1 '\" te
2 .\" Copyright (c) 2007, Sun Microsystems, Inc. All Rights Reserved.
3 .\" Portions Copyright (c) 2001, the Institute of Electrical and Electronics Engineers, Inc. and The Open Group. All Rights Reserved.
4 .\" Portions Copyright (c) 1995 IEEE All Rights Reserved.
5 .\" Sun Microsystems, Inc. gratefully acknowledges The Open Group for permission to reproduce portions of its copyrighted documentation. Original documentation from The Open Group can be obtained online at
6 .\" http://www.opengroup.org/bookstore/.
7 .\" The Institute of Electrical and Electronics Engineers and The Open Group, have given us permission to reprint portions of their documentation. In the following statement, the phrase "this text" refers to portions of the system documentation. Portions of this text are reprinted and reproduced in electronic form in the Sun OS Reference Manual, from IEEE Std 1003.1, 2004 Edition, Standard for Information Technology -- Portable Operating System Interface (POSIX), The Open Group Base Specifications Issue 6, Copyright (C) 2001-2004 by the Institute of Electrical and Electronics Engineers, Inc and The Open Group. In the event of any discrepancy between these versions and the original IEEE and The Open Group Standard, the original IEEE and The Open Group Standard is the referee document. The original Standard can be obtained online at http://www.opengroup.org/unix/online.html.
8 .\" This notice shall appear on any product containing this material.
9 .\" The contents of this file are subject to the terms of the Common Development and Distribution License (the "License"). You may not use this file except in compliance with the License.
10 .\" You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. See the License for the specific language governing permissions and limitations under the License.
11 .\" When distributing Covered Code, include this CDDL HEADER in each file and include the License file at usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your own identifying information: Portions Copyright [yyyy] [name of copyright owner]
12 .TH MUTEX_INIT 3C "Jun 5, 2007"
13 .SH NAME
14 mutex_init, mutex_lock, mutex_trylock, mutex_unlock, mutex_consistent,
15 mutex_destroy \- mutual exclusion locks
16 .SH SYNOPSIS
17 .LP
18 .nf
19 cc -mt [ \fIflag\fR... ] \fIfile\fR... [ \fIlibrary\fR... ]
20 #include <thread.h>
21 #include <synch.h>
22
23 \fBint\fR \fBmutex_init\fR(\fBmutex_t *\fR\fImp\fR, \fBint\fR \fItype\fR, \fBvoid *\fR \fIarg\fR);
24 .fi
25
26 .LP
27 .nf
28 \fBint\fR \fBmutex_lock\fR(\fBmutex_t *\fR\fImp\fR);
29 .fi
30
31 .LP
32 .nf
33 \fBint\fR \fBmutex_trylock\fR(\fBmutex_t *\fR\fImp\fR);
34 .fi
35
36 .LP
37 .nf
38 \fBint\fR \fBmutex_unlock\fR(\fBmutex_t *\fR\fImp\fR);
39 .fi
40
41 .LP
42 .nf
43 \fBint\fR \fBmutex_consistent\fR(\fBmutex_t *\fR\fImp\fR);
44 .fi
45
46 .LP
47 .nf
48 \fBint\fR \fBmutex_destroy\fR(\fBmutex_t *\fR\fImp\fR);
49 .fi
50
51 .SH DESCRIPTION
52 .sp
53 .LP
54 Mutual exclusion locks (mutexes) prevent multiple threads from simultaneously
55 executing critical sections of code that access shared data (that is, mutexes
56 are used to serialize the execution of threads). All mutexes must be global. A
57 successful call for a mutex lock by way of \fBmutex_lock()\fR will cause
58 another thread that is also trying to lock the same mutex to block until the
59 owner thread unlocks it by way of \fBmutex_unlock()\fR. Threads within the
60 same process or within other processes can share mutexes.
61 .sp
62 .LP
63 Mutexes can synchronize threads within the same process or in other processes.
64 Mutexes can be used to synchronize threads between processes if the mutexes are
65 allocated in writable memory and shared among the cooperating processes (see
66 \fBmmap\fR(2)), and have been initialized for this task.
67 .SS "Initialize"
68 .sp
69 .LP
70 Mutexes are either intra-process or inter-process, depending upon the argument
71 passed implicitly or explicitly to the initialization of that mutex. A
72 statically allocated mutex does not need to be explicitly initialized; by
73 default, a statically allocated mutex is initialized with all zeros and its
74 scope is set to be within the calling process.
75 .sp
76 .LP
77 For inter-process synchronization, a mutex needs to be allocated in memory
78 shared between these processes. Since the memory for such a mutex must be
79 allocated dynamically, the mutex needs to be explicitly initialized using
80 \fBmutex_init()\fR.
