1 /* Malloc implementation for multiple threads without lock contention.
2 Copyright (C) 1996, 1997, 1998 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Wolfram Gloger <wmglo@dent.med.uni-muenchen.de>
5 and Doug Lea <dl@cs.oswego.edu>, 1996.
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Library General Public License as
9 published by the Free Software Foundation; either version 2 of the
10 License, or (at your option) any later version.
12 The GNU C Library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Library General Public License for more details.
17 You should have received a copy of the GNU Library General Public
18 License along with the GNU C Library; see the file COPYING.LIB. If not,
19 write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
22 /* V2.6.4-pt3 Thu Feb 20 1997
24 This work is mainly derived from malloc-2.6.4 by Doug Lea
25 <dl@cs.oswego.edu>, which is available from:
27 ftp://g.oswego.edu/pub/misc/malloc.c
29 Most of the original comments are reproduced in the code below.
31 * Why use this malloc?
33 This is not the fastest, most space-conserving, most portable, or
34 most tunable malloc ever written. However it is among the fastest
35 while also being among the most space-conserving, portable and tunable.
36 Consistent balance across these factors results in a good general-purpose
37 allocator. For a high-level description, see
38 http://g.oswego.edu/dl/html/malloc.html
40 On many systems, the standard malloc implementation is by itself not
41 thread-safe, and therefore wrapped with a single global lock around
42 all malloc-related functions. In some applications, especially with
43 multiple available processors, this can lead to contention problems
44 and bad performance. This malloc version was designed with the goal
45 to avoid waiting for locks as much as possible. Statistics indicate
46 that this goal is achieved in many cases.
48 * Synopsis of public routines
50 (Much fuller descriptions are contained in the program documentation below.)
53 Initialize global configuration. When compiled for multiple threads,
54 this function must be called once before any other function in the
55 package. It is not required otherwise. It is called automatically
56 in the Linux/GNU C libray or when compiling with MALLOC_HOOKS.
58 Return a pointer to a newly allocated chunk of at least n bytes, or null
59 if no space is available.
61 Release the chunk of memory pointed to by p, or no effect if p is null.
62 realloc(Void_t* p, size_t n);
63 Return a pointer to a chunk of size n that contains the same data
64 as does chunk p up to the minimum of (n, p's size) bytes, or null
65 if no space is available. The returned pointer may or may not be
66 the same as p. If p is null, equivalent to malloc. Unless the
67 #define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
68 size argument of zero (re)allocates a minimum-sized chunk.
69 memalign(size_t alignment, size_t n);
70 Return a pointer to a newly allocated chunk of n bytes, aligned
71 in accord with the alignment argument, which must be a power of
74 Equivalent to memalign(pagesize, n), where pagesize is the page
75 size of the system (or as near to this as can be figured out from
76 all the includes/defines below.)
78 Equivalent to valloc(minimum-page-that-holds(n)), that is,
79 round up n to nearest pagesize.
80 calloc(size_t unit, size_t quantity);
81 Returns a pointer to quantity * unit bytes, with all locations
84 Equivalent to free(p).
85 malloc_trim(size_t pad);
86 Release all but pad bytes of freed top-most memory back
87 to the system. Return 1 if successful, else 0.
88 malloc_usable_size(Void_t* p);
89 Report the number usable allocated bytes associated with allocated
90 chunk p. This may or may not report more bytes than were requested,
91 due to alignment and minimum size constraints.
93 Prints brief summary statistics on stderr.
95 Returns (by copy) a struct containing various summary statistics.
96 mallopt(int parameter_number, int parameter_value)
97 Changes one of the tunable parameters described below. Returns
98 1 if successful in changing the parameter, else 0.
103 8 byte alignment is currently hardwired into the design. This
104 seems to suffice for all current machines and C compilers.
106 Assumed pointer representation: 4 or 8 bytes
107 Code for 8-byte pointers is untested by me but has worked
108 reliably by Wolfram Gloger, who contributed most of the
109 changes supporting this.
111 Assumed size_t representation: 4 or 8 bytes
112 Note that size_t is allowed to be 4 bytes even if pointers are 8.
114 Minimum overhead per allocated chunk: 4 or 8 bytes
115 Each malloced chunk has a hidden overhead of 4 bytes holding size
116 and status information.
118 Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
119 8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
121 When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
122 ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
123 needed; 4 (8) for a trailing size field
124 and 8 (16) bytes for free list pointers. Thus, the minimum
125 allocatable size is 16/24/32 bytes.
127 Even a request for zero bytes (i.e., malloc(0)) returns a
128 pointer to something of the minimum allocatable size.
130 Maximum allocated size: 4-byte size_t: 2^31 - 8 bytes
131 8-byte size_t: 2^63 - 16 bytes
133 It is assumed that (possibly signed) size_t bit values suffice to
134 represent chunk sizes. `Possibly signed' is due to the fact
135 that `size_t' may be defined on a system as either a signed or
136 an unsigned type. To be conservative, values that would appear
137 as negative numbers are avoided.
138 Requests for sizes with a negative sign bit will return a
141 Maximum overhead wastage per allocated chunk: normally 15 bytes
143 Alignment demands, plus the minimum allocatable size restriction
144 make the normal worst-case wastage 15 bytes (i.e., up to 15
145 more bytes will be allocated than were requested in malloc), with
147 1. Because requests for zero bytes allocate non-zero space,
148 the worst case wastage for a request of zero bytes is 24 bytes.
149 2. For requests >= mmap_threshold that are serviced via
150 mmap(), the worst case wastage is 8 bytes plus the remainder
151 from a system page (the minimal mmap unit); typically 4096 bytes.
155 Here are some features that are NOT currently supported
157 * No automated mechanism for fully checking that all accesses
158 to malloced memory stay within their bounds.
159 * No support for compaction.
161 * Synopsis of compile-time options:
163 People have reported using previous versions of this malloc on all
164 versions of Unix, sometimes by tweaking some of the defines
165 below. It has been tested most extensively on Solaris and
166 Linux. People have also reported adapting this malloc for use in
167 stand-alone embedded systems.
169 The implementation is in straight, hand-tuned ANSI C. Among other
170 consequences, it uses a lot of macros. Because of this, to be at
171 all usable, this code should be compiled using an optimizing compiler
172 (for example gcc -O2) that can simplify expressions and control
175 __STD_C (default: derived from C compiler defines)
176 Nonzero if using ANSI-standard C compiler, a C++ compiler, or
177 a C compiler sufficiently close to ANSI to get away with it.
178 MALLOC_DEBUG (default: NOT defined)
179 Define to enable debugging. Adds fairly extensive assertion-based
180 checking to help track down memory errors, but noticeably slows down
182 MALLOC_HOOKS (default: NOT defined)
183 Define to enable support run-time replacement of the allocation
184 functions through user-defined `hooks'.
185 REALLOC_ZERO_BYTES_FREES (default: NOT defined)
186 Define this if you think that realloc(p, 0) should be equivalent
187 to free(p). Otherwise, since malloc returns a unique pointer for
188 malloc(0), so does realloc(p, 0).
189 HAVE_MEMCPY (default: defined)
190 Define if you are not otherwise using ANSI STD C, but still
191 have memcpy and memset in your C library and want to use them.
192 Otherwise, simple internal versions are supplied.
193 USE_MEMCPY (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
194 Define as 1 if you want the C library versions of memset and
195 memcpy called in realloc and calloc (otherwise macro versions are used).
196 At least on some platforms, the simple macro versions usually
197 outperform libc versions.
198 HAVE_MMAP (default: defined as 1)
199 Define to non-zero to optionally make malloc() use mmap() to
200 allocate very large blocks.
201 HAVE_MREMAP (default: defined as 0 unless Linux libc set)
202 Define to non-zero to optionally make realloc() use mremap() to
203 reallocate very large blocks.
204 malloc_getpagesize (default: derived from system #includes)
205 Either a constant or routine call returning the system page size.
206 HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
207 Optionally define if you are on a system with a /usr/include/malloc.h
208 that declares struct mallinfo. It is not at all necessary to
209 define this even if you do, but will ensure consistency.
210 INTERNAL_SIZE_T (default: size_t)
211 Define to a 32-bit type (probably `unsigned int') if you are on a
212 64-bit machine, yet do not want or need to allow malloc requests of
213 greater than 2^31 to be handled. This saves space, especially for
215 _LIBC (default: NOT defined)
216 Defined only when compiled as part of the Linux libc/glibc.
217 Also note that there is some odd internal name-mangling via defines
218 (for example, internally, `malloc' is named `mALLOc') needed
219 when compiling in this case. These look funny but don't otherwise
221 LACKS_UNISTD_H (default: undefined)
222 Define this if your system does not have a <unistd.h>.
223 MORECORE (default: sbrk)
224 The name of the routine to call to obtain more memory from the system.
225 MORECORE_FAILURE (default: -1)
226 The value returned upon failure of MORECORE.
227 MORECORE_CLEARS (default 1)
228 True (1) if the routine mapped to MORECORE zeroes out memory (which
230 DEFAULT_TRIM_THRESHOLD
232 DEFAULT_MMAP_THRESHOLD
234 Default values of tunable parameters (described in detail below)
235 controlling interaction with host system routines (sbrk, mmap, etc).
236 These values may also be changed dynamically via mallopt(). The
237 preset defaults are those that give best performance for typical
240 When the standard debugging hooks are in place, and a pointer is
241 detected as corrupt, do nothing (0), print an error message (1),
249 * Compile-time options for multiple threads:
251 USE_PTHREADS, USE_THR, USE_SPROC
252 Define one of these as 1 to select the thread interface:
253 POSIX threads, Solaris threads or SGI sproc's, respectively.
254 If none of these is defined as non-zero, you get a `normal'
255 malloc implementation which is not thread-safe. Support for
256 multiple threads requires HAVE_MMAP=1. As an exception, when
257 compiling for GNU libc, i.e. when _LIBC is defined, then none of
258 the USE_... symbols have to be defined.
262 When thread support is enabled, additional `heap's are created
263 with mmap calls. These are limited in size; HEAP_MIN_SIZE should
264 be a multiple of the page size, while HEAP_MAX_SIZE must be a power
265 of two for alignment reasons. HEAP_MAX_SIZE should be at least
266 twice as large as the mmap threshold.
268 When this is defined as non-zero, some statistics on mutex locking
279 #if defined (__STDC__)
286 #endif /*__cplusplus*/
299 # include <stddef.h> /* for size_t */
300 # if defined _LIBC || defined MALLOC_HOOKS
301 # include <stdlib.h> /* for getenv(), abort() */
304 # include <sys/types.h>
307 /* Macros for handling mutexes and thread-specific data. This is
308 included early, because some thread-related header files (such as
309 pthread.h) should be included before any others. */
310 #include "thread-m.h"
316 #include <stdio.h> /* needed for malloc_stats */
327 Because freed chunks may be overwritten with link fields, this
328 malloc will often die when freed memory is overwritten by user
329 programs. This can be very effective (albeit in an annoying way)
330 in helping track down dangling pointers.
332 If you compile with -DMALLOC_DEBUG, a number of assertion checks are
333 enabled that will catch more memory errors. You probably won't be
334 able to make much sense of the actual assertion errors, but they
335 should help you locate incorrectly overwritten memory. The
336 checking is fairly extensive, and will slow down execution
337 noticeably. Calling malloc_stats or mallinfo with MALLOC_DEBUG set will
338 attempt to check every non-mmapped allocated and free chunk in the
339 course of computing the summaries. (By nature, mmapped regions
340 cannot be checked very much automatically.)
342 Setting MALLOC_DEBUG may also be helpful if you are trying to modify
343 this code. The assertions in the check routines spell out in more
344 detail the assumptions and invariants underlying the algorithms.
351 #define assert(x) ((void)0)
356 INTERNAL_SIZE_T is the word-size used for internal bookkeeping
357 of chunk sizes. On a 64-bit machine, you can reduce malloc
358 overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
359 at the expense of not being able to handle requests greater than
360 2^31. This limitation is hardly ever a concern; you are encouraged
361 to set this. However, the default version is the same as size_t.
364 #ifndef INTERNAL_SIZE_T
365 #define INTERNAL_SIZE_T size_t
369 REALLOC_ZERO_BYTES_FREES should be set if a call to
370 realloc with zero bytes should be the same as a call to free.
371 Some people think it should. Otherwise, since this malloc
372 returns a unique pointer for malloc(0), so does realloc(p, 0).
376 /* #define REALLOC_ZERO_BYTES_FREES */
380 HAVE_MEMCPY should be defined if you are not otherwise using
381 ANSI STD C, but still have memcpy and memset in your C library
382 and want to use them in calloc and realloc. Otherwise simple
383 macro versions are defined here.
385 USE_MEMCPY should be defined as 1 if you actually want to
386 have memset and memcpy called. People report that the macro
387 versions are often enough faster than libc versions on many
388 systems that it is better to use them.
392 #define HAVE_MEMCPY 1
402 #if (__STD_C || defined(HAVE_MEMCPY))
405 void* memset(void*, int, size_t);
406 void* memcpy(void*, const void*, size_t);
415 /* The following macros are only invoked with (2n+1)-multiples of
416 INTERNAL_SIZE_T units, with a positive integer n. This is exploited
417 for fast inline execution when n is small. */
419 #define MALLOC_ZERO(charp, nbytes) \
421 INTERNAL_SIZE_T mzsz = (nbytes); \
422 if(mzsz <= 9*sizeof(mzsz)) { \
423 INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
424 if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
426 if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
428 if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
433 } else memset((charp), 0, mzsz); \
436 #define MALLOC_COPY(dest,src,nbytes) \
438 INTERNAL_SIZE_T mcsz = (nbytes); \
439 if(mcsz <= 9*sizeof(mcsz)) { \
440 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
441 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
442 if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
443 *mcdst++ = *mcsrc++; \
444 if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
445 *mcdst++ = *mcsrc++; \
446 if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
447 *mcdst++ = *mcsrc++; }}} \
448 *mcdst++ = *mcsrc++; \
449 *mcdst++ = *mcsrc++; \
451 } else memcpy(dest, src, mcsz); \
454 #else /* !USE_MEMCPY */
456 /* Use Duff's device for good zeroing/copying performance. */
458 #define MALLOC_ZERO(charp, nbytes) \
460 INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
461 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
462 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
464 case 0: for(;;) { *mzp++ = 0; \
465 case 7: *mzp++ = 0; \
466 case 6: *mzp++ = 0; \
467 case 5: *mzp++ = 0; \
468 case 4: *mzp++ = 0; \
469 case 3: *mzp++ = 0; \
470 case 2: *mzp++ = 0; \
471 case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
475 #define MALLOC_COPY(dest,src,nbytes) \
477 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
478 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
479 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
480 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
482 case 0: for(;;) { *mcdst++ = *mcsrc++; \
483 case 7: *mcdst++ = *mcsrc++; \
484 case 6: *mcdst++ = *mcsrc++; \
485 case 5: *mcdst++ = *mcsrc++; \
486 case 4: *mcdst++ = *mcsrc++; \
487 case 3: *mcdst++ = *mcsrc++; \
488 case 2: *mcdst++ = *mcsrc++; \
489 case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
497 Define HAVE_MMAP to optionally make malloc() use mmap() to
498 allocate very large blocks. These will be returned to the
499 operating system immediately after a free().
507 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
508 large blocks. This is currently only possible on Linux with
509 kernel versions newer than 1.3.77.
513 #define HAVE_MREMAP defined(__linux__) && !defined(__arm__)
520 #include <sys/mman.h>
522 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
523 #define MAP_ANONYMOUS MAP_ANON
526 #ifndef MAP_NORESERVE
527 # ifdef MAP_AUTORESRV
528 # define MAP_NORESERVE MAP_AUTORESRV
530 # define MAP_NORESERVE 0
534 #endif /* HAVE_MMAP */
537 Access to system page size. To the extent possible, this malloc
538 manages memory from the system in page-size units.
540 The following mechanics for getpagesize were adapted from
541 bsd/gnu getpagesize.h
544 #ifndef LACKS_UNISTD_H
548 #ifndef malloc_getpagesize
549 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
550 # ifndef _SC_PAGE_SIZE
551 # define _SC_PAGE_SIZE _SC_PAGESIZE
554 # ifdef _SC_PAGE_SIZE
555 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
557 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
558 extern size_t getpagesize();
559 # define malloc_getpagesize getpagesize()
561 # include <sys/param.h>
562 # ifdef EXEC_PAGESIZE
563 # define malloc_getpagesize EXEC_PAGESIZE
567 # define malloc_getpagesize NBPG
569 # define malloc_getpagesize (NBPG * CLSIZE)
573 # define malloc_getpagesize NBPC
576 # define malloc_getpagesize PAGESIZE
578 # define malloc_getpagesize (4096) /* just guess */
591 This version of malloc supports the standard SVID/XPG mallinfo
592 routine that returns a struct containing the same kind of
593 information you can get from malloc_stats. It should work on
594 any SVID/XPG compliant system that has a /usr/include/malloc.h
595 defining struct mallinfo. (If you'd like to install such a thing
596 yourself, cut out the preliminary declarations as described above
597 and below and save them in a malloc.h file. But there's no
598 compelling reason to bother to do this.)
600 The main declaration needed is the mallinfo struct that is returned
601 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
602 bunch of fields, most of which are not even meaningful in this
603 version of malloc. Some of these fields are are instead filled by
604 mallinfo() with other numbers that might possibly be of interest.
606 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
607 /usr/include/malloc.h file that includes a declaration of struct
608 mallinfo. If so, it is included; else an SVID2/XPG2 compliant
609 version is declared below. These must be precisely the same for
614 /* #define HAVE_USR_INCLUDE_MALLOC_H */
616 #if HAVE_USR_INCLUDE_MALLOC_H
617 # include "/usr/include/malloc.h"
622 # include "ptmalloc.h"
628 #ifndef DEFAULT_TRIM_THRESHOLD
629 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
633 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
634 to keep before releasing via malloc_trim in free().
