OpenCores
URL https://opencores.org/ocsvn/openrisc_me/openrisc_me/trunk

Subversion Repositories openrisc_me

[/] [openrisc/] [trunk/] [gnu-src/] [newlib-1.18.0/] [newlib/] [libc/] [stdlib/] [mallocr.c] - Blame information for rev 258

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 207 jeremybenn
#ifdef MALLOC_PROVIDED
2
int _dummy_mallocr = 1;
3
#else
4
/* ---------- To make a malloc.h, start cutting here ------------ */
5
 
6
/*
7
  A version of malloc/free/realloc written by Doug Lea and released to the
8
  public domain.  Send questions/comments/complaints/performance data
9
  to dl@cs.oswego.edu
10
 
11
* VERSION 2.6.4  Thu Nov 28 07:54:55 1996  Doug Lea  (dl at gee)
12
 
13
   Note: There may be an updated version of this malloc obtainable at
14
           ftp://g.oswego.edu/pub/misc/malloc.c
15
         Check before installing!
16
 
17
* Why use this malloc?
18
 
19
  This is not the fastest, most space-conserving, most portable, or
20
  most tunable malloc ever written. However it is among the fastest
21
  while also being among the most space-conserving, portable and tunable.
22
  Consistent balance across these factors results in a good general-purpose
23
  allocator. For a high-level description, see
24
     http://g.oswego.edu/dl/html/malloc.html
25
 
26
* Synopsis of public routines
27
 
28
  (Much fuller descriptions are contained in the program documentation below.)
29
 
30
  malloc(size_t n);
31
     Return a pointer to a newly allocated chunk of at least n bytes, or null
32
     if no space is available.
33
  free(Void_t* p);
34
     Release the chunk of memory pointed to by p, or no effect if p is null.
35
  realloc(Void_t* p, size_t n);
36
     Return a pointer to a chunk of size n that contains the same data
37
     as does chunk p up to the minimum of (n, p's size) bytes, or null
38
     if no space is available. The returned pointer may or may not be
39
     the same as p. If p is null, equivalent to malloc.  Unless the
40
     #define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
41
     size argument of zero (re)allocates a minimum-sized chunk.
42
  memalign(size_t alignment, size_t n);
43
     Return a pointer to a newly allocated chunk of n bytes, aligned
44
     in accord with the alignment argument, which must be a power of
45
     two.
46
  valloc(size_t n);
47
     Equivalent to memalign(pagesize, n), where pagesize is the page
48
     size of the system (or as near to this as can be figured out from
49
     all the includes/defines below.)
50
  pvalloc(size_t n);
51
     Equivalent to valloc(minimum-page-that-holds(n)), that is,
52
     round up n to nearest pagesize.
53
  calloc(size_t unit, size_t quantity);
54
     Returns a pointer to quantity * unit bytes, with all locations
55
     set to zero.
56
  cfree(Void_t* p);
57
     Equivalent to free(p).
58
  malloc_trim(size_t pad);
59
     Release all but pad bytes of freed top-most memory back
60
     to the system. Return 1 if successful, else 0.
61
  malloc_usable_size(Void_t* p);
62
     Report the number usable allocated bytes associated with allocated
63
     chunk p. This may or may not report more bytes than were requested,
64
     due to alignment and minimum size constraints.
65
  malloc_stats();
66
     Prints brief summary statistics on stderr.
67
  mallinfo()
68
     Returns (by copy) a struct containing various summary statistics.
69
  mallopt(int parameter_number, int parameter_value)
70
     Changes one of the tunable parameters described below. Returns
71
     1 if successful in changing the parameter, else 0.
72
 
73
* Vital statistics:
74
 
75
  Alignment:                            8-byte
76
       8 byte alignment is currently hardwired into the design.  This
77
       seems to suffice for all current machines and C compilers.
78
 
79
  Assumed pointer representation:       4 or 8 bytes
80
       Code for 8-byte pointers is untested by me but has worked
81
       reliably by Wolfram Gloger, who contributed most of the
82
       changes supporting this.
83
 
84
  Assumed size_t  representation:       4 or 8 bytes
85
       Note that size_t is allowed to be 4 bytes even if pointers are 8.
86
 
87
  Minimum overhead per allocated chunk: 4 or 8 bytes
88
       Each malloced chunk has a hidden overhead of 4 bytes holding size
89
       and status information.
90
 
91
  Minimum allocated size: 4-byte ptrs:  16 bytes    (including 4 overhead)
92
                          8-byte ptrs:  24/32 bytes (including, 4/8 overhead)
93
 
94
       When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
95
       ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
96
       needed; 4 (8) for a trailing size field
97
       and 8 (16) bytes for free list pointers. Thus, the minimum
98
       allocatable size is 16/24/32 bytes.
99
 
100
       Even a request for zero bytes (i.e., malloc(0)) returns a
101
       pointer to something of the minimum allocatable size.
102
 
103
  Maximum allocated size: 4-byte size_t: 2^31 -  8 bytes
104
                          8-byte size_t: 2^63 - 16 bytes
105
 
106
       It is assumed that (possibly signed) size_t bit values suffice to
107
       represent chunk sizes. `Possibly signed' is due to the fact
108
       that `size_t' may be defined on a system as either a signed or
109
       an unsigned type. To be conservative, values that would appear
110
       as negative numbers are avoided.
111
       Requests for sizes with a negative sign bit will return a
112
       minimum-sized chunk.
113
 
114
  Maximum overhead wastage per allocated chunk: normally 15 bytes
115
 
116
       Alignnment demands, plus the minimum allocatable size restriction
117
       make the normal worst-case wastage 15 bytes (i.e., up to 15
118
       more bytes will be allocated than were requested in malloc), with
119
       two exceptions:
120
         1. Because requests for zero bytes allocate non-zero space,
121
            the worst case wastage for a request of zero bytes is 24 bytes.
122
         2. For requests >= mmap_threshold that are serviced via
123
            mmap(), the worst case wastage is 8 bytes plus the remainder
124
            from a system page (the minimal mmap unit); typically 4096 bytes.
125
 
126
* Limitations
127
 
128
    Here are some features that are NOT currently supported
129
 
130
    * No user-definable hooks for callbacks and the like.
131
    * No automated mechanism for fully checking that all accesses
132
      to malloced memory stay within their bounds.
133
    * No support for compaction.
134
 
135
* Synopsis of compile-time options:
136
 
137
    People have reported using previous versions of this malloc on all
138
    versions of Unix, sometimes by tweaking some of the defines
139
    below. It has been tested most extensively on Solaris and
140
    Linux. It is also reported to work on WIN32 platforms.
141
    People have also reported adapting this malloc for use in
142
    stand-alone embedded systems.
143
 
144
    The implementation is in straight, hand-tuned ANSI C.  Among other
145
    consequences, it uses a lot of macros.  Because of this, to be at
146
    all usable, this code should be compiled using an optimizing compiler
147
    (for example gcc -O2) that can simplify expressions and control
148
    paths.
149
 
150
  __STD_C                  (default: derived from C compiler defines)
151
     Nonzero if using ANSI-standard C compiler, a C++ compiler, or
152
     a C compiler sufficiently close to ANSI to get away with it.
153
  DEBUG                    (default: NOT defined)
154
     Define to enable debugging. Adds fairly extensive assertion-based
155
     checking to help track down memory errors, but noticeably slows down
156
     execution.
157
  SEPARATE_OBJECTS         (default: NOT defined)
158
     Define this to compile into separate .o files.  You must then
159
     compile malloc.c several times, defining a DEFINE_* macro each
160
     time.  The list of DEFINE_* macros appears below.
161
  MALLOC_LOCK              (default: NOT defined)
162
  MALLOC_UNLOCK            (default: NOT defined)
163
     Define these to C expressions which are run to lock and unlock
164
     the malloc data structures.  Calls may be nested; that is,
165
     MALLOC_LOCK may be called more than once before the corresponding
166
     MALLOC_UNLOCK calls.  MALLOC_LOCK must avoid waiting for a lock
167
     that it already holds.
168
  MALLOC_ALIGNMENT          (default: NOT defined)
169
     Define this to 16 if you need 16 byte alignment instead of 8 byte alignment
170
     which is the normal default.
171
  REALLOC_ZERO_BYTES_FREES (default: NOT defined)
172
     Define this if you think that realloc(p, 0) should be equivalent
173
     to free(p). Otherwise, since malloc returns a unique pointer for
174
     malloc(0), so does realloc(p, 0).
175
  HAVE_MEMCPY               (default: defined)
176
     Define if you are not otherwise using ANSI STD C, but still
177
     have memcpy and memset in your C library and want to use them.
178
     Otherwise, simple internal versions are supplied.
179
  USE_MEMCPY               (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
180
     Define as 1 if you want the C library versions of memset and
181
     memcpy called in realloc and calloc (otherwise macro versions are used).
182
     At least on some platforms, the simple macro versions usually
183
     outperform libc versions.
184
  HAVE_MMAP                 (default: defined as 1)
185
     Define to non-zero to optionally make malloc() use mmap() to
186
     allocate very large blocks.
187
  HAVE_MREMAP                 (default: defined as 0 unless Linux libc set)
188
     Define to non-zero to optionally make realloc() use mremap() to
189
     reallocate very large blocks.
190
  malloc_getpagesize        (default: derived from system #includes)
191
     Either a constant or routine call returning the system page size.
192
  HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
193
     Optionally define if you are on a system with a /usr/include/malloc.h
194
     that declares struct mallinfo. It is not at all necessary to
195
     define this even if you do, but will ensure consistency.
196
  INTERNAL_SIZE_T           (default: size_t)
197
     Define to a 32-bit type (probably `unsigned int') if you are on a
198
     64-bit machine, yet do not want or need to allow malloc requests of
199
     greater than 2^31 to be handled. This saves space, especially for
200
     very small chunks.
201
  INTERNAL_LINUX_C_LIB      (default: NOT defined)
202
     Defined only when compiled as part of Linux libc.
203
     Also note that there is some odd internal name-mangling via defines
204
     (for example, internally, `malloc' is named `mALLOc') needed
205
     when compiling in this case. These look funny but don't otherwise
206
     affect anything.
207
  INTERNAL_NEWLIB           (default: NOT defined)
208
     Defined only when compiled as part of the Cygnus newlib
209
     distribution.
210
  WIN32                     (default: undefined)
211
     Define this on MS win (95, nt) platforms to compile in sbrk emulation.
212
  LACKS_UNISTD_H            (default: undefined)
213
     Define this if your system does not have a <unistd.h>.
214
  MORECORE                  (default: sbrk)
215
     The name of the routine to call to obtain more memory from the system.
216
  MORECORE_FAILURE          (default: -1)
217
     The value returned upon failure of MORECORE.
218
  MORECORE_CLEARS           (default 1)
219
     True (1) if the routine mapped to MORECORE zeroes out memory (which
220
     holds for sbrk).
221
  DEFAULT_TRIM_THRESHOLD
222
  DEFAULT_TOP_PAD
223
  DEFAULT_MMAP_THRESHOLD
224
  DEFAULT_MMAP_MAX
225
     Default values of tunable parameters (described in detail below)
226
     controlling interaction with host system routines (sbrk, mmap, etc).
227
     These values may also be changed dynamically via mallopt(). The
228
     preset defaults are those that give best performance for typical
229
     programs/systems.
230
 
231
 
232
*/
233
 
234
 
235
 
236
 
237
/* Preliminaries */
238
 
239
#ifndef __STD_C
240
#ifdef __STDC__
241
#define __STD_C     1
242
#else
243
#if __cplusplus
244
#define __STD_C     1
245
#else
246
#define __STD_C     0
247
#endif /*__cplusplus*/
248
#endif /*__STDC__*/
249
#endif /*__STD_C*/
250
 
251
#ifndef Void_t
252
#if __STD_C
253
#define Void_t      void
254
#else
255
#define Void_t      char
256
#endif
257
#endif /*Void_t*/
258
 
259
#if __STD_C
260
#include <stddef.h>   /* for size_t */
261
#else
262
#include <sys/types.h>
263
#endif
264
 
265
#ifdef __cplusplus
266
extern "C" {
267
#endif
268
 
269
#include <stdio.h>    /* needed for malloc_stats */
270
#include <limits.h>   /* needed for overflow checks */
271
#include <errno.h>    /* needed to set errno to ENOMEM */
272
 
273
#ifdef WIN32
274
#define WIN32_LEAN_AND_MEAN
275
#include <windows.h>
276
#endif
277
 
278
/*
279
  Compile-time options
280
*/
281
 
282
 
283
/*
284
 
285
  Special defines for Cygnus newlib distribution.
286
 
287
 */
288
 
289
#ifdef INTERNAL_NEWLIB
290
 
291
#include <sys/config.h>
292
 
293
/*
294
  In newlib, all the publically visible routines take a reentrancy
295
  pointer.  We don't currently do anything much with it, but we do
296
  pass it to the lock routine.
297
 */
298
 
299
#include <reent.h>
300
 
301
#define POINTER_UINT unsigned _POINTER_INT
302
#define SEPARATE_OBJECTS
303
#define HAVE_MMAP 0
304
#define MORECORE(size) _sbrk_r(reent_ptr, (size))
305
#define MORECORE_CLEARS 0
306
#define MALLOC_LOCK __malloc_lock(reent_ptr)
307
#define MALLOC_UNLOCK __malloc_unlock(reent_ptr)
308
 
309
#ifdef __CYGWIN__
310
# undef _WIN32
311
# undef WIN32
312
#endif
313
 
314
#ifndef _WIN32
315
#ifdef SMALL_MEMORY
316
#define malloc_getpagesize (128)
317
#else
318
#define malloc_getpagesize (4096)
319
#endif
320
#endif
321
 
322
#if __STD_C
323
extern void __malloc_lock(struct _reent *);
324
extern void __malloc_unlock(struct _reent *);
325
#else
326
extern void __malloc_lock();
327
extern void __malloc_unlock();
328
#endif
329
 
330
#if __STD_C
331
#define RARG struct _reent *reent_ptr,
332
#define RONEARG struct _reent *reent_ptr
333
#else
334
#define RARG reent_ptr
335
#define RONEARG reent_ptr
336
#define RDECL struct _reent *reent_ptr;
337
#endif
338
 
339
#define RERRNO reent_ptr->_errno
340
#define RCALL reent_ptr,
341
#define RONECALL reent_ptr
342
 
343
#else /* ! INTERNAL_NEWLIB */
344
 
345
#define POINTER_UINT unsigned long
346
#define RARG
347
#define RONEARG
348
#define RDECL
349
#define RERRNO errno
350
#define RCALL
351
#define RONECALL
352
 
353
#endif /* ! INTERNAL_NEWLIB */
354
 
355
/*
356
    Debugging:
357
 
358
    Because freed chunks may be overwritten with link fields, this
359
    malloc will often die when freed memory is overwritten by user
360
    programs.  This can be very effective (albeit in an annoying way)
361
    in helping track down dangling pointers.
362
 
363
    If you compile with -DDEBUG, a number of assertion checks are
364
    enabled that will catch more memory errors. You probably won't be
365
    able to make much sense of the actual assertion errors, but they
366
    should help you locate incorrectly overwritten memory.  The
367
    checking is fairly extensive, and will slow down execution
368
    noticeably. Calling malloc_stats or mallinfo with DEBUG set will
369
    attempt to check every non-mmapped allocated and free chunk in the
370
    course of computing the summmaries. (By nature, mmapped regions
371
    cannot be checked very much automatically.)
372
 
373
    Setting DEBUG may also be helpful if you are trying to modify
374
    this code. The assertions in the check routines spell out in more
375
    detail the assumptions and invariants underlying the algorithms.
376
 
377
*/
378
 
379
#if DEBUG 
380
#include <assert.h>
381
#else
382
#define assert(x) ((void)0)
383
#endif
384
 
385
 
386
/*
387
  SEPARATE_OBJECTS should be defined if you want each function to go
388
  into a separate .o file.  You must then compile malloc.c once per
389
  function, defining the appropriate DEFINE_ macro.  See below for the
390
  list of macros.
391
 */
392
 
393
#ifndef SEPARATE_OBJECTS
394
#define DEFINE_MALLOC
395
#define DEFINE_FREE
396
#define DEFINE_REALLOC
397
#define DEFINE_CALLOC
398
#define DEFINE_CFREE
399
#define DEFINE_MEMALIGN
400
#define DEFINE_VALLOC
401
#define DEFINE_PVALLOC
402
#define DEFINE_MALLINFO
403
#define DEFINE_MALLOC_STATS
404
#define DEFINE_MALLOC_USABLE_SIZE
405
#define DEFINE_MALLOPT
406
 
407
#define STATIC static
408
#else
409
#define STATIC
410
#endif
411
 
412
/*
413
   Define MALLOC_LOCK and MALLOC_UNLOCK to C expressions to run to
414
   lock and unlock the malloc data structures.  MALLOC_LOCK may be
415
   called recursively.
416
 */
417
 
418
#ifndef MALLOC_LOCK
419
#define MALLOC_LOCK
420
#endif
421
 
422
#ifndef MALLOC_UNLOCK
423
#define MALLOC_UNLOCK
424
#endif
425
 
426
/*
427
  INTERNAL_SIZE_T is the word-size used for internal bookkeeping
428
  of chunk sizes. On a 64-bit machine, you can reduce malloc
429
  overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
430
  at the expense of not being able to handle requests greater than
431
  2^31. This limitation is hardly ever a concern; you are encouraged
432
  to set this. However, the default version is the same as size_t.
433
*/
434
 
