/*

 * Portions Copyright (c) 2008 Nokia Corporation and/or its subsidiary(-ies). All rights reserved.

 *

 * Copyright (c) 1996,1997

 * Silicon Graphics Computer Systems, Inc.

 *

 * Copyright (c) 1997

 * Moscow Center for SPARC Technology

 *

 * Copyright (c) 1999

 * Boris Fomitchev

 *

 * This material is provided "as is", with absolutely no warranty expressed

 * or implied. Any use is at your own risk.

 *

 * Permission to use or copy this software for any purpose is hereby granted

 * without fee, provided the above notices are retained on all copies.

 * Permission to modify the code and to distribute modified code is granted,

 * provided the above notices are retained, and a notice that the code was

 * modified is included with the above copyright notice.

 *

 */

#include "stlport_prefix.h"

#include <memory>

#if defined (__GNUC__) && (defined (__CYGWIN__) || defined (__MINGW32__)) && (!defined (__SYMBIAN32__))

#  include <malloc.h>

//#  define _STLP_MALLOC_USABLE_SIZE(__buf) malloc_usable_size(__buf)

#endif

#if defined (_STLP_PTHREADS) && !defined (_STLP_NO_THREADS)

#  include <pthread_alloc>

#  include <cerrno>

#endif

#include <stl/_threads.h>

#include "lock_free_slist.h"

#if defined(__SYMBIAN32__WSD__)

#include "libstdcppwsd.h"

#define __oom_handler 	get_oom_handler()

#define _S_lock 		get_allocator_S_lock()

#define _S_heap_size 	get_S_heap_size()

#define _S_start_free 	get_S_start_free()

#define _S_end_free		get_S_end_free()

#define _S_free_list 	get_S_free_list()

#define _S_chunk_allocator_lock 	get_S_chunk_allocator_lock()

#define _S_free_per_thread_states	get_S_free_per_thread_states()

#define _S_key						get_S_key()

#define _S_key_initialized			get_S_key_initialized()

#endif

#if defined (__WATCOMC__)

#  pragma warning 13 9

#  pragma warning 367 9

#  pragma warning 368 9

#endif

#if defined (_STLP_SGI_THREADS)

  // We test whether threads are in use before locking.

  // Perhaps this should be moved into stl_threads.h, but that

  // probably makes it harder to avoid the procedure call when

  // it isn't needed.

extern "C" {

  extern int __us_rsthread_malloc;

}

#endif

// Specialised debug form of malloc which does not provide "false"

// memory leaks when run with debug CRT libraries.

#if defined (_STLP_MSVC) && (_STLP_MSVC >= 1020 && defined (_STLP_DEBUG_ALLOC)) && !defined (_STLP_WCE)

#  include <crtdbg.h>

inline void* __stlp_chunk_malloc(size_t __bytes) { _STLP_CHECK_NULL_ALLOC(_malloc_dbg(__bytes, _CRT_BLOCK, __FILE__, __LINE__)); }

inline void __stlp_chunck_free(void* __p) { _free_dbg(__p, _CRT_BLOCK); }

#else  // !_DEBUG

#  ifdef _STLP_NODE_ALLOC_USE_MALLOC

#    include <cstdlib>

inline void* __stlp_chunk_malloc(size_t __bytes) { _STLP_CHECK_NULL_ALLOC(_STLP_VENDOR_CSTD::malloc(__bytes)); }

inline void __stlp_chunck_free(void* __p) { _STLP_VENDOR_CSTD::free(__p); }

#  else

#  endif

#endif  // !_DEBUG

#define _S_FREELIST_INDEX(__bytes) ((__bytes - size_t(1)) >> (int)_ALIGN_SHIFT)

_STLP_BEGIN_NAMESPACE

class __malloc_alloc_impl {

private:

  static void* _S_oom_malloc(size_t __n) {

    __oom_handler_type __my_malloc_handler;

    void * __result;

    for (;;) {

      __my_malloc_handler = __oom_handler;

      if (0 == __my_malloc_handler) { __THROW_BAD_ALLOC; }

      (*__my_malloc_handler)();

      __result = malloc(__n);

      if (__result) return(__result);

    }

#if defined (_STLP_NEED_UNREACHABLE_RETURN)

    return 0;

#endif

  }

#if defined(__SYMBIAN32__WSD__)  

  static _STLP_STATIC_MEMBER_DECLSPEC __oom_handler_type& get_oom_handler();

#else

  static __oom_handler_type __oom_handler;

#endif

public:

  // this one is needed for proper simple_alloc wrapping

  typedef char value_type;

  static void* allocate(size_t& __n) {

    void* __result = malloc(__n);

    if (0 == __result) {

      __result = _S_oom_malloc(__n);

    }

#if defined (_STLP_MALLOC_USABLE_SIZE)

    else {

      size_t __new_n = _STLP_MALLOC_USABLE_SIZE(__result);

      /*

      if (__n != __new_n) {

        printf("requested size %d, usable %d\n", __n, __new_n);

      }

      */

      __n = __new_n;

    }

#endif

    return __result;

  }

  static void deallocate(void* __p, size_t /* __n */) { free((char*)__p); }

  static __oom_handler_type set_malloc_handler(__oom_handler_type __f) {

    __oom_handler_type __old = __oom_handler;

    __oom_handler = __f;

    return __old;

  }

#if defined(__SYMBIAN32__WSD__)  

  friend void ::stdcpp_allocators_init();

#endif

};

#if !defined(__SYMBIAN32__WSD__)

// malloc_alloc out-of-memory handling

__oom_handler_type __malloc_alloc_impl::__oom_handler = __STATIC_CAST(__oom_handler_type, 0);

#endif

void* _STLP_CALL __malloc_alloc::allocate(size_t& __n)

{ return __malloc_alloc_impl::allocate(__n); }

__oom_handler_type _STLP_CALL __malloc_alloc::set_malloc_handler(__oom_handler_type __f)

{ return __malloc_alloc_impl::set_malloc_handler(__f); }

// *******************************************************

// Default node allocator.

// With a reasonable compiler, this should be roughly as fast as the

// original STL class-specific allocators, but with less fragmentation.

//

// Important implementation properties:

// 1. If the client request an object of size > _MAX_BYTES, the resulting

//    object will be obtained directly from malloc.

// 2. In all other cases, we allocate an object of size exactly

//    _S_round_up(requested_size).  Thus the client has enough size

//    information that we can return the object to the proper free list

//    without permanently losing part of the object.

//

#define _STLP_NFREELISTS 16

/*

 * We can only do cleanup of the node allocator memory pool if we are

 * sure that the STLport library is used as a shared one as it guaranties

 * the unicity of the node allocator instance. Without that guaranty node

 * allocator instances might exchange memory blocks making the implementation

 * of a cleaning process much more complicated.

 */

#  define _STLP_DO_CLEAN_NODE_ALLOC

#endif

/* When STLport is used without multi threaded safety we use the node allocator

 * implementation with locks as locks becomes no-op. The lock free implementation

 * always use system specific atomic operations which are slower than 'normal'

 * ones.

 */

#if defined (_STLP_THREADS) && \

    defined (_STLP_HAS_ATOMIC_FREELIST) && defined (_STLP_ATOMIC_ADD)

/*

 * We have an implementation of the atomic freelist (_STLP_atomic_freelist)

 * for this architecture and compiler.  That means we can use the non-blocking

 * implementation of the node-allocation engine.*/

#  define _STLP_USE_LOCK_FREE_IMPLEMENTATION

#endif

#if !defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

#  if defined (_STLP_THREADS)

class _Node_Alloc_Lock {

public:

  _Node_Alloc_Lock() {

#  if defined (_STLP_SGI_THREADS)

    if (__us_rsthread_malloc)

#  endif

      _S_lock._M_acquire_lock();

  }

  ~_Node_Alloc_Lock() {

#  if defined (_STLP_SGI_THREADS)

    if (__us_rsthread_malloc)

#  endif

        _S_lock._M_release_lock();

  }

#if defined (__SYMBIAN32__WSD__)

  static _STLP_STATIC_MUTEX& get_allocator_S_lock();

#else

  static _STLP_STATIC_MUTEX _S_lock;

#endif

};

#if !defined(__SYMBIAN32__WSD__)

_STLP_STATIC_MUTEX _Node_Alloc_Lock::_S_lock _STLP_MUTEX_INITIALIZER;

#endif

#  else

class _Node_Alloc_Lock {

public:

  _Node_Alloc_Lock() { }

  ~_Node_Alloc_Lock() { }

};

#  endif

struct _Node_alloc_obj {

  _Node_alloc_obj * _M_next;

};

#endif

class __node_alloc_impl {

_STLP_PRIVATE:

  static inline size_t _STLP_CALL _S_round_up(size_t __bytes)

