build: Use FAIL_EXIT1 () on failure to exec child [BZ #23990]
[glibc.git] / malloc / malloc.c
blobee87ddbbf9c3b8da9f0f6daa84b42b0a6a859fe0
1 /* Malloc implementation for multiple threads without lock contention.
2 Copyright (C) 1996-2020 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Wolfram Gloger <wg@malloc.de>
5 and Doug Lea <dl@cs.oswego.edu>, 2001.
7 The GNU C Library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public License as
9 published by the Free Software Foundation; either version 2.1 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 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public
18 License along with the GNU C Library; see the file COPYING.LIB. If
19 not, see <https://www.gnu.org/licenses/>. */
22 This is a version (aka ptmalloc2) of malloc/free/realloc written by
23 Doug Lea and adapted to multiple threads/arenas by Wolfram Gloger.
25 There have been substantial changes made after the integration into
26 glibc in all parts of the code. Do not look for much commonality
27 with the ptmalloc2 version.
29 * Version ptmalloc2-20011215
30 based on:
31 VERSION 2.7.0 Sun Mar 11 14:14:06 2001 Doug Lea (dl at gee)
33 * Quickstart
35 In order to compile this implementation, a Makefile is provided with
36 the ptmalloc2 distribution, which has pre-defined targets for some
37 popular systems (e.g. "make posix" for Posix threads). All that is
38 typically required with regard to compiler flags is the selection of
39 the thread package via defining one out of USE_PTHREADS, USE_THR or
40 USE_SPROC. Check the thread-m.h file for what effects this has.
41 Many/most systems will additionally require USE_TSD_DATA_HACK to be
42 defined, so this is the default for "make posix".
44 * Why use this malloc?
46 This is not the fastest, most space-conserving, most portable, or
47 most tunable malloc ever written. However it is among the fastest
48 while also being among the most space-conserving, portable and tunable.
49 Consistent balance across these factors results in a good general-purpose
50 allocator for malloc-intensive programs.
52 The main properties of the algorithms are:
53 * For large (>= 512 bytes) requests, it is a pure best-fit allocator,
54 with ties normally decided via FIFO (i.e. least recently used).
55 * For small (<= 64 bytes by default) requests, it is a caching
56 allocator, that maintains pools of quickly recycled chunks.
57 * In between, and for combinations of large and small requests, it does
58 the best it can trying to meet both goals at once.
59 * For very large requests (>= 128KB by default), it relies on system
60 memory mapping facilities, if supported.
62 For a longer but slightly out of date high-level description, see
63 http://gee.cs.oswego.edu/dl/html/malloc.html
65 You may already by default be using a C library containing a malloc
66 that is based on some version of this malloc (for example in
67 linux). You might still want to use the one in this file in order to
68 customize settings or to avoid overheads associated with library
69 versions.
71 * Contents, described in more detail in "description of public routines" below.
73 Standard (ANSI/SVID/...) functions:
74 malloc(size_t n);
75 calloc(size_t n_elements, size_t element_size);
76 free(void* p);
77 realloc(void* p, size_t n);
78 memalign(size_t alignment, size_t n);
79 valloc(size_t n);
80 mallinfo()
81 mallopt(int parameter_number, int parameter_value)
83 Additional functions:
84 independent_calloc(size_t n_elements, size_t size, void* chunks[]);
85 independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]);
86 pvalloc(size_t n);
87 malloc_trim(size_t pad);
88 malloc_usable_size(void* p);
89 malloc_stats();
91 * Vital statistics:
93 Supported pointer representation: 4 or 8 bytes
94 Supported size_t representation: 4 or 8 bytes
95 Note that size_t is allowed to be 4 bytes even if pointers are 8.
96 You can adjust this by defining INTERNAL_SIZE_T
98 Alignment: 2 * sizeof(size_t) (default)
99 (i.e., 8 byte alignment with 4byte size_t). This suffices for
100 nearly all current machines and C compilers. However, you can
101 define MALLOC_ALIGNMENT to be wider than this if necessary.
103 Minimum overhead per allocated chunk: 4 or 8 bytes
104 Each malloced chunk has a hidden word of overhead holding size
105 and status information.
107 Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
108 8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
110 When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
111 ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
112 needed; 4 (8) for a trailing size field and 8 (16) bytes for
113 free list pointers. Thus, the minimum allocatable size is
114 16/24/32 bytes.
116 Even a request for zero bytes (i.e., malloc(0)) returns a
117 pointer to something of the minimum allocatable size.
119 The maximum overhead wastage (i.e., number of extra bytes
120 allocated than were requested in malloc) is less than or equal
121 to the minimum size, except for requests >= mmap_threshold that
122 are serviced via mmap(), where the worst case wastage is 2 *
123 sizeof(size_t) bytes plus the remainder from a system page (the
124 minimal mmap unit); typically 4096 or 8192 bytes.
126 Maximum allocated size: 4-byte size_t: 2^32 minus about two pages
127 8-byte size_t: 2^64 minus about two pages
129 It is assumed that (possibly signed) size_t values suffice to
130 represent chunk sizes. `Possibly signed' is due to the fact
131 that `size_t' may be defined on a system as either a signed or
132 an unsigned type. The ISO C standard says that it must be
133 unsigned, but a few systems are known not to adhere to this.
134 Additionally, even when size_t is unsigned, sbrk (which is by
135 default used to obtain memory from system) accepts signed
136 arguments, and may not be able to handle size_t-wide arguments
137 with negative sign bit. Generally, values that would
138 appear as negative after accounting for overhead and alignment
139 are supported only via mmap(), which does not have this
140 limitation.
142 Requests for sizes outside the allowed range will perform an optional
143 failure action and then return null. (Requests may also
144 also fail because a system is out of memory.)
146 Thread-safety: thread-safe
148 Compliance: I believe it is compliant with the 1997 Single Unix Specification
149 Also SVID/XPG, ANSI C, and probably others as well.
151 * Synopsis of compile-time options:
153 People have reported using previous versions of this malloc on all
154 versions of Unix, sometimes by tweaking some of the defines
155 below. It has been tested most extensively on Solaris and Linux.
156 People also report using it in stand-alone embedded systems.
158 The implementation is in straight, hand-tuned ANSI C. It is not
159 at all modular. (Sorry!) It uses a lot of macros. To be at all
160 usable, this code should be compiled using an optimizing compiler
161 (for example gcc -O3) that can simplify expressions and control
162 paths. (FAQ: some macros import variables as arguments rather than
163 declare locals because people reported that some debuggers
164 otherwise get confused.)
166 OPTION DEFAULT VALUE
168 Compilation Environment options:
170 HAVE_MREMAP 0
172 Changing default word sizes:
174 INTERNAL_SIZE_T size_t
176 Configuration and functionality options:
178 USE_PUBLIC_MALLOC_WRAPPERS NOT defined
179 USE_MALLOC_LOCK NOT defined
180 MALLOC_DEBUG NOT defined
181 REALLOC_ZERO_BYTES_FREES 1
182 TRIM_FASTBINS 0
184 Options for customizing MORECORE:
186 MORECORE sbrk
187 MORECORE_FAILURE -1
188 MORECORE_CONTIGUOUS 1
189 MORECORE_CANNOT_TRIM NOT defined
190 MORECORE_CLEARS 1
191 MMAP_AS_MORECORE_SIZE (1024 * 1024)
193 Tuning options that are also dynamically changeable via mallopt:
195 DEFAULT_MXFAST 64 (for 32bit), 128 (for 64bit)
196 DEFAULT_TRIM_THRESHOLD 128 * 1024
197 DEFAULT_TOP_PAD 0
198 DEFAULT_MMAP_THRESHOLD 128 * 1024
199 DEFAULT_MMAP_MAX 65536
201 There are several other #defined constants and macros that you
202 probably don't want to touch unless you are extending or adapting malloc. */
205 void* is the pointer type that malloc should say it returns
208 #ifndef void
209 #define void void
210 #endif /*void*/
212 #include <stddef.h> /* for size_t */
213 #include <stdlib.h> /* for getenv(), abort() */
214 #include <unistd.h> /* for __libc_enable_secure */
216 #include <atomic.h>
217 #include <_itoa.h>
218 #include <bits/wordsize.h>
219 #include <sys/sysinfo.h>
221 #include <ldsodefs.h>
223 #include <unistd.h>
224 #include <stdio.h> /* needed for malloc_stats */
225 #include <errno.h>
226 #include <assert.h>
228 #include <shlib-compat.h>
230 /* For uintptr_t. */
231 #include <stdint.h>
233 /* For va_arg, va_start, va_end. */
234 #include <stdarg.h>
236 /* For MIN, MAX, powerof2. */
237 #include <sys/param.h>
239 /* For ALIGN_UP et. al. */
240 #include <libc-pointer-arith.h>
242 /* For DIAG_PUSH/POP_NEEDS_COMMENT et al. */
243 #include <libc-diag.h>
245 #include <malloc/malloc-internal.h>
247 /* For SINGLE_THREAD_P. */
248 #include <sysdep-cancel.h>
251 Debugging:
253 Because freed chunks may be overwritten with bookkeeping fields, this
254 malloc will often die when freed memory is overwritten by user
255 programs. This can be very effective (albeit in an annoying way)
256 in helping track down dangling pointers.
258 If you compile with -DMALLOC_DEBUG, a number of assertion checks are
259 enabled that will catch more memory errors. You probably won't be
260 able to make much sense of the actual assertion errors, but they
261 should help you locate incorrectly overwritten memory. The checking
262 is fairly extensive, and will slow down execution
263 noticeably. Calling malloc_stats or mallinfo with MALLOC_DEBUG set
264 will attempt to check every non-mmapped allocated and free chunk in
265 the course of computing the summmaries. (By nature, mmapped regions
266 cannot be checked very much automatically.)
268 Setting MALLOC_DEBUG may also be helpful if you are trying to modify
269 this code. The assertions in the check routines spell out in more
270 detail the assumptions and invariants underlying the algorithms.
272 Setting MALLOC_DEBUG does NOT provide an automated mechanism for
273 checking that all accesses to malloced memory stay within their
274 bounds. However, there are several add-ons and adaptations of this
275 or other mallocs available that do this.
278 #ifndef MALLOC_DEBUG
279 #define MALLOC_DEBUG 0
280 #endif
282 #ifndef NDEBUG
283 # define __assert_fail(assertion, file, line, function) \
284 __malloc_assert(assertion, file, line, function)
286 extern const char *__progname;
288 static void
289 __malloc_assert (const char *assertion, const char *file, unsigned int line,
290 const char *function)
292 (void) __fxprintf (NULL, "%s%s%s:%u: %s%sAssertion `%s' failed.\n",
293 __progname, __progname[0] ? ": " : "",
294 file, line,
295 function ? function : "", function ? ": " : "",
296 assertion);
297 fflush (stderr);
298 abort ();
300 #endif
302 #if USE_TCACHE
303 /* We want 64 entries. This is an arbitrary limit, which tunables can reduce. */
304 # define TCACHE_MAX_BINS 64
305 # define MAX_TCACHE_SIZE tidx2usize (TCACHE_MAX_BINS-1)
307 /* Only used to pre-fill the tunables. */
308 # define tidx2usize(idx) (((size_t) idx) * MALLOC_ALIGNMENT + MINSIZE - SIZE_SZ)
310 /* When "x" is from chunksize(). */
311 # define csize2tidx(x) (((x) - MINSIZE + MALLOC_ALIGNMENT - 1) / MALLOC_ALIGNMENT)
312 /* When "x" is a user-provided size. */
313 # define usize2tidx(x) csize2tidx (request2size (x))
315 /* With rounding and alignment, the bins are...
316 idx 0 bytes 0..24 (64-bit) or 0..12 (32-bit)
317 idx 1 bytes 25..40 or 13..20
318 idx 2 bytes 41..56 or 21..28
319 etc. */
321 /* This is another arbitrary limit, which tunables can change. Each
322 tcache bin will hold at most this number of chunks. */
323 # define TCACHE_FILL_COUNT 7
325 /* Maximum chunks in tcache bins for tunables. This value must fit the range
326 of tcache->counts[] entries, else they may overflow. */
327 # define MAX_TCACHE_COUNT UINT16_MAX
328 #endif
330 /* Safe-Linking:
331 Use randomness from ASLR (mmap_base) to protect single-linked lists
332 of Fast-Bins and TCache. That is, mask the "next" pointers of the
333 lists' chunks, and also perform allocation alignment checks on them.
334 This mechanism reduces the risk of pointer hijacking, as was done with
335 Safe-Unlinking in the double-linked lists of Small-Bins.
336 It assumes a minimum page size of 4096 bytes (12 bits). Systems with
337 larger pages provide less entropy, although the pointer mangling
338 still works. */
339 #define PROTECT_PTR(pos, ptr) \
340 ((__typeof (ptr)) ((((size_t) pos) >> 12) ^ ((size_t) ptr)))
341 #define REVEAL_PTR(ptr) PROTECT_PTR (&ptr, ptr)
344 REALLOC_ZERO_BYTES_FREES should be set if a call to
345 realloc with zero bytes should be the same as a call to free.
346 This is required by the C standard. Otherwise, since this malloc
347 returns a unique pointer for malloc(0), so does realloc(p, 0).
350 #ifndef REALLOC_ZERO_BYTES_FREES
351 #define REALLOC_ZERO_BYTES_FREES 1
352 #endif
355 TRIM_FASTBINS controls whether free() of a very small chunk can
356 immediately lead to trimming. Setting to true (1) can reduce memory
357 footprint, but will almost always slow down programs that use a lot
358 of small chunks.
360 Define this only if you are willing to give up some speed to more
361 aggressively reduce system-level memory footprint when releasing
362 memory in programs that use many small chunks. You can get
363 essentially the same effect by setting MXFAST to 0, but this can
364 lead to even greater slowdowns in programs using many small chunks.
365 TRIM_FASTBINS is an in-between compile-time option, that disables
366 only those chunks bordering topmost memory from being placed in
367 fastbins.
370 #ifndef TRIM_FASTBINS
371 #define TRIM_FASTBINS 0
372 #endif
375 /* Definition for getting more memory from the OS. */
376 #define MORECORE (*__morecore)
377 #define MORECORE_FAILURE 0
378 void * __default_morecore (ptrdiff_t);
379 void *(*__morecore)(ptrdiff_t) = __default_morecore;
382 #include <string.h>
385 MORECORE-related declarations. By default, rely on sbrk
390 MORECORE is the name of the routine to call to obtain more memory
391 from the system. See below for general guidance on writing
392 alternative MORECORE functions, as well as a version for WIN32 and a
393 sample version for pre-OSX macos.
396 #ifndef MORECORE
397 #define MORECORE sbrk
398 #endif
401 MORECORE_FAILURE is the value returned upon failure of MORECORE
402 as well as mmap. Since it cannot be an otherwise valid memory address,
403 and must reflect values of standard sys calls, you probably ought not
404 try to redefine it.
407 #ifndef MORECORE_FAILURE
408 #define MORECORE_FAILURE (-1)
409 #endif
412 If MORECORE_CONTIGUOUS is true, take advantage of fact that
413 consecutive calls to MORECORE with positive arguments always return
414 contiguous increasing addresses. This is true of unix sbrk. Even
415 if not defined, when regions happen to be contiguous, malloc will
416 permit allocations spanning regions obtained from different
417 calls. But defining this when applicable enables some stronger
418 consistency checks and space efficiencies.
421 #ifndef MORECORE_CONTIGUOUS
422 #define MORECORE_CONTIGUOUS 1
423 #endif
426 Define MORECORE_CANNOT_TRIM if your version of MORECORE
427 cannot release space back to the system when given negative
428 arguments. This is generally necessary only if you are using
429 a hand-crafted MORECORE function that cannot handle negative arguments.
432 /* #define MORECORE_CANNOT_TRIM */
434 /* MORECORE_CLEARS (default 1)
435 The degree to which the routine mapped to MORECORE zeroes out
436 memory: never (0), only for newly allocated space (1) or always
437 (2). The distinction between (1) and (2) is necessary because on
438 some systems, if the application first decrements and then
439 increments the break value, the contents of the reallocated space
440 are unspecified.
443 #ifndef MORECORE_CLEARS
444 # define MORECORE_CLEARS 1
445 #endif
449 MMAP_AS_MORECORE_SIZE is the minimum mmap size argument to use if
450 sbrk fails, and mmap is used as a backup. The value must be a
451 multiple of page size. This backup strategy generally applies only
452 when systems have "holes" in address space, so sbrk cannot perform
453 contiguous expansion, but there is still space available on system.
454 On systems for which this is known to be useful (i.e. most linux
455 kernels), this occurs only when programs allocate huge amounts of
456 memory. Between this, and the fact that mmap regions tend to be
457 limited, the size should be large, to avoid too many mmap calls and
458 thus avoid running out of kernel resources. */
460 #ifndef MMAP_AS_MORECORE_SIZE
461 #define MMAP_AS_MORECORE_SIZE (1024 * 1024)
462 #endif
465 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
466 large blocks.
469 #ifndef HAVE_MREMAP
470 #define HAVE_MREMAP 0
471 #endif
473 /* We may need to support __malloc_initialize_hook for backwards
474 compatibility. */
476 #if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_24)
477 # define HAVE_MALLOC_INIT_HOOK 1
478 #else
479 # define HAVE_MALLOC_INIT_HOOK 0
480 #endif
484 This version of malloc supports the standard SVID/XPG mallinfo
485 routine that returns a struct containing usage properties and
486 statistics. It should work on any SVID/XPG compliant system that has
487 a /usr/include/malloc.h defining struct mallinfo. (If you'd like to
488 install such a thing yourself, cut out the preliminary declarations
489 as described above and below and save them in a malloc.h file. But
490 there's no compelling reason to bother to do this.)
492 The main declaration needed is the mallinfo struct that is returned
493 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
494 bunch of fields that are not even meaningful in this version of
495 malloc. These fields are are instead filled by mallinfo() with
496 other numbers that might be of interest.
500 /* ---------- description of public routines ------------ */
503 malloc(size_t n)
504 Returns a pointer to a newly allocated chunk of at least n bytes, or null
505 if no space is available. Additionally, on failure, errno is
506 set to ENOMEM on ANSI C systems.
508 If n is zero, malloc returns a minumum-sized chunk. (The minimum
509 size is 16 bytes on most 32bit systems, and 24 or 32 bytes on 64bit
510 systems.) On most systems, size_t is an unsigned type, so calls
511 with negative arguments are interpreted as requests for huge amounts
512 of space, which will often fail. The maximum supported value of n
513 differs across systems, but is in all cases less than the maximum
514 representable value of a size_t.
516 void* __libc_malloc(size_t);
517 libc_hidden_proto (__libc_malloc)
520 free(void* p)
521 Releases the chunk of memory pointed to by p, that had been previously
522 allocated using malloc or a related routine such as realloc.
523 It has no effect if p is null. It can have arbitrary (i.e., bad!)
524 effects if p has already been freed.
526 Unless disabled (using mallopt), freeing very large spaces will
527 when possible, automatically trigger operations that give
528 back unused memory to the system, thus reducing program footprint.
530 void __libc_free(void*);
531 libc_hidden_proto (__libc_free)
534 calloc(size_t n_elements, size_t element_size);
535 Returns a pointer to n_elements * element_size bytes, with all locations
536 set to zero.
538 void* __libc_calloc(size_t, size_t);
541 realloc(void* p, size_t n)
542 Returns a pointer to a chunk of size n that contains the same data
543 as does chunk p up to the minimum of (n, p's size) bytes, or null
544 if no space is available.
546 The returned pointer may or may not be the same as p. The algorithm
547 prefers extending p when possible, otherwise it employs the
548 equivalent of a malloc-copy-free sequence.
550 If p is null, realloc is equivalent to malloc.
552 If space is not available, realloc returns null, errno is set (if on
553 ANSI) and p is NOT freed.
555 if n is for fewer bytes than already held by p, the newly unused
556 space is lopped off and freed if possible. Unless the #define
557 REALLOC_ZERO_BYTES_FREES is set, realloc with a size argument of
558 zero (re)allocates a minimum-sized chunk.
560 Large chunks that were internally obtained via mmap will always be
561 grown using malloc-copy-free sequences unless the system supports
562 MREMAP (currently only linux).
564 The old unix realloc convention of allowing the last-free'd chunk
565 to be used as an argument to realloc is not supported.
567 void* __libc_realloc(void*, size_t);
568 libc_hidden_proto (__libc_realloc)
571 memalign(size_t alignment, size_t n);
572 Returns a pointer to a newly allocated chunk of n bytes, aligned
573 in accord with the alignment argument.
575 The alignment argument should be a power of two. If the argument is
576 not a power of two, the nearest greater power is used.
577 8-byte alignment is guaranteed by normal malloc calls, so don't
578 bother calling memalign with an argument of 8 or less.
580 Overreliance on memalign is a sure way to fragment space.
582 void* __libc_memalign(size_t, size_t);
583 libc_hidden_proto (__libc_memalign)
586 valloc(size_t n);
587 Equivalent to memalign(pagesize, n), where pagesize is the page
588 size of the system. If the pagesize is unknown, 4096 is used.
590 void* __libc_valloc(size_t);
595 mallopt(int parameter_number, int parameter_value)
596 Sets tunable parameters The format is to provide a
597 (parameter-number, parameter-value) pair. mallopt then sets the
598 corresponding parameter to the argument value if it can (i.e., so
599 long as the value is meaningful), and returns 1 if successful else
600 0. SVID/XPG/ANSI defines four standard param numbers for mallopt,
601 normally defined in malloc.h. Only one of these (M_MXFAST) is used
602 in this malloc. The others (M_NLBLKS, M_GRAIN, M_KEEP) don't apply,
603 so setting them has no effect. But this malloc also supports four
604 other options in mallopt. See below for details. Briefly, supported
605 parameters are as follows (listed defaults are for "typical"
606 configurations).