435
#ifndef INTERNAL_SIZE_T
436
#define INTERNAL_SIZE_T size_t
437
#endif
438
 
439
/*
440
  Following is needed on implementations whereby long > size_t.
441
  The problem is caused because the code performs subtractions of
442
  size_t values and stores the result in long values.  In the case
443
  where long > size_t and the first value is actually less than
444
  the second value, the resultant value is positive.  For example,
445
  (long)(x - y) where x = 0 and y is 1 ends up being 0x00000000FFFFFFFF
446
  which is 2*31 - 1 instead of 0xFFFFFFFFFFFFFFFF.  This is due to the
447
  fact that assignment from unsigned to signed won't sign extend.
448
*/
449
 
450
#define long_sub_size_t(x, y)                           \
451
  (sizeof (long) > sizeof (INTERNAL_SIZE_T) && x < y    \
452
   ? -(long) (y - x)                                    \
453
   : (long) (x - y))
454
 
455
/*
456
  REALLOC_ZERO_BYTES_FREES should be set if a call to
457
  realloc with zero bytes should be the same as a call to free.
458
  Some people think it should. Otherwise, since this malloc
459
  returns a unique pointer for malloc(0), so does realloc(p, 0).
460
*/
461
 
462
 
463
/*   #define REALLOC_ZERO_BYTES_FREES */
464
 
465
 
466
/*
467
  WIN32 causes an emulation of sbrk to be compiled in
468
  mmap-based options are not currently supported in WIN32.
469
*/
470
 
471
/* #define WIN32 */
472
#ifdef WIN32
473
#define MORECORE wsbrk
474
#define HAVE_MMAP 0
475
#endif
476
 
477
 
478
/*
479
  HAVE_MEMCPY should be defined if you are not otherwise using
480
  ANSI STD C, but still have memcpy and memset in your C library
481
  and want to use them in calloc and realloc. Otherwise simple
482
  macro versions are defined here.
483
 
484
  USE_MEMCPY should be defined as 1 if you actually want to
485
  have memset and memcpy called. People report that the macro
486
  versions are often enough faster than libc versions on many
487
  systems that it is better to use them.
488
 
489
*/
490
 
491
#define HAVE_MEMCPY 
492
 
493
/* Although the original macro is called USE_MEMCPY, newlib actually
494
   uses memmove to handle cases whereby a platform's memcpy implementation
495
   copies backwards and thus destructive overlap may occur in realloc
496
   whereby we are reclaiming free memory prior to the old allocation.  */
497
#ifndef USE_MEMCPY
498
#ifdef HAVE_MEMCPY
499
#define USE_MEMCPY 1
500
#else
501
#define USE_MEMCPY 0
502
#endif
503
#endif
504
 
505
#if (__STD_C || defined(HAVE_MEMCPY)) 
506
 
507
#if __STD_C
508
void* memset(void*, int, size_t);
509
void* memcpy(void*, const void*, size_t);
510
void* memmove(void*, const void*, size_t);
511
#else
512
Void_t* memset();
513
Void_t* memcpy();
514
Void_t* memmove();
515
#endif
516
#endif
517
 
518
#if USE_MEMCPY
519
 
520
/* The following macros are only invoked with (2n+1)-multiples of
521
   INTERNAL_SIZE_T units, with a positive integer n. This is exploited
522
   for fast inline execution when n is small. */
523
 
524
#define MALLOC_ZERO(charp, nbytes)                                            \
525
do {                                                                          \
526
  INTERNAL_SIZE_T mzsz = (nbytes);                                            \
527
  if(mzsz <= 9*sizeof(mzsz)) {                                                \
528
    INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp);                         \
529
    if(mzsz >= 5*sizeof(mzsz)) {     *mz++ = 0;                               \
530
                                     *mz++ = 0;                               \
531
      if(mzsz >= 7*sizeof(mzsz)) {   *mz++ = 0;                               \
532
                                     *mz++ = 0;                               \
533
        if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0;                               \
534
                                     *mz++ = 0; }}}                           \
535
                                     *mz++ = 0;                               \
536
                                     *mz++ = 0;                               \
537
                                     *mz   = 0;                               \
538
  } else memset((charp), 0, mzsz);                                            \
539
} while(0)
540
 
541
#define MALLOC_COPY(dest,src,nbytes)                                          \
542
do {                                                                          \
543
  INTERNAL_SIZE_T mcsz = (nbytes);                                            \
544
  if(mcsz <= 9*sizeof(mcsz)) {                                                \
545
    INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src);                        \
546
    INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest);                       \
547
    if(mcsz >= 5*sizeof(mcsz)) {     *mcdst++ = *mcsrc++;                     \
548
                                     *mcdst++ = *mcsrc++;                     \
549
      if(mcsz >= 7*sizeof(mcsz)) {   *mcdst++ = *mcsrc++;                     \
550
                                     *mcdst++ = *mcsrc++;                     \
551
        if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++;                     \
552
                                     *mcdst++ = *mcsrc++; }}}                 \
553
                                     *mcdst++ = *mcsrc++;                     \
554
                                     *mcdst++ = *mcsrc++;                     \
555
                                     *mcdst   = *mcsrc  ;                     \
556
  } else memmove(dest, src, mcsz);                                             \
557
} while(0)
558
 
559
#else /* !USE_MEMCPY */
560
 
561
/* Use Duff's device for good zeroing/copying performance. */
562
 
563
#define MALLOC_ZERO(charp, nbytes)                                            \
564
do {                                                                          \
565
  INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp);                           \
566
  long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn;                         \
567
  if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
568
  switch (mctmp) {                                                            \
569
    case 0: for(;;) { *mzp++ = 0;                                             \
570
    case 7:           *mzp++ = 0;                                             \
571
    case 6:           *mzp++ = 0;                                             \
572
    case 5:           *mzp++ = 0;                                             \
573
    case 4:           *mzp++ = 0;                                             \
574
    case 3:           *mzp++ = 0;                                             \
575
    case 2:           *mzp++ = 0;                                             \
576
    case 1:           *mzp++ = 0; if(mcn <= 0) break; mcn--; }                \
577
  }                                                                           \
578
} while(0)
579
 
580
#define MALLOC_COPY(dest,src,nbytes)                                          \
581
do {                                                                          \
582
  INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src;                            \
583
  INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest;                           \
584
  long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn;                         \
585
  if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
586
  switch (mctmp) {                                                            \
587
    case 0: for(;;) { *mcdst++ = *mcsrc++;                                    \
588
    case 7:           *mcdst++ = *mcsrc++;                                    \
589
    case 6:           *mcdst++ = *mcsrc++;                                    \
590
    case 5:           *mcdst++ = *mcsrc++;                                    \
591
    case 4:           *mcdst++ = *mcsrc++;                                    \
592
    case 3:           *mcdst++ = *mcsrc++;                                    \
593
    case 2:           *mcdst++ = *mcsrc++;                                    \
594
    case 1:           *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; }       \
595
  }                                                                           \
596
} while(0)
597
 
598
#endif
599
 
600
 
601
/*
602
  Define HAVE_MMAP to optionally make malloc() use mmap() to
603
  allocate very large blocks.  These will be returned to the
604
  operating system immediately after a free().
605
*/
606
 
607
#ifndef HAVE_MMAP
608
#define HAVE_MMAP 1
609
#endif
610
 
611
/*
612
  Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
613
  large blocks.  This is currently only possible on Linux with
614
  kernel versions newer than 1.3.77.
615
*/
616
 
617
#ifndef HAVE_MREMAP
618
#ifdef INTERNAL_LINUX_C_LIB
619
#define HAVE_MREMAP 1
620
#else
621
#define HAVE_MREMAP 0
622
#endif
623
#endif
624
 
625
#if HAVE_MMAP
626
 
627
#include <unistd.h>
628
#include <fcntl.h>
629
#include <sys/mman.h>
630
 
631
#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
632
#define MAP_ANONYMOUS MAP_ANON
633
#endif
634
 
635
#endif /* HAVE_MMAP */
636
 
637
/*
638
  Access to system page size. To the extent possible, this malloc
639
  manages memory from the system in page-size units.
640
 
641
  The following mechanics for getpagesize were adapted from
642
  bsd/gnu getpagesize.h
643
*/
644
 
645
#ifndef LACKS_UNISTD_H
646
#  include <unistd.h>
647
#endif
648
 
649
#ifndef malloc_getpagesize
650
#  ifdef _SC_PAGESIZE         /* some SVR4 systems omit an underscore */
651
#    ifndef _SC_PAGE_SIZE
652
#      define _SC_PAGE_SIZE _SC_PAGESIZE
653
#    endif
654
#  endif
655
#  ifdef _SC_PAGE_SIZE
656
#    define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
657
#  else
658
#    if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
659
       extern size_t getpagesize();
660
#      define malloc_getpagesize getpagesize()
661
#    else
662
#      include <sys/param.h>
663
#      ifdef EXEC_PAGESIZE
664
#        define malloc_getpagesize EXEC_PAGESIZE
665
#      else
666
#        ifdef NBPG
667
#          ifndef CLSIZE
668
#            define malloc_getpagesize NBPG
669
#          else
670
#            define malloc_getpagesize (NBPG * CLSIZE)
671
#          endif
672
#        else 
673
#          ifdef NBPC
674
#            define malloc_getpagesize NBPC
675
#          else
676
#            ifdef PAGESIZE
677
#              define malloc_getpagesize PAGESIZE
678
#            else
679
#              define malloc_getpagesize (4096) /* just guess */
680
#            endif
681
#          endif
682
#        endif 
683
#      endif
684
#    endif 
685
#  endif
686
#endif
687
 
688
 
689
 
690
/*
691
 
692
  This version of malloc supports the standard SVID/XPG mallinfo
693
  routine that returns a struct containing the same kind of
694
  information you can get from malloc_stats. It should work on
695
  any SVID/XPG compliant system that has a /usr/include/malloc.h
696
  defining struct mallinfo. (If you'd like to install such a thing
697
  yourself, cut out the preliminary declarations as described above
698
  and below and save them in a malloc.h file. But there's no
699
  compelling reason to bother to do this.)
700
 
701
  The main declaration needed is the mallinfo struct that is returned
702
  (by-copy) by mallinfo().  The SVID/XPG malloinfo struct contains a
703
  bunch of fields, most of which are not even meaningful in this
704
  version of malloc. Some of these fields are are instead filled by
705
  mallinfo() with other numbers that might possibly be of interest.
706
 
707
  HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
708
  /usr/include/malloc.h file that includes a declaration of struct
709
  mallinfo.  If so, it is included; else an SVID2/XPG2 compliant
710
  version is declared below.  These must be precisely the same for
711
  mallinfo() to work.
712
 
713
*/
714
 
715
/* #define HAVE_USR_INCLUDE_MALLOC_H */
716
 
717
#if HAVE_USR_INCLUDE_MALLOC_H
718
#include "/usr/include/malloc.h"
719
#else
720
 
721
/* SVID2/XPG mallinfo structure */
722
 
723
struct mallinfo {
724
  int arena;    /* total space allocated from system */
725
  int ordblks;  /* number of non-inuse chunks */
726
  int smblks;   /* unused -- always zero */
727
  int hblks;    /* number of mmapped regions */
728
  int hblkhd;   /* total space in mmapped regions */
729
  int usmblks;  /* unused -- always zero */
730
  int fsmblks;  /* unused -- always zero */
731
  int uordblks; /* total allocated space */
732
  int fordblks; /* total non-inuse space */
733
  int keepcost; /* top-most, releasable (via malloc_trim) space */
734
};
735
 
736
/* SVID2/XPG mallopt options */
737
 
738
#define M_MXFAST  1    /* UNUSED in this malloc */
739
#define M_NLBLKS  2    /* UNUSED in this malloc */
740
#define M_GRAIN   3    /* UNUSED in this malloc */
741
#define M_KEEP    4    /* UNUSED in this malloc */
742
 
743
#endif
744
 
745
/* mallopt options that actually do something */
746
 
747
#define M_TRIM_THRESHOLD    -1
748
#define M_TOP_PAD           -2
749
#define M_MMAP_THRESHOLD    -3
750
#define M_MMAP_MAX          -4
751
 
752
 
753
 
754
#ifndef DEFAULT_TRIM_THRESHOLD
755
#define DEFAULT_TRIM_THRESHOLD (128L * 1024L)
756
#endif
757
 
758
/*
759
    M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
760
      to keep before releasing via malloc_trim in free().
761
 
762
      Automatic trimming is mainly useful in long-lived programs.
763
      Because trimming via sbrk can be slow on some systems, and can
764
      sometimes be wasteful (in cases where programs immediately
765
      afterward allocate more large chunks) the value should be high
766
      enough so that your overall system performance would improve by
767
      releasing.
768
 
769
      The trim threshold and the mmap control parameters (see below)
770
      can be traded off with one another. Trimming and mmapping are
771
      two different ways of releasing unused memory back to the
772
      system. Between these two, it is often possible to keep
773
      system-level demands of a long-lived program down to a bare
774
      minimum. For example, in one test suite of sessions measuring
775
      the XF86 X server on Linux, using a trim threshold of 128K and a
776
      mmap threshold of 192K led to near-minimal long term resource
777
      consumption.
778
 
779
      If you are using this malloc in a long-lived program, it should
780
      pay to experiment with these values.  As a rough guide, you
781
      might set to a value close to the average size of a process
782
      (program) running on your system.  Releasing this much memory
783
      would allow such a process to run in memory.  Generally, it's
784
      worth it to tune for trimming rather tham memory mapping when a
785
      program undergoes phases where several large chunks are
786
      allocated and released in ways that can reuse each other's
787
      storage, perhaps mixed with phases where there are no such
788
      chunks at all.  And in well-behaved long-lived programs,
789
      controlling release of large blocks via trimming versus mapping
790
      is usually faster.
791
 
792
      However, in most programs, these parameters serve mainly as
793
      protection against the system-level effects of carrying around
794
      massive amounts of unneeded memory. Since frequent calls to
795
      sbrk, mmap, and munmap otherwise degrade performance, the default
796
      parameters are set to relatively high values that serve only as
797
      safeguards.
798
 
799
      The default trim value is high enough to cause trimming only in
800
      fairly extreme (by current memory consumption standards) cases.
801
      It must be greater than page size to have any useful effect.  To
802
      disable trimming completely, you can set to (unsigned long)(-1);
803
 
804
 
805
*/
806
 
807
 
808
#ifndef DEFAULT_TOP_PAD
809
#define DEFAULT_TOP_PAD        (0)
810
#endif
811
 
812
/*
813
    M_TOP_PAD is the amount of extra `padding' space to allocate or
814
      retain whenever sbrk is called. It is used in two ways internally:
815
 
816
      * When sbrk is called to extend the top of the arena to satisfy
817
        a new malloc request, this much padding is added to the sbrk
818
        request.
819
 
820
      * When malloc_trim is called automatically from free(),
821
        it is used as the `pad' argument.
822
 
823
      In both cases, the actual amount of padding is rounded
824
      so that the end of the arena is always a system page boundary.
825
 
826
      The main reason for using padding is to avoid calling sbrk so
827
      often. Having even a small pad greatly reduces the likelihood
828
      that nearly every malloc request during program start-up (or
829
      after trimming) will invoke sbrk, which needlessly wastes
830
      time.
831
 
832
      Automatic rounding-up to page-size units is normally sufficient
833
      to avoid measurable overhead, so the default is 0.  However, in
834
      systems where sbrk is relatively slow, it can pay to increase
835
      this value, at the expense of carrying around more memory than
836
      the program needs.
837
 
838
*/
839
 
840
 
841
#ifndef DEFAULT_MMAP_THRESHOLD
842
#define DEFAULT_MMAP_THRESHOLD (128 * 1024)
843
#endif
844
 
845
/*
846
 
847
    M_MMAP_THRESHOLD is the request size threshold for using mmap()
848
      to service a request. Requests of at least this size that cannot
849
      be allocated using already-existing space will be serviced via mmap.
850
      (If enough normal freed space already exists it is used instead.)
851
 
852
      Using mmap segregates relatively large chunks of memory so that
853
      they can be individually obtained and released from the host
854
      system. A request serviced through mmap is never reused by any
855
      other request (at least not directly; the system may just so
856
      happen to remap successive requests to the same locations).
857
 
858
      Segregating space in this way has the benefit that mmapped space
859
      can ALWAYS be individually released back to the system, which
860
      helps keep the system level memory demands of a long-lived
861
      program low. Mapped memory can never become `locked' between
862
      other chunks, as can happen with normally allocated chunks, which
863
      menas that even trimming via malloc_trim would not release them.
864
 
865
      However, it has the disadvantages that:
866
 
867
         1. The space cannot be reclaimed, consolidated, and then
868
            used to service later requests, as happens with normal chunks.
869
         2. It can lead to more wastage because of mmap page alignment
870
            requirements
871
         3. It causes malloc performance to be more dependent on host
872
            system memory management support routines which may vary in
873
            implementation quality and may impose arbitrary
874
            limitations. Generally, servicing a request via normal
875
            malloc steps is faster than going through a system's mmap.
876
 
877
      All together, these considerations should lead you to use mmap
878
      only for relatively large requests.
879
 
