Coverage Report

Created: 2023-05-18 19:10

/src/libxml2/dict.c
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.com
17
 */
18
19
#define IN_LIBXML
20
#include "libxml.h"
21
22
#include <limits.h>
23
#include <stdlib.h>
24
#include <time.h>
25
26
#include "private/dict.h"
27
#include "private/threads.h"
28
29
/*
30
 * Following http://www.ocert.org/advisories/ocert-2011-003.html
31
 * it seems that having hash randomization might be a good idea
32
 * when using XML with untrusted data
33
 * Note1: that it works correctly only if compiled with WITH_BIG_KEY
34
 *  which is the default.
35
 * Note2: the fast function used for a small dict won't protect very
36
 *  well but since the attack is based on growing a very big hash
37
 *  list we will use the BigKey algo as soon as the hash size grows
38
 *  over MIN_DICT_SIZE so this actually works
39
 */
40
#if !defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION)
41
#define DICT_RANDOMIZATION
42
#endif
43
44
#include <string.h>
45
#ifdef HAVE_STDINT_H
46
#include <stdint.h>
47
#else
48
#ifdef HAVE_INTTYPES_H
49
#include <inttypes.h>
50
#elif defined(_WIN32)
51
typedef unsigned __int32 uint32_t;
52
#endif
53
#endif
54
#include <libxml/tree.h>
55
#include <libxml/dict.h>
56
#include <libxml/xmlmemory.h>
57
#include <libxml/xmlerror.h>
58
#include <libxml/globals.h>
59
60
/* #define DEBUG_GROW */
61
/* #define DICT_DEBUG_PATTERNS */
62
63
24.3M
#define MAX_HASH_LEN 3
64
391M
#define MIN_DICT_SIZE 128
65
27.7k
#define MAX_DICT_HASH 8 * 2048
66
#define WITH_BIG_KEY
67
68
#ifdef WITH_BIG_KEY
69
#define xmlDictComputeKey(dict, name, len)                              \
70
385M
    (((dict)->size == MIN_DICT_SIZE) ?                                  \
71
385M
     xmlDictComputeFastKey(name, len, (dict)->seed) :                   \
72
385M
     xmlDictComputeBigKey(name, len, (dict)->seed))
73
74
#define xmlDictComputeQKey(dict, prefix, plen, name, len)               \
75
316k
    (((prefix) == NULL) ?                                               \
76
316k
      (xmlDictComputeKey(dict, name, len)) :                             \
77
316k
      (((dict)->size == MIN_DICT_SIZE) ?                                \
78
316k
       xmlDictComputeFastQKey(prefix, plen, name, len, (dict)->seed) :  \
79
316k
       xmlDictComputeBigQKey(prefix, plen, name, len, (dict)->seed)))
80
81
#else /* !WITH_BIG_KEY */
82
#define xmlDictComputeKey(dict, name, len)                              \
83
        xmlDictComputeFastKey(name, len, (dict)->seed)
84
#define xmlDictComputeQKey(dict, prefix, plen, name, len)               \
85
        xmlDictComputeFastQKey(prefix, plen, name, len, (dict)->seed)
86
#endif /* WITH_BIG_KEY */
87
88
/*
89
 * An entry in the dictionary
90
 */
91
typedef struct _xmlDictEntry xmlDictEntry;
92
typedef xmlDictEntry *xmlDictEntryPtr;
93
struct _xmlDictEntry {
94
    struct _xmlDictEntry *next;
95
    const xmlChar *name;
96
    unsigned int len;
97
    int valid;
98
    unsigned long okey;
99
};
100
101
typedef struct _xmlDictStrings xmlDictStrings;
102
typedef xmlDictStrings *xmlDictStringsPtr;
103
struct _xmlDictStrings {
104
    xmlDictStringsPtr next;
105
    xmlChar *free;
106
    xmlChar *end;
107
    size_t size;
108
    size_t nbStrings;
109
    xmlChar array[1];
110
};
111
/*
112
 * The entire dictionary
113
 */
114
struct _xmlDict {
115
    int ref_counter;
116
117
    struct _xmlDictEntry *dict;
118
    size_t size;
119
    unsigned int nbElems;
120
    xmlDictStringsPtr strings;
121
122
    struct _xmlDict *subdict;
123
    /* used for randomization */
124
    int seed;
125
    /* used to impose a limit on size */
126
    size_t limit;
127
};
128
129
/*
130
 * A mutex for modifying the reference counter for shared
131
 * dictionaries.
132
 */
133
static xmlMutex xmlDictMutex;
134
135
#ifdef DICT_RANDOMIZATION
136
#ifdef HAVE_RAND_R
137
/*
138
 * Internal data for random function, protected by xmlDictMutex
139
 */
140
static unsigned int rand_seed = 0;
141
#endif
142
#endif
143
144
/**
145
 * xmlInitializeDict:
146
 *
147
 * DEPRECATED: Alias for xmlInitParser.
148
 */
149
0
int xmlInitializeDict(void) {
150
0
    xmlInitParser();
151
0
    return(0);
152
0
}
153
154
/**
155
 * __xmlInitializeDict:
156
 *
157
 * This function is not public
158
 * Do the dictionary mutex initialization.
159
 */
160
1.02k
int __xmlInitializeDict(void) {
161
1.02k
    xmlInitMutex(&xmlDictMutex);
162
163
#ifdef DICT_RANDOMIZATION
164
#ifdef HAVE_RAND_R
165
    rand_seed = time(NULL);
166
    rand_r(& rand_seed);
167
#else
168
    srand(time(NULL));
169
#endif
170
#endif
171
1.02k
    return(1);
172
1.02k
}
173
174
#ifdef DICT_RANDOMIZATION
175
int __xmlRandom(void) {
176
    int ret;
177
178
    xmlMutexLock(&xmlDictMutex);
179
#ifdef HAVE_RAND_R
180
    ret = rand_r(& rand_seed);
181
#else
182
    ret = rand();
183
#endif
184
    xmlMutexUnlock(&xmlDictMutex);
185
    return(ret);
186
}
187
#endif
188
189
/**
190
 * xmlDictCleanup:
191
 *
192
 * DEPRECATED: This function is a no-op. Call xmlCleanupParser
193
 * to free global state but see the warnings there. xmlCleanupParser
194
 * should be only called once at program exit. In most cases, you don't
195
 * have call cleanup functions at all.
