[PATCH v4] ARM: vexpress: add support for multiple core tiles

[will.deacon@arm.com (Will Deacon)]

Hi Russell,

> > -	/* sanity check */
> > -	if (ncores > NR_CPUS) {
> > +	ct_desc->init_cpu_map();
> > +	if (!num_possible_cpus()) {
> >  		printk(KERN_WARNING
> > -		       "vexpress: no. of cores (%d) greater than configured "
> > -		       "maximum of %d - clipping\n",
> > -		       ncores, NR_CPUS);
> > -		ncores = NR_CPUS;
> > +		       "vexpress: no possible CPUs declared, forcing logical "
> > +		       "CPU 0\n");
> > +		set_cpu_possible(0, true);
> 
> Hmm. The right check is whether the current CPU number is set in the
> possible map - if it's clear, then that's certainly a bug.
> 
> However, this will never happen, unless you do something really silly
> in ->init_cpu_map() such as clearing all possible CPUs, or setting
> them false.  See boot_cpu_init() in init/main.c.
> 
> So I don't think we need this check (and I think it should be in the
> platform code.)

Ok, my initial thought was to lose the check as well. This means
that all we do in smp_init_cpus for vexpress is invoke
ct_desc->init_cpu_map(). 
 
> Alteratively, we could do this in the generic SMP code:
> 
> 	if (num_possible_cpus() > NR_CPUS) {
> 		unsigned int nr = 0, cpu;
> 
> 		pr_warning("SMP: number of CPUs greater than configured %u, disabling CPU",
> 			NR_CPUS);
> 
> 		for_each_possible_cpu(cpu) {
> 			if (nr >= NR_CPUS) {
> 				cpu_set_possible(cpu, false);
> 				pr_cont(" %u", cpu);
> 			}
> 			nr++;
> 		}
> 		pr_cont("\n");
> 	}

Can num_possible_cpus() be > NR_CPUS? I think the bitmap used to calculate
the former is limited by the latter anyway.

Thanks,

Will