From: Kristian Hoffmann (khoff_at_pc-intouch.com)
Date: 2002-04-05 01:01:28 UTC
On Thu, 4 Apr 2002, Jouni Malinen wrote:
> At least I haven't tested the driver much with stations being far from
> the AP or at all with that many stations. Station firmware will
> automatically retry sending frames that do not get acknowledged, but
> this should not cause duplications since the retries use the same
> sequence number and receiver should drop duplicates.
I think the stations we use are, well, stupid, and don't drop the duplicate frames. We have 11mbit lucent stations at one or two places and they have little to no dupliactes.
> Did you ping from another stations associated to the same AP or from
> wired network? Is the AP configured as the default gateway for the
> stations (i.e., no bridging needed; just route the packets with normal
> IP routing) or as a bridge between wired network and the stations?
> What was the interval between the first reply and the duplicates
> (i.e., did they came more than second later or more or less
This is pinging from the AP to the station. The duplicate replies are all 2-3ms higher than the last.
> I haven't seen duplicate packets with the latest versions of the
> driver, but I have to admit I haven't tested anything close to your
> setup. It would be quite useful to be able to duplicate the problems
> in my test environment so that I could try to locate the cause for
We didn't have any problems at close range with a small number of stations. I've been looking over the code trying to find a way to prevent frame retransmission by setting the SHORTRETRYLIMIT and LONGRETRYLIMIT RIDs to 0 or 1. Something smaller in any case. I'm hoping that the higher layers (IP) will take care of retransmitting frames that were actually lost in transit.
I should mention, we're using ZCoMax XWL-450 2Mbit stations for most of our customers. With the FW-based AP driver, we don't have as much problem with duplicates. There is actually an acktime setting on the XWL-450 that you can only set on the serial interface. Does something like this exist in the station firmware in the form of a RID? If there is, I haven't been able to find it. We aren't using power management so I'm assuming the DTIM period would have no affect on this problem.