OS/2 TCP/IP Filtering
Configuring the OS/2 Packet filter
One might ask himself why one would use an expensive OS/2-machine as a firewall, instead of a simple Linux-box, with ipchains/iptables or some other packet filtering tool.
The answer of this lies deep within the structure of OS/2 itself. From the beginning, OS/2 has been thought out as an operating system which should be rock-stable by itself, perform excellently concerning network-connections and be able to open many sockets at the same time without too much of an overhead.
This made OS/2 to be widely known as one of the most stable operating-systems ever developed. When correctly installed and tuned, according to my personal opinion, nothing meets uptime-demands as completely as a finely configured Warp installation, and if there are three things which are typical characteristics of “a good firewall”, they are stability, security and of course, availability. OS/2 meets all of those requirements. Above that, configuring a warp-system is a lot simpler than having to configure yet another Unix-box. So the question is not, why should I use an OS/2-firewall, the question is, why not consider it?
In this paper we will introduce a way to use the packet-filtering-procedures built into the TCP/IP stack of OS/2 itself. This method has been inherited from the older AIX TCP/IP stack. It is already widely used, however, documentation for this kind of packet-filtering doesn’t grow as prosperous on the net as it should. When I first started out at building my OS/2 firewall, I had to search the web for quite a bit of time until I had enough information to get me started, while still knowing what I was doing. So I decided to take my notes together and wrote a paper on the subject, so everyone can start building their firewall with all of the info at hand.
Please understand that this is only a very basic configuration, and needs lots of tuning before it can truly be used in production environments. With this in mind, have fun.
Configuring the Filter
Enabling the correct drivers
First of all, make sure that the following lines are in your config.sys. If they are not present, please add them.
FWIP.SYS is the devic e which will receive packets, check them to the current rule set and choose whether or not to discard them. IPSEC.SYS is used for the IPSEC implementation on OS/2, however it is needed for the installation of your firewall, otherwise you will not be able to use the logging-feature.
Creating the configuration-file and securing the network-devices
The next step in building our packet filter consists of the creation of a configuration file, in which we will list all IP-addresses of the interfaces we need to secure. All these interfaces will go into a deny-all stance after the next reboot. For example, if you have 3 NIC’s (Network Interface Cards) in your machine, with ip’s 22.214.171.124, 192.168.35.21 & 126.96.36.199, create a file called:
188.8.131.52 192.168.35.21 184.108.40.206
You may choose not to add certain interfaces here. For example, if you only want your internet-connection to be filtered, just add the IP address you have on the internet, but not the ip’s of any of the other cards.
Now we are going to create a rule-set config file. We will add rule sets after the next reboot, however you may choose to do so at this stage. My experience is that it is easier to build the rules while you have everything denied, so you can have a better look at the effect it has on the firewall (in production-environments however this may cause your boss to start yelling “why can’t I read my mail” instead of him giving you your well-deserved promotion). If you create the rule sets before rebooting, you will only see the effect of all of them after the next reboot. Both of those methods have their own advantages and disadvantages, though.
To create the rule-set file, make an empty file at:
Now, we reboot the machine before we proceed to the next step. This is done to load-up the drivers we added in config.sys.
Initializing the OS/2 Packet filter
After those preliminary jobs are done... we are ready to boot our firewall For this, you open up an OS/2-console-window and you enter the following command:
cfgfilt –u –i
CFGFILT is a program better known as the “Packet Filter Rules Dump Facility”. The parameter “-u” tells it to update all of the filter-rules, “-i” just initialises the filter-device. Now it starts up and you get a whole lot of interesting status-info about your packet-filter. Normally you should have to bring up the firewall yourself, but OS/2 anticipates this, and brings it up using inetcfg –s firewall 1. Inetcfg is used to configure parameters of the TCP/IP-stack. The only relevant portion of its usage here is the “firewall” parameter, which turns the firewall on or off. This is in a binary format, so 0 means “off”, 1 means “on”. You can also get the status of your firewall with the parameter “–g firewall”.
