Coverage Report

Created: 2025-07-23 06:33

/src/php-src/Zend/zend_string.c
Line
Count
Source (jump to first uncovered line)
1
/*
2
   +----------------------------------------------------------------------+
3
   | Zend Engine                                                          |
4
   +----------------------------------------------------------------------+
5
   | Copyright (c) Zend Technologies Ltd. (http://www.zend.com)           |
6
   +----------------------------------------------------------------------+
7
   | This source file is subject to version 2.00 of the Zend license,     |
8
   | that is bundled with this package in the file LICENSE, and is        |
9
   | available through the world-wide-web at the following url:           |
10
   | http://www.zend.com/license/2_00.txt.                                |
11
   | If you did not receive a copy of the Zend license and are unable to  |
12
   | obtain it through the world-wide-web, please send a note to          |
13
   | license@zend.com so we can mail you a copy immediately.              |
14
   +----------------------------------------------------------------------+
15
   | Authors: Dmitry Stogov <dmitry@php.net>                              |
16
   +----------------------------------------------------------------------+
17
*/
18
19
#include "zend.h"
20
#include "zend_globals.h"
21
22
#ifdef HAVE_VALGRIND
23
# include "valgrind/callgrind.h"
24
#endif
25
26
#if __has_feature(memory_sanitizer)
27
# include <sanitizer/msan_interface.h>
28
#endif
29
30
ZEND_API zend_new_interned_string_func_t zend_new_interned_string;
31
ZEND_API zend_string_init_interned_func_t zend_string_init_interned;
32
ZEND_API zend_string_init_existing_interned_func_t zend_string_init_existing_interned;
33
34
static zend_string* ZEND_FASTCALL zend_new_interned_string_permanent(zend_string *str);
35
static zend_string* ZEND_FASTCALL zend_new_interned_string_request(zend_string *str);
36
static zend_string* ZEND_FASTCALL zend_string_init_interned_permanent(const char *str, size_t size, bool permanent);
37
static zend_string* ZEND_FASTCALL zend_string_init_existing_interned_permanent(const char *str, size_t size, bool permanent);
38
static zend_string* ZEND_FASTCALL zend_string_init_interned_request(const char *str, size_t size, bool permanent);
39
static zend_string* ZEND_FASTCALL zend_string_init_existing_interned_request(const char *str, size_t size, bool permanent);
40
41
/* Any strings interned in the startup phase. Common to all the threads,
42
   won't be free'd until process exit. If we want an ability to
43
   add permanent strings even after startup, it would be still
44
   possible on costs of locking in the thread safe builds. */
45
static HashTable interned_strings_permanent;
46
47
static zend_new_interned_string_func_t interned_string_request_handler = zend_new_interned_string_request;
48
static zend_string_init_interned_func_t interned_string_init_request_handler = zend_string_init_interned_request;
49
static zend_string_init_existing_interned_func_t interned_string_init_existing_request_handler = zend_string_init_existing_interned_request;
50
51
ZEND_API zend_string  *zend_empty_string = NULL;
52
ZEND_API zend_string  *zend_one_char_string[256];
53
ZEND_API zend_string **zend_known_strings = NULL;
54
55
ZEND_API zend_ulong ZEND_FASTCALL zend_string_hash_func(zend_string *str)
56
7.85M
{
57
7.85M
  return ZSTR_H(str) = zend_hash_func(ZSTR_VAL(str), ZSTR_LEN(str));
58
7.85M
}
59
60
ZEND_API zend_ulong ZEND_FASTCALL zend_hash_func(const char *str, size_t len)
61
8.24M
{
62
8.24M
  return zend_inline_hash_func(str, len);
63
8.24M
}
64
65
static void _str_dtor(zval *zv)
66
347k
{
67
347k
  zend_string *str = Z_STR_P(zv);
68
347k
  pefree(str, GC_FLAGS(str) & IS_STR_PERSISTENT);
69
347k
}
70
71
static const char *known_strings[] = {
72
#define _ZEND_STR_DSC(id, str) str,
73
ZEND_KNOWN_STRINGS(_ZEND_STR_DSC)
74
#undef _ZEND_STR_DSC
75
  NULL
76
};
77
78
static zend_always_inline void zend_init_interned_strings_ht(HashTable *interned_strings, bool permanent)
79
268k
{
80
268k
  zend_hash_init(interned_strings, 1024, NULL, _str_dtor, permanent);
81
268k
  if (permanent) {
82
16
    zend_hash_real_init_mixed(interned_strings);
83
16
  }
84
268k
}
85
86
ZEND_API void zend_interned_strings_init(void)
87
16
{
88
16
  char s[2];
89
90
16
  interned_string_request_handler = zend_new_interned_string_request;
91
16
  interned_string_init_request_handler = zend_string_init_interned_request;
92
16
  interned_string_init_existing_request_handler = zend_string_init_existing_interned_request;
93
94
16
  zend_empty_string = NULL;
95
16
  zend_known_strings = NULL;
96
97
16
  zend_init_interned_strings_ht(&interned_strings_permanent, 1);
98
99
16
  zend_new_interned_string = zend_new_interned_string_permanent;
100
16
  zend_string_init_interned = zend_string_init_interned_permanent;
101
16
  zend_string_init_existing_interned = zend_string_init_existing_interned_permanent;
102
103
  /* interned empty string */
104
16
  zend_empty_string = zend_string_init_interned_permanent("", 0, true);
105
16
  GC_ADD_FLAGS(zend_empty_string, IS_STR_VALID_UTF8);
106
107
16
  s[1] = 0;
108
4.11k
  for (size_t i = 0; i < 256; i++) {
109
4.09k
    s[0] = i;
110
4.09k
    zend_one_char_string[i] = zend_string_init_interned_permanent(s, 1, true);
111
4.09k
    if (i < 0x80) {
112
2.04k
      GC_ADD_FLAGS(zend_one_char_string[i], IS_STR_VALID_UTF8);
113
2.04k
    }
114
4.09k
  }
115
116
  /* known strings */
117
16
  zend_known_strings = pemalloc(sizeof(zend_string*) * ((sizeof(known_strings) / sizeof(known_strings[0]) - 1)), 1);
118
1.42k
  for (size_t i = 0; i < (sizeof(known_strings) / sizeof(known_strings[0])) - 1; i++) {
119
1.40k
    zend_known_strings[i] = zend_string_init_interned_permanent(known_strings[i], strlen(known_strings[i]), true);
120
1.40k
    GC_ADD_FLAGS(zend_known_strings[i], IS_STR_VALID_UTF8);
121
1.40k
  }
122
16
}
123
124
ZEND_API void zend_interned_strings_dtor(void)
125
0
{
126
0
  zend_hash_destroy(&interned_strings_permanent);
127
128
0
  free(zend_known_strings);
129
0
  zend_known_strings = NULL;
130
0
}
131
132
static zend_always_inline zend_string *zend_interned_string_ht_lookup_ex(zend_ulong h, const char *str, size_t size, HashTable *interned_strings)
133
2.07M
{
134
2.07M
  uint32_t nIndex;
135
2.07M
  uint32_t idx;
136
2.07M
  Bucket *p;
137
138
2.07M
  nIndex = h | interned_strings->nTableMask;
139
2.07M
  idx = HT_HASH(interned_strings, nIndex);
140
2.42M
  while (idx != HT_INVALID_IDX) {
141
1.41M
    p = HT_HASH_TO_BUCKET(interned_strings, idx);
142
1.41M
