cpu.c 20.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12
/* CPU control.
 * (C) 2001, 2002, 2003, 2004 Rusty Russell
 *
 * This code is licenced under the GPL.
 */
#include <linux/proc_fs.h>
#include <linux/smp.h>
#include <linux/init.h>
#include <linux/notifier.h>
#include <linux/sched.h>
#include <linux/unistd.h>
#include <linux/cpu.h>
13 14
#include <linux/oom.h>
#include <linux/rcupdate.h>
15
#include <linux/export.h>
16
#include <linux/bug.h>
L
Linus Torvalds 已提交
17 18
#include <linux/kthread.h>
#include <linux/stop_machine.h>
19
#include <linux/mutex.h>
20
#include <linux/gfp.h>
21
#include <linux/suspend.h>
22
#include <linux/lockdep.h>
23
#include <linux/tick.h>
24
#include <linux/irq.h>
25
#include <trace/events/power.h>
L
Linus Torvalds 已提交
26

27 28
#include "smpboot.h"

29
#ifdef CONFIG_SMP
30
/* Serializes the updates to cpu_online_mask, cpu_present_mask */
31
static DEFINE_MUTEX(cpu_add_remove_lock);
L
Linus Torvalds 已提交
32

33
/*
34 35 36 37 38
 * The following two APIs (cpu_maps_update_begin/done) must be used when
 * attempting to serialize the updates to cpu_online_mask & cpu_present_mask.
 * The APIs cpu_notifier_register_begin/done() must be used to protect CPU
 * hotplug callback (un)registration performed using __register_cpu_notifier()
 * or __unregister_cpu_notifier().
39 40 41 42 43
 */
void cpu_maps_update_begin(void)
{
	mutex_lock(&cpu_add_remove_lock);
}
44
EXPORT_SYMBOL(cpu_notifier_register_begin);
45 46 47 48 49

void cpu_maps_update_done(void)
{
	mutex_unlock(&cpu_add_remove_lock);
}
50
EXPORT_SYMBOL(cpu_notifier_register_done);
51

52
static RAW_NOTIFIER_HEAD(cpu_chain);
L
Linus Torvalds 已提交
53

54 55 56 57 58
/* If set, cpu_up and cpu_down will return -EBUSY and do nothing.
 * Should always be manipulated under cpu_add_remove_lock
 */
static int cpu_hotplug_disabled;

59 60
#ifdef CONFIG_HOTPLUG_CPU

61 62
static struct {
	struct task_struct *active_writer;
63 64 65 66
	/* wait queue to wake up the active_writer */
	wait_queue_head_t wq;
	/* verifies that no writer will get active while readers are active */
	struct mutex lock;
67 68 69 70
	/*
	 * Also blocks the new readers during
	 * an ongoing cpu hotplug operation.
	 */
71
	atomic_t refcount;
72 73 74 75

#ifdef CONFIG_DEBUG_LOCK_ALLOC
	struct lockdep_map dep_map;
#endif
76 77
} cpu_hotplug = {
	.active_writer = NULL,
78
	.wq = __WAIT_QUEUE_HEAD_INITIALIZER(cpu_hotplug.wq),
79
	.lock = __MUTEX_INITIALIZER(cpu_hotplug.lock),
80 81 82
#ifdef CONFIG_DEBUG_LOCK_ALLOC
	.dep_map = {.name = "cpu_hotplug.lock" },
#endif
83
};
84

85 86
/* Lockdep annotations for get/put_online_cpus() and cpu_hotplug_begin/end() */
#define cpuhp_lock_acquire_read() lock_map_acquire_read(&cpu_hotplug.dep_map)
87 88
#define cpuhp_lock_acquire_tryread() \
				  lock_map_acquire_tryread(&cpu_hotplug.dep_map)
89 90 91
#define cpuhp_lock_acquire()      lock_map_acquire(&cpu_hotplug.dep_map)
#define cpuhp_lock_release()      lock_map_release(&cpu_hotplug.dep_map)

