1 /* Malloc implementation for multiple threads without lock contention.
2 Copyright (C) 1996 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-pt2 Sat Dec 14 1996
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 Alignnment 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() */
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 */
318 /* We must not pollute the name space in the GNU libc. */
320 #define malloc_stats __malloc_stats
321 #define malloc_usable_size __malloc_usable_size
322 #define malloc_trim __malloc_trim
334 Because freed chunks may be overwritten with link fields, this
335 malloc will often die when freed memory is overwritten by user
336 programs. This can be very effective (albeit in an annoying way)
337 in helping track down dangling pointers.
339 If you compile with -DMALLOC_DEBUG, a number of assertion checks are
340 enabled that will catch more memory errors. You probably won't be
341 able to make much sense of the actual assertion errors, but they
342 should help you locate incorrectly overwritten memory. The
343 checking is fairly extensive, and will slow down execution
344 noticeably. Calling malloc_stats or mallinfo with MALLOC_DEBUG set will
345 attempt to check every non-mmapped allocated and free chunk in the
346 course of computing the summmaries. (By nature, mmapped regions
347 cannot be checked very much automatically.)
349 Setting MALLOC_DEBUG may also be helpful if you are trying to modify
350 this code. The assertions in the check routines spell out in more
351 detail the assumptions and invariants underlying the algorithms.
358 #define assert(x) ((void)0)
363 INTERNAL_SIZE_T is the word-size used for internal bookkeeping
364 of chunk sizes. On a 64-bit machine, you can reduce malloc
365 overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
366 at the expense of not being able to handle requests greater than
367 2^31. This limitation is hardly ever a concern; you are encouraged
368 to set this. However, the default version is the same as size_t.
371 #ifndef INTERNAL_SIZE_T
372 #define INTERNAL_SIZE_T size_t
376 REALLOC_ZERO_BYTES_FREES should be set if a call to
377 realloc with zero bytes should be the same as a call to free.
378 Some people think it should. Otherwise, since this malloc
379 returns a unique pointer for malloc(0), so does realloc(p, 0).
383 /* #define REALLOC_ZERO_BYTES_FREES */
387 HAVE_MEMCPY should be defined if you are not otherwise using
388 ANSI STD C, but still have memcpy and memset in your C library
389 and want to use them in calloc and realloc. Otherwise simple
390 macro versions are defined here.
392 USE_MEMCPY should be defined as 1 if you actually want to
393 have memset and memcpy called. People report that the macro
394 versions are often enough faster than libc versions on many
395 systems that it is better to use them.
399 #define HAVE_MEMCPY 1
409 #if (__STD_C || defined(HAVE_MEMCPY))
412 void* memset(void*, int, size_t);
413 void* memcpy(void*, const void*, size_t);
422 /* The following macros are only invoked with (2n+1)-multiples of
423 INTERNAL_SIZE_T units, with a positive integer n. This is exploited
424 for fast inline execution when n is small. */
426 #define MALLOC_ZERO(charp, nbytes) \
428 INTERNAL_SIZE_T mzsz = (nbytes); \
429 if(mzsz <= 9*sizeof(mzsz)) { \
430 INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
431 if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
433 if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
435 if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
440 } else memset((charp), 0, mzsz); \
443 #define MALLOC_COPY(dest,src,nbytes) \
445 INTERNAL_SIZE_T mcsz = (nbytes); \
446 if(mcsz <= 9*sizeof(mcsz)) { \
447 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
448 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
449 if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
450 *mcdst++ = *mcsrc++; \
451 if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
452 *mcdst++ = *mcsrc++; \
453 if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
454 *mcdst++ = *mcsrc++; }}} \
455 *mcdst++ = *mcsrc++; \
456 *mcdst++ = *mcsrc++; \
458 } else memcpy(dest, src, mcsz); \
461 #else /* !USE_MEMCPY */
463 /* Use Duff's device for good zeroing/copying performance. */
465 #define MALLOC_ZERO(charp, nbytes) \
467 INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
468 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
469 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
471 case 0: for(;;) { *mzp++ = 0; \
472 case 7: *mzp++ = 0; \
473 case 6: *mzp++ = 0; \
474 case 5: *mzp++ = 0; \
475 case 4: *mzp++ = 0; \
476 case 3: *mzp++ = 0; \
477 case 2: *mzp++ = 0; \
478 case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
482 #define MALLOC_COPY(dest,src,nbytes) \
484 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
485 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
486 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
487 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
489 case 0: for(;;) { *mcdst++ = *mcsrc++; \
490 case 7: *mcdst++ = *mcsrc++; \
491 case 6: *mcdst++ = *mcsrc++; \
492 case 5: *mcdst++ = *mcsrc++; \
493 case 4: *mcdst++ = *mcsrc++; \
494 case 3: *mcdst++ = *mcsrc++; \
495 case 2: *mcdst++ = *mcsrc++; \
496 case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
504 Define HAVE_MMAP to optionally make malloc() use mmap() to
505 allocate very large blocks. These will be returned to the
506 operating system immediately after a free().
514 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
515 large blocks. This is currently only possible on Linux with
516 kernel versions newer than 1.3.77.
520 #define HAVE_MREMAP defined(__linux__)
527 #include <sys/mman.h>
529 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
530 #define MAP_ANONYMOUS MAP_ANON
533 #endif /* HAVE_MMAP */
536 Access to system page size. To the extent possible, this malloc
537 manages memory from the system in page-size units.
539 The following mechanics for getpagesize were adapted from
540 bsd/gnu getpagesize.h
543 #ifndef LACKS_UNISTD_H
547 #ifndef malloc_getpagesize
548 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
549 # ifndef _SC_PAGE_SIZE
550 # define _SC_PAGE_SIZE _SC_PAGESIZE
553 # ifdef _SC_PAGE_SIZE
554 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
556 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
557 extern size_t getpagesize();
558 # define malloc_getpagesize getpagesize()
560 # include <sys/param.h>
561 # ifdef EXEC_PAGESIZE
562 # define malloc_getpagesize EXEC_PAGESIZE
566 # define malloc_getpagesize NBPG
568 # define malloc_getpagesize (NBPG * CLSIZE)
572 # define malloc_getpagesize NBPC
575 # define malloc_getpagesize PAGESIZE
577 # define malloc_getpagesize (4096) /* just guess */
590 This version of malloc supports the standard SVID/XPG mallinfo
591 routine that returns a struct containing the same kind of
592 information you can get from malloc_stats. It should work on
593 any SVID/XPG compliant system that has a /usr/include/malloc.h
594 defining struct mallinfo. (If you'd like to install such a thing
595 yourself, cut out the preliminary declarations as described above
596 and below and save them in a malloc.h file. But there's no
597 compelling reason to bother to do this.)
599 The main declaration needed is the mallinfo struct that is returned
600 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
601 bunch of fields, most of which are not even meaningful in this
602 version of malloc. Some of these fields are are instead filled by
603 mallinfo() with other numbers that might possibly be of interest.
605 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
606 /usr/include/malloc.h file that includes a declaration of struct
607 mallinfo. If so, it is included; else an SVID2/XPG2 compliant
608 version is declared below. These must be precisely the same for
613 /* #define HAVE_USR_INCLUDE_MALLOC_H */
615 #if HAVE_USR_INCLUDE_MALLOC_H
616 # include "/usr/include/malloc.h"
621 # include "ptmalloc.h"
627 #ifndef DEFAULT_TRIM_THRESHOLD
628 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
632 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
633 to keep before releasing via malloc_trim in free().
635 Automatic trimming is mainly useful in long-lived programs.
636 Because trimming via sbrk can be slow on some systems, and can
637 sometimes be wasteful (in cases where programs immediately
638 afterward allocate more large chunks) the value should be high
639 enough so that your overall system performance would improve by
642 The trim threshold and the mmap control parameters (see below)
643 can be traded off with one another. Trimming and mmapping are
644 two different ways of releasing unused memory back to the
645 system. Between these two, it is often possible to keep
646 system-level demands of a long-lived program down to a bare
647 minimum. For example, in one test suite of sessions measuring
648 the XF86 X server on Linux, using a trim threshold of 128K and a
649 mmap threshold of 192K led to near-minimal long term resource
652 If you are using this malloc in a long-lived program, it should
653 pay to experiment with these values. As a rough guide, you
654 might set to a value close to the average size of a process
655 (program) running on your system. Releasing this much memory
656 would allow such a process to run in memory. Generally, it's
657 worth it to tune for trimming rather tham memory mapping when a
658 program undergoes phases where several large chunks are
659 allocated and released in ways that can reuse each other's
660 storage, perhaps mixed with phases where there are no such
661 chunks at all. And in well-behaved long-lived programs,
662 controlling release of large blocks via trimming versus mapping
665 However, in most programs, these parameters serve mainly as
666 protection against the system-level effects of carrying around
667 massive amounts of unneeded memory. Since frequent calls to
668 sbrk, mmap, and munmap otherwise degrade performance, the default
669 parameters are set to relatively high values that serve only as
672 The default trim value is high enough to cause trimming only in
673 fairly extreme (by current memory consumption standards) cases.
674 It must be greater than page size to have any useful effect. To
675 disable trimming completely, you can set to (unsigned long)(-1);
681 #ifndef DEFAULT_TOP_PAD
682 #define DEFAULT_TOP_PAD (0)
686 M_TOP_PAD is the amount of extra `padding' space to allocate or
687 retain whenever sbrk is called. It is used in two ways internally:
689 * When sbrk is called to extend the top of the arena to satisfy
690 a new malloc request, this much padding is added to the sbrk
693 * When malloc_trim is called automatically from free(),
694 it is used as the `pad' argument.
696 In both cases, the actual amount of padding is rounded
697 so that the end of the arena is always a system page boundary.
699 The main reason for using padding is to avoid calling sbrk so
700 often. Having even a small pad greatly reduces the likelihood
701 that nearly every malloc request during program start-up (or
702 after trimming) will invoke sbrk, which needlessly wastes
705 Automatic rounding-up to page-size units is normally sufficient
706 to avoid measurable overhead, so the default is 0. However, in
707 systems where sbrk is relatively slow, it can pay to increase
708 this value, at the expense of carrying around more memory than
714 #ifndef DEFAULT_MMAP_THRESHOLD
715 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
720 M_MMAP_THRESHOLD is the request size threshold for using mmap()
721 to service a request. Requests of at least this size that cannot
722 be allocated using already-existing space will be serviced via mmap.
723 (If enough normal freed space already exists it is used instead.)
725 Using mmap segregates relatively large chunks of memory so that
726 they can be individually obtained and released from the host
727 system. A request serviced through mmap is never reused by any
728 other request (at least not directly; the system may just so
729 happen to remap successive requests to the same locations).
731 Segregating space in this way has the benefit that mmapped space
732 can ALWAYS be individually released back to the system, which
733 helps keep the system level memory demands of a long-lived
734 program low. Mapped memory can never become `locked' between
735 other chunks, as can happen with normally allocated chunks, which
736 menas that even trimming via malloc_trim would not release them.
738 However, it has the disadvantages that:
740 1. The space cannot be reclaimed, consolidated, and then
741 used to service later requests, as happens with normal chunks.
742 2. It can lead to more wastage because of mmap page alignment
744 3. It causes malloc performance to be more dependent on host
745 system memory management support routines which may vary in
746 implementation quality and may impose arbitrary
747 limitations. Generally, servicing a request via normal
748 malloc steps is faster than going through a system's mmap.
750 All together, these considerations should lead you to use mmap
751 only for relatively large requests.
758 #ifndef DEFAULT_MMAP_MAX
760 #define DEFAULT_MMAP_MAX (1024)
762 #define DEFAULT_MMAP_MAX (0)
767 M_MMAP_MAX is the maximum number of requests to simultaneously
768 service using mmap. This parameter exists because:
770 1. Some systems have a limited number of internal tables for
772 2. In most systems, overreliance on mmap can degrade overall
774 3. If a program allocates many large regions, it is probably
775 better off using normal sbrk-based allocation routines that
776 can reclaim and reallocate normal heap memory. Using a
777 small value allows transition into this mode after the
778 first few allocations.
780 Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
781 the default value is 0, and attempts to set it to non-zero values
782 in mallopt will fail.
787 #ifndef DEFAULT_CHECK_ACTION
788 #define DEFAULT_CHECK_ACTION 1
791 /* What to do if the standard debugging hooks are in place and a
792 corrupt pointer is detected: do nothing (0), print an error message
793 (1), or call abort() (2). */
797 #define HEAP_MIN_SIZE (32*1024)
798 #define HEAP_MAX_SIZE (1024*1024) /* must be a power of two */
800 /* HEAP_MIN_SIZE and HEAP_MAX_SIZE limit the size of mmap()ed heaps
801 that are dynamically created for multi-threaded programs. The
802 maximum size must be a power of two, for fast determination of
803 which heap belongs to a chunk. It should be much larger than
804 the mmap threshold, so that requests with a size just below that
805 threshold can be fulfilled without creating too many heaps.
