xref: /linux/kernel/workqueue.c (revision 606d099cdd1080bbb50ea50dc52d98252f8f10a1)
1 /*
2  * linux/kernel/workqueue.c
3  *
4  * Generic mechanism for defining kernel helper threads for running
5  * arbitrary tasks in process context.
6  *
7  * Started by Ingo Molnar, Copyright (C) 2002
8  *
9  * Derived from the taskqueue/keventd code by:
10  *
11  *   David Woodhouse <dwmw2@infradead.org>
12  *   Andrew Morton <andrewm@uow.edu.au>
13  *   Kai Petzke <wpp@marie.physik.tu-berlin.de>
14  *   Theodore Ts'o <tytso@mit.edu>
15  *
16  * Made to use alloc_percpu by Christoph Lameter <clameter@sgi.com>.
17  */
18 
19 #include <linux/module.h>
20 #include <linux/kernel.h>
21 #include <linux/sched.h>
22 #include <linux/init.h>
23 #include <linux/signal.h>
24 #include <linux/completion.h>
25 #include <linux/workqueue.h>
26 #include <linux/slab.h>
27 #include <linux/cpu.h>
28 #include <linux/notifier.h>
29 #include <linux/kthread.h>
30 #include <linux/hardirq.h>
31 #include <linux/mempolicy.h>
32 #include <linux/freezer.h>
33 #include <linux/kallsyms.h>
34 #include <linux/debug_locks.h>
35 
36 /*
37  * The per-CPU workqueue (if single thread, we always use the first
38  * possible cpu).
39  *
40  * The sequence counters are for flush_scheduled_work().  It wants to wait
41  * until all currently-scheduled works are completed, but it doesn't
42  * want to be livelocked by new, incoming ones.  So it waits until
43  * remove_sequence is >= the insert_sequence which pertained when
44  * flush_scheduled_work() was called.
45  */
46 struct cpu_workqueue_struct {
47 
48 	spinlock_t lock;
49 
50 	long remove_sequence;	/* Least-recently added (next to run) */
51 	long insert_sequence;	/* Next to add */
52 
53 	struct list_head worklist;
54 	wait_queue_head_t more_work;
55 	wait_queue_head_t work_done;
56 
57 	struct workqueue_struct *wq;
58 	struct task_struct *thread;
59 
60 	int run_depth;		/* Detect run_workqueue() recursion depth */
61 
62 	int freezeable;		/* Freeze the thread during suspend */
63 } ____cacheline_aligned;
64 
65 /*
66  * The externally visible workqueue abstraction is an array of
67  * per-CPU workqueues:
68  */
69 struct workqueue_struct {
70 	struct cpu_workqueue_struct *cpu_wq;
71 	const char *name;
72 	struct list_head list; 	/* Empty if single thread */
73 };
74 
75 /* All the per-cpu workqueues on the system, for hotplug cpu to add/remove
76    threads to each one as cpus come/go. */
77 static DEFINE_MUTEX(workqueue_mutex);
78 static LIST_HEAD(workqueues);
79 
80 static int singlethread_cpu;
81 
82 /* If it's single threaded, it isn't in the list of workqueues. */
83 static inline int is_single_threaded(struct workqueue_struct *wq)
84 {
85 	return list_empty(&wq->list);
86 }
87 
88 static inline void set_wq_data(struct work_struct *work, void *wq)
89 {
90 	unsigned long new, old, res;
91 
92 	/* assume the pending flag is already set and that the task has already
93 	 * been queued on this workqueue */
94 	new = (unsigned long) wq | (1UL << WORK_STRUCT_PENDING);
95 	res = work->management;
96 	if (res != new) {
97 		do {
98 			old = res;
99 			new = (unsigned long) wq;
100 			new |= (old & WORK_STRUCT_FLAG_MASK);
101 			res = cmpxchg(&work->management, old, new);
102 		} while (res != old);
103 	}
104 }
105 
106 static inline void *get_wq_data(struct work_struct *work)
107 {
108 	return (void *) (work->management & WORK_STRUCT_WQ_DATA_MASK);
109 }
110 
111 static int __run_work(struct cpu_workqueue_struct *cwq, struct work_struct *work)
112 {
113 	int ret = 0;
114 	unsigned long flags;
115 
116 	spin_lock_irqsave(&cwq->lock, flags);
117 	/*
118 	 * We need to re-validate the work info after we've gotten
119 	 * the cpu_workqueue lock. We can run the work now iff:
120 	 *
121 	 *  - the wq_data still matches the cpu_workqueue_struct
122 	 *  - AND the work is still marked pending
123 	 *  - AND the work is still on a list (which will be this
124 	 *    workqueue_struct list)
125 	 *
126 	 * All these conditions are important, because we
127 	 * need to protect against the work being run right
128 	 * now on another CPU (all but the last one might be
129 	 * true if it's currently running and has not been
130 	 * released yet, for example).
