1 /*
2 * default memory allocator for libavutil
3 * Copyright (c) 2002 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg 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 FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * default memory allocator for libavutil
25 */
26
27 #define _XOPEN_SOURCE 600
28
29 #include "config.h"
30
31 #include <limits.h>
32 #include <stdint.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #if HAVE_MALLOC_H
36 #include <malloc.h>
37 #endif
38
39 #include "avassert.h"
40 #include "avutil.h"
41 #include "common.h"
42 #include "dynarray.h"
43 #include "intreadwrite.h"
44 #include "mem.h"
45
46 #ifdef MALLOC_PREFIX
47
48 #define malloc AV_JOIN(MALLOC_PREFIX, malloc)
49 #define memalign AV_JOIN(MALLOC_PREFIX, memalign)
50 #define posix_memalign AV_JOIN(MALLOC_PREFIX, posix_memalign)
51 #define realloc AV_JOIN(MALLOC_PREFIX, realloc)
52 #define free AV_JOIN(MALLOC_PREFIX, free)
53
54 void *malloc(size_t size);
55 void *memalign(size_t align, size_t size);
56 int posix_memalign(void **ptr, size_t align, size_t size);
57 void *realloc(void *ptr, size_t size);
58 void free(void *ptr);
59
60 #endif /* MALLOC_PREFIX */
61
62 #include "mem_internal.h"
63
64 #define ALIGN (HAVE_AVX512 ? 64 : (HAVE_AVX ? 32 : 16))
65
66 /* NOTE: if you want to override these functions with your own
67 * implementations (not recommended) you have to link libav* as
68 * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
69 * Note that this will cost performance. */
70
71 static size_t max_alloc_size= INT_MAX;
72
av_max_alloc(size_t max)73 void av_max_alloc(size_t max){
74 max_alloc_size = max;
75 }
76
av_malloc(size_t size)77 void *av_malloc(size_t size)
78 {
79 void *ptr = NULL;
80
81 if (size > max_alloc_size)
82 return NULL;
83
84 #if HAVE_POSIX_MEMALIGN
85 if (size) //OS X on SDK 10.6 has a broken posix_memalign implementation
86 if (posix_memalign(&ptr, ALIGN, size))
87 ptr = NULL;
88 #elif HAVE_ALIGNED_MALLOC
89 ptr = _aligned_malloc(size, ALIGN);
90 #elif HAVE_MEMALIGN
91 #ifndef __DJGPP__
92 ptr = memalign(ALIGN, size);
93 #else
94 ptr = memalign(size, ALIGN);
95 #endif
96 /* Why 64?
97 * Indeed, we should align it:
98 * on 4 for 386
99 * on 16 for 486
100 * on 32 for 586, PPro - K6-III
101 * on 64 for K7 (maybe for P3 too).
102 * Because L1 and L2 caches are aligned on those values.
103 * But I don't want to code such logic here!
104 */
105 /* Why 32?
106 * For AVX ASM. SSE / NEON needs only 16.
107 * Why not larger? Because I did not see a difference in benchmarks ...
108 */
109 /* benchmarks with P3
110 * memalign(64) + 1 3071, 3051, 3032
111 * memalign(64) + 2 3051, 3032, 3041
112 * memalign(64) + 4 2911, 2896, 2915
113 * memalign(64) + 8 2545, 2554, 2550
114 * memalign(64) + 16 2543, 2572, 2563
115 * memalign(64) + 32 2546, 2545, 2571
116 * memalign(64) + 64 2570, 2533, 2558
117 *
118 * BTW, malloc seems to do 8-byte alignment by default here.
