|
1 /* |
|
2 $Id: malloc.c,v 1.4 2006/03/30 16:47:29 wg Exp $ |
|
3 |
|
4 This version of malloc.c was adapted for ptmalloc3 by Wolfram Gloger |
|
5 <wg@malloc.de>. Therefore, some of the comments below do not apply |
|
6 for this modified version. However, it is the intention to keep |
|
7 differences to Doug Lea's original version minimal, hence the |
|
8 comments were mostly left unchanged. |
|
9 |
|
10 ----------------------------------------------------------------------- |
|
11 |
|
12 This is a version (aka dlmalloc) of malloc/free/realloc written by |
|
13 Doug Lea and released to the public domain, as explained at |
|
14 http://creativecommons.org/licenses/publicdomain. Send questions, |
|
15 comments, complaints, performance data, etc to dl@cs.oswego.edu |
|
16 |
|
17 * Version pre-2.8.4 Wed Mar 29 19:46:29 2006 (dl at gee) |
|
18 |
|
19 Note: There may be an updated version of this malloc obtainable at |
|
20 ftp://gee.cs.oswego.edu/pub/misc/malloc.c |
|
21 Check before installing! |
|
22 |
|
23 * Quickstart |
|
24 |
|
25 This library is all in one file to simplify the most common usage: |
|
26 ftp it, compile it (-O3), and link it into another program. All of |
|
27 the compile-time options default to reasonable values for use on |
|
28 most platforms. You might later want to step through various |
|
29 compile-time and dynamic tuning options. |
|
30 |
|
31 For convenience, an include file for code using this malloc is at: |
|
32 ftp://gee.cs.oswego.edu/pub/misc/malloc-2.8.3.h |
|
33 You don't really need this .h file unless you call functions not |
|
34 defined in your system include files. The .h file contains only the |
|
35 excerpts from this file needed for using this malloc on ANSI C/C++ |
|
36 systems, so long as you haven't changed compile-time options about |
|
37 naming and tuning parameters. If you do, then you can create your |
|
38 own malloc.h that does include all settings by cutting at the point |
|
39 indicated below. Note that you may already by default be using a C |
|
40 library containing a malloc that is based on some version of this |
|
41 malloc (for example in linux). You might still want to use the one |
|
42 in this file to customize settings or to avoid overheads associated |
|
43 with library versions. |
|
44 |
|
45 * Vital statistics: |
|
46 |
|
47 Supported pointer/size_t representation: 4 or 8 bytes |
|
48 size_t MUST be an unsigned type of the same width as |
|
49 pointers. (If you are using an ancient system that declares |
|
50 size_t as a signed type, or need it to be a different width |
|
51 than pointers, you can use a previous release of this malloc |
|
52 (e.g. 2.7.2) supporting these.) |
|
53 |
|
54 Alignment: 8 bytes (default) |
|
55 This suffices for nearly all current machines and C compilers. |
|
56 However, you can define MALLOC_ALIGNMENT to be wider than this |
|
57 if necessary (up to 128bytes), at the expense of using more space. |
|
58 |
|
59 Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes) |
|
60 8 or 16 bytes (if 8byte sizes) |
|
61 Each malloced chunk has a hidden word of overhead holding size |
|
62 and status information, and additional cross-check word |
|
63 if FOOTERS is defined. |
|
64 |
|
65 Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead) |
|
66 8-byte ptrs: 32 bytes (including overhead) |
|
67 |
|
68 Even a request for zero bytes (i.e., malloc(0)) returns a |
|
69 pointer to something of the minimum allocatable size. |
|
70 The maximum overhead wastage (i.e., number of extra bytes |
|
71 allocated than were requested in malloc) is less than or equal |
|
72 to the minimum size, except for requests >= mmap_threshold that |
|
73 are serviced via mmap(), where the worst case wastage is about |
|
74 32 bytes plus the remainder from a system page (the minimal |
|
75 mmap unit); typically 4096 or 8192 bytes. |
|
76 |
|
77 Security: static-safe; optionally more or less |
|
78 The "security" of malloc refers to the ability of malicious |
|
79 code to accentuate the effects of errors (for example, freeing |
|
80 space that is not currently malloc'ed or overwriting past the |
|
81 ends of chunks) in code that calls malloc. This malloc |
|
82 guarantees not to modify any memory locations below the base of |
|
83 heap, i.e., static variables, even in the presence of usage |
|
84 errors. The routines additionally detect most improper frees |
|
85 and reallocs. All this holds as long as the static bookkeeping |
|
86 for malloc itself is not corrupted by some other means. This |
|
87 is only one aspect of security -- these checks do not, and |
|
88 cannot, detect all possible programming errors. |
|
89 |
|
90 If FOOTERS is defined nonzero, then each allocated chunk |
|
91 carries an additional check word to verify that it was malloced |
|
92 from its space. These check words are the same within each |
|
93 execution of a program using malloc, but differ across |
|
94 executions, so externally crafted fake chunks cannot be |
|
95 freed. This improves security by rejecting frees/reallocs that |
|
96 could corrupt heap memory, in addition to the checks preventing |
|
97 writes to statics that are always on. This may further improve |
|
98 security at the expense of time and space overhead. (Note that |
|
99 FOOTERS may also be worth using with MSPACES.) |
|
100 |
|
101 By default detected errors cause the program to abort (calling |
|
102 "abort()"). You can override this to instead proceed past |
|
103 errors by defining PROCEED_ON_ERROR. In this case, a bad free |
|
104 has no effect, and a malloc that encounters a bad address |
|
105 caused by user overwrites will ignore the bad address by |
|
106 dropping pointers and indices to all known memory. This may |
|
107 be appropriate for programs that should continue if at all |
|
108 possible in the face of programming errors, although they may |
|
109 run out of memory because dropped memory is never reclaimed. |
|
110 |
|
111 If you don't like either of these options, you can define |
|
112 CORRUPTION_ERROR_ACTION and USAGE_ERROR_ACTION to do anything |
|
113 else. And if if you are sure that your program using malloc has |
|
114 no errors or vulnerabilities, you can define INSECURE to 1, |
|
115 which might (or might not) provide a small performance improvement. |
|
116 |
|
117 Thread-safety: NOT thread-safe unless USE_LOCKS defined |
|
118 When USE_LOCKS is defined, each public call to malloc, free, |
|
119 etc is surrounded with either a pthread mutex or a win32 |
|
120 spinlock (depending on WIN32). This is not especially fast, and |
|
121 can be a major bottleneck. It is designed only to provide |
|
122 minimal protection in concurrent environments, and to provide a |
|
123 basis for extensions. If you are using malloc in a concurrent |
|
124 program, consider instead using nedmalloc |
|
125 (http://www.nedprod.com/programs/portable/nedmalloc/) or |
|
126 ptmalloc (See http://www.malloc.de), which are derived |
|
127 from versions of this malloc. |
|
128 |
|
129 System requirements: Any combination of MORECORE and/or MMAP/MUNMAP |
|
130 This malloc can use unix sbrk or any emulation (invoked using |
|
131 the CALL_MORECORE macro) and/or mmap/munmap or any emulation |
|
132 (invoked using CALL_MMAP/CALL_MUNMAP) to get and release system |
|
133 memory. On most unix systems, it tends to work best if both |
|
134 MORECORE and MMAP are enabled. On Win32, it uses emulations |
|
135 based on VirtualAlloc. It also uses common C library functions |
|
136 like memset. |
|
137 |
|
138 Compliance: I believe it is compliant with the Single Unix Specification |
|
139 (See http://www.unix.org). Also SVID/XPG, ANSI C, and probably |
|
140 others as well. |
|
141 |
|
142 * Overview of algorithms |
|
143 |
|
144 This is not the fastest, most space-conserving, most portable, or |
|
145 most tunable malloc ever written. However it is among the fastest |
|
146 while also being among the most space-conserving, portable and |
|
147 tunable. Consistent balance across these factors results in a good |
|
148 general-purpose allocator for malloc-intensive programs. |
|
149 |
|
150 In most ways, this malloc is a best-fit allocator. Generally, it |
|
151 chooses the best-fitting existing chunk for a request, with ties |
|
152 broken in approximately least-recently-used order. (This strategy |
|
153 normally maintains low fragmentation.) However, for requests less |
|
154 than 256bytes, it deviates from best-fit when there is not an |
|
155 exactly fitting available chunk by preferring to use space adjacent |
|
156 to that used for the previous small request, as well as by breaking |
|
157 ties in approximately most-recently-used order. (These enhance |
|
158 locality of series of small allocations.) And for very large requests |
|
159 (>= 256Kb by default), it relies on system memory mapping |
|
160 facilities, if supported. (This helps avoid carrying around and |
|
161 possibly fragmenting memory used only for large chunks.) |
|
162 |
|
163 All operations (except malloc_stats and mallinfo) have execution |
|
164 times that are bounded by a constant factor of the number of bits in |
|
165 a size_t, not counting any clearing in calloc or copying in realloc, |
|
166 or actions surrounding MORECORE and MMAP that have times |
|
167 proportional to the number of non-contiguous regions returned by |
|
168 system allocation routines, which is often just 1. In real-time |
|
169 applications, you can optionally suppress segment traversals using |
|
170 NO_SEGMENT_TRAVERSAL, which assures bounded execution even when |
|
171 system allocators return non-contiguous spaces, at the typical |
|
172 expense of carrying around more memory and increased fragmentation. |
|
173 |
|
174 The implementation is not very modular and seriously overuses |
|
175 macros. Perhaps someday all C compilers will do as good a job |
|
176 inlining modular code as can now be done by brute-force expansion, |
|
177 but now, enough of them seem not to. |
|
178 |
|
179 Some compilers issue a lot of warnings about code that is |
|
180 dead/unreachable only on some platforms, and also about intentional |
|
181 uses of negation on unsigned types. All known cases of each can be |
|
182 ignored. |
|
183 |
|
184 For a longer but out of date high-level description, see |
|
185 http://gee.cs.oswego.edu/dl/html/malloc.html |
|
186 |
|
187 * MSPACES |
|
188 If MSPACES is defined, then in addition to malloc, free, etc., |
|
189 this file also defines mspace_malloc, mspace_free, etc. These |
|
190 are versions of malloc routines that take an "mspace" argument |
|
191 obtained using create_mspace, to control all internal bookkeeping. |
|
192 If ONLY_MSPACES is defined, only these versions are compiled. |
|
193 So if you would like to use this allocator for only some allocations, |
|
194 and your system malloc for others, you can compile with |
|
195 ONLY_MSPACES and then do something like... |
|
196 static mspace mymspace = create_mspace(0,0); // for example |
|
197 #define mymalloc(bytes) mspace_malloc(mymspace, bytes) |
|
198 |
|
199 (Note: If you only need one instance of an mspace, you can instead |
|
200 use "USE_DL_PREFIX" to relabel the global malloc.) |
|
201 |
|
202 You can similarly create thread-local allocators by storing |
|
203 mspaces as thread-locals. For example: |
|
204 static __thread mspace tlms = 0; |
|
205 void* tlmalloc(size_t bytes) { |
|
206 if (tlms == 0) tlms = create_mspace(0, 0); |
|
207 return mspace_malloc(tlms, bytes); |
|
208 } |
|
209 void tlfree(void* mem) { mspace_free(tlms, mem); } |
|
210 |
|
211 Unless FOOTERS is defined, each mspace is completely independent. |
|
212 You cannot allocate from one and free to another (although |
|
213 conformance is only weakly checked, so usage errors are not always |
|
214 caught). If FOOTERS is defined, then each chunk carries around a tag |
|
215 indicating its originating mspace, and frees are directed to their |
|
216 originating spaces. |
|
217 |
|
218 ------------------------- Compile-time options --------------------------- |
|
219 |
|
220 Be careful in setting #define values for numerical constants of type |
|
221 size_t. On some systems, literal values are not automatically extended |
|
222 to size_t precision unless they are explicitly casted. You can also |
|
223 use the symbolic values MAX_SIZE_T, SIZE_T_ONE, etc below. |
|
224 |
|
225 WIN32 default: defined if _WIN32 defined |
|
226 Defining WIN32 sets up defaults for MS environment and compilers. |
|
227 Otherwise defaults are for unix. |
|
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. See |
|
325 near the end of this file for guidelines for creating a custom |
|
326 version of MORECORE. |
|
327 |
|
328 MORECORE_CONTIGUOUS default: 1 (true) if HAVE_MORECORE |
|
329 If true, take advantage of fact that consecutive calls to MORECORE |
|
330 with positive arguments always return contiguous increasing |
|
331 addresses. This is true of unix sbrk. It does not hurt too much to |
|
332 set it true anyway, since malloc copes with non-contiguities. |
|
333 Setting it false when definitely non-contiguous saves time |
|
334 and possibly wasted space it would take to discover this though. |
|
335 |
|
336 MORECORE_CANNOT_TRIM default: NOT defined |
|
337 True if MORECORE cannot release space back to the system when given |
|
338 negative arguments. This is generally necessary only if you are |
|
339 using a hand-crafted MORECORE function that cannot handle negative |
|
340 arguments. |
|
341 |
|
342 NO_SEGMENT_TRAVERSAL default: 0 |
|
343 If non-zero, suppresses traversals of memory segments |
|
344 returned by either MORECORE or CALL_MMAP. This disables |
|
345 merging of segments that are contiguous, and selectively |
|
346 releasing them to the OS if unused, but bounds execution times. |
|
347 |
|
348 HAVE_MMAP default: 1 (true) |
|
349 True if this system supports mmap or an emulation of it. If so, and |
|
350 HAVE_MORECORE is not true, MMAP is used for all system |
|
351 allocation. If set and HAVE_MORECORE is true as well, MMAP is |
|
352 primarily used to directly allocate very large blocks. It is also |
|
353 used as a backup strategy in cases where MORECORE fails to provide |
|
354 space from system. Note: A single call to MUNMAP is assumed to be |
|
355 able to unmap memory that may have be allocated using multiple calls |
|
356 to MMAP, so long as they are adjacent. |
|
357 |
|
358 HAVE_MREMAP default: 1 on linux, else 0 |
|
359 If true realloc() uses mremap() to re-allocate large blocks and |
|
360 extend or shrink allocation spaces. |
|
361 |
|
362 MMAP_CLEARS default: 1 except on WINCE. |
|
363 True if mmap clears memory so calloc doesn't need to. This is true |
|
364 for standard unix mmap using /dev/zero and on WIN32 except for WINCE. |
|
365 |
|
366 USE_BUILTIN_FFS default: 0 (i.e., not used) |
|
367 Causes malloc to use the builtin ffs() function to compute indices. |
|
368 Some compilers may recognize and intrinsify ffs to be faster than the |
|
369 supplied C version. Also, the case of x86 using gcc is special-cased |
|
370 to an asm instruction, so is already as fast as it can be, and so |
|
371 this setting has no effect. Similarly for Win32 under recent MS compilers. |
|
372 (On most x86s, the asm version is only slightly faster than the C version.) |
|
373 |
|
374 malloc_getpagesize default: derive from system includes, or 4096. |
|
375 The system page size. To the extent possible, this malloc manages |
|
376 memory from the system in page-size units. This may be (and |
|
377 usually is) a function rather than a constant. This is ignored |
|
378 if WIN32, where page size is determined using getSystemInfo during |
|
379 initialization. |
|
380 |
|
381 USE_DEV_RANDOM default: 0 (i.e., not used) |
|
382 Causes malloc to use /dev/random to initialize secure magic seed for |
|
383 stamping footers. Otherwise, the current time is used. |
|
384 |
|
385 NO_MALLINFO default: 0 |
|
386 If defined, don't compile "mallinfo". This can be a simple way |
|
387 of dealing with mismatches between system declarations and |
|
388 those in this file. |
|
389 |
|
390 MALLINFO_FIELD_TYPE default: size_t |
|
391 The type of the fields in the mallinfo struct. This was originally |
|
392 defined as "int" in SVID etc, but is more usefully defined as |
|
393 size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set |
|
394 |
|
395 REALLOC_ZERO_BYTES_FREES default: not defined |
|
396 This should be set if a call to realloc with zero bytes should |
|
397 be the same as a call to free. Some people think it should. Otherwise, |
|
398 since this malloc returns a unique pointer for malloc(0), so does |
|
399 realloc(p, 0). |
|
400 |
|
401 LACKS_UNISTD_H, LACKS_FCNTL_H, LACKS_SYS_PARAM_H, LACKS_SYS_MMAN_H |
|
402 LACKS_STRINGS_H, LACKS_STRING_H, LACKS_SYS_TYPES_H, LACKS_ERRNO_H |
|
403 LACKS_STDLIB_H default: NOT defined unless on WIN32 |
|
404 Define these if your system does not have these header files. |
|
405 You might need to manually insert some of the declarations they provide. |
|
406 |
|
407 DEFAULT_GRANULARITY default: page size if MORECORE_CONTIGUOUS, |
|
408 system_info.dwAllocationGranularity in WIN32, |
|
409 otherwise 64K. |
|
410 Also settable using mallopt(M_GRANULARITY, x) |
|
411 The unit for allocating and deallocating memory from the system. On |
|
412 most systems with contiguous MORECORE, there is no reason to |
|
413 make this more than a page. However, systems with MMAP tend to |
|
414 either require or encourage larger granularities. You can increase |
|
415 this value to prevent system allocation functions to be called so |
|
416 often, especially if they are slow. The value must be at least one |
|
417 page and must be a power of two. Setting to 0 causes initialization |
|
418 to either page size or win32 region size. (Note: In previous |
|
419 versions of malloc, the equivalent of this option was called |
|
420 "TOP_PAD") |
|
421 |
|
422 DEFAULT_TRIM_THRESHOLD default: 2MB |
|
423 Also settable using mallopt(M_TRIM_THRESHOLD, x) |
|
424 The maximum amount of unused top-most memory to keep before |
|
425 releasing via malloc_trim in free(). Automatic trimming is mainly |
|
426 useful in long-lived programs using contiguous MORECORE. Because |
|
427 trimming via sbrk can be slow on some systems, and can sometimes be |
|
428 wasteful (in cases where programs immediately afterward allocate |
|
429 more large chunks) the value should be high enough so that your |
|
430 overall system performance would improve by releasing this much |
|
431 memory. As a rough guide, you might set to a value close to the |
|
432 average size of a process (program) running on your system. |
|
433 Releasing this much memory would allow such a process to run in |
|
434 memory. Generally, it is worth tuning trim thresholds when a |
|
435 program undergoes phases where several large chunks are allocated |
|
436 and released in ways that can reuse each other's storage, perhaps |
|
437 mixed with phases where there are no such chunks at all. The trim |
|
438 value must be greater than page size to have any useful effect. To |
|
439 disable trimming completely, you can set to MAX_SIZE_T. Note that the trick |
|
440 some people use of mallocing a huge space and then freeing it at |
|
441 program startup, in an attempt to reserve system memory, doesn't |
|
442 have the intended effect under automatic trimming, since that memory |
|
443 will immediately be returned to the system. |
|
444 |
|
445 DEFAULT_MMAP_THRESHOLD default: 256K |
|
446 Also settable using mallopt(M_MMAP_THRESHOLD, x) |
|
447 The request size threshold for using MMAP to directly service a |
|
448 request. Requests of at least this size that cannot be allocated |
|
449 using already-existing space will be serviced via mmap. (If enough |
|
450 normal freed space already exists it is used instead.) Using mmap |
|
451 segregates relatively large chunks of memory so that they can be |
|
452 individually obtained and released from the host system. A request |
|
453 serviced through mmap is never reused by any other request (at least |
|
454 not directly; the system may just so happen to remap successive |
|
455 requests to the same locations). Segregating space in this way has |
|
456 the benefits that: Mmapped space can always be individually released |
|
457 back to the system, which helps keep the system level memory demands |
|
458 of a long-lived program low. Also, mapped memory doesn't become |
|
459 `locked' between other chunks, as can happen with normally allocated |
|
460 chunks, which means that even trimming via malloc_trim would not |
|
461 release them. However, it has the disadvantage that the space |
|
462 cannot be reclaimed, consolidated, and then used to service later |
|
463 requests, as happens with normal chunks. The advantages of mmap |
|
464 nearly always outweigh disadvantages for "large" chunks, but the |
|
465 value of "large" may vary across systems. The default is an |
|
466 empirically derived value that works well in most systems. You can |
|
467 disable mmap by setting to MAX_SIZE_T. |
|
468 |
|
469 MAX_RELEASE_CHECK_RATE default: 255 unless not HAVE_MMAP |
|
470 The number of consolidated frees between checks to release |
|
471 unused segments when freeing. When using non-contiguous segments, |
|
472 especially with multiple mspaces, checking only for topmost space |
|
473 doesn't always suffice to trigger trimming. To compensate for this, |
|
474 free() will, with a period of MAX_RELEASE_CHECK_RATE (or the |
|
475 current number of segments, if greater) try to release unused |
|
476 segments to the OS when freeing chunks that result in |
|
477 consolidation. The best value for this parameter is a compromise |
|
478 between slowing down frees with relatively costly checks that |
|
479 rarely trigger versus holding on to unused memory. To effectively |
|
480 disable, set to MAX_SIZE_T. This may lead to a very slight speed |
|
481 improvement at the expense of carrying around more memory. |
|
482 */ |
|
483 |
|
484 #ifndef WIN32 |
|
485 #ifdef _WIN32 |
|
486 #define WIN32 1 |
|
487 #endif /* _WIN32 */ |
|
488 #endif /* WIN32 */ |
|
489 #ifdef WIN32 |
|
490 #define WIN32_LEAN_AND_MEAN |
|
491 #include <windows.h> |
|
492 #define HAVE_MMAP 1 |
|
493 #define HAVE_MORECORE 0 |
|
494 #define LACKS_UNISTD_H |
|
495 #define LACKS_SYS_PARAM_H |
|
496 #define LACKS_SYS_MMAN_H |
|
497 #define LACKS_STRING_H |
|
498 #define LACKS_STRINGS_H |
|
499 #define LACKS_SYS_TYPES_H |
|
500 #define LACKS_ERRNO_H |
|
501 #define MALLOC_FAILURE_ACTION |
|
502 #ifdef _WIN32_WCE /* WINCE reportedly does not clear */ |
|
503 #define MMAP_CLEARS 0 |
|
504 #else |
|
505 #define MMAP_CLEARS 1 |
|
506 #endif /* _WIN32_WCE */ |
|
507 #endif /* WIN32 */ |
|
508 |
|
509 #if defined(DARWIN) || defined(_DARWIN) |
|
510 /* Mac OSX docs advise not to use sbrk; it seems better to use mmap */ |
|
511 #ifndef HAVE_MORECORE |
|
512 #define HAVE_MORECORE 0 |
|
513 #define HAVE_MMAP 1 |
|
514 #endif /* HAVE_MORECORE */ |
|
515 #endif /* DARWIN */ |
|
516 |
|
517 #ifndef LACKS_SYS_TYPES_H |
|
518 #include <sys/types.h> /* For size_t */ |
|
519 #endif /* LACKS_SYS_TYPES_H */ |
|
520 |
|
521 /* The maximum possible size_t value has all bits set */ |
|
522 #define MAX_SIZE_T (~(size_t)0) |
|
523 |
|
524 #ifndef ONLY_MSPACES |
|
525 #define ONLY_MSPACES 0 |
|
526 #endif /* ONLY_MSPACES */ |
|
527 #ifndef MSPACES |
|
528 #if ONLY_MSPACES |
|
529 #define MSPACES 1 |
|
530 #else /* ONLY_MSPACES */ |
|
531 #define MSPACES 0 |
|
532 #endif /* ONLY_MSPACES */ |
|
533 #endif /* MSPACES */ |
|
534 #ifndef MALLOC_ALIGNMENT |
|
535 #define MALLOC_ALIGNMENT ((size_t)8U) |
|
536 #endif /* MALLOC_ALIGNMENT */ |
|
537 #ifndef FOOTERS |
|
538 #define FOOTERS 0 |
|
539 #endif /* FOOTERS */ |
|
540 #ifndef ABORT |
|
541 #define ABORT abort() |
|
542 #endif /* ABORT */ |
|
543 #ifndef ABORT_ON_ASSERT_FAILURE |
|
544 #define ABORT_ON_ASSERT_FAILURE 1 |
|
545 #endif /* ABORT_ON_ASSERT_FAILURE */ |
|
546 #ifndef PROCEED_ON_ERROR |
|
547 #define PROCEED_ON_ERROR 0 |
|
548 #endif /* PROCEED_ON_ERROR */ |
|
549 #ifndef USE_LOCKS |
|
550 #define USE_LOCKS 0 |
|
551 #endif /* USE_LOCKS */ |
|
552 #ifndef USE_SPIN_LOCKS |
|
553 #if USE_LOCKS && (defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))) || (defined(_MSC_VER) && _MSC_VER>=1310) |
|
554 #define USE_SPIN_LOCKS 1 |
|
555 #else |
|
556 #define USE_SPIN_LOCKS 0 |
|
557 #endif /* USE_LOCKS && ... */ |
|
558 #endif /* USE_SPIN_LOCKS */ |
|
559 #ifndef INSECURE |
|
560 #define INSECURE 0 |
|
561 #endif /* INSECURE */ |
|
562 #ifndef HAVE_MMAP |
|
563 #define HAVE_MMAP 1 |
|
564 #endif /* HAVE_MMAP */ |
|
565 #ifndef MMAP_CLEARS |
|
566 #define MMAP_CLEARS 1 |
|
567 #endif /* MMAP_CLEARS */ |
|
568 #ifndef HAVE_MREMAP |
|
569 #ifdef linux |
|
570 #define HAVE_MREMAP 1 |
|
571 #else /* linux */ |
|
572 #define HAVE_MREMAP 0 |
|
573 #endif /* linux */ |
|
574 #endif /* HAVE_MREMAP */ |
|
575 #ifndef MALLOC_FAILURE_ACTION |
|
576 #define MALLOC_FAILURE_ACTION errno = ENOMEM; |
|
577 #endif /* MALLOC_FAILURE_ACTION */ |
|
578 #ifndef HAVE_MORECORE |
|
579 #if ONLY_MSPACES |
|
580 #define HAVE_MORECORE 0 |
|
581 #else /* ONLY_MSPACES */ |
|
582 #define HAVE_MORECORE 1 |
|
583 #endif /* ONLY_MSPACES */ |
|
584 #endif /* HAVE_MORECORE */ |
|
585 #if !HAVE_MORECORE |
|
586 #define MORECORE_CONTIGUOUS 0 |
|
587 #else /* !HAVE_MORECORE */ |
|
588 #ifndef MORECORE |
|
589 #define MORECORE sbrk |
|
590 #endif /* MORECORE */ |
|
591 #ifndef MORECORE_CONTIGUOUS |
|
592 #define MORECORE_CONTIGUOUS 1 |
|
593 #endif /* MORECORE_CONTIGUOUS */ |
|
594 #endif /* HAVE_MORECORE */ |
|
595 #ifndef DEFAULT_GRANULARITY |
|
596 #if MORECORE_CONTIGUOUS |
|
597 #define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */ |
|
598 #else /* MORECORE_CONTIGUOUS */ |
|
599 #define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U) |
|
600 #endif /* MORECORE_CONTIGUOUS */ |
|
601 #endif /* DEFAULT_GRANULARITY */ |
|
602 #ifndef DEFAULT_TRIM_THRESHOLD |
|
603 #ifndef MORECORE_CANNOT_TRIM |
|
604 #define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U) |
|
605 #else /* MORECORE_CANNOT_TRIM */ |
|
606 #define DEFAULT_TRIM_THRESHOLD MAX_SIZE_T |
|
607 #endif /* MORECORE_CANNOT_TRIM */ |
|
608 #endif /* DEFAULT_TRIM_THRESHOLD */ |
|
609 #ifndef DEFAULT_MMAP_THRESHOLD |
|
610 #if HAVE_MMAP |
|
611 #define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U) |
|
612 #else /* HAVE_MMAP */ |
|
613 #define DEFAULT_MMAP_THRESHOLD MAX_SIZE_T |
|
614 #endif /* HAVE_MMAP */ |
|
615 #endif /* DEFAULT_MMAP_THRESHOLD */ |
|
616 #ifndef MAX_RELEASE_CHECK_RATE |
|
617 #if HAVE_MMAP |
|
