| 1 | /* |
| 2 | * Copyright (c) 1995, 2018, Oracle and/or its affiliates. All rights reserved. |
| 3 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
| 4 | * |
| 5 | * This code is free software; you can redistribute it and/or modify it |
| 6 | * under the terms of the GNU General Public License version 2 only, as |
| 7 | * published by the Free Software Foundation. Oracle designates this |
| 8 | * particular file as subject to the "Classpath" exception as provided |
| 9 | * by Oracle in the LICENSE file that accompanied this code. |
| 10 | * |
| 11 | * This code is distributed in the hope that it will be useful, but WITHOUT |
| 12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 14 | * version 2 for more details (a copy is included in the LICENSE file that |
| 15 | * accompanied this code). |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License version |
| 18 | * 2 along with this work; if not, write to the Free Software Foundation, |
| 19 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
| 20 | * |
| 21 | * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
| 22 | * or visit www.oracle.com if you need additional information or have any |
| 23 | * questions. |
| 24 | */ |
| 25 | |
| 26 | /* |
| 27 | * Support for reading ZIP/JAR files. |
| 28 | */ |
| 29 | |
| 30 | #include <stdio.h> |
| 31 | #include <stdlib.h> |
| 32 | #include <stddef.h> |
| 33 | #include <string.h> |
| 34 | #include <fcntl.h> |
| 35 | #include <limits.h> |
| 36 | #include <time.h> |
| 37 | #include <ctype.h> |
| 38 | #include <assert.h> |
| 39 | |
| 40 | #include "jni.h" |
| 41 | #include "jni_util.h" |
| 42 | #include "jlong.h" |
| 43 | #include "jvm.h" |
| 44 | #include "io_util.h" |
| 45 | #include "io_util_md.h" |
| 46 | #include "zip_util.h" |
| 47 | #include <zlib.h> |
| 48 | |
| 49 | #ifdef _ALLBSD_SOURCE |
| 50 | #define off64_t off_t |
| 51 | #define mmap64 mmap |
| 52 | #endif |
| 53 | |
| 54 | /* USE_MMAP means mmap the CEN & ENDHDR part of the zip file. */ |
| 55 | #ifdef USE_MMAP |
| 56 | #include <sys/mman.h> |
| 57 | #endif |
| 58 | |
| 59 | #define MAXREFS 0xFFFF /* max number of open zip file references */ |
| 60 | |
| 61 | #define MCREATE() JVM_RawMonitorCreate() |
| 62 | #define MLOCK(lock) JVM_RawMonitorEnter(lock) |
| 63 | #define MUNLOCK(lock) JVM_RawMonitorExit(lock) |
| 64 | #define MDESTROY(lock) JVM_RawMonitorDestroy(lock) |
| 65 | |
| 66 | #define CENSIZE(cen) (CENHDR + CENNAM(cen) + CENEXT(cen) + CENCOM(cen)) |
| 67 | |
| 68 | static jzfile *zfiles = 0; /* currently open zip files */ |
| 69 | static void *zfiles_lock = 0; |
| 70 | |
| 71 | static void freeCEN(jzfile *); |
| 72 | |
| 73 | #ifndef PATH_MAX |
| 74 | #define PATH_MAX 1024 |
| 75 | #endif |
| 76 | |
| 77 | static jint INITIAL_META_COUNT = 2; /* initial number of entries in meta name array */ |
| 78 | |
| 79 | /* |
| 80 | * Declare library specific JNI_Onload entry if static build |
| 81 | */ |
| 82 | #ifdef STATIC_BUILD |
| 83 | DEF_STATIC_JNI_OnLoad |
| 84 | #endif |
| 85 | |
| 86 | /* |
| 87 | * The ZFILE_* functions exist to provide some platform-independence with |
| 88 | * respect to file access needs. |
| 89 | */ |
| 90 | |
| 91 | /* |
| 92 | * Opens the named file for reading, returning a ZFILE. |
| 93 | * |
| 94 | * Compare this with winFileHandleOpen in windows/native/java/io/io_util_md.c. |
| 95 | * This function does not take JNIEnv* and uses CreateFile (instead of |
| 96 | * CreateFileW). The expectation is that this function will be called only |
| 97 | * from ZIP_Open_Generic, which in turn is used by the JVM, where we do not |
| 98 | * need to concern ourselves with wide chars. |
| 99 | */ |
| 100 | static ZFILE |
| 101 | ZFILE_Open(const char *fname, int flags) { |
| 102 | #ifdef WIN32 |
| 103 | WCHAR *wfname, *wprefixed_fname; |
| 104 | size_t converted_chars, fname_length; |
| 105 | jlong fhandle; |
| 106 | const DWORD access = |
| 107 | (flags & O_RDWR) ? (GENERIC_WRITE | GENERIC_READ) : |
| 108 | (flags & O_WRONLY) ? GENERIC_WRITE : |
| 109 | GENERIC_READ; |
| 110 | const DWORD sharing = |
| 111 | FILE_SHARE_READ | FILE_SHARE_WRITE; |
| 112 | const DWORD disposition = |
| 113 | /* Note: O_TRUNC overrides O_CREAT */ |
| 114 | (flags & O_TRUNC) ? CREATE_ALWAYS : |
| 115 | (flags & O_CREAT) ? OPEN_ALWAYS : |
| 116 | OPEN_EXISTING; |
| 117 | const DWORD maybeWriteThrough = |
| 118 | (flags & (O_SYNC | O_DSYNC)) ? |
| 119 | FILE_FLAG_WRITE_THROUGH : |
| 120 | FILE_ATTRIBUTE_NORMAL; |
| 121 | const DWORD maybeDeleteOnClose = |
| 122 | (flags & O_TEMPORARY) ? |
| 123 | FILE_FLAG_DELETE_ON_CLOSE : |
| 124 | FILE_ATTRIBUTE_NORMAL; |
| 125 | const DWORD flagsAndAttributes = maybeWriteThrough | maybeDeleteOnClose; |
| 126 | |
| 127 | fname_length = strlen(fname); |
| 128 | if (fname_length < MAX_PATH) { |
| 129 | return (jlong)CreateFile( |
| 130 | fname, /* path name in multibyte char */ |
| 131 | access, /* Read and/or write permission */ |
| 132 | sharing, /* File sharing flags */ |
| 133 | NULL, /* Security attributes */ |
| 134 | disposition, /* creation disposition */ |
| 135 | flagsAndAttributes, /* flags and attributes */ |
| 136 | NULL); |
| 137 | } else { |
| 138 | if ((wfname = (WCHAR*)malloc((fname_length + 1) * sizeof(WCHAR))) == NULL) |
| 139 | return (jlong)INVALID_HANDLE_VALUE; |
| 140 | |
| 141 | if (mbstowcs_s(&converted_chars, wfname, fname_length + 1, fname, fname_length) != 0) { |
| 142 | free(wfname); |
| 143 | return (jlong)INVALID_HANDLE_VALUE; |
| 144 | } |
| 145 | wprefixed_fname = getPrefixed(wfname, (int)fname_length); |
| 146 | fhandle = (jlong)CreateFileW( |
| 147 | wprefixed_fname, /* Wide char path name */ |
| 148 | access, /* Read and/or write permission */ |
| 149 | sharing, /* File sharing flags */ |
| 150 | NULL, /* Security attributes */ |
| 151 | disposition, /* creation disposition */ |
| 152 | flagsAndAttributes, /* flags and attributes */ |
| 153 | NULL); |
| 154 | free(wfname); |
| 155 | free(wprefixed_fname); |
| 156 | return fhandle; |
| 157 | } |
| 158 | #else |
| 159 | return open(fname, flags, 0); |
| 160 | #endif |
| 161 | } |
| 162 | |
| 163 | /* |
| 164 | * The io_util_md.h files do not provide IO_CLOSE, hence we use platform |
| 165 | * specifics. |
| 166 | */ |
| 167 | static void |
| 168 | ZFILE_Close(ZFILE zfd) { |
| 169 | #ifdef WIN32 |
| 170 | CloseHandle((HANDLE) zfd); |
| 171 | #else |
| 172 | close(zfd); |
| 173 | #endif |
| 174 | } |
| 175 | |
| 176 | static int |
| 177 | ZFILE_read(ZFILE zfd, char *buf, jint nbytes) { |
| 178 | #ifdef WIN32 |
| 179 | return (int) IO_Read(zfd, buf, nbytes); |
| 180 | #else |
| 181 | return read(zfd, buf, nbytes); |
| 182 | #endif |
| 183 | } |
| 184 | |
| 185 | /* |
| 186 | * Initialize zip file support. Return 0 if successful otherwise -1 |
| 187 | * if could not be initialized. |
