wimlib.net Git - wimlib/blob - src/lzx_decompress.c

   1 /*
   2  * lzx_decompress.c
   3  *
   4  * A decompressor for the LZX compression format, as used in WIM files.
   5  */
   6
   7 /*
   8  * Copyright (C) 2012-2016 Eric Biggers
   9  *
  10  * This file is free software; you can redistribute it and/or modify it under
  11  * the terms of the GNU Lesser General Public License as published by the Free
  12  * Software Foundation; either version 3 of the License, or (at your option) any
  13  * later version.
  14  *
  15  * This file is distributed in the hope that it will be useful, but WITHOUT
  16  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
  17  * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
  18  * details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public License
  21  * along with this file; if not, see http://www.gnu.org/licenses/.
  22  */
  23
  24 /*
  25  * LZX is an LZ77 and Huffman-code based compression format that has many
  26  * similarities to DEFLATE (the format used by zlib/gzip).  The compression
  27  * ratio is as good or better than DEFLATE.  See lzx_compress.c for a format
  28  * overview, and see https://en.wikipedia.org/wiki/LZX_(algorithm) for a
  29  * historical overview.  Here I make some pragmatic notes.
  30  *
  31  * The old specification for LZX is the document "Microsoft LZX Data Compression
  32  * Format" (1997).  It defines the LZX format as used in cabinet files.  Allowed
  33  * window sizes are 2^n where 15 <= n <= 21.  However, this document contains
  34  * several errors, so don't read too much into it...
  35  *
  36  * The new specification for LZX is the document "[MS-PATCH]: LZX DELTA
  37  * Compression and Decompression" (2014).  It defines the LZX format as used by
  38  * Microsoft's binary patcher.  It corrects several errors in the 1997 document
  39  * and extends the format in several ways --- namely, optional reference data,
  40  * up to 2^25 byte windows, and longer match lengths.
  41  *
  42  * WIM files use a more restricted form of LZX.  No LZX DELTA extensions are
  43  * present, the window is not "sliding", E8 preprocessing is done
  44  * unconditionally with a fixed file size, and the maximum window size is always
  45  * 2^15 bytes (equal to the size of each "chunk" in a compressed WIM resource).
  46  * This code is primarily intended to implement this form of LZX.  But although
  47  * not compatible with WIMGAPI, this code also supports maximum window sizes up
  48  * to 2^21 bytes.
  49  *
  50  * TODO: Add support for window sizes up to 2^25 bytes.
  51  */
  52
  53 #ifdef HAVE_CONFIG_H
  54 #  include "config.h"
  55 #endif
  56
  57 #include <string.h>
  58
  59 #include "wimlib/decompressor_ops.h"
  60 #include "wimlib/decompress_common.h"
  61 #include "wimlib/error.h"
  62 #include "wimlib/lzx_common.h"
  63 #include "wimlib/util.h"
  64
  65 /* These values are chosen for fast decompression.  */
  66 #define LZX_MAINCODE_TABLEBITS          11
  67 #define LZX_LENCODE_TABLEBITS           10
  68 #define LZX_PRECODE_TABLEBITS           6
  69 #define LZX_ALIGNEDCODE_TABLEBITS       7
  70
  71 #define LZX_READ_LENS_MAX_OVERRUN 50
  72
  73 /* Reusable heap-allocated memory for LZX decompression */
  74 struct lzx_decompressor {
  75
  76         /* Huffman decoding tables, and arrays that map symbols to codeword
  77          * lengths  */
  78
  79         u16 maincode_decode_table[(1 << LZX_MAINCODE_TABLEBITS) +
  80                                         (LZX_MAINCODE_MAX_NUM_SYMBOLS * 2)]
  81                                         _aligned_attribute(DECODE_TABLE_ALIGNMENT);
  82         u8 maincode_lens[LZX_MAINCODE_MAX_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
  83
  84
  85         u16 lencode_decode_table[(1 << LZX_LENCODE_TABLEBITS) +
  86                                         (LZX_LENCODE_NUM_SYMBOLS * 2)]
  87                                         _aligned_attribute(DECODE_TABLE_ALIGNMENT);
  88         u8 lencode_lens[LZX_LENCODE_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
  89
  90         union {
  91                 u16 alignedcode_decode_table[(1 << LZX_ALIGNEDCODE_TABLEBITS) +
  92                                                 (LZX_ALIGNEDCODE_NUM_SYMBOLS * 2)]
  93                                                 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
  94                 u8 alignedcode_lens[LZX_ALIGNEDCODE_NUM_SYMBOLS];
  95         };
  96
  97         union {
  98                 u16 precode_decode_table[(1 << LZX_PRECODE_TABLEBITS) +
  99                                          (LZX_PRECODE_NUM_SYMBOLS * 2)]
 100                                                 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
