1 | /* Malloc implementation for multiple threads without lock contention. |
2 | Copyright (C) 1996-2017 Free Software Foundation, Inc. |
3 | This file is part of the GNU C Library. |
4 | Contributed by Wolfram Gloger <wg@malloc.de> |
5 | and Doug Lea <dl@cs.oswego.edu>, 2001. |
6 | |
7 | The GNU C Library is free software; you can redistribute it and/or |
8 | modify it under the terms of the GNU Lesser General Public License as |
9 | published by the Free Software Foundation; either version 2.1 of the |
10 | License, or (at your option) any later version. |
11 | |
12 | The GNU C Library is distributed in the hope that it will be useful, |
13 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
15 | Lesser General Public License for more details. |
16 | |
17 | You should have received a copy of the GNU Lesser General Public |
18 | License along with the GNU C Library; see the file COPYING.LIB. If |
19 | not, see <http://www.gnu.org/licenses/>. */ |
20 | |
21 | /* |
22 | This is a version (aka ptmalloc2) of malloc/free/realloc written by |
23 | Doug Lea and adapted to multiple threads/arenas by Wolfram Gloger. |
24 | |
25 | There have been substantial changes made after the integration into |
26 | glibc in all parts of the code. Do not look for much commonality |
27 | with the ptmalloc2 version. |
28 | |
29 | * Version ptmalloc2-20011215 |
30 | based on: |
31 | VERSION 2.7.0 Sun Mar 11 14:14:06 2001 Doug Lea (dl at gee) |
32 | |
33 | * Quickstart |
34 | |
35 | In order to compile this implementation, a Makefile is provided with |
36 | the ptmalloc2 distribution, which has pre-defined targets for some |
37 | popular systems (e.g. "make posix" for Posix threads). All that is |
38 | typically required with regard to compiler flags is the selection of |
39 | the thread package via defining one out of USE_PTHREADS, USE_THR or |
40 | USE_SPROC. Check the thread-m.h file for what effects this has. |
41 | Many/most systems will additionally require USE_TSD_DATA_HACK to be |
42 | defined, so this is the default for "make posix". |
43 | |
44 | * Why use this malloc? |
45 | |
46 | This is not the fastest, most space-conserving, most portable, or |
47 | most tunable malloc ever written. However it is among the fastest |
48 | while also being among the most space-conserving, portable and tunable. |
49 | Consistent balance across these factors results in a good general-purpose |
50 | allocator for malloc-intensive programs. |
51 | |
52 | The main properties of the algorithms are: |
53 | * For large (>= 512 bytes) requests, it is a pure best-fit allocator, |
54 | with ties normally decided via FIFO (i.e. least recently used). |
55 | * For small (<= 64 bytes by default) requests, it is a caching |
56 | allocator, that maintains pools of quickly recycled chunks. |
57 | * In between, and for combinations of large and small requests, it does |
58 | the best it can trying to meet both goals at once. |
59 | * For very large requests (>= 128KB by default), it relies on system |
60 | memory mapping facilities, if supported. |
61 | |
62 | For a longer but slightly out of date high-level description, see |
63 | http://gee.cs.oswego.edu/dl/html/malloc.html |
64 | |
65 | You may already by default be using a C library containing a malloc |
66 | that is based on some version of this malloc (for example in |
67 | linux). You might still want to use the one in this file in order to |
68 | customize settings or to avoid overheads associated with library |
69 | versions. |
70 | |
71 | * Contents, described in more detail in "description of public routines" below. |
72 | |
73 | Standard (ANSI/SVID/...) functions: |
74 | malloc(size_t n); |
75 | calloc(size_t n_elements, size_t element_size); |
76 | free(void* p); |
77 | realloc(void* p, size_t n); |
78 | memalign(size_t alignment, size_t n); |
79 | valloc(size_t n); |
80 | mallinfo() |
81 | mallopt(int parameter_number, int parameter_value) |
82 | |
83 | Additional functions: |
84 | independent_calloc(size_t n_elements, size_t size, void* chunks[]); |
85 | independent_comalloc(size_t n_elements, size_t sizes[], void* chunks[]); |
86 | pvalloc(size_t n); |
87 | malloc_trim(size_t pad); |
88 | malloc_usable_size(void* p); |
89 | malloc_stats(); |
90 | |
91 | * Vital statistics: |
92 | |
93 | Supported pointer representation: 4 or 8 bytes |
94 | Supported size_t representation: 4 or 8 bytes |
95 | Note that size_t is allowed to be 4 bytes even if pointers are 8. |
96 | You can adjust this by defining INTERNAL_SIZE_T |
97 | |
98 | Alignment: 2 * sizeof(size_t) (default) |
99 | (i.e., 8 byte alignment with 4byte size_t). This suffices for |
100 | nearly all current machines and C compilers. However, you can |
101 | define MALLOC_ALIGNMENT to be wider than this if necessary. |
102 | |
103 | Minimum overhead per allocated chunk: 4 or 8 bytes |
104 | Each malloced chunk has a hidden word of overhead holding size |
105 | and status information. |
106 | |
107 | Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead) |
108 | 8-byte ptrs: 24/32 bytes (including, 4/8 overhead) |
109 | |
110 | When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte |
111 | ptrs but 4 byte size) or 24 (for 8/8) additional bytes are |
112 | needed; 4 (8) for a trailing size field and 8 (16) bytes for |
113 | free list pointers. Thus, the minimum allocatable size is |
114 | 16/24/32 bytes. |
115 | |
116 | Even a request for zero bytes (i.e., malloc(0)) returns a |
117 | pointer to something of the minimum allocatable size. |
118 | |
119 | The maximum overhead wastage (i.e., number of extra bytes |
120 | allocated than were requested in malloc) is less than or equal |
121 | to the minimum size, except for requests >= mmap_threshold that |
122 | are serviced via mmap(), where the worst case wastage is 2 * |
123 | sizeof(size_t) bytes plus the remainder from a system page (the |
124 | minimal mmap unit); typically 4096 or 8192 bytes. |
125 | |
126 | Maximum allocated size: 4-byte size_t: 2^32 minus about two pages |
127 | 8-byte size_t: 2^64 minus about two pages |
128 | |
129 | It is assumed that (possibly signed) size_t values suffice to |
130 | represent chunk sizes. `Possibly signed' is due to the fact |
131 | that `size_t' may be defined on a system as either a signed or |
132 | an unsigned type. The ISO C standard says that it must be |
133 | unsigned, but a few systems are known not to adhere to this. |
134 | Additionally, even when size_t is unsigned, sbrk (which is by |
135 | default used to obtain memory from system) accepts signed |
136 | arguments, and may not be able to handle size_t-wide arguments |
137 | with negative sign bit. Generally, values that would |
138 | appear as negative after accounting for overhead and alignment |
139 | are supported only via mmap(), which does not have this |
140 | limitation. |
141 | |
142 | Requests for sizes outside the allowed range will perform an optional |
143 | failure action and then return null. (Requests may also |
144 | also fail because a system is out of memory.) |
145 | |
146 | Thread-safety: thread-safe |
147 | |
148 | Compliance: I believe it is compliant with the 1997 Single Unix Specification |
149 | Also SVID/XPG, ANSI C, and probably others as well. |
150 | |
151 | * Synopsis of compile-time options: |
152 | |
153 | People have reported using previous versions of this malloc on all |
154 | versions of Unix, sometimes by tweaking some of the defines |
155 | below. It has been tested most extensively on Solaris and Linux. |
156 | People also report using it in stand-alone embedded systems. |
157 | |
158 | The implementation is in straight, hand-tuned ANSI C. It is not |
159 | at all modular. (Sorry!) It uses a lot of macros. To be at all |
160 | usable, this code should be compiled using an optimizing compiler |
161 | (for example gcc -O3) that can simplify expressions and control |
162 | paths. (FAQ: some macros import variables as arguments rather than |
163 | declare locals because people reported that some debuggers |
164 | otherwise get confused.) |
165 | |
166 | OPTION DEFAULT VALUE |
167 | |
168 | Compilation Environment options: |
169 | |
170 | HAVE_MREMAP 0 |
171 | |
172 | Changing default word sizes: |
173 | |
174 | INTERNAL_SIZE_T size_t |
175 | |
176 | Configuration and functionality options: |
177 | |
178 | USE_PUBLIC_MALLOC_WRAPPERS NOT defined |
179 | USE_MALLOC_LOCK NOT defined |
180 | MALLOC_DEBUG NOT defined |
181 | REALLOC_ZERO_BYTES_FREES 1 |
182 | TRIM_FASTBINS 0 |
183 | |
184 | Options for customizing MORECORE: |
185 | |
186 | MORECORE sbrk |
187 | MORECORE_FAILURE -1 |
188 | MORECORE_CONTIGUOUS 1 |
189 | MORECORE_CANNOT_TRIM NOT defined |
190 | MORECORE_CLEARS 1 |
191 | MMAP_AS_MORECORE_SIZE (1024 * 1024) |
192 | |
193 | Tuning options that are also dynamically changeable via mallopt: |
194 | |
195 | DEFAULT_MXFAST 64 (for 32bit), 128 (for 64bit) |
196 | DEFAULT_TRIM_THRESHOLD 128 * 1024 |
197 | DEFAULT_TOP_PAD 0 |
198 | DEFAULT_MMAP_THRESHOLD 128 * 1024 |
199 | DEFAULT_MMAP_MAX 65536 |
200 | |
201 | There are several other #defined constants and macros that you |
202 | probably don't want to touch unless you are extending or adapting malloc. */ |
203 | |
204 | /* |
205 | void* is the pointer type that malloc should say it returns |
206 | */ |
207 | |
208 | #ifndef void |
209 | #define void void |
210 | #endif /*void*/ |
211 | |
212 | #include <stddef.h> /* for size_t */ |
213 | #include <stdlib.h> /* for getenv(), abort() */ |
214 | #include <unistd.h> /* for __libc_enable_secure */ |
215 | |
216 | #include <atomic.h> |
217 | #include <_itoa.h> |
218 | #include <bits/wordsize.h> |
219 | #include <sys/sysinfo.h> |
220 | |
221 | #include <ldsodefs.h> |
222 | |
223 | #include <unistd.h> |
224 | #include <stdio.h> /* needed for malloc_stats */ |
225 | #include <errno.h> |
226 | |
227 | #include <shlib-compat.h> |
228 | |
229 | /* For uintptr_t. */ |
230 | #include <stdint.h> |
231 | |
232 | /* For va_arg, va_start, va_end. */ |
233 | #include <stdarg.h> |
234 | |
235 | /* For MIN, MAX, powerof2. */ |
236 | #include <sys/param.h> |
237 | |
238 | /* For ALIGN_UP et. al. */ |
239 | #include <libc-pointer-arith.h> |
240 | |
241 | /* For DIAG_PUSH/POP_NEEDS_COMMENT et al. */ |
242 | #include <libc-diag.h> |
243 | |
244 | #include <malloc/malloc-internal.h> |
245 | |
246 | /* For SINGLE_THREAD_P. */ |
247 | #include <sysdep-cancel.h> |
248 | |
249 | /* |
250 | Debugging: |
251 | |
252 | Because freed chunks may be overwritten with bookkeeping fields, this |
253 | malloc will often die when freed memory is overwritten by user |
254 | programs. This can be very effective (albeit in an annoying way) |
255 | in helping track down dangling pointers. |
256 | |
257 | If you compile with -DMALLOC_DEBUG, a number of assertion checks are |
258 | enabled that will catch more memory errors. You probably won't be |
259 | able to make much sense of the actual assertion errors, but they |
260 | should help you locate incorrectly overwritten memory. The checking |
261 | is fairly extensive, and will slow down execution |
262 | noticeably. Calling malloc_stats or mallinfo with MALLOC_DEBUG set |
263 | will attempt to check every non-mmapped allocated and free chunk in |
264 | the course of computing the summmaries. (By nature, mmapped regions |
265 | cannot be checked very much automatically.) |
266 | |
267 | Setting MALLOC_DEBUG may also be helpful if you are trying to modify |
268 | this code. The assertions in the check routines spell out in more |
269 | detail the assumptions and invariants underlying the algorithms. |
270 | |
271 | Setting MALLOC_DEBUG does NOT provide an automated mechanism for |
272 | checking that all accesses to malloced memory stay within their |
273 | bounds. However, there are several add-ons and adaptations of this |
274 | or other mallocs available that do this. |
275 | */ |
276 | |
277 | #ifndef MALLOC_DEBUG |
278 | #define MALLOC_DEBUG 0 |
279 | #endif |
280 | |
281 | #ifdef NDEBUG |
282 | # define assert(expr) ((void) 0) |
283 | #else |
284 | # define assert(expr) \ |
285 | ((expr) \ |
286 | ? ((void) 0) \ |
287 | : __malloc_assert (#expr, __FILE__, __LINE__, __func__)) |
288 | |
289 | extern const char *__progname; |
290 | |
291 | static void |
292 | __malloc_assert (const char *assertion, const char *file, unsigned int line, |
293 | const char *function) |
294 | { |
295 | (void) __fxprintf (NULL, "%s%s%s:%u: %s%sAssertion `%s' failed.\n" , |
296 | __progname, __progname[0] ? ": " : "" , |
297 | file, line, |
298 | function ? function : "" , function ? ": " : "" , |
299 | assertion); |
300 | fflush (stderr); |
301 | abort (); |
302 | } |
303 | #endif |
304 | |
305 | #if USE_TCACHE |
306 | /* We want 64 entries. This is an arbitrary limit, which tunables can reduce. */ |
307 | # define TCACHE_MAX_BINS 64 |
308 | # define MAX_TCACHE_SIZE tidx2usize (TCACHE_MAX_BINS-1) |
309 | |
310 | /* Only used to pre-fill the tunables. */ |
311 | # define tidx2usize(idx) (((size_t) idx) * MALLOC_ALIGNMENT + MINSIZE - SIZE_SZ) |
312 | |
313 | /* When "x" is from chunksize(). */ |
314 | # define csize2tidx(x) (((x) - MINSIZE + MALLOC_ALIGNMENT - 1) / MALLOC_ALIGNMENT) |
315 | /* When "x" is a user-provided size. */ |
316 | # define usize2tidx(x) csize2tidx (request2size (x)) |
317 | |
318 | /* With rounding and alignment, the bins are... |
319 | idx 0 bytes 0..24 (64-bit) or 0..12 (32-bit) |
320 | idx 1 bytes 25..40 or 13..20 |
321 | idx 2 bytes 41..56 or 21..28 |
322 | etc. */ |
323 | |
324 | /* This is another arbitrary limit, which tunables can change. Each |
325 | tcache bin will hold at most this number of chunks. */ |
326 | # define TCACHE_FILL_COUNT 7 |
327 | #endif |
328 | |
329 | |
330 | /* |
331 | REALLOC_ZERO_BYTES_FREES should be set if a call to |
332 | realloc with zero bytes should be the same as a call to free. |
333 | This is required by the C standard. Otherwise, since this malloc |
334 | returns a unique pointer for malloc(0), so does realloc(p, 0). |
335 | */ |
336 | |
337 | #ifndef REALLOC_ZERO_BYTES_FREES |
338 | #define REALLOC_ZERO_BYTES_FREES 1 |
339 | #endif |
340 | |
341 | /* |
342 | TRIM_FASTBINS controls whether free() of a very small chunk can |
343 | immediately lead to trimming. Setting to true (1) can reduce memory |
344 | footprint, but will almost always slow down programs that use a lot |
345 | of small chunks. |
346 | |
347 | Define this only if you are willing to give up some speed to more |
348 | aggressively reduce system-level memory footprint when releasing |
349 | memory in programs that use many small chunks. You can get |
350 | essentially the same effect by setting MXFAST to 0, but this can |
351 | lead to even greater slowdowns in programs using many small chunks. |
352 | TRIM_FASTBINS is an in-between compile-time option, that disables |
353 | only those chunks bordering topmost memory from being placed in |
354 | fastbins. |
355 | */ |
356 | |
357 | #ifndef TRIM_FASTBINS |
358 | #define TRIM_FASTBINS 0 |
359 | #endif |
360 | |
361 | |
362 | /* Definition for getting more memory from the OS. */ |
363 | #define MORECORE (*__morecore) |
364 | #define MORECORE_FAILURE 0 |
365 | void * __default_morecore (ptrdiff_t); |
366 | void *(*__morecore)(ptrdiff_t) = __default_morecore; |
367 | |
368 | |
369 | #include <string.h> |
370 | |
371 | /* |
372 | MORECORE-related declarations. By default, rely on sbrk |
373 | */ |
374 | |
375 | |
376 | /* |
377 | MORECORE is the name of the routine to call to obtain more memory |
378 | from the system. See below for general guidance on writing |
379 | alternative MORECORE functions, as well as a version for WIN32 and a |
380 | sample version for pre-OSX macos. |
381 | */ |
382 | |
383 | #ifndef MORECORE |
384 | #define MORECORE sbrk |
385 | #endif |
386 | |
387 | /* |
388 | MORECORE_FAILURE is the value returned upon failure of MORECORE |
389 | as well as mmap. Since it cannot be an otherwise valid memory address, |
390 | and must reflect values of standard sys calls, you probably ought not |
391 | try to redefine it. |
392 | */ |
393 | |
394 | #ifndef MORECORE_FAILURE |
395 | #define MORECORE_FAILURE (-1) |
396 | #endif |
397 | |
398 | /* |
399 | If MORECORE_CONTIGUOUS is true, take advantage of fact that |
400 | consecutive calls to MORECORE with positive arguments always return |
401 | contiguous increasing addresses. This is true of unix sbrk. Even |
402 | if not defined, when regions happen to be contiguous, malloc will |
403 | permit allocations spanning regions obtained from different |
404 | calls. But defining this when applicable enables some stronger |
405 | consistency checks and space efficiencies. |
406 | */ |
407 | |
408 | #ifndef MORECORE_CONTIGUOUS |
409 | #define MORECORE_CONTIGUOUS 1 |
410 | #endif |
411 | |
412 | /* |
413 | Define MORECORE_CANNOT_TRIM if your version of MORECORE |
414 | cannot release space back to the system when given negative |
415 | arguments. This is generally necessary only if you are using |
416 | a hand-crafted MORECORE function that cannot handle negative arguments. |
417 | */ |
418 | |
419 | /* #define MORECORE_CANNOT_TRIM */ |
420 | |
421 | /* MORECORE_CLEARS (default 1) |
422 | The degree to which the routine mapped to MORECORE zeroes out |
423 | memory: never (0), only for newly allocated space (1) or always |
424 | (2). The distinction between (1) and (2) is necessary because on |
425 | some systems, if the application first decrements and then |
426 | increments the break value, the contents of the reallocated space |
427 | are unspecified. |
428 | */ |
429 | |
430 | #ifndef MORECORE_CLEARS |
431 | # define MORECORE_CLEARS 1 |
432 | #endif |
433 | |
434 | |
435 | /* |
436 | MMAP_AS_MORECORE_SIZE is the minimum mmap size argument to use if |
437 | sbrk fails, and mmap is used as a backup. The value must be a |
438 | multiple of page size. This backup strategy generally applies only |
439 | when systems have "holes" in address space, so sbrk cannot perform |
440 | contiguous expansion, but there is still space available on system. |
441 | On systems for which this is known to be useful (i.e. most linux |
442 | kernels), this occurs only when programs allocate huge amounts of |
443 | memory. Between this, and the fact that mmap regions tend to be |
444 | limited, the size should be large, to avoid too many mmap calls and |
445 | thus avoid running out of kernel resources. */ |
446 | |
447 | #ifndef MMAP_AS_MORECORE_SIZE |
448 | #define MMAP_AS_MORECORE_SIZE (1024 * 1024) |
449 | #endif |
450 | |
451 | /* |
452 | Define HAVE_MREMAP to make realloc() use mremap() to re-allocate |
453 | large blocks. |
454 | */ |
455 | |
456 | #ifndef HAVE_MREMAP |
457 | #define HAVE_MREMAP 0 |
458 | #endif |
459 | |
460 | /* We may need to support __malloc_initialize_hook for backwards |
461 | compatibility. */ |
462 | |
463 | #if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_24) |
464 | # define HAVE_MALLOC_INIT_HOOK 1 |
465 | #else |
466 | # define HAVE_MALLOC_INIT_HOOK 0 |
467 | #endif |
468 | |
469 | |
470 | /* |
471 | This version of malloc supports the standard SVID/XPG mallinfo |
472 | routine that returns a struct containing usage properties and |
473 | statistics. It should work on any SVID/XPG compliant system that has |
474 | a /usr/include/malloc.h defining struct mallinfo. (If you'd like to |
475 | install such a thing yourself, cut out the preliminary declarations |
476 | as described above and below and save them in a malloc.h file. But |
477 | there's no compelling reason to bother to do this.) |
478 | |
479 | The main declaration needed is the mallinfo struct that is returned |
480 | (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a |
481 | bunch of fields that are not even meaningful in this version of |
482 | malloc. These fields are are instead filled by mallinfo() with |
483 | other numbers that might be of interest. |
484 | */ |
485 | |
486 | |
487 | /* ---------- description of public routines ------------ */ |
488 | |
489 | /* |
490 | malloc(size_t n) |
491 | Returns a pointer to a newly allocated chunk of at least n bytes, or null |
492 | if no space is available. Additionally, on failure, errno is |
493 | set to ENOMEM on ANSI C systems. |
494 | |
495 | If n is zero, malloc returns a minumum-sized chunk. (The minimum |
496 | size is 16 bytes on most 32bit systems, and 24 or 32 bytes on 64bit |
497 | systems.) On most systems, size_t is an unsigned type, so calls |
498 | with negative arguments are interpreted as requests for huge amounts |
499 | of space, which will often fail. The maximum supported value of n |
500 | differs across systems, but is in all cases less than the maximum |
501 | representable value of a size_t. |
502 | */ |
503 | void* __libc_malloc(size_t); |
504 | libc_hidden_proto (__libc_malloc) |
505 | |
506 | /* |
507 | free(void* p) |
508 | Releases the chunk of memory pointed to by p, that had been previously |
509 | allocated using malloc or a related routine such as realloc. |
510 | It has no effect if p is null. It can have arbitrary (i.e., bad!) |
511 | effects if p has already been freed. |
512 | |
513 | Unless disabled (using mallopt), freeing very large spaces will |
514 | when possible, automatically trigger operations that give |
515 | back unused memory to the system, thus reducing program footprint. |
516 | */ |
517 | void __libc_free(void*); |
518 | libc_hidden_proto (__libc_free) |
519 | |
520 | /* |
521 | calloc(size_t n_elements, size_t element_size); |
522 | Returns a pointer to n_elements * element_size bytes, with all locations |
523 | set to zero. |
524 | */ |
525 | void* __libc_calloc(size_t, size_t); |
526 | |
527 | /* |
528 | realloc(void* p, size_t n) |
529 | Returns a pointer to a chunk of size n that contains the same data |
530 | as does chunk p up to the minimum of (n, p's size) bytes, or null |
531 | if no space is available. |
532 | |
533 | The returned pointer may or may not be the same as p. The algorithm |
534 | prefers extending p when possible, otherwise it employs the |
535 | equivalent of a malloc-copy-free sequence. |
536 | |
537 | If p is null, realloc is equivalent to malloc. |
538 | |
539 | If space is not available, realloc returns null, errno is set (if on |
540 | ANSI) and p is NOT freed. |
541 | |
542 | if n is for fewer bytes than already held by p, the newly unused |
543 | space is lopped off and freed if possible. Unless the #define |
544 | REALLOC_ZERO_BYTES_FREES is set, realloc with a size argument of |
545 | zero (re)allocates a minimum-sized chunk. |
546 | |
547 | Large chunks that were internally obtained via mmap will always be |
548 | grown using malloc-copy-free sequences unless the system supports |
549 | MREMAP (currently only linux). |
550 | |
551 | The old unix realloc convention of allowing the last-free'd chunk |
552 | to be used as an argument to realloc is not supported. |
553 | */ |
554 | void* __libc_realloc(void*, size_t); |
555 | libc_hidden_proto (__libc_realloc) |
556 | |
557 | /* |
558 | memalign(size_t alignment, size_t n); |
559 | Returns a pointer to a newly allocated chunk of n bytes, aligned |
560 | in accord with the alignment argument. |
561 | |
562 | The alignment argument should be a power of two. If the argument is |
563 | not a power of two, the nearest greater power is used. |
564 | 8-byte alignment is guaranteed by normal malloc calls, so don't |
565 | bother calling memalign with an argument of 8 or less. |
566 | |
567 | Overreliance on memalign is a sure way to fragment space. |
568 | */ |
569 | void* __libc_memalign(size_t, size_t); |
570 | libc_hidden_proto (__libc_memalign) |
571 | |
572 | /* |
573 | valloc(size_t n); |
574 | Equivalent to memalign(pagesize, n), where pagesize is the page |
575 | size of the system. If the pagesize is unknown, 4096 is used. |
576 | */ |
577 | void* __libc_valloc(size_t); |
578 | |
579 | |
580 | |
581 | /* |
582 | mallopt(int parameter_number, int parameter_value) |
583 | Sets tunable parameters The format is to provide a |
584 | (parameter-number, parameter-value) pair. mallopt then sets the |
585 | corresponding parameter to the argument value if it can (i.e., so |
586 | long as the value is meaningful), and returns 1 if successful else |
587 | 0. SVID/XPG/ANSI defines four standard param numbers for mallopt, |
588 | normally defined in malloc.h. Only one of these (M_MXFAST) is used |
589 | in this malloc. The others (M_NLBLKS, M_GRAIN, M_KEEP) don't apply, |
590 | so setting them has no effect. But this malloc also supports four |
591 | other options in mallopt. See below for details. Briefly, supported |
592 | parameters are as follows (listed defaults are for "typical" |
593 | configurations). |
594 | |
595 | Symbol param # default allowed param values |
596 | M_MXFAST 1 64 0-80 (0 disables fastbins) |
597 | M_TRIM_THRESHOLD -1 128*1024 any (-1U disables trimming) |
598 | M_TOP_PAD -2 0 any |
599 | M_MMAP_THRESHOLD -3 128*1024 any (or 0 if no MMAP support) |
600 | M_MMAP_MAX -4 65536 any (0 disables use of mmap) |
601 | */ |
602 | int __libc_mallopt(int, int); |
603 | libc_hidden_proto (__libc_mallopt) |
604 | |
605 | |
606 | /* |
607 | mallinfo() |
608 | Returns (by copy) a struct containing various summary statistics: |
609 | |
610 | arena: current total non-mmapped bytes allocated from system |
611 | ordblks: the number of free chunks |
612 | smblks: the number of fastbin blocks (i.e., small chunks that |