92

93
void get_online_cpus(void)
94
{
95 96
	might_sleep();
	if (cpu_hotplug.active_writer == current)
97
		return;
98
	cpuhp_lock_acquire_read();
99
	mutex_lock(&cpu_hotplug.lock);
100
	atomic_inc(&cpu_hotplug.refcount);
101
	mutex_unlock(&cpu_hotplug.lock);
102
}
103
EXPORT_SYMBOL_GPL(get_online_cpus);
104

105
void put_online_cpus(void)
106
{
107 108
	int refcount;

109
	if (cpu_hotplug.active_writer == current)
110
		return;
111

112 113 114 115 116 117
	refcount = atomic_dec_return(&cpu_hotplug.refcount);
	if (WARN_ON(refcount < 0)) /* try to fix things up */
		atomic_inc(&cpu_hotplug.refcount);

	if (refcount <= 0 && waitqueue_active(&cpu_hotplug.wq))
		wake_up(&cpu_hotplug.wq);
118

119
	cpuhp_lock_release();
120

121
}
122
EXPORT_SYMBOL_GPL(put_online_cpus);
123

124 125 126 127 128 129 130
/*
 * This ensures that the hotplug operation can begin only when the
 * refcount goes to zero.
 *
 * Note that during a cpu-hotplug operation, the new readers, if any,
 * will be blocked by the cpu_hotplug.lock
 *
131 132
 * Since cpu_hotplug_begin() is always called after invoking
 * cpu_maps_update_begin(), we can be sure that only one writer is active.
133 134 135 136 137 138 139 140 141 142
 *
 * Note that theoretically, there is a possibility of a livelock:
 * - Refcount goes to zero, last reader wakes up the sleeping
 *   writer.
 * - Last reader unlocks the cpu_hotplug.lock.
 * - A new reader arrives at this moment, bumps up the refcount.
 * - The writer acquires the cpu_hotplug.lock finds the refcount
 *   non zero and goes to sleep again.
 *
 * However, this is very difficult to achieve in practice since
143
 * get_online_cpus() not an api which is called all that often.
144 145
 *
 */
146
void cpu_hotplug_begin(void)
147
{
148
	DEFINE_WAIT(wait);
149

150
	cpu_hotplug.active_writer = current;
151
	cpuhp_lock_acquire();
152

153 154
	for (;;) {
		mutex_lock(&cpu_hotplug.lock);
155 156 157
		prepare_to_wait(&cpu_hotplug.wq, &wait, TASK_UNINTERRUPTIBLE);
		if (likely(!atomic_read(&cpu_hotplug.refcount)))
				break;
158 159 160
		mutex_unlock(&cpu_hotplug.lock);
		schedule();
	}
161
	finish_wait(&cpu_hotplug.wq, &wait);
162 163
}

164
void cpu_hotplug_done(void)
165 166 167
{
	cpu_hotplug.active_writer = NULL;
	mutex_unlock(&cpu_hotplug.lock);
168
	cpuhp_lock_release();
169
}
170

171 172 173 174 175 176 177 178 179 180
/*
 * Wait for currently running CPU hotplug operations to complete (if any) and
 * disable future CPU hotplug (from sysfs). The 'cpu_add_remove_lock' protects
 * the 'cpu_hotplug_disabled' flag. The same lock is also acquired by the
 * hotplug path before performing hotplug operations. So acquiring that lock
 * guarantees mutual exclusion from any currently running hotplug operations.
 */
void cpu_hotplug_disable(void)
{
	cpu_maps_update_begin();
181
	cpu_hotplug_disabled++;
182 183
	cpu_maps_update_done();
}
184
EXPORT_SYMBOL_GPL(cpu_hotplug_disable);
185 186 187 188

void cpu_hotplug_enable(void)
{
	cpu_maps_update_begin();
189
	WARN_ON(--cpu_hotplug_disabled < 0);
190 191
	cpu_maps_update_done();
}
192
EXPORT_SYMBOL_GPL(cpu_hotplug_enable);
193
#endif	/* CONFIG_HOTPLUG_CPU */
194

L
Linus Torvalds 已提交
195
/* Need to know about CPUs going up/down? */
196
int register_cpu_notifier(struct notifier_block *nb)
L
Linus Torvalds 已提交
197
{
198
	int ret;
199
	cpu_maps_update_begin();
200
	ret = raw_notifier_chain_register(&cpu_chain, nb);
201
	cpu_maps_update_done();
202
	return ret;
L
Linus Torvalds 已提交
203
}
204

