summaryrefslogtreecommitdiffstats
path: root/nsprpub/pr/src/pthreads/ptsynch.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-21 11:44:51 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-21 11:44:51 +0000
commit9e3c08db40b8916968b9f30096c7be3f00ce9647 (patch)
treea68f146d7fa01f0134297619fbe7e33db084e0aa /nsprpub/pr/src/pthreads/ptsynch.c
parentInitial commit. (diff)
downloadthunderbird-9e3c08db40b8916968b9f30096c7be3f00ce9647.tar.xz
thunderbird-9e3c08db40b8916968b9f30096c7be3f00ce9647.zip
Adding upstream version 1:115.7.0.upstream/1%115.7.0upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'nsprpub/pr/src/pthreads/ptsynch.c')
-rw-r--r--nsprpub/pr/src/pthreads/ptsynch.c1295
1 files changed, 1295 insertions, 0 deletions
diff --git a/nsprpub/pr/src/pthreads/ptsynch.c b/nsprpub/pr/src/pthreads/ptsynch.c
new file mode 100644
index 0000000000..26f58b2d0f
--- /dev/null
+++ b/nsprpub/pr/src/pthreads/ptsynch.c
@@ -0,0 +1,1295 @@
+/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
+/* This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+/*
+** File: ptsynch.c
+** Descritpion: Implemenation for thread synchronization using pthreads
+** Exports: prlock.h, prcvar.h, prmon.h, prcmon.h
+*/
+
+#if defined(_PR_PTHREADS)
+
+#include "primpl.h"
+#include "obsolete/prsem.h"
+
+#include <string.h>
+#include <pthread.h>
+#include <sys/time.h>
+
+static pthread_mutexattr_t _pt_mattr;
+static pthread_condattr_t _pt_cvar_attr;
+
+#if defined(DEBUG)
+extern PTDebug pt_debug; /* this is shared between several modules */
+#endif /* defined(DEBUG) */
+
+#if defined(FREEBSD)
+/*
+ * On older versions of FreeBSD, pthread_mutex_trylock returns EDEADLK.
+ * Newer versions return EBUSY. We still need to support both.
+ */
+static int
+pt_pthread_mutex_is_locked(pthread_mutex_t *m)
+{
+ int rv = pthread_mutex_trylock(m);
+ return (EBUSY == rv || EDEADLK == rv);
+}
+#endif
+
+/**************************************************************/
+/**************************************************************/
+/*****************************LOCKS****************************/
+/**************************************************************/
+/**************************************************************/
+
+void _PR_InitLocks(void)
+{
+ int rv;
+ rv = _PT_PTHREAD_MUTEXATTR_INIT(&_pt_mattr);
+ PR_ASSERT(0 == rv);
+
+#if (defined(LINUX) && (__GLIBC__ > 2) || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 2)) || \
+ (defined(FREEBSD) && __FreeBSD_version > 700055)
+ rv = pthread_mutexattr_settype(&_pt_mattr, PTHREAD_MUTEX_ADAPTIVE_NP);
+ PR_ASSERT(0 == rv);
+#endif
+
+ rv = _PT_PTHREAD_CONDATTR_INIT(&_pt_cvar_attr);
+ PR_ASSERT(0 == rv);
+}
+
+static void pt_PostNotifies(PRLock *lock, PRBool unlock)
+{
+ PRIntn index, rv;
+ _PT_Notified post;
+ _PT_Notified *notified, *prev = NULL;
+ /*
+ * Time to actually notify any conditions that were affected
+ * while the lock was held. Get a copy of the list that's in
+ * the lock structure and then zero the original. If it's
+ * linked to other such structures, we own that storage.
+ */
+ post = lock->notified; /* a safe copy; we own the lock */
+
+#if defined(DEBUG)
+ memset(&lock->notified, 0, sizeof(_PT_Notified)); /* reset */
+#else
+ lock->notified.length = 0; /* these are really sufficient */
+ lock->notified.link = NULL;
+#endif
+
+ /* should (may) we release lock before notifying? */
+ if (unlock)
+ {
+ rv = pthread_mutex_unlock(&lock->mutex);
+ PR_ASSERT(0 == rv);
+ }
+
+ notified = &post; /* this is where we start */
+ do
+ {
+ for (index = 0; index < notified->length; ++index)
+ {
+ PRCondVar *cv = notified->cv[index].cv;
+ PR_ASSERT(NULL != cv);
+ PR_ASSERT(0 != notified->cv[index].times);
+ if (-1 == notified->cv[index].times)
+ {
+ rv = pthread_cond_broadcast(&cv->cv);
+ PR_ASSERT(0 == rv);
+ }
+ else
+ {
+ while (notified->cv[index].times-- > 0)
+ {
+ rv = pthread_cond_signal(&cv->cv);
+ PR_ASSERT(0 == rv);
+ }
+ }
+#if defined(DEBUG)
+ pt_debug.cvars_notified += 1;
+ if (0 > PR_ATOMIC_DECREMENT(&cv->notify_pending))
+ {
+ pt_debug.delayed_cv_deletes += 1;
+ PR_DestroyCondVar(cv);
+ }
+#else /* defined(DEBUG) */
+ if (0 > PR_ATOMIC_DECREMENT(&cv->notify_pending)) {
+ PR_DestroyCondVar(cv);
+ }
+#endif /* defined(DEBUG) */
+ }
+ prev = notified;
+ notified = notified->link;
