workqueue.c 20.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 * linux/kernel/workqueue.c
 *
 * Generic mechanism for defining kernel helper threads for running
 * arbitrary tasks in process context.
 *
 * Started by Ingo Molnar, Copyright (C) 2002
 *
 * Derived from the taskqueue/keventd code by:
 *
 *   David Woodhouse <dwmw2@infradead.org>
 *   Andrew Morton <andrewm@uow.edu.au>
 *   Kai Petzke <wpp@marie.physik.tu-berlin.de>
 *   Theodore Ts'o <tytso@mit.edu>
15 16
 *
 * Made to use alloc_percpu by Christoph Lameter <clameter@sgi.com>.
L
Linus Torvalds 已提交
17 18 19 20 21 22 23 24 25 26 27 28 29
 */

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/signal.h>
#include <linux/completion.h>
#include <linux/workqueue.h>
#include <linux/slab.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <linux/kthread.h>
30
#include <linux/hardirq.h>
31
#include <linux/mempolicy.h>
32
#include <linux/freezer.h>
33 34
#include <linux/kallsyms.h>
#include <linux/debug_locks.h>
L
Linus Torvalds 已提交
35 36

/*
37 38
 * The per-CPU workqueue (if single thread, we always use the first
 * possible cpu).
L
Linus Torvalds 已提交
39 40 41 42 43 44 45
 */
struct cpu_workqueue_struct {

	spinlock_t lock;

	struct list_head worklist;
	wait_queue_head_t more_work;
46
	struct work_struct *current_work;
L
Linus Torvalds 已提交
47 48

	struct workqueue_struct *wq;
49
	struct task_struct *thread;
50
	int should_stop;
L
Linus Torvalds 已提交
51 52 53 54 55 56 57 58 59

	int run_depth;		/* Detect run_workqueue() recursion depth */
} ____cacheline_aligned;

/*
 * The externally visible workqueue abstraction is an array of
 * per-CPU workqueues:
 */
struct workqueue_struct {
60
	struct cpu_workqueue_struct *cpu_wq;
61
	struct list_head list;
L
Linus Torvalds 已提交
62
	const char *name;
63
	int singlethread;
64
	int freezeable;		/* Freeze threads during suspend */
L
Linus Torvalds 已提交
65 66 67 68
};

/* All the per-cpu workqueues on the system, for hotplug cpu to add/remove
   threads to each one as cpus come/go. */
69
static DEFINE_MUTEX(workqueue_mutex);
L
Linus Torvalds 已提交
70 71
static LIST_HEAD(workqueues);

72
static int singlethread_cpu __read_mostly;
73
static cpumask_t cpu_singlethread_map __read_mostly;
74 75
/* optimization, we could use cpu_possible_map */
static cpumask_t cpu_populated_map __read_mostly;
76

L
Linus Torvalds 已提交
77 78 79
/* If it's single threaded, it isn't in the list of workqueues. */
static inline int is_single_threaded(struct workqueue_struct *wq)
{
80
	return wq->singlethread;
L
Linus Torvalds 已提交
81 82
}

83 84 85 86 87 88
static const cpumask_t *wq_cpu_map(struct workqueue_struct *wq)
{
	return is_single_threaded(wq)
		? &cpu_singlethread_map : &cpu_populated_map;
}

89 90 91 92 93 94 95 96
static
struct cpu_workqueue_struct *wq_per_cpu(struct workqueue_struct *wq, int cpu)
{
	if (unlikely(is_single_threaded(wq)))
		cpu = singlethread_cpu;
	return per_cpu_ptr(wq->cpu_wq, cpu);
}

97 98 99 100
/*
 * Set the workqueue on which a work item is to be run
 * - Must *only* be called if the pending flag is set
 */
101 102
static inline void set_wq_data(struct work_struct *work,
				struct cpu_workqueue_struct *cwq)
103
{
104 105 106
	unsigned long new;

	BUG_ON(!work_pending(work));
107

108
	new = (unsigned long) cwq | (1UL << WORK_STRUCT_PENDING);
109 110
	new |= WORK_STRUCT_FLAG_MASK & *work_data_bits(work);
	atomic_long_set(&work->data, new);
111 112
}

