FEATURE REQUEST: include linux-phc patch in kernel

Feature request from dapper.

I'll consider this patch, but I can't guarantee it to get in. The reason being that it modifies stock code, and we are trying to avoid that where possible.

The main question I have is if this patch is as useful as it claims (and it seems that it is), why isn't it merged with upstream kernel? As much as it is a pain for users to have to get this themselves, it is also a pain for distro maintainers to support external patches like this just to have happy users.

I strongly suggest talking to the linux-phc maintainers and try to get them to push this stuff upstream so more people can benefit. It helps only a handful of people to have this patch external to the kernel. It would help everyone to have it in mainline.

i talked to the maintainers and they told me that to include it in the mainline kernel is nearly impossible. Look at http://lkml.org/lkml/2006/7/4/246 .

if the patch can't get in ubuntu, maybe a seperate package with the patched speedstep-centrino module will help the users to use this feature...

i'm working on a test package.

Maybe this is just a serious problem for me ... but under 2.6.17 my computer functions and under the latest it does not. Without scaling it overheats and is unusable for any computationally intensive task (my only recourse is to use a conservative governor which is now impossible).

I know this is a bug report but could someone please explain to me or point me to a reference as to why everything worked fine in 2.6.17 but is now broken in 2.6.19?

after you've installed the package you should
- set your voltage settings in /etc/linux-phc/undervolt
- reboot

a package with the module for linux-image-2.6.19-7-generic .

Thanks a lot for these patches. I've had problems with 0.27 starting properly at boot (and I think I've read every thread related to the subject), so what about an 0.28 patch for the stock Edgy kernel (2.6.17-10)? If possible. Thanks again.

Is there any chance of someone releasing a Debian package for the new 2.6.20 kernel that in the current ubuntu feisty? I've tried all the guides and just can't get the kernel to compile after patching with the PHC 0.29 from svn.

I'd be very much interested in a phc package for the feisty kernel as well. Anyone?

Please note I only included the built-in tables portion of the phc patch. I do not want to support the sysfs twiddling. Based on the information I've seen, it can prove dangerous, and I'm not open to allowing our users to accidentally break their hardware. The cpufreq tables should suffice for what most users want.

Is it only a problem when users overvolt their hardware or also when they undervolt the hardware (beside freezes of course!) ?

If only overclocking is a real problem I think that it shouldn't be that hard to implement a limit based on the cpu default tables...

For users with bad acpi tables the build in frequency tables are a solution, but my intention was to have the undervolt functionality in Ubuntu to have much longer battery runtime...

Beside this all: Why should Ubuntu protect it's users more than Microsoft..?

Some of us have laptops with lousy heat control, and undervolting is a great help (not to mention, as Phillip said, that it helps prolong battery life). The only problem I've ever experienced from undervolting too much, by one voltage increment, is a system lockup resolved by rebooting and adjusting the parameters. Can anyone point to solid evidence that someone's fried their system by undervolting? Also, caveats are laid out in any explanation of undervolting, including with the tools built for Windows. Any fool can figure out a way to destroy their computer if they want. If you just read instructions, I think you're totally safe.

If you are not willing to implement the sysfs interace, then there is little point in including this patch. The default tables are *extremely* conservative. There is virtually no gain as far as extended battery life or lower operating temperature when using those tables.

First, thanks to Ben for getting this in.

Second, Pausanias, I have to disagree. I at least have a badly broken BIOS and have been forced to patch the kernel myself to put in exactly the conservative tables you speak of in order to run my 1.6GHz chip at any speed higher than 600MHz. There is, at least for some of us, quite a point in this patch.

I didn't realize that this patch was also used to fix broken BIOSes. Therefore, I take back what I said about there being little point in including this patch.

However I still maintain that *if you have a correctly working BIOS*, you will not get much of a performance boost from the default tables---the performance boost is dramatically greater if you are allowed to lower those voltages another 30% or so, where allowed. Especially at your highest GHz.

I agree with Dreimann and Schneider that the risk of someone destroying their computer with the sysfs interface is nil... But I could see someone handing an op_points table to a friend and then the friend's computer crashing intermittently and it being blamed on Ubuntu. So I get yer point...

Now off I go to compile my kernel yet again.

I don't see that a fix is commited for all the points we talked about.

Ben, why don't you discuss the open points with us here?

Yes, actually I'd like to see any evidence that undervolting can damage a CPU or other hardware. I haven't seen that anywhere. Thanks.

>Yes, actually I'd like to see any evidence that undervolting can damage a CPU or other hardware. I haven't seen that anywhere. >Thanks.

I've been seriously trying to damage some hardware myself with this patch in the past. The only thing that happens is that everything freezes and a cold shutdown is required.

How about re-opening the bug and continue the discussion? (Perhaps I'm mistaken, but as far as I understand the possibility to undervolt the CPU is not enabled in the latest kernel, yet that is precisely what the feature request is about).

its confirmed by ben but should IMHO not be treated as already fixed!

Against linux-source-2.6.20 this bug does not meet the criteria for a stable release update and is being marked as Won't Fix. You can learn more about the stable release update process at https://wiki.ubuntu.com/StableReleaseUpdates .

If this is still a issue for the upcoming Hardy kernel, please provide a link or attachment to the updated patches in question and I'll retarget accordingly. However, note that upsteam mainline submission/acceptance will also be a major factor. Thanks.