Re: [RFC PATCH v4 1/9] CPU hotplug: Provide APIs to prevent CPU offline from atomic context

[Oleg Nesterov <oleg@redhat.com>]

On 12/11, Srivatsa S. Bhat wrote:
>
> IOW, the hotplug readers just increment/decrement their per-cpu refcounts
> when no writer is active.

plus cli/sti ;) and increment/decrement are atomic.

At first glance looks correct to me, but I'll try to read it carefully
later.

A couple of minor nits,

> +static DEFINE_PER_CPU(bool, writer_signal);

Why it needs to be per-cpu? It can be global and __read_mostly to avoid
the false-sharing. OK, perhaps to put reader_percpu_refcnt/writer_signal
into a single cacheline...

> +void get_online_cpus_atomic(void)
> +{
> +	unsigned long flags;
> +
> +	preempt_disable();
> +
> +	if (cpu_hotplug.active_writer == current)
> +		return;
> +
> +	local_irq_save(flags);

Yes... this is still needed, we are going to increment reader_percpu_refcnt
unconditionally and this makes reader_nested_percpu() == T.

But,

> +void put_online_cpus_atomic(void)
> +{
> +	unsigned long flags;
> +
> +	if (cpu_hotplug.active_writer == current)
> +		goto out;
> +
> +	local_irq_save(flags);
> +
> +	/*
> +	 * We never allow heterogeneous nesting of readers. So it is trivial
> +	 * to find out the kind of reader we are, and undo the operation
> +	 * done by our corresponding get_online_cpus_atomic().
> +	 */
> +	if (__this_cpu_read(reader_percpu_refcnt))
> +		__this_cpu_dec(reader_percpu_refcnt);
> +	else
> +		read_unlock(&hotplug_rwlock);
> +
> +	local_irq_restore(flags);
> +out:
> +	preempt_enable();
> +}

Do we really need local_irq_save/restore in put_ ?

Oleg.