Re: [RFC PATCH 2/2] sched: idle: IRQ based next prediction for idle period

[Daniel Lezcano <daniel.lezcano@linaro.org>]

On 01/12/2016 03:26 PM, Thomas Gleixner wrote:
> On Tue, 12 Jan 2016, Daniel Lezcano wrote:
>> On 01/12/2016 02:42 PM, Thomas Gleixner wrote:
>>> On Tue, 12 Jan 2016, Daniel Lezcano wrote:
>>>> On 01/08/2016 04:43 PM, Thomas Gleixner wrote:
>>>>>> +	/*
>>>>>> +	 * Register the setup/free irq callbacks, so new interrupt or
>>>>>> +	 * freed interrupt will update their tracking.
>>>>>> +	 */
>>>>>> +	ret = register_irq_timings(&irqt_ops);
>>>>>> +	if (ret) {
>>>>>> +		pr_err("Failed to register timings ops\n");
>>>>>> +		return ret;
>>>>>> +	}
>>>>>
>>>>> So that stuff is installed unconditionally. Is it used unconditionally
>>>>> as
>>>>> well?
>>>>
>>>> Sorry, I am not sure to understand your question. If the kernel is
>>>> compiled
>>>> with CONFIG_CPU_IDLE_GOV_SCHED=y, this code is enabled and use the irq
>>>> timings. The condition comes from the compilation option.
>>>
>>> The question is whether the option also activates that thing or is there
>>> still
>>> some /sys/whatever/idlegov magic where you can (de)select it.
>>
>> Yes, in the next patches of the series I did not send, we can switch to the
>> cpuidle's governor framework or idle-sched. I will look at how to disable it
>> when switching to the cpuidle's governors.
>
> You better implement the switching part in the cpuidle core first, i.e. proper
> callbacks when a governor is switched in/out. Then make use of this switcheroo
> right away. Doing it the other way round is just wrong.

The problem is this code is not another governor but a 'predictor' where 
the scheduler will use the information to ask the cpuidle to go to a 
specific idle state without going through the governor code, so into the 
governor's callbacks. It is on top of cpuidle. The scheduler will become 
the governor.

The current straightforward code, does the switch in the cpu_idle_loop 
idle_task's function:

[ ... ]

if (cpu_idle_force_poll || tick_check_broadcast_expired())
	cpu_idle_poll();
else {
	if (sched_idle_enabled()) {
		int latency = pm_qos_request(PM_QOS_CPU_DMA_LATENCY);
		s64 duration = sched_idle_next_wakeup();
		sched_idle(duration, latency);
	} else {
		cpuidle_idle_call();
	}
}

Due to the complexity of the code, this first step introduce a mechanism 
to predict the next event and re-use it trivially in the idle task.

Perhaps, it would be acceptable to have cpuidle_idle_call() to be 
replaced by a callback and the switch acts at this level ?



-- 
  <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog