Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * dict.c: dictionary of reusable strings, just used to avoid allocation |
3 | | * and freeing operations. |
4 | | * |
5 | | * Copyright (C) 2003-2012 Daniel Veillard. |
6 | | * |
7 | | * Permission to use, copy, modify, and distribute this software for any |
8 | | * purpose with or without fee is hereby granted, provided that the above |
9 | | * copyright notice and this permission notice appear in all copies. |
10 | | * |
11 | | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED |
12 | | * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF |
13 | | * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE AUTHORS AND |
14 | | * CONTRIBUTORS ACCEPT NO RESPONSIBILITY IN ANY CONCEIVABLE MANNER. |
15 | | * |
16 | | * Author: Daniel Veillard |
17 | | */ |
18 | | |
19 | | #define IN_LIBXML |
20 | | #include "libxml.h" |
21 | | |
22 | | #include <errno.h> |
23 | | #include <limits.h> |
24 | | #include <stdlib.h> |
25 | | #include <string.h> |
26 | | |
27 | | #include "private/dict.h" |
28 | | #include "private/error.h" |
29 | | #include "private/globals.h" |
30 | | #include "private/threads.h" |
31 | | |
32 | | #include <libxml/parser.h> |
33 | | #include <libxml/dict.h> |
34 | | #include <libxml/xmlmemory.h> |
35 | | #include <libxml/xmlstring.h> |
36 | | |
37 | | #ifndef SIZE_MAX |
38 | | #define SIZE_MAX ((size_t) -1) |
39 | | #endif |
40 | | |
41 | 9.40k | #define MAX_FILL_NUM 7 |
42 | 9.40k | #define MAX_FILL_DENOM 8 |
43 | 411 | #define MIN_HASH_SIZE 8 |
44 | 2.71M | #define MAX_HASH_SIZE (1u << 31) |
45 | | |
46 | | typedef struct _xmlDictStrings xmlDictStrings; |
47 | | typedef xmlDictStrings *xmlDictStringsPtr; |
48 | | struct _xmlDictStrings { |
49 | | xmlDictStringsPtr next; |
50 | | xmlChar *free; |
51 | | xmlChar *end; |
52 | | size_t size; |
53 | | size_t nbStrings; |
54 | | xmlChar array[1]; |
55 | | }; |
56 | | |
57 | | typedef xmlHashedString xmlDictEntry; |
58 | | |
59 | | /* |
60 | | * The entire dictionary |
61 | | */ |
62 | | struct _xmlDict { |
63 | | int ref_counter; |
64 | | |
65 | | xmlDictEntry *table; |
66 | | size_t size; |
67 | | unsigned int nbElems; |
68 | | xmlDictStringsPtr strings; |
69 | | |
70 | | struct _xmlDict *subdict; |
71 | | /* used for randomization */ |
72 | | unsigned seed; |
73 | | /* used to impose a limit on size */ |
74 | | size_t limit; |
75 | | }; |
76 | | |
77 | | /* |
78 | | * A mutex for modifying the reference counter for shared |
79 | | * dictionaries. |
80 | | */ |
81 | | static xmlMutex xmlDictMutex; |
82 | | |
83 | | /** |
84 | | * @deprecated Alias for #xmlInitParser. |
85 | | * |
86 | | * @returns 0. |
87 | | */ |
88 | | int |
89 | 0 | xmlInitializeDict(void) { |
90 | 0 | xmlInitParser(); |
91 | 0 | return(0); |
92 | 0 | } |
93 | | |
94 | | /** |
95 | | * Initialize mutex. |
96 | | */ |
97 | | void |
98 | 2 | xmlInitDictInternal(void) { |
99 | 2 | xmlInitMutex(&xmlDictMutex); |
100 | 2 | } |
101 | | |
102 | | /** |
103 | | * @deprecated This function is a no-op. Call #xmlCleanupParser |
104 | | * to free global state but see the warnings there. #xmlCleanupParser |
105 | | * should be only called once at program exit. In most cases, you don't |
106 | | * have call cleanup functions at all. |
107 | | */ |
108 | | void |
109 | 0 | xmlDictCleanup(void) { |
110 | 0 | } |
111 | | |
112 | | /** |
113 | | * Free the dictionary mutex. |
114 | | */ |
115 | | void |
116 | 0 | xmlCleanupDictInternal(void) { |
117 | 0 | xmlCleanupMutex(&xmlDictMutex); |
118 | 0 | } |
119 | | |
120 | | /* |
121 | | * @param dict the dictionary |
122 | | * @param name the name of the userdata |
123 | | * @param len the length of the name |
124 | | * |
125 | | * Add the string to the array[s] |
126 | | * |
127 | | * @returns the pointer of the local string, or NULL in case of error. |
128 | | */ |
129 | | static const xmlChar * |
130 | 9.39k | xmlDictAddString(xmlDictPtr dict, const xmlChar *name, unsigned int namelen) { |
131 | 9.39k | xmlDictStringsPtr pool; |
132 | 9.39k | const xmlChar *ret; |
133 | 9.39k | size_t size = 0; /* + sizeof(_xmlDictStrings) == 1024 */ |
134 | 9.39k | size_t limit = 0; |
135 | | |
136 | 9.39k | pool = dict->strings; |
