Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1355071

[patch 11/15] sched/migration: Move CPU_ONLINE into scheduler state

Show all headers | View raw

The alleged requirement that the migration notifier has a lower priority than
perf is completely undocumented and there is no indication at all that this is
true. perf does not even handle the CPU_ONLINE notification and perf really
has nothing to do with migration.

Move the CPU_ONLINE code into the sched_activate_cpu() state callback.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 kernel/sched/core.c |   33 ++++++++++++++++++++++-----------
 1 file changed, 22 insertions(+), 11 deletions(-)

--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5662,17 +5662,6 @@ migration_call(struct notifier_block *nf
 
 	switch (action & ~CPU_TASKS_FROZEN) {
 
-	case CPU_ONLINE:
-		/* Update our root-domain */
-		raw_spin_lock_irqsave(&rq->lock, flags);
-		if (rq->rd) {
-			BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
-
-			set_rq_online(rq);
-		}
-		raw_spin_unlock_irqrestore(&rq->lock, flags);
-		break;
-
 #ifdef CONFIG_HOTPLUG_CPU
 	case CPU_DYING:
 		sched_ttwu_pending();
@@ -7323,12 +7312,34 @@ static int cpuset_cpu_inactive(unsigned
 
 int sched_cpu_activate(unsigned int cpu)
 {
+	struct rq *rq = cpu_rq(cpu);
+	unsigned long flags;
+
 	set_cpu_active(cpu, true);
 
 	if (sched_smp_initialized) {
 		sched_domains_numa_masks_set(cpu);
 		cpuset_cpu_active();
 	}
+
+	/*
+	 * Put the rq online, if not already. This happens:
+	 *
+	 * 1) In the early boot process, because we build the real domains
+	 *    after all cpus have been brought up.
+	 *
+	 * 2) At runtime, if cpuset_cpu_active() fails to rebuild the
+	 *    domains.
+	 */
+	raw_spin_lock_irqsave(&rq->lock, flags);
+	if (rq->rd) {
+		BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
+		set_rq_online(rq);
+	}
+	raw_spin_unlock_irqrestore(&rq->lock, flags);
+
+	update_max_interval();
+
 	return 0;
 }
 

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

[patch 00/15] cpu/hotplug: Convert scheduler to hotplug state machine Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 14/15] sched/fair: Make ilb_notifier an explicit call Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 07/15] sched/hotplug: Convert cpu_[in]active notifiers to  state machine Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 01/15] cpu/hotplug: Document states better Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 10/15] sched/migration: Move calc_load_migrate() into CPU_DYING Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 13/15] sched/hotplug: Make activate() the last hotplug step Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 05/15] sched: Consolidate the notifier maze Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 09/15] sched/migration: Move prepare transition to  SCHED_STARTING state Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 08/15] sched,  hotplug: Move sync_rcu to be with set_cpu_active(false) Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 06/15] sched: Move sched_domains_numa_masks_clear() to  DOWN_PREPARE Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 11/15] sched/migration: Move CPU_ONLINE into scheduler state Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 03/15] sched: Make set_cpu_rq_start_time() a built in hotplug  state Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 12/15] sched/hotplug: Move migration CPU_DYING to  sched_cpu_dying() Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 15/15] sched: Make hrtick_notifier an explicit call Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:10 +0100
  [patch 04/15] sched: Allow hotplug notifiers to be setup early Thomas Gleixner <tglx@linutronix.de> - 2016-03-10 13:20 +0100

csiph-web