diff options
author | Maurus Cuelenaere <mcuelenaere@gmail.com> | 2009-05-21 19:01:41 +0000 |
---|---|---|
committer | Maurus Cuelenaere <mcuelenaere@gmail.com> | 2009-05-21 19:01:41 +0000 |
commit | cf87597226f5d6b269f1f2c4d6f402aa1eccb852 (patch) | |
tree | 4ba1f3ae53b3bd9cae0e2c6c4dd57836b43a5ece /apps/plugins/lua/malloc.c | |
parent | c483efadc63eaed35b5fb5e4e02c2282daf32470 (diff) | |
download | rockbox-cf87597226f5d6b269f1f2c4d6f402aa1eccb852.tar.gz rockbox-cf87597226f5d6b269f1f2c4d6f402aa1eccb852.zip |
Commit FS#9174: Lua scripting language by Dan Everton
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21020 a1c6a512-1295-4272-9138-f99709370657
Diffstat (limited to 'apps/plugins/lua/malloc.c')
-rw-r--r-- | apps/plugins/lua/malloc.c | 5061 |
1 files changed, 5061 insertions, 0 deletions
diff --git a/apps/plugins/lua/malloc.c b/apps/plugins/lua/malloc.c new file mode 100644 index 0000000000..6dc00c4f9a --- /dev/null +++ b/apps/plugins/lua/malloc.c | |||
@@ -0,0 +1,5061 @@ | |||
1 | /* | ||
2 | This is a version (aka dlmalloc) of malloc/free/realloc written by | ||
3 | Doug Lea and released to the public domain, as explained at | ||
4 | http://creativecommons.org/licenses/publicdomain. Send questions, | ||
5 | comments, complaints, performance data, etc to dl@cs.oswego.edu | ||
6 | |||
7 | * Version 2.8.3 Thu Sep 22 11:16:15 2005 Doug Lea (dl at gee) | ||
8 | |||
9 | Note: There may be an updated version of this malloc obtainable at | ||
10 | ftp://gee.cs.oswego.edu/pub/misc/malloc.c | ||
11 | Check before installing! | ||
12 | |||
13 | * Quickstart | ||
14 | |||
15 | This library is all in one file to simplify the most common usage: | ||
16 | ftp it, compile it (-O3), and link it into another program. All of | ||
17 | the compile-time options default to reasonable values for use on | ||
18 | most platforms. You might later want to step through various | ||
19 | compile-time and dynamic tuning options. | ||
20 | |||
21 | For convenience, an include file for code using this malloc is at: | ||
22 | ftp://gee.cs.oswego.edu/pub/misc/malloc-2.8.3.h | ||
23 | You don't really need this .h file unless you call functions not | ||
24 | defined in your system include files. The .h file contains only the | ||
25 | excerpts from this file needed for using this malloc on ANSI C/C++ | ||
26 | systems, so long as you haven't changed compile-time options about | ||
27 | naming and tuning parameters. If you do, then you can create your | ||
28 | own malloc.h that does include all settings by cutting at the point | ||
29 | indicated below. Note that you may already by default be using a C | ||
30 | library containing a malloc that is based on some version of this | ||
31 | malloc (for example in linux). You might still want to use the one | ||
32 | in this file to customize settings or to avoid overheads associated | ||
33 | with library versions. | ||
34 | |||
35 | * Vital statistics: | ||
36 | |||
37 | Supported pointer/size_t representation: 4 or 8 bytes | ||
38 | size_t MUST be an unsigned type of the same width as | ||
39 | pointers. (If you are using an ancient system that declares | ||
40 | size_t as a signed type, or need it to be a different width | ||
41 | than pointers, you can use a previous release of this malloc | ||
42 | (e.g. 2.7.2) supporting these.) | ||
43 | |||
44 | Alignment: 8 bytes (default) | ||
45 | This suffices for nearly all current machines and C compilers. | ||
46 | However, you can define MALLOC_ALIGNMENT to be wider than this | ||
47 | if necessary (up to 128bytes), at the expense of using more space. | ||
48 | |||
49 | Minimum overhead per allocated chunk: 4 or 8 bytes (if 4byte sizes) | ||
50 | 8 or 16 bytes (if 8byte sizes) | ||
51 | Each malloced chunk has a hidden word of overhead holding size | ||
52 | and status information, and additional cross-check word | ||
53 | if FOOTERS is defined. | ||
54 | |||
55 | Minimum allocated size: 4-byte ptrs: 16 bytes (including overhead) | ||
56 | 8-byte ptrs: 32 bytes (including overhead) | ||
57 | |||
58 | Even a request for zero bytes (i.e., malloc(0)) returns a | ||
59 | pointer to something of the minimum allocatable size. | ||
60 | The maximum overhead wastage (i.e., number of extra bytes | ||
61 | allocated than were requested in malloc) is less than or equal | ||
62 | to the minimum size, except for requests >= mmap_threshold that | ||
63 | are serviced via mmap(), where the worst case wastage is about | ||
64 | 32 bytes plus the remainder from a system page (the minimal | ||
65 | mmap unit); typically 4096 or 8192 bytes. | ||
66 | |||
67 | Security: static-safe; optionally more or less | ||
68 | The "security" of malloc refers to the ability of malicious | ||
69 | code to accentuate the effects of errors (for example, freeing | ||
70 | space that is not currently malloc'ed or overwriting past the | ||
71 | ends of chunks) in code that calls malloc. This malloc | ||
72 | guarantees not to modify any memory locations below the base of | ||
73 | heap, i.e., static variables, even in the presence of usage | ||
74 | errors. The routines additionally detect most improper frees | ||
75 | and reallocs. All this holds as long as the static bookkeeping | ||
76 | for malloc itself is not corrupted by some other means. This | ||
77 | is only one aspect of security -- these checks do not, and | ||
78 | cannot, detect all possible programming errors. | ||
79 | |||
80 | If FOOTERS is defined nonzero, then each allocated chunk | ||
81 | carries an additional check word to verify that it was malloced | ||
82 | from its space. These check words are the same within each | ||
83 | execution of a program using malloc, but differ across | ||
84 | executions, so externally crafted fake chunks cannot be | ||
85 | freed. This improves security by rejecting frees/reallocs that | ||
86 | could corrupt heap memory, in addition to the checks preventing | ||
87 | writes to statics that are always on. This may further improve | ||
88 | security at the expense of time and space overhead. (Note that | ||
89 | FOOTERS may also be worth using with MSPACES.) | ||
90 | |||
91 | By default detected errors cause the program to abort (calling | ||
92 | "abort()"). You can override this to instead proceed past | ||
93 | errors by defining PROCEED_ON_ERROR. In this case, a bad free | ||
94 | has no effect, and a malloc that encounters a bad address | ||
95 | caused by user overwrites will ignore the bad address by | ||
96 | dropping pointers and indices to all known memory. This may | ||
97 | be appropriate for programs that should continue if at all | ||
98 | possible in the face of programming errors, although they may | ||
99 | run out of memory because dropped memory is never reclaimed. | ||
100 | |||
101 | If you don't like either of these options, you can define | ||
102 | CORRUPTION_ERROR_ACTION and USAGE_ERROR_ACTION to do anything | ||
103 | else. And if if you are sure that your program using malloc has | ||
104 | no errors or vulnerabilities, you can define INSECURE to 1, | ||
105 | which might (or might not) provide a small performance improvement. | ||
106 | |||
107 | Thread-safety: NOT thread-safe unless USE_LOCKS defined | ||
108 | When USE_LOCKS is defined, each public call to malloc, free, | ||
109 | etc is surrounded with either a pthread mutex or a win32 | ||
110 | spinlock (depending on WIN32). This is not especially fast, and | ||
111 | can be a major bottleneck. It is designed only to provide | ||
112 | minimal protection in concurrent environments, and to provide a | ||
113 | basis for extensions. If you are using malloc in a concurrent | ||
114 | program, consider instead using ptmalloc, which is derived from | ||
115 | a version of this malloc. (See http://www.malloc.de). | ||
116 | |||
117 | System requirements: Any combination of MORECORE and/or MMAP/MUNMAP | ||
118 | This malloc can use unix sbrk or any emulation (invoked using | ||
119 | the CALL_MORECORE macro) and/or mmap/munmap or any emulation | ||
120 | (invoked using CALL_MMAP/CALL_MUNMAP) to get and release system | ||
121 | memory. On most unix systems, it tends to work best if both | ||
122 | MORECORE and MMAP are enabled. On Win32, it uses emulations | ||
123 | based on VirtualAlloc. It also uses common C library functions | ||
124 | like memset. | ||
125 | |||
126 | Compliance: I believe it is compliant with the Single Unix Specification | ||
127 | (See http://www.unix.org). Also SVID/XPG, ANSI C, and probably | ||
128 | others as well. | ||
129 | |||
130 | * Overview of algorithms | ||
131 | |||
132 | This is not the fastest, most space-conserving, most portable, or | ||
133 | most tunable malloc ever written. However it is among the fastest | ||
134 | while also being among the most space-conserving, portable and | ||
135 | tunable. Consistent balance across these factors results in a good | ||
136 | general-purpose allocator for malloc-intensive programs. | ||
137 | |||
138 | In most ways, this malloc is a best-fit allocator. Generally, it | ||
139 | chooses the best-fitting existing chunk for a request, with ties | ||
140 | broken in approximately least-recently-used order. (This strategy | ||
141 | normally maintains low fragmentation.) However, for requests less | ||
142 | than 256bytes, it deviates from best-fit when there is not an | ||
143 | exactly fitting available chunk by preferring to use space adjacent | ||
144 | to that used for the previous small request, as well as by breaking | ||
145 | ties in approximately most-recently-used order. (These enhance | ||
146 | locality of series of small allocations.) And for very large requests | ||
147 | (>= 256Kb by default), it relies on system memory mapping | ||
148 | facilities, if supported. (This helps avoid carrying around and | ||
149 | possibly fragmenting memory used only for large chunks.) | ||
150 | |||
151 | All operations (except malloc_stats and mallinfo) have execution | ||
152 | times that are bounded by a constant factor of the number of bits in | ||
153 | a size_t, not counting any clearing in calloc or copying in realloc, | ||
154 | or actions surrounding MORECORE and MMAP that have times | ||
155 | proportional to the number of non-contiguous regions returned by | ||
156 | system allocation routines, which is often just 1. | ||
157 | |||
158 | The implementation is not very modular and seriously overuses | ||
159 | macros. Perhaps someday all C compilers will do as good a job | ||
160 | inlining modular code as can now be done by brute-force expansion, | ||
161 | but now, enough of them seem not to. | ||
162 | |||
163 | Some compilers issue a lot of warnings about code that is | ||
164 | dead/unreachable only on some platforms, and also about intentional | ||
165 | uses of negation on unsigned types. All known cases of each can be | ||
166 | ignored. | ||
167 | |||
168 | For a longer but out of date high-level description, see | ||
169 | http://gee.cs.oswego.edu/dl/html/malloc.html | ||
170 | |||
171 | * MSPACES | ||
172 | If MSPACES is defined, then in addition to malloc, free, etc., | ||
173 | this file also defines mspace_malloc, mspace_free, etc. These | ||
174 | are versions of malloc routines that take an "mspace" argument | ||
175 | obtained using create_mspace, to control all internal bookkeeping. | ||
176 | If ONLY_MSPACES is defined, only these versions are compiled. | ||
177 | So if you would like to use this allocator for only some allocations, | ||
178 | and your system malloc for others, you can compile with | ||
179 | ONLY_MSPACES and then do something like... | ||
180 | static mspace mymspace = create_mspace(0,0); // for example | ||
181 | #define mymalloc(bytes) mspace_malloc(mymspace, bytes) | ||
182 | |||
183 | (Note: If you only need one instance of an mspace, you can instead | ||
184 | use "USE_DL_PREFIX" to relabel the global malloc.) | ||
185 | |||
186 | You can similarly create thread-local allocators by storing | ||
187 | mspaces as thread-locals. For example: | ||
188 | static __thread mspace tlms = 0; | ||
189 | void* tlmalloc(size_t bytes) { | ||
190 | if (tlms == 0) tlms = create_mspace(0, 0); | ||
191 | return mspace_malloc(tlms, bytes); | ||
192 | } | ||
193 | void tlfree(void* mem) { mspace_free(tlms, mem); } | ||
194 | |||
195 | Unless FOOTERS is defined, each mspace is completely independent. | ||
196 | You cannot allocate from one and free to another (although | ||
197 | conformance is only weakly checked, so usage errors are not always | ||
198 | caught). If FOOTERS is defined, then each chunk carries around a tag | ||
199 | indicating its originating mspace, and frees are directed to their | ||
200 | originating spaces. | ||
201 | |||
202 | ------------------------- Compile-time options --------------------------- | ||
203 | |||
204 | Be careful in setting #define values for numerical constants of type | ||
205 | size_t. On some systems, literal values are not automatically extended | ||
206 | to size_t precision unless they are explicitly casted. | ||
207 | |||
208 | WIN32 default: defined if _WIN32 defined | ||
209 | Defining WIN32 sets up defaults for MS environment and compilers. | ||
210 | Otherwise defaults are for unix. | ||
211 | |||
212 | MALLOC_ALIGNMENT default: (size_t)8 | ||
213 | Controls the minimum alignment for malloc'ed chunks. It must be a | ||
214 | power of two and at least 8, even on machines for which smaller | ||
215 | alignments would suffice. It may be defined as larger than this | ||
216 | though. Note however that code and data structures are optimized for | ||
217 | the case of 8-byte alignment. | ||
218 | |||
219 | MSPACES default: 0 (false) | ||
220 | If true, compile in support for independent allocation spaces. | ||
221 | This is only supported if HAVE_MMAP is true. | ||
222 | |||
223 | ONLY_MSPACES default: 0 (false) | ||
224 | If true, only compile in mspace versions, not regular versions. | ||
225 | |||
226 | USE_LOCKS default: 0 (false) | ||
227 | Causes each call to each public routine to be surrounded with | ||
228 | pthread or WIN32 mutex lock/unlock. (If set true, this can be | ||
229 | overridden on a per-mspace basis for mspace versions.) | ||
230 | |||
231 | FOOTERS default: 0 | ||
232 | If true, provide extra checking and dispatching by placing | ||
233 | information in the footers of allocated chunks. This adds | ||
234 | space and time overhead. | ||
235 | |||
236 | INSECURE default: 0 | ||
237 | If true, omit checks for usage errors and heap space overwrites. | ||
238 | |||
239 | USE_DL_PREFIX default: NOT defined | ||
240 | Causes compiler to prefix all public routines with the string 'dl'. | ||
241 | This can be useful when you only want to use this malloc in one part | ||
242 | of a program, using your regular system malloc elsewhere. | ||
243 | |||
244 | ABORT default: defined as abort() | ||
245 | Defines how to abort on failed checks. On most systems, a failed | ||
246 | check cannot die with an "assert" or even print an informative | ||
247 | message, because the underlying print routines in turn call malloc, | ||
248 | which will fail again. Generally, the best policy is to simply call | ||
249 | abort(). It's not very useful to do more than this because many | ||
250 | errors due to overwriting will show up as address faults (null, odd | ||
251 | addresses etc) rather than malloc-triggered checks, so will also | ||
252 | abort. Also, most compilers know that abort() does not return, so | ||
253 | can better optimize code conditionally calling it. | ||
254 | |||
255 | PROCEED_ON_ERROR default: defined as 0 (false) | ||
256 | Controls whether detected bad addresses cause them to bypassed | ||
257 | rather than aborting. If set, detected bad arguments to free and | ||
258 | realloc are ignored. And all bookkeeping information is zeroed out | ||
259 | upon a detected overwrite of freed heap space, thus losing the | ||
260 | ability to ever return it from malloc again, but enabling the | ||
261 | application to proceed. If PROCEED_ON_ERROR is defined, the | ||
262 | static variable malloc_corruption_error_count is compiled in | ||
263 | and can be examined to see if errors have occurred. This option | ||
264 | generates slower code than the default abort policy. | ||
265 | |||
266 | DEBUG default: NOT defined | ||
267 | The DEBUG setting is mainly intended for people trying to modify | ||
268 | this code or diagnose problems when porting to new platforms. | ||
269 | However, it may also be able to better isolate user errors than just | ||
270 | using runtime checks. The assertions in the check routines spell | ||
271 | out in more detail the assumptions and invariants underlying the | ||
272 | algorithms. The checking is fairly extensive, and will slow down | ||
273 | execution noticeably. Calling malloc_stats or mallinfo with DEBUG | ||
274 | set will attempt to check every non-mmapped allocated and free chunk | ||
275 | in the course of computing the summaries. | ||
276 | |||
277 | ABORT_ON_ASSERT_FAILURE default: defined as 1 (true) | ||
278 | Debugging assertion failures can be nearly impossible if your | ||
279 | version of the assert macro causes malloc to be called, which will | ||
280 | lead to a cascade of further failures, blowing the runtime stack. | ||
281 | ABORT_ON_ASSERT_FAILURE cause assertions failures to call abort(), | ||
282 | which will usually make debugging easier. | ||
283 | |||
284 | MALLOC_FAILURE_ACTION default: sets errno to ENOMEM, or no-op on win32 | ||
285 | The action to take before "return 0" when malloc fails to be able to | ||
286 | return memory because there is none available. | ||
287 | |||
288 | HAVE_MORECORE default: 1 (true) unless win32 or ONLY_MSPACES | ||
289 | True if this system supports sbrk or an emulation of it. | ||
290 | |||
291 | MORECORE default: sbrk | ||
292 | The name of the sbrk-style system routine to call to obtain more | ||
293 | memory. See below for guidance on writing custom MORECORE | ||
294 | functions. The type of the argument to sbrk/MORECORE varies across | ||
295 | systems. It cannot be size_t, because it supports negative | ||
296 | arguments, so it is normally the signed type of the same width as | ||
297 | size_t (sometimes declared as "intptr_t"). It doesn't much matter | ||
298 | though. Internally, we only call it with arguments less than half | ||
299 | the max value of a size_t, which should work across all reasonable | ||
300 | possibilities, although sometimes generating compiler warnings. See | ||
301 | near the end of this file for guidelines for creating a custom | ||
302 | version of MORECORE. | ||
303 | |||
304 | MORECORE_CONTIGUOUS default: 1 (true) | ||
305 | If true, take advantage of fact that consecutive calls to MORECORE | ||
306 | with positive arguments always return contiguous increasing | ||
307 | addresses. This is true of unix sbrk. It does not hurt too much to | ||
308 | set it true anyway, since malloc copes with non-contiguities. | ||
309 | Setting it false when definitely non-contiguous saves time | ||
310 | and possibly wasted space it would take to discover this though. | ||
311 | |||
312 | MORECORE_CANNOT_TRIM default: NOT defined | ||
313 | True if MORECORE cannot release space back to the system when given | ||
314 | negative arguments. This is generally necessary only if you are | ||
315 | using a hand-crafted MORECORE function that cannot handle negative | ||
316 | arguments. | ||
317 | |||
318 | HAVE_MMAP default: 1 (true) | ||
319 | True if this system supports mmap or an emulation of it. If so, and | ||
320 | HAVE_MORECORE is not true, MMAP is used for all system | ||
321 | allocation. If set and HAVE_MORECORE is true as well, MMAP is | ||
322 | primarily used to directly allocate very large blocks. It is also | ||
323 | used as a backup strategy in cases where MORECORE fails to provide | ||
324 | space from system. Note: A single call to MUNMAP is assumed to be | ||
325 | able to unmap memory that may have be allocated using multiple calls | ||
326 | to MMAP, so long as they are adjacent. | ||
327 | |||
328 | HAVE_MREMAP default: 1 on linux, else 0 | ||
329 | If true realloc() uses mremap() to re-allocate large blocks and | ||
330 | extend or shrink allocation spaces. | ||
331 | |||
332 | MMAP_CLEARS default: 1 on unix | ||
333 | True if mmap clears memory so calloc doesn't need to. This is true | ||
334 | for standard unix mmap using /dev/zero. | ||
335 | |||
336 | USE_BUILTIN_FFS default: 0 (i.e., not used) | ||
337 | Causes malloc to use the builtin ffs() function to compute indices. | ||
338 | Some compilers may recognize and intrinsify ffs to be faster than the | ||
339 | supplied C version. Also, the case of x86 using gcc is special-cased | ||
340 | to an asm instruction, so is already as fast as it can be, and so | ||
341 | this setting has no effect. (On most x86s, the asm version is only | ||
342 | slightly faster than the C version.) | ||
343 | |||
344 | malloc_getpagesize default: derive from system includes, or 4096. | ||
345 | The system page size. To the extent possible, this malloc manages | ||
346 | memory from the system in page-size units. This may be (and | ||
347 | usually is) a function rather than a constant. This is ignored | ||
348 | if WIN32, where page size is determined using getSystemInfo during | ||
349 | initialization. | ||
350 | |||
351 | USE_DEV_RANDOM default: 0 (i.e., not used) | ||
352 | Causes malloc to use /dev/random to initialize secure magic seed for | ||
353 | stamping footers. Otherwise, the current time is used. | ||
354 | |||
355 | NO_MALLINFO default: 0 | ||
356 | If defined, don't compile "mallinfo". This can be a simple way | ||
357 | of dealing with mismatches between system declarations and | ||
358 | those in this file. | ||
359 | |||
360 | MALLINFO_FIELD_TYPE default: size_t | ||
361 | The type of the fields in the mallinfo struct. This was originally | ||
362 | defined as "int" in SVID etc, but is more usefully defined as | ||
363 | size_t. The value is used only if HAVE_USR_INCLUDE_MALLOC_H is not set | ||
364 | |||
365 | REALLOC_ZERO_BYTES_FREES default: not defined | ||
366 | This should be set if a call to realloc with zero bytes should | ||
367 | be the same as a call to free. Some people think it should. Otherwise, | ||
368 | since this malloc returns a unique pointer for malloc(0), so does | ||
369 | realloc(p, 0). | ||
370 | |||
371 | LACKS_UNISTD_H, LACKS_FCNTL_H, LACKS_SYS_PARAM_H, LACKS_SYS_MMAN_H | ||
372 | LACKS_STRINGS_H, LACKS_STRING_H, LACKS_SYS_TYPES_H, LACKS_ERRNO_H | ||
373 | LACKS_STDLIB_H default: NOT defined unless on WIN32 | ||
374 | Define these if your system does not have these header files. | ||
375 | You might need to manually insert some of the declarations they provide. | ||
376 | |||
377 | DEFAULT_GRANULARITY default: page size if MORECORE_CONTIGUOUS, | ||
378 | system_info.dwAllocationGranularity in WIN32, | ||
379 | otherwise 64K. | ||
380 | Also settable using mallopt(M_GRANULARITY, x) | ||
381 | The unit for allocating and deallocating memory from the system. On | ||
382 | most systems with contiguous MORECORE, there is no reason to | ||
383 | make this more than a page. However, systems with MMAP tend to | ||
384 | either require or encourage larger granularities. You can increase | ||
385 | this value to prevent system allocation functions to be called so | ||
386 | often, especially if they are slow. The value must be at least one | ||
387 | page and must be a power of two. Setting to 0 causes initialization | ||
388 | to either page size or win32 region size. (Note: In previous | ||
389 | versions of malloc, the equivalent of this option was called | ||
390 | "TOP_PAD") | ||
391 | |||
392 | DEFAULT_TRIM_THRESHOLD default: 2MB | ||
393 | Also settable using mallopt(M_TRIM_THRESHOLD, x) | ||
394 | The maximum amount of unused top-most memory to keep before | ||
395 | releasing via malloc_trim in free(). Automatic trimming is mainly | ||
396 | useful in long-lived programs using contiguous MORECORE. Because | ||
397 | trimming via sbrk can be slow on some systems, and can sometimes be | ||
398 | wasteful (in cases where programs immediately afterward allocate | ||
399 | more large chunks) the value should be high enough so that your | ||
400 | overall system performance would improve by releasing this much | ||
401 | memory. As a rough guide, you might set to a value close to the | ||
402 | average size of a process (program) running on your system. | ||
403 | Releasing this much memory would allow such a process to run in | ||
404 | memory. Generally, it is worth tuning trim thresholds when a | ||
405 | program undergoes phases where several large chunks are allocated | ||
406 | and released in ways that can reuse each other's storage, perhaps | ||
407 | mixed with phases where there are no such chunks at all. The trim | ||
408 | value must be greater than page size to have any useful effect. To | ||
409 | disable trimming completely, you can set to MAX_SIZE_T. Note that the trick | ||
410 | some people use of mallocing a huge space and then freeing it at | ||
411 | program startup, in an attempt to reserve system memory, doesn't | ||
412 | have the intended effect under automatic trimming, since that memory | ||
413 | will immediately be returned to the system. | ||
414 | |||
415 | DEFAULT_MMAP_THRESHOLD default: 256K | ||
416 | Also settable using mallopt(M_MMAP_THRESHOLD, x) | ||
417 | The request size threshold for using MMAP to directly service a | ||
418 | request. Requests of at least this size that cannot be allocated | ||
419 | using already-existing space will be serviced via mmap. (If enough | ||
420 | normal freed space already exists it is used instead.) Using mmap | ||
421 | segregates relatively large chunks of memory so that they can be | ||
422 | individually obtained and released from the host system. A request | ||
423 | serviced through mmap is never reused by any other request (at least | ||
424 | not directly; the system may just so happen to remap successive | ||
425 | requests to the same locations). Segregating space in this way has | ||
426 | the benefits that: Mmapped space can always be individually released | ||
427 | back to the system, which helps keep the system level memory demands | ||
428 | of a long-lived program low. Also, mapped memory doesn't become | ||
429 | `locked' between other chunks, as can happen with normally allocated | ||
430 | chunks, which means that even trimming via malloc_trim would not | ||
431 | release them. However, it has the disadvantage that the space | ||
432 | cannot be reclaimed, consolidated, and then used to service later | ||
433 | requests, as happens with normal chunks. The advantages of mmap | ||
434 | nearly always outweigh disadvantages for "large" chunks, but the | ||
435 | value of "large" may vary across systems. The default is an | ||
436 | empirically derived value that works well in most systems. You can | ||
437 | disable mmap by setting to MAX_SIZE_T. | ||
438 | |||
439 | */ | ||
440 | |||
