|
| 1 | +"""MSI Handler |
| 2 | +
|
| 3 | +Extracts uses 7z but could migrate to a fully Python-based implementation: |
| 4 | +
|
| 5 | + https://github.com/nightlark/pymsi |
| 6 | + https://github.com/decalage2/olefile |
| 7 | +
|
| 8 | +As of v0.47, olefile does not handle padded MSIs properly so we re-implement |
| 9 | +CFBF header parsing and compute the archive size ourselves. |
| 10 | +""" |
| 11 | + |
| 12 | +import io |
| 13 | +import struct |
| 14 | +from typing import Optional |
| 15 | + |
| 16 | +from structlog import get_logger |
| 17 | + |
| 18 | +from unblob.extractors import Command |
| 19 | + |
| 20 | +from ...file_utils import OffsetFile |
| 21 | +from ...models import ( |
| 22 | + File, |
| 23 | + HandlerDoc, |
| 24 | + HandlerType, |
| 25 | + HexString, |
| 26 | + Reference, |
| 27 | + StructHandler, |
| 28 | + ValidChunk, |
| 29 | +) |
| 30 | + |
| 31 | +logger = get_logger() |
| 32 | + |
| 33 | + |
| 34 | +class MsiHandler(StructHandler): |
| 35 | + NAME = "msi" |
| 36 | + |
| 37 | + PATTERNS = [ |
| 38 | + HexString("D0 CF 11 E0 A1 B1 1A E1") |
| 39 | + ] |
| 40 | + C_DEFINITIONS = r""" |
| 41 | + typedef struct cfbf_header |
| 42 | + { |
| 43 | + // [offset from start (bytes), length (bytes)] |
| 44 | + uint8 signature[8]; // [00H,08] {0xd0, 0xcf, 0x11, 0xe0, 0xa1, 0xb1, |
| 45 | + // 0x1a, 0xe1} for current version |
| 46 | + uint8 clsid[16]; // [08H,16] reserved must be zero (WriteClassStg/ |
| 47 | + // GetClassFile uses root directory class id) |
| 48 | + uint16 minorVersion; // [18H,02] minor version of the format: 33 is |
| 49 | + // written by reference implementation |
| 50 | + uint16 dllVersion; // [1AH,02] major version of the dll/format: 3 for |
| 51 | + // 512-byte sectors, 4 for 4 KB sectors |
| 52 | + uint16 byteOrder; // [1CH,02] 0xFFFE: indicates Intel byte-ordering |
| 53 | + uint16 sectorShift; // [1EH,02] size of sectors in power-of-two; |
| 54 | + // typically 9 indicating 512-byte sectors |
| 55 | + uint16 miniSectorShift; // [20H,02] size of mini-sectors in power-of-two; |
| 56 | + // typically 6 indicating 64-byte mini-sectors |
| 57 | + uint16 reserved; // [22H,02] reserved, must be zero |
| 58 | + uint32 reserved1; // [24H,04] reserved, must be zero |
| 59 | + uint32 csectDir; // [28H,04] must be zero for 512-byte sectors, |
| 60 | + // number of SECTs in directory chain for 4 KB |
| 61 | + // sectors |
| 62 | + uint32 csectFat; // [2CH,04] number of SECTs in the FAT chain |
| 63 | + uint32 sectDirStart; // [30H,04] first SECT in the directory chain |
| 64 | + uint32 txSignature; // [34H,04] signature used for transactions; must |
| 65 | + // be zero. The reference implementation |
| 66 | + // does not support transactions |
| 67 | + uint32 miniSectorCutoff; // [38H,04] maximum size for a mini stream; |
| 68 | + // typically 4096 bytes |
| 69 | + uint32 sectMiniFatStart; // [3CH,04] first SECT in the MiniFAT chain |
| 70 | + uint32 csectMiniFat; // [40H,04] number of SECTs in the MiniFAT chain |
| 71 | + uint32 sectDifStart; // [44H,04] first SECT in the DIFAT chain |
| 72 | + uint32 csectDif; // [48H,04] number of SECTs in the DIFAT chain |
| 73 | + uint32 sectFat[109]; // [4CH,436] the SECTs of first 109 FAT sectors |
| 74 | + } cfbf_header_t; |
| 75 | + """ |
| 76 | + HEADER_STRUCT = "cfbf_header_t" |
| 77 | + |
| 78 | + EXTRACTOR = Command("7z", "x", "-p", "-y", "{inpath}", "-o{outdir}") |
| 79 | + |
| 80 | + DOC = HandlerDoc( |
| 81 | + name="MSI", |
| 82 | + description="Microsoft Installer (MSI) files are used for the installation, maintenance, and removal of software.", |
| 83 | + handler_type=HandlerType.ARCHIVE, |
| 84 | + vendor="Microsoft", |
| 85 | + references=[ |
| 86 | + Reference( |
| 87 | + title="MSI File Format Documentation", |
| 88 | + url="https://docs.microsoft.com/en-us/windows/win32/msi/overview-of-windows-installer", |
| 89 | + ), |
| 90 | + Reference( |
| 91 | + title="Compound File Binary Format", |
| 92 | + url="https://en.wikipedia.org/wiki/Compound_File_Binary_Format", |
| 93 | + ) |
| 94 | + ], |
| 95 | + limitations=[], |
| 96 | + ) |
| 97 | + |
| 98 | + def calculate_chunk(self, file: File, start_offset: int) -> Optional[ValidChunk]: |
| 99 | + file.seek(start_offset) |
| 100 | + header = self.parse_header(file) |
| 101 | + |
| 102 | + # Size of MSI is based on the maximum used sector. Need to walk the |
| 103 | + # DIFAT entries and find the maximum used sector to compute the size. |
| 104 | + sector_size = 2 ** header.sectorShift |
| 105 | + entries_per_sector = sector_size // 4 |
| 106 | + |
| 107 | + max_used_sector = 0 |
| 108 | + |
| 109 | + full_fat = [] |
| 110 | + for i, sect in enumerate(header.sectFat): |
| 111 | + # skip empty |
| 112 | + if sect == 0xFFFFFFFF: |
| 113 | + continue |
| 114 | + |
| 115 | + file.seek(start_offset + 512 + sect * sector_size) |
| 116 | + raw_sector = file.read(sector_size) |
| 117 | + entries = struct.unpack(f'<{entries_per_sector}I', raw_sector) |
| 118 | + |
| 119 | + base_sector_id = i * entries_per_sector |
| 120 | + for i in range(len(entries) - 1, -1, -1): |
| 121 | + if entries[i] == 0xFFFFFFFF: |
| 122 | + continue |
| 123 | + |
| 124 | + # Found the highest id on this page |
| 125 | + max_id = base_sector_id + i |
| 126 | + |
| 127 | + if max_id > max_used_sector: |
| 128 | + max_used_sector = max_id |
| 129 | + |
| 130 | + # Once we have found the first non-empty element, we are done |
| 131 | + # with all IDs in this sector |
| 132 | + break |
| 133 | + |
| 134 | + total_size = 512 + ((max_used_sector + 1) * sector_size) |
| 135 | + |
| 136 | + return ValidChunk( |
| 137 | + start_offset = start_offset, |
| 138 | + end_offset = start_offset + total_size, |
| 139 | + ) |
0 commit comments