From 67178d34afb575d20bbe505123e00dfd05fed903 Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Tue, 11 Aug 2026 10:51:37 +0800 Subject: [PATCH 1/8] feat: Optimize error reporting --- docs/design.md | 9 +- src/unpacker/bytecode.rs | 4 +- src/unpacker/dll.rs | 86 ++++++--- src/unpacker/exe.rs | 346 +++++++++++++++++++++++++++++++------ src/unpacker/mod.rs | 223 ++++++++++++++++++++++-- src/unpacker/parallel.rs | 21 ++- src/unpacker/primitives.rs | 88 ++++++++-- 7 files changed, 664 insertions(+), 113 deletions(-) diff --git a/docs/design.md b/docs/design.md index 7dd6799..d7d5864 100644 --- a/docs/design.md +++ b/docs/design.md @@ -12,7 +12,7 @@ The crate is split into a pure core and a thin CLI shell: - **`src/unpacker/`** — the core. Pure functions over byte slices: no file I/O, no environment access (beyond a few debugging overrides, see [development.md](development.md)), panic-free at the public boundary (all - internal panics are trapped and converted to `UnpackError::Corrupt`). This + internal panics are trapped and converted to `UnpackError::InternalPanic`). This is what the WebAssembly build embeds. - **`src/` (top level)** — the CLI shell: argument parsing, recursive folder scanning, per-run log file, progress bar, Explorer-friendly exit pause, and @@ -120,7 +120,12 @@ threads (WebAssembly) the sequential path is used automatically. The public API never panics: every pipeline runs under a `catch_unwind` wrapper (`catch_unpack`) that converts a trapped panic to -`UnpackError::Corrupt`, with the default panic hook transiently suppressed. +`UnpackError::InternalPanic`, including the Rust source location and panic +payload. The capture context is propagated into section worker threads; panics +outside an active unpack continue through the previously installed panic hook. +Expected validation failures use structured variants carrying the failed stage, +block index, table kind, or invalid range instead of collapsing unrelated causes +into a generic corruption error. Size requests are bounds-checked against a 1 GiB `MAX_IMAGE_SIZE` before allocation so a crafted header cannot abort the process with a huge allocation. In folder mode each file is isolated: one file's failure is logged diff --git a/src/unpacker/bytecode.rs b/src/unpacker/bytecode.rs index f662c6e..30c4a13 100644 --- a/src/unpacker/bytecode.rs +++ b/src/unpacker/bytecode.rs @@ -61,8 +61,8 @@ impl OpsLut { pub fn generate(data: &[u8], offset: u32) -> Option> { // Bounds-checked cursor: a corrupt `data_offset` (bad decrypt_data6 / the // alignment fallback) must yield `None`, not an out-of-bounds panic — the - // panic path would surface as a misleading `UnpackError::Corrupt` instead - // of the precise `BytecodeGenFailed`, and any future caller without a + // panic path would surface as a misleading `UnpackError::InternalPanic` instead + // of the precise `BytecodeGenerationFailed`, and any future caller without a // `catch_unwind` wrapper would abort outright. let mut pos = offset as usize; let mut next = move || { diff --git a/src/unpacker/dll.rs b/src/unpacker/dll.rs index 6bc0894..3cc69c8 100644 --- a/src/unpacker/dll.rs +++ b/src/unpacker/dll.rs @@ -13,9 +13,12 @@ //! CalculateChecksumWithSizeXor -> primitives::calculate_checksum //! CalculateCrc32 -> crc32::compute (via above) -use super::UnpackError; use super::bytecode::{Op, OpsLut, generate}; use super::primitives::{self, *}; +use super::{ + BufferOperation, BytecodeStage, DecompressionStage, DescriptorTable, SectionPipeline, + UnpackError, +}; /// Read a signed 32-bit little-endian value. fn get_i32(d: &[u8], offset: i32) -> i32 { @@ -68,6 +71,7 @@ fn decrypt_data4( key: i32, decomp_params: &[i32; 4], transform: Option<&[Op]>, + stage: DecompressionStage, ) -> Result<(), UnpackError> { let addr = get_i32(d, offset); let size = get_i32(d, offset + 4); @@ -95,7 +99,7 @@ fn decrypt_data4( size as u32, decompressed_size as u32, ) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed(stage)); } } Ok(()) @@ -218,7 +222,11 @@ fn decrypt_and_decompress_data( // Guard: need 16 bytes at section_data_offset in `d` let off = section_data_offset as usize; if off.saturating_add(16) > d.len() { - return Err(UnpackError::OutOfBounds(off)); + return Err(UnpackError::DescriptorOutOfBounds { + table: DescriptorTable::DllSectionBlocks, + offset: off, + image_len: d.len(), + }); } decrypt_data6_shift6(d, section_data_offset, 16); let dest_offset = get_i32(d, section_data_offset); @@ -246,10 +254,10 @@ fn decrypt_and_decompress_data( let lut = OpsLut::new(decrypt_func); let ko0 = decomp_params[0]; let ko2 = decomp_params[2]; - let ks_snap = - primitives::aes_schedule_snapshot(d, ko2 as u32).ok_or(UnpackError::Corrupt)?; + let ks_snap = primitives::aes_schedule_snapshot(d, ko2 as u32) + .ok_or(UnpackError::InvalidAesKeySchedule { offset: ko2 as u32 })?; let tab_snap = primitives::huffman_table_snapshot(d, ko0 as u32) - .ok_or(UnpackError::DecompressFailed)?; + .ok_or(UnpackError::InvalidHuffmanTable { offset: ko0 as u32 })?; let spans: Vec<(usize, usize)> = blocks .iter() .map(|b| { @@ -277,7 +285,10 @@ fn decrypt_and_decompress_data( b.size as u32, b.expected_crc as u32, ) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::SectionDecompressionFailed { + pipeline: SectionPipeline::Dll, + block: i, + }); } } Ok(()) @@ -292,7 +303,11 @@ fn decrypt_and_decompress_data( // decrypts 16 too, so guard 16 (an 8-byte guard would let // decrypt_data6_shift6 index past the end of a truncated descriptor). if off.saturating_add(16) > d.len() { - return Err(UnpackError::OutOfBounds(off)); + return Err(UnpackError::DescriptorOutOfBounds { + table: DescriptorTable::DllZeroFill, + offset: off, + image_len: d.len(), + }); } decrypt_data6_shift6(d, section_data_offset, 16); let zero_offset = get_i32(d, section_data_offset); @@ -305,7 +320,12 @@ fn decrypt_and_decompress_data( for i in 0..zero_size { let idx = (zero_offset + i) as usize; if idx >= d.len() { - return Err(UnpackError::OutOfBounds(idx)); + return Err(UnpackError::BufferRangeOutOfBounds { + operation: BufferOperation::ZeroFill, + offset: idx, + size: 1, + buffer_len: d.len(), + }); } d[idx] = 0; } @@ -328,8 +348,12 @@ pub fn unpack_dll_v(input: &[u8], verbose: bool) -> Result, UnpackError> } fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> { - if input.len() < 4096 { - return Err(UnpackError::InputTooShort(input.len())); + const HEADER_LEN: usize = 4128; + if input.len() < HEADER_LEN { + return Err(UnpackError::InputTooShort { + actual: input.len(), + required: HEADER_LEN, + }); } // `file_data` and `original_file_data` both borrow the same protected input. @@ -348,14 +372,17 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> } if !super::is_supported_magic(keys[1] as u32) { - return Err(UnpackError::DllUnpack( - "Not a Crackproof protected file (KONN magic mismatch)".into(), - )); + return Err(UnpackError::HeaderMagicMismatch { + found: keys[1] as u32, + }); } let pe_offset = get_i32(file_data, 60); if pe_offset < 0 || (pe_offset as usize).saturating_add(84) > file_data.len() { - return Err(UnpackError::DllUnpack("implausible PE offset".into())); + return Err(UnpackError::InvalidPeOffset { + offset: i64::from(pe_offset), + input_len: file_data.len(), + }); } // This pipeline is PE32+-only: its header fixups write the data // directories at PE32+ offsets (pe+144..180, pe+136 for the DD blob). On a @@ -363,14 +390,18 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> // structurally plausible but unloadable file. Reject early with a clear // error so `unpack_auto`'s EXE-pipeline fallback handles PE32 DLLs (that // path is PE32-aware — see run_pe32), instead of us mangling them here. - if get_i32(file_data, pe_offset + 24) & 0xFFFF != 0x20B { - return Err(UnpackError::DllUnpack( - "not a PE32+ image (the DLL pipeline handles 64-bit only)".into(), - )); + let optional_magic = get_u16(file_data, (pe_offset + 24) as u32); + if optional_magic != 0x20B { + return Err(UnpackError::UnsupportedDllPeMagic { + found: optional_magic, + }); } let size_of_image = get_i32(file_data, pe_offset + 80); if size_of_image <= 0 || size_of_image as u64 > super::MAX_IMAGE_SIZE { - return Err(UnpackError::DllUnpack("implausible SizeOfImage".into())); + return Err(UnpackError::InvalidImageSize { + size: i64::from(size_of_image), + max: super::MAX_IMAGE_SIZE, + }); } let mut out = vec![0u8; size_of_image as usize]; let base_offset = keys[6] - keys[3] + 0x2000; @@ -512,6 +543,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> table_val ^ checksum2 ^ (xor_accumulator as i32), &decomp_params, None, + DecompressionStage::DllCodeBlock1, )?; let addr3b = get_i32(&out, decrypted_addr1 + 3728); @@ -537,6 +569,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> crc_xored ^ (xor_accumulator as i32) ^ trailing_val, &decomp_params, None, + DecompressionStage::DllCodeBlock2, )?; let checksum3 = calculate_checksum(&out, (decrypted_addr1 + 3480) as u32) as i32; @@ -549,6 +582,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> (not_val ^ (xor_key as u32)) as i32, &decomp_params, None, + DecompressionStage::DllCodeBlock3, )?; let addr4 = get_i32(&out, addr4_offset); @@ -574,8 +608,9 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> lfsr_seed_val = lfsr_seed_val.wrapping_add(k); } - let decrypt_func = generate(&out, lfsr as u32) - .ok_or_else(|| UnpackError::DllUnpack("Failed to build decryption expression".into()))?; + let decrypt_func = generate(&out, lfsr as u32).ok_or(UnpackError::BytecodeGenerationFailed( + BytecodeStage::DllPrimaryDecryptor, + ))?; let addr5_offset = decrypted_addr1 + 3840; let addr5 = get_i32(&out, addr5_offset); @@ -585,6 +620,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> lfsr_seed_val ^ xor_key ^ checksum4, &decomp_params, Some(&decrypt_func), + DecompressionStage::DllCodeBlock4, )?; if verbose { println!("[7/9] Decrypting code block 4 (addr5)..."); @@ -603,9 +639,9 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> let lfsr2 = metadata_offset + 88; decrypt_data6(&mut out, lfsr2 as u32); - let decrypt_func2 = generate(&out, lfsr2 as u32).ok_or_else(|| { - UnpackError::DllUnpack("Failed to build second decryption expression".into()) - })?; + let decrypt_func2 = generate(&out, lfsr2 as u32).ok_or( + UnpackError::BytecodeGenerationFailed(BytecodeStage::DllSectionDecryptor), + )?; let section_image_base = 4095 - get_i32(original_file_data, 4224); let section_data_offset = get_i32(&out, addr5 + 11976); diff --git a/src/unpacker/exe.rs b/src/unpacker/exe.rs index cdf0470..a76cf1e 100644 --- a/src/unpacker/exe.rs +++ b/src/unpacker/exe.rs @@ -2,13 +2,121 @@ use super::bytecode::{Op, OpsLut, generate}; use super::primitives; use super::primitives::*; -#[derive(Debug, thiserror::Error)] -pub enum UnpackError { - #[error("input too short for header (need at least 4096 bytes, got {0})")] - InputTooShort(usize), +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum DecompressionStage { + ExeStage3, + ExeStage3Secondary, + ExeStage4, + ExeStage5, + Pe32FourthStage, + Pe32FifthStage, + Pe32SeventhStage, + DllCodeBlock1, + DllCodeBlock2, + DllCodeBlock3, + DllCodeBlock4, +} - #[error("info[1] mismatch — corrupt data or wrong offset")] - HeaderMismatch, +impl std::fmt::Display for DecompressionStage { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExeStage3 => "EXE stage3", + Self::ExeStage3Secondary => "EXE secondary stage3", + Self::ExeStage4 => "EXE stage4", + Self::ExeStage5 => "EXE stage5", + Self::Pe32FourthStage => "PE32 fourth stage", + Self::Pe32FifthStage => "PE32 fifth stage", + Self::Pe32SeventhStage => "PE32 seventh stage", + Self::DllCodeBlock1 => "DLL code block 1", + Self::DllCodeBlock2 => "DLL code block 2", + Self::DllCodeBlock3 => "DLL code block 3", + Self::DllCodeBlock4 => "DLL code block 4", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum BytecodeStage { + ExeStage4, + ExeStage5, + Pe32CustomDecryptor, + Pe32FileDecryptor, + DllPrimaryDecryptor, + DllSectionDecryptor, +} + +impl std::fmt::Display for BytecodeStage { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExeStage4 => "EXE stage4", + Self::ExeStage5 => "EXE stage5", + Self::Pe32CustomDecryptor => "PE32 custom decryptor", + Self::Pe32FileDecryptor => "PE32 file decryptor", + Self::DllPrimaryDecryptor => "DLL primary decryptor", + Self::DllSectionDecryptor => "DLL section decryptor", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum SectionPipeline { + ExePe32Plus, + ExePe32, + Dll, +} + +impl std::fmt::Display for SectionPipeline { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExePe32Plus => "PE32+ EXE", + Self::ExePe32 => "PE32 EXE", + Self::Dll => "DLL", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum DescriptorTable { + DllSectionBlocks, + DllZeroFill, +} + +impl std::fmt::Display for DescriptorTable { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::DllSectionBlocks => "DLL section-block", + Self::DllZeroFill => "DLL zero-fill", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum BufferOperation { + Read, + CopySource, + CopyDestination, + ZeroFill, +} + +impl std::fmt::Display for BufferOperation { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::Read => "read", + Self::CopySource => "copy source", + Self::CopyDestination => "copy destination", + Self::ZeroFill => "zero-fill", + }) + } +} + +#[derive(Debug, Clone, PartialEq, Eq, thiserror::Error)] +#[non_exhaustive] +pub enum UnpackError { + #[error("input too short (need at least {required} bytes, got {actual})")] + InputTooShort { actual: usize, required: usize }, + + #[error("decrypted header magic mismatch (got 0x{found:08X})")] + HeaderMagicMismatch { found: u32 }, #[error("anchor field not found — corrupt data or wrong offset")] AnchorNotFound, @@ -22,23 +130,42 @@ pub enum UnpackError { #[error("table_start not found — corrupt data or wrong offset")] TableStartNotFound, - #[error("stage4 bytecode generation failed — corrupt data or wrong offset")] - BytecodeGenFailed, + #[error("{0} bytecode generation failed — corrupt data or wrong offset")] + BytecodeGenerationFailed(BytecodeStage), #[error("stage5 marker not found — this build's layout is not supported by this unpacker")] Stage5MarkerNotFound, - #[error("stage5 bytecode generation failed — corrupt data or wrong offset")] - Stage5BytecodeGenFailed, - - #[error("DLL unpack failed: {0}")] - DllUnpack(String), - #[error("not a Crackproof-protected file")] NotCrackproof, - #[error("out-of-bounds access at offset {0}")] - OutOfBounds(usize), + #[error("invalid PE header offset {offset} for {input_len}-byte input")] + InvalidPeOffset { offset: i64, input_len: usize }, + + #[error("DLL pipeline requires PE32+ optional-header magic, got 0x{found:04X}")] + UnsupportedDllPeMagic { found: u16 }, + + #[error("invalid SizeOfImage {size}; expected 1..={max}")] + InvalidImageSize { size: i64, max: u64 }, + + #[error( + "{operation} range out of bounds (offset {offset}, size {size}, buffer length {buffer_len})" + )] + BufferRangeOutOfBounds { + operation: BufferOperation, + offset: usize, + size: usize, + buffer_len: usize, + }, + + #[error( + "{table} descriptor out of bounds (offset {offset}, size 16, image length {image_len})" + )] + DescriptorOutOfBounds { + table: DescriptorTable, + offset: usize, + image_len: usize, + }, #[error("PE32 tbl not found — corrupt data or wrong offset")] Pe32TblNotFound, @@ -49,20 +176,52 @@ pub enum UnpackError { #[error("PE32 customDecryptor not found in sevenStage")] Pe32CustomDecryptorNotFound, - #[error("PE32 stage bytecode generation failed")] - Pe32BytecodeGenFailed, - #[error("PE32 eighthStageKey not found")] Pe32EighthKeyNotFound, #[error("PE32 file LFSR not found in eighthStage")] Pe32FileLfsrNotFound, - #[error("decompression failed — corrupt data or wrong offset")] - DecompressFailed, + #[error("{0} decompression failed — corrupt data or wrong offset")] + StageDecompressionFailed(DecompressionStage), - #[error("input is corrupt or not a supported Crackproof layout")] - Corrupt, + #[error("{pipeline} section block {block} decompression failed")] + SectionDecompressionFailed { + pipeline: SectionPipeline, + block: usize, + }, + + #[error("AES key schedule is outside the image at offset {offset}")] + InvalidAesKeySchedule { offset: u32 }, + + #[error("Huffman table is outside the image at offset {offset}")] + InvalidHuffmanTable { offset: u32 }, + + #[error( + "PE32 second-stage range is invalid (offset {offset}, size {size}, image length {image_len})" + )] + Pe32SecondStageRangeInvalid { + offset: u32, + size: u32, + image_len: usize, + }, + + #[error("PE32 relocation-data descriptor not found")] + Pe32RelocationDataNotFound, + + #[error("file decryptor candidate failed structural validation")] + FileDecryptorValidationFailed, + + #[error("PE32 memory image could not be rebuilt as a file-layout PE")] + Pe32OutputLayoutInvalid, + + #[error("internal panic at {file}:{line}:{column}: {message}")] + InternalPanic { + message: String, + file: String, + line: u32, + column: u32, + }, } pub fn unpack(input: &[u8]) -> Result, UnpackError> { @@ -93,8 +252,12 @@ fn prot_rva_to_off(file_data: &[u8], pe_header: u32, rva: u32) -> Option { } pub fn unpack_v(input: &[u8], verbose: bool) -> Result, UnpackError> { - if input.len() < 4096 { - return Err(UnpackError::InputTooShort(input.len())); + const HEADER_LEN: usize = 4128; + if input.len() < HEADER_LEN { + return Err(UnpackError::InputTooShort { + actual: input.len(), + required: HEADER_LEN, + }); } // The pipeline chases offsets read out of the decrypted image; on a // truncated/garbled-but-detected file those run out of bounds. Trap any @@ -376,13 +539,25 @@ impl<'a> Unpacker<'a> { println!(" info[7] = 0x{:08X}", u.info[7]); } if !super::is_supported_magic(u.info[1]) { - return Err(UnpackError::HeaderMismatch); + return Err(UnpackError::HeaderMagicMismatch { found: u.info[1] }); } let pe_off = get_u32(u.file_data, 60); + if (pe_off as usize) + .checked_add(84) + .is_none_or(|end| end > u.file_data.len()) + { + return Err(UnpackError::InvalidPeOffset { + offset: i64::from(pe_off), + input_len: u.file_data.len(), + }); + } let size_of_image = get_u32(u.file_data, pe_off.wrapping_add(80)); if size_of_image == 0 || size_of_image as u64 > super::MAX_IMAGE_SIZE { - return Err(UnpackError::Corrupt); + return Err(UnpackError::InvalidImageSize { + size: i64::from(size_of_image), + max: super::MAX_IMAGE_SIZE, + }); } u.decompressed = vec![0u8; size_of_image as usize]; u.decrypt_size = u.info[6].wrapping_sub(u.info[3]).wrapping_add(8192); @@ -671,7 +846,9 @@ impl<'a> Unpacker<'a> { println!(" stage3 = 0x{:08X}", stage3_field); } if !u.decrypt_and_decompress_data(at1, xor_acc ^ chk2 ^ accum, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::ExeStage3, + )); } let at2 = stage1.wrapping_add(stage2_off.wrapping_add(104)); @@ -689,7 +866,9 @@ impl<'a> Unpacker<'a> { .unwrap_or_else(|| stage3_field.wrapping_add(4692)); let v4_val = get_u32(&u.decompressed, v4); if !u.decrypt_and_decompress_data(at2, xor_acc ^ chk3 ^ v4_val, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::ExeStage3Secondary, + )); } let chk4 = u.calculate_checksum(stage1.wrapping_add(chk_src_start.wrapping_add(16))); @@ -708,7 +887,9 @@ impl<'a> Unpacker<'a> { println!(" stage4 = 0x{:08X}", stage4_field); } if !u.decrypt_and_decompress_data(at3, xor_acc ^ chk4 ^ v5_val, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::ExeStage4, + )); } // Inside stage4, two locations vary by build: @@ -758,7 +939,9 @@ impl<'a> Unpacker<'a> { let ops1 = match generate(&u.decompressed, data_offset) { Some(v) => v, None => { - return Err(UnpackError::BytecodeGenFailed); + return Err(UnpackError::BytecodeGenerationFailed( + BytecodeStage::ExeStage4, + )); } }; @@ -771,7 +954,9 @@ impl<'a> Unpacker<'a> { println!(" stage5 = 0x{:08X}", stage5_field); } if !u.decrypt_and_decompress_data(at4, xor_acc ^ chk4 ^ chk5 ^ accum2, Some(&ops1)) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::ExeStage5, + )); } // Inside stage5, the loader stores a table of (ptr, size) pairs at a @@ -921,7 +1106,9 @@ impl<'a> Unpacker<'a> { let ops2 = match generate(&u.decompressed, data_offset2) { Some(v) => v, None => { - return Err(UnpackError::Stage5BytecodeGenFailed); + return Err(UnpackError::BytecodeGenerationFailed( + BytecodeStage::ExeStage5, + )); } }; // The new layout picked its file decryptor by distance (no marker, no @@ -930,7 +1117,7 @@ impl<'a> Unpacker<'a> { // garbling into the output without any error (see the validator). let rebase = (!get_u32(u.file_data, 4224)).wrapping_add(4096); if new_layout && !u.new_layout_file_ops_validate(walk4_slot, &ops2, rebase) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::FileDecryptorValidationFailed); } let at6 = walk4_slot; @@ -980,9 +1167,9 @@ impl<'a> Unpacker<'a> { // Snapshot the shared tables before the fan-out: workers get // disjoint span slices, not the whole buffer. let ks_snap = primitives::aes_schedule_snapshot(&u.decompressed, ko[2]) - .ok_or(UnpackError::Corrupt)?; + .ok_or(UnpackError::InvalidAesKeySchedule { offset: ko[2] })?; let tab_snap = primitives::huffman_table_snapshot(&u.decompressed, ko[0]) - .ok_or(UnpackError::DecompressFailed)?; + .ok_or(UnpackError::InvalidHuffmanTable { offset: ko[0] })?; let spans: Vec<(usize, usize)> = blocks .iter() .map(|b| { @@ -1009,7 +1196,10 @@ impl<'a> Unpacker<'a> { b.len, b.plain_len, ) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::SectionDecompressionFailed { + pipeline: SectionPipeline::ExePe32Plus, + block: i, + }); } } Ok(()) @@ -1475,7 +1665,7 @@ impl<'a> Unpacker<'a> { /// whose fileCS pointer sits just past `info[3]`), not by content. A /// coincidental LFSR-shaped block at a shorter distance would decode to a /// wrong `ops2` translate and silently garble every section block — raw - /// blocks never hit `DecompressFailed`, so the failure would ship as a + /// raw blocks never enter the decompressor, so the failure would ship as a /// plausible but wrong image. Replay the first *compressed* block's full /// transform (raw copy, AES, translate, decompress) on a snapshot and /// require decompression to succeed; restore the region afterwards. @@ -1837,7 +2027,11 @@ impl<'a> Unpacker<'a> { let ss_lo = ss as usize; let ss_hi = ss_lo.wrapping_add(ss_size as usize); if ss_hi < ss_lo || ss_hi > self.decompressed.len() { - return Err(UnpackError::Corrupt); + return Err(UnpackError::Pe32SecondStageRangeInvalid { + offset: ss, + size: ss_size, + image_len: self.decompressed.len(), + }); } let ss_ct: Vec = self.decompressed[ss_lo..ss_hi].to_vec(); // PE32 data dir 5 (BaseReloc) = optional_header(pe+24) + 0x60 + 5*8 = pe+0xA0. @@ -1872,7 +2066,7 @@ impl<'a> Unpacker<'a> { } } if !found { - return Err(UnpackError::Corrupt); + return Err(UnpackError::Pe32RelocationDataNotFound); } if verbose { println!( @@ -1996,7 +2190,9 @@ impl<'a> Unpacker<'a> { let forth_addr = dp_base.wrapping_add(0x40); let fk = header_checksum ^ second_stage_cs ^ forth_stage_key; if !self.decrypt_and_decompress_data(forth_addr, fk, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::Pe32FourthStage, + )); } // ---- FifthStage ---- @@ -2012,7 +2208,9 @@ impl<'a> Unpacker<'a> { ); let fk5 = header_checksum ^ forth_cs ^ fifth_key; if !self.decrypt_and_decompress_data(fifth_addr, fk5, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::Pe32FifthStage, + )); } // ---- SevenStage ---- @@ -2035,7 +2233,9 @@ impl<'a> Unpacker<'a> { ); let fk7 = header_checksum ^ fifth_cs ^ seven_key; if !self.decrypt_and_decompress_data(seven_addr, fk7, None) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::StageDecompressionFailed( + DecompressionStage::Pe32SeventhStage, + )); } // ---- EighthStage ---- @@ -2063,8 +2263,9 @@ impl<'a> Unpacker<'a> { .ok_or(UnpackError::Pe32CustomDecryptorNotFound)?; let custom_dec_addr = seven_start_actual.wrapping_add(custom_dec_off); self.decrypt_data6(custom_dec_addr); - let custom_ops = generate(&self.decompressed, custom_dec_addr) - .ok_or(UnpackError::Pe32BytecodeGenFailed)?; + let custom_ops = generate(&self.decompressed, custom_dec_addr).ok_or( + UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32CustomDecryptor), + )?; let seven_cs = self.calculate_checksum(seven_stage_cs_addr); let eighth_addr = dp_base.wrapping_add(0xC0); @@ -2323,7 +2524,7 @@ impl<'a> Unpacker<'a> { // "legacy loose scan" picked the nearest LFSR-shaped block // by offset distance without any validation — that is // exactly how a wrong file_ops got applied to every data - // block (uncompressed blocks never hit DecompressFailed), + // block (uncompressed blocks never enter the decompressor), // producing a plausible but fully wrong image (the PE32 // .text scramble root cause). Trial-and-validate or error. return Err(UnpackError::Pe32FileLfsrNotFound); @@ -2362,8 +2563,9 @@ impl<'a> Unpacker<'a> { }; let file_dec_addr = eighth_start.wrapping_add(lfsr_off); self.decrypt_data6(file_dec_addr); - let file_ops = generate(&self.decompressed, file_dec_addr) - .ok_or(UnpackError::Pe32BytecodeGenFailed)?; + let file_ops = generate(&self.decompressed, file_dec_addr).ok_or( + UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32FileDecryptor), + )?; // ---- PE32 metadata: EP and data dirs from info[3] ---- let test_val = get_u32(&self.decompressed, info3.wrapping_add(0x10)); @@ -2448,9 +2650,9 @@ impl<'a> Unpacker<'a> { let clean = &self.file_data; let ko = self.key_offsets; let ks_snap = primitives::aes_schedule_snapshot(&self.decompressed, ko[2]) - .ok_or(UnpackError::Corrupt)?; + .ok_or(UnpackError::InvalidAesKeySchedule { offset: ko[2] })?; let tab_snap = primitives::huffman_table_snapshot(&self.decompressed, ko[0]) - .ok_or(UnpackError::DecompressFailed)?; + .ok_or(UnpackError::InvalidHuffmanTable { offset: ko[0] })?; let spans: Vec<(usize, usize)> = blocks .iter() .map(|b| { @@ -2472,7 +2674,10 @@ impl<'a> Unpacker<'a> { if !primitives::decompress_tbl( &tab_snap, span, rel as u32, rel as u32, b.ssz, b.dsz, ) { - return Err(UnpackError::DecompressFailed); + return Err(UnpackError::SectionDecompressionFailed { + pipeline: SectionPipeline::ExePe32, + block: i, + }); } } Ok(()) @@ -2805,8 +3010,43 @@ impl<'a> Unpacker<'a> { if !is_dll && !primitives::pe32_imports_already_match_idata_layout(&mut out, pe_off) { primitives::move_pe32_imports_to_kmiat(&mut out, pe_off); } - let compact = - primitives::compact_memory_image_to_pe(&out, pe_off).ok_or(UnpackError::Corrupt)?; + let compact = primitives::compact_memory_image_to_pe(&out, pe_off) + .ok_or(UnpackError::Pe32OutputLayoutInvalid)?; Ok(compact) } } + +#[cfg(test)] +mod error_tests { + use super::*; + + #[test] + fn short_input_reports_actual_and_required_lengths() { + let error = unpack(&[0; 4096]).expect_err("header must be rejected"); + assert_eq!( + error, + UnpackError::InputTooShort { + actual: 4096, + required: 4128, + } + ); + } + + #[test] + fn structured_errors_include_stage_and_block_context() { + let stage = UnpackError::StageDecompressionFailed(DecompressionStage::ExeStage4); + assert_eq!( + stage.to_string(), + "EXE stage4 decompression failed — corrupt data or wrong offset" + ); + + let block = UnpackError::SectionDecompressionFailed { + pipeline: SectionPipeline::ExePe32, + block: 7, + }; + assert_eq!