3 #if 0 /* Moved to malloc.h */
4 /* ---------- To make a malloc.h, start cutting here ------------ */
7 A version of malloc/free/realloc written by Doug Lea and released to the
8 public domain. Send questions/comments/complaints/performance data
11 * VERSION 2.6.6 Sun Mar 5 19:10:03 2000 Doug Lea (dl at gee)
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!
17 * Why use this malloc?
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
26 * Synopsis of public routines
28 (Much fuller descriptions are contained in the program documentation below.)
31 Return a pointer to a newly allocated chunk of at least n bytes, or null
32 if no space is available.
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
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.)
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
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.
66 Prints brief summary statistics.
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.
76 8 byte alignment is currently hardwired into the design. This
77 seems to suffice for all current machines and C compilers.
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.
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.
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.
91 Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
92 8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
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.
100 Even a request for zero bytes (i.e., malloc(0)) returns a
101 pointer to something of the minimum allocatable size.
103 Maximum allocated size: 4-byte size_t: 2^31 - 8 bytes
104 8-byte size_t: 2^63 - 16 bytes
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.
114 Maximum overhead wastage per allocated chunk: normally 15 bytes
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
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.
128 Here are some features that are NOT currently supported
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.
135 * Synopsis of compile-time options:
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.
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
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
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
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
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
206 DEFAULT_TRIM_THRESHOLD
208 DEFAULT_MMAP_THRESHOLD
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
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.
235 #endif /*__cplusplus*/
240 #if (__STD_C || defined(WIN32))
248 #include <stddef.h> /* for size_t */
250 #include <sys/types.h>
257 #include <stdio.h> /* needed for malloc_stats */
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.
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.)
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.
290 INTERNAL_SIZE_T is the word-size used for internal bookkeeping
291 of chunk sizes. On a 64-bit machine, you can reduce malloc
292 overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
293 at the expense of not being able to handle requests greater than
294 2^31. This limitation is hardly ever a concern; you are encouraged
295 to set this. However, the default version is the same as size_t.
298 #ifndef INTERNAL_SIZE_T
299 #define INTERNAL_SIZE_T size_t
303 REALLOC_ZERO_BYTES_FREES should be set if a call to
304 realloc with zero bytes should be the same as a call to free.
305 Some people think it should. Otherwise, since this malloc
306 returns a unique pointer for malloc(0), so does realloc(p, 0).
310 /* #define REALLOC_ZERO_BYTES_FREES */
314 WIN32 causes an emulation of sbrk to be compiled in
315 mmap-based options are not currently supported in WIN32.
320 #define MORECORE wsbrk
323 #define LACKS_UNISTD_H
324 #define LACKS_SYS_PARAM_H
327 Include 'windows.h' to get the necessary declarations for the
328 Microsoft Visual C++ data structures and routines used in the 'sbrk'
331 Define WIN32_LEAN_AND_MEAN so that only the essential Microsoft
332 Visual C++ header files are included.
334 #define WIN32_LEAN_AND_MEAN
340 HAVE_MEMCPY should be defined if you are not otherwise using
341 ANSI STD C, but still have memcpy and memset in your C library
342 and want to use them in calloc and realloc. Otherwise simple
343 macro versions are defined here.
345 USE_MEMCPY should be defined as 1 if you actually want to
346 have memset and memcpy called. People report that the macro
347 versions are often enough faster than libc versions on many
348 systems that it is better to use them.
362 #if (__STD_C || defined(HAVE_MEMCPY))
365 void* memset(void*, int, size_t);
366 void* memcpy(void*, const void*, size_t);
369 /* On Win32 platforms, 'memset()' and 'memcpy()' are already declared in */
380 /* The following macros are only invoked with (2n+1)-multiples of
381 INTERNAL_SIZE_T units, with a positive integer n. This is exploited
382 for fast inline execution when n is small. */
384 #define MALLOC_ZERO(charp, nbytes) \
386 INTERNAL_SIZE_T mzsz = (nbytes); \
387 if(mzsz <= 9*sizeof(mzsz)) { \
388 INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
389 if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
391 if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
393 if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
398 } else memset((charp), 0, mzsz); \
401 #define MALLOC_COPY(dest,src,nbytes) \
403 INTERNAL_SIZE_T mcsz = (nbytes); \
404 if(mcsz <= 9*sizeof(mcsz)) { \
405 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
406 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
407 if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
408 *mcdst++ = *mcsrc++; \
409 if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
410 *mcdst++ = *mcsrc++; \
411 if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
412 *mcdst++ = *mcsrc++; }}} \
413 *mcdst++ = *mcsrc++; \
414 *mcdst++ = *mcsrc++; \
416 } else memcpy(dest, src, mcsz); \
419 #else /* !USE_MEMCPY */
421 /* Use Duff's device for good zeroing/copying performance. */
423 #define MALLOC_ZERO(charp, nbytes) \
425 INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
426 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
427 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
429 case 0: for(;;) { *mzp++ = 0; \
430 case 7: *mzp++ = 0; \
431 case 6: *mzp++ = 0; \
432 case 5: *mzp++ = 0; \
433 case 4: *mzp++ = 0; \
434 case 3: *mzp++ = 0; \
435 case 2: *mzp++ = 0; \
436 case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
440 #define MALLOC_COPY(dest,src,nbytes) \
442 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
443 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
444 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
445 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
447 case 0: for(;;) { *mcdst++ = *mcsrc++; \
448 case 7: *mcdst++ = *mcsrc++; \
449 case 6: *mcdst++ = *mcsrc++; \
450 case 5: *mcdst++ = *mcsrc++; \
451 case 4: *mcdst++ = *mcsrc++; \
452 case 3: *mcdst++ = *mcsrc++; \
453 case 2: *mcdst++ = *mcsrc++; \
454 case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
462 Define HAVE_MMAP to optionally make malloc() use mmap() to
463 allocate very large blocks. These will be returned to the
464 operating system immediately after a free().
472 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
473 large blocks. This is currently only possible on Linux with
474 kernel versions newer than 1.3.77.
478 #ifdef INTERNAL_LINUX_C_LIB
479 #define HAVE_MREMAP 1
481 #define HAVE_MREMAP 0
489 #include <sys/mman.h>
491 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
492 #define MAP_ANONYMOUS MAP_ANON
495 #endif /* HAVE_MMAP */
498 Access to system page size. To the extent possible, this malloc
499 manages memory from the system in page-size units.
501 The following mechanics for getpagesize were adapted from
502 bsd/gnu getpagesize.h
505 #ifndef LACKS_UNISTD_H
509 #ifndef malloc_getpagesize
510 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
511 # ifndef _SC_PAGE_SIZE
512 # define _SC_PAGE_SIZE _SC_PAGESIZE
515 # ifdef _SC_PAGE_SIZE
516 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
518 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
519 extern size_t getpagesize();
520 # define malloc_getpagesize getpagesize()
523 # define malloc_getpagesize (4096) /* TBD: Use 'GetSystemInfo' instead */
525 # ifndef LACKS_SYS_PARAM_H
526 # include <sys/param.h>
528 # ifdef EXEC_PAGESIZE
529 # define malloc_getpagesize EXEC_PAGESIZE
533 # define malloc_getpagesize NBPG
535 # define malloc_getpagesize (NBPG * CLSIZE)
539 # define malloc_getpagesize NBPC
542 # define malloc_getpagesize PAGESIZE
544 # define malloc_getpagesize (4096) /* just guess */
557 This version of malloc supports the standard SVID/XPG mallinfo
558 routine that returns a struct containing the same kind of
559 information you can get from malloc_stats. It should work on
560 any SVID/XPG compliant system that has a /usr/include/malloc.h
561 defining struct mallinfo. (If you'd like to install such a thing
562 yourself, cut out the preliminary declarations as described above
563 and below and save them in a malloc.h file. But there's no
564 compelling reason to bother to do this.)
566 The main declaration needed is the mallinfo struct that is returned
567 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
568 bunch of fields, most of which are not even meaningful in this
569 version of malloc. Some of these fields are are instead filled by
570 mallinfo() with other numbers that might possibly be of interest.
572 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
573 /usr/include/malloc.h file that includes a declaration of struct
574 mallinfo. If so, it is included; else an SVID2/XPG2 compliant
575 version is declared below. These must be precisely the same for
580 /* #define HAVE_USR_INCLUDE_MALLOC_H */
582 #if HAVE_USR_INCLUDE_MALLOC_H
583 #include "/usr/include/malloc.h"
586 /* SVID2/XPG mallinfo structure */
589 int arena; /* total space allocated from system */
590 int ordblks; /* number of non-inuse chunks */
591 int smblks; /* unused -- always zero */
592 int hblks; /* number of mmapped regions */
593 int hblkhd; /* total space in mmapped regions */
594 int usmblks; /* unused -- always zero */
595 int fsmblks; /* unused -- always zero */
596 int uordblks; /* total allocated space */
597 int fordblks; /* total non-inuse space */
598 int keepcost; /* top-most, releasable (via malloc_trim) space */
601 /* SVID2/XPG mallopt options */
603 #define M_MXFAST 1 /* UNUSED in this malloc */
604 #define M_NLBLKS 2 /* UNUSED in this malloc */
605 #define M_GRAIN 3 /* UNUSED in this malloc */
606 #define M_KEEP 4 /* UNUSED in this malloc */
610 /* mallopt options that actually do something */
612 #define M_TRIM_THRESHOLD -1
614 #define M_MMAP_THRESHOLD -3
615 #define M_MMAP_MAX -4
618 #ifndef DEFAULT_TRIM_THRESHOLD
619 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
623 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
624 to keep before releasing via malloc_trim in free().