811 #define THREAD_STATS 0
814 /* If THREAD_STATS is non-zero, some statistics on mutex locking are
820 Special defines for the Linux/GNU C library.
829 Void_t
* __default_morecore (ptrdiff_t);
830 static Void_t
*(*__morecore
)(ptrdiff_t) = __default_morecore
;
834 Void_t
* __default_morecore ();
835 static Void_t
*(*__morecore
)() = __default_morecore
;
839 #define MORECORE (*__morecore)
840 #define MORECORE_FAILURE 0
841 #define MORECORE_CLEARS 1
843 #define munmap __munmap
844 #define mremap __mremap
845 #undef malloc_getpagesize
846 #define malloc_getpagesize __getpagesize()
851 extern Void_t
* sbrk(ptrdiff_t);
853 extern Void_t
* sbrk();
857 #define MORECORE sbrk
860 #ifndef MORECORE_FAILURE
861 #define MORECORE_FAILURE -1
864 #ifndef MORECORE_CLEARS
865 #define MORECORE_CLEARS 1
872 #define cALLOc __libc_calloc
873 #define fREe __libc_free
874 #define mALLOc __libc_malloc
875 #define mEMALIGn __libc_memalign
876 #define rEALLOc __libc_realloc
877 #define vALLOc __libc_valloc
878 #define pvALLOc __libc_pvalloc
879 #define mALLINFo __libc_mallinfo
880 #define mALLOPt __libc_mallopt
884 #define cALLOc calloc
886 #define mALLOc malloc
887 #define mEMALIGn memalign
888 #define rEALLOc realloc
889 #define vALLOc valloc
890 #define pvALLOc pvalloc
891 #define mALLINFo mallinfo
892 #define mALLOPt mallopt
896 /* Public routines */
901 void ptmalloc_init(void);
903 Void_t
* mALLOc(size_t);
905 Void_t
* rEALLOc(Void_t
*, size_t);
906 Void_t
* mEMALIGn(size_t, size_t);
907 Void_t
* vALLOc(size_t);
908 Void_t
* pvALLOc(size_t);
909 Void_t
* cALLOc(size_t, size_t);
911 int __malloc_trim(size_t);
912 int malloc_trim(size_t);
913 size_t __malloc_usable_size(Void_t
*);
914 size_t malloc_usable_size(Void_t
*);
915 void __malloc_stats(void);
916 void malloc_stats(void);
917 int mALLOPt(int, int);
918 struct mallinfo
mALLINFo(void);
921 void ptmalloc_init();
933 size_t _malloc_usable_size();
934 size_t malloc_usable_size();
935 void __malloc_stats();
938 struct mallinfo
mALLINFo();
943 }; /* end of extern "C" */
946 #if !defined(NO_THREADS) && !HAVE_MMAP
947 "Can't have threads support without mmap"
958 INTERNAL_SIZE_T prev_size
; /* Size of previous chunk (if free). */
959 INTERNAL_SIZE_T size
; /* Size in bytes, including overhead. */
960 struct malloc_chunk
* fd
; /* double links -- used only if free. */
961 struct malloc_chunk
* bk
;
964 typedef struct malloc_chunk
* mchunkptr
;
968 malloc_chunk details:
970 (The following includes lightly edited explanations by Colin Plumb.)
972 Chunks of memory are maintained using a `boundary tag' method as
973 described in e.g., Knuth or Standish. (See the paper by Paul
974 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
975 survey of such techniques.) Sizes of free chunks are stored both
976 in the front of each chunk and at the end. This makes
977 consolidating fragmented chunks into bigger chunks very fast. The
978 size fields also hold bits representing whether chunks are free or
981 An allocated chunk looks like this:
984 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
985 | Size of previous chunk, if allocated | |
986 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
987 | Size of chunk, in bytes |P|
988 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
989 | User data starts here... .
991 . (malloc_usable_space() bytes) .
993 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
995 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
998 Where "chunk" is the front of the chunk for the purpose of most of
999 the malloc code, but "mem" is the pointer that is returned to the
1000 user. "Nextchunk" is the beginning of the next contiguous chunk.
1002 Chunks always begin on even word boundries, so the mem portion
1003 (which is returned to the user) is also on an even word boundary, and
1004 thus double-word aligned.
1006 Free chunks are stored in circular doubly-linked lists, and look like this:
1008 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1009 | Size of previous chunk |
1010 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1011 `head:' | Size of chunk, in bytes |P|
1012 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1013 | Forward pointer to next chunk in list |
1014 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1015 | Back pointer to previous chunk in list |
1016 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1017 | Unused space (may be 0 bytes long) .
1020 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1021 `foot:' | Size of chunk, in bytes |
1022 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1024 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1025 chunk size (which is always a multiple of two words), is an in-use
1026 bit for the *previous* chunk. If that bit is *clear*, then the
1027 word before the current chunk size contains the previous chunk
1028 size, and can be used to find the front of the previous chunk.
1029 (The very first chunk allocated always has this bit set,
1030 preventing access to non-existent (or non-owned) memory.)
1032 Note that the `foot' of the current chunk is actually represented
1033 as the prev_size of the NEXT chunk. (This makes it easier to
1034 deal with alignments etc).
1036 The two exceptions to all this are
1038 1. The special chunk `top', which doesn't bother using the
1039 trailing size field since there is no
1040 next contiguous chunk that would have to index off it. (After
1041 initialization, `top' is forced to always exist. If it would
1042 become less than MINSIZE bytes long, it is replenished via
1045 2. Chunks allocated via mmap, which have the second-lowest-order
1046 bit (IS_MMAPPED) set in their size fields. Because they are
1047 never merged or traversed from any other chunk, they have no
1048 foot size or inuse information.
1050 Available chunks are kept in any of several places (all declared below):
1052 * `av': An array of chunks serving as bin headers for consolidated
1053 chunks. Each bin is doubly linked. The bins are approximately
1054 proportionally (log) spaced. There are a lot of these bins
1055 (128). This may look excessive, but works very well in
1056 practice. All procedures maintain the invariant that no
1057 consolidated chunk physically borders another one. Chunks in
1058 bins are kept in size order, with ties going to the
1059 approximately least recently used chunk.
1061 The chunks in each bin are maintained in decreasing sorted order by
1062 size. This is irrelevant for the small bins, which all contain
1063 the same-sized chunks, but facilitates best-fit allocation for
1064 larger chunks. (These lists are just sequential. Keeping them in
1065 order almost never requires enough traversal to warrant using
1066 fancier ordered data structures.) Chunks of the same size are
1067 linked with the most recently freed at the front, and allocations
1068 are taken from the back. This results in LRU or FIFO allocation
1069 order, which tends to give each chunk an equal opportunity to be
1070 consolidated with adjacent freed chunks, resulting in larger free
1071 chunks and less fragmentation.
1073 * `top': The top-most available chunk (i.e., the one bordering the
1074 end of available memory) is treated specially. It is never
1075 included in any bin, is used only if no other chunk is
1076 available, and is released back to the system if it is very
1077 large (see M_TRIM_THRESHOLD).
1079 * `last_remainder': A bin holding only the remainder of the
1080 most recently split (non-top) chunk. This bin is checked
1081 before other non-fitting chunks, so as to provide better
1082 locality for runs of sequentially allocated chunks.
1084 * Implicitly, through the host system's memory mapping tables.
1085 If supported, requests greater than a threshold are usually
1086 serviced via calls to mmap, and then later released via munmap.
1093 The bins are an array of pairs of pointers serving as the
1094 heads of (initially empty) doubly-linked lists of chunks, laid out
1095 in a way so that each pair can be treated as if it were in a
1096 malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1097 and chunks are the same).
1099 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1100 8 bytes apart. Larger bins are approximately logarithmically
1101 spaced. (See the table below.)
1109 4 bins of size 32768
1110 2 bins of size 262144
1111 1 bin of size what's left
1113 There is actually a little bit of slop in the numbers in bin_index
1114 for the sake of speed. This makes no difference elsewhere.
1116 The special chunks `top' and `last_remainder' get their own bins,
1117 (this is implemented via yet more trickery with the av array),
1118 although `top' is never properly linked to its bin since it is
1119 always handled specially.
1123 #define NAV 128 /* number of bins */
1125 typedef struct malloc_chunk
* mbinptr
;
1127 /* An arena is a configuration of malloc_chunks together with an array
1128 of bins. With multiple threads, it must be locked via a mutex
1129 before changing its data structures. One or more `heaps' are
1130 associated with each arena, except for the main_arena, which is
1131 associated only with the `main heap', i.e. the conventional free
1132 store obtained with calls to MORECORE() (usually sbrk). The `av'
1133 array is never mentioned directly in the code, but instead used via
1134 bin access macros. */
1136 typedef struct _arena
{
1137 mbinptr av
[2*NAV
+ 2];
1138 struct _arena
*next
;
1141 long stat_lock_direct
, stat_lock_loop
, stat_lock_wait
;
1147 /* A heap is a single contiguous memory region holding (coalescable)
1148 malloc_chunks. It is allocated with mmap() and always starts at an
1149 address aligned to HEAP_MAX_SIZE. Not used unless compiling for
1150 multiple threads. */
1152 typedef struct _heap_info
{
1153 arena
*ar_ptr
; /* Arena for this heap. */
1154 struct _heap_info
*prev
; /* Previous heap. */
1155 size_t size
; /* Current size in bytes. */
1156 size_t pad
; /* Make sure the following data is properly aligned. */
1161 Static functions (forward declarations)
1166 static void chunk_free(arena
*ar_ptr
, mchunkptr p
);
1167 static mchunkptr
chunk_alloc(arena
*ar_ptr
, INTERNAL_SIZE_T size
);
1168 static mchunkptr
chunk_realloc(arena
*ar_ptr
, mchunkptr oldp
,
1169 INTERNAL_SIZE_T oldsize
, INTERNAL_SIZE_T nb
);
1170 static mchunkptr
chunk_align(arena
*ar_ptr
, INTERNAL_SIZE_T nb
,
1172 static int main_trim(size_t pad
);
1174 static int heap_trim(heap_info
*heap
, size_t pad
);
1176 #if defined(_LIBC) || defined(MALLOC_HOOKS)
1177 static Void_t
* malloc_check(size_t sz
);
1178 static void free_check(Void_t
* mem
);
1179 static Void_t
* realloc_check(Void_t
* oldmem
, size_t bytes
);
1180 static Void_t
* memalign_check(size_t alignment
, size_t bytes
);
1185 static void chunk_free();
1186 static mchunkptr
chunk_alloc();
1187 static mchunkptr
chunk_realloc();
1188 static mchunkptr
chunk_align();
1189 static int main_trim();
1191 static int heap_trim();
1193 #if defined(_LIBC) || defined(MALLOC_HOOKS)
1194 static Void_t
* malloc_check();
1195 static void free_check();
1196 static Void_t
* realloc_check();
1197 static Void_t
* memalign_check();
1204 /* sizes, alignments */
1206 #define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
1207 #define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
1208 #define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
1209 #define MINSIZE (sizeof(struct malloc_chunk))
1211 /* conversion from malloc headers to user pointers, and back */
1213 #define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1214 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1216 /* pad request bytes into a usable size */
1218 #define request2size(req) \
1219 (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1220 (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1221 (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1223 /* Check if m has acceptable alignment */
1225 #define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1231 Physical chunk operations
1235 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1237 #define PREV_INUSE 0x1
1239 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1241 #define IS_MMAPPED 0x2
1243 /* Bits to mask off when extracting size */
1245 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1248 /* Ptr to next physical malloc_chunk. */
1250 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1252 /* Ptr to previous physical malloc_chunk */
1254 #define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1257 /* Treat space at ptr + offset as a chunk */
1259 #define chunk_at_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
1265 Dealing with use bits
1268 /* extract p's inuse bit */
1271 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1273 /* extract inuse bit of previous chunk */
1275 #define prev_inuse(p) ((p)->size & PREV_INUSE)
1277 /* check for mmap()'ed chunk */
1279 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1281 /* set/clear chunk as in use without otherwise disturbing */
1283 #define set_inuse(p) \
1284 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1286 #define clear_inuse(p) \
1287 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1289 /* check/set/clear inuse bits in known places */
1291 #define inuse_bit_at_offset(p, s)\
1292 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1294 #define set_inuse_bit_at_offset(p, s)\
1295 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1297 #define clear_inuse_bit_at_offset(p, s)\
1298 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1304 Dealing with size fields
1307 /* Get size, ignoring use bits */
1309 #define chunksize(p) ((p)->size & ~(SIZE_BITS))
1311 /* Set size at head, without disturbing its use bit */
1313 #define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1315 /* Set size/use ignoring previous bits in header */
1317 #define set_head(p, s) ((p)->size = (s))
1319 /* Set size at footer (only when chunk is not in use) */
1321 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1329 #define bin_at(a, i) ((mbinptr)((char*)&(((a)->av)[2*(i) + 2]) - 2*SIZE_SZ))
1330 #define init_bin(a, i) ((a)->av[2*i+2] = (a)->av[2*i+3] = bin_at((a), i))
1331 #define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1332 #define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1335 The first 2 bins are never indexed. The corresponding av cells are instead
1336 used for bookkeeping. This is not to save space, but to simplify
1337 indexing, maintain locality, and avoid some initialization tests.
1340 #define binblocks(a) (bin_at(a,0)->size)/* bitvector of nonempty blocks */
1341 #define top(a) (bin_at(a,0)->fd) /* The topmost chunk */
1342 #define last_remainder(a) (bin_at(a,1)) /* remainder from last split */
1345 Because top initially points to its own bin with initial
1346 zero size, thus forcing extension on the first malloc request,
1347 we avoid having any special code in malloc to check whether
1348 it even exists yet. But we still need to in malloc_extend_top.
1351 #define initial_top(a) ((mchunkptr)bin_at(a, 0))
1355 /* field-extraction macros */
1357 #define first(b) ((b)->fd)
1358 #define last(b) ((b)->bk)
1364 #define bin_index(sz) \
1365 (((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3): \
1366 ((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6): \
1367 ((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9): \
1368 ((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12): \
1369 ((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15): \
1370 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1373 bins for chunks < 512 are all spaced 8 bytes apart, and hold
1374 identically sized chunks. This is exploited in malloc.