( + block.to_string(), + "PE32 EXE section block 7 decompression failed" + ); + } +} diff --git a/src/unpacker/mod.rs b/src/unpacker/mod.rs index e4b9c0a..2db55d3 100644 --- a/src/unpacker/mod.rs +++ b/src/unpacker/mod.rs @@ -9,8 +9,14 @@ pub(crate) mod parallel; pub(crate) mod primitives; mod tables; +use std::cell::RefCell; +use std::sync::{Arc, Mutex}; + pub use dll::{unpack_dll, unpack_dll_v}; -pub use exe::{UnpackError, unpack as unpack_exe, unpack_v as unpack_exe_v}; +pub use exe::{ + BufferOperation, BytecodeStage, DecompressionStage, DescriptorTable, SectionPipeline, + UnpackError, unpack as unpack_exe, unpack_v as unpack_exe_v, +}; pub use integrity::{IntegrityReport, check as check_integrity}; /// Maximum plausible PE `SizeOfImage` we are willing to allocate a zero buffer @@ -20,11 +26,139 @@ pub use integrity::{IntegrityReport, check as check_integrity}; /// Real protected binaries are far below this. pub(crate) const MAX_IMAGE_SIZE: u64 = 1 << 30; // 1 GiB +#[derive(Clone)] +pub(crate) struct PanicCapture(Arc>>); + +#[derive(Clone)] +struct PanicDetails { + message: String, + file: String, + line: u32, + column: u32, +} + +thread_local! { + static ACTIVE_PANIC_CAPTURE: RefCell> = const { RefCell::new(None) }; +} + +struct PanicCaptureGuard(Option); + +impl Drop for PanicCaptureGuard { + fn drop(&mut self) { + ACTIVE_PANIC_CAPTURE.with(|slot| { + slot.replace(self.0.take()); + }); + } +} + +impl PanicCapture { + fn new() -> Self { + Self(Arc::new(Mutex::new(None))) + } + + #[cfg(not(target_arch = "wasm32"))] + fn record(&self, info: &std::panic::PanicHookInfo<'_>) { + let location = info.location(); + let details = PanicDetails { + message: panic_message(info.payload()), + file: location + .map(|value| value.file().to_owned()) + .unwrap_or_else(|| "".to_owned()), + line: location.map_or(0, std::panic::Location::line), + column: location.map_or(0, std::panic::Location::column), + }; + let mut captured = self + .0 + .lock() + .unwrap_or_else(std::sync::PoisonError::into_inner); + if captured.is_none() { + *captured = Some(details); + } + } + + fn into_error(self, payload: &(dyn std::any::Any + Send)) -> UnpackError { + let details = self + .0 + .lock() + .unwrap_or_else(std::sync::PoisonError::into_inner) + .clone() + .unwrap_or_else(|| PanicDetails { + message: panic_message(payload), + file: "".to_owned(), + line: 0, + column: 0, + }); + UnpackError::InternalPanic { + message: details.message, + file: details.file, + line: details.line, + column: details.column, + } + } + + fn merge_from(&self, other: &Self) { + let details = other + .0 + .lock() + .unwrap_or_else(std::sync::PoisonError::into_inner) + .clone(); + let Some(details) = details else { return }; + let mut captured = self + .0 + .lock() + .unwrap_or_else(std::sync::PoisonError::into_inner); + if captured.is_none() { + *captured = Some(details); + } + } +} + +fn panic_message(payload: &(dyn std::any::Any + Send)) -> String { + if let Some(message) = payload.downcast_ref::<&str>() { + (*message).to_owned() + } else if let Some(message) = payload.downcast_ref::() { + message.clone() + } else { + "non-string panic payload".to_owned() + } +} + +#[cfg(not(target_arch = "wasm32"))] +fn install_panic_capture_hook() { + static INSTALL: std::sync::Once = std::sync::Once::new(); + INSTALL.call_once(|| { + let previous = std::panic::take_hook(); + std::panic::set_hook(Box::new(move |info| { + let capture = ACTIVE_PANIC_CAPTURE + .try_with(|slot| slot.borrow().clone()) + .ok() + .flatten(); + if let Some(capture) = capture { + capture.record(info); + } else { + previous(info); + } + })); + }); +} + +#[cfg(target_arch = "wasm32")] +fn install_panic_capture_hook() {} + +pub(crate) fn current_panic_capture() -> Option { + ACTIVE_PANIC_CAPTURE.with(|slot| slot.borrow().clone()) +} + +pub(crate) fn with_panic_capture(capture: Option, f: impl FnOnce() -> R) -> R { + let previous = ACTIVE_PANIC_CAPTURE.with(|slot| slot.replace(capture)); + let _guard = PanicCaptureGuard(previous); + f() +} + /// Run an unpack pipeline, converting any internal panic into a clean -/// [`UnpackError::Corrupt`] so the public API stays panic-free on any input -/// (truncated/garbled files chase offsets out of bounds). The default panic -/// hook is suppressed transiently so a trapped panic does not spill a -/// backtrace to stderr. +/// [`UnpackError::InternalPanic`] so the public API stays panic-free on any input +/// (truncated/garbled files chase offsets out of bounds). The panic location and +/// payload are captured for diagnostics without printing a backtrace to stderr. /// /// Note: allocation *failures* abort the process and are NOT caught here; size /// requests are bounds-checked against [`MAX_IMAGE_SIZE`] before allocating. @@ -32,17 +166,17 @@ pub(crate) fn catch_unpack(f: F) -> Result, UnpackError> where F: FnOnce() -> Result, UnpackError>, { - // Hook suppression is skipped on wasm: the prebuilt std cannot unwind - // there, so a panic traps immediately — and the suppressed hook would - // hide the panic message, leaving a bare `unreachable` with no clue. - #[cfg(not(target_arch = "wasm32"))] - let prev = std::panic::take_hook(); - #[cfg(not(target_arch = "wasm32"))] - std::panic::set_hook(Box::new(|_| {})); - let r = std::panic::catch_unwind(std::panic::AssertUnwindSafe(f)); - #[cfg(not(target_arch = "wasm32"))] - std::panic::set_hook(prev); - r.unwrap_or(Err(UnpackError::Corrupt)) + // Hook capture is skipped on wasm: the prebuilt std cannot unwind there, + // so a panic traps immediately. The Web Worker boundary reports that trap. + install_panic_capture_hook(); + let capture = PanicCapture::new(); + let r = with_panic_capture(Some(capture.clone()), || { + std::panic::catch_unwind(std::panic::AssertUnwindSafe(f)) + }); + match r { + Ok(result) => result, + Err(payload) => Err(capture.into_error(payload.as_ref())), + } } /// Crackproof header magic stored in `keys[1]`/`info[1]`. @@ -208,3 +342,60 @@ pub fn unpack_auto_v(input: &[u8], verbose: bool) -> Result<(Kind, Vec), Unp }; Ok((detected.kind, out)) } + +#[cfg(test)] +mod tests { + use super::*; + + #[cfg(not(target_arch = "wasm32"))] + #[test] + fn caught_panic_reports_location_and_message() { + let error = catch_unpack(|| -> Result, UnpackError> { + panic!("test panic"); + }) + .expect_err("panic must become an error"); + let UnpackError::InternalPanic { + message, + file, + line, + column, + } = error + else { + panic!("unexpected error: {error}"); + }; + assert_eq!(message, "test panic"); + assert!(file.ends_with("src/unpacker/mod.rs") || file.ends_with("src\\unpacker\\mod.rs")); + assert!(line > 0); + assert!(column > 0); + } + + #[cfg(not(target_arch = "wasm32"))] + #[test] + fn worker_panic_keeps_the_worker_source_location() { + let error = catch_unpack(|| -> Result, UnpackError> { + let capture = current_panic_capture(); + let result = std::thread::spawn(move || { + with_panic_capture(capture, || panic!("worker panic")); + }) + .join(); + if let Err(payload) = result { + std::panic::resume_unwind(payload); + } + Ok(Vec::new()) + }) + .expect_err("worker panic must become an error"); + let UnpackError::InternalPanic { + message, + file, + line, + column, + } = error + else { + panic!("unexpected error: {error}"); + }; + assert_eq!(message, "worker panic"); + assert!(file.ends_with("src/unpacker/mod.rs") || file.ends_with("src\\unpacker\\mod.rs")); + assert!(line > 0); + assert!(column > 0); + } +} diff --git a/src/unpacker/parallel.rs b/src/unpacker/parallel.rs index e6e0a8a..5f20d3d 100644 --- a/src/unpacker/parallel.rs +++ b/src/unpacker/parallel.rs @@ -46,7 +46,7 @@ pub(crate) fn thread_cap() -> usize { /// /// Returns the first `Err` any block produces; re-raises the first block panic /// on the calling thread (so the pipeline's existing `catch_unpack` still -/// converts it to `UnpackError::Corrupt`). +/// converts it to `UnpackError::InternalPanic`). pub(crate) fn parallel_for( buf: &mut [u8], spans: &[(usize, usize)], @@ -125,6 +125,7 @@ where let stop = AtomicBool::new(false); let first_err: Mutex> = Mutex::new(None); let first_panic: Mutex>> = Mutex::new(None); + let panic_capture = super::current_panic_capture(); std::thread::scope(|scope| { for _ in 0..workers { @@ -133,6 +134,7 @@ where let first_err = &first_err; let first_panic = &first_panic; let f = &f; + let panic_capture = panic_capture.clone(); scope.spawn(move || { loop { if stop.load(Ordering::Relaxed) { @@ -141,9 +143,15 @@ where let next = iter.lock().unwrap().next(); let Some((i, piece)) = next else { break }; let span = piece.unwrap(); - let r = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| { - f(i, spans[i].0, span) - })); + // Keep details local until this panic wins `first_panic`; + // otherwise simultaneous workers could pair one worker's + // location with another worker's propagated payload. + let block_capture = panic_capture.as_ref().map(|_| super::PanicCapture::new()); + let r = super::with_panic_capture(block_capture.clone(), || { + std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| { + f(i, spans[i].0, span) + })) + }); match r { Ok(Ok(())) => {} Ok(Err(e)) => { @@ -157,6 +165,11 @@ where Err(panic) => { let mut slot = first_panic.lock().unwrap(); if slot.is_none() { + if let (Some(parent), Some(block)) = + (&panic_capture, &block_capture) + { + parent.merge_from(block); + } *slot = Some(panic); } stop.store(true, Ordering::Relaxed); diff --git a/src/unpacker/primitives.rs b/src/unpacker/primitives.rs index f003261..1a8ab7f 100644 --- a/src/unpacker/primitives.rs +++ b/src/unpacker/primitives.rs @@ -69,9 +69,22 @@ pub(crate) fn write_u32(data: &mut [u8], offset: u32, value: u32) { #[allow(dead_code)] pub(crate) fn try_u32(d: &[u8], off: usize) -> Result { - d.get(off..off + 4) + let end = off + .checked_add(4) + .ok_or(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::Read, + offset: off, + size: 4, + buffer_len: d.len(), + })?; + d.get(off..end) .map(|s| u32::from_le_bytes(s.try_into().unwrap())) - .ok_or(super::UnpackError::OutOfBounds(off)) + .ok_or(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::Read, + offset: off, + size: 4, + buffer_len: d.len(), + }) } #[allow(dead_code)] @@ -79,7 +92,7 @@ pub(crate) fn try_i32(d: &[u8], off: usize) -> Result { try_u32(d, off).map(|v| v as i32) } -/// Checked copy: returns OutOfBounds if src or dst ranges exceed their respective slices. +/// Checked copy with distinct source and destination range errors. pub(crate) fn try_copy_from_slice( dst: &mut [u8], dst_off: usize, @@ -87,17 +100,39 @@ pub(crate) fn try_copy_from_slice( src: &[u8], src_off: usize, ) -> Result<(), super::UnpackError> { - let dst_end = dst_off - .checked_add(dst_len) - .ok_or(super::UnpackError::OutOfBounds(dst_off))?; - let src_end = src_off - .checked_add(dst_len) - .ok_or(super::UnpackError::OutOfBounds(src_off))?; + let dst_end = + dst_off + .checked_add(dst_len) + .ok_or(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::CopyDestination, + offset: dst_off, + size: dst_len, + buffer_len: dst.len(), + })?; + let src_end = + src_off + .checked_add(dst_len) + .ok_or(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::CopySource, + offset: src_off, + size: dst_len, + buffer_len: src.len(), + })?; if dst_end > dst.len() { - return Err(super::UnpackError::OutOfBounds(dst_off)); + return Err(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::CopyDestination, + offset: dst_off, + size: dst_len, + buffer_len: dst.len(), + }); } if src_end > src.len() { - return Err(super::UnpackError::OutOfBounds(src_off)); + return Err(super::UnpackError::BufferRangeOutOfBounds { + operation: super::BufferOperation::CopySource, + offset: src_off, + size: dst_len, + buffer_len: src.len(), + }); } dst[dst_off..dst_end].copy_from_slice(&src[src_off..src_end]); Ok(()) @@ -2005,6 +2040,37 @@ fn score_dd8_shift( mod tests { use super::*; + #[test] + fn checked_copy_distinguishes_source_and_destination_ranges() { + let mut short_destination = [0u8; 2]; + let source = [1u8; 4]; + let error = try_copy_from_slice(&mut short_destination, 0, 3, &source, 0) + .expect_err("destination must be rejected"); + assert!(matches!( + error, + super::super::UnpackError::BufferRangeOutOfBounds { + operation: super::super::BufferOperation::CopyDestination, + offset: 0, + size: 3, + buffer_len: 2, + } + )); + + let mut destination = [0u8; 4]; + let short_source = [1u8; 2]; + let error = try_copy_from_slice(&mut destination, 0, 3, &short_source, 0) + .expect_err("source must be rejected"); + assert!(matches!( + error, + super::super::UnpackError::BufferRangeOutOfBounds { + operation: super::super::BufferOperation::CopySource, + offset: 0, + size: 3, + buffer_len: 2, + } + )); + } + #[test] fn aes_ks_variant_matches_single_buffer() { // Random-ish key schedule at ko and data block; both variants must From a89900a812cf422ce5641bea1c8501dbb44bf885 Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Tue, 11 Aug 2026 11:42:12 +0800 Subject: [PATCH 2/8] fix(unpacker): refine layout validation and diagnostics --- docs/design.md | 14 +- samples/README.md | 6 +- src/unpacker/dll.rs | 22 ++- src/unpacker/exe.rs | 302 ++++++++++++++++++++++++++++++------- src/unpacker/mod.rs | 14 +- src/unpacker/primitives.rs | 116 ++++++++++---- tests/samples.rs | 24 ++- 7 files changed, 393 insertions(+), 105 deletions(-) diff --git a/docs/design.md b/docs/design.md index d7d5864..653f34d 100644 --- a/docs/design.md +++ b/docs/design.md @@ -93,6 +93,15 @@ Several protected stages are themselves little bytecode programs. The core includes a small VM (`bytecode.rs`) that generates and interprets those programs rather than hardcoding each variant's constants. +The PE32+ configuration block has two observed anchor-relative alignments. +Senbei selects between them by validating the stage1 `(RVA, length)` +descriptor against the image, rather than relying on a version-like word whose +value is not stable across build families. Stage3 seed advancement also varies: +the usual four-round result is tried first, then bounded alternatives are +replayed from the untouched ciphertext and accepted only when decompression +writes the exact target size and the recovered stage has its expected function +tail structure. + ## Integrity check Every produced image passes through `integrity::check` — a static, execution- @@ -125,7 +134,10 @@ payload. The capture context is propagated into section worker threads; panics outside an active unpack continue through the previously installed panic hook. Expected validation failures use structured variants carrying the failed stage, block index, table kind, or invalid range instead of collapsing unrelated causes -into a generic corruption error. +into a generic corruption error. Huffman/LZ failures distinguish invalid code +lengths, tree traversal, pending-length overflow, invalid back-references, +output overflow, and size mismatch. When both DLL parsing and the EXE-layout +fallback fail, the returned error retains both pipeline errors. Size requests are bounds-checked against a 1 GiB `MAX_IMAGE_SIZE` before allocation so a crafted header cannot abort the process with a huge allocation. In folder mode each file is isolated: one file's failure is logged diff --git a/samples/README.md b/samples/README.md index f6f0de3..2a5761d 100644 --- a/samples/README.md +++ b/samples/README.md @@ -8,7 +8,8 @@ committed. ## What to put here -Place protected inputs directly in this folder: +Place protected inputs in this folder, either directly or grouped in +subdirectories: - `*.exe` — Crackproof-protected executables (PE32 or PE32+) - `*.dll` — Crackproof-protected DLLs (native or managed) @@ -19,6 +20,9 @@ keeping the exact `._` suffix on the full file name. The test splices it the same way the CLI does; without it the loader stub alone is meaningless and the splice / export-overlay / TLS-restore code is never exercised. +The corpus scan is recursive and skips directories named `unpack`, matching the +CLI's output-directory rule. + Optionally, place a **golden** next to each input — the known-good unpacked output, named `.golden.`: diff --git a/src/unpacker/dll.rs b/src/unpacker/dll.rs index 3cc69c8..794e5e0 100644 --- a/src/unpacker/dll.rs +++ b/src/unpacker/dll.rs @@ -88,19 +88,17 @@ fn decrypt_data4( OpsLut::new(ops).map_region(d, addr as usize, size as usize); } - if size != decompressed_size { - // decompress reports corruption (after partial writes) via its bool; - // surface it instead of shipping a garbage block. - if !decompress( + if size != decompressed_size + && let Err(reason) = primitives::decompress_detailed( d, addr as u32, compressed_addr as u32, decomp_params[1] as u32, size as u32, decompressed_size as u32, - ) { - return Err(UnpackError::StageDecompressionFailed(stage)); - } + ) + { + return Err(UnpackError::StageDecompressionFailed { stage, reason }); } Ok(()) } @@ -469,6 +467,16 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> println!(" checksum1 = 0x{:08X}", checksum1 as u32); println!(" decrypted_addr1 = 0x{:08X}", decrypted_addr1 as u32); } + let primary_end = decrypted_addr1.checked_add(3856); + if decrypted_addr1 < keys[3] + || primary_end.is_none_or(|end| end < 0 || end as usize > out.len()) + { + return Err(UnpackError::InvalidDllPrimaryDescriptor { + address: decrypted_addr1 as u32, + minimum: keys[3] as u32, + image_len: out.len(), + }); + } let import_offset = get_i32(&out, decrypted_addr1 + 3444); let decrypted_addr2_size = get_i32(&out, decrypted_addr1 + 3632); decrypt_data3( diff --git a/src/unpacker/exe.rs b/src/unpacker/exe.rs index a76cf1e..2dfd5ad 100644 --- a/src/unpacker/exe.rs +++ b/src/unpacker/exe.rs @@ -35,6 +35,42 @@ impl std::fmt::Display for DecompressionStage { } } +#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)] +#[non_exhaustive] +pub enum DecompressionFailure { + #[error("compressed source size {size} exceeds limit {max}")] + SourceTooLarge { size: u32, max: u64 }, + #[error("Huffman code length {bits} is invalid")] + InvalidCodeLength { bits: u8 }, + #[error("Huffman tree traversal exceeded 64 levels")] + HuffmanTraversalLimit, + #[error("pending length accumulator overflowed at {pending}")] + PendingLengthOverflow { pending: u32 }, + #[error("output step {step} at byte {written} exceeds expected size {expected}")] + OutputOverflow { + written: u32, + step: u32, + expected: u32, + }, + #[error("run-fill width {width} reads before output offset 0x{destination:08X}")] + RunFillBeforeOutput { width: u32, destination: u32 }, + #[error("run-fill width {width} is unsupported")] + InvalidRunFillWidth { width: u32 }, + #[error("back-reference distance {distance} exceeds {written} written bytes")] + InvalidBackReference { distance: u32, written: u32 }, + #[error("Huffman symbol consumed no input and produced no output")] + NoProgress, + #[error( + "output size mismatch (wrote {written}/{expected} bytes after consuming {consumed}/{source_size})" + )] + OutputSizeMismatch { + written: u32, + expected: u32, + consumed: u32, + source_size: u32, + }, +} + #[derive(Debug, Clone, Copy, PartialEq, Eq)] pub enum BytecodeStage { ExeStage4, @@ -121,6 +157,9 @@ pub enum UnpackError { #[error("anchor field not found — corrupt data or wrong offset")] AnchorNotFound, + #[error("stage1 descriptor not found near anchor 0x{anchor:08X}")] + Stage1DescriptorNotFound { anchor: u32 }, + #[error("stage2 field not found — corrupt data or wrong offset")] Stage2NotFound, @@ -145,6 +184,15 @@ pub enum UnpackError { #[error("DLL pipeline requires PE32+ optional-header magic, got 0x{found:04X}")] UnsupportedDllPeMagic { found: u16 }, + #[error( + "DLL primary descriptor address 0x{address:08X} is below layout base 0x{minimum:08X} or outside {image_len}-byte image" + )] + InvalidDllPrimaryDescriptor { + address: u32, + minimum: u32, + image_len: usize, + }, + #[error("invalid SizeOfImage {size}; expected 1..={max}")] InvalidImageSize { size: i64, max: u64 }, @@ -182,8 +230,11 @@ pub enum UnpackError { #[error("PE32 file LFSR not found in eighthStage")] Pe32FileLfsrNotFound, - #[error("{0} decompression failed — corrupt data or wrong offset")] - StageDecompressionFailed(DecompressionStage), + #[error("{stage} decompression failed: {reason}")] + StageDecompressionFailed { + stage: DecompressionStage, + reason: DecompressionFailure, + }, #[error("{pipeline} section block {block} decompression failed")] SectionDecompressionFailed { @@ -197,6 +248,12 @@ pub enum UnpackError { #[error("Huffman table is outside the image at offset {offset}")] InvalidHuffmanTable { offset: u32 }, + #[error("DLL pipeline failed: {dll}; EXE fallback failed: {exe}")] + PipelineFallbackFailed { + dll: Box, + exe: Box, + }, + #[error( "PE32 second-stage range is invalid (offset {offset}, size {size}, image length {image_len})" )] @@ -391,8 +448,13 @@ impl<'a> Unpacker<'a> { } // Strategy (a): delegate to primitives::decrypt_and_decompress_data - fn decrypt_and_decompress_data(&mut self, pos: u32, key: u32, custom: Option<&[Op]>) -> bool { - primitives::decrypt_and_decompress_data( + fn decrypt_and_decompress_data( + &mut self, + pos: u32, + key: u32, + custom: Option<&[Op]>, + ) -> Result<(), DecompressionFailure> { + primitives::decrypt_and_decompress_data_detailed( &mut self.decompressed, pos, key, @@ -620,26 +682,34 @@ impl<'a> Unpacker<'a> { } }; - // Detect config-block layout version. Newer Crackproof builds (observed - // across several EXE families) shift every anchor-relative field from - // offset 40 onward by +8 bytes. The config-version stamp sits at - // anchor+104 in the old layout and anchor+112 in the new one. Across - // the whole corpus the stamp's top nibble is always 0x4 (top byte 0x40 - // or 0x44), whereas the +8 layout's anchor+104 holds an inserted small - // count (top nibble 0), so the stamp position is a reliable layout - // discriminator. - let stamp_at = |off: u32| -> bool { - (anchor + off + 4) as usize <= u.decompressed.len() - && (get_u32(&u.decompressed, anchor + off) >> 28) == 0x4 + // Layouts shift the anchor-relative fields by either zero or eight + // bytes. The nearby version-like word is not stable across all build + // families, so validate the stage1 (RVA, length) descriptor itself. + let descriptor_is_valid = |extra: u32| -> bool { + let pos = anchor.wrapping_add(120 + extra); + let Some(end) = (pos as usize).checked_add(8) else { + return false; + }; + if end > u.decompressed.len() { + return false; + } + let base = get_u32(&u.decompressed, pos); + let length = get_u32(&u.decompressed, pos.wrapping_add(4)); + base >= u.info[3] + && length >= 16 + && (base as usize) + .checked_add(length as usize) + .is_some_and(|stage_end| stage_end <= u.decompressed.len()) }; - let magic_off: u32 = if stamp_at(104) { - 104 - } else if stamp_at(112) { - 112 - } else { - 104 - }; - let anchor_extra: u32 = magic_off - 104; + let anchor_extra = [0u32, 8] + .into_iter() + .find(|&extra| descriptor_is_valid(extra)) + .ok_or(UnpackError::Stage1DescriptorNotFound { anchor })?; + + if verbose { + println!(" anchor = 0x{anchor:08X}"); + println!(" anchor layout offset = +0x{anchor_extra:X}"); + } let p1 = get_u32(&u.decompressed, anchor.wrapping_add(8)); let p2 = get_u32(&u.decompressed, anchor.wrapping_add(4)); @@ -667,11 +737,22 @@ impl<'a> Unpacker<'a> { let v_at = anchor.wrapping_add(20); let v = get_u32(&u.decompressed, v_at); let tgt = anchor.wrapping_add(120 + anchor_extra); + let stage1_descriptor = [ + get_u32(&u.decompressed, tgt), + get_u32(&u.decompressed, tgt.wrapping_add(4)), + ]; u.decrypt_data3(tgt, xor_acc ^ chk1 ^ v, 21); let stage1 = get_u32(&u.decompressed, tgt); + let stage1_len = get_u32(&u.decompressed, tgt.wrapping_add(4)); if verbose { println!("[3/9] Locating config layout..."); println!(" stage1 = 0x{:08X}", stage1); + println!(" stage1_len = 0x{stage1_len:08X}"); + println!( + " stage1 descriptor = [0x{:08X}, 0x{:08X}]", + stage1_descriptor[0], stage1_descriptor[1] + ); + println!(" stage1 key = xor 0x{xor_acc:08X} ^ chk 0x{chk1:08X} ^ val 0x{v:08X}"); } // Field offsets inside stage1 vary between Crackproof versions. Locate @@ -680,7 +761,6 @@ impl<'a> Unpacker<'a> { // derive every other field as fixed offsets from there. Observed // stage2_off: 3632 (older EXE builds), 3616 (another old-layout build), // 3624 (managed-assembly builds). - let stage1_len = get_u32(&u.decompressed, tgt.wrapping_add(4)); let info3 = u.info[3]; let info5 = u.info[5]; // Use the full info[3]..info[3]+info[5] range: stage entries may live in @@ -752,6 +832,9 @@ impl<'a> Unpacker<'a> { if verbose { println!("[4/9] Decrypting stage2..."); println!(" stage2 = 0x{:08X}", stage2); + println!(" stage2_off = 0x{stage2_off:04X}"); + println!(" checksum table = stage1+0x{chk_src_start:04X}"); + println!(" stage2 key = 0x{key2:08X}"); } // The stage2 head/walk2 tables shift between Crackproof versions. The 4-entry @@ -783,6 +866,20 @@ impl<'a> Unpacker<'a> { }; let head_off = table_start.wrapping_add(32); let walk2_off = head_off.wrapping_sub(88); + if verbose { + println!(" operation table = stage2+0x{table_start:04X}"); + println!(" head/walk = +0x{head_off:04X}/+0x{walk2_off:04X}"); + for index in 0..2u32 { + let entry = stage2.wrapping_add(head_off + index * 16); + println!( + " operation[{index}] = [0x{:08X}, 0x{:08X}, 0x{:08X}, 0x{:08X}]", + get_u32(&u.decompressed, entry), + get_u32(&u.decompressed, entry.wrapping_add(4)), + get_u32(&u.decompressed, entry.wrapping_add(8)), + get_u32(&u.decompressed, entry.wrapping_add(12)), + ); + } + } let mut head = stage2.wrapping_add(head_off); for _iter in 0..2 { @@ -825,30 +922,110 @@ impl<'a> Unpacker<'a> { } walk2 = walk2.wrapping_add(32); } + if verbose { + println!( + " key offsets = [0x{:08X}, 0x{:08X}, 0x{:08X}, 0x{:08X}]", + u.key_offsets[0], u.key_offsets[1], u.key_offsets[2], u.key_offsets[3] + ); + } let chk2 = u.calculate_checksum(anchor.wrapping_add(48 + anchor_extra)); let accum_at = stage1.wrapping_add(chk_src_start.wrapping_sub(16)); - let mut accum = get_u32(&u.decompressed, accum_at); - for l in 0..4u32 { + let accum_seed = get_u32(&u.decompressed, accum_at); + let mut running_accum = accum_seed; + let mut accum_candidates = vec![(0u32, accum_seed)]; + for l in 0..8u32 { let bound = (l + 1).wrapping_mul(25) << 2; let mut i: u32 = 1; while i <= bound { - accum = accum.wrapping_add(i); + running_accum = running_accum.wrapping_add(i); i = i.wrapping_add(1); } + accum_candidates.push((l + 1, running_accum)); } + let accum = accum_candidates[4].1; let at1 = stage1.wrapping_add(stage2_off.wrapping_add(88)); let stage3_field = get_u32(&u.decompressed, at1); + let stage3_slen = get_u32(&u.decompressed, at1.wrapping_add(4)); + let stage3_dest = get_u32(&u.decompressed, at1.wrapping_add(8)); let stage3_dlen = get_u32(&u.decompressed, at1.wrapping_add(12)); if verbose { println!("[5/9] Decrypting stages 3-5..."); println!(" stage3 = 0x{:08X}", stage3_field); + println!( + " stage3 descriptor = [0x{stage3_field:08X}, 0x{stage3_slen:08X}, 0x{stage3_dest:08X}, 0x{stage3_dlen:08X}]" + ); + println!(" stage3 key = xor 0x{xor_acc:08X} ^ chk 0x{chk2:08X} ^ val 0x{accum:08X}"); + println!