1 #include "first.h" 2 3 #include "buffer.h" 4 5 #include <stdlib.h> 6 #include <string.h> 7 #include "sys-time.h" /* strftime() */ 8 9 static const char hex_chars_lc[] = "0123456789abcdef"; 10 static const char hex_chars_uc[] = "0123456789ABCDEF"; 11 12 13 __attribute_noinline__ 14 buffer* buffer_init(void) { 15 #if 0 /* buffer_init() and chunk_init() can be hot, 16 * so avoid the additional hop of indirection */ 17 return ck_calloc(1, sizeof(buffer)); 18 #else 19 buffer * const b = calloc(1, sizeof(*b)); 20 force_assert(b); 21 return b; 22 #endif 23 } 24 25 void buffer_free(buffer *b) { 26 if (NULL == b) return; 27 free(b->ptr); 28 free(b); 29 } 30 31 void buffer_free_ptr(buffer *b) { 32 free(b->ptr); 33 b->ptr = NULL; 34 b->used = 0; 35 b->size = 0; 36 } 37 38 void buffer_move(buffer * restrict b, buffer * restrict src) { 39 buffer tmp; 40 buffer_clear(b); 41 tmp = *src; *src = *b; *b = tmp; 42 } 43 44 /* make sure buffer is at least "size" big + 1 for '\0'. keep old data */ 45 __attribute_cold__ 46 __attribute_noinline__ 47 __attribute_nonnull__() 48 __attribute_returns_nonnull__ 49 static char* buffer_realloc(buffer * const restrict b, const size_t len) { 50 #define BUFFER_PIECE_SIZE 64uL /*(must be power-of-2)*/ 51 size_t sz = (len + 1 + BUFFER_PIECE_SIZE-1) & ~(BUFFER_PIECE_SIZE-1); 52 force_assert(sz > len); 53 if ((sz & (sz-1)) && sz < INT_MAX) {/* not power-2; huge val not expected */ 54 /*(optimizer should recognize this and use ffs or clz or equivalent)*/ 55 const size_t psz = sz; 56 for (sz = 256; sz < psz; sz <<= 1) ; 57 } 58 sz |= 1; /*(extra +1 for '\0' when needed buffer size is exact power-2)*/ 59 60 b->size = sz; 61 b->ptr = realloc(b->ptr, sz); 62 63 force_assert(NULL != b->ptr); 64 return b->ptr; 65 } 66 67 __attribute_cold__ 68 __attribute_noinline__ 69 __attribute_nonnull__() 70 __attribute_returns_nonnull__ 71 static char* buffer_alloc_replace(buffer * const restrict b, const size_t size) { 72 /*(discard old data so realloc() does not copy)*/ 73 if (NULL != b->ptr) { 74 free(b->ptr); 75 b->ptr = NULL; 76 } 77 /*(note: if size larger than one lshift, use size instead of power-2)*/ 78 const size_t bsize2x = (b->size & ~1uL) << 1; 79 return buffer_realloc(b, bsize2x > size ? bsize2x-1 : size); 80 } 81 82 char* buffer_string_prepare_copy(buffer * const b, const size_t size) { 83 b->used = 0; 84 #ifdef __COVERITY__ /*(b->ptr is not NULL if b->size is not 0)*/ 85 force_assert(size >= b->size || b->ptr); 86 #endif 87 return (size < b->size) 88 ? b->ptr 89 : buffer_alloc_replace(b, size); 90 } 91 92 __attribute_cold__ 93 __attribute_noinline__ 94 __attribute_nonnull__() 95 __attribute_returns_nonnull__ 96 static char* buffer_string_prepare_append_resize(buffer * const restrict b, const size_t size) { 97 if (b->used < 2) { /* buffer_is_blank(b) */ 98 char * const s = buffer_string_prepare_copy(b, size); 99 *s = '\0'; /*(for case (1 == b->used))*/ 100 return s; 101 } 102 103 /* not empty, b->used already includes a terminating 0 */ 104 /*(note: if size larger than one lshift, use size instead of power-2)*/ 105 const size_t bsize2x = (b->size & ~1uL) << 1; 106 const size_t req_size = (bsize2x - b->used > size) 107 ? bsize2x-1 108 : b->used + size; 109 110 /* check for overflow: unsigned overflow is defined to wrap around */ 111 force_assert(req_size >= b->used); 112 113 return buffer_realloc(b, req_size) + b->used - 1; 114 } 115 116 char* buffer_string_prepare_append(buffer * const b, const size_t size) { 117 const uint32_t len = b->used ? b->used-1 : 0; 118 return (b->size - len >= size + 1) 119 ? b->ptr + len 120 : buffer_string_prepare_append_resize(b, size); 121 } 122 123 /*(prefer smaller code than inlining buffer_extend in many places in buffer.c)*/ 124 __attribute_noinline__ 125 char* 126 buffer_extend (buffer * const b, const size_t x) 127 { 128 /* extend buffer to append x (reallocate by power-2 (or larger), if needed) 129 * (combine buffer_string_prepare_append() and buffer_commit()) 130 * (future: might make buffer.h static inline func for HTTP/1.1 performance) 131 * pre-sets '\0' byte and b->used (unlike buffer_string_prepare_append())*/ 132 #if 0 133 char * const s = buffer_string_prepare_append(b, x); 