4 * Kernel scheduler and related syscalls
6 * Copyright (C) 1991-2002 Linus Torvalds
8 * 1996-12-23 Modified by Dave Grothe to fix bugs in semaphores and
9 * make semaphores SMP safe
10 * 1998-11-19 Implemented schedule_timeout() and related stuff
12 * 2002-01-04 New ultra-scalable O(1) scheduler by Ingo Molnar:
13 * hybrid priority-list and round-robin design with
14 * an array-switch method of distributing timeslices
15 * and per-CPU runqueues. Cleanups and useful suggestions
16 * by Davide Libenzi, preemptible kernel bits by Robert Love.
17 * 2003-09-03 Interactivity tuning by Con Kolivas.
18 * 2004-04-02 Scheduler domains code by Nick Piggin
22 #include <linux/module.h>
23 #include <linux/nmi.h>
24 #include <linux/init.h>
25 #include <asm/uaccess.h>
26 #include <linux/highmem.h>
27 #include <linux/smp_lock.h>
28 #include <asm/mmu_context.h>
29 #include <linux/interrupt.h>
30 #include <linux/capability.h>
31 #include <linux/completion.h>
32 #include <linux/kernel_stat.h>
33 #include <linux/security.h>
34 #include <linux/notifier.h>
35 #include <linux/profile.h>
36 #include <linux/suspend.h>
37 #include <linux/vmalloc.h>
38 #include <linux/blkdev.h>
39 #include <linux/delay.h>
40 #include <linux/smp.h>
41 #include <linux/threads.h>
42 #include <linux/timer.h>
43 #include <linux/rcupdate.h>
44 #include <linux/cpu.h>
45 #include <linux/cpuset.h>
46 #include <linux/percpu.h>
47 #include <linux/kthread.h>
48 #include <linux/seq_file.h>
49 #include <linux/syscalls.h>
50 #include <linux/times.h>
51 #include <linux/acct.h>
53 #include <asm/unistd.h>
55 #include <linux/vs_context.h>
56 #include <linux/vs_cvirt.h>
57 #include <linux/vs_sched.h>
60 * Convert user-nice values [ -20 ... 0 ... 19 ]
61 * to static priority [ MAX_RT_PRIO..MAX_PRIO-1 ],
64 #define NICE_TO_PRIO(nice) (MAX_RT_PRIO + (nice) + 20)
65 #define PRIO_TO_NICE(prio) ((prio) - MAX_RT_PRIO - 20)
66 #define TASK_NICE(p) PRIO_TO_NICE((p)->static_prio)
69 * 'User priority' is the nice value converted to something we
70 * can work with better when scaling various scheduler parameters,
71 * it's a [ 0 ... 39 ] range.
73 #define USER_PRIO(p) ((p)-MAX_RT_PRIO)
74 #define TASK_USER_PRIO(p) USER_PRIO((p)->static_prio)
75 #define MAX_USER_PRIO (USER_PRIO(MAX_PRIO))
78 * Some helpers for converting nanosecond timing to jiffy resolution
80 #define NS_TO_JIFFIES(TIME) ((TIME) / (1000000000 / HZ))
81 #define JIFFIES_TO_NS(TIME) ((TIME) * (1000000000 / HZ))
84 * These are the 'tuning knobs' of the scheduler:
86 * Minimum timeslice is 5 msecs (or 1 jiffy, whichever is larger),
87 * default timeslice is 100 msecs, maximum timeslice is 800 msecs.
88 * Timeslices get refilled after they expire.
90 #define MIN_TIMESLICE max(5 * HZ / 1000, 1)
91 #define DEF_TIMESLICE (100 * HZ / 1000)
92 #define ON_RUNQUEUE_WEIGHT 30
93 #define CHILD_PENALTY 95
94 #define PARENT_PENALTY 100
96 #define PRIO_BONUS_RATIO 25
97 #define MAX_BONUS (MAX_USER_PRIO * PRIO_BONUS_RATIO / 100)
98 #define INTERACTIVE_DELTA 2
99 #define MAX_SLEEP_AVG (DEF_TIMESLICE * MAX_BONUS)
100 #define STARVATION_LIMIT (MAX_SLEEP_AVG)
101 #define NS_MAX_SLEEP_AVG (JIFFIES_TO_NS(MAX_SLEEP_AVG))
104 * If a task is 'interactive' then we reinsert it in the active
105 * array after it has expired its current timeslice. (it will not
106 * continue to run immediately, it will still roundrobin with
107 * other interactive tasks.)
109 * This part scales the interactivity limit depending on niceness.
111 * We scale it linearly, offset by the INTERACTIVE_DELTA delta.
112 * Here are a few examples of different nice levels:
114 * TASK_INTERACTIVE(-20): [1,1,1,1,1,1,1,1,1,0,0]
115 * TASK_INTERACTIVE(-10): [1,1,1,1,1,1,1,0,0,0,0]
116 * TASK_INTERACTIVE( 0): [1,1,1,1,0,0,0,0,0,0,0]
117 * TASK_INTERACTIVE( 10): [1,1,0,0,0,0,0,0,0,0,0]
118 * TASK_INTERACTIVE( 19): [0,0,0,0,0,0,0,0,0,0,0]
120 * (the X axis represents the possible -5 ... 0 ... +5 dynamic
121 * priority range a task can explore, a value of '1' means the
122 * task is rated interactive.)
124 * Ie. nice +19 tasks can never get 'interactive' enough to be
125 * reinserted into the active array. And only heavily CPU-hog nice -20
126 * tasks will be expired. Default nice 0 tasks are somewhere between,
127 * it takes some effort for them to get interactive, but it's not
131 #define CURRENT_BONUS(p) \
132 (NS_TO_JIFFIES((p)->sleep_avg) * MAX_BONUS / \
135 #define GRANULARITY (10 * HZ / 1000 ? : 1)
138 #define TIMESLICE_GRANULARITY(p) (GRANULARITY * \
139 (1 << (((MAX_BONUS - CURRENT_BONUS(p)) ? : 1) - 1)) * \
142 #define TIMESLICE_GRANULARITY(p) (GRANULARITY * \
143 (1 << (((MAX_BONUS - CURRENT_BONUS(p)) ? : 1) - 1)))
146 #define SCALE(v1,v1_max,v2_max) \
147 (v1) * (v2_max) / (v1_max)
150 (SCALE(TASK_NICE(p), 40, MAX_BONUS) + INTERACTIVE_DELTA)
152 #define TASK_INTERACTIVE(p) \
153 ((p)->prio <= (p)->static_prio - DELTA(p))
155 #define INTERACTIVE_SLEEP(p) \
156 (JIFFIES_TO_NS(MAX_SLEEP_AVG * \
157 (MAX_BONUS / 2 + DELTA((p)) + 1) / MAX_BONUS - 1))
159 #define TASK_PREEMPTS_CURR(p, rq) \
160 ((p)->prio < (rq)->curr->prio)
163 * task_timeslice() scales user-nice values [ -20 ... 0 ... 19 ]
164 * to time slice values: [800ms ... 100ms ... 5ms]
166 * The higher a thread's priority, the bigger timeslices
167 * it gets during one round of execution. But even the lowest
168 * priority thread gets MIN_TIMESLICE worth of execution time.
171 #define SCALE_PRIO(x, prio) \
172 max(x * (MAX_PRIO - prio) / (MAX_USER_PRIO/2), MIN_TIMESLICE)
174 static unsigned int task_timeslice(task_t *p)
176 if (p->static_prio < NICE_TO_PRIO(0))
177 return SCALE_PRIO(DEF_TIMESLICE*4, p->static_prio);
179 return SCALE_PRIO(DEF_TIMESLICE, p->static_prio);
181 #define task_hot(p, now, sd) ((long long) ((now) - (p)->last_ran) \
182 < (long long) (sd)->cache_hot_time)
185 * These are the runqueue data structures:
188 #define BITMAP_SIZE ((((MAX_PRIO+1+7)/8)+sizeof(long)-1)/sizeof(long))
190 typedef struct runqueue runqueue_t;
193 unsigned int nr_active;
194 unsigned long bitmap[BITMAP_SIZE];
195 struct list_head queue[MAX_PRIO];
199 * This is the main, per-CPU runqueue data structure.
201 * Locking rule: those places that want to lock multiple runqueues
202 * (such as the load balancing or the thread migration code), lock
203 * acquire operations must be ordered by ascending &runqueue.
209 * nr_running and cpu_load should be in the same cacheline because
210 * remote CPUs use both these fields when doing load calculation.
212 unsigned long nr_running;
214 unsigned long cpu_load[3];
216 unsigned long long nr_switches;
219 * This is part of a global counter where only the total sum
220 * over all CPUs matters. A task can increase this counter on
221 * one CPU and if it got migrated afterwards it may decrease
222 * it on another CPU. Always updated under the runqueue lock:
224 unsigned long nr_uninterruptible;
226 unsigned long expired_timestamp;
227 unsigned long long timestamp_last_tick;
229 struct mm_struct *prev_mm;
230 prio_array_t *active, *expired, arrays[2];
231 int best_expired_prio;
235 struct sched_domain *sd;
237 /* For active balancing */
241 task_t *migration_thread;
242 struct list_head migration_queue;
245 #ifdef CONFIG_VSERVER_HARDCPU
246 struct list_head hold_queue;
250 #ifdef CONFIG_SCHEDSTATS
252 struct sched_info rq_sched_info;
254 /* sys_sched_yield() stats */
255 unsigned long yld_exp_empty;
256 unsigned long yld_act_empty;
257 unsigned long yld_both_empty;
258 unsigned long yld_cnt;
260 /* schedule() stats */
261 unsigned long sched_switch;
262 unsigned long sched_cnt;
263 unsigned long sched_goidle;
265 /* try_to_wake_up() stats */
266 unsigned long ttwu_cnt;
267 unsigned long ttwu_local;
271 static DEFINE_PER_CPU(struct runqueue, runqueues);
274 * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
275 * See detach_destroy_domains: synchronize_sched for details.
277 * The domain tree of any CPU may only be accessed from within
278 * preempt-disabled sections.
280 #define for_each_domain(cpu, domain) \
281 for (domain = rcu_dereference(cpu_rq(cpu)->sd); domain; domain = domain->parent)
283 #define cpu_rq(cpu) (&per_cpu(runqueues, (cpu)))
284 #define this_rq() (&__get_cpu_var(runqueues))
285 #define task_rq(p) cpu_rq(task_cpu(p))
286 #define cpu_curr(cpu) (cpu_rq(cpu)->curr)
288 #ifndef prepare_arch_switch
289 # define prepare_arch_switch(next) do { } while (0)
291 #ifndef finish_arch_switch
292 # define finish_arch_switch(prev) do { } while (0)
295 #ifndef __ARCH_WANT_UNLOCKED_CTXSW
296 static inline int task_running(runqueue_t *rq, task_t *p)
298 return rq->curr == p;
301 static inline void prepare_lock_switch(runqueue_t *rq, task_t *next)
305 static inline void finish_lock_switch(runqueue_t *rq, task_t *prev)
307 #ifdef CONFIG_DEBUG_SPINLOCK
308 /* this is a valid case when another task releases the spinlock */
309 rq->lock.owner = current;
311 spin_unlock_irq(&rq->lock);
314 #else /* __ARCH_WANT_UNLOCKED_CTXSW */
315 static inline int task_running(runqueue_t *rq, task_t *p)
320 return rq->curr == p;
324 static inline void prepare_lock_switch(runqueue_t *rq, task_t *next)
328 * We can optimise this out completely for !SMP, because the
329 * SMP rebalancing from interrupt is the only thing that cares
334 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
335 spin_unlock_irq(&rq->lock);
337 spin_unlock(&rq->lock);
341 static inline void finish_lock_switch(runqueue_t *rq, task_t *prev)
345 * After ->oncpu is cleared, the task can be moved to a different CPU.
346 * We must ensure this doesn't happen until the switch is completely
352 #ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
356 #endif /* __ARCH_WANT_UNLOCKED_CTXSW */
359 * task_rq_lock - lock the runqueue a given task resides on and disable
360 * interrupts. Note the ordering: we can safely lookup the task_rq without
361 * explicitly disabling preemption.
363 static inline runqueue_t *task_rq_lock(task_t *p, unsigned long *flags)
369 local_irq_save(*flags);
371 spin_lock(&rq->lock);
372 if (unlikely(rq != task_rq(p))) {
373 spin_unlock_irqrestore(&rq->lock, *flags);
374 goto repeat_lock_task;
379 static inline void task_rq_unlock(runqueue_t *rq, unsigned long *flags)
382 spin_unlock_irqrestore(&rq->lock, *flags);
385 #ifdef CONFIG_SCHEDSTATS
387 * bump this up when changing the output format or the meaning of an existing
388 * format, so that tools can adapt (or abort)
390 #define SCHEDSTAT_VERSION 12
392 static int show_schedstat(struct seq_file *seq, void *v)
396 seq_printf(seq, "version %d\n", SCHEDSTAT_VERSION);
397 seq_printf(seq, "timestamp %lu\n", jiffies);
398 for_each_online_cpu(cpu) {
399 runqueue_t *rq = cpu_rq(cpu);
401 struct sched_domain *sd;
405 /* runqueue-specific stats */
407 "cpu%d %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
408 cpu, rq->yld_both_empty,
409 rq->yld_act_empty, rq->yld_exp_empty, rq->yld_cnt,
410 rq->sched_switch, rq->sched_cnt, rq->sched_goidle,
411 rq->ttwu_cnt, rq->ttwu_local,
412 rq->rq_sched_info.cpu_time,
413 rq->rq_sched_info.run_delay, rq->rq_sched_info.pcnt);
415 seq_printf(seq, "\n");
418 /* domain-specific stats */
420 for_each_domain(cpu, sd) {
421 enum idle_type itype;
422 char mask_str[NR_CPUS];
424 cpumask_scnprintf(mask_str, NR_CPUS, sd->span);
425 seq_printf(seq, "domain%d %s", dcnt++, mask_str);
426 for (itype = SCHED_IDLE; itype < MAX_IDLE_TYPES;
428 seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu",
430 sd->lb_balanced[itype],
431 sd->lb_failed[itype],
432 sd->lb_imbalance[itype],
433 sd->lb_gained[itype],
434 sd->lb_hot_gained[itype],
435 sd->lb_nobusyq[itype],
436 sd->lb_nobusyg[itype]);
438 seq_printf(seq, " %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu\n",
439 sd->alb_cnt, sd->alb_failed, sd->alb_pushed,
440 sd->sbe_cnt, sd->sbe_balanced, sd->sbe_pushed,
441 sd->sbf_cnt, sd->sbf_balanced, sd->sbf_pushed,
442 sd->ttwu_wake_remote, sd->ttwu_move_affine, sd->ttwu_move_balance);
450 static int schedstat_open(struct inode *inode, struct file *file)
452 unsigned int size = PAGE_SIZE * (1 + num_online_cpus() / 32);
453 char *buf = kmalloc(size, GFP_KERNEL);
459 res = single_open(file, show_schedstat, NULL);
461 m = file->private_data;
469 struct file_operations proc_schedstat_operations = {
470 .open = schedstat_open,
473 .release = single_release,
476 # define schedstat_inc(rq, field) do { (rq)->field++; } while (0)
477 # define schedstat_add(rq, field, amt) do { (rq)->field += (amt); } while (0)
478 #else /* !CONFIG_SCHEDSTATS */
479 # define schedstat_inc(rq, field) do { } while (0)
480 # define schedstat_add(rq, field, amt) do { } while (0)
484 * rq_lock - lock a given runqueue and disable interrupts.
486 static inline runqueue_t *this_rq_lock(void)
493 spin_lock(&rq->lock);
498 #ifdef CONFIG_SCHEDSTATS
500 * Called when a process is dequeued from the active array and given
501 * the cpu. We should note that with the exception of interactive
502 * tasks, the expired queue will become the active queue after the active
503 * queue is empty, without explicitly dequeuing and requeuing tasks in the
504 * expired queue. (Interactive tasks may be requeued directly to the
505 * active queue, thus delaying tasks in the expired queue from running;
506 * see scheduler_tick()).
508 * This function is only called from sched_info_arrive(), rather than
509 * dequeue_task(). Even though a task may be queued and dequeued multiple
510 * times as it is shuffled about, we're really interested in knowing how
511 * long it was from the *first* time it was queued to the time that it
514 static inline void sched_info_dequeued(task_t *t)
516 t->sched_info.last_queued = 0;
520 * Called when a task finally hits the cpu. We can now calculate how
521 * long it was waiting to run. We also note when it began so that we
522 * can keep stats on how long its timeslice is.
524 static void sched_info_arrive(task_t *t)
526 unsigned long now = jiffies, diff = 0;
527 struct runqueue *rq = task_rq(t);
529 if (t->sched_info.last_queued)
530 diff = now - t->sched_info.last_queued;
531 sched_info_dequeued(t);
532 t->sched_info.run_delay += diff;
533 t->sched_info.last_arrival = now;
534 t->sched_info.pcnt++;
539 rq->rq_sched_info.run_delay += diff;
540 rq->rq_sched_info.pcnt++;
544 * Called when a process is queued into either the active or expired
545 * array. The time is noted and later used to determine how long we
546 * had to wait for us to reach the cpu. Since the expired queue will
547 * become the active queue after active queue is empty, without dequeuing
548 * and requeuing any tasks, we are interested in queuing to either. It
549 * is unusual but not impossible for tasks to be dequeued and immediately
550 * requeued in the same or another array: this can happen in sched_yield(),
551 * set_user_nice(), and even load_balance() as it moves tasks from runqueue
554 * This function is only called from enqueue_task(), but also only updates
555 * the timestamp if it is already not set. It's assumed that
556 * sched_info_dequeued() will clear that stamp when appropriate.
558 static inline void sched_info_queued(task_t *t)
560 if (!t->sched_info.last_queued)
561 t->sched_info.last_queued = jiffies;
565 * Called when a process ceases being the active-running process, either
566 * voluntarily or involuntarily. Now we can calculate how long we ran.
568 static inline void sched_info_depart(task_t *t)
570 struct runqueue *rq = task_rq(t);
571 unsigned long diff = jiffies - t->sched_info.last_arrival;
573 t->sched_info.cpu_time += diff;
576 rq->rq_sched_info.cpu_time += diff;
580 * Called when tasks are switched involuntarily due, typically, to expiring
581 * their time slice. (This may also be called when switching to or from
582 * the idle task.) We are only called when prev != next.
584 static inline void sched_info_switch(task_t *prev, task_t *next)
586 struct runqueue *rq = task_rq(prev);
589 * prev now departs the cpu. It's not interesting to record
590 * stats about how efficient we were at scheduling the idle
593 if (prev != rq->idle)
594 sched_info_depart(prev);
596 if (next != rq->idle)
597 sched_info_arrive(next);
600 #define sched_info_queued(t) do { } while (0)
601 #define sched_info_switch(t, next) do { } while (0)
602 #endif /* CONFIG_SCHEDSTATS */
605 * Adding/removing a task to/from a priority array:
607 static void dequeue_task(struct task_struct *p, prio_array_t *array)
609 BUG_ON(p->state & TASK_ONHOLD);
611 list_del(&p->run_list);
612 if (list_empty(array->queue + p->prio))
613 __clear_bit(p->prio, array->bitmap);
616 static void enqueue_task(struct task_struct *p, prio_array_t *array)
618 BUG_ON(p->state & TASK_ONHOLD);
619 sched_info_queued(p);
620 list_add_tail(&p->run_list, array->queue + p->prio);
621 __set_bit(p->prio, array->bitmap);
627 * Put task to the end of the run list without the overhead of dequeue
628 * followed by enqueue.
630 static void requeue_task(struct task_struct *p, prio_array_t *array)
632 BUG_ON(p->state & TASK_ONHOLD);
633 list_move_tail(&p->run_list, array->queue + p->prio);
636 static inline void enqueue_task_head(struct task_struct *p, prio_array_t *array)
638 BUG_ON(p->state & TASK_ONHOLD);
639 list_add(&p->run_list, array->queue + p->prio);
640 __set_bit(p->prio, array->bitmap);
646 * effective_prio - return the priority that is based on the static
647 * priority but is modified by bonuses/penalties.
649 * We scale the actual sleep average [0 .... MAX_SLEEP_AVG]
650 * into the -5 ... 0 ... +5 bonus/penalty range.
652 * We use 25% of the full 0...39 priority range so that:
654 * 1) nice +19 interactive tasks do not preempt nice 0 CPU hogs.
655 * 2) nice -20 CPU hogs do not get preempted by nice 0 tasks.
657 * Both properties are important to certain workloads.
659 static int effective_prio(task_t *p)
667 bonus = CURRENT_BONUS(p) - MAX_BONUS / 2;
669 prio = p->static_prio - bonus;
671 if ((vxi = p->vx_info) &&
672 vx_info_flags(vxi, VXF_SCHED_PRIO, 0))
673 prio += vx_effective_vavavoom(vxi, MAX_USER_PRIO);
675 if (prio < MAX_RT_PRIO)
677 if (prio > MAX_PRIO-1)
683 * __activate_task - move a task to the runqueue.
685 static inline void __activate_task(task_t *p, runqueue_t *rq)
687 enqueue_task(p, rq->active);
692 * __activate_idle_task - move idle task to the _front_ of runqueue.
694 static inline void __activate_idle_task(task_t *p, runqueue_t *rq)
696 enqueue_task_head(p, rq->active);
700 static int recalc_task_prio(task_t *p, unsigned long long now)
702 /* Caller must always ensure 'now >= p->timestamp' */
703 unsigned long long __sleep_time = now - p->timestamp;
704 unsigned long sleep_time;
706 if (unlikely(p->policy == SCHED_BATCH))
709 if (__sleep_time > NS_MAX_SLEEP_AVG)
710 sleep_time = NS_MAX_SLEEP_AVG;
712 sleep_time = (unsigned long)__sleep_time;
715 if (likely(sleep_time > 0)) {
717 * User tasks that sleep a long time are categorised as
718 * idle and will get just interactive status to stay active &
719 * prevent them suddenly becoming cpu hogs and starving
722 if (p->mm && p->activated != -1 &&
723 sleep_time > INTERACTIVE_SLEEP(p)) {
724 p->sleep_avg = JIFFIES_TO_NS(MAX_SLEEP_AVG -
728 * The lower the sleep avg a task has the more
729 * rapidly it will rise with sleep time.
