stop_machine.c 14.6 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 * kernel/stop_machine.c
 *
 * Copyright (C) 2008, 2005	IBM Corporation.
 * Copyright (C) 2008, 2005	Rusty Russell rusty@rustcorp.com.au
 * Copyright (C) 2010		SUSE Linux Products GmbH
 * Copyright (C) 2010		Tejun Heo <tj@kernel.org>
 *
 * This file is released under the GPLv2 and any later version.
10
 */
11
#include <linux/completion.h>
Linus Torvalds's avatar
Linus Torvalds committed
12
#include <linux/cpu.h>
13
#include <linux/init.h>
14
#include <linux/kthread.h>
15
#include <linux/export.h>
16
#include <linux/percpu.h>
17 18
#include <linux/sched.h>
#include <linux/stop_machine.h>
19
#include <linux/interrupt.h>
20
#include <linux/kallsyms.h>
21
#include <linux/smpboot.h>
22
#include <linux/atomic.h>
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

/*
 * Structure to determine completion condition and record errors.  May
 * be shared by works on different cpus.
 */
struct cpu_stop_done {
	atomic_t		nr_todo;	/* nr left to execute */
	bool			executed;	/* actually executed? */
	int			ret;		/* collected return value */
	struct completion	completion;	/* fired if nr_todo reaches 0 */
};

/* the actual stopper, one per every possible cpu, enabled on online cpus */
struct cpu_stopper {
	spinlock_t		lock;
38
	bool			enabled;	/* is this stopper enabled? */
39 40 41 42
	struct list_head	works;		/* list of pending works */
};

static DEFINE_PER_CPU(struct cpu_stopper, cpu_stopper);
43
static DEFINE_PER_CPU(struct task_struct *, cpu_stopper_task);
44
static bool stop_machine_initialized = false;
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

static void cpu_stop_init_done(struct cpu_stop_done *done, unsigned int nr_todo)
{
	memset(done, 0, sizeof(*done));
	atomic_set(&done->nr_todo, nr_todo);
	init_completion(&done->completion);
}

/* signal completion unless @done is NULL */
static void cpu_stop_signal_done(struct cpu_stop_done *done, bool executed)
{
	if (done) {
		if (executed)
			done->executed = true;
		if (atomic_dec_and_test(&done->nr_todo))
			complete(&done->completion);
	}
}

/* queue @work to @stopper.  if offline, @work is completed immediately */
65
static void cpu_stop_queue_work(unsigned int cpu, struct cpu_stop_work *work)
66
{
67 68 69
	struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);
	struct task_struct *p = per_cpu(cpu_stopper_task, cpu);

70 71 72 73 74 75
	unsigned long flags;

	spin_lock_irqsave(&stopper->lock, flags);

	if (stopper->enabled) {
		list_add_tail(&work->list, &stopper->works);
76
		wake_up_process(p);
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
	} else
		cpu_stop_signal_done(work->done, false);

	spin_unlock_irqrestore(&stopper->lock, flags);
}

/**
 * stop_one_cpu - stop a cpu
 * @cpu: cpu to stop
 * @fn: function to execute
 * @arg: argument to @fn
 *
 * Execute @fn(@arg) on @cpu.  @fn is run in a process context with
 * the highest priority preempting any task on the cpu and
 * monopolizing it.  This function returns after the execution is
 * complete.
 *
 * This function doesn't guarantee @cpu stays online till @fn
 * completes.  If @cpu goes down in the middle, execution may happen
 * partially or fully on different cpus.  @fn should either be ready
 * for that or the caller should ensure that @cpu stays online until
 * this function completes.
 *
 * CONTEXT:
 * Might sleep.
 *
 * RETURNS:
 * -ENOENT if @fn(@arg) was not executed because @cpu was offline;
 * otherwise, the return value of @fn.
 */
int stop_one_cpu(unsigned int cpu, cpu_stop_fn_t fn, void *arg)
{
	struct cpu_stop_done done;
	struct cpu_stop_work work = { .fn = fn, .arg = arg, .done = &done };

	cpu_stop_init_done(&done, 1);
113
	cpu_stop_queue_work(cpu, &work);
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
	wait_for_completion(&done.completion);
	return done.executed ? done.ret : -ENOENT;
}

