Hostapd and ath9k - AP not working
Stephen Gutknecht hilltx
Sat Mar 28 04:33:56 PDT 2009
2009/3/27 Helena Zhang <helena at astbb.com>:
> Thanks for the help Jouni. I switched to 2.6.29 and it works better now (I
> don't know why but wireless testing does not work). Now I have a new
> problem. When both sides are in a/b/g mode, the connection works fine.
> However, when I switch to na/ng mode, the connection is very fragile. The
> sta can associate with the ap and stays associated, but they cannot ping
> each other unless I put the antenna in a very specific place (I have to
> manually adjust for each connection carefully since the place changes with
> every channel), which is really strange because the two boxes are only two
> feet from each other. Even when I do get them to ping each other the ping
> time is very unstable. When I do iperf, the maximum throughput I managed to
> get is 21 Mb on channel 44, HT40+, which is a lot less than what it should
> be. Has anyone else seen this problem before? What could I be doing wrong?
Yes, users on OpenWrt are seeing similar performance with newest ath9k
driver from compat-wireless. Some kind of timing issues or what not.
The same hardware run with a proprietary binary-only madwifi driver
for 2.6.15 runs far faster.
Information is a bit scattered on the issue, please take the time to
review it all:
1. Up to date testing and feedback thread:
2. Original thread where the ath9k was first ported to AHB bus
drivers, some reference to the original performance achieved. Back in
January I think we were lucky to be getting 5Mbp/s throughput. It has
since improved along lines with the 20Mbp/s you see.
3. A thread on the need to open bugs on kernel.org that hasn't gotten
the attention it needs. We need to be more formal in submitting these
problems now that multiple people confirm them.
Regardless if you use OpenWrt, I think you will find there are at
least a dozen somewhat active people using hostap + ath9k, so a good
More information about the Hostap