[ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS

[ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS

[Joerg Pommnitz]

Hello all,
I have seen a problem using ath9k in an IBSS network formed from
two nodes: one AR5xxx running Madwifi and one A9xxx using ath9k.

The network was running on channel 14 which Madwifi limits to 802.11b
for regulatory reasons (we have permission to use this channel).

The strange part was ath9k: Using "iw dev wlan0 stations dump" I could see
that the TX rate to the Madwifi node was stuck at 1 Mbit/s. An iperf run 
confirmed the miserable data rate. I could force
the rate with "iwconfig wlan0 rate 11M" to something reasonable (once again
measured with iperf). In the latter case, iw still claimed a TX rate of 1Mbit/s
but iperf did show the expected data rate.

The Madwifi node on the other end was happily selecting 11Mbit/s on its own.

Any ideas?
-- 
Regards 
Joerg 

__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verf?gt ?ber einen herausragenden Schutz gegen Massenmails. 
http://mail.yahoo.com 

[ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS

[Luis R. Rodriguez]

On Tue, Nov 17, 2009 at 12:34 AM, Joerg Pommnitz <pommnitz@yahoo.com> wrote:
> Hello all,
> I have seen a problem using ath9k in an IBSS network formed from
> two nodes: one AR5xxx running Madwifi and one A9xxx using ath9k.
>
> The network was running on channel 14 which Madwifi limits to 802.11b
> for regulatory reasons (we have permission to use this channel).
>
> The strange part was ath9k: Using "iw dev wlan0 stations dump" I could see
> that the TX rate to the Madwifi node was stuck at 1 Mbit/s. An iperf run
> confirmed the miserable data rate. I could force
> the rate with "iwconfig wlan0 rate 11M" to something reasonable (once again
> measured with iperf). In the latter case, iw still claimed a TX rate of 1Mbit/s
> but iperf did show the expected data rate.
>
> The Madwifi node on the other end was happily selecting 11Mbit/s on its own.
>
> Any ideas?

Yeah I'm curious can you try on the ath9k station:

iw list

And then please try monitoring the rcstat file as root:

watch cat /sys/kernel/debug/ath9k/phy0/rcstat

When you run iperf what rates do you see being used? After a while of
running iperf (3 times) with UDP:

iperc -c ip-address-of-server -u -i 1 -b 30M

please paste the output of the rcstat file.

  Luis

[ath9k-devel] TX power control, WAS: Rate Selection problems in mixed 802.11b/802.11g IBSS

[Joerg Pommnitz]

Luis,
I'm retracting my observation. Trying to reproduce it with the
logs you asked for did show that the ath9k node tried to increase
the tx data rate from time to time and that the signal strength
was very low. I don't quite understand this, but now I do not
think that the RC algorithm is to blame.

Another question popped up, however. I'm evaluating ath9k and
Atheros 802.11n cards for a project we are working on. Together
with a co-worker we did some measurements of the tx power. What
we found is, that using iwconfig to change the tx power had no
measurable effect on the tx power on the antenna port. Even if I
configured the tx power down to 1db the measured value remained 
constant at 27mW.
So, tx power control seems not to be implemented in ath9k. Is this
correct and if so, will this change?

 -- 
Regards 
Joerg 




----- Urspr?ngliche Mail ----
> Von: Luis R. Rodriguez <mcgrof@gmail.com>
> An: Joerg Pommnitz <pommnitz@yahoo.com>
> CC: ath9k-devel at lists.ath9k.org
> Gesendet: Dienstag, den 17. November 2009, 18:07:12 Uhr
> Betreff: Re: [ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS
> 
> On Tue, Nov 17, 2009 at 12:34 AM, Joerg Pommnitz wrote:
> > Hello all,
> > I have seen a problem using ath9k in an IBSS network formed from
> > two nodes: one AR5xxx running Madwifi and one A9xxx using ath9k.
> >
> > The network was running on channel 14 which Madwifi limits to 802.11b
> > for regulatory reasons (we have permission to use this channel).
> >
> > The strange part was ath9k: Using "iwdevwlan0 stations dump" I could see
> > that the TX rate to the Madwifi node was stuck at 1 Mbit/s. An iperf run
> > confirmed the miserable data rate. I could force
> > the rate with "iwconfigwlan0 rate 11M" to something reasonable (once again
> > measured with iperf). In the latter case, iw still claimed a TX rate of 
> 1Mbit/s
> > but iperf did show the expected data rate.
> >
> > The Madwifi node on the other end was happily selecting 11Mbit/s on its own.
> >
> > Any ideas?
> 
> Yeah I'm curious can you try on the ath9k station:
> 
> iw list
> 
> And then please try monitoring the rcstat file as root:
> 
> watch cat /sys/kernel/debug/ath9k/phy0/rcstat
> 
> When you run iperf what rates do you see being used? After a while of
> running iperf (3 times) with UDP:
> 
> iperc -c ip-address-of-server -u -i 1 -b 30M
> 
> please paste the output of the rcstat file.
> 
>   Luis