205
int __register_cpu_notifier(struct notifier_block *nb)
206 207 208 209
{
	return raw_notifier_chain_register(&cpu_chain, nb);
}

210 211 212
static int __cpu_notify(unsigned long val, void *v, int nr_to_call,
			int *nr_calls)
{
213 214 215
	int ret;

	ret = __raw_notifier_call_chain(&cpu_chain, val, v, nr_to_call,
216
					nr_calls);
217 218

	return notifier_to_errno(ret);
219 220 221 222 223 224 225
}

static int cpu_notify(unsigned long val, void *v)
{
	return __cpu_notify(val, v, -1, NULL);
}

226 227
#ifdef CONFIG_HOTPLUG_CPU

228 229
static void cpu_notify_nofail(unsigned long val, void *v)
{
230
	BUG_ON(cpu_notify(val, v));
231
}
L
Linus Torvalds 已提交
232
EXPORT_SYMBOL(register_cpu_notifier);
233
EXPORT_SYMBOL(__register_cpu_notifier);
L
Linus Torvalds 已提交
234

235
void unregister_cpu_notifier(struct notifier_block *nb)
L
Linus Torvalds 已提交
236
{
237
	cpu_maps_update_begin();
238
	raw_notifier_chain_unregister(&cpu_chain, nb);
239
	cpu_maps_update_done();
L
Linus Torvalds 已提交
240 241 242
}
EXPORT_SYMBOL(unregister_cpu_notifier);

243
void __unregister_cpu_notifier(struct notifier_block *nb)
244 245 246 247 248
{
	raw_notifier_chain_unregister(&cpu_chain, nb);
}
EXPORT_SYMBOL(__unregister_cpu_notifier);

249 250 251 252 253 254 255 256 257 258 259 260
/**
 * clear_tasks_mm_cpumask - Safely clear tasks' mm_cpumask for a CPU
 * @cpu: a CPU id
 *
 * This function walks all processes, finds a valid mm struct for each one and
 * then clears a corresponding bit in mm's cpumask.  While this all sounds
 * trivial, there are various non-obvious corner cases, which this function
 * tries to solve in a safe manner.
 *
 * Also note that the function uses a somewhat relaxed locking scheme, so it may
 * be called only for an already offlined CPU.
 */
261 262 263 264 265 266 267 268 269 270 271
void clear_tasks_mm_cpumask(int cpu)
{
	struct task_struct *p;

	/*
	 * This function is called after the cpu is taken down and marked
	 * offline, so its not like new tasks will ever get this cpu set in
	 * their mm mask. -- Peter Zijlstra
	 * Thus, we may use rcu_read_lock() here, instead of grabbing
	 * full-fledged tasklist_lock.
	 */
272
	WARN_ON(cpu_online(cpu));
273 274 275 276
	rcu_read_lock();
	for_each_process(p) {
		struct task_struct *t;

277 278 279 280
		/*
		 * Main thread might exit, but other threads may still have
		 * a valid mm. Find one.
		 */
281 282 283 284 285 286 287 288 289
		t = find_lock_task_mm(p);
		if (!t)
			continue;
		cpumask_clear_cpu(cpu, mm_cpumask(t->mm));
		task_unlock(t);
	}
	rcu_read_unlock();
}

K
Kirill Tkhai 已提交
290
static inline void check_for_tasks(int dead_cpu)
L
Linus Torvalds 已提交
291
{
K
Kirill Tkhai 已提交
292
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
293

294 295
	read_lock(&tasklist_lock);
	for_each_process_thread(g, p) {
K
Kirill Tkhai 已提交
296 297 298 299 300 301 302 303 304 305 306 307 308 309
		if (!p->on_rq)
			continue;
		/*
		 * We do the check with unlocked task_rq(p)->lock.
		 * Order the reading to do not warn about a task,
		 * which was running on this cpu in the past, and
		 * it's just been woken on another cpu.
		 */
		rmb();
		if (task_cpu(p) != dead_cpu)
			continue;

		pr_warn("Task %s (pid=%d) is on cpu %d (state=%ld, flags=%x)\n",
			p->comm, task_pid_nr(p), dead_cpu, p->state, p->flags);
310 311
	}
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
312 313
}

