perf/core: Convert to hotplug state machine

Actually a nice symmetric startup/teardown pair which fits properly into
the state machine concept. In the long run we should be able to invoke
the startup callback for the boot CPU via the state machine and get
rid of the init function which invokes it on the boot CPU.

Note: This comes actually before the perf hardware callbacks. In the notifier
model the hardware callbacks have a higher priority than the core
callback. But that's solely for CPU offline so that hardware migration of
events happens before the core is notified about the outgoing CPU.

With the symetric state array model we have the following ordering:

 UP:     core -> hardware
 DOWN:   hardware -> core
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Signed-off-by: NAnna-Maria Gleixner <anna-maria@linutronix.de>
Reviewed-by: NSebastian Siewior <bigeasy@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160713153333.587514098@linutronix.deSigned-off-by: NIngo Molnar <mingo@kernel.org>

include/linux/cpuhotplug.h

@@ -4,6 +4,7 @@
 enum cpuhp_state {
 	CPUHP_OFFLINE,
 	CPUHP_CREATE_THREADS,
+	CPUHP_PERF_PREPARE,
 	CPUHP_NOTIFY_PREPARE,
 	CPUHP_BRINGUP_CPU,
 	CPUHP_AP_IDLE_DEAD,
@@ -22,6 +23,7 @@ enum cpuhp_state {
 	CPUHP_AP_ONLINE_IDLE,
 	CPUHP_AP_SMPBOOT_THREADS,
 	CPUHP_AP_X86_VDSO_VMA_ONLINE,
+	CPUHP_AP_PERF_ONLINE,
 	CPUHP_AP_NOTIFY_ONLINE,
 	CPUHP_AP_ONLINE_DYN,
 	CPUHP_AP_ONLINE_DYN_END		= CPUHP_AP_ONLINE_DYN + 30,

include/linux/perf_event.h

@@ -1354,4 +1354,13 @@ _name##_show(struct device *dev,					\
 									\
 static struct device_attribute format_attr_##_name = __ATTR_RO(_name)
 
+/* Performance counter hotplug functions */
+#ifdef CONFIG_PERF_EVENTS
+int perf_event_init_cpu(unsigned int cpu);
+int perf_event_exit_cpu(unsigned int cpu);
+#else
+#define perf_event_init_cpu	NULL
+#define perf_event_exit_cpu	NULL
+#endif
+
 #endif /* _LINUX_PERF_EVENT_H */

kernel/cpu.c

@@ -1180,6 +1180,11 @@ static struct cpuhp_step cpuhp_bp_states[] = {
 		.teardown		= NULL,
 		.cant_stop		= true,
 	},
+	[CPUHP_PERF_PREPARE] = {
+		.name = "perf prepare",
+		.startup = perf_event_init_cpu,
+		.teardown = perf_event_exit_cpu,
+	},
 	/*
 	 * Preparatory and dead notifiers. Will be replaced once the notifiers
 	 * are converted to states.
@@ -1257,6 +1262,12 @@ static struct cpuhp_step cpuhp_ap_states[] = {
 		.startup		= smpboot_unpark_threads,
 		.teardown		= NULL,
 	},
+	[CPUHP_AP_PERF_ONLINE] = {
+		.name = "perf online",
+		.startup = perf_event_init_cpu,
+		.teardown = perf_event_exit_cpu,
+	},
+
 	/*
 	 * Online/down_prepare notifiers. Will be removed once the notifiers
 	 * are converted to states.

kernel/events/core.c

@@ -10255,7 +10255,7 @@ static void __init perf_event_init_all_cpus(void)
 	}
 }
 
-static void perf_event_init_cpu(int cpu)
+int perf_event_init_cpu(unsigned int cpu)
 {
 	struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu);
 
@@ -10268,6 +10268,7 @@ static void perf_event_init_cpu(int cpu)
 		rcu_assign_pointer(swhash->swevent_hlist, hlist);
 	}
 	mutex_unlock(&swhash->hlist_mutex);
+	return 0;
 }
 
 #if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC_CORE
@@ -10299,14 +10300,17 @@ static void perf_event_exit_cpu_context(int cpu)
 	}
 	srcu_read_unlock(&pmus_srcu, idx);
 }
+#else
+
+static void perf_event_exit_cpu_context(int cpu) { }
+
+#endif
 
-static void perf_event_exit_cpu(int cpu)
+int perf_event_exit_cpu(unsigned int cpu)
 {
 	perf_event_exit_cpu_context(cpu);
+	return 0;
 }
-#else
-static inline void perf_event_exit_cpu(int cpu) { }
-#endif
 
 static int
 perf_reboot(struct notifier_block *notifier, unsigned long val, void *v)
@@ -10328,46 +10332,6 @@ static struct notifier_block perf_reboot_notifier = {
 	.priority = INT_MIN,
 };
 
-static int
-perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu)
-{
-	unsigned int cpu = (long)hcpu;
-
-	switch (action & ~CPU_TASKS_FROZEN) {
-
-	case CPU_UP_PREPARE:
-		/*
-		 * This must be done before the CPU comes alive, because the
-		 * moment we can run tasks we can encounter (software) events.
-		 *
-		 * Specifically, someone can have inherited events on kthreadd
-		 * or a pre-existing worker thread that gets re-bound.
-		 */
-		perf_event_init_cpu(cpu);
-		break;
-
-	case CPU_DOWN_PREPARE:
-		/*
-		 * This must be done before the CPU dies because after that an
-		 * active event might want to IPI the CPU and that'll not work
-		 * so great for dead CPUs.
-		 *
-		 * XXX smp_call_function_single() return -ENXIO without a warn
-		 * so we could possibly deal with this.
-		 *
-		 * This is safe against new events arriving because
-		 * sys_perf_event_open() serializes against hotplug using
-		 * get_online_cpus().
-		 */
-		perf_event_exit_cpu(cpu);
-		break;
-	default:
-		break;
-	}
-
-	return NOTIFY_OK;
-}
-
 void __init perf_event_init(void)
 {
 	int ret;
@@ -10380,7 +10344,7 @@ void __init perf_event_init(void)
 	perf_pmu_register(&perf_cpu_clock, NULL, -1);
 	perf_pmu_register(&perf_task_clock, NULL, -1);
 	perf_tp_register();
-	perf_cpu_notifier(perf_cpu_notify);
+	perf_event_init_cpu(smp_processor_id());
 	register_reboot_notifier(&perf_reboot_notifier);
 
 	ret = init_hw_breakpoint();