workqueue.c 20.4 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 46 47
 */
struct cpu_workqueue_struct {

	spinlock_t lock;

	struct list_head worklist;
	wait_queue_head_t more_work;

	struct workqueue_struct *wq;
48
	struct task_struct *thread;
L
Linus Torvalds 已提交
49 50

	int run_depth;		/* Detect run_workqueue() recursion depth */
51 52

	int freezeable;		/* Freeze the thread during suspend */
L
Linus Torvalds 已提交
53 54 55 56 57 58 59
} ____cacheline_aligned;

/*
 * The externally visible workqueue abstraction is an array of
 * per-CPU workqueues:
 */
struct workqueue_struct {
60
	struct cpu_workqueue_struct *cpu_wq;
L
Linus Torvalds 已提交
61 62 63 64 65 66
	const char *name;
	struct list_head list; 	/* Empty if single thread */
};

/* All the per-cpu workqueues on the system, for hotplug cpu to add/remove
   threads to each one as cpus come/go. */
67
static DEFINE_MUTEX(workqueue_mutex);
L
Linus Torvalds 已提交
68 69
static LIST_HEAD(workqueues);

70 71
static int singlethread_cpu;

L
Linus Torvalds 已提交
72 73 74 75 76 77
/* If it's single threaded, it isn't in the list of workqueues. */
static inline int is_single_threaded(struct workqueue_struct *wq)
{
	return list_empty(&wq->list);
}

78 79 80 81
/*
 * Set the workqueue on which a work item is to be run
 * - Must *only* be called if the pending flag is set
 */
82 83
static inline void set_wq_data(struct work_struct *work, void *wq)
{
84 85 86
	unsigned long new;

	BUG_ON(!work_pending(work));
87 88

	new = (unsigned long) wq | (1UL << WORK_STRUCT_PENDING);
89 90
	new |= WORK_STRUCT_FLAG_MASK & *work_data_bits(work);
	atomic_long_set(&work->data, new);
91 92 93 94
}

static inline void *get_wq_data(struct work_struct *work)
{
95
	return (void *) (atomic_long_read(&work->data) & WORK_STRUCT_WQ_DATA_MASK);
96 97
}

98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
static int __run_work(struct cpu_workqueue_struct *cwq, struct work_struct *work)
{
	int ret = 0;
	unsigned long flags;

	spin_lock_irqsave(&cwq->lock, flags);
	/*
	 * We need to re-validate the work info after we've gotten
	 * the cpu_workqueue lock. We can run the work now iff:
	 *
	 *  - the wq_data still matches the cpu_workqueue_struct
	 *  - AND the work is still marked pending
	 *  - AND the work is still on a list (which will be this
	 *    workqueue_struct list)
	 *
	 * All these conditions are important, because we
	 * need to protect against the work being run right
	 * now on another CPU (all but the last one might be
	 * true if it's currently running and has not been
	 * released yet, for example).
	 */
	if (get_wq_data(work) == cwq
	    && work_pending(work)
	    && !list_empty(&work->entry)) {
		work_func_t f = work->func;
		list_del_init(&work->entry);
		spin_unlock_irqrestore(&cwq->lock, flags);

126
		if (!test_bit(WORK_STRUCT_NOAUTOREL, work_data_bits(work)))
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
			work_release(work);
		f(work);

		spin_lock_irqsave(&cwq->lock, flags);
		ret = 1;
	}
	spin_unlock_irqrestore(&cwq->lock, flags);
	return ret;
}

/**
 * run_scheduled_work - run scheduled work synchronously
 * @work: work to run
 *
 * This checks if the work was pending, and runs it
 * synchronously if so. It returns a boolean to indicate
 * whether it had any scheduled work to run or not.
 *
 * NOTE! This _only_ works for normal work_structs. You
 * CANNOT use this for delayed work, because the wq data
 * for delayed work will not point properly to the per-
 * CPU workqueue struct, but will change!
 */
int fastcall run_scheduled_work(struct work_struct *work)
{
	for (;;) {
		struct cpu_workqueue_struct *cwq;

		if (!work_pending(work))
			return 0;
		if (list_empty(&work->entry))
			return 0;
		/* NOTE! This depends intimately on __queue_work! */
		cwq = get_wq_data(work);
		if (!cwq)
			return 0;
		if (__run_work(cwq, work))
			return 1;
	}
}
EXPORT_SYMBOL(run_scheduled_work);

