| 1 | /************************************************* |
| 2 | * Exim - an Internet mail transport agent * |
| 3 | *************************************************/ |
| 4 | |
| 5 | /* Copyright (c) University of Cambridge 1995 - 2018 */ |
| 6 | /* See the file NOTICE for conditions of use and distribution. */ |
| 7 | |
| 8 | /* Miscellaneous string-handling functions. Some are not required for |
| 9 | utilities and tests, and are cut out by the COMPILE_UTILITY macro. */ |
| 10 | |
| 11 | |
| 12 | #include "exim.h" |
| 13 | #include <assert.h> |
| 14 | |
| 15 | static void gstring_rebuffer(gstring * g); |
| 16 | |
| 17 | #ifndef COMPILE_UTILITY |
| 18 | /************************************************* |
| 19 | * Test for IP address * |
| 20 | *************************************************/ |
| 21 | |
| 22 | /* This used just to be a regular expression, but with IPv6 things are a bit |
| 23 | more complicated. If the address contains a colon, it is assumed to be a v6 |
| 24 | address (assuming HAVE_IPV6 is set). If a mask is permitted and one is present, |
| 25 | and maskptr is not NULL, its offset is placed there. |
| 26 | |
| 27 | Arguments: |
| 28 | s a string |
| 29 | maskptr NULL if no mask is permitted to follow |
| 30 | otherwise, points to an int where the offset of '/' is placed |
| 31 | if there is no / followed by trailing digits, *maskptr is set 0 |
| 32 | |
| 33 | Returns: 0 if the string is not a textual representation of an IP address |
| 34 | 4 if it is an IPv4 address |
| 35 | 6 if it is an IPv6 address |
| 36 | */ |
| 37 | |
| 38 | int |
| 39 | string_is_ip_address(const uschar *s, int *maskptr) |
| 40 | { |
| 41 | int yield = 4; |
| 42 | |
| 43 | /* If an optional mask is permitted, check for it. If found, pass back the |
| 44 | offset. */ |
| 45 | |
| 46 | if (maskptr) |
| 47 | { |
| 48 | const uschar *ss = s + Ustrlen(s); |
| 49 | *maskptr = 0; |
| 50 | if (s != ss && isdigit(*(--ss))) |
| 51 | { |
| 52 | while (ss > s && isdigit(ss[-1])) ss--; |
| 53 | if (ss > s && *(--ss) == '/') *maskptr = ss - s; |
| 54 | } |
| 55 | } |
| 56 | |
| 57 | /* A colon anywhere in the string => IPv6 address */ |
| 58 | |
| 59 | if (Ustrchr(s, ':') != NULL) |
| 60 | { |
| 61 | BOOL had_double_colon = FALSE; |
| 62 | BOOL v4end = FALSE; |
| 63 | |
| 64 | yield = 6; |
| 65 | |
| 66 | /* An IPv6 address must start with hex digit or double colon. A single |
| 67 | colon is invalid. */ |
| 68 | |
| 69 | if (*s == ':' && *(++s) != ':') return 0; |
| 70 | |
| 71 | /* Now read up to 8 components consisting of up to 4 hex digits each. There |
| 72 | may be one and only one appearance of double colon, which implies any number |
| 73 | of binary zero bits. The number of preceding components is held in count. */ |
| 74 | |
| 75 | for (int count = 0; count < 8; count++) |
| 76 | { |
| 77 | /* If the end of the string is reached before reading 8 components, the |
| 78 | address is valid provided a double colon has been read. This also applies |
| 79 | if we hit the / that introduces a mask or the % that introduces the |
| 80 | interface specifier (scope id) of a link-local address. */ |
| 81 | |
| 82 | if (*s == 0 || *s == '%' || *s == '/') return had_double_colon ? yield : 0; |
| 83 | |
| 84 | /* If a component starts with an additional colon, we have hit a double |
| 85 | colon. This is permitted to appear once only, and counts as at least |
| 86 | one component. The final component may be of this form. */ |
| 87 | |
| 88 | if (*s == ':') |
| 89 | { |
| 90 | if (had_double_colon) return 0; |
| 91 | had_double_colon = TRUE; |
| 92 | s++; |
| 93 | continue; |
| 94 | } |
| 95 | |
| 96 | /* If the remainder of the string contains a dot but no colons, we |
| 97 | can expect a trailing IPv4 address. This is valid if either there has |
| 98 | been no double-colon and this is the 7th component (with the IPv4 address |
| 99 | being the 7th & 8th components), OR if there has been a double-colon |
| 100 | and fewer than 6 components. */ |
| 101 | |
| 102 | if (Ustrchr(s, ':') == NULL && Ustrchr(s, '.') != NULL) |
| 103 | { |
| 104 | if ((!had_double_colon && count != 6) || |
| 105 | (had_double_colon && count > 6)) return 0; |
| 106 | v4end = TRUE; |
| 107 | yield = 6; |
| 108 | break; |
| 109 | } |
| 110 | |
| 111 | /* Check for at least one and not more than 4 hex digits for this |
| 112 | component. */ |
| 113 | |
| 114 | if (!isxdigit(*s++)) return 0; |
| 115 | if (isxdigit(*s) && isxdigit(*(++s)) && isxdigit(*(++s))) s++; |
| 116 | |
| 117 | /* If the component is terminated by colon and there is more to |
| 118 | follow, skip over the colon. If there is no more to follow the address is |
| 119 | invalid. */ |
| 120 | |
| 121 | if (*s == ':' && *(++s) == 0) return 0; |
| 122 | } |
| 123 | |
| 124 | /* If about to handle a trailing IPv4 address, drop through. Otherwise |
| 125 | all is well if we are at the end of the string or at the mask or at a percent |
| 126 | sign, which introduces the interface specifier (scope id) of a link local |
| 127 | address. */ |
| 128 | |
| 129 | if (!v4end) |
| 130 | return (*s == 0 || *s == '%' || |
| 131 | (*s == '/' && maskptr != NULL && *maskptr != 0))? yield : 0; |
| 132 | } |
| 133 | |
| 134 | /* Test for IPv4 address, which may be the tail-end of an IPv6 address. */ |
| 135 | |
| 136 | for (int i = 0; i < 4; i++) |
| 137 | { |
| 138 | long n; |
| 139 | uschar * end; |
| 140 | |
| 141 | if (i != 0 && *s++ != '.') return 0; |
| 142 | n = strtol(CCS s, CSS &end, 10); |
| 143 | if (n > 255 || n < 0 || end <= s || end > s+3) return 0; |
| 144 | s = end; |
| 145 | } |
| 146 | |
| 147 | return !*s || (*s == '/' && maskptr && *maskptr != 0) ? yield : 0; |
| 148 | } |
| 149 | #endif /* COMPILE_UTILITY */ |
| 150 | |
| 151 | |
| 152 | /************************************************* |
| 153 | * Format message size * |
| 154 | *************************************************/ |
| 155 | |
| 156 | /* Convert a message size in bytes to printing form, rounding |
| 157 | according to the magnitude of the number. A value of zero causes |
| 158 | a string of spaces to be returned. |
| 159 | |
| 160 | Arguments: |
| 161 | size the message size in bytes |
| 162 | buffer where to put the answer |
| 163 | |
| 164 | Returns: pointer to the buffer |
| 165 | a string of exactly 5 characters is normally returned |
| 166 | */ |
| 167 | |
| 168 | uschar * |
| 169 | string_format_size(int size, uschar *buffer) |
| 170 | { |
| 171 | if (size == 0) Ustrcpy(buffer, US" "); |
| 172 | else if (size < 1024) sprintf(CS buffer, "%5d", size); |
| 173 | else if (size < 10*1024) |
| 174 | sprintf(CS buffer, "%4.1fK", (double)size / 1024.0); |
| 175 | else if (size < 1024*1024) |
| 176 | sprintf(CS buffer, "%4dK", (size + 512)/1024); |
| 177 | else if (size < 10*1024*1024) |
| 178 | sprintf(CS buffer, "%4.1fM", (double)size / (1024.0 * 1024.0)); |
| 179 | else |
| 180 | sprintf(CS buffer, "%4dM", (size + 512 * 1024)/(1024*1024)); |
| 181 | return buffer; |
| 182 | } |
| 183 | |
| 184 | |
| 185 | |
| 186 | #ifndef COMPILE_UTILITY |
| 187 | /************************************************* |
| 188 | * Convert a number to base 62 format * |
| 189 | *************************************************/ |
| 190 | |
| 191 | /* Convert a long integer into an ASCII base 62 string. For Cygwin the value of |
| 192 | BASE_62 is actually 36. Always return exactly 6 characters plus zero, in a |
| 193 | static area. |
| 194 | |
| 195 | Argument: a long integer |
| 196 | Returns: pointer to base 62 string |
| 197 | */ |
| 198 | |
| 199 | uschar * |
| 200 | string_base62(unsigned long int value) |
| 201 | { |
| 202 | static uschar yield[7]; |