618 #define MAX_RELEASE_CHECK_RATE 255 |
|
619 #else |
|
620 #define MAX_RELEASE_CHECK_RATE MAX_SIZE_T |
|
621 #endif /* HAVE_MMAP */ |
|
622 #endif /* MAX_RELEASE_CHECK_RATE */ |
|
623 #ifndef USE_BUILTIN_FFS |
|
624 #define USE_BUILTIN_FFS 0 |
|
625 #endif /* USE_BUILTIN_FFS */ |
|
626 #ifndef USE_DEV_RANDOM |
|
627 #define USE_DEV_RANDOM 0 |
|
628 #endif /* USE_DEV_RANDOM */ |
|
629 #ifndef NO_MALLINFO |
|
630 #define NO_MALLINFO 0 |
|
631 #endif /* NO_MALLINFO */ |
|
632 #ifndef MALLINFO_FIELD_TYPE |
|
633 #define MALLINFO_FIELD_TYPE size_t |
|
634 #endif /* MALLINFO_FIELD_TYPE */ |
|
635 #ifndef NO_SEGMENT_TRAVERSAL |
|
636 #define NO_SEGMENT_TRAVERSAL 0 |
|
637 #endif /* NO_SEGMENT_TRAVERSAL */ |
|
638 |
|
639 /* |
|
640 mallopt tuning options. SVID/XPG defines four standard parameter |
|
641 numbers for mallopt, normally defined in malloc.h. None of these |
|
642 are used in this malloc, so setting them has no effect. But this |
|
643 malloc does support the following options. |
|
644 */ |
|
645 |
|
646 #define M_TRIM_THRESHOLD (-1) |
|
647 #define M_GRANULARITY (-2) |
|
648 #define M_MMAP_THRESHOLD (-3) |
|
649 |
|
650 /* ------------------------ Mallinfo declarations ------------------------ */ |
|
651 |
|
652 #if !NO_MALLINFO |
|
653 /* |
|
654 This version of malloc supports the standard SVID/XPG mallinfo |
|
655 routine that returns a struct containing usage properties and |
|
656 statistics. It should work on any system that has a |
|
657 /usr/include/malloc.h defining struct mallinfo. The main |
|
658 declaration needed is the mallinfo struct that is returned (by-copy) |
|
659 by mallinfo(). The malloinfo struct contains a bunch of fields that |
|
660 are not even meaningful in this version of malloc. These fields are |
|
661 are instead filled by mallinfo() with other numbers that might be of |
|
662 interest. |
|
663 |
|
664 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a |
|
665 /usr/include/malloc.h file that includes a declaration of struct |
|
666 mallinfo. If so, it is included; else a compliant version is |
|
667 declared below. These must be precisely the same for mallinfo() to |
|
668 work. The original SVID version of this struct, defined on most |
|
669 systems with mallinfo, declares all fields as ints. But some others |
|
670 define as unsigned long. If your system defines the fields using a |
|
671 type of different width than listed here, you MUST #include your |
|
672 system version and #define HAVE_USR_INCLUDE_MALLOC_H. |
|
673 */ |
|
674 |
|
675 /* #define HAVE_USR_INCLUDE_MALLOC_H */ |
|
676 |
|
677 #ifdef HAVE_USR_INCLUDE_MALLOC_H |
|
678 #include "/usr/include/malloc.h" |
|
679 #else /* HAVE_USR_INCLUDE_MALLOC_H */ |
|
680 |
|
681 struct mallinfo { |
|
682 MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */ |
|
683 MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */ |
|
684 MALLINFO_FIELD_TYPE smblks; /* always 0 */ |
|
685 MALLINFO_FIELD_TYPE hblks; /* always 0 */ |
|
686 MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */ |
|
687 MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */ |
|
688 MALLINFO_FIELD_TYPE fsmblks; /* always 0 */ |
|
689 MALLINFO_FIELD_TYPE uordblks; /* total allocated space */ |
|
690 MALLINFO_FIELD_TYPE fordblks; /* total free space */ |
|
691 MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */ |
|
692 }; |
|
693 |
|
694 #endif /* HAVE_USR_INCLUDE_MALLOC_H */ |
|
695 #endif /* NO_MALLINFO */ |
|
696 |
|
697 /* |
|
698 Try to persuade compilers to inline. The most critical functions for |
|
699 inlining are defined as macros, so these aren't used for them. |
|
700 */ |
|
701 |
|
702 #ifndef FORCEINLINE |
|
703 #if defined(__GNUC__) |
|
704 #define FORCEINLINE __inline __attribute__ ((always_inline)) |
|
705 #elif defined(_MSC_VER) |
|
706 #define FORCEINLINE __forceinline |
|
707 #endif |
|
708 #endif |
|
709 #ifndef NOINLINE |
|
710 #if defined(__GNUC__) |
|
711 #define NOINLINE __attribute__ ((noinline)) |
|
712 #elif defined(_MSC_VER) |
|
713 #define NOINLINE __declspec(noinline) |
|
714 #else |
|
715 #define NOINLINE |
|
716 #endif |
|
717 #endif |
|
718 |
|
719 #ifdef __cplusplus |
|
720 extern "C" { |
|
721 #ifndef FORCEINLINE |
|
722 #define FORCEINLINE inline |
|
723 #endif |
|
724 #endif /* __cplusplus */ |
|
725 #ifndef FORCEINLINE |
|
726 #define FORCEINLINE |
|
727 #endif |
|
728 |
|
729 #if !ONLY_MSPACES |
|
730 |
|
731 /* ------------------- Declarations of public routines ------------------- */ |
|
732 |
|
733 #ifndef USE_DL_PREFIX |
|
734 #define dlcalloc calloc |
|
735 #define dlfree free |
|
736 #define dlmalloc malloc |
|
737 #define dlmemalign memalign |
|
738 #define dlrealloc realloc |
|
739 #define dlvalloc valloc |
|
740 #define dlpvalloc pvalloc |
|
741 #define dlmallinfo mallinfo |
|
742 #define dlmallopt mallopt |
|
743 #define dlmalloc_trim malloc_trim |
|
744 #define dlmalloc_stats malloc_stats |
|
745 #define dlmalloc_usable_size malloc_usable_size |
|
746 #define dlmalloc_footprint malloc_footprint |
|
747 #define dlmalloc_max_footprint malloc_max_footprint |
|
748 #define dlindependent_calloc independent_calloc |
|
749 #define dlindependent_comalloc independent_comalloc |
|
750 #endif /* USE_DL_PREFIX */ |
|
751 |
|
752 |
|
753 /* |
|
754 malloc(size_t n) |
|
755 Returns a pointer to a newly allocated chunk of at least n bytes, or |
|
756 null if no space is available, in which case errno is set to ENOMEM |
|
757 on ANSI C systems. |
|
758 |
|
759 If n is zero, malloc returns a minimum-sized chunk. (The minimum |
|
760 size is 16 bytes on most 32bit systems, and 32 bytes on 64bit |
|
761 systems.) Note that size_t is an unsigned type, so calls with |
|
762 arguments that would be negative if signed are interpreted as |
|
763 requests for huge amounts of space, which will often fail. The |
|
764 maximum supported value of n differs across systems, but is in all |
|
765 cases less than the maximum representable value of a size_t. |
|
766 */ |
|
767 void* dlmalloc(size_t); |
|
768 |
|
769 /* |
|
770 free(void* p) |
|
771 Releases the chunk of memory pointed to by p, that had been previously |
|
772 allocated using malloc or a related routine such as realloc. |
|
773 It has no effect if p is null. If p was not malloced or already |
|
774 freed, free(p) will by default cause the current program to abort. |
|
775 */ |
|
776 void dlfree(void*); |
|
777 |
|
778 /* |
|
779 calloc(size_t n_elements, size_t element_size); |
|
780 Returns a pointer to n_elements * element_size bytes, with all locations |
|
781 set to zero. |
|
782 */ |
|
783 void* dlcalloc(size_t, size_t); |
|
784 |
|
785 /* |
|
786 realloc(void* p, size_t n) |
|
787 Returns a pointer to a chunk of size n that contains the same data |
|
788 as does chunk p up to the minimum of (n, p's size) bytes, or null |
|
789 if no space is available. |
|
790 |
|
791 The returned pointer may or may not be the same as p. The algorithm |
|
792 prefers extending p in most cases when possible, otherwise it |
|
793 employs the equivalent of a malloc-copy-free sequence. |
|
794 |
|
795 If p is null, realloc is equivalent to malloc. |
|
796 |
|
797 If space is not available, realloc returns null, errno is set (if on |
|
798 ANSI) and p is NOT freed. |
|
799 |
|
800 if n is for fewer bytes than already held by p, the newly unused |
|
801 space is lopped off and freed if possible. realloc with a size |
|
802 argument of zero (re)allocates a minimum-sized chunk. |
|
803 |
|
804 The old unix realloc convention of allowing the last-free'd chunk |
|
805 to be used as an argument to realloc is not supported. |
|
806 */ |
|
807 |
|
808 void* dlrealloc(void*, size_t); |
|
809 |
|
810 /* |
|
811 memalign(size_t alignment, size_t n); |
|
812 Returns a pointer to a newly allocated chunk of n bytes, aligned |
|
813 in accord with the alignment argument. |
|
814 |
|
815 The alignment argument should be a power of two. If the argument is |
|
816 not a power of two, the nearest greater power is used. |
|
817 8-byte alignment is guaranteed by normal malloc calls, so don't |
|
818 bother calling memalign with an argument of 8 or less. |
|
819 |
|
820 Overreliance on memalign is a sure way to fragment space. |
|
821 */ |
|
822 void* dlmemalign(size_t, size_t); |
|
823 |
|
824 /* |
|
825 valloc(size_t n); |
|
826 Equivalent to memalign(pagesize, n), where pagesize is the page |
|
827 size of the system. If the pagesize is unknown, 4096 is used. |
|
828 */ |
|
829 void* dlvalloc(size_t); |
|
830 |
|
831 /* |
|
832 mallopt(int parameter_number, int parameter_value) |
|
833 Sets tunable parameters The format is to provide a |
|
834 (parameter-number, parameter-value) pair. mallopt then sets the |
|
835 corresponding parameter to the argument value if it can (i.e., so |
|
836 long as the value is meaningful), and returns 1 if successful else |
|
837 0. SVID/XPG/ANSI defines four standard param numbers for mallopt, |
|
838 normally defined in malloc.h. None of these are use in this malloc, |
|
839 so setting them has no effect. But this malloc also supports other |
|
840 options in mallopt. See below for details. Briefly, supported |
|
841 parameters are as follows (listed defaults are for "typical" |
|
842 configurations). |
|
843 |
|
844 Symbol param # default allowed param values |
|
845 M_TRIM_THRESHOLD -1 2*1024*1024 any (MAX_SIZE_T disables) |
|
846 M_GRANULARITY -2 page size any power of 2 >= page size |
|
847 M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support) |
|
848 */ |
|
849 int dlmallopt(int, int); |
|
850 |
|
851 /* |
|
852 malloc_footprint(); |
|
853 Returns the number of bytes obtained from the system. The total |
|
854 number of bytes allocated by malloc, realloc etc., is less than this |
|
855 value. Unlike mallinfo, this function returns only a precomputed |
|
856 result, so can be called frequently to monitor memory consumption. |
|
857 Even if locks are otherwise defined, this function does not use them, |
|
858 so results might not be up to date. |
|
859 */ |
|
860 size_t dlmalloc_footprint(void); |
|
861 |
|
862 /* |
|
863 malloc_max_footprint(); |
|
864 Returns the maximum number of bytes obtained from the system. This |
|
865 value will be greater than current footprint if deallocated space |
|
866 has been reclaimed by the system. The peak number of bytes allocated |
|
867 by malloc, realloc etc., is less than this value. Unlike mallinfo, |
|
868 this function returns only a precomputed result, so can be called |
|
869 frequently to monitor memory consumption. Even if locks are |
|
870 otherwise defined, this function does not use them, so results might |
|
871 not be up to date. |
|
872 */ |
|
873 size_t dlmalloc_max_footprint(void); |
|
874 |
|
875 #if !NO_MALLINFO |
|
876 /* |
|
877 mallinfo() |
|
878 Returns (by copy) a struct containing various summary statistics: |
|
879 |
|
880 arena: current total non-mmapped bytes allocated from system |
|
881 ordblks: the number of free chunks |
|
882 smblks: always zero. |
|
883 hblks: current number of mmapped regions |
|
884 hblkhd: total bytes held in mmapped regions |
|
885 usmblks: the maximum total allocated space. This will be greater |
|
886 than current total if trimming has occurred. |
|
887 fsmblks: always zero |
|
888 uordblks: current total allocated space (normal or mmapped) |
|
889 fordblks: total free space |
|
890 keepcost: the maximum number of bytes that could ideally be released |
|
891 back to system via malloc_trim. ("ideally" means that |
|
892 it ignores page restrictions etc.) |
|
893 |
|
894 Because these fields are ints, but internal bookkeeping may |
|
895 be kept as longs, the reported values may wrap around zero and |
|
896 thus be inaccurate. |
|
897 */ |
|
898 struct mallinfo dlmallinfo(void); |
|
899 #endif /* NO_MALLINFO */ |
|
900 |
|
901 /* |
|
902 independent_calloc(size_t n_elements, size_t element_size, void* chunks[]); |
|
903 |
|
904 independent_calloc is similar to calloc, but instead of returning a |
|
905 single cleared space, it returns an array of pointers to n_elements |
|
906 independent elements that can hold contents of size elem_size, each |
|
907 of which starts out cleared, and can be independently freed, |
|
908 realloc'ed etc. The elements are guaranteed to be adjacently |
|
909 allocated (this is not guaranteed to occur with multiple callocs or |
|
910 mallocs), which may also improve cache locality in some |
|
911 applications. |
|
912 |
|
913 The "chunks" argument is optional (i.e., may be null, which is |
|
914 probably the most typical usage). If it is null, the returned array |
|
915 is itself dynamically allocated and should also be freed when it is |
|
916 no longer needed. Otherwise, the chunks array must be of at least |
|
917 n_elements in length. It is filled in with the pointers to the |
|
918 chunks. |
|
919 |
|
920 In either case, independent_calloc returns this pointer array, or |
|
921 null if the allocation failed. If n_elements is zero and "chunks" |
|
922 is null, it returns a chunk representing an array with zero elements |
|
923 (which should be freed if not wanted). |
|
924 |
|
925 Each element must be individually freed when it is no longer |
|
926 needed. If you'd like to instead be able to free all at once, you |
|
927 should instead use regular calloc and assign pointers into this |
|
928 space to represent elements. (In this case though, you cannot |
|
929 independently free elements.) |
|
930 |
|
931 independent_calloc simplifies and speeds up implementations of many |
|
932 kinds of pools. It may also be useful when constructing large data |
|
933 structures that initially have a fixed number of fixed-sized nodes, |
|
934 but the number is not known at compile time, and some of the nodes |
|
935 may later need to be freed. For example: |
|
936 |
|
937 struct Node { int item; struct Node* next; }; |
|
938 |
|
939 struct Node* build_list() { |
|
940 struct Node** pool; |
|
941 int n = read_number_of_nodes_needed(); |
|
942 if (n <= 0) return 0; |
|
943 pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0); |
|
944 if (pool == 0) die(); |
|
945 // organize into a linked list... |
|
946 struct Node* first = pool[0]; |
|
947 for (i = 0; i < n-1; ++i) |
|
948 pool[i]->next = pool[i+1]; |
|
949 free(pool); // Can now free the array (or not, if it is needed later) |
|
950 return first; |
|
951 } |
|
952 */ |
|
953 void** dlindependent_calloc(size_t, size_t, void**); |
|
954 |
|
955 /* |
|
956 independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]); |
|
957 |
|
958 independent_comalloc allocates, all at once, a set of n_elements |
|
959 chunks with sizes indicated in the "sizes" array. It returns |
|
960 an array of pointers to these elements, each of which can be |
|
961 independently freed, realloc'ed etc. The elements are guaranteed to |
|
962 be adjacently allocated (this is not guaranteed to occur with |
|
963 multiple callocs or mallocs), which may also improve cache locality |
|
964 in some applications. |
|
965 |
|
966 The "chunks" argument is optional (i.e., may be null). If it is null |
|
967 the returned array is itself dynamically allocated and should also |
|
968 be freed when it is no longer needed. Otherwise, the chunks array |
|
969 must be of at least n_elements in length. It is filled in with the |
|
970 pointers to the chunks. |
|
971 |
|
972 In either case, independent_comalloc returns this pointer array, or |
|
973 null if the allocation failed. If n_elements is zero and chunks is |
|
974 null, it returns a chunk representing an array with zero elements |
|
975 (which should be freed if not wanted). |
|
976 |
|
977 Each element must be individually freed when it is no longer |
|
978 needed. If you'd like to instead be able to free all at once, you |
|
979 should instead use a single regular malloc, and assign pointers at |
|
980 particular offsets in the aggregate space. (In this case though, you |
|
981 cannot independently free elements.) |
|
982 |
|
983 independent_comallac differs from independent_calloc in that each |
|
984 element may have a different size, and also that it does not |
|
985 automatically clear elements. |
|
986 |
|
987 independent_comalloc can be used to speed up allocation in cases |
|
988 where several structs or objects must always be allocated at the |
|
989 same time. For example: |
|
990 |
|
991 struct Head { ... } |
|
992 struct Foot { ... } |
|
993 |
|
994 void send_message(char* msg) { |
|
995 int msglen = strlen(msg); |
|
996 size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) }; |
|
997 void* chunks[3]; |
|
998 if (independent_comalloc(3, sizes, chunks) == 0) |
|
999 die(); |
|
1000 struct Head* head = (struct Head*)(chunks[0]); |
|
1001 char* body = (char*)(chunks[1]); |
|
1002 struct Foot* foot = (struct Foot*)(chunks[2]); |
|
1003 // ... |
|
1004 } |
|
1005 |
|
1006 In general though, independent_comalloc is worth using only for |
|
1007 larger values of n_elements. For small values, you probably won't |
|
1008 detect enough difference from series of malloc calls to bother. |
|
1009 |
|
1010 Overuse of independent_comalloc can increase overall memory usage, |
|
1011 since it cannot reuse existing noncontiguous small chunks that |
|
1012 might be available for some of the elements. |
|
1013 */ |
|
1014 void** dlindependent_comalloc(size_t, size_t*, void**); |
|
1015 |
|
1016 |
|
1017 /* |
|
1018 pvalloc(size_t n); |
|
1019 Equivalent to valloc(minimum-page-that-holds(n)), that is, |
|
1020 round up n to nearest pagesize. |
|
1021 */ |
|
1022 void* dlpvalloc(size_t); |
|
1023 |
|
1024 /* |
|
1025 malloc_trim(size_t pad); |
|
1026 |
|
1027 If possible, gives memory back to the system (via negative arguments |
|
1028 to sbrk) if there is unused memory at the `high' end of the malloc |
|
1029 pool or in unused MMAP segments. You can call this after freeing |
|
1030 large blocks of memory to potentially reduce the system-level memory |
|
1031 requirements of a program. However, it cannot guarantee to reduce |
|
1032 memory. Under some allocation patterns, some large free blocks of |
|
1033 memory will be locked between two used chunks, so they cannot be |
|
1034 given back to the system. |
|
1035 |
|
1036 The `pad' argument to malloc_trim represents the amount of free |
|
1037 trailing space to leave untrimmed. If this argument is zero, only |
|
1038 the minimum amount of memory to maintain internal data structures |
|
1039 will be left. Non-zero arguments can be supplied to maintain enough |
|
1040 trailing space to service future expected allocations without having |
|
1041 to re-obtain memory from the system. |
|
1042 |
|
1043 Malloc_trim returns 1 if it actually released any memory, else 0. |
|
1044 */ |
|
1045 int dlmalloc_trim(size_t); |
|
1046 |
|
1047 /* |
|
1048 malloc_usable_size(void* p); |
|
1049 |
|
1050 Returns the number of bytes you can actually use in |
|
1051 an allocated chunk, which may be more than you requested (although |
|
1052 often not) due to alignment and minimum size constraints. |
|
1053 You can use this many bytes without worrying about |
|
1054 overwriting other allocated objects. This is not a particularly great |
|
1055 programming practice. malloc_usable_size can be more useful in |
|
1056 debugging and assertions, for example: |
|
1057 |
|
1058 p = malloc(n); |
|
1059 assert(malloc_usable_size(p) >= 256); |
|
1060 */ |
|
1061 size_t dlmalloc_usable_size(void*); |
|
1062 |
|
1063 /* |
|
1064 malloc_stats(); |
|
1065 Prints on stderr the amount of space obtained from the system (both |
|
1066 via sbrk and mmap), the maximum amount (which may be more than |
|
1067 current if malloc_trim and/or munmap got called), and the current |
|
1068 number of bytes allocated via malloc (or realloc, etc) but not yet |
|
1069 freed. Note that this is the number of bytes allocated, not the |
|
1070 number requested. It will be larger than the number requested |
|
1071 because of alignment and bookkeeping overhead. Because it includes |
|
1072 alignment wastage as being in use, this figure may be greater than |
|
1073 zero even when no user-level chunks are allocated. |
|
1074 |
|
1075 The reported current and maximum system memory can be inaccurate if |
|
1076 a program makes other calls to system memory allocation functions |
|
1077 (normally sbrk) outside of malloc. |
|
1078 |
|
1079 malloc_stats prints only the most commonly interesting statistics. |
|
1080 More information can be obtained by calling mallinfo. |
|
1081 */ |
|
1082 void dlmalloc_stats(void); |
|
1083 |
|
1084 #endif /* ONLY_MSPACES */ |
|
1085 |
|
1086 #if MSPACES |
|
1087 |
|
1088 /* |
|
1089 mspace is an opaque type representing an independent |
|
1090 region of space that supports mspace_malloc, etc. |
|
1091 */ |
|
1092 typedef void* mspace; |
|
1093 |
|
1094 /* |
|
1095 create_mspace creates and returns a new independent space with the |
|
1096 given initial capacity, or, if 0, the default granularity size. It |
|
1097 returns null if there is no system memory available to create the |
|
1098 space. If argument locked is non-zero, the space uses a separate |
|
1099 lock to control access. The capacity of the space will grow |
|
1100 dynamically as needed to service mspace_malloc requests. You can |
|
1101 control the sizes of incremental increases of this space by |
|
1102 compiling with a different DEFAULT_GRANULARITY or dynamically |
|
1103 setting with mallopt(M_GRANULARITY, value). |
|
1104 */ |
|
1105 mspace create_mspace(size_t capacity, int locked); |
|
1106 |
|
1107 /* |
|
1108 destroy_mspace destroys the given space, and attempts to return all |
|
1109 of its memory back to the system, returning the total number of |
|
1110 bytes freed. After destruction, the results of access to all memory |
|
1111 used by the space become undefined. |
|
1112 */ |
|
1113 size_t destroy_mspace(mspace msp); |
|
1114 |
|
1115 /* |
|
1116 create_mspace_with_base uses the memory supplied as the initial base |
|
1117 of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this |
|
1118 space is used for bookkeeping, so the capacity must be at least this |
|
1119 large. (Otherwise 0 is returned.) When this initial space is |
|
1120 exhausted, additional memory will be obtained from the system. |
|
1121 Destroying this space will deallocate all additionally allocated |
|
1122 space (if possible) but not the initial base. |
|
1123 */ |
|
1124 mspace create_mspace_with_base(void* base, size_t capacity, int locked); |
|
1125 |
|
1126 /* |
|
1127 mspace_malloc behaves as malloc, but operates within |
|
1128 the given space. |
|
1129 */ |
|
1130 void* mspace_malloc(mspace msp, size_t bytes); |
|
1131 |
|
1132 /* |
|
1133 mspace_free behaves as free, but operates within |
|
1134 the given space. |
|
1135 |
|
1136 If compiled with FOOTERS==1, mspace_free is not actually needed. |
|
1137 free may be called instead of mspace_free because freed chunks from |
|
1138 any space are handled by their originating spaces. |
|
1139 */ |
|
1140 void mspace_free(mspace msp, void* mem); |
|
1141 |
|
1142 /* |
|
1143 mspace_realloc behaves as realloc, but operates within |
|
1144 the given space. |
|
1145 |
|
1146 If compiled with FOOTERS==1, mspace_realloc is not actually |
|
1147 needed. realloc may be called instead of mspace_realloc because |
|
1148 realloced chunks from any space are handled by their originating |
|
1149 spaces. |
|
1150 */ |
|
1151 void* mspace_realloc(mspace msp, void* mem, size_t newsize); |
|
1152 |
|
1153 /* |
|
1154 mspace_calloc behaves as calloc, but operates within |
|
1155 the given space. |
|
1156 */ |
|
1157 void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size); |
|
1158 |
|
1159 /* |
|
1160 mspace_memalign behaves as memalign, but operates within |
|
1161 the given space. |
|
1162 */ |
|
1163 void* mspace_memalign(mspace msp, size_t alignment, size_t bytes); |
|
1164 |
|
1165 /* |
|
1166 mspace_independent_calloc behaves as independent_calloc, but |
|
1167 operates within the given space. |
|
1168 */ |
|
1169 void** mspace_independent_calloc(mspace msp, size_t n_elements, |
|
1170 size_t elem_size, void* chunks[]); |
|
1171 |
|
1172 /* |
|
1173 mspace_independent_comalloc behaves as independent_comalloc, but |
|
1174 operates within the given space. |
|
1175 */ |
|
1176 void** mspace_independent_comalloc(mspace msp, size_t n_elements, |
|
1177 size_t sizes[], void* chunks[]); |
|
1178 |
|
1179 /* |
|
1180 mspace_footprint() returns the number of bytes obtained from the |
|
1181 system for this space. |
|
1182 */ |
|
1183 size_t mspace_footprint(mspace msp); |
|
1184 |
|
1185 /* |
|
1186 mspace_max_footprint() returns the peak number of bytes obtained from the |
|
1187 system for this space. |
|
1188 */ |
|
1189 size_t mspace_max_footprint(mspace msp); |
|
1190 |
|
1191 |
|
1192 #if !NO_MALLINFO |
|
1193 /* |
|
1194 mspace_mallinfo behaves as mallinfo, but reports properties of |
|
1195 the given space. |
|
1196 */ |
|
1197 struct mallinfo mspace_mallinfo(mspace msp); |
|
1198 #endif /* NO_MALLINFO */ |
|
1199 |
|
1200 /* |
|
1201 mspace_malloc_stats behaves as malloc_stats, but reports |
|
1202 properties of the given space. |
|
1203 */ |
|
1204 void mspace_malloc_stats(mspace msp); |
|
1205 |
|
1206 /* |
|
1207 mspace_trim behaves as malloc_trim, but |
|
1208 operates within the given space. |
|
1209 */ |
|
1210 int mspace_trim(mspace msp, size_t pad); |
|
1211 |
|
1212 /* |
|
1213 An alias for mallopt. |
|
1214 */ |
|
1215 int mspace_mallopt(int, int); |
|
1216 |
|
1217 #endif /* MSPACES */ |
|
1218 |
|
1219 #ifdef __cplusplus |
|
1220 }; /* end of extern "C" */ |
|
1221 #endif /* __cplusplus */ |
|
1222 |
|
1223 /* |
|
1224 ======================================================================== |
|
1225 To make a fully customizable malloc.h header file, cut everything |
|
1226 above this line, put into file malloc.h, edit to suit, and #include it |
|
1227 on the next line, as well as in programs that use this malloc. |
|
1228 ======================================================================== |
|
1229 */ |
|
1230 |
|
1231 /* #include "malloc.h" */ |
|
1232 |
|
1233 /*------------------------------ internal #includes ---------------------- */ |
|
1234 |
|
1235 #ifdef WIN32 |
|
1236 #pragma warning( disable : 4146 ) /* no "unsigned" warnings */ |
|
1237 #endif /* WIN32 */ |
|
1238 |
|
1239 #include <stdio.h> /* for printing in malloc_stats */ |
|
1240 |
|
1241 #ifndef LACKS_ERRNO_H |
|
1242 #include <errno.h> /* for MALLOC_FAILURE_ACTION */ |
|
1243 #endif /* LACKS_ERRNO_H */ |
|
1244 #if FOOTERS |
|
1245 #include <time.h> /* for magic initialization */ |
|
1246 #endif /* FOOTERS */ |
|
1247 #ifndef LACKS_STDLIB_H |
|
