OpenCores
URL https://opencores.org/ocsvn/or1k/or1k/trunk

Subversion Repositories or1k

[/] [or1k/] [trunk/] [newlib/] [newlib/] [libc/] [stdlib/] [mallocr.c] - Blame information for rev 39

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

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

powered by: WebSVN 2.1.0

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