Merge branch 'jk/perf-no-dups'
[git/git.git] / compat / nedmalloc / malloc.c.h
1 /*
2 This is a version (aka dlmalloc) of malloc/free/realloc written by
3 Doug Lea and released to the public domain, as explained at
4 http://creativecommons.org/licenses/publicdomain. Send questions,
5 comments, complaints, performance data, etc to dl@cs.oswego.edu
6
7 * Version pre-2.8.4 Mon Nov 27 11:22:37 2006 (dl at gee)
8
9 Note: There may be an updated version of this malloc obtainable at
10 ftp://gee.cs.oswego.edu/pub/misc/malloc.c
11 Check before installing!
12
13 * Quickstart
14
15 This library is all in one file to simplify the most common usage:
16 ftp it, compile it (-O3), and link it into another program. All of
17 the compile-time options default to reasonable values for use on
18 most platforms. You might later want to step through various
19 compile-time and dynamic tuning options.
20
21 For convenience, an include file for code using this malloc is at:
22 ftp://gee.cs.oswego.edu/pub/misc/malloc-2.8.4.h
23 You don't really need this .h file unless you call functions not
24 defined in your system include files. The .h file contains only the
25 excerpts from this file needed for using this malloc on ANSI C/C++
26 systems, so long as you haven't changed compile-time options about
27 naming and tuning parameters. If you do, then you can create your
28 own malloc.h that does include all settings by cutting at the point
29 indicated below. Note that you may already by default be using a C
30 library containing a malloc that is based on some version of this
31 malloc (for example in linux). You might still want to use the one
32 in this file to customize settings or to avoid overheads associated
33 with library versions.
34
35 * Vital statistics:
36
37 Supported pointer/size_t representation: 4 or 8 bytes
38 size_t MUST be an unsigned type of the same width as
39 pointers. (If you are using an ancient system that declares
40 size_t as a signed type, or need it to be a different width
41 than pointers, you can use a previous release of this malloc
42 (e.g. 2.7.2) supporting these.)
43
44 Alignment: 8 bytes (default)
45 This suffices for nearly all current machines and C compilers.
46 However, you can define MALLOC_ALIGNMENT to be wider than this
47 if necessary (up to 128bytes), at the expense of using more space.
48
49 Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes)
50 8 or 16 bytes (if 8byte sizes)
51 Each malloced chunk has a hidden word of overhead holding size
52 and status information, and additional cross-check word
53 if FOOTERS is defined.
54
55 Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead)
56 8-byte ptrs: 32 bytes (including overhead)
57
58 Even a request for zero bytes (i.e., malloc(0)) returns a
59 pointer to something of the minimum allocatable size.
60 The maximum overhead wastage (i.e., number of extra bytes
61 allocated than were requested in malloc) is less than or equal
62 to the minimum size, except for requests >= mmap_threshold that
63 are serviced via mmap(), where the worst case wastage is about
64 32 bytes plus the remainder from a system page (the minimal
65 mmap unit); typically 4096 or 8192 bytes.
66
67 Security: static-safe; optionally more or less
68 The "security" of malloc refers to the ability of malicious
69 code to accentuate the effects of errors (for example, freeing
70 space that is not currently malloc'ed or overwriting past the
71 ends of chunks) in code that calls malloc. This malloc
72 guarantees not to modify any memory locations below the base of
73 heap, i.e., static variables, even in the presence of usage
74 errors. The routines additionally detect most improper frees
75 and reallocs. All this holds as long as the static bookkeeping
76 for malloc itself is not corrupted by some other means. This
77 is only one aspect of security -- these checks do not, and
78 cannot, detect all possible programming errors.
79
80 If FOOTERS is defined nonzero, then each allocated chunk
81 carries an additional check word to verify that it was malloced
82 from its space. These check words are the same within each
83 execution of a program using malloc, but differ across
84 executions, so externally crafted fake chunks cannot be
85 freed. This improves security by rejecting frees/reallocs that
86 could corrupt heap memory, in addition to the checks preventing
87 writes to statics that are always on. This may further improve
88 security at the expense of time and space overhead. (Note that
89 FOOTERS may also be worth using with MSPACES.)
90
91 By default detected errors cause the program to abort (calling
92 "abort()"). You can override this to instead proceed past
93 errors by defining PROCEED_ON_ERROR. In this case, a bad free
94 has no effect, and a malloc that encounters a bad address
95 caused by user overwrites will ignore the bad address by
96 dropping pointers and indices to all known memory. This may
97 be appropriate for programs that should continue if at all
98 possible in the face of programming errors, although they may
99 run out of memory because dropped memory is never reclaimed.
100
101 If you don't like either of these options, you can define
102 CORRUPTION_ERROR_ACTION and USAGE_ERROR_ACTION to do anything
103 else. And if you are sure that your program using malloc has
104 no errors or vulnerabilities, you can define INSECURE to 1,
105 which might (or might not) provide a small performance improvement.
106
107 Thread-safety: NOT thread-safe unless USE_LOCKS defined
108 When USE_LOCKS is defined, each public call to malloc, free,
109 etc is surrounded with either a pthread mutex or a win32
110 spinlock (depending on WIN32). This is not especially fast, and
111 can be a major bottleneck. It is designed only to provide
112 minimal protection in concurrent environments, and to provide a
113 basis for extensions. If you are using malloc in a concurrent
114 program, consider instead using nedmalloc
115 (http://www.nedprod.com/programs/portable/nedmalloc/) or
116 ptmalloc (See http://www.malloc.de), which are derived
117 from versions of this malloc.
118
119 System requirements: Any combination of MORECORE and/or MMAP/MUNMAP
120 This malloc can use unix sbrk or any emulation (invoked using
121 the CALL_MORECORE macro) and/or mmap/munmap or any emulation
122 (invoked using CALL_MMAP/CALL_MUNMAP) to get and release system
123 memory. On most unix systems, it tends to work best if both
124 MORECORE and MMAP are enabled. On Win32, it uses emulations
125 based on VirtualAlloc. It also uses common C library functions
126 like memset.
127
128 Compliance: I believe it is compliant with the Single Unix Specification
129 (See http://www.unix.org). Also SVID/XPG, ANSI C, and probably
130 others as well.
131
132 * Overview of algorithms
133
134 This is not the fastest, most space-conserving, most portable, or
135 most tunable malloc ever written. However it is among the fastest
136 while also being among the most space-conserving, portable and
137 tunable. Consistent balance across these factors results in a good
138 general-purpose allocator for malloc-intensive programs.
139
140 In most ways, this malloc is a best-fit allocator. Generally, it
141 chooses the best-fitting existing chunk for a request, with ties
142 broken in approximately least-recently-used order. (This strategy
143 normally maintains low fragmentation.) However, for requests less
144 than 256bytes, it deviates from best-fit when there is not an
145 exactly fitting available chunk by preferring to use space adjacent
146 to that used for the previous small request, as well as by breaking
147 ties in approximately most-recently-used order. (These enhance
148 locality of series of small allocations.) And for very large requests
149 (>= 256Kb by default), it relies on system memory mapping
150 facilities, if supported. (This helps avoid carrying around and
151 possibly fragmenting memory used only for large chunks.)
152
153 All operations (except malloc_stats and mallinfo) have execution
154 times that are bounded by a constant factor of the number of bits in
155 a size_t, not counting any clearing in calloc or copying in realloc,
156 or actions surrounding MORECORE and MMAP that have times
157 proportional to the number of non-contiguous regions returned by
158 system allocation routines, which is often just 1. In real-time
159 applications, you can optionally suppress segment traversals using
160 NO_SEGMENT_TRAVERSAL, which assures bounded execution even when
161 system allocators return non-contiguous spaces, at the typical
162 expense of carrying around more memory and increased fragmentation.
163
164 The implementation is not very modular and seriously overuses
165 macros. Perhaps someday all C compilers will do as good a job
166 inlining modular code as can now be done by brute-force expansion,
167 but now, enough of them seem not to.
168
169 Some compilers issue a lot of warnings about code that is
170 dead/unreachable only on some platforms, and also about intentional
171 uses of negation on unsigned types. All known cases of each can be
172 ignored.
173
174 For a longer but out of date high-level description, see
175 http://gee.cs.oswego.edu/dl/html/malloc.html
176
177 * MSPACES
178 If MSPACES is defined, then in addition to malloc, free, etc.,
179 this file also defines mspace_malloc, mspace_free, etc. These
180 are versions of malloc routines that take an "mspace" argument
181 obtained using create_mspace, to control all internal bookkeeping.
182 If ONLY_MSPACES is defined, only these versions are compiled.
183 So if you would like to use this allocator for only some allocations,
184 and your system malloc for others, you can compile with
185 ONLY_MSPACES and then do something like...
186 static mspace mymspace = create_mspace(0,0); // for example
187 #define mymalloc(bytes) mspace_malloc(mymspace, bytes)
188
189 (Note: If you only need one instance of an mspace, you can instead
190 use "USE_DL_PREFIX" to relabel the global malloc.)
191
192 You can similarly create thread-local allocators by storing
193 mspaces as thread-locals. For example:
194 static __thread mspace tlms = 0;
195 void* tlmalloc(size_t bytes) {
196 if (tlms == 0) tlms = create_mspace(0, 0);
197 return mspace_malloc(tlms, bytes);
198 }
199 void tlfree(void* mem) { mspace_free(tlms, mem); }
200
201 Unless FOOTERS is defined, each mspace is completely independent.
202 You cannot allocate from one and free to another (although
203 conformance is only weakly checked, so usage errors are not always
204 caught). If FOOTERS is defined, then each chunk carries around a tag
205 indicating its originating mspace, and frees are directed to their
206 originating spaces.
207
208 ------------------------- Compile-time options ---------------------------
209
210 Be careful in setting #define values for numerical constants of type
211 size_t. On some systems, literal values are not automatically extended
212 to size_t precision unless they are explicitly casted. You can also
213 use the symbolic values MAX_SIZE_T, SIZE_T_ONE, etc below.
214
215 WIN32 default: defined if _WIN32 defined
216 Defining WIN32 sets up defaults for MS environment and compilers.
217 Otherwise defaults are for unix. Beware that there seem to be some
218 cases where this malloc might not be a pure drop-in replacement for
219 Win32 malloc: Random-looking failures from Win32 GDI API's (eg;
220 SetDIBits()) may be due to bugs in some video driver implementations
221 when pixel buffers are malloc()ed, and the region spans more than
222 one VirtualAlloc()ed region. Because dlmalloc uses a small (64Kb)
223 default granularity, pixel buffers may straddle virtual allocation
224 regions more often than when using the Microsoft allocator. You can
225 avoid this by using VirtualAlloc() and VirtualFree() for all pixel
226 buffers rather than using malloc(). If this is not possible,
227 recompile this malloc with a larger DEFAULT_GRANULARITY.
228
229 MALLOC_ALIGNMENT default: (size_t)8
230 Controls the minimum alignment for malloc'ed chunks. It must be a
231 power of two and at least 8, even on machines for which smaller
232 alignments would suffice. It may be defined as larger than this
233 though. Note however that code and data structures are optimized for
234 the case of 8-byte alignment.
235
236 MSPACES default: 0 (false)
237 If true, compile in support for independent allocation spaces.
238 This is only supported if HAVE_MMAP is true.
239
240 ONLY_MSPACES default: 0 (false)
241 If true, only compile in mspace versions, not regular versions.
242
243 USE_LOCKS default: 0 (false)
244 Causes each call to each public routine to be surrounded with
245 pthread or WIN32 mutex lock/unlock. (If set true, this can be
246 overridden on a per-mspace basis for mspace versions.) If set to a
247 non-zero value other than 1, locks are used, but their
248 implementation is left out, so lock functions must be supplied manually.
249
250 USE_SPIN_LOCKS default: 1 iff USE_LOCKS and on x86 using gcc or MSC
251 If true, uses custom spin locks for locking. This is currently
252 supported only for x86 platforms using gcc or recent MS compilers.
253 Otherwise, posix locks or win32 critical sections are used.
254
255 FOOTERS default: 0
256 If true, provide extra checking and dispatching by placing
257 information in the footers of allocated chunks. This adds
258 space and time overhead.
259
260 INSECURE default: 0
261 If true, omit checks for usage errors and heap space overwrites.
262
263 USE_DL_PREFIX default: NOT defined
264 Causes compiler to prefix all public routines with the string 'dl'.
265 This can be useful when you only want to use this malloc in one part
266 of a program, using your regular system malloc elsewhere.
267
268 ABORT default: defined as abort()
269 Defines how to abort on failed checks. On most systems, a failed
270 check cannot die with an "assert" or even print an informative
271 message, because the underlying print routines in turn call malloc,
272 which will fail again. Generally, the best policy is to simply call
273 abort(). It's not very useful to do more than this because many
274 errors due to overwriting will show up as address faults (null, odd
275 addresses etc) rather than malloc-triggered checks, so will also
276 abort. Also, most compilers know that abort() does not return, so
277 can better optimize code conditionally calling it.
278
279 PROCEED_ON_ERROR default: defined as 0 (false)
280 Controls whether detected bad addresses cause them to bypassed
281 rather than aborting. If set, detected bad arguments to free and
282 realloc are ignored. And all bookkeeping information is zeroed out
283 upon a detected overwrite of freed heap space, thus losing the
284 ability to ever return it from malloc again, but enabling the
285 application to proceed. If PROCEED_ON_ERROR is defined, the
286 static variable malloc_corruption_error_count is compiled in
287 and can be examined to see if errors have occurred. This option
288 generates slower code than the default abort policy.
289
290 DEBUG default: NOT defined
291 The DEBUG setting is mainly intended for people trying to modify
292 this code or diagnose problems when porting to new platforms.
293 However, it may also be able to better isolate user errors than just
294 using runtime checks. The assertions in the check routines spell
295 out in more detail the assumptions and invariants underlying the
296 algorithms. The checking is fairly extensive, and will slow down
297 execution noticeably. Calling malloc_stats or mallinfo with DEBUG
298 set will attempt to check every non-mmapped allocated and free chunk
299 in the course of computing the summaries.
300
301 ABORT_ON_ASSERT_FAILURE default: defined as 1 (true)
302 Debugging assertion failures can be nearly impossible if your
303 version of the assert macro causes malloc to be called, which will
304 lead to a cascade of further failures, blowing the runtime stack.
305 ABORT_ON_ASSERT_FAILURE cause assertions failures to call abort(),
306 which will usually make debugging easier.
307
308 MALLOC_FAILURE_ACTION default: sets errno to ENOMEM, or no-op on win32
309 The action to take before "return 0" when malloc fails to be able to
310 return memory because there is none available.
311
312 HAVE_MORECORE default: 1 (true) unless win32 or ONLY_MSPACES
313 True if this system supports sbrk or an emulation of it.
314
315 MORECORE default: sbrk
316 The name of the sbrk-style system routine to call to obtain more
317 memory. See below for guidance on writing custom MORECORE
318 functions. The type of the argument to sbrk/MORECORE varies across
319 systems. It cannot be size_t, because it supports negative
320 arguments, so it is normally the signed type of the same width as
321 size_t (sometimes declared as "intptr_t"). It doesn't much matter
322 though. Internally, we only call it with arguments less than half
323 the max value of a size_t, which should work across all reasonable
324 possibilities, although sometimes generating compiler warnings.
325
326 MORECORE_CONTIGUOUS default: 1 (true) if HAVE_MORECORE
327 If true, take advantage of fact that consecutive calls to MORECORE
328 with positive arguments always return contiguous increasing
329 addresses. This is true of unix sbrk. It does not hurt too much to
330 set it true anyway, since malloc copes with non-contiguities.
331 Setting it false when definitely non-contiguous saves time
332 and possibly wasted space it would take to discover this though.
333
334 MORECORE_CANNOT_TRIM default: NOT defined
335 True if MORECORE cannot release space back to the system when given
336 negative arguments. This is generally necessary only if you are
337 using a hand-crafted MORECORE function that cannot handle negative
338 arguments.
339
340 NO_SEGMENT_TRAVERSAL default: 0
341 If non-zero, suppresses traversals of memory segments
342 returned by either MORECORE or CALL_MMAP. This disables
343 merging of segments that are contiguous, and selectively
344 releasing them to the OS if unused, but bounds execution times.
345
346 HAVE_MMAP default: 1 (true)
347 True if this system supports mmap or an emulation of it. If so, and
348 HAVE_MORECORE is not true, MMAP is used for all system
349 allocation. If set and HAVE_MORECORE is true as well, MMAP is
350 primarily used to directly allocate very large blocks. It is also
351 used as a backup strategy in cases where MORECORE fails to provide
352 space from system. Note: A single call to MUNMAP is assumed to be
353 able to unmap memory that may have be allocated using multiple calls
354 to MMAP, so long as they are adjacent.
355
356 HAVE_MREMAP default: 1 on linux, else 0
357 If true realloc() uses mremap() to re-allocate large blocks and
358 extend or shrink allocation spaces.
359
360 MMAP_CLEARS default: 1 except on WINCE.
361 True if mmap clears memory so calloc doesn't need to. This is true
362 for standard unix mmap using /dev/zero and on WIN32 except for WINCE.
363
364 USE_BUILTIN_FFS default: 0 (i.e., not used)
365 Causes malloc to use the builtin ffs() function to compute indices.
366 Some compilers may recognize and intrinsify ffs to be faster than the
367 supplied C version. Also, the case of x86 using gcc is special-cased
368 to an asm instruction, so is already as fast as it can be, and so
369 this setting has no effect. Similarly for Win32 under recent MS compilers.
370 (On most x86s, the asm version is only slightly faster than the C version.)
371
372 malloc_getpagesize default: derive from system includes, or 4096.
373 The system page size. To the extent possible, this malloc manages
374 memory from the system in page-size units. This may be (and
375 usually is) a function rather than a constant. This is ignored
376 if WIN32, where page size is determined using getSystemInfo during
377 initialization.
378
379 USE_DEV_RANDOM default: 0 (i.e., not used)
380 Causes malloc to use /dev/random to initialize secure magic seed for
381 stamping footers. Otherwise, the current time is used.
382
383 NO_MALLINFO default: 0
384 If defined, don't compile "mallinfo". This can be a simple way
385 of dealing with mismatches between system declarations and
386 those in this file.
387
388 MALLINFO_FIELD_TYPE default: size_t
389 The type of the fields in the mallinfo struct. This was originally
390 defined as "int" in SVID etc, but is more usefully defined as
391 size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set
392
393 REALLOC_ZERO_BYTES_FREES default: not defined
394 This should be set if a call to realloc with zero bytes should
395 be the same as a call to free. Some people think it should. Otherwise,
396 since this malloc returns a unique pointer for malloc(0), so does
397 realloc(p, 0).
398
399 LACKS_UNISTD_H, LACKS_FCNTL_H, LACKS_SYS_PARAM_H, LACKS_SYS_MMAN_H
400 LACKS_STRINGS_H, LACKS_STRING_H, LACKS_SYS_TYPES_H, LACKS_ERRNO_H
401 LACKS_STDLIB_H default: NOT defined unless on WIN32
402 Define these if your system does not have these header files.
403 You might need to manually insert some of the declarations they provide.
404
405 DEFAULT_GRANULARITY default: page size if MORECORE_CONTIGUOUS,
406 system_info.dwAllocationGranularity in WIN32,
407 otherwise 64K.
408 Also settable using mallopt(M_GRANULARITY, x)
409 The unit for allocating and deallocating memory from the system. On
410 most systems with contiguous MORECORE, there is no reason to
411 make this more than a page. However, systems with MMAP tend to
412 either require or encourage larger granularities. You can increase
413 this value to prevent system allocation functions to be called so
414 often, especially if they are slow. The value must be at least one
415 page and must be a power of two. Setting to 0 causes initialization
416 to either page size or win32 region size. (Note: In previous
417 versions of malloc, the equivalent of this option was called
418 "TOP_PAD")
419
420 DEFAULT_TRIM_THRESHOLD default: 2MB
421 Also settable using mallopt(M_TRIM_THRESHOLD, x)
422 The maximum amount of unused top-most memory to keep before
423 releasing via malloc_trim in free(). Automatic trimming is mainly
424 useful in long-lived programs using contiguous MORECORE. Because
425 trimming via sbrk can be slow on some systems, and can sometimes be
426 wasteful (in cases where programs immediately afterward allocate
427 more large chunks) the value should be high enough so that your
428 overall system performance would improve by releasing this much
429 memory. As a rough guide, you might set to a value close to the
430 average size of a process (program) running on your system.