137 | 9.68k | while (pool != NULL) { |
138 | 9.08k | if ((size_t)(pool->end - pool->free) > namelen) |
139 | 8.79k | goto found_pool; |
140 | 288 | if (pool->size > size) size = pool->size; |
141 | 288 | limit += pool->size; |
142 | 288 | pool = pool->next; |
143 | 288 | } |
144 | | /* |
145 | | * Not found, need to allocate |
146 | | */ |
147 | 600 | if (pool == NULL) { |
148 | 600 | if ((dict->limit > 0) && (limit > dict->limit)) { |
149 | 0 | return(NULL); |
150 | 0 | } |
151 | | |
152 | 600 | if (size == 0) { |
153 | 411 | size = 1000; |
154 | 411 | } else { |
155 | 189 | if (size < (SIZE_MAX - sizeof(xmlDictStrings)) / 4) |
156 | 189 | size *= 4; /* exponential growth */ |
157 | 0 | else |
158 | 0 | size = SIZE_MAX - sizeof(xmlDictStrings); |
159 | 189 | } |
160 | 600 | if (size / 4 < namelen) { |
161 | 133 | if ((size_t) namelen + 0 < (SIZE_MAX - sizeof(xmlDictStrings)) / 4) |
162 | 133 | size = 4 * (size_t) namelen; /* just in case ! */ |
163 | 0 | else |
164 | 0 | return(NULL); |
165 | 133 | } |
166 | 600 | pool = (xmlDictStringsPtr) xmlMalloc(sizeof(xmlDictStrings) + size); |
167 | 600 | if (pool == NULL) |
168 | 0 | return(NULL); |
169 | 600 | pool->size = size; |
170 | 600 | pool->nbStrings = 0; |
171 | 600 | pool->free = &pool->array[0]; |
172 | 600 | pool->end = &pool->array[size]; |
173 | 600 | pool->next = dict->strings; |
174 | 600 | dict->strings = pool; |
175 | 600 | } |
176 | 9.39k | found_pool: |
177 | 9.39k | ret = pool->free; |
178 | 9.39k | memcpy(pool->free, name, namelen); |
179 | 9.39k | pool->free += namelen; |
180 | 9.39k | *(pool->free++) = 0; |
181 | 9.39k | pool->nbStrings++; |
182 | 9.39k | return(ret); |
183 | 600 | } |
184 | | |
185 | | /* |
186 | | * @param dict the dictionary |
187 | | * @param prefix the prefix of the userdata |
188 | | * @param plen the prefix length |
189 | | * @param name the name of the userdata |
190 | | * @param len the length of the name |
191 | | * |
192 | | * Add the QName to the array[s] |
193 | | * |
194 | | * @returns the pointer of the local string, or NULL in case of error. |
195 | | */ |
196 | | static const xmlChar * |
197 | | xmlDictAddQString(xmlDictPtr dict, const xmlChar *prefix, unsigned int plen, |
198 | | const xmlChar *name, unsigned int namelen) |
199 | 417 | { |
200 | 417 | xmlDictStringsPtr pool; |
201 | 417 | const xmlChar *ret; |
202 | 417 | size_t size = 0; /* + sizeof(_xmlDictStrings) == 1024 */ |
203 | 417 | size_t limit = 0; |
204 | | |
205 | 417 | pool = dict->strings; |
206 | 430 | while (pool != NULL) { |
207 | 426 | if ((size_t)(pool->end - pool->free) > namelen + plen + 1) |
208 | 413 | goto found_pool; |
209 | 13 | if (pool->size > size) size = pool->size; |
210 | 13 | limit += pool->size; |
211 | 13 | pool = pool->next; |
212 | 13 | } |
213 | | /* |
214 | | * Not found, need to allocate |
215 | | */ |
216 | 4 | if (pool == NULL) { |
217 | 4 | if ((dict->limit > 0) && (limit > dict->limit)) { |
218 | 0 | return(NULL); |
219 | 0 | } |
220 | | |
221 | 4 | if (size == 0) size = 1000; |
222 | 4 | else size *= 4; /* exponential growth */ |
223 | 4 | if (size < 4 * (namelen + plen + 1)) |
224 | 0 | size = 4 * (namelen + plen + 1); /* just in case ! */ |
225 | 4 | pool = (xmlDictStringsPtr) xmlMalloc(sizeof(xmlDictStrings) + size); |
226 | 4 | if (pool == NULL) |
227 | 0 | return(NULL); |
228 | 4 | pool->size = size; |
229 | 4 | pool->nbStrings = 0; |
230 | 4 | pool->free = &pool->array[0]; |
231 | 4 | pool->end = &pool->array[size]; |
232 | 4 | pool->next = dict->strings; |
233 | 4 | dict->strings = pool; |
234 | 4 | } |
235 | 417 | found_pool: |
236 | 417 | ret = pool->free; |
237 | 417 | memcpy(pool->free, prefix, plen); |
238 | 417 | pool->free += plen; |
239 | 417 | *(pool->free++) = ':'; |
240 | 417 | memcpy(pool->free, name, namelen); |
241 | 417 | pool->free += namelen; |
242 | 417 | *(pool->free++) = 0; |
243 | 417 | pool->nbStrings++; |
244 | 417 | return(ret); |
245 | 4 | } |
246 | | |
247 | | /** |
248 | | * Create a new dictionary |
249 | | * |
250 | | * @returns the newly created dictionary, or NULL if an error occurred. |
251 | | */ |
252 | | xmlDict * |
253 | 799 | xmlDictCreate(void) { |
254 | 799 | xmlDictPtr dict; |
