Bridged Firewalls

Bridged Firewalls Tom Eastep 2007 2009 Thomas M. Eastep Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover, and with no Back-Cover Texts. A copy of the license is included in the section entitled GNU Free Documentation License. This article applies to Shorewall 4.4 and later.

Background Systems where Shorewall runs normally function as routers. In the context of the Open System Interconnect (OSI) reference model, a router operates at layer 3, Shorewall may also be deployed on a GNU Linux System that acts as a bridge. Bridges are layer 2 devices in the OSI model (think of a bridge as an Ethernet switch). Some differences between routers and bridges are: Routers determine packet destination based on the destination IP address, while bridges route traffic based on the destination MAC address in the Ethernet frame. As a consequence of the first difference, routers can be connected to more than one IP network while a bridge/firewall may be part of only a single network (see below). In most configurations, routers don't forward broadcast packets while bridges do. Section 4 of RFC 1812 describes the conditions under which a router may or must forward broadcasts.

Requirements Note that if you need a bridge but do not need to restrict the traffic through the bridge then any version of Shorewall will work. See the Simple Bridge documentation for details. In order to use Shorewall as a bridging firewall: Your kernel must contain bridge support (CONFIG_BRIDGE=m or CONFIG_BRIDGE=y). Your kernel must contain bridge/netfilter integration (CONFIG_BRIDGE_NETFILTER=y). Your kernel must contain Netfilter physdev match support (CONFIG_IP_NF_MATCH_PHYSDEV=m or CONFIG_IP_NF_MATCH_PHYSDEV=y). Physdev match is standard in the 2.6 kernel series but must be patched into the 2.4 kernels (see http://bridge.sf.net). Bering and Bering uCLibc users must find and install ipt_physdev.o for their distribution and add ipt_physdev to /etc/modules. Your iptables must contain physdev match support and must support multiple instances of '-m physdev' in a single rule. iptables 1.3.6 and later contain this support. You must have the bridge utilities (bridge-utils) package installed.

Application The following diagram shows a typical application of a bridge/firewall. There is already an existing router in place whose internal interface supports a network, and you want to insert a firewall between the router, and the systems in the local network. In the example shown, the network uses RFC 1918 addresses but that is not a requirement; the bridge would work exactly the same if public IP addresses were used (remember that the bridge doesn't deal with IP addresses). There are a several key differences in this setup and a normal Shorewall configuration: The Shorewall system (the Bridge/Firewall) has only a single IP address even though it has two Ethernet interfaces! The IP address is configured on the bridge itself, rather than on either of the network cards. The systems connected to the LAN are configured with the router's IP address (192.168.1.254 in the above diagram) as their default gateway. traceroute doesn't detect the Bridge/Firewall as an intermediate router. If the router runs a DHCP server, the hosts connected to the LAN can use that server without having dhcrelay running on the Bridge/Firewall. Inserting a bridge/firewall between a router and a set of local hosts only works if those local hosts form a single IP network. In the above diagram, all of the hosts in the loc zone are in the 192.168.1.0/24 network. If the router is routing between several local networks through the same physical interface (there are multiple IP networks sharing the same LAN), then inserting a bridge/firewall between the router and the local LAN won't work. There are other possibilities here -- there could be a hub or switch between the router and the Bridge/Firewall and there could be other systems connected to that switch. All of the systems on the local side of the router would still be configured with IP addresses in 192.168.1.0/24 as shown below.

Configuring the Bridge Configuring the bridge itself is quite simple and uses the brctl utility from the bridge-utils package. Bridge configuration information may be found at http://bridge.sf.net. Unfortunately, many Linux distributions don't have good bridge configuration tools, and the network configuration GUIs don't detect the presence of bridge devices. Here is an excerpt from a Debian /etc/network/interfaces file for a two-port bridge with a static IP address:

auto br0 iface br0 inet static address 192.168.1.253 netmask 255.255.255.0 network 192.168.1.0 broadcast 192.168.1.255 pre-up /sbin/ip link set eth0 up pre-up /sbin/ip link set eth1 up pre-up /usr/sbin/brctl addbr br0 pre-up /usr/sbin/brctl addif br0 eth0 pre-up /usr/sbin/brctl addif br0 eth1 pre-down /usr/sbin/brctl delif br0 eth0 pre-down /sbin/ip link set eth0 down pre-down /usr/sbin/brctl delif br0 eth1 pre-down /sbin/ip link set eth1 down post-down /usr/sbin/brctl delbr br0