196
 */
197
void
198
0
xmlDictCleanup(void) {
199
0
}
200
201
/**
202
 * xmlCleanupDictInternal:
203
 *
204
 * Free the dictionary mutex.
205
 */
206
void
207
0
xmlCleanupDictInternal(void) {
208
0
    xmlCleanupMutex(&xmlDictMutex);
209
0
}
210
211
/*
212
 * xmlDictAddString:
213
 * @dict: the dictionary
214
 * @name: the name of the userdata
215
 * @len: the length of the name
216
 *
217
 * Add the string to the array[s]
218
 *
219
 * Returns the pointer of the local string, or NULL in case of error.
220
 */
221
static const xmlChar *
222
24.2M
xmlDictAddString(xmlDictPtr dict, const xmlChar *name, unsigned int namelen) {
223
24.2M
    xmlDictStringsPtr pool;
224
24.2M
    const xmlChar *ret;
225
24.2M
    size_t size = 0; /* + sizeof(_xmlDictStrings) == 1024 */
226
24.2M
    size_t limit = 0;
227
228
#ifdef DICT_DEBUG_PATTERNS
229
    fprintf(stderr, "-");
230
#endif
231
24.2M
    pool = dict->strings;
232
24.2M
    while (pool != NULL) {
233
22.9M
  if ((size_t)(pool->end - pool->free) > namelen)
234
22.8M
      goto found_pool;
235
72.0k
  if (pool->size > size) size = pool->size;
236
72.0k
        limit += pool->size;
237
72.0k
  pool = pool->next;
238
72.0k
    }
239
    /*
240
     * Not found, need to allocate
241
     */
242
1.32M
    if (pool == NULL) {
243
1.32M
        if ((dict->limit > 0) && (limit > dict->limit)) {
244
0
            return(NULL);
245
0
        }
246
247
1.32M
        if (size == 0) size = 1000;
248
68.1k
  else size *= 4; /* exponential growth */
249
1.32M
        if (size < 4 * namelen)
250
13.7k
      size = 4 * namelen; /* just in case ! */
251
1.32M
  pool = (xmlDictStringsPtr) xmlMalloc(sizeof(xmlDictStrings) + size);
252
1.32M
  if (pool == NULL)
253
0
      return(NULL);
254
1.32M
  pool->size = size;
255
1.32M
  pool->nbStrings = 0;
256
1.32M
  pool->free = &pool->array[0];
257
1.32M
  pool->end = &pool->array[size];
258
1.32M
  pool->next = dict->strings;
259
1.32M
  dict->strings = pool;
260
#ifdef DICT_DEBUG_PATTERNS
261
        fprintf(stderr, "+");
262
#endif
263
1.32M
    }
264
24.2M
found_pool:
265
24.2M
    ret = pool->free;
266
24.2M
    memcpy(pool->free, name, namelen);
267
24.2M
    pool->free += namelen;
268
24.2M
    *(pool->free++) = 0;
269
24.2M
    pool->nbStrings++;
270
24.2M
    return(ret);
271
1.32M
}
272
273
/*
274
 * xmlDictAddQString:
275
 * @dict: the dictionary
276
 * @prefix: the prefix of the userdata
277
 * @plen: the prefix length
278
 * @name: the name of the userdata
279
 * @len: the length of the name
280
 *
281
 * Add the QName to the array[s]
282
 *
283
 * Returns the pointer of the local string, or NULL in case of error.
284
 */
285
static const xmlChar *
286
xmlDictAddQString(xmlDictPtr dict, const xmlChar *prefix, unsigned int plen,
287
                 const xmlChar *name, unsigned int namelen)
288
94.5k
{
289
94.5k
    xmlDictStringsPtr pool;
290
94.5k
    const xmlChar *ret;
291
94.5k
    size_t size = 0; /* + sizeof(_xmlDictStrings) == 1024 */
292
94.5k
    size_t limit = 0;
293
294
94.5k
    if (prefix == NULL) return(xmlDictAddString(dict, name, namelen));
295
296
#ifdef DICT_DEBUG_PATTERNS
297
    fprintf(stderr, "=");
298
#endif
299
94.5k
    pool = dict->strings;
300
95.0k
    while (pool != NULL) {
301
94.8k
  if ((size_t)(pool->end - pool->free) > namelen + plen + 1)
302
94.2k
      goto found_pool;
303
532
  if (pool->size > size) size = pool->size;
304
532
        limit += pool->size;
305
532
  pool = pool->next;
306
532
    }
307
    /*
308
     * Not found, need to allocate
309
     */
310
278
    if (pool == NULL) {
311
278
        if ((dict->limit > 0) && (limit > dict->limit)) {
312
0
            return(NULL);
313
0
        }
314
315
278
        if (size == 0) size = 1000;
316
278
  else size *= 4; /* exponential growth */
317
278
        if (size < 4 * (namelen + plen + 1))
318
0
      size = 4 * (namelen + plen + 1); /* just in case ! */
319
278
  pool = (xmlDictStringsPtr) xmlMalloc(sizeof(xmlDictStrings) + size);
320
278
  if (pool == NULL)
321
0
      return(NULL);
322
278
  pool->size = size;
323
278
  pool->nbStrings = 0;
324
278
  pool->free = &pool->array[0];
325
278
  pool->end = &pool->array[size];
326
278
  pool->next = dict->strings;
327
278
  dict->strings = pool;
328
#ifdef DICT_DEBUG_PATTERNS
329
        fprintf(stderr, "+");
330
#endif
331
278
    }
332
94.5k
found_pool:
333
94.5k
    ret = pool->free;
334
94.5k
    memcpy(pool->free, prefix, plen);
335
94.5k
    pool->free += plen;
336
94.5k
    *(pool->free++) = ':';
337
94.5k
    memcpy(pool->free, name, namelen);
338
94.5k
    pool->free += namelen;
339
94.5k
    *(pool->free++) = 0;
340
94.5k
    pool->nbStrings++;
341
94.5k
    return(ret);
342
278
}
343
344
#ifdef WITH_BIG_KEY
345
/*
346
 * xmlDictComputeBigKey:
347
 *
348
 * Calculate a hash key using a good hash function that works well for
349
 * larger hash table sizes.