81 .sp
82 .LP
83 The \fBmutex_init()\fR function initializes the mutex referenced by \fImp\fR
84 with the type specified by \fItype\fR. Upon successful initialization the
85 state of the mutex becomes initialized and unlocked. Only the attribute type
86 \fBLOCK_PRIO_PROTECT\fR uses \fIarg\fR. The \fItype\fR argument must be one of
87 the following:
88 .sp
89 .ne 2
90 .na
91 \fB\fBUSYNC_THREAD\fR\fR
92 .ad
93 .sp .6
94 .RS 4n
95 The mutex can synchronize threads only in this process.
96 .RE
97
98 .sp
99 .ne 2
100 .na
101 \fB\fBUSYNC_PROCESS\fR\fR
102 .ad
103 .sp .6
104 .RS 4n
105 The mutex can synchronize threads in this process and other processes. The
106 object initialized with this attribute must be allocated in memory shared
107 between processes, either in System V shared memory (see \fBshmop\fR(2)) or in
108 memory mapped to a file (see \fBmmap\fR(2)). If the object is not allocated in
109 such shared memory, it will not be shared between processes.
110 .RE
111
112 .sp
113 .LP
114 The \fItype\fR argument can be augmented by the bitwise-inclusive-\fBOR\fR of
115 zero or more of the following flags:
116 .sp
117 .ne 2
118 .na
119 \fB\fBLOCK_ROBUST\fR\fR
120 .ad
121 .sp .6
122 .RS 4n
123 The mutex can synchronize threads robustly. At the time of thread or process
124 death, either by calling \fBthr_exit()\fR or \fBexit()\fR or due to process
125 abnormal termination, the lock is unlocked if is held by the thread or process.
126 The next owner of the mutex will acquire it with an error return of
127 \fBEOWNERDEAD\fR. The application must always check the return value from
128 \fBmutex_lock()\fR for a mutex of this type. The new owner of this mutex should
129 then attempt to make the state protected by the mutex consistent, since this
130 state could have been left inconsistent when the last owner died. If the new
131 owner is able to make the state consistent, it should call
132 \fBmutex_consistent()\fR to restore the state of the mutex and then unlock the
133 mutex. All subsequent calls to \fBmutex_lock()\fRwill then behave normally.
134 Only the new owner can make the mutex consistent. If for any reason the new
135 owner is not able to make the state consistent, it should not call
136 \fBmutex_consistent()\fR but should simply unlock the mutex. All waiting
137 processes will be awakened and all subsequent calls to \fBmutex_lock()\fR will
138 fail in acquiring the mutex with an error value of \fBENOTRECOVERABLE\fR. If
139 the thread or process that acquired the lock with \fBEOWNERDEAD\fR terminates
140 without unlocking the mutex, the next owner will acquire the lock with an error
141 value of \fBEOWNERDEAD\fR.
142 .sp
143 The memory for the object to be initialized with this attribute must be zeroed
144 before initialization. Any thread or process interested in the robust lock can
145 call \fBmutex_init()\fR to potentially initialize it, provided that all such
146 callers of \fBmutex_init()\fR specify the same set of attribute flags. In this
147 situation, if \fBmutex_init()\fR is called on a previously initialized robust
148 mutex, \fBmutex_init()\fR will not reinitialize the mutex and will return the
149 error value \fBEBUSY\fR.
150 .RE
151
152 .sp
153 .ne 2
154 .na
155 \fB\fBLOCK_RECURSIVE\fR\fR
156 .ad
157 .sp .6
158 .RS 4n
159 A thread attempting to relock this mutex without first unlocking it will
160 succeed in locking the mutex. The mutex must be unlocked as many times as it is
161 locked.
162 .RE
163
164 .sp
165 .ne 2
166 .na
167 \fB\fBLOCK_ERRORCHECK\fR\fR
168 .ad
169 .sp .6
170 .RS 4n
171 Unless \fBLOCK_RECURSIVE\fR is also set, a thread attempting to relock this
172 mutex without first unlocking it will return with an error rather than
173 deadlocking itself. A thread attempting to unlock this mutex without first
174 owning it will return with an error.
175 .RE
176
177 .sp
178 .ne 2
179 .na
180 \fB\fBLOCK_PRIO_INHERIT\fR\fR
181 .ad
182 .sp .6
183 .RS 4n
184 When a thread is blocking higher priority threads because of owning one or more
185 mutexes with the \fBLOCK_PRIO_INHERIT\fR attribute, it executes at the higher
186 of its priority or the priority of the highest priority thread waiting on any
187 of the mutexes owned by this thread and initialized with this attribute.
188 .RE
189
190 .sp
191 .ne 2
192 .na
193 \fB\fBLOCK_PRIO_PROTECT\fR\fR
194 .ad
195 .sp .6
196 .RS 4n
197 When a thread owns one or more mutexes initialized with the
198 \fBLOCK_PRIO_PROTECT\fR attribute, it executes at the higher of its priority or
199 the highest of the priority ceilings of all the mutexes owned by this thread
200 and initialized with this attribute, regardless of whether other threads are
201 blocked on any of these mutexes. When this attribute is specified, \fIarg\fR
202 must point to an \fBint\fR containing the priority ceiling.