(" stage3 accum seed = 0x{accum_seed:08X}"); } - if !u.decrypt_and_decompress_data(at1, xor_acc ^ chk2 ^ accum, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::ExeStage3, - )); + let stage3_key = xor_acc ^ chk2 ^ accum; + let stage3_source_end = (stage3_field as usize) + .checked_add(stage3_slen as usize) + .filter(|&end| end <= u.decompressed.len()) + .ok_or(UnpackError::BufferRangeOutOfBounds { + operation: BufferOperation::Read, + offset: stage3_field as usize, + size: stage3_slen as usize, + buffer_len: u.decompressed.len(), + })?; + let stage3_dest_end = (stage3_dest as usize) + .checked_add(stage3_dlen as usize) + .filter(|&end| end <= u.decompressed.len()) + .ok_or(UnpackError::BufferRangeOutOfBounds { + operation: BufferOperation::CopyDestination, + offset: stage3_dest as usize, + size: stage3_dlen as usize, + buffer_len: u.decompressed.len(), + })?; + const MAX_STAGE3_TRIAL_BYTES: usize = 16 * 1024 * 1024; + let trial_size = (stage3_slen as usize).checked_add(stage3_dlen as usize); + let stage3_backups = trial_size + .filter(|&size| size <= MAX_STAGE3_TRIAL_BYTES) + .map(|_| { + ( + u.decompressed[stage3_field as usize..stage3_source_end].to_vec(), + u.decompressed[stage3_dest as usize..stage3_dest_end].to_vec(), + ) + }); + let default_result = u.decrypt_and_decompress_data(at1, stage3_key, None); + if let Err(reason) = default_result { + let Some((source_backup, dest_backup)) = stage3_backups else { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage3, + reason, + }); + }; + let restore_stage3 = |data: &mut [u8]| { + data[stage3_field as usize..stage3_source_end].copy_from_slice(&source_backup); + data[stage3_dest as usize..stage3_dest_end].copy_from_slice(&dest_backup); + }; + let mut selected = None; + for (rounds, candidate_accum) in &accum_candidates { + if *rounds == 4 { + continue; + } + restore_stage3(&mut u.decompressed); + let candidate_key = xor_acc ^ chk2 ^ candidate_accum; + let result = u.decrypt_and_decompress_data(at1, candidate_key, None); + if result.is_ok() + && find_v4_offset(&u.decompressed, stage3_field, stage3_dlen).is_some() + { + selected = Some(*rounds); + break; + } + } + if let Some(rounds) = selected { + if verbose { + println!(" selected stage3 accumulator rounds = {rounds}"); + } + } else { + restore_stage3(&mut u.decompressed); + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage3, + reason, + }); + } } let at2 = stage1.wrapping_add(stage2_off.wrapping_add(104)); @@ -865,10 +1042,11 @@ impl<'a> Unpacker<'a> { let v4 = find_v4_offset(&u.decompressed, stage3_field, stage3_dlen) .unwrap_or_else(|| stage3_field.wrapping_add(4692)); let v4_val = get_u32(&u.decompressed, v4); - if !u.decrypt_and_decompress_data(at2, xor_acc ^ chk3 ^ v4_val, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::ExeStage3Secondary, - )); + if let Err(reason) = u.decrypt_and_decompress_data(at2, xor_acc ^ chk3 ^ v4_val, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage3Secondary, + reason, + }); } let chk4 = u.calculate_checksum(stage1.wrapping_add(chk_src_start.wrapping_add(16))); @@ -886,10 +1064,11 @@ impl<'a> Unpacker<'a> { if verbose { println!(" stage4 = 0x{:08X}", stage4_field); } - if !u.decrypt_and_decompress_data(at3, xor_acc ^ chk4 ^ v5_val, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::ExeStage4, - )); + if let Err(reason) = u.decrypt_and_decompress_data(at3, xor_acc ^ chk4 ^ v5_val, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage4, + reason, + }); } // Inside stage4, two locations vary by build: @@ -953,10 +1132,13 @@ impl<'a> Unpacker<'a> { if verbose { println!(" stage5 = 0x{:08X}", stage5_field); } - if !u.decrypt_and_decompress_data(at4, xor_acc ^ chk4 ^ chk5 ^ accum2, Some(&ops1)) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::ExeStage5, - )); + if let Err(reason) = + u.decrypt_and_decompress_data(at4, xor_acc ^ chk4 ^ chk5 ^ accum2, Some(&ops1)) + { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage5, + reason, + }); } // Inside stage5, the loader stores a table of (ptr, size) pairs at a @@ -2189,10 +2371,11 @@ impl<'a> Unpacker<'a> { let dp_base = ss.wrapping_add(dp_base_off); let forth_addr = dp_base.wrapping_add(0x40); let fk = header_checksum ^ second_stage_cs ^ forth_stage_key; - if !self.decrypt_and_decompress_data(forth_addr, fk, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::Pe32FourthStage, - )); + if let Err(reason) = self.decrypt_and_decompress_data(forth_addr, fk, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32FourthStage, + reason, + }); } // ---- FifthStage ---- @@ -2207,10 +2390,11 @@ impl<'a> Unpacker<'a> { .wrapping_sub(4), ); let fk5 = header_checksum ^ forth_cs ^ fifth_key; - if !self.decrypt_and_decompress_data(fifth_addr, fk5, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::Pe32FifthStage, - )); + if let Err(reason) = self.decrypt_and_decompress_data(fifth_addr, fk5, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32FifthStage, + reason, + }); } // ---- SevenStage ---- @@ -2232,10 +2416,11 @@ impl<'a> Unpacker<'a> { cs1_addr.wrapping_add(cs1_size).wrapping_sub(0x10), ); let fk7 = header_checksum ^ fifth_cs ^ seven_key; - if !self.decrypt_and_decompress_data(seven_addr, fk7, None) { - return Err(UnpackError::StageDecompressionFailed( - DecompressionStage::Pe32SeventhStage, - )); + if let Err(reason) = self.decrypt_and_decompress_data(seven_addr, fk7, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32SeventhStage, + reason, + }); } // ---- EighthStage ---- @@ -3034,10 +3219,13 @@ mod error_tests { #[test] fn structured_errors_include_stage_and_block_context() { - let stage = UnpackError::StageDecompressionFailed(DecompressionStage::ExeStage4); + let stage = UnpackError::StageDecompressionFailed { + stage: DecompressionStage::ExeStage4, + reason: DecompressionFailure::NoProgress, + }; assert_eq!( stage.to_string(), - "EXE stage4 decompression failed — corrupt data or wrong offset" + "EXE stage4 decompression failed: Huffman symbol consumed no input and produced no output" ); let block = UnpackError::SectionDecompressionFailed { diff --git a/src/unpacker/mod.rs b/src/unpacker/mod.rs index 2db55d3..35447a2 100644 --- a/src/unpacker/mod.rs +++ b/src/unpacker/mod.rs @@ -14,8 +14,8 @@ use std::sync::{Arc, Mutex}; pub use dll::{unpack_dll, unpack_dll_v}; pub use exe::{ - BufferOperation, BytecodeStage, DecompressionStage, DescriptorTable, SectionPipeline, - UnpackError, unpack as unpack_exe, unpack_v as unpack_exe_v, + BufferOperation, BytecodeStage, DecompressionFailure, DecompressionStage, DescriptorTable, + SectionPipeline, UnpackError, unpack as unpack_exe, unpack_v as unpack_exe_v, }; pub use integrity::{IntegrityReport, check as check_integrity}; @@ -332,10 +332,12 @@ pub fn unpack_auto_v(input: &[u8], verbose: bool) -> Result<(Kind, Vec), Unp Ok(out) => out, Err(dll_err) => match exe::unpack_v(input, verbose) { Ok(out) => out, - // Surface the DLL-pipeline error, not the EXE one: for a - // genuinely corrupt DLL the DLL error is the more relevant - // diagnostic, and the EXE fallback is best-effort. - Err(_) => return Err(dll_err), + Err(exe_err) => { + return Err(UnpackError::PipelineFallbackFailed { + dll: Box::new(dll_err), + exe: Box::new(exe_err), + }); + } }, } } diff --git a/src/unpacker/primitives.rs b/src/unpacker/primitives.rs index 1a8ab7f..886b159 100644 --- a/src/unpacker/primitives.rs +++ b/src/unpacker/primitives.rs @@ -610,24 +610,28 @@ pub(crate) fn calculate_checksum2(d: &[u8], clean: &[u8], pos: u32, start: u32) /// Reads `s_size` bytes from `src`, writes `d_size` bytes to `dest`. /// The Huffman table lives at `key_offset` within `d`. /// -/// Returns `true` when exactly `d_size` bytes were written (full success), -/// `false` on any corruption-triggered early exit. The PE32 eighth-stage key -/// brute force uses this status to discriminate the correct key. -pub(crate) fn decompress( +/// Returns a structured reason when the stream cannot produce exactly +/// `d_size` bytes. +pub(crate) fn decompress_detailed( d: &mut [u8], src: u32, mut dest: u32, key_offset: u32, s_size: u32, d_size: u32, -) -> bool { +) -> Result<(), super::DecompressionFailure> { + use super::DecompressionFailure; + // Bound the scratch allocation: a corrupt descriptor could request a // multi-gigabyte source size, and an allocation failure aborts the process // (uncatchable). Real payloads are far below this. if s_size as u64 > super::MAX_IMAGE_SIZE { - return false; + return Err(DecompressionFailure::SourceTooLarge { + size: s_size, + max: super::MAX_IMAGE_SIZE, + }); } - DECOMPRESS_SCRATCH.with_borrow_mut(|buf| { + DECOMPRESS_SCRATCH.with_borrow_mut(|buf| -> Result<(), DecompressionFailure> { let mut bit_pos: i32 = 0; let need = (s_size as usize).saturating_add(3); if buf.len() < need { @@ -661,7 +665,7 @@ pub(crate) fn decompress( // length byte comes from a corrupt table, and `1 << b2` would // panic (debug) or wrap (release) on it. if b2 >= 32 { - return false; + return Err(DecompressionFailure::InvalidCodeLength { bits: b2 }); } let mut mask: u32 = 1u32 << b2; b2 = b2.wrapping_add(1); @@ -673,7 +677,7 @@ pub(crate) fn decompress( while (t2 & 0x8000) == 0 { depth += 1; if depth > 64 { - return false; + return Err(DecompressionFailure::HuffmanTraversalLimit); } mask <<= 1; b2 = b2.wrapping_add(1); @@ -700,8 +704,7 @@ pub(crate) fn decompress( 0x100 => { step = 0; if pending >= 256 { - // corrupt input: stop decompressing (diagnostics go to caller/log, not stdout) - return false; + return Err(DecompressionFailure::PendingLengthOverflow { pending }); } pending = if pending == 0 { payload @@ -715,7 +718,11 @@ pub(crate) fn decompress( } step = pending.wrapping_mul(payload); if step.wrapping_add(written) > d_size { - return false; + return Err(DecompressionFailure::OutputOverflow { + written, + step, + expected: d_size, + }); } // Run-fill replicates the unit just written before `dest`. A // corrupt stream can emit one of these before anything has been @@ -724,7 +731,10 @@ pub(crate) fn decompress( match payload { 1 => { if dest < 1 { - return false; + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); } let v = d[(dest as usize) - 1]; for k in 0..pending { @@ -733,7 +743,10 @@ pub(crate) fn decompress( } 2 => { if dest < 2 { - return false; + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); } let v = get_u16(d, dest.wrapping_sub(2)); for k in 0..pending { @@ -742,7 +755,10 @@ pub(crate) fn decompress( } 4 => { if dest < 4 { - return false; + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); } let v = get_u32(d, dest.wrapping_sub(4)); for k in 0..pending { @@ -755,7 +771,9 @@ pub(crate) fn decompress( // yet still counted `step` bytes as written, leaving // stale-buffer holes that later stages treated as // plaintext. Report corruption instead. - return false; + return Err(DecompressionFailure::InvalidRunFillWidth { + width: payload, + }); } } pending = 0; @@ -765,7 +783,18 @@ pub(crate) fn decompress( if written.wrapping_add(payload) > d_size || pending.wrapping_add(payload) > written { - return false; + let distance = pending.wrapping_add(payload); + if distance > written { + return Err(DecompressionFailure::InvalidBackReference { + distance, + written, + }); + } + return Err(DecompressionFailure::OutputOverflow { + written, + step: payload, + expected: d_size, + }); } let back = pending.wrapping_add(payload); for k in 0..payload { @@ -783,18 +812,34 @@ pub(crate) fn decompress( // infinite loop (and `catch_unpack` traps panics, not hangs). // Every real symbol consumes ≥ 1 bit, so a valid stream can // never hit this. - return false; + return Err(DecompressionFailure::NoProgress); } } src_consumed += if bit_pos != 0 { 1 } else { 0 }; - // Mismatch in consumed/written sizes indicates corrupt input; the unpack - // result will then fail downstream checks. No stdout diagnostics here — - // the pure core stays I/O-free; surface errors via the caller/logfile. - let _ = src_consumed; - written == d_size + if written != d_size { + return Err(DecompressionFailure::OutputSizeMismatch { + written, + expected: d_size, + consumed: src_consumed.max(0) as u32, + source_size: s_size, + }); + } + Ok(()) }) } +/// Boolean compatibility wrapper used by candidate searches and block fan-out. +pub(crate) fn decompress( + d: &mut [u8], + src: u32, + dest: u32, + key_offset: u32, + s_size: u32, + d_size: u32, +) -> bool { + decompress_detailed(d, src, dest, key_offset, s_size, d_size).is_ok() +} + /// Walk the Huffman table at `key_offset` and snapshot its bytes for /// [`decompress_tbl`]. The table is a forest of 256 root entries (3 bytes /// each); non-terminal entries point at a child index pair. Returns `None` @@ -1874,14 +1919,14 @@ pub(crate) fn decrypt_data7(d: &mut [u8], pos: u32, mut key: u8) { /// /// Returns the decompression success status (always `true` when no /// decompression was needed). The PE32 eighth-stage key search relies on this. -pub(crate) fn decrypt_and_decompress_data( +pub(crate) fn decrypt_and_decompress_data_detailed( d: &mut [u8], pos: u32, key: u32, key1_offset: u32, key3_offset: u32, ops: Option<&[Op]>, -) -> bool { +) -> Result<(), super::DecompressionFailure> { let src = get_u32(d, pos); let src_len = get_u32(d, pos.wrapping_add(4)); aes_decrypt(d, src, src_len, key3_offset); @@ -1894,9 +1939,21 @@ pub(crate) fn decrypt_and_decompress_data( let dest = get_u32(d, pos.wrapping_add(8)); let dest_len = get_u32(d, pos.wrapping_add(12)); if src_len != dest_len { - return decompress(d, src, dest, key1_offset, src_len, dest_len); + return decompress_detailed(d, src, dest, key1_offset, src_len, dest_len); } - true + Ok(()) +} + +/// Boolean compatibility wrapper used by key searches that trial candidates. +pub(crate) fn decrypt_and_decompress_data( + d: &mut [u8], + pos: u32, + key: u32, + key1_offset: u32, + key3_offset: u32, + ops: Option<&[Op]>, +) -> bool { + decrypt_and_decompress_data_detailed(d, pos, key, key1_offset, key3_offset, ops).is_ok() } // --------------------------------------------------------------------------- @@ -2202,7 +2259,10 @@ mod tests { d[0..2].copy_from_slice(&sym.to_le_bytes()); d[2] = 8; // All-zero source -> symbol index 0 -> the invalid run-fill. - assert!(!decompress(&mut d, 0x40, 0x80, 0, 4, 3)); + assert_eq!( + decompress_detailed(&mut d, 0x40, 0x80, 0, 4, 3), + Err(super::super::DecompressionFailure::InvalidRunFillWidth { width: 3 }) + ); } /// Control for the above: a width-1 run-fill is legal and succeeds. diff --git a/tests/samples.rs b/tests/samples.rs index 7adf8fe..782c413 100644 --- a/tests/samples.rs +++ b/tests/samples.rs @@ -27,6 +27,7 @@ mod common; use common::samples_dir; use std::path::Path; +use walkdir::WalkDir; /// An input is a `.exe`/`.dll`/`.dat` whose name doesn't carry the `.golden.` /// marker — those are goldens, not inputs. External companions (`._`) @@ -85,10 +86,19 @@ fn samples_unpack_against_goldens() { return; } - let mut inputs: Vec<_> = std::fs::read_dir(&dir) - .unwrap_or_else(|e| panic!("read {}: {e}", dir.display())) - .filter_map(|e| e.ok().map(|e| e.path())) - .filter(|p| p.is_file() && is_input(p)) + let mut inputs: Vec<_> = WalkDir::new(&dir) + .follow_links(false) + .into_iter() + .filter_entry(|entry| { + entry.depth() == 0 + || !entry + .file_name() + .to_str() + .is_some_and(|name| name.eq_ignore_ascii_case("unpack")) + }) + .map(|entry| entry.unwrap_or_else(|e| panic!("walk {}: {e}", dir.display()))) + .filter(|entry| entry.file_type().is_file() && is_input(entry.path())) + .map(|entry| entry.into_path()) .collect(); inputs.sort(); @@ -107,7 +117,11 @@ fn samples_unpack_against_goldens() { let mut failures: Vec = Vec::new(); for input in &inputs { - let name = input.file_name().unwrap().to_string_lossy().to_string(); + let name = input + .strip_prefix(&dir) + .unwrap_or(input) + .to_string_lossy() + .to_string(); let bytes = match std::fs::read(input) { Ok(b) => b, Err(e) => { From e9ead4dc5f2ca29c72eea047263037c16c331c07 Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Tue, 11 Aug 2026 14:19:56 +0800 Subject: [PATCH 3/8] refactor: init --- .gitattributes | 3 - .github/ISSUE_TEMPLATE/bug_report.yml | 48 - .github/ISSUE_TEMPLATE/config.yml | 1 - .github/ISSUE_TEMPLATE/feature_request.yml | 25 - .github/workflows/ci.yml | 99 - .github/workflows/release.yml | 79 - .gitignore | 7 +- AGENTS.md | 79 - CLAUDE.md | 1 - Cargo.lock | 31 +- Cargo.toml | 45 +- README.md | 77 - docs/design.md | 154 -- docs/development.md | 116 - docs/usage.md | 126 - rust-toolchain.toml | 3 - samples/README.md | 88 - senbei-cli/Cargo.toml | 15 + {src => senbei-cli/src}/main.rs | 29 +- senbei-crypto/Cargo.toml | 9 + .../src}/bytecode.rs | 0 {src/unpacker => senbei-crypto/src}/crc32.rs | 0 senbei-crypto/src/lib.rs | 77 + senbei-crypto/src/primitives.rs | 1098 ++++++++ {src/unpacker => senbei-crypto/src}/tables.rs | 10 +- senbei-io/Cargo.toml | 23 + {src => senbei-io/src}/job.rs | 61 +- {src => senbei-io/src}/lib.rs | 4 +- {src => senbei-io/src}/logfile.rs | 9 +- {src => senbei-io/src}/pause.rs | 0 {src => senbei-io/src}/scan.rs | 10 +- {src => senbei-io/src}/ui.rs | 2 +- senbei-metadata/Cargo.toml | 6 + senbei-metadata/src/lib.rs | 5 + {src => senbei-metadata/src}/metadata.rs | 0 senbei-pe/Cargo.toml | 10 + senbei-pe/src/engine/dll/mod.rs | 3 + .../src/engine/dll/pipeline.rs | 18 +- senbei-pe/src/engine/error.rs | 243 ++ senbei-pe/src/engine/exe/mod.rs | 3 + .../src/engine/exe/pipeline.rs | 1368 +--------- senbei-pe/src/engine/exe/pipeline/pe32.rs | 1063 ++++++++ .../src/engine}/integrity.rs | 0 senbei-pe/src/engine/layout.rs | 14 + senbei-pe/src/engine/layout/dd8.rs | 321 +++ senbei-pe/src/engine/layout/discovery.rs | 507 ++++ senbei-pe/src/engine/layout/image.rs | 481 ++++ {src/unpacker => senbei-pe/src/engine}/mod.rs | 44 +- .../src/engine}/parallel.rs | 2 +- senbei-pe/src/lib.rs | 5 + src/unpacker/primitives.rs | 2394 ----------------- tests/common/mod.rs | 10 - tests/job.rs | 31 - tests/logfile.rs | 47 - tests/run_log.rs | 80 - tests/samples.rs | 229 -- web/Cargo.lock | 441 --- web/Cargo.toml | 19 - web/LICENSE | 662 ----- web/README.md | 75 - web/app.js | 392 --- web/index.html | 78 - web/src/lib.rs | 180 -- web/style.css | 369 --- web/worker.js | 41 - 65 files changed, 4028 insertions(+), 7442 deletions(-) delete mode 100644 .github/ISSUE_TEMPLATE/bug_report.yml delete mode 100644 .github/ISSUE_TEMPLATE/config.yml delete mode 100644 .github/ISSUE_TEMPLATE/feature_request.yml delete mode 100644 .github/workflows/ci.yml delete mode 100644 .github/workflows/release.yml delete mode 100644 AGENTS.md delete mode 120000 CLAUDE.md delete mode 100644 README.md delete mode 100644 docs/design.md delete mode 100644 docs/development.md delete mode 100644 docs/usage.md delete mode 100644 rust-toolchain.toml delete mode 100644 samples/README.md create mode 100644 senbei-cli/Cargo.toml rename {src => senbei-cli/src}/main.rs (80%) create mode 100644 senbei-crypto/Cargo.toml rename {src/unpacker => senbei-crypto/src}/bytecode.rs (100%) rename {src/unpacker => senbei-crypto/src}/crc32.rs (100%) create mode 100644 senbei-crypto/src/lib.rs create mode 100644 senbei-crypto/src/primitives.rs rename {src/unpacker => senbei-crypto/src}/tables.rs (92%) create mode 100644 senbei-io/Cargo.toml rename {src => senbei-io/src}/job.rs (94%) rename {src => senbei-io/src}/lib.rs (59%) rename {src => senbei-io/src}/logfile.rs (92%) rename {src => senbei-io/src}/pause.rs (100%) rename {src => senbei-io/src}/scan.rs (98%) rename {src => senbei-io/src}/ui.rs (97%) create mode 100644 senbei-metadata/Cargo.toml create mode 100644 senbei-metadata/src/lib.rs rename {src => senbei-metadata/src}/metadata.rs (100%) create mode 100644 senbei-pe/Cargo.toml create mode 100644 senbei-pe/src/engine/dll/mod.rs rename src/unpacker/dll.rs => senbei-pe/src/engine/dll/pipeline.rs (98%) create mode 100644 senbei-pe/src/engine/error.rs create mode 100644 senbei-pe/src/engine/exe/mod.rs rename src/unpacker/exe.rs => senbei-pe/src/engine/exe/pipeline.rs (59%) create mode 100644 senbei-pe/src/engine/exe/pipeline/pe32.rs rename {src/unpacker => senbei-pe/src/engine}/integrity.rs (100%) create mode 100644 senbei-pe/src/engine/layout.rs create mode 100644 senbei-pe/src/engine/layout/dd8.rs create mode 100644 senbei-pe/src/engine/layout/discovery.rs create mode 100644 senbei-pe/src/engine/layout/image.rs rename {src/unpacker => senbei-pe/src/engine}/mod.rs (90%) rename {src/unpacker => senbei-pe/src/engine}/parallel.rs (99%) create mode 100644 senbei-pe/src/lib.rs delete mode 100644 src/unpacker/primitives.rs delete mode 100644 tests/common/mod.rs delete mode 100644 tests/job.rs delete mode 100644 tests/logfile.rs delete mode 100644 tests/run_log.rs delete mode 100644 tests/samples.rs delete mode 100644 web/Cargo.lock delete mode 100644 web/Cargo.toml delete mode 100644 web/LICENSE delete mode 100644 web/README.md delete mode 100644 web/app.js delete mode 100644 web/index.html delete mode 100644 web/src/lib.rs delete mode 100644 web/style.css delete mode 100644 web/worker.js diff --git a/.gitattributes b/.gitattributes index 10d65d1..1ad8843 100644 --- a/.gitattributes +++ b/.gitattributes @@ -47,6 +47,3 @@ LICENSE text eol=lf *.gz binary *.xz binary *.dat binary - -# wasm-pack output is generated. -web/pkg/** linguist-generated=true diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml deleted file mode 100644 index 2bddcf0..0000000 --- a/.github/ISSUE_TEMPLATE/bug_report.yml +++ /dev/null @@ -1,48 +0,0 @@ -name: Bug report -description: Report a file that fails to unpack, unpacks incorrectly, or crashes senbei -title: "[Bug]: " -labels: [bug] -body: - - type: markdown - attributes: - value: | - Thanks for the report. Attaching the protected input file helps us - better diagnose the issue. Please only attach files you are - authorized to share. - - type: textarea - id: symptom - attributes: - label: What happened? - description: The command you ran and the output/error you got. Use -v/--verbose and paste the stage log if unpacking failed midway. - placeholder: "senbei app.exe -> error: ..." - validations: - required: true - - type: textarea - id: expected - attributes: - label: What did you expect? - placeholder: "A runnable unpacked image at unpack\\app.unpack.exe" - validations: - required: true - - type: input - id: version - attributes: - label: Senbei version - description: Output of `senbei -V` - validations: - required: true - - type: textarea - id: fileinfo - attributes: - label: About the input file - description: | - Without attaching it: is it an EXE or DLL? 32-bit or 64-bit? Managed - (.NET) or native? Roughly how large? Does it have a `._` companion - file next to it? - - type: checkboxes - id: legal - attributes: - label: Confirmation - options: - - label: I am authorized to analyze and share the attached file(s), and I understand attachments are removed after investigation. - required: true diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml deleted file mode 100644 index 3ba13e0..0000000 --- a/.github/ISSUE_TEMPLATE/config.yml +++ /dev/null @@ -1 +0,0 @@ -blank_issues_enabled: false diff --git a/.github/ISSUE_TEMPLATE/feature_request.yml b/.github/ISSUE_TEMPLATE/feature_request.yml deleted file mode 100644 index 613aacb..0000000 --- a/.github/ISSUE_TEMPLATE/feature_request.yml +++ /dev/null @@ -1,25 +0,0 @@ -name: Feature request -description: Suggest an improvement or support for a new layout -title: "[Feature]: " -labels: [enhancement] -body: - - type: markdown - attributes: - value: | - For a new Crackproof layout, attaching a sample protected file helps - us better understand the request. Please only attach files you are - authorized to share. - - type: textarea - id: idea - attributes: - label: What would you like? - description: The problem it solves for you, and any layout details you can share. - validations: - required: true - - type: checkboxes - id: legal - attributes: - label: Confirmation - options: - - label: This request is for lawful research/interoperability use, and I am authorized to share any file I attach. - required: true diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml deleted file mode 100644 index 65d0959..0000000 --- a/.github/workflows/ci.yml +++ /dev/null @@ -1,99 +0,0 @@ -name: CI - -on: - push: - branches: [main] - pull_request: - -env: - CARGO_TERM_COLOR: always - -jobs: - fmt: - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v4 - - run: cargo fmt --all -- --check - - clippy: - runs-on: windows-latest - steps: - - uses: actions/checkout@v4 - - run: cargo clippy --all-targets -- -D warnings - - test: - # The test suite exercises Windows path semantics, so it runs on Windows. - # The golden corpus (samples/) is user-managed and absent on CI; the - # samples test is a no-op pass there by design. - runs-on: windows-latest - steps: - - uses: actions/checkout@v4 - - run: cargo test --release - - check-portable: - # Build-only portability gate: non-Windows host and the wasm target the - # web build uses. Clippy runs here too, not just on Windows: the platform - # `cfg` branches (the POSIX `localtime_r` path, the non-Windows stubs) are - # invisible to the Windows clippy job, so without this they are never - # linted at all. - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v4 - - run: cargo clippy --all-targets -- -D warnings - - run: cargo check --target wasm32-unknown-unknown - - cli: - strategy: - matrix: - include: - - os: windows-latest - target: x86_64-pc-windows-msvc - bin: senbei.exe - - os: ubuntu-latest - target: x86_64-unknown-linux-gnu - bin: senbei - runs-on: ${{ matrix.os }} - steps: - - uses: actions/checkout@v4 - - run: cargo build --release --locked - - name: Package senbei-- - shell: bash - run: | - set -euo pipefail - version=$(grep -m1 '^version' Cargo.toml | cut -d'"' -f2) - name="senbei-$version-${{ matrix.target }}" - mkdir -p "stage/$name" - cp "target/release/${{ matrix.bin }}" "stage/$name/" - cp LICENSE "stage/$name/" - awk '/^## Legal notice and intended use/{f=1} f && /^## / && !/Legal notice/{exit} f' \ - README.md > "stage/$name/LEGAL-NOTICE.md" - echo "package=$name" >> "$GITHUB_ENV" - # upload-artifact always wraps the upload in its own zip, so the staged - # directory is uploaded loose — pre-compressing here would nest archives. - # The download is already .zip with the folder inside. - - uses: actions/upload-artifact@v4 - with: - name: ${{ env.package }} - path: stage/ - retention-days: 14 - - web: - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v4 - - run: cargo install wasm-pack --locked - # `-- --locked` forwards to cargo: web/Cargo.lock is committed on - # purpose, so the wasm build must be pinned by it rather than silently - # re-resolving (which is how it drifted out of sync with the manifest). - - run: wasm-pack build --target web --release -- --locked - working-directory: web - - uses: actions/upload-artifact@v4 - with: - name: senbei-web - path: | - web/index.html - web/app.js - web/worker.js - web/style.css - web/pkg/ - retention-days: 14 diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml deleted file mode 100644 index ce1c71c..0000000 --- a/.github/workflows/release.yml +++ /dev/null @@ -1,79 +0,0 @@ -name: Release - -# Publishing a GitHub release: -# - cli-assets builds the Windows and Linux CLI binaries and attaches -# senbei--.zip (binary + LICENSE + the legal -# notice extracted from README.md) to the release. -# - deploy-web rebuilds the browser app and pushes it to Cloudflare Pages -# (https://senbei.pages.dev). Requires two repository secrets: -# CLOUDFLARE_API_TOKEN (Pages:Edit) and CLOUDFLARE_ACCOUNT_ID. The Pages -# project is `senbei`; direct-upload projects default to `main` as the -# production branch, which `--branch=main` targets. - -on: - release: - types: [published] - -permissions: - contents: read - -jobs: - cli-assets: - permissions: - contents: write # attach assets to the release - strategy: - matrix: - include: - - os: windows-latest - target: x86_64-pc-windows-msvc - bin: senbei.exe - - os: ubuntu-latest - target: x86_64-unknown-linux-gnu - bin: senbei - runs-on: ${{ matrix.os }} - steps: - - uses: actions/checkout@v4 - - run: cargo build --release --locked - - name: Package senbei-- - shell: bash - run: | - set -euo pipefail - version=$(grep -m1 '^version' Cargo.toml | cut -d'"' -f2) - name="senbei-$version-${{ matrix.target }}" - mkdir -p "stage/$name" - cp "target/release/${{ matrix.bin }}" "stage/$name/" - cp LICENSE "stage/$name/" - awk '/^## Legal notice and intended use/{f=1} f && /^## / && !