1248 #include <stdlib.h> /* for abort() */ |
|
1249 #endif /* LACKS_STDLIB_H */ |
|
1250 #ifdef DEBUG |
|
1251 #if ABORT_ON_ASSERT_FAILURE |
|
1252 #define assert(x) if(!(x)) ABORT |
|
1253 #else /* ABORT_ON_ASSERT_FAILURE */ |
|
1254 #include <assert.h> |
|
1255 #endif /* ABORT_ON_ASSERT_FAILURE */ |
|
1256 #else /* DEBUG */ |
|
1257 #define assert(x) |
|
1258 #endif /* DEBUG */ |
|
1259 #ifndef LACKS_STRING_H |
|
1260 #include <string.h> /* for memset etc */ |
|
1261 #endif /* LACKS_STRING_H */ |
|
1262 #if USE_BUILTIN_FFS |
|
1263 #ifndef LACKS_STRINGS_H |
|
1264 #include <strings.h> /* for ffs */ |
|
1265 #endif /* LACKS_STRINGS_H */ |
|
1266 #endif /* USE_BUILTIN_FFS */ |
|
1267 #if HAVE_MMAP |
|
1268 #ifndef LACKS_SYS_MMAN_H |
|
1269 #include <sys/mman.h> /* for mmap */ |
|
1270 #endif /* LACKS_SYS_MMAN_H */ |
|
1271 #ifndef LACKS_FCNTL_H |
|
1272 #include <fcntl.h> |
|
1273 #endif /* LACKS_FCNTL_H */ |
|
1274 #endif /* HAVE_MMAP */ |
|
1275 #if HAVE_MORECORE |
|
1276 #ifndef LACKS_UNISTD_H |
|
1277 #include <unistd.h> /* for sbrk */ |
|
1278 #else /* LACKS_UNISTD_H */ |
|
1279 #if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__) |
|
1280 extern void* sbrk(ptrdiff_t); |
|
1281 #endif /* FreeBSD etc */ |
|
1282 #endif /* LACKS_UNISTD_H */ |
|
1283 #endif /* HAVE_MMAP */ |
|
1284 |
|
1285 /* Declarations for locking */ |
|
1286 #if USE_LOCKS |
|
1287 #ifndef WIN32 |
|
1288 #include <pthread.h> |
|
1289 #if defined (__SVR4) && defined (__sun) /* solaris */ |
|
1290 #include <thread.h> |
|
1291 #endif /* solaris */ |
|
1292 #else |
|
1293 #ifndef _M_AMD64 |
|
1294 /* These are already defined on AMD64 builds */ |
|
1295 #ifdef __cplusplus |
|
1296 extern "C" { |
|
1297 #endif /* __cplusplus */ |
|
1298 LONG __cdecl _InterlockedCompareExchange(LPLONG volatile Dest, LONG Exchange, LONG Comp); |
|
1299 LONG __cdecl _InterlockedExchange(LPLONG volatile Target, LONG Value); |
|
1300 #ifdef __cplusplus |
|
1301 } |
|
1302 #endif /* __cplusplus */ |
|
1303 #endif /* _M_AMD64 */ |
|
1304 #pragma intrinsic (_InterlockedCompareExchange) |
|
1305 #pragma intrinsic (_InterlockedExchange) |
|
1306 #define interlockedcompareexchange _InterlockedCompareExchange |
|
1307 #define interlockedexchange _InterlockedExchange |
|
1308 #endif /* Win32 */ |
|
1309 #endif /* USE_LOCKS */ |
|
1310 |
|
1311 /* Declarations for bit scanning on win32 */ |
|
1312 #if defined(_MSC_VER) && _MSC_VER>=1300 |
|
1313 #ifndef BitScanForward /* Try to avoid pulling in WinNT.h */ |
|
1314 #ifdef __cplusplus |
|
1315 extern "C" { |
|
1316 #endif /* __cplusplus */ |
|
1317 unsigned char _BitScanForward(unsigned long *index, unsigned long mask); |
|
1318 unsigned char _BitScanReverse(unsigned long *index, unsigned long mask); |
|
1319 #ifdef __cplusplus |
|
1320 } |
|
1321 #endif /* __cplusplus */ |
|
1322 |
|
1323 #define BitScanForward _BitScanForward |
|
1324 #define BitScanReverse _BitScanReverse |
|
1325 #pragma intrinsic(_BitScanForward) |
|
1326 #pragma intrinsic(_BitScanReverse) |
|
1327 #endif /* BitScanForward */ |
|
1328 #endif /* defined(_MSC_VER) && _MSC_VER>=1300 */ |
|
1329 |
|
1330 #ifndef WIN32 |
|
1331 #ifndef malloc_getpagesize |
|
1332 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */ |
|
1333 # ifndef _SC_PAGE_SIZE |
|
1334 # define _SC_PAGE_SIZE _SC_PAGESIZE |
|
1335 # endif |
|
1336 # endif |
|
1337 # ifdef _SC_PAGE_SIZE |
|
1338 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE) |
|
1339 # else |
|
1340 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE) |
|
1341 extern size_t getpagesize(); |
|
1342 # define malloc_getpagesize getpagesize() |
|
1343 # else |
|
1344 # ifdef WIN32 /* use supplied emulation of getpagesize */ |
|
1345 # define malloc_getpagesize getpagesize() |
|
1346 # else |
|
1347 # ifndef LACKS_SYS_PARAM_H |
|
1348 # include <sys/param.h> |
|
1349 # endif |
|
1350 # ifdef EXEC_PAGESIZE |
|
1351 # define malloc_getpagesize EXEC_PAGESIZE |
|
1352 # else |
|
1353 # ifdef NBPG |
|
1354 # ifndef CLSIZE |
|
1355 # define malloc_getpagesize NBPG |
|
1356 # else |
|
1357 # define malloc_getpagesize (NBPG * CLSIZE) |
|
1358 # endif |
|
1359 # else |
|
1360 # ifdef NBPC |
|
1361 # define malloc_getpagesize NBPC |
|
1362 # else |
|
1363 # ifdef PAGESIZE |
|
1364 # define malloc_getpagesize PAGESIZE |
|
1365 # else /* just guess */ |
|
1366 # define malloc_getpagesize ((size_t)4096U) |
|
1367 # endif |
|
1368 # endif |
|
1369 # endif |
|
1370 # endif |
|
1371 # endif |
|
1372 # endif |
|
1373 # endif |
|
1374 #endif |
|
1375 #endif |
|
1376 |
|
1377 |
|
1378 |
|
1379 /* ------------------- size_t and alignment properties -------------------- */ |
|
1380 |
|
1381 /* The byte and bit size of a size_t */ |
|
1382 #define SIZE_T_SIZE (sizeof(size_t)) |
|
1383 #define SIZE_T_BITSIZE (sizeof(size_t) << 3) |
|
1384 |
|
1385 /* Some constants coerced to size_t */ |
|
1386 /* Annoying but necessary to avoid errors on some platforms */ |
|
1387 #define SIZE_T_ZERO ((size_t)0) |
|
1388 #define SIZE_T_ONE ((size_t)1) |
|
1389 #define SIZE_T_TWO ((size_t)2) |
|
1390 #define SIZE_T_FOUR ((size_t)4) |
|
1391 #define TWO_SIZE_T_SIZES (SIZE_T_SIZE<<1) |
|
1392 #define FOUR_SIZE_T_SIZES (SIZE_T_SIZE<<2) |
|
1393 #define SIX_SIZE_T_SIZES (FOUR_SIZE_T_SIZES+TWO_SIZE_T_SIZES) |
|
1394 #define HALF_MAX_SIZE_T (MAX_SIZE_T / 2U) |
|
1395 |
|
1396 /* The bit mask value corresponding to MALLOC_ALIGNMENT */ |
|
1397 #define CHUNK_ALIGN_MASK (MALLOC_ALIGNMENT - SIZE_T_ONE) |
|
1398 |
|
1399 /* True if address a has acceptable alignment */ |
|
1400 #define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0) |
|
1401 |
|
1402 /* the number of bytes to offset an address to align it */ |
|
1403 #define align_offset(A)\ |
|
1404 ((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\ |
|
1405 ((MALLOC_ALIGNMENT - ((size_t)(A) & CHUNK_ALIGN_MASK)) & CHUNK_ALIGN_MASK)) |
|
1406 |
|
1407 /* -------------------------- MMAP preliminaries ------------------------- */ |
|
1408 |
|
1409 /* |
|
1410 If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and |
|
1411 checks to fail so compiler optimizer can delete code rather than |
|
1412 using so many "#if"s. |
|
1413 */ |
|
1414 |
|
1415 |
|
1416 /* MORECORE and MMAP must return MFAIL on failure */ |
|
1417 #define MFAIL ((void*)(MAX_SIZE_T)) |
|
1418 #define CMFAIL ((char*)(MFAIL)) /* defined for convenience */ |
|
1419 |
|
1420 #if !HAVE_MMAP |
|
1421 #define IS_MMAPPED_BIT (SIZE_T_ZERO) |
|
1422 #define USE_MMAP_BIT (SIZE_T_ZERO) |
|
1423 #define CALL_MMAP(s) MFAIL |
|
1424 #define CALL_MUNMAP(a, s) (-1) |
|
1425 #define DIRECT_MMAP(s) MFAIL |
|
1426 |
|
1427 #else /* HAVE_MMAP */ |
|
1428 #define IS_MMAPPED_BIT (SIZE_T_ONE) |
|
1429 #define USE_MMAP_BIT (SIZE_T_ONE) |
|
1430 |
|
1431 #ifndef WIN32 |
|
1432 #define CALL_MUNMAP(a, s) munmap((a), (s)) |
|
1433 #define MMAP_PROT (PROT_READ|PROT_WRITE) |
|
1434 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON) |
|
1435 #define MAP_ANONYMOUS MAP_ANON |
|
1436 #endif /* MAP_ANON */ |
|
1437 #ifdef MAP_ANONYMOUS |
|
1438 #define MMAP_FLAGS (MAP_PRIVATE|MAP_ANONYMOUS) |
|
1439 #define CALL_MMAP(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0) |
|
1440 #else /* MAP_ANONYMOUS */ |
|
1441 /* |
|
1442 Nearly all versions of mmap support MAP_ANONYMOUS, so the following |
|
1443 is unlikely to be needed, but is supplied just in case. |
|
1444 */ |
|
1445 #define MMAP_FLAGS (MAP_PRIVATE) |
|
1446 static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */ |
|
1447 #define CALL_MMAP(s) ((dev_zero_fd < 0) ? \ |
|
1448 (dev_zero_fd = open("/dev/zero", O_RDWR), \ |
|
1449 mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \ |
|
1450 mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) |
|
1451 #endif /* MAP_ANONYMOUS */ |
|
1452 |
|
1453 #define DIRECT_MMAP(s) CALL_MMAP(s) |
|
1454 #else /* WIN32 */ |
|
1455 |
|
1456 /* Win32 MMAP via VirtualAlloc */ |
|
1457 static FORCEINLINE void* win32mmap(size_t size) { |
|
1458 void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE); |
|
1459 return (ptr != 0)? ptr: MFAIL; |
|
1460 } |
|
1461 |
|
1462 /* For direct MMAP, use MEM_TOP_DOWN to minimize interference */ |
|
1463 static FORCEINLINE void* win32direct_mmap(size_t size) { |
|
1464 void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN, |
|
1465 PAGE_READWRITE); |
|
1466 return (ptr != 0)? ptr: MFAIL; |
|
1467 } |
|
1468 |
|
1469 /* This function supports releasing coalesed segments */ |
|
1470 static FORCEINLINE int win32munmap(void* ptr, size_t size) { |
|
1471 MEMORY_BASIC_INFORMATION minfo; |
|
1472 char* cptr = (char*)ptr; |
|
1473 while (size) { |
|
1474 if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0) |
|
1475 return -1; |
|
1476 if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr || |
|
1477 minfo.State != MEM_COMMIT || minfo.RegionSize > size) |
|
1478 return -1; |
|
1479 if (VirtualFree(cptr, 0, MEM_RELEASE) == 0) |
|
1480 return -1; |
|
1481 cptr += minfo.RegionSize; |
|
1482 size -= minfo.RegionSize; |
|
1483 } |
|
1484 return 0; |
|
1485 } |
|
1486 |
|
1487 #define CALL_MMAP(s) win32mmap(s) |
|
1488 #define CALL_MUNMAP(a, s) win32munmap((a), (s)) |
|
1489 #define DIRECT_MMAP(s) win32direct_mmap(s) |
|
1490 #endif /* WIN32 */ |
|
1491 #endif /* HAVE_MMAP */ |
|
1492 |
|
1493 #if HAVE_MMAP && HAVE_MREMAP |
|
1494 #define CALL_MREMAP(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv)) |
|
1495 #else /* HAVE_MMAP && HAVE_MREMAP */ |
|
1496 #define CALL_MREMAP(addr, osz, nsz, mv) ((void)(addr),(void)(osz), \ |
|
1497 (void)(nsz), (void)(mv),MFAIL) |
|
1498 #endif /* HAVE_MMAP && HAVE_MREMAP */ |
|
1499 |
|
1500 #if HAVE_MORECORE |
|
1501 #define CALL_MORECORE(S) MORECORE(S) |
|
1502 #else /* HAVE_MORECORE */ |
|
1503 #define CALL_MORECORE(S) MFAIL |
|
1504 #endif /* HAVE_MORECORE */ |
|
1505 |
|
1506 /* mstate bit set if continguous morecore disabled or failed */ |
|
1507 #define USE_NONCONTIGUOUS_BIT (4U) |
|
1508 |
|
1509 /* segment bit set in create_mspace_with_base */ |
|
1510 #define EXTERN_BIT (8U) |
|
1511 |
|
1512 |
|
1513 /* --------------------------- Lock preliminaries ------------------------ */ |
|
1514 |
|
1515 /* |
|
1516 When locks are defined, there are up to two global locks: |
|
1517 |
|
1518 * If HAVE_MORECORE, morecore_mutex protects sequences of calls to |
|
1519 MORECORE. In many cases sys_alloc requires two calls, that should |
|
1520 not be interleaved with calls by other threads. This does not |
|
1521 protect against direct calls to MORECORE by other threads not |
|
1522 using this lock, so there is still code to cope the best we can on |
|
1523 interference. |
|
1524 |
|
1525 * magic_init_mutex ensures that mparams.magic and other |
|
1526 unique mparams values are initialized only once. |
|
1527 |
|
1528 To enable use in layered extensions, locks are reentrant. |
|
1529 |
|
1530 Because lock-protected regions generally have bounded times, we use |
|
1531 the supplied simple spinlocks in the custom versions for x86. |
|
1532 |
|
1533 If USE_LOCKS is > 1, the definitions of lock routines here are |
|
1534 bypassed, in which case you will need to define at least |
|
1535 INITIAL_LOCK, ACQUIRE_LOCK, RELEASE_LOCK, and |
|
1536 NULL_LOCK_INITIALIZER, and possibly TRY_LOCK and IS_LOCKED |
|
1537 (The latter two are not used in this malloc, but are |
|
1538 commonly needed in extensions.) |
|
1539 */ |
|
1540 |
|
1541 #if USE_LOCKS == 1 |
|
1542 |
|
1543 #if USE_SPIN_LOCKS |
|
1544 #ifndef WIN32 |
|
1545 /* Custom pthread-style spin locks on x86 and x64 for gcc */ |
|
1546 struct pthread_mlock_t |
|
1547 { |
|
1548 volatile pthread_t threadid; |
|
1549 volatile unsigned int c; |
|
1550 volatile unsigned int l; |
|
1551 }; |
|
1552 #define MLOCK_T struct pthread_mlock_t |
|
1553 #define CURRENT_THREAD pthread_self() |
|
1554 #define SPINS_PER_YIELD 63 |
|
1555 static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) { |
|
1556 if(CURRENT_THREAD==sl->threadid) |
|
1557 ++sl->c; |
|
1558 else { |
|
1559 int spins = 0; |
|
1560 for (;;) { |
|
1561 int ret; |
|
1562 __asm__ __volatile__ ("lock cmpxchgl %2,(%1)" : "=a" (ret) : "r" (&sl->l), "r" (1), "a" (0)); |
|
1563 if(!ret) { |
|
1564 assert(!sl->threadid); |
|
1565 sl->threadid=CURRENT_THREAD; |
|
1566 sl->c=1; |
|
1567 break; |
|
1568 } |
|
1569 if ((++spins & SPINS_PER_YIELD) == 0) { |
|
1570 #if defined (__SVR4) && defined (__sun) /* solaris */ |
|
1571 thr_yield(); |
|
1572 #else |
|
1573 #ifdef linux |
|
1574 sched_yield(); |
|
1575 #else /* no-op yield on unknown systems */ |
|
1576 ; |
|
1577 #endif /* linux */ |
|
1578 #endif /* solaris */ |
|
1579 } |
|
1580 } |
|
1581 } |
|
1582 |
|
1583 return 0; |
|
1584 } |
|
1585 |
|
1586 static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) { |
|
1587 int ret; |
|
1588 assert(CURRENT_THREAD==sl->threadid); |
|
1589 if (!--sl->c) { |
|
1590 sl->threadid=0; |
|
1591 __asm__ __volatile__ ("xchgl %2,(%1)" : "=r" (ret) : "r" (&sl->l), "0" (0)); |
|
1592 } |
|
1593 } |
|
1594 |
|
1595 static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) { |
|
1596 int ret; |
|
1597 __asm__ __volatile__ ("lock cmpxchgl %2,(%1)" : "=a" (ret) : "r" (&sl->l), "r" (1), "a" (0)); |
|
1598 if(!ret){ |
|
1599 assert(!sl->threadid); |
|
1600 sl->threadid=CURRENT_THREAD; |
|
1601 sl->c=1; |
|
1602 return 1; |
|
1603 } |
|
1604 return 0; |
|
1605 } |
|
1606 |
|
1607 #define INITIAL_LOCK(sl) (memset((sl), 0, sizeof(MLOCK_T)), 0) |
|
1608 #define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl) |
|
1609 #define RELEASE_LOCK(sl) pthread_release_lock(sl) |
|
1610 #define TRY_LOCK(sl) pthread_try_lock(sl) |
|
1611 #define IS_LOCKED(sl) ((sl)->l) |
|
1612 |
|
1613 static MLOCK_T magic_init_mutex = {0, 0, 0 }; |
|
1614 #if HAVE_MORECORE |
|
1615 static MLOCK_T morecore_mutex = {0, 0, 0 }; |
|
1616 #endif /* HAVE_MORECORE */ |
|
1617 |
|
1618 #else /* WIN32 */ |
|
1619 /* Custom win32-style spin locks on x86 and x64 for MSC */ |
|
1620 struct win32_mlock_t |
|
1621 { |
|
1622 volatile long threadid; |
|
1623 volatile unsigned int c; |
|
1624 long l; |
|
1625 }; |
|
1626 #define MLOCK_T struct win32_mlock_t |
|
1627 #define CURRENT_THREAD GetCurrentThreadId() |
|
1628 #define SPINS_PER_YIELD 63 |
|
1629 static FORCEINLINE int win32_acquire_lock (MLOCK_T *sl) { |
|
1630 long mythreadid=CURRENT_THREAD; |
|
1631 if(mythreadid==sl->threadid) |
|
1632 ++sl->c; |
|
1633 else { |
|
1634 int spins = 0; |
|
1635 for (;;) { |
|
1636 if (!interlockedexchange(&sl->l, 1)) { |
|
1637 assert(!sl->threadid); |
|
1638 sl->threadid=mythreadid; |
|
1639 sl->c=1; |
|
1640 break; |
|
1641 } |
|
1642 if ((++spins & SPINS_PER_YIELD) == 0) |
|
1643 SleepEx(0, FALSE); |
|
1644 } |
|
1645 } |
|
1646 return 0; |
|
1647 } |
|
1648 |
|
1649 static FORCEINLINE void win32_release_lock (MLOCK_T *sl) { |
|
1650 assert(CURRENT_THREAD==sl->threadid); |
|
1651 if (!--sl->c) { |
|
1652 sl->threadid=0; |
|
1653 interlockedexchange (&sl->l, 0); |
|
1654 } |
|
1655 } |
|
1656 |
|
1657 static FORCEINLINE int win32_try_lock (MLOCK_T *sl) { |
|
1658 if (!interlockedexchange(&sl->l, 1)){ |
|
1659 assert(!sl->threadid); |
|
1660 sl->threadid=CURRENT_THREAD; |
|
1661 sl->c=1; |
|
1662 return 1; |
|
1663 } |
|
1664 return 0; |
|
1665 } |
|
1666 |
|
1667 #define INITIAL_LOCK(sl) (memset(sl, 0, sizeof(MLOCK_T)), 0) |
|
1668 #define ACQUIRE_LOCK(sl) win32_acquire_lock(sl) |
|
1669 #define RELEASE_LOCK(sl) win32_release_lock(sl) |
|
1670 #define TRY_LOCK(sl) win32_try_lock(sl) |
|
1671 #define IS_LOCKED(sl) ((sl)->l) |
|
1672 |
|
1673 static MLOCK_T magic_init_mutex = {0, 0 }; |
|
1674 #if HAVE_MORECORE |
|
1675 static MLOCK_T morecore_mutex = {0, 0 }; |
|
1676 #endif /* HAVE_MORECORE */ |
|
1677 |
|
1678 #endif /* WIN32 */ |
|
1679 #else /* USE_SPIN_LOCKS */ |
|
1680 |
|
1681 #ifndef WIN32 |
|
1682 /* pthreads-based locks */ |
|
1683 struct pthread_mlock_t |
|
1684 { |
|
1685 volatile unsigned int c; |
|
1686 pthread_mutex_t l; |
|
1687 }; |
|
1688 #define MLOCK_T struct pthread_mlock_t |
|
1689 #define CURRENT_THREAD pthread_self() |
|
1690 static FORCEINLINE int pthread_acquire_lock (MLOCK_T *sl) { |
|
1691 if(!pthread_mutex_lock(&(sl)->l)){ |
|
1692 sl->c++; |
|
1693 return 0; |
|
1694 } |
|
1695 return 1; |
|
1696 } |
|
1697 |
|
1698 static FORCEINLINE void pthread_release_lock (MLOCK_T *sl) { |
|
1699 --sl->c; |
|
1700 pthread_mutex_unlock(&(sl)->l); |
|
1701 } |
|
1702 |
|
1703 static FORCEINLINE int pthread_try_lock (MLOCK_T *sl) { |
|
1704 if(!pthread_mutex_trylock(&(sl)->l)){ |
|
1705 sl->c++; |
|
1706 return 1; |
|
1707 } |
|
1708 return 0; |
|
1709 } |
|
1710 |
|
1711 static FORCEINLINE int pthread_init_lock (MLOCK_T *sl) { |
|
1712 pthread_mutexattr_t attr; |
|
1713 sl->c=0; |
|
1714 if(pthread_mutexattr_init(&attr)) return 1; |
|
1715 if(pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE)) return 1; |
|
1716 if(pthread_mutex_init(&sl->l, &attr)) return 1; |
|
1717 pthread_mutexattr_destroy(&attr); |
|
1718 return 0; |
|
1719 } |
|
1720 |
|
1721 static FORCEINLINE int pthread_islocked (MLOCK_T *sl) { |
|
1722 if(!pthread_try_lock(sl)){ |
|
1723 int ret = (sl->c != 0); |
|
1724 pthread_mutex_unlock(sl); |
|
1725 return ret; |
|
1726 } |
|
1727 return 0; |
|
1728 } |
|
1729 |
|
1730 #define INITIAL_LOCK(sl) pthread_init_lock(sl) |
|
1731 #define ACQUIRE_LOCK(sl) pthread_acquire_lock(sl) |
|
1732 #define RELEASE_LOCK(sl) pthread_release_lock(sl) |
|
1733 #define TRY_LOCK(sl) pthread_try_lock(sl) |
|
1734 #define IS_LOCKED(sl) pthread_islocked(sl) |
|
1735 |
|
1736 static MLOCK_T magic_init_mutex = {0, PTHREAD_MUTEX_INITIALIZER }; |
|
1737 #if HAVE_MORECORE |
|
1738 static MLOCK_T morecore_mutex = {0, PTHREAD_MUTEX_INITIALIZER }; |
|
1739 #endif /* HAVE_MORECORE */ |
|
1740 |
|
1741 #else /* WIN32 */ |
|
1742 /* Win32 critical sections */ |
|
1743 #define MLOCK_T CRITICAL_SECTION |
|
1744 #define CURRENT_THREAD GetCurrentThreadId() |
|
1745 #define INITIAL_LOCK(s) (!InitializeCriticalSectionAndSpinCount((s), 4000) |
|
1746 #define ACQUIRE_LOCK(s) ( (!((s))->DebugInfo ? INITIAL_LOCK((s)) : 0), !EnterCriticalSection((s)), 0) |
|
1747 #define RELEASE_LOCK(s) ( LeaveCriticalSection((s)), 0 ) |
|
1748 #define TRY_LOCK(s) ( TryEnterCriticalSection((s)) ) |
|
1749 #define IS_LOCKED(s) ( (s)->LockCount >= 0 ) |
|
1750 #define NULL_LOCK_INITIALIZER |
|
1751 static MLOCK_T magic_init_mutex; |
|
1752 #if HAVE_MORECORE |
|
1753 static MLOCK_T morecore_mutex; |
|
1754 #endif /* HAVE_MORECORE */ |
|
1755 #endif /* WIN32 */ |
|
1756 #endif /* USE_SPIN_LOCKS */ |
|
1757 #endif /* USE_LOCKS == 1 */ |
|
1758 |
|
1759 /* ----------------------- User-defined locks ------------------------ */ |
|
1760 |
|
1761 #if USE_LOCKS > 1 |
|
1762 /* Define your own lock implementation here */ |
|
1763 /* #define INITIAL_LOCK(sl) ... */ |
|
1764 /* #define ACQUIRE_LOCK(sl) ... */ |
|
1765 /* #define RELEASE_LOCK(sl) ... */ |
|
1766 /* #define TRY_LOCK(sl) ... */ |
|
1767 /* #define IS_LOCKED(sl) ... */ |
|
1768 /* #define NULL_LOCK_INITIALIZER ... */ |
|
1769 |
|
1770 static MLOCK_T magic_init_mutex = NULL_LOCK_INITIALIZER; |
|
1771 #if HAVE_MORECORE |
|
1772 static MLOCK_T morecore_mutex = NULL_LOCK_INITIALIZER; |
|
1773 #endif /* HAVE_MORECORE */ |
|
1774 #endif /* USE_LOCKS > 1 */ |
|
1775 |
|
1776 /* ----------------------- Lock-based state ------------------------ */ |
|
1777 |
|
1778 |
|
1779 #if USE_LOCKS |
|
1780 #define USE_LOCK_BIT (2U) |
|
1781 #else /* USE_LOCKS */ |
|
1782 #define USE_LOCK_BIT (0U) |
|
1783 #define INITIAL_LOCK(l) |
|
1784 #endif /* USE_LOCKS */ |
|
1785 |
|
1786 #if USE_LOCKS && HAVE_MORECORE |
|
1787 #define ACQUIRE_MORECORE_LOCK() ACQUIRE_LOCK(&morecore_mutex); |
|
1788 #define RELEASE_MORECORE_LOCK() RELEASE_LOCK(&morecore_mutex); |
|
1789 #else /* USE_LOCKS && HAVE_MORECORE */ |
|
1790 #define ACQUIRE_MORECORE_LOCK() |
|
1791 #define RELEASE_MORECORE_LOCK() |
|
1792 #endif /* USE_LOCKS && HAVE_MORECORE */ |
|
1793 |
|
1794 #if USE_LOCKS |
|
1795 #define ACQUIRE_MAGIC_INIT_LOCK() ACQUIRE_LOCK(&magic_init_mutex); |
|
1796 #define RELEASE_MAGIC_INIT_LOCK() RELEASE_LOCK(&magic_init_mutex); |
|
1797 #else /* USE_LOCKS */ |
|
1798 #define ACQUIRE_MAGIC_INIT_LOCK() |
|
1799 #define RELEASE_MAGIC_INIT_LOCK() |
|
1800 #endif /* USE_LOCKS */ |
|
1801 |
|
1802 |
|
1803 /* ----------------------- Chunk representations ------------------------ */ |
|
1804 |
|
1805 /* |
|
1806 (The following includes lightly edited explanations by Colin Plumb.) |
|
1807 |
|
1808 The malloc_chunk declaration below is misleading (but accurate and |
|
1809 necessary). It declares a "view" into memory allowing access to |
|
1810 necessary fields at known offsets from a given base. |
|
1811 |
|
1812 Chunks of memory are maintained using a `boundary tag' method as |
|
1813 originally described by Knuth. (See the paper by Paul Wilson |
|
1814 ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a survey of such |
|
1815 techniques.) Sizes of free chunks are stored both in the front of |
|
1816 each chunk and at the end. This makes consolidating fragmented |
|
1817 chunks into bigger chunks fast. The head fields also hold bits |
|
1818 representing whether chunks are free or in use. |
|
1819 |
|
1820 Here are some pictures to make it clearer. They are "exploded" to |
|
1821 show that the state of a chunk can be thought of as extending from |
|
1822 the high 31 bits of the head field of its header through the |
|
1823 prev_foot and PINUSE_BIT bit of the following chunk header. |
|
1824 |
|
1825 A chunk that's in use looks like: |
|
1826 |
|
1827 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1828 | Size of previous chunk (if P = 1) | |
|
1829 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1830 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P| |
|
1831 | Size of this chunk 1| +-+ |
|
1832 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1833 | | |
|
1834 +- -+ |
|
1835 | | |
|
1836 +- -+ |
|
1837 | : |
|
1838 +- size - sizeof(size_t) available payload bytes -+ |
|
1839 : | |
|
1840 chunk-> +- -+ |
|
1841 | | |
|
1842 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1843 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1| |
|
1844 | Size of next chunk (may or may not be in use) | +-+ |
|
1845 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1846 |
|
1847 And if it's free, it looks like this: |
|
1848 |
|
1849 chunk-> +- -+ |
|
1850 | User payload (must be in use, or we would have merged!) | |
|
1851 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1852 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P| |
|
1853 | Size of this chunk 0| +-+ |
|
1854 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1855 | Next pointer | |
|
1856 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1857 | Prev pointer | |
|
1858 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1859 | : |
|
1860 +- size - sizeof(struct chunk) unused bytes -+ |
|
1861 : | |
|
1862 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1863 | Size of this chunk | |
|
1864 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1865 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0| |
|
1866 | Size of next chunk (must be in use, or we would have merged)| +-+ |
|
1867 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1868 | : |
|
1869 +- User payload -+ |
|
1870 : | |
|
1871 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
1872 |0| |
|
1873 +-+ |
|
1874 Note that since we always merge adjacent free chunks, the chunks |
|
1875 adjacent to a free chunk must be in use. |
|
1876 |
|
1877 Given a pointer to a chunk (which can be derived trivially from the |
|
1878 payload pointer) we can, in O(1) time, find out whether the adjacent |
|
1879 chunks are free, and if so, unlink them from the lists that they |
|
1880 are on and merge them with the current chunk. |
|
1881 |
|
1882 Chunks always begin on even word boundaries, so the mem portion |
|
1883 (which is returned to the user) is also on an even word boundary, and |
|
1884 thus at least double-word aligned. |
|
1885 |
|
1886 The P (PINUSE_BIT) bit, stored in the unused low-order bit of the |
|
1887 chunk size (which is always a multiple of two words), is an in-use |
|
1888 bit for the *previous* chunk. If that bit is *clear*, then the |
|
1889 word before the current chunk size contains the previous chunk |
|
1890 size, and can be used to find the front of the previous chunk. |
|
1891 The very first chunk allocated always has this bit set, preventing |
|
1892 access to non-existent (or non-owned) memory. If pinuse is set for |
|
1893 any given chunk, then you CANNOT determine the size of the |
|
1894 previous chunk, and might even get a memory addressing fault when |
|
1895 trying to do so. |
|
1896 |
|
1897 The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of |
|
1898 the chunk size redundantly records whether the current chunk is |
|
1899 inuse. This redundancy enables usage checks within free and realloc, |
|
1900 and reduces indirection when freeing and consolidating chunks. |
|
1901 |
|
1902 Each freshly allocated chunk must have both cinuse and pinuse set. |
|
1903 That is, each allocated chunk borders either a previously allocated |
|
1904 and still in-use chunk, or the base of its memory arena. This is |
|
1905 ensured by making all allocations from the the `lowest' part of any |
|
1906 found chunk. Further, no free chunk physically borders another one, |
|
1907 so each free chunk is known to be preceded and followed by either |
|
1908 inuse chunks or the ends of memory. |
|
1909 |
|
1910 Note that the `foot' of the current chunk is actually represented |
|
1911 as the prev_foot of the NEXT chunk. This makes it easier to |
|
1912 deal with alignments etc but can be very confusing when trying |
|
1913 to extend or adapt this code. |
|
1914 |
|
1915 The exceptions to all this are |
|
1916 |
|
1917 1. The special chunk `top' is the top-most available chunk (i.e., |
|
1918 the one bordering the end of available memory). It is treated |
|
1919 specially. Top is never included in any bin, is used only if |
|
1920 no other chunk is available, and is released back to the |
|
1921 system if it is very large (see M_TRIM_THRESHOLD). In effect, |
|
1922 the top chunk is treated as larger (and thus less well |
|
1923 fitting) than any other available chunk. The top chunk |
|
1924 doesn't update its trailing size field since there is no next |
|
1925 contiguous chunk that would have to index off it. However, |
|
1926 space is still allocated for it (TOP_FOOT_SIZE) to enable |
|
1927 separation or merging when space is extended. |
|
1928 |
|
1929 3. Chunks allocated via mmap, which have the lowest-order bit |
|
1930 (IS_MMAPPED_BIT) set in their prev_foot fields, and do not set |
|
1931 PINUSE_BIT in their head fields. Because they are allocated |