608 Symbol param # default allowed param values
609 M_MXFAST 1 64 0-80 (0 disables fastbins)
610 M_TRIM_THRESHOLD -1 128*1024 any (-1U disables trimming)
611 M_TOP_PAD -2 0 any
612 M_MMAP_THRESHOLD -3 128*1024 any (or 0 if no MMAP support)
613 M_MMAP_MAX -4 65536 any (0 disables use of mmap)
615 int __libc_mallopt(int, int);
616 libc_hidden_proto (__libc_mallopt)
620 mallinfo()
621 Returns (by copy) a struct containing various summary statistics:
623 arena: current total non-mmapped bytes allocated from system
624 ordblks: the number of free chunks
625 smblks: the number of fastbin blocks (i.e., small chunks that
626 have been freed but not use resused or consolidated)
627 hblks: current number of mmapped regions
628 hblkhd: total bytes held in mmapped regions
629 usmblks: always 0
630 fsmblks: total bytes held in fastbin blocks
631 uordblks: current total allocated space (normal or mmapped)
632 fordblks: total free space
633 keepcost: the maximum number of bytes that could ideally be released
634 back to system via malloc_trim. ("ideally" means that
635 it ignores page restrictions etc.)
637 Because these fields are ints, but internal bookkeeping may
638 be kept as longs, the reported values may wrap around zero and
639 thus be inaccurate.
641 struct mallinfo __libc_mallinfo(void);
645 pvalloc(size_t n);
646 Equivalent to valloc(minimum-page-that-holds(n)), that is,
647 round up n to nearest pagesize.
649 void* __libc_pvalloc(size_t);
652 malloc_trim(size_t pad);
654 If possible, gives memory back to the system (via negative
655 arguments to sbrk) if there is unused memory at the `high' end of
656 the malloc pool. You can call this after freeing large blocks of
657 memory to potentially reduce the system-level memory requirements
658 of a program. However, it cannot guarantee to reduce memory. Under
659 some allocation patterns, some large free blocks of memory will be
660 locked between two used chunks, so they cannot be given back to
661 the system.
663 The `pad' argument to malloc_trim represents the amount of free
664 trailing space to leave untrimmed. If this argument is zero,
665 only the minimum amount of memory to maintain internal data
666 structures will be left (one page or less). Non-zero arguments
667 can be supplied to maintain enough trailing space to service
668 future expected allocations without having to re-obtain memory
669 from the system.
671 Malloc_trim returns 1 if it actually released any memory, else 0.
672 On systems that do not support "negative sbrks", it will always
673 return 0.
675 int __malloc_trim(size_t);
678 malloc_usable_size(void* p);
680 Returns the number of bytes you can actually use in
681 an allocated chunk, which may be more than you requested (although
682 often not) due to alignment and minimum size constraints.
683 You can use this many bytes without worrying about
684 overwriting other allocated objects. This is not a particularly great
685 programming practice. malloc_usable_size can be more useful in
686 debugging and assertions, for example:
688 p = malloc(n);
689 assert(malloc_usable_size(p) >= 256);
692 size_t __malloc_usable_size(void*);
695 malloc_stats();
696 Prints on stderr the amount of space obtained from the system (both
697 via sbrk and mmap), the maximum amount (which may be more than
698 current if malloc_trim and/or munmap got called), and the current
699 number of bytes allocated via malloc (or realloc, etc) but not yet
700 freed. Note that this is the number of bytes allocated, not the
701 number requested. It will be larger than the number requested
702 because of alignment and bookkeeping overhead. Because it includes
703 alignment wastage as being in use, this figure may be greater than
704 zero even when no user-level chunks are allocated.
706 The reported current and maximum system memory can be inaccurate if
707 a program makes other calls to system memory allocation functions
708 (normally sbrk) outside of malloc.
710 malloc_stats prints only the most commonly interesting statistics.
711 More information can be obtained by calling mallinfo.
714 void __malloc_stats(void);
717 posix_memalign(void **memptr, size_t alignment, size_t size);
719 POSIX wrapper like memalign(), checking for validity of size.
721 int __posix_memalign(void **, size_t, size_t);
723 /* mallopt tuning options */
726 M_MXFAST is the maximum request size used for "fastbins", special bins
727 that hold returned chunks without consolidating their spaces. This
728 enables future requests for chunks of the same size to be handled
729 very quickly, but can increase fragmentation, and thus increase the
730 overall memory footprint of a program.
732 This malloc manages fastbins very conservatively yet still
733 efficiently, so fragmentation is rarely a problem for values less
734 than or equal to the default. The maximum supported value of MXFAST
735 is 80. You wouldn't want it any higher than this anyway. Fastbins
736 are designed especially for use with many small structs, objects or
737 strings -- the default handles structs/objects/arrays with sizes up
738 to 8 4byte fields, or small strings representing words, tokens,
739 etc. Using fastbins for larger objects normally worsens
740 fragmentation without improving speed.
742 M_MXFAST is set in REQUEST size units. It is internally used in
743 chunksize units, which adds padding and alignment. You can reduce
744 M_MXFAST to 0 to disable all use of fastbins. This causes the malloc
745 algorithm to be a closer approximation of fifo-best-fit in all cases,
746 not just for larger requests, but will generally cause it to be
747 slower.
751 /* M_MXFAST is a standard SVID/XPG tuning option, usually listed in malloc.h */
752 #ifndef M_MXFAST
753 #define M_MXFAST 1
754 #endif
756 #ifndef DEFAULT_MXFAST
757 #define DEFAULT_MXFAST (64 * SIZE_SZ / 4)
758 #endif
762 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
763 to keep before releasing via malloc_trim in free().
765 Automatic trimming is mainly useful in long-lived programs.
766 Because trimming via sbrk can be slow on some systems, and can
767 sometimes be wasteful (in cases where programs immediately
768 afterward allocate more large chunks) the value should be high
769 enough so that your overall system performance would improve by
770 releasing this much memory.
772 The trim threshold and the mmap control parameters (see below)
773 can be traded off with one another. Trimming and mmapping are
774 two different ways of releasing unused memory back to the
775 system. Between these two, it is often possible to keep
776 system-level demands of a long-lived program down to a bare
777 minimum. For example, in one test suite of sessions measuring
778 the XF86 X server on Linux, using a trim threshold of 128K and a
779 mmap threshold of 192K led to near-minimal long term resource
780 consumption.
782 If you are using this malloc in a long-lived program, it should
783 pay to experiment with these values. As a rough guide, you
784 might set to a value close to the average size of a process
785 (program) running on your system. Releasing this much memory
786 would allow such a process to run in memory. Generally, it's
787 worth it to tune for trimming rather tham memory mapping when a
788 program undergoes phases where several large chunks are
789 allocated and released in ways that can reuse each other's
790 storage, perhaps mixed with phases where there are no such
791 chunks at all. And in well-behaved long-lived programs,
792 controlling release of large blocks via trimming versus mapping
793 is usually faster.
795 However, in most programs, these parameters serve mainly as
796 protection against the system-level effects of carrying around
797 massive amounts of unneeded memory. Since frequent calls to
798 sbrk, mmap, and munmap otherwise degrade performance, the default
799 parameters are set to relatively high values that serve only as
800 safeguards.
802 The trim value It must be greater than page size to have any useful
803 effect. To disable trimming completely, you can set to
804 (unsigned long)(-1)
806 Trim settings interact with fastbin (MXFAST) settings: Unless
807 TRIM_FASTBINS is defined, automatic trimming never takes place upon
808 freeing a chunk with size less than or equal to MXFAST. Trimming is
809 instead delayed until subsequent freeing of larger chunks. However,
810 you can still force an attempted trim by calling malloc_trim.
812 Also, trimming is not generally possible in cases where
813 the main arena is obtained via mmap.
815 Note that the trick some people use of mallocing a huge space and
816 then freeing it at program startup, in an attempt to reserve system
817 memory, doesn't have the intended effect under automatic trimming,
818 since that memory will immediately be returned to the system.
821 #define M_TRIM_THRESHOLD -1
823 #ifndef DEFAULT_TRIM_THRESHOLD
824 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
825 #endif
828 M_TOP_PAD is the amount of extra `padding' space to allocate or
829 retain whenever sbrk is called. It is used in two ways internally:
831 * When sbrk is called to extend the top of the arena to satisfy
832 a new malloc request, this much padding is added to the sbrk
833 request.
835 * When malloc_trim is called automatically from free(),
836 it is used as the `pad' argument.
838 In both cases, the actual amount of padding is rounded
839 so that the end of the arena is always a system page boundary.
841 The main reason for using padding is to avoid calling sbrk so
842 often. Having even a small pad greatly reduces the likelihood
843 that nearly every malloc request during program start-up (or
844 after trimming) will invoke sbrk, which needlessly wastes
845 time.
847 Automatic rounding-up to page-size units is normally sufficient
848 to avoid measurable overhead, so the default is 0. However, in
849 systems where sbrk is relatively slow, it can pay to increase
850 this value, at the expense of carrying around more memory than
851 the program needs.
854 #define M_TOP_PAD -2
856 #ifndef DEFAULT_TOP_PAD
857 #define DEFAULT_TOP_PAD (0)
858 #endif
861 MMAP_THRESHOLD_MAX and _MIN are the bounds on the dynamically
862 adjusted MMAP_THRESHOLD.
865 #ifndef DEFAULT_MMAP_THRESHOLD_MIN
866 #define DEFAULT_MMAP_THRESHOLD_MIN (128 * 1024)
867 #endif
869 #ifndef DEFAULT_MMAP_THRESHOLD_MAX
870 /* For 32-bit platforms we cannot increase the maximum mmap
871 threshold much because it is also the minimum value for the
872 maximum heap size and its alignment. Going above 512k (i.e., 1M
873 for new heaps) wastes too much address space. */
874 # if __WORDSIZE == 32
875 # define DEFAULT_MMAP_THRESHOLD_MAX (512 * 1024)
876 # else
877 # define DEFAULT_MMAP_THRESHOLD_MAX (4 * 1024 * 1024 * sizeof(long))
878 # endif
879 #endif
882 M_MMAP_THRESHOLD is the request size threshold for using mmap()
883 to service a request. Requests of at least this size that cannot
884 be allocated using already-existing space will be serviced via mmap.
885 (If enough normal freed space already exists it is used instead.)
887 Using mmap segregates relatively large chunks of memory so that
888 they can be individually obtained and released from the host
889 system. A request serviced through mmap is never reused by any
890 other request (at least not directly; the system may just so
891 happen to remap successive requests to the same locations).
893 Segregating space in this way has the benefits that:
895 1. Mmapped space can ALWAYS be individually released back
896 to the system, which helps keep the system level memory
897 demands of a long-lived program low.
898 2. Mapped memory can never become `locked' between
899 other chunks, as can happen with normally allocated chunks, which
900 means that even trimming via malloc_trim would not release them.
901 3. On some systems with "holes" in address spaces, mmap can obtain
902 memory that sbrk cannot.
904 However, it has the disadvantages that:
906 1. The space cannot be reclaimed, consolidated, and then
907 used to service later requests, as happens with normal chunks.
908 2. It can lead to more wastage because of mmap page alignment
909 requirements
910 3. It causes malloc performance to be more dependent on host
911 system memory management support routines which may vary in
912 implementation quality and may impose arbitrary
913 limitations. Generally, servicing a request via normal
914 malloc steps is faster than going through a system's mmap.
916 The advantages of mmap nearly always outweigh disadvantages for
917 "large" chunks, but the value of "large" varies across systems. The
918 default is an empirically derived value that works well in most
919 systems.
922 Update in 2006:
923 The above was written in 2001. Since then the world has changed a lot.
924 Memory got bigger. Applications got bigger. The virtual address space
925 layout in 32 bit linux changed.
927 In the new situation, brk() and mmap space is shared and there are no
928 artificial limits on brk size imposed by the kernel. What is more,
929 applications have started using transient allocations larger than the
930 128Kb as was imagined in 2001.
932 The price for mmap is also high now; each time glibc mmaps from the
933 kernel, the kernel is forced to zero out the memory it gives to the
934 application. Zeroing memory is expensive and eats a lot of cache and
935 memory bandwidth. This has nothing to do with the efficiency of the
936 virtual memory system, by doing mmap the kernel just has no choice but
937 to zero.
939 In 2001, the kernel had a maximum size for brk() which was about 800
940 megabytes on 32 bit x86, at that point brk() would hit the first
941 mmaped shared libaries and couldn't expand anymore. With current 2.6
942 kernels, the VA space layout is different and brk() and mmap
943 both can span the entire heap at will.
945 Rather than using a static threshold for the brk/mmap tradeoff,
946 we are now using a simple dynamic one. The goal is still to avoid
947 fragmentation. The old goals we kept are
948 1) try to get the long lived large allocations to use mmap()
949 2) really large allocations should always use mmap()
950 and we're adding now:
951 3) transient allocations should use brk() to avoid forcing the kernel
952 having to zero memory over and over again
954 The implementation works with a sliding threshold, which is by default
955 limited to go between 128Kb and 32Mb (64Mb for 64 bitmachines) and starts
956 out at 128Kb as per the 2001 default.
958 This allows us to satisfy requirement 1) under the assumption that long
959 lived allocations are made early in the process' lifespan, before it has
960 started doing dynamic allocations of the same size (which will
961 increase the threshold).
963 The upperbound on the threshold satisfies requirement 2)
965 The threshold goes up in value when the application frees memory that was
966 allocated with the mmap allocator. The idea is that once the application
967 starts freeing memory of a certain size, it's highly probable that this is
968 a size the application uses for transient allocations. This estimator
969 is there to satisfy the new third requirement.
973 #define M_MMAP_THRESHOLD -3
975 #ifndef DEFAULT_MMAP_THRESHOLD
976 #define DEFAULT_MMAP_THRESHOLD DEFAULT_MMAP_THRESHOLD_MIN
977 #endif
980 M_MMAP_MAX is the maximum number of requests to simultaneously
981 service using mmap. This parameter exists because
982 some systems have a limited number of internal tables for
983 use by mmap, and using more than a few of them may degrade
984 performance.
986 The default is set to a value that serves only as a safeguard.
987 Setting to 0 disables use of mmap for servicing large requests.
990 #define M_MMAP_MAX -4
992 #ifndef DEFAULT_MMAP_MAX
993 #define DEFAULT_MMAP_MAX (65536)
994 #endif
996 #include <malloc.h>
998 #ifndef RETURN_ADDRESS
999 #define RETURN_ADDRESS(X_) (NULL)
1000 #endif
1002 /* Forward declarations. */
1003 struct malloc_chunk;
1004 typedef struct malloc_chunk* mchunkptr;
1006 /* Internal routines. */
1008 static void* _int_malloc(mstate, size_t);
1009 static void _int_free(mstate, mchunkptr, int);
1010 static void* _int_realloc(mstate, mchunkptr, INTERNAL_SIZE_T,
1011 INTERNAL_SIZE_T);
1012 static void* _int_memalign(mstate, size_t, size_t);
1013 static void* _mid_memalign(size_t, size_t, void *);
1015 static void malloc_printerr(const char *str) __attribute__ ((noreturn));
1017 static void* mem2mem_check(void *p, size_t sz);
1018 static void top_check(void);
1019 static void munmap_chunk(mchunkptr p);
1020 #if HAVE_MREMAP
1021 static mchunkptr mremap_chunk(mchunkptr p, size_t new_size);
1022 #endif
1024 static void* malloc_check(size_t sz, const void *caller);
1025 static void free_check(void* mem, const void *caller);
1026 static void* realloc_check(void* oldmem, size_t bytes,
1027 const void *caller);
1028 static void* memalign_check(size_t alignment, size_t bytes,
1029 const void *caller);
1031 /* ------------------ MMAP support ------------------ */
1034 #include <fcntl.h>
1035 #include <sys/mman.h>
1037 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
1038 # define MAP_ANONYMOUS MAP_ANON
1039 #endif
1041 #ifndef MAP_NORESERVE
1042 # define MAP_NORESERVE 0
1043 #endif
1045 #define MMAP(addr, size, prot, flags) \
1046 __mmap((addr), (size), (prot), (flags)|MAP_ANONYMOUS|MAP_PRIVATE, -1, 0)
1050 ----------------------- Chunk representations -----------------------
1055 This struct declaration is misleading (but accurate and necessary).
1056 It declares a "view" into memory allowing access to necessary
1057 fields at known offsets from a given base. See explanation below.
1060 struct malloc_chunk {
1062 INTERNAL_SIZE_T mchunk_prev_size; /* Size of previous chunk (if free). */
1063 INTERNAL_SIZE_T mchunk_size; /* Size in bytes, including overhead. */
1065 struct malloc_chunk* fd; /* double links -- used only if free. */
1066 struct malloc_chunk* bk;
1068 /* Only used for large blocks: pointer to next larger size. */
1069 struct malloc_chunk* fd_nextsize; /* double links -- used only if free. */
1070 struct malloc_chunk* bk_nextsize;
1075 malloc_chunk details:
1077 (The following includes lightly edited explanations by Colin Plumb.)
1079 Chunks of memory are maintained using a `boundary tag' method as
1080 described in e.g., Knuth or Standish. (See the paper by Paul
1081 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1082 survey of such techniques.) Sizes of free chunks are stored both
1083 in the front of each chunk and at the end. This makes
1084 consolidating fragmented chunks into bigger chunks very fast. The
1085 size fields also hold bits representing whether chunks are free or
1086 in use.
1088 An allocated chunk looks like this:
1091 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1092 | Size of previous chunk, if unallocated (P clear) |
1093 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1094 | Size of chunk, in bytes |A|M|P|
1095 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1096 | User data starts here... .
1098 . (malloc_usable_size() bytes) .
1100 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1101 | (size of chunk, but used for application data) |
1102 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1103 | Size of next chunk, in bytes |A|0|1|
1104 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1106 Where "chunk" is the front of the chunk for the purpose of most of
1107 the malloc code, but "mem" is the pointer that is returned to the
1108 user. "Nextchunk" is the beginning of the next contiguous chunk.
1110 Chunks always begin on even word boundaries, so the mem portion
1111 (which is returned to the user) is also on an even word boundary, and
1112 thus at least double-word aligned.
1114 Free chunks are stored in circular doubly-linked lists, and look like this:
1116 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1117 | Size of previous chunk, if unallocated (P clear) |
1118 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1119 `head:' | Size of chunk, in bytes |A|0|P|
1120 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1121 | Forward pointer to next chunk in list |
1122 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1123 | Back pointer to previous chunk in list |
1124 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1125 | Unused space (may be 0 bytes long) .
1128 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1129 `foot:' | Size of chunk, in bytes |
1130 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1131 | Size of next chunk, in bytes |A|0|0|
1132 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1134 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1135 chunk size (which is always a multiple of two words), is an in-use
1136 bit for the *previous* chunk. If that bit is *clear*, then the
1137 word before the current chunk size contains the previous chunk
1138 size, and can be used to find the front of the previous chunk.
1139 The very first chunk allocated always has this bit set,
1140 preventing access to non-existent (or non-owned) memory. If
1141 prev_inuse is set for any given chunk, then you CANNOT determine
1142 the size of the previous chunk, and might even get a memory
1143 addressing fault when trying to do so.
1145 The A (NON_MAIN_ARENA) bit is cleared for chunks on the initial,
1146 main arena, described by the main_arena variable. When additional
1147 threads are spawned, each thread receives its own arena (up to a
1148 configurable limit, after which arenas are reused for multiple
1149 threads), and the chunks in these arenas have the A bit set. To
1150 find the arena for a chunk on such a non-main arena, heap_for_ptr
1151 performs a bit mask operation and indirection through the ar_ptr
1152 member of the per-heap header heap_info (see arena.c).
1154 Note that the `foot' of the current chunk is actually represented
1155 as the prev_size of the NEXT chunk. This makes it easier to
1156 deal with alignments etc but can be very confusing when trying
1157 to extend or adapt this code.
1159 The three exceptions to all this are:
1161 1. The special chunk `top' doesn't bother using the
1162 trailing size field since there is no next contiguous chunk
1163 that would have to index off it. After initialization, `top'
1164 is forced to always exist. If it would become less than
1165 MINSIZE bytes long, it is replenished.
1167 2. Chunks allocated via mmap, which have the second-lowest-order
1168 bit M (IS_MMAPPED) set in their size fields. Because they are
1169 allocated one-by-one, each must contain its own trailing size
1170 field. If the M bit is set, the other bits are ignored
1171 (because mmapped chunks are neither in an arena, nor adjacent
1172 to a freed chunk). The M bit is also used for chunks which
1173 originally came from a dumped heap via malloc_set_state in
1174 hooks.c.
1176 3. Chunks in fastbins are treated as allocated chunks from the
1177 point of view of the chunk allocator. They are consolidated
1178 with their neighbors only in bulk, in malloc_consolidate.
