XO-1 wireless tests

[Tim Moody]

I've spent quite a bit of time testing the recent fix to the libertas driver to determine 

1) does it resolve the issue of access points being missed in scans
2) if so, does that help the problem we had in Haiti, where XO-1s could not see and not connect to access points

On the first point, I ran the scripts provided in the call for testing and returned the files produced.  I am told the results indicate that access points are significantly less likely to be missed in a scan and I believe that to be true.

Confirming the second point was more tricky in that I found it difficult to reproduce the failure we saw in Haiti.  I used three APs that I had in Haiti (2 x TPL 4300 and a VT MP02), but in the house everything worked fine, in spite of the fact that it is a very noisy environment, as many as 14 APs can be visible in the NN.  I didn't see any difference between Kevin Gordon's custom OS based on 11.3 and the 12.1 we had in Haiti (I did not try 13.2 on more than one XO-1).  I didn't see any difference that I could pin on the make and model (all had the same firmware - VT SECN 2.0 RC3d).

But channel makes a big difference, though not in the way I had assumed.  I initially set the AP to 9 to avoid the mesh and other APs and there were no connectivity issues.  Of course my AP was right next to the XOs, so had the strongest signal.  I tried 1 and 11 and still connectivity was 100%, even though there are a number of other APs visible that are on those channels.  But the XOs change the active mesh according to their own inscrutable (by me) logic.  When the mesh was active on channel 1 and the AP was set to 1 the likelihood of a successful connection went down to 50% and below. 

I applied James Cameron's fix.  All XOs successfully connected to the AP even though both the AP and the active mesh were on channel 1.  For more certainty I reverted the fix and connectivity went down again, applied it and connectivity went back to 100%.

I should say that I did not try the workaround of turning off the mesh on all the XOs.

So I think the answer is yes to 1) and yes to 2), especially if you are unlucky enough to have your target AP and the active mesh on the same channel.

The tinycore script I used to apply the fix is at 

http://xsce.activitycentral.com/downloads/TC-libertas-patch.zip (I only tested it with 12.1, but it has logic and files for both 12.1 and 13.2)

unzip it to a usb, insert the usb into the XO-1 and boot (no four button or other button combination)

It contains the modified libertas.ko files, the originals of which are at

http://dev.laptop.org/~quozl/12757/dkms/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.laptop.org/pipermail/devel/attachments/20140215/dda58243/attachment.html>