xref: /openbmc/u-boot/common/dlmalloc.c (revision 544d97e9)
1 #include <common.h>
2 
3 #if 0	/* Moved to malloc.h */
4 /* ---------- To make a malloc.h, start cutting here ------------ */
5 
6 /*
7   A version of malloc/free/realloc written by Doug Lea and released to the
8   public domain.  Send questions/comments/complaints/performance data
9   to dl@cs.oswego.edu
10 
11 * VERSION 2.6.6  Sun Mar  5 19:10:03 2000  Doug Lea  (dl at gee)
12 
13    Note: There may be an updated version of this malloc obtainable at
14 	   ftp://g.oswego.edu/pub/misc/malloc.c
15 	 Check before installing!
16 
17 * Why use this malloc?
18 
19   This is not the fastest, most space-conserving, most portable, or
20   most tunable malloc ever written. However it is among the fastest
21   while also being among the most space-conserving, portable and tunable.
22   Consistent balance across these factors results in a good general-purpose
23   allocator. For a high-level description, see
24      http://g.oswego.edu/dl/html/malloc.html
25 
26 * Synopsis of public routines
27 
28   (Much fuller descriptions are contained in the program documentation below.)
29 
30   malloc(size_t n);
31      Return a pointer to a newly allocated chunk of at least n bytes, or null
32      if no space is available.
33   free(Void_t* p);
34      Release the chunk of memory pointed to by p, or no effect if p is null.
35   realloc(Void_t* p, size_t n);
36      Return a pointer to a chunk of size n that contains the same data
37      as does chunk p up to the minimum of (n, p's size) bytes, or null
38      if no space is available. The returned pointer may or may not be
39      the same as p. If p is null, equivalent to malloc.  Unless the
40      #define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
41      size argument of zero (re)allocates a minimum-sized chunk.
42   memalign(size_t alignment, size_t n);
43      Return a pointer to a newly allocated chunk of n bytes, aligned
44      in accord with the alignment argument, which must be a power of
45      two.
46   valloc(size_t n);
47      Equivalent to memalign(pagesize, n), where pagesize is the page
48      size of the system (or as near to this as can be figured out from
49      all the includes/defines below.)
50   pvalloc(size_t n);
51      Equivalent to valloc(minimum-page-that-holds(n)), that is,
52      round up n to nearest pagesize.
53   calloc(size_t unit, size_t quantity);
54      Returns a pointer to quantity * unit bytes, with all locations
55      set to zero.
56   cfree(Void_t* p);
57      Equivalent to free(p).
58   malloc_trim(size_t pad);
59      Release all but pad bytes of freed top-most memory back
60      to the system. Return 1 if successful, else 0.
61   malloc_usable_size(Void_t* p);
62      Report the number usable allocated bytes associated with allocated
63      chunk p. This may or may not report more bytes than were requested,
64      due to alignment and minimum size constraints.
65   malloc_stats();
66      Prints brief summary statistics.
67   mallinfo()
68      Returns (by copy) a struct containing various summary statistics.
69   mallopt(int parameter_number, int parameter_value)
70      Changes one of the tunable parameters described below. Returns
71      1 if successful in changing the parameter, else 0.
72 
73 * Vital statistics:
74 
75   Alignment:                            8-byte
76        8 byte alignment is currently hardwired into the design.  This
77        seems to suffice for all current machines and C compilers.
78 
79   Assumed pointer representation:       4 or 8 bytes
80        Code for 8-byte pointers is untested by me but has worked
81        reliably by Wolfram Gloger, who contributed most of the
82        changes supporting this.
83 
84   Assumed size_t  representation:       4 or 8 bytes
85        Note that size_t is allowed to be 4 bytes even if pointers are 8.
86 
87   Minimum overhead per allocated chunk: 4 or 8 bytes
88        Each malloced chunk has a hidden overhead of 4 bytes holding size
89        and status information.
90 
91   Minimum allocated size: 4-byte ptrs:  16 bytes    (including 4 overhead)
92 			  8-byte ptrs:  24/32 bytes (including, 4/8 overhead)
93 
94        When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
95        ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
96        needed; 4 (8) for a trailing size field
97        and 8 (16) bytes for free list pointers. Thus, the minimum
98        allocatable size is 16/24/32 bytes.
99 
100        Even a request for zero bytes (i.e., malloc(0)) returns a
101        pointer to something of the minimum allocatable size.
102 
103   Maximum allocated size: 4-byte size_t: 2^31 -  8 bytes
104 			  8-byte size_t: 2^63 - 16 bytes
105 
106        It is assumed that (possibly signed) size_t bit values suffice to
107        represent chunk sizes. `Possibly signed' is due to the fact
108        that `size_t' may be defined on a system as either a signed or
109        an unsigned type. To be conservative, values that would appear
110        as negative numbers are avoided.
111        Requests for sizes with a negative sign bit when the request
112        size is treaded as a long will return null.
113 
114   Maximum overhead wastage per allocated chunk: normally 15 bytes
115 
116        Alignnment demands, plus the minimum allocatable size restriction
117        make the normal worst-case wastage 15 bytes (i.e., up to 15
118        more bytes will be allocated than were requested in malloc), with
119        two exceptions:
120 	 1. Because requests for zero bytes allocate non-zero space,
121 	    the worst case wastage for a request of zero bytes is 24 bytes.
122 	 2. For requests >= mmap_threshold that are serviced via
123 	    mmap(), the worst case wastage is 8 bytes plus the remainder
124 	    from a system page (the minimal mmap unit); typically 4096 bytes.
125 
126 * Limitations
127 
128     Here are some features that are NOT currently supported
129 
130     * No user-definable hooks for callbacks and the like.
131     * No automated mechanism for fully checking that all accesses
132       to malloced memory stay within their bounds.
133     * No support for compaction.
134 
135 * Synopsis of compile-time options:
136 
137     People have reported using previous versions of this malloc on all
138     versions of Unix, sometimes by tweaking some of the defines
139     below. It has been tested most extensively on Solaris and
140     Linux. It is also reported to work on WIN32 platforms.
141     People have also reported adapting this malloc for use in
142     stand-alone embedded systems.
143 
144     The implementation is in straight, hand-tuned ANSI C.  Among other
145     consequences, it uses a lot of macros.  Because of this, to be at
146     all usable, this code should be compiled using an optimizing compiler
147     (for example gcc -O2) that can simplify expressions and control
148     paths.
149 
150   __STD_C                  (default: derived from C compiler defines)
151      Nonzero if using ANSI-standard C compiler, a C++ compiler, or
152      a C compiler sufficiently close to ANSI to get away with it.
153   DEBUG                    (default: NOT defined)
154      Define to enable debugging. Adds fairly extensive assertion-based
155      checking to help track down memory errors, but noticeably slows down
156      execution.
157   REALLOC_ZERO_BYTES_FREES (default: NOT defined)
158      Define this if you think that realloc(p, 0) should be equivalent
159      to free(p). Otherwise, since malloc returns a unique pointer for
160      malloc(0), so does realloc(p, 0).
161   HAVE_MEMCPY               (default: defined)
162      Define if you are not otherwise using ANSI STD C, but still
163      have memcpy and memset in your C library and want to use them.
164      Otherwise, simple internal versions are supplied.
165   USE_MEMCPY               (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
166      Define as 1 if you want the C library versions of memset and
167      memcpy called in realloc and calloc (otherwise macro versions are used).
168      At least on some platforms, the simple macro versions usually
169      outperform libc versions.
170   HAVE_MMAP                 (default: defined as 1)
171      Define to non-zero to optionally make malloc() use mmap() to
172      allocate very large blocks.
173   HAVE_MREMAP                 (default: defined as 0 unless Linux libc set)
174      Define to non-zero to optionally make realloc() use mremap() to
175      reallocate very large blocks.
176   malloc_getpagesize        (default: derived from system #includes)
177      Either a constant or routine call returning the system page size.
178   HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
179      Optionally define if you are on a system with a /usr/include/malloc.h
180      that declares struct mallinfo. It is not at all necessary to
181      define this even if you do, but will ensure consistency.
182   INTERNAL_SIZE_T           (default: size_t)
183      Define to a 32-bit type (probably `unsigned int') if you are on a
184      64-bit machine, yet do not want or need to allow malloc requests of
185      greater than 2^31 to be handled. This saves space, especially for
186      very small chunks.
187   INTERNAL_LINUX_C_LIB      (default: NOT defined)
188      Defined only when compiled as part of Linux libc.
189      Also note that there is some odd internal name-mangling via defines
190      (for example, internally, `malloc' is named `mALLOc') needed
191      when compiling in this case. These look funny but don't otherwise
192      affect anything.
193   WIN32                     (default: undefined)
194      Define this on MS win (95, nt) platforms to compile in sbrk emulation.
195   LACKS_UNISTD_H            (default: undefined if not WIN32)
196      Define this if your system does not have a <unistd.h>.
197   LACKS_SYS_PARAM_H         (default: undefined if not WIN32)
198      Define this if your system does not have a <sys/param.h>.
199   MORECORE                  (default: sbrk)
200      The name of the routine to call to obtain more memory from the system.
201   MORECORE_FAILURE          (default: -1)
202      The value returned upon failure of MORECORE.
203   MORECORE_CLEARS           (default 1)
204      True (1) if the routine mapped to MORECORE zeroes out memory (which
205      holds for sbrk).
206   DEFAULT_TRIM_THRESHOLD
207   DEFAULT_TOP_PAD
208   DEFAULT_MMAP_THRESHOLD
209   DEFAULT_MMAP_MAX
210      Default values of tunable parameters (described in detail below)
211      controlling interaction with host system routines (sbrk, mmap, etc).
212      These values may also be changed dynamically via mallopt(). The
213      preset defaults are those that give best performance for typical
214      programs/systems.
215   USE_DL_PREFIX             (default: undefined)
216      Prefix all public routines with the string 'dl'.  Useful to
217      quickly avoid procedure declaration conflicts and linker symbol
218      conflicts with existing memory allocation routines.
219 
220 
221 */
222 
223 
224 
225 /* Preliminaries */
226 
227 #ifndef __STD_C
228 #ifdef __STDC__
229 #define __STD_C     1
230 #else
231 #if __cplusplus
232 #define __STD_C     1
233 #else
234 #define __STD_C     0
235 #endif /*__cplusplus*/
236 #endif /*__STDC__*/
237 #endif /*__STD_C*/
238 
239 #ifndef Void_t
240 #if (__STD_C || defined(WIN32))
241 #define Void_t      void
242 #else
243 #define Void_t      char
244 #endif
245 #endif /*Void_t*/
246 
247 #if __STD_C
248 #include <stddef.h>   /* for size_t */
249 #else
250 #include <sys/types.h>
251 #endif
252 
253 #ifdef __cplusplus
254 extern "C" {
255 #endif
256 
257 #include <stdio.h>    /* needed for malloc_stats */
258 
259 
260 /*
261   Compile-time options
262 */
263 
264 
265 /*
266     Debugging:
267 
268     Because freed chunks may be overwritten with link fields, this
269     malloc will often die when freed memory is overwritten by user
270     programs.  This can be very effective (albeit in an annoying way)
271     in helping track down dangling pointers.
272 
273     If you compile with -DDEBUG, a number of assertion checks are
274     enabled that will catch more memory errors. You probably won't be
275     able to make much sense of the actual assertion errors, but they
276     should help you locate incorrectly overwritten memory.  The
277     checking is fairly extensive, and will slow down execution
278     noticeably. Calling malloc_stats or mallinfo with DEBUG set will
279     attempt to check every non-mmapped allocated and free chunk in the
280     course of computing the summmaries. (By nature, mmapped regions
281     cannot be checked very much automatically.)
282 
283     Setting DEBUG may also be helpful if you are trying to modify
284     this code. The assertions in the check routines spell out in more
285     detail the assumptions and invariants underlying the algorithms.
286 
287 */
288 
289 #ifdef DEBUG
290 #include <assert.h>
291 #else
292 #define assert(x) ((void)0)
293 #endif
294 
295 
296 /*
297   INTERNAL_SIZE_T is the word-size used for internal bookkeeping
298   of chunk sizes. On a 64-bit machine, you can reduce malloc
299   overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
300   at the expense of not being able to handle requests greater than
301   2^31. This limitation is hardly ever a concern; you are encouraged
302   to set this. However, the default version is the same as size_t.
303 */
304 
305 #ifndef INTERNAL_SIZE_T
306 #define INTERNAL_SIZE_T size_t
307 #endif
308 
309 /*
310   REALLOC_ZERO_BYTES_FREES should be set if a call to
311   realloc with zero bytes should be the same as a call to free.
312   Some people think it should. Otherwise, since this malloc
313   returns a unique pointer for malloc(0), so does realloc(p, 0).
314 */
315 
316 
317 /*   #define REALLOC_ZERO_BYTES_FREES */
318 
319 
320 /*
321   WIN32 causes an emulation of sbrk to be compiled in
322   mmap-based options are not currently supported in WIN32.
323 */
324 
325 /* #define WIN32 */
326 #ifdef WIN32
327 #define MORECORE wsbrk
328 #define HAVE_MMAP 0
329 
330 #define LACKS_UNISTD_H
331 #define LACKS_SYS_PARAM_H
332 
333 /*
334   Include 'windows.h' to get the necessary declarations for the
335   Microsoft Visual C++ data structures and routines used in the 'sbrk'
336   emulation.
337 
338   Define WIN32_LEAN_AND_MEAN so that only the essential Microsoft
339   Visual C++ header files are included.
340 */
341 #define WIN32_LEAN_AND_MEAN
342 #include <windows.h>
343 #endif
344 
345 
346 /*
347   HAVE_MEMCPY should be defined if you are not otherwise using
348   ANSI STD C, but still have memcpy and memset in your C library
349   and want to use them in calloc and realloc. Otherwise simple
350   macro versions are defined here.
351 
352   USE_MEMCPY should be defined as 1 if you actually want to
353   have memset and memcpy called. People report that the macro
354   versions are often enough faster than libc versions on many
355   systems that it is better to use them.