113 114
static inline
struct cpu_workqueue_struct *get_wq_data(struct work_struct *work)
115
{
116
	return (void *) (atomic_long_read(&work->data) & WORK_STRUCT_WQ_DATA_MASK);
117 118
}

O
Oleg Nesterov 已提交
119 120 121 122 123 124 125 126 127 128 129
static void insert_work(struct cpu_workqueue_struct *cwq,
				struct work_struct *work, int tail)
{
	set_wq_data(work, cwq);
	if (tail)
		list_add_tail(&work->entry, &cwq->worklist);
	else
		list_add(&work->entry, &cwq->worklist);
	wake_up(&cwq->more_work);
}

L
Linus Torvalds 已提交
130 131 132 133 134 135 136
/* Preempt must be disabled. */
static void __queue_work(struct cpu_workqueue_struct *cwq,
			 struct work_struct *work)
{
	unsigned long flags;

	spin_lock_irqsave(&cwq->lock, flags);
O
Oleg Nesterov 已提交
137
	insert_work(cwq, work, 1);
L
Linus Torvalds 已提交
138 139 140
	spin_unlock_irqrestore(&cwq->lock, flags);
}

141 142 143 144 145
/**
 * queue_work - queue work on a workqueue
 * @wq: workqueue to use
 * @work: work to queue
 *
A
Alan Stern 已提交
146
 * Returns 0 if @work was already on a queue, non-zero otherwise.
L
Linus Torvalds 已提交
147 148 149 150 151 152
 *
 * We queue the work to the CPU it was submitted, but there is no
 * guarantee that it will be processed by that CPU.
 */
int fastcall queue_work(struct workqueue_struct *wq, struct work_struct *work)
{
153
	int ret = 0;
L
Linus Torvalds 已提交
154

155
	if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
L
Linus Torvalds 已提交
156
		BUG_ON(!list_empty(&work->entry));
157 158
		__queue_work(wq_per_cpu(wq, get_cpu()), work);
		put_cpu();
L
Linus Torvalds 已提交
159 160 161 162
		ret = 1;
	}
	return ret;
}
163
EXPORT_SYMBOL_GPL(queue_work);
L
Linus Torvalds 已提交
164

165
void delayed_work_timer_fn(unsigned long __data)
L
Linus Torvalds 已提交
166
{
167
	struct delayed_work *dwork = (struct delayed_work *)__data;
168 169
	struct cpu_workqueue_struct *cwq = get_wq_data(&dwork->work);
	struct workqueue_struct *wq = cwq->wq;
L
Linus Torvalds 已提交
170

171
	__queue_work(wq_per_cpu(wq, smp_processor_id()), &dwork->work);
L
Linus Torvalds 已提交
172 173
}

174 175 176
/**
 * queue_delayed_work - queue work on a workqueue after delay
 * @wq: workqueue to use
177
 * @dwork: delayable work to queue
178 179
 * @delay: number of jiffies to wait before queueing
 *
A
Alan Stern 已提交
180
 * Returns 0 if @work was already on a queue, non-zero otherwise.
181
 */
L
Linus Torvalds 已提交
182
int fastcall queue_delayed_work(struct workqueue_struct *wq,
183
			struct delayed_work *dwork, unsigned long delay)
L
Linus Torvalds 已提交
184
{
185
	timer_stats_timer_set_start_info(&dwork->timer);
186
	if (delay == 0)
187
		return queue_work(wq, &dwork->work);
L
Linus Torvalds 已提交
188

189
	return queue_delayed_work_on(-1, wq, dwork, delay);
L
Linus Torvalds 已提交
190
}
191
EXPORT_SYMBOL_GPL(queue_delayed_work);
L
Linus Torvalds 已提交
192