| 188 | */ |
| 189 | static jint |
| 190 | InitializeZip() |
| 191 | { |
| 192 | static jboolean inited = JNI_FALSE; |
| 193 | |
| 194 | // Initialize errno to 0. It may be set later (e.g. during memory |
| 195 | // allocation) but we can disregard previous values. |
| 196 | errno = 0; |
| 197 | |
| 198 | if (inited) |
| 199 | return 0; |
| 200 | zfiles_lock = MCREATE(); |
| 201 | if (zfiles_lock == 0) { |
| 202 | return -1; |
| 203 | } |
| 204 | inited = JNI_TRUE; |
| 205 | |
| 206 | return 0; |
| 207 | } |
| 208 | |
| 209 | /* |
| 210 | * Reads len bytes of data into buf. |
| 211 | * Returns 0 if all bytes could be read, otherwise returns -1. |
| 212 | */ |
| 213 | static int |
| 214 | readFully(ZFILE zfd, void *buf, jlong len) { |
| 215 | char *bp = (char *) buf; |
| 216 | |
| 217 | while (len > 0) { |
| 218 | jlong limit = ((((jlong) 1) << 31) - 1); |
| 219 | jint count = (len < limit) ? |
| 220 | (jint) len : |
| 221 | (jint) limit; |
| 222 | jint n = ZFILE_read(zfd, bp, count); |
| 223 | if (n > 0) { |
| 224 | bp += n; |
| 225 | len -= n; |
| 226 | } else if (n == -1 && errno == EINTR) { |
| 227 | /* Retry after EINTR (interrupted by signal). */ |
| 228 | continue; |
| 229 | } else { /* EOF or IO error */ |
| 230 | return -1; |
| 231 | } |
| 232 | } |
| 233 | return 0; |
| 234 | } |
| 235 | |
| 236 | /* |
| 237 | * Reads len bytes of data from the specified offset into buf. |
| 238 | * Returns 0 if all bytes could be read, otherwise returns -1. |
| 239 | */ |
| 240 | static int |
| 241 | readFullyAt(ZFILE zfd, void *buf, jlong len, jlong offset) |
| 242 | { |
| 243 | if (IO_Lseek(zfd, offset, SEEK_SET) == -1) { |
| 244 | return -1; /* lseek failure. */ |
| 245 | } |
| 246 | |
| 247 | return readFully(zfd, buf, len); |
| 248 | } |
| 249 | |
| 250 | /* |
| 251 | * Allocates a new zip file object for the specified file name. |
| 252 | * Returns the zip file object or NULL if not enough memory. |
| 253 | */ |
| 254 | static jzfile * |
| 255 | allocZip(const char *name) |
| 256 | { |
| 257 | jzfile *zip; |
| 258 | if (((zip = calloc(1, sizeof(jzfile))) != NULL) && |
| 259 | ((zip->name = strdup(name)) != NULL) && |
| 260 | ((zip->lock = MCREATE()) != NULL)) { |
| 261 | zip->zfd = -1; |
| 262 | return zip; |
| 263 | } |
| 264 | |
| 265 | if (zip != NULL) { |
| 266 | free(zip->name); |
| 267 | free(zip); |
| 268 | } |
| 269 | return NULL; |
| 270 | } |
| 271 | |
| 272 | /* |
| 273 | * Frees all native resources owned by the specified zip file object. |
| 274 | */ |
| 275 | static void |
| 276 | freeZip(jzfile *zip) |
| 277 | { |
| 278 | /* First free any cached jzentry */ |
| 279 | ZIP_FreeEntry(zip,0); |
| 280 | if (zip->lock != NULL) MDESTROY(zip->lock); |
| 281 | free(zip->name); |
| 282 | freeCEN(zip); |
| 283 | |
| 284 | #ifdef USE_MMAP |
| 285 | if (zip->usemmap) { |
| 286 | if (zip->maddr != NULL) |
| 287 | munmap((char *)zip->maddr, zip->mlen); |
| 288 | } else |
| 289 | #endif |
| 290 | { |
| 291 | free(zip->cencache.data); |
| 292 | } |
| 293 | if (zip->comment != NULL) |
| 294 | free(zip->comment); |
| 295 | if (zip->zfd != -1) ZFILE_Close(zip->zfd); |
| 296 | free(zip); |
| 297 | } |
| 298 | |
| 299 | /* The END header is followed by a variable length comment of size < 64k. */ |
| 300 | static const jlong END_MAXLEN = 0xFFFF + ENDHDR; |
| 301 | |
| 302 | #define READBLOCKSZ 128 |
| 303 | |
| 304 | static jboolean verifyEND(jzfile *zip, jlong endpos, char *endbuf) { |
| 305 | /* ENDSIG matched, however the size of file comment in it does not |
| 306 | match the real size. One "common" cause for this problem is some |
| 307 | "extra" bytes are padded at the end of the zipfile. |
| 308 | Let's do some extra verification, we don't care about the performance |
| 309 | in this situation. |
| 310 | */ |
| 311 | jlong cenpos = endpos - ENDSIZ(endbuf); |
| 312 | jlong locpos = cenpos - ENDOFF(endbuf); |
| 313 | char buf[4]; |
| 314 | return (cenpos >= 0 && |
| 315 | locpos >= 0 && |
| 316 | readFullyAt(zip->zfd, buf, sizeof(buf), cenpos) != -1 && |
| 317 | CENSIG_AT(buf) && |
| 318 | readFullyAt(zip->zfd, buf, sizeof(buf), locpos) != -1 && |
| 319 | LOCSIG_AT(buf)); |
| 320 | } |
| 321 | |
| 322 | /* |
| 323 | * Searches for end of central directory (END) header. The contents of |
| 324 | * the END header will be read and placed in endbuf. Returns the file |
| 325 | * position of the END header, otherwise returns -1 if the END header |
| 326 | * was not found or an error occurred. |
| 327 | */ |
| 328 | static jlong |
| 329 | findEND(jzfile *zip, void *endbuf) |
| 330 | { |
| 331 | char buf[READBLOCKSZ]; |
| 332 | jlong pos; |
| 333 | const jlong len = zip->len; |
| 334 | const ZFILE zfd = zip->zfd; |
| 335 | const jlong minHDR = len - END_MAXLEN > 0 ? len - END_MAXLEN : 0; |
| 336 | const jlong minPos = minHDR - (sizeof(buf)-ENDHDR); |
| 337 | jint clen; |
| 338 | |
| 339 | for (pos = len - sizeof(buf); pos >= minPos; pos -= (sizeof(buf)-ENDHDR)) { |
| 340 | |
| 341 | int i; |
| 342 | jlong off = 0; |
| 343 | if (pos < 0) { |
| 344 | /* Pretend there are some NUL bytes before start of file */ |
| 345 | off = -pos; |
| 346 | memset(buf, '\0', (size_t)off); |
| 347 | } |
| 348 | |
| 349 | if (readFullyAt(zfd, buf + off, sizeof(buf) - off, |
| 350 | pos + off) == -1) { |
| 351 | return -1; /* System error */ |
| 352 | } |
| 353 | |
| 354 | /* Now scan the block backwards for END header signature */ |
| 355 | for (i = sizeof(buf) - ENDHDR; i >= 0; i--) { |
| 356 | if (buf[i+0] == 'P' && |
| 357 | buf[i+1] == 'K' && |
| 358 | buf[i+2] == '\005' && |
| 359 | buf[i+3] == '\006' && |
| 360 | ((pos + i + ENDHDR + ENDCOM(buf + i) == len) |
| 361 | || verifyEND(zip, pos + i, buf + i))) { |
| 362 | /* Found END header */ |
| 363 | memcpy(endbuf, buf + i, ENDHDR); |
| 364 | |
| 365 | clen = ENDCOM(endbuf); |
| 366 | if (clen != 0) { |
| 367 | zip->comment = malloc(clen + 1); |
| 368 | if (zip->comment == NULL) { |
| 369 | return -1; |
| 370 | } |
| 371 | if (readFullyAt(zfd, zip->comment, clen, pos + i + ENDHDR) |
| 372 | == -1) { |
| 373 | free(zip->comment); |
| 374 | zip->comment = NULL; |
| 375 | return -1; |
| 376 | } |
| 377 | zip->comment[clen] = '\0'; |
| 378 | zip->clen = clen; |
| 379 | } |
| 380 | return pos + i; |
| 381 | } |
| 382 | } |
| 383 | } |
| 384 | |
| 385 | return -1; /* END header not found */ |
| 386 | } |
| 387 | |
| 388 | /* |
| 389 | * Searches for the ZIP64 end of central directory (END) header. The |
| 390 | * contents of the ZIP64 END header will be read and placed in end64buf. |
| 391 | * Returns the file position of the ZIP64 END header, otherwise returns |
| 392 | * -1 if the END header was not found or an error occurred. |
| 393 | * |
| 394 | * The ZIP format specifies the "position" of each related record as |