356 
357 */
358 
359 #define HAVE_MEMCPY
360 
361 #ifndef USE_MEMCPY
362 #ifdef HAVE_MEMCPY
363 #define USE_MEMCPY 1
364 #else
365 #define USE_MEMCPY 0
366 #endif
367 #endif
368 
369 #if (__STD_C || defined(HAVE_MEMCPY))
370 
371 #if __STD_C
372 void* memset(void*, int, size_t);
373 void* memcpy(void*, const void*, size_t);
374 #else
375 #ifdef WIN32
376 /* On Win32 platforms, 'memset()' and 'memcpy()' are already declared in */
377 /* 'windows.h' */
378 #else
379 Void_t* memset();
380 Void_t* memcpy();
381 #endif
382 #endif
383 #endif
384 
385 #if USE_MEMCPY
386 
387 /* The following macros are only invoked with (2n+1)-multiples of
388    INTERNAL_SIZE_T units, with a positive integer n. This is exploited
389    for fast inline execution when n is small. */
390 
391 #define MALLOC_ZERO(charp, nbytes)                                            \
392 do {                                                                          \
393   INTERNAL_SIZE_T mzsz = (nbytes);                                            \
394   if(mzsz <= 9*sizeof(mzsz)) {                                                \
395     INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp);                         \
396     if(mzsz >= 5*sizeof(mzsz)) {     *mz++ = 0;                               \
397 				     *mz++ = 0;                               \
398       if(mzsz >= 7*sizeof(mzsz)) {   *mz++ = 0;                               \
399 				     *mz++ = 0;                               \
400 	if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0;                               \
401 				     *mz++ = 0; }}}                           \
402 				     *mz++ = 0;                               \
403 				     *mz++ = 0;                               \
404 				     *mz   = 0;                               \
405   } else memset((charp), 0, mzsz);                                            \
406 } while(0)
407 
408 #define MALLOC_COPY(dest,src,nbytes)                                          \
409 do {                                                                          \
410   INTERNAL_SIZE_T mcsz = (nbytes);                                            \
411   if(mcsz <= 9*sizeof(mcsz)) {                                                \
412     INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src);                        \
413     INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest);                       \
414     if(mcsz >= 5*sizeof(mcsz)) {     *mcdst++ = *mcsrc++;                     \
415 				     *mcdst++ = *mcsrc++;                     \
416       if(mcsz >= 7*sizeof(mcsz)) {   *mcdst++ = *mcsrc++;                     \
417 				     *mcdst++ = *mcsrc++;                     \
418 	if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++;                     \
419 				     *mcdst++ = *mcsrc++; }}}                 \
420 				     *mcdst++ = *mcsrc++;                     \
421 				     *mcdst++ = *mcsrc++;                     \
422 				     *mcdst   = *mcsrc  ;                     \
423   } else memcpy(dest, src, mcsz);                                             \
424 } while(0)
425 
426 #else /* !USE_MEMCPY */
427 
428 /* Use Duff's device for good zeroing/copying performance. */
429 
430 #define MALLOC_ZERO(charp, nbytes)                                            \
431 do {                                                                          \
432   INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp);                           \
433   long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn;                         \
434   if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
435   switch (mctmp) {                                                            \
436     case 0: for(;;) { *mzp++ = 0;                                             \
437     case 7:           *mzp++ = 0;                                             \
438     case 6:           *mzp++ = 0;                                             \
439     case 5:           *mzp++ = 0;                                             \
440     case 4:           *mzp++ = 0;                                             \
441     case 3:           *mzp++ = 0;                                             \
442     case 2:           *mzp++ = 0;                                             \
443     case 1:           *mzp++ = 0; if(mcn <= 0) break; mcn--; }                \
444   }                                                                           \
445 } while(0)
446 
447 #define MALLOC_COPY(dest,src,nbytes)                                          \
448 do {                                                                          \
449   INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src;                            \
450   INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest;                           \
451   long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn;                         \
452   if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
453   switch (mctmp) {                                                            \
454     case 0: for(;;) { *mcdst++ = *mcsrc++;                                    \
455     case 7:           *mcdst++ = *mcsrc++;                                    \
456     case 6:           *mcdst++ = *mcsrc++;                                    \
457     case 5:           *mcdst++ = *mcsrc++;                                    \
458     case 4:           *mcdst++ = *mcsrc++;                                    \
459     case 3:           *mcdst++ = *mcsrc++;                                    \
460     case 2:           *mcdst++ = *mcsrc++;                                    \
461     case 1:           *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; }       \
462   }                                                                           \
463 } while(0)
464 
465 #endif
466 
467 
468 /*
469   Define HAVE_MMAP to optionally make malloc() use mmap() to
470   allocate very large blocks.  These will be returned to the
471   operating system immediately after a free().
472 */
473 
474 #ifndef HAVE_MMAP
475 #define HAVE_MMAP 1
476 #endif
477 
478 /*
479   Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
480   large blocks.  This is currently only possible on Linux with
481   kernel versions newer than 1.3.77.
482 */
483 
484 #ifndef HAVE_MREMAP
485 #ifdef INTERNAL_LINUX_C_LIB
486 #define HAVE_MREMAP 1
487 #else
488 #define HAVE_MREMAP 0
489 #endif
490 #endif
491 
492 #if HAVE_MMAP
493 
494 #include <unistd.h>
495 #include <fcntl.h>
496 #include <sys/mman.h>
497 
498 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
499 #define MAP_ANONYMOUS MAP_ANON
500 #endif
501 
502 #endif /* HAVE_MMAP */
503 
504 /*
505   Access to system page size. To the extent possible, this malloc
506   manages memory from the system in page-size units.
507 
508   The following mechanics for getpagesize were adapted from
509   bsd/gnu getpagesize.h
510 */
511 
512 #ifndef LACKS_UNISTD_H
513 #  include <unistd.h>
514 #endif
515 
516 #ifndef malloc_getpagesize
517 #  ifdef _SC_PAGESIZE         /* some SVR4 systems omit an underscore */
518 #    ifndef _SC_PAGE_SIZE
519 #      define _SC_PAGE_SIZE _SC_PAGESIZE
520 #    endif
521 #  endif
522 #  ifdef _SC_PAGE_SIZE
523 #    define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
524 #  else
525 #    if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
526        extern size_t getpagesize();
527 #      define malloc_getpagesize getpagesize()
528 #    else
529 #      ifdef WIN32
530 #        define malloc_getpagesize (4096) /* TBD: Use 'GetSystemInfo' instead */
531 #      else
532 #        ifndef LACKS_SYS_PARAM_H
533 #          include <sys/param.h>
534 #        endif
535 #        ifdef EXEC_PAGESIZE
536 #          define malloc_getpagesize EXEC_PAGESIZE
537 #        else
538 #          ifdef NBPG
539 #            ifndef CLSIZE
540 #              define malloc_getpagesize NBPG
541 #            else
542 #              define malloc_getpagesize (NBPG * CLSIZE)
543 #            endif
544 #          else
545 #            ifdef NBPC
546 #              define malloc_getpagesize NBPC
547 #            else
548 #              ifdef PAGESIZE
549 #                define malloc_getpagesize PAGESIZE
550 #              else
551 #                define malloc_getpagesize (4096) /* just guess */
552 #              endif
553 #            endif
554 #          endif
555 #        endif
556 #      endif
557 #    endif
558 #  endif
559 #endif
560 
561 
562 /*
563 
564   This version of malloc supports the standard SVID/XPG mallinfo
565   routine that returns a struct containing the same kind of
566   information you can get from malloc_stats. It should work on
567   any SVID/XPG compliant system that has a /usr/include/malloc.h
568   defining struct mallinfo. (If you'd like to install such a thing
569   yourself, cut out the preliminary declarations as described above
570   and below and save them in a malloc.h file. But there's no
571   compelling reason to bother to do this.)
572 
573   The main declaration needed is the mallinfo struct that is returned
574   (by-copy) by mallinfo().  The SVID/XPG malloinfo struct contains a
575   bunch of fields, most of which are not even meaningful in this
576   version of malloc. Some of these fields are are instead filled by
577   mallinfo() with other numbers that might possibly be of interest.
578 
579   HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
580   /usr/include/malloc.h file that includes a declaration of struct
581   mallinfo.  If so, it is included; else an SVID2/XPG2 compliant
582   version is declared below.  These must be precisely the same for
583   mallinfo() to work.
584 
585 */
586 
587 /* #define HAVE_USR_INCLUDE_MALLOC_H */
588 
589 #if HAVE_USR_INCLUDE_MALLOC_H
590 #include "/usr/include/malloc.h"
591 #else
592 
593 /* SVID2/XPG mallinfo structure */
594 
595 struct mallinfo {
596   int arena;    /* total space allocated from system */
597   int ordblks;  /* number of non-inuse chunks */
598   int smblks;   /* unused -- always zero */
599   int hblks;    /* number of mmapped regions */
600   int hblkhd;   /* total space in mmapped regions */
601   int usmblks;  /* unused -- always zero */
602   int fsmblks;  /* unused -- always zero */
603   int uordblks; /* total allocated space */
604   int fordblks; /* total non-inuse space */
605   int keepcost; /* top-most, releasable (via malloc_trim) space */
606 };
607 
608 /* SVID2/XPG mallopt options */
609 
610 #define M_MXFAST  1    /* UNUSED in this malloc */
611 #define M_NLBLKS  2    /* UNUSED in this malloc */
612 #define M_GRAIN   3    /* UNUSED in this malloc */
613 #define M_KEEP    4    /* UNUSED in this malloc */
614 
615 #endif
616 
617 /* mallopt options that actually do something */
618 
619 #define M_TRIM_THRESHOLD    -1
620 #define M_TOP_PAD           -2
621 #define M_MMAP_THRESHOLD    -3
622 #define M_MMAP_MAX          -4
623 
624 
625 #ifndef DEFAULT_TRIM_THRESHOLD
626 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
627 #endif
628 
629 /*
630     M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
631       to keep before releasing via malloc_trim in free().
632 
633       Automatic trimming is mainly useful in long-lived programs.
634       Because trimming via sbrk can be slow on some systems, and can
635       sometimes be wasteful (in cases where programs immediately
636       afterward allocate more large chunks) the value should be high
637       enough so that your overall system performance would improve by
638       releasing.
639 
640       The trim threshold and the mmap control parameters (see below)
641       can be traded off with one another. Trimming and mmapping are
642       two different ways of releasing unused memory back to the
643       system. Between these two, it is often possible to keep
644       system-level demands of a long-lived program down to a bare
645       minimum. For example, in one test suite of sessions measuring
646       the XF86 X server on Linux, using a trim threshold of 128K and a
647       mmap threshold of 192K led to near-minimal long term resource
648       consumption.
649 
650       If you are using this malloc in a long-lived program, it should
651       pay to experiment with these values.  As a rough guide, you
652       might set to a value close to the average size of a process
653       (program) running on your system.  Releasing this much memory
654       would allow such a process to run in memory.  Generally, it's
655       worth it to tune for trimming rather tham memory mapping when a
656       program undergoes phases where several large chunks are
657       allocated and released in ways that can reuse each other's
658       storage, perhaps mixed with phases where there are no such
659       chunks at all.  And in well-behaved long-lived programs,
660       controlling release of large blocks via trimming versus mapping
661       is usually faster.
662 
663       However, in most programs, these parameters serve mainly as
664       protection against the system-level effects of carrying around
665       massive amounts of unneeded memory. Since frequent calls to
666       sbrk, mmap, and munmap otherwise degrade performance, the default
667       parameters are set to relatively high values that serve only as
668       safeguards.
669 
670       The default trim value is high enough to cause trimming only in
671       fairly extreme (by current memory consumption standards) cases.
672       It must be greater than page size to have any useful effect.  To
673       disable trimming completely, you can set to (unsigned long)(-1);
674 
675 
676 */
677 
678 
679 #ifndef DEFAULT_TOP_PAD
680 #define DEFAULT_TOP_PAD        (0)
681 #endif
682 
683 /*
684     M_TOP_PAD is the amount of extra `padding' space to allocate or
685       retain whenever sbrk is called. It is used in two ways internally:
686 
687       * When sbrk is called to extend the top of the arena to satisfy
688 	a new malloc request, this much padding is added to the sbrk
689 	request.
690 
691       * When malloc_trim is called automatically from free(),
692 	it is used as the `pad' argument.
693 
694       In both cases, the actual amount of padding is rounded
695       so that the end of the arena is always a system page boundary.
696 
697       The main reason for using padding is to avoid calling sbrk so
698       often. Having even a small pad greatly reduces the likelihood
699       that nearly every malloc request during program start-up (or
700       after trimming) will invoke sbrk, which needlessly wastes
701       time.
702 
703       Automatic rounding-up to page-size units is normally sufficient
704       to avoid measurable overhead, so the default is 0.  However, in
705       systems where sbrk is relatively slow, it can pay to increase
706       this value, at the expense of carrying around more memory than
707       the program needs.
708 
709 */
710 
711 
712 #ifndef DEFAULT_MMAP_THRESHOLD
713 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
714 #endif
715 
716 /*
717 
718     M_MMAP_THRESHOLD is the request size threshold for using mmap()
719       to service a request. Requests of at least this size that cannot
720       be allocated using already-existing space will be serviced via mmap.
721       (If enough normal freed space already exists it is used instead.)
722 
723       Using mmap segregates relatively large chunks of memory so that
724       they can be individually obtained and released from the host
725       system. A request serviced through mmap is never reused by any
726       other request (at least not directly; the system may just so
727       happen to remap successive requests to the same locations).
728 
729       Segregating space in this way has the benefit that mmapped space
730       can ALWAYS be individually released back to the system, which
731       helps keep the system level memory demands of a long-lived
732       program low. Mapped memory can never become `locked' between
733       other chunks, as can happen with normally allocated chunks, which
734       menas that even trimming via malloc_trim would not release them.
735 
736       However, it has the disadvantages that:
737 
738 	 1. The space cannot be reclaimed, consolidated, and then
739 	    used to service later requests, as happens with normal chunks.
740 	 2. It can lead to more wastage because of mmap page alignment
741 	    requirements
742 	 3. It causes malloc performance to be more dependent on host
743 	    system memory management support routines which may vary in
744 	    implementation quality and may impose arbitrary
745 	    limitations. Generally, servicing a request via normal
746 	    malloc steps is faster than going through a system's mmap.
747 
748       All together, these considerations should lead you to use mmap
749       only for relatively large requests.
750 
751 
752 */
753 
754 
755 #ifndef DEFAULT_MMAP_MAX
756 #if HAVE_MMAP
757 #define DEFAULT_MMAP_MAX       (64)
758 #else
759 #define DEFAULT_MMAP_MAX       (0)
760 #endif
761 #endif
762 
763 /*
764     M_MMAP_MAX is the maximum number of requests to simultaneously
765       service using mmap. This parameter exists because:
766 
767 	 1. Some systems have a limited number of internal tables for
768 	    use by mmap.
769 	 2. In most systems, overreliance on mmap can degrade overall
770 	    performance.
771 	 3. If a program allocates many large regions, it is probably
772 	    better off using normal sbrk-based allocation routines that
773 	    can reclaim and reallocate normal heap memory. Using a
774 	    small value allows transition into this mode after the
775 	    first few allocations.
776 
777       Setting to 0 disables all use of mmap.  If HAVE_MMAP is not set,
778       the default value is 0, and attempts to set it to non-zero values
779       in mallopt will fail.
780 */
781 
782 
783 /*
784     USE_DL_PREFIX will prefix all public routines with the string 'dl'.
785       Useful to quickly avoid procedure declaration conflicts and linker
786       symbol conflicts with existing memory allocation routines.