/**
 * stop_one_cpu_nowait - stop a cpu but don't wait for completion
 * @cpu: cpu to stop
 * @fn: function to execute
 * @arg: argument to @fn
 *
 * Similar to stop_one_cpu() but doesn't wait for completion.  The
 * caller is responsible for ensuring @work_buf is currently unused
 * and will remain untouched until stopper starts executing @fn.
 *
 * CONTEXT:
 * Don't care.
 */
void stop_one_cpu_nowait(unsigned int cpu, cpu_stop_fn_t fn, void *arg,
			struct cpu_stop_work *work_buf)
{
	*work_buf = (struct cpu_stop_work){ .fn = fn, .arg = arg, };
135
	cpu_stop_queue_work(cpu, work_buf);
136 137 138
}

/* static data for stop_cpus */
139
static DEFINE_MUTEX(stop_cpus_mutex);
140 141
static DEFINE_PER_CPU(struct cpu_stop_work, stop_cpus_work);

142 143 144
static void queue_stop_cpus_work(const struct cpumask *cpumask,
				 cpu_stop_fn_t fn, void *arg,
				 struct cpu_stop_done *done)
145 146 147 148 149 150 151 152 153
{
	struct cpu_stop_work *work;
	unsigned int cpu;

	/* initialize works and done */
	for_each_cpu(cpu, cpumask) {
		work = &per_cpu(stop_cpus_work, cpu);
		work->fn = fn;
		work->arg = arg;
154
		work->done = done;
155 156 157 158 159 160 161 162 163
	}

	/*
	 * Disable preemption while queueing to avoid getting
	 * preempted by a stopper which might wait for other stoppers
	 * to enter @fn which can lead to deadlock.
	 */
	preempt_disable();
	for_each_cpu(cpu, cpumask)
164
		cpu_stop_queue_work(cpu, &per_cpu(stop_cpus_work, cpu));
165
	preempt_enable();
166
}
167

168 169 170 171 172 173 174
static int __stop_cpus(const struct cpumask *cpumask,
		       cpu_stop_fn_t fn, void *arg)
{
	struct cpu_stop_done done;

	cpu_stop_init_done(&done, cpumask_weight(cpumask));
	queue_stop_cpus_work(cpumask, fn, arg, &done);
175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
	wait_for_completion(&done.completion);
	return done.executed ? done.ret : -ENOENT;
}

/**
 * stop_cpus - stop multiple cpus
 * @cpumask: cpus to stop
 * @fn: function to execute
 * @arg: argument to @fn
 *
 * Execute @fn(@arg) on online cpus in @cpumask.  On each target cpu,
 * @fn is run in a process context with the highest priority
 * preempting any task on the cpu and monopolizing it.  This function
 * returns after all executions are complete.
 *
 * This function doesn't guarantee the cpus in @cpumask stay online
 * till @fn completes.  If some cpus go down in the middle, execution
 * on the cpu may happen partially or fully on different cpus.  @fn
 * should either be ready for that or the caller should ensure that
 * the cpus stay online until this function completes.
 *
 * All stop_cpus() calls are serialized making it safe for @fn to wait
 * for all cpus to start executing it.
 *
 * CONTEXT:
 * Might sleep.
 *
 * RETURNS:
 * -ENOENT if @fn(@arg) was not executed at all because all cpus in
 * @cpumask were offline; otherwise, 0 if all executions of @fn
 * returned 0, any non zero return value if any returned non zero.
 */
int stop_cpus(const struct cpumask *cpumask, cpu_stop_fn_t fn, void *arg)
{
	int ret;

	/* static works are used, process one request at a time */
	mutex_lock(&stop_cpus_mutex);
	ret = __stop_cpus(cpumask, fn, arg);
	mutex_unlock(&stop_cpus_mutex);
	return ret;
}