880
 
881
*/
882
 
883
 
884
 
885
#ifndef DEFAULT_MMAP_MAX
886
#if HAVE_MMAP
887
#define DEFAULT_MMAP_MAX       (64)
888
#else
889
#define DEFAULT_MMAP_MAX       (0)
890
#endif
891
#endif
892
 
893
/*
894
    M_MMAP_MAX is the maximum number of requests to simultaneously
895
      service using mmap. This parameter exists because:
896
 
897
         1. Some systems have a limited number of internal tables for
898
            use by mmap.
899
         2. In most systems, overreliance on mmap can degrade overall
900
            performance.
901
         3. If a program allocates many large regions, it is probably
902
            better off using normal sbrk-based allocation routines that
903
            can reclaim and reallocate normal heap memory. Using a
904
            small value allows transition into this mode after the
905
            first few allocations.
906
 
907
      Setting to 0 disables all use of mmap.  If HAVE_MMAP is not set,
908
      the default value is 0, and attempts to set it to non-zero values
909
      in mallopt will fail.
910
*/
911
 
912
 
913
 
914
 
915
/*
916
 
917
  Special defines for linux libc
918
 
919
  Except when compiled using these special defines for Linux libc
920
  using weak aliases, this malloc is NOT designed to work in
921
  multithreaded applications.  No semaphores or other concurrency
922
  control are provided to ensure that multiple malloc or free calls
923
  don't run at the same time, which could be disasterous. A single
924
  semaphore could be used across malloc, realloc, and free (which is
925
  essentially the effect of the linux weak alias approach). It would
926
  be hard to obtain finer granularity.
927
 
928
*/
929
 
930
 
931
#ifdef INTERNAL_LINUX_C_LIB
932
 
933
#if __STD_C
934
 
935
Void_t * __default_morecore_init (ptrdiff_t);
936
Void_t *(*__morecore)(ptrdiff_t) = __default_morecore_init;
937
 
938
#else
939
 
940
Void_t * __default_morecore_init ();
941
Void_t *(*__morecore)() = __default_morecore_init;
942
 
943
#endif
944
 
945
#define MORECORE (*__morecore)
946
#define MORECORE_FAILURE 0
947
#define MORECORE_CLEARS 1 
948
 
949
#else /* INTERNAL_LINUX_C_LIB */
950
 
951
#ifndef INTERNAL_NEWLIB
952
#if __STD_C
953
extern Void_t*     sbrk(ptrdiff_t);
954
#else
955
extern Void_t*     sbrk();
956
#endif
957
#endif
958
 
959
#ifndef MORECORE
960
#define MORECORE sbrk
961
#endif
962
 
963
#ifndef MORECORE_FAILURE
964
#define MORECORE_FAILURE -1
965
#endif
966
 
967
#ifndef MORECORE_CLEARS
968
#define MORECORE_CLEARS 1
969
#endif
970
 
971
#endif /* INTERNAL_LINUX_C_LIB */
972
 
973
#if defined(INTERNAL_LINUX_C_LIB) && defined(__ELF__)
974
 
975
#define cALLOc          __libc_calloc
976
#define fREe            __libc_free
977
#define mALLOc          __libc_malloc
978
#define mEMALIGn        __libc_memalign
979
#define rEALLOc         __libc_realloc
980
#define vALLOc          __libc_valloc
981
#define pvALLOc         __libc_pvalloc
982
#define mALLINFo        __libc_mallinfo
983
#define mALLOPt         __libc_mallopt
984
 
985
#pragma weak calloc = __libc_calloc
986
#pragma weak free = __libc_free
987
#pragma weak cfree = __libc_free
988
#pragma weak malloc = __libc_malloc
989
#pragma weak memalign = __libc_memalign
990
#pragma weak realloc = __libc_realloc
991
#pragma weak valloc = __libc_valloc
992
#pragma weak pvalloc = __libc_pvalloc
993
#pragma weak mallinfo = __libc_mallinfo
994
#pragma weak mallopt = __libc_mallopt
995
 
996
#else
997
 
998
#ifdef INTERNAL_NEWLIB
999
 
1000
#define cALLOc          _calloc_r
1001
#define fREe            _free_r
1002
#define mALLOc          _malloc_r
1003
#define mEMALIGn        _memalign_r
1004
#define rEALLOc         _realloc_r
1005
#define vALLOc          _valloc_r
1006
#define pvALLOc         _pvalloc_r
1007
#define mALLINFo        _mallinfo_r
1008
#define mALLOPt         _mallopt_r
1009
 
1010
#define malloc_stats                    _malloc_stats_r
1011
#define malloc_trim                     _malloc_trim_r
1012
#define malloc_usable_size              _malloc_usable_size_r
1013
 
1014
#define malloc_update_mallinfo          __malloc_update_mallinfo
1015
 
1016
#define malloc_av_                      __malloc_av_
1017
#define malloc_current_mallinfo         __malloc_current_mallinfo
1018
#define malloc_max_sbrked_mem           __malloc_max_sbrked_mem
1019
#define malloc_max_total_mem            __malloc_max_total_mem
1020
#define malloc_sbrk_base                __malloc_sbrk_base
1021
#define malloc_top_pad                  __malloc_top_pad
1022
#define malloc_trim_threshold           __malloc_trim_threshold
1023
 
1024
#else /* ! INTERNAL_NEWLIB */
1025
 
1026
#define cALLOc          calloc
1027
#define fREe            free
1028
#define mALLOc          malloc
1029
#define mEMALIGn        memalign
1030
#define rEALLOc         realloc
1031
#define vALLOc          valloc
1032
#define pvALLOc         pvalloc
1033
#define mALLINFo        mallinfo
1034
#define mALLOPt         mallopt
1035
 
1036
#endif /* ! INTERNAL_NEWLIB */
1037
#endif
1038
 
1039
/* Public routines */
1040
 
1041
#if __STD_C
1042
 
1043
Void_t* mALLOc(RARG size_t);
1044
void    fREe(RARG Void_t*);
1045
Void_t* rEALLOc(RARG Void_t*, size_t);
1046
Void_t* mEMALIGn(RARG size_t, size_t);
1047
Void_t* vALLOc(RARG size_t);
1048
Void_t* pvALLOc(RARG size_t);
1049
Void_t* cALLOc(RARG size_t, size_t);
1050
void    cfree(Void_t*);
1051
int     malloc_trim(RARG size_t);
1052
size_t  malloc_usable_size(RARG Void_t*);
1053
void    malloc_stats(RONEARG);
1054
int     mALLOPt(RARG int, int);
1055
struct mallinfo mALLINFo(RONEARG);
1056
#else
1057
Void_t* mALLOc();
1058
void    fREe();
1059
Void_t* rEALLOc();
1060
Void_t* mEMALIGn();
1061
Void_t* vALLOc();
1062
Void_t* pvALLOc();
1063
Void_t* cALLOc();
1064
void    cfree();
1065
int     malloc_trim();
1066
size_t  malloc_usable_size();
1067
void    malloc_stats();
1068
int     mALLOPt();
1069
struct mallinfo mALLINFo();
1070
#endif
1071
 
1072
 
1073
#ifdef __cplusplus
1074
};  /* end of extern "C" */
1075
#endif
1076
 
1077
/* ---------- To make a malloc.h, end cutting here ------------ */
1078
 
1079
 
1080
/*
1081
  Emulation of sbrk for WIN32
1082
  All code within the ifdef WIN32 is untested by me.
1083
*/
1084
 
1085
 
1086
#ifdef WIN32
1087
 
1088
#define AlignPage(add) (((add) + (malloc_getpagesize-1)) & \
1089
~(malloc_getpagesize-1))
1090
 
1091
/* resrve 64MB to insure large contiguous space */
1092
#define RESERVED_SIZE (1024*1024*64)
1093
#define NEXT_SIZE (2048*1024)
1094
#define TOP_MEMORY ((unsigned long)2*1024*1024*1024)
1095
 
1096
struct GmListElement;
1097
typedef struct GmListElement GmListElement;
1098
 
1099
struct GmListElement
1100
{
1101
        GmListElement* next;
1102
        void* base;
1103
};
1104
 
1105
static GmListElement* head = 0;
1106
static unsigned int gNextAddress = 0;
1107
static unsigned int gAddressBase = 0;
1108
static unsigned int gAllocatedSize = 0;
1109
 
1110
static
1111
GmListElement* makeGmListElement (void* bas)
1112
{
1113
        GmListElement* this;
1114
        this = (GmListElement*)(void*)LocalAlloc (0, sizeof (GmListElement));
1115
        ASSERT (this);
1116
        if (this)
1117
        {
1118
                this->base = bas;
1119
                this->next = head;
1120
                head = this;
1121
        }
1122
        return this;
1123
}
1124
 
1125
void gcleanup ()
1126
{
1127
        BOOL rval;
1128
        ASSERT ( (head == NULL) || (head->base == (void*)gAddressBase));
1129
        if (gAddressBase && (gNextAddress - gAddressBase))
1130
        {
1131
                rval = VirtualFree ((void*)gAddressBase,
1132
                                                        gNextAddress - gAddressBase,
1133
                                                        MEM_DECOMMIT);
1134
        ASSERT (rval);
1135
        }
1136
        while (head)
1137
        {
1138
                GmListElement* next = head->next;
1139
                rval = VirtualFree (head->base, 0, MEM_RELEASE);
1140
                ASSERT (rval);
1141
                LocalFree (head);
1142
                head = next;
1143
        }
1144
}
1145
 
1146
static
1147
void* findRegion (void* start_address, unsigned long size)
1148
{
1149
        MEMORY_BASIC_INFORMATION info;
1150
        while ((unsigned long)start_address < TOP_MEMORY)
1151
        {
1152
                VirtualQuery (start_address, &info, sizeof (info));
1153
                if (info.State != MEM_FREE)
1154
                        start_address = (char*)info.BaseAddress + info.RegionSize;
1155
                else if (info.RegionSize >= size)
1156
                        return start_address;
1157
                else
1158
                        start_address = (char*)info.BaseAddress + info.RegionSize;
1159
        }
1160
        return NULL;
1161
 
1162
}
1163
 
1164
 
1165
void* wsbrk (long size)
1166
{
1167
        void* tmp;
1168
        if (size > 0)
1169
        {
1170
                if (gAddressBase == 0)
1171
                {
1172
                        gAllocatedSize = max (RESERVED_SIZE, AlignPage (size));
1173
                        gNextAddress = gAddressBase =
1174
                                (unsigned int)VirtualAlloc (NULL, gAllocatedSize,
1175
                                                                                        MEM_RESERVE, PAGE_NOACCESS);
1176
                } else if (AlignPage (gNextAddress + size) > (gAddressBase +
1177
gAllocatedSize))
1178
                {
1179
                        long new_size = max (NEXT_SIZE, AlignPage (size));
1180
                        void* new_address = (void*)(gAddressBase+gAllocatedSize);
1181
                        do
1182
                        {
1183
                                new_address = findRegion (new_address, new_size);
1184
 
1185
                                if (new_address == 0)
1186
                                        return (void*)-1;
1187
 
1188
                                gAddressBase = gNextAddress =
1189
                                        (unsigned int)VirtualAlloc (new_address, new_size,
1190
                                                                                                MEM_RESERVE, PAGE_NOACCESS);
1191
                                // repeat in case of race condition
1192
                                // The region that we found has been snagged 
1193
                                // by another thread
1194
                        }
1195
                        while (gAddressBase == 0);
1196
 
1197
                        ASSERT (new_address == (void*)gAddressBase);
1198
 
1199
                        gAllocatedSize = new_size;
1200
 
1201
                        if (!makeGmListElement ((void*)gAddressBase))
1202
                                return (void*)-1;
1203
                }
1204
                if ((size + gNextAddress) > AlignPage (gNextAddress))
1205
                {
1206
                        void* res;
1207
                        res = VirtualAlloc ((void*)AlignPage (gNextAddress),
1208
                                                                (size + gNextAddress -
1209
                                                                 AlignPage (gNextAddress)),
1210
                                                                MEM_COMMIT, PAGE_READWRITE);
1211
                        if (res == 0)
1212
                                return (void*)-1;
1213
                }
1214
                tmp = (void*)gNextAddress;
1215
                gNextAddress = (unsigned int)tmp + size;
1216
                return tmp;
1217
        }
1218
        else if (size < 0)
1219
        {
1220
                unsigned int alignedGoal = AlignPage (gNextAddress + size);
1221
                /* Trim by releasing the virtual memory */
1222
                if (alignedGoal >= gAddressBase)
1223
                {
1224
                        VirtualFree ((void*)alignedGoal, gNextAddress - alignedGoal,
1225
                                                 MEM_DECOMMIT);
1226
                        gNextAddress = gNextAddress + size;
1227
                        return (void*)gNextAddress;
1228
                }
1229
                else
1230
                {
1231
                        VirtualFree ((void*)gAddressBase, gNextAddress - gAddressBase,
1232
                                                 MEM_DECOMMIT);
1233
                        gNextAddress = gAddressBase;
1234
                        return (void*)-1;
1235
                }
1236
        }
1237
        else
1238
        {
1239
                return (void*)gNextAddress;
1240
        }
1241
}
1242
 
1243
#endif
1244
 
1245
 
1246
 
1247
/*
1248
  Type declarations
1249
*/
1250
 
1251
 
1252
struct malloc_chunk
1253
{
1254
  INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
1255
  INTERNAL_SIZE_T size;      /* Size in bytes, including overhead. */
1256
  struct malloc_chunk* fd;   /* double links -- used only if free. */
1257
  struct malloc_chunk* bk;
1258
};
1259
 
1260
typedef struct malloc_chunk* mchunkptr;
1261
 
1262
/*
1263
 
1264
   malloc_chunk details:
1265
 
1266
    (The following includes lightly edited explanations by Colin Plumb.)
1267
 
1268
    Chunks of memory are maintained using a `boundary tag' method as
1269
    described in e.g., Knuth or Standish.  (See the paper by Paul
1270
    Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1271
    survey of such techniques.)  Sizes of free chunks are stored both
1272
    in the front of each chunk and at the end.  This makes
1273
    consolidating fragmented chunks into bigger chunks very fast.  The
1274
    size fields also hold bits representing whether chunks are free or
1275
    in use.
1276
 
1277
    An allocated chunk looks like this:
1278
 
1279
 
1280
    chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1281
            |             Size of previous chunk, if allocated            | |
1282
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1283
            |             Size of chunk, in bytes                         |P|
1284
      mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1285
            |             User data starts here...                          .
1286
            .                                                               .
1287
            .             (malloc_usable_space() bytes)                     .
1288
            .                                                               |
1289
nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1290
            |             Size of chunk                                     |
1291
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1292
 
1293
 
1294
    Where "chunk" is the front of the chunk for the purpose of most of
1295
    the malloc code, but "mem" is the pointer that is returned to the
1296
    user.  "Nextchunk" is the beginning of the next contiguous chunk.
1297
 
1298
    Chunks always begin on even word boundries, so the mem portion
1299
    (which is returned to the user) is also on an even word boundary, and
1300
    thus double-word aligned.
1301
 
1302
    Free chunks are stored in circular doubly-linked lists, and look like this:
1303
 
1304
    chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1305
            |             Size of previous chunk                            |
1306
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1307
    `head:' |             Size of chunk, in bytes                         |P|
1308
      mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1309
            |             Forward pointer to next chunk in list             |
1310
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1311
            |             Back pointer to previous chunk in list            |
1312
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1313
            |             Unused space (may be 0 bytes long)                .
1314
            .                                                               .
1315
            .                                                               |
1316
nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1317
    `foot:' |             Size of chunk, in bytes                           |
1318
            +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1319
 
1320
    The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1321
    chunk size (which is always a multiple of two words), is an in-use
1322
    bit for the *previous* chunk.  If that bit is *clear*, then the
1323
    word before the current chunk size contains the previous chunk
1324
    size, and can be used to find the front of the previous chunk.
1325
    (The very first chunk allocated always has this bit set,
1326
    preventing access to non-existent (or non-owned) memory.)
1327
 
1328
    Note that the `foot' of the current chunk is actually represented
1329
    as the prev_size of the NEXT chunk. (This makes it easier to
1330
    deal with alignments etc).
1331
 
1332
    The two exceptions to all this are
1333
 
1334
     1. The special chunk `top', which doesn't bother using the
1335
        trailing size field since there is no
1336
        next contiguous chunk that would have to index off it. (After
1337
        initialization, `top' is forced to always exist.  If it would
1338
        become less than MINSIZE bytes long, it is replenished via
1339
        malloc_extend_top.)
1340
 
1341
     2. Chunks allocated via mmap, which have the second-lowest-order
1342
        bit (IS_MMAPPED) set in their size fields.  Because they are
1343
        never merged or traversed from any other chunk, they have no
1344
        foot size or inuse information.
1345
 
1346
    Available chunks are kept in any of several places (all declared below):
1347
 
1348
    * `av': An array of chunks serving as bin headers for consolidated
1349
       chunks. Each bin is doubly linked.  The bins are approximately
1350
       proportionally (log) spaced.  There are a lot of these bins
1351
       (128). This may look excessive, but works very well in
1352
       practice.  All procedures maintain the invariant that no
1353
       consolidated chunk physically borders another one. Chunks in
1354
       bins are kept in size order, with ties going to the
1355
       approximately least recently used chunk.
1356
 
