xref: /netbsd-src/lib/libpthread/pthread_mutex.c (revision 48360965f30c307b6836d0d898d15ce6c1d9b387)
1 /*	$NetBSD: pthread_mutex.c,v 1.51 2008/08/02 19:46:30 matt Exp $	*/
2 
3 /*-
4  * Copyright (c) 2001, 2003, 2006, 2007, 2008 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Nathan J. Williams, by Jason R. Thorpe, and by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * To track threads waiting for mutexes to be released, we use lockless
34  * lists built on atomic operations and memory barriers.
35  *
36  * A simple spinlock would be faster and make the code easier to
37  * follow, but spinlocks are problematic in userspace.  If a thread is
38  * preempted by the kernel while holding a spinlock, any other thread
39  * attempting to acquire that spinlock will needlessly busy wait.
40  *
41  * There is no good way to know that the holding thread is no longer
42  * running, nor to request a wake-up once it has begun running again.
43  * Of more concern, threads in the SCHED_FIFO class do not have a
44  * limited time quantum and so could spin forever, preventing the
45  * thread holding the spinlock from getting CPU time: it would never
46  * be released.
47  */
48 
49 #include <sys/cdefs.h>
50 __RCSID("$NetBSD: pthread_mutex.c,v 1.51 2008/08/02 19:46:30 matt Exp $");
51 
52 #include <sys/types.h>
53 #include <sys/lwpctl.h>
54 #include <sys/lock.h>
55 
56 #include <errno.h>
57 #include <limits.h>
58 #include <stdlib.h>
59 #include <string.h>
60 #include <stdio.h>
61 
62 #include "pthread.h"
63 #include "pthread_int.h"
64 
65 #define	MUTEX_WAITERS_BIT		((uintptr_t)0x01)
66 #define	MUTEX_RECURSIVE_BIT		((uintptr_t)0x02)
67 #define	MUTEX_DEFERRED_BIT		((uintptr_t)0x04)
68 #define	MUTEX_THREAD			((uintptr_t)-16L)
69 
70 #define	MUTEX_HAS_WAITERS(x)		((uintptr_t)(x) & MUTEX_WAITERS_BIT)
71 #define	MUTEX_RECURSIVE(x)		((uintptr_t)(x) & MUTEX_RECURSIVE_BIT)
72 #define	MUTEX_OWNER(x)			((uintptr_t)(x) & MUTEX_THREAD)
73 
74 #if __GNUC_PREREQ__(3, 0)
75 #define	NOINLINE		__attribute ((noinline))
76 #else
77 #define	NOINLINE		/* nothing */
78 #endif
79 
80 static void	pthread__mutex_wakeup(pthread_t, pthread_mutex_t *);
81 static int	pthread__mutex_lock_slow(pthread_mutex_t *);
82 static int	pthread__mutex_unlock_slow(pthread_mutex_t *);
83 static void	pthread__mutex_pause(void);
84 
85 int		_pthread_mutex_held_np(pthread_mutex_t *);
86 pthread_t	_pthread_mutex_owner_np(pthread_mutex_t *);
87 
88 __weak_alias(pthread_mutex_held_np,_pthread_mutex_held_np)
89 __weak_alias(pthread_mutex_owner_np,_pthread_mutex_owner_np)
90 