350
 *
351
 * Hash function by "One-at-a-Time Hash" see
352
 * http://burtleburtle.net/bob/hash/doobs.html
353
 */
354
355
#ifdef __clang__
356
ATTRIBUTE_NO_SANITIZE("unsigned-integer-overflow")
357
#endif
358
static uint32_t
359
54.0M
xmlDictComputeBigKey(const xmlChar* data, int namelen, int seed) {
360
54.0M
    uint32_t hash;
361
54.0M
    int i;
362
363
54.0M
    if (namelen <= 0 || data == NULL) return(0);
364
365
54.0M
    hash = seed;
366
367
357M
    for (i = 0;i < namelen; i++) {
368
303M
        hash += data[i];
369
303M
  hash += (hash << 10);
370
303M
  hash ^= (hash >> 6);
371
303M
    }
372
54.0M
    hash += (hash << 3);
373
54.0M
    hash ^= (hash >> 11);
374
54.0M
    hash += (hash << 15);
375
376
54.0M
    return hash;
377
54.0M
}
378
379
/*
380
 * xmlDictComputeBigQKey:
381
 *
382
 * Calculate a hash key for two strings using a good hash function
383
 * that works well for larger hash table sizes.
384
 *
385
 * Hash function by "One-at-a-Time Hash" see
386
 * http://burtleburtle.net/bob/hash/doobs.html
387
 *
388
 * Neither of the two strings must be NULL.
389
 */
390
#ifdef __clang__
391
ATTRIBUTE_NO_SANITIZE("unsigned-integer-overflow")
392
#endif
393
static unsigned long
394
xmlDictComputeBigQKey(const xmlChar *prefix, int plen,
395
                      const xmlChar *name, int len, int seed)
396
108k
{
397
108k
    uint32_t hash;
398
108k
    int i;
399
400
108k
    hash = seed;
401
402
453k
    for (i = 0;i < plen; i++) {
403
344k
        hash += prefix[i];
404
344k
  hash += (hash << 10);
405
344k
  hash ^= (hash >> 6);
406
344k
    }
407
108k
    hash += ':';
408
108k
    hash += (hash << 10);
409
108k
    hash ^= (hash >> 6);
410
411
858k
    for (i = 0;i < len; i++) {
412
750k
        hash += name[i];
413
750k
  hash += (hash << 10);
414
750k
  hash ^= (hash >> 6);
415
750k
    }
416
108k
    hash += (hash << 3);
417
108k
    hash ^= (hash >> 11);
418
108k
    hash += (hash << 15);
419
420
108k
    return hash;
421
108k
}
422
#endif /* WITH_BIG_KEY */
423
424
/*
425
 * xmlDictComputeFastKey:
426
 *
427
 * Calculate a hash key using a fast hash function that works well
428
 * for low hash table fill.
429
 */
430
static unsigned long
431
331M
xmlDictComputeFastKey(const xmlChar *name, int namelen, int seed) {
432
331M
    unsigned long value = seed;
433
434
331M
    if (name == NULL) return(0);
435
331M
    value += *name;
436
331M
    value <<= 5;
437
331M
    if (namelen > 10) {
438
13.7M
        value += name[namelen - 1];
439
13.7M
        namelen = 10;
440
13.7M
    }
441
331M
    switch (namelen) {
442
18.4M
        case 10: value += name[9];
443
        /* Falls through. */
444
21.5M
        case 9: value += name[8];
445
        /* Falls through. */
446
27.7M
        case 8: value += name[7];
447
        /* Falls through. */
448
38.2M
        case 7: value += name[6];
449
        /* Falls through. */
450
47.9M
        case 6: value += name[5];
451
        /* Falls through. */
452
65.1M
        case 5: value += name[4];
453
        /* Falls through. */
454
230M
        case 4: value += name[3];
455
        /* Falls through. */
456
259M
        case 3: value += name[2];
457
        /* Falls through. */
458
274M
        case 2: value += name[1];
459
        /* Falls through. */
460
331M
        default: break;
461
331M
    }
462
331M
    return(value);
463
331M
}
464
465
/*
466
 * xmlDictComputeFastQKey:
467
 *
468
 * Calculate a hash key for two strings using a fast hash function
469
 * that works well for low hash table fill.
470
 *
471
 * Neither of the two strings must be NULL.