 101                 u8 precode_lens[LZX_PRECODE_NUM_SYMBOLS];
 102         };
 103
 104         unsigned window_order;
 105         unsigned num_main_syms;
 106 } _aligned_attribute(DECODE_TABLE_ALIGNMENT);
 107
 108 /* Read a Huffman-encoded symbol using the precode. */
 109 static inline unsigned
 110 read_presym(const struct lzx_decompressor *d, struct input_bitstream *is)
 111 {
 112         return read_huffsym(is, d->precode_decode_table,
 113                             LZX_PRECODE_TABLEBITS, LZX_MAX_PRE_CODEWORD_LEN);
 114 }
 115
 116 /* Read a Huffman-encoded symbol using the main code. */
 117 static inline unsigned
 118 read_mainsym(const struct lzx_decompressor *d, struct input_bitstream *is)
 119 {
 120         return read_huffsym(is, d->maincode_decode_table,
 121                             LZX_MAINCODE_TABLEBITS, LZX_MAX_MAIN_CODEWORD_LEN);
 122 }
 123
 124 /* Read a Huffman-encoded symbol using the length code. */
 125 static inline unsigned
 126 read_lensym(const struct lzx_decompressor *d, struct input_bitstream *is)
 127 {
 128         return read_huffsym(is, d->lencode_decode_table,
 129                             LZX_LENCODE_TABLEBITS, LZX_MAX_LEN_CODEWORD_LEN);
 130 }
 131
 132 /* Read a Huffman-encoded symbol using the aligned offset code. */
 133 static inline unsigned
 134 read_alignedsym(const struct lzx_decompressor *d, struct input_bitstream *is)
 135 {
 136         return read_huffsym(is, d->alignedcode_decode_table,
 137                             LZX_ALIGNEDCODE_TABLEBITS, LZX_MAX_ALIGNED_CODEWORD_LEN);
 138 }
 139
 140 /*
 141  * Read a precode from the compressed input bitstream, then use it to decode
 142  * @num_lens codeword length values and write them to @lens.
 143  */
 144 static int
 145 lzx_read_codeword_lens(struct lzx_decompressor *d, struct input_bitstream *is,
 146                        u8 *lens, unsigned num_lens)
 147 {
 148         u8 *len_ptr = lens;
 149         u8 *lens_end = lens + num_lens;
 150
 151         /* Read the lengths of the precode codewords, which are stored
 152          * explicitly. */
 153         for (int i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) {
 154                 d->precode_lens[i] =
 155                         bitstream_read_bits(is, LZX_PRECODE_ELEMENT_SIZE);
 156         }
 157
 158         /* Build the decoding table for the precode. */
 159         if (make_huffman_decode_table(d->precode_decode_table,
 160                                       LZX_PRECODE_NUM_SYMBOLS,
 161                                       LZX_PRECODE_TABLEBITS,
 162                                       d->precode_lens,
 163                                       LZX_MAX_PRE_CODEWORD_LEN))
 164                 return -1;
 165
 166         /* Decode the codeword lengths. */
 167         do {
 168                 unsigned presym;
 169                 u8 len;
 170
 171                 /* Read the next precode symbol. */
 172                 presym = read_presym(d, is);
 173                 if (presym < 17) {
 174                         /* Difference from old length */
 175                         len = *len_ptr - presym;
 176                         if ((s8)len < 0)
 177                                 len += 17;
 178                         *len_ptr++ = len;
 179                 } else {
 180                         /* Special RLE values */
 181
 182                         unsigned run_len;
 183
 184                         if (presym == 17) {
 185                                 /* Run of 0's */
 186                                 run_len = 4 + bitstream_read_bits(is, 4);
 187                                 len = 0;
 188                         } else if (presym == 18) {
 189                                 /* Longer run of 0's */
 190                                 run_len = 20 + bitstream_read_bits(is, 5);
 191                                 len = 0;
 192                         } else {
 193                                 /* Run of identical lengths */
 194                                 run_len = 4 + bitstream_read_bits(is, 1);
 195                                 presym = read_presym(d, is);
 196                                 if (unlikely(presym > 17))
 197                                         return -1;
 198                                 len = *len_ptr - presym;
 199                                 if ((s8)len < 0)
 200                                         len += 17;
 201                         }
 202
 203                         do {
 204                                 *len_ptr++ = len;
 205                         } while (--run_len);
 206                         /*
 207                          * The worst case overrun is when presym == 18,
 208                          * run_len == 20 + 31, and only 1 length was remaining.
