| 1 | /* gzlib.c -- zlib functions common to reading and writing gzip files |
| 2 | * Copyright (C) 2004-2017 Mark Adler |
| 3 | * For conditions of distribution and use, see copyright notice in zlib.h |
| 4 | */ |
| 5 | |
| 6 | #include "zbuild.h" |
| 7 | #include "gzguts.h" |
| 8 | |
| 9 | #if defined(WIN32) && !defined(__BORLANDC__) |
| 10 | # define LSEEK _lseeki64 |
| 11 | #else |
| 12 | #if defined(_LARGEFILE64_SOURCE) && _LFS64_LARGEFILE-0 |
| 13 | # define LSEEK lseek64 |
| 14 | #else |
| 15 | # define LSEEK lseek |
| 16 | #endif |
| 17 | #endif |
| 18 | |
| 19 | /* Local functions */ |
| 20 | static void gz_reset(gz_state *); |
| 21 | static gzFile gz_open(const void *, int, const char *); |
| 22 | |
| 23 | /* Reset gzip file state */ |
| 24 | static void gz_reset(gz_state *state) { |
| 25 | state->x.have = 0; /* no output data available */ |
| 26 | if (state->mode == GZ_READ) { /* for reading ... */ |
| 27 | state->eof = 0; /* not at end of file */ |
| 28 | state->past = 0; /* have not read past end yet */ |
| 29 | state->how = LOOK; /* look for gzip header */ |
| 30 | } |
| 31 | else /* for writing ... */ |
| 32 | state->reset = 0; /* no deflateReset pending */ |
| 33 | state->seek = 0; /* no seek request pending */ |
| 34 | gz_error(state, Z_OK, NULL); /* clear error */ |
| 35 | state->x.pos = 0; /* no uncompressed data yet */ |
| 36 | state->strm.avail_in = 0; /* no input data yet */ |
| 37 | } |
| 38 | |
| 39 | /* Open a gzip file either by name or file descriptor. */ |
| 40 | static gzFile gz_open(const void *path, int fd, const char *mode) { |
| 41 | gz_state *state; |
| 42 | size_t len; |
| 43 | int oflag; |
| 44 | #ifdef O_CLOEXEC |
| 45 | int cloexec = 0; |
| 46 | #endif |
| 47 | #ifdef O_EXCL |
| 48 | int exclusive = 0; |
| 49 | #endif |
| 50 | |
| 51 | /* check input */ |
| 52 | if (path == NULL) |
| 53 | return NULL; |
| 54 | |
| 55 | /* allocate gzFile structure to return */ |
| 56 | state = (gz_state *)malloc(sizeof(gz_state)); |
| 57 | if (state == NULL) |
| 58 | return NULL; |
| 59 | state->size = 0; /* no buffers allocated yet */ |
| 60 | state->want = GZBUFSIZE; /* requested buffer size */ |
| 61 | state->msg = NULL; /* no error message yet */ |
| 62 | |
| 63 | /* interpret mode */ |
| 64 | state->mode = GZ_NONE; |
| 65 | state->level = Z_DEFAULT_COMPRESSION; |
| 66 | state->strategy = Z_DEFAULT_STRATEGY; |
| 67 | state->direct = 0; |
| 68 | while (*mode) { |
| 69 | if (*mode >= '0' && *mode <= '9') { |
| 70 | state->level = *mode - '0'; |
| 71 | } else { |
| 72 | switch (*mode) { |
| 73 | case 'r': |
| 74 | state->mode = GZ_READ; |
| 75 | break; |
| 76 | #ifndef NO_GZCOMPRESS |
| 77 | case 'w': |
| 78 | state->mode = GZ_WRITE; |
| 79 | break; |
| 80 | case 'a': |
| 81 | state->mode = GZ_APPEND; |
| 82 | break; |
| 83 | #endif |
| 84 | case '+': /* can't read and write at the same time */ |
| 85 | free(state); |
| 86 | return NULL; |
| 87 | case 'b': /* ignore -- will request binary anyway */ |
| 88 | break; |
| 89 | #ifdef O_CLOEXEC |
| 90 | case 'e': |
| 91 | cloexec = 1; |
| 92 | break; |
| 93 | #endif |
| 94 | #ifdef O_EXCL |
| 95 | case 'x': |
| 96 | exclusive = 1; |
| 97 | break; |
| 98 | #endif |
| 99 | case 'f': |
| 100 | state->strategy = Z_FILTERED; |
| 101 | break; |
| 102 | case 'h': |
| 103 | state->strategy = Z_HUFFMAN_ONLY; |
| 104 | break; |
| 105 | case 'R': |
| 106 | state->strategy = Z_RLE; |
| 107 | break; |
| 108 | case 'F': |