472
 */
473
static unsigned long
474
xmlDictComputeFastQKey(const xmlChar *prefix, int plen,
475
                       const xmlChar *name, int len, int seed)
476
208k
{
477
208k
    unsigned long value = seed;
478
479
208k
    if (plen == 0)
480
0
  value += 30 * ':';
481
208k
    else
482
208k
  value += 30 * (*prefix);
483
484
208k
    if (len > 10) {
485
14.9k
        int offset = len - (plen + 1 + 1);
486
14.9k
  if (offset < 0)
487
541
      offset = len - (10 + 1);
488
14.9k
  value += name[offset];
489
14.9k
        len = 10;
490
14.9k
  if (plen > 10)
491
691
      plen = 10;
492
14.9k
    }
493
208k
    switch (plen) {
494
1.77k
        case 10: value += prefix[9];
495
        /* Falls through. */
496
2.50k
        case 9: value += prefix[8];
497
        /* Falls through. */
498
4.13k
        case 8: value += prefix[7];
499
        /* Falls through. */
500
5.55k
        case 7: value += prefix[6];
501
        /* Falls through. */
502
7.02k
        case 6: value += prefix[5];
503
        /* Falls through. */
504
16.6k
        case 5: value += prefix[4];
505
        /* Falls through. */
506
39.7k
        case 4: value += prefix[3];
507
        /* Falls through. */
508
158k
        case 3: value += prefix[2];
509
        /* Falls through. */
510
161k
        case 2: value += prefix[1];
511
        /* Falls through. */
512
206k
        case 1: value += prefix[0];
513
        /* Falls through. */
514
208k
        default: break;
515
208k
    }
516
208k
    len -= plen;
517
208k
    if (len > 0) {
518
168k
        value += ':';
519
168k
  len--;
520
168k
    }
521
208k
    switch (len) {
522
0
        case 10: value += name[9];
523
        /* Falls through. */
524
0
        case 9: value += name[8];
525
        /* Falls through. */
526
5.34k
        case 8: value += name[7];
527
        /* Falls through. */
528
7.01k
        case 7: value += name[6];
529
        /* Falls through. */
530
23.9k
        case 6: value += name[5];
531
        /* Falls through. */
532
61.3k
        case 5: value += name[4];
533
        /* Falls through. */
534
72.5k
        case 4: value += name[3];
535
        /* Falls through. */
536
81.5k
        case 3: value += name[2];
537
        /* Falls through. */
538
115k
        case 2: value += name[1];
539
        /* Falls through. */
540
140k
        case 1: value += name[0];
541
        /* Falls through. */
542
208k
        default: break;
543
208k
    }
544
208k
    return(value);
545
208k
}
546
547
/**
548
 * xmlDictCreate:
549
 *
550
 * Create a new dictionary
551
 *
552
 * Returns the newly created dictionary, or NULL if an error occurred.
553
 */
554
xmlDictPtr
555
1.81M
xmlDictCreate(void) {
556
1.81M
    xmlDictPtr dict;
557
558
1.81M
    xmlInitParser();
559
560
#ifdef DICT_DEBUG_PATTERNS
561
    fprintf(stderr, "C");
562
#endif
563
564
1.81M
    dict = xmlMalloc(sizeof(xmlDict));
565
1.81M
    if (dict) {
566
1.81M
        dict->ref_counter = 1;
567
1.81M
        dict->limit = 0;
568
569
1.81M
        dict->size = MIN_DICT_SIZE;
570
1.81M
  dict->nbElems = 0;
571
1.81M
        dict->dict = xmlMalloc(MIN_DICT_SIZE * sizeof(xmlDictEntry));
572
1.81M
  dict->strings = NULL;
573
1.81M
  dict->subdict = NULL;
574
1.81M
        if (dict->dict) {
575
1.81M
      memset(dict->dict, 0, MIN_DICT_SIZE * sizeof(xmlDictEntry));
576
#ifdef DICT_RANDOMIZATION
577
            dict->seed = __xmlRandom();
578
#else
579
1.81M
            dict->seed = 0;
580
1.81M
#endif
581
1.81M
      return(dict);
582
1.81M
        }
583
0
        xmlFree(dict);
584
0
    }
585
0
    return(NULL);
586
1.81M
}
587
588
/**
589
 * xmlDictCreateSub:
590
 * @sub: an existing dictionary
591
 *
592
 * Create a new dictionary, inheriting strings from the read-only
593
 * dictionary @sub. On lookup, strings are first searched in the
594
 * new dictionary, then in @sub, and if not found are created in the
595
 * new dictionary.
596
 *
597
 * Returns the newly created dictionary, or NULL if an error occurred.
598
 */
599
xmlDictPtr
600
0
xmlDictCreateSub(xmlDictPtr sub) {
601
0
    xmlDictPtr dict = xmlDictCreate();
602
603
0
    if ((dict != NULL) && (sub != NULL)) {
604
#ifdef DICT_DEBUG_PATTERNS
605
        fprintf(stderr, "R");
606
#endif
607
0
        dict->seed = sub->seed;
608
0
        dict->subdict = sub;
609
0
  xmlDictReference(dict->subdict);
610
0
    }
611
0
    return(dict);
612
0
}
613
614
/**
615
 * xmlDictReference:
616
 * @dict: the dictionary
617
 *
618
 * Increment the reference counter of a dictionary
619
 *
620
 * Returns 0 in case of success and -1 in case of error
621
 */
622
int
623
1.94M
xmlDictReference(xmlDictPtr dict) {
624
1.94M
    if (dict == NULL) return -1;
625
1.69M
    xmlMutexLock(&xmlDictMutex);
626
1.69M
    dict->ref_counter++;
627
1.69M
    xmlMutexUnlock(&xmlDictMutex);
628
1.69M
    return(0);
629
1.94M
}
630
631
/**
632
 * xmlDictGrow:
633
 * @dict: the dictionary
634
 * @size: the new size of the dictionary
635
 *
636
 * resize the dictionary
637
 *
638
 * Returns 0 in case of success, -1 in case of failure
639
 */
640
static int
641
27.7k
xmlDictGrow(xmlDictPtr dict, size_t size) {
642
27.7k
    unsigned long key, okey;
643
27.7k
    size_t oldsize, i;
644
27.7k
    xmlDictEntryPtr iter, next;
645
27.7k
    struct _xmlDictEntry *olddict;
646
#ifdef DEBUG_GROW
647
    unsigned long nbElem = 0;
648
#endif
649
27.7k
    int ret = 0;
650
27.7k
    int keep_keys = 1;
651
652
27.7k
    if (dict == NULL)
653
0
  return(-1);
654
27.7k
    if (size < 8)
655
0
        return(-1);
656
27.7k
    if (size > 8 * 2048)
657
0
  return(-1);
658
659
#ifdef DICT_DEBUG_PATTERNS
660
    fprintf(stderr, "*");
661
#endif
662
663
27.7k
    oldsize = dict->size;
664
27.7k
    olddict = dict->dict;
665
27.7k
    if (olddict == NULL)
666
0
        return(-1);
667
27.7k
    if (oldsize == MIN_DICT_SIZE)
668
27.5k
        keep_keys = 0;
669
670
27.7k
    dict->dict = xmlMalloc(size * sizeof(xmlDictEntry));
671
27.7k
    if (dict->dict == NULL) {
672
0
  dict->dict = olddict;
673
0
  return(-1);
674
0
    }
675
27.7k
    memset(dict->dict, 0, size * sizeof(xmlDictEntry));
676
27.7k
    dict->size = size;
677
678
    /*  If the two loops are merged, there would be situations where
679
  a new entry needs to allocated and data copied into it from
680
  the main dict. It is nicer to run through the array twice, first
681
  copying all the elements in the main array (less probability of
682
  allocate) and then the rest, so we only free in the second loop.
