Re: [PATCH 2/2] x86/idle: use dynamic halt poll

[Yang Zhang]

On 2017/6/22 22:32, Thomas Gleixner wrote:
> On Thu, 22 Jun 2017, root wrote:
> > @@ -962,6 +962,7 @@ __visible void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
> >  	 * interrupt lock, which is the WrongThing (tm) to do.
> >  	 */
> >  	entering_ack_irq();
> > +	check_poll();
>
> No way, that we sprinkle this function into every interrupt hotpath. There
> are enough genuine ways to do that w/o touching a gazillion of files.

I will find a more correct place to call this function.

> >  #ifdef CONFIG_HYPERVISOR_GUEST
> > +static unsigned int grow_poll_ns(unsigned int old, unsigned int grow,
> > +				      unsigned int max)
> > +{
> > +	unsigned int val;
> > +
> > +	/* 10us as base poll duration */
> > +	if (old == 0 && grow)
> > +		return 10000;
> > +
> > +	val = old * grow;
> > +	if (val > max)
> > +		val = max;
> > +
> > +	return val;
> > +}
> > +
> > +static unsigned int shrink_poll_ns(unsigned int old, unsigned int shrink)
> > +{
> > +	if (shrink == 0)
> > +		return 0;
> > +
> > +	return old / shrink;
> > +}
> > +
> > +void check_poll(void)
> > +{
> > +	unsigned int val, poll_duration;
> > +	unsigned long begin_ns, now_ns;
> > +
> > +	if (!poll_threshold_ns)
> > +		return;
>
> If at all then this needs to be a static key based decision.

Sure, will do it.

> > +
> > +	begin_ns = this_cpu_read(poll_begin_ns);
> > +	/* Not from halt state */
> > +	if (!begin_ns)
> > +		return;
>
> If you integrate this stuff into the proper place, then the whole mess goes
> away. We really do not need another facility to track idle state. We have
> enough already, really.

Agree, I will check current code to find a more proper way to do the check.

> Thanks,
>
> 	tglx


--
Yang
Alibaba Cloud Computing