203 .RE
204
205 .sp
206 .LP
207 See \fBpthread_mutexattr_getrobust\fR(3C) for more information about robust
208 mutexes. The \fBLOCK_ROBUST\fR attribute is the same as the POSIX
209 \fBPTHREAD_MUTEX_ROBUST\fR attribute.
210 .sp
211 .LP
212 See \fBpthread_mutexattr_settype\fR(3C) for more information on recursive and
213 error checking mutex types. The combination (\fBLOCK_RECURSIVE\fR |
214 \fBLOCK_ERRORCHECK\fR) is the same as the POSIX \fBPTHREAD_MUTEX_RECURSIVE\fR
215 type. By itself, \fBLOCK_ERRORCHECK\fR is the same as the \fBPOSIX
216 PTHREAD_MUTEX_ERRORCHECK\fR type.
217 .sp
218 .LP
219 The \fBLOCK_PRIO_INHERIT\fR attribute is the same as the POSIX
220 \fBPTHREAD_PRIO_INHERIT\fR attribute. The \fBLOCK_PRIO_PROTECT\fR attribute is
221 the same as the POSIX \fBPTHREAD_PRIO_PROTECT\fR attribute. See
222 \fBpthread_mutexattr_getprotocol\fR(3C),
223 \fBpthread_mutexattr_getprioceiling\fR(3C), and
224 \fBpthread_mutex_getprioceiling\fR(3C) for a full discussion. The
225 \fBLOCK_PRIO_INHERIT\fR and \fBLOCK_PRIO_PROTECT\fR attributes are mutually
226 exclusive. Specifying both of these attributes causes \fBmutex_init()\fR to
227 fail with \fBEINVAL\fR.
228 .sp
229 .LP
230 Initializing mutexes can also be accomplished by allocating in zeroed memory
231 (default), in which case a \fItype\fR of \fBUSYNC_THREAD\fR is assumed. In
232 general, the following rules apply to mutex initialization:
233 .RS +4
234 .TP
235 .ie t \(bu
236 .el o
237 The same mutex must not be simultaneously initialized by multiple threads.
238 .RE
239 .RS +4
240 .TP
241 .ie t \(bu
242 .el o
243 A mutex lock must not be reinitialized while in use by other threads.
244 .RE
245 .sp
246 .LP
247 These rules do not apply to \fBLOCK_ROBUST\fR mutexes. See the description for
248 \fBLOCK_ROBUST\fR above. If default mutex attributes are used, the macro
249 \fBDEFAULTMUTEX\fR can be used to initialize mutexes that are statically
250 allocated.
251 .sp
252 .LP
253 Default mutex initialization (intra-process):
254 .sp
255 .in +2
256 .nf
257 mutex_t mp;
258 mutex_init(&mp, USYNC_THREAD, NULL);
259 .fi
260 .in -2
261
262 .sp
263 .LP
264 or
265 .sp
266 .in +2
267 .nf
268 mutex_t mp = DEFAULTMUTEX;
269 .fi
270 .in -2
271
272 .sp
273 .LP
274 Customized mutex initialization (inter-process):
275 .sp
276 .in +2
277 .nf
278 mutex_init(&mp, USYNC_PROCESS, NULL);
279 .fi
280 .in -2
281
282 .sp
283 .LP
284 Customized mutex initialization (inter-process robust):
285 .sp
286 .in +2
287 .nf
288 mutex_init(&mp, USYNC_PROCESS | LOCK_ROBUST, NULL);
289 .fi
290 .in -2
291
292 .sp
293 .LP
294 Statically allocated mutexes can also be initialized with macros specifying
295 \fBLOCK_RECURSIVE\fR and/or \fBLOCK_ERRORCHECK\fR:
296 .sp
297 .ne 2
298 .na
299 \fB\fBmutex_t mp = RECURSIVEMUTEX;\fR\fR
300 .ad
301 .sp .6
302 .RS 4n
303 Same as (\fBUSYNC_THREAD\fR | \fBLOCK_RECURSIVE\fR)
304 .RE
305
306 .sp
307 .ne 2
308 .na
309 \fB\fBmutex_t mp = ERRORCHECKMUTEX;\fR\fR
310 .ad
311 .sp .6
312 .RS 4n
313 Same as (\fBUSYNC_THREAD\fR | \fBLOCK_ERRORCHECK\fR)
314 .RE
315
316 .sp
317 .ne 2
318 .na
319 \fB\fBmutex_t mp = RECURSIVE_ERRORCHECKMUTEX;\fR\fR
320 .ad
321 .sp .6
322 .RS 4n
323 Same as (\fBUSYNC_THREAD\fR | \fBLOCK_RECURSIVE\fR | \fBLOCK_ERRORCHECK\fR)
324 .RE
325
326 .SS "Lock and Unlock"
327 .sp
328 .LP
329 A critical section of code is enclosed by a the call to lock the mutex and the
330 call to unlock the mutex to protect it from simultaneous access by multiple
331 threads. Only one thread at a time may possess mutually exclusive access to
332 the critical section of code that is enclosed by the mutex-locking call and the
333 mutex-unlocking call, whether the mutex's scope is intra-process or
334 inter-process. A thread calling to lock the mutex either gets exclusive access
335 to the code starting from the successful locking until its call to unlock the
336 mutex, or it waits until the mutex is unlocked by the thread that locked it.