613 | have been freed but not use resused or consolidated) |
614 | hblks: current number of mmapped regions |
615 | hblkhd: total bytes held in mmapped regions |
616 | usmblks: always 0 |
617 | fsmblks: total bytes held in fastbin blocks |
618 | uordblks: current total allocated space (normal or mmapped) |
619 | fordblks: total free space |
620 | keepcost: the maximum number of bytes that could ideally be released |
621 | back to system via malloc_trim. ("ideally" means that |
622 | it ignores page restrictions etc.) |
623 | |
624 | Because these fields are ints, but internal bookkeeping may |
625 | be kept as longs, the reported values may wrap around zero and |
626 | thus be inaccurate. |
627 | */ |
628 | struct mallinfo __libc_mallinfo(void); |
629 | |
630 | |
631 | /* |
632 | pvalloc(size_t n); |
633 | Equivalent to valloc(minimum-page-that-holds(n)), that is, |
634 | round up n to nearest pagesize. |
635 | */ |
636 | void* __libc_pvalloc(size_t); |
637 | |
638 | /* |
639 | malloc_trim(size_t pad); |
640 | |
641 | If possible, gives memory back to the system (via negative |
642 | arguments to sbrk) if there is unused memory at the `high' end of |
643 | the malloc pool. You can call this after freeing large blocks of |
644 | memory to potentially reduce the system-level memory requirements |
645 | of a program. However, it cannot guarantee to reduce memory. Under |
646 | some allocation patterns, some large free blocks of memory will be |
647 | locked between two used chunks, so they cannot be given back to |
648 | the system. |
649 | |
650 | The `pad' argument to malloc_trim represents the amount of free |
651 | trailing space to leave untrimmed. If this argument is zero, |
652 | only the minimum amount of memory to maintain internal data |
653 | structures will be left (one page or less). Non-zero arguments |
654 | can be supplied to maintain enough trailing space to service |
655 | future expected allocations without having to re-obtain memory |
656 | from the system. |
657 | |
658 | Malloc_trim returns 1 if it actually released any memory, else 0. |
659 | On systems that do not support "negative sbrks", it will always |
660 | return 0. |
661 | */ |
662 | int __malloc_trim(size_t); |
663 | |
664 | /* |
665 | malloc_usable_size(void* p); |
666 | |
667 | Returns the number of bytes you can actually use in |
668 | an allocated chunk, which may be more than you requested (although |
669 | often not) due to alignment and minimum size constraints. |
670 | You can use this many bytes without worrying about |
671 | overwriting other allocated objects. This is not a particularly great |
672 | programming practice. malloc_usable_size can be more useful in |
673 | debugging and assertions, for example: |
674 | |
675 | p = malloc(n); |
676 | assert(malloc_usable_size(p) >= 256); |
677 | |
678 | */ |
679 | size_t __malloc_usable_size(void*); |
680 | |
681 | /* |
682 | malloc_stats(); |
683 | Prints on stderr the amount of space obtained from the system (both |
684 | via sbrk and mmap), the maximum amount (which may be more than |
685 | current if malloc_trim and/or munmap got called), and the current |
686 | number of bytes allocated via malloc (or realloc, etc) but not yet |
687 | freed. Note that this is the number of bytes allocated, not the |
688 | number requested. It will be larger than the number requested |
689 | because of alignment and bookkeeping overhead. Because it includes |
690 | alignment wastage as being in use, this figure may be greater than |
691 | zero even when no user-level chunks are allocated. |
692 | |
693 | The reported current and maximum system memory can be inaccurate if |
694 | a program makes other calls to system memory allocation functions |
695 | (normally sbrk) outside of malloc. |
696 | |
697 | malloc_stats prints only the most commonly interesting statistics. |
698 | More information can be obtained by calling mallinfo. |
699 | |
700 | */ |
701 | void __malloc_stats(void); |
702 | |
703 | /* |
704 | malloc_get_state(void); |
705 | |
706 | Returns the state of all malloc variables in an opaque data |
707 | structure. |
708 | */ |
709 | void* __malloc_get_state(void); |
710 | |
711 | /* |
712 | malloc_set_state(void* state); |
713 | |
714 | Restore the state of all malloc variables from data obtained with |
715 | malloc_get_state(). |
716 | */ |
717 | int __malloc_set_state(void*); |
718 | |
719 | /* |
720 | posix_memalign(void **memptr, size_t alignment, size_t size); |
721 | |
722 | POSIX wrapper like memalign(), checking for validity of size. |
723 | */ |
724 | int __posix_memalign(void **, size_t, size_t); |
725 | |
726 | /* mallopt tuning options */ |
727 | |
728 | /* |
729 | M_MXFAST is the maximum request size used for "fastbins", special bins |
730 | that hold returned chunks without consolidating their spaces. This |
731 | enables future requests for chunks of the same size to be handled |
732 | very quickly, but can increase fragmentation, and thus increase the |
733 | overall memory footprint of a program. |
734 | |
735 | This malloc manages fastbins very conservatively yet still |
736 | efficiently, so fragmentation is rarely a problem for values less |
737 | than or equal to the default. The maximum supported value of MXFAST |
738 | is 80. You wouldn't want it any higher than this anyway. Fastbins |
739 | are designed especially for use with many small structs, objects or |
740 | strings -- the default handles structs/objects/arrays with sizes up |
741 | to 8 4byte fields, or small strings representing words, tokens, |
742 | etc. Using fastbins for larger objects normally worsens |
743 | fragmentation without improving speed. |
744 | |
745 | M_MXFAST is set in REQUEST size units. It is internally used in |
746 | chunksize units, which adds padding and alignment. You can reduce |
747 | M_MXFAST to 0 to disable all use of fastbins. This causes the malloc |
748 | algorithm to be a closer approximation of fifo-best-fit in all cases, |
749 | not just for larger requests, but will generally cause it to be |
750 | slower. |
751 | */ |
752 | |
753 | |
754 | /* M_MXFAST is a standard SVID/XPG tuning option, usually listed in malloc.h */ |
755 | #ifndef M_MXFAST |
756 | #define M_MXFAST 1 |
757 | #endif |
758 | |
759 | #ifndef DEFAULT_MXFAST |
760 | #define DEFAULT_MXFAST (64 * SIZE_SZ / 4) |
761 | #endif |
762 | |
763 | |
764 | /* |
765 | M_TRIM_THRESHOLD is the maximum amount of unused top-most memory |
766 | to keep before releasing via malloc_trim in free(). |
767 | |
768 | Automatic trimming is mainly useful in long-lived programs. |
769 | Because trimming via sbrk can be slow on some systems, and can |
770 | sometimes be wasteful (in cases where programs immediately |
771 | afterward allocate more large chunks) the value should be high |
772 | enough so that your overall system performance would improve by |
773 | releasing this much memory. |
774 | |
775 | The trim threshold and the mmap control parameters (see below) |
776 | can be traded off with one another. Trimming and mmapping are |
777 | two different ways of releasing unused memory back to the |
778 | system. Between these two, it is often possible to keep |
779 | system-level demands of a long-lived program down to a bare |
780 | minimum. For example, in one test suite of sessions measuring |
781 | the XF86 X server on Linux, using a trim threshold of 128K and a |
782 | mmap threshold of 192K led to near-minimal long term resource |
783 | consumption. |
784 | |
785 | If you are using this malloc in a long-lived program, it should |
786 | pay to experiment with these values. As a rough guide, you |
787 | might set to a value close to the average size of a process |
788 | (program) running on your system. Releasing this much memory |
789 | would allow such a process to run in memory. Generally, it's |
790 | worth it to tune for trimming rather tham memory mapping when a |
791 | program undergoes phases where several large chunks are |
792 | allocated and released in ways that can reuse each other's |
793 | storage, perhaps mixed with phases where there are no such |
794 | chunks at all. And in well-behaved long-lived programs, |
795 | controlling release of large blocks via trimming versus mapping |
796 | is usually faster. |
797 | |
798 | However, in most programs, these parameters serve mainly as |
799 | protection against the system-level effects of carrying around |
800 | massive amounts of unneeded memory. Since frequent calls to |
801 | sbrk, mmap, and munmap otherwise degrade performance, the default |
802 | parameters are set to relatively high values that serve only as |
803 | safeguards. |
804 | |
805 | The trim value It must be greater than page size to have any useful |
806 | effect. To disable trimming completely, you can set to |
807 | (unsigned long)(-1) |
808 | |
809 | Trim settings interact with fastbin (MXFAST) settings: Unless |
810 | TRIM_FASTBINS is defined, automatic trimming never takes place upon |
811 | freeing a chunk with size less than or equal to MXFAST. Trimming is |
812 | instead delayed until subsequent freeing of larger chunks. However, |
813 | you can still force an attempted trim by calling malloc_trim. |
814 | |
815 | Also, trimming is not generally possible in cases where |
816 | the main arena is obtained via mmap. |
817 | |
818 | Note that the trick some people use of mallocing a huge space and |
819 | then freeing it at program startup, in an attempt to reserve system |
820 | memory, doesn't have the intended effect under automatic trimming, |
821 | since that memory will immediately be returned to the system. |
822 | */ |
823 | |
824 | #define M_TRIM_THRESHOLD -1 |
825 | |
826 | #ifndef DEFAULT_TRIM_THRESHOLD |
827 | #define DEFAULT_TRIM_THRESHOLD (128 * 1024) |
828 | #endif |
829 | |
830 | /* |
831 | M_TOP_PAD is the amount of extra `padding' space to allocate or |
832 | retain whenever sbrk is called. It is used in two ways internally: |
833 | |
834 | * When sbrk is called to extend the top of the arena to satisfy |
835 | a new malloc request, this much padding is added to the sbrk |
836 | request. |
837 | |
838 | * When malloc_trim is called automatically from free(), |
839 | it is used as the `pad' argument. |
840 | |
841 | In both cases, the actual amount of padding is rounded |
842 | so that the end of the arena is always a system page boundary. |
843 | |
844 | The main reason for using padding is to avoid calling sbrk so |
845 | often. Having even a small pad greatly reduces the likelihood |
846 | that nearly every malloc request during program start-up (or |
847 | after trimming) will invoke sbrk, which needlessly wastes |
848 | time. |
849 | |
850 | Automatic rounding-up to page-size units is normally sufficient |
851 | to avoid measurable overhead, so the default is 0. However, in |
852 | systems where sbrk is relatively slow, it can pay to increase |
853 | this value, at the expense of carrying around more memory than |
854 | the program needs. |
855 | */ |
856 | |
857 | #define M_TOP_PAD -2 |
858 | |
859 | #ifndef DEFAULT_TOP_PAD |
860 | #define DEFAULT_TOP_PAD (0) |
861 | #endif |
862 | |
863 | /* |
864 | MMAP_THRESHOLD_MAX and _MIN are the bounds on the dynamically |
865 | adjusted MMAP_THRESHOLD. |
866 | */ |
867 | |
868 | #ifndef DEFAULT_MMAP_THRESHOLD_MIN |
869 | #define DEFAULT_MMAP_THRESHOLD_MIN (128 * 1024) |
870 | #endif |
871 | |
872 | #ifndef DEFAULT_MMAP_THRESHOLD_MAX |
873 | /* For 32-bit platforms we cannot increase the maximum mmap |
874 | threshold much because it is also the minimum value for the |
875 | maximum heap size and its alignment. Going above 512k (i.e., 1M |
876 | for new heaps) wastes too much address space. */ |
877 | # if __WORDSIZE == 32 |
878 | # define DEFAULT_MMAP_THRESHOLD_MAX (512 * 1024) |
879 | # else |
880 | # define DEFAULT_MMAP_THRESHOLD_MAX (4 * 1024 * 1024 * sizeof(long)) |
881 | # endif |
882 | #endif |
883 | |
884 | /* |
885 | M_MMAP_THRESHOLD is the request size threshold for using mmap() |
886 | to service a request. Requests of at least this size that cannot |
887 | be allocated using already-existing space will be serviced via mmap. |
888 | (If enough normal freed space already exists it is used instead.) |
889 | |
890 | Using mmap segregates relatively large chunks of memory so that |
891 | they can be individually obtained and released from the host |
892 | system. A request serviced through mmap is never reused by any |
893 | other request (at least not directly; the system may just so |
894 | happen to remap successive requests to the same locations). |
895 | |
896 | Segregating space in this way has the benefits that: |
897 | |
898 | 1. Mmapped space can ALWAYS be individually released back |
899 | to the system, which helps keep the system level memory |
900 | demands of a long-lived program low. |
901 | 2. Mapped memory can never become `locked' between |
902 | other chunks, as can happen with normally allocated chunks, which |
903 | means that even trimming via malloc_trim would not release them. |
904 | 3. On some systems with "holes" in address spaces, mmap can obtain |
905 | memory that sbrk cannot. |
906 | |
907 | However, it has the disadvantages that: |
908 | |
909 | 1. The space cannot be reclaimed, consolidated, and then |
910 | used to service later requests, as happens with normal chunks. |
911 | 2. It can lead to more wastage because of mmap page alignment |
912 | requirements |
913 | 3. It causes malloc performance to be more dependent on host |
914 | system memory management support routines which may vary in |
915 | implementation quality and may impose arbitrary |
916 | limitations. Generally, servicing a request via normal |
917 | malloc steps is faster than going through a system's mmap. |
918 | |
919 | The advantages of mmap nearly always outweigh disadvantages for |
920 | "large" chunks, but the value of "large" varies across systems. The |
921 | default is an empirically derived value that works well in most |
922 | systems. |
923 | |
924 | |
925 | Update in 2006: |
926 | The above was written in 2001. Since then the world has changed a lot. |
927 | Memory got bigger. Applications got bigger. The virtual address space |
928 | layout in 32 bit linux changed. |
929 | |
930 | In the new situation, brk() and mmap space is shared and there are no |
931 | artificial limits on brk size imposed by the kernel. What is more, |
932 | applications have started using transient allocations larger than the |
933 | 128Kb as was imagined in 2001. |
934 | |
935 | The price for mmap is also high now; each time glibc mmaps from the |
936 | kernel, the kernel is forced to zero out the memory it gives to the |
937 | application. Zeroing memory is expensive and eats a lot of cache and |
938 | memory bandwidth. This has nothing to do with the efficiency of the |
939 | virtual memory system, by doing mmap the kernel just has no choice but |
940 | to zero. |
941 | |
942 | In 2001, the kernel had a maximum size for brk() which was about 800 |
943 | megabytes on 32 bit x86, at that point brk() would hit the first |
944 | mmaped shared libaries and couldn't expand anymore. With current 2.6 |
945 | kernels, the VA space layout is different and brk() and mmap |
946 | both can span the entire heap at will. |
947 | |
948 | Rather than using a static threshold for the brk/mmap tradeoff, |
949 | we are now using a simple dynamic one. The goal is still to avoid |
950 | fragmentation. The old goals we kept are |
951 | 1) try to get the long lived large allocations to use mmap() |
952 | 2) really large allocations should always use mmap() |
953 | and we're adding now: |
954 | 3) transient allocations should use brk() to avoid forcing the kernel |
955 | having to zero memory over and over again |
956 | |
957 | The implementation works with a sliding threshold, which is by default |
958 | limited to go between 128Kb and 32Mb (64Mb for 64 bitmachines) and starts |
959 | out at 128Kb as per the 2001 default. |
960 | |
961 | This allows us to satisfy requirement 1) under the assumption that long |
962 | lived allocations are made early in the process' lifespan, before it has |
963 | started doing dynamic allocations of the same size (which will |
964 | increase the threshold). |
965 | |
966 | The upperbound on the threshold satisfies requirement 2) |
967 | |
968 | The threshold goes up in value when the application frees memory that was |
969 | allocated with the mmap allocator. The idea is that once the application |
970 | starts freeing memory of a certain size, it's highly probable that this is |
971 | a size the application uses for transient allocations. This estimator |
972 | is there to satisfy the new third requirement. |
973 | |
974 | */ |
975 | |
976 | #define M_MMAP_THRESHOLD -3 |
977 | |
978 | #ifndef DEFAULT_MMAP_THRESHOLD |
979 | #define DEFAULT_MMAP_THRESHOLD DEFAULT_MMAP_THRESHOLD_MIN |
980 | #endif |
981 | |
982 | /* |
983 | M_MMAP_MAX is the maximum number of requests to simultaneously |
984 | service using mmap. This parameter exists because |
985 | some systems have a limited number of internal tables for |
986 | use by mmap, and using more than a few of them may degrade |
987 | performance. |
988 | |
989 | The default is set to a value that serves only as a safeguard. |
990 | Setting to 0 disables use of mmap for servicing large requests. |
991 | */ |
992 | |
993 | #define M_MMAP_MAX -4 |
994 | |
995 | #ifndef DEFAULT_MMAP_MAX |
996 | #define DEFAULT_MMAP_MAX (65536) |
997 | #endif |
998 | |
999 | #include <malloc.h> |
1000 | |
1001 | #ifndef RETURN_ADDRESS |
1002 | #define RETURN_ADDRESS(X_) (NULL) |
1003 | #endif |
1004 | |
1005 | /* On some platforms we can compile internal, not exported functions better. |
1006 | Let the environment provide a macro and define it to be empty if it |
1007 | is not available. */ |
1008 | #ifndef internal_function |
1009 | # define internal_function |
1010 | #endif |
1011 | |
1012 | /* Forward declarations. */ |
1013 | struct malloc_chunk; |
1014 | typedef struct malloc_chunk* mchunkptr; |
1015 | |
1016 | /* Internal routines. */ |
1017 | |
1018 | static void* _int_malloc(mstate, size_t); |
1019 | static void _int_free(mstate, mchunkptr, int); |
1020 | static void* _int_realloc(mstate, mchunkptr, INTERNAL_SIZE_T, |
1021 | INTERNAL_SIZE_T); |
1022 | static void* _int_memalign(mstate, size_t, size_t); |
1023 | static void* _mid_memalign(size_t, size_t, void *); |
1024 | |
1025 | static void malloc_printerr(const char *str) __attribute__ ((noreturn)); |
1026 | |
1027 | static void* internal_function mem2mem_check(void *p, size_t sz); |
1028 | static void top_check (void); |
1029 | static void internal_function munmap_chunk(mchunkptr p); |
1030 | #if HAVE_MREMAP |
1031 | static mchunkptr internal_function mremap_chunk(mchunkptr p, size_t new_size); |
1032 | #endif |
1033 | |
1034 | static void* malloc_check(size_t sz, const void *caller); |
1035 | static void free_check(void* mem, const void *caller); |
1036 | static void* realloc_check(void* oldmem, size_t bytes, |
1037 | const void *caller); |
1038 | static void* memalign_check(size_t alignment, size_t bytes, |
1039 | const void *caller); |
1040 | |
1041 | /* ------------------ MMAP support ------------------ */ |
1042 | |
1043 | |
1044 | #include <fcntl.h> |
1045 | #include <sys/mman.h> |
1046 | |
1047 | #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON) |
1048 | # define MAP_ANONYMOUS MAP_ANON |
1049 | #endif |
1050 | |
1051 | #ifndef MAP_NORESERVE |
1052 | # define MAP_NORESERVE 0 |
1053 | #endif |
1054 | |
1055 | #define MMAP(addr, size, prot, flags) \ |
1056 | __mmap((addr), (size), (prot), (flags)|MAP_ANONYMOUS|MAP_PRIVATE, -1, 0) |
1057 | |
1058 | |
1059 | /* |
1060 | ----------------------- Chunk representations ----------------------- |
1061 | */ |
1062 | |
1063 | |
1064 | /* |
1065 | This struct declaration is misleading (but accurate and necessary). |
1066 | It declares a "view" into memory allowing access to necessary |
1067 | fields at known offsets from a given base. See explanation below. |
1068 | */ |
1069 | |
1070 | struct malloc_chunk { |
1071 | |
1072 | INTERNAL_SIZE_T mchunk_prev_size; /* Size of previous chunk (if free). */ |
1073 | INTERNAL_SIZE_T mchunk_size; /* Size in bytes, including overhead. */ |
1074 | |
1075 | struct malloc_chunk* fd; /* double links -- used only if free. */ |
1076 | struct malloc_chunk* bk; |
1077 | |
1078 | /* Only used for large blocks: pointer to next larger size. */ |
1079 | struct malloc_chunk* fd_nextsize; /* double links -- used only if free. */ |
1080 | struct malloc_chunk* bk_nextsize; |
1081 | }; |
1082 | |
1083 | |
1084 | /* |
1085 | malloc_chunk details: |
1086 | |
1087 | (The following includes lightly edited explanations by Colin Plumb.) |
1088 | |
1089 | Chunks of memory are maintained using a `boundary tag' method as |
1090 | described in e.g., Knuth or Standish. (See the paper by Paul |
1091 | Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a |
1092 | survey of such techniques.) Sizes of free chunks are stored both |
1093 | in the front of each chunk and at the end. This makes |
1094 | consolidating fragmented chunks into bigger chunks very fast. The |
1095 | size fields also hold bits representing whether chunks are free or |
1096 | in use. |
1097 | |
1098 | An allocated chunk looks like this: |
1099 | |
1100 | |
1101 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1102 | | Size of previous chunk, if unallocated (P clear) | |
1103 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1104 | | Size of chunk, in bytes |A|M|P| |
1105 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1106 | | User data starts here... . |
1107 | . . |
1108 | . (malloc_usable_size() bytes) . |
1109 | . | |
1110 | nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1111 | | (size of chunk, but used for application data) | |
1112 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1113 | | Size of next chunk, in bytes |A|0|1| |
1114 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1115 | |
1116 | Where "chunk" is the front of the chunk for the purpose of most of |
1117 | the malloc code, but "mem" is the pointer that is returned to the |
1118 | user. "Nextchunk" is the beginning of the next contiguous chunk. |
1119 | |
1120 | Chunks always begin on even word boundaries, so the mem portion |
1121 | (which is returned to the user) is also on an even word boundary, and |
1122 | thus at least double-word aligned. |
1123 | |
1124 | Free chunks are stored in circular doubly-linked lists, and look like this: |
1125 | |
1126 | chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1127 | | Size of previous chunk, if unallocated (P clear) | |
1128 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1129 | `head:' | Size of chunk, in bytes |A|0|P| |
1130 | mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1131 | | Forward pointer to next chunk in list | |
1132 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1133 | | Back pointer to previous chunk in list | |
1134 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1135 | | Unused space (may be 0 bytes long) . |
1136 | . . |
1137 | . | |
1138 | nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1139 | `foot:' | Size of chunk, in bytes | |
1140 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1141 | | Size of next chunk, in bytes |A|0|0| |
1142 | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |
1143 | |
1144 | The P (PREV_INUSE) bit, stored in the unused low-order bit of the |
1145 | chunk size (which is always a multiple of two words), is an in-use |
1146 | bit for the *previous* chunk. If that bit is *clear*, then the |
1147 | word before the current chunk size contains the previous chunk |
1148 | size, and can be used to find the front of the previous chunk. |
1149 | The very first chunk allocated always has this bit set, |
1150 | preventing access to non-existent (or non-owned) memory. If |
1151 | prev_inuse is set for any given chunk, then you CANNOT determine |
1152 | the size of the previous chunk, and might even get a memory |
1153 | addressing fault when trying to do so. |
1154 | |
1155 | The A (NON_MAIN_ARENA) bit is cleared for chunks on the initial, |
1156 | main arena, described by the main_arena variable. When additional |
1157 | threads are spawned, each thread receives its own arena (up to a |
1158 | configurable limit, after which arenas are reused for multiple |
1159 | threads), and the chunks in these arenas have the A bit set. To |
1160 | find the arena for a chunk on such a non-main arena, heap_for_ptr |
1161 | performs a bit mask operation and indirection through the ar_ptr |
1162 | member of the per-heap header heap_info (see arena.c). |
1163 | |
1164 | Note that the `foot' of the current chunk is actually represented |
1165 | as the prev_size of the NEXT chunk. This makes it easier to |
1166 | deal with alignments etc but can be very confusing when trying |
1167 | to extend or adapt this code. |
1168 | |
1169 | The three exceptions to all this are: |
1170 | |
1171 | 1. The special chunk `top' doesn't bother using the |
1172 | trailing size field since there is no next contiguous chunk |
1173 | that would have to index off it. After initialization, `top' |
1174 | is forced to always exist. If it would become less than |
1175 | MINSIZE bytes long, it is replenished. |
