Using OpenBSD 3.3
As A
Firewall/Gateway for Home DSL or Cable

Shamim Mohamed

Abstract

This is a quick tutorial on how to set up an OpenBSD 3.3 system to run as a firewall, NAT proxy, time and DHCP server on a system connected to the Internet via broadband like DSL or cable. These are the things you should know right now as you set up your machine. You are not expected to be a Unix expert (why would a Unix expert need this how-to?) — if you don't understand something, or something looks intimidating, read on and come back to it. If you want to know something in more detail, read the manpages or the OpenBSD FAQ list. If something doesn't make sense, let me know.

Caveat: There may be instances of hyperbole and irony (perhaps even syncopy) in this document. Brush up your critical reasoning.

Do you want an older version, about setting up a firewall on OpenBSD 3.0 or OpenBSD 2.9? (Not too many changes between this and the other 3.x versions.)

New!

I have a shell script that sets up everything mentioned here. This is still experimental but if you try it, please let me know how it goes. Save this file to disk and run it by typing "sh config33-fw.sh". (Doesn't handle PPPoE [the beast].)

There is a new section called Tips and Stuff where I put things I've found or written that are useful sysadmin tools.

Introduction

Why OpenBSD? It's simple and secure. Your firewall machine should not have lots of things installed on it; therefore no exotic hardware, graphical desktops, X11 servers etc. — put those on your desktop machine. A simpler system is more robust and more secure; this machine only offers SMTP (email), ssh, ping/traceroute and optionally HTTP (web) to the outside world. And since it's running Unix, you can log in to it — securely -- using ssh from anywhere on the Internet and make any changes you need to. (N.B.: never use telnet to connect to a machine over the Internet! Anyone can eavesdrop and grab important information like passwords. Only use ssh, which encrypts all communication so that eavesdroppers don't get any information. And verify those key fingerprints or you leave yourself open to a man-in-the-middle attack.)

The utility and security of having this kind of machine: a firewall protects your data and systems from the Big, Bad Internet. When the bad guys are out to vandalise machines on the Internet, MS-Windows machines of various kinds are prime targets because they suck. Er, I mean, Windows is really hard to secure. (Not that an incompetently run Unix machine is any better, of course.) When you dialled in on the phone, your machine was on the 'net for brief periods; with DSL or cable it's vulnerable all the time.

This tutorial assumes that you have some familiarity with using Unix: what filenames look like, how to copy and edit files etc. There's a decent Unix tutorial on the web. The most important command to remember is man (short for "manual") -- if I say something like "read the documentation for foobar it means you should type man foobar. One other piece of Unix argot: if you hear someone write select(2) it indicates that the manual for select is in section 2, i.e. you would read the manpage by typing man 2 select.

NAT allows you to connect lots of PCs up to one network connection. When any of the machines wants to make a connection to some server out there, the NAT box intercepts that request, and sends the request off as though it came from the NAT machine. When the reply arrives, it is sent off to the machine that made the connection. Neither the server nor the machines on the inside know that all this is going on.

Aside: NAT is also called PAT, for "Port Address Translation." Also, read this interesting article by HRH Prince Philip, Duke of Edinburgh, on setting up PAT and DHCP on Cisco routers. The whole routergod.com site features many celebrities offering helpful tips on various network issues. Even if you don't want plan on having more than one PC at home, NAT is useful, because it allows the machine running your firewall to be different from your main workstation. You probably want to install fancy hardware and software on your machine; but every additional package installed on a firewall makes it more vulnerable.


 
Network Address Translation (NAT)

Note: if you only have one machine on the "inside", you don't need an ethernet hub; use a crossover cable to connect the two machines directly. This also has the advantage that you can get a full-duplex connection between the machines (a hub only allows a half-duplex connection). Or you could use a switch, which allows full-duplex connections between hosts — for less than $50 you can get a 5- or 8-port switch.

Note: you can buy little NAT/DHCP boxes from various manufacturers for about $150, but where's the fun in that? Besides, who knows how strong the security is on those things. With OpenBSD you know you're getting the best.

Building the machine

The machine itself: I prefer to build these machines up from individual components rather than buying a pre-made box. That way I can get name-brand supported components, and it works out slightly cheaper since I don't have to get exotic video cards, sound cards, CD-ROM drives etc. (Not to mention a Fisher-Price operating system that you will be required to pay for.)

Can you build a PC? Well, no one showed me how, but I've managed to put together about 10 or so systems, so it can't be that hard. If you've assembed anything with screwdrivers etc. you'll be fine. There are numerous sites on the web that walk you through building a PC. Go do a Google search and read those. I especially like the one at Acme Labs by Jef Poskanzer. There's also an excellent motherboard finder at Acme.