1357
       The chunks in each bin are maintained in decreasing sorted order by
1358
       size.  This is irrelevant for the small bins, which all contain
1359
       the same-sized chunks, but facilitates best-fit allocation for
1360
       larger chunks. (These lists are just sequential. Keeping them in
1361
       order almost never requires enough traversal to warrant using
1362
       fancier ordered data structures.)  Chunks of the same size are
1363
       linked with the most recently freed at the front, and allocations
1364
       are taken from the back.  This results in LRU or FIFO allocation
1365
       order, which tends to give each chunk an equal opportunity to be
1366
       consolidated with adjacent freed chunks, resulting in larger free
1367
       chunks and less fragmentation.
1368
 
1369
    * `top': The top-most available chunk (i.e., the one bordering the
1370
       end of available memory) is treated specially. It is never
1371
       included in any bin, is used only if no other chunk is
1372
       available, and is released back to the system if it is very
1373
       large (see M_TRIM_THRESHOLD).
1374
 
1375
    * `last_remainder': A bin holding only the remainder of the
1376
       most recently split (non-top) chunk. This bin is checked
1377
       before other non-fitting chunks, so as to provide better
1378
       locality for runs of sequentially allocated chunks.
1379
 
1380
    *  Implicitly, through the host system's memory mapping tables.
1381
       If supported, requests greater than a threshold are usually
1382
       serviced via calls to mmap, and then later released via munmap.
1383
 
1384
*/
1385
 
1386
 
1387
 
1388
 
1389
 
1390
 
1391
/*  sizes, alignments */
1392
 
1393
#define SIZE_SZ                (sizeof(INTERNAL_SIZE_T))
1394
#ifndef MALLOC_ALIGNMENT
1395
#define MALLOC_ALIGN           8
1396
#define MALLOC_ALIGNMENT       (SIZE_SZ < 4 ? 8 : (SIZE_SZ + SIZE_SZ))
1397
#else
1398
#define MALLOC_ALIGN           MALLOC_ALIGNMENT
1399
#endif
1400
#define MALLOC_ALIGN_MASK      (MALLOC_ALIGNMENT - 1)
1401
#define MINSIZE                (sizeof(struct malloc_chunk))
1402
 
1403
/* conversion from malloc headers to user pointers, and back */
1404
 
1405
#define chunk2mem(p)   ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1406
#define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1407
 
1408
/* pad request bytes into a usable size */
1409
 
1410
#define request2size(req) \
1411
 (((unsigned long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1412
  (unsigned long)(MINSIZE + MALLOC_ALIGN_MASK)) ? ((MINSIZE + MALLOC_ALIGN_MASK) & ~(MALLOC_ALIGN_MASK)) : \
1413
   (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1414
 
1415
/* Check if m has acceptable alignment */
1416
 
1417
#define aligned_OK(m)    (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1418
 
1419
 
1420
 
1421
 
1422
/*
1423
  Physical chunk operations
1424
*/
1425
 
1426
 
1427
/* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1428
 
1429
#define PREV_INUSE 0x1 
1430
 
1431
/* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1432
 
1433
#define IS_MMAPPED 0x2
1434
 
1435
/* Bits to mask off when extracting size */
1436
 
1437
#define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1438
 
1439
 
1440
/* Ptr to next physical malloc_chunk. */
1441
 
1442
#define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1443
 
1444
/* Ptr to previous physical malloc_chunk */
1445
 
1446
#define prev_chunk(p)\
1447
   ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1448
 
1449
 
1450
/* Treat space at ptr + offset as a chunk */
1451
 
1452
#define chunk_at_offset(p, s)  ((mchunkptr)(((char*)(p)) + (s)))
1453
 
1454
 
1455
 
1456
 
1457
/*
1458
  Dealing with use bits
1459
*/
1460
 
1461
/* extract p's inuse bit */
1462
 
1463
#define inuse(p)\
1464
((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1465
 
1466
/* extract inuse bit of previous chunk */
1467
 
1468
#define prev_inuse(p)  ((p)->size & PREV_INUSE)
1469
 
1470
/* check for mmap()'ed chunk */
1471
 
1472
#define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1473
 
1474
/* set/clear chunk as in use without otherwise disturbing */
1475
 
1476
#define set_inuse(p)\
1477
((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1478
 
1479
#define clear_inuse(p)\
1480
((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1481
 
1482
/* check/set/clear inuse bits in known places */
1483
 
1484
#define inuse_bit_at_offset(p, s)\
1485
 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1486
 
1487
#define set_inuse_bit_at_offset(p, s)\
1488
 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1489
 
1490
#define clear_inuse_bit_at_offset(p, s)\
1491
 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1492
 
1493
 
1494
 
1495
 
1496
/*
1497
  Dealing with size fields
1498
*/
1499
 
1500
/* Get size, ignoring use bits */
1501
 
1502
#define chunksize(p)          ((p)->size & ~(SIZE_BITS))
1503
 
1504
/* Set size at head, without disturbing its use bit */
1505
 
1506
#define set_head_size(p, s)   ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1507
 
1508
/* Set size/use ignoring previous bits in header */
1509
 
1510
#define set_head(p, s)        ((p)->size = (s))
1511
 
1512
/* Set size at footer (only when chunk is not in use) */
1513
 
1514
#define set_foot(p, s)   (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1515
 
1516
 
1517
 
1518
 
1519
 
1520
/*
1521
   Bins
1522
 
1523
    The bins, `av_' are an array of pairs of pointers serving as the
1524
    heads of (initially empty) doubly-linked lists of chunks, laid out
1525
    in a way so that each pair can be treated as if it were in a
1526
    malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1527
    and chunks are the same).
1528
 
1529
    Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1530
    8 bytes apart. Larger bins are approximately logarithmically
1531
    spaced. (See the table below.) The `av_' array is never mentioned
1532
    directly in the code, but instead via bin access macros.
1533
 
1534
    Bin layout:
1535
 
1536
    64 bins of size       8
1537
    32 bins of size      64
1538
    16 bins of size     512
1539
     8 bins of size    4096
1540
     4 bins of size   32768
1541
     2 bins of size  262144
1542
     1 bin  of size what's left
1543
 
1544
    There is actually a little bit of slop in the numbers in bin_index
1545
    for the sake of speed. This makes no difference elsewhere.
1546
 
1547
    The special chunks `top' and `last_remainder' get their own bins,
1548
    (this is implemented via yet more trickery with the av_ array),
1549
    although `top' is never properly linked to its bin since it is
1550
    always handled specially.
1551
 
1552
*/
1553
 
1554
#ifdef SEPARATE_OBJECTS
1555
#define av_ malloc_av_
1556
#endif
1557
 
1558
#define NAV             128   /* number of bins */
1559
 
1560
typedef struct malloc_chunk* mbinptr;
1561
 
1562
/* access macros */
1563
 
1564
#define bin_at(i)      ((mbinptr)((char*)&(av_[2*(i) + 2]) - 2*SIZE_SZ))
1565
#define next_bin(b)    ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1566
#define prev_bin(b)    ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1567
 
1568
/*
1569
   The first 2 bins are never indexed. The corresponding av_ cells are instead
1570
   used for bookkeeping. This is not to save space, but to simplify
1571
   indexing, maintain locality, and avoid some initialization tests.
1572
*/
1573
 
1574
#define top            (bin_at(0)->fd)   /* The topmost chunk */
1575
#define last_remainder (bin_at(1))       /* remainder from last split */
1576
 
1577
 
1578
/*
1579
   Because top initially points to its own bin with initial
1580
   zero size, thus forcing extension on the first malloc request,
1581
   we avoid having any special code in malloc to check whether
1582
   it even exists yet. But we still need to in malloc_extend_top.
1583
*/
1584
 
1585
#define initial_top    ((mchunkptr)(bin_at(0)))
1586
 
1587
/* Helper macro to initialize bins */
1588
 
1589
#define IAV(i)  bin_at(i), bin_at(i)
1590
 
1591
#ifdef DEFINE_MALLOC
1592
STATIC mbinptr av_[NAV * 2 + 2] = {
1593
 0, 0,
1594
 IAV(0),   IAV(1),   IAV(2),   IAV(3),   IAV(4),   IAV(5),   IAV(6),   IAV(7),
1595
 IAV(8),   IAV(9),   IAV(10),  IAV(11),  IAV(12),  IAV(13),  IAV(14),  IAV(15),
1596
 IAV(16),  IAV(17),  IAV(18),  IAV(19),  IAV(20),  IAV(21),  IAV(22),  IAV(23),
1597
 IAV(24),  IAV(25),  IAV(26),  IAV(27),  IAV(28),  IAV(29),  IAV(30),  IAV(31),
1598
 IAV(32),  IAV(33),  IAV(34),  IAV(35),  IAV(36),  IAV(37),  IAV(38),  IAV(39),
1599
 IAV(40),  IAV(41),  IAV(42),  IAV(43),  IAV(44),  IAV(45),  IAV(46),  IAV(47),
1600
 IAV(48),  IAV(49),  IAV(50),  IAV(51),  IAV(52),  IAV(53),  IAV(54),  IAV(55),
1601
 IAV(56),  IAV(57),  IAV(58),  IAV(59),  IAV(60),  IAV(61),  IAV(62),  IAV(63),
1602
 IAV(64),  IAV(65),  IAV(66),  IAV(67),  IAV(68),  IAV(69),  IAV(70),  IAV(71),
1603
 IAV(72),  IAV(73),  IAV(74),  IAV(75),  IAV(76),  IAV(77),  IAV(78),  IAV(79),
1604
 IAV(80),  IAV(81),  IAV(82),  IAV(83),  IAV(84),  IAV(85),  IAV(86),  IAV(87),
1605
 IAV(88),  IAV(89),  IAV(90),  IAV(91),  IAV(92),  IAV(93),  IAV(94),  IAV(95),
1606
 IAV(96),  IAV(97),  IAV(98),  IAV(99),  IAV(100), IAV(101), IAV(102), IAV(103),
1607
 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1608
 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1609
 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1610
};
1611
#else
1612
extern mbinptr av_[NAV * 2 + 2];
1613
#endif
1614
 
1615
 
1616
 
1617
/* field-extraction macros */
1618
 
1619
#define first(b) ((b)->fd)
1620
#define last(b)  ((b)->bk)
1621
 
1622
/*
1623
  Indexing into bins
1624
*/
1625
 
1626
#define bin_index(sz)                                                          \
1627
(((((unsigned long)(sz)) >> 9) ==    0) ?       (((unsigned long)(sz)) >>  3): \
1628
 ((((unsigned long)(sz)) >> 9) <=    4) ?  56 + (((unsigned long)(sz)) >>  6): \
1629
 ((((unsigned long)(sz)) >> 9) <=   20) ?  91 + (((unsigned long)(sz)) >>  9): \
1630
 ((((unsigned long)(sz)) >> 9) <=   84) ? 110 + (((unsigned long)(sz)) >> 12): \
1631
 ((((unsigned long)(sz)) >> 9) <=  340) ? 119 + (((unsigned long)(sz)) >> 15): \
1632
 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1633
                                          126)
1634
/*
1635
  bins for chunks < 512 are all spaced SMALLBIN_WIDTH bytes apart, and hold
1636
  identically sized chunks. This is exploited in malloc.
1637
*/
1638
 
1639
#define MAX_SMALLBIN_SIZE   512
1640
#define SMALLBIN_WIDTH        8
1641
#define SMALLBIN_WIDTH_BITS   3
1642
#define MAX_SMALLBIN        (MAX_SMALLBIN_SIZE / SMALLBIN_WIDTH) - 1
1643
 
1644
#define smallbin_index(sz)  (((unsigned long)(sz)) >> SMALLBIN_WIDTH_BITS)
1645
 
1646
/*
1647
   Requests are `small' if both the corresponding and the next bin are small
1648
*/
1649
 
1650
#define is_small_request(nb) (nb < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1651
 
1652
 
1653
 
1654
/*
1655
    To help compensate for the large number of bins, a one-level index
1656
    structure is used for bin-by-bin searching.  `binblocks' is a
1657
    one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1658
    have any (possibly) non-empty bins, so they can be skipped over
1659
    all at once during during traversals. The bits are NOT always
1660
    cleared as soon as all bins in a block are empty, but instead only
1661
    when all are noticed to be empty during traversal in malloc.
1662
*/
1663
 
1664
#define BINBLOCKWIDTH     4   /* bins per block */
1665
 
1666
#define binblocks      (bin_at(0)->size) /* bitvector of nonempty blocks */
1667
 
1668
/* bin<->block macros */
1669
 
1670
#define idx2binblock(ix)    ((unsigned long)1 << (ix / BINBLOCKWIDTH))
1671
#define mark_binblock(ii)   (binblocks |= idx2binblock(ii))
1672
#define clear_binblock(ii)  (binblocks &= ~(idx2binblock(ii)))
1673
 
1674
 
1675
 
1676
 
1677
 
1678
/*  Other static bookkeeping data */
1679
 
1680
#ifdef SEPARATE_OBJECTS
1681
#define trim_threshold          malloc_trim_threshold
1682
#define top_pad                 malloc_top_pad
1683
#define n_mmaps_max             malloc_n_mmaps_max
1684
#define mmap_threshold          malloc_mmap_threshold
1685
#define sbrk_base               malloc_sbrk_base
1686
#define max_sbrked_mem          malloc_max_sbrked_mem
1687
#define max_total_mem           malloc_max_total_mem
1688
#define current_mallinfo        malloc_current_mallinfo
1689
#define n_mmaps                 malloc_n_mmaps
1690
#define max_n_mmaps             malloc_max_n_mmaps
1691
#define mmapped_mem             malloc_mmapped_mem
1692
#define max_mmapped_mem         malloc_max_mmapped_mem
1693
#endif
1694
 
1695
/* variables holding tunable values */
1696
 
1697
#ifdef DEFINE_MALLOC
1698
 
1699
STATIC unsigned long trim_threshold   = DEFAULT_TRIM_THRESHOLD;
1700
STATIC unsigned long top_pad          = DEFAULT_TOP_PAD;
1701
#if HAVE_MMAP
1702
STATIC unsigned int  n_mmaps_max      = DEFAULT_MMAP_MAX;
1703
STATIC unsigned long mmap_threshold   = DEFAULT_MMAP_THRESHOLD;
1704
#endif
1705
 
1706
/* The first value returned from sbrk */
1707
STATIC char* sbrk_base = (char*)(-1);
1708
 
1709
/* The maximum memory obtained from system via sbrk */
1710
STATIC unsigned long max_sbrked_mem = 0;
1711
 
1712
/* The maximum via either sbrk or mmap */
1713
STATIC unsigned long max_total_mem = 0;
1714
 
1715
/* internal working copy of mallinfo */
1716
STATIC struct mallinfo current_mallinfo = {  0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
1717
 
1718
#if HAVE_MMAP
1719
 
1720
/* Tracking mmaps */
1721
 
1722
STATIC unsigned int n_mmaps = 0;
1723
STATIC unsigned int max_n_mmaps = 0;
1724
STATIC unsigned long mmapped_mem = 0;
1725
STATIC unsigned long max_mmapped_mem = 0;
1726
 
1727
#endif
1728
 
1729
#else /* ! DEFINE_MALLOC */
1730
 
1731
extern unsigned long trim_threshold;
1732
extern unsigned long top_pad;
1733
#if HAVE_MMAP
1734
extern unsigned int  n_mmaps_max;
1735
extern unsigned long mmap_threshold;
1736
#endif
1737
extern char* sbrk_base;
1738
extern unsigned long max_sbrked_mem;
1739
extern unsigned long max_total_mem;
1740
extern struct mallinfo current_mallinfo;
1741
#if HAVE_MMAP
1742
extern unsigned int n_mmaps;
1743
extern unsigned int max_n_mmaps;
1744
extern unsigned long mmapped_mem;
1745
extern unsigned long max_mmapped_mem;
1746
#endif
1747
 
1748
#endif /* ! DEFINE_MALLOC */
1749
 
1750
/* The total memory obtained from system via sbrk */
1751
#define sbrked_mem  (current_mallinfo.arena)
1752
 
1753
 
1754
 
1755
/*
1756
  Debugging support
1757
*/
1758
 
1759
#if DEBUG
1760
 
1761
 
1762
/*
1763
  These routines make a number of assertions about the states
1764
  of data structures that should be true at all times. If any
1765
  are not true, it's very likely that a user program has somehow
1766
  trashed memory. (It's also possible that there is a coding error
1767
  in malloc. In which case, please report it!)
1768
*/
1769
 
1770
#if __STD_C
1771
static void do_check_chunk(mchunkptr p)
1772
#else
1773
static void do_check_chunk(p) mchunkptr p;
1774
#endif
1775
{
1776
  INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1777
 
1778
  /* No checkable chunk is mmapped */
1779
  assert(!chunk_is_mmapped(p));
1780
 
1781
  /* Check for legal address ... */
1782
  assert((char*)p >= sbrk_base);
1783
  if (p != top)
1784
    assert((char*)p + sz <= (char*)top);
1785
  else
1786
    assert((char*)p + sz <= sbrk_base + sbrked_mem);
1787
 
1788
}
1789
 
1790
 
1791
#if __STD_C
1792
static void do_check_free_chunk(mchunkptr p)
1793
#else
1794
static void do_check_free_chunk(p) mchunkptr p;
1795
#endif
1796
{
1797
  INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1798
  mchunkptr next = chunk_at_offset(p, sz);
1799
 
