1 /////////////////////////////////////////////////////////////////////////////// 2 // 3 /// \file stream_decoder.c 4 /// \brief Decodes .xz Streams 5 // 6 // Author: Lasse Collin 7 // 8 // This file has been put into the public domain. 9 // You can do whatever you want with this file. 10 // 11 /////////////////////////////////////////////////////////////////////////////// 12 13 #include "stream_decoder.h" 14 #include "block_decoder.h" 15 16 17 typedef struct { 18 enum { 19 SEQ_STREAM_HEADER, 20 SEQ_BLOCK_HEADER, 21 SEQ_BLOCK, 22 SEQ_INDEX, 23 SEQ_STREAM_FOOTER, 24 SEQ_STREAM_PADDING, 25 } sequence; 26 27 /// Block or Metadata decoder. This takes little memory and the same 28 /// data structure can be used to decode every Block Header, so it's 29 /// a good idea to have a separate lzma_next_coder structure for it. 30 lzma_next_coder block_decoder; 31 32 /// Block options decoded by the Block Header decoder and used by 33 /// the Block decoder. 34 lzma_block block_options; 35 36 /// Stream Flags from Stream Header 37 lzma_stream_flags stream_flags; 38 39 /// Index is hashed so that it can be compared to the sizes of Blocks 40 /// with O(1) memory usage. 41 lzma_index_hash *index_hash; 42 43 /// Memory usage limit 44 uint64_t memlimit; 45 46 /// Amount of memory actually needed (only an estimate) 47 uint64_t memusage; 48 49 /// If true, LZMA_NO_CHECK is returned if the Stream has 50 /// no integrity check. 51 bool tell_no_check; 52 53 /// If true, LZMA_UNSUPPORTED_CHECK is returned if the Stream has 54 /// an integrity check that isn't supported by this liblzma build. 55 bool tell_unsupported_check; 56 57 /// If true, LZMA_GET_CHECK is returned after decoding Stream Header. 58 bool tell_any_check; 59 60 /// If true, we will tell the Block decoder to skip calculating 61 /// and verifying the integrity check. 62 bool ignore_check; 63 64 /// If true, we will decode concatenated Streams that possibly have 65 /// Stream Padding between or after them. LZMA_STREAM_END is returned 66 /// once the application isn't giving us any new input, and we aren't 67 /// in the middle of a Stream, and possible Stream Padding is a 68 /// multiple of four bytes. 69 bool concatenated; 70 71 /// When decoding concatenated Streams, this is true as long as we 72 /// are decoding the first Stream. This is needed to avoid misleading 73 /// LZMA_FORMAT_ERROR in case the later Streams don't have valid magic 74 /// bytes. 75 bool first_stream; 76 77 /// Write position in buffer[] and position in Stream Padding 78 size_t pos; 79 80 /// Buffer to hold Stream Header, Block Header, and Stream Footer. 81 /// Block Header has biggest maximum size. 82 uint8_t buffer[LZMA_BLOCK_HEADER_SIZE_MAX]; 83 } lzma_stream_coder; 84 85 86 static lzma_ret 87 stream_decoder_reset(lzma_stream_coder *coder, const lzma_allocator *allocator) 88 { 89 // Initialize the Index hash used to verify the Index. 90 coder->index_hash = lzma_index_hash_init(coder->index_hash, allocator); 91 if (coder->index_hash == NULL) 92 return LZMA_MEM_ERROR; 93 94 // Reset the rest of the variables. 95 coder->sequence = SEQ_STREAM_HEADER; 96 coder->pos = 0; 97 98 return LZMA_OK; 99 } 100 101 102 static lzma_ret 103 stream_decode(void *coder_ptr, const lzma_allocator *allocator, 104 const uint8_t *restrict in, size_t *restrict in_pos, 105 size_t in_size, uint8_t *restrict out, 106 size_t *restrict out_pos, size_t out_size, lzma_action action) 107 { 108 lzma_stream_coder *coder = coder_ptr; 109 110 // When decoding the actual Block, it may be able to produce more 111 // output even if we don't give it any new input. 