projects/linux·blob v7.1 · refs

1246 lines · 31.9 KiB

1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * kernel/locking/mutex.c
4 *
5 * Mutexes: blocking mutual exclusion locks
6 *
7 * Started by Ingo Molnar:
8 *
9 * Copyright (C) 2004, 2005, 2006 Red Hat, Inc., Ingo Molnar <[email protected]>
10 *
11 * Many thanks to Arjan van de Ven, Thomas Gleixner, Steven Rostedt and
12 * David Howells for suggestions and improvements.
13 *
14 * - Adaptive spinning for mutexes by Peter Zijlstra. (Ported to mainline
15 * from the -rt tree, where it was originally implemented for rtmutexes
16 * by Steven Rostedt, based on work by Gregory Haskins, Peter Morreale
17 * and Sven Dietrich.
18 *
19 * Also see Documentation/locking/mutex-design.rst.
20 */
21#include <linux/mutex.h>
22#include <linux/ww_mutex.h>
23#include <linux/sched/signal.h>
24#include <linux/sched/rt.h>
25#include <linux/sched/wake_q.h>
26#include <linux/sched/debug.h>
27#include <linux/export.h>
28#include <linux/spinlock.h>
29#include <linux/interrupt.h>
30#include <linux/debug_locks.h>
31#include <linux/osq_lock.h>
32#include <linux/hung_task.h>
33
34#define CREATE_TRACE_POINTS
35#include <trace/events/lock.h>
36
37#ifndef CONFIG_PREEMPT_RT
38#include "mutex.h"
39
40#ifdef CONFIG_DEBUG_MUTEXES
41# define MUTEX_WARN_ON(cond) DEBUG_LOCKS_WARN_ON(cond)
42#else
43# define MUTEX_WARN_ON(cond)
44#endif
45
46static void __mutex_init_generic(struct mutex *lock)
47{
48 atomic_long_set(&lock->owner, 0);
49 scoped_guard (raw_spinlock_init, &lock->wait_lock) {
50 lock->first_waiter = NULL;
51 }
52#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
53 osq_lock_init(&lock->osq);
54#endif
55 debug_mutex_init(lock);
56}
57
58static inline struct task_struct *__owner_task(unsigned long owner)
59{
60 return (struct task_struct *)(owner & ~MUTEX_FLAGS);
61}
62
63bool mutex_is_locked(struct mutex *lock)
64{
65 return __mutex_owner(lock) != NULL;
66}
67EXPORT_SYMBOL(mutex_is_locked);
68
69static inline unsigned long __owner_flags(unsigned long owner)
70{
71 return owner & MUTEX_FLAGS;
72}
73
74/* Do not use the return value as a pointer directly. */
75unsigned long mutex_get_owner(struct mutex *lock)
76{
77 unsigned long owner = atomic_long_read(&lock->owner);
78
79 return (unsigned long)__owner_task(owner);
80}
81
82/*
83 * Returns: __mutex_owner(lock) on failure or NULL on success.
84 */
85static inline struct task_struct *__mutex_trylock_common(struct mutex *lock, bool handoff)
86{
87 unsigned long owner, curr = (unsigned long)current;
88
89 owner = atomic_long_read(&lock->owner);
90 for (;;) { /* must loop, can race against a flag */
91 unsigned long flags = __owner_flags(owner);
92 unsigned long task = owner & ~MUTEX_FLAGS;
93
94 if (task) {
95 if (flags & MUTEX_FLAG_PICKUP) {
96 if (task != curr)
97 break;
98 flags &= ~MUTEX_FLAG_PICKUP;
99 } else if (handoff) {
100 if (flags & MUTEX_FLAG_HANDOFF)
101 break;
102 flags |= MUTEX_FLAG_HANDOFF;
103 } else {
104 break;
105 }
106 } else {
107 MUTEX_WARN_ON(flags & (MUTEX_FLAG_HANDOFF | MUTEX_FLAG_PICKUP));
108 task = curr;
109 }
110
111 if (atomic_long_try_cmpxchg_acquire(&lock->owner, &owner, task | flags)) {
112 if (task == curr)
113 return NULL;
114 break;
115 }
116 }
117
118 return __owner_task(owner);
119}
120
121/*
122 * Trylock or set HANDOFF
123 */
124static inline bool __mutex_trylock_or_handoff(struct mutex *lock, bool handoff)
125{
126 return !__mutex_trylock_common(lock, handoff);
127}
128
129/*
130 * Actual trylock that will work on any unlocked state.
131 */
132static inline bool __mutex_trylock(struct mutex *lock)
133{
134 return !__mutex_trylock_common(lock, false);
135}
136
137#ifndef CONFIG_DEBUG_LOCK_ALLOC
138/*
139 * Lockdep annotations are contained to the slow paths for simplicity.
140 * There is nothing that would stop spreading the lockdep annotations outwards
141 * except more code.
142 */
143void mutex_init_generic(struct mutex *lock)
144{
145 __mutex_init_generic(lock);
146}
147EXPORT_SYMBOL(mutex_init_generic);
148
149/*
150 * Optimistic trylock that only works in the uncontended case. Make sure to
151 * follow with a __mutex_trylock() before failing.