626 Automatic trimming is mainly useful in long-lived programs.
627 Because trimming via sbrk can be slow on some systems, and can
628 sometimes be wasteful (in cases where programs immediately
629 afterward allocate more large chunks) the value should be high
630 enough so that your overall system performance would improve by
633 The trim threshold and the mmap control parameters (see below)
634 can be traded off with one another. Trimming and mmapping are
635 two different ways of releasing unused memory back to the
636 system. Between these two, it is often possible to keep
637 system-level demands of a long-lived program down to a bare
638 minimum. For example, in one test suite of sessions measuring
639 the XF86 X server on Linux, using a trim threshold of 128K and a
640 mmap threshold of 192K led to near-minimal long term resource
643 If you are using this malloc in a long-lived program, it should
644 pay to experiment with these values. As a rough guide, you
645 might set to a value close to the average size of a process
646 (program) running on your system. Releasing this much memory
647 would allow such a process to run in memory. Generally, it's
648 worth it to tune for trimming rather tham memory mapping when a
649 program undergoes phases where several large chunks are
650 allocated and released in ways that can reuse each other's
651 storage, perhaps mixed with phases where there are no such
652 chunks at all. And in well-behaved long-lived programs,
653 controlling release of large blocks via trimming versus mapping
656 However, in most programs, these parameters serve mainly as
657 protection against the system-level effects of carrying around
658 massive amounts of unneeded memory. Since frequent calls to
659 sbrk, mmap, and munmap otherwise degrade performance, the default
660 parameters are set to relatively high values that serve only as
663 The default trim value is high enough to cause trimming only in
664 fairly extreme (by current memory consumption standards) cases.
665 It must be greater than page size to have any useful effect. To
666 disable trimming completely, you can set to (unsigned long)(-1);
672 #ifndef DEFAULT_TOP_PAD
673 #define DEFAULT_TOP_PAD (0)
677 M_TOP_PAD is the amount of extra `padding' space to allocate or
678 retain whenever sbrk is called. It is used in two ways internally:
680 * When sbrk is called to extend the top of the arena to satisfy
681 a new malloc request, this much padding is added to the sbrk
684 * When malloc_trim is called automatically from free(),
685 it is used as the `pad' argument.
687 In both cases, the actual amount of padding is rounded
688 so that the end of the arena is always a system page boundary.
690 The main reason for using padding is to avoid calling sbrk so
691 often. Having even a small pad greatly reduces the likelihood
692 that nearly every malloc request during program start-up (or
693 after trimming) will invoke sbrk, which needlessly wastes
696 Automatic rounding-up to page-size units is normally sufficient
697 to avoid measurable overhead, so the default is 0. However, in
698 systems where sbrk is relatively slow, it can pay to increase
699 this value, at the expense of carrying around more memory than
705 #ifndef DEFAULT_MMAP_THRESHOLD
706 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
711 M_MMAP_THRESHOLD is the request size threshold for using mmap()
712 to service a request. Requests of at least this size that cannot
713 be allocated using already-existing space will be serviced via mmap.
714 (If enough normal freed space already exists it is used instead.)
716 Using mmap segregates relatively large chunks of memory so that
717 they can be individually obtained and released from the host
718 system. A request serviced through mmap is never reused by any
719 other request (at least not directly; the system may just so
720 happen to remap successive requests to the same locations).
722 Segregating space in this way has the benefit that mmapped space
723 can ALWAYS be individually released back to the system, which
724 helps keep the system level memory demands of a long-lived
725 program low. Mapped memory can never become `locked' between
726 other chunks, as can happen with normally allocated chunks, which
727 menas that even trimming via malloc_trim would not release them.
729 However, it has the disadvantages that:
731 1. The space cannot be reclaimed, consolidated, and then
732 used to service later requests, as happens with normal chunks.
733 2. It can lead to more wastage because of mmap page alignment
735 3. It causes malloc performance to be more dependent on host
736 system memory management support routines which may vary in
737 implementation quality and may impose arbitrary
738 limitations. Generally, servicing a request via normal
739 malloc steps is faster than going through a system's mmap.
741 All together, these considerations should lead you to use mmap
742 only for relatively large requests.
748 #ifndef DEFAULT_MMAP_MAX
750 #define DEFAULT_MMAP_MAX (64)
752 #define DEFAULT_MMAP_MAX (0)
757 M_MMAP_MAX is the maximum number of requests to simultaneously
758 service using mmap. This parameter exists because:
760 1. Some systems have a limited number of internal tables for
762 2. In most systems, overreliance on mmap can degrade overall
764 3. If a program allocates many large regions, it is probably
765 better off using normal sbrk-based allocation routines that
766 can reclaim and reallocate normal heap memory. Using a
767 small value allows transition into this mode after the
768 first few allocations.
770 Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
771 the default value is 0, and attempts to set it to non-zero values
772 in mallopt will fail.
777 USE_DL_PREFIX will prefix all public routines with the string 'dl'.
778 Useful to quickly avoid procedure declaration conflicts and linker
779 symbol conflicts with existing memory allocation routines.
783 /* #define USE_DL_PREFIX */
788 Special defines for linux libc
790 Except when compiled using these special defines for Linux libc
791 using weak aliases, this malloc is NOT designed to work in
792 multithreaded applications. No semaphores or other concurrency
793 control are provided to ensure that multiple malloc or free calls
794 don't run at the same time, which could be disasterous. A single
795 semaphore could be used across malloc, realloc, and free (which is
796 essentially the effect of the linux weak alias approach). It would
797 be hard to obtain finer granularity.
802 #ifdef INTERNAL_LINUX_C_LIB
806 Void_t * __default_morecore_init (ptrdiff_t);
807 Void_t *(*__morecore)(ptrdiff_t) = __default_morecore_init;
811 Void_t * __default_morecore_init ();
812 Void_t *(*__morecore)() = __default_morecore_init;
816 #define MORECORE (*__morecore)
817 #define MORECORE_FAILURE 0
818 #define MORECORE_CLEARS 1
820 #else /* INTERNAL_LINUX_C_LIB */
823 extern Void_t* sbrk(ptrdiff_t);
825 extern Void_t* sbrk();
829 #define MORECORE sbrk
832 #ifndef MORECORE_FAILURE
833 #define MORECORE_FAILURE -1
836 #ifndef MORECORE_CLEARS
837 #define MORECORE_CLEARS 1
840 #endif /* INTERNAL_LINUX_C_LIB */
842 #if defined(INTERNAL_LINUX_C_LIB) && defined(__ELF__)
844 #define cALLOc __libc_calloc
845 #define fREe __libc_free
846 #define mALLOc __libc_malloc
847 #define mEMALIGn __libc_memalign
848 #define rEALLOc __libc_realloc
849 #define vALLOc __libc_valloc
850 #define pvALLOc __libc_pvalloc
851 #define mALLINFo __libc_mallinfo
852 #define mALLOPt __libc_mallopt
854 #pragma weak calloc = __libc_calloc
855 #pragma weak free = __libc_free
856 #pragma weak cfree = __libc_free
857 #pragma weak malloc = __libc_malloc
858 #pragma weak memalign = __libc_memalign
859 #pragma weak realloc = __libc_realloc
860 #pragma weak valloc = __libc_valloc
861 #pragma weak pvalloc = __libc_pvalloc
862 #pragma weak mallinfo = __libc_mallinfo
863 #pragma weak mallopt = __libc_mallopt
868 #define cALLOc dlcalloc
870 #define mALLOc dlmalloc
871 #define mEMALIGn dlmemalign
872 #define rEALLOc dlrealloc
873 #define vALLOc dlvalloc
874 #define pvALLOc dlpvalloc
875 #define mALLINFo dlmallinfo
876 #define mALLOPt dlmallopt
877 #else /* USE_DL_PREFIX */
878 #define cALLOc calloc
880 #define mALLOc malloc
881 #define mEMALIGn memalign
882 #define rEALLOc realloc
883 #define vALLOc valloc
884 #define pvALLOc pvalloc
885 #define mALLINFo mallinfo
886 #define mALLOPt mallopt
887 #endif /* USE_DL_PREFIX */
891 /* Public routines */
895 Void_t* mALLOc(size_t);
897 Void_t* rEALLOc(Void_t*, size_t);
898 Void_t* mEMALIGn(size_t, size_t);
899 Void_t* vALLOc(size_t);
900 Void_t* pvALLOc(size_t);
901 Void_t* cALLOc(size_t, size_t);
903 int malloc_trim(size_t);
904 size_t malloc_usable_size(Void_t*);
906 int mALLOPt(int, int);
907 struct mallinfo mALLINFo(void);
918 size_t malloc_usable_size();
921 struct mallinfo mALLINFo();
926 }; /* end of extern "C" */
929 /* ---------- To make a malloc.h, end cutting here ------------ */
930 #endif /* 0 */ /* Moved to malloc.h */
935 static void malloc_update_mallinfo (void);
936 void malloc_stats (void);
938 static void malloc_update_mallinfo ();