  { return (((__bytes) + (size_t)_ALIGN-1) & ~((size_t)_ALIGN - 1)); }

#if defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

  typedef _STLP_atomic_freelist::item   _Obj;

  typedef _STLP_atomic_freelist         _Freelist;

  typedef _STLP_atomic_freelist         _ChunkList;

  // Header of blocks of memory that have been allocated as part of

  // a larger chunk but have not yet been chopped up into nodes.

  struct _FreeBlockHeader : public _STLP_atomic_freelist::item {

    char* _M_end;     // pointer to end of free memory

  };

#else

  typedef _Node_alloc_obj       _Obj;

  typedef _Obj* _STLP_VOLATILE  _Freelist;

  typedef _Obj*                 _ChunkList;

#endif

private:

  // Returns an object of size __n, and optionally adds to size __n free list.

  static _Obj* _S_refill(size_t __n);

  // Allocates a chunk for nobjs of size __p_size.  nobjs may be reduced

  // if it is inconvenient to allocate the requested number.

  static char* _S_chunk_alloc(size_t __p_size, int& __nobjs);

  // Chunk allocation state.

#if defined(__SYMBIAN32__WSD__)  

  static _Freelist* get_S_free_list();

#else

  static _Freelist _S_free_list[_STLP_NFREELISTS];

#endif

  // Amount of total allocated memory

#if defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

  static _STLP_VOLATILE __stl_atomic_t _S_heap_size;

#else

#if defined(__SYMBIAN32__WSD__)

  static size_t& get_S_heap_size();

#else

  static size_t _S_heap_size;

#endif

#endif

#if defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

  // List of blocks of free memory

  static _STLP_atomic_freelist  _S_free_mem_blocks;

#else

#if defined(__SYMBIAN32__WSD__)

  // Start of the current free memory buffer

  static char*& get_S_start_free();

  // End of the current free memory buffer

  static char*& get_S_end_free();

#else

  // Start of the current free memory buffer

  static char* _S_start_free;

  // End of the current free memory buffer

  static char* _S_end_free;

#endif

#endif

#if defined (_STLP_DO_CLEAN_NODE_ALLOC)

public:

  // Methods to report alloc/dealloc calls to the counter system.

#  if defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

  typedef _STLP_VOLATILE __stl_atomic_t _AllocCounter;

#  else

  typedef __stl_atomic_t _AllocCounter;

#  endif

  static _AllocCounter& _STLP_CALL _S_alloc_counter();

  static void _S_alloc_call();

  static void _S_dealloc_call();

private:

  // Free all the allocated chuncks of memory

  static void _S_chunk_dealloc();

  // Beginning of the linked list of allocated chunks of memory

  static _ChunkList _S_chunks;

#endif /* _STLP_DO_CLEAN_NODE_ALLOC */

public:

  /* __n must be > 0      */

  static void* _M_allocate(size_t& __n);

  /* __p may not be 0 */

  static void _M_deallocate(void *__p, size_t __n);

#if defined(__SYMBIAN32__WSD__)

	friend void ::stdcpp_allocators_init();

#endif

};

#if !defined (_STLP_USE_LOCK_FREE_IMPLEMENTATION)

void* __node_alloc_impl::_M_allocate(size_t& __n) {

  __n = _S_round_up(__n);

  _Obj * _STLP_VOLATILE * __my_free_list = _S_free_list + _S_FREELIST_INDEX(__n);

  _Obj *__r;

  // Acquire the lock here with a constructor call.

  // This ensures that it is released in exit or during stack

  // unwinding.

  _Node_Alloc_Lock __lock_instance;

  if ( (__r  = *__my_free_list) != 0 ) {

    *__my_free_list = __r->_M_next;

  } else {

    __r = _S_refill(__n);

  }

#  if defined (_STLP_DO_CLEAN_NODE_ALLOC)

  _S_alloc_call();

#  endif

  // lock is released here

  return __r;

}

void __node_alloc_impl::_M_deallocate(void *__p, size_t __n) {

  _Obj * _STLP_VOLATILE * __my_free_list = _S_free_list + _S_FREELIST_INDEX(__n);

  _Obj * __pobj = __STATIC_CAST(_Obj*, __p);

  // acquire lock

  _Node_Alloc_Lock __lock_instance;

  __pobj->_M_next = *__my_free_list;

  *__my_free_list = __pobj;

#  if defined (_STLP_DO_CLEAN_NODE_ALLOC)

  _S_dealloc_call();

