Component: In which sub-part of the software does it exist? This field is required. Click the word "Component" to see a description of each component. If none seems appropriate, look for a "General" component. OS: Ubuntu 12.04 Summary: My laptop is wasting too much energy reducing from 5 hours battery life to 2,5 in Ubuntu. It's a Lenovo e420s core i5 Sandy Bridge. I have tried all the ACPI options in the documentation: hand no help. Powertop estimates 46.2% (744.3) [Rescheduling interrupts] <kernel IPI> and the file /proc/interrupts indicates that i8042 and Rescheduling interrupts are producing too many wakeups when idle my CPU. I attached the files.
Created attachment 76601 [details] interrupts file
Created attachment 76611 [details] Powertop file
The same problem with Kernel 3.5 Ubuntu 12.10 Alpha 3
12: 668220 0 0 0 IO-APIC-edge i8042 yup, that is a heck of a lot of serial interrupts, at what rate are they arriving? is this device being used? can you disable it in BIOS SETUP? have you always had this problem with all versions of linux, or it is a recent regression?
01.org has a new powertop that you might want to try. It will have better hints & stats, and a more detailed report. Also, please run vmstat 1 for a bit - it will show if you are doing disk IO due to being out of memory... you may find that stopping "tor" will reduce this. are you also actively using samba (nmbd), if not, try testing with it disabled. Also, what happens if you un-plug your ps/2 mouse? it seems to be quite busy on irq12.
(In reply to comment #4) Hi, thank you for your reply. > 12: 668220 0 0 0 IO-APIC-edge i8042 > > yup, that is a heck of a lot of serial interrupts, > at what rate are they arriving? The highest rate I've seen is around that number, so between 660000 and 670000 > > is this device being used? The problem is that I don't know what device is. i8042 seems to be PS2 mouse or keyboard, but I don't have any. Just trackpoint, touchpad and normal keyboard integrated in the lenovo > can you disable it in BIOS SETUP? I can disable with bios the devices but I don't know what is the i8042 device. I tried blacklisting the module i8042 and no luck. > > have you always had this problem with all versions of linux, > or it is a recent regression? I cannot tell you because I just bought last month the laptop and I installed directly the OS.
(In reply to comment #5) > 01.org has a new powertop that you might want to try. > It will have better hints & stats, and a more detailed report. > OK I installed the newest version of powertop 1.97 possible to run for this Ubuntu. I attached screen-shoots of the powertop in idle > Also, please run vmstat 1 for a bit - it will show if > you are doing disk IO due to being out of memory... Doesnt seems to be the problem, I attached a screenshot of it as well. > > you may find that stopping "tor" will reduce this. > are you also actively using samba (nmbd), if not, > try testing with it disabled. I unistalled tor and samba, not a big difference gain maybe 5 min. battery life > > Also, what happens if you un-plug your ps/2 mouse? > it seems to be quite busy on irq12. I have no ps/2 mouse. Only input devices are trackpoint and touchpad.
Created attachment 78391 [details] Powertop 1.97 overview
Created attachment 78401 [details] Powertop 1.97 idle stats
Created attachment 78411 [details] Powertop 1.97 frequency stats
Created attachment 78421 [details] Powertop 1.97 Device stats
Created attachment 78431 [details] Powertop 1.97 Tuneables
Created attachment 78441 [details] vmstat1
I just check again my file /proc/interrupts and the wakeups are reduce by half. On the other hand the battery life still the same. I can't get more than 2,5 hours in comparison to the 5 hours in the OS W7. I attached the new version.
Created attachment 78451 [details] New interrupts
Just as additional info, I disable the Visualization (VT-x) feature for the processor in the BIOS settings and I gained 30 minutes more. So at this point I have 3 hours IDLE battery life.
Carlos, Is this still a problem on latest upstream kernel?
Created attachment 106990 [details] New interrupts Hi, Unfortunately I have not seen any improvement. Still same battery life time and more or less the same number of wakeups. I run the test again today and seems the same. I attached the results to this comment.
Created attachment 106991 [details] New powertop
i8042 may be used to connect your touch panel, you can test that by: 1 make sure you do not touch the panel, see if its interrupt number goes up; 2 then move your finger on the panel and check again, its interrupt number should be higher(maybe a lot higher). If it is indeed for the panel, then the interrupt number is correct. BTW, as Len has said, there is an updated powertop: https://01.org/powertop/downloads https://github.com/fenrus75/powertop.git and it has a pretty switch: --html[=FILENAME] generate an HTML report. If a filename is not specified then the default name "powertop.html" is used. The HTML report can be sent to others to help diagnose power issues.
(In reply to Aaron Lu from comment #20) Hi, > i8042 may be used to connect your touch panel, you can test that by: > 1 make sure you do not touch the panel, see if its interrupt number goes up; > 2 then move your finger on the panel and check again, its interrupt number > should be higher(maybe a lot higher). > If it is indeed for the panel, then the interrupt number is correct. > According to your comments it is correct. The wakeups are static until I touch the touchpad. So I guess it's normal. > BTW, as Len has said, there is an updated powertop: > https://01.org/powertop/downloads > https://github.com/fenrus75/powertop.git > and it has a pretty switch: > --html[=FILENAME] > generate an HTML report. If a filename is not specified then the default > name "powertop.html" is used. The HTML report can be sent to others to help > diagnose power issues. I manage to install the powertop 2.4 I attached the powertop.html file
Created attachment 107010 [details] powertop 2.4 html
I don't see any problem in IDLE state, I can only suggest you try all those tuning options.
Ok, I applied the command suggested by Powertop. It increased 10 minutes the battery life. I guess is not really a bug, just how it handles the power. I may have to live with it. Thanks. I am closing the bug.