X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Flzx-decompress.c;h=4047f956c376abf0e2d35dbe02d5a0ef37adc6ec;hp=d25d188b08b1e8fa6320f17b47127fa617d5198c;hb=157d002da341c9109c5c065893ae82c6dbf5d4e8;hpb=2254a0fc3f1d7af1151ee83f3458f44339b5028b diff --git a/src/lzx-decompress.c b/src/lzx-decompress.c index d25d188b..4047f956 100644 --- a/src/lzx-decompress.c +++ b/src/lzx-decompress.c @@ -38,13 +38,13 @@ * * A compressed WIM resource consists of a table of chunk offsets followed by * the compressed chunks themselves. All compressed chunks except possibly the - * last decompress to WIM_CHUNK_SIZE (= 32768) bytes. This is quite similar to - * the cabinet (.cab) file format, but they are not the same. According to the - * cabinet format documentation, the LZX block size is independent from the - * CFDATA blocks, and a LZX block may span several CFDATA blocks. However, in - * WIMs, LZX blocks do not appear to ever span multiple WIM chunks. Note that - * this means any WIM chunk may be decompressed or compressed independently from - * any other chunk, which is convenient. + * last decompress to 32768 bytes. This is quite similar to the cabinet (.cab) + * file format, but they are not the same. According to the cabinet format + * documentation, the LZX block size is independent from the CFDATA blocks, and + * a LZX block may span several CFDATA blocks. However, in WIMs, LZX blocks do + * not appear to ever span multiple WIM chunks. Note that this means any WIM + * chunk may be decompressed or compressed independently from any other chunk, + * which is convenient. * * A LZX compressed WIM chunk contains one or more LZX blocks of the aligned, * verbatim, or uncompressed block types. For aligned and verbatim blocks, the @@ -120,22 +120,22 @@ /* Huffman decoding tables and maps from symbols to code lengths. */ struct lzx_tables { - u16 maintree_decode_table[(1 << LZX_MAINTREE_TABLEBITS) + - (LZX_MAINTREE_NUM_SYMBOLS * 2)] + u16 maintree_decode_table[(1 << LZX_MAINCODE_TABLEBITS) + + (LZX_MAINCODE_MAX_NUM_SYMBOLS * 2)] _aligned_attribute(DECODE_TABLE_ALIGNMENT); - u8 maintree_lens[LZX_MAINTREE_NUM_SYMBOLS]; + u8 maintree_lens[LZX_MAINCODE_MAX_NUM_SYMBOLS]; - u16 lentree_decode_table[(1 << LZX_LENTREE_TABLEBITS) + - (LZX_LENTREE_NUM_SYMBOLS * 2)] + u16 lentree_decode_table[(1 << LZX_LENCODE_TABLEBITS) + + (LZX_LENCODE_NUM_SYMBOLS * 2)] _aligned_attribute(DECODE_TABLE_ALIGNMENT); - u8 lentree_lens[LZX_LENTREE_NUM_SYMBOLS]; + u8 lentree_lens[LZX_LENCODE_NUM_SYMBOLS]; - u16 alignedtree_decode_table[(1 << LZX_ALIGNEDTREE_TABLEBITS) + - (LZX_ALIGNEDTREE_NUM_SYMBOLS * 2)] + u16 alignedtree_decode_table[(1 << LZX_ALIGNEDCODE_TABLEBITS) + + (LZX_ALIGNEDCODE_NUM_SYMBOLS * 2)] _aligned_attribute(DECODE_TABLE_ALIGNMENT); - u8 alignedtree_lens[LZX_ALIGNEDTREE_NUM_SYMBOLS]; + u8 alignedtree_lens[LZX_ALIGNEDCODE_NUM_SYMBOLS]; } _aligned_attribute(DECODE_TABLE_ALIGNMENT); @@ -148,19 +148,20 @@ read_huffsym_using_pretree(struct input_bitstream *istream, const u8 pretree_lens[], unsigned *n) { return read_huffsym(istream, pretree_decode_table, pretree_lens, - LZX_PRETREE_NUM_SYMBOLS, LZX_PRETREE_TABLEBITS, n, - LZX_MAX_CODEWORD_LEN); + LZX_PRECODE_NUM_SYMBOLS, LZX_PRECODE_TABLEBITS, n, + LZX_MAX_PRE_CODEWORD_LEN); } /* Reads