Re: [patch 14/19] softirq: Make softirq control and processing RT aware

[Frederic Weisbecker]

On Fri, Nov 13, 2020 at 03:02:21PM +0100, Thomas Gleixner wrote:
> +void __local_bh_enable_ip(unsigned long ip, unsigned int cnt)
> +{
> +	bool preempt_on = preemptible();
> +	unsigned long flags;
> +	u32 pending;
> +	int curcnt;
> +
> +	WARN_ON_ONCE(in_irq());
> +	lockdep_assert_irqs_enabled();
> +
> +	local_irq_save(flags);
> +	curcnt = this_cpu_read(softirq_ctrl.cnt);
> +
> +	/*
> +	 * If this is not reenabling soft interrupts, no point in trying to
> +	 * run pending ones.
> +	 */
> +	if (curcnt != cnt)
> +		goto out;
> +
> +	pending = local_softirq_pending();
> +	if (!pending || ksoftirqd_running(pending))
> +		goto out;
> +
> +	/*
> +	 * If this was called from non preemptible context, wake up the
> +	 * softirq daemon.
> +	 */
> +	if (!preempt_on) {
> +		wakeup_softirqd();
> +		goto out;
> +	}
> +
> +	/*
> +	 * Adjust softirq count to SOFTIRQ_OFFSET which makes
> +	 * in_serving_softirq() become true.
> +	 */
> +	cnt = SOFTIRQ_OFFSET;
> +	__local_bh_enable(cnt, false);

But then you enter __do_softirq() with softirq_count() == SOFTIRQ_OFFSET.
__do_softirq() calls softirq_handle_begin() which then sets it back to SOFTIRQ_DISABLE_OFFSET...

> +	__do_softirq();
> +
> +out:
> +	__local_bh_enable(cnt, preempt_on);

You escape from there with a correct preempt_count() but still the softirq executes
under SOFTIRQ_DISABLE_OFFSET and not SOFTIRQ_OFFSET, making in_serving_softirq() false.

> +	local_irq_restore(flags);
> +}
> +EXPORT_SYMBOL(__local_bh_enable_ip);

Thanks.