X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=src%2Fthread%2Fpthread_create.c;h=2df2e9f9d63663a8bf1589c0ddace0f7eb04753d;hb=14992d4384990d40f6d3cfb799780fb696d77762;hp=33808ce5989af851e8c30da5528188e47f0e716d;hpb=a2d3053354432f3e3b5a787c1527d4b414e9acb0;p=musl

diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index 33808ce5..2df2e9f9 100644
--- a/src/thread/pthread_create.c
+++ b/src/thread/pthread_create.c
@@ -17,6 +17,13 @@ weak_alias(dummy_0, __acquire_ptc);
 weak_alias(dummy_0, __release_ptc);
 weak_alias(dummy_0, __pthread_tsd_run_dtors);
 weak_alias(dummy_0, __do_orphaned_stdio_locks);
+weak_alias(dummy_0, __dl_thread_cleanup);
+
+static void *dummy_1(void *p)
+{
+	return 0;
+}
+weak_alias(dummy_1, __start_sched);
 
 _Noreturn void __pthread_exit(void *result)
 {
@@ -36,11 +43,11 @@ _Noreturn void __pthread_exit(void *result)
 
 	__pthread_tsd_run_dtors();
 
-	__lock(self->exitlock);
-
-	/* Mark this thread dead before decrementing count */
-	__lock(self->killlock);
-	self->dead = 1;
+	/* Access to target the exiting thread with syscalls that use
+	 * its kernel tid is controlled by killlock. For detached threads,
+	 * any use past this point would have undefined behavior, but for
+	 * joinable threads it's a valid usage that must be handled. */
+	LOCK(self->killlock);
 
 	/* Block all signals before decrementing the live thread count.
 	 * This is important to ensure that dynamically allocated TLS
@@ -48,30 +55,18 @@ _Noreturn void __pthread_exit(void *result)
 	 * reasons as well. */
 	__block_all_sigs(&set);
 
-	/* Wait to unlock the kill lock, which governs functions like
-	 * pthread_kill which target a thread id, until signals have
-	 * been blocked. This precludes observation of the thread id
-	 * as a live thread (with application code running in it) after
-	 * the thread was reported dead by ESRCH being returned. */
-	__unlock(self->killlock);
-
 	/* It's impossible to determine whether this is "the last thread"
 	 * until performing the atomic decrement, since multiple threads
 	 * could exit at the same time. For the last thread, revert the
-	 * decrement and unblock signals to give the atexit handlers and
-	 * stdio cleanup code a consistent state. */
+	 * decrement, restore the tid, and unblock signals to give the
+	 * atexit handlers and stdio cleanup code a consistent state. */
 	if (a_fetch_add(&libc.threads_minus_1, -1)==0) {
 		libc.threads_minus_1 = 0;
+		UNLOCK(self->killlock);
 		__restore_sigs(&set);
 		exit(0);
 	}
 
-	if (self->locale != &libc.global_locale) {
-		a_dec(&libc.uselocale_cnt);
-		if (self->locale->ctype_utf8)
-			a_dec(&libc.bytelocale_cnt_minus_1);
-	}
-
 	/* Process robust list in userspace to handle non-pshared mutexes
 	 * and the detached thread case where the robust list head will
 	 * be invalid when the kernel would process it. */
@@ -84,7 +79,7 @@ _Noreturn void __pthread_exit(void *result)
 		int priv = (m->_m_type & 128) ^ 128;
 		self->robust_list.pending = rp;
 		self->robust_list.head = *rp;
-		int cont = a_swap(&m->_m_lock, self->tid|0x40000000);
+		int cont = a_swap(&m->_m_lock, 0x40000000);
 		self->robust_list.pending = 0;
 		if (cont < 0 || waiters)
 			__wake(&m->_m_lock, 1, priv);
@@ -92,16 +87,21 @@ _Noreturn void __pthread_exit(void *result)
 	__vm_unlock();
 