| 109 | state->strategy = Z_FIXED; |
| 110 | break; |
| 111 | case 'T': |
| 112 | state->direct = 1; |
| 113 | break; |
| 114 | default: /* could consider as an error, but just ignore */ |
| 115 | {} |
| 116 | } |
| 117 | } |
| 118 | mode++; |
| 119 | } |
| 120 | |
| 121 | /* must provide an "r", "w", or "a" */ |
| 122 | if (state->mode == GZ_NONE) { |
| 123 | free(state); |
| 124 | return NULL; |
| 125 | } |
| 126 | |
| 127 | /* can't force transparent read */ |
| 128 | if (state->mode == GZ_READ) { |
| 129 | if (state->direct) { |
| 130 | free(state); |
| 131 | return NULL; |
| 132 | } |
| 133 | state->direct = 1; /* for empty file */ |
| 134 | } |
| 135 | |
| 136 | /* save the path name for error messages */ |
| 137 | #ifdef WIDECHAR |
| 138 | if (fd == -2) { |
| 139 | len = wcstombs(NULL, path, 0); |
| 140 | if (len == (size_t)-1) |
| 141 | len = 0; |
| 142 | } else |
| 143 | #endif |
| 144 | len = strlen((const char *)path); |
| 145 | state->path = (char *)malloc(len + 1); |
| 146 | if (state->path == NULL) { |
| 147 | free(state); |
| 148 | return NULL; |
| 149 | } |
| 150 | #ifdef WIDECHAR |
| 151 | if (fd == -2) |
| 152 | if (len) { |
| 153 | wcstombs(state->path, path, len + 1); |
| 154 | } else { |
| 155 | *(state->path) = 0; |
| 156 | } |
| 157 | else |
| 158 | #endif |
| 159 | (void)snprintf(state->path, len + 1, "%s" , (const char *)path); |
| 160 | |
| 161 | /* compute the flags for open() */ |
| 162 | oflag = |
| 163 | #ifdef O_LARGEFILE |
| 164 | O_LARGEFILE | |
| 165 | #endif |
| 166 | #ifdef O_BINARY |
| 167 | O_BINARY | |
| 168 | #endif |
| 169 | #ifdef O_CLOEXEC |
| 170 | (cloexec ? O_CLOEXEC : 0) | |
| 171 | #endif |
| 172 | (state->mode == GZ_READ ? |
| 173 | O_RDONLY : |
| 174 | (O_WRONLY | O_CREAT | |
| 175 | #ifdef O_EXCL |
| 176 | (exclusive ? O_EXCL : 0) | |
| 177 | #endif |
| 178 | (state->mode == GZ_WRITE ? |
| 179 | O_TRUNC : |
| 180 | O_APPEND))); |
| 181 | |
| 182 | /* open the file with the appropriate flags (or just use fd) */ |
| 183 | state->fd = fd > -1 ? fd : ( |
| 184 | #if defined(WIN32) || defined(__MINGW__) |
| 185 | fd == -2 ? _wopen(path, oflag, 0666) : |
| 186 | #elif __CYGWIN__ |
| 187 | fd == -2 ? open(state->path, oflag, 0666) : |
| 188 | #endif |
| 189 | open((const char *)path, oflag, 0666)); |
| 190 | if (state->fd == -1) { |
| 191 | free(state->path); |
| 192 | free(state); |
| 193 | return NULL; |
| 194 | } |
| 195 | if (state->mode == GZ_APPEND) { |
| 196 | LSEEK(state->fd, 0, SEEK_END); /* so gzoffset() is correct */ |
| 197 | state->mode = GZ_WRITE; /* simplify later checks */ |
| 198 | } |
| 199 | |
| 200 | /* save the current position for rewinding (only if reading) */ |
| 201 | if (state->mode == GZ_READ) { |
| 202 | state->start = LSEEK(state->fd, 0, SEEK_CUR); |
| 203 | if (state->start == -1) state->start = 0; |
| 204 | } |
| 205 | |
| 206 | /* initialize stream */ |
| 207 | gz_reset(state); |
| 208 | |
| 209 | /* return stream */ |
| 210 | return (gzFile)state; |
| 211 | } |
| 212 | |
| 213 | /* -- see zlib.h -- */ |
| 214 | gzFile ZEXPORT PREFIX(gzopen)(const char *path, const char *mode) { |
| 215 | return gz_open(path, -1, mode); |
| 216 | } |
| 217 | |
| 218 | /* -- see zlib.h -- */ |
| 219 | gzFile ZEXPORT PREFIX(gzopen64)(const char *path, const char *mode) { |
| 220 | return gz_open(path, -1, mode); |
| 221 | } |
| 222 | |
| 223 | /* -- see zlib.h -- */ |
| 224 | gzFile ZEXPORT PREFIX(gzdopen)(int fd, const char *mode) { |