787 
788 */
789 
790 /* #define USE_DL_PREFIX */
791 
792 
793 /*
794 
795   Special defines for linux libc
796 
797   Except when compiled using these special defines for Linux libc
798   using weak aliases, this malloc is NOT designed to work in
799   multithreaded applications.  No semaphores or other concurrency
800   control are provided to ensure that multiple malloc or free calls
801   don't run at the same time, which could be disasterous. A single
802   semaphore could be used across malloc, realloc, and free (which is
803   essentially the effect of the linux weak alias approach). It would
804   be hard to obtain finer granularity.
805 
806 */
807 
808 
809 #ifdef INTERNAL_LINUX_C_LIB
810 
811 #if __STD_C
812 
813 Void_t * __default_morecore_init (ptrdiff_t);
814 Void_t *(*__morecore)(ptrdiff_t) = __default_morecore_init;
815 
816 #else
817 
818 Void_t * __default_morecore_init ();
819 Void_t *(*__morecore)() = __default_morecore_init;
820 
821 #endif
822 
823 #define MORECORE (*__morecore)
824 #define MORECORE_FAILURE 0
825 #define MORECORE_CLEARS 1
826 
827 #else /* INTERNAL_LINUX_C_LIB */
828 
829 #if __STD_C
830 extern Void_t*     sbrk(ptrdiff_t);
831 #else
832 extern Void_t*     sbrk();
833 #endif
834 
835 #ifndef MORECORE
836 #define MORECORE sbrk
837 #endif
838 
839 #ifndef MORECORE_FAILURE
840 #define MORECORE_FAILURE -1
841 #endif
842 
843 #ifndef MORECORE_CLEARS
844 #define MORECORE_CLEARS 1
845 #endif
846 
847 #endif /* INTERNAL_LINUX_C_LIB */
848 
849 #if defined(INTERNAL_LINUX_C_LIB) && defined(__ELF__)
850 
851 #define cALLOc		__libc_calloc
852 #define fREe		__libc_free
853 #define mALLOc		__libc_malloc
854 #define mEMALIGn	__libc_memalign
855 #define rEALLOc		__libc_realloc
856 #define vALLOc		__libc_valloc
857 #define pvALLOc		__libc_pvalloc
858 #define mALLINFo	__libc_mallinfo
859 #define mALLOPt		__libc_mallopt
860 
861 #pragma weak calloc = __libc_calloc
862 #pragma weak free = __libc_free
863 #pragma weak cfree = __libc_free
864 #pragma weak malloc = __libc_malloc
865 #pragma weak memalign = __libc_memalign
866 #pragma weak realloc = __libc_realloc
867 #pragma weak valloc = __libc_valloc
868 #pragma weak pvalloc = __libc_pvalloc
869 #pragma weak mallinfo = __libc_mallinfo
870 #pragma weak mallopt = __libc_mallopt
871 
872 #else
873 
874 #ifdef USE_DL_PREFIX
875 #define cALLOc		dlcalloc
876 #define fREe		dlfree
877 #define mALLOc		dlmalloc
878 #define mEMALIGn	dlmemalign
879 #define rEALLOc		dlrealloc
880 #define vALLOc		dlvalloc
881 #define pvALLOc		dlpvalloc
882 #define mALLINFo	dlmallinfo
883 #define mALLOPt		dlmallopt
884 #else /* USE_DL_PREFIX */
885 #define cALLOc		calloc
886 #define fREe		free
887 #define mALLOc		malloc
888 #define mEMALIGn	memalign
889 #define rEALLOc		realloc
890 #define vALLOc		valloc
891 #define pvALLOc		pvalloc
892 #define mALLINFo	mallinfo
893 #define mALLOPt		mallopt
894 #endif /* USE_DL_PREFIX */
895 
896 #endif
897 
898 /* Public routines */
899 
900 #if __STD_C
901 
902 Void_t* mALLOc(size_t);
903 void    fREe(Void_t*);
904 Void_t* rEALLOc(Void_t*, size_t);
905 Void_t* mEMALIGn(size_t, size_t);
906 Void_t* vALLOc(size_t);
907 Void_t* pvALLOc(size_t);
908 Void_t* cALLOc(size_t, size_t);
909 void    cfree(Void_t*);
910 int     malloc_trim(size_t);
911 size_t  malloc_usable_size(Void_t*);
912 void    malloc_stats();
913 int     mALLOPt(int, int);
914 struct mallinfo mALLINFo(void);
915 #else
916 Void_t* mALLOc();
917 void    fREe();
918 Void_t* rEALLOc();
919 Void_t* mEMALIGn();
920 Void_t* vALLOc();
921 Void_t* pvALLOc();
922 Void_t* cALLOc();
923 void    cfree();
924 int     malloc_trim();
925 size_t  malloc_usable_size();
926 void    malloc_stats();
927 int     mALLOPt();
928 struct mallinfo mALLINFo();
929 #endif
930 
931 
932 #ifdef __cplusplus
933 };  /* end of extern "C" */
934 #endif
935 
936 /* ---------- To make a malloc.h, end cutting here ------------ */
937 #endif	/* 0 */			/* Moved to malloc.h */
938 
939 #include <malloc.h>
940 #ifdef DEBUG
941 #if __STD_C
942 static void malloc_update_mallinfo (void);
943 void malloc_stats (void);
944 #else
945 static void malloc_update_mallinfo ();
946 void malloc_stats();
947 #endif
948 #endif	/* DEBUG */
949 
950 DECLARE_GLOBAL_DATA_PTR;
951 
952 /*
953   Emulation of sbrk for WIN32
954   All code within the ifdef WIN32 is untested by me.
955 
956   Thanks to Martin Fong and others for supplying this.
957 */
958 
959 
960 #ifdef WIN32
961 
962 #define AlignPage(add) (((add) + (malloc_getpagesize-1)) & \
963 ~(malloc_getpagesize-1))
964 #define AlignPage64K(add) (((add) + (0x10000 - 1)) & ~(0x10000 - 1))
965 
966 /* resrve 64MB to insure large contiguous space */
967 #define RESERVED_SIZE (1024*1024*64)
968 #define NEXT_SIZE (2048*1024)
969 #define TOP_MEMORY ((unsigned long)2*1024*1024*1024)
970 
971 struct GmListElement;
972 typedef struct GmListElement GmListElement;
973 
974 struct GmListElement
975 {
976 	GmListElement* next;
977 	void* base;
978 };
979 
980 static GmListElement* head = 0;
981 static unsigned int gNextAddress = 0;
982 static unsigned int gAddressBase = 0;
983 static unsigned int gAllocatedSize = 0;
984 
985 static
986 GmListElement* makeGmListElement (void* bas)
987 {
988 	GmListElement* this;
989 	this = (GmListElement*)(void*)LocalAlloc (0, sizeof (GmListElement));
990 	assert (this);
991 	if (this)
992 	{
993 		this->base = bas;
994 		this->next = head;
995 		head = this;
996 	}
997 	return this;
998 }
999 
1000 void gcleanup ()
1001 {
1002 	BOOL rval;
1003 	assert ( (head == NULL) || (head->base == (void*)gAddressBase));
1004 	if (gAddressBase && (gNextAddress - gAddressBase))
1005 	{
1006 		rval = VirtualFree ((void*)gAddressBase,
1007 							gNextAddress - gAddressBase,
1008 							MEM_DECOMMIT);
1009 	assert (rval);
1010 	}
1011 	while (head)
1012 	{
1013 		GmListElement* next = head->next;
1014 		rval = VirtualFree (head->base, 0, MEM_RELEASE);
1015 		assert (rval);
1016 		LocalFree (head);
1017 		head = next;
1018 	}
1019 }
1020 
1021 static
1022 void* findRegion (void* start_address, unsigned long size)
1023 {
1024 	MEMORY_BASIC_INFORMATION info;
1025 	if (size >= TOP_MEMORY) return NULL;
1026 
1027 	while ((unsigned long)start_address + size < TOP_MEMORY)
1028 	{
1029 		VirtualQuery (start_address, &info, sizeof (info));
1030 		if ((info.State == MEM_FREE) && (info.RegionSize >= size))
1031 			return start_address;
1032 		else
1033 		{
1034 			/* Requested region is not available so see if the */
1035 			/* next region is available.  Set 'start_address' */
1036 			/* to the next region and call 'VirtualQuery()' */
1037 			/* again. */
1038 
1039 			start_address = (char*)info.BaseAddress + info.RegionSize;
1040 
1041 			/* Make sure we start looking for the next region */
1042 			/* on the *next* 64K boundary.  Otherwise, even if */
1043 			/* the new region is free according to */
1044 			/* 'VirtualQuery()', the subsequent call to */
1045 			/* 'VirtualAlloc()' (which follows the call to */
1046 			/* this routine in 'wsbrk()') will round *down* */
1047 			/* the requested address to a 64K boundary which */
1048 			/* we already know is an address in the */
1049 			/* unavailable region.  Thus, the subsequent call */
1050 			/* to 'VirtualAlloc()' will fail and bring us back */
1051 			/* here, causing us to go into an infinite loop. */
1052 
1053 			start_address =
1054 				(void *) AlignPage64K((unsigned long) start_address);
1055 		}
1056 	}
1057 	return NULL;
1058 
1059 }
1060 
1061 
1062 void* wsbrk (long size)
1063 {
1064 	void* tmp;
1065 	if (size > 0)
1066 	{
1067 		if (gAddressBase == 0)
1068 		{
1069 			gAllocatedSize = max (RESERVED_SIZE, AlignPage (size));
1070 			gNextAddress = gAddressBase =
1071 				(unsigned int)VirtualAlloc (NULL, gAllocatedSize,
1072 											MEM_RESERVE, PAGE_NOACCESS);
1073 		} else if (AlignPage (gNextAddress + size) > (gAddressBase +
1074 gAllocatedSize))
1075 		{
1076 			long new_size = max (NEXT_SIZE, AlignPage (size));
1077 			void* new_address = (void*)(gAddressBase+gAllocatedSize);
1078 			do
1079 			{
1080 				new_address = findRegion (new_address, new_size);
1081 
1082 				if (new_address == 0)
1083 					return (void*)-1;
1084 
1085 				gAddressBase = gNextAddress =
1086 					(unsigned int)VirtualAlloc (new_address, new_size,
1087 												MEM_RESERVE, PAGE_NOACCESS);
1088 				/* repeat in case of race condition */
1089 				/* The region that we found has been snagged */
1090 				/* by another thread */
1091 			}
1092 			while (gAddressBase == 0);
1093 
1094 			assert (new_address == (void*)gAddressBase);
1095 
1096 			gAllocatedSize = new_size;
1097 
1098 			if (!makeGmListElement ((void*)gAddressBase))
1099 				return (void*)-1;
1100 		}
1101 		if ((size + gNextAddress) > AlignPage (gNextAddress))
1102 		{
1103 			void* res;
1104 			res = VirtualAlloc ((void*)AlignPage (gNextAddress),
1105 								(size + gNextAddress -
1106 								 AlignPage (gNextAddress)),
1107 								MEM_COMMIT, PAGE_READWRITE);
1108 			if (res == 0)
1109 				return (void*)-1;
1110 		}
1111 		tmp = (void*)gNextAddress;
1112 		gNextAddress = (unsigned int)tmp + size;
1113 		return tmp;
1114 	}
1115 	else if (size < 0)
1116 	{
1117 		unsigned int alignedGoal = AlignPage (gNextAddress + size);
1118 		/* Trim by releasing the virtual memory */
1119 		if (alignedGoal >= gAddressBase)
1120 		{
1121 			VirtualFree ((void*)alignedGoal, gNextAddress - alignedGoal,
1122 						 MEM_DECOMMIT);
1123 			gNextAddress = gNextAddress + size;
1124 			return (void*)gNextAddress;
1125 		}
1126 		else
1127 		{
1128 			VirtualFree ((void*)gAddressBase, gNextAddress - gAddressBase,
1129 						 MEM_DECOMMIT);
1130 			gNextAddress = gAddressBase;
1131 			return (void*)-1;
1132 		}
1133 	}
1134 	else
1135 	{
1136 		return (void*)gNextAddress;
1137 	}
1138 }
1139 
1140 #endif
1141 
1142 
1143 
1144 /*
1145   Type declarations
1146 */
1147 
1148 
1149 struct malloc_chunk
1150 {
1151   INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
1152   INTERNAL_SIZE_T size;      /* Size in bytes, including overhead. */
1153   struct malloc_chunk* fd;   /* double links -- used only if free. */
1154   struct malloc_chunk* bk;
1155 };
1156 
1157 typedef struct malloc_chunk* mchunkptr;
1158 
1159 /*
1160 
1161    malloc_chunk details:
1162 
1163     (The following includes lightly edited explanations by Colin Plumb.)
1164 
1165     Chunks of memory are maintained using a `boundary tag' method as
1166     described in e.g., Knuth or Standish.  (See the paper by Paul
1167     Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1168     survey of such techniques.)  Sizes of free chunks are stored both
1169     in the front of each chunk and at the end.  This makes
1170     consolidating fragmented chunks into bigger chunks very fast.  The
1171     size fields also hold bits representing whether chunks are free or
1172     in use.
1173 
1174     An allocated chunk looks like this:
1175 
1176 
1177     chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1178 	    |             Size of previous chunk, if allocated            | |
1179 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1180 	    |             Size of chunk, in bytes                         |P|
1181       mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1182 	    |             User data starts here...                          .
1183 	    .                                                               .
1184 	    .             (malloc_usable_space() bytes)                     .
1185 	    .                                                               |
1186 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1187 	    |             Size of chunk                                     |
1188 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1189 
1190 
1191     Where "chunk" is the front of the chunk for the purpose of most of
1192     the malloc code, but "mem" is the pointer that is returned to the
1193     user.  "Nextchunk" is the beginning of the next contiguous chunk.
1194 
1195     Chunks always begin on even word boundries, so the mem portion
1196     (which is returned to the user) is also on an even word boundary, and
1197     thus double-word aligned.
1198 
1199     Free chunks are stored in circular doubly-linked lists, and look like this:
1200 
1201     chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1202 	    |             Size of previous chunk                            |
1203 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1204     `head:' |             Size of chunk, in bytes                         |P|
1205       mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1206 	    |             Forward pointer to next chunk in list             |
1207 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1208 	    |             Back pointer to previous chunk in list            |
1209 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1210 	    |             Unused space (may be 0 bytes long)                .
1211 	    .                                                               .
1212 	    .                                                               |
1213 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1214     `foot:' |             Size of chunk, in bytes                           |
1215 	    +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1216 
1217     The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1218     chunk size (which is always a multiple of two words), is an in-use
1219     bit for the *previous* chunk.  If that bit is *clear*, then the
1220     word before the current chunk size contains the previous chunk
1221     size, and can be used to find the front of the previous chunk.
1222     (The very first chunk allocated always has this bit set,
1223     preventing access to non-existent (or non-owned) memory.)
1224 
1225     Note that the `foot' of the current chunk is actually represented
1226     as the prev_size of the NEXT chunk. (This makes it easier to
1227     deal with alignments etc).