+ if (&post != prev) {
+ PR_DELETE(prev);
+ }
+ } while (NULL != notified);
+} /* pt_PostNotifies */
+
+PR_IMPLEMENT(PRLock*) PR_NewLock(void)
+{
+ PRIntn rv;
+ PRLock *lock;
+
+ if (!_pr_initialized) {
+ _PR_ImplicitInitialization();
+ }
+
+ lock = PR_NEWZAP(PRLock);
+ if (lock != NULL)
+ {
+ rv = _PT_PTHREAD_MUTEX_INIT(lock->mutex, _pt_mattr);
+ PR_ASSERT(0 == rv);
+ }
+#if defined(DEBUG)
+ pt_debug.locks_created += 1;
+#endif
+ return lock;
+} /* PR_NewLock */
+
+PR_IMPLEMENT(void) PR_DestroyLock(PRLock *lock)
+{
+ PRIntn rv;
+ PR_ASSERT(NULL != lock);
+ PR_ASSERT(PR_FALSE == lock->locked);
+ PR_ASSERT(0 == lock->notified.length);
+ PR_ASSERT(NULL == lock->notified.link);
+ rv = pthread_mutex_destroy(&lock->mutex);
+ PR_ASSERT(0 == rv);
+#if defined(DEBUG)
+ memset(lock, 0xaf, sizeof(PRLock));
+ pt_debug.locks_destroyed += 1;
+#endif
+ PR_Free(lock);
+} /* PR_DestroyLock */
+
+PR_IMPLEMENT(void) PR_Lock(PRLock *lock)
+{
+ /* Nb: PR_Lock must not call PR_GetCurrentThread to access the |id| or
+ * |tid| field of the current thread's PRThread structure because
+ * _pt_root calls PR_Lock before setting thred->id and thred->tid. */
+ PRIntn rv;
+ PR_ASSERT(lock != NULL);
+ rv = pthread_mutex_lock(&lock->mutex);
+ PR_ASSERT(0 == rv);
+ PR_ASSERT(0 == lock->notified.length);
+ PR_ASSERT(NULL == lock->notified.link);
+ PR_ASSERT(PR_FALSE == lock->locked);
+ /* Nb: the order of the next two statements is not critical to
+ * the correctness of PR_AssertCurrentThreadOwnsLock(), but
+ * this particular order makes the assertion more likely to
+ * catch errors. */
+ lock->owner = pthread_self();
+ lock->locked = PR_TRUE;
+#if defined(DEBUG)
+ pt_debug.locks_acquired += 1;
+#endif
+} /* PR_Lock */
+
+PR_IMPLEMENT(PRStatus) PR_Unlock(PRLock *lock)
+{
+ pthread_t self = pthread_self();
+ PRIntn rv;
+
+ PR_ASSERT(lock != NULL);
+ PR_ASSERT(_PT_PTHREAD_MUTEX_IS_LOCKED(lock->mutex));
+ PR_ASSERT(PR_TRUE == lock->locked);
+ PR_ASSERT(pthread_equal(lock->owner, self));
+
+ if (!lock->locked || !pthread_equal(lock->owner, self)) {
+ return PR_FAILURE;
+ }
+
+ lock->locked = PR_FALSE;
+ if (0 == lock->notified.length) /* shortcut */
+ {
+ rv = pthread_mutex_unlock(&lock->mutex);
+ PR_ASSERT(0 == rv);
+ }
+ else {
+ pt_PostNotifies(lock, PR_TRUE);
+ }
+
+#if defined(DEBUG)
+ pt_debug.locks_released += 1;
+#endif
+ return PR_SUCCESS;
+} /* PR_Unlock */
+
+PR_IMPLEMENT(void) PR_AssertCurrentThreadOwnsLock(PRLock *lock)
+{
+ /* Nb: the order of the |locked| and |owner==me| checks is not critical
+ * to the correctness of PR_AssertCurrentThreadOwnsLock(), but
+ * this particular order makes the assertion more likely to
+ * catch errors. */
+ PR_ASSERT(lock->locked && pthread_equal(lock->owner, pthread_self()));
+}
+
+/**************************************************************/
+/**************************************************************/
+/***************************CONDITIONS*************************/
+/**************************************************************/
+/**************************************************************/
+
+
+/*
+ * This code is used to compute the absolute time for the wakeup.
+ * It's moderately ugly, so it's defined here and called in a
+ * couple of places.
+ */
+#define PT_NANOPERMICRO 1000UL
+#define PT_BILLION 1000000000UL
+
+static PRIntn pt_TimedWait(
+ pthread_cond_t *cv, pthread_mutex_t *ml, PRIntervalTime timeout)
+{
+ int rv;
+ struct timeval now;
+ struct timespec tmo;
+ PRUint32 ticks = PR_TicksPerSecond();
+
+ tmo.tv_sec = (PRInt32)(timeout / ticks);
+ tmo.tv_nsec = (PRInt32)(timeout - (tmo.tv_sec * ticks));
+ tmo.tv_nsec = (PRInt32)PR_IntervalToMicroseconds(PT_NANOPERMICRO * tmo.tv_nsec);
+
+ /* pthreads wants this in absolute time, off we go ... */
+ (void)GETTIMEOFDAY(&now);
+ /* that one's usecs, this one's nsecs - grrrr! */
+ tmo.tv_sec += now.tv_sec;
+ tmo.tv_nsec += (PT_NANOPERMICRO * now.tv_usec);
+ tmo.tv_sec += tmo.tv_nsec / PT_BILLION;
+ tmo.tv_nsec %= PT_BILLION;
+
+ rv = pthread_cond_timedwait(cv, ml, &tmo);
+
+ /* NSPR doesn't report timeouts */
+ return (rv == ETIMEDOUT) ? 0 : rv;
+} /* pt_TimedWait */
+
+
+/*
+ * Notifies just get posted to the protecting mutex. The
+ * actual notification is done when the lock is released so that
+ * MP systems don't contend for a lock that they can't have.
