arean.c

glibc-2.14中的arean.c源代码,供研究malloc和free实现使用:

 /* Malloc implementation for multiple threads without lock contention.
  
 			   Copyright (C) 2001,2002,2003,2004,2005,2006,2007,2009,2010
 
 			   Free Software Foundation, Inc.
 
 			   This file is part of the GNU C Library.
 
 			   Contributed by Wolfram Gloger <wg@malloc.de>, 2001.
 
 
 
 			   The GNU C Library is free software; you can redistribute it and/or
 
 			   modify it under the terms of the GNU Lesser General Public License as
 
 			   published by the Free Software Foundation; either version 2.1 of the
 
 			   License, or (at your option) any later version.
 
 
 
 			   The GNU C Library is distributed in the hope that it will be useful,
 
 			   but WITHOUT ANY WARRANTY; without even the implied warranty of
 
 			   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 
 			   Lesser General Public License for more details.
 
 
 
 			   You should have received a copy of the GNU Lesser General Public
 
 			   License along with the GNU C Library; see the file COPYING.LIB. If not,
 
 			   write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 
 			   Boston, MA 02111-1307, USA. */
 
 
 
 			#include <stdbool.h>
 
 
 
 /* Compile-time constants. */
 
 
 
 			#define HEAP_MIN_SIZE (32*1024)
 
 			#ifndef HEAP_MAX_SIZE
 
 			# ifdef DEFAULT_MMAP_THRESHOLD_MAX
 
 			#  define HEAP_MAX_SIZE (2 * DEFAULT_MMAP_THRESHOLD_MAX)
 
 			# else
 
 			#  define HEAP_MAX_SIZE (1024*1024) /* must be a power of two */
 
 			# endif
 
 			#endif
 
 
 
 /* HEAP_MIN_SIZE and HEAP_MAX_SIZE limit the size of mmap()ed heaps
 
 			   that are dynamically created for multi-threaded programs. The
 
 			   maximum size must be a power of two, for fast determination of
 
 			   which heap belongs to a chunk. It should be much larger than the
 
 			   mmap threshold, so that requests with a size just below that
 
 			   threshold can be fulfilled without creating too many heaps. */
 
 
 
 
 
 			#ifndef THREAD_STATS
 
 			#define THREAD_STATS 0
 
 			#endif
 
 
 
 /* If THREAD_STATS is non-zero, some statistics on mutex locking are
 
 			   computed. */
 
 
 
 /***************************************************************************/
 
 
 
 			#define top(ar_ptr) ((ar_ptr)->top)
 
 
 
 /* A heap is a single contiguous memory region holding (coalesceable)
 
 			   malloc_chunks. It is allocated with mmap() and always starts at an
 
 			   address aligned to HEAP_MAX_SIZE. Not used unless compiling with
 
 			   USE_ARENAS. */
 
 
 
 			typedef struct _heap_info {
 
 			  mstate ar_ptr; /* Arena for this heap. */
 
 			  struct _heap_info *prev; /* Previous heap. */
 
 			  size_t size; /* Current size in bytes. */
 
 			  size_t mprotect_size; /* Size in bytes that has been mprotected
 
 			               PROT_READ|PROT_WRITE. */
 
 /* Make sure the following data is properly aligned, particularly
 
 			     that sizeof (heap_info) + 2 * SIZE_SZ is a multiple of
 
 			     MALLOC_ALIGNMENT. */
 
 			  char pad[-6 * SIZE_SZ & MALLOC_ALIGN_MASK];
 
 } heap_info;
 
 
 
 /* Get a compile-time error if the heap_info padding is not correct
 
 to make alignment work as expected in sYSMALLOc. */
 
 			extern int sanity_check_heap_info_alignment[(sizeof (heap_info)
 
 + 2 * SIZE_SZ) % MALLOC_ALIGNMENT
 
 ? -1 : 1];
 
 
 
 /* Thread specific data */
 
 
 
 			static tsd_key_t arena_key;
 
 			static mutex_t list_lock;
 
 			#ifdef PER_THREAD
 
 			static size_t narenas;
 
 			static mstate free_list;
 
 			#endif
 
 
 
 			#if THREAD_STATS
 
 			static int stat_n_heaps;
 
 			#define THREAD_STAT(x) x
 
 			#else
 
 			#define THREAD_STAT(x) do ; while(0)
 
 			#endif
 
 
 
 /* Mapped memory in non-main arenas (reliable only for NO_THREADS). */
 
 			static unsigned long arena_mem;
 
 
 
 /* Already initialized? */
 
 int __malloc_initialized = -1;
 
 
 
 /**************************************************************************/
 
 
 
 			#if USE_ARENAS
 
 
 
 /* arena_get() acquires an arena and locks the corresponding mutex.
 
 			   First, try the one last locked successfully by this thread. (This
 
 is the common case and handled with a macro for speed.) Then, loop
 
 			   once over the circularly linked list of arenas. If no arena is
 
 			   readily available, create a new one. In this latter case, `size'
 
 is just a hint as to how much memory will be required immediately
 
 in the new arena. */
 
 
 
 			#define arena_get(ptr, size) do { \
 
 			  arena_lookup(ptr); \
 
 			  arena_lock(ptr, size); \
 
 } while(0)
 
 
 
 			#define arena_lookup(ptr) do { \
 
 			  Void_t *vptr = NULL; \
 
 			  ptr = (mstate)tsd_getspecific(arena_key, vptr); \
 
 } while(0)
 
 
 
 			#ifdef PER_THREAD
 
 			#define arena_lock(ptr, size) do { \
 
 if(ptr) \
 
 (void)mutex_lock(&ptr->mutex); \
 
 else \
 
 			    ptr = arena_get2(ptr, (size)); \
 
 } while(0)
 