431 Releasing this much memory would allow such a process to run in
432 memory. Generally, it is worth tuning trim thresholds when a
433 program undergoes phases where several large chunks are allocated
434 and released in ways that can reuse each other's storage, perhaps
435 mixed with phases where there are no such chunks at all. The trim
436 value must be greater than page size to have any useful effect. To
437 disable trimming completely, you can set to MAX_SIZE_T. Note that the trick
438 some people use of mallocing a huge space and then freeing it at
439 program startup, in an attempt to reserve system memory, doesn't
440 have the intended effect under automatic trimming, since that memory
441 will immediately be returned to the system.
442
443 DEFAULT_MMAP_THRESHOLD default: 256K
444 Also settable using mallopt(M_MMAP_THRESHOLD, x)
445 The request size threshold for using MMAP to directly service a
446 request. Requests of at least this size that cannot be allocated
447 using already-existing space will be serviced via mmap. (If enough
448 normal freed space already exists it is used instead.) Using mmap
449 segregates relatively large chunks of memory so that they can be
450 individually obtained and released from the host system. A request
451 serviced through mmap is never reused by any other request (at least
452 not directly; the system may just so happen to remap successive
453 requests to the same locations). Segregating space in this way has
454 the benefits that: Mmapped space can always be individually released
455 back to the system, which helps keep the system level memory demands
456 of a long-lived program low. Also, mapped memory doesn't become
457 `locked' between other chunks, as can happen with normally allocated
458 chunks, which means that even trimming via malloc_trim would not
459 release them. However, it has the disadvantage that the space
460 cannot be reclaimed, consolidated, and then used to service later
461 requests, as happens with normal chunks. The advantages of mmap
462 nearly always outweigh disadvantages for "large" chunks, but the
463 value of "large" may vary across systems. The default is an
464 empirically derived value that works well in most systems. You can
465 disable mmap by setting to MAX_SIZE_T.
466
467 MAX_RELEASE_CHECK_RATE default: 4095 unless not HAVE_MMAP
468 The number of consolidated frees between checks to release
469 unused segments when freeing. When using non-contiguous segments,
470 especially with multiple mspaces, checking only for topmost space
471 doesn't always suffice to trigger trimming. To compensate for this,
472 free() will, with a period of MAX_RELEASE_CHECK_RATE (or the
473 current number of segments, if greater) try to release unused
474 segments to the OS when freeing chunks that result in
475 consolidation. The best value for this parameter is a compromise
476 between slowing down frees with relatively costly checks that
477 rarely trigger versus holding on to unused memory. To effectively
478 disable, set to MAX_SIZE_T. This may lead to a very slight speed
479 improvement at the expense of carrying around more memory.
480 */
481
482 /* Version identifier to allow people to support multiple versions */
483 #ifndef DLMALLOC_VERSION
484 #define DLMALLOC_VERSION 20804
485 #endif /* DLMALLOC_VERSION */
486
487 #if defined(linux)
488 #define _GNU_SOURCE 1
489 #endif
490
491 #ifndef WIN32
492 #ifdef _WIN32
493 #define WIN32 1
494 #endif /* _WIN32 */
495 #ifdef _WIN32_WCE
496 #define LACKS_FCNTL_H
497 #define WIN32 1
498 #endif /* _WIN32_WCE */
499 #endif /* WIN32 */
500 #ifdef WIN32
501 #define WIN32_LEAN_AND_MEAN
502 #ifndef _WIN32_WINNT
503 #define _WIN32_WINNT 0x403
504 #endif
505 #include <windows.h>
506 #define HAVE_MMAP 1
507 #define HAVE_MORECORE 0
508 #define LACKS_UNISTD_H
509 #define LACKS_SYS_PARAM_H
510 #define LACKS_SYS_MMAN_H
511 #define LACKS_STRING_H
512 #define LACKS_STRINGS_H
513 #define LACKS_SYS_TYPES_H
514 #define LACKS_ERRNO_H
515 #ifndef MALLOC_FAILURE_ACTION
516 #define MALLOC_FAILURE_ACTION
517 #endif /* MALLOC_FAILURE_ACTION */
518 #ifdef _WIN32_WCE /* WINCE reportedly does not clear */
519 #define MMAP_CLEARS 0
520 #else
521 #define MMAP_CLEARS 1
522 #endif /* _WIN32_WCE */
523 #endif /* WIN32 */
524
525 #if defined(DARWIN) || defined(_DARWIN)
526 /* Mac OSX docs advise not to use sbrk; it seems better to use mmap */
527 #ifndef HAVE_MORECORE
528 #define HAVE_MORECORE 0
529 #define HAVE_MMAP 1
530 /* OSX allocators provide 16 byte alignment */
531 #ifndef MALLOC_ALIGNMENT
532 #define MALLOC_ALIGNMENT ((size_t)16U)
533 #endif
534 #endif /* HAVE_MORECORE */
535 #endif /* DARWIN */
536
537 #ifndef LACKS_SYS_TYPES_H
538 #include <sys/types.h> /* For size_t */
539 #endif /* LACKS_SYS_TYPES_H */
540
541 /* The maximum possible size_t value has all bits set */
542 #define MAX_SIZE_T (~(size_t)0)
543
544 #ifndef ONLY_MSPACES
545 #define ONLY_MSPACES 0 /* define to a value */
546 #else
547 #define ONLY_MSPACES 1
548 #endif /* ONLY_MSPACES */
549 #ifndef MSPACES
550 #if ONLY_MSPACES
551 #define MSPACES 1
552 #else /* ONLY_MSPACES */
553 #define MSPACES 0
554 #endif /* ONLY_MSPACES */
555 #endif /* MSPACES */
556 #ifndef MALLOC_ALIGNMENT
557 #define MALLOC_ALIGNMENT ((size_t)8U)
558 #endif /* MALLOC_ALIGNMENT */
559 #ifndef FOOTERS
560 #define FOOTERS 0
561 #endif /* FOOTERS */
562 #ifndef ABORT
563 #define ABORT abort()
564 #endif /* ABORT */
565 #ifndef ABORT_ON_ASSERT_FAILURE
566 #define ABORT_ON_ASSERT_FAILURE 1
567 #endif /* ABORT_ON_ASSERT_FAILURE */
568 #ifndef PROCEED_ON_ERROR
569 #define PROCEED_ON_ERROR 0
570 #endif /* PROCEED_ON_ERROR */
571 #ifndef USE_LOCKS
572 #define USE_LOCKS 0
573 #endif /* USE_LOCKS */
574 #ifndef USE_SPIN_LOCKS
575 #if USE_LOCKS && (defined(__GNUC__) && ((defined(__i386__) || defined(__x86_64__)))) || (defined(_MSC_VER) && _MSC_VER>=1310)
576 #define USE_SPIN_LOCKS 1
577 #else
578 #define USE_SPIN_LOCKS 0
579 #endif /* USE_LOCKS && ... */
580 #endif /* USE_SPIN_LOCKS */
581 #ifndef INSECURE
582 #define INSECURE 0
583 #endif /* INSECURE */
584 #ifndef HAVE_MMAP
585 #define HAVE_MMAP 1
586 #endif /* HAVE_MMAP */
587 #ifndef MMAP_CLEARS
588 #define MMAP_CLEARS 1
589 #endif /* MMAP_CLEARS */
590 #ifndef HAVE_MREMAP
591 #ifdef linux
592 #define HAVE_MREMAP 1
593 #else /* linux */
594 #define HAVE_MREMAP 0
595 #endif /* linux */
596 #endif /* HAVE_MREMAP */
597 #ifndef MALLOC_FAILURE_ACTION
598 #define MALLOC_FAILURE_ACTION errno = ENOMEM;
599 #endif /* MALLOC_FAILURE_ACTION */
600 #ifndef HAVE_MORECORE
601 #if ONLY_MSPACES
602 #define HAVE_MORECORE 0
603 #else /* ONLY_MSPACES */
604 #define HAVE_MORECORE 1
605 #endif /* ONLY_MSPACES */
606 #endif /* HAVE_MORECORE */
607 #if !HAVE_MORECORE
608 #define MORECORE_CONTIGUOUS 0
609 #else /* !HAVE_MORECORE */
610 #define MORECORE_DEFAULT sbrk
611 #ifndef MORECORE_CONTIGUOUS
612 #define MORECORE_CONTIGUOUS 1
613 #endif /* MORECORE_CONTIGUOUS */
614 #endif /* HAVE_MORECORE */
615 #ifndef DEFAULT_GRANULARITY
616 #if (MORECORE_CONTIGUOUS || defined(WIN32))
617 #define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */
618 #else /* MORECORE_CONTIGUOUS */
619 #define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U)
620 #endif /* MORECORE_CONTIGUOUS */
621 #endif /* DEFAULT_GRANULARITY */
622 #ifndef DEFAULT_TRIM_THRESHOLD
623 #ifndef MORECORE_CANNOT_TRIM
624 #define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U)
625 #else /* MORECORE_CANNOT_TRIM */
626 #define DEFAULT_TRIM_THRESHOLD MAX_SIZE_T
627 #endif /* MORECORE_CANNOT_TRIM */
628 #endif /* DEFAULT_TRIM_THRESHOLD */
629 #ifndef DEFAULT_MMAP_THRESHOLD
630 #if HAVE_MMAP
631 #define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U)
632 #else /* HAVE_MMAP */
633 #define DEFAULT_MMAP_THRESHOLD MAX_SIZE_T
634 #endif /* HAVE_MMAP */
635 #endif /* DEFAULT_MMAP_THRESHOLD */
636 #ifndef MAX_RELEASE_CHECK_RATE
637 #if HAVE_MMAP
638 #define MAX_RELEASE_CHECK_RATE 4095
639 #else
640 #define MAX_RELEASE_CHECK_RATE MAX_SIZE_T
641 #endif /* HAVE_MMAP */
642 #endif /* MAX_RELEASE_CHECK_RATE */
643 #ifndef USE_BUILTIN_FFS
644 #define USE_BUILTIN_FFS 0
645 #endif /* USE_BUILTIN_FFS */
646 #ifndef USE_DEV_RANDOM
647 #define USE_DEV_RANDOM 0
648 #endif /* USE_DEV_RANDOM */
649 #ifndef NO_MALLINFO
650 #define NO_MALLINFO 0
651 #endif /* NO_MALLINFO */
652 #ifndef MALLINFO_FIELD_TYPE
653 #define MALLINFO_FIELD_TYPE size_t
654 #endif /* MALLINFO_FIELD_TYPE */
655 #ifndef NO_SEGMENT_TRAVERSAL
656 #define NO_SEGMENT_TRAVERSAL 0
657 #endif /* NO_SEGMENT_TRAVERSAL */
658
659 /*
660 mallopt tuning options. SVID/XPG defines four standard parameter
661 numbers for mallopt, normally defined in malloc.h. None of these
662 are used in this malloc, so setting them has no effect. But this
663 malloc does support the following options.
664 */
665
666 #define M_TRIM_THRESHOLD (-1)
667 #define M_GRANULARITY (-2)
668 #define M_MMAP_THRESHOLD (-3)
669
670 /* ------------------------ Mallinfo declarations ------------------------ */
671
672 #if !NO_MALLINFO
673 /*
674 This version of malloc supports the standard SVID/XPG mallinfo
675 routine that returns a struct containing usage properties and
676 statistics. It should work on any system that has a
677 /usr/include/malloc.h defining struct mallinfo. The main
678 declaration needed is the mallinfo struct that is returned (by-copy)
679 by mallinfo(). The malloinfo struct contains a bunch of fields that
680 are not even meaningful in this version of malloc. These fields are
681 are instead filled by mallinfo() with other numbers that might be of
682 interest.
683
684 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
685 /usr/include/malloc.h file that includes a declaration of struct
686 mallinfo. If so, it is included; else a compliant version is
687 declared below. These must be precisely the same for mallinfo() to
688 work. The original SVID version of this struct, defined on most
689 systems with mallinfo, declares all fields as ints. But some others
690 define as unsigned long. If your system defines the fields using a
691 type of different width than listed here, you MUST #include your
692 system version and #define HAVE_USR_INCLUDE_MALLOC_H.
693 */
694
695 /* #define HAVE_USR_INCLUDE_MALLOC_H */
696
697 #ifdef HAVE_USR_INCLUDE_MALLOC_H
698 #include "/usr/include/malloc.h"
699 #else /* HAVE_USR_INCLUDE_MALLOC_H */
700 #ifndef STRUCT_MALLINFO_DECLARED
701 #define STRUCT_MALLINFO_DECLARED 1
702 struct mallinfo {
703 MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */
704 MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */
705 MALLINFO_FIELD_TYPE smblks; /* always 0 */
706 MALLINFO_FIELD_TYPE hblks; /* always 0 */
707 MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */
708 MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */
709 MALLINFO_FIELD_TYPE fsmblks; /* always 0 */
710 MALLINFO_FIELD_TYPE uordblks; /* total allocated space */
711 MALLINFO_FIELD_TYPE fordblks; /* total free space */
712 MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */
713 };
714 #endif /* STRUCT_MALLINFO_DECLARED */
715 #endif /* HAVE_USR_INCLUDE_MALLOC_H */
716 #endif /* NO_MALLINFO */
717
718 /*
719 Try to persuade compilers to inline. The most critical functions for
720 inlining are defined as macros, so these aren't used for them.
721 */
722
723 #ifdef __MINGW64_VERSION_MAJOR
724 #undef FORCEINLINE
725 #endif
726 #ifndef FORCEINLINE
727 #if defined(__GNUC__)
728 #define FORCEINLINE __inline __attribute__ ((always_inline))
729 #elif defined(_MSC_VER)
730 #define FORCEINLINE __forceinline
731 #endif
732 #endif
733 #ifndef NOINLINE
734 #if defined(__GNUC__)
735 #define NOINLINE __attribute__ ((noinline))
736 #elif defined(_MSC_VER)
737 #define NOINLINE __declspec(noinline)
738 #else
739 #define NOINLINE
740 #endif
741 #endif
742
743 #ifdef __cplusplus
744 extern "C" {
745 #ifndef FORCEINLINE
746 #define FORCEINLINE inline
747 #endif
748 #endif /* __cplusplus */
749 #ifndef FORCEINLINE
750 #define FORCEINLINE
751 #endif
752
753 #if !ONLY_MSPACES
754
755 /* ------------------- Declarations of public routines ------------------- */
756
757 #ifndef USE_DL_PREFIX
758 #define dlcalloc calloc
759 #define dlfree free
760 #define dlmalloc malloc
761 #define dlmemalign memalign
762 #define dlrealloc realloc
763 #define dlvalloc valloc
764 #define dlpvalloc pvalloc
765 #define dlmallinfo mallinfo
766 #define dlmallopt mallopt
767 #define dlmalloc_trim malloc_trim
768 #define dlmalloc_stats malloc_stats
769 #define dlmalloc_usable_size malloc_usable_size
770 #define dlmalloc_footprint malloc_footprint
771 #define dlmalloc_max_footprint malloc_max_footprint
772 #define dlindependent_calloc independent_calloc
773 #define dlindependent_comalloc independent_comalloc
774 #endif /* USE_DL_PREFIX */
775
776
777 /*
778 malloc(size_t n)
779 Returns a pointer to a newly allocated chunk of at least n bytes, or
780 null if no space is available, in which case errno is set to ENOMEM
781 on ANSI C systems.
782
783 If n is zero, malloc returns a minimum-sized chunk. (The minimum
784 size is 16 bytes on most 32bit systems, and 32 bytes on 64bit
785 systems.) Note that size_t is an unsigned type, so calls with
786 arguments that would be negative if signed are interpreted as
787 requests for huge amounts of space, which will often fail. The
788 maximum supported value of n differs across systems, but is in all
789 cases less than the maximum representable value of a size_t.
790 */
791 void* dlmalloc(size_t);
792
793 /*
794 free(void* p)
795 Releases the chunk of memory pointed to by p, that had been previously
796 allocated using malloc or a related routine such as realloc.
797 It has no effect if p is null. If p was not malloced or already
798 freed, free(p) will by default cause the current program to abort.
799 */
800 void dlfree(void*);
801
802 /*
803 calloc(size_t n_elements, size_t element_size);
804 Returns a pointer to n_elements * element_size bytes, with all locations
805 set to zero.
806 */
807 void* dlcalloc(size_t, size_t);
808
809 /*
810 realloc(void* p, size_t n)
811 Returns a pointer to a chunk of size n that contains the same data
812 as does chunk p up to the minimum of (n, p's size) bytes, or null
813 if no space is available.
814
815 The returned pointer may or may not be the same as p. The algorithm
816 prefers extending p in most cases when possible, otherwise it
817 employs the equivalent of a malloc-copy-free sequence.
818
819 If p is null, realloc is equivalent to malloc.
820
821 If space is not available, realloc returns null, errno is set (if on
822 ANSI) and p is NOT freed.
823
824 if n is for fewer bytes than already held by p, the newly unused
825 space is lopped off and freed if possible. realloc with a size
826 argument of zero (re)allocates a minimum-sized chunk.
827
828 The old unix realloc convention of allowing the last-free'd chunk
829 to be used as an argument to realloc is not supported.
830 */
831
832 void* dlrealloc(void*, size_t);
833
834 /*
835 memalign(size_t alignment, size_t n);
836 Returns a pointer to a newly allocated chunk of n bytes, aligned
837 in accord with the alignment argument.
838
839 The alignment argument should be a power of two. If the argument is
840 not a power of two, the nearest greater power is used.
841 8-byte alignment is guaranteed by normal malloc calls, so don't
842 bother calling memalign with an argument of 8 or less.
843
844 Overreliance on memalign is a sure way to fragment space.
845 */
846 void* dlmemalign(size_t, size_t);
847
848 /*
849 valloc(size_t n);
850 Equivalent to memalign(pagesize, n), where pagesize is the page
851 size of the system. If the pagesize is unknown, 4096 is used.
852 */
853 void* dlvalloc(size_t);
854
855 /*
856 mallopt(int parameter_number, int parameter_value)
857 Sets tunable parameters The format is to provide a
858 (parameter-number, parameter-value) pair. mallopt then sets the
859 corresponding parameter to the argument value if it can (i.e., so
860 long as the value is meaningful), and returns 1 if successful else
861 0. To workaround the fact that mallopt is specified to use int,
862 not size_t parameters, the value -1 is specially treated as the
863 maximum unsigned size_t value.
864
865 SVID/XPG/ANSI defines four standard param numbers for mallopt,
866 normally defined in malloc.h. None of these are use in this malloc,
867 so setting them has no effect. But this malloc also supports other
868 options in mallopt. See below for details. Briefly, supported
869 parameters are as follows (listed defaults are for "typical"
870 configurations).
871
872 Symbol param # default allowed param values
873 M_TRIM_THRESHOLD -1 2*1024*1024 any (-1 disables)
874 M_GRANULARITY -2 page size any power of 2 >= page size
875 M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support)
876 */
877 int dlmallopt(int, int);
878
879 /*
880 malloc_footprint();
881 Returns the number of bytes obtained from the system. The total
882 number of bytes allocated by malloc, realloc etc., is less than this
883 value. Unlike mallinfo, this function returns only a precomputed
884 result, so can be called frequently to monitor memory consumption.
885 Even if locks are otherwise defined, this function does not use them,
886 so results might not be up to date.
887 */
888 size_t dlmalloc_footprint(void);
889
890 /*
891 malloc_max_footprint();
892 Returns the maximum number of bytes obtained from the system. This
893 value will be greater than current footprint if deallocated space
894 has been reclaimed by the system. The peak number of bytes allocated
895 by malloc, realloc etc., is less than this value. Unlike mallinfo,
896 this function returns only a precomputed result, so can be called
897 frequently to monitor memory consumption. Even if locks are
898 otherwise defined, this function does not use them, so results might
899 not be up to date.
900 */
901 size_t dlmalloc_max_footprint(void);
902
903 #if !NO_MALLINFO
904 /*
905 mallinfo()
906 Returns (by copy) a struct containing various summary statistics:
907
908 arena: current total non-mmapped bytes allocated from system
909 ordblks: the number of free chunks
910 smblks: always zero.
911 hblks: current number of mmapped regions
912 hblkhd: total bytes held in mmapped regions
913 usmblks: the maximum total allocated space. This will be greater
914 than current total if trimming has occurred.
915 fsmblks: always zero
916 uordblks: current total allocated space (normal or mmapped)
917 fordblks: total free space
918 keepcost: the maximum number of bytes that could ideally be released
919 back to system via malloc_trim. ("ideally" means that
920 it ignores page restrictions etc.)
921
922 Because these fields are ints, but internal bookkeeping may
923 be kept as longs, the reported values may wrap around zero and
924 thus be inaccurate.
925 */
926 struct mallinfo dlmallinfo(void);
927 #endif /* NO_MALLINFO */
928
929 /*
930 independent_calloc(size_t n_elements, size_t element_size, void* chunks[]);
931
932 independent_calloc is similar to calloc, but instead of returning a
933 single cleared space, it returns an array of pointers to n_elements
934 independent elements that can hold contents of size elem_size, each
935 of which starts out cleared, and can be independently freed,
936 realloc'ed etc. The elements are guaranteed to be adjacently
937 allocated (this is not guaranteed to occur with multiple callocs or
938 mallocs), which may also improve cache locality in some
939 applications.
