This is an old revision of the document!
Table of Contents
Introduction
We have a pair of WRT54GL's running DD-WRT as a WDS bridge. It works great, except we both have our own cable modem and routers. Each router functions as a DHCP server. Because we're bridged, this means that we have two competing DHCP servers on the same LAN, so sometimes a machine will pick the other router as it's gateway. This means that Internet traffic would be crammed through our tenuous bridge, making much sadness.
The correct solution is to have separate subnets with routing and blah blah blah.
Screw that. The quick and dirty solution is just to ban DHCP traffic from the bridge.
HOWTO
Go to Administration | Commands, and paste this:
echo "begin-base64 644 -" > /tmp/ebt_ip.o.gz.u64 echo "H4sIADwAAAACA5VWz28bVRD+dteON20KG9eq3BJUB21UVy1m2+RQRCO5cX5w" >> /tmp/ebt_ip.o.gz.u64 echo "yMGHHpAQcr32Eq9wHMveIFAPNSkHDq7wIb1H5R+p0gpx7J8QtaH8EBfuSGbm" >> /tmp/ebt_ip.o.gz.u64 echo "vbdh8whEHelp9ps3883Mm/ecPFhZXzUMA7EYsPEPAv5Ikc1BdUGgIt7CWSQl" >> /tmp/ebt_ip.o.gz.u64 echo "i5ejFJY8B2VntjrAecIFPByyLUM2B4cjA43RGDveN6g4HJMD7CxeH9nvo5KT" >> /tmp/ebt_ip.o.gz.u64 echo "XK9HNva9KsUxPo/fiKvyyMZ7N5hrmrgmVUxHcV0hrmm8OrK3FBfbJmC6FnyK" >> /tmp/ebt_ip.o.gz.u64 echo "M92P4At7AUuPBsRZwHMPWBq+TXUOHBN5DJzZsg8T5uM8DnfHqHi4ZMHMmpjL" >> /tmp/ebt_ip.o.gz.u64 echo "+7hWaCKNQW62CFzCy91rTgUW1VSg2gykXdZpwgae3FpDWeRap9oc6qeCPYH5" >> /tmp/ebt_ip.o.gz.u64 echo "+6LqDZQnd0qeCyqPqWI4D5+vQzXHua6rXBvqTIsqlzzffS9LfiZpORc+y8Oh" >> /tmp/ebt_ip.o.gz.u64 echo "oWYD/Dycgj90nWXqaUXkuE887B/bc1gRfGzL8YwdiBmniIf5TMU1hV+GjohZ" >> /tmp/ebt_ip.o.gz.u64 echo "QwbLgsvGgcX+sd1K2A2aCe/xtzybH76fIdsNmtPv47Rr49muTfh9NbcU9kXd" >> /tmp/ebt_ip.o.gz.u64 echo "V1W/WboLT8lvjIee6wR0ZtLOtbnOipgN4x8pBfN58L0zYl47Xh73HLbpdyR7" >> /tmp/ebt_ip.o.gz.u64 echo "dEd21N1l31e78bwpXtTmqP14RjnyUXMS+2yXve9x5beBi7Q+vGKjQL1+655s" >> /tmp/ebt_ip.o.gz.u64 echo "Oy5hF6fKx+a/bX9l8ARvIO/SuvwfXMn9gvX/+wun7LfDRtDpB4tr1XV8EfQ6" >> /tmp/ebt_ip.o.gz.u64 echo "Qbv2ZdDrh1udxZulhdK8x86l/tebUd0nHfWkbsVfvaBdioKvIvnVrEd1lPx+" >> /tmp/ebt_ip.o.gz.u64 echo "n+FG2Pl8S9q7zR5Km83A394o1f1w/iahrabYfwO5LN4aMKFwmVY10Z96wHiH" >> /tmp/ebt_ip.o.gz.u64 echo "1mQi7hMb4LFNqfiUuOvAVYVjipElOXW+OY3vJwrOKz4rwfeB0jHfC+3sY74F" >> /tmp/ebt_ip.o.gz.u64 echo "2UPXTPjlE37Mx/iOqi+WX8lvoPmxLGr1Hdj8VmR9mUR9qxpfK3W8vnjvrvo2" >> /tmp/ebt_ip.o.gz.u64 echo "E37uCXmnE72y0B8XfHYCX5KLZYaKOiB9TtWdUmczqfHRm8GnJ/DpYhyr7c9x" >> /tmp/ebt_ip.o.gz.u64 echo "cs/SoixYGk5reFLDZzQ8oeGM8OB3mBO/fwb99sh3F+M57Q7Ma3hVvyOBH9XC" >> /tmp/ebt_ip.o.gz.u64 echo "bqmBsBNGNXop2+0AjXZQ72x3Y1irRa2wHyOO4BfXj4JebbMeNVrCtN3RjOLu" >> /tmp/ebt_ip.o.gz.u64 echo "pUx5j1jPGFJ/p/QLpfdMqelHGGm6SEXSE1OyzzQtjzGtW4xpkLcZn5N98/6y" >> /tmp/ebt_ip.o.gz.u64 echo "2i/y7InnutKc31T/qjAuKF1W+p7SfwMbOFhxAAkAAA==" >> /tmp/ebt_ip.o.gz.u64 echo "====" >> /tmp/ebt_ip.o.gz.u64 uudecode /tmp/ebt_ip.o.gz.u64 | gzip -cd > /tmp/ebt_ip.o insmod ebtables insmod ebtable_filter insmod /tmp/ebt_ip.o ebtables -I INPUT -i tap0 -p IPv4 --ip-protocol udp --ip-destination-port 67:68 -j DROP
