[PATCH] Add debugging feature /proc/timer_stat
[linux-2.6.git] / kernel / workqueue.c
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 /*
89  * Set the workqueue on which a work item is to be run
90  * - Must *only* be called if the pending flag is set
91  */
92 static inline void set_wq_data(struct work_struct *work, void *wq)
93 {
94         unsigned long new;
95
96         BUG_ON(!work_pending(work));
97
98         new = (unsigned long) wq | (1UL << WORK_STRUCT_PENDING);
99         new |= WORK_STRUCT_FLAG_MASK & *work_data_bits(work);
100         atomic_long_set(&work->data, new);
101 }
102
103 static inline void *get_wq_data(struct work_struct *work)
104 {
105         return (void *) (atomic_long_read(&work->data) & WORK_STRUCT_WQ_DATA_MASK);
106 }
107
108 static int __run_work(struct cpu_workqueue_struct *cwq, struct work_struct *work)
109 {
110         int ret = 0;
111         unsigned long flags;
112
113         spin_lock_irqsave(&cwq->lock, flags);
114         /*
115          * We need to re-validate the work info after we've gotten
116          * the cpu_workqueue lock. We can run the work now iff:
117          *
118          *  - the wq_data still matches the cpu_workqueue_struct
119          *  - AND the work is still marked pending
120          *  - AND the work is still on a list (which will be this
121          *    workqueue_struct list)
122          *
123          * All these conditions are important, because we
124          * need to protect against the work being run right
125          * now on another CPU (all but the last one might be
126          * true if it's currently running and has not been
127          * released yet, for example).
128          */
129         if (get_wq_data(work) == cwq
130             && work_pending(work)
131             && !list_empty(&work->entry)) {
132                 work_func_t f = work->func;
133                 list_del_init(&work->entry);
134                 spin_unlock_irqrestore(&cwq->lock, flags);
135
136                 if (!test_bit(WORK_STRUCT_NOAUTOREL, work_data_bits(work)))
137                         work_release(work);
138                 f(work);
139
140                 spin_lock_irqsave(&cwq->lock, flags);
141                 cwq->remove_sequence++;
142                 wake_up(&cwq->work_done);
143                 ret = 1;
144         }
145         spin_unlock_irqrestore(&cwq->lock, flags);
146         return ret;
147 }
148
149 /**
150  * run_scheduled_work - run scheduled work synchronously
151  * @work: work to run
152  *
153  * This checks if the work was pending, and runs it
154  * synchronously if so. It returns a boolean to indicate
155  * whether it had any scheduled work to run or not.
156  *
157  * NOTE! This _only_ works for normal work_structs. You
158  * CANNOT use this for delayed work, because the wq data
159  * for delayed work will not point properly to the per-
160  * CPU workqueue struct, but will change!
161  */
162 int fastcall run_scheduled_work(struct work_struct *work)
163 {
164         for (;;) {
165                 struct cpu_workqueue_struct *cwq;
166
167                 if (!work_pending(work))
168                         return 0;
169                 if (list_empty(&work->entry))
170                         return 0;
171                 /* NOTE! This depends intimately on __queue_work! */
172                 cwq = get_wq_data(work);
173                 if (!cwq)
174                         return 0;
175                 if (__run_work(cwq, work))
176                         return 1;
177         }
178 }
179 EXPORT_SYMBOL(run_scheduled_work);
180
181 /* Preempt must be disabled. */
182 static void __queue_work(struct cpu_workqueue_struct *cwq,
183                          struct work_struct *work)
184 {
185         unsigned long flags;
186
187         spin_lock_irqsave(&cwq->lock, flags);
188         set_wq_data(work, cwq);
189         list_add_tail(&work->entry, &cwq->worklist);
190         cwq->insert_sequence++;
191         wake_up(&cwq->more_work);
192         spin_unlock_irqrestore(&cwq->lock, flags);
193 }
194
195 /**
196  * queue_work - queue work on a workqueue
197  * @wq: workqueue to use
198  * @work: work to queue
199  *
200  * Returns 0 if @work was already on a queue, non-zero otherwise.