Caveat: specific recommendations will be outdated as soon as I write them! I like to use AMD CPUs because I believe Intel is evil and as far as possible I'd like to not buy their products. I'd get the current not-top-of-the-line CPU i.e. the one that costs about $50 and a compatible motherboard that costs in the range of $70. I stay away from integrated components because they're usually garbage. (For a server that I don't use directly I might get integrated video.) Spend about $30-50 on RAM, $30 on ethernet, $60 on an IDE disk, $30 for a case (with power supply). I usually find the best prices on components at Directron and CompuVest. These have both been non-sleazy (everything was as described in their catalog and shipping was prompt) in all my dealings with them — but let me know if you find any evidence of sleaziness.

All these components add up to around $300 — and that's brand-new stuff. If you have any old components lying around, they will be fine. You don't need a keyboard, mouse or monitor when the system is up and running — all maintenance on it can be done over the network. (While you're installing the OS on the machine you will need to hook up a keyboard, monitor and CD-ROM drive to it, of course.)

While installing the system, I plug in a spare CD-ROM drive, keyboard and monitor. Change the BIOS settings so that the machine will boot without a keyboard etc. Boot off the OpenBSD 3.3 CD and install the system. All the hardware should be recognised without any problems. (The installation guide booklet that comes with the CDs is excellent.)

The easiest way to install OpenBSD is to buy the distribution on CDs. Although you can install it via the network, buying the CD will help make sure that the OpenBSD project will continue to improve and better the system. If you can afford an outlay of US$40, please buy the CDs from the OpenBSD ordering site.

When you're installing OpenBSD, the installer program will ask you for disklabel information (partitions). On a Unix system, a group of files organised together is called a filesystem. The disk is partitioned into various pieces each of which will hold one filesystem. This is the filesystem breakup and partition sizes I'd use for a 12GB disk (if your disk is bigger, you can just increase the size of /var (for web files) or /home (for your personal files) — the system will be more than happy with these sizes for /, /tmp and /usr):

/dev/wd0a      100M     /
/dev/wd0d      400M     /tmp
/dev/wd0e      4GB      /var
/dev/wd0g      2GB      /usr
/dev/wd0h      5GB      /home    

(The convention is that a is always /, b is swap and c is the whole disk.) Your web files will live in /var, and your other files in /home.

This is all overkill; /usr only needs about 600M or so. Say pad it to 1GB. A 2GB disk would be plenty for the system, but if the cheapest disk you can get is 13GB....

Note for Unix newcomers: the disk is named /dev/wd0, and in this case it has 5 partitions with names /dev/wd0a, /dev/wd0d, /dev/wd0e, /dev/wd0g and /dev/wd0h. And the different partitions don't get different "drive letters" as in some primitive operating systems; once the system is installed, it looks to the user that there is just one bunch of files; Unix will figure out the right thing to do. After the system has been installed and you've booted off the hard disk, log in and (this is important!) type man afterboot; it will remind of some things that you need to do to complete the installation — pick passwords, create user accounts, check network settings etc. Also, man hier will introduce you to the way the system is organised — which files live where. In fact, let me say that again:

After the first normal boot of the system, be sure to read these manpages:
$  man afterboot
$  man hier      

There! And make sure you keep reading the manpages — OpenBSD manpages are a thing of beauty, complete, up-to-date and informative. And also read the OpenBSD FAQ on the web -- much of this information is also found there.

Configuring the network

For my outside connection I have DSL and a static IP number (from Speakeasy — I recommend them over PacBell etc. — I'm so happy I switched). Other DSL options are PPPoE that PacBell likes to set people up with, or DHCP which is what you usually get over cable. A completely bogus DSL installation is the USB device they try to foist on customers with Windows. Danger, Will Robinson! They stink; they're unsupported on any free O/S, and even on Windows they work about half the time.

PPPoE users: There is one case you need to handle yourself — setting up DNS. Details will be in the section on DNS.

In *BSD the network cards are named according to the driver used. For the D-Link cards, the driver is called rl, so my two ethernet cards are rl0 and rl1. For the inside network I use the "private" (non-routable) IP numbers 192.168.1.* which will make the inward-facing network card 192.168.1.1. The OpenBSD initialization asks you for IP numbers for the two cards. Enter the appropriate ones - the IP number your ISP gave you for rl0, and 192.168.1.1 for rl1. For PPPoE, the outside interface is tun0 and it will figure out its own IP address. If you're supposed use DHCP on your DSL or cable connection, type in dhcp.

It is important to remember which network will be the outside and which the inside. If the two cards are identical, the easiest way is to look at the MAC number. Every ethernet card ever made has a unique ID called its MAC number. This will be printed on the card, usually as a sticker. When the kernel boots up, it will print the MAC numbers of each card it finds:

rl0 at pci0 dev 9 function 0 "Realtek 8139" rev 0x10: irq 11 address 00:50:ba:44:ab:1c
rl1 at pci0 dev 10 function 0 "Realtek 8139" rev 0x10: irq 10 address 00:50:ba:44:9c:3e     

So the card that has a MAC number ending ab1c is rl0; the other is rl1. (If the two network cards you have are different types, there's no problem, of course. The kernel bootup messages will still be useful to tell you what names the system is using for them.)