1800
  do_check_chunk(p);
1801
 
1802
  /* Check whether it claims to be free ... */
1803
  assert(!inuse(p));
1804
 
1805
  /* Unless a special marker, must have OK fields */
1806
  if ((long)sz >= (long)MINSIZE)
1807
  {
1808
    assert((sz & MALLOC_ALIGN_MASK) == 0);
1809
    assert(aligned_OK(chunk2mem(p)));
1810
    /* ... matching footer field */
1811
    assert(next->prev_size == sz);
1812
    /* ... and is fully consolidated */
1813
    assert(prev_inuse(p));
1814
    assert (next == top || inuse(next));
1815
 
1816
    /* ... and has minimally sane links */
1817
    assert(p->fd->bk == p);
1818
    assert(p->bk->fd == p);
1819
  }
1820
  else /* markers are always of size SIZE_SZ */
1821
    assert(sz == SIZE_SZ);
1822
}
1823
 
1824
#if __STD_C
1825
static void do_check_inuse_chunk(mchunkptr p)
1826
#else
1827
static void do_check_inuse_chunk(p) mchunkptr p;
1828
#endif
1829
{
1830
  mchunkptr next = next_chunk(p);
1831
  do_check_chunk(p);
1832
 
1833
  /* Check whether it claims to be in use ... */
1834
  assert(inuse(p));
1835
 
1836
  /* ... and is surrounded by OK chunks.
1837
    Since more things can be checked with free chunks than inuse ones,
1838
    if an inuse chunk borders them and debug is on, it's worth doing them.
1839
  */
1840
  if (!prev_inuse(p))
1841
  {
1842
    mchunkptr prv = prev_chunk(p);
1843
    assert(next_chunk(prv) == p);
1844
    do_check_free_chunk(prv);
1845
  }
1846
  if (next == top)
1847
  {
1848
    assert(prev_inuse(next));
1849
    assert(chunksize(next) >= MINSIZE);
1850
  }
1851
  else if (!inuse(next))
1852
    do_check_free_chunk(next);
1853
 
1854
}
1855
 
1856
#if __STD_C
1857
static void do_check_malloced_chunk(mchunkptr p, INTERNAL_SIZE_T s)
1858
#else
1859
static void do_check_malloced_chunk(p, s) mchunkptr p; INTERNAL_SIZE_T s;
1860
#endif
1861
{
1862
  INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1863
  long room = long_sub_size_t(sz, s);
1864
 
1865
  do_check_inuse_chunk(p);
1866
 
1867
  /* Legal size ... */
1868
  assert((long)sz >= (long)MINSIZE);
1869
  assert((sz & MALLOC_ALIGN_MASK) == 0);
1870
  assert(room >= 0);
1871
  assert(room < (long)MINSIZE);
1872
 
1873
  /* ... and alignment */
1874
  assert(aligned_OK(chunk2mem(p)));
1875
 
1876
 
1877
  /* ... and was allocated at front of an available chunk */
1878
  assert(prev_inuse(p));
1879
 
1880
}
1881
 
1882
 
1883
#define check_free_chunk(P)  do_check_free_chunk(P)
1884
#define check_inuse_chunk(P) do_check_inuse_chunk(P)
1885
#define check_chunk(P) do_check_chunk(P)
1886
#define check_malloced_chunk(P,N) do_check_malloced_chunk(P,N)
1887
#else
1888
#define check_free_chunk(P) 
1889
#define check_inuse_chunk(P)
1890
#define check_chunk(P)
1891
#define check_malloced_chunk(P,N)
1892
#endif
1893
 
1894
 
1895
 
1896
/*
1897
  Macro-based internal utilities
1898
*/
1899
 
1900
 
1901
/*
1902
  Linking chunks in bin lists.
1903
  Call these only with variables, not arbitrary expressions, as arguments.
1904
*/
1905
 
1906
/*
1907
  Place chunk p of size s in its bin, in size order,
1908
  putting it ahead of others of same size.
1909
*/
1910
 
1911
 
1912
#define frontlink(P, S, IDX, BK, FD)                                          \
1913
{                                                                             \
1914
  if (S < MAX_SMALLBIN_SIZE)                                                  \
1915
  {                                                                           \
1916
    IDX = smallbin_index(S);                                                  \
1917
    mark_binblock(IDX);                                                       \
1918
    BK = bin_at(IDX);                                                         \
1919
    FD = BK->fd;                                                              \
1920
    P->bk = BK;                                                               \
1921
    P->fd = FD;                                                               \
1922
    FD->bk = BK->fd = P;                                                      \
1923
  }                                                                           \
1924
  else                                                                        \
1925
  {                                                                           \
1926
    IDX = bin_index(S);                                                       \
1927
    BK = bin_at(IDX);                                                         \
1928
    FD = BK->fd;                                                              \
1929
    if (FD == BK) mark_binblock(IDX);                                         \
1930
    else                                                                      \
1931
    {                                                                         \
1932
      while (FD != BK && S < chunksize(FD)) FD = FD->fd;                      \
1933
      BK = FD->bk;                                                            \
1934
    }                                                                         \
1935
    P->bk = BK;                                                               \
1936
    P->fd = FD;                                                               \
1937
    FD->bk = BK->fd = P;                                                      \
1938
  }                                                                           \
1939
}
1940
 
1941
 
1942
/* take a chunk off a list */
1943
 
1944
#define unlink(P, BK, FD)                                                     \
1945
{                                                                             \
1946
  BK = P->bk;                                                                 \
1947
  FD = P->fd;                                                                 \
1948
  FD->bk = BK;                                                        \
1949
  BK->fd = FD;                                                        \
1950
}                                                                             \
1951
 
1952
/* Place p as the last remainder */
1953
 
1954
#define link_last_remainder(P)                                                \
1955
{                                                                             \
1956
  last_remainder->fd = last_remainder->bk =  P;                               \
1957
  P->fd = P->bk = last_remainder;                                             \
1958
}
1959
 
1960
/* Clear the last_remainder bin */
1961
 
1962
#define clear_last_remainder \
1963
  (last_remainder->fd = last_remainder->bk = last_remainder)
1964
 
1965
 
1966
 
1967
 
1968
 
1969
 
1970
/* Routines dealing with mmap(). */
1971
 
1972
#if HAVE_MMAP
1973
 
1974
#ifdef DEFINE_MALLOC
1975
 
1976
#if __STD_C
1977
static mchunkptr mmap_chunk(size_t size)
1978
#else
1979
static mchunkptr mmap_chunk(size) size_t size;
1980
#endif
1981
{
1982
  size_t page_mask = malloc_getpagesize - 1;
1983
  mchunkptr p;
1984
 
1985
#ifndef MAP_ANONYMOUS
1986
  static int fd = -1;
1987
#endif
1988
 
1989
  if(n_mmaps >= n_mmaps_max) return 0; /* too many regions */
1990
 
1991
  /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1992
   * there is no following chunk whose prev_size field could be used.
1993
   */
1994
  size = (size + SIZE_SZ + page_mask) & ~page_mask;
1995
 
1996
#ifdef MAP_ANONYMOUS
1997
  p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE,
1998
                      MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
1999
#else /* !MAP_ANONYMOUS */
2000
  if (fd < 0)
2001
  {
2002
    fd = open("/dev/zero", O_RDWR);
2003
    if(fd < 0) return 0;
2004
  }
2005
  p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
2006
#endif
2007
 
2008
  if(p == (mchunkptr)-1) return 0;
2009
 
2010
  n_mmaps++;
2011
  if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
2012
 
2013
  /* We demand that eight bytes into a page must be 8-byte aligned. */
2014
  assert(aligned_OK(chunk2mem(p)));
2015
 
2016
  /* The offset to the start of the mmapped region is stored
2017
   * in the prev_size field of the chunk; normally it is zero,
2018
   * but that can be changed in memalign().
2019
   */
2020
  p->prev_size = 0;
2021
  set_head(p, size|IS_MMAPPED);
2022
 
2023
  mmapped_mem += size;
2024
  if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
2025
    max_mmapped_mem = mmapped_mem;
2026
  if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2027
    max_total_mem = mmapped_mem + sbrked_mem;
2028
  return p;
2029
}
2030
 
2031
#endif /* DEFINE_MALLOC */
2032
 
2033
#ifdef SEPARATE_OBJECTS
2034
#define munmap_chunk malloc_munmap_chunk
2035
#endif
2036
 
2037
#ifdef DEFINE_FREE
2038
 
2039
#if __STD_C
2040
STATIC void munmap_chunk(mchunkptr p)
2041
#else
2042
STATIC void munmap_chunk(p) mchunkptr p;
2043
#endif
2044
{
2045
  INTERNAL_SIZE_T size = chunksize(p);
2046
  int ret;
2047
 
2048
  assert (chunk_is_mmapped(p));
2049
  assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
2050
  assert((n_mmaps > 0));
2051
  assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
2052
 
2053
  n_mmaps--;
2054
  mmapped_mem -= (size + p->prev_size);
2055
 
2056
  ret = munmap((char *)p - p->prev_size, size + p->prev_size);
2057
 
2058
  /* munmap returns non-zero on failure */
2059
  assert(ret == 0);
2060
}
2061
 
2062
#else /* ! DEFINE_FREE */
2063
 
2064
#if __STD_C
2065
extern void munmap_chunk(mchunkptr);
2066
#else
2067
extern void munmap_chunk();
2068
#endif
2069
 
2070
#endif /* ! DEFINE_FREE */
2071
 
2072
#if HAVE_MREMAP
2073
 
2074
#ifdef DEFINE_REALLOC
2075
 
2076
#if __STD_C
2077
static mchunkptr mremap_chunk(mchunkptr p, size_t new_size)
2078
#else
2079
static mchunkptr mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
2080
#endif
2081
{
2082
  size_t page_mask = malloc_getpagesize - 1;
2083
  INTERNAL_SIZE_T offset = p->prev_size;
2084
  INTERNAL_SIZE_T size = chunksize(p);
2085
  char *cp;
2086
 
2087
  assert (chunk_is_mmapped(p));
2088
  assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
2089
  assert((n_mmaps > 0));
2090
  assert(((size + offset) & (malloc_getpagesize-1)) == 0);
2091
 
2092
  /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
2093
  new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
2094
 
2095
  cp = (char *)mremap((char *)p - offset, size + offset, new_size, 1);
2096
 
2097
  if (cp == (char *)-1) return 0;
2098
 
2099
  p = (mchunkptr)(cp + offset);
2100
 
2101
  assert(aligned_OK(chunk2mem(p)));
2102
 
2103
  assert((p->prev_size == offset));
2104
  set_head(p, (new_size - offset)|IS_MMAPPED);
2105
 
2106
  mmapped_mem -= size + offset;
2107
  mmapped_mem += new_size;
2108
  if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
2109
    max_mmapped_mem = mmapped_mem;
2110
  if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2111
    max_total_mem = mmapped_mem + sbrked_mem;
2112
  return p;
2113
}
2114
 
2115
#endif /* DEFINE_REALLOC */
2116
 
2117
#endif /* HAVE_MREMAP */
2118
 
2119
#endif /* HAVE_MMAP */
2120
 
2121
 
2122
 
2123
 
2124
#ifdef DEFINE_MALLOC
2125
 
2126
/*
2127
  Extend the top-most chunk by obtaining memory from system.
2128
  Main interface to sbrk (but see also malloc_trim).
2129
*/
2130
 
2131
#if __STD_C
2132
static void malloc_extend_top(RARG INTERNAL_SIZE_T nb)
2133
#else
2134
static void malloc_extend_top(RARG nb) RDECL INTERNAL_SIZE_T nb;
2135
#endif
2136
{
2137
  char*     brk;                  /* return value from sbrk */
2138
  INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
2139
  INTERNAL_SIZE_T correction;     /* bytes for 2nd sbrk call */
2140
  int correction_failed = 0;      /* whether we should relax the assertion */
2141
  char*     new_brk;              /* return of 2nd sbrk call */
2142
  INTERNAL_SIZE_T top_size;       /* new size of top chunk */
2143
 
2144
  mchunkptr old_top     = top;  /* Record state of old top */
2145
  INTERNAL_SIZE_T old_top_size = chunksize(old_top);
2146
  char*     old_end      = (char*)(chunk_at_offset(old_top, old_top_size));
2147
 
2148
  /* Pad request with top_pad plus minimal overhead */
2149
 
2150
  INTERNAL_SIZE_T    sbrk_size     = nb + top_pad + MINSIZE;
2151
  unsigned long pagesz    = malloc_getpagesize;
2152
 
2153
  /* If not the first time through, round to preserve page boundary */
2154
  /* Otherwise, we need to correct to a page size below anyway. */
2155
  /* (We also correct below if an intervening foreign sbrk call.) */
2156
 
2157
  if (sbrk_base != (char*)(-1))
2158
    sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
2159
 
2160
  brk = (char*)(MORECORE (sbrk_size));
2161
 
2162
  /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2163
  if (brk == (char*)(MORECORE_FAILURE) ||
2164
      (brk < old_end && old_top != initial_top))
2165
    return;
2166
 
2167
  sbrked_mem += sbrk_size;
2168
 
2169
  if (brk == old_end /* can just add bytes to current top, unless
2170
                        previous correction failed */
2171
      && ((POINTER_UINT)old_end & (pagesz - 1)) == 0)
2172
  {
2173
    top_size = sbrk_size + old_top_size;
2174
    set_head(top, top_size | PREV_INUSE);
2175
  }
2176
  else
2177
  {
2178
    if (sbrk_base == (char*)(-1))  /* First time through. Record base */
2179
      sbrk_base = brk;
2180
    else  /* Someone else called sbrk().  Count those bytes as sbrked_mem. */
2181
      sbrked_mem += brk - (char*)old_end;
2182
 
2183
    /* Guarantee alignment of first new chunk made from this space */
2184
    front_misalign = (POINTER_UINT)chunk2mem(brk) & MALLOC_ALIGN_MASK;
2185
    if (front_misalign > 0)
2186
    {
2187
      correction = (MALLOC_ALIGNMENT) - front_misalign;
2188
      brk += correction;
2189
    }
2190
    else
2191
      correction = 0;
2192
 
2193
    /* Guarantee the next brk will be at a page boundary */
2194
    correction += pagesz - ((POINTER_UINT)(brk + sbrk_size) & (pagesz - 1));
2195
 
2196
    /* Allocate correction */
2197
    new_brk = (char*)(MORECORE (correction));
2198
    if (new_brk == (char*)(MORECORE_FAILURE))
2199
      {
2200
        correction = 0;
2201
        correction_failed = 1;
2202
        new_brk = brk;
2203
      }
2204
 
2205
    sbrked_mem += correction;
2206
 
2207
    top = (mchunkptr)brk;
2208
    top_size = new_brk - brk + correction;
2209
    set_head(top, top_size | PREV_INUSE);
2210
 
2211
    if (old_top != initial_top)
2212
    {
2213
 
2214
      /* There must have been an intervening foreign sbrk call. */
2215
      /* A double fencepost is necessary to prevent consolidation */
2216
 
2217
      /* If not enough space to do this, then user did something very wrong */
2218
      if (old_top_size < MINSIZE)
2219
      {
2220
        set_head(top, PREV_INUSE); /* will force null return from malloc */
2221
        return;
2222
      }
2223
 
2224
      /* Also keep size a multiple of MALLOC_ALIGNMENT */
2225
      old_top_size = (old_top_size - 3*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2226
      chunk_at_offset(old_top, old_top_size          )->size =
2227
        SIZE_SZ|PREV_INUSE;
2228
      chunk_at_offset(old_top, old_top_size + SIZE_SZ)->size =
2229
        SIZE_SZ|PREV_INUSE;
2230
      set_head_size(old_top, old_top_size);
2231
      /* If possible, release the rest. */
2232
      if (old_top_size >= MINSIZE)
2233
        fREe(RCALL chunk2mem(old_top));
2234
    }
2235
  }
2236
 
2237
  if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
2238
    max_sbrked_mem = sbrked_mem;
2239
#if HAVE_MMAP
2240
  if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2241
    max_total_mem = mmapped_mem + sbrked_mem;
2242
#else
2243
  if ((unsigned long)(sbrked_mem) > (unsigned long)max_total_mem)
2244
    max_total_mem = sbrked_mem;
2245
#endif
2246
 
2247
  /* We always land on a page boundary */
2248
  assert(((unsigned long)((char*)top + top_size) & (pagesz - 1)) == 0
2249
         || correction_failed);
2250
}
2251
 
2252
#endif /* DEFINE_MALLOC */
2253
 
2254
 
2255
/* Main public routines */
2256
 
2257
#ifdef DEFINE_MALLOC
2258
 
2259
/*
2260
  Malloc Algorthim:
2261
 
2262
    The requested size is first converted into a usable form, `nb'.
2263
    This currently means to add 4 bytes overhead plus possibly more to
2264
    obtain 8-byte alignment and/or to obtain a size of at least
2265
    MINSIZE (currently 16 bytes), the smallest allocatable size.
2266
    (All fits are considered `exact' if they are within MINSIZE bytes.)
2267
 
2268
    From there, the first successful of the following steps is taken:
2269
 
2270
      1. The bin corresponding to the request size is scanned, and if
2271
         a chunk of exactly the right size is found, it is taken.
2272
 