| 203 | uschar *p = yield + sizeof(yield) - 1; |
| 204 | *p = 0; |
| 205 | while (p > yield) |
| 206 | { |
| 207 | *(--p) = base62_chars[value % BASE_62]; |
| 208 | value /= BASE_62; |
| 209 | } |
| 210 | return yield; |
| 211 | } |
| 212 | #endif /* COMPILE_UTILITY */ |
| 213 | |
| 214 | |
| 215 | |
| 216 | /************************************************* |
| 217 | * Interpret escape sequence * |
| 218 | *************************************************/ |
| 219 | |
| 220 | /* This function is called from several places where escape sequences are to be |
| 221 | interpreted in strings. |
| 222 | |
| 223 | Arguments: |
| 224 | pp points a pointer to the initiating "\" in the string; |
| 225 | the pointer gets updated to point to the final character |
| 226 | If the backslash is the last character in the string, it |
| 227 | is not interpreted. |
| 228 | Returns: the value of the character escape |
| 229 | */ |
| 230 | |
| 231 | int |
| 232 | string_interpret_escape(const uschar **pp) |
| 233 | { |
| 234 | #ifdef COMPILE_UTILITY |
| 235 | const uschar *hex_digits= CUS"0123456789abcdef"; |
| 236 | #endif |
| 237 | int ch; |
| 238 | const uschar *p = *pp; |
| 239 | ch = *(++p); |
| 240 | if (ch == '\0') return **pp; |
| 241 | if (isdigit(ch) && ch != '8' && ch != '9') |
| 242 | { |
| 243 | ch -= '0'; |
| 244 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') |
| 245 | { |
| 246 | ch = ch * 8 + *(++p) - '0'; |
| 247 | if (isdigit(p[1]) && p[1] != '8' && p[1] != '9') |
| 248 | ch = ch * 8 + *(++p) - '0'; |
| 249 | } |
| 250 | } |
| 251 | else switch(ch) |
| 252 | { |
| 253 | case 'b': ch = '\b'; break; |
| 254 | case 'f': ch = '\f'; break; |
| 255 | case 'n': ch = '\n'; break; |
| 256 | case 'r': ch = '\r'; break; |
| 257 | case 't': ch = '\t'; break; |
| 258 | case 'v': ch = '\v'; break; |
| 259 | case 'x': |
| 260 | ch = 0; |
| 261 | if (isxdigit(p[1])) |
| 262 | { |
| 263 | ch = ch * 16 + |
| 264 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; |
| 265 | if (isxdigit(p[1])) ch = ch * 16 + |
| 266 | Ustrchr(hex_digits, tolower(*(++p))) - hex_digits; |
| 267 | } |
| 268 | break; |
| 269 | } |
| 270 | *pp = p; |
| 271 | return ch; |
| 272 | } |
| 273 | |
| 274 | |
| 275 | |
| 276 | #ifndef COMPILE_UTILITY |
| 277 | /************************************************* |
| 278 | * Ensure string is printable * |
| 279 | *************************************************/ |
| 280 | |
| 281 | /* This function is called for critical strings. It checks for any |
| 282 | non-printing characters, and if any are found, it makes a new copy |
| 283 | of the string with suitable escape sequences. It is most often called by the |
| 284 | macro string_printing(), which sets allow_tab TRUE. |
| 285 | |
| 286 | Arguments: |
| 287 | s the input string |
| 288 | allow_tab TRUE to allow tab as a printing character |
| 289 | |
| 290 | Returns: string with non-printers encoded as printing sequences |
| 291 | */ |
| 292 | |
| 293 | const uschar * |
| 294 | string_printing2(const uschar *s, BOOL allow_tab) |
| 295 | { |
| 296 | int nonprintcount = 0; |
| 297 | int length = 0; |
| 298 | const uschar *t = s; |
| 299 | uschar *ss, *tt; |
| 300 | |
| 301 | while (*t != 0) |
| 302 | { |
| 303 | int c = *t++; |
| 304 | if (!mac_isprint(c) || (!allow_tab && c == '\t')) nonprintcount++; |
| 305 | length++; |
| 306 | } |
| 307 | |
| 308 | if (nonprintcount == 0) return s; |
| 309 | |
| 310 | /* Get a new block of store guaranteed big enough to hold the |
| 311 | expanded string. */ |
| 312 | |
| 313 | ss = store_get(length + nonprintcount * 3 + 1, is_tainted(s)); |
| 314 | |
| 315 | /* Copy everything, escaping non printers. */ |
| 316 | |
| 317 | t = s; |
| 318 | tt = ss; |
| 319 | |
| 320 | while (*t != 0) |
| 321 | { |
| 322 | int c = *t; |
| 323 | if (mac_isprint(c) && (allow_tab || c != '\t')) *tt++ = *t++; else |
| 324 | { |
| 325 | *tt++ = '\\'; |
| 326 | switch (*t) |
| 327 | { |
| 328 | case '\n': *tt++ = 'n'; break; |
| 329 | case '\r': *tt++ = 'r'; break; |
| 330 | case '\b': *tt++ = 'b'; break; |
| 331 | case '\v': *tt++ = 'v'; break; |
| 332 | case '\f': *tt++ = 'f'; break; |
| 333 | case '\t': *tt++ = 't'; break; |
| 334 | default: sprintf(CS tt, "%03o", *t); tt += 3; break; |
| 335 | } |
| 336 | t++; |
| 337 | } |
| 338 | } |
| 339 | *tt = 0; |
| 340 | return ss; |
| 341 | } |
| 342 | #endif /* COMPILE_UTILITY */ |
| 343 | |
| 344 | /************************************************* |
| 345 | * Undo printing escapes in string * |
| 346 | *************************************************/ |
| 347 | |
| 348 | /* This function is the reverse of string_printing2. It searches for |
| 349 | backslash characters and if any are found, it makes a new copy of the |
| 350 | string with escape sequences parsed. Otherwise it returns the original |
| 351 | string. |
| 352 | |
| 353 | Arguments: |
| 354 | s the input string |
| 355 | |
| 356 | Returns: string with printing escapes parsed back |
| 357 | */ |
| 358 | |
| 359 | uschar * |
| 360 | string_unprinting(uschar *s) |
| 361 | { |
| 362 | uschar *p, *q, *r, *ss; |
| 363 | int len, off; |
| 364 | |
| 365 | p = Ustrchr(s, '\\'); |
| 366 | if (!p) return s; |
| 367 | |
| 368 | len = Ustrlen(s) + 1; |
| 369 | ss = store_get(len, is_tainted(s)); |
| 370 | |
| 371 | q = ss; |
| 372 | off = p - s; |
| 373 | if (off) |
| 374 | { |
| 375 | memcpy(q, s, off); |
| 376 | q += off; |
| 377 | } |
| 378 | |
| 379 | while (*p) |
| 380 | { |
| 381 | if (*p == '\\') |
| 382 | { |
| 383 | *q++ = string_interpret_escape((const uschar **)&p); |
| 384 | p++; |
| 385 | } |
| 386 | else |
| 387 | { |
| 388 | r = Ustrchr(p, '\\'); |
| 389 | if (!r) |
| 390 | { |
| 391 | off = Ustrlen(p); |
| 392 | memcpy(q, p, off); |
| 393 | p += off; |
| 394 | q += off; |
| 395 | break; |
| 396 | } |
| 397 | else |
| 398 | { |
| 399 | off = r - p; |
| 400 | memcpy(q, p, off); |
| 401 | q += off; |
| 402 | p = r; |
| 403 | } |
| 404 | } |
| 405 | } |
| 406 | *q = '\0'; |
| 407 | |
| 408 | return ss; |
| 409 | } |
| 410 | |
| 411 | |
| 412 | |
| 413 | |
| 414 | #if (defined(HAVE_LOCAL_SCAN) || defined(EXPAND_DLFUNC)) \ |
| 415 | && !defined(MACRO_PREDEF) && !defined(COMPILE_UTILITY) |
| 416 | /************************************************* |
| 417 | * Copy and save string * |
| 418 | *************************************************/ |
| 419 | |
| 420 | /* |
| 421 | Argument: string to copy |
| 422 | Returns: copy of string in new store with the same taint status |
| 423 | */ |
| 424 | |
| 425 | uschar * |
| 426 | string_copy_function(const uschar *s) |
| 427 | { |
| 428 | return string_copy_taint(s, is_tainted(s)); |
| 429 | } |
| 430 | |
| 431 | /* This function assumes that memcpy() is faster than strcpy(). |
| 432 | As above, but explicitly specifying the result taint status |
| 433 | */ |
| 434 | |
| 435 | uschar * |
| 436 | string_copy_taint_function(const uschar * s, BOOL tainted) |
| 437 | { |
| 438 | int len = Ustrlen(s) + 1; |
| 439 | uschar *ss = store_get(len, tainted); |
| 440 | memcpy(ss, s, len); |
| 441 | return ss; |
| 442 | } |
| 443 | |
| 444 | |
| 445 | |
| 446 | /************************************************* |
| 447 | * Copy and save string, given length * |
| 448 | *************************************************/ |
| 449 | |
| 450 | /* It is assumed the data contains no zeros. A zero is added |
| 451 | onto the end. |
| 452 | |
| 453 | Arguments: |
| 454 | s string to copy |
| 455 | n number of characters |
| 456 | |
| 457 | Returns: copy of string in new store |
| 458 | */ |
| 459 | |
| 460 | uschar * |
| 461 | string_copyn_function(const uschar *s, int n) |
| 462 | { |