Now the firewall is up and running and all traffic is blocked, unless you chose to specify your rules earlier. Otherwise, now is the time to start defining rules.
As mentioned earlier, all rules are to be saved in the file \MPTN\ETC\SECURITY\FWFILTRS.CNF. Just open the file in any editor. Rules have a special format which offers a lot of flexibility to the administrator. You can filter packets using many different parameters.
A rule looks like this:
todo a.a.a.a. b.b.b.b c.c.c.c. d.d.d.d e f g h i j k l
And here’s the meaning of it all:
|Todo||The action the rule should take. This can either be “permit” or “deny”. In case enter “permit” here, packets which match the rule, will be allowed passage through the firewall. When you enter “deny”, packets will be blocked.|
|a.a.a.a.||This is the source IP address of the packet. For example 220.127.116.11.
Please keep in mind the ending-dot. So it’s 18.104.22.168. when you enter it.
|b.b.b.b.||Source address-mask, eg. 255.255.255.0|
|c.c.c.c.||Destination IP address of the packet.|
|E||The protocol-type of the packet. The filter supports the following protocols:
|F||Here you can select the source-port. Yet again you have a wide range of options:
packets to any port number match the rule
Exact match to the port specified by “port”
Any port except for the one specified by “port”
When it matches any port number lower than “port”
When it matches any port number greater than “port”
Any port number less than or equal to “port”
Any port number greater than or equal to “port”
|g||Here you specify the interface-type. It can be “secure” or “non-secure”
|h||The routing of the packets... “local”, “route”, or “both”
|i||Direction... although closely related to the routing, it still means something different. It doesn’t mind its hosts, but just looks at the interface. Can be “inbound”, “outbound” or “both”
|k||Fragmentation. As you know, large packets can be fragmented on their way to your host. This parameter of your rule defines what to do with such a euhmmm... c hopped packet ;)
|l||Tunnelling... through which tunnel should the packet be sent (if tunnel set)
After you created all of those rules, update the filtering rules by issuing the following command on an OS/2-console:
All rules should now be updated and activated. You can, if you wish, run a plain vanilla “cfgfilt” on an OS/2-prompt to take a look at the rules. For interfaces which are secured, the last rule always is a “deny all”, this means that all packets not explicitly allowed, will be forbidden.
One of the most interesting aspects of a firewall is that you can make it log things that aren’t really welcome. Thanks to the IPSec integration in OS/2, the packet filter of the OS/2 TCP/IP stack also has logging capabilities.
To configure logging, first of all, create a file \MPTN\ETC\FWLOG.CNF
In this file, you can define the level of logging you require. It only contains one line:
level = xx
Where xx is the level of logging you require. You have many options from 10 to 50, where 10 logs the most, 50 the least.
- 10 Debug Mode – log all
- 20 Informational Mode – logg info, warning, error, alert msg’s
- 30 Warning Mode – logg warning, error, alert msg’s
- 40 Error Mode – log error, alert msg’s
- 50 Alert Mode – log alert messages only
Next, we should decide which rules should be logged. Like mentioned in the previous chapter, you add l=y to all of the rules which you want logged. This should be done in \MPTN\ETC\SECURITY\FWFILTRS.CNF Now open an OS/2 console window, and run the FSSD program, by just typing “fssd”. This will boot up the packet-filter syslogd. A log file will be created in \MPTN\ETC . Normally the filename will start with FW, followed by the date it was created, like 1225 for Christmas. Of course, logging is not yet being done, as you haven’t reloaded the rules yet to which you added the logging-parameter. This is to be done by the command
cfgfilt –u –d
Now all rules which you have designated to be logged, will be logged. You may inspect the logfiles by going into the \MPTN\ETC directory (or putting it in your path of course) and running the fwlslog binary. As a parameter, you give “file=filename”, for example:
to take a look at what happened on Christmas-day.