    if ((p->h == h) && zend_string_equals_cstr(p->key, str, size)) {
143
1.06M
      return p->key;
144
1.06M
    }
145
347k
    idx = Z_NEXT(p->val);
146
347k
  }
147
148
1.01M
  return NULL;
149
2.07M
}
150
151
static zend_always_inline zend_string *zend_interned_string_ht_lookup(zend_string *str, HashTable *interned_strings)
152
5.56M
{
153
5.56M
  zend_ulong h = ZSTR_H(str);
154
5.56M
  uint32_t nIndex;
155
5.56M
  uint32_t idx;
156
5.56M
  Bucket *p;
157
158
5.56M
  nIndex = h | interned_strings->nTableMask;
159
5.56M
  idx = HT_HASH(interned_strings, nIndex);
160
7.04M
  while (idx != HT_INVALID_IDX) {
161
4.17M
    p = HT_HASH_TO_BUCKET(interned_strings, idx);
162
4.17M
    if ((p->h == h) && zend_string_equal_content(p->key, str)) {
163
2.69M
      return p->key;
164
2.69M
    }
165
1.48M
    idx = Z_NEXT(p->val);
166
1.48M
  }
167
168
2.87M
  return NULL;
169
5.56M
}
170
171
/* This function might be not thread safe at least because it would update the
172
   hash val in the passed string. Be sure it is called in the appropriate context. */
173
static zend_always_inline zend_string *zend_add_interned_string(zend_string *str, HashTable *interned_strings, uint32_t flags)
174
401k
{
175
401k
  zval val;
176
177
401k
  GC_SET_REFCOUNT(str, 1);
178
401k
  GC_ADD_FLAGS(str, IS_STR_INTERNED | flags);
179
180
401k
  ZVAL_INTERNED_STR(&val, str);
181
182
401k
  zend_hash_add_new(interned_strings, str, &val);
183
184
401k
  return str;
185
401k
}
186
187
ZEND_API zend_string* ZEND_FASTCALL zend_interned_string_find_permanent(zend_string *str)
188
0
{
189
0
  zend_string_hash_val(str);
190
0
  return zend_interned_string_ht_lookup(str, &interned_strings_permanent);
191
0
}
192
193
static zend_string* ZEND_FASTCALL zend_init_string_for_interning(zend_string *str, bool persistent)
194
147k
{
195
147k
  uint32_t flags = ZSTR_GET_COPYABLE_CONCAT_PROPERTIES(str);
196
147k
  zend_ulong h = ZSTR_H(str);
197
147k
  zend_string_delref(str);
198
147k
  str = zend_string_init(ZSTR_VAL(str), ZSTR_LEN(str), persistent);
199
147k
  GC_ADD_FLAGS(str, flags);
200
147k
  ZSTR_H(str) = h;
201
147k
  return str;
202
147k
}
203
204
static zend_string* ZEND_FASTCALL zend_new_interned_string_permanent(zend_string *str)
205
35.4k
{
206
35.4k
  zend_string *ret;
207
208
35.4k
  if (ZSTR_IS_INTERNED(str)) {
209
22.4k
    return str;
210
22.4k
  }
211
212
13.0k
  zend_string_hash_val(str);
213
13.0k
  ret = zend_interned_string_ht_lookup(str, &interned_strings_permanent);
214
13.0k
  if (ret) {
215
4.22k
    zend_string_release(str);
216
4.22k
    return ret;
217
4.22k
  }
218
219
8.85k
  ZEND_ASSERT(GC_FLAGS(str) & GC_PERSISTENT);
220
8.85k
  if (GC_REFCOUNT(str) > 1) {
221
338
    str = zend_init_string_for_interning(str, true);
222
338
  }
223
224
8.85k
  return zend_add_interned_string(str, &interned_strings_permanent, IS_STR_PERMANENT);
225
8.85k
}
226
227
static zend_string* ZEND_FASTCALL zend_new_interned_string_request(zend_string *str)
228
3.71M
{
229
3.71M
  zend_string *ret;
230
231
3.71M
  if (ZSTR_IS_INTERNED(str)) {
232
745k
    return str;
233
745k
  }
234
235
2.96M
  zend_string_hash_val(str);
236
237
  /* Check for permanent strings, the table is readonly at this point. */
238
2.96M
  ret = zend_interned_string_ht_lookup(str, &interned_strings_permanent);
239
