[PATCH 12/14] powerpc/64s: cpuidle no memory barrier after break from idle

[Vaidyanathan Srinivasan]

* Nicholas Piggin <npiggin at gmail.com> [2017-06-12 09:58:33]:

> A memory barrier is not required after the task wakes up,
> only if we clear the polling flag before waking. The case
> where we have work to do is the important one, so optimise
> for it.
> 
> Signed-off-by: Nicholas Piggin <npiggin at gmail.com>

Reviewed-by: Vaidyanathan Srinivasan <svaidy at linux.vnet.ibm.com>


> ---
>  drivers/cpuidle/cpuidle-powernv.c | 11 +++++++++--
>  drivers/cpuidle/cpuidle-pseries.c | 11 +++++++++--
>  2 files changed, 18 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/cpuidle/cpuidle-powernv.c b/drivers/cpuidle/cpuidle-powernv.c
> index 9d03326ac05e..37b0698b7193 100644
> --- a/drivers/cpuidle/cpuidle-powernv.c
> +++ b/drivers/cpuidle/cpuidle-powernv.c
> @@ -59,14 +59,21 @@ static int snooze_loop(struct cpuidle_device *dev,
>  	ppc64_runlatch_off();
>  	HMT_very_low();
>  	while (!need_resched()) {
> -		if (likely(snooze_timeout_en) && get_tb() > snooze_exit_time)
> +		if (likely(snooze_timeout_en) && get_tb() > snooze_exit_time) {
> +			/*
> +			 * Task has not woken up but we are exiting the polling
> +			 * loop anyway. Require a barrier after polling is
> +			 * cleared to order subsequent test of need_resched().
> +			 */
> +			clear_thread_flag(TIF_POLLING_NRFLAG);
> +			smp_mb();
>  			break;
> +		}
>  	}
> 
>  	HMT_medium();
>  	ppc64_runlatch_on();
>  	clear_thread_flag(TIF_POLLING_NRFLAG);
> -	smp_mb();

If we reach here without executing if(snooze_timeout) with the
barrier+break, that means we have seen the need_resched flag and hence
we can avoid the barrier.  Clearing of the polling flag can be seen
(take affect) later as we will exit the idle loop and re-enter anyway
at this point.  The caller also will check for need_resched and exit
the idle loop.

Actually do_idle() has a __current_set_polling() and
__current_clr_polling() in the default idle loop.  Do we really need
to set/clear the TIF_POLLING_NRFLAG again in cpuidle driver?

--Vaidy