source: mainline/kernel/generic/src/proc/scheduler.c@ 82d515e9

lfn serial ticket/834-toolchain-update topic/msim-upgrade topic/simplify-dev-export
Last change on this file since 82d515e9 was 82d515e9, checked in by Jakub Jermar <jakub@…>, 8 years ago

Fix terminology

Objects of slab_cache_t type are caches, not slabs.
  • Property mode set to 100644
File size: 17.7 KB
Line 
1/*
2 * Copyright (c) 2010 Jakub Jermar
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup genericproc
30 * @{
31 */
32
33/**
34 * @file
35 * @brief Scheduler and load balancing.
36 *
37 * This file contains the scheduler and kcpulb kernel thread which
38 * performs load-balancing of per-CPU run queues.
39 */
40
41#include <assert.h>
42#include <proc/scheduler.h>
43#include <proc/thread.h>
44#include <proc/task.h>
45#include <mm/frame.h>
46#include <mm/page.h>
47#include <mm/as.h>
48#include <time/timeout.h>
49#include <time/delay.h>
50#include <arch/asm.h>
51#include <arch/faddr.h>
52#include <arch/cycle.h>
53#include <atomic.h>
54#include <synch/spinlock.h>
55#include <synch/workqueue.h>
56#include <synch/rcu.h>
57#include <config.h>
58#include <context.h>
59#include <fpu_context.h>
60#include <func.h>
61#include <arch.h>
62#include <adt/list.h>
63#include <panic.h>
64#include <cpu.h>
65#include <print.h>
66#include <log.h>
67#include <stacktrace.h>
68
69static void scheduler_separated_stack(void);
70
71atomic_t nrdy; /**< Number of ready threads in the system. */
72
73/** Carry out actions before new task runs. */
74static void before_task_runs(void)
75{
76 before_task_runs_arch();
77}
78
79/** Take actions before new thread runs.
80 *
81 * Perform actions that need to be
82 * taken before the newly selected
83 * thread is passed control.
84 *
85 * THREAD->lock is locked on entry
86 *
87 */
88static void before_thread_runs(void)
89{
90 before_thread_runs_arch();
91 rcu_before_thread_runs();
92
93#ifdef CONFIG_FPU_LAZY
94 if (THREAD == CPU->fpu_owner)
95 fpu_enable();
96 else
97 fpu_disable();
98#elif defined CONFIG_FPU
99 fpu_enable();
100 if (THREAD->fpu_context_exists)
101 fpu_context_restore(THREAD->saved_fpu_context);
102 else {
103 fpu_init();
104 THREAD->fpu_context_exists = true;
105 }
106#endif
107
108#ifdef CONFIG_UDEBUG
109 if (THREAD->btrace) {
110 istate_t *istate = THREAD->udebug.uspace_state;
111 if (istate != NULL) {
112 printf("Thread %" PRIu64 " stack trace:\n", THREAD->tid);
113 stack_trace_istate(istate);
114 }
115
116 THREAD->btrace = false;
117 }
118#endif
119}
120
121/** Take actions after THREAD had run.
122 *
123 * Perform actions that need to be
124 * taken after the running thread
125 * had been preempted by the scheduler.
126 *
127 * THREAD->lock is locked on entry
128 *
129 */
130static void after_thread_ran(void)
131{
132 workq_after_thread_ran();
133 rcu_after_thread_ran();
134 after_thread_ran_arch();
135}
136
137#ifdef CONFIG_FPU_LAZY
138void scheduler_fpu_lazy_request(void)
139{
140restart:
141 fpu_enable();
142 irq_spinlock_lock(&CPU->lock, false);
143
144 /* Save old context */
145 if (CPU->fpu_owner != NULL) {
146 irq_spinlock_lock(&CPU->fpu_owner->lock, false);
147 fpu_context_save(CPU->fpu_owner->saved_fpu_context);
148
149 /* Don't prevent migration */
150 CPU->fpu_owner->fpu_context_engaged = false;
151 irq_spinlock_unlock(&CPU->fpu_owner->lock, false);
152 CPU->fpu_owner = NULL;
153 }
154
155 irq_spinlock_lock(&THREAD->lock, false);
156 if (THREAD->fpu_context_exists) {
157 fpu_context_restore(THREAD->saved_fpu_context);
158 } else {
159 /* Allocate FPU context */
160 if (!THREAD->saved_fpu_context) {
161 /* Might sleep */
162 irq_spinlock_unlock(&THREAD->lock, false);
163 irq_spinlock_unlock(&CPU->lock, false);
164 THREAD->saved_fpu_context =
165 (fpu_context_t *) slab_alloc(fpu_context_cache, 0);
166
167 /* We may have switched CPUs during slab_alloc */
168 goto restart;
169 }
170 fpu_init();
171 THREAD->fpu_context_exists = true;
172 }
173
174 CPU->fpu_owner = THREAD;
175 THREAD->fpu_context_engaged = true;
176 irq_spinlock_unlock(&THREAD->lock, false);
177
178 irq_spinlock_unlock(&CPU->lock, false);
179}
180#endif /* CONFIG_FPU_LAZY */
181
182/** Initialize scheduler
183 *
184 * Initialize kernel scheduler.