1176 | |
1177 | 2. Chunks allocated via mmap, which have the second-lowest-order |
1178 | bit M (IS_MMAPPED) set in their size fields. Because they are |
1179 | allocated one-by-one, each must contain its own trailing size |
1180 | field. If the M bit is set, the other bits are ignored |
1181 | (because mmapped chunks are neither in an arena, nor adjacent |
1182 | to a freed chunk). The M bit is also used for chunks which |
1183 | originally came from a dumped heap via malloc_set_state in |
1184 | hooks.c. |
1185 | |
1186 | 3. Chunks in fastbins are treated as allocated chunks from the |
1187 | point of view of the chunk allocator. They are consolidated |
1188 | with their neighbors only in bulk, in malloc_consolidate. |
1189 | */ |
1190 | |
1191 | /* |
1192 | ---------- Size and alignment checks and conversions ---------- |
1193 | */ |
1194 | |
1195 | /* conversion from malloc headers to user pointers, and back */ |
1196 | |
1197 | #define chunk2mem(p) ((void*)((char*)(p) + 2*SIZE_SZ)) |
1198 | #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ)) |
1199 | |
1200 | /* The smallest possible chunk */ |
1201 | #define MIN_CHUNK_SIZE (offsetof(struct malloc_chunk, fd_nextsize)) |
1202 | |
1203 | /* The smallest size we can malloc is an aligned minimal chunk */ |
1204 | |
1205 | #define MINSIZE \ |
1206 | (unsigned long)(((MIN_CHUNK_SIZE+MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK)) |
1207 | |
1208 | /* Check if m has acceptable alignment */ |
1209 | |
1210 | #define aligned_OK(m) (((unsigned long)(m) & MALLOC_ALIGN_MASK) == 0) |
1211 | |
1212 | #define misaligned_chunk(p) \ |
1213 | ((uintptr_t)(MALLOC_ALIGNMENT == 2 * SIZE_SZ ? (p) : chunk2mem (p)) \ |
1214 | & MALLOC_ALIGN_MASK) |
1215 | |
1216 | |
1217 | /* |
1218 | Check if a request is so large that it would wrap around zero when |
1219 | padded and aligned. To simplify some other code, the bound is made |
1220 | low enough so that adding MINSIZE will also not wrap around zero. |
1221 | */ |
1222 | |
1223 | #define REQUEST_OUT_OF_RANGE(req) \ |
1224 | ((unsigned long) (req) >= \ |
1225 | (unsigned long) (INTERNAL_SIZE_T) (-2 * MINSIZE)) |
1226 | |
1227 | /* pad request bytes into a usable size -- internal version */ |
1228 | |
1229 | #define request2size(req) \ |
1230 | (((req) + SIZE_SZ + MALLOC_ALIGN_MASK < MINSIZE) ? \ |
1231 | MINSIZE : \ |
1232 | ((req) + SIZE_SZ + MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK) |
1233 | |
1234 | /* Same, except also perform an argument and result check. First, we check |
1235 | that the padding done by request2size didn't result in an integer |
1236 | overflow. Then we check (using REQUEST_OUT_OF_RANGE) that the resulting |
1237 | size isn't so large that a later alignment would lead to another integer |
1238 | overflow. */ |
1239 | #define checked_request2size(req, sz) \ |
1240 | ({ \ |
1241 | (sz) = request2size (req); \ |
1242 | if (((sz) < (req)) \ |
1243 | || REQUEST_OUT_OF_RANGE (sz)) \ |
1244 | { \ |
1245 | __set_errno (ENOMEM); \ |
1246 | return 0; \ |
1247 | } \ |
1248 | }) |
1249 | |
1250 | /* |
1251 | --------------- Physical chunk operations --------------- |
1252 | */ |
1253 | |
1254 | |
1255 | /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */ |
1256 | #define PREV_INUSE 0x1 |
1257 | |
1258 | /* extract inuse bit of previous chunk */ |
1259 | #define prev_inuse(p) ((p)->mchunk_size & PREV_INUSE) |
1260 | |
1261 | |
1262 | /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */ |
1263 | #define IS_MMAPPED 0x2 |
1264 | |
1265 | /* check for mmap()'ed chunk */ |
1266 | #define chunk_is_mmapped(p) ((p)->mchunk_size & IS_MMAPPED) |
1267 | |
1268 | |
1269 | /* size field is or'ed with NON_MAIN_ARENA if the chunk was obtained |
1270 | from a non-main arena. This is only set immediately before handing |
1271 | the chunk to the user, if necessary. */ |
1272 | #define NON_MAIN_ARENA 0x4 |
1273 | |
1274 | /* Check for chunk from main arena. */ |
1275 | #define chunk_main_arena(p) (((p)->mchunk_size & NON_MAIN_ARENA) == 0) |
1276 | |
1277 | /* Mark a chunk as not being on the main arena. */ |
1278 | #define set_non_main_arena(p) ((p)->mchunk_size |= NON_MAIN_ARENA) |
1279 | |
1280 | |
1281 | /* |
1282 | Bits to mask off when extracting size |
1283 | |
1284 | Note: IS_MMAPPED is intentionally not masked off from size field in |
1285 | macros for which mmapped chunks should never be seen. This should |
1286 | cause helpful core dumps to occur if it is tried by accident by |
1287 | people extending or adapting this malloc. |
1288 | */ |
1289 | #define SIZE_BITS (PREV_INUSE | IS_MMAPPED | NON_MAIN_ARENA) |
1290 | |
1291 | /* Get size, ignoring use bits */ |
1292 | #define chunksize(p) (chunksize_nomask (p) & ~(SIZE_BITS)) |
1293 | |
1294 | /* Like chunksize, but do not mask SIZE_BITS. */ |
1295 | #define chunksize_nomask(p) ((p)->mchunk_size) |
1296 | |
1297 | /* Ptr to next physical malloc_chunk. */ |
1298 | #define next_chunk(p) ((mchunkptr) (((char *) (p)) + chunksize (p))) |
1299 | |
1300 | /* Size of the chunk below P. Only valid if prev_inuse (P). */ |
1301 | #define prev_size(p) ((p)->mchunk_prev_size) |
1302 | |
1303 | /* Set the size of the chunk below P. Only valid if prev_inuse (P). */ |
1304 | #define set_prev_size(p, sz) ((p)->mchunk_prev_size = (sz)) |
1305 | |
1306 | /* Ptr to previous physical malloc_chunk. Only valid if prev_inuse (P). */ |
1307 | #define prev_chunk(p) ((mchunkptr) (((char *) (p)) - prev_size (p))) |
1308 | |
1309 | /* Treat space at ptr + offset as a chunk */ |
1310 | #define chunk_at_offset(p, s) ((mchunkptr) (((char *) (p)) + (s))) |
1311 | |
1312 | /* extract p's inuse bit */ |
1313 | #define inuse(p) \ |
1314 | ((((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size) & PREV_INUSE) |
1315 | |
1316 | /* set/clear chunk as being inuse without otherwise disturbing */ |
1317 | #define set_inuse(p) \ |
1318 | ((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size |= PREV_INUSE |
1319 | |
1320 | #define clear_inuse(p) \ |
1321 | ((mchunkptr) (((char *) (p)) + chunksize (p)))->mchunk_size &= ~(PREV_INUSE) |
1322 | |
1323 | |
1324 | /* check/set/clear inuse bits in known places */ |
1325 | #define inuse_bit_at_offset(p, s) \ |
1326 | (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size & PREV_INUSE) |
1327 | |
1328 | #define set_inuse_bit_at_offset(p, s) \ |
1329 | (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size |= PREV_INUSE) |
1330 | |
1331 | #define clear_inuse_bit_at_offset(p, s) \ |
1332 | (((mchunkptr) (((char *) (p)) + (s)))->mchunk_size &= ~(PREV_INUSE)) |
1333 | |
1334 | |
1335 | /* Set size at head, without disturbing its use bit */ |
1336 | #define set_head_size(p, s) ((p)->mchunk_size = (((p)->mchunk_size & SIZE_BITS) | (s))) |
1337 | |
1338 | /* Set size/use field */ |
1339 | #define set_head(p, s) ((p)->mchunk_size = (s)) |
1340 | |
1341 | /* Set size at footer (only when chunk is not in use) */ |
1342 | #define (p, s) (((mchunkptr) ((char *) (p) + (s)))->mchunk_prev_size = (s)) |
1343 | |
1344 | |
1345 | #pragma GCC poison mchunk_size |
1346 | #pragma GCC poison mchunk_prev_size |
1347 | |
1348 | /* |
1349 | -------------------- Internal data structures -------------------- |
1350 | |
1351 | All internal state is held in an instance of malloc_state defined |
1352 | below. There are no other static variables, except in two optional |
1353 | cases: |
1354 | * If USE_MALLOC_LOCK is defined, the mALLOC_MUTEx declared above. |
1355 | * If mmap doesn't support MAP_ANONYMOUS, a dummy file descriptor |
1356 | for mmap. |
1357 | |
1358 | Beware of lots of tricks that minimize the total bookkeeping space |
1359 | requirements. The result is a little over 1K bytes (for 4byte |
1360 | pointers and size_t.) |
1361 | */ |
1362 | |
1363 | /* |
1364 | Bins |
1365 | |
1366 | An array of bin headers for free chunks. Each bin is doubly |
1367 | linked. The bins are approximately proportionally (log) spaced. |
1368 | There are a lot of these bins (128). This may look excessive, but |
1369 | works very well in practice. Most bins hold sizes that are |
1370 | unusual as malloc request sizes, but are more usual for fragments |
1371 | and consolidated sets of chunks, which is what these bins hold, so |
1372 | they can be found quickly. All procedures maintain the invariant |
1373 | that no consolidated chunk physically borders another one, so each |
1374 | chunk in a list is known to be preceeded and followed by either |
1375 | inuse chunks or the ends of memory. |
1376 | |
1377 | Chunks in bins are kept in size order, with ties going to the |
1378 | approximately least recently used chunk. Ordering isn't needed |
1379 | for the small bins, which all contain the same-sized chunks, but |
1380 | facilitates best-fit allocation for larger chunks. These lists |
1381 | are just sequential. Keeping them in order almost never requires |
1382 | enough traversal to warrant using fancier ordered data |
1383 | structures. |
1384 | |
1385 | Chunks of the same size are linked with the most |
1386 | recently freed at the front, and allocations are taken from the |
1387 | back. This results in LRU (FIFO) allocation order, which tends |
1388 | to give each chunk an equal opportunity to be consolidated with |
1389 | adjacent freed chunks, resulting in larger free chunks and less |
1390 | fragmentation. |
1391 | |
1392 | To simplify use in double-linked lists, each bin header acts |
1393 | as a malloc_chunk. This avoids special-casing for headers. |
1394 | But to conserve space and improve locality, we allocate |
1395 | only the fd/bk pointers of bins, and then use repositioning tricks |
1396 | to treat these as the fields of a malloc_chunk*. |
1397 | */ |
1398 | |
1399 | typedef struct malloc_chunk *mbinptr; |
1400 | |
1401 | /* addressing -- note that bin_at(0) does not exist */ |
1402 | #define bin_at(m, i) \ |
1403 | (mbinptr) (((char *) &((m)->bins[((i) - 1) * 2])) \ |
1404 | - offsetof (struct malloc_chunk, fd)) |
1405 | |
1406 | /* analog of ++bin */ |
1407 | #define next_bin(b) ((mbinptr) ((char *) (b) + (sizeof (mchunkptr) << 1))) |
1408 | |
1409 | /* Reminders about list directionality within bins */ |
1410 | #define first(b) ((b)->fd) |
1411 | #define last(b) ((b)->bk) |
1412 | |
1413 | /* Take a chunk off a bin list */ |
1414 | #define unlink(AV, P, BK, FD) { \ |
1415 | if (__builtin_expect (chunksize(P) != prev_size (next_chunk(P)), 0)) \ |
1416 | malloc_printerr ("corrupted size vs. prev_size"); \ |
1417 | FD = P->fd; \ |
1418 | BK = P->bk; \ |
1419 | if (__builtin_expect (FD->bk != P || BK->fd != P, 0)) \ |
1420 | malloc_printerr ("corrupted double-linked list"); \ |
1421 | else { \ |
1422 | FD->bk = BK; \ |
1423 | BK->fd = FD; \ |
1424 | if (!in_smallbin_range (chunksize_nomask (P)) \ |
1425 | && __builtin_expect (P->fd_nextsize != NULL, 0)) { \ |
1426 | if (__builtin_expect (P->fd_nextsize->bk_nextsize != P, 0) \ |
1427 | || __builtin_expect (P->bk_nextsize->fd_nextsize != P, 0)) \ |
1428 | malloc_printerr ("corrupted double-linked list (not small)"); \ |
1429 | if (FD->fd_nextsize == NULL) { \ |
1430 | if (P->fd_nextsize == P) \ |
1431 | FD->fd_nextsize = FD->bk_nextsize = FD; \ |
1432 | else { \ |
1433 | FD->fd_nextsize = P->fd_nextsize; \ |
1434 | FD->bk_nextsize = P->bk_nextsize; \ |
1435 | P->fd_nextsize->bk_nextsize = FD; \ |
1436 | P->bk_nextsize->fd_nextsize = FD; \ |
1437 | } \ |
1438 | } else { \ |
1439 | P->fd_nextsize->bk_nextsize = P->bk_nextsize; \ |
1440 | P->bk_nextsize->fd_nextsize = P->fd_nextsize; \ |
1441 | } \ |
1442 | } \ |
1443 | } \ |
1444 | } |
1445 | |
1446 | /* |
1447 | Indexing |
1448 | |
1449 | Bins for sizes < 512 bytes contain chunks of all the same size, spaced |
1450 | 8 bytes apart. Larger bins are approximately logarithmically spaced: |
1451 | |
1452 | 64 bins of size 8 |
1453 | 32 bins of size 64 |
1454 | 16 bins of size 512 |
1455 | 8 bins of size 4096 |
1456 | 4 bins of size 32768 |
1457 | 2 bins of size 262144 |
1458 | 1 bin of size what's left |
1459 | |
1460 | There is actually a little bit of slop in the numbers in bin_index |
1461 | for the sake of speed. This makes no difference elsewhere. |
1462 | |
1463 | The bins top out around 1MB because we expect to service large |
1464 | requests via mmap. |
1465 | |
1466 | Bin 0 does not exist. Bin 1 is the unordered list; if that would be |
1467 | a valid chunk size the small bins are bumped up one. |
1468 | */ |
1469 | |
1470 | #define NBINS 128 |
1471 | #define NSMALLBINS 64 |
1472 | #define SMALLBIN_WIDTH MALLOC_ALIGNMENT |
1473 | #define SMALLBIN_CORRECTION (MALLOC_ALIGNMENT > 2 * SIZE_SZ) |
1474 | #define MIN_LARGE_SIZE ((NSMALLBINS - SMALLBIN_CORRECTION) * SMALLBIN_WIDTH) |
1475 | |
1476 | #define in_smallbin_range(sz) \ |
1477 | ((unsigned long) (sz) < (unsigned long) MIN_LARGE_SIZE) |
1478 | |
1479 | #define smallbin_index(sz) \ |
1480 | ((SMALLBIN_WIDTH == 16 ? (((unsigned) (sz)) >> 4) : (((unsigned) (sz)) >> 3))\ |
1481 | + SMALLBIN_CORRECTION) |
1482 | |
1483 | #define largebin_index_32(sz) \ |
1484 | (((((unsigned long) (sz)) >> 6) <= 38) ? 56 + (((unsigned long) (sz)) >> 6) :\ |
1485 | ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\ |
1486 | ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\ |
1487 | ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\ |
1488 | ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\ |
1489 | 126) |
1490 | |
1491 | #define largebin_index_32_big(sz) \ |
1492 | (((((unsigned long) (sz)) >> 6) <= 45) ? 49 + (((unsigned long) (sz)) >> 6) :\ |
1493 | ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\ |
1494 | ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\ |
1495 | ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\ |
1496 | ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\ |
1497 | 126) |
1498 | |
1499 | // XXX It remains to be seen whether it is good to keep the widths of |
1500 | // XXX the buckets the same or whether it should be scaled by a factor |
1501 | // XXX of two as well. |
1502 | #define largebin_index_64(sz) \ |
1503 | (((((unsigned long) (sz)) >> 6) <= 48) ? 48 + (((unsigned long) (sz)) >> 6) :\ |
1504 | ((((unsigned long) (sz)) >> 9) <= 20) ? 91 + (((unsigned long) (sz)) >> 9) :\ |
1505 | ((((unsigned long) (sz)) >> 12) <= 10) ? 110 + (((unsigned long) (sz)) >> 12) :\ |
1506 | ((((unsigned long) (sz)) >> 15) <= 4) ? 119 + (((unsigned long) (sz)) >> 15) :\ |
1507 | ((((unsigned long) (sz)) >> 18) <= 2) ? 124 + (((unsigned long) (sz)) >> 18) :\ |
1508 | 126) |
1509 | |
1510 | #define largebin_index(sz) \ |
1511 | (SIZE_SZ == 8 ? largebin_index_64 (sz) \ |
1512 | : MALLOC_ALIGNMENT == 16 ? largebin_index_32_big (sz) \ |
1513 | : largebin_index_32 (sz)) |
1514 | |
1515 | #define bin_index(sz) \ |
1516 | ((in_smallbin_range (sz)) ? smallbin_index (sz) : largebin_index (sz)) |
1517 | |
1518 | |
1519 | /* |
1520 | Unsorted chunks |
1521 | |
1522 | All remainders from chunk splits, as well as all returned chunks, |
1523 | are first placed in the "unsorted" bin. They are then placed |
1524 | in regular bins after malloc gives them ONE chance to be used before |
1525 | binning. So, basically, the unsorted_chunks list acts as a queue, |
1526 | with chunks being placed on it in free (and malloc_consolidate), |
1527 | and taken off (to be either used or placed in bins) in malloc. |
1528 | |
1529 | The NON_MAIN_ARENA flag is never set for unsorted chunks, so it |
1530 | does not have to be taken into account in size comparisons. |
1531 | */ |
1532 | |
1533 | /* The otherwise unindexable 1-bin is used to hold unsorted chunks. */ |
1534 | #define unsorted_chunks(M) (bin_at (M, 1)) |
1535 | |
1536 | /* |
1537 | Top |
1538 | |
1539 | The top-most available chunk (i.e., the one bordering the end of |
1540 | available memory) is treated specially. It is never included in |
1541 | any bin, is used only if no other chunk is available, and is |
1542 | released back to the system if it is very large (see |
1543 | M_TRIM_THRESHOLD). Because top initially |
1544 | points to its own bin with initial zero size, thus forcing |
1545 | extension on the first malloc request, we avoid having any special |
1546 | code in malloc to check whether it even exists yet. But we still |
1547 | need to do so when getting memory from system, so we make |
1548 | initial_top treat the bin as a legal but unusable chunk during the |
1549 | interval between initialization and the first call to |
1550 | sysmalloc. (This is somewhat delicate, since it relies on |
1551 | the 2 preceding words to be zero during this interval as well.) |
1552 | */ |
1553 | |
1554 | /* Conveniently, the unsorted bin can be used as dummy top on first call */ |
1555 | #define initial_top(M) (unsorted_chunks (M)) |
1556 | |
1557 | /* |
1558 | Binmap |
1559 | |
1560 | To help compensate for the large number of bins, a one-level index |
1561 | structure is used for bin-by-bin searching. `binmap' is a |
1562 | bitvector recording whether bins are definitely empty so they can |
1563 | be skipped over during during traversals. The bits are NOT always |
1564 | cleared as soon as bins are empty, but instead only |
1565 | when they are noticed to be empty during traversal in malloc. |
1566 | */ |
1567 | |
1568 | /* Conservatively use 32 bits per map word, even if on 64bit system */ |
1569 | #define BINMAPSHIFT 5 |
1570 | #define BITSPERMAP (1U << BINMAPSHIFT) |
1571 | #define BINMAPSIZE (NBINS / BITSPERMAP) |
1572 | |
1573 | #define idx2block(i) ((i) >> BINMAPSHIFT) |
1574 | #define idx2bit(i) ((1U << ((i) & ((1U << BINMAPSHIFT) - 1)))) |
1575 | |
1576 | #define mark_bin(m, i) ((m)->binmap[idx2block (i)] |= idx2bit (i)) |
1577 | #define unmark_bin(m, i) ((m)->binmap[idx2block (i)] &= ~(idx2bit (i))) |
1578 | #define get_binmap(m, i) ((m)->binmap[idx2block (i)] & idx2bit (i)) |
1579 | |
1580 | /* |
1581 | Fastbins |
1582 | |
1583 | An array of lists holding recently freed small chunks. Fastbins |
1584 | are not doubly linked. It is faster to single-link them, and |
1585 | since chunks are never removed from the middles of these lists, |
1586 | double linking is not necessary. Also, unlike regular bins, they |
1587 | are not even processed in FIFO order (they use faster LIFO) since |
1588 | ordering doesn't much matter in the transient contexts in which |
1589 | fastbins are normally used. |
1590 | |
1591 | Chunks in fastbins keep their inuse bit set, so they cannot |
1592 | be consolidated with other free chunks. malloc_consolidate |
1593 | releases all chunks in fastbins and consolidates them with |
1594 | other free chunks. |
1595 | */ |
1596 | |
1597 | typedef struct malloc_chunk *mfastbinptr; |
1598 | #define fastbin(ar_ptr, idx) ((ar_ptr)->fastbinsY[idx]) |
1599 | |
1600 | /* offset 2 to use otherwise unindexable first 2 bins */ |
1601 | #define fastbin_index(sz) \ |
1602 | ((((unsigned int) (sz)) >> (SIZE_SZ == 8 ? 4 : 3)) - 2) |
1603 | |
1604 | |
1605 | /* The maximum fastbin request size we support */ |
1606 | #define MAX_FAST_SIZE (80 * SIZE_SZ / 4) |
1607 | |
1608 | #define NFASTBINS (fastbin_index (request2size (MAX_FAST_SIZE)) + 1) |
1609 | |
1610 | /* |
1611 | FASTBIN_CONSOLIDATION_THRESHOLD is the size of a chunk in free() |
1612 | that triggers automatic consolidation of possibly-surrounding |
1613 | fastbin chunks. This is a heuristic, so the exact value should not |
1614 | matter too much. It is defined at half the default trim threshold as a |
1615 | compromise heuristic to only attempt consolidation if it is likely |
1616 | to lead to trimming. However, it is not dynamically tunable, since |
1617 | consolidation reduces fragmentation surrounding large chunks even |
1618 | if trimming is not used. |
1619 | */ |
1620 | |
1621 | #define FASTBIN_CONSOLIDATION_THRESHOLD (65536UL) |
1622 | |
1623 | /* |
1624 | NONCONTIGUOUS_BIT indicates that MORECORE does not return contiguous |
1625 | regions. Otherwise, contiguity is exploited in merging together, |
1626 | when possible, results from consecutive MORECORE calls. |
1627 | |
1628 | The initial value comes from MORECORE_CONTIGUOUS, but is |
1629 | changed dynamically if mmap is ever used as an sbrk substitute. |
1630 | */ |
1631 | |
1632 | #define NONCONTIGUOUS_BIT (2U) |
1633 | |
1634 | #define contiguous(M) (((M)->flags & NONCONTIGUOUS_BIT) == 0) |
1635 | #define noncontiguous(M) (((M)->flags & NONCONTIGUOUS_BIT) != 0) |
1636 | #define set_noncontiguous(M) ((M)->flags |= NONCONTIGUOUS_BIT) |
1637 | #define set_contiguous(M) ((M)->flags &= ~NONCONTIGUOUS_BIT) |
1638 | |
1639 | /* Maximum size of memory handled in fastbins. */ |
1640 | static INTERNAL_SIZE_T global_max_fast; |
1641 | |
1642 | /* |
1643 | Set value of max_fast. |
1644 | Use impossibly small value if 0. |
1645 | Precondition: there are no existing fastbin chunks. |
1646 | Setting the value clears fastchunk bit but preserves noncontiguous bit. |
1647 | */ |
1648 | |
1649 | #define set_max_fast(s) \ |
1650 | global_max_fast = (((s) == 0) \ |
1651 | ? SMALLBIN_WIDTH : ((s + SIZE_SZ) & ~MALLOC_ALIGN_MASK)) |
1652 | |
1653 | static inline INTERNAL_SIZE_T |
1654 | get_max_fast (void) |
1655 | { |
1656 | /* Tell the GCC optimizers that global_max_fast is never larger |
1657 | than MAX_FAST_SIZE. This avoids out-of-bounds array accesses in |
1658 | _int_malloc after constant propagation of the size parameter. |
1659 | (The code never executes because malloc preserves the |
1660 | global_max_fast invariant, but the optimizers may not recognize |
1661 | this.) */ |
1662 | if (global_max_fast > MAX_FAST_SIZE) |
1663 | __builtin_unreachable (); |
1664 | return global_max_fast; |
1665 | } |
1666 | |
1667 | /* |
1668 | ----------- Internal state representation and initialization ----------- |
1669 | */ |
1670 | |
1671 | /* |
1672 | have_fastchunks indicates that there are probably some fastbin chunks. |
1673 | It is set true on entering a chunk into any fastbin, and cleared early in |
1674 | malloc_consolidate. The value is approximate since it may be set when there |
1675 | are no fastbin chunks, or it may be clear even if there are fastbin chunks |
1676 | available. Given it's sole purpose is to reduce number of redundant calls to |
1677 | malloc_consolidate, it does not affect correctness. As a result we can safely |
1678 | use relaxed atomic accesses. |
1679 | */ |
1680 | |
1681 | |
1682 | struct malloc_state |
1683 | { |
1684 | /* Serialize access. */ |
1685 | __libc_lock_define (, mutex); |
1686 | |
1687 | /* Flags (formerly in max_fast). */ |
1688 | int flags; |
1689 | |
1690 | /* Set if the fastbin chunks contain recently inserted free blocks. */ |
1691 | /* Note this is a bool but not all targets support atomics on booleans. */ |
1692 | int have_fastchunks; |
1693 | |
1694 | /* Fastbins */ |
1695 | mfastbinptr fastbinsY[NFASTBINS]; |
1696 | |
1697 | /* Base of the topmost chunk -- not otherwise kept in a bin */ |
1698 | mchunkptr top; |
1699 | |
1700 | /* The remainder from the most recent split of a small request */ |
1701 | mchunkptr last_remainder; |
1702 | |
1703 | /* Normal bins packed as described above */ |
1704 | mchunkptr bins[NBINS * 2 - 2]; |
1705 | |
1706 | /* Bitmap of bins */ |
1707 | unsigned int binmap[BINMAPSIZE]; |
1708 | |
1709 | /* Linked list */ |
1710 | struct malloc_state *next; |
1711 | |
1712 | /* Linked list for free arenas. Access to this field is serialized |
1713 | by free_list_lock in arena.c. */ |
1714 | struct malloc_state *next_free; |
1715 | |
1716 | /* Number of threads attached to this arena. 0 if the arena is on |
1717 | the free list. Access to this field is serialized by |
1718 | free_list_lock in arena.c. */ |
1719 | INTERNAL_SIZE_T attached_threads; |
1720 | |
1721 | /* Memory allocated from the system in this arena. */ |
1722 | INTERNAL_SIZE_T system_mem; |
1723 | INTERNAL_SIZE_T max_system_mem; |
1724 | }; |
1725 | |
1726 | struct malloc_par |
1727 | { |
1728 | /* Tunable parameters */ |
1729 | unsigned long trim_threshold; |
1730 | INTERNAL_SIZE_T top_pad; |
1731 | INTERNAL_SIZE_T mmap_threshold; |
1732 | INTERNAL_SIZE_T arena_test; |
1733 | INTERNAL_SIZE_T arena_max; |
1734 | |
1735 | /* Memory map support */ |
1736 | int n_mmaps; |
1737 | int n_mmaps_max; |
1738 | int max_n_mmaps; |
1739 | /* the mmap_threshold is dynamic, until the user sets |
1740 | it manually, at which point we need to disable any |
1741 | dynamic behavior. */ |
1742 | int no_dyn_threshold; |
1743 | |
1744 | /* Statistics */ |
1745 | INTERNAL_SIZE_T mmapped_mem; |
1746 | INTERNAL_SIZE_T max_mmapped_mem; |
1747 | |
1748 | /* First address handed out by MORECORE/sbrk. */ |
1749 | char *sbrk_base; |
1750 | |
1751 | #if USE_TCACHE |
1752 | /* Maximum number of buckets to use. */ |
1753 | size_t tcache_bins; |
1754 | size_t tcache_max_bytes; |
1755 | /* Maximum number of chunks in each bucket. */ |
1756 | size_t tcache_count; |
1757 | /* Maximum number of chunks to remove from the unsorted list, which |
1758 | aren't used to prefill the cache. */ |
1759 | size_t tcache_unsorted_limit; |
1760 | #endif |
1761 | }; |
1762 | |
1763 | /* There are several instances of this struct ("arenas") in this |
1764 | malloc. If you are adapting this malloc in a way that does NOT use |
1765 | a static or mmapped malloc_state, you MUST explicitly zero-fill it |
1766 | before using. This malloc relies on the property that malloc_state |
1767 | is initialized to all zeroes (as is true of C statics). */ |
1768 | |
1769 | static struct malloc_state main_arena = |
