Re: [PATCH v2 6/6] stop_machine: kill stop_cpus_lock and lg_double_lock/unlock()

[Oleg Nesterov <oleg@redhat.com>]

On 07/31, Peter Zijlstra wrote:
>
> +	for_each_cpu(cpu, cpumask)
> +		arch_spin_lock((arch_spinlock_t *)&per_cpu(cpu_stopper.lock, cpu));
> +
>  	for_each_cpu(cpu, cpumask) {
>  		work = &per_cpu(cpu_stopper.stop_work, cpu);
>  		work->fn = fn;
>  		work->arg = arg;
>  		work->done = done;
> -		cpu_stop_queue_work(cpu, work);
> +		__cpu_stop_queue_work(cpu, work);
>  	}
> -	lg_global_unlock(&stop_cpus_lock);
> +
> +	for_each_cpu(cpu, cpumask)
> +		arch_spin_unlock((arch_spinlock_t *)&per_cpu(cpu_stopper.lock, cpu));

Of course, we discussed this before and I think this should work too.
However to me this looks more ugly (although better than the current
code), and this is what I tried to avoid.

But! of course "more ugly" is very much subjective, so I won't really
argue if you prefer this change. That said, let me write another email
in reply to your initial review.

Oleg.