193 194 195 196
/**
 * queue_delayed_work_on - queue work on specific CPU after delay
 * @cpu: CPU number to execute work on
 * @wq: workqueue to use
197
 * @dwork: work to queue
198 199
 * @delay: number of jiffies to wait before queueing
 *
A
Alan Stern 已提交
200
 * Returns 0 if @work was already on a queue, non-zero otherwise.
201
 */
202
int queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
203
			struct delayed_work *dwork, unsigned long delay)
204 205
{
	int ret = 0;
206 207
	struct timer_list *timer = &dwork->timer;
	struct work_struct *work = &dwork->work;
208

209
	if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
210 211 212
		BUG_ON(timer_pending(timer));
		BUG_ON(!list_empty(&work->entry));

213
		/* This stores cwq for the moment, for the timer_fn */
214
		set_wq_data(work, wq_per_cpu(wq, raw_smp_processor_id()));
215
		timer->expires = jiffies + delay;
216
		timer->data = (unsigned long)dwork;
217
		timer->function = delayed_work_timer_fn;
218 219 220 221 222

		if (unlikely(cpu >= 0))
			add_timer_on(timer, cpu);
		else
			add_timer(timer);
223 224 225 226
		ret = 1;
	}
	return ret;
}
227
EXPORT_SYMBOL_GPL(queue_delayed_work_on);
L
Linus Torvalds 已提交
228

229
static void run_workqueue(struct cpu_workqueue_struct *cwq)
L
Linus Torvalds 已提交
230
{
231
	spin_lock_irq(&cwq->lock);
L
Linus Torvalds 已提交
232 233 234 235 236 237 238 239 240 241
	cwq->run_depth++;
	if (cwq->run_depth > 3) {
		/* morton gets to eat his hat */
		printk("%s: recursion depth exceeded: %d\n",
			__FUNCTION__, cwq->run_depth);
		dump_stack();
	}
	while (!list_empty(&cwq->worklist)) {
		struct work_struct *work = list_entry(cwq->worklist.next,
						struct work_struct, entry);
242
		work_func_t f = work->func;
L
Linus Torvalds 已提交
243

O
Oleg Nesterov 已提交
244
		cwq->current_work = work;
L
Linus Torvalds 已提交
245
		list_del_init(cwq->worklist.next);
246
		spin_unlock_irq(&cwq->lock);
L
Linus Torvalds 已提交
247

248
		BUG_ON(get_wq_data(work) != cwq);
249
		if (!test_bit(WORK_STRUCT_NOAUTOREL, work_data_bits(work)))
250 251
			work_release(work);
		f(work);
L
Linus Torvalds 已提交
252

253 254 255 256 257 258 259 260 261 262 263
		if (unlikely(in_atomic() || lockdep_depth(current) > 0)) {
			printk(KERN_ERR "BUG: workqueue leaked lock or atomic: "
					"%s/0x%08x/%d\n",
					current->comm, preempt_count(),
				       	current->pid);
			printk(KERN_ERR "    last function: ");
			print_symbol("%s\n", (unsigned long)f);
			debug_show_held_locks(current);
			dump_stack();
		}

264
		spin_lock_irq(&cwq->lock);
O
Oleg Nesterov 已提交
265
		cwq->current_work = NULL;
L
Linus Torvalds 已提交
266 267
	}
	cwq->run_depth--;
268
	spin_unlock_irq(&cwq->lock);
L
Linus Torvalds 已提交
269 270
}

271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
/*
 * NOTE: the caller must not touch *cwq if this func returns true
 */
static int cwq_should_stop(struct cpu_workqueue_struct *cwq)
{
	int should_stop = cwq->should_stop;

	if (unlikely(should_stop)) {
		spin_lock_irq(&cwq->lock);
		should_stop = cwq->should_stop && list_empty(&cwq->worklist);
		if (should_stop)
			cwq->thread = NULL;
		spin_unlock_irq(&cwq->lock);
	}

	return should_stop;
}

L
Linus Torvalds 已提交
289 290 291
static int worker_thread(void *__cwq)
{
	struct cpu_workqueue_struct *cwq = __cwq;
292
	DEFINE_WAIT(wait);
L
Linus Torvalds 已提交
293 294 295
	struct k_sigaction sa;
	sigset_t blocked;