(There's some rule about where the cards are plugged in so which one gets number 0 and which no. 1, but I can never remember that.)

PPPoE

The beast! PPPoE is a pain in the ass but ISPs like it because it makes things simpler for them — they don't have to maintain lists of IP numbers. Also, they can run a crappy service and keep dropping the connection and that's ok, you're expected to reconnect. It's the Micros**t philosophy of "make something really crappy and expect people to just re-start the whole system a couple of times a day." It's a pain in the ass for us because its MTU is 1492 instead of 1500 which used to require changes on every machine inside the network — but now thanks to the "mssfixup" flag we don't have to any more. (You are not expected to understand that.)

Caveat: I do not have access to any PPPoE connections any more, so there may be changes since 3.1. Keep all this in mind as you read this section, and please send me corrections.

The files you will need to change for PPPoE all live in /etc/ppp/. There are other differences: you shouldn't have /etc/mygate; and the file describing the outside interface, /etc/hostname.dc0 in my example, will only have one word in it: up. This tells the system to bring up the interface at boot time, but to do nothing else — the PPP program (daemon) will do the rest.

The network interface when using PPPoE is tun0, not the ethernet card the PPPoE line goes into (dc0 in my example). The PPP daemon handles the network on its own, then passes packets on to the tun0 pseudo-device. That is the name you'll use in any place network devices are specified, like the firewall or network sniffing tools.

The main config file is /etc/ppp/ppp.conf and this is what it should look like:

default:
 set log Phase Chat LCP IPCP CCP tun command
 set redial 15 0
 set reconnect 15 10000

pppoe:
 set device "!/usr/sbin/pppoe -i rl0"
 disable acfcomp protocomp
 deny acfcomp
 set mtu 1492
 set speed sync
 enable lqr
 set lqrperiod 5
 set cd 5
 set dial
 set login
 set timeout 0
 set authname login
 set authkey password
 enable dns
 enable mssfixup
    
Caveat: pay attention to the leading spaces. Use your login name and password where indicated. The "set device" line tells ppp which physical device to use to talk to the outside world. You also have to tell the system to start PPPoE at boot time. That can be done with this little snippet of shell script:
echo -n "Trying to establish PPPoE DSL"; ppp -ddial pppoe
for i in 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0; do
    sleep 5
    echo -n "$i"
    if /usr/local/sbin/adsl-status>/dev/null; then
        break
    fi
done
echo
/usr/local/sbin/adsl-status    
Where adsl-status is a little shell-script that tests to see whether the PPP link has come up properly:
#!/bin/sh
 
IP=$(/sbin/ifconfig tun0 | awk '/netmask/{print $2}')
 
if [ -z "$IP" ]; then
   echo "ADSL link is down."
   exit 1
else
   echo "ADSL is up, IP address $IP"
   exit 0
fi    
Now the question is: where should we put the little loop that tries to get ppp going? The right place to put all these is in /etc/rc.local. However this has the drawback that the outside network hasn't been initialised while the rest of the system is coming up, which causes some scary-looking error messages from NAT to be printed at boot time. So I do something a little un-kosher: I put the ppp initialisation in /etc/netstart right at the end:
...
echo -n ' ADSL... '; ; ppp -ddial pppoe 
for i in 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0; do
    sleep 5
    echo -n.$i"
    if /usr/local/sbin/adsl-status>/dev/null; then
        break
    fi
done
echo
/usr/local/sbin/adsl-status    
Now remember that each time the PPP link goes up or down, the firewall and NAT rules must be re-done. (NAT and firewalls will be covered shortly.) The files /etc/ppp/ppp.linkup and /etc/ppp/linkdown are scripts that get run by ppp. Here's /etc/ppp/ppp.linkup:
MYADDR:
 ! sh -c "/sbin/route del default"
 ! sh -c "/sbin/route add default HISADDR -mtu 1492"
 ! sh -c "/sbin/pfctl -e -f /etc/pf.conf"
 ! sh -c "/usr/local/sbin/ntpd -p /var/run/ntpd.pid"
And this is /etc/ppp/linkdown:
MYADDR:
 ! sh -c "/sbin/pfctl -d"    
Caveat: There is a space before the exclamation points on those lines.

As I said before, I no longer have access to any PPPoE systems; if you know of any inaccuracies or bogosities above, please let me know.

Configure system files

To set up the system, the files you will be editing are:/etc/rc.conf, /etc/myname, /etc/mygate, /etc/pf.conf, /etc/nat.conf, /etc/*.conf, /etc/hostname.interface, /var/named/*.