636 Automatic trimming is mainly useful in long-lived programs.
637 Because trimming via sbrk can be slow on some systems, and can
638 sometimes be wasteful (in cases where programs immediately
639 afterward allocate more large chunks) the value should be high
640 enough so that your overall system performance would improve by
643 The trim threshold and the mmap control parameters (see below)
644 can be traded off with one another. Trimming and mmapping are
645 two different ways of releasing unused memory back to the
646 system. Between these two, it is often possible to keep
647 system-level demands of a long-lived program down to a bare
648 minimum. For example, in one test suite of sessions measuring
649 the XF86 X server on Linux, using a trim threshold of 128K and a
650 mmap threshold of 192K led to near-minimal long term resource
653 If you are using this malloc in a long-lived program, it should
654 pay to experiment with these values. As a rough guide, you
655 might set to a value close to the average size of a process
656 (program) running on your system. Releasing this much memory
657 would allow such a process to run in memory. Generally, it's
658 worth it to tune for trimming rather than memory mapping when a
659 program undergoes phases where several large chunks are
660 allocated and released in ways that can reuse each other's
661 storage, perhaps mixed with phases where there are no such
662 chunks at all. And in well-behaved long-lived programs,
663 controlling release of large blocks via trimming versus mapping
666 However, in most programs, these parameters serve mainly as
667 protection against the system-level effects of carrying around
668 massive amounts of unneeded memory. Since frequent calls to
669 sbrk, mmap, and munmap otherwise degrade performance, the default
670 parameters are set to relatively high values that serve only as
673 The default trim value is high enough to cause trimming only in
674 fairly extreme (by current memory consumption standards) cases.
675 It must be greater than page size to have any useful effect. To
676 disable trimming completely, you can set to (unsigned long)(-1);
682 #ifndef DEFAULT_TOP_PAD
683 #define DEFAULT_TOP_PAD (0)
687 M_TOP_PAD is the amount of extra `padding' space to allocate or
688 retain whenever sbrk is called. It is used in two ways internally:
690 * When sbrk is called to extend the top of the arena to satisfy
691 a new malloc request, this much padding is added to the sbrk
694 * When malloc_trim is called automatically from free(),
695 it is used as the `pad' argument.
697 In both cases, the actual amount of padding is rounded
698 so that the end of the arena is always a system page boundary.
700 The main reason for using padding is to avoid calling sbrk so
701 often. Having even a small pad greatly reduces the likelihood
702 that nearly every malloc request during program start-up (or
703 after trimming) will invoke sbrk, which needlessly wastes
706 Automatic rounding-up to page-size units is normally sufficient
707 to avoid measurable overhead, so the default is 0. However, in
708 systems where sbrk is relatively slow, it can pay to increase
709 this value, at the expense of carrying around more memory than
715 #ifndef DEFAULT_MMAP_THRESHOLD
716 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
721 M_MMAP_THRESHOLD is the request size threshold for using mmap()
722 to service a request. Requests of at least this size that cannot
723 be allocated using already-existing space will be serviced via mmap.
724 (If enough normal freed space already exists it is used instead.)
726 Using mmap segregates relatively large chunks of memory so that
727 they can be individually obtained and released from the host
728 system. A request serviced through mmap is never reused by any
729 other request (at least not directly; the system may just so
730 happen to remap successive requests to the same locations).
732 Segregating space in this way has the benefit that mmapped space
733 can ALWAYS be individually released back to the system, which
734 helps keep the system level memory demands of a long-lived
735 program low. Mapped memory can never become `locked' between
736 other chunks, as can happen with normally allocated chunks, which
737 menas that even trimming via malloc_trim would not release them.
739 However, it has the disadvantages that:
741 1. The space cannot be reclaimed, consolidated, and then
742 used to service later requests, as happens with normal chunks.
743 2. It can lead to more wastage because of mmap page alignment
745 3. It causes malloc performance to be more dependent on host
746 system memory management support routines which may vary in
747 implementation quality and may impose arbitrary
748 limitations. Generally, servicing a request via normal
749 malloc steps is faster than going through a system's mmap.
751 All together, these considerations should lead you to use mmap
752 only for relatively large requests.
759 #ifndef DEFAULT_MMAP_MAX
761 #define DEFAULT_MMAP_MAX (1024)
763 #define DEFAULT_MMAP_MAX (0)
768 M_MMAP_MAX is the maximum number of requests to simultaneously
769 service using mmap. This parameter exists because:
771 1. Some systems have a limited number of internal tables for
773 2. In most systems, overreliance on mmap can degrade overall
775 3. If a program allocates many large regions, it is probably
776 better off using normal sbrk-based allocation routines that
777 can reclaim and reallocate normal heap memory. Using a
778 small value allows transition into this mode after the
779 first few allocations.
781 Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
782 the default value is 0, and attempts to set it to non-zero values
783 in mallopt will fail.
788 #ifndef DEFAULT_CHECK_ACTION
789 #define DEFAULT_CHECK_ACTION 1
792 /* What to do if the standard debugging hooks are in place and a
793 corrupt pointer is detected: do nothing (0), print an error message
794 (1), or call abort() (2). */
798 #define HEAP_MIN_SIZE (32*1024)
799 #define HEAP_MAX_SIZE (1024*1024) /* must be a power of two */
801 /* HEAP_MIN_SIZE and HEAP_MAX_SIZE limit the size of mmap()ed heaps
802 that are dynamically created for multi-threaded programs. The
803 maximum size must be a power of two, for fast determination of
804 which heap belongs to a chunk. It should be much larger than
805 the mmap threshold, so that requests with a size just below that
806 threshold can be fulfilled without creating too many heaps.
812 #define THREAD_STATS 0
815 /* If THREAD_STATS is non-zero, some statistics on mutex locking are
821 Special defines for the Linux/GNU C library.
830 Void_t
* __default_morecore (ptrdiff_t);
831 Void_t
*(*__morecore
)(ptrdiff_t) = __default_morecore
;
835 Void_t
* __default_morecore ();
836 Void_t
*(*__morecore
)() = __default_morecore
;
840 #define MORECORE (*__morecore)
841 #define MORECORE_FAILURE 0
842 #define MORECORE_CLEARS 1
844 #define munmap __munmap
845 #define mremap __mremap
846 #define mprotect __mprotect
847 #undef malloc_getpagesize
848 #define malloc_getpagesize __getpagesize()
853 extern Void_t
* sbrk(ptrdiff_t);
855 extern Void_t
* sbrk();
859 #define MORECORE sbrk
862 #ifndef MORECORE_FAILURE
863 #define MORECORE_FAILURE -1
866 #ifndef MORECORE_CLEARS
867 #define MORECORE_CLEARS 1
874 #define cALLOc __libc_calloc
875 #define fREe __libc_free
876 #define mALLOc __libc_malloc
877 #define mEMALIGn __libc_memalign
878 #define rEALLOc __libc_realloc
879 #define vALLOc __libc_valloc
880 #define pvALLOc __libc_pvalloc
881 #define mALLINFo __libc_mallinfo
882 #define mALLOPt __libc_mallopt
883 #define mALLOC_STATs __malloc_stats
884 #define mALLOC_USABLE_SIZe __malloc_usable_size
885 #define mALLOC_TRIm __malloc_trim
886 #define mALLOC_GET_STATe __malloc_get_state
887 #define mALLOC_SET_STATe __malloc_set_state
891 #define cALLOc calloc
893 #define mALLOc malloc
894 #define mEMALIGn memalign
895 #define rEALLOc realloc
896 #define vALLOc valloc
897 #define pvALLOc pvalloc
898 #define mALLINFo mallinfo
899 #define mALLOPt mallopt
900 #define mALLOC_STATs malloc_stats
901 #define mALLOC_USABLE_SIZe malloc_usable_size
902 #define mALLOC_TRIm malloc_trim
903 #define mALLOC_GET_STATe malloc_get_state
904 #define mALLOC_SET_STATe malloc_set_state
908 /* Public routines */
913 void ptmalloc_init(void);
915 Void_t
* mALLOc(size_t);
917 Void_t
* rEALLOc(Void_t
*, size_t);
918 Void_t
* mEMALIGn(size_t, size_t);
919 Void_t
* vALLOc(size_t);
920 Void_t
* pvALLOc(size_t);
921 Void_t
* cALLOc(size_t, size_t);
923 int mALLOC_TRIm(size_t);
924 size_t mALLOC_USABLE_SIZe(Void_t
*);
925 void mALLOC_STATs(void);
926 int mALLOPt(int, int);
927 struct mallinfo
mALLINFo(void);
928 Void_t
* mALLOC_GET_STATe(void);
929 int mALLOC_SET_STATe(Void_t
*);
934 void ptmalloc_init();
945 size_t mALLOC_USABLE_SIZe();
948 struct mallinfo
mALLINFo();
949 Void_t
* mALLOC_GET_STATe();
950 int mALLOC_SET_STATe();
956 }; /* end of extern "C" */
959 #if !defined(NO_THREADS) && !HAVE_MMAP
960 "Can't have threads support without mmap"
971 INTERNAL_SIZE_T prev_size
; /* Size of previous chunk (if free). */
972 INTERNAL_SIZE_T size
; /* Size in bytes, including overhead. */
973 struct malloc_chunk
* fd
; /* double links -- used only if free. */
974 struct malloc_chunk
* bk
;
977 typedef struct malloc_chunk
* mchunkptr
;
981 malloc_chunk details:
983 (The following includes lightly edited explanations by Colin Plumb.)
985 Chunks of memory are maintained using a `boundary tag' method as
986 described in e.g., Knuth or Standish. (See the paper by Paul
987 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
988 survey of such techniques.) Sizes of free chunks are stored both
989 in the front of each chunk and at the end. This makes
990 consolidating fragmented chunks into bigger chunks very fast. The
991 size fields also hold bits representing whether chunks are free or
994 An allocated chunk looks like this:
997 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
998 | Size of previous chunk, if allocated | |
999 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1000 | Size of chunk, in bytes |P|
1001 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1002 | User data starts here... .
1004 . (malloc_usable_space() bytes) .
1006 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1008 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1011 Where "chunk" is the front of the chunk for the purpose of most of
1012 the malloc code, but "mem" is the pointer that is returned to the
1013 user. "Nextchunk" is the beginning of the next contiguous chunk.
1015 Chunks always begin on even word boundaries, so the mem portion
1016 (which is returned to the user) is also on an even word boundary, and
1017 thus double-word aligned.
1019 Free chunks are stored in circular doubly-linked lists, and look like this:
1021 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1022 | Size of previous chunk |
1023 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1024 `head:' | Size of chunk, in bytes |P|
1025 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1026 | Forward pointer to next chunk in list |
1027 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1028 | Back pointer to previous chunk in list |
1029 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1030 | Unused space (may be 0 bytes long) .
1033 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1034 `foot:' | Size of chunk, in bytes |
1035 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1037 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1038 chunk size (which is always a multiple of two words), is an in-use
1039 bit for the *previous* chunk. If that bit is *clear*, then the
1040 word before the current chunk size contains the previous chunk
1041 size, and can be used to find the front of the previous chunk.
1042 (The very first chunk allocated always has this bit set,
1043 preventing access to non-existent (or non-owned) memory.)
1045 Note that the `foot' of the current chunk is actually represented
1046 as the prev_size of the NEXT chunk. (This makes it easier to
1047 deal with alignments etc).
1049 The two exceptions to all this are
1051 1. The special chunk `top', which doesn't bother using the
1052 trailing size field since there is no
1053 next contiguous chunk that would have to index off it. (After
1054 initialization, `top' is forced to always exist. If it would
1055 become less than MINSIZE bytes long, it is replenished via
1058 2. Chunks allocated via mmap, which have the second-lowest-order
1059 bit (IS_MMAPPED) set in their size fields. Because they are
1060 never merged or traversed from any other chunk, they have no
1061 foot size or inuse information.
1063 Available chunks are kept in any of several places (all declared below):
1065 * `av': An array of chunks serving as bin headers for consolidated
1066 chunks. Each bin is doubly linked. The bins are approximately
1067 proportionally (log) spaced. There are a lot of these bins
1068 (128). This may look excessive, but works very well in
1069 practice. All procedures maintain the invariant that no
1070 consolidated chunk physically borders another one. Chunks in
1071 bins are kept in size order, with ties going to the
1072 approximately least recently used chunk.
1074 The chunks in each bin are maintained in decreasing sorted order by
1075 size. This is irrelevant for the small bins, which all contain
1076 the same-sized chunks, but facilitates best-fit allocation for
1077 larger chunks. (These lists are just sequential. Keeping them in
1078 order almost never requires enough traversal to warrant using
1079 fancier ordered data structures.) Chunks of the same size are
1080 linked with the most recently freed at the front, and allocations
1081 are taken from the back. This results in LRU or FIFO allocation
1082 order, which tends to give each chunk an equal opportunity to be
1083 consolidated with adjacent freed chunks, resulting in larger free
1084 chunks and less fragmentation.
1086 * `top': The top-most available chunk (i.e., the one bordering the
1087 end of available memory) is treated specially. It is never
1088 included in any bin, is used only if no other chunk is
1089 available, and is released back to the system if it is very
1090 large (see M_TRIM_THRESHOLD).
1092 * `last_remainder': A bin holding only the remainder of the
1093 most recently split (non-top) chunk. This bin is checked
1094 before other non-fitting chunks, so as to provide better
1095 locality for runs of sequentially allocated chunks.
1097 * Implicitly, through the host system's memory mapping tables.
1098 If supported, requests greater than a threshold are usually
1099 serviced via calls to mmap, and then later released via munmap.
1106 The bins are an array of pairs of pointers serving as the
1107 heads of (initially empty) doubly-linked lists of chunks, laid out
1108 in a way so that each pair can be treated as if it were in a
1109 malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1110 and chunks are the same).
1112 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1113 8 bytes apart. Larger bins are approximately logarithmically
1114 spaced. (See the table below.)
1122 4 bins of size 32768
1123 2 bins of size 262144
1124 1 bin of size what's left
1126 There is actually a little bit of slop in the numbers in bin_index
1127 for the sake of speed. This makes no difference elsewhere.
1129 The special chunks `top' and `last_remainder' get their own bins,
1130 (this is implemented via yet more trickery with the av array),
1131 although `top' is never properly linked to its bin since it is
1132 always handled specially.
1136 #define NAV 128 /* number of bins */
1138 typedef struct malloc_chunk
* mbinptr
;
1140 /* An arena is a configuration of malloc_chunks together with an array
1141 of bins. With multiple threads, it must be locked via a mutex
1142 before changing its data structures. One or more `heaps' are
1143 associated with each arena, except for the main_arena, which is
1144 associated only with the `main heap', i.e. the conventional free
1145 store obtained with calls to MORECORE() (usually sbrk). The `av'
1146 array is never mentioned directly in the code, but instead used via
1147 bin access macros. */
1149 typedef struct _arena
{
1150 mbinptr av
[2*NAV
+ 2];
1151 struct _arena
*next
;
1154 long stat_lock_direct
, stat_lock_loop
, stat_lock_wait
;
1160 /* A heap is a single contiguous memory region holding (coalesceable)
1161 malloc_chunks. It is allocated with mmap() and always starts at an
1162 address aligned to HEAP_MAX_SIZE. Not used unless compiling for
1163 multiple threads. */
1165 typedef struct _heap_info
{
1166 arena
*ar_ptr
; /* Arena for this heap. */
1167 struct _heap_info
*prev
; /* Previous heap. */
1168 size_t size
; /* Current size in bytes. */
1169 size_t pad
; /* Make sure the following data is properly aligned. */
1174 Static functions (forward declarations)
1179 static void chunk_free(arena
*ar_ptr
, mchunkptr p
) internal_function
;
1180 static mchunkptr
chunk_alloc(arena
*ar_ptr
, INTERNAL_SIZE_T size
)
1182 static mchunkptr
chunk_realloc(arena
*ar_ptr
, mchunkptr oldp
,
1183 INTERNAL_SIZE_T oldsize
, INTERNAL_SIZE_T nb
)
1185 static mchunkptr
chunk_align(arena
*ar_ptr
, INTERNAL_SIZE_T nb
,
1186 size_t alignment
) internal_function
;
1187 static int main_trim(size_t pad
) internal_function
;
1189 static int heap_trim(heap_info
*heap
, size_t pad
) internal_function
;
1191 #if defined _LIBC || defined MALLOC_HOOKS
1192 static Void_t
* malloc_check(size_t sz
, const Void_t
*caller
);
1193 static void free_check(Void_t
* mem
, const Void_t
*caller
);
1194 static Void_t
* realloc_check(Void_t
* oldmem
, size_t bytes
,
1195 const Void_t
*caller
);
1196 static Void_t
* memalign_check(size_t alignment
, size_t bytes
,
1197 const Void_t
*caller
);
1199 static Void_t
* malloc_starter(size_t sz
, const Void_t
*caller
);
1200 static void free_starter(Void_t
* mem
, const Void_t
*caller
);
1201 static Void_t
* malloc_atfork(size_t sz
, const Void_t
*caller
);
1202 static void free_atfork(Void_t
* mem
, const Void_t
*caller
);
1208 static void chunk_free();
1209 static mchunkptr
chunk_alloc();
1210 static mchunkptr
chunk_realloc();
1211 static mchunkptr
chunk_align();
1212 static int main_trim();
1214 static int heap_trim();
1216 #if defined _LIBC || defined MALLOC_HOOKS
1217 static Void_t
* malloc_check();
1218 static void free_check();
1219 static Void_t
* realloc_check();
1220 static Void_t
* memalign_check();
1222 static Void_t
* malloc_starter();
1223 static void free_starter();
1224 static Void_t
* malloc_atfork();
1225 static void free_atfork();
1231 /* On some platforms we can compile internal, not exported functions better.