441 | #ifndef WIN32 | ||
442 | #ifdef _WIN32 | ||
443 | #define WIN32 1 | ||
444 | #endif /* _WIN32 */ | ||
445 | #endif /* WIN32 */ | ||
446 | #ifdef WIN32 | ||
447 | #define WIN32_LEAN_AND_MEAN | ||
448 | #include <windows.h> | ||
449 | #define HAVE_MMAP 1 | ||
450 | #define HAVE_MORECORE 0 | ||
451 | #define LACKS_UNISTD_H | ||
452 | #define LACKS_SYS_PARAM_H | ||
453 | #define LACKS_SYS_MMAN_H | ||
454 | #define LACKS_STRING_H | ||
455 | #define LACKS_STRINGS_H | ||
456 | #define LACKS_SYS_TYPES_H | ||
457 | #define LACKS_ERRNO_H | ||
458 | #define MALLOC_FAILURE_ACTION | ||
459 | #define MMAP_CLEARS 0 /* WINCE and some others apparently don't clear */ | ||
460 | #endif /* WIN32 */ | ||
461 | |||
462 | #if defined(DARWIN) || defined(_DARWIN) | ||
463 | /* Mac OSX docs advise not to use sbrk; it seems better to use mmap */ | ||
464 | #ifndef HAVE_MORECORE | ||
465 | #define HAVE_MORECORE 0 | ||
466 | #define HAVE_MMAP 1 | ||
467 | #endif /* HAVE_MORECORE */ | ||
468 | #endif /* DARWIN */ | ||
469 | |||
470 | #ifndef LACKS_SYS_TYPES_H | ||
471 | #include <sys/types.h> /* For size_t */ | ||
472 | #endif /* LACKS_SYS_TYPES_H */ | ||
473 | |||
474 | /* The maximum possible size_t value has all bits set */ | ||
475 | #define MAX_SIZE_T (~(size_t)0) | ||
476 | |||
477 | #ifndef ONLY_MSPACES | ||
478 | #define ONLY_MSPACES 0 | ||
479 | #endif /* ONLY_MSPACES */ | ||
480 | #ifndef MSPACES | ||
481 | #if ONLY_MSPACES | ||
482 | #define MSPACES 1 | ||
483 | #else /* ONLY_MSPACES */ | ||
484 | #define MSPACES 0 | ||
485 | #endif /* ONLY_MSPACES */ | ||
486 | #endif /* MSPACES */ | ||
487 | #ifndef MALLOC_ALIGNMENT | ||
488 | #define MALLOC_ALIGNMENT ((size_t)8U) | ||
489 | #endif /* MALLOC_ALIGNMENT */ | ||
490 | #ifndef FOOTERS | ||
491 | #define FOOTERS 0 | ||
492 | #endif /* FOOTERS */ | ||
493 | #ifndef ABORT | ||
494 | #define ABORT abort() | ||
495 | #endif /* ABORT */ | ||
496 | #ifndef ABORT_ON_ASSERT_FAILURE | ||
497 | #define ABORT_ON_ASSERT_FAILURE 1 | ||
498 | #endif /* ABORT_ON_ASSERT_FAILURE */ | ||
499 | #ifndef PROCEED_ON_ERROR | ||
500 | #define PROCEED_ON_ERROR 0 | ||
501 | #endif /* PROCEED_ON_ERROR */ | ||
502 | #ifndef USE_LOCKS | ||
503 | #define USE_LOCKS 0 | ||
504 | #endif /* USE_LOCKS */ | ||
505 | #ifndef INSECURE | ||
506 | #define INSECURE 0 | ||
507 | #endif /* INSECURE */ | ||
508 | #ifndef HAVE_MMAP | ||
509 | #define HAVE_MMAP 1 | ||
510 | #endif /* HAVE_MMAP */ | ||
511 | #ifndef MMAP_CLEARS | ||
512 | #define MMAP_CLEARS 1 | ||
513 | #endif /* MMAP_CLEARS */ | ||
514 | #ifndef HAVE_MREMAP | ||
515 | #ifdef linux | ||
516 | #define HAVE_MREMAP 1 | ||
517 | #else /* linux */ | ||
518 | #define HAVE_MREMAP 0 | ||
519 | #endif /* linux */ | ||
520 | #endif /* HAVE_MREMAP */ | ||
521 | #ifndef MALLOC_FAILURE_ACTION | ||
522 | #define MALLOC_FAILURE_ACTION errno = ENOMEM; | ||
523 | #endif /* MALLOC_FAILURE_ACTION */ | ||
524 | #ifndef HAVE_MORECORE | ||
525 | #if ONLY_MSPACES | ||
526 | #define HAVE_MORECORE 0 | ||
527 | #else /* ONLY_MSPACES */ | ||
528 | #define HAVE_MORECORE 1 | ||
529 | #endif /* ONLY_MSPACES */ | ||
530 | #endif /* HAVE_MORECORE */ | ||
531 | #if !HAVE_MORECORE | ||
532 | #define MORECORE_CONTIGUOUS 0 | ||
533 | #else /* !HAVE_MORECORE */ | ||
534 | #ifndef MORECORE | ||
535 | #define MORECORE sbrk | ||
536 | #endif /* MORECORE */ | ||
537 | #ifndef MORECORE_CONTIGUOUS | ||
538 | #define MORECORE_CONTIGUOUS 1 | ||
539 | #endif /* MORECORE_CONTIGUOUS */ | ||
540 | #endif /* HAVE_MORECORE */ | ||
541 | #ifndef DEFAULT_GRANULARITY | ||
542 | #if MORECORE_CONTIGUOUS | ||
543 | #define DEFAULT_GRANULARITY (0) /* 0 means to compute in init_mparams */ | ||
544 | #else /* MORECORE_CONTIGUOUS */ | ||
545 | #define DEFAULT_GRANULARITY ((size_t)64U * (size_t)1024U) | ||
546 | #endif /* MORECORE_CONTIGUOUS */ | ||
547 | #endif /* DEFAULT_GRANULARITY */ | ||
548 | #ifndef DEFAULT_TRIM_THRESHOLD | ||
549 | #ifndef MORECORE_CANNOT_TRIM | ||
550 | #define DEFAULT_TRIM_THRESHOLD ((size_t)2U * (size_t)1024U * (size_t)1024U) | ||
551 | #else /* MORECORE_CANNOT_TRIM */ | ||
552 | #define DEFAULT_TRIM_THRESHOLD MAX_SIZE_T | ||
553 | #endif /* MORECORE_CANNOT_TRIM */ | ||
554 | #endif /* DEFAULT_TRIM_THRESHOLD */ | ||
555 | #ifndef DEFAULT_MMAP_THRESHOLD | ||
556 | #if HAVE_MMAP | ||
557 | #define DEFAULT_MMAP_THRESHOLD ((size_t)256U * (size_t)1024U) | ||
558 | #else /* HAVE_MMAP */ | ||
559 | #define DEFAULT_MMAP_THRESHOLD MAX_SIZE_T | ||
560 | #endif /* HAVE_MMAP */ | ||
561 | #endif /* DEFAULT_MMAP_THRESHOLD */ | ||
562 | #ifndef USE_BUILTIN_FFS | ||
563 | #define USE_BUILTIN_FFS 0 | ||
564 | #endif /* USE_BUILTIN_FFS */ | ||
565 | #ifndef USE_DEV_RANDOM | ||
566 | #define USE_DEV_RANDOM 0 | ||
567 | #endif /* USE_DEV_RANDOM */ | ||
568 | #ifndef NO_MALLINFO | ||
569 | #define NO_MALLINFO 0 | ||
570 | #endif /* NO_MALLINFO */ | ||
571 | #ifndef MALLINFO_FIELD_TYPE | ||
572 | #define MALLINFO_FIELD_TYPE size_t | ||
573 | #endif /* MALLINFO_FIELD_TYPE */ | ||
574 | |||
575 | /* | ||
576 | mallopt tuning options. SVID/XPG defines four standard parameter | ||
577 | numbers for mallopt, normally defined in malloc.h. None of these | ||
578 | are used in this malloc, so setting them has no effect. But this | ||
579 | malloc does support the following options. | ||
580 | */ | ||
581 | |||
582 | #define M_TRIM_THRESHOLD (-1) | ||
583 | #define M_GRANULARITY (-2) | ||
584 | #define M_MMAP_THRESHOLD (-3) | ||
585 | |||
586 | /* ------------------------ Mallinfo declarations ------------------------ */ | ||
587 | |||
588 | #if !NO_MALLINFO | ||
589 | /* | ||
590 | This version of malloc supports the standard SVID/XPG mallinfo | ||
591 | routine that returns a struct containing usage properties and | ||
592 | statistics. It should work on any system that has a | ||
593 | /usr/include/malloc.h defining struct mallinfo. The main | ||
594 | declaration needed is the mallinfo struct that is returned (by-copy) | ||
595 | by mallinfo(). The malloinfo struct contains a bunch of fields that | ||
596 | are not even meaningful in this version of malloc. These fields are | ||
597 | are instead filled by mallinfo() with other numbers that might be of | ||
598 | interest. | ||
599 | |||
600 | HAVE_USR_INCLUDE_MALLOC_H should be set if you have a | ||
601 | /usr/include/malloc.h file that includes a declaration of struct | ||
602 | mallinfo. If so, it is included; else a compliant version is | ||
603 | declared below. These must be precisely the same for mallinfo() to | ||
604 | work. The original SVID version of this struct, defined on most | ||
605 | systems with mallinfo, declares all fields as ints. But some others | ||
606 | define as unsigned long. If your system defines the fields using a | ||
607 | type of different width than listed here, you MUST #include your | ||
608 | system version and #define HAVE_USR_INCLUDE_MALLOC_H. | ||
609 | */ | ||
610 | |||
611 | /* #define HAVE_USR_INCLUDE_MALLOC_H */ | ||
612 | |||
613 | #ifdef HAVE_USR_INCLUDE_MALLOC_H | ||
614 | #include "/usr/include/malloc.h" | ||
615 | #else /* HAVE_USR_INCLUDE_MALLOC_H */ | ||
616 | |||
617 | struct mallinfo { | ||
618 | MALLINFO_FIELD_TYPE arena; /* non-mmapped space allocated from system */ | ||
619 | MALLINFO_FIELD_TYPE ordblks; /* number of free chunks */ | ||
620 | MALLINFO_FIELD_TYPE smblks; /* always 0 */ | ||
621 | MALLINFO_FIELD_TYPE hblks; /* always 0 */ | ||
622 | MALLINFO_FIELD_TYPE hblkhd; /* space in mmapped regions */ | ||
623 | MALLINFO_FIELD_TYPE usmblks; /* maximum total allocated space */ | ||
624 | MALLINFO_FIELD_TYPE fsmblks; /* always 0 */ | ||
625 | MALLINFO_FIELD_TYPE uordblks; /* total allocated space */ | ||
626 | MALLINFO_FIELD_TYPE fordblks; /* total free space */ | ||
627 | MALLINFO_FIELD_TYPE keepcost; /* releasable (via malloc_trim) space */ | ||
628 | }; | ||
629 | |||
630 | #endif /* HAVE_USR_INCLUDE_MALLOC_H */ | ||
631 | #endif /* NO_MALLINFO */ | ||
632 | |||
633 | #ifdef __cplusplus | ||
634 | extern "C" { | ||
635 | #endif /* __cplusplus */ | ||
636 | |||
637 | #if !ONLY_MSPACES | ||
638 | |||
639 | /* ------------------- Declarations of public routines ------------------- */ | ||
640 | |||
641 | #ifndef USE_DL_PREFIX | ||
642 | #define dlcalloc calloc | ||
643 | #define dlfree free | ||
644 | #define dlmalloc malloc | ||
645 | #define dlmemalign memalign | ||
646 | #define dlrealloc realloc | ||
647 | #define dlvalloc valloc | ||
648 | #define dlpvalloc pvalloc | ||
649 | #define dlmallinfo mallinfo | ||
650 | #define dlmallopt mallopt | ||
651 | #define dlmalloc_trim malloc_trim | ||
652 | #define dlmalloc_stats malloc_stats | ||
653 | #define dlmalloc_usable_size malloc_usable_size | ||
654 | #define dlmalloc_footprint malloc_footprint | ||
655 | #define dlmalloc_max_footprint malloc_max_footprint | ||
656 | #define dlindependent_calloc independent_calloc | ||
657 | #define dlindependent_comalloc independent_comalloc | ||
658 | #endif /* USE_DL_PREFIX */ | ||
659 | |||
660 | |||
661 | /* | ||
662 | malloc(size_t n) | ||
663 | Returns a pointer to a newly allocated chunk of at least n bytes, or | ||
664 | null if no space is available, in which case errno is set to ENOMEM | ||
665 | on ANSI C systems. | ||
666 | |||
667 | If n is zero, malloc returns a minimum-sized chunk. (The minimum | ||
668 | size is 16 bytes on most 32bit systems, and 32 bytes on 64bit | ||
669 | systems.) Note that size_t is an unsigned type, so calls with | ||
670 | arguments that would be negative if signed are interpreted as | ||
671 | requests for huge amounts of space, which will often fail. The | ||
672 | maximum supported value of n differs across systems, but is in all | ||
673 | cases less than the maximum representable value of a size_t. | ||
674 | */ | ||
675 | void* dlmalloc(size_t); | ||
676 | |||
677 | /* | ||
678 | free(void* p) | ||
679 | Releases the chunk of memory pointed to by p, that had been previously | ||
680 | allocated using malloc or a related routine such as realloc. | ||
681 | It has no effect if p is null. If p was not malloced or already | ||
682 | freed, free(p) will by default cause the current program to abort. | ||
683 | */ | ||
684 | void dlfree(void*); | ||
685 | |||
686 | /* | ||
687 | calloc(size_t n_elements, size_t element_size); | ||
688 | Returns a pointer to n_elements * element_size bytes, with all locations | ||
689 | set to zero. | ||
690 | */ | ||
691 | void* dlcalloc(size_t, size_t); | ||
692 | |||
693 | /* | ||
694 | realloc(void* p, size_t n) | ||
695 | Returns a pointer to a chunk of size n that contains the same data | ||
696 | as does chunk p up to the minimum of (n, p's size) bytes, or null | ||
697 | if no space is available. | ||
698 | |||
699 | The returned pointer may or may not be the same as p. The algorithm | ||
700 | prefers extending p in most cases when possible, otherwise it | ||
701 | employs the equivalent of a malloc-copy-free sequence. | ||
702 | |||
703 | If p is null, realloc is equivalent to malloc. | ||
704 | |||
705 | If space is not available, realloc returns null, errno is set (if on | ||
706 | ANSI) and p is NOT freed. | ||
707 | |||
708 | if n is for fewer bytes than already held by p, the newly unused | ||
709 | space is lopped off and freed if possible. realloc with a size | ||
710 | argument of zero (re)allocates a minimum-sized chunk. | ||
711 | |||
712 | The old unix realloc convention of allowing the last-free'd chunk | ||
713 | to be used as an argument to realloc is not supported. | ||
714 | */ | ||
715 | |||
716 | void* dlrealloc(void*, size_t); | ||
717 | |||
718 | /* | ||
719 | memalign(size_t alignment, size_t n); | ||
720 | Returns a pointer to a newly allocated chunk of n bytes, aligned | ||
721 | in accord with the alignment argument. | ||
722 | |||
723 | The alignment argument should be a power of two. If the argument is | ||
724 | not a power of two, the nearest greater power is used. | ||
725 | 8-byte alignment is guaranteed by normal malloc calls, so don't | ||
726 | bother calling memalign with an argument of 8 or less. | ||
727 | |||
728 | Overreliance on memalign is a sure way to fragment space. | ||
729 | */ | ||
730 | void* dlmemalign(size_t, size_t); | ||
731 | |||
732 | /* | ||
733 | valloc(size_t n); | ||
734 | Equivalent to memalign(pagesize, n), where pagesize is the page | ||
735 | size of the system. If the pagesize is unknown, 4096 is used. | ||
736 | */ | ||
737 | void* dlvalloc(size_t); | ||
738 | |||
739 | /* | ||
740 | mallopt(int parameter_number, int parameter_value) | ||
741 | Sets tunable parameters The format is to provide a | ||
742 | (parameter-number, parameter-value) pair. mallopt then sets the | ||
743 | corresponding parameter to the argument value if it can (i.e., so | ||
744 | long as the value is meaningful), and returns 1 if successful else | ||
745 | 0. SVID/XPG/ANSI defines four standard param numbers for mallopt, | ||
746 | normally defined in malloc.h. None of these are use in this malloc, | ||
747 | so setting them has no effect. But this malloc also supports other | ||
748 | options in mallopt. See below for details. Briefly, supported | ||
749 | parameters are as follows (listed defaults are for "typical" | ||
750 | configurations). | ||
751 | |||
752 | Symbol param # default allowed param values | ||
753 | M_TRIM_THRESHOLD -1 2*1024*1024 any (MAX_SIZE_T disables) | ||
754 | M_GRANULARITY -2 page size any power of 2 >= page size | ||
755 | M_MMAP_THRESHOLD -3 256*1024 any (or 0 if no MMAP support) | ||
756 | */ | ||
757 | int dlmallopt(int, int); | ||
758 | |||
759 | /* | ||
760 | malloc_footprint(); | ||
761 | Returns the number of bytes obtained from the system. The total | ||
762 | number of bytes allocated by malloc, realloc etc., is less than this | ||
763 | value. Unlike mallinfo, this function returns only a precomputed | ||
764 | result, so can be called frequently to monitor memory consumption. | ||
765 | Even if locks are otherwise defined, this function does not use them, | ||
766 | so results might not be up to date. | ||
767 | */ | ||
768 | size_t dlmalloc_footprint(void); | ||
769 | |||
770 | /* | ||
771 | malloc_max_footprint(); | ||
772 | Returns the maximum number of bytes obtained from the system. This | ||
773 | value will be greater than current footprint if deallocated space | ||
774 | has been reclaimed by the system. The peak number of bytes allocated | ||
775 | by malloc, realloc etc., is less than this value. Unlike mallinfo, | ||
776 | this function returns only a precomputed result, so can be called | ||
777 | frequently to monitor memory consumption. Even if locks are | ||
778 | otherwise defined, this function does not use them, so results might | ||
779 | not be up to date. | ||
780 | */ | ||
781 | size_t dlmalloc_max_footprint(void); | ||
782 | |||
783 | #if !NO_MALLINFO | ||
784 | /* | ||
785 | mallinfo() | ||
786 | Returns (by copy) a struct containing various summary statistics: | ||
787 | |||
788 | arena: current total non-mmapped bytes allocated from system | ||
789 | ordblks: the number of free chunks | ||
790 | smblks: always zero. | ||
791 | hblks: current number of mmapped regions | ||
792 | hblkhd: total bytes held in mmapped regions | ||
793 | usmblks: the maximum total allocated space. This will be greater | ||
794 | than current total if trimming has occurred. | ||
795 | fsmblks: always zero | ||
796 | uordblks: current total allocated space (normal or mmapped) | ||
797 | fordblks: total free space | ||
798 | keepcost: the maximum number of bytes that could ideally be released | ||
799 | back to system via malloc_trim. ("ideally" means that | ||
800 | it ignores page restrictions etc.) | ||
801 | |||
802 | Because these fields are ints, but internal bookkeeping may | ||
803 | be kept as longs, the reported values may wrap around zero and | ||
804 | thus be inaccurate. | ||
805 | */ | ||
806 | struct mallinfo dlmallinfo(void); | ||
807 | #endif /* NO_MALLINFO */ | ||
808 | |||
809 | /* | ||
810 | independent_calloc(size_t n_elements, size_t element_size, void* chunks[]); | ||
811 | |||
812 | independent_calloc is similar to calloc, but instead of returning a | ||
813 | single cleared space, it returns an array of pointers to n_elements | ||
814 | independent elements that can hold contents of size elem_size, each | ||
815 | of which starts out cleared, and can be independently freed, | ||
816 | realloc'ed etc. The elements are guaranteed to be adjacently | ||
817 | allocated (this is not guaranteed to occur with multiple callocs or | ||
818 | mallocs), which may also improve cache locality in some | ||
819 | applications. | ||
820 | |||
821 | The "chunks" argument is optional (i.e., may be null, which is | ||
822 | probably the most typical usage). If it is null, the returned array | ||
823 | is itself dynamically allocated and should also be freed when it is | ||
824 | no longer needed. Otherwise, the chunks array must be of at least | ||
825 | n_elements in length. It is filled in with the pointers to the | ||
826 | chunks. | ||
827 | |||
828 | In either case, independent_calloc returns this pointer array, or | ||
829 | null if the allocation failed. If n_elements is zero and "chunks" | ||
830 | is null, it returns a chunk representing an array with zero elements | ||
831 | (which should be freed if not wanted). | ||
832 | |||
833 | Each element must be individually freed when it is no longer | ||
834 | needed. If you'd like to instead be able to free all at once, you | ||
835 | should instead use regular calloc and assign pointers into this | ||
836 | space to represent elements. (In this case though, you cannot | ||
837 | independently free elements.) | ||
838 | |||
839 | independent_calloc simplifies and speeds up implementations of many | ||
840 | kinds of pools. It may also be useful when constructing large data | ||
841 | structures that initially have a fixed number of fixed-sized nodes, | ||
842 | but the number is not known at compile time, and some of the nodes | ||
843 | may later need to be freed. For example: | ||
844 | |||
845 | struct Node { int item; struct Node* next; }; | ||
846 | |||
847 | struct Node* build_list() { | ||
848 | struct Node** pool; | ||
849 | int n = read_number_of_nodes_needed(); | ||
850 | if (n <= 0) return 0; | ||
851 | pool = (struct Node**)(independent_calloc(n, sizeof(struct Node), 0); | ||
852 | if (pool == 0) die(); | ||
853 | // organize into a linked list... | ||
854 | struct Node* first = pool[0]; | ||
855 | for (i = 0; i < n-1; ++i) | ||
856 | pool[i]->next = pool[i+1]; | ||
857 | free(pool); // Can now free the array (or not, if it is needed later) | ||
858 | return first; | ||
859 | } | ||
860 | */ | ||
861 | void** dlindependent_calloc(size_t, size_t, void**); | ||
862 | |||
863 | /* | ||
864 | independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]); | ||
865 | |||
866 | independent_comalloc allocates, all at once, a set of n_elements | ||
867 | chunks with sizes indicated in the "sizes" array. It returns | ||
868 | an array of pointers to these elements, each of which can be | ||
869 | independently freed, realloc'ed etc. The elements are guaranteed to | ||
870 | be adjacently allocated (this is not guaranteed to occur with | ||
871 | multiple callocs or mallocs), which may also improve cache locality | ||
872 | in some applications. | ||
873 | |||
874 | The "chunks" argument is optional (i.e., may be null). If it is null | ||
875 | the returned array is itself dynamically allocated and should also | ||
876 | be freed when it is no longer needed. Otherwise, the chunks array | ||
877 | must be of at least n_elements in length. It is filled in with the | ||
878 | pointers to the chunks. | ||
879 | |||
880 | In either case, independent_comalloc returns this pointer array, or | ||
881 | null if the allocation failed. If n_elements is zero and chunks is | ||
882 | null, it returns a chunk representing an array with zero elements | ||
883 | (which should be freed if not wanted). | ||
884 | |||
885 | Each element must be individually freed when it is no longer | ||
886 | needed. If you'd like to instead be able to free all at once, you | ||
887 | should instead use a single regular malloc, and assign pointers at | ||
888 | particular offsets in the aggregate space. (In this case though, you | ||
889 | cannot independently free elements.) | ||
890 | |||
891 | independent_comallac differs from independent_calloc in that each | ||
892 | element may have a different size, and also that it does not | ||
893 | automatically clear elements. | ||
894 | |||
895 | independent_comalloc can be used to speed up allocation in cases | ||
896 | where several structs or objects must always be allocated at the | ||
897 | same time. For example: | ||
898 | |||
899 | struct Head { ... } | ||
900 | struct Foot { ... } | ||
901 | |||
902 | void send_message(char* msg) { | ||
903 | int msglen = strlen(msg); | ||
904 | size_t sizes[3] = { sizeof(struct Head), msglen, sizeof(struct Foot) }; | ||
905 | void* chunks[3]; | ||
906 | if (independent_comalloc(3, sizes, chunks) == 0) | ||
907 | die(); | ||
908 | struct Head* head = (struct Head*)(chunks[0]); | ||
909 | char* body = (char*)(chunks[1]); | ||
910 | struct Foot* foot = (struct Foot*)(chunks[2]); | ||
911 | // ... | ||
912 | } | ||
913 | |||
914 | In general though, independent_comalloc is worth using only for | ||
915 | larger values of n_elements. For small values, you probably won't | ||
916 | detect enough difference from series of malloc calls to bother. | ||
917 | |||
918 | Overuse of independent_comalloc can increase overall memory usage, | ||
919 | since it cannot reuse existing noncontiguous small chunks that | ||
920 | might be available for some of the elements. | ||
921 | */ | ||
922 | void** dlindependent_comalloc(size_t, size_t*, void**); | ||
923 | |||
924 | |||
925 | /* | ||
926 | pvalloc(size_t n); | ||
927 | Equivalent to valloc(minimum-page-that-holds(n)), that is, | ||
928 | round up n to nearest pagesize. | ||
929 | */ | ||
930 | void* dlpvalloc(size_t); | ||
931 | |||
932 | /* | ||
933 | malloc_trim(size_t pad); | ||
934 | |||
935 | If possible, gives memory back to the system (via negative arguments | ||
936 | to sbrk) if there is unused memory at the `high' end of the malloc | ||
937 | pool or in unused MMAP segments. You can call this after freeing | ||
938 | large blocks of memory to potentially reduce the system-level memory | ||
939 | requirements of a program. However, it cannot guarantee to reduce | ||
940 | memory. Under some allocation patterns, some large free blocks of | ||
941 | memory will be locked between two used chunks, so they cannot be | ||
942 | given back to the system. | ||
943 | |||
944 | The `pad' argument to malloc_trim represents the amount of free | ||
945 | trailing space to leave untrimmed. If this argument is zero, only | ||
946 | the minimum amount of memory to maintain internal data structures | ||
947 | will be left. Non-zero arguments can be supplied to maintain enough | ||
948 | trailing space to service future expected allocations without having | ||
949 | to re-obtain memory from the system. | ||
950 | |||
951 | Malloc_trim returns 1 if it actually released any memory, else 0. | ||
952 | */ | ||
953 | int dlmalloc_trim(size_t); | ||
954 | |||
955 | /* | ||
956 | malloc_usable_size(void* p); | ||
957 | |||
958 | Returns the number of bytes you can actually use in | ||
959 | an allocated chunk, which may be more than you requested (although | ||
960 | often not) due to alignment and minimum size constraints. | ||
961 | You can use this many bytes without worrying about | ||
962 | overwriting other allocated objects. This is not a particularly great | ||
963 | programming practice. malloc_usable_size can be more useful in | ||
964 | debugging and assertions, for example: | ||
965 | |||
966 | p = malloc(n); | ||
967 | assert(malloc_usable_size(p) >= 256); | ||
968 | */ | ||
969 | size_t dlmalloc_usable_size(void*); | ||
970 | |||
971 | /* | ||
972 | malloc_stats(); | ||
973 | Prints on stderr the amount of space obtained from the system (both | ||
974 | via sbrk and mmap), the maximum amount (which may be more than | ||
975 | current if malloc_trim and/or munmap got called), and the current | ||
976 | number of bytes allocated via malloc (or realloc, etc) but not yet | ||
977 | freed. Note that this is the number of bytes allocated, not the | ||
978 | number requested. It will be larger than the number requested | ||
979 | because of alignment and bookkeeping overhead. Because it includes | ||
980 | alignment wastage as being in use, this figure may be greater than | ||
981 | zero even when no user-level chunks are allocated. | ||
982 | |||
983 | The reported current and maximum system memory can be inaccurate if | ||
984 | a program makes other calls to system memory allocation functions | ||
985 | (normally sbrk) outside of malloc. | ||
986 | |||
987 | malloc_stats prints only the most commonly interesting statistics. | ||
988 | More information can be obtained by calling mallinfo. | ||
989 | */ | ||
990 | void dlmalloc_stats(void); | ||
991 | |||
992 | #endif /* ONLY_MSPACES */ | ||
993 | |||
994 | #if MSPACES | ||
995 | |||
996 | /* | ||
997 | mspace is an opaque type representing an independent | ||
998 | region of space that supports mspace_malloc, etc. | ||
999 | */ | ||
1000 | typedef void* mspace; | ||
1001 | |||
1002 | /* | ||
1003 | create_mspace creates and returns a new independent space with the | ||
1004 | given initial capacity, or, if 0, the default granularity size. It | ||
1005 | returns null if there is no system memory available to create the | ||
1006 | space. If argument locked is non-zero, the space uses a separate | ||
1007 | lock to control access. The capacity of the space will grow | ||
1008 | dynamically as needed to service mspace_malloc requests. You can | ||
1009 | control the sizes of incremental increases of this space by | ||
1010 | compiling with a different DEFAULT_GRANULARITY or dynamically | ||
1011 | setting with mallopt(M_GRANULARITY, value). | ||
1012 | */ | ||
1013 | mspace create_mspace(size_t capacity, int locked); | ||
1014 | |||
1015 | /* | ||
1016 | destroy_mspace destroys the given space, and attempts to return all | ||
1017 | of its memory back to the system, returning the total number of | ||
1018 | bytes freed. After destruction, the results of access to all memory | ||
1019 | used by the space become undefined. | ||
1020 | */ | ||
1021 | size_t destroy_mspace(mspace msp); | ||
1022 | |||
1023 | /* | ||
1024 | create_mspace_with_base uses the memory supplied as the initial base | ||
1025 | of a new mspace. Part (less than 128*sizeof(size_t) bytes) of this | ||
1026 | space is used for bookkeeping, so the capacity must be at least this | ||
1027 | large. (Otherwise 0 is returned.) When this initial space is | ||
1028 | exhausted, additional memory will be obtained from the system. | ||
1029 | Destroying this space will deallocate all additionally allocated | ||
1030 | space (if possible) but not the initial base. | ||