 			#else
 
 			#define arena_lock(ptr, size) do { \
 
 if(ptr && !mutex_trylock(&ptr->mutex)) { \
 
 			    THREAD_STAT(++(ptr->stat_lock_direct)); \
 
 } else \
 
 			    ptr = arena_get2(ptr, (size)); \
 
 } while(0)
 
 			#endif
 
 
 
 /* find the heap and corresponding arena for a given ptr */
 
 
 
 			#define heap_for_ptr(ptr) \
 
 ((heap_info *)((unsigned long)(ptr) & ~(HEAP_MAX_SIZE-1)))
 
 			#define arena_for_chunk(ptr) \
 
 (chunk_non_main_arena(ptr) ? heap_for_ptr(ptr)->ar_ptr : &main_arena)
 
 
 
 			#else /* !USE_ARENAS */
 
 
 
 /* There is only one arena, main_arena. */
 
 
 
 			#if THREAD_STATS
 
 			#define arena_get(ar_ptr, sz) do { \
 
 			  ar_ptr = &main_arena; \
 
 if(!mutex_trylock(&ar_ptr->mutex)) \
 
 ++(ar_ptr->stat_lock_direct); \
 
 else { \
 
 (void)mutex_lock(&ar_ptr->mutex); \
 
 ++(ar_ptr->stat_lock_wait); \
 
 } \
 
 } while(0)
 
 			#else
 
 			#define arena_get(ar_ptr, sz) do { \
 
 			  ar_ptr = &main_arena; \
 
 (void)mutex_lock(&ar_ptr->mutex); \
 
 } while(0)
 
 			#endif
 
 			#define arena_for_chunk(ptr) (&main_arena)
 
 
 
 			#endif /* USE_ARENAS */
 
 
 
 /**************************************************************************/
 
 
 
 			#ifndef NO_THREADS
 
 
 
 /* atfork support. */
 
 
 
 			static __malloc_ptr_t (*save_malloc_hook) (size_t __size,
 
 			                       __const __malloc_ptr_t);
 
 			# if !defined _LIBC || (defined SHARED && !USE___THREAD)
 
 			static __malloc_ptr_t (*save_memalign_hook) (size_t __align, size_t __size,
 
 			                         __const __malloc_ptr_t);
 
 			# endif
 
 			static void (*save_free_hook) (__malloc_ptr_t __ptr,
 
 			                     __const __malloc_ptr_t);
 
 			static Void_t* save_arena;
 
 
 
 			#ifdef ATFORK_MEM
 
 			ATFORK_MEM;
 
 			#endif
 
 
 
 /* Magic value for the thread-specific arena pointer when
 
 			   malloc_atfork() is in use. */
 
 
 
 			#define ATFORK_ARENA_PTR ((Void_t*)-1)
 
 
 
 /* The following hooks are used while the `atfork' handling mechanism
 
 is active. */
 
 
 
 			static Void_t*
 
 			malloc_atfork(size_t sz, const Void_t *caller)
 
 {
 
 			  Void_t *vptr = NULL;
 
 			  Void_t *victim;
 
 
 
 			  tsd_getspecific(arena_key, vptr);
 
 if(vptr == ATFORK_ARENA_PTR) {
 
 /* We are the only thread that may allocate at all. */
 
 if(save_malloc_hook != malloc_check) {
 
 			      return _int_malloc(&main_arena, sz);
 
 } else {
 
 if(top_check()<0)
 
 			    return 0;
 
 			      victim = _int_malloc(&main_arena, sz+1);
 
 			      return mem2mem_check(victim, sz);
 
 }
 
 } else {
 
 /* Suspend the thread until the `atfork' handlers have completed.
 
 			       By that time, the hooks will have been reset as well, so that
 
 			       mALLOc() can be used again. */
 
 (void)mutex_lock(&list_lock);
 
 (void)mutex_unlock(&list_lock);
 
 			    return public_mALLOc(sz);
 
 }
 
 }
 
 
 
 			static void
 
 			free_atfork(Void_t* mem, const Void_t *caller)
 
 {
 
 			  Void_t *vptr = NULL;
 
 			  mstate ar_ptr;
 
 			  mchunkptr p; /* chunk corresponding to mem */
 
 
 
 if (mem == 0) /* free(0) has no effect */
 
 			    return;
 
 
 
 			  p = mem2chunk(mem); /* do not bother to replicate free_check here */
 
 
 
 			#if HAVE_MMAP
 
 if (chunk_is_mmapped(p)) /* release mmapped memory. */
 
 {
 
 			    munmap_chunk(p);
 
 			    return;
 
 }
 
 			#endif
 
 
 
 			#ifdef ATOMIC_FASTBINS
 
 			  ar_ptr = arena_for_chunk(p);
 
 			  tsd_getspecific(arena_key, vptr);
 
 			  _int_free(ar_ptr, p, vptr == ATFORK_ARENA_PTR);
 
 			#else
 
 			  ar_ptr = arena_for_chunk(p);
 
 			  tsd_getspecific(arena_key, vptr);
 
 if(vptr != ATFORK_ARENA_PTR)
 
 (void)mutex_lock(&ar_ptr->mutex);
 
 			  _int_free(ar_ptr, p);
 
 if(vptr != ATFORK_ARENA_PTR)
 
 (void)mutex_unlock(&ar_ptr->mutex);
 
 			#endif
 
 }
 
 
 
 
 
 /* Counter for number of times the list is locked by the same thread. */
 
 			static unsigned int atfork_recursive_cntr;
 
 
 
 /* The following two functions are registered via thread_atfork() to
 
 			   make sure that the mutexes remain in a consistent state in the
 
 			   fork()ed version of a thread. Also adapt the malloc and free hooks
 
 			   temporarily, because the `atfork' handler mechanism may use
 
 			   malloc/free internally (e.g. in LinuxThreads). */
 
 
 
 			static void
 
 			ptmalloc_lock_all (void)
 
 {
 
 			  mstate ar_ptr;
 
 
 
 if(__malloc_initialized < 1)
 
 			    return;
 
 if (mutex_trylock(&list_lock))
 
 {
 
 			      Void_t *my_arena;
 
 			      tsd_getspecific(arena_key, my_arena);
 
 if (my_arena == ATFORK_ARENA_PTR)
 
 /* This is the same thread which already locks the global list.
 