/Legal notice/{exit} f' \ - README.md > "stage/$name/LEGAL-NOTICE.md" - # Release assets are served as-is (unlike CI artifacts, which the - # upload action always re-zips), so the archive is created here. - # Zip on every OS, matching the CI artifacts; 7z is preinstalled on - # both windows-latest and ubuntu-latest. - (cd stage && 7z a "$name.zip" "$name" > /dev/null) - echo "package=$name" >> "$GITHUB_ENV" - - name: Attach to release - shell: bash - env: - GH_TOKEN: ${{ secrets.GITHUB_TOKEN }} - run: gh release upload "${{ github.event.release.tag_name }}" "stage/${{ env.package }}".* --clobber - - deploy-web: - runs-on: ubuntu-latest - steps: - - uses: actions/checkout@v4 - - run: cargo install wasm-pack --locked - # `-- --locked` forwards to cargo: web/Cargo.lock is committed on - # purpose, so the wasm build must be pinned by it rather than silently - # re-resolving (which is how it drifted out of sync with the manifest). - - run: wasm-pack build --target web --release -- --locked - working-directory: web - - name: Stage static site - run: | - mkdir dist - cp web/index.html web/app.js web/worker.js web/style.css dist/ - cp -r web/pkg dist/pkg - - uses: cloudflare/wrangler-action@v3 - with: - apiToken: ${{ secrets.CLOUDFLARE_API_TOKEN }} - accountId: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }} - command: pages deploy dist --project-name=senbei --branch=main diff --git a/.gitignore b/.gitignore index 7285ac2..24a516a 100644 --- a/.gitignore +++ b/.gitignore @@ -118,7 +118,7 @@ debug/ target/ # Cargo.lock is committed on purpose: senbei is a binary, and CI builds with -# --locked (web/Cargo.lock likewise, for the wasm build). +# --locked. # These are backup files generated by rustfmt **/*.rs.bk @@ -151,8 +151,3 @@ target/ # descend into it to find the re-included README). See senbei/samples/README.md. /samples/* !/samples/README.md - -### senbei web build ### -/web/pkg/ -/web/target/ -/web/.playwright-cli \ No newline at end of file diff --git a/AGENTS.md b/AGENTS.md deleted file mode 100644 index 55d7781..0000000 --- a/AGENTS.md +++ /dev/null @@ -1,79 +0,0 @@ -# AGENTS.md - -Guidance for AI coding agents (and human contributors) working in this repo. - -## Project - -Senbei is a static unpacker for Crackproof-protected PE files: a pure, -panic-free, no-I/O unpacker core (`src/unpacker/`) plus a thin CLI shell -(`src/`), an il2cpp metadata de-obfuscator (`src/metadata.rs`), and a -WebAssembly browser frontend (`web/`). Read `docs/design.md` first. - -## Commands - -```cmd -cargo build --release :: CLI -cargo test --release :: full suite (golden corpus: samples/, git-ignored) -cargo clippy --all-targets -- -D warnings -cargo fmt --all -cd web && wasm-pack build --target web --release :: browser build -``` - -The `samples/` corpus is user-managed and absent on CI; without it the -samples test is a no-op pass. `SENBEI_REQUIRE_SAMPLES=1` makes an absent -corpus fail (use this on a private CI that *does* have the corpus). Do not -delete `samples/` with `rm -rf` — it may be a junction; use git -worktree-aware cleanup. - -## Hard rules - -- **The unpacker core stays pure**: no file I/O, no `unsafe`, no panics across - the public boundary, no platform-specific code. It must keep compiling to - `wasm32-unknown-unknown` (`cargo check --target wasm32-unknown-unknown`). -- **`catch_unwind` does not work on wasm** (the prebuilt std can't unwind; a - caught panic becomes a fatal `unreachable` trap). Native code may rely on - `catch_unpack`, but any routing decision must also work without a catchable - panic: spliced companion inputs route straight to the EXE pipeline, and the - web app isolates every unpack in a disposable Web Worker, retrying trapped - DLLs with `job::unpack_bytes_force_exe`. Never make correctness on wasm - depend on catching a panic. -- **Byte-identical output is the contract.** Any pipeline change must re-run - the full golden corpus; a byte mismatch on any golden is a regression. -- **Trial-and-validate, never trust a heuristic.** A silently wrong offset - produces a silently broken binary — worse than an error. Every layout - candidate must be validated (checksum / structural oracle) with fall-through - to the next candidate. -- **Determinism under parallelism.** Block fan-out must stay byte-identical - regardless of thread count (`SENBEI_THREADS=1` is the sequential reference). -- **Folder scanning: deny-list, never allow-list.** Targets are recognised by - content, not extension, and can carry arbitrary names — there is no closed - set of target extensions an allow-list could enumerate. Only known - bulk-asset formats are excluded. -- **No binaries in the repo** — not as fixtures, not in commits. The only - corpus is the local git-ignored `samples/`. (Issue attachments of - protected inputs are fine when the user is authorized to share them, but - never commit them.) - -## Public-repo hygiene (important) - -This is a public research repository. In code comments, docs, tests, and -commit messages: - -- **Never name specific games, publishers, or product codenames.** Refer to - build families generically ("older EXE-64 builds", "the marker-less - layout", "external-companion builds"). Keep offsets/numbers — drop names. -- **Never name specific protected filenames** from real distributions. Test - fixtures use generic names (`app.exe`, `managed.dll`, `daemon.exe`). - Exceptions (platform-standard technology names, allowed): `il2cpp`, - `Unity`, `global-metadata.dat`, the Crackproof magic `KONN`. -- **Never reference other tools, projects, implementations, or paths outside - this repo.** Describe behavior and layout directly; do not mention prior - art, porting, or where any algorithm came from. - -## Conventions - -- Comments explain *why* (layout rationale, observed variants, failure modes), - not *what*. -- Rust 2024 edition; clippy-clean at `-D warnings`; rustfmt default style. -- CLI behavior (flags, exit codes, output naming) is documented in - `docs/usage.md` — update the doc when changing behavior. diff --git a/CLAUDE.md b/CLAUDE.md deleted file mode 120000 index 47dc3e3..0000000 --- a/CLAUDE.md +++ /dev/null @@ -1 +0,0 @@ -AGENTS.md \ No newline at end of file diff --git a/Cargo.lock b/Cargo.lock index 8fb6378..cd7546b 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -208,19 +208,46 @@ dependencies = [ ] [[package]] -name = "senbei" +name = "senbei-cli" +version = "1.0.0" +dependencies = [ + "senbei-io", +] + +[[package]] +name = "senbei-crypto" +version = "1.0.0" +dependencies = [ + "thiserror", +] + +[[package]] +name = "senbei-io" version = "1.0.0" dependencies = [ "anyhow", "indicatif", "libc", "owo-colors", + "senbei-metadata", + "senbei-pe", "tempfile", - "thiserror", "walkdir", "windows", ] +[[package]] +name = "senbei-metadata" +version = "1.0.0" + +[[package]] +name = "senbei-pe" +version = "1.0.0" +dependencies = [ + "senbei-crypto", + "thiserror", +] + [[package]] name = "slab" version = "0.4.12" diff --git a/Cargo.toml b/Cargo.toml index eba1196..fcef7df 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -1,39 +1,36 @@ -[package] -name = "senbei" +[workspace] +members = [ + "senbei-cli", + "senbei-crypto", + "senbei-io", + "senbei-metadata", + "senbei-pe", +] +default-members = ["senbei-cli"] +resolver = "2" + +[workspace.package] version = "1.0.0" edition = "2024" -description = "Static unpacker for Crackproof-protected PE files" license = "AGPL-3.0-only" -keywords = ["unpacker", "reverse-engineering", "pe", "security-research"] -categories = ["command-line-utilities"] -[lib] -name = "senbei" -path = "src/lib.rs" - -[[bin]] -name = "senbei" -path = "src/main.rs" - -[dependencies] +[workspace.dependencies] anyhow = "1" +indicatif = "0.18" +libc = "0.2" +owo-colors = "4" +tempfile = "3" thiserror = "2" walkdir = "2" -indicatif = "0.18" -owo-colors = "4" - -[target.'cfg(windows)'.dependencies] windows = { version = "0.62", features = [ "Win32_Foundation", "Win32_System_Console", "Win32_System_SystemInformation", ] } - -[target.'cfg(all(not(windows), not(target_arch = "wasm32")))'.dependencies] -libc = "0.2" - -[dev-dependencies] -tempfile = "3" +senbei-crypto = { path = "senbei-crypto" } +senbei-io = { path = "senbei-io" } +senbei-metadata = { path = "senbei-metadata" } +senbei-pe = { path = "senbei-pe" } [profile.release] opt-level = 3 diff --git a/README.md b/README.md deleted file mode 100644 index 3c2215b..0000000 --- a/README.md +++ /dev/null @@ -1,77 +0,0 @@ -# Senbei - -A static unpacker for Crackproof-protected 64-bit and 32-bit PE files. Point it -at a file or a folder and it writes decrypted copies — no launch of the -protected program, no kernel driver, no code runs out of the protected binary. - -> _"Crackproof"? It's senbei (煎餅 — rice cracker). Cracks itself._ - -Senbei reads a protected `.exe` or `.dll`, replays the unpacking algorithm -entirely in memory, and writes the recovered image to a new file. The core is a -pure, panic-free library with no file I/O; the CLI wraps it with scanning, a -progress bar, and a run log. A browser version (WebAssembly, fully client-side) -lives in [`web/`](web/). - -## Legal notice and intended use - -**Read this before using Senbei.** - -- Senbei is a research and interoperability tool. It exists to enable lawful - reverse engineering, security research, preservation, and interoperability - with software you already legitimately possess. -- **Only process binaries you own or are explicitly authorized to analyze.** - Depending on your jurisdiction and license agreements, circumventing - technological protection measures may be restricted (for example under - DMCA §1201 in the United States, which contains exemptions for security - research and interoperability). It is your responsibility to ensure your use - is lawful. -- Senbei does not bypass any access control for you: it performs a purely - static transformation of a file already on your disk. It derives everything - it needs from the input file itself, contains no vendor code or secrets, and - distributes no keys, cracks, or copyrighted content. -- Senbei does not enable online play, license fraud, or cheating, and must not - be used to redistribute decrypted binaries. Do not upload outputs anywhere. -- The authors provide this software "as is", without warranty of any kind, and - accept no liability for misuse. See [LICENSE](LICENSE) (AGPL-3.0). -- "Crackproof" is a trademark of its respective owner; this project is not - affiliated with or endorsed by the protection vendor or any software - publisher. Names are used for identification only. - -## What it handles - -| Kind | Description | -| --- | --- | -| `Exe` | Crackproof-protected executable (PE32+ and PE32). | -| `NativeDll` | Protected native (unmanaged) DLL. | -| `ManagedDll` | Protected .NET assembly (has a CLR data directory). | -| `._` companion | Stub + external encrypted payload layout, spliced automatically. | -| `global-metadata.dat` | il2cpp metadata with obfuscated method tokens, de-obfuscated in place. | - -Detection is content-based (header key-table at offset 4096, magic `KONN`), -not extension-based. Anything unrecognized is left untouched. - -## Quick start - -```cmd -cargo build --release - -senbei protected.exe -:: -> unpack\protected.unpack.exe - -senbei "C:\Games\MyGame" -:: -> C:\Games\MyGame\unpack\... (recursive, skips non-targets) -``` - -Every output is sanity-checked statically; structurally broken results are -flagged as suspect rather than silently trusted. - -## Documentation - -- [Usage reference](docs/usage.md) — CLI flags, exit codes, integrity check -- [Design](docs/design.md) — architecture, routing, and error model -- [Development](docs/development.md) — building, testing, environment variables -- [Web version](web/README.md) — run Senbei in a browser - -## License - -[GNU Affero General Public License v3.0](LICENSE) (AGPL-3.0-only). diff --git a/docs/design.md b/docs/design.md deleted file mode 100644 index 653f34d..0000000 --- a/docs/design.md +++ /dev/null @@ -1,154 +0,0 @@ -# Design - -Senbei is a fully static unpacker: it replays the unpacking algorithm on the -file bytes in memory and writes the recovered PE image. No code from the -protected binary is ever executed, no process is launched or attached to, and -no driver or proxy DLL is involved. - -## Crate layout - -The crate is split into a pure core and a thin CLI shell: - -- **`src/unpacker/`** — the core. Pure functions over byte slices: no file - I/O, no environment access (beyond a few debugging overrides, see - [development.md](development.md)), panic-free at the public boundary (all - internal panics are trapped and converted to `UnpackError::InternalPanic`). This - is what the WebAssembly build embeds. -- **`src/` (top level)** — the CLI shell: argument parsing, recursive folder - scanning, per-run log file, progress bar, Explorer-friendly exit pause, and - the single-file/folder orchestration in `job.rs`. -- **`src/metadata.rs`** — il2cpp `global-metadata.dat` method-token - de-obfuscation (format version 31; other versions are left untouched). - -``` -src/ -├── main.rs argument parsing + dispatch -├── lib.rs module roots -├── job.rs single-file + folder orchestration, out-naming, -│ companion splice, stub overlay/TLS restore, -│ pipeline routing (incl. the wasm-safe byte API) -├── scan.rs recursive Crackproof + metadata discovery -├── metadata.rs il2cpp global-metadata.dat de-obfuscation -├── logfile.rs per-run timestamped log -├── ui.rs progress bar + status lines -├── pause.rs Explorer-friendly exit pause -└── unpacker/ pure, panic-free, no-I/O core - ├── mod.rs detection + unpack_auto dispatch - ├── exe.rs EXE pipeline (PE32+ and PE32) - ├── dll.rs native + managed DLL pipeline - ├── integrity.rs static post-unpack sanity check - ├── primitives.rs decrypt_data* steps, key/shift selection - ├── bytecode.rs bytecode VM - ├── parallel.rs deterministic block-parallel fan-out - ├── tables.rs constant tables - └── crc32.rs checksum -``` - -## Detection and routing - -Detection is content-based (`unpacker::detect`), never extension-based: the -key table is derived from the file header and checked against the format -magic, then the PE characteristics classify the input as EXE, native DLL, or -managed DLL. - -`unpack_auto` then dispatches: - -- `Exe` → the EXE pipeline (handles both PE32+ and PE32). -- `NativeDll` / `ManagedDll` → the DLL pipeline first; on failure, the EXE - pipeline as a fallback. Two DLL layouts exist in the wild: an older layout - the DLL pipeline parses, and a newer one that protects DLLs with the - EXE-style shell layout instead. The DLL-first order keeps old-layout outputs - byte-identical (the EXE pipeline also "succeeds" on old-layout DLLs but - produces different bytes); the fallback handles the new layout (including - the managed-DLL .NET metadata restore). - -One routing shortcut bypasses `unpack_auto`: inputs spliced from an external -companion (`job.rs`, both the CLI and the wasm byte API) go **straight to the -EXE pipeline**. The companion layout is definitionally the EXE-style shell, -so the DLL probe can never be right for it — and the probe's rejection of -EXE-shell DLLs relies on a caught panic, which is a fatal trap on targets -without unwinding (WebAssembly). Output bytes are identical to the -probe-then-fallback route. - -## External-companion inputs - -Some builds split a protected module into an on-disk loader stub plus an -encrypted `._` companion. When a `._` sibling matches the stub's header -region, `job.rs` splices the two before unpacking and afterwards overlays the -export table and TLS directory from the stub — pieces the encrypted companion -does not carry. All overlay steps are best-effort no-ops when their inputs -can't be mapped, so a malformed stub can never corrupt an otherwise-good -unpack. - -## Pipelines - -Both pipelines are **heuristic with trial-and-validate**: where a layout -leaves ambiguity (e.g. which block is the real file decryptor, or a page-XOR -shift), the pipeline tries candidates and validates the result structurally -(an entry-stub oracle, checksum stamps, cluster stamps) instead of trusting -the first match. A validation failure falls through to the next candidate -rather than producing silently wrong output. - -Several protected stages are themselves little bytecode programs. The core -includes a small VM (`bytecode.rs`) that generates and interprets those -programs rather than hardcoding each variant's constants. - -The PE32+ configuration block has two observed anchor-relative alignments. -Senbei selects between them by validating the stage1 `(RVA, length)` -descriptor against the image, rather than relying on a version-like word whose -value is not stable across build families. Stage3 seed advancement also varies: -the usual four-round result is tried first, then bounded alternatives are -replayed from the untouched ciphertext and accepted only when decompression -writes the exact target size and the recovered stage has its expected function -tail structure. - -## Integrity check - -Every produced image passes through `integrity::check` — a static, execution- -free sanity check that only flags defects impossible in a correctly unpacked -image (malformed headers, unmapped/non-executable/all-zero/all-int3 entry -point, a native DLL with no base-relocation directory, any import descriptor -whose DLL name is still ciphertext, a managed image whose COR20 header or BSJB -metadata did not survive). See [usage.md](usage.md#integrity-check). -A clean report is not a proof of correctness; a non-clean report is a reliable -"broken" signal. - -## Parallelism - -Section decrypt/decompress blocks write disjoint output spans and read only -immutable input plus snapshotted key tables, so `parallel.rs` fans them out -across worker threads with **byte-identical** output regardless of thread -count. There is no `unsafe`: the buffer is carved with safe `split_at_mut` -chains so the borrow checker proves spans never alias. Overlapping spans (only -possible on corrupt input) degrade to the sequential whole-buffer pass, -preserving the deterministic last-writer-wins behavior of the serial -pipeline. `SENBEI_THREADS=1` forces the sequential path; on targets without -threads (WebAssembly) the sequential path is used automatically. - -## Error model - -The public API never panics: every pipeline runs under a `catch_unwind` -wrapper (`catch_unpack`) that converts a trapped panic to -`UnpackError::InternalPanic`, including the Rust source location and panic -payload. The capture context is propagated into section worker threads; panics -outside an active unpack continue through the previously installed panic hook. -Expected validation failures use structured variants carrying the failed stage, -block index, table kind, or invalid range instead of collapsing unrelated causes -into a generic corruption error. Huffman/LZ failures distinguish invalid code -lengths, tree traversal, pending-length overflow, invalid back-references, -output overflow, and size mismatch. When both DLL parsing and the EXE-layout -fallback fail, the returned error retains both pipeline errors. -Size requests are bounds-checked against a 1 GiB `MAX_IMAGE_SIZE` before -allocation so a crafted header cannot abort the process with a huge -allocation. In folder mode each file is isolated: one file's failure is logged -and counted, never fatal to the run. - -**WebAssembly caveat:** the prebuilt wasm std cannot unwind, so a caught -panic becomes a fatal `unreachable` trap there. The DLL-routing probe relies -on this mechanism to reject EXE-shell-layout DLLs, so the web build routes -around it instead of through it: spliced companion inputs skip the probe -entirely (see "Detection and routing"), and the web app isolates every unpack -in a disposable Web Worker — a trapped DLL is retried once in a fresh worker -with the forced-EXE pipeline (`job::unpack_bytes_force_exe`), reproducing the -probe-then-fallback outcome without a catchable panic. A trap on any other -input is reported as a clean error rather than freezing the page. diff --git a/docs/development.md b/docs/development.md deleted file mode 100644 index 0dbaad8..0000000 --- a/docs/development.md +++ /dev/null @@ -1,116 +0,0 @@ -# Development - -## Building - -Requires a Rust toolchain (MSVC backend is the default on Windows; -`rustup-init.exe` from installs it). The pinned toolchain -and targets are in `rust-toolchain.toml`. - -```cmd -cargo build --release -``` - -Output: `target\release\senbei.exe`. The binary is self-contained — no driver, -no proxy DLL, no external assets. - -The library and CLI also build for Linux/macOS (`cfg`-gated platform code -only) and for `wasm32-unknown-unknown` (see the [web version](../web/README.md)). - -## Testing - -```cmd -cargo test --release -``` - -The suite covers CLI behavior, detection, the folder driver, the run log, and -byte-exact golden tests over `samples/` — a user-managed corpus (git-ignored, -see `samples/README.md`) of real Crackproof inputs plus `.golden.` -reference outputs. Every input goes through `job::unpack_bytes` — the same -routing the CLI uses, so an `._` companion in the corpus is spliced and -the stub export/TLS overlays run — and is gated on **two** checks: the static -integrity check (catches runtime-broken outputs even when a stale golden would -still byte-match) and, when a golden exists, a bit-for-bit comparison. il2cpp -`*.dat` inputs are routed through `metadata::deobfuscate` instead. An empty or -absent corpus is a no-op pass; set `SENBEI_REQUIRE_SAMPLES` to make it fail -instead (useful on a private CI that has the corpus — public CI never does, -since binaries are not committed). - -> **Note:** goldens encode expected *bytes*, not runtime behavior. A golden -> produced before a pipeline fix may byte-match while still being wrong — the -> integrity check is the second gate for exactly this reason. Re-verify -> goldens against real runs when touching the affected pipeline stages. -> -> **The corpus only protects what it contains.** Wire the test to the routing -> the CLI actually takes (it is), and keep a sample for every layout family — -> marker-based, marker-less, external-companion, PE32, PE32+, native, managed, -> metadata. An unrepresented family has no regression gate at all, which is -> how a "re-run the golden corpus" rule can pass while silently covering -> nothing. - -## Debugging levers (environment variables) - -- `DD8_SHIFT` — override the `decrypt_data8` page-XOR shift (`99` skips dd8 - entirely). -- `SEL_DIAG` — print the dd8 selector's scores: the per-shift `0xCC` counts and - the plaintext baseline they are compared against (PE32+), and the per-formula - counts, baseline and net gain (PE32). -- `SENBEI_THREADS` — cap the block-parallel fan-out (`1` forces the fully - sequential path). -- `SENBEI_SCAN_ALL` — same as `--scan-all` (probe every file in a folder). - -## Conventions - -- The `src/unpacker/` core is pure: no file I/O, no panics across the public - boundary, no `unsafe`. Keep it that way — it is what the WebAssembly build - embeds. -- Layout heuristics must **trial-and-validate**: never pick a candidate offset - on shape alone and trust it; validate by decryption/checksum and fall - through to the next candidate on failure. A silent wrong offset produces a - silently broken output, which is worse than an error. -- Output must remain byte-identical against the golden corpus for every - supported layout. When fixing one build family, re-run the full golden - corpus to prove no other family regressed. -- Folder scanning uses a size floor plus an extension **deny**-list, never an - allow-list: targets are recognised by content, not extension, and can carry - arbitrary names, so only known bulk-asset extensions are excluded. The - pre-filter exists because folder-scan cost is per-file I/O latency, not the - walk — probe fewer files, don't parallelize the probe loop. -- `cargo fmt` and `cargo clippy` must stay clean (CI enforces both). - -## Repository layout - -``` -senbei/ -├── Cargo.toml senbei lib + bin package -├── rust-toolchain.toml pinned toolchain + targets -├── src/ CLI shell + pure unpacker core (see docs/design.md) -├── tests/ CLI, detection, golden, and folder tests -├── samples/ local-only test corpus (git-ignored) -├── web/ WebAssembly browser build -├── docs/ usage, design, and development documentation -└── .github/ CI workflows and issue templates -``` - -## Web build - -See [web/README.md](../web/README.md). In short: - -```cmd -cd web -wasm-pack build --target web --release -``` - -then serve `web/` statically and open `index.html`. Everything runs -client-side; no file leaves the browser. - -## Contributing - -Issues and pull requests are welcome. A few ground rules: - -- **Never commit binaries** (protected or decrypted) to the repository — - the only corpus is the local git-ignored `samples/`. Attaching a protected - input file to an issue is welcome if it helps diagnose the problem; only - attach files you are authorized to share. -- Run `cargo test --release`, `cargo clippy`, and `cargo fmt` before - submitting. -- Keep the unpacker core free of I/O, `unsafe`, and platform-specific code. diff --git a/docs/usage.md b/docs/usage.md deleted file mode 100644 index 9bfcc3f..0000000 --- a/docs/usage.md +++ /dev/null @@ -1,126 +0,0 @@ -# Usage - -``` -senbei [--out DIR] [-v|--verbose] [-q|--quiet]... [--scan-all] - [--no-log] [--no-pause] [-V|--version] [-h|--help] -``` - -Real runs print `Senbei ` once at start. Use `-V` / `--version` to -print the version and exit. - -## Single file - -The decrypted image is written under `/unpack/` with `.unpack` inserted -before the extension. A `senbei-.log` is written in the same -directory. With `--out DIR`, both the output and the log go into `DIR` instead: - -```cmd -senbei app.exe -:: -> unpack\app.unpack.exe -:: -> unpack\senbei-YYYYMMDD-HHMMSS.log - -senbei app.exe --out C:\out -:: -> C:\out\app.unpack.exe -:: -> C:\out\senbei-YYYYMMDD-HHMMSS.log -``` - -Pointing senbei directly at an il2cpp `global-metadata.dat` rewrites its -obfuscated method tokens back to the contiguous per-module range il2cpp -expects; the output is `global-metadata.unpack.dat`, written only when tokens -actually changed. Only metadata format version 31 is rewritten; other versions -are reported and left untouched. - -## Folder mode - -Senbei walks the directory recursively, skips any subdirectory literally named -`unpack`, and unpacks every file it recognises as Crackproof-protected (by -content, not extension — renamed files and `.bak` backups are still found). -Results land under `/unpack/` (or `--out DIR`), mirroring the input -tree's relative paths. The run log is written **in that same out directory**: - -```cmd -senbei "C:\Games\MyGame" -:: -> C:\Games\MyGame\unpack\... -:: -> C:\Games\MyGame\unpack\senbei-YYYYMMDD-HHMMSS.log -``` - -Folder mode also picks up `global-metadata.dat` files and external-companion -`._` payloads: a module whose `._` sibling matches its header region is -spliced with the companion automatically (no flag needed) and unpacked as one -image, with the output named for the stub. - -Each file is processed in isolation: an error or panic on one file is caught, -counted, and logged, and the run continues. Folder mode finishes with a summary -line, then duration: - -``` -12 unpacked · 3 skipped · 0 errors · 1 suspect · 2 metadata -done in 1234 ms -``` - -## Integrity check - -A successful unpack is not always a runnable one: a layout heuristic can pick -the wrong offset and leave the entry-point stub or import strings encrypted, so -the pipeline reports success but the OS loader faults at runtime (typically -`0xC0000005`, STATUS_ACCESS_VIOLATION). To catch this, senbei runs a static -sanity check over every output it produces — inspecting the bytes alone, with -no reference image and no execution. - -It flags only defects that cannot occur in a correctly unpacked image: - -- malformed DOS/PE headers, bad optional-header magic, implausible section - count, zero `SizeOfImage`, or section raw-data ranges that run past EOF; -- an entry point that doesn't map into a section, isn't in an executable - section, or whose stub is all zeros or all `0xCC` int3 padding (the classic - left-encrypted symptom); -- a native (unmanaged) DLL with no base-relocation directory — it cannot - survive being mapped at a non-preferred base; -- **any** import descriptor whose DLL name doesn't resolve or isn't readable - ASCII (imports left encrypted) — the whole table is walked, not just the - first entry; -- for a managed assembly, a COR20 header whose `cb` isn't `0x48` or a - MetaData stream missing its `BSJB` signature (the CLR would reject the - image outright). - -The entry-point and import checks are skipped for managed assemblies, whose -native EP and import stub are legitimately not what the native loader expects. - -The check is deliberately conservative: a clean report is **not** a proof of -correctness, but a non-clean report is a reliable "this is broken" signal. A -suspect file is still written (the bytes are the best available) and flagged — -single-file mode prints a warning to stderr, folder mode prints a yellow `!