| 463 | uschar *ss = store_get(n + 1, is_tainted(s)); |
| 464 | Ustrncpy(ss, s, n); |
| 465 | ss[n] = 0; |
| 466 | return ss; |
| 467 | } |
| 468 | #endif |
| 469 | |
| 470 | |
| 471 | /************************************************* |
| 472 | * Copy and save string in malloc'd store * |
| 473 | *************************************************/ |
| 474 | |
| 475 | /* This function assumes that memcpy() is faster than strcpy(). |
| 476 | |
| 477 | Argument: string to copy |
| 478 | Returns: copy of string in new store |
| 479 | */ |
| 480 | |
| 481 | uschar * |
| 482 | string_copy_malloc(const uschar *s) |
| 483 | { |
| 484 | int len = Ustrlen(s) + 1; |
| 485 | uschar *ss = store_malloc(len); |
| 486 | memcpy(ss, s, len); |
| 487 | return ss; |
| 488 | } |
| 489 | |
| 490 | |
| 491 | |
| 492 | /************************************************* |
| 493 | * Copy string if long, inserting newlines * |
| 494 | *************************************************/ |
| 495 | |
| 496 | /* If the given string is longer than 75 characters, it is copied, and within |
| 497 | the copy, certain space characters are converted into newlines. |
| 498 | |
| 499 | Argument: pointer to the string |
| 500 | Returns: pointer to the possibly altered string |
| 501 | */ |
| 502 | |
| 503 | uschar * |
| 504 | string_split_message(uschar *msg) |
| 505 | { |
| 506 | uschar *s, *ss; |
| 507 | |
| 508 | if (msg == NULL || Ustrlen(msg) <= 75) return msg; |
| 509 | s = ss = msg = string_copy(msg); |
| 510 | |
| 511 | for (;;) |
| 512 | { |
| 513 | int i = 0; |
| 514 | while (i < 75 && *ss != 0 && *ss != '\n') ss++, i++; |
| 515 | if (*ss == 0) break; |
| 516 | if (*ss == '\n') |
| 517 | s = ++ss; |
| 518 | else |
| 519 | { |
| 520 | uschar *t = ss + 1; |
| 521 | uschar *tt = NULL; |
| 522 | while (--t > s + 35) |
| 523 | { |
| 524 | if (*t == ' ') |
| 525 | { |
| 526 | if (t[-1] == ':') { tt = t; break; } |
| 527 | if (tt == NULL) tt = t; |
| 528 | } |
| 529 | } |
| 530 | |
| 531 | if (tt == NULL) /* Can't split behind - try ahead */ |
| 532 | { |
| 533 | t = ss + 1; |
| 534 | while (*t != 0) |
| 535 | { |
| 536 | if (*t == ' ' || *t == '\n') |
| 537 | { tt = t; break; } |
| 538 | t++; |
| 539 | } |
| 540 | } |
| 541 | |
| 542 | if (tt == NULL) break; /* Can't find anywhere to split */ |
| 543 | *tt = '\n'; |
| 544 | s = ss = tt+1; |
| 545 | } |
| 546 | } |
| 547 | |
| 548 | return msg; |
| 549 | } |
| 550 | |
| 551 | |
| 552 | |
| 553 | /************************************************* |
| 554 | * Copy returned DNS domain name, de-escaping * |
| 555 | *************************************************/ |
| 556 | |
| 557 | /* If a domain name contains top-bit characters, some resolvers return |
| 558 | the fully qualified name with those characters turned into escapes. The |
| 559 | convention is a backslash followed by _decimal_ digits. We convert these |
| 560 | back into the original binary values. This will be relevant when |
| 561 | allow_utf8_domains is set true and UTF-8 characters are used in domain |
| 562 | names. Backslash can also be used to escape other characters, though we |
| 563 | shouldn't come across them in domain names. |
| 564 | |
| 565 | Argument: the domain name string |
| 566 | Returns: copy of string in new store, de-escaped |
| 567 | */ |
| 568 | |
| 569 | uschar * |
| 570 | string_copy_dnsdomain(uschar *s) |
| 571 | { |
| 572 | uschar *yield; |
| 573 | uschar *ss = yield = store_get(Ustrlen(s) + 1, is_tainted(s)); |
| 574 | |
| 575 | while (*s != 0) |
| 576 | { |
| 577 | if (*s != '\\') |
| 578 | { |
| 579 | *ss++ = *s++; |
| 580 | } |
| 581 | else if (isdigit(s[1])) |
| 582 | { |
| 583 | *ss++ = (s[1] - '0')*100 + (s[2] - '0')*10 + s[3] - '0'; |
| 584 | s += 4; |
| 585 | } |
| 586 | else if (*(++s) != 0) |
| 587 | { |
| 588 | *ss++ = *s++; |
| 589 | } |
| 590 | } |
| 591 | |
| 592 | *ss = 0; |
| 593 | return yield; |
| 594 | } |
| 595 | |
| 596 | |
| 597 | #ifndef COMPILE_UTILITY |
| 598 | /************************************************* |
| 599 | * Copy space-terminated or quoted string * |
| 600 | *************************************************/ |
| 601 | |
| 602 | /* This function copies from a string until its end, or until whitespace is |
| 603 | encountered, unless the string begins with a double quote, in which case the |
| 604 | terminating quote is sought, and escaping within the string is done. The length |
| 605 | of a de-quoted string can be no longer than the original, since escaping always |
| 606 | turns n characters into 1 character. |
| 607 | |
| 608 | Argument: pointer to the pointer to the first character, which gets updated |
| 609 | Returns: the new string |
| 610 | */ |
| 611 | |
| 612 | uschar * |
| 613 | string_dequote(const uschar **sptr) |
| 614 | { |
| 615 | const uschar *s = *sptr; |
| 616 | uschar *t, *yield; |
| 617 | |
| 618 | /* First find the end of the string */ |
| 619 | |
| 620 | if (*s != '\"') |
| 621 | while (*s != 0 && !isspace(*s)) s++; |
| 622 | else |
| 623 | { |
| 624 | s++; |
| 625 | while (*s && *s != '\"') |
| 626 | { |
| 627 | if (*s == '\\') (void)string_interpret_escape(&s); |
| 628 | s++; |
| 629 | } |
| 630 | if (*s) s++; |
| 631 | } |
| 632 | |
| 633 | /* Get enough store to copy into */ |
| 634 | |
| 635 | t = yield = store_get(s - *sptr + 1, is_tainted(*sptr)); |
| 636 | s = *sptr; |
| 637 | |
| 638 | /* Do the copy */ |
| 639 | |
| 640 | if (*s != '\"') |
| 641 | while (*s != 0 && !isspace(*s)) *t++ = *s++; |
| 642 | else |
| 643 | { |
| 644 | s++; |
| 645 | while (*s != 0 && *s != '\"') |
| 646 | { |
| 647 | *t++ = *s == '\\' ? string_interpret_escape(&s) : *s; |
| 648 | s++; |
| 649 | } |
| 650 | if (*s) s++; |
| 651 | } |
| 652 | |
| 653 | /* Update the pointer and return the terminated copy */ |
| 654 | |
| 655 | *sptr = s; |
| 656 | *t = 0; |
| 657 | return yield; |
| 658 | } |
| 659 | #endif /* COMPILE_UTILITY */ |
| 660 | |
| 661 | |
| 662 | |
| 663 | /************************************************* |
| 664 | * Format a string and save it * |
| 665 | *************************************************/ |
| 666 | |
| 667 | /* The formatting is done by string_vformat, which checks the length of |
| 668 | everything. Taint is taken from the worst of the arguments. |
| 669 | |
| 670 | Arguments: |
| 671 | format a printf() format - deliberately char * rather than uschar * |
| 672 | because it will most usually be a literal string |
| 673 | ... arguments for format |
| 674 | |
| 675 | Returns: pointer to fresh piece of store containing sprintf'ed string |
| 676 | */ |
| 677 | |
| 678 | uschar * |
| 679 | string_sprintf_trc(const char *format, const uschar * func, unsigned line, ...) |
| 680 | { |
| 681 | #ifdef COMPILE_UTILITY |
| 682 | uschar buffer[STRING_SPRINTF_BUFFER_SIZE]; |
| 683 | gstring gs = { .size = STRING_SPRINTF_BUFFER_SIZE, .ptr = 0, .s = buffer }; |
| 684 | gstring * g = &gs; |
| 685 | unsigned flags = 0; |
| 686 | #else |
| 687 | gstring * g = NULL; |
| 688 | unsigned flags = SVFMT_REBUFFER|SVFMT_EXTEND; |
| 689 | #endif |
| 690 | |
| 691 | va_list ap; |
| 692 | va_start(ap, line); |
| 693 | g = string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE, |
| 694 | flags, format, ap); |
| 695 | va_end(ap); |
| 696 | |
| 697 | if (!g) |
| 698 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, |
| 699 | "string_sprintf expansion was longer than %d; format string was (%s)\n" |
| 700 | " called from %s %d\n", |
| 701 | STRING_SPRINTF_BUFFER_SIZE, format, func, line); |
| 702 | |
| 703 | #ifdef COMPILE_UTILITY |
| 704 | return string_copyn(g->s, g->ptr); |
| 705 | #else |
| 706 | gstring_release_unused(g); |
| 707 | return string_from_gstring(g); |