119 */
120 #else
121 ptr = malloc(size);
122 #endif
123 if(!ptr && !size) {
124 size = 1;
125 ptr= av_malloc(1);
126 }
127 #if CONFIG_MEMORY_POISONING
128 if (ptr)
129 memset(ptr, FF_MEMORY_POISON, size);
130 #endif
131 return ptr;
132 }
133
av_realloc(void * ptr,size_t size)134 void *av_realloc(void *ptr, size_t size)
135 {
136 if (size > max_alloc_size)
137 return NULL;
138
139 #if HAVE_ALIGNED_MALLOC
140 return _aligned_realloc(ptr, size + !size, ALIGN);
141 #else
142 return realloc(ptr, size + !size);
143 #endif
144 }
145
av_realloc_f(void * ptr,size_t nelem,size_t elsize)146 void *av_realloc_f(void *ptr, size_t nelem, size_t elsize)
147 {
148 size_t size;
149 void *r;
150
151 if (av_size_mult(elsize, nelem, &size)) {
152 av_free(ptr);
153 return NULL;
154 }
155 r = av_realloc(ptr, size);
156 if (!r)
157 av_free(ptr);
158 return r;
159 }
160
av_reallocp(void * ptr,size_t size)161 int av_reallocp(void *ptr, size_t size)
162 {
163 void *val;
164
165 if (!size) {
166 av_freep(ptr);
167 return 0;
168 }
169
170 memcpy(&val, ptr, sizeof(val));
171 val = av_realloc(val, size);
172
173 if (!val) {
174 av_freep(ptr);
175 return AVERROR(ENOMEM);
176 }
177
178 memcpy(ptr, &val, sizeof(val));
179 return 0;
180 }
181
av_malloc_array(size_t nmemb,size_t size)182 void *av_malloc_array(size_t nmemb, size_t size)
183 {
184 size_t result;
185 if (av_size_mult(nmemb, size, &result) < 0)
186 return NULL;
187 return av_malloc(result);
188 }
189
av_mallocz_array(size_t nmemb,size_t size)190 void *av_mallocz_array(size_t nmemb, size_t size)
191 {
192 size_t result;
193 if (av_size_mult(nmemb, size, &result) < 0)
194 return NULL;
195 return av_mallocz(result);
196 }
197
av_realloc_array(void * ptr,size_t nmemb,size_t size)198 void *av_realloc_array(void *ptr, size_t nmemb, size_t size)
199 {
200 size_t result;
201 if (av_size_mult(nmemb, size, &result) < 0)
202 return NULL;
203 return av_realloc(ptr, result);
204 }
205
av_reallocp_array(void * ptr,size_t nmemb,size_t size)206 int av_reallocp_array(void *ptr, size_t nmemb, size_t size)
207 {
208 void *val;
209
210 memcpy(&val, ptr, sizeof(val));
211 val = av_realloc_f(val, nmemb, size);
212 memcpy(ptr, &val, sizeof(val));
213 if (!val && nmemb && size)
214 return AVERROR(ENOMEM);
215
216 return 0;
217 }
218
av_free(void * ptr)219 void av_free(void *ptr)
220 {
221 #if HAVE_ALIGNED_MALLOC
222 _aligned_free(ptr);
223 #else
224 free(ptr);
225 #endif
226 }
227
av_freep(void * arg)228 void av_freep(void *arg)
229 {
230 void *val;
231
232 memcpy(&val, arg, sizeof(val));
233 memcpy(arg, &(void *){ NULL }, sizeof(val));
234 av_free(val);
235 }
236
av_mallocz(size_t size)237 void *av_mallocz(size_t size)
238 {
239 void *ptr = av_malloc(size);
240 if (ptr)
241 memset(ptr, 0, size);
242 return ptr;
243 }
244
av_calloc(size_t nmemb,size_t size)245 void *av_calloc(size_t nmemb, size_t size)
246 {
247 size_t result;
248 if (av_size_mult(nmemb, size, &result) < 0)
