Source

mino-pypy / pypy / translator / c / src / thread_pthread.c

Full commit
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498

/* Posix threads interface (from CPython) */

#include <unistd.h>   /* for the _POSIX_xxx and _POSIX_THREAD_xxx defines */
#include <stdlib.h>
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <errno.h>
#include <assert.h>

/* The following is hopefully equivalent to what CPython does
   (which is trying to compile a snippet of code using it) */
#ifdef PTHREAD_SCOPE_SYSTEM
#  ifndef PTHREAD_SYSTEM_SCHED_SUPPORTED
#    define PTHREAD_SYSTEM_SCHED_SUPPORTED
#  endif
#endif

#if !defined(pthread_attr_default)
#  define pthread_attr_default ((pthread_attr_t *)NULL)
#endif
#if !defined(pthread_mutexattr_default)
#  define pthread_mutexattr_default ((pthread_mutexattr_t *)NULL)
#endif
#if !defined(pthread_condattr_default)
#  define pthread_condattr_default ((pthread_condattr_t *)NULL)
#endif

#define CHECK_STATUS(name)  if (status != 0) { perror(name); error = 1; }

/* The POSIX spec requires that use of pthread_attr_setstacksize
   be conditional on _POSIX_THREAD_ATTR_STACKSIZE being defined. */
#ifdef _POSIX_THREAD_ATTR_STACKSIZE
# ifndef THREAD_STACK_SIZE
#  define THREAD_STACK_SIZE   0   /* use default stack size */
# endif
/* for safety, ensure a viable minimum stacksize */
# define THREAD_STACK_MIN    0x8000  /* 32kB */
#else  /* !_POSIX_THREAD_ATTR_STACKSIZE */
# ifdef THREAD_STACK_SIZE
#  error "THREAD_STACK_SIZE defined but _POSIX_THREAD_ATTR_STACKSIZE undefined"
# endif
#endif

/* XXX This implementation is considered (to quote Tim Peters) "inherently
   hosed" because:
     - It does not guarantee the promise that a non-zero integer is returned.
     - The cast to long is inherently unsafe.
     - It is not clear that the 'volatile' (for AIX?) and ugly casting in the
       latter return statement (for Alpha OSF/1) are any longer necessary.
*/
long RPyThreadGetIdent(void)
{
	volatile pthread_t threadid;
	/* Jump through some hoops for Alpha OSF/1 */
	threadid = pthread_self();

#ifdef __CYGWIN__
	/* typedef __uint32_t pthread_t; */
	return (long) threadid;
#else
	if (sizeof(pthread_t) <= sizeof(long))
		return (long) threadid;
	else
		return (long) *(long *) &threadid;
#endif
}

static long _pypythread_stacksize = 0;

static void *bootstrap_pthread(void *func)
{
  ((void(*)(void))func)();
  return NULL;
}

long RPyThreadStart(void (*func)(void))
{
	pthread_t th;
	int status;
#if defined(THREAD_STACK_SIZE) || defined(PTHREAD_SYSTEM_SCHED_SUPPORTED)
	pthread_attr_t attrs;
#endif
#if defined(THREAD_STACK_SIZE)
	size_t tss;
#endif

#if defined(THREAD_STACK_SIZE) || defined(PTHREAD_SYSTEM_SCHED_SUPPORTED)
	pthread_attr_init(&attrs);
#endif
#ifdef THREAD_STACK_SIZE
	tss = (_pypythread_stacksize != 0) ? _pypythread_stacksize
		: THREAD_STACK_SIZE;
	if (tss != 0)
		pthread_attr_setstacksize(&attrs, tss);
#endif
#if defined(PTHREAD_SYSTEM_SCHED_SUPPORTED) && !defined(__FreeBSD__)
        pthread_attr_setscope(&attrs, PTHREAD_SCOPE_SYSTEM);
#endif

	status = pthread_create(&th, 
#if defined(THREAD_STACK_SIZE) || defined(PTHREAD_SYSTEM_SCHED_SUPPORTED)
				 &attrs,
#else
				 (pthread_attr_t*)NULL,
#endif
				 bootstrap_pthread,
				 (void *)func
				 );

#if defined(THREAD_STACK_SIZE) || defined(PTHREAD_SYSTEM_SCHED_SUPPORTED)
	pthread_attr_destroy(&attrs);
#endif
	if (status != 0)
            return -1;

        pthread_detach(th);

#ifdef __CYGWIN__
	/* typedef __uint32_t pthread_t; */
	return (long) th;
#else
	if (sizeof(pthread_t) <= sizeof(long))
		return (long) th;
	else
		return (long) *(long *) &th;
#endif
}

long RPyThreadGetStackSize(void)
{
	return _pypythread_stacksize;
}

long RPyThreadSetStackSize(long newsize)
{
#if defined(THREAD_STACK_SIZE)
	pthread_attr_t attrs;
	size_t tss_min;
	int rc;
#endif

	if (newsize == 0) {    /* set to default */
		_pypythread_stacksize = 0;
		return 0;
	}