1232 Let the environment provide a macro and define it to be empty if it
1233 is not available. */
1234 #ifndef internal_function
1235 # define internal_function
1240 /* sizes, alignments */
1242 #define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
1243 #define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
1244 #define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
1245 #define MINSIZE (sizeof(struct malloc_chunk))
1247 /* conversion from malloc headers to user pointers, and back */
1249 #define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1250 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1252 /* pad request bytes into a usable size */
1254 #define request2size(req) \
1255 (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1256 (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1257 (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1259 /* Check if m has acceptable alignment */
1261 #define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1267 Physical chunk operations
1271 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1273 #define PREV_INUSE 0x1
1275 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1277 #define IS_MMAPPED 0x2
1279 /* Bits to mask off when extracting size */
1281 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1284 /* Ptr to next physical malloc_chunk. */
1286 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1288 /* Ptr to previous physical malloc_chunk */
1290 #define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1293 /* Treat space at ptr + offset as a chunk */
1295 #define chunk_at_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
1301 Dealing with use bits
1304 /* extract p's inuse bit */
1307 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1309 /* extract inuse bit of previous chunk */
1311 #define prev_inuse(p) ((p)->size & PREV_INUSE)
1313 /* check for mmap()'ed chunk */
1315 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1317 /* set/clear chunk as in use without otherwise disturbing */
1319 #define set_inuse(p) \
1320 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1322 #define clear_inuse(p) \
1323 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1325 /* check/set/clear inuse bits in known places */
1327 #define inuse_bit_at_offset(p, s)\
1328 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1330 #define set_inuse_bit_at_offset(p, s)\
1331 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1333 #define clear_inuse_bit_at_offset(p, s)\
1334 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1340 Dealing with size fields
1343 /* Get size, ignoring use bits */
1345 #define chunksize(p) ((p)->size & ~(SIZE_BITS))
1347 /* Set size at head, without disturbing its use bit */
1349 #define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1351 /* Set size/use ignoring previous bits in header */
1353 #define set_head(p, s) ((p)->size = (s))
1355 /* Set size at footer (only when chunk is not in use) */
1357 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1365 #define bin_at(a, i) ((mbinptr)((char*)&(((a)->av)[2*(i) + 2]) - 2*SIZE_SZ))
1366 #define init_bin(a, i) ((a)->av[2*i+2] = (a)->av[2*i+3] = bin_at((a), i))
1367 #define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1368 #define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1371 The first 2 bins are never indexed. The corresponding av cells are instead
1372 used for bookkeeping. This is not to save space, but to simplify
1373 indexing, maintain locality, and avoid some initialization tests.
1376 #define binblocks(a) (bin_at(a,0)->size)/* bitvector of nonempty blocks */
1377 #define top(a) (bin_at(a,0)->fd) /* The topmost chunk */
1378 #define last_remainder(a) (bin_at(a,1)) /* remainder from last split */
1381 Because top initially points to its own bin with initial
1382 zero size, thus forcing extension on the first malloc request,
1383 we avoid having any special code in malloc to check whether
1384 it even exists yet. But we still need to in malloc_extend_top.
1387 #define initial_top(a) ((mchunkptr)bin_at(a, 0))
1391 /* field-extraction macros */
1393 #define first(b) ((b)->fd)
1394 #define last(b) ((b)->bk)
1400 #define bin_index(sz) \
1401 (((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3):\
1402 ((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6):\
1403 ((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9):\
1404 ((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12):\
1405 ((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15):\
1406 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18):\
1409 bins for chunks < 512 are all spaced 8 bytes apart, and hold
1410 identically sized chunks. This is exploited in malloc.
1413 #define MAX_SMALLBIN 63
1414 #define MAX_SMALLBIN_SIZE 512
1415 #define SMALLBIN_WIDTH 8
1417 #define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
1420 Requests are `small' if both the corresponding and the next bin are small
1423 #define is_small_request(nb) ((nb) < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1428 To help compensate for the large number of bins, a one-level index
1429 structure is used for bin-by-bin searching. `binblocks' is a
1430 one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1431 have any (possibly) non-empty bins, so they can be skipped over
1432 all at once during during traversals. The bits are NOT always
1433 cleared as soon as all bins in a block are empty, but instead only
1434 when all are noticed to be empty during traversal in malloc.
1437 #define BINBLOCKWIDTH 4 /* bins per block */
1439 /* bin<->block macros */
1441 #define idx2binblock(ix) ((unsigned)1 << ((ix) / BINBLOCKWIDTH))
1442 #define mark_binblock(a, ii) (binblocks(a) |= idx2binblock(ii))
1443 #define clear_binblock(a, ii) (binblocks(a) &= ~(idx2binblock(ii)))
1448 /* Static bookkeeping data */
1450 /* Helper macro to initialize bins */
1451 #define IAV(i) bin_at(&main_arena, i), bin_at(&main_arena, i)
1453 static arena main_arena
= {
1456 IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
1457 IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
1458 IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
1459 IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
1460 IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
1461 IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
1462 IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
1463 IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
1464 IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
1465 IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
1466 IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
1467 IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
1468 IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
1469 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1470 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1471 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1473 &main_arena
, /* next */
1476 0, 0, 0, /* stat_lock_direct, stat_lock_loop, stat_lock_wait */
1478 MUTEX_INITIALIZER
/* mutex */
1483 /* Thread specific data */
1486 static tsd_key_t arena_key
;
1487 static mutex_t list_lock
= MUTEX_INITIALIZER
;
1491 static int stat_n_heaps
= 0;
1492 #define THREAD_STAT(x) x
1494 #define THREAD_STAT(x) do ; while(0)
1497 /* variables holding tunable values */
1499 static unsigned long trim_threshold
= DEFAULT_TRIM_THRESHOLD
;
1500 static unsigned long top_pad
= DEFAULT_TOP_PAD
;
1501 static unsigned int n_mmaps_max
= DEFAULT_MMAP_MAX
;
1502 static unsigned long mmap_threshold
= DEFAULT_MMAP_THRESHOLD
;
1503 static int check_action
= DEFAULT_CHECK_ACTION
;
1505 /* The first value returned from sbrk */
1506 static char* sbrk_base
= (char*)(-1);
1508 /* The maximum memory obtained from system via sbrk */
1509 static unsigned long max_sbrked_mem
= 0;
1511 /* The maximum via either sbrk or mmap (too difficult to track with threads) */
1513 static unsigned long max_total_mem
= 0;
1516 /* The total memory obtained from system via sbrk */
1517 #define sbrked_mem (main_arena.size)
1519 /* Tracking mmaps */
1521 static unsigned int n_mmaps
= 0;
1522 static unsigned int max_n_mmaps
= 0;
1523 static unsigned long mmapped_mem
= 0;
1524 static unsigned long max_mmapped_mem
= 0;
1529 #define weak_variable
1531 /* In GNU libc we want the hook variables to be weak definitions to
1532 avoid a problem with Emacs. */
1533 #define weak_variable weak_function
1536 /* Already initialized? */
1537 int __malloc_initialized
= -1;
1542 /* The following two functions are registered via thread_atfork() to
1543 make sure that the mutexes remain in a consistent state in the
1544 fork()ed version of a thread. Also adapt the malloc and free hooks
1545 temporarily, because the `atfork' handler mechanism may use
1546 malloc/free internally (e.g. in LinuxThreads). */
1548 #if defined _LIBC || defined MALLOC_HOOKS
1549 static __malloc_ptr_t (*save_malloc_hook
) __MALLOC_P ((size_t __size
,
1550 const __malloc_ptr_t
));
1551 static void (*save_free_hook
) __MALLOC_P ((__malloc_ptr_t __ptr
,
1552 const __malloc_ptr_t
));
1553 static Void_t
* save_arena
;
1557 ptmalloc_lock_all
__MALLOC_P((void))
1561 (void)mutex_lock(&list_lock
);
1562 for(ar_ptr
= &main_arena
;;) {
1563 (void)mutex_lock(&ar_ptr
->mutex
);
1564 ar_ptr
= ar_ptr
->next
;
1565 if(ar_ptr
== &main_arena
) break;
1567 #if defined _LIBC || defined MALLOC_HOOKS
1568 save_malloc_hook
= __malloc_hook
;
1569 save_free_hook
= __free_hook
;
1570 __malloc_hook
= malloc_atfork
;
1571 __free_hook
= free_atfork
;
1572 /* Only the current thread may perform malloc/free calls now. */
1573 tsd_getspecific(arena_key
, save_arena
);
1574 tsd_setspecific(arena_key
, (Void_t
*)0);
1579 ptmalloc_unlock_all
__MALLOC_P((void))
1583 #if defined _LIBC || defined MALLOC_HOOKS
1584 tsd_setspecific(arena_key
, save_arena
);
1585 __malloc_hook
= save_malloc_hook
;
1586 __free_hook
= save_free_hook
;
1588 for(ar_ptr
= &main_arena
;;) {
1589 (void)mutex_unlock(&ar_ptr
->mutex
);
1590 ar_ptr
= ar_ptr
->next
;
1591 if(ar_ptr
== &main_arena
) break;
1593 (void)mutex_unlock(&list_lock
);
1597 ptmalloc_init_all
__MALLOC_P((void))
1601 #if defined _LIBC || defined MALLOC_HOOKS
1602 tsd_setspecific(arena_key
, save_arena
);
1603 __malloc_hook
= save_malloc_hook
;
1604 __free_hook
= save_free_hook
;
1606 for(ar_ptr
= &main_arena
;;) {
1607 (void)mutex_init(&ar_ptr
->mutex
);
1608 ar_ptr
= ar_ptr
->next
;
1609 if(ar_ptr
== &main_arena
) break;
1611 (void)mutex_init(&list_lock
);
1616 /* Initialization routine. */
1619 static void ptmalloc_init
__MALLOC_P ((void)) __attribute__ ((constructor
));
1623 ptmalloc_init
__MALLOC_P((void))
1626 ptmalloc_init
__MALLOC_P((void))
1629 #if defined _LIBC || defined MALLOC_HOOKS
1633 if(__malloc_initialized
>= 0) return;
1634 __malloc_initialized
= 0;
1636 #if defined _LIBC || defined MALLOC_HOOKS
1637 /* With some threads implementations, creating thread-specific data
1638 or initializing a mutex may call malloc() itself. Provide a
1639 simple starter version (realloc() won't work). */
1640 save_malloc_hook
= __malloc_hook
;
1641 save_free_hook
= __free_hook
;
1642 __malloc_hook
= malloc_starter
;
1643 __free_hook
= free_starter
;
1646 /* Initialize the pthreads interface. */
1647 if (__pthread_initialize
!= NULL
)
1648 __pthread_initialize();
1650 mutex_init(&main_arena
.mutex
);
1651 mutex_init(&list_lock
);
1652 tsd_key_create(&arena_key
, NULL
);
1653 tsd_setspecific(arena_key
, (Void_t
*)&main_arena
);
1654 thread_atfork(ptmalloc_lock_all
, ptmalloc_unlock_all
, ptmalloc_init_all
);
1655 #endif /* !defined NO_THREADS */
1656 #if defined _LIBC || defined MALLOC_HOOKS
1657 if((s
= getenv("MALLOC_TRIM_THRESHOLD_")))
1658 mALLOPt(M_TRIM_THRESHOLD
, atoi(s
));
1659 if((s
= getenv("MALLOC_TOP_PAD_")))
1660 mALLOPt(M_TOP_PAD
, atoi(s
));
1661 if((s
= getenv("MALLOC_MMAP_THRESHOLD_")))
1662 mALLOPt(M_MMAP_THRESHOLD
, atoi(s
));
1663 if((s
= getenv("MALLOC_MMAP_MAX_")))
1664 mALLOPt(M_MMAP_MAX
, atoi(s
));
1665 s
= getenv("MALLOC_CHECK_");
1667 __malloc_hook
= save_malloc_hook
;
1668 __free_hook
= save_free_hook
;
1671 if(s
[0]) mALLOPt(M_CHECK_ACTION
, (int)(s
[0] - '0'));
1672 __malloc_check_init();
1674 if(__malloc_initialize_hook
!= NULL
)
1675 (*__malloc_initialize_hook
)();
1677 __malloc_initialized
= 1;
1680 /* There are platforms (e.g. Hurd) with a link-time hook mechanism. */
1681 #ifdef thread_atfork_static
1682 thread_atfork_static(ptmalloc_lock_all
, ptmalloc_unlock_all
, \
1686 #if defined _LIBC || defined MALLOC_HOOKS
1688 /* Hooks for debugging versions. The initial hooks just call the
1689 initialization routine, then do the normal work. */
1693 malloc_hook_ini(size_t sz
, const __malloc_ptr_t caller
)
1696 malloc_hook_ini(size_t sz
)
1698 malloc_hook_ini(sz
) size_t sz
;
1702 __malloc_hook
= NULL
;
1703 __realloc_hook
= NULL
;
1704 __memalign_hook
= NULL
;
1711 realloc_hook_ini(Void_t
* ptr
, size_t sz
, const __malloc_ptr_t caller
)
1713 realloc_hook_ini(ptr
, sz
, caller
)
1714 Void_t
* ptr
; size_t sz
; const __malloc_ptr_t caller
;
1717 __malloc_hook
= NULL
;
1718 __realloc_hook
= NULL
;
1719 __memalign_hook
= NULL
;
1721 return rEALLOc(ptr
, sz
);
1726 memalign_hook_ini(size_t sz
, size_t alignment
, const __malloc_ptr_t caller
)
1728 memalign_hook_ini(sz
, alignment
, caller
)
1729 size_t sz
; size_t alignment
; const __malloc_ptr_t caller
;
1732 __malloc_hook
= NULL
;
1733 __realloc_hook
= NULL
;
1734 __memalign_hook
= NULL
;
1736 return mEMALIGn(sz
, alignment
);
1739 void weak_variable (*__malloc_initialize_hook
) __MALLOC_P ((void)) = NULL
;
1740 void weak_variable (*__free_hook
) __MALLOC_P ((__malloc_ptr_t __ptr
,
1741 const __malloc_ptr_t
)) = NULL
;
1742 __malloc_ptr_t
weak_variable (*__malloc_hook
)
1743 __MALLOC_P ((size_t __size
, const __malloc_ptr_t
)) = malloc_hook_ini
;
1744 __malloc_ptr_t
weak_variable (*__realloc_hook
)
1745 __MALLOC_P ((__malloc_ptr_t __ptr
, size_t __size
, const __malloc_ptr_t
))
1747 __malloc_ptr_t
weak_variable (*__memalign_hook
)
1748 __MALLOC_P ((size_t __size
, size_t __alignment
, const __malloc_ptr_t
))
1749 = memalign_hook_ini
;
1750 void weak_variable (*__after_morecore_hook
) __MALLOC_P ((void)) = NULL
;
1752 /* Activate a standard set of debugging hooks. */
1754 __malloc_check_init()
1756 __malloc_hook
= malloc_check
;
1757 __free_hook
= free_check
;
1758 __realloc_hook
= realloc_check
;
1759 __memalign_hook
= memalign_check
;
1760 if(check_action
== 1)
1761 fprintf(stderr
, "malloc: using debugging hooks\n");
1770 /* Routines dealing with mmap(). */
1774 #ifndef MAP_ANONYMOUS
1776 static int dev_zero_fd
= -1; /* Cached file descriptor for /dev/zero. */
1778 #define MMAP(size, prot, flags) ((dev_zero_fd < 0) ? \
1779 (dev_zero_fd = open("/dev/zero", O_RDWR), \
1780 mmap(0, (size), (prot), (flags), dev_zero_fd, 0)) : \
1781 mmap(0, (size), (prot), (flags), dev_zero_fd, 0))
1785 #define MMAP(size, prot, flags) \
1786 (mmap(0, (size), (prot), (flags)|MAP_ANONYMOUS, -1, 0))
1790 #if defined __GNUC__ && __GNUC__ >= 2
1791 /* This function is only called from one place, inline it. */
1797 mmap_chunk(size_t size
)
1799 mmap_chunk(size
) size_t size
;
1802 size_t page_mask
= malloc_getpagesize
- 1;
1805 if(n_mmaps
>= n_mmaps_max
) return 0; /* too many regions */
1807 /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1808 * there is no following chunk whose prev_size field could be used.
1810 size
= (size
+ SIZE_SZ
+ page_mask
) & ~page_mask
;
1812 p
= (mchunkptr
)MMAP(size
, PROT_READ
|PROT_WRITE
, MAP_PRIVATE
);
1813 if(p
== (mchunkptr
) MAP_FAILED
) return 0;
1816 if (n_mmaps
> max_n_mmaps
) max_n_mmaps
= n_mmaps
;
1818 /* We demand that eight bytes into a page must be 8-byte aligned. */
1819 assert(aligned_OK(chunk2mem(p
)));
1821 /* The offset to the start of the mmapped region is stored
1822 * in the prev_size field of the chunk; normally it is zero,
1823 * but that can be changed in memalign().