| 395 | * ... |
| 396 | * [central directory] |
| 397 | * [zip64 end of central directory record] |
| 398 | * [zip64 end of central directory locator] |
| 399 | * [end of central directory record] |
| 400 | * |
| 401 | * The offset of zip64 end locator can be calculated from endpos as |
| 402 | * "endpos - ZIP64_LOCHDR". |
| 403 | * The "offset" of zip64 end record is stored in zip64 end locator. |
| 404 | */ |
| 405 | static jlong |
| 406 | findEND64(jzfile *zip, void *end64buf, jlong endpos) |
| 407 | { |
| 408 | char loc64[ZIP64_LOCHDR]; |
| 409 | jlong end64pos; |
| 410 | if (readFullyAt(zip->zfd, loc64, ZIP64_LOCHDR, endpos - ZIP64_LOCHDR) == -1) { |
| 411 | return -1; // end64 locator not found |
| 412 | } |
| 413 | end64pos = ZIP64_LOCOFF(loc64); |
| 414 | if (readFullyAt(zip->zfd, end64buf, ZIP64_ENDHDR, end64pos) == -1) { |
| 415 | return -1; // end64 record not found |
| 416 | } |
| 417 | return end64pos; |
| 418 | } |
| 419 | |
| 420 | /* |
| 421 | * Returns a hash code value for a C-style NUL-terminated string. |
| 422 | */ |
| 423 | static unsigned int |
| 424 | hash(const char *s) |
| 425 | { |
| 426 | int h = 0; |
| 427 | while (*s != '\0') |
| 428 | h = 31*h + *s++; |
| 429 | return h; |
| 430 | } |
| 431 | |
| 432 | /* |
| 433 | * Returns a hash code value for a string of a specified length. |
| 434 | */ |
| 435 | static unsigned int |
| 436 | hashN(const char *s, int length) |
| 437 | { |
| 438 | int h = 0; |
| 439 | while (length-- > 0) |
| 440 | h = 31*h + *s++; |
| 441 | return h; |
| 442 | } |
| 443 | |
| 444 | static unsigned int |
| 445 | hash_append(unsigned int hash, char c) |
| 446 | { |
| 447 | return ((int)hash)*31 + c; |
| 448 | } |
| 449 | |
| 450 | /* |
| 451 | * Returns true if the specified entry's name begins with the string |
| 452 | * "META-INF/" irrespective of case. |
| 453 | */ |
| 454 | static int |
| 455 | isMetaName(const char *name, int length) |
| 456 | { |
| 457 | const char *s; |
| 458 | if (length < (int)sizeof("META-INF/" ) - 1) |
| 459 | return 0; |
| 460 | for (s = "META-INF/" ; *s != '\0'; s++) { |
| 461 | char c = *name++; |
| 462 | // Avoid toupper; it's locale-dependent |
| 463 | if (c >= 'a' && c <= 'z') c += 'A' - 'a'; |
| 464 | if (*s != c) |
| 465 | return 0; |
| 466 | } |
| 467 | return 1; |
| 468 | } |
| 469 | |
| 470 | /* |
| 471 | * Increases the capacity of zip->metanames. |
| 472 | * Returns non-zero in case of allocation error. |
| 473 | */ |
| 474 | static int |
| 475 | growMetaNames(jzfile *zip) |
| 476 | { |
| 477 | jint i; |
| 478 | /* double the meta names array */ |
| 479 | const jint new_metacount = zip->metacount << 1; |
| 480 | zip->metanames = |
| 481 | realloc(zip->metanames, new_metacount * sizeof(zip->metanames[0])); |
| 482 | if (zip->metanames == NULL) return -1; |
| 483 | for (i = zip->metacount; i < new_metacount; i++) |
| 484 | zip->metanames[i] = NULL; |
| 485 | zip->metacurrent = zip->metacount; |
| 486 | zip->metacount = new_metacount; |
| 487 | return 0; |
| 488 | } |
| 489 | |
| 490 | /* |
| 491 | * Adds name to zip->metanames. |
| 492 | * Returns non-zero in case of allocation error. |
| 493 | */ |
| 494 | static int |
| 495 | addMetaName(jzfile *zip, const char *name, int length) |
| 496 | { |
| 497 | jint i; |
| 498 | if (zip->metanames == NULL) { |
| 499 | zip->metacount = INITIAL_META_COUNT; |
| 500 | zip->metanames = calloc(zip->metacount, sizeof(zip->metanames[0])); |
| 501 | if (zip->metanames == NULL) return -1; |
| 502 | zip->metacurrent = 0; |
| 503 | } |
| 504 | |
| 505 | i = zip->metacurrent; |
| 506 | |
| 507 | /* current meta name array isn't full yet. */ |
| 508 | if (i < zip->metacount) { |
| 509 | zip->metanames[i] = (char *) malloc(length+1); |
| 510 | if (zip->metanames[i] == NULL) return -1; |
| 511 | memcpy(zip->metanames[i], name, length); |
| 512 | zip->metanames[i][length] = '\0'; |
| 513 | zip->metacurrent++; |
| 514 | return 0; |
| 515 | } |
| 516 | |
| 517 | /* No free entries in zip->metanames? */ |
| 518 | if (growMetaNames(zip) != 0) return -1; |
| 519 | return addMetaName(zip, name, length); |
| 520 | } |
| 521 | |
| 522 | static void |
| 523 | freeMetaNames(jzfile *zip) |
| 524 | { |
| 525 | if (zip->metanames) { |
| 526 | jint i; |
| 527 | for (i = 0; i < zip->metacount; i++) |
| 528 | free(zip->metanames[i]); |
| 529 | free(zip->metanames); |
| 530 | zip->metanames = NULL; |
| 531 | } |
| 532 | } |
| 533 | |
| 534 | /* Free Zip data allocated by readCEN() */ |
| 535 | static void |
| 536 | freeCEN(jzfile *zip) |
| 537 | { |
| 538 | free(zip->entries); zip->entries = NULL; |
| 539 | free(zip->table); zip->table = NULL; |
| 540 | freeMetaNames(zip); |
| 541 | } |
| 542 | |
| 543 | /* |
| 544 | * Counts the number of CEN headers in a central directory extending |
| 545 | * from BEG to END. Might return a bogus answer if the zip file is |
| 546 | * corrupt, but will not crash. |
| 547 | */ |
| 548 | static jint |
| 549 | (unsigned char *beg, unsigned char *end) |
| 550 | { |
| 551 | jint count = 0; |
| 552 | ptrdiff_t i; |
| 553 | for (i = 0; i + CENHDR <= end - beg; i += CENSIZE(beg + i)) |
| 554 | count++; |
| 555 | return count; |
| 556 | } |
| 557 | |
| 558 | #define ZIP_FORMAT_ERROR(message) \ |
| 559 | if (1) { zip->msg = message; goto Catch; } else ((void)0) |
| 560 | |
| 561 | /* |
| 562 | * Reads zip file central directory. Returns the file position of first |
| 563 | * CEN header, otherwise returns -1 if an error occurred. If zip->msg != NULL |
| 564 | * then the error was a zip format error and zip->msg has the error text. |
| 565 | * Always pass in -1 for knownTotal; it's used for a recursive call. |
| 566 | */ |
| 567 | static jlong |
| 568 | readCEN(jzfile *zip, jint knownTotal) |
| 569 | { |
| 570 | /* Following are unsigned 32-bit */ |
| 571 | jlong endpos, end64pos, cenpos, cenlen, cenoff; |
| 572 | /* Following are unsigned 16-bit */ |
| 573 | jint total, tablelen, i, j; |
| 574 | unsigned char *cenbuf = NULL; |
| 575 | unsigned char *cenend; |
| 576 | unsigned char *cp; |
| 577 | #ifdef USE_MMAP |
| 578 | static jlong pagesize; |
| 579 | jlong offset; |
| 580 | #endif |
| 581 | unsigned char endbuf[ENDHDR]; |
| 582 | jint endhdrlen = ENDHDR; |
| 583 | jzcell *entries; |
| 584 | jint *table; |
| 585 | |
| 586 | /* Clear previous zip error */ |
| 587 | zip->msg = NULL; |
| 588 | /* Get position of END header */ |
| 589 | if ((endpos = findEND(zip, endbuf)) == -1) |
| 590 | return -1; /* no END header or system error */ |
| 591 | |
| 592 | if (endpos == 0) return 0; /* only END header present */ |
| 593 | |
| 594 | freeCEN(zip); |
| 595 | /* Get position and length of central directory */ |
| 596 | cenlen = ENDSIZ(endbuf); |
| 597 | cenoff = ENDOFF(endbuf); |
| 598 | total = ENDTOT(endbuf); |
