4 * A decompressor for the LZX compression format, as used in WIM files.
8 * Copyright (C) 2012-2016 Eric Biggers
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
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
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/.
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.
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...
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.
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
50 * TODO: Add support for window sizes up to 2^25 bytes.
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"
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
71 #define LZX_READ_LENS_MAX_OVERRUN 50
73 struct lzx_decompressor {
75 u16 maincode_decode_table[(1 << LZX_MAINCODE_TABLEBITS) +
76 (LZX_MAINCODE_MAX_NUM_SYMBOLS * 2)]
77 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
78 u8 maincode_lens[LZX_MAINCODE_MAX_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
81 u16 lencode_decode_table[(1 << LZX_LENCODE_TABLEBITS) +
82 (LZX_LENCODE_NUM_SYMBOLS * 2)]
83 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
84 u8 lencode_lens[LZX_LENCODE_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
87 u16 alignedcode_decode_table[(1 << LZX_ALIGNEDCODE_TABLEBITS) +
88 (LZX_ALIGNEDCODE_NUM_SYMBOLS * 2)]
89 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
90 u8 alignedcode_lens[LZX_ALIGNEDCODE_NUM_SYMBOLS];
94 u16 precode_decode_table[(1 << LZX_PRECODE_TABLEBITS) +
95 (LZX_PRECODE_NUM_SYMBOLS * 2)]
96 _aligned_attribute(DECODE_TABLE_ALIGNMENT);
97 u8 precode_lens[LZX_PRECODE_NUM_SYMBOLS];
100 unsigned window_order;
101 unsigned num_main_syms;
102 } _aligned_attribute(DECODE_TABLE_ALIGNMENT);
104 /* Read a Huffman-encoded symbol using the precode. */
105 static inline unsigned
106 read_presym(const struct lzx_decompressor *d, struct input_bitstream *is)
108 return read_huffsym(is, d->precode_decode_table,
109 LZX_PRECODE_TABLEBITS, LZX_MAX_PRE_CODEWORD_LEN);
112 /* Read a Huffman-encoded symbol using the main code. */
113 static inline unsigned
114 read_mainsym(const struct lzx_decompressor *d, struct input_bitstream *is)
116 return read_huffsym(is, d->maincode_decode_table,
117 LZX_MAINCODE_TABLEBITS, LZX_MAX_MAIN_CODEWORD_LEN);
120 /* Read a Huffman-encoded symbol using the length code. */
121 static inline unsigned
122 read_lensym(const struct lzx_decompressor *d, struct input_bitstream *is)
124 return read_huffsym(is, d->lencode_decode_table,
125 LZX_LENCODE_TABLEBITS, LZX_MAX_LEN_CODEWORD_LEN);
128 /* Read a Huffman-encoded symbol using the aligned offset code. */
129 static inline unsigned
130 read_alignedsym(const struct lzx_decompressor *d, struct input_bitstream *is)
132 return read_huffsym(is, d->alignedcode_decode_table,
133 LZX_ALIGNEDCODE_TABLEBITS, LZX_MAX_ALIGNED_CODEWORD_LEN);
137 * Read a precode from the compressed input bitstream, then use it to decode
138 * @num_lens codeword length values and write them to @lens.
141 lzx_read_codeword_lens(struct lzx_decompressor *d, struct input_bitstream *is,
142 u8 *lens, unsigned num_lens)
145 u8 *lens_end = lens + num_lens;
147 /* Read the lengths of the precode codewords. These are stored
149 for (int i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) {
151 bitstream_read_bits(is, LZX_PRECODE_ELEMENT_SIZE);
154 /* Build the decoding table for the precode. */
155 if (make_huffman_decode_table(d->precode_decode_table,
156 LZX_PRECODE_NUM_SYMBOLS,
157 LZX_PRECODE_TABLEBITS,
159 LZX_MAX_PRE_CODEWORD_LEN))
162 /* Decode the codeword lengths. */
167 /* Read the next precode symbol. */
168 presym = read_presym(d, is);
170 /* Difference from old length */
171 len = *len_ptr - presym;
176 /* Special RLE values */
182 run_len = 4 + bitstream_read_bits(is, 4);
184 } else if (presym == 18) {
185 /* Longer run of 0's */
186 run_len = 20 + bitstream_read_bits(is, 5);
189 /* Run of identical lengths */
190 run_len = 4 + bitstream_read_bits(is, 1);
191 presym = read_presym(d, is);
192 if (unlikely(presym > 17))
194 len = *len_ptr - presym;
203 * The worst case overrun is when presym == 18,
204 * run_len == 20 + 31, and only 1 length was remaining.
205 * So LZX_READ_LENS_MAX_OVERRUN == 50.
207 * Overrun while reading the first half of maincode_lens
208 * can corrupt the previous values in the second half.
209 * This doesn't really matter because the resulting
210 * lengths will still be in range, and data that
211 * generates overruns is invalid anyway.
