Re: [PATCH v6 2/6] sched_ext: Implement scx_bpf_now_ns()

[Tejun Heo <tj@kernel.org>]

Hello,

On Fri, Dec 20, 2024 at 03:20:21PM +0900, Changwoo Min wrote:
...
> +__bpf_kfunc u64 scx_bpf_now_ns(void)

Given that the default time unit is ns for the scheduler, the _ns suffix
can probably go.

> +{
> +	struct rq *rq;
> +	u64 clock;
> +
> +	preempt_disable();
> +
> +	/*
> +	 * If the rq clock is valid, use the cached rq clock.
> +	 * Otherwise, return a fresh rq glock.
> +	 *
> +	 * Note that scx_bpf_now_ns() is re-entrant between a process
> +	 * context and an interrupt context (e.g., timer interrupt).
> +	 * However, we don't need to consider the race between them
> +	 * because such race is not observable from a caller.
> +	 */
> +	rq = this_rq();
> +	clock = READ_ONCE(rq->scx.clock);
> +
> +	if (!(READ_ONCE(rq->scx.flags) & SCX_RQ_CLK_VALID)) {
> +		clock = sched_clock_cpu(cpu_of(rq));
> +
> +		/*
> +		 * The rq clock is updated outside of the rq lock.
> +		 * In this case, keep the updated rq clock invalid so the next
> +		 * kfunc call outside the rq lock gets a fresh rq clock.
> +		 */
> +		scx_rq_clock_update(rq, clock, false);

Hmm... what does this update do?

...
> +static inline void scx_rq_clock_update(struct rq *rq, u64 clock, bool valid)
> +{
> +	if (!scx_enabled())
> +		return;
> +	WRITE_ONCE(rq->scx.clock, clock);
> +	if (valid)
> +		WRITE_ONCE(rq->scx.flags, rq->scx.flags | SCX_RQ_CLK_VALID);
> +}

Isn't rq->scx.clock used iff VALID is set? If so, why does !VALID read need
to update rq->scx.clock?

Thanks.

-- 
tejun