1826 set_head(p
, size
|IS_MMAPPED
);
1828 mmapped_mem
+= size
;
1829 if ((unsigned long)mmapped_mem
> (unsigned long)max_mmapped_mem
)
1830 max_mmapped_mem
= mmapped_mem
;
1832 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) > (unsigned long)max_total_mem
)
1833 max_total_mem
= mmapped_mem
+ sbrked_mem
;
1839 static void munmap_chunk(mchunkptr p
)
1841 static void munmap_chunk(p
) mchunkptr p
;
1844 INTERNAL_SIZE_T size
= chunksize(p
);
1847 assert (chunk_is_mmapped(p
));
1848 assert(! ((char*)p
>= sbrk_base
&& (char*)p
< sbrk_base
+ sbrked_mem
));
1849 assert((n_mmaps
> 0));
1850 assert(((p
->prev_size
+ size
) & (malloc_getpagesize
-1)) == 0);
1853 mmapped_mem
-= (size
+ p
->prev_size
);
1855 ret
= munmap((char *)p
- p
->prev_size
, size
+ p
->prev_size
);
1857 /* munmap returns non-zero on failure */
1864 static mchunkptr
mremap_chunk(mchunkptr p
, size_t new_size
)
1866 static mchunkptr
mremap_chunk(p
, new_size
) mchunkptr p
; size_t new_size
;
1869 size_t page_mask
= malloc_getpagesize
- 1;
1870 INTERNAL_SIZE_T offset
= p
->prev_size
;
1871 INTERNAL_SIZE_T size
= chunksize(p
);
1874 assert (chunk_is_mmapped(p
));
1875 assert(! ((char*)p
>= sbrk_base
&& (char*)p
< sbrk_base
+ sbrked_mem
));
1876 assert((n_mmaps
> 0));
1877 assert(((size
+ offset
) & (malloc_getpagesize
-1)) == 0);
1879 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1880 new_size
= (new_size
+ offset
+ SIZE_SZ
+ page_mask
) & ~page_mask
;
1882 cp
= (char *)mremap((char *)p
- offset
, size
+ offset
, new_size
,
1885 if (cp
== (char *)-1) return 0;
1887 p
= (mchunkptr
)(cp
+ offset
);
1889 assert(aligned_OK(chunk2mem(p
)));
1891 assert((p
->prev_size
== offset
));
1892 set_head(p
, (new_size
- offset
)|IS_MMAPPED
);
1894 mmapped_mem
-= size
+ offset
;
1895 mmapped_mem
+= new_size
;
1896 if ((unsigned long)mmapped_mem
> (unsigned long)max_mmapped_mem
)
1897 max_mmapped_mem
= mmapped_mem
;
1899 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) > (unsigned long)max_total_mem
)
1900 max_total_mem
= mmapped_mem
+ sbrked_mem
;
1905 #endif /* HAVE_MREMAP */
1907 #endif /* HAVE_MMAP */
1911 /* Managing heaps and arenas (for concurrent threads) */
1915 /* Create a new heap. size is automatically rounded up to a multiple
1916 of the page size. */
1921 new_heap(size_t size
)
1923 new_heap(size
) size_t size
;
1926 size_t page_mask
= malloc_getpagesize
- 1;
1931 if(size
+top_pad
< HEAP_MIN_SIZE
)
1932 size
= HEAP_MIN_SIZE
;
1933 else if(size
+top_pad
<= HEAP_MAX_SIZE
)
1935 else if(size
> HEAP_MAX_SIZE
)
1938 size
= HEAP_MAX_SIZE
;
1939 size
= (size
+ page_mask
) & ~page_mask
;
1941 /* A memory region aligned to a multiple of HEAP_MAX_SIZE is needed.
1942 No swap space needs to be reserved for the following large
1943 mapping (on Linux, this is the case for all non-writable mappings
1945 p1
= (char *)MMAP(HEAP_MAX_SIZE
<<1, PROT_NONE
, MAP_PRIVATE
|MAP_NORESERVE
);
1946 if(p1
== MAP_FAILED
)
1948 p2
= (char *)(((unsigned long)p1
+ HEAP_MAX_SIZE
) & ~(HEAP_MAX_SIZE
-1));
1951 munmap(p2
+ HEAP_MAX_SIZE
, HEAP_MAX_SIZE
- ul
);
1952 if(mprotect(p2
, size
, PROT_READ
|PROT_WRITE
) != 0) {
1953 munmap(p2
, HEAP_MAX_SIZE
);
1956 h
= (heap_info
*)p2
;
1958 THREAD_STAT(stat_n_heaps
++);
1962 /* Grow or shrink a heap. size is automatically rounded up to a
1963 multiple of the page size if it is positive. */
1967 grow_heap(heap_info
*h
, long diff
)
1969 grow_heap(h
, diff
) heap_info
*h
; long diff
;
1972 size_t page_mask
= malloc_getpagesize
- 1;
1976 diff
= (diff
+ page_mask
) & ~page_mask
;
1977 new_size
= (long)h
->size
+ diff
;
1978 if(new_size
> HEAP_MAX_SIZE
)
1980 if(mprotect((char *)h
+ h
->size
, diff
, PROT_READ
|PROT_WRITE
) != 0)
1983 new_size
= (long)h
->size
+ diff
;
1984 if(new_size
< (long)sizeof(*h
))
1986 if(mprotect((char *)h
+ new_size
, -diff
, PROT_NONE
) != 0)
1993 /* Delete a heap. */
1995 #define delete_heap(heap) munmap((char*)(heap), HEAP_MAX_SIZE)
1997 /* arena_get() acquires an arena and locks the corresponding mutex.
1998 First, try the one last locked successfully by this thread. (This
1999 is the common case and handled with a macro for speed.) Then, loop
2000 once over the circularly linked list of arenas. If no arena is
2001 readily available, create a new one. */
2003 #define arena_get(ptr, size) do { \
2004 Void_t *vptr = NULL; \
2005 ptr = (arena *)tsd_getspecific(arena_key, vptr); \
2006 if(ptr && !mutex_trylock(&ptr->mutex)) { \
2007 THREAD_STAT(++(ptr->stat_lock_direct)); \
2009 ptr = arena_get2(ptr, (size)); \
2015 arena_get2(arena
*a_tsd
, size_t size
)
2017 arena_get2(a_tsd
, size
) arena
*a_tsd
; size_t size
;
2024 unsigned long misalign
;
2027 a
= a_tsd
= &main_arena
;
2031 /* This can only happen while initializing the new arena. */
2032 (void)mutex_lock(&main_arena
.mutex
);
2033 THREAD_STAT(++(main_arena
.stat_lock_wait
));
2038 /* Check the global, circularly linked list for available arenas. */
2041 if(!mutex_trylock(&a
->mutex
)) {
2042 THREAD_STAT(++(a
->stat_lock_loop
));
2043 tsd_setspecific(arena_key
, (Void_t
*)a
);
2047 } while(a
!= a_tsd
);
2049 /* If not even the list_lock can be obtained, try again. This can
2050 happen during `atfork', or for example on systems where thread
2051 creation makes it temporarily impossible to obtain _any_
2053 if(mutex_trylock(&list_lock
)) {
2057 (void)mutex_unlock(&list_lock
);
2059 /* Nothing immediately available, so generate a new arena. */
2060 h
= new_heap(size
+ (sizeof(*h
) + sizeof(*a
) + MALLOC_ALIGNMENT
));
2063 a
= h
->ar_ptr
= (arena
*)(h
+1);
2064 for(i
=0; i
<NAV
; i
++)
2068 tsd_setspecific(arena_key
, (Void_t
*)a
);
2069 mutex_init(&a
->mutex
);
2070 i
= mutex_lock(&a
->mutex
); /* remember result */
2072 /* Set up the top chunk, with proper alignment. */
2073 ptr
= (char *)(a
+ 1);
2074 misalign
= (unsigned long)chunk2mem(ptr
) & MALLOC_ALIGN_MASK
;
2076 ptr
+= MALLOC_ALIGNMENT
- misalign
;
2077 top(a
) = (mchunkptr
)ptr
;
2078 set_head(top(a
), (((char*)h
+ h
->size
) - ptr
) | PREV_INUSE
);
2080 /* Add the new arena to the list. */
2081 (void)mutex_lock(&list_lock
);
2082 a
->next
= main_arena
.next
;
2083 main_arena
.next
= a
;
2084 (void)mutex_unlock(&list_lock
);
2086 if(i
) /* locking failed; keep arena for further attempts later */
2089 THREAD_STAT(++(a
->stat_lock_loop
));
2093 /* find the heap and corresponding arena for a given ptr */
2095 #define heap_for_ptr(ptr) \
2096 ((heap_info *)((unsigned long)(ptr) & ~(HEAP_MAX_SIZE-1)))
2097 #define arena_for_ptr(ptr) \
2098 (((mchunkptr)(ptr) < top(&main_arena) && (char *)(ptr) >= sbrk_base) ? \
2099 &main_arena : heap_for_ptr(ptr)->ar_ptr)
2101 #else /* defined(NO_THREADS) */
2103 /* Without concurrent threads, there is only one arena. */
2105 #define arena_get(ptr, sz) (ptr = &main_arena)
2106 #define arena_for_ptr(ptr) (&main_arena)
2108 #endif /* !defined(NO_THREADS) */
2120 These routines make a number of assertions about the states
2121 of data structures that should be true at all times. If any
2122 are not true, it's very likely that a user program has somehow
2123 trashed memory. (It's also possible that there is a coding error
2124 in malloc. In which case, please report it!)
2128 static void do_check_chunk(arena
*ar_ptr
, mchunkptr p
)
2130 static void do_check_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
2133 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
2135 /* No checkable chunk is mmapped */
2136 assert(!chunk_is_mmapped(p
));
2139 if(ar_ptr
!= &main_arena
) {
2140 heap_info
*heap
= heap_for_ptr(p
);
2141 assert(heap
->ar_ptr
== ar_ptr
);
2142 assert((char *)p
+ sz
<= (char *)heap
+ heap
->size
);
2147 /* Check for legal address ... */
2148 assert((char*)p
>= sbrk_base
);
2149 if (p
!= top(ar_ptr
))
2150 assert((char*)p
+ sz
<= (char*)top(ar_ptr
));
2152 assert((char*)p
+ sz
<= sbrk_base
+ sbrked_mem
);
2158 static void do_check_free_chunk(arena
*ar_ptr
, mchunkptr p
)
2160 static void do_check_free_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
2163 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
2164 mchunkptr next
= chunk_at_offset(p
, sz
);
2166 do_check_chunk(ar_ptr
, p
);
2168 /* Check whether it claims to be free ... */
2171 /* Must have OK size and fields */
2172 assert((long)sz
>= (long)MINSIZE
);
2173 assert((sz
& MALLOC_ALIGN_MASK
) == 0);
2174 assert(aligned_OK(chunk2mem(p
)));
2175 /* ... matching footer field */
2176 assert(next
->prev_size
== sz
);
2177 /* ... and is fully consolidated */
2178 assert(prev_inuse(p
));
2179 assert (next
== top(ar_ptr
) || inuse(next
));
2181 /* ... and has minimally sane links */
2182 assert(p
->fd
->bk
== p
);
2183 assert(p
->bk
->fd
== p
);
2187 static void do_check_inuse_chunk(arena
*ar_ptr
, mchunkptr p
)
2189 static void do_check_inuse_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
2192 mchunkptr next
= next_chunk(p
);
2193 do_check_chunk(ar_ptr
, p
);
2195 /* Check whether it claims to be in use ... */
2198 /* ... whether its size is OK (it might be a fencepost) ... */
2199 assert(chunksize(p
) >= MINSIZE
|| next
->size
== (0|PREV_INUSE
));
2201 /* ... and is surrounded by OK chunks.
2202 Since more things can be checked with free chunks than inuse ones,
2203 if an inuse chunk borders them and debug is on, it's worth doing them.
2207 mchunkptr prv
= prev_chunk(p
);
2208 assert(next_chunk(prv
) == p
);
2209 do_check_free_chunk(ar_ptr
, prv
);
2211 if (next
== top(ar_ptr
))
2213 assert(prev_inuse(next
));
2214 assert(chunksize(next
) >= MINSIZE
);
2216 else if (!inuse(next
))
2217 do_check_free_chunk(ar_ptr
, next
);
2222 static void do_check_malloced_chunk(arena
*ar_ptr
,
2223 mchunkptr p
, INTERNAL_SIZE_T s
)
2225 static void do_check_malloced_chunk(ar_ptr
, p
, s
)
2226 arena
*ar_ptr
; mchunkptr p
; INTERNAL_SIZE_T s
;
2229 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
2232 do_check_inuse_chunk(ar_ptr
, p
);
2234 /* Legal size ... */
2235 assert((long)sz
>= (long)MINSIZE
);
2236 assert((sz
& MALLOC_ALIGN_MASK
) == 0);
2238 assert(room
< (long)MINSIZE
);
2240 /* ... and alignment */
2241 assert(aligned_OK(chunk2mem(p
)));
2244 /* ... and was allocated at front of an available chunk */
2245 assert(prev_inuse(p
));
2250 #define check_free_chunk(A,P) do_check_free_chunk(A,P)
2251 #define check_inuse_chunk(A,P) do_check_inuse_chunk(A,P)
2252 #define check_chunk(A,P) do_check_chunk(A,P)
2253 #define check_malloced_chunk(A,P,N) do_check_malloced_chunk(A,P,N)
2255 #define check_free_chunk(A,P)
2256 #define check_inuse_chunk(A,P)
2257 #define check_chunk(A,P)
2258 #define check_malloced_chunk(A,P,N)
2264 Macro-based internal utilities
2269 Linking chunks in bin lists.
2270 Call these only with variables, not arbitrary expressions, as arguments.
2274 Place chunk p of size s in its bin, in size order,
2275 putting it ahead of others of same size.
2279 #define frontlink(A, P, S, IDX, BK, FD) \
2281 if (S < MAX_SMALLBIN_SIZE) \
2283 IDX = smallbin_index(S); \
2284 mark_binblock(A, IDX); \
2285 BK = bin_at(A, IDX); \
2289 FD->bk = BK->fd = P; \
2293 IDX = bin_index(S); \
2294 BK = bin_at(A, IDX); \
2296 if (FD == BK) mark_binblock(A, IDX); \
2299 while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
2304 FD->bk = BK->fd = P; \
2309 /* take a chunk off a list */
2311 #define unlink(P, BK, FD) \
2319 /* Place p as the last remainder */
2321 #define link_last_remainder(A, P) \
2323 last_remainder(A)->fd = last_remainder(A)->bk = P; \
2324 P->fd = P->bk = last_remainder(A); \
2327 /* Clear the last_remainder bin */
2329 #define clear_last_remainder(A) \
2330 (last_remainder(A)->fd = last_remainder(A)->bk = last_remainder(A))
2337 Extend the top-most chunk by obtaining memory from system.
2338 Main interface to sbrk (but see also malloc_trim).
2341 #if defined __GNUC__ && __GNUC__ >= 2
2342 /* This function is called only from one place, inline it. */
2348 malloc_extend_top(arena
*ar_ptr
, INTERNAL_SIZE_T nb
)
2350 malloc_extend_top(ar_ptr
, nb
) arena
*ar_ptr
; INTERNAL_SIZE_T nb
;
2353 unsigned long pagesz
= malloc_getpagesize
;
2354 mchunkptr old_top
= top(ar_ptr
); /* Record state of old top */
2355 INTERNAL_SIZE_T old_top_size
= chunksize(old_top
);
2356 INTERNAL_SIZE_T top_size
; /* new size of top chunk */
2359 if(ar_ptr
== &main_arena
) {
2362 char* brk
; /* return value from sbrk */
2363 INTERNAL_SIZE_T front_misalign
; /* unusable bytes at front of sbrked space */
2364 INTERNAL_SIZE_T correction
; /* bytes for 2nd sbrk call */
2365 char* new_brk
; /* return of 2nd sbrk call */
2366 char* old_end
= (char*)(chunk_at_offset(old_top
, old_top_size
));
2368 /* Pad request with top_pad plus minimal overhead */
2369 INTERNAL_SIZE_T sbrk_size
= nb
+ top_pad
+ MINSIZE
;
2371 /* If not the first time through, round to preserve page boundary */
2372 /* Otherwise, we need to correct to a page size below anyway. */
2373 /* (We also correct below if an intervening foreign sbrk call.) */
2375 if (sbrk_base
!= (char*)(-1))
2376 sbrk_size
= (sbrk_size
+ (pagesz
- 1)) & ~(pagesz
- 1);
2378 brk
= (char*)(MORECORE (sbrk_size
));
2380 /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2381 if (brk
== (char*)(MORECORE_FAILURE
) ||
2382 (brk
< old_end
&& old_top
!= initial_top(&main_arena
)))
2385 #if defined _LIBC || defined MALLOC_HOOKS
2386 /* Call the `morecore' hook if necessary. */
2387 if (__after_morecore_hook
)
2388 (*__after_morecore_hook
) ();
2391 sbrked_mem
+= sbrk_size
;
2393 if (brk
== old_end
) { /* can just add bytes to current top */
2394 top_size
= sbrk_size
+ old_top_size
;
2395 set_head(old_top
, top_size
| PREV_INUSE
);
2396 old_top
= 0; /* don't free below */
2398 if (sbrk_base
== (char*)(-1)) /* First time through. Record base */
2401 /* Someone else called sbrk(). Count those bytes as sbrked_mem. */
2402 sbrked_mem
+= brk
- (char*)old_end
;
2404 /* Guarantee alignment of first new chunk made from this space */
2405 front_misalign
= (unsigned long)chunk2mem(brk
) & MALLOC_ALIGN_MASK
;
2406 if (front_misalign
> 0) {
2407 correction
= (MALLOC_ALIGNMENT
) - front_misalign
;
2412 /* Guarantee the next brk will be at a page boundary */
2413 correction
+= pagesz
- ((unsigned long)(brk
+ sbrk_size
) & (pagesz
- 1));
2415 /* Allocate correction */
2416 new_brk
= (char*)(MORECORE (correction
));
2417 if (new_brk
== (char*)(MORECORE_FAILURE
)) return;
2419 #if defined _LIBC || defined MALLOC_HOOKS
2420 /* Call the `morecore' hook if necessary. */
2421 if (__after_morecore_hook
)
2422 (*__after_morecore_hook
) ();
2425 sbrked_mem
+= correction
;
2427 top(&main_arena
) = (mchunkptr
)brk
;
2428 top_size
= new_brk
- brk
+ correction
;
2429 set_head(top(&main_arena
), top_size
| PREV_INUSE
);
2431 if (old_top
== initial_top(&main_arena
))
2432 old_top
= 0; /* don't free below */
2435 if ((unsigned long)sbrked_mem
> (unsigned long)max_sbrked_mem
)
2436 max_sbrked_mem
= sbrked_mem
;
2438 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) >
2439 (unsigned long)max_total_mem
)
2440 max_total_mem
= mmapped_mem
+ sbrked_mem
;
2444 } else { /* ar_ptr != &main_arena */
2445 heap_info
*old_heap
, *heap
;
2446 size_t old_heap_size
;
2448 if(old_top_size
< MINSIZE
) /* this should never happen */
2451 /* First try to extend the current heap. */
2452 if(MINSIZE
+ nb
<= old_top_size
)
2454 old_heap
= heap_for_ptr(old_top
);
2455 old_heap_size
= old_heap
->size
;
2456 if(grow_heap(old_heap
, MINSIZE
+ nb
- old_top_size
) == 0) {
2457 ar_ptr
->size
+= old_heap
->size
- old_heap_size
;
2458 top_size
= ((char *)old_heap
+ old_heap
->size
) - (char *)old_top
;
2459 set_head(old_top
, top_size
| PREV_INUSE
);
2463 /* A new heap must be created. */
2464 heap
= new_heap(nb
+ (MINSIZE
+ sizeof(*heap
)));
2467 heap
->ar_ptr
= ar_ptr
;
2468 heap
->prev
= old_heap
;
2469 ar_ptr
->size
+= heap
->size
;
2471 /* Set up the new top, so we can safely use chunk_free() below. */
2472 top(ar_ptr
) = chunk_at_offset(heap
, sizeof(*heap
));
2473 top_size
= heap
->size
- sizeof(*heap
);
2474 set_head(top(ar_ptr
), top_size
| PREV_INUSE
);
2476 #endif /* !defined(NO_THREADS) */
2478 /* We always land on a page boundary */
2479 assert(((unsigned long)((char*)top(ar_ptr
) + top_size
) & (pagesz
-1)) == 0);
2481 /* Setup fencepost and free the old top chunk. */
2483 /* The fencepost takes at least MINSIZE bytes, because it might
2484 become the top chunk again later. Note that a footer is set
2485 up, too, although the chunk is marked in use. */
2486 old_top_size
-= MINSIZE
;
2487 set_head(chunk_at_offset(old_top
, old_top_size
+ 2*SIZE_SZ
), 0|PREV_INUSE
);
2488 if(old_top_size
>= MINSIZE
) {
2489 set_head(chunk_at_offset(old_top
, old_top_size
), (2*SIZE_SZ
)|PREV_INUSE
);
2490 set_foot(chunk_at_offset(old_top
, old_top_size
), (2*SIZE_SZ
));
2491 set_head_size(old_top
, old_top_size
);
2492 chunk_free(ar_ptr
, old_top
);
2494 set_head(old_top
, (old_top_size
+ 2*SIZE_SZ
)|PREV_INUSE
);
2495 set_foot(old_top
, (old_top_size
+ 2*SIZE_SZ
));