1770 | { |
1771 | .mutex = _LIBC_LOCK_INITIALIZER, |
1772 | .next = &main_arena, |
1773 | .attached_threads = 1 |
1774 | }; |
1775 | |
1776 | /* These variables are used for undumping support. Chunked are marked |
1777 | as using mmap, but we leave them alone if they fall into this |
1778 | range. NB: The chunk size for these chunks only includes the |
1779 | initial size field (of SIZE_SZ bytes), there is no trailing size |
1780 | field (unlike with regular mmapped chunks). */ |
1781 | static mchunkptr dumped_main_arena_start; /* Inclusive. */ |
1782 | static mchunkptr dumped_main_arena_end; /* Exclusive. */ |
1783 | |
1784 | /* True if the pointer falls into the dumped arena. Use this after |
1785 | chunk_is_mmapped indicates a chunk is mmapped. */ |
1786 | #define DUMPED_MAIN_ARENA_CHUNK(p) \ |
1787 | ((p) >= dumped_main_arena_start && (p) < dumped_main_arena_end) |
1788 | |
1789 | /* There is only one instance of the malloc parameters. */ |
1790 | |
1791 | static struct malloc_par mp_ = |
1792 | { |
1793 | .top_pad = DEFAULT_TOP_PAD, |
1794 | .n_mmaps_max = DEFAULT_MMAP_MAX, |
1795 | .mmap_threshold = DEFAULT_MMAP_THRESHOLD, |
1796 | .trim_threshold = DEFAULT_TRIM_THRESHOLD, |
1797 | #define NARENAS_FROM_NCORES(n) ((n) * (sizeof (long) == 4 ? 2 : 8)) |
1798 | .arena_test = NARENAS_FROM_NCORES (1) |
1799 | #if USE_TCACHE |
1800 | , |
1801 | .tcache_count = TCACHE_FILL_COUNT, |
1802 | .tcache_bins = TCACHE_MAX_BINS, |
1803 | .tcache_max_bytes = tidx2usize (TCACHE_MAX_BINS-1), |
1804 | .tcache_unsorted_limit = 0 /* No limit. */ |
1805 | #endif |
1806 | }; |
1807 | |
1808 | /* |
1809 | Initialize a malloc_state struct. |
1810 | |
1811 | This is called only from within malloc_consolidate, which needs |
1812 | be called in the same contexts anyway. It is never called directly |
1813 | outside of malloc_consolidate because some optimizing compilers try |
1814 | to inline it at all call points, which turns out not to be an |
1815 | optimization at all. (Inlining it in malloc_consolidate is fine though.) |
1816 | */ |
1817 | |
1818 | static void |
1819 | malloc_init_state (mstate av) |
1820 | { |
1821 | int i; |
1822 | mbinptr bin; |
1823 | |
1824 | /* Establish circular links for normal bins */ |
1825 | for (i = 1; i < NBINS; ++i) |
1826 | { |
1827 | bin = bin_at (av, i); |
1828 | bin->fd = bin->bk = bin; |
1829 | } |
1830 | |
1831 | #if MORECORE_CONTIGUOUS |
1832 | if (av != &main_arena) |
1833 | #endif |
1834 | set_noncontiguous (av); |
1835 | if (av == &main_arena) |
1836 | set_max_fast (DEFAULT_MXFAST); |
1837 | atomic_store_relaxed (&av->have_fastchunks, false); |
1838 | |
1839 | av->top = initial_top (av); |
1840 | } |
1841 | |
1842 | /* |
1843 | Other internal utilities operating on mstates |
1844 | */ |
1845 | |
1846 | static void *sysmalloc (INTERNAL_SIZE_T, mstate); |
1847 | static int systrim (size_t, mstate); |
1848 | static void malloc_consolidate (mstate); |
1849 | |
1850 | |
1851 | /* -------------- Early definitions for debugging hooks ---------------- */ |
1852 | |
1853 | /* Define and initialize the hook variables. These weak definitions must |
1854 | appear before any use of the variables in a function (arena.c uses one). */ |
1855 | #ifndef weak_variable |
1856 | /* In GNU libc we want the hook variables to be weak definitions to |
1857 | avoid a problem with Emacs. */ |
1858 | # define weak_variable weak_function |
1859 | #endif |
1860 | |
1861 | /* Forward declarations. */ |
1862 | static void *malloc_hook_ini (size_t sz, |
1863 | const void *caller) __THROW; |
1864 | static void *realloc_hook_ini (void *ptr, size_t sz, |
1865 | const void *caller) __THROW; |
1866 | static void *memalign_hook_ini (size_t alignment, size_t sz, |
1867 | const void *caller) __THROW; |
1868 | |
1869 | #if HAVE_MALLOC_INIT_HOOK |
1870 | void weak_variable (*__malloc_initialize_hook) (void) = NULL; |
1871 | compat_symbol (libc, __malloc_initialize_hook, |
1872 | __malloc_initialize_hook, GLIBC_2_0); |
1873 | #endif |
1874 | |
1875 | void weak_variable (*__free_hook) (void *__ptr, |
1876 | const void *) = NULL; |
1877 | void *weak_variable (*__malloc_hook) |
1878 | (size_t __size, const void *) = malloc_hook_ini; |
1879 | void *weak_variable (*__realloc_hook) |
1880 | (void *__ptr, size_t __size, const void *) |
1881 | = realloc_hook_ini; |
1882 | void *weak_variable (*__memalign_hook) |
1883 | (size_t __alignment, size_t __size, const void *) |
1884 | = memalign_hook_ini; |
1885 | void weak_variable (*__after_morecore_hook) (void) = NULL; |
1886 | |
1887 | |
1888 | /* ------------------ Testing support ----------------------------------*/ |
1889 | |
1890 | static int perturb_byte; |
1891 | |
1892 | static void |
1893 | alloc_perturb (char *p, size_t n) |
1894 | { |
1895 | if (__glibc_unlikely (perturb_byte)) |
1896 | memset (p, perturb_byte ^ 0xff, n); |
1897 | } |
1898 | |
1899 | static void |
1900 | free_perturb (char *p, size_t n) |
1901 | { |
1902 | if (__glibc_unlikely (perturb_byte)) |
1903 | memset (p, perturb_byte, n); |
1904 | } |
1905 | |
1906 | |
1907 | |
1908 | #include <stap-probe.h> |
1909 | |
1910 | /* ------------------- Support for multiple arenas -------------------- */ |
1911 | #include "arena.c" |
1912 | |
1913 | /* |
1914 | Debugging support |
1915 | |
1916 | These routines make a number of assertions about the states |
1917 | of data structures that should be true at all times. If any |
1918 | are not true, it's very likely that a user program has somehow |
1919 | trashed memory. (It's also possible that there is a coding error |
1920 | in malloc. In which case, please report it!) |
1921 | */ |
1922 | |
1923 | #if !MALLOC_DEBUG |
1924 | |
1925 | # define check_chunk(A, P) |
1926 | # define check_free_chunk(A, P) |
1927 | # define check_inuse_chunk(A, P) |
1928 | # define check_remalloced_chunk(A, P, N) |
1929 | # define check_malloced_chunk(A, P, N) |
1930 | # define check_malloc_state(A) |
1931 | |
1932 | #else |
1933 | |
1934 | # define check_chunk(A, P) do_check_chunk (A, P) |
1935 | # define check_free_chunk(A, P) do_check_free_chunk (A, P) |
1936 | # define check_inuse_chunk(A, P) do_check_inuse_chunk (A, P) |
1937 | # define check_remalloced_chunk(A, P, N) do_check_remalloced_chunk (A, P, N) |
1938 | # define check_malloced_chunk(A, P, N) do_check_malloced_chunk (A, P, N) |
1939 | # define check_malloc_state(A) do_check_malloc_state (A) |
1940 | |
1941 | /* |
1942 | Properties of all chunks |
1943 | */ |
1944 | |
1945 | static void |
1946 | do_check_chunk (mstate av, mchunkptr p) |
1947 | { |
1948 | unsigned long sz = chunksize (p); |
1949 | /* min and max possible addresses assuming contiguous allocation */ |
1950 | char *max_address = (char *) (av->top) + chunksize (av->top); |
1951 | char *min_address = max_address - av->system_mem; |
1952 | |
1953 | if (!chunk_is_mmapped (p)) |
1954 | { |
1955 | /* Has legal address ... */ |
1956 | if (p != av->top) |
1957 | { |
1958 | if (contiguous (av)) |
1959 | { |
1960 | assert (((char *) p) >= min_address); |
1961 | assert (((char *) p + sz) <= ((char *) (av->top))); |
1962 | } |
1963 | } |
1964 | else |
1965 | { |
1966 | /* top size is always at least MINSIZE */ |
1967 | assert ((unsigned long) (sz) >= MINSIZE); |
1968 | /* top predecessor always marked inuse */ |
1969 | assert (prev_inuse (p)); |
1970 | } |
1971 | } |
1972 | else if (!DUMPED_MAIN_ARENA_CHUNK (p)) |
1973 | { |
1974 | /* address is outside main heap */ |
1975 | if (contiguous (av) && av->top != initial_top (av)) |
1976 | { |
1977 | assert (((char *) p) < min_address || ((char *) p) >= max_address); |
1978 | } |
1979 | /* chunk is page-aligned */ |
1980 | assert (((prev_size (p) + sz) & (GLRO (dl_pagesize) - 1)) == 0); |
1981 | /* mem is aligned */ |
1982 | assert (aligned_OK (chunk2mem (p))); |
1983 | } |
1984 | } |
1985 | |
1986 | /* |
1987 | Properties of free chunks |
1988 | */ |
1989 | |
1990 | static void |
1991 | do_check_free_chunk (mstate av, mchunkptr p) |
1992 | { |
1993 | INTERNAL_SIZE_T sz = p->size & ~(PREV_INUSE | NON_MAIN_ARENA); |
1994 | mchunkptr next = chunk_at_offset (p, sz); |
1995 | |
1996 | do_check_chunk (av, p); |
1997 | |
1998 | /* Chunk must claim to be free ... */ |
1999 | assert (!inuse (p)); |
2000 | assert (!chunk_is_mmapped (p)); |
2001 | |
2002 | /* Unless a special marker, must have OK fields */ |
2003 | if ((unsigned long) (sz) >= MINSIZE) |
2004 | { |
2005 | assert ((sz & MALLOC_ALIGN_MASK) == 0); |
2006 | assert (aligned_OK (chunk2mem (p))); |
2007 | /* ... matching footer field */ |
2008 | assert (prev_size (p) == sz); |
2009 | /* ... and is fully consolidated */ |
2010 | assert (prev_inuse (p)); |
2011 | assert (next == av->top || inuse (next)); |
2012 | |
2013 | /* ... and has minimally sane links */ |
2014 | assert (p->fd->bk == p); |
2015 | assert (p->bk->fd == p); |
2016 | } |
2017 | else /* markers are always of size SIZE_SZ */ |
2018 | assert (sz == SIZE_SZ); |
2019 | } |
2020 | |
2021 | /* |
2022 | Properties of inuse chunks |
2023 | */ |
2024 | |
2025 | static void |
2026 | do_check_inuse_chunk (mstate av, mchunkptr p) |
2027 | { |
2028 | mchunkptr next; |
2029 | |
2030 | do_check_chunk (av, p); |
2031 | |
2032 | if (chunk_is_mmapped (p)) |
2033 | return; /* mmapped chunks have no next/prev */ |
2034 | |
2035 | /* Check whether it claims to be in use ... */ |
2036 | assert (inuse (p)); |
2037 | |
2038 | next = next_chunk (p); |
2039 | |
2040 | /* ... and is surrounded by OK chunks. |
2041 | Since more things can be checked with free chunks than inuse ones, |
2042 | if an inuse chunk borders them and debug is on, it's worth doing them. |
2043 | */ |
2044 | if (!prev_inuse (p)) |
2045 | { |
2046 | /* Note that we cannot even look at prev unless it is not inuse */ |
2047 | mchunkptr prv = prev_chunk (p); |
2048 | assert (next_chunk (prv) == p); |
2049 | do_check_free_chunk (av, prv); |
2050 | } |
2051 | |
2052 | if (next == av->top) |
2053 | { |
2054 | assert (prev_inuse (next)); |
2055 | assert (chunksize (next) >= MINSIZE); |
2056 | } |
2057 | else if (!inuse (next)) |
2058 | do_check_free_chunk (av, next); |
2059 | } |
2060 | |
2061 | /* |
2062 | Properties of chunks recycled from fastbins |
2063 | */ |
2064 | |
2065 | static void |
2066 | do_check_remalloced_chunk (mstate av, mchunkptr p, INTERNAL_SIZE_T s) |
2067 | { |
2068 | INTERNAL_SIZE_T sz = p->size & ~(PREV_INUSE | NON_MAIN_ARENA); |
2069 | |
2070 | if (!chunk_is_mmapped (p)) |
2071 | { |
2072 | assert (av == arena_for_chunk (p)); |
2073 | if (chunk_main_arena (p)) |
2074 | assert (av == &main_arena); |
2075 | else |
2076 | assert (av != &main_arena); |
2077 | } |
2078 | |
2079 | do_check_inuse_chunk (av, p); |
2080 | |
2081 | /* Legal size ... */ |
2082 | assert ((sz & MALLOC_ALIGN_MASK) == 0); |
2083 | assert ((unsigned long) (sz) >= MINSIZE); |
2084 | /* ... and alignment */ |
2085 | assert (aligned_OK (chunk2mem (p))); |
2086 | /* chunk is less than MINSIZE more than request */ |
2087 | assert ((long) (sz) - (long) (s) >= 0); |
2088 | assert ((long) (sz) - (long) (s + MINSIZE) < 0); |
2089 | } |
2090 | |
2091 | /* |
2092 | Properties of nonrecycled chunks at the point they are malloced |
2093 | */ |
2094 | |
2095 | static void |
2096 | do_check_malloced_chunk (mstate av, mchunkptr p, INTERNAL_SIZE_T s) |
2097 | { |
2098 | /* same as recycled case ... */ |
2099 | do_check_remalloced_chunk (av, p, s); |
2100 | |
2101 | /* |
2102 | ... plus, must obey implementation invariant that prev_inuse is |
2103 | always true of any allocated chunk; i.e., that each allocated |
2104 | chunk borders either a previously allocated and still in-use |
2105 | chunk, or the base of its memory arena. This is ensured |
2106 | by making all allocations from the `lowest' part of any found |
2107 | chunk. This does not necessarily hold however for chunks |
2108 | recycled via fastbins. |
2109 | */ |
2110 | |
2111 | assert (prev_inuse (p)); |
2112 | } |
2113 | |
2114 | |
2115 | /* |
2116 | Properties of malloc_state. |
2117 | |
2118 | This may be useful for debugging malloc, as well as detecting user |
2119 | programmer errors that somehow write into malloc_state. |
2120 | |
2121 | If you are extending or experimenting with this malloc, you can |
2122 | probably figure out how to hack this routine to print out or |
2123 | display chunk addresses, sizes, bins, and other instrumentation. |
2124 | */ |
2125 | |
2126 | static void |
2127 | do_check_malloc_state (mstate av) |
2128 | { |
2129 | int i; |
2130 | mchunkptr p; |
2131 | mchunkptr q; |
2132 | mbinptr b; |
2133 | unsigned int idx; |
2134 | INTERNAL_SIZE_T size; |
2135 | unsigned long total = 0; |
2136 | int max_fast_bin; |
2137 | |
2138 | /* internal size_t must be no wider than pointer type */ |
2139 | assert (sizeof (INTERNAL_SIZE_T) <= sizeof (char *)); |
2140 | |
2141 | /* alignment is a power of 2 */ |
2142 | assert ((MALLOC_ALIGNMENT & (MALLOC_ALIGNMENT - 1)) == 0); |
2143 | |
2144 | /* cannot run remaining checks until fully initialized */ |
2145 | if (av->top == 0 || av->top == initial_top (av)) |
2146 | return; |
2147 | |
2148 | /* pagesize is a power of 2 */ |
2149 | assert (powerof2(GLRO (dl_pagesize))); |
2150 | |
2151 | /* A contiguous main_arena is consistent with sbrk_base. */ |
2152 | if (av == &main_arena && contiguous (av)) |
2153 | assert ((char *) mp_.sbrk_base + av->system_mem == |
2154 | (char *) av->top + chunksize (av->top)); |
2155 | |
2156 | /* properties of fastbins */ |
2157 | |
2158 | /* max_fast is in allowed range */ |
2159 | assert ((get_max_fast () & ~1) <= request2size (MAX_FAST_SIZE)); |
2160 | |
2161 | max_fast_bin = fastbin_index (get_max_fast ()); |
2162 | |
2163 | for (i = 0; i < NFASTBINS; ++i) |
2164 | { |
2165 | p = fastbin (av, i); |
2166 | |
2167 | /* The following test can only be performed for the main arena. |
2168 | While mallopt calls malloc_consolidate to get rid of all fast |
2169 | bins (especially those larger than the new maximum) this does |
2170 | only happen for the main arena. Trying to do this for any |
2171 | other arena would mean those arenas have to be locked and |
2172 | malloc_consolidate be called for them. This is excessive. And |
2173 | even if this is acceptable to somebody it still cannot solve |
2174 | the problem completely since if the arena is locked a |
2175 | concurrent malloc call might create a new arena which then |
2176 | could use the newly invalid fast bins. */ |
2177 | |
2178 | /* all bins past max_fast are empty */ |
2179 | if (av == &main_arena && i > max_fast_bin) |
2180 | assert (p == 0); |
2181 | |
2182 | while (p != 0) |
2183 | { |
2184 | /* each chunk claims to be inuse */ |
2185 | do_check_inuse_chunk (av, p); |
2186 | total += chunksize (p); |
2187 | /* chunk belongs in this bin */ |
2188 | assert (fastbin_index (chunksize (p)) == i); |
2189 | p = p->fd; |
2190 | } |
2191 | } |
2192 | |
2193 | /* check normal bins */ |
2194 | for (i = 1; i < NBINS; ++i) |
2195 | { |
2196 | b = bin_at (av, i); |
2197 | |
2198 | /* binmap is accurate (except for bin 1 == unsorted_chunks) */ |
2199 | if (i >= 2) |
2200 | { |
2201 | unsigned int binbit = get_binmap (av, i); |
2202 | int empty = last (b) == b; |
2203 | if (!binbit) |
2204 | assert (empty); |
2205 | else if (!empty) |
2206 | assert (binbit); |
2207 | } |
2208 | |
2209 | for (p = last (b); p != b; p = p->bk) |
2210 | { |
2211 | /* each chunk claims to be free */ |
2212 | do_check_free_chunk (av, p); |
2213 | size = chunksize (p); |
2214 | total += size; |
2215 | if (i >= 2) |
2216 | { |
2217 | /* chunk belongs in bin */ |
2218 | idx = bin_index (size); |
2219 | assert (idx == i); |
2220 | /* lists are sorted */ |
2221 | assert (p->bk == b || |
2222 | (unsigned long) chunksize (p->bk) >= (unsigned long) chunksize (p)); |
2223 | |
2224 | if (!in_smallbin_range (size)) |
2225 | { |
2226 | if (p->fd_nextsize != NULL) |
2227 | { |
2228 | if (p->fd_nextsize == p) |
2229 | assert (p->bk_nextsize == p); |
2230 | else |
2231 | { |
2232 | if (p->fd_nextsize == first (b)) |
2233 | assert (chunksize (p) < chunksize (p->fd_nextsize)); |
2234 | else |
2235 | assert (chunksize (p) > chunksize (p->fd_nextsize)); |
2236 | |
2237 | if (p == first (b)) |
2238 | assert (chunksize (p) > chunksize (p->bk_nextsize)); |
2239 | else |
2240 | assert (chunksize (p) < chunksize (p->bk_nextsize)); |
2241 | } |
2242 | } |
2243 | else |
2244 | assert (p->bk_nextsize == NULL); |
2245 | } |
2246 | } |
2247 | else if (!in_smallbin_range (size)) |
2248 | assert (p->fd_nextsize == NULL && p->bk_nextsize == NULL); |
2249 | /* chunk is followed by a legal chain of inuse chunks */ |
2250 | for (q = next_chunk (p); |
2251 | (q != av->top && inuse (q) && |
2252 | (unsigned long) (chunksize (q)) >= MINSIZE); |
2253 | q = next_chunk (q)) |
2254 | do_check_inuse_chunk (av, q); |
2255 | } |
2256 | } |
2257 | |
2258 | /* top chunk is OK */ |
2259 | check_chunk (av, av->top); |
2260 | } |
2261 | #endif |
2262 | |
2263 | |
2264 | /* ----------------- Support for debugging hooks -------------------- */ |
2265 | #include "hooks.c" |
2266 | |
2267 | |
2268 | /* ----------- Routines dealing with system allocation -------------- */ |
2269 | |
2270 | /* |
2271 | sysmalloc handles malloc cases requiring more memory from the system. |
2272 | On entry, it is assumed that av->top does not have enough |
2273 | space to service request for nb bytes, thus requiring that av->top |
2274 | be extended or replaced. |
2275 | */ |
2276 | |
2277 | static void * |
2278 | sysmalloc (INTERNAL_SIZE_T nb, mstate av) |
2279 | { |
2280 | mchunkptr old_top; /* incoming value of av->top */ |
2281 | INTERNAL_SIZE_T old_size; /* its size */ |
2282 | char *old_end; /* its end address */ |
2283 | |
2284 | long size; /* arg to first MORECORE or mmap call */ |
2285 | char *brk; /* return value from MORECORE */ |
2286 | |
2287 | long correction; /* arg to 2nd MORECORE call */ |
2288 | char *snd_brk; /* 2nd return val */ |
2289 | |
2290 | INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of new space */ |
2291 | INTERNAL_SIZE_T end_misalign; /* partial page left at end of new space */ |
2292 | char *aligned_brk; /* aligned offset into brk */ |
2293 | |
2294 | mchunkptr p; /* the allocated/returned chunk */ |
2295 | mchunkptr remainder; /* remainder from allocation */ |
2296 | unsigned long remainder_size; /* its size */ |
2297 | |
2298 | |
2299 | size_t pagesize = GLRO (dl_pagesize); |
2300 | bool tried_mmap = false; |
2301 | |
2302 | |
2303 | /* |
2304 | If have mmap, and the request size meets the mmap threshold, and |
2305 | the system supports mmap, and there are few enough currently |
2306 | allocated mmapped regions, try to directly map this request |
2307 | rather than expanding top. |
2308 | */ |
2309 | |
2310 | if (av == NULL |
2311 | || ((unsigned long) (nb) >= (unsigned long) (mp_.mmap_threshold) |
2312 | && (mp_.n_mmaps < mp_.n_mmaps_max))) |
2313 | { |
2314 | char *mm; /* return value from mmap call*/ |
2315 | |
2316 | try_mmap: |
2317 | /* |
2318 | Round up size to nearest page. For mmapped chunks, the overhead |
2319 | is one SIZE_SZ unit larger than for normal chunks, because there |
2320 | is no following chunk whose prev_size field could be used. |
2321 | |
2322 | See the front_misalign handling below, for glibc there is no |
2323 | need for further alignments unless we have have high alignment. |
2324 | */ |
2325 | if (MALLOC_ALIGNMENT == 2 * SIZE_SZ) |
2326 | size = ALIGN_UP (nb + SIZE_SZ, pagesize); |
2327 | else |
2328 | size = ALIGN_UP (nb + SIZE_SZ + MALLOC_ALIGN_MASK, pagesize); |
2329 | tried_mmap = true; |
2330 | |
2331 | /* Don't try if size wraps around 0 */ |
2332 | if ((unsigned long) (size) > (unsigned long) (nb)) |
2333 | { |
2334 | mm = (char *) (MMAP (0, size, PROT_READ | PROT_WRITE, 0)); |
2335 | |
2336 | if (mm != MAP_FAILED) |
2337 | { |
2338 | /* |
2339 | The offset to the start of the mmapped region is stored |
2340 | in the prev_size field of the chunk. This allows us to adjust |
2341 | returned start address to meet alignment requirements here |
2342 | and in memalign(), and still be able to compute proper |
2343 | address argument for later munmap in free() and realloc(). |
2344 | */ |
2345 | |
2346 | if (MALLOC_ALIGNMENT == 2 * SIZE_SZ) |
2347 | { |
2348 | /* For glibc, chunk2mem increases the address by 2*SIZE_SZ and |
2349 | MALLOC_ALIGN_MASK is 2*SIZE_SZ-1. Each mmap'ed area is page |
2350 | aligned and therefore definitely MALLOC_ALIGN_MASK-aligned. */ |
2351 | assert (((INTERNAL_SIZE_T) chunk2mem (mm) & MALLOC_ALIGN_MASK) == 0); |
2352 | front_misalign = 0; |
2353 | } |
2354 | else |
2355 | front_misalign = (INTERNAL_SIZE_T) chunk2mem (mm) & MALLOC_ALIGN_MASK; |
2356 | if (front_misalign > 0) |
2357 | { |
2358 | correction = MALLOC_ALIGNMENT - front_misalign; |
2359 | p = (mchunkptr) (mm + correction); |
2360 | set_prev_size (p, correction); |
2361 | set_head (p, (size - correction) | IS_MMAPPED); |
2362 | } |
2363 | else |
2364 | { |
2365 | p = (mchunkptr) mm; |
2366 | set_prev_size (p, 0); |
2367 | set_head (p, size | IS_MMAPPED); |
2368 | } |
2369 | |
2370 | /* update statistics */ |
2371 | |
2372 | int new = atomic_exchange_and_add (&mp_.n_mmaps, 1) + 1; |
2373 | atomic_max (&mp_.max_n_mmaps, new); |
2374 | |
2375 | unsigned long sum; |
2376 | sum = atomic_exchange_and_add (&mp_.mmapped_mem, size) + size; |
2377 | atomic_max (&mp_.max_mmapped_mem, sum); |
2378 | |
2379 | check_chunk (av, p); |
2380 | |
2381 | return chunk2mem (p); |
2382 | } |
2383 | } |
2384 | } |
2385 | |
2386 | /* There are no usable arenas and mmap also failed. */ |
2387 | if (av == NULL) |
2388 | return 0; |
2389 | |
2390 | /* Record incoming configuration of top */ |
2391 | |
2392 | old_top = av->top; |
2393 | old_size = chunksize (old_top); |
2394 | old_end = (char *) (chunk_at_offset (old_top, old_size)); |
2395 | |
2396 | brk = snd_brk = (char *) (MORECORE_FAILURE); |
2397 | |
2398 | /* |
2399 | If not the first time through, we require old_size to be |
2400 | at least MINSIZE and to have prev_inuse set. |
2401 | */ |
2402 | |
2403 | assert ((old_top == initial_top (av) && old_size == 0) || |
2404 | ((unsigned long) (old_size) >= MINSIZE && |
2405 | prev_inuse (old_top) && |
2406 | ((unsigned long) old_end & (pagesize - 1)) == 0)); |
2407 | |
2408 | /* Precondition: not enough current space to satisfy nb request */ |
2409 | assert ((unsigned long) (old_size) < (unsigned long) (nb + MINSIZE)); |
2410 | |
2411 | |
2412 | if (av != &main_arena) |
2413 | { |
2414 | heap_info *old_heap, *heap; |
2415 | size_t old_heap_size; |
2416 | |
2417 | /* First try to extend the current heap. */ |
2418 | old_heap = heap_for_ptr (old_top); |
2419 | old_heap_size = old_heap->size; |
2420 | if ((long) (MINSIZE + nb - old_size) > 0 |
2421 | && grow_heap (old_heap, MINSIZE + nb - old_size) == 0) |
2422 | { |
2423 | av->system_mem += old_heap->size - old_heap_size; |
2424 | set_head (old_top, (((char *) old_heap + old_heap->size) - (char *) old_top) |
2425 | | PREV_INUSE); |
2426 | } |
2427 | else if ((heap = new_heap (nb + (MINSIZE + sizeof (*heap)), mp_.top_pad))) |
2428 | { |
2429 | /* Use a newly allocated heap. */ |
2430 | heap->ar_ptr = av; |
2431 | heap->prev = old_heap; |
2432 | av->system_mem += heap->size; |
2433 | /* Set up the new top. */ |
2434 | top (av) = chunk_at_offset (heap, sizeof (*heap)); |
2435 | set_head (top (av), (heap->size - sizeof (*heap)) | PREV_INUSE); |
2436 | |
2437 | /* Setup fencepost and free the old top chunk with a multiple of |
2438 | MALLOC_ALIGNMENT in size. */ |
2439 | /* The fencepost takes at least MINSIZE bytes, because it might |
2440 | become the top chunk again later. Note that a footer is set |
2441 | up, too, although the chunk is marked in use. */ |
2442 | old_size = (old_size - MINSIZE) & ~MALLOC_ALIGN_MASK; |
2443 | set_head (chunk_at_offset (old_top, old_size + 2 * SIZE_SZ), 0 | PREV_INUSE); |
2444 | if (old_size >= MINSIZE) |
2445 | { |
2446 | set_head (chunk_at_offset (old_top, old_size), (2 * SIZE_SZ) | PREV_INUSE); |
2447 | set_foot (chunk_at_offset (old_top, old_size), (2 * SIZE_SZ)); |
2448 | set_head (old_top, old_size | PREV_INUSE | NON_MAIN_ARENA); |
2449 | _int_free (av, old_top, 1); |
2450 | } |
2451 | else |
2452 | { |
2453 | set_head (old_top, (old_size + 2 * SIZE_SZ) | PREV_INUSE); |
2454 | set_foot (old_top, (old_size + 2 * SIZE_SZ)); |
2455 | } |
2456 | } |
2457 | else if (!tried_mmap) |
2458 | /* We can at least try to use to mmap memory. */ |
2459 | goto try_mmap; |
2460 | } |
2461 | else /* av == main_arena */ |
2462 | |
2463 | |
2464 | { /* Request enough space for nb + pad + overhead */ |
2465 | size = nb + mp_.top_pad + MINSIZE; |
2466 | |
2467 | /* |
2468 | If contiguous, we can subtract out existing space that we hope to |
2469 | combine with new space. We add it back later only if |
2470 | we don't actually get contiguous space. |
2471 | */ |
2472 | |
2473 | if (contiguous (av)) |
2474 | size -= old_size; |
2475 | |
2476 | /* |
2477 | Round to a multiple of page size. |
2478 | If MORECORE is not contiguous, this ensures that we only call it |
2479 | with whole-page arguments. And if MORECORE is contiguous and |
2480 | this is not first time through, this preserves page-alignment of |
2481 | previous calls. Otherwise, we correct to page-align below. |
2482 | */ |
2483 | |
2484 | size = ALIGN_UP (size, pagesize); |
2485 | |
2486 | /* |
2487 | Don't try to call MORECORE if argument is so big as to appear |