|
1932 one-by-one, each must carry its own prev_foot field, which is |
|
1933 also used to hold the offset this chunk has within its mmapped |
|
1934 region, which is needed to preserve alignment. Each mmapped |
|
1935 chunk is trailed by the first two fields of a fake next-chunk |
|
1936 for sake of usage checks. |
|
1937 |
|
1938 */ |
|
1939 |
|
1940 struct malloc_chunk { |
|
1941 size_t prev_foot; /* Size of previous chunk (if free). */ |
|
1942 size_t head; /* Size and inuse bits. */ |
|
1943 struct malloc_chunk* fd; /* double links -- used only if free. */ |
|
1944 struct malloc_chunk* bk; |
|
1945 }; |
|
1946 |
|
1947 typedef struct malloc_chunk mchunk; |
|
1948 typedef struct malloc_chunk* mchunkptr; |
|
1949 typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */ |
|
1950 typedef unsigned int bindex_t; /* Described below */ |
|
1951 typedef unsigned int binmap_t; /* Described below */ |
|
1952 typedef unsigned int flag_t; /* The type of various bit flag sets */ |
|
1953 |
|
1954 /* ------------------- Chunks sizes and alignments ----------------------- */ |
|
1955 |
|
1956 #define MCHUNK_SIZE (sizeof(mchunk)) |
|
1957 |
|
1958 #if FOOTERS |
|
1959 #define CHUNK_OVERHEAD (TWO_SIZE_T_SIZES) |
|
1960 #else /* FOOTERS */ |
|
1961 #define CHUNK_OVERHEAD (SIZE_T_SIZE) |
|
1962 #endif /* FOOTERS */ |
|
1963 |
|
1964 /* MMapped chunks need a second word of overhead ... */ |
|
1965 #define MMAP_CHUNK_OVERHEAD (TWO_SIZE_T_SIZES) |
|
1966 /* ... and additional padding for fake next-chunk at foot */ |
|
1967 #define MMAP_FOOT_PAD (FOUR_SIZE_T_SIZES) |
|
1968 |
|
1969 /* The smallest size we can malloc is an aligned minimal chunk */ |
|
1970 #define MIN_CHUNK_SIZE\ |
|
1971 ((MCHUNK_SIZE + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK) |
|
1972 |
|
1973 /* conversion from malloc headers to user pointers, and back */ |
|
1974 #define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES)) |
|
1975 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES)) |
|
1976 /* chunk associated with aligned address A */ |
|
1977 #define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A))) |
|
1978 |
|
1979 /* Bounds on request (not chunk) sizes. */ |
|
1980 #define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2) |
|
1981 #define MIN_REQUEST (MIN_CHUNK_SIZE - CHUNK_OVERHEAD - SIZE_T_ONE) |
|
1982 |
|
1983 /* pad request bytes into a usable size */ |
|
1984 #define pad_request(req) \ |
|
1985 (((req) + CHUNK_OVERHEAD + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK) |
|
1986 |
|
1987 /* pad request, checking for minimum (but not maximum) */ |
|
1988 #define request2size(req) \ |
|
1989 (((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req)) |
|
1990 |
|
1991 |
|
1992 /* ------------------ Operations on head and foot fields ----------------- */ |
|
1993 |
|
1994 /* |
|
1995 The head field of a chunk is or'ed with PINUSE_BIT when previous |
|
1996 adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in |
|
1997 use. If the chunk was obtained with mmap, the prev_foot field has |
|
1998 IS_MMAPPED_BIT set, otherwise holding the offset of the base of the |
|
1999 mmapped region to the base of the chunk. |
|
2000 |
|
2001 FLAG4_BIT is not used by this malloc, but might be useful in extensions. |
|
2002 */ |
|
2003 |
|
2004 #define PINUSE_BIT (SIZE_T_ONE) |
|
2005 #define CINUSE_BIT (SIZE_T_TWO) |
|
2006 #define FLAG4_BIT (SIZE_T_FOUR) |
|
2007 #define INUSE_BITS (PINUSE_BIT|CINUSE_BIT) |
|
2008 #define FLAG_BITS (PINUSE_BIT|CINUSE_BIT|FLAG4_BIT) |
|
2009 |
|
2010 /* Head value for fenceposts */ |
|
2011 #define FENCEPOST_HEAD (INUSE_BITS|SIZE_T_SIZE) |
|
2012 |
|
2013 /* extraction of fields from head words */ |
|
2014 #define cinuse(p) ((p)->head & CINUSE_BIT) |
|
2015 #define pinuse(p) ((p)->head & PINUSE_BIT) |
|
2016 #define chunksize(p) ((p)->head & ~(FLAG_BITS)) |
|
2017 |
|
2018 #define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT) |
|
2019 #define clear_cinuse(p) ((p)->head &= ~CINUSE_BIT) |
|
2020 |
|
2021 /* Treat space at ptr +/- offset as a chunk */ |
|
2022 #define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s))) |
|
2023 #define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s))) |
|
2024 |
|
2025 /* Ptr to next or previous physical malloc_chunk. */ |
|
2026 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~FLAG_BITS))) |
|
2027 #define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) )) |
|
2028 |
|
2029 /* extract next chunk's pinuse bit */ |
|
2030 #define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT) |
|
2031 |
|
2032 /* Get/set size at footer */ |
|
2033 #define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot) |
|
2034 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s)) |
|
2035 |
|
2036 /* Set size, pinuse bit, and foot */ |
|
2037 #define set_size_and_pinuse_of_free_chunk(p, s)\ |
|
2038 ((p)->head = (s|PINUSE_BIT), set_foot(p, s)) |
|
2039 |
|
2040 /* Set size, pinuse bit, foot, and clear next pinuse */ |
|
2041 #define set_free_with_pinuse(p, s, n)\ |
|
2042 (clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s)) |
|
2043 |
|
2044 #define is_mmapped(p)\ |
|
2045 (!((p)->head & PINUSE_BIT) && ((p)->prev_foot & IS_MMAPPED_BIT)) |
|
2046 |
|
2047 /* Get the internal overhead associated with chunk p */ |
|
2048 #define overhead_for(p)\ |
|
2049 (is_mmapped(p)? MMAP_CHUNK_OVERHEAD : CHUNK_OVERHEAD) |
|
2050 |
|
2051 /* Return true if malloced space is not necessarily cleared */ |
|
2052 #if MMAP_CLEARS |
|
2053 #define calloc_must_clear(p) (!is_mmapped(p)) |
|
2054 #else /* MMAP_CLEARS */ |
|
2055 #define calloc_must_clear(p) (1) |
|
2056 #endif /* MMAP_CLEARS */ |
|
2057 |
|
2058 /* ---------------------- Overlaid data structures ----------------------- */ |
|
2059 |
|
2060 /* |
|
2061 When chunks are not in use, they are treated as nodes of either |
|
2062 lists or trees. |
|
2063 |
|
2064 "Small" chunks are stored in circular doubly-linked lists, and look |
|
2065 like this: |
|
2066 |
|
2067 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2068 | Size of previous chunk | |
|
2069 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2070 `head:' | Size of chunk, in bytes |P| |
|
2071 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2072 | Forward pointer to next chunk in list | |
|
2073 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2074 | Back pointer to previous chunk in list | |
|
2075 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2076 | Unused space (may be 0 bytes long) . |
|
2077 . . |
|
2078 . | |
|
2079 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2080 `foot:' | Size of chunk, in bytes | |
|
2081 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2082 |
|
2083 Larger chunks are kept in a form of bitwise digital trees (aka |
|
2084 tries) keyed on chunksizes. Because malloc_tree_chunks are only for |
|
2085 free chunks greater than 256 bytes, their size doesn't impose any |
|
2086 constraints on user chunk sizes. Each node looks like: |
|
2087 |
|
2088 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2089 | Size of previous chunk | |
|
2090 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2091 `head:' | Size of chunk, in bytes |P| |
|
2092 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2093 | Forward pointer to next chunk of same size | |
|
2094 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2095 | Back pointer to previous chunk of same size | |
|
2096 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2097 | Pointer to left child (child[0]) | |
|
2098 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2099 | Pointer to right child (child[1]) | |
|
2100 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2101 | Pointer to parent | |
|
2102 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2103 | bin index of this chunk | |
|
2104 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2105 | Unused space . |
|
2106 . | |
|
2107 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2108 `foot:' | Size of chunk, in bytes | |
|
2109 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
|
2110 |
|
2111 Each tree holding treenodes is a tree of unique chunk sizes. Chunks |
|
2112 of the same size are arranged in a circularly-linked list, with only |
|
2113 the oldest chunk (the next to be used, in our FIFO ordering) |
|
2114 actually in the tree. (Tree members are distinguished by a non-null |
|
2115 parent pointer.) If a chunk with the same size an an existing node |
|
2116 is inserted, it is linked off the existing node using pointers that |
|
2117 work in the same way as fd/bk pointers of small chunks. |
|
2118 |
|
2119 Each tree contains a power of 2 sized range of chunk sizes (the |
|
2120 smallest is 0x100 <= x < 0x180), which is is divided in half at each |
|
2121 tree level, with the chunks in the smaller half of the range (0x100 |
|
2122 <= x < 0x140 for the top nose) in the left subtree and the larger |
|
2123 half (0x140 <= x < 0x180) in the right subtree. This is, of course, |
|
2124 done by inspecting individual bits. |
|
2125 |
|
2126 Using these rules, each node's left subtree contains all smaller |
|
2127 sizes than its right subtree. However, the node at the root of each |
|
2128 subtree has no particular ordering relationship to either. (The |
|
2129 dividing line between the subtree sizes is based on trie relation.) |
|
2130 If we remove the last chunk of a given size from the interior of the |
|
2131 tree, we need to replace it with a leaf node. The tree ordering |
|
2132 rules permit a node to be replaced by any leaf below it. |
|
2133 |
|
2134 The smallest chunk in a tree (a common operation in a best-fit |
|
2135 allocator) can be found by walking a path to the leftmost leaf in |
|
2136 the tree. Unlike a usual binary tree, where we follow left child |
|
2137 pointers until we reach a null, here we follow the right child |
|
2138 pointer any time the left one is null, until we reach a leaf with |
|
2139 both child pointers null. The smallest chunk in the tree will be |
|
2140 somewhere along that path. |
|
2141 |
|
2142 The worst case number of steps to add, find, or remove a node is |
|
2143 bounded by the number of bits differentiating chunks within |
|
2144 bins. Under current bin calculations, this ranges from 6 up to 21 |
|
2145 (for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case |
|
2146 is of course much better. |
|
2147 */ |
|
2148 |
|
2149 struct malloc_tree_chunk { |
|
2150 /* The first four fields must be compatible with malloc_chunk */ |
|
2151 size_t prev_foot; |
|
2152 size_t head; |
|
2153 struct malloc_tree_chunk* fd; |
|
2154 struct malloc_tree_chunk* bk; |
|
2155 |
|
2156 struct malloc_tree_chunk* child[2]; |
|
2157 struct malloc_tree_chunk* parent; |
|
2158 bindex_t index; |
|
2159 }; |
|
2160 |
|
2161 typedef struct malloc_tree_chunk tchunk; |
|
2162 typedef struct malloc_tree_chunk* tchunkptr; |
|
2163 typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */ |
|
2164 |
|
2165 /* A little helper macro for trees */ |
|
2166 #define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1]) |
|
2167 |
|
2168 /* ----------------------------- Segments -------------------------------- */ |
|
2169 |
|
2170 /* |
|
2171 Each malloc space may include non-contiguous segments, held in a |
|
2172 list headed by an embedded malloc_segment record representing the |
|
2173 top-most space. Segments also include flags holding properties of |
|
2174 the space. Large chunks that are directly allocated by mmap are not |
|
2175 included in this list. They are instead independently created and |
|
2176 destroyed without otherwise keeping track of them. |
|
2177 |
|
2178 Segment management mainly comes into play for spaces allocated by |
|
2179 MMAP. Any call to MMAP might or might not return memory that is |
|
2180 adjacent to an existing segment. MORECORE normally contiguously |
|
2181 extends the current space, so this space is almost always adjacent, |
|
2182 which is simpler and faster to deal with. (This is why MORECORE is |
|
2183 used preferentially to MMAP when both are available -- see |
|
2184 sys_alloc.) When allocating using MMAP, we don't use any of the |
|
2185 hinting mechanisms (inconsistently) supported in various |
|
2186 implementations of unix mmap, or distinguish reserving from |
|
2187 committing memory. Instead, we just ask for space, and exploit |
|
2188 contiguity when we get it. It is probably possible to do |
|
2189 better than this on some systems, but no general scheme seems |
|
2190 to be significantly better. |
|
2191 |
|
2192 Management entails a simpler variant of the consolidation scheme |
|
2193 used for chunks to reduce fragmentation -- new adjacent memory is |
|
2194 normally prepended or appended to an existing segment. However, |
|
2195 there are limitations compared to chunk consolidation that mostly |
|
2196 reflect the fact that segment processing is relatively infrequent |
|
2197 (occurring only when getting memory from system) and that we |
|
2198 don't expect to have huge numbers of segments: |
|
2199 |
|
2200 * Segments are not indexed, so traversal requires linear scans. (It |
|
2201 would be possible to index these, but is not worth the extra |
|
2202 overhead and complexity for most programs on most platforms.) |
|
2203 * New segments are only appended to old ones when holding top-most |
|
2204 memory; if they cannot be prepended to others, they are held in |
|
2205 different segments. |
|
2206 |
|
2207 Except for the top-most segment of an mstate, each segment record |
|
2208 is kept at the tail of its segment. Segments are added by pushing |
|
2209 segment records onto the list headed by &mstate.seg for the |
|
2210 containing mstate. |
|
2211 |
|
2212 Segment flags control allocation/merge/deallocation policies: |
|
2213 * If EXTERN_BIT set, then we did not allocate this segment, |
|
2214 and so should not try to deallocate or merge with others. |
|
2215 (This currently holds only for the initial segment passed |
|
2216 into create_mspace_with_base.) |
|
2217 * If IS_MMAPPED_BIT set, the segment may be merged with |
|
2218 other surrounding mmapped segments and trimmed/de-allocated |
|
2219 using munmap. |
|
2220 * If neither bit is set, then the segment was obtained using |
|
2221 MORECORE so can be merged with surrounding MORECORE'd segments |
|
2222 and deallocated/trimmed using MORECORE with negative arguments. |
|
2223 */ |
|
2224 |
|
2225 struct malloc_segment { |
|
2226 char* base; /* base address */ |
|
2227 size_t size; /* allocated size */ |
|
2228 struct malloc_segment* next; /* ptr to next segment */ |
|
2229 flag_t sflags; /* mmap and extern flag */ |
|
2230 }; |
|
2231 |
|
2232 #define is_mmapped_segment(S) ((S)->sflags & IS_MMAPPED_BIT) |
|
2233 #define is_extern_segment(S) ((S)->sflags & EXTERN_BIT) |
|
2234 |
|
2235 typedef struct malloc_segment msegment; |
|
2236 typedef struct malloc_segment* msegmentptr; |
|
2237 |
|
2238 /* ---------------------------- malloc_state ----------------------------- */ |
|
2239 |
|
2240 /* |
|
2241 A malloc_state holds all of the bookkeeping for a space. |
|
2242 The main fields are: |
|
2243 |
|
2244 Top |
|
2245 The topmost chunk of the currently active segment. Its size is |
|
2246 cached in topsize. The actual size of topmost space is |
|
2247 topsize+TOP_FOOT_SIZE, which includes space reserved for adding |
|
2248 fenceposts and segment records if necessary when getting more |
|
2249 space from the system. The size at which to autotrim top is |
|
2250 cached from mparams in trim_check, except that it is disabled if |
|
2251 an autotrim fails. |
|
2252 |
|
2253 Designated victim (dv) |
|
2254 This is the preferred chunk for servicing small requests that |
|
2255 don't have exact fits. It is normally the chunk split off most |
|
2256 recently to service another small request. Its size is cached in |
|
2257 dvsize. The link fields of this chunk are not maintained since it |
|
2258 is not kept in a bin. |
|
2259 |
|
2260 SmallBins |
|
2261 An array of bin headers for free chunks. These bins hold chunks |
|
2262 with sizes less than MIN_LARGE_SIZE bytes. Each bin contains |
|
2263 chunks of all the same size, spaced 8 bytes apart. To simplify |
|
2264 use in double-linked lists, each bin header acts as a malloc_chunk |
|
2265 pointing to the real first node, if it exists (else pointing to |
|
2266 itself). This avoids special-casing for headers. But to avoid |
|
2267 waste, we allocate only the fd/bk pointers of bins, and then use |
|
2268 repositioning tricks to treat these as the fields of a chunk. |
|
2269 |
|
2270 TreeBins |
|
2271 Treebins are pointers to the roots of trees holding a range of |
|
2272 sizes. There are 2 equally spaced treebins for each power of two |
|
2273 from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything |
|
2274 larger. |
|
2275 |
|
2276 Bin maps |
|
2277 There is one bit map for small bins ("smallmap") and one for |
|
2278 treebins ("treemap). Each bin sets its bit when non-empty, and |
|
2279 clears the bit when empty. Bit operations are then used to avoid |
|
2280 bin-by-bin searching -- nearly all "search" is done without ever |
|
2281 looking at bins that won't be selected. The bit maps |
|
2282 conservatively use 32 bits per map word, even if on 64bit system. |
|
2283 For a good description of some of the bit-based techniques used |
|
2284 here, see Henry S. Warren Jr's book "Hacker's Delight" (and |
|
2285 supplement at http://hackersdelight.org/). Many of these are |
|
2286 intended to reduce the branchiness of paths through malloc etc, as |
|
2287 well as to reduce the number of memory locations read or written. |
|
2288 |
|
2289 Segments |
|
2290 A list of segments headed by an embedded malloc_segment record |
|
2291 representing the initial space. |
|
2292 |
|
2293 Address check support |
|
2294 The least_addr field is the least address ever obtained from |
|
2295 MORECORE or MMAP. Attempted frees and reallocs of any address less |
|
2296 than this are trapped (unless INSECURE is defined). |
|
2297 |
|
2298 Magic tag |
|
2299 A cross-check field that should always hold same value as mparams.magic. |
|
2300 |
|
2301 Flags |
|
2302 Bits recording whether to use MMAP, locks, or contiguous MORECORE |
|
2303 |
|
2304 Statistics |
|
2305 Each space keeps track of current and maximum system memory |
|
2306 obtained via MORECORE or MMAP. |
|
2307 |
|
2308 Trim support |
|
2309 Fields holding the amount of unused topmost memory that should trigger |
|
2310 timming, and a counter to force periodic scanning to release unused |
|
2311 non-topmost segments. |
|
2312 |
|
2313 Locking |
|
2314 If USE_LOCKS is defined, the "mutex" lock is acquired and released |
|
2315 around every public call using this mspace. |
|
2316 |
|
2317 Extension support |
|
2318 A void* pointer and a size_t field that can be used to help implement |
|
2319 extensions to this malloc. |
|
2320 */ |
|
2321 |
|
2322 /* Bin types, widths and sizes */ |
|
2323 #define NSMALLBINS (32U) |
|
2324 #define NTREEBINS (32U) |
|
2325 #define SMALLBIN_SHIFT (3U) |
|
2326 #define SMALLBIN_WIDTH (SIZE_T_ONE << SMALLBIN_SHIFT) |
|
2327 #define TREEBIN_SHIFT (8U) |
|
2328 #define MIN_LARGE_SIZE (SIZE_T_ONE << TREEBIN_SHIFT) |
|
2329 #define MAX_SMALL_SIZE (MIN_LARGE_SIZE - SIZE_T_ONE) |
|
2330 #define MAX_SMALL_REQUEST (MAX_SMALL_SIZE - CHUNK_ALIGN_MASK - CHUNK_OVERHEAD) |
|
2331 |
|
2332 struct malloc_state { |
|
2333 binmap_t smallmap; |
|
2334 binmap_t treemap; |
|
2335 size_t dvsize; |
|
2336 size_t topsize; |
|
2337 char* least_addr; |
|
2338 mchunkptr dv; |
|
2339 mchunkptr top; |
|
2340 size_t trim_check; |
|
2341 size_t release_checks; |
|
2342 size_t magic; |
|
2343 mchunkptr smallbins[(NSMALLBINS+1)*2]; |
|
2344 tbinptr treebins[NTREEBINS]; |
|
2345 size_t footprint; |
|
2346 size_t max_footprint; |
|
2347 flag_t mflags; |
|
2348 #if USE_LOCKS |
|
2349 MLOCK_T mutex; /* locate lock among fields that rarely change */ |
|
2350 #endif /* USE_LOCKS */ |
|
2351 msegment seg; |
|
2352 void* extp; /* Unused but available for extensions */ |
|
2353 size_t exts; |
|
2354 }; |
|
2355 |
|
2356 typedef struct malloc_state* mstate; |
|
2357 |
|
2358 /* ------------- Global malloc_state and malloc_params ------------------- */ |
|
2359 |
|
2360 /* |
|
2361 malloc_params holds global properties, including those that can be |
|
2362 dynamically set using mallopt. There is a single instance, mparams, |
|
2363 initialized in init_mparams. |
|
2364 */ |
|
2365 |
|
2366 struct malloc_params { |
|
2367 size_t magic; |
|
2368 size_t page_size; |
|
2369 size_t granularity; |
|
2370 size_t mmap_threshold; |
|
2371 size_t trim_threshold; |
|
2372 flag_t default_mflags; |
|
2373 }; |
|
2374 |
|
2375 static struct malloc_params mparams; |
|
2376 |
|
2377 #if !ONLY_MSPACES |
|
2378 |
|
2379 /* The global malloc_state used for all non-"mspace" calls */ |
|
2380 static struct malloc_state _gm_; |
|
2381 #define gm (&_gm_) |
|
2382 #define is_global(M) ((M) == &_gm_) |
|
2383 |
|
2384 #endif /* !ONLY_MSPACES */ |
|
2385 |
|
2386 #define is_initialized(M) ((M)->top != 0) |
|
2387 |
|
2388 /* -------------------------- system alloc setup ------------------------- */ |
|
2389 |
|
2390 /* Operations on mflags */ |
|
2391 |
|
2392 #define use_lock(M) ((M)->mflags & USE_LOCK_BIT) |
|
2393 #define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT) |
|
2394 #define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT) |
|
2395 |
|
2396 #define use_mmap(M) ((M)->mflags & USE_MMAP_BIT) |
|
2397 #define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT) |
|
2398 #define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT) |
|
2399 |
|
2400 #define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT) |
|
2401 #define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT) |
|
2402 |
|
2403 #define set_lock(M,L)\ |
|
2404 ((M)->mflags = (L)?\ |
|
2405 ((M)->mflags | USE_LOCK_BIT) :\ |
|
2406 ((M)->mflags & ~USE_LOCK_BIT)) |
|
2407 |
|
2408 /* page-align a size */ |
|
2409 #define page_align(S)\ |
|
2410 (((S) + (mparams.page_size - SIZE_T_ONE)) & ~(mparams.page_size - SIZE_T_ONE)) |
|
2411 |
|
2412 /* granularity-align a size */ |
|
2413 #define granularity_align(S)\ |
|
2414 (((S) + (mparams.granularity - SIZE_T_ONE))\ |
|
2415 & ~(mparams.granularity - SIZE_T_ONE)) |
|
2416 |
|
2417 |
|
2418 /* For mmap, use granularity alignment on windows, else page-align */ |
|
2419 #ifdef WIN32 |
|
2420 #define mmap_align(S) granularity_align(S) |
|
2421 #else |
|
2422 #define mmap_align(S) page_align(S) |
|
2423 #endif |
|
2424 |
|
2425 #define is_page_aligned(S)\ |
|
2426 (((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0) |
|
2427 #define is_granularity_aligned(S)\ |
|
2428 (((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0) |
|
2429 |
|
2430 /* True if segment S holds address A */ |
|
2431 #define segment_holds(S, A)\ |
|
2432 ((char*)(A) >= S->base && (char*)(A) < S->base + S->size) |
|
2433 |
|
2434 /* Return segment holding given address */ |
|
2435 static msegmentptr segment_holding(mstate m, char* addr) { |
|
2436 msegmentptr sp = &m->seg; |
|
2437 for (;;) { |
|
2438 if (addr >= sp->base && addr < sp->base + sp->size) |
|
2439 return sp; |
|
2440 if ((sp = sp->next) == 0) |
|
2441 return 0; |
|
2442 } |
|
2443 } |
|
2444 |
|
2445 /* Return true if segment contains a segment link */ |
|
2446 static int has_segment_link(mstate m, msegmentptr ss) { |
|
2447 msegmentptr sp = &m->seg; |
|
2448 for (;;) { |
|
2449 if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size) |
|
2450 return 1; |
|
2451 if ((sp = sp->next) == 0) |
|
2452 return 0; |
|
2453 } |
|
2454 } |
|
2455 |
|
2456 #ifndef MORECORE_CANNOT_TRIM |
|
2457 #define should_trim(M,s) ((s) > (M)->trim_check) |
|
2458 #else /* MORECORE_CANNOT_TRIM */ |
|
2459 #define should_trim(M,s) (0) |
|
2460 #endif /* MORECORE_CANNOT_TRIM */ |
|
2461 |
|
2462 /* |
|
2463 TOP_FOOT_SIZE is padding at the end of a segment, including space |
|
2464 that may be needed to place segment records and fenceposts when new |
|
2465 noncontiguous segments are added. |
|
2466 */ |
|
2467 #define TOP_FOOT_SIZE\ |
|
2468 (align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE) |
|
2469 |
|
2470 |
|
2471 /* ------------------------------- Hooks -------------------------------- */ |
|
2472 |
|
2473 /* |
|
2474 PREACTION should be defined to return 0 on success, and nonzero on |
|
2475 failure. If you are not using locking, you can redefine these to do |
|
2476 anything you like. |
|
2477 */ |
|
2478 |
|
2479 #if USE_LOCKS |
|
2480 |
|
2481 /* Ensure locks are initialized */ |
|
2482 #define GLOBALLY_INITIALIZE() (mparams.page_size == 0 && init_mparams()) |
|
2483 |
|
2484 #define PREACTION(M) ((GLOBALLY_INITIALIZE() || use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0) |
|
2485 #define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); } |
|
2486 #else /* USE_LOCKS */ |
|
2487 |
|
2488 #ifndef PREACTION |
|
2489 #define PREACTION(M) (0) |
|
2490 #endif /* PREACTION */ |
|
2491 |
|
2492 #ifndef POSTACTION |
|
2493 #define POSTACTION(M) |
|
2494 #endif /* POSTACTION */ |
|
2495 |
|
2496 #endif /* USE_LOCKS */ |
|
2497 |
|
2498 /* |
|
2499 CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses. |
|
2500 USAGE_ERROR_ACTION is triggered on detected bad frees and |
|
2501 reallocs. The argument p is an address that might have triggered the |
|
2502 fault. It is ignored by the two predefined actions, but might be |
|
2503 useful in custom actions that try to help diagnose errors. |
|
2504 */ |
|
2505 |
|
2506 #if PROCEED_ON_ERROR |
|
2507 |
|
2508 /* A count of the number of corruption errors causing resets */ |
|
2509 int malloc_corruption_error_count; |
|
2510 |
|
2511 /* default corruption action */ |
|
2512 static void reset_on_error(mstate m); |
|
2513 |
|
2514 #define CORRUPTION_ERROR_ACTION(m) reset_on_error(m) |
|
2515 #define USAGE_ERROR_ACTION(m, p) |
|
2516 |
|
2517 #else /* PROCEED_ON_ERROR */ |
|
2518 |
|
2519 #ifndef CORRUPTION_ERROR_ACTION |
|
2520 #define CORRUPTION_ERROR_ACTION(m) ABORT |
|
2521 #endif /* CORRUPTION_ERROR_ACTION */ |
|
2522 |
|
2523 #ifndef USAGE_ERROR_ACTION |
|
2524 #define USAGE_ERROR_ACTION(m,p) ABORT |
|
2525 #endif /* USAGE_ERROR_ACTION */ |
|
2526 |
|
2527 #endif /* PROCEED_ON_ERROR */ |
|
2528 |
|
2529 /* -------------------------- Debugging setup ---------------------------- */ |
|
2530 |
|
2531 #if ! DEBUG |
|
2532 |
|
2533 #define check_free_chunk(M,P) |
|
2534 #define check_inuse_chunk(M,P) |
|
2535 #define check_malloced_chunk(M,P,N) |
|
2536 #define check_mmapped_chunk(M,P) |
|
2537 #define check_malloc_state(M) |
|