| 708 | #endif |
| 709 | } |
| 710 | |
| 711 | |
| 712 | |
| 713 | /************************************************* |
| 714 | * Case-independent strncmp() function * |
| 715 | *************************************************/ |
| 716 | |
| 717 | /* |
| 718 | Arguments: |
| 719 | s first string |
| 720 | t second string |
| 721 | n number of characters to compare |
| 722 | |
| 723 | Returns: < 0, = 0, or > 0, according to the comparison |
| 724 | */ |
| 725 | |
| 726 | int |
| 727 | strncmpic(const uschar *s, const uschar *t, int n) |
| 728 | { |
| 729 | while (n--) |
| 730 | { |
| 731 | int c = tolower(*s++) - tolower(*t++); |
| 732 | if (c) return c; |
| 733 | } |
| 734 | return 0; |
| 735 | } |
| 736 | |
| 737 | |
| 738 | /************************************************* |
| 739 | * Case-independent strcmp() function * |
| 740 | *************************************************/ |
| 741 | |
| 742 | /* |
| 743 | Arguments: |
| 744 | s first string |
| 745 | t second string |
| 746 | |
| 747 | Returns: < 0, = 0, or > 0, according to the comparison |
| 748 | */ |
| 749 | |
| 750 | int |
| 751 | strcmpic(const uschar *s, const uschar *t) |
| 752 | { |
| 753 | while (*s != 0) |
| 754 | { |
| 755 | int c = tolower(*s++) - tolower(*t++); |
| 756 | if (c != 0) return c; |
| 757 | } |
| 758 | return *t; |
| 759 | } |
| 760 | |
| 761 | |
| 762 | /************************************************* |
| 763 | * Case-independent strstr() function * |
| 764 | *************************************************/ |
| 765 | |
| 766 | /* The third argument specifies whether whitespace is required |
| 767 | to follow the matched string. |
| 768 | |
| 769 | Arguments: |
| 770 | s string to search |
| 771 | t substring to search for |
| 772 | space_follows if TRUE, match only if whitespace follows |
| 773 | |
| 774 | Returns: pointer to substring in string, or NULL if not found |
| 775 | */ |
| 776 | |
| 777 | uschar * |
| 778 | strstric(uschar *s, uschar *t, BOOL space_follows) |
| 779 | { |
| 780 | uschar *p = t; |
| 781 | uschar *yield = NULL; |
| 782 | int cl = tolower(*p); |
| 783 | int cu = toupper(*p); |
| 784 | |
| 785 | while (*s) |
| 786 | { |
| 787 | if (*s == cl || *s == cu) |
| 788 | { |
| 789 | if (yield == NULL) yield = s; |
| 790 | if (*(++p) == 0) |
| 791 | { |
| 792 | if (!space_follows || s[1] == ' ' || s[1] == '\n' ) return yield; |
| 793 | yield = NULL; |
| 794 | p = t; |
| 795 | } |
| 796 | cl = tolower(*p); |
| 797 | cu = toupper(*p); |
| 798 | s++; |
| 799 | } |
| 800 | else if (yield != NULL) |
| 801 | { |
| 802 | yield = NULL; |
| 803 | p = t; |
| 804 | cl = tolower(*p); |
| 805 | cu = toupper(*p); |
| 806 | } |
| 807 | else s++; |
| 808 | } |
| 809 | return NULL; |
| 810 | } |
| 811 | |
| 812 | |
| 813 | |
| 814 | #ifdef COMPILE_UTILITY |
| 815 | /* Dummy version for this function; it should never be called */ |
| 816 | static void |
| 817 | gstring_grow(gstring * g, int count) |
| 818 | { |
| 819 | assert(FALSE); |
| 820 | } |
| 821 | #endif |
| 822 | |
| 823 | |
| 824 | |
| 825 | #ifndef COMPILE_UTILITY |
| 826 | /************************************************* |
| 827 | * Get next string from separated list * |
| 828 | *************************************************/ |
| 829 | |
| 830 | /* Leading and trailing space is removed from each item. The separator in the |
| 831 | list is controlled by the int pointed to by the separator argument as follows: |
| 832 | |
| 833 | If the value is > 0 it is used as the separator. This is typically used for |
| 834 | sublists such as slash-separated options. The value is always a printing |
| 835 | character. |
| 836 | |
| 837 | (If the value is actually > UCHAR_MAX there is only one item in the list. |
| 838 | This is used for some cases when called via functions that sometimes |
| 839 | plough through lists, and sometimes are given single items.) |
| 840 | |
| 841 | If the value is <= 0, the string is inspected for a leading <x, where x is an |
| 842 | ispunct() or an iscntrl() character. If found, x is used as the separator. If |
| 843 | not found: |
| 844 | |
| 845 | (a) if separator == 0, ':' is used |
| 846 | (b) if separator <0, -separator is used |
| 847 | |
| 848 | In all cases the value of the separator that is used is written back to the |
| 849 | int so that it is used on subsequent calls as we progress through the list. |
| 850 | |
| 851 | A literal ispunct() separator can be represented in an item by doubling, but |
| 852 | there is no way to include an iscntrl() separator as part of the data. |
| 853 | |
| 854 | Arguments: |
| 855 | listptr points to a pointer to the current start of the list; the |
| 856 | pointer gets updated to point after the end of the next item |
| 857 | separator a pointer to the separator character in an int (see above) |
| 858 | buffer where to put a copy of the next string in the list; or |
| 859 | NULL if the next string is returned in new memory |
| 860 | buflen when buffer is not NULL, the size of buffer; otherwise ignored |
| 861 | |
| 862 | Returns: pointer to buffer, containing the next substring, |
| 863 | or NULL if no more substrings |
| 864 | */ |
| 865 | |
| 866 | uschar * |
| 867 | string_nextinlist(const uschar **listptr, int *separator, uschar *buffer, int buflen) |
| 868 | { |
| 869 | int sep = *separator; |
| 870 | const uschar *s = *listptr; |
| 871 | BOOL sep_is_special; |
| 872 | |
| 873 | if (!s) return NULL; |
| 874 | |
| 875 | /* This allows for a fixed specified separator to be an iscntrl() character, |
| 876 | but at the time of implementation, this is never the case. However, it's best |
| 877 | to be conservative. */ |
| 878 | |
| 879 | while (isspace(*s) && *s != sep) s++; |
| 880 | |
| 881 | /* A change of separator is permitted, so look for a leading '<' followed by an |
| 882 | allowed character. */ |
| 883 | |
| 884 | if (sep <= 0) |
| 885 | { |
| 886 | if (*s == '<' && (ispunct(s[1]) || iscntrl(s[1]))) |
| 887 | { |
| 888 | sep = s[1]; |
| 889 | if (*++s) ++s; |
| 890 | while (isspace(*s) && *s != sep) s++; |
| 891 | } |
| 892 | else |
| 893 | sep = sep ? -sep : ':'; |
| 894 | *separator = sep; |
| 895 | } |
| 896 | |
| 897 | /* An empty string has no list elements */ |
| 898 | |
| 899 | if (!*s) return NULL; |
| 900 | |
| 901 | /* Note whether whether or not the separator is an iscntrl() character. */ |
| 902 | |
| 903 | sep_is_special = iscntrl(sep); |
| 904 | |
| 905 | /* Handle the case when a buffer is provided. */ |
| 906 | |
| 907 | if (buffer) |
| 908 | { |
| 909 | int p = 0; |
| 910 | for (; *s; s++) |
| 911 | { |
| 912 | if (*s == sep && (*(++s) != sep || sep_is_special)) break; |
| 913 | if (p < buflen - 1) buffer[p++] = *s; |
| 914 | } |
| 915 | while (p > 0 && isspace(buffer[p-1])) p--; |
| 916 | buffer[p] = '\0'; |
| 917 | } |
| 918 | |
| 919 | /* Handle the case when a buffer is not provided. */ |
| 920 | |
| 921 | else |
| 922 | { |
| 923 | gstring * g = NULL; |
| 924 | |
| 925 | /* We know that *s != 0 at this point. However, it might be pointing to a |
| 926 | separator, which could indicate an empty string, or (if an ispunct() |
| 927 | character) could be doubled to indicate a separator character as data at the |
| 928 | start of a string. Avoid getting working memory for an empty item. */ |
| 929 | |
| 930 | if (*s == sep) |
| 931 | { |
| 932 | s++; |
| 933 | if (*s != sep || sep_is_special) |
| 934 | { |
| 935 | *listptr = s; |
| 936 | return string_copy(US""); |
| 937 | } |
| 938 | } |
| 939 | |
| 940 | /* Not an empty string; the first character is guaranteed to be a data |
| 941 | character. */ |
| 942 | |
| 943 | for (;;) |
| 944 | { |
| 945 | const uschar * ss; |