940
941 The "chunks" argument is optional (i.e., may be null, which is
942 probably the most typical usage). If it is null, the returned array
943 is itself dynamically allocated and should also be freed when it is
944 no longer needed. Otherwise, the chunks array must be of at least
945 n_elements in length. It is filled in with the pointers to the
946 chunks.
947
948 In either case, independent_calloc returns this pointer array, or
949 null if the allocation failed. If n_elements is zero and "chunks"
950 is null, it returns a chunk representing an array with zero elements
951 (which should be freed if not wanted).
952
953 Each element must be individually freed when it is no longer
954 needed. If you'd like to instead be able to free all at once, you
955 should instead use regular calloc and assign pointers into this
956 space to represent elements. (In this case though, you cannot
957 independently free elements.)
958
959 independent_calloc simplifies and speeds up implementations of many
960 kinds of pools. It may also be useful when constructing large data
961 structures that initially have a fixed number of fixed-sized nodes,
962 but the number is not known at compile time, and some of the nodes
963 may later need to be freed. For example:
964
965 struct Node { int item; struct Node* next; };
966
967 struct Node* build_list() {
968 struct Node** pool;
969 int n = read_number_of_nodes_needed();
970 if (n <= 0) return 0;
971 pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0);
972 if (pool == 0) die();
973 // organize into a linked list...
974 struct Node* first = pool[0];
975 for (i = 0; i < n-1; ++i)
976 pool[i]->next = pool[i+1];
977 free(pool); // Can now free the array (or not, if it is needed later)
978 return first;
979 }
980 */
981 void** dlindependent_calloc(size_t, size_t, void**);
982
983 /*
984 independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]);
985
986 independent_comalloc allocates, all at once, a set of n_elements
987 chunks with sizes indicated in the "sizes" array. It returns
988 an array of pointers to these elements, each of which can be
989 independently freed, realloc'ed etc. The elements are guaranteed to
990 be adjacently allocated (this is not guaranteed to occur with
991 multiple callocs or mallocs), which may also improve cache locality
992 in some applications.
993
994 The "chunks" argument is optional (i.e., may be null). If it is null
995 the returned array is itself dynamically allocated and should also
996 be freed when it is no longer needed. Otherwise, the chunks array
997 must be of at least n_elements in length. It is filled in with the
998 pointers to the chunks.
999
1000 In either case, independent_comalloc returns this pointer array, or
1001 null if the allocation failed. If n_elements is zero and chunks is
1002 null, it returns a chunk representing an array with zero elements
1003 (which should be freed if not wanted).
1004
1005 Each element must be individually freed when it is no longer
1006 needed. If you'd like to instead be able to free all at once, you
1007 should instead use a single regular malloc, and assign pointers at
1008 particular offsets in the aggregate space. (In this case though, you
1009 cannot independently free elements.)
1010
1011 independent_comallac differs from independent_calloc in that each
1012 element may have a different size, and also that it does not
1013 automatically clear elements.
1014
1015 independent_comalloc can be used to speed up allocation in cases
1016 where several structs or objects must always be allocated at the
1017 same time. For example:
1018
1019 struct Head { ... }
1020 struct Foot { ... }
1021
1022 void send_message(char* msg) {
1023 int msglen = strlen(msg);
1024 size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) };
1025 void* chunks[3];
1026 if (independent_comalloc(3, sizes, chunks) == 0)
1027 die();
1028 struct Head* head = (struct Head*)(chunks[0]);
1029 char* body = (char*)(chunks[1]);
1030 struct Foot* foot = (struct Foot*)(chunks[2]);
1031 // ...
1032 }
1033
1034 In general though, independent_comalloc is worth using only for
1035 larger values of n_elements. For small values, you probably won't
1036 detect enough difference from series of malloc calls to bother.
1037
1038 Overuse of independent_comalloc can increase overall memory usage,
1039 since it cannot reuse existing noncontiguous small chunks that
1040 might be available for some of the elements.
1041 */
1042 void** dlindependent_comalloc(size_t, size_t*, void**);
1043
1044
1045 /*
1046 pvalloc(size_t n);
1047 Equivalent to valloc(minimum-page-that-holds(n)), that is,
1048 round up n to nearest pagesize.
1049 */
1050 void* dlpvalloc(size_t);
1051
1052 /*
1053 malloc_trim(size_t pad);
1054
1055 If possible, gives memory back to the system (via negative arguments
1056 to sbrk) if there is unused memory at the `high' end of the malloc
1057 pool or in unused MMAP segments. You can call this after freeing
1058 large blocks of memory to potentially reduce the system-level memory
1059 requirements of a program. However, it cannot guarantee to reduce
1060 memory. Under some allocation patterns, some large free blocks of
1061 memory will be locked between two used chunks, so they cannot be
1062 given back to the system.
1063
1064 The `pad' argument to malloc_trim represents the amount of free
1065 trailing space to leave untrimmed. If this argument is zero, only
1066 the minimum amount of memory to maintain internal data structures
1067 will be left. Non-zero arguments can be supplied to maintain enough
1068 trailing space to service future expected allocations without having
1069 to re-obtain memory from the system.
1070
1071 Malloc_trim returns 1 if it actually released any memory, else 0.
1072 */
1073 int dlmalloc_trim(size_t);
1074
1075 /*
1076 malloc_stats();
1077 Prints on stderr the amount of space obtained from the system (both
1078 via sbrk and mmap), the maximum amount (which may be more than
1079 current if malloc_trim and/or munmap got called), and the current
1080 number of bytes allocated via malloc (or realloc, etc) but not yet
1081 freed. Note that this is the number of bytes allocated, not the
1082 number requested. It will be larger than the number requested
1083 because of alignment and bookkeeping overhead. Because it includes
1084 alignment wastage as being in use, this figure may be greater than
1085 zero even when no user-level chunks are allocated.
1086
1087 The reported current and maximum system memory can be inaccurate if
1088 a program makes other calls to system memory allocation functions
1089 (normally sbrk) outside of malloc.
1090
1091 malloc_stats prints only the most commonly interesting statistics.
1092 More information can be obtained by calling mallinfo.
1093 */
1094 void dlmalloc_stats(void);
1095
1096 #endif /* ONLY_MSPACES */
1097
1098 /*
1099 malloc_usable_size(void* p);
1100
1101 Returns the number of bytes you can actually use in
1102 an allocated chunk, which may be more than you requested (although
1103 often not) due to alignment and minimum size constraints.
1104 You can use this many bytes without worrying about
1105 overwriting other allocated objects. This is not a particularly great
1106 programming practice. malloc_usable_size can be more useful in
1107 debugging and assertions, for example:
1108
1109 p = malloc(n);
1110 assert(malloc_usable_size(p) >= 256);
1111 */
1112 size_t dlmalloc_usable_size(void*);
1113
1114
1115 #if MSPACES
1116
1117 /*
1118 mspace is an opaque type representing an independent
1119 region of space that supports mspace_malloc, etc.
1120 */
1121 typedef void* mspace;
1122
1123 /*
1124 create_mspace creates and returns a new independent space with the
1125 given initial capacity, or, if 0, the default granularity size. It
1126 returns null if there is no system memory available to create the
1127 space. If argument locked is non-zero, the space uses a separate
1128 lock to control access. The capacity of the space will grow
1129 dynamically as needed to service mspace_malloc requests. You can
1130 control the sizes of incremental increases of this space by
1131 compiling with a different DEFAULT_GRANULARITY or dynamically
1132 setting with mallopt(M_GRANULARITY, value).
1133 */
1134 mspace create_mspace(size_t capacity, int locked);
1135
1136 /*
1137 destroy_mspace destroys the given space, and attempts to return all
1138 of its memory back to the system, returning the total number of
1139 bytes freed. After destruction, the results of access to all memory
1140 used by the space become undefined.
1141 */
1142 size_t destroy_mspace(mspace msp);
1143
1144 /*
1145 create_mspace_with_base uses the memory supplied as the initial base
1146 of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this
1147 space is used for bookkeeping, so the capacity must be at least this
1148 large. (Otherwise 0 is returned.) When this initial space is
1149 exhausted, additional memory will be obtained from the system.
1150 Destroying this space will deallocate all additionally allocated
1151 space (if possible) but not the initial base.
1152 */
1153 mspace create_mspace_with_base(void* base, size_t capacity, int locked);
1154
1155 /*
1156 mspace_mmap_large_chunks controls whether requests for large chunks
1157 are allocated in their own mmapped regions, separate from others in
1158 this mspace. By default this is enabled, which reduces
1159 fragmentation. However, such chunks are not necessarily released to
1160 the system upon destroy_mspace. Disabling by setting to false may
1161 increase fragmentation, but avoids leakage when relying on
1162 destroy_mspace to release all memory allocated using this space.
1163 */
1164 int mspace_mmap_large_chunks(mspace msp, int enable);
1165
1166
1167 /*
1168 mspace_malloc behaves as malloc, but operates within
1169 the given space.
1170 */
1171 void* mspace_malloc(mspace msp, size_t bytes);
1172
1173 /*
1174 mspace_free behaves as free, but operates within
1175 the given space.
1176
1177 If compiled with FOOTERS==1, mspace_free is not actually needed.
1178 free may be called instead of mspace_free because freed chunks from
1179 any space are handled by their originating spaces.
1180 */
1181 void mspace_free(mspace msp, void* mem);
1182
1183 /*
1184 mspace_realloc behaves as realloc, but operates within
1185 the given space.
1186
1187 If compiled with FOOTERS==1, mspace_realloc is not actually
1188 needed. realloc may be called instead of mspace_realloc because
1189 realloced chunks from any space are handled by their originating
1190 spaces.
1191 */
1192 void* mspace_realloc(mspace msp, void* mem, size_t newsize);
1193
1194 /*
1195 mspace_calloc behaves as calloc, but operates within
1196 the given space.
1197 */
1198 void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size);
1199
1200 /*
1201 mspace_memalign behaves as memalign, but operates within
1202 the given space.
1203 */
1204 void* mspace_memalign(mspace msp, size_t alignment, size_t bytes);
1205
1206 /*
1207 mspace_independent_calloc behaves as independent_calloc, but
1208 operates within the given space.
1209 */
1210 void** mspace_independent_calloc(mspace msp, size_t n_elements,
1211 size_t elem_size, void* chunks[]);
1212
1213 /*
1214 mspace_independent_comalloc behaves as independent_comalloc, but
1215 operates within the given space.
1216 */
1217 void** mspace_independent_comalloc(mspace msp, size_t n_elements,
1218 size_t sizes[], void* chunks[]);
1219
1220 /*
1221 mspace_footprint() returns the number of bytes obtained from the
1222 system for this space.
1223 */
1224 size_t mspace_footprint(mspace msp);
1225
1226 /*
1227 mspace_max_footprint() returns the peak number of bytes obtained from the
1228 system for this space.
1229 */
1230 size_t mspace_max_footprint(mspace msp);
1231
1232
1233 #if !NO_MALLINFO
1234 /*
1235 mspace_mallinfo behaves as mallinfo, but reports properties of
1236 the given space.
1237 */
1238 struct mallinfo mspace_mallinfo(mspace msp);
1239 #endif /* NO_MALLINFO */
1240
1241 /*
1242 malloc_usable_size(void* p) behaves the same as malloc_usable_size;
1243 */
1244 size_t mspace_usable_size(void* mem);
1245
1246 /*
1247 mspace_malloc_stats behaves as malloc_stats, but reports
1248 properties of the given space.
1249 */
1250 void mspace_malloc_stats(mspace msp);
1251
1252 /*
1253 mspace_trim behaves as malloc_trim, but
1254 operates within the given space.
1255 */
1256 int mspace_trim(mspace msp, size_t pad);
1257
1258 /*
1259 An alias for mallopt.
1260 */
1261 int mspace_mallopt(int, int);
1262
1263 #endif /* MSPACES */
1264
1265 #ifdef __cplusplus
1266 }; /* end of extern "C" */
1267 #endif /* __cplusplus */
1268
1269 /*
1270 ========================================================================
1271 To make a fully customizable malloc.h header file, cut everything
1272 above this line, put into file malloc.h, edit to suit, and #include it
1273 on the next line, as well as in programs that use this malloc.
1274 ========================================================================
1275 */
1276
1277 /* #include "malloc.h" */
1278
1279 /*------------------------------ internal #includes ---------------------- */
1280
1281 #ifdef WIN32
1282 #ifndef __GNUC__
1283 #pragma warning( disable : 4146 ) /* no "unsigned" warnings */
1284 #endif
1285 #endif /* WIN32 */
1286
1287 #include <stdio.h> /* for printing in malloc_stats */
1288
1289 #ifndef LACKS_ERRNO_H
1290 #include <errno.h> /* for MALLOC_FAILURE_ACTION */
1291 #endif /* LACKS_ERRNO_H */
1292 #if FOOTERS
1293 #include <time.h> /* for magic initialization */
1294 #endif /* FOOTERS */
1295 #ifndef LACKS_STDLIB_H
1296 #include <stdlib.h> /* for abort() */
1297 #endif /* LACKS_STDLIB_H */
1298 #ifdef DEBUG
1299 #if ABORT_ON_ASSERT_FAILURE
1300 #define assert(x) if(!(x)) ABORT
1301 #else /* ABORT_ON_ASSERT_FAILURE */
1302 #include <assert.h>
1303 #endif /* ABORT_ON_ASSERT_FAILURE */
1304 #else /* DEBUG */
1305 #ifndef assert
1306 #define assert(x)
1307 #endif
1308 #define DEBUG 0
1309 #endif /* DEBUG */
1310 #ifndef LACKS_STRING_H
1311 #include <string.h> /* for memset etc */
1312 #endif /* LACKS_STRING_H */
1313 #if USE_BUILTIN_FFS
1314 #ifndef LACKS_STRINGS_H
1315 #include <strings.h> /* for ffs */
1316 #endif /* LACKS_STRINGS_H */
1317 #endif /* USE_BUILTIN_FFS */
1318 #if HAVE_MMAP
1319 #ifndef LACKS_SYS_MMAN_H
1320 #include <sys/mman.h> /* for mmap */
1321 #endif /* LACKS_SYS_MMAN_H */
1322 #ifndef LACKS_FCNTL_H
1323 #include <fcntl.h>
1324 #endif /* LACKS_FCNTL_H */
1325 #endif /* HAVE_MMAP */
1326 #ifndef LACKS_UNISTD_H
1327 #include <unistd.h> /* for sbrk, sysconf */
1328 #else /* LACKS_UNISTD_H */
1329 #if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
1330 extern void* sbrk(ptrdiff_t);
1331 #endif /* FreeBSD etc */
1332 #endif /* LACKS_UNISTD_H */
1333
1334 /* Declarations for locking */
1335 #if USE_LOCKS
1336 #ifndef WIN32
1337 #include <pthread.h>
1338 #if defined (__SVR4) && defined (__sun) /* solaris */
1339 #include <thread.h>
1340 #endif /* solaris */
1341 #else
1342 #ifndef _M_AMD64
1343 /* These are already defined on AMD64 builds */
1344 #ifdef __cplusplus
1345 extern "C" {
1346 #endif /* __cplusplus */
1347 #ifndef __MINGW32__
1348 LONG __cdecl _InterlockedCompareExchange(LONG volatile *Dest, LONG Exchange, LONG Comp);
1349 LONG __cdecl _InterlockedExchange(LONG volatile *Target, LONG Value);
1350 #endif
1351 #ifdef __cplusplus
1352 }
1353 #endif /* __cplusplus */
1354 #endif /* _M_AMD64 */
1355 #ifndef __MINGW32__
1356 #pragma intrinsic (_InterlockedCompareExchange)
1357 #pragma intrinsic (_InterlockedExchange)
1358 #else
1359 /* --[ start GCC compatibility ]----------------------------------------------
1360 * Compatibility <intrin_x86.h> header for GCC -- GCC equivalents of intrinsic
1361 * Microsoft Visual C++ functions. Originally developed for the ReactOS
1362 * (<http://www.reactos.org/>) and TinyKrnl (<http://www.tinykrnl.org/>)
1363 * projects.
1364 *
1365 * Copyright (c) 2006 KJK::Hyperion <hackbunny@reactos.com>
1366 *
1367 * Permission is hereby granted, free of charge, to any person obtaining a
1368 * copy of this software and associated documentation files (the "Software"),
1369 * to deal in the Software without restriction, including without limitation
1370 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
1371 * and/or sell copies of the Software, and to permit persons to whom the
1372 * Software is furnished to do so, subject to the following conditions:
1373 *
1374 * The above copyright notice and this permission notice shall be included in
1375 * all copies or substantial portions of the Software.
1376 *
1377 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
1378 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
1379 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
1380 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
1381 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
1382 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
1383 * DEALINGS IN THE SOFTWARE.
1384 */
1385
1386 /*** Atomic operations ***/
1387 #if (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) > 40100
1388 #undef _ReadWriteBarrier
1389 #define _ReadWriteBarrier() __sync_synchronize()
1390 #else
1391 static __inline__ __attribute__((always_inline)) long __sync_lock_test_and_set(volatile long * const Target, const long Value)
1392 {
1393 long res;
1394 __asm__ __volatile__("xchg%z0 %2, %0" : "=g" (*(Target)), "=r" (res) : "1" (Value));
1395 return res;
1396 }
1397 static void __inline__ __attribute__((always_inline)) _MemoryBarrier(void)
1398 {
1399 __asm__ __volatile__("" : : : "memory");
1400 }
1401 #define _ReadWriteBarrier() _MemoryBarrier()
1402 #endif
1403 /* BUGBUG: GCC only supports full barriers */
1404 static __inline__ __attribute__((always_inline)) long _InterlockedExchange(volatile long * const Target, const long Value)
1405 {
1406 /* NOTE: __sync_lock_test_and_set would be an acquire barrier, so we force a full barrier */
1407 _ReadWriteBarrier();
1408 return __sync_lock_test_and_set(Target, Value);
1409 }
1410 /* --[ end GCC compatibility ]---------------------------------------------- */
1411 #endif
1412 #define interlockedcompareexchange _InterlockedCompareExchange
1413 #define interlockedexchange _InterlockedExchange
1414 #endif /* Win32 */
1415 #endif /* USE_LOCKS */
1416
1417 /* Declarations for bit scanning on win32 */
1418 #if defined(_MSC_VER) && _MSC_VER>=1300
1419 #ifndef BitScanForward /* Try to avoid pulling in WinNT.h */
1420 #ifdef __cplusplus
1421 extern "C" {
1422 #endif /* __cplusplus */
1423 unsigned char _BitScanForward(unsigned long *index, unsigned long mask);
1424 unsigned char _BitScanReverse(unsigned long *index, unsigned long mask);
1425 #ifdef __cplusplus
1426 }
1427 #endif /* __cplusplus */
1428
1429 #define BitScanForward _BitScanForward
1430 #define BitScanReverse _BitScanReverse
1431 #pragma intrinsic(_BitScanForward)
1432 #pragma intrinsic(_BitScanReverse)
1433 #endif /* BitScanForward */
1434 #endif /* defined(_MSC_VER) && _MSC_VER>=1300 */
1435
1436 #ifndef WIN32
1437 #ifndef malloc_getpagesize
1438 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
1439 # ifndef _SC_PAGE_SIZE
1440 # define _SC_PAGE_SIZE _SC_PAGESIZE
1441 # endif
1442 # endif
1443 # ifdef _SC_PAGE_SIZE
1444 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
1445 # else
1446 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
1447 extern size_t getpagesize();
1448 # define malloc_getpagesize getpagesize()
1449 # else
1450 # ifdef WIN32 /* use supplied emulation of getpagesize */
1451 # define malloc_getpagesize getpagesize()
1452 # else
1453 # ifndef LACKS_SYS_PARAM_H
1454 # include <sys/param.h>
1455 # endif
1456 # ifdef EXEC_PAGESIZE
1457 # define malloc_getpagesize EXEC_PAGESIZE
1458 # else
1459 # ifdef NBPG
1460 # ifndef CLSIZE
1461 # define malloc_getpagesize NBPG
1462 # else
1463 # define malloc_getpagesize (NBPG * CLSIZE)
1464 # endif
1465 # else
1466 # ifdef NBPC
1467 # define malloc_getpagesize NBPC
1468 # else
1469 # ifdef PAGESIZE
1470 # define malloc_getpagesize PAGESIZE
1471 # else /* just guess */
1472 # define malloc_getpagesize ((size_t)4096U)
1473 # endif
1474 # endif
1475 # endif
1476 # endif
1477 # endif
1478 # endif
1479 # endif
1480 #endif
1481 #endif
1482
1483
1484
1485 /* ------------------- size_t and alignment properties -------------------- */
1486
1487 /* The byte and bit size of a size_t */
1488 #define SIZE_T_SIZE (sizeof(size_t))
1489 #define SIZE_T_BITSIZE (sizeof(size_t) << 3)
1490
1491 /* Some constants coerced to size_t */
1492 /* Annoying but necessary to avoid errors on some platforms */
1493 #define SIZE_T_ZERO ((size_t)0)
1494 #define SIZE_T_ONE ((size_t)1)
1495 #define SIZE_T_TWO ((size_t)2)
1496 #define SIZE_T_FOUR ((size_t)4)
1497 #define TWO_SIZE_T_SIZES (SIZE_T_SIZE<<1)
1498 #define FOUR_SIZE_T_SIZES (SIZE_T_SIZE<<2)
1499 #define SIX_SIZE_T_SIZES (FOUR_SIZE_T_SIZES+TWO_SIZE_T_SIZES)
1500 #define HALF_MAX_SIZE_T (MAX_SIZE_T / 2U)
1501
1502 /* The bit mask value corresponding to MALLOC_ALIGNMENT */
1503 #define CHUNK_ALIGN_MASK (MALLOC_ALIGNMENT - SIZE_T_ONE)
1504
1505 /* True if address a has acceptable alignment */
1506 #define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0)
1507
1508 /* the number of bytes to offset an address to align it */
1509 #define align_offset(A)\
1510 ((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\
1511 ((MALLOC_ALIGNMENT - ((size_t)(A) & CHUNK_ALIGN_MASK)) & CHUNK_ALIGN_MASK))
1512
1513 /* -------------------------- MMAP preliminaries ------------------------- */
1514
1515 /*
1516 If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and
1517 checks to fail so compiler optimizer can delete code rather than
1518 using so many "#if"s.
