x86/percpu, sched/fair: Avoid local_clock()

Nadav reported that code-gen changed because of the this_cpu_*()
constraints, avoid this for select_idle_cpu() because that runs with
preemption (and IRQs) disabled anyway.
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

kernel/sched/fair.c

@@ -6189,6 +6189,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
 	u64 time, cost;
 	s64 delta;
 	int cpu, nr = INT_MAX;
+	int this = smp_processor_id();
 
 	this_sd = rcu_dereference(*this_cpu_ptr(&sd_llc));
 	if (!this_sd)
@@ -6212,7 +6213,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
 			nr = 4;
 	}
 
-	time = local_clock();
+	time = cpu_clock(this);
 
 	for_each_cpu_wrap(cpu, sched_domain_span(sd), target) {
 		if (!--nr)
@@ -6223,7 +6224,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, int t
 			break;
 	}
 
-	time = local_clock() - time;
+	time = cpu_clock(this) - time;
 	cost = this_sd->avg_scan_cost;
 	delta = (s64)(time - cost) / 8;
 	this_sd->avg_scan_cost += delta;