185 *
186 */
187void scheduler_init(void)
188{
189}
190
191/** Get thread to be scheduled
192 *
193 * Get the optimal thread to be scheduled
194 * according to thread accounting and scheduler
195 * policy.
196 *
197 * @return Thread to be scheduled.
198 *
199 */
200static thread_t *find_best_thread(void)
201{
202 assert(CPU != NULL);
203
204loop:
205
206 if (atomic_get(&CPU->nrdy) == 0) {
207 /*
208 * For there was nothing to run, the CPU goes to sleep
209 * until a hardware interrupt or an IPI comes.
210 * This improves energy saving and hyperthreading.
211 */
212 irq_spinlock_lock(&CPU->lock, false);
213 CPU->idle = true;
214 irq_spinlock_unlock(&CPU->lock, false);
215 interrupts_enable();
216
217 /*
218 * An interrupt might occur right now and wake up a thread.
219 * In such case, the CPU will continue to go to sleep
220 * even though there is a runnable thread.
221 */
222 cpu_sleep();
223 interrupts_disable();
224 goto loop;
225 }
226
227 assert(!CPU->idle);
228
229 unsigned int i;
230 for (i = 0; i < RQ_COUNT; i++) {
231 irq_spinlock_lock(&(CPU->rq[i].lock), false);
232 if (CPU->rq[i].n == 0) {
233 /*
234 * If this queue is empty, try a lower-priority queue.
235 */
236 irq_spinlock_unlock(&(CPU->rq[i].lock), false);
237 continue;
238 }
239
240 atomic_dec(&CPU->nrdy);
241 atomic_dec(&nrdy);
242 CPU->rq[i].n--;
243
244 /*
245 * Take the first thread from the queue.
246 */
247 thread_t *thread = list_get_instance(
248 list_first(&CPU->rq[i].rq), thread_t, rq_link);
249 list_remove(&thread->rq_link);
250
251 irq_spinlock_pass(&(CPU->rq[i].lock), &thread->lock);
252
253 thread->cpu = CPU;
254 thread->ticks = us2ticks((i + 1) * 10000);
255 thread->priority = i; /* Correct rq index */
256
257 /*
258 * Clear the stolen flag so that it can be migrated
259 * when load balancing needs emerge.
260 */
261 thread->stolen = false;
262 irq_spinlock_unlock(&thread->lock, false);
263
264 return thread;
265 }
266
267 goto loop;
268}
269
270/** Prevent rq starvation
271 *
272 * Prevent low priority threads from starving in rq's.
273 *
274 * When the function decides to relink rq's, it reconnects
275 * respective pointers so that in result threads with 'pri'
276 * greater or equal start are moved to a higher-priority queue.
277 *
278 * @param start Threshold priority.
279 *
280 */
281static void relink_rq(int start)
282{
283 list_t list;
284
285 list_initialize(&list);
286 irq_spinlock_lock(&CPU->lock, false);
287
288 if (CPU->needs_relink > NEEDS_RELINK_MAX) {
289 int i;
290 for (i = start; i < RQ_COUNT - 1; i++) {
291 /* Remember and empty rq[i + 1] */
292
293 irq_spinlock_lock(&CPU->rq[i + 1].lock, false);
294 list_concat(&list, &CPU->rq[i + 1].rq);
295 size_t n = CPU->rq[i + 1].n;
296 CPU->rq[i + 1].n = 0;
297 irq_spinlock_unlock(&CPU->rq[i + 1].lock, false);
298
299 /* Append rq[i + 1] to rq[i] */
300
301 irq_spinlock_lock(&CPU->rq[i].lock, false);
302 list_concat(&CPU->rq[i].rq, &list);
303 CPU->rq[i].n += n;
304 irq_spinlock_unlock(&CPU->rq[i].lock, false);
305 }
306
307 CPU->needs_relink = 0;
308 }
309
310 irq_spinlock_unlock(&CPU->lock, false);
311}
312
313/** The scheduler
314 *
315 * The thread scheduling procedure.