/**
 * try_stop_cpus - try to stop multiple cpus
 * @cpumask: cpus to stop
 * @fn: function to execute
 * @arg: argument to @fn
 *
 * Identical to stop_cpus() except that it fails with -EAGAIN if
 * someone else is already using the facility.
 *
 * CONTEXT:
 * Might sleep.
 *
 * RETURNS:
 * -EAGAIN if someone else is already stopping cpus, -ENOENT if
 * @fn(@arg) was not executed at all because all cpus in @cpumask were
 * offline; otherwise, 0 if all executions of @fn returned 0, any non
 * zero return value if any returned non zero.
 */
int try_stop_cpus(const struct cpumask *cpumask, cpu_stop_fn_t fn, void *arg)
{
	int ret;

	/* static works are used, process one request at a time */
	if (!mutex_trylock(&stop_cpus_mutex))
		return -EAGAIN;
	ret = __stop_cpus(cpumask, fn, arg);
	mutex_unlock(&stop_cpus_mutex);
	return ret;
}

248 249 250 251 252 253 254 255 256 257 258 259 260
static int cpu_stop_should_run(unsigned int cpu)
{
	struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);
	unsigned long flags;
	int run;

	spin_lock_irqsave(&stopper->lock, flags);
	run = !list_empty(&stopper->works);
	spin_unlock_irqrestore(&stopper->lock, flags);
	return run;
}

static void cpu_stopper_thread(unsigned int cpu)
261
{
262
	struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);
263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
	struct cpu_stop_work *work;
	int ret;

repeat:
	work = NULL;
	spin_lock_irq(&stopper->lock);
	if (!list_empty(&stopper->works)) {
		work = list_first_entry(&stopper->works,
					struct cpu_stop_work, list);
		list_del_init(&work->list);
	}
	spin_unlock_irq(&stopper->lock);

	if (work) {
		cpu_stop_fn_t fn = work->fn;
		void *arg = work->arg;
		struct cpu_stop_done *done = work->done;
280
		char ksym_buf[KSYM_NAME_LEN] __maybe_unused;
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296

		/* cpu stop callbacks are not allowed to sleep */
		preempt_disable();

		ret = fn(arg);
		if (ret)
			done->ret = ret;

		/* restore preemption and check it's still balanced */
		preempt_enable();
		WARN_ONCE(preempt_count(),
			  "cpu_stop: %s(%p) leaked preempt count\n",
			  kallsyms_lookup((unsigned long)fn, NULL, NULL, NULL,
					  ksym_buf), arg);

		cpu_stop_signal_done(done, true);
297 298
		goto repeat;
	}
299 300
}

301 302
extern void sched_set_stop_task(int cpu, struct task_struct *stop);

303 304 305 306 307 308
static void cpu_stop_create(unsigned int cpu)
{
	sched_set_stop_task(cpu, per_cpu(cpu_stopper_task, cpu));
}

static void cpu_stop_park(unsigned int cpu)
309 310
{
	struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);
311 312
	struct cpu_stop_work *work;
	unsigned long flags;
313

314 315 316 317 318 319 320
	/* drain remaining works */
	spin_lock_irqsave(&stopper->lock, flags);
	list_for_each_entry(work, &stopper->works, list)
		cpu_stop_signal_done(work->done, false);
	stopper->enabled = false;
	spin_unlock_irqrestore(&stopper->lock, flags);
}
321

322 323 324 325 326 327 328
static void cpu_stop_unpark(unsigned int cpu)
{
	struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);

	spin_lock_irq(&stopper->lock);
	stopper->enabled = true;
	spin_unlock_irq(&stopper->lock);
329 330
}

331 332 333 334 335 336 337 338
static struct smp_hotplug_thread cpu_stop_threads = {
	.store			= &cpu_stopper_task,
	.thread_should_run	= cpu_stop_should_run,
	.thread_fn		= cpu_stopper_thread,
	.thread_comm		= "migration/%u",
	.create			= cpu_stop_create,
	.setup			= cpu_stop_unpark,
	.park			= cpu_stop_park,
339
	.pre_unpark		= cpu_stop_unpark,
340
	.selfparking		= true,
341 342 343 344 345 346 347 348 349 350 351 352 353
};

static int __init cpu_stop_init(void)
{
	unsigned int cpu;

	for_each_possible_cpu(cpu) {
		struct cpu_stopper *stopper = &per_cpu(cpu_stopper, cpu);

		spin_lock_init(&stopper->lock);
		INIT_LIST_HEAD(&stopper->works);
	}

354
	BUG_ON(smpboot_register_percpu_thread(&cpu_stop_threads));
355
	stop_machine_initialized = true;
356 357 358
	return 0;
}
early_initcall(cpu_stop_init);
Linus Torvalds's avatar
Linus Torvalds committed
359