a Huffman-encoded symbol using the main tree. */ static inline int read_huffsym_using_maintree(struct input_bitstream *istream, const struct lzx_tables *tables, - unsigned *n) + unsigned *n, + unsigned num_main_syms) { return read_huffsym(istream, tables->maintree_decode_table, - tables->maintree_lens, LZX_MAINTREE_NUM_SYMBOLS, - LZX_MAINTREE_TABLEBITS, n, LZX_MAX_CODEWORD_LEN); + tables->maintree_lens, num_main_syms, + LZX_MAINCODE_TABLEBITS, n, LZX_MAX_MAIN_CODEWORD_LEN); } /* Reads a Huffman-encoded symbol using the length tree. */ @@ -170,8 +171,8 @@ read_huffsym_using_lentree(struct input_bitstream *istream, unsigned *n) { return read_huffsym(istream, tables->lentree_decode_table, - tables->lentree_lens, LZX_LENTREE_NUM_SYMBOLS, - LZX_LENTREE_TABLEBITS, n, LZX_MAX_CODEWORD_LEN); + tables->lentree_lens, LZX_LENCODE_NUM_SYMBOLS, + LZX_LENCODE_TABLEBITS, n, LZX_MAX_LEN_CODEWORD_LEN); } /* Reads a Huffman-encoded symbol using the aligned offset tree. */ @@ -182,8 +183,9 @@ read_huffsym_using_alignedtree(struct input_bitstream *istream, { return read_huffsym(istream, tables->alignedtree_decode_table, tables->alignedtree_lens, - LZX_ALIGNEDTREE_NUM_SYMBOLS, - LZX_ALIGNEDTREE_TABLEBITS, n, 8); + LZX_ALIGNEDCODE_NUM_SYMBOLS, + LZX_ALIGNEDCODE_TABLEBITS, n, + LZX_MAX_ALIGNED_CODEWORD_LEN); } /* @@ -203,31 +205,31 @@ lzx_read_code_lens(struct input_bitstream *istream, u8 lens[], unsigned num_lens) { /* Declare the decoding table and length table for the pretree. */ - u16 pretree_decode_table[(1 << LZX_PRETREE_TABLEBITS) + - (LZX_PRETREE_NUM_SYMBOLS * 2)] + u16 pretree_decode_table[(1 << LZX_PRECODE_TABLEBITS) + + (LZX_PRECODE_NUM_SYMBOLS * 2)] _aligned_attribute(DECODE_TABLE_ALIGNMENT); - u8 pretree_lens[LZX_PRETREE_NUM_SYMBOLS]; + u8 pretree_lens[LZX_PRECODE_NUM_SYMBOLS]; unsigned i; unsigned len; int ret; /* Read the code lengths of the pretree codes. There are 20 lengths of * 4 bits each. */ - for (i = 0; i < LZX_PRETREE_NUM_SYMBOLS; i++) { - ret = bitstream_read_bits(istream, LZX_PRETREE_ELEMENT_SIZE, + for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) { + ret = bitstream_read_bits(istream, LZX_PRECODE_ELEMENT_SIZE, &len); - if (ret != 0) + if (ret) return ret; pretree_lens[i] = len; } /* Make the decoding table for the pretree. */ ret = make_huffman_decode_table(pretree_decode_table, - LZX_PRETREE_NUM_SYMBOLS, - LZX_PRETREE_TABLEBITS, + LZX_PRECODE_NUM_SYMBOLS, + LZX_PRECODE_TABLEBITS, pretree_lens, - LZX_MAX_CODEWORD_LEN); - if (ret != 0) + LZX_MAX_PRE_CODEWORD_LEN); + if (ret) return ret; /* Pointer past the last length value that needs to be filled in. */ @@ -249,12 +251,12 @@ lzx_read_code_lens(struct input_bitstream *istream, u8 lens[], ret = read_huffsym_using_pretree(istream, pretree_decode_table, pretree_lens, &tree_code); - if (ret != 0) + if (ret) return ret; switch (tree_code) { case 17: /* Run of 0's */ ret = bitstream_read_bits(istream, 4, &num_zeroes); - if (ret != 0) + if (ret) return ret; num_zeroes += 4; while (num_zeroes--) { @@ -265,7 +267,7 @@ lzx_read_code_lens(struct