1519 */
1520
1521
1522 /* MORECORE and MMAP must return MFAIL on failure */
1523 #define MFAIL ((void*)(MAX_SIZE_T))
1524 #define CMFAIL ((char*)(MFAIL)) /* defined for convenience */
1525
1526 #if HAVE_MMAP
1527
1528 #ifndef WIN32
1529 #define MUNMAP_DEFAULT(a, s) munmap((a), (s))
1530 #define MMAP_PROT (PROT_READ|PROT_WRITE)
1531 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
1532 #define MAP_ANONYMOUS MAP_ANON
1533 #endif /* MAP_ANON */
1534 #ifdef MAP_ANONYMOUS
1535 #define MMAP_FLAGS (MAP_PRIVATE|MAP_ANONYMOUS)
1536 #define MMAP_DEFAULT(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0)
1537 #else /* MAP_ANONYMOUS */
1538 /*
1539 Nearly all versions of mmap support MAP_ANONYMOUS, so the following
1540 is unlikely to be needed, but is supplied just in case.
1541 */
1542 #define MMAP_FLAGS (MAP_PRIVATE)
1543 static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */
1544 #define MMAP_DEFAULT(s) ((dev_zero_fd < 0) ? \
1545 (dev_zero_fd = open("/dev/zero", O_RDWR), \
1546 mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \
1547 mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0))
1548 #endif /* MAP_ANONYMOUS */
1549
1550 #define DIRECT_MMAP_DEFAULT(s) MMAP_DEFAULT(s)
1551
1552 #else /* WIN32 */
1553
1554 /* Win32 MMAP via VirtualAlloc */
1555 static FORCEINLINE void* win32mmap(size_t size) {
1556 void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE);
1557 return (ptr != 0)? ptr: MFAIL;
1558 }
1559
1560 /* For direct MMAP, use MEM_TOP_DOWN to minimize interference */
1561 static FORCEINLINE void* win32direct_mmap(size_t size) {
1562 void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN,
1563 PAGE_READWRITE);
1564 return (ptr != 0)? ptr: MFAIL;
1565 }
1566
1567 /* This function supports releasing coalesed segments */
1568 static FORCEINLINE int win32munmap(void* ptr, size_t size) {
1569 MEMORY_BASIC_INFORMATION minfo;
1570 char* cptr = (char*)ptr;
1571 while (size) {
1572 if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0)
1573 return -1;
1574 if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr ||
1575 minfo.State != MEM_COMMIT || minfo.RegionSize > size)
1576 return -1;
1577 if (VirtualFree(cptr, 0, MEM_RELEASE) == 0)
1578 return -1;
1579 cptr += minfo.RegionSize;
1580 size -= minfo.RegionSize;
1581 }
1582 return 0;
1583 }
1584
1585 #define MMAP_DEFAULT(s) win32mmap(s)
1586 #define MUNMAP_DEFAULT(a, s) win32munmap((a), (s))
1587 #define DIRECT_MMAP_DEFAULT(s) win32direct_mmap(s)
1588 #endif /* WIN32 */
1589 #endif /* HAVE_MMAP */
1590
1591 #if HAVE_MREMAP
1592 #ifndef WIN32
1593 #define MREMAP_DEFAULT(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv))
1594 #endif /* WIN32 */
1595 #endif /* HAVE_MREMAP */
1596
1597
1598 /**
1599 * Define CALL_MORECORE
1600 */
1601 #if HAVE_MORECORE
1602 #ifdef MORECORE
1603 #define CALL_MORECORE(S) MORECORE(S)
1604 #else /* MORECORE */
1605 #define CALL_MORECORE(S) MORECORE_DEFAULT(S)
1606 #endif /* MORECORE */
1607 #else /* HAVE_MORECORE */
1608 #define CALL_MORECORE(S) MFAIL
1609 #endif /* HAVE_MORECORE */
1610
1611 /**
1612 * Define CALL_MMAP/CALL_MUNMAP/CALL_DIRECT_MMAP
1613 */
1614 #if HAVE_MMAP
1615 #define IS_MMAPPED_BIT (SIZE_T_ONE)
1616 #define USE_MMAP_BIT (SIZE_T_ONE)
1617
1618 #ifdef MMAP
1619 #define CALL_MMAP(s) MMAP(s)
1620 #else /* MMAP */
1621 #define CALL_MMAP(s) MMAP_DEFAULT(s)
1622 #endif /* MMAP */
1623 #ifdef MUNMAP
1624 #define CALL_MUNMAP(a, s) MUNMAP((a), (s))
1625 #else /* MUNMAP */
1626 #define CALL_MUNMAP(a, s) MUNMAP_DEFAULT((a), (s))
1627 #endif /* MUNMAP */
1628 #ifdef DIRECT_MMAP
1629 #define CALL_DIRECT_MMAP(s) DIRECT_MMAP(s)
1630 #else /* DIRECT_MMAP */
1631 #define CALL_DIRECT_MMAP(s) DIRECT_MMAP_DEFAULT(s)
1632 #endif /* DIRECT_MMAP */
1633 #else /* HAVE_MMAP */
1634 #define IS_MMAPPED_BIT (SIZE_T_ZERO)
1635 #define USE_MMAP_BIT (SIZE_T_ZERO)
1636
1637 #define MMAP(s) MFAIL
1638 #define MUNMAP(a, s) (-1)
1639 #define DIRECT_MMAP(s) MFAIL
1640 #define CALL_DIRECT_MMAP(s) DIRECT_MMAP(s)
1641 #define CALL_MMAP(s) MMAP(s)
1642 #define CALL_MUNMAP(a, s) MUNMAP((a), (s))
1643 #endif /* HAVE_MMAP */
1644
1645 /**
1646 * Define CALL_MREMAP
1647 */
1648 #if HAVE_MMAP && HAVE_MREMAP
1649 #ifdef MREMAP
1650 #define CALL_MREMAP(addr, osz, nsz, mv) MREMAP((addr), (osz), (nsz), (mv))
1651 #else /* MREMAP */
1652 #define CALL_MREMAP(addr, osz, nsz, mv) MREMAP_DEFAULT((addr), (osz), (nsz), (mv))
1653 #endif /* MREMAP */
1654 #else /* HAVE_MMAP && HAVE_MREMAP */
1655 #define CALL_MREMAP(addr, osz, nsz, mv) MFAIL
1656 #endif /* HAVE_MMAP && HAVE_MREMAP */
1657
1658 /* mstate bit set if continguous morecore disabled or failed */
1659 #define USE_NONCONTIGUOUS_BIT (4U)
1660
1661 /* segment bit set in create_mspace_with_base */
1662 #define EXTERN_BIT (8U)
1663
1664
1665 /* --------------------------- Lock preliminaries ------------------------ */
1666
1667 /*
1668 When locks are defined, there is one global lock, plus
1669 one per-mspace lock.
1670
1671 The global lock_ensures that mparams.magic and other unique
1672 mparams values are initialized only once. It also protects
1673 sequences of calls to MORECORE. In many cases sys_alloc requires
1674 two calls, that should not be interleaved with calls by other
1675 threads. This does not protect against direct calls to MORECORE
1676 by other threads not using this lock, so there is still code to
1677 cope the best we can on interference.
1678
1679 Per-mspace locks surround calls to malloc, free, etc. To enable use
1680 in layered extensions, per-mspace locks are reentrant.
1681
1682 Because lock-protected regions generally have bounded times, it is
1683 OK to use the supplied simple spinlocks in the custom versions for
1684 x86.
1685
1686 If USE_LOCKS is > 1, the definitions of lock routines here are
1687 bypassed, in which case you will need to define at least
1688 INITIAL_LOCK, ACQUIRE_LOCK, RELEASE_LOCK and possibly TRY_LOCK
1689 (which is not used in this malloc, but commonly needed in
1690 extensions.)
1691 */
1692
1693 #if USE_LOCKS == 1
1694
1695 #if USE_SPIN_LOCKS
1696 #ifndef WIN32
1697
1698 /* Custom pthread-style spin locks on x86 and x64 for gcc */
1699 struct pthread_mlock_t {
1700 volatile unsigned int l;
1701 volatile unsigned int c;
1702 volatile pthread_t threadid;
1703 };
1704 #define MLOCK_T struct pthread_mlock_t
1705 #define CURRENT_THREAD pthread_self()
1706 #define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), 0)
1707 #define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl)
1708 #define RELEASE_LOCK(sl) pthread_release_lock(sl)
1709 #define TRY_LOCK(sl) pthread_try_lock(sl)
1710 #define SPINS_PER_YIELD 63
1711
1712 static MLOCK_T malloc_global_mutex = { 0, 0, 0};
1713
1714 static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) {
1715 int spins = 0;
1716 volatile unsigned int* lp = &sl->l;
1717 for (;;) {
1718 if (*lp != 0) {
1719 if (sl->threadid == CURRENT_THREAD) {
1720 ++sl->c;
1721 return 0;
1722 }
1723 }
1724 else {
1725 /* place args to cmpxchgl in locals to evade oddities in some gccs */
1726 int cmp = 0;
1727 int val = 1;
1728 int ret;
1729 __asm__ __volatile__ ("lock; cmpxchgl %1, %2"
1730 : "=a" (ret)
1731 : "r" (val), "m" (*(lp)), "0"(cmp)
1732 : "memory", "cc");
1733 if (!ret) {
1734 assert(!sl->threadid);
1735 sl->c = 1;
1736 sl->threadid = CURRENT_THREAD;
1737 return 0;
1738 }
1739 if ((++spins & SPINS_PER_YIELD) == 0) {
1740 #if defined (__SVR4) && defined (__sun) /* solaris */
1741 thr_yield();
1742 #else
1743 #if defined(__linux__) || defined(__FreeBSD__) || defined(__APPLE__)
1744 sched_yield();
1745 #else /* no-op yield on unknown systems */
1746 ;
1747 #endif /* __linux__ || __FreeBSD__ || __APPLE__ */
1748 #endif /* solaris */
1749 }
1750 }
1751 }
1752 }
1753
1754 static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) {
1755 assert(sl->l != 0);
1756 assert(sl->threadid == CURRENT_THREAD);
1757 if (--sl->c == 0) {
1758 volatile unsigned int* lp = &sl->l;
1759 int prev = 0;
1760 int ret;
1761 sl->threadid = 0;
1762 __asm__ __volatile__ ("lock; xchgl %0, %1"
1763 : "=r" (ret)
1764 : "m" (*(lp)), "0"(prev)
1765 : "memory");
1766 }
1767 }
1768
1769 static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) {
1770 volatile unsigned int* lp = &sl->l;
1771 if (*lp != 0) {
1772 if (sl->threadid == CURRENT_THREAD) {
1773 ++sl->c;
1774 return 1;
1775 }
1776 }
1777 else {
1778 int cmp = 0;
1779 int val = 1;
1780 int ret;
1781 __asm__ __volatile__ ("lock; cmpxchgl %1, %2"
1782 : "=a" (ret)
1783 : "r" (val), "m" (*(lp)), "0"(cmp)
1784 : "memory", "cc");
1785 if (!ret) {
1786 assert(!sl->threadid);
1787 sl->c = 1;
1788 sl->threadid = CURRENT_THREAD;
1789 return 1;
1790 }
1791 }
1792 return 0;
1793 }
1794
1795
1796 #else /* WIN32 */
1797 /* Custom win32-style spin locks on x86 and x64 for MSC */
1798 struct win32_mlock_t
1799 {
1800 volatile long l;
1801 volatile unsigned int c;
1802 volatile long threadid;
1803 };
1804
1805 static inline int return_0(int i) { return 0; }
1806 #define MLOCK_T struct win32_mlock_t
1807 #define CURRENT_THREAD win32_getcurrentthreadid()
1808 #define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), return_0(0))
1809 #define ACQUIRE_LOCK(sl) win32_acquire_lock(sl)
1810 #define RELEASE_LOCK(sl) win32_release_lock(sl)
1811 #define TRY_LOCK(sl) win32_try_lock(sl)
1812 #define SPINS_PER_YIELD 63
1813
1814 static MLOCK_T malloc_global_mutex = { 0, 0, 0};
1815
1816 static FORCEINLINE long win32_getcurrentthreadid(void) {
1817 #ifdef _MSC_VER
1818 #if defined(_M_IX86)
1819 long *threadstruct=(long *)__readfsdword(0x18);
1820 long threadid=threadstruct[0x24/sizeof(long)];
1821 return threadid;
1822 #elif defined(_M_X64)
1823 /* todo */
1824 return GetCurrentThreadId();
1825 #else
1826 return GetCurrentThreadId();
1827 #endif
1828 #else
1829 return GetCurrentThreadId();
1830 #endif
1831 }
1832
1833 static FORCEINLINE int win32_acquire_lock (MLOCK_T *sl) {
1834 int spins = 0;
1835 for (;;) {
1836 if (sl->l != 0) {
1837 if (sl->threadid == CURRENT_THREAD) {
1838 ++sl->c;
1839 return 0;
1840 }
1841 }
1842 else {
1843 if (!interlockedexchange(&sl->l, 1)) {
1844 assert(!sl->threadid);
1845 sl->c=CURRENT_THREAD;
1846 sl->threadid = CURRENT_THREAD;
1847 sl->c = 1;
1848 return 0;
1849 }
1850 }
1851 if ((++spins & SPINS_PER_YIELD) == 0)
1852 SleepEx(0, FALSE);
1853 }
1854 }
1855
1856 static FORCEINLINE void win32_release_lock (MLOCK_T *sl) {
1857 assert(sl->threadid == CURRENT_THREAD);
1858 assert(sl->l != 0);
1859 if (--sl->c == 0) {
1860 sl->threadid = 0;
1861 interlockedexchange (&sl->l, 0);
1862 }
1863 }
1864
1865 static FORCEINLINE int win32_try_lock (MLOCK_T *sl) {
1866 if(sl->l != 0) {
1867 if (sl->threadid == CURRENT_THREAD) {
1868 ++sl->c;
1869 return 1;
1870 }
1871 }
1872 else {
1873 if (!interlockedexchange(&sl->l, 1)){
1874 assert(!sl->threadid);
1875 sl->threadid = CURRENT_THREAD;
1876 sl->c = 1;
1877 return 1;
1878 }
1879 }
1880 return 0;
1881 }
1882
1883 #endif /* WIN32 */
1884 #else /* USE_SPIN_LOCKS */
1885
1886 #ifndef WIN32
1887 /* pthreads-based locks */
1888
1889 #define MLOCK_T pthread_mutex_t
1890 #define CURRENT_THREAD pthread_self()
1891 #define INITIAL_LOCK(sl) pthread_init_lock(sl)
1892 #define ACQUIRE_LOCK(sl) pthread_mutex_lock(sl)
1893 #define RELEASE_LOCK(sl) pthread_mutex_unlock(sl)
1894 #define TRY_LOCK(sl) (!pthread_mutex_trylock(sl))
1895
1896 static MLOCK_T malloc_global_mutex = PTHREAD_MUTEX_INITIALIZER;
1897
1898 /* Cope with old-style linux recursive lock initialization by adding */
1899 /* skipped internal declaration from pthread.h */
1900 #ifdef linux
1901 #ifndef PTHREAD_MUTEX_RECURSIVE
1902 extern int pthread_mutexattr_setkind_np __P ((pthread_mutexattr_t *__attr,
1903 int __kind));
1904 #define PTHREAD_MUTEX_RECURSIVE PTHREAD_MUTEX_RECURSIVE_NP
1905 #define pthread_mutexattr_settype(x,y) pthread_mutexattr_setkind_np(x,y)
1906 #endif
1907 #endif
1908
1909 static int pthread_init_lock (MLOCK_T *sl) {
1910 pthread_mutexattr_t attr;
1911 if (pthread_mutexattr_init(&attr)) return 1;
1912 if (pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) return 1;
1913 if (pthread_mutex_init(sl, &attr)) return 1;
1914 if (pthread_mutexattr_destroy(&attr)) return 1;
1915 return 0;
1916 }
1917
1918 #else /* WIN32 */
1919 /* Win32 critical sections */
1920 #define MLOCK_T CRITICAL_SECTION
1921 #define CURRENT_THREAD GetCurrentThreadId()
1922 #define INITIAL_LOCK(s) (!InitializeCriticalSectionAndSpinCount((s), 0x80000000|4000))
1923 #define ACQUIRE_LOCK(s) (EnterCriticalSection(s), 0)
1924 #define RELEASE_LOCK(s) LeaveCriticalSection(s)
1925 #define TRY_LOCK(s) TryEnterCriticalSection(s)
1926 #define NEED_GLOBAL_LOCK_INIT
1927
1928 static MLOCK_T malloc_global_mutex;
1929 static volatile long malloc_global_mutex_status;
1930
1931 /* Use spin loop to initialize global lock */
1932 static void init_malloc_global_mutex() {
1933 for (;;) {
1934 long stat = malloc_global_mutex_status;
1935 if (stat > 0)
1936 return;
1937 /* transition to < 0 while initializing, then to > 0) */
1938 if (stat == 0 &&
1939 interlockedcompareexchange(&malloc_global_mutex_status, -1, 0) == 0) {
1940 InitializeCriticalSection(&malloc_global_mutex);
1941 interlockedexchange(&malloc_global_mutex_status,1);
1942 return;
1943 }
1944 SleepEx(0, FALSE);
1945 }
1946 }
1947
1948 #endif /* WIN32 */
1949 #endif /* USE_SPIN_LOCKS */
1950 #endif /* USE_LOCKS == 1 */
1951
1952 /* ----------------------- User-defined locks ------------------------ */
1953
1954 #if USE_LOCKS > 1
1955 /* Define your own lock implementation here */
1956 /* #define INITIAL_LOCK(sl) ... */
1957 /* #define ACQUIRE_LOCK(sl) ... */
1958 /* #define RELEASE_LOCK(sl) ... */
1959 /* #define TRY_LOCK(sl) ... */
1960 /* static MLOCK_T malloc_global_mutex = ... */
1961 #endif /* USE_LOCKS > 1 */
1962
1963 /* ----------------------- Lock-based state ------------------------ */
1964
1965 #if USE_LOCKS
1966 #define USE_LOCK_BIT (2U)
1967 #else /* USE_LOCKS */
1968 #define USE_LOCK_BIT (0U)
1969 #define INITIAL_LOCK(l)
1970 #endif /* USE_LOCKS */
1971
1972 #if USE_LOCKS
1973 #define ACQUIRE_MALLOC_GLOBAL_LOCK() ACQUIRE_LOCK(&malloc_global_mutex);
1974 #define RELEASE_MALLOC_GLOBAL_LOCK() RELEASE_LOCK(&malloc_global_mutex);
1975 #else /* USE_LOCKS */
1976 #define ACQUIRE_MALLOC_GLOBAL_LOCK()
1977 #define RELEASE_MALLOC_GLOBAL_LOCK()
1978 #endif /* USE_LOCKS */
1979
1980
1981 /* ----------------------- Chunk representations ------------------------ */
1982
1983 /*
1984 (The following includes lightly edited explanations by Colin Plumb.)