#if defined(THREAD_STACK_SIZE)
# if defined(PTHREAD_STACK_MIN)
	tss_min = PTHREAD_STACK_MIN > THREAD_STACK_MIN ? PTHREAD_STACK_MIN
		: THREAD_STACK_MIN;
# else
	tss_min = THREAD_STACK_MIN;
# endif
	if (newsize >= tss_min) {
		/* validate stack size by setting thread attribute */
		if (pthread_attr_init(&attrs) == 0) {
			rc = pthread_attr_setstacksize(&attrs, newsize);
			pthread_attr_destroy(&attrs);
			if (rc == 0) {
				_pypythread_stacksize = newsize;
				return 0;
			}
		}
	}
	return -1;
#else
	return -2;
#endif
}

/************************************************************/
#ifdef USE_SEMAPHORES
/************************************************************/

#include <semaphore.h>

void RPyThreadAfterFork(void)
{
}

int RPyThreadLockInit(struct RPyOpaque_ThreadLock *lock)
{
	int status, error = 0;
	lock->initialized = 0;
	status = sem_init(&lock->sem, 0, 1);
	CHECK_STATUS("sem_init");
	if (error)
		return 0;
	lock->initialized = 1;
	return 1;
}

void RPyOpaqueDealloc_ThreadLock(struct RPyOpaque_ThreadLock *lock)
{
	int status, error = 0;
	if (lock->initialized) {
		status = sem_destroy(&lock->sem);
		CHECK_STATUS("sem_destroy");
		/* 'error' is ignored;
		   CHECK_STATUS already printed an error message */
	}
}

/*
 * As of February 2002, Cygwin thread implementations mistakenly report error
 * codes in the return value of the sem_ calls (like the pthread_ functions).
 * Correct implementations return -1 and put the code in errno. This supports
 * either.
 */
static int
rpythread_fix_status(int status)
{
	return (status == -1) ? errno : status;
}

int RPyThreadAcquireLock(struct RPyOpaque_ThreadLock *lock, int waitflag)
{
	int success;
	sem_t *thelock = &lock->sem;
	int status, error = 0;

	do {
		if (waitflag)
			status = rpythread_fix_status(sem_wait(thelock));
		else
			status = rpythread_fix_status(sem_trywait(thelock));
	} while (status == EINTR); /* Retry if interrupted by a signal */

	if (waitflag) {
		CHECK_STATUS("sem_wait");
	} else if (status != EAGAIN) {
		CHECK_STATUS("sem_trywait");
	}
	
	success = (status == 0) ? 1 : 0;
	return success;
}

void RPyThreadReleaseLock(struct RPyOpaque_ThreadLock *lock)
{
	sem_t *thelock = &lock->sem;
	int status, error = 0;

	status = sem_post(thelock);
	CHECK_STATUS("sem_post");
}

/************************************************************/
#else                                      /* no semaphores */
/************************************************************/

struct RPyOpaque_ThreadLock *alllocks;   /* doubly-linked list */

void RPyThreadAfterFork(void)
{
	/* Mess.  We have no clue about how it works on CPython on OSX,
	   but the issue is that the state of mutexes is not really
	   preserved across a fork().  So we need to walk over all lock
	   objects here, and rebuild their mutex and condition variable.

	   See e.g. http://hackage.haskell.org/trac/ghc/ticket/1391 for
	   a similar bug about GHC.
	*/
	struct RPyOpaque_ThreadLock *p = alllocks;
	alllocks = NULL;
	while (p) {
		struct RPyOpaque_ThreadLock *next = p->next;
		int was_locked = p->locked;
		RPyThreadLockInit(p);
		p->locked = was_locked;
		p = next;
	}
}

int RPyThreadLockInit(struct RPyOpaque_ThreadLock *lock)
{
	int status, error = 0;

	lock->initialized = 0;
	lock->locked = 0;

	status = pthread_mutex_init(&lock->mut,
				    pthread_mutexattr_default);
	CHECK_STATUS("pthread_mutex_init");

	status = pthread_cond_init(&lock->lock_released,
				   pthread_condattr_default);
	CHECK_STATUS("pthread_cond_init");

	if (error)
		return 0;
	lock->initialized = 1;
	/* add 'lock' in the doubly-linked list */
	if (alllocks)
		alllocks->prev = lock;
	lock->next = alllocks;
	lock->prev = NULL;
	alllocks = lock;
	return 1;
}

void RPyOpaqueDealloc_ThreadLock(struct RPyOpaque_ThreadLock *lock)
{
	int status, error = 0;
	if (lock->initialized) {
		/* remove 'lock' from the doubly-linked list */
		if (lock->prev)
			lock->prev->next = lock->next;
		else {
			assert(alllocks == lock);
			alllocks = lock->next;
		}
		if (lock->next)
			lock->next->prev = lock->prev;

		status = pthread_mutex_destroy(&lock->mut);
		CHECK_STATUS("pthread_mutex_destroy");

		status = pthread_cond_destroy(&lock->lock_released);
		CHECK_STATUS("pthread_cond_destroy");