360 361
#ifdef CONFIG_STOP_MACHINE

Rusty Russell's avatar
Rusty Russell committed
362
/* This controls the threads on each CPU. */
Linus Torvalds's avatar
Linus Torvalds committed
363
enum stopmachine_state {
Rusty Russell's avatar
Rusty Russell committed
364 365 366
	/* Dummy starting state for thread. */
	STOPMACHINE_NONE,
	/* Awaiting everyone to be scheduled. */
Linus Torvalds's avatar
Linus Torvalds committed
367
	STOPMACHINE_PREPARE,
Rusty Russell's avatar
Rusty Russell committed
368
	/* Disable interrupts. */
Linus Torvalds's avatar
Linus Torvalds committed
369
	STOPMACHINE_DISABLE_IRQ,
Rusty Russell's avatar
Rusty Russell committed
370
	/* Run the function */
371
	STOPMACHINE_RUN,
Rusty Russell's avatar
Rusty Russell committed
372
	/* Exit */
Linus Torvalds's avatar
Linus Torvalds committed
373 374 375
	STOPMACHINE_EXIT,
};

376
struct stop_machine_data {
377 378 379 380 381 382 383 384
	int			(*fn)(void *);
	void			*data;
	/* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */
	unsigned int		num_threads;
	const struct cpumask	*active_cpus;

	enum stopmachine_state	state;
	atomic_t		thread_ack;
Rusty Russell's avatar
Rusty Russell committed
385
};
386

387 388
static void set_state(struct stop_machine_data *smdata,
		      enum stopmachine_state newstate)
Linus Torvalds's avatar
Linus Torvalds committed
389
{
Rusty Russell's avatar
Rusty Russell committed
390
	/* Reset ack counter. */
391
	atomic_set(&smdata->thread_ack, smdata->num_threads);
Rusty Russell's avatar
Rusty Russell committed
392
	smp_wmb();
393
	smdata->state = newstate;
Linus Torvalds's avatar
Linus Torvalds committed
394 395
}

Rusty Russell's avatar
Rusty Russell committed
396
/* Last one to ack a state moves to the next state. */
397
static void ack_state(struct stop_machine_data *smdata)
Linus Torvalds's avatar
Linus Torvalds committed
398
{
399 400
	if (atomic_dec_and_test(&smdata->thread_ack))
		set_state(smdata, smdata->state + 1);
Linus Torvalds's avatar
Linus Torvalds committed
401 402
}

403 404
/* This is the cpu_stop function which stops the CPU. */
static int stop_machine_cpu_stop(void *data)
Linus Torvalds's avatar
Linus Torvalds committed
405
{
406
	struct stop_machine_data *smdata = data;
Rusty Russell's avatar
Rusty Russell committed
407
	enum stopmachine_state curstate = STOPMACHINE_NONE;
408
	int cpu = smp_processor_id(), err = 0;
409
	unsigned long flags;
410 411
	bool is_active;

412 413 414 415 416 417
	/*
	 * When called from stop_machine_from_inactive_cpu(), irq might
	 * already be disabled.  Save the state and restore it on exit.
	 */
	local_save_flags(flags);

418 419 420 421
	if (!smdata->active_cpus)
		is_active = cpu == cpumask_first(cpu_online_mask);
	else
		is_active = cpumask_test_cpu(cpu, smdata->active_cpus);
422

Rusty Russell's avatar
Rusty Russell committed
423 424 425
	/* Simple state machine */
	do {
		/* Chill out and ensure we re-read stopmachine_state. */
426
		cpu_relax();
427 428
		if (smdata->state != curstate) {
			curstate = smdata->state;
Rusty Russell's avatar
Rusty Russell committed
429 430 431 432 433 434
			switch (curstate) {
			case STOPMACHINE_DISABLE_IRQ:
				local_irq_disable();
				hard_irq_disable();
				break;
			case STOPMACHINE_RUN:
435 436
				if (is_active)
					err = smdata->fn(smdata->data);
Rusty Russell's avatar
Rusty Russell committed
437 438 439 440
				break;
			default:
				break;
			}
441
			ack_state(smdata);
Rusty Russell's avatar
Rusty Russell committed
442 443
		}
	} while (curstate != STOPMACHINE_EXIT);
Linus Torvalds's avatar
Linus Torvalds committed
444