+ */
+static void pt_PostNotifyToCvar(PRCondVar *cvar, PRBool broadcast)
+{
+ PRIntn index = 0;
+ _PT_Notified *notified = &cvar->lock->notified;
+
+ PR_ASSERT(PR_TRUE == cvar->lock->locked);
+ PR_ASSERT(pthread_equal(cvar->lock->owner, pthread_self()));
+ PR_ASSERT(_PT_PTHREAD_MUTEX_IS_LOCKED(cvar->lock->mutex));
+
+ while (1)
+ {
+ for (index = 0; index < notified->length; ++index)
+ {
+ if (notified->cv[index].cv == cvar)
+ {
+ if (broadcast) {
+ notified->cv[index].times = -1;
+ }
+ else if (-1 != notified->cv[index].times) {
+ notified->cv[index].times += 1;
+ }
+ return; /* we're finished */
+ }
+ }
+ /* if not full, enter new CV in this array */
+ if (notified->length < PT_CV_NOTIFIED_LENGTH) {
+ break;
+ }
+
+ /* if there's no link, create an empty array and link it */
+ if (NULL == notified->link) {
+ notified->link = PR_NEWZAP(_PT_Notified);
+ }
+ notified = notified->link;
+ }
+
+ /* A brand new entry in the array */
+ (void)PR_ATOMIC_INCREMENT(&cvar->notify_pending);
+ notified->cv[index].times = (broadcast) ? -1 : 1;
+ notified->cv[index].cv = cvar;
+ notified->length += 1;
+} /* pt_PostNotifyToCvar */
+
+PR_IMPLEMENT(PRCondVar*) PR_NewCondVar(PRLock *lock)
+{
+ PRCondVar *cv = PR_NEW(PRCondVar);
+ PR_ASSERT(lock != NULL);
+ if (cv != NULL)
+ {
+ int rv = _PT_PTHREAD_COND_INIT(cv->cv, _pt_cvar_attr);
+ PR_ASSERT(0 == rv);
+ if (0 == rv)
+ {
+ cv->lock = lock;
+ cv->notify_pending = 0;
+#if defined(DEBUG)
+ pt_debug.cvars_created += 1;
+#endif
+ }
+ else
+ {
+ PR_DELETE(cv);
+ cv = NULL;
+ }
+ }
+ return cv;
+} /* PR_NewCondVar */
+
+PR_IMPLEMENT(void) PR_DestroyCondVar(PRCondVar *cvar)
+{
+ if (0 > PR_ATOMIC_DECREMENT(&cvar->notify_pending))
+ {
+ PRIntn rv = pthread_cond_destroy(&cvar->cv);
+#if defined(DEBUG)
+ PR_ASSERT(0 == rv);
+ memset(cvar, 0xaf, sizeof(PRCondVar));
+ pt_debug.cvars_destroyed += 1;
+#else
+ (void)rv;
+#endif
+ PR_Free(cvar);
+ }
+} /* PR_DestroyCondVar */
+
+PR_IMPLEMENT(PRStatus) PR_WaitCondVar(PRCondVar *cvar, PRIntervalTime timeout)
+{
+ PRIntn rv;
+ PRThread *thred = PR_GetCurrentThread();
+
+ PR_ASSERT(cvar != NULL);
+ /* We'd better be locked */
+ PR_ASSERT(_PT_PTHREAD_MUTEX_IS_LOCKED(cvar->lock->mutex));
+ PR_ASSERT(PR_TRUE == cvar->lock->locked);
+ /* and it better be by us */
+ PR_ASSERT(pthread_equal(cvar->lock->owner, pthread_self()));
+
+ if (_PT_THREAD_INTERRUPTED(thred)) {
+ goto aborted;
+ }
+
+ /*
+ * The thread waiting is used for PR_Interrupt
+ */
+ thred->waiting = cvar; /* this is where we're waiting */
+
+ /*
+ * If we have pending notifies, post them now.
+ *
+ * This is not optimal. We're going to post these notifies
+ * while we're holding the lock. That means on MP systems
+ * that they are going to collide for the lock that we will
+ * hold until we actually wait.
+ */
+ if (0 != cvar->lock->notified.length) {
+ pt_PostNotifies(cvar->lock, PR_FALSE);
+ }
+
+ /*
+ * We're surrendering the lock, so clear out the locked field.