2538 #define check_top_chunk(M,P) |
|
2539 |
|
2540 #else /* DEBUG */ |
|
2541 #define check_free_chunk(M,P) do_check_free_chunk(M,P) |
|
2542 #define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P) |
|
2543 #define check_top_chunk(M,P) do_check_top_chunk(M,P) |
|
2544 #define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N) |
|
2545 #define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P) |
|
2546 #define check_malloc_state(M) do_check_malloc_state(M) |
|
2547 |
|
2548 static void do_check_any_chunk(mstate m, mchunkptr p); |
|
2549 static void do_check_top_chunk(mstate m, mchunkptr p); |
|
2550 static void do_check_mmapped_chunk(mstate m, mchunkptr p); |
|
2551 static void do_check_inuse_chunk(mstate m, mchunkptr p); |
|
2552 static void do_check_free_chunk(mstate m, mchunkptr p); |
|
2553 static void do_check_malloced_chunk(mstate m, void* mem, size_t s); |
|
2554 static void do_check_tree(mstate m, tchunkptr t); |
|
2555 static void do_check_treebin(mstate m, bindex_t i); |
|
2556 static void do_check_smallbin(mstate m, bindex_t i); |
|
2557 static void do_check_malloc_state(mstate m); |
|
2558 static int bin_find(mstate m, mchunkptr x); |
|
2559 static size_t traverse_and_check(mstate m); |
|
2560 #endif /* DEBUG */ |
|
2561 |
|
2562 /* ---------------------------- Indexing Bins ---------------------------- */ |
|
2563 |
|
2564 #define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS) |
|
2565 #define small_index(s) ((s) >> SMALLBIN_SHIFT) |
|
2566 #define small_index2size(i) ((i) << SMALLBIN_SHIFT) |
|
2567 #define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE)) |
|
2568 |
|
2569 /* addressing by index. See above about smallbin repositioning */ |
|
2570 #define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1]))) |
|
2571 #define treebin_at(M,i) (&((M)->treebins[i])) |
|
2572 |
|
2573 /* assign tree index for size S to variable I */ |
|
2574 #if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) |
|
2575 #define compute_tree_index(S, I)\ |
|
2576 {\ |
|
2577 unsigned int X = S >> TREEBIN_SHIFT;\ |
|
2578 if (X == 0)\ |
|
2579 I = 0;\ |
|
2580 else if (X > 0xFFFF)\ |
|
2581 I = NTREEBINS-1;\ |
|
2582 else {\ |
|
2583 unsigned int K;\ |
|
2584 __asm__("bsrl\t%1, %0\n\t" : "=r" (K) : "g" (X));\ |
|
2585 I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\ |
|
2586 }\ |
|
2587 } |
|
2588 |
|
2589 #elif defined(_MSC_VER) && _MSC_VER>=1300 |
|
2590 #define compute_tree_index(S, I)\ |
|
2591 {\ |
|
2592 size_t X = S >> TREEBIN_SHIFT;\ |
|
2593 if (X == 0)\ |
|
2594 I = 0;\ |
|
2595 else if (X > 0xFFFF)\ |
|
2596 I = NTREEBINS-1;\ |
|
2597 else {\ |
|
2598 unsigned int K;\ |
|
2599 _BitScanReverse((DWORD *) &K, X);\ |
|
2600 I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\ |
|
2601 }\ |
|
2602 } |
|
2603 #else /* GNUC */ |
|
2604 #define compute_tree_index(S, I)\ |
|
2605 {\ |
|
2606 size_t X = S >> TREEBIN_SHIFT;\ |
|
2607 if (X == 0)\ |
|
2608 I = 0;\ |
|
2609 else if (X > 0xFFFF)\ |
|
2610 I = NTREEBINS-1;\ |
|
2611 else {\ |
|
2612 unsigned int Y = (unsigned int)X;\ |
|
2613 unsigned int N = ((Y - 0x100) >> 16) & 8;\ |
|
2614 unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\ |
|
2615 N += K;\ |
|
2616 N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\ |
|
2617 K = 14 - N + ((Y <<= K) >> 15);\ |
|
2618 I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\ |
|
2619 }\ |
|
2620 } |
|
2621 #endif /* GNUC */ |
|
2622 |
|
2623 /* Bit representing maximum resolved size in a treebin at i */ |
|
2624 #define bit_for_tree_index(i) \ |
|
2625 (i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2) |
|
2626 |
|
2627 /* Shift placing maximum resolved bit in a treebin at i as sign bit */ |
|
2628 #define leftshift_for_tree_index(i) \ |
|
2629 ((i == NTREEBINS-1)? 0 : \ |
|
2630 ((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2))) |
|
2631 |
|
2632 /* The size of the smallest chunk held in bin with index i */ |
|
2633 #define minsize_for_tree_index(i) \ |
|
2634 ((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \ |
|
2635 (((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1))) |
|
2636 |
|
2637 |
|
2638 /* ------------------------ Operations on bin maps ----------------------- */ |
|
2639 |
|
2640 /* bit corresponding to given index */ |
|
2641 #define idx2bit(i) ((binmap_t)(1) << (i)) |
|
2642 |
|
2643 /* Mark/Clear bits with given index */ |
|
2644 #define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i)) |
|
2645 #define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i)) |
|
2646 #define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i)) |
|
2647 |
|
2648 #define mark_treemap(M,i) ((M)->treemap |= idx2bit(i)) |
|
2649 #define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i)) |
|
2650 #define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i)) |
|
2651 |
|
2652 /* index corresponding to given bit */ |
|
2653 |
|
2654 #if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) |
|
2655 #define compute_bit2idx(X, I)\ |
|
2656 {\ |
|
2657 unsigned int J;\ |
|
2658 __asm__("bsfl\t%1, %0\n\t" : "=r" (J) : "g" (X));\ |
|
2659 I = (bindex_t)J;\ |
|
2660 } |
|
2661 #elif defined(_MSC_VER) && _MSC_VER>=1300 |
|
2662 #define compute_bit2idx(X, I)\ |
|
2663 {\ |
|
2664 unsigned int J;\ |
|
2665 _BitScanForward((DWORD *) &J, X);\ |
|
2666 I = (bindex_t)J;\ |
|
2667 } |
|
2668 |
|
2669 #else /* GNUC */ |
|
2670 #if USE_BUILTIN_FFS |
|
2671 #define compute_bit2idx(X, I) I = ffs(X)-1 |
|
2672 |
|
2673 #else /* USE_BUILTIN_FFS */ |
|
2674 #define compute_bit2idx(X, I)\ |
|
2675 {\ |
|
2676 unsigned int Y = X - 1;\ |
|
2677 unsigned int K = Y >> (16-4) & 16;\ |
|
2678 unsigned int N = K; Y >>= K;\ |
|
2679 N += K = Y >> (8-3) & 8; Y >>= K;\ |
|
2680 N += K = Y >> (4-2) & 4; Y >>= K;\ |
|
2681 N += K = Y >> (2-1) & 2; Y >>= K;\ |
|
2682 N += K = Y >> (1-0) & 1; Y >>= K;\ |
|
2683 I = (bindex_t)(N + Y);\ |
|
2684 } |
|
2685 #endif /* USE_BUILTIN_FFS */ |
|
2686 #endif /* GNUC */ |
|
2687 |
|
2688 /* isolate the least set bit of a bitmap */ |
|
2689 #define least_bit(x) ((x) & -(x)) |
|
2690 |
|
2691 /* mask with all bits to left of least bit of x on */ |
|
2692 #define left_bits(x) ((x<<1) | -(x<<1)) |
|
2693 |
|
2694 /* mask with all bits to left of or equal to least bit of x on */ |
|
2695 #define same_or_left_bits(x) ((x) | -(x)) |
|
2696 |
|
2697 |
|
2698 /* ----------------------- Runtime Check Support ------------------------- */ |
|
2699 |
|
2700 /* |
|
2701 For security, the main invariant is that malloc/free/etc never |
|
2702 writes to a static address other than malloc_state, unless static |
|
2703 malloc_state itself has been corrupted, which cannot occur via |
|
2704 malloc (because of these checks). In essence this means that we |
|
2705 believe all pointers, sizes, maps etc held in malloc_state, but |
|
2706 check all of those linked or offsetted from other embedded data |
|
2707 structures. These checks are interspersed with main code in a way |
|
2708 that tends to minimize their run-time cost. |
|
2709 |
|
2710 When FOOTERS is defined, in addition to range checking, we also |
|
2711 verify footer fields of inuse chunks, which can be used guarantee |
|
2712 that the mstate controlling malloc/free is intact. This is a |
|
2713 streamlined version of the approach described by William Robertson |
|
2714 et al in "Run-time Detection of Heap-based Overflows" LISA'03 |
|
2715 http://www.usenix.org/events/lisa03/tech/robertson.html The footer |
|
2716 of an inuse chunk holds the xor of its mstate and a random seed, |
|
2717 that is checked upon calls to free() and realloc(). This is |
|
2718 (probablistically) unguessable from outside the program, but can be |
|
2719 computed by any code successfully malloc'ing any chunk, so does not |
|
2720 itself provide protection against code that has already broken |
|
2721 security through some other means. Unlike Robertson et al, we |
|
2722 always dynamically check addresses of all offset chunks (previous, |
|
2723 next, etc). This turns out to be cheaper than relying on hashes. |
|
2724 */ |
|
2725 |
|
2726 #if !INSECURE |
|
2727 /* Check if address a is at least as high as any from MORECORE or MMAP */ |
|
2728 #define ok_address(M, a) ((char*)(a) >= (M)->least_addr) |
|
2729 /* Check if address of next chunk n is higher than base chunk p */ |
|
2730 #define ok_next(p, n) ((char*)(p) < (char*)(n)) |
|
2731 /* Check if p has its cinuse bit on */ |
|
2732 #define ok_cinuse(p) cinuse(p) |
|
2733 /* Check if p has its pinuse bit on */ |
|
2734 #define ok_pinuse(p) pinuse(p) |
|
2735 |
|
2736 #else /* !INSECURE */ |
|
2737 #define ok_address(M, a) (1) |
|
2738 #define ok_next(b, n) (1) |
|
2739 #define ok_cinuse(p) (1) |
|
2740 #define ok_pinuse(p) (1) |
|
2741 #endif /* !INSECURE */ |
|
2742 |
|
2743 #if (FOOTERS && !INSECURE) |
|
2744 /* Check if (alleged) mstate m has expected magic field */ |
|
2745 #define ok_magic(M) ((M)->magic == mparams.magic) |
|
2746 #else /* (FOOTERS && !INSECURE) */ |
|
2747 #define ok_magic(M) (1) |
|
2748 #endif /* (FOOTERS && !INSECURE) */ |
|
2749 |
|
2750 |
|
2751 /* In gcc, use __builtin_expect to minimize impact of checks */ |
|
2752 #if !INSECURE |
|
2753 #if defined(__GNUC__) && __GNUC__ >= 3 |
|
2754 #define RTCHECK(e) __builtin_expect(e, 1) |
|
2755 #else /* GNUC */ |
|
2756 #define RTCHECK(e) (e) |
|
2757 #endif /* GNUC */ |
|
2758 #else /* !INSECURE */ |
|
2759 #define RTCHECK(e) (1) |
|
2760 #endif /* !INSECURE */ |
|
2761 |
|
2762 /* macros to set up inuse chunks with or without footers */ |
|
2763 |
|
2764 #if !FOOTERS |
|
2765 |
|
2766 #define mark_inuse_foot(M,p,s) |
|
2767 |
|
2768 /* Set cinuse bit and pinuse bit of next chunk */ |
|
2769 #define set_inuse(M,p,s)\ |
|
2770 ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\ |
|
2771 ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT) |
|
2772 |
|
2773 /* Set cinuse and pinuse of this chunk and pinuse of next chunk */ |
|
2774 #define set_inuse_and_pinuse(M,p,s)\ |
|
2775 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ |
|
2776 ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT) |
|
2777 |
|
2778 /* Set size, cinuse and pinuse bit of this chunk */ |
|
2779 #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\ |
|
2780 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT)) |
|
2781 |
|
2782 #else /* FOOTERS */ |
|
2783 |
|
2784 /* Set foot of inuse chunk to be xor of mstate and seed */ |
|
2785 #define mark_inuse_foot(M,p,s)\ |
|
2786 (((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic)) |
|
2787 |
|
2788 #define get_mstate_for(p)\ |
|
2789 ((mstate)(((mchunkptr)((char*)(p) +\ |
|
2790 (chunksize(p))))->prev_foot ^ mparams.magic)) |
|
2791 |
|
2792 #define set_inuse(M,p,s)\ |
|
2793 ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\ |
|
2794 (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \ |
|
2795 mark_inuse_foot(M,p,s)) |
|
2796 |
|
2797 #define set_inuse_and_pinuse(M,p,s)\ |
|
2798 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ |
|
2799 (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\ |
|
2800 mark_inuse_foot(M,p,s)) |
|
2801 |
|
2802 #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\ |
|
2803 ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ |
|
2804 mark_inuse_foot(M, p, s)) |
|
2805 |
|
2806 #endif /* !FOOTERS */ |
|
2807 |
|
2808 /* ---------------------------- setting mparams -------------------------- */ |
|
2809 |
|
2810 /* Initialize mparams */ |
|
2811 static int init_mparams(void) { |
|
2812 if (mparams.page_size == 0) { |
|
2813 size_t s; |
|
2814 |
|
2815 mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD; |
|
2816 mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD; |
|
2817 #if MORECORE_CONTIGUOUS |
|
2818 mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT; |
|
2819 #else /* MORECORE_CONTIGUOUS */ |
|
2820 mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT|USE_NONCONTIGUOUS_BIT; |
|
2821 #endif /* MORECORE_CONTIGUOUS */ |
|
2822 |
|
2823 #if (FOOTERS && !INSECURE) |
|
2824 { |
|
2825 #if USE_DEV_RANDOM |
|
2826 int fd; |
|
2827 unsigned char buf[sizeof(size_t)]; |
|
2828 /* Try to use /dev/urandom, else fall back on using time */ |
|
2829 if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 && |
|
2830 read(fd, buf, sizeof(buf)) == sizeof(buf)) { |
|
2831 s = *((size_t *) buf); |
|
2832 close(fd); |
|
2833 } |
|
2834 else |
|
2835 #endif /* USE_DEV_RANDOM */ |
|
2836 s = (size_t)(time(0) ^ (size_t)0x55555555U); |
|
2837 |
|
2838 s |= (size_t)8U; /* ensure nonzero */ |
|
2839 s &= ~(size_t)7U; /* improve chances of fault for bad values */ |
|
2840 |
|
2841 } |
|
2842 #else /* (FOOTERS && !INSECURE) */ |
|
2843 s = (size_t)0x58585858U; |
|
2844 #endif /* (FOOTERS && !INSECURE) */ |
|
2845 ACQUIRE_MAGIC_INIT_LOCK(); |
|
2846 if (mparams.magic == 0) { |
|
2847 mparams.magic = s; |
|
2848 #if !ONLY_MSPACES |
|
2849 /* Set up lock for main malloc area */ |
|
2850 INITIAL_LOCK(&gm->mutex); |
|
2851 gm->mflags = mparams.default_mflags; |
|
2852 #endif |
|
2853 } |
|
2854 RELEASE_MAGIC_INIT_LOCK(); |
|
2855 |
|
2856 #ifndef WIN32 |
|
2857 mparams.page_size = malloc_getpagesize; |
|
2858 mparams.granularity = ((DEFAULT_GRANULARITY != 0)? |
|
2859 DEFAULT_GRANULARITY : mparams.page_size); |
|
2860 #else /* WIN32 */ |
|
2861 { |
|
2862 SYSTEM_INFO system_info; |
|
2863 GetSystemInfo(&system_info); |
|
2864 mparams.page_size = system_info.dwPageSize; |
|
2865 mparams.granularity = system_info.dwAllocationGranularity; |
|
2866 } |
|
2867 #endif /* WIN32 */ |
|
2868 |
|
2869 /* Sanity-check configuration: |
|
2870 size_t must be unsigned and as wide as pointer type. |
|
2871 ints must be at least 4 bytes. |
|
2872 alignment must be at least 8. |
|
2873 Alignment, min chunk size, and page size must all be powers of 2. |
|
2874 */ |
|
2875 if ((sizeof(size_t) != sizeof(char*)) || |
|
2876 (MAX_SIZE_T < MIN_CHUNK_SIZE) || |
|
2877 (sizeof(int) < 4) || |
|
2878 (MALLOC_ALIGNMENT < (size_t)8U) || |
|
2879 ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT-SIZE_T_ONE)) != 0) || |
|
2880 ((MCHUNK_SIZE & (MCHUNK_SIZE-SIZE_T_ONE)) != 0) || |
|
2881 ((mparams.granularity & (mparams.granularity-SIZE_T_ONE)) != 0) || |
|
2882 ((mparams.page_size & (mparams.page_size-SIZE_T_ONE)) != 0)) |
|
2883 ABORT; |
|
2884 } |
|
2885 return 0; |
|
2886 } |
|
2887 |
|
2888 /* support for mallopt */ |
|
2889 static int change_mparam(int param_number, int value) { |
|
2890 size_t val = (size_t)value; |
|
2891 init_mparams(); |
|
2892 switch(param_number) { |
|
2893 case M_TRIM_THRESHOLD: |
|
2894 mparams.trim_threshold = val; |
|
2895 return 1; |
|
2896 case M_GRANULARITY: |
|
2897 if (val >= mparams.page_size && ((val & (val-1)) == 0)) { |
|
2898 mparams.granularity = val; |
|
2899 return 1; |
|
2900 } |
|
2901 else |
|
2902 return 0; |
|
2903 case M_MMAP_THRESHOLD: |
|
2904 mparams.mmap_threshold = val; |
|
2905 return 1; |
|
2906 default: |
|
2907 return 0; |
|
2908 } |
|
2909 } |
|
2910 |
|
2911 #if DEBUG |
|
2912 /* ------------------------- Debugging Support --------------------------- */ |
|
2913 |
|
2914 /* Check properties of any chunk, whether free, inuse, mmapped etc */ |
|
2915 static void do_check_any_chunk(mstate m, mchunkptr p) { |
|
2916 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); |
|
2917 assert(ok_address(m, p)); |
|
2918 } |
|
2919 |
|
2920 /* Check properties of top chunk */ |
|
2921 static void do_check_top_chunk(mstate m, mchunkptr p) { |
|
2922 msegmentptr sp = segment_holding(m, (char*)p); |
|
2923 size_t sz = p->head & ~INUSE_BITS; /* third-lowest bit can be set! */ |
|
2924 assert(sp != 0); |
|
2925 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); |
|
2926 assert(ok_address(m, p)); |
|
2927 assert(sz == m->topsize); |
|
2928 assert(sz > 0); |
|
2929 assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE); |
|
2930 assert(pinuse(p)); |
|
2931 assert(!pinuse(chunk_plus_offset(p, sz))); |
|
2932 } |
|
2933 |
|
2934 /* Check properties of (inuse) mmapped chunks */ |
|
2935 static void do_check_mmapped_chunk(mstate m, mchunkptr p) { |
|
2936 size_t sz = chunksize(p); |
|
2937 size_t len = (sz + (p->prev_foot & ~IS_MMAPPED_BIT) + MMAP_FOOT_PAD); |
|
2938 assert(is_mmapped(p)); |
|
2939 assert(use_mmap(m)); |
|
2940 assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); |
|
2941 assert(ok_address(m, p)); |
|
2942 assert(!is_small(sz)); |
|
2943 assert((len & (mparams.page_size-SIZE_T_ONE)) == 0); |
|
2944 assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD); |
|
2945 assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0); |
|
2946 } |
|
2947 |
|
2948 /* Check properties of inuse chunks */ |
|
2949 static void do_check_inuse_chunk(mstate m, mchunkptr p) { |
|
2950 do_check_any_chunk(m, p); |
|
2951 assert(cinuse(p)); |
|
2952 assert(next_pinuse(p)); |
|
2953 /* If not pinuse and not mmapped, previous chunk has OK offset */ |
|
2954 assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p); |
|
2955 if (is_mmapped(p)) |
|
2956 do_check_mmapped_chunk(m, p); |
|
2957 } |
|
2958 |
|
2959 /* Check properties of free chunks */ |
|
2960 static void do_check_free_chunk(mstate m, mchunkptr p) { |
|
2961 size_t sz = chunksize(p); |
|
2962 mchunkptr next = chunk_plus_offset(p, sz); |
|
2963 do_check_any_chunk(m, p); |
|
2964 assert(!cinuse(p)); |
|
2965 assert(!next_pinuse(p)); |
|
2966 assert (!is_mmapped(p)); |
|
2967 if (p != m->dv && p != m->top) { |
|
2968 if (sz >= MIN_CHUNK_SIZE) { |
|
2969 assert((sz & CHUNK_ALIGN_MASK) == 0); |
|
2970 assert(is_aligned(chunk2mem(p))); |
|
2971 assert(next->prev_foot == sz); |
|
2972 assert(pinuse(p)); |
|
2973 assert (next == m->top || cinuse(next)); |
|
2974 assert(p->fd->bk == p); |
|
2975 assert(p->bk->fd == p); |
|
2976 } |
|
2977 else /* markers are always of size SIZE_T_SIZE */ |
|
2978 assert(sz == SIZE_T_SIZE); |
|
2979 } |
|
2980 } |
|
2981 |
|
2982 /* Check properties of malloced chunks at the point they are malloced */ |
|
2983 static void do_check_malloced_chunk(mstate m, void* mem, size_t s) { |
|
2984 if (mem != 0) { |
|
2985 mchunkptr p = mem2chunk(mem); |
|
2986 size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT); |
|
2987 do_check_inuse_chunk(m, p); |
|
2988 assert((sz & CHUNK_ALIGN_MASK) == 0); |
|
2989 assert(sz >= MIN_CHUNK_SIZE); |
|
2990 assert(sz >= s); |
|
2991 /* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */ |
|
2992 assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE)); |
|
2993 } |
|
2994 } |
|
2995 |
|
2996 /* Check a tree and its subtrees. */ |
|
2997 static void do_check_tree(mstate m, tchunkptr t) { |
|
2998 tchunkptr head = 0; |
|
2999 tchunkptr u = t; |
|
3000 bindex_t tindex = t->index; |
|
3001 size_t tsize = chunksize(t); |
|
3002 bindex_t idx; |
|
3003 compute_tree_index(tsize, idx); |
|
3004 assert(tindex == idx); |
|
3005 assert(tsize >= MIN_LARGE_SIZE); |
|
3006 assert(tsize >= minsize_for_tree_index(idx)); |
|
3007 assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1)))); |
|
3008 |
|
3009 do { /* traverse through chain of same-sized nodes */ |
|
3010 do_check_any_chunk(m, ((mchunkptr)u)); |
|
3011 assert(u->index == tindex); |
|
3012 assert(chunksize(u) == tsize); |
|
3013 assert(!cinuse(u)); |
|
3014 assert(!next_pinuse(u)); |
|
3015 assert(u->fd->bk == u); |
|
3016 assert(u->bk->fd == u); |
|
3017 if (u->parent == 0) { |
|
3018 assert(u->child[0] == 0); |
|
3019 assert(u->child[1] == 0); |
|
3020 } |
|
3021 else { |
|
3022 assert(head == 0); /* only one node on chain has parent */ |
|
3023 head = u; |
|
3024 assert(u->parent != u); |
|
3025 assert (u->parent->child[0] == u || |
|
3026 u->parent->child[1] == u || |
|
3027 *((tbinptr*)(u->parent)) == u); |
|
3028 if (u->child[0] != 0) { |
|
3029 assert(u->child[0]->parent == u); |
|
3030 assert(u->child[0] != u); |
|
3031 do_check_tree(m, u->child[0]); |
|
3032 } |
|
3033 if (u->child[1] != 0) { |
|
3034 assert(u->child[1]->parent == u); |
|
3035 assert(u->child[1] != u); |
|
3036 do_check_tree(m, u->child[1]); |
|
3037 } |
|
3038 if (u->child[0] != 0 && u->child[1] != 0) { |
|
3039 assert(chunksize(u->child[0]) < chunksize(u->child[1])); |
|
3040 } |
|
3041 } |
|
3042 u = u->fd; |
|
3043 } while (u != t); |
|
3044 assert(head != 0); |
|
3045 } |
|
3046 |
|
3047 /* Check all the chunks in a treebin. */ |
|
3048 static void do_check_treebin(mstate m, bindex_t i) { |
|
3049 tbinptr* tb = treebin_at(m, i); |
|
3050 tchunkptr t = *tb; |
|
3051 int empty = (m->treemap & (1U << i)) == 0; |
|
3052 if (t == 0) |
|
3053 assert(empty); |
|
3054 if (!empty) |
|
3055 do_check_tree(m, t); |
|
3056 } |
|
3057 |
|
3058 /* Check all the chunks in a smallbin. */ |
|
3059 static void do_check_smallbin(mstate m, bindex_t i) { |
|
3060 sbinptr b = smallbin_at(m, i); |
|
3061 mchunkptr p = b->bk; |
|
3062 unsigned int empty = (m->smallmap & (1U << i)) == 0; |
|
3063 if (p == b) |
|
3064 assert(empty); |
|
3065 if (!empty) { |
|
3066 for (; p != b; p = p->bk) { |
|
3067 size_t size = chunksize(p); |
|
3068 mchunkptr q; |
|
3069 /* each chunk claims to be free */ |
|
3070 do_check_free_chunk(m, p); |
|
3071 /* chunk belongs in bin */ |
|
3072 assert(small_index(size) == i); |
|
3073 assert(p->bk == b || chunksize(p->bk) == chunksize(p)); |
|
3074 /* chunk is followed by an inuse chunk */ |
|
3075 q = next_chunk(p); |
|
3076 if (q->head != FENCEPOST_HEAD) |
|
3077 do_check_inuse_chunk(m, q); |
|
3078 } |
|
3079 } |
|
3080 } |
|
3081 |
|
3082 /* Find x in a bin. Used in other check functions. */ |
|
3083 static int bin_find(mstate m, mchunkptr x) { |
|
3084 size_t size = chunksize(x); |
|
3085 if (is_small(size)) { |
|
3086 bindex_t sidx = small_index(size); |
|
3087 sbinptr b = smallbin_at(m, sidx); |
|
3088 if (smallmap_is_marked(m, sidx)) { |
|
3089 mchunkptr p = b; |
|
3090 do { |
|
3091 if (p == x) |
|
3092 return 1; |
|
3093 } while ((p = p->fd) != b); |
|
3094 } |
|
3095 } |
|
3096 else { |
|
3097 bindex_t tidx; |
|
3098 compute_tree_index(size, tidx); |
|
3099 if (treemap_is_marked(m, tidx)) { |
|
3100 tchunkptr t = *treebin_at(m, tidx); |
|
3101 size_t sizebits = size << leftshift_for_tree_index(tidx); |
|
3102 while (t != 0 && chunksize(t) != size) { |
|
3103 t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]; |
|
3104 sizebits <<= 1; |
|
3105 } |
|
3106 if (t != 0) { |
|
3107 tchunkptr u = t; |
|
3108 do { |
|
3109 if (u == (tchunkptr)x) |
|
3110 return 1; |
|
3111 } while ((u = u->fd) != t); |
|
3112 } |
|
3113 } |
|
3114 } |
|
3115 return 0; |
|
3116 } |
|
3117 |
|
3118 /* Traverse each chunk and check it; return total */ |
|
3119 static size_t traverse_and_check(mstate m) { |
|
3120 size_t sum = 0; |
|
3121 if (is_initialized(m)) { |
|
3122 msegmentptr s = &m->seg; |
|
3123 sum += m->topsize + TOP_FOOT_SIZE; |
|
3124 while (s != 0) { |
|
3125 mchunkptr q = align_as_chunk(s->base); |
|
3126 mchunkptr lastq = 0; |
|
3127 assert(pinuse(q)); |
|
3128 while (segment_holds(s, q) && |
|
3129 q != m->top && q->head != FENCEPOST_HEAD) { |
|
3130 sum += chunksize(q); |
|
3131 if (cinuse(q)) { |
|
3132 assert(!bin_find(m, q)); |
|
3133 do_check_inuse_chunk(m, q); |
|
3134 } |
|
3135 else { |
|
3136 assert(q == m->dv || bin_find(m, q)); |
|
3137 assert(lastq == 0 || cinuse(lastq)); /* Not 2 consecutive free */ |
|
3138 do_check_free_chunk(m, q); |
|
3139 } |
|
3140 lastq = q; |
|
3141 q = next_chunk(q); |
|
3142 } |
|
3143 s = s->next; |
|
3144 } |
|
3145 } |
|
3146 return sum; |
|
3147 } |
|
3148 |
|
3149 /* Check all properties of malloc_state. */ |
|
3150 static void do_check_malloc_state(mstate m) { |
|
3151 bindex_t i; |
|
3152 size_t total; |
|
3153 /* check bins */ |
|
3154 for (i = 0; i < NSMALLBINS; ++i) |
|
3155 do_check_smallbin(m, i); |
|
3156 for (i = 0; i < NTREEBINS; ++i) |
|
3157 do_check_treebin(m, i); |
|
3158 |
|
3159 if (m->dvsize != 0) { /* check dv chunk */ |
|
3160 do_check_any_chunk(m, m->dv); |
|
3161 assert(m->dvsize == chunksize(m->dv)); |
|
3162 assert(m->dvsize >= MIN_CHUNK_SIZE); |
|
3163 assert(bin_find(m, m->dv) == 0); |
|
3164 } |
|
3165 |
|
3166 if (m->top != 0) { /* check top chunk */ |
|
3167 do_check_top_chunk(m, m->top); |
|
3168 /*assert(m->topsize == chunksize(m->top)); redundant */ |
|
3169 assert(m->topsize > 0); |
|
3170 assert(bin_find(m, m->top) == 0); |
|
3171 } |
|
3172 |
|
3173 total = traverse_and_check(m); |
|
3174 assert(total <= m->footprint); |
|
3175 assert(m->footprint <= m->max_footprint); |
|
3176 } |
|
3177 #endif /* DEBUG */ |
|
3178 |
|
3179 /* ----------------------------- statistics ------------------------------ */ |
|
3180 |
|
3181 #if !NO_MALLINFO |
|
3182 static struct mallinfo internal_mallinfo(mstate m) { |
|
3183 struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; |
|
3184 if (!PREACTION(m)) { |
|
3185 check_malloc_state(m); |
|
3186 if (is_initialized(m)) { |
|
3187 size_t nfree = SIZE_T_ONE; /* top always free */ |
|
3188 size_t mfree = m->topsize + TOP_FOOT_SIZE; |
|
3189 size_t sum = mfree; |
|
3190 msegmentptr s = &m->seg; |
|
3191 while (s != 0) { |
|
3192 mchunkptr q = align_as_chunk(s->base); |
|
3193 while (segment_holds(s, q) && |
|
3194 q != m->top && q->head != FENCEPOST_HEAD) { |
|
3195 size_t sz = chunksize(q); |
|
3196 sum += sz; |
|
3197 if (!cinuse(q)) { |
|
3198 mfree += sz; |
|
3199 ++nfree; |
|
3200 } |
|
3201 q = next_chunk(q); |
|
3202 } |
|
3203 s = s->next; |
|
3204 } |
|
3205 |
|
3206 nm.arena = sum; |
|
3207 nm.ordblks = nfree; |
|
3208 nm.hblkhd = m->footprint - sum; |
|
3209 nm.usmblks = m->max_footprint; |
|
3210 nm.uordblks = m->footprint - mfree; |
|
3211 nm.fordblks = mfree; |
|
3212 nm.keepcost = m->topsize; |
|
3213 } |
|
3214 |
|
3215 POSTACTION(m); |
|
3216 } |
|
3217 return nm; |
|
3218 } |
|
3219 #endif /* !NO_MALLINFO */ |
|
3220 |
|
3221 static void internal_malloc_stats(mstate m) { |
|
3222 if (!PREACTION(m)) { |
|
3223 size_t maxfp = 0; |
|
3224 size_t fp = 0; |
|
3225 size_t used = 0; |
|
3226 check_malloc_state(m); |
|
3227 if (is_initialized(m)) { |
|
3228 msegmentptr s = &m->seg; |
|
3229 maxfp = m->max_footprint; |
|
3230 fp = m->footprint; |
|
3231 used = fp - (m->topsize + TOP_FOOT_SIZE); |
|
3232 |
|
3233 while (s != 0) { |
|
3234 mchunkptr q = align_as_chunk(s->base); |
|
3235 while (segment_holds(s, q) && |
|
3236 q != m->top && q->head != FENCEPOST_HEAD) { |
|
3237 if (!cinuse(q)) |
|
3238 used -= chunksize(q); |
|
3239 q = next_chunk(q); |
|
3240 } |
|
3241 s = s->next; |
|
3242 } |
|
3243 } |
|
3244 |
|
3245 fprintf(stderr, "max system bytes = %10lu\n", (unsigned long)(maxfp)); |
|
3246 fprintf(stderr, "system bytes = %10lu\n", (unsigned long)(fp)); |
|
3247 fprintf(stderr, "in use bytes = %10lu\n", (unsigned long)(used)); |
|
3248 |
|
3249 POSTACTION(m); |
|
3250 } |
|
3251 } |
|
3252 |
|