 			       Just bump the counter. */
 
 			    goto out;
 
 
 
 /* This thread has to wait its turn. */
 
 (void)mutex_lock(&list_lock);
 
 }
 
 for(ar_ptr = &main_arena;;) {
 
 (void)mutex_lock(&ar_ptr->mutex);
 
 			    ar_ptr = ar_ptr->next;
 
 if(ar_ptr == &main_arena) break;
 
 }
 
 			  save_malloc_hook = __malloc_hook;
 
 			  save_free_hook = __free_hook;
 
 			  __malloc_hook = malloc_atfork;
 
 			  __free_hook = free_atfork;
 
 /* Only the current thread may perform malloc/free calls now. */
 
 			  tsd_getspecific(arena_key, save_arena);
 
 			  tsd_setspecific(arena_key, ATFORK_ARENA_PTR);
 
 			 out:
 
 ++atfork_recursive_cntr;
 
 }
 
 
 
 			static void
 
 			ptmalloc_unlock_all (void)
 
 {
 
 			  mstate ar_ptr;
 
 
 
 if(__malloc_initialized < 1)
 
 			    return;
 
 if (--atfork_recursive_cntr != 0)
 
 			    return;
 
 			  tsd_setspecific(arena_key, save_arena);
 
 			  __malloc_hook = save_malloc_hook;
 
 			  __free_hook = save_free_hook;
 
 for(ar_ptr = &main_arena;;) {
 
 (void)mutex_unlock(&ar_ptr->mutex);
 
 			    ar_ptr = ar_ptr->next;
 
 if(ar_ptr == &main_arena) break;
 
 }
 
 (void)mutex_unlock(&list_lock);
 
 }
 
 
 
 			#ifdef __linux__
 
 
 
 /* In NPTL, unlocking a mutex in the child process after a
 
 			   fork() is currently unsafe, whereas re-initializing it is safe and
 
 			   does not leak resources. Therefore, a special atfork handler is
 
 			   installed for the child. */
 
 
 
 			static void
 
 			ptmalloc_unlock_all2 (void)
 
 {
 
 			  mstate ar_ptr;
 
 
 
 if(__malloc_initialized < 1)
 
 			    return;
 
 			#if defined _LIBC || defined MALLOC_HOOKS
 
 			  tsd_setspecific(arena_key, save_arena);
 
 			  __malloc_hook = save_malloc_hook;
 
 			  __free_hook = save_free_hook;
 
 			#endif
 
 			#ifdef PER_THREAD
 
 			  free_list = NULL;
 
 			#endif
 
 for(ar_ptr = &main_arena;;) {
 
 			    mutex_init(&ar_ptr->mutex);
 
 			#ifdef PER_THREAD
 
 if (ar_ptr != save_arena) {
 
 			      ar_ptr->next_free = free_list;
 
 			      free_list = ar_ptr;
 
 }
 
 			#endif
 
 			    ar_ptr = ar_ptr->next;
 
 if(ar_ptr == &main_arena) break;
 
 }
 
 			  mutex_init(&list_lock);
 
 			  atfork_recursive_cntr = 0;
 
 }
 
 
 
 			#else
 
 
 
 			#define ptmalloc_unlock_all2 ptmalloc_unlock_all
 
 
 
 			#endif
 
 
 
 			#endif /* !defined NO_THREADS */
 
 
 
 /* Initialization routine. */
 
 			#ifdef _LIBC
 
 			#include <string.h>
 
 			extern char **_environ;
 
 
 
 			static char *
 
 			internal_function
 
 			next_env_entry (char ***position)
 
 {
 
 			  char **current = *position;
 
 			  char *result = NULL;
 
 
 
 while (*current != NULL)
 
 {
 
 if (__builtin_expect ((*current)[0] == 'M', 0)
 
 && (*current)[1] == 'A'
 
 && (*current)[2] == 'L'
 
 && (*current)[3] == 'L'
 
 && (*current)[4] == 'O'
 
 && (*current)[5] == 'C'
 
 && (*current)[6] == '_')
 
 {
 
 			      result = &(*current)[7];
 
 
 
 /* Save current position for next visit. */
 
 *position = ++current;
 
 
 
 			      break;
 
 }
 
 
 
 ++current;
 
 }
 
 
 
 			  return result;
 
 }
 
 			#endif /* _LIBC */
 
 
 
 /* Set up basic state so that _int_malloc et al can work. */
 
 			static void
 
 			ptmalloc_init_minimal (void)
 
 {
 
 			#if DEFAULT_TOP_PAD != 0
 
 			  mp_.top_pad = DEFAULT_TOP_PAD;
 
 			#endif
 
 			  mp_.n_mmaps_max = DEFAULT_MMAP_MAX;
 
 			  mp_.mmap_threshold = DEFAULT_MMAP_THRESHOLD;
 
 			  mp_.trim_threshold = DEFAULT_TRIM_THRESHOLD;
 
 			  mp_.pagesize = malloc_getpagesize;
 
 			#ifdef PER_THREAD
 
 			# define NARENAS_FROM_NCORES(n) ((n) * (sizeof(long) == 4 ? 2 : 8))
 
 			  mp_.arena_test = NARENAS_FROM_NCORES (1);
 
 			  narenas = 1;
 
 			#endif
 
 }
 
 
 
 
 