1228 
1229     The two exceptions to all this are
1230 
1231      1. The special chunk `top', which doesn't bother using the
1232 	trailing size field since there is no
1233 	next contiguous chunk that would have to index off it. (After
1234 	initialization, `top' is forced to always exist.  If it would
1235 	become less than MINSIZE bytes long, it is replenished via
1236 	malloc_extend_top.)
1237 
1238      2. Chunks allocated via mmap, which have the second-lowest-order
1239 	bit (IS_MMAPPED) set in their size fields.  Because they are
1240 	never merged or traversed from any other chunk, they have no
1241 	foot size or inuse information.
1242 
1243     Available chunks are kept in any of several places (all declared below):
1244 
1245     * `av': An array of chunks serving as bin headers for consolidated
1246        chunks. Each bin is doubly linked.  The bins are approximately
1247        proportionally (log) spaced.  There are a lot of these bins
1248        (128). This may look excessive, but works very well in
1249        practice.  All procedures maintain the invariant that no
1250        consolidated chunk physically borders another one. Chunks in
1251        bins are kept in size order, with ties going to the
1252        approximately least recently used chunk.
1253 
1254        The chunks in each bin are maintained in decreasing sorted order by
1255        size.  This is irrelevant for the small bins, which all contain
1256        the same-sized chunks, but facilitates best-fit allocation for
1257        larger chunks. (These lists are just sequential. Keeping them in
1258        order almost never requires enough traversal to warrant using
1259        fancier ordered data structures.)  Chunks of the same size are
1260        linked with the most recently freed at the front, and allocations
1261        are taken from the back.  This results in LRU or FIFO allocation
1262        order, which tends to give each chunk an equal opportunity to be
1263        consolidated with adjacent freed chunks, resulting in larger free
1264        chunks and less fragmentation.
1265 
1266     * `top': The top-most available chunk (i.e., the one bordering the
1267        end of available memory) is treated specially. It is never
1268        included in any bin, is used only if no other chunk is
1269        available, and is released back to the system if it is very
1270        large (see M_TRIM_THRESHOLD).
1271 
1272     * `last_remainder': A bin holding only the remainder of the
1273        most recently split (non-top) chunk. This bin is checked
1274        before other non-fitting chunks, so as to provide better
1275        locality for runs of sequentially allocated chunks.
1276 
1277     *  Implicitly, through the host system's memory mapping tables.
1278        If supported, requests greater than a threshold are usually
1279        serviced via calls to mmap, and then later released via munmap.
1280 
1281 */
1282 
1283 /*  sizes, alignments */
1284 
1285 #define SIZE_SZ                (sizeof(INTERNAL_SIZE_T))
1286 #define MALLOC_ALIGNMENT       (SIZE_SZ + SIZE_SZ)
1287 #define MALLOC_ALIGN_MASK      (MALLOC_ALIGNMENT - 1)
1288 #define MINSIZE                (sizeof(struct malloc_chunk))
1289 
1290 /* conversion from malloc headers to user pointers, and back */
1291 
1292 #define chunk2mem(p)   ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1293 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1294 
1295 /* pad request bytes into a usable size */
1296 
1297 #define request2size(req) \
1298  (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1299   (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1300    (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1301 
1302 /* Check if m has acceptable alignment */
1303 
1304 #define aligned_OK(m)    (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1305 
1306 
1307 
1308 
1309 /*
1310   Physical chunk operations
1311 */
1312 
1313 
1314 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1315 
1316 #define PREV_INUSE 0x1
1317 
1318 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1319 
1320 #define IS_MMAPPED 0x2
1321 
1322 /* Bits to mask off when extracting size */
1323 
1324 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1325 
1326 
1327 /* Ptr to next physical malloc_chunk. */
1328 
1329 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1330 
1331 /* Ptr to previous physical malloc_chunk */
1332 
1333 #define prev_chunk(p)\
1334    ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1335 
1336 
1337 /* Treat space at ptr + offset as a chunk */
1338 
1339 #define chunk_at_offset(p, s)  ((mchunkptr)(((char*)(p)) + (s)))
1340 
1341 
1342 
1343 
1344 /*
1345   Dealing with use bits
1346 */
1347 
1348 /* extract p's inuse bit */
1349 
1350 #define inuse(p)\
1351 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1352 
1353 /* extract inuse bit of previous chunk */
1354 
1355 #define prev_inuse(p)  ((p)->size & PREV_INUSE)
1356 
1357 /* check for mmap()'ed chunk */
1358 
1359 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1360 
1361 /* set/clear chunk as in use without otherwise disturbing */
1362 
1363 #define set_inuse(p)\
1364 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1365 
1366 #define clear_inuse(p)\
1367 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1368 
1369 /* check/set/clear inuse bits in known places */
1370 
1371 #define inuse_bit_at_offset(p, s)\
1372  (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1373 
1374 #define set_inuse_bit_at_offset(p, s)\
1375  (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1376 
1377 #define clear_inuse_bit_at_offset(p, s)\
1378  (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1379 
1380 
1381 
1382 
1383 /*
1384   Dealing with size fields
1385 */
1386 
1387 /* Get size, ignoring use bits */
1388 
1389 #define chunksize(p)          ((p)->size & ~(SIZE_BITS))
1390 
1391 /* Set size at head, without disturbing its use bit */
1392 
1393 #define set_head_size(p, s)   ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1394 
1395 /* Set size/use ignoring previous bits in header */
1396 
1397 #define set_head(p, s)        ((p)->size = (s))
1398 
1399 /* Set size at footer (only when chunk is not in use) */
1400 
1401 #define set_foot(p, s)   (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1402 
1403 
1404 
1405 
1406 
1407 /*
1408    Bins
1409 
1410     The bins, `av_' are an array of pairs of pointers serving as the
1411     heads of (initially empty) doubly-linked lists of chunks, laid out
1412     in a way so that each pair can be treated as if it were in a
1413     malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1414     and chunks are the same).
1415 
1416     Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1417     8 bytes apart. Larger bins are approximately logarithmically
1418     spaced. (See the table below.) The `av_' array is never mentioned
1419     directly in the code, but instead via bin access macros.
1420 
1421     Bin layout:
1422 
1423     64 bins of size       8
1424     32 bins of size      64
1425     16 bins of size     512
1426      8 bins of size    4096
1427      4 bins of size   32768
1428      2 bins of size  262144
1429      1 bin  of size what's left
1430 
1431     There is actually a little bit of slop in the numbers in bin_index
1432     for the sake of speed. This makes no difference elsewhere.
1433 
1434     The special chunks `top' and `last_remainder' get their own bins,
1435     (this is implemented via yet more trickery with the av_ array),
1436     although `top' is never properly linked to its bin since it is
1437     always handled specially.
1438 
1439 */
1440 
1441 #define NAV             128   /* number of bins */
1442 
1443 typedef struct malloc_chunk* mbinptr;
1444 
1445 /* access macros */
1446 
1447 #define bin_at(i)      ((mbinptr)((char*)&(av_[2*(i) + 2]) - 2*SIZE_SZ))
1448 #define next_bin(b)    ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1449 #define prev_bin(b)    ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1450 
1451 /*
1452    The first 2 bins are never indexed. The corresponding av_ cells are instead
1453    used for bookkeeping. This is not to save space, but to simplify
1454    indexing, maintain locality, and avoid some initialization tests.
1455 */
1456 
1457 #define top            (av_[2])          /* The topmost chunk */
1458 #define last_remainder (bin_at(1))       /* remainder from last split */
1459 
1460 
1461 /*
1462    Because top initially points to its own bin with initial
1463    zero size, thus forcing extension on the first malloc request,
1464    we avoid having any special code in malloc to check whether
1465    it even exists yet. But we still need to in malloc_extend_top.
1466 */
1467 
1468 #define initial_top    ((mchunkptr)(bin_at(0)))
1469 
1470 /* Helper macro to initialize bins */
1471 
1472 #define IAV(i)  bin_at(i), bin_at(i)
1473 
1474 static mbinptr av_[NAV * 2 + 2] = {
1475  0, 0,
1476  IAV(0),   IAV(1),   IAV(2),   IAV(3),   IAV(4),   IAV(5),   IAV(6),   IAV(7),
1477  IAV(8),   IAV(9),   IAV(10),  IAV(11),  IAV(12),  IAV(13),  IAV(14),  IAV(15),
1478  IAV(16),  IAV(17),  IAV(18),  IAV(19),  IAV(20),  IAV(21),  IAV(22),  IAV(23),
1479  IAV(24),  IAV(25),  IAV(26),  IAV(27),  IAV(28),  IAV(29),  IAV(30),  IAV(31),
1480  IAV(32),  IAV(33),  IAV(34),  IAV(35),  IAV(36),  IAV(37),  IAV(38),  IAV(39),
1481  IAV(40),  IAV(41),  IAV(42),  IAV(43),  IAV(44),  IAV(45),  IAV(46),  IAV(47),
1482  IAV(48),  IAV(49),  IAV(50),  IAV(51),  IAV(52),  IAV(53),  IAV(54),  IAV(55),
1483  IAV(56),  IAV(57),  IAV(58),  IAV(59),  IAV(60),  IAV(61),  IAV(62),  IAV(63),
1484  IAV(64),  IAV(65),  IAV(66),  IAV(67),  IAV(68),  IAV(69),  IAV(70),  IAV(71),
1485  IAV(72),  IAV(73),  IAV(74),  IAV(75),  IAV(76),  IAV(77),  IAV(78),  IAV(79),
1486  IAV(80),  IAV(81),  IAV(82),  IAV(83),  IAV(84),  IAV(85),  IAV(86),  IAV(87),
1487  IAV(88),  IAV(89),  IAV(90),  IAV(91),  IAV(92),  IAV(93),  IAV(94),  IAV(95),
1488  IAV(96),  IAV(97),  IAV(98),  IAV(99),  IAV(100), IAV(101), IAV(102), IAV(103),
1489  IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1490  IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1491  IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1492 };
1493 
1494 #ifndef CONFIG_RELOC_FIXUP_WORKS
1495 void malloc_bin_reloc (void)
1496 {
1497 	unsigned long *p = (unsigned long *)(&av_[2]);
1498 	int i;
1499 	for (i=2; i<(sizeof(av_)/sizeof(mbinptr)); ++i) {
1500 		*p++ += gd->reloc_off;
1501 	}
1502 }
1503 #endif
1504 
1505 ulong mem_malloc_start = 0;
1506 ulong mem_malloc_end = 0;
1507 ulong mem_malloc_brk = 0;
1508 
1509 void *sbrk(ptrdiff_t increment)
1510 {
1511 	ulong old = mem_malloc_brk;
1512 	ulong new = old + increment;
1513 
1514 	if ((new < mem_malloc_start) || (new > mem_malloc_end))
1515 		return (void *)MORECORE_FAILURE;
1516 
1517 	mem_malloc_brk = new;
1518 
1519 	return (void *)old;
1520 }
1521 
1522 void mem_malloc_init(ulong start, ulong size)
1523 {
1524 	mem_malloc_start = start;
1525 	mem_malloc_end = start + size;
1526 	mem_malloc_brk = start;
1527 
1528 	memset((void *)mem_malloc_start, 0, size);
1529 }
1530 
1531 /* field-extraction macros */
1532 
1533 #define first(b) ((b)->fd)
1534 #define last(b)  ((b)->bk)
1535 
1536 /*
1537   Indexing into bins
1538 */
1539 
1540 #define bin_index(sz)                                                          \
1541 (((((unsigned long)(sz)) >> 9) ==    0) ?       (((unsigned long)(sz)) >>  3): \
1542  ((((unsigned long)(sz)) >> 9) <=    4) ?  56 + (((unsigned long)(sz)) >>  6): \
1543  ((((unsigned long)(sz)) >> 9) <=   20) ?  91 + (((unsigned long)(sz)) >>  9): \
1544  ((((unsigned long)(sz)) >> 9) <=   84) ? 110 + (((unsigned long)(sz)) >> 12): \
1545  ((((unsigned long)(sz)) >> 9) <=  340) ? 119 + (((unsigned long)(sz)) >> 15): \
1546  ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1547 					  126)
1548 /*
1549   bins for chunks < 512 are all spaced 8 bytes apart, and hold
1550   identically sized chunks. This is exploited in malloc.
1551 */
1552 
1553 #define MAX_SMALLBIN         63
1554 #define MAX_SMALLBIN_SIZE   512
1555 #define SMALLBIN_WIDTH        8
1556 
1557 #define smallbin_index(sz)  (((unsigned long)(sz)) >> 3)
1558 
1559 /*
1560    Requests are `small' if both the corresponding and the next bin are small
1561 */
1562 
1563 #define is_small_request(nb) (nb < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1564 
1565 
1566 
1567 /*
1568     To help compensate for the large number of bins, a one-level index
1569     structure is used for bin-by-bin searching.  `binblocks' is a
1570     one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1571     have any (possibly) non-empty bins, so they can be skipped over
1572     all at once during during traversals. The bits are NOT always
1573     cleared as soon as all bins in a block are empty, but instead only
1574     when all are noticed to be empty during traversal in malloc.
1575 */
1576 
1577 #define BINBLOCKWIDTH     4   /* bins per block */
1578 
1579 #define binblocks_r     ((INTERNAL_SIZE_T)av_[1]) /* bitvector of nonempty blocks */
1580 #define binblocks_w     (av_[1])
1581 
1582 /* bin<->block macros */
1583 
1584 #define idx2binblock(ix)    ((unsigned)1 << (ix / BINBLOCKWIDTH))
1585 #define mark_binblock(ii)   (binblocks_w = (mbinptr)(binblocks_r | idx2binblock(ii)))
1586 #define clear_binblock(ii)  (binblocks_w = (mbinptr)(binblocks_r & ~(idx2binblock(ii))))
1587 
1588 
1589 
1590 
1591 
1592 /*  Other static bookkeeping data */
1593 
1594 /* variables holding tunable values */
1595 
1596 static unsigned long trim_threshold   = DEFAULT_TRIM_THRESHOLD;
1597 static unsigned long top_pad          = DEFAULT_TOP_PAD;
1598 static unsigned int  n_mmaps_max      = DEFAULT_MMAP_MAX;
1599 static unsigned long mmap_threshold   = DEFAULT_MMAP_THRESHOLD;
1600 
1601 /* The first value returned from sbrk */
1602 static char* sbrk_base = (char*)(-1);
1603 
1604 /* The maximum memory obtained from system via sbrk */
1605 static unsigned long max_sbrked_mem = 0;
1606 
1607 /* The maximum via either sbrk or mmap */
1608 static unsigned long max_total_mem = 0;
1609 
1610 /* internal working copy of mallinfo */
1611 static struct mallinfo current_mallinfo = {  0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
1612 
1613 /* The total memory obtained from system via sbrk */
1614 #define sbrked_mem  (current_mallinfo.arena)
1615 
1616 /* Tracking mmaps */
1617 
1618 #ifdef DEBUG
1619 static unsigned int n_mmaps = 0;
1620 #endif	/* DEBUG */
1621 static unsigned long mmapped_mem = 0;
1622 #if HAVE_MMAP
1623 static unsigned int max_n_mmaps = 0;
1624 static unsigned long max_mmapped_mem = 0;
1625 #endif
1626 
1627 
1628 
1629 /*
1630   Debugging support
1631 */
1632 
1633 #ifdef DEBUG
1634 
1635 
1636 /*
1637   These routines make a number of assertions about the states
1638   of data structures that should be true at all times. If any
1639   are not true, it's very likely that a user program has somehow
1640   trashed memory. (It's also possible that there is a coding error
1641   in malloc. In which case, please report it!)