| 599 | if (cenlen == ZIP64_MAGICVAL || cenoff == ZIP64_MAGICVAL || |
| 600 | total == ZIP64_MAGICCOUNT) { |
| 601 | unsigned char end64buf[ZIP64_ENDHDR]; |
| 602 | if ((end64pos = findEND64(zip, end64buf, endpos)) != -1) { |
| 603 | cenlen = ZIP64_ENDSIZ(end64buf); |
| 604 | cenoff = ZIP64_ENDOFF(end64buf); |
| 605 | total = (jint)ZIP64_ENDTOT(end64buf); |
| 606 | endpos = end64pos; |
| 607 | endhdrlen = ZIP64_ENDHDR; |
| 608 | } |
| 609 | } |
| 610 | |
| 611 | if (cenlen > endpos) { |
| 612 | ZIP_FORMAT_ERROR("invalid END header (bad central directory size)" ); |
| 613 | } |
| 614 | cenpos = endpos - cenlen; |
| 615 | |
| 616 | /* Get position of first local file (LOC) header, taking into |
| 617 | * account that there may be a stub prefixed to the zip file. */ |
| 618 | zip->locpos = cenpos - cenoff; |
| 619 | if (zip->locpos < 0) { |
| 620 | ZIP_FORMAT_ERROR("invalid END header (bad central directory offset)" ); |
| 621 | } |
| 622 | #ifdef USE_MMAP |
| 623 | if (zip->usemmap) { |
| 624 | /* On Solaris & Linux prior to JDK 6, we used to mmap the whole jar file to |
| 625 | * read the jar file contents. However, this greatly increased the perceived |
| 626 | * footprint numbers because the mmap'ed pages were adding into the totals shown |
| 627 | * by 'ps' and 'top'. We switched to mmaping only the central directory of jar |
| 628 | * file while calling 'read' to read the rest of jar file. Here are a list of |
| 629 | * reasons apart from above of why we are doing so: |
| 630 | * 1. Greatly reduces mmap overhead after startup complete; |
| 631 | * 2. Avoids dual path code maintainance; |
| 632 | * 3. Greatly reduces risk of address space (not virtual memory) exhaustion. |
| 633 | */ |
| 634 | if (pagesize == 0) { |
| 635 | pagesize = (jlong)sysconf(_SC_PAGESIZE); |
| 636 | if (pagesize == 0) goto Catch; |
| 637 | } |
| 638 | if (cenpos > pagesize) { |
| 639 | offset = cenpos & ~(pagesize - 1); |
| 640 | } else { |
| 641 | offset = 0; |
| 642 | } |
| 643 | /* When we are not calling recursively, knownTotal is -1. */ |
| 644 | if (knownTotal == -1) { |
| 645 | void* mappedAddr; |
| 646 | /* Mmap the CEN and END part only. We have to figure |
| 647 | out the page size in order to make offset to be multiples of |
| 648 | page size. |
| 649 | */ |
| 650 | zip->mlen = cenpos - offset + cenlen + endhdrlen; |
| 651 | zip->offset = offset; |
| 652 | mappedAddr = mmap64(0, zip->mlen, PROT_READ, MAP_SHARED, zip->zfd, (off64_t) offset); |
| 653 | zip->maddr = (mappedAddr == (void*) MAP_FAILED) ? NULL : |
| 654 | (unsigned char*)mappedAddr; |
| 655 | |
| 656 | if (zip->maddr == NULL) { |
| 657 | jio_fprintf(stderr, "mmap failed for CEN and END part of zip file\n" ); |
| 658 | goto Catch; |
| 659 | } |
| 660 | } |
| 661 | cenbuf = zip->maddr + cenpos - offset; |
| 662 | } else |
| 663 | #endif |
| 664 | { |
| 665 | if ((cenbuf = malloc((size_t) cenlen)) == NULL || |
| 666 | (readFullyAt(zip->zfd, cenbuf, cenlen, cenpos) == -1)) |
| 667 | goto Catch; |
| 668 | } |
| 669 | |
| 670 | cenend = cenbuf + cenlen; |
| 671 | |
| 672 | /* Initialize zip file data structures based on the total number |
| 673 | * of central directory entries as stored in ENDTOT. Since this |
| 674 | * is a 2-byte field, but we (and other zip implementations) |
| 675 | * support approx. 2**31 entries, we do not trust ENDTOT, but |
| 676 | * treat it only as a strong hint. When we call ourselves |
| 677 | * recursively, knownTotal will have the "true" value. |
| 678 | * |
| 679 | * Keep this path alive even with the Zip64 END support added, just |
| 680 | * for zip files that have more than 0xffff entries but don't have |
| 681 | * the Zip64 enabled. |
| 682 | */ |
| 683 | total = (knownTotal != -1) ? knownTotal : total; |
| 684 | entries = zip->entries = calloc(total, sizeof(entries[0])); |
| 685 | tablelen = zip->tablelen = ((total/2) | 1); // Odd -> fewer collisions |
| 686 | table = zip->table = malloc(tablelen * sizeof(table[0])); |
| 687 | /* According to ISO C it is perfectly legal for malloc to return zero |
| 688 | * if called with a zero argument. We check this for 'entries' but not |
| 689 | * for 'table' because 'tablelen' can't be zero (see computation above). */ |
| 690 | if ((entries == NULL && total != 0) || table == NULL) goto Catch; |
| 691 | for (j = 0; j < tablelen; j++) |
| 692 | table[j] = ZIP_ENDCHAIN; |
| 693 | |
| 694 | /* Iterate through the entries in the central directory */ |
| 695 | for (i = 0, cp = cenbuf; cp <= cenend - CENHDR; i++, cp += CENSIZE(cp)) { |
| 696 | /* Following are unsigned 16-bit */ |
| 697 | jint method, nlen; |
| 698 | unsigned int hsh; |
| 699 | |
| 700 | if (i >= total) { |
| 701 | /* This will only happen if the zip file has an incorrect |
| 702 | * ENDTOT field, which usually means it contains more than |
| 703 | * 65535 entries. */ |
| 704 | cenpos = readCEN(zip, countCENHeaders(cenbuf, cenend)); |
| 705 | goto Finally; |
| 706 | } |
| 707 | |
| 708 | method = CENHOW(cp); |
| 709 | nlen = CENNAM(cp); |
| 710 | |
| 711 | if (!CENSIG_AT(cp)) { |
| 712 | ZIP_FORMAT_ERROR("invalid CEN header (bad signature)" ); |
| 713 | } |
| 714 | if (CENFLG(cp) & 1) { |
| 715 | ZIP_FORMAT_ERROR("invalid CEN header (encrypted entry)" ); |
| 716 | } |
| 717 | if (method != STORED && method != DEFLATED) { |
| 718 | ZIP_FORMAT_ERROR("invalid CEN header (bad compression method)" ); |
| 719 | } |
| 720 | if (cp + CENHDR + nlen > cenend) { |
| 721 | ZIP_FORMAT_ERROR("invalid CEN header (bad header size)" ); |
| 722 | } |
| 723 | /* if the entry is metadata add it to our metadata names */ |
| 724 | if (isMetaName((char *)cp+CENHDR, nlen)) |
| 725 | if (addMetaName(zip, (char *)cp+CENHDR, nlen) != 0) |
| 726 | goto Catch; |
| 727 | |
| 728 | /* Record the CEN offset and the name hash in our hash cell. */ |
| 729 | entries[i].cenpos = cenpos + (cp - cenbuf); |
| 730 | entries[i].hash = hashN((char *)cp+CENHDR, nlen); |
| 731 | |
| 732 | /* Add the entry to the hash table */ |
| 733 | hsh = entries[i].hash % tablelen; |
| 734 | entries[i].next = table[hsh]; |
| 735 | table[hsh] = i; |
| 736 | } |
| 737 | if (cp != cenend) { |
| 738 | ZIP_FORMAT_ERROR("invalid CEN header (bad header size)" ); |
| 739 | } |
| 740 | zip->total = i; |
| 741 | goto Finally; |
| 742 | |
| 743 | Catch: |
| 744 | freeCEN(zip); |
| 745 | cenpos = -1; |
| 746 | |
| 747 | Finally: |
| 748 | #ifdef USE_MMAP |
| 749 | if (!zip->usemmap) |
| 750 | #endif |
| 751 | free(cenbuf); |
| 752 | |
| 753 | return cenpos; |
| 754 | } |
| 755 | |
| 756 | /* |
| 757 | * Opens a zip file with the specified mode. Returns the jzfile object |
| 758 | * or NULL if an error occurred. If a zip error occurred then *pmsg will |
| 759 | * be set to the error message text if pmsg != 0. Otherwise, *pmsg will be |
