Force HPET patch for 2.6.22 kernel?

[Kyle Cunningham]

I was able to successfully apply the hrt6 patch to a 2.6.22 kernel.
Unfortunately it doesn't patch cleanly against 2.6.22.6.  Nonetheless,
I must say that the difference is amazing.  My current powertop numbers 
are below.
Even though there are an annoying number of libata interrupts on a 
normal workload
(web browsing and such) the processor has had an avg residency in C4 > 
50%.  I also
have attached my original numbers (at bottom) and data after disabling 
DRI. 
Thanks again everyone for the help!

Cheers,
Kyle

Cn                Avg residency       P-states (frequencies)
C0 (cpu running)        ( 8.8%)         1.60 Ghz     0.0%
C1                0.0ms ( 0.0%)         1333 Mhz     0.0%
C2                0.1ms ( 0.1%)         1067 Mhz     0.0%
C3                0.6ms ( 0.9%)          800 Mhz   100.0%
C4                1.8ms (90.2%)

Wakeups-from-idle per second : 516.0    interval: 3.0s
Power usage (ACPI estimate): 14.8W (1.3 hours)

Top causes for wakeups:
  81.9% (420.7)       <interrupt> : PS/2 keyboard/mouse/touchpad
   3.4% ( 17.7)       <interrupt> : libata
   3.4% ( 17.3)              Xorg : do_setitimer (it_real_fn)
   2.5% ( 12.7)              Xorg : schedule_timeout (process_timeout)
   2.2% ( 11.3)       <interrupt> : ipw2200, Intel ICH6
   1.9% ( 10.0)   hald-addon-cpuf : cpufreq_governor_dbs 
(delayed_work_timer_fn)
   1.4% (  7.3)            kicker : schedule_timeout (process_timeout)
   0.3% (  1.7)              kwin : schedule_timeout (process_timeout)
   0.3% (  1.7)    wpa_supplicant : schedule_timeout (process_timeout)
   0.3% (  1.3)     <kernel core> : neigh_table_init_no_netlink 
(neigh_periodic_timer)
   0.2% (  1.0)          ifconfig : tg3_open (tg3_timer)
   0.2% (  1.0)     <kernel core> : queue_delayed_work_on 
(delayed_work_timer_fn)
   0.2% (  1.0)          kwrapper : do_nanosleep (hrtimer_wakeup)
   0.2% (  1.0)            dhcdbd : schedule_timeout (process_timeout)
   0.2% (  1.0)         dhclient3 : ipw_net_hard_start_xmit 
(delayed_work_timer_fn)
   0.2% (  1.0)         powernowd : do_nanosleep (hrtimer_wakeup)
   0.2% (  1.0)          kdesktop : schedule_timeout (process_timeout)
   0.2% (  1.0)   guidance-power- : schedule_timeout (process_timeout)
   0.1% (  0.7)           konsole : schedule_timeout (process_timeout)
   0.1% (  0.7)   hald-addon-stor : schedule_timeout (process_timeout)
   0.1% (  0.7)   knetworkmanager : schedule_timeout (process_timeout)
   0.1% (  0.3)     <kernel core> : end_that_request_last (laptop_timer_fn)




*** After Disabling DRI ***

Cn                Avg residency       P-states (frequencies)
C0 (cpu running)        ( 3.3%)         1.60 Ghz     0.0%
C1                0.0ms ( 0.0%)         1333 Mhz     0.0%
C2                8.5ms (29.8%)         1067 Mhz     0.0%
C3               14.2ms (27.0%)          800 Mhz   100.0%
C4               10.5ms (39.9%)

Wakeups-from-idle per second : 92.0     interval: 3.0s
Power usage (ACPI estimate): 14.6W (0.9 hours)

Top causes for wakeups:
  24.8% ( 17.3)       <interrupt> : libata
  15.2% ( 10.7)       firefox-bin : futex_wait (hrtimer_wakeup)
  14.3% ( 10.0)   hald-addon-cpuf : cpufreq_governor_dbs 
(delayed_work_timer_fn)
   8.6% (  6.0)            kicker : schedule_timeout (process_timeout)
   7.6% (  5.3)       <interrupt> : ipw2200, Intel ICH6
   4.8% (  3.3)              Xorg : do_setitimer (it_real_fn)
   3.3% (  2.3)   thunderbird-bin : futex_wait (hrtimer_wakeup)
   2.4% (  1.7)    wpa_supplicant : schedule_timeout (process_timeout)
   1.9% (  1.3)              Xorg : schedule_timeout (process_timeout)
   1.4% (  1.0)          ifconfig : tg3_open (tg3_timer)
   1.4% (  1.0)              kwin : schedule_timeout (process_timeout)
   1.4% (  1.0)          kdesktop : schedule_timeout (process_timeout)
   1.4% (  1.0)    NetworkManager : schedule_timeout (process_timeout)
   1.4% (  1.0)          kwrapper : do_nanosleep (hrtimer_wakeup)
   1.4% (  1.0)     <kernel core> : queue_delayed_work_on 
(delayed_work_timer_fn)
   1.4% (  1.0)            dhcdbd : schedule_timeout (process_timeout)





> Hi All,
> Is there a patch that applies cleanly to 2.6.22.x kernels to
> force HPET on.  I have an Intel ICH6 chipset and as far as
> I know it should have this functionality but it isn't being
> enabled by default.
>
> In addition I have a strange problem with an ethernet interrupt
> which is being triggered even though it isn't connected to any
> network.  The ouput from powertop is below.
>
>
> 35.9% ( 66.0)       <interrupt> : extra timer interrupt
> 23.2% ( 42.7)       <interrupt> : eth0, i915 at pci:0000:00:02.0
> 11.3% ( 20.7)       <interrupt> : libata
> 5.4% ( 10.0)       <interrupt> : Intel ICH6, ipw2200
> 5.3% (  9.7)   hald-addon-cpuf : cpufreq_governor_dbs 
> (delayed_work_timer_fn)
> 4.0% (  7.3)       <interrupt> : acpi
> 3.4% (  6.3)            kicker : schedule_timeout (process_timeout)
> 2.0% (  3.7)              Xorg : do_setitimer (it_real_fn)
> 1.1% (  2.0)   thunderbird-bin : futex_wait (hrtimer_wakeup)
> 0.7% (  1.3)              Xorg : schedule_timeout (process_timeout)
> 0.7% (  1.3)    wpa_supplicant : schedule_timeout (process_timeout)
> 0.5% (  1.0)          kdesktop : schedule_timeout (process_timeout)
> 0.5% (  1.0)            dhcdbd : schedule_timeout (process_timeout)
> 0.5% (  1.0)          kwrapper : do_nanosleep (hrtimer_wakeup)
> 0.5% (  1.0)     <kernel core> : neigh_table_init_no_netlink 
> (neigh_periodic_timer)
> 0.5% (  1.0)          ifconfig : tg3_open (tg3_timer)
>
> If anyone has any ideas about these I would really appreciate the help.
>
> Cheers,
> Kyle
>
> _______________________________________________
> Power mailing list
> Power at bughost.org
> http://www.bughost.org/mailman/listinfo/power