1642 */
1643 
1644 #if __STD_C
1645 static void do_check_chunk(mchunkptr p)
1646 #else
1647 static void do_check_chunk(p) mchunkptr p;
1648 #endif
1649 {
1650 #if 0	/* causes warnings because assert() is off */
1651   INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1652 #endif	/* 0 */
1653 
1654   /* No checkable chunk is mmapped */
1655   assert(!chunk_is_mmapped(p));
1656 
1657   /* Check for legal address ... */
1658   assert((char*)p >= sbrk_base);
1659   if (p != top)
1660     assert((char*)p + sz <= (char*)top);
1661   else
1662     assert((char*)p + sz <= sbrk_base + sbrked_mem);
1663 
1664 }
1665 
1666 
1667 #if __STD_C
1668 static void do_check_free_chunk(mchunkptr p)
1669 #else
1670 static void do_check_free_chunk(p) mchunkptr p;
1671 #endif
1672 {
1673   INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1674 #if 0	/* causes warnings because assert() is off */
1675   mchunkptr next = chunk_at_offset(p, sz);
1676 #endif	/* 0 */
1677 
1678   do_check_chunk(p);
1679 
1680   /* Check whether it claims to be free ... */
1681   assert(!inuse(p));
1682 
1683   /* Unless a special marker, must have OK fields */
1684   if ((long)sz >= (long)MINSIZE)
1685   {
1686     assert((sz & MALLOC_ALIGN_MASK) == 0);
1687     assert(aligned_OK(chunk2mem(p)));
1688     /* ... matching footer field */
1689     assert(next->prev_size == sz);
1690     /* ... and is fully consolidated */
1691     assert(prev_inuse(p));
1692     assert (next == top || inuse(next));
1693 
1694     /* ... and has minimally sane links */
1695     assert(p->fd->bk == p);
1696     assert(p->bk->fd == p);
1697   }
1698   else /* markers are always of size SIZE_SZ */
1699     assert(sz == SIZE_SZ);
1700 }
1701 
1702 #if __STD_C
1703 static void do_check_inuse_chunk(mchunkptr p)
1704 #else
1705 static void do_check_inuse_chunk(p) mchunkptr p;
1706 #endif
1707 {
1708   mchunkptr next = next_chunk(p);
1709   do_check_chunk(p);
1710 
1711   /* Check whether it claims to be in use ... */
1712   assert(inuse(p));
1713 
1714   /* ... and is surrounded by OK chunks.
1715     Since more things can be checked with free chunks than inuse ones,
1716     if an inuse chunk borders them and debug is on, it's worth doing them.
1717   */
1718   if (!prev_inuse(p))
1719   {
1720     mchunkptr prv = prev_chunk(p);
1721     assert(next_chunk(prv) == p);
1722     do_check_free_chunk(prv);
1723   }
1724   if (next == top)
1725   {
1726     assert(prev_inuse(next));
1727     assert(chunksize(next) >= MINSIZE);
1728   }
1729   else if (!inuse(next))
1730     do_check_free_chunk(next);
1731 
1732 }
1733 
1734 #if __STD_C
1735 static void do_check_malloced_chunk(mchunkptr p, INTERNAL_SIZE_T s)
1736 #else
1737 static void do_check_malloced_chunk(p, s) mchunkptr p; INTERNAL_SIZE_T s;
1738 #endif
1739 {
1740 #if 0	/* causes warnings because assert() is off */
1741   INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1742   long room = sz - s;
1743 #endif	/* 0 */
1744 
1745   do_check_inuse_chunk(p);
1746 
1747   /* Legal size ... */
1748   assert((long)sz >= (long)MINSIZE);
1749   assert((sz & MALLOC_ALIGN_MASK) == 0);
1750   assert(room >= 0);
1751   assert(room < (long)MINSIZE);
1752 
1753   /* ... and alignment */
1754   assert(aligned_OK(chunk2mem(p)));
1755 
1756 
1757   /* ... and was allocated at front of an available chunk */
1758   assert(prev_inuse(p));
1759 
1760 }
1761 
1762 
1763 #define check_free_chunk(P)  do_check_free_chunk(P)
1764 #define check_inuse_chunk(P) do_check_inuse_chunk(P)
1765 #define check_chunk(P) do_check_chunk(P)
1766 #define check_malloced_chunk(P,N) do_check_malloced_chunk(P,N)
1767 #else
1768 #define check_free_chunk(P)
1769 #define check_inuse_chunk(P)
1770 #define check_chunk(P)
1771 #define check_malloced_chunk(P,N)
1772 #endif
1773 
1774 
1775 
1776 /*
1777   Macro-based internal utilities
1778 */
1779 
1780 
1781 /*
1782   Linking chunks in bin lists.
1783   Call these only with variables, not arbitrary expressions, as arguments.
1784 */
1785 
1786 /*
1787   Place chunk p of size s in its bin, in size order,
1788   putting it ahead of others of same size.
1789 */
1790 
1791 
1792 #define frontlink(P, S, IDX, BK, FD)                                          \
1793 {                                                                             \
1794   if (S < MAX_SMALLBIN_SIZE)                                                  \
1795   {                                                                           \
1796     IDX = smallbin_index(S);                                                  \
1797     mark_binblock(IDX);                                                       \
1798     BK = bin_at(IDX);                                                         \
1799     FD = BK->fd;                                                              \
1800     P->bk = BK;                                                               \
1801     P->fd = FD;                                                               \
1802     FD->bk = BK->fd = P;                                                      \
1803   }                                                                           \
1804   else                                                                        \
1805   {                                                                           \
1806     IDX = bin_index(S);                                                       \
1807     BK = bin_at(IDX);                                                         \
1808     FD = BK->fd;                                                              \
1809     if (FD == BK) mark_binblock(IDX);                                         \
1810     else                                                                      \
1811     {                                                                         \
1812       while (FD != BK && S < chunksize(FD)) FD = FD->fd;                      \
1813       BK = FD->bk;                                                            \
1814     }                                                                         \
1815     P->bk = BK;                                                               \
1816     P->fd = FD;                                                               \
1817     FD->bk = BK->fd = P;                                                      \
1818   }                                                                           \
1819 }
1820 
1821 
1822 /* take a chunk off a list */
1823 
1824 #define unlink(P, BK, FD)                                                     \
1825 {                                                                             \
1826   BK = P->bk;                                                                 \
1827   FD = P->fd;                                                                 \
1828   FD->bk = BK;                                                                \
1829   BK->fd = FD;                                                                \
1830 }                                                                             \
1831 
1832 /* Place p as the last remainder */
1833 
1834 #define link_last_remainder(P)                                                \
1835 {                                                                             \
1836   last_remainder->fd = last_remainder->bk =  P;                               \
1837   P->fd = P->bk = last_remainder;                                             \
1838 }
1839 
1840 /* Clear the last_remainder bin */
1841 
1842 #define clear_last_remainder \
1843   (last_remainder->fd = last_remainder->bk = last_remainder)
1844 
1845 
1846 
1847 
1848 
1849 /* Routines dealing with mmap(). */
1850 
1851 #if HAVE_MMAP
1852 
1853 #if __STD_C
1854 static mchunkptr mmap_chunk(size_t size)
1855 #else
1856 static mchunkptr mmap_chunk(size) size_t size;
1857 #endif
1858 {
1859   size_t page_mask = malloc_getpagesize - 1;
1860   mchunkptr p;
1861 
1862 #ifndef MAP_ANONYMOUS
1863   static int fd = -1;
1864 #endif
1865 
1866   if(n_mmaps >= n_mmaps_max) return 0; /* too many regions */
1867 
1868   /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1869    * there is no following chunk whose prev_size field could be used.
1870    */
1871   size = (size + SIZE_SZ + page_mask) & ~page_mask;
1872 
1873 #ifdef MAP_ANONYMOUS
1874   p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE,
1875 		      MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
1876 #else /* !MAP_ANONYMOUS */
1877   if (fd < 0)
1878   {
1879     fd = open("/dev/zero", O_RDWR);
1880     if(fd < 0) return 0;
1881   }
1882   p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
1883 #endif
1884 
1885   if(p == (mchunkptr)-1) return 0;
1886 
1887   n_mmaps++;
1888   if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
1889 
1890   /* We demand that eight bytes into a page must be 8-byte aligned. */
1891   assert(aligned_OK(chunk2mem(p)));
1892 
1893   /* The offset to the start of the mmapped region is stored
1894    * in the prev_size field of the chunk; normally it is zero,
1895    * but that can be changed in memalign().
1896    */
1897   p->prev_size = 0;
1898   set_head(p, size|IS_MMAPPED);
1899 
1900   mmapped_mem += size;
1901   if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1902     max_mmapped_mem = mmapped_mem;
1903   if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1904     max_total_mem = mmapped_mem + sbrked_mem;
1905   return p;
1906 }
1907 
1908 #if __STD_C
1909 static void munmap_chunk(mchunkptr p)
1910 #else
1911 static void munmap_chunk(p) mchunkptr p;
1912 #endif
1913 {
1914   INTERNAL_SIZE_T size = chunksize(p);
1915   int ret;
1916 
1917   assert (chunk_is_mmapped(p));
1918   assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1919   assert((n_mmaps > 0));
1920   assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
1921 
1922   n_mmaps--;
1923   mmapped_mem -= (size + p->prev_size);
1924 
1925   ret = munmap((char *)p - p->prev_size, size + p->prev_size);
1926 
1927   /* munmap returns non-zero on failure */
1928   assert(ret == 0);
1929 }
1930 
1931 #if HAVE_MREMAP
1932 
1933 #if __STD_C
1934 static mchunkptr mremap_chunk(mchunkptr p, size_t new_size)
1935 #else
1936 static mchunkptr mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
1937 #endif
1938 {
1939   size_t page_mask = malloc_getpagesize - 1;
1940   INTERNAL_SIZE_T offset = p->prev_size;
1941   INTERNAL_SIZE_T size = chunksize(p);
1942   char *cp;
1943 
1944   assert (chunk_is_mmapped(p));
1945   assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1946   assert((n_mmaps > 0));
1947   assert(((size + offset) & (malloc_getpagesize-1)) == 0);
1948 
1949   /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1950   new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
1951 
1952   cp = (char *)mremap((char *)p - offset, size + offset, new_size, 1);
1953 
1954   if (cp == (char *)-1) return 0;
1955 
1956   p = (mchunkptr)(cp + offset);
1957 
1958   assert(aligned_OK(chunk2mem(p)));
1959 
1960   assert((p->prev_size == offset));
1961   set_head(p, (new_size - offset)|IS_MMAPPED);
1962 
1963   mmapped_mem -= size + offset;
1964   mmapped_mem += new_size;
1965   if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1966     max_mmapped_mem = mmapped_mem;
1967   if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1968     max_total_mem = mmapped_mem + sbrked_mem;
1969   return p;
1970 }
1971 
1972 #endif /* HAVE_MREMAP */
1973 
1974 #endif /* HAVE_MMAP */
1975 
1976 
1977 
1978 
1979 /*
1980   Extend the top-most chunk by obtaining memory from system.
1981   Main interface to sbrk (but see also malloc_trim).
1982 */
1983 
1984 #if __STD_C
1985 static void malloc_extend_top(INTERNAL_SIZE_T nb)
1986 #else
1987 static void malloc_extend_top(nb) INTERNAL_SIZE_T nb;
1988 #endif
1989 {
1990   char*     brk;                  /* return value from sbrk */
1991   INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
1992   INTERNAL_SIZE_T correction;     /* bytes for 2nd sbrk call */
1993   char*     new_brk;              /* return of 2nd sbrk call */
1994   INTERNAL_SIZE_T top_size;       /* new size of top chunk */
1995 
1996   mchunkptr old_top     = top;  /* Record state of old top */
1997   INTERNAL_SIZE_T old_top_size = chunksize(old_top);
1998   char*     old_end      = (char*)(chunk_at_offset(old_top, old_top_size));
1999 
2000   /* Pad request with top_pad plus minimal overhead */
2001 
2002   INTERNAL_SIZE_T    sbrk_size     = nb + top_pad + MINSIZE;
2003   unsigned long pagesz    = malloc_getpagesize;
2004 
2005   /* If not the first time through, round to preserve page boundary */
2006   /* Otherwise, we need to correct to a page size below anyway. */
2007   /* (We also correct below if an intervening foreign sbrk call.) */
2008 
2009   if (sbrk_base != (char*)(-1))
2010     sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
2011 
2012   brk = (char*)(MORECORE (sbrk_size));
2013 
2014   /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2015   if (brk == (char*)(MORECORE_FAILURE) ||
2016       (brk < old_end && old_top != initial_top))
2017     return;
2018 
2019   sbrked_mem += sbrk_size;
2020 
2021   if (brk == old_end) /* can just add bytes to current top */
2022   {
2023     top_size = sbrk_size + old_top_size;
2024     set_head(top, top_size | PREV_INUSE);
2025   }
2026   else
2027   {
2028     if (sbrk_base == (char*)(-1))  /* First time through. Record base */
2029       sbrk_base = brk;
2030     else  /* Someone else called sbrk().  Count those bytes as sbrked_mem. */
2031       sbrked_mem += brk - (char*)old_end;
2032 
2033     /* Guarantee alignment of first new chunk made from this space */
2034     front_misalign = (unsigned long)chunk2mem(brk) & MALLOC_ALIGN_MASK;
2035     if (front_misalign > 0)
2036     {
2037       correction = (MALLOC_ALIGNMENT) - front_misalign;
2038       brk += correction;
2039     }
2040     else
2041       correction = 0;
2042 
2043     /* Guarantee the next brk will be at a page boundary */
2044 
2045     correction += ((((unsigned long)(brk + sbrk_size))+(pagesz-1)) &
2046 		   ~(pagesz - 1)) - ((unsigned long)(brk + sbrk_size));
2047 
2048     /* Allocate correction */
2049     new_brk = (char*)(MORECORE (correction));
2050     if (new_brk == (char*)(MORECORE_FAILURE)) return;
2051 
2052     sbrked_mem += correction;
2053 
2054     top = (mchunkptr)brk;
2055     top_size = new_brk - brk + correction;
2056     set_head(top, top_size | PREV_INUSE);
2057 
2058     if (old_top != initial_top)
2059     {
2060 
2061       /* There must have been an intervening foreign sbrk call. */
2062       /* A double fencepost is necessary to prevent consolidation */
2063 
2064       /* If not enough space to do this, then user did something very wrong */
2065       if (old_top_size < MINSIZE)
2066       {
2067 	set_head(top, PREV_INUSE); /* will force null return from malloc */
2068 	return;
2069       }
2070 
2071       /* Also keep size a multiple of MALLOC_ALIGNMENT */
2072       old_top_size = (old_top_size - 3*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2073       set_head_size(old_top, old_top_size);
2074       chunk_at_offset(old_top, old_top_size          )->size =
2075 	SIZE_SZ|PREV_INUSE;
2076       chunk_at_offset(old_top, old_top_size + SIZE_SZ)->size =
2077 	SIZE_SZ|PREV_INUSE;
2078       /* If possible, release the rest. */
2079       if (old_top_size >= MINSIZE)
2080 	fREe(chunk2mem(old_top));
2081     }
2082   }
2083 
2084   if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
2085     max_sbrked_mem = sbrked_mem;
2086   if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2087     max_total_mem = mmapped_mem + sbrked_mem;
2088 
2089   /* We always land on a page boundary */
2090   assert(((unsigned long)((char*)top + top_size) & (pagesz - 1)) == 0);
2091 }
2092 
2093 
2094 
2095 
2096 /* Main public routines */
2097 
2098 
2099 /*
2100   Malloc Algorthim:
2101 
2102     The requested size is first converted into a usable form, `nb'.
2103     This currently means to add 4 bytes overhead plus possibly more to
2104     obtain 8-byte alignment and/or to obtain a size of at least
2105     MINSIZE (currently 16 bytes), the smallest allocatable size.
2106     (All fits are considered `exact' if they are within MINSIZE bytes.)