201  *
202  * We queue the work to the CPU it was submitted, but there is no
203  * guarantee that it will be processed by that CPU.
204  */
205 int fastcall queue_work(struct workqueue_struct *wq, struct work_struct *work)
206 {
207         int ret = 0, cpu = get_cpu();
208
209         if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
210                 if (unlikely(is_single_threaded(wq)))
211                         cpu = singlethread_cpu;
212                 BUG_ON(!list_empty(&work->entry));
213                 __queue_work(per_cpu_ptr(wq->cpu_wq, cpu), work);
214                 ret = 1;
215         }
216         put_cpu();
217         return ret;
218 }
219 EXPORT_SYMBOL_GPL(queue_work);
220
221 void delayed_work_timer_fn(unsigned long __data)
222 {
223         struct delayed_work *dwork = (struct delayed_work *)__data;
224         struct workqueue_struct *wq = get_wq_data(&dwork->work);
225         int cpu = smp_processor_id();
226
227         if (unlikely(is_single_threaded(wq)))
228                 cpu = singlethread_cpu;
229
230         __queue_work(per_cpu_ptr(wq->cpu_wq, cpu), &dwork->work);
231 }
232
233 /**
234  * queue_delayed_work - queue work on a workqueue after delay
235  * @wq: workqueue to use
236  * @dwork: delayable work to queue
237  * @delay: number of jiffies to wait before queueing
238  *
239  * Returns 0 if @work was already on a queue, non-zero otherwise.
240  */
241 int fastcall queue_delayed_work(struct workqueue_struct *wq,
242                         struct delayed_work *dwork, unsigned long delay)
243 {
244         int ret = 0;
245         struct timer_list *timer = &dwork->timer;
246         struct work_struct *work = &dwork->work;
247
248         timer_stats_timer_set_start_info(timer);
249         if (delay == 0)
250                 return queue_work(wq, work);
251
252         if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
253                 BUG_ON(timer_pending(timer));
254                 BUG_ON(!list_empty(&work->entry));
255
256                 /* This stores wq for the moment, for the timer_fn */
257                 set_wq_data(work, wq);
258                 timer->expires = jiffies + delay;
259                 timer->data = (unsigned long)dwork;
260                 timer->function = delayed_work_timer_fn;
261                 add_timer(timer);
262                 ret = 1;
263         }
264         return ret;
265 }
266 EXPORT_SYMBOL_GPL(queue_delayed_work);
267
268 /**
269  * queue_delayed_work_on - queue work on specific CPU after delay
270  * @cpu: CPU number to execute work on
271  * @wq: workqueue to use
272  * @dwork: work to queue
273  * @delay: number of jiffies to wait before queueing
274  *
275  * Returns 0 if @work was already on a queue, non-zero otherwise.
276  */
277 int queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
278                         struct delayed_work *dwork, unsigned long delay)
279 {
280         int ret = 0;
281         struct timer_list *timer = &dwork->timer;
282         struct work_struct *work = &dwork->work;
283
284         if (!test_and_set_bit(WORK_STRUCT_PENDING, work_data_bits(work))) {
285                 BUG_ON(timer_pending(timer));
286                 BUG_ON(!list_empty(&work->entry));
287
288                 /* This stores wq for the moment, for the timer_fn */
289                 set_wq_data(work, wq);
290                 timer->expires = jiffies + delay;
291                 timer->data = (unsigned long)dwork;
292                 timer->function = delayed_work_timer_fn;
293                 add_timer_on(timer, cpu);
294                 ret = 1;
295         }
296         return ret;
297 }
298 EXPORT_SYMBOL_GPL(queue_delayed_work_on);
299
300 static void run_workqueue(struct cpu_workqueue_struct *cwq)
301 {
302         unsigned long flags;
303
304         /*
305          * Keep taking off work from the queue until
306          * done.