1182 ---------- Size and alignment checks and conversions ----------
1185 /* conversion from malloc headers to user pointers, and back */
1187 #define chunk2mem(p) ((void*)((char*)(p) + 2*SIZE_SZ))
1188 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1190 /* The smallest possible chunk */
1191 #define MIN_CHUNK_SIZE (offsetof(struct malloc_chunk, fd_nextsize))
1193 /* The smallest size we can malloc is an aligned minimal chunk */
1195 #define MINSIZE \
1196 (unsigned long)(((MIN_CHUNK_SIZE+MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK))
1198 /* Check if m has acceptable alignment */
1200 #define aligned_OK(m) (((unsigned long)(m) & MALLOC_ALIGN_MASK) == 0)
1202 #define misaligned_chunk(p) \
1203 ((uintptr_t)(MALLOC_ALIGNMENT == 2 * SIZE_SZ ? (p) : chunk2mem (p)) \
1204 & MALLOC_ALIGN_MASK)
1206 /* pad request bytes into a usable size -- internal version */
1208 #define request2size(req) \
1209 (((req) + SIZE_SZ + MALLOC_ALIGN_MASK < MINSIZE) ? \
1210 MINSIZE : \
1211 ((req) + SIZE_SZ + MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK)
1213 /* Check if REQ overflows when padded and aligned and if the resulting value
1214 is less than PTRDIFF_T. Returns TRUE and the requested size or MINSIZE in
1215 case the value is less than MINSIZE on SZ or false if any of the previous
1216 check fail. */
1217 static inline bool
1218 checked_request2size (size_t req, size_t *sz) __nonnull (1)
1220 if (__glibc_unlikely (req > PTRDIFF_MAX))
1221 return false;
1222 *sz = request2size (req);
1223 return true;
1227 --------------- Physical chunk operations ---------------
1231 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1232 #define PREV_INUSE 0x1
1234 /* extract inuse bit of previous chunk */
1235 #define prev_inuse(p) ((p)->mchunk_size & PREV_INUSE)
1238 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1239 #define IS_MMAPPED 0x2
1241 /* check for mmap()'ed chunk */
1242 #define chunk_is_mmapped(p) ((p)->mchunk_size & IS_MMAPPED)
1245 /* size field is or'ed with NON_MAIN_ARENA if the chunk was obtained
1246 from a non-main arena. This is only set immediately before handing
1247 the chunk to the user, if necessary. */
1248 #define NON_MAIN_ARENA 0x4
1250 /* Check for chunk from main arena. */
1251 #define chunk_main_arena(p) (((p)->mchunk_size & NON_MAIN_ARENA) == 0)
1253 /* Mark a chunk as not being on the main arena. */
1254 #define set_non_main_arena(p) ((p)->mchunk_size |= NON_MAIN_ARENA)
1258 Bits to mask off when extracting size
1260 Note: IS_MMAPPED is intentionally not masked off from size field in
1261 macros for which mmapped chunks should never be seen. This should
1262 cause helpful core dumps to occur if it is tried by accident by
1263 people extending or adapting this malloc.
1265 #define SIZE_BITS (PREV_INUSE | IS_MMAPPED | NON_MAIN_ARENA)
1267 /* Get size, ignoring use bits */
1268 #define chunksize(p) (chunksize_nomask (p) & ~(SIZE_BITS))
1270 /* Like chunksize, but do not mask SIZE_BITS. */
1271 #define chunksize_nomask(p) ((p)->mchunk_size)
1273 /* Ptr to next physical malloc_chunk. */
1274 #define next_chunk(p) ((mchunkptr) (((char *) (p)) + chunksize (p)))
1276 /* Size of the chunk below P. Only valid if !prev_inuse (P). */
1277 #define prev_size(p) ((p)->mchunk_prev_size)
1279 /* Set the size of the chunk below P. Only valid if !prev_inuse (P). */
1280 #define set_prev_size(p, sz) ((p)->mchunk_prev_size = (sz))
1282 /* Ptr to previous physical malloc_chunk. Only valid if !prev_inuse (P). */
1283 #define prev_chunk(p) ((mchunkptr) (((char *) (p)) - prev_size (p)))
1285 /* Treat space at ptr + offset as a chunk */
1286 #define chunk_at_offset(p, s) ((mchunkptr) (((char *) (p)) + (s)))
1288 /* extract p's inuse bit */
1289 #define inuse(p) \
1290 ((((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size) & PREV_INUSE)
1292 /* set/clear chunk as being inuse without otherwise disturbing */
1293 #define set_inuse(p) \
1294 ((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size |= PREV_INUSE
1296 #define clear_inuse(p) \
1297 ((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size &= ~(PREV_INUSE)
1300 /* check/set/clear inuse bits in known places */
1301 #define inuse_bit_at_offset(p, s) \
1302 (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size & PREV_INUSE)
1304 #define set_inuse_bit_at_offset(p, s) \
1305 (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size |= PREV_INUSE)
1307 #define clear_inuse_bit_at_offset(p, s) \
1308 (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size &= ~(PREV_INUSE))
1311 /* Set size at head, without disturbing its use bit */
1312 #define set_head_size(p, s) ((p)->mchunk_size = (((p)->mchunk_size & SIZE_BITS) | (s)))
1314 /* Set size/use field */
1315 #define set_head(p, s) ((p)->mchunk_size = (s))
1317 /* Set size at footer (only when chunk is not in use) */
1318 #define set_foot(p, s) (((mchunkptr) ((char *) (p) + (s)))->mchunk_prev_size = (s))
1321 #pragma GCC poison mchunk_size
1322 #pragma GCC poison mchunk_prev_size
1325 -------------------- Internal data structures --------------------
1327 All internal state is held in an instance of malloc_state defined
1328 below. There are no other static variables, except in two optional
1329 cases:
1330 * If USE_MALLOC_LOCK is defined, the mALLOC_MUTEx declared above.
1331 * If mmap doesn't support MAP_ANONYMOUS, a dummy file descriptor
1332 for mmap.
1334 Beware of lots of tricks that minimize the total bookkeeping space
1335 requirements. The result is a little over 1K bytes (for 4byte
1336 pointers and size_t.)
1340 Bins
1342 An array of bin headers for free chunks. Each bin is doubly
1343 linked. The bins are approximately proportionally (log) spaced.
1344 There are a lot of these bins (128). This may look excessive, but
1345 works very well in practice. Most bins hold sizes that are
1346 unusual as malloc request sizes, but are more usual for fragments
1347 and consolidated sets of chunks, which is what these bins hold, so
1348 they can be found quickly. All procedures maintain the invariant
1349 that no consolidated chunk physically borders another one, so each
1350 chunk in a list is known to be preceeded and followed by either
1351 inuse chunks or the ends of memory.
1353 Chunks in bins are kept in size order, with ties going to the
1354 approximately least recently used chunk. Ordering isn't needed
1355 for the small bins, which all contain the same-sized chunks, but
1356 facilitates best-fit allocation for larger chunks. These lists
1357 are just sequential. Keeping them in order almost never requires
1358 enough traversal to warrant using fancier ordered data
1359 structures.
1361 Chunks of the same size are linked with the most
1362 recently freed at the front, and allocations are taken from the
1363 back. This results in LRU (FIFO) allocation order, which tends
1364 to give each chunk an equal opportunity to be consolidated with
1365 adjacent freed chunks, resulting in larger free chunks and less
1366 fragmentation.
1368 To simplify use in double-linked lists, each bin header acts
1369 as a malloc_chunk. This avoids special-casing for headers.
1370 But to conserve space and improve locality, we allocate
1371 only the fd/bk pointers of bins, and then use repositioning tricks
1372 to treat these as the fields of a malloc_chunk*.
1375 typedef struct malloc_chunk *mbinptr;
1377 /* addressing -- note that bin_at(0) does not exist */
1378 #define bin_at(m, i) \
1379 (mbinptr) (((char *) &((m)->bins[((i) - 1) * 2])) \
1380 - offsetof (struct malloc_chunk, fd))
1382 /* analog of ++bin */
1383 #define next_bin(b) ((mbinptr) ((char *) (b) + (sizeof (mchunkptr) << 1)))
1385 /* Reminders about list directionality within bins */
1386 #define first(b) ((b)->fd)
1387 #define last(b) ((b)->bk)
1390 Indexing
1392 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1393 8 bytes apart. Larger bins are approximately logarithmically spaced:
1395 64 bins of size 8
1396 32 bins of size 64
1397 16 bins of size 512
1398 8 bins of size 4096
1399 4 bins of size 32768
1400 2 bins of size 262144
1401 1 bin of size what's left
1403 There is actually a little bit of slop in the numbers in bin_index
1404 for the sake of speed. This makes no difference elsewhere.
1406 The bins top out around 1MB because we expect to service large
1407 requests via mmap.
1409 Bin 0 does not exist. Bin 1 is the unordered list; if that would be
1410 a valid chunk size the small bins are bumped up one.
1413 #define NBINS 128
1414 #define NSMALLBINS 64
1415 #define SMALLBIN_WIDTH MALLOC_ALIGNMENT
1416 #define SMALLBIN_CORRECTION (MALLOC_ALIGNMENT > 2 * SIZE_SZ)
1417 #define MIN_LARGE_SIZE ((NSMALLBINS - SMALLBIN_CORRECTION) * SMALLBIN_WIDTH)
1419 #define in_smallbin_range(sz) \
1420 ((unsigned long) (sz) < (unsigned long) MIN_LARGE_SIZE)
1422 #define smallbin_index(sz) \
1423 ((SMALLBIN_WIDTH == 16 ? (((unsigned) (sz)) >> 4) : (((unsigned) (sz)) >> 3))\
1424 + SMALLBIN_CORRECTION)
1426 #define largebin_index_32(sz) \
1427 (((((unsigned long) (sz)) >> 6) <= 38) ? 56 + (((unsigned long) (sz)) >> 6) :\
1428 ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\
1429 ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\
1430 ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\
1431 ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\
1432 126)
1434 #define largebin_index_32_big(sz) \
1435 (((((unsigned long) (sz)) >> 6) <= 45) ? 49 + (((unsigned long) (sz)) >> 6) :\
1436 ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\
1437 ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\
1438 ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\
1439 ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\
1440 126)
1442 // XXX It remains to be seen whether it is good to keep the widths of
1443 // XXX the buckets the same or whether it should be scaled by a factor
1444 // XXX of two as well.
1445 #define largebin_index_64(sz) \
1446 (((((unsigned long) (sz)) >> 6) <= 48) ? 48 + (((unsigned long) (sz)) >> 6) :\
1447 ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\
1448 ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\
1449 ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\
1450 ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\
1451 126)
1453 #define largebin_index(sz) \
1454 (SIZE_SZ == 8 ? largebin_index_64 (sz) \
1455 : MALLOC_ALIGNMENT == 16 ? largebin_index_32_big (sz) \
1456 : largebin_index_32 (sz))
1458 #define bin_index(sz) \
1459 ((in_smallbin_range (sz)) ? smallbin_index (sz) : largebin_index (sz))
1461 /* Take a chunk off a bin list. */
1462 static void
1463 unlink_chunk (mstate av, mchunkptr p)
1465 if (chunksize (p) != prev_size (next_chunk (p)))
1466 malloc_printerr ("corrupted size vs. prev_size");
1468 mchunkptr fd = p->fd;
1469 mchunkptr bk = p->bk;
1471 if (__builtin_expect (fd->bk != p || bk->fd != p, 0))
1472 malloc_printerr ("corrupted double-linked list");
1474 fd->bk = bk;
1475 bk->fd = fd;
1476 if (!in_smallbin_range (chunksize_nomask (p)) && p->fd_nextsize != NULL)
1478 if (p->fd_nextsize->bk_nextsize != p
1479 || p->bk_nextsize->fd_nextsize != p)
1480 malloc_printerr ("corrupted double-linked list (not small)");
1482 if (fd->fd_nextsize == NULL)
1484 if (p->fd_nextsize == p)
1485 fd->fd_nextsize = fd->bk_nextsize = fd;
1486 else
1488 fd->fd_nextsize = p->fd_nextsize;
1489 fd->bk_nextsize = p->bk_nextsize;
1490 p->fd_nextsize->bk_nextsize = fd;
1491 p->bk_nextsize->fd_nextsize = fd;
1494 else
1496 p->fd_nextsize->bk_nextsize = p->bk_nextsize;
1497 p->bk_nextsize->fd_nextsize = p->fd_nextsize;
1503 Unsorted chunks
1505 All remainders from chunk splits, as well as all returned chunks,
1506 are first placed in the "unsorted" bin. They are then placed
1507 in regular bins after malloc gives them ONE chance to be used before
1508 binning. So, basically, the unsorted_chunks list acts as a queue,
1509 with chunks being placed on it in free (and malloc_consolidate),
1510 and taken off (to be either used or placed in bins) in malloc.
1512 The NON_MAIN_ARENA flag is never set for unsorted chunks, so it
1513 does not have to be taken into account in size comparisons.
1516 /* The otherwise unindexable 1-bin is used to hold unsorted chunks. */
1517 #define unsorted_chunks(M) (bin_at (M, 1))
1522 The top-most available chunk (i.e., the one bordering the end of
1523 available memory) is treated specially. It is never included in
1524 any bin, is used only if no other chunk is available, and is
1525 released back to the system if it is very large (see
1526 M_TRIM_THRESHOLD). Because top initially
1527 points to its own bin with initial zero size, thus forcing
1528 extension on the first malloc request, we avoid having any special
1529 code in malloc to check whether it even exists yet. But we still
1530 need to do so when getting memory from system, so we make
1531 initial_top treat the bin as a legal but unusable chunk during the
1532 interval between initialization and the first call to
1533 sysmalloc. (This is somewhat delicate, since it relies on
1534 the 2 preceding words to be zero during this interval as well.)
1537 /* Conveniently, the unsorted bin can be used as dummy top on first call */
1538 #define initial_top(M) (unsorted_chunks (M))
1541 Binmap
1543 To help compensate for the large number of bins, a one-level index
1544 structure is used for bin-by-bin searching. `binmap' is a
1545 bitvector recording whether bins are definitely empty so they can
1546 be skipped over during during traversals. The bits are NOT always
1547 cleared as soon as bins are empty, but instead only
1548 when they are noticed to be empty during traversal in malloc.
1551 /* Conservatively use 32 bits per map word, even if on 64bit system */
1552 #define BINMAPSHIFT 5
1553 #define BITSPERMAP (1U << BINMAPSHIFT)
1554 #define BINMAPSIZE (NBINS / BITSPERMAP)
1556 #define idx2block(i) ((i) >> BINMAPSHIFT)
1557 #define idx2bit(i) ((1U << ((i) & ((1U << BINMAPSHIFT) - 1))))
1559 #define mark_bin(m, i) ((m)->binmap[idx2block (i)] |= idx2bit (i))
1560 #define unmark_bin(m, i) ((m)->binmap[idx2block (i)] &= ~(idx2bit (i)))
1561 #define get_binmap(m, i) ((m)->binmap[idx2block (i)] & idx2bit (i))
1564 Fastbins
1566 An array of lists holding recently freed small chunks. Fastbins
1567 are not doubly linked. It is faster to single-link them, and
1568 since chunks are never removed from the middles of these lists,
1569 double linking is not necessary. Also, unlike regular bins, they
1570 are not even processed in FIFO order (they use faster LIFO) since
1571 ordering doesn't much matter in the transient contexts in which
1572 fastbins are normally used.
1574 Chunks in fastbins keep their inuse bit set, so they cannot
1575 be consolidated with other free chunks. malloc_consolidate
1576 releases all chunks in fastbins and consolidates them with
1577 other free chunks.
1580 typedef struct malloc_chunk *mfastbinptr;
1581 #define fastbin(ar_ptr, idx) ((ar_ptr)->fastbinsY[idx])
1583 /* offset 2 to use otherwise unindexable first 2 bins */
1584 #define fastbin_index(sz) \
1585 ((((unsigned int) (sz)) >> (SIZE_SZ == 8 ? 4 : 3)) - 2)
1588 /* The maximum fastbin request size we support */
1589 #define MAX_FAST_SIZE (80 * SIZE_SZ / 4)
1591 #define NFASTBINS (fastbin_index (request2size (MAX_FAST_SIZE)) + 1)
1594 FASTBIN_CONSOLIDATION_THRESHOLD is the size of a chunk in free()
1595 that triggers automatic consolidation of possibly-surrounding
1596 fastbin chunks. This is a heuristic, so the exact value should not
1597 matter too much. It is defined at half the default trim threshold as a
1598 compromise heuristic to only attempt consolidation if it is likely
1599 to lead to trimming. However, it is not dynamically tunable, since
1600 consolidation reduces fragmentation surrounding large chunks even
1601 if trimming is not used.
1604 #define FASTBIN_CONSOLIDATION_THRESHOLD (65536UL)
1607 NONCONTIGUOUS_BIT indicates that MORECORE does not return contiguous
1608 regions. Otherwise, contiguity is exploited in merging together,
1609 when possible, results from consecutive MORECORE calls.
1611 The initial value comes from MORECORE_CONTIGUOUS, but is
1612 changed dynamically if mmap is ever used as an sbrk substitute.
1615 #define NONCONTIGUOUS_BIT (2U)
1617 #define contiguous(M) (((M)->flags & NONCONTIGUOUS_BIT) == 0)
1618 #define noncontiguous(M) (((M)->flags & NONCONTIGUOUS_BIT) != 0)
1619 #define set_noncontiguous(M) ((M)->flags |= NONCONTIGUOUS_BIT)
1620 #define set_contiguous(M) ((M)->flags &= ~NONCONTIGUOUS_BIT)
1622 /* Maximum size of memory handled in fastbins. */
1623 static INTERNAL_SIZE_T global_max_fast;
1626 Set value of max_fast.
1627 Use impossibly small value if 0.
1628 Precondition: there are no existing fastbin chunks in the main arena.
1629 Since do_check_malloc_state () checks this, we call malloc_consolidate ()
1630 before changing max_fast. Note other arenas will leak their fast bin
1631 entries if max_fast is reduced.
1634 #define set_max_fast(s) \
1635 global_max_fast = (((size_t) (s) <= MALLOC_ALIGN_MASK - SIZE_SZ) \
1636 ? MIN_CHUNK_SIZE / 2 : ((s + SIZE_SZ) & ~MALLOC_ALIGN_MASK))
1638 static inline INTERNAL_SIZE_T
1639 get_max_fast (void)
1641 /* Tell the GCC optimizers that global_max_fast is never larger
1642 than MAX_FAST_SIZE. This avoids out-of-bounds array accesses in
1643 _int_malloc after constant propagation of the size parameter.
1644 (The code never executes because malloc preserves the
1645 global_max_fast invariant, but the optimizers may not recognize
1646 this.) */
1647 if (global_max_fast > MAX_FAST_SIZE)
1648 __builtin_unreachable ();
1649 return global_max_fast;
1653 ----------- Internal state representation and initialization -----------
1657 have_fastchunks indicates that there are probably some fastbin chunks.
1658 It is set true on entering a chunk into any fastbin, and cleared early in
1659 malloc_consolidate. The value is approximate since it may be set when there
1660 are no fastbin chunks, or it may be clear even if there are fastbin chunks
1661 available. Given it's sole purpose is to reduce number of redundant calls to
1662 malloc_consolidate, it does not affect correctness. As a result we can safely
1663 use relaxed atomic accesses.
1667 struct malloc_state
1669 /* Serialize access. */
1670 __libc_lock_define (, mutex);
1672 /* Flags (formerly in max_fast). */
1673 int flags;
1675 /* Set if the fastbin chunks contain recently inserted free blocks. */
1676 /* Note this is a bool but not all targets support atomics on booleans. */
1677 int have_fastchunks;
1679 /* Fastbins */
1680 mfastbinptr fastbinsY[NFASTBINS];
1682 /* Base of the topmost chunk -- not otherwise kept in a bin */
1683 mchunkptr top;
1685 /* The remainder from the most recent split of a small request */
1686 mchunkptr last_remainder;
1688 /* Normal bins packed as described above */
1689 mchunkptr bins[NBINS * 2 - 2];
1691 /* Bitmap of bins */
1692 unsigned int binmap[BINMAPSIZE];
1694 /* Linked list */
1695 struct malloc_state *next;
1697 /* Linked list for free arenas. Access to this field is serialized
1698 by free_list_lock in arena.c. */
1699 struct malloc_state *next_free;
1701 /* Number of threads attached to this arena. 0 if the arena is on
1702 the free list. Access to this field is serialized by
1703 free_list_lock in arena.c. */
1704 INTERNAL_SIZE_T attached_threads;
1706 /* Memory allocated from the system in this arena. */
1707 INTERNAL_SIZE_T system_mem;
1708 INTERNAL_SIZE_T max_system_mem;
1711 struct malloc_par
1713 /* Tunable parameters */
1714 unsigned long trim_threshold;
1715 INTERNAL_SIZE_T top_pad;
1716 INTERNAL_SIZE_T mmap_threshold;
1717 INTERNAL_SIZE_T arena_test;
1718 INTERNAL_SIZE_T arena_max;
1720 /* Memory map support */
1721 int n_mmaps;
1722 int n_mmaps_max;
1723 int max_n_mmaps;
1724 /* the mmap_threshold is dynamic, until the user sets
1725 it manually, at which point we need to disable any
1726 dynamic behavior. */
1727 int no_dyn_threshold;
1729 /* Statistics */
1730 INTERNAL_SIZE_T mmapped_mem;
1731 INTERNAL_SIZE_T max_mmapped_mem;
1733 /* First address handed out by MORECORE/sbrk. */
1734 char *sbrk_base;
1736 #if USE_TCACHE
1737 /* Maximum number of buckets to use. */
1738 size_t tcache_bins;
1739 size_t tcache_max_bytes;
1740 /* Maximum number of chunks in each bucket. */
1741 size_t tcache_count;
1742 /* Maximum number of chunks to remove from the unsorted list, which
1743 aren't used to prefill the cache. */
1744 size_t tcache_unsorted_limit;
1745 #endif
1748 /* There are several instances of this struct ("arenas") in this
1749 malloc. If you are adapting this malloc in a way that does NOT use
1750 a static or mmapped malloc_state, you MUST explicitly zero-fill it
1751 before using. This malloc relies on the property that malloc_state
1752 is initialized to all zeroes (as is true of C statics). */
1754 static struct malloc_state main_arena =
1756 .mutex = _LIBC_LOCK_INITIALIZER,
1757 .next = &main_arena,
1758 .attached_threads = 1
1761 /* These variables are used for undumping support. Chunked are marked
1762 as using mmap, but we leave them alone if they fall into this
1763 range. NB: The chunk size for these chunks only includes the
1764 initial size field (of SIZE_SZ bytes), there is no trailing size
1765 field (unlike with regular mmapped chunks). */
1766 static mchunkptr dumped_main_arena_start; /* Inclusive. */
1767 static mchunkptr dumped_main_arena_end; /* Exclusive. */
1769 /* True if the pointer falls into the dumped arena. Use this after
1770 chunk_is_mmapped indicates a chunk is mmapped. */
1771 #define DUMPED_MAIN_ARENA_CHUNK(p) \
1772 ((p) >= dumped_main_arena_start && (p) < dumped_main_arena_end)
1774 /* There is only one instance of the malloc parameters. */
1776 static struct malloc_par mp_ =
1778 .top_pad = DEFAULT_TOP_PAD,
1779 .n_mmaps_max = DEFAULT_MMAP_MAX,
1780 .mmap_threshold = DEFAULT_MMAP_THRESHOLD,
1781 .trim_threshold = DEFAULT_TRIM_THRESHOLD,
1782 #define NARENAS_FROM_NCORES(n) ((n) * (sizeof (long) == 4 ? 2 : 8))
1783 .arena_test = NARENAS_FROM_NCORES (1)
1784 #if USE_TCACHE
1786 .tcache_count = TCACHE_FILL_COUNT,
1787 .tcache_bins = TCACHE_MAX_BINS,
1788 .tcache_max_bytes = tidx2usize (TCACHE_MAX_BINS-1),
1789 .tcache_unsorted_limit = 0 /* No limit. */
1790 #endif
1794 Initialize a malloc_state struct.
1796 This is called from ptmalloc_init () or from _int_new_arena ()
1797 when creating a new arena.