A
Avi Kivity 已提交
314 315 316 317 318
struct take_cpu_down_param {
	unsigned long mod;
	void *hcpu;
};

L
Linus Torvalds 已提交
319
/* Take this CPU down. */
320
static int take_cpu_down(void *_param)
L
Linus Torvalds 已提交
321
{
A
Avi Kivity 已提交
322
	struct take_cpu_down_param *param = _param;
L
Linus Torvalds 已提交
323 324 325 326 327
	int err;

	/* Ensure this CPU doesn't handle any more interrupts. */
	err = __cpu_disable();
	if (err < 0)
Z
Zwane Mwaikambo 已提交
328
		return err;
L
Linus Torvalds 已提交
329

330
	cpu_notify(CPU_DYING | param->mod, param->hcpu);
331 332
	/* Give up timekeeping duties */
	tick_handover_do_timer();
333
	/* Park the stopper thread */
334
	stop_machine_park((long)param->hcpu);
Z
Zwane Mwaikambo 已提交
335
	return 0;
L
Linus Torvalds 已提交
336 337
}

338
/* Requires cpu_add_remove_lock to be held */
339
static int _cpu_down(unsigned int cpu, int tasks_frozen)
L
Linus Torvalds 已提交
340
{
341 342
	int err, nr_calls = 0;
	void *hcpu = (void *)(long)cpu;
343
	unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0;
A
Avi Kivity 已提交
344 345 346 347
	struct take_cpu_down_param tcd_param = {
		.mod = mod,
		.hcpu = hcpu,
	};
L
Linus Torvalds 已提交
348

349 350
	if (num_online_cpus() == 1)
		return -EBUSY;
L
Linus Torvalds 已提交
351

352 353
	if (!cpu_online(cpu))
		return -EINVAL;
L
Linus Torvalds 已提交
354

355
	cpu_hotplug_begin();
356

357
	err = __cpu_notify(CPU_DOWN_PREPARE | mod, hcpu, -1, &nr_calls);
358
	if (err) {
359
		nr_calls--;
360
		__cpu_notify(CPU_DOWN_FAILED | mod, hcpu, nr_calls, NULL);
361 362
		pr_warn("%s: attempt to take down CPU %u failed\n",
			__func__, cpu);
363
		goto out_release;
L
Linus Torvalds 已提交
364 365
	}

366 367 368 369 370 371
	/*
	 * By now we've cleared cpu_active_mask, wait for all preempt-disabled
	 * and RCU users of this state to go away such that all new such users
	 * will observe it.
	 *
	 * For CONFIG_PREEMPT we have preemptible RCU and its sync_rcu() might
372
	 * not imply sync_sched(), so wait for both.
373 374
	 *
	 * Do sync before park smpboot threads to take care the rcu boost case.
375
	 */
376 377 378 379
	if (IS_ENABLED(CONFIG_PREEMPT))
		synchronize_rcu_mult(call_rcu, call_rcu_sched);
	else
		synchronize_rcu();
380

381 382
	smpboot_park_threads(cpu);

383
	/*
384 385
	 * Prevent irq alloc/free while the dying cpu reorganizes the
	 * interrupt affinities.
386
	 */
387
	irq_lock_sparse();
388

389 390 391
	/*
	 * So now all preempt/rcu users must observe !cpu_active().
	 */
392
	err = stop_machine(take_cpu_down, &tcd_param, cpumask_of(cpu));
393
	if (err) {
L
Linus Torvalds 已提交
394
		/* CPU didn't die: tell everyone.  Can't complain. */
395
		cpu_notify_nofail(CPU_DOWN_FAILED | mod, hcpu);
396
		irq_unlock_sparse();
397
		goto out_release;
398
	}
399
	BUG_ON(cpu_online(cpu));
L
Linus Torvalds 已提交
400

