Re: [PATCH v2 2/5] sched/numa: Replace runnable_load_avg by load_avg

[Mel Gorman <mgorman@suse.de>]

On Tue, Feb 18, 2020 at 01:37:45PM +0100, Dietmar Eggemann wrote:
> On 14/02/2020 16:27, Vincent Guittot wrote:
> 
> [...]
> 
> >  	/*
> >  	 * The load is corrected for the CPU capacity available on each node.
> >  	 *
> > @@ -1788,10 +1831,10 @@ static int task_numa_migrate(struct task_struct *p)
> >  	dist = env.dist = node_distance(env.src_nid, env.dst_nid);
> >  	taskweight = task_weight(p, env.src_nid, dist);
> >  	groupweight = group_weight(p, env.src_nid, dist);
> > -	update_numa_stats(&env.src_stats, env.src_nid);
> > +	update_numa_stats(&env, &env.src_stats, env.src_nid);
> 
> This looks strange. Can you do:
> 
> -static void update_numa_stats(struct task_numa_env *env,
> +static void update_numa_stats(unsigned int imbalance_pct,
>                               struct numa_stats *ns, int nid)
> 
> -    update_numa_stats(&env, &env.src_stats, env.src_nid);
> +    update_numa_stats(env.imbalance_pct, &env.src_stats, env.src_nid);
> 

You'd also have to pass in env->p and while it could be done, I do not
think its worthwhile.

> [...]
> 
> > +static unsigned long cpu_runnable_load(struct rq *rq)
> > +{
> > +	return cfs_rq_runnable_load_avg(&rq->cfs);
> > +}
> > +
> 
> Why not remove cpu_runnable_load() in this patch rather moving it?
> 
> kernel/sched/fair.c:5492:22: warning: ???cpu_runnable_load??? defined but
> not used [-Wunused-function]
>  static unsigned long cpu_runnable_load(struct rq *rq)
> 

I took the liberty of addressing that when I picked up Vincent's patches
for "Reconcile NUMA balancing decisions with the load balancer v3" to fix
a build warning. I did not highlight it when I posted because it was such
a trivial change.

-- 
Mel Gorman
SUSE Labs