249 return NULL;
250 return av_mallocz(result);
251 }
252
av_strdup(const char * s)253 char *av_strdup(const char *s)
254 {
255 char *ptr = NULL;
256 if (s) {
257 size_t len = strlen(s) + 1;
258 ptr = av_realloc(NULL, len);
259 if (ptr)
260 memcpy(ptr, s, len);
261 }
262 return ptr;
263 }
264
av_strndup(const char * s,size_t len)265 char *av_strndup(const char *s, size_t len)
266 {
267 char *ret = NULL, *end;
268
269 if (!s)
270 return NULL;
271
272 end = memchr(s, 0, len);
273 if (end)
274 len = end - s;
275
276 ret = av_realloc(NULL, len + 1);
277 if (!ret)
278 return NULL;
279
280 memcpy(ret, s, len);
281 ret[len] = 0;
282 return ret;
283 }
284
av_memdup(const void * p,size_t size)285 void *av_memdup(const void *p, size_t size)
286 {
287 void *ptr = NULL;
288 if (p) {
289 ptr = av_malloc(size);
290 if (ptr)
291 memcpy(ptr, p, size);
292 }
293 return ptr;
294 }
295
av_dynarray_add_nofree(void * tab_ptr,int * nb_ptr,void * elem)296 int av_dynarray_add_nofree(void *tab_ptr, int *nb_ptr, void *elem)
297 {
298 void **tab;
299 memcpy(&tab, tab_ptr, sizeof(tab));
300
301 FF_DYNARRAY_ADD(INT_MAX, sizeof(*tab), tab, *nb_ptr, {
302 tab[*nb_ptr] = elem;
303 memcpy(tab_ptr, &tab, sizeof(tab));
304 }, {
305 return AVERROR(ENOMEM);
306 });
307 return 0;
308 }
309
av_dynarray_add(void * tab_ptr,int * nb_ptr,void * elem)310 void av_dynarray_add(void *tab_ptr, int *nb_ptr, void *elem)
311 {
312 void **tab;
313 memcpy(&tab, tab_ptr, sizeof(tab));
314
315 FF_DYNARRAY_ADD(INT_MAX, sizeof(*tab), tab, *nb_ptr, {
316 tab[*nb_ptr] = elem;
317 memcpy(tab_ptr, &tab, sizeof(tab));
318 }, {
319 *nb_ptr = 0;
320 av_freep(tab_ptr);
321 });
322 }
323
av_dynarray2_add(void ** tab_ptr,int * nb_ptr,size_t elem_size,const uint8_t * elem_data)324 void *av_dynarray2_add(void **tab_ptr, int *nb_ptr, size_t elem_size,
325 const uint8_t *elem_data)
326 {
327 uint8_t *tab_elem_data = NULL;
328
329 FF_DYNARRAY_ADD(INT_MAX, elem_size, *tab_ptr, *nb_ptr, {
330 tab_elem_data = (uint8_t *)*tab_ptr + (*nb_ptr) * elem_size;
331 if (elem_data)
332 memcpy(tab_elem_data, elem_data, elem_size);
333 else if (CONFIG_MEMORY_POISONING)
334 memset(tab_elem_data, FF_MEMORY_POISON, elem_size);
335 }, {
336 av_freep(tab_ptr);
337 *nb_ptr = 0;
338 });
339 return tab_elem_data;
340 }
341
fill16(uint8_t * dst,int len)342 static void fill16(uint8_t *dst, int len)
343 {
344 uint32_t v = AV_RN16(dst - 2);
345
346 v |= v << 16;
347
348 while (len >= 4) {
349 AV_WN32(dst, v);
350 dst += 4;
351 len -= 4;
352 }
353
354 while (len--) {
355 *dst = dst[-2];
356 dst++;
357 }
358 }
359
fill24(uint8_t * dst,int len)360 static void fill24(uint8_t *dst, int len)
361 {
362 #if HAVE_BIGENDIAN
363 uint32_t v = AV_RB24(dst - 3);
364 uint32_t a = v << 8 | v >> 16;
365 uint32_t b = v << 16 | v >> 8;
366 uint32_t c = v << 24 | v;
367 #else
368 uint32_t v = AV_RL24(dst - 3);
369 uint32_t a = v | v << 24;