+ */
+ cvar->lock->locked = PR_FALSE;
+
+ if (timeout == PR_INTERVAL_NO_TIMEOUT) {
+ rv = pthread_cond_wait(&cvar->cv, &cvar->lock->mutex);
+ }
+ else {
+ rv = pt_TimedWait(&cvar->cv, &cvar->lock->mutex, timeout);
+ }
+
+ /* We just got the lock back - this better be empty */
+ PR_ASSERT(PR_FALSE == cvar->lock->locked);
+ cvar->lock->locked = PR_TRUE;
+ cvar->lock->owner = pthread_self();
+
+ PR_ASSERT(0 == cvar->lock->notified.length);
+ thred->waiting = NULL; /* and now we're not */
+ if (_PT_THREAD_INTERRUPTED(thred)) {
+ goto aborted;
+ }
+ if (rv != 0)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(rv);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+
+aborted:
+ PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0);
+ thred->state &= ~PT_THREAD_ABORTED;
+ return PR_FAILURE;
+} /* PR_WaitCondVar */
+
+PR_IMPLEMENT(PRStatus) PR_NotifyCondVar(PRCondVar *cvar)
+{
+ PR_ASSERT(cvar != NULL);
+ pt_PostNotifyToCvar(cvar, PR_FALSE);
+ return PR_SUCCESS;
+} /* PR_NotifyCondVar */
+
+PR_IMPLEMENT(PRStatus) PR_NotifyAllCondVar(PRCondVar *cvar)
+{
+ PR_ASSERT(cvar != NULL);
+ pt_PostNotifyToCvar(cvar, PR_TRUE);
+ return PR_SUCCESS;
+} /* PR_NotifyAllCondVar */
+
+/**************************************************************/
+/**************************************************************/
+/***************************MONITORS***************************/
+/**************************************************************/
+/**************************************************************/
+
+/*
+ * Notifies just get posted to the monitor. The actual notification is done
+ * when the monitor is fully exited so that MP systems don't contend for a
+ * monitor that they can't enter.
+ */
+static void pt_PostNotifyToMonitor(PRMonitor *mon, PRBool broadcast)
+{
+ PR_ASSERT(NULL != mon);
+ PR_ASSERT_CURRENT_THREAD_IN_MONITOR(mon);
+
+ /* mon->notifyTimes is protected by the monitor, so we don't need to
+ * acquire mon->lock.
+ */
+ if (broadcast) {
+ mon->notifyTimes = -1;
+ }
+ else if (-1 != mon->notifyTimes) {
+ mon->notifyTimes += 1;
+ }
+} /* pt_PostNotifyToMonitor */
+
+static void pt_PostNotifiesFromMonitor(pthread_cond_t *cv, PRIntn times)
+{
+ PRIntn rv;
+
+ /*
+ * Time to actually notify any waits that were affected while the monitor
+ * was entered.
+ */
+ PR_ASSERT(NULL != cv);
+ PR_ASSERT(0 != times);
+ if (-1 == times)
+ {
+ rv = pthread_cond_broadcast(cv);
+ PR_ASSERT(0 == rv);
+ }
+ else
+ {
+ while (times-- > 0)
+ {
+ rv = pthread_cond_signal(cv);
+ PR_ASSERT(0 == rv);
+ }
+ }
+} /* pt_PostNotifiesFromMonitor */
+
+PR_IMPLEMENT(PRMonitor*) PR_NewMonitor(void)
+{
+ PRMonitor *mon;
+ int rv;
+
+ if (!_pr_initialized) {
+ _PR_ImplicitInitialization();
+ }
+
+ mon = PR_NEWZAP(PRMonitor);
+ if (mon == NULL)
+ {
+ PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0);
+ return NULL;
+ }
+
+ rv = _PT_PTHREAD_MUTEX_INIT(mon->lock, _pt_mattr);
+ PR_ASSERT(0 == rv);
+ if (0 != rv) {
+ goto error1;
+ }
+
+ _PT_PTHREAD_INVALIDATE_THR_HANDLE(mon->owner);
+
+ rv = _PT_PTHREAD_COND_INIT(mon->entryCV, _pt_cvar_attr);
+ PR_ASSERT(0 == rv);
+ if (0 != rv) {
+ goto error2;
+ }
+
+ rv = _PT_PTHREAD_COND_INIT(mon->waitCV, _pt_cvar_attr);
+ PR_ASSERT(0 == rv);
+ if (0 != rv) {
+ goto error3;
+ }
+
+ mon->notifyTimes = 0;
+ mon->entryCount = 0;
+ mon->refCount = 1;
+ mon->name = NULL;
+ return mon;
+
+error3:
+ pthread_cond_destroy(&mon->entryCV);
+error2:
+ pthread_mutex_destroy(&mon->lock);
+error1:
+ PR_Free(mon);
+ _PR_MD_MAP_DEFAULT_ERROR(rv);
+ return NULL;
+} /* PR_NewMonitor */
+
+PR_IMPLEMENT(PRMonitor*) PR_NewNamedMonitor(const char* name)
+{
+ PRMonitor* mon = PR_NewMonitor();
+ if (mon) {
+ mon->name = name;
+ }
+ return mon;
+}
+
+PR_IMPLEMENT(void) PR_DestroyMonitor(PRMonitor *mon)
+{
+ int rv;
+
+ PR_ASSERT(mon != NULL);
+ if (PR_ATOMIC_DECREMENT(&mon->refCount) == 0)
+ {
+ rv = pthread_cond_destroy(&mon->waitCV); PR_ASSERT(0 == rv);
+ rv = pthread_cond_destroy(&mon->entryCV); PR_ASSERT(0 == rv);