307          */
308         spin_lock_irqsave(&cwq->lock, flags);
309         cwq->run_depth++;
310         if (cwq->run_depth > 3) {
311                 /* morton gets to eat his hat */
312                 printk("%s: recursion depth exceeded: %d\n",
313                         __FUNCTION__, cwq->run_depth);
314                 dump_stack();
315         }
316         while (!list_empty(&cwq->worklist)) {
317                 struct work_struct *work = list_entry(cwq->worklist.next,
318                                                 struct work_struct, entry);
319                 work_func_t f = work->func;
320
321                 list_del_init(cwq->worklist.next);
322                 spin_unlock_irqrestore(&cwq->lock, flags);
323
324                 BUG_ON(get_wq_data(work) != cwq);
325                 if (!test_bit(WORK_STRUCT_NOAUTOREL, work_data_bits(work)))
326                         work_release(work);
327                 f(work);
328
329                 if (unlikely(in_atomic() || lockdep_depth(current) > 0)) {
330                         printk(KERN_ERR "BUG: workqueue leaked lock or atomic: "
331                                         "%s/0x%08x/%d\n",
332                                         current->comm, preempt_count(),
333                                         current->pid);
334                         printk(KERN_ERR "    last function: ");
335                         print_symbol("%s\n", (unsigned long)f);
336                         debug_show_held_locks(current);
337                         dump_stack();
338                 }
339
340                 spin_lock_irqsave(&cwq->lock, flags);
341                 cwq->remove_sequence++;
342                 wake_up(&cwq->work_done);
343         }
344         cwq->run_depth--;
345         spin_unlock_irqrestore(&cwq->lock, flags);
346 }
347
348 static int worker_thread(void *__cwq)
349 {
350         struct cpu_workqueue_struct *cwq = __cwq;
351         DECLARE_WAITQUEUE(wait, current);
352         struct k_sigaction sa;
353         sigset_t blocked;
354
355         if (!cwq->freezeable)
356                 current->flags |= PF_NOFREEZE;
357
358         set_user_nice(current, -5);
359
360         /* Block and flush all signals */
361         sigfillset(&blocked);
362         sigprocmask(SIG_BLOCK, &blocked, NULL);
363         flush_signals(current);
364
365         /*
366          * We inherited MPOL_INTERLEAVE from the booting kernel.
367          * Set MPOL_DEFAULT to insure node local allocations.
368          */
369         numa_default_policy();
370
371         /* SIG_IGN makes children autoreap: see do_notify_parent(). */
372         sa.sa.sa_handler = SIG_IGN;
373         sa.sa.sa_flags = 0;
374         siginitset(&sa.sa.sa_mask, sigmask(SIGCHLD));
375         do_sigaction(SIGCHLD, &sa, (struct k_sigaction *)0);
376
377         set_current_state(TASK_INTERRUPTIBLE);
378         while (!kthread_should_stop()) {
379                 if (cwq->freezeable)
380                         try_to_freeze();
381
382                 add_wait_queue(&cwq->more_work, &wait);
383                 if (list_empty(&cwq->worklist))
384                         schedule();
385                 else
386                         __set_current_state(TASK_RUNNING);
387                 remove_wait_queue(&cwq->more_work, &wait);
388
389                 if (!list_empty(&cwq->worklist))
390                         run_workqueue(cwq);
391                 set_current_state(TASK_INTERRUPTIBLE);
392         }
393         __set_current_state(TASK_RUNNING);
394         return 0;
395 }
396
397 static void flush_cpu_workqueue(struct cpu_workqueue_struct *cwq)
398 {
399         if (cwq->thread == current) {
400                 /*
401                  * Probably keventd trying to flush its own queue. So simply run
402                  * it by hand rather than deadlocking.