2273
      2. The most recently remaindered chunk is used if it is big
2274
         enough.  This is a form of (roving) first fit, used only in
2275
         the absence of exact fits. Runs of consecutive requests use
2276
         the remainder of the chunk used for the previous such request
2277
         whenever possible. This limited use of a first-fit style
2278
         allocation strategy tends to give contiguous chunks
2279
         coextensive lifetimes, which improves locality and can reduce
2280
         fragmentation in the long run.
2281
 
2282
      3. Other bins are scanned in increasing size order, using a
2283
         chunk big enough to fulfill the request, and splitting off
2284
         any remainder.  This search is strictly by best-fit; i.e.,
2285
         the smallest (with ties going to approximately the least
2286
         recently used) chunk that fits is selected.
2287
 
2288
      4. If large enough, the chunk bordering the end of memory
2289
         (`top') is split off. (This use of `top' is in accord with
2290
         the best-fit search rule.  In effect, `top' is treated as
2291
         larger (and thus less well fitting) than any other available
2292
         chunk since it can be extended to be as large as necessary
2293
         (up to system limitations).
2294
 
2295
      5. If the request size meets the mmap threshold and the
2296
         system supports mmap, and there are few enough currently
2297
         allocated mmapped regions, and a call to mmap succeeds,
2298
         the request is allocated via direct memory mapping.
2299
 
2300
      6. Otherwise, the top of memory is extended by
2301
         obtaining more space from the system (normally using sbrk,
2302
         but definable to anything else via the MORECORE macro).
2303
         Memory is gathered from the system (in system page-sized
2304
         units) in a way that allows chunks obtained across different
2305
         sbrk calls to be consolidated, but does not require
2306
         contiguous memory. Thus, it should be safe to intersperse
2307
         mallocs with other sbrk calls.
2308
 
2309
 
2310
      All allocations are made from the the `lowest' part of any found
2311
      chunk. (The implementation invariant is that prev_inuse is
2312
      always true of any allocated chunk; i.e., that each allocated
2313
      chunk borders either a previously allocated and still in-use chunk,
2314
      or the base of its memory arena.)
2315
 
2316
*/
2317
 
2318
#if __STD_C
2319
Void_t* mALLOc(RARG size_t bytes)
2320
#else
2321
Void_t* mALLOc(RARG bytes) RDECL size_t bytes;
2322
#endif
2323
{
2324
#ifdef MALLOC_PROVIDED
2325
 
2326
  return malloc (bytes); // Make sure that the pointer returned by malloc is returned back.
2327
 
2328
#else
2329
 
2330
  mchunkptr victim;                  /* inspected/selected chunk */
2331
  INTERNAL_SIZE_T victim_size;       /* its size */
2332
  int       idx;                     /* index for bin traversal */
2333
  mbinptr   bin;                     /* associated bin */
2334
  mchunkptr remainder;               /* remainder from a split */
2335
  long      remainder_size;          /* its size */
2336
  int       remainder_index;         /* its bin index */
2337
  unsigned long block;               /* block traverser bit */
2338
  int       startidx;                /* first bin of a traversed block */
2339
  mchunkptr fwd;                     /* misc temp for linking */
2340
  mchunkptr bck;                     /* misc temp for linking */
2341
  mbinptr q;                         /* misc temp */
2342
 
2343
  INTERNAL_SIZE_T nb  = request2size(bytes);  /* padded request size; */
2344
 
2345
  /* Check for overflow and just fail, if so. */
2346
  if (nb > INT_MAX || nb < bytes)
2347
  {
2348
    RERRNO = ENOMEM;
2349
    return 0;
2350
  }
2351
 
2352
  MALLOC_LOCK;
2353
 
2354
  /* Check for exact match in a bin */
2355
 
2356
  if (is_small_request(nb))  /* Faster version for small requests */
2357
  {
2358
    idx = smallbin_index(nb);
2359
 
2360
    /* No traversal or size check necessary for small bins.  */
2361
 
2362
    q = bin_at(idx);
2363
    victim = last(q);
2364
 
2365
#if MALLOC_ALIGN != 16
2366
    /* Also scan the next one, since it would have a remainder < MINSIZE */
2367
    if (victim == q)
2368
    {
2369
      q = next_bin(q);
2370
      victim = last(q);
2371
    }
2372
#endif
2373
    if (victim != q)
2374
    {
2375
      victim_size = chunksize(victim);
2376
      unlink(victim, bck, fwd);
2377
      set_inuse_bit_at_offset(victim, victim_size);
2378
      check_malloced_chunk(victim, nb);
2379
      MALLOC_UNLOCK;
2380
      return chunk2mem(victim);
2381
    }
2382
 
2383
    idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
2384
 
2385
  }
2386
  else
2387
  {
2388
    idx = bin_index(nb);
2389
    bin = bin_at(idx);
2390
 
2391
    for (victim = last(bin); victim != bin; victim = victim->bk)
2392
    {
2393
      victim_size = chunksize(victim);
2394
      remainder_size = long_sub_size_t(victim_size, nb);
2395
 
2396
      if (remainder_size >= (long)MINSIZE) /* too big */
2397
      {
2398
        --idx; /* adjust to rescan below after checking last remainder */
2399
        break;
2400
      }
2401
 
2402
      else if (remainder_size >= 0) /* exact fit */
2403
      {
2404
        unlink(victim, bck, fwd);
2405
        set_inuse_bit_at_offset(victim, victim_size);
2406
        check_malloced_chunk(victim, nb);
2407
        MALLOC_UNLOCK;
2408
        return chunk2mem(victim);
2409
      }
2410
    }
2411
 
2412
    ++idx;
2413
 
2414
  }
2415
 
2416
  /* Try to use the last split-off remainder */
2417
 
2418
  if ( (victim = last_remainder->fd) != last_remainder)
2419
  {
2420
    victim_size = chunksize(victim);
2421
    remainder_size = long_sub_size_t(victim_size, nb);
2422
 
2423
    if (remainder_size >= (long)MINSIZE) /* re-split */
2424
    {
2425
      remainder = chunk_at_offset(victim, nb);
2426
      set_head(victim, nb | PREV_INUSE);
2427
      link_last_remainder(remainder);
2428
      set_head(remainder, remainder_size | PREV_INUSE);
2429
      set_foot(remainder, remainder_size);
2430
      check_malloced_chunk(victim, nb);
2431
      MALLOC_UNLOCK;
2432
      return chunk2mem(victim);
2433
    }
2434
 
2435
    clear_last_remainder;
2436
 
2437
    if (remainder_size >= 0)  /* exhaust */
2438
    {
2439
      set_inuse_bit_at_offset(victim, victim_size);
2440
      check_malloced_chunk(victim, nb);
2441
      MALLOC_UNLOCK;
2442
      return chunk2mem(victim);
2443
    }
2444
 
2445
    /* Else place in bin */
2446
 
2447
    frontlink(victim, victim_size, remainder_index, bck, fwd);
2448
  }
2449
 
2450
  /*
2451
     If there are any possibly nonempty big-enough blocks,
2452
     search for best fitting chunk by scanning bins in blockwidth units.
2453
  */
2454
 
2455
  if ( (block = idx2binblock(idx)) <= binblocks)
2456
  {
2457
 
2458
    /* Get to the first marked block */
2459
 
2460
    if ( (block & binblocks) == 0)
2461
    {
2462
      /* force to an even block boundary */
2463
      idx = (idx & ~(BINBLOCKWIDTH - 1)) + BINBLOCKWIDTH;
2464
      block <<= 1;
2465
      while ((block & binblocks) == 0)
2466
      {
2467
        idx += BINBLOCKWIDTH;
2468
        block <<= 1;
2469
      }
2470
    }
2471
 
2472
    /* For each possibly nonempty block ... */
2473
    for (;;)
2474
    {
2475
      startidx = idx;          /* (track incomplete blocks) */
2476
      q = bin = bin_at(idx);
2477
 
2478
      /* For each bin in this block ... */
2479
      do
2480
      {
2481
        /* Find and use first big enough chunk ... */
2482
 
2483
        for (victim = last(bin); victim != bin; victim = victim->bk)
2484
        {
2485
          victim_size = chunksize(victim);
2486
          remainder_size = long_sub_size_t(victim_size, nb);
2487
 
2488
          if (remainder_size >= (long)MINSIZE) /* split */
2489
          {
2490
            remainder = chunk_at_offset(victim, nb);
2491
            set_head(victim, nb | PREV_INUSE);
2492
            unlink(victim, bck, fwd);
2493
            link_last_remainder(remainder);
2494
            set_head(remainder, remainder_size | PREV_INUSE);
2495
            set_foot(remainder, remainder_size);
2496
            check_malloced_chunk(victim, nb);
2497
            MALLOC_UNLOCK;
2498
            return chunk2mem(victim);
2499
          }
2500
 
2501
          else if (remainder_size >= 0)  /* take */
2502
          {
2503
            set_inuse_bit_at_offset(victim, victim_size);
2504
            unlink(victim, bck, fwd);
2505
            check_malloced_chunk(victim, nb);
2506
            MALLOC_UNLOCK;
2507
            return chunk2mem(victim);
2508
          }
2509
 
2510
        }
2511
 
2512
       bin = next_bin(bin);
2513
 
2514
#if MALLOC_ALIGN == 16
2515
       if (idx < MAX_SMALLBIN)
2516
         {
2517
           bin = next_bin(bin);
2518
           ++idx;
2519
         }
2520
#endif
2521
      } while ((++idx & (BINBLOCKWIDTH - 1)) != 0);
2522
 
2523
      /* Clear out the block bit. */
2524
 
2525
      do   /* Possibly backtrack to try to clear a partial block */
2526
      {
2527
        if ((startidx & (BINBLOCKWIDTH - 1)) == 0)
2528
        {
2529
          binblocks &= ~block;
2530
          break;
2531
        }
2532
        --startidx;
2533
       q = prev_bin(q);
2534
      } while (first(q) == q);
2535
 
2536
      /* Get to the next possibly nonempty block */
2537
 
2538
      if ( (block <<= 1) <= binblocks && (block != 0) )
2539
      {
2540
        while ((block & binblocks) == 0)
2541
        {
2542
          idx += BINBLOCKWIDTH;
2543
          block <<= 1;
2544
        }
2545
      }
2546
      else
2547
        break;
2548
    }
2549
  }
2550
 
2551
 
2552
  /* Try to use top chunk */
2553
 
2554
  /* Require that there be a remainder, ensuring top always exists  */
2555
  remainder_size = long_sub_size_t(chunksize(top), nb);
2556
  if (chunksize(top) < nb || remainder_size < (long)MINSIZE)
2557
  {
2558
 
2559
#if HAVE_MMAP
2560
    /* If big and would otherwise need to extend, try to use mmap instead */
2561
    if ((unsigned long)nb >= (unsigned long)mmap_threshold &&
2562
        (victim = mmap_chunk(nb)) != 0)
2563
    {
2564
      MALLOC_UNLOCK;
2565
      return chunk2mem(victim);
2566
    }
2567
#endif
2568
 
2569
    /* Try to extend */
2570
    malloc_extend_top(RCALL nb);
2571
    remainder_size = long_sub_size_t(chunksize(top), nb);
2572
    if (chunksize(top) < nb || remainder_size < (long)MINSIZE)
2573
    {
2574
      MALLOC_UNLOCK;
2575
      return 0; /* propagate failure */
2576
    }
2577
  }
2578
 
2579
  victim = top;
2580
  set_head(victim, nb | PREV_INUSE);
2581
  top = chunk_at_offset(victim, nb);
2582
  set_head(top, remainder_size | PREV_INUSE);
2583
  check_malloced_chunk(victim, nb);
2584
  MALLOC_UNLOCK;
2585
  return chunk2mem(victim);
2586
 
2587
#endif /* MALLOC_PROVIDED */
2588
}
2589
 
2590
#endif /* DEFINE_MALLOC */
2591
 
2592
#ifdef DEFINE_FREE
2593
 
2594
/*
2595
 
2596
  free() algorithm :
2597
 
2598
    cases:
2599
 
2600
       1. free(0) has no effect.
2601
 
2602
       2. If the chunk was allocated via mmap, it is release via munmap().
2603
 
2604
       3. If a returned chunk borders the current high end of memory,
2605
          it is consolidated into the top, and if the total unused
2606
          topmost memory exceeds the trim threshold, malloc_trim is
2607
          called.
2608
 
2609
       4. Other chunks are consolidated as they arrive, and
2610
          placed in corresponding bins. (This includes the case of
2611
          consolidating with the current `last_remainder').
2612
 
2613
*/
2614
 
2615
 
2616
#if __STD_C
2617
void fREe(RARG Void_t* mem)
2618
#else
2619
void fREe(RARG mem) RDECL Void_t* mem;
2620
#endif
2621
{
2622
#ifdef MALLOC_PROVIDED
2623
 
2624
  free (mem);
2625
 
2626
#else
2627
 
2628
  mchunkptr p;         /* chunk corresponding to mem */
2629
  INTERNAL_SIZE_T hd;  /* its head field */
2630
  INTERNAL_SIZE_T sz;  /* its size */
2631
  int       idx;       /* its bin index */
2632
  mchunkptr next;      /* next contiguous chunk */
2633
  INTERNAL_SIZE_T nextsz; /* its size */
2634
  INTERNAL_SIZE_T prevsz; /* size of previous contiguous chunk */
2635
  mchunkptr bck;       /* misc temp for linking */
2636
  mchunkptr fwd;       /* misc temp for linking */
2637
  int       islr;      /* track whether merging with last_remainder */
2638
 
2639
  if (mem == 0)                              /* free(0) has no effect */
2640
    return;
2641
 
2642
  MALLOC_LOCK;
2643
 
2644
  p = mem2chunk(mem);
2645
  hd = p->size;
2646
 
2647
#if HAVE_MMAP
2648
  if (hd & IS_MMAPPED)                       /* release mmapped memory. */
2649
  {
2650
    munmap_chunk(p);
2651
    MALLOC_UNLOCK;
2652
    return;
2653
  }
2654
#endif
2655
 
2656
  check_inuse_chunk(p);
2657
 
2658
  sz = hd & ~PREV_INUSE;
2659
  next = chunk_at_offset(p, sz);
2660
  nextsz = chunksize(next);
2661
 
2662
  if (next == top)                            /* merge with top */
2663
  {
2664
    sz += nextsz;
2665
 
2666
    if (!(hd & PREV_INUSE))                    /* consolidate backward */
2667
    {
2668
      prevsz = p->prev_size;
2669
      p = chunk_at_offset(p, -prevsz);
2670
      sz += prevsz;
2671
      unlink(p, bck, fwd);
2672
    }
2673
 
2674
    set_head(p, sz | PREV_INUSE);
2675
    top = p;
2676
    if ((unsigned long)(sz) >= (unsigned long)trim_threshold)
2677
      malloc_trim(RCALL top_pad);
2678
    MALLOC_UNLOCK;
2679
    return;
2680
  }
2681
 
2682
  set_head(next, nextsz);                    /* clear inuse bit */
2683
 
2684
  islr = 0;
2685
 
2686
  if (!(hd & PREV_INUSE))                    /* consolidate backward */
2687
  {
2688
    prevsz = p->prev_size;
2689
    p = chunk_at_offset(p, -prevsz);
2690
    sz += prevsz;
2691
 
2692
    if (p->fd == last_remainder)             /* keep as last_remainder */
2693
      islr = 1;
2694
    else
2695
      unlink(p, bck, fwd);
2696
  }
2697
 
2698
  if (!(inuse_bit_at_offset(next, nextsz)))   /* consolidate forward */
2699
  {
2700
    sz += nextsz;
2701
 
2702
    if (!islr && next->fd == last_remainder)  /* re-insert last_remainder */
2703
    {
2704
      islr = 1;
2705
      link_last_remainder(p);
2706
    }
2707
    else
2708
      unlink(next, bck, fwd);
2709
  }
2710
 
2711
 
2712
  set_head(p, sz | PREV_INUSE);
2713
  set_foot(p, sz);
2714
  if (!islr)
2715
    frontlink(p, sz, idx, bck, fwd);
2716
 
2717
  MALLOC_UNLOCK;
2718
 
2719
#endif /* MALLOC_PROVIDED */
2720
}
2721
 
2722
#endif /* DEFINE_FREE */
2723
 
2724
#ifdef DEFINE_REALLOC
2725
 
2726
/*
2727
 
2728
  Realloc algorithm:
2729
 
2730
    Chunks that were obtained via mmap cannot be extended or shrunk
2731
    unless HAVE_MREMAP is defined, in which case mremap is used.
2732
    Otherwise, if their reallocation is for additional space, they are
2733
    copied.  If for less, they are just left alone.
2734
 
2735
    Otherwise, if the reallocation is for additional space, and the
2736
    chunk can be extended, it is, else a malloc-copy-free sequence is
2737
    taken.  There are several different ways that a chunk could be
2738
    extended. All are tried:
2739
 
2740
       * Extending forward into following adjacent free chunk.
2741
       * Shifting backwards, joining preceding adjacent space
2742
       * Both shifting backwards and extending forward.
2743
       * Extending into newly sbrked space
2744
 
2745
    Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2746
    size argument of zero (re)allocates a minimum-sized chunk.
2747
 
2748
    If the reallocation is for less space, and the new request is for
2749
    a `small' (<512 bytes) size, then the newly unused space is lopped
2750
    off and freed.
2751
 