1377 #define MAX_SMALLBIN 63
1378 #define MAX_SMALLBIN_SIZE 512
1379 #define SMALLBIN_WIDTH 8
1381 #define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
1384 Requests are `small' if both the corresponding and the next bin are small
1387 #define is_small_request(nb) ((nb) < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1392 To help compensate for the large number of bins, a one-level index
1393 structure is used for bin-by-bin searching. `binblocks' is a
1394 one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1395 have any (possibly) non-empty bins, so they can be skipped over
1396 all at once during during traversals. The bits are NOT always
1397 cleared as soon as all bins in a block are empty, but instead only
1398 when all are noticed to be empty during traversal in malloc.
1401 #define BINBLOCKWIDTH 4 /* bins per block */
1403 /* bin<->block macros */
1405 #define idx2binblock(ix) ((unsigned)1 << ((ix) / BINBLOCKWIDTH))
1406 #define mark_binblock(a, ii) (binblocks(a) |= idx2binblock(ii))
1407 #define clear_binblock(a, ii) (binblocks(a) &= ~(idx2binblock(ii)))
1412 /* Static bookkeeping data */
1414 /* Helper macro to initialize bins */
1415 #define IAV(i) bin_at(&main_arena, i), bin_at(&main_arena, i)
1417 static arena main_arena
= {
1420 IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
1421 IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
1422 IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
1423 IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
1424 IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
1425 IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
1426 IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
1427 IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
1428 IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
1429 IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
1430 IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
1431 IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
1432 IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
1433 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1434 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1435 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1440 0, 0, 0, /* stat_lock_direct, stat_lock_loop, stat_lock_wait */
1442 MUTEX_INITIALIZER
/* mutex */
1447 /* Thread specific data */
1450 static tsd_key_t arena_key
;
1451 static mutex_t list_lock
= MUTEX_INITIALIZER
;
1455 static int stat_n_heaps
= 0;
1456 #define THREAD_STAT(x) x
1458 #define THREAD_STAT(x) do ; while(0)
1461 /* variables holding tunable values */
1463 static unsigned long trim_threshold
= DEFAULT_TRIM_THRESHOLD
;
1464 static unsigned long top_pad
= DEFAULT_TOP_PAD
;
1465 static unsigned int n_mmaps_max
= DEFAULT_MMAP_MAX
;
1466 static unsigned long mmap_threshold
= DEFAULT_MMAP_THRESHOLD
;
1467 static int check_action
= DEFAULT_CHECK_ACTION
;
1469 /* The first value returned from sbrk */
1470 static char* sbrk_base
= (char*)(-1);
1472 /* The maximum memory obtained from system via sbrk */
1473 static unsigned long max_sbrked_mem
= 0;
1475 /* The maximum via either sbrk or mmap (too difficult to track with threads) */
1477 static unsigned long max_total_mem
= 0;
1480 /* The total memory obtained from system via sbrk */
1481 #define sbrked_mem (main_arena.size)
1483 /* Tracking mmaps */
1485 static unsigned int n_mmaps
= 0;
1486 static unsigned int max_n_mmaps
= 0;
1487 static unsigned long mmapped_mem
= 0;
1488 static unsigned long max_mmapped_mem
= 0;
1494 /* Initialization routine. */
1497 static void ptmalloc_init
__MALLOC_P ((void)) __attribute__ ((constructor
));
1501 ptmalloc_init
__MALLOC_P((void))
1504 ptmalloc_init
__MALLOC_P((void))
1507 static int first
= 1;
1508 #if defined(_LIBC) || defined(MALLOC_HOOKS)
1515 /* Initialize the pthreads interface. */
1516 if (__pthread_initialize
!= NULL
)
1517 __pthread_initialize();
1520 mutex_init(&main_arena
.mutex
);
1521 mutex_init(&list_lock
);
1522 tsd_key_create(&arena_key
, NULL
);
1523 tsd_setspecific(arena_key
, (Void_t
*)&main_arena
);
1525 #if defined(_LIBC) || defined(MALLOC_HOOKS)
1526 s
= getenv("MALLOC_CHECK_");
1528 if(s
[0]) mallopt(M_CHECK_ACTION
, (int)(s
[0] - '0'));
1529 malloc_check_init();
1531 if(__malloc_initialize_hook
!= NULL
)
1532 (*__malloc_initialize_hook
)();
1536 #if defined(_LIBC) || defined(MALLOC_HOOKS)
1538 /* Hooks for debugging versions. The initial hooks just call the
1539 initialization routine, then do the normal work. */
1543 malloc_hook_ini(size_t sz
)
1545 malloc_hook_ini(sz
) size_t sz
;
1548 __malloc_hook
= NULL
;
1549 __realloc_hook
= NULL
;
1550 __memalign_hook
= NULL
;
1557 realloc_hook_ini(Void_t
* ptr
, size_t sz
)
1559 realloc_hook_ini(ptr
, sz
) Void_t
* ptr
; size_t sz
;
1562 __malloc_hook
= NULL
;
1563 __realloc_hook
= NULL
;
1564 __memalign_hook
= NULL
;
1566 return rEALLOc(ptr
, sz
);
1571 memalign_hook_ini(size_t sz
, size_t alignment
)
1573 memalign_hook_ini(sz
, alignment
) size_t sz
; size_t alignment
;
1576 __malloc_hook
= NULL
;
1577 __realloc_hook
= NULL
;
1578 __memalign_hook
= NULL
;
1580 return mEMALIGn(sz
, alignment
);
1583 void (*__malloc_initialize_hook
) __MALLOC_P ((void)) = NULL
;
1584 void (*__free_hook
) __MALLOC_P ((__malloc_ptr_t __ptr
)) = NULL
;
1585 __malloc_ptr_t (*__malloc_hook
)
1586 __MALLOC_P ((size_t __size
)) = malloc_hook_ini
;
1587 __malloc_ptr_t (*__realloc_hook
)
1588 __MALLOC_P ((__malloc_ptr_t __ptr
, size_t __size
)) = realloc_hook_ini
;
1589 __malloc_ptr_t (*__memalign_hook
)
1590 __MALLOC_P ((size_t __size
, size_t __alignment
)) = memalign_hook_ini
;
1592 /* Activate a standard set of debugging hooks. */
1596 __malloc_hook
= malloc_check
;
1597 __free_hook
= free_check
;
1598 __realloc_hook
= realloc_check
;
1599 __memalign_hook
= memalign_check
;
1600 fprintf(stderr
, "Using debugging hooks\n");
1609 /* Routines dealing with mmap(). */
1613 #ifndef MAP_ANONYMOUS
1615 static int dev_zero_fd
= -1; /* Cached file descriptor for /dev/zero. */
1617 #define MMAP(size, prot) ((dev_zero_fd < 0) ? \
1618 (dev_zero_fd = open("/dev/zero", O_RDWR), \
1619 mmap(0, (size), (prot), MAP_PRIVATE, dev_zero_fd, 0)) : \
1620 mmap(0, (size), (prot), MAP_PRIVATE, dev_zero_fd, 0))
1624 #define MMAP(size, prot) \
1625 (mmap(0, (size), (prot), MAP_PRIVATE|MAP_ANONYMOUS, -1, 0))
1630 static mchunkptr
mmap_chunk(size_t size
)
1632 static mchunkptr
mmap_chunk(size
) size_t size
;
1635 size_t page_mask
= malloc_getpagesize
- 1;
1638 if(n_mmaps
>= n_mmaps_max
) return 0; /* too many regions */
1640 /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1641 * there is no following chunk whose prev_size field could be used.
1643 size
= (size
+ SIZE_SZ
+ page_mask
) & ~page_mask
;
1645 p
= (mchunkptr
)MMAP(size
, PROT_READ
|PROT_WRITE
);
1646 if(p
== (mchunkptr
)-1) return 0;
1649 if (n_mmaps
> max_n_mmaps
) max_n_mmaps
= n_mmaps
;
1651 /* We demand that eight bytes into a page must be 8-byte aligned. */
1652 assert(aligned_OK(chunk2mem(p
)));
1654 /* The offset to the start of the mmapped region is stored
1655 * in the prev_size field of the chunk; normally it is zero,
1656 * but that can be changed in memalign().
1659 set_head(p
, size
|IS_MMAPPED
);
1661 mmapped_mem
+= size
;
1662 if ((unsigned long)mmapped_mem
> (unsigned long)max_mmapped_mem
)
1663 max_mmapped_mem
= mmapped_mem
;
1665 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) > (unsigned long)max_total_mem
)
1666 max_total_mem
= mmapped_mem
+ sbrked_mem
;
1672 static void munmap_chunk(mchunkptr p
)
1674 static void munmap_chunk(p
) mchunkptr p
;
1677 INTERNAL_SIZE_T size
= chunksize(p
);
1680 assert (chunk_is_mmapped(p
));
1681 assert(! ((char*)p
>= sbrk_base
&& (char*)p
< sbrk_base
+ sbrked_mem
));
1682 assert((n_mmaps
> 0));
1683 assert(((p
->prev_size
+ size
) & (malloc_getpagesize
-1)) == 0);
1686 mmapped_mem
-= (size
+ p
->prev_size
);
1688 ret
= munmap((char *)p
- p
->prev_size
, size
+ p
->prev_size
);
1690 /* munmap returns non-zero on failure */
1697 static mchunkptr
mremap_chunk(mchunkptr p
, size_t new_size
)
1699 static mchunkptr
mremap_chunk(p
, new_size
) mchunkptr p
; size_t new_size
;
1702 size_t page_mask
= malloc_getpagesize
- 1;
1703 INTERNAL_SIZE_T offset
= p
->prev_size
;
1704 INTERNAL_SIZE_T size
= chunksize(p
);
1707 assert (chunk_is_mmapped(p
));
1708 assert(! ((char*)p
>= sbrk_base
&& (char*)p
< sbrk_base
+ sbrked_mem
));
1709 assert((n_mmaps
> 0));
1710 assert(((size
+ offset
) & (malloc_getpagesize
-1)) == 0);
1712 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1713 new_size
= (new_size
+ offset
+ SIZE_SZ
+ page_mask
) & ~page_mask
;
1715 cp
= (char *)mremap((char *)p
- offset
, size
+ offset
, new_size
,
1718 if (cp
== (char *)-1) return 0;
1720 p
= (mchunkptr
)(cp
+ offset
);
1722 assert(aligned_OK(chunk2mem(p
)));
1724 assert((p
->prev_size
== offset
));
1725 set_head(p
, (new_size
- offset
)|IS_MMAPPED
);
1727 mmapped_mem
-= size
+ offset
;
1728 mmapped_mem
+= new_size
;
1729 if ((unsigned long)mmapped_mem
> (unsigned long)max_mmapped_mem
)
1730 max_mmapped_mem
= mmapped_mem
;
1732 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) > (unsigned long)max_total_mem
)
1733 max_total_mem
= mmapped_mem
+ sbrked_mem
;
1738 #endif /* HAVE_MREMAP */
1740 #endif /* HAVE_MMAP */
1744 /* Managing heaps and arenas (for concurrent threads) */
1748 /* Create a new heap. size is automatically rounded up to a multiple
1749 of the page size. */
1753 new_heap(size_t size
)
1755 new_heap(size
) size_t size
;
1758 size_t page_mask
= malloc_getpagesize
- 1;
1763 if(size
< HEAP_MIN_SIZE
)
1764 size
= HEAP_MIN_SIZE
;
1765 size
= (size
+ page_mask
) & ~page_mask
;
1766 if(size
> HEAP_MAX_SIZE
)
1768 p1
= (char *)MMAP(HEAP_MAX_SIZE
<<1, PROT_NONE
);
1769 if(p1
== (char *)-1)
1771 p2
= (char *)(((unsigned long)p1
+ HEAP_MAX_SIZE
) & ~(HEAP_MAX_SIZE
-1));
1774 munmap(p2
+ HEAP_MAX_SIZE
, HEAP_MAX_SIZE
- ul
);
1775 if(mprotect(p2
, size
, PROT_READ
|PROT_WRITE
) != 0) {
1776 munmap(p2
, HEAP_MAX_SIZE
);
1779 h
= (heap_info
*)p2
;
1781 THREAD_STAT(stat_n_heaps
++);
1785 /* Grow or shrink a heap. size is automatically rounded up to a
1786 multiple of the page size if it is positive. */
1790 grow_heap(heap_info
*h
, long diff
)
1792 grow_heap(h
, diff
) heap_info
*h
; long diff
;
1795 size_t page_mask
= malloc_getpagesize
- 1;
1799 diff
= (diff
+ page_mask
) & ~page_mask
;
1800 new_size
= (long)h
->size
+ diff
;
1801 if(new_size
> HEAP_MAX_SIZE
)
1803 if(mprotect((char *)h
+ h
->size
, diff
, PROT_READ
|PROT_WRITE
) != 0)
1806 new_size
= (long)h
->size
+ diff
;
1807 if(new_size
< (long)sizeof(*h
))
1809 if(mprotect((char *)h
+ new_size
, -diff
, PROT_NONE
) != 0)
1816 /* Delete a heap. */
1818 #define delete_heap(heap) munmap((char*)(heap), HEAP_MAX_SIZE)
1820 /* arena_get() acquires an arena and locks the corresponding mutex.
