Re: [PATCH] tick/nohz: Make the idle_exittime update correctly

[Yunfeng Ye <yeyunfeng@huawei.com>]

On 2020/12/17 21:59, Frederic Weisbecker wrote:
> On Thu, Dec 17, 2020 at 02:51:58PM +0800, Yunfeng Ye wrote:
>>
>>
>> On 2020/12/15 22:47, Frederic Weisbecker wrote:
>>> On Tue, Dec 15, 2020 at 08:06:34PM +0800, Yunfeng Ye wrote:
>>>> The idle_exittime field of tick_sched is used to record the time when
>>>> the idle state was left. but currently the idle_exittime is updated in
>>>> the function tick_nohz_restart_sched_tick(), which is not always in idle
>>>> state when nohz_full is configured.
>>>>
>>>>   tick_irq_exit
>>>>     tick_nohz_irq_exit
>>>>       tick_nohz_full_update_tick
>>>>         tick_nohz_restart_sched_tick
>>>>           ts->idle_exittime = now;
>>>>
>>>> So move to tick_nohz_stop_idle() to make the idle_exittime update
>>>> correctly.
>>>>
>>>> Signed-off-by: Yunfeng Ye <yeyunfeng@huawei.com>
>>>> ---
>>>>  kernel/time/tick-sched.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
>>>> index 749ec2a583de..be2e5d772d50 100644
>>>> --- a/kernel/time/tick-sched.c
>>>> +++ b/kernel/time/tick-sched.c
>>>> @@ -591,6 +591,7 @@ static void tick_nohz_stop_idle(struct tick_sched *ts, ktime_t now)
>>>>  {
>>>>  	update_ts_time_stats(smp_processor_id(), ts, now, NULL);
>>>>  	ts->idle_active = 0;
>>>> +	ts->idle_exittime = now;
>>>
>>> This changes a bit the meaning of idle_exittime then since this is also called
>>> from idle interrupt entry.
>>>
>>> __tick_nohz_idle_restart_tick() would be a better place.
>>>
>> So is it necessary to modify the comment "@idle_exittime:      Time when the idle state was left" ?
>>
>> On the other hand, if the patch "nohz: Update tick instead of restarting tick in tick_nohz_idle_exit()"
>> (https://www.spinics.net/lists/kernel/msg3747039.html ) applied, __tick_nohz_idle_restart_tick will not
>> be called always, So is it put here also a better place?
> 
> Right but I need to re-order some code before. That's ok, I'll integrate this
> patch inside the changes.
> 
Ok, thanks.

> Thanks.
> .
>