While it is not a requirement to give the bridge an IP address, doing so allows the bridge/firewall to access other systems and allows the bridge/firewall to be managed remotely. The bridge must also have an IP address for REJECT rules and policies to work correctly — otherwise REJECT behaves the same as DROP. It is also a requirement for bridges to have an IP address if they are part of a bridge/router. Get your bridge configuration working first, including bridge startup at boot, before you configure and start Shorewall. The bridge may have its IP address assigned via DHCP. Here's an example of an /etc/sysconfig/network/ifcfg-br0 file from a SUSE system:

BOOTPROTO='dhcp' REMOTE_IPADDR='' STARTMODE='onboot' UNIQUE='3hqH.MjuOqWfSZ+C' WIRELESS='no' MTU=''

Here's an /etc/sysconfig/network-scripts/ifcfg-br0 file for a Mandriva system:

DEVICE=br0 BOOTPROTO=dhcp ONBOOT=yes

On both the SUSE and Mandriva systems, a separate script is required to configure the bridge itself. Here are scripts that I used on a SUSE 9.1 system.

/etc/sysconfig/network/ifcfg-br0 BOOTPROTO='dhcp' REMOTE_IPADDR='' STARTMODE='onboot' UNIQUE='3hqH.MjuOqWfSZ+C' WIRELESS='no' MTU='' /etc/init.d/bridge#!/bin/sh ################################################################################ # Script to create a bridge # # (c) 2004 - Tom Eastep (teastep@shorewall.net) # # Modify the following variables to match your configuration # #### BEGIN INIT INFO # Provides: bridge # Required-Start: coldplug # Required-Stop: # Default-Start: 2 3 5 # Default-Stop: 0 1 6 # Description: starts and stops a bridge ### END INIT INFO # # chkconfig: 2345 05 89 # description: GRE/IP Tunnel # ################################################################################ PATH=$PATH:/sbin:/usr/sbin:/usr/local/sbin INTERFACES="eth1 eth0" BRIDGE="br0" MODULES="tulip" do_stop() { echo "Stopping Bridge $BRIDGE" brctl delbr $BRIDGE for interface in $INTERFACES; do ip link set $interface down done } do_start() { echo "Starting Bridge $BRIDGE" for module in $MODULES; do modprobe $module done sleep 5 for interface in $INTERFACES; do ip link set $interface up done brctl addbr $BRIDGE for interface in $INTERFACES; do brctl addif $BRIDGE $interface done } case "$1" in start) do_start ;; stop) do_stop ;; restart) do_stop sleep 1 do_start ;; *) echo "Usage: $0 {start|stop|restart}" exit 1 esac exit 0

Axel Westerhold has contributed this example of configuring a bridge with a static IP address on a Fedora System (Core 1 and Core 2 Test 1). Note that these files also configure the bridge itself, so there is no need for a separate bridge config script.

/etc/sysconfig/network-scripts/ifcfg-br0: DEVICE=br0 TYPE=Bridge IPADDR=192.168.50.14 NETMASK=255.255.255.0 ONBOOT=yes /etc/sysconfig/network-scripts/ifcfg-eth0:DEVICE=eth0 TYPE=ETHER BRIDGE=br0 ONBOOT=yes/etc/sysconfig/network-scripts/ifcfg-eth1:DEVICE=eth1 TYPE=ETHER BRIDGE=br0 ONBOOT=yes

Florin Grad at Mandriva provides this script for configuring a bridge:

#!/bin/sh # chkconfig: 2345 05 89 # description: Layer 2 Bridge # [ -f /etc/sysconfig/bridge ] && . /etc/sysconfig/bridge PATH=$PATH:/sbin:/usr/sbin:/usr/local/sbin do_stop() { echo "Stopping Bridge" for i in $INTERFACES $BRIDGE_INTERFACE ; do ip link set $i down done brctl delbr $BRIDGE_INTERFACE } do_start() { echo "Starting Bridge" for i in $INTERFACES ; do ip link set $i up done brctl addbr br0 for i in $INTERFACES ; do ip link set $i up brctl addif br0 $i done ifup $BRIDGE_INTERFACE } case "$1" in start) do_start ;; stop) do_stop ;; restart) do_stop sleep 1 do_start ;; *) echo "Usage: $0 {start|stop|restart}" exit 1 esac exit 0 The /etc/sysconfig/bridge file: BRIDGE_INTERFACE=br0 #The name of your Bridge INTERFACES="eth0 eth1" #The physical interfaces to be bridged

Andrzej Szelachowski contributed the following.