112 while (true) 113 switch (coder->sequence) { 114 case SEQ_STREAM_HEADER: { 115 // Copy the Stream Header to the internal buffer. 116 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, 117 LZMA_STREAM_HEADER_SIZE); 118 119 // Return if we didn't get the whole Stream Header yet. 120 if (coder->pos < LZMA_STREAM_HEADER_SIZE) 121 return LZMA_OK; 122 123 coder->pos = 0; 124 125 // Decode the Stream Header. 126 const lzma_ret ret = lzma_stream_header_decode( 127 &coder->stream_flags, coder->buffer); 128 if (ret != LZMA_OK) 129 return ret == LZMA_FORMAT_ERROR && !coder->first_stream 130 ? LZMA_DATA_ERROR : ret; 131 132 // If we are decoding concatenated Streams, and the later 133 // Streams have invalid Header Magic Bytes, we give 134 // LZMA_DATA_ERROR instead of LZMA_FORMAT_ERROR. 135 coder->first_stream = false; 136 137 // Copy the type of the Check so that Block Header and Block 138 // decoders see it. 139 coder->block_options.check = coder->stream_flags.check; 140 141 // Even if we return LZMA_*_CHECK below, we want 142 // to continue from Block Header decoding. 143 coder->sequence = SEQ_BLOCK_HEADER; 144 145 // Detect if there's no integrity check or if it is 146 // unsupported if those were requested by the application. 147 if (coder->tell_no_check && coder->stream_flags.check 148 == LZMA_CHECK_NONE) 149 return LZMA_NO_CHECK; 150 151 if (coder->tell_unsupported_check 152 && !lzma_check_is_supported( 153 coder->stream_flags.check)) 154 return LZMA_UNSUPPORTED_CHECK; 155 156 if (coder->tell_any_check) 157 return LZMA_GET_CHECK; 158 } 159 160 // Fall through 161 162 case SEQ_BLOCK_HEADER: { 163 if (*in_pos >= in_size) 164 return LZMA_OK; 165 166 if (coder->pos == 0) { 167 // Detect if it's Index. 168 if (in[*in_pos] == 0x00) { 169 coder->sequence = SEQ_INDEX; 170 break; 171 } 172 173 // Calculate the size of the Block Header. Note that 174 // Block Header decoder wants to see this byte too 175 // so don't advance *in_pos. 176 coder->block_options.header_size 177 = lzma_block_header_size_decode( 178 in[*in_pos]); 179 } 180 181 // Copy the Block Header to the internal buffer. 182 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, 183 coder->block_options.header_size); 184 185 // Return if we didn't get the whole Block Header yet. 186 if (coder->pos < coder->block_options.header_size) 187 return LZMA_OK; 188 189 coder->pos = 0; 190 191 // Version 1 is needed to support the .ignore_check option. 192 coder->block_options.version = 1; 193 194 // Set up a buffer to hold the filter chain. Block Header 195 // decoder will initialize all members of this array so 196 // we don't need to do it here. 197 lzma_filter filters[LZMA_FILTERS_MAX + 1]; 198 coder->block_options.filters = filters; 199 200 // Decode the Block Header. 201 return_if_error(lzma_block_header_decode(&coder->block_options, 202 allocator, coder->buffer)); 203 204 // If LZMA_IGNORE_CHECK was used, this flag needs to be set. 205 // It has to be set after lzma_block_header_decode() because 206 // it always resets this to false. 207 coder->block_options.ignore_check = coder->ignore_check; 208 209 // Check the memory usage limit. 210 const uint64_t memusage = lzma_raw_decoder_memusage(filters); 211 lzma_ret ret; 212 213 if (memusage == UINT64_MAX) { 214 // One or more unknown Filter IDs. 215 ret = LZMA_OPTIONS_ERROR; 216 } else { 217 // Now we can set coder->memusage since we know that 218 // the filter chain is valid. We don't want 219 // lzma_memusage() to return UINT64_MAX in case of 220 // invalid filter chain. 