731 sleep_time *= (MAX_BONUS - CURRENT_BONUS(p)) ? : 1;
734 * Tasks waking from uninterruptible sleep are
735 * limited in their sleep_avg rise as they
736 * are likely to be waiting on I/O
738 if (p->activated == -1 && p->mm) {
739 if (p->sleep_avg >= INTERACTIVE_SLEEP(p))
741 else if (p->sleep_avg + sleep_time >=
742 INTERACTIVE_SLEEP(p)) {
743 p->sleep_avg = INTERACTIVE_SLEEP(p);
749 * This code gives a bonus to interactive tasks.
751 * The boost works by updating the 'average sleep time'
752 * value here, based on ->timestamp. The more time a
753 * task spends sleeping, the higher the average gets -
754 * and the higher the priority boost gets as well.
756 p->sleep_avg += sleep_time;
758 if (p->sleep_avg > NS_MAX_SLEEP_AVG)
759 p->sleep_avg = NS_MAX_SLEEP_AVG;
763 return effective_prio(p);
767 * activate_task - move a task to the runqueue and do priority recalculation
769 * Update all the scheduling statistics stuff. (sleep average
770 * calculation, priority modifiers, etc.)
772 static void activate_task(task_t *p, runqueue_t *rq, int local)
774 unsigned long long now;
779 /* Compensate for drifting sched_clock */
780 runqueue_t *this_rq = this_rq();
781 now = (now - this_rq->timestamp_last_tick)
782 + rq->timestamp_last_tick;
787 p->prio = recalc_task_prio(p, now);
790 * This checks to make sure it's not an uninterruptible task
791 * that is now waking up.
795 * Tasks which were woken up by interrupts (ie. hw events)
796 * are most likely of interactive nature. So we give them
797 * the credit of extending their sleep time to the period
798 * of time they spend on the runqueue, waiting for execution
799 * on a CPU, first time around:
805 * Normal first-time wakeups get a credit too for
806 * on-runqueue time, but it will be weighted down:
814 __activate_task(p, rq);
818 * deactivate_task - remove a task from the runqueue.
820 static void __deactivate_task(struct task_struct *p, runqueue_t *rq)
823 dequeue_task(p, p->array);
828 void deactivate_task(struct task_struct *p, runqueue_t *rq)
830 vx_deactivate_task(p);
831 __deactivate_task(p, rq);
835 #ifdef CONFIG_VSERVER_HARDCPU
837 * vx_hold_task - put a task on the hold queue
840 void vx_hold_task(struct vx_info *vxi,
841 struct task_struct *p, runqueue_t *rq)
843 __deactivate_task(p, rq);
844 p->state |= TASK_ONHOLD;
845 /* a new one on hold */
847 list_add_tail(&p->run_list, &rq->hold_queue);
851 * vx_unhold_task - put a task back to the runqueue
854 void vx_unhold_task(struct vx_info *vxi,
855 struct task_struct *p, runqueue_t *rq)
857 list_del(&p->run_list);
858 /* one less waiting */
860 p->state &= ~TASK_ONHOLD;
861 enqueue_task(p, rq->expired);
864 if (p->static_prio < rq->best_expired_prio)
865 rq->best_expired_prio = p->static_prio;
869 void vx_hold_task(struct vx_info *vxi,
870 struct task_struct *p, runqueue_t *rq)
876 void vx_unhold_task(struct vx_info *vxi,
877 struct task_struct *p, runqueue_t *rq)
881 #endif /* CONFIG_VSERVER_HARDCPU */
885 * resched_task - mark a task 'to be rescheduled now'.
887 * On UP this means the setting of the need_resched flag, on SMP it
888 * might also involve a cross-CPU call to trigger the scheduler on
892 static void resched_task(task_t *p)
896 assert_spin_locked(&task_rq(p)->lock);
898 if (unlikely(test_tsk_thread_flag(p, TIF_NEED_RESCHED)))
901 set_tsk_thread_flag(p, TIF_NEED_RESCHED);
904 if (cpu == smp_processor_id())
907 /* NEED_RESCHED must be visible before we test POLLING_NRFLAG */
909 if (!test_tsk_thread_flag(p, TIF_POLLING_NRFLAG))
910 smp_send_reschedule(cpu);
913 static inline void resched_task(task_t *p)
915 assert_spin_locked(&task_rq(p)->lock);
916 set_tsk_need_resched(p);
921 * task_curr - is this task currently executing on a CPU?
922 * @p: the task in question.
924 inline int task_curr(const task_t *p)
926 return cpu_curr(task_cpu(p)) == p;
931 struct list_head list;
936 struct completion done;
940 * The task's runqueue lock must be held.
941 * Returns true if you have to wait for migration thread.
943 static int migrate_task(task_t *p, int dest_cpu, migration_req_t *req)
945 runqueue_t *rq = task_rq(p);
948 * If the task is not on a runqueue (and not running), then
949 * it is sufficient to simply update the task's cpu field.
951 if (!p->array && !task_running(rq, p)) {
952 set_task_cpu(p, dest_cpu);
956 init_completion(&req->done);
958 req->dest_cpu = dest_cpu;
959 list_add(&req->list, &rq->migration_queue);
964 * wait_task_inactive - wait for a thread to unschedule.
966 * The caller must ensure that the task *will* unschedule sometime soon,
967 * else this function might spin for a *long* time. This function can't
968 * be called with interrupts off, or it may introduce deadlock with
969 * smp_call_function() if an IPI is sent by the same process we are
970 * waiting to become inactive.
972 void wait_task_inactive(task_t *p)
979 rq = task_rq_lock(p, &flags);
980 /* Must be off runqueue entirely, not preempted. */
981 if (unlikely(p->array || task_running(rq, p))) {
982 /* If it's preempted, we yield. It could be a while. */
983 preempted = !task_running(rq, p);
984 task_rq_unlock(rq, &flags);
990 task_rq_unlock(rq, &flags);
994 * kick_process - kick a running thread to enter/exit the kernel
995 * @p: the to-be-kicked thread
997 * Cause a process which is running on another CPU to enter
998 * kernel-mode, without any delay. (to get signals handled.)
1000 * NOTE: this function doesnt have to take the runqueue lock,
1001 * because all it wants to ensure is that the remote task enters
1002 * the kernel. If the IPI races and the task has been migrated
1003 * to another CPU then no harm is done and the purpose has been
1006 void kick_process(task_t *p)
1012 if ((cpu != smp_processor_id()) && task_curr(p))
1013 smp_send_reschedule(cpu);
1018 * Return a low guess at the load of a migration-source cpu.
1020 * We want to under-estimate the load of migration sources, to
1021 * balance conservatively.
1023 static inline unsigned long source_load(int cpu, int type)
1025 runqueue_t *rq = cpu_rq(cpu);
1026 unsigned long load_now = rq->nr_running * SCHED_LOAD_SCALE;
1030 return min(rq->cpu_load[type-1], load_now);
1034 * Return a high guess at the load of a migration-target cpu
1036 static inline unsigned long target_load(int cpu, int type)
1038 runqueue_t *rq = cpu_rq(cpu);
1039 unsigned long load_now = rq->nr_running * SCHED_LOAD_SCALE;
1043 return max(rq->cpu_load[type-1], load_now);
1047 * find_idlest_group finds and returns the least busy CPU group within the
1050 static struct sched_group *
1051 find_idlest_group(struct sched_domain *sd, struct task_struct *p, int this_cpu)
1053 struct sched_group *idlest = NULL, *this = NULL, *group = sd->groups;
1054 unsigned long min_load = ULONG_MAX, this_load = 0;
1055 int load_idx = sd->forkexec_idx;
1056 int imbalance = 100 + (sd->imbalance_pct-100)/2;
1059 unsigned long load, avg_load;
1063 /* Skip over this group if it has no CPUs allowed */
1064 if (!cpus_intersects(group->cpumask, p->cpus_allowed))
1067 local_group = cpu_isset(this_cpu, group->cpumask);
1069 /* Tally up the load of all CPUs in the group */
1072 for_each_cpu_mask(i, group->cpumask) {
1073 /* Bias balancing toward cpus of our domain */
1075 load = source_load(i, load_idx);
1077 load = target_load(i, load_idx);
1082 /* Adjust by relative CPU power of the group */
1083 avg_load = (avg_load * SCHED_LOAD_SCALE) / group->cpu_power;
1086 this_load = avg_load;
1088 } else if (avg_load < min_load) {
1089 min_load = avg_load;
1093 group = group->next;
1094 } while (group != sd->groups);
1096 if (!idlest || 100*this_load < imbalance*min_load)
1102 * find_idlest_queue - find the idlest runqueue among the cpus in group.
1105 find_idlest_cpu(struct sched_group *group, struct task_struct *p, int this_cpu)
1108 unsigned long load, min_load = ULONG_MAX;
1112 /* Traverse only the allowed CPUs */
1113 cpus_and(tmp, group->cpumask, p->cpus_allowed);
1115 for_each_cpu_mask(i, tmp) {
1116 load = source_load(i, 0);
1118 if (load < min_load || (load == min_load && i == this_cpu)) {
1128 * sched_balance_self: balance the current task (running on cpu) in domains
1129 * that have the 'flag' flag set. In practice, this is SD_BALANCE_FORK and
1132 * Balance, ie. select the least loaded group.
1134 * Returns the target CPU number, or the same CPU if no balancing is needed.
1136 * preempt must be disabled.
1138 static int sched_balance_self(int cpu, int flag)
1140 struct task_struct *t = current;
1141 struct sched_domain *tmp, *sd = NULL;
1143 for_each_domain(cpu, tmp)
1144 if (tmp->flags & flag)
1149 struct sched_group *group;
1154 group = find_idlest_group(sd, t, cpu);
1158 new_cpu = find_idlest_cpu(group, t, cpu);
1159 if (new_cpu == -1 || new_cpu == cpu)
1162 /* Now try balancing at a lower domain level */
1166 weight = cpus_weight(span);
1167 for_each_domain(cpu, tmp) {
1168 if (weight <= cpus_weight(tmp->span))
1170 if (tmp->flags & flag)
1173 /* while loop will break here if sd == NULL */
1179 #endif /* CONFIG_SMP */
1182 * wake_idle() will wake a task on an idle cpu if task->cpu is
1183 * not idle and an idle cpu is available. The span of cpus to
1184 * search starts with cpus closest then further out as needed,
1185 * so we always favor a closer, idle cpu.
1187 * Returns the CPU we should wake onto.
1189 #if defined(ARCH_HAS_SCHED_WAKE_IDLE)
1190 static int wake_idle(int cpu, task_t *p)
1193 struct sched_domain *sd;
1199 for_each_domain(cpu, sd) {
1200 if (sd->flags & SD_WAKE_IDLE) {
1201 cpus_and(tmp, sd->span, p->cpus_allowed);
1202 for_each_cpu_mask(i, tmp) {
1213 static inline int wake_idle(int cpu, task_t *p)
1220 * try_to_wake_up - wake up a thread
1221 * @p: the to-be-woken-up thread
1222 * @state: the mask of task states that can be woken
1223 * @sync: do a synchronous wakeup?
1225 * Put it on the run-queue if it's not already there. The "current"
1226 * thread is always on the run-queue (except when the actual
1227 * re-schedule is in progress), and as such you're allowed to do
1228 * the simpler "current->state = TASK_RUNNING" to mark yourself
1229 * runnable without the overhead of this.
1231 * returns failure only if the task is already active.
1233 static int try_to_wake_up(task_t *p, unsigned int state, int sync)
1235 int cpu, this_cpu, success = 0;
1236 unsigned long flags;
1240 unsigned long load, this_load;
1241 struct sched_domain *sd, *this_sd = NULL;
1245 rq = task_rq_lock(p, &flags);
1246 old_state = p->state;
1248 /* we need to unhold suspended tasks */
1249 if (old_state & TASK_ONHOLD) {
1250 vx_unhold_task(p->vx_info, p, rq);
1251 old_state = p->state;
1253 if (!(old_state & state))
1260 this_cpu = smp_processor_id();
1263 if (unlikely(task_running(rq, p)))
1268 schedstat_inc(rq, ttwu_cnt);
1269 if (cpu == this_cpu) {
1270 schedstat_inc(rq, ttwu_local);
1274 for_each_domain(this_cpu, sd) {
1275 if (cpu_isset(cpu, sd->span)) {
1276 schedstat_inc(sd, ttwu_wake_remote);
1282 if (unlikely(!cpu_isset(this_cpu, p->cpus_allowed)))
1286 * Check for affine wakeup and passive balancing possibilities.
1289 int idx = this_sd->wake_idx;
1290 unsigned int imbalance;
1292 imbalance = 100 + (this_sd->imbalance_pct - 100) / 2;
1294 load = source_load(cpu, idx);
1295 this_load = target_load(this_cpu, idx);
1297 new_cpu = this_cpu; /* Wake to this CPU if we can */
1299 if (this_sd->flags & SD_WAKE_AFFINE) {
1300 unsigned long tl = this_load;
1302 * If sync wakeup then subtract the (maximum possible)
1303 * effect of the currently running task from the load
1304 * of the current CPU:
1307 tl -= SCHED_LOAD_SCALE;
1310 tl + target_load(cpu, idx) <= SCHED_LOAD_SCALE) ||
1311 100*(tl + SCHED_LOAD_SCALE) <= imbalance*load) {
1313 * This domain has SD_WAKE_AFFINE and
1314 * p is cache cold in this domain, and
1315 * there is no bad imbalance.
1317 schedstat_inc(this_sd, ttwu_move_affine);
1323 * Start passive balancing when half the imbalance_pct
1326 if (this_sd->flags & SD_WAKE_BALANCE) {
1327 if (imbalance*this_load <= 100*load) {
1328 schedstat_inc(this_sd, ttwu_move_balance);
1334 new_cpu = cpu; /* Could not wake to this_cpu. Wake to cpu instead */
1336 new_cpu = wake_idle(new_cpu, p);
1337 if (new_cpu != cpu) {
1338 set_task_cpu(p, new_cpu);
1339 task_rq_unlock(rq, &flags);
1340 /* might preempt at this point */
1341 rq = task_rq_lock(p, &flags);
1342 old_state = p->state;
1343 if (!(old_state & state))
1348 this_cpu = smp_processor_id();
1353 #endif /* CONFIG_SMP */
1354 if (old_state == TASK_UNINTERRUPTIBLE) {
1355 rq->nr_uninterruptible--;
1357 * Tasks on involuntary sleep don't earn
1358 * sleep_avg beyond just interactive state.
1364 * Tasks that have marked their sleep as noninteractive get
1365 * woken up without updating their sleep average. (i.e. their
1366 * sleep is handled in a priority-neutral manner, no priority
1367 * boost and no penalty.)
1369 if (old_state & TASK_NONINTERACTIVE) {
1370 vx_activate_task(p);
1371 __activate_task(p, rq);
1373 activate_task(p, rq, cpu == this_cpu);
1375 /* this is to get the accounting behind the load update */
1376 if (old_state & TASK_UNINTERRUPTIBLE)
1377 vx_uninterruptible_dec(p);
1380 * Sync wakeups (i.e. those types of wakeups where the waker
1381 * has indicated that it will leave the CPU in short order)
1382 * don't trigger a preemption, if the woken up task will run on
1383 * this cpu. (in this case the 'I will reschedule' promise of
1384 * the waker guarantees that the freshly woken up task is going
1385 * to be considered on this CPU.)
1387 if (!sync || cpu != this_cpu) {
1388 if (TASK_PREEMPTS_CURR(p, rq))
1389 resched_task(rq->curr);
1394 p->state = TASK_RUNNING;
1396 task_rq_unlock(rq, &flags);
1401 int fastcall wake_up_process(task_t *p)
1403 return try_to_wake_up(p, TASK_STOPPED | TASK_TRACED |
1404 TASK_INTERRUPTIBLE | TASK_UNINTERRUPTIBLE, 0);
1407 EXPORT_SYMBOL(wake_up_process);
1409 int fastcall wake_up_state(task_t *p, unsigned int state)
1411 return try_to_wake_up(p, state, 0);
1415 * Perform scheduler related setup for a newly forked process p.
1416 * p is forked by current.
1418 void fastcall sched_fork(task_t *p, int clone_flags)
1420 int cpu = get_cpu();
1423 cpu = sched_balance_self(cpu, SD_BALANCE_FORK);
1425 set_task_cpu(p, cpu);
1428 * We mark the process as running here, but have not actually
1429 * inserted it onto the runqueue yet. This guarantees that
1430 * nobody will actually run it, and a signal or other external
1431 * event cannot wake it up and insert it on the runqueue either.
1433 p->state = TASK_RUNNING;
1434 INIT_LIST_HEAD(&p->run_list);
1436 #ifdef CONFIG_SCHEDSTATS
1437 memset(&p->sched_info, 0, sizeof(p->sched_info));
1439 #if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
1442 #ifdef CONFIG_PREEMPT
1443 /* Want to start with kernel preemption disabled. */
1444 task_thread_info(p)->preempt_count = 1;
1447 * Share the timeslice between parent and child, thus the
1448 * total amount of pending timeslices in the system doesn't change,
1449 * resulting in more scheduling fairness.
1451 local_irq_disable();
1452 p->time_slice = (current->time_slice + 1) >> 1;
1454 * The remainder of the first timeslice might be recovered by
1455 * the parent if the child exits early enough.
1457 p->first_time_slice = 1;
1458 current->time_slice >>= 1;
1459 p->timestamp = sched_clock();
1460 if (unlikely(!current->time_slice)) {
1462 * This case is rare, it happens when the parent has only
1463 * a single jiffy left from its timeslice. Taking the
1464 * runqueue lock is not a problem.
1466 current->time_slice = 1;
1474 * wake_up_new_task - wake up a newly created task for the first time.
1476 * This function will do some initial scheduler statistics housekeeping
1477 * that must be done for every newly created context, then puts the task
1478 * on the runqueue and wakes it.
1480 void fastcall wake_up_new_task(task_t *p, unsigned long clone_flags)
1482 unsigned long flags;
1484 runqueue_t *rq, *this_rq;
1486 rq = task_rq_lock(p, &flags);
1487 BUG_ON(p->state != TASK_RUNNING);
1488 this_cpu = smp_processor_id();
1492 * We decrease the sleep average of forking parents
1493 * and children as well, to keep max-interactive tasks
1494 * from forking tasks that are max-interactive. The parent
1495 * (current) is done further down, under its lock.
1497 p->sleep_avg = JIFFIES_TO_NS(CURRENT_BONUS(p) *
1498 CHILD_PENALTY / 100 * MAX_SLEEP_AVG / MAX_BONUS);
1500 p->prio = effective_prio(p);
1502 vx_activate_task(p);
1503 if (likely(cpu == this_cpu)) {
1504 if (!(clone_flags & CLONE_VM)) {
1506 * The VM isn't cloned, so we're in a good position to
1507 * do child-runs-first in anticipation of an exec. This
1508 * usually avoids a lot of COW overhead.
1510 if (unlikely(!current->array))
1511 __activate_task(p, rq);
1513 p->prio = current->prio;
1514 BUG_ON(p->state & TASK_ONHOLD);
1515 list_add_tail(&p->run_list, ¤t->run_list);
1516 p->array = current->array;
1517 p->array->nr_active++;
1522 /* Run child last */
1523 __activate_task(p, rq);
1525 * We skip the following code due to cpu == this_cpu
1527 * task_rq_unlock(rq, &flags);
1528 * this_rq = task_rq_lock(current, &flags);
1532 this_rq = cpu_rq(this_cpu);
1535 * Not the local CPU - must adjust timestamp. This should
1536 * get optimised away in the !CONFIG_SMP case.
1538 p->timestamp = (p->timestamp - this_rq->timestamp_last_tick)
1539 + rq->timestamp_last_tick;
1540 __activate_task(p, rq);
1541 if (TASK_PREEMPTS_CURR(p, rq))
1542 resched_task(rq->curr);
1545 * Parent and child are on different CPUs, now get the
1546 * parent runqueue to update the parent's ->sleep_avg:
1548 task_rq_unlock(rq, &flags);
1549 this_rq = task_rq_lock(current, &flags);
1551 current->sleep_avg = JIFFIES_TO_NS(CURRENT_BONUS(current) *
1552 PARENT_PENALTY / 100 * MAX_SLEEP_AVG / MAX_BONUS);
1553 task_rq_unlock(this_rq, &flags);
1557 * Potentially available exiting-child timeslices are
1558 * retrieved here - this way the parent does not get
1559 * penalized for creating too many threads.
1561 * (this cannot be used to 'generate' timeslices
1562 * artificially, because any timeslice recovered here
1563 * was given away by the parent in the first place.)
1565 void fastcall sched_exit(task_t *p)
1567 unsigned long flags;
1571 * If the child was a (relative-) CPU hog then decrease
1572 * the sleep_avg of the parent as well.
1574 rq = task_rq_lock(p->parent, &flags);
1575 if (p->first_time_slice && task_cpu(p) == task_cpu(p->parent)) {
1576 p->parent->time_slice += p->time_slice;
1577 if (unlikely(p->parent->time_slice > task_timeslice(p)))
1578 p->parent->time_slice = task_timeslice(p);
1580 if (p->sleep_avg < p->parent->sleep_avg)
1581 p->parent->sleep_avg = p->parent->sleep_avg /
1582 (EXIT_WEIGHT + 1) * EXIT_WEIGHT + p->sleep_avg /
1584 task_rq_unlock(rq, &flags);
1588 * prepare_task_switch - prepare to switch tasks
1589 * @rq: the runqueue preparing to switch
1590 * @next: the task we are going to switch to.