296
	if (!cwq->wq->freezeable)
297
		current->flags |= PF_NOFREEZE;
L
Linus Torvalds 已提交
298 299 300 301 302 303 304 305

	set_user_nice(current, -5);

	/* Block and flush all signals */
	sigfillset(&blocked);
	sigprocmask(SIG_BLOCK, &blocked, NULL);
	flush_signals(current);

306 307 308 309 310 311
	/*
	 * We inherited MPOL_INTERLEAVE from the booting kernel.
	 * Set MPOL_DEFAULT to insure node local allocations.
	 */
	numa_default_policy();

L
Linus Torvalds 已提交
312 313 314 315 316 317
	/* SIG_IGN makes children autoreap: see do_notify_parent(). */
	sa.sa.sa_handler = SIG_IGN;
	sa.sa.sa_flags = 0;
	siginitset(&sa.sa.sa_mask, sigmask(SIGCHLD));
	do_sigaction(SIGCHLD, &sa, (struct k_sigaction *)0);

318
	for (;;) {
319
		if (cwq->wq->freezeable)
320 321
			try_to_freeze();

322 323
		prepare_to_wait(&cwq->more_work, &wait, TASK_INTERRUPTIBLE);
		if (!cwq->should_stop && list_empty(&cwq->worklist))
L
Linus Torvalds 已提交
324
			schedule();
325 326 327 328
		finish_wait(&cwq->more_work, &wait);

		if (cwq_should_stop(cwq))
			break;
L
Linus Torvalds 已提交
329

330
		run_workqueue(cwq);
L
Linus Torvalds 已提交
331
	}
332

L
Linus Torvalds 已提交
333 334 335
	return 0;
}

O
Oleg Nesterov 已提交
336 337 338 339 340 341 342 343 344 345 346
struct wq_barrier {
	struct work_struct	work;
	struct completion	done;
};

static void wq_barrier_func(struct work_struct *work)
{
	struct wq_barrier *barr = container_of(work, struct wq_barrier, work);
	complete(&barr->done);
}

347 348
static void insert_wq_barrier(struct cpu_workqueue_struct *cwq,
					struct wq_barrier *barr, int tail)
O
Oleg Nesterov 已提交
349 350 351 352 353
{
	INIT_WORK(&barr->work, wq_barrier_func);
	__set_bit(WORK_STRUCT_PENDING, work_data_bits(&barr->work));

	init_completion(&barr->done);
354 355

	insert_work(cwq, &barr->work, tail);
O
Oleg Nesterov 已提交
356 357
}

L
Linus Torvalds 已提交
358 359 360 361 362 363 364 365 366
static void flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
{
	if (cwq->thread == current) {
		/*
		 * Probably keventd trying to flush its own queue. So simply run
		 * it by hand rather than deadlocking.
		 */
		run_workqueue(cwq);
	} else {
O
Oleg Nesterov 已提交
367
		struct wq_barrier barr;
368
		int active = 0;
L
Linus Torvalds 已提交
369

370 371 372 373 374 375
		spin_lock_irq(&cwq->lock);
		if (!list_empty(&cwq->worklist) || cwq->current_work != NULL) {
			insert_wq_barrier(cwq, &barr, 1);
			active = 1;
		}
		spin_unlock_irq(&cwq->lock);
L
Linus Torvalds 已提交
376

377
		if (active)
378
			wait_for_completion(&barr.done);
L
Linus Torvalds 已提交
379 380 381
	}
}

382
/**
L
Linus Torvalds 已提交
383
 * flush_workqueue - ensure that any scheduled work has run to completion.
384
 * @wq: workqueue to flush
L
Linus Torvalds 已提交
385 386 387 388
 *
 * Forces execution of the workqueue and blocks until its completion.
 * This is typically used in driver shutdown handlers.
 *
O
Oleg Nesterov 已提交
389 390
 * We sleep until all works which were queued on entry have been handled,
 * but we are not livelocked by new incoming ones.
L
Linus Torvalds 已提交
391 392 393 394 395 396
 *
 * This function used to run the workqueues itself.  Now we just wait for the
 * helper threads to do it.
 */
void fastcall flush_workqueue(struct workqueue_struct *wq)
{
397
	const cpumask_t *cpu_map = wq_cpu_map(wq);
398
	int cpu;
L
Linus Torvalds 已提交
399