1985
1986 The malloc_chunk declaration below is misleading (but accurate and
1987 necessary). It declares a "view" into memory allowing access to
1988 necessary fields at known offsets from a given base.
1989
1990 Chunks of memory are maintained using a `boundary tag' method as
1991 originally described by Knuth. (See the paper by Paul Wilson
1992 ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a survey of such
1993 techniques.) Sizes of free chunks are stored both in the front of
1994 each chunk and at the end. This makes consolidating fragmented
1995 chunks into bigger chunks fast. The head fields also hold bits
1996 representing whether chunks are free or in use.
1997
1998 Here are some pictures to make it clearer. They are "exploded" to
1999 show that the state of a chunk can be thought of as extending from
2000 the high 31 bits of the head field of its header through the
2001 prev_foot and PINUSE_BIT bit of the following chunk header.
2002
2003 A chunk that's in use looks like:
2004
2005 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2006 | Size of previous chunk (if P = 0) |
2007 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2008 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
2009 | Size of this chunk 1| +-+
2010 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2011 | |
2012 +- -+
2013 | |
2014 +- -+
2015 | :
2016 +- size - sizeof(size_t) available payload bytes -+
2017 : |
2018 chunk-> +- -+
2019 | |
2020 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2021 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1|
2022 | Size of next chunk (may or may not be in use) | +-+
2023 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2024
2025 And if it's free, it looks like this:
2026
2027 chunk-> +- -+
2028 | User payload (must be in use, or we would have merged!) |
2029 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2030 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P|
2031 | Size of this chunk 0| +-+
2032 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2033 | Next pointer |
2034 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2035 | Prev pointer |
2036 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2037 | :
2038 +- size - sizeof(struct chunk) unused bytes -+
2039 : |
2040 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2041 | Size of this chunk |
2042 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2043 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0|
2044 | Size of next chunk (must be in use, or we would have merged)| +-+
2045 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2046 | :
2047 +- User payload -+
2048 : |
2049 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2050 |0|
2051 +-+
2052 Note that since we always merge adjacent free chunks, the chunks
2053 adjacent to a free chunk must be in use.
2054
2055 Given a pointer to a chunk (which can be derived trivially from the
2056 payload pointer) we can, in O(1) time, find out whether the adjacent
2057 chunks are free, and if so, unlink them from the lists that they
2058 are on and merge them with the current chunk.
2059
2060 Chunks always begin on even word boundaries, so the mem portion
2061 (which is returned to the user) is also on an even word boundary, and
2062 thus at least double-word aligned.
2063
2064 The P (PINUSE_BIT) bit, stored in the unused low-order bit of the
2065 chunk size (which is always a multiple of two words), is an in-use
2066 bit for the *previous* chunk. If that bit is *clear*, then the
2067 word before the current chunk size contains the previous chunk
2068 size, and can be used to find the front of the previous chunk.
2069 The very first chunk allocated always has this bit set, preventing
2070 access to non-existent (or non-owned) memory. If pinuse is set for
2071 any given chunk, then you CANNOT determine the size of the
2072 previous chunk, and might even get a memory addressing fault when
2073 trying to do so.
2074
2075 The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of
2076 the chunk size redundantly records whether the current chunk is
2077 inuse. This redundancy enables usage checks within free and realloc,
2078 and reduces indirection when freeing and consolidating chunks.
2079
2080 Each freshly allocated chunk must have both cinuse and pinuse set.
2081 That is, each allocated chunk borders either a previously allocated
2082 and still in-use chunk, or the base of its memory arena. This is
2083 ensured by making all allocations from the `lowest' part of any
2084 found chunk. Further, no free chunk physically borders another one,
2085 so each free chunk is known to be preceded and followed by either
2086 inuse chunks or the ends of memory.
2087
2088 Note that the `foot' of the current chunk is actually represented
2089 as the prev_foot of the NEXT chunk. This makes it easier to
2090 deal with alignments etc but can be very confusing when trying
2091 to extend or adapt this code.
2092
2093 The exceptions to all this are
2094
2095 1. The special chunk `top' is the top-most available chunk (i.e.,
2096 the one bordering the end of available memory). It is treated
2097 specially. Top is never included in any bin, is used only if
2098 no other chunk is available, and is released back to the
2099 system if it is very large (see M_TRIM_THRESHOLD). In effect,
2100 the top chunk is treated as larger (and thus less well
2101 fitting) than any other available chunk. The top chunk
2102 doesn't update its trailing size field since there is no next
2103 contiguous chunk that would have to index off it. However,
2104 space is still allocated for it (TOP_FOOT_SIZE) to enable
2105 separation or merging when space is extended.
2106
2107 3. Chunks allocated via mmap, which have the lowest-order bit
2108 (IS_MMAPPED_BIT) set in their prev_foot fields, and do not set
2109 PINUSE_BIT in their head fields. Because they are allocated
2110 one-by-one, each must carry its own prev_foot field, which is
2111 also used to hold the offset this chunk has within its mmapped
2112 region, which is needed to preserve alignment. Each mmapped
2113 chunk is trailed by the first two fields of a fake next-chunk
2114 for sake of usage checks.
2115
2116 */
2117
2118 struct malloc_chunk {
2119 size_t prev_foot; /* Size of previous chunk (if free). */
2120 size_t head; /* Size and inuse bits. */
2121 struct malloc_chunk* fd; /* double links -- used only if free. */
2122 struct malloc_chunk* bk;
2123 };
2124
2125 typedef struct malloc_chunk mchunk;
2126 typedef struct malloc_chunk* mchunkptr;
2127 typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */
2128 typedef unsigned int bindex_t; /* Described below */
2129 typedef unsigned int binmap_t; /* Described below */
2130 typedef unsigned int flag_t; /* The type of various bit flag sets */
2131
2132 /* ------------------- Chunks sizes and alignments ----------------------- */
2133
2134 #define MCHUNK_SIZE (sizeof(mchunk))
2135
2136 #if FOOTERS
2137 #define CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)
2138 #else /* FOOTERS */
2139 #define CHUNK_OVERHEAD (SIZE_T_SIZE)
2140 #endif /* FOOTERS */
2141
2142 /* MMapped chunks need a second word of overhead ... */
2143 #define MMAP_CHUNK_OVERHEAD (TWO_SIZE_T_SIZES)
2144 /* ... and additional padding for fake next-chunk at foot */
2145 #define MMAP_FOOT_PAD (FOUR_SIZE_T_SIZES)
2146
2147 /* The smallest size we can malloc is an aligned minimal chunk */
2148 #define MIN_CHUNK_SIZE\
2149 ((MCHUNK_SIZE + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)
2150
2151 /* conversion from malloc headers to user pointers, and back */
2152 #define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES))
2153 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES))
2154 /* chunk associated with aligned address A */
2155 #define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A)))
2156
2157 /* Bounds on request (not chunk) sizes. */
2158 #define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2)
2159 #define MIN_REQUEST (MIN_CHUNK_SIZE - CHUNK_OVERHEAD - SIZE_T_ONE)
2160
2161 /* pad request bytes into a usable size */
2162 #define pad_request(req) \
2163 (((req) + CHUNK_OVERHEAD + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK)
2164
2165 /* pad request, checking for minimum (but not maximum) */
2166 #define request2size(req) \
2167 (((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req))
2168
2169
2170 /* ------------------ Operations on head and foot fields ----------------- */
2171
2172 /*
2173 The head field of a chunk is or'ed with PINUSE_BIT when previous
2174 adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in
2175 use. If the chunk was obtained with mmap, the prev_foot field has
2176 IS_MMAPPED_BIT set, otherwise holding the offset of the base of the
2177 mmapped region to the base of the chunk.
2178
2179 FLAG4_BIT is not used by this malloc, but might be useful in extensions.
2180 */
2181
2182 #define PINUSE_BIT (SIZE_T_ONE)
2183 #define CINUSE_BIT (SIZE_T_TWO)
2184 #define FLAG4_BIT (SIZE_T_FOUR)
2185 #define INUSE_BITS (PINUSE_BIT|CINUSE_BIT)
2186 #define FLAG_BITS (PINUSE_BIT|CINUSE_BIT|FLAG4_BIT)
2187
2188 /* Head value for fenceposts */
2189 #define FENCEPOST_HEAD (INUSE_BITS|SIZE_T_SIZE)
2190
2191 /* extraction of fields from head words */
2192 #define cinuse(p) ((p)->head & CINUSE_BIT)
2193 #define pinuse(p) ((p)->head & PINUSE_BIT)
2194 #define chunksize(p) ((p)->head & ~(FLAG_BITS))
2195
2196 #define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT)
2197 #define clear_cinuse(p) ((p)->head &= ~CINUSE_BIT)
2198
2199 /* Treat space at ptr +/- offset as a chunk */
2200 #define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
2201 #define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s)))
2202
2203 /* Ptr to next or previous physical malloc_chunk. */
2204 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~FLAG_BITS)))
2205 #define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) ))
2206
2207 /* extract next chunk's pinuse bit */
2208 #define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT)
2209
2210 /* Get/set size at footer */
2211 #define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot)
2212 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s))
2213
2214 /* Set size, pinuse bit, and foot */
2215 #define set_size_and_pinuse_of_free_chunk(p, s)\
2216 ((p)->head = (s|PINUSE_BIT), set_foot(p, s))
2217
2218 /* Set size, pinuse bit, foot, and clear next pinuse */
2219 #define set_free_with_pinuse(p, s, n)\
2220 (clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s))
2221
2222 #define is_mmapped(p)\
2223 (!((p)->head & PINUSE_BIT) && ((p)->prev_foot & IS_MMAPPED_BIT))
2224
2225 /* Get the internal overhead associated with chunk p */
2226 #define overhead_for(p)\
2227 (is_mmapped(p)? MMAP_CHUNK_OVERHEAD : CHUNK_OVERHEAD)
2228
2229 /* Return true if malloced space is not necessarily cleared */
2230 #if MMAP_CLEARS
2231 #define calloc_must_clear(p) (!is_mmapped(p))
2232 #else /* MMAP_CLEARS */
2233 #define calloc_must_clear(p) (1)
2234 #endif /* MMAP_CLEARS */
2235
2236 /* ---------------------- Overlaid data structures ----------------------- */
2237
2238 /*
2239 When chunks are not in use, they are treated as nodes of either
2240 lists or trees.
2241
2242 "Small" chunks are stored in circular doubly-linked lists, and look
2243 like this:
2244
2245 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2246 | Size of previous chunk |
2247 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2248 `head:' | Size of chunk, in bytes |P|
2249 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2250 | Forward pointer to next chunk in list |
2251 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2252 | Back pointer to previous chunk in list |
2253 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2254 | Unused space (may be 0 bytes long) .
2255 . .
2256 . |
2257 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2258 `foot:' | Size of chunk, in bytes |
2259 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2260
2261 Larger chunks are kept in a form of bitwise digital trees (aka
2262 tries) keyed on chunksizes. Because malloc_tree_chunks are only for
2263 free chunks greater than 256 bytes, their size doesn't impose any
2264 constraints on user chunk sizes. Each node looks like:
2265
2266 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2267 | Size of previous chunk |
2268 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2269 `head:' | Size of chunk, in bytes |P|
2270 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2271 | Forward pointer to next chunk of same size |
2272 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2273 | Back pointer to previous chunk of same size |
2274 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2275 | Pointer to left child (child[0]) |
2276 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2277 | Pointer to right child (child[1]) |
2278 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2279 | Pointer to parent |
2280 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2281 | bin index of this chunk |
2282 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2283 | Unused space .
2284 . |
2285 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2286 `foot:' | Size of chunk, in bytes |
2287 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
2288
2289 Each tree holding treenodes is a tree of unique chunk sizes. Chunks
2290 of the same size are arranged in a circularly-linked list, with only
2291 the oldest chunk (the next to be used, in our FIFO ordering)
2292 actually in the tree. (Tree members are distinguished by a non-null
2293 parent pointer.) If a chunk with the same size as an existing node
2294 is inserted, it is linked off the existing node using pointers that
2295 work in the same way as fd/bk pointers of small chunks.
2296
2297 Each tree contains a power of 2 sized range of chunk sizes (the
2298 smallest is 0x100 <= x < 0x180), which is divided in half at each
2299 tree level, with the chunks in the smaller half of the range (0x100
2300 <= x < 0x140 for the top nose) in the left subtree and the larger
2301 half (0x140 <= x < 0x180) in the right subtree. This is, of course,
2302 done by inspecting individual bits.
2303
2304 Using these rules, each node's left subtree contains all smaller
2305 sizes than its right subtree. However, the node at the root of each
2306 subtree has no particular ordering relationship to either. (The
2307 dividing line between the subtree sizes is based on trie relation.)
2308 If we remove the last chunk of a given size from the interior of the
2309 tree, we need to replace it with a leaf node. The tree ordering
2310 rules permit a node to be replaced by any leaf below it.
2311
2312 The smallest chunk in a tree (a common operation in a best-fit
2313 allocator) can be found by walking a path to the leftmost leaf in
2314 the tree. Unlike a usual binary tree, where we follow left child
2315 pointers until we reach a null, here we follow the right child
2316 pointer any time the left one is null, until we reach a leaf with
2317 both child pointers null. The smallest chunk in the tree will be
2318 somewhere along that path.
2319
2320 The worst case number of steps to add, find, or remove a node is
2321 bounded by the number of bits differentiating chunks within
2322 bins. Under current bin calculations, this ranges from 6 up to 21
2323 (for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case
2324 is of course much better.
2325 */
2326
2327 struct malloc_tree_chunk {
2328 /* The first four fields must be compatible with malloc_chunk */
2329 size_t prev_foot;
2330 size_t head;
2331 struct malloc_tree_chunk* fd;
2332 struct malloc_tree_chunk* bk;
2333
2334 struct malloc_tree_chunk* child[2];
2335 struct malloc_tree_chunk* parent;
2336 bindex_t index;
2337 };
2338
2339 typedef struct malloc_tree_chunk tchunk;
2340 typedef struct malloc_tree_chunk* tchunkptr;
2341 typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */
2342
2343 /* A little helper macro for trees */
2344 #define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1])
2345
2346 /* ----------------------------- Segments -------------------------------- */
2347
2348 /*
2349 Each malloc space may include non-contiguous segments, held in a
2350 list headed by an embedded malloc_segment record representing the
2351 top-most space. Segments also include flags holding properties of
2352 the space. Large chunks that are directly allocated by mmap are not
2353 included in this list. They are instead independently created and
2354 destroyed without otherwise keeping track of them.
2355
2356 Segment management mainly comes into play for spaces allocated by
2357 MMAP. Any call to MMAP might or might not return memory that is
2358 adjacent to an existing segment. MORECORE normally contiguously
2359 extends the current space, so this space is almost always adjacent,
2360 which is simpler and faster to deal with. (This is why MORECORE is
2361 used preferentially to MMAP when both are available -- see
2362 sys_alloc.) When allocating using MMAP, we don't use any of the
2363 hinting mechanisms (inconsistently) supported in various
2364 implementations of unix mmap, or distinguish reserving from
2365 committing memory. Instead, we just ask for space, and exploit
2366 contiguity when we get it. It is probably possible to do
2367 better than this on some systems, but no general scheme seems
2368 to be significantly better.
2369
2370 Management entails a simpler variant of the consolidation scheme
2371 used for chunks to reduce fragmentation -- new adjacent memory is
2372 normally prepended or appended to an existing segment. However,
2373 there are limitations compared to chunk consolidation that mostly
2374 reflect the fact that segment processing is relatively infrequent
2375 (occurring only when getting memory from system) and that we
2376 don't expect to have huge numbers of segments:
2377
2378 * Segments are not indexed, so traversal requires linear scans. (It
2379 would be possible to index these, but is not worth the extra
2380 overhead and complexity for most programs on most platforms.)
2381 * New segments are only appended to old ones when holding top-most
2382 memory; if they cannot be prepended to others, they are held in
2383 different segments.
2384
2385 Except for the top-most segment of an mstate, each segment record
2386 is kept at the tail of its segment. Segments are added by pushing
2387 segment records onto the list headed by &mstate.seg for the
2388 containing mstate.
2389
2390 Segment flags control allocation/merge/deallocation policies:
2391 * If EXTERN_BIT set, then we did not allocate this segment,
2392 and so should not try to deallocate or merge with others.
2393 (This currently holds only for the initial segment passed
2394 into create_mspace_with_base.)
2395 * If IS_MMAPPED_BIT set, the segment may be merged with
2396 other surrounding mmapped segments and trimmed/de-allocated
2397 using munmap.
2398 * If neither bit is set, then the segment was obtained using
2399 MORECORE so can be merged with surrounding MORECORE'd segments
2400 and deallocated/trimmed using MORECORE with negative arguments.
2401 */
2402
2403 struct malloc_segment {
2404 char* base; /* base address */
2405 size_t size; /* allocated size */
2406 struct malloc_segment* next; /* ptr to next segment */
2407 flag_t sflags; /* mmap and extern flag */
2408 };
2409
2410 #define is_mmapped_segment(S) ((S)->sflags & IS_MMAPPED_BIT)
2411 #define is_extern_segment(S) ((S)->sflags & EXTERN_BIT)
2412
2413 typedef struct malloc_segment msegment;
2414 typedef struct malloc_segment* msegmentptr;
2415
2416 /* ---------------------------- malloc_state ----------------------------- */
2417
2418 /*
2419 A malloc_state holds all of the bookkeeping for a space.
2420 The main fields are:
2421
2422 Top
2423 The topmost chunk of the currently active segment. Its size is
2424 cached in topsize. The actual size of topmost space is
2425 topsize+TOP_FOOT_SIZE, which includes space reserved for adding
2426 fenceposts and segment records if necessary when getting more
2427 space from the system. The size at which to autotrim top is
2428 cached from mparams in trim_check, except that it is disabled if
2429 an autotrim fails.
2430
2431 Designated victim (dv)
2432 This is the preferred chunk for servicing small requests that
2433 don't have exact fits. It is normally the chunk split off most
2434 recently to service another small request. Its size is cached in
2435 dvsize. The link fields of this chunk are not maintained since it
2436 is not kept in a bin.
2437
2438 SmallBins
2439 An array of bin headers for free chunks. These bins hold chunks
2440 with sizes less than MIN_LARGE_SIZE bytes. Each bin contains
2441 chunks of all the same size, spaced 8 bytes apart. To simplify
2442 use in double-linked lists, each bin header acts as a malloc_chunk
2443 pointing to the real first node, if it exists (else pointing to
2444 itself). This avoids special-casing for headers. But to avoid
2445 waste, we allocate only the fd/bk pointers of bins, and then use
2446 repositioning tricks to treat these as the fields of a chunk.
2447
2448 TreeBins
2449 Treebins are pointers to the roots of trees holding a range of
2450 sizes. There are 2 equally spaced treebins for each power of two
2451 from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything
2452 larger.
2453
2454 Bin maps
2455 There is one bit map for small bins ("smallmap") and one for
2456 treebins ("treemap). Each bin sets its bit when non-empty, and
2457 clears the bit when empty. Bit operations are then used to avoid
2458 bin-by-bin searching -- nearly all "search" is done without ever
2459 looking at bins that won't be selected. The bit maps
2460 conservatively use 32 bits per map word, even if on 64bit system.
2461 For a good description of some of the bit-based techniques used
2462 here, see Henry S. Warren Jr's book "Hacker's Delight" (and
2463 supplement at http://hackersdelight.org/). Many of these are
2464 intended to reduce the branchiness of paths through malloc etc, as
2465 well as to reduce the number of memory locations read or written.
2466
2467 Segments
2468 A list of segments headed by an embedded malloc_segment record
2469 representing the initial space.
2470
2471 Address check support
2472 The least_addr field is the least address ever obtained from
2473 MORECORE or MMAP. Attempted frees and reallocs of any address less
2474 than this are trapped (unless INSECURE is defined).
2475
2476 Magic tag
2477 A cross-check field that should always hold same value as mparams.magic.
2478
2479 Flags
2480 Bits recording whether to use MMAP, locks, or contiguous MORECORE
2481
2482 Statistics
2483 Each space keeps track of current and maximum system memory
2484 obtained via MORECORE or MMAP.
2485
2486 Trim support
2487 Fields holding the amount of unused topmost memory that should trigger
2488 timming, and a counter to force periodic scanning to release unused
2489 non-topmost segments.
2490
2491 Locking
2492 If USE_LOCKS is defined, the "mutex" lock is acquired and released
2493 around every public call using this mspace.
2494
2495 Extension support
2496 A void* pointer and a size_t field that can be used to help implement
2497 extensions to this malloc.