316 * Passes control directly to
317 * scheduler_separated_stack().
318 *
319 */
320void scheduler(void)
321{
322 volatile ipl_t ipl;
323
324 assert(CPU != NULL);
325
326 ipl = interrupts_disable();
327
328 if (atomic_get(&haltstate))
329 halt();
330
331 if (THREAD) {
332 irq_spinlock_lock(&THREAD->lock, false);
333
334 /* Update thread kernel accounting */
335 THREAD->kcycles += get_cycle() - THREAD->last_cycle;
336
337#if (defined CONFIG_FPU) && (!defined CONFIG_FPU_LAZY)
338 fpu_context_save(THREAD->saved_fpu_context);
339#endif
340 if (!context_save(&THREAD->saved_context)) {
341 /*
342 * This is the place where threads leave scheduler();
343 */
344
345 /* Save current CPU cycle */
346 THREAD->last_cycle = get_cycle();
347
348 irq_spinlock_unlock(&THREAD->lock, false);
349 interrupts_restore(THREAD->saved_context.ipl);
350
351 return;
352 }
353
354 /*
355 * Interrupt priority level of preempted thread is recorded
356 * here to facilitate scheduler() invocations from
357 * interrupts_disable()'d code (e.g. waitq_sleep_timeout()).
358 *
359 */
360 THREAD->saved_context.ipl = ipl;
361 }
362
363 /*
364 * Through the 'THE' structure, we keep track of THREAD, TASK, CPU, AS
365 * and preemption counter. At this point THE could be coming either
366 * from THREAD's or CPU's stack.
367 *
368 */
369 the_copy(THE, (the_t *) CPU->stack);
370
371 /*
372 * We may not keep the old stack.
373 * Reason: If we kept the old stack and got blocked, for instance, in
374 * find_best_thread(), the old thread could get rescheduled by another
375 * CPU and overwrite the part of its own stack that was also used by
376 * the scheduler on this CPU.
377 *
378 * Moreover, we have to bypass the compiler-generated POP sequence
379 * which is fooled by SP being set to the very top of the stack.
380 * Therefore the scheduler() function continues in
381 * scheduler_separated_stack().
382 *
383 */
384 context_save(&CPU->saved_context);
385 context_set(&CPU->saved_context, FADDR(scheduler_separated_stack),
386 (uintptr_t) CPU->stack, STACK_SIZE);
387 context_restore(&CPU->saved_context);
388
389 /* Not reached */
390}
391
392/** Scheduler stack switch wrapper
393 *
394 * Second part of the scheduler() function
395 * using new stack. Handling the actual context
396 * switch to a new thread.
397 *
398 */
399void scheduler_separated_stack(void)
400{
401 DEADLOCK_PROBE_INIT(p_joinwq);
402 task_t *old_task = TASK;
403 as_t *old_as = AS;
404
405 assert((!THREAD) || (irq_spinlock_locked(&THREAD->lock)));
406 assert(CPU != NULL);
407 assert(interrupts_disabled());
408
409 /*
410 * Hold the current task and the address space to prevent their
411 * possible destruction should thread_destroy() be called on this or any
412 * other processor while the scheduler is still using them.
413 */
414 if (old_task)
415 task_hold(old_task);
416
417 if (old_as)
418 as_hold(old_as);
419
420 if (THREAD) {
421 /* Must be run after the switch to scheduler stack */
422 after_thread_ran();
423
424 switch (THREAD->state) {
425 case Running:
426 irq_spinlock_unlock(&THREAD->lock, false);
427 thread_ready(THREAD);
428 break;
429
430 case Exiting:
431 rcu_thread_exiting();
432repeat:
433 if (THREAD->detached) {
434 thread_destroy(THREAD, false);
435 } else {
436 /*
437 * The thread structure is kept allocated until
438 * somebody calls thread_detach() on it.
439 */
440 if (!irq_spinlock_trylock(&THREAD->join_wq.lock)) {
441 /*
442 * Avoid deadlock.
443 */
444 irq_spinlock_unlock(&THREAD->lock, false);
445 delay(HZ);
446 irq_spinlock_lock(&THREAD->lock, false);
447 DEADLOCK_PROBE(p_joinwq,
448 DEADLOCK_THRESHOLD);
449 goto repeat;
450 }
451 _waitq_wakeup_unsafe(&THREAD->join_wq,
452 WAKEUP_FIRST);
453 irq_spinlock_unlock(&THREAD->join_wq.lock, false);
454
455 THREAD->state = Lingering;
456 irq_spinlock_unlock(&THREAD->lock, false);
457 }
458 break;
459
460 case Sleeping:
461 /*
462 * Prefer the thread after it's woken up.