1821 First, try the one last locked successfully by this thread. (This
1822 is the common case and handled with a macro for speed.) Then, loop
1823 over the singly linked list of arenas. If no arena is readily
1824 available, create a new one. */
1826 #define arena_get(ptr, size) do { \
1827 Void_t *vptr = NULL; \
1828 ptr = (arena *)tsd_getspecific(arena_key, vptr); \
1829 if(ptr && !mutex_trylock(&ptr->mutex)) { \
1830 THREAD_STAT(++(ptr->stat_lock_direct)); \
1832 ptr = arena_get2(ptr, (size)); \
1838 arena_get2(arena
*a_tsd
, size_t size
)
1840 arena_get2(a_tsd
, size
) arena
*a_tsd
; size_t size
;
1847 unsigned long misalign
;
1849 /* Check the singly-linked list for unlocked arenas. */
1851 for(a
= a_tsd
->next
; a
; a
= a
->next
) {
1852 if(!mutex_trylock(&a
->mutex
))
1856 for(a
= &main_arena
; a
!= a_tsd
; a
= a
->next
) {
1857 if(!mutex_trylock(&a
->mutex
))
1861 /* Nothing immediately available, so generate a new arena. */
1862 h
= new_heap(size
+ (sizeof(*h
) + sizeof(*a
) + MALLOC_ALIGNMENT
));
1865 a
= h
->ar_ptr
= (arena
*)(h
+1);
1866 for(i
=0; i
<NAV
; i
++)
1869 mutex_init(&a
->mutex
);
1870 i
= mutex_lock(&a
->mutex
); /* remember result */
1872 /* Set up the top chunk, with proper alignment. */
1873 ptr
= (char *)(a
+ 1);
1874 misalign
= (unsigned long)chunk2mem(ptr
) & MALLOC_ALIGN_MASK
;
1876 ptr
+= MALLOC_ALIGNMENT
- misalign
;
1877 top(a
) = (mchunkptr
)ptr
;
1878 set_head(top(a
), (((char*)h
+ h
->size
) - ptr
) | PREV_INUSE
);
1880 /* Add the new arena to the list. */
1881 (void)mutex_lock(&list_lock
);
1882 a
->next
= main_arena
.next
;
1883 main_arena
.next
= a
;
1884 (void)mutex_unlock(&list_lock
);
1886 if(i
) /* locking failed; keep arena for further attempts later */
1890 THREAD_STAT(++(a
->stat_lock_loop
));
1891 tsd_setspecific(arena_key
, (Void_t
*)a
);
1895 /* find the heap and corresponding arena for a given ptr */
1897 #define heap_for_ptr(ptr) \
1898 ((heap_info *)((unsigned long)(ptr) & ~(HEAP_MAX_SIZE-1)))
1899 #define arena_for_ptr(ptr) \
1900 (((mchunkptr)(ptr) < top(&main_arena) && (char *)(ptr) >= sbrk_base) ? \
1901 &main_arena : heap_for_ptr(ptr)->ar_ptr)
1903 #else /* defined(NO_THREADS) */
1905 /* Without concurrent threads, there is only one arena. */
1907 #define arena_get(ptr, sz) (ptr = &main_arena)
1908 #define arena_for_ptr(ptr) (&main_arena)
1910 #endif /* !defined(NO_THREADS) */
1922 These routines make a number of assertions about the states
1923 of data structures that should be true at all times. If any
1924 are not true, it's very likely that a user program has somehow
1925 trashed memory. (It's also possible that there is a coding error
1926 in malloc. In which case, please report it!)
1930 static void do_check_chunk(arena
*ar_ptr
, mchunkptr p
)
1932 static void do_check_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
1935 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
1937 /* No checkable chunk is mmapped */
1938 assert(!chunk_is_mmapped(p
));
1941 if(ar_ptr
!= &main_arena
) {
1942 heap_info
*heap
= heap_for_ptr(p
);
1943 assert(heap
->ar_ptr
== ar_ptr
);
1944 assert((char *)p
+ sz
<= (char *)heap
+ heap
->size
);
1949 /* Check for legal address ... */
1950 assert((char*)p
>= sbrk_base
);
1951 if (p
!= top(ar_ptr
))
1952 assert((char*)p
+ sz
<= (char*)top(ar_ptr
));
1954 assert((char*)p
+ sz
<= sbrk_base
+ sbrked_mem
);
1960 static void do_check_free_chunk(arena
*ar_ptr
, mchunkptr p
)
1962 static void do_check_free_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
1965 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
1966 mchunkptr next
= chunk_at_offset(p
, sz
);
1968 do_check_chunk(ar_ptr
, p
);
1970 /* Check whether it claims to be free ... */
1973 /* Must have OK size and fields */
1974 assert((long)sz
>= (long)MINSIZE
);
1975 assert((sz
& MALLOC_ALIGN_MASK
) == 0);
1976 assert(aligned_OK(chunk2mem(p
)));
1977 /* ... matching footer field */
1978 assert(next
->prev_size
== sz
);
1979 /* ... and is fully consolidated */
1980 assert(prev_inuse(p
));
1981 assert (next
== top(ar_ptr
) || inuse(next
));
1983 /* ... and has minimally sane links */
1984 assert(p
->fd
->bk
== p
);
1985 assert(p
->bk
->fd
== p
);
1989 static void do_check_inuse_chunk(arena
*ar_ptr
, mchunkptr p
)
1991 static void do_check_inuse_chunk(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
1994 mchunkptr next
= next_chunk(p
);
1995 do_check_chunk(ar_ptr
, p
);
1997 /* Check whether it claims to be in use ... */
2000 /* ... whether its size is OK (it might be a fencepost) ... */
2001 assert(chunksize(p
) >= MINSIZE
|| next
->size
== (0|PREV_INUSE
));
2003 /* ... and is surrounded by OK chunks.
2004 Since more things can be checked with free chunks than inuse ones,
2005 if an inuse chunk borders them and debug is on, it's worth doing them.
2009 mchunkptr prv
= prev_chunk(p
);
2010 assert(next_chunk(prv
) == p
);
2011 do_check_free_chunk(ar_ptr
, prv
);
2013 if (next
== top(ar_ptr
))
2015 assert(prev_inuse(next
));
2016 assert(chunksize(next
) >= MINSIZE
);
2018 else if (!inuse(next
))
2019 do_check_free_chunk(ar_ptr
, next
);
2024 static void do_check_malloced_chunk(arena
*ar_ptr
,
2025 mchunkptr p
, INTERNAL_SIZE_T s
)
2027 static void do_check_malloced_chunk(ar_ptr
, p
, s
)
2028 arena
*ar_ptr
; mchunkptr p
; INTERNAL_SIZE_T s
;
2031 INTERNAL_SIZE_T sz
= p
->size
& ~PREV_INUSE
;
2034 do_check_inuse_chunk(ar_ptr
, p
);
2036 /* Legal size ... */
2037 assert((long)sz
>= (long)MINSIZE
);
2038 assert((sz
& MALLOC_ALIGN_MASK
) == 0);
2040 assert(room
< (long)MINSIZE
);
2042 /* ... and alignment */
2043 assert(aligned_OK(chunk2mem(p
)));
2046 /* ... and was allocated at front of an available chunk */
2047 assert(prev_inuse(p
));
2052 #define check_free_chunk(A,P) do_check_free_chunk(A,P)
2053 #define check_inuse_chunk(A,P) do_check_inuse_chunk(A,P)
2054 #define check_chunk(A,P) do_check_chunk(A,P)
2055 #define check_malloced_chunk(A,P,N) do_check_malloced_chunk(A,P,N)
2057 #define check_free_chunk(A,P)
2058 #define check_inuse_chunk(A,P)
2059 #define check_chunk(A,P)
2060 #define check_malloced_chunk(A,P,N)
2066 Macro-based internal utilities
2071 Linking chunks in bin lists.
2072 Call these only with variables, not arbitrary expressions, as arguments.
2076 Place chunk p of size s in its bin, in size order,
2077 putting it ahead of others of same size.
2081 #define frontlink(A, P, S, IDX, BK, FD) \
2083 if (S < MAX_SMALLBIN_SIZE) \
2085 IDX = smallbin_index(S); \
2086 mark_binblock(A, IDX); \
2087 BK = bin_at(A, IDX); \
2091 FD->bk = BK->fd = P; \
2095 IDX = bin_index(S); \
2096 BK = bin_at(A, IDX); \
2098 if (FD == BK) mark_binblock(A, IDX); \
2101 while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
2106 FD->bk = BK->fd = P; \
2111 /* take a chunk off a list */
2113 #define unlink(P, BK, FD) \
2121 /* Place p as the last remainder */
2123 #define link_last_remainder(A, P) \
2125 last_remainder(A)->fd = last_remainder(A)->bk = P; \
2126 P->fd = P->bk = last_remainder(A); \
2129 /* Clear the last_remainder bin */
2131 #define clear_last_remainder(A) \
2132 (last_remainder(A)->fd = last_remainder(A)->bk = last_remainder(A))
2139 Extend the top-most chunk by obtaining memory from system.
2140 Main interface to sbrk (but see also malloc_trim).
2144 static void malloc_extend_top(arena
*ar_ptr
, INTERNAL_SIZE_T nb
)
2146 static void malloc_extend_top(ar_ptr
, nb
) arena
*ar_ptr
; INTERNAL_SIZE_T nb
;
2149 unsigned long pagesz
= malloc_getpagesize
;
2150 mchunkptr old_top
= top(ar_ptr
); /* Record state of old top */
2151 INTERNAL_SIZE_T old_top_size
= chunksize(old_top
);
2152 INTERNAL_SIZE_T top_size
; /* new size of top chunk */
2155 if(ar_ptr
== &main_arena
) {
2158 char* brk
; /* return value from sbrk */
2159 INTERNAL_SIZE_T front_misalign
; /* unusable bytes at front of sbrked space */
2160 INTERNAL_SIZE_T correction
; /* bytes for 2nd sbrk call */
2161 char* new_brk
; /* return of 2nd sbrk call */
2162 char* old_end
= (char*)(chunk_at_offset(old_top
, old_top_size
));
2164 /* Pad request with top_pad plus minimal overhead */
2165 INTERNAL_SIZE_T sbrk_size
= nb
+ top_pad
+ MINSIZE
;
2167 /* If not the first time through, round to preserve page boundary */
2168 /* Otherwise, we need to correct to a page size below anyway. */
2169 /* (We also correct below if an intervening foreign sbrk call.) */
2171 if (sbrk_base
!= (char*)(-1))
2172 sbrk_size
= (sbrk_size
+ (pagesz
- 1)) & ~(pagesz
- 1);
2174 brk
= (char*)(MORECORE (sbrk_size
));
2176 /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2177 if (brk
== (char*)(MORECORE_FAILURE
) ||
2178 (brk
< old_end
&& old_top
!= initial_top(&main_arena
)))
2181 sbrked_mem
+= sbrk_size
;
2183 if (brk
== old_end
) { /* can just add bytes to current top */
2184 top_size
= sbrk_size
+ old_top_size
;
2185 set_head(old_top
, top_size
| PREV_INUSE
);
2186 old_top
= 0; /* don't free below */
2188 if (sbrk_base
== (char*)(-1)) /* First time through. Record base */
2191 /* Someone else called sbrk(). Count those bytes as sbrked_mem. */
2192 sbrked_mem
+= brk
- (char*)old_end
;
2194 /* Guarantee alignment of first new chunk made from this space */
2195 front_misalign
= (unsigned long)chunk2mem(brk
) & MALLOC_ALIGN_MASK
;
2196 if (front_misalign
> 0) {
2197 correction
= (MALLOC_ALIGNMENT
) - front_misalign
;
2202 /* Guarantee the next brk will be at a page boundary */
2203 correction
+= pagesz
- ((unsigned long)(brk
+ sbrk_size
) & (pagesz
- 1));
2205 /* Allocate correction */
2206 new_brk
= (char*)(MORECORE (correction
));
2207 if (new_brk
== (char*)(MORECORE_FAILURE
)) return;
2209 sbrked_mem
+= correction
;
2211 top(&main_arena
) = (mchunkptr
)brk
;
2212 top_size
= new_brk
- brk
+ correction
;
2213 set_head(top(&main_arena
), top_size
| PREV_INUSE
);
2215 if (old_top
== initial_top(&main_arena
))
2216 old_top
= 0; /* don't free below */
2219 if ((unsigned long)sbrked_mem
> (unsigned long)max_sbrked_mem
)
2220 max_sbrked_mem
= sbrked_mem
;
2222 if ((unsigned long)(mmapped_mem
+ sbrked_mem
) >
2223 (unsigned long)max_total_mem
)
2224 max_total_mem
= mmapped_mem
+ sbrked_mem
;
2228 } else { /* ar_ptr != &main_arena */
2229 heap_info
*old_heap
, *heap
;
2230 size_t old_heap_size
;
2232 if(old_top_size
< MINSIZE
) /* this should never happen */
2235 /* First try to extend the current heap. */
2236 if(MINSIZE
+ nb
<= old_top_size
)
2238 old_heap
= heap_for_ptr(old_top
);
2239 old_heap_size
= old_heap
->size
;
2240 if(grow_heap(old_heap
, MINSIZE
+ nb
- old_top_size
) == 0) {
2241 ar_ptr
->size
+= old_heap
->size
- old_heap_size
;
2242 top_size
= ((char *)old_heap
+ old_heap
->size
) - (char *)old_top
;
2243 set_head(old_top
, top_size
| PREV_INUSE
);
2247 /* A new heap must be created. */
2248 heap
= new_heap(nb
+ top_pad
+ (MINSIZE
+ sizeof(*heap
)));
2251 heap
->ar_ptr
= ar_ptr
;
2252 heap
->prev
= old_heap
;
2253 ar_ptr
->size
+= heap
->size
;
2255 /* Set up the new top, so we can safely use chunk_free() below. */
2256 top(ar_ptr
) = chunk_at_offset(heap
, sizeof(*heap
));
2257 top_size
= heap
->size
- sizeof(*heap
);
2258 set_head(top(ar_ptr
), top_size
| PREV_INUSE
);
2260 #endif /* !defined(NO_THREADS) */
2262 /* We always land on a page boundary */
2263 assert(((unsigned long)((char*)top(ar_ptr
) + top_size
) & (pagesz
-1)) == 0);
2265 /* Setup fencepost and free the old top chunk. */
2267 /* The fencepost takes at least MINSIZE bytes, because it might
2268 become the top chunk again later. Note that a footer is set
2269 up, too, although the chunk is marked in use. */
2270 old_top_size
-= MINSIZE
;
2271 set_head(chunk_at_offset(old_top
, old_top_size
+ 2*SIZE_SZ
), 0|PREV_INUSE
);
2272 if(old_top_size
>= MINSIZE
) {
2273 set_head(chunk_at_offset(old_top
, old_top_size
), (2*SIZE_SZ
)|PREV_INUSE
);
2274 set_foot(chunk_at_offset(old_top
, old_top_size
), (2*SIZE_SZ
));
2275 set_head_size(old_top
, old_top_size
);
2276 chunk_free(ar_ptr
, old_top
);
2278 set_head(old_top
, (old_top_size
+ 2*SIZE_SZ
)|PREV_INUSE
);
2279 set_foot(old_top
, (old_top_size
+ 2*SIZE_SZ
));
2287 /* Main public routines */
2293 The requested size is first converted into a usable form, `nb'.
2294 This currently means to add 4 bytes overhead plus possibly more to
2295 obtain 8-byte alignment and/or to obtain a size of at least
2296 MINSIZE (currently 16, 24, or 32 bytes), the smallest allocatable
2297 size. (All fits are considered `exact' if they are within MINSIZE
2300 From there, the first successful of the following steps is taken:
2302 1. The bin corresponding to the request size is scanned, and if
2303 a chunk of exactly the right size is found, it is taken.