Edit /etc/rc.conf. On my servers I run SMTP, Apache, and ssh. In other words, from the outside it handles email, web acess and secure shell for remote logins. For convenience, on the inside I have a private name server (DNS) and NTP server for accurate time. To get sendmail, NTP, httpd, and NAT to work, these are the lines to change:

sendmail_flags="-bd -q30m"      # for normal use: "-bd -q30m"
named_flags=""                  # for normal use: ""
ntpdate_flags="put.server.here" # for normal use: NTP server; run before ntpd starts
httpd_flags=""                  # for normal use: "" (or "-DSSL" after reading ssl(8))
dhcpd_flags=-q                  # for normal use: "-q"
pf=YES                          # Packet filter / NAT
ntpd=YES                        # run ntpd if it exists       
pf_rules=/etc/pf.conf           # Packet filter rules file    
Make sure that /etc/sysctl.conf has this line in it:
net.inet.ip.forwarding=1        # 1=Permit forwarding (routing) of packets    
Get the names of NTP servers close to where you are and put that name in the ntpdate value. Here's a list of public NTP servers.

The system should already have setup /etc/hostname.dc0 and /etc/hostname.fxp0 (or whatever your network device names are) for you. Each file will have the IP number and netmask. This is what these files would look like:

 
bash-2.05$ cat /etc/hostname.fxp0
inet 192.168.1.1 255.255.255.0 NONE 
bash-2.05$ cat /etc/hostname.dc0
inet 123.45.67.89 255.255.255.0 NONE    
(The bash-2.05$ is the prompt; cat types a file out to the output.) If you're using DHCP, the outside interface's hostname file will say dhcp. If PPPoE, then just the word up.

Other important files: /etc/myname contains your hostname; /etc/mygate — your default gateway to the outside world (your ISP told you what this should be — it's usually the same as your IP number except that the last number is replaced with a 1 or 254) — except if you're on PPPoE, in which case you don't have one.

NAT and firewall rules

OpenBSD 3.x has a new packet filter — 2.9 used ipf but 3.x has a re-written from scratch one called pf. The details are not important; pf config files are much simpler.

My outside interface is dc0 and the inside interface is fxp0. (If you're using PPPoE, the outside interface will be tun0.) I also have a wireless network; since 802.11b wireless ethernet is not particularly secure, and I wanted my network to be freely accessible by anyone, the inside machines need to be protected from it. The best way to do this is to make the wireless and inside networks completely different; a third network card rl0 is connected to an 802.11b access point and is assigned the network 192.168.2.0/24. I also block outbound email (port 25) from the wireless network, since otherwise anyone on the street could use my resources to send spam. (Regular people that use some sort of web-based email service will not be affected.)

Firewall rules (they tell the gateway what kind of network traffic should be allowed into the internal network) live in /etc/pf.conf; NAT configuration is also in this file.

Here's a sample /etc/pf.conf — very little is accessible from the outside, but machines on the inside can go out with no restrictions. In your files you'd edit the lines near the top of the file with the names of your outward- and inward-facing ethernet cards, and wireless card if any. Read it, understand it, modify it for your specific needs. Security is not a spectator sport.

#####################################################################
#
# IP packet filtering rules (firewall)
# Shamim Mohamed 3/2002, 5/2003

# See pf.conf(5) for syntax and examples

# If you change this file, run
#    pfctl -f /etc/pf.conf
# to update kernel tables (also run "pfctl -e" if pf was not running)

# Network interfaces (Remember, if using PPPoE the ext. interface is tun0)
internal = "fxp0"
external = "dc0"
wireless = "rl0"
unsafe = "{ dc0, rl0 }"

# Services visible from the outside — remove any you're not using
services = "{ ssh, http, https, smtp, domain }"

# The wireless interface is not allowed to send anything to the inside
# network. It can send anything out except smtp since we don't
# want being used as a spam relay. Yes, this is paranoid. Better safe
# than sorry.

# You shouldn't need to change anything below this line
#####################################################################

# Non-routable IP numbers
nonroutable = "{ 192.168.0.0/16, 127.0.0.0/8, 172.16.0.0/12, 10.0.0.0/8,
    0.0.0.0/8, 169.254.0.0/16, 192.0.2.0/24, 204.152.64.0/23, 224.0.0.0/3,
    255.255.255.255/32 }"

# All rules are "quick" so go strictly top to bottom

# Fix fragmented packets
scrub in all

# Create two packet queues: one for regular traffic, another for
# high priority: TCP ACKs and packets with ToS 'lowdelay'
altq on $external priq bandwidth 125Kb queue { highpri_q, default_q }
queue highpri_q priority 7
queue default_q priority 1 priq(default)


# NAT

# nat: packets going out through dc0 with source addr 192.168.1.0/24
# will get translated as coming from our external address. State is
# created for such packets, and incoming packets will be redirected to
# the internal address.