2107 
2108     From there, the first successful of the following steps is taken:
2109 
2110       1. The bin corresponding to the request size is scanned, and if
2111 	 a chunk of exactly the right size is found, it is taken.
2112 
2113       2. The most recently remaindered chunk is used if it is big
2114 	 enough.  This is a form of (roving) first fit, used only in
2115 	 the absence of exact fits. Runs of consecutive requests use
2116 	 the remainder of the chunk used for the previous such request
2117 	 whenever possible. This limited use of a first-fit style
2118 	 allocation strategy tends to give contiguous chunks
2119 	 coextensive lifetimes, which improves locality and can reduce
2120 	 fragmentation in the long run.
2121 
2122       3. Other bins are scanned in increasing size order, using a
2123 	 chunk big enough to fulfill the request, and splitting off
2124 	 any remainder.  This search is strictly by best-fit; i.e.,
2125 	 the smallest (with ties going to approximately the least
2126 	 recently used) chunk that fits is selected.
2127 
2128       4. If large enough, the chunk bordering the end of memory
2129 	 (`top') is split off. (This use of `top' is in accord with
2130 	 the best-fit search rule.  In effect, `top' is treated as
2131 	 larger (and thus less well fitting) than any other available
2132 	 chunk since it can be extended to be as large as necessary
2133 	 (up to system limitations).
2134 
2135       5. If the request size meets the mmap threshold and the
2136 	 system supports mmap, and there are few enough currently
2137 	 allocated mmapped regions, and a call to mmap succeeds,
2138 	 the request is allocated via direct memory mapping.
2139 
2140       6. Otherwise, the top of memory is extended by
2141 	 obtaining more space from the system (normally using sbrk,
2142 	 but definable to anything else via the MORECORE macro).
2143 	 Memory is gathered from the system (in system page-sized
2144 	 units) in a way that allows chunks obtained across different
2145 	 sbrk calls to be consolidated, but does not require
2146 	 contiguous memory. Thus, it should be safe to intersperse
2147 	 mallocs with other sbrk calls.
2148 
2149 
2150       All allocations are made from the the `lowest' part of any found
2151       chunk. (The implementation invariant is that prev_inuse is
2152       always true of any allocated chunk; i.e., that each allocated
2153       chunk borders either a previously allocated and still in-use chunk,
2154       or the base of its memory arena.)
2155 
2156 */
2157 
2158 #if __STD_C
2159 Void_t* mALLOc(size_t bytes)
2160 #else
2161 Void_t* mALLOc(bytes) size_t bytes;
2162 #endif
2163 {
2164   mchunkptr victim;                  /* inspected/selected chunk */
2165   INTERNAL_SIZE_T victim_size;       /* its size */
2166   int       idx;                     /* index for bin traversal */
2167   mbinptr   bin;                     /* associated bin */
2168   mchunkptr remainder;               /* remainder from a split */
2169   long      remainder_size;          /* its size */
2170   int       remainder_index;         /* its bin index */
2171   unsigned long block;               /* block traverser bit */
2172   int       startidx;                /* first bin of a traversed block */
2173   mchunkptr fwd;                     /* misc temp for linking */
2174   mchunkptr bck;                     /* misc temp for linking */
2175   mbinptr q;                         /* misc temp */
2176 
2177   INTERNAL_SIZE_T nb;
2178 
2179   /* check if mem_malloc_init() was run */
2180   if ((mem_malloc_start == 0) && (mem_malloc_end == 0)) {
2181     /* not initialized yet */
2182     return 0;
2183   }
2184 
2185   if ((long)bytes < 0) return 0;
2186 
2187   nb = request2size(bytes);  /* padded request size; */
2188 
2189   /* Check for exact match in a bin */
2190 
2191   if (is_small_request(nb))  /* Faster version for small requests */
2192   {
2193     idx = smallbin_index(nb);
2194 
2195     /* No traversal or size check necessary for small bins.  */
2196 
2197     q = bin_at(idx);
2198     victim = last(q);
2199 
2200     /* Also scan the next one, since it would have a remainder < MINSIZE */
2201     if (victim == q)
2202     {
2203       q = next_bin(q);
2204       victim = last(q);
2205     }
2206     if (victim != q)
2207     {
2208       victim_size = chunksize(victim);
2209       unlink(victim, bck, fwd);
2210       set_inuse_bit_at_offset(victim, victim_size);
2211       check_malloced_chunk(victim, nb);
2212       return chunk2mem(victim);
2213     }
2214 
2215     idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
2216 
2217   }
2218   else
2219   {
2220     idx = bin_index(nb);
2221     bin = bin_at(idx);
2222 
2223     for (victim = last(bin); victim != bin; victim = victim->bk)
2224     {
2225       victim_size = chunksize(victim);
2226       remainder_size = victim_size - nb;
2227 
2228       if (remainder_size >= (long)MINSIZE) /* too big */
2229       {
2230 	--idx; /* adjust to rescan below after checking last remainder */
2231 	break;
2232       }
2233 
2234       else if (remainder_size >= 0) /* exact fit */
2235       {
2236 	unlink(victim, bck, fwd);
2237 	set_inuse_bit_at_offset(victim, victim_size);
2238 	check_malloced_chunk(victim, nb);
2239 	return chunk2mem(victim);
2240       }
2241     }
2242 
2243     ++idx;
2244 
2245   }
2246 
2247   /* Try to use the last split-off remainder */
2248 
2249   if ( (victim = last_remainder->fd) != last_remainder)
2250   {
2251     victim_size = chunksize(victim);
2252     remainder_size = victim_size - nb;
2253 
2254     if (remainder_size >= (long)MINSIZE) /* re-split */
2255     {
2256       remainder = chunk_at_offset(victim, nb);
2257       set_head(victim, nb | PREV_INUSE);
2258       link_last_remainder(remainder);
2259       set_head(remainder, remainder_size | PREV_INUSE);
2260       set_foot(remainder, remainder_size);
2261       check_malloced_chunk(victim, nb);
2262       return chunk2mem(victim);
2263     }
2264 
2265     clear_last_remainder;
2266 
2267     if (remainder_size >= 0)  /* exhaust */
2268     {
2269       set_inuse_bit_at_offset(victim, victim_size);
2270       check_malloced_chunk(victim, nb);
2271       return chunk2mem(victim);
2272     }
2273 
2274     /* Else place in bin */
2275 
2276     frontlink(victim, victim_size, remainder_index, bck, fwd);
2277   }
2278 
2279   /*
2280      If there are any possibly nonempty big-enough blocks,
2281      search for best fitting chunk by scanning bins in blockwidth units.
2282   */
2283 
2284   if ( (block = idx2binblock(idx)) <= binblocks_r)
2285   {
2286 
2287     /* Get to the first marked block */
2288 
2289     if ( (block & binblocks_r) == 0)
2290     {
2291       /* force to an even block boundary */
2292       idx = (idx & ~(BINBLOCKWIDTH - 1)) + BINBLOCKWIDTH;
2293       block <<= 1;
2294       while ((block & binblocks_r) == 0)
2295       {
2296 	idx += BINBLOCKWIDTH;
2297 	block <<= 1;
2298       }
2299     }
2300 
2301     /* For each possibly nonempty block ... */
2302     for (;;)
2303     {
2304       startidx = idx;          /* (track incomplete blocks) */
2305       q = bin = bin_at(idx);
2306 
2307       /* For each bin in this block ... */
2308       do
2309       {
2310 	/* Find and use first big enough chunk ... */
2311 
2312 	for (victim = last(bin); victim != bin; victim = victim->bk)
2313 	{
2314 	  victim_size = chunksize(victim);
2315 	  remainder_size = victim_size - nb;
2316 
2317 	  if (remainder_size >= (long)MINSIZE) /* split */
2318 	  {
2319 	    remainder = chunk_at_offset(victim, nb);
2320 	    set_head(victim, nb | PREV_INUSE);
2321 	    unlink(victim, bck, fwd);
2322 	    link_last_remainder(remainder);
2323 	    set_head(remainder, remainder_size | PREV_INUSE);
2324 	    set_foot(remainder, remainder_size);
2325 	    check_malloced_chunk(victim, nb);
2326 	    return chunk2mem(victim);
2327 	  }
2328 
2329 	  else if (remainder_size >= 0)  /* take */
2330 	  {
2331 	    set_inuse_bit_at_offset(victim, victim_size);
2332 	    unlink(victim, bck, fwd);
2333 	    check_malloced_chunk(victim, nb);
2334 	    return chunk2mem(victim);
2335 	  }
2336 
2337 	}
2338 
2339        bin = next_bin(bin);
2340 
2341       } while ((++idx & (BINBLOCKWIDTH - 1)) != 0);
2342 
2343       /* Clear out the block bit. */
2344 
2345       do   /* Possibly backtrack to try to clear a partial block */
2346       {
2347 	if ((startidx & (BINBLOCKWIDTH - 1)) == 0)
2348 	{
2349 	  av_[1] = (mbinptr)(binblocks_r & ~block);
2350 	  break;
2351 	}
2352 	--startidx;
2353        q = prev_bin(q);
2354       } while (first(q) == q);
2355 
2356       /* Get to the next possibly nonempty block */
2357 
2358       if ( (block <<= 1) <= binblocks_r && (block != 0) )
2359       {
2360 	while ((block & binblocks_r) == 0)
2361 	{
2362 	  idx += BINBLOCKWIDTH;
2363 	  block <<= 1;
2364 	}
2365       }
2366       else
2367 	break;
2368     }
2369   }
2370 
2371 
2372   /* Try to use top chunk */
2373 
2374   /* Require that there be a remainder, ensuring top always exists  */
2375   if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2376   {
2377 
2378 #if HAVE_MMAP
2379     /* If big and would otherwise need to extend, try to use mmap instead */
2380     if ((unsigned long)nb >= (unsigned long)mmap_threshold &&
2381 	(victim = mmap_chunk(nb)) != 0)
2382       return chunk2mem(victim);
2383 #endif
2384 
2385     /* Try to extend */
2386     malloc_extend_top(nb);
2387     if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2388       return 0; /* propagate failure */
2389   }
2390 
2391   victim = top;
2392   set_head(victim, nb | PREV_INUSE);
2393   top = chunk_at_offset(victim, nb);
2394   set_head(top, remainder_size | PREV_INUSE);
2395   check_malloced_chunk(victim, nb);
2396   return chunk2mem(victim);
2397 
2398 }
2399 
2400 
2401 
2402 
2403 /*
2404 
2405   free() algorithm :
2406 
2407     cases:
2408 
2409        1. free(0) has no effect.
2410 
2411        2. If the chunk was allocated via mmap, it is release via munmap().
2412 
2413        3. If a returned chunk borders the current high end of memory,
2414 	  it is consolidated into the top, and if the total unused
2415 	  topmost memory exceeds the trim threshold, malloc_trim is
2416 	  called.
2417 
2418        4. Other chunks are consolidated as they arrive, and
2419 	  placed in corresponding bins. (This includes the case of
2420 	  consolidating with the current `last_remainder').
2421 
2422 */
2423 
2424 
2425 #if __STD_C
2426 void fREe(Void_t* mem)
2427 #else
2428 void fREe(mem) Void_t* mem;
2429 #endif
2430 {
2431   mchunkptr p;         /* chunk corresponding to mem */
2432   INTERNAL_SIZE_T hd;  /* its head field */
2433   INTERNAL_SIZE_T sz;  /* its size */
2434   int       idx;       /* its bin index */
2435   mchunkptr next;      /* next contiguous chunk */
2436   INTERNAL_SIZE_T nextsz; /* its size */
2437   INTERNAL_SIZE_T prevsz; /* size of previous contiguous chunk */
2438   mchunkptr bck;       /* misc temp for linking */
2439   mchunkptr fwd;       /* misc temp for linking */
2440   int       islr;      /* track whether merging with last_remainder */
2441 
2442   if (mem == 0)                              /* free(0) has no effect */
2443     return;
2444 
2445   p = mem2chunk(mem);
2446   hd = p->size;
2447 
2448 #if HAVE_MMAP
2449   if (hd & IS_MMAPPED)                       /* release mmapped memory. */
2450   {
2451     munmap_chunk(p);
2452     return;
2453   }
2454 #endif
2455 
2456   check_inuse_chunk(p);
2457 
2458   sz = hd & ~PREV_INUSE;
2459   next = chunk_at_offset(p, sz);
2460   nextsz = chunksize(next);
2461 
2462   if (next == top)                            /* merge with top */
2463   {
2464     sz += nextsz;
2465 
2466     if (!(hd & PREV_INUSE))                    /* consolidate backward */
2467     {
2468       prevsz = p->prev_size;
2469       p = chunk_at_offset(p, -((long) prevsz));
2470       sz += prevsz;
2471       unlink(p, bck, fwd);
2472     }
2473 
2474     set_head(p, sz | PREV_INUSE);
2475     top = p;
2476     if ((unsigned long)(sz) >= (unsigned long)trim_threshold)
2477       malloc_trim(top_pad);
2478     return;
2479   }
2480 
2481   set_head(next, nextsz);                    /* clear inuse bit */
2482 
2483   islr = 0;
2484 
2485   if (!(hd & PREV_INUSE))                    /* consolidate backward */
2486   {
2487     prevsz = p->prev_size;
2488     p = chunk_at_offset(p, -((long) prevsz));
2489     sz += prevsz;
2490 
2491     if (p->fd == last_remainder)             /* keep as last_remainder */
2492       islr = 1;
2493     else
2494       unlink(p, bck, fwd);
2495   }
2496 
2497   if (!(inuse_bit_at_offset(next, nextsz)))   /* consolidate forward */
2498   {
2499     sz += nextsz;
2500 
2501     if (!islr && next->fd == last_remainder)  /* re-insert last_remainder */
2502     {
2503       islr = 1;
2504       link_last_remainder(p);
2505     }
2506     else
2507       unlink(next, bck, fwd);
2508   }
2509 
2510 
2511   set_head(p, sz | PREV_INUSE);
2512   set_foot(p, sz);
2513   if (!islr)
2514     frontlink(p, sz, idx, bck, fwd);
2515 }
2516 
2517 
2518 
2519 
2520 
2521 /*
2522 
2523   Realloc algorithm:
2524 
2525     Chunks that were obtained via mmap cannot be extended or shrunk
2526     unless HAVE_MREMAP is defined, in which case mremap is used.