2.96M
  if (ret) {
240
385k
    zend_string_release(str);
241
385k
    return ret;
242
385k
  }
243
244
2.58M
  ret = zend_interned_string_ht_lookup(str, &CG(interned_strings));
245
2.58M
  if (ret) {
246
2.30M
    zend_string_release(str);
247
2.30M
    return ret;
248
2.30M
  }
249
250
  /* Create a short living interned, freed after the request. */
251
#if ZEND_RC_DEBUG
252
  if (zend_rc_debug) {
253
    /* PHP shouldn't create persistent interned string during request,
254
     * but at least dl() may do this */
255
    ZEND_ASSERT(!(GC_FLAGS(str) & GC_PERSISTENT));
256
  }
257
#endif
258
280k
  if (GC_REFCOUNT(str) > 1) {
259
147k
    str = zend_init_string_for_interning(str, false);
260
147k
  }
261
262
280k
  ret = zend_add_interned_string(str, &CG(interned_strings), 0);
263
264
280k
  return ret;
265
2.58M
}
266
267
static zend_string* ZEND_FASTCALL zend_string_init_interned_permanent(const char *str, size_t size, bool permanent)
268
64.3k
{
269
64.3k
  zend_string *ret;
270
64.3k
  zend_ulong h = zend_inline_hash_func(str, size);
271
272
64.3k
  ret = zend_interned_string_ht_lookup_ex(h, str, size, &interned_strings_permanent);
273
64.3k
  if (ret) {
274
19.1k
    return ret;
275
19.1k
  }
276
277
45.2k
  ZEND_ASSERT(permanent);
278
45.2k
  ret = zend_string_init(str, size, permanent);
279
45.2k
  ZSTR_H(ret) = h;
280
45.2k
  return zend_add_interned_string(ret, &interned_strings_permanent, IS_STR_PERMANENT);
281
45.2k
}
282
283
static zend_string* ZEND_FASTCALL zend_string_init_existing_interned_permanent(const char *str, size_t size, bool permanent)
284
0
{
285
0
  zend_ulong h = zend_inline_hash_func(str, size);
286
0
  zend_string *ret = zend_interned_string_ht_lookup_ex(h, str, size, &interned_strings_permanent);
287
0
  if (ret) {
288
0
    return ret;
289
0
  }
290
291
0
  ZEND_ASSERT(permanent);
292
0
  ret = zend_string_init(str, size, permanent);
293
0
  ZSTR_H(ret) = h;
294
0
  return ret;
295
0
}
296
297
static zend_string* ZEND_FASTCALL zend_string_init_interned_request(const char *str, size_t size, bool permanent)
298
880k
{
299
880k
  zend_string *ret;
300
880k
  zend_ulong h = zend_inline_hash_func(str, size);
301
302
  /* Check for permanent strings, the table is readonly at this point. */
303
880k
  ret = zend_interned_string_ht_lookup_ex(h, str, size, &interned_strings_permanent);
304
880k
  if (ret) {
305
511k
    return ret;
306
511k
  }
307
308
368k
  ret = zend_interned_string_ht_lookup_ex(h, str, size, &CG(interned_strings));
309
368k
  if (ret) {
310
302k
    return ret;
311
302k
  }
312
313
#if ZEND_RC_DEBUG
314
  if (zend_rc_debug) {
315
    /* PHP shouldn't create persistent interned string during request,
316
     * but at least dl() may do this */
317
    ZEND_ASSERT(!permanent);
318
  }
319
#endif
320
66.6k
  ret = zend_string_init(str, size, permanent);
321
66.6k
  ZSTR_H(ret) = h;
322
323
  /* Create a short living interned, freed after the request. */
324
66.6k
  return zend_add_interned_string(ret, &CG(interned_strings), 0);
325
368k
}
326
327
static zend_string* ZEND_FASTCALL zend_string_init_existing_interned_request(const char *str, size_t size, bool permanent)
328
489k
{
329
489k
  zend_ulong h = zend_inline_hash_func(str, size);
330
489k
  zend_string *ret = zend_interned_string_ht_lookup_ex(h, str, size, &interned_strings_permanent);
