Re: [RFC PATCH 3/6] sched: Add over-utilization/tipping point indicator

[Peter Zijlstra]

(I know there is a new version out; but I was reading through this to
catch up with the discussion)

On Tue, Mar 20, 2018 at 09:43:09AM +0000, Dietmar Eggemann wrote:
> +static inline int sd_overutilized(struct sched_domain *sd)
> +{
> +	return READ_ONCE(sd->shared->overutilized);
> +}
> +
> +static inline void update_overutilized_status(struct rq *rq)
> +{
> +	struct sched_domain *sd;
> +
> +	rcu_read_lock();
> +	sd = rcu_dereference(rq->sd);
> +	if (sd && !sd_overutilized(sd) && cpu_overutilized(rq->cpu))
> +		WRITE_ONCE(sd->shared->overutilized, 1);
> +	rcu_read_unlock();
> +}
> +#else

I think you ought to go have a look at the end of
kernel/sched/topology.c:sd_init(), where it says:

	/*
	 * For all levels sharing cache; connect a sched_domain_shared
	 * instance.
	 */
	if (sd->flags & SD_SHARE_PKG_RESOURCES) {
		sd->shared = *per_cpu_ptr(sdd->sds, sd_id);
		atomic_inc(&sd->shared->ref);
		atomic_set(&sd->shared->nr_busy_cpus, sd_weight);
	}

Because if I read all this correctly, your code assumes sd->shared
exists unconditionally, while the quoted bit only ensures it does so <=
LLC.