2498 */
2499
2500 /* Bin types, widths and sizes */
2501 #define NSMALLBINS (32U)
2502 #define NTREEBINS (32U)
2503 #define SMALLBIN_SHIFT (3U)
2504 #define SMALLBIN_WIDTH (SIZE_T_ONE << SMALLBIN_SHIFT)
2505 #define TREEBIN_SHIFT (8U)
2506 #define MIN_LARGE_SIZE (SIZE_T_ONE << TREEBIN_SHIFT)
2507 #define MAX_SMALL_SIZE (MIN_LARGE_SIZE - SIZE_T_ONE)
2508 #define MAX_SMALL_REQUEST (MAX_SMALL_SIZE - CHUNK_ALIGN_MASK - CHUNK_OVERHEAD)
2509
2510 struct malloc_state {
2511 binmap_t smallmap;
2512 binmap_t treemap;
2513 size_t dvsize;
2514 size_t topsize;
2515 char* least_addr;
2516 mchunkptr dv;
2517 mchunkptr top;
2518 size_t trim_check;
2519 size_t release_checks;
2520 size_t magic;
2521 mchunkptr smallbins[(NSMALLBINS+1)*2];
2522 tbinptr treebins[NTREEBINS];
2523 size_t footprint;
2524 size_t max_footprint;
2525 flag_t mflags;
2526 #if USE_LOCKS
2527 MLOCK_T mutex; /* locate lock among fields that rarely change */
2528 #endif /* USE_LOCKS */
2529 msegment seg;
2530 void* extp; /* Unused but available for extensions */
2531 size_t exts;
2532 };
2533
2534 typedef struct malloc_state* mstate;
2535
2536 /* ------------- Global malloc_state and malloc_params ------------------- */
2537
2538 /*
2539 malloc_params holds global properties, including those that can be
2540 dynamically set using mallopt. There is a single instance, mparams,
2541 initialized in init_mparams. Note that the non-zeroness of "magic"
2542 also serves as an initialization flag.
2543 */
2544
2545 struct malloc_params {
2546 volatile size_t magic;
2547 size_t page_size;
2548 size_t granularity;
2549 size_t mmap_threshold;
2550 size_t trim_threshold;
2551 flag_t default_mflags;
2552 };
2553
2554 static struct malloc_params mparams;
2555
2556 /* Ensure mparams initialized */
2557 #define ensure_initialization() ((void)(mparams.magic != 0 || init_mparams()))
2558
2559 #if !ONLY_MSPACES
2560
2561 /* The global malloc_state used for all non-"mspace" calls */
2562 static struct malloc_state _gm_;
2563 #define gm (&_gm_)
2564 #define is_global(M) ((M) == &_gm_)
2565
2566 #endif /* !ONLY_MSPACES */
2567
2568 #define is_initialized(M) ((M)->top != 0)
2569
2570 /* -------------------------- system alloc setup ------------------------- */
2571
2572 /* Operations on mflags */
2573
2574 #define use_lock(M) ((M)->mflags & USE_LOCK_BIT)
2575 #define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT)
2576 #define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT)
2577
2578 #define use_mmap(M) ((M)->mflags & USE_MMAP_BIT)
2579 #define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT)
2580 #define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT)
2581
2582 #define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT)
2583 #define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT)
2584
2585 #define set_lock(M,L)\
2586 ((M)->mflags = (L)?\
2587 ((M)->mflags | USE_LOCK_BIT) :\
2588 ((M)->mflags & ~USE_LOCK_BIT))
2589
2590 /* page-align a size */
2591 #define page_align(S)\
2592 (((S) + (mparams.page_size - SIZE_T_ONE)) & ~(mparams.page_size - SIZE_T_ONE))
2593
2594 /* granularity-align a size */
2595 #define granularity_align(S)\
2596 (((S) + (mparams.granularity - SIZE_T_ONE))\
2597 & ~(mparams.granularity - SIZE_T_ONE))
2598
2599
2600 /* For mmap, use granularity alignment on windows, else page-align */
2601 #ifdef WIN32
2602 #define mmap_align(S) granularity_align(S)
2603 #else
2604 #define mmap_align(S) page_align(S)
2605 #endif
2606
2607 /* For sys_alloc, enough padding to ensure can malloc request on success */
2608 #define SYS_ALLOC_PADDING (TOP_FOOT_SIZE + MALLOC_ALIGNMENT)
2609
2610 #define is_page_aligned(S)\
2611 (((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0)
2612 #define is_granularity_aligned(S)\
2613 (((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0)
2614
2615 /* True if segment S holds address A */
2616 #define segment_holds(S, A)\
2617 ((char*)(A) >= S->base && (char*)(A) < S->base + S->size)
2618
2619 /* Return segment holding given address */
2620 static msegmentptr segment_holding(mstate m, char* addr) {
2621 msegmentptr sp = &m->seg;
2622 for (;;) {
2623 if (addr >= sp->base && addr < sp->base + sp->size)
2624 return sp;
2625 if ((sp = sp->next) == 0)
2626 return 0;
2627 }
2628 }
2629
2630 /* Return true if segment contains a segment link */
2631 static int has_segment_link(mstate m, msegmentptr ss) {
2632 msegmentptr sp = &m->seg;
2633 for (;;) {
2634 if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size)
2635 return 1;
2636 if ((sp = sp->next) == 0)
2637 return 0;
2638 }
2639 }
2640
2641 #ifndef MORECORE_CANNOT_TRIM
2642 #define should_trim(M,s) ((s) > (M)->trim_check)
2643 #else /* MORECORE_CANNOT_TRIM */
2644 #define should_trim(M,s) (0)
2645 #endif /* MORECORE_CANNOT_TRIM */
2646
2647 /*
2648 TOP_FOOT_SIZE is padding at the end of a segment, including space
2649 that may be needed to place segment records and fenceposts when new
2650 noncontiguous segments are added.
2651 */
2652 #define TOP_FOOT_SIZE\
2653 (align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE)
2654
2655
2656 /* ------------------------------- Hooks -------------------------------- */
2657
2658 /*
2659 PREACTION should be defined to return 0 on success, and nonzero on
2660 failure. If you are not using locking, you can redefine these to do
2661 anything you like.
2662 */
2663
2664 #if USE_LOCKS
2665
2666 #define PREACTION(M) ((use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0)
2667 #define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); }
2668 #else /* USE_LOCKS */
2669
2670 #ifndef PREACTION
2671 #define PREACTION(M) (0)
2672 #endif /* PREACTION */
2673
2674 #ifndef POSTACTION
2675 #define POSTACTION(M)
2676 #endif /* POSTACTION */
2677
2678 #endif /* USE_LOCKS */
2679
2680 /*
2681 CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses.
2682 USAGE_ERROR_ACTION is triggered on detected bad frees and
2683 reallocs. The argument p is an address that might have triggered the
2684 fault. It is ignored by the two predefined actions, but might be
2685 useful in custom actions that try to help diagnose errors.
2686 */
2687
2688 #if PROCEED_ON_ERROR
2689
2690 /* A count of the number of corruption errors causing resets */
2691 int malloc_corruption_error_count;
2692
2693 /* default corruption action */
2694 static void reset_on_error(mstate m);
2695
2696 #define CORRUPTION_ERROR_ACTION(m) reset_on_error(m)
2697 #define USAGE_ERROR_ACTION(m, p)
2698
2699 #else /* PROCEED_ON_ERROR */
2700
2701 #ifndef CORRUPTION_ERROR_ACTION
2702 #define CORRUPTION_ERROR_ACTION(m) ABORT
2703 #endif /* CORRUPTION_ERROR_ACTION */
2704
2705 #ifndef USAGE_ERROR_ACTION
2706 #define USAGE_ERROR_ACTION(m,p) ABORT
2707 #endif /* USAGE_ERROR_ACTION */
2708
2709 #endif /* PROCEED_ON_ERROR */
2710
2711 /* -------------------------- Debugging setup ---------------------------- */
2712
2713 #if ! DEBUG
2714
2715 #define check_free_chunk(M,P)
2716 #define check_inuse_chunk(M,P)
2717 #define check_malloced_chunk(M,P,N)
2718 #define check_mmapped_chunk(M,P)
2719 #define check_malloc_state(M)
2720 #define check_top_chunk(M,P)
2721
2722 #else /* DEBUG */
2723 #define check_free_chunk(M,P) do_check_free_chunk(M,P)
2724 #define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P)
2725 #define check_top_chunk(M,P) do_check_top_chunk(M,P)
2726 #define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N)
2727 #define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P)
2728 #define check_malloc_state(M) do_check_malloc_state(M)
2729
2730 static void do_check_any_chunk(mstate m, mchunkptr p);
2731 static void do_check_top_chunk(mstate m, mchunkptr p);
2732 static void do_check_mmapped_chunk(mstate m, mchunkptr p);
2733 static void do_check_inuse_chunk(mstate m, mchunkptr p);
2734 static void do_check_free_chunk(mstate m, mchunkptr p);
2735 static void do_check_malloced_chunk(mstate m, void* mem, size_t s);
2736 static void do_check_tree(mstate m, tchunkptr t);
2737 static void do_check_treebin(mstate m, bindex_t i);
2738 static void do_check_smallbin(mstate m, bindex_t i);
2739 static void do_check_malloc_state(mstate m);
2740 static int bin_find(mstate m, mchunkptr x);
2741 static size_t traverse_and_check(mstate m);
2742 #endif /* DEBUG */
2743
2744 /* ---------------------------- Indexing Bins ---------------------------- */
2745
2746 #define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS)
2747 #define small_index(s) ((s) >> SMALLBIN_SHIFT)
2748 #define small_index2size(i) ((i) << SMALLBIN_SHIFT)
2749 #define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE))
2750
2751 /* addressing by index. See above about smallbin repositioning */
2752 #define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1])))
2753 #define treebin_at(M,i) (&((M)->treebins[i]))
2754
2755 /* assign tree index for size S to variable I. Use x86 asm if possible */
2756 #if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
2757 #define compute_tree_index(S, I)\
2758 {\
2759 unsigned int X = S >> TREEBIN_SHIFT;\
2760 if (X == 0)\
2761 I = 0;\
2762 else if (X > 0xFFFF)\
2763 I = NTREEBINS-1;\
2764 else {\
2765 unsigned int K;\
2766 __asm__("bsrl\t%1, %0\n\t" : "=r" (K) : "rm" (X));\
2767 I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2768 }\
2769 }
2770
2771 #elif defined (__INTEL_COMPILER)
2772 #define compute_tree_index(S, I)\
2773 {\
2774 size_t X = S >> TREEBIN_SHIFT;\
2775 if (X == 0)\
2776 I = 0;\
2777 else if (X > 0xFFFF)\
2778 I = NTREEBINS-1;\
2779 else {\
2780 unsigned int K = _bit_scan_reverse (X); \
2781 I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2782 }\
2783 }
2784
2785 #elif defined(_MSC_VER) && _MSC_VER>=1300
2786 #define compute_tree_index(S, I)\
2787 {\
2788 size_t X = S >> TREEBIN_SHIFT;\
2789 if (X == 0)\
2790 I = 0;\
2791 else if (X > 0xFFFF)\
2792 I = NTREEBINS-1;\
2793 else {\
2794 unsigned int K;\
2795 _BitScanReverse((DWORD *) &K, X);\
2796 I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\
2797 }\
2798 }
2799
2800 #else /* GNUC */
2801 #define compute_tree_index(S, I)\
2802 {\
2803 size_t X = S >> TREEBIN_SHIFT;\
2804 if (X == 0)\
2805 I = 0;\
2806 else if (X > 0xFFFF)\
2807 I = NTREEBINS-1;\
2808 else {\
2809 unsigned int Y = (unsigned int)X;\
2810 unsigned int N = ((Y - 0x100) >> 16) & 8;\
2811 unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\
2812 N += K;\
2813 N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\
2814 K = 14 - N + ((Y <<= K) >> 15);\
2815 I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\
2816 }\
2817 }
2818 #endif /* GNUC */
2819
2820 /* Bit representing maximum resolved size in a treebin at i */
2821 #define bit_for_tree_index(i) \
2822 (i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2)
2823
2824 /* Shift placing maximum resolved bit in a treebin at i as sign bit */
2825 #define leftshift_for_tree_index(i) \
2826 ((i == NTREEBINS-1)? 0 : \
2827 ((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2)))
2828
2829 /* The size of the smallest chunk held in bin with index i */
2830 #define minsize_for_tree_index(i) \
2831 ((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \
2832 (((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1)))
2833
2834
2835 /* ------------------------ Operations on bin maps ----------------------- */
2836
2837 /* bit corresponding to given index */
2838 #define idx2bit(i) ((binmap_t)(1) << (i))
2839
2840 /* Mark/Clear bits with given index */
2841 #define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i))
2842 #define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i))
2843 #define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i))
2844
2845 #define mark_treemap(M,i) ((M)->treemap |= idx2bit(i))
2846 #define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i))
2847 #define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i))
2848
2849 /* isolate the least set bit of a bitmap */
2850 #define least_bit(x) ((x) & -(x))
2851
2852 /* mask with all bits to left of least bit of x on */
2853 #define left_bits(x) ((x<<1) | -(x<<1))
2854
2855 /* mask with all bits to left of or equal to least bit of x on */
2856 #define same_or_left_bits(x) ((x) | -(x))
2857
2858 /* index corresponding to given bit. Use x86 asm if possible */
2859
2860 #if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
2861 #define compute_bit2idx(X, I)\
2862 {\
2863 unsigned int J;\
2864 __asm__("bsfl\t%1, %0\n\t" : "=r" (J) : "rm" (X));\
2865 I = (bindex_t)J;\
2866 }
2867
2868 #elif defined (__INTEL_COMPILER)
2869 #define compute_bit2idx(X, I)\
2870 {\
2871 unsigned int J;\
2872 J = _bit_scan_forward (X); \
2873 I = (bindex_t)J;\
2874 }
2875
2876 #elif defined(_MSC_VER) && _MSC_VER>=1300
2877 #define compute_bit2idx(X, I)\
2878 {\
2879 unsigned int J;\
2880 _BitScanForward((DWORD *) &J, X);\
2881 I = (bindex_t)J;\
2882 }
2883
2884 #elif USE_BUILTIN_FFS
2885 #define compute_bit2idx(X, I) I = ffs(X)-1
2886
2887 #else
2888 #define compute_bit2idx(X, I)\
2889 {\
2890 unsigned int Y = X - 1;\
2891 unsigned int K = Y >> (16-4) & 16;\
2892 unsigned int N = K; Y >>= K;\
2893 N += K = Y >> (8-3) & 8; Y >>= K;\
2894 N += K = Y >> (4-2) & 4; Y >>= K;\
2895 N += K = Y >> (2-1) & 2; Y >>= K;\
2896 N += K = Y >> (1-0) & 1; Y >>= K;\
2897 I = (bindex_t)(N + Y);\
2898 }
2899 #endif /* GNUC */
2900
2901
2902 /* ----------------------- Runtime Check Support ------------------------- */
2903
2904 /*
2905 For security, the main invariant is that malloc/free/etc never
2906 writes to a static address other than malloc_state, unless static
2907 malloc_state itself has been corrupted, which cannot occur via
2908 malloc (because of these checks). In essence this means that we
2909 believe all pointers, sizes, maps etc held in malloc_state, but
2910 check all of those linked or offsetted from other embedded data
2911 structures. These checks are interspersed with main code in a way
2912 that tends to minimize their run-time cost.
2913
2914 When FOOTERS is defined, in addition to range checking, we also
2915 verify footer fields of inuse chunks, which can be used guarantee
2916 that the mstate controlling malloc/free is intact. This is a
2917 streamlined version of the approach described by William Robertson
2918 et al in "Run-time Detection of Heap-based Overflows" LISA'03
2919 http://www.usenix.org/events/lisa03/tech/robertson.html The footer
2920 of an inuse chunk holds the xor of its mstate and a random seed,
2921 that is checked upon calls to free() and realloc(). This is
2922 (probablistically) unguessable from outside the program, but can be
2923 computed by any code successfully malloc'ing any chunk, so does not
2924 itself provide protection against code that has already broken
2925 security through some other means. Unlike Robertson et al, we
2926 always dynamically check addresses of all offset chunks (previous,
2927 next, etc). This turns out to be cheaper than relying on hashes.
2928 */
2929
2930 #if !INSECURE
2931 /* Check if address a is at least as high as any from MORECORE or MMAP */
2932 #define ok_address(M, a) ((char*)(a) >= (M)->least_addr)
2933 /* Check if address of next chunk n is higher than base chunk p */
2934 #define ok_next(p, n) ((char*)(p) < (char*)(n))
2935 /* Check if p has its cinuse bit on */
2936 #define ok_cinuse(p) cinuse(p)
2937 /* Check if p has its pinuse bit on */
2938 #define ok_pinuse(p) pinuse(p)
2939
2940 #else /* !INSECURE */
2941 #define ok_address(M, a) (1)
2942 #define ok_next(b, n) (1)
2943 #define ok_cinuse(p) (1)
2944 #define ok_pinuse(p) (1)
2945 #endif /* !INSECURE */
2946
2947 #if (FOOTERS && !INSECURE)
2948 /* Check if (alleged) mstate m has expected magic field */
2949 #define ok_magic(M) ((M)->magic == mparams.magic)
2950 #else /* (FOOTERS && !INSECURE) */
2951 #define ok_magic(M) (1)
2952 #endif /* (FOOTERS && !INSECURE) */
2953
2954
2955 /* In gcc, use __builtin_expect to minimize impact of checks */
2956 #if !INSECURE
2957 #if defined(__GNUC__) && __GNUC__ >= 3
2958 #define RTCHECK(e) __builtin_expect(e, 1)
2959 #else /* GNUC */
2960 #define RTCHECK(e) (e)
2961 #endif /* GNUC */
2962 #else /* !INSECURE */
2963 #define RTCHECK(e) (1)
2964 #endif /* !INSECURE */
2965
2966 /* macros to set up inuse chunks with or without footers */
2967
2968 #if !FOOTERS
2969
2970 #define mark_inuse_foot(M,p,s)
2971
2972 /* Set cinuse bit and pinuse bit of next chunk */
2973 #define set_inuse(M,p,s)\
2974 ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
2975 ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
2976
2977 /* Set cinuse and pinuse of this chunk and pinuse of next chunk */
2978 #define set_inuse_and_pinuse(M,p,s)\
2979 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
2980 ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT)
2981
2982 /* Set size, cinuse and pinuse bit of this chunk */
2983 #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
2984 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT))
2985
2986 #else /* FOOTERS */
2987
2988 /* Set foot of inuse chunk to be xor of mstate and seed */
2989 #define mark_inuse_foot(M,p,s)\
2990 (((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic))
2991
2992 #define get_mstate_for(p)\
2993 ((mstate)(((mchunkptr)((char*)(p) +\
2994 (chunksize(p))))->prev_foot ^ mparams.magic))
2995
2996 #define set_inuse(M,p,s)\
2997 ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\
2998 (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \
2999 mark_inuse_foot(M,p,s))
3000
3001 #define set_inuse_and_pinuse(M,p,s)\
3002 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
3003 (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\
3004 mark_inuse_foot(M,p,s))
3005
3006 #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\
3007 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\
3008 mark_inuse_foot(M, p, s))
3009
3010 #endif /* !FOOTERS */
3011
3012 /* ---------------------------- setting mparams -------------------------- */
3013
3014 /* Initialize mparams */
3015 static int init_mparams(void) {
3016 #ifdef NEED_GLOBAL_LOCK_INIT
3017 if (malloc_global_mutex_status <= 0)
3018 init_malloc_global_mutex();
3019 #endif
3020
3021 ACQUIRE_MALLOC_GLOBAL_LOCK();
3022 if (mparams.magic == 0) {
3023 size_t magic;
3024 size_t psize;
3025 size_t gsize;
3026
3027 #ifndef WIN32
3028 psize = malloc_getpagesize;
3029 gsize = ((DEFAULT_GRANULARITY != 0)? DEFAULT_GRANULARITY : psize);
3030 #else /* WIN32 */
3031 {
3032 SYSTEM_INFO system_info;
3033 GetSystemInfo(&system_info);
3034 psize = system_info.dwPageSize;
3035 gsize = ((DEFAULT_GRANULARITY != 0)?
3036 DEFAULT_GRANULARITY : system_info.dwAllocationGranularity);
3037 }
3038 #endif /* WIN32 */
3039
3040 /* Sanity-check configuration:
3041 size_t must be unsigned and as wide as pointer type.
3042 ints must be at least 4 bytes.
3043 alignment must be at least 8.