 			#ifdef _LIBC
 
 			# ifdef SHARED
 
 			static void *
 
 			__failing_morecore (ptrdiff_t d)
 
 {
 
 			  return (void *) MORECORE_FAILURE;
 
 }
 
 
 
 			extern struct dl_open_hook *_dl_open_hook;
 
 			libc_hidden_proto (_dl_open_hook);
 
 			# endif
 
 
 
 			# if defined SHARED && !USE___THREAD
 
 /* This is called by __pthread_initialize_minimal when it needs to use
 
 			   malloc to set up the TLS state. We cannot do the full work of
 
 			   ptmalloc_init (below) until __pthread_initialize_minimal has finished,
 
 			   so it has to switch to using the special startup-time hooks while doing
 
 			   those allocations. */
 
 			void
 
 			__libc_malloc_pthread_startup (bool first_time)
 
 {
 
 if (first_time)
 
 {
 
 			      ptmalloc_init_minimal ();
 
 			      save_malloc_hook = __malloc_hook;
 
 			      save_memalign_hook = __memalign_hook;
 
 			      save_free_hook = __free_hook;
 
 			      __malloc_hook = malloc_starter;
 
 			      __memalign_hook = memalign_starter;
 
 			      __free_hook = free_starter;
 
 }
 
 else
 
 {
 
 			      __malloc_hook = save_malloc_hook;
 
 			      __memalign_hook = save_memalign_hook;
 
 			      __free_hook = save_free_hook;
 
 }
 
 }
 
 			# endif
 
 			#endif
 
 
 
 			static void
 
 			ptmalloc_init (void)
 
 {
 
 			#if __STD_C
 
 const char* s;
 
 			#else
 
 			  char* s;
 
 			#endif
 
 int secure = 0;
 
 
 
 if(__malloc_initialized >= 0) return;
 
 			  __malloc_initialized = 0;
 
 
 
 			#ifdef _LIBC
 
 			# if defined SHARED && !USE___THREAD
 
 /* ptmalloc_init_minimal may already have been called via
 
 			     __libc_malloc_pthread_startup, above. */
 
 if (mp_.pagesize == 0)
 
 			# endif
 
 			#endif
 
 			    ptmalloc_init_minimal();
 
 
 
 			#ifndef NO_THREADS
 
 			# if defined _LIBC
 
 /* We know __pthread_initialize_minimal has already been called,
 
 and that is enough. */
 
 			#   define NO_STARTER
 
 			# endif
 
 			# ifndef NO_STARTER
 
 /* With some threads implementations, creating thread-specific data
 
 or initializing a mutex may call malloc() itself. Provide a
 
 			     simple starter version (realloc() won't work). */
 
 			  save_malloc_hook = __malloc_hook;
 
 			  save_memalign_hook = __memalign_hook;
 
 			  save_free_hook = __free_hook;
 
 			  __malloc_hook = malloc_starter;
 
 			  __memalign_hook = memalign_starter;
 
 			  __free_hook = free_starter;
 
 			#  ifdef _LIBC
 
 /* Initialize the pthreads interface. */
 
 if (__pthread_initialize != NULL)
 
 			    __pthread_initialize();
 
 			#  endif /* !defined _LIBC */
 
 			# endif    /* !defined NO_STARTER */
 
 			#endif /* !defined NO_THREADS */
 
 			  mutex_init(&main_arena.mutex);
 
 			  main_arena.next = &main_arena;
 
 
 
 			#if defined _LIBC && defined SHARED
 
 /* In case this libc copy is in a non-default namespace, never use brk.
 
 			     Likewise if dlopened from statically linked program. */
 
 			  Dl_info di;
 
 			  struct link_map *l;
 
 
 
 if (_dl_open_hook != NULL
 
 || (_dl_addr (ptmalloc_init, &di, &l, NULL) != 0
 
 && l->l_ns != LM_ID_BASE))
 
 			    __morecore = __failing_morecore;
 
 			#endif
 
 
 
 			  mutex_init(&list_lock);
 
 			  tsd_key_create(&arena_key, NULL);
 
 			  tsd_setspecific(arena_key, (Void_t *)&main_arena);
 
 			  thread_atfork(ptmalloc_lock_all, ptmalloc_unlock_all, ptmalloc_unlock_all2);
 
 			#ifndef NO_THREADS
 
 			# ifndef NO_STARTER
 
 			  __malloc_hook = save_malloc_hook;
 
 			  __memalign_hook = save_memalign_hook;
 
 			  __free_hook = save_free_hook;
 
 			# else
 
 			#  undef NO_STARTER
 
 			# endif
 
 			#endif
 
 			#ifdef _LIBC
 
 			  secure = __libc_enable_secure;
 
 			  s = NULL;
 
 if (__builtin_expect (_environ != NULL, 1))
 
 {
 
 			      char **runp = _environ;
 
 			      char *envline;
 
 
 
 while (__builtin_expect ((envline = next_env_entry (&runp)) != NULL,
 
 			                   0))
 
 {
 
 			      size_t len = strcspn (envline, "=");
 
 
 
 if (envline[len] != '=')
 
 /* This is a "MALLOC_" variable at the end of the string
 
 			           without a '=' character. Ignore it since otherwise we
 
 			           will access invalid memory below. */
 
 			        continue;
 
 
 
 			      switch (len)
 
 {
 
 case 6:
 
 if (memcmp (envline, "CHECK_", 6) == 0)
 
 			        s = &envline[7];
 
 			          break;
 
 case 8:
 
 if (! secure)
 
 {
 
 if (memcmp (envline, "TOP_PAD_", 8) == 0)
 
 			            mALLOPt(M_TOP_PAD, atoi(&envline[9]));
 
 else if (memcmp (envline, "PERTURB_", 8) == 0)
 
 			            mALLOPt(M_PERTURB, atoi(&envline[9]));
 