2527     Otherwise, if their reallocation is for additional space, they are
2528     copied.  If for less, they are just left alone.
2529 
2530     Otherwise, if the reallocation is for additional space, and the
2531     chunk can be extended, it is, else a malloc-copy-free sequence is
2532     taken.  There are several different ways that a chunk could be
2533     extended. All are tried:
2534 
2535        * Extending forward into following adjacent free chunk.
2536        * Shifting backwards, joining preceding adjacent space
2537        * Both shifting backwards and extending forward.
2538        * Extending into newly sbrked space
2539 
2540     Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2541     size argument of zero (re)allocates a minimum-sized chunk.
2542 
2543     If the reallocation is for less space, and the new request is for
2544     a `small' (<512 bytes) size, then the newly unused space is lopped
2545     off and freed.
2546 
2547     The old unix realloc convention of allowing the last-free'd chunk
2548     to be used as an argument to realloc is no longer supported.
2549     I don't know of any programs still relying on this feature,
2550     and allowing it would also allow too many other incorrect
2551     usages of realloc to be sensible.
2552 
2553 
2554 */
2555 
2556 
2557 #if __STD_C
2558 Void_t* rEALLOc(Void_t* oldmem, size_t bytes)
2559 #else
2560 Void_t* rEALLOc(oldmem, bytes) Void_t* oldmem; size_t bytes;
2561 #endif
2562 {
2563   INTERNAL_SIZE_T    nb;      /* padded request size */
2564 
2565   mchunkptr oldp;             /* chunk corresponding to oldmem */
2566   INTERNAL_SIZE_T    oldsize; /* its size */
2567 
2568   mchunkptr newp;             /* chunk to return */
2569   INTERNAL_SIZE_T    newsize; /* its size */
2570   Void_t*   newmem;           /* corresponding user mem */
2571 
2572   mchunkptr next;             /* next contiguous chunk after oldp */
2573   INTERNAL_SIZE_T  nextsize;  /* its size */
2574 
2575   mchunkptr prev;             /* previous contiguous chunk before oldp */
2576   INTERNAL_SIZE_T  prevsize;  /* its size */
2577 
2578   mchunkptr remainder;        /* holds split off extra space from newp */
2579   INTERNAL_SIZE_T  remainder_size;   /* its size */
2580 
2581   mchunkptr bck;              /* misc temp for linking */
2582   mchunkptr fwd;              /* misc temp for linking */
2583 
2584 #ifdef REALLOC_ZERO_BYTES_FREES
2585   if (bytes == 0) { fREe(oldmem); return 0; }
2586 #endif
2587 
2588   if ((long)bytes < 0) return 0;
2589 
2590   /* realloc of null is supposed to be same as malloc */
2591   if (oldmem == 0) return mALLOc(bytes);
2592 
2593   newp    = oldp    = mem2chunk(oldmem);
2594   newsize = oldsize = chunksize(oldp);
2595 
2596 
2597   nb = request2size(bytes);
2598 
2599 #if HAVE_MMAP
2600   if (chunk_is_mmapped(oldp))
2601   {
2602 #if HAVE_MREMAP
2603     newp = mremap_chunk(oldp, nb);
2604     if(newp) return chunk2mem(newp);
2605 #endif
2606     /* Note the extra SIZE_SZ overhead. */
2607     if(oldsize - SIZE_SZ >= nb) return oldmem; /* do nothing */
2608     /* Must alloc, copy, free. */
2609     newmem = mALLOc(bytes);
2610     if (newmem == 0) return 0; /* propagate failure */
2611     MALLOC_COPY(newmem, oldmem, oldsize - 2*SIZE_SZ);
2612     munmap_chunk(oldp);
2613     return newmem;
2614   }
2615 #endif
2616 
2617   check_inuse_chunk(oldp);
2618 
2619   if ((long)(oldsize) < (long)(nb))
2620   {
2621 
2622     /* Try expanding forward */
2623 
2624     next = chunk_at_offset(oldp, oldsize);
2625     if (next == top || !inuse(next))
2626     {
2627       nextsize = chunksize(next);
2628 
2629       /* Forward into top only if a remainder */
2630       if (next == top)
2631       {
2632 	if ((long)(nextsize + newsize) >= (long)(nb + MINSIZE))
2633 	{
2634 	  newsize += nextsize;
2635 	  top = chunk_at_offset(oldp, nb);
2636 	  set_head(top, (newsize - nb) | PREV_INUSE);
2637 	  set_head_size(oldp, nb);
2638 	  return chunk2mem(oldp);
2639 	}
2640       }
2641 
2642       /* Forward into next chunk */
2643       else if (((long)(nextsize + newsize) >= (long)(nb)))
2644       {
2645 	unlink(next, bck, fwd);
2646 	newsize  += nextsize;
2647 	goto split;
2648       }
2649     }
2650     else
2651     {
2652       next = 0;
2653       nextsize = 0;
2654     }
2655 
2656     /* Try shifting backwards. */
2657 
2658     if (!prev_inuse(oldp))
2659     {
2660       prev = prev_chunk(oldp);
2661       prevsize = chunksize(prev);
2662 
2663       /* try forward + backward first to save a later consolidation */
2664 
2665       if (next != 0)
2666       {
2667 	/* into top */
2668 	if (next == top)
2669 	{
2670 	  if ((long)(nextsize + prevsize + newsize) >= (long)(nb + MINSIZE))
2671 	  {
2672 	    unlink(prev, bck, fwd);
2673 	    newp = prev;
2674 	    newsize += prevsize + nextsize;
2675 	    newmem = chunk2mem(newp);
2676 	    MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2677 	    top = chunk_at_offset(newp, nb);
2678 	    set_head(top, (newsize - nb) | PREV_INUSE);
2679 	    set_head_size(newp, nb);
2680 	    return newmem;
2681 	  }
2682 	}
2683 
2684 	/* into next chunk */
2685 	else if (((long)(nextsize + prevsize + newsize) >= (long)(nb)))
2686 	{
2687 	  unlink(next, bck, fwd);
2688 	  unlink(prev, bck, fwd);
2689 	  newp = prev;
2690 	  newsize += nextsize + prevsize;
2691 	  newmem = chunk2mem(newp);
2692 	  MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2693 	  goto split;
2694 	}
2695       }
2696 
2697       /* backward only */
2698       if (prev != 0 && (long)(prevsize + newsize) >= (long)nb)
2699       {
2700 	unlink(prev, bck, fwd);
2701 	newp = prev;
2702 	newsize += prevsize;
2703 	newmem = chunk2mem(newp);
2704 	MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2705 	goto split;
2706       }
2707     }
2708 
2709     /* Must allocate */
2710 
2711     newmem = mALLOc (bytes);
2712 
2713     if (newmem == 0)  /* propagate failure */
2714       return 0;
2715 
2716     /* Avoid copy if newp is next chunk after oldp. */
2717     /* (This can only happen when new chunk is sbrk'ed.) */
2718 
2719     if ( (newp = mem2chunk(newmem)) == next_chunk(oldp))
2720     {
2721       newsize += chunksize(newp);
2722       newp = oldp;
2723       goto split;
2724     }
2725 
2726     /* Otherwise copy, free, and exit */
2727     MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2728     fREe(oldmem);
2729     return newmem;
2730   }
2731 
2732 
2733  split:  /* split off extra room in old or expanded chunk */
2734 
2735   if (newsize - nb >= MINSIZE) /* split off remainder */
2736   {
2737     remainder = chunk_at_offset(newp, nb);
2738     remainder_size = newsize - nb;
2739     set_head_size(newp, nb);
2740     set_head(remainder, remainder_size | PREV_INUSE);
2741     set_inuse_bit_at_offset(remainder, remainder_size);
2742     fREe(chunk2mem(remainder)); /* let free() deal with it */
2743   }
2744   else
2745   {
2746     set_head_size(newp, newsize);
2747     set_inuse_bit_at_offset(newp, newsize);
2748   }
2749 
2750   check_inuse_chunk(newp);
2751   return chunk2mem(newp);
2752 }
2753 
2754 
2755 
2756 
2757 /*
2758 
2759   memalign algorithm:
2760 
2761     memalign requests more than enough space from malloc, finds a spot
2762     within that chunk that meets the alignment request, and then
2763     possibly frees the leading and trailing space.
2764 
2765     The alignment argument must be a power of two. This property is not
2766     checked by memalign, so misuse may result in random runtime errors.
2767 
2768     8-byte alignment is guaranteed by normal malloc calls, so don't
2769     bother calling memalign with an argument of 8 or less.
2770 
2771     Overreliance on memalign is a sure way to fragment space.
2772 
2773 */
2774 
2775 
2776 #if __STD_C
2777 Void_t* mEMALIGn(size_t alignment, size_t bytes)
2778 #else
2779 Void_t* mEMALIGn(alignment, bytes) size_t alignment; size_t bytes;
2780 #endif
2781 {
2782   INTERNAL_SIZE_T    nb;      /* padded  request size */
2783   char*     m;                /* memory returned by malloc call */
2784   mchunkptr p;                /* corresponding chunk */
2785   char*     brk;              /* alignment point within p */
2786   mchunkptr newp;             /* chunk to return */
2787   INTERNAL_SIZE_T  newsize;   /* its size */
2788   INTERNAL_SIZE_T  leadsize;  /* leading space befor alignment point */
2789   mchunkptr remainder;        /* spare room at end to split off */
2790   long      remainder_size;   /* its size */
2791 
2792   if ((long)bytes < 0) return 0;
2793 
2794   /* If need less alignment than we give anyway, just relay to malloc */
2795 
2796   if (alignment <= MALLOC_ALIGNMENT) return mALLOc(bytes);
2797 
2798   /* Otherwise, ensure that it is at least a minimum chunk size */
2799 
2800   if (alignment <  MINSIZE) alignment = MINSIZE;
2801 
2802   /* Call malloc with worst case padding to hit alignment. */
2803 
2804   nb = request2size(bytes);
2805   m  = (char*)(mALLOc(nb + alignment + MINSIZE));
2806 
2807   if (m == 0) return 0; /* propagate failure */
2808 
2809   p = mem2chunk(m);
2810 
2811   if ((((unsigned long)(m)) % alignment) == 0) /* aligned */
2812   {
2813 #if HAVE_MMAP
2814     if(chunk_is_mmapped(p))
2815       return chunk2mem(p); /* nothing more to do */
2816 #endif
2817   }
2818   else /* misaligned */
2819   {
2820     /*
2821       Find an aligned spot inside chunk.
2822       Since we need to give back leading space in a chunk of at
2823       least MINSIZE, if the first calculation places us at
2824       a spot with less than MINSIZE leader, we can move to the
2825       next aligned spot -- we've allocated enough total room so that
2826       this is always possible.
2827     */
2828 
2829     brk = (char*)mem2chunk(((unsigned long)(m + alignment - 1)) & -((signed) alignment));
2830     if ((long)(brk - (char*)(p)) < MINSIZE) brk = brk + alignment;
2831 
2832     newp = (mchunkptr)brk;
2833     leadsize = brk - (char*)(p);
2834     newsize = chunksize(p) - leadsize;
2835 
2836 #if HAVE_MMAP
2837     if(chunk_is_mmapped(p))
2838     {
2839       newp->prev_size = p->prev_size + leadsize;
2840       set_head(newp, newsize|IS_MMAPPED);
2841       return chunk2mem(newp);
2842     }
2843 #endif
2844 
2845     /* give back leader, use the rest */
2846 
2847     set_head(newp, newsize | PREV_INUSE);
2848     set_inuse_bit_at_offset(newp, newsize);
2849     set_head_size(p, leadsize);
2850     fREe(chunk2mem(p));
2851     p = newp;
2852 
2853     assert (newsize >= nb && (((unsigned long)(chunk2mem(p))) % alignment) == 0);
2854   }
2855 
2856   /* Also give back spare room at the end */
2857 
2858   remainder_size = chunksize(p) - nb;
2859 
2860   if (remainder_size >= (long)MINSIZE)
2861   {
2862     remainder = chunk_at_offset(p, nb);
2863     set_head(remainder, remainder_size | PREV_INUSE);
2864     set_head_size(p, nb);
2865     fREe(chunk2mem(remainder));
2866   }
2867 
2868   check_inuse_chunk(p);
2869   return chunk2mem(p);
2870 
2871 }
2872 
2873 
2874 
2875 
2876 /*
2877     valloc just invokes memalign with alignment argument equal
2878     to the page size of the system (or as near to this as can
2879     be figured out from all the includes/defines above.)
2880 */
2881 
2882 #if __STD_C
2883 Void_t* vALLOc(size_t bytes)
2884 #else
2885 Void_t* vALLOc(bytes) size_t bytes;
2886 #endif
2887 {
2888   return mEMALIGn (malloc_getpagesize, bytes);
2889 }
2890 
2891 /*
2892   pvalloc just invokes valloc for the nearest pagesize
2893   that will accommodate request
2894 */
2895 
2896 
2897 #if __STD_C
2898 Void_t* pvALLOc(size_t bytes)
2899 #else
2900 Void_t* pvALLOc(bytes) size_t bytes;
2901 #endif
2902 {
2903   size_t pagesize = malloc_getpagesize;
2904   return mEMALIGn (pagesize, (bytes + pagesize - 1) & ~(pagesize - 1));
2905 }
2906 
2907 /*
2908 
2909   calloc calls malloc, then zeroes out the allocated chunk.