2752
    The old unix realloc convention of allowing the last-free'd chunk
2753
    to be used as an argument to realloc is no longer supported.
2754
    I don't know of any programs still relying on this feature,
2755
    and allowing it would also allow too many other incorrect
2756
    usages of realloc to be sensible.
2757
 
2758
 
2759
*/
2760
 
2761
 
2762
#if __STD_C
2763
Void_t* rEALLOc(RARG Void_t* oldmem, size_t bytes)
2764
#else
2765
Void_t* rEALLOc(RARG oldmem, bytes) RDECL Void_t* oldmem; size_t bytes;
2766
#endif
2767
{
2768
#ifdef MALLOC_PROVIDED
2769
 
2770
  realloc (oldmem, bytes);
2771
 
2772
#else
2773
 
2774
  INTERNAL_SIZE_T    nb;      /* padded request size */
2775
 
2776
  mchunkptr oldp;             /* chunk corresponding to oldmem */
2777
  INTERNAL_SIZE_T    oldsize; /* its size */
2778
 
2779
  mchunkptr newp;             /* chunk to return */
2780
  INTERNAL_SIZE_T    newsize; /* its size */
2781
  Void_t*   newmem;           /* corresponding user mem */
2782
 
2783
  mchunkptr next;             /* next contiguous chunk after oldp */
2784
  INTERNAL_SIZE_T  nextsize;  /* its size */
2785
 
2786
  mchunkptr prev;             /* previous contiguous chunk before oldp */
2787
  INTERNAL_SIZE_T  prevsize;  /* its size */
2788
 
2789
  mchunkptr remainder;        /* holds split off extra space from newp */
2790
  INTERNAL_SIZE_T  remainder_size;   /* its size */
2791
 
2792
  mchunkptr bck;              /* misc temp for linking */
2793
  mchunkptr fwd;              /* misc temp for linking */
2794
 
2795
#ifdef REALLOC_ZERO_BYTES_FREES
2796
  if (bytes == 0) { fREe(RCALL oldmem); return 0; }
2797
#endif
2798
 
2799
 
2800
  /* realloc of null is supposed to be same as malloc */
2801
  if (oldmem == 0) return mALLOc(RCALL bytes);
2802
 
2803
  MALLOC_LOCK;
2804
 
2805
  newp    = oldp    = mem2chunk(oldmem);
2806
  newsize = oldsize = chunksize(oldp);
2807
 
2808
 
2809
  nb = request2size(bytes);
2810
 
2811
  /* Check for overflow and just fail, if so. */
2812
  if (nb > INT_MAX || nb < bytes)
2813
  {
2814
    RERRNO = ENOMEM;
2815
    return 0;
2816
  }
2817
 
2818
#if HAVE_MMAP
2819
  if (chunk_is_mmapped(oldp))
2820
  {
2821
#if HAVE_MREMAP
2822
    newp = mremap_chunk(oldp, nb);
2823
    if(newp)
2824
    {
2825
      MALLOC_UNLOCK;
2826
      return chunk2mem(newp);
2827
    }
2828
#endif
2829
    /* Note the extra SIZE_SZ overhead. */
2830
    if(oldsize - SIZE_SZ >= nb)
2831
    {
2832
      MALLOC_UNLOCK;
2833
      return oldmem; /* do nothing */
2834
    }
2835
    /* Must alloc, copy, free. */
2836
    newmem = mALLOc(RCALL bytes);
2837
    if (newmem == 0)
2838
    {
2839
      MALLOC_UNLOCK;
2840
      return 0; /* propagate failure */
2841
    }
2842
    MALLOC_COPY(newmem, oldmem, oldsize - 2*SIZE_SZ);
2843
    munmap_chunk(oldp);
2844
    MALLOC_UNLOCK;
2845
    return newmem;
2846
  }
2847
#endif
2848
 
2849
  check_inuse_chunk(oldp);
2850
 
2851
  if ((long)(oldsize) < (long)(nb))
2852
  {
2853
 
2854
    /* Try expanding forward */
2855
 
2856
    next = chunk_at_offset(oldp, oldsize);
2857
    if (next == top || !inuse(next))
2858
    {
2859
      nextsize = chunksize(next);
2860
 
2861
      /* Forward into top only if a remainder */
2862
      if (next == top)
2863
      {
2864
        if ((long)(nextsize + newsize) >= (long)(nb + MINSIZE))
2865
        {
2866
          newsize += nextsize;
2867
          top = chunk_at_offset(oldp, nb);
2868
          set_head(top, (newsize - nb) | PREV_INUSE);
2869
          set_head_size(oldp, nb);
2870
          MALLOC_UNLOCK;
2871
          return chunk2mem(oldp);
2872
        }
2873
      }
2874
 
2875
      /* Forward into next chunk */
2876
      else if (((long)(nextsize + newsize) >= (long)(nb)))
2877
      {
2878
        unlink(next, bck, fwd);
2879
        newsize  += nextsize;
2880
        goto split;
2881
      }
2882
    }
2883
    else
2884
    {
2885
      next = 0;
2886
      nextsize = 0;
2887
    }
2888
 
2889
    /* Try shifting backwards. */
2890
 
2891
    if (!prev_inuse(oldp))
2892
    {
2893
      prev = prev_chunk(oldp);
2894
      prevsize = chunksize(prev);
2895
 
2896
      /* try forward + backward first to save a later consolidation */
2897
 
2898
      if (next != 0)
2899
      {
2900
        /* into top */
2901
        if (next == top)
2902
        {
2903
          if ((long)(nextsize + prevsize + newsize) >= (long)(nb + MINSIZE))
2904
          {
2905
            unlink(prev, bck, fwd);
2906
            newp = prev;
2907
            newsize += prevsize + nextsize;
2908
            newmem = chunk2mem(newp);
2909
            MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2910
            top = chunk_at_offset(newp, nb);
2911
            set_head(top, (newsize - nb) | PREV_INUSE);
2912
            set_head_size(newp, nb);
2913
            MALLOC_UNLOCK;
2914
            return newmem;
2915
          }
2916
        }
2917
 
2918
        /* into next chunk */
2919
        else if (((long)(nextsize + prevsize + newsize) >= (long)(nb)))
2920
        {
2921
          unlink(next, bck, fwd);
2922
          unlink(prev, bck, fwd);
2923
          newp = prev;
2924
          newsize += nextsize + prevsize;
2925
          newmem = chunk2mem(newp);
2926
          MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2927
          goto split;
2928
        }
2929
      }
2930
 
2931
      /* backward only */
2932
      if (prev != 0 && (long)(prevsize + newsize) >= (long)nb)
2933
      {
2934
        unlink(prev, bck, fwd);
2935
        newp = prev;
2936
        newsize += prevsize;
2937
        newmem = chunk2mem(newp);
2938
        MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2939
        goto split;
2940
      }
2941
    }
2942
 
2943
    /* Must allocate */
2944
 
2945
    newmem = mALLOc (RCALL bytes);
2946
 
2947
    if (newmem == 0)  /* propagate failure */
2948
    {
2949
      MALLOC_UNLOCK;
2950
      return 0;
2951
    }
2952
 
2953
    /* Avoid copy if newp is next chunk after oldp. */
2954
    /* (This can only happen when new chunk is sbrk'ed.) */
2955
 
2956
    if ( (newp = mem2chunk(newmem)) == next_chunk(oldp))
2957
    {
2958
      newsize += chunksize(newp);
2959
      newp = oldp;
2960
      goto split;
2961
    }
2962
 
2963
    /* Otherwise copy, free, and exit */
2964
    MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2965
    fREe(RCALL oldmem);
2966
    MALLOC_UNLOCK;
2967
    return newmem;
2968
  }
2969
 
2970
 
2971
 split:  /* split off extra room in old or expanded chunk */
2972
 
2973
  remainder_size = long_sub_size_t(newsize, nb);
2974
 
2975
  if (remainder_size >= (long)MINSIZE) /* split off remainder */
2976
  {
2977
    remainder = chunk_at_offset(newp, nb);
2978
    set_head_size(newp, nb);
2979
    set_head(remainder, remainder_size | PREV_INUSE);
2980
    set_inuse_bit_at_offset(remainder, remainder_size);
2981
    fREe(RCALL chunk2mem(remainder)); /* let free() deal with it */
2982
  }
2983
  else
2984
  {
2985
    set_head_size(newp, newsize);
2986
    set_inuse_bit_at_offset(newp, newsize);
2987
  }
2988
 
2989
  check_inuse_chunk(newp);
2990
  MALLOC_UNLOCK;
2991
  return chunk2mem(newp);
2992
 
2993
#endif /* MALLOC_PROVIDED */
2994
}
2995
 
2996
#endif /* DEFINE_REALLOC */
2997
 
2998
#ifdef DEFINE_MEMALIGN
2999
 
3000
/*
3001
 
3002
  memalign algorithm:
3003
 
3004
    memalign requests more than enough space from malloc, finds a spot
3005
    within that chunk that meets the alignment request, and then
3006
    possibly frees the leading and trailing space.
3007
 
3008
    The alignment argument must be a power of two. This property is not
3009
    checked by memalign, so misuse may result in random runtime errors.
3010
 
3011
    8-byte alignment is guaranteed by normal malloc calls, so don't
3012
    bother calling memalign with an argument of 8 or less.
3013
 
3014
    Overreliance on memalign is a sure way to fragment space.
3015
 
3016
*/
3017
 
3018
 
3019
#if __STD_C
3020
Void_t* mEMALIGn(RARG size_t alignment, size_t bytes)
3021
#else
3022
Void_t* mEMALIGn(RARG alignment, bytes) RDECL size_t alignment; size_t bytes;
3023
#endif
3024
{
3025
  INTERNAL_SIZE_T    nb;      /* padded  request size */
3026
  char*     m;                /* memory returned by malloc call */
3027
  mchunkptr p;                /* corresponding chunk */
3028
  char*     brk;              /* alignment point within p */
3029
  mchunkptr newp;             /* chunk to return */
3030
  INTERNAL_SIZE_T  newsize;   /* its size */
3031
  INTERNAL_SIZE_T  leadsize;  /* leading space befor alignment point */
3032
  mchunkptr remainder;        /* spare room at end to split off */
3033
  long      remainder_size;   /* its size */
3034
 
3035
  /* If need less alignment than we give anyway, just relay to malloc */
3036
 
3037
  if (alignment <= MALLOC_ALIGNMENT) return mALLOc(RCALL bytes);
3038
 
3039
  /* Otherwise, ensure that it is at least a minimum chunk size */
3040
 
3041
  if (alignment <  MINSIZE) alignment = MINSIZE;
3042
 
3043
  /* Call malloc with worst case padding to hit alignment. */
3044
 
3045
  nb = request2size(bytes);
3046
 
3047
  /* Check for overflow. */
3048
  if (nb > INT_MAX || nb < bytes)
3049
  {
3050
    RERRNO = ENOMEM;
3051
    return 0;
3052
  }
3053
 
3054
  m  = (char*)(mALLOc(RCALL nb + alignment + MINSIZE));
3055
 
3056
  if (m == 0) return 0; /* propagate failure */
3057
 
3058
  MALLOC_LOCK;
3059
 
3060
  p = mem2chunk(m);
3061
 
3062
  if ((((unsigned long)(m)) % alignment) == 0) /* aligned */
3063
  {
3064
#if HAVE_MMAP
3065
    if(chunk_is_mmapped(p))
3066
    {
3067
      MALLOC_UNLOCK;
3068
      return chunk2mem(p); /* nothing more to do */
3069
    }
3070
#endif
3071
  }
3072
  else /* misaligned */
3073
  {
3074
    /*
3075
      Find an aligned spot inside chunk.
3076
      Since we need to give back leading space in a chunk of at
3077
      least MINSIZE, if the first calculation places us at
3078
      a spot with less than MINSIZE leader, we can move to the
3079
      next aligned spot -- we've allocated enough total room so that
3080
      this is always possible.
3081
    */
3082
 
3083
    brk = (char*)mem2chunk(((unsigned long)(m + alignment - 1)) & -alignment);
3084
    if ((long)(brk - (char*)(p)) < (long)MINSIZE) brk = brk + alignment;
3085
 
3086
    newp = (mchunkptr)brk;
3087
    leadsize = brk - (char*)(p);
3088
    newsize = chunksize(p) - leadsize;
3089
 
3090
#if HAVE_MMAP
3091
    if(chunk_is_mmapped(p))
3092
    {
3093
      newp->prev_size = p->prev_size + leadsize;
3094
      set_head(newp, newsize|IS_MMAPPED);
3095
      MALLOC_UNLOCK;
3096
      return chunk2mem(newp);
3097
    }
3098
#endif
3099
 
3100
    /* give back leader, use the rest */
3101
 
3102
    set_head(newp, newsize | PREV_INUSE);
3103
    set_inuse_bit_at_offset(newp, newsize);
3104
    set_head_size(p, leadsize);
3105
    fREe(RCALL chunk2mem(p));
3106
    p = newp;
3107
 
3108
    assert (newsize >= nb && (((unsigned long)(chunk2mem(p))) % alignment) == 0);
3109
  }
3110
 
3111
  /* Also give back spare room at the end */
3112
 
3113
  remainder_size = long_sub_size_t(chunksize(p), nb);
3114
 
3115
  if (remainder_size >= (long)MINSIZE)
3116
  {
3117
    remainder = chunk_at_offset(p, nb);
3118
    set_head(remainder, remainder_size | PREV_INUSE);
3119
    set_head_size(p, nb);
3120
    fREe(RCALL chunk2mem(remainder));
3121
  }
3122
 
3123
  check_inuse_chunk(p);
3124
  MALLOC_UNLOCK;
3125
  return chunk2mem(p);
3126
 
3127
}
3128
 
3129
#endif /* DEFINE_MEMALIGN */
3130
 
3131
#ifdef DEFINE_VALLOC
3132
 
3133
/*
3134
    valloc just invokes memalign with alignment argument equal
3135
    to the page size of the system (or as near to this as can
3136
    be figured out from all the includes/defines above.)
3137
*/
3138
 
3139
#if __STD_C
3140
Void_t* vALLOc(RARG size_t bytes)
3141
#else
3142
Void_t* vALLOc(RARG bytes) RDECL size_t bytes;
3143
#endif
3144
{
3145
  return mEMALIGn (RCALL malloc_getpagesize, bytes);
3146
}
3147
 
3148
#endif /* DEFINE_VALLOC */
3149
 
3150
#ifdef DEFINE_PVALLOC
3151
 
3152
/*
3153
  pvalloc just invokes valloc for the nearest pagesize
3154
  that will accommodate request
3155
*/
3156
 
3157
 
3158
#if __STD_C
3159
Void_t* pvALLOc(RARG size_t bytes)
3160
#else
3161
Void_t* pvALLOc(RARG bytes) RDECL size_t bytes;
3162
#endif
3163
{
3164
  size_t pagesize = malloc_getpagesize;
3165
  return mEMALIGn (RCALL pagesize, (bytes + pagesize - 1) & ~(pagesize - 1));
3166
}
3167
 
3168
#endif /* DEFINE_PVALLOC */
3169
 
3170
#ifdef DEFINE_CALLOC
3171
 
3172
/*
3173
 
3174
  calloc calls malloc, then zeroes out the allocated chunk.
3175
 
3176
*/
3177
 
3178
#if __STD_C
3179
Void_t* cALLOc(RARG size_t n, size_t elem_size)
3180
#else
3181
Void_t* cALLOc(RARG n, elem_size) RDECL size_t n; size_t elem_size;
3182
#endif
3183
{
3184
  mchunkptr p;
3185
  INTERNAL_SIZE_T csz;
3186
 
3187
  INTERNAL_SIZE_T sz = n * elem_size;
3188
 
3189
#if MORECORE_CLEARS
3190
  mchunkptr oldtop;
3191
  INTERNAL_SIZE_T oldtopsize;
3192
#endif
3193
  Void_t* mem;
3194
 
3195
  /* check if expand_top called, in which case don't need to clear */
3196
#if MORECORE_CLEARS
3197
  MALLOC_LOCK;
3198
  oldtop = top;
3199
  oldtopsize = chunksize(top);
3200
#endif
3201
 
3202
  mem = mALLOc (RCALL sz);
3203
 
3204
  if (mem == 0)
3205
  {
3206
#if MORECORE_CLEARS
3207
    MALLOC_UNLOCK;
3208
#endif
3209
    return 0;
3210
  }
3211
  else
3212
  {
3213
    p = mem2chunk(mem);
3214
 
3215
    /* Two optional cases in which clearing not necessary */
3216
 
3217
 
3218
#if HAVE_MMAP
3219
    if (chunk_is_mmapped(p))
3220
    {
3221
#if MORECORE_CLEARS
3222
      MALLOC_UNLOCK;
3223
#endif
3224
      return mem;
3225
    }
3226
#endif
3227
 
3228
    csz = chunksize(p);
3229
 
3230
#if MORECORE_CLEARS
3231
    if (p == oldtop && csz > oldtopsize)
3232
    {
3233
      /* clear only the bytes from non-freshly-sbrked memory */
3234
      csz = oldtopsize;
3235
    }
3236
    MALLOC_UNLOCK;
3237
#endif
3238
 