3044 Alignment, min chunk size, and page size must all be powers of 2.
3045 */
3046 if ((sizeof(size_t) != sizeof(char*)) ||
3047 (MAX_SIZE_T < MIN_CHUNK_SIZE) ||
3048 (sizeof(int) < 4) ||
3049 (MALLOC_ALIGNMENT < (size_t)8U) ||
3050 ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT-SIZE_T_ONE)) != 0) ||
3051 ((MCHUNK_SIZE & (MCHUNK_SIZE-SIZE_T_ONE)) != 0) ||
3052 ((gsize & (gsize-SIZE_T_ONE)) != 0) ||
3053 ((psize & (psize-SIZE_T_ONE)) != 0))
3054 ABORT;
3055
3056 mparams.granularity = gsize;
3057 mparams.page_size = psize;
3058 mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD;
3059 mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD;
3060 #if MORECORE_CONTIGUOUS
3061 mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT;
3062 #else /* MORECORE_CONTIGUOUS */
3063 mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT|USE_NONCONTIGUOUS_BIT;
3064 #endif /* MORECORE_CONTIGUOUS */
3065
3066 #if !ONLY_MSPACES
3067 /* Set up lock for main malloc area */
3068 gm->mflags = mparams.default_mflags;
3069 (void)INITIAL_LOCK(&gm->mutex);
3070 #endif
3071
3072 #if (FOOTERS && !INSECURE)
3073 {
3074 #if USE_DEV_RANDOM
3075 int fd;
3076 unsigned char buf[sizeof(size_t)];
3077 /* Try to use /dev/urandom, else fall back on using time */
3078 if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 &&
3079 read(fd, buf, sizeof(buf)) == sizeof(buf)) {
3080 magic = *((size_t *) buf);
3081 close(fd);
3082 }
3083 else
3084 #endif /* USE_DEV_RANDOM */
3085 #ifdef WIN32
3086 magic = (size_t)(GetTickCount() ^ (size_t)0x55555555U);
3087 #else
3088 magic = (size_t)(time(0) ^ (size_t)0x55555555U);
3089 #endif
3090 magic |= (size_t)8U; /* ensure nonzero */
3091 magic &= ~(size_t)7U; /* improve chances of fault for bad values */
3092 }
3093 #else /* (FOOTERS && !INSECURE) */
3094 magic = (size_t)0x58585858U;
3095 #endif /* (FOOTERS && !INSECURE) */
3096
3097 mparams.magic = magic;
3098 }
3099
3100 RELEASE_MALLOC_GLOBAL_LOCK();
3101 return 1;
3102 }
3103
3104 /* support for mallopt */
3105 static int change_mparam(int param_number, int value) {
3106 size_t val = (value == -1)? MAX_SIZE_T : (size_t)value;
3107 ensure_initialization();
3108 switch(param_number) {
3109 case M_TRIM_THRESHOLD:
3110 mparams.trim_threshold = val;
3111 return 1;
3112 case M_GRANULARITY:
3113 if (val >= mparams.page_size && ((val & (val-1)) == 0)) {
3114 mparams.granularity = val;
3115 return 1;
3116 }
3117 else
3118 return 0;
3119 case M_MMAP_THRESHOLD:
3120 mparams.mmap_threshold = val;
3121 return 1;
3122 default:
3123 return 0;
3124 }
3125 }
3126
3127 #if DEBUG
3128 /* ------------------------- Debugging Support --------------------------- */
3129
3130 /* Check properties of any chunk, whether free, inuse, mmapped etc */
3131 static void do_check_any_chunk(mstate m, mchunkptr p) {
3132 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3133 assert(ok_address(m, p));
3134 }
3135
3136 /* Check properties of top chunk */
3137 static void do_check_top_chunk(mstate m, mchunkptr p) {
3138 msegmentptr sp = segment_holding(m, (char*)p);
3139 size_t sz = p->head & ~INUSE_BITS; /* third-lowest bit can be set! */
3140 assert(sp != 0);
3141 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3142 assert(ok_address(m, p));
3143 assert(sz == m->topsize);
3144 assert(sz > 0);
3145 assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE);
3146 assert(pinuse(p));
3147 assert(!pinuse(chunk_plus_offset(p, sz)));
3148 }
3149
3150 /* Check properties of (inuse) mmapped chunks */
3151 static void do_check_mmapped_chunk(mstate m, mchunkptr p) {
3152 size_t sz = chunksize(p);
3153 size_t len = (sz + (p->prev_foot & ~IS_MMAPPED_BIT) + MMAP_FOOT_PAD);
3154 assert(is_mmapped(p));
3155 assert(use_mmap(m));
3156 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD));
3157 assert(ok_address(m, p));
3158 assert(!is_small(sz));
3159 assert((len & (mparams.page_size-SIZE_T_ONE)) == 0);
3160 assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD);
3161 assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0);
3162 }
3163
3164 /* Check properties of inuse chunks */
3165 static void do_check_inuse_chunk(mstate m, mchunkptr p) {
3166 do_check_any_chunk(m, p);
3167 assert(cinuse(p));
3168 assert(next_pinuse(p));
3169 /* If not pinuse and not mmapped, previous chunk has OK offset */
3170 assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p);
3171 if (is_mmapped(p))
3172 do_check_mmapped_chunk(m, p);
3173 }
3174
3175 /* Check properties of free chunks */
3176 static void do_check_free_chunk(mstate m, mchunkptr p) {
3177 size_t sz = chunksize(p);
3178 mchunkptr next = chunk_plus_offset(p, sz);
3179 do_check_any_chunk(m, p);
3180 assert(!cinuse(p));
3181 assert(!next_pinuse(p));
3182 assert (!is_mmapped(p));
3183 if (p != m->dv && p != m->top) {
3184 if (sz >= MIN_CHUNK_SIZE) {
3185 assert((sz & CHUNK_ALIGN_MASK) == 0);
3186 assert(is_aligned(chunk2mem(p)));
3187 assert(next->prev_foot == sz);
3188 assert(pinuse(p));
3189 assert (next == m->top || cinuse(next));
3190 assert(p->fd->bk == p);
3191 assert(p->bk->fd == p);
3192 }
3193 else /* markers are always of size SIZE_T_SIZE */
3194 assert(sz == SIZE_T_SIZE);
3195 }
3196 }
3197
3198 /* Check properties of malloced chunks at the point they are malloced */
3199 static void do_check_malloced_chunk(mstate m, void* mem, size_t s) {
3200 if (mem != 0) {
3201 mchunkptr p = mem2chunk(mem);
3202 size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT);
3203 do_check_inuse_chunk(m, p);
3204 assert((sz & CHUNK_ALIGN_MASK) == 0);
3205 assert(sz >= MIN_CHUNK_SIZE);
3206 assert(sz >= s);
3207 /* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */
3208 assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE));
3209 }
3210 }
3211
3212 /* Check a tree and its subtrees. */
3213 static void do_check_tree(mstate m, tchunkptr t) {
3214 tchunkptr head = 0;
3215 tchunkptr u = t;
3216 bindex_t tindex = t->index;
3217 size_t tsize = chunksize(t);
3218 bindex_t idx;
3219 compute_tree_index(tsize, idx);
3220 assert(tindex == idx);
3221 assert(tsize >= MIN_LARGE_SIZE);
3222 assert(tsize >= minsize_for_tree_index(idx));
3223 assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1))));
3224
3225 do { /* traverse through chain of same-sized nodes */
3226 do_check_any_chunk(m, ((mchunkptr)u));
3227 assert(u->index == tindex);
3228 assert(chunksize(u) == tsize);
3229 assert(!cinuse(u));
3230 assert(!next_pinuse(u));
3231 assert(u->fd->bk == u);
3232 assert(u->bk->fd == u);
3233 if (u->parent == 0) {
3234 assert(u->child[0] == 0);
3235 assert(u->child[1] == 0);
3236 }
3237 else {
3238 assert(head == 0); /* only one node on chain has parent */
3239 head = u;
3240 assert(u->parent != u);
3241 assert (u->parent->child[0] == u ||
3242 u->parent->child[1] == u ||
3243 *((tbinptr*)(u->parent)) == u);
3244 if (u->child[0] != 0) {
3245 assert(u->child[0]->parent == u);
3246 assert(u->child[0] != u);
3247 do_check_tree(m, u->child[0]);
3248 }
3249 if (u->child[1] != 0) {
3250 assert(u->child[1]->parent == u);
3251 assert(u->child[1] != u);
3252 do_check_tree(m, u->child[1]);
3253 }
3254 if (u->child[0] != 0 && u->child[1] != 0) {
3255 assert(chunksize(u->child[0]) < chunksize(u->child[1]));
3256 }
3257 }
3258 u = u->fd;
3259 } while (u != t);
3260 assert(head != 0);
3261 }
3262
3263 /* Check all the chunks in a treebin. */
3264 static void do_check_treebin(mstate m, bindex_t i) {
3265 tbinptr* tb = treebin_at(m, i);
3266 tchunkptr t = *tb;
3267 int empty = (m->treemap & (1U << i)) == 0;
3268 if (t == 0)
3269 assert(empty);
3270 if (!empty)
3271 do_check_tree(m, t);
3272 }
3273
3274 /* Check all the chunks in a smallbin. */
3275 static void do_check_smallbin(mstate m, bindex_t i) {
3276 sbinptr b = smallbin_at(m, i);
3277 mchunkptr p = b->bk;
3278 unsigned int empty = (m->smallmap & (1U << i)) == 0;
3279 if (p == b)
3280 assert(empty);
3281 if (!empty) {
3282 for (; p != b; p = p->bk) {
3283 size_t size = chunksize(p);
3284 mchunkptr q;
3285 /* each chunk claims to be free */
3286 do_check_free_chunk(m, p);
3287 /* chunk belongs in bin */
3288 assert(small_index(size) == i);
3289 assert(p->bk == b || chunksize(p->bk) == chunksize(p));
3290 /* chunk is followed by an inuse chunk */
3291 q = next_chunk(p);
3292 if (q->head != FENCEPOST_HEAD)
3293 do_check_inuse_chunk(m, q);
3294 }
3295 }
3296 }
3297
3298 /* Find x in a bin. Used in other check functions. */
3299 static int bin_find(mstate m, mchunkptr x) {
3300 size_t size = chunksize(x);
3301 if (is_small(size)) {
3302 bindex_t sidx = small_index(size);
3303 sbinptr b = smallbin_at(m, sidx);
3304 if (smallmap_is_marked(m, sidx)) {
3305 mchunkptr p = b;
3306 do {
3307 if (p == x)
3308 return 1;
3309 } while ((p = p->fd) != b);
3310 }
3311 }
3312 else {
3313 bindex_t tidx;
3314 compute_tree_index(size, tidx);
3315 if (treemap_is_marked(m, tidx)) {
3316 tchunkptr t = *treebin_at(m, tidx);
3317 size_t sizebits = size << leftshift_for_tree_index(tidx);
3318 while (t != 0 && chunksize(t) != size) {
3319 t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1];
3320 sizebits <<= 1;
3321 }
3322 if (t != 0) {
3323 tchunkptr u = t;
3324 do {
3325 if (u == (tchunkptr)x)
3326 return 1;
3327 } while ((u = u->fd) != t);
3328 }
3329 }
3330 }
3331 return 0;
3332 }
3333
3334 /* Traverse each chunk and check it; return total */
3335 static size_t traverse_and_check(mstate m) {
3336 size_t sum = 0;
3337 if (is_initialized(m)) {
3338 msegmentptr s = &m->seg;
3339 sum += m->topsize + TOP_FOOT_SIZE;
3340 while (s != 0) {
3341 mchunkptr q = align_as_chunk(s->base);
3342 mchunkptr lastq = 0;
3343 assert(pinuse(q));
3344 while (segment_holds(s, q) &&
3345 q != m->top && q->head != FENCEPOST_HEAD) {
3346 sum += chunksize(q);
3347 if (cinuse(q)) {
3348 assert(!bin_find(m, q));
3349 do_check_inuse_chunk(m, q);
3350 }
3351 else {
3352 assert(q == m->dv || bin_find(m, q));
3353 assert(lastq == 0 || cinuse(lastq)); /* Not 2 consecutive free */
3354 do_check_free_chunk(m, q);
3355 }
3356 lastq = q;
3357 q = next_chunk(q);
3358 }
3359 s = s->next;
3360 }
3361 }
3362 return sum;
3363 }
3364
3365 /* Check all properties of malloc_state. */
3366 static void do_check_malloc_state(mstate m) {
3367 bindex_t i;
3368 size_t total;
3369 /* check bins */
3370 for (i = 0; i < NSMALLBINS; ++i)
3371 do_check_smallbin(m, i);
3372 for (i = 0; i < NTREEBINS; ++i)
3373 do_check_treebin(m, i);
3374
3375 if (m->dvsize != 0) { /* check dv chunk */
3376 do_check_any_chunk(m, m->dv);
3377 assert(m->dvsize == chunksize(m->dv));
3378 assert(m->dvsize >= MIN_CHUNK_SIZE);
3379 assert(bin_find(m, m->dv) == 0);
3380 }
3381
3382 if (m->top != 0) { /* check top chunk */
3383 do_check_top_chunk(m, m->top);
3384 /*assert(m->topsize == chunksize(m->top)); redundant */
3385 assert(m->topsize > 0);
3386 assert(bin_find(m, m->top) == 0);
3387 }
3388
3389 total = traverse_and_check(m);
3390 assert(total <= m->footprint);
3391 assert(m->footprint <= m->max_footprint);
3392 }
3393 #endif /* DEBUG */
3394
3395 /* ----------------------------- statistics ------------------------------ */
3396
3397 #if !NO_MALLINFO
3398 static struct mallinfo internal_mallinfo(mstate m) {
3399 struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
3400 ensure_initialization();
3401 if (!PREACTION(m)) {
3402 check_malloc_state(m);
3403 if (is_initialized(m)) {
3404 size_t nfree = SIZE_T_ONE; /* top always free */
3405 size_t mfree = m->topsize + TOP_FOOT_SIZE;
3406 size_t sum = mfree;
3407 msegmentptr s = &m->seg;
3408 while (s != 0) {
3409 mchunkptr q = align_as_chunk(s->base);
3410 while (segment_holds(s, q) &&
3411 q != m->top && q->head != FENCEPOST_HEAD) {
3412 size_t sz = chunksize(q);
3413 sum += sz;
3414 if (!cinuse(q)) {
3415 mfree += sz;
3416 ++nfree;
3417 }
3418 q = next_chunk(q);
3419 }
3420 s = s->next;
3421 }
3422
3423 nm.arena = sum;
3424 nm.ordblks = nfree;
3425 nm.hblkhd = m->footprint - sum;
3426 nm.usmblks = m->max_footprint;
3427 nm.uordblks = m->footprint - mfree;
3428 nm.fordblks = mfree;
3429 nm.keepcost = m->topsize;
3430 }
3431
3432 POSTACTION(m);
3433 }
3434 return nm;
3435 }
3436 #endif /* !NO_MALLINFO */
3437
3438 static void internal_malloc_stats(mstate m) {
3439 ensure_initialization();
3440 if (!PREACTION(m)) {
3441 size_t maxfp = 0;
3442 size_t fp = 0;
3443 size_t used = 0;
3444 check_malloc_state(m);
3445 if (is_initialized(m)) {
3446 msegmentptr s = &m->seg;
3447 maxfp = m->max_footprint;
3448 fp = m->footprint;
3449 used = fp - (m->topsize + TOP_FOOT_SIZE);
3450
3451 while (s != 0) {
3452 mchunkptr q = align_as_chunk(s->base);
3453 while (segment_holds(s, q) &&
3454 q != m->top && q->head != FENCEPOST_HEAD) {
3455 if (!cinuse(q))
3456 used -= chunksize(q);
3457 q = next_chunk(q);
3458 }
3459 s = s->next;
3460 }
3461 }
3462
3463 fprintf(stderr, "max system bytes = %10lu\n", (unsigned long)(maxfp));
3464 fprintf(stderr, "system bytes = %10lu\n", (unsigned long)(fp));
3465 fprintf(stderr, "in use bytes = %10lu\n", (unsigned long)(used));
3466
3467 POSTACTION(m);
3468 }
3469 }
3470
3471 /* ----------------------- Operations on smallbins ----------------------- */
3472
3473 /*
3474 Various forms of linking and unlinking are defined as macros. Even
3475 the ones for trees, which are very long but have very short typical
3476 paths. This is ugly but reduces reliance on inlining support of
3477 compilers.
3478 */
3479
3480 /* Link a free chunk into a smallbin */
3481 #define insert_small_chunk(M, P, S) {\
3482 bindex_t I = small_index(S);\
3483 mchunkptr B = smallbin_at(M, I);\
3484 mchunkptr F = B;\
3485 assert(S >= MIN_CHUNK_SIZE);\
3486 if (!smallmap_is_marked(M, I))\
3487 mark_smallmap(M, I);\
3488 else if (RTCHECK(ok_address(M, B->fd)))\
3489 F = B->fd;\
3490 else {\
3491 CORRUPTION_ERROR_ACTION(M);\
3492 }\
3493 B->fd = P;\
3494 F->bk = P;\
3495 P->fd = F;\
3496 P->bk = B;\
3497 }
3498
3499 /* Unlink a chunk from a smallbin */
3500 #define unlink_small_chunk(M, P, S) {\
3501 mchunkptr F = P->fd;\
3502 mchunkptr B = P->bk;\
3503 bindex_t I = small_index(S);\
3504 assert(P != B);\
3505 assert(P != F);\
3506 assert(chunksize(P) == small_index2size(I));\
3507 if (F == B)\
3508 clear_smallmap(M, I);\
3509 else if (RTCHECK((F == smallbin_at(M,I) || ok_address(M, F)) &&\
3510 (B == smallbin_at(M,I) || ok_address(M, B)))) {\
3511 F->bk = B;\
3512 B->fd = F;\
3513 }\
3514 else {\
3515 CORRUPTION_ERROR_ACTION(M);\
3516 }\
3517 }
3518
3519 /* Unlink the first chunk from a smallbin */
3520 #define unlink_first_small_chunk(M, B, P, I) {\
3521 mchunkptr F = P->fd;\
3522 assert(P != B);\
3523 assert(P != F);\
3524 assert(chunksize(P) == small_index2size(I));\
3525 if (B == F)\
3526 clear_smallmap(M, I);\
3527 else if (RTCHECK(ok_address(M, F))) {\
3528 B->fd = F;\
3529 F->bk = B;\
3530 }\
3531 else {\
3532 CORRUPTION_ERROR_ACTION(M);\
3533 }\
3534 }
3535
3536
3537
3538 /* Replace dv node, binning the old one */
3539 /* Used only when dvsize known to be small */
3540 #define replace_dv(M, P, S) {\
3541 size_t DVS = M->dvsize;\
3542 if (DVS != 0) {\
3543 mchunkptr DV = M->dv;\
3544 assert(is_small(DVS));\
3545 insert_small_chunk(M, DV, DVS);\
3546 }\
3547 M->dvsize = S;\
3548 M->dv = P;\
3549 }
3550
3551 /* ------------------------- Operations on trees ------------------------- */
3552
3553 /* Insert chunk into tree */
3554 #define insert_large_chunk(M, X, S) {\
3555 tbinptr* H;\
3556 bindex_t I;\
3557 compute_tree_index(S, I);\
3558 H = treebin_at(M, I);\
3559 X->index = I;\
3560 X->child[0] = X->child[1] = 0;\
3561 if (!treemap_is_marked(M, I)) {\
3562 mark_treemap(M, I);\
3563 *H = X;\
3564 X->parent = (tchunkptr)H;\
3565 X->fd = X->bk = X;\
3566 }\
3567 else {\
3568 tchunkptr T = *H;\
3569 size_t K = S << leftshift_for_tree_index(I);\
3570 for (;;) {\
3571 if (chunksize(T) != S) {\
3572 tchunkptr* C = &(T->child[(K >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]);\
3573 K <<= 1;\
3574 if (*C != 0)\
3575 T = *C;\
3576 else if (RTCHECK(ok_address(M, C))) {\
3577 *C = X;\
3578 X->parent = T;\
3579 X->fd = X->bk = X;\
3580 break;\
3581 }\
3582 else {\
3583 CORRUPTION_ERROR_ACTION(M);\
3584 break;\
3585 }\
3586 }\
3587 else {\
3588 tchunkptr F = T->fd;\
3589 if (RTCHECK(ok_address(M, T) && ok_address(M, F))) {\
3590 T->fd = F->bk = X;\
3591 X->fd = F;\
3592 X->bk = T;\
3593 X->parent = 0;\
3594 break;\
3595 }\
3596 else {\
3597 CORRUPTION_ERROR_ACTION(M);\
3598 break;\
3599 }\
3600 }\
3601 }\
3602 }\
3603 }
3604
3605 /*
3606 Unlink steps:
3607
3608 1. If x is a chained node, unlink it from its same-sized fd/bk links
3609 and choose its bk node as its replacement.