| 760 | * set to NULL. Caller is responsible to free the error message. |
| 761 | */ |
| 762 | jzfile * |
| 763 | ZIP_Open_Generic(const char *name, char **pmsg, int mode, jlong lastModified) |
| 764 | { |
| 765 | jzfile *zip = NULL; |
| 766 | |
| 767 | /* Clear zip error message */ |
| 768 | if (pmsg != NULL) { |
| 769 | *pmsg = NULL; |
| 770 | } |
| 771 | |
| 772 | zip = ZIP_Get_From_Cache(name, pmsg, lastModified); |
| 773 | |
| 774 | if (zip == NULL && pmsg != NULL && *pmsg == NULL) { |
| 775 | ZFILE zfd = ZFILE_Open(name, mode); |
| 776 | zip = ZIP_Put_In_Cache(name, zfd, pmsg, lastModified); |
| 777 | } |
| 778 | return zip; |
| 779 | } |
| 780 | |
| 781 | /* |
| 782 | * Returns the jzfile corresponding to the given file name from the cache of |
| 783 | * zip files, or NULL if the file is not in the cache. If the name is longer |
| 784 | * than PATH_MAX or a zip error occurred then *pmsg will be set to the error |
| 785 | * message text if pmsg != 0. Otherwise, *pmsg will be set to NULL. Caller |
| 786 | * is responsible to free the error message. |
| 787 | */ |
| 788 | jzfile * |
| 789 | ZIP_Get_From_Cache(const char *name, char **pmsg, jlong lastModified) |
| 790 | { |
| 791 | char buf[PATH_MAX]; |
| 792 | jzfile *zip; |
| 793 | |
| 794 | if (InitializeZip()) { |
| 795 | return NULL; |
| 796 | } |
| 797 | |
| 798 | /* Clear zip error message */ |
| 799 | if (pmsg != 0) { |
| 800 | *pmsg = NULL; |
| 801 | } |
| 802 | |
| 803 | if (strlen(name) >= PATH_MAX) { |
| 804 | if (pmsg) { |
| 805 | *pmsg = strdup("zip file name too long" ); |
| 806 | } |
| 807 | return NULL; |
| 808 | } |
| 809 | strcpy(buf, name); |
| 810 | JVM_NativePath(buf); |
| 811 | name = buf; |
| 812 | |
| 813 | MLOCK(zfiles_lock); |
| 814 | for (zip = zfiles; zip != NULL; zip = zip->next) { |
| 815 | if (strcmp(name, zip->name) == 0 |
| 816 | && (zip->lastModified == lastModified || zip->lastModified == 0) |
| 817 | && zip->refs < MAXREFS) { |
| 818 | zip->refs++; |
| 819 | break; |
| 820 | } |
| 821 | } |
| 822 | MUNLOCK(zfiles_lock); |
| 823 | return zip; |
| 824 | } |
| 825 | |
| 826 | /* |
| 827 | * Reads data from the given file descriptor to create a jzfile, puts the |
| 828 | * jzfile in a cache, and returns that jzfile. Returns NULL in case of error. |
| 829 | * If a zip error occurs, then *pmsg will be set to the error message text if |
| 830 | * pmsg != 0. Otherwise, *pmsg will be set to NULL. Caller is responsible to |
| 831 | * free the error message. |
| 832 | */ |
| 833 | |
| 834 | jzfile * |
| 835 | ZIP_Put_In_Cache(const char *name, ZFILE zfd, char **pmsg, jlong lastModified) |
| 836 | { |
| 837 | return ZIP_Put_In_Cache0(name, zfd, pmsg, lastModified, JNI_TRUE); |
| 838 | } |
| 839 | |
| 840 | jzfile * |
| 841 | ZIP_Put_In_Cache0(const char *name, ZFILE zfd, char **pmsg, jlong lastModified, |
| 842 | jboolean usemmap) |
| 843 | { |
| 844 | char errbuf[256]; |
| 845 | jlong len; |
| 846 | jzfile *zip; |
| 847 | |
| 848 | if ((zip = allocZip(name)) == NULL) { |
| 849 | return NULL; |
| 850 | } |
| 851 | |
| 852 | #ifdef USE_MMAP |
| 853 | zip->usemmap = usemmap; |
| 854 | #endif |
| 855 | zip->refs = 1; |
| 856 | zip->lastModified = lastModified; |
| 857 | |
| 858 | if (zfd == -1) { |
| 859 | if (pmsg && getLastErrorString(errbuf, sizeof(errbuf)) > 0) |
| 860 | *pmsg = strdup(errbuf); |
| 861 | freeZip(zip); |
| 862 | return NULL; |
| 863 | } |
| 864 | |
| 865 | // Assumption, zfd refers to start of file. Trivially, reuse errbuf. |
| 866 | if (readFully(zfd, errbuf, 4) != -1) { // errors will be handled later |
| 867 | zip->locsig = LOCSIG_AT(errbuf) ? JNI_TRUE : JNI_FALSE; |
| 868 | } |
| 869 | |
| 870 | len = zip->len = IO_Lseek(zfd, 0, SEEK_END); |
| 871 | if (len <= 0) { |
| 872 | if (len == 0) { /* zip file is empty */ |
| 873 | if (pmsg) { |
| 874 | *pmsg = strdup("zip file is empty" ); |
| 875 | } |
| 876 | } else { /* error */ |
| 877 | if (pmsg && getLastErrorString(errbuf, sizeof(errbuf)) > 0) |
| 878 | *pmsg = strdup(errbuf); |
| 879 | } |
| 880 | ZFILE_Close(zfd); |
| 881 | freeZip(zip); |
| 882 | return NULL; |
| 883 | } |
| 884 | |
| 885 | zip->zfd = zfd; |
| 886 | if (readCEN(zip, -1) < 0) { |
| 887 | /* An error occurred while trying to read the zip file */ |
| 888 | if (pmsg != 0) { |
| 889 | /* Set the zip error message */ |
| 890 | if (zip->msg != NULL) |
| 891 | *pmsg = strdup(zip->msg); |
| 892 | } |
| 893 | freeZip(zip); |
| 894 | return NULL; |
| 895 | } |
| 896 | MLOCK(zfiles_lock); |
| 897 | zip->next = zfiles; |
| 898 | zfiles = zip; |
| 899 | MUNLOCK(zfiles_lock); |
| 900 | |
| 901 | return zip; |
| 902 | } |
| 903 | |
| 904 | /* |
| 905 | * Opens a zip file for reading. Returns the jzfile object or NULL |
| 906 | * if an error occurred. If a zip error occurred then *msg will be |
| 907 | * set to the error message text if msg != 0. Otherwise, *msg will be |
| 908 | * set to NULL. Caller doesn't need to free the error message. |
| 909 | */ |
| 910 | JNIEXPORT jzfile * |
| 911 | ZIP_Open(const char *name, char **pmsg) |
| 912 | { |
| 913 | jzfile *file = ZIP_Open_Generic(name, pmsg, O_RDONLY, 0); |
| 914 | if (file == NULL && pmsg != NULL && *pmsg != NULL) { |
| 915 | free(*pmsg); |
| 916 | *pmsg = "Zip file open error" ; |
| 917 | } |
| 918 | return file; |
| 919 | } |
| 920 | |
| 921 | /* |
| 922 | * Closes the specified zip file object. |
| 923 | */ |
| 924 | JNIEXPORT void |
| 925 | ZIP_Close(jzfile *zip) |
| 926 | { |
| 927 | MLOCK(zfiles_lock); |
| 928 | if (--zip->refs > 0) { |
| 929 | /* Still more references so just return */ |
| 930 | MUNLOCK(zfiles_lock); |
| 931 | return; |
| 932 | } |
| 933 | /* No other references so close the file and remove from list */ |
| 934 | if (zfiles == zip) { |
| 935 | zfiles = zfiles->next; |
| 936 | } else { |
| 937 | jzfile *zp; |
| 938 | for (zp = zfiles; zp->next != 0; zp = zp->next) { |
| 939 | if (zp->next == zip) { |
| 940 | zp->next = zip->next; |
| 941 | break; |
| 942 | } |
| 943 | } |
| 944 | } |
| 945 | MUNLOCK(zfiles_lock); |
| 946 | freeZip(zip); |
| 947 | return; |
| 948 | } |
| 949 | |
| 950 | /* Empirically, most CEN headers are smaller than this. */ |
| 951 | #define 160 |
| 952 | |
| 953 | /* A good buffer size when we want to read CEN headers sequentially. */ |
| 954 | #define CENCACHE_PAGESIZE 8192 |
| 955 | |
| 956 | static char * |
| 957 | (jzfile *zip, jlong cenpos, jint bufsize) |
| 958 | { |
| 959 | jint censize; |
| 960 | ZFILE zfd = zip->zfd; |
| 961 | char *cen; |
| 962 | if (bufsize > zip->len - cenpos) |
| 963 | bufsize = (jint)(zip->len - cenpos); |
| 964 | if ((cen = malloc(bufsize)) == NULL) goto Catch; |
| 965 | if (readFullyAt(zfd, cen, bufsize, cenpos) == -1) goto Catch; |
| 966 | censize = CENSIZE(cen); |
| 967 | if (censize <= bufsize) return cen; |