337 .sp
338 .LP
339 Mutexes have ownership, unlike semaphores. Although any thread, within the
340 scope of a mutex, can get an unlocked mutex and lock access to the same
341 critical section of code, only the thread that locked a mutex should unlock it.
342 .sp
343 .LP
344 If a thread waiting for a mutex receives a signal, upon return from the signal
345 handler, the thread resumes waiting for the mutex as if there was no interrupt.
346 A mutex protects code, not data; therefore, strongly bind a mutex with the
347 data by putting both within the same structure, or at least within the same
348 procedure.
349 .sp
350 .LP
351 A call to \fBmutex_lock()\fR locks the mutex object referenced by \fImp\fR. If
352 the mutex is already locked, the calling thread blocks until the mutex is
353 freed; this will return with the mutex object referenced by \fImp\fR in the
354 locked state with the calling thread as its owner. If the current owner of a
355 mutex tries to relock the mutex, it will result in deadlock.
356 .sp
357 .LP
358 The \fBmutex_trylock()\fR function is the same as \fBmutex_lock()\fR,
359 respectively, except that if the mutex object referenced by \fImp\fR is locked
360 (by any thread, including the current thread), the call returns immediately
361 with an error.
362 .sp
363 .LP
364 The \fBmutex_unlock()\fR function are called by the owner of the mutex object
365 referenced by \fImp\fR to release it. The mutex must be locked and the calling
366 thread must be the one that last locked the mutex (the owner). If there are
367 threads blocked on the mutex object referenced by \fImp\fR when
368 \fBmutex_unlock()\fR is called, the \fImp\fR is freed, and the scheduling
369 policy will determine which thread gets the mutex. If the calling thread is not
370 the owner of the lock, no error status is returned, and the behavior of the
371 program is undefined.
372 .SS "Destroy"
373 .sp
374 .LP
375 The \fBmutex_destroy()\fR function destroys the mutex object referenced by
376 \fImp\fR. The mutex object becomes uninitialized. The space used by the
377 destroyed mutex variable is not freed. It needs to be explicitly reclaimed.
378 .SH RETURN VALUES
379 .sp
380 .LP
381 If successful, these functions return \fB0\fR. Otherwise, an error number is
382 returned.
383 .SH ERRORS
384 .sp
385 .LP
386 The \fBmutex_init()\fR function will fail if:
387 .sp
388 .ne 2
389 .na
390 \fB\fBEINVAL\fR\fR
391 .ad
392 .RS 10n
393 The value specified by \fItype\fR is invalid, or the \fBLOCK_PRIO_INHERIT\fR
394 and \fBLOCK_PRIO_PROTECT\fR attributes are both specified.
395 .RE
396
397 .sp
398 .LP
399 The \fBmutex_init()\fR function will fail for \fBLOCK_ROBUST\fR type mutex if:
400 .sp
401 .ne 2
402 .na
403 \fB\fBEBUSY\fR\fR
404 .ad
405 .RS 10n
406 The mutex pointed to by \fImp\fR was previously initialized and has not yet
407 been destroyed.
408 .RE
409
410 .sp
411 .ne 2
412 .na
413 \fB\fBEINVAL\fR\fR
414 .ad
415 .RS 10n
416 The mutex pointed to by \fImp\fR was previously initialized with a different
417 set of attribute flags.
418 .RE
419
420 .sp
421 .LP
422 The \fBmutex_trylock()\fR function will fail if:
423 .sp
424 .ne 2
425 .na
426 \fB\fBEBUSY\fR\fR
427 .ad
428 .RS 9n
429 The mutex pointed to by \fImp\fR is already locked.
430 .RE
431
432 .sp
433 .LP
434 The \fBmutex_lock()\fR and \fBmutex_trylock()\fR functions will fail for a
435 \fBLOCK_RECURSIVE\fR mutex if:
436 .sp
437 .ne 2
438 .na
439 \fB\fBEAGAIN\fR\fR
440 .ad
441 .RS 10n
442 The mutex could not be acquired because the maximum number of recursive locks
443 for the mutex has been reached.