2488 | negative. Note that since mmap takes size_t arg, it may succeed |
2489 | below even if we cannot call MORECORE. |
2490 | */ |
2491 | |
2492 | if (size > 0) |
2493 | { |
2494 | brk = (char *) (MORECORE (size)); |
2495 | LIBC_PROBE (memory_sbrk_more, 2, brk, size); |
2496 | } |
2497 | |
2498 | if (brk != (char *) (MORECORE_FAILURE)) |
2499 | { |
2500 | /* Call the `morecore' hook if necessary. */ |
2501 | void (*hook) (void) = atomic_forced_read (__after_morecore_hook); |
2502 | if (__builtin_expect (hook != NULL, 0)) |
2503 | (*hook)(); |
2504 | } |
2505 | else |
2506 | { |
2507 | /* |
2508 | If have mmap, try using it as a backup when MORECORE fails or |
2509 | cannot be used. This is worth doing on systems that have "holes" in |
2510 | address space, so sbrk cannot extend to give contiguous space, but |
2511 | space is available elsewhere. Note that we ignore mmap max count |
2512 | and threshold limits, since the space will not be used as a |
2513 | segregated mmap region. |
2514 | */ |
2515 | |
2516 | /* Cannot merge with old top, so add its size back in */ |
2517 | if (contiguous (av)) |
2518 | size = ALIGN_UP (size + old_size, pagesize); |
2519 | |
2520 | /* If we are relying on mmap as backup, then use larger units */ |
2521 | if ((unsigned long) (size) < (unsigned long) (MMAP_AS_MORECORE_SIZE)) |
2522 | size = MMAP_AS_MORECORE_SIZE; |
2523 | |
2524 | /* Don't try if size wraps around 0 */ |
2525 | if ((unsigned long) (size) > (unsigned long) (nb)) |
2526 | { |
2527 | char *mbrk = (char *) (MMAP (0, size, PROT_READ | PROT_WRITE, 0)); |
2528 | |
2529 | if (mbrk != MAP_FAILED) |
2530 | { |
2531 | /* We do not need, and cannot use, another sbrk call to find end */ |
2532 | brk = mbrk; |
2533 | snd_brk = brk + size; |
2534 | |
2535 | /* |
2536 | Record that we no longer have a contiguous sbrk region. |
2537 | After the first time mmap is used as backup, we do not |
2538 | ever rely on contiguous space since this could incorrectly |
2539 | bridge regions. |
2540 | */ |
2541 | set_noncontiguous (av); |
2542 | } |
2543 | } |
2544 | } |
2545 | |
2546 | if (brk != (char *) (MORECORE_FAILURE)) |
2547 | { |
2548 | if (mp_.sbrk_base == 0) |
2549 | mp_.sbrk_base = brk; |
2550 | av->system_mem += size; |
2551 | |
2552 | /* |
2553 | If MORECORE extends previous space, we can likewise extend top size. |
2554 | */ |
2555 | |
2556 | if (brk == old_end && snd_brk == (char *) (MORECORE_FAILURE)) |
2557 | set_head (old_top, (size + old_size) | PREV_INUSE); |
2558 | |
2559 | else if (contiguous (av) && old_size && brk < old_end) |
2560 | /* Oops! Someone else killed our space.. Can't touch anything. */ |
2561 | malloc_printerr ("break adjusted to free malloc space" ); |
2562 | |
2563 | /* |
2564 | Otherwise, make adjustments: |
2565 | |
2566 | * If the first time through or noncontiguous, we need to call sbrk |
2567 | just to find out where the end of memory lies. |
2568 | |
2569 | * We need to ensure that all returned chunks from malloc will meet |
2570 | MALLOC_ALIGNMENT |
2571 | |
2572 | * If there was an intervening foreign sbrk, we need to adjust sbrk |
2573 | request size to account for fact that we will not be able to |
2574 | combine new space with existing space in old_top. |
2575 | |
2576 | * Almost all systems internally allocate whole pages at a time, in |
2577 | which case we might as well use the whole last page of request. |
2578 | So we allocate enough more memory to hit a page boundary now, |
2579 | which in turn causes future contiguous calls to page-align. |
2580 | */ |
2581 | |
2582 | else |
2583 | { |
2584 | front_misalign = 0; |
2585 | end_misalign = 0; |
2586 | correction = 0; |
2587 | aligned_brk = brk; |
2588 | |
2589 | /* handle contiguous cases */ |
2590 | if (contiguous (av)) |
2591 | { |
2592 | /* Count foreign sbrk as system_mem. */ |
2593 | if (old_size) |
2594 | av->system_mem += brk - old_end; |
2595 | |
2596 | /* Guarantee alignment of first new chunk made from this space */ |
2597 | |
2598 | front_misalign = (INTERNAL_SIZE_T) chunk2mem (brk) & MALLOC_ALIGN_MASK; |
2599 | if (front_misalign > 0) |
2600 | { |
2601 | /* |
2602 | Skip over some bytes to arrive at an aligned position. |
2603 | We don't need to specially mark these wasted front bytes. |
2604 | They will never be accessed anyway because |
2605 | prev_inuse of av->top (and any chunk created from its start) |
2606 | is always true after initialization. |
2607 | */ |
2608 | |
2609 | correction = MALLOC_ALIGNMENT - front_misalign; |
2610 | aligned_brk += correction; |
2611 | } |
2612 | |
2613 | /* |
2614 | If this isn't adjacent to existing space, then we will not |
2615 | be able to merge with old_top space, so must add to 2nd request. |
2616 | */ |
2617 | |
2618 | correction += old_size; |
2619 | |
2620 | /* Extend the end address to hit a page boundary */ |
2621 | end_misalign = (INTERNAL_SIZE_T) (brk + size + correction); |
2622 | correction += (ALIGN_UP (end_misalign, pagesize)) - end_misalign; |
2623 | |
2624 | assert (correction >= 0); |
2625 | snd_brk = (char *) (MORECORE (correction)); |
2626 | |
2627 | /* |
2628 | If can't allocate correction, try to at least find out current |
2629 | brk. It might be enough to proceed without failing. |
2630 | |
2631 | Note that if second sbrk did NOT fail, we assume that space |
2632 | is contiguous with first sbrk. This is a safe assumption unless |
2633 | program is multithreaded but doesn't use locks and a foreign sbrk |
2634 | occurred between our first and second calls. |
2635 | */ |
2636 | |
2637 | if (snd_brk == (char *) (MORECORE_FAILURE)) |
2638 | { |
2639 | correction = 0; |
2640 | snd_brk = (char *) (MORECORE (0)); |
2641 | } |
2642 | else |
2643 | { |
2644 | /* Call the `morecore' hook if necessary. */ |
2645 | void (*hook) (void) = atomic_forced_read (__after_morecore_hook); |
2646 | if (__builtin_expect (hook != NULL, 0)) |
2647 | (*hook)(); |
2648 | } |
2649 | } |
2650 | |
2651 | /* handle non-contiguous cases */ |
2652 | else |
2653 | { |
2654 | if (MALLOC_ALIGNMENT == 2 * SIZE_SZ) |
2655 | /* MORECORE/mmap must correctly align */ |
2656 | assert (((unsigned long) chunk2mem (brk) & MALLOC_ALIGN_MASK) == 0); |
2657 | else |
2658 | { |
2659 | front_misalign = (INTERNAL_SIZE_T) chunk2mem (brk) & MALLOC_ALIGN_MASK; |
2660 | if (front_misalign > 0) |
2661 | { |
2662 | /* |
2663 | Skip over some bytes to arrive at an aligned position. |
2664 | We don't need to specially mark these wasted front bytes. |
2665 | They will never be accessed anyway because |
2666 | prev_inuse of av->top (and any chunk created from its start) |
2667 | is always true after initialization. |
2668 | */ |
2669 | |
2670 | aligned_brk += MALLOC_ALIGNMENT - front_misalign; |
2671 | } |
2672 | } |
2673 | |
2674 | /* Find out current end of memory */ |
2675 | if (snd_brk == (char *) (MORECORE_FAILURE)) |
2676 | { |
2677 | snd_brk = (char *) (MORECORE (0)); |
2678 | } |
2679 | } |
2680 | |
2681 | /* Adjust top based on results of second sbrk */ |
2682 | if (snd_brk != (char *) (MORECORE_FAILURE)) |
2683 | { |
2684 | av->top = (mchunkptr) aligned_brk; |
2685 | set_head (av->top, (snd_brk - aligned_brk + correction) | PREV_INUSE); |
2686 | av->system_mem += correction; |
2687 | |
2688 | /* |
2689 | If not the first time through, we either have a |
2690 | gap due to foreign sbrk or a non-contiguous region. Insert a |
2691 | double fencepost at old_top to prevent consolidation with space |
2692 | we don't own. These fenceposts are artificial chunks that are |
2693 | marked as inuse and are in any case too small to use. We need |
2694 | two to make sizes and alignments work out. |
2695 | */ |
2696 | |
2697 | if (old_size != 0) |
2698 | { |
2699 | /* |
2700 | Shrink old_top to insert fenceposts, keeping size a |
2701 | multiple of MALLOC_ALIGNMENT. We know there is at least |
2702 | enough space in old_top to do this. |
2703 | */ |
2704 | old_size = (old_size - 4 * SIZE_SZ) & ~MALLOC_ALIGN_MASK; |
2705 | set_head (old_top, old_size | PREV_INUSE); |
2706 | |
2707 | /* |
2708 | Note that the following assignments completely overwrite |
2709 | old_top when old_size was previously MINSIZE. This is |
2710 | intentional. We need the fencepost, even if old_top otherwise gets |
2711 | lost. |
2712 | */ |
2713 | set_head (chunk_at_offset (old_top, old_size), |
2714 | (2 * SIZE_SZ) | PREV_INUSE); |
2715 | set_head (chunk_at_offset (old_top, old_size + 2 * SIZE_SZ), |
2716 | (2 * SIZE_SZ) | PREV_INUSE); |
2717 | |
2718 | /* If possible, release the rest. */ |
2719 | if (old_size >= MINSIZE) |
2720 | { |
2721 | _int_free (av, old_top, 1); |
2722 | } |
2723 | } |
2724 | } |
2725 | } |
2726 | } |
2727 | } /* if (av != &main_arena) */ |
2728 | |
2729 | if ((unsigned long) av->system_mem > (unsigned long) (av->max_system_mem)) |
2730 | av->max_system_mem = av->system_mem; |
2731 | check_malloc_state (av); |
2732 | |
2733 | /* finally, do the allocation */ |
2734 | p = av->top; |
2735 | size = chunksize (p); |
2736 | |
2737 | /* check that one of the above allocation paths succeeded */ |
2738 | if ((unsigned long) (size) >= (unsigned long) (nb + MINSIZE)) |
2739 | { |
2740 | remainder_size = size - nb; |
2741 | remainder = chunk_at_offset (p, nb); |
2742 | av->top = remainder; |
2743 | set_head (p, nb | PREV_INUSE | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
2744 | set_head (remainder, remainder_size | PREV_INUSE); |
2745 | check_malloced_chunk (av, p, nb); |
2746 | return chunk2mem (p); |
2747 | } |
2748 | |
2749 | /* catch all failure paths */ |
2750 | __set_errno (ENOMEM); |
2751 | return 0; |
2752 | } |
2753 | |
2754 | |
2755 | /* |
2756 | systrim is an inverse of sorts to sysmalloc. It gives memory back |
2757 | to the system (via negative arguments to sbrk) if there is unused |
2758 | memory at the `high' end of the malloc pool. It is called |
2759 | automatically by free() when top space exceeds the trim |
2760 | threshold. It is also called by the public malloc_trim routine. It |
2761 | returns 1 if it actually released any memory, else 0. |
2762 | */ |
2763 | |
2764 | static int |
2765 | systrim (size_t pad, mstate av) |
2766 | { |
2767 | long top_size; /* Amount of top-most memory */ |
2768 | long ; /* Amount to release */ |
2769 | long released; /* Amount actually released */ |
2770 | char *current_brk; /* address returned by pre-check sbrk call */ |
2771 | char *new_brk; /* address returned by post-check sbrk call */ |
2772 | size_t pagesize; |
2773 | long top_area; |
2774 | |
2775 | pagesize = GLRO (dl_pagesize); |
2776 | top_size = chunksize (av->top); |
2777 | |
2778 | top_area = top_size - MINSIZE - 1; |
2779 | if (top_area <= pad) |
2780 | return 0; |
2781 | |
2782 | /* Release in pagesize units and round down to the nearest page. */ |
2783 | extra = ALIGN_DOWN(top_area - pad, pagesize); |
2784 | |
2785 | if (extra == 0) |
2786 | return 0; |
2787 | |
2788 | /* |
2789 | Only proceed if end of memory is where we last set it. |
2790 | This avoids problems if there were foreign sbrk calls. |
2791 | */ |
2792 | current_brk = (char *) (MORECORE (0)); |
2793 | if (current_brk == (char *) (av->top) + top_size) |
2794 | { |
2795 | /* |
2796 | Attempt to release memory. We ignore MORECORE return value, |
2797 | and instead call again to find out where new end of memory is. |
2798 | This avoids problems if first call releases less than we asked, |
2799 | of if failure somehow altered brk value. (We could still |
2800 | encounter problems if it altered brk in some very bad way, |
2801 | but the only thing we can do is adjust anyway, which will cause |
2802 | some downstream failure.) |
2803 | */ |
2804 | |
2805 | MORECORE (-extra); |
2806 | /* Call the `morecore' hook if necessary. */ |
2807 | void (*hook) (void) = atomic_forced_read (__after_morecore_hook); |
2808 | if (__builtin_expect (hook != NULL, 0)) |
2809 | (*hook)(); |
2810 | new_brk = (char *) (MORECORE (0)); |
2811 | |
2812 | LIBC_PROBE (memory_sbrk_less, 2, new_brk, extra); |
2813 | |
2814 | if (new_brk != (char *) MORECORE_FAILURE) |
2815 | { |
2816 | released = (long) (current_brk - new_brk); |
2817 | |
2818 | if (released != 0) |
2819 | { |
2820 | /* Success. Adjust top. */ |
2821 | av->system_mem -= released; |
2822 | set_head (av->top, (top_size - released) | PREV_INUSE); |
2823 | check_malloc_state (av); |
2824 | return 1; |
2825 | } |
2826 | } |
2827 | } |
2828 | return 0; |
2829 | } |
2830 | |
2831 | static void |
2832 | internal_function |
2833 | munmap_chunk (mchunkptr p) |
2834 | { |
2835 | INTERNAL_SIZE_T size = chunksize (p); |
2836 | |
2837 | assert (chunk_is_mmapped (p)); |
2838 | |
2839 | /* Do nothing if the chunk is a faked mmapped chunk in the dumped |
2840 | main arena. We never free this memory. */ |
2841 | if (DUMPED_MAIN_ARENA_CHUNK (p)) |
2842 | return; |
2843 | |
2844 | uintptr_t block = (uintptr_t) p - prev_size (p); |
2845 | size_t total_size = prev_size (p) + size; |
2846 | /* Unfortunately we have to do the compilers job by hand here. Normally |
2847 | we would test BLOCK and TOTAL-SIZE separately for compliance with the |
2848 | page size. But gcc does not recognize the optimization possibility |
2849 | (in the moment at least) so we combine the two values into one before |
2850 | the bit test. */ |
2851 | if (__builtin_expect (((block | total_size) & (GLRO (dl_pagesize) - 1)) != 0, 0)) |
2852 | malloc_printerr ("munmap_chunk(): invalid pointer" ); |
2853 | |
2854 | atomic_decrement (&mp_.n_mmaps); |
2855 | atomic_add (&mp_.mmapped_mem, -total_size); |
2856 | |
2857 | /* If munmap failed the process virtual memory address space is in a |
2858 | bad shape. Just leave the block hanging around, the process will |
2859 | terminate shortly anyway since not much can be done. */ |
2860 | __munmap ((char *) block, total_size); |
2861 | } |
2862 | |
2863 | #if HAVE_MREMAP |
2864 | |
2865 | static mchunkptr |
2866 | internal_function |
2867 | mremap_chunk (mchunkptr p, size_t new_size) |
2868 | { |
2869 | size_t pagesize = GLRO (dl_pagesize); |
2870 | INTERNAL_SIZE_T offset = prev_size (p); |
2871 | INTERNAL_SIZE_T size = chunksize (p); |
2872 | char *cp; |
2873 | |
2874 | assert (chunk_is_mmapped (p)); |
2875 | assert (((size + offset) & (GLRO (dl_pagesize) - 1)) == 0); |
2876 | |
2877 | /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */ |
2878 | new_size = ALIGN_UP (new_size + offset + SIZE_SZ, pagesize); |
2879 | |
2880 | /* No need to remap if the number of pages does not change. */ |
2881 | if (size + offset == new_size) |
2882 | return p; |
2883 | |
2884 | cp = (char *) __mremap ((char *) p - offset, size + offset, new_size, |
2885 | MREMAP_MAYMOVE); |
2886 | |
2887 | if (cp == MAP_FAILED) |
2888 | return 0; |
2889 | |
2890 | p = (mchunkptr) (cp + offset); |
2891 | |
2892 | assert (aligned_OK (chunk2mem (p))); |
2893 | |
2894 | assert (prev_size (p) == offset); |
2895 | set_head (p, (new_size - offset) | IS_MMAPPED); |
2896 | |
2897 | INTERNAL_SIZE_T new; |
2898 | new = atomic_exchange_and_add (&mp_.mmapped_mem, new_size - size - offset) |
2899 | + new_size - size - offset; |
2900 | atomic_max (&mp_.max_mmapped_mem, new); |
2901 | return p; |
2902 | } |
2903 | #endif /* HAVE_MREMAP */ |
2904 | |
2905 | /*------------------------ Public wrappers. --------------------------------*/ |
2906 | |
2907 | #if USE_TCACHE |
2908 | |
2909 | /* We overlay this structure on the user-data portion of a chunk when |
2910 | the chunk is stored in the per-thread cache. */ |
2911 | typedef struct tcache_entry |
2912 | { |
2913 | struct tcache_entry *next; |
2914 | } tcache_entry; |
2915 | |
2916 | /* There is one of these for each thread, which contains the |
2917 | per-thread cache (hence "tcache_perthread_struct"). Keeping |
2918 | overall size low is mildly important. Note that COUNTS and ENTRIES |
2919 | are redundant (we could have just counted the linked list each |
2920 | time), this is for performance reasons. */ |
2921 | typedef struct tcache_perthread_struct |
2922 | { |
2923 | char counts[TCACHE_MAX_BINS]; |
2924 | tcache_entry *entries[TCACHE_MAX_BINS]; |
2925 | } tcache_perthread_struct; |
2926 | |
2927 | #define MAX_TCACHE_COUNT 127 /* Maximum value of counts[] entries. */ |
2928 | |
2929 | static __thread bool tcache_shutting_down = false; |
2930 | static __thread tcache_perthread_struct *tcache = NULL; |
2931 | |
2932 | /* Caller must ensure that we know tc_idx is valid and there's room |
2933 | for more chunks. */ |
2934 | static __always_inline void |
2935 | tcache_put (mchunkptr chunk, size_t tc_idx) |
2936 | { |
2937 | tcache_entry *e = (tcache_entry *) chunk2mem (chunk); |
2938 | assert (tc_idx < TCACHE_MAX_BINS); |
2939 | e->next = tcache->entries[tc_idx]; |
2940 | tcache->entries[tc_idx] = e; |
2941 | ++(tcache->counts[tc_idx]); |
2942 | } |
2943 | |
2944 | /* Caller must ensure that we know tc_idx is valid and there's |
2945 | available chunks to remove. */ |
2946 | static __always_inline void * |
2947 | tcache_get (size_t tc_idx) |
2948 | { |
2949 | tcache_entry *e = tcache->entries[tc_idx]; |
2950 | assert (tc_idx < TCACHE_MAX_BINS); |
2951 | assert (tcache->entries[tc_idx] > 0); |
2952 | tcache->entries[tc_idx] = e->next; |
2953 | --(tcache->counts[tc_idx]); |
2954 | return (void *) e; |
2955 | } |
2956 | |
2957 | static void __attribute__ ((section ("__libc_thread_freeres_fn" ))) |
2958 | tcache_thread_freeres (void) |
2959 | { |
2960 | int i; |
2961 | tcache_perthread_struct *tcache_tmp = tcache; |
2962 | |
2963 | if (!tcache) |
2964 | return; |
2965 | |
2966 | /* Disable the tcache and prevent it from being reinitialized. */ |
2967 | tcache = NULL; |
2968 | tcache_shutting_down = true; |
2969 | |
2970 | /* Free all of the entries and the tcache itself back to the arena |
2971 | heap for coalescing. */ |
2972 | for (i = 0; i < TCACHE_MAX_BINS; ++i) |
2973 | { |
2974 | while (tcache_tmp->entries[i]) |
2975 | { |
2976 | tcache_entry *e = tcache_tmp->entries[i]; |
2977 | tcache_tmp->entries[i] = e->next; |
2978 | __libc_free (e); |
2979 | } |
2980 | } |
2981 | |
2982 | __libc_free (tcache_tmp); |
2983 | } |
2984 | text_set_element (__libc_thread_subfreeres, tcache_thread_freeres); |
2985 | |
2986 | static void |
2987 | tcache_init(void) |
2988 | { |
2989 | mstate ar_ptr; |
2990 | void *victim = 0; |
2991 | const size_t bytes = sizeof (tcache_perthread_struct); |
2992 | |
2993 | if (tcache_shutting_down) |
2994 | return; |
2995 | |
2996 | arena_get (ar_ptr, bytes); |
2997 | victim = _int_malloc (ar_ptr, bytes); |
2998 | if (!victim && ar_ptr != NULL) |
2999 | { |
3000 | ar_ptr = arena_get_retry (ar_ptr, bytes); |
3001 | victim = _int_malloc (ar_ptr, bytes); |
3002 | } |
3003 | |
3004 | |
3005 | if (ar_ptr != NULL) |
3006 | __libc_lock_unlock (ar_ptr->mutex); |
3007 | |
3008 | /* In a low memory situation, we may not be able to allocate memory |
3009 | - in which case, we just keep trying later. However, we |
3010 | typically do this very early, so either there is sufficient |
3011 | memory, or there isn't enough memory to do non-trivial |
3012 | allocations anyway. */ |
3013 | if (victim) |
3014 | { |
3015 | tcache = (tcache_perthread_struct *) victim; |
3016 | memset (tcache, 0, sizeof (tcache_perthread_struct)); |
3017 | } |
3018 | |
3019 | } |
3020 | |
3021 | #define MAYBE_INIT_TCACHE() \ |
3022 | if (__glibc_unlikely (tcache == NULL)) \ |
3023 | tcache_init(); |
3024 | |
3025 | #else |
3026 | #define MAYBE_INIT_TCACHE() |
3027 | #endif |
3028 | |
3029 | void * |
3030 | __libc_malloc (size_t bytes) |
3031 | { |
3032 | mstate ar_ptr; |
3033 | void *victim; |
3034 | |
3035 | void *(*hook) (size_t, const void *) |
3036 | = atomic_forced_read (__malloc_hook); |
3037 | if (__builtin_expect (hook != NULL, 0)) |
3038 | return (*hook)(bytes, RETURN_ADDRESS (0)); |
3039 | #if USE_TCACHE |
3040 | /* int_free also calls request2size, be careful to not pad twice. */ |
3041 | size_t tbytes; |
3042 | checked_request2size (bytes, tbytes); |
3043 | size_t tc_idx = csize2tidx (tbytes); |
3044 | |
3045 | MAYBE_INIT_TCACHE (); |
3046 | |
3047 | DIAG_PUSH_NEEDS_COMMENT; |
3048 | if (tc_idx < mp_.tcache_bins |
3049 | /*&& tc_idx < TCACHE_MAX_BINS*/ /* to appease gcc */ |
3050 | && tcache |
3051 | && tcache->entries[tc_idx] != NULL) |
3052 | { |
3053 | return tcache_get (tc_idx); |
3054 | } |
3055 | DIAG_POP_NEEDS_COMMENT; |
3056 | #endif |
3057 | |
3058 | if (SINGLE_THREAD_P) |
3059 | { |
3060 | victim = _int_malloc (&main_arena, bytes); |
3061 | assert (!victim || chunk_is_mmapped (mem2chunk (victim)) || |
3062 | &main_arena == arena_for_chunk (mem2chunk (victim))); |
3063 | return victim; |
3064 | } |
3065 | |
3066 | arena_get (ar_ptr, bytes); |
3067 | |
3068 | victim = _int_malloc (ar_ptr, bytes); |
3069 | /* Retry with another arena only if we were able to find a usable arena |
3070 | before. */ |
3071 | if (!victim && ar_ptr != NULL) |
3072 | { |
3073 | LIBC_PROBE (memory_malloc_retry, 1, bytes); |
3074 | ar_ptr = arena_get_retry (ar_ptr, bytes); |
3075 | victim = _int_malloc (ar_ptr, bytes); |
3076 | } |
3077 | |
3078 | if (ar_ptr != NULL) |
3079 | __libc_lock_unlock (ar_ptr->mutex); |
3080 | |
3081 | assert (!victim || chunk_is_mmapped (mem2chunk (victim)) || |
3082 | ar_ptr == arena_for_chunk (mem2chunk (victim))); |
3083 | return victim; |
3084 | } |
3085 | libc_hidden_def (__libc_malloc) |
3086 | |
3087 | void |
3088 | __libc_free (void *mem) |
3089 | { |
3090 | mstate ar_ptr; |
3091 | mchunkptr p; /* chunk corresponding to mem */ |
3092 | |
3093 | void (*hook) (void *, const void *) |
3094 | = atomic_forced_read (__free_hook); |
3095 | if (__builtin_expect (hook != NULL, 0)) |
3096 | { |
3097 | (*hook)(mem, RETURN_ADDRESS (0)); |
3098 | return; |
3099 | } |
3100 | |
3101 | if (mem == 0) /* free(0) has no effect */ |
3102 | return; |
3103 | |
3104 | p = mem2chunk (mem); |
3105 | |
3106 | if (chunk_is_mmapped (p)) /* release mmapped memory. */ |
3107 | { |
3108 | /* See if the dynamic brk/mmap threshold needs adjusting. |
3109 | Dumped fake mmapped chunks do not affect the threshold. */ |
3110 | if (!mp_.no_dyn_threshold |
3111 | && chunksize_nomask (p) > mp_.mmap_threshold |
3112 | && chunksize_nomask (p) <= DEFAULT_MMAP_THRESHOLD_MAX |
3113 | && !DUMPED_MAIN_ARENA_CHUNK (p)) |
3114 | { |
3115 | mp_.mmap_threshold = chunksize (p); |
3116 | mp_.trim_threshold = 2 * mp_.mmap_threshold; |
3117 | LIBC_PROBE (memory_mallopt_free_dyn_thresholds, 2, |
3118 | mp_.mmap_threshold, mp_.trim_threshold); |
3119 | } |
3120 | munmap_chunk (p); |
3121 | return; |
3122 | } |
3123 | |
3124 | MAYBE_INIT_TCACHE (); |
3125 | |
3126 | ar_ptr = arena_for_chunk (p); |
3127 | _int_free (ar_ptr, p, 0); |
3128 | } |
3129 | libc_hidden_def (__libc_free) |
3130 | |
3131 | void * |
3132 | __libc_realloc (void *oldmem, size_t bytes) |
3133 | { |
3134 | mstate ar_ptr; |
3135 | INTERNAL_SIZE_T nb; /* padded request size */ |
3136 | |
3137 | void *newp; /* chunk to return */ |
3138 | |
3139 | void *(*hook) (void *, size_t, const void *) = |
3140 | atomic_forced_read (__realloc_hook); |
3141 | if (__builtin_expect (hook != NULL, 0)) |
3142 | return (*hook)(oldmem, bytes, RETURN_ADDRESS (0)); |
3143 | |
3144 | #if REALLOC_ZERO_BYTES_FREES |
3145 | if (bytes == 0 && oldmem != NULL) |
3146 | { |
3147 | __libc_free (oldmem); return 0; |
3148 | } |
3149 | #endif |
3150 | |
3151 | /* realloc of null is supposed to be same as malloc */ |
3152 | if (oldmem == 0) |
3153 | return __libc_malloc (bytes); |
3154 | |
3155 | /* chunk corresponding to oldmem */ |
3156 | const mchunkptr oldp = mem2chunk (oldmem); |
3157 | /* its size */ |
3158 | const INTERNAL_SIZE_T oldsize = chunksize (oldp); |
3159 | |
3160 | if (chunk_is_mmapped (oldp)) |
3161 | ar_ptr = NULL; |
3162 | else |
3163 | { |
3164 | MAYBE_INIT_TCACHE (); |
3165 | ar_ptr = arena_for_chunk (oldp); |
3166 | } |
3167 | |
3168 | /* Little security check which won't hurt performance: the allocator |
3169 | never wrapps around at the end of the address space. Therefore |
3170 | we can exclude some size values which might appear here by |
3171 | accident or by "design" from some intruder. We need to bypass |
3172 | this check for dumped fake mmap chunks from the old main arena |
3173 | because the new malloc may provide additional alignment. */ |
3174 | if ((__builtin_expect ((uintptr_t) oldp > (uintptr_t) -oldsize, 0) |
3175 | || __builtin_expect (misaligned_chunk (oldp), 0)) |
3176 | && !DUMPED_MAIN_ARENA_CHUNK (oldp)) |
3177 | malloc_printerr ("realloc(): invalid pointer" ); |
3178 | |
3179 | checked_request2size (bytes, nb); |
3180 | |
3181 | if (chunk_is_mmapped (oldp)) |
3182 | { |
3183 | /* If this is a faked mmapped chunk from the dumped main arena, |
3184 | always make a copy (and do not free the old chunk). */ |
3185 | if (DUMPED_MAIN_ARENA_CHUNK (oldp)) |
3186 | { |
3187 | /* Must alloc, copy, free. */ |
3188 | void *newmem = __libc_malloc (bytes); |
3189 | if (newmem == 0) |
3190 | return NULL; |
3191 | /* Copy as many bytes as are available from the old chunk |
3192 | and fit into the new size. NB: The overhead for faked |
3193 | mmapped chunks is only SIZE_SZ, not 2 * SIZE_SZ as for |
3194 | regular mmapped chunks. */ |
3195 | if (bytes > oldsize - SIZE_SZ) |
3196 | bytes = oldsize - SIZE_SZ; |
3197 | memcpy (newmem, oldmem, bytes); |
3198 | return newmem; |
3199 | } |
3200 | |
3201 | void *newmem; |
3202 | |
3203 | #if HAVE_MREMAP |
3204 | newp = mremap_chunk (oldp, nb); |
3205 | if (newp) |
3206 | return chunk2mem (newp); |
3207 | #endif |