2503 /* Main public routines */
2509 The requested size is first converted into a usable form, `nb'.
2510 This currently means to add 4 bytes overhead plus possibly more to
2511 obtain 8-byte alignment and/or to obtain a size of at least
2512 MINSIZE (currently 16, 24, or 32 bytes), the smallest allocatable
2513 size. (All fits are considered `exact' if they are within MINSIZE
2516 From there, the first successful of the following steps is taken:
2518 1. The bin corresponding to the request size is scanned, and if
2519 a chunk of exactly the right size is found, it is taken.
2521 2. The most recently remaindered chunk is used if it is big
2522 enough. This is a form of (roving) first fit, used only in
2523 the absence of exact fits. Runs of consecutive requests use
2524 the remainder of the chunk used for the previous such request
2525 whenever possible. This limited use of a first-fit style
2526 allocation strategy tends to give contiguous chunks
2527 coextensive lifetimes, which improves locality and can reduce
2528 fragmentation in the long run.
2530 3. Other bins are scanned in increasing size order, using a
2531 chunk big enough to fulfill the request, and splitting off
2532 any remainder. This search is strictly by best-fit; i.e.,
2533 the smallest (with ties going to approximately the least
2534 recently used) chunk that fits is selected.
2536 4. If large enough, the chunk bordering the end of memory
2537 (`top') is split off. (This use of `top' is in accord with
2538 the best-fit search rule. In effect, `top' is treated as
2539 larger (and thus less well fitting) than any other available
2540 chunk since it can be extended to be as large as necessary
2541 (up to system limitations).
2543 5. If the request size meets the mmap threshold and the
2544 system supports mmap, and there are few enough currently
2545 allocated mmapped regions, and a call to mmap succeeds,
2546 the request is allocated via direct memory mapping.
2548 6. Otherwise, the top of memory is extended by
2549 obtaining more space from the system (normally using sbrk,
2550 but definable to anything else via the MORECORE macro).
2551 Memory is gathered from the system (in system page-sized
2552 units) in a way that allows chunks obtained across different
2553 sbrk calls to be consolidated, but does not require
2554 contiguous memory. Thus, it should be safe to intersperse
2555 mallocs with other sbrk calls.
2558 All allocations are made from the `lowest' part of any found
2559 chunk. (The implementation invariant is that prev_inuse is
2560 always true of any allocated chunk; i.e., that each allocated
2561 chunk borders either a previously allocated and still in-use chunk,
2562 or the base of its memory arena.)
2567 Void_t
* mALLOc(size_t bytes
)
2569 Void_t
* mALLOc(bytes
) size_t bytes
;
2573 INTERNAL_SIZE_T nb
; /* padded request size */
2576 #if defined _LIBC || defined MALLOC_HOOKS
2577 if (__malloc_hook
!= NULL
) {
2580 #if defined __GNUC__ && __GNUC__ >= 2
2581 result
= (*__malloc_hook
)(bytes
, __builtin_return_address (0));
2583 result
= (*__malloc_hook
)(bytes
, NULL
);
2589 nb
= request2size(bytes
);
2590 arena_get(ar_ptr
, nb
);
2593 victim
= chunk_alloc(ar_ptr
, nb
);
2594 (void)mutex_unlock(&ar_ptr
->mutex
);
2596 /* Maybe the failure is due to running out of mmapped areas. */
2597 if(ar_ptr
!= &main_arena
) {
2598 (void)mutex_lock(&main_arena
.mutex
);
2599 victim
= chunk_alloc(&main_arena
, nb
);
2600 (void)mutex_unlock(&main_arena
.mutex
);
2602 if(!victim
) return 0;
2604 return chunk2mem(victim
);
2610 chunk_alloc(arena
*ar_ptr
, INTERNAL_SIZE_T nb
)
2612 chunk_alloc(ar_ptr
, nb
) arena
*ar_ptr
; INTERNAL_SIZE_T nb
;
2615 mchunkptr victim
; /* inspected/selected chunk */
2616 INTERNAL_SIZE_T victim_size
; /* its size */
2617 int idx
; /* index for bin traversal */
2618 mbinptr bin
; /* associated bin */
2619 mchunkptr remainder
; /* remainder from a split */
2620 long remainder_size
; /* its size */
2621 int remainder_index
; /* its bin index */
2622 unsigned long block
; /* block traverser bit */
2623 int startidx
; /* first bin of a traversed block */
2624 mchunkptr fwd
; /* misc temp for linking */
2625 mchunkptr bck
; /* misc temp for linking */
2626 mbinptr q
; /* misc temp */
2629 /* Check for exact match in a bin */
2631 if (is_small_request(nb
)) /* Faster version for small requests */
2633 idx
= smallbin_index(nb
);
2635 /* No traversal or size check necessary for small bins. */
2637 q
= bin_at(ar_ptr
, idx
);
2640 /* Also scan the next one, since it would have a remainder < MINSIZE */
2648 victim_size
= chunksize(victim
);
2649 unlink(victim
, bck
, fwd
);
2650 set_inuse_bit_at_offset(victim
, victim_size
);
2651 check_malloced_chunk(ar_ptr
, victim
, nb
);
2655 idx
+= 2; /* Set for bin scan below. We've already scanned 2 bins. */
2660 idx
= bin_index(nb
);
2661 bin
= bin_at(ar_ptr
, idx
);
2663 for (victim
= last(bin
); victim
!= bin
; victim
= victim
->bk
)
2665 victim_size
= chunksize(victim
);
2666 remainder_size
= victim_size
- nb
;
2668 if (remainder_size
>= (long)MINSIZE
) /* too big */
2670 --idx
; /* adjust to rescan below after checking last remainder */
2674 else if (remainder_size
>= 0) /* exact fit */
2676 unlink(victim
, bck
, fwd
);
2677 set_inuse_bit_at_offset(victim
, victim_size
);
2678 check_malloced_chunk(ar_ptr
, victim
, nb
);
2687 /* Try to use the last split-off remainder */
2689 if ( (victim
= last_remainder(ar_ptr
)->fd
) != last_remainder(ar_ptr
))
2691 victim_size
= chunksize(victim
);
2692 remainder_size
= victim_size
- nb
;
2694 if (remainder_size
>= (long)MINSIZE
) /* re-split */
2696 remainder
= chunk_at_offset(victim
, nb
);
2697 set_head(victim
, nb
| PREV_INUSE
);
2698 link_last_remainder(ar_ptr
, remainder
);
2699 set_head(remainder
, remainder_size
| PREV_INUSE
);
2700 set_foot(remainder
, remainder_size
);
2701 check_malloced_chunk(ar_ptr
, victim
, nb
);
2705 clear_last_remainder(ar_ptr
);
2707 if (remainder_size
>= 0) /* exhaust */
2709 set_inuse_bit_at_offset(victim
, victim_size
);
2710 check_malloced_chunk(ar_ptr
, victim
, nb
);
2714 /* Else place in bin */
2716 frontlink(ar_ptr
, victim
, victim_size
, remainder_index
, bck
, fwd
);
2720 If there are any possibly nonempty big-enough blocks,
2721 search for best fitting chunk by scanning bins in blockwidth units.
2724 if ( (block
= idx2binblock(idx
)) <= binblocks(ar_ptr
))
2727 /* Get to the first marked block */
2729 if ( (block
& binblocks(ar_ptr
)) == 0)
2731 /* force to an even block boundary */
2732 idx
= (idx
& ~(BINBLOCKWIDTH
- 1)) + BINBLOCKWIDTH
;
2734 while ((block
& binblocks(ar_ptr
)) == 0)
2736 idx
+= BINBLOCKWIDTH
;
2741 /* For each possibly nonempty block ... */
2744 startidx
= idx
; /* (track incomplete blocks) */
2745 q
= bin
= bin_at(ar_ptr
, idx
);
2747 /* For each bin in this block ... */
2750 /* Find and use first big enough chunk ... */
2752 for (victim
= last(bin
); victim
!= bin
; victim
= victim
->bk
)
2754 victim_size
= chunksize(victim
);
2755 remainder_size
= victim_size
- nb
;
2757 if (remainder_size
>= (long)MINSIZE
) /* split */
2759 remainder
= chunk_at_offset(victim
, nb
);
2760 set_head(victim
, nb
| PREV_INUSE
);
2761 unlink(victim
, bck
, fwd
);
2762 link_last_remainder(ar_ptr
, remainder
);
2763 set_head(remainder
, remainder_size
| PREV_INUSE
);
2764 set_foot(remainder
, remainder_size
);
2765 check_malloced_chunk(ar_ptr
, victim
, nb
);
2769 else if (remainder_size
>= 0) /* take */
2771 set_inuse_bit_at_offset(victim
, victim_size
);
2772 unlink(victim
, bck
, fwd
);
2773 check_malloced_chunk(ar_ptr
, victim
, nb
);
2779 bin
= next_bin(bin
);
2781 } while ((++idx
& (BINBLOCKWIDTH
- 1)) != 0);
2783 /* Clear out the block bit. */
2785 do /* Possibly backtrack to try to clear a partial block */
2787 if ((startidx
& (BINBLOCKWIDTH
- 1)) == 0)
2789 binblocks(ar_ptr
) &= ~block
;
2794 } while (first(q
) == q
);
2796 /* Get to the next possibly nonempty block */
2798 if ( (block
<<= 1) <= binblocks(ar_ptr
) && (block
!= 0) )
2800 while ((block
& binblocks(ar_ptr
)) == 0)
2802 idx
+= BINBLOCKWIDTH
;
2812 /* Try to use top chunk */
2814 /* Require that there be a remainder, ensuring top always exists */
2815 if ( (remainder_size
= chunksize(top(ar_ptr
)) - nb
) < (long)MINSIZE
)
2819 /* If big and would otherwise need to extend, try to use mmap instead */
2820 if ((unsigned long)nb
>= (unsigned long)mmap_threshold
&&
2821 (victim
= mmap_chunk(nb
)) != 0)
2826 malloc_extend_top(ar_ptr
, nb
);
2827 if ((remainder_size
= chunksize(top(ar_ptr
)) - nb
) < (long)MINSIZE
)
2828 return 0; /* propagate failure */
2831 victim
= top(ar_ptr
);
2832 set_head(victim
, nb
| PREV_INUSE
);
2833 top(ar_ptr
) = chunk_at_offset(victim
, nb
);
2834 set_head(top(ar_ptr
), remainder_size
| PREV_INUSE
);
2835 check_malloced_chunk(ar_ptr
, victim
, nb
);
2849 1. free(0) has no effect.
2851 2. If the chunk was allocated via mmap, it is released via munmap().
2853 3. If a returned chunk borders the current high end of memory,
2854 it is consolidated into the top, and if the total unused
2855 topmost memory exceeds the trim threshold, malloc_trim is
2858 4. Other chunks are consolidated as they arrive, and
2859 placed in corresponding bins. (This includes the case of
2860 consolidating with the current `last_remainder').
2866 void fREe(Void_t
* mem
)
2868 void fREe(mem
) Void_t
* mem
;
2872 mchunkptr p
; /* chunk corresponding to mem */
2874 #if defined _LIBC || defined MALLOC_HOOKS
2875 if (__free_hook
!= NULL
) {
2876 #if defined __GNUC__ && __GNUC__ >= 2
2877 (*__free_hook
)(mem
, __builtin_return_address (0));
2879 (*__free_hook
)(mem
, NULL
);
2885 if (mem
== 0) /* free(0) has no effect */
2891 if (chunk_is_mmapped(p
)) /* release mmapped memory. */
2898 ar_ptr
= arena_for_ptr(p
);
2900 if(!mutex_trylock(&ar_ptr
->mutex
))
2901 ++(ar_ptr
->stat_lock_direct
);
2903 (void)mutex_lock(&ar_ptr
->mutex
);
2904 ++(ar_ptr
->stat_lock_wait
);
2907 (void)mutex_lock(&ar_ptr
->mutex
);
2909 chunk_free(ar_ptr
, p
);
2910 (void)mutex_unlock(&ar_ptr
->mutex
);
2916 chunk_free(arena
*ar_ptr
, mchunkptr p
)
2918 chunk_free(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
2921 INTERNAL_SIZE_T hd
= p
->size
; /* its head field */
2922 INTERNAL_SIZE_T sz
; /* its size */
2923 int idx
; /* its bin index */
2924 mchunkptr next
; /* next contiguous chunk */
2925 INTERNAL_SIZE_T nextsz
; /* its size */
2926 INTERNAL_SIZE_T prevsz
; /* size of previous contiguous chunk */
2927 mchunkptr bck
; /* misc temp for linking */
2928 mchunkptr fwd
; /* misc temp for linking */
2929 int islr
; /* track whether merging with last_remainder */
2931 check_inuse_chunk(ar_ptr
, p
);
2933 sz
= hd
& ~PREV_INUSE
;
2934 next
= chunk_at_offset(p
, sz
);
2935 nextsz
= chunksize(next
);
2937 if (next
== top(ar_ptr
)) /* merge with top */
2941 if (!(hd
& PREV_INUSE
)) /* consolidate backward */
2943 prevsz
= p
->prev_size
;
2944 p
= chunk_at_offset(p
, -prevsz
);
2946 unlink(p
, bck
, fwd
);
2949 set_head(p
, sz
| PREV_INUSE
);
2953 if(ar_ptr
== &main_arena
) {
2955 if ((unsigned long)(sz
) >= (unsigned long)trim_threshold
)
2959 heap_info
*heap
= heap_for_ptr(p
);
2961 assert(heap
->ar_ptr
== ar_ptr
);
2963 /* Try to get rid of completely empty heaps, if possible. */
2964 if((unsigned long)(sz
) >= (unsigned long)trim_threshold
||
2965 p
== chunk_at_offset(heap
, sizeof(*heap
)))
2966 heap_trim(heap
, top_pad
);
2974 if (!(hd
& PREV_INUSE
)) /* consolidate backward */
2976 prevsz
= p
->prev_size
;
2977 p
= chunk_at_offset(p
, -prevsz
);
2980 if (p
->fd
== last_remainder(ar_ptr
)) /* keep as last_remainder */
2983 unlink(p
, bck
, fwd
);
2986 if (!(inuse_bit_at_offset(next
, nextsz
))) /* consolidate forward */
2990 if (!islr
&& next
->fd
== last_remainder(ar_ptr
))
2991 /* re-insert last_remainder */
2994 link_last_remainder(ar_ptr
, p
);
2997 unlink(next
, bck
, fwd
);
2999 next
= chunk_at_offset(p
, sz
);
3002 set_head(next
, nextsz
); /* clear inuse bit */
3004 set_head(p
, sz
| PREV_INUSE
);
3005 next
->prev_size
= sz
;
3007 frontlink(ar_ptr
, p
, sz
, idx
, bck
, fwd
);
3010 /* Check whether the heap containing top can go away now. */
3011 if(next
->size
< MINSIZE
&&
3012 (unsigned long)sz
> trim_threshold
&&
3013 ar_ptr
!= &main_arena
) { /* fencepost */
3014 heap_info
* heap
= heap_for_ptr(top(ar_ptr
));
3016 if(top(ar_ptr
) == chunk_at_offset(heap
, sizeof(*heap
)) &&
3017 heap
->prev
== heap_for_ptr(p
))
3018 heap_trim(heap
, top_pad
);
3031 Chunks that were obtained via mmap cannot be extended or shrunk
3032 unless HAVE_MREMAP is defined, in which case mremap is used.