		/* 'error' is ignored;
		   CHECK_STATUS already printed an error message */
	}
}

int RPyThreadAcquireLock(struct RPyOpaque_ThreadLock *lock, int waitflag)
{
	int success;
	int status, error = 0;

	status = pthread_mutex_lock( &lock->mut );
	CHECK_STATUS("pthread_mutex_lock[1]");
	success = lock->locked == 0;

	if ( !success && waitflag ) {
		/* continue trying until we get the lock */

		/* mut must be locked by me -- part of the condition
		 * protocol */
		while ( lock->locked ) {
			status = pthread_cond_wait(&lock->lock_released,
						   &lock->mut);
			CHECK_STATUS("pthread_cond_wait");
		}
		success = 1;
	}
	if (success) lock->locked = 1;
	status = pthread_mutex_unlock( &lock->mut );
	CHECK_STATUS("pthread_mutex_unlock[1]");

	if (error) success = 0;
	return success;
}

void RPyThreadReleaseLock(struct RPyOpaque_ThreadLock *lock)
{
	int status, error = 0;

	status = pthread_mutex_lock( &lock->mut );
	CHECK_STATUS("pthread_mutex_lock[3]");

	lock->locked = 0;

	status = pthread_mutex_unlock( &lock->mut );
	CHECK_STATUS("pthread_mutex_unlock[3]");

	/* wake up someone (anyone, if any) waiting on the lock */
	status = pthread_cond_signal( &lock->lock_released );
	CHECK_STATUS("pthread_cond_signal");
}

/************************************************************/
#endif                                     /* no semaphores */
/************************************************************/


/************************************************************/
/* GIL code                                                 */
/************************************************************/

#ifdef __llvm__
#  define HAS_ATOMIC_ADD
#endif

#ifdef __GNUC__
#  if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 1)
#    define HAS_ATOMIC_ADD
#  endif
#endif

#ifdef HAS_ATOMIC_ADD
#  define atomic_add __sync_fetch_and_add
#else
#  if defined(__amd64__)
#    define atomic_add(ptr, value)  asm volatile ("lock addq %0, %1"        \
                                 : : "ri"(value), "m"(*(ptr)) : "memory")
#  elif defined(__i386__)
#    define atomic_add(ptr, value)  asm volatile ("lock addl %0, %1"        \
                                 : : "ri"(value), "m"(*(ptr)) : "memory")
#  else
#    error "Please use gcc >= 4.1 or write a custom 'asm' for your CPU."
#  endif
#endif

#define ASSERT_STATUS(call)                             \
    if (call != 0) {                                    \
        fprintf(stderr, "Fatal error: " #call "\n");    \
        abort();                                        \
    }

static void _debug_print(const char *msg)
{
#if 0
    int col = (int)pthread_self();
    col = 31 + ((col / 8) % 8);
    fprintf(stderr, "\033[%dm%s\033[0m", col, msg);
#endif
}

static volatile long pending_acquires = -1;
static pthread_mutex_t mutex_gil;
static pthread_cond_t cond_gil;

static void assert_has_the_gil(void)
{
#ifdef RPY_ASSERT
    assert(pthread_mutex_trylock(&mutex_gil) != 0);
    assert(pending_acquires >= 0);
#endif
}

long RPyGilAllocate(void)
{
    int status, error = 0;
    _debug_print("RPyGilAllocate\n");
    pending_acquires = -1;

    status = pthread_mutex_init(&mutex_gil,
                                pthread_mutexattr_default);
    CHECK_STATUS("pthread_mutex_init[GIL]");

    status = pthread_cond_init(&cond_gil,
                               pthread_condattr_default);
    CHECK_STATUS("pthread_cond_init[GIL]");

    if (error == 0) {
        pending_acquires = 0;
        RPyGilAcquire();
    }
    return (error == 0);
}

long RPyGilYieldThread(void)
{
    /* can be called even before RPyGilAllocate(), but in this case,
       pending_acquires will be -1 */
#ifdef RPY_ASSERT
    if (pending_acquires >= 0)
        assert_has_the_gil();
#endif
    if (pending_acquires <= 0)
        return 0;
    atomic_add(&pending_acquires, 1L);
    _debug_print("{");
    ASSERT_STATUS(pthread_cond_signal(&cond_gil));
    ASSERT_STATUS(pthread_cond_wait(&cond_gil, &mutex_gil));
    _debug_print("}");
    atomic_add(&pending_acquires, -1L);
    assert_has_the_gil();
    return 1;
}

void RPyGilRelease(void)
{
    _debug_print("RPyGilRelease\n");
#ifdef RPY_ASSERT
    assert(pending_acquires >= 0);
#endif
    assert_has_the_gil();
    ASSERT_STATUS(pthread_mutex_unlock(&mutex_gil));
    ASSERT_STATUS(pthread_cond_signal(&cond_gil));
}

void RPyGilAcquire(void)
{
    _debug_print("about to RPyGilAcquire...\n");
#ifdef RPY_ASSERT
    assert(pending_acquires >= 0);
#endif
    atomic_add(&pending_acquires, 1L);
    ASSERT_STATUS(pthread_mutex_lock(&mutex_gil));
    atomic_add(&pending_acquires, -1L);
    assert_has_the_gil();
    _debug_print("RPyGilAcquire\n");
}