1592 * This is called with the rq lock held and interrupts off. It must
1593 * be paired with a subsequent finish_task_switch after the context
1596 * prepare_task_switch sets up locking and calls architecture specific
1599 static inline void prepare_task_switch(runqueue_t *rq, task_t *next)
1601 prepare_lock_switch(rq, next);
1602 prepare_arch_switch(next);
1606 * finish_task_switch - clean up after a task-switch
1607 * @rq: runqueue associated with task-switch
1608 * @prev: the thread we just switched away from.
1610 * finish_task_switch must be called after the context switch, paired
1611 * with a prepare_task_switch call before the context switch.
1612 * finish_task_switch will reconcile locking set up by prepare_task_switch,
1613 * and do any other architecture-specific cleanup actions.
1615 * Note that we may have delayed dropping an mm in context_switch(). If
1616 * so, we finish that here outside of the runqueue lock. (Doing it
1617 * with the lock held can cause deadlocks; see schedule() for
1620 static inline void finish_task_switch(runqueue_t *rq, task_t *prev)
1621 __releases(rq->lock)
1623 struct mm_struct *mm = rq->prev_mm;
1624 unsigned long prev_task_flags;
1629 * A task struct has one reference for the use as "current".
1630 * If a task dies, then it sets EXIT_ZOMBIE in tsk->exit_state and
1631 * calls schedule one last time. The schedule call will never return,
1632 * and the scheduled task must drop that reference.
1633 * The test for EXIT_ZOMBIE must occur while the runqueue locks are
1634 * still held, otherwise prev could be scheduled on another cpu, die
1635 * there before we look at prev->state, and then the reference would
1637 * Manfred Spraul <manfred@colorfullife.com>
1639 prev_task_flags = prev->flags;
1640 finish_arch_switch(prev);
1641 finish_lock_switch(rq, prev);
1644 if (unlikely(prev_task_flags & PF_DEAD))
1645 put_task_struct(prev);
1649 * schedule_tail - first thing a freshly forked thread must call.
1650 * @prev: the thread we just switched away from.
1652 asmlinkage void schedule_tail(task_t *prev)
1653 __releases(rq->lock)
1655 runqueue_t *rq = this_rq();
1656 finish_task_switch(rq, prev);
1657 #ifdef __ARCH_WANT_UNLOCKED_CTXSW
1658 /* In this case, finish_task_switch does not reenable preemption */
1661 if (current->set_child_tid)
1662 put_user(current->pid, current->set_child_tid);
1666 * context_switch - switch to the new MM and the new
1667 * thread's register state.
1670 task_t * context_switch(runqueue_t *rq, task_t *prev, task_t *next)
1672 struct mm_struct *mm = next->mm;
1673 struct mm_struct *oldmm = prev->active_mm;
1675 if (unlikely(!mm)) {
1676 next->active_mm = oldmm;
1677 atomic_inc(&oldmm->mm_count);
1678 enter_lazy_tlb(oldmm, next);
1680 switch_mm(oldmm, mm, next);
1682 if (unlikely(!prev->mm)) {
1683 prev->active_mm = NULL;
1684 WARN_ON(rq->prev_mm);
1685 rq->prev_mm = oldmm;
1688 /* Here we just switch the register state and the stack. */
1689 switch_to(prev, next, prev);
1695 * nr_running, nr_uninterruptible and nr_context_switches:
1697 * externally visible scheduler statistics: current number of runnable
1698 * threads, current number of uninterruptible-sleeping threads, total
1699 * number of context switches performed since bootup.
1701 unsigned long nr_running(void)
1703 unsigned long i, sum = 0;
1705 for_each_online_cpu(i)
1706 sum += cpu_rq(i)->nr_running;
1711 unsigned long nr_uninterruptible(void)
1713 unsigned long i, sum = 0;
1716 sum += cpu_rq(i)->nr_uninterruptible;
1719 * Since we read the counters lockless, it might be slightly
1720 * inaccurate. Do not allow it to go below zero though:
1722 if (unlikely((long)sum < 0))
1728 unsigned long long nr_context_switches(void)
1730 unsigned long long i, sum = 0;
1733 sum += cpu_rq(i)->nr_switches;
1738 unsigned long nr_iowait(void)
1740 unsigned long i, sum = 0;
1743 sum += atomic_read(&cpu_rq(i)->nr_iowait);
1751 * double_rq_lock - safely lock two runqueues
1753 * We must take them in cpu order to match code in
1754 * dependent_sleeper and wake_dependent_sleeper.
1756 * Note this does not disable interrupts like task_rq_lock,
1757 * you need to do so manually before calling.
1759 static void double_rq_lock(runqueue_t *rq1, runqueue_t *rq2)
1760 __acquires(rq1->lock)
1761 __acquires(rq2->lock)
1764 spin_lock(&rq1->lock);
1765 __acquire(rq2->lock); /* Fake it out ;) */
1767 if (rq1->cpu < rq2->cpu) {
1768 spin_lock(&rq1->lock);
1769 spin_lock(&rq2->lock);
1771 spin_lock(&rq2->lock);
1772 spin_lock(&rq1->lock);
1778 * double_rq_unlock - safely unlock two runqueues
1780 * Note this does not restore interrupts like task_rq_unlock,
1781 * you need to do so manually after calling.
1783 static void double_rq_unlock(runqueue_t *rq1, runqueue_t *rq2)
1784 __releases(rq1->lock)
1785 __releases(rq2->lock)
1787 spin_unlock(&rq1->lock);
1789 spin_unlock(&rq2->lock);
1791 __release(rq2->lock);
1795 * double_lock_balance - lock the busiest runqueue, this_rq is locked already.
1797 static void double_lock_balance(runqueue_t *this_rq, runqueue_t *busiest)
1798 __releases(this_rq->lock)
1799 __acquires(busiest->lock)
1800 __acquires(this_rq->lock)
1802 if (unlikely(!spin_trylock(&busiest->lock))) {
1803 if (busiest->cpu < this_rq->cpu) {
1804 spin_unlock(&this_rq->lock);
1805 spin_lock(&busiest->lock);
1806 spin_lock(&this_rq->lock);
1808 spin_lock(&busiest->lock);
1813 * If dest_cpu is allowed for this process, migrate the task to it.
1814 * This is accomplished by forcing the cpu_allowed mask to only
1815 * allow dest_cpu, which will force the cpu onto dest_cpu. Then
1816 * the cpu_allowed mask is restored.
1818 static void sched_migrate_task(task_t *p, int dest_cpu)
1820 migration_req_t req;
1822 unsigned long flags;
1824 rq = task_rq_lock(p, &flags);
1825 if (!cpu_isset(dest_cpu, p->cpus_allowed)
1826 || unlikely(cpu_is_offline(dest_cpu)))
1829 /* force the process onto the specified CPU */
1830 if (migrate_task(p, dest_cpu, &req)) {
1831 /* Need to wait for migration thread (might exit: take ref). */
1832 struct task_struct *mt = rq->migration_thread;
1833 get_task_struct(mt);
1834 task_rq_unlock(rq, &flags);
1835 wake_up_process(mt);
1836 put_task_struct(mt);
1837 wait_for_completion(&req.done);
1841 task_rq_unlock(rq, &flags);
1845 * sched_exec - execve() is a valuable balancing opportunity, because at
1846 * this point the task has the smallest effective memory and cache footprint.
1848 void sched_exec(void)
1850 int new_cpu, this_cpu = get_cpu();
1851 new_cpu = sched_balance_self(this_cpu, SD_BALANCE_EXEC);
1853 if (new_cpu != this_cpu)
1854 sched_migrate_task(current, new_cpu);
1858 * pull_task - move a task from a remote runqueue to the local runqueue.
1859 * Both runqueues must be locked.
1862 void pull_task(runqueue_t *src_rq, prio_array_t *src_array, task_t *p,
1863 runqueue_t *this_rq, prio_array_t *this_array, int this_cpu)
1865 dequeue_task(p, src_array);
1866 src_rq->nr_running--;
1867 set_task_cpu(p, this_cpu);
1868 this_rq->nr_running++;
1869 enqueue_task(p, this_array);
1870 p->timestamp = (p->timestamp - src_rq->timestamp_last_tick)
1871 + this_rq->timestamp_last_tick;
1873 * Note that idle threads have a prio of MAX_PRIO, for this test
1874 * to be always true for them.
1876 if (TASK_PREEMPTS_CURR(p, this_rq))
1877 resched_task(this_rq->curr);
1881 * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
1884 int can_migrate_task(task_t *p, runqueue_t *rq, int this_cpu,
1885 struct sched_domain *sd, enum idle_type idle,
1889 * We do not migrate tasks that are:
1890 * 1) running (obviously), or
1891 * 2) cannot be migrated to this CPU due to cpus_allowed, or
1892 * 3) are cache-hot on their current CPU.
1894 if (!cpu_isset(this_cpu, p->cpus_allowed))
1898 if (task_running(rq, p))
1902 * Aggressive migration if:
1903 * 1) task is cache cold, or
1904 * 2) too many balance attempts have failed.
1907 if (sd->nr_balance_failed > sd->cache_nice_tries)
1910 if (task_hot(p, rq->timestamp_last_tick, sd))
1916 * move_tasks tries to move up to max_nr_move tasks from busiest to this_rq,
1917 * as part of a balancing operation within "domain". Returns the number of
1920 * Called with both runqueues locked.
1922 static int move_tasks(runqueue_t *this_rq, int this_cpu, runqueue_t *busiest,
1923 unsigned long max_nr_move, struct sched_domain *sd,
1924 enum idle_type idle, int *all_pinned)
1926 prio_array_t *array, *dst_array;
1927 struct list_head *head, *curr;
1928 int idx, pulled = 0, pinned = 0;
1931 if (max_nr_move == 0)
1937 * We first consider expired tasks. Those will likely not be
1938 * executed in the near future, and they are most likely to
1939 * be cache-cold, thus switching CPUs has the least effect
1942 if (busiest->expired->nr_active) {
1943 array = busiest->expired;
1944 dst_array = this_rq->expired;
1946 array = busiest->active;
1947 dst_array = this_rq->active;
1951 /* Start searching at priority 0: */
1955 idx = sched_find_first_bit(array->bitmap);
1957 idx = find_next_bit(array->bitmap, MAX_PRIO, idx);
1958 if (idx >= MAX_PRIO) {
1959 if (array == busiest->expired && busiest->active->nr_active) {
1960 array = busiest->active;
1961 dst_array = this_rq->active;
1967 head = array->queue + idx;
1970 tmp = list_entry(curr, task_t, run_list);
1974 if (!can_migrate_task(tmp, busiest, this_cpu, sd, idle, &pinned)) {
1981 #ifdef CONFIG_SCHEDSTATS
1982 if (task_hot(tmp, busiest->timestamp_last_tick, sd))
1983 schedstat_inc(sd, lb_hot_gained[idle]);
1986 pull_task(busiest, array, tmp, this_rq, dst_array, this_cpu);
1989 /* We only want to steal up to the prescribed number of tasks. */
1990 if (pulled < max_nr_move) {
1998 * Right now, this is the only place pull_task() is called,
1999 * so we can safely collect pull_task() stats here rather than
2000 * inside pull_task().
2002 schedstat_add(sd, lb_gained[idle], pulled);
2005 *all_pinned = pinned;
2010 * find_busiest_group finds and returns the busiest CPU group within the
2011 * domain. It calculates and returns the number of tasks which should be
2012 * moved to restore balance via the imbalance parameter.
2014 static struct sched_group *
2015 find_busiest_group(struct sched_domain *sd, int this_cpu,
2016 unsigned long *imbalance, enum idle_type idle, int *sd_idle,
2019 struct sched_group *busiest = NULL, *this = NULL, *group = sd->groups;
2020 unsigned long max_load, avg_load, total_load, this_load, total_pwr;
2021 unsigned long max_pull;
2024 max_load = this_load = total_load = total_pwr = 0;
2025 if (idle == NOT_IDLE)
2026 load_idx = sd->busy_idx;
2027 else if (idle == NEWLY_IDLE)
2028 load_idx = sd->newidle_idx;
2030 load_idx = sd->idle_idx;
2037 local_group = cpu_isset(this_cpu, group->cpumask);
2039 /* Tally up the load of all CPUs in the group */
2042 for_each_cpu_mask(i, group->cpumask) {
2043 if (!cpu_isset(i, *cpus))
2046 if (*sd_idle && !idle_cpu(i))
2049 /* Bias balancing toward cpus of our domain */
2051 load = target_load(i, load_idx);
2053 load = source_load(i, load_idx);
2058 total_load += avg_load;
2059 total_pwr += group->cpu_power;
2061 /* Adjust by relative CPU power of the group */
2062 avg_load = (avg_load * SCHED_LOAD_SCALE) / group->cpu_power;
2065 this_load = avg_load;
2067 } else if (avg_load > max_load) {
2068 max_load = avg_load;
2071 group = group->next;
2072 } while (group != sd->groups);
2074 if (!busiest || this_load >= max_load || max_load <= SCHED_LOAD_SCALE)
2077 avg_load = (SCHED_LOAD_SCALE * total_load) / total_pwr;
2079 if (this_load >= avg_load ||
2080 100*max_load <= sd->imbalance_pct*this_load)
2084 * We're trying to get all the cpus to the average_load, so we don't
2085 * want to push ourselves above the average load, nor do we wish to
2086 * reduce the max loaded cpu below the average load, as either of these
2087 * actions would just result in more rebalancing later, and ping-pong
2088 * tasks around. Thus we look for the minimum possible imbalance.
2089 * Negative imbalances (*we* are more loaded than anyone else) will
2090 * be counted as no imbalance for these purposes -- we can't fix that
2091 * by pulling tasks to us. Be careful of negative numbers as they'll
2092 * appear as very large values with unsigned longs.
2095 /* Don't want to pull so many tasks that a group would go idle */
2096 max_pull = min(max_load - avg_load, max_load - SCHED_LOAD_SCALE);
2098 /* How much load to actually move to equalise the imbalance */
2099 *imbalance = min(max_pull * busiest->cpu_power,
2100 (avg_load - this_load) * this->cpu_power)
2103 if (*imbalance < SCHED_LOAD_SCALE) {
2104 unsigned long pwr_now = 0, pwr_move = 0;
2107 if (max_load - this_load >= SCHED_LOAD_SCALE*2) {
2113 * OK, we don't have enough imbalance to justify moving tasks,
2114 * however we may be able to increase total CPU power used by
2118 pwr_now += busiest->cpu_power*min(SCHED_LOAD_SCALE, max_load);
2119 pwr_now += this->cpu_power*min(SCHED_LOAD_SCALE, this_load);
2120 pwr_now /= SCHED_LOAD_SCALE;
2122 /* Amount of load we'd subtract */
2123 tmp = SCHED_LOAD_SCALE*SCHED_LOAD_SCALE/busiest->cpu_power;
2125 pwr_move += busiest->cpu_power*min(SCHED_LOAD_SCALE,
2128 /* Amount of load we'd add */
2129 if (max_load*busiest->cpu_power <
2130 SCHED_LOAD_SCALE*SCHED_LOAD_SCALE)
2131 tmp = max_load*busiest->cpu_power/this->cpu_power;
2133 tmp = SCHED_LOAD_SCALE*SCHED_LOAD_SCALE/this->cpu_power;
2134 pwr_move += this->cpu_power*min(SCHED_LOAD_SCALE, this_load + tmp);
2135 pwr_move /= SCHED_LOAD_SCALE;
2137 /* Move if we gain throughput */
2138 if (pwr_move <= pwr_now)
2145 /* Get rid of the scaling factor, rounding down as we divide */
2146 *imbalance = *imbalance / SCHED_LOAD_SCALE;
2156 * find_busiest_queue - find the busiest runqueue among the cpus in group.
2158 static runqueue_t *find_busiest_queue(struct sched_group *group,
2159 enum idle_type idle, cpumask_t *cpus)
2161 unsigned long load, max_load = 0;
2162 runqueue_t *busiest = NULL;
2165 for_each_cpu_mask(i, group->cpumask) {
2166 if (!cpu_isset(i, *cpus))
2169 load = source_load(i, 0);
2171 if (load > max_load) {
2173 busiest = cpu_rq(i);
2181 * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but
2182 * so long as it is large enough.
2184 #define MAX_PINNED_INTERVAL 512
2187 * Check this_cpu to ensure it is balanced within domain. Attempt to move
2188 * tasks if there is an imbalance.
2190 * Called with this_rq unlocked.
2192 static int load_balance(int this_cpu, runqueue_t *this_rq,
2193 struct sched_domain *sd, enum idle_type idle)
2195 struct sched_group *group;
2196 runqueue_t *busiest;
2197 unsigned long imbalance;
2198 int nr_moved, all_pinned = 0;
2199 int active_balance = 0;
2201 cpumask_t cpus = CPU_MASK_ALL;
2203 if (idle != NOT_IDLE && sd->flags & SD_SHARE_CPUPOWER)
2206 schedstat_inc(sd, lb_cnt[idle]);
2209 group = find_busiest_group(sd, this_cpu, &imbalance, idle,
2212 schedstat_inc(sd, lb_nobusyg[idle]);
2216 busiest = find_busiest_queue(group, idle, &cpus);
2218 schedstat_inc(sd, lb_nobusyq[idle]);
2222 BUG_ON(busiest == this_rq);
2224 schedstat_add(sd, lb_imbalance[idle], imbalance);
2227 if (busiest->nr_running > 1) {
2229 * Attempt to move tasks. If find_busiest_group has found
2230 * an imbalance but busiest->nr_running <= 1, the group is
2231 * still unbalanced. nr_moved simply stays zero, so it is
2232 * correctly treated as an imbalance.
2234 double_rq_lock(this_rq, busiest);
2235 nr_moved = move_tasks(this_rq, this_cpu, busiest,
2236 imbalance, sd, idle, &all_pinned);
2237 double_rq_unlock(this_rq, busiest);
2239 /* All tasks on this runqueue were pinned by CPU affinity */
2240 if (unlikely(all_pinned)) {
2241 cpu_clear(busiest->cpu, cpus);
2242 if (!cpus_empty(cpus))
2249 schedstat_inc(sd, lb_failed[idle]);
2250 sd->nr_balance_failed++;
2252 if (unlikely(sd->nr_balance_failed > sd->cache_nice_tries+2)) {
2254 spin_lock(&busiest->lock);
2256 /* don't kick the migration_thread, if the curr
2257 * task on busiest cpu can't be moved to this_cpu
2259 if (!cpu_isset(this_cpu, busiest->curr->cpus_allowed)) {
2260 spin_unlock(&busiest->lock);
2262 goto out_one_pinned;
2265 if (!busiest->active_balance) {
2266 busiest->active_balance = 1;
2267 busiest->push_cpu = this_cpu;
2270 spin_unlock(&busiest->lock);
2272 wake_up_process(busiest->migration_thread);
2275 * We've kicked active balancing, reset the failure
2278 sd->nr_balance_failed = sd->cache_nice_tries+1;
2281 sd->nr_balance_failed = 0;
2283 if (likely(!active_balance)) {
2284 /* We were unbalanced, so reset the balancing interval */
2285 sd->balance_interval = sd->min_interval;
2288 * If we've begun active balancing, start to back off. This
2289 * case may not be covered by the all_pinned logic if there
2290 * is only 1 task on the busy runqueue (because we don't call
2293 if (sd->balance_interval < sd->max_interval)
2294 sd->balance_interval *= 2;
2297 if (!nr_moved && !sd_idle && sd->flags & SD_SHARE_CPUPOWER)
2302 schedstat_inc(sd, lb_balanced[idle]);
2304 sd->nr_balance_failed = 0;
2307 /* tune up the balancing interval */
2308 if ((all_pinned && sd->balance_interval < MAX_PINNED_INTERVAL) ||
2309 (sd->balance_interval < sd->max_interval))
2310 sd->balance_interval *= 2;
2312 if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER)
2318 * Check this_cpu to ensure it is balanced within domain. Attempt to move
2319 * tasks if there is an imbalance.
2321 * Called from schedule when this_rq is about to become idle (NEWLY_IDLE).
2322 * this_rq is locked.
2324 static int load_balance_newidle(int this_cpu, runqueue_t *this_rq,
2325 struct sched_domain *sd)
2327 struct sched_group *group;
2328 runqueue_t *busiest = NULL;
2329 unsigned long imbalance;
2332 cpumask_t cpus = CPU_MASK_ALL;
2334 if (sd->flags & SD_SHARE_CPUPOWER)
2337 schedstat_inc(sd, lb_cnt[NEWLY_IDLE]);
2339 group = find_busiest_group(sd, this_cpu, &imbalance, NEWLY_IDLE,
2342 schedstat_inc(sd, lb_nobusyg[NEWLY_IDLE]);
2346 busiest = find_busiest_queue(group, NEWLY_IDLE, &cpus);
2348 schedstat_inc(sd, lb_nobusyq[NEWLY_IDLE]);
2352 BUG_ON(busiest == this_rq);
2354 schedstat_add(sd, lb_imbalance[NEWLY_IDLE], imbalance);
2357 if (busiest->nr_running > 1) {
2358 /* Attempt to move tasks */
2359 double_lock_balance(this_rq, busiest);
2360 nr_moved = move_tasks(this_rq, this_cpu, busiest,
2361 imbalance, sd, NEWLY_IDLE, NULL);
2362 spin_unlock(&busiest->lock);
2365 cpu_clear(busiest->cpu, cpus);
2366 if (!cpus_empty(cpus))
2372 schedstat_inc(sd, lb_failed[NEWLY_IDLE]);
2373 if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER)
2376 sd->nr_balance_failed = 0;
2381 schedstat_inc(sd, lb_balanced[NEWLY_IDLE]);
2382 if (!sd_idle && sd->flags & SD_SHARE_CPUPOWER)
2384 sd->nr_balance_failed = 0;
2389 * idle_balance is called by schedule() if this_cpu is about to become
2390 * idle. Attempts to pull tasks from other CPUs.
