/* newlib_syscalls.c - newlib syscalls for ESP8266
 *
 * Part of esp-open-rtos
 * Copyright (C) 2105 Superhouse Automation Pty Ltd
 * BSD Licensed as described in the file LICENSE
 */
#include <sys/reent.h>
#include <sys/types.h>
#include <sys/errno.h>
#include <espressif/sdk_private.h>
#include <common_macros.h>
#include <xtensa_ops.h>
#include <esp/uart.h>
#include <stdlib.h>
#include <stdout_redirect.h>
#include <sys/time.h>
#include <lwip/sockets.h>
#include <sys/lock.h>
#include <FreeRTOS.h>
#include <semphr.h>
#include <esp/hwrand.h>

/*
 * The file descriptor index space is allocated in blocks. The first block of 3
 * is for newlib I/O the stdin stdout and stderr. The next block of
 * MEMP_NUM_NETCONN is allocated for lwip sockets, and the remainer to file
 * system descriptors. The newlib default FD_SETSIZE is 64.
 */
#if LWIP_SOCKET_OFFSET < 3
#error Expecting a LWIP_SOCKET_OFFSET >= 3, to allow room for the standard I/O descriptors.
#endif
#define FILE_DESCRIPTOR_OFFSET (LWIP_SOCKET_OFFSET + MEMP_NUM_NETCONN)
#if FILE_DESCRIPTOR_OFFSET > FD_SETSIZE
#error Too many lwip sockets for the FD_SETSIZE.
#endif

extern void *xPortSupervisorStackPointer;

IRAM void *_sbrk_r (struct _reent *r, ptrdiff_t incr)
{
    extern char   _heap_start; /* linker script defined */
    static char * heap_end;
    char *        prev_heap_end;

    if (heap_end == NULL)
	heap_end = &_heap_start;
    prev_heap_end = heap_end;

    intptr_t sp = (intptr_t)xPortSupervisorStackPointer;
    if(sp == 0) /* scheduler not started */
        SP(sp);

    if ((intptr_t)heap_end + incr >= sp)
    {
        r->_errno = ENOMEM;
        return (caddr_t)-1;
    }

    heap_end += incr;

    return (caddr_t) prev_heap_end;
}


/* Insert a disjoint region into the nano malloc pool. Create a malloc chunk,
 * filling the size as newlib nano malloc expects, and then free it. */
void nano_malloc_insert_chunk(void *start, size_t size) {
    *(uint32_t *)start = size;
    free(start + sizeof(size_t));
}

/* syscall implementation for stdio write to UART */
__attribute__((weak)) ssize_t _write_stdout_r(struct _reent *r, int fd, const void *ptr, size_t len )
{
    for(int i = 0; i < len; i++) {
        /* Auto convert CR to CRLF, ignore other LFs (compatible with Espressif SDK behaviour) */
        if(((char *)ptr)[i] == '\r')
            continue;
        if(((char *)ptr)[i] == '\n')
            uart_putc(0, '\r');
        uart_putc(0, ((char *)ptr)[i]);
    }
    return len;
}

static _WriteFunction *current_stdout_write_r = &_write_stdout_r;

void set_write_stdout(_WriteFunction *f)
{
    if  (f != NULL) {
        current_stdout_write_r = f;
    } else {
        current_stdout_write_r = &_write_stdout_r;
    }
}

_WriteFunction *get_write_stdout()
{
    return current_stdout_write_r;
}

/* default implementation, replace in a filesystem */
__attribute__((weak)) ssize_t _write_filesystem_r(struct _reent *r, int fd, const void *ptr, size_t len)
{
    r->_errno = EBADF;
    return -1;
}

__attribute__((weak)) ssize_t _write_r(struct _reent *r, int fd, const void *ptr, size_t len)
{
    if (fd >= FILE_DESCRIPTOR_OFFSET) {
        return _write_filesystem_r(r, fd, ptr, len);
    }
    if (fd >= LWIP_SOCKET_OFFSET) {
        return lwip_write(fd, ptr, len);
    }
    if (fd == r->_stdout->_file) {
        return current_stdout_write_r(r, fd, ptr, len);
    }
    r->_errno = EBADF;
    return -1;
}

/* syscall implementation for stdio read from UART */
__attribute__((weak)) ssize_t _read_stdin_r(struct _reent *r, int fd, void *ptr, size_t len)
{
    int ch, i;
    uart_rxfifo_wait(0, 1);
    for(i = 0; i < len; i++) {
        ch = uart_getc_nowait(0);
        if (ch < 0) break;
        ((char *)ptr)[i] = ch;
    }
    return i;
}

/* default implementation, replace in a filesystem */
__attribute__((weak)) ssize_t _read_filesystem_r( struct _reent *r, int fd, void *ptr, size_t len )
{
    r->_errno = EBADF;
    return -1;
}

__attribute__((weak)) ssize_t _read_r( struct _reent *r, int fd, void *ptr, size_t len )
{
    if (fd >= FILE_DESCRIPTOR_OFFSET) {
        return _read_filesystem_r(r, fd, ptr, len);
    }
    if (fd >= LWIP_SOCKET_OFFSET) {
        return lwip_read(fd, ptr, len);
    }
    if (fd == r->_stdin->_file) {
        return _read_stdin_r(r, fd, ptr, len);
    }
    r->_errno = EBADF;
    return -1;
}

/* default implementation, replace in a filesystem */
__attribute__((weak)) int _close_filesystem_r(struct _reent *r, int fd)
{
    r->_errno = EBADF;
    return -1;
}