L
Linus Torvalds 已提交
169 170 171 172 173 174 175
/* 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);
176
	set_wq_data(work, cwq);
L
Linus Torvalds 已提交
177 178 179 180 181
	list_add_tail(&work->entry, &cwq->worklist);
	wake_up(&cwq->more_work);
	spin_unlock_irqrestore(&cwq->lock, flags);
}

182 183 184 185 186
/**
 * queue_work - queue work on a workqueue
 * @wq: workqueue to use
 * @work: work to queue
 *
A
Alan Stern 已提交
187
 * Returns 0 if @work was already on a queue, non-zero otherwise.
L
Linus Torvalds 已提交
188 189 190 191 192 193 194 195
 *
 * 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)
{
	int ret = 0, cpu = get_cpu();

196
	if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
L
Linus Torvalds 已提交
197
		if (unlikely(is_single_threaded(wq)))
198
			cpu = singlethread_cpu;
L
Linus Torvalds 已提交
199
		BUG_ON(!list_empty(&work->entry));
200
		__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
L
Linus Torvalds 已提交
201 202 203 204 205
		ret = 1;
	}
	put_cpu();
	return ret;
}
206
EXPORT_SYMBOL_GPL(queue_work);
L
Linus Torvalds 已提交
207

208
void delayed_work_timer_fn(unsigned long __data)
L
Linus Torvalds 已提交
209
{
210
	struct delayed_work *dwork = (struct delayed_work *)__data;
211
	struct workqueue_struct *wq = get_wq_data(&dwork->work);
L
Linus Torvalds 已提交
212 213 214
	int cpu = smp_processor_id();

	if (unlikely(is_single_threaded(wq)))
215
		cpu = singlethread_cpu;
L
Linus Torvalds 已提交
216

217
	__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), &dwork->work);
L
Linus Torvalds 已提交
218 219
}

220 221 222
/**
 * queue_delayed_work - queue work on a workqueue after delay
 * @wq: workqueue to use
223
 * @dwork: delayable work to queue
224 225
 * @delay: number of jiffies to wait before queueing
 *
A
Alan Stern 已提交
226
 * Returns 0 if @work was already on a queue, non-zero otherwise.
227
 */
L
Linus Torvalds 已提交
228
int fastcall queue_delayed_work(struct workqueue_struct *wq,
229
			struct delayed_work *dwork, unsigned long delay)
L
Linus Torvalds 已提交
230 231
{
	int ret = 0;
232 233 234
	struct timer_list *timer = &dwork->timer;
	struct work_struct *work = &dwork->work;

235
	timer_stats_timer_set_start_info(timer);
236 237
	if (delay == 0)
		return queue_work(wq, work);
L
Linus Torvalds 已提交
238

239
	if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
L
Linus Torvalds 已提交
240 241 242 243
		BUG_ON(timer_pending(timer));
		BUG_ON(!list_empty(&work->entry));

		/* This stores wq for the moment, for the timer_fn */
244
		set_wq_data(work, wq);
L
Linus Torvalds 已提交
245
		timer->expires = jiffies + delay;
246
		timer->data = (unsigned long)dwork;
L
Linus Torvalds 已提交
247 248 249 250 251 252
		timer->function = delayed_work_timer_fn;
		add_timer(timer);
		ret = 1;
	}
	return ret;
}
253
EXPORT_SYMBOL_GPL(queue_delayed_work);
L
Linus Torvalds 已提交
254

255 256 257 258
/**
 * queue_delayed_work_on - queue work on specific CPU after delay
 * @cpu: CPU number to execute work on
 * @wq: workqueue to use
259
 * @dwork: work to queue
260 261
 * @delay: number of jiffies to wait before queueing
 *
A
Alan Stern 已提交
262
 * Returns 0 if @work was already on a queue, non-zero otherwise.
263
 */
264
int queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
265
			struct delayed_work *dwork, unsigned long delay)
266 267
{
	int ret = 0;
268 269
	struct timer_list *timer = &dwork->timer;
	struct work_struct *work = &dwork->work;
270

271
	if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
272 273 274 275
		BUG_ON(timer_pending(timer));
		BUG_ON(!list_empty(&work->entry));

