#define _GNU_SOURCE
#include "pthread_impl.h"
#include "stdio_impl.h"
#include "libc.h"
#include <sys/mman.h>
#include <string.h>
#include <threads.h>

void *__mmap(void *, size_t, int, int, int, off_t);
int __munmap(void *, size_t);
int __mprotect(void *, size_t, int);
void __thread_enable(void);
_Noreturn void __pthread_exit(void *);
void *__copy_tls(unsigned char *);
extern volatile size_t __pthread_tsd_size;

static void dummy_0()
{
}
weak_alias(dummy_0, __acquire_ptc);
weak_alias(dummy_0, __release_ptc);

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);
	pthread_exit(self->start(self->start_arg));
	return 0;
}

#define ROUND(x) (((x)+PAGE_SIZE-1)&-PAGE_SIZE)

int pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict attrp, void *(*entry)(void *), void *restrict arg)
{
	int ret;
	size_t size, guard = 0;
	struct pthread *self, *new;
	unsigned char *map = 0, *stack = 0, *tsd = 0, *stack_limit;
	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};

	if (!libc.can_do_threads) return ENOSYS;
	self = __pthread_self();
	if (!libc.threaded) __thread_enable();
	if (attrp) attr = *attrp;

	__acquire_ptc();

	if (attr._a_stackaddr) {
		size_t need = libc.tls_size + __pthread_tsd_size;
		size = attr._a_stacksize + DEFAULT_STACK_SIZE;
		stack = (void *)(attr._a_stackaddr & -16);
		stack_limit = (void *)(attr._a_stackaddr - size);
		/* Use application-provided stack for TLS only when
		 * it does not take more than ~12% or 2k of the
		 * application's stack space. */
		if (need < size/8 && need < 2048) {
			tsd = stack - __pthread_tsd_size;
			stack = tsd - libc.tls_size;
			memset(stack, 0, need);
		} else {
			size = ROUND(need);
			guard = 0;
		}
	} else {
		guard = ROUND(DEFAULT_GUARD_SIZE + attr._a_guardsize);
		size = guard + ROUND(DEFAULT_STACK_SIZE + attr._a_stacksize
			+ libc.tls_size +  __pthread_tsd_size);
	}

	if (!tsd) {
		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)) {
				__munmap(map, size);
				goto fail;
			}
		} else {
			map = __mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANON, -1, 0);
			if (map == MAP_FAILED) goto fail;
		}
		tsd = map + size - __pthread_tsd_size;
		if (!stack) {
			stack = tsd - libc.tls_size;
			stack_limit = map + guard;
		}
	}

	new = __copy_tls(tsd - libc.tls_size);
	new->map_base = map;
	new->map_size = size;
	new->stack = stack;
	new->stack_size = stack - stack_limit;
	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;
		flags -= CLONE_CHILD_CLEARTID;
	}
	if (attr._a_sched) {
		do_sched = new->startlock[0] = 1;
		__block_app_sigs(new->sigmask);
	}
	new->unblock_cancel = self->cancel;
	new->canary = self->canary;

	a_inc(&libc.threads_minus_1);
	ret = __clone(start, stack, flags, new, &new->tid, TP_ADJ(new), &new->tid);

	__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 : 0);
		__wake(new->startlock, 1, 1);
		if (ret < 0) return -ret;
	}

	*res = new;
	return 0;
fail:
	__release_ptc();
	return EAGAIN;
}