src/liblzma/common/stream_decoder.c

   1 ///////////////////////////////////////////////////////////////////////////////
   2 //
   3 /// \file       stream_decoder.c
   4 /// \brief      Decodes .lzma Streams
   5 //
   6 //  Copyright (C) 2007 Lasse Collin
   7 //
   8 //  This library is free software; you can redistribute it and/or
   9 //  modify it under the terms of the GNU Lesser General Public
  10 //  License as published by the Free Software Foundation; either
  11 //  version 2.1 of the License, or (at your option) any later version.
  12 //
  13 //  This library is distributed in the hope that it will be useful,
  14 //  but WITHOUT ANY WARRANTY; without even the implied warranty of
  15 //  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16 //  Lesser General Public License for more details.
  17 //
  18 ///////////////////////////////////////////////////////////////////////////////
  19
  20 #include "stream_decoder.h"
  21 #include "stream_flags_common.h"
  22 #include "check.h"
  23 #include "stream_flags_decoder.h"
  24 #include "block_decoder.h"
  25
  26
  27 struct lzma_coder_s {
  28         enum {
  29                 SEQ_STREAM_HEADER,
  30                 SEQ_BLOCK_HEADER,
  31                 SEQ_BLOCK,
  32                 SEQ_INDEX,
  33                 SEQ_STREAM_FOOTER,
  34                 SEQ_STREAM_PADDING,
  35         } sequence;
  36
  37         /// Block or Metadata decoder. This takes little memory and the same
  38         /// data structure can be used to decode every Block Header, so it's
  39         /// a good idea to have a separate lzma_next_coder structure for it.
  40         lzma_next_coder block_decoder;
  41
  42         /// Block options decoded by the Block Header decoder and used by
  43         /// the Block decoder.
  44         lzma_block block_options;
  45
  46         /// Stream Flags from Stream Header
  47         lzma_stream_flags stream_flags;
  48
  49         /// Index is hashed so that it can be compared to the sizes of Blocks
  50         /// with O(1) memory usage.
  51         lzma_index_hash *index_hash;
  52
  53         /// Memory usage limit
  54         uint64_t memlimit;
  55
  56         /// If true, LZMA_NO_CHECK is returned if the Stream has
  57         /// no integrity check.
  58         bool warn_no_check;
  59
  60         /// If true, LZMA_UNSUPPORTED_CHECK is returned if the Stream has
  61         /// an integrity check that isn't supported by this liblzma build.
  62         bool warn_unsupported_check;
  63
  64         /// If true, LZMA_SEE_CHECK is returned after decoding Stream Header.
  65         bool tell_check;
  66
  67         /// If true, we will decode concatenated Streams that possibly have
  68         /// Stream Padding between or after them. LZMA_STREAM_END is returned
  69         /// once the application isn't giving us any new input, and we aren't
  70         /// in the middle of a Stream, and possible Stream Padding is a
  71         /// multiple of four bytes.
  72         bool concatenated;
  73
  74         /// When decoding concatenated Streams, this is true as long as we
  75         /// are decoding the first Stream. This is needed to avoid misleading
  76         /// LZMA_FORMAT_ERROR in case the later Streams don't have valid magic
  77         /// bytes.
  78         bool first_stream;
  79
  80         /// Write position in buffer[] and position in Stream Padding
  81         size_t pos;
  82
  83         /// Buffer to hold Stream Header, Block Header, and Stream Footer.
  84         /// Block Header has biggest maximum size.
  85         uint8_t buffer[LZMA_BLOCK_HEADER_SIZE_MAX];
  86 };
  87
  88
  89 static lzma_ret
  90 stream_decoder_reset(lzma_coder *coder, lzma_allocator *allocator)
  91 {
  92         // Initialize the Index hash used to verify the Index.
  93         coder->index_hash = lzma_index_hash_init(coder->index_hash, allocator);
  94         if (coder->index_hash == NULL)
  95                 return LZMA_MEM_ERROR;
  96
  97         // Reset the rest of the variables.