463 */
464 THREAD->priority = -1;
465
466 /*
467 * We need to release wq->lock which we locked in
468 * waitq_sleep(). Address of wq->lock is kept in
469 * THREAD->sleep_queue.
470 */
471 irq_spinlock_unlock(&THREAD->sleep_queue->lock, false);
472
473 irq_spinlock_unlock(&THREAD->lock, false);
474 break;
475
476 default:
477 /*
478 * Entering state is unexpected.
479 */
480 panic("tid%" PRIu64 ": unexpected state %s.",
481 THREAD->tid, thread_states[THREAD->state]);
482 break;
483 }
484
485 THREAD = NULL;
486 }
487
488 THREAD = find_best_thread();
489
490 irq_spinlock_lock(&THREAD->lock, false);
491 int priority = THREAD->priority;
492 irq_spinlock_unlock(&THREAD->lock, false);
493
494 relink_rq(priority);
495
496 /*
497 * If both the old and the new task are the same,
498 * lots of work is avoided.
499 */
500 if (TASK != THREAD->task) {
501 as_t *new_as = THREAD->task->as;
502
503 /*
504 * Note that it is possible for two tasks
505 * to share one address space.
506 */
507 if (old_as != new_as) {
508 /*
509 * Both tasks and address spaces are different.
510 * Replace the old one with the new one.
511 */
512 as_switch(old_as, new_as);
513 }
514
515 TASK = THREAD->task;
516 before_task_runs();
517 }
518
519 if (old_task)
520 task_release(old_task);
521
522 if (old_as)
523 as_release(old_as);
524
525 irq_spinlock_lock(&THREAD->lock, false);
526 THREAD->state = Running;
527
528#ifdef SCHEDULER_VERBOSE
529 log(LF_OTHER, LVL_DEBUG,
530 "cpu%u: tid %" PRIu64 " (priority=%d, ticks=%" PRIu64
531 ", nrdy=%" PRIua ")", CPU->id, THREAD->tid, THREAD->priority,
532 THREAD->ticks, atomic_get(&CPU->nrdy));
533#endif
534
535 /*
536 * Some architectures provide late kernel PA2KA(identity)
537 * mapping in a page fault handler. However, the page fault
538 * handler uses the kernel stack of the running thread and
539 * therefore cannot be used to map it. The kernel stack, if
540 * necessary, is to be mapped in before_thread_runs(). This
541 * function must be executed before the switch to the new stack.
542 */
543 before_thread_runs();
544
545 /*
546 * Copy the knowledge of CPU, TASK, THREAD and preemption counter to
547 * thread's stack.
548 */
549 the_copy(THE, (the_t *) THREAD->kstack);
550
551 context_restore(&THREAD->saved_context);
552
553 /* Not reached */
554}
555
556#ifdef CONFIG_SMP
557/** Load balancing thread
558 *
559 * SMP load balancing thread, supervising thread supplies
560 * for the CPU it's wired to.
561 *
562 * @param arg Generic thread argument (unused).
563 *
564 */
565void kcpulb(void *arg)
566{
567 atomic_count_t average;
568 atomic_count_t rdy;
569
570 /*
571 * Detach kcpulb as nobody will call thread_join_timeout() on it.
572 */
573 thread_detach(THREAD);
574
575loop:
576 /*
577 * Work in 1s intervals.
578 */
579 thread_sleep(1);
580
581not_satisfied:
582 /*
583 * Calculate the number of threads that will be migrated/stolen from
584 * other CPU's. Note that situation can have changed between two
585 * passes. Each time get the most up to date counts.
586 *
587 */
588 average = atomic_get(&nrdy) / config.cpu_active + 1;
589 rdy = atomic_get(&CPU->nrdy);
590
591 if (average <= rdy)
592 goto satisfied;
593
594 atomic_count_t count = average - rdy;
595
596 /*
597 * Searching least priority queues on all CPU's first and most priority
598 * queues on all CPU's last.
599 */
600 size_t acpu;
601 size_t acpu_bias = 0;
602 int rq;
603
604 for (rq = RQ_COUNT - 1; rq >= 0; rq--) {
605 for (acpu = 0; acpu < config.cpu_active; acpu++) {
606 cpu_t *cpu = &cpus[(acpu + acpu_bias) % config.cpu_active];
607
608 /*
609 * Not interested in ourselves.
610 * Doesn't require interrupt disabling for kcpulb has
611 * THREAD_FLAG_WIRED.