2392 static void idle_balance(int this_cpu, runqueue_t *this_rq)
2394 struct sched_domain *sd;
2396 for_each_domain(this_cpu, sd) {
2397 if (sd->flags & SD_BALANCE_NEWIDLE) {
2398 if (load_balance_newidle(this_cpu, this_rq, sd)) {
2399 /* We've pulled tasks over so stop searching */
2407 * active_load_balance is run by migration threads. It pushes running tasks
2408 * off the busiest CPU onto idle CPUs. It requires at least 1 task to be
2409 * running on each physical CPU where possible, and avoids physical /
2410 * logical imbalances.
2412 * Called with busiest_rq locked.
2414 static void active_load_balance(runqueue_t *busiest_rq, int busiest_cpu)
2416 struct sched_domain *sd;
2417 runqueue_t *target_rq;
2418 int target_cpu = busiest_rq->push_cpu;
2420 if (busiest_rq->nr_running <= 1)
2421 /* no task to move */
2424 target_rq = cpu_rq(target_cpu);
2427 * This condition is "impossible", if it occurs
2428 * we need to fix it. Originally reported by
2429 * Bjorn Helgaas on a 128-cpu setup.
2431 BUG_ON(busiest_rq == target_rq);
2433 /* move a task from busiest_rq to target_rq */
2434 double_lock_balance(busiest_rq, target_rq);
2436 /* Search for an sd spanning us and the target CPU. */
2437 for_each_domain(target_cpu, sd)
2438 if ((sd->flags & SD_LOAD_BALANCE) &&
2439 cpu_isset(busiest_cpu, sd->span))
2442 if (unlikely(sd == NULL))
2445 schedstat_inc(sd, alb_cnt);
2447 if (move_tasks(target_rq, target_cpu, busiest_rq, 1, sd, SCHED_IDLE, NULL))
2448 schedstat_inc(sd, alb_pushed);
2450 schedstat_inc(sd, alb_failed);
2452 spin_unlock(&target_rq->lock);
2456 * rebalance_tick will get called every timer tick, on every CPU.
2458 * It checks each scheduling domain to see if it is due to be balanced,
2459 * and initiates a balancing operation if so.
2461 * Balancing parameters are set up in arch_init_sched_domains.
2464 /* Don't have all balancing operations going off at once */
2465 #define CPU_OFFSET(cpu) (HZ * cpu / NR_CPUS)
2467 static void rebalance_tick(int this_cpu, runqueue_t *this_rq,
2468 enum idle_type idle)
2470 unsigned long old_load, this_load;
2471 unsigned long j = jiffies + CPU_OFFSET(this_cpu);
2472 struct sched_domain *sd;
2475 this_load = this_rq->nr_running * SCHED_LOAD_SCALE;
2476 /* Update our load */
2477 for (i = 0; i < 3; i++) {
2478 unsigned long new_load = this_load;
2480 old_load = this_rq->cpu_load[i];
2482 * Round up the averaging division if load is increasing. This
2483 * prevents us from getting stuck on 9 if the load is 10, for
2486 if (new_load > old_load)
2487 new_load += scale-1;
2488 this_rq->cpu_load[i] = (old_load*(scale-1) + new_load) / scale;
2491 for_each_domain(this_cpu, sd) {
2492 unsigned long interval;
2494 if (!(sd->flags & SD_LOAD_BALANCE))
2497 interval = sd->balance_interval;
2498 if (idle != SCHED_IDLE)
2499 interval *= sd->busy_factor;
2501 /* scale ms to jiffies */
2502 interval = msecs_to_jiffies(interval);
2503 if (unlikely(!interval))
2506 if (j - sd->last_balance >= interval) {
2507 if (load_balance(this_cpu, this_rq, sd, idle)) {
2509 * We've pulled tasks over so either we're no
2510 * longer idle, or one of our SMT siblings is
2515 sd->last_balance += interval;
2521 * on UP we do not need to balance between CPUs:
2523 static inline void rebalance_tick(int cpu, runqueue_t *rq, enum idle_type idle)
2526 static inline void idle_balance(int cpu, runqueue_t *rq)
2531 static inline int wake_priority_sleeper(runqueue_t *rq)
2534 #ifdef CONFIG_SCHED_SMT
2535 spin_lock(&rq->lock);
2537 * If an SMT sibling task has been put to sleep for priority
2538 * reasons reschedule the idle task to see if it can now run.
2540 if (rq->nr_running) {
2541 resched_task(rq->idle);
2544 spin_unlock(&rq->lock);
2549 DEFINE_PER_CPU(struct kernel_stat, kstat);
2551 EXPORT_PER_CPU_SYMBOL(kstat);
2554 * This is called on clock ticks and on context switches.
2555 * Bank in p->sched_time the ns elapsed since the last tick or switch.
2557 static inline void update_cpu_clock(task_t *p, runqueue_t *rq,
2558 unsigned long long now)
2560 unsigned long long last = max(p->timestamp, rq->timestamp_last_tick);
2561 p->sched_time += now - last;
2565 * Return current->sched_time plus any more ns on the sched_clock
2566 * that have not yet been banked.
2568 unsigned long long current_sched_time(const task_t *tsk)
2570 unsigned long long ns;
2571 unsigned long flags;
2572 local_irq_save(flags);
2573 ns = max(tsk->timestamp, task_rq(tsk)->timestamp_last_tick);
2574 ns = tsk->sched_time + (sched_clock() - ns);
2575 local_irq_restore(flags);
2580 * We place interactive tasks back into the active array, if possible.
2582 * To guarantee that this does not starve expired tasks we ignore the
2583 * interactivity of a task if the first expired task had to wait more
2584 * than a 'reasonable' amount of time. This deadline timeout is
2585 * load-dependent, as the frequency of array switched decreases with
2586 * increasing number of running tasks. We also ignore the interactivity
2587 * if a better static_prio task has expired:
2589 #define EXPIRED_STARVING(rq) \
2590 ((STARVATION_LIMIT && ((rq)->expired_timestamp && \
2591 (jiffies - (rq)->expired_timestamp >= \
2592 STARVATION_LIMIT * ((rq)->nr_running) + 1))) || \
2593 ((rq)->curr->static_prio > (rq)->best_expired_prio))
2596 * Account user cpu time to a process.
2597 * @p: the process that the cpu time gets accounted to
2598 * @hardirq_offset: the offset to subtract from hardirq_count()
2599 * @cputime: the cpu time spent in user space since the last update
2601 void account_user_time(struct task_struct *p, cputime_t cputime)
2603 struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
2604 struct vx_info *vxi = p->vx_info; /* p is _always_ current */
2606 int nice = (TASK_NICE(p) > 0);
2608 p->utime = cputime_add(p->utime, cputime);
2609 vx_account_user(vxi, cputime, nice);
2611 /* Add user time to cpustat. */
2612 tmp = cputime_to_cputime64(cputime);
2614 cpustat->nice = cputime64_add(cpustat->nice, tmp);
2616 cpustat->user = cputime64_add(cpustat->user, tmp);
2620 * Account system cpu time to a process.
2621 * @p: the process that the cpu time gets accounted to
2622 * @hardirq_offset: the offset to subtract from hardirq_count()
2623 * @cputime: the cpu time spent in kernel space since the last update
2625 void account_system_time(struct task_struct *p, int hardirq_offset,
2628 struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
2629 struct vx_info *vxi = p->vx_info; /* p is _always_ current */
2630 runqueue_t *rq = this_rq();
2633 p->stime = cputime_add(p->stime, cputime);
2634 vx_account_system(vxi, cputime, (p == rq->idle));
2636 /* Add system time to cpustat. */
2637 tmp = cputime_to_cputime64(cputime);
2638 if (hardirq_count() - hardirq_offset)
2639 cpustat->irq = cputime64_add(cpustat->irq, tmp);
2640 else if (softirq_count())
2641 cpustat->softirq = cputime64_add(cpustat->softirq, tmp);
2642 else if (p != rq->idle)
2643 cpustat->system = cputime64_add(cpustat->system, tmp);
2644 else if (atomic_read(&rq->nr_iowait) > 0)
2645 cpustat->iowait = cputime64_add(cpustat->iowait, tmp);
2647 cpustat->idle = cputime64_add(cpustat->idle, tmp);
2648 /* Account for system time used */
2649 acct_update_integrals(p);
2653 * Account for involuntary wait time.
2654 * @p: the process from which the cpu time has been stolen
2655 * @steal: the cpu time spent in involuntary wait
2657 void account_steal_time(struct task_struct *p, cputime_t steal)
2659 struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
2660 cputime64_t tmp = cputime_to_cputime64(steal);
2661 runqueue_t *rq = this_rq();
2663 if (p == rq->idle) {
2664 p->stime = cputime_add(p->stime, steal);
2665 if (atomic_read(&rq->nr_iowait) > 0)
2666 cpustat->iowait = cputime64_add(cpustat->iowait, tmp);
2668 cpustat->idle = cputime64_add(cpustat->idle, tmp);
2670 cpustat->steal = cputime64_add(cpustat->steal, tmp);
2674 * This function gets called by the timer code, with HZ frequency.
2675 * We call it with interrupts disabled.
2677 * It also gets called by the fork code, when changing the parent's
2680 void scheduler_tick(void)
2682 int cpu = smp_processor_id();
2683 runqueue_t *rq = this_rq();
2684 task_t *p = current;
2685 unsigned long long now = sched_clock();
2687 update_cpu_clock(p, rq, now);
2689 rq->timestamp_last_tick = now;
2691 #if defined(CONFIG_VSERVER_HARDCPU) && defined(CONFIG_VSERVER_ACB_SCHED)
2692 vx_scheduler_tick();
2695 if (p == rq->idle) {
2696 if (wake_priority_sleeper(rq))
2698 #ifdef CONFIG_VSERVER_HARDCPU_IDLE
2699 if (!--rq->idle_tokens && !list_empty(&rq->hold_queue))
2702 rebalance_tick(cpu, rq, SCHED_IDLE);
2706 /* Task might have expired already, but not scheduled off yet */
2707 if (p->array != rq->active) {
2708 set_tsk_need_resched(p);
2711 spin_lock(&rq->lock);
2713 * The task was running during this tick - update the
2714 * time slice counter. Note: we do not update a thread's
2715 * priority until it either goes to sleep or uses up its
2716 * timeslice. This makes it possible for interactive tasks
2717 * to use up their timeslices at their highest priority levels.
2721 * RR tasks need a special form of timeslice management.
2722 * FIFO tasks have no timeslices.
2724 if ((p->policy == SCHED_RR) && vx_need_resched(p)) {
2725 p->time_slice = task_timeslice(p);
2726 p->first_time_slice = 0;
2727 set_tsk_need_resched(p);
2729 /* put it at the end of the queue: */
2730 requeue_task(p, rq->active);
2734 if (vx_need_resched(p)) {
2735 dequeue_task(p, rq->active);
2736 set_tsk_need_resched(p);
2737 p->prio = effective_prio(p);
2738 p->time_slice = task_timeslice(p);
2739 p->first_time_slice = 0;
2741 if (!rq->expired_timestamp)
2742 rq->expired_timestamp = jiffies;
2743 if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) {
2744 enqueue_task(p, rq->expired);
2745 if (p->static_prio < rq->best_expired_prio)
2746 rq->best_expired_prio = p->static_prio;
2748 enqueue_task(p, rq->active);
2751 * Prevent a too long timeslice allowing a task to monopolize
2752 * the CPU. We do this by splitting up the timeslice into
2755 * Note: this does not mean the task's timeslices expire or
2756 * get lost in any way, they just might be preempted by
2757 * another task of equal priority. (one with higher
2758 * priority would have preempted this task already.) We
2759 * requeue this task to the end of the list on this priority
2760 * level, which is in essence a round-robin of tasks with
2763 * This only applies to tasks in the interactive
2764 * delta range with at least TIMESLICE_GRANULARITY to requeue.
2766 if (TASK_INTERACTIVE(p) && !((task_timeslice(p) -
2767 p->time_slice) % TIMESLICE_GRANULARITY(p)) &&
2768 (p->time_slice >= TIMESLICE_GRANULARITY(p)) &&
2769 (p->array == rq->active)) {
2771 requeue_task(p, rq->active);
2772 set_tsk_need_resched(p);
2776 spin_unlock(&rq->lock);
2778 rebalance_tick(cpu, rq, NOT_IDLE);
2781 #ifdef CONFIG_SCHED_SMT
2782 static inline void wakeup_busy_runqueue(runqueue_t *rq)
2784 /* If an SMT runqueue is sleeping due to priority reasons wake it up */
2785 if (rq->curr == rq->idle && rq->nr_running)
2786 resched_task(rq->idle);
2789 static void wake_sleeping_dependent(int this_cpu, runqueue_t *this_rq)
2791 struct sched_domain *tmp, *sd = NULL;
2792 cpumask_t sibling_map;
2795 for_each_domain(this_cpu, tmp)
2796 if (tmp->flags & SD_SHARE_CPUPOWER)
2803 * Unlock the current runqueue because we have to lock in
2804 * CPU order to avoid deadlocks. Caller knows that we might
2805 * unlock. We keep IRQs disabled.
2807 spin_unlock(&this_rq->lock);
2809 sibling_map = sd->span;
2811 for_each_cpu_mask(i, sibling_map)
2812 spin_lock(&cpu_rq(i)->lock);
2814 * We clear this CPU from the mask. This both simplifies the
2815 * inner loop and keps this_rq locked when we exit:
2817 cpu_clear(this_cpu, sibling_map);
2819 for_each_cpu_mask(i, sibling_map) {
2820 runqueue_t *smt_rq = cpu_rq(i);
2822 wakeup_busy_runqueue(smt_rq);
2825 for_each_cpu_mask(i, sibling_map)
2826 spin_unlock(&cpu_rq(i)->lock);
2828 * We exit with this_cpu's rq still held and IRQs
2834 * number of 'lost' timeslices this task wont be able to fully
2835 * utilize, if another task runs on a sibling. This models the
2836 * slowdown effect of other tasks running on siblings:
2838 static inline unsigned long smt_slice(task_t *p, struct sched_domain *sd)
2840 return p->time_slice * (100 - sd->per_cpu_gain) / 100;
2843 static int dependent_sleeper(int this_cpu, runqueue_t *this_rq)
2845 struct sched_domain *tmp, *sd = NULL;
2846 cpumask_t sibling_map;
2847 prio_array_t *array;
2851 for_each_domain(this_cpu, tmp)
2852 if (tmp->flags & SD_SHARE_CPUPOWER)
2859 * The same locking rules and details apply as for
2860 * wake_sleeping_dependent():
2862 spin_unlock(&this_rq->lock);
2863 sibling_map = sd->span;
2864 for_each_cpu_mask(i, sibling_map)
2865 spin_lock(&cpu_rq(i)->lock);
2866 cpu_clear(this_cpu, sibling_map);
2869 * Establish next task to be run - it might have gone away because
2870 * we released the runqueue lock above:
2872 if (!this_rq->nr_running)
2874 array = this_rq->active;
2875 if (!array->nr_active)
2876 array = this_rq->expired;
2877 BUG_ON(!array->nr_active);
2879 p = list_entry(array->queue[sched_find_first_bit(array->bitmap)].next,
2882 for_each_cpu_mask(i, sibling_map) {
2883 runqueue_t *smt_rq = cpu_rq(i);
2884 task_t *smt_curr = smt_rq->curr;
2886 /* Kernel threads do not participate in dependent sleeping */
2887 if (!p->mm || !smt_curr->mm || rt_task(p))
2888 goto check_smt_task;
2891 * If a user task with lower static priority than the
2892 * running task on the SMT sibling is trying to schedule,
2893 * delay it till there is proportionately less timeslice
2894 * left of the sibling task to prevent a lower priority
2895 * task from using an unfair proportion of the
2896 * physical cpu's resources. -ck
2898 if (rt_task(smt_curr)) {
2900 * With real time tasks we run non-rt tasks only
2901 * per_cpu_gain% of the time.
2903 if ((jiffies % DEF_TIMESLICE) >
2904 (sd->per_cpu_gain * DEF_TIMESLICE / 100))
2907 if (smt_curr->static_prio < p->static_prio &&
2908 !TASK_PREEMPTS_CURR(p, smt_rq) &&
2909 smt_slice(smt_curr, sd) > task_timeslice(p))
2913 if ((!smt_curr->mm && smt_curr != smt_rq->idle) ||
2917 wakeup_busy_runqueue(smt_rq);
2922 * Reschedule a lower priority task on the SMT sibling for
2923 * it to be put to sleep, or wake it up if it has been put to
2924 * sleep for priority reasons to see if it should run now.
2927 if ((jiffies % DEF_TIMESLICE) >
2928 (sd->per_cpu_gain * DEF_TIMESLICE / 100))
2929 resched_task(smt_curr);
2931 if (TASK_PREEMPTS_CURR(p, smt_rq) &&
2932 smt_slice(p, sd) > task_timeslice(smt_curr))
2933 resched_task(smt_curr);
2935 wakeup_busy_runqueue(smt_rq);
2939 for_each_cpu_mask(i, sibling_map)
2940 spin_unlock(&cpu_rq(i)->lock);
2944 static inline void wake_sleeping_dependent(int this_cpu, runqueue_t *this_rq)
2948 static inline int dependent_sleeper(int this_cpu, runqueue_t *this_rq)
2954 #if defined(CONFIG_PREEMPT) && defined(CONFIG_DEBUG_PREEMPT)
2956 void fastcall add_preempt_count(int val)
2961 BUG_ON((preempt_count() < 0));
2962 preempt_count() += val;
2964 * Spinlock count overflowing soon?
2966 BUG_ON((preempt_count() & PREEMPT_MASK) >= PREEMPT_MASK-10);
2968 EXPORT_SYMBOL(add_preempt_count);
2970 void fastcall sub_preempt_count(int val)
2975 BUG_ON(val > preempt_count());
2977 * Is the spinlock portion underflowing?
2979 BUG_ON((val < PREEMPT_MASK) && !(preempt_count() & PREEMPT_MASK));
2980 preempt_count() -= val;
2982 EXPORT_SYMBOL(sub_preempt_count);
2987 * schedule() is the main scheduler function.
2989 asmlinkage void __sched schedule(void)
2992 task_t *prev, *next;
2994 prio_array_t *array;
2995 struct list_head *queue;
2996 unsigned long long now;
2997 unsigned long run_time;
2998 int cpu, idx, new_prio;
2999 struct vx_info *vxi;
3000 #ifdef CONFIG_VSERVER_HARDCPU
3002 # ifdef CONFIG_VSERVER_ACB_SCHED
3003 int min_guarantee_ticks = VX_INVALID_TICKS;
3004 int min_best_effort_ticks = VX_INVALID_TICKS;
3009 * Test if we are atomic. Since do_exit() needs to call into
3010 * schedule() atomically, we ignore that path for now.
3011 * Otherwise, whine if we are scheduling when we should not be.
3013 if (likely(!current->exit_state)) {
3014 if (unlikely(in_atomic())) {
3015 printk(KERN_ERR "scheduling while atomic: "
3017 current->comm, preempt_count(), current->pid);
3021 profile_hit(SCHED_PROFILING, __builtin_return_address(0));
3026 release_kernel_lock(prev);
3027 need_resched_nonpreemptible:
3031 * The idle thread is not allowed to schedule!
3032 * Remove this check after it has been exercised a bit.
3034 if (unlikely(prev == rq->idle) && prev->state != TASK_RUNNING) {
3035 printk(KERN_ERR "bad: scheduling from the idle thread!\n");
3039 schedstat_inc(rq, sched_cnt);
3040 now = sched_clock();
3041 if (likely((long long)(now - prev->timestamp) < NS_MAX_SLEEP_AVG)) {
3042 run_time = now - prev->timestamp;
3043 if (unlikely((long long)(now - prev->timestamp) < 0))
3046 run_time = NS_MAX_SLEEP_AVG;
3049 * Tasks charged proportionately less run_time at high sleep_avg to
3050 * delay them losing their interactive status
3052 run_time /= (CURRENT_BONUS(prev) ? : 1);
3054 spin_lock_irq(&rq->lock);
3056 if (unlikely(prev->flags & PF_DEAD))
3057 prev->state = EXIT_DEAD;
3059 switch_count = &prev->nivcsw;
3060 if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
3061 switch_count = &prev->nvcsw;
3062 if (unlikely((prev->state & TASK_INTERRUPTIBLE) &&
3063 unlikely(signal_pending(prev))))
3064 prev->state = TASK_RUNNING;
3066 if (prev->state == TASK_UNINTERRUPTIBLE) {
3067 rq->nr_uninterruptible++;
3068 vx_uninterruptible_inc(prev);
3070 deactivate_task(prev, rq);
3074 #ifdef CONFIG_VSERVER_HARDCPU
3075 # ifdef CONFIG_VSERVER_ACB_SCHED
3078 min_guarantee_ticks = VX_INVALID_TICKS;
3079 min_best_effort_ticks = VX_INVALID_TICKS;
3082 if (!list_empty(&rq->hold_queue)) {
3083 struct list_head *l, *n;
3087 list_for_each_safe(l, n, &rq->hold_queue) {
3088 next = list_entry(l, task_t, run_list);
3089 if (vxi == next->vx_info)
3092 vxi = next->vx_info;
3093 ret = vx_tokens_recalc(vxi);
3096 vx_unhold_task(vxi, next, rq);
3099 if ((ret < 0) && (maxidle < ret))
3101 # ifdef CONFIG_VSERVER_ACB_SCHED
3103 if (IS_BEST_EFFORT(vxi)) {
3104 if (min_best_effort_ticks < ret)
3105 min_best_effort_ticks = ret;
3107 if (min_guarantee_ticks < ret)
3108 min_guarantee_ticks = ret;
3114 rq->idle_tokens = -maxidle;
3119 cpu = smp_processor_id();
3120 if (unlikely(!rq->nr_running)) {
3122 idle_balance(cpu, rq);
3123 if (!rq->nr_running) {
3125 rq->expired_timestamp = 0;
3126 wake_sleeping_dependent(cpu, rq);
3128 * wake_sleeping_dependent() might have released
3129 * the runqueue, so break out if we got new
3132 if (!rq->nr_running)
3136 if (dependent_sleeper(cpu, rq)) {
3141 * dependent_sleeper() releases and reacquires the runqueue
3142 * lock, hence go into the idle loop if the rq went
3145 if (unlikely(!rq->nr_running))
3150 if (unlikely(!array->nr_active)) {
3152 * Switch the active and expired arrays.