		/* This stores wq for the moment, for the timer_fn */
276
		set_wq_data(work, wq);
277
		timer->expires = jiffies + delay;
278
		timer->data = (unsigned long)dwork;
279 280 281 282 283 284
		timer->function = delayed_work_timer_fn;
		add_timer_on(timer, cpu);
		ret = 1;
	}
	return ret;
}
285
EXPORT_SYMBOL_GPL(queue_delayed_work_on);
L
Linus Torvalds 已提交
286

287
static void run_workqueue(struct cpu_workqueue_struct *cwq)
L
Linus Torvalds 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
{
	unsigned long flags;

	/*
	 * Keep taking off work from the queue until
	 * done.
	 */
	spin_lock_irqsave(&cwq->lock, flags);
	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);
306
		work_func_t f = work->func;
L
Linus Torvalds 已提交
307 308 309 310

		list_del_init(cwq->worklist.next);
		spin_unlock_irqrestore(&cwq->lock, flags);

311
		BUG_ON(get_wq_data(work) != cwq);
312
		if (!test_bit(WORK_STRUCT_NOAUTOREL, work_data_bits(work)))
313 314
			work_release(work);
		f(work);
L
Linus Torvalds 已提交
315

316 317 318 319 320 321 322 323 324 325 326
		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();
		}

L
Linus Torvalds 已提交
327 328 329 330 331 332 333 334 335 336 337 338 339
		spin_lock_irqsave(&cwq->lock, flags);
	}
	cwq->run_depth--;
	spin_unlock_irqrestore(&cwq->lock, flags);
}

static int worker_thread(void *__cwq)
{
	struct cpu_workqueue_struct *cwq = __cwq;
	DECLARE_WAITQUEUE(wait, current);
	struct k_sigaction sa;
	sigset_t blocked;

340 341
	if (!cwq->freezeable)
		current->flags |= PF_NOFREEZE;
L
Linus Torvalds 已提交
342 343 344 345 346 347 348 349

	set_user_nice(current, -5);

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

350 351 352 353 354 355
	/*
	 * 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
356 357 358 359 360 361 362 363
	/* 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);

	set_current_state(TASK_INTERRUPTIBLE);
	while (!kthread_should_stop()) {
364 365 366
		if (cwq->freezeable)
			try_to_freeze();

L
Linus Torvalds 已提交
367 368 369 370 371 372 373 374 375 376 377 378 379 380 381
		add_wait_queue(&cwq->more_work, &wait);
		if (list_empty(&cwq->worklist))
			schedule();
		else
			__set_current_state(TASK_RUNNING);
		remove_wait_queue(&cwq->more_work, &wait);

		if (!list_empty(&cwq->worklist))
			run_workqueue(cwq);
		set_current_state(TASK_INTERRUPTIBLE);
	}
	__set_current_state(TASK_RUNNING);
	return 0;
}

O
Oleg Nesterov 已提交
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
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);
}

static inline void init_wq_barrier(struct wq_barrier *barr)
{
	INIT_WORK(&barr->work, wq_barrier_func);
	__set_bit(WORK_STRUCT_PENDING, work_data_bits(&barr->work));

	init_completion(&barr->done);
}

L
Linus Torvalds 已提交
401 402 403 404 405 406 407
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.
		 */
O
Oleg Nesterov 已提交
408
		mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
409
		run_workqueue(cwq);
O
Oleg Nesterov 已提交
410
		mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
411
	} else {
O
Oleg Nesterov 已提交
412
		struct wq_barrier barr;
L
Linus Torvalds 已提交
413

O
Oleg Nesterov 已提交
414 415
		init_wq_barrier(&barr);
		__queue_work(cwq, &barr.work);
L
Linus Torvalds 已提交
416

O
Oleg Nesterov 已提交
417 418 419
		mutex_unlock(&workqueue_mutex);
		wait_for_completion(&barr.done);
		mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
420 421 422
	}
}

423
/**
L
Linus Torvalds 已提交
424
 * flush_workqueue - ensure that any scheduled work has run to completion.
425
 * @wq: workqueue to flush
L
Linus Torvalds 已提交
426 427 428 429
 *
 * Forces execution of the workqueue and blocks until its completion.
 * This is typically used in driver shutdown handlers.
 *
O
Oleg Nesterov 已提交
430 431
 * 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
432 433 434 435 436 437
 *
 * 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)
{
O
Oleg Nesterov 已提交
438
	mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
439
	if (is_single_threaded(wq)) {
440
		/* Always use first cpu's area. */
441
		flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, singlethread_cpu));
L
Linus Torvalds 已提交
442 443 444 445
	} else {
		int cpu;