3033 Otherwise, if their reallocation is for additional space, they are
3034 copied. If for less, they are just left alone.
3036 Otherwise, if the reallocation is for additional space, and the
3037 chunk can be extended, it is, else a malloc-copy-free sequence is
3038 taken. There are several different ways that a chunk could be
3039 extended. All are tried:
3041 * Extending forward into following adjacent free chunk.
3042 * Shifting backwards, joining preceding adjacent space
3043 * Both shifting backwards and extending forward.
3044 * Extending into newly sbrked space
3046 Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
3047 size argument of zero (re)allocates a minimum-sized chunk.
3049 If the reallocation is for less space, and the new request is for
3050 a `small' (<512 bytes) size, then the newly unused space is lopped
3053 The old unix realloc convention of allowing the last-free'd chunk
3054 to be used as an argument to realloc is no longer supported.
3055 I don't know of any programs still relying on this feature,
3056 and allowing it would also allow too many other incorrect
3057 usages of realloc to be sensible.
3064 Void_t
* rEALLOc(Void_t
* oldmem
, size_t bytes
)
3066 Void_t
* rEALLOc(oldmem
, bytes
) Void_t
* oldmem
; size_t bytes
;
3070 INTERNAL_SIZE_T nb
; /* padded request size */
3072 mchunkptr oldp
; /* chunk corresponding to oldmem */
3073 INTERNAL_SIZE_T oldsize
; /* its size */
3075 mchunkptr newp
; /* chunk to return */
3077 #if defined _LIBC || defined MALLOC_HOOKS
3078 if (__realloc_hook
!= NULL
) {
3081 #if defined __GNUC__ && __GNUC__ >= 2
3082 result
= (*__realloc_hook
)(oldmem
, bytes
, __builtin_return_address (0));
3084 result
= (*__realloc_hook
)(oldmem
, bytes
, NULL
);
3090 #ifdef REALLOC_ZERO_BYTES_FREES
3091 if (bytes
== 0) { fREe(oldmem
); return 0; }
3094 /* realloc of null is supposed to be same as malloc */
3095 if (oldmem
== 0) return mALLOc(bytes
);
3097 oldp
= mem2chunk(oldmem
);
3098 oldsize
= chunksize(oldp
);
3100 nb
= request2size(bytes
);
3103 if (chunk_is_mmapped(oldp
))
3108 newp
= mremap_chunk(oldp
, nb
);
3109 if(newp
) return chunk2mem(newp
);
3111 /* Note the extra SIZE_SZ overhead. */
3112 if(oldsize
- SIZE_SZ
>= nb
) return oldmem
; /* do nothing */
3113 /* Must alloc, copy, free. */
3114 newmem
= mALLOc(bytes
);
3115 if (newmem
== 0) return 0; /* propagate failure */
3116 MALLOC_COPY(newmem
, oldmem
, oldsize
- 2*SIZE_SZ
);
3122 ar_ptr
= arena_for_ptr(oldp
);
3124 if(!mutex_trylock(&ar_ptr
->mutex
))
3125 ++(ar_ptr
->stat_lock_direct
);
3127 (void)mutex_lock(&ar_ptr
->mutex
);
3128 ++(ar_ptr
->stat_lock_wait
);
3131 (void)mutex_lock(&ar_ptr
->mutex
);
3135 /* As in malloc(), remember this arena for the next allocation. */
3136 tsd_setspecific(arena_key
, (Void_t
*)ar_ptr
);
3139 newp
= chunk_realloc(ar_ptr
, oldp
, oldsize
, nb
);
3141 (void)mutex_unlock(&ar_ptr
->mutex
);
3142 return newp
? chunk2mem(newp
) : NULL
;
3148 chunk_realloc(arena
* ar_ptr
, mchunkptr oldp
, INTERNAL_SIZE_T oldsize
,
3151 chunk_realloc(ar_ptr
, oldp
, oldsize
, nb
)
3152 arena
* ar_ptr
; mchunkptr oldp
; INTERNAL_SIZE_T oldsize
, nb
;
3155 mchunkptr newp
= oldp
; /* chunk to return */
3156 INTERNAL_SIZE_T newsize
= oldsize
; /* its size */
3158 mchunkptr next
; /* next contiguous chunk after oldp */
3159 INTERNAL_SIZE_T nextsize
; /* its size */
3161 mchunkptr prev
; /* previous contiguous chunk before oldp */
3162 INTERNAL_SIZE_T prevsize
; /* its size */
3164 mchunkptr remainder
; /* holds split off extra space from newp */
3165 INTERNAL_SIZE_T remainder_size
; /* its size */
3167 mchunkptr bck
; /* misc temp for linking */
3168 mchunkptr fwd
; /* misc temp for linking */
3170 check_inuse_chunk(ar_ptr
, oldp
);
3172 if ((long)(oldsize
) < (long)(nb
))
3175 /* Try expanding forward */
3177 next
= chunk_at_offset(oldp
, oldsize
);
3178 if (next
== top(ar_ptr
) || !inuse(next
))
3180 nextsize
= chunksize(next
);
3182 /* Forward into top only if a remainder */
3183 if (next
== top(ar_ptr
))
3185 if ((long)(nextsize
+ newsize
) >= (long)(nb
+ MINSIZE
))
3187 newsize
+= nextsize
;
3188 top(ar_ptr
) = chunk_at_offset(oldp
, nb
);
3189 set_head(top(ar_ptr
), (newsize
- nb
) | PREV_INUSE
);
3190 set_head_size(oldp
, nb
);
3195 /* Forward into next chunk */
3196 else if (((long)(nextsize
+ newsize
) >= (long)(nb
)))
3198 unlink(next
, bck
, fwd
);
3199 newsize
+= nextsize
;
3209 /* Try shifting backwards. */
3211 if (!prev_inuse(oldp
))
3213 prev
= prev_chunk(oldp
);
3214 prevsize
= chunksize(prev
);
3216 /* try forward + backward first to save a later consolidation */
3221 if (next
== top(ar_ptr
))
3223 if ((long)(nextsize
+ prevsize
+ newsize
) >= (long)(nb
+ MINSIZE
))
3225 unlink(prev
, bck
, fwd
);
3227 newsize
+= prevsize
+ nextsize
;
3228 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3229 top(ar_ptr
) = chunk_at_offset(newp
, nb
);
3230 set_head(top(ar_ptr
), (newsize
- nb
) | PREV_INUSE
);
3231 set_head_size(newp
, nb
);
3236 /* into next chunk */
3237 else if (((long)(nextsize
+ prevsize
+ newsize
) >= (long)(nb
)))
3239 unlink(next
, bck
, fwd
);
3240 unlink(prev
, bck
, fwd
);
3242 newsize
+= nextsize
+ prevsize
;
3243 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3249 if (prev
!= 0 && (long)(prevsize
+ newsize
) >= (long)nb
)
3251 unlink(prev
, bck
, fwd
);
3253 newsize
+= prevsize
;
3254 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3261 newp
= chunk_alloc (ar_ptr
, nb
);
3264 /* Maybe the failure is due to running out of mmapped areas. */
3265 if (ar_ptr
!= &main_arena
) {
3266 (void)mutex_lock(&main_arena
.mutex
);
3267 newp
= chunk_alloc(&main_arena
, nb
);
3268 (void)mutex_unlock(&main_arena
.mutex
);
3270 if (newp
== 0) /* propagate failure */
3274 /* Avoid copy if newp is next chunk after oldp. */
3275 /* (This can only happen when new chunk is sbrk'ed.) */
3277 if ( newp
== next_chunk(oldp
))
3279 newsize
+= chunksize(newp
);
3284 /* Otherwise copy, free, and exit */
3285 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3286 chunk_free(ar_ptr
, oldp
);
3291 split
: /* split off extra room in old or expanded chunk */
3293 if (newsize
- nb
>= MINSIZE
) /* split off remainder */
3295 remainder
= chunk_at_offset(newp
, nb
);
3296 remainder_size
= newsize
- nb
;
3297 set_head_size(newp
, nb
);
3298 set_head(remainder
, remainder_size
| PREV_INUSE
);
3299 set_inuse_bit_at_offset(remainder
, remainder_size
);
3300 chunk_free(ar_ptr
, remainder
);
3304 set_head_size(newp
, newsize
);
3305 set_inuse_bit_at_offset(newp
, newsize
);
3308 check_inuse_chunk(ar_ptr
, newp
);
3319 memalign requests more than enough space from malloc, finds a spot
3320 within that chunk that meets the alignment request, and then
3321 possibly frees the leading and trailing space.
3323 The alignment argument must be a power of two. This property is not
3324 checked by memalign, so misuse may result in random runtime errors.
3326 8-byte alignment is guaranteed by normal malloc calls, so don't
3327 bother calling memalign with an argument of 8 or less.
3329 Overreliance on memalign is a sure way to fragment space.
3335 Void_t
* mEMALIGn(size_t alignment
, size_t bytes
)
3337 Void_t
* mEMALIGn(alignment
, bytes
) size_t alignment
; size_t bytes
;
3341 INTERNAL_SIZE_T nb
; /* padded request size */
3344 #if defined _LIBC || defined MALLOC_HOOKS
3345 if (__memalign_hook
!= NULL
) {
3348 #if defined __GNUC__ && __GNUC__ >= 2
3349 result
= (*__memalign_hook
)(alignment
, bytes
,
3350 __builtin_return_address (0));
3352 result
= (*__memalign_hook
)(alignment
, bytes
, NULL
);
3358 /* If need less alignment than we give anyway, just relay to malloc */
3360 if (alignment
<= MALLOC_ALIGNMENT
) return mALLOc(bytes
);
3362 /* Otherwise, ensure that it is at least a minimum chunk size */
3364 if (alignment
< MINSIZE
) alignment
= MINSIZE
;
3366 nb
= request2size(bytes
);
3367 arena_get(ar_ptr
, nb
+ alignment
+ MINSIZE
);
3370 p
= chunk_align(ar_ptr
, nb
, alignment
);
3371 (void)mutex_unlock(&ar_ptr
->mutex
);
3373 /* Maybe the failure is due to running out of mmapped areas. */
3374 if(ar_ptr
!= &main_arena
) {
3375 (void)mutex_lock(&main_arena
.mutex
);
3376 p
= chunk_align(&main_arena
, nb
, alignment
);
3377 (void)mutex_unlock(&main_arena
.mutex
);
3381 return chunk2mem(p
);
3387 chunk_align(arena
* ar_ptr
, INTERNAL_SIZE_T nb
, size_t alignment
)
3389 chunk_align(ar_ptr
, nb
, alignment
)
3390 arena
* ar_ptr
; INTERNAL_SIZE_T nb
; size_t alignment
;
3393 char* m
; /* memory returned by malloc call */
3394 mchunkptr p
; /* corresponding chunk */
3395 char* brk
; /* alignment point within p */
3396 mchunkptr newp
; /* chunk to return */
3397 INTERNAL_SIZE_T newsize
; /* its size */
3398 INTERNAL_SIZE_T leadsize
; /* leading space befor alignment point */
3399 mchunkptr remainder
; /* spare room at end to split off */
3400 long remainder_size
; /* its size */
3402 /* Call chunk_alloc with worst case padding to hit alignment. */
3403 p
= chunk_alloc(ar_ptr
, nb
+ alignment
+ MINSIZE
);
3405 return 0; /* propagate failure */
3409 if ((((unsigned long)(m
)) % alignment
) == 0) /* aligned */
3412 if(chunk_is_mmapped(p
)) {
3413 return p
; /* nothing more to do */
3417 else /* misaligned */
3420 Find an aligned spot inside chunk.
3421 Since we need to give back leading space in a chunk of at
3422 least MINSIZE, if the first calculation places us at
3423 a spot with less than MINSIZE leader, we can move to the
3424 next aligned spot -- we've allocated enough total room so that
3425 this is always possible.
3428 brk
= (char*)mem2chunk(((unsigned long)(m
+ alignment
- 1)) & -alignment
);
3429 if ((long)(brk
- (char*)(p
)) < (long)MINSIZE
) brk
+= alignment
;
3431 newp
= (mchunkptr
)brk
;
3432 leadsize
= brk
- (char*)(p
);
3433 newsize
= chunksize(p
) - leadsize
;
3436 if(chunk_is_mmapped(p
))
3438 newp
->prev_size
= p
->prev_size
+ leadsize
;
3439 set_head(newp
, newsize
|IS_MMAPPED
);
3444 /* give back leader, use the rest */
3446 set_head(newp
, newsize
| PREV_INUSE
);
3447 set_inuse_bit_at_offset(newp
, newsize
);
3448 set_head_size(p
, leadsize
);
3449 chunk_free(ar_ptr
, p
);
3452 assert (newsize
>=nb
&& (((unsigned long)(chunk2mem(p
))) % alignment
) == 0);
3455 /* Also give back spare room at the end */
3457 remainder_size
= chunksize(p
) - nb
;
3459 if (remainder_size
>= (long)MINSIZE
)
3461 remainder
= chunk_at_offset(p
, nb
);
3462 set_head(remainder
, remainder_size
| PREV_INUSE
);
3463 set_head_size(p
, nb
);
3464 chunk_free(ar_ptr
, remainder
);
3467 check_inuse_chunk(ar_ptr
, p
);
3475 valloc just invokes memalign with alignment argument equal
3476 to the page size of the system (or as near to this as can
3477 be figured out from all the includes/defines above.)
3481 Void_t
* vALLOc(size_t bytes
)
3483 Void_t
* vALLOc(bytes
) size_t bytes
;
3486 return mEMALIGn (malloc_getpagesize
, bytes
);
3490 pvalloc just invokes valloc for the nearest pagesize
3491 that will accommodate request
3496 Void_t
* pvALLOc(size_t bytes
)
3498 Void_t
* pvALLOc(bytes
) size_t bytes
;
3501 size_t pagesize
= malloc_getpagesize
;
3502 return mEMALIGn (pagesize
, (bytes
+ pagesize
- 1) & ~(pagesize
- 1));
3507 calloc calls chunk_alloc, then zeroes out the allocated chunk.
3512 Void_t
* cALLOc(size_t n
, size_t elem_size
)
3514 Void_t
* cALLOc(n
, elem_size
) size_t n
; size_t elem_size
;
3518 mchunkptr p
, oldtop
;
3519 INTERNAL_SIZE_T sz
, csz
, oldtopsize
;
3522 #if defined _LIBC || defined MALLOC_HOOKS
3523 if (__malloc_hook
!= NULL
) {
3525 #if defined __GNUC__ && __GNUC__ >= 2
3526 mem
= (*__malloc_hook
)(sz
, __builtin_return_address (0));
3528 mem
= (*__malloc_hook
)(sz
, NULL
);
3533 return memset(mem
, 0, sz
);
3535 while(sz
> 0) ((char*)mem
)[--sz
] = 0; /* rather inefficient */
3541 sz
= request2size(n
* elem_size
);
3542 arena_get(ar_ptr
, sz
);
3546 /* check if expand_top called, in which case don't need to clear */
3548 oldtop
= top(ar_ptr
);
3549 oldtopsize
= chunksize(top(ar_ptr
));
3551 p
= chunk_alloc (ar_ptr
, sz
);
3553 /* Only clearing follows, so we can unlock early. */
3554 (void)mutex_unlock(&ar_ptr
->mutex
);
3557 /* Maybe the failure is due to running out of mmapped areas. */
3558 if(ar_ptr
!= &main_arena
) {
3559 (void)mutex_lock(&main_arena
.mutex
);
3560 p
= chunk_alloc(&main_arena
, sz
);
3561 (void)mutex_unlock(&main_arena
.mutex
);
3563 if (p
== 0) return 0;
3567 /* Two optional cases in which clearing not necessary */
3570 if (chunk_is_mmapped(p
)) return mem
;
3576 if (p
== oldtop
&& csz
> oldtopsize
) {
3577 /* clear only the bytes from non-freshly-sbrked memory */
3582 MALLOC_ZERO(mem
, csz
- SIZE_SZ
);
3588 cfree just calls free. It is needed/defined on some systems
3589 that pair it with calloc, presumably for odd historical reasons.