3208 | /* Note the extra SIZE_SZ overhead. */ |
3209 | if (oldsize - SIZE_SZ >= nb) |
3210 | return oldmem; /* do nothing */ |
3211 | |
3212 | /* Must alloc, copy, free. */ |
3213 | newmem = __libc_malloc (bytes); |
3214 | if (newmem == 0) |
3215 | return 0; /* propagate failure */ |
3216 | |
3217 | memcpy (newmem, oldmem, oldsize - 2 * SIZE_SZ); |
3218 | munmap_chunk (oldp); |
3219 | return newmem; |
3220 | } |
3221 | |
3222 | if (SINGLE_THREAD_P) |
3223 | { |
3224 | newp = _int_realloc (ar_ptr, oldp, oldsize, nb); |
3225 | assert (!newp || chunk_is_mmapped (mem2chunk (newp)) || |
3226 | ar_ptr == arena_for_chunk (mem2chunk (newp))); |
3227 | |
3228 | return newp; |
3229 | } |
3230 | |
3231 | __libc_lock_lock (ar_ptr->mutex); |
3232 | |
3233 | newp = _int_realloc (ar_ptr, oldp, oldsize, nb); |
3234 | |
3235 | __libc_lock_unlock (ar_ptr->mutex); |
3236 | assert (!newp || chunk_is_mmapped (mem2chunk (newp)) || |
3237 | ar_ptr == arena_for_chunk (mem2chunk (newp))); |
3238 | |
3239 | if (newp == NULL) |
3240 | { |
3241 | /* Try harder to allocate memory in other arenas. */ |
3242 | LIBC_PROBE (memory_realloc_retry, 2, bytes, oldmem); |
3243 | newp = __libc_malloc (bytes); |
3244 | if (newp != NULL) |
3245 | { |
3246 | memcpy (newp, oldmem, oldsize - SIZE_SZ); |
3247 | _int_free (ar_ptr, oldp, 0); |
3248 | } |
3249 | } |
3250 | |
3251 | return newp; |
3252 | } |
3253 | libc_hidden_def (__libc_realloc) |
3254 | |
3255 | void * |
3256 | __libc_memalign (size_t alignment, size_t bytes) |
3257 | { |
3258 | void *address = RETURN_ADDRESS (0); |
3259 | return _mid_memalign (alignment, bytes, address); |
3260 | } |
3261 | |
3262 | static void * |
3263 | _mid_memalign (size_t alignment, size_t bytes, void *address) |
3264 | { |
3265 | mstate ar_ptr; |
3266 | void *p; |
3267 | |
3268 | void *(*hook) (size_t, size_t, const void *) = |
3269 | atomic_forced_read (__memalign_hook); |
3270 | if (__builtin_expect (hook != NULL, 0)) |
3271 | return (*hook)(alignment, bytes, address); |
3272 | |
3273 | /* If we need less alignment than we give anyway, just relay to malloc. */ |
3274 | if (alignment <= MALLOC_ALIGNMENT) |
3275 | return __libc_malloc (bytes); |
3276 | |
3277 | /* Otherwise, ensure that it is at least a minimum chunk size */ |
3278 | if (alignment < MINSIZE) |
3279 | alignment = MINSIZE; |
3280 | |
3281 | /* If the alignment is greater than SIZE_MAX / 2 + 1 it cannot be a |
3282 | power of 2 and will cause overflow in the check below. */ |
3283 | if (alignment > SIZE_MAX / 2 + 1) |
3284 | { |
3285 | __set_errno (EINVAL); |
3286 | return 0; |
3287 | } |
3288 | |
3289 | /* Check for overflow. */ |
3290 | if (bytes > SIZE_MAX - alignment - MINSIZE) |
3291 | { |
3292 | __set_errno (ENOMEM); |
3293 | return 0; |
3294 | } |
3295 | |
3296 | |
3297 | /* Make sure alignment is power of 2. */ |
3298 | if (!powerof2 (alignment)) |
3299 | { |
3300 | size_t a = MALLOC_ALIGNMENT * 2; |
3301 | while (a < alignment) |
3302 | a <<= 1; |
3303 | alignment = a; |
3304 | } |
3305 | |
3306 | if (SINGLE_THREAD_P) |
3307 | { |
3308 | p = _int_memalign (&main_arena, alignment, bytes); |
3309 | assert (!p || chunk_is_mmapped (mem2chunk (p)) || |
3310 | &main_arena == arena_for_chunk (mem2chunk (p))); |
3311 | |
3312 | return p; |
3313 | } |
3314 | |
3315 | arena_get (ar_ptr, bytes + alignment + MINSIZE); |
3316 | |
3317 | p = _int_memalign (ar_ptr, alignment, bytes); |
3318 | if (!p && ar_ptr != NULL) |
3319 | { |
3320 | LIBC_PROBE (memory_memalign_retry, 2, bytes, alignment); |
3321 | ar_ptr = arena_get_retry (ar_ptr, bytes); |
3322 | p = _int_memalign (ar_ptr, alignment, bytes); |
3323 | } |
3324 | |
3325 | if (ar_ptr != NULL) |
3326 | __libc_lock_unlock (ar_ptr->mutex); |
3327 | |
3328 | assert (!p || chunk_is_mmapped (mem2chunk (p)) || |
3329 | ar_ptr == arena_for_chunk (mem2chunk (p))); |
3330 | return p; |
3331 | } |
3332 | /* For ISO C11. */ |
3333 | weak_alias (__libc_memalign, aligned_alloc) |
3334 | libc_hidden_def (__libc_memalign) |
3335 | |
3336 | void * |
3337 | __libc_valloc (size_t bytes) |
3338 | { |
3339 | if (__malloc_initialized < 0) |
3340 | ptmalloc_init (); |
3341 | |
3342 | void *address = RETURN_ADDRESS (0); |
3343 | size_t pagesize = GLRO (dl_pagesize); |
3344 | return _mid_memalign (pagesize, bytes, address); |
3345 | } |
3346 | |
3347 | void * |
3348 | __libc_pvalloc (size_t bytes) |
3349 | { |
3350 | if (__malloc_initialized < 0) |
3351 | ptmalloc_init (); |
3352 | |
3353 | void *address = RETURN_ADDRESS (0); |
3354 | size_t pagesize = GLRO (dl_pagesize); |
3355 | size_t rounded_bytes = ALIGN_UP (bytes, pagesize); |
3356 | |
3357 | /* Check for overflow. */ |
3358 | if (bytes > SIZE_MAX - 2 * pagesize - MINSIZE) |
3359 | { |
3360 | __set_errno (ENOMEM); |
3361 | return 0; |
3362 | } |
3363 | |
3364 | return _mid_memalign (pagesize, rounded_bytes, address); |
3365 | } |
3366 | |
3367 | void * |
3368 | __libc_calloc (size_t n, size_t elem_size) |
3369 | { |
3370 | mstate av; |
3371 | mchunkptr oldtop, p; |
3372 | INTERNAL_SIZE_T bytes, sz, csz, oldtopsize; |
3373 | void *mem; |
3374 | unsigned long clearsize; |
3375 | unsigned long nclears; |
3376 | INTERNAL_SIZE_T *d; |
3377 | |
3378 | /* size_t is unsigned so the behavior on overflow is defined. */ |
3379 | bytes = n * elem_size; |
3380 | #define HALF_INTERNAL_SIZE_T \ |
3381 | (((INTERNAL_SIZE_T) 1) << (8 * sizeof (INTERNAL_SIZE_T) / 2)) |
3382 | if (__builtin_expect ((n | elem_size) >= HALF_INTERNAL_SIZE_T, 0)) |
3383 | { |
3384 | if (elem_size != 0 && bytes / elem_size != n) |
3385 | { |
3386 | __set_errno (ENOMEM); |
3387 | return 0; |
3388 | } |
3389 | } |
3390 | |
3391 | void *(*hook) (size_t, const void *) = |
3392 | atomic_forced_read (__malloc_hook); |
3393 | if (__builtin_expect (hook != NULL, 0)) |
3394 | { |
3395 | sz = bytes; |
3396 | mem = (*hook)(sz, RETURN_ADDRESS (0)); |
3397 | if (mem == 0) |
3398 | return 0; |
3399 | |
3400 | return memset (mem, 0, sz); |
3401 | } |
3402 | |
3403 | sz = bytes; |
3404 | |
3405 | MAYBE_INIT_TCACHE (); |
3406 | |
3407 | if (SINGLE_THREAD_P) |
3408 | av = &main_arena; |
3409 | else |
3410 | arena_get (av, sz); |
3411 | |
3412 | if (av) |
3413 | { |
3414 | /* Check if we hand out the top chunk, in which case there may be no |
3415 | need to clear. */ |
3416 | #if MORECORE_CLEARS |
3417 | oldtop = top (av); |
3418 | oldtopsize = chunksize (top (av)); |
3419 | # if MORECORE_CLEARS < 2 |
3420 | /* Only newly allocated memory is guaranteed to be cleared. */ |
3421 | if (av == &main_arena && |
3422 | oldtopsize < mp_.sbrk_base + av->max_system_mem - (char *) oldtop) |
3423 | oldtopsize = (mp_.sbrk_base + av->max_system_mem - (char *) oldtop); |
3424 | # endif |
3425 | if (av != &main_arena) |
3426 | { |
3427 | heap_info *heap = heap_for_ptr (oldtop); |
3428 | if (oldtopsize < (char *) heap + heap->mprotect_size - (char *) oldtop) |
3429 | oldtopsize = (char *) heap + heap->mprotect_size - (char *) oldtop; |
3430 | } |
3431 | #endif |
3432 | } |
3433 | else |
3434 | { |
3435 | /* No usable arenas. */ |
3436 | oldtop = 0; |
3437 | oldtopsize = 0; |
3438 | } |
3439 | mem = _int_malloc (av, sz); |
3440 | |
3441 | assert (!mem || chunk_is_mmapped (mem2chunk (mem)) || |
3442 | av == arena_for_chunk (mem2chunk (mem))); |
3443 | |
3444 | if (!SINGLE_THREAD_P) |
3445 | { |
3446 | if (mem == 0 && av != NULL) |
3447 | { |
3448 | LIBC_PROBE (memory_calloc_retry, 1, sz); |
3449 | av = arena_get_retry (av, sz); |
3450 | mem = _int_malloc (av, sz); |
3451 | } |
3452 | |
3453 | if (av != NULL) |
3454 | __libc_lock_unlock (av->mutex); |
3455 | } |
3456 | |
3457 | /* Allocation failed even after a retry. */ |
3458 | if (mem == 0) |
3459 | return 0; |
3460 | |
3461 | p = mem2chunk (mem); |
3462 | |
3463 | /* Two optional cases in which clearing not necessary */ |
3464 | if (chunk_is_mmapped (p)) |
3465 | { |
3466 | if (__builtin_expect (perturb_byte, 0)) |
3467 | return memset (mem, 0, sz); |
3468 | |
3469 | return mem; |
3470 | } |
3471 | |
3472 | csz = chunksize (p); |
3473 | |
3474 | #if MORECORE_CLEARS |
3475 | if (perturb_byte == 0 && (p == oldtop && csz > oldtopsize)) |
3476 | { |
3477 | /* clear only the bytes from non-freshly-sbrked memory */ |
3478 | csz = oldtopsize; |
3479 | } |
3480 | #endif |
3481 | |
3482 | /* Unroll clear of <= 36 bytes (72 if 8byte sizes). We know that |
3483 | contents have an odd number of INTERNAL_SIZE_T-sized words; |
3484 | minimally 3. */ |
3485 | d = (INTERNAL_SIZE_T *) mem; |
3486 | clearsize = csz - SIZE_SZ; |
3487 | nclears = clearsize / sizeof (INTERNAL_SIZE_T); |
3488 | assert (nclears >= 3); |
3489 | |
3490 | if (nclears > 9) |
3491 | return memset (d, 0, clearsize); |
3492 | |
3493 | else |
3494 | { |
3495 | *(d + 0) = 0; |
3496 | *(d + 1) = 0; |
3497 | *(d + 2) = 0; |
3498 | if (nclears > 4) |
3499 | { |
3500 | *(d + 3) = 0; |
3501 | *(d + 4) = 0; |
3502 | if (nclears > 6) |
3503 | { |
3504 | *(d + 5) = 0; |
3505 | *(d + 6) = 0; |
3506 | if (nclears > 8) |
3507 | { |
3508 | *(d + 7) = 0; |
3509 | *(d + 8) = 0; |
3510 | } |
3511 | } |
3512 | } |
3513 | } |
3514 | |
3515 | return mem; |
3516 | } |
3517 | |
3518 | /* |
3519 | ------------------------------ malloc ------------------------------ |
3520 | */ |
3521 | |
3522 | static void * |
3523 | _int_malloc (mstate av, size_t bytes) |
3524 | { |
3525 | INTERNAL_SIZE_T nb; /* normalized request size */ |
3526 | unsigned int idx; /* associated bin index */ |
3527 | mbinptr bin; /* associated bin */ |
3528 | |
3529 | mchunkptr victim; /* inspected/selected chunk */ |
3530 | INTERNAL_SIZE_T size; /* its size */ |
3531 | int victim_index; /* its bin index */ |
3532 | |
3533 | mchunkptr remainder; /* remainder from a split */ |
3534 | unsigned long remainder_size; /* its size */ |
3535 | |
3536 | unsigned int block; /* bit map traverser */ |
3537 | unsigned int bit; /* bit map traverser */ |
3538 | unsigned int map; /* current word of binmap */ |
3539 | |
3540 | mchunkptr fwd; /* misc temp for linking */ |
3541 | mchunkptr bck; /* misc temp for linking */ |
3542 | |
3543 | #if USE_TCACHE |
3544 | size_t tcache_unsorted_count; /* count of unsorted chunks processed */ |
3545 | #endif |
3546 | |
3547 | /* |
3548 | Convert request size to internal form by adding SIZE_SZ bytes |
3549 | overhead plus possibly more to obtain necessary alignment and/or |
3550 | to obtain a size of at least MINSIZE, the smallest allocatable |
3551 | size. Also, checked_request2size traps (returning 0) request sizes |
3552 | that are so large that they wrap around zero when padded and |
3553 | aligned. |
3554 | */ |
3555 | |
3556 | checked_request2size (bytes, nb); |
3557 | |
3558 | /* There are no usable arenas. Fall back to sysmalloc to get a chunk from |
3559 | mmap. */ |
3560 | if (__glibc_unlikely (av == NULL)) |
3561 | { |
3562 | void *p = sysmalloc (nb, av); |
3563 | if (p != NULL) |
3564 | alloc_perturb (p, bytes); |
3565 | return p; |
3566 | } |
3567 | |
3568 | /* |
3569 | If the size qualifies as a fastbin, first check corresponding bin. |
3570 | This code is safe to execute even if av is not yet initialized, so we |
3571 | can try it without checking, which saves some time on this fast path. |
3572 | */ |
3573 | |
3574 | #define REMOVE_FB(fb, victim, pp) \ |
3575 | do \ |
3576 | { \ |
3577 | victim = pp; \ |
3578 | if (victim == NULL) \ |
3579 | break; \ |
3580 | } \ |
3581 | while ((pp = catomic_compare_and_exchange_val_acq (fb, victim->fd, victim)) \ |
3582 | != victim); \ |
3583 | |
3584 | if ((unsigned long) (nb) <= (unsigned long) (get_max_fast ())) |
3585 | { |
3586 | idx = fastbin_index (nb); |
3587 | mfastbinptr *fb = &fastbin (av, idx); |
3588 | mchunkptr pp; |
3589 | victim = *fb; |
3590 | |
3591 | if (victim != NULL) |
3592 | { |
3593 | if (SINGLE_THREAD_P) |
3594 | *fb = victim->fd; |
3595 | else |
3596 | REMOVE_FB (fb, pp, victim); |
3597 | if (__glibc_likely (victim != NULL)) |
3598 | { |
3599 | size_t victim_idx = fastbin_index (chunksize (victim)); |
3600 | if (__builtin_expect (victim_idx != idx, 0)) |
3601 | malloc_printerr ("malloc(): memory corruption (fast)" ); |
3602 | check_remalloced_chunk (av, victim, nb); |
3603 | #if USE_TCACHE |
3604 | /* While we're here, if we see other chunks of the same size, |
3605 | stash them in the tcache. */ |
3606 | size_t tc_idx = csize2tidx (nb); |
3607 | if (tcache && tc_idx < mp_.tcache_bins) |
3608 | { |
3609 | mchunkptr tc_victim; |
3610 | |
3611 | /* While bin not empty and tcache not full, copy chunks. */ |
3612 | while (tcache->counts[tc_idx] < mp_.tcache_count |
3613 | && (tc_victim = *fb) != NULL) |
3614 | { |
3615 | if (SINGLE_THREAD_P) |
3616 | *fb = tc_victim->fd; |
3617 | else |
3618 | { |
3619 | REMOVE_FB (fb, pp, tc_victim); |
3620 | if (__glibc_unlikely (tc_victim == NULL)) |
3621 | break; |
3622 | } |
3623 | tcache_put (tc_victim, tc_idx); |
3624 | } |
3625 | } |
3626 | #endif |
3627 | void *p = chunk2mem (victim); |
3628 | alloc_perturb (p, bytes); |
3629 | return p; |
3630 | } |
3631 | } |
3632 | } |
3633 | |
3634 | /* |
3635 | If a small request, check regular bin. Since these "smallbins" |
3636 | hold one size each, no searching within bins is necessary. |
3637 | (For a large request, we need to wait until unsorted chunks are |
3638 | processed to find best fit. But for small ones, fits are exact |
3639 | anyway, so we can check now, which is faster.) |
3640 | */ |
3641 | |
3642 | if (in_smallbin_range (nb)) |
3643 | { |
3644 | idx = smallbin_index (nb); |
3645 | bin = bin_at (av, idx); |
3646 | |
3647 | if ((victim = last (bin)) != bin) |
3648 | { |
3649 | if (victim == 0) /* initialization check */ |
3650 | malloc_consolidate (av); |
3651 | else |
3652 | { |
3653 | bck = victim->bk; |
3654 | if (__glibc_unlikely (bck->fd != victim)) |
3655 | malloc_printerr |
3656 | ("malloc(): smallbin double linked list corrupted" ); |
3657 | set_inuse_bit_at_offset (victim, nb); |
3658 | bin->bk = bck; |
3659 | bck->fd = bin; |
3660 | |
3661 | if (av != &main_arena) |
3662 | set_non_main_arena (victim); |
3663 | check_malloced_chunk (av, victim, nb); |
3664 | #if USE_TCACHE |
3665 | /* While we're here, if we see other chunks of the same size, |
3666 | stash them in the tcache. */ |
3667 | size_t tc_idx = csize2tidx (nb); |
3668 | if (tcache && tc_idx < mp_.tcache_bins) |
3669 | { |
3670 | mchunkptr tc_victim; |
3671 | |
3672 | /* While bin not empty and tcache not full, copy chunks over. */ |
3673 | while (tcache->counts[tc_idx] < mp_.tcache_count |
3674 | && (tc_victim = last (bin)) != bin) |
3675 | { |
3676 | if (tc_victim != 0) |
3677 | { |
3678 | bck = tc_victim->bk; |
3679 | set_inuse_bit_at_offset (tc_victim, nb); |
3680 | if (av != &main_arena) |
3681 | set_non_main_arena (tc_victim); |
3682 | bin->bk = bck; |
3683 | bck->fd = bin; |
3684 | |
3685 | tcache_put (tc_victim, tc_idx); |
3686 | } |
3687 | } |
3688 | } |
3689 | #endif |
3690 | void *p = chunk2mem (victim); |
3691 | alloc_perturb (p, bytes); |
3692 | return p; |
3693 | } |
3694 | } |
3695 | } |
3696 | |
3697 | /* |
3698 | If this is a large request, consolidate fastbins before continuing. |
3699 | While it might look excessive to kill all fastbins before |
3700 | even seeing if there is space available, this avoids |
3701 | fragmentation problems normally associated with fastbins. |
3702 | Also, in practice, programs tend to have runs of either small or |
3703 | large requests, but less often mixtures, so consolidation is not |
3704 | invoked all that often in most programs. And the programs that |
3705 | it is called frequently in otherwise tend to fragment. |
3706 | */ |
3707 | |
3708 | else |
3709 | { |
3710 | idx = largebin_index (nb); |
3711 | if (atomic_load_relaxed (&av->have_fastchunks)) |
3712 | malloc_consolidate (av); |
3713 | } |
3714 | |
3715 | /* |
3716 | Process recently freed or remaindered chunks, taking one only if |
3717 | it is exact fit, or, if this a small request, the chunk is remainder from |
3718 | the most recent non-exact fit. Place other traversed chunks in |
3719 | bins. Note that this step is the only place in any routine where |
3720 | chunks are placed in bins. |
3721 | |
3722 | The outer loop here is needed because we might not realize until |
3723 | near the end of malloc that we should have consolidated, so must |
3724 | do so and retry. This happens at most once, and only when we would |
3725 | otherwise need to expand memory to service a "small" request. |
3726 | */ |
3727 | |
3728 | #if USE_TCACHE |
3729 | INTERNAL_SIZE_T tcache_nb = 0; |
3730 | size_t tc_idx = csize2tidx (nb); |
3731 | if (tcache && tc_idx < mp_.tcache_bins) |
3732 | tcache_nb = nb; |
3733 | int return_cached = 0; |
3734 | |
3735 | tcache_unsorted_count = 0; |
3736 | #endif |
3737 | |
3738 | for (;; ) |
3739 | { |
3740 | int iters = 0; |
3741 | while ((victim = unsorted_chunks (av)->bk) != unsorted_chunks (av)) |
3742 | { |
3743 | bck = victim->bk; |
3744 | if (__builtin_expect (chunksize_nomask (victim) <= 2 * SIZE_SZ, 0) |
3745 | || __builtin_expect (chunksize_nomask (victim) |
3746 | > av->system_mem, 0)) |
3747 | malloc_printerr ("malloc(): memory corruption" ); |
3748 | size = chunksize (victim); |
3749 | |
3750 | /* |
3751 | If a small request, try to use last remainder if it is the |
3752 | only chunk in unsorted bin. This helps promote locality for |
3753 | runs of consecutive small requests. This is the only |
3754 | exception to best-fit, and applies only when there is |
3755 | no exact fit for a small chunk. |
3756 | */ |
3757 | |
3758 | if (in_smallbin_range (nb) && |
3759 | bck == unsorted_chunks (av) && |
3760 | victim == av->last_remainder && |
3761 | (unsigned long) (size) > (unsigned long) (nb + MINSIZE)) |
3762 | { |
3763 | /* split and reattach remainder */ |
3764 | remainder_size = size - nb; |
3765 | remainder = chunk_at_offset (victim, nb); |
3766 | unsorted_chunks (av)->bk = unsorted_chunks (av)->fd = remainder; |
3767 | av->last_remainder = remainder; |
3768 | remainder->bk = remainder->fd = unsorted_chunks (av); |
3769 | if (!in_smallbin_range (remainder_size)) |
3770 | { |
3771 | remainder->fd_nextsize = NULL; |
3772 | remainder->bk_nextsize = NULL; |
3773 | } |
3774 | |
3775 | set_head (victim, nb | PREV_INUSE | |
3776 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
3777 | set_head (remainder, remainder_size | PREV_INUSE); |
3778 | set_foot (remainder, remainder_size); |
3779 | |
3780 | check_malloced_chunk (av, victim, nb); |
3781 | void *p = chunk2mem (victim); |
3782 | alloc_perturb (p, bytes); |
3783 | return p; |
3784 | } |
3785 | |
3786 | /* remove from unsorted list */ |
3787 | unsorted_chunks (av)->bk = bck; |
3788 | bck->fd = unsorted_chunks (av); |
3789 | |
3790 | /* Take now instead of binning if exact fit */ |
3791 | |
3792 | if (size == nb) |
3793 | { |
3794 | set_inuse_bit_at_offset (victim, size); |
3795 | if (av != &main_arena) |
3796 | set_non_main_arena (victim); |
3797 | #if USE_TCACHE |
3798 | /* Fill cache first, return to user only if cache fills. |
3799 | We may return one of these chunks later. */ |
3800 | if (tcache_nb |
3801 | && tcache->counts[tc_idx] < mp_.tcache_count) |
3802 | { |
3803 | tcache_put (victim, tc_idx); |
3804 | return_cached = 1; |
3805 | continue; |
3806 | } |
3807 | else |
3808 | { |
3809 | #endif |
3810 | check_malloced_chunk (av, victim, nb); |
3811 | void *p = chunk2mem (victim); |
3812 | alloc_perturb (p, bytes); |
3813 | return p; |
3814 | #if USE_TCACHE |
3815 | } |
3816 | #endif |
3817 | } |
3818 | |
3819 | /* place chunk in bin */ |
3820 | |
3821 | if (in_smallbin_range (size)) |
3822 | { |
3823 | victim_index = smallbin_index (size); |
3824 | bck = bin_at (av, victim_index); |
3825 | fwd = bck->fd; |
3826 | } |
3827 | else |
3828 | { |
3829 | victim_index = largebin_index (size); |
3830 | bck = bin_at (av, victim_index); |
3831 | fwd = bck->fd; |
3832 | |
3833 | /* maintain large bins in sorted order */ |
3834 | if (fwd != bck) |
3835 | { |
3836 | /* Or with inuse bit to speed comparisons */ |
3837 | size |= PREV_INUSE; |
3838 | /* if smaller than smallest, bypass loop below */ |
3839 | assert (chunk_main_arena (bck->bk)); |
3840 | if ((unsigned long) (size) |
3841 | < (unsigned long) chunksize_nomask (bck->bk)) |
3842 | { |
3843 | fwd = bck; |
3844 | bck = bck->bk; |
3845 | |
3846 | victim->fd_nextsize = fwd->fd; |
3847 | victim->bk_nextsize = fwd->fd->bk_nextsize; |
3848 | fwd->fd->bk_nextsize = victim->bk_nextsize->fd_nextsize = victim; |
3849 | } |
3850 | else |
3851 | { |
3852 | assert (chunk_main_arena (fwd)); |
3853 | while ((unsigned long) size < chunksize_nomask (fwd)) |
3854 | { |
3855 | fwd = fwd->fd_nextsize; |
3856 | assert (chunk_main_arena (fwd)); |
3857 | } |
3858 | |
3859 | if ((unsigned long) size |
3860 | == (unsigned long) chunksize_nomask (fwd)) |
3861 | /* Always insert in the second position. */ |
3862 | fwd = fwd->fd; |
3863 | else |
3864 | { |
3865 | victim->fd_nextsize = fwd; |
3866 | victim->bk_nextsize = fwd->bk_nextsize; |
3867 | fwd->bk_nextsize = victim; |
3868 | victim->bk_nextsize->fd_nextsize = victim; |
3869 | } |
3870 | bck = fwd->bk; |
3871 | } |
3872 | } |
3873 | else |
3874 | victim->fd_nextsize = victim->bk_nextsize = victim; |
3875 | } |
3876 | |
3877 | mark_bin (av, victim_index); |
3878 | victim->bk = bck; |
3879 | victim->fd = fwd; |
3880 | fwd->bk = victim; |
3881 | bck->fd = victim; |
3882 | |
3883 | #if USE_TCACHE |
3884 | /* If we've processed as many chunks as we're allowed while |
3885 | filling the cache, return one of the cached ones. */ |
3886 | ++tcache_unsorted_count; |
3887 | if (return_cached |
3888 | && mp_.tcache_unsorted_limit > 0 |
3889 | && tcache_unsorted_count > mp_.tcache_unsorted_limit) |
3890 | { |
3891 | return tcache_get (tc_idx); |
3892 | } |
3893 | #endif |
3894 | |
3895 | #define MAX_ITERS 10000 |
3896 | if (++iters >= MAX_ITERS) |
3897 | break; |
3898 | } |
3899 | |
3900 | #if USE_TCACHE |
3901 | /* If all the small chunks we found ended up cached, return one now. */ |
3902 | if (return_cached) |
3903 | { |
3904 | return tcache_get (tc_idx); |
3905 | } |
3906 | #endif |
3907 | |
3908 | /* |
3909 | If a large request, scan through the chunks of current bin in |
3910 | sorted order to find smallest that fits. Use the skip list for this. |
3911 | */ |
3912 | |
3913 | if (!in_smallbin_range (nb)) |
3914 | { |
3915 | bin = bin_at (av, idx); |
3916 | |
3917 | /* skip scan if empty or largest chunk is too small */ |
3918 | if ((victim = first (bin)) != bin |
3919 | && (unsigned long) chunksize_nomask (victim) |
3920 | >= (unsigned long) (nb)) |
3921 | { |
3922 | victim = victim->bk_nextsize; |
3923 | while (((unsigned long) (size = chunksize (victim)) < |
3924 | (unsigned long) (nb))) |
3925 | victim = victim->bk_nextsize; |
3926 | |
3927 | /* Avoid removing the first entry for a size so that the skip |
3928 | list does not have to be rerouted. */ |
3929 | if (victim != last (bin) |
3930 | && chunksize_nomask (victim) |
3931 | == chunksize_nomask (victim->fd)) |
3932 | victim = victim->fd; |
3933 | |
3934 | remainder_size = size - nb; |
3935 | unlink (av, victim, bck, fwd); |
3936 | |
3937 | /* Exhaust */ |
3938 | if (remainder_size < MINSIZE) |
3939 | { |
3940 | set_inuse_bit_at_offset (victim, size); |
3941 | if (av != &main_arena) |
3942 | set_non_main_arena (victim); |
3943 | } |
3944 | /* Split */ |
3945 | else |
3946 | { |
3947 | remainder = chunk_at_offset (victim, nb); |
3948 | /* We cannot assume the unsorted list is empty and therefore |
3949 | have to perform a complete insert here. */ |
3950 | bck = unsorted_chunks (av); |
3951 | fwd = bck->fd; |
3952 | if (__glibc_unlikely (fwd->bk != bck)) |
3953 | malloc_printerr ("malloc(): corrupted unsorted chunks" ); |
3954 | remainder->bk = bck; |
3955 | remainder->fd = fwd; |
3956 | bck->fd = remainder; |
3957 | fwd->bk = remainder; |
3958 | if (!in_smallbin_range (remainder_size)) |
3959 | { |
3960 | remainder->fd_nextsize = NULL; |
3961 | remainder->bk_nextsize = NULL; |
3962 | } |
3963 | set_head (victim, nb | PREV_INUSE | |
3964 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
3965 | set_head (remainder, remainder_size | PREV_INUSE); |
3966 | set_foot (remainder, remainder_size); |
3967 | } |
3968 | check_malloced_chunk (av, victim, nb); |
3969 | void *p = chunk2mem (victim); |
3970 | alloc_perturb (p, bytes); |
3971 | return p; |
3972 | } |
3973 | } |
3974 | |
3975 | /* |
3976 | Search for a chunk by scanning bins, starting with next largest |
3977 | bin. This search is strictly by best-fit; i.e., the smallest |
3978 | (with ties going to approximately the least recently used) chunk |
3979 | that fits is selected. |
3980 | |
3981 | The bitmap avoids needing to check that most blocks are nonempty. |
3982 | The particular case of skipping all bins during warm-up phases |
3983 | when no chunks have been returned yet is faster than it might look. |
3984 | */ |
3985 | |
3986 | ++idx; |
3987 | bin = bin_at (av, idx); |
3988 | block = idx2block (idx); |
3989 | map = av->binmap[block]; |
3990 | bit = idx2bit (idx); |
3991 | |
3992 | for (;; ) |
3993 | { |
3994 | /* Skip rest of block if there are no more set bits in this block. */ |
3995 | if (bit > map || bit == 0) |
3996 | { |
3997 | do |
3998 | { |
3999 | if (++block >= BINMAPSIZE) /* out of bins */ |
4000 | goto use_top; |
4001 | } |
4002 | while ((map = av->binmap[block]) == 0); |
4003 | |
4004 | bin = bin_at (av, (block << BINMAPSHIFT)); |
4005 | bit = 1; |
4006 | } |
4007 | |