2305 2. The most recently remaindered chunk is used if it is big
2306 enough. This is a form of (roving) first fit, used only in
2307 the absence of exact fits. Runs of consecutive requests use
2308 the remainder of the chunk used for the previous such request
2309 whenever possible. This limited use of a first-fit style
2310 allocation strategy tends to give contiguous chunks
2311 coextensive lifetimes, which improves locality and can reduce
2312 fragmentation in the long run.
2314 3. Other bins are scanned in increasing size order, using a
2315 chunk big enough to fulfill the request, and splitting off
2316 any remainder. This search is strictly by best-fit; i.e.,
2317 the smallest (with ties going to approximately the least
2318 recently used) chunk that fits is selected.
2320 4. If large enough, the chunk bordering the end of memory
2321 (`top') is split off. (This use of `top' is in accord with
2322 the best-fit search rule. In effect, `top' is treated as
2323 larger (and thus less well fitting) than any other available
2324 chunk since it can be extended to be as large as necessary
2325 (up to system limitations).
2327 5. If the request size meets the mmap threshold and the
2328 system supports mmap, and there are few enough currently
2329 allocated mmapped regions, and a call to mmap succeeds,
2330 the request is allocated via direct memory mapping.
2332 6. Otherwise, the top of memory is extended by
2333 obtaining more space from the system (normally using sbrk,
2334 but definable to anything else via the MORECORE macro).
2335 Memory is gathered from the system (in system page-sized
2336 units) in a way that allows chunks obtained across different
2337 sbrk calls to be consolidated, but does not require
2338 contiguous memory. Thus, it should be safe to intersperse
2339 mallocs with other sbrk calls.
2342 All allocations are made from the the `lowest' part of any found
2343 chunk. (The implementation invariant is that prev_inuse is
2344 always true of any allocated chunk; i.e., that each allocated
2345 chunk borders either a previously allocated and still in-use chunk,
2346 or the base of its memory arena.)
2351 Void_t
* mALLOc(size_t bytes
)
2353 Void_t
* mALLOc(bytes
) size_t bytes
;
2357 INTERNAL_SIZE_T nb
; /* padded request size */
2360 #if defined(_LIBC) || defined(MALLOC_HOOKS)
2361 if (__malloc_hook
!= NULL
) {
2364 result
= (*__malloc_hook
)(bytes
);
2369 nb
= request2size(bytes
);
2370 arena_get(ar_ptr
, nb
+ top_pad
);
2373 victim
= chunk_alloc(ar_ptr
, nb
);
2374 (void)mutex_unlock(&ar_ptr
->mutex
);
2375 return victim
? chunk2mem(victim
) : 0;
2380 chunk_alloc(arena
*ar_ptr
, INTERNAL_SIZE_T nb
)
2382 chunk_alloc(ar_ptr
, nb
) arena
*ar_ptr
; INTERNAL_SIZE_T nb
;
2385 mchunkptr victim
; /* inspected/selected chunk */
2386 INTERNAL_SIZE_T victim_size
; /* its size */
2387 int idx
; /* index for bin traversal */
2388 mbinptr bin
; /* associated bin */
2389 mchunkptr remainder
; /* remainder from a split */
2390 long remainder_size
; /* its size */
2391 int remainder_index
; /* its bin index */
2392 unsigned long block
; /* block traverser bit */
2393 int startidx
; /* first bin of a traversed block */
2394 mchunkptr fwd
; /* misc temp for linking */
2395 mchunkptr bck
; /* misc temp for linking */
2396 mbinptr q
; /* misc temp */
2399 /* Check for exact match in a bin */
2401 if (is_small_request(nb
)) /* Faster version for small requests */
2403 idx
= smallbin_index(nb
);
2405 /* No traversal or size check necessary for small bins. */
2407 q
= bin_at(ar_ptr
, idx
);
2410 /* Also scan the next one, since it would have a remainder < MINSIZE */
2418 victim_size
= chunksize(victim
);
2419 unlink(victim
, bck
, fwd
);
2420 set_inuse_bit_at_offset(victim
, victim_size
);
2421 check_malloced_chunk(ar_ptr
, victim
, nb
);
2425 idx
+= 2; /* Set for bin scan below. We've already scanned 2 bins. */
2430 idx
= bin_index(nb
);
2431 bin
= bin_at(ar_ptr
, idx
);
2433 for (victim
= last(bin
); victim
!= bin
; victim
= victim
->bk
)
2435 victim_size
= chunksize(victim
);
2436 remainder_size
= victim_size
- nb
;
2438 if (remainder_size
>= (long)MINSIZE
) /* too big */
2440 --idx
; /* adjust to rescan below after checking last remainder */
2444 else if (remainder_size
>= 0) /* exact fit */
2446 unlink(victim
, bck
, fwd
);
2447 set_inuse_bit_at_offset(victim
, victim_size
);
2448 check_malloced_chunk(ar_ptr
, victim
, nb
);
2457 /* Try to use the last split-off remainder */
2459 if ( (victim
= last_remainder(ar_ptr
)->fd
) != last_remainder(ar_ptr
))
2461 victim_size
= chunksize(victim
);
2462 remainder_size
= victim_size
- nb
;
2464 if (remainder_size
>= (long)MINSIZE
) /* re-split */
2466 remainder
= chunk_at_offset(victim
, nb
);
2467 set_head(victim
, nb
| PREV_INUSE
);
2468 link_last_remainder(ar_ptr
, remainder
);
2469 set_head(remainder
, remainder_size
| PREV_INUSE
);
2470 set_foot(remainder
, remainder_size
);
2471 check_malloced_chunk(ar_ptr
, victim
, nb
);
2475 clear_last_remainder(ar_ptr
);
2477 if (remainder_size
>= 0) /* exhaust */
2479 set_inuse_bit_at_offset(victim
, victim_size
);
2480 check_malloced_chunk(ar_ptr
, victim
, nb
);
2484 /* Else place in bin */
2486 frontlink(ar_ptr
, victim
, victim_size
, remainder_index
, bck
, fwd
);
2490 If there are any possibly nonempty big-enough blocks,
2491 search for best fitting chunk by scanning bins in blockwidth units.
2494 if ( (block
= idx2binblock(idx
)) <= binblocks(ar_ptr
))
2497 /* Get to the first marked block */
2499 if ( (block
& binblocks(ar_ptr
)) == 0)
2501 /* force to an even block boundary */
2502 idx
= (idx
& ~(BINBLOCKWIDTH
- 1)) + BINBLOCKWIDTH
;
2504 while ((block
& binblocks(ar_ptr
)) == 0)
2506 idx
+= BINBLOCKWIDTH
;
2511 /* For each possibly nonempty block ... */
2514 startidx
= idx
; /* (track incomplete blocks) */
2515 q
= bin
= bin_at(ar_ptr
, idx
);
2517 /* For each bin in this block ... */
2520 /* Find and use first big enough chunk ... */
2522 for (victim
= last(bin
); victim
!= bin
; victim
= victim
->bk
)
2524 victim_size
= chunksize(victim
);
2525 remainder_size
= victim_size
- nb
;
2527 if (remainder_size
>= (long)MINSIZE
) /* split */
2529 remainder
= chunk_at_offset(victim
, nb
);
2530 set_head(victim
, nb
| PREV_INUSE
);
2531 unlink(victim
, bck
, fwd
);
2532 link_last_remainder(ar_ptr
, remainder
);
2533 set_head(remainder
, remainder_size
| PREV_INUSE
);
2534 set_foot(remainder
, remainder_size
);
2535 check_malloced_chunk(ar_ptr
, victim
, nb
);
2539 else if (remainder_size
>= 0) /* take */
2541 set_inuse_bit_at_offset(victim
, victim_size
);
2542 unlink(victim
, bck
, fwd
);
2543 check_malloced_chunk(ar_ptr
, victim
, nb
);
2549 bin
= next_bin(bin
);
2551 } while ((++idx
& (BINBLOCKWIDTH
- 1)) != 0);
2553 /* Clear out the block bit. */
2555 do /* Possibly backtrack to try to clear a partial block */
2557 if ((startidx
& (BINBLOCKWIDTH
- 1)) == 0)
2559 binblocks(ar_ptr
) &= ~block
;
2564 } while (first(q
) == q
);
2566 /* Get to the next possibly nonempty block */
2568 if ( (block
<<= 1) <= binblocks(ar_ptr
) && (block
!= 0) )
2570 while ((block
& binblocks(ar_ptr
)) == 0)
2572 idx
+= BINBLOCKWIDTH
;
2582 /* Try to use top chunk */
2584 /* Require that there be a remainder, ensuring top always exists */
2585 if ( (remainder_size
= chunksize(top(ar_ptr
)) - nb
) < (long)MINSIZE
)
2589 /* If big and would otherwise need to extend, try to use mmap instead */
2590 if ((unsigned long)nb
>= (unsigned long)mmap_threshold
&&
2591 (victim
= mmap_chunk(nb
)) != 0)
2596 malloc_extend_top(ar_ptr
, nb
);
2597 if ((remainder_size
= chunksize(top(ar_ptr
)) - nb
) < (long)MINSIZE
)
2598 return 0; /* propagate failure */
2601 victim
= top(ar_ptr
);
2602 set_head(victim
, nb
| PREV_INUSE
);
2603 top(ar_ptr
) = chunk_at_offset(victim
, nb
);
2604 set_head(top(ar_ptr
), remainder_size
| PREV_INUSE
);
2605 check_malloced_chunk(ar_ptr
, victim
, nb
);
2619 1. free(0) has no effect.
2621 2. If the chunk was allocated via mmap, it is released via munmap().
2623 3. If a returned chunk borders the current high end of memory,
2624 it is consolidated into the top, and if the total unused
2625 topmost memory exceeds the trim threshold, malloc_trim is
2628 4. Other chunks are consolidated as they arrive, and
2629 placed in corresponding bins. (This includes the case of
2630 consolidating with the current `last_remainder').
2636 void fREe(Void_t
* mem
)
2638 void fREe(mem
) Void_t
* mem
;
2642 mchunkptr p
; /* chunk corresponding to mem */
2644 #if defined(_LIBC) || defined(MALLOC_HOOKS)
2645 if (__free_hook
!= NULL
) {
2646 (*__free_hook
)(mem
);
2651 if (mem
== 0) /* free(0) has no effect */
2657 if (chunk_is_mmapped(p
)) /* release mmapped memory. */
2664 ar_ptr
= arena_for_ptr(p
);
2666 if(!mutex_trylock(&ar_ptr
->mutex
))
2667 ++(ar_ptr
->stat_lock_direct
);
2669 (void)mutex_lock(&ar_ptr
->mutex
);
2670 ++(ar_ptr
->stat_lock_wait
);
2673 (void)mutex_lock(&ar_ptr
->mutex
);
2675 chunk_free(ar_ptr
, p
);
2676 (void)mutex_unlock(&ar_ptr
->mutex
);
2681 chunk_free(arena
*ar_ptr
, mchunkptr p
)
2683 chunk_free(ar_ptr
, p
) arena
*ar_ptr
; mchunkptr p
;
2686 INTERNAL_SIZE_T hd
= p
->size
; /* its head field */
2687 INTERNAL_SIZE_T sz
; /* its size */
2688 int idx
; /* its bin index */
2689 mchunkptr next
; /* next contiguous chunk */
2690 INTERNAL_SIZE_T nextsz
; /* its size */
2691 INTERNAL_SIZE_T prevsz
; /* size of previous contiguous chunk */
2692 mchunkptr bck
; /* misc temp for linking */
2693 mchunkptr fwd
; /* misc temp for linking */
2694 int islr
; /* track whether merging with last_remainder */
2696 check_inuse_chunk(ar_ptr
, p
);
2698 sz
= hd
& ~PREV_INUSE
;
2699 next
= chunk_at_offset(p
, sz
);
2700 nextsz
= chunksize(next
);
2702 if (next
== top(ar_ptr
)) /* merge with top */
2706 if (!(hd
& PREV_INUSE
)) /* consolidate backward */
2708 prevsz
= p
->prev_size
;
2709 p
= chunk_at_offset(p
, -prevsz
);
2711 unlink(p
, bck
, fwd
);
2714 set_head(p
, sz
| PREV_INUSE
);
2718 if(ar_ptr
== &main_arena
) {
2720 if ((unsigned long)(sz
) >= (unsigned long)trim_threshold
)
2724 heap_info
*heap
= heap_for_ptr(p
);
2726 assert(heap
->ar_ptr
== ar_ptr
);
2728 /* Try to get rid of completely empty heaps, if possible. */
2729 if((unsigned long)(sz
) >= (unsigned long)trim_threshold
||
2730 p
== chunk_at_offset(heap
, sizeof(*heap
)))
2731 heap_trim(heap
, top_pad
);
2737 set_head(next
, nextsz
); /* clear inuse bit */
2741 if (!(hd
& PREV_INUSE
)) /* consolidate backward */
2743 prevsz
= p
->prev_size
;
2744 p
= chunk_at_offset(p
, -prevsz
);
2747 if (p
->fd
== last_remainder(ar_ptr
)) /* keep as last_remainder */
2750 unlink(p
, bck
, fwd
);
2753 if (!(inuse_bit_at_offset(next
, nextsz
))) /* consolidate forward */
2757 if (!islr
&& next
->fd
== last_remainder(ar_ptr
))
2758 /* re-insert last_remainder */
2761 link_last_remainder(ar_ptr
, p
);
2764 unlink(next
, bck
, fwd
);
2767 set_head(p
, sz
| PREV_INUSE
);
2770 frontlink(ar_ptr
, p
, sz
, idx
, bck
, fwd
);
2781 Chunks that were obtained via mmap cannot be extended or shrunk
2782 unless HAVE_MREMAP is defined, in which case mremap is used.
