Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2008 The Android Open Source Project |
| 3 | * |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | * you may not use this file except in compliance with the License. |
| 6 | * You may obtain a copy of the License at |
| 7 | * |
| 8 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | * |
| 10 | * Unless required by applicable law or agreed to in writing, software |
| 11 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | * See the License for the specific language governing permissions and |
| 14 | * limitations under the License. |
| 15 | */ |
| 16 | |
| 17 | /* |
| 18 | * Read-only access to Zip archives, with minimal heap allocation. |
| 19 | */ |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 20 | |
| 21 | #include <assert.h> |
| 22 | #include <errno.h> |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 23 | #include <fcntl.h> |
| 24 | #include <inttypes.h> |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 25 | #include <limits.h> |
| 26 | #include <log/log.h> |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 27 | #include <stdlib.h> |
| 28 | #include <string.h> |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 29 | #include <unistd.h> |
Mark Salyzyn | 51d562d | 2014-05-05 14:38:05 -0700 | [diff] [blame] | 30 | #include <utils/Compat.h> |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 31 | #include <utils/FileMap.h> |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 32 | #include <zlib.h> |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 33 | |
| 34 | #include <JNIHelp.h> // TEMP_FAILURE_RETRY may or may not be in unistd |
| 35 | |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 36 | #include "ziparchive/zip_archive.h" |
| 37 | |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 38 | // This is for windows. If we don't open a file in binary mode, weirds |
| 39 | // things will happen. |
| 40 | #ifndef O_BINARY |
| 41 | #define O_BINARY 0 |
| 42 | #endif |
| 43 | |
| 44 | /* |
| 45 | * Zip file constants. |
| 46 | */ |
Narayan Kamath | 4f6b499 | 2014-06-03 13:59:23 +0100 | [diff] [blame^] | 47 | static const uint32_t kEOCDSignature = 0x06054b50; |
| 48 | static const uint32_t kEOCDLen = 2; |
| 49 | static const uint32_t kEOCDNumEntries = 8; // number of entries in the archive |
| 50 | static const uint32_t kEOCDSize = 12; // size of the central directory |
| 51 | static const uint32_t kEOCDFileOffset = 16; // offset to central directory |
| 52 | static const uint32_t kEOCDCommentLen = 20; // length of the EOCD comment |
| 53 | static const uint32_t kEOCDComment = 22; // offset of the EOCD comment |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 54 | |
| 55 | static const uint32_t kMaxCommentLen = 65535; // longest possible in ushort |
| 56 | static const uint32_t kMaxEOCDSearch = (kMaxCommentLen + kEOCDLen); |
| 57 | |
| 58 | static const uint32_t kLFHSignature = 0x04034b50; |
| 59 | static const uint32_t kLFHLen = 30; // excluding variable-len fields |
| 60 | static const uint32_t kLFHGPBFlags = 6; // general purpose bit flags |
| 61 | static const uint32_t kLFHCRC = 14; // offset to CRC |
| 62 | static const uint32_t kLFHCompLen = 18; // offset to compressed length |
| 63 | static const uint32_t kLFHUncompLen = 22; // offset to uncompressed length |
| 64 | static const uint32_t kLFHNameLen = 26; // offset to filename length |
| 65 | static const uint32_t kLFHExtraLen = 28; // offset to extra length |
| 66 | |
| 67 | static const uint32_t kCDESignature = 0x02014b50; |
| 68 | static const uint32_t kCDELen = 46; // excluding variable-len fields |
| 69 | static const uint32_t kCDEMethod = 10; // offset to compression method |
| 70 | static const uint32_t kCDEModWhen = 12; // offset to modification timestamp |
| 71 | static const uint32_t kCDECRC = 16; // offset to entry CRC |
| 72 | static const uint32_t kCDECompLen = 20; // offset to compressed length |
| 73 | static const uint32_t kCDEUncompLen = 24; // offset to uncompressed length |
| 74 | static const uint32_t kCDENameLen = 28; // offset to filename length |
| 75 | static const uint32_t kCDEExtraLen = 30; // offset to extra length |
| 76 | static const uint32_t kCDECommentLen = 32; // offset to comment length |
| 77 | static const uint32_t kCDELocalOffset = 42; // offset to local hdr |
| 78 | |
| 79 | static const uint32_t kDDOptSignature = 0x08074b50; // *OPTIONAL* data descriptor signature |
| 80 | static const uint32_t kDDSignatureLen = 4; |
| 81 | static const uint32_t kDDLen = 12; |
| 82 | static const uint32_t kDDMaxLen = 16; // max of 16 bytes with a signature, 12 bytes without |
| 83 | static const uint32_t kDDCrc32 = 0; // offset to crc32 |
| 84 | static const uint32_t kDDCompLen = 4; // offset to compressed length |
| 85 | static const uint32_t kDDUncompLen = 8; // offset to uncompressed length |
| 86 | |
| 87 | static const uint32_t kGPBDDFlagMask = 0x0008; // mask value that signifies that the entry has a DD |
| 88 | |
| 89 | static const uint32_t kMaxErrorLen = 1024; |
| 90 | |
| 91 | static const char* kErrorMessages[] = { |
| 92 | "Unknown return code.", |
Narayan Kamath | eb41ad2 | 2013-12-09 16:26:36 +0000 | [diff] [blame] | 93 | "Iteration ended", |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 94 | "Zlib error", |
| 95 | "Invalid file", |
| 96 | "Invalid handle", |
| 97 | "Duplicate entries in archive", |
| 98 | "Empty archive", |
| 99 | "Entry not found", |
| 100 | "Invalid offset", |
| 101 | "Inconsistent information", |
| 102 | "Invalid entry name", |
Narayan Kamath | eb41ad2 | 2013-12-09 16:26:36 +0000 | [diff] [blame] | 103 | "I/O Error", |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 104 | "File mapping failed" |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 105 | }; |
| 106 | |
| 107 | static const int32_t kErrorMessageUpperBound = 0; |
| 108 | |
Narayan Kamath | eb41ad2 | 2013-12-09 16:26:36 +0000 | [diff] [blame] | 109 | static const int32_t kIterationEnd = -1; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 110 | |
