ath5k with two NICs runing IBSS and AP mode simultaneously causes error

ath5k with two NICs runing IBSS and AP mode simultaneously causes error

[Joerg Pommnitz]

Hello all,
I have a configuration with two Atheros NICs (AR5212). To evaluate whether ath5k is now good enough to replace Madwifi I run the current wireless-testing branch and did some tests:
One interface (wlan0) is in IBSS mode joined to another host. I run iperf in both directions simultaneously over night. This worked very well (except for the known IBSS join bug).
Additionally I built hostapd 0.6.9 and run it with a very minmal configuration (interface=wlan1, driver=nl80211, ssid=MyNetwork, hw_mode=g, channel=1) and tried to connect from a WinXP client. I had not yet setup a DHCP server, so this was not actually expected to succeed.
I think immediately after starting hostapd the iperf run stopped and I got the following error message:
ath5k phy0: no further txbuf available, dropping packet.
The timing makes it very likely that the hostapd on wlan1 disturbed the iperf run on wlan0. Is this a known problem? Is there something I can do to help track this down?

 -- 
Regards 
Joerg 


      

Re: ath5k with two NICs runing IBSS and AP mode simultaneously causes error

[Bob Copeland]

On Wed, Oct 21, 2009 at 7:01 AM, Joerg Pommnitz <pommnitz@yahoo.com> wrote:

> ath5k phy0: no further txbuf available, dropping packet.
> The timing makes it very likely that the hostapd on wlan1 disturbed the iperf run on wlan0. Is this a known problem? Is there something I can do to help track this down?

It is a known problem - please do try to track it down.

I think a good place to start is looking at the creation
of tx descriptors, particularly for the beacon and CAB
queues, and make sure they are properly released.  Maybe
printing out a high-water mark of the created vs available
descriptors when transmitting packets would be a useful
data point.

BTW the reason all traffic stops is that we stop the queues
but don't re-enable them in that case -- we should probably
set a timer or something to recover from that situation.

-- 
Bob Copeland %% www.bobcopeland.com

RE: ath5k with two NICs runing IBSS and AP mode simultaneously causes error

[Joerg Pommnitz]

Interesting. I tried to reproduce this problem and couldn't so I assumed that this was just a fluke.

By the way, please have a look at http://marc.info/?l=linux-wireless&m=125621556504279&w=2
I don't have an ath5k trace (yet), but I think this might be a problem shared by ath5k and ath9k.

 -- 
Regards 
Joerg 




----- Ursprüngliche Mail ----
Von: Bob Copeland <me@bobcopeland.com>
An: Joerg Pommnitz <pommnitz@yahoo.com>
CC: linux-wireless@vger.kernel.org; ath5k-devel@lists.ath5k.org
Gesendet: Donnerstag, den 22. Oktober 2009, 15:08:32 Uhr
Betreff: Re: ath5k with two NICs runing IBSS and AP mode simultaneously causes  error

On Wed, Oct 21, 2009 at 7:01 AM, Joerg Pommnitz <pommnitz@yahoo.com> wrote:

> ath5k phy0: no further txbuf available, dropping packet.
> The timing makes it very likely that the hostapd on wlan1 disturbed the iperf run on wlan0. Is this a known problem? Is there something I can do to help track this down?

It is a known problem - please do try to track it down.

I think a good place to start is looking at the creation
of tx descriptors, particularly for the beacon and CAB
queues, and make sure they are properly released.  Maybe
printing out a high-water mark of the created vs available
descriptors when transmitting packets would be a useful
data point.

BTW the reason all traffic stops is that we stop the queues
but don't re-enable them in that case -- we should probably
set a timer or something to recover from that situation.

-- 
Bob Copeland %% www.bobcopeland.com