+ rv = pthread_mutex_destroy(&mon->lock); PR_ASSERT(0 == rv);
+#if defined(DEBUG)
+ memset(mon, 0xaf, sizeof(PRMonitor));
+#endif
+ PR_Free(mon);
+ }
+} /* PR_DestroyMonitor */
+
+/* The GC uses this; it is quite arguably a bad interface. I'm just
+ * duplicating it for now - XXXMB
+ */
+PR_IMPLEMENT(PRIntn) PR_GetMonitorEntryCount(PRMonitor *mon)
+{
+ pthread_t self = pthread_self();
+ PRIntn rv;
+ PRIntn count = 0;
+
+ rv = pthread_mutex_lock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ if (pthread_equal(mon->owner, self)) {
+ count = mon->entryCount;
+ }
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ return count;
+}
+
+PR_IMPLEMENT(void) PR_AssertCurrentThreadInMonitor(PRMonitor *mon)
+{
+#if defined(DEBUG) || defined(FORCE_PR_ASSERT)
+ PRIntn rv;
+
+ rv = pthread_mutex_lock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ PR_ASSERT(mon->entryCount != 0 &&
+ pthread_equal(mon->owner, pthread_self()));
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+#endif
+}
+
+PR_IMPLEMENT(void) PR_EnterMonitor(PRMonitor *mon)
+{
+ pthread_t self = pthread_self();
+ PRIntn rv;
+
+ PR_ASSERT(mon != NULL);
+ rv = pthread_mutex_lock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ if (mon->entryCount != 0)
+ {
+ if (pthread_equal(mon->owner, self)) {
+ goto done;
+ }
+ while (mon->entryCount != 0)
+ {
+ rv = pthread_cond_wait(&mon->entryCV, &mon->lock);
+ PR_ASSERT(0 == rv);
+ }
+ }
+ /* and now I have the monitor */
+ PR_ASSERT(0 == mon->notifyTimes);
+ PR_ASSERT(_PT_PTHREAD_THR_HANDLE_IS_INVALID(mon->owner));
+ _PT_PTHREAD_COPY_THR_HANDLE(self, mon->owner);
+
+done:
+ mon->entryCount += 1;
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+} /* PR_EnterMonitor */
+
+PR_IMPLEMENT(PRStatus) PR_ExitMonitor(PRMonitor *mon)
+{
+ pthread_t self = pthread_self();
+ PRIntn rv;
+ PRBool notifyEntryWaiter = PR_FALSE;
+ PRIntn notifyTimes = 0;
+
+ PR_ASSERT(mon != NULL);
+ rv = pthread_mutex_lock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ /* the entries should be > 0 and we'd better be the owner */
+ PR_ASSERT(mon->entryCount > 0);
+ PR_ASSERT(pthread_equal(mon->owner, self));
+ if (mon->entryCount == 0 || !pthread_equal(mon->owner, self))
+ {
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ return PR_FAILURE;
+ }
+
+ mon->entryCount -= 1; /* reduce by one */
+ if (mon->entryCount == 0)
+ {
+ /* and if it transitioned to zero - notify an entry waiter */
+ /* make the owner unknown */
+ _PT_PTHREAD_INVALIDATE_THR_HANDLE(mon->owner);
+ notifyEntryWaiter = PR_TRUE;
+ notifyTimes = mon->notifyTimes;
+ mon->notifyTimes = 0;
+ /* We will access the members of 'mon' after unlocking mon->lock.
+ * Add a reference. */
+ PR_ATOMIC_INCREMENT(&mon->refCount);
+ }
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ if (notifyEntryWaiter)
+ {
+ if (notifyTimes) {
+ pt_PostNotifiesFromMonitor(&mon->waitCV, notifyTimes);
+ }
+ rv = pthread_cond_signal(&mon->entryCV);
+ PR_ASSERT(0 == rv);
+ /* We are done accessing the members of 'mon'. Release the
+ * reference. */
+ PR_DestroyMonitor(mon);
+ }
+ return PR_SUCCESS;
+} /* PR_ExitMonitor */
+
+PR_IMPLEMENT(PRStatus) PR_Wait(PRMonitor *mon, PRIntervalTime timeout)
+{
+ PRStatus rv;
+ PRUint32 saved_entries;
+ pthread_t saved_owner;
+
+ PR_ASSERT(mon != NULL);
+ rv = pthread_mutex_lock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ /* the entries better be positive */
+ PR_ASSERT(mon->entryCount > 0);
+ /* and it better be owned by us */
+ PR_ASSERT(pthread_equal(mon->owner, pthread_self()));
+
+ /* tuck these away 'till later */
+ saved_entries = mon->entryCount;
+ mon->entryCount = 0;
+ _PT_PTHREAD_COPY_THR_HANDLE(mon->owner, saved_owner);
+ _PT_PTHREAD_INVALIDATE_THR_HANDLE(mon->owner);
+ /*
+ * If we have pending notifies, post them now.
+ *
+ * This is not optimal. We're going to post these notifies
+ * while we're holding the lock. That means on MP systems
+ * that they are going to collide for the lock that we will
+ * hold until we actually wait.