400 401 402
	might_sleep();
	for_each_cpu_mask(cpu, *cpu_map)
		flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu));
L
Linus Torvalds 已提交
403
}
404
EXPORT_SYMBOL_GPL(flush_workqueue);
L
Linus Torvalds 已提交
405

O
Oleg Nesterov 已提交
406 407 408 409 410 411 412 413
static void wait_on_work(struct cpu_workqueue_struct *cwq,
				struct work_struct *work)
{
	struct wq_barrier barr;
	int running = 0;

	spin_lock_irq(&cwq->lock);
	if (unlikely(cwq->current_work == work)) {
414
		insert_wq_barrier(cwq, &barr, 0);
O
Oleg Nesterov 已提交
415 416 417 418
		running = 1;
	}
	spin_unlock_irq(&cwq->lock);

419
	if (unlikely(running))
O
Oleg Nesterov 已提交
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
		wait_for_completion(&barr.done);
}

/**
 * flush_work - block until a work_struct's callback has terminated
 * @wq: the workqueue on which the work is queued
 * @work: the work which is to be flushed
 *
 * flush_work() will attempt to cancel the work if it is queued.  If the work's
 * callback appears to be running, flush_work() will block until it has
 * completed.
 *
 * flush_work() is designed to be used when the caller is tearing down data
 * structures which the callback function operates upon.  It is expected that,
 * prior to calling flush_work(), the caller has arranged for the work to not
 * be requeued.
 */
void flush_work(struct workqueue_struct *wq, struct work_struct *work)
{
439
	const cpumask_t *cpu_map = wq_cpu_map(wq);
O
Oleg Nesterov 已提交
440
	struct cpu_workqueue_struct *cwq;
441
	int cpu;
O
Oleg Nesterov 已提交
442

443 444
	might_sleep();

O
Oleg Nesterov 已提交
445 446 447
	cwq = get_wq_data(work);
	/* Was it ever queued ? */
	if (!cwq)
448
		return;
O
Oleg Nesterov 已提交
449 450

	/*
451 452
	 * This work can't be re-queued, no need to re-check that
	 * get_wq_data() is still the same when we take cwq->lock.
O
Oleg Nesterov 已提交
453 454 455 456 457 458
	 */
	spin_lock_irq(&cwq->lock);
	list_del_init(&work->entry);
	work_release(work);
	spin_unlock_irq(&cwq->lock);

459 460
	for_each_cpu_mask(cpu, *cpu_map)
		wait_on_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
O
Oleg Nesterov 已提交
461 462 463
}
EXPORT_SYMBOL_GPL(flush_work);

L
Linus Torvalds 已提交
464 465 466

static struct workqueue_struct *keventd_wq;

467 468 469 470 471 472
/**
 * schedule_work - put work task in global workqueue
 * @work: job to be done
 *
 * This puts a job in the kernel-global workqueue.
 */
L
Linus Torvalds 已提交
473 474 475 476
int fastcall schedule_work(struct work_struct *work)
{
	return queue_work(keventd_wq, work);
}
477
EXPORT_SYMBOL(schedule_work);
L
Linus Torvalds 已提交
478

479 480
/**
 * schedule_delayed_work - put work task in global workqueue after delay
481 482
 * @dwork: job to be done
 * @delay: number of jiffies to wait or 0 for immediate execution
483 484 485 486
 *
 * After waiting for a given time this puts a job in the kernel-global
 * workqueue.
 */
487 488
int fastcall schedule_delayed_work(struct delayed_work *dwork,
					unsigned long delay)
L
Linus Torvalds 已提交
489
{
490
	timer_stats_timer_set_start_info(&dwork->timer);
491
	return queue_delayed_work(keventd_wq, dwork, delay);
L
Linus Torvalds 已提交
492
}
493
EXPORT_SYMBOL(schedule_delayed_work);
L
Linus Torvalds 已提交
494