 }
 
 			          break;
 
 case 9:
 
 if (! secure)
 
 {
 
 if (memcmp (envline, "MMAP_MAX_", 9) == 0)
 
 			            mALLOPt(M_MMAP_MAX, atoi(&envline[10]));
 
 			#ifdef PER_THREAD
 
 else if (memcmp (envline, "ARENA_MAX", 9) == 0)
 
 			            mALLOPt(M_ARENA_MAX, atoi(&envline[10]));
 
 			#endif
 
 }
 
 			          break;
 
 			#ifdef PER_THREAD
 
 case 10:
 
 if (! secure)
 
 {
 
 if (memcmp (envline, "ARENA_TEST", 10) == 0)
 
 			            mALLOPt(M_ARENA_TEST, atoi(&envline[11]));
 
 }
 
 			          break;
 
 			#endif
 
 case 15:
 
 if (! secure)
 
 {
 
 if (memcmp (envline, "TRIM_THRESHOLD_", 15) == 0)
 
 			            mALLOPt(M_TRIM_THRESHOLD, atoi(&envline[16]));
 
 else if (memcmp (envline, "MMAP_THRESHOLD_", 15) == 0)
 
 			            mALLOPt(M_MMAP_THRESHOLD, atoi(&envline[16]));
 
 }
 
 			          break;
 
 			        default:
 
 			          break;
 
 }
 
 }
 
 }
 
 			#else
 
 if (! secure)
 
 {
 
 if((s = getenv("MALLOC_TRIM_THRESHOLD_")))
 
 			    mALLOPt(M_TRIM_THRESHOLD, atoi(s));
 
 if((s = getenv("MALLOC_TOP_PAD_")))
 
 			    mALLOPt(M_TOP_PAD, atoi(s));
 
 if((s = getenv("MALLOC_PERTURB_")))
 
 			    mALLOPt(M_PERTURB, atoi(s));
 
 if((s = getenv("MALLOC_MMAP_THRESHOLD_")))
 
 			    mALLOPt(M_MMAP_THRESHOLD, atoi(s));
 
 if((s = getenv("MALLOC_MMAP_MAX_")))
 
 			    mALLOPt(M_MMAP_MAX, atoi(s));
 
 }
 
 			  s = getenv("MALLOC_CHECK_");
 
 			#endif
 
 if(s && s[0]) {
 
 			    mALLOPt(M_CHECK_ACTION, (int)(s[0] - '0'));
 
 if (check_action != 0)
 
 			      __malloc_check_init();
 
 }
 
 			  void (*hook) (void) = force_reg (__malloc_initialize_hook);
 
 if (hook != NULL)
 
 (*hook)();
 
 			  __malloc_initialized = 1;
 
 }
 
 
 
 /* There are platforms (e.g. Hurd) with a link-time hook mechanism. */
 
 			#ifdef thread_atfork_static
 
 			thread_atfork_static(ptmalloc_lock_all, ptmalloc_unlock_all, \
 
 			             ptmalloc_unlock_all2)
 
 			#endif
 
 
 
 
 
 
 
 /* Managing heaps and arenas (for concurrent threads) */
 
 
 
 			#if USE_ARENAS
 
 
 
 			#if MALLOC_DEBUG > 1
 
 
 
 /* Print the complete contents of a single heap to stderr. */
 
 
 
 			static void
 
 			#if __STD_C
 
 			dump_heap(heap_info *heap)
 
 			#else
 
 			dump_heap(heap) heap_info *heap;
 
 			#endif
 
 {
 
 			  char *ptr;
 
 			  mchunkptr p;
 
 
 
 			  fprintf(stderr, "Heap %p, size %10lx:\n", heap, (long)heap->size);
 
 			  ptr = (heap->ar_ptr != (mstate)(heap+1)) ?
 
 (char*)(heap + 1) : (char*)(heap + 1) + sizeof(struct malloc_state);
 
 			  p = (mchunkptr)(((unsigned long)ptr + MALLOC_ALIGN_MASK) &
 
 ~MALLOC_ALIGN_MASK);
 
 for(;;) {
 
 			    fprintf(stderr, "chunk %p size %10lx", p, (long)p->size);
 
 if(p == top(heap->ar_ptr)) {
 
 			      fprintf(stderr, " (top)\n");
 
 			      break;
 
 } else if(p->size == (0|PREV_INUSE)) {
 
 			      fprintf(stderr, " (fence)\n");
 
 			      break;
 
 }
 
 			    fprintf(stderr, "\n");
 
 			    p = next_chunk(p);
 
 }
 
 }
 
 
 
 			#endif /* MALLOC_DEBUG > 1 */
 
 
 
 /* If consecutive mmap (0, HEAP_MAX_SIZE << 1, ...) calls return decreasing
 
 			   addresses as opposed to increasing, new_heap would badly fragment the
 
 			   address space. In that case remember the second HEAP_MAX_SIZE part
 
 			   aligned to HEAP_MAX_SIZE from last mmap (0, HEAP_MAX_SIZE << 1, ...)
 
 call (if it is already aligned) and try to reuse it next time. We need
 
 			   no locking for it, as kernel ensures the atomicity for us - worst case
 
 			   we'll call mmap (addr, HEAP_MAX_SIZE, ...) for some value of addr in
 
 			   multiple threads, but only one will succeed. */
 
 			static char *aligned_heap_area;
 
 
 
 /* Create a new heap. size is automatically rounded up to a multiple
 
 			   of the page size. */
 
 
 
 			static heap_info *
 
 			internal_function
 
 			#if __STD_C
 
 			new_heap(size_t size, size_t top_pad)
 
 			#else
 
 			new_heap(size, top_pad) size_t size, top_pad;
 
 			#endif
 
 {
 
 			  size_t page_mask = malloc_getpagesize - 1;
 
 			  char *p1, *p2;
 
 			  unsigned long ul;
 
 			  heap_info *h;
 
 
 
 if(size+top_pad < HEAP_MIN_SIZE)
 
 			    size = HEAP_MIN_SIZE;
 
 else if(size+top_pad <= HEAP_MAX_SIZE)
 
 			    size += top_pad;
 
 else if(size > HEAP_MAX_SIZE)
 
 			    return 0;
 
 else
 
 			    size = HEAP_MAX_SIZE;
 
 			  size = (size + page_mask) & ~page_mask;
 
 
 
 /* A memory region aligned to a multiple of HEAP_MAX_SIZE is needed.
 