152 */
153static __always_inline bool __mutex_trylock_fast(struct mutex *lock)
154 __cond_acquires(true, lock)
155{
156 unsigned long curr = (unsigned long)current;
157 unsigned long zero = 0UL;
158
159 MUTEX_WARN_ON(lock->magic != lock);
160
161 if (atomic_long_try_cmpxchg_acquire(&lock->owner, &zero, curr))
162 return true;
163
164 return false;
165}
166
167static __always_inline bool __mutex_unlock_fast(struct mutex *lock)
168 __cond_releases(true, lock)
169{
170 unsigned long curr = (unsigned long)current;
171
172 return atomic_long_try_cmpxchg_release(&lock->owner, &curr, 0UL);
173}
174
175#else /* !CONFIG_DEBUG_LOCK_ALLOC */
176
177void mutex_init_lockdep(struct mutex *lock, const char *name, struct lock_class_key *key)
178{
179 __mutex_init_generic(lock);
180
181 /*
182 * Make sure we are not reinitializing a held lock:
183 */
184 debug_check_no_locks_freed((void *)lock, sizeof(*lock));
185 lockdep_init_map_wait(&lock->dep_map, name, key, 0, LD_WAIT_SLEEP);
186}
187EXPORT_SYMBOL(mutex_init_lockdep);
188#endif /* !CONFIG_DEBUG_LOCK_ALLOC */
189
190static inline void __mutex_set_flag(struct mutex *lock, unsigned long flag)
191{
192 atomic_long_or(flag, &lock->owner);
193}
194
195static inline void __mutex_clear_flag(struct mutex *lock, unsigned long flag)
196{
197 atomic_long_andnot(flag, &lock->owner);
198}
199
200/*
201 * Add @waiter to the @lock wait_list and set the FLAG_WAITERS flag if it's
202 * the first waiter.
203 *
204 * When @pos, @waiter is added before the waiter indicated by @pos. Otherwise
205 * @waiter will be added to the tail of the list.
206 */
207static void
208__mutex_add_waiter(struct mutex *lock, struct mutex_waiter *waiter,
209 struct mutex_waiter *pos)
210 __must_hold(&lock->wait_lock)
211{
212 struct mutex_waiter *first = lock->first_waiter;
213
214 hung_task_set_blocker(lock, BLOCKER_TYPE_MUTEX);
215 debug_mutex_add_waiter(lock, waiter, current);
216
217 if (pos) {
218 /*
219 * Insert @waiter before @pos.
220 */
221 list_add_tail(&waiter->list, &pos->list);
222 /*
223 * If @pos == @first, then @waiter will be the new first.
224 */
225 if (pos == first)
226 lock->first_waiter = waiter;
227 return;
228 }
229
230 if (first) {
231 list_add_tail(&waiter->list, &first->list);
232 return;
233 }
234
235 INIT_LIST_HEAD(&waiter->list);
236 lock->first_waiter = waiter;
237 __mutex_set_flag(lock, MUTEX_FLAG_WAITERS);
238}
239
240static void
241__mutex_remove_waiter(struct mutex *lock, struct mutex_waiter *waiter)
242 __must_hold(&lock->wait_lock)
243{
244 if (list_empty(&waiter->list)) {
245 __mutex_clear_flag(lock, MUTEX_FLAGS);
246 lock->first_waiter = NULL;
247 } else {
248 if (lock->first_waiter == waiter)
249 lock->first_waiter = list_next_entry(waiter, list);
250 list_del(&waiter->list);
251 }
252
253 debug_mutex_remove_waiter(lock, waiter, current);
254 hung_task_clear_blocker();
255}
256
257/*
258 * Give up ownership to a specific task, when @task = NULL, this is equivalent
259 * to a regular unlock. Sets PICKUP on a handoff, clears HANDOFF, preserves
260 * WAITERS. Provides RELEASE semantics like a regular unlock, the
261 * __mutex_trylock() provides a matching ACQUIRE semantics for the handoff.
262 */
263static void __mutex_handoff(struct mutex *lock, struct task_struct *task)
264{
265 unsigned long owner = atomic_long_read(&lock->owner);
266
267 for (;;) {
268 unsigned long new;
269
270 MUTEX_WARN_ON(__owner_task(owner) != current);
271 MUTEX_WARN_ON(owner & MUTEX_FLAG_PICKUP);
272
273 new = (owner & MUTEX_FLAG_WAITERS);
274 new |= (unsigned long)task;
275 if (task)
276 new |= MUTEX_FLAG_PICKUP;
277
278 if (atomic_long_try_cmpxchg_release(&lock->owner, &owner, new))
279 break;
280 }
281}
282
283#ifndef CONFIG_DEBUG_LOCK_ALLOC
284/*
285 * We split the mutex lock/unlock logic into separate fastpath and
286 * slowpath functions, to reduce the register pressure on the fastpath.
287 * We also put the fastpath first in the kernel image, to make sure the
288 * branch is predicted by the CPU as default-untaken.
289 */
290static void __sched __mutex_lock_slowpath(struct mutex *lock)
291 __acquires(lock);
292
293/**
294 * mutex_lock - acquire the mutex
295 * @lock: the mutex to be acquired
296 *
297 * Lock the mutex exclusively for this task. If the mutex is not
298 * available right now, it will sleep until it can get it.
299 *
300 * The mutex must later on be released by the same task that
301 * acquired it. Recursive locking is not allowed. The task
302 * may not exit without first unlocking the mutex. Also, kernel
303 * memory where the mutex resides must not be freed with
304 * the mutex still locked. The mutex must first be initialized
305 * (or statically defined) before it can be locked. memset()-ing
306 * the mutex to 0 is not allowed.