2783 Otherwise, if their reallocation is for additional space, they are
2784 copied. If for less, they are just left alone.
2786 Otherwise, if the reallocation is for additional space, and the
2787 chunk can be extended, it is, else a malloc-copy-free sequence is
2788 taken. There are several different ways that a chunk could be
2789 extended. All are tried:
2791 * Extending forward into following adjacent free chunk.
2792 * Shifting backwards, joining preceding adjacent space
2793 * Both shifting backwards and extending forward.
2794 * Extending into newly sbrked space
2796 Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2797 size argument of zero (re)allocates a minimum-sized chunk.
2799 If the reallocation is for less space, and the new request is for
2800 a `small' (<512 bytes) size, then the newly unused space is lopped
2803 The old unix realloc convention of allowing the last-free'd chunk
2804 to be used as an argument to realloc is no longer supported.
2805 I don't know of any programs still relying on this feature,
2806 and allowing it would also allow too many other incorrect
2807 usages of realloc to be sensible.
2814 Void_t
* rEALLOc(Void_t
* oldmem
, size_t bytes
)
2816 Void_t
* rEALLOc(oldmem
, bytes
) Void_t
* oldmem
; size_t bytes
;
2820 INTERNAL_SIZE_T nb
; /* padded request size */
2822 mchunkptr oldp
; /* chunk corresponding to oldmem */
2823 INTERNAL_SIZE_T oldsize
; /* its size */
2825 mchunkptr newp
; /* chunk to return */
2827 #if defined(_LIBC) || defined(MALLOC_HOOKS)
2828 if (__realloc_hook
!= NULL
) {
2831 result
= (*__realloc_hook
)(oldmem
, bytes
);
2836 #ifdef REALLOC_ZERO_BYTES_FREES
2837 if (bytes
== 0) { fREe(oldmem
); return 0; }
2840 /* realloc of null is supposed to be same as malloc */
2841 if (oldmem
== 0) return mALLOc(bytes
);
2843 oldp
= mem2chunk(oldmem
);
2844 oldsize
= chunksize(oldp
);
2846 nb
= request2size(bytes
);
2849 if (chunk_is_mmapped(oldp
))
2854 newp
= mremap_chunk(oldp
, nb
);
2855 if(newp
) return chunk2mem(newp
);
2857 /* Note the extra SIZE_SZ overhead. */
2858 if(oldsize
- SIZE_SZ
>= nb
) return oldmem
; /* do nothing */
2859 /* Must alloc, copy, free. */
2860 newmem
= mALLOc(bytes
);
2861 if (newmem
== 0) return 0; /* propagate failure */
2862 MALLOC_COPY(newmem
, oldmem
, oldsize
- 2*SIZE_SZ
);
2868 ar_ptr
= arena_for_ptr(oldp
);
2870 if(!mutex_trylock(&ar_ptr
->mutex
))
2871 ++(ar_ptr
->stat_lock_direct
);
2873 (void)mutex_lock(&ar_ptr
->mutex
);
2874 ++(ar_ptr
->stat_lock_wait
);
2877 (void)mutex_lock(&ar_ptr
->mutex
);
2880 /* As in malloc(), remember this arena for the next allocation. */
2881 tsd_setspecific(arena_key
, (Void_t
*)ar_ptr
);
2883 newp
= chunk_realloc(ar_ptr
, oldp
, oldsize
, nb
);
2885 (void)mutex_unlock(&ar_ptr
->mutex
);
2886 return newp
? chunk2mem(newp
) : NULL
;
2891 chunk_realloc(arena
* ar_ptr
, mchunkptr oldp
, INTERNAL_SIZE_T oldsize
,
2894 chunk_realloc(ar_ptr
, oldp
, oldsize
, nb
)
2895 arena
* ar_ptr
; mchunkptr oldp
; INTERNAL_SIZE_T oldsize
, nb
;
2898 mchunkptr newp
= oldp
; /* chunk to return */
2899 INTERNAL_SIZE_T newsize
= oldsize
; /* its size */
2901 mchunkptr next
; /* next contiguous chunk after oldp */
2902 INTERNAL_SIZE_T nextsize
; /* its size */
2904 mchunkptr prev
; /* previous contiguous chunk before oldp */
2905 INTERNAL_SIZE_T prevsize
; /* its size */
2907 mchunkptr remainder
; /* holds split off extra space from newp */
2908 INTERNAL_SIZE_T remainder_size
; /* its size */
2910 mchunkptr bck
; /* misc temp for linking */
2911 mchunkptr fwd
; /* misc temp for linking */
2913 check_inuse_chunk(ar_ptr
, oldp
);
2915 if ((long)(oldsize
) < (long)(nb
))
2918 /* Try expanding forward */
2920 next
= chunk_at_offset(oldp
, oldsize
);
2921 if (next
== top(ar_ptr
) || !inuse(next
))
2923 nextsize
= chunksize(next
);
2925 /* Forward into top only if a remainder */
2926 if (next
== top(ar_ptr
))
2928 if ((long)(nextsize
+ newsize
) >= (long)(nb
+ MINSIZE
))
2930 newsize
+= nextsize
;
2931 top(ar_ptr
) = chunk_at_offset(oldp
, nb
);
2932 set_head(top(ar_ptr
), (newsize
- nb
) | PREV_INUSE
);
2933 set_head_size(oldp
, nb
);
2938 /* Forward into next chunk */
2939 else if (((long)(nextsize
+ newsize
) >= (long)(nb
)))
2941 unlink(next
, bck
, fwd
);
2942 newsize
+= nextsize
;
2952 /* Try shifting backwards. */
2954 if (!prev_inuse(oldp
))
2956 prev
= prev_chunk(oldp
);
2957 prevsize
= chunksize(prev
);
2959 /* try forward + backward first to save a later consolidation */
2964 if (next
== top(ar_ptr
))
2966 if ((long)(nextsize
+ prevsize
+ newsize
) >= (long)(nb
+ MINSIZE
))
2968 unlink(prev
, bck
, fwd
);
2970 newsize
+= prevsize
+ nextsize
;
2971 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
2972 top(ar_ptr
) = chunk_at_offset(newp
, nb
);
2973 set_head(top(ar_ptr
), (newsize
- nb
) | PREV_INUSE
);
2974 set_head_size(newp
, nb
);
2979 /* into next chunk */
2980 else if (((long)(nextsize
+ prevsize
+ newsize
) >= (long)(nb
)))
2982 unlink(next
, bck
, fwd
);
2983 unlink(prev
, bck
, fwd
);
2985 newsize
+= nextsize
+ prevsize
;
2986 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
2992 if (prev
!= 0 && (long)(prevsize
+ newsize
) >= (long)nb
)
2994 unlink(prev
, bck
, fwd
);
2996 newsize
+= prevsize
;
2997 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3004 newp
= chunk_alloc (ar_ptr
, nb
);
3006 if (newp
== 0) /* propagate failure */
3009 /* Avoid copy if newp is next chunk after oldp. */
3010 /* (This can only happen when new chunk is sbrk'ed.) */
3012 if ( newp
== next_chunk(oldp
))
3014 newsize
+= chunksize(newp
);
3019 /* Otherwise copy, free, and exit */
3020 MALLOC_COPY(chunk2mem(newp
), chunk2mem(oldp
), oldsize
- SIZE_SZ
);
3021 chunk_free(ar_ptr
, oldp
);
3026 split
: /* split off extra room in old or expanded chunk */
3028 if (newsize
- nb
>= MINSIZE
) /* split off remainder */
3030 remainder
= chunk_at_offset(newp
, nb
);
3031 remainder_size
= newsize
- nb
;
3032 set_head_size(newp
, nb
);
3033 set_head(remainder
, remainder_size
| PREV_INUSE
);
3034 set_inuse_bit_at_offset(remainder
, remainder_size
);
3035 chunk_free(ar_ptr
, remainder
);
3039 set_head_size(newp
, newsize
);
3040 set_inuse_bit_at_offset(newp
, newsize
);
3043 check_inuse_chunk(ar_ptr
, newp
);
3054 memalign requests more than enough space from malloc, finds a spot
3055 within that chunk that meets the alignment request, and then
3056 possibly frees the leading and trailing space.
3058 The alignment argument must be a power of two. This property is not
3059 checked by memalign, so misuse may result in random runtime errors.
3061 8-byte alignment is guaranteed by normal malloc calls, so don't
3062 bother calling memalign with an argument of 8 or less.
3064 Overreliance on memalign is a sure way to fragment space.
3070 Void_t
* mEMALIGn(size_t alignment
, size_t bytes
)
3072 Void_t
* mEMALIGn(alignment
, bytes
) size_t alignment
; size_t bytes
;
3076 INTERNAL_SIZE_T nb
; /* padded request size */
3079 #if defined(_LIBC) || defined(MALLOC_HOOKS)
3080 if (__memalign_hook
!= NULL
) {
3083 result
= (*__memalign_hook
)(alignment
, bytes
);
3088 /* If need less alignment than we give anyway, just relay to malloc */
3090 if (alignment
<= MALLOC_ALIGNMENT
) return mALLOc(bytes
);
3092 /* Otherwise, ensure that it is at least a minimum chunk size */
3094 if (alignment
< MINSIZE
) alignment
= MINSIZE
;
3096 nb
= request2size(bytes
);
3097 arena_get(ar_ptr
, nb
+ alignment
+ MINSIZE
);
3100 p
= chunk_align(ar_ptr
, nb
, alignment
);
3101 (void)mutex_unlock(&ar_ptr
->mutex
);
3102 return p
? chunk2mem(p
) : NULL
;
3107 chunk_align(arena
* ar_ptr
, INTERNAL_SIZE_T nb
, size_t alignment
)
3109 chunk_align(ar_ptr
, nb
, alignment
)
3110 arena
* ar_ptr
; INTERNAL_SIZE_T nb
; size_t alignment
;
3113 char* m
; /* memory returned by malloc call */
3114 mchunkptr p
; /* corresponding chunk */
3115 char* brk
; /* alignment point within p */
3116 mchunkptr newp
; /* chunk to return */
3117 INTERNAL_SIZE_T newsize
; /* its size */
3118 INTERNAL_SIZE_T leadsize
; /* leading space befor alignment point */
3119 mchunkptr remainder
; /* spare room at end to split off */
3120 long remainder_size
; /* its size */
3122 /* Call chunk_alloc with worst case padding to hit alignment. */
3123 p
= chunk_alloc(ar_ptr
, nb
+ alignment
+ MINSIZE
);
3125 return 0; /* propagate failure */
3129 if ((((unsigned long)(m
)) % alignment
) == 0) /* aligned */
3132 if(chunk_is_mmapped(p
)) {
3133 return p
; /* nothing more to do */
3137 else /* misaligned */
3140 Find an aligned spot inside chunk.
3141 Since we need to give back leading space in a chunk of at
3142 least MINSIZE, if the first calculation places us at
3143 a spot with less than MINSIZE leader, we can move to the
3144 next aligned spot -- we've allocated enough total room so that
3145 this is always possible.
3148 brk
= (char*)mem2chunk(((unsigned long)(m
+ alignment
- 1)) & -alignment
);
3149 if ((long)(brk
- (char*)(p
)) < (long)MINSIZE
) brk
+= alignment
;
3151 newp
= (mchunkptr
)brk
;
3152 leadsize
= brk
- (char*)(p
);
3153 newsize
= chunksize(p
) - leadsize
;
3156 if(chunk_is_mmapped(p
))
3158 newp
->prev_size
= p
->prev_size
+ leadsize
;
3159 set_head(newp
, newsize
|IS_MMAPPED
);
3164 /* give back leader, use the rest */
3166 set_head(newp
, newsize
| PREV_INUSE
);
3167 set_inuse_bit_at_offset(newp
, newsize
);
3168 set_head_size(p
, leadsize
);
3169 chunk_free(ar_ptr
, p
);
3172 assert (newsize
>=nb
&& (((unsigned long)(chunk2mem(p
))) % alignment
) == 0);
3175 /* Also give back spare room at the end */
3177 remainder_size
= chunksize(p
) - nb
;
3179 if (remainder_size
>= (long)MINSIZE
)
3181 remainder
= chunk_at_offset(p
, nb
);
3182 set_head(remainder
, remainder_size
| PREV_INUSE
);
3183 set_head_size(p
, nb
);
3184 chunk_free(ar_ptr
, remainder
);
3187 check_inuse_chunk(ar_ptr
, p
);
3195 valloc just invokes memalign with alignment argument equal
3196 to the page size of the system (or as near to this as can
3197 be figured out from all the includes/defines above.)