+ */
+ if (0 != mon->notifyTimes)
+ {
+ pt_PostNotifiesFromMonitor(&mon->waitCV, mon->notifyTimes);
+ mon->notifyTimes = 0;
+ }
+ rv = pthread_cond_signal(&mon->entryCV);
+ PR_ASSERT(0 == rv);
+
+ if (timeout == PR_INTERVAL_NO_TIMEOUT) {
+ rv = pthread_cond_wait(&mon->waitCV, &mon->lock);
+ }
+ else {
+ rv = pt_TimedWait(&mon->waitCV, &mon->lock, timeout);
+ }
+ PR_ASSERT(0 == rv);
+
+ while (mon->entryCount != 0)
+ {
+ rv = pthread_cond_wait(&mon->entryCV, &mon->lock);
+ PR_ASSERT(0 == rv);
+ }
+ PR_ASSERT(0 == mon->notifyTimes);
+ /* reinstate the interesting information */
+ mon->entryCount = saved_entries;
+ _PT_PTHREAD_COPY_THR_HANDLE(saved_owner, mon->owner);
+
+ rv = pthread_mutex_unlock(&mon->lock);
+ PR_ASSERT(0 == rv);
+ return rv;
+} /* PR_Wait */
+
+PR_IMPLEMENT(PRStatus) PR_Notify(PRMonitor *mon)
+{
+ pt_PostNotifyToMonitor(mon, PR_FALSE);
+ return PR_SUCCESS;
+} /* PR_Notify */
+
+PR_IMPLEMENT(PRStatus) PR_NotifyAll(PRMonitor *mon)
+{
+ pt_PostNotifyToMonitor(mon, PR_TRUE);
+ return PR_SUCCESS;
+} /* PR_NotifyAll */
+
+/**************************************************************/
+/**************************************************************/
+/**************************SEMAPHORES**************************/
+/**************************************************************/
+/**************************************************************/
+PR_IMPLEMENT(void) PR_PostSem(PRSemaphore *semaphore)
+{
+ static PRBool unwarned = PR_TRUE;
+ if (unwarned) unwarned = _PR_Obsolete(
+ "PR_PostSem", "locks & condition variables");
+ PR_Lock(semaphore->cvar->lock);
+ PR_NotifyCondVar(semaphore->cvar);
+ semaphore->count += 1;
+ PR_Unlock(semaphore->cvar->lock);
+} /* PR_PostSem */
+
+PR_IMPLEMENT(PRStatus) PR_WaitSem(PRSemaphore *semaphore)
+{
+ PRStatus status = PR_SUCCESS;
+ static PRBool unwarned = PR_TRUE;
+ if (unwarned) unwarned = _PR_Obsolete(
+ "PR_WaitSem", "locks & condition variables");
+ PR_Lock(semaphore->cvar->lock);
+ while ((semaphore->count == 0) && (PR_SUCCESS == status)) {
+ status = PR_WaitCondVar(semaphore->cvar, PR_INTERVAL_NO_TIMEOUT);
+ }
+ if (PR_SUCCESS == status) {
+ semaphore->count -= 1;
+ }
+ PR_Unlock(semaphore->cvar->lock);
+ return status;
+} /* PR_WaitSem */
+
+PR_IMPLEMENT(void) PR_DestroySem(PRSemaphore *semaphore)
+{
+ static PRBool unwarned = PR_TRUE;
+ if (unwarned) unwarned = _PR_Obsolete(
+ "PR_DestroySem", "locks & condition variables");
+ PR_DestroyLock(semaphore->cvar->lock);
+ PR_DestroyCondVar(semaphore->cvar);
+ PR_Free(semaphore);
+} /* PR_DestroySem */
+
+PR_IMPLEMENT(PRSemaphore*) PR_NewSem(PRUintn value)
+{
+ PRSemaphore *semaphore;
+ static PRBool unwarned = PR_TRUE;
+ if (!_pr_initialized) {
+ _PR_ImplicitInitialization();
+ }
+
+ if (unwarned) unwarned = _PR_Obsolete(
+ "PR_NewSem", "locks & condition variables");
+
+ semaphore = PR_NEWZAP(PRSemaphore);
+ if (NULL != semaphore)
+ {
+ PRLock *lock = PR_NewLock();
+ if (NULL != lock)
+ {
+ semaphore->cvar = PR_NewCondVar(lock);
+ if (NULL != semaphore->cvar)
+ {
+ semaphore->count = value;
+ return semaphore;
+ }
+ PR_DestroyLock(lock);
+ }
+ PR_Free(semaphore);
+ }
+ return NULL;
+}
+
+/*
+ * Define the interprocess named semaphore functions.
+ * There are three implementations:
+ * 1. POSIX semaphore based;
+ * 2. System V semaphore based;
+ * 3. unsupported (fails with PR_NOT_IMPLEMENTED_ERROR).
+ */
+
+#ifdef _PR_HAVE_POSIX_SEMAPHORES
+#include <fcntl.h>
+
+PR_IMPLEMENT(PRSem *) PR_OpenSemaphore(
+ const char *name,
+ PRIntn flags,
+ PRIntn mode,
+ PRUintn value)
+{
+ PRSem *sem;
+ char osname[PR_IPC_NAME_SIZE];
+
+ if (_PR_MakeNativeIPCName(name, osname, sizeof(osname), _PRIPCSem)
+ == PR_FAILURE)
+ {
+ return NULL;
+ }
+
+ sem = PR_NEW(PRSem);
+ if (NULL == sem)
+ {
+ PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0);
+ return NULL;
+ }
+
+ if (flags & PR_SEM_CREATE)
+ {
+ int oflag = O_CREAT;
+
+ if (flags & PR_SEM_EXCL) {
+ oflag |= O_EXCL;
+ }
+ sem->sem = sem_open(osname, oflag, mode, value);
+ }
+ else
+ {
+#ifdef HPUX
+ /* Pass 0 as the mode and value arguments to work around a bug. */
+ sem->sem = sem_open(osname, 0, 0, 0);
+#else
+ sem->sem = sem_open(osname, 0);
+#endif
+ }
+ if ((sem_t *) -1 == sem->sem)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ PR_Free(sem);
+ return NULL;
+ }
+ return sem;
+}
+
+PR_IMPLEMENT(PRStatus) PR_WaitSemaphore(PRSem *sem)
+{
+ int rv;
+ rv = sem_wait(sem->sem);
+ if (0 != rv)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_PostSemaphore(PRSem *sem)
+{
+ int rv;
+ rv = sem_post(sem->sem);
+ if (0 != rv)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_CloseSemaphore(PRSem *sem)
+{
+ int rv;
+ rv = sem_close(sem->sem);
+ if (0 != rv)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ PR_Free(sem);
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_DeleteSemaphore(const char *name)
+{
+ int rv;
+ char osname[PR_IPC_NAME_SIZE];
+
+ if (_PR_MakeNativeIPCName(name, osname, sizeof(osname), _PRIPCSem)
+ == PR_FAILURE)
+ {
+ return PR_FAILURE;
+ }
+ rv = sem_unlink(osname);
+ if (0 != rv)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+#elif defined(_PR_HAVE_SYSV_SEMAPHORES)
+
+#include <fcntl.h>
+#include <sys/sem.h>
+
+/*
+ * From the semctl(2) man page in glibc 2.0
+ */
+#if (defined(__GNU_LIBRARY__) && !defined(_SEM_SEMUN_UNDEFINED)) \
+ || (defined(FREEBSD) && __FreeBSD_version < 1200059) \
+ || defined(OPENBSD) || defined(BSDI) \
+ || defined(DARWIN)
+/* union semun is defined by including <sys/sem.h> */
+#else
+/* according to X/OPEN we have to define it ourselves */
+union semun {
+ int val;
+ struct semid_ds *buf;
+ unsigned short *array;
+};
+#endif
+
+/*
+ * 'a' (97) is the final closing price of NSCP stock.