307 *
308 * (The CONFIG_DEBUG_MUTEXES .config option turns on debugging
309 * checks that will enforce the restrictions and will also do
310 * deadlock debugging)
311 *
312 * This function is similar to (but not equivalent to) down().
313 */
314void __sched mutex_lock(struct mutex *lock)
315{
316 might_sleep();
317
318 if (!__mutex_trylock_fast(lock))
319 __mutex_lock_slowpath(lock);
320}
321EXPORT_SYMBOL(mutex_lock);
322#endif
323
324#include "ww_mutex.h"
325
326#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
327
328/*
329 * Trylock variant that returns the owning task on failure.
330 */
331static inline struct task_struct *__mutex_trylock_or_owner(struct mutex *lock)
332{
333 return __mutex_trylock_common(lock, false);
334}
335
336static inline
337bool ww_mutex_spin_on_owner(struct mutex *lock, struct ww_acquire_ctx *ww_ctx,
338 struct mutex_waiter *waiter)
339{
340 struct ww_mutex *ww;
341
342 ww = container_of(lock, struct ww_mutex, base);
343
344 /*
345 * If ww->ctx is set the contents are undefined, only
346 * by acquiring wait_lock there is a guarantee that
347 * they are not invalid when reading.
348 *
349 * As such, when deadlock detection needs to be
350 * performed the optimistic spinning cannot be done.
351 *
352 * Check this in every inner iteration because we may
353 * be racing against another thread's ww_mutex_lock.
354 */
355 if (ww_ctx->acquired > 0 && READ_ONCE(ww->ctx))
356 return false;
357
358 /*
359 * If we aren't on the wait list yet, cancel the spin
360 * if there are waiters. We want to avoid stealing the
361 * lock from a waiter with an earlier stamp, since the
362 * other thread may already own a lock that we also
363 * need.
364 */
365 if (!waiter && (atomic_long_read(&lock->owner) & MUTEX_FLAG_WAITERS))
366 return false;
367
368 /*
369 * Similarly, stop spinning if we are no longer the
370 * first waiter.
371 */
372 if (waiter && data_race(lock->first_waiter != waiter))
373 return false;
374
375 return true;
376}
377
378/*
379 * Look out! "owner" is an entirely speculative pointer access and not
380 * reliable.
381 *
382 * "noinline" so that this function shows up on perf profiles.
383 */
384static noinline
385bool mutex_spin_on_owner(struct mutex *lock, struct task_struct *owner,
386 struct ww_acquire_ctx *ww_ctx, struct mutex_waiter *waiter)
387{
388 bool ret = true;
389
390 lockdep_assert_preemption_disabled();
391
392 while (__mutex_owner(lock) == owner) {
393 /*
394 * Ensure we emit the owner->on_cpu, dereference _after_
395 * checking lock->owner still matches owner. And we already
396 * disabled preemption which is equal to the RCU read-side
397 * crital section in optimistic spinning code. Thus the
398 * task_strcut structure won't go away during the spinning
399 * period
400 */
401 barrier();
402
403 /*
404 * Use vcpu_is_preempted to detect lock holder preemption issue.
405 */
406 if (!owner_on_cpu(owner) || need_resched()) {
407 ret = false;
408 break;
409 }
410
411 if (ww_ctx && !ww_mutex_spin_on_owner(lock, ww_ctx, waiter)) {
412 ret = false;
413 break;
414 }
415
416 cpu_relax();
417 }
418
419 return ret;
420}
421
422/*
423 * Initial check for entering the mutex spinning loop
424 */
425static inline int mutex_can_spin_on_owner(struct mutex *lock)
426{
427 struct task_struct *owner;
428 int retval = 1;
429
430 lockdep_assert_preemption_disabled();
431
432 if (need_resched())
433 return 0;
434
435 /*
436 * We already disabled preemption which is equal to the RCU read-side
437 * crital section in optimistic spinning code. Thus the task_strcut
438 * structure won't go away during the spinning period.
439 */
440 owner = __mutex_owner(lock);
441 if (owner)
442 retval = owner_on_cpu(owner);
443
444 /*
445 * If lock->owner is not set, the mutex has been released. Return true
446 * such that we'll trylock in the spin path, which is a faster option
447 * than the blocking slow path.
448 */
449 return retval;
450}
451
452/*
453 * Optimistic spinning.
454 *
455 * We try to spin for acquisition when we find that the lock owner
456 * is currently running on a (different) CPU and while we don't
457 * need to reschedule. The rationale is that if the lock owner is
458 * running, it is likely to release the lock soon.
459 *
460 * The mutex spinners are queued up using MCS lock so that only one
461 * spinner can compete for the mutex. However, if mutex spinning isn't
462 * going to happen, there is no point in going through the lock/unlock
463 * overhead.
464 *
465 * Returns true when the lock was taken, otherwise false, indicating
466 * that we need to jump to the slowpath and sleep.
467 *
468 * The waiter flag is set to true if the spinner is a waiter in the wait
469 * queue. The waiter-spinner will spin on the lock directly and concurrently
470 * with the spinner at the head of the OSQ, if present, until the owner is
471 * changed to itself.
472 */
473static __always_inline bool
474mutex_optimistic_spin(struct mutex *lock, struct ww_acquire_ctx *ww_ctx,
475 struct mutex_waiter *waiter)
476{
477 if (!waiter) {
478 /*
479 * The purpose of the mutex_can_spin_on_owner() function is
480 * to eliminate the overhead of osq_lock() and osq_unlock()
481 * in case spinning isn't possible. As a waiter-spinner
482 * is not going to take OSQ lock anyway, there is no need
483 * to call mutex_can_spin_on_owner().