131 	 */
132 	if (get_wq_data(work) == cwq
133 	    && work_pending(work)
134 	    && !list_empty(&work->entry)) {
135 		work_func_t f = work->func;
136 		list_del_init(&work->entry);
137 		spin_unlock_irqrestore(&cwq->lock, flags);
138 
139 		if (!test_bit(WORK_STRUCT_NOAUTOREL, &work->management))
140 			work_release(work);
141 		f(work);
142 
143 		spin_lock_irqsave(&cwq->lock, flags);
144 		cwq->remove_sequence++;
145 		wake_up(&cwq->work_done);
146 		ret = 1;
147 	}
148 	spin_unlock_irqrestore(&cwq->lock, flags);
149 	return ret;
150 }
151 
152 /**
153  * run_scheduled_work - run scheduled work synchronously
154  * @work: work to run
155  *
156  * This checks if the work was pending, and runs it
157  * synchronously if so. It returns a boolean to indicate
158  * whether it had any scheduled work to run or not.
159  *
160  * NOTE! This _only_ works for normal work_structs. You
161  * CANNOT use this for delayed work, because the wq data
162  * for delayed work will not point properly to the per-
163  * CPU workqueue struct, but will change!
164  */
165 int fastcall run_scheduled_work(struct work_struct *work)
166 {
167 	for (;;) {
168 		struct cpu_workqueue_struct *cwq;
169 
170 		if (!work_pending(work))
171 			return 0;
172 		if (list_empty(&work->entry))
173 			return 0;
174 		/* NOTE! This depends intimately on __queue_work! */
175 		cwq = get_wq_data(work);
176 		if (!cwq)
177 			return 0;
178 		if (__run_work(cwq, work))
179 			return 1;
180 	}
181 }
182 EXPORT_SYMBOL(run_scheduled_work);
183 
184 /* Preempt must be disabled. */
185 static void __queue_work(struct cpu_workqueue_struct *cwq,
186 			 struct work_struct *work)
187 {
188 	unsigned long flags;
189 
190 	spin_lock_irqsave(&cwq->lock, flags);
191 	set_wq_data(work, cwq);
192 	list_add_tail(&work->entry, &cwq->worklist);
193 	cwq->insert_sequence++;
194 	wake_up(&cwq->more_work);
195 	spin_unlock_irqrestore(&cwq->lock, flags);
196 }
197 
198 /**
199  * queue_work - queue work on a workqueue
200  * @wq: workqueue to use
201  * @work: work to queue
202  *
203  * Returns 0 if @work was already on a queue, non-zero otherwise.
204  *
205  * We queue the work to the CPU it was submitted, but there is no
206  * guarantee that it will be processed by that CPU.
207  */
208 int fastcall queue_work(struct workqueue_struct *wq, struct work_struct *work)
209 {
210 	int ret = 0, cpu = get_cpu();
211 
212 	if (!test_and_set_bit(WORK_STRUCT_PENDING, &work->management)) {
213 		if (unlikely(is_single_threaded(wq)))
214 			cpu = singlethread_cpu;
215 		BUG_ON(!list_empty(&work->entry));
216 		__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
217 		ret = 1;
218 	}
219 	put_cpu();
220 	return ret;
221 }
222 EXPORT_SYMBOL_GPL(queue_work);
223 
224 static void delayed_work_timer_fn(unsigned long __data)
225 {
226 	struct delayed_work *dwork = (struct delayed_work *)__data;
227 	struct workqueue_struct *wq = get_wq_data(&dwork->work);
228 	int cpu = smp_processor_id();
229 
230 	if (unlikely(is_single_threaded(wq)))
231 		cpu = singlethread_cpu;
232 
233 	__queue_work(per_cpu_ptr(wq->cpu_wq, cpu), &dwork->work);
234 }
235 
236 /**
237  * queue_delayed_work - queue work on a workqueue after delay
238  * @wq: workqueue to use
239  * @work: delayable work to queue
240  * @delay: number of jiffies to wait before queueing
241  *
242  * Returns 0 if @work was already on a queue, non-zero otherwise.