444 .RE
445
446 .sp
447 .LP
448 The \fBmutex_lock()\fR function will fail for a \fBLOCK_ERRORCHECK\fR and
449 non-\fBLOCK_RECURSIVE\fR mutex if:
450 .sp
451 .ne 2
452 .na
453 \fB\fBEDEADLK\fR\fR
454 .ad
455 .RS 11n
456 The caller already owns the mutex.
457 .RE
458
459 .sp
460 .LP
461 The \fBmutex_lock()\fR function may fail for a non-\fBLOCK_ERRORCHECK\fR and
462 non-\fBLOCK_RECURSIVE\fR mutex if:
463 .sp
464 .ne 2
465 .na
466 \fB\fBEDEADLK\fR\fR
467 .ad
468 .RS 11n
469 The caller already owns the mutex.
470 .RE
471
472 .sp
473 .LP
474 The \fBmutex_unlock()\fR function will fail for a \fBLOCK_ERRORCHECK\fR mutex
475 if:
476 .sp
477 .ne 2
478 .na
479 \fB\fBEPERM\fR\fR
480 .ad
481 .RS 9n
482 The caller does not own the mutex.
483 .RE
484
485 .sp
486 .LP
487 The \fBmutex_lock()\fR or \fBmutex_trylock()\fR functions will fail for
488 \fBLOCK_ROBUST\fR type mutex if:
489 .sp
490 .ne 2
491 .na
492 \fB\fBEOWNERDEAD\fR\fR
493 .ad
494 .RS 19n
495 The last owner of this mutex died while holding the mutex. This mutex is now
496 owned by the caller. The caller must now attempt to make the state protected by
497 the mutex consistent. If it is able to clean up the state, then it should
498 restore the state of the mutex by calling \fBmutex_consistent()\fR and unlock
499 the mutex. Subsequent calls to \fBmutex_lock()\fR will behave normally, as
500 before. If the caller is not able to clean up the state,
501 \fBmutex_consistent()\fR should not be called but the mutex should be unlocked.
502 Subsequent calls to \fBmutex_lock()\fR will fail to acquire the mutex,
503 returning with the error value \fBENOTRECOVERABLE\fR. If the owner who acquired
504 the lock with \fBEOWNERDEAD\fR dies, the next owner will acquire the lock with
505 \fBEOWNERDEAD\fR.
506 .RE
507
508 .sp
509 .ne 2
510 .na
511 \fB\fBENOTRECOVERABLE\fR\fR
512 .ad
513 .RS 19n
514 The mutex trying to be acquired was protecting the state that has been left
515 unrecoverable when the mutex's last owner could not make the state protected by
516 the mutex consistent. The mutex has not been acquired. This condition occurs
517 when the lock was previously acquired with \fBEOWNERDEAD\fR and the owner was
518 not able to clean up the state and unlocked the mutex without calling
519 \fBmutex_consistent()\fR.
520 .RE
521
522 .sp
523 .LP
524 The \fBmutex_consistent()\fR function will fail if:
525 .sp
526 .ne 2
527 .na
528 \fB\fBEINVAL\fR\fR
529 .ad
530 .RS 10n
531 The caller does not own the mutex or the mutex is not a \fBLOCK_ROBUST\fR mutex
532 having an inconsistent state (\fBEOWNERDEAD\fR).
533 .RE
534
535 .SH EXAMPLES
536 .SS "Single Gate"
537 .sp
538 .LP
539 The following example uses one global mutex as a gate-keeper to permit each
540 thread exclusive sequential access to the code within the user-defined
541 function "change_global_data." This type of synchronization will protect the
542 state of shared data, but it also prohibits parallelism.
543 .sp
544 .in +2
545 .nf
546 /* cc thisfile.c -lthread */
547 #define _REENTRANT
548 #include <stdio.h>
549 #include <thread.h>
550 #define NUM_THREADS 12
551 void *change_global_data(void *); /* for thr_create() */
552 main(int argc,char * argv[]) {
553 int i=0;
554 for (i=0; i< NUM_THREADS; i++) {
555 thr_create(NULL, 0, change_global_data, NULL, 0, NULL);
556 }
557 while ((thr_join(NULL, NULL, NULL) == 0));
558 }
559
560 void * change_global_data(void *null){
561 static mutex_t Global_mutex;
562 static int Global_data = 0;
563 mutex_lock(&Global_mutex);
564 Global_data++;
565 sleep(1);
566 printf("%d is global data\en",Global_data);
567 mutex_unlock(&Global_mutex);
568 return NULL;
569 }
570 .fi
571 .in -2
572
573 .SS "Multiple Instruction Single Data"
574 .sp
575 .LP
576 The previous example, the mutex, the code it owns, and the data it protects was
577 enclosed in one function. The next example uses C++ features to accommodate
578 many functions that use just one mutex to protect one data:
579 .sp
580 .in +2
581 .nf
582 /* CC thisfile.c -lthread use C++ to compile*/
583
584 #define _REENTRANT
585 #include <stdlib.h>
586 #include <stdio.h>
587 #include <thread.h>
588 #include <errno.h>
589 #include <iostream.h>
590 #define NUM_THREADS 16
591 void *change_global_data(void *); /* for thr_create() */
592
593 class Mutected {
594 private:
595 static mutex_t Global_mutex;
596 static int Global_data;
597 public:
598 static int add_to_global_data(void);
599 static int subtract_from_global_data(void);
600 };
601
602 int Mutected::Global_data = 0;
603 mutex_t Mutected::Global_mutex;
604
605 int Mutected::add_to_global_data() {
606 mutex_lock(&Global_mutex);
