Home » Mailing list archives » linux-kernel » 2007 » September » 2

Re: Linux, 2.6.23-rc5, powersaving degradation, (time spend in C0 goes up after a while)

Previous thread: [PATCH] use VM_READ/WRITE/EXEC to set vm_page_prot by Coly Li on Sunday, September 2, 2007 - 1:38 pm. (2 messages)

Next thread: [PATCH][Trivial] CIFS: Use ssleep() instead of msleep() when sleeping multiple seconds by Jesper Juhl on Sunday, September 2, 2007 - 1:44 pm. (1 message)
[view original message]
From: Christian Leber <christian@...>
To: <linux-kernel@...>
Subject: Linux, 2.6.23-rc5, powersaving degradation, (time spend in C0 goes up after a while)
Date: Sunday, September 2, 2007 - 1:40 pm

Hello,


I have observed some oddness in the powersaving behavior in rc5 and the

rcs before and i suspect this may be scheduler related.


After booting or resuming from drisk or ram(!) the wakeups are about

45 per second (about 1% C0 (cpu runng)) and after

about half an hour to hours of doing nothing with the laptop besides

having a powertop in konsole the wakeups rise to about 6200 and the cpu

is running 55% of the time.


(with rc3 this was odder, it took three time pretty exactly 43min until

this occured)


I appended the output of powertop and the .config.

But powertop doesn't even show the reason for the wakeups, probably they

don't exist, but the power usage really goes up.


The hardware is btw. Dell latitude d830 (intel T7300, intel wlan and

gfx)


It's not a bad bug, it's just odd and eats about 4W of power.


Christian Leber


--

http://rettetdieti.vde-uni-mannheim.de/


-
[view original message]
From: Arjan van de Ven <arjan@...>
To: Christian Leber <christian@...>
Cc: <linux-kernel@...>
Subject: Re: Linux, 2.6.23-rc5, powersaving degradation, (time spend in C0 goes up after a while)
Date: Sunday, September 2, 2007 - 2:40 pm

On Sun, 2 Sep 2007 19:40:38 +0200


can you give us an lspci -vvxx of the system when it's in this "bad"

state? We have a very vague suspicion on something like this, the lspci

would help us a lot (this is a really rare thing that we can't really

reproduce, but it has been reported a few times before)

-
[view original message]
From: Christian Leber <christian@...>
To: Arjan van de Ven <arjan@...>
Cc: <linux-kernel@...>
Subject: Re: Linux, 2.6.23-rc5, powersaving degradation, (time spend in C0 goes up after a while)
Date: Tuesday, September 4, 2007 - 6:25 am

On Sun, Sep 02, 2007 at 11:40:17AM -0700, Arjan van de Ven wrote:


sure, took some time, I waited yesterday 4h next to the laptop and it

didn't happen ;-)

(i usually run some virtualization and then wakeups are screwed anyway)


This time i also did not forget the powertop output.


The lspcivvxx_good was done with the same linux running, after a resume

from ram, then the wakeup behavior was also ok again.

The difference is:


Secondary status: 66MHz- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- <SERR- <PERR-

that is in the good case, and in the bad it is MAbort+


Christian Leber


--

http://rettetdieti.vde-uni-mannheim.de/
[view original message]
From: Rene Treffer <treffer@...>
To: Linux Kernel Mailing List <linux-kernel@...>
Subject: Re: Linux, 2.6.23-rc5, powersaving degradation, (time spend in C0 goes up after a while)
Date: Sunday, September 2, 2007 - 4:38 pm

Hi,


Same problem here, lspci output attached, I can also pm you a lshw output.
Previous thread: [PATCH] use VM_READ/WRITE/EXEC to set vm_page_prot by Coly Li on Sunday, September 2, 2007 - 1:38 pm. (2 messages)

Next thread: [PATCH][Trivial] CIFS: Use ssleep() instead of msleep() when sleeping multiple seconds by Jesper Juhl on Sunday, September 2, 2007 - 1:44 pm. (1 message)