403                  */
404                 run_workqueue(cwq);
405         } else {
406                 DEFINE_WAIT(wait);
407                 long sequence_needed;
408
409                 spin_lock_irq(&cwq->lock);
410                 sequence_needed = cwq->insert_sequence;
411
412                 while (sequence_needed - cwq->remove_sequence > 0) {
413                         prepare_to_wait(&cwq->work_done, &wait,
414                                         TASK_UNINTERRUPTIBLE);
415                         spin_unlock_irq(&cwq->lock);
416                         schedule();
417                         spin_lock_irq(&cwq->lock);
418                 }
419                 finish_wait(&cwq->work_done, &wait);
420                 spin_unlock_irq(&cwq->lock);
421         }
422 }
423
424 /**
425  * flush_workqueue - ensure that any scheduled work has run to completion.
426  * @wq: workqueue to flush
427  *
428  * Forces execution of the workqueue and blocks until its completion.
429  * This is typically used in driver shutdown handlers.
430  *
431  * This function will sample each workqueue's current insert_sequence number and
432  * will sleep until the head sequence is greater than or equal to that.  This
433  * means that we sleep until all works which were queued on entry have been
434  * handled, but we are not livelocked by new incoming ones.
435  *
436  * This function used to run the workqueues itself.  Now we just wait for the
437  * helper threads to do it.
438  */
439 void fastcall flush_workqueue(struct workqueue_struct *wq)
440 {
441         might_sleep();
442
443         if (is_single_threaded(wq)) {
444                 /* Always use first cpu's area. */
445                 flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, singlethread_cpu));
446         } else {
447                 int cpu;
448
449                 mutex_lock(&workqueue_mutex);
450                 for_each_online_cpu(cpu)
451                         flush_cpu_workqueue(per_cpu_ptr(wq->cpu_wq, cpu));
452                 mutex_unlock(&workqueue_mutex);
453         }
454 }
455 EXPORT_SYMBOL_GPL(flush_workqueue);
456
457 static struct task_struct *create_workqueue_thread(struct workqueue_struct *wq,
458                                                    int cpu, int freezeable)
459 {
460         struct cpu_workqueue_struct *cwq = per_cpu_ptr(wq->cpu_wq, cpu);
461         struct task_struct *p;
462
463         spin_lock_init(&cwq->lock);
464         cwq->wq = wq;
465         cwq->thread = NULL;
466         cwq->insert_sequence = 0;
467         cwq->remove_sequence = 0;
468         cwq->freezeable = freezeable;
469         INIT_LIST_HEAD(&cwq->worklist);
470         init_waitqueue_head(&cwq->more_work);
471         init_waitqueue_head(&cwq->work_done);
472
473         if (is_single_threaded(wq))
474                 p = kthread_create(worker_thread, cwq, "%s", wq->name);
475         else
476                 p = kthread_create(worker_thread, cwq, "%s/%d", wq->name, cpu);
477         if (IS_ERR(p))
478                 return NULL;
479         cwq->thread = p;
480         return p;
481 }
482
483 struct workqueue_struct *__create_workqueue(const char *name,
484                                             int singlethread, int freezeable)
485 {
486         int cpu, destroy = 0;
487         struct workqueue_struct *wq;
488         struct task_struct *p;
489
490         wq = kzalloc(sizeof(*wq), GFP_KERNEL);
491         if (!wq)
492                 return NULL;
493
494         wq->cpu_wq = alloc_percpu(struct cpu_workqueue_struct);
495         if (!wq->cpu_wq) {
496                 kfree(wq);
497                 return NULL;
498         }
499
500         wq->name = name;
501         mutex_lock(&workqueue_mutex);
502         if (singlethread) {
503                 INIT_LIST_HEAD(&wq->list);
504                 p = create_workqueue_thread(wq, singlethread_cpu, freezeable);
505                 if (!p)
506                         destroy = 1;
507                 else
508                         wake_up_process(p);
509         } else {
510                 list_add(&wq->list, &workqueues);
511                 for_each_online_cpu(cpu) {
512                         p = create_workqueue_thread(wq, cpu, freezeable);
513                         if (p) {
514                                 kthread_bind(p, cpu);
515                                 wake_up_process(p);
516                         } else
517                                 destroy = 1;
518                 }
519         }
520         mutex_unlock(&workqueue_mutex);
521
522         /*
523          * Was there any error during startup? If yes then clean up:
524          */
525         if (destroy) {
526                 destroy_workqueue(wq);
527                 wq = NULL;
528         }
529         return wq;
530 }
531 EXPORT_SYMBOL_GPL(__create_workqueue);
532
533 static void cleanup_workqueue_thread(struct workqueue_struct *wq, int cpu)
534 {
535         struct cpu_workqueue_struct *cwq;
536         unsigned long flags;
537         struct task_struct *p;
538
539         cwq = per_cpu_ptr(wq->cpu_wq, cpu);
540         spin_lock_irqsave(&cwq->lock, flags);
541         p = cwq->thread;
542         cwq->thread = NULL;
543         spin_unlock_irqrestore(&cwq->lock, flags);
544         if (p)
545                 kthread_stop(p);
546 }
547
548 /**
549  * destroy_workqueue - safely terminate a workqueue
550  * @wq: target workqueue
551  *
552  * Safely destroy a workqueue. All work currently pending will be done first.