1031 | */ | ||
1032 | mspace create_mspace_with_base(void* base, size_t capacity, int locked); | ||
1033 | |||
1034 | /* | ||
1035 | mspace_malloc behaves as malloc, but operates within | ||
1036 | the given space. | ||
1037 | */ | ||
1038 | void* mspace_malloc(mspace msp, size_t bytes); | ||
1039 | |||
1040 | /* | ||
1041 | mspace_free behaves as free, but operates within | ||
1042 | the given space. | ||
1043 | |||
1044 | If compiled with FOOTERS==1, mspace_free is not actually needed. | ||
1045 | free may be called instead of mspace_free because freed chunks from | ||
1046 | any space are handled by their originating spaces. | ||
1047 | */ | ||
1048 | void mspace_free(mspace msp, void* mem); | ||
1049 | |||
1050 | /* | ||
1051 | mspace_realloc behaves as realloc, but operates within | ||
1052 | the given space. | ||
1053 | |||
1054 | If compiled with FOOTERS==1, mspace_realloc is not actually | ||
1055 | needed. realloc may be called instead of mspace_realloc because | ||
1056 | realloced chunks from any space are handled by their originating | ||
1057 | spaces. | ||
1058 | */ | ||
1059 | void* mspace_realloc(mspace msp, void* mem, size_t newsize); | ||
1060 | |||
1061 | /* | ||
1062 | mspace_calloc behaves as calloc, but operates within | ||
1063 | the given space. | ||
1064 | */ | ||
1065 | void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size); | ||
1066 | |||
1067 | /* | ||
1068 | mspace_memalign behaves as memalign, but operates within | ||
1069 | the given space. | ||
1070 | */ | ||
1071 | void* mspace_memalign(mspace msp, size_t alignment, size_t bytes); | ||
1072 | |||
1073 | /* | ||
1074 | mspace_independent_calloc behaves as independent_calloc, but | ||
1075 | operates within the given space. | ||
1076 | */ | ||
1077 | void** mspace_independent_calloc(mspace msp, size_t n_elements, | ||
1078 | size_t elem_size, void* chunks[]); | ||
1079 | |||
1080 | /* | ||
1081 | mspace_independent_comalloc behaves as independent_comalloc, but | ||
1082 | operates within the given space. | ||
1083 | */ | ||
1084 | void** mspace_independent_comalloc(mspace msp, size_t n_elements, | ||
1085 | size_t sizes[], void* chunks[]); | ||
1086 | |||
1087 | /* | ||
1088 | mspace_footprint() returns the number of bytes obtained from the | ||
1089 | system for this space. | ||
1090 | */ | ||
1091 | size_t mspace_footprint(mspace msp); | ||
1092 | |||
1093 | /* | ||
1094 | mspace_max_footprint() returns the peak number of bytes obtained from the | ||
1095 | system for this space. | ||
1096 | */ | ||
1097 | size_t mspace_max_footprint(mspace msp); | ||
1098 | |||
1099 | |||
1100 | #if !NO_MALLINFO | ||
1101 | /* | ||
1102 | mspace_mallinfo behaves as mallinfo, but reports properties of | ||
1103 | the given space. | ||
1104 | */ | ||
1105 | struct mallinfo mspace_mallinfo(mspace msp); | ||
1106 | #endif /* NO_MALLINFO */ | ||
1107 | |||
1108 | /* | ||
1109 | mspace_malloc_stats behaves as malloc_stats, but reports | ||
1110 | properties of the given space. | ||
1111 | */ | ||
1112 | void mspace_malloc_stats(mspace msp); | ||
1113 | |||
1114 | /* | ||
1115 | mspace_trim behaves as malloc_trim, but | ||
1116 | operates within the given space. | ||
1117 | */ | ||
1118 | int mspace_trim(mspace msp, size_t pad); | ||
1119 | |||
1120 | /* | ||
1121 | An alias for mallopt. | ||
1122 | */ | ||
1123 | int mspace_mallopt(int, int); | ||
1124 | |||
1125 | #endif /* MSPACES */ | ||
1126 | |||
1127 | #ifdef __cplusplus | ||
1128 | }; /* end of extern "C" */ | ||
1129 | #endif /* __cplusplus */ | ||
1130 | |||
1131 | /* | ||
1132 | ======================================================================== | ||
1133 | To make a fully customizable malloc.h header file, cut everything | ||
1134 | above this line, put into file malloc.h, edit to suit, and #include it | ||
1135 | on the next line, as well as in programs that use this malloc. | ||
1136 | ======================================================================== | ||
1137 | */ | ||
1138 | |||
1139 | // #include "malloc.h" | ||
1140 | |||
1141 | /*------------------------------ internal #includes ---------------------- */ | ||
1142 | |||
1143 | #ifdef WIN32 | ||
1144 | #pragma warning( disable : 4146 ) /* no "unsigned" warnings */ | ||
1145 | #endif /* WIN32 */ | ||
1146 | |||
1147 | #include <stdio.h> /* for printing in malloc_stats */ | ||
1148 | |||
1149 | #ifndef LACKS_ERRNO_H | ||
1150 | #include <errno.h> /* for MALLOC_FAILURE_ACTION */ | ||
1151 | #endif /* LACKS_ERRNO_H */ | ||
1152 | #if FOOTERS | ||
1153 | #include <time.h> /* for magic initialization */ | ||
1154 | #endif /* FOOTERS */ | ||
1155 | #ifndef LACKS_STDLIB_H | ||
1156 | #include <stdlib.h> /* for abort() */ | ||
1157 | #endif /* LACKS_STDLIB_H */ | ||
1158 | #ifdef DEBUG | ||
1159 | #if ABORT_ON_ASSERT_FAILURE | ||
1160 | #define assert(x) if(!(x)) ABORT | ||
1161 | #else /* ABORT_ON_ASSERT_FAILURE */ | ||
1162 | #include <assert.h> | ||
1163 | #endif /* ABORT_ON_ASSERT_FAILURE */ | ||
1164 | #else /* DEBUG */ | ||
1165 | #define assert(x) | ||
1166 | #endif /* DEBUG */ | ||
1167 | #ifndef LACKS_STRING_H | ||
1168 | #include <string.h> /* for memset etc */ | ||
1169 | #endif /* LACKS_STRING_H */ | ||
1170 | #if USE_BUILTIN_FFS | ||
1171 | #ifndef LACKS_STRINGS_H | ||
1172 | #include <strings.h> /* for ffs */ | ||
1173 | #endif /* LACKS_STRINGS_H */ | ||
1174 | #endif /* USE_BUILTIN_FFS */ | ||
1175 | #if HAVE_MMAP | ||
1176 | #ifndef LACKS_SYS_MMAN_H | ||
1177 | #include <sys/mman.h> /* for mmap */ | ||
1178 | #endif /* LACKS_SYS_MMAN_H */ | ||
1179 | #ifndef LACKS_FCNTL_H | ||
1180 | #include <fcntl.h> | ||
1181 | #endif /* LACKS_FCNTL_H */ | ||
1182 | #endif /* HAVE_MMAP */ | ||
1183 | #if HAVE_MORECORE | ||
1184 | #ifndef LACKS_UNISTD_H | ||
1185 | #include <unistd.h> /* for sbrk */ | ||
1186 | #else /* LACKS_UNISTD_H */ | ||
1187 | #if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__) | ||
1188 | extern void* sbrk(ptrdiff_t); | ||
1189 | #endif /* FreeBSD etc */ | ||
1190 | #endif /* LACKS_UNISTD_H */ | ||
1191 | #endif /* HAVE_MMAP */ | ||
1192 | |||
1193 | #ifndef WIN32 | ||
1194 | #ifndef malloc_getpagesize | ||
1195 | # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */ | ||
1196 | # ifndef _SC_PAGE_SIZE | ||
1197 | # define _SC_PAGE_SIZE _SC_PAGESIZE | ||
1198 | # endif | ||
1199 | # endif | ||
1200 | # ifdef _SC_PAGE_SIZE | ||
1201 | # define malloc_getpagesize sysconf(_SC_PAGE_SIZE) | ||
1202 | # else | ||
1203 | # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE) | ||
1204 | extern size_t getpagesize(); | ||
1205 | # define malloc_getpagesize getpagesize() | ||
1206 | # else | ||
1207 | # ifdef WIN32 /* use supplied emulation of getpagesize */ | ||
1208 | # define malloc_getpagesize getpagesize() | ||
1209 | # else | ||
1210 | # ifndef LACKS_SYS_PARAM_H | ||
1211 | # include <sys/param.h> | ||
1212 | # endif | ||
1213 | # ifdef EXEC_PAGESIZE | ||
1214 | # define malloc_getpagesize EXEC_PAGESIZE | ||
1215 | # else | ||
1216 | # ifdef NBPG | ||
1217 | # ifndef CLSIZE | ||
1218 | # define malloc_getpagesize NBPG | ||
1219 | # else | ||
1220 | # define malloc_getpagesize (NBPG * CLSIZE) | ||
1221 | # endif | ||
1222 | # else | ||
1223 | # ifdef NBPC | ||
1224 | # define malloc_getpagesize NBPC | ||
1225 | # else | ||
1226 | # ifdef PAGESIZE | ||
1227 | # define malloc_getpagesize PAGESIZE | ||
1228 | # else /* just guess */ | ||
1229 | # define malloc_getpagesize ((size_t)4096U) | ||
1230 | # endif | ||
1231 | # endif | ||
1232 | # endif | ||
1233 | # endif | ||
1234 | # endif | ||
1235 | # endif | ||
1236 | # endif | ||
1237 | #endif | ||
1238 | #endif | ||
1239 | |||
1240 | /* ------------------- size_t and alignment properties -------------------- */ | ||
1241 | |||
1242 | /* The byte and bit size of a size_t */ | ||
1243 | #define SIZE_T_SIZE (sizeof(size_t)) | ||
1244 | #define SIZE_T_BITSIZE (sizeof(size_t) << 3) | ||
1245 | |||
1246 | /* Some constants coerced to size_t */ | ||
1247 | /* Annoying but necessary to avoid errors on some plaftorms */ | ||
1248 | #define SIZE_T_ZERO ((size_t)0) | ||
1249 | #define SIZE_T_ONE ((size_t)1) | ||
1250 | #define SIZE_T_TWO ((size_t)2) | ||
1251 | #define TWO_SIZE_T_SIZES (SIZE_T_SIZE<<1) | ||
1252 | #define FOUR_SIZE_T_SIZES (SIZE_T_SIZE<<2) | ||
1253 | #define SIX_SIZE_T_SIZES (FOUR_SIZE_T_SIZES+TWO_SIZE_T_SIZES) | ||
1254 | #define HALF_MAX_SIZE_T (MAX_SIZE_T / 2U) | ||
1255 | |||
1256 | /* The bit mask value corresponding to MALLOC_ALIGNMENT */ | ||
1257 | #define CHUNK_ALIGN_MASK (MALLOC_ALIGNMENT - SIZE_T_ONE) | ||
1258 | |||
1259 | /* True if address a has acceptable alignment */ | ||
1260 | #define is_aligned(A) (((size_t)((A)) & (CHUNK_ALIGN_MASK)) == 0) | ||
1261 | |||
1262 | /* the number of bytes to offset an address to align it */ | ||
1263 | #define align_offset(A)\ | ||
1264 | ((((size_t)(A) & CHUNK_ALIGN_MASK) == 0)? 0 :\ | ||
1265 | ((MALLOC_ALIGNMENT - ((size_t)(A) & CHUNK_ALIGN_MASK)) & CHUNK_ALIGN_MASK)) | ||
1266 | |||
1267 | /* -------------------------- MMAP preliminaries ------------------------- */ | ||
1268 | |||
1269 | /* | ||
1270 | If HAVE_MORECORE or HAVE_MMAP are false, we just define calls and | ||
1271 | checks to fail so compiler optimizer can delete code rather than | ||
1272 | using so many "#if"s. | ||
1273 | */ | ||
1274 | |||
1275 | |||
1276 | /* MORECORE and MMAP must return MFAIL on failure */ | ||
1277 | #define MFAIL ((void*)(MAX_SIZE_T)) | ||
1278 | #define CMFAIL ((char*)(MFAIL)) /* defined for convenience */ | ||
1279 | |||
1280 | #if !HAVE_MMAP | ||
1281 | #define IS_MMAPPED_BIT (SIZE_T_ZERO) | ||
1282 | #define USE_MMAP_BIT (SIZE_T_ZERO) | ||
1283 | #define CALL_MMAP(s) MFAIL | ||
1284 | #define CALL_MUNMAP(a, s) (-1) | ||
1285 | #define DIRECT_MMAP(s) MFAIL | ||
1286 | |||
1287 | #else /* HAVE_MMAP */ | ||
1288 | #define IS_MMAPPED_BIT (SIZE_T_ONE) | ||
1289 | #define USE_MMAP_BIT (SIZE_T_ONE) | ||
1290 | |||
1291 | #ifndef WIN32 | ||
1292 | #define CALL_MUNMAP(a, s) munmap((a), (s)) | ||
1293 | #define MMAP_PROT (PROT_READ|PROT_WRITE) | ||
1294 | #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON) | ||
1295 | #define MAP_ANONYMOUS MAP_ANON | ||
1296 | #endif /* MAP_ANON */ | ||
1297 | #ifdef MAP_ANONYMOUS | ||
1298 | #define MMAP_FLAGS (MAP_PRIVATE|MAP_ANONYMOUS) | ||
1299 | #define CALL_MMAP(s) mmap(0, (s), MMAP_PROT, MMAP_FLAGS, -1, 0) | ||
1300 | #else /* MAP_ANONYMOUS */ | ||
1301 | /* | ||
1302 | Nearly all versions of mmap support MAP_ANONYMOUS, so the following | ||
1303 | is unlikely to be needed, but is supplied just in case. | ||
1304 | */ | ||
1305 | #define MMAP_FLAGS (MAP_PRIVATE) | ||
1306 | static int dev_zero_fd = -1; /* Cached file descriptor for /dev/zero. */ | ||
1307 | #define CALL_MMAP(s) ((dev_zero_fd < 0) ? \ | ||
1308 | (dev_zero_fd = open("/dev/zero", O_RDWR), \ | ||
1309 | mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) : \ | ||
1310 | mmap(0, (s), MMAP_PROT, MMAP_FLAGS, dev_zero_fd, 0)) | ||
1311 | #endif /* MAP_ANONYMOUS */ | ||
1312 | |||
1313 | #define DIRECT_MMAP(s) CALL_MMAP(s) | ||
1314 | #else /* WIN32 */ | ||
1315 | |||
1316 | /* Win32 MMAP via VirtualAlloc */ | ||
1317 | static void* win32mmap(size_t size) { | ||
1318 | void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT, PAGE_READWRITE); | ||
1319 | return (ptr != 0)? ptr: MFAIL; | ||
1320 | } | ||
1321 | |||
1322 | /* For direct MMAP, use MEM_TOP_DOWN to minimize interference */ | ||
1323 | static void* win32direct_mmap(size_t size) { | ||
1324 | void* ptr = VirtualAlloc(0, size, MEM_RESERVE|MEM_COMMIT|MEM_TOP_DOWN, | ||
1325 | PAGE_READWRITE); | ||
1326 | return (ptr != 0)? ptr: MFAIL; | ||
1327 | } | ||
1328 | |||
1329 | /* This function supports releasing coalesed segments */ | ||
1330 | static int win32munmap(void* ptr, size_t size) { | ||
1331 | MEMORY_BASIC_INFORMATION minfo; | ||
1332 | char* cptr = ptr; | ||
1333 | while (size) { | ||
1334 | if (VirtualQuery(cptr, &minfo, sizeof(minfo)) == 0) | ||
1335 | return -1; | ||
1336 | if (minfo.BaseAddress != cptr || minfo.AllocationBase != cptr || | ||
1337 | minfo.State != MEM_COMMIT || minfo.RegionSize > size) | ||
1338 | return -1; | ||
1339 | if (VirtualFree(cptr, 0, MEM_RELEASE) == 0) | ||
1340 | return -1; | ||
1341 | cptr += minfo.RegionSize; | ||
1342 | size -= minfo.RegionSize; | ||
1343 | } | ||
1344 | return 0; | ||
1345 | } | ||
1346 | |||
1347 | #define CALL_MMAP(s) win32mmap(s) | ||
1348 | #define CALL_MUNMAP(a, s) win32munmap((a), (s)) | ||
1349 | #define DIRECT_MMAP(s) win32direct_mmap(s) | ||
1350 | #endif /* WIN32 */ | ||
1351 | #endif /* HAVE_MMAP */ | ||
1352 | |||
1353 | #if HAVE_MMAP && HAVE_MREMAP | ||
1354 | #define CALL_MREMAP(addr, osz, nsz, mv) mremap((addr), (osz), (nsz), (mv)) | ||
1355 | #else /* HAVE_MMAP && HAVE_MREMAP */ | ||
1356 | #define CALL_MREMAP(addr, osz, nsz, mv) MFAIL | ||
1357 | #endif /* HAVE_MMAP && HAVE_MREMAP */ | ||
1358 | |||
1359 | #if HAVE_MORECORE | ||
1360 | #define CALL_MORECORE(S) MORECORE(S) | ||
1361 | #else /* HAVE_MORECORE */ | ||
1362 | #define CALL_MORECORE(S) MFAIL | ||
1363 | #endif /* HAVE_MORECORE */ | ||
1364 | |||
1365 | /* mstate bit set if continguous morecore disabled or failed */ | ||
1366 | #define USE_NONCONTIGUOUS_BIT (4U) | ||
1367 | |||
1368 | /* segment bit set in create_mspace_with_base */ | ||
1369 | #define EXTERN_BIT (8U) | ||
1370 | |||
1371 | |||
1372 | /* --------------------------- Lock preliminaries ------------------------ */ | ||
1373 | |||
1374 | #if USE_LOCKS | ||
1375 | |||
1376 | /* | ||
1377 | When locks are defined, there are up to two global locks: | ||
1378 | |||
1379 | * If HAVE_MORECORE, morecore_mutex protects sequences of calls to | ||
1380 | MORECORE. In many cases sys_alloc requires two calls, that should | ||
1381 | not be interleaved with calls by other threads. This does not | ||
1382 | protect against direct calls to MORECORE by other threads not | ||
1383 | using this lock, so there is still code to cope the best we can on | ||
1384 | interference. | ||
1385 | |||
1386 | * magic_init_mutex ensures that mparams.magic and other | ||
1387 | unique mparams values are initialized only once. | ||
1388 | */ | ||
1389 | |||
1390 | #ifndef WIN32 | ||
1391 | /* By default use posix locks */ | ||
1392 | #include <pthread.h> | ||
1393 | #define MLOCK_T pthread_mutex_t | ||
1394 | #define INITIAL_LOCK(l) pthread_mutex_init(l, NULL) | ||
1395 | #define ACQUIRE_LOCK(l) pthread_mutex_lock(l) | ||
1396 | #define RELEASE_LOCK(l) pthread_mutex_unlock(l) | ||
1397 | |||
1398 | #if HAVE_MORECORE | ||
1399 | static MLOCK_T morecore_mutex = PTHREAD_MUTEX_INITIALIZER; | ||
1400 | #endif /* HAVE_MORECORE */ | ||
1401 | |||
1402 | static MLOCK_T magic_init_mutex = PTHREAD_MUTEX_INITIALIZER; | ||
1403 | |||
1404 | #else /* WIN32 */ | ||
1405 | /* | ||
1406 | Because lock-protected regions have bounded times, and there | ||
1407 | are no recursive lock calls, we can use simple spinlocks. | ||
1408 | */ | ||
1409 | |||
1410 | #define MLOCK_T long | ||
1411 | static int win32_acquire_lock (MLOCK_T *sl) { | ||
1412 | for (;;) { | ||
1413 | #ifdef InterlockedCompareExchangePointer | ||
1414 | if (!InterlockedCompareExchange(sl, 1, 0)) | ||
1415 | return 0; | ||
1416 | #else /* Use older void* version */ | ||
1417 | if (!InterlockedCompareExchange((void**)sl, (void*)1, (void*)0)) | ||
1418 | return 0; | ||
1419 | #endif /* InterlockedCompareExchangePointer */ | ||
1420 | Sleep (0); | ||
1421 | } | ||
1422 | } | ||
1423 | |||
1424 | static void win32_release_lock (MLOCK_T *sl) { | ||
1425 | InterlockedExchange (sl, 0); | ||
1426 | } | ||
1427 | |||
1428 | #define INITIAL_LOCK(l) *(l)=0 | ||
1429 | #define ACQUIRE_LOCK(l) win32_acquire_lock(l) | ||
1430 | #define RELEASE_LOCK(l) win32_release_lock(l) | ||
1431 | #if HAVE_MORECORE | ||
1432 | static MLOCK_T morecore_mutex; | ||
1433 | #endif /* HAVE_MORECORE */ | ||
1434 | static MLOCK_T magic_init_mutex; | ||
1435 | #endif /* WIN32 */ | ||
1436 | |||
1437 | #define USE_LOCK_BIT (2U) | ||
1438 | #else /* USE_LOCKS */ | ||
1439 | #define USE_LOCK_BIT (0U) | ||
1440 | #define INITIAL_LOCK(l) | ||
1441 | #endif /* USE_LOCKS */ | ||
1442 | |||
1443 | #if USE_LOCKS && HAVE_MORECORE | ||
1444 | #define ACQUIRE_MORECORE_LOCK() ACQUIRE_LOCK(&morecore_mutex); | ||
1445 | #define RELEASE_MORECORE_LOCK() RELEASE_LOCK(&morecore_mutex); | ||
1446 | #else /* USE_LOCKS && HAVE_MORECORE */ | ||
1447 | #define ACQUIRE_MORECORE_LOCK() | ||
1448 | #define RELEASE_MORECORE_LOCK() | ||
1449 | #endif /* USE_LOCKS && HAVE_MORECORE */ | ||
1450 | |||
1451 | #if USE_LOCKS | ||
1452 | #define ACQUIRE_MAGIC_INIT_LOCK() ACQUIRE_LOCK(&magic_init_mutex); | ||
1453 | #define RELEASE_MAGIC_INIT_LOCK() RELEASE_LOCK(&magic_init_mutex); | ||
1454 | #else /* USE_LOCKS */ | ||
1455 | #define ACQUIRE_MAGIC_INIT_LOCK() | ||
1456 | #define RELEASE_MAGIC_INIT_LOCK() | ||
1457 | #endif /* USE_LOCKS */ | ||
1458 | |||
1459 | |||
1460 | /* ----------------------- Chunk representations ------------------------ */ | ||
1461 | |||
1462 | /* | ||
1463 | (The following includes lightly edited explanations by Colin Plumb.) | ||
1464 | |||
1465 | The malloc_chunk declaration below is misleading (but accurate and | ||
1466 | necessary). It declares a "view" into memory allowing access to | ||
1467 | necessary fields at known offsets from a given base. | ||
1468 | |||
1469 | Chunks of memory are maintained using a `boundary tag' method as | ||
1470 | originally described by Knuth. (See the paper by Paul Wilson | ||
1471 | ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a survey of such | ||
1472 | techniques.) Sizes of free chunks are stored both in the front of | ||
1473 | each chunk and at the end. This makes consolidating fragmented | ||
1474 | chunks into bigger chunks fast. The head fields also hold bits | ||
1475 | representing whether chunks are free or in use. | ||
1476 | |||
1477 | Here are some pictures to make it clearer. They are "exploded" to | ||
1478 | show that the state of a chunk can be thought of as extending from | ||
1479 | the high 31 bits of the head field of its header through the | ||
1480 | prev_foot and PINUSE_BIT bit of the following chunk header. | ||
1481 | |||
1482 | A chunk that's in use looks like: | ||
1483 | |||
1484 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1485 | | Size of previous chunk (if P = 1) | | ||
1486 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1487 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P| | ||
1488 | | Size of this chunk 1| +-+ | ||
1489 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1490 | | | | ||
1491 | +- -+ | ||
1492 | | | | ||
1493 | +- -+ | ||
1494 | | : | ||
1495 | +- size - sizeof(size_t) available payload bytes -+ | ||
1496 | : | | ||
1497 | chunk-> +- -+ | ||
1498 | | | | ||
1499 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1500 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |1| | ||
1501 | | Size of next chunk (may or may not be in use) | +-+ | ||
1502 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1503 | |||
1504 | And if it's free, it looks like this: | ||
1505 | |||
1506 | chunk-> +- -+ | ||
1507 | | User payload (must be in use, or we would have merged!) | | ||
1508 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1509 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |P| | ||
1510 | | Size of this chunk 0| +-+ | ||
1511 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1512 | | Next pointer | | ||
1513 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1514 | | Prev pointer | | ||
1515 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1516 | | : | ||
1517 | +- size - sizeof(struct chunk) unused bytes -+ | ||
1518 | : | | ||
1519 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1520 | | Size of this chunk | | ||
1521 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1522 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0| | ||
1523 | | Size of next chunk (must be in use, or we would have merged)| +-+ | ||
1524 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1525 | | : | ||
1526 | +- User payload -+ | ||
1527 | : | | ||
1528 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1529 | |0| | ||
1530 | +-+ | ||
1531 | Note that since we always merge adjacent free chunks, the chunks | ||
1532 | adjacent to a free chunk must be in use. | ||
1533 | |||
1534 | Given a pointer to a chunk (which can be derived trivially from the | ||
1535 | payload pointer) we can, in O(1) time, find out whether the adjacent | ||
1536 | chunks are free, and if so, unlink them from the lists that they | ||
1537 | are on and merge them with the current chunk. | ||
1538 | |||
1539 | Chunks always begin on even word boundaries, so the mem portion | ||
1540 | (which is returned to the user) is also on an even word boundary, and | ||
1541 | thus at least double-word aligned. | ||
1542 | |||
1543 | The P (PINUSE_BIT) bit, stored in the unused low-order bit of the | ||
1544 | chunk size (which is always a multiple of two words), is an in-use | ||
1545 | bit for the *previous* chunk. If that bit is *clear*, then the | ||
1546 | word before the current chunk size contains the previous chunk | ||
1547 | size, and can be used to find the front of the previous chunk. | ||
1548 | The very first chunk allocated always has this bit set, preventing | ||
1549 | access to non-existent (or non-owned) memory. If pinuse is set for | ||
1550 | any given chunk, then you CANNOT determine the size of the | ||
1551 | previous chunk, and might even get a memory addressing fault when | ||
1552 | trying to do so. | ||
1553 | |||
1554 | The C (CINUSE_BIT) bit, stored in the unused second-lowest bit of | ||
1555 | the chunk size redundantly records whether the current chunk is | ||
1556 | inuse. This redundancy enables usage checks within free and realloc, | ||
1557 | and reduces indirection when freeing and consolidating chunks. | ||
1558 | |||
1559 | Each freshly allocated chunk must have both cinuse and pinuse set. | ||
1560 | That is, each allocated chunk borders either a previously allocated | ||
1561 | and still in-use chunk, or the base of its memory arena. This is | ||
1562 | ensured by making all allocations from the the `lowest' part of any | ||
1563 | found chunk. Further, no free chunk physically borders another one, | ||
1564 | so each free chunk is known to be preceded and followed by either | ||
1565 | inuse chunks or the ends of memory. | ||
1566 | |||
1567 | Note that the `foot' of the current chunk is actually represented | ||
1568 | as the prev_foot of the NEXT chunk. This makes it easier to | ||
1569 | deal with alignments etc but can be very confusing when trying | ||
1570 | to extend or adapt this code. | ||
1571 | |||
1572 | The exceptions to all this are | ||
1573 | |||
1574 | 1. The special chunk `top' is the top-most available chunk (i.e., | ||
1575 | the one bordering the end of available memory). It is treated | ||
1576 | specially. Top is never included in any bin, is used only if | ||
1577 | no other chunk is available, and is released back to the | ||
1578 | system if it is very large (see M_TRIM_THRESHOLD). In effect, | ||
1579 | the top chunk is treated as larger (and thus less well | ||
1580 | fitting) than any other available chunk. The top chunk | ||
1581 | doesn't update its trailing size field since there is no next | ||
1582 | contiguous chunk that would have to index off it. However, | ||
1583 | space is still allocated for it (TOP_FOOT_SIZE) to enable | ||
1584 | separation or merging when space is extended. | ||
1585 | |||
1586 | 3. Chunks allocated via mmap, which have the lowest-order bit | ||
1587 | (IS_MMAPPED_BIT) set in their prev_foot fields, and do not set | ||
1588 | PINUSE_BIT in their head fields. Because they are allocated | ||
1589 | one-by-one, each must carry its own prev_foot field, which is | ||
1590 | also used to hold the offset this chunk has within its mmapped | ||
1591 | region, which is needed to preserve alignment. Each mmapped | ||
1592 | chunk is trailed by the first two fields of a fake next-chunk | ||
1593 | for sake of usage checks. | ||
1594 | |||
1595 | */ | ||
1596 | |||
1597 | struct malloc_chunk { | ||
1598 | size_t prev_foot; /* Size of previous chunk (if free). */ | ||
1599 | size_t head; /* Size and inuse bits. */ | ||
1600 | struct malloc_chunk* fd; /* double links -- used only if free. */ | ||
1601 | struct malloc_chunk* bk; | ||
1602 | }; | ||
1603 | |||
1604 | typedef struct malloc_chunk mchunk; | ||
1605 | typedef struct malloc_chunk* mchunkptr; | ||
1606 | typedef struct malloc_chunk* sbinptr; /* The type of bins of chunks */ | ||
1607 | typedef unsigned int bindex_t; /* Described below */ | ||
1608 | typedef unsigned int binmap_t; /* Described below */ | ||
1609 | typedef unsigned int flag_t; /* The type of various bit flag sets */ | ||
1610 | |||
1611 | /* ------------------- Chunks sizes and alignments ----------------------- */ | ||
1612 | |||
1613 | #define MCHUNK_SIZE (sizeof(mchunk)) | ||
1614 | |||
1615 | #if FOOTERS | ||
1616 | #define CHUNK_OVERHEAD (TWO_SIZE_T_SIZES) | ||
1617 | #else /* FOOTERS */ | ||
1618 | #define CHUNK_OVERHEAD (SIZE_T_SIZE) | ||
1619 | #endif /* FOOTERS */ | ||
1620 | |||
1621 | /* MMapped chunks need a second word of overhead ... */ | ||
1622 | #define MMAP_CHUNK_OVERHEAD (TWO_SIZE_T_SIZES) | ||
1623 | /* ... and additional padding for fake next-chunk at foot */ | ||
1624 | #define MMAP_FOOT_PAD (FOUR_SIZE_T_SIZES) | ||
1625 | |||
1626 | /* The smallest size we can malloc is an aligned minimal chunk */ | ||
1627 | #define MIN_CHUNK_SIZE\ | ||
1628 | ((MCHUNK_SIZE + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK) | ||
1629 | |||
1630 | /* conversion from malloc headers to user pointers, and back */ | ||
1631 | #define chunk2mem(p) ((void*)((char*)(p) + TWO_SIZE_T_SIZES)) | ||
1632 | #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - TWO_SIZE_T_SIZES)) | ||
1633 | /* chunk associated with aligned address A */ | ||
1634 | #define align_as_chunk(A) (mchunkptr)((A) + align_offset(chunk2mem(A))) | ||
1635 | |||
1636 | /* Bounds on request (not chunk) sizes. */ | ||
1637 | #define MAX_REQUEST ((-MIN_CHUNK_SIZE) << 2) | ||
1638 | #define MIN_REQUEST (MIN_CHUNK_SIZE - CHUNK_OVERHEAD - SIZE_T_ONE) | ||
1639 | |||
1640 | /* pad request bytes into a usable size */ | ||
1641 | #define pad_request(req) \ | ||
1642 | (((req) + CHUNK_OVERHEAD + CHUNK_ALIGN_MASK) & ~CHUNK_ALIGN_MASK) | ||
1643 | |||
1644 | /* pad request, checking for minimum (but not maximum) */ | ||
1645 | #define request2size(req) \ | ||
1646 | (((req) < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(req)) | ||
1647 | |||
1648 | |||
1649 | /* ------------------ Operations on head and foot fields ----------------- */ | ||
1650 | |||
1651 | /* | ||
1652 | The head field of a chunk is or'ed with PINUSE_BIT when previous | ||
1653 | adjacent chunk in use, and or'ed with CINUSE_BIT if this chunk is in | ||
1654 | use. If the chunk was obtained with mmap, the prev_foot field has | ||