243  */
244 int fastcall queue_delayed_work(struct workqueue_struct *wq,
245 			struct delayed_work *dwork, unsigned long delay)
246 {
247 	int ret = 0;
248 	struct timer_list *timer = &dwork->timer;
249 	struct work_struct *work = &dwork->work;
250 
251 	if (delay == 0)
252 		return queue_work(wq, work);
253 
254 	if (!test_and_set_bit(WORK_STRUCT_PENDING, &work->management)) {
255 		BUG_ON(timer_pending(timer));
256 		BUG_ON(!list_empty(&work->entry));
257 
258 		/* This stores wq for the moment, for the timer_fn */
259 		set_wq_data(work, wq);
260 		timer->expires = jiffies + delay;
261 		timer->data = (unsigned long)dwork;
262 		timer->function = delayed_work_timer_fn;
263 		add_timer(timer);
264 		ret = 1;
265 	}
266 	return ret;
267 }
268 EXPORT_SYMBOL_GPL(queue_delayed_work);
269 
270 /**
271  * queue_delayed_work_on - queue work on specific CPU after delay
272  * @cpu: CPU number to execute work on
273  * @wq: workqueue to use
274  * @work: work to queue
275  * @delay: number of jiffies to wait before queueing
276  *
277  * Returns 0 if @work was already on a queue, non-zero otherwise.
278  */
279 int queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
280 			struct delayed_work *dwork, unsigned long delay)
281 {
282 	int ret = 0;
283 	struct timer_list *timer = &dwork->timer;
284 	struct work_struct *work = &dwork->work;
285 
286 	if (!test_and_set_bit(WORK_STRUCT_PENDING, &work->management)) {
287 		BUG_ON(timer_pending(timer));
288 		BUG_ON(!list_empty(&work->entry));
289 
290 		/* This stores wq for the moment, for the timer_fn */
291 		set_wq_data(work, wq);
292 		timer->expires = jiffies + delay;
293 		timer->data = (unsigned long)dwork;
294 		timer->function = delayed_work_timer_fn;
295 		add_timer_on(timer, cpu);
296 		ret = 1;
297 	}
298 	return ret;
299 }
300 EXPORT_SYMBOL_GPL(queue_delayed_work_on);
301 
302 static void run_workqueue(struct cpu_workqueue_struct *cwq)
303 {
304 	unsigned long flags;
305 
306 	/*
307 	 * Keep taking off work from the queue until
308 	 * done.
309 	 */
310 	spin_lock_irqsave(&cwq->lock, flags);
311 	cwq->run_depth++;
312 	if (cwq->run_depth > 3) {
313 		/* morton gets to eat his hat */
314 		printk("%s: recursion depth exceeded: %d\n",
315 			__FUNCTION__, cwq->run_depth);
316 		dump_stack();
317 	}
318 	while (!list_empty(&cwq->worklist)) {
319 		struct work_struct *work = list_entry(cwq->worklist.next,
320 						struct work_struct, entry);
321 		work_func_t f = work->func;
322 
323 		list_del_init(cwq->worklist.next);
324 		spin_unlock_irqrestore(&cwq->lock, flags);
325 
326 		BUG_ON(get_wq_data(work) != cwq);
327 		if (!test_bit(WORK_STRUCT_NOAUTOREL, &work->management))
328 			work_release(work);
329 		f(work);
330 
331 		if (unlikely(in_atomic() || lockdep_depth(current) > 0)) {
332 			printk(KERN_ERR "BUG: workqueue leaked lock or atomic: "
333 					"%s/0x%08x/%d\n",
334 					current->comm, preempt_count(),
335 				       	current->pid);
336 			printk(KERN_ERR "    last function: ");
337 			print_symbol("%s\n", (unsigned long)f);
338 			debug_show_held_locks(current);
339 			dump_stack();
340 		}
341 
342 		spin_lock_irqsave(&cwq->lock, flags);
343 		cwq->remove_sequence++;
344 		wake_up(&cwq->work_done);
345 	}
346 	cwq->run_depth--;
347 	spin_unlock_irqrestore(&cwq->lock, flags);
348 }
349 
350 static int worker_thread(void *__cwq)
351 {
352 	struct cpu_workqueue_struct *cwq = __cwq;
353 	DECLARE_WAITQUEUE(wait, current);
354 	struct k_sigaction sa;
355 	sigset_t blocked;
356 
357 	if (!cwq->freezeable)
358 		current->flags |= PF_NOFREEZE;
359 
360 	set_user_nice(current, -5);
361 
362 	/* Block and flush all signals */
363 	sigfillset(&blocked);
364 	sigprocmask(SIG_BLOCK, &blocked, NULL);
365 	flush_signals(current);
366 
367 	/*
368 	 * We inherited MPOL_INTERLEAVE from the booting kernel.