| 111 | // We encountered a Zlib error when inflating a stream from this file. |
| 112 | // Usually indicates file corruption. |
| 113 | static const int32_t kZlibError = -2; |
| 114 | |
| 115 | // The input file cannot be processed as a zip archive. Usually because |
| 116 | // it's too small, too large or does not have a valid signature. |
| 117 | static const int32_t kInvalidFile = -3; |
| 118 | |
| 119 | // An invalid iteration / ziparchive handle was passed in as an input |
| 120 | // argument. |
| 121 | static const int32_t kInvalidHandle = -4; |
| 122 | |
| 123 | // The zip archive contained two (or possibly more) entries with the same |
| 124 | // name. |
| 125 | static const int32_t kDuplicateEntry = -5; |
| 126 | |
| 127 | // The zip archive contains no entries. |
| 128 | static const int32_t kEmptyArchive = -6; |
| 129 | |
| 130 | // The specified entry was not found in the archive. |
| 131 | static const int32_t kEntryNotFound = -7; |
| 132 | |
| 133 | // The zip archive contained an invalid local file header pointer. |
| 134 | static const int32_t kInvalidOffset = -8; |
| 135 | |
| 136 | // The zip archive contained inconsistent entry information. This could |
| 137 | // be because the central directory & local file header did not agree, or |
| 138 | // if the actual uncompressed length or crc32 do not match their declared |
| 139 | // values. |
| 140 | static const int32_t kInconsistentInformation = -9; |
| 141 | |
| 142 | // An invalid entry name was encountered. |
| 143 | static const int32_t kInvalidEntryName = -10; |
| 144 | |
Narayan Kamath | eb41ad2 | 2013-12-09 16:26:36 +0000 | [diff] [blame] | 145 | // An I/O related system call (read, lseek, ftruncate, map) failed. |
| 146 | static const int32_t kIoError = -11; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 147 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 148 | // We were not able to mmap the central directory or entry contents. |
| 149 | static const int32_t kMmapFailed = -12; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 150 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 151 | static const int32_t kErrorMessageLowerBound = -13; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 152 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 153 | static const char kTempMappingFileName[] = "zip: ExtractFileToFile"; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 154 | |
| 155 | /* |
| 156 | * A Read-only Zip archive. |
| 157 | * |
| 158 | * We want "open" and "find entry by name" to be fast operations, and |
| 159 | * we want to use as little memory as possible. We memory-map the zip |
| 160 | * central directory, and load a hash table with pointers to the filenames |
| 161 | * (which aren't null-terminated). The other fields are at a fixed offset |
| 162 | * from the filename, so we don't need to extract those (but we do need |
| 163 | * to byte-read and endian-swap them every time we want them). |
| 164 | * |
| 165 | * It's possible that somebody has handed us a massive (~1GB) zip archive, |
| 166 | * so we can't expect to mmap the entire file. |
| 167 | * |
| 168 | * To speed comparisons when doing a lookup by name, we could make the mapping |
| 169 | * "private" (copy-on-write) and null-terminate the filenames after verifying |
| 170 | * the record structure. However, this requires a private mapping of |
| 171 | * every page that the Central Directory touches. Easier to tuck a copy |
| 172 | * of the string length into the hash table entry. |
| 173 | */ |
| 174 | struct ZipArchive { |
| 175 | /* open Zip archive */ |
| 176 | int fd; |
| 177 | |
| 178 | /* mapped central directory area */ |
| 179 | off64_t directory_offset; |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 180 | android::FileMap* directory_map; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 181 | |
| 182 | /* number of entries in the Zip archive */ |
| 183 | uint16_t num_entries; |
| 184 | |
| 185 | /* |
| 186 | * We know how many entries are in the Zip archive, so we can have a |
| 187 | * fixed-size hash table. We define a load factor of 0.75 and overallocat |
| 188 | * so the maximum number entries can never be higher than |
| 189 | * ((4 * UINT16_MAX) / 3 + 1) which can safely fit into a uint32_t. |
| 190 | */ |
| 191 | uint32_t hash_table_size; |
| 192 | ZipEntryName* hash_table; |
| 193 | }; |
| 194 | |
| 195 | // Returns 0 on success and negative values on failure. |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 196 | static android::FileMap* MapFileSegment(const int fd, const off64_t start, |
| 197 | const size_t length, const bool read_only, |
| 198 | const char* debug_file_name) { |
| 199 | android::FileMap* file_map = new android::FileMap; |
| 200 | const bool success = file_map->create(debug_file_name, fd, start, length, read_only); |
| 201 | if (!success) { |
| 202 | file_map->release(); |
| 203 | return NULL; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 204 | } |
| 205 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 206 | return file_map; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 207 | } |
| 208 | |
| 209 | static int32_t CopyFileToFile(int fd, uint8_t* begin, const uint32_t length, uint64_t *crc_out) { |
| 210 | static const uint32_t kBufSize = 32768; |
| 211 | uint8_t buf[kBufSize]; |
| 212 | |
| 213 | uint32_t count = 0; |
| 214 | uint64_t crc = 0; |
Narayan Kamath | 58aaf46 | 2013-12-10 16:47:14 +0000 | [diff] [blame] | 215 | while (count < length) { |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 216 | uint32_t remaining = length - count; |
| 217 | |
| 218 | // Safe conversion because kBufSize is narrow enough for a 32 bit signed |
| 219 | // value. |
| 220 | ssize_t get_size = (remaining > kBufSize) ? kBufSize : remaining; |
| 221 | ssize_t actual = TEMP_FAILURE_RETRY(read(fd, buf, get_size)); |
| 222 | |
| 223 | if (actual != get_size) { |
Mark Salyzyn | 51d562d | 2014-05-05 14:38:05 -0700 | [diff] [blame] | 224 | ALOGW("CopyFileToFile: copy read failed (" ZD " vs " ZD ")", actual, get_size); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 225 | return kIoError; |