+ */
+#define NSPR_IPC_KEY_ID 'a' /* the id argument for ftok() */
+
+#define NSPR_SEM_MODE 0666
+
+PR_IMPLEMENT(PRSem *) PR_OpenSemaphore(
+ const char *name,
+ PRIntn flags,
+ PRIntn mode,
+ PRUintn value)
+{
+ PRSem *sem;
+ key_t key;
+ union semun arg;
+ struct sembuf sop;
+ struct semid_ds seminfo;
+#define MAX_TRIES 60
+ PRIntn i;
+ char osname[PR_IPC_NAME_SIZE];
+
+ if (_PR_MakeNativeIPCName(name, osname, sizeof(osname), _PRIPCSem)
+ == PR_FAILURE)
+ {
+ return NULL;
+ }
+
+ /* Make sure the file exists before calling ftok. */
+ if (flags & PR_SEM_CREATE)
+ {
+ int osfd = open(osname, O_RDWR|O_CREAT, mode);
+ if (-1 == osfd)
+ {
+ _PR_MD_MAP_OPEN_ERROR(errno);
+ return NULL;
+ }
+ if (close(osfd) == -1)
+ {
+ _PR_MD_MAP_CLOSE_ERROR(errno);
+ return NULL;
+ }
+ }
+ key = ftok(osname, NSPR_IPC_KEY_ID);
+ if ((key_t)-1 == key)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return NULL;
+ }
+
+ sem = PR_NEW(PRSem);
+ if (NULL == sem)
+ {
+ PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0);
+ return NULL;
+ }
+
+ if (flags & PR_SEM_CREATE)
+ {
+ sem->semid = semget(key, 1, mode|IPC_CREAT|IPC_EXCL);
+ if (sem->semid >= 0)
+ {
+ /* creator of a semaphore is responsible for initializing it */
+ arg.val = 0;
+ if (semctl(sem->semid, 0, SETVAL, arg) == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ PR_Free(sem);
+ return NULL;
+ }
+ /* call semop to set sem_otime to nonzero */
+ sop.sem_num = 0;
+ sop.sem_op = value;
+ sop.sem_flg = 0;
+ if (semop(sem->semid, &sop, 1) == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ PR_Free(sem);
+ return NULL;
+ }
+ return sem;
+ }
+
+ if (errno != EEXIST || flags & PR_SEM_EXCL)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ PR_Free(sem);
+ return NULL;
+ }
+ }
+
+ sem->semid = semget(key, 1, NSPR_SEM_MODE);
+ if (sem->semid == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ PR_Free(sem);
+ return NULL;
+ }
+ for (i = 0; i < MAX_TRIES; i++)
+ {
+ arg.buf = &seminfo;
+ semctl(sem->semid, 0, IPC_STAT, arg);
+ if (seminfo.sem_otime != 0) {
+ break;
+ }
+ sleep(1);
+ }
+ if (i == MAX_TRIES)
+ {
+ PR_SetError(PR_IO_TIMEOUT_ERROR, 0);
+ PR_Free(sem);
+ return NULL;
+ }
+ return sem;
+}
+
+PR_IMPLEMENT(PRStatus) PR_WaitSemaphore(PRSem *sem)
+{
+ struct sembuf sop;
+
+ sop.sem_num = 0;
+ sop.sem_op = -1;
+ sop.sem_flg = 0;
+ if (semop(sem->semid, &sop, 1) == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_PostSemaphore(PRSem *sem)
+{
+ struct sembuf sop;
+
+ sop.sem_num = 0;
+ sop.sem_op = 1;
+ sop.sem_flg = 0;
+ if (semop(sem->semid, &sop, 1) == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_CloseSemaphore(PRSem *sem)
+{
+ PR_Free(sem);
+ return PR_SUCCESS;
+}
+
+PR_IMPLEMENT(PRStatus) PR_DeleteSemaphore(const char *name)
+{
+ key_t key;
+ int semid;
+ /* On some systems (e.g., glibc 2.0) semctl requires a fourth argument */
+ union semun unused;
+ char osname[PR_IPC_NAME_SIZE];
+
+ if (_PR_MakeNativeIPCName(name, osname, sizeof(osname), _PRIPCSem)
+ == PR_FAILURE)
+ {
+ return PR_FAILURE;
+ }
+ key = ftok(osname, NSPR_IPC_KEY_ID);
+ if ((key_t) -1 == key)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ if (unlink(osname) == -1)
+ {
+ _PR_MD_MAP_UNLINK_ERROR(errno);
+ return PR_FAILURE;
+ }
+ semid = semget(key, 1, NSPR_SEM_MODE);
+ if (-1 == semid)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ unused.val = 0;