#  endif

  // lock is released here

}

/* We allocate memory in large chunks in order to avoid fragmenting     */

/* the malloc heap too much.                                            */

/* We assume that size is properly aligned.                             */

/* We hold the allocation lock.                                         */

char* __node_alloc_impl::_S_chunk_alloc(size_t _p_size, int& __nobjs) {

  char* __result;

  size_t __total_bytes = _p_size * __nobjs;

  size_t __bytes_left = _S_end_free - _S_start_free;

  if (__bytes_left > 0) {

    if (__bytes_left >= __total_bytes) {

      __result = _S_start_free;

      _S_start_free += __total_bytes;

      return __result;

    }

    if (__bytes_left >= _p_size) {

      __nobjs = (int)(__bytes_left / _p_size);

      __total_bytes = _p_size * __nobjs;

      __result = _S_start_free;

      _S_start_free += __total_bytes;

      return __result;

    }

    // Try to make use of the left-over piece.

    _Obj* _STLP_VOLATILE* __my_free_list = _S_free_list + _S_FREELIST_INDEX(__bytes_left);

    __REINTERPRET_CAST(_Obj*, _S_start_free)->_M_next = *__my_free_list;

    *__my_free_list = __REINTERPRET_CAST(_Obj*, _S_start_free);

  }

  size_t __bytes_to_get =

    2 * __total_bytes + _S_round_up(_S_heap_size >> 4)

#  if defined (_STLP_DO_CLEAN_NODE_ALLOC)

    + sizeof(_Obj)

#  endif

    ;

  _S_start_free = __STATIC_CAST(char*, __stlp_chunk_malloc(__bytes_to_get));

  if (0 == _S_start_free) {

    _Obj* _STLP_VOLATILE* __my_free_list;

    _Obj* __p;

    // Try to do with what we have.  That can't hurt.

    // We do not try smaller requests, since that tends

    // to result in disaster on multi-process machines.

    for (size_t __i = _p_size; __i <= (size_t)_MAX_BYTES; __i += (size_t)_ALIGN) {

      __my_free_list = _S_free_list + _S_FREELIST_INDEX(__i);

      __p = *__my_free_list;

      if (0 != __p) {

        *__my_free_list = __p -> _M_next;

        _S_start_free = __REINTERPRET_CAST(char*, __p);

        _S_end_free = _S_start_free + __i;

        return _S_chunk_alloc(_p_size, __nobjs);

        // Any leftover piece will eventually make it to the

        // right free list.

      }

    }

    _S_end_free = 0;    // In case of exception.

    _S_start_free = __STATIC_CAST(char*, __stlp_chunk_malloc(__bytes_to_get));

    /*

    (char*)malloc_alloc::allocate(__bytes_to_get);

    */

    // This should either throw an

    // exception or remedy the situation.  Thus we assume it

    // succeeded.

  }

  _S_heap_size += __bytes_to_get;

#  if defined (_STLP_DO_CLEAN_NODE_ALLOC)

  __REINTERPRET_CAST(_Obj*, _S_start_free)->_M_next = _S_chunks;

  _S_chunks = __REINTERPRET_CAST(_Obj*, _S_start_free);

#  endif

  _S_end_free = _S_start_free + __bytes_to_get;

#  if defined (_STLP_DO_CLEAN_NODE_ALLOC)

  _S_start_free += sizeof(_Obj);

#  endif

  return _S_chunk_alloc(_p_size, __nobjs);

}

/* Returns an object of size __n, and optionally adds to size __n free list.*/

/* We assume that __n is properly aligned.                                  */

/* We hold the allocation lock.                                             */

_Node_alloc_obj* __node_alloc_impl::_S_refill(size_t __n) {

  int __nobjs = 20;

  char* __chunk = _S_chunk_alloc(__n, __nobjs);

  if (1 == __nobjs) return __REINTERPRET_CAST(_Obj*, __chunk);

  _Obj* _STLP_VOLATILE* __my_free_list = _S_free_list + _S_FREELIST_INDEX(__n);

  _Obj* __result;

  _Obj* __current_obj;

  _Obj* __next_obj;

  /* Build free list in chunk */

  __result = __REINTERPRET_CAST(_Obj*, __chunk);

  *__my_free_list = __next_obj = __REINTERPRET_CAST(_Obj*, __chunk + __n);

  for (--__nobjs; --__nobjs; ) {

    __current_obj = __next_obj;

    __next_obj = __REINTERPRET_CAST(_Obj*, __REINTERPRET_CAST(char*, __next_obj) + __n);

    __current_obj->_M_next = __next_obj;