|
| 1 | +import io |
| 2 | +from enum import Enum |
| 3 | +from pathlib import Path |
| 4 | + |
| 5 | +import lzfse |
| 6 | + |
| 7 | +from unblob.file_utils import ( |
| 8 | + Endian, |
| 9 | + File, |
| 10 | + FileSystem, |
| 11 | + InvalidInputFormat, |
| 12 | + StructParser, |
| 13 | +) |
| 14 | +from unblob.models import ( |
| 15 | + Extractor, |
| 16 | + ExtractResult, |
| 17 | + Handler, |
| 18 | + HandlerDoc, |
| 19 | + HandlerType, |
| 20 | + HexString, |
| 21 | + Reference, |
| 22 | + ValidChunk, |
| 23 | +) |
| 24 | + |
| 25 | + |
| 26 | +class LZFSEMagic(bytes, Enum): |
| 27 | + END = b"bvx$" # end-of-stream block |
| 28 | + UNCOMPRESSED = b"bvx-" # raw block |
| 29 | + LZVN = b"bvxn" # LZVN-compressed block |
| 30 | + LZFSE_V1 = b"bvx1" # LZFSE v1 block (legacy) |
| 31 | + LZFSE_V2 = b"bvx2" # LZFSE v2 block |
| 32 | + |
| 33 | + |
| 34 | +# sizeof(lzfse_compressed_block_header_v1), including struct alignment padding |
| 35 | +_V1_HEADER_SIZE = 772 |
| 36 | + |
| 37 | +# length of every LZFSE block magic |
| 38 | +MAGIC_LEN = 4 |
| 39 | + |
| 40 | +# a stream decoding to zero bytes: an empty raw block followed by the end block, |
| 41 | +# as emitted by lzfse.compress(b"") |
| 42 | +_EMPTY_STREAM = LZFSEMagic.UNCOMPRESSED + b"\x00\x00\x00\x00" + LZFSEMagic.END |
| 43 | + |
| 44 | +C_DEFINITIONS = r""" |
| 45 | + typedef struct lzfse_uncompressed { |
| 46 | + char magic[4]; |
| 47 | + uint32 n_raw_bytes; |
| 48 | + } lzfse_uncompressed_t; |
| 49 | +
|
| 50 | + typedef struct lzvn_compressed { |
| 51 | + char magic[4]; |
| 52 | + uint32 n_raw_bytes; |
| 53 | + uint32 n_payload_bytes; |
| 54 | + } lzvn_compressed_t; |
| 55 | +
|
| 56 | + typedef struct lzfse_v1 { |
| 57 | + char magic[4]; |
| 58 | + uint32 n_raw_bytes; |
| 59 | + uint32 n_payload_bytes; |
| 60 | + uint32 n_literals; |
| 61 | + uint32 n_matches; |
| 62 | + uint32 n_literal_payload_bytes; |
| 63 | + uint32 n_lmd_payload_bytes; |
| 64 | + } lzfse_v1_t; |
| 65 | +
|
| 66 | + typedef struct lzfse_v2 { |
| 67 | + char magic[4]; |
| 68 | + uint32 n_raw_bytes; |
| 69 | + uint64 packed_fields_0; |
| 70 | + uint64 packed_fields_1; |
| 71 | + uint64 packed_fields_2; |
| 72 | + } lzfse_v2_t; |
| 73 | +""" |
| 74 | + |
| 75 | +_parser = StructParser(C_DEFINITIONS) |
| 76 | + |
| 77 | + |
| 78 | +class LZFSEExtractor(Extractor): |
| 79 | + def extract(self, inpath: Path, outdir: Path) -> ExtractResult | None: |
| 80 | + fs = FileSystem(outdir) |
| 81 | + with File.from_path(inpath) as file: |
| 82 | + content = file.read() |
| 83 | + # lzfse.decompress raises on a stream that decodes to zero bytes, |
| 84 | + # so write the empty output ourselves. |
| 85 | + decompressed = ( |
| 86 | + b"" if content == _EMPTY_STREAM else lzfse.decompress(content) |
| 87 | + ) |
| 88 | + fs.write_bytes(Path(f"{inpath.stem}.bin"), decompressed) |
| 89 | + return ExtractResult(reports=fs.problems) |
| 90 | + |
| 91 | + |
| 92 | +class LZFSEHandler(Handler): |
| 93 | + NAME = "lzfse" |
| 94 | + |
| 95 | + PATTERNS = [ |
| 96 | + HexString("62 76 78 2D"), # "bvx-" uncompressed block |
| 97 | + HexString("62 76 78 31"), # "bvx1" LZFSE v1 compressed block (legacy) |
