While checking powertop in my Dell Inspiron 640m, an 82801GBM (ICH7-M) machine running linux-x86_64 with a 2.6.22-rc4-hrt6 kernel, I noticed this strange output:
Idle machine, working ipw3945 network: Wakeups-from-idle per second : 35.3 Top causes for wakeups: 34.7% ( 24.4) <interrupt> : ipw3945 24.5% ( 17.2) <interrupt> : [EMAIL PROTECTED]:0000:00:02.0 13.2% ( 9.3) firefox-bin : futex_wait (hrtimer_wakeup) 5.1% ( 3.6) hald-addon-cpuf : queue_delayed_work_on (delayed_work_timer_fn 3.1% ( 2.2) X : do_setitimer (it_real_fn) 3.0% ( 2.1) soffice.bin : schedule_timeout (process_timeout) CTRL-PgDown in the gnome-terminal app, sudo modprobe -r ipw3945, Ctrl-PgUp, wait one minute for the machine to stabilize again: Wakeups-from-idle per second : 29.1 Top causes for wakeups: 25.8% ( 11.3) <interrupt> : [EMAIL PROTECTED]:0000:00:02.0 21.1% ( 9.2) firefox-bin : futex_wait (hrtimer_wakeup) 9.2% ( 4.0) hald-addon-cpuf : queue_delayed_work_on (delayed_work_timer_fn 6.5% ( 2.9) <interrupt> : extra timer interrupt As you can see, the number of i915 interrupts (irq16) while idle is 11.3 when ipw3945 is *not* running, and jumps up to 17.2 when ipw3945 (irq17) is running (idling actually). In less controlled conditions, with active ipw3945 moving bytes and interrupting way more, i915 goes up to ~40 interrupts, spoiling a lot of the potential power savings. I can reproduce those results consistently, in fact this is one of the times I've seen smaller difference when modprobing ipw3945, maybe because the network was almost completely inactive. I wonder what can cause this raise in interrupts. I am no expert, but it looks to me this is hiding a bug somewhere, be it the chipset, the BIOS or the kernel drm code. Any clue? Regards Santiago _______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