| 226 | } |
| 227 | |
| 228 | memcpy(begin + count, buf, get_size); |
| 229 | crc = crc32(crc, buf, get_size); |
| 230 | count += get_size; |
| 231 | } |
| 232 | |
| 233 | *crc_out = crc; |
| 234 | |
| 235 | return 0; |
| 236 | } |
| 237 | |
| 238 | /* |
| 239 | * Round up to the next highest power of 2. |
| 240 | * |
| 241 | * Found on http://graphics.stanford.edu/~seander/bithacks.html. |
| 242 | */ |
| 243 | static uint32_t RoundUpPower2(uint32_t val) { |
| 244 | val--; |
| 245 | val |= val >> 1; |
| 246 | val |= val >> 2; |
| 247 | val |= val >> 4; |
| 248 | val |= val >> 8; |
| 249 | val |= val >> 16; |
| 250 | val++; |
| 251 | |
| 252 | return val; |
| 253 | } |
| 254 | |
| 255 | static uint32_t ComputeHash(const char* str, uint16_t len) { |
| 256 | uint32_t hash = 0; |
| 257 | |
| 258 | while (len--) { |
| 259 | hash = hash * 31 + *str++; |
| 260 | } |
| 261 | |
| 262 | return hash; |
| 263 | } |
| 264 | |
| 265 | /* |
| 266 | * Convert a ZipEntry to a hash table index, verifying that it's in a |
| 267 | * valid range. |
| 268 | */ |
| 269 | static int64_t EntryToIndex(const ZipEntryName* hash_table, |
| 270 | const uint32_t hash_table_size, |
| 271 | const char* name, uint16_t length) { |
| 272 | const uint32_t hash = ComputeHash(name, length); |
| 273 | |
| 274 | // NOTE: (hash_table_size - 1) is guaranteed to be non-negative. |
| 275 | uint32_t ent = hash & (hash_table_size - 1); |
| 276 | while (hash_table[ent].name != NULL) { |
| 277 | if (hash_table[ent].name_length == length && |
| 278 | memcmp(hash_table[ent].name, name, length) == 0) { |
| 279 | return ent; |
| 280 | } |
| 281 | |
| 282 | ent = (ent + 1) & (hash_table_size - 1); |
| 283 | } |
| 284 | |
Colin Cross | f4b0b79 | 2014-02-06 20:07:15 -0800 | [diff] [blame] | 285 | ALOGV("Zip: Unable to find entry %.*s", length, name); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 286 | return kEntryNotFound; |
| 287 | } |
| 288 | |
| 289 | /* |
| 290 | * Add a new entry to the hash table. |
| 291 | */ |
| 292 | static int32_t AddToHash(ZipEntryName *hash_table, const uint64_t hash_table_size, |
| 293 | const char* name, uint16_t length) { |
| 294 | const uint64_t hash = ComputeHash(name, length); |
| 295 | uint32_t ent = hash & (hash_table_size - 1); |
| 296 | |
| 297 | /* |
| 298 | * We over-allocated the table, so we're guaranteed to find an empty slot. |
| 299 | * Further, we guarantee that the hashtable size is not 0. |
| 300 | */ |
| 301 | while (hash_table[ent].name != NULL) { |
| 302 | if (hash_table[ent].name_length == length && |
| 303 | memcmp(hash_table[ent].name, name, length) == 0) { |
| 304 | // We've found a duplicate entry. We don't accept it |
| 305 | ALOGW("Zip: Found duplicate entry %.*s", length, name); |
| 306 | return kDuplicateEntry; |
| 307 | } |
| 308 | ent = (ent + 1) & (hash_table_size - 1); |
| 309 | } |
| 310 | |
| 311 | hash_table[ent].name = name; |
| 312 | hash_table[ent].name_length = length; |
| 313 | return 0; |
| 314 | } |
| 315 | |
| 316 | /* |
| 317 | * Get 2 little-endian bytes. |
| 318 | */ |
| 319 | static uint16_t get2LE(const uint8_t* src) { |
| 320 | return src[0] | (src[1] << 8); |
| 321 | } |
| 322 | |
| 323 | /* |
| 324 | * Get 4 little-endian bytes. |
| 325 | */ |
| 326 | static uint32_t get4LE(const uint8_t* src) { |
| 327 | uint32_t result; |
| 328 | |
| 329 | result = src[0]; |
| 330 | result |= src[1] << 8; |
| 331 | result |= src[2] << 16; |
| 332 | result |= src[3] << 24; |
| 333 | |
| 334 | return result; |
| 335 | } |
| 336 | |
| 337 | static int32_t MapCentralDirectory0(int fd, const char* debug_file_name, |
| 338 | ZipArchive* archive, off64_t file_length, |
| 339 | uint32_t read_amount, uint8_t* scan_buffer) { |
| 340 | const off64_t search_start = file_length - read_amount; |
| 341 | |
| 342 | if (lseek64(fd, search_start, SEEK_SET) != search_start) { |
Mark Salyzyn | 96c5c99 | 2014-05-08 19:16:40 -0700 | [diff] [blame] | 343 | ALOGW("Zip: seek %" PRId64 " failed: %s", (int64_t)search_start, strerror(errno)); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 344 | return kIoError; |
| 345 | } |
| 346 | ssize_t actual = TEMP_FAILURE_RETRY(read(fd, scan_buffer, read_amount)); |
| 347 | if (actual != (ssize_t) read_amount) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 348 | ALOGW("Zip: read %" PRIu32 " failed: %s", read_amount, strerror(errno)); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 349 | return kIoError; |
| 350 | } |
| 351 | |
| 352 | /* |
| 353 | * Scan backward for the EOCD magic. In an archive without a trailing |
| 354 | * comment, we'll find it on the first try. (We may want to consider |
| 355 | * doing an initial minimal read; if we don't find it, retry with a |
| 356 | * second read as above.) |
| 357 | */ |
| 358 | int i; |
| 359 | for (i = read_amount - kEOCDLen; i >= 0; i--) { |
| 360 | if (scan_buffer[i] == 0x50 && get4LE(&scan_buffer[i]) == kEOCDSignature) { |
| 361 | ALOGV("+++ Found EOCD at buf+%d", i); |
| 362 | break; |
| 363 | } |
| 364 | } |
| 365 | if (i < 0) { |
| 366 | ALOGD("Zip: EOCD not found, %s is not zip", debug_file_name); |
| 367 | return kInvalidFile; |
| 368 | } |
| 369 | |
| 370 | const off64_t eocd_offset = search_start + i; |
| 371 | const uint8_t* eocd_ptr = scan_buffer + i; |
| 372 | |
| 373 | assert(eocd_offset < file_length); |
| 374 | |
| 375 | /* |
| 376 | * Grab the CD offset and size, and the number of entries in the |
| 377 | * archive. Verify that they look reasonable. Widen dir_size and |
| 378 | * dir_offset to the file offset type. |
| 379 | */ |
| 380 | const uint16_t num_entries = get2LE(eocd_ptr + kEOCDNumEntries); |
| 381 | const off64_t dir_size = get4LE(eocd_ptr + kEOCDSize); |
| 382 | const off64_t dir_offset = get4LE(eocd_ptr + kEOCDFileOffset); |
Narayan Kamath | 4f6b499 | 2014-06-03 13:59:23 +0100 | [diff] [blame^] | 383 | const uint16_t comment_length = get2LE(eocd_ptr + kEOCDCommentLen); |
| 384 | |
| 385 | if (eocd_offset + comment_length + kEOCDCommentOffset != file_length) { |
| 386 | ALOGW("Zip: %" PRId64 " extraneous bytes at the end of the central directory", |