683
    */
684
3.71M
    for (i = 0; i < oldsize; i++) {
685
3.68M
  if (olddict[i].valid == 0)
686
1.12M
      continue;
687
688
2.56M
  if (keep_keys)
689
72.2k
      okey = olddict[i].okey;
690
2.48M
  else
691
2.48M
      okey = xmlDictComputeKey(dict, olddict[i].name, olddict[i].len);
692
2.56M
  key = okey % dict->size;
693
694
2.56M
  if (dict->dict[key].valid == 0) {
695
2.37M
      memcpy(&(dict->dict[key]), &(olddict[i]), sizeof(xmlDictEntry));
696
2.37M
      dict->dict[key].next = NULL;
697
2.37M
      dict->dict[key].okey = okey;
698
2.37M
  } else {
699
184k
      xmlDictEntryPtr entry;
700
701
184k
      entry = xmlMalloc(sizeof(xmlDictEntry));
702
184k
      if (entry != NULL) {
703
184k
    entry->name = olddict[i].name;
704
184k
    entry->len = olddict[i].len;
705
184k
    entry->okey = okey;
706
184k
    entry->next = dict->dict[key].next;
707
184k
    entry->valid = 1;
708
184k
    dict->dict[key].next = entry;
709
184k
      } else {
710
          /*
711
     * we don't have much ways to alert from here
712
     * result is losing an entry and unicity guarantee
713
     */
714
0
          ret = -1;
715
0
      }
716
184k
  }
717
#ifdef DEBUG_GROW
718
  nbElem++;
719
#endif
720
2.56M
    }
721
722
3.71M
    for (i = 0; i < oldsize; i++) {
723
3.68M
  iter = olddict[i].next;
724
6.20M
  while (iter) {
725
2.51M
      next = iter->next;
726
727
      /*
728
       * put back the entry in the new dict
729
       */
730
731
2.51M
      if (keep_keys)
732
23.3k
    okey = iter->okey;
733
2.49M
      else
734
2.49M
    okey = xmlDictComputeKey(dict, iter->name, iter->len);
735
2.51M
      key = okey % dict->size;
736
2.51M
      if (dict->dict[key].valid == 0) {
737
1.98M
    memcpy(&(dict->dict[key]), iter, sizeof(xmlDictEntry));
738
1.98M
    dict->dict[key].next = NULL;
739
1.98M
    dict->dict[key].valid = 1;
740
1.98M
    dict->dict[key].okey = okey;
741
1.98M
    xmlFree(iter);
742
1.98M
      } else {
743
533k
    iter->next = dict->dict[key].next;
744
533k
    iter->okey = okey;
745
533k
    dict->dict[key].next = iter;
746
533k
      }
747
748
#ifdef DEBUG_GROW
749
      nbElem++;
750
#endif
751
752
2.51M
      iter = next;
753
2.51M
  }
754
3.68M
    }
755
756
27.7k
    xmlFree(olddict);
757
758
#ifdef DEBUG_GROW
759
    xmlGenericError(xmlGenericErrorContext,
760
      "xmlDictGrow : from %lu to %lu, %u elems\n", oldsize, size, nbElem);
761
#endif
762
763
27.7k
    return(ret);
764
27.7k
}
765
766
/**
767
 * xmlDictFree:
768
 * @dict: the dictionary
769
 *
770
 * Free the hash @dict and its contents. The userdata is
771
 * deallocated with @f if provided.
772
 */
773
void
774
3.51M
xmlDictFree(xmlDictPtr dict) {
775
3.51M
    size_t i;
776
3.51M
    xmlDictEntryPtr iter;
777
3.51M
    xmlDictEntryPtr next;
778
3.51M
    int inside_dict = 0;
779
3.51M
    xmlDictStringsPtr pool, nextp;
780
781
3.51M
    if (dict == NULL)
782
0
  return;
783
784
    /* decrement the counter, it may be shared by a parser and docs */
785
3.51M
    xmlMutexLock(&xmlDictMutex);
786
3.51M
    dict->ref_counter--;
787
3.51M
    if (dict->ref_counter > 0) {
788
1.69M
        xmlMutexUnlock(&xmlDictMutex);
789
1.69M
        return;
790
1.69M
    }
791
792
1.81M
    xmlMutexUnlock(&xmlDictMutex);
793
794
1.81M
    if (dict->subdict != NULL) {
795
0
        xmlDictFree(dict->subdict);
796
0
    }
797
798
1.81M
    if (dict->dict) {
799
148M
  for(i = 0; ((i < dict->size) && (dict->nbElems > 0)); i++) {
800
146M
      iter = &(dict->dict[i]);
801
146M
      if (iter->valid == 0)
802
128M
    continue;
803
18.0M
      inside_dict = 1;
804
42.3M
      while (iter) {
805
24.3M
    next = iter->next;
806
24.3M
    if (!inside_dict)
807
6.26M
        xmlFree(iter);
808
24.3M
    dict->nbElems--;
809
24.3M
    inside_dict = 0;
810
24.3M
    iter = next;
811
24.3M
      }
812
18.0M
  }
813
1.81M
  xmlFree(dict->dict);
814
1.81M
    }
815
1.81M
    pool = dict->strings;
816
3.14M
    while (pool != NULL) {
817
1.32M
        nextp = pool->next;
818
1.32M
  xmlFree(pool);
819
1.32M
  pool = nextp;
820
1.32M
    }
821
1.81M
    xmlFree(dict);
822
1.81M
}
823
824
/**
825
 * xmlDictLookup:
826
 * @dict: the dictionary
827
 * @name: the name of the userdata
828
 * @len: the length of the name, if -1 it is recomputed
829
 *
830
 * Add the @name to the dictionary @dict if not present.