91 __strong_alias(__libc_mutex_init,pthread_mutex_init)
92 __strong_alias(__libc_mutex_lock,pthread_mutex_lock)
93 __strong_alias(__libc_mutex_trylock,pthread_mutex_trylock)
94 __strong_alias(__libc_mutex_unlock,pthread_mutex_unlock)
95 __strong_alias(__libc_mutex_destroy,pthread_mutex_destroy)
96 
97 __strong_alias(__libc_mutexattr_init,pthread_mutexattr_init)
98 __strong_alias(__libc_mutexattr_destroy,pthread_mutexattr_destroy)
99 __strong_alias(__libc_mutexattr_settype,pthread_mutexattr_settype)
100 
101 __strong_alias(__libc_thr_once,pthread_once)
102 
103 int
104 pthread_mutex_init(pthread_mutex_t *ptm, const pthread_mutexattr_t *attr)
105 {
106 	intptr_t type;
107 
108 	if (attr == NULL)
109 		type = PTHREAD_MUTEX_NORMAL;
110 	else
111 		type = (intptr_t)attr->ptma_private;
112 
113 	switch (type) {
114 	case PTHREAD_MUTEX_ERRORCHECK:
115 		__cpu_simple_lock_set(&ptm->ptm_errorcheck);
116 		ptm->ptm_owner = NULL;
117 		break;
118 	case PTHREAD_MUTEX_RECURSIVE:
119 		__cpu_simple_lock_clear(&ptm->ptm_errorcheck);
120 		ptm->ptm_owner = (void *)MUTEX_RECURSIVE_BIT;
121 		break;
122 	default:
123 		__cpu_simple_lock_clear(&ptm->ptm_errorcheck);
124 		ptm->ptm_owner = NULL;
125 		break;
126 	}
127 
128 	ptm->ptm_magic = _PT_MUTEX_MAGIC;
129 	ptm->ptm_waiters = NULL;
130 	ptm->ptm_recursed = 0;
131 
132 	return 0;
133 }
134 
135 
136 int
137 pthread_mutex_destroy(pthread_mutex_t *ptm)
138 {
139 
140 	pthread__error(EINVAL, "Invalid mutex",
141 	    ptm->ptm_magic == _PT_MUTEX_MAGIC);
142 	pthread__error(EBUSY, "Destroying locked mutex",
143 	    MUTEX_OWNER(ptm->ptm_owner) == 0);
144 
145 	ptm->ptm_magic = _PT_MUTEX_DEAD;
146 	return 0;
147 }
148 
149 int
150 pthread_mutex_lock(pthread_mutex_t *ptm)
151 {
152 	pthread_t self;
153 	void *val;
154 
155 	self = pthread__self();
156 	val = atomic_cas_ptr(&ptm->ptm_owner, NULL, self);
157 	if (__predict_true(val == NULL)) {
158 #ifndef PTHREAD__ATOMIC_IS_MEMBAR
159 		membar_enter();
160 #endif
161 		return 0;
162 	}
163 	return pthread__mutex_lock_slow(ptm);
164 }
165 
166 /* We want function call overhead. */
167 NOINLINE static void
168 pthread__mutex_pause(void)
169 {
170 
171 	pthread__smt_pause();
172 }
173 
174 /*
175  * Spin while the holder is running.  'lwpctl' gives us the true
176  * status of the thread.  pt_blocking is set by libpthread in order
177  * to cut out system call and kernel spinlock overhead on remote CPUs
178  * (could represent many thousands of clock cycles).  pt_blocking also
179  * makes this thread yield if the target is calling sched_yield().
