Hi Paul,
On Thu, Feb 24, 2022 at 06:44:23AM -0800, Paul E. McKenney wrote:
[...]
> > > > > Rafael replied with one possible option. Alternatively when planing to
> > > > > enter deep idle, set P-state to min with a callback like we do in
> > > > > offline callback.
> > > >
> > > > Yes, if the system is going to idle, it makes sense to goto a lower
> > > > cpufreq first (also what my debug patch will essentially lead to).
> > > >
> > > > Given cprfreq-util's normal running frequency is every 10ms, doing
> > > > this before entering idle is not a big extra burden.
> > >
> > > But this is not related to idle as such, but to the fact that idle
> > > sometimes stops the scheduler tick which otherwise would run the
> > > cpufreq governor callback on a regular basis.
> > >
> > > It is stopping the tick that gets us into trouble, so I would avoid
> > > doing it if the current performance state is too aggressive.
> >
> > I've tried to simulate Doug's environment by using his kconfig, and
> > offline my 36 CPUs Desktop to leave 12 CPUs online, and on it I can
> > still see Local timer interrupts when there is no active load, with
> > the longest interval between 2 timer interrupts is 4 seconds, while
> > idle class's task_tick_idle() will do nothing, and CFS'
> > task_tick_fair() will in turn call cfs_rq_util_change()
>
> Every four seconds? Could you please post your .config?
Aha, I didn't make it clear, that the timer interrupt was not always
coming every 4 seconds, but when system is silent, the maxim interval
between 2 timer interrupts was 4 seconds.
When initially I checked this, I doubted if the timer interrupt are
too few on the system, so I used Doug's config and tried to make my
desktop silent (like disabling GUI), following is some trace_printk
log, though I figured out later when idle thread is running, the
idle class' scheduler tick will not help as it doesn't call cpufreq
callback.
<idle>-0 [009] d.h1. 235.980053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.981054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.982053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.983053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.984053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.985053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.986054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.987054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.988054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.989054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.990054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.991053: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.992054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.993054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 235.994054: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 236.331126: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 236.460130: hrtimer_interrupt: enter
<idle>-0 [009] d.s5. 236.460147: intel_pstate_update_util: old_state=48 new=27
<idle>-0 [009] d.h1. 238.380130: hrtimer_interrupt: enter
<idle>-0 [009] d.s5. 238.380147: intel_pstate_update_util: old_state=27 new=12
<idle>-0 [009] d.h1. 240.331133: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 240.364133: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 244.331135: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 248.331139: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 252.331138: hrtimer_interrupt: enter
<...>-1167 [009] d.h.. 254.860056: hrtimer_interrupt: enter
snapd-1128 [009] d.h.. 254.861054: hrtimer_interrupt: enter
snapd-1128 [009] d.h.. 254.862055: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 254.863056: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 254.864056: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 254.865055: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 256.331133: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 260.331127: hrtimer_interrupt: enter
<idle>-0 [009] d.h1. 264.331135: hrtimer_interrupt: enter
Thanks,
Feng
> Thanx, Paul
