Using Firewall Builder, Part I
Like Red Hat, SuSE has not yet incorporated Firewall Builder into its official release. SuSE 8.1 RPMs (albeit unofficial contributed ones) are available from the Firewall Builder download page (sourceforge.net/project/showfiles.php?group_id=5314).
You'll need the fwbuilder and libfwbuilder packages, plus one or more of fwbuilder-ipt, fwbuilder-ipf or fwbuilder-pf. You'll also need to have installed these standard SuSE packages: gcc, gdk_pixbuf, glib, glibc-2.2.4, gtk, gtkmm, libsigc++, libstdc++, libxml2, libxslt, libz, openssl-0.9.6b, ucdsnmp and xshared.
Once its packages are installed, you're ready to run Firewall Builder. There's only one command to remember: fwbuilder. You must have the X Window System running to run this command. You don't need to be root, though; in fact, I recommend against it, because you always should avoid needlessly doing things as root.
Once the fwbuilder window is up, you can start defining objects (Figure 2). The whole point of Firewall Builder is to be able to write rules using reusable, drag-and-drop objects, so obviously, objects must exist before rules may be written. Even Firewall Builder's automatic-policy-generating druids require that objects already exist.
You can use objects to represent hosts, networks (identified by IP address and subnet mask), address ranges, TCP/IP services, firewalls (both multi-homed firewalls proper and bastion hosts), time-ranges and groups of other objects. You may define as many or as few objects as you think you'll use in your rulesets. At a minimum, you'll need a firewall object and at least one network or host object. Predefined service objects are provided for many popular TCP/IP services.
Objects are created from the dialogs in Firewall Builder's Insert menu. Figure 2 shows the Insert host dialog for creating a host object. A host object's defining characteristic for rule-creating purposes is its IP address. If you wish to write rules that match hosts by MAC/Ethernet address, you can define that too. As you can see, you may enter the IP address manually or by DNS lookup. The latter feature is handy, but it works only for hosts whose names are resolvable by the system on which Firewall Builder is running.
Figure 3 shows the Insert network dialog. Unlike Insert host, which pops up a separate window, Insert network simply opens a blank New object form in the right-hand portion of the main window. This dialog actually is simpler than the Insert host dialog; all you need to enter are the network's IP address and subnet mask, a name for the network object and, optionally, a comment.
The most complicated object by far is the firewall object. Initial setup isn't too hairy in itself; simply define the firewall's interface or interfaces by IP address and subnet mask. But once the firewall object has been added, and it appears in the list of user-defined objects on the left-hand side of the main fwbuilder window, click on its icon. Five pages (tabs) of information should appear on the right-hand side of the window (Figure 4).
On the General screen, we see the hostname entered in when the firewall object was created. It's important, though, to select appropriate Host OS and Platform options, so Firewall Builder will know which of its compiler engines to use when converting policies to firewall scripts for this firewall.
The SysInfo screen applies only to SNMP data (see Sidebar). The Compile/Install screen is where you can set up automatic installation of your firewall policies. If you intend to install them manually, you can leave this screen alone. Someday, hopefully soon, Firewall Builder will support the automatic transfer and installation of firewall scripts using SSL. As of this writing, however, the fwbd dæmon that must run on a target firewall for this method to work has not been released.
If you leave the Compile/Install screen's Installer option at its default of fwbd, even though this feature isn't yet supported, nothing bad will happen; compiled firewall rules still are saved to your home directory. The Install option in the Rules menu will be grayed out, though. If, on the other hand, you set Installer to Install Script, you then can specify the path to a custom script in the Policy Install Script field, with optional command-line parameters below. The custom script will be executed when you select Rules®Install after compiling a policy.
This method is a handy way to script, for example, an scp command that securely copies your policies to their target firewalls. Sample installation scripts, notably fwb_install, are available under contrib at the Firewall Builder download site (sourceforge.net/project/showfiles.php?group_id=5314).
Regardless of what you set Installer to, Firewall Builder writes the script it compiles for this firewall to a local ASCII file with the same name as the firewall object and a suffix of .fw. For example, scripts generated for the firewall Trillian in Figure 4 are named trillian.fw.
Continuing with Firewall's object properties, the Firewall tab is used to configure options specific to the platform you selected in the General screen, in our case Netfilter/iptables-specific options (Figure 5). The defaults here work fine for many if not most users, but a couple of the options are worth discussing.
In the Global Logging Parameters section you can control how Firewall Builder writes log entries. The default Log Level of 6: Info is okay. Personally, I log only dropped and rejected packets, so I like to bump this up to 4: Warning.
In the Options section of the Firewall screen you may wish to select Assume firewall object is part of Any. The default is for the built-in Source/Destination object Any to be interpreted as “Any host except the firewall”. This is not atypical in firewall policy builders, but it can cause some surprising behaviors.
For example, if the last rule in your policy is a cleanup rule that sets source=any, destination=any, service=any, action=drop and logging=on, you'd expect any attempted connections to the firewall not matching previous rules to be logged and dropped, right? Indeed, they will be dropped, but not because of this rule. They'll be dropped by the INPUT chain's default policy, which Firewall Builder always sets to DROP. This example cleanup rule is triggered only by attempted connections through the firewall. As the firewall itself is not assumed to be part of Any, your cleanup rule is implemented only in the FORWARD chain, not in the INPUT or OUTPUT chains.
Selecting Assume firewall object as part of Any reverses this behavior, and it causes such a cleanup rule to behave the way you'd expect. However, it may complicate other things, such as anti-spoofing rules specific to firewall interfaces. In short, it's a trade-off. My own preference is to leave this option deselected. Then, I either tweak my Firewall Builder scripts to include the log and drop lines to at least the INPUT chain, or I add an extra Firewall input log and drop rule to my policy.
If in doubt, test and tinker with this setting. You can use the Log all dropped packets in the Global Logging Parameters section, but it requires your firewall to have had Netfilter compiled with the Patch-O-Matic Dropped Table patch. This may not be the case if you installed a kernel provided with your Linux distribution.
The last screen of Firewall object properties is Network. This contains settings specific to the Host OS you specified in the General screen. These options directly alter your kernel's behavior; if that frightens you, ignore this screen. But if your firewall is a full-blown, entire-network-protecting firewall rather than simply a hardened host, make sure you set Packet Forwarding to On.
- Resurrecting the Armadillo
- High-Availability Storage with HA-LVM
- March 2015 Issue of Linux Journal: System Administration
- Real-Time Rogue Wireless Access Point Detection with the Raspberry Pi
- DNSMasq, the Pint-Sized Super Dæmon!
- Localhost DNS Cache
- Days Between Dates: the Counting
- The Usability of GNOME
- Linux for Astronomers
- You're the Boss with UBOS