		for_each_online_cpu(cpu)
446
			flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu));
L
Linus Torvalds 已提交
447
	}
O
Oleg Nesterov 已提交
448
	mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
449
}
450
EXPORT_SYMBOL_GPL(flush_workqueue);
L
Linus Torvalds 已提交
451 452

static struct task_struct *create_workqueue_thread(struct workqueue_struct *wq,
453
						   int cpu, int freezeable)
L
Linus Torvalds 已提交
454
{
455
	struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
L
Linus Torvalds 已提交
456 457 458 459 460
	struct task_struct *p;

	spin_lock_init(&cwq->lock);
	cwq->wq = wq;
	cwq->thread = NULL;
461
	cwq->freezeable = freezeable;
L
Linus Torvalds 已提交
462 463 464 465 466 467 468 469 470 471 472 473 474 475
	INIT_LIST_HEAD(&cwq->worklist);
	init_waitqueue_head(&cwq->more_work);

	if (is_single_threaded(wq))
		p = kthread_create(worker_thread, cwq, "%s", wq->name);
	else
		p = kthread_create(worker_thread, cwq, "%s/%d", wq->name, cpu);
	if (IS_ERR(p))
		return NULL;
	cwq->thread = p;
	return p;
}

struct workqueue_struct *__create_workqueue(const char *name,
476
					    int singlethread, int freezeable)
L
Linus Torvalds 已提交
477 478 479 480 481
{
	int cpu, destroy = 0;
	struct workqueue_struct *wq;
	struct task_struct *p;

482
	wq = kzalloc(sizeof(*wq), GFP_KERNEL);
L
Linus Torvalds 已提交
483 484 485
	if (!wq)
		return NULL;

486
	wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
487 488 489 490 491
	if (!wq->cpu_wq) {
		kfree(wq);
		return NULL;
	}

L
Linus Torvalds 已提交
492
	wq->name = name;
493
	mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
494 495
	if (singlethread) {
		INIT_LIST_HEAD(&wq->list);
496
		p = create_workqueue_thread(wq, singlethread_cpu, freezeable);
L
Linus Torvalds 已提交
497 498 499 500 501 502 503
		if (!p)
			destroy = 1;
		else
			wake_up_process(p);
	} else {
		list_add(&wq->list, &workqueues);
		for_each_online_cpu(cpu) {
504
			p = create_workqueue_thread(wq, cpu, freezeable);
L
Linus Torvalds 已提交
505 506 507 508 509 510 511
			if (p) {
				kthread_bind(p, cpu);
				wake_up_process(p);
			} else
				destroy = 1;
		}
	}
512
	mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521 522

	/*
	 * Was there any error during startup? If yes then clean up:
	 */
	if (destroy) {
		destroy_workqueue(wq);
		wq = NULL;
	}
	return wq;
}
523
EXPORT_SYMBOL_GPL(__create_workqueue);
L
Linus Torvalds 已提交
524 525 526 527 528 529 530

static void cleanup_workqueue_thread(struct workqueue_struct *wq, int cpu)
{
	struct cpu_workqueue_struct *cwq;
	unsigned long flags;
	struct task_struct *p;

531
	cwq = per_cpu_ptr(wq->cpu_wq, cpu);
L
Linus Torvalds 已提交
532 533 534 535 536 537 538 539
	spin_lock_irqsave(&cwq->lock, flags);
	p = cwq->thread;
	cwq->thread = NULL;
	spin_unlock_irqrestore(&cwq->lock, flags);
	if (p)
		kthread_stop(p);
}

540 541 542 543 544 545
/**
 * destroy_workqueue - safely terminate a workqueue
 * @wq: target workqueue
 *
 * Safely destroy a workqueue. All work currently pending will be done first.
 */
L
Linus Torvalds 已提交
546 547 548 549 550 551 552
void destroy_workqueue(struct workqueue_struct *wq)
{
	int cpu;

	flush_workqueue(wq);

	/* We don't need the distraction of CPUs appearing and vanishing. */
553
	mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
554
	if (is_single_threaded(wq))
555
		cleanup_workqueue_thread(wq, singlethread_cpu);
L
Linus Torvalds 已提交
556 557 558 559 560
	else {
		for_each_online_cpu(cpu)
			cleanup_workqueue_thread(wq, cpu);
		list_del(&wq->list);
	}
561
	mutex_unlock(&workqueue_mutex);
562
	free_percpu(wq->cpu_wq);
L
Linus Torvalds 已提交
563 564
	kfree(wq);
}
565
EXPORT_SYMBOL_GPL(destroy_workqueue);
L
Linus Torvalds 已提交
566 567 568

static struct workqueue_struct *keventd_wq;