134 b->used += x + (0 == b->used); 135 #else 136 const uint32_t len = b->used ? b->used-1 : 0; 137 char * const s = (b->size - len >= x + 1) 138 ? b->ptr + len 139 : buffer_string_prepare_append_resize(b, x); 140 b->used = len+x+1; 141 #endif 142 s[x] = '\0'; 143 return s; 144 } 145 146 void buffer_commit(buffer *b, size_t size) 147 { 148 size_t sz = b->used; 149 if (0 == sz) sz = 1; 150 151 if (size > 0) { 152 /* check for overflow: unsigned overflow is defined to wrap around */ 153 sz += size; 154 force_assert(sz > size); 155 } 156 157 b->used = sz; 158 b->ptr[sz - 1] = '\0'; 159 } 160 161 __attribute_cold__ /*(reduce code size due to inlining)*/ 162 void buffer_copy_string(buffer * restrict b, const char * restrict s) { 163 if (__builtin_expect( (NULL == s), 0)) s = ""; 164 buffer_copy_string_len(b, s, strlen(s)); 165 } 166 167 void buffer_copy_string_len(buffer * const restrict b, const char * const restrict s, const size_t len) { 168 b->used = len + 1; 169 char * const restrict d = (len < b->size) 170 ? b->ptr 171 : buffer_alloc_replace(b, len); 172 d[len] = '\0'; 173 memcpy(d, s, len); 174 } 175 176 __attribute_cold__ /*(reduce code size due to inlining)*/ 177 void buffer_append_string(buffer * restrict b, const char * restrict s) { 178 if (__builtin_expect( (NULL == s), 0)) s = ""; 179 buffer_append_string_len(b, s, strlen(s)); 180 } 181 182 /** 183 * append a string to the end of the buffer 184 * 185 * the resulting buffer is terminated with a '\0' 186 * s is treated as a un-terminated string (a \0 is handled a normal character) 187 * 188 * @param b a buffer 189 * @param s the string 190 * @param s_len size of the string (without the terminating \0) 191 */ 192 193 void buffer_append_string_len(buffer * const restrict b, const char * const restrict s, const size_t len) { 194 memcpy(buffer_extend(b, len), s, len); 195 } 196 197 void buffer_append_str2(buffer * const restrict b, const char * const s1, const size_t len1, const char * const s2, const size_t len2) { 198 char * const restrict s = buffer_extend(b, len1+len2); 199 #ifdef HAVE_MEMPCPY 200 mempcpy(mempcpy(s, s1, len1), s2, len2); 201 #else 202 memcpy(s, s1, len1); 203 memcpy(s+len1, s2, len2); 204 #endif 205 } 206 207 void buffer_append_str3(buffer * const restrict b, const char * const s1, const size_t len1, const char * const s2, const size_t len2, const char * const s3, const size_t len3) { 208 char * restrict s = buffer_extend(b, len1+len2+len3); 209 #ifdef HAVE_MEMPCPY 210 mempcpy(mempcpy(mempcpy(s, s1, len1), s2, len2), s3, len3); 211 #else 212 memcpy(s, s1, len1); 213 memcpy((s+=len1), s2, len2); 214 memcpy((s+=len2), s3, len3); 215 #endif 216 } 217 218 void buffer_append_iovec(buffer * const restrict b, const struct const_iovec * const iov, const size_t n) { 219 size_t len = 0; 220 for (size_t i = 0; i < n; ++i) 221 len += iov[i].iov_len; 222 char *s = buffer_extend(b, len); 223 for (size_t i = 0; i < n; ++i) { 224 if (0 == iov[i].iov_len) continue; 225 #ifdef HAVE_MEMPCPY 226 s = mempcpy(s, iov[i].iov_base, iov[i].iov_len); 227 #else 228 memcpy(s, iov[i].iov_base, iov[i].iov_len); 229 s += iov[i].iov_len; 230 #endif 231 } 232 } 233 234 void buffer_append_path_len(buffer * restrict b, const char * restrict a, size_t alen) { 235 char * restrict s = buffer_string_prepare_append(b, alen+1); 236 const int aslash = (alen && a[0] == '/'); 237 if (b->used > 1 && s[-1] == '/') { 238 if (aslash) { 239 ++a; 240 --alen; 241 } 242 } 243 else { 244 if (0 == b->used) b->used = 1; 245 if (!aslash) { 246 *s++ = '/'; 247 ++b->used; 248 } 249 } 250 b->used += alen; 251 s[alen] = '\0'; 252 memcpy(s, a, alen); 253 } 254 255 void 256 buffer_copy_path_len2 (buffer * const restrict b, const char * const restrict s1, size_t len1, const char * const restrict s2, size_t len2) 257 { 258 /*(similar to buffer_copy_string_len(b, s1, len1) but combined allocation)*/ 259 memcpy(buffer_string_prepare_copy(b, len1+len2+1), s1, len1); 260 b->used = len1 + 1; /*('\0' byte will be written below)*/ 261 262 buffer_append_path_len(b, s2, len2);/*(choice: not inlined, special-cased)*/ 263 } 264 265 void 266 buffer_copy_string_len_lc (buffer * const restrict b, const char * const restrict s, const size_t len) 267 { 268 char * const restrict d = buffer_string_prepare_copy(b, len); 269 b->used = len+1; 270 d[len] = '\0'; 271 for (size_t i = 0; i < len; ++i) 272 d[i] = (!light_isupper(s[i])) ? s[i] : s[i] | 0x20; 273 } 274 275 void buffer_append_uint_hex_lc(buffer *b, uintmax_t value) { 276 char *buf; 277 unsigned int shift = 0; 278 279 { 280 uintmax_t copy = value; 281 do { 282 copy >>= 8; 283 shift += 8; /* counting bits */ 284 } while (0 != copy); 285 } 286 287 buf = buffer_extend(b, shift >> 2); /*nibbles (4 bits)*/ 288 289 while (shift > 0) { 290 shift -= 4; 291 *(buf++) = hex_chars_lc[(value >> shift) & 0x0F]; 292 } 293 } 294 295 __attribute_nonnull__() 296 __attribute_returns_nonnull__ 297 static char* utostr(char buf[LI_ITOSTRING_LENGTH], uintmax_t val) { 298 char *cur = buf+LI_ITOSTRING_LENGTH; 299 uintmax_t x; 300 do { 301 *(--cur) = (char) ('0' + (int)(val - (x = val/10) * 10)); 302 } while (0 != (val = x)); /* val % 10 */ 303 return cur; 304 } 305 306 __attribute_nonnull__() 307 __attribute_returns_nonnull__ 308 static char* itostr(char buf[LI_ITOSTRING_LENGTH], intmax_t val) { 309 /* absolute value not defined for INTMAX_MIN, but can take absolute 310 * value of any negative number via twos complement cast to unsigned. 311 * negative sign is prepended after (now unsigned) value is converted 312 * to string */ 313 uintmax_t uval = val >= 0 ? (uintmax_t)val : ((uintmax_t)~val) + 1; 314 char *cur = utostr(buf, uval); 315 if (val < 0) *(--cur) = '-'; 316 317 return cur; 318 } 319 320 void buffer_append_int(buffer *b, intmax_t val) { 321 char buf[LI_ITOSTRING_LENGTH]; 322 const char * const str = itostr(buf, val); 323 buffer_append_string_len(b, str, buf+sizeof(buf) - str); 324 } 325 326 void buffer_append_strftime(buffer * const restrict b, const char * const restrict format, const struct tm * const restrict tm) { 327 /*(localtime_r() or gmtime_r() producing tm should not have failed)*/ 328 if (__builtin_expect( (NULL == tm), 0)) return; 329 330 /*(expecting typical format strings to result in < 64 bytes needed; 331 * skipping buffer_string_space() calculation and providing fixed size)*/ 332 size_t rv = strftime(buffer_string_prepare_append(b, 63), 64, format, tm); 333 334 /* 0 (in some apis) signals the string may have been too small; 335 * but the format could also just have lead to an empty string */ 336 if (__builtin_expect( (0 == rv), 0) || __builtin_expect( (rv > 63), 0)) { 337 /* unexpected; give it a second try with a larger string */ 338 rv = strftime(buffer_string_prepare_append(b, 4095), 4096, format, tm); 339 if (__builtin_expect( (rv > 4095), 0))/*(input format was ridiculous)*/ 340 return; 341 } 342 343 /*buffer_commit(b, rv);*/ 344 b->used += (uint32_t)rv + (0 == b->used); 345 } 346 347 348 size_t li_itostrn(char *buf, size_t buf_len, intmax_t val) { 349 char p_buf[LI_ITOSTRING_LENGTH]; 350 char* const str = itostr(p_buf, val); 351 size_t len = (size_t)(p_buf+sizeof(p_buf)-str); 352 force_assert(len <= buf_len); 353 memcpy(buf, str, len); 354 return len; 355 } 356 357 size_t li_utostrn(char *buf, size_t buf_len, uintmax_t val) { 358 char p_buf[LI_ITOSTRING_LENGTH]; 359 char* const str = utostr(p_buf, val); 360 size_t len = (size_t)(p_buf+sizeof(p_buf)-str); 361 force_assert(len <= buf_len); 362 memcpy(buf, str, len); 363 return len; 364 } 365 366 #define li_ntox_lc(n) ((n) <= 9 ? (n) + '0' : (n) + 'a' - 10) 367 368 /* c (char) and n (nibble) MUST be unsigned integer types */ 369 #define li_cton(c,n) \ 370 (((n) = (c) - '0') <= 9 || (((n) = ((c)&0xdf) - 'A') <= 5 ? ((n) += 10) : 0)) 371 372 /* converts hex char (0-9, A-Z, a-z) to decimal. 373 * returns 0xFF on invalid input. 