1800 static void
1801 malloc_init_state (mstate av)
1803 int i;
1804 mbinptr bin;
1806 /* Establish circular links for normal bins */
1807 for (i = 1; i < NBINS; ++i)
1809 bin = bin_at (av, i);
1810 bin->fd = bin->bk = bin;
1813 #if MORECORE_CONTIGUOUS
1814 if (av != &main_arena)
1815 #endif
1816 set_noncontiguous (av);
1817 if (av == &main_arena)
1818 set_max_fast (DEFAULT_MXFAST);
1819 atomic_store_relaxed (&av->have_fastchunks, false);
1821 av->top = initial_top (av);
1825 Other internal utilities operating on mstates
1828 static void *sysmalloc (INTERNAL_SIZE_T, mstate);
1829 static int systrim (size_t, mstate);
1830 static void malloc_consolidate (mstate);
1833 /* -------------- Early definitions for debugging hooks ---------------- */
1835 /* Define and initialize the hook variables. These weak definitions must
1836 appear before any use of the variables in a function (arena.c uses one). */
1837 #ifndef weak_variable
1838 /* In GNU libc we want the hook variables to be weak definitions to
1839 avoid a problem with Emacs. */
1840 # define weak_variable weak_function
1841 #endif
1843 /* Forward declarations. */
1844 static void *malloc_hook_ini (size_t sz,
1845 const void *caller) __THROW;
1846 static void *realloc_hook_ini (void *ptr, size_t sz,
1847 const void *caller) __THROW;
1848 static void *memalign_hook_ini (size_t alignment, size_t sz,
1849 const void *caller) __THROW;
1851 #if HAVE_MALLOC_INIT_HOOK
1852 void weak_variable (*__malloc_initialize_hook) (void) = NULL;
1853 compat_symbol (libc, __malloc_initialize_hook,
1854 __malloc_initialize_hook, GLIBC_2_0);
1855 #endif
1857 void weak_variable (*__free_hook) (void *__ptr,
1858 const void *) = NULL;
1859 void *weak_variable (*__malloc_hook)
1860 (size_t __size, const void *) = malloc_hook_ini;
1861 void *weak_variable (*__realloc_hook)
1862 (void *__ptr, size_t __size, const void *)
1863 = realloc_hook_ini;
1864 void *weak_variable (*__memalign_hook)
1865 (size_t __alignment, size_t __size, const void *)
1866 = memalign_hook_ini;
1867 void weak_variable (*__after_morecore_hook) (void) = NULL;
1869 /* This function is called from the arena shutdown hook, to free the
1870 thread cache (if it exists). */
1871 static void tcache_thread_shutdown (void);
1873 /* ------------------ Testing support ----------------------------------*/
1875 static int perturb_byte;
1877 static void
1878 alloc_perturb (char *p, size_t n)
1880 if (__glibc_unlikely (perturb_byte))
1881 memset (p, perturb_byte ^ 0xff, n);
1884 static void
1885 free_perturb (char *p, size_t n)
1887 if (__glibc_unlikely (perturb_byte))
1888 memset (p, perturb_byte, n);
1893 #include <stap-probe.h>
1895 /* ------------------- Support for multiple arenas -------------------- */
1896 #include "arena.c"
1899 Debugging support
1901 These routines make a number of assertions about the states
1902 of data structures that should be true at all times. If any
1903 are not true, it's very likely that a user program has somehow
1904 trashed memory. (It's also possible that there is a coding error
1905 in malloc. In which case, please report it!)
1908 #if !MALLOC_DEBUG
1910 # define check_chunk(A, P)
1911 # define check_free_chunk(A, P)
1912 # define check_inuse_chunk(A, P)
1913 # define check_remalloced_chunk(A, P, N)
1914 # define check_malloced_chunk(A, P, N)
1915 # define check_malloc_state(A)
1917 #else
1919 # define check_chunk(A, P) do_check_chunk (A, P)
1920 # define check_free_chunk(A, P) do_check_free_chunk (A, P)
1921 # define check_inuse_chunk(A, P) do_check_inuse_chunk (A, P)
1922 # define check_remalloced_chunk(A, P, N) do_check_remalloced_chunk (A, P, N)
1923 # define check_malloced_chunk(A, P, N) do_check_malloced_chunk (A, P, N)
1924 # define check_malloc_state(A) do_check_malloc_state (A)
1927 Properties of all chunks
1930 static void
1931 do_check_chunk (mstate av, mchunkptr p)
1933 unsigned long sz = chunksize (p);
1934 /* min and max possible addresses assuming contiguous allocation */
1935 char *max_address = (char *) (av->top) + chunksize (av->top);
1936 char *min_address = max_address - av->system_mem;
1938 if (!chunk_is_mmapped (p))
1940 /* Has legal address ... */
1941 if (p != av->top)
1943 if (contiguous (av))
1945 assert (((char *) p) >= min_address);
1946 assert (((char *) p + sz) <= ((char *) (av->top)));
1949 else
1951 /* top size is always at least MINSIZE */
1952 assert ((unsigned long) (sz) >= MINSIZE);
1953 /* top predecessor always marked inuse */
1954 assert (prev_inuse (p));
1957 else if (!DUMPED_MAIN_ARENA_CHUNK (p))
1959 /* address is outside main heap */
1960 if (contiguous (av) && av->top != initial_top (av))
1962 assert (((char *) p) < min_address || ((char *) p) >= max_address);
1964 /* chunk is page-aligned */
1965 assert (((prev_size (p) + sz) & (GLRO (dl_pagesize) - 1)) == 0);
1966 /* mem is aligned */
1967 assert (aligned_OK (chunk2mem (p)));
1972 Properties of free chunks
1975 static void
1976 do_check_free_chunk (mstate av, mchunkptr p)
1978 INTERNAL_SIZE_T sz = chunksize_nomask (p) & ~(PREV_INUSE | NON_MAIN_ARENA);
1979 mchunkptr next = chunk_at_offset (p, sz);
1981 do_check_chunk (av, p);
1983 /* Chunk must claim to be free ... */
1984 assert (!inuse (p));
1985 assert (!chunk_is_mmapped (p));
1987 /* Unless a special marker, must have OK fields */
1988 if ((unsigned long) (sz) >= MINSIZE)
1990 assert ((sz & MALLOC_ALIGN_MASK) == 0);
1991 assert (aligned_OK (chunk2mem (p)));
1992 /* ... matching footer field */
1993 assert (prev_size (next_chunk (p)) == sz);
1994 /* ... and is fully consolidated */
1995 assert (prev_inuse (p));
1996 assert (next == av->top || inuse (next));
1998 /* ... and has minimally sane links */
1999 assert (p->fd->bk == p);
2000 assert (p->bk->fd == p);
2002 else /* markers are always of size SIZE_SZ */
2003 assert (sz == SIZE_SZ);
2007 Properties of inuse chunks
2010 static void
2011 do_check_inuse_chunk (mstate av, mchunkptr p)
2013 mchunkptr next;
2015 do_check_chunk (av, p);
2017 if (chunk_is_mmapped (p))
2018 return; /* mmapped chunks have no next/prev */
2020 /* Check whether it claims to be in use ... */
2021 assert (inuse (p));
2023 next = next_chunk (p);
2025 /* ... and is surrounded by OK chunks.
2026 Since more things can be checked with free chunks than inuse ones,
2027 if an inuse chunk borders them and debug is on, it's worth doing them.
2029 if (!prev_inuse (p))
2031 /* Note that we cannot even look at prev unless it is not inuse */
2032 mchunkptr prv = prev_chunk (p);
2033 assert (next_chunk (prv) == p);
2034 do_check_free_chunk (av, prv);
2037 if (next == av->top)
2039 assert (prev_inuse (next));
2040 assert (chunksize (next) >= MINSIZE);
2042 else if (!inuse (next))
2043 do_check_free_chunk (av, next);
2047 Properties of chunks recycled from fastbins
2050 static void
2051 do_check_remalloced_chunk (mstate av, mchunkptr p, INTERNAL_SIZE_T s)
2053 INTERNAL_SIZE_T sz = chunksize_nomask (p) & ~(PREV_INUSE | NON_MAIN_ARENA);
2055 if (!chunk_is_mmapped (p))
2057 assert (av == arena_for_chunk (p));
2058 if (chunk_main_arena (p))
2059 assert (av == &main_arena);
2060 else
2061 assert (av != &main_arena);
2064 do_check_inuse_chunk (av, p);
2066 /* Legal size ... */
2067 assert ((sz & MALLOC_ALIGN_MASK) == 0);
2068 assert ((unsigned long) (sz) >= MINSIZE);
2069 /* ... and alignment */
2070 assert (aligned_OK (chunk2mem (p)));
2071 /* chunk is less than MINSIZE more than request */
2072 assert ((long) (sz) - (long) (s) >= 0);
2073 assert ((long) (sz) - (long) (s + MINSIZE) < 0);
2077 Properties of nonrecycled chunks at the point they are malloced
2080 static void
2081 do_check_malloced_chunk (mstate av, mchunkptr p, INTERNAL_SIZE_T s)
2083 /* same as recycled case ... */
2084 do_check_remalloced_chunk (av, p, s);
2087 ... plus, must obey implementation invariant that prev_inuse is
2088 always true of any allocated chunk; i.e., that each allocated
2089 chunk borders either a previously allocated and still in-use
2090 chunk, or the base of its memory arena. This is ensured
2091 by making all allocations from the `lowest' part of any found
2092 chunk. This does not necessarily hold however for chunks
2093 recycled via fastbins.
2096 assert (prev_inuse (p));
2101 Properties of malloc_state.
2103 This may be useful for debugging malloc, as well as detecting user
2104 programmer errors that somehow write into malloc_state.
2106 If you are extending or experimenting with this malloc, you can
2107 probably figure out how to hack this routine to print out or
2108 display chunk addresses, sizes, bins, and other instrumentation.
2111 static void
2112 do_check_malloc_state (mstate av)
2114 int i;
2115 mchunkptr p;
2116 mchunkptr q;
2117 mbinptr b;
2118 unsigned int idx;
2119 INTERNAL_SIZE_T size;
2120 unsigned long total = 0;
2121 int max_fast_bin;
2123 /* internal size_t must be no wider than pointer type */
2124 assert (sizeof (INTERNAL_SIZE_T) <= sizeof (char *));
2126 /* alignment is a power of 2 */
2127 assert ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT - 1)) == 0);
2129 /* Check the arena is initialized. */
2130 assert (av->top != 0);
2132 /* No memory has been allocated yet, so doing more tests is not possible. */
2133 if (av->top == initial_top (av))
2134 return;
2136 /* pagesize is a power of 2 */
2137 assert (powerof2(GLRO (dl_pagesize)));
2139 /* A contiguous main_arena is consistent with sbrk_base. */
2140 if (av == &main_arena && contiguous (av))
2141 assert ((char *) mp_.sbrk_base + av->system_mem ==
2142 (char *) av->top + chunksize (av->top));
2144 /* properties of fastbins */
2146 /* max_fast is in allowed range */
2147 assert ((get_max_fast () & ~1) <= request2size (MAX_FAST_SIZE));
2149 max_fast_bin = fastbin_index (get_max_fast ());
2151 for (i = 0; i < NFASTBINS; ++i)
2153 p = fastbin (av, i);
2155 /* The following test can only be performed for the main arena.
2156 While mallopt calls malloc_consolidate to get rid of all fast
2157 bins (especially those larger than the new maximum) this does
2158 only happen for the main arena. Trying to do this for any
2159 other arena would mean those arenas have to be locked and
2160 malloc_consolidate be called for them. This is excessive. And
2161 even if this is acceptable to somebody it still cannot solve
2162 the problem completely since if the arena is locked a
2163 concurrent malloc call might create a new arena which then
2164 could use the newly invalid fast bins. */
2166 /* all bins past max_fast are empty */
2167 if (av == &main_arena && i > max_fast_bin)
2168 assert (p == 0);
2170 while (p != 0)
2172 if (__glibc_unlikely (misaligned_chunk (p)))
2173 malloc_printerr ("do_check_malloc_state(): "
2174 "unaligned fastbin chunk detected");
2175 /* each chunk claims to be inuse */
2176 do_check_inuse_chunk (av, p);
2177 total += chunksize (p);
2178 /* chunk belongs in this bin */
2179 assert (fastbin_index (chunksize (p)) == i);
2180 p = REVEAL_PTR (p->fd);
2184 /* check normal bins */
2185 for (i = 1; i < NBINS; ++i)
2187 b = bin_at (av, i);
2189 /* binmap is accurate (except for bin 1 == unsorted_chunks) */
2190 if (i >= 2)
2192 unsigned int binbit = get_binmap (av, i);
2193 int empty = last (b) == b;
2194 if (!binbit)
2195 assert (empty);
2196 else if (!empty)
2197 assert (binbit);
2200 for (p = last (b); p != b; p = p->bk)
2202 /* each chunk claims to be free */
2203 do_check_free_chunk (av, p);
2204 size = chunksize (p);
2205 total += size;
2206 if (i >= 2)
2208 /* chunk belongs in bin */
2209 idx = bin_index (size);
2210 assert (idx == i);
2211 /* lists are sorted */
2212 assert (p->bk == b ||
2213 (unsigned long) chunksize (p->bk) >= (unsigned long) chunksize (p));
2215 if (!in_smallbin_range (size))
2217 if (p->fd_nextsize != NULL)
2219 if (p->fd_nextsize == p)
2220 assert (p->bk_nextsize == p);
2221 else
2223 if (p->fd_nextsize == first (b))
2224 assert (chunksize (p) < chunksize (p->fd_nextsize));
2225 else
2226 assert (chunksize (p) > chunksize (p->fd_nextsize));
2228 if (p == first (b))
2229 assert (chunksize (p) > chunksize (p->bk_nextsize));
2230 else
2231 assert (chunksize (p) < chunksize (p->bk_nextsize));
2234 else
2235 assert (p->bk_nextsize == NULL);
2238 else if (!in_smallbin_range (size))
2239 assert (p->fd_nextsize == NULL && p->bk_nextsize == NULL);
2240 /* chunk is followed by a legal chain of inuse chunks */
2241 for (q = next_chunk (p);
2242 (q != av->top && inuse (q) &&
2243 (unsigned long) (chunksize (q)) >= MINSIZE);
2244 q = next_chunk (q))
2245 do_check_inuse_chunk (av, q);
2249 /* top chunk is OK */
2250 check_chunk (av, av->top);
2252 #endif
2255 /* ----------------- Support for debugging hooks -------------------- */
2256 #include "hooks.c"
2259 /* ----------- Routines dealing with system allocation -------------- */
2262 sysmalloc handles malloc cases requiring more memory from the system.
2263 On entry, it is assumed that av->top does not have enough
2264 space to service request for nb bytes, thus requiring that av->top
2265 be extended or replaced.
2268 static void *
2269 sysmalloc (INTERNAL_SIZE_T nb, mstate av)
2271 mchunkptr old_top; /* incoming value of av->top */
2272 INTERNAL_SIZE_T old_size; /* its size */
2273 char *old_end; /* its end address */
2275 long size; /* arg to first MORECORE or mmap call */
2276 char *brk; /* return value from MORECORE */
2278 long correction; /* arg to 2nd MORECORE call */
2279 char *snd_brk; /* 2nd return val */
2281 INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of new space */
2282 INTERNAL_SIZE_T end_misalign; /* partial page left at end of new space */
2283 char *aligned_brk; /* aligned offset into brk */
2285 mchunkptr p; /* the allocated/returned chunk */
2286 mchunkptr remainder; /* remainder from allocation */
2287 unsigned long remainder_size; /* its size */
2290 size_t pagesize = GLRO (dl_pagesize);
2291 bool tried_mmap = false;
2295 If have mmap, and the request size meets the mmap threshold, and
2296 the system supports mmap, and there are few enough currently
2297 allocated mmapped regions, try to directly map this request
2298 rather than expanding top.
2301 if (av == NULL
2302 || ((unsigned long) (nb) >= (unsigned long) (mp_.mmap_threshold)
2303 && (mp_.n_mmaps < mp_.n_mmaps_max)))
2305 char *mm; /* return value from mmap call*/
2307 try_mmap:
2309 Round up size to nearest page. For mmapped chunks, the overhead
2310 is one SIZE_SZ unit larger than for normal chunks, because there
2311 is no following chunk whose prev_size field could be used.
2313 See the front_misalign handling below, for glibc there is no
2314 need for further alignments unless we have have high alignment.
2316 if (MALLOC_ALIGNMENT == 2 * SIZE_SZ)
2317 size = ALIGN_UP (nb + SIZE_SZ, pagesize);
2318 else
2319 size = ALIGN_UP (nb + SIZE_SZ + MALLOC_ALIGN_MASK, pagesize);
2320 tried_mmap = true;
2322 /* Don't try if size wraps around 0 */
2323 if ((unsigned long) (size) > (unsigned long) (nb))
2325 mm = (char *) (MMAP (0, size, PROT_READ | PROT_WRITE, 0));
2327 if (mm != MAP_FAILED)
2330 The offset to the start of the mmapped region is stored
2331 in the prev_size field of the chunk. This allows us to adjust
2332 returned start address to meet alignment requirements here
2333 and in memalign(), and still be able to compute proper
2334 address argument for later munmap in free() and realloc().
2337 if (MALLOC_ALIGNMENT == 2 * SIZE_SZ)
2339 /* For glibc, chunk2mem increases the address by 2*SIZE_SZ and
2340 MALLOC_ALIGN_MASK is 2*SIZE_SZ-1. Each mmap'ed area is page
2341 aligned and therefore definitely MALLOC_ALIGN_MASK-aligned. */
2342 assert (((INTERNAL_SIZE_T) chunk2mem (mm) & MALLOC_ALIGN_MASK) == 0);
2343 front_misalign = 0;
2345 else
2346 front_misalign = (INTERNAL_SIZE_T) chunk2mem (mm) & MALLOC_ALIGN_MASK;
2347 if (front_misalign > 0)
2349 correction = MALLOC_ALIGNMENT - front_misalign;
2350 p = (mchunkptr) (mm + correction);
2351 set_prev_size (p, correction);
2352 set_head (p, (size - correction) | IS_MMAPPED);
2354 else
2356 p = (mchunkptr) mm;
2357 set_prev_size (p, 0);
2358 set_head (p, size | IS_MMAPPED);
2361 /* update statistics */
2363 int new = atomic_exchange_and_add (&mp_.n_mmaps, 1) + 1;
2364 atomic_max (&mp_.max_n_mmaps, new);
2366 unsigned long sum;
2367 sum = atomic_exchange_and_add (&mp_.mmapped_mem, size) + size;
2368 atomic_max (&mp_.max_mmapped_mem, sum);
2370 check_chunk (av, p);
2372 return chunk2mem (p);
2377 /* There are no usable arenas and mmap also failed. */
2378 if (av == NULL)
2379 return 0;
2381 /* Record incoming configuration of top */
2383 old_top = av->top;
2384 old_size = chunksize (old_top);
2385 old_end = (char *) (chunk_at_offset (old_top, old_size));
2387 brk = snd_brk = (char *) (MORECORE_FAILURE);
2390 If not the first time through, we require old_size to be
2391 at least MINSIZE and to have prev_inuse set.
2394 assert ((old_top == initial_top (av) && old_size == 0) ||
2395 ((unsigned long) (old_size) >= MINSIZE &&
2396 prev_inuse (old_top) &&
2397 ((unsigned long) old_end & (pagesize - 1)) == 0));
2399 /* Precondition: not enough current space to satisfy nb request */
2400 assert ((unsigned long) (old_size) < (unsigned long) (nb + MINSIZE));
2403 if (av != &main_arena)
2405 heap_info *old_heap, *heap;
2406 size_t old_heap_size;
2408 /* First try to extend the current heap. */
2409 old_heap = heap_for_ptr (old_top);
2410 old_heap_size = old_heap->size;
2411 if ((long) (MINSIZE + nb - old_size) > 0
2412 && grow_heap (old_heap, MINSIZE + nb - old_size) == 0)
2414 av->system_mem += old_heap->size - old_heap_size;
2415 set_head (old_top, (((char *) old_heap + old_heap->size) - (char *) old_top)
2416 | PREV_INUSE);
2418 else if ((heap = new_heap (nb + (MINSIZE + sizeof (*heap)), mp_.top_pad)))
2420 /* Use a newly allocated heap. */
2421 heap->ar_ptr = av;
2422 heap->prev = old_heap;
2423 av->system_mem += heap->size;
2424 /* Set up the new top. */
2425 top (av) = chunk_at_offset (heap, sizeof (*heap));
2426 set_head (top (av), (heap->size - sizeof (*heap)) | PREV_INUSE);
2428 /* Setup fencepost and free the old top chunk with a multiple of
2429 MALLOC_ALIGNMENT in size. */
2430 /* The fencepost takes at least MINSIZE bytes, because it might
2431 become the top chunk again later. Note that a footer is set
2432 up, too, although the chunk is marked in use. */
2433 old_size = (old_size - MINSIZE) & ~MALLOC_ALIGN_MASK;
2434 set_head (chunk_at_offset (old_top, old_size + 2 * SIZE_SZ), 0 | PREV_INUSE);
2435 if (old_size >= MINSIZE)
2437 set_head (chunk_at_offset (old_top, old_size), (2 * SIZE_SZ) | PREV_INUSE);
2438 set_foot (chunk_at_offset (old_top, old_size), (2 * SIZE_SZ));
2439 set_head (old_top, old_size | PREV_INUSE | NON_MAIN_ARENA);
2440 _int_free (av, old_top, 1);
2442 else
2444 set_head (old_top, (old_size + 2 * SIZE_SZ) | PREV_INUSE);
2445 set_foot (old_top, (old_size + 2 * SIZE_SZ));
2448 else if (!tried_mmap)
2449 /* We can at least try to use to mmap memory. */
2450 goto try_mmap;
2452 else /* av == main_arena */
2455 { /* Request enough space for nb + pad + overhead */
2456 size = nb + mp_.top_pad + MINSIZE;
2459 If contiguous, we can subtract out existing space that we hope to
2460 combine with new space. We add it back later only if
2461 we don't actually get contiguous space.