3154 schedstat_inc(rq, sched_switch);
3155 rq->active = rq->expired;
3156 rq->expired = array;
3158 rq->expired_timestamp = 0;
3159 rq->best_expired_prio = MAX_PRIO;
3162 idx = sched_find_first_bit(array->bitmap);
3163 queue = array->queue + idx;
3164 next = list_entry(queue->next, task_t, run_list);
3166 vxi = next->vx_info;
3167 #ifdef CONFIG_VSERVER_HARDCPU
3168 if (vx_info_flags(vxi, VXF_SCHED_PAUSE|VXF_SCHED_HARD, 0)) {
3169 int ret = vx_tokens_recalc(vxi);
3171 if (unlikely(ret <= 0)) {
3173 if ((rq->idle_tokens > -ret))
3174 rq->idle_tokens = -ret;
3175 # ifdef CONFIG_VSERVER_ACB_SCHED
3176 if (IS_BEST_EFFORT(vxi)) {
3177 if (min_best_effort_ticks < ret)
3178 min_best_effort_ticks = ret;
3180 if (min_guarantee_ticks < ret)
3181 min_guarantee_ticks = ret;
3185 vx_hold_task(vxi, next, rq);
3188 } else /* well, looks ugly but not as ugly as the ifdef-ed version */
3190 if (vx_info_flags(vxi, VXF_SCHED_PRIO, 0))
3191 vx_tokens_recalc(vxi);
3193 if (!rt_task(next) && next->activated > 0) {
3194 unsigned long long delta = now - next->timestamp;
3195 if (unlikely((long long)(now - next->timestamp) < 0))
3198 if (next->activated == 1)
3199 delta = delta * (ON_RUNQUEUE_WEIGHT * 128 / 100) / 128;
3201 array = next->array;
3202 new_prio = recalc_task_prio(next, next->timestamp + delta);
3204 if (unlikely(next->prio != new_prio)) {
3205 dequeue_task(next, array);
3206 next->prio = new_prio;
3207 enqueue_task(next, array);
3209 requeue_task(next, array);
3211 next->activated = 0;
3213 #if defined(CONFIG_VSERVER_HARDCPU) && defined(CONFIG_VSERVER_ACB_SCHED)
3214 if (next == rq->idle && !list_empty(&rq->hold_queue)) {
3215 if (min_best_effort_ticks != VX_INVALID_TICKS) {
3216 vx_advance_best_effort_ticks(-min_best_effort_ticks);
3217 goto drain_hold_queue;
3219 if (min_guarantee_ticks != VX_INVALID_TICKS) {
3220 vx_advance_guaranteed_ticks(-min_guarantee_ticks);
3221 goto drain_hold_queue;
3225 if (next == rq->idle)
3226 schedstat_inc(rq, sched_goidle);
3228 prefetch_stack(next);
3229 clear_tsk_need_resched(prev);
3230 rcu_qsctr_inc(task_cpu(prev));
3232 update_cpu_clock(prev, rq, now);
3234 prev->sleep_avg -= run_time;
3235 if ((long)prev->sleep_avg <= 0)
3236 prev->sleep_avg = 0;
3237 prev->timestamp = prev->last_ran = now;
3239 sched_info_switch(prev, next);
3240 if (likely(prev != next)) {
3241 next->timestamp = now;
3246 prepare_task_switch(rq, next);
3247 prev = context_switch(rq, prev, next);
3250 * this_rq must be evaluated again because prev may have moved
3251 * CPUs since it called schedule(), thus the 'rq' on its stack
3252 * frame will be invalid.
3254 finish_task_switch(this_rq(), prev);
3256 spin_unlock_irq(&rq->lock);
3259 if (unlikely(reacquire_kernel_lock(prev) < 0))
3260 goto need_resched_nonpreemptible;
3261 preempt_enable_no_resched();
3262 if (unlikely(test_thread_flag(TIF_NEED_RESCHED)))
3266 EXPORT_SYMBOL(schedule);
3268 #ifdef CONFIG_PREEMPT
3270 * this is is the entry point to schedule() from in-kernel preemption
3271 * off of preempt_enable. Kernel preemptions off return from interrupt
3272 * occur there and call schedule directly.
3274 asmlinkage void __sched preempt_schedule(void)
3276 struct thread_info *ti = current_thread_info();
3277 #ifdef CONFIG_PREEMPT_BKL
3278 struct task_struct *task = current;
3279 int saved_lock_depth;
3282 * If there is a non-zero preempt_count or interrupts are disabled,
3283 * we do not want to preempt the current task. Just return..
3285 if (unlikely(ti->preempt_count || irqs_disabled()))
3289 add_preempt_count(PREEMPT_ACTIVE);
3291 * We keep the big kernel semaphore locked, but we
3292 * clear ->lock_depth so that schedule() doesnt
3293 * auto-release the semaphore:
3295 #ifdef CONFIG_PREEMPT_BKL
3296 saved_lock_depth = task->lock_depth;
3297 task->lock_depth = -1;
3300 #ifdef CONFIG_PREEMPT_BKL
3301 task->lock_depth = saved_lock_depth;
3303 sub_preempt_count(PREEMPT_ACTIVE);
3305 /* we could miss a preemption opportunity between schedule and now */
3307 if (unlikely(test_thread_flag(TIF_NEED_RESCHED)))
3311 EXPORT_SYMBOL(preempt_schedule);
3314 * this is is the entry point to schedule() from kernel preemption
3315 * off of irq context.
3316 * Note, that this is called and return with irqs disabled. This will
3317 * protect us against recursive calling from irq.
3319 asmlinkage void __sched preempt_schedule_irq(void)
3321 struct thread_info *ti = current_thread_info();
3322 #ifdef CONFIG_PREEMPT_BKL
3323 struct task_struct *task = current;
3324 int saved_lock_depth;
3326 /* Catch callers which need to be fixed*/
3327 BUG_ON(ti->preempt_count || !irqs_disabled());
3330 add_preempt_count(PREEMPT_ACTIVE);
3332 * We keep the big kernel semaphore locked, but we
3333 * clear ->lock_depth so that schedule() doesnt
3334 * auto-release the semaphore:
3336 #ifdef CONFIG_PREEMPT_BKL
3337 saved_lock_depth = task->lock_depth;
3338 task->lock_depth = -1;
3342 local_irq_disable();
3343 #ifdef CONFIG_PREEMPT_BKL
3344 task->lock_depth = saved_lock_depth;
3346 sub_preempt_count(PREEMPT_ACTIVE);
3348 /* we could miss a preemption opportunity between schedule and now */
3350 if (unlikely(test_thread_flag(TIF_NEED_RESCHED)))
3354 #endif /* CONFIG_PREEMPT */
3356 int default_wake_function(wait_queue_t *curr, unsigned mode, int sync,
3359 task_t *p = curr->private;
3360 return try_to_wake_up(p, mode, sync);
3363 EXPORT_SYMBOL(default_wake_function);
3366 * The core wakeup function. Non-exclusive wakeups (nr_exclusive == 0) just
3367 * wake everything up. If it's an exclusive wakeup (nr_exclusive == small +ve
3368 * number) then we wake all the non-exclusive tasks and one exclusive task.
3370 * There are circumstances in which we can try to wake a task which has already
3371 * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
3372 * zero in this (rare) case, and we handle it by continuing to scan the queue.
3374 static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
3375 int nr_exclusive, int sync, void *key)
3377 struct list_head *tmp, *next;
3379 list_for_each_safe(tmp, next, &q->task_list) {
3382 curr = list_entry(tmp, wait_queue_t, task_list);
3383 flags = curr->flags;
3384 if (curr->func(curr, mode, sync, key) &&
3385 (flags & WQ_FLAG_EXCLUSIVE) &&
3392 * __wake_up - wake up threads blocked on a waitqueue.
3394 * @mode: which threads
3395 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3396 * @key: is directly passed to the wakeup function
3398 void fastcall __wake_up(wait_queue_head_t *q, unsigned int mode,
3399 int nr_exclusive, void *key)
3401 unsigned long flags;
3403 spin_lock_irqsave(&q->lock, flags);
3404 __wake_up_common(q, mode, nr_exclusive, 0, key);
3405 spin_unlock_irqrestore(&q->lock, flags);
3408 EXPORT_SYMBOL(__wake_up);
3411 * Same as __wake_up but called with the spinlock in wait_queue_head_t held.
3413 void fastcall __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
3415 __wake_up_common(q, mode, 1, 0, NULL);
3419 * __wake_up_sync - wake up threads blocked on a waitqueue.
3421 * @mode: which threads
3422 * @nr_exclusive: how many wake-one or wake-many threads to wake up
3424 * The sync wakeup differs that the waker knows that it will schedule
3425 * away soon, so while the target thread will be woken up, it will not
3426 * be migrated to another CPU - ie. the two threads are 'synchronized'
3427 * with each other. This can prevent needless bouncing between CPUs.
3429 * On UP it can prevent extra preemption.
3432 __wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr_exclusive)
3434 unsigned long flags;
3440 if (unlikely(!nr_exclusive))
3443 spin_lock_irqsave(&q->lock, flags);
3444 __wake_up_common(q, mode, nr_exclusive, sync, NULL);
3445 spin_unlock_irqrestore(&q->lock, flags);
3447 EXPORT_SYMBOL_GPL(__wake_up_sync); /* For internal use only */
3449 void fastcall complete(struct completion *x)
3451 unsigned long flags;
3453 spin_lock_irqsave(&x->wait.lock, flags);
3455 __wake_up_common(&x->wait, TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE,
3457 spin_unlock_irqrestore(&x->wait.lock, flags);
3459 EXPORT_SYMBOL(complete);
3461 void fastcall complete_all(struct completion *x)
3463 unsigned long flags;
3465 spin_lock_irqsave(&x->wait.lock, flags);
3466 x->done += UINT_MAX/2;
3467 __wake_up_common(&x->wait, TASK_UNINTERRUPTIBLE | TASK_INTERRUPTIBLE,
3469 spin_unlock_irqrestore(&x->wait.lock, flags);
3471 EXPORT_SYMBOL(complete_all);
3473 void fastcall __sched wait_for_completion(struct completion *x)
3476 spin_lock_irq(&x->wait.lock);
3478 DECLARE_WAITQUEUE(wait, current);
3480 wait.flags |= WQ_FLAG_EXCLUSIVE;
3481 __add_wait_queue_tail(&x->wait, &wait);
3483 __set_current_state(TASK_UNINTERRUPTIBLE);
3484 spin_unlock_irq(&x->wait.lock);
3486 spin_lock_irq(&x->wait.lock);
3488 __remove_wait_queue(&x->wait, &wait);
3491 spin_unlock_irq(&x->wait.lock);
3493 EXPORT_SYMBOL(wait_for_completion);
3495 unsigned long fastcall __sched
3496 wait_for_completion_timeout(struct completion *x, unsigned long timeout)
3500 spin_lock_irq(&x->wait.lock);
3502 DECLARE_WAITQUEUE(wait, current);
3504 wait.flags |= WQ_FLAG_EXCLUSIVE;
3505 __add_wait_queue_tail(&x->wait, &wait);
3507 __set_current_state(TASK_UNINTERRUPTIBLE);
3508 spin_unlock_irq(&x->wait.lock);
3509 timeout = schedule_timeout(timeout);
3510 spin_lock_irq(&x->wait.lock);
3512 __remove_wait_queue(&x->wait, &wait);
3516 __remove_wait_queue(&x->wait, &wait);
3520 spin_unlock_irq(&x->wait.lock);
3523 EXPORT_SYMBOL(wait_for_completion_timeout);
3525 int fastcall __sched wait_for_completion_interruptible(struct completion *x)
3531 spin_lock_irq(&x->wait.lock);
3533 DECLARE_WAITQUEUE(wait, current);
3535 wait.flags |= WQ_FLAG_EXCLUSIVE;
3536 __add_wait_queue_tail(&x->wait, &wait);
3538 if (signal_pending(current)) {
3540 __remove_wait_queue(&x->wait, &wait);
3543 __set_current_state(TASK_INTERRUPTIBLE);
3544 spin_unlock_irq(&x->wait.lock);
3546 spin_lock_irq(&x->wait.lock);
3548 __remove_wait_queue(&x->wait, &wait);
3552 spin_unlock_irq(&x->wait.lock);
3556 EXPORT_SYMBOL(wait_for_completion_interruptible);
3558 unsigned long fastcall __sched
3559 wait_for_completion_interruptible_timeout(struct completion *x,
3560 unsigned long timeout)
3564 spin_lock_irq(&x->wait.lock);
3566 DECLARE_WAITQUEUE(wait, current);
3568 wait.flags |= WQ_FLAG_EXCLUSIVE;
3569 __add_wait_queue_tail(&x->wait, &wait);
3571 if (signal_pending(current)) {
3572 timeout = -ERESTARTSYS;
3573 __remove_wait_queue(&x->wait, &wait);
3576 __set_current_state(TASK_INTERRUPTIBLE);
3577 spin_unlock_irq(&x->wait.lock);
3578 timeout = schedule_timeout(timeout);
3579 spin_lock_irq(&x->wait.lock);
3581 __remove_wait_queue(&x->wait, &wait);
3585 __remove_wait_queue(&x->wait, &wait);
3589 spin_unlock_irq(&x->wait.lock);
3592 EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
3595 #define SLEEP_ON_VAR \
3596 unsigned long flags; \
3597 wait_queue_t wait; \
3598 init_waitqueue_entry(&wait, current);
3600 #define SLEEP_ON_HEAD \
3601 spin_lock_irqsave(&q->lock,flags); \
3602 __add_wait_queue(q, &wait); \
3603 spin_unlock(&q->lock);
3605 #define SLEEP_ON_TAIL \
3606 spin_lock_irq(&q->lock); \
3607 __remove_wait_queue(q, &wait); \
3608 spin_unlock_irqrestore(&q->lock, flags);
3610 void fastcall __sched interruptible_sleep_on(wait_queue_head_t *q)
3614 current->state = TASK_INTERRUPTIBLE;
3621 EXPORT_SYMBOL(interruptible_sleep_on);
3623 long fastcall __sched
3624 interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
3628 current->state = TASK_INTERRUPTIBLE;
3631 timeout = schedule_timeout(timeout);
3637 EXPORT_SYMBOL(interruptible_sleep_on_timeout);
3639 void fastcall __sched sleep_on(wait_queue_head_t *q)
3643 current->state = TASK_UNINTERRUPTIBLE;
3650 EXPORT_SYMBOL(sleep_on);
3652 long fastcall __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
3656 current->state = TASK_UNINTERRUPTIBLE;
3659 timeout = schedule_timeout(timeout);
3665 EXPORT_SYMBOL(sleep_on_timeout);
3667 void set_user_nice(task_t *p, long nice)
3669 unsigned long flags;
3670 prio_array_t *array;
3672 int old_prio, new_prio, delta;
3674 if (TASK_NICE(p) == nice || nice < -20 || nice > 19)
3677 * We have to be careful, if called from sys_setpriority(),
3678 * the task might be in the middle of scheduling on another CPU.
3680 rq = task_rq_lock(p, &flags);
3682 * The RT priorities are set via sched_setscheduler(), but we still
3683 * allow the 'normal' nice value to be set - but as expected
3684 * it wont have any effect on scheduling until the task is
3685 * not SCHED_NORMAL/SCHED_BATCH:
3688 p->static_prio = NICE_TO_PRIO(nice);
3693 dequeue_task(p, array);
3696 new_prio = NICE_TO_PRIO(nice);
3697 delta = new_prio - old_prio;
3698 p->static_prio = NICE_TO_PRIO(nice);
3702 enqueue_task(p, array);
3704 * If the task increased its priority or is running and
3705 * lowered its priority, then reschedule its CPU:
3707 if (delta < 0 || (delta > 0 && task_running(rq, p)))
3708 resched_task(rq->curr);
3711 task_rq_unlock(rq, &flags);
3714 EXPORT_SYMBOL(set_user_nice);
3717 * can_nice - check if a task can reduce its nice value
3721 int can_nice(const task_t *p, const int nice)
3723 /* convert nice value [19,-20] to rlimit style value [1,40] */
3724 int nice_rlim = 20 - nice;
3725 return (nice_rlim <= p->signal->rlim[RLIMIT_NICE].rlim_cur ||
3726 capable(CAP_SYS_NICE));
3729 #ifdef __ARCH_WANT_SYS_NICE
3732 * sys_nice - change the priority of the current process.
3733 * @increment: priority increment
3735 * sys_setpriority is a more generic, but much slower function that
3736 * does similar things.
3738 asmlinkage long sys_nice(int increment)
3744 * Setpriority might change our priority at the same moment.
3745 * We don't have to worry. Conceptually one call occurs first
3746 * and we have a single winner.
3748 if (increment < -40)
3753 nice = PRIO_TO_NICE(current->static_prio) + increment;
3759 if (increment < 0 && !can_nice(current, nice))
3760 return vx_flags(VXF_IGNEG_NICE, 0) ? 0 : -EPERM;
3762 retval = security_task_setnice(current, nice);
3766 set_user_nice(current, nice);
3773 * task_prio - return the priority value of a given task.
3774 * @p: the task in question.
3776 * This is the priority value as seen by users in /proc.
3777 * RT tasks are offset by -200. Normal tasks are centered
3778 * around 0, value goes from -16 to +15.
3780 int task_prio(const task_t *p)
3782 return p->prio - MAX_RT_PRIO;
3786 * task_nice - return the nice value of a given task.
3787 * @p: the task in question.
3789 int task_nice(const task_t *p)
3791 return TASK_NICE(p);
3793 EXPORT_SYMBOL_GPL(task_nice);
3796 * idle_cpu - is a given cpu idle currently?
3797 * @cpu: the processor in question.
3799 int idle_cpu(int cpu)
3801 return cpu_curr(cpu) == cpu_rq(cpu)->idle;
3805 * idle_task - return the idle task for a given cpu.
3806 * @cpu: the processor in question.
3808 task_t *idle_task(int cpu)
3810 return cpu_rq(cpu)->idle;
3814 * find_process_by_pid - find a process with a matching PID value.
3815 * @pid: the pid in question.
3817 static inline task_t *find_process_by_pid(pid_t pid)
3819 return pid ? find_task_by_pid(pid) : current;
3822 /* Actually do priority change: must hold rq lock. */
3823 static void __setscheduler(struct task_struct *p, int policy, int prio)
3827 p->rt_priority = prio;
3828 if (policy != SCHED_NORMAL && policy != SCHED_BATCH) {
3829 p->prio = MAX_RT_PRIO-1 - p->rt_priority;
3831 p->prio = p->static_prio;
3833 * SCHED_BATCH tasks are treated as perpetual CPU hogs:
3835 if (policy == SCHED_BATCH)
3841 * sched_setscheduler - change the scheduling policy and/or RT priority of
3843 * @p: the task in question.
3844 * @policy: new policy.
3845 * @param: structure containing the new RT priority.
3847 int sched_setscheduler(struct task_struct *p, int policy,
3848 struct sched_param *param)
3851 int oldprio, oldpolicy = -1;
3852 prio_array_t *array;
3853 unsigned long flags;
3857 /* double check policy once rq lock held */
3859 policy = oldpolicy = p->policy;
3860 else if (policy != SCHED_FIFO && policy != SCHED_RR &&
3861 policy != SCHED_NORMAL && policy != SCHED_BATCH)
3864 * Valid priorities for SCHED_FIFO and SCHED_RR are
3865 * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL and
3868 if (param->sched_priority < 0 ||
3869 (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
3870 (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
3872 if ((policy == SCHED_NORMAL || policy == SCHED_BATCH)
3873 != (param->sched_priority == 0))
3877 * Allow unprivileged RT tasks to decrease priority:
3879 if (!capable(CAP_SYS_NICE)) {
3881 * can't change policy, except between SCHED_NORMAL
3884 if (((policy != SCHED_NORMAL && p->policy != SCHED_BATCH) &&
3885 (policy != SCHED_BATCH && p->policy != SCHED_NORMAL)) &&
3886 !p->signal->rlim[RLIMIT_RTPRIO].rlim_cur)
3888 /* can't increase priority */
3889 if ((policy != SCHED_NORMAL && policy != SCHED_BATCH) &&
3890 param->sched_priority > p->rt_priority &&
3891 param->sched_priority >
3892 p->signal->rlim[RLIMIT_RTPRIO].rlim_cur)
3894 /* can't change other user's priorities */
3895 if ((current->euid != p->euid) &&
3896 (current->euid != p->uid))
3900 retval = security_task_setscheduler(p, policy, param);
3904 * To be able to change p->policy safely, the apropriate
3905 * runqueue lock must be held.
3907 rq = task_rq_lock(p, &flags);
3908 /* recheck policy now with rq lock held */
3909 if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
3910 policy = oldpolicy = -1;
3911 task_rq_unlock(rq, &flags);
3916 deactivate_task(p, rq);
3918 __setscheduler(p, policy, param->sched_priority);
3920 vx_activate_task(p);
3921 __activate_task(p, rq);
3923 * Reschedule if we are currently running on this runqueue and
3924 * our priority decreased, or if we are not currently running on
3925 * this runqueue and our priority is higher than the current's
3927 if (task_running(rq, p)) {
3928 if (p->prio > oldprio)
3929 resched_task(rq->curr);
3930 } else if (TASK_PREEMPTS_CURR(p, rq))
3931 resched_task(rq->curr);
3933 task_rq_unlock(rq, &flags);
3936 EXPORT_SYMBOL_GPL(sched_setscheduler);
3939 do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
3942 struct sched_param lparam;
3943 struct task_struct *p;
3945 if (!param || pid < 0)
3947 if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
3949 read_lock_irq(&tasklist_lock);
3950 p = find_process_by_pid(pid);
3952 read_unlock_irq(&tasklist_lock);
3955 retval = sched_setscheduler(p, policy, &lparam);
3956 read_unlock_irq(&tasklist_lock);
3961 * sys_sched_setscheduler - set/change the scheduler policy and RT priority
3962 * @pid: the pid in question.