374 */ 375 char hex2int(unsigned char hex) { 376 unsigned char n; 377 return li_cton(hex,n) ? (char)n : 0xFF; 378 } 379 380 int li_hex2bin (unsigned char * const bin, const size_t binlen, const char * const hexstr, const size_t len) 381 { 382 /* validate and transform 32-byte MD5 hex string to 16-byte binary MD5, 383 * or 64-byte SHA-256 or SHA-512-256 hex string to 32-byte binary digest */ 384 if (len > (binlen << 1)) return -1; 385 for (int i = 0, ilen = (int)len; i < ilen; i+=2) { 386 int hi = hexstr[i]; 387 int lo = hexstr[i+1]; 388 if ('0' <= hi && hi <= '9') hi -= '0'; 389 else if ((uint32_t)(hi |= 0x20)-'a' <= 'f'-'a')hi += -'a' + 10; 390 else return -1; 391 if ('0' <= lo && lo <= '9') lo -= '0'; 392 else if ((uint32_t)(lo |= 0x20)-'a' <= 'f'-'a')lo += -'a' + 10; 393 else return -1; 394 bin[(i >> 1)] = (unsigned char)((hi << 4) | lo); 395 } 396 return 0; 397 } 398 399 400 __attribute_noinline__ 401 int buffer_eq_icase_ssn(const char * const a, const char * const b, const size_t len) { 402 for (size_t i = 0; i < len; ++i) { 403 unsigned int ca = ((unsigned char *)a)[i]; 404 unsigned int cb = ((unsigned char *)b)[i]; 405 if (ca != cb && ((ca ^ cb) != 0x20 || !light_isalpha(ca))) return 0; 406 } 407 return 1; 408 } 409 410 int buffer_eq_icase_ss(const char * const a, const size_t alen, const char * const b, const size_t blen) { 411 /* 1 = equal; 0 = not equal */ /* short string sizes expected (< INT_MAX) */ 412 return (alen == blen) ? buffer_eq_icase_ssn(a, b, blen) : 0; 413 } 414 415 int buffer_eq_icase_slen(const buffer * const b, const char * const s, const size_t slen) { 416 /* Note: b must be initialized, i.e. 0 != b->used; uninitialized is not eq*/ 417 /* 1 = equal; 0 = not equal */ /* short string sizes expected (< INT_MAX) */ 418 return (b->used == slen + 1) ? buffer_eq_icase_ssn(b->ptr, s, slen) : 0; 419 } 420 421 int buffer_eq_slen(const buffer * const b, const char * const s, const size_t slen) { 422 /* Note: b must be initialized, i.e. 0 != b->used; uninitialized is not eq*/ 423 /* 1 = equal; 0 = not equal */ /* short string sizes expected (< INT_MAX) */ 424 return (b->used == slen + 1 && 0 == memcmp(b->ptr, s, slen)); 425 } 426 427 428 /** 429 * check if two buffer contain the same data 430 */ 431 432 int buffer_is_equal(const buffer *a, const buffer *b) { 433 /* 1 = equal; 0 = not equal */ 434 return (a->used == b->used && 0 == memcmp(a->ptr, b->ptr, a->used)); 435 } 436 437 438 void li_tohex_lc(char * const restrict buf, size_t buf_len, const char * const restrict s, size_t s_len) { 439 force_assert(2 * s_len > s_len); 440 force_assert(2 * s_len < buf_len); 441 442 for (size_t i = 0; i < s_len; ++i) { 443 buf[2*i] = hex_chars_lc[(s[i] >> 4) & 0x0F]; 444 buf[2*i+1] = hex_chars_lc[s[i] & 0x0F]; 445 } 446 buf[2*s_len] = '\0'; 447 } 448 449 void li_tohex_uc(char * const restrict buf, size_t buf_len, const char * const restrict s, size_t s_len) { 450 force_assert(2 * s_len > s_len); 451 force_assert(2 * s_len < buf_len); 452 453 for (size_t i = 0; i < s_len; ++i) { 454 buf[2*i] = hex_chars_uc[(s[i] >> 4) & 0x0F]; 455 buf[2*i+1] = hex_chars_uc[s[i] & 0x0F]; 456 } 457 buf[2*s_len] = '\0'; 458 } 459 460 461 void buffer_substr_replace (buffer * const restrict b, const size_t offset, 462 const size_t len, const buffer * const restrict replace) 463 { 464 const size_t blen = buffer_clen(b); 465 const size_t rlen = buffer_clen(replace); 466 467 if (rlen > len) { 468 buffer_extend(b, rlen-len); 469 memmove(b->ptr+offset+rlen, b->ptr+offset+len, blen-offset-len); 470 } 471 472 memcpy(b->ptr+offset, replace->ptr, rlen); 473 474 if (rlen < len) { 475 memmove(b->ptr+offset+rlen, b->ptr+offset+len, blen-offset-len); 476 buffer_truncate(b, blen-len+rlen); 477 } 478 } 479 480 481 void buffer_append_string_encoded_hex_lc(buffer * const restrict b, const char * const restrict s, size_t len) { 482 unsigned char * const p = (unsigned char *)buffer_extend(b, len*2); 483 for (size_t i = 0; i < len; ++i) { 484 p[(i<<1)] = hex_chars_lc[(s[i] >> 4) & 0x0F]; 485 p[(i<<1)+1] = hex_chars_lc[(s[i]) & 0x0F]; 486 } 487 } 488 489 void buffer_append_string_encoded_hex_uc(buffer * const restrict b, const char * const restrict s, size_t len) { 490 unsigned char * const p = (unsigned char *)buffer_extend(b, len*2); 491 for (size_t i = 0; i < len; ++i) { 492 p[(i<<1)] = hex_chars_uc[(s[i] >> 4) & 0x0F]; 493 p[(i<<1)+1] = hex_chars_uc[(s[i]) & 0x0F]; 494 } 495 } 496 497 498 /* everything except: ! ( ) * - . 