[ltp] Re: Interference of wlan usage and hdaps?

Elias Oltmanns linux-thinkpad@linux-thinkpad.org
Tue, 22 Sep 2009 21:55:23 +0200


Henrique de Moraes Holschuh <hmh@hmh.eng.br> wrote:
> On Tue, 22 Sep 2009, Elias Oltmanns wrote:
>> The rather annoying result of it all is that hdapsd frequently parks the
>
>> disk heads, sometimes preventing all I/O for tens of seconds without
>> interruption. Clearly, this strange behaviour only occurs once I have
>> connected to my wlan and goes on persistently afterwards, irrespective
>> of the current state of the wlan interface.
>
> Is it a function of the interrupt handling of the kernel driver (i.e. normal
> behaviour is restored if you rmmod the kernel driver for the wlan card), or
> of the hardware itself?

Well, I have done very little testing on the current machine yet, but
the problem persisted even after unloading the driver last time. Still,
after my experience with the atheros driver, I am inclined to believe
that this is a software issue and probably very similar too. It may, of
course, be hidden in the binary firmware image loaded by the hotplug
agent.

>
>> actual culprit and to point the finger. Unfortunately, the problem
>> turned out to be inherently linked with the design of the atheros driver
>> (or parts of it) and nontrivial to solve. Now, I'm lead to believe that
>
> Can you give me any hints about what the problem was with the Atheros card?

Well, at the time (2.6.27) regular noise floor calibration was performed
from sw interrupt context. The problem was that it takes some time for
the hardware to perform that calibration and the driver keeps busy
waiting until a hw register indicates completion. This often takes much
longer than is any good in sw interrupt context. A very brief glance at
more recent code (2.6.30.7) suggests that the relevant snippets of code
haven't changed much since then, I haven't checked whether they are still
called from sw interrupt context though. Besides, I seem to remember
that there were more examples for that sort of design flaw, but perhaps
not with such grave consequences.

Perhaps the following link provides a reasonable summary of the
investigations a the time:
http://marc.info/?l=linux-wireless&m=122427744200942&w=2
Please note that the timer code has been subject to further changes
since because Thomas' patch lead to other problems. The important bit
remains though, ath5s has to be fixed. And now I rather suspect that the
same applies to ipw2200 or the firmware.

>
>> Apparently, this issue has not come up so far, which rather surprises
>> me. Perhaps there aren't that many hdaps users after all, or they aren't
>> heavy wlan usres. But then, it is perfectly possible that I have missed
>> something as I'm following developments and discussions rather
>> sluggishly these days due to time constraints.
>
> Well, I have never seen anything like it in my T43 with an IPW2950 wlan
> card.

Interesting, perhaps I'm barking up the wrong tree after all.

Regards,

Elias