Bug 53051 - Conservative governor get's sometimes stuck at highest frequency - HP Pavilion dv6 CORE i5
Summary: Conservative governor get's sometimes stuck at highest frequency - HP Pavilio...
Status: CLOSED OBSOLETE
Alias: None
Product: Power Management
Classification: Unclassified
Component: cpufreq (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: cpufreq
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-01-25 19:12 UTC by cem.aydin
Modified: 2013-05-12 15:56 UTC (History)
2 users (show)

See Also:
Kernel Version: 3.6.11
Subsystem:
Regression: No
Bisected commit-id:


Attachments
Zipped files as described in description. (3.29 KB, application/zip)
2013-01-25 19:12 UTC, cem.aydin
Details

Description cem.aydin 2013-01-25 19:12:27 UTC
Created attachment 91801 [details]
Zipped files as described in description.

About once a week, when booting daily, the CPU (all cores) get's stuck at highest frequency.

Unfortunately, I can't tell exactly when this appeared but it was about during the time, when I switched to systemd and cpufrequtils got replaced cpupower, about half a year ago.

The kernel modules are loaded properly and the start scripts are executed. I can not find any difference, when this occurs or when it doesn't. Exept that the frequency is stucked at highest level of course. When this occurs I wasn't able to set the frequency manually (by the userspace governor). Reloading the conservative module or reexuting the (systemd) start script did not have any effect. The only thing I can do to resolve the issue is reboot.

I use an up to date Arch Linux with the regular kernel and systemd.

Additional Infos:
proc/cpuinfo reports the wrong frequency (lowest). I think this is mentioned in another bug. However cpupower -c all frequency-info (executed as root) reports the correct highest frequency. I can clearly see that this is the true frequency because the temperatures reported by sensors are almost 10°C higher than under normal circumstances. I don't have any process that could cause the governor to raise the frequencies. CPU usage is below 1% and load is between 0.0 and 0.05.

Hardware:
A HP Pavilion dv6 CORE i5

Attachments:
- cpupower (/usr/lib/systemd/scripts/cpupower, the start up scripts used by systemd to set the frequencies)
- cpupower.conf (/etc/conf.d/cpupower)
- cpupower-service.txt (output from systemctl status cpupower.service)
- cpupower.txt (output from cpupower -c all frequency-info)
- journal.txt (output from journalctl -b | grep cpufreq)
- lsmod.txt (output from lsmod |grep cpufreq)
- proc_cpuinfo.txt (output from cat /proc/cpuinfo (wrong frequencies))

Please let me know if you need additional info.

Regards
Comment 1 Aaron Lu 2013-02-21 07:30:41 UTC
Hello,

I'm not sure if this is possible, since you said this problem happened once a week. But it would be helpful to identify which kernel begins to have this problem, and if the latest mainline kernel fixed it already. Thanks.
Comment 2 Aaron Lu 2013-04-01 05:53:55 UTC
Hi Cem,

Do you still have this problem on latest upstream kernel? Thanks.
Comment 3 cem.aydin 2013-04-02 14:00:44 UTC
Hello

Yes, the problem occured once last week. However it does not occur often. Therefor it's a bit hard to tell.

At the moment I'm on 3.8.4-1-ARCH.

I'll keep having an eye on this and report back if it occurs on 3.8.5

Regards
Comment 4 Aaron Lu 2013-04-02 14:05:31 UTC
Thanks Cem. It would be good to test latest Linus' tree too.
Comment 5 Zhang Rui 2013-04-25 02:17:30 UTC
Cem, is there any update on this?
Comment 6 cem.aydin 2013-05-05 13:48:31 UTC
Hi

Since my last comment (2013-04-02) this didn't occur anymore !

Using 3.8.11-1-ARCH now.

However I'd like to keep this open for one or two more weeks to make sure it doesn't appear anymore.

Regards
Comment 7 cem.aydin 2013-05-12 15:56:02 UTC
Ok

This bug didn't occur anymore.

And by todays update to 3.9.2-1-ARCH acpi-cpufreq got replaced by the new intel-pstate and it's powersave governor.

I'm closing as obsolete.

Thanks and Regards

Note You need to log in before you can comment on or make changes to this bug.