0-9 A-Z _ a-z */ 499 static const char encoded_chars_rel_uri_part[] = { 500 /* 501 0 1 2 3 4 5 6 7 8 9 A B C D E F 502 */ 503 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 00 - 0F control chars */ 504 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 10 - 1F */ 505 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, /* 20 - 2F space " # $ % & ' + , / */ 506 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, /* 30 - 3F : ; < = > ? */ 507 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 40 - 4F @ */ 508 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, /* 50 - 5F [ \ ] ^ */ 509 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 60 - 6F ` */ 510 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, /* 70 - 7F { | } DEL */ 511 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 80 - 8F */ 512 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 90 - 9F */ 513 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* A0 - AF */ 514 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* B0 - BF */ 515 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* C0 - CF */ 516 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* D0 - DF */ 517 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* E0 - EF */ 518 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* F0 - FF */ 519 }; 520 521 /* everything except: ! ( ) * - . / 0-9 A-Z _ a-z */ 522 static const char encoded_chars_rel_uri[] = { 523 /* 524 0 1 2 3 4 5 6 7 8 9 A B C D E F 525 */ 526 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 00 - 0F control chars */ 527 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 10 - 1F */ 528 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, /* 20 - 2F space " # $ % & ' + , */ 529 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, /* 30 - 3F : ; < = > ? */ 530 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 40 - 4F @ */ 531 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, /* 50 - 5F [ \ ] ^ */ 532 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 60 - 6F ` */ 533 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, /* 70 - 7F { | } DEL */ 534 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 80 - 8F */ 535 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 90 - 9F */ 536 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* A0 - AF */ 537 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* B0 - BF */ 538 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* C0 - CF */ 539 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* D0 - DF */ 540 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* E0 - EF */ 541 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* F0 - FF */ 542 }; 543 544 static const char encoded_chars_html[] = { 545 /* 546 0 1 2 3 4 5 6 7 8 9 A B C D E F 547 */ 548 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 00 - 0F control chars */ 549 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 10 - 1F */ 550 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, /* 20 - 2F " & ' */ 551 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, /* 30 - 3F < > */ 552 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 40 - 4F */ 553 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 50 - 5F */ 554 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 60 - 6F ` */ 555 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, /* 70 - 7F DEL */ 556 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 80 - 8F */ 557 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 90 - 9F */ 558 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* A0 - AF */ 559 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* B0 - BF */ 560 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* C0 - CF */ 561 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* D0 - DF */ 562 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* E0 - EF */ 563 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* F0 - FF */ 564 }; 565 566 static const char encoded_chars_minimal_xml[] = { 567 /* 568 0 1 2 3 4 5 6 7 8 9 A B C D E F 569 */ 570 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 00 - 0F control chars */ 571 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* 10 - 1F */ 572 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, /* 20 - 2F " & ' */ 573 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, /* 30 - 3F < > */ 574 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 