| 387 | (int64_t) (file_length - (eocd_offset + comment_length + kEOCDCommentOffset))); |
| 388 | return kInvalidFile; |
| 389 | } |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 390 | |
| 391 | if (dir_offset + dir_size > eocd_offset) { |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 392 | ALOGW("Zip: bad offsets (dir %" PRId64 ", size %" PRId64 ", eocd %" PRId64 ")", |
Mark Salyzyn | 96c5c99 | 2014-05-08 19:16:40 -0700 | [diff] [blame] | 393 | (int64_t)dir_offset, (int64_t)dir_size, (int64_t)eocd_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 394 | return kInvalidOffset; |
| 395 | } |
| 396 | if (num_entries == 0) { |
| 397 | ALOGW("Zip: empty archive?"); |
| 398 | return kEmptyArchive; |
| 399 | } |
| 400 | |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 401 | ALOGV("+++ num_entries=%d dir_size=%" PRId64 " dir_offset=%" PRId64, |
Mark Salyzyn | 96c5c99 | 2014-05-08 19:16:40 -0700 | [diff] [blame] | 402 | num_entries, (int64_t)dir_size, (int64_t)dir_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 403 | |
| 404 | /* |
| 405 | * It all looks good. Create a mapping for the CD, and set the fields |
| 406 | * in archive. |
| 407 | */ |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 408 | android::FileMap* map = MapFileSegment(fd, dir_offset, dir_size, |
| 409 | true /* read only */, debug_file_name); |
| 410 | if (map == NULL) { |
| 411 | archive->directory_map = NULL; |
| 412 | return kMmapFailed; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 413 | } |
| 414 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 415 | archive->directory_map = map; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 416 | archive->num_entries = num_entries; |
| 417 | archive->directory_offset = dir_offset; |
| 418 | |
| 419 | return 0; |
| 420 | } |
| 421 | |
| 422 | /* |
| 423 | * Find the zip Central Directory and memory-map it. |
| 424 | * |
| 425 | * On success, returns 0 after populating fields from the EOCD area: |
| 426 | * directory_offset |
| 427 | * directory_map |
| 428 | * num_entries |
| 429 | */ |
| 430 | static int32_t MapCentralDirectory(int fd, const char* debug_file_name, |
| 431 | ZipArchive* archive) { |
| 432 | |
| 433 | // Test file length. We use lseek64 to make sure the file |
| 434 | // is small enough to be a zip file (Its size must be less than |
| 435 | // 0xffffffff bytes). |
| 436 | off64_t file_length = lseek64(fd, 0, SEEK_END); |
| 437 | if (file_length == -1) { |
| 438 | ALOGV("Zip: lseek on fd %d failed", fd); |
| 439 | return kInvalidFile; |
| 440 | } |
| 441 | |
| 442 | if (file_length > (off64_t) 0xffffffff) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 443 | ALOGV("Zip: zip file too long %" PRId64, (int64_t)file_length); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 444 | return kInvalidFile; |
| 445 | } |
| 446 | |
| 447 | if (file_length < (int64_t) kEOCDLen) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 448 | ALOGV("Zip: length %" PRId64 " is too small to be zip", (int64_t)file_length); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 449 | return kInvalidFile; |
| 450 | } |
| 451 | |
| 452 | /* |
| 453 | * Perform the traditional EOCD snipe hunt. |
| 454 | * |
| 455 | * We're searching for the End of Central Directory magic number, |
| 456 | * which appears at the start of the EOCD block. It's followed by |
| 457 | * 18 bytes of EOCD stuff and up to 64KB of archive comment. We |
| 458 | * need to read the last part of the file into a buffer, dig through |
| 459 | * it to find the magic number, parse some values out, and use those |
| 460 | * to determine the extent of the CD. |
| 461 | * |
| 462 | * We start by pulling in the last part of the file. |
| 463 | */ |
| 464 | uint32_t read_amount = kMaxEOCDSearch; |
| 465 | if (file_length < (off64_t) read_amount) { |
| 466 | read_amount = file_length; |
| 467 | } |
| 468 | |
| 469 | uint8_t* scan_buffer = (uint8_t*) malloc(read_amount); |
| 470 | int32_t result = MapCentralDirectory0(fd, debug_file_name, archive, |
| 471 | file_length, read_amount, scan_buffer); |
| 472 | |
| 473 | free(scan_buffer); |
| 474 | return result; |
| 475 | } |
| 476 | |
| 477 | /* |
| 478 | * Parses the Zip archive's Central Directory. Allocates and populates the |
| 479 | * hash table. |
| 480 | * |
| 481 | * Returns 0 on success. |
| 482 | */ |
| 483 | static int32_t ParseZipArchive(ZipArchive* archive) { |
| 484 | int32_t result = -1; |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 485 | const uint8_t* cd_ptr = (const uint8_t*) archive->directory_map->getDataPtr(); |
| 486 | size_t cd_length = archive->directory_map->getDataLength(); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 487 | uint16_t num_entries = archive->num_entries; |
| 488 | |
| 489 | /* |
| 490 | * Create hash table. We have a minimum 75% load factor, possibly as |
| 491 | * low as 50% after we round off to a power of 2. There must be at |
| 492 | * least one unused entry to avoid an infinite loop during creation. |
| 493 | */ |
| 494 | archive->hash_table_size = RoundUpPower2(1 + (num_entries * 4) / 3); |
| 495 | archive->hash_table = (ZipEntryName*) calloc(archive->hash_table_size, |
| 496 | sizeof(ZipEntryName)); |
| 497 | |
| 498 | /* |
| 499 | * Walk through the central directory, adding entries to the hash |
| 500 | * table and verifying values. |
| 501 | */ |
| 502 | const uint8_t* ptr = cd_ptr; |
| 503 | for (uint16_t i = 0; i < num_entries; i++) { |
| 504 | if (get4LE(ptr) != kCDESignature) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 505 | ALOGW("Zip: missed a central dir sig (at %" PRIu16 ")", i); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 506 | goto bail; |
| 507 | } |
| 508 | |
| 509 | if (ptr + kCDELen > cd_ptr + cd_length) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 510 | ALOGW("Zip: ran off the end (at %" PRIu16 ")", i); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 511 | goto bail; |
| 512 | } |
| 513 | |
| 514 | const off64_t local_header_offset = get4LE(ptr + kCDELocalOffset); |