369 	 * Set MPOL_DEFAULT to insure node local allocations.
370 	 */
371 	numa_default_policy();
372 
373 	/* SIG_IGN makes children autoreap: see do_notify_parent(). */
374 	sa.sa.sa_handler = SIG_IGN;
375 	sa.sa.sa_flags = 0;
376 	siginitset(&sa.sa.sa_mask, sigmask(SIGCHLD));
377 	do_sigaction(SIGCHLD, &sa, (struct k_sigaction *)0);
378 
379 	set_current_state(TASK_INTERRUPTIBLE);
380 	while (!kthread_should_stop()) {
381 		if (cwq->freezeable)
382 			try_to_freeze();
383 
384 		add_wait_queue(&cwq->more_work, &wait);
385 		if (list_empty(&cwq->worklist))
386 			schedule();
387 		else
388 			__set_current_state(TASK_RUNNING);
389 		remove_wait_queue(&cwq->more_work, &wait);
390 
391 		if (!list_empty(&cwq->worklist))
392 			run_workqueue(cwq);
393 		set_current_state(TASK_INTERRUPTIBLE);
394 	}
395 	__set_current_state(TASK_RUNNING);
396 	return 0;
397 }
398 
399 static void flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
400 {
401 	if (cwq->thread == current) {
402 		/*
403 		 * Probably keventd trying to flush its own queue. So simply run
404 		 * it by hand rather than deadlocking.
405 		 */
406 		run_workqueue(cwq);
407 	} else {
408 		DEFINE_WAIT(wait);
409 		long sequence_needed;
410 
411 		spin_lock_irq(&cwq->lock);
412 		sequence_needed = cwq->insert_sequence;
413 
414 		while (sequence_needed - cwq->remove_sequence > 0) {
415 			prepare_to_wait(&cwq->work_done, &wait,
416 					TASK_UNINTERRUPTIBLE);
417 			spin_unlock_irq(&cwq->lock);
418 			schedule();
419 			spin_lock_irq(&cwq->lock);
420 		}
421 		finish_wait(&cwq->work_done, &wait);
422 		spin_unlock_irq(&cwq->lock);
423 	}
424 }
425 
426 /**
427  * flush_workqueue - ensure that any scheduled work has run to completion.
428  * @wq: workqueue to flush
429  *
430  * Forces execution of the workqueue and blocks until its completion.
431  * This is typically used in driver shutdown handlers.
432  *
433  * This function will sample each workqueue's current insert_sequence number and
434  * will sleep until the head sequence is greater than or equal to that.  This
435  * means that we sleep until all works which were queued on entry have been
436  * handled, but we are not livelocked by new incoming ones.
437  *
438  * This function used to run the workqueues itself.  Now we just wait for the
439  * helper threads to do it.