370 uint32_t b = v >> 8 | v << 16;
371 uint32_t c = v >> 16 | v << 8;
372 #endif
373
374 while (len >= 12) {
375 AV_WN32(dst, a);
376 AV_WN32(dst + 4, b);
377 AV_WN32(dst + 8, c);
378 dst += 12;
379 len -= 12;
380 }
381
382 if (len >= 4) {
383 AV_WN32(dst, a);
384 dst += 4;
385 len -= 4;
386 }
387
388 if (len >= 4) {
389 AV_WN32(dst, b);
390 dst += 4;
391 len -= 4;
392 }
393
394 while (len--) {
395 *dst = dst[-3];
396 dst++;
397 }
398 }
399
fill32(uint8_t * dst,int len)400 static void fill32(uint8_t *dst, int len)
401 {
402 uint32_t v = AV_RN32(dst - 4);
403
404 #if HAVE_FAST_64BIT
405 uint64_t v2= v + ((uint64_t)v<<32);
406 while (len >= 32) {
407 AV_WN64(dst , v2);
408 AV_WN64(dst+ 8, v2);
409 AV_WN64(dst+16, v2);
410 AV_WN64(dst+24, v2);
411 dst += 32;
412 len -= 32;
413 }
414 #endif
415
416 while (len >= 4) {
417 AV_WN32(dst, v);
418 dst += 4;
419 len -= 4;
420 }
421
422 while (len--) {
423 *dst = dst[-4];
424 dst++;
425 }
426 }
427
av_memcpy_backptr(uint8_t * dst,int back,int cnt)428 void av_memcpy_backptr(uint8_t *dst, int back, int cnt)
429 {
430 const uint8_t *src = &dst[-back];
431 if (!back)
432 return;
433
434 if (back == 1) {
435 memset(dst, *src, cnt);
436 } else if (back == 2) {
437 fill16(dst, cnt);
438 } else if (back == 3) {
439 fill24(dst, cnt);
440 } else if (back == 4) {
441 fill32(dst, cnt);
442 } else {
443 if (cnt >= 16) {
444 int blocklen = back;
445 while (cnt > blocklen) {
446 memcpy(dst, src, blocklen);
447 dst += blocklen;
448 cnt -= blocklen;
449 blocklen <<= 1;
450 }
451 memcpy(dst, src, cnt);
452 return;
453 }
454 if (cnt >= 8) {
455 AV_COPY32U(dst, src);
456 AV_COPY32U(dst + 4, src + 4);
457 src += 8;
458 dst += 8;
459 cnt -= 8;
460 }
461 if (cnt >= 4) {
462 AV_COPY32U(dst, src);
463 src += 4;
464 dst += 4;
465 cnt -= 4;
466 }
467 if (cnt >= 2) {
468 AV_COPY16U(dst, src);
469 src += 2;
470 dst += 2;
471 cnt -= 2;
472 }
473 if (cnt)
474 *dst = *src;
475 }
476 }
477
av_fast_realloc(void * ptr,unsigned int * size,size_t min_size)478 void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size)
479 {
480 if (min_size <= *size)
481 return ptr;
482
483 if (min_size > max_alloc_size) {
484 *size = 0;
485 return NULL;
486 }
487
488 min_size = FFMIN(max_alloc_size, FFMAX(min_size + min_size / 16 + 32, min_size));
489
490 ptr = av_realloc(ptr, min_size);
491 /* we could set this to the unmodified min_size but this is safer
492 * if the user lost the ptr and uses NULL now
493 */
494 if (!ptr)
495 min_size = 0;
496
497 *size = min_size;
498
499 return ptr;
500 }
501
av_fast_malloc(void * ptr,unsigned int * size,size_t min_size)502 void av_fast_malloc(void *ptr, unsigned int *size, size_t min_size)
503 {
504 ff_fast_malloc(ptr, size, min_size, 0);
505 }
506
av_fast_mallocz(void * ptr,unsigned int * size,size_t min_size)507 void av_fast_mallocz(void *ptr, unsigned int *size, size_t min_size)
508 {
509 ff_fast_malloc(ptr, size, min_size, 1);
510 }
511