612 *
613 */
614 if (CPU == cpu)
615 continue;
616
617 if (atomic_get(&cpu->nrdy) <= average)
618 continue;
619
620 irq_spinlock_lock(&(cpu->rq[rq].lock), true);
621 if (cpu->rq[rq].n == 0) {
622 irq_spinlock_unlock(&(cpu->rq[rq].lock), true);
623 continue;
624 }
625
626 thread_t *thread = NULL;
627
628 /* Search rq from the back */
629 link_t *link = cpu->rq[rq].rq.head.prev;
630
631 while (link != &(cpu->rq[rq].rq.head)) {
632 thread = (thread_t *) list_get_instance(link,
633 thread_t, rq_link);
634
635 /*
636 * Do not steal CPU-wired threads, threads
637 * already stolen, threads for which migration
638 * was temporarily disabled or threads whose
639 * FPU context is still in the CPU.
640 */
641 irq_spinlock_lock(&thread->lock, false);
642
643 if ((!thread->wired) && (!thread->stolen) &&
644 (!thread->nomigrate) &&
645 (!thread->fpu_context_engaged)) {
646 /*
647 * Remove thread from ready queue.
648 */
649 irq_spinlock_unlock(&thread->lock,
650 false);
651
652 atomic_dec(&cpu->nrdy);
653 atomic_dec(&nrdy);
654
655 cpu->rq[rq].n--;
656 list_remove(&thread->rq_link);
657
658 break;
659 }
660
661 irq_spinlock_unlock(&thread->lock, false);
662
663 link = link->prev;
664 thread = NULL;
665 }
666
667 if (thread) {
668 /*
669 * Ready thread on local CPU
670 */
671
672 irq_spinlock_pass(&(cpu->rq[rq].lock),
673 &thread->lock);
674
675#ifdef KCPULB_VERBOSE
676 log(LF_OTHER, LVL_DEBUG,
677 "kcpulb%u: TID %" PRIu64 " -> cpu%u, "
678 "nrdy=%ld, avg=%ld", CPU->id, t->tid,
679 CPU->id, atomic_get(&CPU->nrdy),
680 atomic_get(&nrdy) / config.cpu_active);
681#endif
682
683 thread->stolen = true;
684 thread->state = Entering;
685
686 irq_spinlock_unlock(&thread->lock, true);
687 thread_ready(thread);
688
689 if (--count == 0)
690 goto satisfied;
691
692 /*
693 * We are not satisfied yet, focus on another
694 * CPU next time.
695 *
696 */
697 acpu_bias++;
698
699 continue;
700 } else
701 irq_spinlock_unlock(&(cpu->rq[rq].lock), true);
702
703 }
704 }
705
706 if (atomic_get(&CPU->nrdy)) {
707 /*
708 * Be a little bit light-weight and let migrated threads run.
709 *
710 */
711 scheduler();
712 } else {
713 /*
714 * We failed to migrate a single thread.
715 * Give up this turn.
716 *
717 */
718 goto loop;
719 }
720
721 goto not_satisfied;
722
723satisfied:
724 goto loop;
725}
726#endif /* CONFIG_SMP */
727
728/** Print information about threads & scheduler queues
729 *
730 */
731void sched_print_list(void)
732{
733 size_t cpu;
734 for (cpu = 0; cpu < config.cpu_count; cpu++) {
735 if (!cpus[cpu].active)
736 continue;
737
738 irq_spinlock_lock(&cpus[cpu].lock, true);
739
740 printf("cpu%u: address=%p, nrdy=%" PRIua ", needs_relink=%zu\n",
741 cpus[cpu].id, &cpus[cpu], atomic_get(&cpus[cpu].nrdy),
742 cpus[cpu].needs_relink);
743
744 unsigned int i;
745 for (i = 0; i < RQ_COUNT; i++) {
746 irq_spinlock_lock(&(cpus[cpu].rq[i].lock), false);
747 if (cpus[cpu].rq[i].n == 0) {
748 irq_spinlock_unlock(&(cpus[cpu].rq[i].lock), false);
749 continue;
750 }
751
752 printf("\trq[%u]: ", i);
753 list_foreach(cpus[cpu].rq[i].rq, rq_link, thread_t,
754 thread) {
755 printf("%" PRIu64 "(%s) ", thread->tid,
756 thread_states[thread->state]);
757 }
758 printf("\n");
759
760 irq_spinlock_unlock(&(cpus[cpu].rq[i].lock), false);
761 }
762
763 irq_spinlock_unlock(&cpus[cpu].lock, true);
764 }
765}
766
767/** @}
768 */
Note: See TracBrowser for help on using the repository browser.