  98         coder->sequence = SEQ_STREAM_HEADER;
  99         coder->block_options.filters = NULL;
 100         coder->pos = 0;
 101
 102         return LZMA_OK;
 103 }
 104
 105
 106 static lzma_ret
 107 stream_decode(lzma_coder *coder, lzma_allocator *allocator,
 108                 const uint8_t *restrict in, size_t *restrict in_pos,
 109                 size_t in_size, uint8_t *restrict out,
 110                 size_t *restrict out_pos, size_t out_size, lzma_action action)
 111 {
 112         // When decoding the actual Block, it may be able to produce more
 113         // output even if we don't give it any new input.
 114         while (true)
 115         switch (coder->sequence) {
 116         case SEQ_STREAM_HEADER: {
 117                 // Copy the Stream Header to the internal buffer.
 118                 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
 119                                 LZMA_STREAM_HEADER_SIZE);
 120
 121                 // Return if we didn't get the whole Stream Header yet.
 122                 if (coder->pos < LZMA_STREAM_HEADER_SIZE)
 123                         return LZMA_OK;
 124
 125                 coder->pos = 0;
 126
 127                 // Decode the Stream Header.
 128                 const lzma_ret ret = lzma_stream_header_decode(
 129                                 &coder->stream_flags, coder->buffer);
 130                 if (ret != LZMA_OK)
 131                         return ret == LZMA_FORMAT_ERROR && !coder->first_stream
 132                                         ? LZMA_DATA_ERROR : ret;
 133
 134                 // Copy the type of the Check so that Block Header and Block
 135                 // decoders see it.
 136                 coder->block_options.check = coder->stream_flags.check;
 137
 138                 // Even if we return LZMA_*_CHECK below, we want
 139                 // to continue from Block Header decoding.
 140                 coder->sequence = SEQ_BLOCK_HEADER;
 141
 142                 // Detect if there's no integrity check or if it is
 143                 // unsupported if those were requested by the application.
 144                 if (coder->warn_no_check && coder->stream_flags.check
 145                                 == LZMA_CHECK_NONE)
 146                         return LZMA_NO_CHECK;
 147
 148                 if (coder->warn_unsupported_check
 149                                 && !lzma_check_is_supported(
 150                                         coder->stream_flags.check))
 151                         return LZMA_UNSUPPORTED_CHECK;
 152
 153                 if (coder->tell_check)
 154                         return LZMA_SEE_CHECK;
 155         }
 156
 157         // Fall through
 158
 159         case SEQ_BLOCK_HEADER: {
 160                 if (*in_pos >= in_size)
 161                         return LZMA_OK;
 162
 163                 if (coder->pos == 0) {
 164                         // Detect if it's Index.
 165                         if (in[*in_pos] == 0x00) {
 166                                 coder->sequence = SEQ_INDEX;
 167                                 break;
 168                         }
 169
 170                         // Calculate the size of the Block Header. Note that
 171                         // Block Header decoder wants to see this byte too
 172                         // so don't advance *in_pos.
 173                         coder->block_options.header_size
 174                                         = lzma_block_header_size_decode(
 175                                                 in[*in_pos]);
 176                 }
 177
 178                 // Copy the Block Header to the internal buffer.
 179                 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
 180                                 coder->block_options.header_size);
 181
 182                 // Return if we didn't get the whole Block Header yet.
 183                 if (coder->pos < coder->block_options.header_size)
 184                         return LZMA_OK;
 185
 186                 coder->pos = 0;
 187
 188                 // Set up a buffer to hold the filter chain. Block Header
 189                 // decoder will initialize all members of this array so
 190                 // we don't need to do it here.
 191                 lzma_filter filters[LZMA_BLOCK_FILTERS_MAX + 1];
 192                 coder->block_options.filters = filters;
 193
 194                 // Decode the Block Header.
 195                 return_if_error(lzma_block_header_decode(&coder->block_options,
 196                                 allocator, coder->buffer));
 197
 198                 // Check the memory usage limit.
 199                 const uint64_t memusage = lzma_memusage_decoder(filters);
 200                 lzma_ret ret;
 201
 202                 if (memusage == UINT64_MAX) {
 203                         // One or more unknown Filter IDs.
 204                         ret = LZMA_HEADER_ERROR;
 205                 } else if (memusage > coder->memlimit) {
 206                         // The chain would need too much memory.
