#include "pthread_impl.h"
#include "stdio_impl.h"
#include "libc.h"
+#include "lock.h"
#include <sys/mman.h>
#include <string.h>
#include <stddef.h>
-void *__mmap(void *, size_t, int, int, int, off_t);
-int __munmap(void *, size_t);
-int __mprotect(void *, size_t, int);
-
static void dummy_0()
{
}
__pthread_tsd_run_dtors();
- LOCK(self->exitlock);
-
- /* Mark this thread dead before decrementing count */
+ /* 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);
- self->dead = 1;
/* Block all signals before decrementing the live thread count.
* This is important to ensure that dynamically allocated TLS
* 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);
}
__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. */
__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);
}
__pthread_self()->cancelbuf = cb->__next;
}
+struct start_args {
+ void *(*start_func)(void *);
+ void *start_arg;
+ pthread_attr_t *attr;
+ volatile int *perr;
+ unsigned long sig_mask[_NSIG/8/sizeof(long)];
+};
+
static int start(void *p)
{
- pthread_t self = p;
- /* States for startlock:
- * 0 = no need for start sync
- * 1 = waiting for parent to do work
- * 2 = failure in parent, child must abort
- * 3 = success in parent, child must restore sigmask */
- if (self->startlock[0]) {
- __wait(self->startlock, 0, 1, 1);
- if (self->startlock[0] == 2) {
- self->detached = 2;
- pthread_exit(0);
+ struct start_args *args = p;
+ if (args->attr) {
+ pthread_t self = __pthread_self();
+ int ret = -__syscall(SYS_sched_setscheduler, self->tid,
+ args->attr->_a_policy, &args->attr->_a_prio);
+ if (a_swap(args->perr, ret)==-2)
+ __wake(args->perr, 1, 1);
+ if (ret) {
+ self->detach_state = DT_DYNAMIC;
+ __pthread_exit(0);
}
- __restore_sigs(self->sigmask);
}
- if (self->unblock_cancel)
- __syscall(SYS_rt_sigprocmask, SIG_UNBLOCK,
- SIGPT_SET, 0, _NSIG/8);
- __pthread_exit(self->start(self->start_arg));
+ __syscall(SYS_rt_sigprocmask, SIG_SETMASK, &args->sig_mask, 0, _NSIG/8);
+ __pthread_exit(args->start_func(args->start_arg));
return 0;
}
static int start_c11(void *p)
{
- pthread_t self = p;
- int (*start)(void*) = (int(*)(void*)) self->start;
- __pthread_exit((void *)(uintptr_t)start(self->start_arg));
+ struct start_args *args = p;
+ int (*start)(void*) = (int(*)(void*)) args->start_func;
+ __pthread_exit((void *)(uintptr_t)start(args->start_arg));
return 0;
}
weak_alias(dummy_tsd, __pthread_tsd_main);
volatile int __block_new_threads = 0;
-size_t __default_stacksize = DEFAULT_STACK_SIZE;
-size_t __default_guardsize = DEFAULT_GUARD_SIZE;
static FILE *volatile dummy_file = 0;
weak_alias(dummy_file, __stdin_used);
if (f && f->lock<0) f->lock = 0;
}
-void *__copy_tls(unsigned char *);
-
int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict attrp, void *(*entry)(void *), void *restrict arg)
{
int ret, c11 = (attrp == __ATTRP_C11_THREAD);
unsigned flags = CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND
| CLONE_THREAD | CLONE_SYSVSEM | CLONE_SETTLS
| CLONE_PARENT_SETTID | CLONE_CHILD_CLEARTID | CLONE_DETACHED;
- int do_sched = 0;
pthread_attr_t attr = { 0 };
+ sigset_t set;
+ volatile int err = -1;
if (!libc.can_do_threads) return ENOSYS;
self = __pthread_self();
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;
- }
- if (attr._a_sched) {
- do_sched = new->startlock[0] = 1;
- __block_app_sigs(new->sigmask);
+ } else {
+ new->detach_state = DT_JOINABLE;
}
new->robust_list.head = &new->robust_list.head;
- new->unblock_cancel = self->cancel;
new->CANARY = self->CANARY;
+ /* Setup argument structure for the new thread on its stack. */
+ stack -= (uintptr_t)stack % sizeof(uintptr_t);
+ stack -= sizeof(struct start_args);
+ struct start_args *args = (void *)stack;
+ args->start_func = entry;
+ args->start_arg = arg;
+ if (attr._a_sched) {
+ args->attr = &attr;
+ args->perr = &err;
+ } else {
+ args->attr = 0;
+ args->perr = 0;
+ }
+
+ __block_app_sigs(&set);
+
+ /* Ensure SIGCANCEL is unblocked in new thread. This requires
+ * working with a copy of the set so we can restore the
+ * original mask in the calling thread. */
+ memcpy(&args->sig_mask, &set, sizeof args->sig_mask);
+ args->sig_mask[(SIGCANCEL-1)/8/sizeof(long)] &=
+ ~(1UL<<((SIGCANCEL-1)%(8*sizeof(long))));
+
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, args, &new->tid, TP_ADJ(new), &new->detach_state);
+ __restore_sigs(&set);
__release_ptc();
- if (do_sched) {
- __restore_sigs(new->sigmask);
- }
-
if (ret < 0) {
a_dec(&libc.threads_minus_1);
if (map) __munmap(map, size);
return EAGAIN;
}
- if (do_sched) {
- ret = __syscall(SYS_sched_setscheduler, new->tid,
- attr._a_policy, &attr._a_prio);
- a_store(new->startlock, ret<0 ? 2 : 3);
- __wake(new->startlock, 1, 1);
- if (ret < 0) return -ret;
+ if (attr._a_sched) {
+ if (a_cas(&err, -1, -2)==-1)
+ __wait(&err, 0, -2, 1);
+ ret = err;
+ if (ret) return ret;
}
*res = new;