Most recent kernel where this bug did not occur: 2.6.15-mm1 and 2.6.15 Distribution: Debian testing Hardware Environment: IBM Thinkpad T43 I reported on January 7th that acpi-cpufreq did not work anymore on my IBM Thinkpad T43 (fails to load with ENODEV). The problem appeared first in 2.6.15-mm2 and has been reported in the following thread on LKML: http://lkml.org/lkml/2006/1/7/181 The issue has been discussed a little bit with Len Brown and Venkatesh Pallipadi on linux-acpi, for instance in http://marc.theaimsgroup.com/?l=linux-acpi&m=113770495819445&w=2 and in private mails. Andrew Morton tracked the ENODEV down in this post http://marc.theaimsgroup.com/?l=linux-kernel&m=113672322608497&w=2 The last news I heard was that I should use speedstep-centrino since it is not broken and is actually faster. I now use speedstep-centrino and it works fine. But Andrew Morton does not like this solution. The acpi-cpufreq driver should either fixed or removed. That's why I am opening a bug here.
Andrew is correct, and there is a fix for this bug in the works.
I have also been expierencing the same issues. I cannot load the acpi-cpufreq module anymore. I have tested this with the newest development kernel (2.6.17_rc4) I can provide dmesg output and config if requested. Last known kernel to have acpi work was 2.6.15-gentoo-r1 for me. Thanks.
Gentoo have several reports of this regression. If there is anything we can do to help (more info, acpi dumps, testing patches), ask away.
Created attachment 8257 [details] acpi-cpufreq, speedstep-centrino merged driver All ACPI related things have been removed from the speedstep-centrino driver and added to the acpi-cpiufreq one. acpi-cpufreq now combines MSR and SystemIO capabilities.
Should this bug be closed? I don't see the patch in the cpufreq git tree or any branch of the acpi git tree.
Denis, which kernel does the patch apply to? Most hunks fail on 2.6.16 and 2.6.17 for me.
Can someone test a recent -git snapshot to confirm this bug still exists?
Since 2.6.18-rc1, acpi-cpufreq is able to load again. But it does not work. It seems to acquire the device but no cpufreq capability seems to be exported. speedstep-centrino still works fine.
Brice: I suggest you open a new bug about that.
As of 2.6.20, speedstep-centrino is depricated and acpi-cpufreq is preferred. If either of them don't work in 2.6.20, please re-open. Going forward, the ACPI part of speedstep-centrino will be removed and it will rot on the vine as a legacy-only driver with hard-coded tables for non-ACPI mode.