401 402 403 404
	/*
	 * The migration_call() CPU_DYING callback will have removed all
	 * runnable tasks from the cpu, there's only the idle task left now
	 * that the migration thread is done doing the stop_machine thing.
P
Peter Zijlstra 已提交
405 406
	 *
	 * Wait for the stop thread to go away.
407
	 */
408
	while (!per_cpu(cpu_dead_idle, cpu))
P
Peter Zijlstra 已提交
409
		cpu_relax();
410 411
	smp_mb(); /* Read from cpu_dead_idle before __cpu_die(). */
	per_cpu(cpu_dead_idle, cpu) = false;
L
Linus Torvalds 已提交
412

413 414 415
	/* Interrupts are moved away from the dying cpu, reenable alloc/free */
	irq_unlock_sparse();

416
	hotplug_cpu__broadcast_tick_pull(cpu);
L
Linus Torvalds 已提交
417 418 419 420
	/* This actually kills the CPU. */
	__cpu_die(cpu);

	/* CPU is completely dead: tell everyone.  Too late to complain. */
421
	tick_cleanup_dead_cpu(cpu);
422
	cpu_notify_nofail(CPU_DEAD | mod, hcpu);
L
Linus Torvalds 已提交
423 424 425

	check_for_tasks(cpu);

426
out_release:
427
	cpu_hotplug_done();
428 429
	if (!err)
		cpu_notify_nofail(CPU_POST_DEAD | mod, hcpu);
430 431 432
	return err;
}

433
int cpu_down(unsigned int cpu)
434
{
435
	int err;
436

437
	cpu_maps_update_begin();
438 439

	if (cpu_hotplug_disabled) {
440
		err = -EBUSY;
441 442 443 444
		goto out;
	}

	err = _cpu_down(cpu, 0);
445

446
out:
447
	cpu_maps_update_done();
L
Linus Torvalds 已提交
448 449
	return err;
}
450
EXPORT_SYMBOL(cpu_down);
L
Linus Torvalds 已提交
451 452
#endif /*CONFIG_HOTPLUG_CPU*/

453 454 455 456 457 458 459 460 461 462
/*
 * Unpark per-CPU smpboot kthreads at CPU-online time.
 */
static int smpboot_thread_call(struct notifier_block *nfb,
			       unsigned long action, void *hcpu)
{
	int cpu = (long)hcpu;

	switch (action & ~CPU_TASKS_FROZEN) {

463
	case CPU_DOWN_FAILED:
464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
	case CPU_ONLINE:
		smpboot_unpark_threads(cpu);
		break;

	default:
		break;
	}

	return NOTIFY_OK;
}

static struct notifier_block smpboot_thread_notifier = {
	.notifier_call = smpboot_thread_call,
	.priority = CPU_PRI_SMPBOOT,
};

480
void smpboot_thread_init(void)
481 482 483 484
{
	register_cpu_notifier(&smpboot_thread_notifier);
}

485
/* Requires cpu_add_remove_lock to be held */
486
static int _cpu_up(unsigned int cpu, int tasks_frozen)
L
Linus Torvalds 已提交
487
{
488
	int ret, nr_calls = 0;
L
Linus Torvalds 已提交
489
	void *hcpu = (void *)(long)cpu;
490
	unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0;
491
	struct task_struct *idle;
L
Linus Torvalds 已提交
492

493
	cpu_hotplug_begin();
494

495 496 497 498 499
	if (cpu_online(cpu) || !cpu_present(cpu)) {
		ret = -EINVAL;
		goto out;
	}

500 501 502
	idle = idle_thread_get(cpu);
	if (IS_ERR(idle)) {
		ret = PTR_ERR(idle);
503
		goto out;
504
	}
505

506 507 508 509
	ret = smpboot_create_threads(cpu);
	if (ret)
		goto out;

510
	ret = __cpu_notify(CPU_UP_PREPARE | mod, hcpu, -1, &nr_calls);
511
	if (ret) {
512
		nr_calls--;
513 514
		pr_warn("%s: attempt to bring up CPU %u failed\n",
			__func__, cpu);
L
Linus Torvalds 已提交
515 516 517 518
		goto out_notify;
	}