1655 | IS_MMAPPED_BIT set, otherwise holding the offset of the base of the | ||
1656 | mmapped region to the base of the chunk. | ||
1657 | */ | ||
1658 | |||
1659 | #define PINUSE_BIT (SIZE_T_ONE) | ||
1660 | #define CINUSE_BIT (SIZE_T_TWO) | ||
1661 | #define INUSE_BITS (PINUSE_BIT|CINUSE_BIT) | ||
1662 | |||
1663 | /* Head value for fenceposts */ | ||
1664 | #define FENCEPOST_HEAD (INUSE_BITS|SIZE_T_SIZE) | ||
1665 | |||
1666 | /* extraction of fields from head words */ | ||
1667 | #define cinuse(p) ((p)->head & CINUSE_BIT) | ||
1668 | #define pinuse(p) ((p)->head & PINUSE_BIT) | ||
1669 | #define chunksize(p) ((p)->head & ~(INUSE_BITS)) | ||
1670 | |||
1671 | #define clear_pinuse(p) ((p)->head &= ~PINUSE_BIT) | ||
1672 | #define clear_cinuse(p) ((p)->head &= ~CINUSE_BIT) | ||
1673 | |||
1674 | /* Treat space at ptr +/- offset as a chunk */ | ||
1675 | #define chunk_plus_offset(p, s) ((mchunkptr)(((char*)(p)) + (s))) | ||
1676 | #define chunk_minus_offset(p, s) ((mchunkptr)(((char*)(p)) - (s))) | ||
1677 | |||
1678 | /* Ptr to next or previous physical malloc_chunk. */ | ||
1679 | #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->head & ~INUSE_BITS))) | ||
1680 | #define prev_chunk(p) ((mchunkptr)( ((char*)(p)) - ((p)->prev_foot) )) | ||
1681 | |||
1682 | /* extract next chunk's pinuse bit */ | ||
1683 | #define next_pinuse(p) ((next_chunk(p)->head) & PINUSE_BIT) | ||
1684 | |||
1685 | /* Get/set size at footer */ | ||
1686 | #define get_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot) | ||
1687 | #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_foot = (s)) | ||
1688 | |||
1689 | /* Set size, pinuse bit, and foot */ | ||
1690 | #define set_size_and_pinuse_of_free_chunk(p, s)\ | ||
1691 | ((p)->head = (s|PINUSE_BIT), set_foot(p, s)) | ||
1692 | |||
1693 | /* Set size, pinuse bit, foot, and clear next pinuse */ | ||
1694 | #define set_free_with_pinuse(p, s, n)\ | ||
1695 | (clear_pinuse(n), set_size_and_pinuse_of_free_chunk(p, s)) | ||
1696 | |||
1697 | #define is_mmapped(p)\ | ||
1698 | (!((p)->head & PINUSE_BIT) && ((p)->prev_foot & IS_MMAPPED_BIT)) | ||
1699 | |||
1700 | /* Get the internal overhead associated with chunk p */ | ||
1701 | #define overhead_for(p)\ | ||
1702 | (is_mmapped(p)? MMAP_CHUNK_OVERHEAD : CHUNK_OVERHEAD) | ||
1703 | |||
1704 | /* Return true if malloced space is not necessarily cleared */ | ||
1705 | #if MMAP_CLEARS | ||
1706 | #define calloc_must_clear(p) (!is_mmapped(p)) | ||
1707 | #else /* MMAP_CLEARS */ | ||
1708 | #define calloc_must_clear(p) (1) | ||
1709 | #endif /* MMAP_CLEARS */ | ||
1710 | |||
1711 | /* ---------------------- Overlaid data structures ----------------------- */ | ||
1712 | |||
1713 | /* | ||
1714 | When chunks are not in use, they are treated as nodes of either | ||
1715 | lists or trees. | ||
1716 | |||
1717 | "Small" chunks are stored in circular doubly-linked lists, and look | ||
1718 | like this: | ||
1719 | |||
1720 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1721 | | Size of previous chunk | | ||
1722 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1723 | `head:' | Size of chunk, in bytes |P| | ||
1724 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1725 | | Forward pointer to next chunk in list | | ||
1726 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1727 | | Back pointer to previous chunk in list | | ||
1728 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1729 | | Unused space (may be 0 bytes long) . | ||
1730 | . . | ||
1731 | . | | ||
1732 | nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1733 | `foot:' | Size of chunk, in bytes | | ||
1734 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1735 | |||
1736 | Larger chunks are kept in a form of bitwise digital trees (aka | ||
1737 | tries) keyed on chunksizes. Because malloc_tree_chunks are only for | ||
1738 | free chunks greater than 256 bytes, their size doesn't impose any | ||
1739 | constraints on user chunk sizes. Each node looks like: | ||
1740 | |||
1741 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1742 | | Size of previous chunk | | ||
1743 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1744 | `head:' | Size of chunk, in bytes |P| | ||
1745 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1746 | | Forward pointer to next chunk of same size | | ||
1747 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1748 | | Back pointer to previous chunk of same size | | ||
1749 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1750 | | Pointer to left child (child[0]) | | ||
1751 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1752 | | Pointer to right child (child[1]) | | ||
1753 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1754 | | Pointer to parent | | ||
1755 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1756 | | bin index of this chunk | | ||
1757 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1758 | | Unused space . | ||
1759 | . | | ||
1760 | nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1761 | `foot:' | Size of chunk, in bytes | | ||
1762 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | ||
1763 | |||
1764 | Each tree holding treenodes is a tree of unique chunk sizes. Chunks | ||
1765 | of the same size are arranged in a circularly-linked list, with only | ||
1766 | the oldest chunk (the next to be used, in our FIFO ordering) | ||
1767 | actually in the tree. (Tree members are distinguished by a non-null | ||
1768 | parent pointer.) If a chunk with the same size an an existing node | ||
1769 | is inserted, it is linked off the existing node using pointers that | ||
1770 | work in the same way as fd/bk pointers of small chunks. | ||
1771 | |||
1772 | Each tree contains a power of 2 sized range of chunk sizes (the | ||
1773 | smallest is 0x100 <= x < 0x180), which is is divided in half at each | ||
1774 | tree level, with the chunks in the smaller half of the range (0x100 | ||
1775 | <= x < 0x140 for the top nose) in the left subtree and the larger | ||
1776 | half (0x140 <= x < 0x180) in the right subtree. This is, of course, | ||
1777 | done by inspecting individual bits. | ||
1778 | |||
1779 | Using these rules, each node's left subtree contains all smaller | ||
1780 | sizes than its right subtree. However, the node at the root of each | ||
1781 | subtree has no particular ordering relationship to either. (The | ||
1782 | dividing line between the subtree sizes is based on trie relation.) | ||
1783 | If we remove the last chunk of a given size from the interior of the | ||
1784 | tree, we need to replace it with a leaf node. The tree ordering | ||
1785 | rules permit a node to be replaced by any leaf below it. | ||
1786 | |||
1787 | The smallest chunk in a tree (a common operation in a best-fit | ||
1788 | allocator) can be found by walking a path to the leftmost leaf in | ||
1789 | the tree. Unlike a usual binary tree, where we follow left child | ||
1790 | pointers until we reach a null, here we follow the right child | ||
1791 | pointer any time the left one is null, until we reach a leaf with | ||
1792 | both child pointers null. The smallest chunk in the tree will be | ||
1793 | somewhere along that path. | ||
1794 | |||
1795 | The worst case number of steps to add, find, or remove a node is | ||
1796 | bounded by the number of bits differentiating chunks within | ||
1797 | bins. Under current bin calculations, this ranges from 6 up to 21 | ||
1798 | (for 32 bit sizes) or up to 53 (for 64 bit sizes). The typical case | ||
1799 | is of course much better. | ||
1800 | */ | ||
1801 | |||
1802 | struct malloc_tree_chunk { | ||
1803 | /* The first four fields must be compatible with malloc_chunk */ | ||
1804 | size_t prev_foot; | ||
1805 | size_t head; | ||
1806 | struct malloc_tree_chunk* fd; | ||
1807 | struct malloc_tree_chunk* bk; | ||
1808 | |||
1809 | struct malloc_tree_chunk* child[2]; | ||
1810 | struct malloc_tree_chunk* parent; | ||
1811 | bindex_t index; | ||
1812 | }; | ||
1813 | |||
1814 | typedef struct malloc_tree_chunk tchunk; | ||
1815 | typedef struct malloc_tree_chunk* tchunkptr; | ||
1816 | typedef struct malloc_tree_chunk* tbinptr; /* The type of bins of trees */ | ||
1817 | |||
1818 | /* A little helper macro for trees */ | ||
1819 | #define leftmost_child(t) ((t)->child[0] != 0? (t)->child[0] : (t)->child[1]) | ||
1820 | |||
1821 | /* ----------------------------- Segments -------------------------------- */ | ||
1822 | |||
1823 | /* | ||
1824 | Each malloc space may include non-contiguous segments, held in a | ||
1825 | list headed by an embedded malloc_segment record representing the | ||
1826 | top-most space. Segments also include flags holding properties of | ||
1827 | the space. Large chunks that are directly allocated by mmap are not | ||
1828 | included in this list. They are instead independently created and | ||
1829 | destroyed without otherwise keeping track of them. | ||
1830 | |||
1831 | Segment management mainly comes into play for spaces allocated by | ||
1832 | MMAP. Any call to MMAP might or might not return memory that is | ||
1833 | adjacent to an existing segment. MORECORE normally contiguously | ||
1834 | extends the current space, so this space is almost always adjacent, | ||
1835 | which is simpler and faster to deal with. (This is why MORECORE is | ||
1836 | used preferentially to MMAP when both are available -- see | ||
1837 | sys_alloc.) When allocating using MMAP, we don't use any of the | ||
1838 | hinting mechanisms (inconsistently) supported in various | ||
1839 | implementations of unix mmap, or distinguish reserving from | ||
1840 | committing memory. Instead, we just ask for space, and exploit | ||
1841 | contiguity when we get it. It is probably possible to do | ||
1842 | better than this on some systems, but no general scheme seems | ||
1843 | to be significantly better. | ||
1844 | |||
1845 | Management entails a simpler variant of the consolidation scheme | ||
1846 | used for chunks to reduce fragmentation -- new adjacent memory is | ||
1847 | normally prepended or appended to an existing segment. However, | ||
1848 | there are limitations compared to chunk consolidation that mostly | ||
1849 | reflect the fact that segment processing is relatively infrequent | ||
1850 | (occurring only when getting memory from system) and that we | ||
1851 | don't expect to have huge numbers of segments: | ||
1852 | |||
1853 | * Segments are not indexed, so traversal requires linear scans. (It | ||
1854 | would be possible to index these, but is not worth the extra | ||
1855 | overhead and complexity for most programs on most platforms.) | ||
1856 | * New segments are only appended to old ones when holding top-most | ||
1857 | memory; if they cannot be prepended to others, they are held in | ||
1858 | different segments. | ||
1859 | |||
1860 | Except for the top-most segment of an mstate, each segment record | ||
1861 | is kept at the tail of its segment. Segments are added by pushing | ||
1862 | segment records onto the list headed by &mstate.seg for the | ||
1863 | containing mstate. | ||
1864 | |||
1865 | Segment flags control allocation/merge/deallocation policies: | ||
1866 | * If EXTERN_BIT set, then we did not allocate this segment, | ||
1867 | and so should not try to deallocate or merge with others. | ||
1868 | (This currently holds only for the initial segment passed | ||
1869 | into create_mspace_with_base.) | ||
1870 | * If IS_MMAPPED_BIT set, the segment may be merged with | ||
1871 | other surrounding mmapped segments and trimmed/de-allocated | ||
1872 | using munmap. | ||
1873 | * If neither bit is set, then the segment was obtained using | ||
1874 | MORECORE so can be merged with surrounding MORECORE'd segments | ||
1875 | and deallocated/trimmed using MORECORE with negative arguments. | ||
1876 | */ | ||
1877 | |||
1878 | struct malloc_segment { | ||
1879 | char* base; /* base address */ | ||
1880 | size_t size; /* allocated size */ | ||
1881 | struct malloc_segment* next; /* ptr to next segment */ | ||
1882 | flag_t sflags; /* mmap and extern flag */ | ||
1883 | }; | ||
1884 | |||
1885 | #define is_mmapped_segment(S) ((S)->sflags & IS_MMAPPED_BIT) | ||
1886 | #define is_extern_segment(S) ((S)->sflags & EXTERN_BIT) | ||
1887 | |||
1888 | typedef struct malloc_segment msegment; | ||
1889 | typedef struct malloc_segment* msegmentptr; | ||
1890 | |||
1891 | /* ---------------------------- malloc_state ----------------------------- */ | ||
1892 | |||
1893 | /* | ||
1894 | A malloc_state holds all of the bookkeeping for a space. | ||
1895 | The main fields are: | ||
1896 | |||
1897 | Top | ||
1898 | The topmost chunk of the currently active segment. Its size is | ||
1899 | cached in topsize. The actual size of topmost space is | ||
1900 | topsize+TOP_FOOT_SIZE, which includes space reserved for adding | ||
1901 | fenceposts and segment records if necessary when getting more | ||
1902 | space from the system. The size at which to autotrim top is | ||
1903 | cached from mparams in trim_check, except that it is disabled if | ||
1904 | an autotrim fails. | ||
1905 | |||
1906 | Designated victim (dv) | ||
1907 | This is the preferred chunk for servicing small requests that | ||
1908 | don't have exact fits. It is normally the chunk split off most | ||
1909 | recently to service another small request. Its size is cached in | ||
1910 | dvsize. The link fields of this chunk are not maintained since it | ||
1911 | is not kept in a bin. | ||
1912 | |||
1913 | SmallBins | ||
1914 | An array of bin headers for free chunks. These bins hold chunks | ||
1915 | with sizes less than MIN_LARGE_SIZE bytes. Each bin contains | ||
1916 | chunks of all the same size, spaced 8 bytes apart. To simplify | ||
1917 | use in double-linked lists, each bin header acts as a malloc_chunk | ||
1918 | pointing to the real first node, if it exists (else pointing to | ||
1919 | itself). This avoids special-casing for headers. But to avoid | ||
1920 | waste, we allocate only the fd/bk pointers of bins, and then use | ||
1921 | repositioning tricks to treat these as the fields of a chunk. | ||
1922 | |||
1923 | TreeBins | ||
1924 | Treebins are pointers to the roots of trees holding a range of | ||
1925 | sizes. There are 2 equally spaced treebins for each power of two | ||
1926 | from TREE_SHIFT to TREE_SHIFT+16. The last bin holds anything | ||
1927 | larger. | ||
1928 | |||
1929 | Bin maps | ||
1930 | There is one bit map for small bins ("smallmap") and one for | ||
1931 | treebins ("treemap). Each bin sets its bit when non-empty, and | ||
1932 | clears the bit when empty. Bit operations are then used to avoid | ||
1933 | bin-by-bin searching -- nearly all "search" is done without ever | ||
1934 | looking at bins that won't be selected. The bit maps | ||
1935 | conservatively use 32 bits per map word, even if on 64bit system. | ||
1936 | For a good description of some of the bit-based techniques used | ||
1937 | here, see Henry S. Warren Jr's book "Hacker's Delight" (and | ||
1938 | supplement at http://hackersdelight.org/). Many of these are | ||
1939 | intended to reduce the branchiness of paths through malloc etc, as | ||
1940 | well as to reduce the number of memory locations read or written. | ||
1941 | |||
1942 | Segments | ||
1943 | A list of segments headed by an embedded malloc_segment record | ||
1944 | representing the initial space. | ||
1945 | |||
1946 | Address check support | ||
1947 | The least_addr field is the least address ever obtained from | ||
1948 | MORECORE or MMAP. Attempted frees and reallocs of any address less | ||
1949 | than this are trapped (unless INSECURE is defined). | ||
1950 | |||
1951 | Magic tag | ||
1952 | A cross-check field that should always hold same value as mparams.magic. | ||
1953 | |||
1954 | Flags | ||
1955 | Bits recording whether to use MMAP, locks, or contiguous MORECORE | ||
1956 | |||
1957 | Statistics | ||
1958 | Each space keeps track of current and maximum system memory | ||
1959 | obtained via MORECORE or MMAP. | ||
1960 | |||
1961 | Locking | ||
1962 | If USE_LOCKS is defined, the "mutex" lock is acquired and released | ||
1963 | around every public call using this mspace. | ||
1964 | */ | ||
1965 | |||
1966 | /* Bin types, widths and sizes */ | ||
1967 | #define NSMALLBINS (32U) | ||
1968 | #define NTREEBINS (32U) | ||
1969 | #define SMALLBIN_SHIFT (3U) | ||
1970 | #define SMALLBIN_WIDTH (SIZE_T_ONE << SMALLBIN_SHIFT) | ||
1971 | #define TREEBIN_SHIFT (8U) | ||
1972 | #define MIN_LARGE_SIZE (SIZE_T_ONE << TREEBIN_SHIFT) | ||
1973 | #define MAX_SMALL_SIZE (MIN_LARGE_SIZE - SIZE_T_ONE) | ||
1974 | #define MAX_SMALL_REQUEST (MAX_SMALL_SIZE - CHUNK_ALIGN_MASK - CHUNK_OVERHEAD) | ||
1975 | |||
1976 | struct malloc_state { | ||
1977 | binmap_t smallmap; | ||
1978 | binmap_t treemap; | ||
1979 | size_t dvsize; | ||
1980 | size_t topsize; | ||
1981 | char* least_addr; | ||
1982 | mchunkptr dv; | ||
1983 | mchunkptr top; | ||
1984 | size_t trim_check; | ||
1985 | size_t magic; | ||
1986 | mchunkptr smallbins[(NSMALLBINS+1)*2]; | ||
1987 | tbinptr treebins[NTREEBINS]; | ||
1988 | size_t footprint; | ||
1989 | size_t max_footprint; | ||
1990 | flag_t mflags; | ||
1991 | #if USE_LOCKS | ||
1992 | MLOCK_T mutex; /* locate lock among fields that rarely change */ | ||
1993 | #endif /* USE_LOCKS */ | ||
1994 | msegment seg; | ||
1995 | }; | ||
1996 | |||
1997 | typedef struct malloc_state* mstate; | ||
1998 | |||
1999 | /* ------------- Global malloc_state and malloc_params ------------------- */ | ||
2000 | |||
2001 | /* | ||
2002 | malloc_params holds global properties, including those that can be | ||
2003 | dynamically set using mallopt. There is a single instance, mparams, | ||
2004 | initialized in init_mparams. | ||
2005 | */ | ||
2006 | |||
2007 | struct malloc_params { | ||
2008 | size_t magic; | ||
2009 | size_t page_size; | ||
2010 | size_t granularity; | ||
2011 | size_t mmap_threshold; | ||
2012 | size_t trim_threshold; | ||
2013 | flag_t default_mflags; | ||
2014 | }; | ||
2015 | |||
2016 | static struct malloc_params mparams; | ||
2017 | |||
2018 | /* The global malloc_state used for all non-"mspace" calls */ | ||
2019 | static struct malloc_state _gm_; | ||
2020 | #define gm (&_gm_) | ||
2021 | #define is_global(M) ((M) == &_gm_) | ||
2022 | #define is_initialized(M) ((M)->top != 0) | ||
2023 | |||
2024 | /* -------------------------- system alloc setup ------------------------- */ | ||
2025 | |||
2026 | /* Operations on mflags */ | ||
2027 | |||
2028 | #define use_lock(M) ((M)->mflags & USE_LOCK_BIT) | ||
2029 | #define enable_lock(M) ((M)->mflags |= USE_LOCK_BIT) | ||
2030 | #define disable_lock(M) ((M)->mflags &= ~USE_LOCK_BIT) | ||
2031 | |||
2032 | #define use_mmap(M) ((M)->mflags & USE_MMAP_BIT) | ||
2033 | #define enable_mmap(M) ((M)->mflags |= USE_MMAP_BIT) | ||
2034 | #define disable_mmap(M) ((M)->mflags &= ~USE_MMAP_BIT) | ||
2035 | |||
2036 | #define use_noncontiguous(M) ((M)->mflags & USE_NONCONTIGUOUS_BIT) | ||
2037 | #define disable_contiguous(M) ((M)->mflags |= USE_NONCONTIGUOUS_BIT) | ||
2038 | |||
2039 | #define set_lock(M,L)\ | ||
2040 | ((M)->mflags = (L)?\ | ||
2041 | ((M)->mflags | USE_LOCK_BIT) :\ | ||
2042 | ((M)->mflags & ~USE_LOCK_BIT)) | ||
2043 | |||
2044 | /* page-align a size */ | ||
2045 | #define page_align(S)\ | ||
2046 | (((S) + (mparams.page_size)) & ~(mparams.page_size - SIZE_T_ONE)) | ||
2047 | |||
2048 | /* granularity-align a size */ | ||
2049 | #define granularity_align(S)\ | ||
2050 | (((S) + (mparams.granularity)) & ~(mparams.granularity - SIZE_T_ONE)) | ||
2051 | |||
2052 | #define is_page_aligned(S)\ | ||
2053 | (((size_t)(S) & (mparams.page_size - SIZE_T_ONE)) == 0) | ||
2054 | #define is_granularity_aligned(S)\ | ||
2055 | (((size_t)(S) & (mparams.granularity - SIZE_T_ONE)) == 0) | ||
2056 | |||
2057 | /* True if segment S holds address A */ | ||
2058 | #define segment_holds(S, A)\ | ||
2059 | ((char*)(A) >= S->base && (char*)(A) < S->base + S->size) | ||
2060 | |||
2061 | /* Return segment holding given address */ | ||
2062 | static msegmentptr segment_holding(mstate m, char* addr) { | ||
2063 | msegmentptr sp = &m->seg; | ||
2064 | for (;;) { | ||
2065 | if (addr >= sp->base && addr < sp->base + sp->size) | ||
2066 | return sp; | ||
2067 | if ((sp = sp->next) == 0) | ||
2068 | return 0; | ||
2069 | } | ||
2070 | } | ||
2071 | |||
2072 | /* Return true if segment contains a segment link */ | ||
2073 | static int has_segment_link(mstate m, msegmentptr ss) { | ||
2074 | msegmentptr sp = &m->seg; | ||
2075 | for (;;) { | ||
2076 | if ((char*)sp >= ss->base && (char*)sp < ss->base + ss->size) | ||
2077 | return 1; | ||
2078 | if ((sp = sp->next) == 0) | ||
2079 | return 0; | ||
2080 | } | ||
2081 | } | ||
2082 | |||
2083 | #ifndef MORECORE_CANNOT_TRIM | ||
2084 | #define should_trim(M,s) ((s) > (M)->trim_check) | ||
2085 | #else /* MORECORE_CANNOT_TRIM */ | ||
2086 | #define should_trim(M,s) (0) | ||
2087 | #endif /* MORECORE_CANNOT_TRIM */ | ||
2088 | |||
2089 | /* | ||
2090 | TOP_FOOT_SIZE is padding at the end of a segment, including space | ||
2091 | that may be needed to place segment records and fenceposts when new | ||
2092 | noncontiguous segments are added. | ||
2093 | */ | ||
2094 | #define TOP_FOOT_SIZE\ | ||
2095 | (align_offset(chunk2mem(0))+pad_request(sizeof(struct malloc_segment))+MIN_CHUNK_SIZE) | ||
2096 | |||
2097 | |||
2098 | /* ------------------------------- Hooks -------------------------------- */ | ||
2099 | |||
2100 | /* | ||
2101 | PREACTION should be defined to return 0 on success, and nonzero on | ||
2102 | failure. If you are not using locking, you can redefine these to do | ||
2103 | anything you like. | ||
2104 | */ | ||
2105 | |||
2106 | #if USE_LOCKS | ||
2107 | |||
2108 | /* Ensure locks are initialized */ | ||
2109 | #define GLOBALLY_INITIALIZE() (mparams.page_size == 0 && init_mparams()) | ||
2110 | |||
2111 | #define PREACTION(M) ((GLOBALLY_INITIALIZE() || use_lock(M))? ACQUIRE_LOCK(&(M)->mutex) : 0) | ||
2112 | #define POSTACTION(M) { if (use_lock(M)) RELEASE_LOCK(&(M)->mutex); } | ||
2113 | #else /* USE_LOCKS */ | ||
2114 | |||
2115 | #ifndef PREACTION | ||
2116 | #define PREACTION(M) (0) | ||
2117 | #endif /* PREACTION */ | ||
2118 | |||
2119 | #ifndef POSTACTION | ||
2120 | #define POSTACTION(M) | ||
2121 | #endif /* POSTACTION */ | ||
2122 | |||
2123 | #endif /* USE_LOCKS */ | ||
2124 | |||
2125 | /* | ||
2126 | CORRUPTION_ERROR_ACTION is triggered upon detected bad addresses. | ||
2127 | USAGE_ERROR_ACTION is triggered on detected bad frees and | ||
2128 | reallocs. The argument p is an address that might have triggered the | ||
2129 | fault. It is ignored by the two predefined actions, but might be | ||
2130 | useful in custom actions that try to help diagnose errors. | ||
2131 | */ | ||
2132 | |||
2133 | #if PROCEED_ON_ERROR | ||
2134 | |||
2135 | /* A count of the number of corruption errors causing resets */ | ||
2136 | int malloc_corruption_error_count; | ||
2137 | |||
2138 | /* default corruption action */ | ||
2139 | static void reset_on_error(mstate m); | ||
2140 | |||
2141 | #define CORRUPTION_ERROR_ACTION(m) reset_on_error(m) | ||
2142 | #define USAGE_ERROR_ACTION(m, p) | ||
2143 | |||
2144 | #else /* PROCEED_ON_ERROR */ | ||
2145 | |||
2146 | #ifndef CORRUPTION_ERROR_ACTION | ||
2147 | #define CORRUPTION_ERROR_ACTION(m) ABORT | ||
2148 | #endif /* CORRUPTION_ERROR_ACTION */ | ||
2149 | |||
2150 | #ifndef USAGE_ERROR_ACTION | ||
2151 | #define USAGE_ERROR_ACTION(m,p) ABORT | ||
2152 | #endif /* USAGE_ERROR_ACTION */ | ||
2153 | |||
2154 | #endif /* PROCEED_ON_ERROR */ | ||
2155 | |||
2156 | /* -------------------------- Debugging setup ---------------------------- */ | ||
2157 | |||
2158 | #if ! DEBUG | ||
2159 | |||
2160 | #define check_free_chunk(M,P) | ||
2161 | #define check_inuse_chunk(M,P) | ||
2162 | #define check_malloced_chunk(M,P,N) | ||
2163 | #define check_mmapped_chunk(M,P) | ||
2164 | #define check_malloc_state(M) | ||
2165 | #define check_top_chunk(M,P) | ||
2166 | |||
2167 | #else /* DEBUG */ | ||
2168 | #define check_free_chunk(M,P) do_check_free_chunk(M,P) | ||
2169 | #define check_inuse_chunk(M,P) do_check_inuse_chunk(M,P) | ||
2170 | #define check_top_chunk(M,P) do_check_top_chunk(M,P) | ||
2171 | #define check_malloced_chunk(M,P,N) do_check_malloced_chunk(M,P,N) | ||
2172 | #define check_mmapped_chunk(M,P) do_check_mmapped_chunk(M,P) | ||
2173 | #define check_malloc_state(M) do_check_malloc_state(M) | ||
2174 | |||
2175 | static void do_check_any_chunk(mstate m, mchunkptr p); | ||
2176 | static void do_check_top_chunk(mstate m, mchunkptr p); | ||
2177 | static void do_check_mmapped_chunk(mstate m, mchunkptr p); | ||
2178 | static void do_check_inuse_chunk(mstate m, mchunkptr p); | ||
2179 | static void do_check_free_chunk(mstate m, mchunkptr p); | ||
2180 | static void do_check_malloced_chunk(mstate m, void* mem, size_t s); | ||
2181 | static void do_check_tree(mstate m, tchunkptr t); | ||
2182 | static void do_check_treebin(mstate m, bindex_t i); | ||
2183 | static void do_check_smallbin(mstate m, bindex_t i); | ||
2184 | static void do_check_malloc_state(mstate m); | ||
2185 | static int bin_find(mstate m, mchunkptr x); | ||
2186 | static size_t traverse_and_check(mstate m); | ||
2187 | #endif /* DEBUG */ | ||
2188 | |||
2189 | /* ---------------------------- Indexing Bins ---------------------------- */ | ||
2190 | |||
2191 | #define is_small(s) (((s) >> SMALLBIN_SHIFT) < NSMALLBINS) | ||
2192 | #define small_index(s) ((s) >> SMALLBIN_SHIFT) | ||
2193 | #define small_index2size(i) ((i) << SMALLBIN_SHIFT) | ||
2194 | #define MIN_SMALL_INDEX (small_index(MIN_CHUNK_SIZE)) | ||
2195 | |||
2196 | /* addressing by index. See above about smallbin repositioning */ | ||
2197 | #define smallbin_at(M, i) ((sbinptr)((char*)&((M)->smallbins[(i)<<1]))) | ||
2198 | #define treebin_at(M,i) (&((M)->treebins[i])) | ||
2199 | |||
2200 | /* assign tree index for size S to variable I */ | ||
2201 | #if defined(__GNUC__) && defined(i386) | ||
2202 | #define compute_tree_index(S, I)\ | ||
2203 | {\ | ||
2204 | size_t X = S >> TREEBIN_SHIFT;\ | ||
2205 | if (X == 0)\ | ||
2206 | I = 0;\ | ||
2207 | else if (X > 0xFFFF)\ | ||
2208 | I = NTREEBINS-1;\ | ||
2209 | else {\ | ||
2210 | unsigned int K;\ | ||
2211 | __asm__("bsrl %1,%0\n\t" : "=r" (K) : "rm" (X));\ | ||
2212 | I = (bindex_t)((K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1)));\ | ||
2213 | }\ | ||
2214 | } | ||
2215 | #else /* GNUC */ | ||
2216 | #define compute_tree_index(S, I)\ | ||
2217 | {\ | ||
2218 | size_t X = S >> TREEBIN_SHIFT;\ | ||
2219 | if (X == 0)\ | ||
2220 | I = 0;\ | ||
2221 | else if (X > 0xFFFF)\ | ||
2222 | I = NTREEBINS-1;\ | ||
2223 | else {\ | ||
2224 | unsigned int Y = (unsigned int)X;\ | ||
2225 | unsigned int N = ((Y - 0x100) >> 16) & 8;\ | ||
2226 | unsigned int K = (((Y <<= N) - 0x1000) >> 16) & 4;\ | ||
2227 | N += K;\ | ||
2228 | N += K = (((Y <<= K) - 0x4000) >> 16) & 2;\ | ||
2229 | K = 14 - N + ((Y <<= K) >> 15);\ | ||
2230 | I = (K << 1) + ((S >> (K + (TREEBIN_SHIFT-1)) & 1));\ | ||