255 | | |
256 | 799 | xmlInitParser(); |
257 | | |
258 | 799 | dict = xmlMalloc(sizeof(xmlDict)); |
259 | 799 | if (dict == NULL) |
260 | 0 | return(NULL); |
261 | 799 | dict->ref_counter = 1; |
262 | 799 | dict->limit = 0; |
263 | | |
264 | 799 | dict->size = 0; |
265 | 799 | dict->nbElems = 0; |
266 | 799 | dict->table = NULL; |
267 | 799 | dict->strings = NULL; |
268 | 799 | dict->subdict = NULL; |
269 | 799 | dict->seed = xmlRandom(); |
270 | 799 | #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION |
271 | 799 | dict->seed = 0; |
272 | 799 | #endif |
273 | 799 | return(dict); |
274 | 799 | } |
275 | | |
276 | | /** |
277 | | * Create a new dictionary, inheriting strings from the read-only |
278 | | * dictionary `sub`. On lookup, strings are first searched in the |
279 | | * new dictionary, then in `sub`, and if not found are created in the |
280 | | * new dictionary. |
281 | | * |
282 | | * @param sub an existing dictionary |
283 | | * @returns the newly created dictionary, or NULL if an error occurred. |
284 | | */ |
285 | | xmlDict * |
286 | 397 | xmlDictCreateSub(xmlDict *sub) { |
287 | 397 | xmlDictPtr dict = xmlDictCreate(); |
288 | | |
289 | 397 | if ((dict != NULL) && (sub != NULL)) { |
290 | 397 | dict->seed = sub->seed; |
291 | 397 | dict->subdict = sub; |
292 | 397 | xmlDictReference(dict->subdict); |
293 | 397 | } |
294 | 397 | return(dict); |
295 | 397 | } |
296 | | |
297 | | /** |
298 | | * Increment the reference counter of a dictionary |
299 | | * |
300 | | * @param dict the dictionary |
301 | | * @returns 0 in case of success and -1 in case of error |
302 | | */ |
303 | | int |
304 | 6.45k | xmlDictReference(xmlDict *dict) { |
305 | 6.45k | if (dict == NULL) return -1; |
306 | 6.45k | xmlMutexLock(&xmlDictMutex); |
307 | 6.45k | dict->ref_counter++; |
308 | 6.45k | xmlMutexUnlock(&xmlDictMutex); |
309 | 6.45k | return(0); |
310 | 6.45k | } |
311 | | |
312 | | /** |
313 | | * Free the hash `dict` and its contents. The userdata is |
314 | | * deallocated with `f` if provided. |
315 | | * |
316 | | * @param dict the dictionary |
317 | | */ |
318 | | void |
319 | 7.25k | xmlDictFree(xmlDict *dict) { |
320 | 7.25k | xmlDictStringsPtr pool, nextp; |
321 | | |
322 | 7.25k | if (dict == NULL) |
323 | 0 | return; |
324 | | |
325 | | /* decrement the counter, it may be shared by a parser and docs */ |
326 | 7.25k | xmlMutexLock(&xmlDictMutex); |
327 | 7.25k | dict->ref_counter--; |
328 | 7.25k | if (dict->ref_counter > 0) { |
329 | 6.45k | xmlMutexUnlock(&xmlDictMutex); |
330 | 6.45k | return; |
331 | 6.45k | } |
332 | | |
333 | 797 | xmlMutexUnlock(&xmlDictMutex); |
334 | | |
335 | 797 | if (dict->subdict != NULL) { |
336 | 397 | xmlDictFree(dict->subdict); |
337 | 397 | } |
338 | | |
339 | 797 | if (dict->table) { |
340 | 411 | xmlFree(dict->table); |
341 | 411 | } |
342 | 797 | pool = dict->strings; |
343 | 1.40k | while (pool != NULL) { |
344 | 604 | nextp = pool->next; |
345 | 604 | xmlFree(pool); |
346 | 604 | pool = nextp; |
347 | 604 | } |
348 | 797 | xmlFree(dict); |
349 | 797 | } |
350 | | |
351 | | /** |
352 | | * check if a string is owned by the dictionary |
353 | | * |
354 | | * @param dict the dictionary |
355 | | * @param str the string |
356 | | * @returns 1 if true, 0 if false and -1 in case of error |
357 | | * -1 in case of error |
358 | | */ |
359 | | int |
360 | 2.81M | xmlDictOwns(xmlDict *dict, const xmlChar *str) { |
361 | 2.81M | xmlDictStringsPtr pool; |
362 | | |
363 | 2.81M | if ((dict == NULL) || (str == NULL)) |
364 | 0 | return(-1); |
365 | 2.81M | pool = dict->strings; |
366 | 3.55M | while (pool != NULL) { |
367 | 3.21M | if ((str >= &pool->array[0]) && (str <= pool->free)) |
368 | 2.48M | return(1); |
369 | 736k | pool = pool->next; |
370 | 736k | } |
371 | 333k | if (dict->subdict) |
372 | 17.0k | return(xmlDictOwns(dict->subdict, str)); |
373 | 316k | return(0); |
374 | 333k | } |
375 | | |
376 | | /** |
377 | | * Query the number of elements installed in the hash `dict`. |
378 | | * |
379 | | * @param dict the dictionary |
380 | | * @returns the number of elements in the dictionary or |
381 | | * -1 in case of error |
382 | | */ |
383 | | int |