3253 /* ----------------------- Operations on smallbins ----------------------- */ |
|
3254 |
|
3255 /* |
|
3256 Various forms of linking and unlinking are defined as macros. Even |
|
3257 the ones for trees, which are very long but have very short typical |
|
3258 paths. This is ugly but reduces reliance on inlining support of |
|
3259 compilers. |
|
3260 */ |
|
3261 |
|
3262 /* Link a free chunk into a smallbin */ |
|
3263 #define insert_small_chunk(M, P, S) {\ |
|
3264 bindex_t I = small_index(S);\ |
|
3265 mchunkptr B = smallbin_at(M, I);\ |
|
3266 mchunkptr F = B;\ |
|
3267 assert(S >= MIN_CHUNK_SIZE);\ |
|
3268 if (!smallmap_is_marked(M, I))\ |
|
3269 mark_smallmap(M, I);\ |
|
3270 else if (RTCHECK(ok_address(M, B->fd)))\ |
|
3271 F = B->fd;\ |
|
3272 else {\ |
|
3273 CORRUPTION_ERROR_ACTION(M);\ |
|
3274 }\ |
|
3275 B->fd = P;\ |
|
3276 F->bk = P;\ |
|
3277 P->fd = F;\ |
|
3278 P->bk = B;\ |
|
3279 } |
|
3280 |
|
3281 /* Unlink a chunk from a smallbin */ |
|
3282 #define unlink_small_chunk(M, P, S) {\ |
|
3283 mchunkptr F = P->fd;\ |
|
3284 mchunkptr B = P->bk;\ |
|
3285 bindex_t I = small_index(S);\ |
|
3286 assert(P != B);\ |
|
3287 assert(P != F);\ |
|
3288 assert(chunksize(P) == small_index2size(I));\ |
|
3289 if (F == B)\ |
|
3290 clear_smallmap(M, I);\ |
|
3291 else if (RTCHECK((F == smallbin_at(M,I) || ok_address(M, F)) &&\ |
|
3292 (B == smallbin_at(M,I) || ok_address(M, B)))) {\ |
|
3293 F->bk = B;\ |
|
3294 B->fd = F;\ |
|
3295 }\ |
|
3296 else {\ |
|
3297 CORRUPTION_ERROR_ACTION(M);\ |
|
3298 }\ |
|
3299 } |
|
3300 |
|
3301 /* Unlink the first chunk from a smallbin */ |
|
3302 #define unlink_first_small_chunk(M, B, P, I) {\ |
|
3303 mchunkptr F = P->fd;\ |
|
3304 assert(P != B);\ |
|
3305 assert(P != F);\ |
|
3306 assert(chunksize(P) == small_index2size(I));\ |
|
3307 if (B == F)\ |
|
3308 clear_smallmap(M, I);\ |
|
3309 else if (RTCHECK(ok_address(M, F))) {\ |
|
3310 B->fd = F;\ |
|
3311 F->bk = B;\ |
|
3312 }\ |
|
3313 else {\ |
|
3314 CORRUPTION_ERROR_ACTION(M);\ |
|
3315 }\ |
|
3316 } |
|
3317 |
|
3318 /* Replace dv node, binning the old one */ |
|
3319 /* Used only when dvsize known to be small */ |
|
3320 #define replace_dv(M, P, S) {\ |
|
3321 size_t DVS = M->dvsize;\ |
|
3322 if (DVS != 0) {\ |
|
3323 mchunkptr DV = M->dv;\ |
|
3324 assert(is_small(DVS));\ |
|
3325 insert_small_chunk(M, DV, DVS);\ |
|
3326 }\ |
|
3327 M->dvsize = S;\ |
|
3328 M->dv = P;\ |
|
3329 } |
|
3330 |
|
3331 /* ------------------------- Operations on trees ------------------------- */ |
|
3332 |
|
3333 /* Insert chunk into tree */ |
|
3334 #define insert_large_chunk(M, X, S) {\ |
|
3335 tbinptr* H;\ |
|
3336 bindex_t I;\ |
|
3337 compute_tree_index(S, I);\ |
|
3338 H = treebin_at(M, I);\ |
|
3339 X->index = I;\ |
|
3340 X->child[0] = X->child[1] = 0;\ |
|
3341 if (!treemap_is_marked(M, I)) {\ |
|
3342 mark_treemap(M, I);\ |
|
3343 *H = X;\ |
|
3344 X->parent = (tchunkptr)H;\ |
|
3345 X->fd = X->bk = X;\ |
|
3346 }\ |
|
3347 else {\ |
|
3348 tchunkptr T = *H;\ |
|
3349 size_t K = S << leftshift_for_tree_index(I);\ |
|
3350 for (;;) {\ |
|
3351 if (chunksize(T) != S) {\ |
|
3352 tchunkptr* C = &(T->child[(K >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]);\ |
|
3353 K <<= 1;\ |
|
3354 if (*C != 0)\ |
|
3355 T = *C;\ |
|
3356 else if (RTCHECK(ok_address(M, C))) {\ |
|
3357 *C = X;\ |
|
3358 X->parent = T;\ |
|
3359 X->fd = X->bk = X;\ |
|
3360 break;\ |
|
3361 }\ |
|
3362 else {\ |
|
3363 CORRUPTION_ERROR_ACTION(M);\ |
|
3364 break;\ |
|
3365 }\ |
|
3366 }\ |
|
3367 else {\ |
|
3368 tchunkptr F = T->fd;\ |
|
3369 if (RTCHECK(ok_address(M, T) && ok_address(M, F))) {\ |
|
3370 T->fd = F->bk = X;\ |
|
3371 X->fd = F;\ |
|
3372 X->bk = T;\ |
|
3373 X->parent = 0;\ |
|
3374 break;\ |
|
3375 }\ |
|
3376 else {\ |
|
3377 CORRUPTION_ERROR_ACTION(M);\ |
|
3378 break;\ |
|
3379 }\ |
|
3380 }\ |
|
3381 }\ |
|
3382 }\ |
|
3383 } |
|
3384 |
|
3385 /* |
|
3386 Unlink steps: |
|
3387 |
|
3388 1. If x is a chained node, unlink it from its same-sized fd/bk links |
|
3389 and choose its bk node as its replacement. |
|
3390 2. If x was the last node of its size, but not a leaf node, it must |
|
3391 be replaced with a leaf node (not merely one with an open left or |
|
3392 right), to make sure that lefts and rights of descendents |
|
3393 correspond properly to bit masks. We use the rightmost descendent |
|
3394 of x. We could use any other leaf, but this is easy to locate and |
|
3395 tends to counteract removal of leftmosts elsewhere, and so keeps |
|
3396 paths shorter than minimally guaranteed. This doesn't loop much |
|
3397 because on average a node in a tree is near the bottom. |
|
3398 3. If x is the base of a chain (i.e., has parent links) relink |
|
3399 x's parent and children to x's replacement (or null if none). |
|
3400 */ |
|
3401 |
|
3402 #define unlink_large_chunk(M, X) {\ |
|
3403 tchunkptr XP = X->parent;\ |
|
3404 tchunkptr R;\ |
|
3405 if (X->bk != X) {\ |
|
3406 tchunkptr F = X->fd;\ |
|
3407 R = X->bk;\ |
|
3408 if (RTCHECK(ok_address(M, F))) {\ |
|
3409 F->bk = R;\ |
|
3410 R->fd = F;\ |
|
3411 }\ |
|
3412 else {\ |
|
3413 CORRUPTION_ERROR_ACTION(M);\ |
|
3414 }\ |
|
3415 }\ |
|
3416 else {\ |
|
3417 tchunkptr* RP;\ |
|
3418 if (((R = *(RP = &(X->child[1]))) != 0) ||\ |
|
3419 ((R = *(RP = &(X->child[0]))) != 0)) {\ |
|
3420 tchunkptr* CP;\ |
|
3421 while ((*(CP = &(R->child[1])) != 0) ||\ |
|
3422 (*(CP = &(R->child[0])) != 0)) {\ |
|
3423 R = *(RP = CP);\ |
|
3424 }\ |
|
3425 if (RTCHECK(ok_address(M, RP)))\ |
|
3426 *RP = 0;\ |
|
3427 else {\ |
|
3428 CORRUPTION_ERROR_ACTION(M);\ |
|
3429 }\ |
|
3430 }\ |
|
3431 }\ |
|
3432 if (XP != 0) {\ |
|
3433 tbinptr* H = treebin_at(M, X->index);\ |
|
3434 if (X == *H) {\ |
|
3435 if ((*H = R) == 0) \ |
|
3436 clear_treemap(M, X->index);\ |
|
3437 }\ |
|
3438 else if (RTCHECK(ok_address(M, XP))) {\ |
|
3439 if (XP->child[0] == X) \ |
|
3440 XP->child[0] = R;\ |
|
3441 else \ |
|
3442 XP->child[1] = R;\ |
|
3443 }\ |
|
3444 else\ |
|
3445 CORRUPTION_ERROR_ACTION(M);\ |
|
3446 if (R != 0) {\ |
|
3447 if (RTCHECK(ok_address(M, R))) {\ |
|
3448 tchunkptr C0, C1;\ |
|
3449 R->parent = XP;\ |
|
3450 if ((C0 = X->child[0]) != 0) {\ |
|
3451 if (RTCHECK(ok_address(M, C0))) {\ |
|
3452 R->child[0] = C0;\ |
|
3453 C0->parent = R;\ |
|
3454 }\ |
|
3455 else\ |
|
3456 CORRUPTION_ERROR_ACTION(M);\ |
|
3457 }\ |
|
3458 if ((C1 = X->child[1]) != 0) {\ |
|
3459 if (RTCHECK(ok_address(M, C1))) {\ |
|
3460 R->child[1] = C1;\ |
|
3461 C1->parent = R;\ |
|
3462 }\ |
|
3463 else\ |
|
3464 CORRUPTION_ERROR_ACTION(M);\ |
|
3465 }\ |
|
3466 }\ |
|
3467 else\ |
|
3468 CORRUPTION_ERROR_ACTION(M);\ |
|
3469 }\ |
|
3470 }\ |
|
3471 } |
|
3472 |
|
3473 /* Relays to large vs small bin operations */ |
|
3474 |
|
3475 #define insert_chunk(M, P, S)\ |
|
3476 if (is_small(S)) insert_small_chunk(M, P, S)\ |
|
3477 else { tchunkptr TP = (tchunkptr)(P); insert_large_chunk(M, TP, S); } |
|
3478 |
|
3479 #define unlink_chunk(M, P, S)\ |
|
3480 if (is_small(S)) unlink_small_chunk(M, P, S)\ |
|
3481 else { tchunkptr TP = (tchunkptr)(P); unlink_large_chunk(M, TP); } |
|
3482 |
|
3483 |
|
3484 /* Relays to internal calls to malloc/free from realloc, memalign etc */ |
|
3485 |
|
3486 #if ONLY_MSPACES |
|
3487 #define internal_malloc(m, b) mspace_malloc(m, b) |
|
3488 #define internal_free(m, mem) mspace_free(m,mem); |
|
3489 #else /* ONLY_MSPACES */ |
|
3490 #if MSPACES |
|
3491 #define internal_malloc(m, b)\ |
|
3492 (m == gm)? dlmalloc(b) : mspace_malloc(m, b) |
|
3493 #define internal_free(m, mem)\ |
|
3494 if (m == gm) dlfree(mem); else mspace_free(m,mem); |
|
3495 #else /* MSPACES */ |
|
3496 #define internal_malloc(m, b) dlmalloc(b) |
|
3497 #define internal_free(m, mem) dlfree(mem) |
|
3498 #endif /* MSPACES */ |
|
3499 #endif /* ONLY_MSPACES */ |
|
3500 |
|
3501 /* ----------------------- Direct-mmapping chunks ----------------------- */ |
|
3502 |
|
3503 /* |
|
3504 Directly mmapped chunks are set up with an offset to the start of |
|
3505 the mmapped region stored in the prev_foot field of the chunk. This |
|
3506 allows reconstruction of the required argument to MUNMAP when freed, |
|
3507 and also allows adjustment of the returned chunk to meet alignment |
|
3508 requirements (especially in memalign). There is also enough space |
|
3509 allocated to hold a fake next chunk of size SIZE_T_SIZE to maintain |
|
3510 the PINUSE bit so frees can be checked. |
|
3511 */ |
|
3512 |
|
3513 /* Malloc using mmap */ |
|
3514 static void* mmap_alloc(mstate m, size_t nb) { |
|
3515 size_t mmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK); |
|
3516 if (mmsize > nb) { /* Check for wrap around 0 */ |
|
3517 char* mm = (char*)(DIRECT_MMAP(mmsize)); |
|
3518 if (mm != CMFAIL) { |
|
3519 size_t offset = align_offset(chunk2mem(mm)); |
|
3520 size_t psize = mmsize - offset - MMAP_FOOT_PAD; |
|
3521 mchunkptr p = (mchunkptr)(mm + offset); |
|
3522 p->prev_foot = offset | IS_MMAPPED_BIT; |
|
3523 (p)->head = (psize|CINUSE_BIT); |
|
3524 mark_inuse_foot(m, p, psize); |
|
3525 chunk_plus_offset(p, psize)->head = FENCEPOST_HEAD; |
|
3526 chunk_plus_offset(p, psize+SIZE_T_SIZE)->head = 0; |
|
3527 |
|
3528 if (mm < m->least_addr) |
|
3529 m->least_addr = mm; |
|
3530 if ((m->footprint += mmsize) > m->max_footprint) |
|
3531 m->max_footprint = m->footprint; |
|
3532 assert(is_aligned(chunk2mem(p))); |
|
3533 check_mmapped_chunk(m, p); |
|
3534 return chunk2mem(p); |
|
3535 } |
|
3536 } |
|
3537 return 0; |
|
3538 } |
|
3539 |
|
3540 /* Realloc using mmap */ |
|
3541 static mchunkptr mmap_resize(mstate m, mchunkptr oldp, size_t nb) { |
|
3542 size_t oldsize = chunksize(oldp); |
|
3543 if (is_small(nb)) /* Can't shrink mmap regions below small size */ |
|
3544 return 0; |
|
3545 /* Keep old chunk if big enough but not too big */ |
|
3546 if (oldsize >= nb + SIZE_T_SIZE && |
|
3547 (oldsize - nb) <= (mparams.granularity << 1)) |
|
3548 return oldp; |
|
3549 else { |
|
3550 size_t offset = oldp->prev_foot & ~IS_MMAPPED_BIT; |
|
3551 size_t oldmmsize = oldsize + offset + MMAP_FOOT_PAD; |
|
3552 size_t newmmsize = mmap_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK); |
|
3553 char* cp = (char*)CALL_MREMAP((char*)oldp - offset, |
|
3554 oldmmsize, newmmsize, 1); |
|
3555 if (cp != CMFAIL) { |
|
3556 mchunkptr newp = (mchunkptr)(cp + offset); |
|
3557 size_t psize = newmmsize - offset - MMAP_FOOT_PAD; |
|
3558 newp->head = (psize|CINUSE_BIT); |
|
3559 mark_inuse_foot(m, newp, psize); |
|
3560 chunk_plus_offset(newp, psize)->head = FENCEPOST_HEAD; |
|
3561 chunk_plus_offset(newp, psize+SIZE_T_SIZE)->head = 0; |
|
3562 |
|
3563 if (cp < m->least_addr) |
|
3564 m->least_addr = cp; |
|
3565 if ((m->footprint += newmmsize - oldmmsize) > m->max_footprint) |
|
3566 m->max_footprint = m->footprint; |
|
3567 check_mmapped_chunk(m, newp); |
|
3568 return newp; |
|
3569 } |
|
3570 } |
|
3571 return 0; |
|
3572 } |
|
3573 |
|
3574 /* -------------------------- mspace management -------------------------- */ |
|
3575 |
|
3576 /* Initialize top chunk and its size */ |
|
3577 static void init_top(mstate m, mchunkptr p, size_t psize) { |
|
3578 /* Ensure alignment */ |
|
3579 size_t offset = align_offset(chunk2mem(p)); |
|
3580 p = (mchunkptr)((char*)p + offset); |
|
3581 psize -= offset; |
|
3582 |
|
3583 m->top = p; |
|
3584 m->topsize = psize; |
|
3585 p->head = psize | PINUSE_BIT; |
|
3586 /* set size of fake trailing chunk holding overhead space only once */ |
|
3587 chunk_plus_offset(p, psize)->head = TOP_FOOT_SIZE; |
|
3588 m->trim_check = mparams.trim_threshold; /* reset on each update */ |
|
3589 } |
|
3590 |
|
3591 /* Initialize bins for a new mstate that is otherwise zeroed out */ |
|
3592 static void init_bins(mstate m) { |
|
3593 /* Establish circular links for smallbins */ |
|
3594 bindex_t i; |
|
3595 for (i = 0; i < NSMALLBINS; ++i) { |
|
3596 sbinptr bin = smallbin_at(m,i); |
|
3597 bin->fd = bin->bk = bin; |
|
3598 } |
|
3599 } |
|
3600 |
|
3601 #if PROCEED_ON_ERROR |
|
3602 |
|
3603 /* default corruption action */ |
|
3604 static void reset_on_error(mstate m) { |
|
3605 int i; |
|
3606 ++malloc_corruption_error_count; |
|
3607 /* Reinitialize fields to forget about all memory */ |
|
3608 m->smallbins = m->treebins = 0; |
|
3609 m->dvsize = m->topsize = 0; |
|
3610 m->seg.base = 0; |
|
3611 m->seg.size = 0; |
|
3612 m->seg.next = 0; |
|
3613 m->top = m->dv = 0; |
|
3614 for (i = 0; i < NTREEBINS; ++i) |
|
3615 *treebin_at(m, i) = 0; |
|
3616 init_bins(m); |
|
3617 } |
|
3618 #endif /* PROCEED_ON_ERROR */ |
|
3619 |
|
3620 /* Allocate chunk and prepend remainder with chunk in successor base. */ |
|
3621 static void* prepend_alloc(mstate m, char* newbase, char* oldbase, |
|
3622 size_t nb) { |
|
3623 mchunkptr p = align_as_chunk(newbase); |
|
3624 mchunkptr oldfirst = align_as_chunk(oldbase); |
|
3625 size_t psize = (char*)oldfirst - (char*)p; |
|
3626 mchunkptr q = chunk_plus_offset(p, nb); |
|
3627 size_t qsize = psize - nb; |
|
3628 set_size_and_pinuse_of_inuse_chunk(m, p, nb); |
|
3629 |
|
3630 assert((char*)oldfirst > (char*)q); |
|
3631 assert(pinuse(oldfirst)); |
|
3632 assert(qsize >= MIN_CHUNK_SIZE); |
|
3633 |
|
3634 /* consolidate remainder with first chunk of old base */ |
|
3635 if (oldfirst == m->top) { |
|
3636 size_t tsize = m->topsize += qsize; |
|
3637 m->top = q; |
|
3638 q->head = tsize | PINUSE_BIT; |
|
3639 check_top_chunk(m, q); |
|
3640 } |
|
3641 else if (oldfirst == m->dv) { |
|
3642 size_t dsize = m->dvsize += qsize; |
|
3643 m->dv = q; |
|
3644 set_size_and_pinuse_of_free_chunk(q, dsize); |
|
3645 } |
|
3646 else { |
|
3647 if (!cinuse(oldfirst)) { |
|
3648 size_t nsize = chunksize(oldfirst); |
|
3649 unlink_chunk(m, oldfirst, nsize); |
|
3650 oldfirst = chunk_plus_offset(oldfirst, nsize); |
|
3651 qsize += nsize; |
|
3652 } |
|
3653 set_free_with_pinuse(q, qsize, oldfirst); |
|
3654 insert_chunk(m, q, qsize); |
|
3655 check_free_chunk(m, q); |
|
3656 } |
|
3657 |
|
3658 check_malloced_chunk(m, chunk2mem(p), nb); |
|
3659 return chunk2mem(p); |
|
3660 } |
|
3661 |
|
3662 /* Add a segment to hold a new noncontiguous region */ |
|
3663 static void add_segment(mstate m, char* tbase, size_t tsize, flag_t mmapped) { |
|
3664 /* Determine locations and sizes of segment, fenceposts, old top */ |
|
3665 char* old_top = (char*)m->top; |
|
3666 msegmentptr oldsp = segment_holding(m, old_top); |
|
3667 char* old_end = oldsp->base + oldsp->size; |
|
3668 size_t ssize = pad_request(sizeof(struct malloc_segment)); |
|
3669 char* rawsp = old_end - (ssize + FOUR_SIZE_T_SIZES + CHUNK_ALIGN_MASK); |
|
3670 size_t offset = align_offset(chunk2mem(rawsp)); |
|
3671 char* asp = rawsp + offset; |
|
3672 char* csp = (asp < (old_top + MIN_CHUNK_SIZE))? old_top : asp; |
|
3673 mchunkptr sp = (mchunkptr)csp; |
|
3674 msegmentptr ss = (msegmentptr)(chunk2mem(sp)); |
|
3675 mchunkptr tnext = chunk_plus_offset(sp, ssize); |
|
3676 mchunkptr p = tnext; |
|
3677 int nfences = 0; |
|
3678 |
|
3679 /* reset top to new space */ |
|
3680 init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE); |
|
3681 |
|
3682 /* Set up segment record */ |
|
3683 assert(is_aligned(ss)); |
|
3684 set_size_and_pinuse_of_inuse_chunk(m, sp, ssize); |
|
3685 *ss = m->seg; /* Push current record */ |
|
3686 m->seg.base = tbase; |
|
3687 m->seg.size = tsize; |
|
3688 m->seg.sflags = mmapped; |
|
3689 m->seg.next = ss; |
|
3690 |
|
3691 /* Insert trailing fenceposts */ |
|
3692 for (;;) { |
|
3693 mchunkptr nextp = chunk_plus_offset(p, SIZE_T_SIZE); |
|
3694 p->head = FENCEPOST_HEAD; |
|
3695 ++nfences; |
|
3696 if ((char*)(&(nextp->head)) < old_end) |
|
3697 p = nextp; |
|
3698 else |
|
3699 break; |
|
3700 } |
|
3701 assert(nfences >= 2); |
|
3702 |
|
3703 /* Insert the rest of old top into a bin as an ordinary free chunk */ |
|
3704 if (csp != old_top) { |
|
3705 mchunkptr q = (mchunkptr)old_top; |
|
3706 size_t psize = csp - old_top; |
|
3707 mchunkptr tn = chunk_plus_offset(q, psize); |
|
3708 set_free_with_pinuse(q, psize, tn); |
|
3709 insert_chunk(m, q, psize); |
|
3710 } |
|
3711 |
|
3712 check_top_chunk(m, m->top); |
|
3713 } |
|
3714 |
|
3715 /* -------------------------- System allocation -------------------------- */ |
|
3716 |
|
3717 /* Get memory from system using MORECORE or MMAP */ |
|
3718 static void* sys_alloc(mstate m, size_t nb) { |
|
3719 char* tbase = CMFAIL; |
|
3720 size_t tsize = 0; |
|
3721 flag_t mmap_flag = 0; |
|
3722 |
|
3723 init_mparams(); |
|
3724 |
|
3725 /* Directly map large chunks */ |
|
3726 if (use_mmap(m) && nb >= mparams.mmap_threshold) { |
|
3727 void* mem = mmap_alloc(m, nb); |
|
3728 if (mem != 0) |
|
3729 return mem; |
|
3730 } |
|
3731 |
|
3732 /* |
|
3733 Try getting memory in any of three ways (in most-preferred to |
|
3734 least-preferred order): |
|
3735 1. A call to MORECORE that can normally contiguously extend memory. |
|
3736 (disabled if not MORECORE_CONTIGUOUS or not HAVE_MORECORE or |
|
3737 or main space is mmapped or a previous contiguous call failed) |
|
3738 2. A call to MMAP new space (disabled if not HAVE_MMAP). |
|
3739 Note that under the default settings, if MORECORE is unable to |
|
3740 fulfill a request, and HAVE_MMAP is true, then mmap is |
|
3741 used as a noncontiguous system allocator. This is a useful backup |
|
3742 strategy for systems with holes in address spaces -- in this case |
|
3743 sbrk cannot contiguously expand the heap, but mmap may be able to |
|
3744 find space. |
|
3745 3. A call to MORECORE that cannot usually contiguously extend memory. |
|
3746 (disabled if not HAVE_MORECORE) |
|
3747 */ |
|
3748 |
|
3749 if (MORECORE_CONTIGUOUS && !use_noncontiguous(m)) { |
|
3750 char* br = CMFAIL; |
|
3751 msegmentptr ss = (m->top == 0)? 0 : segment_holding(m, (char*)m->top); |
|
3752 size_t asize = 0; |
|
3753 ACQUIRE_MORECORE_LOCK(); |
|
3754 |
|
3755 if (ss == 0) { /* First time through or recovery */ |
|
3756 char* base = (char*)CALL_MORECORE(0); |
|
3757 if (base != CMFAIL) { |
|
3758 asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE); |
|
3759 /* Adjust to end on a page boundary */ |
|
3760 if (!is_page_aligned(base)) |
|
3761 asize += (page_align((size_t)base) - (size_t)base); |
|
3762 /* Can't call MORECORE if size is negative when treated as signed */ |
|
3763 if (asize < HALF_MAX_SIZE_T && |
|
3764 (br = (char*)(CALL_MORECORE(asize))) == base) { |
|
3765 tbase = base; |
|
3766 tsize = asize; |
|
3767 } |
|
3768 } |
|
3769 } |
|
3770 else { |
|
3771 /* Subtract out existing available top space from MORECORE request. */ |
|
3772 asize = granularity_align(nb - m->topsize + TOP_FOOT_SIZE + SIZE_T_ONE); |
|
3773 /* Use mem here only if it did continuously extend old space */ |
|
3774 if (asize < HALF_MAX_SIZE_T && |
|
3775 (br = (char*)(CALL_MORECORE(asize))) == ss->base+ss->size) { |
|
3776 tbase = br; |
|
3777 tsize = asize; |
|
3778 } |
|
3779 } |
|
3780 |
|
3781 if (tbase == CMFAIL) { /* Cope with partial failure */ |
|
3782 if (br != CMFAIL) { /* Try to use/extend the space we did get */ |
|
3783 if (asize < HALF_MAX_SIZE_T && |
|
3784 asize < nb + TOP_FOOT_SIZE + SIZE_T_ONE) { |
|
3785 size_t esize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE - asize); |
|
3786 if (esize < HALF_MAX_SIZE_T) { |
|
3787 char* end = (char*)CALL_MORECORE(esize); |
|
3788 if (end != CMFAIL) |
|
3789 asize += esize; |
|
3790 else { /* Can't use; try to release */ |
|
3791 (void) CALL_MORECORE(-asize); |
|
3792 br = CMFAIL; |
|
3793 } |
|
3794 } |
|
3795 } |
|
3796 } |
|
3797 if (br != CMFAIL) { /* Use the space we did get */ |
|
3798 tbase = br; |
|
3799 tsize = asize; |
|
3800 } |
|
3801 else |
|
3802 disable_contiguous(m); /* Don't try contiguous path in the future */ |
|
3803 } |
|
3804 |
|
3805 RELEASE_MORECORE_LOCK(); |
|
3806 } |
|
3807 |
|
3808 if (HAVE_MMAP && tbase == CMFAIL) { /* Try MMAP */ |
|
3809 size_t req = nb + TOP_FOOT_SIZE + SIZE_T_ONE; |
|
3810 size_t rsize = granularity_align(req); |
|
3811 if (rsize > nb) { /* Fail if wraps around zero */ |
|
3812 char* mp = (char*)(CALL_MMAP(rsize)); |
|
3813 if (mp != CMFAIL) { |
|
3814 tbase = mp; |
|
3815 tsize = rsize; |
|
3816 mmap_flag = IS_MMAPPED_BIT; |
|
3817 } |
|
3818 } |
|
3819 } |
|
3820 |
|
3821 if (HAVE_MORECORE && tbase == CMFAIL) { /* Try noncontiguous MORECORE */ |
|
3822 size_t asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE); |
|
3823 if (asize < HALF_MAX_SIZE_T) { |
|
3824 char* br = CMFAIL; |
|
3825 char* end = CMFAIL; |
|
3826 ACQUIRE_MORECORE_LOCK(); |
|
3827 br = (char*)(CALL_MORECORE(asize)); |
|
3828 end = (char*)(CALL_MORECORE(0)); |
|
3829 RELEASE_MORECORE_LOCK(); |
|
3830 if (br != CMFAIL && end != CMFAIL && br < end) { |
|
3831 size_t ssize = end - br; |
|
3832 if (ssize > nb + TOP_FOOT_SIZE) { |
|
3833 tbase = br; |
|
3834 tsize = ssize; |
|
3835 } |
|
3836 } |
|
3837 } |
|
3838 } |
|
3839 |
|
3840 if (tbase != CMFAIL) { |
|
3841 |
|
3842 if ((m->footprint += tsize) > m->max_footprint) |
|
3843 m->max_footprint = m->footprint; |
|
3844 |
|
3845 if (!is_initialized(m)) { /* first-time initialization */ |
|
3846 m->seg.base = m->least_addr = tbase; |
|
3847 m->seg.size = tsize; |
|
3848 m->seg.sflags = mmap_flag; |
|
3849 m->magic = mparams.magic; |
|
3850 m->release_checks = MAX_RELEASE_CHECK_RATE; |
|
3851 init_bins(m); |
|
3852 #if !ONLY_MSPACES |
|
3853 if (is_global(m)) |
|
3854 init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE); |
|
3855 else |
|
3856 #endif |
|
3857 { |
|
3858 /* Offset top by embedded malloc_state */ |
|
3859 mchunkptr mn = next_chunk(mem2chunk(m)); |
|
3860 init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) -TOP_FOOT_SIZE); |
|
3861 } |
|
3862 } |
|
3863 |
|
3864 else { |
|
3865 /* Try to merge with an existing segment */ |
|
3866 msegmentptr sp = &m->seg; |
|
3867 /* Only consider most recent segment if traversal suppressed */ |
|
3868 while (sp != 0 && tbase != sp->base + sp->size) |
|
3869 sp = (NO_SEGMENT_TRAVERSAL) ? 0 : sp->next; |
|
3870 if (sp != 0 && |
|
3871 !is_extern_segment(sp) && |
|
3872 (sp->sflags & IS_MMAPPED_BIT) == mmap_flag && |
|
3873 segment_holds(sp, m->top)) { /* append */ |
|
3874 sp->size += tsize; |
|
3875 init_top(m, m->top, m->topsize + tsize); |
|
3876 } |
|
3877 else { |
|
3878 if (tbase < m->least_addr) |
|
3879 m->least_addr = tbase; |
|
3880 sp = &m->seg; |
|
3881 while (sp != 0 && sp->base != tbase + tsize) |
|
3882 sp = (NO_SEGMENT_TRAVERSAL) ? 0 : sp->next; |
|
3883 if (sp != 0 && |
|
3884 !is_extern_segment(sp) && |
|
3885 (sp->sflags & IS_MMAPPED_BIT) == mmap_flag) { |
|
3886 char* oldbase = sp->base; |
|
3887 sp->base = tbase; |
|
3888 sp->size += tsize; |
|
3889 return prepend_alloc(m, tbase, oldbase, nb); |
|
3890 } |
|
3891 else |
|
3892 add_segment(m, tbase, tsize, mmap_flag); |
|
3893 } |
|
3894 } |
|
3895 |
|
3896 if (nb < m->topsize) { /* Allocate from new or extended top space */ |
|
3897 size_t rsize = m->topsize -= nb; |
|
3898 mchunkptr p = m->top; |
|
3899 mchunkptr r = m->top = chunk_plus_offset(p, nb); |
|
3900 r->head = rsize | PINUSE_BIT; |
|
3901 set_size_and_pinuse_of_inuse_chunk(m, p, nb); |
|
3902 check_top_chunk(m, m->top); |
|
3903 check_malloced_chunk(m, chunk2mem(p), nb); |
|
3904 return chunk2mem(p); |
|
3905 } |
|
3906 } |
|
3907 |
|
3908 MALLOC_FAILURE_ACTION; |
|
3909 return 0; |
|
3910 } |
|
3911 |
|
3912 /* ----------------------- system deallocation -------------------------- */ |
|
3913 |
|
3914 /* Unmap and unlink any mmapped segments that don't contain used chunks */ |
|
3915 static size_t release_unused_segments(mstate m) { |
|
3916 size_t released = 0; |
|
3917 int nsegs = 0; |
|
3918 msegmentptr pred = &m->seg; |
|
3919 msegmentptr sp = pred->next; |
|
3920 while (sp != 0) { |
|
3921 char* base = sp->base; |
|
3922 size_t size = sp->size; |
|
3923 msegmentptr next = sp->next; |
|
3924 ++nsegs; |
|
3925 if (is_mmapped_segment(sp) && !is_extern_segment(sp)) { |
|
3926 mchunkptr p = align_as_chunk(base); |
|
3927 size_t psize = chunksize(p); |
|
3928 /* Can unmap if first chunk holds entire segment and not pinned */ |
|
3929 if (!cinuse(p) && (char*)p + psize >= base + size - TOP_FOOT_SIZE) { |
|
3930 tchunkptr tp = (tchunkptr)p; |
|
3931 assert(segment_holds(sp, (char*)sp)); |
|
3932 if (p == m->dv) { |
|
3933 m->dv = 0; |
|
3934 m->dvsize = 0; |
|
3935 } |
|
3936 else { |
|
3937 unlink_large_chunk(m, tp); |
|
3938 } |
|
3939 if (CALL_MUNMAP(base, size) == 0) { |
|
3940 released += size; |
|
3941 m->footprint -= size; |
|
3942 /* unlink obsoleted record */ |
|
3943 sp = pred; |
|
3944 sp->next = next; |
|
3945 } |
|
3946 else { /* back out if cannot unmap */ |
|
3947 insert_large_chunk(m, tp, psize); |
|
3948 } |
|
3949 } |
|
3950 } |
|
3951 if (NO_SEGMENT_TRAVERSAL) /* scan only first segment */ |
|
3952 break; |
|
3953 pred = sp; |
|
3954 sp = next; |
|
3955 } |
|
3956 /* Reset check counter */ |
|
3957 m->release_checks = ((nsegs > MAX_RELEASE_CHECK_RATE)? |
|
3958 nsegs : MAX_RELEASE_CHECK_RATE); |
|
3959 return released; |
|
3960 } |
|
3961 |
|
3962 static int sys_trim(mstate m, size_t pad) { |
|
3963 size_t released = 0; |
|
3964 if (pad < MAX_REQUEST && is_initialized(m)) { |
|
3965 pad += TOP_FOOT_SIZE; /* ensure enough room for segment overhead */ |
|
3966 |
|
3967 if (m->topsize > pad) { |
|
3968 /* Shrink top space in granularity-size units, keeping at least one */ |
|
3969 size_t unit = mparams.granularity; |
|
3970 size_t extra = ((m->topsize - pad + (unit - SIZE_T_ONE)) / unit - |
|
3971 SIZE_T_ONE) * unit; |
|
3972 msegmentptr sp = segment_holding(m, (char*)m->top); |
|
3973 |
|
3974 if (!is_extern_segment(sp)) { |
|
3975 if (is_mmapped_segment(sp)) { |
|
3976 if (HAVE_MMAP && |
|
3977 sp->size >= extra && |
|
3978 !has_segment_link(m, sp)) { /* can't shrink if pinned */ |
|
3979 size_t newsize = sp->size - extra; |
|
3980 /* Prefer mremap, fall back to munmap */ |
|
3981 if ((CALL_MREMAP(sp->base, sp->size, newsize, 0) != MFAIL) || |
|
3982 (CALL_MUNMAP(sp->base + newsize, extra) == 0)) { |
|
3983 released = extra; |
|
3984 } |
|
3985 } |
|
3986 } |
|
3987 else if (HAVE_MORECORE) { |
|
3988 if (extra >= HALF_MAX_SIZE_T) /* Avoid wrapping negative */ |
|
3989 extra = (HALF_MAX_SIZE_T) + SIZE_T_ONE - unit; |
|
3990 ACQUIRE_MORECORE_LOCK(); |
|
3991 { |
|
3992 /* Make sure end of memory is where we last set it. */ |
|
3993 char* old_br = (char*)(CALL_MORECORE(0)); |
|
3994 if (old_br == sp->base + sp->size) { |
|
3995 char* rel_br = (char*)(CALL_MORECORE(-extra)); |
|
3996 char* new_br = (char*)(CALL_MORECORE(0)); |
|
3997 if (rel_br != CMFAIL && new_br < old_br) |
|
3998 released = old_br - new_br; |
|
3999 } |
|
4000 } |
|
4001 RELEASE_MORECORE_LOCK(); |
|
4002 } |
|
4003 } |
|
4004 |
|
4005 if (released != 0) { |
|