# I have an experimental web server on an inside machine; I can test it
# from the outside by connecting to port 8042.
rdr on $external inet proto tcp to port 8042 -> 192.168.1.12 port 80

# NAT: we need a rule for the inside network as well as the wireless.
nat on $external from 192.168.1.0/24 to any -> $external
nat on $external from 192.168.2.0/24 to any -> $external

########################################################################

#  Don't bug loopback
#
pass out quick on lo0 from any to any
pass in quick on lo0 from any to any

#  Don't bother the inside interface either
#
pass out quick on $internal from any to any
pass in quick on $internal from any to any

#####################################################################
#
#  First, we deal with bogus packets.
#

#  Block any inherently bad packets coming in from the outside world.
#  These include ICMP redirect packets and IP fragments so short the
#  filtering rules won't be able to examine the whole UDP/TCP header.
#
block in log quick on $unsafe inet proto icmp from any to any icmp-type redir

#  Block any IP spoofing atempts.  (Packets "from" non-routable
#  addresses shouldn't be coming in from the outside).
#
block in quick on $external from $nonroutable to any

#  Don't allow non-routable packets to leave our network
#
block out quick on $external from any to $nonroutable

#####################################################################
#
# Wireless:  block SMTP from wireless - spam threat
#
block in quick on $wireless inet proto tcp from any to any port smtp
#
#
#####################################################################

#####################################################################
#
#  The normal filtering rules
#

#  ICMP: allow incoming ping and traceroute only
#
pass in quick on $unsafe inet proto icmp from any to any icmp-type { \
    echorep, echoreq, timex, unreach }
block in log quick on $unsafe inet proto icmp from any to any

#  TCP: Allow ssh, smtp, http and https incoming. Only match
#  SYN packets, and allow the state table to handle the rest of the
#  connection. ACKs and ToS "lowdelay" are given priority.
#
pass in quick on $external inet proto tcp from any to any port $services \
     flags S/SA keep state queue (default_q, highpri_q)

# UDP: allow DNS since I run a public nameserver (remove if you don't!)
pass in quick on $unsafe inet proto udp from any to any port domain

################
# Wireless
#
# allow connections from 192.168.2.0/24, the inside wired network.
pass out quick on $wireless inet proto tcp from any to any \
     flags S/SA keep state queue (default_q, highpri_q)

# Everyone is allowed to send UDP and ICMP out
pass out quick on $external inet proto udp  all keep state
pass out quick on $external inet proto icmp from any to any keep state

# Block wireless -> inside network
block in quick on $wireless from any to $nonroutable

# Everything else is ok
pass in quick on $wireless from any to any

##################
#  Of course we need to allow packets coming in as replies to our
#  connections so we keep state. Strictly speaking, with packets
#  coming from our network we don't have to only match SYN, but
#  what the hell. It allows us to put those packets in the high
#  priority queue.
#
pass out quick on $external inet proto tcp from any to any \
     flags S/SA keep state queue (default_q, highpri_q)
pass out quick on $external inet proto udp  all keep state
pass out quick on $external inet proto icmp from any to any keep state


#  End of rules. Block everything to all ports, all protocols and return
#  RST (TCP) or ICMP/port-unreachable (UDP).
#
block return-rst in log quick on $unsafe inet proto tcp from any to any
block return-icmp in log quick on $unsafe inet proto udp from any to any
block in quick on $unsafe all

#
#  End of file
#
#####################################################################

Configuring email

(I'm now using postfix instead of sendmail since it has easier human-readable setup; I'll soon document that too.)

Sendmail should have been setup automatically since you edited /etc/rc.conf but I've occasionally had to make one change in /etc/mail/sendmail.cf:

 
Djmy-domain-name.com
    
(If you don't own a domain, or plan on having it point to your DSL machine, you don't need sendmail.)

You should have a normal user account that you're going to use (never log in as root! Always use su or sudo). Administrative email should be forwarded to you; if your normal username is zippy edit /etc/mail/aliases and make sure you make the appropriate lines look like this:

# Well-known aliases — these should be filled in!
root: zippy
manager: zippy
dumper: zippy
    
One thing you should consider is being an email handler for friends. My DSL service goes down too often -- every few months. This is too unreliable for my tastes. What I do is collaborate with friends to accept and queue email for them, and they do the same for me. For my domain foo.com the primary mail exchanger is gateway.foo.com, the OpenBSD firewall/gateway. But a secondary mail exchanger (which someone out there on the network will use if my DSL is down) will be gateway.bar.com, and email will wait on that site until my machine is back on the network. I want to perform the same service for my friend — if gateway.bar.com is down, I want people to be able to send my machine the email destined for bar.com. This goes in the file /etc/mail/relay-domains:
bar.com
fubar.org    
Now the machine will accept email for bar.com and fubar.org as well as for itself and then forward the messages on. If the machine it's trying to forward to is down, it will put them in the queue and keep re-trying for a while.

Setting up DNS

You probably shouldn't be running the primary DNS server for your domain on your DSL box; DSL may not be reliable enough for that. Get someone else to do it for you for free, like www.zoneedit.com.

However, it is nice to have a local private DNS because lots of daemons (services that run in the background, like the web server) like to do reverse lookups of IP numbers, so we should have a DNS server for the private network. Also, this installation will give you a caching nameserver which should improve your browsing speed. PPPoE users: don't forget, you have a few extra steps.