2910 
2911 */
2912 
2913 #if __STD_C
2914 Void_t* cALLOc(size_t n, size_t elem_size)
2915 #else
2916 Void_t* cALLOc(n, elem_size) size_t n; size_t elem_size;
2917 #endif
2918 {
2919   mchunkptr p;
2920   INTERNAL_SIZE_T csz;
2921 
2922   INTERNAL_SIZE_T sz = n * elem_size;
2923 
2924 
2925   /* check if expand_top called, in which case don't need to clear */
2926 #if MORECORE_CLEARS
2927   mchunkptr oldtop = top;
2928   INTERNAL_SIZE_T oldtopsize = chunksize(top);
2929 #endif
2930   Void_t* mem = mALLOc (sz);
2931 
2932   if ((long)n < 0) return 0;
2933 
2934   if (mem == 0)
2935     return 0;
2936   else
2937   {
2938     p = mem2chunk(mem);
2939 
2940     /* Two optional cases in which clearing not necessary */
2941 
2942 
2943 #if HAVE_MMAP
2944     if (chunk_is_mmapped(p)) return mem;
2945 #endif
2946 
2947     csz = chunksize(p);
2948 
2949 #if MORECORE_CLEARS
2950     if (p == oldtop && csz > oldtopsize)
2951     {
2952       /* clear only the bytes from non-freshly-sbrked memory */
2953       csz = oldtopsize;
2954     }
2955 #endif
2956 
2957     MALLOC_ZERO(mem, csz - SIZE_SZ);
2958     return mem;
2959   }
2960 }
2961 
2962 /*
2963 
2964   cfree just calls free. It is needed/defined on some systems
2965   that pair it with calloc, presumably for odd historical reasons.
2966 
2967 */
2968 
2969 #if !defined(INTERNAL_LINUX_C_LIB) || !defined(__ELF__)
2970 #if __STD_C
2971 void cfree(Void_t *mem)
2972 #else
2973 void cfree(mem) Void_t *mem;
2974 #endif
2975 {
2976   fREe(mem);
2977 }
2978 #endif
2979 
2980 
2981 
2982 /*
2983 
2984     Malloc_trim gives memory back to the system (via negative
2985     arguments to sbrk) if there is unused memory at the `high' end of
2986     the malloc pool. You can call this after freeing large blocks of
2987     memory to potentially reduce the system-level memory requirements
2988     of a program. However, it cannot guarantee to reduce memory. Under
2989     some allocation patterns, some large free blocks of memory will be
2990     locked between two used chunks, so they cannot be given back to
2991     the system.
2992 
2993     The `pad' argument to malloc_trim represents the amount of free
2994     trailing space to leave untrimmed. If this argument is zero,
2995     only the minimum amount of memory to maintain internal data
2996     structures will be left (one page or less). Non-zero arguments
2997     can be supplied to maintain enough trailing space to service
2998     future expected allocations without having to re-obtain memory
2999     from the system.
3000 
3001     Malloc_trim returns 1 if it actually released any memory, else 0.
3002 
3003 */
3004 
3005 #if __STD_C
3006 int malloc_trim(size_t pad)
3007 #else
3008 int malloc_trim(pad) size_t pad;
3009 #endif
3010 {
3011   long  top_size;        /* Amount of top-most memory */
3012   long  extra;           /* Amount to release */
3013   char* current_brk;     /* address returned by pre-check sbrk call */
3014   char* new_brk;         /* address returned by negative sbrk call */
3015 
3016   unsigned long pagesz = malloc_getpagesize;
3017 
3018   top_size = chunksize(top);
3019   extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
3020 
3021   if (extra < (long)pagesz)  /* Not enough memory to release */
3022     return 0;
3023 
3024   else
3025   {
3026     /* Test to make sure no one else called sbrk */
3027     current_brk = (char*)(MORECORE (0));
3028     if (current_brk != (char*)(top) + top_size)
3029       return 0;     /* Apparently we don't own memory; must fail */
3030 
3031     else
3032     {
3033       new_brk = (char*)(MORECORE (-extra));
3034 
3035       if (new_brk == (char*)(MORECORE_FAILURE)) /* sbrk failed? */
3036       {
3037 	/* Try to figure out what we have */
3038 	current_brk = (char*)(MORECORE (0));
3039 	top_size = current_brk - (char*)top;
3040 	if (top_size >= (long)MINSIZE) /* if not, we are very very dead! */
3041 	{
3042 	  sbrked_mem = current_brk - sbrk_base;
3043 	  set_head(top, top_size | PREV_INUSE);
3044 	}
3045 	check_chunk(top);
3046 	return 0;
3047       }
3048 
3049       else
3050       {
3051 	/* Success. Adjust top accordingly. */
3052 	set_head(top, (top_size - extra) | PREV_INUSE);
3053 	sbrked_mem -= extra;
3054 	check_chunk(top);
3055 	return 1;
3056       }
3057     }
3058   }
3059 }
3060 
3061 
3062 
3063 /*
3064   malloc_usable_size:
3065 
3066     This routine tells you how many bytes you can actually use in an
3067     allocated chunk, which may be more than you requested (although
3068     often not). You can use this many bytes without worrying about
3069     overwriting other allocated objects. Not a particularly great
3070     programming practice, but still sometimes useful.
3071 
3072 */
3073 
3074 #if __STD_C
3075 size_t malloc_usable_size(Void_t* mem)
3076 #else
3077 size_t malloc_usable_size(mem) Void_t* mem;
3078 #endif
3079 {
3080   mchunkptr p;
3081   if (mem == 0)
3082     return 0;
3083   else
3084   {
3085     p = mem2chunk(mem);
3086     if(!chunk_is_mmapped(p))
3087     {
3088       if (!inuse(p)) return 0;
3089       check_inuse_chunk(p);
3090       return chunksize(p) - SIZE_SZ;
3091     }
3092     return chunksize(p) - 2*SIZE_SZ;
3093   }
3094 }
3095 
3096 
3097 
3098 
3099 /* Utility to update current_mallinfo for malloc_stats and mallinfo() */
3100 
3101 #ifdef DEBUG
3102 static void malloc_update_mallinfo()
3103 {
3104   int i;
3105   mbinptr b;
3106   mchunkptr p;
3107 #ifdef DEBUG
3108   mchunkptr q;
3109 #endif
3110 
3111   INTERNAL_SIZE_T avail = chunksize(top);
3112   int   navail = ((long)(avail) >= (long)MINSIZE)? 1 : 0;
3113 
3114   for (i = 1; i < NAV; ++i)
3115   {
3116     b = bin_at(i);
3117     for (p = last(b); p != b; p = p->bk)
3118     {
3119 #ifdef DEBUG
3120       check_free_chunk(p);
3121       for (q = next_chunk(p);
3122 	   q < top && inuse(q) && (long)(chunksize(q)) >= (long)MINSIZE;
3123 	   q = next_chunk(q))
3124 	check_inuse_chunk(q);
3125 #endif
3126       avail += chunksize(p);
3127       navail++;
3128     }
3129   }
3130 
3131   current_mallinfo.ordblks = navail;
3132   current_mallinfo.uordblks = sbrked_mem - avail;
3133   current_mallinfo.fordblks = avail;
3134   current_mallinfo.hblks = n_mmaps;
3135   current_mallinfo.hblkhd = mmapped_mem;
3136   current_mallinfo.keepcost = chunksize(top);
3137 
3138 }
3139 #endif	/* DEBUG */
3140 
3141 
3142 
3143 /*
3144 
3145   malloc_stats:
3146 
3147     Prints on the amount of space obtain from the system (both
3148     via sbrk and mmap), the maximum amount (which may be more than
3149     current if malloc_trim and/or munmap got called), the maximum
3150     number of simultaneous mmap regions used, and the current number
3151     of bytes allocated via malloc (or realloc, etc) but not yet
3152     freed. (Note that this is the number of bytes allocated, not the
3153     number requested. It will be larger than the number requested
3154     because of alignment and bookkeeping overhead.)
3155 
3156 */
3157 
3158 #ifdef DEBUG
3159 void malloc_stats()
3160 {
3161   malloc_update_mallinfo();
3162   printf("max system bytes = %10u\n",
3163 	  (unsigned int)(max_total_mem));
3164   printf("system bytes     = %10u\n",
3165 	  (unsigned int)(sbrked_mem + mmapped_mem));
3166   printf("in use bytes     = %10u\n",
3167 	  (unsigned int)(current_mallinfo.uordblks + mmapped_mem));
3168 #if HAVE_MMAP
3169   printf("max mmap regions = %10u\n",
3170 	  (unsigned int)max_n_mmaps);
3171 #endif
3172 }
3173 #endif	/* DEBUG */
3174 
3175 /*
3176   mallinfo returns a copy of updated current mallinfo.
3177 */
3178 
3179 #ifdef DEBUG
3180 struct mallinfo mALLINFo()
3181 {
3182   malloc_update_mallinfo();
3183   return current_mallinfo;
3184 }
3185 #endif	/* DEBUG */
3186 
3187 
3188 
3189 
3190 /*
3191   mallopt:
3192 
3193     mallopt is the general SVID/XPG interface to tunable parameters.
3194     The format is to provide a (parameter-number, parameter-value) pair.
3195     mallopt then sets the corresponding parameter to the argument
3196     value if it can (i.e., so long as the value is meaningful),
3197     and returns 1 if successful else 0.
3198 
3199     See descriptions of tunable parameters above.
3200 
3201 */
3202 
3203 #if __STD_C
3204 int mALLOPt(int param_number, int value)
3205 #else
3206 int mALLOPt(param_number, value) int param_number; int value;
3207 #endif
3208 {
3209   switch(param_number)
3210   {
3211     case M_TRIM_THRESHOLD:
3212       trim_threshold = value; return 1;
3213     case M_TOP_PAD:
3214       top_pad = value; return 1;
3215     case M_MMAP_THRESHOLD:
3216       mmap_threshold = value; return 1;
3217     case M_MMAP_MAX:
3218 #if HAVE_MMAP
3219       n_mmaps_max = value; return 1;
3220 #else
3221       if (value != 0) return 0; else  n_mmaps_max = value; return 1;
3222 #endif
3223 
3224     default:
3225       return 0;
3226   }
3227 }
3228 
3229 /*
3230 
3231 History:
3232 
3233     V2.6.6 Sun Dec  5 07:42:19 1999  Doug Lea  (dl at gee)
3234       * return null for negative arguments
3235       * Added Several WIN32 cleanups from Martin C. Fong <mcfong@yahoo.com>
3236 	 * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h'
3237 	  (e.g. WIN32 platforms)
3238 	 * Cleanup up header file inclusion for WIN32 platforms
3239 	 * Cleanup code to avoid Microsoft Visual C++ compiler complaints
3240 	 * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing
3241 	   memory allocation routines
3242 	 * Set 'malloc_getpagesize' for WIN32 platforms (needs more work)
3243 	 * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to
3244 	   usage of 'assert' in non-WIN32 code
3245 	 * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to
3246 	   avoid infinite loop
3247       * Always call 'fREe()' rather than 'free()'
3248 
3249     V2.6.5 Wed Jun 17 15:57:31 1998  Doug Lea  (dl at gee)
3250       * Fixed ordering problem with boundary-stamping
3251 
3252     V2.6.3 Sun May 19 08:17:58 1996  Doug Lea  (dl at gee)
3253       * Added pvalloc, as recommended by H.J. Liu
3254       * Added 64bit pointer support mainly from Wolfram Gloger
3255       * Added anonymously donated WIN32 sbrk emulation
3256       * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3257       * malloc_extend_top: fix mask error that caused wastage after
3258 	foreign sbrks
3259       * Add linux mremap support code from HJ Liu
3260 
3261     V2.6.2 Tue Dec  5 06:52:55 1995  Doug Lea  (dl at gee)
3262       * Integrated most documentation with the code.
3263       * Add support for mmap, with help from
3264 	Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3265       * Use last_remainder in more cases.
3266       * Pack bins using idea from  colin@nyx10.cs.du.edu
3267       * Use ordered bins instead of best-fit threshhold
3268       * Eliminate block-local decls to simplify tracing and debugging.
3269       * Support another case of realloc via move into top
3270       * Fix error occuring when initial sbrk_base not word-aligned.
3271       * Rely on page size for units instead of SBRK_UNIT to
3272 	avoid surprises about sbrk alignment conventions.
3273       * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3274 	(raymond@es.ele.tue.nl) for the suggestion.
3275       * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3276       * More precautions for cases where other routines call sbrk,
3277 	courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3278       * Added macros etc., allowing use in linux libc from
3279 	H.J. Lu (hjl@gnu.ai.mit.edu)
3280       * Inverted this history list
3281 
3282     V2.6.1 Sat Dec  2 14:10:57 1995  Doug Lea  (dl at gee)
3283       * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3284       * Removed all preallocation code since under current scheme
3285 	the work required to undo bad preallocations exceeds
3286 	the work saved in good cases for most test programs.
3287       * No longer use return list or unconsolidated bins since
3288 	no scheme using them consistently outperforms those that don't
3289 	given above changes.
3290       * Use best fit for very large chunks to prevent some worst-cases.
3291       * Added some support for debugging
3292 
3293     V2.6.0 Sat Nov  4 07:05:23 1995  Doug Lea  (dl at gee)
3294       * Removed footers when chunks are in use. Thanks to
3295 	Paul Wilson (wilson@cs.texas.edu) for the suggestion.
3296 
3297     V2.5.4 Wed Nov  1 07:54:51 1995  Doug Lea  (dl at gee)
3298       * Added malloc_trim, with help from Wolfram Gloger
3299 	(wmglo@Dent.MED.Uni-Muenchen.DE).
3300 
3301     V2.5.3 Tue Apr 26 10:16:01 1994  Doug Lea  (dl at g)
3302 
3303     V2.5.2 Tue Apr  5 16:20:40 1994  Doug Lea  (dl at g)
3304       * realloc: try to expand in both directions
3305       * malloc: swap order of clean-bin strategy;
3306       * realloc: only conditionally expand backwards
3307       * Try not to scavenge used bins
3308       * Use bin counts as a guide to preallocation
3309       * Occasionally bin return list chunks in first scan
3310       * Add a few optimizations from colin@nyx10.cs.du.edu
3311 
3312     V2.5.1 Sat Aug 14 15:40:43 1993  Doug Lea  (dl at g)
3313       * faster bin computation & slightly different binning
3314       * merged all consolidations to one part of malloc proper
3315 	 (eliminating old malloc_find_space & malloc_clean_bin)
3316       * Scan 2 returns chunks (not just 1)
3317       * Propagate failure in realloc if malloc returns 0
3318       * Add stuff to allow compilation on non-ANSI compilers
3319 	  from kpv@research.att.com
3320 
3321     V2.5 Sat Aug  7 07:41:59 1993  Doug Lea  (dl at g.oswego.edu)
3322       * removed potential for odd address access in prev_chunk
3323       * removed dependency on getpagesize.h
3324       * misc cosmetics and a bit more internal documentation
3325       * anticosmetics: mangled names in macros to evade debugger strangeness
3326       * tested on sparc, hp-700, dec-mips, rs6000
3327 	  with gcc & native cc (hp, dec only) allowing
3328 	  Detlefs & Zorn comparison study (in SIGPLAN Notices.)
3329 
3330     Trial version Fri Aug 28 13:14:29 1992  Doug Lea  (dl at g.oswego.edu)
3331       * Based loosely on libg++-1.2X malloc. (It retains some of the overall
3332 	 structure of old version,  but most details differ.)
3333 
3334 */
3335