495 496 497
/**
 * schedule_delayed_work_on - queue work in global workqueue on CPU after delay
 * @cpu: cpu to use
498
 * @dwork: job to be done
499 500 501 502 503
 * @delay: number of jiffies to wait
 *
 * After waiting for a given time this puts a job in the kernel-global
 * workqueue on the specified CPU.
 */
L
Linus Torvalds 已提交
504
int schedule_delayed_work_on(int cpu,
505
			struct delayed_work *dwork, unsigned long delay)
L
Linus Torvalds 已提交
506
{
507
	return queue_delayed_work_on(cpu, keventd_wq, dwork, delay);
L
Linus Torvalds 已提交
508
}
509
EXPORT_SYMBOL(schedule_delayed_work_on);
L
Linus Torvalds 已提交
510

511 512 513 514 515 516 517 518 519 520 521
/**
 * schedule_on_each_cpu - call a function on each online CPU from keventd
 * @func: the function to call
 *
 * Returns zero on success.
 * Returns -ve errno on failure.
 *
 * Appears to be racy against CPU hotplug.
 *
 * schedule_on_each_cpu() is very slow.
 */
522
int schedule_on_each_cpu(work_func_t func)
523 524
{
	int cpu;
525
	struct work_struct *works;
526

527 528
	works = alloc_percpu(struct work_struct);
	if (!works)
529
		return -ENOMEM;
530

531
	preempt_disable();		/* CPU hotplug */
532
	for_each_online_cpu(cpu) {
533 534 535 536 537
		struct work_struct *work = per_cpu_ptr(works, cpu);

		INIT_WORK(work, func);
		set_bit(WORK_STRUCT_PENDING, work_data_bits(work));
		__queue_work(per_cpu_ptr(keventd_wq->cpu_wq, cpu), work);
538
	}
539
	preempt_enable();
540
	flush_workqueue(keventd_wq);
541
	free_percpu(works);
542 543 544
	return 0;
}

L
Linus Torvalds 已提交
545 546 547 548
void flush_scheduled_work(void)
{
	flush_workqueue(keventd_wq);
}
549
EXPORT_SYMBOL(flush_scheduled_work);
L
Linus Torvalds 已提交
550

O
Oleg Nesterov 已提交
551 552 553 554 555 556
void flush_work_keventd(struct work_struct *work)
{
	flush_work(keventd_wq, work);
}
EXPORT_SYMBOL(flush_work_keventd);

L
Linus Torvalds 已提交
557
/**
558
 * cancel_rearming_delayed_work - kill off a delayed work whose handler rearms the delayed work.
559
 * @dwork: the delayed work struct
560 561 562
 *
 * Note that the work callback function may still be running on return from
 * cancel_delayed_work(). Run flush_workqueue() or flush_work() to wait on it.
L
Linus Torvalds 已提交
563
 */
564
void cancel_rearming_delayed_work(struct delayed_work *dwork)
L
Linus Torvalds 已提交
565
{
566
	struct cpu_workqueue_struct *cwq = get_wq_data(&dwork->work);
567

568 569 570
	/* Was it ever queued ? */
	if (cwq != NULL) {
		struct workqueue_struct *wq = cwq->wq;
L
Linus Torvalds 已提交
571

572 573 574
		while (!cancel_delayed_work(dwork))
			flush_workqueue(wq);
	}
L
Linus Torvalds 已提交
575 576 577
}
EXPORT_SYMBOL(cancel_rearming_delayed_work);

578 579 580 581 582 583 584 585 586 587 588 589
/**
 * execute_in_process_context - reliably execute the routine with user context
 * @fn:		the function to execute
 * @ew:		guaranteed storage for the execute work structure (must
 *		be available when the work executes)
 *
 * Executes the function immediately if process context is available,
 * otherwise schedules the function for delayed execution.
 *
 * Returns:	0 - function was executed
 *		1 - function was scheduled for execution
 */
590
int execute_in_process_context(work_func_t fn, struct execute_work *ew)
591 592
{
	if (!in_interrupt()) {
593
		fn(&ew->work);
594 595 596
		return 0;
	}

