lkml.org 
[lkml]   [2008]   [Mar]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Too many timer interrupts in NO_HZ
* Arjan van de Ven <arjan@infradead.org> [2008-03-02 11:57:09]:

> On Mon, 3 Mar 2008 01:18:13 +0530
> Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> wrote:
>
> > Hi,
> >
> > I am trying to play around with the tickless kernel and
> > sched_mc_power_savings tunables to increase the idle duration of the
> > CPU in an SMP server. sched_mc_power_savings provides good
> > consolidation of long running jobs when the system is lightly loaded.
> > However I am interested to see CPU idle times for couple minutes on
> > atleast few CPUs in an completely idle SMP system.
> >
> > I have posted related instrumentation results at
> > http://lkml.org/lkml/2008/1/8/243
> >
> > The experimental setup is identical. Just that I have tried to
> > collect more data and also bias the wakeup logic.
> >
> > The kernel is 2.6.25-rc3 with NO_HZ, SCHED_MC, HIGH_RES_TIMERS, and
> > HPET turned on.
> >
> > I have tried to collect interesting data from tick-sched.c, sched.c
> > and softirq.c.
> >
> > As suggested by Ingo in the previous thread, I have hacked the sched.c
> > try_to_wake_up logic so that I can force all wake-ups to happen on
> > CPU0. This is just and experiment, in reality we will have to
> > carefully choose a domain and CPU to bias all wakeups on an idle
> > system.
>
>
> please show us powertop -d output; that way we can at least see sort of what
> is going on...

I have some more information on the timer interrupt activities in the
system. I have obtained the following trace from irq_exit and local
timer is the only interrupt on CPUs 1, 2 & 3.

Trace CPU/function Timestamp
number

U2336208 C1 irq_exit EXIT 4452403.687322 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336212 C3 irq_exit EXIT 4452403.722253 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336214 C2 irq_exit EXIT 4452403.722668 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336236 C1 irq_exit EXIT 4452807.345543 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336241 C3 irq_exit EXIT 4452807.381894 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336242 C2 irq_exit EXIT 4452807.382144 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336248 C1 irq_exit EXIT 4453211.003638 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336253 C3 irq_exit EXIT 4453211.041529 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336254 C2 irq_exit EXIT 4453211.041703 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336260 C1 irq_exit EXIT 4453614.662043 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336265 C3 irq_exit EXIT 4453614.701429 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336266 C2 irq_exit EXIT 4453614.701474 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336293 C1 irq_exit EXIT 4454018.320078 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336298 C3 irq_exit EXIT 4454018.361100 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336299 C2 irq_exit EXIT 4454018.361142 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336313 C1 irq_exit EXIT 4454421.978047 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336318 C2 irq_exit EXIT 4454422.020756 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336319 C3 irq_exit EXIT 4454422.020807 ms in_intr 0 in_idle 1 need_resched 0 pid 0


U2336357 C1 irq_exit EXIT 4454825.636341 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336369 C2 irq_exit EXIT 4454825.681058 ms in_intr 0 in_idle 1 need_resched 0 pid 0
U2336370 C3 irq_exit EXIT 4454825.681191 ms in_intr 0 in_idle 1 need_resched 0 pid 0

This is over a short interval where I complain that the tick is not
really getting stopped. Actually there is a pattern of 400ms timer
being programmed.

Apart from these there are 4ms pattern on one of the CPU that is used
to drive tick_sched_timer. Most of the missing trace entries are CPU0
which takes care of all the idle daemon process.

There are wakeups due to kondemand and other kernel threads on CPUs 1,
2 & 3, but the time is in seconds.

Please help me find the source for this timer.

Thanks,
Vaidy



\
 
 \ /
  Last update: 2008-03-05 16:41    [W:0.188 / U:0.524 seconds]
©2003-2014 Jasper Spaans. hosted at Digital Ocean