On Wed, 3 Nov 2010, Thomas Renninger wrote: > On Wednesday 03 November 2010 07:53:08 Vitus Jensen wrote: >> On Tue, 2 Nov 2010, Auke Kok wrote: >>> On 11/02/10 07:40, Vitus Jensen wrote: >>>> >>>> Since some kernel versions I notice very high wakeup rates during idle ... >>>> while the reasons for the wakeup stay at reasonable low numbers. In the >>>> 'powertop -d' output below it's 27463 wakeups/sec but the highest number >>>> for any reason is 48. How do I find out what causes the wakeups? They >>>> are actually there because the machine is using ~2W more than usual. >>>> >>>> My usual kernel is 2.6.27, it's shows ~100 wakeups (X11, Wifi+powersave) >>>> >>>> On this 2.6.36 powertop most of the time reports 1.000-200.000 wakeups >>>> during idle. It may get to 100 (X11, Wifi) or ~50 (no X11, no Wifi), >>>> seldomly. >>> >>> I've seen this on specific pieces of hardware (in my case, I had a >>> Lenovo T43 that went apenuts anyway, and I got rid of it). >>> >>> Is this older hardware? It's not uncommon for those - but shouldn't >>> happen anymore on core2 level hardware and newer. >> >> It's a Thinkpad R51e with Pentium M 1.73 >> (http://www.thinkwiki.org/wiki/Category:R51e). > > I remember the R51e to have ACPI GPE address issues. > You may want to look at: > https://bugzilla.kernel.org/show_bug.cgi?id=8246
Ah, now I remember that bug entry! There occurs indeed the address mismatch warning which is cured by "acpi=rsdt". But for the wakeups this doesn't change anything. As for other ideas from bugzilla: /sys/firmware/acpi/interrupts/* -> counts between 0 and 2 /proc/interrupts -> 30/s on timer, 41/s on wlan It's acpi_idle on this machine, not intel_idle, how do I switch? Next I will try to remove NOHZ as suggested in that other bugzilla entry. Bye, Vitus -- Vitus Jensen, Hannover, Germany, Universe (current) pgp public key available from keyservers _______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
