Re: [PATCH 1/3] ath9k: prevent calibration during off-channel activity

["Luis R. Rodriguez" <lrodriguez@atheros.com>]

On Wed, Jul 28, 2010 at 3:28 PM, Felix Fietkau <nbd@openwrt.org> wrote:
> On 2010-07-29 12:12 AM, Luis R. Rodriguez wrote:
>> On Wed, Jul 28, 2010 at 2:08 PM, Felix Fietkau <nbd@openwrt.org> wrote:
>>> On 2010-07-28 10:43 PM, Luis R. Rodriguez wrote:
>>>> On Wed, Jul 28, 2010 at 10:45 AM, Felix Fietkau <nbd@openwrt.org> wrote:
>>>>> Previously the software scan callback was used to indicate to the hardware,
>>>>> when it was safe to calibrate. This didn't really work properly, because it
>>>>> depends on a specific order of software scan callbacks vs. channel changes.
>>>>> Also, software scans are not the only thing that triggers off-channel
>>>>> activity, so it's better to use the newly added indication from mac80211 for
>>>>> this and not use the software scan callback for anything calibration related.
>>>>>
>>>>> This fixes at least some of the invalid noise floor readings that I've seen
>>>>> in AP mode on AR9160
>>>>
>>>> Neat!
>>>>
>>>>> --- a/drivers/net/wireless/ath/ath9k/main.c
>>>>> +++ b/drivers/net/wireless/ath/ath9k/main.c
>>>>> @@ -154,6 +154,27 @@ void ath9k_ps_restore(struct ath_softc *sc)
>>>>>        spin_unlock_irqrestore(&sc->sc_pm_lock, flags);
>>>>>  }
>>>>>        spin_unlock_irqrestore(&sc->sc_pm_lock, flags);
>>>>>  }
>>>>>
>>>>> +static void ath_start_ani(struct ath_common *common)
>>>>> +{
>>>>> +       struct ath_hw *ah = common->ah;
>>>>> +       unsigned long timestamp = jiffies_to_msecs(jiffies);
>>>>> +       struct ath_softc *sc = (struct ath_softc *) common->priv;
>>>>> +
>>>>> +       if (!(sc->sc_flags & SC_OP_ANI_RUN))
>>>>> +               return;
>>>>> +
>>>>> +       if (sc->sc_flags & SC_OP_OFFCHANNEL)
>>>>> +               return;
>>>>> +
>>>>> +       common->ani.longcal_timer = timestamp;
>>>>> +       common->ani.shortcal_timer = timestamp;
>>>>> +       common->ani.checkani_timer = timestamp;
>>>>> +
>>>>> +       mod_timer(&common->ani.timer,
>>>>> +                 jiffies +
>>>>> +                       msecs_to_jiffies((u32)ah->config.ani_poll_interval));
>>>>> +}
>>>>
>>>> I would prefer if you do this sort of code shift in a separate patch.
>>>> In this case its pretty easy to see the code is the same so I think
>>>> its fine the way it is now but for next time please. Otherwise looks
>>>> good, thanks!!
>>> If it had been a bigger function, I'd have made a separate patch, but I
>>> figured for something as trivial as this it wouldn't matter.
>>
>> iw event -t while  pinging and then issuing a scan:
>>
>> 1280354915.820607: wlan32 (phy #0): scan started
>> 1280354920.390438: wlan32 (phy #0): scan finished: 2412 2417 2422 2427
>> 2432 2437 2442 2447 2452 2457 2462 5180 5200 5220 5240 5260 5280 5300
>> 5320 5500 5520 5540 5560 5580 5660 5680 5700 5745 5765 5785 5805 5825,
>> ""
>> 1280354923.103628: wlan32 (phy #0): deauth FOO -> BAR reason 4:
>> Disassociated due to inactivity
>> 1280354923.103736: wlan32 (phy #0): disconnected (local request)
>> 1280354923.111251: phy #0: regulatory domain change: set to world
>> roaming by the wireless core upon initialization request
>>
>> So this seems to re-introduce the same issue I was seeing before.
>
> Are you sure it's this change?

Just reverted the changes and it works without these issues.

> Can you make a log with debug=0x8?

Sure, one sec let me open that branch up again and recompile again.

> What's the specific symptom here? Does the rx path turn deaf?

Not sure exactly, but what I immediately see is disconnects right after scans.

  Luis