Here is how I configured bridge in Slackware: 1) I had to compile bridge-utils (It's not in the standard distribution) 2) I've created rc.bridge in /etc/rc.d: ######################### #! /bin/sh ifconfig eth0 0.0.0.0 ifconfig eth1 0.0.0.0 #ifconfig lo 127.0.0.1 #this line should be uncommented if you don't use rc.inet1 brctl addbr most brctl addif most eth0 brctl addif most eth1 ifconfig most 192.168.1.31 netmask 255.255.255.0 up #route add default gw 192.168.1.1 metric 1 #this line should be uncommented if #you don't use rc.inet1 ######################### 3) I made rc.bridge executable and added the following line to /etc/rc.d/rc.local /etc/rc.d/rc.bridge

Joshua Schmidlkofer writes:

Bridge Setup for Gentoo #install bridge-utils emerge bridge-utils ## create a link for net.br0 cd /etc/init.d ln -s net.eth0 net.br0 # Remove net.eth*, add net.br0 and bridge. rc-update del net.eth0 rc-update del net.eth1 rc-update add net.br0 default rc-update add bridge boot /etc/conf.d/bridge: #bridge contains the name of each bridge you want created. bridge="br0" # bridge_<bridge>_devices contains the devices to use at bridge startup. bridge_br0_devices="eth0 eth1" /etc/conf.d/net iface_br0="10.0.0.1 broadcast 10.0.0.255 netmask 255.255.255.0" #for dhcp: #iface_br0="dhcp" #comment this out if you use dhcp. gateway="eth0/10.0.0.1"

Users who successfully configure bridges on other distributions, with static or dynamic IP addresses, are encouraged to send me their configuration so I can post it here.

Configuring Shorewall As described above, Shorewall bridge support requires the physdev match feature of Netfilter/iptables. Physdev match allows rules to be triggered based on the bridge port that a packet arrived on and/or the bridge port that a packet will be sent over. The latter has proved to be problematic because it requires that the evaluation of rules be deferred until the destination bridge port is known. This deferral has the unfortunate side effect that it makes IPSEC Netfilter filtration incompatible with bridges. To work around this problem, in kernel version 2.6.20 the Netfilter developers decided to remove the deferred processing in two cases: When a packet being sent through a bridge entered the firewall on another interface and was being forwarded to the bridge. When a packet originating on the firewall itself is being sent through a bridge. Notice that physdev match was only weakened with respect to the destination bridge port -- it remains fully functional with respect to the source bridge port. To deal with the asymmetric nature of the new physdev match, Shorewall supports a new type of zone - a Bridge Port (BP) zone. Bridge port zones have a number of restrictions: BP zones may only be associated with bridge ports. All ports associated with a given BP zone must be on the same bridge. Policies from a non-BP zone to a BP are disallowed. Rules where the SOURCE is a non-BP zone and the DEST is a BP zone are disallowed. In /etc/shorewall/zones, BP zones are specified using the bport (or bport4) keyword. If your version of shorewall.conf contains the BRIDGING option, it must be set to No. In the scenario pictured above, there would probably be two BP zones defined -- one for the Internet and one for the local LAN so in /etc/shorewall/zones: #ZONE TYPE OPTIONS fw firewall world ipv4 net:world bport loc:world bport #LAST LINE - ADD YOUR ENTRIES ABOVE THIS ONE - DO NOT REMOVE The world zone can be used when defining rules whose source zone is the firewall itself (remember that fw-><BP zone> rules are not allowed). A conventional two-zone policy file is appropriate here — /etc/shorewall/policy: #SOURCE DEST POLICY LOG LIMIT:BURST loc net ACCEPT net all DROP info all all REJECT info #LAST LINE - ADD YOUR ENTRIES ABOVE THIS ONE - DO NOT REMOVE In /etc/shorewall/shorewall.conf: IMPLICIT_CONTINUE=No Bridges use a special syntax in /etc/shorewall/interfaces. Assuming that the router is connected to eth0 and the switch to eth1: #ZONE INTERFACE BROADCAST OPTIONS world br0 detect bridge net br0:eth0 loc br0:eth1 #LAST LINE -- ADD YOUR ENTRIES BEFORE THIS ONE -- DO NOT REMOVE The world zone is associated with the bridge itself which is defined with the bridge option. Bridge port entries may not have any OPTIONS. When a bridge is configured without an IP address, the option must also be specified. When Shorewall is stopped, you want to allow only local traffic through the bridge — /etc/shorewall/routestopped: #INTERFACE HOST(S) OPTIONS br0 192.168.1.0/24 routeback #LAST LINE -- ADD YOUR ENTRIES BEFORE THIS ONE -- DO NOT REMOVE The /etc/shorewall/rules file from the two-interface sample is a good place to start for defining a set of firewall rules.