943 DECLARE_GLOBAL_DATA_PTR;
946 Emulation of sbrk for WIN32
947 All code within the ifdef WIN32 is untested by me.
949 Thanks to Martin Fong and others for supplying this.
955 #define AlignPage(add) (((add) + (malloc_getpagesize-1)) & \
956 ~(malloc_getpagesize-1))
957 #define AlignPage64K(add) (((add) + (0x10000 - 1)) & ~(0x10000 - 1))
959 /* resrve 64MB to insure large contiguous space */
960 #define RESERVED_SIZE (1024*1024*64)
961 #define NEXT_SIZE (2048*1024)
962 #define TOP_MEMORY ((unsigned long)2*1024*1024*1024)
964 struct GmListElement;
965 typedef struct GmListElement GmListElement;
973 static GmListElement* head = 0;
974 static unsigned int gNextAddress = 0;
975 static unsigned int gAddressBase = 0;
976 static unsigned int gAllocatedSize = 0;
979 GmListElement* makeGmListElement (void* bas)
982 this = (GmListElement*)(void*)LocalAlloc (0, sizeof (GmListElement));
996 assert ( (head == NULL) || (head->base == (void*)gAddressBase));
997 if (gAddressBase && (gNextAddress - gAddressBase))
999 rval = VirtualFree ((void*)gAddressBase,
1000 gNextAddress - gAddressBase,
1006 GmListElement* next = head->next;
1007 rval = VirtualFree (head->base, 0, MEM_RELEASE);
1015 void* findRegion (void* start_address, unsigned long size)
1017 MEMORY_BASIC_INFORMATION info;
1018 if (size >= TOP_MEMORY) return NULL;
1020 while ((unsigned long)start_address + size < TOP_MEMORY)
1022 VirtualQuery (start_address, &info, sizeof (info));
1023 if ((info.State == MEM_FREE) && (info.RegionSize >= size))
1024 return start_address;
1027 /* Requested region is not available so see if the */
1028 /* next region is available. Set 'start_address' */
1029 /* to the next region and call 'VirtualQuery()' */
1032 start_address = (char*)info.BaseAddress + info.RegionSize;
1034 /* Make sure we start looking for the next region */
1035 /* on the *next* 64K boundary. Otherwise, even if */
1036 /* the new region is free according to */
1037 /* 'VirtualQuery()', the subsequent call to */
1038 /* 'VirtualAlloc()' (which follows the call to */
1039 /* this routine in 'wsbrk()') will round *down* */
1040 /* the requested address to a 64K boundary which */
1041 /* we already know is an address in the */
1042 /* unavailable region. Thus, the subsequent call */
1043 /* to 'VirtualAlloc()' will fail and bring us back */
1044 /* here, causing us to go into an infinite loop. */
1047 (void *) AlignPage64K((unsigned long) start_address);
1055 void* wsbrk (long size)
1060 if (gAddressBase == 0)
1062 gAllocatedSize = max (RESERVED_SIZE, AlignPage (size));
1063 gNextAddress = gAddressBase =
1064 (unsigned int)VirtualAlloc (NULL, gAllocatedSize,
1065 MEM_RESERVE, PAGE_NOACCESS);
1066 } else if (AlignPage (gNextAddress + size) > (gAddressBase +
1069 long new_size = max (NEXT_SIZE, AlignPage (size));
1070 void* new_address = (void*)(gAddressBase+gAllocatedSize);
1073 new_address = findRegion (new_address, new_size);
1075 if (new_address == 0)
1078 gAddressBase = gNextAddress =
1079 (unsigned int)VirtualAlloc (new_address, new_size,
1080 MEM_RESERVE, PAGE_NOACCESS);
1081 /* repeat in case of race condition */
1082 /* The region that we found has been snagged */
1083 /* by another thread */
1085 while (gAddressBase == 0);
1087 assert (new_address == (void*)gAddressBase);
1089 gAllocatedSize = new_size;
1091 if (!makeGmListElement ((void*)gAddressBase))
1094 if ((size + gNextAddress) > AlignPage (gNextAddress))
1097 res = VirtualAlloc ((void*)AlignPage (gNextAddress),
1098 (size + gNextAddress -
1099 AlignPage (gNextAddress)),
1100 MEM_COMMIT, PAGE_READWRITE);
1104 tmp = (void*)gNextAddress;
1105 gNextAddress = (unsigned int)tmp + size;
1110 unsigned int alignedGoal = AlignPage (gNextAddress + size);
1111 /* Trim by releasing the virtual memory */
1112 if (alignedGoal >= gAddressBase)
1114 VirtualFree ((void*)alignedGoal, gNextAddress - alignedGoal,
1116 gNextAddress = gNextAddress + size;
1117 return (void*)gNextAddress;
1121 VirtualFree ((void*)gAddressBase, gNextAddress - gAddressBase,
1123 gNextAddress = gAddressBase;
1129 return (void*)gNextAddress;
1144 INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
1145 INTERNAL_SIZE_T size; /* Size in bytes, including overhead. */
1146 struct malloc_chunk* fd; /* double links -- used only if free. */
1147 struct malloc_chunk* bk;
1148 } __attribute__((__may_alias__)) ;
1150 typedef struct malloc_chunk* mchunkptr;
1154 malloc_chunk details:
1156 (The following includes lightly edited explanations by Colin Plumb.)
1158 Chunks of memory are maintained using a `boundary tag' method as
1159 described in e.g., Knuth or Standish. (See the paper by Paul
1160 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1161 survey of such techniques.) Sizes of free chunks are stored both
1162 in the front of each chunk and at the end. This makes
1163 consolidating fragmented chunks into bigger chunks very fast. The
1164 size fields also hold bits representing whether chunks are free or
1167 An allocated chunk looks like this:
1170 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1171 | Size of previous chunk, if allocated | |
1172 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1173 | Size of chunk, in bytes |P|
1174 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1175 | User data starts here... .
1177 . (malloc_usable_space() bytes) .
1179 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1181 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1184 Where "chunk" is the front of the chunk for the purpose of most of
1185 the malloc code, but "mem" is the pointer that is returned to the
1186 user. "Nextchunk" is the beginning of the next contiguous chunk.
1188 Chunks always begin on even word boundries, so the mem portion
1189 (which is returned to the user) is also on an even word boundary, and
1190 thus double-word aligned.
1192 Free chunks are stored in circular doubly-linked lists, and look like this:
1194 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1195 | Size of previous chunk |
1196 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1197 `head:' | Size of chunk, in bytes |P|
1198 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1199 | Forward pointer to next chunk in list |
1200 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1201 | Back pointer to previous chunk in list |
1202 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1203 | Unused space (may be 0 bytes long) .
1206 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1207 `foot:' | Size of chunk, in bytes |
1208 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1210 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1211 chunk size (which is always a multiple of two words), is an in-use
1212 bit for the *previous* chunk. If that bit is *clear*, then the
1213 word before the current chunk size contains the previous chunk
1214 size, and can be used to find the front of the previous chunk.
1215 (The very first chunk allocated always has this bit set,
1216 preventing access to non-existent (or non-owned) memory.)
1218 Note that the `foot' of the current chunk is actually represented
1219 as the prev_size of the NEXT chunk. (This makes it easier to
1220 deal with alignments etc).
1222 The two exceptions to all this are
1224 1. The special chunk `top', which doesn't bother using the
1225 trailing size field since there is no
1226 next contiguous chunk that would have to index off it. (After
1227 initialization, `top' is forced to always exist. If it would
1228 become less than MINSIZE bytes long, it is replenished via
1231 2. Chunks allocated via mmap, which have the second-lowest-order
1232 bit (IS_MMAPPED) set in their size fields. Because they are
1233 never merged or traversed from any other chunk, they have no
1234 foot size or inuse information.
1236 Available chunks are kept in any of several places (all declared below):
1238 * `av': An array of chunks serving as bin headers for consolidated
1239 chunks. Each bin is doubly linked. The bins are approximately
1240 proportionally (log) spaced. There are a lot of these bins
1241 (128). This may look excessive, but works very well in
1242 practice. All procedures maintain the invariant that no
1243 consolidated chunk physically borders another one. Chunks in
1244 bins are kept in size order, with ties going to the
1245 approximately least recently used chunk.
1247 The chunks in each bin are maintained in decreasing sorted order by
1248 size. This is irrelevant for the small bins, which all contain
1249 the same-sized chunks, but facilitates best-fit allocation for
1250 larger chunks. (These lists are just sequential. Keeping them in
1251 order almost never requires enough traversal to warrant using
1252 fancier ordered data structures.) Chunks of the same size are
1253 linked with the most recently freed at the front, and allocations
1254 are taken from the back. This results in LRU or FIFO allocation
1255 order, which tends to give each chunk an equal opportunity to be
1256 consolidated with adjacent freed chunks, resulting in larger free
1257 chunks and less fragmentation.
1259 * `top': The top-most available chunk (i.e., the one bordering the
1260 end of available memory) is treated specially. It is never
1261 included in any bin, is used only if no other chunk is
1262 available, and is released back to the system if it is very
1263 large (see M_TRIM_THRESHOLD).
1265 * `last_remainder': A bin holding only the remainder of the
1266 most recently split (non-top) chunk. This bin is checked
1267 before other non-fitting chunks, so as to provide better
1268 locality for runs of sequentially allocated chunks.
1270 * Implicitly, through the host system's memory mapping tables.
1271 If supported, requests greater than a threshold are usually
1272 serviced via calls to mmap, and then later released via munmap.
1276 /* sizes, alignments */
1278 #define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
1279 #define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
1280 #define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
1281 #define MINSIZE (sizeof(struct malloc_chunk))
1283 /* conversion from malloc headers to user pointers, and back */
1285 #define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1286 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1288 /* pad request bytes into a usable size */
1290 #define request2size(req) \
1291 (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1292 (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1293 (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1295 /* Check if m has acceptable alignment */
1297 #define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1303 Physical chunk operations
1307 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1309 #define PREV_INUSE 0x1
1311 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1313 #define IS_MMAPPED 0x2
1315 /* Bits to mask off when extracting size */
1317 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1320 /* Ptr to next physical malloc_chunk. */
1322 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1324 /* Ptr to previous physical malloc_chunk */
1326 #define prev_chunk(p)\
1327 ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1330 /* Treat space at ptr + offset as a chunk */
1332 #define chunk_at_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
1338 Dealing with use bits
1341 /* extract p's inuse bit */
1344 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1346 /* extract inuse bit of previous chunk */
1348 #define prev_inuse(p) ((p)->size & PREV_INUSE)
1350 /* check for mmap()'ed chunk */
1352 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1354 /* set/clear chunk as in use without otherwise disturbing */
1356 #define set_inuse(p)\
1357 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1359 #define clear_inuse(p)\
1360 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1362 /* check/set/clear inuse bits in known places */
1364 #define inuse_bit_at_offset(p, s)\
1365 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1367 #define set_inuse_bit_at_offset(p, s)\
1368 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1370 #define clear_inuse_bit_at_offset(p, s)\
1371 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1377 Dealing with size fields
1380 /* Get size, ignoring use bits */
1382 #define chunksize(p) ((p)->size & ~(SIZE_BITS))
1384 /* Set size at head, without disturbing its use bit */
1386 #define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1388 /* Set size/use ignoring previous bits in header */
1390 #define set_head(p, s) ((p)->size = (s))
1392 /* Set size at footer (only when chunk is not in use) */
1394 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1403 The bins, `av_' are an array of pairs of pointers serving as the
1404 heads of (initially empty) doubly-linked lists of chunks, laid out
1405 in a way so that each pair can be treated as if it were in a
1406 malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1407 and chunks are the same).
