Re: [PATCH V2 1/4] cpufreq: stats: Defer stats update to cpufreq_stats_record_transition()

[Lukasz Luba]

On 9/16/20 7:45 AM, Viresh Kumar wrote:
> In order to prepare for lock-less stats update, add support to defer any
> updates to it until cpufreq_stats_record_transition() is called.
>
> Signed-off-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
> ---
>   drivers/cpufreq/cpufreq_stats.c | 75 ++++++++++++++++++++++++---------
>   1 file changed, 56 insertions(+), 19 deletions(-)

[snip]

> @@ -228,10 +264,11 @@ void cpufreq_stats_record_transition(struct cpufreq_policy *policy,
>   	struct cpufreq_stats *stats = policy->stats;
>   	int old_index, new_index;
>   
> -	if (!stats) {
> -		pr_debug("%s: No stats found\n", __func__);
> +	if (!stats)
>   		return;
> -	}
> +
> +	if (READ_ONCE(stats->reset_pending))
> +		cpufreq_stats_reset_table(stats);
>   

This is in the hot path code, called from the scheduler. I wonder if we
avoid it or make that branch 'unlikely'?

if (unlikely(READ_ONCE(stats->reset_pending)))

Probably the CPU (when it has good prefetcher) would realize about it,
but maybe we can help a bit here.