Linux 3.2 Is Still Looking To Be Power Hungry
Phoronix: Linux 3.2 Is Still Looking To Be Power Hungry
The PCI subsystem pull for the Linux 3.2 kernel was published on Friday evening. If you were hoping it would rework PCI-E ASPM (Active-State Power Management) to be more like the Windows implementation or for more PCI drivers to be setting the bits directly to support it (effectively white-listing drivers/hardware), it didn't happen yet...
how to force it
To set the boot option to save the power....
Open up and edit "/etc/default/grub"
Edit the following line to include "pcie_aspm=force" (stuff in quotes varies)
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash pcie_aspm=force"
Then run "sudo update-grub" from the command line for the changes to take effect.
Isn't this aspm bug the hardware lying to the os? Where the hardware says it can't use aspm, but would work if it was told to use it? I fail to see what else the kernel could do that would result in better behavior other than a list of hardware that says it doesn't workk but does anyways.
Ok. Here is the problem with the current behavior.
1. All the motherboards for new laptops have power saving.
2. Only 1 manufacture sometimes has aspm report correctly.
So with the kernel switching from a default of on - if it isn't being reported - to off means that the power saving stops working for almost all laptops and netbooks in existence.
Whats worse is that it is the major way to save power on a laptop. With it off, I loose half of my battery life (eee pc).
Because of the change of the default behavior, linux is a very negative experience for those who try it out on a laptop. And if the person who is trying it out, doesn't frequent phoronix (not likely), won't know about the command line and config files that he/she has to manually edit to get their battery life back where it should be.
so its not a kernel bug but a distri bug
Why does the Distries nut just set this kernel-flag by default if there are only 5-10 notebooks out there who dont have this feature?
That's a possibility, too. Bottom line is that the community somehow has to find a way to deal with such nasty issues.
Originally Posted by blackiwid
This shows again how little hardware manufacturers care about Linux (hell, even on my modern Thinkpad the BIOS doesn't advertise ASPM properly). And this is important to note in the context of the Secure Boot discussions...
Right but doesn't setting this to "on" when the hardware really doesn't support cause breakage? If it does, you have to remember that the linux kernel supports something like 15 years worth of hardware and the goal is for the hardware to work out of the box. You are barking up the wrong tree. Call your laptop's support line and request an update that correctly reports this.
Originally Posted by ua=42
Also you should take a look back to when ACPI replaced APM there were a lot of boards that needed/need all sorts of workaround because the DSDT tables are bad/incorrect. The vendors didn't seem to care much because their windows drivers knew better. The kernels stance then was "go talk to the vendor to get an update, or see this guide for fixing your dsdt table. We will not be adding workarounds to fix this".
cynyr: The vendors are not fixing it, so what are we suppose to do in the meantime?
This is based on the original assumption of the problem, which turned out to be wrong. The kernel devs are now saying that they were misunderstanding the spec when they decided to turn it off the first time, and that it's actually OK to turn it on.
Originally Posted by cynyr
The problem is ACPI, which is a poorly documented clusterfuck of a standard Intel created, and then they allowed Microsoft to twist the arms of motherboard manufacturers to use it as a tool to make it difficult to support various hardware with an x86 OS. There are famous emails subpoenaed from Bill Gates himself during the Microsoft anti-trust hearings in the 90s, discussing ways to use ACPI to sabotage Linux. It obviously worked, as we have these kind of bugs in the mainline kernel, because motherboard manufacturers won't reveal how their ACPI implementation works for fear of what the Microsoft mafia will do to them. Of course, nobody thinks to blame Microsoft or Intel for the problem, it must always be Linus' fault.