 			     No swap space needs to be reserved for the following large
 
 			     mapping (on Linux, this is the case for all non-writable mappings
 
 			     anyway). */
 
 			  p2 = MAP_FAILED;
 
 if(aligned_heap_area) {
 
 			    p2 = (char *)MMAP(aligned_heap_area, HEAP_MAX_SIZE, PROT_NONE,
 
 			              MAP_PRIVATE|MAP_NORESERVE);
 
 			    aligned_heap_area = NULL;
 
 if (p2 != MAP_FAILED && ((unsigned long)p2 & (HEAP_MAX_SIZE-1))) {
 
 			      munmap(p2, HEAP_MAX_SIZE);
 
 			      p2 = MAP_FAILED;
 
 }
 
 }
 
 if(p2 == MAP_FAILED) {
 
 			    p1 = (char *)MMAP(0, HEAP_MAX_SIZE<<1, PROT_NONE,
 
 			              MAP_PRIVATE|MAP_NORESERVE);
 
 if(p1 != MAP_FAILED) {
 
 			      p2 = (char *)(((unsigned long)p1 + (HEAP_MAX_SIZE-1))
 
 & ~(HEAP_MAX_SIZE-1));
 
 			      ul = p2 - p1;
 
 if (ul)
 
 			    munmap(p1, ul);
 
 else
 
 			    aligned_heap_area = p2 + HEAP_MAX_SIZE;
 
 			      munmap(p2 + HEAP_MAX_SIZE, HEAP_MAX_SIZE - ul);
 
 } else {
 
 /* Try to take the chance that an allocation of only HEAP_MAX_SIZE
 
 is already aligned. */
 
 			      p2 = (char *)MMAP(0, HEAP_MAX_SIZE, PROT_NONE, MAP_PRIVATE|MAP_NORESERVE);
 
 if(p2 == MAP_FAILED)
 
 			    return 0;
 
 if((unsigned long)p2 & (HEAP_MAX_SIZE-1)) {
 
 			    munmap(p2, HEAP_MAX_SIZE);
 
 			    return 0;
 
 }
 
 }
 
 }
 
 if(mprotect(p2, size, PROT_READ|PROT_WRITE) != 0) {
 
 			    munmap(p2, HEAP_MAX_SIZE);
 
 			    return 0;
 
 }
 
 			  h = (heap_info *)p2;
 
 			  h->size = size;
 
 			  h->mprotect_size = size;
 
 			  THREAD_STAT(stat_n_heaps++);
 
 			  return h;
 
 }
 
 
 
 /* Grow a heap. size is automatically rounded up to a
 
 			   multiple of the page size. */
 
 
 
 			static int
 
 			#if __STD_C
 
 			grow_heap(heap_info *h, long diff)
 
 			#else
 
 			grow_heap(h, diff) heap_info *h; long diff;
 
 			#endif
 
 {
 
 			  size_t page_mask = malloc_getpagesize - 1;
 
 			  long new_size;
 
 
 
 			  diff = (diff + page_mask) & ~page_mask;
 
 			  new_size = (long)h->size + diff;
 
 if((unsigned long) new_size > (unsigned long) HEAP_MAX_SIZE)
 
 			    return -1;
 
 if((unsigned long) new_size > h->mprotect_size) {
 
 if (mprotect((char *)h + h->mprotect_size,
 
 (unsigned long) new_size - h->mprotect_size,
 
 			         PROT_READ|PROT_WRITE) != 0)
 
 			      return -2;
 
 			    h->mprotect_size = new_size;
 
 }
 
 
 
 			  h->size = new_size;
 
 			  return 0;
 
 }
 
 
 
 /* Shrink a heap. */
 
 
 
 			static int
 
 			#if __STD_C
 
 			shrink_heap(heap_info *h, long diff)
 
 			#else
 
 			shrink_heap(h, diff) heap_info *h; long diff;
 
 			#endif
 
 {
 
 			  long new_size;
 
 
 
 			  new_size = (long)h->size - diff;
 
 if(new_size < (long)sizeof(*h))
 
 			    return -1;
 
 /* Try to re-map the extra heap space freshly to save memory, and
 
 			     make it inaccessible. */
 
 			#ifdef _LIBC
 
 if (__builtin_expect (__libc_enable_secure, 0))
 
 			#else
 
 if (1)
 
 			#endif
 
 {
 
 if((char *)MMAP((char *)h + new_size, diff, PROT_NONE,
 
 			              MAP_PRIVATE|MAP_FIXED) == (char *) MAP_FAILED)
 
 			    return -2;
 
 			      h->mprotect_size = new_size;
 
 }
 
 			#ifdef _LIBC
 
 else
 
 			    madvise ((char *)h + new_size, diff, MADV_DONTNEED);
 
 			#endif
 
 /*fprintf(stderr, "shrink %p %08lx\n", h, new_size);*/
 
 
 
 			  h->size = new_size;
 
 			  return 0;
 
 }
 
 
 
 /* Delete a heap. */
 
 
 
