Re: Regression with suspicious RCU usage splats with cpu_pm change

From: Alex Shi
Date: Sun Jul 16 2017 - 19:24:43 EST




On 07/13/2017 08:43 PM, Paul E. McKenney wrote:
> On Thu, Jul 13, 2017 at 01:50:26PM +0200, Rafael J. Wysocki wrote:
>> On Thu, Jul 13, 2017 at 11:43 AM, Alex Shi <alex.shi@xxxxxxxxxx> wrote:
>>>
>>> On 07/13/2017 03:07 PM, Tony Lindgren wrote:
>>>> Hi,
>>>>
>>>> Looks like next-20170713 gives me a bunch of "suspicious RCU usage"
>>>> splats with cpuidle_coupled on duovero, see below. I bisected it down
>>>> to commit 2f027e003d05 ("cpu_pm: replace raw_notifier with
>>>> atomic_notifier").
>>
>> OK, so I'm dropping this commit.
>
> You can surround idle-loop RCU-reading code with RCU_NONIDLE().
> This will tell RCU to pay attention even though the CPU is otherwise
> idle.
>
> Thanx, Paul
>


Thanks a lot, Paul! :)
I reused the rcu_irq_enter_irqson() from RCU_NONIDLE to avoid this issue.
It works fine.

Tony, Could you like to give a tested-by if this patch works for you.

Sebastian,
May I keep your acked-by with new fixed patch, since the main thing remained? :)


Thanks everyone!

======