2464 if (contiguous (av))
2465 size -= old_size;
2468 Round to a multiple of page size.
2469 If MORECORE is not contiguous, this ensures that we only call it
2470 with whole-page arguments. And if MORECORE is contiguous and
2471 this is not first time through, this preserves page-alignment of
2472 previous calls. Otherwise, we correct to page-align below.
2475 size = ALIGN_UP (size, pagesize);
2478 Don't try to call MORECORE if argument is so big as to appear
2479 negative. Note that since mmap takes size_t arg, it may succeed
2480 below even if we cannot call MORECORE.
2483 if (size > 0)
2485 brk = (char *) (MORECORE (size));
2486 LIBC_PROBE (memory_sbrk_more, 2, brk, size);
2489 if (brk != (char *) (MORECORE_FAILURE))
2491 /* Call the `morecore' hook if necessary. */
2492 void (*hook) (void) = atomic_forced_read (__after_morecore_hook);
2493 if (__builtin_expect (hook != NULL, 0))
2494 (*hook)();
2496 else
2499 If have mmap, try using it as a backup when MORECORE fails or
2500 cannot be used. This is worth doing on systems that have "holes" in
2501 address space, so sbrk cannot extend to give contiguous space, but
2502 space is available elsewhere. Note that we ignore mmap max count
2503 and threshold limits, since the space will not be used as a
2504 segregated mmap region.
2507 /* Cannot merge with old top, so add its size back in */
2508 if (contiguous (av))
2509 size = ALIGN_UP (size + old_size, pagesize);
2511 /* If we are relying on mmap as backup, then use larger units */
2512 if ((unsigned long) (size) < (unsigned long) (MMAP_AS_MORECORE_SIZE))
2513 size = MMAP_AS_MORECORE_SIZE;
2515 /* Don't try if size wraps around 0 */
2516 if ((unsigned long) (size) > (unsigned long) (nb))
2518 char *mbrk = (char *) (MMAP (0, size, PROT_READ | PROT_WRITE, 0));
2520 if (mbrk != MAP_FAILED)
2522 /* We do not need, and cannot use, another sbrk call to find end */
2523 brk = mbrk;
2524 snd_brk = brk + size;
2527 Record that we no longer have a contiguous sbrk region.
2528 After the first time mmap is used as backup, we do not
2529 ever rely on contiguous space since this could incorrectly
2530 bridge regions.
2532 set_noncontiguous (av);
2537 if (brk != (char *) (MORECORE_FAILURE))
2539 if (mp_.sbrk_base == 0)
2540 mp_.sbrk_base = brk;
2541 av->system_mem += size;
2544 If MORECORE extends previous space, we can likewise extend top size.
2547 if (brk == old_end && snd_brk == (char *) (MORECORE_FAILURE))
2548 set_head (old_top, (size + old_size) | PREV_INUSE);
2550 else if (contiguous (av) && old_size && brk < old_end)
2551 /* Oops! Someone else killed our space.. Can't touch anything. */
2552 malloc_printerr ("break adjusted to free malloc space");
2555 Otherwise, make adjustments:
2557 * If the first time through or noncontiguous, we need to call sbrk
2558 just to find out where the end of memory lies.
2560 * We need to ensure that all returned chunks from malloc will meet
2561 MALLOC_ALIGNMENT
2563 * If there was an intervening foreign sbrk, we need to adjust sbrk
2564 request size to account for fact that we will not be able to
2565 combine new space with existing space in old_top.
2567 * Almost all systems internally allocate whole pages at a time, in
2568 which case we might as well use the whole last page of request.
2569 So we allocate enough more memory to hit a page boundary now,
2570 which in turn causes future contiguous calls to page-align.
2573 else
2575 front_misalign = 0;
2576 end_misalign = 0;
2577 correction = 0;
2578 aligned_brk = brk;
2580 /* handle contiguous cases */
2581 if (contiguous (av))
2583 /* Count foreign sbrk as system_mem. */
2584 if (old_size)
2585 av->system_mem += brk - old_end;
2587 /* Guarantee alignment of first new chunk made from this space */
2589 front_misalign = (INTERNAL_SIZE_T) chunk2mem (brk) & MALLOC_ALIGN_MASK;
2590 if (front_misalign > 0)
2593 Skip over some bytes to arrive at an aligned position.
2594 We don't need to specially mark these wasted front bytes.
2595 They will never be accessed anyway because
2596 prev_inuse of av->top (and any chunk created from its start)
2597 is always true after initialization.
2600 correction = MALLOC_ALIGNMENT - front_misalign;
2601 aligned_brk += correction;
2605 If this isn't adjacent to existing space, then we will not
2606 be able to merge with old_top space, so must add to 2nd request.
2609 correction += old_size;
2611 /* Extend the end address to hit a page boundary */
2612 end_misalign = (INTERNAL_SIZE_T) (brk + size + correction);
2613 correction += (ALIGN_UP (end_misalign, pagesize)) - end_misalign;
2615 assert (correction >= 0);
2616 snd_brk = (char *) (MORECORE (correction));
2619 If can't allocate correction, try to at least find out current
2620 brk. It might be enough to proceed without failing.
2622 Note that if second sbrk did NOT fail, we assume that space
2623 is contiguous with first sbrk. This is a safe assumption unless
2624 program is multithreaded but doesn't use locks and a foreign sbrk
2625 occurred between our first and second calls.
2628 if (snd_brk == (char *) (MORECORE_FAILURE))
2630 correction = 0;
2631 snd_brk = (char *) (MORECORE (0));
2633 else
2635 /* Call the `morecore' hook if necessary. */
2636 void (*hook) (void) = atomic_forced_read (__after_morecore_hook);
2637 if (__builtin_expect (hook != NULL, 0))
2638 (*hook)();
2642 /* handle non-contiguous cases */
2643 else
2645 if (MALLOC_ALIGNMENT == 2 * SIZE_SZ)
2646 /* MORECORE/mmap must correctly align */
2647 assert (((unsigned long) chunk2mem (brk) & MALLOC_ALIGN_MASK) == 0);
2648 else
2650 front_misalign = (INTERNAL_SIZE_T) chunk2mem (brk) & MALLOC_ALIGN_MASK;
2651 if (front_misalign > 0)
2654 Skip over some bytes to arrive at an aligned position.
2655 We don't need to specially mark these wasted front bytes.
2656 They will never be accessed anyway because
2657 prev_inuse of av->top (and any chunk created from its start)
2658 is always true after initialization.
2661 aligned_brk += MALLOC_ALIGNMENT - front_misalign;
2665 /* Find out current end of memory */
2666 if (snd_brk == (char *) (MORECORE_FAILURE))
2668 snd_brk = (char *) (MORECORE (0));
2672 /* Adjust top based on results of second sbrk */
2673 if (snd_brk != (char *) (MORECORE_FAILURE))
2675 av->top = (mchunkptr) aligned_brk;
2676 set_head (av->top, (snd_brk - aligned_brk + correction) | PREV_INUSE);
2677 av->system_mem += correction;
2680 If not the first time through, we either have a
2681 gap due to foreign sbrk or a non-contiguous region. Insert a
2682 double fencepost at old_top to prevent consolidation with space
2683 we don't own. These fenceposts are artificial chunks that are
2684 marked as inuse and are in any case too small to use. We need
2685 two to make sizes and alignments work out.
2688 if (old_size != 0)
2691 Shrink old_top to insert fenceposts, keeping size a
2692 multiple of MALLOC_ALIGNMENT. We know there is at least
2693 enough space in old_top to do this.
2695 old_size = (old_size - 4 * SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2696 set_head (old_top, old_size | PREV_INUSE);
2699 Note that the following assignments completely overwrite
2700 old_top when old_size was previously MINSIZE. This is
2701 intentional. We need the fencepost, even if old_top otherwise gets
2702 lost.
2704 set_head (chunk_at_offset (old_top, old_size),
2705 (2 * SIZE_SZ) | PREV_INUSE);
2706 set_head (chunk_at_offset (old_top, old_size + 2 * SIZE_SZ),
2707 (2 * SIZE_SZ) | PREV_INUSE);
2709 /* If possible, release the rest. */
2710 if (old_size >= MINSIZE)
2712 _int_free (av, old_top, 1);
2718 } /* if (av != &main_arena) */
2720 if ((unsigned long) av->system_mem > (unsigned long) (av->max_system_mem))
2721 av->max_system_mem = av->system_mem;
2722 check_malloc_state (av);
2724 /* finally, do the allocation */
2725 p = av->top;
2726 size = chunksize (p);
2728 /* check that one of the above allocation paths succeeded */
2729 if ((unsigned long) (size) >= (unsigned long) (nb + MINSIZE))
2731 remainder_size = size - nb;
2732 remainder = chunk_at_offset (p, nb);
2733 av->top = remainder;
2734 set_head (p, nb | PREV_INUSE | (av != &main_arena ? NON_MAIN_ARENA : 0));
2735 set_head (remainder, remainder_size | PREV_INUSE);
2736 check_malloced_chunk (av, p, nb);
2737 return chunk2mem (p);
2740 /* catch all failure paths */
2741 __set_errno (ENOMEM);
2742 return 0;
2747 systrim is an inverse of sorts to sysmalloc. It gives memory back
2748 to the system (via negative arguments to sbrk) if there is unused
2749 memory at the `high' end of the malloc pool. It is called
2750 automatically by free() when top space exceeds the trim
2751 threshold. It is also called by the public malloc_trim routine. It
2752 returns 1 if it actually released any memory, else 0.
2755 static int
2756 systrim (size_t pad, mstate av)
2758 long top_size; /* Amount of top-most memory */
2759 long extra; /* Amount to release */
2760 long released; /* Amount actually released */
2761 char *current_brk; /* address returned by pre-check sbrk call */
2762 char *new_brk; /* address returned by post-check sbrk call */
2763 size_t pagesize;
2764 long top_area;
2766 pagesize = GLRO (dl_pagesize);
2767 top_size = chunksize (av->top);
2769 top_area = top_size - MINSIZE - 1;
2770 if (top_area <= pad)
2771 return 0;
2773 /* Release in pagesize units and round down to the nearest page. */
2774 extra = ALIGN_DOWN(top_area - pad, pagesize);
2776 if (extra == 0)
2777 return 0;
2780 Only proceed if end of memory is where we last set it.
2781 This avoids problems if there were foreign sbrk calls.
2783 current_brk = (char *) (MORECORE (0));
2784 if (current_brk == (char *) (av->top) + top_size)
2787 Attempt to release memory. We ignore MORECORE return value,
2788 and instead call again to find out where new end of memory is.
2789 This avoids problems if first call releases less than we asked,
2790 of if failure somehow altered brk value. (We could still
2791 encounter problems if it altered brk in some very bad way,
2792 but the only thing we can do is adjust anyway, which will cause
2793 some downstream failure.)
2796 MORECORE (-extra);
2797 /* Call the `morecore' hook if necessary. */
2798 void (*hook) (void) = atomic_forced_read (__after_morecore_hook);
2799 if (__builtin_expect (hook != NULL, 0))
2800 (*hook)();
2801 new_brk = (char *) (MORECORE (0));
2803 LIBC_PROBE (memory_sbrk_less, 2, new_brk, extra);
2805 if (new_brk != (char *) MORECORE_FAILURE)
2807 released = (long) (current_brk - new_brk);
2809 if (released != 0)
2811 /* Success. Adjust top. */
2812 av->system_mem -= released;
2813 set_head (av->top, (top_size - released) | PREV_INUSE);
2814 check_malloc_state (av);
2815 return 1;
2819 return 0;
2822 static void
2823 munmap_chunk (mchunkptr p)
2825 size_t pagesize = GLRO (dl_pagesize);
2826 INTERNAL_SIZE_T size = chunksize (p);
2828 assert (chunk_is_mmapped (p));
2830 /* Do nothing if the chunk is a faked mmapped chunk in the dumped
2831 main arena. We never free this memory. */
2832 if (DUMPED_MAIN_ARENA_CHUNK (p))
2833 return;
2835 uintptr_t mem = (uintptr_t) chunk2mem (p);
2836 uintptr_t block = (uintptr_t) p - prev_size (p);
2837 size_t total_size = prev_size (p) + size;
2838 /* Unfortunately we have to do the compilers job by hand here. Normally
2839 we would test BLOCK and TOTAL-SIZE separately for compliance with the
2840 page size. But gcc does not recognize the optimization possibility
2841 (in the moment at least) so we combine the two values into one before
2842 the bit test. */
2843 if (__glibc_unlikely ((block | total_size) & (pagesize - 1)) != 0
2844 || __glibc_unlikely (!powerof2 (mem & (pagesize - 1))))
2845 malloc_printerr ("munmap_chunk(): invalid pointer");
2847 atomic_decrement (&mp_.n_mmaps);
2848 atomic_add (&mp_.mmapped_mem, -total_size);
2850 /* If munmap failed the process virtual memory address space is in a
2851 bad shape. Just leave the block hanging around, the process will
2852 terminate shortly anyway since not much can be done. */
2853 __munmap ((char *) block, total_size);
2856 #if HAVE_MREMAP
2858 static mchunkptr
2859 mremap_chunk (mchunkptr p, size_t new_size)
2861 size_t pagesize = GLRO (dl_pagesize);
2862 INTERNAL_SIZE_T offset = prev_size (p);
2863 INTERNAL_SIZE_T size = chunksize (p);
2864 char *cp;
2866 assert (chunk_is_mmapped (p));
2868 uintptr_t block = (uintptr_t) p - offset;
2869 uintptr_t mem = (uintptr_t) chunk2mem(p);
2870 size_t total_size = offset + size;
2871 if (__glibc_unlikely ((block | total_size) & (pagesize - 1)) != 0
2872 || __glibc_unlikely (!powerof2 (mem & (pagesize - 1))))
2873 malloc_printerr("mremap_chunk(): invalid pointer");
2875 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
2876 new_size = ALIGN_UP (new_size + offset + SIZE_SZ, pagesize);
2878 /* No need to remap if the number of pages does not change. */
2879 if (total_size == new_size)
2880 return p;
2882 cp = (char *) __mremap ((char *) block, total_size, new_size,
2883 MREMAP_MAYMOVE);
2885 if (cp == MAP_FAILED)
2886 return 0;
2888 p = (mchunkptr) (cp + offset);
2890 assert (aligned_OK (chunk2mem (p)));
2892 assert (prev_size (p) == offset);
2893 set_head (p, (new_size - offset) | IS_MMAPPED);
2895 INTERNAL_SIZE_T new;
2896 new = atomic_exchange_and_add (&mp_.mmapped_mem, new_size - size - offset)
2897 + new_size - size - offset;
2898 atomic_max (&mp_.max_mmapped_mem, new);
2899 return p;
2901 #endif /* HAVE_MREMAP */
2903 /*------------------------ Public wrappers. --------------------------------*/
2905 #if USE_TCACHE
2907 /* We overlay this structure on the user-data portion of a chunk when
2908 the chunk is stored in the per-thread cache. */
2909 typedef struct tcache_entry
2911 struct tcache_entry *next;
2912 /* This field exists to detect double frees. */
2913 struct tcache_perthread_struct *key;
2914 } tcache_entry;
2916 /* There is one of these for each thread, which contains the
2917 per-thread cache (hence "tcache_perthread_struct"). Keeping
2918 overall size low is mildly important. Note that COUNTS and ENTRIES
2919 are redundant (we could have just counted the linked list each
2920 time), this is for performance reasons. */
2921 typedef struct tcache_perthread_struct
2923 uint16_t counts[TCACHE_MAX_BINS];
2924 tcache_entry *entries[TCACHE_MAX_BINS];
2925 } tcache_perthread_struct;
2927 static __thread bool tcache_shutting_down = false;
2928 static __thread tcache_perthread_struct *tcache = NULL;
2930 /* Caller must ensure that we know tc_idx is valid and there's room
2931 for more chunks. */
2932 static __always_inline void
2933 tcache_put (mchunkptr chunk, size_t tc_idx)
2935 tcache_entry *e = (tcache_entry *) chunk2mem (chunk);
2937 /* Mark this chunk as "in the tcache" so the test in _int_free will
2938 detect a double free. */
2939 e->key = tcache;
2941 e->next = PROTECT_PTR (&e->next, tcache->entries[tc_idx]);
2942 tcache->entries[tc_idx] = e;
2943 ++(tcache->counts[tc_idx]);
2946 /* Caller must ensure that we know tc_idx is valid and there's
2947 available chunks to remove. */
2948 static __always_inline void *
2949 tcache_get (size_t tc_idx)
2951 tcache_entry *e = tcache->entries[tc_idx];
2952 if (__glibc_unlikely (!aligned_OK (e)))
2953 malloc_printerr ("malloc(): unaligned tcache chunk detected");
2954 tcache->entries[tc_idx] = REVEAL_PTR (e->next);
2955 --(tcache->counts[tc_idx]);
2956 e->key = NULL;
2957 return (void *) e;
2960 static void
2961 tcache_thread_shutdown (void)
2963 int i;
2964 tcache_perthread_struct *tcache_tmp = tcache;
2966 if (!tcache)
2967 return;
2969 /* Disable the tcache and prevent it from being reinitialized. */
2970 tcache = NULL;
2971 tcache_shutting_down = true;
2973 /* Free all of the entries and the tcache itself back to the arena
2974 heap for coalescing. */
2975 for (i = 0; i < TCACHE_MAX_BINS; ++i)
2977 while (tcache_tmp->entries[i])
2979 tcache_entry *e = tcache_tmp->entries[i];
2980 if (__glibc_unlikely (!aligned_OK (e)))
2981 malloc_printerr ("tcache_thread_shutdown(): "
2982 "unaligned tcache chunk detected");
2983 tcache_tmp->entries[i] = REVEAL_PTR (e->next);
2984 __libc_free (e);
2988 __libc_free (tcache_tmp);
2991 static void
2992 tcache_init(void)
2994 mstate ar_ptr;
2995 void *victim = 0;
2996 const size_t bytes = sizeof (tcache_perthread_struct);
2998 if (tcache_shutting_down)
2999 return;
3001 arena_get (ar_ptr, bytes);
3002 victim = _int_malloc (ar_ptr, bytes);
3003 if (!victim && ar_ptr != NULL)
3005 ar_ptr = arena_get_retry (ar_ptr, bytes);
3006 victim = _int_malloc (ar_ptr, bytes);
3010 if (ar_ptr != NULL)
3011 __libc_lock_unlock (ar_ptr->mutex);
3013 /* In a low memory situation, we may not be able to allocate memory
3014 - in which case, we just keep trying later. However, we
3015 typically do this very early, so either there is sufficient
3016 memory, or there isn't enough memory to do non-trivial
3017 allocations anyway. */
3018 if (victim)
3020 tcache = (tcache_perthread_struct *) victim;
3021 memset (tcache, 0, sizeof (tcache_perthread_struct));
3026 # define MAYBE_INIT_TCACHE() \
3027 if (__glibc_unlikely (tcache == NULL)) \
3028 tcache_init();
3030 #else /* !USE_TCACHE */
3031 # define MAYBE_INIT_TCACHE()
3033 static void
3034 tcache_thread_shutdown (void)
3036 /* Nothing to do if there is no thread cache. */
3039 #endif /* !USE_TCACHE */
3041 void *
3042 __libc_malloc (size_t bytes)
3044 mstate ar_ptr;
3045 void *victim;
3047 _Static_assert (PTRDIFF_MAX <= SIZE_MAX / 2,
3048 "PTRDIFF_MAX is not more than half of SIZE_MAX");
3050 void *(*hook) (size_t, const void *)
3051 = atomic_forced_read (__malloc_hook);
3052 if (__builtin_expect (hook != NULL, 0))
3053 return (*hook)(bytes, RETURN_ADDRESS (0));
3054 #if USE_TCACHE
3055 /* int_free also calls request2size, be careful to not pad twice. */
3056 size_t tbytes;
3057 if (!checked_request2size (bytes, &tbytes))
3059 __set_errno (ENOMEM);
3060 return NULL;
3062 size_t tc_idx = csize2tidx (tbytes);
3064 MAYBE_INIT_TCACHE ();
3066 DIAG_PUSH_NEEDS_COMMENT;
3067 if (tc_idx < mp_.tcache_bins
3068 && tcache
3069 && tcache->counts[tc_idx] > 0)
3071 return tcache_get (tc_idx);
3073 DIAG_POP_NEEDS_COMMENT;
3074 #endif
3076 if (SINGLE_THREAD_P)
3078 victim = _int_malloc (&main_arena, bytes);
3079 assert (!victim || chunk_is_mmapped (mem2chunk (victim)) ||
3080 &main_arena == arena_for_chunk (mem2chunk (victim)));
3081 return victim;
3084 arena_get (ar_ptr, bytes);
3086 victim = _int_malloc (ar_ptr, bytes);
3087 /* Retry with another arena only if we were able to find a usable arena
3088 before. */
3089 if (!victim && ar_ptr != NULL)
3091 LIBC_PROBE (memory_malloc_retry, 1, bytes);
3092 ar_ptr = arena_get_retry (ar_ptr, bytes);
3093 victim = _int_malloc (ar_ptr, bytes);
3096 if (ar_ptr != NULL)
3097 __libc_lock_unlock (ar_ptr->mutex);
3099 assert (!victim || chunk_is_mmapped (mem2chunk (victim)) ||
3100 ar_ptr == arena_for_chunk (mem2chunk (victim)));
3101 return victim;
3103 libc_hidden_def (__libc_malloc)
3105 void
3106 __libc_free (void *mem)
3108 mstate ar_ptr;
3109 mchunkptr p; /* chunk corresponding to mem */
3111 void (*hook) (void *, const void *)
3112 = atomic_forced_read (__free_hook);
3113 if (__builtin_expect (hook != NULL, 0))
3115 (*hook)(mem, RETURN_ADDRESS (0));
3116 return;
3119 if (mem == 0) /* free(0) has no effect */
3120 return;
3122 p = mem2chunk (mem);
3124 if (chunk_is_mmapped (p)) /* release mmapped memory. */
3126 /* See if the dynamic brk/mmap threshold needs adjusting.