| 946 | for (ss = s + 1; *ss && *ss != sep; ) ss++; |
| 947 | g = string_catn(g, s, ss-s); |
| 948 | s = ss; |
| 949 | if (!*s || *++s != sep || sep_is_special) break; |
| 950 | } |
| 951 | while (g->ptr > 0 && isspace(g->s[g->ptr-1])) g->ptr--; |
| 952 | buffer = string_from_gstring(g); |
| 953 | gstring_release_unused(g); |
| 954 | } |
| 955 | |
| 956 | /* Update the current pointer and return the new string */ |
| 957 | |
| 958 | *listptr = s; |
| 959 | return buffer; |
| 960 | } |
| 961 | |
| 962 | |
| 963 | static const uschar * |
| 964 | Ustrnchr(const uschar * s, int c, unsigned * len) |
| 965 | { |
| 966 | unsigned siz = *len; |
| 967 | while (siz) |
| 968 | { |
| 969 | if (!*s) return NULL; |
| 970 | if (*s == c) |
| 971 | { |
| 972 | *len = siz; |
| 973 | return s; |
| 974 | } |
| 975 | s++; |
| 976 | siz--; |
| 977 | } |
| 978 | return NULL; |
| 979 | } |
| 980 | |
| 981 | |
| 982 | /************************************************ |
| 983 | * Add element to separated list * |
| 984 | ************************************************/ |
| 985 | /* This function is used to build a list, returning an allocated null-terminated |
| 986 | growable string. The given element has any embedded separator characters |
| 987 | doubled. |
| 988 | |
| 989 | Despite having the same growable-string interface as string_cat() the list is |
| 990 | always returned null-terminated. |
| 991 | |
| 992 | Arguments: |
| 993 | list expanding-string for the list that is being built, or NULL |
| 994 | if this is a new list that has no contents yet |
| 995 | sep list separator character |
| 996 | ele new element to be appended to the list |
| 997 | |
| 998 | Returns: pointer to the start of the list, changed if copied for expansion. |
| 999 | */ |
| 1000 | |
| 1001 | gstring * |
| 1002 | string_append_listele(gstring * list, uschar sep, const uschar * ele) |
| 1003 | { |
| 1004 | uschar * sp; |
| 1005 | |
| 1006 | if (list && list->ptr) |
| 1007 | list = string_catn(list, &sep, 1); |
| 1008 | |
| 1009 | while((sp = Ustrchr(ele, sep))) |
| 1010 | { |
| 1011 | list = string_catn(list, ele, sp-ele+1); |
| 1012 | list = string_catn(list, &sep, 1); |
| 1013 | ele = sp+1; |
| 1014 | } |
| 1015 | list = string_cat(list, ele); |
| 1016 | (void) string_from_gstring(list); |
| 1017 | return list; |
| 1018 | } |
| 1019 | |
| 1020 | |
| 1021 | gstring * |
| 1022 | string_append_listele_n(gstring * list, uschar sep, const uschar * ele, |
| 1023 | unsigned len) |
| 1024 | { |
| 1025 | const uschar * sp; |
| 1026 | |
| 1027 | if (list && list->ptr) |
| 1028 | list = string_catn(list, &sep, 1); |
| 1029 | |
| 1030 | while((sp = Ustrnchr(ele, sep, &len))) |
| 1031 | { |
| 1032 | list = string_catn(list, ele, sp-ele+1); |
| 1033 | list = string_catn(list, &sep, 1); |
| 1034 | ele = sp+1; |
| 1035 | len--; |
| 1036 | } |
| 1037 | list = string_catn(list, ele, len); |
| 1038 | (void) string_from_gstring(list); |
| 1039 | return list; |
| 1040 | } |
| 1041 | |
| 1042 | |
| 1043 | |
| 1044 | /* A slightly-bogus listmaker utility; the separator is a string so |
| 1045 | can be multiple chars - there is no checking for the element content |
| 1046 | containing any of the separator. */ |
| 1047 | |
| 1048 | gstring * |
| 1049 | string_append2_listele_n(gstring * list, const uschar * sepstr, |
| 1050 | const uschar * ele, unsigned len) |
| 1051 | { |
| 1052 | if (list && list->ptr) |
| 1053 | list = string_cat(list, sepstr); |
| 1054 | |
| 1055 | list = string_catn(list, ele, len); |
| 1056 | (void) string_from_gstring(list); |
| 1057 | return list; |
| 1058 | } |
| 1059 | |
| 1060 | |
| 1061 | |
| 1062 | /************************************************/ |
| 1063 | /* Add more space to a growable-string. The caller should check |
| 1064 | first if growth is required. The gstring struct is modified on |
| 1065 | return; specifically, the string-base-pointer may have been changed. |
| 1066 | |
| 1067 | Arguments: |
| 1068 | g the growable-string |
| 1069 | count amount needed for g->ptr to increase by |
| 1070 | */ |
| 1071 | |
| 1072 | static void |
| 1073 | gstring_grow(gstring * g, int count) |
| 1074 | { |
| 1075 | int p = g->ptr; |
| 1076 | int oldsize = g->size; |
| 1077 | BOOL tainted = is_tainted(g->s); |
| 1078 | |
| 1079 | /* Mostly, string_cat() is used to build small strings of a few hundred |
| 1080 | characters at most. There are times, however, when the strings are very much |
| 1081 | longer (for example, a lookup that returns a vast number of alias addresses). |
| 1082 | To try to keep things reasonable, we use increments whose size depends on the |
| 1083 | existing length of the string. */ |
| 1084 | |
| 1085 | unsigned inc = oldsize < 4096 ? 127 : 1023; |
| 1086 | |
| 1087 | if (count <= 0) return; |
| 1088 | g->size = (p + count + inc + 1) & ~inc; /* one for a NUL */ |
| 1089 | |
| 1090 | /* Try to extend an existing allocation. If the result of calling |
| 1091 | store_extend() is false, either there isn't room in the current memory block, |
| 1092 | or this string is not the top item on the dynamic store stack. We then have |
| 1093 | to get a new chunk of store and copy the old string. When building large |
| 1094 | strings, it is helpful to call store_release() on the old string, to release |
| 1095 | memory blocks that have become empty. (The block will be freed if the string |
| 1096 | is at its start.) However, we can do this only if we know that the old string |
| 1097 | was the last item on the dynamic memory stack. This is the case if it matches |
| 1098 | store_last_get. */ |
| 1099 | |
| 1100 | if (!store_extend(g->s, tainted, oldsize, g->size)) |
| 1101 | g->s = store_newblock(g->s, tainted, g->size, p); |
| 1102 | } |
| 1103 | |
| 1104 | |
| 1105 | |
| 1106 | /************************************************* |
| 1107 | * Add chars to string * |
| 1108 | *************************************************/ |
| 1109 | /* This function is used when building up strings of unknown length. Room is |
| 1110 | always left for a terminating zero to be added to the string that is being |
| 1111 | built. This function does not require the string that is being added to be NUL |
| 1112 | terminated, because the number of characters to add is given explicitly. It is |
| 1113 | sometimes called to extract parts of other strings. |
| 1114 | |
| 1115 | Arguments: |
| 1116 | string points to the start of the string that is being built, or NULL |
| 1117 | if this is a new string that has no contents yet |
| 1118 | s points to characters to add |
| 1119 | count count of characters to add; must not exceed the length of s, if s |
| 1120 | is a C string. |
| 1121 | |
| 1122 | Returns: pointer to the start of the string, changed if copied for expansion. |
| 1123 | Note that a NUL is not added, though space is left for one. This is |
| 1124 | because string_cat() is often called multiple times to build up a |
| 1125 | string - there's no point adding the NUL till the end. |
| 1126 | |
| 1127 | */ |
| 1128 | /* coverity[+alloc] */ |
| 1129 | |
| 1130 | gstring * |
| 1131 | string_catn(gstring * g, const uschar *s, int count) |
| 1132 | { |
| 1133 | int p; |
| 1134 | BOOL srctaint = is_tainted(s); |
| 1135 | |
| 1136 | if (!g) |
| 1137 | { |
| 1138 | unsigned inc = count < 4096 ? 127 : 1023; |
| 1139 | unsigned size = ((count + inc) & ~inc) + 1; |
| 1140 | g = string_get_tainted(size, srctaint); |
| 1141 | } |
| 1142 | else if (srctaint && !is_tainted(g->s)) |
| 1143 | gstring_rebuffer(g); |
| 1144 | |
| 1145 | p = g->ptr; |
| 1146 | if (p + count >= g->size) |
| 1147 | gstring_grow(g, count); |
| 1148 | |
| 1149 | /* Because we always specify the exact number of characters to copy, we can |