1409 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1410 8 bytes apart. Larger bins are approximately logarithmically
1411 spaced. (See the table below.) The `av_' array is never mentioned
1412 directly in the code, but instead via bin access macros.
1420 4 bins of size 32768
1421 2 bins of size 262144
1422 1 bin of size what's left
1424 There is actually a little bit of slop in the numbers in bin_index
1425 for the sake of speed. This makes no difference elsewhere.
1427 The special chunks `top' and `last_remainder' get their own bins,
1428 (this is implemented via yet more trickery with the av_ array),
1429 although `top' is never properly linked to its bin since it is
1430 always handled specially.
1434 #define NAV 128 /* number of bins */
1436 typedef struct malloc_chunk* mbinptr;
1440 #define bin_at(i) ((mbinptr)((char*)&(av_[2*(i) + 2]) - 2*SIZE_SZ))
1441 #define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1442 #define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1445 The first 2 bins are never indexed. The corresponding av_ cells are instead
1446 used for bookkeeping. This is not to save space, but to simplify
1447 indexing, maintain locality, and avoid some initialization tests.
1450 #define top (av_[2]) /* The topmost chunk */
1451 #define last_remainder (bin_at(1)) /* remainder from last split */
1455 Because top initially points to its own bin with initial
1456 zero size, thus forcing extension on the first malloc request,
1457 we avoid having any special code in malloc to check whether
1458 it even exists yet. But we still need to in malloc_extend_top.
1461 #define initial_top ((mchunkptr)(bin_at(0)))
1463 /* Helper macro to initialize bins */
1465 #define IAV(i) bin_at(i), bin_at(i)
1467 static mbinptr av_[NAV * 2 + 2] = {
1469 IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
1470 IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
1471 IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
1472 IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
1473 IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
1474 IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
1475 IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
1476 IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
1477 IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
1478 IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
1479 IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
1480 IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
1481 IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
1482 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1483 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1484 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1487 #ifdef CONFIG_NEEDS_MANUAL_RELOC
1488 static void malloc_bin_reloc(void)
1490 mbinptr *p = &av_[2];
1493 for (i = 2; i < ARRAY_SIZE(av_); ++i, ++p)
1494 *p = (mbinptr)((ulong)*p + gd->reloc_off);
1497 static inline void malloc_bin_reloc(void) {}
1500 ulong mem_malloc_start = 0;
1501 ulong mem_malloc_end = 0;
1502 ulong mem_malloc_brk = 0;
1504 void *sbrk(ptrdiff_t increment)
1506 ulong old = mem_malloc_brk;
1507 ulong new = old + increment;
1510 * if we are giving memory back make sure we clear it out since
1511 * we set MORECORE_CLEARS to 1
1514 memset((void *)new, 0, -increment);
1516 if ((new < mem_malloc_start) || (new > mem_malloc_end))
1517 return (void *)MORECORE_FAILURE;
1519 mem_malloc_brk = new;
1524 void mem_malloc_init(ulong start, ulong size)
1526 mem_malloc_start = start;
1527 mem_malloc_end = start + size;
1528 mem_malloc_brk = start;
1530 memset((void *)mem_malloc_start, 0, size);
1535 /* field-extraction macros */
1537 #define first(b) ((b)->fd)
1538 #define last(b) ((b)->bk)
1544 #define bin_index(sz) \
1545 (((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3): \
1546 ((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6): \
1547 ((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9): \
1548 ((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12): \
1549 ((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15): \
1550 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1553 bins for chunks < 512 are all spaced 8 bytes apart, and hold
1554 identically sized chunks. This is exploited in malloc.
1557 #define MAX_SMALLBIN 63
1558 #define MAX_SMALLBIN_SIZE 512
1559 #define SMALLBIN_WIDTH 8
1561 #define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
1564 Requests are `small' if both the corresponding and the next bin are small
1567 #define is_small_request(nb) (nb < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1572 To help compensate for the large number of bins, a one-level index
1573 structure is used for bin-by-bin searching. `binblocks' is a
1574 one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1575 have any (possibly) non-empty bins, so they can be skipped over
1576 all at once during during traversals. The bits are NOT always
1577 cleared as soon as all bins in a block are empty, but instead only
1578 when all are noticed to be empty during traversal in malloc.
1581 #define BINBLOCKWIDTH 4 /* bins per block */
1583 #define binblocks_r ((INTERNAL_SIZE_T)av_[1]) /* bitvector of nonempty blocks */
1584 #define binblocks_w (av_[1])
1586 /* bin<->block macros */
1588 #define idx2binblock(ix) ((unsigned)1 << (ix / BINBLOCKWIDTH))
1589 #define mark_binblock(ii) (binblocks_w = (mbinptr)(binblocks_r | idx2binblock(ii)))
1590 #define clear_binblock(ii) (binblocks_w = (mbinptr)(binblocks_r & ~(idx2binblock(ii))))
1596 /* Other static bookkeeping data */
1598 /* variables holding tunable values */
1600 static unsigned long trim_threshold = DEFAULT_TRIM_THRESHOLD;
1601 static unsigned long top_pad = DEFAULT_TOP_PAD;
1602 static unsigned int n_mmaps_max = DEFAULT_MMAP_MAX;
1603 static unsigned long mmap_threshold = DEFAULT_MMAP_THRESHOLD;
1605 /* The first value returned from sbrk */
1606 static char* sbrk_base = (char*)(-1);
1608 /* The maximum memory obtained from system via sbrk */
1609 static unsigned long max_sbrked_mem = 0;
1611 /* The maximum via either sbrk or mmap */
1612 static unsigned long max_total_mem = 0;
1614 /* internal working copy of mallinfo */
1615 static struct mallinfo current_mallinfo = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
1617 /* The total memory obtained from system via sbrk */
1618 #define sbrked_mem (current_mallinfo.arena)
1620 /* Tracking mmaps */
1623 static unsigned int n_mmaps = 0;
1625 static unsigned long mmapped_mem = 0;
1627 static unsigned int max_n_mmaps = 0;
1628 static unsigned long max_mmapped_mem = 0;
1641 These routines make a number of assertions about the states
1642 of data structures that should be true at all times. If any
1643 are not true, it's very likely that a user program has somehow
1644 trashed memory. (It's also possible that there is a coding error
1645 in malloc. In which case, please report it!)
1649 static void do_check_chunk(mchunkptr p)
1651 static void do_check_chunk(p) mchunkptr p;
1654 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1656 /* No checkable chunk is mmapped */
1657 assert(!chunk_is_mmapped(p));
1659 /* Check for legal address ... */
1660 assert((char*)p >= sbrk_base);
1662 assert((char*)p + sz <= (char*)top);
1664 assert((char*)p + sz <= sbrk_base + sbrked_mem);
1670 static void do_check_free_chunk(mchunkptr p)
1672 static void do_check_free_chunk(p) mchunkptr p;
1675 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1676 mchunkptr next = chunk_at_offset(p, sz);
1680 /* Check whether it claims to be free ... */
1683 /* Unless a special marker, must have OK fields */
1684 if ((long)sz >= (long)MINSIZE)
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));
1694 /* ... and has minimally sane links */
1695 assert(p->fd->bk == p);
1696 assert(p->bk->fd == p);
1698 else /* markers are always of size SIZE_SZ */
1699 assert(sz == SIZE_SZ);
1703 static void do_check_inuse_chunk(mchunkptr p)
1705 static void do_check_inuse_chunk(p) mchunkptr p;
1708 mchunkptr next = next_chunk(p);
1711 /* Check whether it claims to be in use ... */
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.
1720 mchunkptr prv = prev_chunk(p);
1721 assert(next_chunk(prv) == p);
1722 do_check_free_chunk(prv);
1726 assert(prev_inuse(next));
1727 assert(chunksize(next) >= MINSIZE);
1729 else if (!inuse(next))
1730 do_check_free_chunk(next);
1735 static void do_check_malloced_chunk(mchunkptr p, INTERNAL_SIZE_T s)
1737 static void do_check_malloced_chunk(p, s) mchunkptr p; INTERNAL_SIZE_T s;
1740 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1743 do_check_inuse_chunk(p);
1745 /* Legal size ... */
1746 assert((long)sz >= (long)MINSIZE);
1747 assert((sz & MALLOC_ALIGN_MASK) == 0);
1749 assert(room < (long)MINSIZE);
1751 /* ... and alignment */
1752 assert(aligned_OK(chunk2mem(p)));
1755 /* ... and was allocated at front of an available chunk */
1756 assert(prev_inuse(p));
1761 #define check_free_chunk(P) do_check_free_chunk(P)
1762 #define check_inuse_chunk(P) do_check_inuse_chunk(P)
1763 #define check_chunk(P) do_check_chunk(P)
1764 #define check_malloced_chunk(P,N) do_check_malloced_chunk(P,N)
1766 #define check_free_chunk(P)
1767 #define check_inuse_chunk(P)
1768 #define check_chunk(P)
1769 #define check_malloced_chunk(P,N)
1775 Macro-based internal utilities
1780 Linking chunks in bin lists.
1781 Call these only with variables, not arbitrary expressions, as arguments.
1785 Place chunk p of size s in its bin, in size order,
1786 putting it ahead of others of same size.