3595 void cfree(Void_t
*mem
)
3597 void cfree(mem
) Void_t
*mem
;
3608 Malloc_trim gives memory back to the system (via negative
3609 arguments to sbrk) if there is unused memory at the `high' end of
3610 the malloc pool. You can call this after freeing large blocks of
3611 memory to potentially reduce the system-level memory requirements
3612 of a program. However, it cannot guarantee to reduce memory. Under
3613 some allocation patterns, some large free blocks of memory will be
3614 locked between two used chunks, so they cannot be given back to
3617 The `pad' argument to malloc_trim represents the amount of free
3618 trailing space to leave untrimmed. If this argument is zero,
3619 only the minimum amount of memory to maintain internal data
3620 structures will be left (one page or less). Non-zero arguments
3621 can be supplied to maintain enough trailing space to service
3622 future expected allocations without having to re-obtain memory
3625 Malloc_trim returns 1 if it actually released any memory, else 0.
3630 int mALLOC_TRIm(size_t pad
)
3632 int mALLOC_TRIm(pad
) size_t pad
;
3637 (void)mutex_lock(&main_arena
.mutex
);
3638 res
= main_trim(pad
);
3639 (void)mutex_unlock(&main_arena
.mutex
);
3643 /* Trim the main arena. */
3648 main_trim(size_t pad
)
3650 main_trim(pad
) size_t pad
;
3653 mchunkptr top_chunk
; /* The current top chunk */
3654 long top_size
; /* Amount of top-most memory */
3655 long extra
; /* Amount to release */
3656 char* current_brk
; /* address returned by pre-check sbrk call */
3657 char* new_brk
; /* address returned by negative sbrk call */
3659 unsigned long pagesz
= malloc_getpagesize
;
3661 top_chunk
= top(&main_arena
);
3662 top_size
= chunksize(top_chunk
);
3663 extra
= ((top_size
- pad
- MINSIZE
+ (pagesz
-1)) / pagesz
- 1) * pagesz
;
3665 if (extra
< (long)pagesz
) /* Not enough memory to release */
3668 /* Test to make sure no one else called sbrk */
3669 current_brk
= (char*)(MORECORE (0));
3670 if (current_brk
!= (char*)(top_chunk
) + top_size
)
3671 return 0; /* Apparently we don't own memory; must fail */
3673 new_brk
= (char*)(MORECORE (-extra
));
3675 #if defined _LIBC || defined MALLOC_HOOKS
3676 /* Call the `morecore' hook if necessary. */
3677 if (__after_morecore_hook
)
3678 (*__after_morecore_hook
) ();
3681 if (new_brk
== (char*)(MORECORE_FAILURE
)) { /* sbrk failed? */
3682 /* Try to figure out what we have */
3683 current_brk
= (char*)(MORECORE (0));
3684 top_size
= current_brk
- (char*)top_chunk
;
3685 if (top_size
>= (long)MINSIZE
) /* if not, we are very very dead! */
3687 sbrked_mem
= current_brk
- sbrk_base
;
3688 set_head(top_chunk
, top_size
| PREV_INUSE
);
3690 check_chunk(&main_arena
, top_chunk
);
3693 sbrked_mem
-= extra
;
3695 /* Success. Adjust top accordingly. */
3696 set_head(top_chunk
, (top_size
- extra
) | PREV_INUSE
);
3697 check_chunk(&main_arena
, top_chunk
);
3706 heap_trim(heap_info
*heap
, size_t pad
)
3708 heap_trim(heap
, pad
) heap_info
*heap
; size_t pad
;
3711 unsigned long pagesz
= malloc_getpagesize
;
3712 arena
*ar_ptr
= heap
->ar_ptr
;
3713 mchunkptr top_chunk
= top(ar_ptr
), p
, bck
, fwd
;
3714 heap_info
*prev_heap
;
3715 long new_size
, top_size
, extra
;
3717 /* Can this heap go away completely ? */
3718 while(top_chunk
== chunk_at_offset(heap
, sizeof(*heap
))) {
3719 prev_heap
= heap
->prev
;
3720 p
= chunk_at_offset(prev_heap
, prev_heap
->size
- (MINSIZE
-2*SIZE_SZ
));
3721 assert(p
->size
== (0|PREV_INUSE
)); /* must be fencepost */
3723 new_size
= chunksize(p
) + (MINSIZE
-2*SIZE_SZ
);
3724 assert(new_size
>0 && new_size
<(long)(2*MINSIZE
));
3726 new_size
+= p
->prev_size
;
3727 assert(new_size
>0 && new_size
<HEAP_MAX_SIZE
);
3728 if(new_size
+ (HEAP_MAX_SIZE
- prev_heap
->size
) < pad
+ MINSIZE
+ pagesz
)
3730 ar_ptr
->size
-= heap
->size
;
3733 if(!prev_inuse(p
)) { /* consolidate backward */
3735 unlink(p
, bck
, fwd
);
3737 assert(((unsigned long)((char*)p
+ new_size
) & (pagesz
-1)) == 0);
3738 assert( ((char*)p
+ new_size
) == ((char*)heap
+ heap
->size
) );
3739 top(ar_ptr
) = top_chunk
= p
;
3740 set_head(top_chunk
, new_size
| PREV_INUSE
);
3741 check_chunk(ar_ptr
, top_chunk
);
3743 top_size
= chunksize(top_chunk
);
3744 extra
= ((top_size
- pad
- MINSIZE
+ (pagesz
-1))/pagesz
- 1) * pagesz
;
3745 if(extra
< (long)pagesz
)
3747 /* Try to shrink. */
3748 if(grow_heap(heap
, -extra
) != 0)
3750 ar_ptr
->size
-= extra
;
3752 /* Success. Adjust top accordingly. */
3753 set_head(top_chunk
, (top_size
- extra
) | PREV_INUSE
);
3754 check_chunk(ar_ptr
, top_chunk
);
3765 This routine tells you how many bytes you can actually use in an
3766 allocated chunk, which may be more than you requested (although
3767 often not). You can use this many bytes without worrying about
3768 overwriting other allocated objects. Not a particularly great
3769 programming practice, but still sometimes useful.
3774 size_t mALLOC_USABLE_SIZe(Void_t
* mem
)
3776 size_t mALLOC_USABLE_SIZe(mem
) Void_t
* mem
;
3786 if(!chunk_is_mmapped(p
))
3788 if (!inuse(p
)) return 0;
3789 check_inuse_chunk(arena_for_ptr(mem
), p
);
3790 return chunksize(p
) - SIZE_SZ
;
3792 return chunksize(p
) - 2*SIZE_SZ
;
3799 /* Utility to update mallinfo for malloc_stats() and mallinfo() */
3803 malloc_update_mallinfo(arena
*ar_ptr
, struct mallinfo
*mi
)
3805 malloc_update_mallinfo(ar_ptr
, mi
) arena
*ar_ptr
; struct mallinfo
*mi
;
3814 INTERNAL_SIZE_T avail
;
3816 (void)mutex_lock(&ar_ptr
->mutex
);
3817 avail
= chunksize(top(ar_ptr
));
3818 navail
= ((long)(avail
) >= (long)MINSIZE
)? 1 : 0;
3820 for (i
= 1; i
< NAV
; ++i
)
3822 b
= bin_at(ar_ptr
, i
);
3823 for (p
= last(b
); p
!= b
; p
= p
->bk
)
3826 check_free_chunk(ar_ptr
, p
);
3827 for (q
= next_chunk(p
);
3828 q
!= top(ar_ptr
) && inuse(q
) && (long)chunksize(q
) > 0;
3830 check_inuse_chunk(ar_ptr
, q
);
3832 avail
+= chunksize(p
);
3837 mi
->arena
= ar_ptr
->size
;
3838 mi
->ordblks
= navail
;
3839 mi
->smblks
= mi
->usmblks
= mi
->fsmblks
= 0; /* clear unused fields */
3840 mi
->uordblks
= ar_ptr
->size
- avail
;
3841 mi
->fordblks
= avail
;
3842 mi
->hblks
= n_mmaps
;
3843 mi
->hblkhd
= mmapped_mem
;
3844 mi
->keepcost
= chunksize(top(ar_ptr
));
3846 (void)mutex_unlock(&ar_ptr
->mutex
);
3849 #if !defined(NO_THREADS) && MALLOC_DEBUG > 1
3851 /* Print the complete contents of a single heap to stderr. */
3855 dump_heap(heap_info
*heap
)
3857 dump_heap(heap
) heap_info
*heap
;
3863 fprintf(stderr
, "Heap %p, size %10lx:\n", heap
, (long)heap
->size
);
3864 ptr
= (heap
->ar_ptr
!= (arena
*)(heap
+1)) ?
3865 (char*)(heap
+ 1) : (char*)(heap
+ 1) + sizeof(arena
);
3866 p
= (mchunkptr
)(((unsigned long)ptr
+ MALLOC_ALIGN_MASK
) &
3867 ~MALLOC_ALIGN_MASK
);
3869 fprintf(stderr
, "chunk %p size %10lx", p
, (long)p
->size
);
3870 if(p
== top(heap
->ar_ptr
)) {
3871 fprintf(stderr
, " (top)\n");
3873 } else if(p
->size
== (0|PREV_INUSE
)) {
3874 fprintf(stderr
, " (fence)\n");
3877 fprintf(stderr
, "\n");
3890 For all arenas separately and in total, prints on stderr the
3891 amount of space obtained from the system, and the current number
3892 of bytes allocated via malloc (or realloc, etc) but not yet
3893 freed. (Note that this is the number of bytes allocated, not the
3894 number requested. It will be larger than the number requested
3895 because of alignment and bookkeeping overhead.) When not compiled
3896 for multiple threads, the maximum amount of allocated memory
3897 (which may be more than current if malloc_trim and/or munmap got
3898 called) is also reported. When using mmap(), prints the maximum
3899 number of simultaneous mmap regions used, too.
3908 unsigned int in_use_b
= mmapped_mem
, system_b
= in_use_b
;
3910 long stat_lock_direct
= 0, stat_lock_loop
= 0, stat_lock_wait
= 0;
3913 for(i
=0, ar_ptr
= &main_arena
;; i
++) {
3914 malloc_update_mallinfo(ar_ptr
, &mi
);
3915 fprintf(stderr
, "Arena %d:\n", i
);
3916 fprintf(stderr
, "system bytes = %10u\n", (unsigned int)mi
.arena
);
3917 fprintf(stderr
, "in use bytes = %10u\n", (unsigned int)mi
.uordblks
);
3918 system_b
+= mi
.arena
;
3919 in_use_b
+= mi
.uordblks
;
3921 stat_lock_direct
+= ar_ptr
->stat_lock_direct
;
3922 stat_lock_loop
+= ar_ptr
->stat_lock_loop
;
3923 stat_lock_wait
+= ar_ptr
->stat_lock_wait
;
3925 #if !defined(NO_THREADS) && MALLOC_DEBUG > 1
3926 if(ar_ptr
!= &main_arena
) {
3928 (void)mutex_lock(&ar_ptr
->mutex
);
3929 heap
= heap_for_ptr(top(ar_ptr
));
3930 while(heap
) { dump_heap(heap
); heap
= heap
->prev
; }
3931 (void)mutex_unlock(&ar_ptr
->mutex
);
3934 ar_ptr
= ar_ptr
->next
;
3935 if(ar_ptr
== &main_arena
) break;
3938 fprintf(stderr
, "Total (incl. mmap):\n");
3940 fprintf(stderr
, "Total:\n");
3942 fprintf(stderr
, "system bytes = %10u\n", system_b
);
3943 fprintf(stderr
, "in use bytes = %10u\n", in_use_b
);
3945 fprintf(stderr
, "max system bytes = %10u\n", (unsigned int)max_total_mem
);
3948 fprintf(stderr
, "max mmap regions = %10u\n", (unsigned int)max_n_mmaps
);
3949 fprintf(stderr
, "max mmap bytes = %10lu\n", max_mmapped_mem
);
3952 fprintf(stderr
, "heaps created = %10d\n", stat_n_heaps
);
3953 fprintf(stderr
, "locked directly = %10ld\n", stat_lock_direct
);
3954 fprintf(stderr
, "locked in loop = %10ld\n", stat_lock_loop
);
3955 fprintf(stderr
, "locked waiting = %10ld\n", stat_lock_wait
);
3956 fprintf(stderr
, "locked total = %10ld\n",
3957 stat_lock_direct
+ stat_lock_loop
+ stat_lock_wait
);
3962 mallinfo returns a copy of updated current mallinfo.
3963 The information reported is for the arena last used by the thread.
3966 struct mallinfo
mALLINFo()
3969 Void_t
*vptr
= NULL
;
3972 tsd_getspecific(arena_key
, vptr
);
3974 malloc_update_mallinfo((vptr
? (arena
*)vptr
: &main_arena
), &mi
);
3984 mallopt is the general SVID/XPG interface to tunable parameters.
3985 The format is to provide a (parameter-number, parameter-value) pair.
3986 mallopt then sets the corresponding parameter to the argument
3987 value if it can (i.e., so long as the value is meaningful),
3988 and returns 1 if successful else 0.
3990 See descriptions of tunable parameters above.
3995 int mALLOPt(int param_number
, int value
)
3997 int mALLOPt(param_number
, value
) int param_number
; int value
;
4000 switch(param_number
)
4002 case M_TRIM_THRESHOLD
:
4003 trim_threshold
= value
; return 1;
4005 top_pad
= value
; return 1;
4006 case M_MMAP_THRESHOLD
:
4008 /* Forbid setting the threshold too high. */
4009 if((unsigned long)value
> HEAP_MAX_SIZE
/2) return 0;
4011 mmap_threshold
= value
; return 1;
4014 n_mmaps_max
= value
; return 1;
4016 if (value
!= 0) return 0; else n_mmaps_max
= value
; return 1;
4018 case M_CHECK_ACTION
:
4019 check_action
= value
; return 1;
4028 /* Get/set state: malloc_get_state() records the current state of all
4029 malloc variables (_except_ for the actual heap contents and `hook'
4030 function pointers) in a system dependent, opaque data structure.
4031 This data structure is dynamically allocated and can be free()d
4032 after use. malloc_set_state() restores the state of all malloc
4033 variables to the previously obtained state. This is especially
4034 useful when using this malloc as part of a shared library, and when
4035 the heap contents are saved/restored via some other method. The
4036 primary example for this is GNU Emacs with its `dumping' procedure.