331
489k
  if (ret) {
332
214k
    return ret;
333
214k
  }
334
335
274k
  ret = zend_interned_string_ht_lookup_ex(h, str, size, &CG(interned_strings));
336
274k
  if (ret) {
337
15.5k
    return ret;
338
15.5k
  }
339
340
259k
  ZEND_ASSERT(!permanent);
341
259k
  ret = zend_string_init(str, size, permanent);
342
259k
  ZSTR_H(ret) = h;
343
259k
  return ret;
344
259k
}
345
346
ZEND_API void zend_interned_strings_activate(void)
347
268k
{
348
268k
  zend_init_interned_strings_ht(&CG(interned_strings), 0);
349
268k
}
350
351
ZEND_API void zend_interned_strings_deactivate(void)
352
268k
{
353
268k
  zend_hash_destroy(&CG(interned_strings));
354
268k
}
355
356
ZEND_API void zend_interned_strings_set_request_storage_handlers(zend_new_interned_string_func_t handler, zend_string_init_interned_func_t init_handler, zend_string_init_existing_interned_func_t init_existing_handler)
357
4
{
358
4
  interned_string_request_handler = handler;
359
4
  interned_string_init_request_handler = init_handler;
360
4
  interned_string_init_existing_request_handler = init_existing_handler;
361
4
}
362
363
ZEND_API void zend_interned_strings_switch_storage(bool request)
364
16
{
365
16
  if (request) {
366
16
    zend_new_interned_string = interned_string_request_handler;
367
16
    zend_string_init_interned = interned_string_init_request_handler;
368
16
    zend_string_init_existing_interned = interned_string_init_existing_request_handler;
369
16
  } else {
370
0
    zend_new_interned_string = zend_new_interned_string_permanent;
371
0
    zend_string_init_interned = zend_string_init_interned_permanent;
372
0
    zend_string_init_existing_interned = zend_string_init_existing_interned_permanent;
373
0
  }
374
16
}
375
376
#if defined(__GNUC__) && (defined(__i386__) || (defined(__x86_64__) && !defined(__ILP32__)))
377
/* Even if we don't build with valgrind support, include the symbol so that valgrind available
378
 * only at runtime will not result in false positives. */
379
#ifndef I_REPLACE_SONAME_FNNAME_ZU
380
# define I_REPLACE_SONAME_FNNAME_ZU(soname, fnname) _vgr00000ZU_ ## soname ## _ ## fnname
381
#endif
382
383
/* See GH-9068 */
384
#if __has_attribute(noipa)
385
# define NOIPA __attribute__((noipa))
386
#else
387
# define NOIPA
388
#endif
389
390
ZEND_API bool ZEND_FASTCALL I_REPLACE_SONAME_FNNAME_ZU(NONE,zend_string_equal_val)(const zend_string *s1, const zend_string *s2)
391
0
{
392
0
  return !memcmp(ZSTR_VAL(s1), ZSTR_VAL(s2), ZSTR_LEN(s1));
393
0
}
394
#endif
395
396
#if defined(__GNUC__) && defined(__i386__)
397
ZEND_API zend_never_inline NOIPA bool ZEND_FASTCALL zend_string_equal_val(const zend_string *s1, const zend_string *s2)
398
{
399
  const char *ptr = ZSTR_VAL(s1);
400
  uintptr_t delta = (uintptr_t) s2 - (uintptr_t) s1;
401
  size_t len = ZSTR_LEN(s1);
402
  zend_ulong ret;
403
404
  __asm__ (
405
    "0:\n\t"
406
    "movl (%2,%3), %0\n\t"
407
    "xorl (%2), %0\n\t"
408
    "jne 1f\n\t"
409
    "addl $0x4, %2\n\t"
410
    "subl $0x4, %1\n\t"
411
    "ja 0b\n\t"
412
    "movl $0x1, %0\n\t"
413
    "jmp 3f\n\t"
414
    "1:\n\t"
415
    "cmpl $0x4,%1\n\t"
416
    "jb 2f\n\t"
417
    "xorl %0, %0\n\t"
418
    "jmp 3f\n\t"
419
    "2:\n\t"
420
    "negl %1\n\t"
421
    "lea 0x20(,%1,8), %1\n\t"
422
    "shll %b1, %0\n\t"
