On Tuesday 06 December 2011 15:18:03 Marko Panger - AGB Lab wrote:
Sven,
Please don't mail me in private without a good reason. At least Cc b.a.t.m.a.n@lists.open-mesh.org for batman-adv related things.
This is a very important info. Gigabyte is the provider of the HW and if this is related to the driver I have to report them. We had some problems already with this driver as it doesn't set the cell number correctly.
It sounds more like Ralink... but that is not important right now ;) But maybe you can provide more information about the driver (which exact driver did you use, ...) because some people on the mailing list may have used the same driver and know a workaround/fix.
Could you give me a hint how to provide such information to them ?
The same way you contacted us... with a lot of information. But try to use a small tool to send unicast packets in one direction and look if you can create the same situation without batman-adv involved. IP is not the best solution because it has to send more packets and therefore create more problems when you just want to test a simple scenario. I usually use a small program that directly sends one small ethernet frame to another node over the selected interface while the other side only waits for this packet on a specific interface. So nothing really complicated.
The receiver side has to start the program like this:
$ ./rawsend ra0
and the sender has to start it with the MAC of the receiver
$ ./rawsend ra0 ca:fe:ba:de:ba:be
The receiver will exit the rawsend after it finished the receive.
So start the setup without batman-adv. Try to contact the nodes, restart the one node you always restarted in your tests and try to communicat with it. You can also use broadcast (just use the mac ff:ff:ff:ff:ff:ff as target) to test if unicast is the only problem.
I hope this helps to document the problem. This should allow the driver author to debug/reproduce the whole scenario. But please send us more information when this is not the problem and really batman-adv's fault.
Kind regards, Sven
b.a.t.m.a.n@lists.open-mesh.org