 209                          * So LZX_READ_LENS_MAX_OVERRUN == 50.
 210                          *
 211                          * Overrun while reading the first half of maincode_lens
 212                          * can corrupt the previous values in the second half.
 213                          * This doesn't really matter because the resulting
 214                          * lengths will still be in range, and data that
 215                          * generates overruns is invalid anyway.
 216                          */
 217                 }
 218         } while (len_ptr < lens_end);
 219
 220         return 0;
 221 }
 222
 223 /*
 224  * Read the header of an LZX block and save the block type and size in
 225  * *block_type_ret and *block_size_ret, respectively.
 226  *
 227  * If the block is compressed, the nalso initialize the decode tables for the
 228  * new Huffman codes.
 229  *
 230  * If the block is uncompressed, then also update the recent offsets queue.
 231  *
 232  * Return 0 on success, or -1 if the data was invalid.
 233  */
 234 static int
 235 lzx_read_block_header(struct lzx_decompressor *d, struct input_bitstream *is,
 236                       u32 recent_offsets[], int *block_type_ret,
 237                       u32 *block_size_ret)
 238 {
 239         int block_type;
 240         u32 block_size;
 241
 242         bitstream_ensure_bits(is, 4);
 243
 244         /* Read the block type. */
 245         block_type = bitstream_pop_bits(is, 3);
 246
 247         /* Read the block size. */
 248         if (bitstream_pop_bits(is, 1)) {
 249                 block_size = LZX_DEFAULT_BLOCK_SIZE;
 250         } else {
 251                 u32 tmp;
 252                 block_size = 0;
 253
 254                 tmp = bitstream_read_bits(is, 8);
 255                 block_size |= tmp;
 256                 tmp = bitstream_read_bits(is, 8);
 257                 block_size <<= 8;
 258                 block_size |= tmp;
 259
 260                 if (d->window_order >= 16) {
 261                         tmp = bitstream_read_bits(is, 8);
 262                         block_size <<= 8;
 263                         block_size |= tmp;
 264                 }
 265         }
 266
 267         switch (block_type) {
 268
 269         case LZX_BLOCKTYPE_ALIGNED:
 270
 271                 /* Read the aligned offset code and build its decode table. */
 272
 273                 for (int i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) {
 274                         d->alignedcode_lens[i] =
 275                                 bitstream_read_bits(is,
 276                                                     LZX_ALIGNEDCODE_ELEMENT_SIZE);
 277                 }
 278
 279                 if (make_huffman_decode_table(d->alignedcode_decode_table,
 280                                               LZX_ALIGNEDCODE_NUM_SYMBOLS,
 281                                               LZX_ALIGNEDCODE_TABLEBITS,
 282                                               d->alignedcode_lens,
 283                                               LZX_MAX_ALIGNED_CODEWORD_LEN))
 284                         return -1;
 285
 286                 /* Fall though, since the rest of the header for aligned offset
 287                  * blocks is the same as that for verbatim blocks.  */
 288
 289         case LZX_BLOCKTYPE_VERBATIM:
 290
 291                 /* Read the main code and build its decode table.  The codeword
 292                  * lengths in the main code are encoded in two parts: one part
 293                  * for literal symbols, and one part for match symbols. */
 294
 295                 if (lzx_read_codeword_lens(d, is, d->maincode_lens,
 296                                            LZX_NUM_CHARS))
 297                         return -1;
 298
 299                 if (lzx_read_codeword_lens(d, is, d->maincode_lens + LZX_NUM_CHARS,
 300                                            d->num_main_syms - LZX_NUM_CHARS))
 301                         return -1;
 302
 303                 if (make_huffman_decode_table(d->maincode_decode_table,
 304                                               d->num_main_syms,
 305                                               LZX_MAINCODE_TABLEBITS,
 306                                               d->maincode_lens,
 307                                               LZX_MAX_MAIN_CODEWORD_LEN))
 308                         return -1;
 309
 310                 /* Read the length code and build its decode table. */
 311
 312                 if (lzx_read_codeword_lens(d, is, d->lencode_lens,
 313                                            LZX_LENCODE_NUM_SYMBOLS))
 314                         return -1;
 315
 316                 if (make_huffman_decode_table(d->lencode_decode_table,
 317                                               LZX_LENCODE_NUM_SYMBOLS,
 318                                               LZX_LENCODE_TABLEBITS,
 319                                               d->lencode_lens,
 320                                               LZX_MAX_LEN_CODEWORD_LEN))
 321                         return -1;
 322
 323                 break;
 324
 325         case LZX_BLOCKTYPE_UNCOMPRESSED:
 326                 /*
 327                  * The header of an uncompressed block contains new values for
 328                  * the recent offsets queue, starting on the next 16-bit
 329                  * boundary in the bitstream.  Careful: if the stream is
 330                  * *already* aligned, the correct thing to do is to throw away
 331                  * the next 16 bits (this is probably a mistake in the format).