440  */
441 void fastcall flush_workqueue(struct workqueue_struct *wq)
442 {
443 	might_sleep();
444 
445 	if (is_single_threaded(wq)) {
446 		/* Always use first cpu's area. */
447 		flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, singlethread_cpu));
448 	} else {
449 		int cpu;
450 
451 		mutex_lock(&workqueue_mutex);
452 		for_each_online_cpu(cpu)
453 			flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu));
454 		mutex_unlock(&workqueue_mutex);
455 	}
456 }
457 EXPORT_SYMBOL_GPL(flush_workqueue);
458 
459 static struct task_struct *create_workqueue_thread(struct workqueue_struct *wq,
460 						   int cpu, int freezeable)
461 {
462 	struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
463 	struct task_struct *p;
464 
465 	spin_lock_init(&cwq->lock);
466 	cwq->wq = wq;
467 	cwq->thread = NULL;
468 	cwq->insert_sequence = 0;
469 	cwq->remove_sequence = 0;
470 	cwq->freezeable = freezeable;
471 	INIT_LIST_HEAD(&cwq->worklist);
472 	init_waitqueue_head(&cwq->more_work);
473 	init_waitqueue_head(&cwq->work_done);
474 
475 	if (is_single_threaded(wq))
476 		p = kthread_create(worker_thread, cwq, "%s", wq->name);
477 	else
478 		p = kthread_create(worker_thread, cwq, "%s/%d", wq->name, cpu);
479 	if (IS_ERR(p))
480 		return NULL;
481 	cwq->thread = p;
482 	return p;
483 }
484 
485 struct workqueue_struct *__create_workqueue(const char *name,
486 					    int singlethread, int freezeable)
487 {
488 	int cpu, destroy = 0;
489 	struct workqueue_struct *wq;
490 	struct task_struct *p;
491 
492 	wq = kzalloc(sizeof(*wq), GFP_KERNEL);
493 	if (!wq)
494 		return NULL;
495 
496 	wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
497 	if (!wq->cpu_wq) {
498 		kfree(wq);
499 		return NULL;
500 	}
501 
502 	wq->name = name;
503 	mutex_lock(&workqueue_mutex);
504 	if (singlethread) {
505 		INIT_LIST_HEAD(&wq->list);
506 		p = create_workqueue_thread(wq, singlethread_cpu, freezeable);
507 		if (!p)
508 			destroy = 1;
509 		else
510 			wake_up_process(p);
511 	} else {
512 		list_add(&wq->list, &workqueues);
513 		for_each_online_cpu(cpu) {
514 			p = create_workqueue_thread(wq, cpu, freezeable);
515 			if (p) {
516 				kthread_bind(p, cpu);
517 				wake_up_process(p);
518 			} else
519 				destroy = 1;
520 		}
521 	}
522 	mutex_unlock(&workqueue_mutex);
523 
524 	/*
525 	 * Was there any error during startup? If yes then clean up:
526 	 */
527 	if (destroy) {
528 		destroy_workqueue(wq);
529 		wq = NULL;
530 	}
531 	return wq;
532 }
533 EXPORT_SYMBOL_GPL(__create_workqueue);
534 
535 static void cleanup_workqueue_thread(struct workqueue_struct *wq, int cpu)
536 {
537 	struct cpu_workqueue_struct *cwq;
538 	unsigned long flags;
539 	struct task_struct *p;
540 
541 	cwq = per_cpu_ptr(wq->cpu_wq, cpu);
542 	spin_lock_irqsave(&cwq->lock, flags);
543 	p = cwq->thread;
544 	cwq->thread = NULL;
545 	spin_unlock_irqrestore(&cwq->lock, flags);
546 	if (p)
547 		kthread_stop(p);
548 }
549 
550 /**
551  * destroy_workqueue - safely terminate a workqueue
552  * @wq: target workqueue
553  *
554  * Safely destroy a workqueue. All work currently pending will be done first.
555  */
556 void destroy_workqueue(struct workqueue_struct *wq)
557 {
558 	int cpu;
559 
560 	flush_workqueue(wq);
561 
562 	/* We don't need the distraction of CPUs appearing and vanishing. */
563 	mutex_lock(&workqueue_mutex);
564 	if (is_single_threaded(wq))
565 		cleanup_workqueue_thread(wq, singlethread_cpu);
566 	else {
567 		for_each_online_cpu(cpu)
568 			cleanup_workqueue_thread(wq, cpu);
569 		list_del(&wq->list);
570 	}
571 	mutex_unlock(&workqueue_mutex);
572 	free_percpu(wq->cpu_wq);
573 	kfree(wq);
574 }
575 EXPORT_SYMBOL_GPL(destroy_workqueue);
576 
577 static struct workqueue_struct *keventd_wq;
578 
579 /**
580  * schedule_work - put work task in global workqueue
581  * @work: job to be done
582  *
583  * This puts a job in the kernel-global workqueue.