 	__do_orphaned_stdio_locks();
+	__dl_thread_cleanup();
 
-	if (self->detached && self->map_base) {
+	/* This atomic potentially competes with a concurrent pthread_detach
+	 * call; the loser is responsible for freeing thread resources. */
+	int state = a_cas(&self->detach_state, DT_JOINABLE, DT_EXITING);
+
+	if (state>=DT_DETACHED && self->map_base) {
 		/* Detached threads must avoid the kernel clear_child_tid
 		 * feature, since the virtual address will have been
 		 * unmapped and possibly already reused by a new mapping
 		 * at the time the kernel would perform the write. In
 		 * the case of threads that started out detached, the
 		 * initial clone flags are correct, but if the thread was
-		 * detached later (== 2), we need to clear it here. */
-		if (self->detached == 2) __syscall(SYS_set_tid_address, 0);
+		 * detached later, we need to clear it here. */
+		if (state == DT_DYNAMIC) __syscall(SYS_set_tid_address, 0);
 
 		/* Robust list will no longer be valid, and was already
 		 * processed above, so unregister it with the kernel. */
@@ -117,12 +117,17 @@ _Noreturn void __pthread_exit(void *result)
 		__unmapself(self->map_base, self->map_size);
 	}
 
+	/* After the kernel thread exits, its tid may be reused. Clear it
+	 * to prevent inadvertent use and inform functions that would use
+	 * it that it's no longer available. */
+	self->tid = 0;
+	UNLOCK(self->killlock);
+
 	for (;;) __syscall(SYS_exit, 0);
 }
 
 void __do_cleanup_push(struct __ptcb *cb)
 {
-	if (!libc.has_thread_pointer) return;
 	struct pthread *self = __pthread_self();
 	cb->__next = self->cancelbuf;
 	self->cancelbuf = cb;
@@ -130,21 +135,12 @@ void __do_cleanup_push(struct __ptcb *cb)
 
 void __do_cleanup_pop(struct __ptcb *cb)
 {
-	if (!libc.has_thread_pointer) return;
 	__pthread_self()->cancelbuf = cb->__next;
 }
 
 static int start(void *p)
 {
 	pthread_t self = p;
-	if (self->startlock[0]) {
-		__wait(self->startlock, 0, 1, 1);
-		if (self->startlock[0]) {
-			self->detached = 2;
-			pthread_exit(0);
-		}
-		__restore_sigs(self->sigmask);
-	}
 	if (self->unblock_cancel)
 		__syscall(SYS_rt_sigprocmask, SIG_UNBLOCK,
 			SIGPT_SET, 0, _NSIG/8);
@@ -169,6 +165,8 @@ static void *dummy_tsd[1] = { 0 };
 weak_alias(dummy_tsd, __pthread_tsd_main);
 
 volatile int __block_new_threads = 0;
+extern size_t __default_stacksize;
+extern size_t __default_guardsize;
 
 static FILE *volatile dummy_file = 0;
 weak_alias(dummy_file, __stdin_used);
@@ -192,13 +190,15 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 		| CLONE_THREAD | CLONE_SYSVSEM | CLONE_SETTLS
 		| CLONE_PARENT_SETTID | CLONE_CHILD_CLEARTID | CLONE_DETACHED;
 	int do_sched = 0;
-	pthread_attr_t attr = {0};
+	pthread_attr_t attr = { 0 };
+	struct start_sched_args ssa;
 
 	if (!libc.can_do_threads) return ENOSYS;
 	self = __pthread_self();
 	if (!libc.threaded) {
-		for (FILE *f=libc.ofl_head; f; f=f->next)
+		for (FILE *f=*__ofl_lock(); f; f=f->next)
 			init_file_lock(f);
+		__ofl_unlock();
 		init_file_lock(__stdin_used);
 		init_file_lock(__stdout_used);
 		init_file_lock(__stderr_used);
@@ -209,11 +209,16 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 	if (attrp && !c11) attr = *attrp;
 