180  */
181 NOINLINE static void *
182 pthread__mutex_spin(pthread_mutex_t *ptm, pthread_t owner)
183 {
184 	pthread_t thread;
185 	unsigned int count, i;
186 
187 	for (count = 2;; owner = ptm->ptm_owner) {
188 		thread = (pthread_t)MUTEX_OWNER(owner);
189 		if (thread == NULL)
190 			break;
191 		if (thread->pt_lwpctl->lc_curcpu == LWPCTL_CPU_NONE ||
192 		    thread->pt_blocking)
193 			break;
194 		if (count < 128)
195 			count += count;
196 		for (i = count; i != 0; i--)
197 			pthread__mutex_pause();
198 	}
199 
200 	return owner;
201 }
202 
203 NOINLINE static int
204 pthread__mutex_lock_slow(pthread_mutex_t *ptm)
205 {
206 	void *waiters, *new, *owner, *next;
207 	pthread_t self;
208 
209 	pthread__error(EINVAL, "Invalid mutex",
210 	    ptm->ptm_magic == _PT_MUTEX_MAGIC);
211 
212 	owner = ptm->ptm_owner;
213 	self = pthread__self();
214 
215 	/* Recursive or errorcheck? */
216 	if (MUTEX_OWNER(owner) == (uintptr_t)self) {
217 		if (MUTEX_RECURSIVE(owner)) {
218 			if (ptm->ptm_recursed == INT_MAX)
219 				return EAGAIN;
220 			ptm->ptm_recursed++;
221 			return 0;
222 		}
223 		if (__SIMPLELOCK_LOCKED_P(&ptm->ptm_errorcheck))
224 			return EDEADLK;
225 	}
226 
227 	for (;; owner = ptm->ptm_owner) {
228 		/* Spin while the owner is running. */
229 		owner = pthread__mutex_spin(ptm, owner);
230 
231 		/* If it has become free, try to acquire it again. */
232 		if (MUTEX_OWNER(owner) == 0) {
233 			do {
234 				new = (void *)
235 				    ((uintptr_t)self | (uintptr_t)owner);
236 				next = atomic_cas_ptr(&ptm->ptm_owner, owner,
237 				    new);
238 				if (next == owner) {
239 #ifndef PTHREAD__ATOMIC_IS_MEMBAR
240 					membar_enter();
241 #endif
242 					return 0;
243 				}
244 				owner = next;
245 			} while (MUTEX_OWNER(owner) == 0);
246 			/*
247 			 * We have lost the race to acquire the mutex.
248 			 * The new owner could be running on another
249 			 * CPU, in which case we should spin and avoid
250 			 * the overhead of blocking.
251 			 */
252 			continue;
253 		}
254 
255 		/*
256 		 * Nope, still held.  Add thread to the list of waiters.
257 		 * Issue a memory barrier to ensure mutexwait/mutexnext
258 		 * are visible before we enter the waiters list.
259 		 */
260 		self->pt_mutexwait = 1;
261 		for (waiters = ptm->ptm_waiters;; waiters = next) {
262 			self->pt_mutexnext = waiters;
263 			membar_producer();
264 			next = atomic_cas_ptr(&ptm->ptm_waiters, waiters, self);
265 			if (next == waiters)
266 			    	break;
267 		}
268 
269 		/*
270 		 * Set the waiters bit and block.
271 		 *
272 		 * Note that the mutex can become unlocked before we set
273 		 * the waiters bit.  If that happens it's not safe to sleep
274 		 * as we may never be awoken: we must remove the current
275 		 * thread from the waiters list and try again.
276 		 *
277 		 * Because we are doing this atomically, we can't remove
278 		 * one waiter: we must remove all waiters and awken them,
279 		 * then sleep in _lwp_park() until we have been awoken.
280 		 *
281 		 * Issue a memory barrier to ensure that we are reading
282 		 * the value of ptm_owner/pt_mutexwait after we have entered
283 		 * the waiters list (the CAS itself must be atomic).
284 		 */
285 		membar_consumer();
286 		for (owner = ptm->ptm_owner;; owner = next) {
287 			if (MUTEX_HAS_WAITERS(owner))
288 				break;
289 			if (MUTEX_OWNER(owner) == 0) {
290 				pthread__mutex_wakeup(self, ptm);
291 				break;
292 			}
293 			new = (void *)((uintptr_t)owner | MUTEX_WAITERS_BIT);
294 			next = atomic_cas_ptr(&ptm->ptm_owner, owner, new);
295 			if (next == owner) {
296 				/*
297 				 * pthread_mutex_unlock() can do a
298 				 * non-interlocked CAS.  We cannot
299 				 * know if our attempt to set the
300 				 * waiters bit has succeeded while
301 				 * the holding thread is running.
302 				 * There are many assumptions; see
303 				 * sys/kern/kern_mutex.c for details.
304 				 * In short, we must spin if we see
305 				 * that the holder is running again.