Then hit "Save commands". Finally, reboot. You can check to see if it works by telnetting in and typing "ebtables -L". You should see:
Bridge table: filter Bridge chain: INPUT, entries: 1, policy: ACCEPT -p IPv4 -i tap0 --ip-proto udp --ip-dport 67:68 -j DROP Bridge chain: FORWARD, entries: 0, policy: ACCEPT Bridge chain: OUTPUT, entries: 0, policy: ACCEPT
What does that crap do?
Because we need to hook into the bridging system, iptables is insufficient. This thread explains that we need to use ebtables to do that. This requires 2 included modules: ebtables
and ebtable_filter
, plus one module that's not included: ebt_ip.o
(attached in the thread). How can we add this extra module, given that the flash filesystem is read only? Answer: we crap it into the NVRAM as a series of commands.
I took the original ebt_ip.o, and ran "gzip -c ebt_ip.o | uuencode -m -", which means "compress with gzip, then uuencode with base-64 mode". This turns the binary file into safe ASCII. We then wrap all this ASCII in that giant pile of "echo … » /tmp/ebt_ip.o.gz.u64" commands. We then decode it back into binary with "gzip -cd … | uudecode -". Then we load the required modules: ebtables
, ebtable_filter
, and our hacked up /tmp/ebt_ip.o
. Finally, we input the ebtables rule to drop all the DHCP packets it sees.
WRONG INFORMATION
This was my first pass; it didn't work. Preserved for comedy value.
After some trial and error, I found the iptables rules that would do that:
iptables -t mangle -I PREROUTING 1 -p UDP --sport 67 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --sport 68 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --dport 67 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --dport 68 -j DROP
The only question then is how to make these rules get applied automatically at bootup. This article covers that: just use the 'rc_firewall' value in the nvram. At the router's command prompt:
nvram set rc_firewall="iptables -t mangle -I PREROUTING 1 -p UDP --sport 67 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --sport 68 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --dport 67 -j DROP iptables -t mangle -I PREROUTING 1 -p UDP --dport 68 -j DROP" nvram commit
Reboot, and presto, the bridge is eating all DHCP traffic:
# iptables -t mangle -vL Chain PREROUTING (policy ACCEPT 262 packets, 50464 bytes) pkts bytes target prot opt in out source destination 0 0 DROP udp -- any any anywhere anywhere udp dpt:bootpc 37 14776 DROP udp -- any any anywhere anywhere udp dpt:bootps <=== Counter shows 37 DHCP packets have been eaten! 0 0 DROP udp -- any any anywhere anywhere udp spt:bootpc 0 0 DROP udp -- any any anywhere anywhere udp spt:bootps