597
	INIT_WORK(&ew->work, fn);
598 599 600 601 602 603
	schedule_work(&ew->work);

	return 1;
}
EXPORT_SYMBOL_GPL(execute_in_process_context);

L
Linus Torvalds 已提交
604 605 606 607 608 609 610 611 612 613 614 615 616
int keventd_up(void)
{
	return keventd_wq != NULL;
}

int current_is_keventd(void)
{
	struct cpu_workqueue_struct *cwq;
	int cpu = smp_processor_id();	/* preempt-safe: keventd is per-cpu */
	int ret = 0;

	BUG_ON(!keventd_wq);

617
	cwq = per_cpu_ptr(keventd_wq->cpu_wq, cpu);
L
Linus Torvalds 已提交
618 619 620 621 622 623 624
	if (current == cwq->thread)
		ret = 1;

	return ret;

}

625 626
static struct cpu_workqueue_struct *
init_cpu_workqueue(struct workqueue_struct *wq, int cpu)
L
Linus Torvalds 已提交
627
{
628
	struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
L
Linus Torvalds 已提交
629

630 631 632 633 634 635
	cwq->wq = wq;
	spin_lock_init(&cwq->lock);
	INIT_LIST_HEAD(&cwq->worklist);
	init_waitqueue_head(&cwq->more_work);

	return cwq;
L
Linus Torvalds 已提交
636 637
}

638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
static int create_workqueue_thread(struct cpu_workqueue_struct *cwq, int cpu)
{
	struct workqueue_struct *wq = cwq->wq;
	const char *fmt = is_single_threaded(wq) ? "%s" : "%s/%d";
	struct task_struct *p;

	p = kthread_create(worker_thread, cwq, fmt, wq->name, cpu);
	/*
	 * Nobody can add the work_struct to this cwq,
	 *	if (caller is __create_workqueue)
	 *		nobody should see this wq
	 *	else // caller is CPU_UP_PREPARE
	 *		cpu is not on cpu_online_map
	 * so we can abort safely.
	 */
	if (IS_ERR(p))
		return PTR_ERR(p);

	cwq->thread = p;
	cwq->should_stop = 0;

	return 0;
}

662 663 664 665 666 667 668 669 670 671 672
static void start_workqueue_thread(struct cpu_workqueue_struct *cwq, int cpu)
{
	struct task_struct *p = cwq->thread;

	if (p != NULL) {
		if (cpu >= 0)
			kthread_bind(p, cpu);
		wake_up_process(p);
	}
}

673 674
struct workqueue_struct *__create_workqueue(const char *name,
					    int singlethread, int freezeable)
L
Linus Torvalds 已提交
675 676
{
	struct workqueue_struct *wq;
677 678
	struct cpu_workqueue_struct *cwq;
	int err = 0, cpu;
L
Linus Torvalds 已提交
679

680 681 682 683 684 685 686 687 688 689 690
	wq = kzalloc(sizeof(*wq), GFP_KERNEL);
	if (!wq)
		return NULL;

	wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
	if (!wq->cpu_wq) {
		kfree(wq);
		return NULL;
	}

	wq->name = name;
691
	wq->singlethread = singlethread;
692
	wq->freezeable = freezeable;
693
	INIT_LIST_HEAD(&wq->list);
694 695 696 697

	if (singlethread) {
		cwq = init_cpu_workqueue(wq, singlethread_cpu);
		err = create_workqueue_thread(cwq, singlethread_cpu);
698
		start_workqueue_thread(cwq, -1);
699
	} else {
700
		mutex_lock(&workqueue_mutex);
701 702 703 704 705 706 707
		list_add(&wq->list, &workqueues);

		for_each_possible_cpu(cpu) {
			cwq = init_cpu_workqueue(wq, cpu);
			if (err || !cpu_online(cpu))
				continue;
			err = create_workqueue_thread(cwq, cpu);
708
			start_workqueue_thread(cwq, cpu);
L
Linus Torvalds 已提交
709
		}
710 711 712 713 714 715 716 717 718 719
		mutex_unlock(&workqueue_mutex);
	}

	if (err) {
		destroy_workqueue(wq);
		wq = NULL;
	}
	return wq;
}
EXPORT_SYMBOL_GPL(__create_workqueue);
L
Linus Torvalds 已提交
720

