2011-01-22 21:01:52

by Richard Schütz

[permalink] [raw]
Subject: ath9k prevents CPU from entering C4 state (kernel 2.6.37 regression)

I noticed that from kernel 2.6.37 on (i've tested 2.6.38-rc2, too) ath9k
prevents my CPU (Intel Atom N450) from entering the C4 power state. It
only goes down to C2 (there's no C3) instead. In 2.6.36(.3) everything
is still fine. The CPU will only go down to C4 with 2.6.37 when the
wireless interface is brought down.

used wireless device:
05:00.0 Network controller: Atheros Communications Inc. AR9285 Wireless
Network Adapter (PCI-Express) (rev 01)

--
Regards,
Richard Sch?tz


2011-01-22 23:57:41

by Thomas Bächler

[permalink] [raw]
Subject: Re: ath9k prevents CPU from entering C4 state (kernel 2.6.37 regression)

Am 22.01.2011 23:52, schrieb Christian Lamparter:
> On Saturday 22 January 2011 21:38:47 Richard Sch?tz wrote:
>> I noticed that from kernel 2.6.37 on (i've tested 2.6.38-rc2, too) ath9k
>> prevents my CPU (Intel Atom N450) from entering the C4 power state. It
>> only goes down to C2 (there's no C3) instead. In 2.6.36(.3) everything
>> is still fine. The CPU will only go down to C4 with 2.6.37 when the
>> wireless interface is brought down.
>
> "ath9k: Fix a DMA latency issue for Intel Pinetrail platforms."
> http://www.spinics.net/lists/linux-wireless/msg58264.html

I experience the same problem, and I do not have a Pinetrail. I run
Intel Core i5-520M (Arrandale) with an AR9280 Rev 2.


Attachments:
signature.asc (262.00 B)
OpenPGP digital signature

2011-01-22 22:52:52

by Christian Lamparter

[permalink] [raw]
Subject: Re: ath9k prevents CPU from entering C4 state (kernel 2.6.37 regression)

On Saturday 22 January 2011 21:38:47 Richard Sch?tz wrote:
> I noticed that from kernel 2.6.37 on (i've tested 2.6.38-rc2, too) ath9k
> prevents my CPU (Intel Atom N450) from entering the C4 power state. It
> only goes down to C2 (there's no C3) instead. In 2.6.36(.3) everything
> is still fine. The CPU will only go down to C4 with 2.6.37 when the
> wireless interface is brought down.

"ath9k: Fix a DMA latency issue for Intel Pinetrail platforms."
http://www.spinics.net/lists/linux-wireless/msg58264.html

the pm_qos value will be configurable with 2.6.38:

"ath9k: Make PM-QOS value as user configurable"
http://www.spinics.net/lists/linux-wireless/msg61259.html

"This will help our customers to configure the pm-qos value according
to the effect in throughput due to the DMA latency problem which was
observed in Intel Pinetrail platforms."