 			#define delete_heap(heap) \
 
 do { \
 
 if ((char *)(heap) + HEAP_MAX_SIZE == aligned_heap_area) \
 
 			      aligned_heap_area = NULL; \
 
 			    munmap((char*)(heap), HEAP_MAX_SIZE); \
 
 } while (0)
 
 
 
 			static int
 
 			internal_function
 
 			#if __STD_C
 
 			heap_trim(heap_info *heap, size_t pad)
 
 			#else
 
 			heap_trim(heap, pad) heap_info *heap; size_t pad;
 
 			#endif
 
 {
 
 			  mstate ar_ptr = heap->ar_ptr;
 
 			  unsigned long pagesz = mp_.pagesize;
 
 			  mchunkptr top_chunk = top(ar_ptr), p, bck, fwd;
 
 			  heap_info *prev_heap;
 
 			  long new_size, top_size, extra;
 
 
 
 /* Can this heap go away completely? */
 
 while(top_chunk == chunk_at_offset(heap, sizeof(*heap))) {
 
 			    prev_heap = heap->prev;
 
 			    p = chunk_at_offset(prev_heap, prev_heap->size - (MINSIZE-2*SIZE_SZ));
 
 			    assert(p->size == (0|PREV_INUSE)); /* must be fencepost */
 
 			    p = prev_chunk(p);
 
 			    new_size = chunksize(p) + (MINSIZE-2*SIZE_SZ);
 
 			    assert(new_size>0 && new_size<(long)(2*MINSIZE));
 
 if(!prev_inuse(p))
 
 			      new_size += p->prev_size;
 
 			    assert(new_size>0 && new_size<HEAP_MAX_SIZE);
 
 if(new_size + (HEAP_MAX_SIZE - prev_heap->size) < pad + MINSIZE + pagesz)
 
 			      break;
 
 			    ar_ptr->system_mem -= heap->size;
 
 			    arena_mem -= heap->size;
 
 			    delete_heap(heap);
 
 			    heap = prev_heap;
 
 if(!prev_inuse(p)) { /* consolidate backward */
 
 			      p = prev_chunk(p);
 
 			      unlink(p, bck, fwd);
 
 }
 
 			    assert(((unsigned long)((char*)p + new_size) & (pagesz-1)) == 0);
 
 			    assert( ((char*)p + new_size) == ((char*)heap + heap->size) );
 
 			    top(ar_ptr) = top_chunk = p;
 
 			    set_head(top_chunk, new_size | PREV_INUSE);
 
 /*check_chunk(ar_ptr, top_chunk);*/
 
 }
 
 			  top_size = chunksize(top_chunk);
 
 			  extra = (top_size - pad - MINSIZE - 1) & ~(pagesz - 1);
 
 if(extra < (long)pagesz)
 
 			    return 0;
 
 /* Try to shrink. */
 
 if(shrink_heap(heap, extra) != 0)
 
 			    return 0;
 
 			  ar_ptr->system_mem -= extra;
 
 			  arena_mem -= extra;
 
 
 
 /* Success. Adjust top accordingly. */
 
 			  set_head(top_chunk, (top_size - extra) | PREV_INUSE);
 
 /*check_chunk(ar_ptr, top_chunk);*/
 
 			  return 1;
 
 }
 
 
 
 /* Create a new arena with initial size "size". */
 
 
 
 			static mstate
 
 			_int_new_arena(size_t size)
 
 {
 
 			  mstate a;
 
 			  heap_info *h;
 
 			  char *ptr;
 
 			  unsigned long misalign;
 
 
 
 			  h = new_heap(size + (sizeof(*h) + sizeof(*a) + MALLOC_ALIGNMENT),
 
 			           mp_.top_pad);
 
 if(!h) {
 
 /* Maybe size is too large to fit in a single heap. So, just try
 
 to create a minimally-sized arena and let _int_malloc() attempt
 
 to deal with the large request via mmap_chunk(). */
 
 			    h = new_heap(sizeof(*h) + sizeof(*a) + MALLOC_ALIGNMENT, mp_.top_pad);
 
 if(!h)
 
 			      return 0;
 
 }
 
 			  a = h->ar_ptr = (mstate)(h+1);
 
 			  malloc_init_state(a);
 
 /*a->next = NULL;*/
 
 			  a->system_mem = a->max_system_mem = h->size;
 
 			  arena_mem += h->size;
 
 			#ifdef NO_THREADS
 
 if((unsigned long)(mp_.mmapped_mem + arena_mem + main_arena.system_mem) >
 
 			     mp_.max_total_mem)
 
 			    mp_.max_total_mem = mp_.mmapped_mem + arena_mem + main_arena.system_mem;
 
 			#endif
 
 
 
 /* Set up the top chunk, with proper alignment. */
 
 			  ptr = (char *)(a + 1);
 
 			  misalign = (unsigned long)chunk2mem(ptr) & MALLOC_ALIGN_MASK;
 
 if (misalign > 0)
 
 			    ptr += MALLOC_ALIGNMENT - misalign;
 
 			  top(a) = (mchunkptr)ptr;
 
 			  set_head(top(a), (((char*)h + h->size) - ptr) | PREV_INUSE);
 
 
 
 			  tsd_setspecific(arena_key, (Void_t *)a);
 
 			  mutex_init(&a->mutex);
 
 (void)mutex_lock(&a->mutex);
 
 
 
 			#ifdef PER_THREAD
 
 (void)mutex_lock(&list_lock);
 
 			#endif
 
 
 
 /* Add the new arena to the global list. */
 
 			  a->next = main_arena.next;
 
 			  atomic_write_barrier ();
 
 			  main_arena.next = a;
 
 
 
 			#ifdef PER_THREAD
 
 ++narenas;
 
 
 
 (void)mutex_unlock(&list_lock);
 
 			#endif
 
 
 
 			  THREAD_STAT(++(a->stat_lock_loop));
 
