Rescheduling Interrupts

[Thomas Meyer]

Arjan van de Ven schrieb:
> Erik Andrén wrote:
>   
>> Hi,
>>
>> 2007/12/9, Arjan van de Ven <arjan at linux.intel.com 
>> <mailto:arjan at linux.intel.com>>:
>>
>>     Peter Hüwe wrote:
>>      > Hmm but the cpu is almost 100% idle ?
>>      > Or do you mean by overloaded that too much interrupts are
>>     occuring and that
>>
>>     No I mean that the scheduler thinks the current CPU is "too busy" and
>>     that it needs to overflow to another one
>>
>>     Personally I think we are seeing a scheduler bug here (and yes I'm
>>     mostly seeing the same behavior; I need to trace it down somehow)
>>
>>
>>
>> Did you manage to track this issue down?
>> Otherwise, shouldn't a kernel bug be opened to track down this issue 
>> before the 2.6.24-final is released?
>>
>>     
>
> at this point it'll be a 2.6.25 issue I suspect, but yes it'll be good to report this to lkml at least
Still happens with 2.6.25-rc1:

# powertop -d
PowerTOP 1.9    (C) 2007 Intel Corporation

Sammle Daten für 15 Sekunden
Cn                 Verweildauer
C0 (Prozessor läuft)    ( 0,3%)
C1                0,0ms ( 0,0%)
C2                0,0ms ( 0,0%)
C3               21,5ms (99,7%)
P-States (Frequenzen)
  1,84 GHz   100,0%
  1,67 GHz     0,0%
  1500 MHz     0,0%
  1333 MHz     0,0%
Aufwachen pro Sekunde : 46,5    Intervall: 15,0s
Keine ACPI Stromverbrauch-Schätzung verfügbar
Häufigste Ursachen für das Aufwachen:
  36,9% ( 24,4)      <kernel IPI> : Rescheduling interrupts
  23,1% ( 15,3)       <interrupt> : wifi0
  18,1% ( 12,0)   <kernel module> : usb_hcd_poll_rh_status (rh_timer_func)
  11,6% (  7,7)     group_balance : schedule_timeout (process_timeout)
   1,5% (  1,0)           ifplugd : schedule_timeout (process_timeout)
   1,5% (  1,0)              kwin : schedule_timeout (process_timeout)
   1,5% (  1,0)          kwrapper : do_nanosleep (hrtimer_wakeup)
   0,9% (  0,6)                 X : do_setitimer (it_real_fn)
   0,8% (  0,5)              kded : schedule_timeout (process_timeout)
   0,7% (  0,5)         gpg-agent : schedule_timeout (process_timeout)
   0,4% (  0,3)           konsole : schedule_timeout (process_timeout)
   0,4% (  0,3)         klauncher : schedule_timeout (process_timeout)
   0,4% (  0,3)          kdesktop : schedule_timeout (process_timeout)
   0,3% (  0,2)                ip : __netdev_watchdog_up (dev_watchdog)
   0,3% (  0,2)              init : schedule_timeout (process_timeout)
   0,3% (  0,2)              nscd : futex_wait (hrtimer_wakeup)
   0,2% (  0,1)                 X : schedule_timeout (process_timeout)
   0,2% (  0,1)     <kernel core> : neigh_table_init_no_netlink 
(neigh_periodic_timer)
   0,1% (  0,1)       <interrupt> : uhci_hcd:usb1, ehci_hcd:usb5
   0,1% (  0,1)      <kernel IPI> : function call interrupts
   0,1% (  0,1)     <kernel core> : input_handle_event (input_repeat_key)
   0,1% (  0,1)   <kernel module> : ieee80211_node_attach 
(ieee80211_node_timeout)
   0,1% (  0,1)     <kernel core> : init_nonfatal_mce_checker 
(delayed_work_timer_fn)
   0,1% (  0,1)     <kernel core> : ip_rt_init (delayed_work_timer_fn)
   0,1% (  0,1)              nscd : neigh_add_timer (neigh_timer_handler)
   0,1% (  0,1)   USB Gerät  1-2 : Apple Internal Keyboard / Trackpad 
(Apple Computer)


Any new hints?

mfg
thomas