Re: [PATCH] hrtimer: avoid retrigger_next_event IPI

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, Apr 07 2021 at 10:53, Marcelo Tosatti wrote:
> Setting the realtime clock triggers an IPI to all CPUs to reprogram
> hrtimers.
>
> However, only base, boottime and tai clocks have their offsets updated

base clock? And why boottime? Boottime is not affected by a clock
realtime set. It's clock REALTIME and TAI, nothing else.

> +#define CLOCK_SET_BASES ((1U << HRTIMER_BASE_REALTIME)|		\
> +			 (1U << HRTIMER_BASE_REALTIME_SOFT)|	\
> +			 (1U << HRTIMER_BASE_BOOTTIME)|		\
> +			 (1U << HRTIMER_BASE_BOOTTIME_SOFT)|	\
> +			 (1U << HRTIMER_BASE_TAI)|		\
> +			 (1U << HRTIMER_BASE_TAI_SOFT))
> +
> +static bool need_reprogram_timer(struct hrtimer_cpu_base *cpu_base)
> +{
> +	unsigned int active = 0;
> +
> +	if (!cpu_base->softirq_activated)
> +		active = cpu_base->active_bases & HRTIMER_ACTIVE_SOFT;
> +
> +	active = active | (cpu_base->active_bases & HRTIMER_ACTIVE_HARD);
> +
> +	if ((active & CLOCK_SET_BASES) == 0)
> +		return false;
> +
> +	return true;
> +}

Errm. 

> +	/* Avoid interrupting nohz_full CPUs if possible */
> +	preempt_disable();
> +	for_each_online_cpu(cpu) {
> +		if (tick_nohz_full_cpu(cpu)) {
> +			unsigned long flags;
> +			struct hrtimer_cpu_base *cpu_base = &per_cpu(hrtimer_bases, cpu);
> +
> +			raw_spin_lock_irqsave(&cpu_base->lock, flags);
> +			if (need_reprogram_timer(cpu_base))
> +				cpumask_set_cpu(cpu, mask);
> +			else
> +				hrtimer_update_base(cpu_base);
> +			raw_spin_unlock_irqrestore(&cpu_base->lock, flags);
> +		}
> +	}

How is that supposed to be correct?

CPU0                    	CPU1

clock_was_set()                 hrtimer_start(CLOCK_REALTIME)

  if (!active_mask[CPU1] & XXX)
  	continue;
                                active_mask |= REALTIME;

---> fail because that newly started timer is on the old offset.

Thanks,

        tglx