607 Global_data++;
608 mutex_unlock(&Global_mutex);
609 return Global_data;
610 }
611
612 int Mutected::subtract_from_global_data() {
613 mutex_lock(&Global_mutex);
614 Global_data--;
615 mutex_unlock(&Global_mutex);
616 return Global_data;
617 }
618
619 void
620 main(int argc,char * argv[]) {
621 int i=0;
622 for (i=0;i< NUM_THREADS;i++) {
623 thr_create(NULL,0,change_global_data,NULL,0,NULL);
624 }
625 while ((thr_join(NULL,NULL,NULL) == 0));
626 }
627
628 void * change_global_data(void *) {
629 static int switcher = 0;
630 if ((switcher++ % 3) == 0) /* one-in-three threads subtracts */
631 cout << Mutected::subtract_from_global_data() << endl;
632 else
633 cout << Mutected::add_to_global_data() << endl;
634 return NULL;
635 }
636 .fi
637 .in -2
638
639 .SS "Interprocess Locking"
640 .sp
641 .LP
642 A mutex can protect data that is shared among processes. The mutex would need
643 to be initialized as \fBUSYNC_PROCESS\fR. One process initializes the
644 process-shared mutex and writes it to a file to be mapped into memory by all
645 cooperating processes (see \fBmmap\fR(2)). Afterwards, other independent
646 processes can run the same program (whether concurrently or not) and share
647 mutex-protected data.
648 .sp
649 .in +2
650 .nf
651 /* cc thisfile.c -lthread */
652 /* To execute, run the command line "a.out 0 &; a.out 1" */
653
654 #define _REENTRANT
655 #include <sys/types.h>
656 #include <sys/mman.h>
657 #include <sys/stat.h>
658 #include <fcntl.h>
659 #include <stdio.h>
660 #include <thread.h>
661 #define INTERPROCESS_FILE "ipc-sharedfile"
662 #define NUM_ADDTHREADS 12
663 #define NUM_SUBTRACTTHREADS 10
664 #define INCREMENT '0'
665 #define DECREMENT '1'
666 typedef struct {
667 mutex_t Interprocess_mutex;
668 int Interprocess_data;
669 } buffer_t;
670 buffer_t *buffer;
671
672 void *add_interprocess_data(), *subtract_interprocess_data();
673 void create_shared_memory(), test_argv();
674 int zeroed[sizeof(buffer_t)];
675 int ipc_fd, i=0;
676
677 void
678 main(int argc,char * argv[]){
679 test_argv(argv[1]);
680
681 switch (*argv[1]) {
682 case INCREMENT:
683 /* Initializes the process-shared mutex */
684 /* Should be run prior to running a DECREMENT process */
685 create_shared_memory();
686 ipc_fd = open(INTERPROCESS_FILE, O_RDWR);
687 buffer = (buffer_t *)mmap(NULL, sizeof(buffer_t),
688 PROT_READ | PROT_WRITE, MAP_SHARED, ipc_fd, 0);
689 buffer->Interprocess_data = 0;
690 mutex_init(&buffer->Interprocess_mutex, USYNC_PROCESS,0);
691 for (i=0; i< NUM_ADDTHREADS; i++)
692 thr_create(NULL, 0, add_interprocess_data, argv[1],
693 0, NULL);
694 break;
695
696 case DECREMENT:
697 /* Should be run after the INCREMENT process has run. */
698 while(ipc_fd = open(INTERPROCESS_FILE, O_RDWR)) == -1)
699 sleep(1);
700 buffer = (buffer_t *)mmap(NULL, sizeof(buffer_t),
701 PROT_READ | PROT_WRITE, MAP_SHARED, ipc_fd, 0);
702 for (i=0; i< NUM_SUBTRACTTHREADS; i++)
703 thr_create(NULL, 0, subtract_interprocess_data, argv[1],
704 0, NULL);
705 break;
706 } /* end switch */
707
708 while ((thr_join(NULL,NULL,NULL) == 0));
709 } /* end main */
710
711 void *add_interprocess_data(char argv_1[]){
712 mutex_lock(&buffer->Interprocess_mutex);
713 buffer->Interprocess_data++;
714 sleep(2);
715 printf("%d is add-interprocess data, and %c is argv1\en",
716 buffer->Interprocess_data, argv_1[0]);
717 mutex_unlock(&buffer->Interprocess_mutex);
718 return NULL;
719 }
720
721 void *subtract_interprocess_data(char argv_1[]) {
722 mutex_lock(&buffer->Interprocess_mutex);
723 buffer->Interprocess_data--;
724 sleep(2);
725 printf("%d is subtract-interprocess data, and %c is argv1\en",
726 buffer->Interprocess_data, argv_1[0]);
727 mutex_unlock(&buffer->Interprocess_mutex);
728 return NULL;
729 }
730
731 void create_shared_memory(){
732 int i;
733 ipc_fd = creat(INTERPROCESS_FILE, O_CREAT | O_RDWR );
734 for (i=0; i<sizeof(buffer_t); i++){
735 zeroed[i] = 0;
736 write(ipc_fd, &zeroed[i],2);
737 }
738 close(ipc_fd);
739 chmod(INTERPROCESS_FILE, S_IRWXU | S_IRWXG | S_IRWXO);
740 }
741
742 void test_argv(char argv1[]) {
743 if (argv1 == NULL) {
744 printf("use 0 as arg1 for initial process\en \e
745 or use 1 as arg1 for the second process\en");
746 exit(NULL);
747 }
748 }
749 .fi
750 .in -2
751
752 .SS "Solaris Interprocess Robust Locking"
753 .sp
754 .LP
755 A mutex can protect data that is shared among processes robustly. The mutex
756 would need to be initialized as \fBUSYNC_PROCESS\fR | \fBLOCK_ROBUST\fR. One
757 process initializes the robust process-shared mutex and writes it to a file to
758 be mapped into memory by all cooperating processes (see \fBmmap\fR(2)).