445
	local_irq_restore(flags);
446
	return err;
Linus Torvalds's avatar
Linus Torvalds committed
447 448
}

449
int __stop_machine(int (*fn)(void *), void *data, const struct cpumask *cpus)
Linus Torvalds's avatar
Linus Torvalds committed
450
{
451 452 453 454
	struct stop_machine_data smdata = { .fn = fn, .data = data,
					    .num_threads = num_online_cpus(),
					    .active_cpus = cpus };

455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
	if (!stop_machine_initialized) {
		/*
		 * Handle the case where stop_machine() is called
		 * early in boot before stop_machine() has been
		 * initialized.
		 */
		unsigned long flags;
		int ret;

		WARN_ON_ONCE(smdata.num_threads != 1);

		local_irq_save(flags);
		hard_irq_disable();
		ret = (*fn)(data);
		local_irq_restore(flags);

		return ret;
	}

474 475 476
	/* Set the initial state and stop all online cpus. */
	set_state(&smdata, STOPMACHINE_PREPARE);
	return stop_cpus(cpu_online_mask, stop_machine_cpu_stop, &smdata);
Linus Torvalds's avatar
Linus Torvalds committed
477 478
}

479
int stop_machine(int (*fn)(void *), void *data, const struct cpumask *cpus)
Linus Torvalds's avatar
Linus Torvalds committed
480 481 482 483
{
	int ret;

	/* No CPUs can come up or down during this. */
484
	get_online_cpus();
485
	ret = __stop_machine(fn, data, cpus);
486
	put_online_cpus();
Linus Torvalds's avatar
Linus Torvalds committed
487 488
	return ret;
}
489
EXPORT_SYMBOL_GPL(stop_machine);
490

491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
/**
 * stop_machine_from_inactive_cpu - stop_machine() from inactive CPU
 * @fn: the function to run
 * @data: the data ptr for the @fn()
 * @cpus: the cpus to run the @fn() on (NULL = any online cpu)
 *
 * This is identical to stop_machine() but can be called from a CPU which
 * is not active.  The local CPU is in the process of hotplug (so no other
 * CPU hotplug can start) and not marked active and doesn't have enough
 * context to sleep.
 *
 * This function provides stop_machine() functionality for such state by
 * using busy-wait for synchronization and executing @fn directly for local
 * CPU.
 *
 * CONTEXT:
 * Local CPU is inactive.  Temporarily stops all active CPUs.
 *
 * RETURNS:
 * 0 if all executions of @fn returned 0, any non zero return value if any
 * returned non zero.
 */
int stop_machine_from_inactive_cpu(int (*fn)(void *), void *data,
				  const struct cpumask *cpus)
{
	struct stop_machine_data smdata = { .fn = fn, .data = data,
					    .active_cpus = cpus };
	struct cpu_stop_done done;
	int ret;

	/* Local CPU must be inactive and CPU hotplug in progress. */
	BUG_ON(cpu_active(raw_smp_processor_id()));
	smdata.num_threads = num_active_cpus() + 1;	/* +1 for local */

	/* No proper task established and can't sleep - busy wait for lock. */
	while (!mutex_trylock(&stop_cpus_mutex))
		cpu_relax();

	/* Schedule work on other CPUs and execute directly for local CPU */
	set_state(&smdata, STOPMACHINE_PREPARE);
	cpu_stop_init_done(&done, num_active_cpus());
	queue_stop_cpus_work(cpu_active_mask, stop_machine_cpu_stop, &smdata,
			     &done);
	ret = stop_machine_cpu_stop(&smdata);

	/* Busy wait for completion. */
	while (!completion_done(&done.completion))
		cpu_relax();

	mutex_unlock(&stop_cpus_mutex);
	return ret ?: done.ret;
}

544
#endif	/* CONFIG_STOP_MACHINE */