Subject: Re: ath seems still buggy
To: Marcin Jessa <>
From: Sam Leffler <>
List: current-users
Date: 10/19/2005 09:21:04
Marcin Jessa wrote:
> On Wed, 19 Oct 2005 10:09:35 -0400
> Thor Lancelot Simon <> wrote:
>>On Wed, Oct 19, 2005 at 09:28:35AM +0000, Marcin Jessa wrote:
>>>On Tue, 18 Oct 2005 21:29:35 -0400
>>>Thor Lancelot Simon <> wrote:
>>>: In fact, I'm typing this through a Soekris with an Atheros
>>>wireless : module in hostap mode right now.
>>>I did not mean occasional usage on a laptop but in real life
>>>enviroment running as AP with clients connected to it.
>>What part of "in hostap mode" did you miss?
> What part of "real life enviroment" did you miss?
> I am not talking of home-brewed small scale soekris boxes which are
> useless/too slow for any modern WLAN Access Point. 
> An example of sysctl values letting boost up connection speed to
> 6Mbit/s on FreeBSD that I cannot setup on NetBSD: 
> dev.ath.0.tpscale=1
> dev.ath.0.tpc=1
> dev.ath.0.diversity=0

Odd, that support is in the hal that's in cvs.  I noticed while 
diff'ing that the tpc stuff wasn't in the driver but since that hal 
doesn't do per-packet tpc correctly anyway it shouldn't matter.  Turning 
off diversity should reduce your performance; not improve it.  I'd be 
curious to see numbers for each change separately.  It may be that the 
antenna flipping logic is hurting you (it needs some fixups).

> Another one 
> ifconfig ath0 protmod off
> ifconfig ath0 mtu 2290

I would be curious to see if you're actually getting frames close to the 
mtu.  I just realized recently that I allowed the larger mtu but I think 
the driver tosses frames larger than a cluster because the logic isn't 
in the rx intr handler to chain rx descriptors (it is in another version 
of code where I need it for Atheros' fast frames).

Turning off protection presumably means you're operating in 11g and only 
supporting 11g stations?  If so there's a better way to do this; just 
set the pureg config so 11b stations are rejected by the ap.