| 515 | if (local_header_offset >= archive->directory_offset) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 516 | ALOGW("Zip: bad LFH offset %" PRId64 " at entry %" PRIu16, (int64_t)local_header_offset, i); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 517 | goto bail; |
| 518 | } |
| 519 | |
| 520 | const uint16_t file_name_length = get2LE(ptr + kCDENameLen); |
| 521 | const uint16_t extra_length = get2LE(ptr + kCDEExtraLen); |
| 522 | const uint16_t comment_length = get2LE(ptr + kCDECommentLen); |
| 523 | |
| 524 | /* add the CDE filename to the hash table */ |
| 525 | const int add_result = AddToHash(archive->hash_table, |
| 526 | archive->hash_table_size, (const char*) ptr + kCDELen, file_name_length); |
| 527 | if (add_result) { |
| 528 | ALOGW("Zip: Error adding entry to hash table %d", add_result); |
| 529 | result = add_result; |
| 530 | goto bail; |
| 531 | } |
| 532 | |
| 533 | ptr += kCDELen + file_name_length + extra_length + comment_length; |
| 534 | if ((size_t)(ptr - cd_ptr) > cd_length) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 535 | ALOGW("Zip: bad CD advance (%tu vs %zu) at entry %" PRIu16, |
| 536 | ptr - cd_ptr, cd_length, i); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 537 | goto bail; |
| 538 | } |
| 539 | } |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 540 | ALOGV("+++ zip good scan %" PRIu16 " entries", num_entries); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 541 | |
| 542 | result = 0; |
| 543 | |
| 544 | bail: |
| 545 | return result; |
| 546 | } |
| 547 | |
| 548 | static int32_t OpenArchiveInternal(ZipArchive* archive, |
| 549 | const char* debug_file_name) { |
| 550 | int32_t result = -1; |
| 551 | if ((result = MapCentralDirectory(archive->fd, debug_file_name, archive))) { |
| 552 | return result; |
| 553 | } |
| 554 | |
| 555 | if ((result = ParseZipArchive(archive))) { |
| 556 | return result; |
| 557 | } |
| 558 | |
| 559 | return 0; |
| 560 | } |
| 561 | |
| 562 | int32_t OpenArchiveFd(int fd, const char* debug_file_name, |
| 563 | ZipArchiveHandle* handle) { |
| 564 | ZipArchive* archive = (ZipArchive*) malloc(sizeof(ZipArchive)); |
| 565 | memset(archive, 0, sizeof(*archive)); |
| 566 | *handle = archive; |
| 567 | |
| 568 | archive->fd = fd; |
| 569 | |
| 570 | return OpenArchiveInternal(archive, debug_file_name); |
| 571 | } |
| 572 | |
| 573 | int32_t OpenArchive(const char* fileName, ZipArchiveHandle* handle) { |
| 574 | ZipArchive* archive = (ZipArchive*) malloc(sizeof(ZipArchive)); |
| 575 | memset(archive, 0, sizeof(*archive)); |
| 576 | *handle = archive; |
| 577 | |
| 578 | const int fd = open(fileName, O_RDONLY | O_BINARY, 0); |
| 579 | if (fd < 0) { |
| 580 | ALOGW("Unable to open '%s': %s", fileName, strerror(errno)); |
| 581 | return kIoError; |
| 582 | } else { |
| 583 | archive->fd = fd; |
| 584 | } |
| 585 | |
| 586 | return OpenArchiveInternal(archive, fileName); |
| 587 | } |
| 588 | |
| 589 | /* |
| 590 | * Close a ZipArchive, closing the file and freeing the contents. |
| 591 | */ |
| 592 | void CloseArchive(ZipArchiveHandle handle) { |
| 593 | ZipArchive* archive = (ZipArchive*) handle; |
| 594 | ALOGV("Closing archive %p", archive); |
| 595 | |
| 596 | if (archive->fd >= 0) { |
| 597 | close(archive->fd); |
| 598 | } |
| 599 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 600 | if (archive->directory_map != NULL) { |
| 601 | archive->directory_map->release(); |
| 602 | } |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 603 | free(archive->hash_table); |
Mathieu Chartier | 5f98b12 | 2014-03-04 17:39:38 -0800 | [diff] [blame] | 604 | free(archive); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 605 | } |
| 606 | |
| 607 | static int32_t UpdateEntryFromDataDescriptor(int fd, |
| 608 | ZipEntry *entry) { |
| 609 | uint8_t ddBuf[kDDMaxLen]; |
| 610 | ssize_t actual = TEMP_FAILURE_RETRY(read(fd, ddBuf, sizeof(ddBuf))); |
| 611 | if (actual != sizeof(ddBuf)) { |
| 612 | return kIoError; |
| 613 | } |
| 614 | |
| 615 | const uint32_t ddSignature = get4LE(ddBuf); |
| 616 | uint16_t ddOffset = 0; |
| 617 | if (ddSignature == kDDOptSignature) { |
| 618 | ddOffset = 4; |
| 619 | } |
| 620 | |
| 621 | entry->crc32 = get4LE(ddBuf + ddOffset + kDDCrc32); |
| 622 | entry->compressed_length = get4LE(ddBuf + ddOffset + kDDCompLen); |
| 623 | entry->uncompressed_length = get4LE(ddBuf + ddOffset + kDDUncompLen); |
| 624 | |
| 625 | return 0; |
| 626 | } |
| 627 | |
| 628 | // Attempts to read |len| bytes into |buf| at offset |off|. |
| 629 | // |
| 630 | // This method uses pread64 on platforms that support it and |
| 631 | // lseek64 + read on platforms that don't. This implies that |
| 632 | // callers should not rely on the |fd| offset being incremented |
| 633 | // as a side effect of this call. |
| 634 | static inline ssize_t ReadAtOffset(int fd, uint8_t* buf, size_t len, |
| 635 | off64_t off) { |
| 636 | #ifdef HAVE_PREAD |
| 637 | return TEMP_FAILURE_RETRY(pread64(fd, buf, len, off)); |
| 638 | #else |
| 639 | // The only supported platform that doesn't support pread at the moment |
| 640 | // is Windows. Only recent versions of windows support unix like forks, |
| 641 | // and even there the semantics are quite different. |
| 642 | if (lseek64(fd, off, SEEK_SET) != off) { |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 643 | ALOGW("Zip: failed seek to offset %" PRId64, off); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 644 | return kIoError; |
| 645 | } |
| 646 | |
| 647 | return TEMP_FAILURE_RETRY(read(fd, buf, len)); |
| 648 | #endif // HAVE_PREAD |
| 649 | } |
| 650 | |
| 651 | static int32_t FindEntry(const ZipArchive* archive, const int ent, |
| 652 | ZipEntry* data) { |
| 653 | const uint16_t nameLen = archive->hash_table[ent].name_length; |
| 654 | const char* name = archive->hash_table[ent].name; |
| 655 | |
| 656 | // Recover the start of the central directory entry from the filename |
| 657 | // pointer. The filename is the first entry past the fixed-size data, |
| 658 | // so we can just subtract back from that. |