3127 Dumped fake mmapped chunks do not affect the threshold. */
3128 if (!mp_.no_dyn_threshold
3129 && chunksize_nomask (p) > mp_.mmap_threshold
3130 && chunksize_nomask (p) <= DEFAULT_MMAP_THRESHOLD_MAX
3131 && !DUMPED_MAIN_ARENA_CHUNK (p))
3133 mp_.mmap_threshold = chunksize (p);
3134 mp_.trim_threshold = 2 * mp_.mmap_threshold;
3135 LIBC_PROBE (memory_mallopt_free_dyn_thresholds, 2,
3136 mp_.mmap_threshold, mp_.trim_threshold);
3138 munmap_chunk (p);
3139 return;
3142 MAYBE_INIT_TCACHE ();
3144 ar_ptr = arena_for_chunk (p);
3145 _int_free (ar_ptr, p, 0);
3147 libc_hidden_def (__libc_free)
3149 void *
3150 __libc_realloc (void *oldmem, size_t bytes)
3152 mstate ar_ptr;
3153 INTERNAL_SIZE_T nb; /* padded request size */
3155 void *newp; /* chunk to return */
3157 void *(*hook) (void *, size_t, const void *) =
3158 atomic_forced_read (__realloc_hook);
3159 if (__builtin_expect (hook != NULL, 0))
3160 return (*hook)(oldmem, bytes, RETURN_ADDRESS (0));
3162 #if REALLOC_ZERO_BYTES_FREES
3163 if (bytes == 0 && oldmem != NULL)
3165 __libc_free (oldmem); return 0;
3167 #endif
3169 /* realloc of null is supposed to be same as malloc */
3170 if (oldmem == 0)
3171 return __libc_malloc (bytes);
3173 /* chunk corresponding to oldmem */
3174 const mchunkptr oldp = mem2chunk (oldmem);
3175 /* its size */
3176 const INTERNAL_SIZE_T oldsize = chunksize (oldp);
3178 if (chunk_is_mmapped (oldp))
3179 ar_ptr = NULL;
3180 else
3182 MAYBE_INIT_TCACHE ();
3183 ar_ptr = arena_for_chunk (oldp);
3186 /* Little security check which won't hurt performance: the allocator
3187 never wrapps around at the end of the address space. Therefore
3188 we can exclude some size values which might appear here by
3189 accident or by "design" from some intruder. We need to bypass
3190 this check for dumped fake mmap chunks from the old main arena
3191 because the new malloc may provide additional alignment. */
3192 if ((__builtin_expect ((uintptr_t) oldp > (uintptr_t) -oldsize, 0)
3193 || __builtin_expect (misaligned_chunk (oldp), 0))
3194 && !DUMPED_MAIN_ARENA_CHUNK (oldp))
3195 malloc_printerr ("realloc(): invalid pointer");
3197 if (!checked_request2size (bytes, &nb))
3199 __set_errno (ENOMEM);
3200 return NULL;
3203 if (chunk_is_mmapped (oldp))
3205 /* If this is a faked mmapped chunk from the dumped main arena,
3206 always make a copy (and do not free the old chunk). */
3207 if (DUMPED_MAIN_ARENA_CHUNK (oldp))
3209 /* Must alloc, copy, free. */
3210 void *newmem = __libc_malloc (bytes);
3211 if (newmem == 0)
3212 return NULL;
3213 /* Copy as many bytes as are available from the old chunk
3214 and fit into the new size. NB: The overhead for faked
3215 mmapped chunks is only SIZE_SZ, not 2 * SIZE_SZ as for
3216 regular mmapped chunks. */
3217 if (bytes > oldsize - SIZE_SZ)
3218 bytes = oldsize - SIZE_SZ;
3219 memcpy (newmem, oldmem, bytes);
3220 return newmem;
3223 void *newmem;
3225 #if HAVE_MREMAP
3226 newp = mremap_chunk (oldp, nb);
3227 if (newp)
3228 return chunk2mem (newp);
3229 #endif
3230 /* Note the extra SIZE_SZ overhead. */
3231 if (oldsize - SIZE_SZ >= nb)
3232 return oldmem; /* do nothing */
3234 /* Must alloc, copy, free. */
3235 newmem = __libc_malloc (bytes);
3236 if (newmem == 0)
3237 return 0; /* propagate failure */
3239 memcpy (newmem, oldmem, oldsize - 2 * SIZE_SZ);
3240 munmap_chunk (oldp);
3241 return newmem;
3244 if (SINGLE_THREAD_P)
3246 newp = _int_realloc (ar_ptr, oldp, oldsize, nb);
3247 assert (!newp || chunk_is_mmapped (mem2chunk (newp)) ||
3248 ar_ptr == arena_for_chunk (mem2chunk (newp)));
3250 return newp;
3253 __libc_lock_lock (ar_ptr->mutex);
3255 newp = _int_realloc (ar_ptr, oldp, oldsize, nb);
3257 __libc_lock_unlock (ar_ptr->mutex);
3258 assert (!newp || chunk_is_mmapped (mem2chunk (newp)) ||
3259 ar_ptr == arena_for_chunk (mem2chunk (newp)));
3261 if (newp == NULL)
3263 /* Try harder to allocate memory in other arenas. */
3264 LIBC_PROBE (memory_realloc_retry, 2, bytes, oldmem);
3265 newp = __libc_malloc (bytes);
3266 if (newp != NULL)
3268 memcpy (newp, oldmem, oldsize - SIZE_SZ);
3269 _int_free (ar_ptr, oldp, 0);
3273 return newp;
3275 libc_hidden_def (__libc_realloc)
3277 void *
3278 __libc_memalign (size_t alignment, size_t bytes)
3280 void *address = RETURN_ADDRESS (0);
3281 return _mid_memalign (alignment, bytes, address);
3284 static void *
3285 _mid_memalign (size_t alignment, size_t bytes, void *address)
3287 mstate ar_ptr;
3288 void *p;
3290 void *(*hook) (size_t, size_t, const void *) =
3291 atomic_forced_read (__memalign_hook);
3292 if (__builtin_expect (hook != NULL, 0))
3293 return (*hook)(alignment, bytes, address);
3295 /* If we need less alignment than we give anyway, just relay to malloc. */
3296 if (alignment <= MALLOC_ALIGNMENT)
3297 return __libc_malloc (bytes);
3299 /* Otherwise, ensure that it is at least a minimum chunk size */
3300 if (alignment < MINSIZE)
3301 alignment = MINSIZE;
3303 /* If the alignment is greater than SIZE_MAX / 2 + 1 it cannot be a
3304 power of 2 and will cause overflow in the check below. */
3305 if (alignment > SIZE_MAX / 2 + 1)
3307 __set_errno (EINVAL);
3308 return 0;
3312 /* Make sure alignment is power of 2. */
3313 if (!powerof2 (alignment))
3315 size_t a = MALLOC_ALIGNMENT * 2;
3316 while (a < alignment)
3317 a <<= 1;
3318 alignment = a;
3321 if (SINGLE_THREAD_P)
3323 p = _int_memalign (&main_arena, alignment, bytes);
3324 assert (!p || chunk_is_mmapped (mem2chunk (p)) ||
3325 &main_arena == arena_for_chunk (mem2chunk (p)));
3327 return p;
3330 arena_get (ar_ptr, bytes + alignment + MINSIZE);
3332 p = _int_memalign (ar_ptr, alignment, bytes);
3333 if (!p && ar_ptr != NULL)
3335 LIBC_PROBE (memory_memalign_retry, 2, bytes, alignment);
3336 ar_ptr = arena_get_retry (ar_ptr, bytes);
3337 p = _int_memalign (ar_ptr, alignment, bytes);
3340 if (ar_ptr != NULL)
3341 __libc_lock_unlock (ar_ptr->mutex);
3343 assert (!p || chunk_is_mmapped (mem2chunk (p)) ||
3344 ar_ptr == arena_for_chunk (mem2chunk (p)));
3345 return p;
3347 /* For ISO C11. */
3348 weak_alias (__libc_memalign, aligned_alloc)
3349 libc_hidden_def (__libc_memalign)
3351 void *
3352 __libc_valloc (size_t bytes)
3354 if (__malloc_initialized < 0)
3355 ptmalloc_init ();
3357 void *address = RETURN_ADDRESS (0);
3358 size_t pagesize = GLRO (dl_pagesize);
3359 return _mid_memalign (pagesize, bytes, address);
3362 void *
3363 __libc_pvalloc (size_t bytes)
3365 if (__malloc_initialized < 0)
3366 ptmalloc_init ();
3368 void *address = RETURN_ADDRESS (0);
3369 size_t pagesize = GLRO (dl_pagesize);
3370 size_t rounded_bytes;
3371 /* ALIGN_UP with overflow check. */
3372 if (__glibc_unlikely (__builtin_add_overflow (bytes,
3373 pagesize - 1,
3374 &rounded_bytes)))
3376 __set_errno (ENOMEM);
3377 return 0;
3379 rounded_bytes = rounded_bytes & -(pagesize - 1);
3381 return _mid_memalign (pagesize, rounded_bytes, address);
3384 void *
3385 __libc_calloc (size_t n, size_t elem_size)
3387 mstate av;
3388 mchunkptr oldtop, p;
3389 INTERNAL_SIZE_T sz, csz, oldtopsize;
3390 void *mem;
3391 unsigned long clearsize;
3392 unsigned long nclears;
3393 INTERNAL_SIZE_T *d;
3394 ptrdiff_t bytes;
3396 if (__glibc_unlikely (__builtin_mul_overflow (n, elem_size, &bytes)))
3398 __set_errno (ENOMEM);
3399 return NULL;
3401 sz = bytes;
3403 void *(*hook) (size_t, const void *) =
3404 atomic_forced_read (__malloc_hook);
3405 if (__builtin_expect (hook != NULL, 0))
3407 mem = (*hook)(sz, RETURN_ADDRESS (0));
3408 if (mem == 0)
3409 return 0;
3411 return memset (mem, 0, sz);
3414 MAYBE_INIT_TCACHE ();
3416 if (SINGLE_THREAD_P)
3417 av = &main_arena;
3418 else
3419 arena_get (av, sz);
3421 if (av)
3423 /* Check if we hand out the top chunk, in which case there may be no
3424 need to clear. */
3425 #if MORECORE_CLEARS
3426 oldtop = top (av);
3427 oldtopsize = chunksize (top (av));
3428 # if MORECORE_CLEARS < 2
3429 /* Only newly allocated memory is guaranteed to be cleared. */
3430 if (av == &main_arena &&
3431 oldtopsize < mp_.sbrk_base + av->max_system_mem - (char *) oldtop)
3432 oldtopsize = (mp_.sbrk_base + av->max_system_mem - (char *) oldtop);
3433 # endif
3434 if (av != &main_arena)
3436 heap_info *heap = heap_for_ptr (oldtop);
3437 if (oldtopsize < (char *) heap + heap->mprotect_size - (char *) oldtop)
3438 oldtopsize = (char *) heap + heap->mprotect_size - (char *) oldtop;
3440 #endif
3442 else
3444 /* No usable arenas. */
3445 oldtop = 0;
3446 oldtopsize = 0;
3448 mem = _int_malloc (av, sz);
3450 assert (!mem || chunk_is_mmapped (mem2chunk (mem)) ||
3451 av == arena_for_chunk (mem2chunk (mem)));
3453 if (!SINGLE_THREAD_P)
3455 if (mem == 0 && av != NULL)
3457 LIBC_PROBE (memory_calloc_retry, 1, sz);
3458 av = arena_get_retry (av, sz);
3459 mem = _int_malloc (av, sz);
3462 if (av != NULL)
3463 __libc_lock_unlock (av->mutex);
3466 /* Allocation failed even after a retry. */
3467 if (mem == 0)
3468 return 0;
3470 p = mem2chunk (mem);
3472 /* Two optional cases in which clearing not necessary */
3473 if (chunk_is_mmapped (p))
3475 if (__builtin_expect (perturb_byte, 0))
3476 return memset (mem, 0, sz);
3478 return mem;
3481 csz = chunksize (p);
3483 #if MORECORE_CLEARS
3484 if (perturb_byte == 0 && (p == oldtop && csz > oldtopsize))
3486 /* clear only the bytes from non-freshly-sbrked memory */
3487 csz = oldtopsize;
3489 #endif
3491 /* Unroll clear of <= 36 bytes (72 if 8byte sizes). We know that
3492 contents have an odd number of INTERNAL_SIZE_T-sized words;
3493 minimally 3. */
3494 d = (INTERNAL_SIZE_T *) mem;
3495 clearsize = csz - SIZE_SZ;
3496 nclears = clearsize / sizeof (INTERNAL_SIZE_T);
3497 assert (nclears >= 3);
3499 if (nclears > 9)
3500 return memset (d, 0, clearsize);
3502 else
3504 *(d + 0) = 0;
3505 *(d + 1) = 0;
3506 *(d + 2) = 0;
3507 if (nclears > 4)
3509 *(d + 3) = 0;
3510 *(d + 4) = 0;
3511 if (nclears > 6)
3513 *(d + 5) = 0;
3514 *(d + 6) = 0;
3515 if (nclears > 8)
3517 *(d + 7) = 0;
3518 *(d + 8) = 0;
3524 return mem;
3528 ------------------------------ malloc ------------------------------
3531 static void *
3532 _int_malloc (mstate av, size_t bytes)
3534 INTERNAL_SIZE_T nb; /* normalized request size */
3535 unsigned int idx; /* associated bin index */
3536 mbinptr bin; /* associated bin */
3538 mchunkptr victim; /* inspected/selected chunk */
3539 INTERNAL_SIZE_T size; /* its size */
3540 int victim_index; /* its bin index */
3542 mchunkptr remainder; /* remainder from a split */
3543 unsigned long remainder_size; /* its size */
3545 unsigned int block; /* bit map traverser */
3546 unsigned int bit; /* bit map traverser */
3547 unsigned int map; /* current word of binmap */
3549 mchunkptr fwd; /* misc temp for linking */
3550 mchunkptr bck; /* misc temp for linking */
3552 #if USE_TCACHE
3553 size_t tcache_unsorted_count; /* count of unsorted chunks processed */
3554 #endif
3557 Convert request size to internal form by adding SIZE_SZ bytes
3558 overhead plus possibly more to obtain necessary alignment and/or
3559 to obtain a size of at least MINSIZE, the smallest allocatable
3560 size. Also, checked_request2size returns false for request sizes
3561 that are so large that they wrap around zero when padded and
3562 aligned.
3565 if (!checked_request2size (bytes, &nb))
3567 __set_errno (ENOMEM);
3568 return NULL;
3571 /* There are no usable arenas. Fall back to sysmalloc to get a chunk from
3572 mmap. */
3573 if (__glibc_unlikely (av == NULL))
3575 void *p = sysmalloc (nb, av);
3576 if (p != NULL)
3577 alloc_perturb (p, bytes);
3578 return p;
3582 If the size qualifies as a fastbin, first check corresponding bin.
3583 This code is safe to execute even if av is not yet initialized, so we
3584 can try it without checking, which saves some time on this fast path.
3587 #define REMOVE_FB(fb, victim, pp) \
3588 do \
3590 victim = pp; \
3591 if (victim == NULL) \
3592 break; \
3593 pp = REVEAL_PTR (victim->fd); \
3594 if (__glibc_unlikely (pp != NULL && misaligned_chunk (pp))) \
3595 malloc_printerr ("malloc(): unaligned fastbin chunk detected"); \
3597 while ((pp = catomic_compare_and_exchange_val_acq (fb, pp, victim)) \
3598 != victim); \
3600 if ((unsigned long) (nb) <= (unsigned long) (get_max_fast ()))
3602 idx = fastbin_index (nb);
3603 mfastbinptr *fb = &fastbin (av, idx);
3604 mchunkptr pp;
3605 victim = *fb;
3607 if (victim != NULL)
3609 if (__glibc_unlikely (misaligned_chunk (victim)))
3610 malloc_printerr ("malloc(): unaligned fastbin chunk detected 2");
3612 if (SINGLE_THREAD_P)
3613 *fb = REVEAL_PTR (victim->fd);
3614 else
3615 REMOVE_FB (fb, pp, victim);
3616 if (__glibc_likely (victim != NULL))
3618 size_t victim_idx = fastbin_index (chunksize (victim));
3619 if (__builtin_expect (victim_idx != idx, 0))
3620 malloc_printerr ("malloc(): memory corruption (fast)");
3621 check_remalloced_chunk (av, victim, nb);
3622 #if USE_TCACHE
3623 /* While we're here, if we see other chunks of the same size,
3624 stash them in the tcache. */
3625 size_t tc_idx = csize2tidx (nb);
3626 if (tcache && tc_idx < mp_.tcache_bins)
3628 mchunkptr tc_victim;
3630 /* While bin not empty and tcache not full, copy chunks. */
3631 while (tcache->counts[tc_idx] < mp_.tcache_count
3632 && (tc_victim = *fb) != NULL)
3634 if (__glibc_unlikely (misaligned_chunk (tc_victim)))
3635 malloc_printerr ("malloc(): unaligned fastbin chunk detected 3");
3636 if (SINGLE_THREAD_P)
3637 *fb = REVEAL_PTR (tc_victim->fd);
3638 else
3640 REMOVE_FB (fb, pp, tc_victim);
3641 if (__glibc_unlikely (tc_victim == NULL))
3642 break;
3644 tcache_put (tc_victim, tc_idx);
3647 #endif
3648 void *p = chunk2mem (victim);
3649 alloc_perturb (p, bytes);
3650 return p;
3656 If a small request, check regular bin. Since these "smallbins"
3657 hold one size each, no searching within bins is necessary.
3658 (For a large request, we need to wait until unsorted chunks are
3659 processed to find best fit. But for small ones, fits are exact
3660 anyway, so we can check now, which is faster.)
3663 if (in_smallbin_range (nb))
3665 idx = smallbin_index (nb);
3666 bin = bin_at (av, idx);
3668 if ((victim = last (bin)) != bin)
3670 bck = victim->bk;
3671 if (__glibc_unlikely (bck->fd != victim))
3672 malloc_printerr ("malloc(): smallbin double linked list corrupted");
3673 set_inuse_bit_at_offset (victim, nb);
3674 bin->bk = bck;
3675 bck->fd = bin;
3677 if (av != &main_arena)
3678 set_non_main_arena (victim);
3679 check_malloced_chunk (av, victim, nb);
3680 #if USE_TCACHE
3681 /* While we're here, if we see other chunks of the same size,
3682 stash them in the tcache. */
3683 size_t tc_idx = csize2tidx (nb);
3684 if (tcache && tc_idx < mp_.tcache_bins)
3686 mchunkptr tc_victim;
3688 /* While bin not empty and tcache not full, copy chunks over. */
3689 while (tcache->counts[tc_idx] < mp_.tcache_count
3690 && (tc_victim = last (bin)) != bin)
3692 if (tc_victim != 0)
3694 bck = tc_victim->bk;
3695 set_inuse_bit_at_offset (tc_victim, nb);
3696 if (av != &main_arena)
3697 set_non_main_arena (tc_victim);
3698 bin->bk = bck;
3699 bck->fd = bin;
3701 tcache_put (tc_victim, tc_idx);
3705 #endif
3706 void *p = chunk2mem (victim);
3707 alloc_perturb (p, bytes);
3708 return p;
3713 If this is a large request, consolidate fastbins before continuing.
3714 While it might look excessive to kill all fastbins before
3715 even seeing if there is space available, this avoids
3716 fragmentation problems normally associated with fastbins.
3717 Also, in practice, programs tend to have runs of either small or
3718 large requests, but less often mixtures, so consolidation is not
3719 invoked all that often in most programs. And the programs that
3720 it is called frequently in otherwise tend to fragment.
3723 else
3725 idx = largebin_index (nb);
3726 if (atomic_load_relaxed (&av->have_fastchunks))
3727 malloc_consolidate (av);
3731 Process recently freed or remaindered chunks, taking one only if
3732 it is exact fit, or, if this a small request, the chunk is remainder from
3733 the most recent non-exact fit. Place other traversed chunks in
3734 bins. Note that this step is the only place in any routine where
3735 chunks are placed in bins.
3737 The outer loop here is needed because we might not realize until
3738 near the end of malloc that we should have consolidated, so must
3739 do so and retry. This happens at most once, and only when we would
3740 otherwise need to expand memory to service a "small" request.