384 | 0 | xmlDictSize(xmlDict *dict) { |
385 | 0 | if (dict == NULL) |
386 | 0 | return(-1); |
387 | 0 | if (dict->subdict) |
388 | 0 | return(dict->nbElems + dict->subdict->nbElems); |
389 | 0 | return(dict->nbElems); |
390 | 0 | } |
391 | | |
392 | | /** |
393 | | * Set a size limit for the dictionary |
394 | | * Added in 2.9.0 |
395 | | * |
396 | | * @param dict the dictionary |
397 | | * @param limit the limit in bytes |
398 | | * @returns the previous limit of the dictionary or 0 |
399 | | */ |
400 | | size_t |
401 | 398 | xmlDictSetLimit(xmlDict *dict, size_t limit) { |
402 | 398 | size_t ret; |
403 | | |
404 | 398 | if (dict == NULL) |
405 | 0 | return(0); |
406 | 398 | ret = dict->limit; |
407 | 398 | dict->limit = limit; |
408 | 398 | return(ret); |
409 | 398 | } |
410 | | |
411 | | /** |
412 | | * Get how much memory is used by a dictionary for strings |
413 | | * Added in 2.9.0 |
414 | | * |
415 | | * @param dict the dictionary |
416 | | * @returns the amount of strings allocated |
417 | | */ |
418 | | size_t |
419 | 0 | xmlDictGetUsage(xmlDict *dict) { |
420 | 0 | xmlDictStringsPtr pool; |
421 | 0 | size_t limit = 0; |
422 | |
|
423 | 0 | if (dict == NULL) |
424 | 0 | return(0); |
425 | 0 | pool = dict->strings; |
426 | 0 | while (pool != NULL) { |
427 | 0 | limit += pool->size; |
428 | 0 | pool = pool->next; |
429 | 0 | } |
430 | 0 | return(limit); |
431 | 0 | } |
432 | | |
433 | | /***************************************************************** |
434 | | * |
435 | | * The code below was rewritten and is additionally licensed under |
436 | | * the main license in file 'Copyright'. |
437 | | * |
438 | | *****************************************************************/ |
439 | | |
440 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
441 | | static unsigned |
442 | | xmlDictHashName(unsigned seed, const xmlChar* data, size_t maxLen, |
443 | 2.69M | size_t *plen) { |
444 | 2.69M | unsigned h1, h2; |
445 | 2.69M | size_t i; |
446 | | |
447 | 2.69M | HASH_INIT(h1, h2, seed); |
448 | | |
449 | 90.0M | for (i = 0; i < maxLen && data[i]; i++) { |
450 | 87.3M | HASH_UPDATE(h1, h2, data[i]); |
451 | 87.3M | } |
452 | | |
453 | 2.69M | HASH_FINISH(h1, h2); |
454 | | |
455 | 2.69M | *plen = i; |
456 | 2.69M | return(h2 | MAX_HASH_SIZE); |
457 | 2.69M | } |
458 | | |
459 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
460 | | static unsigned |
461 | | xmlDictHashQName(unsigned seed, const xmlChar *prefix, const xmlChar *name, |
462 | 11.5k | size_t *pplen, size_t *plen) { |
463 | 11.5k | unsigned h1, h2; |
464 | 11.5k | size_t i; |
465 | | |
466 | 11.5k | HASH_INIT(h1, h2, seed); |
467 | | |
468 | 519k | for (i = 0; prefix[i] != 0; i++) { |
469 | 508k | HASH_UPDATE(h1, h2, prefix[i]); |
470 | 508k | } |
471 | 11.5k | *pplen = i; |
472 | | |
473 | 11.5k | HASH_UPDATE(h1, h2, ':'); |
474 | | |
475 | 66.4k | for (i = 0; name[i] != 0; i++) { |
476 | 54.9k | HASH_UPDATE(h1, h2, name[i]); |
477 | 54.9k | } |
478 | 11.5k | *plen = i; |
479 | | |
480 | 11.5k | HASH_FINISH(h1, h2); |
481 | | |
482 | | /* |
483 | | * Always set the upper bit of hash values since 0 means an unoccupied |
484 | | * bucket. |
485 | | */ |
486 | 11.5k | return(h2 | MAX_HASH_SIZE); |
487 | 11.5k | } |
488 | | |
489 | | /** |
490 | | * Compute the hash value of a C string. |
491 | | * |
492 | | * @param dict dictionary |
493 | | * @param string C string |
494 | | * @returns the hash value. |
495 | | */ |
496 | | unsigned |
497 | 5.07k | xmlDictComputeHash(const xmlDict *dict, const xmlChar *string) { |
498 | 5.07k | size_t len; |
499 | 5.07k | return(xmlDictHashName(dict->seed, string, SIZE_MAX, &len)); |
500 | 5.07k | } |
501 | | |
502 | 3.83k | #define HASH_ROL31(x,n) ((x) << (n) | ((x) & 0x7FFFFFFF) >> (31 - (n))) |
503 | | |
504 | | /** |
505 | | * Combine two hash values. |
506 | | * |
507 | | * @param v1 first hash value |
508 | | * @param v2 second hash value |
509 | | * @returns the combined hash value. |
510 | | */ |
511 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
512 | | unsigned |
513 | 3.83k | xmlDictCombineHash(unsigned v1, unsigned v2) { |
514 | | /* |
515 | | * The upper bit of hash values is always set, so we have to operate on |