1790 #define frontlink(P, S, IDX, BK, FD) \
1792 if (S < MAX_SMALLBIN_SIZE) \
1794 IDX = smallbin_index(S); \
1795 mark_binblock(IDX); \
1800 FD->bk = BK->fd = P; \
1804 IDX = bin_index(S); \
1807 if (FD == BK) mark_binblock(IDX); \
1810 while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
1815 FD->bk = BK->fd = P; \
1820 /* take a chunk off a list */
1822 #define unlink(P, BK, FD) \
1830 /* Place p as the last remainder */
1832 #define link_last_remainder(P) \
1834 last_remainder->fd = last_remainder->bk = P; \
1835 P->fd = P->bk = last_remainder; \
1838 /* Clear the last_remainder bin */
1840 #define clear_last_remainder \
1841 (last_remainder->fd = last_remainder->bk = last_remainder)
1847 /* Routines dealing with mmap(). */
1852 static mchunkptr mmap_chunk(size_t size)
1854 static mchunkptr mmap_chunk(size) size_t size;
1857 size_t page_mask = malloc_getpagesize - 1;
1860 #ifndef MAP_ANONYMOUS
1864 if(n_mmaps >= n_mmaps_max) return 0; /* too many regions */
1866 /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1867 * there is no following chunk whose prev_size field could be used.
1869 size = (size + SIZE_SZ + page_mask) & ~page_mask;
1871 #ifdef MAP_ANONYMOUS
1872 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE,
1873 MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
1874 #else /* !MAP_ANONYMOUS */
1877 fd = open("/dev/zero", O_RDWR);
1878 if(fd < 0) return 0;
1880 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
1883 if(p == (mchunkptr)-1) return 0;
1886 if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
1888 /* We demand that eight bytes into a page must be 8-byte aligned. */
1889 assert(aligned_OK(chunk2mem(p)));
1891 /* The offset to the start of the mmapped region is stored
1892 * in the prev_size field of the chunk; normally it is zero,
1893 * but that can be changed in memalign().
1896 set_head(p, size|IS_MMAPPED);
1898 mmapped_mem += size;
1899 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1900 max_mmapped_mem = mmapped_mem;
1901 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1902 max_total_mem = mmapped_mem + sbrked_mem;
1907 static void munmap_chunk(mchunkptr p)
1909 static void munmap_chunk(p) mchunkptr p;
1912 INTERNAL_SIZE_T size = chunksize(p);
1915 assert (chunk_is_mmapped(p));
1916 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1917 assert((n_mmaps > 0));
1918 assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
1921 mmapped_mem -= (size + p->prev_size);
1923 ret = munmap((char *)p - p->prev_size, size + p->prev_size);
1925 /* munmap returns non-zero on failure */
1932 static mchunkptr mremap_chunk(mchunkptr p, size_t new_size)
1934 static mchunkptr mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
1937 size_t page_mask = malloc_getpagesize - 1;
1938 INTERNAL_SIZE_T offset = p->prev_size;
1939 INTERNAL_SIZE_T size = chunksize(p);
1942 assert (chunk_is_mmapped(p));
1943 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1944 assert((n_mmaps > 0));
1945 assert(((size + offset) & (malloc_getpagesize-1)) == 0);
1947 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1948 new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
1950 cp = (char *)mremap((char *)p - offset, size + offset, new_size, 1);
1952 if (cp == (char *)-1) return 0;
1954 p = (mchunkptr)(cp + offset);
1956 assert(aligned_OK(chunk2mem(p)));
1958 assert((p->prev_size == offset));
1959 set_head(p, (new_size - offset)|IS_MMAPPED);
1961 mmapped_mem -= size + offset;
1962 mmapped_mem += new_size;
1963 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1964 max_mmapped_mem = mmapped_mem;
1965 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1966 max_total_mem = mmapped_mem + sbrked_mem;
1970 #endif /* HAVE_MREMAP */
1972 #endif /* HAVE_MMAP */
1978 Extend the top-most chunk by obtaining memory from system.
1979 Main interface to sbrk (but see also malloc_trim).
1983 static void malloc_extend_top(INTERNAL_SIZE_T nb)
1985 static void malloc_extend_top(nb) INTERNAL_SIZE_T nb;
1988 char* brk; /* return value from sbrk */
1989 INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
1990 INTERNAL_SIZE_T correction; /* bytes for 2nd sbrk call */
1991 char* new_brk; /* return of 2nd sbrk call */
1992 INTERNAL_SIZE_T top_size; /* new size of top chunk */
1994 mchunkptr old_top = top; /* Record state of old top */
1995 INTERNAL_SIZE_T old_top_size = chunksize(old_top);
1996 char* old_end = (char*)(chunk_at_offset(old_top, old_top_size));
1998 /* Pad request with top_pad plus minimal overhead */
2000 INTERNAL_SIZE_T sbrk_size = nb + top_pad + MINSIZE;
2001 unsigned long pagesz = malloc_getpagesize;
2003 /* If not the first time through, round to preserve page boundary */
2004 /* Otherwise, we need to correct to a page size below anyway. */
2005 /* (We also correct below if an intervening foreign sbrk call.) */
2007 if (sbrk_base != (char*)(-1))
2008 sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
2010 brk = (char*)(MORECORE (sbrk_size));
2012 /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2013 if (brk == (char*)(MORECORE_FAILURE) ||
2014 (brk < old_end && old_top != initial_top))
2017 sbrked_mem += sbrk_size;
2019 if (brk == old_end) /* can just add bytes to current top */
2021 top_size = sbrk_size + old_top_size;
2022 set_head(top, top_size | PREV_INUSE);
2026 if (sbrk_base == (char*)(-1)) /* First time through. Record base */
2028 else /* Someone else called sbrk(). Count those bytes as sbrked_mem. */
2029 sbrked_mem += brk - (char*)old_end;
2031 /* Guarantee alignment of first new chunk made from this space */
2032 front_misalign = (unsigned long)chunk2mem(brk) & MALLOC_ALIGN_MASK;
2033 if (front_misalign > 0)
2035 correction = (MALLOC_ALIGNMENT) - front_misalign;
2041 /* Guarantee the next brk will be at a page boundary */
2043 correction += ((((unsigned long)(brk + sbrk_size))+(pagesz-1)) &
2044 ~(pagesz - 1)) - ((unsigned long)(brk + sbrk_size));
2046 /* Allocate correction */
2047 new_brk = (char*)(MORECORE (correction));
2048 if (new_brk == (char*)(MORECORE_FAILURE)) return;
2050 sbrked_mem += correction;
2052 top = (mchunkptr)brk;
2053 top_size = new_brk - brk + correction;
2054 set_head(top, top_size | PREV_INUSE);
2056 if (old_top != initial_top)
2059 /* There must have been an intervening foreign sbrk call. */
2060 /* A double fencepost is necessary to prevent consolidation */
2062 /* If not enough space to do this, then user did something very wrong */
2063 if (old_top_size < MINSIZE)
2065 set_head(top, PREV_INUSE); /* will force null return from malloc */
2069 /* Also keep size a multiple of MALLOC_ALIGNMENT */
2070 old_top_size = (old_top_size - 3*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2071 set_head_size(old_top, old_top_size);
2072 chunk_at_offset(old_top, old_top_size )->size =
2074 chunk_at_offset(old_top, old_top_size + SIZE_SZ)->size =
2076 /* If possible, release the rest. */
2077 if (old_top_size >= MINSIZE)
2078 fREe(chunk2mem(old_top));
2082 if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
2083 max_sbrked_mem = sbrked_mem;
2084 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2085 max_total_mem = mmapped_mem + sbrked_mem;
2087 /* We always land on a page boundary */
2088 assert(((unsigned long)((char*)top + top_size) & (pagesz - 1)) == 0);
2094 /* Main public routines */
2100 The requested size is first converted into a usable form, `nb'.
2101 This currently means to add 4 bytes overhead plus possibly more to
2102 obtain 8-byte alignment and/or to obtain a size of at least
2103 MINSIZE (currently 16 bytes), the smallest allocatable size.
2104 (All fits are considered `exact' if they are within MINSIZE bytes.)
2106 From there, the first successful of the following steps is taken:
2108 1. The bin corresponding to the request size is scanned, and if
2109 a chunk of exactly the right size is found, it is taken.
2111 2. The most recently remaindered chunk is used if it is big
2112 enough. This is a form of (roving) first fit, used only in
2113 the absence of exact fits. Runs of consecutive requests use
2114 the remainder of the chunk used for the previous such request
2115 whenever possible. This limited use of a first-fit style
2116 allocation strategy tends to give contiguous chunks
2117 coextensive lifetimes, which improves locality and can reduce
2118 fragmentation in the long run.
2120 3. Other bins are scanned in increasing size order, using a
2121 chunk big enough to fulfill the request, and splitting off
2122 any remainder. This search is strictly by best-fit; i.e.,
2123 the smallest (with ties going to approximately the least
2124 recently used) chunk that fits is selected.
2126 4. If large enough, the chunk bordering the end of memory
2127 (`top') is split off. (This use of `top' is in accord with
2128 the best-fit search rule. In effect, `top' is treated as
2129 larger (and thus less well fitting) than any other available
2130 chunk since it can be extended to be as large as necessary
2131 (up to system limitations).
2133 5. If the request size meets the mmap threshold and the
2134 system supports mmap, and there are few enough currently
2135 allocated mmapped regions, and a call to mmap succeeds,
2136 the request is allocated via direct memory mapping.
2138 6. Otherwise, the top of memory is extended by
2139 obtaining more space from the system (normally using sbrk,
2140 but definable to anything else via the MORECORE macro).
2141 Memory is gathered from the system (in system page-sized
2142 units) in a way that allows chunks obtained across different
2143 sbrk calls to be consolidated, but does not require
2144 contiguous memory. Thus, it should be safe to intersperse
2145 mallocs with other sbrk calls.
2148 All allocations are made from the the `lowest' part of any found
2149 chunk. (The implementation invariant is that prev_inuse is
2150 always true of any allocated chunk; i.e., that each allocated
2151 chunk borders either a previously allocated and still in-use chunk,
2152 or the base of its memory arena.)