3743 #if USE_TCACHE
3744 INTERNAL_SIZE_T tcache_nb = 0;
3745 size_t tc_idx = csize2tidx (nb);
3746 if (tcache && tc_idx < mp_.tcache_bins)
3747 tcache_nb = nb;
3748 int return_cached = 0;
3750 tcache_unsorted_count = 0;
3751 #endif
3753 for (;; )
3755 int iters = 0;
3756 while ((victim = unsorted_chunks (av)->bk) != unsorted_chunks (av))
3758 bck = victim->bk;
3759 size = chunksize (victim);
3760 mchunkptr next = chunk_at_offset (victim, size);
3762 if (__glibc_unlikely (size <= 2 * SIZE_SZ)
3763 || __glibc_unlikely (size > av->system_mem))
3764 malloc_printerr ("malloc(): invalid size (unsorted)");
3765 if (__glibc_unlikely (chunksize_nomask (next) < 2 * SIZE_SZ)
3766 || __glibc_unlikely (chunksize_nomask (next) > av->system_mem))
3767 malloc_printerr ("malloc(): invalid next size (unsorted)");
3768 if (__glibc_unlikely ((prev_size (next) & ~(SIZE_BITS)) != size))
3769 malloc_printerr ("malloc(): mismatching next->prev_size (unsorted)");
3770 if (__glibc_unlikely (bck->fd != victim)
3771 || __glibc_unlikely (victim->fd != unsorted_chunks (av)))
3772 malloc_printerr ("malloc(): unsorted double linked list corrupted");
3773 if (__glibc_unlikely (prev_inuse (next)))
3774 malloc_printerr ("malloc(): invalid next->prev_inuse (unsorted)");
3777 If a small request, try to use last remainder if it is the
3778 only chunk in unsorted bin. This helps promote locality for
3779 runs of consecutive small requests. This is the only
3780 exception to best-fit, and applies only when there is
3781 no exact fit for a small chunk.
3784 if (in_smallbin_range (nb) &&
3785 bck == unsorted_chunks (av) &&
3786 victim == av->last_remainder &&
3787 (unsigned long) (size) > (unsigned long) (nb + MINSIZE))
3789 /* split and reattach remainder */
3790 remainder_size = size - nb;
3791 remainder = chunk_at_offset (victim, nb);
3792 unsorted_chunks (av)->bk = unsorted_chunks (av)->fd = remainder;
3793 av->last_remainder = remainder;
3794 remainder->bk = remainder->fd = unsorted_chunks (av);
3795 if (!in_smallbin_range (remainder_size))
3797 remainder->fd_nextsize = NULL;
3798 remainder->bk_nextsize = NULL;
3801 set_head (victim, nb | PREV_INUSE |
3802 (av != &main_arena ? NON_MAIN_ARENA : 0));
3803 set_head (remainder, remainder_size | PREV_INUSE);
3804 set_foot (remainder, remainder_size);
3806 check_malloced_chunk (av, victim, nb);
3807 void *p = chunk2mem (victim);
3808 alloc_perturb (p, bytes);
3809 return p;
3812 /* remove from unsorted list */
3813 if (__glibc_unlikely (bck->fd != victim))
3814 malloc_printerr ("malloc(): corrupted unsorted chunks 3");
3815 unsorted_chunks (av)->bk = bck;
3816 bck->fd = unsorted_chunks (av);
3818 /* Take now instead of binning if exact fit */
3820 if (size == nb)
3822 set_inuse_bit_at_offset (victim, size);
3823 if (av != &main_arena)
3824 set_non_main_arena (victim);
3825 #if USE_TCACHE
3826 /* Fill cache first, return to user only if cache fills.
3827 We may return one of these chunks later. */
3828 if (tcache_nb
3829 && tcache->counts[tc_idx] < mp_.tcache_count)
3831 tcache_put (victim, tc_idx);
3832 return_cached = 1;
3833 continue;
3835 else
3837 #endif
3838 check_malloced_chunk (av, victim, nb);
3839 void *p = chunk2mem (victim);
3840 alloc_perturb (p, bytes);
3841 return p;
3842 #if USE_TCACHE
3844 #endif
3847 /* place chunk in bin */
3849 if (in_smallbin_range (size))
3851 victim_index = smallbin_index (size);
3852 bck = bin_at (av, victim_index);
3853 fwd = bck->fd;
3855 else
3857 victim_index = largebin_index (size);
3858 bck = bin_at (av, victim_index);
3859 fwd = bck->fd;
3861 /* maintain large bins in sorted order */
3862 if (fwd != bck)
3864 /* Or with inuse bit to speed comparisons */
3865 size |= PREV_INUSE;
3866 /* if smaller than smallest, bypass loop below */
3867 assert (chunk_main_arena (bck->bk));
3868 if ((unsigned long) (size)
3869 < (unsigned long) chunksize_nomask (bck->bk))
3871 fwd = bck;
3872 bck = bck->bk;
3874 victim->fd_nextsize = fwd->fd;
3875 victim->bk_nextsize = fwd->fd->bk_nextsize;
3876 fwd->fd->bk_nextsize = victim->bk_nextsize->fd_nextsize = victim;
3878 else
3880 assert (chunk_main_arena (fwd));
3881 while ((unsigned long) size < chunksize_nomask (fwd))
3883 fwd = fwd->fd_nextsize;
3884 assert (chunk_main_arena (fwd));
3887 if ((unsigned long) size
3888 == (unsigned long) chunksize_nomask (fwd))
3889 /* Always insert in the second position. */
3890 fwd = fwd->fd;
3891 else
3893 victim->fd_nextsize = fwd;
3894 victim->bk_nextsize = fwd->bk_nextsize;
3895 if (__glibc_unlikely (fwd->bk_nextsize->fd_nextsize != fwd))
3896 malloc_printerr ("malloc(): largebin double linked list corrupted (nextsize)");
3897 fwd->bk_nextsize = victim;
3898 victim->bk_nextsize->fd_nextsize = victim;
3900 bck = fwd->bk;
3901 if (bck->fd != fwd)
3902 malloc_printerr ("malloc(): largebin double linked list corrupted (bk)");
3905 else
3906 victim->fd_nextsize = victim->bk_nextsize = victim;
3909 mark_bin (av, victim_index);
3910 victim->bk = bck;
3911 victim->fd = fwd;
3912 fwd->bk = victim;
3913 bck->fd = victim;
3915 #if USE_TCACHE
3916 /* If we've processed as many chunks as we're allowed while
3917 filling the cache, return one of the cached ones. */
3918 ++tcache_unsorted_count;
3919 if (return_cached
3920 && mp_.tcache_unsorted_limit > 0
3921 && tcache_unsorted_count > mp_.tcache_unsorted_limit)
3923 return tcache_get (tc_idx);
3925 #endif
3927 #define MAX_ITERS 10000
3928 if (++iters >= MAX_ITERS)
3929 break;
3932 #if USE_TCACHE
3933 /* If all the small chunks we found ended up cached, return one now. */
3934 if (return_cached)
3936 return tcache_get (tc_idx);
3938 #endif
3941 If a large request, scan through the chunks of current bin in
3942 sorted order to find smallest that fits. Use the skip list for this.
3945 if (!in_smallbin_range (nb))
3947 bin = bin_at (av, idx);
3949 /* skip scan if empty or largest chunk is too small */
3950 if ((victim = first (bin)) != bin
3951 && (unsigned long) chunksize_nomask (victim)
3952 >= (unsigned long) (nb))
3954 victim = victim->bk_nextsize;
3955 while (((unsigned long) (size = chunksize (victim)) <
3956 (unsigned long) (nb)))
3957 victim = victim->bk_nextsize;
3959 /* Avoid removing the first entry for a size so that the skip
3960 list does not have to be rerouted. */
3961 if (victim != last (bin)
3962 && chunksize_nomask (victim)
3963 == chunksize_nomask (victim->fd))
3964 victim = victim->fd;
3966 remainder_size = size - nb;
3967 unlink_chunk (av, victim);
3969 /* Exhaust */
3970 if (remainder_size < MINSIZE)
3972 set_inuse_bit_at_offset (victim, size);
3973 if (av != &main_arena)
3974 set_non_main_arena (victim);
3976 /* Split */
3977 else
3979 remainder = chunk_at_offset (victim, nb);
3980 /* We cannot assume the unsorted list is empty and therefore
3981 have to perform a complete insert here. */
3982 bck = unsorted_chunks (av);
3983 fwd = bck->fd;
3984 if (__glibc_unlikely (fwd->bk != bck))
3985 malloc_printerr ("malloc(): corrupted unsorted chunks");
3986 remainder->bk = bck;
3987 remainder->fd = fwd;
3988 bck->fd = remainder;
3989 fwd->bk = remainder;
3990 if (!in_smallbin_range (remainder_size))
3992 remainder->fd_nextsize = NULL;
3993 remainder->bk_nextsize = NULL;
3995 set_head (victim, nb | PREV_INUSE |
3996 (av != &main_arena ? NON_MAIN_ARENA : 0));
3997 set_head (remainder, remainder_size | PREV_INUSE);
3998 set_foot (remainder, remainder_size);
4000 check_malloced_chunk (av, victim, nb);
4001 void *p = chunk2mem (victim);
4002 alloc_perturb (p, bytes);
4003 return p;
4008 Search for a chunk by scanning bins, starting with next largest
4009 bin. This search is strictly by best-fit; i.e., the smallest
4010 (with ties going to approximately the least recently used) chunk
4011 that fits is selected.
4013 The bitmap avoids needing to check that most blocks are nonempty.
4014 The particular case of skipping all bins during warm-up phases
4015 when no chunks have been returned yet is faster than it might look.
4018 ++idx;
4019 bin = bin_at (av, idx);
4020 block = idx2block (idx);
4021 map = av->binmap[block];
4022 bit = idx2bit (idx);
4024 for (;; )
4026 /* Skip rest of block if there are no more set bits in this block. */
4027 if (bit > map || bit == 0)
4031 if (++block >= BINMAPSIZE) /* out of bins */
4032 goto use_top;
4034 while ((map = av->binmap[block]) == 0);
4036 bin = bin_at (av, (block << BINMAPSHIFT));
4037 bit = 1;
4040 /* Advance to bin with set bit. There must be one. */
4041 while ((bit & map) == 0)
4043 bin = next_bin (bin);
4044 bit <<= 1;
4045 assert (bit != 0);
4048 /* Inspect the bin. It is likely to be non-empty */
4049 victim = last (bin);
4051 /* If a false alarm (empty bin), clear the bit. */
4052 if (victim == bin)
4054 av->binmap[block] = map &= ~bit; /* Write through */
4055 bin = next_bin (bin);
4056 bit <<= 1;
4059 else
4061 size = chunksize (victim);
4063 /* We know the first chunk in this bin is big enough to use. */
4064 assert ((unsigned long) (size) >= (unsigned long) (nb));
4066 remainder_size = size - nb;
4068 /* unlink */
4069 unlink_chunk (av, victim);
4071 /* Exhaust */
4072 if (remainder_size < MINSIZE)
4074 set_inuse_bit_at_offset (victim, size);
4075 if (av != &main_arena)
4076 set_non_main_arena (victim);
4079 /* Split */
4080 else
4082 remainder = chunk_at_offset (victim, nb);
4084 /* We cannot assume the unsorted list is empty and therefore
4085 have to perform a complete insert here. */
4086 bck = unsorted_chunks (av);
4087 fwd = bck->fd;
4088 if (__glibc_unlikely (fwd->bk != bck))
4089 malloc_printerr ("malloc(): corrupted unsorted chunks 2");
4090 remainder->bk = bck;
4091 remainder->fd = fwd;
4092 bck->fd = remainder;
4093 fwd->bk = remainder;
4095 /* advertise as last remainder */
4096 if (in_smallbin_range (nb))
4097 av->last_remainder = remainder;
4098 if (!in_smallbin_range (remainder_size))
4100 remainder->fd_nextsize = NULL;
4101 remainder->bk_nextsize = NULL;
4103 set_head (victim, nb | PREV_INUSE |
4104 (av != &main_arena ? NON_MAIN_ARENA : 0));
4105 set_head (remainder, remainder_size | PREV_INUSE);
4106 set_foot (remainder, remainder_size);
4108 check_malloced_chunk (av, victim, nb);
4109 void *p = chunk2mem (victim);
4110 alloc_perturb (p, bytes);
4111 return p;
4115 use_top:
4117 If large enough, split off the chunk bordering the end of memory
4118 (held in av->top). Note that this is in accord with the best-fit
4119 search rule. In effect, av->top is treated as larger (and thus
4120 less well fitting) than any other available chunk since it can
4121 be extended to be as large as necessary (up to system
4122 limitations).
4124 We require that av->top always exists (i.e., has size >=
4125 MINSIZE) after initialization, so if it would otherwise be
4126 exhausted by current request, it is replenished. (The main
4127 reason for ensuring it exists is that we may need MINSIZE space
4128 to put in fenceposts in sysmalloc.)
4131 victim = av->top;
4132 size = chunksize (victim);
4134 if (__glibc_unlikely (size > av->system_mem))
4135 malloc_printerr ("malloc(): corrupted top size");
4137 if ((unsigned long) (size) >= (unsigned long) (nb + MINSIZE))
4139 remainder_size = size - nb;
4140 remainder = chunk_at_offset (victim, nb);
4141 av->top = remainder;
4142 set_head (victim, nb | PREV_INUSE |
4143 (av != &main_arena ? NON_MAIN_ARENA : 0));
4144 set_head (remainder, remainder_size | PREV_INUSE);
4146 check_malloced_chunk (av, victim, nb);
4147 void *p = chunk2mem (victim);
4148 alloc_perturb (p, bytes);
4149 return p;
4152 /* When we are using atomic ops to free fast chunks we can get
4153 here for all block sizes. */
4154 else if (atomic_load_relaxed (&av->have_fastchunks))
4156 malloc_consolidate (av);
4157 /* restore original bin index */
4158 if (in_smallbin_range (nb))
4159 idx = smallbin_index (nb);
4160 else
4161 idx = largebin_index (nb);
4165 Otherwise, relay to handle system-dependent cases
4167 else
4169 void *p = sysmalloc (nb, av);
4170 if (p != NULL)
4171 alloc_perturb (p, bytes);
4172 return p;
4178 ------------------------------ free ------------------------------
4181 static void
4182 _int_free (mstate av, mchunkptr p, int have_lock)
4184 INTERNAL_SIZE_T size; /* its size */
4185 mfastbinptr *fb; /* associated fastbin */
4186 mchunkptr nextchunk; /* next contiguous chunk */
4187 INTERNAL_SIZE_T nextsize; /* its size */
4188 int nextinuse; /* true if nextchunk is used */
4189 INTERNAL_SIZE_T prevsize; /* size of previous contiguous chunk */
4190 mchunkptr bck; /* misc temp for linking */
4191 mchunkptr fwd; /* misc temp for linking */
4193 size = chunksize (p);
4195 /* Little security check which won't hurt performance: the
4196 allocator never wrapps around at the end of the address space.
4197 Therefore we can exclude some size values which might appear
4198 here by accident or by "design" from some intruder. */
4199 if (__builtin_expect ((uintptr_t) p > (uintptr_t) -size, 0)
4200 || __builtin_expect (misaligned_chunk (p), 0))
4201 malloc_printerr ("free(): invalid pointer");
4202 /* We know that each chunk is at least MINSIZE bytes in size or a
4203 multiple of MALLOC_ALIGNMENT. */
4204 if (__glibc_unlikely (size < MINSIZE || !aligned_OK (size)))
4205 malloc_printerr ("free(): invalid size");
4207 check_inuse_chunk(av, p);
4209 #if USE_TCACHE
4211 size_t tc_idx = csize2tidx (size);
4212 if (tcache != NULL && tc_idx < mp_.tcache_bins)
4214 /* Check to see if it's already in the tcache. */
4215 tcache_entry *e = (tcache_entry *) chunk2mem (p);
4217 /* This test succeeds on double free. However, we don't 100%
4218 trust it (it also matches random payload data at a 1 in
4219 2^<size_t> chance), so verify it's not an unlikely
4220 coincidence before aborting. */
4221 if (__glibc_unlikely (e->key == tcache))
4223 tcache_entry *tmp;
4224 LIBC_PROBE (memory_tcache_double_free, 2, e, tc_idx);
4225 for (tmp = tcache->entries[tc_idx];
4226 tmp;
4227 tmp = REVEAL_PTR (tmp->next))
4229 if (__glibc_unlikely (!aligned_OK (tmp)))
4230 malloc_printerr ("free(): unaligned chunk detected in tcache 2");
4231 if (tmp == e)
4232 malloc_printerr ("free(): double free detected in tcache 2");
4233 /* If we get here, it was a coincidence. We've wasted a
4234 few cycles, but don't abort. */
4238 if (tcache->counts[tc_idx] < mp_.tcache_count)
4240 tcache_put (p, tc_idx);
4241 return;
4245 #endif
4248 If eligible, place chunk on a fastbin so it can be found
4249 and used quickly in malloc.
4252 if ((unsigned long)(size) <= (unsigned long)(get_max_fast ())
4254 #if TRIM_FASTBINS
4256 If TRIM_FASTBINS set, don't place chunks
4257 bordering top into fastbins
4259 && (chunk_at_offset(p, size) != av->top)
4260 #endif
4263 if (__builtin_expect (chunksize_nomask (chunk_at_offset (p, size))
4264 <= 2 * SIZE_SZ, 0)
4265 || __builtin_expect (chunksize (chunk_at_offset (p, size))
4266 >= av->system_mem, 0))
4268 bool fail = true;
4269 /* We might not have a lock at this point and concurrent modifications
4270 of system_mem might result in a false positive. Redo the test after
4271 getting the lock. */
4272 if (!have_lock)
4274 __libc_lock_lock (av->mutex);
4275 fail = (chunksize_nomask (chunk_at_offset (p, size)) <= 2 * SIZE_SZ
4276 || chunksize (chunk_at_offset (p, size)) >= av->system_mem);
4277 __libc_lock_unlock (av->mutex);
4280 if (fail)
4281 malloc_printerr ("free(): invalid next size (fast)");
4284 free_perturb (chunk2mem(p), size - 2 * SIZE_SZ);
4286 atomic_store_relaxed (&av->have_fastchunks, true);
4287 unsigned int idx = fastbin_index(size);
4288 fb = &fastbin (av, idx);
4290 /* Atomically link P to its fastbin: P->FD = *FB; *FB = P; */
4291 mchunkptr old = *fb, old2;
4293 if (SINGLE_THREAD_P)
4295 /* Check that the top of the bin is not the record we are going to
4296 add (i.e., double free). */
4297 if (__builtin_expect (old == p, 0))
4298 malloc_printerr ("double free or corruption (fasttop)");
4299 p->fd = PROTECT_PTR (&p->fd, old);
4300 *fb = p;
4302 else
4305 /* Check that the top of the bin is not the record we are going to
4306 add (i.e., double free). */
4307 if (__builtin_expect (old == p, 0))
4308 malloc_printerr ("double free or corruption (fasttop)");
4309 old2 = old;
4310 p->fd = PROTECT_PTR (&p->fd, old);
4312 while ((old = catomic_compare_and_exchange_val_rel (fb, p, old2))
4313 != old2);
4315 /* Check that size of fastbin chunk at the top is the same as
4316 size of the chunk that we are adding. We can dereference OLD
4317 only if we have the lock, otherwise it might have already been
4318 allocated again. */
4319 if (have_lock && old != NULL
4320 && __builtin_expect (fastbin_index (chunksize (old)) != idx, 0))
4321 malloc_printerr ("invalid fastbin entry (free)");
4325 Consolidate other non-mmapped chunks as they arrive.
4328 else if (!chunk_is_mmapped(p)) {
4330 /* If we're single-threaded, don't lock the arena. */
4331 if (SINGLE_THREAD_P)
4332 have_lock = true;
4334 if (!have_lock)
4335 __libc_lock_lock (av->mutex);
4337 nextchunk = chunk_at_offset(p, size);
4339 /* Lightweight tests: check whether the block is already the
4340 top block. */
4341 if (__glibc_unlikely (p == av->top))
4342 malloc_printerr ("double free or corruption (top)");
4343 /* Or whether the next chunk is beyond the boundaries of the arena. */
4344 if (__builtin_expect (contiguous (av)
4345 && (char *) nextchunk
4346 >= ((char *) av->top + chunksize(av->top)), 0))
4347 malloc_printerr ("double free or corruption (out)");
4348 /* Or whether the block is actually not marked used. */
4349 if (__glibc_unlikely (!prev_inuse(nextchunk)))
4350 malloc_printerr ("double free or corruption (!prev)");
4352 nextsize = chunksize(nextchunk);
4353 if (__builtin_expect (chunksize_nomask (nextchunk) <= 2 * SIZE_SZ, 0)
4354 || __builtin_expect (nextsize >= av->system_mem, 0))
4355 malloc_printerr ("free(): invalid next size (normal)");
4357 free_perturb (chunk2mem(p), size - 2 * SIZE_SZ);
4359 /* consolidate backward */
4360 if (!prev_inuse(p)) {
4361 prevsize = prev_size (p);
4362 size += prevsize;
4363 p = chunk_at_offset(p, -((long) prevsize));
4364 if (__glibc_unlikely (chunksize(p) != prevsize))
4365 malloc_printerr ("corrupted size vs. prev_size while consolidating");
4366 unlink_chunk (av, p);
4369 if (nextchunk != av->top) {
4370 /* get and clear inuse bit */
4371 nextinuse = inuse_bit_at_offset(nextchunk, nextsize);
4373 /* consolidate forward */
4374 if (!nextinuse) {
4375 unlink_chunk (av, nextchunk);
4376 size += nextsize;
4377 } else
4378 clear_inuse_bit_at_offset(nextchunk, 0);
4381 Place the chunk in unsorted chunk list. Chunks are
4382 not placed into regular bins until after they have
4383 been given one chance to be used in malloc.
4386 bck = unsorted_chunks(av);
4387 fwd = bck->fd;
4388 if (__glibc_unlikely (fwd->bk != bck))
4389 malloc_printerr ("free(): corrupted unsorted chunks");
4390 p->fd = fwd;
4391 p->bk = bck;
4392 if (!in_smallbin_range(size))
4394 p->fd_nextsize = NULL;
4395 p->bk_nextsize = NULL;
4397 bck->fd = p;
4398 fwd->bk = p;
4400 set_head(p, size | PREV_INUSE);
4401 set_foot(p, size);
4403 check_free_chunk(av, p);
4407 If the chunk borders the current high end of memory,
4408 consolidate into top
4411 else {
4412 size += nextsize;
4413 set_head(p, size | PREV_INUSE);
4414 av->top = p;
4415 check_chunk(av, p);
4419 If freeing a large space, consolidate possibly-surrounding
4420 chunks. Then, if the total unused topmost memory exceeds trim
4421 threshold, ask malloc_trim to reduce top.