584  */
585 int fastcall schedule_work(struct work_struct *work)
586 {
587 	return queue_work(keventd_wq, work);
588 }
589 EXPORT_SYMBOL(schedule_work);
590 
591 /**
592  * schedule_delayed_work - put work task in global workqueue after delay
593  * @dwork: job to be done
594  * @delay: number of jiffies to wait or 0 for immediate execution
595  *
596  * After waiting for a given time this puts a job in the kernel-global
597  * workqueue.
598  */
599 int fastcall schedule_delayed_work(struct delayed_work *dwork, unsigned long delay)
600 {
601 	return queue_delayed_work(keventd_wq, dwork, delay);
602 }
603 EXPORT_SYMBOL(schedule_delayed_work);
604 
605 /**
606  * schedule_delayed_work_on - queue work in global workqueue on CPU after delay
607  * @cpu: cpu to use
608  * @dwork: job to be done
609  * @delay: number of jiffies to wait
610  *
611  * After waiting for a given time this puts a job in the kernel-global
612  * workqueue on the specified CPU.
613  */
614 int schedule_delayed_work_on(int cpu,
615 			struct delayed_work *dwork, unsigned long delay)
616 {
617 	return queue_delayed_work_on(cpu, keventd_wq, dwork, delay);
618 }
619 EXPORT_SYMBOL(schedule_delayed_work_on);
620 
621 /**
622  * schedule_on_each_cpu - call a function on each online CPU from keventd
623  * @func: the function to call
624  *
625  * Returns zero on success.
626  * Returns -ve errno on failure.
627  *
628  * Appears to be racy against CPU hotplug.
629  *
630  * schedule_on_each_cpu() is very slow.
631  */
632 int schedule_on_each_cpu(work_func_t func)
633 {
634 	int cpu;
635 	struct work_struct *works;
636 
637 	works = alloc_percpu(struct work_struct);
638 	if (!works)
639 		return -ENOMEM;
640 
641 	mutex_lock(&workqueue_mutex);
642 	for_each_online_cpu(cpu) {
643 		INIT_WORK(per_cpu_ptr(works, cpu), func);
644 		__queue_work(per_cpu_ptr(keventd_wq->cpu_wq, cpu),
645 				per_cpu_ptr(works, cpu));
646 	}
647 	mutex_unlock(&workqueue_mutex);
648 	flush_workqueue(keventd_wq);
649 	free_percpu(works);
650 	return 0;
651 }
652 
653 void flush_scheduled_work(void)
654 {
655 	flush_workqueue(keventd_wq);
656 }
657 EXPORT_SYMBOL(flush_scheduled_work);
658 
659 /**
660  * cancel_rearming_delayed_workqueue - reliably kill off a delayed
661  *			work whose handler rearms the delayed work.
662  * @wq:   the controlling workqueue structure
663  * @dwork: the delayed work struct
664  */
665 void cancel_rearming_delayed_workqueue(struct workqueue_struct *wq,
666 				       struct delayed_work *dwork)
667 {
668 	while (!cancel_delayed_work(dwork))
669 		flush_workqueue(wq);
670 }
671 EXPORT_SYMBOL(cancel_rearming_delayed_workqueue);
672 
673 /**
674  * cancel_rearming_delayed_work - reliably kill off a delayed keventd
675  *			work whose handler rearms the delayed work.
676  * @dwork: the delayed work struct
677  */
678 void cancel_rearming_delayed_work(struct delayed_work *dwork)
679 {
680 	cancel_rearming_delayed_workqueue(keventd_wq, dwork);
681 }
682 EXPORT_SYMBOL(cancel_rearming_delayed_work);
683 
684 /**
685  * execute_in_process_context - reliably execute the routine with user context
686  * @fn:		the function to execute
687  * @ew:		guaranteed storage for the execute work structure (must
688  *		be available when the work executes)
689  *
690  * Executes the function immediately if process context is available,
691  * otherwise schedules the function for delayed execution.