4006 sp->size -= released; |
|
4007 m->footprint -= released; |
|
4008 init_top(m, m->top, m->topsize - released); |
|
4009 check_top_chunk(m, m->top); |
|
4010 } |
|
4011 } |
|
4012 |
|
4013 /* Unmap any unused mmapped segments */ |
|
4014 if (HAVE_MMAP) |
|
4015 released += release_unused_segments(m); |
|
4016 |
|
4017 /* On failure, disable autotrim to avoid repeated failed future calls */ |
|
4018 if (released == 0 && m->topsize > m->trim_check) |
|
4019 m->trim_check = MAX_SIZE_T; |
|
4020 } |
|
4021 |
|
4022 return (released != 0)? 1 : 0; |
|
4023 } |
|
4024 |
|
4025 /* ---------------------------- malloc support --------------------------- */ |
|
4026 |
|
4027 /* allocate a large request from the best fitting chunk in a treebin */ |
|
4028 static void* tmalloc_large(mstate m, size_t nb) { |
|
4029 tchunkptr v = 0; |
|
4030 size_t rsize = -nb; /* Unsigned negation */ |
|
4031 tchunkptr t; |
|
4032 bindex_t idx; |
|
4033 compute_tree_index(nb, idx); |
|
4034 |
|
4035 if ((t = *treebin_at(m, idx)) != 0) { |
|
4036 /* Traverse tree for this bin looking for node with size == nb */ |
|
4037 size_t sizebits = nb << leftshift_for_tree_index(idx); |
|
4038 tchunkptr rst = 0; /* The deepest untaken right subtree */ |
|
4039 for (;;) { |
|
4040 tchunkptr rt; |
|
4041 size_t trem = chunksize(t) - nb; |
|
4042 if (trem < rsize) { |
|
4043 v = t; |
|
4044 if ((rsize = trem) == 0) |
|
4045 break; |
|
4046 } |
|
4047 rt = t->child[1]; |
|
4048 t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]; |
|
4049 if (rt != 0 && rt != t) |
|
4050 rst = rt; |
|
4051 if (t == 0) { |
|
4052 t = rst; /* set t to least subtree holding sizes > nb */ |
|
4053 break; |
|
4054 } |
|
4055 sizebits <<= 1; |
|
4056 } |
|
4057 } |
|
4058 |
|
4059 if (t == 0 && v == 0) { /* set t to root of next non-empty treebin */ |
|
4060 binmap_t leftbits = left_bits(idx2bit(idx)) & m->treemap; |
|
4061 if (leftbits != 0) { |
|
4062 bindex_t i; |
|
4063 binmap_t leastbit = least_bit(leftbits); |
|
4064 compute_bit2idx(leastbit, i); |
|
4065 t = *treebin_at(m, i); |
|
4066 } |
|
4067 } |
|
4068 |
|
4069 while (t != 0) { /* find smallest of tree or subtree */ |
|
4070 size_t trem = chunksize(t) - nb; |
|
4071 if (trem < rsize) { |
|
4072 rsize = trem; |
|
4073 v = t; |
|
4074 } |
|
4075 t = leftmost_child(t); |
|
4076 } |
|
4077 |
|
4078 /* If dv is a better fit, return 0 so malloc will use it */ |
|
4079 if (v != 0 && rsize < (size_t)(m->dvsize - nb)) { |
|
4080 if (RTCHECK(ok_address(m, v))) { /* split */ |
|
4081 mchunkptr r = chunk_plus_offset(v, nb); |
|
4082 assert(chunksize(v) == rsize + nb); |
|
4083 if (RTCHECK(ok_next(v, r))) { |
|
4084 unlink_large_chunk(m, v); |
|
4085 if (rsize < MIN_CHUNK_SIZE) |
|
4086 set_inuse_and_pinuse(m, v, (rsize + nb)); |
|
4087 else { |
|
4088 set_size_and_pinuse_of_inuse_chunk(m, v, nb); |
|
4089 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4090 insert_chunk(m, r, rsize); |
|
4091 } |
|
4092 return chunk2mem(v); |
|
4093 } |
|
4094 } |
|
4095 CORRUPTION_ERROR_ACTION(m); |
|
4096 } |
|
4097 return 0; |
|
4098 } |
|
4099 |
|
4100 /* allocate a small request from the best fitting chunk in a treebin */ |
|
4101 static void* tmalloc_small(mstate m, size_t nb) { |
|
4102 tchunkptr t, v; |
|
4103 size_t rsize; |
|
4104 bindex_t i; |
|
4105 binmap_t leastbit = least_bit(m->treemap); |
|
4106 compute_bit2idx(leastbit, i); |
|
4107 |
|
4108 v = t = *treebin_at(m, i); |
|
4109 rsize = chunksize(t) - nb; |
|
4110 |
|
4111 while ((t = leftmost_child(t)) != 0) { |
|
4112 size_t trem = chunksize(t) - nb; |
|
4113 if (trem < rsize) { |
|
4114 rsize = trem; |
|
4115 v = t; |
|
4116 } |
|
4117 } |
|
4118 |
|
4119 if (RTCHECK(ok_address(m, v))) { |
|
4120 mchunkptr r = chunk_plus_offset(v, nb); |
|
4121 assert(chunksize(v) == rsize + nb); |
|
4122 if (RTCHECK(ok_next(v, r))) { |
|
4123 unlink_large_chunk(m, v); |
|
4124 if (rsize < MIN_CHUNK_SIZE) |
|
4125 set_inuse_and_pinuse(m, v, (rsize + nb)); |
|
4126 else { |
|
4127 set_size_and_pinuse_of_inuse_chunk(m, v, nb); |
|
4128 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4129 replace_dv(m, r, rsize); |
|
4130 } |
|
4131 return chunk2mem(v); |
|
4132 } |
|
4133 } |
|
4134 |
|
4135 CORRUPTION_ERROR_ACTION(m); |
|
4136 return 0; |
|
4137 } |
|
4138 |
|
4139 /* --------------------------- realloc support --------------------------- */ |
|
4140 |
|
4141 static void* internal_realloc(mstate m, void* oldmem, size_t bytes) { |
|
4142 if (bytes >= MAX_REQUEST) { |
|
4143 MALLOC_FAILURE_ACTION; |
|
4144 return 0; |
|
4145 } |
|
4146 if (!PREACTION(m)) { |
|
4147 mchunkptr oldp = mem2chunk(oldmem); |
|
4148 size_t oldsize = chunksize(oldp); |
|
4149 mchunkptr next = chunk_plus_offset(oldp, oldsize); |
|
4150 mchunkptr newp = 0; |
|
4151 void* extra = 0; |
|
4152 |
|
4153 /* Try to either shrink or extend into top. Else malloc-copy-free */ |
|
4154 |
|
4155 if (RTCHECK(ok_address(m, oldp) && ok_cinuse(oldp) && |
|
4156 ok_next(oldp, next) && ok_pinuse(next))) { |
|
4157 size_t nb = request2size(bytes); |
|
4158 if (is_mmapped(oldp)) |
|
4159 newp = mmap_resize(m, oldp, nb); |
|
4160 else if (oldsize >= nb) { /* already big enough */ |
|
4161 size_t rsize = oldsize - nb; |
|
4162 newp = oldp; |
|
4163 if (rsize >= MIN_CHUNK_SIZE) { |
|
4164 mchunkptr remainder = chunk_plus_offset(newp, nb); |
|
4165 set_inuse(m, newp, nb); |
|
4166 set_inuse(m, remainder, rsize); |
|
4167 extra = chunk2mem(remainder); |
|
4168 } |
|
4169 } |
|
4170 else if (next == m->top && oldsize + m->topsize > nb) { |
|
4171 /* Expand into top */ |
|
4172 size_t newsize = oldsize + m->topsize; |
|
4173 size_t newtopsize = newsize - nb; |
|
4174 mchunkptr newtop = chunk_plus_offset(oldp, nb); |
|
4175 set_inuse(m, oldp, nb); |
|
4176 newtop->head = newtopsize |PINUSE_BIT; |
|
4177 m->top = newtop; |
|
4178 m->topsize = newtopsize; |
|
4179 newp = oldp; |
|
4180 } |
|
4181 } |
|
4182 else { |
|
4183 USAGE_ERROR_ACTION(m, oldmem); |
|
4184 POSTACTION(m); |
|
4185 return 0; |
|
4186 } |
|
4187 |
|
4188 POSTACTION(m); |
|
4189 |
|
4190 if (newp != 0) { |
|
4191 if (extra != 0) { |
|
4192 internal_free(m, extra); |
|
4193 } |
|
4194 check_inuse_chunk(m, newp); |
|
4195 return chunk2mem(newp); |
|
4196 } |
|
4197 else { |
|
4198 void* newmem = internal_malloc(m, bytes); |
|
4199 if (newmem != 0) { |
|
4200 size_t oc = oldsize - overhead_for(oldp); |
|
4201 memcpy(newmem, oldmem, (oc < bytes)? oc : bytes); |
|
4202 internal_free(m, oldmem); |
|
4203 } |
|
4204 return newmem; |
|
4205 } |
|
4206 } |
|
4207 return 0; |
|
4208 } |
|
4209 |
|
4210 /* --------------------------- memalign support -------------------------- */ |
|
4211 |
|
4212 static void* internal_memalign(mstate m, size_t alignment, size_t bytes) { |
|
4213 if (alignment <= MALLOC_ALIGNMENT) /* Can just use malloc */ |
|
4214 return internal_malloc(m, bytes); |
|
4215 if (alignment < MIN_CHUNK_SIZE) /* must be at least a minimum chunk size */ |
|
4216 alignment = MIN_CHUNK_SIZE; |
|
4217 if ((alignment & (alignment-SIZE_T_ONE)) != 0) {/* Ensure a power of 2 */ |
|
4218 size_t a = MALLOC_ALIGNMENT << 1; |
|
4219 while (a < alignment) a <<= 1; |
|
4220 alignment = a; |
|
4221 } |
|
4222 |
|
4223 if (bytes >= MAX_REQUEST - alignment) { |
|
4224 if (m != 0) { /* Test isn't needed but avoids compiler warning */ |
|
4225 MALLOC_FAILURE_ACTION; |
|
4226 } |
|
4227 } |
|
4228 else { |
|
4229 size_t nb = request2size(bytes); |
|
4230 size_t req = nb + alignment + MIN_CHUNK_SIZE - CHUNK_OVERHEAD; |
|
4231 char* mem = (char*)internal_malloc(m, req); |
|
4232 if (mem != 0) { |
|
4233 void* leader = 0; |
|
4234 void* trailer = 0; |
|
4235 mchunkptr p = mem2chunk(mem); |
|
4236 |
|
4237 if (PREACTION(m)) return 0; |
|
4238 if ((((size_t)(mem)) % alignment) != 0) { /* misaligned */ |
|
4239 /* |
|
4240 Find an aligned spot inside chunk. Since we need to give |
|
4241 back leading space in a chunk of at least MIN_CHUNK_SIZE, if |
|
4242 the first calculation places us at a spot with less than |
|
4243 MIN_CHUNK_SIZE leader, we can move to the next aligned spot. |
|
4244 We've allocated enough total room so that this is always |
|
4245 possible. |
|
4246 */ |
|
4247 char* br = (char*)mem2chunk((size_t)(((size_t)(mem + |
|
4248 alignment - |
|
4249 SIZE_T_ONE)) & |
|
4250 -alignment)); |
|
4251 char* pos = ((size_t)(br - (char*)(p)) >= MIN_CHUNK_SIZE)? |
|
4252 br : br+alignment; |
|
4253 mchunkptr newp = (mchunkptr)pos; |
|
4254 size_t leadsize = pos - (char*)(p); |
|
4255 size_t newsize = chunksize(p) - leadsize; |
|
4256 |
|
4257 if (is_mmapped(p)) { /* For mmapped chunks, just adjust offset */ |
|
4258 newp->prev_foot = p->prev_foot + leadsize; |
|
4259 newp->head = (newsize|CINUSE_BIT); |
|
4260 } |
|
4261 else { /* Otherwise, give back leader, use the rest */ |
|
4262 set_inuse(m, newp, newsize); |
|
4263 set_inuse(m, p, leadsize); |
|
4264 leader = chunk2mem(p); |
|
4265 } |
|
4266 p = newp; |
|
4267 } |
|
4268 |
|
4269 /* Give back spare room at the end */ |
|
4270 if (!is_mmapped(p)) { |
|
4271 size_t size = chunksize(p); |
|
4272 if (size > nb + MIN_CHUNK_SIZE) { |
|
4273 size_t remainder_size = size - nb; |
|
4274 mchunkptr remainder = chunk_plus_offset(p, nb); |
|
4275 set_inuse(m, p, nb); |
|
4276 set_inuse(m, remainder, remainder_size); |
|
4277 trailer = chunk2mem(remainder); |
|
4278 } |
|
4279 } |
|
4280 |
|
4281 assert (chunksize(p) >= nb); |
|
4282 assert((((size_t)(chunk2mem(p))) % alignment) == 0); |
|
4283 check_inuse_chunk(m, p); |
|
4284 POSTACTION(m); |
|
4285 if (leader != 0) { |
|
4286 internal_free(m, leader); |
|
4287 } |
|
4288 if (trailer != 0) { |
|
4289 internal_free(m, trailer); |
|
4290 } |
|
4291 return chunk2mem(p); |
|
4292 } |
|
4293 } |
|
4294 return 0; |
|
4295 } |
|
4296 |
|
4297 /* ------------------------ comalloc/coalloc support --------------------- */ |
|
4298 |
|
4299 static void** ialloc(mstate m, |
|
4300 size_t n_elements, |
|
4301 size_t* sizes, |
|
4302 int opts, |
|
4303 void* chunks[]) { |
|
4304 /* |
|
4305 This provides common support for independent_X routines, handling |
|
4306 all of the combinations that can result. |
|
4307 |
|
4308 The opts arg has: |
|
4309 bit 0 set if all elements are same size (using sizes[0]) |
|
4310 bit 1 set if elements should be zeroed |
|
4311 */ |
|
4312 |
|
4313 size_t element_size; /* chunksize of each element, if all same */ |
|
4314 size_t contents_size; /* total size of elements */ |
|
4315 size_t array_size; /* request size of pointer array */ |
|
4316 void* mem; /* malloced aggregate space */ |
|
4317 mchunkptr p; /* corresponding chunk */ |
|
4318 size_t remainder_size; /* remaining bytes while splitting */ |
|
4319 void** marray; /* either "chunks" or malloced ptr array */ |
|
4320 mchunkptr array_chunk; /* chunk for malloced ptr array */ |
|
4321 flag_t was_enabled; /* to disable mmap */ |
|
4322 size_t size; |
|
4323 size_t i; |
|
4324 |
|
4325 /* compute array length, if needed */ |
|
4326 if (chunks != 0) { |
|
4327 if (n_elements == 0) |
|
4328 return chunks; /* nothing to do */ |
|
4329 marray = chunks; |
|
4330 array_size = 0; |
|
4331 } |
|
4332 else { |
|
4333 /* if empty req, must still return chunk representing empty array */ |
|
4334 if (n_elements == 0) |
|
4335 return (void**)internal_malloc(m, 0); |
|
4336 marray = 0; |
|
4337 array_size = request2size(n_elements * (sizeof(void*))); |
|
4338 } |
|
4339 |
|
4340 /* compute total element size */ |
|
4341 if (opts & 0x1) { /* all-same-size */ |
|
4342 element_size = request2size(*sizes); |
|
4343 contents_size = n_elements * element_size; |
|
4344 } |
|
4345 else { /* add up all the sizes */ |
|
4346 element_size = 0; |
|
4347 contents_size = 0; |
|
4348 for (i = 0; i != n_elements; ++i) |
|
4349 contents_size += request2size(sizes[i]); |
|
4350 } |
|
4351 |
|
4352 size = contents_size + array_size; |
|
4353 |
|
4354 /* |
|
4355 Allocate the aggregate chunk. First disable direct-mmapping so |
|
4356 malloc won't use it, since we would not be able to later |
|
4357 free/realloc space internal to a segregated mmap region. |
|
4358 */ |
|
4359 was_enabled = use_mmap(m); |
|
4360 disable_mmap(m); |
|
4361 mem = internal_malloc(m, size - CHUNK_OVERHEAD); |
|
4362 if (was_enabled) |
|
4363 enable_mmap(m); |
|
4364 if (mem == 0) |
|
4365 return 0; |
|
4366 |
|
4367 if (PREACTION(m)) return 0; |
|
4368 p = mem2chunk(mem); |
|
4369 remainder_size = chunksize(p); |
|
4370 |
|
4371 assert(!is_mmapped(p)); |
|
4372 |
|
4373 if (opts & 0x2) { /* optionally clear the elements */ |
|
4374 memset((size_t*)mem, 0, remainder_size - SIZE_T_SIZE - array_size); |
|
4375 } |
|
4376 |
|
4377 /* If not provided, allocate the pointer array as final part of chunk */ |
|
4378 if (marray == 0) { |
|
4379 size_t array_chunk_size; |
|
4380 array_chunk = chunk_plus_offset(p, contents_size); |
|
4381 array_chunk_size = remainder_size - contents_size; |
|
4382 marray = (void**) (chunk2mem(array_chunk)); |
|
4383 set_size_and_pinuse_of_inuse_chunk(m, array_chunk, array_chunk_size); |
|
4384 remainder_size = contents_size; |
|
4385 } |
|
4386 |
|
4387 /* split out elements */ |
|
4388 for (i = 0; ; ++i) { |
|
4389 marray[i] = chunk2mem(p); |
|
4390 if (i != n_elements-1) { |
|
4391 if (element_size != 0) |
|
4392 size = element_size; |
|
4393 else |
|
4394 size = request2size(sizes[i]); |
|
4395 remainder_size -= size; |
|
4396 set_size_and_pinuse_of_inuse_chunk(m, p, size); |
|
4397 p = chunk_plus_offset(p, size); |
|
4398 } |
|
4399 else { /* the final element absorbs any overallocation slop */ |
|
4400 set_size_and_pinuse_of_inuse_chunk(m, p, remainder_size); |
|
4401 break; |
|
4402 } |
|
4403 } |
|
4404 |
|
4405 #if DEBUG |
|
4406 if (marray != chunks) { |
|
4407 /* final element must have exactly exhausted chunk */ |
|
4408 if (element_size != 0) { |
|
4409 assert(remainder_size == element_size); |
|
4410 } |
|
4411 else { |
|
4412 assert(remainder_size == request2size(sizes[i])); |
|
4413 } |
|
4414 check_inuse_chunk(m, mem2chunk(marray)); |
|
4415 } |
|
4416 for (i = 0; i != n_elements; ++i) |
|
4417 check_inuse_chunk(m, mem2chunk(marray[i])); |
|
4418 |
|
4419 #endif /* DEBUG */ |
|
4420 |
|
4421 POSTACTION(m); |
|
4422 return marray; |
|
4423 } |
|
4424 |
|
4425 |
|
4426 /* -------------------------- public routines ---------------------------- */ |
|
4427 |
|
4428 #if !ONLY_MSPACES |
|
4429 |
|
4430 void* dlmalloc(size_t bytes) { |
|
4431 /* |
|
4432 Basic algorithm: |
|
4433 If a small request (< 256 bytes minus per-chunk overhead): |
|
4434 1. If one exists, use a remainderless chunk in associated smallbin. |
|
4435 (Remainderless means that there are too few excess bytes to |
|
4436 represent as a chunk.) |
|
4437 2. If it is big enough, use the dv chunk, which is normally the |
|
4438 chunk adjacent to the one used for the most recent small request. |
|
4439 3. If one exists, split the smallest available chunk in a bin, |
|
4440 saving remainder in dv. |
|
4441 4. If it is big enough, use the top chunk. |
|
4442 5. If available, get memory from system and use it |
|
4443 Otherwise, for a large request: |
|
4444 1. Find the smallest available binned chunk that fits, and use it |
|
4445 if it is better fitting than dv chunk, splitting if necessary. |
|
4446 2. If better fitting than any binned chunk, use the dv chunk. |
|
4447 3. If it is big enough, use the top chunk. |
|
4448 4. If request size >= mmap threshold, try to directly mmap this chunk. |
|
4449 5. If available, get memory from system and use it |
|
4450 |
|
4451 The ugly goto's here ensure that postaction occurs along all paths. |
|
4452 */ |
|
4453 |
|
4454 if (!PREACTION(gm)) { |
|
4455 void* mem; |
|
4456 size_t nb; |
|
4457 if (bytes <= MAX_SMALL_REQUEST) { |
|
4458 bindex_t idx; |
|
4459 binmap_t smallbits; |
|
4460 nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes); |
|
4461 idx = small_index(nb); |
|
4462 smallbits = gm->smallmap >> idx; |
|
4463 |
|
4464 if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */ |
|
4465 mchunkptr b, p; |
|
4466 idx += ~smallbits & 1; /* Uses next bin if idx empty */ |
|
4467 b = smallbin_at(gm, idx); |
|
4468 p = b->fd; |
|
4469 assert(chunksize(p) == small_index2size(idx)); |
|
4470 unlink_first_small_chunk(gm, b, p, idx); |
|
4471 set_inuse_and_pinuse(gm, p, small_index2size(idx)); |
|
4472 mem = chunk2mem(p); |
|
4473 check_malloced_chunk(gm, mem, nb); |
|
4474 goto postaction; |
|
4475 } |
|
4476 |
|
4477 else if (nb > gm->dvsize) { |
|
4478 if (smallbits != 0) { /* Use chunk in next nonempty smallbin */ |
|
4479 mchunkptr b, p, r; |
|
4480 size_t rsize; |
|
4481 bindex_t i; |
|
4482 binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx)); |
|
4483 binmap_t leastbit = least_bit(leftbits); |
|
4484 compute_bit2idx(leastbit, i); |
|
4485 b = smallbin_at(gm, i); |
|
4486 p = b->fd; |
|
4487 assert(chunksize(p) == small_index2size(i)); |
|
4488 unlink_first_small_chunk(gm, b, p, i); |
|
4489 rsize = small_index2size(i) - nb; |
|
4490 /* Fit here cannot be remainderless if 4byte sizes */ |
|
4491 if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE) |
|
4492 set_inuse_and_pinuse(gm, p, small_index2size(i)); |
|
4493 else { |
|
4494 set_size_and_pinuse_of_inuse_chunk(gm, p, nb); |
|
4495 r = chunk_plus_offset(p, nb); |
|
4496 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4497 replace_dv(gm, r, rsize); |
|
4498 } |
|
4499 mem = chunk2mem(p); |
|
4500 check_malloced_chunk(gm, mem, nb); |
|
4501 goto postaction; |
|
4502 } |
|
4503 |
|
4504 else if (gm->treemap != 0 && (mem = tmalloc_small(gm, nb)) != 0) { |
|
4505 check_malloced_chunk(gm, mem, nb); |
|
4506 goto postaction; |
|
4507 } |
|
4508 } |
|
4509 } |
|
4510 else if (bytes >= MAX_REQUEST) |
|
4511 nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */ |
|
4512 else { |
|
4513 nb = pad_request(bytes); |
|
4514 if (gm->treemap != 0 && (mem = tmalloc_large(gm, nb)) != 0) { |
|
4515 check_malloced_chunk(gm, mem, nb); |
|
4516 goto postaction; |
|
4517 } |
|
4518 } |
|
4519 |
|
4520 if (nb <= gm->dvsize) { |
|
4521 size_t rsize = gm->dvsize - nb; |
|
4522 mchunkptr p = gm->dv; |
|
4523 if (rsize >= MIN_CHUNK_SIZE) { /* split dv */ |
|
4524 mchunkptr r = gm->dv = chunk_plus_offset(p, nb); |
|
4525 gm->dvsize = rsize; |
|
4526 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4527 set_size_and_pinuse_of_inuse_chunk(gm, p, nb); |
|
4528 } |
|
4529 else { /* exhaust dv */ |
|
4530 size_t dvs = gm->dvsize; |
|
4531 gm->dvsize = 0; |
|
4532 gm->dv = 0; |
|
4533 set_inuse_and_pinuse(gm, p, dvs); |
|
4534 } |
|
4535 mem = chunk2mem(p); |
|
4536 check_malloced_chunk(gm, mem, nb); |
|
4537 goto postaction; |
|
4538 } |
|
4539 |
|
4540 else if (nb < gm->topsize) { /* Split top */ |
|
4541 size_t rsize = gm->topsize -= nb; |
|
4542 mchunkptr p = gm->top; |
|
4543 mchunkptr r = gm->top = chunk_plus_offset(p, nb); |
|
4544 r->head = rsize | PINUSE_BIT; |
|
4545 set_size_and_pinuse_of_inuse_chunk(gm, p, nb); |
|
4546 mem = chunk2mem(p); |
|
4547 check_top_chunk(gm, gm->top); |
|
4548 check_malloced_chunk(gm, mem, nb); |
|
4549 goto postaction; |
|
4550 } |
|
4551 |
|
4552 mem = sys_alloc(gm, nb); |
|
4553 |
|
4554 postaction: |
|
4555 POSTACTION(gm); |
|
4556 return mem; |
|
4557 } |
|
4558 |
|
4559 return 0; |
|
4560 } |
|
4561 |
|
4562 void dlfree(void* mem) { |
|
4563 /* |
|
4564 Consolidate freed chunks with preceeding or succeeding bordering |
|
4565 free chunks, if they exist, and then place in a bin. Intermixed |
|
4566 with special cases for top, dv, mmapped chunks, and usage errors. |
|
4567 */ |
|
4568 |
|
4569 if (mem != 0) { |
|
4570 mchunkptr p = mem2chunk(mem); |
|
4571 #if FOOTERS |
|
4572 mstate fm = get_mstate_for(p); |
|
4573 if (!ok_magic(fm)) { |
|
4574 USAGE_ERROR_ACTION(fm, p); |
|
4575 return; |
|
4576 } |
|
4577 #else /* FOOTERS */ |
|
4578 #define fm gm |
|
4579 #endif /* FOOTERS */ |
|
4580 if (!PREACTION(fm)) { |
|
4581 check_inuse_chunk(fm, p); |
|
4582 if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) { |
|
4583 size_t psize = chunksize(p); |
|
4584 mchunkptr next = chunk_plus_offset(p, psize); |
|
4585 if (!pinuse(p)) { |
|
4586 size_t prevsize = p->prev_foot; |
|
4587 if ((prevsize & IS_MMAPPED_BIT) != 0) { |
|
4588 prevsize &= ~IS_MMAPPED_BIT; |
|
4589 psize += prevsize + MMAP_FOOT_PAD; |
|
4590 if (CALL_MUNMAP((char*)p - prevsize, psize) == 0) |
|
4591 fm->footprint -= psize; |
|
4592 goto postaction; |
|
4593 } |
|
4594 else { |
|
4595 mchunkptr prev = chunk_minus_offset(p, prevsize); |
|
4596 psize += prevsize; |
|
4597 p = prev; |
|
4598 if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */ |
|
4599 if (p != fm->dv) { |
|
4600 unlink_chunk(fm, p, prevsize); |
|
4601 } |
|
4602 else if ((next->head & INUSE_BITS) == INUSE_BITS) { |
|
4603 fm->dvsize = psize; |
|
4604 set_free_with_pinuse(p, psize, next); |
|
4605 goto postaction; |
|
4606 } |
|
4607 } |
|
4608 else |
|
4609 goto erroraction; |
|
4610 } |
|
4611 } |
|
4612 |
|
4613 if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) { |
|
4614 if (!cinuse(next)) { /* consolidate forward */ |
|
4615 if (next == fm->top) { |
|
4616 size_t tsize = fm->topsize += psize; |
|
4617 fm->top = p; |
|
4618 p->head = tsize | PINUSE_BIT; |
|
4619 if (p == fm->dv) { |
|
4620 fm->dv = 0; |
|
4621 fm->dvsize = 0; |
|
4622 } |
|
4623 if (should_trim(fm, tsize)) |
|
4624 sys_trim(fm, 0); |
|
4625 goto postaction; |
|
4626 } |
|
4627 else if (next == fm->dv) { |
|
4628 size_t dsize = fm->dvsize += psize; |
|
4629 fm->dv = p; |
|
4630 set_size_and_pinuse_of_free_chunk(p, dsize); |
|
4631 goto postaction; |
|
4632 } |
|
4633 else { |
|
4634 size_t nsize = chunksize(next); |
|
4635 psize += nsize; |
|
4636 unlink_chunk(fm, next, nsize); |
|
4637 set_size_and_pinuse_of_free_chunk(p, psize); |
|
4638 if (p == fm->dv) { |
|
4639 fm->dvsize = psize; |
|
4640 goto postaction; |
|
4641 } |
|
4642 } |
|
4643 } |
|
4644 else |
|
4645 set_free_with_pinuse(p, psize, next); |
|
4646 |
|
4647 if (is_small(psize)) { |
|
4648 insert_small_chunk(fm, p, psize); |
|
4649 check_free_chunk(fm, p); |
|
4650 } |
|
4651 else { |
|
4652 tchunkptr tp = (tchunkptr)p; |
|
4653 insert_large_chunk(fm, tp, psize); |
|
4654 check_free_chunk(fm, p); |
|
4655 if (--fm->release_checks == 0) |
|
4656 release_unused_segments(fm); |
|
4657 } |
|
4658 goto postaction; |
|
4659 } |
|
4660 } |
|
4661 erroraction: |
|
4662 USAGE_ERROR_ACTION(fm, p); |
|
4663 postaction: |
|
4664 POSTACTION(fm); |
|
4665 } |
|
4666 } |
|
4667 #if !FOOTERS |
|
4668 #undef fm |
|
4669 #endif /* FOOTERS */ |
|
4670 } |
|
4671 |
|
4672 void* dlcalloc(size_t n_elements, size_t elem_size) { |
|
4673 void* mem; |
|
4674 size_t req = 0; |
|
4675 if (n_elements != 0) { |
|
4676 req = n_elements * elem_size; |
|
4677 if (((n_elements | elem_size) & ~(size_t)0xffff) && |
|
4678 (req / n_elements != elem_size)) |
|
4679 req = MAX_SIZE_T; /* force downstream failure on overflow */ |
|
4680 } |
|
4681 mem = dlmalloc(req); |
|
4682 if (mem != 0 && calloc_must_clear(mem2chunk(mem))) |
|
4683 memset(mem, 0, req); |
|
4684 return mem; |
|
4685 } |
|
4686 |
|
4687 void* dlrealloc(void* oldmem, size_t bytes) { |
|
4688 if (oldmem == 0) |
|
4689 return dlmalloc(bytes); |
|
4690 #ifdef REALLOC_ZERO_BYTES_FREES |
|
4691 if (bytes == 0) { |
|
4692 dlfree(oldmem); |
|
4693 return 0; |
|
4694 } |
|
4695 #endif /* REALLOC_ZERO_BYTES_FREES */ |
|
4696 else { |
|
4697 #if ! FOOTERS |
|
4698 mstate m = gm; |
|
4699 #else /* FOOTERS */ |
|
4700 mstate m = get_mstate_for(mem2chunk(oldmem)); |
|
4701 if (!ok_magic(m)) { |
|
4702 USAGE_ERROR_ACTION(m, oldmem); |
|
4703 return 0; |
|
4704 } |
|
4705 #endif /* FOOTERS */ |
|
4706 return internal_realloc(m, oldmem, bytes); |
|
4707 } |
|
4708 } |
|
4709 |
|
4710 void* dlmemalign(size_t alignment, size_t bytes) { |
|
4711 return internal_memalign(gm, alignment, bytes); |
|
4712 } |
|
4713 |
|
4714 void** dlindependent_calloc(size_t n_elements, size_t elem_size, |
|
4715 void* chunks[]) { |
|
4716 size_t sz = elem_size; /* serves as 1-element array */ |
|
4717 return ialloc(gm, n_elements, &sz, 3, chunks); |
|
4718 } |
|
4719 |
|
4720 void** dlindependent_comalloc(size_t n_elements, size_t sizes[], |
|
4721 void* chunks[]) { |
|
4722 return ialloc(gm, n_elements, sizes, 0, chunks); |
|
4723 } |
|
4724 |
|
4725 void* dlvalloc(size_t bytes) { |
|
4726 size_t pagesz; |
|
4727 init_mparams(); |
|
4728 pagesz = mparams.page_size; |
|
4729 return dlmemalign(pagesz, bytes); |
|
4730 } |
|
4731 |
|
4732 void* dlpvalloc(size_t bytes) { |
|
4733 size_t pagesz; |
|
4734 init_mparams(); |
|
4735 pagesz = mparams.page_size; |
|
4736 return dlmemalign(pagesz, (bytes + pagesz - SIZE_T_ONE) & ~(pagesz - SIZE_T_ONE)); |
|
4737 } |
|
4738 |
|
4739 int dlmalloc_trim(size_t pad) { |
|
4740 int result = 0; |
|
4741 if (!PREACTION(gm)) { |
|
4742 result = sys_trim(gm, pad); |
|
4743 POSTACTION(gm); |
|
4744 } |
|
4745 return result; |
|
4746 } |
|
4747 |
|
4748 size_t dlmalloc_footprint(void) { |
|
4749 return gm->footprint; |
|
4750 } |
|
4751 |
|
4752 size_t dlmalloc_max_footprint(void) { |
|
4753 return gm->max_footprint; |
|
4754 } |
|
4755 |
|
4756 #if !NO_MALLINFO |
|
4757 struct mallinfo dlmallinfo(void) { |
|
4758 return internal_mallinfo(gm); |
|
4759 } |
|
4760 #endif /* NO_MALLINFO */ |
|
4761 |
|
4762 void dlmalloc_stats() { |
|
4763 internal_malloc_stats(gm); |
|
4764 } |
|
4765 |
|
4766 size_t dlmalloc_usable_size(void* mem) { |
|
4767 if (mem != 0) { |
|
4768 mchunkptr p = mem2chunk(mem); |
|
4769 if (cinuse(p)) |
|
4770 return chunksize(p) - overhead_for(p); |
|
4771 } |
|
4772 return 0; |
|
4773 } |
|
4774 |
|
4775 int dlmallopt(int param_number, int value) { |
|
4776 return change_mparam(param_number, value); |
|
4777 } |
|
4778 |
|
4779 #endif /* !ONLY_MSPACES */ |
|
4780 |
|
4781 /* ----------------------------- user mspaces ---------------------------- */ |
|
4782 |
|
4783 #if MSPACES |
|
4784 |
|
4785 static mstate init_user_mstate(char* tbase, size_t tsize) { |
|
4786 size_t msize = pad_request(sizeof(struct malloc_state)); |
|
4787 mchunkptr mn; |
|
4788 mchunkptr msp = align_as_chunk(tbase); |
|
4789 mstate m = (mstate)(chunk2mem(msp)); |
|
4790 memset(m, 0, msize); |
|
4791 INITIAL_LOCK(&m->mutex); |
|
4792 msp->head = (msize|PINUSE_BIT|CINUSE_BIT); |
|
4793 m->seg.base = m->least_addr = tbase; |
|
4794 m->seg.size = m->footprint = m->max_footprint = tsize; |
|
4795 m->magic = mparams.magic; |
|
4796 m->release_checks = MAX_RELEASE_CHECK_RATE; |
|
4797 m->mflags = mparams.default_mflags; |
|
4798 m->extp = 0; |
|
4799 m->exts = 0; |
|
4800 disable_contiguous(m); |
|
4801 init_bins(m); |
|
4802 mn = next_chunk(mem2chunk(m)); |
|
4803 init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) - TOP_FOOT_SIZE); |
|