484 */
485 if (!mutex_can_spin_on_owner(lock))
486 goto fail;
487
488 /*
489 * In order to avoid a stampede of mutex spinners trying to
490 * acquire the mutex all at once, the spinners need to take a
491 * MCS (queued) lock first before spinning on the owner field.
492 */
493 if (!osq_lock(&lock->osq))
494 goto fail;
495 }
496
497 for (;;) {
498 struct task_struct *owner;
499
500 /* Try to acquire the mutex... */
501 owner = __mutex_trylock_or_owner(lock);
502 if (!owner)
503 break;
504
505 /*
506 * There's an owner, wait for it to either
507 * release the lock or go to sleep.
508 */
509 if (!mutex_spin_on_owner(lock, owner, ww_ctx, waiter))
510 goto fail_unlock;
511
512 /*
513 * The cpu_relax() call is a compiler barrier which forces
514 * everything in this loop to be re-loaded. We don't need
515 * memory barriers as we'll eventually observe the right
516 * values at the cost of a few extra spins.
517 */
518 cpu_relax();
519 }
520
521 if (!waiter)
522 osq_unlock(&lock->osq);
523
524 return true;
525
526
527fail_unlock:
528 if (!waiter)
529 osq_unlock(&lock->osq);
530
531fail:
532 /*
533 * If we fell out of the spin path because of need_resched(),
534 * reschedule now, before we try-lock the mutex. This avoids getting
535 * scheduled out right after we obtained the mutex.
536 */
537 if (need_resched()) {
538 /*
539 * We _should_ have TASK_RUNNING here, but just in case
540 * we do not, make it so, otherwise we might get stuck.
541 */
542 __set_current_state(TASK_RUNNING);
543 schedule_preempt_disabled();
544 }
545
546 return false;
547}
548#else
549static __always_inline bool
550mutex_optimistic_spin(struct mutex *lock, struct ww_acquire_ctx *ww_ctx,
551 struct mutex_waiter *waiter)
552{
553 return false;
554}
555#endif
556
557static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock, unsigned long ip)
558 __releases(lock);
559
560/**
561 * mutex_unlock - release the mutex
562 * @lock: the mutex to be released
563 *
564 * Unlock a mutex that has been locked by this task previously.
565 *
566 * This function must not be used in interrupt context. Unlocking
567 * of a not locked mutex is not allowed.
568 *
569 * The caller must ensure that the mutex stays alive until this function has
570 * returned - mutex_unlock() can NOT directly be used to release an object such
571 * that another concurrent task can free it.
572 * Mutexes are different from spinlocks & refcounts in this aspect.
573 *
574 * This function is similar to (but not equivalent to) up().
575 */
576void __sched mutex_unlock(struct mutex *lock)
577{
578#ifndef CONFIG_DEBUG_LOCK_ALLOC
579 if (__mutex_unlock_fast(lock))
580 return;
581#endif
582 __mutex_unlock_slowpath(lock, _RET_IP_);
583}
584EXPORT_SYMBOL(mutex_unlock);
585
586/**
587 * ww_mutex_unlock - release the w/w mutex
588 * @lock: the mutex to be released
589 *
590 * Unlock a mutex that has been locked by this task previously with any of the
591 * ww_mutex_lock* functions (with or without an acquire context). It is
592 * forbidden to release the locks after releasing the acquire context.
593 *
594 * This function must not be used in interrupt context. Unlocking
595 * of a unlocked mutex is not allowed.
596 */
597void __sched ww_mutex_unlock(struct ww_mutex *lock)
598 __no_context_analysis
599{
600 __ww_mutex_unlock(lock);
601 mutex_unlock(&lock->base);
602}
603EXPORT_SYMBOL(ww_mutex_unlock);
604
605/*
606 * Lock a mutex (possibly interruptible), slowpath:
607 */
608static __always_inline int __sched
609__mutex_lock_common(struct mutex *lock, unsigned int state, unsigned int subclass,
610 struct lockdep_map *nest_lock, unsigned long ip,
611 struct ww_acquire_ctx *ww_ctx, const bool use_ww_ctx)
612 __cond_acquires(0, lock)
613{
614 DEFINE_WAKE_Q(wake_q);
615 struct mutex_waiter waiter;
616 struct ww_mutex *ww;
617 unsigned long flags;
618 int ret;
619
620 if (!use_ww_ctx)
621 ww_ctx = NULL;
622
623 might_sleep();
624
625 MUTEX_WARN_ON(lock->magic != lock);
626
627 ww = container_of(lock, struct ww_mutex, base);
628 if (ww_ctx) {
629 if (unlikely(ww_ctx == READ_ONCE(ww->ctx)))
630 return -EALREADY;
631
632 /*
633 * Reset the wounded flag after a kill. No other process can
634 * race and wound us here since they can't have a valid owner
635 * pointer if we don't have any locks held.