214 } while (len_ptr < lens_end);
220 * Read the header of an LZX block. For all block types, the block type and
221 * size is saved in *block_type_ret and *block_size_ret, respectively. For
222 * compressed blocks, the codeword lengths are also saved. For uncompressed
223 * blocks, the recent offsets queue is also updated.
226 lzx_read_block_header(struct lzx_decompressor *d, struct input_bitstream *is,
227 u32 recent_offsets[], int *block_type_ret,
233 bitstream_ensure_bits(is, 4);
235 /* Read the block type. */
236 block_type = bitstream_pop_bits(is, 3);
238 /* Read the block size. */
239 if (bitstream_pop_bits(is, 1)) {
240 block_size = LZX_DEFAULT_BLOCK_SIZE;
242 block_size = bitstream_read_bits(is, 16);
243 if (d->window_order >= 16) {
245 block_size |= bitstream_read_bits(is, 8);
249 switch (block_type) {
251 case LZX_BLOCKTYPE_ALIGNED:
253 /* Read the aligned offset codeword lengths. */
255 for (int i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) {
256 d->alignedcode_lens[i] =
257 bitstream_read_bits(is,
258 LZX_ALIGNEDCODE_ELEMENT_SIZE);
261 /* Fall though, since the rest of the header for aligned offset
262 * blocks is the same as that for verbatim blocks. */
264 case LZX_BLOCKTYPE_VERBATIM:
266 /* Read the main codeword lengths, which are divided into two
267 * parts: literal symbols and match headers. */
269 if (lzx_read_codeword_lens(d, is, d->maincode_lens,
273 if (lzx_read_codeword_lens(d, is, d->maincode_lens + LZX_NUM_CHARS,
274 d->num_main_syms - LZX_NUM_CHARS))
278 /* Read the length codeword lengths. */
280 if (lzx_read_codeword_lens(d, is, d->lencode_lens,
281 LZX_LENCODE_NUM_SYMBOLS))
286 case LZX_BLOCKTYPE_UNCOMPRESSED:
288 * The header of an uncompressed block contains new values for
289 * the recent offsets queue, starting on the next 16-bit
290 * boundary in the bitstream. Careful: if the stream is
291 * *already* aligned, the correct thing to do is to throw away
292 * the next 16 bits (this is probably a mistake in the format).
294 bitstream_ensure_bits(is, 1);
296 recent_offsets[0] = bitstream_read_u32(is);
297 recent_offsets[1] = bitstream_read_u32(is);
298 recent_offsets[2] = bitstream_read_u32(is);
300 /* Offsets of 0 are invalid. */
301 if (recent_offsets[0] == 0 || recent_offsets[1] == 0 ||
302 recent_offsets[2] == 0)
307 /* Unrecognized block type. */
311 *block_type_ret = block_type;
312 *block_size_ret = block_size;
316 /* Decompress a block of LZX-compressed data. */
318 lzx_decompress_block(struct lzx_decompressor *d, struct input_bitstream *is,
319 int block_type, u32 block_size,
320 u8 * const out_begin, u8 *out_next, u32 recent_offsets[])
322 u8 * const block_end = out_next + block_size;
323 unsigned min_aligned_offset_slot;
326 * Build the Huffman decode tables. We always need to build the main
327 * and length decode tables. For aligned blocks we additionally need to
328 * build the aligned offset decode table.
331 if (make_huffman_decode_table(d->maincode_decode_table,
333 LZX_MAINCODE_TABLEBITS,
335 LZX_MAX_MAIN_CODEWORD_LEN))
338 if (make_huffman_decode_table(d->lencode_decode_table,
339 LZX_LENCODE_NUM_SYMBOLS,
340 LZX_LENCODE_TABLEBITS,
342 LZX_MAX_LEN_CODEWORD_LEN))
345 if (block_type == LZX_BLOCKTYPE_ALIGNED) {
346 if (make_huffman_decode_table(d->alignedcode_decode_table,
347 LZX_ALIGNEDCODE_NUM_SYMBOLS,
348 LZX_ALIGNEDCODE_TABLEBITS,
350 LZX_MAX_ALIGNED_CODEWORD_LEN))
352 min_aligned_offset_slot = 8;
354 min_aligned_offset_slot = LZX_MAX_OFFSET_SLOTS;
357 /* Decode the literals and matches. */
363 unsigned offset_slot;
364 unsigned num_extra_bits;
366 mainsym = read_mainsym(d, is);
367 if (mainsym < LZX_NUM_CHARS) {
369 *out_next++ = mainsym;
375 /* Decode the length header and offset slot. */
376 STATIC_ASSERT(LZX_NUM_CHARS % LZX_NUM_LEN_HEADERS == 0);
377 length = mainsym % LZX_NUM_LEN_HEADERS;
378 offset_slot = (mainsym - LZX_NUM_CHARS) / LZX_NUM_LEN_HEADERS;
380 /* If needed, read a length symbol to decode the full length. */