221 coder->memusage = memusage; 222 223 if (memusage > coder->memlimit) { 224 // The chain would need too much memory. 225 ret = LZMA_MEMLIMIT_ERROR; 226 } else { 227 // Memory usage is OK. 228 // Initialize the Block decoder. 229 ret = lzma_block_decoder_init( 230 &coder->block_decoder, 231 allocator, 232 &coder->block_options); 233 } 234 } 235 236 // Free the allocated filter options since they are needed 237 // only to initialize the Block decoder. 238 for (size_t i = 0; i < LZMA_FILTERS_MAX; ++i) 239 lzma_free(filters[i].options, allocator); 240 241 coder->block_options.filters = NULL; 242 243 // Check if memory usage calculation and Block enocoder 244 // initialization succeeded. 245 if (ret != LZMA_OK) 246 return ret; 247 248 coder->sequence = SEQ_BLOCK; 249 } 250 251 // Fall through 252 253 case SEQ_BLOCK: { 254 const lzma_ret ret = coder->block_decoder.code( 255 coder->block_decoder.coder, allocator, 256 in, in_pos, in_size, out, out_pos, out_size, 257 action); 258 259 if (ret != LZMA_STREAM_END) 260 return ret; 261 262 // Block decoded successfully. Add the new size pair to 263 // the Index hash. 264 return_if_error(lzma_index_hash_append(coder->index_hash, 265 lzma_block_unpadded_size( 266 &coder->block_options), 267 coder->block_options.uncompressed_size)); 268 269 coder->sequence = SEQ_BLOCK_HEADER; 270 break; 271 } 272 273 case SEQ_INDEX: { 274 // If we don't have any input, don't call 275 // lzma_index_hash_decode() since it would return 276 // LZMA_BUF_ERROR, which we must not do here. 277 if (*in_pos >= in_size) 278 return LZMA_OK; 279 280 // Decode the Index and compare it to the hash calculated 281 // from the sizes of the Blocks (if any). 282 const lzma_ret ret = lzma_index_hash_decode(coder->index_hash, 283 in, in_pos, in_size); 284 if (ret != LZMA_STREAM_END) 285 return ret; 286 287 coder->sequence = SEQ_STREAM_FOOTER; 288 } 289 290 // Fall through 291 292 case SEQ_STREAM_FOOTER: { 293 // Copy the Stream Footer to the internal buffer. 294 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos, 295 LZMA_STREAM_HEADER_SIZE); 296 297 // Return if we didn't get the whole Stream Footer yet. 298 if (coder->pos < LZMA_STREAM_HEADER_SIZE) 299 return LZMA_OK; 300 301 coder->pos = 0; 302 303 // Decode the Stream Footer. The decoder gives 304 // LZMA_FORMAT_ERROR if the magic bytes don't match, 305 // so convert that return code to LZMA_DATA_ERROR. 306 lzma_stream_flags footer_flags; 307 const lzma_ret ret = lzma_stream_footer_decode( 308 &footer_flags, coder->buffer); 309 if (ret != LZMA_OK) 310 return ret == LZMA_FORMAT_ERROR 311 ? LZMA_DATA_ERROR : ret; 312 313 // Check that Index Size stored in the Stream Footer matches 314 // the real size of the Index field. 315 if (lzma_index_hash_size(coder->index_hash) 316 != footer_flags.backward_size) 317 return LZMA_DATA_ERROR; 318 319 // Compare that the Stream Flags fields are identical in 320 // both Stream Header and Stream Footer. 321 return_if_error(lzma_stream_flags_compare( 322 &coder->stream_flags, &footer_flags)); 323 324 if (!coder->concatenated) 325 return LZMA_STREAM_END; 326 327 coder->sequence = SEQ_STREAM_PADDING; 328 } 329 330 // Fall through 331 332 case SEQ_STREAM_PADDING: 333 assert(coder->concatenated); 334 335 // Skip over possible Stream Padding. 