636 */
637 if (ww_ctx->acquired == 0)
638 ww_ctx->wounded = 0;
639
640#ifdef CONFIG_DEBUG_LOCK_ALLOC
641 nest_lock = &ww_ctx->dep_map;
642#endif
643 }
644
645 preempt_disable();
646 mutex_acquire_nest(&lock->dep_map, subclass, 0, nest_lock, ip);
647
648 trace_contention_begin(lock, LCB_F_MUTEX | LCB_F_SPIN);
649 if (__mutex_trylock(lock) ||
650 mutex_optimistic_spin(lock, ww_ctx, NULL)) {
651 /* got the lock, yay! */
652 lock_acquired(&lock->dep_map, ip);
653 if (ww_ctx)
654 ww_mutex_set_context_fastpath(ww, ww_ctx);
655 trace_contention_end(lock, 0);
656 preempt_enable();
657 return 0;
658 }
659
660 raw_spin_lock_irqsave(&lock->wait_lock, flags);
661 /*
662 * After waiting to acquire the wait_lock, try again.
663 */
664 if (__mutex_trylock(lock)) {
665 if (ww_ctx)
666 __ww_mutex_check_waiters(lock, ww_ctx, &wake_q);
667
668 goto skip_wait;
669 }
670
671 debug_mutex_lock_common(lock, &waiter);
672 waiter.task = current;
673 if (use_ww_ctx)
674 waiter.ww_ctx = ww_ctx;
675
676 lock_contended(&lock->dep_map, ip);
677
678 if (!use_ww_ctx) {
679 /* add waiting tasks to the end of the waitqueue (FIFO): */
680 __mutex_add_waiter(lock, &waiter, NULL);
681 } else {
682 /*
683 * Add in stamp order, waking up waiters that must kill
684 * themselves.
685 */
686 ret = __ww_mutex_add_waiter(&waiter, lock, ww_ctx, &wake_q);
687 if (ret)
688 goto err_early_kill;
689 }
690
691 raw_spin_lock(&current->blocked_lock);
692 __set_task_blocked_on(current, lock);
693 set_current_state(state);
694 trace_contention_begin(lock, LCB_F_MUTEX);
695 for (;;) {
696 bool first;
697
698 /*
699 * Once we hold wait_lock, we're serialized against
700 * mutex_unlock() handing the lock off to us, do a trylock
701 * before testing the error conditions to make sure we pick up
702 * the handoff.
703 */
704 if (__mutex_trylock(lock))
705 break;
706
707 raw_spin_unlock(&current->blocked_lock);
708 /*
709 * Check for signals and kill conditions while holding
710 * wait_lock. This ensures the lock cancellation is ordered
711 * against mutex_unlock() and wake-ups do not go missing.
712 */
713 if (signal_pending_state(state, current)) {
714 ret = -EINTR;
715 goto err;
716 }
717
718 if (ww_ctx) {
719 ret = __ww_mutex_check_kill(lock, &waiter, ww_ctx);
720 if (ret)
721 goto err;
722 }
723
724 raw_spin_unlock_irqrestore_wake(&lock->wait_lock, flags, &wake_q);
725
726 schedule_preempt_disabled();
727
728 first = lock->first_waiter == &waiter;
729
730 raw_spin_lock_irqsave(&lock->wait_lock, flags);
731 raw_spin_lock(&current->blocked_lock);
732 /*
733 * As we likely have been woken up by task
734 * that has cleared our blocked_on state, re-set
735 * it to the lock we are trying to acquire.
736 */
737 __set_task_blocked_on(current, lock);
738 set_current_state(state);
739 /*
740 * Here we order against unlock; we must either see it change
741 * state back to RUNNING and fall through the next schedule(),
742 * or we must see its unlock and acquire.
743 */
744 if (__mutex_trylock_or_handoff(lock, first))
745 break;
746
747 if (first) {
748 bool opt_acquired;
749
750 /*
751 * mutex_optimistic_spin() can call schedule(), so
752 * we need to release these locks before calling it,
753 * and clear blocked on so we don't become unselectable
754 * to run.
755 */
756 __clear_task_blocked_on(current, lock);
757 raw_spin_unlock(&current->blocked_lock);
758 raw_spin_unlock_irqrestore(&lock->wait_lock, flags);
759
760 trace_contention_begin(lock, LCB_F_MUTEX | LCB_F_SPIN);
761 opt_acquired = mutex_optimistic_spin(lock, ww_ctx, &waiter);
762
763 raw_spin_lock_irqsave(&lock->wait_lock, flags);
764 raw_spin_lock(&current->blocked_lock);
765 __set_task_blocked_on(current, lock);
766
767 if (opt_acquired)
768 break;
769 trace_contention_begin(lock, LCB_F_MUTEX);
770 }
771 }
772 __clear_task_blocked_on(current, lock);
773 __set_current_state(TASK_RUNNING);
774 raw_spin_unlock(&current->blocked_lock);
775
776 if (ww_ctx) {
777 /*
778 * Wound-Wait; we stole the lock (!first_waiter), check the
779 * waiters as anyone might want to wound us.