306 				 */
307 				membar_sync();
308 				next = pthread__mutex_spin(ptm, owner);
309 			}
310 		}
311 
312 		/*
313 		 * We may have been awoken by the current thread above,
314 		 * or will be awoken by the current holder of the mutex.
315 		 * The key requirement is that we must not proceed until
316 		 * told that we are no longer waiting (via pt_mutexwait
317 		 * being set to zero).  Otherwise it is unsafe to re-enter
318 		 * the thread onto the waiters list.
319 		 */
320 		while (self->pt_mutexwait) {
321 			self->pt_blocking++;
322 			(void)_lwp_park(NULL, self->pt_unpark,
323 			    __UNVOLATILE(&ptm->ptm_waiters),
324 			    __UNVOLATILE(&ptm->ptm_waiters));
325 			self->pt_unpark = 0;
326 			self->pt_blocking--;
327 			membar_sync();
328 		}
329 	}
330 }
331 
332 int
333 pthread_mutex_trylock(pthread_mutex_t *ptm)
334 {
335 	pthread_t self;
336 	void *val, *new, *next;
337 
338 	self = pthread__self();
339 	val = atomic_cas_ptr(&ptm->ptm_owner, NULL, self);
340 	if (__predict_true(val == NULL)) {
341 #ifndef PTHREAD__ATOMIC_IS_MEMBAR
342 		membar_enter();
343 #endif
344 		return 0;
345 	}
346 
347 	if (MUTEX_RECURSIVE(val)) {
348 		if (MUTEX_OWNER(val) == 0) {
349 			new = (void *)((uintptr_t)self | (uintptr_t)val);
350 			next = atomic_cas_ptr(&ptm->ptm_owner, val, new);
351 			if (__predict_true(next == val)) {
352 #ifndef PTHREAD__ATOMIC_IS_MEMBAR
353 				membar_enter();
354 #endif
355 				return 0;
356 			}
357 		}
358 		if (MUTEX_OWNER(val) == (uintptr_t)self) {
359 			if (ptm->ptm_recursed == INT_MAX)
360 				return EAGAIN;
361 			ptm->ptm_recursed++;
362 			return 0;
363 		}
364 	}
365 
366 	return EBUSY;
367 }
368 
369 int
370 pthread_mutex_unlock(pthread_mutex_t *ptm)
371 {
372 	pthread_t self;
373 	void *value;
374 
375 	/*
376 	 * Note this may be a non-interlocked CAS.  See lock_slow()
377 	 * above and sys/kern/kern_mutex.c for details.
378 	 */
379 #ifndef PTHREAD__ATOMIC_IS_MEMBAR
380 	membar_exit();
381 #endif
382 	self = pthread__self();
383 	value = atomic_cas_ptr_ni(&ptm->ptm_owner, self, NULL);
384 	if (__predict_true(value == self))
385 		return 0;
386 	return pthread__mutex_unlock_slow(ptm);
387 }
388 
389 NOINLINE static int
390 pthread__mutex_unlock_slow(pthread_mutex_t *ptm)
391 {
392 	pthread_t self, owner, new;
393 	int weown, error, deferred;
394 
395 	pthread__error(EINVAL, "Invalid mutex",
396 	    ptm->ptm_magic == _PT_MUTEX_MAGIC);
397 
398 	self = pthread__self();
399 	owner = ptm->ptm_owner;
400 	weown = (MUTEX_OWNER(owner) == (uintptr_t)self);
401 	deferred = (int)((uintptr_t)owner & MUTEX_DEFERRED_BIT);
402 	error = 0;
403 
404 	if (__SIMPLELOCK_LOCKED_P(&ptm->ptm_errorcheck)) {
405 		if (!weown) {
406 			error = EPERM;
407 			new = owner;
408 		} else {
409 			new = NULL;
410 		}
411 	} else if (MUTEX_RECURSIVE(owner)) {
412 		if (!weown) {
413 			error = EPERM;
414 			new = owner;
415 		} else if (ptm->ptm_recursed) {
416 			ptm->ptm_recursed--;
417 			new = owner;
418 		} else {
419 			new = (pthread_t)MUTEX_RECURSIVE_BIT;
420 		}
421 	} else {
422 		pthread__error(EPERM,
423 		    "Unlocking unlocked mutex", (owner != NULL));
424 		pthread__error(EPERM,
425 		    "Unlocking mutex owned by another thread", weown);
426 		new = NULL;
427 	}
428 
429 	/*
430 	 * Release the mutex.  If there appear to be waiters, then
431 	 * wake them up.