 
 
 			  return a;
 
 }
 
 
 
 
 
 			#ifdef PER_THREAD
 
 			static mstate
 
 			get_free_list (void)
 
 {
 
 			  mstate result = free_list;
 
 if (result != NULL)
 
 {
 
 (void)mutex_lock(&list_lock);
 
 			      result = free_list;
 
 if (result != NULL)
 
 			    free_list = result->next_free;
 
 (void)mutex_unlock(&list_lock);
 
 
 
 if (result != NULL)
 
 {
 
 (void)mutex_lock(&result->mutex);
 
 			      tsd_setspecific(arena_key, (Void_t *)result);
 
 			      THREAD_STAT(++(result->stat_lock_loop));
 
 }
 
 }
 
 
 
 			  return result;
 
 }
 
 
 
 
 
 			static mstate
 
 			reused_arena (void)
 
 {
 
 if (narenas <= mp_.arena_test)
 
 			    return NULL;
 
 
 
 			  static int narenas_limit;
 
 if (narenas_limit == 0)
 
 {
 
 if (mp_.arena_max != 0)
 
 			    narenas_limit = mp_.arena_max;
 
 else
 
 {
 
 int n = __get_nprocs ();
 
 
 
 if (n >= 1)
 
 			        narenas_limit = NARENAS_FROM_NCORES (n);
 
 else
 
 /* We have no information about the system. Assume two
 
 			           cores. */
 
 			        narenas_limit = NARENAS_FROM_NCORES (2);
 
 }
 
 }
 
 
 
 if (narenas < narenas_limit)
 
 			    return NULL;
 
 
 
 			  mstate result;
 
 			  static mstate next_to_use;
 
 if (next_to_use == NULL)
 
 			    next_to_use = &main_arena;
 
 
 
 			  result = next_to_use;
 
 do
 
 {
 
 if (!mutex_trylock(&result->mutex))
 
 			    goto out;
 
 
 
 			      result = result->next;
 
 }
 
 while (result != next_to_use);
 
 
 
 /* No arena available. Wait for the next in line. */
 
 (void)mutex_lock(&result->mutex);
 
 
 
 			 out:
 
 			  tsd_setspecific(arena_key, (Void_t *)result);
 
 			  THREAD_STAT(++(result->stat_lock_loop));
 
 			  next_to_use = result->next;
 
 
 
 			  return result;
 
 }
 
 			#endif
 
 
 
 			static mstate
 
 			internal_function
 
 			#if __STD_C
 
 			arena_get2(mstate a_tsd, size_t size)
 
 			#else
 
 			arena_get2(a_tsd, size) mstate a_tsd; size_t size;
 
 			#endif
 
 {
 
 			  mstate a;
 
 
 
 			#ifdef PER_THREAD
 
 if ((a = get_free_list ()) == NULL
 
 && (a = reused_arena ()) == NULL)
 
 /* Nothing immediately available, so generate a new arena. */
 
 			    a = _int_new_arena(size);
 
 			#else
 
 if(!a_tsd)
 
 			    a = a_tsd = &main_arena;
 
 else {
 
 			    a = a_tsd->next;
 
 if(!a) {
 
 /* This can only happen while initializing the new arena. */
 
 (void)mutex_lock(&main_arena.mutex);
 
 			      THREAD_STAT(++(main_arena.stat_lock_wait));
 
 			      return &main_arena;
 
 }
 
 }
 
 
 
 /* Check the global, circularly linked list for available arenas. */
 
 			  bool retried = false;
 
 			 repeat:
 
 do {
 
 if(!mutex_trylock(&a->mutex)) {
 
 if (retried)
 
 (void)mutex_unlock(&list_lock);
 
 			      THREAD_STAT(++(a->stat_lock_loop));
 
 			      tsd_setspecific(arena_key, (Void_t *)a);
 
 			      return a;
 
 }
 
 			    a = a->next;
 
 } while(a != a_tsd);
 
 
 
 /* If not even the list_lock can be obtained, try again. This can
 
 			     happen during `atfork', or for example on systems where thread
 
 			     creation makes it temporarily impossible to obtain _any_
 
 			     locks. */
 
 if(!retried && mutex_trylock(&list_lock)) {
 
 /* We will block to not run in a busy loop. */
 
 (void)mutex_lock(&list_lock);
 
 
 
 /* Since we blocked there might be an arena available now. */
 
 			    retried = true;
 
 			    a = a_tsd;
 
 			    goto repeat;
 
 }
 
 
 
 /* Nothing immediately available, so generate a new arena. */
 
 			  a = _int_new_arena(size);
 
 (void)mutex_unlock(&list_lock);
 
 			#endif
 
 
 
 			  return a;
 
 }
 
 
 
 			#ifdef PER_THREAD
 
 			static void __attribute__ ((section ("__libc_thread_freeres_fn")))
 
 			arena_thread_freeres (void)
 
 {
 
 			  Void_t *vptr = NULL;
 
 			  mstate a = tsd_getspecific(arena_key, vptr);
 
 			  tsd_setspecific(arena_key, NULL);
 
 
 
 if (a != NULL)
 
 {
 
 (void)mutex_lock(&list_lock);
 
 			      a->next_free = free_list;
 
 			      free_list = a;
 
 (void)mutex_unlock(&list_lock);
 
 }
 
 }
 
 			text_set_element (__libc_thread_subfreeres, arena_thread_freeres);
 
 			#endif
 
 
 
 			#endif /* USE_ARENAS */
 
 
 
 /*
 
 * Local variables:
 
 * c-basic-offset: 2
 
 * End:
 
 */ 

本文参与腾讯云自媒体分享计划,欢迎正在阅读的你也加入,一起分享。

发表于

我来说两句

0 条评论
登录 后参与评论

扫码关注云+社区

领取腾讯云代金券