780 */
781 if (!ww_ctx->is_wait_die && lock->first_waiter != &waiter)
782 __ww_mutex_check_waiters(lock, ww_ctx, &wake_q);
783 }
784
785 __mutex_remove_waiter(lock, &waiter);
786
787 debug_mutex_free_waiter(&waiter);
788
789skip_wait:
790 /* got the lock - cleanup and rejoice! */
791 lock_acquired(&lock->dep_map, ip);
792 trace_contention_end(lock, 0);
793
794 if (ww_ctx)
795 ww_mutex_lock_acquired(ww, ww_ctx);
796
797 raw_spin_unlock_irqrestore_wake(&lock->wait_lock, flags, &wake_q);
798 preempt_enable();
799 return 0;
800
801err:
802 clear_task_blocked_on(current, lock);
803 __set_current_state(TASK_RUNNING);
804 __mutex_remove_waiter(lock, &waiter);
805err_early_kill:
806 WARN_ON(get_task_blocked_on(current));
807 trace_contention_end(lock, ret);
808 raw_spin_unlock_irqrestore_wake(&lock->wait_lock, flags, &wake_q);
809 debug_mutex_free_waiter(&waiter);
810 mutex_release(&lock->dep_map, ip);
811 preempt_enable();
812 return ret;
813}
814
815static int __sched
816__mutex_lock(struct mutex *lock, unsigned int state, unsigned int subclass,
817 struct lockdep_map *nest_lock, unsigned long ip)
818 __cond_acquires(0, lock)
819{
820 return __mutex_lock_common(lock, state, subclass, nest_lock, ip, NULL, false);
821}
822
823static int __sched
824__ww_mutex_lock(struct mutex *lock, unsigned int state, unsigned int subclass,
825 unsigned long ip, struct ww_acquire_ctx *ww_ctx)
826 __cond_acquires(0, lock)
827{
828 return __mutex_lock_common(lock, state, subclass, NULL, ip, ww_ctx, true);
829}
830
831/**
832 * ww_mutex_trylock - tries to acquire the w/w mutex with optional acquire context
833 * @ww: mutex to lock
834 * @ww_ctx: optional w/w acquire context
835 *
836 * Trylocks a mutex with the optional acquire context; no deadlock detection is
837 * possible. Returns 1 if the mutex has been acquired successfully, 0 otherwise.
838 *
839 * Unlike ww_mutex_lock, no deadlock handling is performed. However, if a @ctx is
840 * specified, -EALREADY handling may happen in calls to ww_mutex_trylock.
841 *
842 * A mutex acquired with this function must be released with ww_mutex_unlock.
843 */
844int ww_mutex_trylock(struct ww_mutex *ww, struct ww_acquire_ctx *ww_ctx)
845{
846 if (!ww_ctx)
847 return mutex_trylock(&ww->base);
848
849 MUTEX_WARN_ON(ww->base.magic != &ww->base);
850
851 /*
852 * Reset the wounded flag after a kill. No other process can
853 * race and wound us here, since they can't have a valid owner
854 * pointer if we don't have any locks held.
855 */
856 if (ww_ctx->acquired == 0)
857 ww_ctx->wounded = 0;
858
859 if (__mutex_trylock(&ww->base)) {
860 ww_mutex_set_context_fastpath(ww, ww_ctx);
861 mutex_acquire_nest(&ww->base.dep_map, 0, 1, &ww_ctx->dep_map, _RET_IP_);
862 return 1;
863 }
864
865 return 0;
866}
867EXPORT_SYMBOL(ww_mutex_trylock);
868
869#ifdef CONFIG_DEBUG_LOCK_ALLOC
870void __sched
871mutex_lock_nested(struct mutex *lock, unsigned int subclass)
872{
873 __mutex_lock(lock, TASK_UNINTERRUPTIBLE, subclass, NULL, _RET_IP_);
874 __acquire(lock);
875}
876
877EXPORT_SYMBOL_GPL(mutex_lock_nested);
878
879void __sched
880_mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest)
881{
882 __mutex_lock(lock, TASK_UNINTERRUPTIBLE, 0, nest, _RET_IP_);
883 __acquire(lock);
884}
885EXPORT_SYMBOL_GPL(_mutex_lock_nest_lock);
886
887int __sched
888_mutex_lock_killable(struct mutex *lock, unsigned int subclass,
889 struct lockdep_map *nest)
890{
891 return __mutex_lock(lock, TASK_KILLABLE, subclass, nest, _RET_IP_);
892}
893EXPORT_SYMBOL_GPL(_mutex_lock_killable);
894
895int __sched
896mutex_lock_interruptible_nested(struct mutex *lock, unsigned int subclass)
897{
898 return __mutex_lock(lock, TASK_INTERRUPTIBLE, subclass, NULL, _RET_IP_);
899}
900EXPORT_SYMBOL_GPL(mutex_lock_interruptible_nested);
901
902void __sched
903mutex_lock_io_nested(struct mutex *lock, unsigned int subclass)
904{
905 int token;
906
907 might_sleep();
908
909 token = io_schedule_prepare();
910 __mutex_lock_common(lock, TASK_UNINTERRUPTIBLE,
911 subclass, NULL, _RET_IP_, NULL, 0);
912 __acquire(lock);
913 io_schedule_finish(token);
914}
915EXPORT_SYMBOL_GPL(mutex_lock_io_nested);
916
917static inline int
918ww_mutex_deadlock_injection(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
919 __cond_releases(nonzero, lock)
920{
921#ifdef CONFIG_DEBUG_WW_MUTEX_SLOWPATH
922 unsigned tmp;
923
924 if (ctx->deadlock_inject_countdown-- == 0) {
925 tmp = ctx->deadlock_inject_interval;
926 if (tmp > UINT_MAX/4)
927 tmp = UINT_MAX;
928 else
929 tmp = tmp*2 + tmp + tmp/2;
930
931 ctx->deadlock_inject_interval = tmp;
932 ctx->deadlock_inject_countdown = tmp;
933 ctx->contending_lock = lock;
934
935 ww_mutex_unlock(lock);
936
937 return -EDEADLK;
938 }
939#endif
940
941 return 0;
942}
943
944int __sched
945ww_mutex_lock(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
946{
947 int ret;
948
949 might_sleep();
950 ret = __ww_mutex_lock(&lock->base, TASK_UNINTERRUPTIBLE,
951 0, _RET_IP_, ctx);
952 if (!ret && ctx && ctx->acquired > 1)
953 return ww_mutex_deadlock_injection(lock, ctx);
954
955 return ret;
956}
957EXPORT_SYMBOL_GPL(ww_mutex_lock);
958
959int __sched
960ww_mutex_lock_interruptible(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
961{
962 int ret;
963
964 might_sleep();
965 ret = __ww_mutex_lock(&lock->base, TASK_INTERRUPTIBLE,
966 0, _RET_IP_, ctx);
967
968 if (!ret && ctx && ctx->acquired > 1)
969 return ww_mutex_deadlock_injection(lock, ctx);
970
971 return ret;
972}
973EXPORT_SYMBOL_GPL(ww_mutex_lock_interruptible);
974
975#endif
976
977/*
978 * Release the lock, slowpath:
979 */
980static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock, unsigned long ip)
981 __releases(lock)
982{
983 struct task_struct *next = NULL;
984 struct mutex_waiter *waiter;
985 DEFINE_WAKE_Q(wake_q);
986 unsigned long owner;
987 unsigned long flags;
988
989 mutex_release(&lock->dep_map, ip);
990 __release(lock);
991
992 /*
993 * Release the lock before (potentially) taking the spinlock such that
994 * other contenders can get on with things ASAP.