Multiple Bridges with Wildcard Ports It is sometimes required to configure multiple bridges on a single firewall/gateway. The following seemingly valid configuration results in a compile-time error ERROR: Duplicate Interface Name (p+) /etc/shorewall/zones: #ZONE TYPE fw firewall world ipv4 z1:world bport4 z2:world bport4 /etc/shorewall/interfaces: #ZONE INTERFACE BROADCAST OPTIONS world br0 - bridge world br1 - bridge z1 br0:p+ z2 br1:p+ The reason is that the Shorewall implementation requires each bridge port to have a unique name. The interface option was added in Shorewall 4.4.4 to work around this problem. The above configuration may be defined using the following in /etc/shorewall/interfaces: #ZONE INTERFACE BROADCAST OPTIONS world br0 - bridge world br1 - bridge z1 br0:x+ - physical=p+ z2 br1:y+ - physical=p+ In this configuration, 'x+' is the logical name for ports p+ on bridge br0 while 'y+' is the logical name for ports p+ on bridge br1. If you need to refer to a particular port on br1 (for example p1023), you write it as y1023; Shorewall will translate that name to p1023 when needed. Example from /etc/shorewall/rules: #ACTION SOURCE DEST PROTO DEST # PORT(S) REJECT z1:x1023 z1:x1024 tcp 1234

Combination Router/Bridge A system running Shorewall doesn't have to be exclusively a bridge or a router -- it can act as both, which is also know as a brouter. Here's an example: This is basically the same setup as shown in the Shorewall Setup Guide with the exception that the DMZ is bridged rather than using Proxy ARP. Changes in the configuration shown in the Setup Guide are as follows: The /etc/shorewall/proxyarp file is empty in this configuration. The /etc/shorewall/zones file is modified: #ZONE TYPE OPTIONS fw firewall pub ipv4 #zone containing all public addresses net:pub bport4 dmz:pub bport4 loc ipv4 The /etc/shorewall/interfaces file is as follows:#ZONE INTERFACE BROADCAST OPTIONS pub br0 detect routefilter,bridge net br0:eth0 dmz br0:eth2 loc eth1 detect The DMZ systems need a route to the 192.168.201.0/24 network via 192.0.2.176 to enable them to communicate with the local network. This configuration does not support separate fw->dmz and fw->net policies/rules; similarly, it does not support separate loc->dmz and loc->net rules. This will make it a bit trickier to configure the rules. I suggest something like the following: /etc/shorewall/params: SERVERS=192.0.2.177,192.0.2.178 #IP Addresses of hosts in the DMZ DMZ=pub:$SERVERS #Use in place of 'dmz' in rule DEST NET=pub:!$SERVERS #Use in place of 'net' in rule DEST /etc/shorewall/policy: #SOURCE DEST POLICY LEVEL loc pub ACCEPT loc $FW REJECT info loc all REJECT info $FW pub REJECT info $FW loc REJECT info $FW all REJECT info dmz net REJECT info dmz $FW REJECT info dmz loc REJECT info dmz all REJECT info net dmz DROP info net $FW DROP info net loc DROP info net all DROP info # THE FOLLOWING POLICY MUST BE LAST all all REJECT info /etc/shorewall/rules: #ACTION SOURCE DEST PROTO DEST SOURCE # PORT(S) PORT(S) ACCEPT all all icmp 8 ACCEPT loc $DMZ tcp 25,53,80,443,... ACCEPT loc $DMZ udp 53 ACCEPT loc $NET ACCEPT $FW $DMZ udp 53 ACCEPT $FW $DMZ tcp 53