| 659 | const unsigned char* ptr = (const unsigned char*) name; |
| 660 | ptr -= kCDELen; |
| 661 | |
| 662 | // This is the base of our mmapped region, we have to sanity check that |
| 663 | // the name that's in the hash table is a pointer to a location within |
| 664 | // this mapped region. |
| 665 | const unsigned char* base_ptr = (const unsigned char*) |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 666 | archive->directory_map->getDataPtr(); |
| 667 | if (ptr < base_ptr || ptr > base_ptr + archive->directory_map->getDataLength()) { |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 668 | ALOGW("Zip: Invalid entry pointer"); |
| 669 | return kInvalidOffset; |
| 670 | } |
| 671 | |
| 672 | // The offset of the start of the central directory in the zipfile. |
| 673 | // We keep this lying around so that we can sanity check all our lengths |
| 674 | // and our per-file structures. |
| 675 | const off64_t cd_offset = archive->directory_offset; |
| 676 | |
| 677 | // Fill out the compression method, modification time, crc32 |
| 678 | // and other interesting attributes from the central directory. These |
| 679 | // will later be compared against values from the local file header. |
| 680 | data->method = get2LE(ptr + kCDEMethod); |
| 681 | data->mod_time = get4LE(ptr + kCDEModWhen); |
| 682 | data->crc32 = get4LE(ptr + kCDECRC); |
| 683 | data->compressed_length = get4LE(ptr + kCDECompLen); |
| 684 | data->uncompressed_length = get4LE(ptr + kCDEUncompLen); |
| 685 | |
| 686 | // Figure out the local header offset from the central directory. The |
| 687 | // actual file data will begin after the local header and the name / |
| 688 | // extra comments. |
| 689 | const off64_t local_header_offset = get4LE(ptr + kCDELocalOffset); |
| 690 | if (local_header_offset + (off64_t) kLFHLen >= cd_offset) { |
| 691 | ALOGW("Zip: bad local hdr offset in zip"); |
| 692 | return kInvalidOffset; |
| 693 | } |
| 694 | |
| 695 | uint8_t lfh_buf[kLFHLen]; |
| 696 | ssize_t actual = ReadAtOffset(archive->fd, lfh_buf, sizeof(lfh_buf), |
| 697 | local_header_offset); |
| 698 | if (actual != sizeof(lfh_buf)) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 699 | ALOGW("Zip: failed reading lfh name from offset %" PRId64, (int64_t)local_header_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 700 | return kIoError; |
| 701 | } |
| 702 | |
| 703 | if (get4LE(lfh_buf) != kLFHSignature) { |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 704 | ALOGW("Zip: didn't find signature at start of lfh, offset=%" PRId64, |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 705 | (int64_t)local_header_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 706 | return kInvalidOffset; |
| 707 | } |
| 708 | |
| 709 | // Paranoia: Match the values specified in the local file header |
| 710 | // to those specified in the central directory. |
| 711 | const uint16_t lfhGpbFlags = get2LE(lfh_buf + kLFHGPBFlags); |
| 712 | const uint16_t lfhNameLen = get2LE(lfh_buf + kLFHNameLen); |
| 713 | const uint16_t lfhExtraLen = get2LE(lfh_buf + kLFHExtraLen); |
| 714 | |
| 715 | if ((lfhGpbFlags & kGPBDDFlagMask) == 0) { |
| 716 | const uint32_t lfhCrc = get4LE(lfh_buf + kLFHCRC); |
| 717 | const uint32_t lfhCompLen = get4LE(lfh_buf + kLFHCompLen); |
| 718 | const uint32_t lfhUncompLen = get4LE(lfh_buf + kLFHUncompLen); |
| 719 | |
| 720 | data->has_data_descriptor = 0; |
| 721 | if (data->compressed_length != lfhCompLen || data->uncompressed_length != lfhUncompLen |
| 722 | || data->crc32 != lfhCrc) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 723 | ALOGW("Zip: size/crc32 mismatch. expected {%" PRIu32 ", %" PRIu32 |
| 724 | ", %" PRIx32 "}, was {%" PRIu32 ", %" PRIu32 ", %" PRIx32 "}", |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 725 | data->compressed_length, data->uncompressed_length, data->crc32, |
| 726 | lfhCompLen, lfhUncompLen, lfhCrc); |
| 727 | return kInconsistentInformation; |
| 728 | } |
| 729 | } else { |
| 730 | data->has_data_descriptor = 1; |
| 731 | } |
| 732 | |
| 733 | // Check that the local file header name matches the declared |
| 734 | // name in the central directory. |
| 735 | if (lfhNameLen == nameLen) { |
| 736 | const off64_t name_offset = local_header_offset + kLFHLen; |
| 737 | if (name_offset + lfhNameLen >= cd_offset) { |
| 738 | ALOGW("Zip: Invalid declared length"); |
| 739 | return kInvalidOffset; |
| 740 | } |
| 741 | |
| 742 | uint8_t* name_buf = (uint8_t*) malloc(nameLen); |
| 743 | ssize_t actual = ReadAtOffset(archive->fd, name_buf, nameLen, |
| 744 | name_offset); |
| 745 | |
| 746 | if (actual != nameLen) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 747 | ALOGW("Zip: failed reading lfh name from offset %" PRId64, (int64_t)name_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 748 | free(name_buf); |
| 749 | return kIoError; |
| 750 | } |
| 751 | |
| 752 | if (memcmp(name, name_buf, nameLen)) { |
| 753 | free(name_buf); |
| 754 | return kInconsistentInformation; |
| 755 | } |
| 756 | |
| 757 | free(name_buf); |
| 758 | } else { |
| 759 | ALOGW("Zip: lfh name did not match central directory."); |
| 760 | return kInconsistentInformation; |
| 761 | } |
| 762 | |
| 763 | const off64_t data_offset = local_header_offset + kLFHLen + lfhNameLen + lfhExtraLen; |
Narayan Kamath | 48953a1 | 2014-01-24 12:32:39 +0000 | [diff] [blame] | 764 | if (data_offset > cd_offset) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 765 | ALOGW("Zip: bad data offset %" PRId64 " in zip", (int64_t)data_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 766 | return kInvalidOffset; |
| 767 | } |
| 768 | |
| 769 | if ((off64_t)(data_offset + data->compressed_length) > cd_offset) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 770 | ALOGW("Zip: bad compressed length in zip (%" PRId64 " + %" PRIu32 " > %" PRId64 ")", |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 771 | (int64_t)data_offset, data->compressed_length, (int64_t)cd_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 772 | return kInvalidOffset; |
| 773 | } |
| 774 | |
