Re: [ath9k-devel] ath9k: massive unexplained latency in 2.6.27 (rc5, rc6, probably others)

[Steven Noonan]

2008/9/23 Steven Noonan <steven@xxxxxxxxxxxxxx>:
> On Mon, Sep 22, 2008 at 9:01 AM, Luis R. Rodriguez
> <lrodriguez@xxxxxxxxxxx> wrote:
>> On Mon, Sep 22, 2008 at 8:11 AM, Steven Noonan <steven@xxxxxxxxxxxxxx> wrote:
>>> On Mon, Sep 22, 2008 at 1:06 AM, Luis R. Rodriguez
>>> <lrodriguez@xxxxxxxxxxx> wrote:
>>>> Please test the following patch.
>>>>
>>>> diff --git a/drivers/net/wireless/ath9k/core.c b/drivers/net/wireless/ath9k/core.c
>>>> index c262ef2..9a51739 100644
>>>> --- a/drivers/net/wireless/ath9k/core.c
>>>> +++ b/drivers/net/wireless/ath9k/core.c
>>>> @@ -681,10 +681,12 @@ int ath_open(struct ath_softc *sc, struct ath9k_channel *initial_chan)
>>>>         * Enable MIB interrupts when there are hardware phy counters.
>>>>         * Note we only do this (at the moment) for station mode.
>>>>         */
>>>> +#if 0
>>>>        if (ath9k_hw_phycounters(ah) &&
>>>>            ((sc->sc_ah->ah_opmode == ATH9K_M_STA) ||
>>>>             (sc->sc_ah->ah_opmode == ATH9K_M_IBSS)))
>>>>                sc->sc_imask |= ATH9K_INT_MIB;
>>>> +#endif
>>>>        /*
>>>>         * Some hardware processes the TIM IE and fires an
>>>>         * interrupt when the TIM bit is set.  For hardware
>>>>
>>>
>>> It didn't apply to -rc7, but I managed to apply it manually
>>> (apparently you made this for wireless-testing?) and added the #if 0.
>>> After 7 hours uptime with the driver, no apparent interrupt storm.
>>>
>>> I'll do more thorough testing later, but I have a 3 hour drive today,
>>> and some packing to do, so it'll need to wait a bit.
>>
>> Yeah I did it for wireless-testing. Good to hear this so far has
>> solved the issue. I'll port it to 27 and post it once I get your
>> blessings that this fixed it by a Tested-by.
>>
>
> Okay. I'm in Seattle now.
>
> Yes, it solved the issue, but surely an #if 0 isn't a proper solution.
> What's the actual bug here? I'm afraid the meaning of what's going on
> here isn't exactly intuitive.
>
> Also, I'm doing an 8 hour run with this patch tonight. Maybe more,
> depends on when I wake up. ;)
>
> I think if this resolves the issue on two separate ≈8 hour runs, it
> should be considered the solution to the issue, unless it resurfaces.
>

I am only 25 minutes into the run, but I feel this is worth reporting
before I forget. No IRQ storm thus far, but the connection keeps
dropping. Or rather, it doesn't disassociate, but while I was
transferring a large file, it had points of time where it suddenly
lost and reacquired the connection (signal strength fluctuates wildly
between 70-90% and then 0% for a second, and then back to
fluctuating). Perhaps this is related to the lack of aggregation?

- Steven
Х╨{.nг+┴╥÷╝┴╜├+%┼кlzwm┘Иb²К╖╡ФЛr╦⌡zX╖╩╝w╔┼{ay╨й┤з≥К,j╜╒fё╒╥h ▀Юz╧╝w╔╒╦╒╥╕j:+v┴╗┼wХjьm╤÷Ъ╬╚▒ЙГzZ+┐Ы ▌┼щ╒j"²З!╤iO∙Ф╛z╥ vь^╤m╖ЪПцnф┼ЮЧY&≈