516 | | * 31-bit hashes here. |
517 | | */ |
518 | 3.83k | v1 ^= v2; |
519 | 3.83k | v1 += HASH_ROL31(v2, 5); |
520 | | |
521 | 3.83k | return((v1 & 0xFFFFFFFF) | 0x80000000); |
522 | 3.83k | } |
523 | | |
524 | | /** |
525 | | * Try to find a matching hash table entry. If an entry was found, set |
526 | | * `found` to 1 and return the entry. Otherwise, set `found` to 0 and return |
527 | | * the location where a new entry should be inserted. |
528 | | * |
529 | | * @param dict dict |
530 | | * @param prefix optional QName prefix |
531 | | * @param name string |
532 | | * @param len length of string |
533 | | * @param hashValue valid hash value of string |
534 | | * @param pfound result of search |
535 | | */ |
536 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
537 | | static xmlDictEntry * |
538 | | xmlDictFindEntry(const xmlDict *dict, const xmlChar *prefix, |
539 | | const xmlChar *name, int len, unsigned hashValue, |
540 | 2.70M | int *pfound) { |
541 | 2.70M | xmlDictEntry *entry; |
542 | 2.70M | unsigned mask, pos, displ; |
543 | 2.70M | int found = 0; |
544 | | |
545 | 2.70M | mask = dict->size - 1; |
546 | 2.70M | pos = hashValue & mask; |
547 | 2.70M | entry = &dict->table[pos]; |
548 | | |
549 | 2.70M | if (entry->hashValue != 0) { |
550 | | /* |
551 | | * Robin hood hashing: abort if the displacement of the entry |
552 | | * is smaller than the displacement of the key we look for. |
553 | | * This also stops at the correct position when inserting. |
554 | | */ |
555 | 2.70M | displ = 0; |
556 | | |
557 | 3.21M | do { |
558 | 3.21M | if (entry->hashValue == hashValue) { |
559 | 2.69M | if (prefix == NULL) { |
560 | | /* |
561 | | * name is not necessarily null-terminated. |
562 | | */ |
563 | 2.68M | if ((strncmp((const char *) entry->name, |
564 | 2.68M | (const char *) name, len) == 0) && |
565 | 2.68M | (entry->name[len] == 0)) { |
566 | 2.68M | found = 1; |
567 | 2.68M | break; |
568 | 2.68M | } |
569 | 2.68M | } else { |
570 | 11.1k | if (xmlStrQEqual(prefix, name, entry->name)) { |
571 | 11.1k | found = 1; |
572 | 11.1k | break; |
573 | 11.1k | } |
574 | 11.1k | } |
575 | 2.69M | } |
576 | | |
577 | 519k | displ++; |
578 | 519k | pos++; |
579 | 519k | entry++; |
580 | 519k | if ((pos & mask) == 0) |
581 | 5.94k | entry = dict->table; |
582 | 519k | } while ((entry->hashValue != 0) && |
583 | 519k | (((pos - entry->hashValue) & mask) >= displ)); |
584 | 2.70M | } |
585 | | |
586 | 0 | *pfound = found; |
587 | 2.70M | return(entry); |
588 | 2.70M | } |
589 | | |
590 | | /** |
591 | | * Resize the dictionary hash table. |
592 | | * |
593 | | * @param dict dictionary |
594 | | * @param size new size of the dictionary |
595 | | * @returns 0 in case of success, -1 if a memory allocation failed. |
596 | | */ |
597 | | static int |
598 | 1.09k | xmlDictGrow(xmlDictPtr dict, unsigned size) { |
599 | 1.09k | const xmlDictEntry *oldentry, *oldend, *end; |
600 | 1.09k | xmlDictEntry *table; |
601 | 1.09k | unsigned oldsize, i; |
602 | | |
603 | | /* Add 0 to avoid spurious -Wtype-limits warning on 64-bit GCC */ |
604 | 1.09k | if ((size_t) size + 0 > SIZE_MAX / sizeof(table[0])) |
605 | 0 | return(-1); |
606 | 1.09k | table = xmlMalloc(size * sizeof(table[0])); |
607 | 1.09k | if (table == NULL) |
608 | 0 | return(-1); |
609 | 1.09k | memset(table, 0, size * sizeof(table[0])); |
610 | | |
611 | 1.09k | oldsize = dict->size; |
612 | 1.09k | if (oldsize == 0) |
613 | 411 | goto done; |
614 | | |
615 | 687 | oldend = &dict->table[oldsize]; |
616 | 687 | end = &table[size]; |
617 | | |
618 | | /* |
619 | | * Robin Hood sorting order is maintained if we |
620 | | * |
621 | | * - compute dict indices with modulo |
622 | | * - resize by an integer factor |
623 | | * - start to copy from the beginning of a probe sequence |
624 | | */ |
625 | 687 | oldentry = dict->table; |
626 | 4.49k | while (oldentry->hashValue != 0) { |
627 | 3.80k | if (++oldentry >= oldend) |
628 | 0 | oldentry = dict->table; |
629 | 3.80k | } |
630 | | |
631 | 13.6k | for (i = 0; i < oldsize; i++) { |
632 | 12.9k | if (oldentry->hashValue != 0) { |
633 | 11.3k | xmlDictEntry *entry = &table[oldentry->hashValue & (size - 1)]; |
