[Prism54-devel] Re: Strange problem with wds in prism54

[Jean-Baptiste Note]

Hello Sergio

"Sergio Ammirata" <ammirata at econointl.com> said :
> He is absolutely right. I have spanning tree protocol enabled on the hostap
> AP.

Additionally, i re-checked, and the frame is a standard 802.3 frame,
with length field 00 2e, then some magic i don't understand but hope the
upper layers will :)

So it seems part of the WDS logic that's in the code is there for
nothing. Can you try and simply remove :

*** islpci_eth.c:
islpci_eth_transmit[456]
		src = skb->data + 6;
		memmove(skb->data, src, skb->len - 6);
		skb_trim(skb, skb->len - 6);

This should be okay at least for ping and spanning-tree packets. Please
test, and leave the debugging on for *all* packets, whether or not the
wds_find_device returns null (this will fill the logs quickly if all the
rest works all right, but this is less than sure...)

This is simple. Now for the delicacies; an explanation of why above
change may not be sufficient :

* maybe we only need to remove this in case wds_find_device returns null
(don't know, for you hitherto all calls returned null...) ; this is the
reason why i'm asking for logging of the packets to be turned on in all
cases.

* maybe the tx path needs some adjustments too ! there's no way to be
sure... (what you can do : with WDS up on the prism54, try and ping one
of your STAtions, and see with ethereal/tcpdump (on the station) if they
receive some frames, and whether those frames are correctly
formatted. There's really no need to do this if the above makes the ping
work !).

This is getting complicated. I can provide more detailed explanations in
case you need it. Maybe we also should only leave the other listeners to
this thread (Jouni, eg) out of the intricacies of this debugging ;)

JB

-- 
Jean-Baptiste Note
+33 (0)6 83 03 42 38
jean-baptiste.note at wanadoo.fr