995 *
996 * Except when HANDOFF, in that case we must not clear the owner field,
997 * but instead set it to the top waiter.
998 */
999 owner = atomic_long_read(&lock->owner);
1000 for (;;) {
1001 MUTEX_WARN_ON(__owner_task(owner) != current);
1002 MUTEX_WARN_ON(owner & MUTEX_FLAG_PICKUP);
1003
1004 if (owner & MUTEX_FLAG_HANDOFF)
1005 break;
1006
1007 if (atomic_long_try_cmpxchg_release(&lock->owner, &owner, __owner_flags(owner))) {
1008 if (owner & MUTEX_FLAG_WAITERS)
1009 break;
1010
1011 return;
1012 }
1013 }
1014
1015 raw_spin_lock_irqsave(&lock->wait_lock, flags);
1016 debug_mutex_unlock(lock);
1017 waiter = lock->first_waiter;
1018 if (waiter) {
1019 next = waiter->task;
1020
1021 debug_mutex_wake_waiter(lock, waiter);
1022 set_task_blocked_on_waking(next, lock);
1023 wake_q_add(&wake_q, next);
1024 }
1025
1026 if (owner & MUTEX_FLAG_HANDOFF)
1027 __mutex_handoff(lock, next);
1028
1029 raw_spin_unlock_irqrestore_wake(&lock->wait_lock, flags, &wake_q);
1030}
1031
1032#ifndef CONFIG_DEBUG_LOCK_ALLOC
1033/*
1034 * Here come the less common (and hence less performance-critical) APIs:
1035 * mutex_lock_interruptible() and mutex_trylock().
1036 */
1037static noinline int __sched
1038__mutex_lock_killable_slowpath(struct mutex *lock);
1039
1040static noinline int __sched
1041__mutex_lock_interruptible_slowpath(struct mutex *lock);
1042
1043/**
1044 * mutex_lock_interruptible() - Acquire the mutex, interruptible by signals.
1045 * @lock: The mutex to be acquired.
1046 *
1047 * Lock the mutex like mutex_lock(). If a signal is delivered while the
1048 * process is sleeping, this function will return without acquiring the
1049 * mutex.
1050 *
1051 * Context: Process context.
1052 * Return: 0 if the lock was successfully acquired or %-EINTR if a
1053 * signal arrived.
1054 */
1055int __sched mutex_lock_interruptible(struct mutex *lock)
1056{
1057 might_sleep();
1058
1059 if (__mutex_trylock_fast(lock))
1060 return 0;
1061
1062 return __mutex_lock_interruptible_slowpath(lock);
1063}
1064
1065EXPORT_SYMBOL(mutex_lock_interruptible);
1066
1067/**
1068 * mutex_lock_killable() - Acquire the mutex, interruptible by fatal signals.
1069 * @lock: The mutex to be acquired.
1070 *
1071 * Lock the mutex like mutex_lock(). If a signal which will be fatal to
1072 * the current process is delivered while the process is sleeping, this
1073 * function will return without acquiring the mutex.
1074 *
1075 * Context: Process context.
1076 * Return: 0 if the lock was successfully acquired or %-EINTR if a
1077 * fatal signal arrived.
1078 */
1079int __sched mutex_lock_killable(struct mutex *lock)
1080{
1081 might_sleep();
1082
1083 if (__mutex_trylock_fast(lock))
1084 return 0;
1085
1086 return __mutex_lock_killable_slowpath(lock);
1087}
1088EXPORT_SYMBOL(mutex_lock_killable);
1089
1090/**
1091 * mutex_lock_io() - Acquire the mutex and mark the process as waiting for I/O
1092 * @lock: The mutex to be acquired.
1093 *
1094 * Lock the mutex like mutex_lock(). While the task is waiting for this
1095 * mutex, it will be accounted as being in the IO wait state by the
1096 * scheduler.
1097 *
1098 * Context: Process context.
