BFS 420: fix ttwu stat

[Hillf Danton]

Due to monolithic runqueue used by BFS, the ttwu stat is bogus as we could not
tell/predict on which CPU a woken task will run after it is activated. And we
have to delay updating ttwu stat until the stage when task is taken off run
queue. To achieve that, a field, wakeup_cpu, is added in the task struct, and is
inited to be invalid CPU number for each new task.

When updating ttwu stat, now it is simply to mark the delayed operation; and
when taking task off runqueue, the stat is updated.


--- a/include/linux/sched.h	Fri May 18 20:51:06 2012
+++ b/include/linux/sched.h	Fri May 18 20:32:20 2012
@@ -1255,6 +1255,11 @@ struct task_struct {
 	struct list_head run_list;
 	u64 last_ran;
 	u64 sched_time; /* sched_clock time spent running */
+#ifdef CONFIG_SCHEDSTATS
+#ifdef CONFIG_SMP
+	int wakeup_cpu; //for ttwu stat
+#endif
+#endif
 #ifdef CONFIG_SMP
 	bool sticky; /* Soft affined flag */
 #endif
--- a/kernel/sched/bfs.c	Mon May 14 20:50:38 2012
+++ b/kernel/sched/bfs.c	Fri May 18 20:46:22 2012
@@ -1145,6 +1145,30 @@ static inline void unstick_task(struct r
  */
 static inline void take_task(int cpu, struct task_struct *p)
 {
+#ifdef CONFIG_SCHEDSTATS
+#ifdef CONFIG_SMP
+	if (p->wakeup_cpu == -1)
+		goto skip;
+
+	if (cpu == p->wakeup_cpu) {
+		schedstat_inc(cpu_rq(cpu), ttwu_local);
+	}
+	else if (cpu_online(p->wakeup_cpu)) {
+		struct sched_domain *sd;
+
+		rcu_read_lock();
+		for_each_domain(p->wakeup_cpu, sd) {
+			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
+				schedstat_inc(sd, ttwu_wake_remote);
+				break;
+			}
+		}
+		rcu_read_unlock();
+	}
+	p->wakeup_cpu = -1;
+skip:
+#endif
+#endif
 	set_task_cpu(p, cpu);
 	dequeue_task(p);
 	clear_sticky(p);
@@ -1485,30 +1509,13 @@ static void try_preempt(struct task_stru
 }
 #endif /* CONFIG_SMP */

-static void
-ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
+static void ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
 {
 #ifdef CONFIG_SCHEDSTATS
 	struct rq *rq = this_rq();

 #ifdef CONFIG_SMP
-	int this_cpu = smp_processor_id();
-
-	if (cpu == this_cpu)
-		schedstat_inc(rq, ttwu_local);
-	else {
-		struct sched_domain *sd;
-
-		rcu_read_lock();
-		for_each_domain(this_cpu, sd) {
-			if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
-				schedstat_inc(sd, ttwu_wake_remote);
-				break;
-			}
-		}
-		rcu_read_unlock();
-	}
-
+	p->wakeup_cpu = smp_processor_id();
 #endif /* CONFIG_SMP */

 	schedstat_inc(rq, ttwu_count);
@@ -1717,6 +1724,12 @@ void sched_fork(struct task_struct *p)
 #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
 	if (unlikely(sched_info_on()))
 		memset(&p->sched_info, 0, sizeof(p->sched_info));
+#endif
+
+#ifdef CONFIG_SCHEDSTATS
+#ifdef CONFIG_SMP
+	p->wakeup_cpu = -1;
+#endif
 #endif

 	p->on_cpu = false;
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/