Gargoyle Forum

Just wanted to post a heads-up that I explained (and posted steps for) patching OpenWRT with the Gargoyle patch for ARPNAT wireless bridging, on my blog:

http://projectgus.com/wireless-client-b ... h-openwrt/

Pretty much all the hard work behind that blog post belongs to the Gargoyle patch, so let me know if you don't think there's sufficient kudos there.

The only reason I didn't choose Gargoyle over OpenWRT is that I'm a big fan of ssh + editing text files, and I don't need any of the other nifty features that Gargoyle provides.

The blog post also links to a small patch for 8.09 that Gargoyle might find useful. It's a fix for /lib/network/config.sh not setting the MAC address on the bridge correctly if it's overridden with "option macaddr".

That bug is particularly important if you're using a DHCP client to set the IP on your bridge, because the bridge MAC determines which physical interface that the DHCP request goes out on,

FWIW, I also did some rough benchmarking (which is in the blog post), and I couldn't find any evidence for the OpenWRT dev's statements about ebtables/ARPNAT crippling performance. ARPNAT certainly costs performance, but it didn't seem excessive to me.

Great! Glad you find my patch helpful. Do be aware that it currently only works with the 8.09 branch of OpenWrt.

Also, thanks for the bridge patch for 8.09. I'll test it out within the next couple of days, and it will likely be integrated into Gargoyle.

projectgus
I saw that post & went for gargoyle for simplicity...

It did occur to me that the 'performance' hit is no different to running a NAT router (or indeed any router code) on the units as each packet still has to be examined in software, changed & checksums etc recalculated... The only performance hit would be in comparison to running a pure switched config, which you couldn't do with the wireless in sta mode anyway as it woudl require arpnat....

David

I think you're probably right, David.

However, it t is always possible one of IP routing, NAT, and/or WDS has significantly less load. Due to implementation optimisations, or offloading a bit more of the work to hardware.

If I had all the time in the world, i'd measure them all and compare. However, I'm content that ARPNAT performs fine for my needs, and gives me the simplest network layout.