40 - 4F */ 575 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 50 - 5F */ 576 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 60 - 6F ` */ 577 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, /* 70 - 7F DEL */ 578 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 80 - 8F */ 579 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 90 - 9F */ 580 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* A0 - AF */ 581 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* B0 - BF */ 582 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* C0 - CF */ 583 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* D0 - DF */ 584 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* E0 - EF */ 585 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* F0 - FF */ 586 }; 587 588 589 590 void buffer_append_string_encoded(buffer * const restrict b, const char * const restrict s, size_t s_len, buffer_encoding_t encoding) { 591 unsigned char *ds, *d; 592 size_t d_len, ndx; 593 const char *map = NULL; 594 595 switch(encoding) { 596 case ENCODING_REL_URI: 597 map = encoded_chars_rel_uri; 598 break; 599 case ENCODING_REL_URI_PART: 600 map = encoded_chars_rel_uri_part; 601 break; 602 case ENCODING_HTML: 603 map = encoded_chars_html; 604 break; 605 case ENCODING_MINIMAL_XML: 606 map = encoded_chars_minimal_xml; 607 break; 608 } 609 610 /* count to-be-encoded-characters */ 611 for (ds = (unsigned char *)s, d_len = 0, ndx = 0; ndx < s_len; ds++, ndx++) { 612 if (map[*ds & 0xFF]) { 613 switch(encoding) { 614 case ENCODING_REL_URI: 615 case ENCODING_REL_URI_PART: 616 d_len += 3; 617 break; 618 case ENCODING_HTML: 619 case ENCODING_MINIMAL_XML: 620 d_len += 6; 621 break; 622 } 623 } else { 624 d_len++; 625 } 626 } 627 628 d = (unsigned char*) buffer_extend(b, d_len); 629 630 if (d_len == s_len) { /*(short-circuit; nothing to encoded)*/ 631 memcpy(d, s, s_len); 632 return; 633 } 634 635 for (ds = (unsigned char *)s, d_len = 0, ndx = 0; ndx < s_len; ds++, ndx++) { 636 if (map[*ds & 0xFF]) { 637 switch(encoding) { 638 case ENCODING_REL_URI: 639 case ENCODING_REL_URI_PART: 640 d[d_len++] = '%'; 641 d[d_len++] = hex_chars_uc[((*ds) >> 4) & 0x0F]; 642 d[d_len++] = hex_chars_uc[(*ds) & 0x0F]; 643 break; 644 case ENCODING_HTML: 645 case ENCODING_MINIMAL_XML: 646 d[d_len++] = '&'; 647 d[d_len++] = '#'; 648 d[d_len++] = 'x'; 649 d[d_len++] = hex_chars_uc[((*ds) >> 4) & 0x0F]; 650 d[d_len++] = hex_chars_uc[(*ds) & 0x0F]; 651 d[d_len++] = ';'; 652 break; 653 } 654 } else { 655 d[d_len++] = *ds; 656 } 657 } 658 } 659 660 void buffer_append_string_c_escaped(buffer * const restrict b, const char * const restrict s, size_t s_len) { 661 unsigned char *ds, *d; 662 size_t d_len, ndx; 663 664 /* count to-be-encoded-characters */ 665 for (ds = (unsigned char *)s, d_len = 0, ndx = 0; ndx < s_len; ds++, ndx++) { 666 if (__builtin_expect( (*ds >= ' ' && *ds <= '~'), 1)) 667 d_len++; 668 else { /* CTLs or non-ASCII characters */ 669 switch (*ds) { 670 case '\t': 671 case '\r': 672 case '\n': 673 d_len += 2; 674 break; 675 default: 676 d_len += 4; /* \xCC */ 677 break; 678 } 679 } 680 } 681 682 d = (unsigned char*) buffer_extend(b, d_len); 683 684 if (d_len == s_len) { /*(short-circuit; nothing to encoded)*/ 685 memcpy(d, s, s_len); 686 return; 687 } 688 689 for (ds = (unsigned char *)s, d_len = 0, ndx = 0; ndx < s_len; ds++, ndx++) { 690 if (__builtin_expect( (*ds >= ' ' && *ds <= '~'), 1)) 691 d[d_len++] = *ds; 692 else { /* CTLs or non-ASCII characters */ 693 d[d_len++] = '\\'; 694 switch (*ds) { 695 case '\t': 696 d[d_len++] = 't'; 697 break; 698 case '\r': 699 d[d_len++] = 'r'; 700 break; 701 case '\n': 702 d[d_len++] = 'n'; 703 break; 704 default: 705 d[d_len++] = 'x'; 706 d[d_len++] = hex_chars_lc[(*ds) >> 4]; 707 d[d_len++] = hex_chars_lc[(*ds) & 0x0F]; 708 break; 709 } 710 } 711 } 712 } 713 714 715 void 716 buffer_append_bs_escaped (buffer * const restrict b, 717 const char * restrict s, const size_t len) 718 { 719 /* replaces non-printable chars with escaped string 720 * default: \xHH where HH is the hex representation of the byte 721 * exceptions: " => \", \ => \\, whitespace chars => \n \t etc. */ 722 /* Intended for use escaping