input_bitstream *istream, u8 lens[], break; case 18: /* Longer run of 0's */ ret = bitstream_read_bits(istream, 5, &num_zeroes); - if (ret != 0) + if (ret) return ret; num_zeroes += 20; while (num_zeroes--) { @@ -276,14 +278,14 @@ lzx_read_code_lens(struct input_bitstream *istream, u8 lens[], break; case 19: /* Run of identical lengths */ ret = bitstream_read_bits(istream, 1, &num_same); - if (ret != 0) + if (ret) return ret; num_same += 4; ret = read_huffsym_using_pretree(istream, pretree_decode_table, pretree_lens, &code); - if (ret != 0) + if (ret) return ret; value = (signed char)*lens - (signed char)code; if (value < 0) @@ -323,6 +325,8 @@ lzx_read_code_lens(struct input_bitstream *istream, u8 lens[], */ static int lzx_read_block_header(struct input_bitstream *istream, + unsigned num_main_syms, + unsigned max_window_size, unsigned *block_size_ret, unsigned *block_type_ret, struct lzx_tables *tables, @@ -335,7 +339,7 @@ lzx_read_block_header(struct input_bitstream *istream, unsigned i; unsigned len; - ret = bitstream_ensure_bits(istream, LZX_BLOCKTYPE_NBITS + 1); + ret = bitstream_ensure_bits(istream, 4); if (ret) { LZX_DEBUG("LZX input stream overrun"); return ret; @@ -343,20 +347,36 @@ lzx_read_block_header(struct input_bitstream *istream, /* The first three bits tell us what kind of block it is, and are one * of the LZX_BLOCKTYPE_* values. */ - block_type = bitstream_read_bits_nocheck(istream, LZX_BLOCKTYPE_NBITS); + block_type = bitstream_read_bits_nocheck(istream, 3); - /* The next bit indicates whether the block size is the default (32768), - * indicated by a 1 bit, or whether the block size is given by the next - * 16 bits, indicated by a 0 bit. */ + /* Read the block size. This mirrors the behavior + * lzx_write_compressed_block() in lzx-compress.c; see that for more + * details. */ s = bitstream_read_bits_nocheck(istream, 1); - if (s) { - block_size = 32768; + block_size = LZX_DEFAULT_BLOCK_SIZE; } else { - ret = bitstream_read_bits(istream, LZX_BLOCKSIZE_NBITS, &block_size); + unsigned tmp; + block_size = 0; + + ret = bitstream_read_bits(istream, 8, &tmp); + if (ret) + return ret; + block_size |= tmp; + + ret = bitstream_read_bits(istream, 8, &tmp); if (ret) return ret; - block_size = le16_to_cpu(block_size); + block_size <<= 8; + block_size |= tmp; + + if (max_window_size >= 65536) { + ret = bitstream_read_bits(istream, 8, &tmp); + if (ret) + return ret; + block_size <<= 8; + block_size |= tmp; + } } switch (block_type) { @@ -364,9 +384,9 @@ lzx_read_block_header(struct input_bitstream *istream, /* Read the path lengths for the elements of the aligned tree, * then build it. */ - for (i = 0; i < LZX_ALIGNEDTREE_NUM_SYMBOLS; i++) { + for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { ret = bitstream_read_bits(istream, - LZX_ALIGNEDTREE_ELEMENT_SIZE, + LZX_ALIGNEDCODE_ELEMENT_SIZE, &len); if (ret) return ret; @@ -375,10 +395,10 @@ lzx_read_block_header(struct input_bitstream *istream, LZX_DEBUG("Building the aligned tree."); ret = make_huffman_decode_table(tables->alignedtree_decode_table, - LZX_ALIGNEDTREE_NUM_SYMBOLS, - LZX_ALIGNEDTREE_TABLEBITS, + LZX_ALIGNEDCODE_NUM_SYMBOLS, + LZX_ALIGNEDCODE_TABLEBITS, tables->alignedtree_lens, - 