available as OpenBSD PR number 4217. I have promised to test the card again with a new snapshot - as soon as I can locate the card again. From the Dell we could see an amazing number of networks, almost all unsecured, but that's another story entirely.

I decided i wanted to try ath cards, and bought a D-Link DWL-G520, which I then managed to misplace while moving house. Next, I bought a DWL-G520+, thinking that the plus sign must mean it's better. Unfortunately, the plus meant a whole different chipset was used, the TI ACX111, which comes with a low price tag but with no documentation accessible to free software developers. Fortunately the store let me return the card for a refund with no trouble at all. At this point, I was getting rather frustrated, and went all across town to a shop which had several DWL-AG520 cards in stock. It was a bit more expensive than the others, but it did work right away. A couple of weeks later the G520 turned up, and of course that worked too. My laptop (which runs FreeBSD) came with a Realtek 8180 wireless mini-PCI card, but for some reason I could not get it to work. I ended up buying DWL-AG650 cardbus card, which works flawlessly with the ath driver. In general, my advice is, if you shop online, keep the man pages available in another tab or window, and if you go to a physical store, make sure to tell the clerks you will be using a BSD, and if you're not sure about the card they are trying to sell you, see if you can borrow a machine to browse the online man pages. Telling the clerks up front could end up making it easier to get a refund if the part does not work, and telling them the card did work is good advocacy.


The relevant RFCs are mainly RFC1123 and RFC2821.


The original Harris paper and a number of other useful articles and resources can be found at the web site.


Note that on FreeBSD, spamd is a port, /usr/ports/mail/spamd/. If you are running PF on FreeBSD 5.n or newer, you need to install the port, follow the directions given by the port's messages and return here.