759 Afterwards, other independent processes can run the same program (whether
760 concurrently or not) and share mutex-protected data.
761 .sp
762 .LP
763 The following example shows how to use a \fBUSYNC_PROCESS\fR |
764 \fBLOCK_ROBUST\fR type mutex.
765 .sp
766 .in +2
767 .nf
768 /* cc thisfile.c -lthread */
769 /* To execute, run the command line "a.out & a.out 1" */
770 #include <sys/types.h>
771 #include <sys/mman.h>
772 #include <fcntl.h>
773 #include <stdio.h>
774 #include <thread.h>
775 #define INTERPROCESS_FILE "ipc-sharedfile"
776 typedef struct {
777 mutex_t Interprocess_mutex;
778 int Interprocess_data;
779 } buffer_t;
780 buffer_t *buffer;
781 int make_date_consistent();
782 void create_shared_memory();
783 int zeroed[sizeof(buffer_t)];
784 int ipc_fd, i=0;
785 main(int argc,char * argv[]) {
786 int rc;
787 if (argc > 1) {
788 while((ipc_fd = open(INTERPROCESS_FILE, O_RDWR)) == -1)
789 sleep(1);
790 buffer = (buffer_t *)mmap(NULL, sizeof(buffer_t),
791 PROT_READ | PROT_WRITE, MAP_SHARED, ipc_fd, 0);
792 mutex_init(&buffer->Interprocess_mutex,
793 USYNC_PROCESS | LOCK_ROBUST,0);
794 } else {
795 create_shared_memory();
796 ipc_fd = open(INTERPROCESS_FILE, O_RDWR);
797 buffer = (buffer_t *)mmap(NULL, sizeof(buffer_t),
798 PROT_READ | PROT_WRITE, MAP_SHARED, ipc_fd, 0);
799 buffer->Interprocess_data = 0;
800 mutex_init(&buffer->Interprocess_mutex,
801 USYNC_PROCESS | LOCK_ROBUST,0);
802 }
803 for(;;) {
804 rc = mutex_lock(&buffer->Interprocess_mutex);
805 switch (rc) {
806 case EOWNERDEAD:
807 /*
808 * The lock is acquired.
809 * The last owner died holding the lock.
810 * Try to make the state associated with
811 * the mutex consistent.
812 * If successful, make the robust lock consistent.
813 */
814 if (make_data_consistent())
815 mutex_consistent(&buffer->Interprocess_mutex);
816 mutex_unlock(&buffer->Interprocess_mutex);
817 break;
818 case ENOTRECOVERABLE:
819 /*
820 * The lock is not acquired.
821 * The last owner got the mutex with EOWNERDEAD
822 * and failed to make the data consistent.
823 * There is no way to recover, so just exit.
824 */
825 exit(1);
826 case 0:
827 /*
828 * There is no error - data is consistent.
829 * Do something with data.
830 */
831 mutex_unlock(&buffer->Interprocess_mutex);
832 break;
833 }
834 }
835 } /* end main */
836 void create_shared_memory() {
837 int i;
838 ipc_fd = creat(INTERPROCESS_FILE, O_CREAT | O_RDWR );
839 for (i=0; i<sizeof(buffer_t); i++) {
840 zeroed[i] = 0;
841 write(ipc_fd, &zeroed[i],2);
842 }
843 close(ipc_fd);
844 chmod(INTERPROCESS_FILE, S_IRWXU | S_IRWXG | S_IRWXO);
845 }
846
847 /* return 1 if able to make data consistent, otherwise 0. */
848 int make_data_consistent () {
849 buffer->Interprocess_data = 0;
850 return (1);
851 }
852 .fi
853 .in -2
854
855 .SS "Dynamically Allocated Mutexes"
856 .sp
857 .LP
858 The following example allocates and frees memory in which a mutex is embedded.
859 .sp
860 .in +2
861 .nf
862 struct record {
863 int field1;
864 int field2;
865 mutex_t m;
866 } *r;
867 r = malloc(sizeof(struct record));
868 mutex_init(&r->m, USYNC_THREAD, NULL);
869 /*
870 * The fields in this record are accessed concurrently
871 * by acquiring the embedded lock.