| 775 | if (data->method == kCompressStored && |
| 776 | (off64_t)(data_offset + data->uncompressed_length) > cd_offset) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 777 | ALOGW("Zip: bad uncompressed length in zip (%" PRId64 " + %" PRIu32 " > %" PRId64 ")", |
Mark Salyzyn | 96c5c99 | 2014-05-08 19:16:40 -0700 | [diff] [blame] | 778 | (int64_t)data_offset, data->uncompressed_length, (int64_t)cd_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 779 | return kInvalidOffset; |
| 780 | } |
| 781 | |
| 782 | data->offset = data_offset; |
| 783 | return 0; |
| 784 | } |
| 785 | |
| 786 | struct IterationHandle { |
| 787 | uint32_t position; |
| 788 | const char* prefix; |
| 789 | uint16_t prefix_len; |
| 790 | ZipArchive* archive; |
| 791 | }; |
| 792 | |
| 793 | int32_t StartIteration(ZipArchiveHandle handle, void** cookie_ptr, const char* prefix) { |
| 794 | ZipArchive* archive = (ZipArchive *) handle; |
| 795 | |
| 796 | if (archive == NULL || archive->hash_table == NULL) { |
| 797 | ALOGW("Zip: Invalid ZipArchiveHandle"); |
| 798 | return kInvalidHandle; |
| 799 | } |
| 800 | |
| 801 | IterationHandle* cookie = (IterationHandle*) malloc(sizeof(IterationHandle)); |
| 802 | cookie->position = 0; |
| 803 | cookie->prefix = prefix; |
| 804 | cookie->archive = archive; |
| 805 | if (prefix != NULL) { |
| 806 | cookie->prefix_len = strlen(prefix); |
| 807 | } |
| 808 | |
| 809 | *cookie_ptr = cookie ; |
| 810 | return 0; |
| 811 | } |
| 812 | |
| 813 | int32_t FindEntry(const ZipArchiveHandle handle, const char* entryName, |
| 814 | ZipEntry* data) { |
| 815 | const ZipArchive* archive = (ZipArchive*) handle; |
| 816 | const int nameLen = strlen(entryName); |
| 817 | if (nameLen == 0 || nameLen > 65535) { |
| 818 | ALOGW("Zip: Invalid filename %s", entryName); |
| 819 | return kInvalidEntryName; |
| 820 | } |
| 821 | |
| 822 | const int64_t ent = EntryToIndex(archive->hash_table, |
| 823 | archive->hash_table_size, entryName, nameLen); |
| 824 | |
| 825 | if (ent < 0) { |
Narayan Kamath | a1ff801 | 2013-12-31 10:27:59 +0000 | [diff] [blame] | 826 | ALOGV("Zip: Could not find entry %.*s", nameLen, entryName); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 827 | return ent; |
| 828 | } |
| 829 | |
| 830 | return FindEntry(archive, ent, data); |
| 831 | } |
| 832 | |
| 833 | int32_t Next(void* cookie, ZipEntry* data, ZipEntryName* name) { |
| 834 | IterationHandle* handle = (IterationHandle *) cookie; |
| 835 | if (handle == NULL) { |
| 836 | return kInvalidHandle; |
| 837 | } |
| 838 | |
| 839 | ZipArchive* archive = handle->archive; |
| 840 | if (archive == NULL || archive->hash_table == NULL) { |
| 841 | ALOGW("Zip: Invalid ZipArchiveHandle"); |
| 842 | return kInvalidHandle; |
| 843 | } |
| 844 | |
| 845 | const uint32_t currentOffset = handle->position; |
| 846 | const uint32_t hash_table_length = archive->hash_table_size; |
| 847 | const ZipEntryName *hash_table = archive->hash_table; |
| 848 | |
| 849 | for (uint32_t i = currentOffset; i < hash_table_length; ++i) { |
| 850 | if (hash_table[i].name != NULL && |
| 851 | (handle->prefix == NULL || |
| 852 | (memcmp(handle->prefix, hash_table[i].name, handle->prefix_len) == 0))) { |
| 853 | handle->position = (i + 1); |
| 854 | const int error = FindEntry(archive, i, data); |
| 855 | if (!error) { |
| 856 | name->name = hash_table[i].name; |
| 857 | name->name_length = hash_table[i].name_length; |
| 858 | } |
| 859 | |
| 860 | return error; |
| 861 | } |
| 862 | } |
| 863 | |
| 864 | handle->position = 0; |
| 865 | return kIterationEnd; |
| 866 | } |
| 867 | |
| 868 | static int32_t InflateToFile(int fd, const ZipEntry* entry, |
| 869 | uint8_t* begin, uint32_t length, |
| 870 | uint64_t* crc_out) { |
| 871 | int32_t result = -1; |
| 872 | const uint32_t kBufSize = 32768; |
| 873 | uint8_t read_buf[kBufSize]; |
| 874 | uint8_t write_buf[kBufSize]; |
| 875 | z_stream zstream; |
| 876 | int zerr; |
| 877 | |
| 878 | /* |
| 879 | * Initialize the zlib stream struct. |
| 880 | */ |
| 881 | memset(&zstream, 0, sizeof(zstream)); |
| 882 | zstream.zalloc = Z_NULL; |
| 883 | zstream.zfree = Z_NULL; |
| 884 | zstream.opaque = Z_NULL; |
| 885 | zstream.next_in = NULL; |
| 886 | zstream.avail_in = 0; |
| 887 | zstream.next_out = (Bytef*) write_buf; |
| 888 | zstream.avail_out = kBufSize; |
| 889 | zstream.data_type = Z_UNKNOWN; |
| 890 | |
| 891 | /* |
| 892 | * Use the undocumented "negative window bits" feature to tell zlib |
| 893 | * that there's no zlib header waiting for it. |
| 894 | */ |
| 895 | zerr = inflateInit2(&zstream, -MAX_WBITS); |
| 896 | if (zerr != Z_OK) { |
| 897 | if (zerr == Z_VERSION_ERROR) { |
| 898 | ALOGE("Installed zlib is not compatible with linked version (%s)", |
| 899 | ZLIB_VERSION); |
| 900 | } else { |
| 901 | ALOGW("Call to inflateInit2 failed (zerr=%d)", zerr); |
| 902 | } |
| 903 | |
| 904 | return kZlibError; |
| 905 | } |
| 906 | |
| 907 | const uint32_t uncompressed_length = entry->uncompressed_length; |
| 908 | |
| 909 | uint32_t compressed_length = entry->compressed_length; |
| 910 | uint32_t write_count = 0; |
| 911 | do { |
| 912 | /* read as much as we can */ |
| 913 | if (zstream.avail_in == 0) { |
Mark Salyzyn | 51d562d | 2014-05-05 14:38:05 -0700 | [diff] [blame] | 914 | const ZD_TYPE getSize = (compressed_length > kBufSize) ? kBufSize : compressed_length; |
| 915 | const ZD_TYPE actual = TEMP_FAILURE_RETRY(read(fd, read_buf, getSize)); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 916 | if (actual != getSize) { |
Mark Salyzyn | 51d562d | 2014-05-05 14:38:05 -0700 | [diff] [blame] | 917 | ALOGW("Zip: inflate read failed (" ZD " vs " ZD ")", actual, getSize); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 918 | result = kIoError; |
| 919 | goto z_bail; |
| 920 | } |
| 921 | |
| 922 | compressed_length -= getSize; |
| 923 | |
| 924 | zstream.next_in = read_buf; |
| 925 | zstream.avail_in = getSize; |
| 926 | } |
| 927 | |
| 928 | /* uncompress the data */ |
| 929 | zerr = inflate(&zstream, Z_NO_FLUSH); |
| 930 | if (zerr != Z_OK && zerr != Z_STREAM_END) { |
