Re: [PATCH 1/2] iwlwifi: fix set_tx_power vs scan

[Stanislaw Gruszka <sgruszka@redhat.com>]

On Thu, Oct 21, 2010 at 03:13:48PM +0200, Stanislaw Gruszka wrote:
> On Thu, Oct 14, 2010 at 10:32:05AM +0200, Stanislaw Gruszka wrote:
> > > Looks good, the only thing is if priv->tx_power_user_lmt ==
> > > priv->tx_power_next, we don't even have to call set_tx_power, but I
> > > guess calling it won't hurt, so its your decision check or not.
> > 
> > I'll will call iwl_set_tx_power( ... , false); what seems to be right
> > thing to do.
> 
> Set tx power have to be forces. Without that I get  
> 
> iwlagn 0000:40:00.0: low ack count detected, restart firmware
> iwlagn 0000:40:00.0: On demand firmware reload
> iwlagn 0000:40:00.0: Stopping AGG while state not ON or starting
> iwlagn 0000:40:00.0: queue number out of range: 0, must be 10 to 19
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 0
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 1
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 0
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 0
> iwlagn 0000:40:00.0: iwlagn_tx_agg_start on ra = 00:23:69:35:d1:3f tid = 0
> iwlagn 0000:40:00.0: low ack count detected, restart firmware
> iwlagn 0000:40:00.0: On demand firmware reload
> iwlagn 0000:40:00.0: Stopping AGG while state not ON or starting
> iwlagn 0000:40:00.0: queue number out of range: 0, must be 10 to 19
> iwlagn 0000:40:00.0: iwlagn_tx_agg_start on ra = 00:23:69:35:d1:3f tid = 0
> iwlagn 0000:40:00.0: low ack count detected, restart firmware
> iwlagn 0000:40:00.0: On demand firmware reload
> iwlagn 0000:40:00.0: Stopping AGG while state not ON or starting
> iwlagn 0000:40:00.0: queue number out of range: 0, must be 10 to 19
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 4
> iwlagn 0000:40:00.0: Aggregation not enabled for tid 0 because load = 0
> 
> on 5300 and device was unusable in general. Hence I will only comment
> that forcing send tx power after scan is needed.

Not true. I have the same problem with vanilla 2.6.36 without any of my
patches (further investigation show this is 2.6.34 -> 2.6.35
regression). Problem happens only on 5Ghz, that confused me 
because NetworkManager in some cases connects to 5Ghz, in others to
2.4Ghz on my system.

Since my patches do not screw up (at least not that thing :) I'm going
to repost them, also further bisect this "low ack count detected" issue.

Stanislaw