872 */
873 .fi
874 .in -2
875
876 .sp
877 .LP
878 The thread execution in this example is as follows:
879 .sp
880 .in +2
881 .nf
882 \fIThread 1 executes:\fR \fIThread 2 executes:\fR
883
884 \&... ...
885 mutex_lock(&r->m); mutex_lock(&r->m);
886 r->field1++; localvar = r->field1;
887 mutex_unlock(&r->m); mutex_unlock(&r->m);
888 \&... ...
889 .fi
890 .in -2
891
892 .sp
893 .LP
894 Later, when a thread decides to free the memory pointed to by \fIr\fR, the
895 thread should call \fBmutex_destroy\fR(\|) on the mutexes in this memory.
896 .sp
897 .LP
898 In the following example, the main thread can do a \fBthr_join\fR(\|) on both
899 of the above threads. If there are no other threads using the memory in
900 \fIr\fR, the main thread can now safely free \fIr\fR:
901 .sp
902 .in +2
903 .nf
904 for (i = 0; i < 2; i++)
905 thr_join(0, 0, 0);
906 mutex_destroy(&r->m); /* first destroy mutex */
907 free(r); /* then free memory */
908 .fi
909 .in -2
910
911 .sp
912 .LP
913 If the mutex is not destroyed, the program could have memory leaks.
914 .SH ATTRIBUTES
915 .sp
916 .LP
917 See \fBattributes\fR(5) for descriptions of the following attributes:
918 .sp
919
920 .sp
921 .TS
922 box;
923 c | c
924 l | l .
925 ATTRIBUTE TYPE ATTRIBUTE VALUE
926 _
927 Interface Stability Stable
928 _
929 MT-Level MT-Safe
930 .TE
931
932 .SH SEE ALSO
933 .sp
934 .LP
935 \fBmmap\fR(2), \fBshmop\fR(2), \fBpthread_mutexattr_getprioceiling\fR(3C),
936 \fBpthread_mutexattr_getprotocol\fR(3C), \fBpthread_mutexattr_getrobust\fR(3C),
937 \fBpthread_mutexattr_gettype\fR(3C), \fBpthread_mutex_getprioceiling\fR(3C),
938 \fBpthread_mutex_init\fR(3C), \fBattributes\fR(5), \fBmutex\fR(5),
939 \fBstandards\fR(5)
940 .SH NOTES
941 .sp
942 .LP
943 Previous releases of Solaris provided the \fBUSYNC_PROCESS_ROBUST\fR mutex
944 type. This type is now deprecated but is still supported for source and binary
945 compatibility. When passed to \fBmutex_init()\fR, it is transformed into
946 (\fBUSYNC_PROCESS\fR | \fBLOCK_ROBUST\fR). The former method for restoring a
947 \fBUSYNC_PROCESS_ROBUST\fR mutex to a consistent state was to reinitialize it
948 by calling \fBmutex_init()\fR. This method is still supported for source and
949 binary compatibility, but the proper method is to call
950 \fBmutex_consistent()\fR.
951 .sp
952 .LP
953 The \fBUSYNC_PROCESS_ROBUST\fR type permitted an alternate error value,
954 \fBELOCKUNMAPPED\fR, to be returned by \fBmutex_lock()\fR if the process
955 containing a locked robust mutex unmapped the memory containing the mutex or
956 performed one of the \fBexec\fR(2) functions. The \fBELOCKUNMAPPED\fR error
957 value implies all of the consequences of the \fBEOWNERDEAD\fR error value and
958 as such is just a synonym for \fBEOWNERDEAD\fR. For full source and binary
959 compatibility, the \fBELOCKUNMAPPED\fR error value is still returned from
960 \fBmutex_lock()\fR in these circumstances, but only if the mutex was
961 initialized with the \fBUSYNC_PROCESS_ROBUST\fR type. Otherwise,
962 \fBEOWNERDEAD\fR is returned in these circumstances.
963 .sp
964 .LP
965 The \fBmutex_lock()\fR, \fBmutex_unlock()\fR, and \fBmutex_trylock()\fR
966 functions do not validate the mutex type. An uninitialized mutex or a mutex
967 with an invalid type does not return \fBEINVAL\fR. Interfaces for mutexes with
968 an invalid type have unspecified behavior.
969 .sp
970 .LP
971 Uninitialized mutexes that are allocated locally could contain junk data. Such
972 mutexes need to be initialized using \fBmutex_init()\fR.
973 .sp
974 .LP
975 By default, if multiple threads are waiting for a mutex, the order of
976 acquisition is undefined.