569 570 571 572 573 574
/**
 * 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
575 576 577 578
int fastcall schedule_work(struct work_struct *work)
{
	return queue_work(keventd_wq, work);
}
579
EXPORT_SYMBOL(schedule_work);
L
Linus Torvalds 已提交
580

581 582
/**
 * schedule_delayed_work - put work task in global workqueue after delay
583 584
 * @dwork: job to be done
 * @delay: number of jiffies to wait or 0 for immediate execution
585 586 587 588
 *
 * After waiting for a given time this puts a job in the kernel-global
 * workqueue.
 */
589 590
int fastcall schedule_delayed_work(struct delayed_work *dwork,
					unsigned long delay)
L
Linus Torvalds 已提交
591
{
592
	timer_stats_timer_set_start_info(&dwork->timer);
593
	return queue_delayed_work(keventd_wq, dwork, delay);
L
Linus Torvalds 已提交
594
}
595
EXPORT_SYMBOL(schedule_delayed_work);
L
Linus Torvalds 已提交
596

597 598 599
/**
 * schedule_delayed_work_on - queue work in global workqueue on CPU after delay
 * @cpu: cpu to use
600
 * @dwork: job to be done
601 602 603 604 605
 * @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
606
int schedule_delayed_work_on(int cpu,
607
			struct delayed_work *dwork, unsigned long delay)
L
Linus Torvalds 已提交
608
{
609
	return queue_delayed_work_on(cpu, keventd_wq, dwork, delay);
L
Linus Torvalds 已提交
610
}
611
EXPORT_SYMBOL(schedule_delayed_work_on);
L
Linus Torvalds 已提交
612

613 614 615 616 617 618 619 620 621 622 623
/**
 * 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.
 */
624
int schedule_on_each_cpu(work_func_t func)
625 626
{
	int cpu;
627
	struct work_struct *works;
628

629 630
	works = alloc_percpu(struct work_struct);
	if (!works)
631
		return -ENOMEM;
632

633
	preempt_disable();		/* CPU hotplug */
634
	for_each_online_cpu(cpu) {
635 636 637 638 639
		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);
640
	}
641
	preempt_enable();
642
	flush_workqueue(keventd_wq);
643
	free_percpu(works);
644 645 646
	return 0;
}

L
Linus Torvalds 已提交
647 648 649 650
void flush_scheduled_work(void)
{
	flush_workqueue(keventd_wq);
}
651
EXPORT_SYMBOL(flush_scheduled_work);
L
Linus Torvalds 已提交
652 653

/**
654
 * cancel_rearming_delayed_workqueue - reliably kill off a delayed work whose handler rearms the delayed work.
L
Linus Torvalds 已提交
655
 * @wq:   the controlling workqueue structure
656
 * @dwork: the delayed work struct
L
Linus Torvalds 已提交
657
 */
658
void cancel_rearming_delayed_workqueue(struct workqueue_struct *wq,
659
				       struct delayed_work *dwork)
L
Linus Torvalds 已提交
660
{
661
	while (!cancel_delayed_work(dwork))
L
Linus Torvalds 已提交
662 663
		flush_workqueue(wq);
}
664
EXPORT_SYMBOL(cancel_rearming_delayed_workqueue);
L
Linus Torvalds 已提交
665 666

/**
667
 * cancel_rearming_delayed_work - reliably kill off a delayed keventd work whose handler rearms the delayed work.
668
 * @dwork: the delayed work struct
L
Linus Torvalds 已提交
669
 */
670
void cancel_rearming_delayed_work(struct delayed_work *dwork)
L
Linus Torvalds 已提交
671
{
672
	cancel_rearming_delayed_workqueue(keventd_wq, dwork);
L
Linus Torvalds 已提交
673 674 675
}
EXPORT_SYMBOL(cancel_rearming_delayed_work);

676 677 678 679 680 681 682 683 684 685 686 687
/**
 * 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
 */
688
int execute_in_process_context(work_func_t fn, struct execute_work *ew)
689 690
{
	if (!in_interrupt()) {
691
		fn(&ew->work);
692 693 694
		return 0;
	}