3963 * @policy: new policy.
3964 * @param: structure containing the new RT priority.
3966 asmlinkage long sys_sched_setscheduler(pid_t pid, int policy,
3967 struct sched_param __user *param)
3969 /* negative values for policy are not valid */
3973 return do_sched_setscheduler(pid, policy, param);
3977 * sys_sched_setparam - set/change the RT priority of a thread
3978 * @pid: the pid in question.
3979 * @param: structure containing the new RT priority.
3981 asmlinkage long sys_sched_setparam(pid_t pid, struct sched_param __user *param)
3983 return do_sched_setscheduler(pid, -1, param);
3987 * sys_sched_getscheduler - get the policy (scheduling class) of a thread
3988 * @pid: the pid in question.
3990 asmlinkage long sys_sched_getscheduler(pid_t pid)
3992 int retval = -EINVAL;
3999 read_lock(&tasklist_lock);
4000 p = find_process_by_pid(pid);
4002 retval = security_task_getscheduler(p);
4006 read_unlock(&tasklist_lock);
4013 * sys_sched_getscheduler - get the RT priority of a thread
4014 * @pid: the pid in question.
4015 * @param: structure containing the RT priority.
4017 asmlinkage long sys_sched_getparam(pid_t pid, struct sched_param __user *param)
4019 struct sched_param lp;
4020 int retval = -EINVAL;
4023 if (!param || pid < 0)
4026 read_lock(&tasklist_lock);
4027 p = find_process_by_pid(pid);
4032 retval = security_task_getscheduler(p);
4036 lp.sched_priority = p->rt_priority;
4037 read_unlock(&tasklist_lock);
4040 * This one might sleep, we cannot do it with a spinlock held ...
4042 retval = copy_to_user(param, &lp, sizeof(*param)) ? -EFAULT : 0;
4048 read_unlock(&tasklist_lock);
4052 long sched_setaffinity(pid_t pid, cpumask_t new_mask)
4056 cpumask_t cpus_allowed;
4059 read_lock(&tasklist_lock);
4061 p = find_process_by_pid(pid);
4063 read_unlock(&tasklist_lock);
4064 unlock_cpu_hotplug();
4069 * It is not safe to call set_cpus_allowed with the
4070 * tasklist_lock held. We will bump the task_struct's
4071 * usage count and then drop tasklist_lock.
4074 read_unlock(&tasklist_lock);
4077 if ((current->euid != p->euid) && (current->euid != p->uid) &&
4078 !capable(CAP_SYS_NICE))
4081 cpus_allowed = cpuset_cpus_allowed(p);
4082 cpus_and(new_mask, new_mask, cpus_allowed);
4083 retval = set_cpus_allowed(p, new_mask);
4087 unlock_cpu_hotplug();
4091 static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
4092 cpumask_t *new_mask)
4094 if (len < sizeof(cpumask_t)) {
4095 memset(new_mask, 0, sizeof(cpumask_t));
4096 } else if (len > sizeof(cpumask_t)) {
4097 len = sizeof(cpumask_t);
4099 return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
4103 * sys_sched_setaffinity - set the cpu affinity of a process
4104 * @pid: pid of the process
4105 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
4106 * @user_mask_ptr: user-space pointer to the new cpu mask
4108 asmlinkage long sys_sched_setaffinity(pid_t pid, unsigned int len,
4109 unsigned long __user *user_mask_ptr)
4114 retval = get_user_cpu_mask(user_mask_ptr, len, &new_mask);
4118 return sched_setaffinity(pid, new_mask);
4122 * Represents all cpu's present in the system
4123 * In systems capable of hotplug, this map could dynamically grow
4124 * as new cpu's are detected in the system via any platform specific
4125 * method, such as ACPI for e.g.
4128 cpumask_t cpu_present_map __read_mostly;
4129 EXPORT_SYMBOL(cpu_present_map);
4132 cpumask_t cpu_online_map __read_mostly = CPU_MASK_ALL;
4133 cpumask_t cpu_possible_map __read_mostly = CPU_MASK_ALL;
4136 long sched_getaffinity(pid_t pid, cpumask_t *mask)
4142 read_lock(&tasklist_lock);
4145 p = find_process_by_pid(pid);
4150 cpus_and(*mask, p->cpus_allowed, cpu_online_map);
4153 read_unlock(&tasklist_lock);
4154 unlock_cpu_hotplug();
4162 * sys_sched_getaffinity - get the cpu affinity of a process
4163 * @pid: pid of the process
4164 * @len: length in bytes of the bitmask pointed to by user_mask_ptr
4165 * @user_mask_ptr: user-space pointer to hold the current cpu mask
4167 asmlinkage long sys_sched_getaffinity(pid_t pid, unsigned int len,
4168 unsigned long __user *user_mask_ptr)
4173 if (len < sizeof(cpumask_t))
4176 ret = sched_getaffinity(pid, &mask);
4180 if (copy_to_user(user_mask_ptr, &mask, sizeof(cpumask_t)))
4183 return sizeof(cpumask_t);
4187 * sys_sched_yield - yield the current processor to other threads.
4189 * this function yields the current CPU by moving the calling thread
4190 * to the expired array. If there are no other threads running on this
4191 * CPU then this function will return.
4193 asmlinkage long sys_sched_yield(void)
4195 runqueue_t *rq = this_rq_lock();
4196 prio_array_t *array = current->array;
4197 prio_array_t *target = rq->expired;
4199 schedstat_inc(rq, yld_cnt);
4201 * We implement yielding by moving the task into the expired
4204 * (special rule: RT tasks will just roundrobin in the active
4207 if (rt_task(current))
4208 target = rq->active;
4210 if (array->nr_active == 1) {
4211 schedstat_inc(rq, yld_act_empty);
4212 if (!rq->expired->nr_active)
4213 schedstat_inc(rq, yld_both_empty);
4214 } else if (!rq->expired->nr_active)
4215 schedstat_inc(rq, yld_exp_empty);
4217 if (array != target) {
4218 dequeue_task(current, array);
4219 enqueue_task(current, target);
4222 * requeue_task is cheaper so perform that if possible.
4224 requeue_task(current, array);
4227 * Since we are going to call schedule() anyway, there's
4228 * no need to preempt or enable interrupts:
4230 __release(rq->lock);
4231 _raw_spin_unlock(&rq->lock);
4232 preempt_enable_no_resched();
4239 static inline void __cond_resched(void)
4242 * The BKS might be reacquired before we have dropped
4243 * PREEMPT_ACTIVE, which could trigger a second
4244 * cond_resched() call.
4246 if (unlikely(preempt_count()))
4248 if (unlikely(system_state != SYSTEM_RUNNING))
4251 add_preempt_count(PREEMPT_ACTIVE);
4253 sub_preempt_count(PREEMPT_ACTIVE);
4254 } while (need_resched());
4257 int __sched cond_resched(void)
4259 if (need_resched()) {
4266 EXPORT_SYMBOL(cond_resched);
4269 * cond_resched_lock() - if a reschedule is pending, drop the given lock,
4270 * call schedule, and on return reacquire the lock.
4272 * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
4273 * operations here to prevent schedule() from being called twice (once via
4274 * spin_unlock(), once by hand).
4276 int cond_resched_lock(spinlock_t *lock)
4280 if (need_lockbreak(lock)) {
4286 if (need_resched()) {
4287 _raw_spin_unlock(lock);
4288 preempt_enable_no_resched();
4296 EXPORT_SYMBOL(cond_resched_lock);
4298 int __sched cond_resched_softirq(void)
4300 BUG_ON(!in_softirq());
4302 if (need_resched()) {
4303 __local_bh_enable();
4311 EXPORT_SYMBOL(cond_resched_softirq);
4315 * yield - yield the current processor to other threads.
4317 * this is a shortcut for kernel-space yielding - it marks the
4318 * thread runnable and calls sys_sched_yield().
4320 void __sched yield(void)
4322 set_current_state(TASK_RUNNING);
4326 EXPORT_SYMBOL(yield);
4329 * This task is about to go to sleep on IO. Increment rq->nr_iowait so
4330 * that process accounting knows that this is a task in IO wait state.
4332 * But don't do that if it is a deliberate, throttling IO wait (this task
4333 * has set its backing_dev_info: the queue against which it should throttle)
4335 void __sched io_schedule(void)
4337 struct runqueue *rq = &per_cpu(runqueues, raw_smp_processor_id());
4339 atomic_inc(&rq->nr_iowait);
4341 atomic_dec(&rq->nr_iowait);
4344 EXPORT_SYMBOL(io_schedule);
4346 long __sched io_schedule_timeout(long timeout)
4348 struct runqueue *rq = &per_cpu(runqueues, raw_smp_processor_id());
4351 atomic_inc(&rq->nr_iowait);
4352 ret = schedule_timeout(timeout);
4353 atomic_dec(&rq->nr_iowait);
4358 * sys_sched_get_priority_max - return maximum RT priority.
4359 * @policy: scheduling class.
4361 * this syscall returns the maximum rt_priority that can be used
4362 * by a given scheduling class.
4364 asmlinkage long sys_sched_get_priority_max(int policy)
4371 ret = MAX_USER_RT_PRIO-1;
4382 * sys_sched_get_priority_min - return minimum RT priority.
4383 * @policy: scheduling class.
4385 * this syscall returns the minimum rt_priority that can be used
4386 * by a given scheduling class.
4388 asmlinkage long sys_sched_get_priority_min(int policy)
4405 * sys_sched_rr_get_interval - return the default timeslice of a process.
4406 * @pid: pid of the process.
4407 * @interval: userspace pointer to the timeslice value.
4409 * this syscall writes the default timeslice value of a given process
4410 * into the user-space timespec buffer. A value of '0' means infinity.
4413 long sys_sched_rr_get_interval(pid_t pid, struct timespec __user *interval)
4415 int retval = -EINVAL;
4423 read_lock(&tasklist_lock);
4424 p = find_process_by_pid(pid);
4428 retval = security_task_getscheduler(p);
4432 jiffies_to_timespec(p->policy & SCHED_FIFO ?
4433 0 : task_timeslice(p), &t);
4434 read_unlock(&tasklist_lock);
4435 retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
4439 read_unlock(&tasklist_lock);
4443 static inline struct task_struct *eldest_child(struct task_struct *p)
4445 if (list_empty(&p->children)) return NULL;
4446 return list_entry(p->children.next,struct task_struct,sibling);
4449 static inline struct task_struct *older_sibling(struct task_struct *p)
4451 if (p->sibling.prev==&p->parent->children) return NULL;
4452 return list_entry(p->sibling.prev,struct task_struct,sibling);
4455 static inline struct task_struct *younger_sibling(struct task_struct *p)
4457 if (p->sibling.next==&p->parent->children) return NULL;
4458 return list_entry(p->sibling.next,struct task_struct,sibling);
4461 static void show_task(task_t *p)
4465 unsigned long free = 0;
4466 static const char *stat_nam[] = { "R", "S", "D", "T", "t", "Z", "X" };
4468 printk("%-13.13s ", p->comm);
4469 state = p->state ? __ffs(p->state) + 1 : 0;
4470 if (state < ARRAY_SIZE(stat_nam))
4471 printk(stat_nam[state]);
4474 #if (BITS_PER_LONG == 32)
4475 if (state == TASK_RUNNING)
4476 printk(" running ");
4478 printk(" %08lX ", thread_saved_pc(p));
4480 if (state == TASK_RUNNING)
4481 printk(" running task ");
4483 printk(" %016lx ", thread_saved_pc(p));
4485 #ifdef CONFIG_DEBUG_STACK_USAGE
4487 unsigned long *n = end_of_stack(p);
4490 free = (unsigned long)n - (unsigned long)end_of_stack(p);
4493 printk("%5lu %5d %6d ", free, p->pid, p->parent->pid);
4494 if ((relative = eldest_child(p)))
4495 printk("%5d ", relative->pid);
4498 if ((relative = younger_sibling(p)))
4499 printk("%7d", relative->pid);
4502 if ((relative = older_sibling(p)))
4503 printk(" %5d", relative->pid);
4507 printk(" (L-TLB)\n");
4509 printk(" (NOTLB)\n");
4511 if (state != TASK_RUNNING)
4512 show_stack(p, NULL);
4515 void show_state(void)
4519 #if (BITS_PER_LONG == 32)
4522 printk(" task PC pid father child younger older\n");
4526 printk(" task PC pid father child younger older\n");
4528 read_lock(&tasklist_lock);
4529 do_each_thread(g, p) {
4531 * reset the NMI-timeout, listing all files on a slow
4532 * console might take alot of time:
4534 touch_nmi_watchdog();
4536 } while_each_thread(g, p);
4538 read_unlock(&tasklist_lock);
4539 mutex_debug_show_all_locks();
4543 * init_idle - set up an idle thread for a given CPU
4544 * @idle: task in question
4545 * @cpu: cpu the idle task belongs to
4547 * NOTE: this function does not set the idle thread's NEED_RESCHED
4548 * flag, to make booting more robust.
4550 void __devinit init_idle(task_t *idle, int cpu)
4552 runqueue_t *rq = cpu_rq(cpu);
4553 unsigned long flags;
4555 idle->timestamp = sched_clock();
4556 idle->sleep_avg = 0;
4558 idle->prio = MAX_PRIO;
4559 idle->state = TASK_RUNNING;
4560 idle->cpus_allowed = cpumask_of_cpu(cpu);
4561 set_task_cpu(idle, cpu);
4563 spin_lock_irqsave(&rq->lock, flags);
4564 rq->curr = rq->idle = idle;
4565 #if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
4568 spin_unlock_irqrestore(&rq->lock, flags);
4570 /* Set the preempt count _outside_ the spinlocks! */
4571 #if defined(CONFIG_PREEMPT) && !defined(CONFIG_PREEMPT_BKL)
4572 task_thread_info(idle)->preempt_count = (idle->lock_depth >= 0);
4574 task_thread_info(idle)->preempt_count = 0;
4579 * In a system that switches off the HZ timer nohz_cpu_mask
4580 * indicates which cpus entered this state. This is used
4581 * in the rcu update to wait only for active cpus. For system
4582 * which do not switch off the HZ timer nohz_cpu_mask should
4583 * always be CPU_MASK_NONE.
4585 cpumask_t nohz_cpu_mask = CPU_MASK_NONE;
4589 * This is how migration works:
4591 * 1) we queue a migration_req_t structure in the source CPU's
4592 * runqueue and wake up that CPU's migration thread.
4593 * 2) we down() the locked semaphore => thread blocks.
4594 * 3) migration thread wakes up (implicitly it forces the migrated
4595 * thread off the CPU)
4596 * 4) it gets the migration request and checks whether the migrated
4597 * task is still in the wrong runqueue.
4598 * 5) if it's in the wrong runqueue then the migration thread removes
4599 * it and puts it into the right queue.
4600 * 6) migration thread up()s the semaphore.
4601 * 7) we wake up and the migration is done.
4605 * Change a given task's CPU affinity. Migrate the thread to a
4606 * proper CPU and schedule it away if the CPU it's executing on
4607 * is removed from the allowed bitmask.
4609 * NOTE: the caller must have a valid reference to the task, the
4610 * task must not exit() & deallocate itself prematurely. The
4611 * call is not atomic; no spinlocks may be held.
4613 int set_cpus_allowed(task_t *p, cpumask_t new_mask)
4615 unsigned long flags;
4617 migration_req_t req;
4620 rq = task_rq_lock(p, &flags);
4621 if (!cpus_intersects(new_mask, cpu_online_map)) {
4626 p->cpus_allowed = new_mask;
4627 /* Can the task run on the task's current CPU? If so, we're done */
4628 if (cpu_isset(task_cpu(p), new_mask))
4631 if (migrate_task(p, any_online_cpu(new_mask), &req)) {
4632 /* Need help from migration thread: drop lock and wait. */
4633 task_rq_unlock(rq, &flags);
4634 wake_up_process(rq->migration_thread);
4635 wait_for_completion(&req.done);
4636 tlb_migrate_finish(p->mm);
4640 task_rq_unlock(rq, &flags);
4644 EXPORT_SYMBOL_GPL(set_cpus_allowed);
4647 * Move (not current) task off this cpu, onto dest cpu. We're doing
4648 * this because either it can't run here any more (set_cpus_allowed()
4649 * away from this CPU, or CPU going down), or because we're
4650 * attempting to rebalance this task on exec (sched_exec).
4652 * So we race with normal scheduler movements, but that's OK, as long
4653 * as the task is no longer on this CPU.
4655 static void __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
4657 runqueue_t *rq_dest, *rq_src;
4659 if (unlikely(cpu_is_offline(dest_cpu)))
4662 rq_src = cpu_rq(src_cpu);
4663 rq_dest = cpu_rq(dest_cpu);
4665 double_rq_lock(rq_src, rq_dest);
4666 /* Already moved. */
4667 if (task_cpu(p) != src_cpu)
4669 /* Affinity changed (again). */
4670 if (!cpu_isset(dest_cpu, p->cpus_allowed))
4673 set_task_cpu(p, dest_cpu);
4676 * Sync timestamp with rq_dest's before activating.
4677 * The same thing could be achieved by doing this step
4678 * afterwards, and pretending it was a local activate.
4679 * This way is cleaner and logically correct.
4681 p->timestamp = p->timestamp - rq_src->timestamp_last_tick
4682 + rq_dest->timestamp_last_tick;
4683 deactivate_task(p, rq_src);
4684 activate_task(p, rq_dest, 0);
4685 if (TASK_PREEMPTS_CURR(p, rq_dest))
4686 resched_task(rq_dest->curr);
4690 double_rq_unlock(rq_src, rq_dest);
4694 * migration_thread - this is a highprio system thread that performs
4695 * thread migration by bumping thread off CPU then 'pushing' onto
4698 static int migration_thread(void *data)
4701 int cpu = (long)data;
4704 BUG_ON(rq->migration_thread != current);
4706 set_current_state(TASK_INTERRUPTIBLE);
4707 while (!kthread_should_stop()) {
4708 struct list_head *head;
4709 migration_req_t *req;
4713 spin_lock_irq(&rq->lock);
4715 if (cpu_is_offline(cpu)) {
4716 spin_unlock_irq(&rq->lock);
4720 if (rq->active_balance) {
4721 active_load_balance(rq, cpu);
4722 rq->active_balance = 0;
4725 head = &rq->migration_queue;
4727 if (list_empty(head)) {
4728 spin_unlock_irq(&rq->lock);
4730 set_current_state(TASK_INTERRUPTIBLE);
4733 req = list_entry(head->next, migration_req_t, list);
4734 list_del_init(head->next);
4736 spin_unlock(&rq->lock);
4737 __migrate_task(req->task, cpu, req->dest_cpu);
4740 complete(&req->done);
4742 __set_current_state(TASK_RUNNING);
4746 /* Wait for kthread_stop */
4747 set_current_state(TASK_INTERRUPTIBLE);
4748 while (!kthread_should_stop()) {
4750 set_current_state(TASK_INTERRUPTIBLE);
4752 __set_current_state(TASK_RUNNING);
4756 #ifdef CONFIG_HOTPLUG_CPU
4757 /* Figure out where task on dead CPU should go, use force if neccessary. */
4758 static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *tsk)
4764 mask = node_to_cpumask(cpu_to_node(dead_cpu));
4765 cpus_and(mask, mask, tsk->cpus_allowed);
4766 dest_cpu = any_online_cpu(mask);
4768 /* On any allowed CPU? */
4769 if (dest_cpu == NR_CPUS)
4770 dest_cpu = any_online_cpu(tsk->cpus_allowed);
4772 /* No more Mr. Nice Guy. */
4773 if (dest_cpu == NR_CPUS) {
4774 cpus_setall(tsk->cpus_allowed);
4775 dest_cpu = any_online_cpu(tsk->cpus_allowed);
4778 * Don't tell them about moving exiting tasks or
4779 * kernel threads (both mm NULL), since they never
4782 if (tsk->mm && printk_ratelimit())
4783 printk(KERN_INFO "process %d (%s) no "
4784 "longer affine to cpu%d\n",
4785 tsk->pid, tsk->comm, dead_cpu);
4787 __migrate_task(tsk, dead_cpu, dest_cpu);
4791 * While a dead CPU has no uninterruptible tasks queued at this point,
4792 * it might still have a nonzero ->nr_uninterruptible counter, because
4793 * for performance reasons the counter is not stricly tracking tasks to
4794 * their home CPUs. So we just add the counter to another CPU's counter,
4795 * to keep the global sum constant after CPU-down:
4797 static void migrate_nr_uninterruptible(runqueue_t *rq_src)
4799 runqueue_t *rq_dest = cpu_rq(any_online_cpu(CPU_MASK_ALL));
4800 unsigned long flags;
4802 local_irq_save(flags);
4803 double_rq_lock(rq_src, rq_dest);
4804 rq_dest->nr_uninterruptible += rq_src->nr_uninterruptible;
4805 rq_src->nr_uninterruptible = 0;
4806 double_rq_unlock(rq_src, rq_dest);
4807 local_irq_restore(flags);
4810 /* Run through task list and migrate tasks from the dead cpu. */
4811 static void migrate_live_tasks(int src_cpu)
4813 struct task_struct *tsk, *t;
4815 write_lock_irq(&tasklist_lock);
4817 do_each_thread(t, tsk) {
4821 if (task_cpu(tsk) == src_cpu)
4822 move_task_off_dead_cpu(src_cpu, tsk);
4823 } while_each_thread(t, tsk);
4825 write_unlock_irq(&tasklist_lock);
4828 /* Schedules idle task to be the next runnable task on current CPU.