	/* Arch-specific enabling code. */
519
	ret = __cpu_up(cpu, idle);
520

L
Linus Torvalds 已提交
521 522
	if (ret != 0)
		goto out_notify;
523
	BUG_ON(!cpu_online(cpu));
L
Linus Torvalds 已提交
524 525

	/* Now call notifier in preparation. */
526
	cpu_notify(CPU_ONLINE | mod, hcpu);
L
Linus Torvalds 已提交
527 528 529

out_notify:
	if (ret != 0)
530
		__cpu_notify(CPU_UP_CANCELED | mod, hcpu, nr_calls, NULL);
531
out:
532
	cpu_hotplug_done();
533 534 535 536

	return ret;
}

537
int cpu_up(unsigned int cpu)
538 539
{
	int err = 0;
540

R
Rusty Russell 已提交
541
	if (!cpu_possible(cpu)) {
542 543
		pr_err("can't online cpu %d because it is not configured as may-hotadd at boot time\n",
		       cpu);
544
#if defined(CONFIG_IA64)
545
		pr_err("please check additional_cpus= boot parameter\n");
546 547 548
#endif
		return -EINVAL;
	}
549

550 551 552
	err = try_online_node(cpu_to_node(cpu));
	if (err)
		return err;
553

554
	cpu_maps_update_begin();
555 556

	if (cpu_hotplug_disabled) {
557
		err = -EBUSY;
558 559 560 561 562 563
		goto out;
	}

	err = _cpu_up(cpu, 0);

out:
564
	cpu_maps_update_done();
565 566
	return err;
}
P
Paul E. McKenney 已提交
567
EXPORT_SYMBOL_GPL(cpu_up);
568

569
#ifdef CONFIG_PM_SLEEP_SMP
R
Rusty Russell 已提交
570
static cpumask_var_t frozen_cpus;
571 572 573

int disable_nonboot_cpus(void)
{
574
	int cpu, first_cpu, error = 0;
575

576
	cpu_maps_update_begin();
R
Rusty Russell 已提交
577
	first_cpu = cpumask_first(cpu_online_mask);
578 579
	/*
	 * We take down all of the non-boot CPUs in one shot to avoid races
580 581
	 * with the userspace trying to use the CPU hotplug at the same time
	 */
R
Rusty Russell 已提交
582
	cpumask_clear(frozen_cpus);
583

584
	pr_info("Disabling non-boot CPUs ...\n");
585 586 587
	for_each_online_cpu(cpu) {
		if (cpu == first_cpu)
			continue;
588
		trace_suspend_resume(TPS("CPU_OFF"), cpu, true);
589
		error = _cpu_down(cpu, 1);
590
		trace_suspend_resume(TPS("CPU_OFF"), cpu, false);
591
		if (!error)
R
Rusty Russell 已提交
592
			cpumask_set_cpu(cpu, frozen_cpus);
593
		else {
594
			pr_err("Error taking CPU%d down: %d\n", cpu, error);
595 596 597
			break;
		}
	}
598

599
	if (!error)
600
		BUG_ON(num_online_cpus() > 1);
601
	else
602
		pr_err("Non-boot CPUs are not disabled\n");
603 604 605 606 607 608 609 610

	/*
	 * Make sure the CPUs won't be enabled by someone else. We need to do
	 * this even in case of failure as all disable_nonboot_cpus() users are
	 * supposed to do enable_nonboot_cpus() on the failure path.
	 */
	cpu_hotplug_disabled++;

611
	cpu_maps_update_done();
612 613 614
	return error;
}

615 616 617 618 619 620 621 622
void __weak arch_enable_nonboot_cpus_begin(void)
{
}

void __weak arch_enable_nonboot_cpus_end(void)
{
}

623
void enable_nonboot_cpus(void)
624 625 626 627
{
	int cpu, error;

	/* Allow everyone to use the CPU hotplug again */
628
	cpu_maps_update_begin();
629
	WARN_ON(--cpu_hotplug_disabled < 0);
R
Rusty Russell 已提交
630
	if (cpumask_empty(frozen_cpus))
631
		goto out;
632

633
	pr_info("Enabling non-boot CPUs ...\n");
634 635 636

	arch_enable_nonboot_cpus_begin();