__attribute__((weak)) int _close_r(struct _reent *r, int fd)
{
    if (fd >= FILE_DESCRIPTOR_OFFSET) {
        return _close_filesystem_r(r, fd);
    }
    if (fd >= LWIP_SOCKET_OFFSET) {
        return lwip_close(fd);
    }
    r->_errno = EBADF;
    return -1;
}

/* Stub syscall implementations follow, to allow compiling newlib functions that
   pull these in via various codepaths
*/
__attribute__((weak, alias("syscall_returns_enosys"))) 
int _open_r(struct _reent *r, const char *pathname, int flags, int mode);

__attribute__((weak, alias("syscall_returns_enosys"))) 
int _unlink_r(struct _reent *r, const char *path);

__attribute__((weak, alias("syscall_returns_enosys"))) 
int _fstat_r(struct _reent *r, int fd, struct stat *buf);

__attribute__((weak, alias("syscall_returns_enosys"))) 
int _stat_r(struct _reent *r, const char *pathname, struct stat *buf);

__attribute__((weak, alias("syscall_returns_enosys"))) 
off_t _lseek_r(struct _reent *r, int fd, off_t offset, int whence);

__attribute__((weak, alias("_gettimeofday_r")))
int _gettimeofday_r _PARAMS ((struct _reent *r, struct timeval *now, void *p)) {
  now->tv_sec = 0;
  now->tv_usec = 0;
  errno = ENOSYS;
  return -1;
}

/* Generic stub for any newlib syscall that fails with errno ENOSYS
   ("Function not implemented") and a return value equivalent to
   (int)-1. */
static int syscall_returns_enosys(struct _reent *r)
{
    r->_errno=ENOSYS;
    return -1;
}

int getentropy(void *ptr, size_t n)
{
    hwrand_fill(ptr, n);
    return 0;
}

void _arc4random_getentropy_fail(void)
{
}

void _exit(int status)
{
    while(1);
}

/*
 * Newlib lock implementation. Some newlib locks are statically allocated, but
 * can not be statically initialized so are set to NULL and initialized at
 * startup. The malloc lock is used before it can be initialized so there are
 * runtime checks on the functions that use it early.
 */
static int locks_initialized = 0;

extern _lock_t __arc4random_mutex;
extern _lock_t __at_quick_exit_mutex;
//extern _lock_t __dd_hash_mutex;
extern _lock_t __tz_mutex;

extern _lock_t __atexit_recursive_mutex;
extern _lock_t __env_recursive_mutex;
extern _lock_t __malloc_recursive_mutex;
extern _lock_t __sfp_recursive_mutex;
extern _lock_t __sinit_recursive_mutex;

void init_newlib_locks()
{
    _lock_init(&__arc4random_mutex);

#if 0
    /* Separate mutex for each lock.
     * Each mutex uses about 96 bytes which adds up. */
    _lock_init(&__at_quick_exit_mutex);
    //_lock_init(&__dd_hash_mutex);
    _lock_init(&__tz_mutex);

    _lock_init_recursive(&__atexit_recursive_mutex);
    _lock_init_recursive(&__env_recursive_mutex);
    _lock_init_recursive(&__malloc_recursive_mutex);
    _lock_init_recursive(&__sfp_recursive_mutex);
    _lock_init_recursive(&__sinit_recursive_mutex);
#else
    /* Reuse the same mutex for all these, reducing memory usage. Newlib
     * will still allocate other locks dynamically and some of those need
     * to be separate such as the file lock where a thread might block with
     * them held. */
    __at_quick_exit_mutex = __arc4random_mutex;
    //__dd_hash_mutex = __arc4random_mutex;
    __tz_mutex = __arc4random_mutex;

    __atexit_recursive_mutex = __arc4random_mutex;
    __env_recursive_mutex  = __arc4random_mutex;
    __malloc_recursive_mutex = __arc4random_mutex;
    __sfp_recursive_mutex = __arc4random_mutex;
    __sinit_recursive_mutex = __arc4random_mutex;
#endif

    locks_initialized = 1;
}

void _lock_init(_lock_t *lock) {
    *lock = (_lock_t)xSemaphoreCreateMutex();
}

void _lock_init_recursive(_lock_t *lock) {
    *lock = (_lock_t)xSemaphoreCreateRecursiveMutex();
}

void _lock_close(_lock_t *lock) {
    vSemaphoreDelete((QueueHandle_t)*lock);
    *lock = 0;
}

void _lock_close_recursive(_lock_t *lock) {
    vSemaphoreDelete((QueueHandle_t)*lock);
    *lock = 0;
}

void _lock_acquire(_lock_t *lock) {
    xSemaphoreTake((QueueHandle_t)*lock, portMAX_DELAY);
}

void _lock_acquire_recursive(_lock_t *lock) {
    if (locks_initialized) {
        xSemaphoreTakeRecursive((QueueHandle_t)*lock, portMAX_DELAY);
    }
}

int _lock_try_acquire(_lock_t *lock) {
    return xSemaphoreTake((QueueHandle_t)*lock, 0);
}

int _lock_try_acquire_recursive(_lock_t *lock) {
    return xSemaphoreTakeRecursive((QueueHandle_t)*lock, 0);
}

void _lock_release(_lock_t *lock) {
    xSemaphoreGive((QueueHandle_t)*lock);
}

void _lock_release_recursive(_lock_t *lock) {
    if (locks_initialized) {
        xSemaphoreGiveRecursive((QueueHandle_t)*lock);
    }
}