8); + LZX_MAX_ALIGNED_CODEWORD_LEN); if (ret) { LZX_DEBUG("Failed to make the decode table for the " "aligned offset tree"); @@ -407,10 +427,10 @@ lzx_read_block_header(struct input_bitstream *istream, * tree. */ LZX_DEBUG("Reading path lengths for remaining elements of " "main tree (%d elements).", - LZX_MAINTREE_NUM_SYMBOLS - LZX_NUM_CHARS); + num_main_syms - LZX_NUM_CHARS); ret = lzx_read_code_lens(istream, tables->maintree_lens + LZX_NUM_CHARS, - LZX_MAINTREE_NUM_SYMBOLS - LZX_NUM_CHARS); + num_main_syms - LZX_NUM_CHARS); if (ret) { LZX_DEBUG("Failed to read the path lengths for the " "remaining elements of the main tree"); @@ -421,10 +441,10 @@ lzx_read_block_header(struct input_bitstream *istream, "table for the main tree."); ret = make_huffman_decode_table(tables->maintree_decode_table, - LZX_MAINTREE_NUM_SYMBOLS, - LZX_MAINTREE_TABLEBITS, + num_main_syms, + LZX_MAINCODE_TABLEBITS, tables->maintree_lens, - LZX_MAX_CODEWORD_LEN); + LZX_MAX_MAIN_CODEWORD_LEN); if (ret) { LZX_DEBUG("Failed to make the decode " "table for the main tree"); @@ -433,7 +453,7 @@ lzx_read_block_header(struct input_bitstream *istream, LZX_DEBUG("Reading path lengths for the length tree."); ret = lzx_read_code_lens(istream, tables->lentree_lens, - LZX_LENTREE_NUM_SYMBOLS); + LZX_LENCODE_NUM_SYMBOLS); if (ret) { LZX_DEBUG("Failed to read the path " "lengths for the length tree"); @@ -442,10 +462,10 @@ lzx_read_block_header(struct input_bitstream *istream, LZX_DEBUG("Building the length tree."); ret = make_huffman_decode_table(tables->lentree_decode_table, - LZX_LENTREE_NUM_SYMBOLS, - LZX_LENTREE_TABLEBITS, + LZX_LENCODE_NUM_SYMBOLS, + LZX_LENCODE_TABLEBITS, tables->lentree_lens, - LZX_MAX_CODEWORD_LEN); + LZX_MAX_LEN_CODEWORD_LEN); if (ret) { LZX_DEBUG("Failed to build the length Huffman tree"); return ret; @@ -478,9 +498,9 @@ lzx_read_block_header(struct input_bitstream *istream, istream->bitsleft = 0; istream->bitbuf = 0; } - queue->R0 = le32_to_cpu(*(u32*)(istream->data + 0)); - queue->R1 = le32_to_cpu(*(u32*)(istream->data + 4)); - queue->R2 = le32_to_cpu(*(u32*)(istream->data + 8)); + queue->R[0] = le32_to_cpu(*(u32*)(istream->data + 0)); + queue->R[1] = le32_to_cpu(*(u32*)(istream->data + 4)); + queue->R[2] = le32_to_cpu(*(u32*)(istream->data + 8)); istream->data += 12; istream->data_bytes_left -= 12; /* The uncompressed data of this block directly follows and will @@ -561,14 +581,15 @@ lzx_decode_match(unsigned main_element, int block_type, position_slot = main_element >> 3; /* If the length_header is less than LZX_NUM_PRIMARY_LENS (= 7), it - * gives the match length as the offset from LZX_MIN_MATCH. Otherwise, - * the length is given by an additional symbol encoded using the length - * tree, offset by 9 (LZX_MIN_MATCH + LZX_NUM_PRIMARY_LENS) */ - match_len = LZX_MIN_MATCH + length_header; + * gives the match length as the offset from LZX_MIN_MATCH_LEN. + * Otherwise, the length is given by an additional symbol encoded using + * the length tree, offset by 9 (LZX_MIN_MATCH_LEN + + * LZX_NUM_PRIMARY_LENS) */ + match_len = LZX_MIN_MATCH_LEN + length_header; if (length_header == LZX_NUM_PRIMARY_LENS) { ret = read_huffsym_using_lentree(istream, tables, &additional_len); - if (ret != 0) + if (ret) return ret; match_len += additional_len; } @@ -579,18 +600,18 @@ lzx_decode_match(unsigned main_element, int block_type, * queue. */ switch (position_slot) { case 0: - match_offset = queue->R0; + match_offset = queue->R[0]; break; case 1: - match_offset = queue->R1; - swap(queue->R0, queue->R1); + match_offset = queue->R[1]; + swap(queue->R[0], queue->R[1]); break; case 2: /* The queue doesn't work quite the same as a real LRU queue, * since using the R2 offset doesn't bump the R1 offset down to * R2. */ - match_offset = queue->R2; - swap(queue->R0, queue->R2); + match_offset = queue->R[2]; + swap(queue->R[0], queue->R[2]); break; default: /* Otherwise, the offset was not encoded as one the offsets in @@ -617,14 +638,14 @@ lzx_decode_match(unsigned main_element, int block_type, * will just set it to 0. ) */ ret = bitstream_read_bits(istream, num_extra_bits - 3, &verbatim_bits); - if (ret != 0) + if (ret) return ret; verbatim_bits <<= 3; ret = read_huffsym_using_alignedtree(istream, tables, &aligned_bits); - if (ret != 0) + if (ret) return ret; } else { /* For non-aligned blocks, or for aligned blocks with @@ -633,7 +654,7 @@ lzx_decode_match(unsigned main_element, int block_type, * the alignment is taken to be 0. */ ret = bitstream_read_bits(istream, num_extra_bits, &verbatim_bits); - if (ret != 0) + if (ret) return ret; aligned_bits = 0; @@ -641,12 +662,12 @@ lzx_decode_match(unsigned main_element, int block_type, /* Calculate the match offset. */ match_offset = lzx_position_base[position_slot] + - verbatim_bits + aligned_bits - 2; + verbatim_bits + aligned_bits - LZX_OFFSET_OFFSET; /* Update the LRU queue. */ - queue->R2 = queue->R1; - queue->R1 = queue->R0; - queue->R0 = match_offset; + queue->R[2] = queue->R[1]; + queue->R[1] = queue->R[0]; + queue->R[0] = match_offset; break; } @@ -752,6 +773,7 @@ undo_call_insn_preprocessing(u8 uncompressed_data[], int uncompressed_data_len) * @block_type: The type of the block (LZX_BLOCKTYPE_VERBATIM or * LZX_BLOCKTYPE_ALIGNED) * @block_size: The size of the block, in bytes. + * @num_main_syms: Number of symbols in the main alphabet. * @window: Pointer to the decompression window. * @window_pos: The current position in the window. Will be 0 for the first * block. @@ -762,6 +784,7 @@ undo_call_insn_preprocessing(u8 uncompressed_data[], int uncompressed_data_len) */ static int lzx_decompress_block(int block_type, unsigned block_size, + unsigned num_main_syms, u8 *window, unsigned window_pos, const struct lzx_tables *tables, @@ -776,7 +799,8 @@ lzx_decompress_block(int block_type, unsigned block_size, end = window_pos + block_size; while (window_pos < end) { ret = read_huffsym_using_maintree(istream, tables, - &main_element); + &main_element, + num_main_syms); if (ret) return ret; @@ -784,7 +808,7 @@ lzx_decompress_block(int block_type, unsigned block_size, /* literal: 0 to LZX_NUM_CHARS - 1 */ window[window_pos++] = main_element; } else { - /* match: LZX_NUM_CHARS to LZX_MAINTREE_NUM_SYMBOLS - 1 */ + /* match: LZX_NUM_CHARS to num_main_syms - 1 */ match_len = lzx_decode_match(main_element, block_type, end - window_pos, @@ -801,10 +825,10 @@ lzx_decompress_block(int block_type, unsigned block_size, return 0; } -/* API function documented in wimlib.h */ WIMLIBAPI int -wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, - void *uncompressed_data, unsigned uncompressed_len) +wimlib_lzx_decompress2(const void *compressed_data, unsigned compressed_len, + void *uncompressed_data, unsigned uncompressed_len, + u32 max_window_size) { struct lzx_tables tables; struct input_bitstream istream; @@ -812,21 +836,34 @@ wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, unsigned window_pos; unsigned block_size; unsigned block_type; + unsigned num_main_syms; int ret; bool e8_preprocessing_done; - LZX_DEBUG("lzx_decompress (compressed_data = %p, compressed_len = %d, " - "uncompressed_data = %p, uncompressed_len = %d).", + LZX_DEBUG("compressed_data = %p, compressed_len = %u, " + "uncompressed_data = %p, uncompressed_len = %u, " + "max_window_size=%u).", compressed_data, compressed_len, - uncompressed_data, uncompressed_len); + uncompressed_data, uncompressed_len, max_window_size); - wimlib_assert(uncompressed_len <= 32768); + if (!lzx_window_size_valid(max_window_size)) { + LZX_DEBUG("Window size of %u is invalid!", + max_window_size); + return -1; + } + + num_main_syms = lzx_get_num_main_syms(max_window_size); + + if (uncompressed_len > max_window_size) { + LZX_DEBUG("Uncompressed chunk size of %u exceeds " + "window size of %u!", + uncompressed_len, max_window_size); + return -1; + } memset(tables.maintree_lens, 0, sizeof(tables.maintree_lens)); memset(tables.lentree_lens, 0, sizeof(tables.lentree_lens)); - queue.R0 = 1; - queue.R1 = 1; - queue.R2 = 1; + lzx_lru_queue_init(&queue); init_input_bitstream(&istream, compressed_data, compressed_len); e8_preprocessing_done = false; /* Set to true if there may be 0xe8 bytes @@ -842,7 +879,8 @@ wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, window_pos += block_size) { LZX_DEBUG("Reading block header."); - ret = lzx_read_block_header(&istream, &block_size, + ret = lzx_read_block_header(&istream, num_main_syms, + max_window_size, &block_size, &block_type, &tables, &queue); if (ret) return ret; @@ -866,6 +904,7 @@ wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, LZX_DEBUG("LZX_BLOCKTYPE_ALIGNED"); ret = lzx_decompress_block(block_type, block_size, + num_main_syms, uncompressed_data, window_pos, &tables, @@ -873,6 +912,7 @@ wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, &istream); if (ret) return ret; + if (tables.maintree_lens[0xe8] != 0) e8_preprocessing_done = true; break; @@ -903,3 +943,13 @@ wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, undo_call_insn_preprocessing(uncompressed_data, uncompressed_len); return 0; } + +/* API function documented in wimlib.h */ +WIMLIBAPI int +wimlib_lzx_decompress(const void *compressed_data, unsigned compressed_len, + void *uncompressed_data, unsigned uncompressed_len) +{ + return wimlib_lzx_decompress2(compressed_data, compressed_len, + uncompressed_data, uncompressed_len, + 32768); +}