R
Rusty Russell 已提交
637
	for_each_cpu(cpu, frozen_cpus) {
638
		trace_suspend_resume(TPS("CPU_ON"), cpu, true);
639
		error = _cpu_up(cpu, 1);
640
		trace_suspend_resume(TPS("CPU_ON"), cpu, false);
641
		if (!error) {
642
			pr_info("CPU%d is up\n", cpu);
643 644
			continue;
		}
645
		pr_warn("Error taking CPU%d up: %d\n", cpu, error);
646
	}
647 648 649

	arch_enable_nonboot_cpus_end();

R
Rusty Russell 已提交
650
	cpumask_clear(frozen_cpus);
651
out:
652
	cpu_maps_update_done();
L
Linus Torvalds 已提交
653
}
R
Rusty Russell 已提交
654

655
static int __init alloc_frozen_cpus(void)
R
Rusty Russell 已提交
656 657 658 659 660 661
{
	if (!alloc_cpumask_var(&frozen_cpus, GFP_KERNEL|__GFP_ZERO))
		return -ENOMEM;
	return 0;
}
core_initcall(alloc_frozen_cpus);
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681

/*
 * When callbacks for CPU hotplug notifications are being executed, we must
 * ensure that the state of the system with respect to the tasks being frozen
 * or not, as reported by the notification, remains unchanged *throughout the
 * duration* of the execution of the callbacks.
 * Hence we need to prevent the freezer from racing with regular CPU hotplug.
 *
 * This synchronization is implemented by mutually excluding regular CPU
 * hotplug and Suspend/Hibernate call paths by hooking onto the Suspend/
 * Hibernate notifications.
 */
static int
cpu_hotplug_pm_callback(struct notifier_block *nb,
			unsigned long action, void *ptr)
{
	switch (action) {

	case PM_SUSPEND_PREPARE:
	case PM_HIBERNATION_PREPARE:
682
		cpu_hotplug_disable();
683 684 685 686
		break;

	case PM_POST_SUSPEND:
	case PM_POST_HIBERNATION:
687
		cpu_hotplug_enable();
688 689 690 691 692 693 694 695 696 697
		break;

	default:
		return NOTIFY_DONE;
	}

	return NOTIFY_OK;
}


698
static int __init cpu_hotplug_pm_sync_init(void)
699
{
700 701 702 703 704
	/*
	 * cpu_hotplug_pm_callback has higher priority than x86
	 * bsp_pm_callback which depends on cpu_hotplug_pm_callback
	 * to disable cpu hotplug to avoid cpu hotplug race.
	 */
705 706 707 708 709
	pm_notifier(cpu_hotplug_pm_callback, 0);
	return 0;
}
core_initcall(cpu_hotplug_pm_sync_init);

710
#endif /* CONFIG_PM_SLEEP_SMP */
711

712 713 714 715 716 717 718 719
/**
 * notify_cpu_starting(cpu) - call the CPU_STARTING notifiers
 * @cpu: cpu that just started
 *
 * This function calls the cpu_chain notifiers with CPU_STARTING.
 * It must be called by the arch code on the new cpu, before the new cpu
 * enables interrupts and before the "boot" cpu returns from __cpu_up().
 */
720
void notify_cpu_starting(unsigned int cpu)
721 722 723 724
{
	unsigned long val = CPU_STARTING;

#ifdef CONFIG_PM_SLEEP_SMP
R
Rusty Russell 已提交
725
	if (frozen_cpus != NULL && cpumask_test_cpu(cpu, frozen_cpus))
726 727
		val = CPU_STARTING_FROZEN;
#endif /* CONFIG_PM_SLEEP_SMP */
728
	cpu_notify(val, (void *)(long)cpu);
729 730
}

731
#endif /* CONFIG_SMP */
732

733 734 735 736
/*
 * cpu_bit_bitmap[] is a special, "compressed" data structure that
 * represents all NR_CPUS bits binary values of 1<<nr.
 *
R
Rusty Russell 已提交
737
 * It is used by cpumask_of() to get a constant address to a CPU
738 739
 * mask value that has a single bit set only.
 */
740