692  *
693  * Returns:	0 - function was executed
694  *		1 - function was scheduled for execution
695  */
696 int execute_in_process_context(work_func_t fn, struct execute_work *ew)
697 {
698 	if (!in_interrupt()) {
699 		fn(&ew->work);
700 		return 0;
701 	}
702 
703 	INIT_WORK(&ew->work, fn);
704 	schedule_work(&ew->work);
705 
706 	return 1;
707 }
708 EXPORT_SYMBOL_GPL(execute_in_process_context);
709 
710 int keventd_up(void)
711 {
712 	return keventd_wq != NULL;
713 }
714 
715 int current_is_keventd(void)
716 {
717 	struct cpu_workqueue_struct *cwq;
718 	int cpu = smp_processor_id();	/* preempt-safe: keventd is per-cpu */
719 	int ret = 0;
720 
721 	BUG_ON(!keventd_wq);
722 
723 	cwq = per_cpu_ptr(keventd_wq->cpu_wq, cpu);
724 	if (current == cwq->thread)
725 		ret = 1;
726 
727 	return ret;
728 
729 }
730 
731 /* Take the work from this (downed) CPU. */
732 static void take_over_work(struct workqueue_struct *wq, unsigned int cpu)
733 {
734 	struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
735 	struct list_head list;
736 	struct work_struct *work;
737 
738 	spin_lock_irq(&cwq->lock);
739 	list_replace_init(&cwq->worklist, &list);
740 
741 	while (!list_empty(&list)) {
742 		printk("Taking work for %s\n", wq->name);
743 		work = list_entry(list.next,struct work_struct,entry);
744 		list_del(&work->entry);
745 		__queue_work(per_cpu_ptr(wq->cpu_wq, smp_processor_id()), work);
746 	}
747 	spin_unlock_irq(&cwq->lock);
748 }
749 
750 /* We're holding the cpucontrol mutex here */
751 static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
752 				  unsigned long action,
753 				  void *hcpu)
754 {
755 	unsigned int hotcpu = (unsigned long)hcpu;
756 	struct workqueue_struct *wq;
757 
758 	switch (action) {
759 	case CPU_UP_PREPARE:
760 		mutex_lock(&workqueue_mutex);
761 		/* Create a new workqueue thread for it. */
762 		list_for_each_entry(wq, &workqueues, list) {
763 			if (!create_workqueue_thread(wq, hotcpu, 0)) {
764 				printk("workqueue for %i failed\n", hotcpu);
765 				return NOTIFY_BAD;
766 			}
767 		}
768 		break;
769 
770 	case CPU_ONLINE:
771 		/* Kick off worker threads. */
772 		list_for_each_entry(wq, &workqueues, list) {
773 			struct cpu_workqueue_struct *cwq;
774 
775 			cwq = per_cpu_ptr(wq->cpu_wq, hotcpu);
776 			kthread_bind(cwq->thread, hotcpu);
777 			wake_up_process(cwq->thread);
778 		}
779 		mutex_unlock(&workqueue_mutex);
780 		break;
781 
782 	case CPU_UP_CANCELED:
783 		list_for_each_entry(wq, &workqueues, list) {
784 			if (!per_cpu_ptr(wq->cpu_wq, hotcpu)->thread)
785 				continue;
786 			/* Unbind so it can run. */
787 			kthread_bind(per_cpu_ptr(wq->cpu_wq, hotcpu)->thread,
788 				     any_online_cpu(cpu_online_map));
789 			cleanup_workqueue_thread(wq, hotcpu);
790 		}
791 		mutex_unlock(&workqueue_mutex);
792 		break;
793 
794 	case CPU_DOWN_PREPARE:
795 		mutex_lock(&workqueue_mutex);
796 		break;
797 
798 	case CPU_DOWN_FAILED:
799 		mutex_unlock(&workqueue_mutex);
800 		break;
801 
802 	case CPU_DEAD:
803 		list_for_each_entry(wq, &workqueues, list)
804 			cleanup_workqueue_thread(wq, hotcpu);
805 		list_for_each_entry(wq, &workqueues, list)
806 			take_over_work(wq, hotcpu);
807 		mutex_unlock(&workqueue_mutex);
808 		break;
809 	}
810 
811 	return NOTIFY_OK;
812 }
813 
814 void init_workqueues(void)
815 {
816 	singlethread_cpu = first_cpu(cpu_possible_map);
817 	hotcpu_notifier(workqueue_cpu_callback, 0);
818 	keventd_wq = create_workqueue("events");
819 	BUG_ON(!keventd_wq);
820 }
821 
822