2231 | }\ | ||
2232 | } | ||
2233 | #endif /* GNUC */ | ||
2234 | |||
2235 | /* Bit representing maximum resolved size in a treebin at i */ | ||
2236 | #define bit_for_tree_index(i) \ | ||
2237 | (i == NTREEBINS-1)? (SIZE_T_BITSIZE-1) : (((i) >> 1) + TREEBIN_SHIFT - 2) | ||
2238 | |||
2239 | /* Shift placing maximum resolved bit in a treebin at i as sign bit */ | ||
2240 | #define leftshift_for_tree_index(i) \ | ||
2241 | ((i == NTREEBINS-1)? 0 : \ | ||
2242 | ((SIZE_T_BITSIZE-SIZE_T_ONE) - (((i) >> 1) + TREEBIN_SHIFT - 2))) | ||
2243 | |||
2244 | /* The size of the smallest chunk held in bin with index i */ | ||
2245 | #define minsize_for_tree_index(i) \ | ||
2246 | ((SIZE_T_ONE << (((i) >> 1) + TREEBIN_SHIFT)) | \ | ||
2247 | (((size_t)((i) & SIZE_T_ONE)) << (((i) >> 1) + TREEBIN_SHIFT - 1))) | ||
2248 | |||
2249 | |||
2250 | /* ------------------------ Operations on bin maps ----------------------- */ | ||
2251 | |||
2252 | /* bit corresponding to given index */ | ||
2253 | #define idx2bit(i) ((binmap_t)(1) << (i)) | ||
2254 | |||
2255 | /* Mark/Clear bits with given index */ | ||
2256 | #define mark_smallmap(M,i) ((M)->smallmap |= idx2bit(i)) | ||
2257 | #define clear_smallmap(M,i) ((M)->smallmap &= ~idx2bit(i)) | ||
2258 | #define smallmap_is_marked(M,i) ((M)->smallmap & idx2bit(i)) | ||
2259 | |||
2260 | #define mark_treemap(M,i) ((M)->treemap |= idx2bit(i)) | ||
2261 | #define clear_treemap(M,i) ((M)->treemap &= ~idx2bit(i)) | ||
2262 | #define treemap_is_marked(M,i) ((M)->treemap & idx2bit(i)) | ||
2263 | |||
2264 | /* index corresponding to given bit */ | ||
2265 | |||
2266 | #if defined(__GNUC__) && defined(i386) | ||
2267 | #define compute_bit2idx(X, I)\ | ||
2268 | {\ | ||
2269 | unsigned int J;\ | ||
2270 | __asm__("bsfl %1,%0\n\t" : "=r" (J) : "rm" (X));\ | ||
2271 | I = (bindex_t)J;\ | ||
2272 | } | ||
2273 | |||
2274 | #else /* GNUC */ | ||
2275 | #if USE_BUILTIN_FFS | ||
2276 | #define compute_bit2idx(X, I) I = ffs(X)-1 | ||
2277 | |||
2278 | #else /* USE_BUILTIN_FFS */ | ||
2279 | #define compute_bit2idx(X, I)\ | ||
2280 | {\ | ||
2281 | unsigned int Y = X - 1;\ | ||
2282 | unsigned int K = Y >> (16-4) & 16;\ | ||
2283 | unsigned int N = K; Y >>= K;\ | ||
2284 | N += K = Y >> (8-3) & 8; Y >>= K;\ | ||
2285 | N += K = Y >> (4-2) & 4; Y >>= K;\ | ||
2286 | N += K = Y >> (2-1) & 2; Y >>= K;\ | ||
2287 | N += K = Y >> (1-0) & 1; Y >>= K;\ | ||
2288 | I = (bindex_t)(N + Y);\ | ||
2289 | } | ||
2290 | #endif /* USE_BUILTIN_FFS */ | ||
2291 | #endif /* GNUC */ | ||
2292 | |||
2293 | /* isolate the least set bit of a bitmap */ | ||
2294 | #define least_bit(x) ((x) & -(x)) | ||
2295 | |||
2296 | /* mask with all bits to left of least bit of x on */ | ||
2297 | #define left_bits(x) ((x<<1) | -(x<<1)) | ||
2298 | |||
2299 | /* mask with all bits to left of or equal to least bit of x on */ | ||
2300 | #define same_or_left_bits(x) ((x) | -(x)) | ||
2301 | |||
2302 | |||
2303 | /* ----------------------- Runtime Check Support ------------------------- */ | ||
2304 | |||
2305 | /* | ||
2306 | For security, the main invariant is that malloc/free/etc never | ||
2307 | writes to a static address other than malloc_state, unless static | ||
2308 | malloc_state itself has been corrupted, which cannot occur via | ||
2309 | malloc (because of these checks). In essence this means that we | ||
2310 | believe all pointers, sizes, maps etc held in malloc_state, but | ||
2311 | check all of those linked or offsetted from other embedded data | ||
2312 | structures. These checks are interspersed with main code in a way | ||
2313 | that tends to minimize their run-time cost. | ||
2314 | |||
2315 | When FOOTERS is defined, in addition to range checking, we also | ||
2316 | verify footer fields of inuse chunks, which can be used guarantee | ||
2317 | that the mstate controlling malloc/free is intact. This is a | ||
2318 | streamlined version of the approach described by William Robertson | ||
2319 | et al in "Run-time Detection of Heap-based Overflows" LISA'03 | ||
2320 | http://www.usenix.org/events/lisa03/tech/robertson.html The footer | ||
2321 | of an inuse chunk holds the xor of its mstate and a random seed, | ||
2322 | that is checked upon calls to free() and realloc(). This is | ||
2323 | (probablistically) unguessable from outside the program, but can be | ||
2324 | computed by any code successfully malloc'ing any chunk, so does not | ||
2325 | itself provide protection against code that has already broken | ||
2326 | security through some other means. Unlike Robertson et al, we | ||
2327 | always dynamically check addresses of all offset chunks (previous, | ||
2328 | next, etc). This turns out to be cheaper than relying on hashes. | ||
2329 | */ | ||
2330 | |||
2331 | #if !INSECURE | ||
2332 | /* Check if address a is at least as high as any from MORECORE or MMAP */ | ||
2333 | #define ok_address(M, a) ((char*)(a) >= (M)->least_addr) | ||
2334 | /* Check if address of next chunk n is higher than base chunk p */ | ||
2335 | #define ok_next(p, n) ((char*)(p) < (char*)(n)) | ||
2336 | /* Check if p has its cinuse bit on */ | ||
2337 | #define ok_cinuse(p) cinuse(p) | ||
2338 | /* Check if p has its pinuse bit on */ | ||
2339 | #define ok_pinuse(p) pinuse(p) | ||
2340 | |||
2341 | #else /* !INSECURE */ | ||
2342 | #define ok_address(M, a) (1) | ||
2343 | #define ok_next(b, n) (1) | ||
2344 | #define ok_cinuse(p) (1) | ||
2345 | #define ok_pinuse(p) (1) | ||
2346 | #endif /* !INSECURE */ | ||
2347 | |||
2348 | #if (FOOTERS && !INSECURE) | ||
2349 | /* Check if (alleged) mstate m has expected magic field */ | ||
2350 | #define ok_magic(M) ((M)->magic == mparams.magic) | ||
2351 | #else /* (FOOTERS && !INSECURE) */ | ||
2352 | #define ok_magic(M) (1) | ||
2353 | #endif /* (FOOTERS && !INSECURE) */ | ||
2354 | |||
2355 | |||
2356 | /* In gcc, use __builtin_expect to minimize impact of checks */ | ||
2357 | #if !INSECURE | ||
2358 | #if defined(__GNUC__) && __GNUC__ >= 3 | ||
2359 | #define RTCHECK(e) __builtin_expect(e, 1) | ||
2360 | #else /* GNUC */ | ||
2361 | #define RTCHECK(e) (e) | ||
2362 | #endif /* GNUC */ | ||
2363 | #else /* !INSECURE */ | ||
2364 | #define RTCHECK(e) (1) | ||
2365 | #endif /* !INSECURE */ | ||
2366 | |||
2367 | /* macros to set up inuse chunks with or without footers */ | ||
2368 | |||
2369 | #if !FOOTERS | ||
2370 | |||
2371 | #define mark_inuse_foot(M,p,s) | ||
2372 | |||
2373 | /* Set cinuse bit and pinuse bit of next chunk */ | ||
2374 | #define set_inuse(M,p,s)\ | ||
2375 | ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\ | ||
2376 | ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT) | ||
2377 | |||
2378 | /* Set cinuse and pinuse of this chunk and pinuse of next chunk */ | ||
2379 | #define set_inuse_and_pinuse(M,p,s)\ | ||
2380 | ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ | ||
2381 | ((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT) | ||
2382 | |||
2383 | /* Set size, cinuse and pinuse bit of this chunk */ | ||
2384 | #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\ | ||
2385 | ((p)->head = (s|PINUSE_BIT|CINUSE_BIT)) | ||
2386 | |||
2387 | #else /* FOOTERS */ | ||
2388 | |||
2389 | /* Set foot of inuse chunk to be xor of mstate and seed */ | ||
2390 | #define mark_inuse_foot(M,p,s)\ | ||
2391 | (((mchunkptr)((char*)(p) + (s)))->prev_foot = ((size_t)(M) ^ mparams.magic)) | ||
2392 | |||
2393 | #define get_mstate_for(p)\ | ||
2394 | ((mstate)(((mchunkptr)((char*)(p) +\ | ||
2395 | (chunksize(p))))->prev_foot ^ mparams.magic)) | ||
2396 | |||
2397 | #define set_inuse(M,p,s)\ | ||
2398 | ((p)->head = (((p)->head & PINUSE_BIT)|s|CINUSE_BIT),\ | ||
2399 | (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT), \ | ||
2400 | mark_inuse_foot(M,p,s)) | ||
2401 | |||
2402 | #define set_inuse_and_pinuse(M,p,s)\ | ||
2403 | ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ | ||
2404 | (((mchunkptr)(((char*)(p)) + (s)))->head |= PINUSE_BIT),\ | ||
2405 | mark_inuse_foot(M,p,s)) | ||
2406 | |||
2407 | #define set_size_and_pinuse_of_inuse_chunk(M, p, s)\ | ||
2408 | ((p)->head = (s|PINUSE_BIT|CINUSE_BIT),\ | ||
2409 | mark_inuse_foot(M, p, s)) | ||
2410 | |||
2411 | #endif /* !FOOTERS */ | ||
2412 | |||
2413 | /* ---------------------------- setting mparams -------------------------- */ | ||
2414 | |||
2415 | /* Initialize mparams */ | ||
2416 | static int init_mparams(void) { | ||
2417 | if (mparams.page_size == 0) { | ||
2418 | size_t s; | ||
2419 | |||
2420 | mparams.mmap_threshold = DEFAULT_MMAP_THRESHOLD; | ||
2421 | mparams.trim_threshold = DEFAULT_TRIM_THRESHOLD; | ||
2422 | #if MORECORE_CONTIGUOUS | ||
2423 | mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT; | ||
2424 | #else /* MORECORE_CONTIGUOUS */ | ||
2425 | mparams.default_mflags = USE_LOCK_BIT|USE_MMAP_BIT|USE_NONCONTIGUOUS_BIT; | ||
2426 | #endif /* MORECORE_CONTIGUOUS */ | ||
2427 | |||
2428 | #if (FOOTERS && !INSECURE) | ||
2429 | { | ||
2430 | #if USE_DEV_RANDOM | ||
2431 | int fd; | ||
2432 | unsigned char buf[sizeof(size_t)]; | ||
2433 | /* Try to use /dev/urandom, else fall back on using time */ | ||
2434 | if ((fd = open("/dev/urandom", O_RDONLY)) >= 0 && | ||
2435 | read(fd, buf, sizeof(buf)) == sizeof(buf)) { | ||
2436 | s = *((size_t *) buf); | ||
2437 | close(fd); | ||
2438 | } | ||
2439 | else | ||
2440 | #endif /* USE_DEV_RANDOM */ | ||
2441 | s = (size_t)(time(0) ^ (size_t)0x55555555U); | ||
2442 | |||
2443 | s |= (size_t)8U; /* ensure nonzero */ | ||
2444 | s &= ~(size_t)7U; /* improve chances of fault for bad values */ | ||
2445 | |||
2446 | } | ||
2447 | #else /* (FOOTERS && !INSECURE) */ | ||
2448 | s = (size_t)0x58585858U; | ||
2449 | #endif /* (FOOTERS && !INSECURE) */ | ||
2450 | ACQUIRE_MAGIC_INIT_LOCK(); | ||
2451 | if (mparams.magic == 0) { | ||
2452 | mparams.magic = s; | ||
2453 | /* Set up lock for main malloc area */ | ||
2454 | INITIAL_LOCK(&gm->mutex); | ||
2455 | gm->mflags = mparams.default_mflags; | ||
2456 | } | ||
2457 | RELEASE_MAGIC_INIT_LOCK(); | ||
2458 | |||
2459 | #ifndef WIN32 | ||
2460 | mparams.page_size = malloc_getpagesize; | ||
2461 | mparams.granularity = ((DEFAULT_GRANULARITY != 0)? | ||
2462 | DEFAULT_GRANULARITY : mparams.page_size); | ||
2463 | #else /* WIN32 */ | ||
2464 | { | ||
2465 | SYSTEM_INFO system_info; | ||
2466 | GetSystemInfo(&system_info); | ||
2467 | mparams.page_size = system_info.dwPageSize; | ||
2468 | mparams.granularity = system_info.dwAllocationGranularity; | ||
2469 | } | ||
2470 | #endif /* WIN32 */ | ||
2471 | |||
2472 | /* Sanity-check configuration: | ||
2473 | size_t must be unsigned and as wide as pointer type. | ||
2474 | ints must be at least 4 bytes. | ||
2475 | alignment must be at least 8. | ||
2476 | Alignment, min chunk size, and page size must all be powers of 2. | ||
2477 | */ | ||
2478 | if ((sizeof(size_t) != sizeof(char*)) || | ||
2479 | (MAX_SIZE_T < MIN_CHUNK_SIZE) || | ||
2480 | (sizeof(int) < 4) || | ||
2481 | (MALLOC_ALIGNMENT < (size_t)8U) || | ||
2482 | ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT-SIZE_T_ONE)) != 0) || | ||
2483 | ((MCHUNK_SIZE & (MCHUNK_SIZE-SIZE_T_ONE)) != 0) || | ||
2484 | ((mparams.granularity & (mparams.granularity-SIZE_T_ONE)) != 0) || | ||
2485 | ((mparams.page_size & (mparams.page_size-SIZE_T_ONE)) != 0)) | ||
2486 | ABORT; | ||
2487 | } | ||
2488 | return 0; | ||
2489 | } | ||
2490 | |||
2491 | /* support for mallopt */ | ||
2492 | static int change_mparam(int param_number, int value) { | ||
2493 | size_t val = (size_t)value; | ||
2494 | init_mparams(); | ||
2495 | switch(param_number) { | ||
2496 | case M_TRIM_THRESHOLD: | ||
2497 | mparams.trim_threshold = val; | ||
2498 | return 1; | ||
2499 | case M_GRANULARITY: | ||
2500 | if (val >= mparams.page_size && ((val & (val-1)) == 0)) { | ||
2501 | mparams.granularity = val; | ||
2502 | return 1; | ||
2503 | } | ||
2504 | else | ||
2505 | return 0; | ||
2506 | case M_MMAP_THRESHOLD: | ||
2507 | mparams.mmap_threshold = val; | ||
2508 | return 1; | ||
2509 | default: | ||
2510 | return 0; | ||
2511 | } | ||
2512 | } | ||
2513 | |||
2514 | #if DEBUG | ||
2515 | /* ------------------------- Debugging Support --------------------------- */ | ||
2516 | |||
2517 | /* Check properties of any chunk, whether free, inuse, mmapped etc */ | ||
2518 | static void do_check_any_chunk(mstate m, mchunkptr p) { | ||
2519 | assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); | ||
2520 | assert(ok_address(m, p)); | ||
2521 | } | ||
2522 | |||
2523 | /* Check properties of top chunk */ | ||
2524 | static void do_check_top_chunk(mstate m, mchunkptr p) { | ||
2525 | msegmentptr sp = segment_holding(m, (char*)p); | ||
2526 | size_t sz = chunksize(p); | ||
2527 | assert(sp != 0); | ||
2528 | assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); | ||
2529 | assert(ok_address(m, p)); | ||
2530 | assert(sz == m->topsize); | ||
2531 | assert(sz > 0); | ||
2532 | assert(sz == ((sp->base + sp->size) - (char*)p) - TOP_FOOT_SIZE); | ||
2533 | assert(pinuse(p)); | ||
2534 | assert(!next_pinuse(p)); | ||
2535 | } | ||
2536 | |||
2537 | /* Check properties of (inuse) mmapped chunks */ | ||
2538 | static void do_check_mmapped_chunk(mstate m, mchunkptr p) { | ||
2539 | size_t sz = chunksize(p); | ||
2540 | size_t len = (sz + (p->prev_foot & ~IS_MMAPPED_BIT) + MMAP_FOOT_PAD); | ||
2541 | assert(is_mmapped(p)); | ||
2542 | assert(use_mmap(m)); | ||
2543 | assert((is_aligned(chunk2mem(p))) || (p->head == FENCEPOST_HEAD)); | ||
2544 | assert(ok_address(m, p)); | ||
2545 | assert(!is_small(sz)); | ||
2546 | assert((len & (mparams.page_size-SIZE_T_ONE)) == 0); | ||
2547 | assert(chunk_plus_offset(p, sz)->head == FENCEPOST_HEAD); | ||
2548 | assert(chunk_plus_offset(p, sz+SIZE_T_SIZE)->head == 0); | ||
2549 | } | ||
2550 | |||
2551 | /* Check properties of inuse chunks */ | ||
2552 | static void do_check_inuse_chunk(mstate m, mchunkptr p) { | ||
2553 | do_check_any_chunk(m, p); | ||
2554 | assert(cinuse(p)); | ||
2555 | assert(next_pinuse(p)); | ||
2556 | /* If not pinuse and not mmapped, previous chunk has OK offset */ | ||
2557 | assert(is_mmapped(p) || pinuse(p) || next_chunk(prev_chunk(p)) == p); | ||
2558 | if (is_mmapped(p)) | ||
2559 | do_check_mmapped_chunk(m, p); | ||
2560 | } | ||
2561 | |||
2562 | /* Check properties of free chunks */ | ||
2563 | static void do_check_free_chunk(mstate m, mchunkptr p) { | ||
2564 | size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT); | ||
2565 | mchunkptr next = chunk_plus_offset(p, sz); | ||
2566 | do_check_any_chunk(m, p); | ||
2567 | assert(!cinuse(p)); | ||
2568 | assert(!next_pinuse(p)); | ||
2569 | assert (!is_mmapped(p)); | ||
2570 | if (p != m->dv && p != m->top) { | ||
2571 | if (sz >= MIN_CHUNK_SIZE) { | ||
2572 | assert((sz & CHUNK_ALIGN_MASK) == 0); | ||
2573 | assert(is_aligned(chunk2mem(p))); | ||
2574 | assert(next->prev_foot == sz); | ||
2575 | assert(pinuse(p)); | ||
2576 | assert (next == m->top || cinuse(next)); | ||
2577 | assert(p->fd->bk == p); | ||
2578 | assert(p->bk->fd == p); | ||
2579 | } | ||
2580 | else /* markers are always of size SIZE_T_SIZE */ | ||
2581 | assert(sz == SIZE_T_SIZE); | ||
2582 | } | ||
2583 | } | ||
2584 | |||
2585 | /* Check properties of malloced chunks at the point they are malloced */ | ||
2586 | static void do_check_malloced_chunk(mstate m, void* mem, size_t s) { | ||
2587 | if (mem != 0) { | ||
2588 | mchunkptr p = mem2chunk(mem); | ||
2589 | size_t sz = p->head & ~(PINUSE_BIT|CINUSE_BIT); | ||
2590 | do_check_inuse_chunk(m, p); | ||
2591 | assert((sz & CHUNK_ALIGN_MASK) == 0); | ||
2592 | assert(sz >= MIN_CHUNK_SIZE); | ||
2593 | assert(sz >= s); | ||
2594 | /* unless mmapped, size is less than MIN_CHUNK_SIZE more than request */ | ||
2595 | assert(is_mmapped(p) || sz < (s + MIN_CHUNK_SIZE)); | ||
2596 | } | ||
2597 | } | ||
2598 | |||
2599 | /* Check a tree and its subtrees. */ | ||
2600 | static void do_check_tree(mstate m, tchunkptr t) { | ||
2601 | tchunkptr head = 0; | ||
2602 | tchunkptr u = t; | ||
2603 | bindex_t tindex = t->index; | ||
2604 | size_t tsize = chunksize(t); | ||
2605 | bindex_t idx; | ||
2606 | compute_tree_index(tsize, idx); | ||
2607 | assert(tindex == idx); | ||
2608 | assert(tsize >= MIN_LARGE_SIZE); | ||
2609 | assert(tsize >= minsize_for_tree_index(idx)); | ||
2610 | assert((idx == NTREEBINS-1) || (tsize < minsize_for_tree_index((idx+1)))); | ||
2611 | |||
2612 | do { /* traverse through chain of same-sized nodes */ | ||
2613 | do_check_any_chunk(m, ((mchunkptr)u)); | ||
2614 | assert(u->index == tindex); | ||
2615 | assert(chunksize(u) == tsize); | ||
2616 | assert(!cinuse(u)); | ||
2617 | assert(!next_pinuse(u)); | ||
2618 | assert(u->fd->bk == u); | ||
2619 | assert(u->bk->fd == u); | ||
2620 | if (u->parent == 0) { | ||
2621 | assert(u->child[0] == 0); | ||
2622 | assert(u->child[1] == 0); | ||
2623 | } | ||
2624 | else { | ||
2625 | assert(head == 0); /* only one node on chain has parent */ | ||
2626 | head = u; | ||
2627 | assert(u->parent != u); | ||
2628 | assert (u->parent->child[0] == u || | ||
2629 | u->parent->child[1] == u || | ||
2630 | *((tbinptr*)(u->parent)) == u); | ||
2631 | if (u->child[0] != 0) { | ||
2632 | assert(u->child[0]->parent == u); | ||
2633 | assert(u->child[0] != u); | ||
2634 | do_check_tree(m, u->child[0]); | ||
2635 | } | ||
2636 | if (u->child[1] != 0) { | ||
2637 | assert(u->child[1]->parent == u); | ||
2638 | assert(u->child[1] != u); | ||
2639 | do_check_tree(m, u->child[1]); | ||
2640 | } | ||
2641 | if (u->child[0] != 0 && u->child[1] != 0) { | ||
2642 | assert(chunksize(u->child[0]) < chunksize(u->child[1])); | ||
2643 | } | ||
2644 | } | ||
2645 | u = u->fd; | ||
2646 | } while (u != t); | ||
2647 | assert(head != 0); | ||
2648 | } | ||
2649 | |||
2650 | /* Check all the chunks in a treebin. */ | ||
2651 | static void do_check_treebin(mstate m, bindex_t i) { | ||
2652 | tbinptr* tb = treebin_at(m, i); | ||
2653 | tchunkptr t = *tb; | ||
2654 | int empty = (m->treemap & (1U << i)) == 0; | ||
2655 | if (t == 0) | ||
2656 | assert(empty); | ||
2657 | if (!empty) | ||
2658 | do_check_tree(m, t); | ||
2659 | } | ||
2660 | |||
2661 | /* Check all the chunks in a smallbin. */ | ||
2662 | static void do_check_smallbin(mstate m, bindex_t i) { | ||
2663 | sbinptr b = smallbin_at(m, i); | ||
2664 | mchunkptr p = b->bk; | ||
2665 | unsigned int empty = (m->smallmap & (1U << i)) == 0; | ||
2666 | if (p == b) | ||
2667 | assert(empty); | ||
2668 | if (!empty) { | ||
2669 | for (; p != b; p = p->bk) { | ||
2670 | size_t size = chunksize(p); | ||
2671 | mchunkptr q; | ||
2672 | /* each chunk claims to be free */ | ||
2673 | do_check_free_chunk(m, p); | ||
2674 | /* chunk belongs in bin */ | ||
2675 | assert(small_index(size) == i); | ||
2676 | assert(p->bk == b || chunksize(p->bk) == chunksize(p)); | ||
2677 | /* chunk is followed by an inuse chunk */ | ||
2678 | q = next_chunk(p); | ||
2679 | if (q->head != FENCEPOST_HEAD) | ||
2680 | do_check_inuse_chunk(m, q); | ||
2681 | } | ||
2682 | } | ||
2683 | } | ||
2684 | |||
2685 | /* Find x in a bin. Used in other check functions. */ | ||
2686 | static int bin_find(mstate m, mchunkptr x) { | ||
2687 | size_t size = chunksize(x); | ||
2688 | if (is_small(size)) { | ||
2689 | bindex_t sidx = small_index(size); | ||
2690 | sbinptr b = smallbin_at(m, sidx); | ||
2691 | if (smallmap_is_marked(m, sidx)) { | ||
2692 | mchunkptr p = b; | ||
2693 | do { | ||
2694 | if (p == x) | ||
2695 | return 1; | ||
2696 | } while ((p = p->fd) != b); | ||
2697 | } | ||
2698 | } | ||
2699 | else { | ||
2700 | bindex_t tidx; | ||
2701 | compute_tree_index(size, tidx); | ||
2702 | if (treemap_is_marked(m, tidx)) { | ||
2703 | tchunkptr t = *treebin_at(m, tidx); | ||
2704 | size_t sizebits = size << leftshift_for_tree_index(tidx); | ||
2705 | while (t != 0 && chunksize(t) != size) { | ||
2706 | t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]; | ||
2707 | sizebits <<= 1; | ||
2708 | } | ||
2709 | if (t != 0) { | ||
2710 | tchunkptr u = t; | ||
2711 | do { | ||
2712 | if (u == (tchunkptr)x) | ||
2713 | return 1; | ||
2714 | } while ((u = u->fd) != t); | ||
2715 | } | ||
2716 | } | ||
2717 | } | ||
2718 | return 0; | ||
2719 | } | ||
2720 | |||
2721 | /* Traverse each chunk and check it; return total */ | ||
2722 | static size_t traverse_and_check(mstate m) { | ||
2723 | size_t sum = 0; | ||
2724 | if (is_initialized(m)) { | ||
2725 | msegmentptr s = &m->seg; | ||
2726 | sum += m->topsize + TOP_FOOT_SIZE; | ||
2727 | while (s != 0) { | ||
2728 | mchunkptr q = align_as_chunk(s->base); | ||
2729 | mchunkptr lastq = 0; | ||
2730 | assert(pinuse(q)); | ||
2731 | while (segment_holds(s, q) && | ||
2732 | q != m->top && q->head != FENCEPOST_HEAD) { | ||
2733 | sum += chunksize(q); | ||
2734 | if (cinuse(q)) { | ||
2735 | assert(!bin_find(m, q)); | ||
2736 | do_check_inuse_chunk(m, q); | ||
2737 | } | ||
2738 | else { | ||
2739 | assert(q == m->dv || bin_find(m, q)); | ||
2740 | assert(lastq == 0 || cinuse(lastq)); /* Not 2 consecutive free */ | ||
2741 | do_check_free_chunk(m, q); | ||
2742 | } | ||
2743 | lastq = q; | ||
2744 | q = next_chunk(q); | ||
2745 | } | ||
2746 | s = s->next; | ||
2747 | } | ||
2748 | } | ||
2749 | return sum; | ||
2750 | } | ||
2751 | |||
2752 | /* Check all properties of malloc_state. */ | ||
2753 | static void do_check_malloc_state(mstate m) { | ||
2754 | bindex_t i; | ||
2755 | size_t total; | ||
2756 | /* check bins */ | ||
2757 | for (i = 0; i < NSMALLBINS; ++i) | ||
2758 | do_check_smallbin(m, i); | ||
2759 | for (i = 0; i < NTREEBINS; ++i) | ||
2760 | do_check_treebin(m, i); | ||
2761 | |||
2762 | if (m->dvsize != 0) { /* check dv chunk */ | ||
2763 | do_check_any_chunk(m, m->dv); | ||
2764 | assert(m->dvsize == chunksize(m->dv)); | ||
2765 | assert(m->dvsize >= MIN_CHUNK_SIZE); | ||
2766 | assert(bin_find(m, m->dv) == 0); | ||
2767 | } | ||
2768 | |||
2769 | if (m->top != 0) { /* check top chunk */ | ||
2770 | do_check_top_chunk(m, m->top); | ||
2771 | assert(m->topsize == chunksize(m->top)); | ||
2772 | assert(m->topsize > 0); | ||
2773 | assert(bin_find(m, m->top) == 0); | ||
2774 | } | ||
2775 | |||
2776 | total = traverse_and_check(m); | ||
2777 | assert(total <= m->footprint); | ||
2778 | assert(m->footprint <= m->max_footprint); | ||
2779 | } | ||
2780 | #endif /* DEBUG */ | ||
2781 | |||
2782 | /* ----------------------------- statistics ------------------------------ */ | ||
2783 | |||
2784 | #if !NO_MALLINFO | ||
2785 | static struct mallinfo internal_mallinfo(mstate m) { | ||
2786 | struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; | ||
2787 | if (!PREACTION(m)) { | ||
2788 | check_malloc_state(m); | ||
2789 | if (is_initialized(m)) { | ||
2790 | size_t nfree = SIZE_T_ONE; /* top always free */ | ||
2791 | size_t mfree = m->topsize + TOP_FOOT_SIZE; | ||
2792 | size_t sum = mfree; | ||
2793 | msegmentptr s = &m->seg; | ||
2794 | while (s != 0) { | ||
2795 | mchunkptr q = align_as_chunk(s->base); | ||
2796 | while (segment_holds(s, q) && | ||
2797 | q != m->top && q->head != FENCEPOST_HEAD) { | ||
2798 | size_t sz = chunksize(q); | ||
2799 | sum += sz; | ||
2800 | if (!cinuse(q)) { | ||
2801 | mfree += sz; | ||
2802 | ++nfree; | ||
2803 | } | ||
2804 | q = next_chunk(q); | ||
2805 | } | ||
2806 | s = s->next; | ||
2807 | } | ||
2808 | |||
2809 | nm.arena = sum; | ||
2810 | nm.ordblks = nfree; | ||
2811 | nm.hblkhd = m->footprint - sum; | ||
2812 | nm.usmblks = m->max_footprint; | ||
2813 | nm.uordblks = m->footprint - mfree; | ||
2814 | nm.fordblks = mfree; | ||
2815 | nm.keepcost = m->topsize; | ||
2816 | } | ||
2817 | |||
2818 | POSTACTION(m); | ||
2819 | } | ||
2820 | return nm; | ||
2821 | } | ||
2822 | #endif /* !NO_MALLINFO */ | ||
2823 | |||
2824 | static void internal_malloc_stats(mstate m) { | ||
2825 | if (!PREACTION(m)) { | ||
2826 | size_t maxfp = 0; | ||
2827 | size_t fp = 0; | ||
2828 | size_t used = 0; | ||
2829 | check_malloc_state(m); | ||
2830 | if (is_initialized(m)) { | ||
2831 | msegmentptr s = &m->seg; | ||
2832 | maxfp = m->max_footprint; | ||
2833 | fp = m->footprint; | ||
2834 | used = fp - (m->topsize + TOP_FOOT_SIZE); | ||
2835 | |||
2836 | while (s != 0) { | ||
2837 | mchunkptr q = align_as_chunk(s->base); | ||
2838 | while (segment_holds(s, q) && | ||
2839 | q != m->top && q->head != FENCEPOST_HEAD) { | ||
2840 | if (!cinuse(q)) | ||
2841 | used -= chunksize(q); | ||
2842 | q = next_chunk(q); | ||
2843 | } | ||
2844 | s = s->next; | ||
2845 | } | ||
2846 | } | ||
2847 | |||
2848 | DEBUGF("max system bytes = %10lu\n", (unsigned long)(maxfp)); | ||
2849 | DEBUGF("system bytes = %10lu\n", (unsigned long)(fp)); | ||
2850 | DEBUGF("in use bytes = %10lu\n", (unsigned long)(used)); | ||
2851 | |||
2852 | POSTACTION(m); | ||
2853 | } | ||
2854 | } | ||
2855 | |||
2856 | /* ----------------------- Operations on smallbins ----------------------- */ | ||
2857 | |||
2858 | /* | ||
2859 | Various forms of linking and unlinking are defined as macros. Even | ||
2860 | the ones for trees, which are very long but have very short typical | ||
2861 | paths. This is ugly but reduces reliance on inlining support of | ||
2862 | compilers. | ||
2863 | */ | ||
2864 | |||
2865 | /* Link a free chunk into a smallbin */ | ||
2866 | #define insert_small_chunk(M, P, S) {\ | ||
2867 | bindex_t I = small_index(S);\ | ||
2868 | mchunkptr B = smallbin_at(M, I);\ | ||
2869 | mchunkptr F = B;\ | ||
2870 | assert(S >= MIN_CHUNK_SIZE);\ | ||
2871 | if (!smallmap_is_marked(M, I))\ | ||
2872 | mark_smallmap(M, I);\ | ||
2873 | else if (RTCHECK(ok_address(M, B->fd)))\ | ||
2874 | F = B->fd;\ | ||
2875 | else {\ | ||
2876 | CORRUPTION_ERROR_ACTION(M);\ | ||
2877 | }\ | ||
2878 | B->fd = P;\ | ||
2879 | F->bk = P;\ | ||
2880 | P->fd = F;\ | ||
2881 | P->bk = B;\ | ||
2882 | } | ||
2883 | |||
2884 | /* Unlink a chunk from a smallbin */ | ||
2885 | #define unlink_small_chunk(M, P, S) {\ | ||
2886 | mchunkptr F = P->fd;\ | ||
2887 | mchunkptr B = P->bk;\ | ||
2888 | bindex_t I = small_index(S);\ | ||
2889 | assert(P != B);\ | ||
2890 | assert(P != F);\ | ||
2891 | assert(chunksize(P) == small_index2size(I));\ | ||
2892 | if (F == B)\ | ||
2893 | clear_smallmap(M, I);\ | ||
2894 | else if (RTCHECK((F == smallbin_at(M,I) || ok_address(M, F)) &&\ | ||
2895 | (B == smallbin_at(M,I) || ok_address(M, B)))) {\ | ||
2896 | F->bk = B;\ | ||
2897 | B->fd = F;\ | ||
2898 | }\ | ||
2899 | else {\ | ||
2900 | CORRUPTION_ERROR_ACTION(M);\ | ||
2901 | }\ | ||
2902 | } | ||
2903 | |||
2904 | /* Unlink the first chunk from a smallbin */ | ||
2905 | #define unlink_first_small_chunk(M, B, P, I) {\ | ||
2906 | mchunkptr F = P->fd;\ | ||
2907 | assert(P != B);\ | ||
2908 | assert(P != F);\ | ||
2909 | assert(chunksize(P) == small_index2size(I));\ | ||
2910 | if (B == F)\ | ||
2911 | clear_smallmap(M, I);\ | ||
2912 | else if (RTCHECK(ok_address(M, F))) {\ | ||
2913 | B->fd = F;\ | ||
2914 | F->bk = B;\ | ||
2915 | }\ | ||
2916 | else {\ | ||
2917 | CORRUPTION_ERROR_ACTION(M);\ | ||
2918 | }\ | ||
2919 | } | ||
2920 | |||
2921 | /* Replace dv node, binning the old one */ | ||