| 1150 | use memcpy(), which is likely to be more efficient than strncopy() because the |
| 1151 | latter has to check for zero bytes. */ |
| 1152 | |
| 1153 | memcpy(g->s + p, s, count); |
| 1154 | g->ptr = p + count; |
| 1155 | return g; |
| 1156 | } |
| 1157 | |
| 1158 | |
| 1159 | gstring * |
| 1160 | string_cat(gstring *string, const uschar *s) |
| 1161 | { |
| 1162 | return string_catn(string, s, Ustrlen(s)); |
| 1163 | } |
| 1164 | |
| 1165 | |
| 1166 | |
| 1167 | /************************************************* |
| 1168 | * Append strings to another string * |
| 1169 | *************************************************/ |
| 1170 | |
| 1171 | /* This function can be used to build a string from many other strings. |
| 1172 | It calls string_cat() to do the dirty work. |
| 1173 | |
| 1174 | Arguments: |
| 1175 | string expanding-string that is being built, or NULL |
| 1176 | if this is a new string that has no contents yet |
| 1177 | count the number of strings to append |
| 1178 | ... "count" uschar* arguments, which must be valid zero-terminated |
| 1179 | C strings |
| 1180 | |
| 1181 | Returns: pointer to the start of the string, changed if copied for expansion. |
| 1182 | The string is not zero-terminated - see string_cat() above. |
| 1183 | */ |
| 1184 | |
| 1185 | __inline__ gstring * |
| 1186 | string_append(gstring *string, int count, ...) |
| 1187 | { |
| 1188 | va_list ap; |
| 1189 | |
| 1190 | va_start(ap, count); |
| 1191 | while (count-- > 0) |
| 1192 | { |
| 1193 | uschar *t = va_arg(ap, uschar *); |
| 1194 | string = string_cat(string, t); |
| 1195 | } |
| 1196 | va_end(ap); |
| 1197 | |
| 1198 | return string; |
| 1199 | } |
| 1200 | #endif |
| 1201 | |
| 1202 | |
| 1203 | |
| 1204 | /************************************************* |
| 1205 | * Format a string with length checks * |
| 1206 | *************************************************/ |
| 1207 | |
| 1208 | /* This function is used to format a string with checking of the length of the |
| 1209 | output for all conversions. It protects Exim from absent-mindedness when |
| 1210 | calling functions like debug_printf and string_sprintf, and elsewhere. There |
| 1211 | are two different entry points to what is actually the same function, depending |
| 1212 | on whether the variable length list of data arguments are given explicitly or |
| 1213 | as a va_list item. |
| 1214 | |
| 1215 | The formats are the usual printf() ones, with some omissions (never used) and |
| 1216 | three additions for strings: %S forces lower case, %T forces upper case, and |
| 1217 | %#s or %#S prints nothing for a NULL string. Without the # "NULL" is printed |
| 1218 | (useful in debugging). There is also the addition of %D and %M, which insert |
| 1219 | the date in the form used for datestamped log files. |
| 1220 | |
| 1221 | Arguments: |
| 1222 | buffer a buffer in which to put the formatted string |
| 1223 | buflen the length of the buffer |
| 1224 | format the format string - deliberately char * and not uschar * |
| 1225 | ... or ap variable list of supplementary arguments |
| 1226 | |
| 1227 | Returns: TRUE if the result fitted in the buffer |
| 1228 | */ |
| 1229 | |
| 1230 | BOOL |
| 1231 | string_format_trc(uschar * buffer, int buflen, |
| 1232 | const uschar * func, unsigned line, const char * format, ...) |
| 1233 | { |
| 1234 | gstring g = { .size = buflen, .ptr = 0, .s = buffer }, *gp; |
| 1235 | va_list ap; |
| 1236 | va_start(ap, format); |
| 1237 | gp = string_vformat_trc(&g, func, line, STRING_SPRINTF_BUFFER_SIZE, |
| 1238 | 0, format, ap); |
| 1239 | va_end(ap); |
| 1240 | g.s[g.ptr] = '\0'; |
| 1241 | return !!gp; |
| 1242 | } |
| 1243 | |
| 1244 | |
| 1245 | |
| 1246 | /* Copy the content of a string to tainted memory */ |
| 1247 | static void |
| 1248 | gstring_rebuffer(gstring * g) |
| 1249 | { |
| 1250 | uschar * s = store_get(g->size, TRUE); |
| 1251 | memcpy(s, g->s, g->ptr); |
| 1252 | g->s = s; |
| 1253 | } |
| 1254 | |
| 1255 | |
| 1256 | |
| 1257 | /* Build or append to a growing-string, sprintf-style. |
| 1258 | |
| 1259 | If the "extend" flag is true, the string passed in can be NULL, |
| 1260 | empty, or non-empty. Growing is subject to an overall limit given |
| 1261 | by the size_limit argument. |
| 1262 | |
| 1263 | If the "extend" flag is false, the string passed in may not be NULL, |
| 1264 | will not be grown, and is usable in the original place after return. |
| 1265 | The return value can be NULL to signify overflow. |
| 1266 | |
| 1267 | Returns the possibly-new (if copy for growth or taint-handling was needed) |
| 1268 | string, not nul-terminated. |
| 1269 | */ |
| 1270 | |
| 1271 | gstring * |
| 1272 | string_vformat_trc(gstring * g, const uschar * func, unsigned line, |
| 1273 | unsigned size_limit, unsigned flags, const char *format, va_list ap) |
| 1274 | { |
| 1275 | enum ltypes { L_NORMAL=1, L_SHORT=2, L_LONG=3, L_LONGLONG=4, L_LONGDOUBLE=5, L_SIZE=6 }; |
| 1276 | |
| 1277 | int width, precision, off, lim, need; |
| 1278 | const char * fp = format; /* Deliberately not unsigned */ |
| 1279 | BOOL dest_tainted = FALSE; |
| 1280 | |
| 1281 | string_datestamp_offset = -1; /* Datestamp not inserted */ |
| 1282 | string_datestamp_length = 0; /* Datestamp not inserted */ |
| 1283 | string_datestamp_type = 0; /* Datestamp not inserted */ |
| 1284 | |
| 1285 | #ifdef COMPILE_UTILITY |
| 1286 | assert(!(flags & SVFMT_EXTEND)); |
| 1287 | assert(g); |
| 1288 | #else |
| 1289 | |
| 1290 | /* Ensure we have a string, to save on checking later */ |
| 1291 | if (!g) g = string_get(16); |
| 1292 | else if (!(flags & SVFMT_TAINT_NOCHK)) dest_tainted = is_tainted(g->s); |
| 1293 | |
| 1294 | if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(format)) |
| 1295 | { |
| 1296 | #ifndef MACRO_PREDEF |
| 1297 | if (!(flags & SVFMT_REBUFFER)) |
| 1298 | die_tainted(US"string_vformat", func, line); |
| 1299 | #endif |
| 1300 | gstring_rebuffer(g); |
| 1301 | dest_tainted = TRUE; |
| 1302 | } |
| 1303 | #endif /*!COMPILE_UTILITY*/ |
| 1304 | |
| 1305 | lim = g->size - 1; /* leave one for a nul */ |
| 1306 | off = g->ptr; /* remember initial offset in gstring */ |
| 1307 | |
| 1308 | /* Scan the format and handle the insertions */ |
| 1309 | |
| 1310 | while (*fp) |
| 1311 | { |
| 1312 | int length = L_NORMAL; |
| 1313 | int *nptr; |
| 1314 | int slen; |
| 1315 | const char *null = "NULL"; /* ) These variables */ |
| 1316 | const char *item_start, *s; /* ) are deliberately */ |
| 1317 | char newformat[16]; /* ) not unsigned */ |
| 1318 | char * gp = CS g->s + g->ptr; /* ) */ |
| 1319 | |
| 1320 | /* Non-% characters just get copied verbatim */ |
| 1321 | |
| 1322 | if (*fp != '%') |
| 1323 | { |
| 1324 | /* Avoid string_copyn() due to COMPILE_UTILITY */ |
| 1325 | if ((need = g->ptr + 1) > lim) |
| 1326 | { |
| 1327 | if (!(flags & SVFMT_EXTEND) || need > size_limit) return NULL; |
| 1328 | gstring_grow(g, 1); |
| 1329 | lim = g->size - 1; |
| 1330 | } |
| 1331 | g->s[g->ptr++] = (uschar) *fp++; |
| 1332 | continue; |
| 1333 | } |
| 1334 | |
| 1335 | /* Deal with % characters. Pick off the width and precision, for checking |
| 1336 | strings, skipping over the flag and modifier characters. */ |
| 1337 | |
| 1338 | item_start = fp; |
| 1339 | width = precision = -1; |
| 1340 | |
| 1341 | if (strchr("-+ #0", *(++fp)) != NULL) |
| 1342 | { |
| 1343 | if (*fp == '#') null = ""; |
| 1344 | fp++; |
| 1345 | } |
| 1346 | |
| 1347 | if (isdigit((uschar)*fp)) |
| 1348 | { |
| 1349 | width = *fp++ - '0'; |
| 1350 | while (isdigit((uschar)*fp)) width = width * 10 + *fp++ - '0'; |
| 1351 | } |