634 | | |
635 | 14.7k | while (entry->hashValue != 0) { |
636 | 3.42k | if (++entry >= end) |
637 | 147 | entry = table; |
638 | 3.42k | } |
639 | 11.3k | *entry = *oldentry; |
640 | 11.3k | } |
641 | | |
642 | 12.9k | if (++oldentry >= oldend) |
643 | 687 | oldentry = dict->table; |
644 | 12.9k | } |
645 | | |
646 | 687 | xmlFree(dict->table); |
647 | | |
648 | 1.09k | done: |
649 | 1.09k | dict->table = table; |
650 | 1.09k | dict->size = size; |
651 | | |
652 | 1.09k | return(0); |
653 | 687 | } |
654 | | |
655 | | /** |
656 | | * Internal lookup and update function. |
657 | | * |
658 | | * @param dict dict |
659 | | * @param prefix optional QName prefix |
660 | | * @param name string |
661 | | * @param maybeLen length of string or -1 if unknown |
662 | | * @param update whether the string should be added |
663 | | */ |
664 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
665 | | static const xmlDictEntry * |
666 | | xmlDictLookupInternal(xmlDict *dict, const xmlChar *prefix, |
667 | 2.70M | const xmlChar *name, int maybeLen, int update) { |
668 | 2.70M | xmlDictEntry *entry = NULL; |
669 | 2.70M | const xmlChar *ret; |
670 | 2.70M | unsigned hashValue, newSize; |
671 | 2.70M | size_t maxLen, len, plen, klen; |
672 | 2.70M | int found = 0; |
673 | | |
674 | 2.70M | if ((dict == NULL) || (name == NULL)) |
675 | 0 | return(NULL); |
676 | | |
677 | 2.70M | maxLen = (maybeLen < 0) ? SIZE_MAX : (size_t) maybeLen; |
678 | | |
679 | 2.70M | if (prefix == NULL) { |
680 | 2.69M | hashValue = xmlDictHashName(dict->seed, name, maxLen, &len); |
681 | 2.69M | if (len > INT_MAX / 2) |
682 | 0 | return(NULL); |
683 | 2.69M | klen = len; |
684 | 2.69M | } else { |
685 | 11.5k | hashValue = xmlDictHashQName(dict->seed, prefix, name, &plen, &len); |
686 | 11.5k | if ((len > INT_MAX / 2) || (plen >= INT_MAX / 2 - len)) |
687 | 0 | return(NULL); |
688 | 11.5k | klen = plen + 1 + len; |
689 | 11.5k | } |
690 | | |
691 | 2.70M | if ((dict->limit > 0) && (klen >= dict->limit)) |
692 | 0 | return(NULL); |
693 | | |
694 | | /* |
695 | | * Check for an existing entry |
696 | | */ |
697 | 2.70M | if (dict->size == 0) { |
698 | 411 | newSize = MIN_HASH_SIZE; |
699 | 2.70M | } else { |
700 | 2.70M | entry = xmlDictFindEntry(dict, prefix, name, klen, hashValue, &found); |
701 | 2.70M | if (found) |
702 | 2.69M | return(entry); |
703 | | |
704 | 9.40k | if (dict->nbElems + 1 > dict->size / MAX_FILL_DENOM * MAX_FILL_NUM) { |
705 | 687 | if (dict->size >= MAX_HASH_SIZE) |
706 | 0 | return(NULL); |
707 | 687 | newSize = dict->size * 2; |
708 | 8.71k | } else { |
709 | 8.71k | newSize = 0; |
710 | 8.71k | } |
711 | 9.40k | } |
712 | | |
713 | 9.81k | if ((dict->subdict != NULL) && (dict->subdict->size > 0)) { |
714 | 0 | xmlDictEntry *subEntry; |
715 | 0 | unsigned subHashValue; |
716 | |
|
717 | 0 | if (prefix == NULL) |
718 | 0 | subHashValue = xmlDictHashName(dict->subdict->seed, name, len, |
719 | 0 | &len); |
720 | 0 | else |
721 | 0 | subHashValue = xmlDictHashQName(dict->subdict->seed, prefix, name, |
722 | 0 | &plen, &len); |
723 | 0 | subEntry = xmlDictFindEntry(dict->subdict, prefix, name, klen, |
724 | 0 | subHashValue, &found); |
725 | 0 | if (found) |
726 | 0 | return(subEntry); |
727 | 0 | } |
728 | | |
729 | 9.81k | if (!update) |
730 | 0 | return(NULL); |
731 | | |
732 | | /* |
733 | | * Grow the hash table if needed |
734 | | */ |
735 | 9.81k | if (newSize > 0) { |
736 | 1.09k | unsigned mask, displ, pos; |
737 | | |
738 | 1.09k | if (xmlDictGrow(dict, newSize) != 0) |
739 | 0 | return(NULL); |
740 | | |
741 | | /* |
742 | | * Find new entry |
743 | | */ |
744 | 1.09k | mask = dict->size - 1; |
745 | 1.09k | displ = 0; |
746 | 1.09k | pos = hashValue & mask; |
747 | 1.09k | entry = &dict->table[pos]; |
748 | | |
749 | 1.42k | while ((entry->hashValue != 0) && |
750 | 1.42k | ((pos - entry->hashValue) & mask) >= displ) { |
751 | 325 | displ++; |
752 | 325 | pos++; |
753 | 325 | entry++; |
754 | 325 | if ((pos & mask) == 0) |
755 | 7 | entry = dict->table; |
756 | 325 | } |
757 | 1.09k | } |
758 | | |
759 | 9.81k | if (prefix == NULL) |
760 | 9.39k | ret = xmlDictAddString(dict, name, len); |
761 | 417 | else |