| 968 | if ((cen = realloc(cen, censize)) == NULL) goto Catch; |
| 969 | if (readFully(zfd, cen+bufsize, censize-bufsize) == -1) goto Catch; |
| 970 | return cen; |
| 971 | |
| 972 | Catch: |
| 973 | free(cen); |
| 974 | return NULL; |
| 975 | } |
| 976 | |
| 977 | static char * |
| 978 | (jzfile *zip, jlong cenpos) |
| 979 | { |
| 980 | cencache *cache = &zip->cencache; |
| 981 | char *cen; |
| 982 | if (cache->data != NULL |
| 983 | && (cenpos >= cache->pos) |
| 984 | && (cenpos + CENHDR <= cache->pos + CENCACHE_PAGESIZE)) |
| 985 | { |
| 986 | cen = cache->data + cenpos - cache->pos; |
| 987 | if (cenpos + CENSIZE(cen) <= cache->pos + CENCACHE_PAGESIZE) |
| 988 | /* A cache hit */ |
| 989 | return cen; |
| 990 | } |
| 991 | |
| 992 | if ((cen = readCENHeader(zip, cenpos, CENCACHE_PAGESIZE)) == NULL) |
| 993 | return NULL; |
| 994 | free(cache->data); |
| 995 | cache->data = cen; |
| 996 | cache->pos = cenpos; |
| 997 | return cen; |
| 998 | } |
| 999 | |
| 1000 | typedef enum { ACCESS_RANDOM, ACCESS_SEQUENTIAL } AccessHint; |
| 1001 | |
| 1002 | /* |
| 1003 | * Return a new initialized jzentry corresponding to a given hash cell. |
| 1004 | * In case of error, returns NULL. |
| 1005 | * We already sanity-checked all the CEN headers for ZIP format errors |
| 1006 | * in readCEN(), so we don't check them again here. |
| 1007 | * The ZIP lock should be held here. |
| 1008 | */ |
| 1009 | static jzentry * |
| 1010 | newEntry(jzfile *zip, jzcell *zc, AccessHint accessHint) |
| 1011 | { |
| 1012 | jlong locoff; |
| 1013 | jint nlen, elen, clen; |
| 1014 | jzentry *ze; |
| 1015 | char *cen; |
| 1016 | |
| 1017 | if ((ze = (jzentry *) malloc(sizeof(jzentry))) == NULL) return NULL; |
| 1018 | ze->name = NULL; |
| 1019 | ze->extra = NULL; |
| 1020 | ze->comment = NULL; |
| 1021 | |
| 1022 | #ifdef USE_MMAP |
| 1023 | if (zip->usemmap) { |
| 1024 | cen = (char*) zip->maddr + zc->cenpos - zip->offset; |
| 1025 | } else |
| 1026 | #endif |
| 1027 | { |
| 1028 | if (accessHint == ACCESS_RANDOM) |
| 1029 | cen = readCENHeader(zip, zc->cenpos, AMPLE_CEN_HEADER_SIZE); |
| 1030 | else |
| 1031 | cen = sequentialAccessReadCENHeader(zip, zc->cenpos); |
| 1032 | if (cen == NULL) goto Catch; |
| 1033 | } |
| 1034 | |
| 1035 | nlen = CENNAM(cen); |
| 1036 | elen = CENEXT(cen); |
| 1037 | clen = CENCOM(cen); |
| 1038 | ze->time = CENTIM(cen); |
| 1039 | ze->size = CENLEN(cen); |
| 1040 | ze->csize = (CENHOW(cen) == STORED) ? 0 : CENSIZ(cen); |
| 1041 | ze->crc = CENCRC(cen); |
| 1042 | locoff = CENOFF(cen); |
| 1043 | ze->pos = -(zip->locpos + locoff); |
| 1044 | ze->flag = CENFLG(cen); |
| 1045 | |
| 1046 | if ((ze->name = malloc(nlen + 1)) == NULL) goto Catch; |
| 1047 | memcpy(ze->name, cen + CENHDR, nlen); |
| 1048 | ze->name[nlen] = '\0'; |
| 1049 | ze->nlen = nlen; |
| 1050 | if (elen > 0) { |
| 1051 | char * = cen + CENHDR + nlen; |
| 1052 | |
| 1053 | /* This entry has "extra" data */ |
| 1054 | if ((ze->extra = malloc(elen + 2)) == NULL) goto Catch; |
| 1055 | ze->extra[0] = (unsigned char) elen; |
| 1056 | ze->extra[1] = (unsigned char) (elen >> 8); |
| 1057 | memcpy(ze->extra+2, extra, elen); |
| 1058 | if (ze->csize == ZIP64_MAGICVAL || ze->size == ZIP64_MAGICVAL || |
| 1059 | locoff == ZIP64_MAGICVAL) { |
| 1060 | jint off = 0; |
| 1061 | while ((off + 4) < elen) { // spec: HeaderID+DataSize+Data |
| 1062 | jint sz = SH(extra, off + 2); |
| 1063 | if (SH(extra, off) == ZIP64_EXTID) { |
| 1064 | off += 4; |
| 1065 | if (ze->size == ZIP64_MAGICVAL) { |
| 1066 | // if invalid zip64 extra fields, just skip |
| 1067 | if (sz < 8 || (off + 8) > elen) |
| 1068 | break; |
| 1069 | ze->size = LL(extra, off); |
| 1070 | sz -= 8; |
| 1071 | off += 8; |
| 1072 | } |
| 1073 | if (ze->csize == ZIP64_MAGICVAL) { |
| 1074 | if (sz < 8 || (off + 8) > elen) |
| 1075 | break; |
| 1076 | ze->csize = LL(extra, off); |
| 1077 | sz -= 8; |
| 1078 | off += 8; |
| 1079 | } |
| 1080 | if (locoff == ZIP64_MAGICVAL) { |
| 1081 | if (sz < 8 || (off + 8) > elen) |
| 1082 | break; |
| 1083 | ze->pos = -(zip->locpos + LL(extra, off)); |
| 1084 | sz -= 8; |
| 1085 | off += 8; |
| 1086 | } |
| 1087 | break; |
| 1088 | } |
| 1089 | off += (sz + 4); |
| 1090 | } |
| 1091 | } |
| 1092 | } |
| 1093 | |
| 1094 | if (clen > 0) { |
| 1095 | /* This entry has a comment */ |
| 1096 | if ((ze->comment = malloc(clen + 1)) == NULL) goto Catch; |
| 1097 | memcpy(ze->comment, cen + CENHDR + nlen + elen, clen); |
| 1098 | ze->comment[clen] = '\0'; |
| 1099 | } |
| 1100 | goto Finally; |
| 1101 | |
| 1102 | Catch: |
| 1103 | free(ze->name); |
| 1104 | free(ze->extra); |
| 1105 | free(ze->comment); |
| 1106 | free(ze); |
| 1107 | ze = NULL; |
| 1108 | |
| 1109 | Finally: |
| 1110 | #ifdef USE_MMAP |
| 1111 | if (!zip->usemmap) |
| 1112 | #endif |
| 1113 | if (cen != NULL && accessHint == ACCESS_RANDOM) free(cen); |
| 1114 | return ze; |
| 1115 | } |
| 1116 | |
| 1117 | /* |
| 1118 | * Free the given jzentry. |
| 1119 | * In fact we maintain a one-entry cache of the most recently used |
| 1120 | * jzentry for each zip. This optimizes a common access pattern. |
| 1121 | */ |
| 1122 | |
| 1123 | void |
| 1124 | ZIP_FreeEntry(jzfile *jz, jzentry *ze) |
| 1125 | { |
| 1126 | jzentry *last; |
| 1127 | ZIP_Lock(jz); |
| 1128 | last = jz->cache; |
| 1129 | jz->cache = ze; |
| 1130 | ZIP_Unlock(jz); |
| 1131 | if (last != NULL) { |
| 1132 | /* Free the previously cached jzentry */ |
| 1133 | free(last->name); |
| 1134 | if (last->extra) free(last->extra); |
| 1135 | if (last->comment) free(last->comment); |
| 1136 | free(last); |
| 1137 | } |
| 1138 | } |
| 1139 | |
| 1140 | /* |
| 1141 | * Returns the zip entry corresponding to the specified name, or |
| 1142 | * NULL if not found. |
| 1143 | */ |
| 1144 | jzentry * |
| 1145 | ZIP_GetEntry(jzfile *zip, char *name, jint ulen) |
| 1146 | { |
| 1147 | if (ulen == 0) { |
| 1148 | return ZIP_GetEntry2(zip, name, (jint)strlen(name), JNI_FALSE); |
| 1149 | } |
| 1150 | return ZIP_GetEntry2(zip, name, ulen, JNI_TRUE); |
| 1151 | } |
| 1152 | |
| 1153 | jboolean equals(char* name1, int len1, char* name2, int len2) { |
| 1154 | if (len1 != len2) { |
| 1155 | return JNI_FALSE; |
| 1156 | } |
| 1157 | while (len1-- > 0) { |
| 1158 | if (*name1++ != *name2++) { |
| 1159 | return JNI_FALSE; |
| 1160 | } |
| 1161 | } |
| 1162 | return JNI_TRUE; |
| 1163 | } |
| 1164 | |
| 1165 | /* |
| 1166 | * Returns the zip entry corresponding to the specified name, or |
| 1167 | * NULL if not found. |
| 1168 | * This method supports embedded null character in "name", use ulen |
| 1169 | * for the length of "name". |
| 1170 | */ |
| 1171 | jzentry * |
| 1172 | ZIP_GetEntry2(jzfile *zip, char *name, jint ulen, jboolean addSlash) |
| 1173 | { |
| 1174 | unsigned int hsh = hashN(name, ulen); |
| 1175 | jint idx; |