4423 Unless max_fast is 0, we don't know if there are fastbins
4424 bordering top, so we cannot tell for sure whether threshold
4425 has been reached unless fastbins are consolidated. But we
4426 don't want to consolidate on each free. As a compromise,
4427 consolidation is performed if FASTBIN_CONSOLIDATION_THRESHOLD
4428 is reached.
4431 if ((unsigned long)(size) >= FASTBIN_CONSOLIDATION_THRESHOLD) {
4432 if (atomic_load_relaxed (&av->have_fastchunks))
4433 malloc_consolidate(av);
4435 if (av == &main_arena) {
4436 #ifndef MORECORE_CANNOT_TRIM
4437 if ((unsigned long)(chunksize(av->top)) >=
4438 (unsigned long)(mp_.trim_threshold))
4439 systrim(mp_.top_pad, av);
4440 #endif
4441 } else {
4442 /* Always try heap_trim(), even if the top chunk is not
4443 large, because the corresponding heap might go away. */
4444 heap_info *heap = heap_for_ptr(top(av));
4446 assert(heap->ar_ptr == av);
4447 heap_trim(heap, mp_.top_pad);
4451 if (!have_lock)
4452 __libc_lock_unlock (av->mutex);
4455 If the chunk was allocated via mmap, release via munmap().
4458 else {
4459 munmap_chunk (p);
4464 ------------------------- malloc_consolidate -------------------------
4466 malloc_consolidate is a specialized version of free() that tears
4467 down chunks held in fastbins. Free itself cannot be used for this
4468 purpose since, among other things, it might place chunks back onto
4469 fastbins. So, instead, we need to use a minor variant of the same
4470 code.
4473 static void malloc_consolidate(mstate av)
4475 mfastbinptr* fb; /* current fastbin being consolidated */
4476 mfastbinptr* maxfb; /* last fastbin (for loop control) */
4477 mchunkptr p; /* current chunk being consolidated */
4478 mchunkptr nextp; /* next chunk to consolidate */
4479 mchunkptr unsorted_bin; /* bin header */
4480 mchunkptr first_unsorted; /* chunk to link to */
4482 /* These have same use as in free() */
4483 mchunkptr nextchunk;
4484 INTERNAL_SIZE_T size;
4485 INTERNAL_SIZE_T nextsize;
4486 INTERNAL_SIZE_T prevsize;
4487 int nextinuse;
4489 atomic_store_relaxed (&av->have_fastchunks, false);
4491 unsorted_bin = unsorted_chunks(av);
4494 Remove each chunk from fast bin and consolidate it, placing it
4495 then in unsorted bin. Among other reasons for doing this,
4496 placing in unsorted bin avoids needing to calculate actual bins
4497 until malloc is sure that chunks aren't immediately going to be
4498 reused anyway.
4501 maxfb = &fastbin (av, NFASTBINS - 1);
4502 fb = &fastbin (av, 0);
4503 do {
4504 p = atomic_exchange_acq (fb, NULL);
4505 if (p != 0) {
4506 do {
4508 if (__glibc_unlikely (misaligned_chunk (p)))
4509 malloc_printerr ("malloc_consolidate(): "
4510 "unaligned fastbin chunk detected");
4512 unsigned int idx = fastbin_index (chunksize (p));
4513 if ((&fastbin (av, idx)) != fb)
4514 malloc_printerr ("malloc_consolidate(): invalid chunk size");
4517 check_inuse_chunk(av, p);
4518 nextp = REVEAL_PTR (p->fd);
4520 /* Slightly streamlined version of consolidation code in free() */
4521 size = chunksize (p);
4522 nextchunk = chunk_at_offset(p, size);
4523 nextsize = chunksize(nextchunk);
4525 if (!prev_inuse(p)) {
4526 prevsize = prev_size (p);
4527 size += prevsize;
4528 p = chunk_at_offset(p, -((long) prevsize));
4529 if (__glibc_unlikely (chunksize(p) != prevsize))
4530 malloc_printerr ("corrupted size vs. prev_size in fastbins");
4531 unlink_chunk (av, p);
4534 if (nextchunk != av->top) {
4535 nextinuse = inuse_bit_at_offset(nextchunk, nextsize);
4537 if (!nextinuse) {
4538 size += nextsize;
4539 unlink_chunk (av, nextchunk);
4540 } else
4541 clear_inuse_bit_at_offset(nextchunk, 0);
4543 first_unsorted = unsorted_bin->fd;
4544 unsorted_bin->fd = p;
4545 first_unsorted->bk = p;
4547 if (!in_smallbin_range (size)) {
4548 p->fd_nextsize = NULL;
4549 p->bk_nextsize = NULL;
4552 set_head(p, size | PREV_INUSE);
4553 p->bk = unsorted_bin;
4554 p->fd = first_unsorted;
4555 set_foot(p, size);
4558 else {
4559 size += nextsize;
4560 set_head(p, size | PREV_INUSE);
4561 av->top = p;
4564 } while ( (p = nextp) != 0);
4567 } while (fb++ != maxfb);
4571 ------------------------------ realloc ------------------------------
4574 void*
4575 _int_realloc(mstate av, mchunkptr oldp, INTERNAL_SIZE_T oldsize,
4576 INTERNAL_SIZE_T nb)
4578 mchunkptr newp; /* chunk to return */
4579 INTERNAL_SIZE_T newsize; /* its size */
4580 void* newmem; /* corresponding user mem */
4582 mchunkptr next; /* next contiguous chunk after oldp */
4584 mchunkptr remainder; /* extra space at end of newp */
4585 unsigned long remainder_size; /* its size */
4587 /* oldmem size */
4588 if (__builtin_expect (chunksize_nomask (oldp) <= 2 * SIZE_SZ, 0)
4589 || __builtin_expect (oldsize >= av->system_mem, 0))
4590 malloc_printerr ("realloc(): invalid old size");
4592 check_inuse_chunk (av, oldp);
4594 /* All callers already filter out mmap'ed chunks. */
4595 assert (!chunk_is_mmapped (oldp));
4597 next = chunk_at_offset (oldp, oldsize);
4598 INTERNAL_SIZE_T nextsize = chunksize (next);
4599 if (__builtin_expect (chunksize_nomask (next) <= 2 * SIZE_SZ, 0)
4600 || __builtin_expect (nextsize >= av->system_mem, 0))
4601 malloc_printerr ("realloc(): invalid next size");
4603 if ((unsigned long) (oldsize) >= (unsigned long) (nb))
4605 /* already big enough; split below */
4606 newp = oldp;
4607 newsize = oldsize;
4610 else
4612 /* Try to expand forward into top */
4613 if (next == av->top &&
4614 (unsigned long) (newsize = oldsize + nextsize) >=
4615 (unsigned long) (nb + MINSIZE))
4617 set_head_size (oldp, nb | (av != &main_arena ? NON_MAIN_ARENA : 0));
4618 av->top = chunk_at_offset (oldp, nb);
4619 set_head (av->top, (newsize - nb) | PREV_INUSE);
4620 check_inuse_chunk (av, oldp);
4621 return chunk2mem (oldp);
4624 /* Try to expand forward into next chunk; split off remainder below */
4625 else if (next != av->top &&
4626 !inuse (next) &&
4627 (unsigned long) (newsize = oldsize + nextsize) >=
4628 (unsigned long) (nb))
4630 newp = oldp;
4631 unlink_chunk (av, next);
4634 /* allocate, copy, free */
4635 else
4637 newmem = _int_malloc (av, nb - MALLOC_ALIGN_MASK);
4638 if (newmem == 0)
4639 return 0; /* propagate failure */
4641 newp = mem2chunk (newmem);
4642 newsize = chunksize (newp);
4645 Avoid copy if newp is next chunk after oldp.
4647 if (newp == next)
4649 newsize += oldsize;
4650 newp = oldp;
4652 else
4654 memcpy (newmem, chunk2mem (oldp), oldsize - SIZE_SZ);
4655 _int_free (av, oldp, 1);
4656 check_inuse_chunk (av, newp);
4657 return chunk2mem (newp);
4662 /* If possible, free extra space in old or extended chunk */
4664 assert ((unsigned long) (newsize) >= (unsigned long) (nb));
4666 remainder_size = newsize - nb;
4668 if (remainder_size < MINSIZE) /* not enough extra to split off */
4670 set_head_size (newp, newsize | (av != &main_arena ? NON_MAIN_ARENA : 0));
4671 set_inuse_bit_at_offset (newp, newsize);
4673 else /* split remainder */
4675 remainder = chunk_at_offset (newp, nb);
4676 set_head_size (newp, nb | (av != &main_arena ? NON_MAIN_ARENA : 0));
4677 set_head (remainder, remainder_size | PREV_INUSE |
4678 (av != &main_arena ? NON_MAIN_ARENA : 0));
4679 /* Mark remainder as inuse so free() won't complain */
4680 set_inuse_bit_at_offset (remainder, remainder_size);
4681 _int_free (av, remainder, 1);
4684 check_inuse_chunk (av, newp);
4685 return chunk2mem (newp);
4689 ------------------------------ memalign ------------------------------
4692 static void *
4693 _int_memalign (mstate av, size_t alignment, size_t bytes)
4695 INTERNAL_SIZE_T nb; /* padded request size */
4696 char *m; /* memory returned by malloc call */
4697 mchunkptr p; /* corresponding chunk */
4698 char *brk; /* alignment point within p */
4699 mchunkptr newp; /* chunk to return */
4700 INTERNAL_SIZE_T newsize; /* its size */
4701 INTERNAL_SIZE_T leadsize; /* leading space before alignment point */
4702 mchunkptr remainder; /* spare room at end to split off */
4703 unsigned long remainder_size; /* its size */
4704 INTERNAL_SIZE_T size;
4708 if (!checked_request2size (bytes, &nb))
4710 __set_errno (ENOMEM);
4711 return NULL;
4715 Strategy: find a spot within that chunk that meets the alignment
4716 request, and then possibly free the leading and trailing space.
4719 /* Call malloc with worst case padding to hit alignment. */
4721 m = (char *) (_int_malloc (av, nb + alignment + MINSIZE));
4723 if (m == 0)
4724 return 0; /* propagate failure */
4726 p = mem2chunk (m);
4728 if ((((unsigned long) (m)) % alignment) != 0) /* misaligned */
4730 { /*
4731 Find an aligned spot inside chunk. Since we need to give back
4732 leading space in a chunk of at least MINSIZE, if the first
4733 calculation places us at a spot with less than MINSIZE leader,
4734 we can move to the next aligned spot -- we've allocated enough
4735 total room so that this is always possible.
4737 brk = (char *) mem2chunk (((unsigned long) (m + alignment - 1)) &
4738 - ((signed long) alignment));
4739 if ((unsigned long) (brk - (char *) (p)) < MINSIZE)
4740 brk += alignment;
4742 newp = (mchunkptr) brk;
4743 leadsize = brk - (char *) (p);
4744 newsize = chunksize (p) - leadsize;
4746 /* For mmapped chunks, just adjust offset */
4747 if (chunk_is_mmapped (p))
4749 set_prev_size (newp, prev_size (p) + leadsize);
4750 set_head (newp, newsize | IS_MMAPPED);
4751 return chunk2mem (newp);
4754 /* Otherwise, give back leader, use the rest */
4755 set_head (newp, newsize | PREV_INUSE |
4756 (av != &main_arena ? NON_MAIN_ARENA : 0));
4757 set_inuse_bit_at_offset (newp, newsize);
4758 set_head_size (p, leadsize | (av != &main_arena ? NON_MAIN_ARENA : 0));
4759 _int_free (av, p, 1);
4760 p = newp;
4762 assert (newsize >= nb &&
4763 (((unsigned long) (chunk2mem (p))) % alignment) == 0);
4766 /* Also give back spare room at the end */
4767 if (!chunk_is_mmapped (p))
4769 size = chunksize (p);
4770 if ((unsigned long) (size) > (unsigned long) (nb + MINSIZE))
4772 remainder_size = size - nb;
4773 remainder = chunk_at_offset (p, nb);
4774 set_head (remainder, remainder_size | PREV_INUSE |
4775 (av != &main_arena ? NON_MAIN_ARENA : 0));
4776 set_head_size (p, nb);
4777 _int_free (av, remainder, 1);
4781 check_inuse_chunk (av, p);
4782 return chunk2mem (p);
4787 ------------------------------ malloc_trim ------------------------------
4790 static int
4791 mtrim (mstate av, size_t pad)
4793 /* Ensure all blocks are consolidated. */
4794 malloc_consolidate (av);
4796 const size_t ps = GLRO (dl_pagesize);
4797 int psindex = bin_index (ps);
4798 const size_t psm1 = ps - 1;
4800 int result = 0;
4801 for (int i = 1; i < NBINS; ++i)
4802 if (i == 1 || i >= psindex)
4804 mbinptr bin = bin_at (av, i);
4806 for (mchunkptr p = last (bin); p != bin; p = p->bk)
4808 INTERNAL_SIZE_T size = chunksize (p);
4810 if (size > psm1 + sizeof (struct malloc_chunk))
4812 /* See whether the chunk contains at least one unused page. */
4813 char *paligned_mem = (char *) (((uintptr_t) p
4814 + sizeof (struct malloc_chunk)
4815 + psm1) & ~psm1);
4817 assert ((char *) chunk2mem (p) + 4 * SIZE_SZ <= paligned_mem);
4818 assert ((char *) p + size > paligned_mem);
4820 /* This is the size we could potentially free. */
4821 size -= paligned_mem - (char *) p;
4823 if (size > psm1)
4825 #if MALLOC_DEBUG
4826 /* When debugging we simulate destroying the memory
4827 content. */
4828 memset (paligned_mem, 0x89, size & ~psm1);
4829 #endif
4830 __madvise (paligned_mem, size & ~psm1, MADV_DONTNEED);
4832 result = 1;
4838 #ifndef MORECORE_CANNOT_TRIM
4839 return result | (av == &main_arena ? systrim (pad, av) : 0);
4841 #else
4842 return result;
4843 #endif
4848 __malloc_trim (size_t s)
4850 int result = 0;
4852 if (__malloc_initialized < 0)
4853 ptmalloc_init ();
4855 mstate ar_ptr = &main_arena;
4858 __libc_lock_lock (ar_ptr->mutex);
4859 result |= mtrim (ar_ptr, s);
4860 __libc_lock_unlock (ar_ptr->mutex);
4862 ar_ptr = ar_ptr->next;
4864 while (ar_ptr != &main_arena);
4866 return result;
4871 ------------------------- malloc_usable_size -------------------------
4874 static size_t
4875 musable (void *mem)
4877 mchunkptr p;
4878 if (mem != 0)
4880 p = mem2chunk (mem);
4882 if (__builtin_expect (using_malloc_checking == 1, 0))
4883 return malloc_check_get_size (p);
4885 if (chunk_is_mmapped (p))
4887 if (DUMPED_MAIN_ARENA_CHUNK (p))
4888 return chunksize (p) - SIZE_SZ;
4889 else
4890 return chunksize (p) - 2 * SIZE_SZ;
4892 else if (inuse (p))
4893 return chunksize (p) - SIZE_SZ;
4895 return 0;
4899 size_t
4900 __malloc_usable_size (void *m)
4902 size_t result;
4904 result = musable (m);
4905 return result;
4909 ------------------------------ mallinfo ------------------------------
4910 Accumulate malloc statistics for arena AV into M.
4913 static void
4914 int_mallinfo (mstate av, struct mallinfo *m)
4916 size_t i;
4917 mbinptr b;
4918 mchunkptr p;
4919 INTERNAL_SIZE_T avail;
4920 INTERNAL_SIZE_T fastavail;
4921 int nblocks;
4922 int nfastblocks;
4924 check_malloc_state (av);
4926 /* Account for top */
4927 avail = chunksize (av->top);
4928 nblocks = 1; /* top always exists */
4930 /* traverse fastbins */
4931 nfastblocks = 0;
4932 fastavail = 0;
4934 for (i = 0; i < NFASTBINS; ++i)
4936 for (p = fastbin (av, i);
4937 p != 0;
4938 p = REVEAL_PTR (p->fd))
4940 if (__glibc_unlikely (misaligned_chunk (p)))
4941 malloc_printerr ("int_mallinfo(): "
4942 "unaligned fastbin chunk detected");
4943 ++nfastblocks;
4944 fastavail += chunksize (p);
4948 avail += fastavail;
4950 /* traverse regular bins */
4951 for (i = 1; i < NBINS; ++i)
4953 b = bin_at (av, i);
4954 for (p = last (b); p != b; p = p->bk)
4956 ++nblocks;
4957 avail += chunksize (p);
4961 m->smblks += nfastblocks;
4962 m->ordblks += nblocks;
4963 m->fordblks += avail;
4964 m->uordblks += av->system_mem - avail;
4965 m->arena += av->system_mem;
4966 m->fsmblks += fastavail;
4967 if (av == &main_arena)
4969 m->hblks = mp_.n_mmaps;
4970 m->hblkhd = mp_.mmapped_mem;
4971 m->usmblks = 0;
4972 m->keepcost = chunksize (av->top);
4977 struct mallinfo
4978 __libc_mallinfo (void)
4980 struct mallinfo m;
4981 mstate ar_ptr;
4983 if (__malloc_initialized < 0)
4984 ptmalloc_init ();
4986 memset (&m, 0, sizeof (m));
4987 ar_ptr = &main_arena;
4990 __libc_lock_lock (ar_ptr->mutex);
4991 int_mallinfo (ar_ptr, &m);
4992 __libc_lock_unlock (ar_ptr->mutex);
4994 ar_ptr = ar_ptr->next;
4996 while (ar_ptr != &main_arena);
4998 return m;
5002 ------------------------------ malloc_stats ------------------------------
5005 void
5006 __malloc_stats (void)
5008 int i;
5009 mstate ar_ptr;
5010 unsigned int in_use_b = mp_.mmapped_mem, system_b = in_use_b;
5012 if (__malloc_initialized < 0)
5013 ptmalloc_init ();
5014 _IO_flockfile (stderr);
5015 int old_flags2 = stderr->_flags2;
5016 stderr->_flags2 |= _IO_FLAGS2_NOTCANCEL;
5017 for (i = 0, ar_ptr = &main_arena;; i++)
5019 struct mallinfo mi;
5021 memset (&mi, 0, sizeof (mi));
5022 __libc_lock_lock (ar_ptr->mutex);
5023 int_mallinfo (ar_ptr, &mi);
5024 fprintf (stderr, "Arena %d:\n", i);
5025 fprintf (stderr, "system bytes = %10u\n", (unsigned int) mi.arena);
5026 fprintf (stderr, "in use bytes = %10u\n", (unsigned int) mi.uordblks);
5027 #if MALLOC_DEBUG > 1
5028 if (i > 0)
5029 dump_heap (heap_for_ptr (top (ar_ptr)));
5030 #endif
5031 system_b += mi.arena;
5032 in_use_b += mi.uordblks;
5033 __libc_lock_unlock (ar_ptr->mutex);
5034 ar_ptr = ar_ptr->next;
5035 if (ar_ptr == &main_arena)
5036 break;
5038 fprintf (stderr, "Total (incl. mmap):\n");
5039 fprintf (stderr, "system bytes = %10u\n", system_b);
5040 fprintf (stderr, "in use bytes = %10u\n", in_use_b);
5041 fprintf (stderr, "max mmap regions = %10u\n", (unsigned int) mp_.max_n_mmaps);
5042 fprintf (stderr, "max mmap bytes = %10lu\n",
5043 (unsigned long) mp_.max_mmapped_mem);
5044 stderr->_flags2 = old_flags2;
5045 _IO_funlockfile (stderr);
5050 ------------------------------ mallopt ------------------------------
5052 static __always_inline int
5053 do_set_trim_threshold (size_t value)
5055 LIBC_PROBE (memory_mallopt_trim_threshold, 3, value, mp_.trim_threshold,
5056 mp_.no_dyn_threshold);
5057 mp_.trim_threshold = value;
5058 mp_.no_dyn_threshold = 1;
5059 return 1;
5062 static __always_inline int
5063 do_set_top_pad (size_t value)
5065 LIBC_PROBE (memory_mallopt_top_pad, 3, value, mp_.top_pad,
5066 mp_.no_dyn_threshold);
5067 mp_.top_pad = value;
5068 mp_.no_dyn_threshold = 1;
5069 return 1;
5072 static __always_inline int
5073 do_set_mmap_threshold (size_t value)
5075 /* Forbid setting the threshold too high. */
5076 if (value <= HEAP_MAX_SIZE / 2)
5078 LIBC_PROBE (memory_mallopt_mmap_threshold, 3, value, mp_.mmap_threshold,
5079 mp_.no_dyn_threshold);
5080 mp_.mmap_threshold = value;
5081 mp_.no_dyn_threshold = 1;
5082 return 1;
5084 return 0;
5087 static __always_inline int
5088 do_set_mmaps_max (int32_t value)
5090 LIBC_PROBE (memory_mallopt_mmap_max, 3, value, mp_.n_mmaps_max,
5091 mp_.no_dyn_threshold);
5092 mp_.n_mmaps_max = value;
5093 mp_.no_dyn_threshold = 1;
5094 return 1;
5097 static __always_inline int
5098 do_set_mallopt_check (int32_t value)
5100 return 1;
5103 static __always_inline int
5104 do_set_perturb_byte (int32_t value)
5106 LIBC_PROBE (memory_mallopt_perturb, 2, value, perturb_byte);
5107 perturb_byte = value;
5108 return 1;
5111 static __always_inline int
5112 do_set_arena_test (size_t value)
5114 LIBC_PROBE (memory_mallopt_arena_test, 2, value, mp_.arena_test);
5115 mp_.arena_test = value;
5116 return 1;
5119 static __always_inline int
5120 do_set_arena_max (size_t value)
<