 207                         ret = LZMA_MEMLIMIT_ERROR;
 208                 } else {
 209                         // Memory usage is OK. Initialize the Block decoder.
 210                         ret = lzma_block_decoder_init(
 211                                         &coder->block_decoder,
 212                                         allocator, &coder->block_options);
 213                 }
 214
 215                 // Free the allocated filter options since they are needed
 216                 // only to initialize the Block decoder.
 217                 for (size_t i = 0; i < LZMA_BLOCK_FILTERS_MAX; ++i)
 218                         lzma_free(filters[i].options, allocator);
 219
 220                 coder->block_options.filters = NULL;
 221
 222                 // Check if memory usage calculation and Block enocoder
 223                 // initialization succeeded.
 224                 if (ret != LZMA_OK)
 225                         return ret;
 226
 227                 coder->sequence = SEQ_BLOCK;
 228         }
 229
 230         // Fall through
 231
 232         case SEQ_BLOCK: {
 233                 const lzma_ret ret = coder->block_decoder.code(
 234                                 coder->block_decoder.coder, allocator,
 235                                 in, in_pos, in_size, out, out_pos, out_size,
 236                                 action);
 237
 238                 if (ret != LZMA_STREAM_END)
 239                         return ret;
 240
 241                 // Block decoded successfully. Add the new size pair to
 242                 // the Index hash.
 243                 return_if_error(lzma_index_hash_append(coder->index_hash,
 244                                 lzma_block_total_size_get(
 245                                         &coder->block_options),
 246                                 coder->block_options.uncompressed_size));
 247
 248                 coder->sequence = SEQ_BLOCK_HEADER;
 249                 break;
 250         }
 251
 252         case SEQ_INDEX: {
 253                 // If we don't have any input, don't call
 254                 // lzma_index_hash_decode() since it would return
 255                 // LZMA_BUF_ERROR, which we must not do here.
 256                 if (*in_pos >= in_size)
 257                         return LZMA_OK;
 258
 259                 // Decode the Index and compare it to the hash calculated
 260                 // from the sizes of the Blocks (if any).
 261                 const lzma_ret ret = lzma_index_hash_decode(coder->index_hash,
 262                                 in, in_pos, in_size);
 263                 if (ret != LZMA_STREAM_END)
 264                         return ret;
 265
 266                 coder->sequence = SEQ_STREAM_FOOTER;
 267         }
 268
 269         // Fall through
 270
 271         case SEQ_STREAM_FOOTER:
 272                 // Copy the Stream Footer to the internal buffer.
 273                 lzma_bufcpy(in, in_pos, in_size, coder->buffer, &coder->pos,
 274                                 LZMA_STREAM_HEADER_SIZE);
 275
 276                 // Return if we didn't get the whole Stream Footer yet.
 277                 if (coder->pos < LZMA_STREAM_HEADER_SIZE)
 278                         return LZMA_OK;
 279
 280                 coder->pos = 0;
 281
 282                 // Decode the Stream Footer. The decoder gives
 283                 // LZMA_FORMAT_ERROR if the magic bytes don't match,
 284                 // so convert that return code to LZMA_DATA_ERROR.
 285                 lzma_stream_flags footer_flags;
 286                 const lzma_ret ret = lzma_stream_footer_decode(
 287                                 &footer_flags, coder->buffer);
 288                 if (ret != LZMA_OK)
 289                         return ret == LZMA_FORMAT_ERROR
 290                                         ? LZMA_DATA_ERROR : ret;
 291
 292                 // Check that Index Size stored in the Stream Footer matches
 293                 // the real size of the Index field.
 294                 if (lzma_index_hash_size(coder->index_hash)
 295                                 != footer_flags.backward_size)
 296                         return LZMA_DATA_ERROR;
 297
 298                 // Compare that the Stream Flags fields are identical in
 299                 // both Stream Header and Stream Footer.