| 225 | char *path; /* identifier for error messages */ |
| 226 | gzFile gz; |
| 227 | |
| 228 | if (fd == -1 || (path = (char *)malloc(7 + 3 * sizeof(int))) == NULL) |
| 229 | return NULL; |
| 230 | (void)snprintf(path, 7 + 3 * sizeof(int), "<fd:%d>" , fd); /* for debugging */ |
| 231 | gz = gz_open(path, fd, mode); |
| 232 | free(path); |
| 233 | return gz; |
| 234 | } |
| 235 | |
| 236 | /* -- see zlib.h -- */ |
| 237 | #ifdef WIDECHAR |
| 238 | gzFile ZEXPORT PREFIX(gzopen_w)(const wchar_t *path, const char *mode) { |
| 239 | return gz_open(path, -2, mode); |
| 240 | } |
| 241 | #endif |
| 242 | |
| 243 | /* -- see zlib.h -- */ |
| 244 | int ZEXPORT PREFIX(gzbuffer)(gzFile file, unsigned size) { |
| 245 | gz_state *state; |
| 246 | |
| 247 | /* get internal structure and check integrity */ |
| 248 | if (file == NULL) |
| 249 | return -1; |
| 250 | state = (gz_state *)file; |
| 251 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 252 | return -1; |
| 253 | |
| 254 | /* make sure we haven't already allocated memory */ |
| 255 | if (state->size != 0) |
| 256 | return -1; |
| 257 | |
| 258 | /* check and set requested size */ |
| 259 | if ((size << 1) < size) |
| 260 | return -1; /* need to be able to double it */ |
| 261 | if (size < 2) |
| 262 | size = 2; /* need two bytes to check magic header */ |
| 263 | state->want = size; |
| 264 | return 0; |
| 265 | } |
| 266 | |
| 267 | /* -- see zlib.h -- */ |
| 268 | int ZEXPORT PREFIX(gzrewind)(gzFile file) { |
| 269 | gz_state *state; |
| 270 | |
| 271 | /* get internal structure */ |
| 272 | if (file == NULL) |
| 273 | return -1; |
| 274 | state = (gz_state *)file; |
| 275 | |
| 276 | /* check that we're reading and that there's no error */ |
| 277 | if (state->mode != GZ_READ || (state->err != Z_OK && state->err != Z_BUF_ERROR)) |
| 278 | return -1; |
| 279 | |
| 280 | /* back up and start over */ |
| 281 | if (LSEEK(state->fd, state->start, SEEK_SET) == -1) |
| 282 | return -1; |
| 283 | gz_reset(state); |
| 284 | return 0; |
| 285 | } |
| 286 | |
| 287 | /* -- see zlib.h -- */ |
| 288 | z_off64_t ZEXPORT PREFIX(gzseek64)(gzFile file, z_off64_t offset, int whence) { |
| 289 | unsigned n; |
| 290 | z_off64_t ret; |
| 291 | gz_state *state; |
| 292 | |
| 293 | /* get internal structure and check integrity */ |
| 294 | if (file == NULL) |
| 295 | return -1; |
| 296 | state = (gz_state *)file; |
| 297 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 298 | return -1; |
| 299 | |
| 300 | /* check that there's no error */ |
| 301 | if (state->err != Z_OK && state->err != Z_BUF_ERROR) |
| 302 | return -1; |
| 303 | |
| 304 | /* can only seek from start or relative to current position */ |
| 305 | if (whence != SEEK_SET && whence != SEEK_CUR) |
| 306 | return -1; |
| 307 | |
| 308 | /* normalize offset to a SEEK_CUR specification */ |
| 309 | if (whence == SEEK_SET) |
| 310 | offset -= state->x.pos; |
| 311 | else if (state->seek) |
| 312 | offset += state->skip; |
| 313 | state->seek = 0; |
| 314 | |
| 315 | /* if within raw area while reading, just go there */ |
| 316 | if (state->mode == GZ_READ && state->how == COPY && state->x.pos + offset >= 0) { |
| 317 | ret = LSEEK(state->fd, offset - (z_off64_t)state->x.have, SEEK_CUR); |
| 318 | if (ret == -1) |
| 319 | return -1; |
| 320 | state->x.have = 0; |
| 321 | state->eof = 0; |
| 322 | state->past = 0; |
| 323 | state->seek = 0; |
| 324 | gz_error(state, Z_OK, NULL); |
| 325 | state->strm.avail_in = 0; |