string to be surrounded by double-quotes */ 723 /* Performs single pass over string and is optimized for ASCII; 724 * non-ASCII escaping might be slightly sped up by walking input twice, 725 * first to calculate escaped length and extend the destination b, and 726 * second to do the escaping. (This non-ASCII optim is not done here) */ 727 buffer_string_prepare_append(b, len); 728 for (const char * const end = s+len; s < end; ++s) { 729 unsigned int c; 730 const char * const ptr = s; 731 do { 732 c = *(const unsigned char *)s; 733 } while (c >= ' ' && c <= '~' && c != '"' && c != '\\' && ++s < end); 734 if (s - ptr) buffer_append_string_len(b, ptr, s - ptr); 735 736 if (s == end) 737 return; 738 739 /* ('\a', '\v' shortcuts are technically not json-escaping) */ 740 /* ('\0' is also omitted due to the possibility of string corruption if 741 * the receiver supports decoding octal escapes (\000) and the escaped 742 * string contains \0 followed by two digits not part of escaping)*/ 743 744 char *d; 745 switch (c) { 746 case '\a':case '\b':case '\t':case '\n':case '\v':case '\f':case '\r': 747 c = "0000000abtnvfr"[c]; 748 __attribute_fallthrough__ 749 case '"': case '\\': 750 d = buffer_extend(b, 2); 751 d[0] = '\\'; 752 d[1] = c; 753 break; 754 default: 755 /* non printable char => \xHH */ 756 d = buffer_extend(b, 4); 757 d[0] = '\\'; 758 d[1] = 'x'; 759 d[2] = hex_chars_uc[c >> 4]; 760 d[3] = hex_chars_uc[c & 0xF]; 761 break; 762 } 763 } 764 } 765 766 767 void 768 buffer_append_bs_escaped_json (buffer * const restrict b, 769 const char * restrict s, const size_t len) 770 { 771 /* replaces non-printable chars with escaped string 772 * json: \u00HH where HH is the hex representation of the byte 773 * exceptions: " => \", \ => \\, whitespace chars => \n \t etc. */ 774 /* Intended for use escaping string to be surrounded by double-quotes */ 775 buffer_string_prepare_append(b, len); 776 for (const char * const end = s+len; s < end; ++s) { 777 unsigned int c; 778 const char * const ptr = s; 779 do { 780 c = *(const unsigned char *)s; 781 } while (c >= ' ' && c != '"' && c != '\\' && ++s < end); 782 if (s - ptr) buffer_append_string_len(b, ptr, s - ptr); 783 784 if (s == end) 785 return; 786 787 /* ('\a', '\v' shortcuts are technically not json-escaping) */ 788 /* ('\0' is also omitted due to the possibility of string corruption if 789 * the receiver supports decoding octal escapes (\000) and the escaped 790 * string contains \0 followed by two digits not part of escaping)*/ 791 792 char *d; 793 switch (c) { 794 case '\a':case '\b':case '\t':case '\n':case '\v':case '\f':case '\r': 795 c = "0000000abtnvfr"[c]; 796 __attribute_fallthrough__ 797 case '"': case '\\': 798 d = buffer_extend(b, 2); 799 d[0] = '\\'; 800 d[1] = c; 801 break; 802 default: 803 d = buffer_extend(b, 6); 804 d[0] = '\\'; 805 d[1] = 'u'; 806 d[2] = '0'; 807 d[3] = '0'; 808 d[4] = hex_chars_uc[c >> 4]; 809 d[5] = hex_chars_uc[c & 0xF]; 810 break; 811 } 812 } 813 } 814 815 816 /* decodes url-special-chars inplace. 817 * replaces non-printable characters with '_' 818 * (If this is used on a portion of query string, then query string should be 819 * split on '&', and '+' replaced with ' ' before calling this routine) 820 */ 821 822 void buffer_urldecode_path(buffer * const b) { 823 const size_t len = buffer_clen(b); 824 char *src = len ? memchr(b->ptr, '%', len) : NULL; 825 if (NULL == src) return; 826 827 char *dst = src; 828 do { 829 /* *src == '%' */ 830 unsigned char high = ((unsigned char *)src)[1]; 831 unsigned char low = high ? hex2int(((unsigned char *)src)[2]) : 0xFF; 832 if (0xFF != (high = hex2int(high)) && 0xFF != low) { 833 high = (high << 4) | low; /* map ctrls to '_' */ 834 *dst = (high >= 32 && high != 127) ? high : '_'; 835 src += 2; 836 } /* else ignore this '%'; leave as-is and move on */ 837 838 while ((*++dst = *++src) != '%' && *src) ; 839 } while (*src); 840 b->used = (dst - b->ptr) + 1; 841 } 842 843 int buffer_is_valid_UTF8(const buffer *b) { 844 /* https://www.w3.org/International/questions/qa-forms-utf-8 */ 845 /*assert(b->used);*//*(b->ptr must exist and be '\0'-terminated)*/ 846 const unsigned char *c = (unsigned char *)b->ptr; 847 