 300                 if (!lzma_stream_flags_equal(&coder->stream_flags,
 301                                 &footer_flags))
 302                         return LZMA_DATA_ERROR;
 303
 304                 if (!coder->concatenated)
 305                         return LZMA_STREAM_END;
 306
 307                 coder->sequence = SEQ_STREAM_PADDING;
 308
 309         // Fall through
 310
 311         case SEQ_STREAM_PADDING:
 312                 assert(coder->concatenated);
 313
 314                 // Skip over possible Stream Padding.
 315                 while (true) {
 316                         if (*in_pos >= in_size) {
 317                                 // Unless LZMA_FINISH was used, we cannot
 318                                 // know if there's more input coming later.
 319                                 if (action != LZMA_FINISH)
 320                                         return LZMA_OK;
 321
 322                                 // Stream Padding must be a multiple of
 323                                 // four bytes.
 324                                 return coder->pos == 0
 325                                                 ? LZMA_STREAM_END
 326                                                 : LZMA_DATA_ERROR;
 327                         }
 328
 329                         // If the byte is not zero, it probably indicates
 330                         // beginning of a new Stream (or the file is corrupt).
 331                         if (in[*in_pos] != 0x00)
 332                                 break;
 333
 334                         ++*in_pos;
 335                         coder->pos = (coder->pos + 1) & 3;
 336                 }
 337
 338                 // Stream Padding must be a multiple of four bytes (empty
 339                 // Stream Padding is OK).
 340                 if (coder->pos != 0) {
 341                         ++*in_pos;
 342                         return LZMA_DATA_ERROR;
 343                 }
 344
 345                 // Prepare to decode the next Stream.
 346                 return_if_error(stream_decoder_reset(coder, allocator));
 347                 break;
 348
 349         default:
 350                 assert(0);
 351                 return LZMA_PROG_ERROR;
 352         }
 353
 354         return LZMA_OK;
 355 }
 356
 357
 358 static void
 359 stream_decoder_end(lzma_coder *coder, lzma_allocator *allocator)
 360 {
 361         lzma_next_end(&coder->block_decoder, allocator);
 362         lzma_index_hash_end(coder->index_hash, allocator);
 363         lzma_free(coder, allocator);
 364         return;
 365 }
 366
 367
 368 static lzma_check
 369 stream_decoder_see_check(const lzma_coder *coder)
 370 {
 371         return coder->stream_flags.check;
 372 }
 373
 374
 375 extern lzma_ret
 376 lzma_stream_decoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 377                 uint64_t memlimit, uint32_t flags)
 378 {
 379         lzma_next_coder_init(lzma_stream_decoder_init, next, allocator);
 380
 381         if (flags & ~LZMA_SUPPORTED_FLAGS)
 382                 return LZMA_HEADER_ERROR;
 383
 384         if (next->coder == NULL) {
 385                 next->coder = lzma_alloc(sizeof(lzma_coder), allocator);
 386                 if (next->coder == NULL)
 387                         return LZMA_MEM_ERROR;
 388
 389                 next->code = &stream_decode;
 390                 next->end = &stream_decoder_end;
 391                 next->see_check = &stream_decoder_see_check;
 392
 393                 next->coder->block_decoder = LZMA_NEXT_CODER_INIT;
 394                 next->coder->index_hash = NULL;
 395         }
 396
 397         next->coder->memlimit = memlimit;
 398         next->coder->warn_no_check = (flags & LZMA_WARN_NO_CHECK) != 0;
 399         next->coder->warn_unsupported_check
 400                         = (flags & LZMA_WARN_UNSUPPORTED_CHECK) != 0;
 401         next->coder->tell_check = (flags & LZMA_TELL_CHECK) != 0;
 402         next->coder->concatenated
 403                         = (flags & LZMA_CONCATENATED) != 0;
 404
 405         return stream_decoder_reset(next->coder, allocator);
 406 }
 407
 408
 409 extern LZMA_API lzma_ret
 410 lzma_stream_decoder(lzma_stream *strm, uint64_t memlimit, uint32_t flags)
 411 {
 412         lzma_next_strm_init(lzma_stream_decoder_init, strm, memlimit, flags);
 413
 414         strm->internal->supported_actions[LZMA_RUN] = true;
 415         strm->internal->supported_actions[LZMA_FINISH] = true;
 416
 417         return LZMA_OK;
 418 }