4829 * It does so by boosting its priority to highest possible and adding it to
4830 * the _front_ of runqueue. Used by CPU offline code.
4832 void sched_idle_next(void)
4834 int cpu = smp_processor_id();
4835 runqueue_t *rq = this_rq();
4836 struct task_struct *p = rq->idle;
4837 unsigned long flags;
4839 /* cpu has to be offline */
4840 BUG_ON(cpu_online(cpu));
4842 /* Strictly not necessary since rest of the CPUs are stopped by now
4843 * and interrupts disabled on current cpu.
4845 spin_lock_irqsave(&rq->lock, flags);
4847 __setscheduler(p, SCHED_FIFO, MAX_RT_PRIO-1);
4848 /* Add idle task to _front_ of it's priority queue */
4849 __activate_idle_task(p, rq);
4851 spin_unlock_irqrestore(&rq->lock, flags);
4854 /* Ensures that the idle task is using init_mm right before its cpu goes
4857 void idle_task_exit(void)
4859 struct mm_struct *mm = current->active_mm;
4861 BUG_ON(cpu_online(smp_processor_id()));
4864 switch_mm(mm, &init_mm, current);
4868 static void migrate_dead(unsigned int dead_cpu, task_t *tsk)
4870 struct runqueue *rq = cpu_rq(dead_cpu);
4872 /* Must be exiting, otherwise would be on tasklist. */
4873 BUG_ON(tsk->exit_state != EXIT_ZOMBIE && tsk->exit_state != EXIT_DEAD);
4875 /* Cannot have done final schedule yet: would have vanished. */
4876 BUG_ON(tsk->flags & PF_DEAD);
4878 get_task_struct(tsk);
4881 * Drop lock around migration; if someone else moves it,
4882 * that's OK. No task can be added to this CPU, so iteration is
4885 spin_unlock_irq(&rq->lock);
4886 move_task_off_dead_cpu(dead_cpu, tsk);
4887 spin_lock_irq(&rq->lock);
4889 put_task_struct(tsk);
4892 /* release_task() removes task from tasklist, so we won't find dead tasks. */
4893 static void migrate_dead_tasks(unsigned int dead_cpu)
4896 struct runqueue *rq = cpu_rq(dead_cpu);
4898 for (arr = 0; arr < 2; arr++) {
4899 for (i = 0; i < MAX_PRIO; i++) {
4900 struct list_head *list = &rq->arrays[arr].queue[i];
4901 while (!list_empty(list))
4902 migrate_dead(dead_cpu,
4903 list_entry(list->next, task_t,
4908 #endif /* CONFIG_HOTPLUG_CPU */
4911 * migration_call - callback that gets triggered when a CPU is added.
4912 * Here we can start up the necessary migration thread for the new CPU.
4914 static int migration_call(struct notifier_block *nfb, unsigned long action,
4917 int cpu = (long)hcpu;
4918 struct task_struct *p;
4919 struct runqueue *rq;
4920 unsigned long flags;
4923 case CPU_UP_PREPARE:
4924 p = kthread_create(migration_thread, hcpu, "migration/%d",cpu);
4927 p->flags |= PF_NOFREEZE;
4928 kthread_bind(p, cpu);
4929 /* Must be high prio: stop_machine expects to yield to it. */
4930 rq = task_rq_lock(p, &flags);
4931 __setscheduler(p, SCHED_FIFO, MAX_RT_PRIO-1);
4932 task_rq_unlock(rq, &flags);
4933 cpu_rq(cpu)->migration_thread = p;
4936 /* Strictly unneccessary, as first user will wake it. */
4937 wake_up_process(cpu_rq(cpu)->migration_thread);
4939 #ifdef CONFIG_HOTPLUG_CPU
4940 case CPU_UP_CANCELED:
4941 /* Unbind it from offline cpu so it can run. Fall thru. */
4942 kthread_bind(cpu_rq(cpu)->migration_thread,
4943 any_online_cpu(cpu_online_map));
4944 kthread_stop(cpu_rq(cpu)->migration_thread);
4945 cpu_rq(cpu)->migration_thread = NULL;
4948 migrate_live_tasks(cpu);
4950 kthread_stop(rq->migration_thread);
4951 rq->migration_thread = NULL;
4952 /* Idle task back to normal (off runqueue, low prio) */
4953 rq = task_rq_lock(rq->idle, &flags);
4954 deactivate_task(rq->idle, rq);
4955 rq->idle->static_prio = MAX_PRIO;
4956 __setscheduler(rq->idle, SCHED_NORMAL, 0);
4957 migrate_dead_tasks(cpu);
4958 task_rq_unlock(rq, &flags);
4959 migrate_nr_uninterruptible(rq);
4960 BUG_ON(rq->nr_running != 0);
4962 /* No need to migrate the tasks: it was best-effort if
4963 * they didn't do lock_cpu_hotplug(). Just wake up
4964 * the requestors. */
4965 spin_lock_irq(&rq->lock);
4966 while (!list_empty(&rq->migration_queue)) {
4967 migration_req_t *req;
4968 req = list_entry(rq->migration_queue.next,
4969 migration_req_t, list);
4970 list_del_init(&req->list);
4971 complete(&req->done);
4973 spin_unlock_irq(&rq->lock);
4980 /* Register at highest priority so that task migration (migrate_all_tasks)
4981 * happens before everything else.
4983 static struct notifier_block __devinitdata migration_notifier = {
4984 .notifier_call = migration_call,
4988 int __init migration_init(void)
4990 void *cpu = (void *)(long)smp_processor_id();
4991 /* Start one for boot CPU. */
4992 migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
4993 migration_call(&migration_notifier, CPU_ONLINE, cpu);
4994 register_cpu_notifier(&migration_notifier);
5000 #undef SCHED_DOMAIN_DEBUG
5001 #ifdef SCHED_DOMAIN_DEBUG
5002 static void sched_domain_debug(struct sched_domain *sd, int cpu)
5007 printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
5011 printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);
5016 struct sched_group *group = sd->groups;
5017 cpumask_t groupmask;
5019 cpumask_scnprintf(str, NR_CPUS, sd->span);
5020 cpus_clear(groupmask);
5023 for (i = 0; i < level + 1; i++)
5025 printk("domain %d: ", level);
5027 if (!(sd->flags & SD_LOAD_BALANCE)) {
5028 printk("does not load-balance\n");
5030 printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain has parent");
5034 printk("span %s\n", str);
5036 if (!cpu_isset(cpu, sd->span))
5037 printk(KERN_ERR "ERROR: domain->span does not contain CPU%d\n", cpu);
5038 if (!cpu_isset(cpu, group->cpumask))
5039 printk(KERN_ERR "ERROR: domain->groups does not contain CPU%d\n", cpu);
5042 for (i = 0; i < level + 2; i++)
5048 printk(KERN_ERR "ERROR: group is NULL\n");
5052 if (!group->cpu_power) {
5054 printk(KERN_ERR "ERROR: domain->cpu_power not set\n");
5057 if (!cpus_weight(group->cpumask)) {
5059 printk(KERN_ERR "ERROR: empty group\n");
5062 if (cpus_intersects(groupmask, group->cpumask)) {
5064 printk(KERN_ERR "ERROR: repeated CPUs\n");
5067 cpus_or(groupmask, groupmask, group->cpumask);
5069 cpumask_scnprintf(str, NR_CPUS, group->cpumask);
5072 group = group->next;
5073 } while (group != sd->groups);
5076 if (!cpus_equal(sd->span, groupmask))
5077 printk(KERN_ERR "ERROR: groups don't span domain->span\n");
5083 if (!cpus_subset(groupmask, sd->span))
5084 printk(KERN_ERR "ERROR: parent span is not a superset of domain->span\n");
5090 #define sched_domain_debug(sd, cpu) {}
5093 static int sd_degenerate(struct sched_domain *sd)
5095 if (cpus_weight(sd->span) == 1)
5098 /* Following flags need at least 2 groups */
5099 if (sd->flags & (SD_LOAD_BALANCE |
5100 SD_BALANCE_NEWIDLE |
5103 if (sd->groups != sd->groups->next)
5107 /* Following flags don't use groups */
5108 if (sd->flags & (SD_WAKE_IDLE |
5116 static int sd_parent_degenerate(struct sched_domain *sd,
5117 struct sched_domain *parent)
5119 unsigned long cflags = sd->flags, pflags = parent->flags;
5121 if (sd_degenerate(parent))
5124 if (!cpus_equal(sd->span, parent->span))
5127 /* Does parent contain flags not in child? */
5128 /* WAKE_BALANCE is a subset of WAKE_AFFINE */
5129 if (cflags & SD_WAKE_AFFINE)
5130 pflags &= ~SD_WAKE_BALANCE;
5131 /* Flags needing groups don't count if only 1 group in parent */
5132 if (parent->groups == parent->groups->next) {
5133 pflags &= ~(SD_LOAD_BALANCE |
5134 SD_BALANCE_NEWIDLE |
5138 if (~cflags & pflags)
5145 * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
5146 * hold the hotplug lock.
5148 static void cpu_attach_domain(struct sched_domain *sd, int cpu)
5150 runqueue_t *rq = cpu_rq(cpu);
5151 struct sched_domain *tmp;
5153 /* Remove the sched domains which do not contribute to scheduling. */
5154 for (tmp = sd; tmp; tmp = tmp->parent) {
5155 struct sched_domain *parent = tmp->parent;
5158 if (sd_parent_degenerate(tmp, parent))
5159 tmp->parent = parent->parent;
5162 if (sd && sd_degenerate(sd))
5165 sched_domain_debug(sd, cpu);
5167 rcu_assign_pointer(rq->sd, sd);
5170 /* cpus with isolated domains */
5171 static cpumask_t __devinitdata cpu_isolated_map = CPU_MASK_NONE;
5173 /* Setup the mask of cpus configured for isolated domains */
5174 static int __init isolated_cpu_setup(char *str)
5176 int ints[NR_CPUS], i;
5178 str = get_options(str, ARRAY_SIZE(ints), ints);
5179 cpus_clear(cpu_isolated_map);
5180 for (i = 1; i <= ints[0]; i++)
5181 if (ints[i] < NR_CPUS)
5182 cpu_set(ints[i], cpu_isolated_map);
5186 __setup ("isolcpus=", isolated_cpu_setup);
5189 * init_sched_build_groups takes an array of groups, the cpumask we wish
5190 * to span, and a pointer to a function which identifies what group a CPU
5191 * belongs to. The return value of group_fn must be a valid index into the
5192 * groups[] array, and must be >= 0 and < NR_CPUS (due to the fact that we
5193 * keep track of groups covered with a cpumask_t).
5195 * init_sched_build_groups will build a circular linked list of the groups
5196 * covered by the given span, and will set each group's ->cpumask correctly,
5197 * and ->cpu_power to 0.
5199 static void init_sched_build_groups(struct sched_group groups[], cpumask_t span,
5200 int (*group_fn)(int cpu))
5202 struct sched_group *first = NULL, *last = NULL;
5203 cpumask_t covered = CPU_MASK_NONE;
5206 for_each_cpu_mask(i, span) {
5207 int group = group_fn(i);
5208 struct sched_group *sg = &groups[group];
5211 if (cpu_isset(i, covered))
5214 sg->cpumask = CPU_MASK_NONE;
5217 for_each_cpu_mask(j, span) {
5218 if (group_fn(j) != group)
5221 cpu_set(j, covered);
5222 cpu_set(j, sg->cpumask);
5233 #define SD_NODES_PER_DOMAIN 16
5236 * Self-tuning task migration cost measurement between source and target CPUs.
5238 * This is done by measuring the cost of manipulating buffers of varying
5239 * sizes. For a given buffer-size here are the steps that are taken:
5241 * 1) the source CPU reads+dirties a shared buffer
5242 * 2) the target CPU reads+dirties the same shared buffer
5244 * We measure how long they take, in the following 4 scenarios:
5246 * - source: CPU1, target: CPU2 | cost1
5247 * - source: CPU2, target: CPU1 | cost2
5248 * - source: CPU1, target: CPU1 | cost3
5249 * - source: CPU2, target: CPU2 | cost4
5251 * We then calculate the cost3+cost4-cost1-cost2 difference - this is
5252 * the cost of migration.
5254 * We then start off from a small buffer-size and iterate up to larger
5255 * buffer sizes, in 5% steps - measuring each buffer-size separately, and
5256 * doing a maximum search for the cost. (The maximum cost for a migration
5257 * normally occurs when the working set size is around the effective cache
5260 #define SEARCH_SCOPE 2
5261 #define MIN_CACHE_SIZE (64*1024U)
5262 #define DEFAULT_CACHE_SIZE (5*1024*1024U)
5263 #define ITERATIONS 1
5264 #define SIZE_THRESH 130
5265 #define COST_THRESH 130
5268 * The migration cost is a function of 'domain distance'. Domain
5269 * distance is the number of steps a CPU has to iterate down its
5270 * domain tree to share a domain with the other CPU. The farther
5271 * two CPUs are from each other, the larger the distance gets.
5273 * Note that we use the distance only to cache measurement results,
5274 * the distance value is not used numerically otherwise. When two
5275 * CPUs have the same distance it is assumed that the migration
5276 * cost is the same. (this is a simplification but quite practical)
5278 #define MAX_DOMAIN_DISTANCE 32
5280 static unsigned long long migration_cost[MAX_DOMAIN_DISTANCE] =
5281 { [ 0 ... MAX_DOMAIN_DISTANCE-1 ] =
5283 * Architectures may override the migration cost and thus avoid
5284 * boot-time calibration. Unit is nanoseconds. Mostly useful for
5285 * virtualized hardware:
5287 #ifdef CONFIG_DEFAULT_MIGRATION_COST
5288 CONFIG_DEFAULT_MIGRATION_COST
5295 * Allow override of migration cost - in units of microseconds.
5296 * E.g. migration_cost=1000,2000,3000 will set up a level-1 cost
5297 * of 1 msec, level-2 cost of 2 msecs and level3 cost of 3 msecs:
5299 static int __init migration_cost_setup(char *str)
5301 int ints[MAX_DOMAIN_DISTANCE+1], i;
5303 str = get_options(str, ARRAY_SIZE(ints), ints);
5305 printk("#ints: %d\n", ints[0]);
5306 for (i = 1; i <= ints[0]; i++) {
5307 migration_cost[i-1] = (unsigned long long)ints[i]*1000;
5308 printk("migration_cost[%d]: %Ld\n", i-1, migration_cost[i-1]);
5313 __setup ("migration_cost=", migration_cost_setup);
5316 * Global multiplier (divisor) for migration-cutoff values,
5317 * in percentiles. E.g. use a value of 150 to get 1.5 times
5318 * longer cache-hot cutoff times.
5320 * (We scale it from 100 to 128 to long long handling easier.)
5323 #define MIGRATION_FACTOR_SCALE 128
5325 static unsigned int migration_factor = MIGRATION_FACTOR_SCALE;
5327 static int __init setup_migration_factor(char *str)
5329 get_option(&str, &migration_factor);
5330 migration_factor = migration_factor * MIGRATION_FACTOR_SCALE / 100;
5334 __setup("migration_factor=", setup_migration_factor);
5337 * Estimated distance of two CPUs, measured via the number of domains
5338 * we have to pass for the two CPUs to be in the same span:
5340 static unsigned long domain_distance(int cpu1, int cpu2)
5342 unsigned long distance = 0;
5343 struct sched_domain *sd;
5345 for_each_domain(cpu1, sd) {
5346 WARN_ON(!cpu_isset(cpu1, sd->span));
5347 if (cpu_isset(cpu2, sd->span))
5351 if (distance >= MAX_DOMAIN_DISTANCE) {
5353 distance = MAX_DOMAIN_DISTANCE-1;
5359 static unsigned int migration_debug;
5361 static int __init setup_migration_debug(char *str)
5363 get_option(&str, &migration_debug);
5367 __setup("migration_debug=", setup_migration_debug);
5370 * Maximum cache-size that the scheduler should try to measure.
5371 * Architectures with larger caches should tune this up during
5372 * bootup. Gets used in the domain-setup code (i.e. during SMP
5375 unsigned int max_cache_size;
5377 static int __init setup_max_cache_size(char *str)
5379 get_option(&str, &max_cache_size);
5383 __setup("max_cache_size=", setup_max_cache_size);
5386 * Dirty a big buffer in a hard-to-predict (for the L2 cache) way. This
5387 * is the operation that is timed, so we try to generate unpredictable
5388 * cachemisses that still end up filling the L2 cache:
5390 static void touch_cache(void *__cache, unsigned long __size)
5392 unsigned long size = __size/sizeof(long), chunk1 = size/3,
5394 unsigned long *cache = __cache;
5397 for (i = 0; i < size/6; i += 8) {
5400 case 1: cache[size-1-i]++;
5401 case 2: cache[chunk1-i]++;
5402 case 3: cache[chunk1+i]++;
5403 case 4: cache[chunk2-i]++;
5404 case 5: cache[chunk2+i]++;
5410 * Measure the cache-cost of one task migration. Returns in units of nsec.
5412 static unsigned long long measure_one(void *cache, unsigned long size,
5413 int source, int target)
5415 cpumask_t mask, saved_mask;
5416 unsigned long long t0, t1, t2, t3, cost;
5418 saved_mask = current->cpus_allowed;
5421 * Flush source caches to RAM and invalidate them:
5426 * Migrate to the source CPU:
5428 mask = cpumask_of_cpu(source);
5429 set_cpus_allowed(current, mask);
5430 WARN_ON(smp_processor_id() != source);
5433 * Dirty the working set:
5436 touch_cache(cache, size);
5440 * Migrate to the target CPU, dirty the L2 cache and access
5441 * the shared buffer. (which represents the working set
5442 * of a migrated task.)
5444 mask = cpumask_of_cpu(target);
5445 set_cpus_allowed(current, mask);
5446 WARN_ON(smp_processor_id() != target);
5449 touch_cache(cache, size);
5452 cost = t1-t0 + t3-t2;
5454 if (migration_debug >= 2)
5455 printk("[%d->%d]: %8Ld %8Ld %8Ld => %10Ld.\n",
5456 source, target, t1-t0, t1-t0, t3-t2, cost);
5458 * Flush target caches to RAM and invalidate them:
5462 set_cpus_allowed(current, saved_mask);
5468 * Measure a series of task migrations and return the average
5469 * result. Since this code runs early during bootup the system
5470 * is 'undisturbed' and the average latency makes sense.
5472 * The algorithm in essence auto-detects the relevant cache-size,
5473 * so it will properly detect different cachesizes for different
5474 * cache-hierarchies, depending on how the CPUs are connected.
5476 * Architectures can prime the upper limit of the search range via
5477 * max_cache_size, otherwise the search range defaults to 20MB...64K.
5479 static unsigned long long
5480 measure_cost(int cpu1, int cpu2, void *cache, unsigned int size)
5482 unsigned long long cost1, cost2;
5486 * Measure the migration cost of 'size' bytes, over an
5487 * average of 10 runs:
5489 * (We perturb the cache size by a small (0..4k)
5490 * value to compensate size/alignment related artifacts.
5491 * We also subtract the cost of the operation done on
5497 * dry run, to make sure we start off cache-cold on cpu1,
5498 * and to get any vmalloc pagefaults in advance:
5500 measure_one(cache, size, cpu1, cpu2);
5501 for (i = 0; i < ITERATIONS; i++)
5502 cost1 += measure_one(cache, size - i*1024, cpu1, cpu2);
5504 measure_one(cache, size, cpu2, cpu1);
5505 for (i = 0; i < ITERATIONS; i++)
5506 cost1 += measure_one(cache, size - i*1024, cpu2, cpu1);
5509 * (We measure the non-migrating [cached] cost on both
5510 * cpu1 and cpu2, to handle CPUs with different speeds)
5514 measure_one(cache, size, cpu1, cpu1);
5515 for (i = 0; i < ITERATIONS; i++)
5516 cost2 += measure_one(cache, size - i*1024, cpu1, cpu1);
5518 measure_one(cache, size, cpu2, cpu2);
5519 for (i = 0; i < ITERATIONS; i++)
5520 cost2 += measure_one(cache, size - i*1024, cpu2, cpu2);
5523 * Get the per-iteration migration cost:
5525 do_div(cost1, 2*ITERATIONS);
5526 do_div(cost2, 2*ITERATIONS);
5528 return cost1 - cost2;
5531 static unsigned long long measure_migration_cost(int cpu1, int cpu2)
5533 unsigned long long max_cost = 0, fluct = 0, avg_fluct = 0;
5534 unsigned int max_size, size, size_found = 0;
5535 long long cost = 0, prev_cost;
5539 * Search from max_cache_size*5 down to 64K - the real relevant
5540 * cachesize has to lie somewhere inbetween.