 	__acquire_ptc();
+	if (!attrp || c11) {
+		attr._a_stacksize = __default_stacksize;
+		attr._a_guardsize = __default_guardsize;
+	}
+
 	if (__block_new_threads) __wait(&__block_new_threads, 0, 1, 1);
 
 	if (attr._a_stackaddr) {
 		size_t need = libc.tls_size + __pthread_tsd_size;
-		size = attr._a_stacksize + DEFAULT_STACK_SIZE;
+		size = attr._a_stacksize;
 		stack = (void *)(attr._a_stackaddr & -16);
 		stack_limit = (void *)(attr._a_stackaddr - size);
 		/* Use application-provided stack for TLS only when
@@ -225,11 +230,11 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 			memset(stack, 0, need);
 		} else {
 			size = ROUND(need);
-			guard = 0;
 		}
+		guard = 0;
 	} else {
-		guard = ROUND(DEFAULT_GUARD_SIZE + attr._a_guardsize);
-		size = guard + ROUND(DEFAULT_STACK_SIZE + attr._a_stacksize
+		guard = ROUND(attr._a_guardsize);
+		size = guard + ROUND(attr._a_stacksize
 			+ libc.tls_size +  __pthread_tsd_size);
 	}
 
@@ -237,7 +242,8 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 		if (guard) {
 			map = __mmap(0, size, PROT_NONE, MAP_PRIVATE|MAP_ANON, -1, 0);
 			if (map == MAP_FAILED) goto fail;
-			if (__mprotect(map+guard, size-guard, PROT_READ|PROT_WRITE)) {
+			if (__mprotect(map+guard, size-guard, PROT_READ|PROT_WRITE)
+			    && errno != ENOSYS) {
 				__munmap(map, size);
 				goto fail;
 			}
@@ -257,30 +263,39 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 	new->map_size = size;
 	new->stack = stack;
 	new->stack_size = stack - stack_limit;
+	new->guard_size = guard;
 	new->start = entry;
 	new->start_arg = arg;
 	new->self = new;
 	new->tsd = (void *)tsd;
 	new->locale = &libc.global_locale;
 	if (attr._a_detach) {
-		new->detached = 1;
+		new->detach_state = DT_DETACHED;
 		flags -= CLONE_CHILD_CLEARTID;
+	} else {
+		new->detach_state = DT_JOINABLE;
 	}
 	if (attr._a_sched) {
-		do_sched = new->startlock[0] = 1;
-		__block_app_sigs(new->sigmask);
+		do_sched = 1;
+		ssa.futex = -1;
+		ssa.start_fn = new->start;
+		ssa.start_arg = new->start_arg;
+		ssa.attr = &attr;
+		new->start = __start_sched;
+		new->start_arg = &ssa;
+		__block_app_sigs(&ssa.mask);
 	}
 	new->robust_list.head = &new->robust_list.head;
 	new->unblock_cancel = self->cancel;
-	new->canary = self->canary;
+	new->CANARY = self->CANARY;
 
 	a_inc(&libc.threads_minus_1);
-	ret = __clone((c11 ? start_c11 : start), stack, flags, new, &new->tid, TP_ADJ(new), &new->tid);
+	ret = __clone((c11 ? start_c11 : start), stack, flags, new, &new->tid, TP_ADJ(new), &new->detach_state);
 
 	__release_ptc();
 
 	if (do_sched) {
-		__restore_sigs(new->sigmask);
+		__restore_sigs(&ssa.mask);
 	}
 
 	if (ret < 0) {
@@ -290,11 +305,8 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 	}
 
 	if (do_sched) {
-		ret = __syscall(SYS_sched_setscheduler, new->tid,
-			attr._a_policy, &attr._a_prio);
-		a_store(new->startlock, ret<0 ? 2 : 0);
-		__wake(new->startlock, 1, 1);
-		if (ret < 0) return -ret;
+		__futexwait(&ssa.futex, -1, 1);
+		if (ret) return ret;
 	}
 
 	*res = new;