| 326 | state->x.pos += offset; |
| 327 | return state->x.pos; |
| 328 | } |
| 329 | |
| 330 | /* calculate skip amount, rewinding if needed for back seek when reading */ |
| 331 | if (offset < 0) { |
| 332 | if (state->mode != GZ_READ) /* writing -- can't go backwards */ |
| 333 | return -1; |
| 334 | offset += state->x.pos; |
| 335 | if (offset < 0) /* before start of file! */ |
| 336 | return -1; |
| 337 | if (PREFIX(gzrewind)(file) == -1) /* rewind, then skip to offset */ |
| 338 | return -1; |
| 339 | } |
| 340 | |
| 341 | /* if reading, skip what's in output buffer (one less gzgetc() check) */ |
| 342 | if (state->mode == GZ_READ) { |
| 343 | n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > offset ? (unsigned)offset : state->x.have; |
| 344 | state->x.have -= n; |
| 345 | state->x.next += n; |
| 346 | state->x.pos += n; |
| 347 | offset -= n; |
| 348 | } |
| 349 | |
| 350 | /* request skip (if not zero) */ |
| 351 | if (offset) { |
| 352 | state->seek = 1; |
| 353 | state->skip = offset; |
| 354 | } |
| 355 | return state->x.pos + offset; |
| 356 | } |
| 357 | |
| 358 | /* -- see zlib.h -- */ |
| 359 | z_off_t ZEXPORT PREFIX(gzseek)(gzFile file, z_off_t offset, int whence) { |
| 360 | z_off64_t ret; |
| 361 | |
| 362 | ret = PREFIX(gzseek64)(file, (z_off64_t)offset, whence); |
| 363 | return ret == (z_off_t)ret ? (z_off_t)ret : -1; |
| 364 | } |
| 365 | |
| 366 | /* -- see zlib.h -- */ |
| 367 | z_off64_t ZEXPORT PREFIX(gztell64)(gzFile file) { |
| 368 | gz_state *state; |
| 369 | |
| 370 | /* get internal structure and check integrity */ |
| 371 | if (file == NULL) |
| 372 | return -1; |
| 373 | state = (gz_state *)file; |
| 374 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 375 | return -1; |
| 376 | |
| 377 | /* return position */ |
| 378 | return state->x.pos + (state->seek ? state->skip : 0); |
| 379 | } |
| 380 | |
| 381 | /* -- see zlib.h -- */ |
| 382 | z_off_t ZEXPORT PREFIX(gztell)(gzFile file) { |
| 383 | z_off64_t ret; |
| 384 | |
| 385 | ret = PREFIX(gztell64)(file); |
| 386 | return ret == (z_off_t)ret ? (z_off_t)ret : -1; |
| 387 | } |
| 388 | |
| 389 | /* -- see zlib.h -- */ |
| 390 | z_off64_t ZEXPORT PREFIX(gzoffset64)(gzFile file) { |
| 391 | z_off64_t offset; |
| 392 | gz_state *state; |
| 393 | |
| 394 | /* get internal structure and check integrity */ |
| 395 | if (file == NULL) |
| 396 | return -1; |
| 397 | state = (gz_state *)file; |
| 398 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 399 | return -1; |
| 400 | |
| 401 | /* compute and return effective offset in file */ |
| 402 | offset = LSEEK(state->fd, 0, SEEK_CUR); |
| 403 | if (offset == -1) |
| 404 | return -1; |
| 405 | if (state->mode == GZ_READ) /* reading */ |
| 406 | offset -= state->strm.avail_in; /* don't count buffered input */ |
| 407 | return offset; |
| 408 | } |
| 409 | |
| 410 | /* -- see zlib.h -- */ |
| 411 | z_off_t ZEXPORT PREFIX(gzoffset)(gzFile file) { |
| 412 | z_off64_t ret; |
| 413 | |
| 414 | ret = PREFIX(gzoffset64)(file); |
| 415 | return ret == (z_off_t)ret ? (z_off_t)ret : -1; |
| 416 | } |
| 417 | |
| 418 | /* -- see zlib.h -- */ |
| 419 | int ZEXPORT PREFIX(gzeof)(gzFile file) { |
| 420 | gz_state *state; |
| 421 | |
| 422 | /* get internal structure and check integrity */ |
| 423 | if (file == NULL) |
| 424 | return 0; |
| 425 | state = (gz_state *)file; |
| 426 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 427 | return 0; |