5542 if (max_cache_size) {
5543 max_size = max(max_cache_size * SEARCH_SCOPE, MIN_CACHE_SIZE);
5544 size = max(max_cache_size / SEARCH_SCOPE, MIN_CACHE_SIZE);
5547 * Since we have no estimation about the relevant
5550 max_size = DEFAULT_CACHE_SIZE * SEARCH_SCOPE;
5551 size = MIN_CACHE_SIZE;
5554 if (!cpu_online(cpu1) || !cpu_online(cpu2)) {
5555 printk("cpu %d and %d not both online!\n", cpu1, cpu2);
5560 * Allocate the working set:
5562 cache = vmalloc(max_size);
5564 printk("could not vmalloc %d bytes for cache!\n", 2*max_size);
5565 return 1000000; // return 1 msec on very small boxen
5568 while (size <= max_size) {
5570 cost = measure_cost(cpu1, cpu2, cache, size);
5576 if (max_cost < cost) {
5582 * Calculate average fluctuation, we use this to prevent
5583 * noise from triggering an early break out of the loop:
5585 fluct = abs(cost - prev_cost);
5586 avg_fluct = (avg_fluct + fluct)/2;
5588 if (migration_debug)
5589 printk("-> [%d][%d][%7d] %3ld.%ld [%3ld.%ld] (%ld): (%8Ld %8Ld)\n",
5591 (long)cost / 1000000,
5592 ((long)cost / 100000) % 10,
5593 (long)max_cost / 1000000,
5594 ((long)max_cost / 100000) % 10,
5595 domain_distance(cpu1, cpu2),
5599 * If we iterated at least 20% past the previous maximum,
5600 * and the cost has dropped by more than 20% already,
5601 * (taking fluctuations into account) then we assume to
5602 * have found the maximum and break out of the loop early:
5604 if (size_found && (size*100 > size_found*SIZE_THRESH))
5605 if (cost+avg_fluct <= 0 ||
5606 max_cost*100 > (cost+avg_fluct)*COST_THRESH) {
5608 if (migration_debug)
5609 printk("-> found max.\n");
5613 * Increase the cachesize in 10% steps:
5615 size = size * 10 / 9;
5618 if (migration_debug)
5619 printk("[%d][%d] working set size found: %d, cost: %Ld\n",
5620 cpu1, cpu2, size_found, max_cost);
5625 * A task is considered 'cache cold' if at least 2 times
5626 * the worst-case cost of migration has passed.
5628 * (this limit is only listened to if the load-balancing
5629 * situation is 'nice' - if there is a large imbalance we
5630 * ignore it for the sake of CPU utilization and
5631 * processing fairness.)
5633 return 2 * max_cost * migration_factor / MIGRATION_FACTOR_SCALE;
5636 static void calibrate_migration_costs(const cpumask_t *cpu_map)
5638 int cpu1 = -1, cpu2 = -1, cpu, orig_cpu = raw_smp_processor_id();
5639 unsigned long j0, j1, distance, max_distance = 0;
5640 struct sched_domain *sd;
5645 * First pass - calculate the cacheflush times:
5647 for_each_cpu_mask(cpu1, *cpu_map) {
5648 for_each_cpu_mask(cpu2, *cpu_map) {
5651 distance = domain_distance(cpu1, cpu2);
5652 max_distance = max(max_distance, distance);
5654 * No result cached yet?
5656 if (migration_cost[distance] == -1LL)
5657 migration_cost[distance] =
5658 measure_migration_cost(cpu1, cpu2);
5662 * Second pass - update the sched domain hierarchy with
5663 * the new cache-hot-time estimations:
5665 for_each_cpu_mask(cpu, *cpu_map) {
5667 for_each_domain(cpu, sd) {
5668 sd->cache_hot_time = migration_cost[distance];
5675 if (migration_debug)
5676 printk("migration: max_cache_size: %d, cpu: %d MHz:\n",
5684 if (system_state == SYSTEM_BOOTING) {
5685 printk("migration_cost=");
5686 for (distance = 0; distance <= max_distance; distance++) {
5689 printk("%ld", (long)migration_cost[distance] / 1000);
5694 if (migration_debug)
5695 printk("migration: %ld seconds\n", (j1-j0)/HZ);
5698 * Move back to the original CPU. NUMA-Q gets confused
5699 * if we migrate to another quad during bootup.
5701 if (raw_smp_processor_id() != orig_cpu) {
5702 cpumask_t mask = cpumask_of_cpu(orig_cpu),
5703 saved_mask = current->cpus_allowed;
5705 set_cpus_allowed(current, mask);
5706 set_cpus_allowed(current, saved_mask);
5713 * find_next_best_node - find the next node to include in a sched_domain
5714 * @node: node whose sched_domain we're building
5715 * @used_nodes: nodes already in the sched_domain
5717 * Find the next node to include in a given scheduling domain. Simply
5718 * finds the closest node not already in the @used_nodes map.
5720 * Should use nodemask_t.
5722 static int find_next_best_node(int node, unsigned long *used_nodes)
5724 int i, n, val, min_val, best_node = 0;
5728 for (i = 0; i < MAX_NUMNODES; i++) {
5729 /* Start at @node */
5730 n = (node + i) % MAX_NUMNODES;
5732 if (!nr_cpus_node(n))
5735 /* Skip already used nodes */
5736 if (test_bit(n, used_nodes))
5739 /* Simple min distance search */
5740 val = node_distance(node, n);
5742 if (val < min_val) {
5748 set_bit(best_node, used_nodes);
5753 * sched_domain_node_span - get a cpumask for a node's sched_domain
5754 * @node: node whose cpumask we're constructing
5755 * @size: number of nodes to include in this span
5757 * Given a node, construct a good cpumask for its sched_domain to span. It
5758 * should be one that prevents unnecessary balancing, but also spreads tasks
5761 static cpumask_t sched_domain_node_span(int node)
5764 cpumask_t span, nodemask;
5765 DECLARE_BITMAP(used_nodes, MAX_NUMNODES);
5768 bitmap_zero(used_nodes, MAX_NUMNODES);
5770 nodemask = node_to_cpumask(node);
5771 cpus_or(span, span, nodemask);
5772 set_bit(node, used_nodes);
5774 for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
5775 int next_node = find_next_best_node(node, used_nodes);
5776 nodemask = node_to_cpumask(next_node);
5777 cpus_or(span, span, nodemask);
5785 * At the moment, CONFIG_SCHED_SMT is never defined, but leave it in so we
5786 * can switch it on easily if needed.
5788 #ifdef CONFIG_SCHED_SMT
5789 static DEFINE_PER_CPU(struct sched_domain, cpu_domains);
5790 static struct sched_group sched_group_cpus[NR_CPUS];
5791 static int cpu_to_cpu_group(int cpu)
5797 static DEFINE_PER_CPU(struct sched_domain, phys_domains);
5798 static struct sched_group sched_group_phys[NR_CPUS];
5799 static int cpu_to_phys_group(int cpu)
5801 #ifdef CONFIG_SCHED_SMT
5802 return first_cpu(cpu_sibling_map[cpu]);
5810 * The init_sched_build_groups can't handle what we want to do with node
5811 * groups, so roll our own. Now each node has its own list of groups which
5812 * gets dynamically allocated.
5814 static DEFINE_PER_CPU(struct sched_domain, node_domains);
5815 static struct sched_group **sched_group_nodes_bycpu[NR_CPUS];
5817 static DEFINE_PER_CPU(struct sched_domain, allnodes_domains);
5818 static struct sched_group *sched_group_allnodes_bycpu[NR_CPUS];
5820 static int cpu_to_allnodes_group(int cpu)
5822 return cpu_to_node(cpu);
5827 * Build sched domains for a given set of cpus and attach the sched domains
5828 * to the individual cpus
5830 void build_sched_domains(const cpumask_t *cpu_map)
5834 struct sched_group **sched_group_nodes = NULL;
5835 struct sched_group *sched_group_allnodes = NULL;
5838 * Allocate the per-node list of sched groups
5840 sched_group_nodes = kmalloc(sizeof(struct sched_group*)*MAX_NUMNODES,
5842 if (!sched_group_nodes) {
5843 printk(KERN_WARNING "Can not alloc sched group node list\n");
5846 sched_group_nodes_bycpu[first_cpu(*cpu_map)] = sched_group_nodes;
5850 * Set up domains for cpus specified by the cpu_map.
5852 for_each_cpu_mask(i, *cpu_map) {
5854 struct sched_domain *sd = NULL, *p;
5855 cpumask_t nodemask = node_to_cpumask(cpu_to_node(i));
5857 cpus_and(nodemask, nodemask, *cpu_map);
5860 if (cpus_weight(*cpu_map)
5861 > SD_NODES_PER_DOMAIN*cpus_weight(nodemask)) {
5862 if (!sched_group_allnodes) {
5863 sched_group_allnodes
5864 = kmalloc(sizeof(struct sched_group)
5867 if (!sched_group_allnodes) {
5869 "Can not alloc allnodes sched group\n");
5872 sched_group_allnodes_bycpu[i]
5873 = sched_group_allnodes;
5875 sd = &per_cpu(allnodes_domains, i);
5876 *sd = SD_ALLNODES_INIT;
5877 sd->span = *cpu_map;
5878 group = cpu_to_allnodes_group(i);
5879 sd->groups = &sched_group_allnodes[group];
5884 sd = &per_cpu(node_domains, i);
5886 sd->span = sched_domain_node_span(cpu_to_node(i));
5888 cpus_and(sd->span, sd->span, *cpu_map);
5892 sd = &per_cpu(phys_domains, i);
5893 group = cpu_to_phys_group(i);
5895 sd->span = nodemask;
5897 sd->groups = &sched_group_phys[group];
5899 #ifdef CONFIG_SCHED_SMT
5901 sd = &per_cpu(cpu_domains, i);
5902 group = cpu_to_cpu_group(i);
5903 *sd = SD_SIBLING_INIT;
5904 sd->span = cpu_sibling_map[i];
5905 cpus_and(sd->span, sd->span, *cpu_map);
5907 sd->groups = &sched_group_cpus[group];
5911 #ifdef CONFIG_SCHED_SMT
5912 /* Set up CPU (sibling) groups */
5913 for_each_cpu_mask(i, *cpu_map) {
5914 cpumask_t this_sibling_map = cpu_sibling_map[i];
5915 cpus_and(this_sibling_map, this_sibling_map, *cpu_map);
5916 if (i != first_cpu(this_sibling_map))
5919 init_sched_build_groups(sched_group_cpus, this_sibling_map,
5924 /* Set up physical groups */
5925 for (i = 0; i < MAX_NUMNODES; i++) {
5926 cpumask_t nodemask = node_to_cpumask(i);
5928 cpus_and(nodemask, nodemask, *cpu_map);
5929 if (cpus_empty(nodemask))
5932 init_sched_build_groups(sched_group_phys, nodemask,
5933 &cpu_to_phys_group);
5937 /* Set up node groups */
5938 if (sched_group_allnodes)
5939 init_sched_build_groups(sched_group_allnodes, *cpu_map,
5940 &cpu_to_allnodes_group);
5942 for (i = 0; i < MAX_NUMNODES; i++) {
5943 /* Set up node groups */
5944 struct sched_group *sg, *prev;
5945 cpumask_t nodemask = node_to_cpumask(i);
5946 cpumask_t domainspan;
5947 cpumask_t covered = CPU_MASK_NONE;
5950 cpus_and(nodemask, nodemask, *cpu_map);
5951 if (cpus_empty(nodemask)) {
5952 sched_group_nodes[i] = NULL;
5956 domainspan = sched_domain_node_span(i);
5957 cpus_and(domainspan, domainspan, *cpu_map);
5959 sg = kmalloc(sizeof(struct sched_group), GFP_KERNEL);
5960 sched_group_nodes[i] = sg;
5961 for_each_cpu_mask(j, nodemask) {
5962 struct sched_domain *sd;
5963 sd = &per_cpu(node_domains, j);
5965 if (sd->groups == NULL) {
5966 /* Turn off balancing if we have no groups */
5972 "Can not alloc domain group for node %d\n", i);
5976 sg->cpumask = nodemask;
5977 cpus_or(covered, covered, nodemask);
5980 for (j = 0; j < MAX_NUMNODES; j++) {
5981 cpumask_t tmp, notcovered;
5982 int n = (i + j) % MAX_NUMNODES;
5984 cpus_complement(notcovered, covered);
5985 cpus_and(tmp, notcovered, *cpu_map);
5986 cpus_and(tmp, tmp, domainspan);
5987 if (cpus_empty(tmp))
5990 nodemask = node_to_cpumask(n);
5991 cpus_and(tmp, tmp, nodemask);
5992 if (cpus_empty(tmp))
5995 sg = kmalloc(sizeof(struct sched_group), GFP_KERNEL);
5998 "Can not alloc domain group for node %d\n", j);
6003 cpus_or(covered, covered, tmp);
6007 prev->next = sched_group_nodes[i];
6011 /* Calculate CPU power for physical packages and nodes */
6012 for_each_cpu_mask(i, *cpu_map) {
6014 struct sched_domain *sd;
6015 #ifdef CONFIG_SCHED_SMT
6016 sd = &per_cpu(cpu_domains, i);
6017 power = SCHED_LOAD_SCALE;
6018 sd->groups->cpu_power = power;
6021 sd = &per_cpu(phys_domains, i);
6022 power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE *
6023 (cpus_weight(sd->groups->cpumask)-1) / 10;
6024 sd->groups->cpu_power = power;
6027 sd = &per_cpu(allnodes_domains, i);
6029 power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE *
6030 (cpus_weight(sd->groups->cpumask)-1) / 10;
6031 sd->groups->cpu_power = power;
6037 for (i = 0; i < MAX_NUMNODES; i++) {
6038 struct sched_group *sg = sched_group_nodes[i];
6044 for_each_cpu_mask(j, sg->cpumask) {
6045 struct sched_domain *sd;
6048 sd = &per_cpu(phys_domains, j);
6049 if (j != first_cpu(sd->groups->cpumask)) {
6051 * Only add "power" once for each
6056 power = SCHED_LOAD_SCALE + SCHED_LOAD_SCALE *
6057 (cpus_weight(sd->groups->cpumask)-1) / 10;
6059 sg->cpu_power += power;
6062 if (sg != sched_group_nodes[i])
6067 /* Attach the domains */
6068 for_each_cpu_mask(i, *cpu_map) {
6069 struct sched_domain *sd;
6070 #ifdef CONFIG_SCHED_SMT
6071 sd = &per_cpu(cpu_domains, i);
6073 sd = &per_cpu(phys_domains, i);
6075 cpu_attach_domain(sd, i);
6078 * Tune cache-hot values:
6080 calibrate_migration_costs(cpu_map);
6083 * Set up scheduler domains and groups. Callers must hold the hotplug lock.
6085 static void arch_init_sched_domains(const cpumask_t *cpu_map)
6087 cpumask_t cpu_default_map;
6090 * Setup mask for cpus without special case scheduling requirements.
6091 * For now this just excludes isolated cpus, but could be used to
6092 * exclude other special cases in the future.
6094 cpus_andnot(cpu_default_map, *cpu_map, cpu_isolated_map);
6096 build_sched_domains(&cpu_default_map);
6099 static void arch_destroy_sched_domains(const cpumask_t *cpu_map)
6105 for_each_cpu_mask(cpu, *cpu_map) {
6106 struct sched_group *sched_group_allnodes
6107 = sched_group_allnodes_bycpu[cpu];
6108 struct sched_group **sched_group_nodes
6109 = sched_group_nodes_bycpu[cpu];
6111 if (sched_group_allnodes) {
6112 kfree(sched_group_allnodes);
6113 sched_group_allnodes_bycpu[cpu] = NULL;
6116 if (!sched_group_nodes)
6119 for (i = 0; i < MAX_NUMNODES; i++) {
6120 cpumask_t nodemask = node_to_cpumask(i);
6121 struct sched_group *oldsg, *sg = sched_group_nodes[i];
6123 cpus_and(nodemask, nodemask, *cpu_map);
6124 if (cpus_empty(nodemask))
6134 if (oldsg != sched_group_nodes[i])
6137 kfree(sched_group_nodes);
6138 sched_group_nodes_bycpu[cpu] = NULL;
6144 * Detach sched domains from a group of cpus specified in cpu_map
6145 * These cpus will now be attached to the NULL domain
6147 static void detach_destroy_domains(const cpumask_t *cpu_map)
6151 for_each_cpu_mask(i, *cpu_map)
6152 cpu_attach_domain(NULL, i);
6153 synchronize_sched();
6154 arch_destroy_sched_domains(cpu_map);
6158 * Partition sched domains as specified by the cpumasks below.
6159 * This attaches all cpus from the cpumasks to the NULL domain,
6160 * waits for a RCU quiescent period, recalculates sched
6161 * domain information and then attaches them back to the
6162 * correct sched domains
6163 * Call with hotplug lock held
6165 void partition_sched_domains(cpumask_t *partition1, cpumask_t *partition2)
6167 cpumask_t change_map;
6169 cpus_and(*partition1, *partition1, cpu_online_map);
6170 cpus_and(*partition2, *partition2, cpu_online_map);
6171 cpus_or(change_map, *partition1, *partition2);
6173 /* Detach sched domains from all of the affected cpus */
6174 detach_destroy_domains(&change_map);
6175 if (!cpus_empty(*partition1))
6176 build_sched_domains(partition1);
6177 if (!cpus_empty(*partition2))
6178 build_sched_domains(partition2);
6181 #ifdef CONFIG_HOTPLUG_CPU
6183 * Force a reinitialization of the sched domains hierarchy. The domains
6184 * and groups cannot be updated in place without racing with the balancing
6185 * code, so we temporarily attach all running cpus to the NULL domain
6186 * which will prevent rebalancing while the sched domains are recalculated.
6188 static int update_sched_domains(struct notifier_block *nfb,
6189 unsigned long action, void *hcpu)
6192 case CPU_UP_PREPARE:
6193 case CPU_DOWN_PREPARE:
6194 detach_destroy_domains(&cpu_online_map);
6197 case CPU_UP_CANCELED:
6198 case CPU_DOWN_FAILED:
6202 * Fall through and re-initialise the domains.
6209 /* The hotplug lock is already held by cpu_up/cpu_down */
6210 arch_init_sched_domains(&cpu_online_map);
6216 void __init sched_init_smp(void)
6219 arch_init_sched_domains(&cpu_online_map);
6220 unlock_cpu_hotplug();
6221 /* XXX: Theoretical race here - CPU may be hotplugged now */
6222 hotcpu_notifier(update_sched_domains, 0);
6225 void __init sched_init_smp(void)
6228 #endif /* CONFIG_SMP */
6230 int in_sched_functions(unsigned long addr)
6232 /* Linker adds these: start and end of __sched functions */
6233 extern char __sched_text_start[], __sched_text_end[];
6234 return in_lock_functions(addr) ||
6235 (addr >= (unsigned long)__sched_text_start
6236 && addr < (unsigned long)__sched_text_end);
6239 void __init sched_init(void)
6245 prio_array_t *array;
6248 spin_lock_init(&rq->lock);
6250 rq->active = rq->arrays;
6251 rq->expired = rq->arrays + 1;
6252 rq->best_expired_prio = MAX_PRIO;
6256 for (j = 1; j < 3; j++)
6257 rq->cpu_load[j] = 0;
6258 rq->active_balance = 0;
6261 rq->migration_thread = NULL;
6262 INIT_LIST_HEAD(&rq->migration_queue);
6265 atomic_set(&rq->nr_iowait, 0);
6266 #ifdef CONFIG_VSERVER_HARDCPU
6267 INIT_LIST_HEAD(&rq->hold_queue);
6270 for (j = 0; j < 2; j++) {
6271 array = rq->arrays + j;
6272 for (k = 0; k < MAX_PRIO; k++) {
6273 INIT_LIST_HEAD(array->queue + k);
6274 __clear_bit(k, array->bitmap);
6276 // delimiter for bitsearch
6277 __set_bit(MAX_PRIO, array->bitmap);
6282 * The boot idle thread does lazy MMU switching as well:
6284 atomic_inc(&init_mm.mm_count);
6285 enter_lazy_tlb(&init_mm, current);
6288 * Make us the idle thread. Technically, schedule() should not be
6289 * called from this thread, however somewhere below it might be,
6290 * but because we are the idle thread, we just pick up running again
6291 * when this runqueue becomes "idle".
6293 init_idle(current, smp_processor_id());
6296 #ifdef CONFIG_DEBUG_SPINLOCK_SLEEP
6297 void __might_sleep(char *file, int line)
6299 #if defined(in_atomic)
6300 static unsigned long prev_jiffy; /* ratelimiting */
6302 if ((in_atomic() || irqs_disabled()) &&
6303 system_state == SYSTEM_RUNNING && !oops_in_progress) {
6304 if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
6306 prev_jiffy = jiffies;
6307 printk(KERN_ERR "Debug: sleeping function called from invalid"
6308 " context at %s:%d\n", file, line);
6309 printk("in_atomic():%d, irqs_disabled():%d\n",
6310 in_atomic(), irqs_disabled());
6315 EXPORT_SYMBOL(__might_sleep);
6318 #ifdef CONFIG_MAGIC_SYSRQ
6319 void normalize_rt_tasks(void)
6321 struct task_struct *p;
6322 prio_array_t *array;
6323 unsigned long flags;
6326 read_lock_irq(&tasklist_lock);
6327 for_each_process (p) {
6331 rq = task_rq_lock(p, &flags);
6335 deactivate_task(p, task_rq(p));
6336 __setscheduler(p, SCHED_NORMAL, 0);
6338 vx_activate_task(p);
6339 __activate_task(p, task_rq(p));
6340 resched_task(rq->curr);
6343 task_rq_unlock(rq, &flags);
6345 read_unlock_irq(&tasklist_lock);
6348 #endif /* CONFIG_MAGIC_SYSRQ */
6352 * These functions are only useful for the IA64 MCA handling.
6354 * They can only be called when the whole system has been
6355 * stopped - every CPU needs to be quiescent, and no scheduling
6356 * activity can take place. Using them for anything else would
6357 * be a serious bug, and as a result, they aren't even visible
6358 * under any other configuration.
6362 * curr_task - return the current task for a given cpu.
6363 * @cpu: the processor in question.
6365 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6367 task_t *curr_task(int cpu)
6369 return cpu_curr(cpu);
6373 * set_curr_task - set the current task for a given cpu.
6374 * @cpu: the processor in question.
6375 * @p: the task pointer to set.
6377 * Description: This function must only be used when non-maskable interrupts
6378 * are serviced on a separate stack. It allows the architecture to switch the
6379 * notion of the current task on a cpu in a non-blocking manner. This function
6380 * must be called with all CPU's synchronized, and interrupts disabled, the
6381 * and caller must save the original value of the current task (see
6382 * curr_task() above) and restore that value before reenabling interrupts and
6383 * re-starting the system.
6385 * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
6387 void set_curr_task(int cpu, task_t *p)