2157 Void_t* mALLOc(size_t bytes)
2159 Void_t* mALLOc(bytes) size_t bytes;
2162 mchunkptr victim; /* inspected/selected chunk */
2163 INTERNAL_SIZE_T victim_size; /* its size */
2164 int idx; /* index for bin traversal */
2165 mbinptr bin; /* associated bin */
2166 mchunkptr remainder; /* remainder from a split */
2167 long remainder_size; /* its size */
2168 int remainder_index; /* its bin index */
2169 unsigned long block; /* block traverser bit */
2170 int startidx; /* first bin of a traversed block */
2171 mchunkptr fwd; /* misc temp for linking */
2172 mchunkptr bck; /* misc temp for linking */
2173 mbinptr q; /* misc temp */
2177 /* check if mem_malloc_init() was run */
2178 if ((mem_malloc_start == 0) && (mem_malloc_end == 0)) {
2179 /* not initialized yet */
2183 if ((long)bytes < 0) return NULL;
2185 nb = request2size(bytes); /* padded request size; */
2187 /* Check for exact match in a bin */
2189 if (is_small_request(nb)) /* Faster version for small requests */
2191 idx = smallbin_index(nb);
2193 /* No traversal or size check necessary for small bins. */
2198 /* Also scan the next one, since it would have a remainder < MINSIZE */
2206 victim_size = chunksize(victim);
2207 unlink(victim, bck, fwd);
2208 set_inuse_bit_at_offset(victim, victim_size);
2209 check_malloced_chunk(victim, nb);
2210 return chunk2mem(victim);
2213 idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
2218 idx = bin_index(nb);
2221 for (victim = last(bin); victim != bin; victim = victim->bk)
2223 victim_size = chunksize(victim);
2224 remainder_size = victim_size - nb;
2226 if (remainder_size >= (long)MINSIZE) /* too big */
2228 --idx; /* adjust to rescan below after checking last remainder */
2232 else if (remainder_size >= 0) /* exact fit */
2234 unlink(victim, bck, fwd);
2235 set_inuse_bit_at_offset(victim, victim_size);
2236 check_malloced_chunk(victim, nb);
2237 return chunk2mem(victim);
2245 /* Try to use the last split-off remainder */
2247 if ( (victim = last_remainder->fd) != last_remainder)
2249 victim_size = chunksize(victim);
2250 remainder_size = victim_size - nb;
2252 if (remainder_size >= (long)MINSIZE) /* re-split */
2254 remainder = chunk_at_offset(victim, nb);
2255 set_head(victim, nb | PREV_INUSE);
2256 link_last_remainder(remainder);
2257 set_head(remainder, remainder_size | PREV_INUSE);
2258 set_foot(remainder, remainder_size);
2259 check_malloced_chunk(victim, nb);
2260 return chunk2mem(victim);
2263 clear_last_remainder;
2265 if (remainder_size >= 0) /* exhaust */
2267 set_inuse_bit_at_offset(victim, victim_size);
2268 check_malloced_chunk(victim, nb);
2269 return chunk2mem(victim);
2272 /* Else place in bin */
2274 frontlink(victim, victim_size, remainder_index, bck, fwd);
2278 If there are any possibly nonempty big-enough blocks,
2279 search for best fitting chunk by scanning bins in blockwidth units.
2282 if ( (block = idx2binblock(idx)) <= binblocks_r)
2285 /* Get to the first marked block */
2287 if ( (block & binblocks_r) == 0)
2289 /* force to an even block boundary */
2290 idx = (idx & ~(BINBLOCKWIDTH - 1)) + BINBLOCKWIDTH;
2292 while ((block & binblocks_r) == 0)
2294 idx += BINBLOCKWIDTH;
2299 /* For each possibly nonempty block ... */
2302 startidx = idx; /* (track incomplete blocks) */
2303 q = bin = bin_at(idx);
2305 /* For each bin in this block ... */
2308 /* Find and use first big enough chunk ... */
2310 for (victim = last(bin); victim != bin; victim = victim->bk)
2312 victim_size = chunksize(victim);
2313 remainder_size = victim_size - nb;
2315 if (remainder_size >= (long)MINSIZE) /* split */
2317 remainder = chunk_at_offset(victim, nb);
2318 set_head(victim, nb | PREV_INUSE);
2319 unlink(victim, bck, fwd);
2320 link_last_remainder(remainder);
2321 set_head(remainder, remainder_size | PREV_INUSE);
2322 set_foot(remainder, remainder_size);
2323 check_malloced_chunk(victim, nb);
2324 return chunk2mem(victim);
2327 else if (remainder_size >= 0) /* take */
2329 set_inuse_bit_at_offset(victim, victim_size);
2330 unlink(victim, bck, fwd);
2331 check_malloced_chunk(victim, nb);
2332 return chunk2mem(victim);
2337 bin = next_bin(bin);
2339 } while ((++idx & (BINBLOCKWIDTH - 1)) != 0);
2341 /* Clear out the block bit. */
2343 do /* Possibly backtrack to try to clear a partial block */
2345 if ((startidx & (BINBLOCKWIDTH - 1)) == 0)
2347 av_[1] = (mbinptr)(binblocks_r & ~block);
2352 } while (first(q) == q);
2354 /* Get to the next possibly nonempty block */
2356 if ( (block <<= 1) <= binblocks_r && (block != 0) )
2358 while ((block & binblocks_r) == 0)
2360 idx += BINBLOCKWIDTH;
2370 /* Try to use top chunk */
2372 /* Require that there be a remainder, ensuring top always exists */
2373 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2377 /* If big and would otherwise need to extend, try to use mmap instead */
2378 if ((unsigned long)nb >= (unsigned long)mmap_threshold &&
2379 (victim = mmap_chunk(nb)) != 0)
2380 return chunk2mem(victim);
2384 malloc_extend_top(nb);
2385 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2386 return NULL; /* propagate failure */
2390 set_head(victim, nb | PREV_INUSE);
2391 top = chunk_at_offset(victim, nb);
2392 set_head(top, remainder_size | PREV_INUSE);
2393 check_malloced_chunk(victim, nb);
2394 return chunk2mem(victim);
2407 1. free(0) has no effect.
2409 2. If the chunk was allocated via mmap, it is release via munmap().
2411 3. If a returned chunk borders the current high end of memory,
2412 it is consolidated into the top, and if the total unused
2413 topmost memory exceeds the trim threshold, malloc_trim is
2416 4. Other chunks are consolidated as they arrive, and
2417 placed in corresponding bins. (This includes the case of
2418 consolidating with the current `last_remainder').
2424 void fREe(Void_t* mem)
2426 void fREe(mem) Void_t* mem;
2429 mchunkptr p; /* chunk corresponding to mem */
2430 INTERNAL_SIZE_T hd; /* its head field */
2431 INTERNAL_SIZE_T sz; /* its size */
2432 int idx; /* its bin index */
2433 mchunkptr next; /* next contiguous chunk */
2434 INTERNAL_SIZE_T nextsz; /* its size */
2435 INTERNAL_SIZE_T prevsz; /* size of previous contiguous chunk */
2436 mchunkptr bck; /* misc temp for linking */
2437 mchunkptr fwd; /* misc temp for linking */
2438 int islr; /* track whether merging with last_remainder */
2440 if (mem == NULL) /* free(0) has no effect */
2447 if (hd & IS_MMAPPED) /* release mmapped memory. */
2454 check_inuse_chunk(p);
2456 sz = hd & ~PREV_INUSE;
2457 next = chunk_at_offset(p, sz);
2458 nextsz = chunksize(next);
2460 if (next == top) /* merge with top */
2464 if (!(hd & PREV_INUSE)) /* consolidate backward */
2466 prevsz = p->prev_size;
2467 p = chunk_at_offset(p, -((long) prevsz));
2469 unlink(p, bck, fwd);
2472 set_head(p, sz | PREV_INUSE);
2474 if ((unsigned long)(sz) >= (unsigned long)trim_threshold)
2475 malloc_trim(top_pad);
2479 set_head(next, nextsz); /* clear inuse bit */
2483 if (!(hd & PREV_INUSE)) /* consolidate backward */
2485 prevsz = p->prev_size;
2486 p = chunk_at_offset(p, -((long) prevsz));
2489 if (p->fd == last_remainder) /* keep as last_remainder */
2492 unlink(p, bck, fwd);
2495 if (!(inuse_bit_at_offset(next, nextsz))) /* consolidate forward */
2499 if (!islr && next->fd == last_remainder) /* re-insert last_remainder */
2502 link_last_remainder(p);
2505 unlink(next, bck, fwd);
2509 set_head(p, sz | PREV_INUSE);
2512 frontlink(p, sz, idx, bck, fwd);
2523 Chunks that were obtained via mmap cannot be extended or shrunk
2524 unless HAVE_MREMAP is defined, in which case mremap is used.
2525 Otherwise, if their reallocation is for additional space, they are
2526 copied. If for less, they are just left alone.
2528 Otherwise, if the reallocation is for additional space, and the
2529 chunk can be extended, it is, else a malloc-copy-free sequence is
2530 taken. There are several different ways that a chunk could be
2531 extended. All are tried:
2533 * Extending forward into following adjacent free chunk.
2534 * Shifting backwards, joining preceding adjacent space
2535 * Both shifting backwards and extending forward.
2536 * Extending into newly sbrked space
2538 Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2539 size argument of zero (re)allocates a minimum-sized chunk.
2541 If the reallocation is for less space, and the new request is for
2542 a `small' (<512 bytes) size, then the newly unused space is lopped
2545 The old unix realloc convention of allowing the last-free'd chunk
2546 to be used as an argument to realloc is no longer supported.
2547 I don't know of any programs still relying on this feature,
2548 and allowing it would also allow too many other incorrect
2549 usages of realloc to be sensible.
