Re: [PATCH 2.6.16-mm2 4/4] sched_domain: Allocate sched_group structures dynamically

["Siddha, Suresh B" <suresh.b.siddha@intel.com>]

On Sun, Apr 02, 2006 at 11:39:30AM +1000, Nick Piggin wrote:
> 
> Srivatsa Vaddagiri wrote:
> >  /*
> > @@ -6113,6 +6125,10 @@ next_sg:
> >  static int build_sched_domains(const cpumask_t *cpu_map)
> >  {
> >  	int i;
> > +	struct sched_group *sched_group_phys = NULL;
> > +#ifdef CONFIG_SCHED_MC
> > +	struct sched_group *sched_group_core = NULL;
> > +#endif
> >  #ifdef CONFIG_NUMA
> >  	struct sched_group **sched_group_nodes = NULL;
> >  	struct sched_group *sched_group_allnodes = NULL;
> > @@ -6171,6 +6187,18 @@ static int build_sched_domains(const cpu
> >  		cpus_and(sd->span, sd->span, *cpu_map);
> >  #endif
> >  
> > +		if (!sched_group_phys) {
> > +			sched_group_phys
> > +				= kmalloc(sizeof(struct sched_group) * NR_CPUS,
> > +					  GFP_KERNEL);
> > +			if (!sched_group_phys) {
> > +				printk (KERN_WARNING "Can not alloc phys sched"
> > +						     "group\n");
> > +				goto error;
> > +			}
> > +			sched_group_phys_bycpu[i] = sched_group_phys;
> > +		}
> 
> Doesn't the last assignment have to be outside the if statement?
> 
> Hmm.. this design seems like the best way to go for now. Suresh?

Only thing I see in this is, even if there are very few cpus in the
exclusive cpuset, we end up allocating NR_CPUS groups and waste memory.

thanks,
suresh