3610 2. If x was the last node of its size, but not a leaf node, it must
3611 be replaced with a leaf node (not merely one with an open left or
3612 right), to make sure that lefts and rights of descendants
3613 correspond properly to bit masks. We use the rightmost descendant
3614 of x. We could use any other leaf, but this is easy to locate and
3615 tends to counteract removal of leftmosts elsewhere, and so keeps
3616 paths shorter than minimally guaranteed. This doesn't loop much
3617 because on average a node in a tree is near the bottom.
3618 3. If x is the base of a chain (i.e., has parent links) relink
3619 x's parent and children to x's replacement (or null if none).
3620 */
3621
3622 #define unlink_large_chunk(M, X) {\
3623 tchunkptr XP = X->parent;\
3624 tchunkptr R;\
3625 if (X->bk != X) {\
3626 tchunkptr F = X->fd;\
3627 R = X->bk;\
3628 if (RTCHECK(ok_address(M, F))) {\
3629 F->bk = R;\
3630 R->fd = F;\
3631 }\
3632 else {\
3633 CORRUPTION_ERROR_ACTION(M);\
3634 }\
3635 }\
3636 else {\
3637 tchunkptr* RP;\
3638 if (((R = *(RP = &(X->child[1]))) != 0) ||\
3639 ((R = *(RP = &(X->child[0]))) != 0)) {\
3640 tchunkptr* CP;\
3641 while ((*(CP = &(R->child[1])) != 0) ||\
3642 (*(CP = &(R->child[0])) != 0)) {\
3643 R = *(RP = CP);\
3644 }\
3645 if (RTCHECK(ok_address(M, RP)))\
3646 *RP = 0;\
3647 else {\
3648 CORRUPTION_ERROR_ACTION(M);\
3649 }\
3650 }\
3651 }\
3652 if (XP != 0) {\
3653 tbinptr* H = treebin_at(M, X->index);\
3654 if (X == *H) {\
3655 if ((*H = R) == 0) \
3656 clear_treemap(M, X->index);\
3657 }\
3658 else if (RTCHECK(ok_address(M, XP))) {\
3659 if (XP->child[0] == X) \
3660 XP->child[0] = R;\
3661 else \
3662 XP->child[1] = R;\
3663 }\
3664 else\
3665 CORRUPTION_ERROR_ACTION(M);\
3666 if (R != 0) {\
3667 if (RTCHECK(ok_address(M, R))) {\
3668 tchunkptr C0, C1;\
3669 R->parent = XP;\
3670 if ((C0 = X->child[0]) != 0) {\
3671 if (RTCHECK(ok_address(M, C0))) {\
3672 R->child[0] = C0;\
3673 C0->parent = R;\
3674 }\
3675 else\
3676 CORRUPTION_ERROR_ACTION(M);\
3677 }\
3678 if ((C1 = X->child[1]) != 0) {\
3679 if (RTCHECK(ok_address(M, C1))) {\
3680 R->child[1] = C1;\
3681 C1->parent = R;\
3682 }\
3683 else\
3684 CORRUPTION_ERROR_ACTION(M);\
3685 }\
3686 }\
3687 else\
3688 CORRUPTION_ERROR_ACTION(M);\
3689 }\
3690 }\
3691 }
3692
3693 /* Relays to large vs small bin operations */
3694
3695 #define insert_chunk(M, P, S)\
3696 if (is_small(S)) insert_small_chunk(M, P, S)\
3697 else { tchunkptr TP = (tchunkptr)(P); insert_large_chunk(M, TP, S); }
3698
3699 #define unlink_chunk(M, P, S)\
3700 if (is_small(S)) unlink_small_chunk(M, P, S)\
3701 else { tchunkptr TP = (tchunkptr)(P); unlink_large_chunk(M, TP); }
3702
3703
3704 /* Relays to internal calls to malloc/free from realloc, memalign etc */
3705
3706 #if ONLY_MSPACES
3707 #define internal_malloc(m, b) mspace_malloc(m, b)
3708 #define internal_free(m, mem) mspace_free(m,mem);
3709 #else /* ONLY_MSPACES */
3710 #if MSPACES
3711 #define internal_malloc(m, b)\
3712 (m == gm)? dlmalloc(b) : mspace_malloc(m, b)
3713 #define internal_free(m, mem)\
3714 if (m == gm) dlfree(mem); else mspace_free(m,mem);
3715 #else /* MSPACES */
3716 #define internal_malloc(m, b) dlmalloc(b)
3717 #define internal_free(m, mem) dlfree(mem)
3718 #endif /* MSPACES */
3719 #endif /* ONLY_MSPACES */
3720
3721 /* ----------------------- Direct-mmapping chunks ----------------------- */
3722
3723 /*
3724 Directly mmapped chunks are set up with an offset to the start of
3725 the mmapped region stored in the prev_foot field of the chunk. This
3726 allows reconstruction of the required argument to MUNMAP when freed,
3727 and also allows adjustment of the returned chunk to meet alignment
3728 requirements (especially in memalign). There is also enough space
3729 allocated to hold a fake next chunk of size SIZE_T_SIZE to maintain
3730 the PINUSE bit so frees can be checked.
3731 */
3732
3733 /* Malloc using mmap */
3734 static void* mmap_alloc(mstate m, size_t nb) {
3735 size_t mmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3736 if (mmsize > nb) { /* Check for wrap around 0 */
3737 char* mm = (char*)(CALL_DIRECT_MMAP(mmsize));
3738 if (mm != CMFAIL) {
3739 size_t offset = align_offset(chunk2mem(mm));
3740 size_t psize = mmsize - offset - MMAP_FOOT_PAD;
3741 mchunkptr p = (mchunkptr)(mm + offset);
3742 p->prev_foot = offset | IS_MMAPPED_BIT;
3743 (p)->head = (psize|CINUSE_BIT);
3744 mark_inuse_foot(m, p, psize);
3745 chunk_plus_offset(p, psize)->head = FENCEPOST_HEAD;
3746 chunk_plus_offset(p, psize+SIZE_T_SIZE)->head = 0;
3747
3748 if (mm < m->least_addr)
3749 m->least_addr = mm;
3750 if ((m->footprint += mmsize) > m->max_footprint)
3751 m->max_footprint = m->footprint;
3752 assert(is_aligned(chunk2mem(p)));
3753 check_mmapped_chunk(m, p);
3754 return chunk2mem(p);
3755 }
3756 }
3757 return 0;
3758 }
3759
3760 /* Realloc using mmap */
3761 static mchunkptr mmap_resize(mstate m, mchunkptr oldp, size_t nb) {
3762 size_t oldsize = chunksize(oldp);
3763 if (is_small(nb)) /* Can't shrink mmap regions below small size */
3764 return 0;
3765 /* Keep old chunk if big enough but not too big */
3766 if (oldsize >= nb + SIZE_T_SIZE &&
3767 (oldsize - nb) <= (mparams.granularity << 1))
3768 return oldp;
3769 else {
3770 size_t offset = oldp->prev_foot & ~IS_MMAPPED_BIT;
3771 size_t oldmmsize = oldsize + offset + MMAP_FOOT_PAD;
3772 size_t newmmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3773 char* cp = (char*)CALL_MREMAP((char*)oldp - offset,
3774 oldmmsize, newmmsize, 1);
3775 if (cp != CMFAIL) {
3776 mchunkptr newp = (mchunkptr)(cp + offset);
3777 size_t psize = newmmsize - offset - MMAP_FOOT_PAD;
3778 newp->head = (psize|CINUSE_BIT);
3779 mark_inuse_foot(m, newp, psize);
3780 chunk_plus_offset(newp, psize)->head = FENCEPOST_HEAD;
3781 chunk_plus_offset(newp, psize+SIZE_T_SIZE)->head = 0;
3782
3783 if (cp < m->least_addr)
3784 m->least_addr = cp;
3785 if ((m->footprint += newmmsize - oldmmsize) > m->max_footprint)
3786 m->max_footprint = m->footprint;
3787 check_mmapped_chunk(m, newp);
3788 return newp;
3789 }
3790 }
3791 return 0;
3792 }
3793
3794 /* -------------------------- mspace management -------------------------- */
3795
3796 /* Initialize top chunk and its size */
3797 static void init_top(mstate m, mchunkptr p, size_t psize) {
3798 /* Ensure alignment */
3799 size_t offset = align_offset(chunk2mem(p));
3800 p = (mchunkptr)((char*)p + offset);
3801 psize -= offset;
3802
3803 m->top = p;
3804 m->topsize = psize;
3805 p->head = psize | PINUSE_BIT;
3806 /* set size of fake trailing chunk holding overhead space only once */
3807 chunk_plus_offset(p, psize)->head = TOP_FOOT_SIZE;
3808 m->trim_check = mparams.trim_threshold; /* reset on each update */
3809 }
3810
3811 /* Initialize bins for a new mstate that is otherwise zeroed out */
3812 static void init_bins(mstate m) {
3813 /* Establish circular links for smallbins */
3814 bindex_t i;
3815 for (i = 0; i < NSMALLBINS; ++i) {
3816 sbinptr bin = smallbin_at(m,i);
3817 bin->fd = bin->bk = bin;
3818 }
3819 }
3820
3821 #if PROCEED_ON_ERROR
3822
3823 /* default corruption action */
3824 static void reset_on_error(mstate m) {
3825 int i;
3826 ++malloc_corruption_error_count;
3827 /* Reinitialize fields to forget about all memory */
3828 m->smallbins = m->treebins = 0;
3829 m->dvsize = m->topsize = 0;
3830 m->seg.base = 0;
3831 m->seg.size = 0;
3832 m->seg.next = 0;
3833 m->top = m->dv = 0;
3834 for (i = 0; i < NTREEBINS; ++i)
3835 *treebin_at(m, i) = 0;
3836 init_bins(m);
3837 }
3838 #endif /* PROCEED_ON_ERROR */
3839
3840 /* Allocate chunk and prepend remainder with chunk in successor base. */
3841 static void* prepend_alloc(mstate m, char* newbase, char* oldbase,
3842 size_t nb) {
3843 mchunkptr p = align_as_chunk(newbase);
3844 mchunkptr oldfirst = align_as_chunk(oldbase);
3845 size_t psize = (char*)oldfirst - (char*)p;
3846 mchunkptr q = chunk_plus_offset(p, nb);
3847 size_t qsize = psize - nb;
3848 set_size_and_pinuse_of_inuse_chunk(m, p, nb);
3849
3850 assert((char*)oldfirst > (char*)q);
3851 assert(pinuse(oldfirst));
3852 assert(qsize >= MIN_CHUNK_SIZE);
3853
3854 /* consolidate remainder with first chunk of old base */
3855 if (oldfirst == m->top) {
3856 size_t tsize = m->topsize += qsize;
3857 m->top = q;
3858 q->head = tsize | PINUSE_BIT;
3859 check_top_chunk(m, q);
3860 }
3861 else if (oldfirst == m->dv) {
3862 size_t dsize = m->dvsize += qsize;
3863 m->dv = q;
3864 set_size_and_pinuse_of_free_chunk(q, dsize);
3865 }
3866 else {
3867 if (!cinuse(oldfirst)) {
3868 size_t nsize = chunksize(oldfirst);
3869 unlink_chunk(m, oldfirst, nsize);
3870 oldfirst = chunk_plus_offset(oldfirst, nsize);
3871 qsize += nsize;
3872 }
3873 set_free_with_pinuse(q, qsize, oldfirst);
3874 insert_chunk(m, q, qsize);
3875 check_free_chunk(m, q);
3876 }
3877
3878 check_malloced_chunk(m, chunk2mem(p), nb);
3879 return chunk2mem(p);
3880 }
3881
3882 /* Add a segment to hold a new noncontiguous region */
3883 static void add_segment(mstate m, char* tbase, size_t tsize, flag_t mmapped) {
3884 /* Determine locations and sizes of segment, fenceposts, old top */
3885 char* old_top = (char*)m->top;
3886 msegmentptr oldsp = segment_holding(m, old_top);
3887 char* old_end = oldsp->base + oldsp->size;
3888 size_t ssize = pad_request(sizeof(struct malloc_segment));
3889 char* rawsp = old_end - (ssize + FOUR_SIZE_T_SIZES + CHUNK_ALIGN_MASK);
3890 size_t offset = align_offset(chunk2mem(rawsp));
3891 char* asp = rawsp + offset;
3892 char* csp = (asp < (old_top + MIN_CHUNK_SIZE))? old_top : asp;
3893 mchunkptr sp = (mchunkptr)csp;
3894 msegmentptr ss = (msegmentptr)(chunk2mem(sp));
3895 mchunkptr tnext = chunk_plus_offset(sp, ssize);
3896 mchunkptr p = tnext;
3897 int nfences = 0;
3898
3899 /* reset top to new space */
3900 init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);
3901
3902 /* Set up segment record */
3903 assert(is_aligned(ss));
3904 set_size_and_pinuse_of_inuse_chunk(m, sp, ssize);
3905 *ss = m->seg; /* Push current record */
3906 m->seg.base = tbase;
3907 m->seg.size = tsize;
3908 m->seg.sflags = mmapped;
3909 m->seg.next = ss;
3910
3911 /* Insert trailing fenceposts */
3912 for (;;) {
3913 mchunkptr nextp = chunk_plus_offset(p, SIZE_T_SIZE);
3914 p->head = FENCEPOST_HEAD;
3915 ++nfences;
3916 if ((char*)(&(nextp->head)) < old_end)
3917 p = nextp;
3918 else
3919 break;
3920 }
3921 assert(nfences >= 2);
3922
3923 /* Insert the rest of old top into a bin as an ordinary free chunk */
3924 if (csp != old_top) {
3925 mchunkptr q = (mchunkptr)old_top;
3926 size_t psize = csp - old_top;
3927 mchunkptr tn = chunk_plus_offset(q, psize);
3928 set_free_with_pinuse(q, psize, tn);
3929 insert_chunk(m, q, psize);
3930 }
3931
3932 check_top_chunk(m, m->top);
3933 }
3934
3935 /* -------------------------- System allocation -------------------------- */
3936
3937 /* Get memory from system using MORECORE or MMAP */
3938 static void* sys_alloc(mstate m, size_t nb) {
3939 char* tbase = CMFAIL;
3940 size_t tsize = 0;
3941 flag_t mmap_flag = 0;
3942
3943 ensure_initialization();
3944
3945 /* Directly map large chunks */
3946 if (use_mmap(m) && nb >= mparams.mmap_threshold) {
3947 void* mem = mmap_alloc(m, nb);
3948 if (mem != 0)
3949 return mem;
3950 }
3951
3952 /*
3953 Try getting memory in any of three ways (in most-preferred to
3954 least-preferred order):
3955 1. A call to MORECORE that can normally contiguously extend memory.
3956 (disabled if not MORECORE_CONTIGUOUS or not HAVE_MORECORE or
3957 main space is mmapped or a previous contiguous call failed)
3958 2. A call to MMAP new space (disabled if not HAVE_MMAP).
3959 Note that under the default settings, if MORECORE is unable to
3960 fulfill a request, and HAVE_MMAP is true, then mmap is
3961 used as a noncontiguous system allocator. This is a useful backup
3962 strategy for systems with holes in address spaces -- in this case
3963 sbrk cannot contiguously expand the heap, but mmap may be able to
3964 find space.
3965 3. A call to MORECORE that cannot usually contiguously extend memory.
3966 (disabled if not HAVE_MORECORE)
3967
3968 In all cases, we need to request enough bytes from system to ensure
3969 we can malloc nb bytes upon success, so pad with enough space for
3970 top_foot, plus alignment-pad to make sure we don't lose bytes if
3971 not on boundary, and round this up to a granularity unit.
3972 */
3973
3974 if (MORECORE_CONTIGUOUS && !use_noncontiguous(m)) {
3975 char* br = CMFAIL;
3976 msegmentptr ss = (m->top == 0)? 0 : segment_holding(m, (char*)m->top);
3977 size_t asize = 0;
3978 ACQUIRE_MALLOC_GLOBAL_LOCK();
3979
3980 if (ss == 0) { /* First time through or recovery */
3981 char* base = (char*)CALL_MORECORE(0);
3982 if (base != CMFAIL) {
3983 asize = granularity_align(nb + SYS_ALLOC_PADDING);
3984 /* Adjust to end on a page boundary */
3985 if (!is_page_aligned(base))
3986 asize += (page_align((size_t)base) - (size_t)base);
3987 /* Can't call MORECORE if size is negative when treated as signed */
3988 if (asize < HALF_MAX_SIZE_T &&
3989 (br = (char*)(CALL_MORECORE(asize))) == base) {
3990 tbase = base;
3991 tsize = asize;
3992 }
3993 }
3994 }
3995 else {
3996 /* Subtract out existing available top space from MORECORE request. */
3997 asize = granularity_align(nb - m->topsize + SYS_ALLOC_PADDING);
3998 /* Use mem here only if it did continuously extend old space */
3999 if (asize < HALF_MAX_SIZE_T &&
4000 (br = (char*)(CALL_MORECORE(asize))) == ss->base+ss->size) {
4001 tbase = br;
4002 tsize = asize;
4003 }
4004 }
4005
4006 if (tbase == CMFAIL) { /* Cope with partial failure */
4007 if (br != CMFAIL) { /* Try to use/extend the space we did get */
4008 if (asize < HALF_MAX_SIZE_T &&
4009 asize < nb + SYS_ALLOC_PADDING) {
4010 size_t esize = granularity_align(nb + SYS_ALLOC_PADDING - asize);
4011 if (esize < HALF_MAX_SIZE_T) {
4012 char* end = (char*)CALL_MORECORE(esize);
4013 if (end != CMFAIL)
4014 asize += esize;
4015 else { /* Can't use; try to release */
4016 (void) CALL_MORECORE(-asize);
4017 br = CMFAIL;
4018 }
4019 }
4020 }
4021 }
4022 if (br != CMFAIL) { /* Use the space we did get */
4023 tbase = br;
4024 tsize = asize;
4025 }
4026 else
4027 disable_contiguous(m); /* Don't try contiguous path in the future */
4028 }
4029
4030 RELEASE_MALLOC_GLOBAL_LOCK();
4031 }
4032
4033 if (HAVE_MMAP && tbase == CMFAIL) { /* Try MMAP */
4034 size_t rsize = granularity_align(nb + SYS_ALLOC_PADDING);
4035 if (rsize > nb) { /* Fail if wraps around zero */
4036 char* mp = (char*)(CALL_MMAP(rsize));
4037 if (mp != CMFAIL) {
4038 tbase = mp;
4039 tsize = rsize;
4040 mmap_flag = IS_MMAPPED_BIT;
4041 }
4042 }
4043 }
4044
4045 if (HAVE_MORECORE && tbase == CMFAIL) { /* Try noncontiguous MORECORE */
4046 size_t asize = granularity_align(nb + SYS_ALLOC_PADDING);
4047 if (asize < HALF_MAX_SIZE_T) {
4048 char* br = CMFAIL;
4049 char* end = CMFAIL;
4050 ACQUIRE_MALLOC_GLOBAL_LOCK();
4051 br = (char*)(CALL_MORECORE(asize));
4052 end = (char*)(CALL_MORECORE(0));
4053 RELEASE_MALLOC_GLOBAL_LOCK();
4054 if (br != CMFAIL && end != CMFAIL && br < end) {
4055 size_t ssize = end - br;
4056 if (ssize > nb + TOP_FOOT_SIZE) {
4057 tbase = br;
4058 tsize = ssize;
4059 }
4060 }
4061 }
4062 }
4063
4064 if (tbase != CMFAIL) {
4065
4066 if ((m->footprint += tsize) > m->max_footprint)
4067 m->max_footprint = m->footprint;
4068
4069 if (!is_initialized(m)) { /* first-time initialization */
4070 m->seg.base = m->least_addr = tbase;
4071 m->seg.size = tsize;
4072 m->seg.sflags = mmap_flag;
4073 m->magic = mparams.magic;
4074 m->release_checks = MAX_RELEASE_CHECK_RATE;
4075 init_bins(m);
4076 #if !ONLY_MSPACES
4077 if (is_global(m))
4078 init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE);
4079 else
4080 #endif
4081 {
4082 /* Offset top by embedded malloc_state */
4083 mchunkptr mn = next_chunk(mem2chunk(m));
4084 init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) -TOP_FOOT_SIZE);
4085 }
4086 }
4087
4088 else {
4089 /* Try to merge with an existing segment */
4090 msegmentptr sp = &m->seg;
4091 /* Only consider most recent segment if traversal suppressed */
4092 while (sp != 0 && tbase != sp->base + sp->size)
4093 sp = (NO_SEGMENT_TRAVERSAL) ? 0 : sp->next;