3201 Void_t
* vALLOc(size_t bytes
)
3203 Void_t
* vALLOc(bytes
) size_t bytes
;
3206 return mEMALIGn (malloc_getpagesize
, bytes
);
3210 pvalloc just invokes valloc for the nearest pagesize
3211 that will accommodate request
3216 Void_t
* pvALLOc(size_t bytes
)
3218 Void_t
* pvALLOc(bytes
) size_t bytes
;
3221 size_t pagesize
= malloc_getpagesize
;
3222 return mEMALIGn (pagesize
, (bytes
+ pagesize
- 1) & ~(pagesize
- 1));
3227 calloc calls chunk_alloc, then zeroes out the allocated chunk.
3232 Void_t
* cALLOc(size_t n
, size_t elem_size
)
3234 Void_t
* cALLOc(n
, elem_size
) size_t n
; size_t elem_size
;
3238 mchunkptr p
, oldtop
;
3239 INTERNAL_SIZE_T sz
, csz
, oldtopsize
;
3242 #if defined(_LIBC) || defined(MALLOC_HOOKS)
3243 if (__malloc_hook
!= NULL
) {
3245 mem
= (*__malloc_hook
)(sz
);
3249 while(sz
> 0) mem
[--sz
] = 0; /* rather inefficient */
3255 sz
= request2size(n
* elem_size
);
3256 arena_get(ar_ptr
, sz
);
3260 /* check if expand_top called, in which case don't need to clear */
3262 oldtop
= top(ar_ptr
);
3263 oldtopsize
= chunksize(top(ar_ptr
));
3265 p
= chunk_alloc (ar_ptr
, sz
);
3267 /* Only clearing follows, so we can unlock early. */
3268 (void)mutex_unlock(&ar_ptr
->mutex
);
3276 /* Two optional cases in which clearing not necessary */
3279 if (chunk_is_mmapped(p
)) return mem
;
3285 if (p
== oldtop
&& csz
> oldtopsize
)
3287 /* clear only the bytes from non-freshly-sbrked memory */
3292 MALLOC_ZERO(mem
, csz
- SIZE_SZ
);
3299 cfree just calls free. It is needed/defined on some systems
3300 that pair it with calloc, presumably for odd historical reasons.
3306 void cfree(Void_t
*mem
)
3308 void cfree(mem
) Void_t
*mem
;
3319 Malloc_trim gives memory back to the system (via negative
3320 arguments to sbrk) if there is unused memory at the `high' end of
3321 the malloc pool. You can call this after freeing large blocks of
3322 memory to potentially reduce the system-level memory requirements
3323 of a program. However, it cannot guarantee to reduce memory. Under
3324 some allocation patterns, some large free blocks of memory will be
3325 locked between two used chunks, so they cannot be given back to
3328 The `pad' argument to malloc_trim represents the amount of free
3329 trailing space to leave untrimmed. If this argument is zero,
3330 only the minimum amount of memory to maintain internal data
3331 structures will be left (one page or less). Non-zero arguments
3332 can be supplied to maintain enough trailing space to service
3333 future expected allocations without having to re-obtain memory
3336 Malloc_trim returns 1 if it actually released any memory, else 0.
3341 int malloc_trim(size_t pad
)
3343 int malloc_trim(pad
) size_t pad
;
3348 (void)mutex_lock(&main_arena
.mutex
);
3349 res
= main_trim(pad
);
3350 (void)mutex_unlock(&main_arena
.mutex
);
3354 /* Trim the main arena. */
3358 main_trim(size_t pad
)
3360 main_trim(pad
) size_t pad
;
3363 mchunkptr top_chunk
; /* The current top chunk */
3364 long top_size
; /* Amount of top-most memory */
3365 long extra
; /* Amount to release */
3366 char* current_brk
; /* address returned by pre-check sbrk call */
3367 char* new_brk
; /* address returned by negative sbrk call */
3369 unsigned long pagesz
= malloc_getpagesize
;
3371 top_chunk
= top(&main_arena
);
3372 top_size
= chunksize(top_chunk
);
3373 extra
= ((top_size
- pad
- MINSIZE
+ (pagesz
-1)) / pagesz
- 1) * pagesz
;
3375 if (extra
< (long)pagesz
) /* Not enough memory to release */
3378 /* Test to make sure no one else called sbrk */
3379 current_brk
= (char*)(MORECORE (0));
3380 if (current_brk
!= (char*)(top_chunk
) + top_size
)
3381 return 0; /* Apparently we don't own memory; must fail */
3383 new_brk
= (char*)(MORECORE (-extra
));
3385 if (new_brk
== (char*)(MORECORE_FAILURE
)) { /* sbrk failed? */
3386 /* Try to figure out what we have */
3387 current_brk
= (char*)(MORECORE (0));
3388 top_size
= current_brk
- (char*)top_chunk
;
3389 if (top_size
>= (long)MINSIZE
) /* if not, we are very very dead! */
3391 sbrked_mem
= current_brk
- sbrk_base
;
3392 set_head(top_chunk
, top_size
| PREV_INUSE
);
3394 check_chunk(&main_arena
, top_chunk
);
3397 sbrked_mem
-= extra
;
3399 /* Success. Adjust top accordingly. */
3400 set_head(top_chunk
, (top_size
- extra
) | PREV_INUSE
);
3401 check_chunk(&main_arena
, top_chunk
);
3409 heap_trim(heap_info
*heap
, size_t pad
)
3411 heap_trim(heap
, pad
) heap_info
*heap
; size_t pad
;
3414 unsigned long pagesz
= malloc_getpagesize
;
3415 arena
*ar_ptr
= heap
->ar_ptr
;
3416 mchunkptr top_chunk
= top(ar_ptr
), p
, bck
, fwd
;
3417 heap_info
*prev_heap
;
3418 long new_size
, top_size
, extra
;
3420 /* Can this heap go away completely ? */
3421 while(top_chunk
== chunk_at_offset(heap
, sizeof(*heap
))) {
3422 prev_heap
= heap
->prev
;
3423 p
= chunk_at_offset(prev_heap
, prev_heap
->size
- (MINSIZE
-2*SIZE_SZ
));
3424 assert(p
->size
== (0|PREV_INUSE
)); /* must be fencepost */
3426 new_size
= chunksize(p
) + (MINSIZE
-2*SIZE_SZ
);
3427 assert(new_size
>0 && new_size
<(long)(2*MINSIZE
));
3429 new_size
+= p
->prev_size
;
3430 assert(new_size
>0 && new_size
<HEAP_MAX_SIZE
);
3431 if(new_size
+ (HEAP_MAX_SIZE
- prev_heap
->size
) < pad
+ MINSIZE
+ pagesz
)
3433 ar_ptr
->size
-= heap
->size
;
3436 if(!prev_inuse(p
)) { /* consolidate backward */
3438 unlink(p
, bck
, fwd
);
3440 assert(((unsigned long)((char*)p
+ new_size
) & (pagesz
-1)) == 0);
3441 assert( ((char*)p
+ new_size
) == ((char*)heap
+ heap
->size
) );
3442 top(ar_ptr
) = top_chunk
= p
;
3443 set_head(top_chunk
, new_size
| PREV_INUSE
);
3444 check_chunk(ar_ptr
, top_chunk
);
3446 top_size
= chunksize(top_chunk
);
3447 extra
= ((top_size
- pad
- MINSIZE
+ (pagesz
-1))/pagesz
- 1) * pagesz
;
3448 if(extra
< (long)pagesz
)
3450 /* Try to shrink. */
3451 if(grow_heap(heap
, -extra
) != 0)
3453 ar_ptr
->size
-= extra
;
3455 /* Success. Adjust top accordingly. */
3456 set_head(top_chunk
, (top_size
- extra
) | PREV_INUSE
);
3457 check_chunk(ar_ptr
, top_chunk
);
3468 This routine tells you how many bytes you can actually use in an
3469 allocated chunk, which may be more than you requested (although
3470 often not). You can use this many bytes without worrying about
3471 overwriting other allocated objects. Not a particularly great
3472 programming practice, but still sometimes useful.
3477 size_t malloc_usable_size(Void_t
* mem
)
3479 size_t malloc_usable_size(mem
) Void_t
* mem
;
3489 if(!chunk_is_mmapped(p
))
3491 if (!inuse(p
)) return 0;
3492 check_inuse_chunk(arena_for_ptr(mem
), p
);
3493 return chunksize(p
) - SIZE_SZ
;
3495 return chunksize(p
) - 2*SIZE_SZ
;
3502 /* Utility to update mallinfo for malloc_stats() and mallinfo() */
3506 malloc_update_mallinfo(arena
*ar_ptr
, struct mallinfo
*mi
)
3508 malloc_update_mallinfo(ar_ptr
, mi
) arena
*ar_ptr
; struct mallinfo
*mi
;
3517 INTERNAL_SIZE_T avail
;
3519 (void)mutex_lock(&ar_ptr
->mutex
);
3520 avail
= chunksize(top(ar_ptr
));
3521 navail
= ((long)(avail
) >= (long)MINSIZE
)? 1 : 0;
3523 for (i
= 1; i
< NAV
; ++i
)
3525 b
= bin_at(ar_ptr
, i
);
3526 for (p
= last(b
); p
!= b
; p
= p
->bk
)
3529 check_free_chunk(ar_ptr
, p
);
3530 for (q
= next_chunk(p
);
3531 q
!= top(ar_ptr
) && inuse(q
) && (long)chunksize(q
) > 0;
3533 check_inuse_chunk(ar_ptr
, q
);
3535 avail
+= chunksize(p
);
3540 mi
->arena
= ar_ptr
->size
;
3541 mi
->ordblks
= navail
;
3542 mi
->uordblks
= ar_ptr
->size
- avail
;
3543 mi
->fordblks
= avail
;
3544 mi
->hblks
= n_mmaps
;
3545 mi
->hblkhd
= mmapped_mem
;
3546 mi
->keepcost
= chunksize(top(ar_ptr
));
3548 (void)mutex_unlock(&ar_ptr
->mutex
);
3551 #if !defined(NO_THREADS) && MALLOC_DEBUG > 1
3553 /* Print the complete contents of a single heap to stderr. */
3557 dump_heap(heap_info
*heap
)
3559 dump_heap(heap
) heap_info
*heap
;
3565 fprintf(stderr
, "Heap %p, size %10lx:\n", heap
, (long)heap
->size
);
3566 ptr
= (heap
->ar_ptr
!= (arena
*)(heap
+1)) ?
3567 (char*)(heap
+ 1) : (char*)(heap
+ 1) + sizeof(arena
);
3568 p
= (mchunkptr
)(((unsigned long)ptr
+ MALLOC_ALIGN_MASK
) &
3569 ~MALLOC_ALIGN_MASK
);
3571 fprintf(stderr
, "chunk %p size %10lx", p
, (long)p
->size
);
3572 if(p
== top(heap
->ar_ptr
)) {
3573 fprintf(stderr
, " (top)\n");
3575 } else if(p
->size
== (0|PREV_INUSE
)) {
3576 fprintf(stderr
, " (fence)\n");
3579 fprintf(stderr
, "\n");
3592 For all arenas seperately and in total, prints on stderr the
3593 amount of space obtained from the system, and the current number
3594 of bytes allocated via malloc (or realloc, etc) but not yet
3595 freed. (Note that this is the number of bytes allocated, not the
3596 number requested. It will be larger than the number requested
3597 because of alignment and bookkeeping overhead.) When not compiled
3598 for multiple threads, the maximum amount of allocated memory
3599 (which may be more than current if malloc_trim and/or munmap got
3600 called) is also reported. When using mmap(), prints the maximum
3601 number of simultaneous mmap regions used, too.
3610 unsigned int in_use_b
= mmapped_mem
, system_b
= in_use_b
;
3612 long stat_lock_direct
= 0, stat_lock_loop
= 0, stat_lock_wait
= 0;
3615 for(i
=0, ar_ptr
= &main_arena
; ar_ptr
; ar_ptr
= ar_ptr
->next
, i
++) {
3616 malloc_update_mallinfo(ar_ptr
, &mi
);
3617 fprintf(stderr
, "Arena %d:\n", i
);
3618 fprintf(stderr
, "system bytes = %10u\n", (unsigned int)mi
.arena
);
3619 fprintf(stderr
, "in use bytes = %10u\n", (unsigned int)mi
.uordblks
);
3620 system_b
+= mi
.arena
;
3621 in_use_b
+= mi
.uordblks
;
3623 stat_lock_direct
+= ar_ptr
->stat_lock_direct
;
3624 stat_lock_loop
+= ar_ptr
->stat_lock_loop
;
3625 stat_lock_wait
+= ar_ptr
->stat_lock_wait
;
3627 #if !defined(NO_THREADS) && MALLOC_DEBUG > 1
3628 if(ar_ptr
!= &main_arena
) {
3629 heap_info
*heap
= heap_for_ptr(top(ar_ptr
));
3630 while(heap
) { dump_heap(heap
); heap
= heap
->prev
; }
3634 fprintf(stderr
, "Total (incl. mmap):\n");
3635 fprintf(stderr
, "system bytes = %10u\n", system_b
);
3636 fprintf(stderr
, "in use bytes = %10u\n", in_use_b
);
3638 fprintf(stderr
, "max system bytes = %10u\n", (unsigned int)max_total_mem
);
3641 fprintf(stderr
, "max mmap regions = %10u\n", (unsigned int)max_n_mmaps
);
3644 fprintf(stderr
, "heaps created = %10d\n", stat_n_heaps
);
3645 fprintf(stderr
, "locked directly = %10ld\n", stat_lock_direct
);
3646 fprintf(stderr
, "locked in loop = %10ld\n", stat_lock_loop
);
3647 fprintf(stderr
, "locked waiting = %10ld\n", stat_lock_wait
);
3648 fprintf(stderr
, "locked total = %10ld\n",
3649 stat_lock_direct
+ stat_lock_loop
+ stat_lock_wait
);
3654 mallinfo returns a copy of updated current mallinfo.