2922 | /* Used only when dvsize known to be small */ | ||
2923 | #define replace_dv(M, P, S) {\ | ||
2924 | size_t DVS = M->dvsize;\ | ||
2925 | if (DVS != 0) {\ | ||
2926 | mchunkptr DV = M->dv;\ | ||
2927 | assert(is_small(DVS));\ | ||
2928 | insert_small_chunk(M, DV, DVS);\ | ||
2929 | }\ | ||
2930 | M->dvsize = S;\ | ||
2931 | M->dv = P;\ | ||
2932 | } | ||
2933 | |||
2934 | /* ------------------------- Operations on trees ------------------------- */ | ||
2935 | |||
2936 | /* Insert chunk into tree */ | ||
2937 | #define insert_large_chunk(M, X, S) {\ | ||
2938 | tbinptr* H;\ | ||
2939 | bindex_t I;\ | ||
2940 | compute_tree_index(S, I);\ | ||
2941 | H = treebin_at(M, I);\ | ||
2942 | X->index = I;\ | ||
2943 | X->child[0] = X->child[1] = 0;\ | ||
2944 | if (!treemap_is_marked(M, I)) {\ | ||
2945 | mark_treemap(M, I);\ | ||
2946 | *H = X;\ | ||
2947 | X->parent = (tchunkptr)H;\ | ||
2948 | X->fd = X->bk = X;\ | ||
2949 | }\ | ||
2950 | else {\ | ||
2951 | tchunkptr T = *H;\ | ||
2952 | size_t K = S << leftshift_for_tree_index(I);\ | ||
2953 | for (;;) {\ | ||
2954 | if (chunksize(T) != S) {\ | ||
2955 | tchunkptr* C = &(T->child[(K >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]);\ | ||
2956 | K <<= 1;\ | ||
2957 | if (*C != 0)\ | ||
2958 | T = *C;\ | ||
2959 | else if (RTCHECK(ok_address(M, C))) {\ | ||
2960 | *C = X;\ | ||
2961 | X->parent = T;\ | ||
2962 | X->fd = X->bk = X;\ | ||
2963 | break;\ | ||
2964 | }\ | ||
2965 | else {\ | ||
2966 | CORRUPTION_ERROR_ACTION(M);\ | ||
2967 | break;\ | ||
2968 | }\ | ||
2969 | }\ | ||
2970 | else {\ | ||
2971 | tchunkptr F = T->fd;\ | ||
2972 | if (RTCHECK(ok_address(M, T) && ok_address(M, F))) {\ | ||
2973 | T->fd = F->bk = X;\ | ||
2974 | X->fd = F;\ | ||
2975 | X->bk = T;\ | ||
2976 | X->parent = 0;\ | ||
2977 | break;\ | ||
2978 | }\ | ||
2979 | else {\ | ||
2980 | CORRUPTION_ERROR_ACTION(M);\ | ||
2981 | break;\ | ||
2982 | }\ | ||
2983 | }\ | ||
2984 | }\ | ||
2985 | }\ | ||
2986 | } | ||
2987 | |||
2988 | /* | ||
2989 | Unlink steps: | ||
2990 | |||
2991 | 1. If x is a chained node, unlink it from its same-sized fd/bk links | ||
2992 | and choose its bk node as its replacement. | ||
2993 | 2. If x was the last node of its size, but not a leaf node, it must | ||
2994 | be replaced with a leaf node (not merely one with an open left or | ||
2995 | right), to make sure that lefts and rights of descendents | ||
2996 | correspond properly to bit masks. We use the rightmost descendent | ||
2997 | of x. We could use any other leaf, but this is easy to locate and | ||
2998 | tends to counteract removal of leftmosts elsewhere, and so keeps | ||
2999 | paths shorter than minimally guaranteed. This doesn't loop much | ||
3000 | because on average a node in a tree is near the bottom. | ||
3001 | 3. If x is the base of a chain (i.e., has parent links) relink | ||
3002 | x's parent and children to x's replacement (or null if none). | ||
3003 | */ | ||
3004 | |||
3005 | #define unlink_large_chunk(M, X) {\ | ||
3006 | tchunkptr XP = X->parent;\ | ||
3007 | tchunkptr R;\ | ||
3008 | if (X->bk != X) {\ | ||
3009 | tchunkptr F = X->fd;\ | ||
3010 | R = X->bk;\ | ||
3011 | if (RTCHECK(ok_address(M, F))) {\ | ||
3012 | F->bk = R;\ | ||
3013 | R->fd = F;\ | ||
3014 | }\ | ||
3015 | else {\ | ||
3016 | CORRUPTION_ERROR_ACTION(M);\ | ||
3017 | }\ | ||
3018 | }\ | ||
3019 | else {\ | ||
3020 | tchunkptr* RP;\ | ||
3021 | if (((R = *(RP = &(X->child[1]))) != 0) ||\ | ||
3022 | ((R = *(RP = &(X->child[0]))) != 0)) {\ | ||
3023 | tchunkptr* CP;\ | ||
3024 | while ((*(CP = &(R->child[1])) != 0) ||\ | ||
3025 | (*(CP = &(R->child[0])) != 0)) {\ | ||
3026 | R = *(RP = CP);\ | ||
3027 | }\ | ||
3028 | if (RTCHECK(ok_address(M, RP)))\ | ||
3029 | *RP = 0;\ | ||
3030 | else {\ | ||
3031 | CORRUPTION_ERROR_ACTION(M);\ | ||
3032 | }\ | ||
3033 | }\ | ||
3034 | }\ | ||
3035 | if (XP != 0) {\ | ||
3036 | tbinptr* H = treebin_at(M, X->index);\ | ||
3037 | if (X == *H) {\ | ||
3038 | if ((*H = R) == 0) \ | ||
3039 | clear_treemap(M, X->index);\ | ||
3040 | }\ | ||
3041 | else if (RTCHECK(ok_address(M, XP))) {\ | ||
3042 | if (XP->child[0] == X) \ | ||
3043 | XP->child[0] = R;\ | ||
3044 | else \ | ||
3045 | XP->child[1] = R;\ | ||
3046 | }\ | ||
3047 | else\ | ||
3048 | CORRUPTION_ERROR_ACTION(M);\ | ||
3049 | if (R != 0) {\ | ||
3050 | if (RTCHECK(ok_address(M, R))) {\ | ||
3051 | tchunkptr C0, C1;\ | ||
3052 | R->parent = XP;\ | ||
3053 | if ((C0 = X->child[0]) != 0) {\ | ||
3054 | if (RTCHECK(ok_address(M, C0))) {\ | ||
3055 | R->child[0] = C0;\ | ||
3056 | C0->parent = R;\ | ||
3057 | }\ | ||
3058 | else\ | ||
3059 | CORRUPTION_ERROR_ACTION(M);\ | ||
3060 | }\ | ||
3061 | if ((C1 = X->child[1]) != 0) {\ | ||
3062 | if (RTCHECK(ok_address(M, C1))) {\ | ||
3063 | R->child[1] = C1;\ | ||
3064 | C1->parent = R;\ | ||
3065 | }\ | ||
3066 | else\ | ||
3067 | CORRUPTION_ERROR_ACTION(M);\ | ||
3068 | }\ | ||
3069 | }\ | ||
3070 | else\ | ||
3071 | CORRUPTION_ERROR_ACTION(M);\ | ||
3072 | }\ | ||
3073 | }\ | ||
3074 | } | ||
3075 | |||
3076 | /* Relays to large vs small bin operations */ | ||
3077 | |||
3078 | #define insert_chunk(M, P, S)\ | ||
3079 | if (is_small(S)) insert_small_chunk(M, P, S)\ | ||
3080 | else { tchunkptr TP = (tchunkptr)(P); insert_large_chunk(M, TP, S); } | ||
3081 | |||
3082 | #define unlink_chunk(M, P, S)\ | ||
3083 | if (is_small(S)) unlink_small_chunk(M, P, S)\ | ||
3084 | else { tchunkptr TP = (tchunkptr)(P); unlink_large_chunk(M, TP); } | ||
3085 | |||
3086 | |||
3087 | /* Relays to internal calls to malloc/free from realloc, memalign etc */ | ||
3088 | |||
3089 | #if ONLY_MSPACES | ||
3090 | #define internal_malloc(m, b) mspace_malloc(m, b) | ||
3091 | #define internal_free(m, mem) mspace_free(m,mem); | ||
3092 | #else /* ONLY_MSPACES */ | ||
3093 | #if MSPACES | ||
3094 | #define internal_malloc(m, b)\ | ||
3095 | (m == gm)? dlmalloc(b) : mspace_malloc(m, b) | ||
3096 | #define internal_free(m, mem)\ | ||
3097 | if (m == gm) dlfree(mem); else mspace_free(m,mem); | ||
3098 | #else /* MSPACES */ | ||
3099 | #define internal_malloc(m, b) dlmalloc(b) | ||
3100 | #define internal_free(m, mem) dlfree(mem) | ||
3101 | #endif /* MSPACES */ | ||
3102 | #endif /* ONLY_MSPACES */ | ||
3103 | |||
3104 | /* ----------------------- Direct-mmapping chunks ----------------------- */ | ||
3105 | |||
3106 | /* | ||
3107 | Directly mmapped chunks are set up with an offset to the start of | ||
3108 | the mmapped region stored in the prev_foot field of the chunk. This | ||
3109 | allows reconstruction of the required argument to MUNMAP when freed, | ||
3110 | and also allows adjustment of the returned chunk to meet alignment | ||
3111 | requirements (especially in memalign). There is also enough space | ||
3112 | allocated to hold a fake next chunk of size SIZE_T_SIZE to maintain | ||
3113 | the PINUSE bit so frees can be checked. | ||
3114 | */ | ||
3115 | |||
3116 | /* Malloc using mmap */ | ||
3117 | static void* mmap_alloc(mstate m, size_t nb) { | ||
3118 | size_t mmsize = granularity_align(nb + SIX_SIZE_T_SIZES + CHUNK_ALIGN_MASK); | ||
3119 | if (mmsize > nb) { /* Check for wrap around 0 */ | ||
3120 | char* mm = (char*)(DIRECT_MMAP(mmsize)); | ||
3121 | if (mm != CMFAIL) { | ||
3122 | size_t offset = align_offset(chunk2mem(mm)); | ||
3123 | size_t psize = mmsize - offset - MMAP_FOOT_PAD; | ||
3124 | mchunkptr p = (mchunkptr)(mm + offset); | ||
3125 | p->prev_foot = offset | IS_MMAPPED_BIT; | ||
3126 | (p)->head = (psize|CINUSE_BIT); | ||
3127 | mark_inuse_foot(m, p, psize); | ||
3128 | chunk_plus_offset(p, psize)->head = FENCEPOST_HEAD; | ||
3129 | chunk_plus_offset(p, psize+SIZE_T_SIZE)->head = 0; | ||
3130 | |||
3131 | if (mm < m->least_addr) | ||
3132 | m->least_addr = mm; | ||
3133 | if ((m->footprint += mmsize) > m->max_footprint) | ||
3134 | m->max_footprint = m->footprint; | ||
3135 | assert(is_aligned(chunk2mem(p))); | ||
3136 | check_mmapped_chunk(m, p); | ||
3137 | return chunk2mem(p); | ||
3138 | } | ||
3139 | } | ||
3140 | return 0; | ||
3141 | } | ||
3142 | |||
3143 | /* Realloc using mmap */ | ||
3144 | static mchunkptr mmap_resize(mstate m, mchunkptr oldp, size_t nb) { | ||
3145 | size_t oldsize = chunksize(oldp); | ||
3146 | if (is_small(nb)) /* Can't shrink mmap regions below small size */ | ||
3147 | return 0; | ||
3148 | /* Keep old chunk if big enough but not too big */ | ||
3149 | if (oldsize >= nb + SIZE_T_SIZE && | ||
3150 | (oldsize - nb) <= (mparams.granularity << 1)) | ||
3151 | return oldp; | ||
3152 | else { | ||
3153 | size_t offset = oldp->prev_foot & ~IS_MMAPPED_BIT; | ||
3154 | size_t oldmmsize = oldsize + offset + MMAP_FOOT_PAD; | ||
3155 | size_t newmmsize = granularity_align(nb + SIX_SIZE_T_SIZES + | ||
3156 | CHUNK_ALIGN_MASK); | ||
3157 | char* cp = (char*)CALL_MREMAP((char*)oldp - offset, | ||
3158 | oldmmsize, newmmsize, 1); | ||
3159 | if (cp != CMFAIL) { | ||
3160 | mchunkptr newp = (mchunkptr)(cp + offset); | ||
3161 | size_t psize = newmmsize - offset - MMAP_FOOT_PAD; | ||
3162 | newp->head = (psize|CINUSE_BIT); | ||
3163 | mark_inuse_foot(m, newp, psize); | ||
3164 | chunk_plus_offset(newp, psize)->head = FENCEPOST_HEAD; | ||
3165 | chunk_plus_offset(newp, psize+SIZE_T_SIZE)->head = 0; | ||
3166 | |||
3167 | if (cp < m->least_addr) | ||
3168 | m->least_addr = cp; | ||
3169 | if ((m->footprint += newmmsize - oldmmsize) > m->max_footprint) | ||
3170 | m->max_footprint = m->footprint; | ||
3171 | check_mmapped_chunk(m, newp); | ||
3172 | return newp; | ||
3173 | } | ||
3174 | } | ||
3175 | return 0; | ||
3176 | } | ||
3177 | |||
3178 | /* -------------------------- mspace management -------------------------- */ | ||
3179 | |||
3180 | /* Initialize top chunk and its size */ | ||
3181 | static void init_top(mstate m, mchunkptr p, size_t psize) { | ||
3182 | /* Ensure alignment */ | ||
3183 | size_t offset = align_offset(chunk2mem(p)); | ||
3184 | p = (mchunkptr)((char*)p + offset); | ||
3185 | psize -= offset; | ||
3186 | |||
3187 | m->top = p; | ||
3188 | m->topsize = psize; | ||
3189 | p->head = psize | PINUSE_BIT; | ||
3190 | /* set size of fake trailing chunk holding overhead space only once */ | ||
3191 | chunk_plus_offset(p, psize)->head = TOP_FOOT_SIZE; | ||
3192 | m->trim_check = mparams.trim_threshold; /* reset on each update */ | ||
3193 | } | ||
3194 | |||
3195 | /* Initialize bins for a new mstate that is otherwise zeroed out */ | ||
3196 | static void init_bins(mstate m) { | ||
3197 | /* Establish circular links for smallbins */ | ||
3198 | bindex_t i; | ||
3199 | for (i = 0; i < NSMALLBINS; ++i) { | ||
3200 | sbinptr bin = smallbin_at(m,i); | ||
3201 | bin->fd = bin->bk = bin; | ||
3202 | } | ||
3203 | } | ||
3204 | |||
3205 | #if PROCEED_ON_ERROR | ||
3206 | |||
3207 | /* default corruption action */ | ||
3208 | static void reset_on_error(mstate m) { | ||
3209 | int i; | ||
3210 | ++malloc_corruption_error_count; | ||
3211 | /* Reinitialize fields to forget about all memory */ | ||
3212 | m->smallbins = m->treebins = 0; | ||
3213 | m->dvsize = m->topsize = 0; | ||
3214 | m->seg.base = 0; | ||
3215 | m->seg.size = 0; | ||
3216 | m->seg.next = 0; | ||
3217 | m->top = m->dv = 0; | ||
3218 | for (i = 0; i < NTREEBINS; ++i) | ||
3219 | *treebin_at(m, i) = 0; | ||
3220 | init_bins(m); | ||
3221 | } | ||
3222 | #endif /* PROCEED_ON_ERROR */ | ||
3223 | |||
3224 | /* Allocate chunk and prepend remainder with chunk in successor base. */ | ||
3225 | static void* prepend_alloc(mstate m, char* newbase, char* oldbase, | ||
3226 | size_t nb) { | ||
3227 | mchunkptr p = align_as_chunk(newbase); | ||
3228 | mchunkptr oldfirst = align_as_chunk(oldbase); | ||
3229 | size_t psize = (char*)oldfirst - (char*)p; | ||
3230 | mchunkptr q = chunk_plus_offset(p, nb); | ||
3231 | size_t qsize = psize - nb; | ||
3232 | set_size_and_pinuse_of_inuse_chunk(m, p, nb); | ||
3233 | |||
3234 | assert((char*)oldfirst > (char*)q); | ||
3235 | assert(pinuse(oldfirst)); | ||
3236 | assert(qsize >= MIN_CHUNK_SIZE); | ||
3237 | |||
3238 | /* consolidate remainder with first chunk of old base */ | ||
3239 | if (oldfirst == m->top) { | ||
3240 | size_t tsize = m->topsize += qsize; | ||
3241 | m->top = q; | ||
3242 | q->head = tsize | PINUSE_BIT; | ||
3243 | check_top_chunk(m, q); | ||
3244 | } | ||
3245 | else if (oldfirst == m->dv) { | ||
3246 | size_t dsize = m->dvsize += qsize; | ||
3247 | m->dv = q; | ||
3248 | set_size_and_pinuse_of_free_chunk(q, dsize); | ||
3249 | } | ||
3250 | else { | ||
3251 | if (!cinuse(oldfirst)) { | ||
3252 | size_t nsize = chunksize(oldfirst); | ||
3253 | unlink_chunk(m, oldfirst, nsize); | ||
3254 | oldfirst = chunk_plus_offset(oldfirst, nsize); | ||
3255 | qsize += nsize; | ||
3256 | } | ||
3257 | set_free_with_pinuse(q, qsize, oldfirst); | ||
3258 | insert_chunk(m, q, qsize); | ||
3259 | check_free_chunk(m, q); | ||
3260 | } | ||
3261 | |||
3262 | check_malloced_chunk(m, chunk2mem(p), nb); | ||
3263 | return chunk2mem(p); | ||
3264 | } | ||
3265 | |||
3266 | |||
3267 | /* Add a segment to hold a new noncontiguous region */ | ||
3268 | static void add_segment(mstate m, char* tbase, size_t tsize, flag_t mmapped) { | ||
3269 | /* Determine locations and sizes of segment, fenceposts, old top */ | ||
3270 | char* old_top = (char*)m->top; | ||
3271 | msegmentptr oldsp = segment_holding(m, old_top); | ||
3272 | char* old_end = oldsp->base + oldsp->size; | ||
3273 | size_t ssize = pad_request(sizeof(struct malloc_segment)); | ||
3274 | char* rawsp = old_end - (ssize + FOUR_SIZE_T_SIZES + CHUNK_ALIGN_MASK); | ||
3275 | size_t offset = align_offset(chunk2mem(rawsp)); | ||
3276 | char* asp = rawsp + offset; | ||
3277 | char* csp = (asp < (old_top + MIN_CHUNK_SIZE))? old_top : asp; | ||
3278 | mchunkptr sp = (mchunkptr)csp; | ||
3279 | msegmentptr ss = (msegmentptr)(chunk2mem(sp)); | ||
3280 | mchunkptr tnext = chunk_plus_offset(sp, ssize); | ||
3281 | mchunkptr p = tnext; | ||
3282 | int nfences = 0; | ||
3283 | |||
3284 | /* reset top to new space */ | ||
3285 | init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE); | ||
3286 | |||
3287 | /* Set up segment record */ | ||
3288 | assert(is_aligned(ss)); | ||
3289 | set_size_and_pinuse_of_inuse_chunk(m, sp, ssize); | ||
3290 | *ss = m->seg; /* Push current record */ | ||
3291 | m->seg.base = tbase; | ||
3292 | m->seg.size = tsize; | ||
3293 | m->seg.sflags = mmapped; | ||
3294 | m->seg.next = ss; | ||
3295 | |||
3296 | /* Insert trailing fenceposts */ | ||
3297 | for (;;) { | ||
3298 | mchunkptr nextp = chunk_plus_offset(p, SIZE_T_SIZE); | ||
3299 | p->head = FENCEPOST_HEAD; | ||
3300 | ++nfences; | ||
3301 | if ((char*)(&(nextp->head)) < old_end) | ||
3302 | p = nextp; | ||
3303 | else | ||
3304 | break; | ||
3305 | } | ||
3306 | assert(nfences >= 2); | ||
3307 | |||
3308 | /* Insert the rest of old top into a bin as an ordinary free chunk */ | ||
3309 | if (csp != old_top) { | ||
3310 | mchunkptr q = (mchunkptr)old_top; | ||
3311 | size_t psize = csp - old_top; | ||
3312 | mchunkptr tn = chunk_plus_offset(q, psize); | ||
3313 | set_free_with_pinuse(q, psize, tn); | ||
3314 | insert_chunk(m, q, psize); | ||
3315 | } | ||
3316 | |||
3317 | check_top_chunk(m, m->top); | ||
3318 | } | ||
3319 | |||
3320 | /* -------------------------- System allocation -------------------------- */ | ||
3321 | |||
3322 | /* Get memory from system using MORECORE or MMAP */ | ||
3323 | static void* sys_alloc(mstate m, size_t nb) { | ||
3324 | char* tbase = CMFAIL; | ||
3325 | size_t tsize = 0; | ||
3326 | flag_t mmap_flag = 0; | ||
3327 | |||
3328 | init_mparams(); | ||
3329 | |||
3330 | /* Directly map large chunks */ | ||
3331 | if (use_mmap(m) && nb >= mparams.mmap_threshold) { | ||
3332 | void* mem = mmap_alloc(m, nb); | ||
3333 | if (mem != 0) | ||
3334 | return mem; | ||
3335 | } | ||
3336 | |||
3337 | /* | ||
3338 | Try getting memory in any of three ways (in most-preferred to | ||
3339 | least-preferred order): | ||
3340 | 1. A call to MORECORE that can normally contiguously extend memory. | ||
3341 | (disabled if not MORECORE_CONTIGUOUS or not HAVE_MORECORE or | ||
3342 | or main space is mmapped or a previous contiguous call failed) | ||
3343 | 2. A call to MMAP new space (disabled if not HAVE_MMAP). | ||
3344 | Note that under the default settings, if MORECORE is unable to | ||
3345 | fulfill a request, and HAVE_MMAP is true, then mmap is | ||
3346 | used as a noncontiguous system allocator. This is a useful backup | ||
3347 | strategy for systems with holes in address spaces -- in this case | ||
3348 | sbrk cannot contiguously expand the heap, but mmap may be able to | ||
3349 | find space. | ||
3350 | 3. A call to MORECORE that cannot usually contiguously extend memory. | ||
3351 | (disabled if not HAVE_MORECORE) | ||
3352 | */ | ||
3353 | |||
3354 | if (MORECORE_CONTIGUOUS && !use_noncontiguous(m)) { | ||
3355 | char* br = CMFAIL; | ||
3356 | msegmentptr ss = (m->top == 0)? 0 : segment_holding(m, (char*)m->top); | ||
3357 | size_t asize = 0; | ||
3358 | ACQUIRE_MORECORE_LOCK(); | ||
3359 | |||
3360 | if (ss == 0) { /* First time through or recovery */ | ||
3361 | char* base = (char*)CALL_MORECORE(0); | ||
3362 | if (base != CMFAIL) { | ||
3363 | asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE); | ||
3364 | /* Adjust to end on a page boundary */ | ||
3365 | if (!is_page_aligned(base)) | ||
3366 | asize += (page_align((size_t)base) - (size_t)base); | ||
3367 | /* Can't call MORECORE if size is negative when treated as signed */ | ||
3368 | if (asize < HALF_MAX_SIZE_T && | ||
3369 | (br = (char*)(CALL_MORECORE(asize))) == base) { | ||
3370 | tbase = base; | ||
3371 | tsize = asize; | ||
3372 | } | ||
3373 | } | ||
3374 | } | ||
3375 | else { | ||
3376 | /* Subtract out existing available top space from MORECORE request. */ | ||
3377 | asize = granularity_align(nb - m->topsize + TOP_FOOT_SIZE + SIZE_T_ONE); | ||
3378 | /* Use mem here only if it did continuously extend old space */ | ||
3379 | if (asize < HALF_MAX_SIZE_T && | ||
3380 | (br = (char*)(CALL_MORECORE(asize))) == ss->base+ss->size) { | ||
3381 | tbase = br; | ||
3382 | tsize = asize; | ||
3383 | } | ||
3384 | } | ||
3385 | |||
3386 | if (tbase == CMFAIL) { /* Cope with partial failure */ | ||
3387 | if (br != CMFAIL) { /* Try to use/extend the space we did get */ | ||
3388 | if (asize < HALF_MAX_SIZE_T && | ||
3389 | asize < nb + TOP_FOOT_SIZE + SIZE_T_ONE) { | ||
3390 | size_t esize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE - asize); | ||
3391 | if (esize < HALF_MAX_SIZE_T) { | ||
3392 | char* end = (char*)CALL_MORECORE(esize); | ||
3393 | if (end != CMFAIL) | ||
3394 | asize += esize; | ||
3395 | else { /* Can't use; try to release */ | ||
3396 | CALL_MORECORE(-asize); | ||
3397 | br = CMFAIL; | ||
3398 | } | ||
3399 | } | ||
3400 | } | ||
3401 | } | ||
3402 | if (br != CMFAIL) { /* Use the space we did get */ | ||
3403 | tbase = br; | ||
3404 | tsize = asize; | ||
3405 | } | ||
3406 | else | ||
3407 | disable_contiguous(m); /* Don't try contiguous path in the future */ | ||
3408 | } | ||
3409 | |||
3410 | RELEASE_MORECORE_LOCK(); | ||
3411 | } | ||
3412 | |||
3413 | if (HAVE_MMAP && tbase == CMFAIL) { /* Try MMAP */ | ||
3414 | size_t req = nb + TOP_FOOT_SIZE + SIZE_T_ONE; | ||
3415 | size_t rsize = granularity_align(req); | ||
3416 | if (rsize > nb) { /* Fail if wraps around zero */ | ||
3417 | char* mp = (char*)(CALL_MMAP(rsize)); | ||
3418 | if (mp != CMFAIL) { | ||
3419 | tbase = mp; | ||
3420 | tsize = rsize; | ||
3421 | mmap_flag = IS_MMAPPED_BIT; | ||
3422 | } | ||
3423 | } | ||
3424 | } | ||
3425 | |||
3426 | if (HAVE_MORECORE && tbase == CMFAIL) { /* Try noncontiguous MORECORE */ | ||
3427 | size_t asize = granularity_align(nb + TOP_FOOT_SIZE + SIZE_T_ONE); | ||
3428 | if (asize < HALF_MAX_SIZE_T) { | ||
3429 | char* br = CMFAIL; | ||
3430 | char* end = CMFAIL; | ||
3431 | ACQUIRE_MORECORE_LOCK(); | ||
3432 | br = (char*)(CALL_MORECORE(asize)); | ||
3433 | end = (char*)(CALL_MORECORE(0)); | ||
3434 | RELEASE_MORECORE_LOCK(); | ||
3435 | if (br != CMFAIL && end != CMFAIL && br < end) { | ||
3436 | size_t ssize = end - br; | ||
3437 | if (ssize > nb + TOP_FOOT_SIZE) { | ||
3438 | tbase = br; | ||
3439 | tsize = ssize; | ||
3440 | } | ||
3441 | } | ||
3442 | } | ||
3443 | } | ||
3444 | |||
3445 | if (tbase != CMFAIL) { | ||
3446 | |||
3447 | if ((m->footprint += tsize) > m->max_footprint) | ||
3448 | m->max_footprint = m->footprint; | ||
3449 | |||
3450 | if (!is_initialized(m)) { /* first-time initialization */ | ||
3451 | m->seg.base = m->least_addr = tbase; | ||
3452 | m->seg.size = tsize; | ||
3453 | m->seg.sflags = mmap_flag; | ||
3454 | m->magic = mparams.magic; | ||
3455 | init_bins(m); | ||
3456 | if (is_global(m)) | ||
3457 | init_top(m, (mchunkptr)tbase, tsize - TOP_FOOT_SIZE); | ||
3458 | else { | ||
3459 | /* Offset top by embedded malloc_state */ | ||
3460 | mchunkptr mn = next_chunk(mem2chunk(m)); | ||
3461 | init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) -TOP_FOOT_SIZE); | ||
3462 | } | ||
3463 | } | ||
3464 | |||
3465 | else { | ||
3466 | /* Try to merge with an existing segment */ | ||
3467 | msegmentptr sp = &m->seg; | ||
3468 | while (sp != 0 && tbase != sp->base + sp->size) | ||
3469 | sp = sp->next; | ||
3470 | if (sp != 0 && | ||
3471 | !is_extern_segment(sp) && | ||
3472 | (sp->sflags & IS_MMAPPED_BIT) == mmap_flag && | ||
3473 | segment_holds(sp, m->top)) { /* append */ | ||
3474 | sp->size += tsize; | ||
3475 | init_top(m, m->top, m->topsize + tsize); | ||
3476 | } | ||
3477 | else { | ||
3478 | if (tbase < m->least_addr) | ||
3479 | m->least_addr = tbase; | ||
3480 | sp = &m->seg; | ||
3481 | while (sp != 0 && sp->base != tbase + tsize) | ||
3482 | sp = sp->next; | ||
3483 | if (sp != 0 && | ||
3484 | !is_extern_segment(sp) && | ||
3485 | (sp->sflags & IS_MMAPPED_BIT) == mmap_flag) { | ||
3486 | char* oldbase = sp->base; | ||
3487 | sp->base = tbase; | ||
3488 | sp->size += tsize; | ||
3489 | return prepend_alloc(m, tbase, oldbase, nb); | ||
3490 | } | ||
3491 | else | ||
3492 | add_segment(m, tbase, tsize, mmap_flag); | ||
3493 | } | ||
3494 | } | ||
3495 | |||
3496 | if (nb < m->topsize) { /* Allocate from new or extended top space */ | ||
3497 | size_t rsize = m->topsize -= nb; | ||
3498 | mchunkptr p = m->top; | ||
3499 | mchunkptr r = m->top = chunk_plus_offset(p, nb); | ||
3500 | r->head = rsize | PINUSE_BIT; | ||
3501 | set_size_and_pinuse_of_inuse_chunk(m, p, nb); | ||
3502 | check_top_chunk(m, m->top); | ||
3503 | check_malloced_chunk(m, chunk2mem(p), nb); | ||
3504 | return chunk2mem(p); | ||
3505 | } | ||
3506 | } | ||
3507 | |||
3508 | MALLOC_FAILURE_ACTION; | ||
3509 | return 0; | ||
3510 | } | ||
3511 | |||
3512 | /* ----------------------- system deallocation -------------------------- */ | ||
3513 | |||
3514 | /* Unmap and unlink any mmapped segments that don't contain used chunks */ | ||
3515 | static size_t release_unused_segments(mstate m) { | ||
3516 | size_t released = 0; | ||
3517 | msegmentptr pred = &m->seg; | ||
3518 | msegmentptr sp = pred->next; | ||
3519 | while (sp != 0) { | ||
3520 | char* base = sp->base; | ||
3521 | size_t size = sp->size; | ||
3522 | msegmentptr next = sp->next; | ||
3523 | if (is_mmapped_segment(sp) && !is_extern_segment(sp)) { | ||
3524 | mchunkptr p = align_as_chunk(base); | ||
3525 | size_t psize = chunksize(p); | ||
3526 | /* Can unmap if first chunk holds entire segment and not pinned */ | ||
3527 | if (!cinuse(p) && (char*)p + psize >= base + size - TOP_FOOT_SIZE) { | ||
3528 | tchunkptr tp = (tchunkptr)p; | ||
3529 | assert(segment_holds(sp, (char*)sp)); | ||
3530 | if (p == m->dv) { | ||
3531 | m->dv = 0; | ||
3532 | m->dvsize = 0; | ||
3533 | } | ||
3534 | else { | ||
3535 | unlink_large_chunk(m, tp); | ||
3536 | } | ||
3537 | if (CALL_MUNMAP(base, size) == 0) { | ||
3538 | released += size; | ||
3539 | m->footprint -= size; | ||
3540 | /* unlink obsoleted record */ | ||
3541 | sp = pred; | ||
3542 | sp->next = next; | ||
3543 | } | ||
3544 | else { /* back out if cannot unmap */ | ||
3545 | insert_large_chunk(m, tp, psize); | ||
3546 | } | ||
3547 | } | ||
3548 | } | ||
3549 | pred = sp; | ||
3550 | sp = next; | ||
3551 | } | ||
3552 | return released; | ||
3553 | } | ||
3554 | |||
3555 | static int sys_trim(mstate m, size_t pad) { | ||
3556 | size_t released = 0; | ||
3557 | if (pad < MAX_REQUEST && is_initialized(m)) { | ||
3558 | pad += TOP_FOOT_SIZE; /* ensure enough room for segment overhead */ | ||
3559 | |||
3560 | if (m->topsize > pad) { | ||
3561 | /* Shrink top space in granularity-size units, keeping at least one */ | ||
3562 | size_t unit = mparams.granularity; | ||
3563 | size_t extra = ((m->topsize - pad + (unit - SIZE_T_ONE)) / unit - | ||
3564 | SIZE_T_ONE) * unit; | ||
3565 | msegmentptr sp = segment_holding(m, (char*)m->top); | ||
3566 | |||
3567 | if (!is_extern_segment(sp)) { | ||
3568 | if (is_mmapped_segment(sp)) { | ||
3569 | if (HAVE_MMAP && | ||
3570 | sp->size >= extra && | ||
3571 | !has_segment_link(m, sp)) { /* can't shrink if pinned */ | ||
3572 | // size_t newsize = sp->size - extra; | ||
3573 | /* Prefer mremap, fall back to munmap */ | ||
3574 | if ((CALL_MREMAP(sp->base, sp->size, newsize, 0) != MFAIL) || | ||
3575 | (CALL_MUNMAP(sp->base + newsize, extra) == 0)) { | ||
3576 | released = extra; | ||
3577 | } | ||
3578 | } | ||
3579 | } | ||
3580 | else if (HAVE_MORECORE) { | ||
3581 | if (extra >= HALF_MAX_SIZE_T) /* Avoid wrapping negative */ | ||
3582 | extra = (HALF_MAX_SIZE_T) + SIZE_T_ONE - unit; | ||
3583 | ACQUIRE_MORECORE_LOCK(); | ||
3584 | { | ||
3585 | /* Make sure end of memory is where we last set it. */ | ||
3586 | char* old_br = (char*)(CALL_MORECORE(0)); | ||
3587 | if (old_br == sp->base + sp->size) { | ||
3588 | char* rel_br = (char*)(CALL_MORECORE(-extra)); | ||
3589 | char* new_br = (char*)(CALL_MORECORE(0)); | ||
3590 | if (rel_br != CMFAIL && new_br < old_br) | ||
3591 | released = old_br - new_br; | ||
3592 | } | ||
3593 | } | ||
3594 | RELEASE_MORECORE_LOCK(); | ||
3595 | } | ||
3596 | } | ||
3597 | |||
3598 | if (released != 0) { | ||
3599 | sp->size -= released; | ||
3600 | m->footprint -= released; | ||
3601 | init_top(m, m->top, m->topsize - released); | ||
3602 | check_top_chunk(m, m->top); | ||
3603 | } | ||
3604 | } | ||
3605 | |||
3606 | /* Unmap any unused mmapped segments */ | ||
3607 | if (HAVE_MMAP) | ||
3608 | released += release_unused_segments(m); | ||
3609 | |||
3610 | /* On failure, disable autotrim to avoid repeated failed future calls */ | ||
3611 | if (released == 0) | ||
3612 | m->trim_check = MAX_SIZE_T; | ||
3613 | } | ||
3614 | |||
3615 | return (released != 0)? 1 : 0; | ||
3616 | } | ||
3617 | |||
3618 | /* ---------------------------- malloc support --------------------------- */ | ||
3619 | |||
3620 | /* allocate a large request from the best fitting chunk in a treebin */ | ||
3621 | static void* tmalloc_large(mstate m, size_t nb) { | ||
3622 | tchunkptr v = 0; | ||
3623 | size_t rsize = -nb; /* Unsigned negation */ | ||
3624 | tchunkptr t; | ||
3625 | bindex_t idx; | ||