553  */
554 void destroy_workqueue(struct workqueue_struct *wq)
555 {
556         int cpu;
557
558         flush_workqueue(wq);
559
560         /* We don't need the distraction of CPUs appearing and vanishing. */
561         mutex_lock(&workqueue_mutex);
562         if (is_single_threaded(wq))
563                 cleanup_workqueue_thread(wq, singlethread_cpu);
564         else {
565                 for_each_online_cpu(cpu)
566                         cleanup_workqueue_thread(wq, cpu);
567                 list_del(&wq->list);
568         }
569         mutex_unlock(&workqueue_mutex);
570         free_percpu(wq->cpu_wq);
571         kfree(wq);
572 }
573 EXPORT_SYMBOL_GPL(destroy_workqueue);
574
575 static struct workqueue_struct *keventd_wq;
576
577 /**
578  * schedule_work - put work task in global workqueue
579  * @work: job to be done
580  *
581  * This puts a job in the kernel-global workqueue.
582  */
583 int fastcall schedule_work(struct work_struct *work)
584 {
585         return queue_work(keventd_wq, work);
586 }
587 EXPORT_SYMBOL(schedule_work);
588
589 /**
590  * schedule_delayed_work - put work task in global workqueue after delay
591  * @dwork: job to be done
592  * @delay: number of jiffies to wait or 0 for immediate execution
593  *
594  * After waiting for a given time this puts a job in the kernel-global
595  * workqueue.
596  */
597 int fastcall schedule_delayed_work(struct delayed_work *dwork,
598                                         unsigned long delay)
599 {
600         timer_stats_timer_set_start_info(&dwork->timer);
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                 struct work_struct *work = per_cpu_ptr(works, cpu);
644
645                 INIT_WORK(work, func);
646                 set_bit(WORK_STRUCT_PENDING, work_data_bits(work));
647                 __queue_work(per_cpu_ptr(keventd_wq->cpu_wq, cpu), work);
648         }
649         mutex_unlock(&workqueue_mutex);
650         flush_workqueue(keventd_wq);
651         free_percpu(works);
652         return 0;
653 }
654
655 void flush_scheduled_work(void)
656 {
657         flush_workqueue(keventd_wq);
658 }
659 EXPORT_SYMBOL(flush_scheduled_work);
660
661 /**
662  * cancel_rearming_delayed_workqueue - reliably kill off a delayed work whose handler rearms the delayed work.
663  * @wq:   the controlling workqueue structure
664  * @dwork: the delayed work struct
665  */
666 void cancel_rearming_delayed_workqueue(struct workqueue_struct *wq,
667                                        struct delayed_work *dwork)
668 {
669         while (!cancel_delayed_work(dwork))
670                 flush_workqueue(wq);
671 }
672 EXPORT_SYMBOL(cancel_rearming_delayed_workqueue);
673
674 /**
675  * cancel_rearming_delayed_work - reliably kill off a delayed keventd 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