+ if (semctl(semid, 0, IPC_RMID, unused) == -1)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(errno);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+}
+
+#else /* neither POSIX nor System V semaphores are available */
+
+PR_IMPLEMENT(PRSem *) PR_OpenSemaphore(
+ const char *name,
+ PRIntn flags,
+ PRIntn mode,
+ PRUintn value)
+{
+ PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0);
+ return NULL;
+}
+
+PR_IMPLEMENT(PRStatus) PR_WaitSemaphore(PRSem *sem)
+{
+ PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0);
+ return PR_FAILURE;
+}
+
+PR_IMPLEMENT(PRStatus) PR_PostSemaphore(PRSem *sem)
+{
+ PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0);
+ return PR_FAILURE;
+}
+
+PR_IMPLEMENT(PRStatus) PR_CloseSemaphore(PRSem *sem)
+{
+ PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0);
+ return PR_FAILURE;
+}
+
+PR_IMPLEMENT(PRStatus) PR_DeleteSemaphore(const char *name)
+{
+ PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0);
+ return PR_FAILURE;
+}
+
+#endif /* end of interprocess named semaphore functions */
+
+/**************************************************************/
+/**************************************************************/
+/******************ROUTINES FOR DCE EMULATION******************/
+/**************************************************************/
+/**************************************************************/
+
+#include "prpdce.h"
+
+PR_IMPLEMENT(PRStatus) PRP_TryLock(PRLock *lock)
+{
+ PRIntn rv = pthread_mutex_trylock(&lock->mutex);
+ if (rv == 0)
+ {
+ PR_ASSERT(PR_FALSE == lock->locked);
+ lock->locked = PR_TRUE;
+ lock->owner = pthread_self();
+ }
+ /* XXX set error code? */
+ return (0 == rv) ? PR_SUCCESS : PR_FAILURE;
+} /* PRP_TryLock */
+
+PR_IMPLEMENT(PRCondVar*) PRP_NewNakedCondVar(void)
+{
+ PRCondVar *cv;
+
+ if (!_pr_initialized) {
+ _PR_ImplicitInitialization();
+ }
+
+ cv = PR_NEW(PRCondVar);
+ if (cv != NULL)
+ {
+ int rv;
+ rv = _PT_PTHREAD_COND_INIT(cv->cv, _pt_cvar_attr);
+ PR_ASSERT(0 == rv);
+ if (0 == rv)
+ {
+ cv->lock = _PR_NAKED_CV_LOCK;
+ }
+ else
+ {
+ PR_DELETE(cv);
+ cv = NULL;
+ }
+ }
+ return cv;
+} /* PRP_NewNakedCondVar */
+
+PR_IMPLEMENT(void) PRP_DestroyNakedCondVar(PRCondVar *cvar)
+{
+ int rv;
+ rv = pthread_cond_destroy(&cvar->cv); PR_ASSERT(0 == rv);
+#if defined(DEBUG)
+ memset(cvar, 0xaf, sizeof(PRCondVar));
+#endif
+ PR_Free(cvar);
+} /* PRP_DestroyNakedCondVar */
+
+PR_IMPLEMENT(PRStatus) PRP_NakedWait(
+ PRCondVar *cvar, PRLock *ml, PRIntervalTime timeout)
+{
+ PRIntn rv;
+ PR_ASSERT(cvar != NULL);
+ /* XXX do we really want to assert this in a naked wait? */
+ PR_ASSERT(_PT_PTHREAD_MUTEX_IS_LOCKED(ml->mutex));
+ if (timeout == PR_INTERVAL_NO_TIMEOUT) {
+ rv = pthread_cond_wait(&cvar->cv, &ml->mutex);
+ }
+ else {
+ rv = pt_TimedWait(&cvar->cv, &ml->mutex, timeout);
+ }
+ if (rv != 0)
+ {
+ _PR_MD_MAP_DEFAULT_ERROR(rv);
+ return PR_FAILURE;
+ }
+ return PR_SUCCESS;
+} /* PRP_NakedWait */
+
+PR_IMPLEMENT(PRStatus) PRP_NakedNotify(PRCondVar *cvar)
+{
+ int rv;
+ PR_ASSERT(cvar != NULL);
+ rv = pthread_cond_signal(&cvar->cv);
+ PR_ASSERT(0 == rv);
+ return PR_SUCCESS;
+} /* PRP_NakedNotify */
+
+PR_IMPLEMENT(PRStatus) PRP_NakedBroadcast(PRCondVar *cvar)
+{
+ int rv;
+ PR_ASSERT(cvar != NULL);
+ rv = pthread_cond_broadcast(&cvar->cv);
+ PR_ASSERT(0 == rv);
+ return PR_SUCCESS;
+} /* PRP_NakedBroadcast */
+
+#endif /* defined(_PR_PTHREADS) */
+
+/* ptsynch.c */