#ifndef _PTHREAD_IMPL_H
#define _PTHREAD_IMPL_H

#include <pthread.h>
#include <signal.h>
#include <errno.h>
#include <limits.h>
#include "libc.h"
#include "syscall.h"
#include "atomic.h"
#include "futex.h"

#define pthread __pthread

struct pthread {
	struct pthread *self;
	void **dtv, *unused1, *unused2;
	uintptr_t sysinfo;
	uintptr_t canary;
	pid_t tid, pid;
	int tsd_used, errno_val;
	volatile int cancel, canceldisable, cancelasync;
	int detached;
	unsigned char *map_base;
	size_t map_size;
	void *stack;
	size_t stack_size;
	void *start_arg;
	void *(*start)(void *);
	void *result;
	struct __ptcb *cancelbuf;
	void **tsd;
	pthread_attr_t attr;
	volatile int dead;
	struct {
		void **head;
		long off;
		void *pending;
	} robust_list;
	int unblock_cancel;
	int timer_id;
	locale_t locale;
	int killlock[2];
	int exitlock[2];
	int startlock[2];
	unsigned long sigmask[_NSIG/8/sizeof(long)];
};

struct __timer {
	int timerid;
	pthread_t thread;
};

/* For the moment this type has the same layout as
   pthread_mutex_t. This is subject to change in later versions of the
   C thread implementation, be it for the simple reason that this
   would reduce the size of the structure by half. */
struct __mtx_on_heap {
	int _mt_typ;
	int _mt_lck;
	int _mt_wts;
	int _mt_pad3;
	int _mt_pad4;
	int _mt_cnt;
	int _mt_rfs;
	int _mt_pad7;
	int _mt_pad8;
	int _mt_pad9;
};

static inline
void __mtx_addref(struct __mtx_on_heap *m) {
	if (m) {
		a_inc(&m->_mt_rfs);
	}
}

struct __cnd_on_heap {
	int rfs;
	int tok;
	int wts;
	int seq;
	struct __mtx_on_heap* mtx;
};

static inline
void __cnd_addref(struct __cnd_on_heap *c) {
	if (c) {
		a_inc(&c->rfs);
	}
}

/* Unless told otherwise this C thread implementation will only use
   private futexes. */
#ifdef __MTX_SHARED
#define THRD_PRIVATE 0
#else
#define THRD_PRIVATE 128
#endif

#define __SU (sizeof(size_t)/sizeof(int))

#define _a_stacksize __u.__s[0]
#define _a_guardsize __u.__s[1]
#define _a_stackaddr __u.__s[2]
#define _a_detach __u.__i[3*__SU+0]
#define _a_sched __u.__i[3*__SU+1]
#define _a_policy __u.__i[3*__SU+2]
#define _a_prio __u.__i[3*__SU+3]
#define _m_type __u.__i[0]
#define _m_lock __u.__i[1]
#define _m_waiters __u.__i[2]
#define _m_prev __u.__p[3]
#define _m_next __u.__p[4]
#define _m_count __u.__i[5]
#define _mx_type __u.__i[0]
#define _mx_lock __u.__i[1]
#define _mx_mtx __u.__p[2]
#define _mx_prev __u.__p[3] /* unused */
#define _mx_next __u.__p[4] /* unused */
#define _mx_count __u.__i[5] /* unused */
#define _c_mutex __u.__p[0]
#define _c_seq __u.__i[2]
#define _c_waiters __u.__i[3]
#define _c_clock __u.__i[4]
#define _c_lock __u.__i[5]
#define _c_lockwait __u.__i[6]
#define _c_waiters2 __u.__i[7]
#define _c_destroy __u.__i[8]
#define _cx_cnd __u.__p[0]
#define _cx_seq __u.__i[2]     /* unused */
#define _cx_waiters __u.__i[3] /* unused */
#define _cx_clock __u.__i[4]   /* unused */
#define _cx_lock __u.__i[5]
#define _cx_lockwait __u.__i[6]
#define _cx_waiters2 __u.__i[7]/* unused */
#define _cx_destroy __u.__i[8] /* unused */
#define _rw_lock __u.__i[0]
#define _rw_waiters __u.__i[1]
#define _b_lock __u.__i[0]
#define _b_waiters __u.__i[1]
#define _b_limit __u.__i[2]
#define _b_count __u.__i[3]
#define _b_waiters2 __u.__i[4]
#define _b_inst __u.__p[3]

#if defined(__ATOMIC_ACQ_REL) && defined(__ATOMIC_ACQUIRE) && defined(__ATOMIC_RELEASE)
#define a_tas(X) __atomic_test_and_set((X), __ATOMIC_ACQ_REL)
#define a_aqf() __atomic_thread_fence(__ATOMIC_ACQUIRE)
#define a_spunl(X)                              \
({                                              \
	__asm__ volatile ("# end CS");          \
	int volatile* _lck = (X);               \
	__atomic_clear(_lck, __ATOMIC_RELEASE); \
})
#else
#define a_tas(X) __sync_lock_test_and_set((X), 1)
#define a_aqf() __sync_synchronize()
#define a_spunl(X)                              \
({                                              \
	__asm__ volatile ("# end CS");          \
	int volatile* _lck = (X);               \
	__sync_lock_release(_lck);              \
})
#endif

#define a_splck(X)                              \
({                                              \
  int volatile* _lck = (X);                     \
  while (a_tas(_lck)) (void)0;                  \
  a_aqf();                                      \
  __asm__ volatile ("# start CS");              \
 })

#include "pthread_arch.h"

#define SIGTIMER 32
#define SIGCANCEL 33
#define SIGSYNCCALL 34

#define SIGALL_SET ((sigset_t *)(const unsigned long long [2]){ -1,-1 })
#define SIGPT_SET \
	((sigset_t *)(const unsigned long [_NSIG/8/sizeof(long)]){ \
	[sizeof(long)==4] = 3UL<<(32*(sizeof(long)>4)) })
#define SIGTIMER_SET \
	((sigset_t *)(const unsigned long [_NSIG/8/sizeof(long)]){ \
	 0x80000000 })

pthread_t __pthread_self_init(void);

int __clone(int (*)(void *), void *, int, void *, ...);
int __set_thread_area(void *);
int __libc_sigaction(int, const struct sigaction *, struct sigaction *);
int __libc_sigprocmask(int, const sigset_t *, sigset_t *);
void __lock(volatile int *);
void __unmapself(void *, size_t);

int __timedwait(volatile int *, int, clockid_t, const struct timespec *, void (*)(void *), void *, int);
void __wait(volatile int *, volatile int *, int, int);
#define __wake(addr, cnt, priv) \
	__syscall(SYS_futex, addr, FUTEX_WAKE, (cnt)<0?INT_MAX:(cnt))

void __wait_priv(volatile int *, volatile int *, int);
#define __wake_priv(addr, cnt) \
	__syscall(SYS_futex, addr, FUTEX_WAKE|THRD_PRIVATE, (cnt)<0?INT_MAX:(cnt))

void __acquire_ptc();
void __release_ptc();
void __inhibit_ptc();

void __block_all_sigs(void *);
void __block_app_sigs(void *);
void __restore_sigs(void *);

#define DEFAULT_STACK_SIZE 81920
#define DEFAULT_GUARD_SIZE PAGE_SIZE

#endif