3239
    MALLOC_ZERO(mem, csz - SIZE_SZ);
3240
    return mem;
3241
  }
3242
}
3243
 
3244
#endif /* DEFINE_CALLOC */
3245
 
3246
#if defined(DEFINE_CFREE) && !defined(__CYGWIN__)
3247
 
3248
/*
3249
 
3250
  cfree just calls free. It is needed/defined on some systems
3251
  that pair it with calloc, presumably for odd historical reasons.
3252
 
3253
*/
3254
 
3255
#if !defined(INTERNAL_LINUX_C_LIB) || !defined(__ELF__)
3256
#if !defined(INTERNAL_NEWLIB) || !defined(_REENT_ONLY)
3257
#if __STD_C
3258
void cfree(Void_t *mem)
3259
#else
3260
void cfree(mem) Void_t *mem;
3261
#endif
3262
{
3263
#ifdef INTERNAL_NEWLIB
3264
  fREe(_REENT, mem);
3265
#else
3266
  fREe(mem);
3267
#endif
3268
}
3269
#endif
3270
#endif
3271
 
3272
#endif /* DEFINE_CFREE */
3273
 
3274
#ifdef DEFINE_FREE
3275
 
3276
/*
3277
 
3278
    Malloc_trim gives memory back to the system (via negative
3279
    arguments to sbrk) if there is unused memory at the `high' end of
3280
    the malloc pool. You can call this after freeing large blocks of
3281
    memory to potentially reduce the system-level memory requirements
3282
    of a program. However, it cannot guarantee to reduce memory. Under
3283
    some allocation patterns, some large free blocks of memory will be
3284
    locked between two used chunks, so they cannot be given back to
3285
    the system.
3286
 
3287
    The `pad' argument to malloc_trim represents the amount of free
3288
    trailing space to leave untrimmed. If this argument is zero,
3289
    only the minimum amount of memory to maintain internal data
3290
    structures will be left (one page or less). Non-zero arguments
3291
    can be supplied to maintain enough trailing space to service
3292
    future expected allocations without having to re-obtain memory
3293
    from the system.
3294
 
3295
    Malloc_trim returns 1 if it actually released any memory, else 0.
3296
 
3297
*/
3298
 
3299
#if __STD_C
3300
int malloc_trim(RARG size_t pad)
3301
#else
3302
int malloc_trim(RARG pad) RDECL size_t pad;
3303
#endif
3304
{
3305
  long  top_size;        /* Amount of top-most memory */
3306
  long  extra;           /* Amount to release */
3307
  char* current_brk;     /* address returned by pre-check sbrk call */
3308
  char* new_brk;         /* address returned by negative sbrk call */
3309
 
3310
  unsigned long pagesz = malloc_getpagesize;
3311
 
3312
  MALLOC_LOCK;
3313
 
3314
  top_size = chunksize(top);
3315
  extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
3316
 
3317
  if (extra < (long)pagesz)  /* Not enough memory to release */
3318
  {
3319
    MALLOC_UNLOCK;
3320
    return 0;
3321
  }
3322
 
3323
  else
3324
  {
3325
    /* Test to make sure no one else called sbrk */
3326
    current_brk = (char*)(MORECORE (0));
3327
    if (current_brk != (char*)(top) + top_size)
3328
    {
3329
      MALLOC_UNLOCK;
3330
      return 0;     /* Apparently we don't own memory; must fail */
3331
    }
3332
 
3333
    else
3334
    {
3335
      new_brk = (char*)(MORECORE (-extra));
3336
 
3337
      if (new_brk == (char*)(MORECORE_FAILURE)) /* sbrk failed? */
3338
      {
3339
        /* Try to figure out what we have */
3340
        current_brk = (char*)(MORECORE (0));
3341
        top_size = current_brk - (char*)top;
3342
        if (top_size >= (long)MINSIZE) /* if not, we are very very dead! */
3343
        {
3344
          sbrked_mem = current_brk - sbrk_base;
3345
          set_head(top, top_size | PREV_INUSE);
3346
        }
3347
        check_chunk(top);
3348
        MALLOC_UNLOCK;
3349
        return 0;
3350
      }
3351
 
3352
      else
3353
      {
3354
        /* Success. Adjust top accordingly. */
3355
        set_head(top, (top_size - extra) | PREV_INUSE);
3356
        sbrked_mem -= extra;
3357
        check_chunk(top);
3358
        MALLOC_UNLOCK;
3359
        return 1;
3360
      }
3361
    }
3362
  }
3363
}
3364
 
3365
#endif /* DEFINE_FREE */
3366
 
3367
#ifdef DEFINE_MALLOC_USABLE_SIZE
3368
 
3369
/*
3370
  malloc_usable_size:
3371
 
3372
    This routine tells you how many bytes you can actually use in an
3373
    allocated chunk, which may be more than you requested (although
3374
    often not). You can use this many bytes without worrying about
3375
    overwriting other allocated objects. Not a particularly great
3376
    programming practice, but still sometimes useful.
3377
 
3378
*/
3379
 
3380
#if __STD_C
3381
size_t malloc_usable_size(RARG Void_t* mem)
3382
#else
3383
size_t malloc_usable_size(RARG mem) RDECL Void_t* mem;
3384
#endif
3385
{
3386
  mchunkptr p;
3387
  if (mem == 0)
3388
    return 0;
3389
  else
3390
  {
3391
    p = mem2chunk(mem);
3392
    if(!chunk_is_mmapped(p))
3393
    {
3394
      if (!inuse(p)) return 0;
3395
#if DEBUG
3396
      MALLOC_LOCK;
3397
      check_inuse_chunk(p);
3398
      MALLOC_UNLOCK;
3399
#endif
3400
      return chunksize(p) - SIZE_SZ;
3401
    }
3402
    return chunksize(p) - 2*SIZE_SZ;
3403
  }
3404
}
3405
 
3406
#endif /* DEFINE_MALLOC_USABLE_SIZE */
3407
 
3408
#ifdef DEFINE_MALLINFO
3409
 
3410
/* Utility to update current_mallinfo for malloc_stats and mallinfo() */
3411
 
3412
STATIC void malloc_update_mallinfo()
3413
{
3414
  int i;
3415
  mbinptr b;
3416
  mchunkptr p;
3417
#if DEBUG
3418
  mchunkptr q;
3419
#endif
3420
 
3421
  INTERNAL_SIZE_T avail = chunksize(top);
3422
  int   navail = ((long)(avail) >= (long)MINSIZE)? 1 : 0;
3423
 
3424
  for (i = 1; i < NAV; ++i)
3425
  {
3426
    b = bin_at(i);
3427
    for (p = last(b); p != b; p = p->bk)
3428
    {
3429
#if DEBUG
3430
      check_free_chunk(p);
3431
      for (q = next_chunk(p);
3432
           q < top && inuse(q) && (long)(chunksize(q)) >= (long)MINSIZE;
3433
           q = next_chunk(q))
3434
        check_inuse_chunk(q);
3435
#endif
3436
      avail += chunksize(p);
3437
      navail++;
3438
    }
3439
  }
3440
 
3441
  current_mallinfo.ordblks = navail;
3442
  current_mallinfo.uordblks = sbrked_mem - avail;
3443
  current_mallinfo.fordblks = avail;
3444
#if HAVE_MMAP
3445
  current_mallinfo.hblks = n_mmaps;
3446
  current_mallinfo.hblkhd = mmapped_mem;
3447
#endif
3448
  current_mallinfo.keepcost = chunksize(top);
3449
 
3450
}
3451
 
3452
#else /* ! DEFINE_MALLINFO */
3453
 
3454
#if __STD_C
3455
extern void malloc_update_mallinfo(void);
3456
#else
3457
extern void malloc_update_mallinfo();
3458
#endif
3459
 
3460
#endif /* ! DEFINE_MALLINFO */
3461
 
3462
#ifdef DEFINE_MALLOC_STATS
3463
 
3464
/*
3465
 
3466
  malloc_stats:
3467
 
3468
    Prints on stderr the amount of space obtain from the system (both
3469
    via sbrk and mmap), the maximum amount (which may be more than
3470
    current if malloc_trim and/or munmap got called), the maximum
3471
    number of simultaneous mmap regions used, and the current number
3472
    of bytes allocated via malloc (or realloc, etc) but not yet
3473
    freed. (Note that this is the number of bytes allocated, not the
3474
    number requested. It will be larger than the number requested
3475
    because of alignment and bookkeeping overhead.)
3476
 
3477
*/
3478
 
3479
#if __STD_C
3480
void malloc_stats(RONEARG)
3481
#else
3482
void malloc_stats(RONEARG) RDECL
3483
#endif
3484
{
3485
  unsigned long local_max_total_mem;
3486
  int local_sbrked_mem;
3487
  struct mallinfo local_mallinfo;
3488
#if HAVE_MMAP
3489
  unsigned long local_mmapped_mem, local_max_n_mmaps;
3490
#endif
3491
  FILE *fp;
3492
 
3493
  MALLOC_LOCK;
3494
  malloc_update_mallinfo();
3495
  local_max_total_mem = max_total_mem;
3496
  local_sbrked_mem = sbrked_mem;
3497
  local_mallinfo = current_mallinfo;
3498
#if HAVE_MMAP
3499
  local_mmapped_mem = mmapped_mem;
3500
  local_max_n_mmaps = max_n_mmaps;
3501
#endif
3502
  MALLOC_UNLOCK;
3503
 
3504
#ifdef INTERNAL_NEWLIB
3505
  _REENT_SMALL_CHECK_INIT(reent_ptr);
3506
  fp = _stderr_r(reent_ptr);
3507
#define fprintf fiprintf
3508
#else
3509
  fp = stderr;
3510
#endif
3511
 
3512
  fprintf(fp, "max system bytes = %10u\n",
3513
          (unsigned int)(local_max_total_mem));
3514
#if HAVE_MMAP
3515
  fprintf(fp, "system bytes     = %10u\n",
3516
          (unsigned int)(local_sbrked_mem + local_mmapped_mem));
3517
  fprintf(fp, "in use bytes     = %10u\n",
3518
          (unsigned int)(local_mallinfo.uordblks + local_mmapped_mem));
3519
#else
3520
  fprintf(fp, "system bytes     = %10u\n",
3521
          (unsigned int)local_sbrked_mem);
3522
  fprintf(fp, "in use bytes     = %10u\n",
3523
          (unsigned int)local_mallinfo.uordblks);
3524
#endif
3525
#if HAVE_MMAP
3526
  fprintf(fp, "max mmap regions = %10u\n",
3527
          (unsigned int)local_max_n_mmaps);
3528
#endif
3529
}
3530
 
3531
#endif /* DEFINE_MALLOC_STATS */
3532
 
3533
#ifdef DEFINE_MALLINFO
3534
 
3535
/*
3536
  mallinfo returns a copy of updated current mallinfo.
3537
*/
3538
 
3539
#if __STD_C
3540
struct mallinfo mALLINFo(RONEARG)
3541
#else
3542
struct mallinfo mALLINFo(RONEARG) RDECL
3543
#endif
3544
{
3545
  struct mallinfo ret;
3546
 
3547
  MALLOC_LOCK;
3548
  malloc_update_mallinfo();
3549
  ret = current_mallinfo;
3550
  MALLOC_UNLOCK;
3551
  return ret;
3552
}
3553
 
3554
#endif /* DEFINE_MALLINFO */
3555
 
3556
#ifdef DEFINE_MALLOPT
3557
 
3558
/*
3559
  mallopt:
3560
 
3561
    mallopt is the general SVID/XPG interface to tunable parameters.
3562
    The format is to provide a (parameter-number, parameter-value) pair.
3563
    mallopt then sets the corresponding parameter to the argument
3564
    value if it can (i.e., so long as the value is meaningful),
3565
    and returns 1 if successful else 0.
3566
 
3567
    See descriptions of tunable parameters above.
3568
 
3569
*/
3570
 
3571
#if __STD_C
3572
int mALLOPt(RARG int param_number, int value)
3573
#else
3574
int mALLOPt(RARG param_number, value) RDECL int param_number; int value;
3575
#endif
3576
{
3577
  MALLOC_LOCK;
3578
  switch(param_number)
3579
  {
3580
    case M_TRIM_THRESHOLD:
3581
      trim_threshold = value; MALLOC_UNLOCK; return 1;
3582
    case M_TOP_PAD:
3583
      top_pad = value; MALLOC_UNLOCK; return 1;
3584
    case M_MMAP_THRESHOLD:
3585
#if HAVE_MMAP
3586
      mmap_threshold = value;
3587
#endif
3588
      MALLOC_UNLOCK;
3589
      return 1;
3590
    case M_MMAP_MAX:
3591
#if HAVE_MMAP
3592
      n_mmaps_max = value; MALLOC_UNLOCK; return 1;
3593
#else
3594
      MALLOC_UNLOCK; return value == 0;
3595
#endif
3596
 
3597
    default:
3598
      MALLOC_UNLOCK;
3599
      return 0;
3600
  }
3601
}
3602
 
3603
#endif /* DEFINE_MALLOPT */
3604
 
3605
/*
3606
 
3607
History:
3608
 
3609
    V2.6.3 Sun May 19 08:17:58 1996  Doug Lea  (dl at gee)
3610
      * Added pvalloc, as recommended by H.J. Liu
3611
      * Added 64bit pointer support mainly from Wolfram Gloger
3612
      * Added anonymously donated WIN32 sbrk emulation
3613
      * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3614
      * malloc_extend_top: fix mask error that caused wastage after
3615
        foreign sbrks
3616
      * Add linux mremap support code from HJ Liu
3617
 
3618
    V2.6.2 Tue Dec  5 06:52:55 1995  Doug Lea  (dl at gee)
3619
      * Integrated most documentation with the code.
3620
      * Add support for mmap, with help from
3621
        Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3622
      * Use last_remainder in more cases.
3623
      * Pack bins using idea from  colin@nyx10.cs.du.edu
3624
      * Use ordered bins instead of best-fit threshhold
3625
      * Eliminate block-local decls to simplify tracing and debugging.
3626
      * Support another case of realloc via move into top
3627
      * Fix error occuring when initial sbrk_base not word-aligned.
3628
      * Rely on page size for units instead of SBRK_UNIT to
3629
        avoid surprises about sbrk alignment conventions.
3630
      * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3631
        (raymond@es.ele.tue.nl) for the suggestion.
3632
      * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3633
      * More precautions for cases where other routines call sbrk,
3634
        courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3635
      * Added macros etc., allowing use in linux libc from
3636
        H.J. Lu (hjl@gnu.ai.mit.edu)
3637
      * Inverted this history list
3638
 
3639
    V2.6.1 Sat Dec  2 14:10:57 1995  Doug Lea  (dl at gee)
3640
      * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3641
      * Removed all preallocation code since under current scheme
3642
        the work required to undo bad preallocations exceeds
3643
        the work saved in good cases for most test programs.
3644
      * No longer use return list or unconsolidated bins since
3645
        no scheme using them consistently outperforms those that don't
3646
        given above changes.
3647
      * Use best fit for very large chunks to prevent some worst-cases.
3648
      * Added some support for debugging
3649
 
3650
    V2.6.0 Sat Nov  4 07:05:23 1995  Doug Lea  (dl at gee)
3651
      * Removed footers when chunks are in use. Thanks to
3652
        Paul Wilson (wilson@cs.texas.edu) for the suggestion.
3653
 
3654
    V2.5.4 Wed Nov  1 07:54:51 1995  Doug Lea  (dl at gee)
3655
      * Added malloc_trim, with help from Wolfram Gloger
3656
        (wmglo@Dent.MED.Uni-Muenchen.DE).
3657
 
3658
    V2.5.3 Tue Apr 26 10:16:01 1994  Doug Lea  (dl at g)
3659
 
3660
    V2.5.2 Tue Apr  5 16:20:40 1994  Doug Lea  (dl at g)
3661
      * realloc: try to expand in both directions
3662
      * malloc: swap order of clean-bin strategy;
3663
      * realloc: only conditionally expand backwards
3664
      * Try not to scavenge used bins
3665
      * Use bin counts as a guide to preallocation
3666
      * Occasionally bin return list chunks in first scan
3667
      * Add a few optimizations from colin@nyx10.cs.du.edu
3668
 
3669
    V2.5.1 Sat Aug 14 15:40:43 1993  Doug Lea  (dl at g)
3670
      * faster bin computation & slightly different binning
3671
      * merged all consolidations to one part of malloc proper
3672
         (eliminating old malloc_find_space & malloc_clean_bin)
3673
      * Scan 2 returns chunks (not just 1)
3674
      * Propagate failure in realloc if malloc returns 0
3675
      * Add stuff to allow compilation on non-ANSI compilers
3676
          from kpv@research.att.com
3677
 
3678
    V2.5 Sat Aug  7 07:41:59 1993  Doug Lea  (dl at g.oswego.edu)
3679
      * removed potential for odd address access in prev_chunk
3680
      * removed dependency on getpagesize.h
3681
      * misc cosmetics and a bit more internal documentation
3682
      * anticosmetics: mangled names in macros to evade debugger strangeness
3683
      * tested on sparc, hp-700, dec-mips, rs6000
3684
          with gcc & native cc (hp, dec only) allowing
3685
          Detlefs & Zorn comparison study (in SIGPLAN Notices.)
3686
 
3687
    Trial version Fri Aug 28 13:14:29 1992  Doug Lea  (dl at g.oswego.edu)
3688
      * Based loosely on libg++-1.2X malloc. (It retains some of the overall
3689
         structure of old version,  but most details differ.)
3690
 
3691
*/
3692
#endif

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.