Madwifi Notes

From ME2000

Jump to: navigation, search

All was grand with madwifi until we had a wee get-together to test madwifi at distance. There's a small article here.


Contents

20 Jun 2006 - MadWifi Tests

This evening, I tried a test to a Lucent WavepointII to which my neighbours are hooked up to. I got a 2.4.31 kernel with madwifi 0.9.0 going on the laptop. Couldnt associate. Couldnt get a mac address from the AP. However, using an orinoco card, I got association no problem. 8-12 SNR. pinged the ap no problem. Went back to the CM9, and no association. I then did

 iwlist ath0 ap

and got 4 access points, one of them was the AP I was trying to connect to, with an SNR of 10dB. So, there's something seriously wrong if madwifi wont associate to Staros enabled CM9s, Orinoco's in WPII's with their default settings.

I also tried

diversity 0
rxantenna 2
txantenna 2

and it made no difference.

21 Jun 2006 - More MadWifi Tests

I got a clean 2.4.31 kernel for the laptop (it has an embedded CM9), along with madwifi 0.9.0. Same thing as the ME2000 builds, no association. I had a script to do the settings, will post it up later. Couldnt associate to an ap about a mile away. iwlist showed the ap at about 15dB SNR. An orinoco card had no problem associating.

22 Jun 2006 - Narrowing down the issue

Got the latest svn snapshow. nwid issue is gone, but same issue. I set up an ME2000v2 with an orinoco card in Master mode and, sure enough, the madwifi drivers on the laptop exhibited exactly the same problem. I've added some info to a ticked on madwifi.org here. Athdebug shows the station repeatedly trying to associate but failing. I gotta get into how to read the athdebug output and do a bit more debugging.

There's definetly a problem with madwifi associating to orinoco cards in master mode on 802.11b.

23 Jun 2006 - mailing list question

I've sent a list to the madwifi-users mailing list:

Hi all.

I'm wondering if anyone is successfully associating to hermes chipset based access 
points successfully using madwifi 0.9.0 or later. I have several outdoor links and 
cannot connect to any of the hermes chipset access points. For test purposes, I 
built a new box with a WaveLAN Silver card in master mode (using HermesAP), with 
similar results. WaveLAN card is at firmware version 8.72. Both 0.9.0 and r1648 were 
tested using a Wistron CM9 and a 2.4.31 kernel. I've posted some of my results on 
http://madwifi.org/ticket/428, but I'm only just getting into how to analyse the 
athdebug output :)

I've got a setup that allows me to easily test this now, so if there's any information 
that I can provide that would help nail  down the issue, then let me know.
 
Also, I have madwifi compiled on wrap (x86), routerboard (mips), and Gateworks Avila 
2348-4 (xscale), so I'll be able to test  on all three platforms.

Cheers,
Dave.


It seems there's a patch out. It may be included in the next release. It's part of ticked 698 at http://madwifi.org/ticket/698


Well, I just tried it with great success. Associated to the me2000v2 on my desk, as well as MEELICK1 access point (wavepointII) about a mile away. Madwifi now associates no problem. I took out the annoying "agere beacon" messages, and it's actually usable now. Next is to test to the StarOS CM9 ap's that we have arount the place.


24 Jun 2006 - Success - Gallows to Kilmaley

Had a successful test from the top of Gallows Hill to Kilmaley (about 30 miles). Patched Madwifi on the laptop associated to irishwan.org-km3 at about 15dB, and I was able to pull 200KB/sec from 10.6.1.1

Personal tools