| 1352 | else if (*fp == '*') |
| 1353 | { |
| 1354 | width = va_arg(ap, int); |
| 1355 | fp++; |
| 1356 | } |
| 1357 | |
| 1358 | if (*fp == '.') |
| 1359 | if (*(++fp) == '*') |
| 1360 | { |
| 1361 | precision = va_arg(ap, int); |
| 1362 | fp++; |
| 1363 | } |
| 1364 | else |
| 1365 | for (precision = 0; isdigit((uschar)*fp); fp++) |
| 1366 | precision = precision*10 + *fp - '0'; |
| 1367 | |
| 1368 | /* Skip over 'h', 'L', 'l', 'll' and 'z', remembering the item length */ |
| 1369 | |
| 1370 | if (*fp == 'h') |
| 1371 | { fp++; length = L_SHORT; } |
| 1372 | else if (*fp == 'L') |
| 1373 | { fp++; length = L_LONGDOUBLE; } |
| 1374 | else if (*fp == 'l') |
| 1375 | if (fp[1] == 'l') |
| 1376 | { fp += 2; length = L_LONGLONG; } |
| 1377 | else |
| 1378 | { fp++; length = L_LONG; } |
| 1379 | else if (*fp == 'z') |
| 1380 | { fp++; length = L_SIZE; } |
| 1381 | |
| 1382 | /* Handle each specific format type. */ |
| 1383 | |
| 1384 | switch (*fp++) |
| 1385 | { |
| 1386 | case 'n': |
| 1387 | nptr = va_arg(ap, int *); |
| 1388 | *nptr = g->ptr - off; |
| 1389 | break; |
| 1390 | |
| 1391 | case 'd': |
| 1392 | case 'o': |
| 1393 | case 'u': |
| 1394 | case 'x': |
| 1395 | case 'X': |
| 1396 | width = length > L_LONG ? 24 : 12; |
| 1397 | if ((need = g->ptr + width) > lim) |
| 1398 | { |
| 1399 | if (!(flags & SVFMT_EXTEND) || need >= size_limit) return NULL; |
| 1400 | gstring_grow(g, width); |
| 1401 | lim = g->size - 1; |
| 1402 | gp = CS g->s + g->ptr; |
| 1403 | } |
| 1404 | strncpy(newformat, item_start, fp - item_start); |
| 1405 | newformat[fp - item_start] = 0; |
| 1406 | |
| 1407 | /* Short int is promoted to int when passing through ..., so we must use |
| 1408 | int for va_arg(). */ |
| 1409 | |
| 1410 | switch(length) |
| 1411 | { |
| 1412 | case L_SHORT: |
| 1413 | case L_NORMAL: |
| 1414 | g->ptr += sprintf(gp, newformat, va_arg(ap, int)); break; |
| 1415 | case L_LONG: |
| 1416 | g->ptr += sprintf(gp, newformat, va_arg(ap, long int)); break; |
| 1417 | case L_LONGLONG: |
| 1418 | g->ptr += sprintf(gp, newformat, va_arg(ap, LONGLONG_T)); break; |
| 1419 | case L_SIZE: |
| 1420 | g->ptr += sprintf(gp, newformat, va_arg(ap, size_t)); break; |
| 1421 | } |
| 1422 | break; |
| 1423 | |
| 1424 | case 'p': |
| 1425 | { |
| 1426 | void * ptr; |
| 1427 | if ((need = g->ptr + 24) > lim) |
| 1428 | { |
| 1429 | if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; |
| 1430 | gstring_grow(g, 24); |
| 1431 | lim = g->size - 1; |
| 1432 | gp = CS g->s + g->ptr; |
| 1433 | } |
| 1434 | /* sprintf() saying "(nil)" for a null pointer seems unreliable. |
| 1435 | Handle it explicitly. */ |
| 1436 | if ((ptr = va_arg(ap, void *))) |
| 1437 | { |
| 1438 | strncpy(newformat, item_start, fp - item_start); |
| 1439 | newformat[fp - item_start] = 0; |
| 1440 | g->ptr += sprintf(gp, newformat, ptr); |
| 1441 | } |
| 1442 | else |
| 1443 | g->ptr += sprintf(gp, "(nil)"); |
| 1444 | } |
| 1445 | break; |
| 1446 | |
| 1447 | /* %f format is inherently insecure if the numbers that it may be |
| 1448 | handed are unknown (e.g. 1e300). However, in Exim, %f is used for |
| 1449 | printing load averages, and these are actually stored as integers |
| 1450 | (load average * 1000) so the size of the numbers is constrained. |
| 1451 | It is also used for formatting sending rates, where the simplicity |
| 1452 | of the format prevents overflow. */ |
| 1453 | |
| 1454 | case 'f': |
| 1455 | case 'e': |
| 1456 | case 'E': |
| 1457 | case 'g': |
| 1458 | case 'G': |
| 1459 | if (precision < 0) precision = 6; |
| 1460 | if ((need = g->ptr + precision + 8) > lim) |
| 1461 | { |
| 1462 | if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; |
| 1463 | gstring_grow(g, precision+8); |
| 1464 | lim = g->size - 1; |
| 1465 | gp = CS g->s + g->ptr; |
| 1466 | } |
| 1467 | strncpy(newformat, item_start, fp - item_start); |
| 1468 | newformat[fp-item_start] = 0; |
| 1469 | if (length == L_LONGDOUBLE) |
| 1470 | g->ptr += sprintf(gp, newformat, va_arg(ap, long double)); |
| 1471 | else |
| 1472 | g->ptr += sprintf(gp, newformat, va_arg(ap, double)); |
| 1473 | break; |
| 1474 | |
| 1475 | /* String types */ |
| 1476 | |
| 1477 | case '%': |
| 1478 | if ((need = g->ptr + 1) > lim) |
| 1479 | { |
| 1480 | if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; |
| 1481 | gstring_grow(g, 1); |
| 1482 | lim = g->size - 1; |
| 1483 | } |
| 1484 | g->s[g->ptr++] = (uschar) '%'; |
| 1485 | break; |
| 1486 | |
| 1487 | case 'c': |
| 1488 | if ((need = g->ptr + 1) > lim) |
| 1489 | { |
| 1490 | if (!(flags & SVFMT_EXTEND || need >= size_limit)) return NULL; |
| 1491 | gstring_grow(g, 1); |
| 1492 | lim = g->size - 1; |
| 1493 | } |
| 1494 | g->s[g->ptr++] = (uschar) va_arg(ap, int); |
| 1495 | break; |
| 1496 | |
| 1497 | case 'D': /* Insert daily datestamp for log file names */ |
| 1498 | s = CS tod_stamp(tod_log_datestamp_daily); |
| 1499 | string_datestamp_offset = g->ptr; /* Passed back via global */ |
| 1500 | string_datestamp_length = Ustrlen(s); /* Passed back via global */ |
| 1501 | string_datestamp_type = tod_log_datestamp_daily; |
| 1502 | slen = string_datestamp_length; |
| 1503 | goto INSERT_STRING; |
| 1504 | |
| 1505 | case 'M': /* Insert monthly datestamp for log file names */ |
| 1506 | s = CS tod_stamp(tod_log_datestamp_monthly); |
| 1507 | string_datestamp_offset = g->ptr; /* Passed back via global */ |
| 1508 | string_datestamp_length = Ustrlen(s); /* Passed back via global */ |
| 1509 | string_datestamp_type = tod_log_datestamp_monthly; |
| 1510 | slen = string_datestamp_length; |
| 1511 | goto INSERT_STRING; |
| 1512 | |
| 1513 | case 's': |
| 1514 | case 'S': /* Forces *lower* case */ |
| 1515 | case 'T': /* Forces *upper* case */ |
| 1516 | s = va_arg(ap, char *); |
| 1517 | |
| 1518 | if (!s) s = null; |
| 1519 | slen = Ustrlen(s); |
| 1520 | |
| 1521 | if (!(flags & SVFMT_TAINT_NOCHK) && !dest_tainted && is_tainted(s)) |
| 1522 | if (flags & SVFMT_REBUFFER) |
| 1523 | { |
| 1524 | gstring_rebuffer(g); |
| 1525 | gp = CS g->s + g->ptr; |
| 1526 | dest_tainted = TRUE; |
| 1527 | } |
| 1528 | #ifndef MACRO_PREDEF |
| 1529 | else |
| 1530 | die_tainted(US"string_vformat", func, line); |
| 1531 | #endif |
| 1532 | |
| 1533 | INSERT_STRING: /* Come to from %D or %M above */ |
| 1534 | |
| 1535 | { |
| 1536 | BOOL truncated = FALSE; |
| 1537 | |
| 1538 | /* If the width is specified, check that there is a precision |
| 1539 | set; if not, set it to the width to prevent overruns of long |
| 1540 | strings. */ |
| 1541 | |
| 1542 | if (width >= 0) |
| 1543 | { |
| 1544 | if (precision < 0) precision = width; |
| 1545 | } |
| 1546 | |
| 1547 | /* If a width is not specified and the precision is specified, set |
| 1548 | the width to the precision, or the string length if shorted. */ |
| 1549 | |
| 1550 | else if (precision >= 0) |
| 1551 | width = precision < slen ? precision : slen; |
| 1552 | |
| 1553 | /* If neither are specified, set them both to the string length. */ |
| 1554 | |
| 1555 | else |
| 1556 | width = precision = slen; |
| 1557 | |
| 1558 | if ((need = g->ptr + width) >= size_limit || !(flags & SVFMT_EXTEND)) |
| 1559 | { |
| 1560 | if (g->ptr == lim) return NULL; |
| 1561 | if (need > lim) |
| 1562 | { |
| 1563 | truncated = TRUE; |
| 1564 | width = precision = lim - g->ptr - 1; |
| 1565 | if (width < 0) width = 0; |
| 1566 | if (precision < 0) precision = 0; |
| 1567 | } |
| 1568 | } |
| 1569 | else if (need > lim) |
| 1570 | { |
| 1571 | gstring_grow(g, width); |
| 1572 | lim = g->size - 1; |
| 1573 | gp = CS g->s + g->ptr; |
| 1574 | } |
| 1575 | |
| 1576 | g->ptr += sprintf(gp, "%*.*s", width, precision, s); |
