On Sun, May 17, 2009 at 2:30 AM, Stefan Lesicnik <[email protected]> wrote:
> > > On Sun, May 17, 2009 at 2:08 AM, Evan <[email protected]> wrote: > >> 1. What does Ubuntu use to scale CPU frequencies? Does it use the kernel >> method, as described at [1], or something else? >> 2. Did the method change between Intrepid and Jaunty? >> >> I ask these two questions because in Intrepid it worked fine >> out-of-the-box, and in Jaunty it's been doing some really creepy things. >> >> I originally started looking at this because the ignore_nice_load default >> value (discussed previously) appeared to have changed. I wanted to change it >> back, so as per the instructions at [1] I added the proper lines to my >> /etc/rc.local file. The value was ignored, and I filed a bug regarding it at >> [2]. >> >> I recently decided to change the default state of my cores from ondemand >> to powersave, and since I saw no easier way to do it, I added another few >> lines to rc.local. These also appeared to be ignored. Since I can change the >> governor manually using the same cmd I added to rc.local, and I know >> rc.local is being properly read, I knew something was fishy, so I did a few >> experiments. >> >> This is what I determined: every few minutes, something changes the >> governor on my CPU cores back to ondemand. I have a relatively vanilla >> install of Jaunty 64, and I know I haven't installed anything extra that >> should be touching my cpu governors (feel free to prove me wrong on this). >> Both the gnome monitor applet and /sys/devices have been used to change the >> governor, and both agree that it changes back. The ignore_nice_load value >> doesn't seem to change, but the fact that it's being ignored seems like it's >> probably related. >> >> Is something funny going on, or am I barking up the wrong tree? >> >> Evan >> >> [1] http://www.pantz.org/software/cpufreq/usingcpufreqonlinux.html >> [2] https://bugs.launchpad.net/ubuntu/+source/cpufreqd/+bug/368809 >> >> >> > Hi Evan, > > I am not sure (i looked at it briefly a while back), but I think its using > HAL. Thats why /usr/lib/hal/hald-addon-cpufreq runs as a process. > Therefore the corresponding stuff is probably in > /usr/share/hal/fdi/policy/10osvendor/10-cpufreq.fdi, and you can use the > hal-get-property and hal-set-property commands to query and set. > > If you want to do other stuff, i think killing that hald cpufreq process > should let you then... (although there may be a safer saner way to do this) > > Stefan I don't know if that's right. I've attached my cpufreq.fdi file, and as you can see, it's pretty empty. I've never worked with Hal before so I can't say for sure, but I don't see anything relating to governors or frequencies in there. I looked up Hal on google and found something interesting at [1], but I keep getting DBus access denied errors even when I run dbus-send with sudo. I've never worked with DBus before either, so just to be sure I've copied the whole text of my call and the response below: eapa...@evan-ubuntu:~$ sudo dbus-send --system --dest=org.freedesktop.Hal > --print-reply /org/freedesktop/Hal/Device/CPUFreq > "org.freedesktop.Hal.SetCPUFreqConsiderNice" boolean:false > Error org.freedesktop.DBus.Error.AccessDenied: Rejected send message, 1 > matched rules; type="method_call", sender=":1.83" (uid=0 pid=4979 > comm="dbus-send --system --dest=org.freedesktop.Hal --pr") > interface="org.freedesktop.Hal" member="SetCPUFreqConsiderNice" error > name="(unset)" requested_reply=0 destination="org.freedesktop.Hal" (uid=0 > pid=2935 comm="/usr/sbin/hald ")) > I read somewhere that Karmic will be migrating from Hal to devicekit. Does that mean that this whole thing is going to change again when I upgrade? Evan [1] http://people.freedesktop.org/~david/hal-spec/hal-spec.html#interface-cpufreq<http://people.freedesktop.org/%7Edavid/hal-spec/hal-spec.html#interface-cpufreq>
10-cpufreq.fdi
Description: Binary data
-- Ubuntu-devel-discuss mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