4008 | /* Advance to bin with set bit. There must be one. */ |
4009 | while ((bit & map) == 0) |
4010 | { |
4011 | bin = next_bin (bin); |
4012 | bit <<= 1; |
4013 | assert (bit != 0); |
4014 | } |
4015 | |
4016 | /* Inspect the bin. It is likely to be non-empty */ |
4017 | victim = last (bin); |
4018 | |
4019 | /* If a false alarm (empty bin), clear the bit. */ |
4020 | if (victim == bin) |
4021 | { |
4022 | av->binmap[block] = map &= ~bit; /* Write through */ |
4023 | bin = next_bin (bin); |
4024 | bit <<= 1; |
4025 | } |
4026 | |
4027 | else |
4028 | { |
4029 | size = chunksize (victim); |
4030 | |
4031 | /* We know the first chunk in this bin is big enough to use. */ |
4032 | assert ((unsigned long) (size) >= (unsigned long) (nb)); |
4033 | |
4034 | remainder_size = size - nb; |
4035 | |
4036 | /* unlink */ |
4037 | unlink (av, victim, bck, fwd); |
4038 | |
4039 | /* Exhaust */ |
4040 | if (remainder_size < MINSIZE) |
4041 | { |
4042 | set_inuse_bit_at_offset (victim, size); |
4043 | if (av != &main_arena) |
4044 | set_non_main_arena (victim); |
4045 | } |
4046 | |
4047 | /* Split */ |
4048 | else |
4049 | { |
4050 | remainder = chunk_at_offset (victim, nb); |
4051 | |
4052 | /* We cannot assume the unsorted list is empty and therefore |
4053 | have to perform a complete insert here. */ |
4054 | bck = unsorted_chunks (av); |
4055 | fwd = bck->fd; |
4056 | if (__glibc_unlikely (fwd->bk != bck)) |
4057 | malloc_printerr ("malloc(): corrupted unsorted chunks 2" ); |
4058 | remainder->bk = bck; |
4059 | remainder->fd = fwd; |
4060 | bck->fd = remainder; |
4061 | fwd->bk = remainder; |
4062 | |
4063 | /* advertise as last remainder */ |
4064 | if (in_smallbin_range (nb)) |
4065 | av->last_remainder = remainder; |
4066 | if (!in_smallbin_range (remainder_size)) |
4067 | { |
4068 | remainder->fd_nextsize = NULL; |
4069 | remainder->bk_nextsize = NULL; |
4070 | } |
4071 | set_head (victim, nb | PREV_INUSE | |
4072 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4073 | set_head (remainder, remainder_size | PREV_INUSE); |
4074 | set_foot (remainder, remainder_size); |
4075 | } |
4076 | check_malloced_chunk (av, victim, nb); |
4077 | void *p = chunk2mem (victim); |
4078 | alloc_perturb (p, bytes); |
4079 | return p; |
4080 | } |
4081 | } |
4082 | |
4083 | use_top: |
4084 | /* |
4085 | If large enough, split off the chunk bordering the end of memory |
4086 | (held in av->top). Note that this is in accord with the best-fit |
4087 | search rule. In effect, av->top is treated as larger (and thus |
4088 | less well fitting) than any other available chunk since it can |
4089 | be extended to be as large as necessary (up to system |
4090 | limitations). |
4091 | |
4092 | We require that av->top always exists (i.e., has size >= |
4093 | MINSIZE) after initialization, so if it would otherwise be |
4094 | exhausted by current request, it is replenished. (The main |
4095 | reason for ensuring it exists is that we may need MINSIZE space |
4096 | to put in fenceposts in sysmalloc.) |
4097 | */ |
4098 | |
4099 | victim = av->top; |
4100 | size = chunksize (victim); |
4101 | |
4102 | if ((unsigned long) (size) >= (unsigned long) (nb + MINSIZE)) |
4103 | { |
4104 | remainder_size = size - nb; |
4105 | remainder = chunk_at_offset (victim, nb); |
4106 | av->top = remainder; |
4107 | set_head (victim, nb | PREV_INUSE | |
4108 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4109 | set_head (remainder, remainder_size | PREV_INUSE); |
4110 | |
4111 | check_malloced_chunk (av, victim, nb); |
4112 | void *p = chunk2mem (victim); |
4113 | alloc_perturb (p, bytes); |
4114 | return p; |
4115 | } |
4116 | |
4117 | /* When we are using atomic ops to free fast chunks we can get |
4118 | here for all block sizes. */ |
4119 | else if (atomic_load_relaxed (&av->have_fastchunks)) |
4120 | { |
4121 | malloc_consolidate (av); |
4122 | /* restore original bin index */ |
4123 | if (in_smallbin_range (nb)) |
4124 | idx = smallbin_index (nb); |
4125 | else |
4126 | idx = largebin_index (nb); |
4127 | } |
4128 | |
4129 | /* |
4130 | Otherwise, relay to handle system-dependent cases |
4131 | */ |
4132 | else |
4133 | { |
4134 | void *p = sysmalloc (nb, av); |
4135 | if (p != NULL) |
4136 | alloc_perturb (p, bytes); |
4137 | return p; |
4138 | } |
4139 | } |
4140 | } |
4141 | |
4142 | /* |
4143 | ------------------------------ free ------------------------------ |
4144 | */ |
4145 | |
4146 | static void |
4147 | _int_free (mstate av, mchunkptr p, int have_lock) |
4148 | { |
4149 | INTERNAL_SIZE_T size; /* its size */ |
4150 | mfastbinptr *fb; /* associated fastbin */ |
4151 | mchunkptr nextchunk; /* next contiguous chunk */ |
4152 | INTERNAL_SIZE_T nextsize; /* its size */ |
4153 | int nextinuse; /* true if nextchunk is used */ |
4154 | INTERNAL_SIZE_T prevsize; /* size of previous contiguous chunk */ |
4155 | mchunkptr bck; /* misc temp for linking */ |
4156 | mchunkptr fwd; /* misc temp for linking */ |
4157 | |
4158 | size = chunksize (p); |
4159 | |
4160 | /* Little security check which won't hurt performance: the |
4161 | allocator never wrapps around at the end of the address space. |
4162 | Therefore we can exclude some size values which might appear |
4163 | here by accident or by "design" from some intruder. */ |
4164 | if (__builtin_expect ((uintptr_t) p > (uintptr_t) -size, 0) |
4165 | || __builtin_expect (misaligned_chunk (p), 0)) |
4166 | malloc_printerr ("free(): invalid pointer" ); |
4167 | /* We know that each chunk is at least MINSIZE bytes in size or a |
4168 | multiple of MALLOC_ALIGNMENT. */ |
4169 | if (__glibc_unlikely (size < MINSIZE || !aligned_OK (size))) |
4170 | malloc_printerr ("free(): invalid size" ); |
4171 | |
4172 | check_inuse_chunk(av, p); |
4173 | |
4174 | #if USE_TCACHE |
4175 | { |
4176 | size_t tc_idx = csize2tidx (size); |
4177 | |
4178 | if (tcache |
4179 | && tc_idx < mp_.tcache_bins |
4180 | && tcache->counts[tc_idx] < mp_.tcache_count) |
4181 | { |
4182 | tcache_put (p, tc_idx); |
4183 | return; |
4184 | } |
4185 | } |
4186 | #endif |
4187 | |
4188 | /* |
4189 | If eligible, place chunk on a fastbin so it can be found |
4190 | and used quickly in malloc. |
4191 | */ |
4192 | |
4193 | if ((unsigned long)(size) <= (unsigned long)(get_max_fast ()) |
4194 | |
4195 | #if TRIM_FASTBINS |
4196 | /* |
4197 | If TRIM_FASTBINS set, don't place chunks |
4198 | bordering top into fastbins |
4199 | */ |
4200 | && (chunk_at_offset(p, size) != av->top) |
4201 | #endif |
4202 | ) { |
4203 | |
4204 | if (__builtin_expect (chunksize_nomask (chunk_at_offset (p, size)) |
4205 | <= 2 * SIZE_SZ, 0) |
4206 | || __builtin_expect (chunksize (chunk_at_offset (p, size)) |
4207 | >= av->system_mem, 0)) |
4208 | { |
4209 | bool fail = true; |
4210 | /* We might not have a lock at this point and concurrent modifications |
4211 | of system_mem might result in a false positive. Redo the test after |
4212 | getting the lock. */ |
4213 | if (!have_lock) |
4214 | { |
4215 | __libc_lock_lock (av->mutex); |
4216 | fail = (chunksize_nomask (chunk_at_offset (p, size)) <= 2 * SIZE_SZ |
4217 | || chunksize (chunk_at_offset (p, size)) >= av->system_mem); |
4218 | __libc_lock_unlock (av->mutex); |
4219 | } |
4220 | |
4221 | if (fail) |
4222 | malloc_printerr ("free(): invalid next size (fast)" ); |
4223 | } |
4224 | |
4225 | free_perturb (chunk2mem(p), size - 2 * SIZE_SZ); |
4226 | |
4227 | atomic_store_relaxed (&av->have_fastchunks, true); |
4228 | unsigned int idx = fastbin_index(size); |
4229 | fb = &fastbin (av, idx); |
4230 | |
4231 | /* Atomically link P to its fastbin: P->FD = *FB; *FB = P; */ |
4232 | mchunkptr old = *fb, old2; |
4233 | |
4234 | if (SINGLE_THREAD_P) |
4235 | { |
4236 | /* Check that the top of the bin is not the record we are going to |
4237 | add (i.e., double free). */ |
4238 | if (__builtin_expect (old == p, 0)) |
4239 | malloc_printerr ("double free or corruption (fasttop)" ); |
4240 | p->fd = old; |
4241 | *fb = p; |
4242 | } |
4243 | else |
4244 | do |
4245 | { |
4246 | /* Check that the top of the bin is not the record we are going to |
4247 | add (i.e., double free). */ |
4248 | if (__builtin_expect (old == p, 0)) |
4249 | malloc_printerr ("double free or corruption (fasttop)" ); |
4250 | p->fd = old2 = old; |
4251 | } |
4252 | while ((old = catomic_compare_and_exchange_val_rel (fb, p, old2)) |
4253 | != old2); |
4254 | |
4255 | /* Check that size of fastbin chunk at the top is the same as |
4256 | size of the chunk that we are adding. We can dereference OLD |
4257 | only if we have the lock, otherwise it might have already been |
4258 | allocated again. */ |
4259 | if (have_lock && old != NULL |
4260 | && __builtin_expect (fastbin_index (chunksize (old)) != idx, 0)) |
4261 | malloc_printerr ("invalid fastbin entry (free)" ); |
4262 | } |
4263 | |
4264 | /* |
4265 | Consolidate other non-mmapped chunks as they arrive. |
4266 | */ |
4267 | |
4268 | else if (!chunk_is_mmapped(p)) { |
4269 | |
4270 | /* If we're single-threaded, don't lock the arena. */ |
4271 | if (SINGLE_THREAD_P) |
4272 | have_lock = true; |
4273 | |
4274 | if (!have_lock) |
4275 | __libc_lock_lock (av->mutex); |
4276 | |
4277 | nextchunk = chunk_at_offset(p, size); |
4278 | |
4279 | /* Lightweight tests: check whether the block is already the |
4280 | top block. */ |
4281 | if (__glibc_unlikely (p == av->top)) |
4282 | malloc_printerr ("double free or corruption (top)" ); |
4283 | /* Or whether the next chunk is beyond the boundaries of the arena. */ |
4284 | if (__builtin_expect (contiguous (av) |
4285 | && (char *) nextchunk |
4286 | >= ((char *) av->top + chunksize(av->top)), 0)) |
4287 | malloc_printerr ("double free or corruption (out)" ); |
4288 | /* Or whether the block is actually not marked used. */ |
4289 | if (__glibc_unlikely (!prev_inuse(nextchunk))) |
4290 | malloc_printerr ("double free or corruption (!prev)" ); |
4291 | |
4292 | nextsize = chunksize(nextchunk); |
4293 | if (__builtin_expect (chunksize_nomask (nextchunk) <= 2 * SIZE_SZ, 0) |
4294 | || __builtin_expect (nextsize >= av->system_mem, 0)) |
4295 | malloc_printerr ("free(): invalid next size (normal)" ); |
4296 | |
4297 | free_perturb (chunk2mem(p), size - 2 * SIZE_SZ); |
4298 | |
4299 | /* consolidate backward */ |
4300 | if (!prev_inuse(p)) { |
4301 | prevsize = prev_size (p); |
4302 | size += prevsize; |
4303 | p = chunk_at_offset(p, -((long) prevsize)); |
4304 | unlink(av, p, bck, fwd); |
4305 | } |
4306 | |
4307 | if (nextchunk != av->top) { |
4308 | /* get and clear inuse bit */ |
4309 | nextinuse = inuse_bit_at_offset(nextchunk, nextsize); |
4310 | |
4311 | /* consolidate forward */ |
4312 | if (!nextinuse) { |
4313 | unlink(av, nextchunk, bck, fwd); |
4314 | size += nextsize; |
4315 | } else |
4316 | clear_inuse_bit_at_offset(nextchunk, 0); |
4317 | |
4318 | /* |
4319 | Place the chunk in unsorted chunk list. Chunks are |
4320 | not placed into regular bins until after they have |
4321 | been given one chance to be used in malloc. |
4322 | */ |
4323 | |
4324 | bck = unsorted_chunks(av); |
4325 | fwd = bck->fd; |
4326 | if (__glibc_unlikely (fwd->bk != bck)) |
4327 | malloc_printerr ("free(): corrupted unsorted chunks" ); |
4328 | p->fd = fwd; |
4329 | p->bk = bck; |
4330 | if (!in_smallbin_range(size)) |
4331 | { |
4332 | p->fd_nextsize = NULL; |
4333 | p->bk_nextsize = NULL; |
4334 | } |
4335 | bck->fd = p; |
4336 | fwd->bk = p; |
4337 | |
4338 | set_head(p, size | PREV_INUSE); |
4339 | set_foot(p, size); |
4340 | |
4341 | check_free_chunk(av, p); |
4342 | } |
4343 | |
4344 | /* |
4345 | If the chunk borders the current high end of memory, |
4346 | consolidate into top |
4347 | */ |
4348 | |
4349 | else { |
4350 | size += nextsize; |
4351 | set_head(p, size | PREV_INUSE); |
4352 | av->top = p; |
4353 | check_chunk(av, p); |
4354 | } |
4355 | |
4356 | /* |
4357 | If freeing a large space, consolidate possibly-surrounding |
4358 | chunks. Then, if the total unused topmost memory exceeds trim |
4359 | threshold, ask malloc_trim to reduce top. |
4360 | |
4361 | Unless max_fast is 0, we don't know if there are fastbins |
4362 | bordering top, so we cannot tell for sure whether threshold |
4363 | has been reached unless fastbins are consolidated. But we |
4364 | don't want to consolidate on each free. As a compromise, |
4365 | consolidation is performed if FASTBIN_CONSOLIDATION_THRESHOLD |
4366 | is reached. |
4367 | */ |
4368 | |
4369 | if ((unsigned long)(size) >= FASTBIN_CONSOLIDATION_THRESHOLD) { |
4370 | if (atomic_load_relaxed (&av->have_fastchunks)) |
4371 | malloc_consolidate(av); |
4372 | |
4373 | if (av == &main_arena) { |
4374 | #ifndef MORECORE_CANNOT_TRIM |
4375 | if ((unsigned long)(chunksize(av->top)) >= |
4376 | (unsigned long)(mp_.trim_threshold)) |
4377 | systrim(mp_.top_pad, av); |
4378 | #endif |
4379 | } else { |
4380 | /* Always try heap_trim(), even if the top chunk is not |
4381 | large, because the corresponding heap might go away. */ |
4382 | heap_info *heap = heap_for_ptr(top(av)); |
4383 | |
4384 | assert(heap->ar_ptr == av); |
4385 | heap_trim(heap, mp_.top_pad); |
4386 | } |
4387 | } |
4388 | |
4389 | if (!have_lock) |
4390 | __libc_lock_unlock (av->mutex); |
4391 | } |
4392 | /* |
4393 | If the chunk was allocated via mmap, release via munmap(). |
4394 | */ |
4395 | |
4396 | else { |
4397 | munmap_chunk (p); |
4398 | } |
4399 | } |
4400 | |
4401 | /* |
4402 | ------------------------- malloc_consolidate ------------------------- |
4403 | |
4404 | malloc_consolidate is a specialized version of free() that tears |
4405 | down chunks held in fastbins. Free itself cannot be used for this |
4406 | purpose since, among other things, it might place chunks back onto |
4407 | fastbins. So, instead, we need to use a minor variant of the same |
4408 | code. |
4409 | |
4410 | Also, because this routine needs to be called the first time through |
4411 | malloc anyway, it turns out to be the perfect place to trigger |
4412 | initialization code. |
4413 | */ |
4414 | |
4415 | static void malloc_consolidate(mstate av) |
4416 | { |
4417 | mfastbinptr* fb; /* current fastbin being consolidated */ |
4418 | mfastbinptr* maxfb; /* last fastbin (for loop control) */ |
4419 | mchunkptr p; /* current chunk being consolidated */ |
4420 | mchunkptr nextp; /* next chunk to consolidate */ |
4421 | mchunkptr unsorted_bin; /* bin header */ |
4422 | mchunkptr first_unsorted; /* chunk to link to */ |
4423 | |
4424 | /* These have same use as in free() */ |
4425 | mchunkptr nextchunk; |
4426 | INTERNAL_SIZE_T size; |
4427 | INTERNAL_SIZE_T nextsize; |
4428 | INTERNAL_SIZE_T prevsize; |
4429 | int nextinuse; |
4430 | mchunkptr bck; |
4431 | mchunkptr fwd; |
4432 | |
4433 | /* |
4434 | If max_fast is 0, we know that av hasn't |
4435 | yet been initialized, in which case do so below |
4436 | */ |
4437 | |
4438 | if (get_max_fast () != 0) { |
4439 | atomic_store_relaxed (&av->have_fastchunks, false); |
4440 | |
4441 | unsorted_bin = unsorted_chunks(av); |
4442 | |
4443 | /* |
4444 | Remove each chunk from fast bin and consolidate it, placing it |
4445 | then in unsorted bin. Among other reasons for doing this, |
4446 | placing in unsorted bin avoids needing to calculate actual bins |
4447 | until malloc is sure that chunks aren't immediately going to be |
4448 | reused anyway. |
4449 | */ |
4450 | |
4451 | maxfb = &fastbin (av, NFASTBINS - 1); |
4452 | fb = &fastbin (av, 0); |
4453 | do { |
4454 | p = atomic_exchange_acq (fb, NULL); |
4455 | if (p != 0) { |
4456 | do { |
4457 | check_inuse_chunk(av, p); |
4458 | nextp = p->fd; |
4459 | |
4460 | /* Slightly streamlined version of consolidation code in free() */ |
4461 | size = chunksize (p); |
4462 | nextchunk = chunk_at_offset(p, size); |
4463 | nextsize = chunksize(nextchunk); |
4464 | |
4465 | if (!prev_inuse(p)) { |
4466 | prevsize = prev_size (p); |
4467 | size += prevsize; |
4468 | p = chunk_at_offset(p, -((long) prevsize)); |
4469 | unlink(av, p, bck, fwd); |
4470 | } |
4471 | |
4472 | if (nextchunk != av->top) { |
4473 | nextinuse = inuse_bit_at_offset(nextchunk, nextsize); |
4474 | |
4475 | if (!nextinuse) { |
4476 | size += nextsize; |
4477 | unlink(av, nextchunk, bck, fwd); |
4478 | } else |
4479 | clear_inuse_bit_at_offset(nextchunk, 0); |
4480 | |
4481 | first_unsorted = unsorted_bin->fd; |
4482 | unsorted_bin->fd = p; |
4483 | first_unsorted->bk = p; |
4484 | |
4485 | if (!in_smallbin_range (size)) { |
4486 | p->fd_nextsize = NULL; |
4487 | p->bk_nextsize = NULL; |
4488 | } |
4489 | |
4490 | set_head(p, size | PREV_INUSE); |
4491 | p->bk = unsorted_bin; |
4492 | p->fd = first_unsorted; |
4493 | set_foot(p, size); |
4494 | } |
4495 | |
4496 | else { |
4497 | size += nextsize; |
4498 | set_head(p, size | PREV_INUSE); |
4499 | av->top = p; |
4500 | } |
4501 | |
4502 | } while ( (p = nextp) != 0); |
4503 | |
4504 | } |
4505 | } while (fb++ != maxfb); |
4506 | } |
4507 | else { |
4508 | malloc_init_state(av); |
4509 | check_malloc_state(av); |
4510 | } |
4511 | } |
4512 | |
4513 | /* |
4514 | ------------------------------ realloc ------------------------------ |
4515 | */ |
4516 | |
4517 | void* |
4518 | _int_realloc(mstate av, mchunkptr oldp, INTERNAL_SIZE_T oldsize, |
4519 | INTERNAL_SIZE_T nb) |
4520 | { |
4521 | mchunkptr newp; /* chunk to return */ |
4522 | INTERNAL_SIZE_T newsize; /* its size */ |
4523 | void* newmem; /* corresponding user mem */ |
4524 | |
4525 | mchunkptr next; /* next contiguous chunk after oldp */ |
4526 | |
4527 | mchunkptr remainder; /* extra space at end of newp */ |
4528 | unsigned long remainder_size; /* its size */ |
4529 | |
4530 | mchunkptr bck; /* misc temp for linking */ |
4531 | mchunkptr fwd; /* misc temp for linking */ |
4532 | |
4533 | /* oldmem size */ |
4534 | if (__builtin_expect (chunksize_nomask (oldp) <= 2 * SIZE_SZ, 0) |
4535 | || __builtin_expect (oldsize >= av->system_mem, 0)) |
4536 | malloc_printerr ("realloc(): invalid old size" ); |
4537 | |
4538 | check_inuse_chunk (av, oldp); |
4539 | |
4540 | /* All callers already filter out mmap'ed chunks. */ |
4541 | assert (!chunk_is_mmapped (oldp)); |
4542 | |
4543 | next = chunk_at_offset (oldp, oldsize); |
4544 | INTERNAL_SIZE_T nextsize = chunksize (next); |
4545 | if (__builtin_expect (chunksize_nomask (next) <= 2 * SIZE_SZ, 0) |
4546 | || __builtin_expect (nextsize >= av->system_mem, 0)) |
4547 | malloc_printerr ("realloc(): invalid next size" ); |
4548 | |
4549 | if ((unsigned long) (oldsize) >= (unsigned long) (nb)) |
4550 | { |
4551 | /* already big enough; split below */ |
4552 | newp = oldp; |
4553 | newsize = oldsize; |
4554 | } |
4555 | |
4556 | else |
4557 | { |
4558 | /* Try to expand forward into top */ |
4559 | if (next == av->top && |
4560 | (unsigned long) (newsize = oldsize + nextsize) >= |
4561 | (unsigned long) (nb + MINSIZE)) |
4562 | { |
4563 | set_head_size (oldp, nb | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4564 | av->top = chunk_at_offset (oldp, nb); |
4565 | set_head (av->top, (newsize - nb) | PREV_INUSE); |
4566 | check_inuse_chunk (av, oldp); |
4567 | return chunk2mem (oldp); |
4568 | } |
4569 | |
4570 | /* Try to expand forward into next chunk; split off remainder below */ |
4571 | else if (next != av->top && |
4572 | !inuse (next) && |
4573 | (unsigned long) (newsize = oldsize + nextsize) >= |
4574 | (unsigned long) (nb)) |
4575 | { |
4576 | newp = oldp; |
4577 | unlink (av, next, bck, fwd); |
4578 | } |
4579 | |
4580 | /* allocate, copy, free */ |
4581 | else |
4582 | { |
4583 | newmem = _int_malloc (av, nb - MALLOC_ALIGN_MASK); |
4584 | if (newmem == 0) |
4585 | return 0; /* propagate failure */ |
4586 | |
4587 | newp = mem2chunk (newmem); |
4588 | newsize = chunksize (newp); |
4589 | |
4590 | /* |
4591 | Avoid copy if newp is next chunk after oldp. |
4592 | */ |
4593 | if (newp == next) |
4594 | { |
4595 | newsize += oldsize; |
4596 | newp = oldp; |
4597 | } |
4598 | else |
4599 | { |
4600 | memcpy (newmem, chunk2mem (oldp), oldsize - SIZE_SZ); |
4601 | _int_free (av, oldp, 1); |
4602 | check_inuse_chunk (av, newp); |
4603 | return chunk2mem (newp); |
4604 | } |
4605 | } |
4606 | } |
4607 | |
4608 | /* If possible, free extra space in old or extended chunk */ |
4609 | |
4610 | assert ((unsigned long) (newsize) >= (unsigned long) (nb)); |
4611 | |
4612 | remainder_size = newsize - nb; |
4613 | |
4614 | if (remainder_size < MINSIZE) /* not enough extra to split off */ |
4615 | { |
4616 | set_head_size (newp, newsize | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4617 | set_inuse_bit_at_offset (newp, newsize); |
4618 | } |
4619 | else /* split remainder */ |
4620 | { |
4621 | remainder = chunk_at_offset (newp, nb); |
4622 | set_head_size (newp, nb | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4623 | set_head (remainder, remainder_size | PREV_INUSE | |
4624 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4625 | /* Mark remainder as inuse so free() won't complain */ |
4626 | set_inuse_bit_at_offset (remainder, remainder_size); |
4627 | _int_free (av, remainder, 1); |
4628 | } |
4629 | |
4630 | check_inuse_chunk (av, newp); |
4631 | return chunk2mem (newp); |
4632 | } |
4633 | |
4634 | /* |
4635 | ------------------------------ memalign ------------------------------ |
4636 | */ |
4637 | |
4638 | static void * |
4639 | _int_memalign (mstate av, size_t alignment, size_t bytes) |
4640 | { |
4641 | INTERNAL_SIZE_T nb; /* padded request size */ |
4642 | char *m; /* memory returned by malloc call */ |
4643 | mchunkptr p; /* corresponding chunk */ |
4644 | char *brk; /* alignment point within p */ |
4645 | mchunkptr newp; /* chunk to return */ |
4646 | INTERNAL_SIZE_T newsize; /* its size */ |
4647 | INTERNAL_SIZE_T leadsize; /* leading space before alignment point */ |
4648 | mchunkptr remainder; /* spare room at end to split off */ |
4649 | unsigned long remainder_size; /* its size */ |
4650 | INTERNAL_SIZE_T size; |
4651 | |
4652 | |
4653 | |
4654 | checked_request2size (bytes, nb); |
4655 | |
4656 | /* |
4657 | Strategy: find a spot within that chunk that meets the alignment |
4658 | request, and then possibly free the leading and trailing space. |
4659 | */ |
4660 | |
4661 | |
4662 | /* Check for overflow. */ |
4663 | if (nb > SIZE_MAX - alignment - MINSIZE) |
4664 | { |
4665 | __set_errno (ENOMEM); |
4666 | return 0; |
4667 | } |
4668 | |
4669 | /* Call malloc with worst case padding to hit alignment. */ |
4670 | |
4671 | m = (char *) (_int_malloc (av, nb + alignment + MINSIZE)); |
4672 | |
4673 | if (m == 0) |
4674 | return 0; /* propagate failure */ |
4675 | |
4676 | p = mem2chunk (m); |
4677 | |
4678 | if ((((unsigned long) (m)) % alignment) != 0) /* misaligned */ |
4679 | |
4680 | { /* |
4681 | Find an aligned spot inside chunk. Since we need to give back |
4682 | leading space in a chunk of at least MINSIZE, if the first |
4683 | calculation places us at a spot with less than MINSIZE leader, |
4684 | we can move to the next aligned spot -- we've allocated enough |
4685 | total room so that this is always possible. |
4686 | */ |
4687 | brk = (char *) mem2chunk (((unsigned long) (m + alignment - 1)) & |
4688 | - ((signed long) alignment)); |
4689 | if ((unsigned long) (brk - (char *) (p)) < MINSIZE) |
4690 | brk += alignment; |
4691 | |
4692 | newp = (mchunkptr) brk; |
4693 | leadsize = brk - (char *) (p); |
4694 | newsize = chunksize (p) - leadsize; |
4695 | |
4696 | /* For mmapped chunks, just adjust offset */ |
4697 | if (chunk_is_mmapped (p)) |
4698 | { |
4699 | set_prev_size (newp, prev_size (p) + leadsize); |
4700 | set_head (newp, newsize | IS_MMAPPED); |
4701 | return chunk2mem (newp); |
4702 | } |
4703 | |
4704 | /* Otherwise, give back leader, use the rest */ |
4705 | set_head (newp, newsize | PREV_INUSE | |
4706 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4707 | set_inuse_bit_at_offset (newp, newsize); |
4708 | set_head_size (p, leadsize | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4709 | _int_free (av, p, 1); |
4710 | p = newp; |
4711 | |
4712 | assert (newsize >= nb && |
4713 | (((unsigned long) (chunk2mem (p))) % alignment) == 0); |
4714 | } |
4715 | |
4716 | /* Also give back spare room at the end */ |
4717 | if (!chunk_is_mmapped (p)) |
4718 | { |
4719 | size = chunksize (p); |
4720 | if ((unsigned long) (size) > (unsigned long) (nb + MINSIZE)) |
4721 | { |
4722 | remainder_size = size - nb; |
4723 | remainder = chunk_at_offset (p, nb); |
4724 | set_head (remainder, remainder_size | PREV_INUSE | |
4725 | (av != &main_arena ? NON_MAIN_ARENA : 0)); |
4726 | set_head_size (p, nb); |
4727 | _int_free (av, remainder, 1); |
4728 | } |
4729 | } |
4730 | |
4731 | check_inuse_chunk (av, p); |
4732 | return chunk2mem (p); |
4733 | } |
4734 | |
4735 | |
4736 | /* |
4737 | ------------------------------ malloc_trim ------------------------------ |
4738 | */ |
4739 | |
4740 | static int |
4741 | mtrim (mstate av, size_t pad) |
4742 | { |
4743 | /* Ensure initialization/consolidation */ |
4744 | malloc_consolidate (av); |
4745 | |
4746 | const size_t ps = GLRO (dl_pagesize); |
4747 | int psindex = bin_index (ps); |
4748 | const size_t psm1 = ps - 1; |
4749 | |
4750 | int result = 0; |
4751 | for (int i = 1; i < NBINS; ++i) |
4752 | if (i == 1 || i >= psindex) |
4753 | { |
4754 | mbinptr bin = bin_at (av, i); |
4755 | |
4756 | for (mchunkptr p = last (bin); p != bin; p = p->bk) |
4757 | { |
4758 | INTERNAL_SIZE_T size = chunksize (p); |
4759 | |