| 1176 | jzentry *ze = 0; |
| 1177 | |
| 1178 | ZIP_Lock(zip); |
| 1179 | if (zip->total == 0) { |
| 1180 | goto Finally; |
| 1181 | } |
| 1182 | |
| 1183 | idx = zip->table[hsh % zip->tablelen]; |
| 1184 | |
| 1185 | /* |
| 1186 | * This while loop is an optimization where a double lookup |
| 1187 | * for name and name+/ is being performed. The name char |
| 1188 | * array has enough room at the end to try again with a |
| 1189 | * slash appended if the first table lookup does not succeed. |
| 1190 | */ |
| 1191 | while(1) { |
| 1192 | |
| 1193 | /* Check the cached entry first */ |
| 1194 | ze = zip->cache; |
| 1195 | if (ze && equals(ze->name, ze->nlen, name, ulen)) { |
| 1196 | /* Cache hit! Remove and return the cached entry. */ |
| 1197 | zip->cache = 0; |
| 1198 | ZIP_Unlock(zip); |
| 1199 | return ze; |
| 1200 | } |
| 1201 | ze = 0; |
| 1202 | |
| 1203 | /* |
| 1204 | * Search down the target hash chain for a cell whose |
| 1205 | * 32 bit hash matches the hashed name. |
| 1206 | */ |
| 1207 | while (idx != ZIP_ENDCHAIN) { |
| 1208 | jzcell *zc = &zip->entries[idx]; |
| 1209 | |
| 1210 | if (zc->hash == hsh) { |
| 1211 | /* |
| 1212 | * OK, we've found a ZIP entry whose 32 bit hashcode |
| 1213 | * matches the name we're looking for. Try to read |
| 1214 | * its entry information from the CEN. If the CEN |
| 1215 | * name matches the name we're looking for, we're |
| 1216 | * done. |
| 1217 | * If the names don't match (which should be very rare) |
| 1218 | * we keep searching. |
| 1219 | */ |
| 1220 | ze = newEntry(zip, zc, ACCESS_RANDOM); |
| 1221 | if (ze && equals(ze->name, ze->nlen, name, ulen)) { |
| 1222 | break; |
| 1223 | } |
| 1224 | if (ze != 0) { |
| 1225 | /* We need to release the lock across the free call */ |
| 1226 | ZIP_Unlock(zip); |
| 1227 | ZIP_FreeEntry(zip, ze); |
| 1228 | ZIP_Lock(zip); |
| 1229 | } |
| 1230 | ze = 0; |
| 1231 | } |
| 1232 | idx = zc->next; |
| 1233 | } |
| 1234 | |
| 1235 | /* Entry found, return it */ |
| 1236 | if (ze != 0) { |
| 1237 | break; |
| 1238 | } |
| 1239 | |
| 1240 | /* If no need to try appending slash, we are done */ |
| 1241 | if (!addSlash) { |
| 1242 | break; |
| 1243 | } |
| 1244 | |
| 1245 | /* Slash is already there? */ |
| 1246 | if (ulen > 0 && name[ulen - 1] == '/') { |
| 1247 | break; |
| 1248 | } |
| 1249 | |
| 1250 | /* Add slash and try once more */ |
| 1251 | name[ulen++] = '/'; |
| 1252 | name[ulen] = '\0'; |
| 1253 | hsh = hash_append(hsh, '/'); |
| 1254 | idx = zip->table[hsh % zip->tablelen]; |
| 1255 | addSlash = JNI_FALSE; |
| 1256 | } |
| 1257 | |
| 1258 | Finally: |
| 1259 | ZIP_Unlock(zip); |
| 1260 | return ze; |
| 1261 | } |
| 1262 | |
| 1263 | /* |
| 1264 | * Returns the n'th (starting at zero) zip file entry, or NULL if the |
| 1265 | * specified index was out of range. |
| 1266 | */ |
| 1267 | JNIEXPORT jzentry * |
| 1268 | ZIP_GetNextEntry(jzfile *zip, jint n) |
| 1269 | { |
| 1270 | jzentry *result; |
| 1271 | if (n < 0 || n >= zip->total) { |
| 1272 | return 0; |
| 1273 | } |
| 1274 | ZIP_Lock(zip); |
| 1275 | result = newEntry(zip, &zip->entries[n], ACCESS_SEQUENTIAL); |
| 1276 | ZIP_Unlock(zip); |
| 1277 | return result; |
| 1278 | } |
| 1279 | |
| 1280 | /* |
| 1281 | * Locks the specified zip file for reading. |
| 1282 | */ |
| 1283 | void |
| 1284 | ZIP_Lock(jzfile *zip) |
| 1285 | { |
| 1286 | MLOCK(zip->lock); |
| 1287 | } |
| 1288 | |
| 1289 | /* |
| 1290 | * Unlocks the specified zip file. |
| 1291 | */ |
| 1292 | void |
| 1293 | ZIP_Unlock(jzfile *zip) |
| 1294 | { |
| 1295 | MUNLOCK(zip->lock); |
| 1296 | } |
| 1297 | |
| 1298 | /* |
| 1299 | * Returns the offset of the entry data within the zip file. |
| 1300 | * Returns -1 if an error occurred, in which case zip->msg will |
| 1301 | * contain the error text. |
| 1302 | */ |
| 1303 | jlong |
| 1304 | ZIP_GetEntryDataOffset(jzfile *zip, jzentry *entry) |
| 1305 | { |
| 1306 | /* The Zip file spec explicitly allows the LOC extra data size to |
| 1307 | * be different from the CEN extra data size, although the JDK |
| 1308 | * never creates such zip files. Since we cannot trust the CEN |
| 1309 | * extra data size, we need to read the LOC to determine the entry |
| 1310 | * data offset. We do this lazily to avoid touching the virtual |
| 1311 | * memory page containing the LOC when initializing jzentry |
| 1312 | * objects. (This speeds up javac by a factor of 10 when the JDK |
| 1313 | * is installed on a very slow filesystem.) |
| 1314 | */ |
| 1315 | if (entry->pos <= 0) { |
| 1316 | unsigned char loc[LOCHDR]; |
| 1317 | if (readFullyAt(zip->zfd, loc, LOCHDR, -(entry->pos)) == -1) { |
| 1318 | zip->msg = "error reading zip file" ; |
| 1319 | return -1; |
| 1320 | } |
| 1321 | if (!LOCSIG_AT(loc)) { |
| 1322 | zip->msg = "invalid LOC header (bad signature)" ; |
| 1323 | return -1; |
| 1324 | } |
| 1325 | entry->pos = (- entry->pos) + LOCHDR + LOCNAM(loc) + LOCEXT(loc); |
| 1326 | } |
| 1327 | return entry->pos; |
| 1328 | } |
| 1329 | |
| 1330 | /* |
| 1331 | * Reads bytes from the specified zip entry. Assumes that the zip |
| 1332 | * file had been previously locked with ZIP_Lock(). Returns the |
| 1333 | * number of bytes read, or -1 if an error occurred. If zip->msg != 0 |
| 1334 | * then a zip error occurred and zip->msg contains the error text. |
| 1335 | * |
| 1336 | * The current implementation does not support reading an entry that |
| 1337 | * has the size bigger than 2**32 bytes in ONE invocation. |
| 1338 | */ |
| 1339 | jint |
| 1340 | ZIP_Read(jzfile *zip, jzentry *entry, jlong pos, void *buf, jint len) |
| 1341 | { |
| 1342 | jlong entry_size; |
| 1343 | jlong start; |
| 1344 | |
| 1345 | if (zip == 0) { |
| 1346 | return -1; |
| 1347 | } |
| 1348 | |
| 1349 | /* Clear previous zip error */ |
| 1350 | zip->msg = NULL; |
| 1351 | |
| 1352 | if (entry == 0) { |
| 1353 | zip->msg = "ZIP_Read: jzentry is NULL" ; |
| 1354 | return -1; |
| 1355 | } |
| 1356 | |
| 1357 | entry_size = (entry->csize != 0) ? entry->csize : entry->size; |
| 1358 | |
| 1359 | /* Check specified position */ |
| 1360 | if (pos < 0 || pos > entry_size - 1) { |
| 1361 | zip->msg = "ZIP_Read: specified offset out of range" ; |
| 1362 | return -1; |
| 1363 | } |
| 1364 | |
| 1365 | /* Check specified length */ |
| 1366 | if (len <= 0) |
| 1367 | return 0; |
| 1368 | if (len > entry_size - pos) |
| 1369 | len = (jint)(entry_size - pos); |
| 1370 | |
| 1371 | /* Get file offset to start reading data */ |
| 1372 | start = ZIP_GetEntryDataOffset(zip, entry); |
| 1373 | if (start < 0) |
| 1374 | return -1; |
| 1375 | start += pos; |
| 1376 | |
| 1377 | if (start + len > zip->len) { |
| 1378 | zip->msg = "ZIP_Read: corrupt zip file: invalid entry size" ; |
| 1379 | return -1; |
| 1380 | } |
| 1381 | |