 332                  */
 333                 bitstream_ensure_bits(is, 1);
 334                 bitstream_align(is);
 335                 recent_offsets[0] = bitstream_read_u32(is);
 336                 recent_offsets[1] = bitstream_read_u32(is);
 337                 recent_offsets[2] = bitstream_read_u32(is);
 338
 339                 /* Offsets of 0 are invalid. */
 340                 if (recent_offsets[0] == 0 || recent_offsets[1] == 0 ||
 341                     recent_offsets[2] == 0)
 342                         return -1;
 343                 break;
 344
 345         default:
 346                 /* Unrecognized block type */
 347                 return -1;
 348         }
 349
 350         *block_type_ret = block_type;
 351         *block_size_ret = block_size;
 352         return 0;
 353 }
 354
 355 /* Decompress a block of LZX-compressed data. */
 356 static int
 357 lzx_decompress_block(const struct lzx_decompressor *d,
 358                      struct input_bitstream *is, int block_type, u32 block_size,
 359                      u8 * const out_begin, u8 *out_next, u32 recent_offsets[])
 360 {
 361         bool is_aligned_block = (block_type == LZX_BLOCKTYPE_ALIGNED);
 362         u8 * const block_end = out_next + block_size;
 363
 364         do {
 365                 unsigned mainsym;
 366                 unsigned length;
 367                 u32 offset;
 368                 unsigned offset_slot;
 369                 unsigned num_extra_bits;
 370
 371                 mainsym = read_mainsym(d, is);
 372                 if (mainsym < LZX_NUM_CHARS) {
 373                         /* Literal */
 374                         *out_next++ = mainsym;
 375                         continue;
 376                 }
 377
 378                 /* Match */
 379
 380                 /* Decode the length header and offset slot.  */
 381                 mainsym -= LZX_NUM_CHARS;
 382                 length = mainsym % LZX_NUM_LEN_HEADERS;
 383                 offset_slot = mainsym / LZX_NUM_LEN_HEADERS;
 384
 385                 /* If needed, read a length symbol to decode the full length. */
 386                 if (length == LZX_NUM_PRIMARY_LENS)
 387                         length += read_lensym(d, is);
 388                 length += LZX_MIN_MATCH_LEN;
 389
 390                 if (offset_slot < LZX_NUM_RECENT_OFFSETS) {
 391                         /* Repeat offset  */
 392
 393                         /* Note: This isn't a real LRU queue, since using the R2
 394                          * offset doesn't bump the R1 offset down to R2.  This
 395                          * quirk allows all 3 recent offsets to be handled by
 396                          * the same code.  (For R0, the swap is a no-op.)  */
 397                         offset = recent_offsets[offset_slot];
 398                         recent_offsets[offset_slot] = recent_offsets[0];
 399                 } else {
 400                         /* Explicit offset  */
 401
 402                         /* Look up the number of extra bits that need to be read
 403                          * to decode offsets with this offset slot.  */
 404                         num_extra_bits = lzx_extra_offset_bits[offset_slot];
 405
 406                         /* Start with the offset slot base value.  */
 407                         offset = lzx_offset_slot_base[offset_slot];
 408
 409                         /* In aligned offset blocks, the low-order 3 bits of
 410                          * each offset are encoded using the aligned offset
 411                          * code.  Otherwise, all the extra bits are literal.  */
 412
 413                         if (is_aligned_block && offset_slot >= 8) {
 414                                 offset +=
 415                                         bitstream_read_bits(is,
 416                                                             num_extra_bits -
 417                                                                 LZX_NUM_ALIGNED_OFFSET_BITS)
 418                                                         << LZX_NUM_ALIGNED_OFFSET_BITS;
 419                                 offset += read_alignedsym(d, is);
 420                         } else {
 421                                 offset += bitstream_read_bits(is, num_extra_bits);
 422                         }
 423
 424                         /* Adjust the offset.  */
 425                         offset -= LZX_OFFSET_ADJUSTMENT;
 426
 427                         /* Update the match offset LRU queue.  */
 428                         STATIC_ASSERT(LZX_NUM_RECENT_OFFSETS == 3);
