powertop suggests a patch to save power

Not so sure this will get in. This sort of patch is a case-by-case type, and I'm a little reluctant to include it for fear of regressions.

do you mean performance regression? because of the patch aim or
because the patch is not well done?
And also, why a case-by-case type? in the sense that some users could
avoid using the patch because their BIOS allows that option? It could
be important to know how many laptop users can edit this option!

On Thu, 2007-05-31 at 16:45 +0000, Nicolò Chieffo wrote:
> do you mean performance regression? because of the patch aim or
> because the patch is not well done?
> And also, why a case-by-case type? in the sense that some users could
> avoid using the patch because their BIOS allows that option? It could
> be important to know how many laptop users can edit this option!

When I mention regressions and case-by-base, I mean the same thing.
Basically that force enabling HPET on some systems will cause bad things
to happen. So we can't just include this by default.

ok, understood.
I hope bad things are not bad crashes!

Have you got any ideas why intel proposed this patch without telling
anything about instability?
Is it possible to know which systems may be bad effected?
Do you know when this patch can be enabled safely?

thank you for your explanations

On Thu, 2007-05-31 at 17:41 +0000, Nicolò Chieffo wrote:
> ok, understood.
> I hope bad things are not bad crashes!
>
> Have you got any ideas why intel proposed this patch without telling
> anything about instability?
> Is it possible to know which systems may be bad effected?
> Do you know when this patch can be enabled safely?

The patch is fine, it's just not well tested. When it gets suitable
testing, it'll probably go to upstream kernel and makes its way down to
our kernel.

I've got an ASUS A6Km Notebook and I've got the same problems as Nicolò Chieffo. Maybe there would be a way to include this patch and activate it only by a special boot option? Because the High Precision Event Timer seems to show a noticeable effect on some systems if it's aktivated.

The "force enable HPET" patch-set is mentioned in Outstanding Issues of https://wiki.ubuntu.com/power-management-in-Ubuntu

The patchs in PowerTOP's site aren't really updated.
Better is to grab them where they are maintained, at http://www.tglx.de/projects/hrtimers/2.6.22/broken-out/ (plus there's more patchs related to HPET improvments there).

ubuntu_demon : it may be more in point to attach such bugs to the blueprint (since the Launchpad interface allows that, and because there's quite a lot of bugs related to power efficiency, it's more efficiently handled linked / tracked down from blueprint than from the wiki) : https://blueprints.launchpad.net/ubuntu/+spec/power-management-in-ubuntu
I've added it there, already.

to Benjamin Pineau :

Thanks I didn't know you could assign bugs to blueprints. (I already added the bug to the wiki page)

Quick update: those patches made their way in Linus' git repository, and are all merged in 2.6.24-rc1 and up.

So HPET is now (as of 2.6.24-rc1) force enabled by default for all chipsets known to support such timers, even when this functionality is hidden by the BIOS. But only if the HPET timer is officially documented in the chipset's specifications.

Also merged, the patches for the remaining chipsets (nVidia nforce4, Via VT8235 and VT8237, Intel ICH4), where an HPET timer has been found, is known to work, as been tested by the powertop and hrtimers userbase, but is not officially documented. But in this later case, the HPET timer isn't activated by default, in the stock kernel : you have to pass the "hpet=force" kernel boot option to enable it. See "hpet=" in Documentation/kernel-parameters.txt (for a kernel > 2.6.24-rc1).

So this bug should be closed as soon as Ubuntu rebase on 2.6.24. Or even right now : it wouldn't make much sense to backport it in 2.6.22 or .23, we'd better wait a bit. Admittedly, a remaining question would be: should we force enable undocumented hpet timers when they are supported (but not enabled by default) in the vanilla kernel ?

That sounds great. A hpet force grub option would be the best since it should be regression free..

2.6.24 is now available in Ubuntu 8.04 development version, so HPET should be there. The question about whether to force-enable it for undocumented but supporting platforms is on the power management wiki page.

Per a decision made by the Ubuntu Kernel Team, bugs will longer be assigned to the ubuntu-kernel-team in Launchpad as part of the bug triage process. The ubuntu-kernel-team is being unassigned from this bug report. Refer to https://wiki.ubuntu.com/KernelTeamBugPolicies for more information. Thanks.