695
	INIT_WORK(&ew->work, fn);
696 697 698 699 700 701
	schedule_work(&ew->work);

	return 1;
}
EXPORT_SYMBOL_GPL(execute_in_process_context);

L
Linus Torvalds 已提交
702 703 704 705 706 707 708 709 710 711 712 713 714
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);

715
	cwq = per_cpu_ptr(keventd_wq->cpu_wq, cpu);
L
Linus Torvalds 已提交
716 717 718 719 720 721 722 723 724 725
	if (current == cwq->thread)
		ret = 1;

	return ret;

}

/* Take the work from this (downed) CPU. */
static void take_over_work(struct workqueue_struct *wq, unsigned int cpu)
{
726
	struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
727
	struct list_head list;
L
Linus Torvalds 已提交
728 729 730
	struct work_struct *work;

	spin_lock_irq(&cwq->lock);
731
	list_replace_init(&cwq->worklist, &list);
L
Linus Torvalds 已提交
732 733 734 735 736

	while (!list_empty(&list)) {
		printk("Taking work for %s\n", wq->name);
		work = list_entry(list.next,struct work_struct,entry);
		list_del(&work->entry);
737
		__queue_work(per_cpu_ptr(wq->cpu_wq, smp_processor_id()), work);
L
Linus Torvalds 已提交
738 739 740 741 742
	}
	spin_unlock_irq(&cwq->lock);
}

/* We're holding the cpucontrol mutex here */
743
static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
L
Linus Torvalds 已提交
744 745 746 747 748 749 750 751
				  unsigned long action,
				  void *hcpu)
{
	unsigned int hotcpu = (unsigned long)hcpu;
	struct workqueue_struct *wq;

	switch (action) {
	case CPU_UP_PREPARE:
752
		mutex_lock(&workqueue_mutex);
L
Linus Torvalds 已提交
753 754
		/* Create a new workqueue thread for it. */
		list_for_each_entry(wq, &workqueues, list) {
755
			if (!create_workqueue_thread(wq, hotcpu, 0)) {
L
Linus Torvalds 已提交
756 757 758 759 760 761 762 763 764
				printk("workqueue for %i failed\n", hotcpu);
				return NOTIFY_BAD;
			}
		}
		break;

	case CPU_ONLINE:
		/* Kick off worker threads. */
		list_for_each_entry(wq, &workqueues, list) {
765 766 767 768 769
			struct cpu_workqueue_struct *cwq;

			cwq = per_cpu_ptr(wq->cpu_wq, hotcpu);
			kthread_bind(cwq->thread, hotcpu);
			wake_up_process(cwq->thread);
L
Linus Torvalds 已提交
770
		}
771
		mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
772 773 774 775
		break;

	case CPU_UP_CANCELED:
		list_for_each_entry(wq, &workqueues, list) {
776 777
			if (!per_cpu_ptr(wq->cpu_wq, hotcpu)->thread)
				continue;
L
Linus Torvalds 已提交
778
			/* Unbind so it can run. */
779
			kthread_bind(per_cpu_ptr(wq->cpu_wq, hotcpu)->thread,
780
				     any_online_cpu(cpu_online_map));
L
Linus Torvalds 已提交
781 782
			cleanup_workqueue_thread(wq, hotcpu);
		}
783 784 785 786 787 788 789 790 791
		mutex_unlock(&workqueue_mutex);
		break;

	case CPU_DOWN_PREPARE:
		mutex_lock(&workqueue_mutex);
		break;

	case CPU_DOWN_FAILED:
		mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
792 793 794 795 796 797 798
		break;

	case CPU_DEAD:
		list_for_each_entry(wq, &workqueues, list)
			cleanup_workqueue_thread(wq, hotcpu);
		list_for_each_entry(wq, &workqueues, list)
			take_over_work(wq, hotcpu);
799
		mutex_unlock(&workqueue_mutex);
L
Linus Torvalds 已提交
800 801 802 803 804 805 806 807
		break;
	}

	return NOTIFY_OK;
}

void init_workqueues(void)
{
808
	singlethread_cpu = first_cpu(cpu_possible_map);
L
Linus Torvalds 已提交
809 810 811 812 813
	hotcpu_notifier(workqueue_cpu_callback, 0);
	keventd_wq = create_workqueue("events");
	BUG_ON(!keventd_wq);
}