4037 `Hook' function pointers are never saved or restored by these
4040 #define MALLOC_STATE_MAGIC 0x444c4541l
4041 #define MALLOC_STATE_VERSION (0*0x100l + 0l) /* major*0x100 + minor */
4043 struct malloc_state
{
4046 mbinptr av
[NAV
* 2 + 2];
4048 int sbrked_mem_bytes
;
4049 unsigned long trim_threshold
;
4050 unsigned long top_pad
;
4051 unsigned int n_mmaps_max
;
4052 unsigned long mmap_threshold
;
4054 unsigned long max_sbrked_mem
;
4055 unsigned long max_total_mem
;
4056 unsigned int n_mmaps
;
4057 unsigned int max_n_mmaps
;
4058 unsigned long mmapped_mem
;
4059 unsigned long max_mmapped_mem
;
4066 struct malloc_state
* ms
;
4071 (void)mutex_lock(&main_arena
.mutex
);
4072 victim
= chunk_alloc(&main_arena
, request2size(sizeof(*ms
)));
4074 (void)mutex_unlock(&main_arena
.mutex
);
4077 ms
= (struct malloc_state
*)chunk2mem(victim
);
4078 ms
->magic
= MALLOC_STATE_MAGIC
;
4079 ms
->version
= MALLOC_STATE_VERSION
;
4080 ms
->av
[0] = main_arena
.av
[0];
4081 ms
->av
[1] = main_arena
.av
[1];
4082 for(i
=0; i
<NAV
; i
++) {
4083 b
= bin_at(&main_arena
, i
);
4085 ms
->av
[2*i
+2] = ms
->av
[2*i
+3] = 0; /* empty bin (or initial top) */
4087 ms
->av
[2*i
+2] = first(b
);
4088 ms
->av
[2*i
+3] = last(b
);
4091 ms
->sbrk_base
= sbrk_base
;
4092 ms
->sbrked_mem_bytes
= sbrked_mem
;
4093 ms
->trim_threshold
= trim_threshold
;
4094 ms
->top_pad
= top_pad
;
4095 ms
->n_mmaps_max
= n_mmaps_max
;
4096 ms
->mmap_threshold
= mmap_threshold
;
4097 ms
->check_action
= check_action
;
4098 ms
->max_sbrked_mem
= max_sbrked_mem
;
4100 ms
->max_total_mem
= max_total_mem
;
4102 ms
->max_total_mem
= 0;
4104 ms
->n_mmaps
= n_mmaps
;
4105 ms
->max_n_mmaps
= max_n_mmaps
;
4106 ms
->mmapped_mem
= mmapped_mem
;
4107 ms
->max_mmapped_mem
= max_mmapped_mem
;
4108 (void)mutex_unlock(&main_arena
.mutex
);
4114 mALLOC_SET_STATe(Void_t
* msptr
)
4116 mALLOC_SET_STATe(msptr
) Void_t
* msptr
;
4119 struct malloc_state
* ms
= (struct malloc_state
*)msptr
;
4124 if(ms
->magic
!= MALLOC_STATE_MAGIC
) return -1;
4125 /* Must fail if the major version is too high. */
4126 if((ms
->version
& ~0xffl
) > (MALLOC_STATE_VERSION
& ~0xffl
)) return -2;
4127 (void)mutex_lock(&main_arena
.mutex
);
4128 main_arena
.av
[0] = ms
->av
[0];
4129 main_arena
.av
[1] = ms
->av
[1];
4130 for(i
=0; i
<NAV
; i
++) {
4131 b
= bin_at(&main_arena
, i
);
4132 if(ms
->av
[2*i
+2] == 0)
4133 first(b
) = last(b
) = b
;
4135 first(b
) = ms
->av
[2*i
+2];
4136 last(b
) = ms
->av
[2*i
+3];
4138 /* Make sure the links to the `av'-bins in the heap are correct. */
4144 sbrk_base
= ms
->sbrk_base
;
4145 sbrked_mem
= ms
->sbrked_mem_bytes
;
4146 trim_threshold
= ms
->trim_threshold
;
4147 top_pad
= ms
->top_pad
;
4148 n_mmaps_max
= ms
->n_mmaps_max
;
4149 mmap_threshold
= ms
->mmap_threshold
;
4150 check_action
= ms
->check_action
;
4151 max_sbrked_mem
= ms
->max_sbrked_mem
;
4153 max_total_mem
= ms
->max_total_mem
;
4155 n_mmaps
= ms
->n_mmaps
;
4156 max_n_mmaps
= ms
->max_n_mmaps
;
4157 mmapped_mem
= ms
->mmapped_mem
;
4158 max_mmapped_mem
= ms
->max_mmapped_mem
;
4159 /* add version-dependent code here */
4160 (void)mutex_unlock(&main_arena
.mutex
);
4166 #if defined _LIBC || defined MALLOC_HOOKS
4168 /* A simple, standard set of debugging hooks. Overhead is `only' one
4169 byte per chunk; still this will catch most cases of double frees or
4170 overruns. The goal here is to avoid obscure crashes due to invalid
4171 usage, unlike in the MALLOC_DEBUG code. */
4173 #define MAGICBYTE(p) ( ( ((size_t)p >> 3) ^ ((size_t)p >> 11)) & 0xFF )
4175 /* Instrument a chunk with overrun detector byte(s) and convert it
4176 into a user pointer with requested size sz. */
4180 chunk2mem_check(mchunkptr p
, size_t sz
)
4182 chunk2mem_check(p
, sz
) mchunkptr p
; size_t sz
;
4185 unsigned char* m_ptr
= (unsigned char*)chunk2mem(p
);
4188 for(i
= chunksize(p
) - (chunk_is_mmapped(p
) ? 2*SIZE_SZ
+1 : SIZE_SZ
+1);
4192 m_ptr
[i
] = (unsigned char)(i
-sz
);
4197 m_ptr
[sz
] = MAGICBYTE(p
);
4198 return (Void_t
*)m_ptr
;
4201 /* Convert a pointer to be free()d or realloc()ed to a valid chunk
4202 pointer. If the provided pointer is not valid, return NULL. */
4207 mem2chunk_check(Void_t
* mem
)
4209 mem2chunk_check(mem
) Void_t
* mem
;
4213 INTERNAL_SIZE_T sz
, c
;
4214 unsigned char magic
;
4217 if(!aligned_OK(p
)) return NULL
;
4218 if( (char*)p
>=sbrk_base
&& (char*)p
<(sbrk_base
+sbrked_mem
) ) {
4219 /* Must be a chunk in conventional heap memory. */
4220 if(chunk_is_mmapped(p
) ||
4221 ( (sz
= chunksize(p
)), ((char*)p
+ sz
)>=(sbrk_base
+sbrked_mem
) ) ||
4222 sz
<MINSIZE
|| sz
&MALLOC_ALIGN_MASK
|| !inuse(p
) ||
4223 ( !prev_inuse(p
) && (p
->prev_size
&MALLOC_ALIGN_MASK
||
4224 (long)prev_chunk(p
)<(long)sbrk_base
||
4225 next_chunk(prev_chunk(p
))!=p
) ))
4227 magic
= MAGICBYTE(p
);
4228 for(sz
+= SIZE_SZ
-1; (c
= ((unsigned char*)p
)[sz
]) != magic
; sz
-= c
) {
4229 if(c
<=0 || sz
<(c
+2*SIZE_SZ
)) return NULL
;
4231 ((unsigned char*)p
)[sz
] ^= 0xFF;
4233 unsigned long offset
, page_mask
= malloc_getpagesize
-1;
4235 /* mmap()ed chunks have MALLOC_ALIGNMENT or higher power-of-two
4236 alignment relative to the beginning of a page. Check this
4238 offset
= (unsigned long)mem
& page_mask
;
4239 if((offset
!=MALLOC_ALIGNMENT
&& offset
!=0 && offset
!=0x10 &&
4240 offset
!=0x20 && offset
!=0x40 && offset
!=0x80 && offset
!=0x100 &&
4241 offset
!=0x200 && offset
!=0x400 && offset
!=0x800 && offset
!=0x1000 &&
4243 !chunk_is_mmapped(p
) || (p
->size
& PREV_INUSE
) ||
4244 ( (((unsigned long)p
- p
->prev_size
) & page_mask
) != 0 ) ||
4245 ( (sz
= chunksize(p
)), ((p
->prev_size
+ sz
) & page_mask
) != 0 ) )
4247 magic
= MAGICBYTE(p
);
4248 for(sz
-= 1; (c
= ((unsigned char*)p
)[sz
]) != magic
; sz
-= c
) {
4249 if(c
<=0 || sz
<(c
+2*SIZE_SZ
)) return NULL
;
4251 ((unsigned char*)p
)[sz
] ^= 0xFF;
4256 /* Check for corruption of the top chunk, and try to recover if
4266 mchunkptr t
= top(&main_arena
);
4267 char* brk
, * new_brk
;
4268 INTERNAL_SIZE_T front_misalign
, sbrk_size
;
4269 unsigned long pagesz
= malloc_getpagesize
;
4271 if((char*)t
+ chunksize(t
) == sbrk_base
+ sbrked_mem
||
4272 t
== initial_top(&main_arena
)) return 0;
4274 switch(check_action
) {
4276 fprintf(stderr
, "malloc: top chunk is corrupt\n");
4281 /* Try to set up a new top chunk. */
4283 front_misalign
= (unsigned long)chunk2mem(brk
) & MALLOC_ALIGN_MASK
;
4284 if (front_misalign
> 0)
4285 front_misalign
= MALLOC_ALIGNMENT
- front_misalign
;
4286 sbrk_size
= front_misalign
+ top_pad
+ MINSIZE
;
4287 sbrk_size
+= pagesz
- ((unsigned long)(brk
+ sbrk_size
) & (pagesz
- 1));
4288 new_brk
= (char*)(MORECORE (sbrk_size
));
4289 if (new_brk
== (char*)(MORECORE_FAILURE
)) return -1;
4290 sbrked_mem
= (new_brk
- sbrk_base
) + sbrk_size
;
4292 top(&main_arena
) = (mchunkptr
)(brk
+ front_misalign
);
4293 set_head(top(&main_arena
), (sbrk_size
- front_misalign
) | PREV_INUSE
);
4300 malloc_check(size_t sz
, const Void_t
*caller
)
4302 malloc_check(sz
, caller
) size_t sz
; const Void_t
*caller
;
4306 INTERNAL_SIZE_T nb
= request2size(sz
+ 1);
4308 (void)mutex_lock(&main_arena
.mutex
);
4309 victim
= (top_check() >= 0) ? chunk_alloc(&main_arena
, nb
) : NULL
;
4310 (void)mutex_unlock(&main_arena
.mutex
);
4311 if(!victim
) return NULL
;
4312 return chunk2mem_check(victim
, sz
);
4317 free_check(Void_t
* mem
, const Void_t
*caller
)
4319 free_check(mem
, caller
) Void_t
* mem
; const Void_t
*caller
;
4325 (void)mutex_lock(&main_arena
.mutex
);
4326 p
= mem2chunk_check(mem
);
4328 (void)mutex_unlock(&main_arena
.mutex
);
4329 switch(check_action
) {
4331 fprintf(stderr
, "free(): invalid pointer %lx!\n", (long)(mem
));
4339 if (chunk_is_mmapped(p
)) {
4340 (void)mutex_unlock(&main_arena
.mutex
);
4345 #if 0 /* Erase freed memory. */
4346 memset(mem
, 0, chunksize(p
) - (SIZE_SZ
+1));
4348 chunk_free(&main_arena
, p
);
4349 (void)mutex_unlock(&main_arena
.mutex
);
4354 realloc_check(Void_t
* oldmem
, size_t bytes
, const Void_t
*caller
)
4356 realloc_check(oldmem
, bytes
, caller
)
4357 Void_t
* oldmem
; size_t bytes
; const Void_t
*caller
;
4360 mchunkptr oldp
, newp
;
4361 INTERNAL_SIZE_T nb
, oldsize
;
4363 if (oldmem
== 0) return malloc_check(bytes
, NULL
);
4364 (void)mutex_lock(&main_arena
.mutex
);
4365 oldp
= mem2chunk_check(oldmem
);
4367 (void)mutex_unlock(&main_arena
.mutex
);
4368 switch(check_action
) {
4370 fprintf(stderr
, "realloc(): invalid pointer %lx!\n", (long)(oldmem
));
4375 return malloc_check(bytes
, NULL
);
4377 oldsize
= chunksize(oldp
);
4379 nb
= request2size(bytes
+1);
4382 if (chunk_is_mmapped(oldp
)) {
4384 newp
= mremap_chunk(oldp
, nb
);
4387 /* Note the extra SIZE_SZ overhead. */
4388 if(oldsize
- SIZE_SZ
>= nb
) newp
= oldp
; /* do nothing */
4390 /* Must alloc, copy, free. */
4391 newp
= (top_check() >= 0) ? chunk_alloc(&main_arena
, nb
) : NULL
;
4393 MALLOC_COPY(chunk2mem(newp
), oldmem
, oldsize
- 2*SIZE_SZ
);
4401 #endif /* HAVE_MMAP */
4402 newp
= (top_check() >= 0) ?
4403 chunk_realloc(&main_arena
, oldp
, oldsize
, nb
) : NULL
;
4404 #if 0 /* Erase freed memory. */
4405 nb
= chunksize(newp
);
4406 if(oldp
<newp
|| oldp
>=chunk_at_offset(newp
, nb
)) {
4407 memset((char*)oldmem
+ 2*sizeof(mbinptr
), 0,
4408 oldsize
- (2*sizeof(mbinptr
)+2*SIZE_SZ
+1));
4409 } else if(nb
> oldsize
+SIZE_SZ
) {
4410 memset((char*)chunk2mem(newp
) + oldsize
, 0, nb
- (oldsize
+SIZE_SZ
));
4416 (void)mutex_unlock(&main_arena
.mutex
);
4418 if(!newp
) return NULL
;
4419 return chunk2mem_check(newp
, bytes
);
4424 memalign_check(size_t alignment
, size_t bytes
, const Void_t
*caller
)
4426 memalign_check(alignment
, bytes
, caller
)
4427 size_t alignment
; size_t bytes
; const Void_t
*caller
;
4433 if (alignment
<= MALLOC_ALIGNMENT
) return malloc_check(bytes
, NULL
);
4434 if (alignment
< MINSIZE
) alignment
= MINSIZE
;
4436 nb
= request2size(bytes
+1);
4437 (void)mutex_lock(&main_arena
.mutex
);
4438 p
= (top_check() >= 0) ? chunk_align(&main_arena
, nb
, alignment
) : NULL
;
4439 (void)mutex_unlock(&main_arena
.mutex
);
4441 return chunk2mem_check(p
, bytes
);
4446 /* The following hooks are used when the global initialization in
4447 ptmalloc_init() hasn't completed yet. */
4451 malloc_starter(size_t sz
, const Void_t
*caller
)
4453 malloc_starter(sz
, caller
) size_t sz
; const Void_t
*caller
;
4456 mchunkptr victim
= chunk_alloc(&main_arena
, request2size(sz
));
4458 return victim
? chunk2mem(victim
) : 0;
4463 free_starter(Void_t
* mem
, const Void_t
*caller
)
4465 free_starter(mem
, caller
) Void_t
* mem
; const Void_t
*caller
;
4473 if (chunk_is_mmapped(p
)) {
4478 chunk_free(&main_arena
, p
);
4481 /* The following hooks are used while the `atfork' handling mechanism
4486 malloc_atfork (size_t sz
, const Void_t
*caller
)
4488 malloc_atfork(sz
, caller
) size_t sz
; const Void_t
*caller
;
4491 Void_t
*vptr
= NULL
;
4493 tsd_getspecific(arena_key
, vptr
);
4495 mchunkptr victim
= chunk_alloc(&main_arena
, request2size(sz
));
4496 return victim
? chunk2mem(victim
) : 0;
4498 /* Suspend the thread until the `atfork' handlers have completed.
4499 By that time, the hooks will have been reset as well, so that
4500 mALLOc() can be used again. */
4501 (void)mutex_lock(&list_lock
);
4502 (void)mutex_unlock(&list_lock
);
4509 free_atfork(Void_t
* mem
, const Void_t
*caller
)
4511 free_atfork(mem
, caller
) Void_t
* mem
; const Void_t
*caller
;
4514 Void_t
*vptr
= NULL
;
4516 mchunkptr p
; /* chunk corresponding to mem */
4518 if (mem
== 0) /* free(0) has no effect */
4524 if (chunk_is_mmapped(p
)) /* release mmapped memory. */
4531 ar_ptr
= arena_for_ptr(p
);
4532 tsd_getspecific(arena_key
, vptr
);
4534 (void)mutex_lock(&ar_ptr
->mutex
);
4535 chunk_free(ar_ptr
, p
);
4537 (void)mutex_unlock(&ar_ptr
->mutex
);
4542 #endif /* defined _LIBC || defined MALLOC_HOOKS */
4547 weak_alias (__libc_calloc
, __calloc
) weak_alias (__libc_calloc
, calloc
)
4548 weak_alias (__libc_free
, __cfree
) weak_alias (__libc_free
, cfree
)
4549 weak_alias (__libc_free
, __free
) weak_alias (__libc_free
, free
)
4550 weak_alias (__libc_malloc
, __malloc
) weak_alias (__libc_malloc
, malloc
)
4551 weak_alias (__libc_memalign
, __memalign
) weak_alias (__libc_memalign
, memalign
)
4552 weak_alias (__libc_realloc
, __realloc
) weak_alias (__libc_realloc
, realloc
)
4553 weak_alias (__libc_valloc
, __valloc
) weak_alias (__libc_valloc
, valloc
)
4554 weak_alias (__libc_pvalloc
, __pvalloc
) weak_alias (__libc_pvalloc
, pvalloc
)
4555 weak_alias (__libc_mallinfo
, __mallinfo
) weak_alias (__libc_mallinfo
, mallinfo
)
4556 weak_alias (__libc_mallopt
, __mallopt
) weak_alias (__libc_mallopt
, mallopt
)
4558 weak_alias (__malloc_stats
, malloc_stats
)
4559 weak_alias (__malloc_usable_size
, malloc_usable_size
)
4560 weak_alias (__malloc_trim
, malloc_trim
)
4561 weak_alias (__malloc_get_state
, malloc_get_state
)
4562 weak_alias (__malloc_set_state
, malloc_set_state
)
4569 V2.6.4-pt3 Thu Feb 20 1997 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
4570 * Added malloc_get/set_state() (mainly for use in GNU emacs),
4571 using interface from Marcus Daniels
4572 * All parameters are now adjustable via environment variables
4574 V2.6.4-pt2 Sat Dec 14 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
4575 * Added debugging hooks
4576 * Fixed possible deadlock in realloc() when out of memory
4577 * Don't pollute namespace in glibc: use __getpagesize, __mmap, etc.
4579 V2.6.4-pt Wed Dec 4 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
4580 * Very minor updates from the released 2.6.4 version.
4581 * Trimmed include file down to exported data structures.
4582 * Changes from H.J. Lu for glibc-2.0.
4584 V2.6.3i-pt Sep 16 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
4585 * Many changes for multiple threads
4586 * Introduced arenas and heaps
4588 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)
4589 * Added pvalloc, as recommended by H.J. Liu
4590 * Added 64bit pointer support mainly from Wolfram Gloger
4591 * Added anonymously donated WIN32 sbrk emulation
4592 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
4593 * malloc_extend_top: fix mask error that caused wastage after
4595 * Add linux mremap support code from HJ Liu
4597 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)
4598 * Integrated most documentation with the code.
4599 * Add support for mmap, with help from
4600 Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
4601 * Use last_remainder in more cases.
4602 * Pack bins using idea from colin@nyx10.cs.du.edu
4603 * Use ordered bins instead of best-fit threshold
4604 * Eliminate block-local decls to simplify tracing and debugging.
4605 * Support another case of realloc via move into top
4606 * Fix error occurring when initial sbrk_base not word-aligned.
4607 * Rely on page size for units instead of SBRK_UNIT to
4608 avoid surprises about sbrk alignment conventions.
4609 * Add mallinfo, mallopt. Thanks to Raymond Nijssen
4610 (raymond@es.ele.tue.nl) for the suggestion.
4611 * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
4612 * More precautions for cases where other routines call sbrk,
4613 courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
4614 * Added macros etc., allowing use in linux libc from
4615 H.J. Lu (hjl@gnu.ai.mit.edu)
4616 * Inverted this history list
4618 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)
4619 * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
4620 * Removed all preallocation code since under current scheme
4621 the work required to undo bad preallocations exceeds
4622 the work saved in good cases for most test programs.
4623 * No longer use return list or unconsolidated bins since
4624 no scheme using them consistently outperforms those that don't
4625 given above changes.
4626 * Use best fit for very large chunks to prevent some worst-cases.
4627 * Added some support for debugging
4629 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)
4630 * Removed footers when chunks are in use. Thanks to
4631 Paul Wilson (wilson@cs.texas.edu) for the suggestion.
4633 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)
4634 * Added malloc_trim, with help from Wolfram Gloger
4635 (wmglo@Dent.MED.Uni-Muenchen.DE).
4637 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)
4639 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)
4640 * realloc: try to expand in both directions
4641 * malloc: swap order of clean-bin strategy;
4642 * realloc: only conditionally expand backwards
4643 * Try not to scavenge used bins
4644 * Use bin counts as a guide to preallocation
4645 * Occasionally bin return list chunks in first scan
4646 * Add a few optimizations from colin@nyx10.cs.du.edu
4648 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)
4649 * faster bin computation & slightly different binning
4650 * merged all consolidations to one part of malloc proper
4651 (eliminating old malloc_find_space & malloc_clean_bin)
4652 * Scan 2 returns chunks (not just 1)
4653 * Propagate failure in realloc if malloc returns 0
4654 * Add stuff to allow compilation on non-ANSI compilers
4655 from kpv@research.att.com
4657 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)
4658 * removed potential for odd address access in prev_chunk
4659 * removed dependency on getpagesize.h
4660 * misc cosmetics and a bit more internal documentation
4661 * anticosmetics: mangled names in macros to evade debugger strangeness
4662 * tested on sparc, hp-700, dec-mips, rs6000
4663 with gcc & native cc (hp, dec only) allowing
4664 Detlefs & Zorn comparison study (in SIGPLAN Notices.)
4666 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)
4667 * Based loosely on libg++-1.2X malloc. (It retains some of the overall
4668 structure of old version, but most details differ.)