741
/* cpu_bit_bitmap[0] is empty - so we can back into it */
742
#define MASK_DECLARE_1(x)	[x+1][0] = (1UL << (x))
743 744 745
#define MASK_DECLARE_2(x)	MASK_DECLARE_1(x), MASK_DECLARE_1(x+1)
#define MASK_DECLARE_4(x)	MASK_DECLARE_2(x), MASK_DECLARE_2(x+2)
#define MASK_DECLARE_8(x)	MASK_DECLARE_4(x), MASK_DECLARE_4(x+4)
746

747 748 749 750 751 752 753
const unsigned long cpu_bit_bitmap[BITS_PER_LONG+1][BITS_TO_LONGS(NR_CPUS)] = {

	MASK_DECLARE_8(0),	MASK_DECLARE_8(8),
	MASK_DECLARE_8(16),	MASK_DECLARE_8(24),
#if BITS_PER_LONG > 32
	MASK_DECLARE_8(32),	MASK_DECLARE_8(40),
	MASK_DECLARE_8(48),	MASK_DECLARE_8(56),
754 755
#endif
};
756
EXPORT_SYMBOL_GPL(cpu_bit_bitmap);
757 758 759

const DECLARE_BITMAP(cpu_all_bits, NR_CPUS) = CPU_BITS_ALL;
EXPORT_SYMBOL(cpu_all_bits);
760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780

#ifdef CONFIG_INIT_ALL_POSSIBLE
static DECLARE_BITMAP(cpu_possible_bits, CONFIG_NR_CPUS) __read_mostly
	= CPU_BITS_ALL;
#else
static DECLARE_BITMAP(cpu_possible_bits, CONFIG_NR_CPUS) __read_mostly;
#endif
const struct cpumask *const cpu_possible_mask = to_cpumask(cpu_possible_bits);
EXPORT_SYMBOL(cpu_possible_mask);

static DECLARE_BITMAP(cpu_online_bits, CONFIG_NR_CPUS) __read_mostly;
const struct cpumask *const cpu_online_mask = to_cpumask(cpu_online_bits);
EXPORT_SYMBOL(cpu_online_mask);

static DECLARE_BITMAP(cpu_present_bits, CONFIG_NR_CPUS) __read_mostly;
const struct cpumask *const cpu_present_mask = to_cpumask(cpu_present_bits);
EXPORT_SYMBOL(cpu_present_mask);

static DECLARE_BITMAP(cpu_active_bits, CONFIG_NR_CPUS) __read_mostly;
const struct cpumask *const cpu_active_mask = to_cpumask(cpu_active_bits);
EXPORT_SYMBOL(cpu_active_mask);
781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799

void set_cpu_possible(unsigned int cpu, bool possible)
{
	if (possible)
		cpumask_set_cpu(cpu, to_cpumask(cpu_possible_bits));
	else
		cpumask_clear_cpu(cpu, to_cpumask(cpu_possible_bits));
}

void set_cpu_present(unsigned int cpu, bool present)
{
	if (present)
		cpumask_set_cpu(cpu, to_cpumask(cpu_present_bits));
	else
		cpumask_clear_cpu(cpu, to_cpumask(cpu_present_bits));
}

void set_cpu_online(unsigned int cpu, bool online)
{
800
	if (online) {
801
		cpumask_set_cpu(cpu, to_cpumask(cpu_online_bits));
802 803
		cpumask_set_cpu(cpu, to_cpumask(cpu_active_bits));
	} else {
804
		cpumask_clear_cpu(cpu, to_cpumask(cpu_online_bits));
805
	}
806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
}

void set_cpu_active(unsigned int cpu, bool active)
{
	if (active)
		cpumask_set_cpu(cpu, to_cpumask(cpu_active_bits));
	else
		cpumask_clear_cpu(cpu, to_cpumask(cpu_active_bits));
}

void init_cpu_present(const struct cpumask *src)
{
	cpumask_copy(to_cpumask(cpu_present_bits), src);
}

void init_cpu_possible(const struct cpumask *src)
{
	cpumask_copy(to_cpumask(cpu_possible_bits), src);
}

void init_cpu_online(const struct cpumask *src)
{
	cpumask_copy(to_cpumask(cpu_online_bits), src);
}