4760 | if (size > psm1 + sizeof (struct malloc_chunk)) |
4761 | { |
4762 | /* See whether the chunk contains at least one unused page. */ |
4763 | char *paligned_mem = (char *) (((uintptr_t) p |
4764 | + sizeof (struct malloc_chunk) |
4765 | + psm1) & ~psm1); |
4766 | |
4767 | assert ((char *) chunk2mem (p) + 4 * SIZE_SZ <= paligned_mem); |
4768 | assert ((char *) p + size > paligned_mem); |
4769 | |
4770 | /* This is the size we could potentially free. */ |
4771 | size -= paligned_mem - (char *) p; |
4772 | |
4773 | if (size > psm1) |
4774 | { |
4775 | #if MALLOC_DEBUG |
4776 | /* When debugging we simulate destroying the memory |
4777 | content. */ |
4778 | memset (paligned_mem, 0x89, size & ~psm1); |
4779 | #endif |
4780 | __madvise (paligned_mem, size & ~psm1, MADV_DONTNEED); |
4781 | |
4782 | result = 1; |
4783 | } |
4784 | } |
4785 | } |
4786 | } |
4787 | |
4788 | #ifndef MORECORE_CANNOT_TRIM |
4789 | return result | (av == &main_arena ? systrim (pad, av) : 0); |
4790 | |
4791 | #else |
4792 | return result; |
4793 | #endif |
4794 | } |
4795 | |
4796 | |
4797 | int |
4798 | __malloc_trim (size_t s) |
4799 | { |
4800 | int result = 0; |
4801 | |
4802 | if (__malloc_initialized < 0) |
4803 | ptmalloc_init (); |
4804 | |
4805 | mstate ar_ptr = &main_arena; |
4806 | do |
4807 | { |
4808 | __libc_lock_lock (ar_ptr->mutex); |
4809 | result |= mtrim (ar_ptr, s); |
4810 | __libc_lock_unlock (ar_ptr->mutex); |
4811 | |
4812 | ar_ptr = ar_ptr->next; |
4813 | } |
4814 | while (ar_ptr != &main_arena); |
4815 | |
4816 | return result; |
4817 | } |
4818 | |
4819 | |
4820 | /* |
4821 | ------------------------- malloc_usable_size ------------------------- |
4822 | */ |
4823 | |
4824 | static size_t |
4825 | musable (void *mem) |
4826 | { |
4827 | mchunkptr p; |
4828 | if (mem != 0) |
4829 | { |
4830 | p = mem2chunk (mem); |
4831 | |
4832 | if (__builtin_expect (using_malloc_checking == 1, 0)) |
4833 | return malloc_check_get_size (p); |
4834 | |
4835 | if (chunk_is_mmapped (p)) |
4836 | { |
4837 | if (DUMPED_MAIN_ARENA_CHUNK (p)) |
4838 | return chunksize (p) - SIZE_SZ; |
4839 | else |
4840 | return chunksize (p) - 2 * SIZE_SZ; |
4841 | } |
4842 | else if (inuse (p)) |
4843 | return chunksize (p) - SIZE_SZ; |
4844 | } |
4845 | return 0; |
4846 | } |
4847 | |
4848 | |
4849 | size_t |
4850 | __malloc_usable_size (void *m) |
4851 | { |
4852 | size_t result; |
4853 | |
4854 | result = musable (m); |
4855 | return result; |
4856 | } |
4857 | |
4858 | /* |
4859 | ------------------------------ mallinfo ------------------------------ |
4860 | Accumulate malloc statistics for arena AV into M. |
4861 | */ |
4862 | |
4863 | static void |
4864 | int_mallinfo (mstate av, struct mallinfo *m) |
4865 | { |
4866 | size_t i; |
4867 | mbinptr b; |
4868 | mchunkptr p; |
4869 | INTERNAL_SIZE_T avail; |
4870 | INTERNAL_SIZE_T fastavail; |
4871 | int nblocks; |
4872 | int nfastblocks; |
4873 | |
4874 | /* Ensure initialization */ |
4875 | if (av->top == 0) |
4876 | malloc_consolidate (av); |
4877 | |
4878 | check_malloc_state (av); |
4879 | |
4880 | /* Account for top */ |
4881 | avail = chunksize (av->top); |
4882 | nblocks = 1; /* top always exists */ |
4883 | |
4884 | /* traverse fastbins */ |
4885 | nfastblocks = 0; |
4886 | fastavail = 0; |
4887 | |
4888 | for (i = 0; i < NFASTBINS; ++i) |
4889 | { |
4890 | for (p = fastbin (av, i); p != 0; p = p->fd) |
4891 | { |
4892 | ++nfastblocks; |
4893 | fastavail += chunksize (p); |
4894 | } |
4895 | } |
4896 | |
4897 | avail += fastavail; |
4898 | |
4899 | /* traverse regular bins */ |
4900 | for (i = 1; i < NBINS; ++i) |
4901 | { |
4902 | b = bin_at (av, i); |
4903 | for (p = last (b); p != b; p = p->bk) |
4904 | { |
4905 | ++nblocks; |
4906 | avail += chunksize (p); |
4907 | } |
4908 | } |
4909 | |
4910 | m->smblks += nfastblocks; |
4911 | m->ordblks += nblocks; |
4912 | m->fordblks += avail; |
4913 | m->uordblks += av->system_mem - avail; |
4914 | m->arena += av->system_mem; |
4915 | m->fsmblks += fastavail; |
4916 | if (av == &main_arena) |
4917 | { |
4918 | m->hblks = mp_.n_mmaps; |
4919 | m->hblkhd = mp_.mmapped_mem; |
4920 | m->usmblks = 0; |
4921 | m->keepcost = chunksize (av->top); |
4922 | } |
4923 | } |
4924 | |
4925 | |
4926 | struct mallinfo |
4927 | __libc_mallinfo (void) |
4928 | { |
4929 | struct mallinfo m; |
4930 | mstate ar_ptr; |
4931 | |
4932 | if (__malloc_initialized < 0) |
4933 | ptmalloc_init (); |
4934 | |
4935 | memset (&m, 0, sizeof (m)); |
4936 | ar_ptr = &main_arena; |
4937 | do |
4938 | { |
4939 | __libc_lock_lock (ar_ptr->mutex); |
4940 | int_mallinfo (ar_ptr, &m); |
4941 | __libc_lock_unlock (ar_ptr->mutex); |
4942 | |
4943 | ar_ptr = ar_ptr->next; |
4944 | } |
4945 | while (ar_ptr != &main_arena); |
4946 | |
4947 | return m; |
4948 | } |
4949 | |
4950 | /* |
4951 | ------------------------------ malloc_stats ------------------------------ |
4952 | */ |
4953 | |
4954 | void |
4955 | __malloc_stats (void) |
4956 | { |
4957 | int i; |
4958 | mstate ar_ptr; |
4959 | unsigned int in_use_b = mp_.mmapped_mem, system_b = in_use_b; |
4960 | |
4961 | if (__malloc_initialized < 0) |
4962 | ptmalloc_init (); |
4963 | _IO_flockfile (stderr); |
4964 | int old_flags2 = ((_IO_FILE *) stderr)->_flags2; |
4965 | ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; |
4966 | for (i = 0, ar_ptr = &main_arena;; i++) |
4967 | { |
4968 | struct mallinfo mi; |
4969 | |
4970 | memset (&mi, 0, sizeof (mi)); |
4971 | __libc_lock_lock (ar_ptr->mutex); |
4972 | int_mallinfo (ar_ptr, &mi); |
4973 | fprintf (stderr, "Arena %d:\n" , i); |
4974 | fprintf (stderr, "system bytes = %10u\n" , (unsigned int) mi.arena); |
4975 | fprintf (stderr, "in use bytes = %10u\n" , (unsigned int) mi.uordblks); |
4976 | #if MALLOC_DEBUG > 1 |
4977 | if (i > 0) |
4978 | dump_heap (heap_for_ptr (top (ar_ptr))); |
4979 | #endif |
4980 | system_b += mi.arena; |
4981 | in_use_b += mi.uordblks; |
4982 | __libc_lock_unlock (ar_ptr->mutex); |
4983 | ar_ptr = ar_ptr->next; |
4984 | if (ar_ptr == &main_arena) |
4985 | break; |
4986 | } |
4987 | fprintf (stderr, "Total (incl. mmap):\n" ); |
4988 | fprintf (stderr, "system bytes = %10u\n" , system_b); |
4989 | fprintf (stderr, "in use bytes = %10u\n" , in_use_b); |
4990 | fprintf (stderr, "max mmap regions = %10u\n" , (unsigned int) mp_.max_n_mmaps); |
4991 | fprintf (stderr, "max mmap bytes = %10lu\n" , |
4992 | (unsigned long) mp_.max_mmapped_mem); |
4993 | ((_IO_FILE *) stderr)->_flags2 |= old_flags2; |
4994 | _IO_funlockfile (stderr); |
4995 | } |
4996 | |
4997 | |
4998 | /* |
4999 | ------------------------------ mallopt ------------------------------ |
5000 | */ |
5001 | static inline int |
5002 | __always_inline |
5003 | do_set_trim_threshold (size_t value) |
5004 | { |
5005 | LIBC_PROBE (memory_mallopt_trim_threshold, 3, value, mp_.trim_threshold, |
5006 | mp_.no_dyn_threshold); |
5007 | mp_.trim_threshold = value; |
5008 | mp_.no_dyn_threshold = 1; |
5009 | return 1; |
5010 | } |
5011 | |
5012 | static inline int |
5013 | __always_inline |
5014 | do_set_top_pad (size_t value) |
5015 | { |
5016 | LIBC_PROBE (memory_mallopt_top_pad, 3, value, mp_.top_pad, |
5017 | mp_.no_dyn_threshold); |
5018 | mp_.top_pad = value; |
5019 | mp_.no_dyn_threshold = 1; |
5020 | return 1; |
5021 | } |
5022 | |
5023 | static inline int |
5024 | __always_inline |
5025 | do_set_mmap_threshold (size_t value) |
5026 | { |
5027 | /* Forbid setting the threshold too high. */ |
5028 | if (value <= HEAP_MAX_SIZE / 2) |
5029 | { |
5030 | LIBC_PROBE (memory_mallopt_mmap_threshold, 3, value, mp_.mmap_threshold, |
5031 | mp_.no_dyn_threshold); |
5032 | mp_.mmap_threshold = value; |
5033 | mp_.no_dyn_threshold = 1; |
5034 | return 1; |
5035 | } |
5036 | return 0; |
5037 | } |
5038 | |
5039 | static inline int |
5040 | __always_inline |
5041 | do_set_mmaps_max (int32_t value) |
5042 | { |
5043 | LIBC_PROBE (memory_mallopt_mmap_max, 3, value, mp_.n_mmaps_max, |
5044 | mp_.no_dyn_threshold); |
5045 | mp_.n_mmaps_max = value; |
5046 | mp_.no_dyn_threshold = 1; |
5047 | return 1; |
5048 | } |
5049 | |
5050 | static inline int |
5051 | __always_inline |
5052 | do_set_mallopt_check (int32_t value) |
5053 | { |
5054 | return 1; |
5055 | } |
5056 | |
5057 | static inline int |
5058 | __always_inline |
5059 | do_set_perturb_byte (int32_t value) |
5060 | { |
5061 | LIBC_PROBE (memory_mallopt_perturb, 2, value, perturb_byte); |
5062 | perturb_byte = value; |
5063 | return 1; |
5064 | } |
5065 | |
5066 | static inline int |
5067 | __always_inline |
5068 | do_set_arena_test (size_t value) |
5069 | { |
5070 | LIBC_PROBE (memory_mallopt_arena_test, 2, value, mp_.arena_test); |
5071 | mp_.arena_test = value; |
5072 | return 1; |
5073 | } |
5074 | |
5075 | static inline int |
5076 | __always_inline |
5077 | do_set_arena_max (size_t value) |
5078 | { |
5079 | LIBC_PROBE (memory_mallopt_arena_max, 2, value, mp_.arena_max); |
5080 | mp_.arena_max = value; |
5081 | return 1; |
5082 | } |
5083 | |
5084 | #if USE_TCACHE |
5085 | static inline int |
5086 | __always_inline |
5087 | do_set_tcache_max (size_t value) |
5088 | { |
5089 | if (value >= 0 && value <= MAX_TCACHE_SIZE) |
5090 | { |
5091 | LIBC_PROBE (memory_tunable_tcache_max_bytes, 2, value, mp_.tcache_max_bytes); |
5092 | mp_.tcache_max_bytes = value; |
5093 | mp_.tcache_bins = csize2tidx (request2size(value)) + 1; |
5094 | } |
5095 | return 1; |
5096 | } |
5097 | |
5098 | static inline int |
5099 | __always_inline |
5100 | do_set_tcache_count (size_t value) |
5101 | { |
5102 | if (value <= MAX_TCACHE_COUNT) |
5103 | { |
5104 | LIBC_PROBE (memory_tunable_tcache_count, 2, value, mp_.tcache_count); |
5105 | mp_.tcache_count = value; |
5106 | } |
5107 | return 1; |
5108 | } |
5109 | |
5110 | static inline int |
5111 | __always_inline |
5112 | do_set_tcache_unsorted_limit (size_t value) |
5113 | { |
5114 | LIBC_PROBE (memory_tunable_tcache_unsorted_limit, 2, value, mp_.tcache_unsorted_limit); |
5115 | mp_.tcache_unsorted_limit = value; |
5116 | return 1; |
5117 | } |
5118 | #endif |
5119 | |
5120 | int |
5121 | __libc_mallopt (int param_number, int value) |
5122 | { |
5123 | mstate av = &main_arena; |
5124 | int res = 1; |
5125 | |
5126 | if (__malloc_initialized < 0) |
5127 | ptmalloc_init (); |
5128 | __libc_lock_lock (av->mutex); |
5129 | /* Ensure initialization/consolidation */ |
5130 | malloc_consolidate (av); |
5131 | |
5132 | LIBC_PROBE (memory_mallopt, 2, param_number, value); |
5133 | |
5134 | switch (param_number) |
5135 | { |
5136 | case M_MXFAST: |
5137 | if (value >= 0 && value <= MAX_FAST_SIZE) |
5138 | { |
5139 | LIBC_PROBE (memory_mallopt_mxfast, 2, value, get_max_fast ()); |
5140 | set_max_fast (value); |
5141 | } |
5142 | else |
5143 | res = 0; |
5144 | break; |
5145 | |
5146 | case M_TRIM_THRESHOLD: |
5147 | do_set_trim_threshold (value); |
5148 | break; |
5149 | |
5150 | case M_TOP_PAD: |
5151 | do_set_top_pad (value); |
5152 | break; |
5153 | |
5154 | case M_MMAP_THRESHOLD: |
5155 | res = do_set_mmap_threshold (value); |
5156 | break; |
5157 | |
5158 | case M_MMAP_MAX: |
5159 | do_set_mmaps_max (value); |
5160 | break; |
5161 | |
5162 | case M_CHECK_ACTION: |
5163 | do_set_mallopt_check (value); |
5164 | break; |
5165 | |
5166 | case M_PERTURB: |
5167 | do_set_perturb_byte (value); |
5168 | break; |
5169 | |
5170 | case M_ARENA_TEST: |
5171 | if (value > 0) |
5172 | do_set_arena_test (value); |
5173 | break; |
5174 | |
5175 | case M_ARENA_MAX: |
5176 | if (value > 0) |
5177 | do_set_arena_max (value); |
5178 | break; |
5179 | } |
5180 | __libc_lock_unlock (av->mutex); |
5181 | return res; |
5182 | } |
5183 | libc_hidden_def (__libc_mallopt) |
5184 | |
5185 | |
5186 | /* |
5187 | -------------------- Alternative MORECORE functions -------------------- |
5188 | */ |
5189 | |
5190 | |
5191 | /* |
5192 | General Requirements for MORECORE. |
5193 | |
5194 | The MORECORE function must have the following properties: |
5195 | |
5196 | If MORECORE_CONTIGUOUS is false: |
5197 | |
5198 | * MORECORE must allocate in multiples of pagesize. It will |
5199 | only be called with arguments that are multiples of pagesize. |
5200 | |
5201 | * MORECORE(0) must return an address that is at least |
5202 | MALLOC_ALIGNMENT aligned. (Page-aligning always suffices.) |
5203 | |
5204 | else (i.e. If MORECORE_CONTIGUOUS is true): |
5205 | |
5206 | * Consecutive calls to MORECORE with positive arguments |
5207 | return increasing addresses, indicating that space has been |
5208 | contiguously extended. |
5209 | |
5210 | * MORECORE need not allocate in multiples of pagesize. |
5211 | Calls to MORECORE need not have args of multiples of pagesize. |
5212 | |
5213 | * MORECORE need not page-align. |
5214 | |
5215 | In either case: |
5216 | |
5217 | * MORECORE may allocate more memory than requested. (Or even less, |
5218 | but this will generally result in a malloc failure.) |
5219 | |
5220 | * MORECORE must not allocate memory when given argument zero, but |
5221 | instead return one past the end address of memory from previous |
5222 | nonzero call. This malloc does NOT call MORECORE(0) |
5223 | until at least one call with positive arguments is made, so |
5224 | the initial value returned is not important. |
5225 | |
5226 | * Even though consecutive calls to MORECORE need not return contiguous |
5227 | addresses, it must be OK for malloc'ed chunks to span multiple |
5228 | regions in those cases where they do happen to be contiguous. |
5229 | |
5230 | * MORECORE need not handle negative arguments -- it may instead |
5231 | just return MORECORE_FAILURE when given negative arguments. |
5232 | Negative arguments are always multiples of pagesize. MORECORE |
5233 | must not misinterpret negative args as large positive unsigned |
5234 | args. You can suppress all such calls from even occurring by defining |
5235 | MORECORE_CANNOT_TRIM, |
5236 | |
5237 | There is some variation across systems about the type of the |
5238 | argument to sbrk/MORECORE. If size_t is unsigned, then it cannot |
5239 | actually be size_t, because sbrk supports negative args, so it is |
5240 | normally the signed type of the same width as size_t (sometimes |
5241 | declared as "intptr_t", and sometimes "ptrdiff_t"). It doesn't much |
5242 | matter though. Internally, we use "long" as arguments, which should |
5243 | work across all reasonable possibilities. |
5244 | |
5245 | Additionally, if MORECORE ever returns failure for a positive |
5246 | request, then mmap is used as a noncontiguous system allocator. This |
5247 | is a useful backup strategy for systems with holes in address spaces |
5248 | -- in this case sbrk cannot contiguously expand the heap, but mmap |
5249 | may be able to map noncontiguous space. |
5250 | |
5251 | If you'd like mmap to ALWAYS be used, you can define MORECORE to be |
5252 | a function that always returns MORECORE_FAILURE. |
5253 | |
5254 | If you are using this malloc with something other than sbrk (or its |
5255 | emulation) to supply memory regions, you probably want to set |
5256 | MORECORE_CONTIGUOUS as false. As an example, here is a custom |
5257 | allocator kindly contributed for pre-OSX macOS. It uses virtually |
5258 | but not necessarily physically contiguous non-paged memory (locked |
5259 | in, present and won't get swapped out). You can use it by |
5260 | uncommenting this section, adding some #includes, and setting up the |
5261 | appropriate defines above: |
5262 | |
5263 | *#define MORECORE osMoreCore |
5264 | *#define MORECORE_CONTIGUOUS 0 |
5265 | |
5266 | There is also a shutdown routine that should somehow be called for |
5267 | cleanup upon program exit. |
5268 | |
5269 | *#define MAX_POOL_ENTRIES 100 |
5270 | *#define MINIMUM_MORECORE_SIZE (64 * 1024) |
5271 | static int next_os_pool; |
5272 | void *our_os_pools[MAX_POOL_ENTRIES]; |
5273 | |
5274 | void *osMoreCore(int size) |
5275 | { |
5276 | void *ptr = 0; |
5277 | static void *sbrk_top = 0; |
5278 | |
5279 | if (size > 0) |
5280 | { |
5281 | if (size < MINIMUM_MORECORE_SIZE) |
5282 | size = MINIMUM_MORECORE_SIZE; |
5283 | if (CurrentExecutionLevel() == kTaskLevel) |
5284 | ptr = PoolAllocateResident(size + RM_PAGE_SIZE, 0); |
5285 | if (ptr == 0) |
5286 | { |
5287 | return (void *) MORECORE_FAILURE; |
5288 | } |
5289 | // save ptrs so they can be freed during cleanup |
5290 | our_os_pools[next_os_pool] = ptr; |
5291 | next_os_pool++; |
5292 | ptr = (void *) ((((unsigned long) ptr) + RM_PAGE_MASK) & ~RM_PAGE_MASK); |
5293 | sbrk_top = (char *) ptr + size; |
5294 | return ptr; |
5295 | } |
5296 | else if (size < 0) |
5297 | { |
5298 | // we don't currently support shrink behavior |
5299 | return (void *) MORECORE_FAILURE; |
5300 | } |
5301 | else |
5302 | { |
5303 | return sbrk_top; |
5304 | } |
5305 | } |
5306 | |
5307 | // cleanup any allocated memory pools |
5308 | // called as last thing before shutting down driver |
5309 | |
5310 | void osCleanupMem(void) |
5311 | { |
5312 | void **ptr; |
5313 | |
5314 | for (ptr = our_os_pools; ptr < &our_os_pools[MAX_POOL_ENTRIES]; ptr++) |
5315 | if (*ptr) |
5316 | { |
5317 | PoolDeallocate(*ptr); |
5318 | * ptr = 0; |
5319 | } |
5320 | } |
5321 | |
5322 | */ |
5323 | |
5324 | |
5325 | /* Helper code. */ |
5326 | |
5327 | extern char **__libc_argv attribute_hidden; |
5328 | |
5329 | static void |
5330 | malloc_printerr (const char *str) |
5331 | { |
5332 | __libc_message (do_abort, "%s\n" , str); |
5333 | __builtin_unreachable (); |
5334 | } |
5335 | |
5336 | /* We need a wrapper function for one of the additions of POSIX. */ |
5337 | int |
5338 | __posix_memalign (void **memptr, size_t alignment, size_t size) |
5339 | { |
5340 | void *mem; |
5341 | |
5342 | /* Test whether the SIZE argument is valid. It must be a power of |
5343 | two multiple of sizeof (void *). */ |
5344 | if (alignment % sizeof (void *) != 0 |
5345 | || !powerof2 (alignment / sizeof (void *)) |
5346 | || alignment == 0) |
5347 | return EINVAL; |
5348 | |
5349 | |
5350 | void *address = RETURN_ADDRESS (0); |
5351 | mem = _mid_memalign (alignment, size, address); |
5352 | |
5353 | if (mem != NULL) |
5354 | { |
5355 | *memptr = mem; |
5356 | return 0; |
5357 | } |
5358 | |
5359 | return ENOMEM; |
5360 | } |
5361 | weak_alias (__posix_memalign, posix_memalign) |
5362 | |
5363 | |
5364 | int |
5365 | __malloc_info (int options, FILE *fp) |
5366 | { |
5367 | /* For now, at least. */ |
5368 | if (options != 0) |
5369 | return EINVAL; |
5370 | |
5371 | int n = 0; |
5372 | size_t total_nblocks = 0; |
5373 | size_t total_nfastblocks = 0; |
5374 | size_t total_avail = 0; |
5375 | size_t total_fastavail = 0; |
5376 | size_t total_system = 0; |
5377 | size_t total_max_system = 0; |
5378 | size_t total_aspace = 0; |
5379 | size_t total_aspace_mprotect = 0; |
5380 | |
5381 | |
5382 | |
5383 | if (__malloc_initialized < 0) |
5384 | ptmalloc_init (); |
5385 | |
5386 | fputs ("<malloc version=\"1\">\n" , fp); |
5387 | |
5388 | /* Iterate over all arenas currently in use. */ |
5389 | mstate ar_ptr = &main_arena; |
5390 | do |
5391 | { |
5392 | fprintf (fp, "<heap nr=\"%d\">\n<sizes>\n" , n++); |
5393 | |
5394 | size_t nblocks = 0; |
5395 | size_t nfastblocks = 0; |
5396 | size_t avail = 0; |
5397 | size_t fastavail = 0; |
5398 | struct |
5399 | { |
5400 | size_t from; |
5401 | size_t to; |
5402 | size_t total; |
5403 | size_t count; |
5404 | } sizes[NFASTBINS + NBINS - 1]; |
5405 | #define nsizes (sizeof (sizes) / sizeof (sizes[0])) |
5406 | |
5407 | __libc_lock_lock (ar_ptr->mutex); |
5408 | |
5409 | for (size_t i = 0; i < NFASTBINS; ++i) |
5410 | { |
5411 | mchunkptr p = fastbin (ar_ptr, i); |
5412 | if (p != NULL) |
5413 | { |
5414 | size_t nthissize = 0; |
5415 | size_t thissize = chunksize (p); |
5416 | |
5417 | while (p != NULL) |
5418 | { |
5419 | ++nthissize; |
5420 | p = p->fd; |
5421 | } |
5422 | |
5423 | fastavail += nthissize * thissize; |
5424 | nfastblocks += nthissize; |
5425 | sizes[i].from = thissize - (MALLOC_ALIGNMENT - 1); |
5426 | sizes[i].to = thissize; |
5427 | sizes[i].count = nthissize; |
5428 | } |
5429 | else |
5430 | sizes[i].from = sizes[i].to = sizes[i].count = 0; |
5431 | |
5432 | sizes[i].total = sizes[i].count * sizes[i].to; |
5433 | } |
5434 | |
5435 | |
5436 | mbinptr bin; |
5437 | struct malloc_chunk *r; |
5438 | |
5439 | for (size_t i = 1; i < NBINS; ++i) |
5440 | { |
5441 | bin = bin_at (ar_ptr, i); |
5442 | r = bin->fd; |
5443 | sizes[NFASTBINS - 1 + i].from = ~((size_t) 0); |
5444 | sizes[NFASTBINS - 1 + i].to = sizes[NFASTBINS - 1 + i].total |
5445 | = sizes[NFASTBINS - 1 + i].count = 0; |
5446 | |
5447 | if (r != NULL) |
5448 | while (r != bin) |
5449 | { |
5450 | size_t r_size = chunksize_nomask (r); |
5451 | ++sizes[NFASTBINS - 1 + i].count; |
5452 | sizes[NFASTBINS - 1 + i].total += r_size; |
5453 | sizes[NFASTBINS - 1 + i].from |
5454 | = MIN (sizes[NFASTBINS - 1 + i].from, r_size); |
5455 | sizes[NFASTBINS - 1 + i].to = MAX (sizes[NFASTBINS - 1 + i].to, |
5456 | r_size); |
5457 | |
5458 | r = r->fd; |
5459 | } |
5460 | |
5461 | if (sizes[NFASTBINS - 1 + i].count == 0) |
5462 | sizes[NFASTBINS - 1 + i].from = 0; |
5463 | nblocks += sizes[NFASTBINS - 1 + i].count; |
5464 | avail += sizes[NFASTBINS - 1 + i].total; |
5465 | } |
5466 | |
5467 | __libc_lock_unlock (ar_ptr->mutex); |
5468 | |
5469 | total_nfastblocks += nfastblocks; |
5470 | total_fastavail += fastavail; |
5471 | |
5472 | total_nblocks += nblocks; |
5473 | total_avail += avail; |
5474 | |
5475 | for (size_t i = 0; i < nsizes; ++i) |
5476 | if (sizes[i].count != 0 && i != NFASTBINS) |
5477 | fprintf (fp, " \ |
5478 | <size from=\"%zu\" to=\"%zu\" total=\"%zu\" count=\"%zu\"/>\n" , |
5479 | sizes[i].from, sizes[i].to, sizes[i].total, sizes[i].count); |
5480 | |
5481 | if (sizes[NFASTBINS].count != 0) |
5482 | fprintf (fp, "\ |
5483 | <unsorted from=\"%zu\" to=\"%zu\" total=\"%zu\" count=\"%zu\"/>\n" , |
5484 | sizes[NFASTBINS].from, sizes[NFASTBINS].to, |
5485 | sizes[NFASTBINS].total, sizes[NFASTBINS].count); |
5486 | |
5487 | total_system += ar_ptr->system_mem; |
5488 | total_max_system += ar_ptr->max_system_mem; |
5489 | |
5490 | fprintf (fp, |
5491 | "</sizes>\n<total type=\"fast\" count=\"%zu\" size=\"%zu\"/>\n" |
5492 | "<total type=\"rest\" count=\"%zu\" size=\"%zu\"/>\n" |
5493 | "<system type=\"current\" size=\"%zu\"/>\n" |
5494 | "<system type=\"max\" size=\"%zu\"/>\n" , |
5495 | nfastblocks, fastavail, nblocks, avail, |
5496 | ar_ptr->system_mem, ar_ptr->max_system_mem); |
5497 | |
5498 | if (ar_ptr != &main_arena) |
5499 | { |
5500 | heap_info *heap = heap_for_ptr (top (ar_ptr)); |
5501 | fprintf (fp, |
5502 | "<aspace type=\"total\" size=\"%zu\"/>\n" |
5503 | "<aspace type=\"mprotect\" size=\"%zu\"/>\n" , |
5504 | heap->size, heap->mprotect_size); |
5505 | total_aspace += heap->size; |
5506 | total_aspace_mprotect += heap->mprotect_size; |
5507 | } |
5508 | else |
5509 | { |
5510 | fprintf (fp, |
5511 | "<aspace type=\"total\" size=\"%zu\"/>\n" |
5512 | "<aspace type=\"mprotect\" size=\"%zu\"/>\n" , |
5513 | ar_ptr->system_mem, ar_ptr->system_mem); |
5514 | total_aspace += ar_ptr->system_mem; |
5515 | total_aspace_mprotect += ar_ptr->system_mem; |
5516 | } |
5517 | |
5518 | fputs ("</heap>\n" , fp); |
5519 | ar_ptr = ar_ptr->next; |
5520 | } |
5521 | while (ar_ptr != &main_arena); |
5522 | |
5523 | fprintf (fp, |
5524 | "<total type=\"fast\" count=\"%zu\" size=\"%zu\"/>\n" |
5525 | "<total type=\"rest\" count=\"%zu\" size=\"%zu\"/>\n" |
5526 | "<total type=\"mmap\" count=\"%d\" size=\"%zu\"/>\n" |
5527 | "<system type=\"current\" size=\"%zu\"/>\n" |
5528 | "<system type=\"max\" size=\"%zu\"/>\n" |
5529 | "<aspace type=\"total\" size=\"%zu\"/>\n" |
5530 | "<aspace type=\"mprotect\" size=\"%zu\"/>\n" |
5531 | "</malloc>\n" , |
5532 | total_nfastblocks, total_fastavail, total_nblocks, total_avail, |
5533 | mp_.n_mmaps, mp_.mmapped_mem, |
5534 | total_system, total_max_system, |
5535 | total_aspace, total_aspace_mprotect); |
5536 | |
5537 | return 0; |
5538 | } |
5539 | weak_alias (__malloc_info, malloc_info) |
5540 | |
5541 | |
5542 | strong_alias (__libc_calloc, __calloc) weak_alias (__libc_calloc, calloc) |
5543 | strong_alias (__libc_free, __free) strong_alias (__libc_free, free) |
5544 | strong_alias (__libc_malloc, __malloc) strong_alias (__libc_malloc, malloc) |
5545 | strong_alias (__libc_memalign, __memalign) |
5546 | weak_alias (__libc_memalign, memalign) |
5547 | strong_alias (__libc_realloc, __realloc) strong_alias (__libc_realloc, realloc) |
5548 | strong_alias (__libc_valloc, __valloc) weak_alias (__libc_valloc, valloc) |
5549 | strong_alias (__libc_pvalloc, __pvalloc) weak_alias (__libc_pvalloc, pvalloc) |
5550 | strong_alias (__libc_mallinfo, __mallinfo) |
5551 | weak_alias (__libc_mallinfo, mallinfo) |
5552 | strong_alias (__libc_mallopt, __mallopt) weak_alias (__libc_mallopt, mallopt) |
5553 | |
5554 | weak_alias (__malloc_stats, malloc_stats) |
5555 | weak_alias (__malloc_usable_size, malloc_usable_size) |
5556 | weak_alias (__malloc_trim, malloc_trim) |
5557 | |
5558 | #if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_26) |
5559 | compat_symbol (libc, __libc_free, cfree, GLIBC_2_0); |
5560 | #endif |
5561 | |
5562 | /* ------------------------------------------------------------ |
5563 | History: |
5564 | |
5565 | [see ftp://g.oswego.edu/pub/misc/malloc.c for the history of dlmalloc] |
5566 | |
5567 | */ |
5568 | /* |
5569 | * Local variables: |
5570 | * c-basic-offset: 2 |
5571 | * End: |
5572 | */ |
5573 | |