| 1577 | if (fp[-1] == 'S') |
| 1578 | while (*gp) { *gp = tolower(*gp); gp++; } |
| 1579 | else if (fp[-1] == 'T') |
| 1580 | while (*gp) { *gp = toupper(*gp); gp++; } |
| 1581 | |
| 1582 | if (truncated) return NULL; |
| 1583 | break; |
| 1584 | } |
| 1585 | |
| 1586 | /* Some things are never used in Exim; also catches junk. */ |
| 1587 | |
| 1588 | default: |
| 1589 | strncpy(newformat, item_start, fp - item_start); |
| 1590 | newformat[fp-item_start] = 0; |
| 1591 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, "string_format: unsupported type " |
| 1592 | "in \"%s\" in \"%s\"", newformat, format); |
| 1593 | break; |
| 1594 | } |
| 1595 | } |
| 1596 | |
| 1597 | if (g->ptr > g->size) |
| 1598 | log_write(0, LOG_MAIN|LOG_PANIC_DIE, |
| 1599 | "string_format internal error: caller %s %d", func, line); |
| 1600 | return g; |
| 1601 | } |
| 1602 | |
| 1603 | |
| 1604 | |
| 1605 | #ifndef COMPILE_UTILITY |
| 1606 | /************************************************* |
| 1607 | * Generate an "open failed" message * |
| 1608 | *************************************************/ |
| 1609 | |
| 1610 | /* This function creates a message after failure to open a file. It includes a |
| 1611 | string supplied as data, adds the strerror() text, and if the failure was |
| 1612 | "Permission denied", reads and includes the euid and egid. |
| 1613 | |
| 1614 | Arguments: |
| 1615 | eno the value of errno after the failure |
| 1616 | format a text format string - deliberately not uschar * |
| 1617 | ... arguments for the format string |
| 1618 | |
| 1619 | Returns: a message, in dynamic store |
| 1620 | */ |
| 1621 | |
| 1622 | uschar * |
| 1623 | string_open_failed_trc(int eno, const uschar * func, unsigned line, |
| 1624 | const char *format, ...) |
| 1625 | { |
| 1626 | va_list ap; |
| 1627 | gstring * g = string_get(1024); |
| 1628 | |
| 1629 | g = string_catn(g, US"failed to open ", 15); |
| 1630 | |
| 1631 | /* Use the checked formatting routine to ensure that the buffer |
| 1632 | does not overflow. It should not, since this is called only for internally |
| 1633 | specified messages. If it does, the message just gets truncated, and there |
| 1634 | doesn't seem much we can do about that. */ |
| 1635 | |
| 1636 | va_start(ap, format); |
| 1637 | (void) string_vformat_trc(g, func, line, STRING_SPRINTF_BUFFER_SIZE, |
| 1638 | 0, format, ap); |
| 1639 | string_from_gstring(g); |
| 1640 | gstring_release_unused(g); |
| 1641 | va_end(ap); |
| 1642 | |
| 1643 | return eno == EACCES |
| 1644 | ? string_sprintf("%s: %s (euid=%ld egid=%ld)", g->s, strerror(eno), |
| 1645 | (long int)geteuid(), (long int)getegid()) |
| 1646 | : string_sprintf("%s: %s", g->s, strerror(eno)); |
| 1647 | } |
| 1648 | #endif /* COMPILE_UTILITY */ |
| 1649 | |
| 1650 | |
| 1651 | |
| 1652 | |
| 1653 | |
| 1654 | #ifndef COMPILE_UTILITY |
| 1655 | /* qsort(3), currently used to sort the environment variables |
| 1656 | for -bP environment output, needs a function to compare two pointers to string |
| 1657 | pointers. Here it is. */ |
| 1658 | |
| 1659 | int |
| 1660 | string_compare_by_pointer(const void *a, const void *b) |
| 1661 | { |
| 1662 | return Ustrcmp(* CUSS a, * CUSS b); |
| 1663 | } |
| 1664 | #endif /* COMPILE_UTILITY */ |
| 1665 | |
| 1666 | |
| 1667 | |
| 1668 | |
| 1669 | /************************************************* |
| 1670 | ************************************************** |
| 1671 | * Stand-alone test program * |
| 1672 | ************************************************** |
| 1673 | *************************************************/ |
| 1674 | |
| 1675 | #ifdef STAND_ALONE |
| 1676 | int main(void) |
| 1677 | { |
| 1678 | uschar buffer[256]; |
| 1679 | |
| 1680 | printf("Testing is_ip_address\n"); |
| 1681 | |
| 1682 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) |
| 1683 | { |
| 1684 | int offset; |
| 1685 | buffer[Ustrlen(buffer) - 1] = 0; |
| 1686 | printf("%d\n", string_is_ip_address(buffer, NULL)); |
| 1687 | printf("%d %d %s\n", string_is_ip_address(buffer, &offset), offset, buffer); |
| 1688 | } |
| 1689 | |
| 1690 | printf("Testing string_nextinlist\n"); |
| 1691 | |
| 1692 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) |
| 1693 | { |
| 1694 | uschar *list = buffer; |
| 1695 | uschar *lp1, *lp2; |
| 1696 | uschar item[256]; |
| 1697 | int sep1 = 0; |
| 1698 | int sep2 = 0; |
| 1699 | |
| 1700 | if (*list == '<') |
| 1701 | { |
| 1702 | sep1 = sep2 = list[1]; |
| 1703 | list += 2; |
| 1704 | } |
| 1705 | |
| 1706 | lp1 = lp2 = list; |
| 1707 | for (;;) |
| 1708 | { |
| 1709 | uschar *item1 = string_nextinlist(&lp1, &sep1, item, sizeof(item)); |
| 1710 | uschar *item2 = string_nextinlist(&lp2, &sep2, NULL, 0); |
| 1711 | |
| 1712 | if (item1 == NULL && item2 == NULL) break; |
| 1713 | if (item == NULL || item2 == NULL || Ustrcmp(item1, item2) != 0) |
| 1714 | { |
| 1715 | printf("***ERROR\nitem1=\"%s\"\nitem2=\"%s\"\n", |
| 1716 | (item1 == NULL)? "NULL" : CS item1, |
| 1717 | (item2 == NULL)? "NULL" : CS item2); |
| 1718 | break; |
| 1719 | } |
| 1720 | else printf(" \"%s\"\n", CS item1); |
| 1721 | } |
| 1722 | } |
| 1723 | |
| 1724 | /* This is a horrible lash-up, but it serves its purpose. */ |
| 1725 | |
| 1726 | printf("Testing string_format\n"); |
| 1727 | |
| 1728 | while (fgets(CS buffer, sizeof(buffer), stdin) != NULL) |
| 1729 | { |
| 1730 | void *args[3]; |
| 1731 | long long llargs[3]; |
| 1732 | double dargs[3]; |
| 1733 | int dflag = 0; |
| 1734 | int llflag = 0; |
| 1735 | int n = 0; |
| 1736 | int count; |
| 1737 | int countset = 0; |
| 1738 | uschar format[256]; |
| 1739 | uschar outbuf[256]; |
| 1740 | uschar *s; |
| 1741 | buffer[Ustrlen(buffer) - 1] = 0; |
| 1742 | |
| 1743 | s = Ustrchr(buffer, ','); |
| 1744 | if (s == NULL) s = buffer + Ustrlen(buffer); |
| 1745 | |
| 1746 | Ustrncpy(format, buffer, s - buffer); |
| 1747 | format[s-buffer] = 0; |
| 1748 | |
| 1749 | if (*s == ',') s++; |
| 1750 | |
| 1751 | while (*s != 0) |
| 1752 | { |
| 1753 | uschar *ss = s; |
| 1754 | s = Ustrchr(ss, ','); |
| 1755 | if (s == NULL) s = ss + Ustrlen(ss); |
| 1756 | |
| 1757 | if (isdigit(*ss)) |
| 1758 | { |
| 1759 | Ustrncpy(outbuf, ss, s-ss); |
| 1760 | if (Ustrchr(outbuf, '.') != NULL) |
| 1761 | { |
| 1762 | dflag = 1; |
| 1763 | dargs[n++] = Ustrtod(outbuf, NULL); |
| 1764 | } |
| 1765 | else if (Ustrstr(outbuf, "ll") != NULL) |
| 1766 | { |
| 1767 | llflag = 1; |
| 1768 | llargs[n++] = strtoull(CS outbuf, NULL, 10); |
| 1769 | } |
| 1770 | else |
| 1771 | { |
| 1772 | args[n++] = (void *)Uatoi(outbuf); |
| 1773 | } |
| 1774 | } |
| 1775 | |
| 1776 | else if (Ustrcmp(ss, "*") == 0) |
| 1777 | { |
| 1778 | args[n++] = (void *)(&count); |
| 1779 | countset = 1; |
| 1780 | } |
| 1781 | |
| 1782 | else |
| 1783 | { |
| 1784 | uschar *sss = malloc(s - ss + 1); |
| 1785 | Ustrncpy(sss, ss, s-ss); |
| 1786 | args[n++] = sss; |
| 1787 | } |
| 1788 | |
| 1789 | if (*s == ',') s++; |
| 1790 | } |
| 1791 | |
| 1792 | if (!dflag && !llflag) |
| 1793 | printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, |
| 1794 | args[0], args[1], args[2])? "True" : "False"); |
| 1795 | |
| 1796 | else if (dflag) |
| 1797 | printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, |
| 1798 | dargs[0], dargs[1], dargs[2])? "True" : "False"); |
| 1799 | |
| 1800 | else printf("%s\n", string_format(outbuf, sizeof(outbuf), CS format, |
| 1801 | llargs[0], llargs[1], llargs[2])? "True" : "False"); |
| 1802 | |
| 1803 | printf("%s\n", CS outbuf); |
| 1804 | if (countset) printf("count=%d\n", count); |
| 1805 | } |
| 1806 | |
| 1807 | return 0; |
| 1808 | } |
| 1809 | #endif |
| 1810 | |
| 1811 | /* End of string.c */ |