| 931 | ALOGW("Zip: inflate zerr=%d (nIn=%p aIn=%u nOut=%p aOut=%u)", |
| 932 | zerr, zstream.next_in, zstream.avail_in, |
| 933 | zstream.next_out, zstream.avail_out); |
| 934 | result = kZlibError; |
| 935 | goto z_bail; |
| 936 | } |
| 937 | |
| 938 | /* write when we're full or when we're done */ |
| 939 | if (zstream.avail_out == 0 || |
| 940 | (zerr == Z_STREAM_END && zstream.avail_out != kBufSize)) { |
| 941 | const size_t write_size = zstream.next_out - write_buf; |
| 942 | // The file might have declared a bogus length. |
| 943 | if (write_size + write_count > length) { |
| 944 | goto z_bail; |
| 945 | } |
| 946 | memcpy(begin + write_count, write_buf, write_size); |
| 947 | write_count += write_size; |
| 948 | |
| 949 | zstream.next_out = write_buf; |
| 950 | zstream.avail_out = kBufSize; |
| 951 | } |
| 952 | } while (zerr == Z_OK); |
| 953 | |
| 954 | assert(zerr == Z_STREAM_END); /* other errors should've been caught */ |
| 955 | |
| 956 | // stream.adler holds the crc32 value for such streams. |
| 957 | *crc_out = zstream.adler; |
| 958 | |
| 959 | if (zstream.total_out != uncompressed_length || compressed_length != 0) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 960 | ALOGW("Zip: size mismatch on inflated file (%lu vs %" PRIu32 ")", |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 961 | zstream.total_out, uncompressed_length); |
| 962 | result = kInconsistentInformation; |
| 963 | goto z_bail; |
| 964 | } |
| 965 | |
| 966 | result = 0; |
| 967 | |
| 968 | z_bail: |
| 969 | inflateEnd(&zstream); /* free up any allocated structures */ |
| 970 | |
| 971 | return result; |
| 972 | } |
| 973 | |
| 974 | int32_t ExtractToMemory(ZipArchiveHandle handle, |
| 975 | ZipEntry* entry, uint8_t* begin, uint32_t size) { |
| 976 | ZipArchive* archive = (ZipArchive*) handle; |
| 977 | const uint16_t method = entry->method; |
| 978 | off64_t data_offset = entry->offset; |
| 979 | |
| 980 | if (lseek64(archive->fd, data_offset, SEEK_SET) != data_offset) { |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 981 | ALOGW("Zip: lseek to data at %" PRId64 " failed", (int64_t)data_offset); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 982 | return kIoError; |
| 983 | } |
| 984 | |
| 985 | // this should default to kUnknownCompressionMethod. |
| 986 | int32_t return_value = -1; |
| 987 | uint64_t crc = 0; |
| 988 | if (method == kCompressStored) { |
| 989 | return_value = CopyFileToFile(archive->fd, begin, size, &crc); |
| 990 | } else if (method == kCompressDeflated) { |
| 991 | return_value = InflateToFile(archive->fd, entry, begin, size, &crc); |
| 992 | } |
| 993 | |
| 994 | if (!return_value && entry->has_data_descriptor) { |
| 995 | return_value = UpdateEntryFromDataDescriptor(archive->fd, entry); |
| 996 | if (return_value) { |
| 997 | return return_value; |
| 998 | } |
| 999 | } |
| 1000 | |
| 1001 | // TODO: Fix this check by passing the right flags to inflate2 so that |
| 1002 | // it calculates the CRC for us. |
| 1003 | if (entry->crc32 != crc && false) { |
Mark Salyzyn | 088bf90 | 2014-05-08 16:02:20 -0700 | [diff] [blame] | 1004 | ALOGW("Zip: crc mismatch: expected %" PRIu32 ", was %" PRIu64, entry->crc32, crc); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 1005 | return kInconsistentInformation; |
| 1006 | } |
| 1007 | |
| 1008 | return return_value; |
| 1009 | } |
| 1010 | |
| 1011 | int32_t ExtractEntryToFile(ZipArchiveHandle handle, |
| 1012 | ZipEntry* entry, int fd) { |
| 1013 | const int32_t declared_length = entry->uncompressed_length; |
| 1014 | |
Narayan Kamath | 00a258c | 2013-12-13 16:06:19 +0000 | [diff] [blame] | 1015 | const off64_t current_offset = lseek64(fd, 0, SEEK_CUR); |
| 1016 | if (current_offset == -1) { |
| 1017 | ALOGW("Zip: unable to seek to current location on fd %d: %s", fd, |
| 1018 | strerror(errno)); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 1019 | return kIoError; |
| 1020 | } |
| 1021 | |
Narayan Kamath | 00a258c | 2013-12-13 16:06:19 +0000 | [diff] [blame] | 1022 | int result = TEMP_FAILURE_RETRY(ftruncate(fd, declared_length + current_offset)); |
| 1023 | if (result == -1) { |
Mark Salyzyn | 99ef991 | 2014-03-14 14:26:22 -0700 | [diff] [blame] | 1024 | ALOGW("Zip: unable to truncate file to %" PRId64 ": %s", |
Mark Salyzyn | 56a90a0 | 2014-05-08 17:20:55 -0700 | [diff] [blame] | 1025 | (int64_t)(declared_length + current_offset), strerror(errno)); |
Narayan Kamath | 00a258c | 2013-12-13 16:06:19 +0000 | [diff] [blame] | 1026 | return kIoError; |
| 1027 | } |
| 1028 | |
Narayan Kamath | 48953a1 | 2014-01-24 12:32:39 +0000 | [diff] [blame] | 1029 | // Don't attempt to map a region of length 0. We still need the |
| 1030 | // ftruncate() though, since the API guarantees that we will truncate |
| 1031 | // the file to the end of the uncompressed output. |
| 1032 | if (declared_length == 0) { |
| 1033 | return 0; |
| 1034 | } |
| 1035 | |
Narayan Kamath | 00a258c | 2013-12-13 16:06:19 +0000 | [diff] [blame] | 1036 | android::FileMap* map = MapFileSegment(fd, current_offset, declared_length, |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 1037 | false, kTempMappingFileName); |
| 1038 | if (map == NULL) { |
| 1039 | return kMmapFailed; |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 1040 | } |
| 1041 | |
Narayan Kamath | eaf9885 | 2013-12-11 14:51:51 +0000 | [diff] [blame] | 1042 | const int32_t error = ExtractToMemory(handle, entry, |
| 1043 | reinterpret_cast<uint8_t*>(map->getDataPtr()), |
| 1044 | map->getDataLength()); |
| 1045 | map->release(); |
Narayan Kamath | 7462f02 | 2013-11-21 13:05:04 +0000 | [diff] [blame] | 1046 | return error; |
| 1047 | } |
| 1048 | |
| 1049 | const char* ErrorCodeString(int32_t error_code) { |
| 1050 | if (error_code > kErrorMessageLowerBound && error_code < kErrorMessageUpperBound) { |
| 1051 | return kErrorMessages[error_code * -1]; |
| 1052 | } |
| 1053 | |
| 1054 | return kErrorMessages[0]; |
| 1055 | } |
| 1056 | |
| 1057 | int GetFileDescriptor(const ZipArchiveHandle handle) { |
| 1058 | return ((ZipArchive*) handle)->fd; |
| 1059 | } |
| 1060 | |