762 | 417 | ret = xmlDictAddQString(dict, prefix, plen, name, len); |
763 | 9.81k | if (ret == NULL) |
764 | 0 | return(NULL); |
765 | | |
766 | | /* |
767 | | * Shift the remainder of the probe sequence to the right |
768 | | */ |
769 | 9.81k | if (entry->hashValue != 0) { |
770 | 2.81k | const xmlDictEntry *end = &dict->table[dict->size]; |
771 | 2.81k | const xmlDictEntry *cur = entry; |
772 | | |
773 | 11.3k | do { |
774 | 11.3k | cur++; |
775 | 11.3k | if (cur >= end) |
776 | 537 | cur = dict->table; |
777 | 11.3k | } while (cur->hashValue != 0); |
778 | | |
779 | 2.81k | if (cur < entry) { |
780 | | /* |
781 | | * If we traversed the end of the buffer, handle the part |
782 | | * at the start of the buffer. |
783 | | */ |
784 | 537 | memmove(&dict->table[1], dict->table, |
785 | 537 | (char *) cur - (char *) dict->table); |
786 | 537 | cur = end - 1; |
787 | 537 | dict->table[0] = *cur; |
788 | 537 | } |
789 | | |
790 | 2.81k | memmove(&entry[1], entry, (char *) cur - (char *) entry); |
791 | 2.81k | } |
792 | | |
793 | | /* |
794 | | * Populate entry |
795 | | */ |
796 | 9.81k | entry->hashValue = hashValue; |
797 | 9.81k | entry->name = ret; |
798 | | |
799 | 9.81k | dict->nbElems++; |
800 | | |
801 | 9.81k | return(entry); |
802 | 9.81k | } |
803 | | |
804 | | /** |
805 | | * Lookup a string and add it to the dictionary if it wasn't found. |
806 | | * |
807 | | * @param dict dictionary |
808 | | * @param name string key |
809 | | * @param len length of the key, if -1 it is recomputed |
810 | | * @returns the interned copy of the string or NULL if a memory allocation |
811 | | * failed. |
812 | | */ |
813 | | const xmlChar * |
814 | 346k | xmlDictLookup(xmlDict *dict, const xmlChar *name, int len) { |
815 | 346k | const xmlDictEntry *entry; |
816 | | |
817 | 346k | entry = xmlDictLookupInternal(dict, NULL, name, len, 1); |
818 | 346k | if (entry == NULL) |
819 | 0 | return(NULL); |
820 | 346k | return(entry->name); |
821 | 346k | } |
822 | | |
823 | | /** |
824 | | * Lookup a dictionary entry and add the string to the dictionary if |
825 | | * it wasn't found. |
826 | | * |
827 | | * @param dict dictionary |
828 | | * @param name string key |
829 | | * @param len length of the key, if -1 it is recomputed |
830 | | * @returns the dictionary entry. |
831 | | */ |
832 | | xmlHashedString |
833 | 2.34M | xmlDictLookupHashed(xmlDict *dict, const xmlChar *name, int len) { |
834 | 2.34M | const xmlDictEntry *entry; |
835 | 2.34M | xmlHashedString ret; |
836 | | |
837 | 2.34M | entry = xmlDictLookupInternal(dict, NULL, name, len, 1); |
838 | | |
839 | 2.34M | if (entry == NULL) { |
840 | 0 | ret.name = NULL; |
841 | 0 | ret.hashValue = 0; |
842 | 2.34M | } else { |
843 | 2.34M | ret = *entry; |
844 | 2.34M | } |
845 | | |
846 | 2.34M | return(ret); |
847 | 2.34M | } |
848 | | |
849 | | /** |
850 | | * Check if a string exists in the dictionary. |
851 | | * |
852 | | * @param dict the dictionary |
853 | | * @param name the name of the userdata |
854 | | * @param len the length of the name, if -1 it is recomputed |
855 | | * @returns the internal copy of the name or NULL if not found. |
856 | | */ |
857 | | const xmlChar * |
858 | 0 | xmlDictExists(xmlDict *dict, const xmlChar *name, int len) { |
859 | 0 | const xmlDictEntry *entry; |
860 | |
|
861 | 0 | entry = xmlDictLookupInternal(dict, NULL, name, len, 0); |
862 | 0 | if (entry == NULL) |
863 | 0 | return(NULL); |
864 | 0 | return(entry->name); |
865 | 0 | } |
866 | | |
867 | | /** |
868 | | * Lookup the QName `prefix:name` and add it to the dictionary if |
869 | | * it wasn't found. |
870 | | * |
871 | | * @param dict the dictionary |
872 | | * @param prefix the prefix |
873 | | * @param name the name |
874 | | * @returns the interned copy of the string or NULL if a memory allocation |
875 | | * failed. |
876 | | */ |
877 | | const xmlChar * |
878 | 11.5k | xmlDictQLookup(xmlDict *dict, const xmlChar *prefix, const xmlChar *name) { |
879 | 11.5k | const xmlDictEntry *entry; |
880 | | |
881 | 11.5k | entry = xmlDictLookupInternal(dict, prefix, name, -1, 1); |
882 | 11.5k | if (entry == NULL) |
883 | 0 | return(NULL); |
884 | 11.5k | return(entry->name); |
885 | 11.5k | } |
886 | | |