Current versions of OpenBSD use BIND 9, which is different from before: older versions used BIND 4. The config file name and syntax are different. The files live in /var/named. Here's a sample named.conf:


// Update this list to include only the networks for which you want
// to execute recursive queries. The default setting allows all hosts
// on any IPv4 networks for which the system has an interface, and
// the IPv6 localhost address.
acl clients {
        localnets;
        ::1;
};

options {
        version "";     // remove this to allow version queries

        listen-on    { any; };
        listen-on-v6 { any; };

        allow-recursion { clients; };

        forwarders {
        // Make sure you put your ISP's nameservers here!
                64.128.32.16;
                4.8.16.64;
        };
};

// Configuration for rndc, the namesever control program
key "rndc-key" {
        algorithm hmac-md5;
        // Choose a Base64-encoded 128-bit random key here!
        secret "hd5I7tJGKp7sC9bg6ddsbQ==";
};

controls {
        inet 127.0.0.1 port 953
                allow { 127.0.0.1; } keys { "rndc-key"; };
};

logging {
        category lame-servers { null; };
};

// Standard zones
//
zone "." {
        type hint;
        file "standard/root.hint";
};

zone "localhost" {
        type master;
        file "standard/localhost";
        allow-transfer { localhost; };
};

zone "127.in-addr.arpa" {
        type master;
        file "standard/loopback";
        allow-transfer { localhost; };
};

zone "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.ip6.arpa" {
        type master;
        file "standard/loopback6.arpa";
        allow-transfer { localhost; };
};

zone "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.ip6.int" {
        type master;
        file "standard/loopback6.int";
        allow-transfer { localhost; };
};

// Master zones
//

zone "my-domain.org" {
        type master;
        file "master/my-domain.org";
};

zone "drones.com" {
        type master;
        file "master/drones.com";
};

// Slave zones
//

zone "other-domain.org" {
        type slave;
        file "slave/other-domain.org";
        masters {
                64.1.23.45;       // Master server for other-domain.org
        };
};    
(Anything starting with // is a comment.) Change the IP numbers in the forwarders section to the nameservers your ISP told you to use. To control BIND 9, a program called rndc is used; a secret key is specified in the named.conf file that rndc uses to authenticate itself to the nameserver process. I use md5 -s to hash a quasi-random string of my choosing to get 16 bytes of key, or read 16 bytes directly from /dev/urandom using dd(1). Other files you need are the "zone files" for the domains you are master for, like /var/named/master/my-domain.org:
$TTL 1d
@       IN      SOA     gateway.my-domain.org.     root.my-domain.org. (
                                14      ; Serial
                                10800   ; Refresh
                                3600    ; Retry
                                604800  ; Expire
                                86400 ) ; Minimum
 
        IN      NS      gateway.my-domain.org.
 
 
gateway IN      A       192.168.1.1
libelle IN      A       192.168.1.2
discus  IN      A       192.168.1.4
ventus  IN      A       192.168.1.3
wander  IN      A       192.168.1.5
brad    IN      A       192.168.1.12
jack    IN      A       192.168.1.13

; your static IP number
dsl     IN      A       123.45.67.89
 
www     IN      CNAME   dsl
mail    IN      CNAME   dsl
    
In this network, there are six machines on the inside and those are their names and IP Number assignments. The OpenBSD gateway machine is named "gateway". Change these entries to names of the machines on your private network. You can give them any IP number that starts with 192.168.1. Of course if you have three machines on your network, there will only by three entries.)

To control the nameserver, the program rndc is used. It has a few simple commands: the only one you need isrndc reload which you should do after changing any configs or zone files. This updates the server with your changes.

PPPoE

Yes, another special case for PPPoE: You don't know what your ISP's DNS servers are. (And they could change which machines you're supposed to use each time you connect! In which case you may be screwed.) What you have to do is: connect "by hand" one time, and see which DNS servers you got. After ppp.conf has been written, you can run ppp -ddial pppoe and pray. If all goes well, ifconfig tun0 should show you two lines:
$  /sbin/ifconfig tun0
tun0: flags=11 mtu 1492
        inet 63.201.32.40 --> 63.201.39.254 netmask 0xff000000    
That means everything worked. Now look at /etc/resolv.conf — there should be one or more lines in there that say which nameservers should be used. Put these IP numbers in the forwarders line in /var/named/named.boot. One other wrinkle: the /etc/resolv.conf that PPP makes for you doesn't know about your domain, or that you're running a nameserver on your machine. To get around these problems, I created another file /etc/resolv.conf-working:
nameserver 192.168.1.1
lookup file bind
search fake-domain.org    
In /etc/ppp/ppp.linkup I tell it to overwrite the created resolv.conf with this one:
 ! sh -c "cp /etc/resolv.conf-working /etc/resolv.conf"    
(Add that to the end of the file that you've already created.) This allows all programs running on the machine to be able to use all the good things about a local caching nameserver — things like being able to refer to internal hosts by short name etc.