` -line, adds a `SUSPECT` entry to the run log, and counts it in the summary's -`suspect` total (which is additive to `unpacked`). - -## Flags - -| Flag | Behavior | -| --- | --- | -| `--out DIR` | Write outputs (and the log, unless `--no-log`) under `DIR`. | -| `-v`, `--verbose` | Print detailed `[N/9]` per-stage unpack progress (and the destination path) for each file. In folder mode this replaces the progress bar. | -| `-q`, `--quiet` | Once: hide progress bar and per-file lines; keep banner, summary, and duration. Twice (`-q -q`): suppress all stdio (exit code only). | -| `--no-log` | Do not write `senbei-*.log`. Console output is unchanged by this flag alone. | -| `--scan-all` | Probe every file in a folder, including ones the scan pre-filter skips (under 4128 bytes, or a bulk-asset extension like `.ab`/`.xml`/`.acb`). Much slower on large game trees; finds the same targets in practice. | -| `--no-pause` | Skip the "Press Enter to exit" prompt (for scripted runs). | -| `-V`, `--version` | Print `Senbei ` and exit. | -| `-h`, `--help` | Show usage. | - -On Windows, when launched from Explorer (the process owns its console) senbei -pauses for Enter before exiting so the window doesn't vanish. `--no-pause` -disables this; it has no effect when stdout is piped or run from another -process. - -## Exit codes - -| Code | Meaning | -| --- | --- | -| `0` | Success (single file unpacked, or folder run with no errors). | -| `1` | At least one file failed, a scan probe was unreadable, or a single-file unpack errored. | -| `2` | Usage error: no path given, unknown option, missing `--out` value, or multiple input paths (help printed). | - -A folder run also fails with `1` when parts of the tree could not be scanned -(unreadable directory entries or files that failed the content probe) — those -are potential missed targets, not clean skips. An il2cpp metadata blob whose -format version senbei does not handle is *not* an error: it is reported, left -untouched, and counted as skipped. diff --git a/rust-toolchain.toml b/rust-toolchain.toml deleted file mode 100644 index f45a9a0..0000000 --- a/rust-toolchain.toml +++ /dev/null @@ -1,3 +0,0 @@ -[toolchain] -channel = "stable" -targets = ["x86_64-pc-windows-msvc", "wasm32-unknown-unknown"] diff --git a/samples/README.md b/samples/README.md deleted file mode 100644 index 2a5761d..0000000 --- a/samples/README.md +++ /dev/null @@ -1,88 +0,0 @@ -# senbei/samples - -Drop-in corpus for the `samples` integration test (`tests/samples.rs`). - -This folder is **git-ignored** (only this `README.md` is tracked), so it holds -whatever Crackproof binaries happen to be on your machine. Nothing here is -committed. - -## What to put here - -Place protected inputs in this folder, either directly or grouped in -subdirectories: - -- `*.exe` — Crackproof-protected executables (PE32 or PE32+) -- `*.dll` — Crackproof-protected DLLs (native or managed) -- `*.dat` — il2cpp `global-metadata.dat` blobs (method-token de-obfuscation) - -For an **external-companion** module, copy the `._` payload in as well, -keeping the exact `._` suffix on the full file name. The test splices it the -same way the CLI does; without it the loader stub alone is meaningless and the -splice / export-overlay / TLS-restore code is never exercised. - -The corpus scan is recursive and skips directories named `unpack`, matching the -CLI's output-directory rule. - -Optionally, place a **golden** next to each input — the known-good unpacked -output, named `.golden.`: - -``` -samples/ - app.exe <- input - app.golden.exe <- golden (optional) - managed.dll <- input - managed.golden.dll <- golden (optional) - stub.dll <- input (external-companion layout) - stub.dll._ <- its encrypted payload (NOT an input itself) - stub.golden.dll <- golden - global-metadata.dat <- input - global-metadata.golden.dat<- golden - mystery.exe <- input, no golden -``` - -The type (EXE vs native/managed DLL vs metadata) is auto-detected from the file -contents, not the extension, so you don't need to classify anything by hand. - -Since the corpus is the only regression gate on byte-identical output, keep it -broad: each build family, each layout (marker-based and marker-less), and at -least one external-companion pair. A family with no sample here is a family no -test protects. - -## How the test treats each input - -Run with: - -``` -cargo test --release --test samples -``` - -For every input file, the test runs the same routing the CLI uses -(`job::unpack_bytes`, so companions splice and the stub overlays run) — or -`metadata::deobfuscate` for an il2cpp blob — and then: - -| Situation | Result | -| ------------------------------------------- | ------------------------------- | -| Golden present, bytes **identical** | **pass** | -| Golden present, bytes **differ** | **fail** (test fails) | -| **No golden** found | **warning** (needs manual check)| -| Unpack errored / file unreadable | **fail** | - -Warnings are printed but do not fail the test — they flag outputs you should -eyeball or promote to a golden once verified. Failures fail the test. An empty -or absent folder is a no-op pass. - -To see the per-file warning/pass/fail summary, run with output shown: - -``` -cargo test --release --test samples -- --nocapture -``` - -## Naming rules - -- An **input** is any `*.exe` / `*.dll` / `*.dat` whose name does **not** - contain the `.golden.` segment. -- A **golden** is `.golden.` sitting next to its input. Files with - `.golden.` in the name are never treated as inputs. -- A **companion** is `._` (e.g. `stub.dll._` for `stub.dll`). - Its extension is `_`, so it is never picked up as an input of its own; it is - read only when its base module is processed. diff --git a/senbei-cli/Cargo.toml b/senbei-cli/Cargo.toml new file mode 100644 index 0000000..f1beb2b --- /dev/null +++ b/senbei-cli/Cargo.toml @@ -0,0 +1,15 @@ +[package] +name = "senbei-cli" +version.workspace = true +edition.workspace = true +description = "Command-line entry point for Senbei" +license.workspace = true +keywords = ["unpacker", "reverse-engineering", "pe", "security-research"] +categories = ["command-line-utilities"] + +[[bin]] +name = "senbei" +path = "src/main.rs" + +[dependencies] +senbei-io.workspace = true diff --git a/src/main.rs b/senbei-cli/src/main.rs similarity index 80% rename from src/main.rs rename to senbei-cli/src/main.rs index da93dcc..f861f10 100644 --- a/src/main.rs +++ b/senbei-cli/src/main.rs @@ -1,4 +1,4 @@ -use senbei::{job, pause}; +use senbei_io::{job, pause, scan}; use std::path::Path; fn main() -> std::process::ExitCode { @@ -27,9 +27,6 @@ fn main() -> std::process::ExitCode { "--no-log" => no_log = true, "--scan-all" => scan_all = true, "--out" => match args.next() { - // Reject a missing value (and a following flag swallowed as the - // value): previously `--out` at end of argv silently fell back - // to the default output directory. Some(v) if !v.starts_with('-') => out = Some(v), _ => { eprintln!("error: --out requires a directory argument"); @@ -42,7 +39,6 @@ fn main() -> std::process::ExitCode { return std::process::ExitCode::from(2); } other => { - // Previously the last positional silently won. if let Some(prev) = &path { eprintln!("error: multiple input paths given ('{prev}' and '{other}')"); return std::process::ExitCode::from(2); @@ -63,33 +59,36 @@ fn main() -> std::process::ExitCode { } let p = Path::new(&p); let out_path = out.as_deref().map(Path::new); - let r = if p.is_dir() { + let result = if p.is_dir() { job::run_folder_opts( p, out_path, quiet, verbose, no_log, - scan_all || senbei::scan::scan_all_env(), + scan_all || scan::scan_all_env(), ) } else { job::run_file_v(p, out_path, quiet, verbose, no_log) }; - match r { - Ok(s) => { + match result { + Ok(summary) => { if quiet < 2 { println!( "{} unpacked · {} skipped · {} errors · {} suspect · {} metadata", - s.unpacked, s.skipped, s.errors, s.suspect, s.metadata + summary.unpacked, + summary.skipped, + summary.errors, + summary.suspect, + summary.metadata ); - println!("done in {} ms", s.duration_ms); + println!("done in {} ms", summary.duration_ms); } - if s.errors > 0 { 1 } else { 0 } + if summary.errors > 0 { 1 } else { 0 } } - Err(e) => { - // Fatal: out-dir/log create, etc. + Err(error) => { if quiet < 2 { - eprintln!("error: {e:#}"); + eprintln!("error: {error:#}"); } 1 } diff --git a/senbei-crypto/Cargo.toml b/senbei-crypto/Cargo.toml new file mode 100644 index 0000000..3ee7e4b --- /dev/null +++ b/senbei-crypto/Cargo.toml @@ -0,0 +1,9 @@ +[package] +name = "senbei-crypto" +version.workspace = true +edition.workspace = true +license.workspace = true +description = "Cryptographic and compression primitives for Senbei" + +[dependencies] +thiserror.workspace = true diff --git a/src/unpacker/bytecode.rs b/senbei-crypto/src/bytecode.rs similarity index 100% rename from src/unpacker/bytecode.rs rename to senbei-crypto/src/bytecode.rs diff --git a/src/unpacker/crc32.rs b/senbei-crypto/src/crc32.rs similarity index 100% rename from src/unpacker/crc32.rs rename to senbei-crypto/src/crc32.rs diff --git a/senbei-crypto/src/lib.rs b/senbei-crypto/src/lib.rs new file mode 100644 index 0000000..2f99e7f --- /dev/null +++ b/senbei-crypto/src/lib.rs @@ -0,0 +1,77 @@ +//! Cryptographic, checksum, compression, and bytecode primitives. + +pub mod bytecode; +pub mod crc32; +pub mod primitives; +mod tables; + +/// Maximum buffer size accepted by allocation-sensitive transforms. +pub const MAX_IMAGE_SIZE: u64 = 1 << 30; + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum BufferOperation { + Read, + CopySource, + CopyDestination, + ZeroFill, +} + +impl std::fmt::Display for BufferOperation { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::Read => "read", + Self::CopySource => "copy source", + Self::CopyDestination => "copy destination", + Self::ZeroFill => "zero-fill", + }) + } +} + +#[derive(Debug, Clone, PartialEq, Eq, thiserror::Error)] +pub enum Error { + #[error( + "{operation} range out of bounds (offset {offset}, size {size}, buffer length {buffer_len})" + )] + BufferRangeOutOfBounds { + operation: BufferOperation, + offset: usize, + size: usize, + buffer_len: usize, + }, +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)] +#[non_exhaustive] +pub enum DecompressionFailure { + #[error("compressed source size {size} exceeds limit {max}")] + SourceTooLarge { size: u32, max: u64 }, + #[error("Huffman code length {bits} is invalid")] + InvalidCodeLength { bits: u8 }, + #[error("Huffman tree traversal exceeded 64 levels")] + HuffmanTraversalLimit, + #[error("pending length accumulator overflowed at {pending}")] + PendingLengthOverflow { pending: u32 }, + #[error("output step {step} at byte {written} exceeds expected size {expected}")] + OutputOverflow { + written: u32, + step: u32, + expected: u32, + }, + #[error("run-fill width {width} reads before output offset 0x{destination:08X}")] + RunFillBeforeOutput { width: u32, destination: u32 }, + #[error("run-fill width {width} is unsupported")] + InvalidRunFillWidth { width: u32 }, + #[error("back-reference distance {distance} exceeds {written} written bytes")] + InvalidBackReference { distance: u32, written: u32 }, + #[error("Huffman symbol consumed no input and produced no output")] + NoProgress, + #[error( + "output size mismatch (wrote {written}/{expected} bytes after consuming {consumed}/{source_size})" + )] + OutputSizeMismatch { + written: u32, + expected: u32, + consumed: u32, + source_size: u32, + }, +} diff --git a/senbei-crypto/src/primitives.rs b/senbei-crypto/src/primitives.rs new file mode 100644 index 0000000..236f137 --- /dev/null +++ b/senbei-crypto/src/primitives.rs @@ -0,0 +1,1098 @@ +//! Shared crypto primitives and helper utilities. +//! +//! These functions form the low-level API used by the PE pipelines. +//! Each free function is self-contained: it takes the relevant byte buffer(s) +//! and parameters explicitly, with no coupling to the EXE `Unpacker` struct. + +use crate::bytecode::{Op, OpsLut}; +use crate::crc32; +use crate::tables::{COLUMMIX1, COLUMMIX2, COLUMMIX3, COLUMMIX4, SBOX}; +use std::cell::RefCell; + +thread_local! { + /// Reusable scratch for `decompress`. A single unpack runs `decompress` + /// hundreds of times over small blocks; reusing one growable buffer avoids a + /// fresh allocation each call. Thread-local, so it stays correct (one buffer + /// per worker) under the parallel block fan-out. + static DECOMPRESS_SCRATCH: RefCell> = const { RefCell::new(Vec::new()) }; +} + +// --------------------------------------------------------------------------- +// Byte-order accessors +// --------------------------------------------------------------------------- + +pub fn get_u16(data: &[u8], offset: u32) -> u16 { + let i = offset as usize; + u16::from_le_bytes([data[i], data[i + 1]]) +} + +pub fn get_u32(data: &[u8], offset: u32) -> u32 { + let i = offset as usize; + u32::from_le_bytes([data[i], data[i + 1], data[i + 2], data[i + 3]]) +} + +pub fn get_u64(data: &[u8], offset: u32) -> u64 { + let i = offset as usize; + u64::from_le_bytes([ + data[i], + data[i + 1], + data[i + 2], + data[i + 3], + data[i + 4], + data[i + 5], + data[i + 6], + data[i + 7], + ]) +} + +pub fn write_u16(data: &mut [u8], offset: u32, value: u32) { + let i = offset as usize; + let v = value as u16; + let b = v.to_le_bytes(); + data[i] = b[0]; + data[i + 1] = b[1]; +} + +pub fn write_u32(data: &mut [u8], offset: u32, value: u32) { + let i = offset as usize; + let b = value.to_le_bytes(); + data[i] = b[0]; + data[i + 1] = b[1]; + data[i + 2] = b[2]; + data[i + 3] = b[3]; +} + +// --------------------------------------------------------------------------- +// Checked accessors (return Err instead of panicking on OOB) +// --------------------------------------------------------------------------- + +#[allow(dead_code)] +pub fn try_u32(d: &[u8], off: usize) -> Result { + let end = off + .checked_add(4) + .ok_or(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::Read, + offset: off, + size: 4, + buffer_len: d.len(), + })?; + d.get(off..end) + .map(|s| u32::from_le_bytes(s.try_into().unwrap())) + .ok_or(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::Read, + offset: off, + size: 4, + buffer_len: d.len(), + }) +} + +#[allow(dead_code)] +pub fn try_i32(d: &[u8], off: usize) -> Result { + try_u32(d, off).map(|v| v as i32) +} + +/// Checked copy with distinct source and destination range errors. +pub fn try_copy_from_slice( + dst: &mut [u8], + dst_off: usize, + dst_len: usize, + src: &[u8], + src_off: usize, +) -> Result<(), crate::Error> { + let dst_end = dst_off + .checked_add(dst_len) + .ok_or(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopyDestination, + offset: dst_off, + size: dst_len, + buffer_len: dst.len(), + })?; + let src_end = src_off + .checked_add(dst_len) + .ok_or(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopySource, + offset: src_off, + size: dst_len, + buffer_len: src.len(), + })?; + if dst_end > dst.len() { + return Err(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopyDestination, + offset: dst_off, + size: dst_len, + buffer_len: dst.len(), + }); + } + if src_end > src.len() { + return Err(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopySource, + offset: src_off, + size: dst_len, + buffer_len: src.len(), + }); + } + dst[dst_off..dst_end].copy_from_slice(&src[src_off..src_end]); + Ok(()) +} + +/// Reproduce the LFSR keystream that decrypt_data6 XORs in. Used to +/// trial-decrypt candidate bytecode positions without mutating the buffer. +pub fn lfsr_keystream(out: &mut [u8]) { + let mut state: u32 = 1; + for byte in out.iter_mut() { + let mut b: u8 = 0; + for k in 0..8u32 { + b |= ((state & 1) << k) as u8; + state <<= 1; + if state & 0x8000 != 0 { + state ^= 0x8003; + } + } + *byte = b; + } +} + +// --------------------------------------------------------------------------- +// AES primitives +// --------------------------------------------------------------------------- + +/// One AES-CBC-like round over a 16-byte block in `d` at `pos`, using the +/// expanded key schedule stored in `d` at `key_offset`. Works entirely within +/// the single `d` buffer (both ciphertext and key schedule live there). +pub fn aes_round(d: &mut [u8], pos: u32, key_offset: u32, round: u32) { + let cm1 = &COLUMMIX1; + let cm2 = &COLUMMIX2; + let cm3 = &COLUMMIX3; + let cm4 = &COLUMMIX4; + let sbox = &SBOX; + + let mut n0 = get_u32(d, pos).swap_bytes() ^ get_u32(d, key_offset); + let mut n1 = + get_u32(d, pos.wrapping_add(4)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(4)); + let mut n2 = + get_u32(d, pos.wrapping_add(8)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(8)); + let mut n3 = + get_u32(d, pos.wrapping_add(12)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(12)); + + let mut r = 1u32; + while r < round { + let off = key_offset.wrapping_add(r.wrapping_mul(16)); + let a = get_u32(cm2, ((n3 >> 16) & 0xFF) * 4) + ^ get_u32(cm3, ((n2 >> 8) & 0xFF) * 4) + ^ get_u32(cm1, ((n0 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n1 & 0xFF) * 4) + ^ get_u32(d, off); + let b = get_u32(cm2, ((n0 >> 16) & 0xFF) * 4) + ^ get_u32(cm1, ((n1 >> 24) & 0xFF) * 4) + ^ get_u32(cm3, ((n3 >> 8) & 0xFF) * 4) + ^ get_u32(cm4, (n2 & 0xFF) * 4) + ^ get_u32(d, off.wrapping_add(4)); + let c = get_u32(cm2, ((n1 >> 16) & 0xFF) * 4) + ^ get_u32(cm3, ((n0 >> 8) & 0xFF) * 4) + ^ get_u32(cm1, ((n2 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n3 & 0xFF) * 4) + ^ get_u32(d, off.wrapping_add(8)); + let e = get_u32(cm3, ((n1 >> 8) & 0xFF) * 4) + ^ get_u32(cm2, ((n2 >> 16) & 0xFF) * 4) + ^ get_u32(cm1, ((n3 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n0 & 0xFF) * 4) + ^ get_u32(d, off.wrapping_add(12)); + n0 = a; + n1 = b; + n2 = c; + n3 = e; + r = r.wrapping_add(1); + } + + let s0 = (get_u32(sbox, ((n0 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n3 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n2 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n1 & 0xFF) * 4) & 0x0000_00FF); + let s1 = (get_u32(sbox, ((n1 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n0 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n3 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n2 & 0xFF) * 4) & 0x0000_00FF); + let s2 = (get_u32(sbox, ((n2 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n1 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n0 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n3 & 0xFF) * 4) & 0x0000_00FF); + let s3 = (get_u32(sbox, ((n3 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n2 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n1 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n0 & 0xFF) * 4) & 0x0000_00FF); + + let last = key_offset.wrapping_add(round.wrapping_mul(16)); + n0 = s0 ^ get_u32(d, last); + n1 = s1 ^ get_u32(d, last.wrapping_add(4)); + n2 = s2 ^ get_u32(d, last.wrapping_add(8)); + n3 = s3 ^ get_u32(d, last.wrapping_add(12)); + + write_u32(d, pos, n0.swap_bytes()); + write_u32(d, pos.wrapping_add(4), n1.swap_bytes()); + write_u32(d, pos.wrapping_add(8), n2.swap_bytes()); + write_u32(d, pos.wrapping_add(12), n3.swap_bytes()); +} + +/// AES-CBC-like decryption over `size` bytes starting at `pos` in `d`. +/// The key schedule lives at `key_offset` within the same buffer `d`. +pub fn aes_decrypt(d: &mut [u8], pos: u32, size: u32, key_offset: u32) { + let mut prev = [0u8; 16]; + let mut cur = [0u8; 16]; + let round = get_u16(d, key_offset.wrapping_add(2)) as u32; + let blocks = size >> 4; + for i in 0..blocks { + let p = pos.wrapping_add(i.wrapping_mul(16)); + let pi = p as usize; + cur.copy_from_slice(&d[pi..pi + 16]); + aes_round(d, p, key_offset.wrapping_add(4), round); + for j in 0..16 { + d[pi + j] ^= prev[j]; + } + prev = cur; + } +} + +/// [`aes_decrypt`] variant reading the key schedule from a separate snapshot +/// slice instead of the data buffer. `ks` is a snapshot of `d[key_offset..]` +/// taken by [`aes_schedule_snapshot`] (round count at `ks[2]`, round keys from +/// `ks[4]`), so the schedule extent is exactly right by construction. Used by +/// the parallel block fan-out, where each worker owns a disjoint `&mut` span +/// of the image and cannot read the schedule out of the shared buffer. +pub fn aes_decrypt_ks(ks: &[u8], d: &mut [u8], pos: u32, size: u32) { + let mut prev = [0u8; 16]; + let mut cur = [0u8; 16]; + let round = u16::from_le_bytes([ks[2], ks[3]]) as u32; + let sched = &ks[4..]; + let blocks = size >> 4; + for i in 0..blocks { + let p = pos.wrapping_add(i.wrapping_mul(16)); + let pi = p as usize; + cur.copy_from_slice(&d[pi..pi + 16]); + aes_round_ks(sched, d, p, round); + for j in 0..16 { + d[pi + j] ^= prev[j]; + } + prev = cur; + } +} + +/// [`aes_round`] with the round keys in a separate slice (see +/// [`aes_decrypt_ks`]). Identical math; only the key source differs. +fn aes_round_ks(ks: &[u8], d: &mut [u8], pos: u32, round: u32) { + let cm1 = &COLUMMIX1; + let cm2 = &COLUMMIX2; + let cm3 = &COLUMMIX3; + let cm4 = &COLUMMIX4; + let sbox = &SBOX; + let k = |i: u32| get_u32(ks, i); + + let mut n0 = get_u32(d, pos).swap_bytes() ^ k(0); + let mut n1 = get_u32(d, pos.wrapping_add(4)).swap_bytes() ^ k(4); + let mut n2 = get_u32(d, pos.wrapping_add(8)).swap_bytes() ^ k(8); + let mut n3 = get_u32(d, pos.wrapping_add(12)).swap_bytes() ^ k(12); + + let mut r = 1u32; + while r < round { + let off = r.wrapping_mul(16); + let a = get_u32(cm2, ((n3 >> 16) & 0xFF) * 4) + ^ get_u32(cm3, ((n2 >> 8) & 0xFF) * 4) + ^ get_u32(cm1, ((n0 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n1 & 0xFF) * 4) + ^ k(off); + let b = get_u32(cm2, ((n0 >> 16) & 0xFF) * 4) + ^ get_u32(cm1, ((n1 >> 24) & 0xFF) * 4) + ^ get_u32(cm3, ((n3 >> 8) & 0xFF) * 4) + ^ get_u32(cm4, (n2 & 0xFF) * 4) + ^ k(off.wrapping_add(4)); + let c = get_u32(cm2, ((n1 >> 16) & 0xFF) * 4) + ^ get_u32(cm3, ((n0 >> 8) & 0xFF) * 4) + ^ get_u32(cm1, ((n2 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n3 & 0xFF) * 4) + ^ k(off.wrapping_add(8)); + let e = get_u32(cm3, ((n1 >> 8) & 0xFF) * 4) + ^ get_u32(cm2, ((n2 >> 16) & 0xFF) * 4) + ^ get_u32(cm1, ((n3 >> 24) & 0xFF) * 4) + ^ get_u32(cm4, (n0 & 0xFF) * 4) + ^ k(off.wrapping_add(12)); + n0 = a; + n1 = b; + n2 = c; + n3 = e; + r = r.wrapping_add(1); + } + + let s0 = (get_u32(sbox, ((n0 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n3 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n2 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n1 & 0xFF) * 4) & 0x0000_00FF); + let s1 = (get_u32(sbox, ((n1 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n0 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n3 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n2 & 0xFF) * 4) & 0x0000_00FF); + let s2 = (get_u32(sbox, ((n2 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n1 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n0 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n3 & 0xFF) * 4) & 0x0000_00FF); + let s3 = (get_u32(sbox, ((n3 >> 24) & 0xFF) * 4) & 0xFF00_0000) + | (get_u32(sbox, ((n2 >> 16) & 0xFF) * 4) & 0x00FF_0000) + | (get_u32(sbox, ((n1 >> 8) & 0xFF) * 4) & 0x0000_FF00) + | (get_u32(sbox, (n0 & 0xFF) * 4) & 0x0000_00FF); + + let last = round.wrapping_mul(16); + n0 = s0 ^ k(last); + n1 = s1 ^ k(last.wrapping_add(4)); + n2 = s2 ^ k(last.wrapping_add(8)); + n3 = s3 ^ k(last.wrapping_add(12)); + + write_u32(d, pos, n0.swap_bytes()); + write_u32(d, pos.wrapping_add(4), n1.swap_bytes()); + write_u32(d, pos.wrapping_add(8), n2.swap_bytes()); + write_u32(d, pos.wrapping_add(12), n3.swap_bytes()); +} + +/// Snapshot the AES key schedule at `key_offset` for [`aes_decrypt_ks`]: +/// `d[key_offset .. key_offset + 4 + (round+1)*16]` where `round` is read from +/// the schedule header. Returns `None` when the header is truncated or the +/// round count is implausible (corrupt input — the same bytes would otherwise +/// drive reads past the buffer). +pub fn aes_schedule_snapshot(d: &[u8], key_offset: u32) -> Option> { + let base = key_offset as usize; + let round = u16::from_le_bytes([*d.get(base + 2)?, *d.get(base + 3)?]) as usize; + if round > 64 { + return None; + } + let end = base.checked_add(4 + (round + 1) * 16)?; + if end > d.len() { + return None; + } + Some(d[base..end].to_vec()) +} + +// --------------------------------------------------------------------------- +// Checksum primitives +// --------------------------------------------------------------------------- + +/// CRC32-based checksum over a (offset, length) descriptor pair embedded in +/// `d` at `pos`. Returns `crc32(d[offset..offset+length]) ^ length`. +pub fn calculate_checksum(d: &[u8], pos: u32) -> u32 { + let offset = get_u32(d, pos); + let length = get_u32(d, pos.wrapping_add(4)); + crc32::compute(&d[offset as usize..(offset + length) as usize]) ^ length +} + +/// CRC32 chained checksum. The (offset, length) descriptor at `pos` is read +/// from `d`; the bytes themselves are read from the separate `clean` buffer +/// (the original file image). `start` is the initial CRC accumulator. +pub fn calculate_checksum2(d: &[u8], clean: &[u8], pos: u32, start: u32) -> u32 { + let offset = get_u32(d, pos); + let length = get_u32(d, pos.wrapping_add(4)); + crc32::append(start, &clean[offset as usize..