432 	 */
433 	if (new != owner) {
434 		owner = atomic_swap_ptr(&ptm->ptm_owner, new);
435 		if (MUTEX_HAS_WAITERS(owner) != 0) {
436 			pthread__mutex_wakeup(self, ptm);
437 			return 0;
438 		}
439 	}
440 
441 	/*
442 	 * There were no waiters, but we may have deferred waking
443 	 * other threads until mutex unlock - we must wake them now.
444 	 */
445 	if (!deferred)
446 		return error;
447 
448 	if (self->pt_nwaiters == 1) {
449 		/*
450 		 * If the calling thread is about to block, defer
451 		 * unparking the target until _lwp_park() is called.
452 		 */
453 		if (self->pt_willpark && self->pt_unpark == 0) {
454 			self->pt_unpark = self->pt_waiters[0];
455 		} else {
456 			(void)_lwp_unpark(self->pt_waiters[0],
457 			    __UNVOLATILE(&ptm->ptm_waiters));
458 		}
459 	} else {
460 		(void)_lwp_unpark_all(self->pt_waiters, self->pt_nwaiters,
461 		    __UNVOLATILE(&ptm->ptm_waiters));
462 	}
463 	self->pt_nwaiters = 0;
464 
465 	return error;
466 }
467 
468 static void
469 pthread__mutex_wakeup(pthread_t self, pthread_mutex_t *ptm)
470 {
471 	pthread_t thread, next;
472 	ssize_t n, rv;
473 
474 	/*
475 	 * Take ownership of the current set of waiters.  No
476 	 * need for a memory barrier following this, all loads
477 	 * are dependent upon 'thread'.
478 	 */
479 	thread = atomic_swap_ptr(&ptm->ptm_waiters, NULL);
480 
481 	for (;;) {
482 		/*
483 		 * Pull waiters from the queue and add to our list.
484 		 * Use a memory barrier to ensure that we safely
485 		 * read the value of pt_mutexnext before 'thread'
486 		 * sees pt_mutexwait being cleared.
487 		 */
488 		for (n = self->pt_nwaiters, self->pt_nwaiters = 0;
489 		    n < pthread__unpark_max && thread != NULL;
490 		    thread = next) {
491 		    	next = thread->pt_mutexnext;
492 		    	if (thread != self) {
493 				self->pt_waiters[n++] = thread->pt_lid;
494 				membar_sync();
495 			}
496 			thread->pt_mutexwait = 0;
497 			/* No longer safe to touch 'thread' */
498 		}
499 
500 		switch (n) {
501 		case 0:
502 			return;
503 		case 1:
504 			/*
505 			 * If the calling thread is about to block,
506 			 * defer unparking the target until _lwp_park()
507 			 * is called.