__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verf?gt ?ber einen herausragenden Schutz gegen Massenmails. 
http://mail.yahoo.com 

[ath9k-devel] TX power control, WAS: Rate Selection problems in mixed 802.11b/802.11g IBSS

[Luis R. Rodriguez]

On Fri, Nov 20, 2009 at 5:05 AM, Joerg Pommnitz <pommnitz@yahoo.com> wrote:
> Luis,
> I'm retracting my observation. Trying to reproduce it with the
> logs you asked for did show that the ath9k node tried to increase
> the tx data rate from time to time and that the signal strength
> was very low. I don't quite understand this, but now I do not
> think that the RC algorithm is to blame.

Oh ok, good to hear thanks for the report back.

> Another question popped up, however. I'm evaluating ath9k and
> Atheros 802.11n cards for a project we are working on. Together
> with a co-worker we did some measurements of the tx power. What
> we found is, that using iwconfig to change the tx power had no
> measurable effect on the tx power on the antenna port. Even if I
> configured the tx power down to 1db the measured value remained
> constant at 27mW.
> So, tx power control seems not to be implemented in ath9k. Is this
> correct and if so, will this change?

It is indeed but it was just reported a couple of days ago it only
takes effect after a scan. This is because we cache the tx power
setting and only apply it later. See:

http://wireless.kernel.org/en/users/Drivers/ath9k/bugs

  Luis

[ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS

[Joerg Pommnitz]

Luis,
I'm retracting my retraction :-). And I think I can provide an easy way to reproduce
the stuck rate:
* connect two nodes in an IBSS network in 802.11g mode
* run iperf and watch the data rate on the client (e.g. the sender) with "iw dev wlan0 station dump"
* change the IBSS network to 802.11a

* run iperf and watch the data rate on the client (e.g. the sender) with "iw dev wlan0 station dump"

After the switch to 802.11a the rate seems to be stuck at 1Mbit/s. Watching /sys/kernel/debug/ath9k/phy0/rcstat shows that the RC algorithm sees the statistics, but it never tries to send at a rate 
above 1Mbit/s.

Can you reproduce this? Attached is the script I'm using to create the IBSS network.
  
-- 
Regards 
Joerg 




----- Urspr?ngliche Mail ----
> Von: Luis R. Rodriguez <mcgrof@gmail.com>
> An: Joerg Pommnitz <pommnitz@yahoo.com>
> CC: ath9k-devel at lists.ath9k.org
> Gesendet: Dienstag, den 17. November 2009, 18:07:12 Uhr
> Betreff: Re: [ath9k-devel] Rate Selection problems in mixed 802.11b/802.11g IBSS
> 
> On Tue, Nov 17, 2009 at 12:34 AM, Joerg Pommnitz wrote:
> > Hello all,
> > I have seen a problem using ath9k in an IBSS network formed from
> > two nodes: one AR5xxx running Madwifi and one A9xxx using ath9k.
> >
> > The network was running on channel 14 which Madwifi limits to 802.11b
> > for regulatory reasons (we have permission to use this channel).
> >
> > The strange part was ath9k: Using "iwdevwlan0 stations dump" I could see
> > that the TX rate to the Madwifi node was stuck at 1 Mbit/s. An iperf run
> > confirmed the miserable data rate. I could force
> > the rate with "iwconfigwlan0 rate 11M" to something reasonable (once again
> > measured with iperf). In the latter case, iw still claimed a TX rate of 
> 1Mbit/s
> > but iperf did show the expected data rate.
> >
> > The Madwifi node on the other end was happily selecting 11Mbit/s on its own.
> >
> > Any ideas?
> 
> Yeah I'm curious can you try on the ath9k station:
> 
> iw list
> 
> And then please try monitoring the rcstat file as root:
> 
> watch cat /sys/kernel/debug/ath9k/phy0/rcstat
> 
> When you run iperf what rates do you see being used? After a while of
> running iperf (3 times) with UDP:
> 
> iperc -c ip-address-of-server -u -i 1 -b 30M
> 
> please paste the output of the rcstat file.
> 
>   Luis


__________________________________________________
Do You Yahoo!?
Sie sind Spam leid? Yahoo! Mail verf?gt ?ber einen herausragenden Schutz gegen Massenmails. 
http://mail.yahoo.com 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ibss.sh
Type: application/x-sh
Size: 296 bytes
Desc: not available
Url : http://lists.ath9k.org/pipermail/ath9k-devel/attachments/20091123/13bd915a/attachment.sh