(offset + length) as usize]) +} + +// --------------------------------------------------------------------------- +// Decompression (Huffman/LZ) +// --------------------------------------------------------------------------- + +/// Huffman/LZ decompression operating entirely within a single `d` buffer. +/// Reads `s_size` bytes from `src`, writes `d_size` bytes to `dest`. +/// The Huffman table lives at `key_offset` within `d`. +/// +/// Returns a structured reason when the stream cannot produce exactly +/// `d_size` bytes. +pub fn decompress_detailed( + d: &mut [u8], + src: u32, + mut dest: u32, + key_offset: u32, + s_size: u32, + d_size: u32, +) -> Result<(), crate::DecompressionFailure> { + use crate::DecompressionFailure; + + // Bound the scratch allocation: a corrupt descriptor could request a + // multi-gigabyte source size, and an allocation failure aborts the process + // (uncatchable). Real payloads are far below this. + if s_size as u64 > crate::MAX_IMAGE_SIZE { + return Err(DecompressionFailure::SourceTooLarge { + size: s_size, + max: crate::MAX_IMAGE_SIZE, + }); + } + DECOMPRESS_SCRATCH.with_borrow_mut(|buf| -> Result<(), DecompressionFailure> { + let mut bit_pos: i32 = 0; + let need = (s_size as usize).saturating_add(3); + if buf.len() < need { + buf.resize(need, 0); + } + let mut buf_off: u32 = 0; + let mut src_consumed: i32 = 0; + let mut pending: u32 = 0; + let mut written: u32 = 0; + let src_u = src as usize; + let s_size_u = s_size as usize; + // The bit-reader's final get_u32 may read up to 3 bytes past s_size; those + // must be zero. Reused scratch can hold stale bytes there, so zero them + // before copying the (exactly s_size) source over the head. + buf[s_size_u] = 0; + buf[s_size_u + 1] = 0; + buf[s_size_u + 2] = 0; + buf[..s_size_u].copy_from_slice(&d[src_u..src_u + s_size_u]); + + while (src_consumed as u32) < s_size && written < d_size { + let word = get_u32(&buf[..], buf_off) >> bit_pos; + let tab_addr = key_offset.wrapping_add((word & 0xFF).wrapping_mul(3)); + let mut tab = get_u16(d, tab_addr); + let bits: u8; + if (tab & 0x8000) != 0 { + tab &= 0x7FFF; + bits = d[tab_addr as usize + 2]; + } else { + let mut b2 = d[tab_addr as usize + 2]; + // A Huffman code longer than 32 bits cannot exist; a larger + // length byte comes from a corrupt table, and `1 << b2` would + // panic (debug) or wrap (release) on it. + if b2 >= 32 { + return Err(DecompressionFailure::InvalidCodeLength { bits: b2 }); + } + let mut mask: u32 = 1u32 << b2; + b2 = b2.wrapping_add(1); + let mut idx = (tab & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; + let mut t2 = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); + // A corrupt table can form a non-terminal cycle; cap the walk so it + // fails instead of spinning forever. + let mut depth = 0u32; + while (t2 & 0x8000) == 0 { + depth += 1; + if depth > 64 { + return Err(DecompressionFailure::HuffmanTraversalLimit); + } + mask <<= 1; + b2 = b2.wrapping_add(1); + idx = (t2 & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; + t2 = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); + } + tab = t2 & 0x7FFF; + bits = b2; + } + bit_pos += bits as i32; + let advance = bit_pos / 8; + buf_off = buf_off.wrapping_add(advance as u32); + src_consumed += advance; + bit_pos %= 8; + + let mode = (tab as u32) & 0x300; + let payload = (tab as u32) & 0xFF; + let step: u32; + match mode { + 0 => { + step = 1; + d[dest as usize] = payload as u8; + } + 0x100 => { + step = 0; + if pending >= 256 { + return Err(DecompressionFailure::PendingLengthOverflow { pending }); + } + pending = if pending == 0 { + payload + } else { + (pending << 8) | payload + }; + } + 0x200 => { + if pending == 0 { + pending = 1; + } + step = pending.wrapping_mul(payload); + if step.wrapping_add(written) > d_size { + return Err(DecompressionFailure::OutputOverflow { + written, + step, + expected: d_size, + }); + } + // Run-fill replicates the unit just written before `dest`. A + // corrupt stream can emit one of these before anything has been + // written, so guard against reading before the buffer start + // (an unsigned underflow would index astronomically far OOB). + match payload { + 1 => { + if dest < 1 { + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); + } + let v = d[(dest as usize) - 1]; + for k in 0..pending { + d[(dest + k) as usize] = v; + } + } + 2 => { + if dest < 2 { + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); + } + let v = get_u16(d, dest.wrapping_sub(2)); + for k in 0..pending { + write_u16(d, dest.wrapping_add(k.wrapping_mul(2)), v as u32); + } + } + 4 => { + if dest < 4 { + return Err(DecompressionFailure::RunFillBeforeOutput { + width: payload, + destination: dest, + }); + } + let v = get_u32(d, dest.wrapping_sub(4)); + for k in 0..pending { + write_u32(d, dest.wrapping_add(k.wrapping_mul(4)), v); + } + } + _ => { + // Only unit widths 1/2/4 exist. Any other payload comes + // from a corrupt stream: previously this wrote nothing + // yet still counted `step` bytes as written, leaving + // stale-buffer holes that later stages treated as + // plaintext. Report corruption instead. + return Err(DecompressionFailure::InvalidRunFillWidth { + width: payload, + }); + } + } + pending = 0; + } + _ => { + step = payload; + if written.wrapping_add(payload) > d_size + || pending.wrapping_add(payload) > written + { + let distance = pending.wrapping_add(payload); + if distance > written { + return Err(DecompressionFailure::InvalidBackReference { + distance, + written, + }); + } + return Err(DecompressionFailure::OutputOverflow { + written, + step: payload, + expected: d_size, + }); + } + let back = pending.wrapping_add(payload); + for k in 0..payload { + d[(dest + k) as usize] = d[(dest + k - back) as usize]; + } + pending = 0; + } + } + + dest = dest.wrapping_add(step); + written = written.wrapping_add(step); + if bits == 0 && step == 0 { + // Corrupt table: no input bits consumed and no output bytes + // written, so the loop condition can never advance — an + // infinite loop (and `catch_unpack` traps panics, not hangs). + // Every real symbol consumes ≥ 1 bit, so a valid stream can + // never hit this. + return Err(DecompressionFailure::NoProgress); + } + } + src_consumed += if bit_pos != 0 { 1 } else { 0 }; + if written != d_size { + return Err(DecompressionFailure::OutputSizeMismatch { + written, + expected: d_size, + consumed: src_consumed.max(0) as u32, + source_size: s_size, + }); + } + Ok(()) + }) +} + +/// Boolean compatibility wrapper used by candidate searches and block fan-out. +pub fn decompress( + d: &mut [u8], + src: u32, + dest: u32, + key_offset: u32, + s_size: u32, + d_size: u32, +) -> bool { + decompress_detailed(d, src, dest, key_offset, s_size, d_size).is_ok() +} + +/// Walk the Huffman table at `key_offset` and snapshot its bytes for +/// [`decompress_tbl`]. The table is a forest of 256 root entries (3 bytes +/// each); non-terminal entries point at a child index pair. Returns `None` +/// when the table is truncated or self-referential past the buffer (corrupt +/// input — the same bytes would otherwise drive reads out of bounds). +pub fn huffman_table_snapshot(d: &[u8], key_offset: u32) -> Option> { + let mut visited = vec![false; 0x1_0000usize]; + let mut stack: Vec = (0..256).collect(); + let mut max_idx: u32 = 255; + while let Some(idx) = stack.pop() { + if idx >= 0x1_0000 || visited[idx as usize] { + continue; + } + visited[idx as usize] = true; + let off = key_offset as usize + idx as usize * 3; + if off + 3 > d.len() { + return None; + } + let t = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); + if (t & 0x8000) == 0 { + let child = (t & 0x7FFF) as u32; + max_idx = max_idx.max(child).max(child.wrapping_add(1)); + stack.push(child); + stack.push(child.wrapping_add(1)); + } + } + let end = key_offset as usize + (max_idx as usize + 1) * 3; + if end > d.len() { + return None; + } + Some(d[key_offset as usize..end].to_vec()) +} + +/// [`decompress`] variant reading the Huffman table from a separate snapshot +/// slice (see [`huffman_table_snapshot`]) instead of the data buffer. Used by +/// the parallel block fan-out, where each worker owns a disjoint `&mut` span +/// and cannot read the table out of the shared image. Table reads are bounds +/// checked against the snapshot — past-the-end means corrupt table, reported +/// as `false` rather than a panic. +pub fn decompress_tbl( + tab: &[u8], + d: &mut [u8], + src: u32, + mut dest: u32, + s_size: u32, + d_size: u32, +) -> bool { + if s_size as u64 > crate::MAX_IMAGE_SIZE { + return false; + } + DECOMPRESS_SCRATCH.with_borrow_mut(|buf| { + // Table reads, bounds-checked against the snapshot. + let tab16 = |addr: usize| -> Option { + let b = tab.get(addr..addr + 3)?; + Some(u16::from_le_bytes([b[0], b[1]])) + }; + let tab8 = |addr: usize| -> Option { tab.get(addr + 2).copied() }; + + let mut bit_pos: i32 = 0; + let need = (s_size as usize).saturating_add(3); + if buf.len() < need { + buf.resize(need, 0); + } + let mut buf_off: u32 = 0; + let mut src_consumed: i32 = 0; + let mut pending: u32 = 0; + let mut written: u32 = 0; + let src_u = src as usize; + let s_size_u = s_size as usize; + buf[s_size_u] = 0; + buf[s_size_u + 1] = 0; + buf[s_size_u + 2] = 0; + buf[..s_size_u].copy_from_slice(&d[src_u..src_u + s_size_u]); + + while (src_consumed as u32) < s_size && written < d_size { + let word = get_u32(&buf[..], buf_off) >> bit_pos; + let tab_addr = ((word & 0xFF).wrapping_mul(3)) as usize; + let mut tab = match tab16(tab_addr) { + Some(t) => t, + None => { + return false; + } + }; + let bits: u8; + if (tab & 0x8000) != 0 { + tab &= 0x7FFF; + bits = match tab8(tab_addr) { + Some(b) => b, + None => return false, + }; + } else { + let mut b2 = match tab8(tab_addr) { + Some(b) => b, + None => return false, + }; + if b2 >= 32 { + return false; + } + let mut mask: u32 = 1u32 << b2; + b2 = b2.wrapping_add(1); + let mut idx = (tab & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; + let mut t2 = match tab16(idx as usize * 3) { + Some(t) => t, + None => { + return false; + } + }; + // A corrupt table can form a non-terminal cycle; cap the walk so it + // fails instead of spinning forever. + let mut depth = 0u32; + while (t2 & 0x8000) == 0 { + depth += 1; + if depth > 64 { + return false; + } + mask <<= 1; + b2 = b2.wrapping_add(1); + idx = (t2 & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; + t2 = match tab16(idx as usize * 3) { + Some(t) => t, + None => { + return false; + } + }; + } + tab = t2 & 0x7FFF; + bits = b2; + } + bit_pos += bits as i32; + let advance = bit_pos / 8; + buf_off = buf_off.wrapping_add(advance as u32); + src_consumed += advance; + bit_pos %= 8; + + let mode = (tab as u32) & 0x300; + let payload = (tab as u32) & 0xFF; + let step: u32; + match mode { + 0 => { + step = 1; + d[dest as usize] = payload as u8; + } + 0x100 => { + step = 0; + if pending >= 256 { + return false; + } + pending = if pending == 0 { + payload + } else { + (pending << 8) | payload + }; + } + 0x200 => { + if pending == 0 { + pending = 1; + } + step = pending.wrapping_mul(payload); + if step.wrapping_add(written) > d_size { + return false; + } + // Run-fill replicates the unit just written before `dest` + // (see `decompress` for the underflow rationale). + match payload { + 1 => { + if dest < 1 { + return false; + } + let v = d[(dest as usize) - 1]; + for k in 0..pending { + d[(dest + k) as usize] = v; + } + } + 2 => { + if dest < 2 { + return false; + } + let v = get_u16(d, dest.wrapping_sub(2)); + for k in 0..pending { + write_u16(d, dest.wrapping_add(k.wrapping_mul(2)), v as u32); + } + } + 4 => { + if dest < 4 { + return false; + } + let v = get_u32(d, dest.wrapping_sub(4)); + for k in 0..pending { + write_u32(d, dest.wrapping_add(k.wrapping_mul(4)), v); + } + } + _ => { + return false; + } + } + pending = 0; + } + _ => { + step = payload; + if written.wrapping_add(payload) > d_size + || pending.wrapping_add(payload) > written + { + return false; + } + let back = pending.wrapping_add(payload); + for k in 0..payload { + d[(dest + k) as usize] = d[(dest + k - back) as usize]; + } + pending = 0; + } + } + + dest = dest.wrapping_add(step); + written = written.wrapping_add(step); + if bits == 0 && step == 0 { + return false; + } + } + src_consumed += if bit_pos != 0 { 1 } else { 0 }; + let _ = src_consumed; + written == d_size + }) +} +// Decrypt primitives (free-function wrappers) +// --------------------------------------------------------------------------- + +/// decrypt_data3: XOR+rotate cipher. Reads/writes dwords in `d` starting at +/// the address stored at `d[pos]`, for `d[pos+4]>>2` words. `shift` is the +/// right-rotate amount (19 or 21 depending on caller). +pub fn decrypt_data3(d: &mut [u8], pos: u32, mut key: u32, shift: u32) { + let base_addr = get_u32(d, pos); + let length = get_u32(d, pos.wrapping_add(4)); + let words = length >> 2; + for i in 0..words { + let off = base_addr.wrapping_add(i.wrapping_mul(4)); + let v = get_u32(d, off) ^ key; + key = key.wrapping_add(i); + let rotated = v.rotate_right(shift); + write_u32(d, off, rotated.wrapping_sub(i)); + } +} + +/// decrypt_data1 (called `decrypt_data` in the original): decode the 8-dword +/// info header from `file_data` at offset 4096 and write results into `info`. +pub fn decrypt_data1(file_data: &[u8], info: &mut [u32; 8]) { + info[0] = get_u32(file_data, 4096); + let mut k = get_u32(file_data, 4096); + for i in 0..7u32 { + let off = i.wrapping_mul(4).wrapping_add(4); + let cell = get_u32(file_data, 4096u32.wrapping_add(off)); + info[(i + 1) as usize] = k ^ cell; + k = i.wrapping_mul(i) ^ (k.wrapping_add(cell).wrapping_sub(i)); + } +} + +/// decrypt_data6: LFSR XOR decryption of a bytecode block at `pos` in `d`. +/// The block length is read from `d[pos + 95]`. +pub fn decrypt_data6(d: &mut [u8], pos: u32) { + let len = d[(pos + 95) as usize] as usize; + // The keystream is exactly `lfsr_keystream`'s — generate it once (len is a + // byte, so 256 always covers it) instead of keeping a second copy of the + // LFSR that a future poly fix would have to update separately. + let mut ks = [0u8; 256]; + lfsr_keystream(&mut ks); + let pos = pos as usize; + for i in 0..len { + d[pos + i] ^= ks[i]; + } +} + +/// decrypt_data7: nibble-swap + key-rolling byte cipher applied to a +/// null-terminated string in `d` starting at `pos`. +pub fn decrypt_data7(d: &mut [u8], pos: u32, mut key: u8) { + let mut i: u32 = 0; + loop { + let idx = (pos + i) as usize; + if d[idx] == 0 { + break; + } + let mut b = d[idx]; + b = b.rotate_right(4); + b = b.wrapping_sub(key); + if b == 0 { + b = 0u8.wrapping_sub(key); + } + d[idx] = b; + key = key.wrapping_add(67); + i += 1; + } +} + +// --------------------------------------------------------------------------- +// Higher-level composite: AES + decrypt3 + optional bytecode + decompress +// --------------------------------------------------------------------------- + +/// Decrypt and optionally decompress a stage payload descriptor. +/// `pos` points to a (src, src_len, dest, dest_len) quad of dwords in `d`. +/// - AES-decrypts `src..src+src_len` using key at `key3_offset` +/// - XOR+rotate-decrypts with `decrypt_data3(pos, key, 19)` +/// - Applies optional custom `ops` bytecode per-byte +/// - If `src_len != dest_len`, Huffman/LZ-decompresses `src..` → `dest..` +/// +/// Returns the decompression success status (always `true` when no +/// decompression was needed). The PE32 eighth-stage key search relies on this. +pub fn decrypt_and_decompress_data_detailed( + d: &mut [u8], + pos: u32, + key: u32, + key1_offset: u32, + key3_offset: u32, + ops: Option<&[Op]>, +) -> Result<(), crate::DecompressionFailure> { + let src = get_u32(d, pos); + let src_len = get_u32(d, pos.wrapping_add(4)); + aes_decrypt(d, src, src_len, key3_offset); + decrypt_data3(d, pos, key, 19); + if let Some(ops) = ops + && src_len != 0 + { + OpsLut::new(ops).map_region(d, src as usize, src_len as usize); + } + let dest = get_u32(d, pos.wrapping_add(8)); + let dest_len = get_u32(d, pos.wrapping_add(12)); + if src_len != dest_len { + return decompress_detailed(d, src, dest, key1_offset, src_len, dest_len); + } + Ok(()) +} + +/// Boolean compatibility wrapper used by key searches that trial candidates. +pub fn decrypt_and_decompress_data( + d: &mut [u8], + pos: u32, + key: u32, + key1_offset: u32, + key3_offset: u32, + ops: Option<&[Op]>, +) -> bool { + decrypt_and_decompress_data_detailed(d, pos, key, key1_offset, key3_offset, ops).is_ok() +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn checked_copy_distinguishes_source_and_destination_ranges() { + let mut short_destination = [0u8; 2]; + let source = [1u8; 4]; + let error = try_copy_from_slice(&mut short_destination, 0, 3, &source, 0) + .expect_err("destination must be rejected"); + assert!(matches!( + error, + crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopyDestination, + offset: 0, + size: 3, + buffer_len: 2, + } + )); + + let mut destination = [0u8; 4]; + let short_source = [1u8; 2]; + let error = try_copy_from_slice(&mut destination, 0, 3, &short_source, 0) + .expect_err("source must be rejected"); + assert!(matches!( + error, + crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::CopySource, + offset: 0, + size: 3, + buffer_len: 2, + } + )); + } + + #[test] + fn aes_ks_variant_matches_single_buffer() { + // Random-ish key schedule at ko and data block; both variants must + // produce identical output. + let ko: usize = 0x40; + let mut d = vec![0u8; 0x400]; + let mut x: u32 = 0x12345678; + for b in d.iter_mut() { + x = x.wrapping_mul(1664525).wrapping_add(1013904223); + *b = (x >> 24) as u8; + } + d[ko + 2] = 10; // round count = 10 + d[ko + 3] = 0; + let snap = aes_schedule_snapshot(&d, ko as u32).expect("snapshot"); + + let mut a = d.clone(); + aes_decrypt(&mut a, 0x100, 0x80, ko as u32); + let mut b = d.clone(); + aes_decrypt_ks(&snap, &mut b, 0x100, 0x80); + if a != b { + let idx = (0..a.len()).find(|&i| a[i] != b[i]).unwrap(); + panic!( + "first diff at {idx:#x}: a={:02x} b={:02x}\n a[..]: {:02x?}\n b[..]: {:02x?}", + a[idx], + b[idx], + &a[idx..idx + 16], + &b[idx..idx + 16] + ); + } + } + + #[test] + fn dtbl_variant_matches_single_buffer() { + // Real table + real compressed block lifted from an actual unpack is + // covered by the golden suite; here we just check a trivial stream: + // build a table where every byte is a literal (mode 0, 8 bits), then + // a source stream of N bytes should expand to N identical bytes. + let ko: usize = 0x100; + let mut d = vec![0u8; 0x1000]; + for e in 0..256usize { + let off = ko + e * 3; + let sym = 0x8000u16 | (e as u16 & 0xFF); // terminal, mode 0, payload=e + d[off] = (sym & 0xFF) as u8; + d[off + 1] = (sym >> 8) as u8; + d[off + 2] = 8; // 8 bits per symbol + } + // Source: 16 bytes 0x00..0x0F at src. + let src = 0x600u32; + for i in 0..16u32 { + d[(src + i) as usize] = i as u8; + } + let snap = huffman_table_snapshot(&d, ko as u32).expect("table snapshot"); + + let mut a = vec![0u8; 0x1000]; + a[..d.len()].copy_from_slice(&d); + assert!(decompress(&mut a, src, 0x800, ko as u32, 16, 16)); + let mut b = d.clone(); + assert!(decompress_tbl(&snap, &mut b, src, 0x800, 16, 16)); + assert_eq!(&a[0x800..0x810], &b[0x800..0x810]); + assert_eq!(&b[0x800..0x810], &(0u8..16).collect::>()[..]); + } + + /// Review regression: a run-fill token with a unit width other than 1/2/4 + /// comes from a corrupt stream and must report failure — previously it + /// wrote nothing yet still counted the bytes as written, leaving stale + /// holes that later stages treated as plaintext. + #[test] + fn decompress_rejects_unknown_run_fill_width() { + // Huffman table at key_offset 0, entry 0: terminal symbol with + // mode 0x200 (run-fill), payload 3 (invalid width), code length 8. + let mut d = vec![0u8; 0x100]; + let sym: u16 = 0x8000 | 0x203; + d[0..2].copy_from_slice(&sym.to_le_bytes()); + d[2] = 8; + // All-zero source -> symbol index 0 -> the invalid run-fill. + assert_eq!( + decompress_detailed(&mut d, 0x40, 0x80, 0, 4, 3), + Err(crate::DecompressionFailure::InvalidRunFillWidth { width: 3 }) + ); + } + + /// Control for the above: a width-1 run-fill is legal and succeeds. + #[test] + fn decompress_accepts_width1_run_fill() { + let mut d = vec![0u8; 0x100]; + d[0x7F] = 0x5A; // unit to replicate + let sym: u16 = 0x8000 | 0x201; + d[0..2].copy_from_slice(&sym.to_le_bytes()); + d[2] = 8; + assert!(decompress(&mut d, 0x40, 0x80, 0, 4, 3)); + assert_eq!(&d[0x80..0x83], &[0x5A, 0x5A, 0x5A]); + } +} diff --git a/src/unpacker/tables.rs b/senbei-crypto/src/tables.rs similarity index 92% rename from src/unpacker/tables.rs rename to senbei-crypto/src/tables.rs index 04dcc9b..29fb9fd 100644 --- a/src/unpacker/tables.rs +++ b/senbei-crypto/src/tables.rs @@ -146,11 +146,11 @@ mod tests { #[test] fn generated_tables_match_committed_bytes() { assert_eq!(COLUMMIX1.len(), 1024); - assert_eq!(super::super::crc32::compute(&COLUMMIX1), 0x7e8d_5d5f); - assert_eq!(super::super::crc32::compute(&COLUMMIX2), 0xfcc4_acfc); - assert_eq!(super::super::crc32::compute(&COLUMMIX3), 0x637a_f0cd); - assert_eq!(super::super::crc32::compute(&COLUMMIX4), 0x1e7b_c381); - assert_eq!(super::super::crc32::compute(&SBOX), 0x10fd_6dc1); + assert_eq!(crate::crc32::compute(&COLUMMIX1), 0x7e8d_5d5f); + assert_eq!(crate::crc32::compute(&COLUMMIX2), 0xfcc4_acfc); + assert_eq!(crate::crc32::compute(&COLUMMIX3), 0x637a_f0cd); + assert_eq!(crate::crc32::compute(&COLUMMIX4), 0x1e7b_c381); + assert_eq!(crate::crc32::compute(&SBOX), 0x10fd_6dc1); // Spot-check the first dword of each (matches the original first row). assert_eq!(&COLUMMIX1[..4], &[0x50, 0xa7, 0xf4, 0x51]); assert_eq!(&COLUMMIX2[..4], &[0xa7, 0xf4, 0x51, 0x50]); diff --git a/senbei-io/Cargo.toml b/senbei-io/Cargo.toml new file mode 100644 index 0000000..e5eb407 --- /dev/null +++ b/senbei-io/Cargo.toml @@ -0,0 +1,23 @@ +[package] +name = "senbei-io" +version.workspace = true +edition.workspace = true +license.workspace = true +description = "Filesystem, scanning, logging, and CLI orchestration for Senbei" + +[dependencies] +anyhow.workspace = true +indicatif.workspace = true +owo-colors.workspace = true +senbei-metadata.workspace = true +senbei-pe.workspace = true +walkdir.workspace = true + +[target.'cfg(windows)'.dependencies] +windows.workspace = true + +[target.'cfg(not(windows))'.dependencies] +libc.workspace = true + +[dev-dependencies] +tempfile.workspace = true diff --git a/src/job.rs b/senbei-io/src/job.rs similarity index 94% rename from src/job.rs rename to senbei-io/src/job.rs index 5d3d22d..32d315a 100644 --- a/src/job.rs +++ b/senbei-io/src/job.rs @@ -1,4 +1,4 @@ -use crate::unpacker; +use senbei_pe as unpacker; use std::path::{Path, PathBuf}; /// Crackproof header key table lives at this fixed file offset. For the @@ -518,7 +518,7 @@ pub fn run_folder_opts( // il2cpp metadata pass. Crackproof's `-GMD` option obfuscates the method // tokens in `global-metadata.dat`; de-obfuscate any we find so the unpacked // il2cpp game assembly resolves methods instead of indexing its per-module - // tables out of bounds (see [`crate::metadata`]). This is additive to the + // tables out of bounds (see [`senbei_metadata`]). This is additive to the // Crackproof module unpack above — the metadata blob is not itself a // Crackproof file. for meta in metas { @@ -660,7 +660,7 @@ pub fn run_file_v( let mut buf = [0u8; 4]; std::fs::File::open(input) .and_then(|mut f| f.read_exact(&mut buf)) - .map(|_| crate::metadata::is_metadata(&buf)) + .map(|_| senbei_metadata::is_metadata(&buf)) .unwrap_or(false) }; @@ -801,13 +801,13 @@ fn panic_payload(panic: &(dyn std::any::Any + Send)) -> String { } } -/// If `e`'s chain contains [`crate::metadata::Error::UnsupportedVersion`], +/// If `e`'s chain contains [`senbei_metadata::Error::UnsupportedVersion`], /// return the version. Used to apply the folder-mode "leave untouched, don't /// fail the run" policy to metadata versions this build can't de-obfuscate. fn unsupported_version(e: &anyhow::Error) -> Option { for cause in e.chain() { - if let Some(crate::metadata::Error::UnsupportedVersion(v)) = - cause.downcast_ref::() + if let Some(senbei_metadata::Error::UnsupportedVersion(v)) = + cause.downcast_ref::() { return Some(*v); } @@ -831,26 +831,20 @@ fn write_atomic(dest: &Path, bytes: &[u8]) -> std::io::Result<()> { r } -/// Detect `bytes` and run the right pipeline. The EXE pipeline is invoked -/// directly (no DLL-pipeline probe) when the input was spliced from an -/// external companion (`spliced`) or when the caller forces it (`force_exe` -/// — the web app's recovery path after a DLL-probe trap; see -/// [`unpack_bytes_force_exe`]). +/// Detect `bytes` and run the right pipeline. Spliced external companions use +/// the EXE pipeline directly because that layout is definitionally EXE-style. /// /// Routing spliced inputs straight to the EXE pipeline is safe: the /// companion layout is definitionally the EXE-style shell (the runtime /// loader maps the companion and runs the standard shell unpack), so the DLL -/// pipeline probe can never be right for it — and probing is not a no-op on -/// targets without unwinding (wasm), where the probe's caught panic becomes -/// a fatal trap. Output bytes are identical to the dll-first + exe-fallback -/// route for every input that route handles. +/// pipeline probe can never be right for it. Output bytes are identical to the +/// DLL-first + EXE-fallback route for every input that route handles. fn unpack_spliced_or_auto( bytes: &[u8], spliced: bool, - force_exe: bool, verbose: bool, ) -> Result<(unpacker::Kind, Vec), unpacker::UnpackError> { - if spliced || force_exe { + if spliced { let detected = unpacker::detect(bytes).ok_or(unpacker::UnpackError::NotCrackproof)?; let out = unpacker::unpack_exe_v(bytes, verbose)?; return Ok((detected.kind, out)); @@ -880,38 +874,23 @@ pub struct UnpackedImage { /// Unpack in-memory `input` bytes, optionally paired with an external /// companion payload `companion` (the `._` file's contents). /// -/// This is the I/O-free counterpart of [`unpack_one_v`], used by the -/// WebAssembly build: splice (when the companion's first 32 bytes match the -/// stub header), unpack, overlay the export table and TLS directory from the -/// stub, then run the static integrity check. +/// This is the in-memory counterpart of [`unpack_one_v`]: splice a matching +/// companion, unpack, overlay the export table and TLS directory from the stub, +/// then run the static integrity check. pub fn unpack_bytes( input: &[u8], companion: Option<&[u8]>, ) -> Result { - unpack_bytes_impl(input, companion, false) -} - -/// Like [`unpack_bytes`], but forces the EXE pipeline (no DLL-pipeline -/// probe). This is the web app's recovery path: the DLL-first probe relies -/// on `catch_unwind` to reject EXE-shell-layout DLLs, and panics cannot be -/// caught on wasm — the probe traps the whole call. The web app runs each -/// unpack in a disposable Web Worker and retries trapped DLLs with this -/// entry point, reproducing the CLI's dll-first/exe-fallback routing. -pub fn unpack_bytes_force_exe( - input: &[u8], - companion: Option<&[u8]>, -) -> Result { - unpack_bytes_impl(input, companion, true) + unpack_bytes_impl(input, companion) } fn unpack_bytes_impl( input: &[u8], companion: Option<&[u8]>, - force_exe: bool, ) -> Result { let spliced = companion.and_then(|c| splice_companion(input, c)); let bytes: &[u8] = spliced.as_deref().unwrap_or(input); - let (kind, mut out) = unpack_spliced_or_auto(bytes, spliced.is_some(), force_exe, false)?; + let (kind, mut out) = unpack_spliced_or_auto(bytes, spliced.is_some(), false)?; if spliced.is_some() { overlay_exports_from_stub(&mut out, input); restore_tls_from_stub(&mut out, input); @@ -933,7 +912,7 @@ pub fn unpack_one_v( verbose: bool, ) -> anyhow::Result<(unpacker::Kind, unpacker::IntegrityReport)> { let UnpackerInput { bytes, stub } = read_unpacker_input(input)?; - let (kind, mut out) = unpack_spliced_or_auto(&bytes, stub.is_some(), false, verbose)?; + let (kind, mut out) = unpack_spliced_or_auto(&bytes, stub.is_some(), verbose)?; // External-companion layout: restore the export table from the stub, which // the encrypted companion does not carry (the loader rebuilds it at runtime). if let Some(stub) = stub { @@ -960,7 +939,7 @@ pub fn unpack_one_v( /// into a sparse, original-metadata-style value; il2cpp expects the contiguous /// per-module index it indexes its codegen tables with, so a statically-unpacked /// il2cpp game assembly reads garbage and crashes during init. This rewrites -/// the tokens back to their canonical form (see [`crate::metadata::deobfuscate`]). +/// the tokens back to their canonical form (see [`senbei_metadata::deobfuscate`]). /// /// The output is written only when something actually changed /// (`report.remapped > 0`); an already-clean metadata is left untouched and no @@ -970,11 +949,11 @@ pub fn deobfuscate_metadata_to( input: &Path, dest: &Path, verbose: bool, -) -> anyhow::Result { +) -> anyhow::Result { let data = std::fs::read(input)?; // Preserve the metadata::Error in the chain (rather than stringifying it) // so the folder driver can apply its unsupported-version policy. - let (out, report) = crate::metadata::deobfuscate(&data) + let (out, report) = senbei_metadata::deobfuscate(&data) .map_err(|e| anyhow::Error::new(e).context(format!