system stays on full CPU clock

[Kai F. Lahmann]

Binary package hint: linux-image-2.6.28-9-generic

(critical regression from today's update)

the system now starts with the CPU governor set to "performance" each time, so we waste electrical power like hell and after 5 minutes my ears get killed by the CPU fan...

[Kai F. Lahmann (kfl)]

as nobody reacts, some more details: Asus M2N (intel Pentium M, 855 Chipset); no known ACPI issues since ages. Nothing (except that nforce2-message) from cpufreq in any of the logs...

[Scott James Remnant (Canonical) (canonical-scott)]

We are missing the piece that sets the driver back to "ondemand" after boot, that's in my court

[Scott James Remnant (Canonical) (canonical-scott)]

Will be fixed with an init script

[Launchpad Janitor (janitor)]

This bug was fixed in the package sysvinit - 2.86.ds1-61ubuntu9

---------------
sysvinit (2.86.ds1-61ubuntu9) jaunty; urgency=low

  * debian/initscripts/etc/init.d/ondemand: Sleep for 60 seconds, then
    set CPU Frequency Scaling governor to "ondemand". LP: #341573.

 -- Scott James Remnant <email address hidden> Wed, 18 Mar 2009 11:27:23 +0000

[Roger (r-wiberg)]

Seems not to be working here for some reason. I've got the latest sysvinit and kernel 2.6.28-11-generic installed now, and I still have to swith to Ondemand manually.

[sasha1024 (sasha1024)]

To Roger: Have you updated the "initscripts" package?

[Roger (r-wiberg)]

Yes, I have. The only thing I can think of that's a little bit out of the ordinary on this machine is that I've been experimenting with VDPAU, so I have NVIDIA 185.13 installed. The machine is a Compaq 8710p laptop.

[Scott James Remnant (Canonical) (canonical-scott)]

Roger, please supply the output of:

  cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_{driver,governor}

[Roger (r-wiberg)]

Here it is:

~$ cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_{driver,governor}
acpi-cpufreq
acpi-cpufreq
performance
performance

Come to think of it, I also have ext4 and an encrypted home directory, though I can't how that would interfere.

Roger

[Scott James Remnant (Canonical) (canonical-scott)]

And ls /etc/rc2.d

[Roger (r-wiberg)]

~$ ls /etc/rc2.d
K08vmware S20winbind S89anacron
README S24hal S89atd
S01policykit S25bluetooth S89cron
S10acpid S30gdm S90binfmt-support
S10sysklogd S50avahi-daemon S98usplash
S11klogd S50cups S99acpi-support
S12dbus S50NetworkManager S99laptop-mode
S16ssh S50pulseaudio S99ondemand
S19vmware S50rsync S99rc.local
S20apport S50system-tools-backends S99rmnologin
S20dkms_autoinstaller S70bootlogs.sh S99stop-readahead
S20hotkey-setup S70dns-clean S99timidity
S20nvidia-kernel S70pppd-dns

[Roger (r-wiberg)]

Hey, it must have been at least almost 25 years since I did any programming at all, but I just had a look at the 599ondemand file. It seems it doesn't end with exit 0 after the case-esac clause, at least not on my computer. Could that be it? Or am I grasping at straws?

[Kai F. Lahmann (kfl)]

Also doesn#t work here. Manually running "/etc/init.d/ondemand start" also doesn't work, but fails silently. Putting the echo-Command manually works. I'm no bash guru, so no idea, where the problem sits.

[ChrisS (yseeley)]

As I said in the duplicate bug #344875, it's the "sleep 60" in /etc/init.d/ondemand that seems to be causing the problem.

[Scott James Remnant (Canonical) (canonical-scott)]

I suspect this is the same bug we once saw with upstart-logd (and indeed, why we don't use it anymore), that the output terminal going away results in the process dying

[Scott James Remnant (Canonical) (canonical-scott)]

Just done an upload that should fix it

[Roger (r-wiberg)]

Excellent! Problem solved here. Thanks.

[Denis Bezrodnykh (bee)]

Why this "bug" was fixed so strange hard way? We always can make planet green using /etc/sysfs.conf or something like. I have spent some time to find init.d/ondemand that resets my governor

[karlrt (karlrt)]

the problem with the performance governor staying after boot occurs to me too.

Running /etc/init.d/ondemand start manually solves the problem, but during start, the script is somehow not started!?

[Julien Olivier (julo)]

Here's what I get on a freshly installed Lucid:

julien@julien:~$ ls /etc/rc2.d/
README S20winbind S50rsync S91apache2
S20fancontrol S25bluetooth S50saned S99acpi-support
S20kerneloops S31atieventsd S70dns-clean S99grub-common
S20speech-dispatcher S50cups S70pppd-dns S99rc.local
S20virtualbox-ose S50pulseaudio S90binfmt-support

Why is there no S99ondemand?
Running "S99ondemand" fixes it, but it is not launched on boot.