508 			 */
509 			if (self->pt_willpark && self->pt_unpark == 0) {
510 				self->pt_unpark = self->pt_waiters[0];
511 				return;
512 			}
513 			rv = (ssize_t)_lwp_unpark(self->pt_waiters[0],
514 			    __UNVOLATILE(&ptm->ptm_waiters));
515 			if (rv != 0 && errno != EALREADY && errno != EINTR &&
516 			    errno != ESRCH) {
517 				pthread__errorfunc(__FILE__, __LINE__,
518 				    __func__, "_lwp_unpark failed");
519 			}
520 			return;
521 		default:
522 			rv = _lwp_unpark_all(self->pt_waiters, (size_t)n,
523 			    __UNVOLATILE(&ptm->ptm_waiters));
524 			if (rv != 0 && errno != EINTR) {
525 				pthread__errorfunc(__FILE__, __LINE__,
526 				    __func__, "_lwp_unpark_all failed");
527 			}
528 			break;
529 		}
530 	}
531 }
532 int
533 pthread_mutexattr_init(pthread_mutexattr_t *attr)
534 {
535 
536 	attr->ptma_magic = _PT_MUTEXATTR_MAGIC;
537 	attr->ptma_private = (void *)PTHREAD_MUTEX_DEFAULT;
538 	return 0;
539 }
540 
541 int
542 pthread_mutexattr_destroy(pthread_mutexattr_t *attr)
543 {
544 
545 	pthread__error(EINVAL, "Invalid mutex attribute",
546 	    attr->ptma_magic == _PT_MUTEXATTR_MAGIC);
547 
548 	return 0;
549 }
550 
551 
552 int
553 pthread_mutexattr_gettype(const pthread_mutexattr_t *attr, int *typep)
554 {
555 
556 	pthread__error(EINVAL, "Invalid mutex attribute",
557 	    attr->ptma_magic == _PT_MUTEXATTR_MAGIC);
558 
559 	*typep = (int)(intptr_t)attr->ptma_private;
560 	return 0;
561 }
562 
563 
564 int
565 pthread_mutexattr_settype(pthread_mutexattr_t *attr, int type)
566 {
567 
568 	pthread__error(EINVAL, "Invalid mutex attribute",
569 	    attr->ptma_magic == _PT_MUTEXATTR_MAGIC);
570 
571 	switch (type) {
572 	case PTHREAD_MUTEX_NORMAL:
573 	case PTHREAD_MUTEX_ERRORCHECK:
574 	case PTHREAD_MUTEX_RECURSIVE:
575 		attr->ptma_private = (void *)(intptr_t)type;
576 		return 0;
577 	default:
578 		return EINVAL;
579 	}
580 }
581 
582 
583 static void
584 once_cleanup(void *closure)
585 {
586 
587        pthread_mutex_unlock((pthread_mutex_t *)closure);
588 }
589 
590 
591 int
592 pthread_once(pthread_once_t *once_control, void (*routine)(void))
593 {
594 
595 	if (once_control->pto_done == 0) {
596 		pthread_mutex_lock(&once_control->pto_mutex);
597 		pthread_cleanup_push(&once_cleanup, &once_control->pto_mutex);
598 		if (once_control->pto_done == 0) {
599 			routine();
600 			once_control->pto_done = 1;
601 		}
602 		pthread_cleanup_pop(1);
603 	}
604 
605 	return 0;
606 }
607 
608 void
609 pthread__mutex_deferwake(pthread_t self, pthread_mutex_t *ptm)
610 {
611 
612 	if (__predict_false(ptm == NULL ||
613 	    MUTEX_OWNER(ptm->ptm_owner) != (uintptr_t)self)) {
614 	    	(void)_lwp_unpark_all(self->pt_waiters, self->pt_nwaiters,
615 	    	    __UNVOLATILE(&ptm->ptm_waiters));
616 	    	self->pt_nwaiters = 0;
617 	} else {
618 		atomic_or_ulong((volatile unsigned long *)
619 		    (uintptr_t)&ptm->ptm_owner,
620 		    (unsigned long)MUTEX_DEFERRED_BIT);
621 	}
622 }
623 
624 int
625 _pthread_mutex_held_np(pthread_mutex_t *ptm)
626 {
627 
628 	return MUTEX_OWNER(ptm->ptm_owner) == (uintptr_t)pthread__self();
629 }
630 
631 pthread_t
632 _pthread_mutex_owner_np(pthread_mutex_t *ptm)
633 {
634 
635 	return (pthread_t)MUTEX_OWNER(ptm->ptm_owner);
636 }
637