336 while (true) { 337 if (*in_pos >= in_size) { 338 // Unless LZMA_FINISH was used, we cannot 339 // know if there's more input coming later. 340 if (action != LZMA_FINISH) 341 return LZMA_OK; 342 343 // Stream Padding must be a multiple of 344 // four bytes. 345 return coder->pos == 0 346 ? LZMA_STREAM_END 347 : LZMA_DATA_ERROR; 348 } 349 350 // If the byte is not zero, it probably indicates 351 // beginning of a new Stream (or the file is corrupt). 352 if (in[*in_pos] != 0x00) 353 break; 354 355 ++*in_pos; 356 coder->pos = (coder->pos + 1) & 3; 357 } 358 359 // Stream Padding must be a multiple of four bytes (empty 360 // Stream Padding is OK). 361 if (coder->pos != 0) { 362 ++*in_pos; 363 return LZMA_DATA_ERROR; 364 } 365 366 // Prepare to decode the next Stream. 367 return_if_error(stream_decoder_reset(coder, allocator)); 368 break; 369 370 default: 371 assert(0); 372 return LZMA_PROG_ERROR; 373 } 374 375 // Never reached 376 } 377 378 379 static void 380 stream_decoder_end(void *coder_ptr, const lzma_allocator *allocator) 381 { 382 lzma_stream_coder *coder = coder_ptr; 383 lzma_next_end(&coder->block_decoder, allocator); 384 lzma_index_hash_end(coder->index_hash, allocator); 385 lzma_free(coder, allocator); 386 return; 387 } 388 389 390 static lzma_check 391 stream_decoder_get_check(const void *coder_ptr) 392 { 393 const lzma_stream_coder *coder = coder_ptr; 394 return coder->stream_flags.check; 395 } 396 397 398 static lzma_ret 399 stream_decoder_memconfig(void *coder_ptr, uint64_t *memusage, 400 uint64_t *old_memlimit, uint64_t new_memlimit) 401 { 402 lzma_stream_coder *coder = coder_ptr; 403 404 *memusage = coder->memusage; 405 *old_memlimit = coder->memlimit; 406 407 if (new_memlimit != 0) { 408 if (new_memlimit < coder->memusage) 409 return LZMA_MEMLIMIT_ERROR; 410 411 coder->memlimit = new_memlimit; 412 } 413 414 return LZMA_OK; 415 } 416 417 418 extern lzma_ret 419 lzma_stream_decoder_init( 420 lzma_next_coder *next, const lzma_allocator *allocator, 421 uint64_t memlimit, uint32_t flags) 422 { 423 lzma_next_coder_init(&lzma_stream_decoder_init, next, allocator); 424 425 if (flags & ~LZMA_SUPPORTED_FLAGS) 426 return LZMA_OPTIONS_ERROR; 427 428 lzma_stream_coder *coder = next->coder; 429 if (coder == NULL) { 430 coder = lzma_alloc(sizeof(lzma_stream_coder), allocator); 431 if (coder == NULL) 432 return LZMA_MEM_ERROR; 433 434 next->coder = coder; 435 next->code = &stream_decode; 436 next->end = &stream_decoder_end; 437 next->get_check = &stream_decoder_get_check; 438 next->memconfig = &stream_decoder_memconfig; 439 440 coder->block_decoder = LZMA_NEXT_CODER_INIT; 441 coder->index_hash = NULL; 442 } 443 444 coder->memlimit = my_max(1, memlimit); 445 coder->memusage = LZMA_MEMUSAGE_BASE; 446 coder->tell_no_check = (flags & LZMA_TELL_NO_CHECK) != 0; 447 coder->tell_unsupported_check 448 = (flags & LZMA_TELL_UNSUPPORTED_CHECK) != 0; 449 coder->tell_any_check = (flags & LZMA_TELL_ANY_CHECK) != 0; 450 coder->ignore_check = (flags & LZMA_IGNORE_CHECK) != 0; 451 coder->concatenated = (flags & LZMA_CONCATENATED) != 0; 452 coder->first_stream = true; 453 454 return stream_decoder_reset(coder, allocator); 455 } 456 457 458 extern LZMA_API(lzma_ret) 459 lzma_stream_decoder(lzma_stream *strm, uint64_t memlimit, uint32_t flags) 460 { 461 lzma_next_strm_init(lzma_stream_decoder_init, strm, memlimit, flags); 462 463 strm->internal->supported_actions[LZMA_RUN] = true; 464 strm->internal->supported_actions[LZMA_FINISH] = true; 465 466 return LZMA_OK; 467 } 468