("{input:?}")))?; if report.remapped > 0 { if let Some(parent) = dest.parent() { diff --git a/src/lib.rs b/senbei-io/src/lib.rs similarity index 59% rename from src/lib.rs rename to senbei-io/src/lib.rs index 270c372..5147a04 100644 --- a/src/lib.rs +++ b/senbei-io/src/lib.rs @@ -1,7 +1,7 @@ +//! Filesystem and command-line orchestration. + pub mod job; pub mod logfile; -pub mod metadata; pub mod pause; pub mod scan; pub mod ui; -pub mod unpacker; diff --git a/src/logfile.rs b/senbei-io/src/logfile.rs similarity index 92% rename from src/logfile.rs rename to senbei-io/src/logfile.rs index 7027520..ead7ed2 100644 --- a/src/logfile.rs +++ b/senbei-io/src/logfile.rs @@ -79,7 +79,7 @@ fn local_parts() -> LocalParts { second: st.wSecond as u32, } } - #[cfg(all(not(windows), not(target_arch = "wasm32")))] + #[cfg(not(windows))] { // Local wall clock via POSIX localtime_r — same semantics as Windows GetLocalTime. use std::time::{SystemTime, UNIX_EPOCH}; @@ -102,13 +102,6 @@ fn local_parts() -> LocalParts { second: tm.tm_sec as u32, } } - #[cfg(all(not(windows), target_arch = "wasm32"))] - { - // wasm has no local timezone database and SystemTime::now() panics - // without a JS time source. The run log is a CLI concern — the wasm - // build never writes one — so a fixed epoch stamp suffices. - utc_parts(0) - } } /// Convert Unix UTC seconds to civil Y-M-D h:m:s (Howard Hinnant). diff --git a/src/pause.rs b/senbei-io/src/pause.rs similarity index 100% rename from src/pause.rs rename to senbei-io/src/pause.rs diff --git a/src/scan.rs b/senbei-io/src/scan.rs similarity index 98% rename from src/scan.rs rename to senbei-io/src/scan.rs index d0308f4..43f46ee 100644 --- a/src/scan.rs +++ b/senbei-io/src/scan.rs @@ -1,4 +1,4 @@ -use crate::unpacker::detect; +use senbei_pe::detect; use std::io::Read; use std::path::{Path, PathBuf}; use walkdir::WalkDir; @@ -16,12 +16,12 @@ const DETECT_PREFIX: u64 = 8 * 1024; /// Smallest file that can possibly be a target, so anything shorter is skipped /// without ever being opened. /// -/// A Crackproof module needs ≥ 4128 bytes for [`crate::unpacker::detect`]'s key +/// A Crackproof module needs ≥ 4128 bytes for [`senbei_pe::detect`]'s key /// table (it reads the dword at 4124), so the bound is exact for the unpack /// path. An il2cpp `global-metadata.dat` only needs 4 bytes to match its magic, /// but its header alone runs to offset 0xB0 and the images/types/methods tables /// it indexes make every real one megabytes long — a sub-4 KiB "metadata" could -/// only ever fail [`crate::metadata::deobfuscate`] with `Malformed`, so nothing +/// only ever fail [`senbei_metadata::deobfuscate`] with `Malformed`, so nothing /// processable is lost. const MIN_SIZE: u64 = 4128; @@ -223,7 +223,7 @@ pub fn find_targets_opts(root: &Path, scan_all: bool) -> (Vec, Vec> = vec![Some(Class::None); n]; - let workers = crate::unpacker::parallel::thread_cap().clamp(1, n.max(1)); + let workers = senbei_pe::thread_cap().clamp(1, n.max(1)); if workers <= 1 { for (p, c) in paths.iter().zip(class.iter_mut()) { *c = classify(p); @@ -304,7 +304,7 @@ fn classify(path: &Path) -> Option { let r = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| { if detect(&head).is_some() { Class::Crackproof - } else if crate::metadata::is_metadata(&head) { + } else if senbei_metadata::is_metadata(&head) { Class::Metadata } else { Class::None diff --git a/src/ui.rs b/senbei-io/src/ui.rs similarity index 97% rename from src/ui.rs rename to senbei-io/src/ui.rs index d882745..8d49ce7 100644 --- a/src/ui.rs +++ b/senbei-io/src/ui.rs @@ -1,6 +1,6 @@ -use crate::unpacker::{IntegrityReport, Kind}; use indicatif::{ProgressBar, ProgressStyle}; use owo_colors::OwoColorize; +use senbei_pe::{IntegrityReport, Kind}; use std::path::Path; /// Create a progress bar for `n` items. Hidden when `quiet` is true. diff --git a/senbei-metadata/Cargo.toml b/senbei-metadata/Cargo.toml new file mode 100644 index 0000000..c9d49e2 --- /dev/null +++ b/senbei-metadata/Cargo.toml @@ -0,0 +1,6 @@ +[package] +name = "senbei-metadata" +version.workspace = true +edition.workspace = true +license.workspace = true +description = "Unity il2cpp metadata de-obfuscation for Senbei" diff --git a/senbei-metadata/src/lib.rs b/senbei-metadata/src/lib.rs new file mode 100644 index 0000000..18e6fc7 --- /dev/null +++ b/senbei-metadata/src/lib.rs @@ -0,0 +1,5 @@ +//! Unity il2cpp metadata de-obfuscation. + +mod metadata; + +pub use metadata::*; diff --git a/src/metadata.rs b/senbei-metadata/src/metadata.rs similarity index 100% rename from src/metadata.rs rename to senbei-metadata/src/metadata.rs diff --git a/senbei-pe/Cargo.toml b/senbei-pe/Cargo.toml new file mode 100644 index 0000000..8790c8b --- /dev/null +++ b/senbei-pe/Cargo.toml @@ -0,0 +1,10 @@ +[package] +name = "senbei-pe" +version.workspace = true +edition.workspace = true +license.workspace = true +description = "PE detection, unpacking, and validation for Senbei" + +[dependencies] +senbei-crypto.workspace = true +thiserror.workspace = true diff --git a/senbei-pe/src/engine/dll/mod.rs b/senbei-pe/src/engine/dll/mod.rs new file mode 100644 index 0000000..bf70ed1 --- /dev/null +++ b/senbei-pe/src/engine/dll/mod.rs @@ -0,0 +1,3 @@ +mod pipeline; + +pub use pipeline::*; diff --git a/src/unpacker/dll.rs b/senbei-pe/src/engine/dll/pipeline.rs similarity index 98% rename from src/unpacker/dll.rs rename to senbei-pe/src/engine/dll/pipeline.rs index 794e5e0..b14c4b0 100644 --- a/src/unpacker/dll.rs +++ b/senbei-pe/src/engine/dll/pipeline.rs @@ -13,12 +13,12 @@ //! CalculateChecksumWithSizeXor -> primitives::calculate_checksum //! CalculateCrc32 -> crc32::compute (via above) -use super::bytecode::{Op, OpsLut, generate}; -use super::primitives::{self, *}; -use super::{ +use super::super::{ BufferOperation, BytecodeStage, DecompressionStage, DescriptorTable, SectionPipeline, UnpackError, }; +use senbei_crypto::bytecode::{Op, OpsLut, generate}; +use senbei_crypto::primitives::{self, *}; /// Read a signed 32-bit little-endian value. fn get_i32(d: &[u8], offset: i32) -> i32 { @@ -291,7 +291,7 @@ fn decrypt_and_decompress_data( } Ok(()) }; - super::parallel::parallel_for(d, &spans, 1, do_block)?; + super::super::parallel::parallel_for(d, &spans, 1, do_block)?; } // Zero-fill loop. @@ -342,7 +342,7 @@ pub fn unpack_dll(input: &[u8]) -> Result, UnpackError> { pub fn unpack_dll_v(input: &[u8], verbose: bool) -> Result, UnpackError> { // Trap any out-of-bounds panic from a truncated/garbled file and report it // as a clean error so the public API stays panic-free. - super::catch_unpack(move || unpack_dll_inner(input, verbose)) + super::super::catch_unpack(move || unpack_dll_inner(input, verbose)) } fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> { @@ -369,7 +369,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> println!(" keys[6] anchor = 0x{:08X}", keys[6] as u32); } - if !super::is_supported_magic(keys[1] as u32) { + if !super::super::is_supported_magic(keys[1] as u32) { return Err(UnpackError::HeaderMagicMismatch { found: keys[1] as u32, }); @@ -395,10 +395,10 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> }); } let size_of_image = get_i32(file_data, pe_offset + 80); - if size_of_image <= 0 || size_of_image as u64 > super::MAX_IMAGE_SIZE { + if size_of_image <= 0 || size_of_image as u64 > super::super::MAX_IMAGE_SIZE { return Err(UnpackError::InvalidImageSize { size: i64::from(size_of_image), - max: super::MAX_IMAGE_SIZE, + max: super::super::MAX_IMAGE_SIZE, }); } let mut out = vec![0u8; size_of_image as usize]; @@ -567,7 +567,7 @@ fn unpack_dll_inner(input: &[u8], verbose: bool) -> Result, UnpackError> let crc_val = { let a = crc_data_addr as usize; let n = crc_data_size as usize; - super::crc32::compute(&out[a..a + n]) as i32 + senbei_crypto::crc32::compute(&out[a..a + n]) as i32 }; let crc_xored = crc_data_size ^ crc_val; let trailing_val = get_i32(&out, crc_data_addr + crc_data_size - 4); diff --git a/senbei-pe/src/engine/error.rs b/senbei-pe/src/engine/error.rs new file mode 100644 index 0000000..3e7cfa6 --- /dev/null +++ b/senbei-pe/src/engine/error.rs @@ -0,0 +1,243 @@ +pub use senbei_crypto::{BufferOperation, DecompressionFailure}; + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum DecompressionStage { + ExeStage3, + ExeStage3Secondary, + ExeStage4, + ExeStage5, + Pe32FourthStage, + Pe32FifthStage, + Pe32SeventhStage, + DllCodeBlock1, + DllCodeBlock2, + DllCodeBlock3, + DllCodeBlock4, +} + +impl std::fmt::Display for DecompressionStage { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExeStage3 => "EXE stage3", + Self::ExeStage3Secondary => "EXE secondary stage3", + Self::ExeStage4 => "EXE stage4", + Self::ExeStage5 => "EXE stage5", + Self::Pe32FourthStage => "PE32 fourth stage", + Self::Pe32FifthStage => "PE32 fifth stage", + Self::Pe32SeventhStage => "PE32 seventh stage", + Self::DllCodeBlock1 => "DLL code block 1", + Self::DllCodeBlock2 => "DLL code block 2", + Self::DllCodeBlock3 => "DLL code block 3", + Self::DllCodeBlock4 => "DLL code block 4", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum BytecodeStage { + ExeStage4, + ExeStage5, + Pe32CustomDecryptor, + Pe32FileDecryptor, + DllPrimaryDecryptor, + DllSectionDecryptor, +} + +impl std::fmt::Display for BytecodeStage { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExeStage4 => "EXE stage4", + Self::ExeStage5 => "EXE stage5", + Self::Pe32CustomDecryptor => "PE32 custom decryptor", + Self::Pe32FileDecryptor => "PE32 file decryptor", + Self::DllPrimaryDecryptor => "DLL primary decryptor", + Self::DllSectionDecryptor => "DLL section decryptor", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum SectionPipeline { + ExePe32Plus, + ExePe32, + Dll, +} + +impl std::fmt::Display for SectionPipeline { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::ExePe32Plus => "PE32+ EXE", + Self::ExePe32 => "PE32 EXE", + Self::Dll => "DLL", + }) + } +} + +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +pub enum DescriptorTable { + DllSectionBlocks, + DllZeroFill, +} + +impl std::fmt::Display for DescriptorTable { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + f.write_str(match self { + Self::DllSectionBlocks => "DLL section-block", + Self::DllZeroFill => "DLL zero-fill", + }) + } +} + +#[derive(Debug, Clone, PartialEq, Eq, thiserror::Error)] +#[non_exhaustive] +pub enum UnpackError { + #[error("input too short (need at least {required} bytes, got {actual})")] + InputTooShort { actual: usize, required: usize }, + + #[error("decrypted header magic mismatch (got 0x{found:08X})")] + HeaderMagicMismatch { found: u32 }, + + #[error("anchor field not found — corrupt data or wrong offset")] + AnchorNotFound, + + #[error("stage1 descriptor not found near anchor 0x{anchor:08X}")] + Stage1DescriptorNotFound { anchor: u32 }, + + #[error("stage2 field not found — corrupt data or wrong offset")] + Stage2NotFound, + + #[error("chk_src_start not found — corrupt data or wrong offset")] + ChkSrcStartNotFound, + + #[error("table_start not found — corrupt data or wrong offset")] + TableStartNotFound, + + #[error("{0} bytecode generation failed — corrupt data or wrong offset")] + BytecodeGenerationFailed(BytecodeStage), + + #[error("stage5 marker not found — this build's layout is not supported by this unpacker")] + Stage5MarkerNotFound, + + #[error("not a Crackproof-protected file")] + NotCrackproof, + + #[error("invalid PE header offset {offset} for {input_len}-byte input")] + InvalidPeOffset { offset: i64, input_len: usize }, + + #[error("DLL pipeline requires PE32+ optional-header magic, got 0x{found:04X}")] + UnsupportedDllPeMagic { found: u16 }, + + #[error( + "DLL primary descriptor address 0x{address:08X} is below layout base 0x{minimum:08X} or outside {image_len}-byte image" + )] + InvalidDllPrimaryDescriptor { + address: u32, + minimum: u32, + image_len: usize, + }, + + #[error("invalid SizeOfImage {size}; expected 1..={max}")] + InvalidImageSize { size: i64, max: u64 }, + + #[error( + "{operation} range out of bounds (offset {offset}, size {size}, buffer length {buffer_len})" + )] + BufferRangeOutOfBounds { + operation: BufferOperation, + offset: usize, + size: usize, + buffer_len: usize, + }, + + #[error( + "{table} descriptor out of bounds (offset {offset}, size 16, image length {image_len})" + )] + DescriptorOutOfBounds { + table: DescriptorTable, + offset: usize, + image_len: usize, + }, + + #[error("PE32 tbl not found — corrupt data or wrong offset")] + Pe32TblNotFound, + + #[error("PE32 thirdStage decrypt failed — corrupt data or wrong offset")] + Pe32ThirdStageFailed, + + #[error("PE32 customDecryptor not found in sevenStage")] + Pe32CustomDecryptorNotFound, + + #[error("PE32 eighthStageKey not found")] + Pe32EighthKeyNotFound, + + #[error("PE32 file LFSR not found in eighthStage")] + Pe32FileLfsrNotFound, + + #[error("{stage} decompression failed: {reason}")] + StageDecompressionFailed { + stage: DecompressionStage, + reason: DecompressionFailure, + }, + + #[error("{pipeline} section block {block} decompression failed")] + SectionDecompressionFailed { + pipeline: SectionPipeline, + block: usize, + }, + + #[error("AES key schedule is outside the image at offset {offset}")] + InvalidAesKeySchedule { offset: u32 }, + + #[error("Huffman table is outside the image at offset {offset}")] + InvalidHuffmanTable { offset: u32 }, + + #[error("DLL pipeline failed: {dll}; EXE fallback failed: {exe}")] + PipelineFallbackFailed { + dll: Box, + exe: Box, + }, + + #[error( + "PE32 second-stage range is invalid (offset {offset}, size {size}, image length {image_len})" + )] + Pe32SecondStageRangeInvalid { + offset: u32, + size: u32, + image_len: usize, + }, + + #[error("PE32 relocation-data descriptor not found")] + Pe32RelocationDataNotFound, + + #[error("file decryptor candidate failed structural validation")] + FileDecryptorValidationFailed, + + #[error("PE32 memory image could not be rebuilt as a file-layout PE")] + Pe32OutputLayoutInvalid, + + #[error("internal panic at {file}:{line}:{column}: {message}")] + InternalPanic { + message: String, + file: String, + line: u32, + column: u32, + }, +} + +impl From for UnpackError { + fn from(error: senbei_crypto::Error) -> Self { + match error { + senbei_crypto::Error::BufferRangeOutOfBounds { + operation, + offset, + size, + buffer_len, + } => Self::BufferRangeOutOfBounds { + operation, + offset, + size, + buffer_len, + }, + } + } +} diff --git a/senbei-pe/src/engine/exe/mod.rs b/senbei-pe/src/engine/exe/mod.rs new file mode 100644 index 0000000..bf70ed1 --- /dev/null +++ b/senbei-pe/src/engine/exe/mod.rs @@ -0,0 +1,3 @@ +mod pipeline; + +pub use pipeline::*; diff --git a/src/unpacker/exe.rs b/senbei-pe/src/engine/exe/pipeline.rs similarity index 59% rename from src/unpacker/exe.rs rename to senbei-pe/src/engine/exe/pipeline.rs index 2dfd5ad..cac25da 100644 --- a/src/unpacker/exe.rs +++ b/senbei-pe/src/engine/exe/pipeline.rs @@ -1,285 +1,11 @@ -use super::bytecode::{Op, OpsLut, generate}; -use super::primitives; -use super::primitives::*; +use senbei_crypto::bytecode::{Op, OpsLut, generate}; +use senbei_crypto::primitives; +use senbei_crypto::primitives::*; -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum DecompressionStage { - ExeStage3, - ExeStage3Secondary, - ExeStage4, - ExeStage5, - Pe32FourthStage, - Pe32FifthStage, - Pe32SeventhStage, - DllCodeBlock1, - DllCodeBlock2, - DllCodeBlock3, - DllCodeBlock4, -} +use super::super::error::*; +use super::super::layout::{self, *}; -impl std::fmt::Display for DecompressionStage { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.write_str(match self { - Self::ExeStage3 => "EXE stage3", - Self::ExeStage3Secondary => "EXE secondary stage3", - Self::ExeStage4 => "EXE stage4", - Self::ExeStage5 => "EXE stage5", - Self::Pe32FourthStage => "PE32 fourth stage", - Self::Pe32FifthStage => "PE32 fifth stage", - Self::Pe32SeventhStage => "PE32 seventh stage", - Self::DllCodeBlock1 => "DLL code block 1", - Self::DllCodeBlock2 => "DLL code block 2", - Self::DllCodeBlock3 => "DLL code block 3", - Self::DllCodeBlock4 => "DLL code block 4", - }) - } -} - -#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)] -#[non_exhaustive] -pub enum DecompressionFailure { - #[error("compressed source size {size} exceeds limit {max}")] - SourceTooLarge { size: u32, max: u64 }, - #[error("Huffman code length {bits} is invalid")] - InvalidCodeLength { bits: u8 }, - #[error("Huffman tree traversal exceeded 64 levels")] - HuffmanTraversalLimit, - #[error("pending length accumulator overflowed at {pending}")] - PendingLengthOverflow { pending: u32 }, - #[error("output step {step} at byte {written} exceeds expected size {expected}")] - OutputOverflow { - written: u32, - step: u32, - expected: u32, - }, - #[error("run-fill width {width} reads before output offset 0x{destination:08X}")] - RunFillBeforeOutput { width: u32, destination: u32 }, - #[error("run-fill width {width} is unsupported")] - InvalidRunFillWidth { width: u32 }, - #[error("back-reference distance {distance} exceeds {written} written bytes")] - InvalidBackReference { distance: u32, written: u32 }, - #[error("Huffman symbol consumed no input and produced no output")] - NoProgress, - #[error( - "output size mismatch (wrote {written}/{expected} bytes after consuming {consumed}/{source_size})" - )] - OutputSizeMismatch { - written: u32, - expected: u32, - consumed: u32, - source_size: u32, - }, -} - -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum BytecodeStage { - ExeStage4, - ExeStage5, - Pe32CustomDecryptor, - Pe32FileDecryptor, - DllPrimaryDecryptor, - DllSectionDecryptor, -} - -impl std::fmt::Display for BytecodeStage { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.write_str(match self { - Self::ExeStage4 => "EXE stage4", - Self::ExeStage5 => "EXE stage5", - Self::Pe32CustomDecryptor => "PE32 custom decryptor", - Self::Pe32FileDecryptor => "PE32 file decryptor", - Self::DllPrimaryDecryptor => "DLL primary decryptor", - Self::DllSectionDecryptor => "DLL section decryptor", - }) - } -} - -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum SectionPipeline { - ExePe32Plus, - ExePe32, - Dll, -} - -impl std::fmt::Display for SectionPipeline { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.write_str(match self { - Self::ExePe32Plus => "PE32+ EXE", - Self::ExePe32 => "PE32 EXE", - Self::Dll => "DLL", - }) - } -} - -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum DescriptorTable { - DllSectionBlocks, - DllZeroFill, -} - -impl std::fmt::Display for DescriptorTable { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.write_str(match self { - Self::DllSectionBlocks => "DLL section-block", - Self::DllZeroFill => "DLL zero-fill", - }) - } -} - -#[derive(Debug, Clone, Copy, PartialEq, Eq)] -pub enum BufferOperation { - Read, - CopySource, - CopyDestination, - ZeroFill, -} - -impl std::fmt::Display for BufferOperation { - fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { - f.write_str(match self { - Self::Read => "read", - Self::CopySource => "copy source", - Self::CopyDestination => "copy destination", - Self::ZeroFill => "zero-fill", - }) - } -} - -#[derive(Debug, Clone, PartialEq, Eq, thiserror::Error)] -#[non_exhaustive] -pub enum UnpackError { - #[error("input too short (need at least {required} bytes, got {actual})")] - InputTooShort { actual: usize, required: usize }, - - #[error("decrypted header magic mismatch (got 0x{found:08X})")] - HeaderMagicMismatch { found: u32 }, - - #[error("anchor field not found — corrupt data or wrong offset")] - AnchorNotFound, - - #[error("stage1 descriptor not found near anchor 0x{anchor:08X}")] - Stage1DescriptorNotFound { anchor: u32 }, - - #[error("stage2 field not found — corrupt data or wrong offset")] - Stage2NotFound, - - #[error("chk_src_start not found — corrupt data or wrong offset")] - ChkSrcStartNotFound, - - #[error("table_start not found — corrupt data or wrong offset")] - TableStartNotFound, - - #[error("{0} bytecode generation failed — corrupt data or wrong offset")] - BytecodeGenerationFailed(BytecodeStage), - - #[error("stage5 marker not found — this build's layout is not supported by this unpacker")] - Stage5MarkerNotFound, - - #[error("not a Crackproof-protected file")] - NotCrackproof, - - #[error("invalid PE header offset {offset} for {input_len}-byte input")] - InvalidPeOffset { offset: i64, input_len: usize }, - - #[error("DLL pipeline requires PE32+ optional-header magic, got 0x{found:04X}")] - UnsupportedDllPeMagic { found: u16 }, - - #[error( - "DLL primary descriptor address 0x{address:08X} is below layout base 0x{minimum:08X} or outside {image_len}-byte image" - )] - InvalidDllPrimaryDescriptor { - address: u32, - minimum: u32, - image_len: usize, - }, - - #[error("invalid SizeOfImage {size}; expected 1..={max}")] - InvalidImageSize { size: i64, max: u64 }, - - #[error( - "{operation} range out of bounds (offset {offset}, size {size}, buffer length {buffer_len})" - )] - BufferRangeOutOfBounds { - operation: BufferOperation, - offset: usize, - size: usize, - buffer_len: usize, - }, - - #[error( - "{table} descriptor out of bounds (offset {offset}, size 16, image length {image_len})" - )] - DescriptorOutOfBounds { - table: DescriptorTable, - offset: usize, - image_len: usize, - }, - - #[error("PE32 tbl not found — corrupt data or wrong offset")] - Pe32TblNotFound, - - #[error("PE32 thirdStage decrypt failed — corrupt data or wrong offset")] - Pe32ThirdStageFailed, - - #[error("PE32 customDecryptor not found in sevenStage")] - Pe32CustomDecryptorNotFound, - - #[error("PE32 eighthStageKey not found")] - Pe32EighthKeyNotFound, - - #[error("PE32 file LFSR not found in eighthStage")] - Pe32FileLfsrNotFound, - - #[error("{stage} decompression failed: {reason}")] - StageDecompressionFailed { - stage: DecompressionStage, - reason: DecompressionFailure, - }, - - #[error("{pipeline} section block {block} decompression failed")] - SectionDecompressionFailed { - pipeline: SectionPipeline, - block: usize, - }, - - #[error("AES key schedule is outside the image at offset {offset}")] - InvalidAesKeySchedule { offset: u32 }, - - #[error("Huffman table is outside the image at offset {offset}")] - InvalidHuffmanTable { offset: u32 }, - - #[error("DLL pipeline failed: {dll}; EXE fallback failed: {exe}")] - PipelineFallbackFailed { - dll: Box, - exe: Box, - }, - - #[error( - "PE32 second-stage range is invalid (offset {offset}, size {size}, image length {image_len})" - )] - Pe32SecondStageRangeInvalid { - offset: u32, - size: u32, - image_len: usize, - }, - - #[error("PE32 relocation-data descriptor not found")] - Pe32RelocationDataNotFound, - - #[error("file decryptor candidate failed structural validation")] - FileDecryptorValidationFailed, - - #[error("PE32 memory image could not be rebuilt as a file-layout PE")] - Pe32OutputLayoutInvalid, - - #[error("internal panic at {file}:{line}:{column}: {message}")] - InternalPanic { - message: String, - file: String, - line: u32, - column: u32, - }, -} +mod pe32; pub fn unpack(input: &[u8]) -> Result, UnpackError> { unpack_v(input, false) @@ -324,7 +50,7 @@ pub fn unpack_v(input: &[u8], verbose: bool) -> Result, UnpackError> { // separate `decompressed` buffer), so the unpacker borrows it directly — no // owned copy is made here. catch_unwind uses AssertUnwindSafe, so a borrowing // (non-'static) closure is fine. - super::catch_unpack(move || Unpacker::run(input, verbose)) + super::super::catch_unpack(move || Unpacker::run(input, verbose)) } struct Unpacker<'a> { @@ -600,7 +326,7 @@ impl<'a> Unpacker<'a> { println!(" info[6] end_mark = 0x{:08X}", u.info[6]); println!(" info[7] = 0x{:08X}", u.info[7]); } - if !super::is_supported_magic(u.info[1]) { + if !super::super::is_supported_magic(u.info[1]) { return Err(UnpackError::HeaderMagicMismatch { found: u.info[1] }); } @@ -615,10 +341,10 @@ impl<'a> Unpacker<'a> { }); } let size_of_image = get_u32(u.file_data, pe_off.wrapping_add(80)); - if size_of_image == 0 || size_of_image as u64 > super::MAX_IMAGE_SIZE { + if size_of_image == 0 || size_of_image as u64 > super::super::MAX_IMAGE_SIZE { return Err(UnpackError::InvalidImageSize { size: i64::from(size_of_image), - max: super::MAX_IMAGE_SIZE, + max: super::super::MAX_IMAGE_SIZE, }); } u.decompressed = vec![0u8; size_of_image as usize]; @@ -1210,7 +936,7 @@ impl<'a> Unpacker<'a> { // structurally. fileCS stays at bc2_off-0x58 as in the old layout. if new_layout { let compress_data_offset = (!get_u32(u.file_data, 0x1080)).wrapping_add(0x1000); - let slots = primitives::discover_eighth_slots( + let slots = layout::discover_eighth_slots( &u.decompressed, stage5_field, stage5_dlen, @@ -1386,7 +1112,7 @@ impl<'a> Unpacker<'a> { } Ok(()) }; - super::parallel::parallel_for(&mut u.decompressed, &spans, 1, do_block)?; + super::super::parallel::parallel_for(&mut u.decompressed, &spans, 1, do_block)?; } loop { u.decrypt_data5(walk4, 16); @@ -1578,9 +1304,7 @@ impl<'a> Unpacker<'a> { let dd8_shift: u32 = match std::env::var("DD8_SHIFT").ok().and_then(|s| s.parse().ok()) { Some(s) => s, - None => { - primitives::select_dd8_shift(&u.decompressed, text_va, text_size, u.info[3]) - } + None => layout::select_dd8_shift(&u.decompressed, text_va, text_size, u.info[3]), }; if dd8_shift != 99 { let mut page = text_va >> 12; @@ -1700,7 +1424,7 @@ impl<'a> Unpacker<'a> { let shift = match std::env::var("DD8_SHIFT").ok().and_then(|s| s.parse().ok()) { Some(s) => s, None => { - primitives::select_dd8_shift(&u.decompressed, text_va, text_size, u.info[3]) + layout::select_dd8_shift(&u.decompressed, text_va, text_size, u.info[3]) } }; if shift != 99 { @@ -1888,7 +1612,8 @@ impl<'a> Unpacker<'a> { self.aes_decrypt(dst, len, self.key_offsets[2]); for k in 0..len { let idx = (dst + k) as usize; - self.decompressed[idx] = super::bytecode::apply(ops, self.decompressed[idx]); + self.decompressed[idx] = + senbei_crypto::bytecode::apply(ops, self.decompressed[idx]); } let ok = primitives::decompress( &mut self.decompressed, @@ -1980,7 +1705,8 @@ impl<'a> Unpacker<'a> { self.aes_decrypt(dst2, s_sz2, self.key_offsets[2]); for k in 0..s_sz2 { let idx = (dst2 + k) as usize; - self.decompressed[idx] = super::bytecode::apply(&ops, self.decompressed[idx]); + self.decompressed[idx] = + senbei_crypto::bytecode::apply(&ops, self.decompressed[idx]); } let ok = primitives::decompress( &mut self.decompressed, @@ -2141,1064 +1867,6 @@ impl<'a> Unpacker<'a> { self.decompressed[d..d + 24].copy_from_slice(&self.file_data[src..src + 24]); true } - - /// PE32 (32-bit) unpack pipeline. The shared Stage 1/2 setup (info decrypt, - /// payload decrypt, raw copy, header restore) has already run in `run()` - /// before dispatch; this takes over from "Locating shell offsets". - fn run_pe32(&mut self, pe_off: u32, verbose: bool) -> Result, UnpackError> { - let info = self.info; - let info3 = info[3]; - - // advance_key: replays the packer's per-iteration key walk. - let advance_key = |mut key: u32, iterations: u32| -> u32 { - for m in 0..iterations { - let bound = (m + 1).wrapping_mul(25) << 2; - let mut n: u32 = 1; - while n <= bound { - key = key.wrapping_add(n); - n += 1; - } - } - key - }; - - // ---- Locate tbl in shell ---- - let tbl = primitives::find_tbl_pe32(&self.decompressed, &info) - .ok_or(UnpackError::Pe32TblNotFound)?; - if verbose { - println!