Other machines on the network

Go to the other machines on your network and set them up with the static IP numbers you assigned above, e.g. the machine wander gets an IP number of 192.168.1.5. All the machines should use 192.168.1.1 for the gateway and use 192.168.1.1 for the DNS server. For more details on DNS, read the excellent O'Reilly book "DNS and BIND"; for more on setting up slightly more complex DNS servers than the one described here, go to the OpenBSD — DNS site maintained by Samiuela LV Taufa.

Setting up DHCP

Above in the DNS setup all internal machines are assigned their own IP numbers. Running DHCP allows guest machines to hook up to the network without fuss. Depending on your comfort level with setting up your other machines, you might also prefer to use DHCP over assigning static IPs. The wireless network also needs DHCP service so guests can get IP addresses use the network. This is what /etc/dhcpd.conf should look like:

#       $OpenBSD: dhcpd.conf,v 1.1 1998/08/19 04:25:45 form Exp $
#
# DHCP server options.
# See dhcpd.conf(5) and dhcpd(8) for more information.
#
shared-network LOCAL-NET {
        option  domain-name "my-domain.org";
        option  domain-name-servers 192.168.1.1;

        subnet 192.168.1.0 netmask 255.255.255.0 {
                option routers 192.168.1.1;

                range 192.168.1.32 192.168.1.127;
        }

        # the laptop with a PCMCIA ethernet card
        host janus {
                fixed-address janus.my-domain.org;
                # This is the Linksys card
                # hardware ethernet 0:e0:98:6:38:0;
                # This is the 3Com card
                hardware ethernet 0:50:4:cd:33:fa;
        }
}

shared-network WIRELESS-NET {
        option  domain-name "my-domain.org";
        option  domain-name-servers 192.168.2.1;

        subnet 192.168.2.0 netmask 255.255.255.0 {
                option routers 192.168.2.1;

                range 192.168.2.32 192.168.2.127;
        }

        # the laptop with the PC-card wireless ethernet
        host janus {
                fixed-address janus-w.my-domain.org;
                hardware ethernet 0:6:25:ab:fe:d;
        }
}    
This will allow up to 96 machines on your internal network, which should be more than sufficient. Machines you know about can be assigned fixed addresses by associating them with the MAC address of the network, as shown above. Create an empty temporary file for dhcpd to use:
bash-2.05# touch /var/db/dhcpd.leases    
If you make any changes to this file, run dhcpd fxp0 rl0 (or whatever your inside network(s) is/are). (Or you can reboot the machine — but that's the Windows way, in the Unix world we prefer to never reboot any machines.)

On boot, the DHCP server needs to know which interfaces it should serve. Create a file /etc/dhcpd.interfaces with their names:

#       $OpenBSD: dhcpd.interfaces,v 1.1 1998/08/19 04:25:45 form Exp $
#
# List of network interfaces served by dhcpd(8).
#
fxp0 rl0

Install "ports"

"Ports" is a *BSD term for a tree of Makefiles for all the software out there that's not part of the standard install. I recommend this highly. It is on CD No. 2 of the OpenBSD 3.3 CD-ROM set as ports.tar.gz. Please read the Ports and Packages page on the OpenBSD web site. You install it by typing (as root)

bash-2.05#  mount /dev/cd0a /mnt
bash-2.05#  cd /usr
bash-2.05#  tar xzf /mnt/ports.tar.gz    
Once you've done this, if you want to install a package, you cd to the appropriate directory and simply type make all install — it will ftp the source from the appopriate site, handle all dependencies, apply any required patches, configure, build and install the tool.

Getting time from the Internet

Set up NTP so that your machine will always have accurate time. Pick two servers from the public NTP server list and make sure /etc/ntp.conf looks like this:

server ntp.server.first
server ntp.server.second    
Since xntpd is not part of the standard install, you have to compile xntpd from source.
bash-2.05#  cd /usr/ports/sysutils/xntpd
bash-2.05#  make all install    
The tools will be installed into /usr/local/sbin/ntpd.

Run ntpdate -b server where you pick a server from the list — this will perform a coarse adjustment of the system clock. The next time the machine reboots, it will sync your clock and record how much your clock drifts.

Setting up other hosts with NTP

On Unix hosts, use the appropriate NTP client; on Linux, it's xntpd. Set them up to use 192.168.1.1 as the NTP server. On Windows, use AboutTime — a free NTP client. In its configuration make sure it uses only SNTP as the protocol, with 192.168.1.1 as the server. Put AboutTime in the Startup folder so it's started automatically.

For more details, go to Robert Mooney's OpenBSD NTP site.