381 if (length == LZX_NUM_PRIMARY_LENS)
382 length += read_lensym(d, is);
383 length += LZX_MIN_MATCH_LEN;
385 if (offset_slot < LZX_NUM_RECENT_OFFSETS) {
388 /* Note: This isn't a real LRU queue, since using the R2
389 * offset doesn't bump the R1 offset down to R2. */
390 offset = recent_offsets[offset_slot];
391 recent_offsets[offset_slot] = recent_offsets[0];
393 /* Explicit offset */
395 /* Look up the number of extra bits that need to be read
396 * to decode offsets with this offset slot. */
397 num_extra_bits = lzx_extra_offset_bits[offset_slot];
399 /* Start with the offset slot base value. */
400 offset = lzx_offset_slot_base[offset_slot];
402 /* In aligned offset blocks, the low-order 3 bits of
403 * each offset are encoded using the aligned offset
404 * code. Otherwise, all the extra bits are literal. */
406 if (offset_slot >= min_aligned_offset_slot) {
408 bitstream_read_bits(is,
410 LZX_NUM_ALIGNED_OFFSET_BITS)
411 << LZX_NUM_ALIGNED_OFFSET_BITS;
412 offset += read_alignedsym(d, is);
414 offset += bitstream_read_bits(is, num_extra_bits);
417 /* Update the match offset LRU queue. */
418 STATIC_ASSERT(LZX_NUM_RECENT_OFFSETS == 3);
419 recent_offsets[2] = recent_offsets[1];
420 recent_offsets[1] = recent_offsets[0];
422 recent_offsets[0] = offset;
424 /* Validate the match, then copy it to the current position. */
426 if (unlikely(length > block_end - out_next))
429 if (unlikely(offset > out_next - out_begin))
432 lz_copy(out_next, length, offset, block_end, LZX_MIN_MATCH_LEN);
436 } while (out_next != block_end);
442 lzx_decompress(const void *restrict compressed_data, size_t compressed_size,
443 void *restrict uncompressed_data, size_t uncompressed_size,
446 struct lzx_decompressor *d = _d;
447 u8 * const out_begin = uncompressed_data;
448 u8 *out_next = out_begin;
449 u8 * const out_end = out_begin + uncompressed_size;
450 struct input_bitstream is;
451 STATIC_ASSERT(LZX_NUM_RECENT_OFFSETS == 3);
452 u32 recent_offsets[LZX_NUM_RECENT_OFFSETS] = {1, 1, 1};
453 unsigned may_have_e8_byte = 0;
455 init_input_bitstream(&is, compressed_data, compressed_size);
457 /* Codeword lengths begin as all 0's for delta encoding purposes. */
458 memset(d->maincode_lens, 0, d->num_main_syms);
459 memset(d->lencode_lens, 0, LZX_LENCODE_NUM_SYMBOLS);
461 /* Decompress blocks until we have all the uncompressed data. */
463 while (out_next != out_end) {
467 if (lzx_read_block_header(d, &is, recent_offsets,
468 &block_type, &block_size))
471 if (block_size < 1 || block_size > out_end - out_next)
474 if (likely(block_type != LZX_BLOCKTYPE_UNCOMPRESSED)) {
476 /* Compressed block */
477 if (lzx_decompress_block(d, &is, block_type, block_size,
482 /* If the first E8 byte was in this block, then it must
483 * have been encoded as a literal using mainsym E8. */
484 may_have_e8_byte |= d->maincode_lens[0xE8];
487 /* Uncompressed block */
488 if (bitstream_read_bytes(&is, out_next, block_size))
491 /* Re-align the bitstream if needed. */
493 bitstream_read_byte(&is);
495 /* There may have been an E8 byte in the block. */
496 may_have_e8_byte = 1;
498 out_next += block_size;
501 /* Postprocess the data unless it cannot possibly contain E8 bytes. */
502 if (may_have_e8_byte)
503 lzx_postprocess(uncompressed_data, uncompressed_size);
509 lzx_create_decompressor(size_t max_block_size, void **d_ret)
511 unsigned window_order;
512 struct lzx_decompressor *d;
514 window_order = lzx_get_window_order(max_block_size);
515 if (window_order == 0)
516 return WIMLIB_ERR_INVALID_PARAM;
518 d = ALIGNED_MALLOC(sizeof(*d), DECODE_TABLE_ALIGNMENT);
520 return WIMLIB_ERR_NOMEM;
522 d->window_order = window_order;
523 d->num_main_syms = lzx_get_num_main_syms(window_order);
530 lzx_free_decompressor(void *_d)
535 const struct decompressor_ops lzx_decompressor_ops = {
536 .create_decompressor = lzx_create_decompressor,
537 .decompress = lzx_decompress,
538 .free_decompressor = lzx_free_decompressor,