| 428 | |
| 429 | /* return end-of-file state */ |
| 430 | return state->mode == GZ_READ ? state->past : 0; |
| 431 | } |
| 432 | |
| 433 | /* -- see zlib.h -- */ |
| 434 | const char * ZEXPORT PREFIX(gzerror)(gzFile file, int *errnum) { |
| 435 | gz_state *state; |
| 436 | |
| 437 | /* get internal structure and check integrity */ |
| 438 | if (file == NULL) |
| 439 | return NULL; |
| 440 | state = (gz_state *)file; |
| 441 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 442 | return NULL; |
| 443 | |
| 444 | /* return error information */ |
| 445 | if (errnum != NULL) |
| 446 | *errnum = state->err; |
| 447 | return state->err == Z_MEM_ERROR ? "out of memory" : (state->msg == NULL ? "" : state->msg); |
| 448 | } |
| 449 | |
| 450 | /* -- see zlib.h -- */ |
| 451 | void ZEXPORT PREFIX(gzclearerr)(gzFile file) { |
| 452 | gz_state *state; |
| 453 | |
| 454 | /* get internal structure and check integrity */ |
| 455 | if (file == NULL) |
| 456 | return; |
| 457 | state = (gz_state *)file; |
| 458 | if (state->mode != GZ_READ && state->mode != GZ_WRITE) |
| 459 | return; |
| 460 | |
| 461 | /* clear error and end-of-file */ |
| 462 | if (state->mode == GZ_READ) { |
| 463 | state->eof = 0; |
| 464 | state->past = 0; |
| 465 | } |
| 466 | gz_error(state, Z_OK, NULL); |
| 467 | } |
| 468 | |
| 469 | /* Create an error message in allocated memory and set state->err and |
| 470 | state->msg accordingly. Free any previous error message already there. Do |
| 471 | not try to free or allocate space if the error is Z_MEM_ERROR (out of |
| 472 | memory). Simply save the error message as a static string. If there is an |
| 473 | allocation failure constructing the error message, then convert the error to |
| 474 | out of memory. */ |
| 475 | void ZLIB_INTERNAL gz_error(gz_state *state, int err, const char *msg) { |
| 476 | /* free previously allocated message and clear */ |
| 477 | if (state->msg != NULL) { |
| 478 | if (state->err != Z_MEM_ERROR) |
| 479 | free(state->msg); |
| 480 | state->msg = NULL; |
| 481 | } |
| 482 | |
| 483 | /* if fatal, set state->x.have to 0 so that the gzgetc() macro fails */ |
| 484 | if (err != Z_OK && err != Z_BUF_ERROR) |
| 485 | state->x.have = 0; |
| 486 | |
| 487 | /* set error code, and if no message, then done */ |
| 488 | state->err = err; |
| 489 | if (msg == NULL) |
| 490 | return; |
| 491 | |
| 492 | /* for an out of memory error, return literal string when requested */ |
| 493 | if (err == Z_MEM_ERROR) |
| 494 | return; |
| 495 | |
| 496 | /* construct error message with path */ |
| 497 | if ((state->msg = (char *)malloc(strlen(state->path) + strlen(msg) + 3)) == NULL) { |
| 498 | state->err = Z_MEM_ERROR; |
| 499 | return; |
| 500 | } |
| 501 | (void)snprintf(state->msg, strlen(state->path) + strlen(msg) + 3, "%s%s%s" , state->path, ": " , msg); |
| 502 | } |
| 503 | |
| 504 | #ifndef INT_MAX |
| 505 | /* portably return maximum value for an int (when limits.h presumed not |
| 506 | available) -- we need to do this to cover cases where 2's complement not |
| 507 | used, since C standard permits 1's complement and sign-bit representations, |
| 508 | otherwise we could just use ((unsigned)-1) >> 1 */ |
| 509 | unsigned ZLIB_INTERNAL gz_intmax() { |
| 510 | unsigned p, q; |
| 511 | |
| 512 | p = 1; |
| 513 | do { |
| 514 | q = p; |
| 515 | p <<= 1; |
| 516 | p++; |
| 517 | } while (p > q); |
| 518 | return q >> 1; |
| 519 | } |
| 520 | #endif |
| 521 | |