 

Tips and Stuff

I have a useful shell script called pkg_install that's a front-end to pkg_add — here's an example of it being used:
bash-2.05# pkg_install tex
These files match:
gettext-0.10.40.tgz
jadetex-3.11.tgz
latex2html-97.1.tgz
php4-4.0.6p1-gettext-imap-mhash-no_x11-mcrypt-mysql.tgz
php4-4.0.6p1-gettext-imap-mhash-no_x11-mcrypt-postgresql.tgz
php4-4.0.6p1-gettext.tgz
teTeX_texmf-1.0.2.tgz
texi2html-1.64.tgz
textutils-2.0.tgz
bash-2.05# pkg_install -n 4 texi
Using ftp5.usa.openbsd.org/pub/OpenBSD
+ pkg_add -v ftp://ftp5.usa.openbsd.org/pub/OpenBSD/3.3/packages/i386//texi2html-1.64.tgz
Trying to fetch ftp://ftp5.usa.openbsd.org/pub/OpenBSD/3.3/packages/i386//texi2html-1.64.tgz.
Extracting from FTP connection into /var/tmp/instmp.BVMJM29414
>>> ftp -o — ftp://ftp5.usa.openbsd.org/pub/OpenBSD/3.3/packages/i386//texi2html-1.64.tgz...     

It has a list of all the pre-compiled packages that are available. You type in a string and it installs the package. If more than one name matches, it shows you their names. (It uses egrep(1) so you can use regular expressions.) Save it to /usr/local/bin. It handles dependencies by recursively installing them also.

New in this version is in -n flag. The script has a list of mirrors, and this option picks one of the mirrors. (Currently in progress: it needs bash, and it needs some error checking but it works.) Don't forget to edit the file — read http://www.openbsd.org/ftp.html and choosea list of mirrors closest to you.

Running chroot Apache

Starting with 3.3, Apache (httpd) runs in a chroot environment, i.e. the httpd process can only see files under the /var/www/ directory. For serving up static files this doesn't change anything; however CGI scripts and environments need to be modified. In my case, I write CGI programs in a language called Unicon, which is interpreted; this means the interpreter iconx — and any dynamic libraries it uses — must be replicated under /var/www/. (Note: CGI scripts and programs must always be approached with caution, since it's very easy to make a huge security hole by accident. There's a good reason that OpenBSD runs Apache in the chroot environment!)


bash-2.05$ cd /var/www; ls -R
...
bin
conf
tmp
usr

./bin:
sed
sh

./conf:
php.ini

./usr
lib
libexec
local

usr/lib:
libc.so.29.0
libcrypto.so.9.0
libssl.so.7.0
libwrap.so.3.0

usr/libexec:
ld.so

usr/local:
bin
unicon

usr/local/unicon:
bin

usr/local/unicon/bin:
iconx

A common environment for web applications is PHP. Similar considerations apply: for instance, the mail() function of PHP expects to run sendmail; this is not good. Much better for it to use SMTP on the local machine, i.e. connect to port 25 on localhost. Look in php.ini.

Setting up a CVS server

(This section is probably not of interest to most people; you only need this if you want to set up a cvs server so you can put files you're working on under source control. So it's a little terse too.)

The changes I made: added a user and group named cvs. All users of CVS should be in the cvs group. Create a directory for the repository: I put it in /var/cvsroot, you might put it in /home or wherever. This directory should be group writable (group cvs). Add a line to /etc/services:

      cvspserver	2401/tcp		# CVS pserver      
Add this line to /etc/inetd.conf:
      cvspserver	stream	tcp	nowait	root	/usr/bin/cvs cvs -f --allow-root=/var/cvsroot -T /var/tmp pserver      

The server uses /var/tmp as its temp directory instead of /tmp since my root partitions are small, but I always make /var large. Now run cvs init in the cvs repository and restart inetd. Voila! Import your directory of files from a client machine, using a pserver CVSROOT and cvs import.

When importing a large set of files, you might want to put a .cvswrappers file in the directory you're importing so CVS won't try to put RCS ID strings inside your JPEG files etc. The syntax is:

      *.jpg -k 'b'
      *.png -k 'b'
      *.tgz -k 'b'      
Coming soon: using ssh for CVS_RSH.

References

  1. The OpenBSD FAQ — everything written here is covered (perhaps better) in the FAQ.
  2. A Step-by-Step Guide to Building an OpenBSD PPPoE Gateway, with Firewall — tutorial by Real Ouellet on setting up OpenBSD gateway/firewalls. This document inspired parts of the page you are reading.
  3. A Brief Overview of Unix.
  4. Build Your Own PC
    Build A PC
    Build [another] PC
    Build Your Own PC (yet another)
    Build Your PC (in RockvilleLiving)
  5. Putty — a free ssh client for Windows machines.
  6. NiftyTelnet SSH — a free ssh client for MacOS.
  7. NTP on OpenBSD by Robert Mooney.
  8. PPPoE and OpenBSD.
  9. DNS on OpenBSD by Samiuela LV Taufa.

Glossary

(Coming soon.)

 

   


Copyright © 1999-2003 Shamim Mohamed
Creative Commons License
This work is licensed under a Creative Commons License. This document may be redistributed only in its entirety and as long as all copyright notices remain intact. File format changes — e.g. converting from HTML to \TeX — are allowed, but no other modifications are.
Last modified: Tue Mar 15 17:22:53 PST 2005
CVS $Date: 2005/07/23 20:37:44 $ $Revision: 1.22 $
 

 

Many thanks to Bill Graham for excellent feedback and suggestions.