linux-kernel - Re: [PATCH 2/3] cpuidle: Add some comments in the cpuidle

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 15 Apr 2015 18:07:07 +0200
From:	Daniel Lezcano <daniel.lezcano@...aro.org>
To:	"Rafael J. Wysocki" <rafael@...nel.org>
CC:	Peter Zijlstra <peterz@...radead.org>,
	"Rafael J. Wysocki" <rjw@...ysocki.net>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	nicolas.pitre@...aro.org
Subject: Re: [PATCH 2/3] cpuidle: Add some comments in the cpuidle_enter function

On 04/15/2015 03:46 PM, Rafael J. Wysocki wrote:
> On Wed, Apr 15, 2015 at 12:00 PM, Daniel Lezcano
> <daniel.lezcano@...aro.org> wrote:
>> The code is a bit poor in comments. Fix that by adding some comments in the
>> cpuidle enter function.
>>
>> Signed-off-by: Daniel Lezcano <daniel.lezcano@...aro.org>
>> ---
>>   drivers/cpuidle/cpuidle.c | 31 +++++++++++++++++++++++++++++++
>>   1 file changed, 31 insertions(+)
>>
>> diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
>> index 1220dac..5e6c6be 100644
>> --- a/drivers/cpuidle/cpuidle.c
>> +++ b/drivers/cpuidle/cpuidle.c
>> @@ -162,19 +162,50 @@ int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv,
>>
>>          trace_cpu_idle_rcuidle(index, dev->cpu);
>>
>> +       /*
>> +        * Store the idle start time for this cpu, this information
>> +        * will be used by cpuidle to measure how long the cpu has
>> +        * been idle and by the scheduler to prevent to wake it up too
>> +        * early
>> +        */
>>          target_state->idle_stamp = ktime_to_us(ktime_get());
>>
>> +       /*
>> +        * The enter to the low level idle routine. This call will block
>> +        * until an interrupt occurs meaning it is the end of the idle
>> +        * period
>> +        */
>>          entered_state = target_state->enter(dev, drv, index);
>>
>> +       /*
>> +        * Measure as soon as possible the duration of the idle
>> +        * period. It MUST be done before re-enabling the interrupt in
>> +        * order to prevent to add in the idle time measurement the
>> +        * interrupt handling duration
>
> That's unless ->enter() itself re-enables interrupts which it may do, right?

Except I missed a change somewhere, I believe I removed all the 
interrupts re-enablement in all the driver with latest commit 554c06ba3.

This is why we were able to move the local_irq_enable in this code path 
and finally remove the CPUIDLE_FLAG_TIME_VALID.

> Which is why we made governors use next_timer_us as the "measured"
> value if the measured value itself is greater.
>
> I'd just say "Compute the idle duration here to avoid adding interrupt
> handling time to the idle time in case an interrupt occurs as soon as
> re-enabled".

Ok, sounds good.

>> +        */
>>          diff = ktime_to_us(ktime_sub_us(ktime_get(), target_state->idle_stamp));
>>
>> +       /*
>> +        * Reset the idle time stamp as the scheduler may think the cpu is idle
>> +        * while it is in the process of waking up
>> +        */
>>          target_state->idle_stamp = 0;
>>
>>          trace_cpu_idle_rcuidle(PWR_EVENT_EXIT, dev->cpu);
>>
>> +       /*
>> +        * The cpuidle_enter_coupled uses the cpuidle_enter function.
>> +        * Don't re-enable the interrupts and let the enter_coupled
>
> "let the enter_coupled function wait" (without the "to").

Ok, thanks.

>> +        * function to wait for all cpus to sync and to enable the
>> +        * interrupts again from there
>> +        */
>
> And I would just say "In the coupled case interrupts will be enabled
> by cpuidle_enter_coupled()".

Ok.

>>          if (!cpuidle_state_is_coupled(dev, drv, entered_state))
>>                  local_irq_enable();
>>
>> +       /*
>> +        * The idle duration will be casted to an integer, prevent to
>
> "will be cast" (like "Cast Away").

Yep.

>> +        * overflow by setting a boundary to INT_MAX
>> +        */
>>          if (diff > INT_MAX)
>>                  diff = INT_MAX;
>
> But anyway, instead of adding that comment, why not to change the code
> like this:
>
> dev->last_residency = diff < INT_MAX ? diff : INT_MAX;
>
> and it now should be clear what happens without the comment.

Ok.

Thanks for the review.

   -- Daniel


-- 
  <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

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/