| 1382 | if (readFullyAt(zip->zfd, buf, len, start) == -1) { |
| 1383 | zip->msg = "ZIP_Read: error reading zip file" ; |
| 1384 | return -1; |
| 1385 | } |
| 1386 | return len; |
| 1387 | } |
| 1388 | |
| 1389 | |
| 1390 | /* The maximum size of a stack-allocated buffer. |
| 1391 | */ |
| 1392 | #define BUF_SIZE 4096 |
| 1393 | |
| 1394 | /* |
| 1395 | * This function is used by the runtime system to load compressed entries |
| 1396 | * from ZIP/JAR files specified in the class path. It is defined here |
| 1397 | * so that it can be dynamically loaded by the runtime if the zip library |
| 1398 | * is found. |
| 1399 | * |
| 1400 | * The current implementation does not support reading an entry that |
| 1401 | * has the size bigger than 2**32 bytes in ONE invocation. |
| 1402 | */ |
| 1403 | jboolean |
| 1404 | InflateFully(jzfile *zip, jzentry *entry, void *buf, char **msg) |
| 1405 | { |
| 1406 | z_stream strm; |
| 1407 | char tmp[BUF_SIZE]; |
| 1408 | jlong pos = 0; |
| 1409 | jlong count = entry->csize; |
| 1410 | |
| 1411 | *msg = 0; /* Reset error message */ |
| 1412 | |
| 1413 | if (count == 0) { |
| 1414 | *msg = "inflateFully: entry not compressed" ; |
| 1415 | return JNI_FALSE; |
| 1416 | } |
| 1417 | |
| 1418 | memset(&strm, 0, sizeof(z_stream)); |
| 1419 | if (inflateInit2(&strm, -MAX_WBITS) != Z_OK) { |
| 1420 | *msg = strm.msg; |
| 1421 | return JNI_FALSE; |
| 1422 | } |
| 1423 | |
| 1424 | strm.next_out = buf; |
| 1425 | strm.avail_out = (uInt)entry->size; |
| 1426 | |
| 1427 | while (count > 0) { |
| 1428 | jint n = count > (jlong)sizeof(tmp) ? (jint)sizeof(tmp) : (jint)count; |
| 1429 | ZIP_Lock(zip); |
| 1430 | n = ZIP_Read(zip, entry, pos, tmp, n); |
| 1431 | ZIP_Unlock(zip); |
| 1432 | if (n <= 0) { |
| 1433 | if (n == 0) { |
| 1434 | *msg = "inflateFully: Unexpected end of file" ; |
| 1435 | } |
| 1436 | inflateEnd(&strm); |
| 1437 | return JNI_FALSE; |
| 1438 | } |
| 1439 | pos += n; |
| 1440 | count -= n; |
| 1441 | strm.next_in = (Bytef *)tmp; |
| 1442 | strm.avail_in = n; |
| 1443 | do { |
| 1444 | switch (inflate(&strm, Z_PARTIAL_FLUSH)) { |
| 1445 | case Z_OK: |
| 1446 | break; |
| 1447 | case Z_STREAM_END: |
| 1448 | if (count != 0 || strm.total_out != (uInt)entry->size) { |
| 1449 | *msg = "inflateFully: Unexpected end of stream" ; |
| 1450 | inflateEnd(&strm); |
| 1451 | return JNI_FALSE; |
| 1452 | } |
| 1453 | break; |
| 1454 | default: |
| 1455 | break; |
| 1456 | } |
| 1457 | } while (strm.avail_in > 0); |
| 1458 | } |
| 1459 | |
| 1460 | inflateEnd(&strm); |
| 1461 | return JNI_TRUE; |
| 1462 | } |
| 1463 | |
| 1464 | /* |
| 1465 | * The current implementation does not support reading an entry that |
| 1466 | * has the size bigger than 2**32 bytes in ONE invocation. |
| 1467 | */ |
| 1468 | JNIEXPORT jzentry * |
| 1469 | ZIP_FindEntry(jzfile *zip, char *name, jint *sizeP, jint *nameLenP) |
| 1470 | { |
| 1471 | jzentry *entry = ZIP_GetEntry(zip, name, 0); |
| 1472 | if (entry) { |
| 1473 | *sizeP = (jint)entry->size; |
| 1474 | *nameLenP = (jint)strlen(entry->name); |
| 1475 | } |
| 1476 | return entry; |
| 1477 | } |
| 1478 | |
| 1479 | /* |
| 1480 | * Reads a zip file entry into the specified byte array |
| 1481 | * When the method completes, it releases the jzentry. |
| 1482 | * Note: this is called from the separately delivered VM (hotspot/classic) |
| 1483 | * so we have to be careful to maintain the expected behaviour. |
| 1484 | */ |
| 1485 | JNIEXPORT jboolean |
| 1486 | ZIP_ReadEntry(jzfile *zip, jzentry *entry, unsigned char *buf, char *entryname) |
| 1487 | { |
| 1488 | char *msg; |
| 1489 | char tmpbuf[1024]; |
| 1490 | |
| 1491 | if (entry == 0) { |
| 1492 | jio_fprintf(stderr, "jzentry was invalid" ); |
| 1493 | return JNI_FALSE; |
| 1494 | } |
| 1495 | |
| 1496 | strcpy(entryname, entry->name); |
| 1497 | if (entry->csize == 0) { |
| 1498 | /* Entry is stored */ |
| 1499 | jlong pos = 0; |
| 1500 | jlong size = entry->size; |
| 1501 | while (pos < size) { |
| 1502 | jint n; |
| 1503 | jlong limit = ((((jlong) 1) << 31) - 1); |
| 1504 | jint count = (size - pos < limit) ? |
| 1505 | /* These casts suppress a VC++ Internal Compiler Error */ |
| 1506 | (jint) (size - pos) : |
| 1507 | (jint) limit; |
| 1508 | ZIP_Lock(zip); |
| 1509 | n = ZIP_Read(zip, entry, pos, buf, count); |
| 1510 | msg = zip->msg; |
| 1511 | ZIP_Unlock(zip); |
| 1512 | if (n == -1) { |
| 1513 | if (msg == 0) { |
| 1514 | getErrorString(errno, tmpbuf, sizeof(tmpbuf)); |
| 1515 | msg = tmpbuf; |
| 1516 | } |
| 1517 | jio_fprintf(stderr, "%s: %s\n" , zip->name, msg); |
| 1518 | return JNI_FALSE; |
| 1519 | } |
| 1520 | buf += n; |
| 1521 | pos += n; |
| 1522 | } |
| 1523 | } else { |
| 1524 | /* Entry is compressed */ |
| 1525 | int ok = InflateFully(zip, entry, buf, &msg); |
| 1526 | if (!ok) { |
| 1527 | if ((msg == NULL) || (*msg == 0)) { |
| 1528 | msg = zip->msg; |
| 1529 | } |
| 1530 | if (msg == 0) { |
| 1531 | getErrorString(errno, tmpbuf, sizeof(tmpbuf)); |
| 1532 | msg = tmpbuf; |
| 1533 | } |
| 1534 | jio_fprintf(stderr, "%s: %s\n" , zip->name, msg); |
| 1535 | return JNI_FALSE; |
| 1536 | } |
| 1537 | } |
| 1538 | |
| 1539 | ZIP_FreeEntry(zip, entry); |
| 1540 | |
| 1541 | return JNI_TRUE; |
| 1542 | } |
| 1543 | |
| 1544 | JNIEXPORT jboolean |
| 1545 | ZIP_InflateFully(void *inBuf, jlong inLen, void *outBuf, jlong outLen, char **pmsg) |
| 1546 | { |
| 1547 | z_stream strm; |
| 1548 | int i = 0; |
| 1549 | memset(&strm, 0, sizeof(z_stream)); |
| 1550 | |
| 1551 | *pmsg = 0; /* Reset error message */ |
| 1552 | |
| 1553 | if (inflateInit2(&strm, MAX_WBITS) != Z_OK) { |
| 1554 | *pmsg = strm.msg; |
| 1555 | return JNI_FALSE; |
| 1556 | } |
| 1557 | |
| 1558 | strm.next_out = (Bytef *) outBuf; |
| 1559 | strm.avail_out = (uInt)outLen; |
| 1560 | strm.next_in = (Bytef *) inBuf; |
| 1561 | strm.avail_in = (uInt)inLen; |
| 1562 | |
| 1563 | do { |
| 1564 | switch (inflate(&strm, Z_PARTIAL_FLUSH)) { |
| 1565 | case Z_OK: |
| 1566 | break; |
| 1567 | case Z_STREAM_END: |
| 1568 | if (strm.total_out != (uInt)outLen) { |
| 1569 | *pmsg = "INFLATER_inflateFully: Unexpected end of stream" ; |
| 1570 | inflateEnd(&strm); |
| 1571 | return JNI_FALSE; |
| 1572 | } |
| 1573 | break; |
| 1574 | case Z_DATA_ERROR: |
| 1575 | *pmsg = "INFLATER_inflateFully: Compressed data corrupted" ; |
| 1576 | inflateEnd(&strm); |
| 1577 | return JNI_FALSE; |
| 1578 | case Z_MEM_ERROR: |
| 1579 | *pmsg = "INFLATER_inflateFully: out of memory" ; |
| 1580 | inflateEnd(&strm); |
| 1581 | return JNI_FALSE; |
| 1582 | default: |
| 1583 | *pmsg = "INFLATER_inflateFully: internal error" ; |
| 1584 | inflateEnd(&strm); |
| 1585 | return JNI_FALSE; |
| 1586 | } |
| 1587 | } while (strm.avail_in > 0); |
| 1588 | |
| 1589 | inflateEnd(&strm); |
| 1590 | return JNI_TRUE; |
| 1591 | } |
| 1592 | |