831
 *
832
 * Returns the internal copy of the name or NULL in case of internal error
833
 */
834
const xmlChar *
835
390M
xmlDictLookup(xmlDictPtr dict, const xmlChar *name, int len) {
836
390M
    unsigned long key, okey, nbi = 0;
837
390M
    xmlDictEntryPtr entry;
838
390M
    xmlDictEntryPtr insert;
839
390M
    const xmlChar *ret;
840
390M
    unsigned int l;
841
842
390M
    if ((dict == NULL) || (name == NULL))
843
9.45M
  return(NULL);
844
845
380M
    if (len < 0)
846
43.3M
        l = strlen((const char *) name);
847
337M
    else
848
337M
        l = len;
849
850
380M
    if (((dict->limit > 0) && (l >= dict->limit)) ||
851
380M
        (l > INT_MAX / 2))
852
0
        return(NULL);
853
854
    /*
855
     * Check for duplicate and insertion location.
856
     */
857
380M
    okey = xmlDictComputeKey(dict, name, l);
858
380M
    key = okey % dict->size;
859
380M
    if (dict->dict[key].valid == 0) {
860
16.1M
  insert = NULL;
861
364M
    } else {
862
442M
  for (insert = &(dict->dict[key]); insert->next != NULL;
863
364M
       insert = insert->next) {
864
121M
#ifdef __GNUC__
865
121M
      if ((insert->okey == okey) && (insert->len == l)) {
866
45.5M
    if (!memcmp(insert->name, name, l))
867
43.9M
        return(insert->name);
868
45.5M
      }
869
#else
870
      if ((insert->okey == okey) && (insert->len == l) &&
871
          (!xmlStrncmp(insert->name, name, l)))
872
    return(insert->name);
873
#endif
874
77.5M
      nbi++;
875
77.5M
  }
876
320M
#ifdef __GNUC__
877
320M
  if ((insert->okey == okey) && (insert->len == l)) {
878
313M
      if (!memcmp(insert->name, name, l))
879
312M
    return(insert->name);
880
313M
  }
881
#else
882
  if ((insert->okey == okey) && (insert->len == l) &&
883
      (!xmlStrncmp(insert->name, name, l)))
884
      return(insert->name);
885
#endif
886
320M
    }
887
888
24.2M
    if (dict->subdict) {
889
0
        unsigned long skey;
890
891
        /* we cannot always reuse the same okey for the subdict */
892
0
        if (((dict->size == MIN_DICT_SIZE) &&
893
0
       (dict->subdict->size != MIN_DICT_SIZE)) ||
894
0
            ((dict->size != MIN_DICT_SIZE) &&
895
0
       (dict->subdict->size == MIN_DICT_SIZE)))
896
0
      skey = xmlDictComputeKey(dict->subdict, name, l);
897
0
  else
898
0
      skey = okey;
899
900
0
  key = skey % dict->subdict->size;
901
0
  if (dict->subdict->dict[key].valid != 0) {
902
0
      xmlDictEntryPtr tmp;
903
904
0
      for (tmp = &(dict->subdict->dict[key]); tmp->next != NULL;
905
0
     tmp = tmp->next) {
906
0
#ifdef __GNUC__
907
0
    if ((tmp->okey == skey) && (tmp->len == l)) {
908
0
        if (!memcmp(tmp->name, name, l))
909
0
      return(tmp->name);
910
0
    }
911
#else
912
    if ((tmp->okey == skey) && (tmp->len == l) &&
913
        (!xmlStrncmp(tmp->name, name, l)))
914
        return(tmp->name);
915
#endif
916
0
    nbi++;
917
0
      }
918
0
#ifdef __GNUC__
919
0
      if ((tmp->okey == skey) && (tmp->len == l)) {
920
0
    if (!memcmp(tmp->name, name, l))
921
0
        return(tmp->name);
922
0
      }
923
#else
924
      if ((tmp->okey == skey) && (tmp->len == l) &&
925
    (!xmlStrncmp(tmp->name, name, l)))
926
    return(tmp->name);
927
#endif
928
0
  }
929
0
  key = okey % dict->size;
930
0
    }
931
932
24.2M
    ret = xmlDictAddString(dict, name, l);
933
24.2M
    if (ret == NULL)
934
0
        return(NULL);
935
24.2M
    if (insert == NULL) {
936
16.1M
  entry = &(dict->dict[key]);
937
16.1M
    } else {
938
8.04M
  entry = xmlMalloc(sizeof(xmlDictEntry));
939
8.04M
  if (entry == NULL)
940
0
       return(NULL);
941
8.04M
    }
942
24.2M
    entry->name = ret;
943
24.2M
    entry->len = l;
944
24.2M
    entry->next = NULL;
945
24.2M
    entry->valid = 1;
946
24.2M
    entry->okey = okey;
947
948
949
24.2M
    if (insert != NULL)
950
8.04M
  insert->next = entry;
951
952
24.2M
    dict->nbElems++;
953
954
24.2M
    if ((nbi > MAX_HASH_LEN) &&
955
24.2M
        (dict->size <= ((MAX_DICT_HASH / 2) / MAX_HASH_LEN))) {
956
27.0k
  if (xmlDictGrow(dict, MAX_HASH_LEN * 2 * dict->size) != 0)
957
0
      return(NULL);
958
27.0k
    }
959
    /* Note that entry may have been freed at this point by xmlDictGrow */
960
961
24.2M
    return(ret);
962
24.2M
}
963
964
/**
965
 * xmlDictExists:
966
 * @dict: the dictionary
967
 * @name: the name of the userdata
968
 * @len: the length of the name, if -1 it is recomputed
969
 *
970
 * Check if the @name exists in the dictionary @dict.