3655 The information reported is for the arena last used by the thread.
3658 struct mallinfo
mALLINFo()
3661 Void_t
*vptr
= NULL
;
3663 tsd_getspecific(arena_key
, vptr
);
3664 malloc_update_mallinfo((vptr
? (arena
*)vptr
: &main_arena
), &mi
);
3674 mallopt is the general SVID/XPG interface to tunable parameters.
3675 The format is to provide a (parameter-number, parameter-value) pair.
3676 mallopt then sets the corresponding parameter to the argument
3677 value if it can (i.e., so long as the value is meaningful),
3678 and returns 1 if successful else 0.
3680 See descriptions of tunable parameters above.
3685 int mALLOPt(int param_number
, int value
)
3687 int mALLOPt(param_number
, value
) int param_number
; int value
;
3690 switch(param_number
)
3692 case M_TRIM_THRESHOLD
:
3693 trim_threshold
= value
; return 1;
3695 top_pad
= value
; return 1;
3696 case M_MMAP_THRESHOLD
:
3698 /* Forbid setting the threshold too high. */
3699 if((unsigned long)value
> HEAP_MAX_SIZE
/2) return 0;
3701 mmap_threshold
= value
; return 1;
3704 n_mmaps_max
= value
; return 1;
3706 if (value
!= 0) return 0; else n_mmaps_max
= value
; return 1;
3708 case M_CHECK_ACTION
:
3709 check_action
= value
; return 1;
3717 weak_alias (__libc_calloc
, __calloc
) weak_alias (__libc_calloc
, calloc
)
3718 weak_alias (__libc_free
, __cfree
) weak_alias (__libc_free
, cfree
)
3719 weak_alias (__libc_free
, __free
) weak_alias (__libc_free
, free
)
3720 weak_alias (__libc_malloc
, __malloc
) weak_alias (__libc_malloc
, malloc
)
3721 weak_alias (__libc_memalign
, __memalign
) weak_alias (__libc_memalign
, memalign
)
3722 weak_alias (__libc_realloc
, __realloc
) weak_alias (__libc_realloc
, realloc
)
3723 weak_alias (__libc_valloc
, __valloc
) weak_alias (__libc_valloc
, valloc
)
3724 weak_alias (__libc_pvalloc
, __pvalloc
) weak_alias (__libc_pvalloc
, pvalloc
)
3725 weak_alias (__libc_mallinfo
, __mallinfo
) weak_alias (__libc_mallinfo
, mallinfo
)
3726 weak_alias (__libc_mallopt
, __mallopt
) weak_alias (__libc_mallopt
, mallopt
)
3729 weak_alias (__malloc_stats
, malloc_stats
)
3730 #undef malloc_usable_size
3731 weak_alias (__malloc_usable_size
, malloc_usable_size
)
3733 weak_alias (__malloc_trim
, malloc_trim
)
3737 #if defined(_LIBC) || defined(MALLOC_HOOKS)
3739 /* A simple, standard set of debugging hooks. Overhead is `only' one
3740 byte per chunk; still this will catch most cases of double frees or
3743 #define MAGICBYTE ((char)0xd7)
3745 /* Convert a pointer to be free()d or realloc()ed to a valid chunk
3746 pointer. If the provided pointer is not valid, return NULL. The
3747 goal here is to avoid crashes, unlike in the MALLOC_DEBUG code. */
3751 mem2chunk_check(Void_t
* mem
)
3753 mem2chunk_check(mem
) Void_t
* mem
;
3760 if(!aligned_OK(p
)) return NULL
;
3761 if( (char*)p
>=sbrk_base
&& (char*)p
<(sbrk_base
+sbrked_mem
) ) {
3762 /* Must be a chunk in conventional memory. */
3763 if(chunk_is_mmapped(p
) ||
3764 ( (sz
= chunksize(p
)), ((char*)p
+ sz
)>=(sbrk_base
+sbrked_mem
) ) ||
3765 sz
<MINSIZE
|| sz
&MALLOC_ALIGN_MASK
|| !inuse(p
) ) return NULL
;
3766 if(*((char*)p
+ sz
+ (SIZE_SZ
-1)) != MAGICBYTE
) return NULL
;
3767 *((char*)p
+ sz
+ (SIZE_SZ
-1)) = 0;
3769 unsigned long offset
, page_mask
= malloc_getpagesize
-1;
3771 /* mmap()ed chunks have MALLOC_ALIGNMENT or higher power-of two
3772 alignment relative to the beginning of a page. Check this
3774 offset
= (unsigned long)mem
& page_mask
;
3775 if((offset
!=MALLOC_ALIGNMENT
&& offset
!=0 && offset
!=0x10 &&
3776 offset
!=0x20 && offset
!=0x40 && offset
!=0x80 && offset
!=0x100 &&
3777 offset
!=0x200 && offset
!=0x400 && offset
!=0x800 && offset
!=0x1000 &&
3779 !chunk_is_mmapped(p
) || (p
->size
& PREV_INUSE
) ||
3780 ( (((unsigned long)p
- p
->prev_size
) & page_mask
) != 0 ) ||
3781 ( (sz
= chunksize(p
)), ((p
->prev_size
+ sz
) & page_mask
) != 0 ) )
3783 if(*((char*)p
+ sz
- 1) != MAGICBYTE
) return NULL
;
3784 *((char*)p
+ sz
- 1) = 0;
3791 malloc_check(size_t sz
)
3793 malloc_check(sz
) size_t sz
;
3797 INTERNAL_SIZE_T nb
= request2size(sz
+ 1);
3799 (void)mutex_lock(&main_arena
.mutex
);
3800 victim
= chunk_alloc(&main_arena
, nb
);
3801 (void)mutex_unlock(&main_arena
.mutex
);
3802 if(!victim
) return NULL
;
3803 nb
= chunksize(victim
);
3804 if(chunk_is_mmapped(victim
))
3808 *((char*)victim
+ nb
) = MAGICBYTE
;
3809 return chunk2mem(victim
);
3814 free_check(Void_t
* mem
)
3816 free_check(mem
) Void_t
* mem
;
3822 p
= mem2chunk_check(mem
);
3824 switch(check_action
) {
3826 fprintf(stderr
, "free(): invalid pointer %lx!\n", (long)(mem
));
3834 if (chunk_is_mmapped(p
)) {
3839 (void)mutex_lock(&main_arena
.mutex
);
3840 chunk_free(&main_arena
, p
);
3841 (void)mutex_unlock(&main_arena
.mutex
);
3846 realloc_check(Void_t
* oldmem
, size_t bytes
)
3848 realloc_check(oldmem
, bytes
) Void_t
* oldmem
; size_t bytes
;
3851 mchunkptr oldp
, newp
;
3852 INTERNAL_SIZE_T nb
, oldsize
;
3854 if (oldmem
== 0) return malloc_check(bytes
);
3855 oldp
= mem2chunk_check(oldmem
);
3857 switch(check_action
) {
3859 fprintf(stderr
, "realloc(): invalid pointer %lx!\n", (long)(oldmem
));
3864 return malloc_check(bytes
);
3866 oldsize
= chunksize(oldp
);
3868 nb
= request2size(bytes
+1);
3870 (void)mutex_lock(&main_arena
.mutex
);
3872 if (chunk_is_mmapped(oldp
)) {
3874 newp
= mremap_chunk(oldp
, nb
);
3877 /* Note the extra SIZE_SZ overhead. */
3878 if(oldsize
- SIZE_SZ
>= nb
) newp
= oldp
; /* do nothing */
3880 /* Must alloc, copy, free. */
3881 newp
= chunk_alloc(&main_arena
, nb
);
3883 MALLOC_COPY(chunk2mem(newp
), oldmem
, oldsize
- 2*SIZE_SZ
);
3891 #endif /* HAVE_MMAP */
3892 newp
= chunk_realloc(&main_arena
, oldp
, oldsize
, nb
);
3893 (void)mutex_unlock(&main_arena
.mutex
);
3895 if(!newp
) return NULL
;
3896 nb
= chunksize(newp
);
3897 if(chunk_is_mmapped(newp
))
3901 *((char*)newp
+ nb
) = MAGICBYTE
;
3902 return chunk2mem(newp
);
3907 memalign_check(size_t alignment
, size_t bytes
)
3909 memalign_check(alignment
, bytes
) size_t alignment
; size_t bytes
;
3915 if (alignment
<= MALLOC_ALIGNMENT
) return malloc_check(bytes
);
3916 if (alignment
< MINSIZE
) alignment
= MINSIZE
;
3918 nb
= request2size(bytes
+1);
3919 (void)mutex_lock(&main_arena
.mutex
);
3920 p
= chunk_align(&main_arena
, nb
, alignment
);
3921 (void)mutex_unlock(&main_arena
.mutex
);
3924 if(chunk_is_mmapped(p
))
3928 *((char*)p
+ nb
) = MAGICBYTE
;
3929 return chunk2mem(p
);
3932 #endif /* defined(_LIBC) || defined(MALLOC_HOOKS) */
3938 V2.6.4-pt2 Sat Dec 14 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
3939 * Added debugging hooks
3940 * Fixed possible deadlock in realloc() when out of memory
3941 * Don't pollute namespace in glibc: use __getpagesize, __mmap, etc.
3943 V2.6.4-pt Wed Dec 4 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
3944 * Very minor updates from the released 2.6.4 version.
3945 * Trimmed include file down to exported data structures.
3946 * Changes from H.J. Lu for glibc-2.0.
3948 V2.6.3i-pt Sep 16 1996 Wolfram Gloger (wmglo@dent.med.uni-muenchen.de)
3949 * Many changes for multiple threads
3950 * Introduced arenas and heaps
3952 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)
3953 * Added pvalloc, as recommended by H.J. Liu
3954 * Added 64bit pointer support mainly from Wolfram Gloger
3955 * Added anonymously donated WIN32 sbrk emulation
3956 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3957 * malloc_extend_top: fix mask error that caused wastage after
3959 * Add linux mremap support code from HJ Liu
3961 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)
3962 * Integrated most documentation with the code.
3963 * Add support for mmap, with help from
3964 Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3965 * Use last_remainder in more cases.
3966 * Pack bins using idea from colin@nyx10.cs.du.edu
3967 * Use ordered bins instead of best-fit threshhold
3968 * Eliminate block-local decls to simplify tracing and debugging.
3969 * Support another case of realloc via move into top
3970 * Fix error occuring when initial sbrk_base not word-aligned.
3971 * Rely on page size for units instead of SBRK_UNIT to
3972 avoid surprises about sbrk alignment conventions.
3973 * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3974 (raymond@es.ele.tue.nl) for the suggestion.
3975 * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3976 * More precautions for cases where other routines call sbrk,
3977 courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3978 * Added macros etc., allowing use in linux libc from
3979 H.J. Lu (hjl@gnu.ai.mit.edu)
3980 * Inverted this history list
3982 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)
3983 * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3984 * Removed all preallocation code since under current scheme
3985 the work required to undo bad preallocations exceeds
3986 the work saved in good cases for most test programs.
3987 * No longer use return list or unconsolidated bins since
3988 no scheme using them consistently outperforms those that don't
3989 given above changes.
3990 * Use best fit for very large chunks to prevent some worst-cases.
3991 * Added some support for debugging
3993 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)
3994 * Removed footers when chunks are in use. Thanks to
3995 Paul Wilson (wilson@cs.texas.edu) for the suggestion.
3997 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)
3998 * Added malloc_trim, with help from Wolfram Gloger
3999 (wmglo@Dent.MED.Uni-Muenchen.DE).
4001 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)
4003 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)
4004 * realloc: try to expand in both directions
4005 * malloc: swap order of clean-bin strategy;
4006 * realloc: only conditionally expand backwards
4007 * Try not to scavenge used bins
4008 * Use bin counts as a guide to preallocation
4009 * Occasionally bin return list chunks in first scan
4010 * Add a few optimizations from colin@nyx10.cs.du.edu
4012 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)
4013 * faster bin computation & slightly different binning
4014 * merged all consolidations to one part of malloc proper
4015 (eliminating old malloc_find_space & malloc_clean_bin)
4016 * Scan 2 returns chunks (not just 1)
4017 * Propagate failure in realloc if malloc returns 0
4018 * Add stuff to allow compilation on non-ANSI compilers
4019 from kpv@research.att.com
4021 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)
4022 * removed potential for odd address access in prev_chunk
4023 * removed dependency on getpagesize.h
4024 * misc cosmetics and a bit more internal documentation
4025 * anticosmetics: mangled names in macros to evade debugger strangeness
4026 * tested on sparc, hp-700, dec-mips, rs6000
4027 with gcc & native cc (hp, dec only) allowing
4028 Detlefs & Zorn comparison study (in SIGPLAN Notices.)
4030 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)
4031 * Based loosely on libg++-1.2X malloc. (It retains some of the overall
4032 structure of old version, but most details differ.)