423
    "sete %b0\n\t"
424
    "movzbl %b0, %0\n\t"
425
    "3:\n"
426
    : "=&a"(ret),
427
      "+c"(len),
428
      "+r"(ptr)
429
    : "r"(delta)
430
    : "cc");
431
  return ret;
432
}
433
434
#elif defined(__GNUC__) && defined(__x86_64__) && !defined(__ILP32__)
435
ZEND_API zend_never_inline NOIPA bool ZEND_FASTCALL zend_string_equal_val(const zend_string *s1, const zend_string *s2)
436
7.02M
{
437
7.02M
  const char *ptr = ZSTR_VAL(s1);
438
7.02M
  uintptr_t delta = (uintptr_t) s2 - (uintptr_t) s1;
439
7.02M
  size_t len = ZSTR_LEN(s1);
440
7.02M
  zend_ulong ret;
441
442
7.02M
  __asm__ (
443
7.02M
    "0:\n\t"
444
7.02M
    "movq (%2,%3), %0\n\t"
445
7.02M
    "xorq (%2), %0\n\t"
446
7.02M
    "jne 1f\n\t"
447
7.02M
    "addq $0x8, %2\n\t"
448
7.02M
    "subq $0x8, %1\n\t"
449
7.02M
    "ja 0b\n\t"
450
7.02M
    "movq $0x1, %0\n\t"
451
7.02M
    "jmp 3f\n\t"
452
7.02M
    "1:\n\t"
453
7.02M
    "cmpq $0x8,%1\n\t"
454
7.02M
    "jb 2f\n\t"
455
7.02M
    "xorq %0, %0\n\t"
456
7.02M
    "jmp 3f\n\t"
457
7.02M
    "2:\n\t"
458
7.02M
    "negq %1\n\t"
459
7.02M
    "lea 0x40(,%1,8), %1\n\t"
460
7.02M
    "shlq %b1, %0\n\t"
461
7.02M
    "sete %b0\n\t"
462
7.02M
    "movzbq %b0, %0\n\t"
463
7.02M
    "3:\n"
464
7.02M
    : "=&a"(ret),
465
7.02M
      "+c"(len),
466
7.02M
      "+r"(ptr)
467
7.02M
    : "r"(delta)
468
7.02M
    : "cc");
469
7.02M
  return ret;
470
7.02M
}
471
#endif
472
473
ZEND_API zend_string *zend_string_concat2(
474
    const char *str1, size_t str1_len,
475
    const char *str2, size_t str2_len)
476
263k
{
477
263k
  size_t len = str1_len + str2_len;
478
263k
  zend_string *res = zend_string_alloc(len, 0);
479
480
263k
  memcpy(ZSTR_VAL(res), str1, str1_len);
481
263k
  memcpy(ZSTR_VAL(res) + str1_len, str2, str2_len);
482
263k
  ZSTR_VAL(res)[len] = '\0';
483
484
263k
  return res;
485
263k
}
486
487
ZEND_API zend_string *zend_string_concat3(
488
    const char *str1, size_t str1_len,
489
    const char *str2, size_t str2_len,
490
    const char *str3, size_t str3_len)
491
1.27M
{
492
1.27M
  size_t len = str1_len + str2_len + str3_len;
493
1.27M
  zend_string *res = zend_string_alloc(len, 0);
494
495
1.27M
  memcpy(ZSTR_VAL(res), str1, str1_len);
496
1.27M
  memcpy(ZSTR_VAL(res) + str1_len, str2, str2_len);
497
1.27M
  memcpy(ZSTR_VAL(res) + str1_len + str2_len, str3, str3_len);
498
1.27M
  ZSTR_VAL(res)[len] = '\0';
499
500
1.27M
  return res;
501
1.27M
}
502
503
/* strlcpy and strlcat are not intercepted by msan, so we need to do it ourselves. */
504
#if __has_feature(memory_sanitizer)
505
static size_t (*libc_strlcpy)(char *__restrict, const char *__restrict, size_t);
506
size_t strlcpy(char *__restrict dest, const char *__restrict src, size_t n)
507
{
508
  if (!libc_strlcpy) {
509
    libc_strlcpy = dlsym(RTLD_NEXT, "strlcpy");
510
  }
511
  size_t result = libc_strlcpy(dest, src, n);
512
  __msan_unpoison_string(dest);
513
  return result;
514
}
515
static size_t (*libc_strlcat)(char *__restrict, const char *__restrict, size_t);
516
size_t strlcat (char *__restrict dest, const char *restrict src, size_t n)
517
{
518
  if (!libc_strlcat) {
519
    libc_strlcat = dlsym(RTLD_NEXT, "strlcat");
520
  }
521
  size_t result = libc_strlcat(dest, src, n);
522
  __msan_unpoison_string(dest);
523
  return result;
524
}
525
#endif