Using Back-to-back veth Devices to Interface with a Bridge Beginning with Shorewall 4.4.26, Shorewall has limited support for using back-to-back veth devices to interface with a bridge. This approach has the advantage that traffic between any pair of zones can be filtered. The disadvantage is the complexity of the approach. This configuration is shown in the following diagram. In this configuration, veth0 is assigned the internal IP address; br0 does not have an IP address. Traffic from the net and fw zones to the zonei zones goes thru veth0->veth1->ethN->. Traffic from the zonei zones to the fw and net zones takes the reverse path: ethN->veth1->veth0. As a consequence, traffic between net,fw and zonei goes through Netfilter twice: once in the routed firewall (eth0,veth0) and once in the bridged firewall (eth1,eth2,eth3,veth1). The back-to-back veth devices (veth0 and veth1) are created using this command: ip link add type veth If you have veth devices and want to assign specific names to the created devices, use this format: ip link add name FOO type veth peer name BAR Here's an /etc/network/interfaces stanza that configures veth0, veth1 and the bridge: auto veth0 iface veth0 inet static address 10.10.10.1 netmask 255.255.255.0 network 10.10.10.0 broadcast 10.10.10.255 pre-up /sbin/ip link add name veth0 type veth peer name veth1 pre-up /sbin/ip link set eth1 up pre-up /sbin/ip link set eth2 up pre-up /sbin/ip link set eth3 up pre-up /sbin/ip link set veth1 up pre-up /usr/sbin/brctl addbr br0 pre-up /usr/sbin/brctl addif br0 eth1 pre-up /usr/sbin/brctl addif br0 eth2 pre-up /usr/sbin/brctl addif br0 eth3 pre-up /usr/sbin/brctl addif br0 veth1 pre-down /usr/sbin/brctl delif br0 eth1 pre-down /sbin/ip link set eth2 down pre-down /usr/sbin/brctl delif br0 eth2 pre-down /sbin/ip link set eth2 down pre-down /usr/sbin/brctl delif br0 eth3 pre-down /sbin/ip link set eth3 down pre-down /usr/sbin/brctl delif br0 veth1 pre-down /sbin/ip link set veth1 down post-down /usr/sbin/brctl delbr br0 post-down /sbin/ip link del veth0 In shorewall.conf (5), we need this: ZONE_BITS=3 This does two things: It enables automatic packet marking. It allows up to 8 marked zones (2**3). Zones are marked unless they have in the OPTIONS column of their entry in shorewall-zones (5). Packets originating in a marked zone have a mark assigned automatically by Shorewall. For this configuration, we need several additional zones as shown here: #ZONE TYPE OPTIONS IN OUT # OPTIONS OPTIONS fw firewall net ipv4 zone1 bport zone2 bport zone3 bport loc ipv4 nomark col ipv4 nomark col is loc spelled backward. #ZONE INTERFACES BROADCAST OPTIONS net eth0 ... - br0 ... zone1 br0:eth1 ... zone2 br0:eth2 ... zone3 br0:eth3 ... loc veth0 ... col br0:veth1 ... Several things to note here We have defined two unmarked zones: loc and col. This allows traffic from the zonei zones to the fw and net zones to retain the mark of their originating bport zones. It also allows traffic from the fw and net zones to the zonei zones to retain the fw and net marks respectively. That means that traffic entering the bridge on veth1 will have a different mark value, depending on whether it originated in the net zone or in the fw zone. Similarly, traffic arriving on the veth0 interface will have a mark that indicates which of the zonei zones each packet originated on. The basic idea here is that we want to filter traffic to the zonei zones as it leaves veth1 and we want to filter traffic from those zones as it leaves veth0. So we use this type of polices: #SOURCE DEST POLICY fw loc ACCEPT net loc ACCEPT net all DROP:info zone1 col ACCEPT zone2 col ACCEPT zone3 col ACCEPT all all REJECT:info Rules allowing traffic from the net to zone2 look like this: #ACTION SOURCE DEST PROTO DEST SOURCE ORIGINAL RATE USER/ MARK # PORT(S) PORT(S) DEST LIMIT GROUP ACCEPT col zone2 tcp 22 - - - - net or more compactly: #ACTION SOURCE DEST PROTO DEST # PORT(S) ACCEPT col zone2 tcp 22 ; mark=net Similarly, rules allowing traffic from the firewall to zone3: #ACTION SOURCE DEST PROTO DEST # PORT(S) ACCEPT col zone3 tcp 22 ; mark=fw The important point here is that, when ZONE_BITS is non-zero, you are allowed to place zone names in the MARK column. Shorewall will automatically replae the name with the zone's mark value. Suppose that you want to forward tcp port 80 to 192.168.4.45 in zone3: #ACTION SOURCE DEST PROTO DEST SOURCE ORIGINAL RATE USER/ MARK # PORT(S) PORT(S) DEST LIMIT GROUP DNAT- net loc:172.168.4.45 tcp 80 ACCEPT col zone3:172.168.4.45 tcp 80 - - - - net Rules allowing traffic from the zonei zones to the net zone look like this: #ACTION SOURCE DEST PROTO DEST SOURCE ORIGINAL RATE USER/ MARK # PORT(S) PORT(S) DEST LIMIT GROUP ACCEPT loc net tcp 21 - - - - zone1 And to the firewall: #ACTION SOURCE DEST PROTO DEST SOURCE ORIGINAL RATE USER/ MARK # PORT(S) PORT(S) DEST LIMIT GROUP ACCEPT zone2 col tcp - - - - zone2

Limitations Bridging doesn't work with some wireless cards — see http://bridge.sf.net.