1099 */
1100void __sched mutex_lock_io(struct mutex *lock)
1101{
1102 int token;
1103
1104 token = io_schedule_prepare();
1105 mutex_lock(lock);
1106 io_schedule_finish(token);
1107}
1108EXPORT_SYMBOL_GPL(mutex_lock_io);
1109
1110static noinline void __sched
1111__mutex_lock_slowpath(struct mutex *lock)
1112 __acquires(lock)
1113{
1114 __mutex_lock(lock, TASK_UNINTERRUPTIBLE, 0, NULL, _RET_IP_);
1115 __acquire(lock);
1116}
1117
1118static noinline int __sched
1119__mutex_lock_killable_slowpath(struct mutex *lock)
1120 __cond_acquires(0, lock)
1121{
1122 return __mutex_lock(lock, TASK_KILLABLE, 0, NULL, _RET_IP_);
1123}
1124
1125static noinline int __sched
1126__mutex_lock_interruptible_slowpath(struct mutex *lock)
1127 __cond_acquires(0, lock)
1128{
1129 return __mutex_lock(lock, TASK_INTERRUPTIBLE, 0, NULL, _RET_IP_);
1130}
1131
1132static noinline int __sched
1133__ww_mutex_lock_slowpath(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
1134 __cond_acquires(0, lock)
1135{
1136 return __ww_mutex_lock(&lock->base, TASK_UNINTERRUPTIBLE, 0,
1137 _RET_IP_, ctx);
1138}
1139
1140static noinline int __sched
1141__ww_mutex_lock_interruptible_slowpath(struct ww_mutex *lock,
1142 struct ww_acquire_ctx *ctx)
1143 __cond_acquires(0, lock)
1144{
1145 return __ww_mutex_lock(&lock->base, TASK_INTERRUPTIBLE, 0,
1146 _RET_IP_, ctx);
1147}
1148
1149#endif
1150
1151#ifndef CONFIG_DEBUG_LOCK_ALLOC
1152/**
1153 * mutex_trylock - try to acquire the mutex, without waiting
1154 * @lock: the mutex to be acquired
1155 *
1156 * Try to acquire the mutex atomically. Returns 1 if the mutex
1157 * has been acquired successfully, and 0 on contention.
1158 *
1159 * NOTE: this function follows the spin_trylock() convention, so
1160 * it is negated from the down_trylock() return values! Be careful
1161 * about this when converting semaphore users to mutexes.
1162 *
1163 * This function must not be used in interrupt context. The
1164 * mutex must be released by the same task that acquired it.
1165 */
1166int __sched mutex_trylock(struct mutex *lock)
1167{
1168 MUTEX_WARN_ON(lock->magic != lock);
1169 return __mutex_trylock(lock);
1170}
1171EXPORT_SYMBOL(mutex_trylock);
1172#else
1173int __sched _mutex_trylock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock)
1174{
1175 bool locked;
1176
1177 MUTEX_WARN_ON(lock->magic != lock);
1178 locked = __mutex_trylock(lock);
1179 if (locked)
1180 mutex_acquire_nest(&lock->dep_map, 0, 1, nest_lock, _RET_IP_);
1181
1182 return locked;
1183}
1184EXPORT_SYMBOL(_mutex_trylock_nest_lock);
1185#endif
1186
1187#ifndef CONFIG_DEBUG_LOCK_ALLOC
1188int __sched
1189ww_mutex_lock(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
1190{
1191 might_sleep();
1192
1193 if (__mutex_trylock_fast(&lock->base)) {
1194 if (ctx)
1195 ww_mutex_set_context_fastpath(lock, ctx);
1196 return 0;
1197 }
1198
1199 return __ww_mutex_lock_slowpath(lock, ctx);
1200}
1201EXPORT_SYMBOL(ww_mutex_lock);
1202
1203int __sched
1204ww_mutex_lock_interruptible(struct ww_mutex *lock, struct ww_acquire_ctx *ctx)
1205{
1206 might_sleep();
1207
1208 if (__mutex_trylock_fast(&lock->base)) {
1209 if (ctx)
1210 ww_mutex_set_context_fastpath(lock, ctx);
1211 return 0;
1212 }
1213
1214 return __ww_mutex_lock_interruptible_slowpath(lock, ctx);
1215}
1216EXPORT_SYMBOL(ww_mutex_lock_interruptible);
1217
1218#endif /* !CONFIG_DEBUG_LOCK_ALLOC */
1219#endif /* !CONFIG_PREEMPT_RT */
1220
1221EXPORT_TRACEPOINT_SYMBOL_GPL(contention_begin);
1222EXPORT_TRACEPOINT_SYMBOL_GPL(contention_end);
1223
1224/**
1225 * atomic_dec_and_mutex_lock - return holding mutex if we dec to 0
1226 * @cnt: the atomic which we are to dec
1227 * @lock: the mutex to return holding if we dec to 0
1228 *
1229 * return true and hold lock if we dec to 0, return false otherwise
1230 */
1231int atomic_dec_and_mutex_lock(atomic_t *cnt, struct mutex *lock)
1232{
1233 /* dec if we can't possibly hit 0 */
1234 if (atomic_add_unless(cnt, -1, 1))
1235 return 0;
1236 /* we might hit 0, so take the lock */
1237 mutex_lock(lock);
1238 if (!atomic_dec_and_test(cnt)) {
1239 /* when we actually did the dec, we didn't hit 0 */
1240 mutex_unlock(lock);
1241 return 0;
1242 }
1243 /* we hit 0, and we hold the lock */
1244 return 1;
1245}
1246EXPORT_SYMBOL(atomic_dec_and_mutex_lock);