2556 Void_t* rEALLOc(Void_t* oldmem, size_t bytes)
2558 Void_t* rEALLOc(oldmem, bytes) Void_t* oldmem; size_t bytes;
2561 INTERNAL_SIZE_T nb; /* padded request size */
2563 mchunkptr oldp; /* chunk corresponding to oldmem */
2564 INTERNAL_SIZE_T oldsize; /* its size */
2566 mchunkptr newp; /* chunk to return */
2567 INTERNAL_SIZE_T newsize; /* its size */
2568 Void_t* newmem; /* corresponding user mem */
2570 mchunkptr next; /* next contiguous chunk after oldp */
2571 INTERNAL_SIZE_T nextsize; /* its size */
2573 mchunkptr prev; /* previous contiguous chunk before oldp */
2574 INTERNAL_SIZE_T prevsize; /* its size */
2576 mchunkptr remainder; /* holds split off extra space from newp */
2577 INTERNAL_SIZE_T remainder_size; /* its size */
2579 mchunkptr bck; /* misc temp for linking */
2580 mchunkptr fwd; /* misc temp for linking */
2582 #ifdef REALLOC_ZERO_BYTES_FREES
2583 if (bytes == 0) { fREe(oldmem); return 0; }
2586 if ((long)bytes < 0) return NULL;
2588 /* realloc of null is supposed to be same as malloc */
2589 if (oldmem == NULL) return mALLOc(bytes);
2591 newp = oldp = mem2chunk(oldmem);
2592 newsize = oldsize = chunksize(oldp);
2595 nb = request2size(bytes);
2598 if (chunk_is_mmapped(oldp))
2601 newp = mremap_chunk(oldp, nb);
2602 if(newp) return chunk2mem(newp);
2604 /* Note the extra SIZE_SZ overhead. */
2605 if(oldsize - SIZE_SZ >= nb) return oldmem; /* do nothing */
2606 /* Must alloc, copy, free. */
2607 newmem = mALLOc(bytes);
2608 if (newmem == 0) return 0; /* propagate failure */
2609 MALLOC_COPY(newmem, oldmem, oldsize - 2*SIZE_SZ);
2615 check_inuse_chunk(oldp);
2617 if ((long)(oldsize) < (long)(nb))
2620 /* Try expanding forward */
2622 next = chunk_at_offset(oldp, oldsize);
2623 if (next == top || !inuse(next))
2625 nextsize = chunksize(next);
2627 /* Forward into top only if a remainder */
2630 if ((long)(nextsize + newsize) >= (long)(nb + MINSIZE))
2632 newsize += nextsize;
2633 top = chunk_at_offset(oldp, nb);
2634 set_head(top, (newsize - nb) | PREV_INUSE);
2635 set_head_size(oldp, nb);
2636 return chunk2mem(oldp);
2640 /* Forward into next chunk */
2641 else if (((long)(nextsize + newsize) >= (long)(nb)))
2643 unlink(next, bck, fwd);
2644 newsize += nextsize;
2654 /* Try shifting backwards. */
2656 if (!prev_inuse(oldp))
2658 prev = prev_chunk(oldp);
2659 prevsize = chunksize(prev);
2661 /* try forward + backward first to save a later consolidation */
2668 if ((long)(nextsize + prevsize + newsize) >= (long)(nb + MINSIZE))
2670 unlink(prev, bck, fwd);
2672 newsize += prevsize + nextsize;
2673 newmem = chunk2mem(newp);
2674 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2675 top = chunk_at_offset(newp, nb);
2676 set_head(top, (newsize - nb) | PREV_INUSE);
2677 set_head_size(newp, nb);
2682 /* into next chunk */
2683 else if (((long)(nextsize + prevsize + newsize) >= (long)(nb)))
2685 unlink(next, bck, fwd);
2686 unlink(prev, bck, fwd);
2688 newsize += nextsize + prevsize;
2689 newmem = chunk2mem(newp);
2690 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2696 if (prev != NULL && (long)(prevsize + newsize) >= (long)nb)
2698 unlink(prev, bck, fwd);
2700 newsize += prevsize;
2701 newmem = chunk2mem(newp);
2702 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2709 newmem = mALLOc (bytes);
2711 if (newmem == NULL) /* propagate failure */
2714 /* Avoid copy if newp is next chunk after oldp. */
2715 /* (This can only happen when new chunk is sbrk'ed.) */
2717 if ( (newp = mem2chunk(newmem)) == next_chunk(oldp))
2719 newsize += chunksize(newp);
2724 /* Otherwise copy, free, and exit */
2725 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2731 split: /* split off extra room in old or expanded chunk */
2733 if (newsize - nb >= MINSIZE) /* split off remainder */
2735 remainder = chunk_at_offset(newp, nb);
2736 remainder_size = newsize - nb;
2737 set_head_size(newp, nb);
2738 set_head(remainder, remainder_size | PREV_INUSE);
2739 set_inuse_bit_at_offset(remainder, remainder_size);
2740 fREe(chunk2mem(remainder)); /* let free() deal with it */
2744 set_head_size(newp, newsize);
2745 set_inuse_bit_at_offset(newp, newsize);
2748 check_inuse_chunk(newp);
2749 return chunk2mem(newp);
2759 memalign requests more than enough space from malloc, finds a spot
2760 within that chunk that meets the alignment request, and then
2761 possibly frees the leading and trailing space.
2763 The alignment argument must be a power of two. This property is not
2764 checked by memalign, so misuse may result in random runtime errors.
2766 8-byte alignment is guaranteed by normal malloc calls, so don't
2767 bother calling memalign with an argument of 8 or less.
2769 Overreliance on memalign is a sure way to fragment space.
2775 Void_t* mEMALIGn(size_t alignment, size_t bytes)
2777 Void_t* mEMALIGn(alignment, bytes) size_t alignment; size_t bytes;
2780 INTERNAL_SIZE_T nb; /* padded request size */
2781 char* m; /* memory returned by malloc call */
2782 mchunkptr p; /* corresponding chunk */
2783 char* brk; /* alignment point within p */
2784 mchunkptr newp; /* chunk to return */
2785 INTERNAL_SIZE_T newsize; /* its size */
2786 INTERNAL_SIZE_T leadsize; /* leading space befor alignment point */
2787 mchunkptr remainder; /* spare room at end to split off */
2788 long remainder_size; /* its size */
2790 if ((long)bytes < 0) return NULL;
2792 /* If need less alignment than we give anyway, just relay to malloc */
2794 if (alignment <= MALLOC_ALIGNMENT) return mALLOc(bytes);
2796 /* Otherwise, ensure that it is at least a minimum chunk size */
2798 if (alignment < MINSIZE) alignment = MINSIZE;
2800 /* Call malloc with worst case padding to hit alignment. */
2802 nb = request2size(bytes);
2803 m = (char*)(mALLOc(nb + alignment + MINSIZE));
2805 if (m == NULL) return NULL; /* propagate failure */
2809 if ((((unsigned long)(m)) % alignment) == 0) /* aligned */
2812 if(chunk_is_mmapped(p))
2813 return chunk2mem(p); /* nothing more to do */
2816 else /* misaligned */
2819 Find an aligned spot inside chunk.
2820 Since we need to give back leading space in a chunk of at
2821 least MINSIZE, if the first calculation places us at
2822 a spot with less than MINSIZE leader, we can move to the
2823 next aligned spot -- we've allocated enough total room so that
2824 this is always possible.
2827 brk = (char*)mem2chunk(((unsigned long)(m + alignment - 1)) & -((signed) alignment));
2828 if ((long)(brk - (char*)(p)) < MINSIZE) brk = brk + alignment;
2830 newp = (mchunkptr)brk;
2831 leadsize = brk - (char*)(p);
2832 newsize = chunksize(p) - leadsize;
2835 if(chunk_is_mmapped(p))
2837 newp->prev_size = p->prev_size + leadsize;
2838 set_head(newp, newsize|IS_MMAPPED);
2839 return chunk2mem(newp);
2843 /* give back leader, use the rest */
2845 set_head(newp, newsize | PREV_INUSE);
2846 set_inuse_bit_at_offset(newp, newsize);
2847 set_head_size(p, leadsize);
2851 assert (newsize >= nb && (((unsigned long)(chunk2mem(p))) % alignment) == 0);
2854 /* Also give back spare room at the end */
2856 remainder_size = chunksize(p) - nb;
2858 if (remainder_size >= (long)MINSIZE)
2860 remainder = chunk_at_offset(p, nb);
2861 set_head(remainder, remainder_size | PREV_INUSE);
2862 set_head_size(p, nb);
2863 fREe(chunk2mem(remainder));
2866 check_inuse_chunk(p);
2867 return chunk2mem(p);
2875 valloc just invokes memalign with alignment argument equal
2876 to the page size of the system (or as near to this as can
2877 be figured out from all the includes/defines above.)
2881 Void_t* vALLOc(size_t bytes)
2883 Void_t* vALLOc(bytes) size_t bytes;
2886 return mEMALIGn (malloc_getpagesize, bytes);
2890 pvalloc just invokes valloc for the nearest pagesize
2891 that will accommodate request
2896 Void_t* pvALLOc(size_t bytes)
2898 Void_t* pvALLOc(bytes) size_t bytes;
2901 size_t pagesize = malloc_getpagesize;
2902 return mEMALIGn (pagesize, (bytes + pagesize - 1) & ~(pagesize - 1));
2907 calloc calls malloc, then zeroes out the allocated chunk.
2912 Void_t* cALLOc(size_t n, size_t elem_size)
2914 Void_t* cALLOc(n, elem_size) size_t n; size_t elem_size;
2918 INTERNAL_SIZE_T csz;
2920 INTERNAL_SIZE_T sz = n * elem_size;
2923 /* check if expand_top called, in which case don't need to clear */
2925 mchunkptr oldtop = top;
2926 INTERNAL_SIZE_T oldtopsize = chunksize(top);
2928 Void_t* mem = mALLOc (sz);
2930 if ((long)n < 0) return NULL;
2938 /* Two optional cases in which clearing not necessary */
2942 if (chunk_is_mmapped(p)) return mem;
2948 if (p == oldtop && csz > oldtopsize)
2950 /* clear only the bytes from non-freshly-sbrked memory */
2955 MALLOC_ZERO(mem, csz - SIZE_SZ);
2962 cfree just calls free. It is needed/defined on some systems
2963 that pair it with calloc, presumably for odd historical reasons.