| 98 | + HexString("62 76 78 6E"), # "bvxn" LZVN compressed block |
| 99 | + HexString("62 76 78 32"), # "bvx2" LZFSE v2 compressed block |
| 100 | + ] |
| 101 | + |
| 102 | + EXTRACTOR = LZFSEExtractor() |
| 103 | + |
| 104 | + DOC = HandlerDoc( |
| 105 | + name="LZFSE", |
| 106 | + description="LZFSE is a lossless compression algorithm developed by Apple and open-sourced in 2016. It combines Lempel-Ziv back-references with Finite State Entropy coding and is the default compression format used in iOS and macOS firmware images.", |
| 107 | + handler_type=HandlerType.COMPRESSION, |
| 108 | + vendor="Apple", |
| 109 | + references=[ |
| 110 | + Reference( |
| 111 | + title="lzfse - Apple open-source LZFSE library", |
| 112 | + url="https://github.com/lzfse/lzfse", |
| 113 | + ), |
| 114 | + ], |
| 115 | + limitations=[], |
| 116 | + ) |
| 117 | + |
| 118 | + def calculate_chunk(self, file: File, start_offset: int) -> ValidChunk | None: |
| 119 | + # An LZFSE stream is a sequence of blocks terminated by an end-of-stream |
| 120 | + # block. Walk the blocks using each header's declared size instead of |
| 121 | + # scanning for "bvx$", which could otherwise be matched inside payload data. |
| 122 | + offset = start_offset |
| 123 | + magic = file[offset : offset + MAGIC_LEN] |
| 124 | + while magic != LZFSEMagic.END: |
| 125 | + if len(magic) < MAGIC_LEN: |
| 126 | + raise InvalidInputFormat("Truncated LZFSE stream: no end block") |
| 127 | + block_size = self._block_size(file, offset, magic) |
| 128 | + if block_size <= 0: |
| 129 | + raise InvalidInputFormat("Invalid LZFSE block size") |
| 130 | + offset += block_size |
| 131 | + magic = file[offset : offset + MAGIC_LEN] |
| 132 | + |
| 133 | + return ValidChunk(start_offset=start_offset, end_offset=offset + MAGIC_LEN) |
| 134 | + |
| 135 | + @staticmethod |
| 136 | + def _block_size(file: File, offset: int, magic: bytes) -> int: |
| 137 | + """Size in bytes of the LZFSE block at offset, including its header.""" |
| 138 | + file.seek(offset, io.SEEK_SET) |
| 139 | + match magic: |
| 140 | + case LZFSEMagic.UNCOMPRESSED: |
| 141 | + header = _parser.parse("lzfse_uncompressed_t", file, Endian.LITTLE) |
| 142 | + size = 8 + header.n_raw_bytes |
| 143 | + case LZFSEMagic.LZVN: |
| 144 | + header = _parser.parse("lzvn_compressed_t", file, Endian.LITTLE) |
| 145 | + size = 12 + header.n_payload_bytes |
| 146 | + case LZFSEMagic.LZFSE_V1: |
| 147 | + header = _parser.parse("lzfse_v1_t", file, Endian.LITTLE) |
| 148 | + size = ( |
| 149 | + _V1_HEADER_SIZE |
| 150 | + + header.n_literal_payload_bytes |
| 151 | + + header.n_lmd_payload_bytes |
| 152 | + ) |
| 153 | + case LZFSEMagic.LZFSE_V2: |
| 154 | + header = _parser.parse("lzfse_v2_t", file, Endian.LITTLE) |
| 155 | + # v2 packs the sizes into bit-fields across three uint64s. |
| 156 | + n_literal = (header.packed_fields_0 >> 20) & 0xFFFFF |
| 157 | + n_lmd = (header.packed_fields_1 >> 40) & 0xFFFFF |
| 158 | + header_size = header.packed_fields_2 & 0xFFFFFFFF |
| 159 | + size = header_size + n_literal + n_lmd |
| 160 | + case _: |
| 161 | + raise InvalidInputFormat(f"Unknown LZFSE block magic: {magic!r}") |
| 162 | + return size |
0 commit comments