4804 check_top_chunk(m, m->top); |
|
4805 return m; |
|
4806 } |
|
4807 |
|
4808 mspace create_mspace(size_t capacity, int locked) { |
|
4809 mstate m = 0; |
|
4810 size_t msize = pad_request(sizeof(struct malloc_state)); |
|
4811 init_mparams(); /* Ensure pagesize etc initialized */ |
|
4812 |
|
4813 if (capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) { |
|
4814 size_t rs = ((capacity == 0)? mparams.granularity : |
|
4815 (capacity + TOP_FOOT_SIZE + msize)); |
|
4816 size_t tsize = granularity_align(rs); |
|
4817 char* tbase = (char*)(CALL_MMAP(tsize)); |
|
4818 if (tbase != CMFAIL) { |
|
4819 m = init_user_mstate(tbase, tsize); |
|
4820 m->seg.sflags = IS_MMAPPED_BIT; |
|
4821 set_lock(m, locked); |
|
4822 } |
|
4823 } |
|
4824 return (mspace)m; |
|
4825 } |
|
4826 |
|
4827 mspace create_mspace_with_base(void* base, size_t capacity, int locked) { |
|
4828 mstate m = 0; |
|
4829 size_t msize = pad_request(sizeof(struct malloc_state)); |
|
4830 init_mparams(); /* Ensure pagesize etc initialized */ |
|
4831 |
|
4832 if (capacity > msize + TOP_FOOT_SIZE && |
|
4833 capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) { |
|
4834 m = init_user_mstate((char*)base, capacity); |
|
4835 m->seg.sflags = EXTERN_BIT; |
|
4836 set_lock(m, locked); |
|
4837 } |
|
4838 return (mspace)m; |
|
4839 } |
|
4840 |
|
4841 size_t destroy_mspace(mspace msp) { |
|
4842 size_t freed = 0; |
|
4843 mstate ms = (mstate)msp; |
|
4844 if (ok_magic(ms)) { |
|
4845 msegmentptr sp = &ms->seg; |
|
4846 while (sp != 0) { |
|
4847 char* base = sp->base; |
|
4848 size_t size = sp->size; |
|
4849 flag_t flag = sp->sflags; |
|
4850 sp = sp->next; |
|
4851 if ((flag & IS_MMAPPED_BIT) && !(flag & EXTERN_BIT) && |
|
4852 CALL_MUNMAP(base, size) == 0) |
|
4853 freed += size; |
|
4854 } |
|
4855 } |
|
4856 else { |
|
4857 USAGE_ERROR_ACTION(ms,ms); |
|
4858 } |
|
4859 return freed; |
|
4860 } |
|
4861 |
|
4862 /* |
|
4863 mspace versions of routines are near-clones of the global |
|
4864 versions. This is not so nice but better than the alternatives. |
|
4865 */ |
|
4866 |
|
4867 |
|
4868 void* mspace_malloc(mspace msp, size_t bytes) { |
|
4869 mstate ms = (mstate)msp; |
|
4870 if (!ok_magic(ms)) { |
|
4871 USAGE_ERROR_ACTION(ms,ms); |
|
4872 return 0; |
|
4873 } |
|
4874 if (!PREACTION(ms)) { |
|
4875 void* mem; |
|
4876 size_t nb; |
|
4877 if (bytes <= MAX_SMALL_REQUEST) { |
|
4878 bindex_t idx; |
|
4879 binmap_t smallbits; |
|
4880 nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes); |
|
4881 idx = small_index(nb); |
|
4882 smallbits = ms->smallmap >> idx; |
|
4883 |
|
4884 if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */ |
|
4885 mchunkptr b, p; |
|
4886 idx += ~smallbits & 1; /* Uses next bin if idx empty */ |
|
4887 b = smallbin_at(ms, idx); |
|
4888 p = b->fd; |
|
4889 assert(chunksize(p) == small_index2size(idx)); |
|
4890 unlink_first_small_chunk(ms, b, p, idx); |
|
4891 set_inuse_and_pinuse(ms, p, small_index2size(idx)); |
|
4892 mem = chunk2mem(p); |
|
4893 check_malloced_chunk(ms, mem, nb); |
|
4894 goto postaction; |
|
4895 } |
|
4896 |
|
4897 else if (nb > ms->dvsize) { |
|
4898 if (smallbits != 0) { /* Use chunk in next nonempty smallbin */ |
|
4899 mchunkptr b, p, r; |
|
4900 size_t rsize; |
|
4901 bindex_t i; |
|
4902 binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx)); |
|
4903 binmap_t leastbit = least_bit(leftbits); |
|
4904 compute_bit2idx(leastbit, i); |
|
4905 b = smallbin_at(ms, i); |
|
4906 p = b->fd; |
|
4907 assert(chunksize(p) == small_index2size(i)); |
|
4908 unlink_first_small_chunk(ms, b, p, i); |
|
4909 rsize = small_index2size(i) - nb; |
|
4910 /* Fit here cannot be remainderless if 4byte sizes */ |
|
4911 if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE) |
|
4912 set_inuse_and_pinuse(ms, p, small_index2size(i)); |
|
4913 else { |
|
4914 set_size_and_pinuse_of_inuse_chunk(ms, p, nb); |
|
4915 r = chunk_plus_offset(p, nb); |
|
4916 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4917 replace_dv(ms, r, rsize); |
|
4918 } |
|
4919 mem = chunk2mem(p); |
|
4920 check_malloced_chunk(ms, mem, nb); |
|
4921 goto postaction; |
|
4922 } |
|
4923 |
|
4924 else if (ms->treemap != 0 && (mem = tmalloc_small(ms, nb)) != 0) { |
|
4925 check_malloced_chunk(ms, mem, nb); |
|
4926 goto postaction; |
|
4927 } |
|
4928 } |
|
4929 } |
|
4930 else if (bytes >= MAX_REQUEST) |
|
4931 nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */ |
|
4932 else { |
|
4933 nb = pad_request(bytes); |
|
4934 if (ms->treemap != 0 && (mem = tmalloc_large(ms, nb)) != 0) { |
|
4935 check_malloced_chunk(ms, mem, nb); |
|
4936 goto postaction; |
|
4937 } |
|
4938 } |
|
4939 |
|
4940 if (nb <= ms->dvsize) { |
|
4941 size_t rsize = ms->dvsize - nb; |
|
4942 mchunkptr p = ms->dv; |
|
4943 if (rsize >= MIN_CHUNK_SIZE) { /* split dv */ |
|
4944 mchunkptr r = ms->dv = chunk_plus_offset(p, nb); |
|
4945 ms->dvsize = rsize; |
|
4946 set_size_and_pinuse_of_free_chunk(r, rsize); |
|
4947 set_size_and_pinuse_of_inuse_chunk(ms, p, nb); |
|
4948 } |
|
4949 else { /* exhaust dv */ |
|
4950 size_t dvs = ms->dvsize; |
|
4951 ms->dvsize = 0; |
|
4952 ms->dv = 0; |
|
4953 set_inuse_and_pinuse(ms, p, dvs); |
|
4954 } |
|
4955 mem = chunk2mem(p); |
|
4956 check_malloced_chunk(ms, mem, nb); |
|
4957 goto postaction; |
|
4958 } |
|
4959 |
|
4960 else if (nb < ms->topsize) { /* Split top */ |
|
4961 size_t rsize = ms->topsize -= nb; |
|
4962 mchunkptr p = ms->top; |
|
4963 mchunkptr r = ms->top = chunk_plus_offset(p, nb); |
|
4964 r->head = rsize | PINUSE_BIT; |
|
4965 set_size_and_pinuse_of_inuse_chunk(ms, p, nb); |
|
4966 mem = chunk2mem(p); |
|
4967 check_top_chunk(ms, ms->top); |
|
4968 check_malloced_chunk(ms, mem, nb); |
|
4969 goto postaction; |
|
4970 } |
|
4971 |
|
4972 mem = sys_alloc(ms, nb); |
|
4973 |
|
4974 postaction: |
|
4975 POSTACTION(ms); |
|
4976 return mem; |
|
4977 } |
|
4978 |
|
4979 return 0; |
|
4980 } |
|
4981 |
|
4982 void mspace_free(mspace msp, void* mem) { |
|
4983 if (mem != 0) { |
|
4984 mchunkptr p = mem2chunk(mem); |
|
4985 #if FOOTERS |
|
4986 mstate fm = get_mstate_for(p); |
|
4987 #else /* FOOTERS */ |
|
4988 mstate fm = (mstate)msp; |
|
4989 #endif /* FOOTERS */ |
|
4990 if (!ok_magic(fm)) { |
|
4991 USAGE_ERROR_ACTION(fm, p); |
|
4992 return; |
|
4993 } |
|
4994 if (!PREACTION(fm)) { |
|
4995 check_inuse_chunk(fm, p); |
|
4996 if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) { |
|
4997 size_t psize = chunksize(p); |
|
4998 mchunkptr next = chunk_plus_offset(p, psize); |
|
4999 if (!pinuse(p)) { |
|
5000 size_t prevsize = p->prev_foot; |
|
5001 if ((prevsize & IS_MMAPPED_BIT) != 0) { |
|
5002 prevsize &= ~IS_MMAPPED_BIT; |
|
5003 psize += prevsize + MMAP_FOOT_PAD; |
|
5004 if (CALL_MUNMAP((char*)p - prevsize, psize) == 0) |
|
5005 fm->footprint -= psize; |
|
5006 goto postaction; |
|
5007 } |
|
5008 else { |
|
5009 mchunkptr prev = chunk_minus_offset(p, prevsize); |
|
5010 psize += prevsize; |
|
5011 p = prev; |
|
5012 if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */ |
|
5013 if (p != fm->dv) { |
|
5014 unlink_chunk(fm, p, prevsize); |
|
5015 } |
|
5016 else if ((next->head & INUSE_BITS) == INUSE_BITS) { |
|
5017 fm->dvsize = psize; |
|
5018 set_free_with_pinuse(p, psize, next); |
|
5019 goto postaction; |
|
5020 } |
|
5021 } |
|
5022 else |
|
5023 goto erroraction; |
|
5024 } |
|
5025 } |
|
5026 |
|
5027 if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) { |
|
5028 if (!cinuse(next)) { /* consolidate forward */ |
|
5029 if (next == fm->top) { |
|
5030 size_t tsize = fm->topsize += psize; |
|
5031 fm->top = p; |
|
5032 p->head = tsize | PINUSE_BIT; |
|
5033 if (p == fm->dv) { |
|
5034 fm->dv = 0; |
|
5035 fm->dvsize = 0; |
|
5036 } |
|
5037 if (should_trim(fm, tsize)) |
|
5038 sys_trim(fm, 0); |
|
5039 goto postaction; |
|
5040 } |
|
5041 else if (next == fm->dv) { |
|
5042 size_t dsize = fm->dvsize += psize; |
|
5043 fm->dv = p; |
|
5044 set_size_and_pinuse_of_free_chunk(p, dsize); |
|
5045 goto postaction; |
|
5046 } |
|
5047 else { |
|
5048 size_t nsize = chunksize(next); |
|
5049 psize += nsize; |
|
5050 unlink_chunk(fm, next, nsize); |
|
5051 set_size_and_pinuse_of_free_chunk(p, psize); |
|
5052 if (p == fm->dv) { |
|
5053 fm->dvsize = psize; |
|
5054 goto postaction; |
|
5055 } |
|
5056 } |
|
5057 } |
|
5058 else |
|
5059 set_free_with_pinuse(p, psize, next); |
|
5060 |
|
5061 if (is_small(psize)) { |
|
5062 insert_small_chunk(fm, p, psize); |
|
5063 check_free_chunk(fm, p); |
|
5064 } |
|
5065 else { |
|
5066 tchunkptr tp = (tchunkptr)p; |
|
5067 insert_large_chunk(fm, tp, psize); |
|
5068 check_free_chunk(fm, p); |
|
5069 if (--fm->release_checks == 0) |
|
5070 release_unused_segments(fm); |
|
5071 } |
|
5072 goto postaction; |
|
5073 } |
|
5074 } |
|
5075 erroraction: |
|
5076 USAGE_ERROR_ACTION(fm, p); |
|
5077 postaction: |
|
5078 POSTACTION(fm); |
|
5079 } |
|
5080 } |
|
5081 } |
|
5082 |
|
5083 void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size) { |
|
5084 void* mem; |
|
5085 size_t req = 0; |
|
5086 mstate ms = (mstate)msp; |
|
5087 if (!ok_magic(ms)) { |
|
5088 USAGE_ERROR_ACTION(ms,ms); |
|
5089 return 0; |
|
5090 } |
|
5091 if (n_elements != 0) { |
|
5092 req = n_elements * elem_size; |
|
5093 if (((n_elements | elem_size) & ~(size_t)0xffff) && |
|
5094 (req / n_elements != elem_size)) |
|
5095 req = MAX_SIZE_T; /* force downstream failure on overflow */ |
|
5096 } |
|
5097 mem = internal_malloc(ms, req); |
|
5098 if (mem != 0 && calloc_must_clear(mem2chunk(mem))) |
|
5099 memset(mem, 0, req); |
|
5100 return mem; |
|
5101 } |
|
5102 |
|
5103 void* mspace_realloc(mspace msp, void* oldmem, size_t bytes) { |
|
5104 if (oldmem == 0) |
|
5105 return mspace_malloc(msp, bytes); |
|
5106 #ifdef REALLOC_ZERO_BYTES_FREES |
|
5107 if (bytes == 0) { |
|
5108 mspace_free(msp, oldmem); |
|
5109 return 0; |
|
5110 } |
|
5111 #endif /* REALLOC_ZERO_BYTES_FREES */ |
|
5112 else { |
|
5113 #if FOOTERS |
|
5114 mchunkptr p = mem2chunk(oldmem); |
|
5115 mstate ms = get_mstate_for(p); |
|
5116 #else /* FOOTERS */ |
|
5117 mstate ms = (mstate)msp; |
|
5118 #endif /* FOOTERS */ |
|
5119 if (!ok_magic(ms)) { |
|
5120 USAGE_ERROR_ACTION(ms,ms); |
|
5121 return 0; |
|
5122 } |
|
5123 return internal_realloc(ms, oldmem, bytes); |
|
5124 } |
|
5125 } |
|
5126 |
|
5127 void* mspace_memalign(mspace msp, size_t alignment, size_t bytes) { |
|
5128 mstate ms = (mstate)msp; |
|
5129 if (!ok_magic(ms)) { |
|
5130 USAGE_ERROR_ACTION(ms,ms); |
|
5131 return 0; |
|
5132 } |
|
5133 return internal_memalign(ms, alignment, bytes); |
|
5134 } |
|
5135 |
|
5136 void** mspace_independent_calloc(mspace msp, size_t n_elements, |
|
5137 size_t elem_size, void* chunks[]) { |
|
5138 size_t sz = elem_size; /* serves as 1-element array */ |
|
5139 mstate ms = (mstate)msp; |
|
5140 if (!ok_magic(ms)) { |
|
5141 USAGE_ERROR_ACTION(ms,ms); |
|
5142 return 0; |
|
5143 } |
|
5144 return ialloc(ms, n_elements, &sz, 3, chunks); |
|
5145 } |
|
5146 |
|
5147 void** mspace_independent_comalloc(mspace msp, size_t n_elements, |
|
5148 size_t sizes[], void* chunks[]) { |
|
5149 mstate ms = (mstate)msp; |
|
5150 if (!ok_magic(ms)) { |
|
5151 USAGE_ERROR_ACTION(ms,ms); |
|
5152 return 0; |
|
5153 } |
|
5154 return ialloc(ms, n_elements, sizes, 0, chunks); |
|
5155 } |
|
5156 |
|
5157 int mspace_trim(mspace msp, size_t pad) { |
|
5158 int result = 0; |
|
5159 mstate ms = (mstate)msp; |
|
5160 if (ok_magic(ms)) { |
|
5161 if (!PREACTION(ms)) { |
|
5162 result = sys_trim(ms, pad); |
|
5163 POSTACTION(ms); |
|
5164 } |
|
5165 } |
|
5166 else { |
|
5167 USAGE_ERROR_ACTION(ms,ms); |
|
5168 } |
|
5169 return result; |
|
5170 } |
|
5171 |
|
5172 void mspace_malloc_stats(mspace msp) { |
|
5173 mstate ms = (mstate)msp; |
|
5174 if (ok_magic(ms)) { |
|
5175 internal_malloc_stats(ms); |
|
5176 } |
|
5177 else { |
|
5178 USAGE_ERROR_ACTION(ms,ms); |
|
5179 } |
|
5180 } |
|
5181 |
|
5182 size_t mspace_footprint(mspace msp) { |
|
5183 size_t result = 0; |
|
5184 mstate ms = (mstate)msp; |
|
5185 if (ok_magic(ms)) { |
|
5186 result = ms->footprint; |
|
5187 } |
|
5188 else { |
|
5189 USAGE_ERROR_ACTION(ms,ms); |
|
5190 } |
|
5191 return result; |
|
5192 } |
|
5193 |
|
5194 |
|
5195 size_t mspace_max_footprint(mspace msp) { |
|
5196 size_t result = 0; |
|
5197 mstate ms = (mstate)msp; |
|
5198 if (ok_magic(ms)) { |
|
5199 result = ms->max_footprint; |
|
5200 } |
|
5201 else { |
|
5202 USAGE_ERROR_ACTION(ms,ms); |
|
5203 } |
|
5204 return result; |
|
5205 } |
|
5206 |
|
5207 |
|
5208 #if !NO_MALLINFO |
|
5209 struct mallinfo mspace_mallinfo(mspace msp) { |
|
5210 mstate ms = (mstate)msp; |
|
5211 if (!ok_magic(ms)) { |
|
5212 USAGE_ERROR_ACTION(ms,ms); |
|
5213 } |
|
5214 return internal_mallinfo(ms); |
|
5215 } |
|
5216 #endif /* NO_MALLINFO */ |
|
5217 |
|
5218 size_t mspace_usable_size(void* mem) { |
|
5219 if (mem != 0) { |
|
5220 mchunkptr p = mem2chunk(mem); |
|
5221 if (cinuse(p)) |
|
5222 return chunksize(p) - overhead_for(p); |
|
5223 } |
|
5224 return 0; |
|
5225 } |
|
5226 |
|
5227 int mspace_mallopt(int param_number, int value) { |
|
5228 return change_mparam(param_number, value); |
|
5229 } |
|
5230 |
|
5231 #endif /* MSPACES */ |
|
5232 |
|
5233 /* -------------------- Alternative MORECORE functions ------------------- */ |
|
5234 |
|
5235 /* |
|
5236 Guidelines for creating a custom version of MORECORE: |
|
5237 |
|
5238 * For best performance, MORECORE should allocate in multiples of pagesize. |
|
5239 * MORECORE may allocate more memory than requested. (Or even less, |
|
5240 but this will usually result in a malloc failure.) |
|
5241 * MORECORE must not allocate memory when given argument zero, but |
|
5242 instead return one past the end address of memory from previous |
|
5243 nonzero call. |
|
5244 * For best performance, consecutive calls to MORECORE with positive |
|
5245 arguments should return increasing addresses, indicating that |
|
5246 space has been contiguously extended. |
|
5247 * Even though consecutive calls to MORECORE need not return contiguous |
|
5248 addresses, it must be OK for malloc'ed chunks to span multiple |
|
5249 regions in those cases where they do happen to be contiguous. |
|
5250 * MORECORE need not handle negative arguments -- it may instead |
|
5251 just return MFAIL when given negative arguments. |
|
5252 Negative arguments are always multiples of pagesize. MORECORE |
|
5253 must not misinterpret negative args as large positive unsigned |
|
5254 args. You can suppress all such calls from even occurring by defining |
|
5255 MORECORE_CANNOT_TRIM, |
|
5256 |
|
5257 As an example alternative MORECORE, here is a custom allocator |
|
5258 kindly contributed for pre-OSX macOS. It uses virtually but not |
|
5259 necessarily physically contiguous non-paged memory (locked in, |
|
5260 present and won't get swapped out). You can use it by uncommenting |
|
5261 this section, adding some #includes, and setting up the appropriate |
|
5262 defines above: |
|
5263 |
|
5264 #define MORECORE osMoreCore |
|
5265 |
|
5266 There is also a shutdown routine that should somehow be called for |
|
5267 cleanup upon program exit. |
|
5268 |
|
5269 #define MAX_POOL_ENTRIES 100 |
|
5270 #define MINIMUM_MORECORE_SIZE (64 * 1024U) |
|
5271 static int next_os_pool; |
|
5272 void *our_os_pools[MAX_POOL_ENTRIES]; |
|
5273 |
|
5274 void *osMoreCore(int size) |
|
5275 { |
|
5276 void *ptr = 0; |
|
5277 static void *sbrk_top = 0; |
|
5278 |
|
5279 if (size > 0) |
|
5280 { |
|
5281 if (size < MINIMUM_MORECORE_SIZE) |
|
5282 size = MINIMUM_MORECORE_SIZE; |
|
5283 if (CurrentExecutionLevel() == kTaskLevel) |
|
5284 ptr = PoolAllocateResident(size + RM_PAGE_SIZE, 0); |
|
5285 if (ptr == 0) |
|
5286 { |
|
5287 return (void *) MFAIL; |
|
5288 } |
|
5289 // save ptrs so they can be freed during cleanup |
|
5290 our_os_pools[next_os_pool] = ptr; |
|
5291 next_os_pool++; |
|
5292 ptr = (void *) ((((size_t) ptr) + RM_PAGE_MASK) & ~RM_PAGE_MASK); |
|
5293 sbrk_top = (char *) ptr + size; |
|
5294 return ptr; |
|
5295 } |
|
5296 else if (size < 0) |
|
5297 { |
|
5298 // we don't currently support shrink behavior |
|
5299 return (void *) MFAIL; |
|
5300 } |
|
5301 else |
|
5302 { |
|
5303 return sbrk_top; |
|
5304 } |
|
5305 } |
|
5306 |
|
5307 // cleanup any allocated memory pools |
|
5308 // called as last thing before shutting down driver |
|
5309 |
|
5310 void osCleanupMem(void) |
|
5311 { |
|
5312 void **ptr; |
|
5313 |
|
5314 for (ptr = our_os_pools; ptr < &our_os_pools[MAX_POOL_ENTRIES]; ptr++) |
|
5315 if (*ptr) |
|
5316 { |
|
5317 PoolDeallocate(*ptr); |
|
5318 *ptr = 0; |
|
5319 } |
|
5320 } |
|
5321 |
|
5322 */ |
|
5323 |
|
5324 |
|
5325 /* ----------------------------------------------------------------------- |
|
5326 History: |
|
5327 V2.8.4 (not yet released) |
|
5328 * Fix bad error check in mspace_footprint |
|
5329 * Adaptations for ptmalloc, courtesy of Wolfram Gloger. |
|
5330 * Reentrant spin locks, courtesy of Earl Chew and others |
|
5331 * Win32 improvements, courtesy of Niall Douglas and Earl Chew |
|
5332 * Add NO_SEGMENT_TRAVERSAL and MAX_RELEASE_CHECK_RATE options |
|
5333 * Various small adjustments to reduce warnings on some compilers |
|
5334 * Extension hook in malloc_state |
|
5335 |
|
5336 V2.8.3 Thu Sep 22 11:16:32 2005 Doug Lea (dl at gee) |
|
5337 * Add max_footprint functions |
|
5338 * Ensure all appropriate literals are size_t |
|
5339 * Fix conditional compilation problem for some #define settings |
|
5340 * Avoid concatenating segments with the one provided |
|
5341 in create_mspace_with_base |
|
5342 * Rename some variables to avoid compiler shadowing warnings |
|
5343 * Use explicit lock initialization. |
|
5344 * Better handling of sbrk interference. |
|
5345 * Simplify and fix segment insertion, trimming and mspace_destroy |
|
5346 * Reinstate REALLOC_ZERO_BYTES_FREES option from 2.7.x |
|
5347 * Thanks especially to Dennis Flanagan for help on these. |
|
5348 |
|
5349 V2.8.2 Sun Jun 12 16:01:10 2005 Doug Lea (dl at gee) |
|
5350 * Fix memalign brace error. |
|
5351 |
|
5352 V2.8.1 Wed Jun 8 16:11:46 2005 Doug Lea (dl at gee) |
|
5353 * Fix improper #endif nesting in C++ |
|
5354 * Add explicit casts needed for C++ |
|
5355 |
|
5356 V2.8.0 Mon May 30 14:09:02 2005 Doug Lea (dl at gee) |
|
5357 * Use trees for large bins |
|
5358 * Support mspaces |
|
5359 * Use segments to unify sbrk-based and mmap-based system allocation, |
|
5360 removing need for emulation on most platforms without sbrk. |
|
5361 * Default safety checks |
|
5362 * Optional footer checks. Thanks to William Robertson for the idea. |
|
5363 * Internal code refactoring |
|
5364 * Incorporate suggestions and platform-specific changes. |
|
5365 Thanks to Dennis Flanagan, Colin Plumb, Niall Douglas, |
|
5366 Aaron Bachmann, Emery Berger, and others. |
|
5367 * Speed up non-fastbin processing enough to remove fastbins. |
|
5368 * Remove useless cfree() to avoid conflicts with other apps. |
|
5369 * Remove internal memcpy, memset. Compilers handle builtins better. |
|
5370 * Remove some options that no one ever used and rename others. |
|
5371 |
|
5372 V2.7.2 Sat Aug 17 09:07:30 2002 Doug Lea (dl at gee) |
|
5373 * Fix malloc_state bitmap array misdeclaration |
|
5374 |
|
5375 V2.7.1 Thu Jul 25 10:58:03 2002 Doug Lea (dl at gee) |
|
5376 * Allow tuning of FIRST_SORTED_BIN_SIZE |
|
5377 * Use PTR_UINT as type for all ptr->int casts. Thanks to John Belmonte. |
|
5378 * Better detection and support for non-contiguousness of MORECORE. |
|
5379 Thanks to Andreas Mueller, Conal Walsh, and Wolfram Gloger |
|
5380 * Bypass most of malloc if no frees. Thanks To Emery Berger. |
|
5381 * Fix freeing of old top non-contiguous chunk im sysmalloc. |
|
5382 * Raised default trim and map thresholds to 256K. |
|
5383 * Fix mmap-related #defines. Thanks to Lubos Lunak. |
|
5384 * Fix copy macros; added LACKS_FCNTL_H. Thanks to Neal Walfield. |
|
5385 * Branch-free bin calculation |
|
5386 * Default trim and mmap thresholds now 256K. |
|
5387 |
|
5388 V2.7.0 Sun Mar 11 14:14:06 2001 Doug Lea (dl at gee) |
|
5389 * Introduce independent_comalloc and independent_calloc. |
|
5390 Thanks to Michael Pachos for motivation and help. |
|
5391 * Make optional .h file available |
|
5392 * Allow > 2GB requests on 32bit systems. |
|
5393 * new WIN32 sbrk, mmap, munmap, lock code from <Walter@GeNeSys-e.de>. |
|
5394 Thanks also to Andreas Mueller <a.mueller at paradatec.de>, |
|
5395 and Anonymous. |
|
5396 * Allow override of MALLOC_ALIGNMENT (Thanks to Ruud Waij for |
|
5397 helping test this.) |
|
5398 * memalign: check alignment arg |
|
5399 * realloc: don't try to shift chunks backwards, since this |
|
5400 leads to more fragmentation in some programs and doesn't |
|
5401 seem to help in any others. |
|
5402 * Collect all cases in malloc requiring system memory into sysmalloc |
|
5403 * Use mmap as backup to sbrk |
|
5404 * Place all internal state in malloc_state |
|
5405 * Introduce fastbins (although similar to 2.5.1) |
|
5406 * Many minor tunings and cosmetic improvements |
|
5407 * Introduce USE_PUBLIC_MALLOC_WRAPPERS, USE_MALLOC_LOCK |
|
5408 * Introduce MALLOC_FAILURE_ACTION, MORECORE_CONTIGUOUS |
|
5409 Thanks to Tony E. Bennett <tbennett@nvidia.com> and others. |
|
5410 * Include errno.h to support default failure action. |
|
5411 |
|
5412 V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee) |
|
5413 * return null for negative arguments |
|
5414 * Added Several WIN32 cleanups from Martin C. Fong <mcfong at yahoo.com> |
|
5415 * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h' |
|
5416 (e.g. WIN32 platforms) |
|
5417 * Cleanup header file inclusion for WIN32 platforms |
|
5418 * Cleanup code to avoid Microsoft Visual C++ compiler complaints |
|
5419 * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing |
|
5420 memory allocation routines |
|
5421 * Set 'malloc_getpagesize' for WIN32 platforms (needs more work) |
|
5422 * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to |
|
5423 usage of 'assert' in non-WIN32 code |
|
5424 * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to |
|
5425 avoid infinite loop |
|
5426 * Always call 'fREe()' rather than 'free()' |
|
5427 |
|
5428 V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee) |
|
5429 * Fixed ordering problem with boundary-stamping |
|
5430 |
|
5431 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee) |
|
5432 * Added pvalloc, as recommended by H.J. Liu |
|
5433 * Added 64bit pointer support mainly from Wolfram Gloger |
|
5434 * Added anonymously donated WIN32 sbrk emulation |
|
5435 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen |
|
5436 * malloc_extend_top: fix mask error that caused wastage after |
|
5437 foreign sbrks |
|
5438 * Add linux mremap support code from HJ Liu |
|
5439 |
|
5440 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee) |
|
5441 * Integrated most documentation with the code. |
|
5442 * Add support for mmap, with help from |
|
5443 Wolfram Gloger (Gloger@lrz.uni-muenchen.de). |
|
5444 * Use last_remainder in more cases. |
|
5445 * Pack bins using idea from colin@nyx10.cs.du.edu |
|
5446 * Use ordered bins instead of best-fit threshhold |
|
5447 * Eliminate block-local decls to simplify tracing and debugging. |
|
5448 * Support another case of realloc via move into top |
|
5449 * Fix error occuring when initial sbrk_base not word-aligned. |
|
5450 * Rely on page size for units instead of SBRK_UNIT to |
|
5451 avoid surprises about sbrk alignment conventions. |
|
5452 * Add mallinfo, mallopt. Thanks to Raymond Nijssen |
|
5453 (raymond@es.ele.tue.nl) for the suggestion. |
|
5454 * Add `pad' argument to malloc_trim and top_pad mallopt parameter. |
|
5455 * More precautions for cases where other routines call sbrk, |
|
5456 courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de). |
|
5457 * Added macros etc., allowing use in linux libc from |
|
5458 H.J. Lu (hjl@gnu.ai.mit.edu) |
|
5459 * Inverted this history list |
|
5460 |
|
5461 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee) |
|
5462 * Re-tuned and fixed to behave more nicely with V2.6.0 changes. |
|
5463 * Removed all preallocation code since under current scheme |
|
5464 the work required to undo bad preallocations exceeds |
|
5465 the work saved in good cases for most test programs. |
|
5466 * No longer use return list or unconsolidated bins since |
|
5467 no scheme using them consistently outperforms those that don't |
|
5468 given above changes. |
|
5469 * Use best fit for very large chunks to prevent some worst-cases. |
|
5470 * Added some support for debugging |
|
5471 |
|
5472 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee) |
|
5473 * Removed footers when chunks are in use. Thanks to |
|
5474 Paul Wilson (wilson@cs.texas.edu) for the suggestion. |
|
5475 |
|
5476 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee) |
|
5477 * Added malloc_trim, with help from Wolfram Gloger |
|
5478 (wmglo@Dent.MED.Uni-Muenchen.DE). |
|
5479 |
|
5480 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g) |
|
5481 |
|
5482 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g) |
|
5483 * realloc: try to expand in both directions |
|
5484 * malloc: swap order of clean-bin strategy; |
|
5485 * realloc: only conditionally expand backwards |
|
5486 * Try not to scavenge used bins |
|
5487 * Use bin counts as a guide to preallocation |
|
5488 * Occasionally bin return list chunks in first scan |
|
5489 * Add a few optimizations from colin@nyx10.cs.du.edu |
|
5490 |
|
5491 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g) |
|
5492 * faster bin computation & slightly different binning |
|
5493 * merged all consolidations to one part of malloc proper |
|
5494 (eliminating old malloc_find_space & malloc_clean_bin) |
|
5495 * Scan 2 returns chunks (not just 1) |
|
5496 * Propagate failure in realloc if malloc returns 0 |
|
5497 * Add stuff to allow compilation on non-ANSI compilers |
|
5498 from kpv@research.att.com |
|
5499 |
|
5500 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu) |
|
5501 * removed potential for odd address access in prev_chunk |
|
5502 * removed dependency on getpagesize.h |
|
5503 * misc cosmetics and a bit more internal documentation |
|
5504 * anticosmetics: mangled names in macros to evade debugger strangeness |
|
5505 * tested on sparc, hp-700, dec-mips, rs6000 |
|
5506 with gcc & native cc (hp, dec only) allowing |
|
5507 Detlefs & Zorn comparison study (in SIGPLAN Notices.) |
|
5508 |
|
5509 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu) |
|
5510 * Based loosely on libg++-1.2X malloc. (It retains some of the overall |
|
5511 structure of old version, but most details differ.) |
|
5512 |
|
5513 */ |
|
5514 |
|
5515 |