2967 #if !defined(INTERNAL_LINUX_C_LIB) || !defined(__ELF__)
2969 void cfree(Void_t *mem)
2971 void cfree(mem) Void_t *mem;
2982 Malloc_trim gives memory back to the system (via negative
2983 arguments to sbrk) if there is unused memory at the `high' end of
2984 the malloc pool. You can call this after freeing large blocks of
2985 memory to potentially reduce the system-level memory requirements
2986 of a program. However, it cannot guarantee to reduce memory. Under
2987 some allocation patterns, some large free blocks of memory will be
2988 locked between two used chunks, so they cannot be given back to
2991 The `pad' argument to malloc_trim represents the amount of free
2992 trailing space to leave untrimmed. If this argument is zero,
2993 only the minimum amount of memory to maintain internal data
2994 structures will be left (one page or less). Non-zero arguments
2995 can be supplied to maintain enough trailing space to service
2996 future expected allocations without having to re-obtain memory
2999 Malloc_trim returns 1 if it actually released any memory, else 0.
3004 int malloc_trim(size_t pad)
3006 int malloc_trim(pad) size_t pad;
3009 long top_size; /* Amount of top-most memory */
3010 long extra; /* Amount to release */
3011 char* current_brk; /* address returned by pre-check sbrk call */
3012 char* new_brk; /* address returned by negative sbrk call */
3014 unsigned long pagesz = malloc_getpagesize;
3016 top_size = chunksize(top);
3017 extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
3019 if (extra < (long)pagesz) /* Not enough memory to release */
3024 /* Test to make sure no one else called sbrk */
3025 current_brk = (char*)(MORECORE (0));
3026 if (current_brk != (char*)(top) + top_size)
3027 return 0; /* Apparently we don't own memory; must fail */
3031 new_brk = (char*)(MORECORE (-extra));
3033 if (new_brk == (char*)(MORECORE_FAILURE)) /* sbrk failed? */
3035 /* Try to figure out what we have */
3036 current_brk = (char*)(MORECORE (0));
3037 top_size = current_brk - (char*)top;
3038 if (top_size >= (long)MINSIZE) /* if not, we are very very dead! */
3040 sbrked_mem = current_brk - sbrk_base;
3041 set_head(top, top_size | PREV_INUSE);
3049 /* Success. Adjust top accordingly. */
3050 set_head(top, (top_size - extra) | PREV_INUSE);
3051 sbrked_mem -= extra;
3064 This routine tells you how many bytes you can actually use in an
3065 allocated chunk, which may be more than you requested (although
3066 often not). You can use this many bytes without worrying about
3067 overwriting other allocated objects. Not a particularly great
3068 programming practice, but still sometimes useful.
3073 size_t malloc_usable_size(Void_t* mem)
3075 size_t malloc_usable_size(mem) Void_t* mem;
3084 if(!chunk_is_mmapped(p))
3086 if (!inuse(p)) return 0;
3087 check_inuse_chunk(p);
3088 return chunksize(p) - SIZE_SZ;
3090 return chunksize(p) - 2*SIZE_SZ;
3097 /* Utility to update current_mallinfo for malloc_stats and mallinfo() */
3100 static void malloc_update_mallinfo()
3109 INTERNAL_SIZE_T avail = chunksize(top);
3110 int navail = ((long)(avail) >= (long)MINSIZE)? 1 : 0;
3112 for (i = 1; i < NAV; ++i)
3115 for (p = last(b); p != b; p = p->bk)
3118 check_free_chunk(p);
3119 for (q = next_chunk(p);
3120 q < top && inuse(q) && (long)(chunksize(q)) >= (long)MINSIZE;
3122 check_inuse_chunk(q);
3124 avail += chunksize(p);
3129 current_mallinfo.ordblks = navail;
3130 current_mallinfo.uordblks = sbrked_mem - avail;
3131 current_mallinfo.fordblks = avail;
3132 current_mallinfo.hblks = n_mmaps;
3133 current_mallinfo.hblkhd = mmapped_mem;
3134 current_mallinfo.keepcost = chunksize(top);
3145 Prints on the amount of space obtain from the system (both
3146 via sbrk and mmap), the maximum amount (which may be more than
3147 current if malloc_trim and/or munmap got called), the maximum
3148 number of simultaneous mmap regions used, and the current number
3149 of bytes allocated via malloc (or realloc, etc) but not yet
3150 freed. (Note that this is the number of bytes allocated, not the
3151 number requested. It will be larger than the number requested
3152 because of alignment and bookkeeping overhead.)
3159 malloc_update_mallinfo();
3160 printf("max system bytes = %10u\n",
3161 (unsigned int)(max_total_mem));
3162 printf("system bytes = %10u\n",
3163 (unsigned int)(sbrked_mem + mmapped_mem));
3164 printf("in use bytes = %10u\n",
3165 (unsigned int)(current_mallinfo.uordblks + mmapped_mem));
3167 printf("max mmap regions = %10u\n",
3168 (unsigned int)max_n_mmaps);
3174 mallinfo returns a copy of updated current mallinfo.
3178 struct mallinfo mALLINFo()
3180 malloc_update_mallinfo();
3181 return current_mallinfo;
3191 mallopt is the general SVID/XPG interface to tunable parameters.
3192 The format is to provide a (parameter-number, parameter-value) pair.
3193 mallopt then sets the corresponding parameter to the argument
3194 value if it can (i.e., so long as the value is meaningful),
3195 and returns 1 if successful else 0.
3197 See descriptions of tunable parameters above.
3202 int mALLOPt(int param_number, int value)
3204 int mALLOPt(param_number, value) int param_number; int value;
3207 switch(param_number)
3209 case M_TRIM_THRESHOLD:
3210 trim_threshold = value; return 1;
3212 top_pad = value; return 1;
3213 case M_MMAP_THRESHOLD:
3214 mmap_threshold = value; return 1;
3217 n_mmaps_max = value; return 1;
3219 if (value != 0) return 0; else n_mmaps_max = value; return 1;
3231 V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee)
3232 * return null for negative arguments
3233 * Added Several WIN32 cleanups from Martin C. Fong <mcfong@yahoo.com>
3234 * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h'
3235 (e.g. WIN32 platforms)
3236 * Cleanup up header file inclusion for WIN32 platforms
3237 * Cleanup code to avoid Microsoft Visual C++ compiler complaints
3238 * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing
3239 memory allocation routines
3240 * Set 'malloc_getpagesize' for WIN32 platforms (needs more work)
3241 * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to
3242 usage of 'assert' in non-WIN32 code
3243 * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to
3245 * Always call 'fREe()' rather than 'free()'
3247 V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee)
3248 * Fixed ordering problem with boundary-stamping
3250 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)
3251 * Added pvalloc, as recommended by H.J. Liu
3252 * Added 64bit pointer support mainly from Wolfram Gloger
3253 * Added anonymously donated WIN32 sbrk emulation
3254 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3255 * malloc_extend_top: fix mask error that caused wastage after
3257 * Add linux mremap support code from HJ Liu
3259 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)
3260 * Integrated most documentation with the code.
3261 * Add support for mmap, with help from
3262 Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3263 * Use last_remainder in more cases.
3264 * Pack bins using idea from colin@nyx10.cs.du.edu
3265 * Use ordered bins instead of best-fit threshhold
3266 * Eliminate block-local decls to simplify tracing and debugging.
3267 * Support another case of realloc via move into top
3268 * Fix error occuring when initial sbrk_base not word-aligned.
3269 * Rely on page size for units instead of SBRK_UNIT to
3270 avoid surprises about sbrk alignment conventions.
3271 * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3272 (raymond@es.ele.tue.nl) for the suggestion.
3273 * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3274 * More precautions for cases where other routines call sbrk,
3275 courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3276 * Added macros etc., allowing use in linux libc from
3277 H.J. Lu (hjl@gnu.ai.mit.edu)
3278 * Inverted this history list
3280 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)
3281 * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3282 * Removed all preallocation code since under current scheme
3283 the work required to undo bad preallocations exceeds
3284 the work saved in good cases for most test programs.
3285 * No longer use return list or unconsolidated bins since
3286 no scheme using them consistently outperforms those that don't
3287 given above changes.
3288 * Use best fit for very large chunks to prevent some worst-cases.
3289 * Added some support for debugging
3291 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)
3292 * Removed footers when chunks are in use. Thanks to
3293 Paul Wilson (wilson@cs.texas.edu) for the suggestion.
3295 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)
3296 * Added malloc_trim, with help from Wolfram Gloger
3297 (wmglo@Dent.MED.Uni-Muenchen.DE).
3299 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)
3301 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)
3302 * realloc: try to expand in both directions
3303 * malloc: swap order of clean-bin strategy;
3304 * realloc: only conditionally expand backwards
3305 * Try not to scavenge used bins
3306 * Use bin counts as a guide to preallocation
3307 * Occasionally bin return list chunks in first scan
3308 * Add a few optimizations from colin@nyx10.cs.du.edu
3310 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)
3311 * faster bin computation & slightly different binning
3312 * merged all consolidations to one part of malloc proper
3313 (eliminating old malloc_find_space & malloc_clean_bin)
3314 * Scan 2 returns chunks (not just 1)
3315 * Propagate failure in realloc if malloc returns 0
3316 * Add stuff to allow compilation on non-ANSI compilers
3317 from kpv@research.att.com
3319 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)
3320 * removed potential for odd address access in prev_chunk
3321 * removed dependency on getpagesize.h
3322 * misc cosmetics and a bit more internal documentation
3323 * anticosmetics: mangled names in macros to evade debugger strangeness
3324 * tested on sparc, hp-700, dec-mips, rs6000
3325 with gcc & native cc (hp, dec only) allowing
3326 Detlefs & Zorn comparison study (in SIGPLAN Notices.)
3328 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)
3329 * Based loosely on libg++-1.2X malloc. (It retains some of the overall
3330 structure of old version, but most details differ.)