while (*c) { 848 849 /*(note: includes ctrls)*/ 850 if ( c[0] < 0x80 ) { ++c; continue; } 851 852 if ( 0xc2 <= c[0] && c[0] <= 0xdf 853 && 0x80 <= c[1] && c[1] <= 0xbf ) { c+=2; continue; } 854 855 if ( ( ( 0xe0 == c[0] 856 && 0xa0 <= c[1] && c[1] <= 0xbf) 857 || ( 0xe1 <= c[0] && c[0] <= 0xef && c[0] != 0xed 858 && 0x80 <= c[1] && c[1] <= 0xbf) 859 || ( 0xed == c[0] 860 && 0x80 <= c[1] && c[1] <= 0x9f) ) 861 && 0x80 <= c[2] && c[2] <= 0xbf ) { c+=3; continue; } 862 863 if ( ( ( 0xf0 == c[0] 864 && 0x90 <= c[1] && c[1] <= 0xbf) 865 || ( 0xf1 <= c[0] && c[0] <= 0xf3 866 && 0x80 <= c[1] && c[1] <= 0xbf) 867 || ( 0xf4 == c[0] 868 && 0x80 <= c[1] && c[1] <= 0x8f) ) 869 && 0x80 <= c[2] && c[2] <= 0xbf 870 && 0x80 <= c[3] && c[3] <= 0xbf ) { c+=4; continue; } 871 872 return 0; /* invalid */ 873 } 874 return 1; /* valid */ 875 } 876 877 /* - special case: empty string returns empty string 878 * - on windows or cygwin: replace \ with / 879 * - strip leading spaces 880 * - prepends "/" if not present already 881 * - resolve "/../", "//" and "/./" the usual way: 882 * the first one removes a preceding component, the other two 883 * get compressed to "/". 884 * - "/." and "/.." at the end are similar, but always leave a trailing 885 * "/" 886 * 887 * /blah/.. gets / 888 * /blah/../foo gets /foo 889 * /abc/./xyz gets /abc/xyz 890 * /abc//xyz gets /abc/xyz 891 */ 892 893 void buffer_path_simplify(buffer *b) 894 { 895 char *out = b->ptr; 896 char * const end = b->ptr + b->used - 1; 897 898 if (__builtin_expect( (buffer_is_blank(b)), 0)) { 899 buffer_blank(b); 900 return; 901 } 902 903 #if defined(__WIN32) || defined(__CYGWIN__) 904 /* cygwin is treating \ and / the same, so we have to that too */ 905 for (char *p = b->ptr; *p; p++) { 906 if (*p == '\\') *p = '/'; 907 } 908 #endif 909 910 *end = '/'; /*(end of path modified to avoid need to check '\0')*/ 911 912 char *walk = out; 913 if (__builtin_expect( (*walk == '/'), 1)) { 914 /* scan to detect (potential) need for path simplification 915 * (repeated '/' or "/.") */ 916 do { 917 if (*++walk == '.' || *walk == '/') 918 break; 919 do { ++walk; } while (*walk != '/'); 920 } while (walk != end); 921 if (__builtin_expect( (walk == end), 1)) { 922 /* common case: no repeated '/' or "/." */ 923 *end = '\0'; /* overwrite extra '/' added to end of path */ 924 return; 925 } 926 out = walk-1; 927 } 928 else { 929 if (walk[0] == '.' && walk[1] == '/') 930 *out = *++walk; 931 else if (walk[0] == '.' && walk[1] == '.' && walk[2] == '/') 932 *out = *(walk += 2); 933 else { 934 while (*++walk != '/') ; 935 out = walk; 936 } 937 ++walk; 938 } 939 940 while (walk <= end) { 941 /* previous char is '/' at this point (or start of string w/o '/') */ 942 if (__builtin_expect( (walk[0] == '/'), 0)) { 943 /* skip repeated '/' (e.g. "///" -> "/") */ 944 if (++walk < end) 945 continue; 946 else { 947 ++out; 948 break; 949 } 950 } 951 else if (__builtin_expect( (walk[0] == '.'), 0)) { 952 /* handle "./" and "../" */ 953 if (walk[1] == '.' && walk[2] == '/') { 954 /* handle "../" */ 955 while (out > b->ptr && *--out != '/') ; 956 *out = '/'; /*(in case path had not started with '/')*/ 957 if ((walk += 3) >= end) { 958 ++out; 959 break; 960 } 961 else 962 continue; 963 } 964 else if (walk[1] == '/') { 965 /* handle "./" */ 966 if ((walk += 2) >= end) { 967 ++out; 968 break; 969 } 970 continue; 971 } 972 else { 973 /* accept "." if not part of "../" or "./" */ 974 *++out = '.'; 975 ++walk; 976 } 977 } 978 979 while ((*++out = *walk++) != '/') ; 980 } 981 *out = *end = '\0'; /* overwrite extra '/' added to end of path */ 982 b->used = (out - b->ptr) + 1; 983 /*buffer_truncate(b, out - b->ptr);*/ 984 } 985 986 void buffer_to_lower(buffer * const b) { 987 unsigned char * const restrict s = (unsigned char *)b->ptr; 988 const uint_fast32_t used = b->used; 989 for (uint_fast32_t i = 0; i < used; ++i) { 990 if (light_isupper(s[i])) s[i] |= 0x20; 991 } 992 } 993 994 995 void buffer_to_upper(buffer * const b) { 996 unsigned char * const restrict s = (unsigned char *)b->ptr; 997 const uint_fast32_t used = b->used; 998 for (uint_fast32_t i = 0; i < used; ++i) { 999 if (light_islower(s[i])) s[i] &= 0xdf; 1000 } 1001 } 1002