721 722 723 724
static void cleanup_workqueue_thread(struct cpu_workqueue_struct *cwq, int cpu)
{
	struct wq_barrier barr;
	int alive = 0;
725

726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
	spin_lock_irq(&cwq->lock);
	if (cwq->thread != NULL) {
		insert_wq_barrier(cwq, &barr, 1);
		cwq->should_stop = 1;
		alive = 1;
	}
	spin_unlock_irq(&cwq->lock);

	if (alive) {
		wait_for_completion(&barr.done);

		while (unlikely(cwq->thread != NULL))
			cpu_relax();
		/*
		 * Wait until cwq->thread unlocks cwq->lock,
		 * it won't touch *cwq after that.
		 */
		smp_rmb();
		spin_unlock_wait(&cwq->lock);
	}
}

/**
 * destroy_workqueue - safely terminate a workqueue
 * @wq: target workqueue
 *
 * Safely destroy a workqueue. All work currently pending will be done first.
 */
void destroy_workqueue(struct workqueue_struct *wq)
{
756
	const cpumask_t *cpu_map = wq_cpu_map(wq);
757
	struct cpu_workqueue_struct *cwq;
758
	int cpu;
759

760 761 762
	mutex_lock(&workqueue_mutex);
	list_del(&wq->list);
	mutex_unlock(&workqueue_mutex);
763

764 765 766
	for_each_cpu_mask(cpu, *cpu_map) {
		cwq = per_cpu_ptr(wq->cpu_wq, cpu);
		cleanup_workqueue_thread(cwq, cpu);
767
	}
768

769 770 771 772 773 774 775 776 777 778 779 780 781 782 783
	free_percpu(wq->cpu_wq);
	kfree(wq);
}
EXPORT_SYMBOL_GPL(destroy_workqueue);

static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
						unsigned long action,
						void *hcpu)
{
	unsigned int cpu = (unsigned long)hcpu;
	struct cpu_workqueue_struct *cwq;
	struct workqueue_struct *wq;

	switch (action) {
	case CPU_LOCK_ACQUIRE:
784
		mutex_lock(&workqueue_mutex);
785
		return NOTIFY_OK;
786

787
	case CPU_LOCK_RELEASE:
788
		mutex_unlock(&workqueue_mutex);
789
		return NOTIFY_OK;
L
Linus Torvalds 已提交
790

791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
	case CPU_UP_PREPARE:
		cpu_set(cpu, cpu_populated_map);
	}

	list_for_each_entry(wq, &workqueues, list) {
		cwq = per_cpu_ptr(wq->cpu_wq, cpu);

		switch (action) {
		case CPU_UP_PREPARE:
			if (!create_workqueue_thread(cwq, cpu))
				break;
			printk(KERN_ERR "workqueue for %i failed\n", cpu);
			return NOTIFY_BAD;

		case CPU_ONLINE:
806
			start_workqueue_thread(cwq, cpu);
807 808 809
			break;

		case CPU_UP_CANCELED:
810
			start_workqueue_thread(cwq, -1);
811 812 813 814
		case CPU_DEAD:
			cleanup_workqueue_thread(cwq, cpu);
			break;
		}
L
Linus Torvalds 已提交
815 816 817 818 819
	}

	return NOTIFY_OK;
}

820
void __init init_workqueues(void)
L
Linus Torvalds 已提交
821
{
822
	cpu_populated_map = cpu_online_map;
823
	singlethread_cpu = first_cpu(cpu_possible_map);
824
	cpu_singlethread_map = cpumask_of_cpu(singlethread_cpu);
L
Linus Torvalds 已提交
825 826 827 828
	hotcpu_notifier(workqueue_cpu_callback, 0);
	keventd_wq = create_workqueue("events");
	BUG_ON(!keventd_wq);
}