3626 | compute_tree_index(nb, idx); | ||
3627 | |||
3628 | if ((t = *treebin_at(m, idx)) != 0) { | ||
3629 | /* Traverse tree for this bin looking for node with size == nb */ | ||
3630 | size_t sizebits = nb << leftshift_for_tree_index(idx); | ||
3631 | tchunkptr rst = 0; /* The deepest untaken right subtree */ | ||
3632 | for (;;) { | ||
3633 | tchunkptr rt; | ||
3634 | size_t trem = chunksize(t) - nb; | ||
3635 | if (trem < rsize) { | ||
3636 | v = t; | ||
3637 | if ((rsize = trem) == 0) | ||
3638 | break; | ||
3639 | } | ||
3640 | rt = t->child[1]; | ||
3641 | t = t->child[(sizebits >> (SIZE_T_BITSIZE-SIZE_T_ONE)) & 1]; | ||
3642 | if (rt != 0 && rt != t) | ||
3643 | rst = rt; | ||
3644 | if (t == 0) { | ||
3645 | t = rst; /* set t to least subtree holding sizes > nb */ | ||
3646 | break; | ||
3647 | } | ||
3648 | sizebits <<= 1; | ||
3649 | } | ||
3650 | } | ||
3651 | |||
3652 | if (t == 0 && v == 0) { /* set t to root of next non-empty treebin */ | ||
3653 | binmap_t leftbits = left_bits(idx2bit(idx)) & m->treemap; | ||
3654 | if (leftbits != 0) { | ||
3655 | bindex_t i; | ||
3656 | binmap_t leastbit = least_bit(leftbits); | ||
3657 | compute_bit2idx(leastbit, i); | ||
3658 | t = *treebin_at(m, i); | ||
3659 | } | ||
3660 | } | ||
3661 | |||
3662 | while (t != 0) { /* find smallest of tree or subtree */ | ||
3663 | size_t trem = chunksize(t) - nb; | ||
3664 | if (trem < rsize) { | ||
3665 | rsize = trem; | ||
3666 | v = t; | ||
3667 | } | ||
3668 | t = leftmost_child(t); | ||
3669 | } | ||
3670 | |||
3671 | /* If dv is a better fit, return 0 so malloc will use it */ | ||
3672 | if (v != 0 && rsize < (size_t)(m->dvsize - nb)) { | ||
3673 | if (RTCHECK(ok_address(m, v))) { /* split */ | ||
3674 | mchunkptr r = chunk_plus_offset(v, nb); | ||
3675 | assert(chunksize(v) == rsize + nb); | ||
3676 | if (RTCHECK(ok_next(v, r))) { | ||
3677 | unlink_large_chunk(m, v); | ||
3678 | if (rsize < MIN_CHUNK_SIZE) | ||
3679 | set_inuse_and_pinuse(m, v, (rsize + nb)); | ||
3680 | else { | ||
3681 | set_size_and_pinuse_of_inuse_chunk(m, v, nb); | ||
3682 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
3683 | insert_chunk(m, r, rsize); | ||
3684 | } | ||
3685 | return chunk2mem(v); | ||
3686 | } | ||
3687 | } | ||
3688 | CORRUPTION_ERROR_ACTION(m); | ||
3689 | } | ||
3690 | return 0; | ||
3691 | } | ||
3692 | |||
3693 | /* allocate a small request from the best fitting chunk in a treebin */ | ||
3694 | static void* tmalloc_small(mstate m, size_t nb) { | ||
3695 | tchunkptr t, v; | ||
3696 | size_t rsize; | ||
3697 | bindex_t i; | ||
3698 | binmap_t leastbit = least_bit(m->treemap); | ||
3699 | compute_bit2idx(leastbit, i); | ||
3700 | |||
3701 | v = t = *treebin_at(m, i); | ||
3702 | rsize = chunksize(t) - nb; | ||
3703 | |||
3704 | while ((t = leftmost_child(t)) != 0) { | ||
3705 | size_t trem = chunksize(t) - nb; | ||
3706 | if (trem < rsize) { | ||
3707 | rsize = trem; | ||
3708 | v = t; | ||
3709 | } | ||
3710 | } | ||
3711 | |||
3712 | if (RTCHECK(ok_address(m, v))) { | ||
3713 | mchunkptr r = chunk_plus_offset(v, nb); | ||
3714 | assert(chunksize(v) == rsize + nb); | ||
3715 | if (RTCHECK(ok_next(v, r))) { | ||
3716 | unlink_large_chunk(m, v); | ||
3717 | if (rsize < MIN_CHUNK_SIZE) | ||
3718 | set_inuse_and_pinuse(m, v, (rsize + nb)); | ||
3719 | else { | ||
3720 | set_size_and_pinuse_of_inuse_chunk(m, v, nb); | ||
3721 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
3722 | replace_dv(m, r, rsize); | ||
3723 | } | ||
3724 | return chunk2mem(v); | ||
3725 | } | ||
3726 | } | ||
3727 | |||
3728 | CORRUPTION_ERROR_ACTION(m); | ||
3729 | return 0; | ||
3730 | } | ||
3731 | |||
3732 | /* --------------------------- realloc support --------------------------- */ | ||
3733 | |||
3734 | static void* internal_realloc(mstate m, void* oldmem, size_t bytes) { | ||
3735 | if (bytes >= MAX_REQUEST) { | ||
3736 | MALLOC_FAILURE_ACTION; | ||
3737 | return 0; | ||
3738 | } | ||
3739 | if (!PREACTION(m)) { | ||
3740 | mchunkptr oldp = mem2chunk(oldmem); | ||
3741 | size_t oldsize = chunksize(oldp); | ||
3742 | mchunkptr next = chunk_plus_offset(oldp, oldsize); | ||
3743 | mchunkptr newp = 0; | ||
3744 | void* extra = 0; | ||
3745 | |||
3746 | /* Try to either shrink or extend into top. Else malloc-copy-free */ | ||
3747 | |||
3748 | if (RTCHECK(ok_address(m, oldp) && ok_cinuse(oldp) && | ||
3749 | ok_next(oldp, next) && ok_pinuse(next))) { | ||
3750 | size_t nb = request2size(bytes); | ||
3751 | if (is_mmapped(oldp)) | ||
3752 | newp = mmap_resize(m, oldp, nb); | ||
3753 | else if (oldsize >= nb) { /* already big enough */ | ||
3754 | size_t rsize = oldsize - nb; | ||
3755 | newp = oldp; | ||
3756 | if (rsize >= MIN_CHUNK_SIZE) { | ||
3757 | mchunkptr remainder = chunk_plus_offset(newp, nb); | ||
3758 | set_inuse(m, newp, nb); | ||
3759 | set_inuse(m, remainder, rsize); | ||
3760 | extra = chunk2mem(remainder); | ||
3761 | } | ||
3762 | } | ||
3763 | else if (next == m->top && oldsize + m->topsize > nb) { | ||
3764 | /* Expand into top */ | ||
3765 | size_t newsize = oldsize + m->topsize; | ||
3766 | size_t newtopsize = newsize - nb; | ||
3767 | mchunkptr newtop = chunk_plus_offset(oldp, nb); | ||
3768 | set_inuse(m, oldp, nb); | ||
3769 | newtop->head = newtopsize |PINUSE_BIT; | ||
3770 | m->top = newtop; | ||
3771 | m->topsize = newtopsize; | ||
3772 | newp = oldp; | ||
3773 | } | ||
3774 | } | ||
3775 | else { | ||
3776 | USAGE_ERROR_ACTION(m, oldmem); | ||
3777 | POSTACTION(m); | ||
3778 | return 0; | ||
3779 | } | ||
3780 | |||
3781 | POSTACTION(m); | ||
3782 | |||
3783 | if (newp != 0) { | ||
3784 | if (extra != 0) { | ||
3785 | internal_free(m, extra); | ||
3786 | } | ||
3787 | check_inuse_chunk(m, newp); | ||
3788 | return chunk2mem(newp); | ||
3789 | } | ||
3790 | else { | ||
3791 | void* newmem = internal_malloc(m, bytes); | ||
3792 | if (newmem != 0) { | ||
3793 | size_t oc = oldsize - overhead_for(oldp); | ||
3794 | memcpy(newmem, oldmem, (oc < bytes)? oc : bytes); | ||
3795 | internal_free(m, oldmem); | ||
3796 | } | ||
3797 | return newmem; | ||
3798 | } | ||
3799 | } | ||
3800 | return 0; | ||
3801 | } | ||
3802 | |||
3803 | /* --------------------------- memalign support -------------------------- */ | ||
3804 | |||
3805 | static void* internal_memalign(mstate m, size_t alignment, size_t bytes) { | ||
3806 | if (alignment <= MALLOC_ALIGNMENT) /* Can just use malloc */ | ||
3807 | return internal_malloc(m, bytes); | ||
3808 | if (alignment < MIN_CHUNK_SIZE) /* must be at least a minimum chunk size */ | ||
3809 | alignment = MIN_CHUNK_SIZE; | ||
3810 | if ((alignment & (alignment-SIZE_T_ONE)) != 0) {/* Ensure a power of 2 */ | ||
3811 | size_t a = MALLOC_ALIGNMENT << 1; | ||
3812 | while (a < alignment) a <<= 1; | ||
3813 | alignment = a; | ||
3814 | } | ||
3815 | |||
3816 | if (bytes >= MAX_REQUEST - alignment) { | ||
3817 | if (m != 0) { /* Test isn't needed but avoids compiler warning */ | ||
3818 | MALLOC_FAILURE_ACTION; | ||
3819 | } | ||
3820 | } | ||
3821 | else { | ||
3822 | size_t nb = request2size(bytes); | ||
3823 | size_t req = nb + alignment + MIN_CHUNK_SIZE - CHUNK_OVERHEAD; | ||
3824 | char* mem = (char*)internal_malloc(m, req); | ||
3825 | if (mem != 0) { | ||
3826 | void* leader = 0; | ||
3827 | void* trailer = 0; | ||
3828 | mchunkptr p = mem2chunk(mem); | ||
3829 | |||
3830 | if (PREACTION(m)) return 0; | ||
3831 | if ((((size_t)(mem)) % alignment) != 0) { /* misaligned */ | ||
3832 | /* | ||
3833 | Find an aligned spot inside chunk. Since we need to give | ||
3834 | back leading space in a chunk of at least MIN_CHUNK_SIZE, if | ||
3835 | the first calculation places us at a spot with less than | ||
3836 | MIN_CHUNK_SIZE leader, we can move to the next aligned spot. | ||
3837 | We've allocated enough total room so that this is always | ||
3838 | possible. | ||
3839 | */ | ||
3840 | char* br = (char*)mem2chunk((size_t)(((size_t)(mem + | ||
3841 | alignment - | ||
3842 | SIZE_T_ONE)) & | ||
3843 | -alignment)); | ||
3844 | char* pos = ((size_t)(br - (char*)(p)) >= MIN_CHUNK_SIZE)? | ||
3845 | br : br+alignment; | ||
3846 | mchunkptr newp = (mchunkptr)pos; | ||
3847 | size_t leadsize = pos - (char*)(p); | ||
3848 | size_t newsize = chunksize(p) - leadsize; | ||
3849 | |||
3850 | if (is_mmapped(p)) { /* For mmapped chunks, just adjust offset */ | ||
3851 | newp->prev_foot = p->prev_foot + leadsize; | ||
3852 | newp->head = (newsize|CINUSE_BIT); | ||
3853 | } | ||
3854 | else { /* Otherwise, give back leader, use the rest */ | ||
3855 | set_inuse(m, newp, newsize); | ||
3856 | set_inuse(m, p, leadsize); | ||
3857 | leader = chunk2mem(p); | ||
3858 | } | ||
3859 | p = newp; | ||
3860 | } | ||
3861 | |||
3862 | /* Give back spare room at the end */ | ||
3863 | if (!is_mmapped(p)) { | ||
3864 | size_t size = chunksize(p); | ||
3865 | if (size > nb + MIN_CHUNK_SIZE) { | ||
3866 | size_t remainder_size = size - nb; | ||
3867 | mchunkptr remainder = chunk_plus_offset(p, nb); | ||
3868 | set_inuse(m, p, nb); | ||
3869 | set_inuse(m, remainder, remainder_size); | ||
3870 | trailer = chunk2mem(remainder); | ||
3871 | } | ||
3872 | } | ||
3873 | |||
3874 | assert (chunksize(p) >= nb); | ||
3875 | assert((((size_t)(chunk2mem(p))) % alignment) == 0); | ||
3876 | check_inuse_chunk(m, p); | ||
3877 | POSTACTION(m); | ||
3878 | if (leader != 0) { | ||
3879 | internal_free(m, leader); | ||
3880 | } | ||
3881 | if (trailer != 0) { | ||
3882 | internal_free(m, trailer); | ||
3883 | } | ||
3884 | return chunk2mem(p); | ||
3885 | } | ||
3886 | } | ||
3887 | return 0; | ||
3888 | } | ||
3889 | |||
3890 | /* ------------------------ comalloc/coalloc support --------------------- */ | ||
3891 | |||
3892 | static void** ialloc(mstate m, | ||
3893 | size_t n_elements, | ||
3894 | size_t* sizes, | ||
3895 | int opts, | ||
3896 | void* chunks[]) { | ||
3897 | /* | ||
3898 | This provides common support for independent_X routines, handling | ||
3899 | all of the combinations that can result. | ||
3900 | |||
3901 | The opts arg has: | ||
3902 | bit 0 set if all elements are same size (using sizes[0]) | ||
3903 | bit 1 set if elements should be zeroed | ||
3904 | */ | ||
3905 | |||
3906 | size_t element_size; /* chunksize of each element, if all same */ | ||
3907 | size_t contents_size; /* total size of elements */ | ||
3908 | size_t array_size; /* request size of pointer array */ | ||
3909 | void* mem; /* malloced aggregate space */ | ||
3910 | mchunkptr p; /* corresponding chunk */ | ||
3911 | size_t remainder_size; /* remaining bytes while splitting */ | ||
3912 | void** marray; /* either "chunks" or malloced ptr array */ | ||
3913 | mchunkptr array_chunk; /* chunk for malloced ptr array */ | ||
3914 | flag_t was_enabled; /* to disable mmap */ | ||
3915 | size_t size; | ||
3916 | size_t i; | ||
3917 | |||
3918 | /* compute array length, if needed */ | ||
3919 | if (chunks != 0) { | ||
3920 | if (n_elements == 0) | ||
3921 | return chunks; /* nothing to do */ | ||
3922 | marray = chunks; | ||
3923 | array_size = 0; | ||
3924 | } | ||
3925 | else { | ||
3926 | /* if empty req, must still return chunk representing empty array */ | ||
3927 | if (n_elements == 0) | ||
3928 | return (void**)internal_malloc(m, 0); | ||
3929 | marray = 0; | ||
3930 | array_size = request2size(n_elements * (sizeof(void*))); | ||
3931 | } | ||
3932 | |||
3933 | /* compute total element size */ | ||
3934 | if (opts & 0x1) { /* all-same-size */ | ||
3935 | element_size = request2size(*sizes); | ||
3936 | contents_size = n_elements * element_size; | ||
3937 | } | ||
3938 | else { /* add up all the sizes */ | ||
3939 | element_size = 0; | ||
3940 | contents_size = 0; | ||
3941 | for (i = 0; i != n_elements; ++i) | ||
3942 | contents_size += request2size(sizes[i]); | ||
3943 | } | ||
3944 | |||
3945 | size = contents_size + array_size; | ||
3946 | |||
3947 | /* | ||
3948 | Allocate the aggregate chunk. First disable direct-mmapping so | ||
3949 | malloc won't use it, since we would not be able to later | ||
3950 | free/realloc space internal to a segregated mmap region. | ||
3951 | */ | ||
3952 | was_enabled = use_mmap(m); | ||
3953 | disable_mmap(m); | ||
3954 | mem = internal_malloc(m, size - CHUNK_OVERHEAD); | ||
3955 | if (was_enabled) | ||
3956 | enable_mmap(m); | ||
3957 | if (mem == 0) | ||
3958 | return 0; | ||
3959 | |||
3960 | if (PREACTION(m)) return 0; | ||
3961 | p = mem2chunk(mem); | ||
3962 | remainder_size = chunksize(p); | ||
3963 | |||
3964 | assert(!is_mmapped(p)); | ||
3965 | |||
3966 | if (opts & 0x2) { /* optionally clear the elements */ | ||
3967 | memset((size_t*)mem, 0, remainder_size - SIZE_T_SIZE - array_size); | ||
3968 | } | ||
3969 | |||
3970 | /* If not provided, allocate the pointer array as final part of chunk */ | ||
3971 | if (marray == 0) { | ||
3972 | size_t array_chunk_size; | ||
3973 | array_chunk = chunk_plus_offset(p, contents_size); | ||
3974 | array_chunk_size = remainder_size - contents_size; | ||
3975 | marray = (void**) (chunk2mem(array_chunk)); | ||
3976 | set_size_and_pinuse_of_inuse_chunk(m, array_chunk, array_chunk_size); | ||
3977 | remainder_size = contents_size; | ||
3978 | } | ||
3979 | |||
3980 | /* split out elements */ | ||
3981 | for (i = 0; ; ++i) { | ||
3982 | marray[i] = chunk2mem(p); | ||
3983 | if (i != n_elements-1) { | ||
3984 | if (element_size != 0) | ||
3985 | size = element_size; | ||
3986 | else | ||
3987 | size = request2size(sizes[i]); | ||
3988 | remainder_size -= size; | ||
3989 | set_size_and_pinuse_of_inuse_chunk(m, p, size); | ||
3990 | p = chunk_plus_offset(p, size); | ||
3991 | } | ||
3992 | else { /* the final element absorbs any overallocation slop */ | ||
3993 | set_size_and_pinuse_of_inuse_chunk(m, p, remainder_size); | ||
3994 | break; | ||
3995 | } | ||
3996 | } | ||
3997 | |||
3998 | #if DEBUG | ||
3999 | if (marray != chunks) { | ||
4000 | /* final element must have exactly exhausted chunk */ | ||
4001 | if (element_size != 0) { | ||
4002 | assert(remainder_size == element_size); | ||
4003 | } | ||
4004 | else { | ||
4005 | assert(remainder_size == request2size(sizes[i])); | ||
4006 | } | ||
4007 | check_inuse_chunk(m, mem2chunk(marray)); | ||
4008 | } | ||
4009 | for (i = 0; i != n_elements; ++i) | ||
4010 | check_inuse_chunk(m, mem2chunk(marray[i])); | ||
4011 | |||
4012 | #endif /* DEBUG */ | ||
4013 | |||
4014 | POSTACTION(m); | ||
4015 | return marray; | ||
4016 | } | ||
4017 | |||
4018 | |||
4019 | /* -------------------------- public routines ---------------------------- */ | ||
4020 | |||
4021 | #if !ONLY_MSPACES | ||
4022 | |||
4023 | void* dlmalloc(size_t bytes) { | ||
4024 | /* | ||
4025 | Basic algorithm: | ||
4026 | If a small request (< 256 bytes minus per-chunk overhead): | ||
4027 | 1. If one exists, use a remainderless chunk in associated smallbin. | ||
4028 | (Remainderless means that there are too few excess bytes to | ||
4029 | represent as a chunk.) | ||
4030 | 2. If it is big enough, use the dv chunk, which is normally the | ||
4031 | chunk adjacent to the one used for the most recent small request. | ||
4032 | 3. If one exists, split the smallest available chunk in a bin, | ||
4033 | saving remainder in dv. | ||
4034 | 4. If it is big enough, use the top chunk. | ||
4035 | 5. If available, get memory from system and use it | ||
4036 | Otherwise, for a large request: | ||
4037 | 1. Find the smallest available binned chunk that fits, and use it | ||
4038 | if it is better fitting than dv chunk, splitting if necessary. | ||
4039 | 2. If better fitting than any binned chunk, use the dv chunk. | ||
4040 | 3. If it is big enough, use the top chunk. | ||
4041 | 4. If request size >= mmap threshold, try to directly mmap this chunk. | ||
4042 | 5. If available, get memory from system and use it | ||
4043 | |||
4044 | The ugly goto's here ensure that postaction occurs along all paths. | ||
4045 | */ | ||
4046 | |||
4047 | if (!PREACTION(gm)) { | ||
4048 | void* mem; | ||
4049 | size_t nb; | ||
4050 | if (bytes <= MAX_SMALL_REQUEST) { | ||
4051 | bindex_t idx; | ||
4052 | binmap_t smallbits; | ||
4053 | nb = (bytes < MIN_REQUEST)? MIN_CHUNK_SIZE : pad_request(bytes); | ||
4054 | idx = small_index(nb); | ||
4055 | smallbits = gm->smallmap >> idx; | ||
4056 | |||
4057 | if ((smallbits & 0x3U) != 0) { /* Remainderless fit to a smallbin. */ | ||
4058 | mchunkptr b, p; | ||
4059 | idx += ~smallbits & 1; /* Uses next bin if idx empty */ | ||
4060 | b = smallbin_at(gm, idx); | ||
4061 | p = b->fd; | ||
4062 | assert(chunksize(p) == small_index2size(idx)); | ||
4063 | unlink_first_small_chunk(gm, b, p, idx); | ||
4064 | set_inuse_and_pinuse(gm, p, small_index2size(idx)); | ||
4065 | mem = chunk2mem(p); | ||
4066 | check_malloced_chunk(gm, mem, nb); | ||
4067 | goto postaction; | ||
4068 | } | ||
4069 | |||
4070 | else if (nb > gm->dvsize) { | ||
4071 | if (smallbits != 0) { /* Use chunk in next nonempty smallbin */ | ||
4072 | mchunkptr b, p, r; | ||
4073 | size_t rsize; | ||
4074 | bindex_t i; | ||
4075 | binmap_t leftbits = (smallbits << idx) & left_bits(idx2bit(idx)); | ||
4076 | binmap_t leastbit = least_bit(leftbits); | ||
4077 | compute_bit2idx(leastbit, i); | ||
4078 | b = smallbin_at(gm, i); | ||
4079 | p = b->fd; | ||
4080 | assert(chunksize(p) == small_index2size(i)); | ||
4081 | unlink_first_small_chunk(gm, b, p, i); | ||
4082 | rsize = small_index2size(i) - nb; | ||
4083 | /* Fit here cannot be remainderless if 4byte sizes */ | ||
4084 | if (SIZE_T_SIZE != 4 && rsize < MIN_CHUNK_SIZE) | ||
4085 | set_inuse_and_pinuse(gm, p, small_index2size(i)); | ||
4086 | else { | ||
4087 | set_size_and_pinuse_of_inuse_chunk(gm, p, nb); | ||
4088 | r = chunk_plus_offset(p, nb); | ||
4089 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
4090 | replace_dv(gm, r, rsize); | ||
4091 | } | ||
4092 | mem = chunk2mem(p); | ||
4093 | check_malloced_chunk(gm, mem, nb); | ||
4094 | goto postaction; | ||
4095 | } | ||
4096 | |||
4097 | else if (gm->treemap != 0 && (mem = tmalloc_small(gm, nb)) != 0) { | ||
4098 | check_malloced_chunk(gm, mem, nb); | ||
4099 | goto postaction; | ||
4100 | } | ||
4101 | } | ||
4102 | } | ||
4103 | else if (bytes >= MAX_REQUEST) | ||
4104 | nb = MAX_SIZE_T; /* Too big to allocate. Force failure (in sys alloc) */ | ||
4105 | else { | ||
4106 | nb = pad_request(bytes); | ||
4107 | if (gm->treemap != 0 && (mem = tmalloc_large(gm, nb)) != 0) { | ||
4108 | check_malloced_chunk(gm, mem, nb); | ||
4109 | goto postaction; | ||
4110 | } | ||
4111 | } | ||
4112 | |||
4113 | if (nb <= gm->dvsize) { | ||
4114 | size_t rsize = gm->dvsize - nb; | ||
4115 | mchunkptr p = gm->dv; | ||
4116 | if (rsize >= MIN_CHUNK_SIZE) { /* split dv */ | ||
4117 | mchunkptr r = gm->dv = chunk_plus_offset(p, nb); | ||
4118 | gm->dvsize = rsize; | ||
4119 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
4120 | set_size_and_pinuse_of_inuse_chunk(gm, p, nb); | ||
4121 | } | ||
4122 | else { /* exhaust dv */ | ||
4123 | size_t dvs = gm->dvsize; | ||
4124 | gm->dvsize = 0; | ||
4125 | gm->dv = 0; | ||
4126 | set_inuse_and_pinuse(gm, p, dvs); | ||
4127 | } | ||
4128 | mem = chunk2mem(p); | ||
4129 | check_malloced_chunk(gm, mem, nb); | ||
4130 | goto postaction; | ||
4131 | } | ||
4132 | |||
4133 | else if (nb < gm->topsize) { /* Split top */ | ||
4134 | size_t rsize = gm->topsize -= nb; | ||
4135 | mchunkptr p = gm->top; | ||
4136 | mchunkptr r = gm->top = chunk_plus_offset(p, nb); | ||
4137 | r->head = rsize | PINUSE_BIT; | ||
4138 | set_size_and_pinuse_of_inuse_chunk(gm, p, nb); | ||
4139 | mem = chunk2mem(p); | ||
4140 | check_top_chunk(gm, gm->top); | ||
4141 | check_malloced_chunk(gm, mem, nb); | ||
4142 | goto postaction; | ||
4143 | } | ||
4144 | |||
4145 | mem = sys_alloc(gm, nb); | ||
4146 | |||
4147 | postaction: | ||
4148 | POSTACTION(gm); | ||
4149 | return mem; | ||
4150 | } | ||
4151 | |||
4152 | return 0; | ||
4153 | } | ||
4154 | |||
4155 | void dlfree(void* mem) { | ||
4156 | /* | ||
4157 | Consolidate freed chunks with preceeding or succeeding bordering | ||
4158 | free chunks, if they exist, and then place in a bin. Intermixed | ||
4159 | with special cases for top, dv, mmapped chunks, and usage errors. | ||
4160 | */ | ||
4161 | |||
4162 | if (mem != 0) { | ||
4163 | mchunkptr p = mem2chunk(mem); | ||
4164 | #if FOOTERS | ||
4165 | mstate fm = get_mstate_for(p); | ||
4166 | if (!ok_magic(fm)) { | ||
4167 | USAGE_ERROR_ACTION(fm, p); | ||
4168 | return; | ||
4169 | } | ||
4170 | #else /* FOOTERS */ | ||
4171 | #define fm gm | ||
4172 | #endif /* FOOTERS */ | ||
4173 | if (!PREACTION(fm)) { | ||
4174 | check_inuse_chunk(fm, p); | ||
4175 | if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) { | ||
4176 | size_t psize = chunksize(p); | ||
4177 | mchunkptr next = chunk_plus_offset(p, psize); | ||
4178 | if (!pinuse(p)) { | ||
4179 | size_t prevsize = p->prev_foot; | ||
4180 | if ((prevsize & IS_MMAPPED_BIT) != 0) { | ||
4181 | prevsize &= ~IS_MMAPPED_BIT; | ||
4182 | psize += prevsize + MMAP_FOOT_PAD; | ||
4183 | if (CALL_MUNMAP((char*)p - prevsize, psize) == 0) | ||
4184 | fm->footprint -= psize; | ||
4185 | goto postaction; | ||
4186 | } | ||
4187 | else { | ||
4188 | mchunkptr prev = chunk_minus_offset(p, prevsize); | ||
4189 | psize += prevsize; | ||
4190 | p = prev; | ||
4191 | if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */ | ||
4192 | if (p != fm->dv) { | ||
4193 | unlink_chunk(fm, p, prevsize); | ||
4194 | } | ||
4195 | else if ((next->head & INUSE_BITS) == INUSE_BITS) { | ||
4196 | fm->dvsize = psize; | ||
4197 | set_free_with_pinuse(p, psize, next); | ||
4198 | goto postaction; | ||
4199 | } | ||
4200 | } | ||
4201 | else | ||
4202 | goto erroraction; | ||
4203 | } | ||
4204 | } | ||
4205 | |||
4206 | if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) { | ||
4207 | if (!cinuse(next)) { /* consolidate forward */ | ||
4208 | if (next == fm->top) { | ||
4209 | size_t tsize = fm->topsize += psize; | ||
4210 | fm->top = p; | ||
4211 | p->head = tsize | PINUSE_BIT; | ||
4212 | if (p == fm->dv) { | ||
4213 | fm->dv = 0; | ||
4214 | fm->dvsize = 0; | ||
4215 | } | ||
4216 | if (should_trim(fm, tsize)) | ||
4217 | sys_trim(fm, 0); | ||
4218 | goto postaction; | ||
4219 | } | ||
4220 | else if (next == fm->dv) { | ||
4221 | size_t dsize = fm->dvsize += psize; | ||
4222 | fm->dv = p; | ||
4223 | set_size_and_pinuse_of_free_chunk(p, dsize); | ||
4224 | goto postaction; | ||
4225 | } | ||
4226 | else { | ||
4227 | size_t nsize = chunksize(next); | ||
4228 | psize += nsize; | ||
4229 | unlink_chunk(fm, next, nsize); | ||
4230 | set_size_and_pinuse_of_free_chunk(p, psize); | ||
4231 | if (p == fm->dv) { | ||
4232 | fm->dvsize = psize; | ||
4233 | goto postaction; | ||
4234 | } | ||
4235 | } | ||
4236 | } | ||
4237 | else | ||
4238 | set_free_with_pinuse(p, psize, next); | ||
4239 | insert_chunk(fm, p, psize); | ||
4240 | check_free_chunk(fm, p); | ||
4241 | goto postaction; | ||
4242 | } | ||
4243 | } | ||
4244 | erroraction: | ||
4245 | USAGE_ERROR_ACTION(fm, p); | ||
4246 | postaction: | ||
4247 | POSTACTION(fm); | ||
4248 | } | ||
4249 | } | ||
4250 | #if !FOOTERS | ||
4251 | #undef fm | ||
4252 | #endif /* FOOTERS */ | ||
4253 | } | ||
4254 | |||
4255 | void* dlcalloc(size_t n_elements, size_t elem_size) { | ||
4256 | void* mem; | ||
4257 | size_t req = 0; | ||
4258 | if (n_elements != 0) { | ||
4259 | req = n_elements * elem_size; | ||
4260 | if (((n_elements | elem_size) & ~(size_t)0xffff) && | ||
4261 | (req / n_elements != elem_size)) | ||
4262 | req = MAX_SIZE_T; /* force downstream failure on overflow */ | ||
4263 | } | ||
4264 | mem = dlmalloc(req); | ||
4265 | if (mem != 0 && calloc_must_clear(mem2chunk(mem))) | ||
4266 | memset(mem, 0, req); | ||
4267 | return mem; | ||
4268 | } | ||
4269 | |||
4270 | void* dlrealloc(void* oldmem, size_t bytes) { | ||
4271 | if (oldmem == 0) | ||
4272 | return dlmalloc(bytes); | ||
4273 | #ifdef REALLOC_ZERO_BYTES_FREES | ||
4274 | if (bytes == 0) { | ||
4275 | dlfree(oldmem); | ||
4276 | return 0; | ||
4277 | } | ||
4278 | #endif /* REALLOC_ZERO_BYTES_FREES */ | ||
4279 | else { | ||
4280 | #if ! FOOTERS | ||
4281 | mstate m = gm; | ||
4282 | #else /* FOOTERS */ | ||
4283 | mstate m = get_mstate_for(mem2chunk(oldmem)); | ||
4284 | if (!ok_magic(m)) { | ||
4285 | USAGE_ERROR_ACTION(m, oldmem); | ||
4286 | return 0; | ||
4287 | } | ||
4288 | #endif /* FOOTERS */ | ||
4289 | return internal_realloc(m, oldmem, bytes); | ||
4290 | } | ||
4291 | } | ||
4292 | |||
4293 | void* dlmemalign(size_t alignment, size_t bytes) { | ||
4294 | return internal_memalign(gm, alignment, bytes); | ||
4295 | } | ||
4296 | |||
4297 | void** dlindependent_calloc(size_t n_elements, size_t elem_size, | ||
4298 | void* chunks[]) { | ||
4299 | size_t sz = elem_size; /* serves as 1-element array */ | ||
4300 | return ialloc(gm, n_elements, &sz, 3, chunks); | ||
4301 | } | ||
4302 | |||
4303 | void** dlindependent_comalloc(size_t n_elements, size_t sizes[], | ||
4304 | void* chunks[]) { | ||
4305 | return ialloc(gm, n_elements, sizes, 0, chunks); | ||
4306 | } | ||
4307 | |||
4308 | void* dlvalloc(size_t bytes) { | ||
4309 | size_t pagesz; | ||
4310 | init_mparams(); | ||
4311 | pagesz = mparams.page_size; | ||
4312 | return dlmemalign(pagesz, bytes); | ||
4313 | } | ||
4314 | |||
4315 | void* dlpvalloc(size_t bytes) { | ||
4316 | size_t pagesz; | ||
4317 | init_mparams(); | ||
4318 | pagesz = mparams.page_size; | ||
4319 | return dlmemalign(pagesz, (bytes + pagesz - SIZE_T_ONE) & ~(pagesz - SIZE_T_ONE)); | ||
4320 | } | ||
4321 | |||
4322 | int dlmalloc_trim(size_t pad) { | ||
4323 | int result = 0; | ||
4324 | if (!PREACTION(gm)) { | ||
4325 | result = sys_trim(gm, pad); | ||
4326 | POSTACTION(gm); | ||
4327 | } | ||
4328 | return result; | ||
4329 | } | ||
4330 | |||
4331 | size_t dlmalloc_footprint(void) { | ||
4332 | return gm->footprint; | ||
4333 | } | ||
4334 | |||
4335 | size_t dlmalloc_max_footprint(void) { | ||
4336 | return gm->max_footprint; | ||
4337 | } | ||
4338 | |||
4339 | #if !NO_MALLINFO | ||
4340 | struct mallinfo dlmallinfo(void) { | ||
4341 | return internal_mallinfo(gm); | ||
4342 | } | ||
4343 | #endif /* NO_MALLINFO */ | ||
4344 | |||
4345 | void dlmalloc_stats() { | ||
4346 | internal_malloc_stats(gm); | ||
4347 | } | ||
4348 | |||
4349 | size_t dlmalloc_usable_size(void* mem) { | ||
4350 | if (mem != 0) { | ||
4351 | mchunkptr p = mem2chunk(mem); | ||
4352 | if (cinuse(p)) | ||
4353 | return chunksize(p) - overhead_for(p); | ||
4354 | } | ||
4355 | return 0; | ||
4356 | } | ||
4357 | |||
4358 | int dlmallopt(int param_number, int value) { | ||
4359 | return change_mparam(param_number, value); | ||
4360 | } | ||
4361 | |||
4362 | #endif /* !ONLY_MSPACES */ | ||
4363 | |||
4364 | /* ----------------------------- user mspaces ---------------------------- */ | ||
4365 | |||
4366 | #if MSPACES | ||
4367 | |||
4368 | static mstate init_user_mstate(char* tbase, size_t tsize) { | ||
4369 | size_t msize = pad_request(sizeof(struct malloc_state)); | ||
4370 | mchunkptr mn; | ||
4371 | mchunkptr msp = align_as_chunk(tbase); | ||
4372 | mstate m = (mstate)(chunk2mem(msp)); | ||