 429                         recent_offsets[2] = recent_offsets[1];
 430                         recent_offsets[1] = recent_offsets[0];
 431                 }
 432                 recent_offsets[0] = offset;
 433
 434                 /* Validate the match, then copy it to the current position.  */
 435
 436                 if (unlikely(length > block_end - out_next))
 437                         return -1;
 438
 439                 if (unlikely(offset > out_next - out_begin))
 440                         return -1;
 441
 442                 lz_copy(out_next, length, offset, block_end, LZX_MIN_MATCH_LEN);
 443
 444                 out_next += length;
 445
 446         } while (out_next != block_end);
 447
 448         return 0;
 449 }
 450
 451 static int
 452 lzx_decompress(const void *restrict compressed_data, size_t compressed_size,
 453                void *restrict uncompressed_data, size_t uncompressed_size,
 454                void *restrict _d)
 455 {
 456         struct lzx_decompressor *d = _d;
 457         u8 * const out_begin = uncompressed_data;
 458         u8 *out_next = out_begin;
 459         u8 * const out_end = out_begin + uncompressed_size;
 460         struct input_bitstream is;
 461         STATIC_ASSERT(LZX_NUM_RECENT_OFFSETS == 3);
 462         u32 recent_offsets[] = {1, 1, 1};
 463         unsigned e8_status = 0;
 464
 465         init_input_bitstream(&is, compressed_data, compressed_size);
 466
 467         /* Codeword lengths begin as all 0's for delta encoding purposes. */
 468         memset(d->maincode_lens, 0, d->num_main_syms);
 469         memset(d->lencode_lens, 0, LZX_LENCODE_NUM_SYMBOLS);
 470
 471         /* Decompress blocks until we have all the uncompressed data. */
 472
 473         while (out_next != out_end) {
 474                 int block_type;
 475                 u32 block_size;
 476
 477                 if (lzx_read_block_header(d, &is, recent_offsets,
 478                                           &block_type, &block_size))
 479                         return -1;
 480
 481                 if (block_size < 1 || block_size > out_end - out_next)
 482                         return -1;
 483
 484                 if (block_type != LZX_BLOCKTYPE_UNCOMPRESSED) {
 485
 486                         /* Compressed block */
 487                         if (lzx_decompress_block(d, &is, block_type, block_size,
 488                                                  out_begin, out_next,
 489                                                  recent_offsets))
 490                                 return -1;
 491
 492                         /* If the first E8 byte was in this block, then it must
 493                          * have been encoded as a literal using mainsym E8. */
 494                         e8_status |= d->maincode_lens[0xE8];
 495                 } else {
 496
 497                         /* Uncompressed block */
 498                         if (bitstream_read_bytes(&is, out_next, block_size))
 499                                 return -1;
 500
 501                         /* Re-align the bitstream if needed. */
 502                         if (block_size & 1)
 503                                 bitstream_read_byte(&is);
 504
 505                         /* There may have been an E8 byte in the block. */
 506                         e8_status = 1;
 507                 }
 508                 out_next += block_size;
 509         }
 510
 511         /* Postprocess the data unless it cannot possibly contain E8 bytes. */
 512         if (e8_status)
 513                 lzx_postprocess(uncompressed_data, uncompressed_size);
 514
 515         return 0;
 516 }
 517
 518 static void
 519 lzx_free_decompressor(void *_d)
 520 {
 521         ALIGNED_FREE(_d);
 522 }
 523
 524 static int
 525 lzx_create_decompressor(size_t max_block_size, void **d_ret)
 526 {
 527         unsigned window_order;
 528         struct lzx_decompressor *d;
 529
 530         window_order = lzx_get_window_order(max_block_size);
 531         if (window_order == 0)
 532                 return WIMLIB_ERR_INVALID_PARAM;
 533
 534         d = ALIGNED_MALLOC(sizeof(*d), DECODE_TABLE_ALIGNMENT);
 535         if (!d)
 536                 return WIMLIB_ERR_NOMEM;
 537
 538         d->window_order = window_order;
 539         d->num_main_syms = lzx_get_num_main_syms(window_order);
 540
 541         *d_ret = d;
 542         return 0;
 543 }
 544
 545 const struct decompressor_ops lzx_decompressor_ops = {
 546         .create_decompressor = lzx_create_decompressor,
 547         .decompress          = lzx_decompress,
 548         .free_decompressor   = lzx_free_decompressor,
 549 };