971
 *
972
 * Returns the internal copy of the name or NULL if not found.
973
 */
974
const xmlChar *
975
0
xmlDictExists(xmlDictPtr dict, const xmlChar *name, int len) {
976
0
    unsigned long key, okey;
977
0
    xmlDictEntryPtr insert;
978
0
    unsigned int l;
979
980
0
    if ((dict == NULL) || (name == NULL))
981
0
  return(NULL);
982
983
0
    if (len < 0)
984
0
        l = strlen((const char *) name);
985
0
    else
986
0
        l = len;
987
0
    if (((dict->limit > 0) && (l >= dict->limit)) ||
988
0
        (l > INT_MAX / 2))
989
0
        return(NULL);
990
991
    /*
992
     * Check for duplicate and insertion location.
993
     */
994
0
    okey = xmlDictComputeKey(dict, name, l);
995
0
    key = okey % dict->size;
996
0
    if (dict->dict[key].valid == 0) {
997
0
  insert = NULL;
998
0
    } else {
999
0
  for (insert = &(dict->dict[key]); insert->next != NULL;
1000
0
       insert = insert->next) {
1001
0
#ifdef __GNUC__
1002
0
      if ((insert->okey == okey) && (insert->len == l)) {
1003
0
    if (!memcmp(insert->name, name, l))
1004
0
        return(insert->name);
1005
0
      }
1006
#else
1007
      if ((insert->okey == okey) && (insert->len == l) &&
1008
          (!xmlStrncmp(insert->name, name, l)))
1009
    return(insert->name);
1010
#endif
1011
0
  }
1012
0
#ifdef __GNUC__
1013
0
  if ((insert->okey == okey) && (insert->len == l)) {
1014
0
      if (!memcmp(insert->name, name, l))
1015
0
    return(insert->name);
1016
0
  }
1017
#else
1018
  if ((insert->okey == okey) && (insert->len == l) &&
1019
      (!xmlStrncmp(insert->name, name, l)))
1020
      return(insert->name);
1021
#endif
1022
0
    }
1023
1024
0
    if (dict->subdict) {
1025
0
        unsigned long skey;
1026
1027
        /* we cannot always reuse the same okey for the subdict */
1028
0
        if (((dict->size == MIN_DICT_SIZE) &&
1029
0
       (dict->subdict->size != MIN_DICT_SIZE)) ||
1030
0
            ((dict->size != MIN_DICT_SIZE) &&
1031
0
       (dict->subdict->size == MIN_DICT_SIZE)))
1032
0
      skey = xmlDictComputeKey(dict->subdict, name, l);
1033
0
  else
1034
0
      skey = okey;
1035
1036
0
  key = skey % dict->subdict->size;
1037
0
  if (dict->subdict->dict[key].valid != 0) {
1038
0
      xmlDictEntryPtr tmp;
1039
1040
0
      for (tmp = &(dict->subdict->dict[key]); tmp->next != NULL;
1041
0
     tmp = tmp->next) {
1042
0
#ifdef __GNUC__
1043
0
    if ((tmp->okey == skey) && (tmp->len == l)) {
1044
0
        if (!memcmp(tmp->name, name, l))
1045
0
      return(tmp->name);
1046
0
    }
1047
#else
1048
    if ((tmp->okey == skey) && (tmp->len == l) &&
1049
        (!xmlStrncmp(tmp->name, name, l)))
1050
        return(tmp->name);
1051
#endif
1052
0
      }
1053
0
#ifdef __GNUC__
1054
0
      if ((tmp->okey == skey) && (tmp->len == l)) {
1055
0
    if (!memcmp(tmp->name, name, l))
1056
0
        return(tmp->name);
1057
0
      }
1058
#else
1059
      if ((tmp->okey == skey) && (tmp->len == l) &&
1060
    (!xmlStrncmp(tmp->name, name, l)))
1061
    return(tmp->name);
1062
#endif
1063
0
  }
1064
0
    }
1065
1066
    /* not found */
1067
0
    return(NULL);
1068
0
}
1069
1070
/**
1071
 * xmlDictQLookup:
1072
 * @dict: the dictionary
1073
 * @prefix: the prefix
1074
 * @name: the name
1075
 *
1076
 * Add the QName @prefix:@name to the hash @dict if not present.
1077
 *
1078
 * Returns the internal copy of the QName or NULL in case of internal error
1079
 */
1080
const xmlChar *
1081
316k
xmlDictQLookup(xmlDictPtr dict, const xmlChar *prefix, const xmlChar *name) {
1082
316k
    unsigned long okey, key, nbi = 0;
1083
316k
    xmlDictEntryPtr entry;
1084
316k
    xmlDictEntryPtr insert;
1085
316k
    const xmlChar *ret;
1086
316k
    unsigned int len, plen, l;
1087
1088
316k
    if ((dict == NULL) || (name == NULL))
1089
0
  return(NULL);
1090
316k
    if (prefix == NULL)
1091
0
        return(xmlDictLookup(dict, name, -1));
1092
1093
316k
    l = len = strlen((const char *) name);
1094
316k
    plen = strlen((const char *) prefix);
1095
316k
    len += 1 + plen;
1096
1097
    /*
1098
     * Check for duplicate and insertion location.