4373 | memset(m, 0, msize); | ||
4374 | INITIAL_LOCK(&m->mutex); | ||
4375 | msp->head = (msize|PINUSE_BIT|CINUSE_BIT); | ||
4376 | m->seg.base = m->least_addr = tbase; | ||
4377 | m->seg.size = m->footprint = m->max_footprint = tsize; | ||
4378 | m->magic = mparams.magic; | ||
4379 | m->mflags = mparams.default_mflags; | ||
4380 | disable_contiguous(m); | ||
4381 | init_bins(m); | ||
4382 | mn = next_chunk(mem2chunk(m)); | ||
4383 | init_top(m, mn, (size_t)((tbase + tsize) - (char*)mn) - TOP_FOOT_SIZE); | ||
4384 | check_top_chunk(m, m->top); | ||
4385 | return m; | ||
4386 | } | ||
4387 | |||
4388 | mspace create_mspace(size_t capacity, int locked) { | ||
4389 | mstate m = 0; | ||
4390 | size_t msize = pad_request(sizeof(struct malloc_state)); | ||
4391 | init_mparams(); /* Ensure pagesize etc initialized */ | ||
4392 | |||
4393 | if (capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) { | ||
4394 | size_t rs = ((capacity == 0)? mparams.granularity : | ||
4395 | (capacity + TOP_FOOT_SIZE + msize)); | ||
4396 | size_t tsize = granularity_align(rs); | ||
4397 | char* tbase = (char*)(CALL_MMAP(tsize)); | ||
4398 | if (tbase != CMFAIL) { | ||
4399 | m = init_user_mstate(tbase, tsize); | ||
4400 | m->seg.sflags = IS_MMAPPED_BIT; | ||
4401 | set_lock(m, locked); | ||
4402 | } | ||
4403 | } | ||
4404 | return (mspace)m; | ||
4405 | } | ||
4406 | |||
4407 | mspace create_mspace_with_base(void* base, size_t capacity, int locked) { | ||
4408 | mstate m = 0; | ||
4409 | size_t msize = pad_request(sizeof(struct malloc_state)); | ||
4410 | init_mparams(); /* Ensure pagesize etc initialized */ | ||
4411 | |||
4412 | if (capacity > msize + TOP_FOOT_SIZE && | ||
4413 | capacity < (size_t) -(msize + TOP_FOOT_SIZE + mparams.page_size)) { | ||
4414 | m = init_user_mstate((char*)base, capacity); | ||
4415 | m->seg.sflags = EXTERN_BIT; | ||
4416 | set_lock(m, locked); | ||
4417 | } | ||
4418 | return (mspace)m; | ||
4419 | } | ||
4420 | |||
4421 | size_t destroy_mspace(mspace msp) { | ||
4422 | size_t freed = 0; | ||
4423 | mstate ms = (mstate)msp; | ||
4424 | if (ok_magic(ms)) { | ||
4425 | msegmentptr sp = &ms->seg; | ||
4426 | while (sp != 0) { | ||
4427 | char* base = sp->base; | ||
4428 | size_t size = sp->size; | ||
4429 | flag_t flag = sp->sflags; | ||
4430 | sp = sp->next; | ||
4431 | if ((flag & IS_MMAPPED_BIT) && !(flag & EXTERN_BIT) && | ||
4432 | CALL_MUNMAP(base, size) == 0) | ||
4433 | freed += size; | ||
4434 | } | ||
4435 | } | ||
4436 | else { | ||
4437 | USAGE_ERROR_ACTION(ms,ms); | ||
4438 | } | ||
4439 | return freed; | ||
4440 | } | ||
4441 | |||
4442 | /* | ||
4443 | mspace versions of routines are near-clones of the global | ||
4444 | versions. This is not so nice but better than the alternatives. | ||
4445 | */ | ||
4446 | |||
4447 | |||
4448 | void* mspace_malloc(mspace msp, size_t bytes) { | ||
4449 | mstate ms = (mstate)msp; | ||
4450 | if (!ok_magic(ms)) { | ||
4451 | USAGE_ERROR_ACTION(ms,ms); | ||
4452 | return 0; | ||
4453 | } | ||
4454 | if (!PREACTION(ms)) { | ||
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 = ms->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(ms, idx); | ||
4468 | p = b->fd; | ||
4469 | assert(chunksize(p) == small_index2size(idx)); | ||
4470 | unlink_first_small_chunk(ms, b, p, idx); | ||
4471 | set_inuse_and_pinuse(ms, p, small_index2size(idx)); | ||
4472 | mem = chunk2mem(p); | ||
4473 | check_malloced_chunk(ms, mem, nb); | ||
4474 | goto postaction; | ||
4475 | } | ||
4476 | |||
4477 | else if (nb > ms->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(ms, i); | ||
4486 | p = b->fd; | ||
4487 | assert(chunksize(p) == small_index2size(i)); | ||
4488 | unlink_first_small_chunk(ms, 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(ms, p, small_index2size(i)); | ||
4493 | else { | ||
4494 | set_size_and_pinuse_of_inuse_chunk(ms, p, nb); | ||
4495 | r = chunk_plus_offset(p, nb); | ||
4496 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
4497 | replace_dv(ms, r, rsize); | ||
4498 | } | ||
4499 | mem = chunk2mem(p); | ||
4500 | check_malloced_chunk(ms, mem, nb); | ||
4501 | goto postaction; | ||
4502 | } | ||
4503 | |||
4504 | else if (ms->treemap != 0 && (mem = tmalloc_small(ms, nb)) != 0) { | ||
4505 | check_malloced_chunk(ms, 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 (ms->treemap != 0 && (mem = tmalloc_large(ms, nb)) != 0) { | ||
4515 | check_malloced_chunk(ms, mem, nb); | ||
4516 | goto postaction; | ||
4517 | } | ||
4518 | } | ||
4519 | |||
4520 | if (nb <= ms->dvsize) { | ||
4521 | size_t rsize = ms->dvsize - nb; | ||
4522 | mchunkptr p = ms->dv; | ||
4523 | if (rsize >= MIN_CHUNK_SIZE) { /* split dv */ | ||
4524 | mchunkptr r = ms->dv = chunk_plus_offset(p, nb); | ||
4525 | ms->dvsize = rsize; | ||
4526 | set_size_and_pinuse_of_free_chunk(r, rsize); | ||
4527 | set_size_and_pinuse_of_inuse_chunk(ms, p, nb); | ||
4528 | } | ||
4529 | else { /* exhaust dv */ | ||
4530 | size_t dvs = ms->dvsize; | ||
4531 | ms->dvsize = 0; | ||
4532 | ms->dv = 0; | ||
4533 | set_inuse_and_pinuse(ms, p, dvs); | ||
4534 | } | ||
4535 | mem = chunk2mem(p); | ||
4536 | check_malloced_chunk(ms, mem, nb); | ||
4537 | goto postaction; | ||
4538 | } | ||
4539 | |||
4540 | else if (nb < ms->topsize) { /* Split top */ | ||
4541 | size_t rsize = ms->topsize -= nb; | ||
4542 | mchunkptr p = ms->top; | ||
4543 | mchunkptr r = ms->top = chunk_plus_offset(p, nb); | ||
4544 | r->head = rsize | PINUSE_BIT; | ||
4545 | set_size_and_pinuse_of_inuse_chunk(ms, p, nb); | ||
4546 | mem = chunk2mem(p); | ||
4547 | check_top_chunk(ms, ms->top); | ||
4548 | check_malloced_chunk(ms, mem, nb); | ||
4549 | goto postaction; | ||
4550 | } | ||
4551 | |||
4552 | mem = sys_alloc(ms, nb); | ||
4553 | |||
4554 | postaction: | ||
4555 | POSTACTION(ms); | ||
4556 | return mem; | ||
4557 | } | ||
4558 | |||
4559 | return 0; | ||
4560 | } | ||
4561 | |||
4562 | void mspace_free(mspace msp, void* mem) { | ||
4563 | if (mem != 0) { | ||
4564 | mchunkptr p = mem2chunk(mem); | ||
4565 | #if FOOTERS | ||
4566 | mstate fm = get_mstate_for(p); | ||
4567 | #else /* FOOTERS */ | ||
4568 | mstate fm = (mstate)msp; | ||
4569 | #endif /* FOOTERS */ | ||
4570 | if (!ok_magic(fm)) { | ||
4571 | USAGE_ERROR_ACTION(fm, p); | ||
4572 | return; | ||
4573 | } | ||
4574 | if (!PREACTION(fm)) { | ||
4575 | check_inuse_chunk(fm, p); | ||
4576 | if (RTCHECK(ok_address(fm, p) && ok_cinuse(p))) { | ||
4577 | size_t psize = chunksize(p); | ||
4578 | mchunkptr next = chunk_plus_offset(p, psize); | ||
4579 | if (!pinuse(p)) { | ||
4580 | size_t prevsize = p->prev_foot; | ||
4581 | if ((prevsize & IS_MMAPPED_BIT) != 0) { | ||
4582 | prevsize &= ~IS_MMAPPED_BIT; | ||
4583 | psize += prevsize + MMAP_FOOT_PAD; | ||
4584 | if (CALL_MUNMAP((char*)p - prevsize, psize) == 0) | ||
4585 | fm->footprint -= psize; | ||
4586 | goto postaction; | ||
4587 | } | ||
4588 | else { | ||
4589 | mchunkptr prev = chunk_minus_offset(p, prevsize); | ||
4590 | psize += prevsize; | ||
4591 | p = prev; | ||
4592 | if (RTCHECK(ok_address(fm, prev))) { /* consolidate backward */ | ||
4593 | if (p != fm->dv) { | ||
4594 | unlink_chunk(fm, p, prevsize); | ||
4595 | } | ||
4596 | else if ((next->head & INUSE_BITS) == INUSE_BITS) { | ||
4597 | fm->dvsize = psize; | ||
4598 | set_free_with_pinuse(p, psize, next); | ||
4599 | goto postaction; | ||
4600 | } | ||
4601 | } | ||
4602 | else | ||
4603 | goto erroraction; | ||
4604 | } | ||
4605 | } | ||
4606 | |||
4607 | if (RTCHECK(ok_next(p, next) && ok_pinuse(next))) { | ||
4608 | if (!cinuse(next)) { /* consolidate forward */ | ||
4609 | if (next == fm->top) { | ||
4610 | size_t tsize = fm->topsize += psize; | ||
4611 | fm->top = p; | ||
4612 | p->head = tsize | PINUSE_BIT; | ||
4613 | if (p == fm->dv) { | ||
4614 | fm->dv = 0; | ||
4615 | fm->dvsize = 0; | ||
4616 | } | ||
4617 | if (should_trim(fm, tsize)) | ||
4618 | sys_trim(fm, 0); | ||
4619 | goto postaction; | ||
4620 | } | ||
4621 | else if (next == fm->dv) { | ||
4622 | size_t dsize = fm->dvsize += psize; | ||
4623 | fm->dv = p; | ||
4624 | set_size_and_pinuse_of_free_chunk(p, dsize); | ||
4625 | goto postaction; | ||
4626 | } | ||
4627 | else { | ||
4628 | size_t nsize = chunksize(next); | ||
4629 | psize += nsize; | ||
4630 | unlink_chunk(fm, next, nsize); | ||
4631 | set_size_and_pinuse_of_free_chunk(p, psize); | ||
4632 | if (p == fm->dv) { | ||
4633 | fm->dvsize = psize; | ||
4634 | goto postaction; | ||
4635 | } | ||
4636 | } | ||
4637 | } | ||
4638 | else | ||
4639 | set_free_with_pinuse(p, psize, next); | ||
4640 | insert_chunk(fm, p, psize); | ||
4641 | check_free_chunk(fm, p); | ||
4642 | goto postaction; | ||
4643 | } | ||
4644 | } | ||
4645 | erroraction: | ||
4646 | USAGE_ERROR_ACTION(fm, p); | ||
4647 | postaction: | ||
4648 | POSTACTION(fm); | ||
4649 | } | ||
4650 | } | ||
4651 | } | ||
4652 | |||
4653 | void* mspace_calloc(mspace msp, size_t n_elements, size_t elem_size) { | ||
4654 | void* mem; | ||
4655 | size_t req = 0; | ||
4656 | mstate ms = (mstate)msp; | ||
4657 | if (!ok_magic(ms)) { | ||
4658 | USAGE_ERROR_ACTION(ms,ms); | ||
4659 | return 0; | ||
4660 | } | ||
4661 | if (n_elements != 0) { | ||
4662 | req = n_elements * elem_size; | ||
4663 | if (((n_elements | elem_size) & ~(size_t)0xffff) && | ||
4664 | (req / n_elements != elem_size)) | ||
4665 | req = MAX_SIZE_T; /* force downstream failure on overflow */ | ||
4666 | } | ||
4667 | mem = internal_malloc(ms, req); | ||
4668 | if (mem != 0 && calloc_must_clear(mem2chunk(mem))) | ||
4669 | memset(mem, 0, req); | ||
4670 | return mem; | ||
4671 | } | ||
4672 | |||
4673 | void* mspace_realloc(mspace msp, void* oldmem, size_t bytes) { | ||
4674 | if (oldmem == 0) | ||
4675 | return mspace_malloc(msp, bytes); | ||
4676 | #ifdef REALLOC_ZERO_BYTES_FREES | ||
4677 | if (bytes == 0) { | ||
4678 | mspace_free(msp, oldmem); | ||
4679 | return 0; | ||
4680 | } | ||
4681 | #endif /* REALLOC_ZERO_BYTES_FREES */ | ||
4682 | else { | ||
4683 | #if FOOTERS | ||
4684 | mchunkptr p = mem2chunk(oldmem); | ||
4685 | mstate ms = get_mstate_for(p); | ||
4686 | #else /* FOOTERS */ | ||
4687 | mstate ms = (mstate)msp; | ||
4688 | #endif /* FOOTERS */ | ||
4689 | if (!ok_magic(ms)) { | ||
4690 | USAGE_ERROR_ACTION(ms,ms); | ||
4691 | return 0; | ||
4692 | } | ||
4693 | return internal_realloc(ms, oldmem, bytes); | ||
4694 | } | ||
4695 | } | ||
4696 | |||
4697 | void* mspace_memalign(mspace msp, size_t alignment, size_t bytes) { | ||
4698 | mstate ms = (mstate)msp; | ||
4699 | if (!ok_magic(ms)) { | ||
4700 | USAGE_ERROR_ACTION(ms,ms); | ||
4701 | return 0; | ||
4702 | } | ||
4703 | return internal_memalign(ms, alignment, bytes); | ||
4704 | } | ||
4705 | |||
4706 | void** mspace_independent_calloc(mspace msp, size_t n_elements, | ||
4707 | size_t elem_size, void* chunks[]) { | ||
4708 | size_t sz = elem_size; /* serves as 1-element array */ | ||
4709 | mstate ms = (mstate)msp; | ||
4710 | if (!ok_magic(ms)) { | ||
4711 | USAGE_ERROR_ACTION(ms,ms); | ||
4712 | return 0; | ||
4713 | } | ||
4714 | return ialloc(ms, n_elements, &sz, 3, chunks); | ||
4715 | } | ||
4716 | |||
4717 | void** mspace_independent_comalloc(mspace msp, size_t n_elements, | ||
4718 | size_t sizes[], void* chunks[]) { | ||
4719 | mstate ms = (mstate)msp; | ||
4720 | if (!ok_magic(ms)) { | ||
4721 | USAGE_ERROR_ACTION(ms,ms); | ||
4722 | return 0; | ||
4723 | } | ||
4724 | return ialloc(ms, n_elements, sizes, 0, chunks); | ||
4725 | } | ||
4726 | |||
4727 | int mspace_trim(mspace msp, size_t pad) { | ||
4728 | int result = 0; | ||
4729 | mstate ms = (mstate)msp; | ||
4730 | if (ok_magic(ms)) { | ||
4731 | if (!PREACTION(ms)) { | ||
4732 | result = sys_trim(ms, pad); | ||
4733 | POSTACTION(ms); | ||
4734 | } | ||
4735 | } | ||
4736 | else { | ||
4737 | USAGE_ERROR_ACTION(ms,ms); | ||
4738 | } | ||
4739 | return result; | ||
4740 | } | ||
4741 | |||
4742 | void mspace_malloc_stats(mspace msp) { | ||
4743 | mstate ms = (mstate)msp; | ||
4744 | if (ok_magic(ms)) { | ||
4745 | internal_malloc_stats(ms); | ||
4746 | } | ||
4747 | else { | ||
4748 | USAGE_ERROR_ACTION(ms,ms); | ||
4749 | } | ||
4750 | } | ||
4751 | |||
4752 | size_t mspace_footprint(mspace msp) { | ||
4753 | size_t result; | ||
4754 | mstate ms = (mstate)msp; | ||
4755 | if (ok_magic(ms)) { | ||
4756 | result = ms->footprint; | ||
4757 | } | ||
4758 | USAGE_ERROR_ACTION(ms,ms); | ||
4759 | return result; | ||
4760 | } | ||
4761 | |||
4762 | |||
4763 | size_t mspace_max_footprint(mspace msp) { | ||
4764 | size_t result; | ||
4765 | mstate ms = (mstate)msp; | ||
4766 | if (ok_magic(ms)) { | ||
4767 | result = ms->max_footprint; | ||
4768 | } | ||
4769 | USAGE_ERROR_ACTION(ms,ms); | ||
4770 | return result; | ||
4771 | } | ||
4772 | |||
4773 | |||
4774 | #if !NO_MALLINFO | ||
4775 | struct mallinfo mspace_mallinfo(mspace msp) { | ||
4776 | mstate ms = (mstate)msp; | ||
4777 | if (!ok_magic(ms)) { | ||
4778 | USAGE_ERROR_ACTION(ms,ms); | ||
4779 | } | ||
4780 | return internal_mallinfo(ms); | ||
4781 | } | ||
4782 | #endif /* NO_MALLINFO */ | ||
4783 | |||
4784 | int mspace_mallopt(int param_number, int value) { | ||
4785 | return change_mparam(param_number, value); | ||
4786 | } | ||
4787 | |||
4788 | #endif /* MSPACES */ | ||
4789 | |||
4790 | /* -------------------- Alternative MORECORE functions ------------------- */ | ||
4791 | |||
4792 | /* | ||
4793 | Guidelines for creating a custom version of MORECORE: | ||
4794 | |||
4795 | * For best performance, MORECORE should allocate in multiples of pagesize. | ||
4796 | * MORECORE may allocate more memory than requested. (Or even less, | ||
4797 | but this will usually result in a malloc failure.) | ||
4798 | * MORECORE must not allocate memory when given argument zero, but | ||
4799 | instead return one past the end address of memory from previous | ||
4800 | nonzero call. | ||
4801 | * For best performance, consecutive calls to MORECORE with positive | ||
4802 | arguments should return increasing addresses, indicating that | ||
4803 | space has been contiguously extended. | ||
4804 | * Even though consecutive calls to MORECORE need not return contiguous | ||
4805 | addresses, it must be OK for malloc'ed chunks to span multiple | ||
4806 | regions in those cases where they do happen to be contiguous. | ||
4807 | * MORECORE need not handle negative arguments -- it may instead | ||
4808 | just return MFAIL when given negative arguments. | ||
4809 | Negative arguments are always multiples of pagesize. MORECORE | ||
4810 | must not misinterpret negative args as large positive unsigned | ||
4811 | args. You can suppress all such calls from even occurring by defining | ||
4812 | MORECORE_CANNOT_TRIM, | ||
4813 | |||
4814 | As an example alternative MORECORE, here is a custom allocator | ||
4815 | kindly contributed for pre-OSX macOS. It uses virtually but not | ||
4816 | necessarily physically contiguous non-paged memory (locked in, | ||
4817 | present and won't get swapped out). You can use it by uncommenting | ||
4818 | this section, adding some #includes, and setting up the appropriate | ||
4819 | defines above: | ||
4820 | |||
4821 | #define MORECORE osMoreCore | ||
4822 | |||
4823 | There is also a shutdown routine that should somehow be called for | ||
4824 | cleanup upon program exit. | ||
4825 | |||
4826 | #define MAX_POOL_ENTRIES 100 | ||
4827 | #define MINIMUM_MORECORE_SIZE (64 * 1024U) | ||
4828 | static int next_os_pool; | ||
4829 | void *our_os_pools[MAX_POOL_ENTRIES]; | ||
4830 | |||
4831 | void *osMoreCore(int size) | ||
4832 | { | ||
4833 | void *ptr = 0; | ||
4834 | static void *sbrk_top = 0; | ||
4835 | |||
4836 | if (size > 0) | ||
4837 | { | ||
4838 | if (size < MINIMUM_MORECORE_SIZE) | ||
4839 | size = MINIMUM_MORECORE_SIZE; | ||
4840 | if (CurrentExecutionLevel() == kTaskLevel) | ||
4841 | ptr = PoolAllocateResident(size + RM_PAGE_SIZE, 0); | ||
4842 | if (ptr == 0) | ||
4843 | { | ||
4844 | return (void *) MFAIL; | ||
4845 | } | ||
4846 | // save ptrs so they can be freed during cleanup | ||
4847 | our_os_pools[next_os_pool] = ptr; | ||
4848 | next_os_pool++; | ||
4849 | ptr = (void *) ((((size_t) ptr) + RM_PAGE_MASK) & ~RM_PAGE_MASK); | ||
4850 | sbrk_top = (char *) ptr + size; | ||
4851 | return ptr; | ||
4852 | } | ||
4853 | else if (size < 0) | ||
4854 | { | ||
4855 | // we don't currently support shrink behavior | ||
4856 | return (void *) MFAIL; | ||
4857 | } | ||
4858 | else | ||
4859 | { | ||
4860 | return sbrk_top; | ||
4861 | } | ||
4862 | } | ||
4863 | |||
4864 | // cleanup any allocated memory pools | ||
4865 | // called as last thing before shutting down driver | ||
4866 | |||
4867 | void osCleanupMem(void) | ||
4868 | { | ||
4869 | void **ptr; | ||
4870 | |||
4871 | for (ptr = our_os_pools; ptr < &our_os_pools[MAX_POOL_ENTRIES]; ptr++) | ||
4872 | if (*ptr) | ||
4873 | { | ||
4874 | PoolDeallocate(*ptr); | ||
4875 | *ptr = 0; | ||
4876 | } | ||
4877 | } | ||
4878 | |||
4879 | */ | ||
4880 | |||
4881 | |||
4882 | /* ----------------------------------------------------------------------- | ||
4883 | History: | ||
4884 | V2.8.3 Thu Sep 22 11:16:32 2005 Doug Lea (dl at gee) | ||
4885 | * Add max_footprint functions | ||
4886 | * Ensure all appropriate literals are size_t | ||
4887 | * Fix conditional compilation problem for some #define settings | ||
4888 | * Avoid concatenating segments with the one provided | ||
4889 | in create_mspace_with_base | ||
4890 | * Rename some variables to avoid compiler shadowing warnings | ||
4891 | * Use explicit lock initialization. | ||
4892 | * Better handling of sbrk interference. | ||
4893 | * Simplify and fix segment insertion, trimming and mspace_destroy | ||
4894 | * Reinstate REALLOC_ZERO_BYTES_FREES option from 2.7.x | ||
4895 | * Thanks especially to Dennis Flanagan for help on these. | ||
4896 | |||
4897 | V2.8.2 Sun Jun 12 16:01:10 2005 Doug Lea (dl at gee) | ||
4898 | * Fix memalign brace error. | ||
4899 | |||
4900 | V2.8.1 Wed Jun 8 16:11:46 2005 Doug Lea (dl at gee) | ||
4901 | * Fix improper #endif nesting in C++ | ||
4902 | * Add explicit casts needed for C++ | ||
4903 | |||
4904 | V2.8.0 Mon May 30 14:09:02 2005 Doug Lea (dl at gee) | ||
4905 | * Use trees for large bins | ||
4906 | * Support mspaces | ||
4907 | * Use segments to unify sbrk-based and mmap-based system allocation, | ||
4908 | removing need for emulation on most platforms without sbrk. | ||
4909 | * Default safety checks | ||
4910 | * Optional footer checks. Thanks to William Robertson for the idea. | ||
4911 | * Internal code refactoring | ||
4912 | * Incorporate suggestions and platform-specific changes. | ||
4913 | Thanks to Dennis Flanagan, Colin Plumb, Niall Douglas, | ||
4914 | Aaron Bachmann, Emery Berger, and others. | ||
4915 | * Speed up non-fastbin processing enough to remove fastbins. | ||
4916 | * Remove useless cfree() to avoid conflicts with other apps. | ||
4917 | * Remove internal memcpy, memset. Compilers handle builtins better. | ||
4918 | * Remove some options that no one ever used and rename others. | ||
4919 | |||
4920 | V2.7.2 Sat Aug 17 09:07:30 2002 Doug Lea (dl at gee) | ||
4921 | * Fix malloc_state bitmap array misdeclaration | ||
4922 | |||
4923 | V2.7.1 Thu Jul 25 10:58:03 2002 Doug Lea (dl at gee) | ||
4924 | * Allow tuning of FIRST_SORTED_BIN_SIZE | ||
4925 | * Use PTR_UINT as type for all ptr->int casts. Thanks to John Belmonte. | ||
4926 | * Better detection and support for non-contiguousness of MORECORE. | ||
4927 | Thanks to Andreas Mueller, Conal Walsh, and Wolfram Gloger | ||
4928 | * Bypass most of malloc if no frees. Thanks To Emery Berger. | ||
4929 | * Fix freeing of old top non-contiguous chunk im sysmalloc. | ||
4930 | * Raised default trim and map thresholds to 256K. | ||
4931 | * Fix mmap-related #defines. Thanks to Lubos Lunak. | ||
4932 | * Fix copy macros; added LACKS_FCNTL_H. Thanks to Neal Walfield. | ||
4933 | * Branch-free bin calculation | ||
4934 | * Default trim and mmap thresholds now 256K. | ||
4935 | |||
4936 | V2.7.0 Sun Mar 11 14:14:06 2001 Doug Lea (dl at gee) | ||
4937 | * Introduce independent_comalloc and independent_calloc. | ||
4938 | Thanks to Michael Pachos for motivation and help. | ||
4939 | * Make optional .h file available | ||
4940 | * Allow > 2GB requests on 32bit systems. | ||
4941 | * new WIN32 sbrk, mmap, munmap, lock code from <Walter@GeNeSys-e.de>. | ||
4942 | Thanks also to Andreas Mueller <a.mueller at paradatec.de>, | ||
4943 | and Anonymous. | ||
4944 | * Allow override of MALLOC_ALIGNMENT (Thanks to Ruud Waij for | ||
4945 | helping test this.) | ||
4946 | * memalign: check alignment arg | ||
4947 | * realloc: don't try to shift chunks backwards, since this | ||
4948 | leads to more fragmentation in some programs and doesn't | ||
4949 | seem to help in any others. | ||
4950 | * Collect all cases in malloc requiring system memory into sysmalloc | ||
4951 | * Use mmap as backup to sbrk | ||
4952 | * Place all internal state in malloc_state | ||
4953 | * Introduce fastbins (although similar to 2.5.1) | ||
4954 | * Many minor tunings and cosmetic improvements | ||
4955 | * Introduce USE_PUBLIC_MALLOC_WRAPPERS, USE_MALLOC_LOCK | ||
4956 | * Introduce MALLOC_FAILURE_ACTION, MORECORE_CONTIGUOUS | ||
4957 | Thanks to Tony E. Bennett <tbennett@nvidia.com> and others. | ||
4958 | * Include errno.h to support default failure action. | ||
4959 | |||
4960 | V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee) | ||
4961 | * return null for negative arguments | ||
4962 | * Added Several WIN32 cleanups from Martin C. Fong <mcfong at yahoo.com> | ||
4963 | * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h' | ||
4964 | (e.g. WIN32 platforms) | ||
4965 | * Cleanup header file inclusion for WIN32 platforms | ||
4966 | * Cleanup code to avoid Microsoft Visual C++ compiler complaints | ||
4967 | * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing | ||
4968 | memory allocation routines | ||
4969 | * Set 'malloc_getpagesize' for WIN32 platforms (needs more work) | ||
4970 | * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to | ||
4971 | usage of 'assert' in non-WIN32 code | ||
4972 | * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to | ||
4973 | avoid infinite loop | ||
4974 | * Always call 'fREe()' rather than 'free()' | ||
4975 | |||
4976 | V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee) | ||
4977 | * Fixed ordering problem with boundary-stamping | ||
4978 | |||
4979 | V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee) | ||
4980 | * Added pvalloc, as recommended by H.J. Liu | ||
4981 | * Added 64bit pointer support mainly from Wolfram Gloger | ||
4982 | * Added anonymously donated WIN32 sbrk emulation | ||
4983 | * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen | ||
4984 | * malloc_extend_top: fix mask error that caused wastage after | ||
4985 | foreign sbrks | ||
4986 | * Add linux mremap support code from HJ Liu | ||
4987 | |||
4988 | V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee) | ||
4989 | * Integrated most documentation with the code. | ||
4990 | * Add support for mmap, with help from | ||
4991 | Wolfram Gloger (Gloger@lrz.uni-muenchen.de). | ||
4992 | * Use last_remainder in more cases. | ||
4993 | * Pack bins using idea from colin@nyx10.cs.du.edu | ||
4994 | * Use ordered bins instead of best-fit threshhold | ||
4995 | * Eliminate block-local decls to simplify tracing and debugging. | ||
4996 | * Support another case of realloc via move into top | ||
4997 | * Fix error occuring when initial sbrk_base not word-aligned. | ||
4998 | * Rely on page size for units instead of SBRK_UNIT to | ||
4999 | avoid surprises about sbrk alignment conventions. | ||
5000 | * Add mallinfo, mallopt. Thanks to Raymond Nijssen | ||
5001 | (raymond@es.ele.tue.nl) for the suggestion. | ||
5002 | * Add `pad' argument to malloc_trim and top_pad mallopt parameter. | ||
5003 | * More precautions for cases where other routines call sbrk, | ||
5004 | courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de). | ||
5005 | * Added macros etc., allowing use in linux libc from | ||
5006 | H.J. Lu (hjl@gnu.ai.mit.edu) | ||
5007 | * Inverted this history list | ||
5008 | |||
5009 | V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee) | ||
5010 | * Re-tuned and fixed to behave more nicely with V2.6.0 changes. | ||
5011 | * Removed all preallocation code since under current scheme | ||
5012 | the work required to undo bad preallocations exceeds | ||
5013 | the work saved in good cases for most test programs. | ||
5014 | * No longer use return list or unconsolidated bins since | ||
5015 | no scheme using them consistently outperforms those that don't | ||
5016 | given above changes. | ||
5017 | * Use best fit for very large chunks to prevent some worst-cases. | ||
5018 | * Added some support for debugging | ||
5019 | |||
5020 | V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee) | ||
5021 | * Removed footers when chunks are in use. Thanks to | ||
5022 | Paul Wilson (wilson@cs.texas.edu) for the suggestion. | ||
5023 | |||
5024 | V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee) | ||
5025 | * Added malloc_trim, with help from Wolfram Gloger | ||
5026 | (wmglo@Dent.MED.Uni-Muenchen.DE). | ||
5027 | |||
5028 | V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g) | ||
5029 | |||
5030 | V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g) | ||
5031 | * realloc: try to expand in both directions | ||
5032 | * malloc: swap order of clean-bin strategy; | ||
5033 | * realloc: only conditionally expand backwards | ||
5034 | * Try not to scavenge used bins | ||
5035 | * Use bin counts as a guide to preallocation | ||
5036 | * Occasionally bin return list chunks in first scan | ||
5037 | * Add a few optimizations from colin@nyx10.cs.du.edu | ||
5038 | |||
5039 | V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g) | ||
5040 | * faster bin computation & slightly different binning | ||
5041 | * merged all consolidations to one part of malloc proper | ||
5042 | (eliminating old malloc_find_space & malloc_clean_bin) | ||
5043 | * Scan 2 returns chunks (not just 1) | ||
5044 | * Propagate failure in realloc if malloc returns 0 | ||
5045 | * Add stuff to allow compilation on non-ANSI compilers | ||
5046 | from kpv@research.att.com | ||
5047 | |||
5048 | V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu) | ||
5049 | * removed potential for odd address access in prev_chunk | ||
5050 | * removed dependency on getpagesize.h | ||
5051 | * misc cosmetics and a bit more internal documentation | ||
5052 | * anticosmetics: mangled names in macros to evade debugger strangeness | ||
5053 | * tested on sparc, hp-700, dec-mips, rs6000 | ||
5054 | with gcc & native cc (hp, dec only) allowing | ||
5055 | Detlefs & Zorn comparison study (in SIGPLAN Notices.) | ||
5056 | |||
5057 | Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu) | ||
5058 | * Based loosely on libg++-1.2X malloc. (It retains some of the overall | ||
5059 | structure of old version, but most details differ.) | ||
5060 | |||
5061 | */ | ||