Hello all,
the following is my analysis of the problem I encountered while trying to get
AP mode with ath5k/ath9k to work.
Software versions
=================
Kernel/drivers: wireless-testing tagged master-2009-10-16 (e.g. the latest as of today)
hostapd: 0.6.9
For comparison I used madwifi-0.9.4 with the same hostapd binary
Client: Windows Vista
Configuration
=============
The simplest possible hostapd configuration, no encryption:
interface=wlan0
driver=nl80211
ssid=mac80211ap
hw_mode=g
Observations
============
Initially I just couldn't connect to the ath5k/ath9k AP. To investigate the problem
I run a tcpdump trace on a different machine in monitoring mode. For comparison I
did an equivalent trace with an AP based on Madwifi-0.9.4.
Comparing the two traces I found two suspicious differences:
1) The Authentication Response and the Association Response frames from the ath9k AP
had a sequence number of 0 (zero) while the equivalent frames from Madwifi had
sequence numbers in sync with previous management frames. Both of these frames
were probably injected by hostapd which doesn't know the current sequence number,
right? Who should assign the correct one: mac80211 or the driver?
2) The Association Request (e.g. *FROM* the client) had the "Short Preamble" capability
bit set zo 0 (zero) despite the fact that it was advertising extended rates. I have
no idea why it did choose so. The equivalent frame from the Madwifi trace had the
"Short Preamble" flag set to 1 as expected.
Can anybody shed some light on the problems I found?
Attached are the traces for reference.
Thanks in advance
Joerg
> Can anybody shed some light on the problems I found?
You need some patience, because most the WIFI developers are the
the japanese kernel summit.
--
http://www.holgerschurig.de