Re: [Patch 3/8] cpu delay collection via schedstats

[Andrew Morton <akpm@osdl.org>]

Balbir Singh <balbir@in.ibm.com> wrote:
>
> +/*
> + * Expects runqueue lock to be held for atomicity of update
> + */
> +static inline void rq_sched_info_arrive(struct runqueue *rq,
> +						unsigned long diff)
> +{
> +	if (rq) {
> +		rq->rq_sched_info.run_delay += diff;
> +		rq->rq_sched_info.pcnt++;
> +	}
> +}
> +
> +/*
> + * Expects runqueue lock to be held for atomicity of update
> + */
> +static inline void rq_sched_info_depart(struct runqueue *rq,
> +						unsigned long diff)
> +{
> +	if (rq)
> +		rq->rq_sched_info.cpu_time += diff;
> +}

The kernel has many different units of time - jiffies, cpu ticks, ns, us,
ms, etc.  So the reader of these functions doesn't have a clue what "diff"
is.

A good way to remove all doubt in all cases is to include the units in the
variable's name.  Something like delta_jiffies, perhaps.