887 | | /* |
888 | | * Pseudo-random generator |
889 | | */ |
890 | | |
891 | | #ifdef _WIN32 |
892 | | #define WIN32_LEAN_AND_MEAN |
893 | | #include <windows.h> |
894 | | #include <bcrypt.h> |
895 | | #else |
896 | | #if HAVE_DECL_GETENTROPY |
897 | | /* POSIX 2024 */ |
898 | | #include <unistd.h> |
899 | | /* Older platforms */ |
900 | | #include <sys/random.h> |
901 | | #endif |
902 | | #include <time.h> |
903 | | #endif |
904 | | |
905 | | static xmlMutex xmlRngMutex; |
906 | | |
907 | | static unsigned globalRngState[2]; |
908 | | |
909 | | /* |
910 | | * |
911 | | * Initialize the PRNG. |
912 | | */ |
913 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
914 | | void |
915 | 2 | xmlInitRandom(void) { |
916 | 2 | xmlInitMutex(&xmlRngMutex); |
917 | | |
918 | 2 | { |
919 | | #ifdef _WIN32 |
920 | | NTSTATUS status; |
921 | | |
922 | | /* |
923 | | * You can find many (recent as of 2025) discussions how |
924 | | * to get a pseudo-random seed on Windows in projects like |
925 | | * Golang, Rust, Chromium and Firefox. |
926 | | * |
927 | | * TODO: Support ProcessPrng available since Windows 10. |
928 | | */ |
929 | | status = BCryptGenRandom(NULL, (unsigned char *) globalRngState, |
930 | | sizeof(globalRngState), |
931 | | BCRYPT_USE_SYSTEM_PREFERRED_RNG); |
932 | | if (!BCRYPT_SUCCESS(status)) |
933 | | xmlAbort("libxml2: BCryptGenRandom failed with error code %lu\n", |
934 | | GetLastError()); |
935 | | #else |
936 | 2 | int var; |
937 | | |
938 | 2 | #if HAVE_DECL_GETENTROPY |
939 | 2 | while (1) { |
940 | 2 | if (getentropy(globalRngState, sizeof(globalRngState)) == 0) |
941 | 2 | return; |
942 | | |
943 | | /* |
944 | | * This most likely means that libxml2 was compiled on |
945 | | * a system supporting certain system calls and is running |
946 | | * on a system that doesn't support these calls, as can |
947 | | * be the case on Linux. |
948 | | */ |
949 | 0 | if (errno == ENOSYS) |
950 | 0 | break; |
951 | | |
952 | | /* |
953 | | * We really don't want to fallback to the unsafe PRNG |
954 | | * for possibly accidental reasons, so we abort on any |
955 | | * unknown error. |
956 | | */ |
957 | 0 | if (errno != EINTR) |
958 | 0 | xmlAbort("libxml2: getentropy failed with error code %d\n", |
959 | 0 | errno); |
960 | 0 | } |
961 | 0 | #endif |
962 | | |
963 | | /* |
964 | | * TODO: Fallback to /dev/urandom for older POSIX systems. |
965 | | */ |
966 | 0 | globalRngState[0] = |
967 | 0 | (unsigned) time(NULL) ^ |
968 | 0 | HASH_ROL((unsigned) ((size_t) &xmlInitRandom & 0xFFFFFFFF), 8); |
969 | 0 | globalRngState[1] = |
970 | 0 | HASH_ROL((unsigned) ((size_t) &xmlRngMutex & 0xFFFFFFFF), 16) ^ |
971 | 0 | HASH_ROL((unsigned) ((size_t) &var & 0xFFFFFFFF), 24); |
972 | 0 | #endif |
973 | 0 | } |
974 | 0 | } |
975 | | |
976 | | /* |
977 | | * |
978 | | * Clean up PRNG globals. |
979 | | */ |
980 | | void |
981 | 0 | xmlCleanupRandom(void) { |
982 | 0 | xmlCleanupMutex(&xmlRngMutex); |
983 | 0 | } |
984 | | |
985 | | ATTRIBUTE_NO_SANITIZE_INTEGER |
986 | | static unsigned |
987 | 3.67k | xoroshiro64ss(unsigned *s) { |
988 | 3.67k | unsigned s0 = s[0]; |
989 | 3.67k | unsigned s1 = s[1]; |
990 | 3.67k | unsigned result = HASH_ROL(s0 * 0x9E3779BB, 5) * 5; |
991 | | |
992 | 3.67k | s1 ^= s0; |
993 | 3.67k | s[0] = HASH_ROL(s0, 26) ^ s1 ^ (s1 << 9); |
994 | 3.67k | s[1] = HASH_ROL(s1, 13); |
995 | | |
996 | 3.67k | return(result & 0xFFFFFFFF); |
997 | 3.67k | } |
998 | | |
999 | | /* |
1000 | | * |
1001 | | * Generate a pseudo-random value using the global PRNG. |
1002 | | * |
1003 | | * @returns a random value. |
1004 | | */ |
1005 | | unsigned |
1006 | 4 | xmlGlobalRandom(void) { |
1007 | 4 | unsigned ret; |
1008 | | |
1009 | 4 | xmlMutexLock(&xmlRngMutex); |
1010 | 4 | ret = xoroshiro64ss(globalRngState); |
1011 | 4 | xmlMutexUnlock(&xmlRngMutex); |
1012 | | |
1013 | 4 | return(ret); |
1014 | 4 | } |
1015 | | |
1016 | | /* |
1017 | | * |
1018 | | * Generate a pseudo-random value using the thread-local PRNG. |
1019 | | * |
1020 | | * @returns a random value. |
1021 | | */ |
1022 | | unsigned |
1023 | 3.66k | xmlRandom(void) { |
1024 | 3.66k | return(xoroshiro64ss(xmlGetLocalRngState())); |
1025 | 3.66k | } |
1026 | | |