1099
     */
1100
316k
    okey = xmlDictComputeQKey(dict, prefix, plen, name, l);
1101
316k
    key = okey % dict->size;
1102
316k
    if (dict->dict[key].valid == 0) {
1103
72.3k
  insert = NULL;
1104
244k
    } else {
1105
310k
  for (insert = &(dict->dict[key]); insert->next != NULL;
1106
244k
       insert = insert->next) {
1107
144k
      if ((insert->okey == okey) && (insert->len == len) &&
1108
144k
          (xmlStrQEqual(prefix, name, insert->name)))
1109
78.7k
    return(insert->name);
1110
66.2k
      nbi++;
1111
66.2k
  }
1112
165k
  if ((insert->okey == okey) && (insert->len == len) &&
1113
165k
      (xmlStrQEqual(prefix, name, insert->name)))
1114
143k
      return(insert->name);
1115
165k
    }
1116
1117
94.5k
    if (dict->subdict) {
1118
0
        unsigned long skey;
1119
1120
        /* we cannot always reuse the same okey for the subdict */
1121
0
        if (((dict->size == MIN_DICT_SIZE) &&
1122
0
       (dict->subdict->size != MIN_DICT_SIZE)) ||
1123
0
            ((dict->size != MIN_DICT_SIZE) &&
1124
0
       (dict->subdict->size == MIN_DICT_SIZE)))
1125
0
      skey = xmlDictComputeQKey(dict->subdict, prefix, plen, name, l);
1126
0
  else
1127
0
      skey = okey;
1128
1129
0
  key = skey % dict->subdict->size;
1130
0
  if (dict->subdict->dict[key].valid != 0) {
1131
0
      xmlDictEntryPtr tmp;
1132
0
      for (tmp = &(dict->subdict->dict[key]); tmp->next != NULL;
1133
0
     tmp = tmp->next) {
1134
0
    if ((tmp->okey == skey) && (tmp->len == len) &&
1135
0
        (xmlStrQEqual(prefix, name, tmp->name)))
1136
0
        return(tmp->name);
1137
0
    nbi++;
1138
0
      }
1139
0
      if ((tmp->okey == skey) && (tmp->len == len) &&
1140
0
    (xmlStrQEqual(prefix, name, tmp->name)))
1141
0
    return(tmp->name);
1142
0
  }
1143
0
  key = okey % dict->size;
1144
0
    }
1145
1146
94.5k
    ret = xmlDictAddQString(dict, prefix, plen, name, l);
1147
94.5k
    if (ret == NULL)
1148
0
        return(NULL);
1149
94.5k
    if (insert == NULL) {
1150
72.3k
  entry = &(dict->dict[key]);
1151
72.3k
    } else {
1152
22.2k
  entry = xmlMalloc(sizeof(xmlDictEntry));
1153
22.2k
  if (entry == NULL)
1154
0
       return(NULL);
1155
22.2k
    }
1156
94.5k
    entry->name = ret;
1157
94.5k
    entry->len = len;
1158
94.5k
    entry->next = NULL;
1159
94.5k
    entry->valid = 1;
1160
94.5k
    entry->okey = okey;
1161
1162
94.5k
    if (insert != NULL)
1163
22.2k
  insert->next = entry;
1164
1165
94.5k
    dict->nbElems++;
1166
1167
94.5k
    if ((nbi > MAX_HASH_LEN) &&
1168
94.5k
        (dict->size <= ((MAX_DICT_HASH / 2) / MAX_HASH_LEN)))
1169
688
  xmlDictGrow(dict, MAX_HASH_LEN * 2 * dict->size);
1170
    /* Note that entry may have been freed at this point by xmlDictGrow */
1171
1172
94.5k
    return(ret);
1173
94.5k
}
1174
1175
/**
1176
 * xmlDictOwns:
1177
 * @dict: the dictionary
1178
 * @str: the string
1179
 *
1180
 * check if a string is owned by the dictionary
1181
 *
1182
 * Returns 1 if true, 0 if false and -1 in case of error
1183
 * -1 in case of error
1184
 */
1185
int
1186
238M
xmlDictOwns(xmlDictPtr dict, const xmlChar *str) {
1187
238M
    xmlDictStringsPtr pool;
1188
1189
238M
    if ((dict == NULL) || (str == NULL))
1190
0
  return(-1);
1191
238M
    pool = dict->strings;
1192
424M
    while (pool != NULL) {
1193
321M
        if ((str >= &pool->array[0]) && (str <= pool->free))
1194
136M
      return(1);
1195
185M
  pool = pool->next;
1196
185M
    }
1197
102M
    if (dict->subdict)
1198
0
        return(xmlDictOwns(dict->subdict, str));
1199
102M
    return(0);
1200
102M
}
1201
1202
/**
1203
 * xmlDictSize:
1204
 * @dict: the dictionary
1205
 *
1206
 * Query the number of elements installed in the hash @dict.
1207
 *
1208
 * Returns the number of elements in the dictionary or
1209
 * -1 in case of error
1210
 */
1211
int
1212
0
xmlDictSize(xmlDictPtr dict) {
1213
0
    if (dict == NULL)
1214
0
  return(-1);
1215
0
    if (dict->subdict)
1216
0
        return(dict->nbElems + dict->subdict->nbElems);
1217
0
    return(dict->nbElems);
1218
0
}
1219
1220
/**
1221
 * xmlDictSetLimit:
1222
 * @dict: the dictionary
1223
 * @limit: the limit in bytes
1224
 *
1225
 * Set a size limit for the dictionary
1226
 * Added in 2.9.0
1227
 *
1228
 * Returns the previous limit of the dictionary or 0
1229
 */
1230
size_t
1231
2.01M
xmlDictSetLimit(xmlDictPtr dict, size_t limit) {
1232
2.01M
    size_t ret;
1233
1234
2.01M
    if (dict == NULL)
1235
0
  return(0);
1236
2.01M
    ret = dict->limit;
1237
2.01M
    dict->limit = limit;
1238
2.01M
    return(ret);
1239
2.01M
}
1240
1241
/**
1242
 * xmlDictGetUsage:
1243
 * @dict: the dictionary
1244
 *
1245
 * Get how much memory is used by a dictionary for strings
1246
 * Added in 2.9.0
1247
 *
1248
 * Returns the amount of strings allocated
1249
 */
1250
size_t
1251
0
xmlDictGetUsage(xmlDictPtr dict) {
1252
0
    xmlDictStringsPtr pool;
1253
0
    size_t limit = 0;
1254
1255
0
    if (dict == NULL)
1256
0
  return(0);
1257
0
    pool = dict->strings;
1258
0
    while (pool != NULL) {
1259
0
        limit += pool->size;
1260
0
  pool = pool->next;
1261
0
    }
1262
0
    return(limit);
1263
0
}
1264