("[3/9] Locating config layout (PE32)..."); - println!(" tbl = 0x{:X}", tbl); - } - - // ---- PE header restore ---- - let val_bc = get_u32(&self.decompressed, tbl.wrapping_add(0xBC)); - let val_c8 = get_u32(&self.decompressed, tbl.wrapping_add(0xC8)); - let val_cc = get_u32(&self.decompressed, tbl.wrapping_add(0xCC)); - write_u32(&mut self.decompressed, pe_off.wrapping_add(0x80), val_bc); - write_u32(&mut self.decompressed, pe_off.wrapping_add(0x88), val_c8); - write_u32(&mut self.decompressed, pe_off.wrapping_add(0x8C), val_cc); - write_u32(&mut self.decompressed, pe_off.wrapping_add(0xB0), 0); - write_u32(&mut self.decompressed, pe_off.wrapping_add(0xB4), 0); - - // ---- Header-independent checksum inputs ---- - let first_stage_cs = self.calculate_checksum(tbl.wrapping_add(0xA8)); - let second_stage_key = get_u32(&self.decompressed, tbl.wrapping_add(0x40)); - - // ---- Stage 3: SecondStage ---- - // - // ss_key = headerChecksum ^ firstStageCS ^ secondStageKey, where the - // header checksum (a XOR of crc32(region)^size over the sub-regions at - // tbl+0x58) is taken over the *original* pre-pack PE header. For EXEs the - // import/resource restore above reconstructs that header exactly. Native - // DLLs additionally carry a packer-added BaseReloc data-directory entry - // (dir 5) that was absent from the checksummed original, so the header - // checksum only matches once that entry is treated as zero. EXEs have no - // dir-5 entry, so zeroing it is a no-op for them. - // - // Rather than branch on EXE-vs-DLL, try the header as-is and, on failure, - // with the BaseReloc entry zeroed; keep whichever ss_key decrypts a - // SecondStage whose ThirdStage (off,size) pair lands inside the image. - // This uses the same shift/key trial-and-validate the later stages - // already use, and keeps EXE output byte-identical (the as-is variant - // wins first). - let ss_pair = tbl.wrapping_add(0x98); - let ss = get_u32(&self.decompressed, ss_pair); - let ss_size = get_u32(&self.decompressed, ss_pair.wrapping_add(4)); - let ss_shift = ss_size.wrapping_sub(0xBC0); - // Back up the SecondStage ciphertext so a failed trial can be retried. - let ss_lo = ss as usize; - let ss_hi = ss_lo.wrapping_add(ss_size as usize); - if ss_hi < ss_lo || ss_hi > self.decompressed.len() { - return Err(UnpackError::Pe32SecondStageRangeInvalid { - offset: ss, - size: ss_size, - image_len: self.decompressed.len(), - }); - } - let ss_ct: Vec = self.decompressed[ss_lo..ss_hi].to_vec(); - // PE32 data dir 5 (BaseReloc) = optional_header(pe+24) + 0x60 + 5*8 = pe+0xA0. - let reloc_dir = pe_off.wrapping_add(0xA0); - let len = self.decompressed.len() as u64; - let pair_off = 0xB8Cu32.wrapping_add(ss_shift); - let mut found = false; - // Holds the winning variant's header checksum; the later stages - // (Forth/Fifth/Seven/Eighth) reuse it as a key component. - let mut header_checksum: u32 = 0; - for zero_reloc in [false, true] { - if zero_reloc { - write_u32(&mut self.decompressed, reloc_dir, 0); - write_u32(&mut self.decompressed, reloc_dir.wrapping_add(4), 0); - } - let mut hcs_addr = tbl.wrapping_add(0x58); - header_checksum = 0; - while get_u32(&self.decompressed, hcs_addr.wrapping_add(4)) != 0 { - header_checksum ^= self.calculate_checksum(hcs_addr); - hcs_addr = hcs_addr.wrapping_add(8); - } - let ss_key = header_checksum ^ first_stage_cs ^ second_stage_key; - self.decompressed[ss_lo..ss_hi].copy_from_slice(&ss_ct); - self.decrypt_data3(ss_pair, ss_key, 21); - // Validate: the ThirdStage (off,size) pair must reference the image. - let pair = ss.wrapping_add(pair_off); - let off = get_u32(&self.decompressed, pair) as u64; - let sz = get_u32(&self.decompressed, pair.wrapping_add(4)) as u64; - if off > 0x1000 && off < len && sz >= 4 && off.saturating_add(sz) <= len { - found = true; - break; - } - } - if !found { - return Err(UnpackError::Pe32RelocationDataNotFound); - } - if verbose { - println!( - " ss = 0x{:08X}, size = 0x{:X}, shift = 0x{:X}", - ss, ss_size, ss_shift - ); - } - - // ---- PE32 fixed offsets ---- - let third_key_off = 0x968u32.wrapping_add(ss_shift); - let forth_key_off = 0x964u32.wrapping_add(ss_shift); - let cs_base_off = 0x96Cu32.wrapping_add(ss_shift); - let dp_base_off = 0xA9Cu32.wrapping_add(ss_shift); - - // ---- Stage 4: ThirdStage (brute-force the rotate shift) ---- - let third_pair_off = 0xB8Cu32.wrapping_add(ss_shift); - let key = get_u32(&self.decompressed, ss.wrapping_add(third_key_off)); - let pair_addr = ss.wrapping_add(third_pair_off); - let ts_addr = get_u32(&self.decompressed, pair_addr); - let ts_size_raw = get_u32(&self.decompressed, pair_addr.wrapping_add(4)); - let backup: Vec = - self.decompressed[ts_addr as usize..(ts_addr + ts_size_raw) as usize].to_vec(); - let mut info_table: Option = None; - let mut keys_addr: u32 = 0; - let mut ts: u32 = 0; - for &shift in &[19u32, 21, 17, 23, 15, 25, 13, 11] { - self.decompressed[ts_addr as usize..(ts_addr + ts_size_raw) as usize] - .copy_from_slice(&backup); - write_u32(&mut self.decompressed, pair_addr, ts_addr); - write_u32( - &mut self.decompressed, - pair_addr.wrapping_add(4), - ts_size_raw, - ); - self.decrypt_data3(pair_addr, key, shift); - let mut off = 0u32; - while off + 32 < ts_size_raw { - let t0 = get_u32(&self.decompressed, ts_addr.wrapping_add(off)); - if t0 == 1 || t0 == 0x11 { - let t1 = get_u32(&self.decompressed, ts_addr.wrapping_add(off + 16)); - if t1 == 2 { - let addr0 = get_u32(&self.decompressed, ts_addr.wrapping_add(off + 4)); - if 0x1000 < addr0 && (addr0 as usize) < self.decompressed.len() { - let it = ts_addr.wrapping_add(off); - info_table = Some(it); - keys_addr = it.wrapping_sub(0x58); - ts = ts_addr; - break; - } - } - } - off = off.wrapping_add(4); - } - if info_table.is_some() { - break; - } - } - let info_table = info_table.ok_or(UnpackError::Pe32ThirdStageFailed)?; - if verbose { - println!("[4/9] Decrypting stages (PE32)..."); - println!( - " thirdStage start = 0x{:X}, infoTable = 0x{:X}", - ts, info_table - ); - } - - // ---- Process infoTable ---- - let mut it_addr = info_table; - for _ in 0..2 { - let tval = get_u32(&self.decompressed, it_addr); - if tval == 1 || tval == 0x11 { - self.decrypt_data4(it_addr.wrapping_add(4)); - } else if tval == 2 { - let mut copy_addr = get_u32(&self.decompressed, it_addr.wrapping_add(4)); - loop { - self.decrypt_data5(copy_addr, 16); - let s_a = get_u32(&self.decompressed, copy_addr); - let s_sz = get_u32(&self.decompressed, copy_addr.wrapping_add(4)); - let d_a = get_u32(&self.decompressed, copy_addr.wrapping_add(8)); - let d_sz = get_u32(&self.decompressed, copy_addr.wrapping_add(12)); - copy_addr = copy_addr.wrapping_add(16); - if s_sz == 0 { - break; - } - if s_a != 0 && d_a != 0 && d_sz == s_sz { - let sa = s_a as usize; - let da = d_a as usize; - let n = s_sz as usize; - self.decompressed.copy_within(sa..sa + n, da); - } - } - } - it_addr = it_addr.wrapping_add(16); - } - - // ---- keyOffsets ---- - let mut ka = keys_addr; - for k in 0..2usize { - let mut ka2 = ka; - for l in 0..2usize { - self.decrypt_data4(ka2); - self.key_offsets[k * 2 + l] = get_u32(&self.decompressed, ka2); - ka2 = ka2.wrapping_add(8); - } - ka = ka.wrapping_add(32); - } - - // ---- Checksum addresses ---- - let second_stage_cs_addr = tbl.wrapping_add(0xB0); - let forth_stage_cs_addr = ss.wrapping_add(cs_base_off); - let fifth_stage_cs_addr = ss.wrapping_add(cs_base_off).wrapping_add(0x08); - let seven_stage_cs_addr = ss.wrapping_add(cs_base_off).wrapping_add(0x10); - - // ---- ForthStage ---- - let second_stage_cs = self.calculate_checksum(second_stage_cs_addr); - let forth_stage_key = advance_key( - get_u32(&self.decompressed, ss.wrapping_add(forth_key_off)), - 4, - ); - let dp_base = ss.wrapping_add(dp_base_off); - let forth_addr = dp_base.wrapping_add(0x40); - let fk = header_checksum ^ second_stage_cs ^ forth_stage_key; - if let Err(reason) = self.decrypt_and_decompress_data(forth_addr, fk, None) { - return Err(UnpackError::StageDecompressionFailed { - stage: DecompressionStage::Pe32FourthStage, - reason, - }); - } - - // ---- FifthStage ---- - let fifth_addr = dp_base.wrapping_add(0x50); - let forth_cs = self.calculate_checksum(forth_stage_cs_addr); - let forth_region_off = get_u32(&self.decompressed, forth_stage_cs_addr); - let forth_region_sz = get_u32(&self.decompressed, forth_stage_cs_addr.wrapping_add(4)); - let fifth_key = get_u32( - &self.decompressed, - forth_region_off - .wrapping_add(forth_region_sz) - .wrapping_sub(4), - ); - let fk5 = header_checksum ^ forth_cs ^ fifth_key; - if let Err(reason) = self.decrypt_and_decompress_data(fifth_addr, fk5, None) { - return Err(UnpackError::StageDecompressionFailed { - stage: DecompressionStage::Pe32FifthStage, - reason, - }); - } - - // ---- SevenStage ---- - let seven_addr = dp_base.wrapping_add(0x70); - let seven_dsz = get_u32(&self.decompressed, seven_addr.wrapping_add(12)); - let fifth_cs = self.calculate_checksum(fifth_stage_cs_addr); - let cs1_addr = get_u32( - &self.decompressed, - ss.wrapping_add(cs_base_off).wrapping_add(0x08), - ); - let cs1_size = get_u32( - &self.decompressed, - ss.wrapping_add(cs_base_off) - .wrapping_add(0x08) - .wrapping_add(4), - ); - let seven_key = !get_u32( - &self.decompressed, - cs1_addr.wrapping_add(cs1_size).wrapping_sub(0x10), - ); - let fk7 = header_checksum ^ fifth_cs ^ seven_key; - if let Err(reason) = self.decrypt_and_decompress_data(seven_addr, fk7, None) { - return Err(UnpackError::StageDecompressionFailed { - stage: DecompressionStage::Pe32SeventhStage, - reason, - }); - } - - // ---- EighthStage ---- - let seven_start_actual = get_u32(&self.decompressed, seven_addr); - if verbose { - println!("[5/9] Decrypting eighthStage (PE32)..."); - println!( - " sevenStart = 0x{:X}, sevenDsz = 0x{:X}", - seven_start_actual, seven_dsz - ); - } - // Locate the customDecryptor LFSR block (scan backward from middle, then - // forward as fallback). - let scan_start = seven_dsz / 2; - let custom_dec_off = primitives::find_lfsr_block( - &self.decompressed, - seven_start_actual, - seven_dsz, - scan_start, - true, - ) - .or_else(|| { - primitives::find_lfsr_block(&self.decompressed, seven_start_actual, seven_dsz, 0, false) - }) - .ok_or(UnpackError::Pe32CustomDecryptorNotFound)?; - let custom_dec_addr = seven_start_actual.wrapping_add(custom_dec_off); - self.decrypt_data6(custom_dec_addr); - let custom_ops = generate(&self.decompressed, custom_dec_addr).ok_or( - UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32CustomDecryptor), - )?; - - let seven_cs = self.calculate_checksum(seven_stage_cs_addr); - let eighth_addr = dp_base.wrapping_add(0xC0); - let eighth_dsz = get_u32(&self.decompressed, eighth_addr.wrapping_add(12)); - let eighth_src = get_u32(&self.decompressed, eighth_addr); - let eighth_ssz = get_u32(&self.decompressed, eighth_addr.wrapping_add(4)); - let eighth_backup: Vec = - self.decompressed[eighth_src as usize..(eighth_src + eighth_ssz) as usize].to_vec(); - let eighth_pair_bak: Vec = - self.decompressed[eighth_addr as usize..(eighth_addr + 16) as usize].to_vec(); - let data_len = self.decompressed.len() as u32; - - // Build the eighthStageKey candidate list (offsets relative to - // sevenStart) using gap heuristics + scan. - let mut candidates: Vec = Vec::new(); - let push_cand = |c: &mut Vec, off: u32| { - if !c.contains(&off) { - c.push(off); - } - }; - for &end_gap in &[0xD0u32, 0xC0, 0xE0, 0xB0, 0xA0, 0xF0, 0x100] { - if end_gap <= seven_dsz { - let off = seven_dsz - end_gap; - if off < seven_dsz { - let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); - if val != 0 && val != 0xCCCC_CCCC { - push_cand(&mut candidates, off); - } - } - } - } - for &gap in &[ - 0x70u32, 0xD0, 0x28, 0x50, 0x48, 0x30, 0x40, 0x58, 0x60, 0x20, 0x38, 0x80, 0x90, 0xA0, - 0xB0, - ] { - if gap <= custom_dec_off { - let off = custom_dec_off - gap; - if off + 4 <= seven_dsz && !candidates.contains(&off) { - let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); - if val != 0 && val != 0xCCCC_CCCC { - push_cand(&mut candidates, off); - } - } - } - } - let scan_lo = custom_dec_off.saturating_sub(0x100); - let mut off = scan_lo; - while off < custom_dec_off { - if !candidates.contains(&off) { - let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); - let all_printable = (0..4u32).all(|i| { - let b = (val >> (i * 8)) & 0xFF; - (32..127).contains(&b) - }); - if val != 0 && val != 0xCCCC_CCCC && !all_printable { - push_cand(&mut candidates, off); - } - } - off = off.wrapping_add(4); - } - - let k1 = self.key_offsets[1]; - let k3 = self.key_offsets[3]; - let mut eighth_ok = false; - for ek_off in candidates { - self.decompressed[eighth_src as usize..(eighth_src + eighth_ssz) as usize] - .copy_from_slice(&eighth_backup); - self.decompressed[eighth_addr as usize..(eighth_addr + 16) as usize] - .copy_from_slice(&eighth_pair_bak); - let raw = get_u32(&self.decompressed, seven_start_actual.wrapping_add(ek_off)); - let test_key = advance_key(raw, 3); - let fk8 = header_checksum ^ fifth_cs ^ seven_cs ^ test_key; - let result = primitives::decrypt_and_decompress_data( - &mut self.decompressed, - eighth_addr, - fk8, - k1, - k3, - Some(&custom_ops), - ); - if result { - let est = get_u32(&self.decompressed, eighth_addr); - if 0x1000 < est && est < data_len { - eighth_ok = true; - break; - } - } - } - if !eighth_ok { - return Err(UnpackError::Pe32EighthKeyNotFound); - } - let eighth_start = get_u32(&self.decompressed, eighth_addr); - if verbose { - println!( - " eighthStart = 0x{:08X}, dsz = 0x{:X}", - eighth_start, eighth_dsz - ); - } - - // ---- Final processing offsets (anchored on the eighthStage config cluster) ---- - // - // The eighthStage holds a config cluster — importTable, fileCS, - // compressedInfo, zeroList — at fixed offsets from a cluster base - // (base+0x18 / +0x30 / +0x40 / +0x48) with the fileLFSR at +0x4B4. - // Classic builds stamp a 0x00007679 dword at that base; native DLLs and - // some older PE32 EXEs (ss_size=0xBE8) omit the stamp. - // Locate the cluster by stamp when present (validated by fileCS at - // base+0x30 pointing past info[3]); otherwise fall back to finding the - // fileCS slot by shape — (addr, size) with addr just past info[3] and a - // small 16-aligned size — and back-derive base = fileCS_off - 0x30. - // Hardcoded eighthStart-relative constants remain as a last-resort - // fallback for builds where neither discovery path fires. - let marker = { - let mut m: Option = None; - let hi = eighth_dsz.saturating_sub(0x4C); - let mut o = 0u32; - while o < hi { - if get_u32(&self.decompressed, eighth_start.wrapping_add(o)) == 0x7679 { - let fc = get_u32(&self.decompressed, eighth_start.wrapping_add(o + 0x30)); - if fc > info3 && (fc as usize) < self.decompressed.len() { - m = Some(o); - break; - } - } - o = o.wrapping_add(4); - } - if m.is_none() { - // fileCS-shaped slot: addr in (info3, info3+0x2000], size in - // 0x10..=0x200 and 16-aligned. Prefer the candidate whose addr - // is closest to (but past) info3 — matches every observed - // build (one PE32 EXE family dist ~0x1C0, another ~0x1A0). - let mut best: Option<(u32 /*dist*/, u32 /*off*/)> = None; - let mut o = 0u32; - let dlen = self.decompressed.len() as u32; - while o + 8 <= eighth_dsz.saturating_sub(0x4B4u32.saturating_sub(0x30)) { - let fc = get_u32(&self.decompressed, eighth_start.wrapping_add(o)); - let sz = get_u32(&self.decompressed, eighth_start.wrapping_add(o + 4)); - if fc > info3 - && fc <= info3.wrapping_add(0x2000) - && fc < dlen - && (0x10..=0x200).contains(&sz) - && (sz & 0xF) == 0 - { - // Cluster base must leave room for the +0x4B4 LFSR slot - // (even if the exact LFSR is later adjusted by scan). - if o >= 0x30 { - let base = o - 0x30; - if base.wrapping_add(0x4C) <= eighth_dsz { - let dist = fc - info3; - match best { - None => best = Some((dist, base)), - Some((bd, _)) if dist < bd => best = Some((dist, base)), - _ => {} - } - } - } - } - o = o.wrapping_add(4); - } - if let Some((dist, base)) = best { - if verbose { - println!( - " pe32 cluster via fileCS (no 0x7679): base=+0x{:X} dist_info3=0x{:X}", - base, dist - ); - } - m = Some(base); - } - } - m - }; - let (off_import_table, off_file_cs, off_compressed_info, off_zero_list, off_file_lfsr) = - match marker { - Some(m) => (m + 0x18, m + 0x30, m + 0x40, m + 0x48, m + 0x4B4), - None => ( - 0x3C50u32.wrapping_add(ss_shift), - 0x3C68u32.wrapping_add(ss_shift), - 0x3C78u32.wrapping_add(ss_shift), - 0x3C80u32.wrapping_add(ss_shift), - 0x40ECu32.wrapping_add(ss_shift), - ), - }; - - // ---- File checksums (permanent decrypt) ---- - let file_cs_addr_ptr = eighth_start.wrapping_add(off_file_cs); - let mut file_cs_addr = get_u32(&self.decompressed, file_cs_addr_ptr); - let file_cs_size = get_u32(&self.decompressed, file_cs_addr_ptr.wrapping_add(4)); - if file_cs_size > 0 { - let file_cs_end = file_cs_addr.wrapping_add(file_cs_size); - while file_cs_addr < file_cs_end { - self.decrypt_data5(file_cs_addr, 16); - file_cs_addr = file_cs_addr.wrapping_add(16); - } - } else { - while get_u32(&self.decompressed, file_cs_addr.wrapping_add(4)) != 0 { - self.decrypt_data5(file_cs_addr, 16); - file_cs_addr = file_cs_addr.wrapping_add(16); - } - } - - // ---- File decryptor LFSR ---- - // - // When the marker-relative off_file_lfsr is in range, try that slot - // first (exact). If it is not a valid LFSR block, trial-and-validate - // candidates from off_zero_list forward — required for older PE32 EXEs - // without the 0x7679 stamp where the expected slot is empty and a loose - // decoded[0]+0xC3 nearest-hit picks the wrong decryptor. Fall back to - // the legacy loose scan only if no candidate trial-decompresses. Native - // DLLs have a smaller eighthStage where off_file_lfsr lands out of range - // and use the same trial-validate scan from just past the cluster (else - // branch). - let lfsr_off = if off_file_lfsr.wrapping_add(96) <= eighth_dsz { - let mut lfsr_off = off_file_lfsr; - let exact = primitives::find_lfsr_block( - &self.decompressed, - eighth_start, - eighth_dsz, - off_file_lfsr, - false, - ); - if exact != Some(off_file_lfsr) { - // Prefer trial-and-validate (same as the DLL branch): a loose - // decoded[0]+0xC3 scan can land on coincidental LFSR-shaped - // blocks that decode to a wrong file_ops and scramble every - // compressed block. Observed on older PE32 EXEs without the - // 0x7679 cluster stamp: the expected slot is empty and the - // nearest loose hit is not the real decryptor. - let ci_slot = eighth_start.wrapping_add(off_compressed_info); - let mut scan = off_zero_list; - let mut chosen: Option = None; - let mut considered = 0u32; - while let Some(cand) = primitives::find_lfsr_block( - &self.decompressed, - eighth_start, - eighth_dsz, - scan, - false, - ) { - considered = considered.wrapping_add(1); - if self.pe32_file_lfsr_validates(eighth_start.wrapping_add(cand), ci_slot) { - chosen = Some(cand); - break; - } - scan = cand + 1; - } - if let Some(c) = chosen { - if verbose { - println!( - " pe32 fileLFSR via trial-validate: +0x{:X} (expected +0x{:X}, considered {})", - c, off_file_lfsr, considered - ); - } - lfsr_off = c; - } else { - // No candidate trial-decompresses: fail loudly. The old - // "legacy loose scan" picked the nearest LFSR-shaped block - // by offset distance without any validation — that is - // exactly how a wrong file_ops got applied to every data - // block (uncompressed blocks never enter the decompressor), - // producing a plausible but fully wrong image (the PE32 - // .text scramble root cause). Trial-and-validate or error. - return Err(UnpackError::Pe32FileLfsrNotFound); - } - } - lfsr_off - } else { - // Native DLL: the marker-relative off_file_lfsr (EXE-tuned, marker + - // 0x4B4) overshoots the smaller DLL eighthStage, so the exact slot is - // unavailable. A plain forward scan returns the FIRST valid-opcode - // block, but the DLL eighthStage contains coincidental valid-opcode - // blocks that decode to trivial programs (e.g. a constant byte add) - // ahead of the real file decryptor. A wrong file_ops corrupts the - // per-block translate (applied before decompression), so every data - // block fails to decompress. Enumerate every candidate forward and - // keep the first whose decoded file_ops actually decompresses the - // first compressed data block — trial-and-validate, same idea as the - // D1/D2 fixes. Non-DLL (EXE) builds never reach this branch. - let ci_slot = eighth_start.wrapping_add(off_compressed_info); - let mut scan = off_zero_list.wrapping_add(8); - let mut chosen: Option = None; - while let Some(cand) = primitives::find_lfsr_block( - &self.decompressed, - eighth_start, - eighth_dsz, - scan, - false, - ) { - if self.pe32_file_lfsr_validates(eighth_start.wrapping_add(cand), ci_slot) { - chosen = Some(cand); - break; - } - scan = cand + 1; - } - chosen.ok_or(UnpackError::Pe32FileLfsrNotFound)? - }; - let file_dec_addr = eighth_start.wrapping_add(lfsr_off); - self.decrypt_data6(file_dec_addr); - let file_ops = generate(&self.decompressed, file_dec_addr).ok_or( - UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32FileDecryptor), - )?; - - // ---- PE32 metadata: EP and data dirs from info[3] ---- - let test_val = get_u32(&self.decompressed, info3.wrapping_add(0x10)); - let metadata_ep: u32; - let mut metadata_dirs = [0u8; 128]; - if test_val > 0x10000 { - // Layout B - let s = info3.wrapping_add(0x10) as usize; - let backup_meta = self.decompressed[s..s + 0x290].to_vec(); - self.decrypt_data5(info3.wrapping_add(0x10), 0x290); - metadata_ep = get_u32(&self.decompressed, info3.wrapping_add(0x20)); - let d = info3.wrapping_add(0x30) as usize; - metadata_dirs.copy_from_slice(&self.decompressed[d..d + 128]); - self.decompressed[s..s + 0x290].copy_from_slice(&backup_meta); - } else { - // Layout A - let s = info3.wrapping_add(0x40) as usize; - let backup_meta = self.decompressed[s..s + 144].to_vec(); - self.decrypt_data5(info3.wrapping_add(0x40), 144); - metadata_ep = get_u32(&self.decompressed, info3.wrapping_add(0x40)); - let d = info3.wrapping_add(0x50) as usize; - metadata_dirs.copy_from_slice(&self.decompressed[d..d + 128]); - self.decompressed[s..s + 144].copy_from_slice(&backup_meta); - } - - // ---- Zero-out list (runs BEFORE decompression) ---- - let zero_list_addr = eighth_start.wrapping_add(off_zero_list); - let mut zero_ptr = get_u32(&self.decompressed, zero_list_addr); - loop { - self.decrypt_data5(zero_ptr, 16); - let src3 = get_u32(&self.decompressed, zero_ptr); - let s_sz3 = get_u32(&self.decompressed, zero_ptr.wrapping_add(4)); - zero_ptr = zero_ptr.wrapping_add(16); - if s_sz3 == 0 { - break; - } - if src3.wrapping_add(s_sz3) as usize > self.decompressed.len() { - break; - } - for b in &mut self.decompressed[src3 as usize..(src3 + s_sz3) as usize] { - *b = 0; - } - } - - // ---- File data decompression ---- - if verbose { - println!("[6/9] Loading and decompressing file data (PE32)..."); - } - let compress_data_offset = (!get_u32(self.file_data, 0x1080)).wrapping_add(0x1000); - let compressed_info_addr = eighth_start.wrapping_add(off_compressed_info); - let mut compressed_info = get_u32(&self.decompressed, compressed_info_addr); - // Pass 1 (sequential): position-keyed descriptor chain (decrypt_data5), - // terminated by a zero source-size record. - struct Blk { - src: u32, - ssz: u32, - dst: u32, - dsz: u32, - } - let mut blocks: Vec = Vec::new(); - loop { - self.decrypt_data5(compressed_info, 16); - let src2 = get_u32(&self.decompressed, compressed_info); - let s_sz2 = get_u32(&self.decompressed, compressed_info.wrapping_add(4)); - let dst2 = get_u32(&self.decompressed, compressed_info.wrapping_add(8)); - let d_sz2 = get_u32(&self.decompressed, compressed_info.wrapping_add(12)); - compressed_info = compressed_info.wrapping_add(16); - if s_sz2 == 0 { - break; - } - blocks.push(Blk { - src: src2, - ssz: s_sz2, - dst: dst2, - dsz: d_sz2, - }); - } - // Pass 2: independent per-block work over disjoint dst spans (see - // `parallel_for` for how the spans are carved safely). - { - let lut = OpsLut::new(&file_ops); - let clean = &self.file_data; - let ko = self.key_offsets; - let ks_snap = primitives::aes_schedule_snapshot(&self.decompressed, ko[2]) - .ok_or(UnpackError::InvalidAesKeySchedule { offset: ko[2] })?; - let tab_snap = primitives::huffman_table_snapshot(&self.decompressed, ko[0]) - .ok_or(UnpackError::InvalidHuffmanTable { offset: ko[0] })?; - let spans: Vec<(usize, usize)> = blocks - .iter() - .map(|b| { - let s = b.dst as usize; - (s, s + b.ssz.max(b.dsz) as usize) - }) - .collect(); - let do_block = |i: usize, base: usize, span: &mut [u8]| -> Result<(), UnpackError> { - let b = &blocks[i]; - let file_src = b.src.wrapping_add(compress_data_offset) as usize; - let rel = b.dst as usize - base; - let n = b.ssz as usize; - span[rel..rel + n].copy_from_slice(&clean[file_src..file_src + n]); - primitives::aes_decrypt_ks(&ks_snap, span, rel as u32, b.ssz); - lut.map_region(span, rel, n); - if b.ssz != b.dsz { - // decompress reports corruption (after partial writes) via - // its bool; surface it instead of shipping a garbage block. - if !primitives::decompress_tbl( - &tab_snap, span, rel as u32, rel as u32, b.ssz, b.dsz, - ) { - return Err(UnpackError::SectionDecompressionFailed { - pipeline: SectionPipeline::ExePe32, - block: i, - }); - } - } - Ok(()) - }; - super::parallel::parallel_for(&mut self.decompressed, &spans, 1, do_block)?; - } - - // ---- Section fixup ---- - self.decompressed[..0x1000].copy_from_slice(&self.file_data[..0x1000]); - let opt_hdr_size = get_u16(self.file_data, pe_off.wrapping_add(20)) as u32; - let sec_hdr_table = pe_off.wrapping_add(24).wrapping_add(opt_hdr_size); - let export_va = get_u32( - self.file_data, - pe_off - .wrapping_add(24) - .wrapping_add(opt_hdr_size) - .wrapping_sub(128), - ); - let export_size = get_u32( - self.file_data, - pe_off - .wrapping_add(24) - .wrapping_add(opt_hdr_size) - .wrapping_sub(124), - ); - let mut export_file_off: u32 = 0; - let mut text_off: u32 = 0; - let mut text_size: u32 = 0; - // Walk by NumberOfSections (PE has no zero-VS sentinel; a real - // VirtualSize==0 section would truncate these fixups early), stopping - // at the all-zero padding in case NumberOfSections is overstated. - let num_sections = get_u16(self.file_data, pe_off.wrapping_add(6)) as u32; - for i in 0..num_sections.min(96) { - let sec_hdr = sec_hdr_table.wrapping_add(i.wrapping_mul(40)); - if self.file_data[sec_hdr as usize..sec_hdr as usize + 8] - .iter() - .all(|&b| b == 0) - { - break; - } - let va = get_u32(self.file_data, sec_hdr.wrapping_add(12)); - let sz = get_u32(self.file_data, sec_hdr.wrapping_add(8)); - let f_off = get_u32(self.file_data, sec_hdr.wrapping_add(20)); - let name = section_name(self.file_data, sec_hdr); - if name.starts_with(".text") { - text_size = sz; - text_off = va; - } - if export_size != 0 - && export_va >= va - && export_va.wrapping_add(export_size) <= va.wrapping_add(sz) - { - export_file_off = export_va.wrapping_sub(va).wrapping_add(f_off); - } - write_u32(&mut self.decompressed, sec_hdr.wrapping_add(16), sz); - write_u32(&mut self.decompressed, sec_hdr.wrapping_add(20), va); - if name.starts_with(".idata") { - write_u32( - &mut self.decompressed, - sec_hdr.wrapping_add(36), - 0xC000_0040, - ); - } - } - if export_size != 0 && export_file_off != 0 { - let d = export_va as usize; - let s = export_file_off as usize; - let n = export_size as usize; - self.decompressed[d..d + n].copy_from_slice(&self.file_data[s..s + n]); - } - - // ---- .text decrypt with decrypt_data8 (PE32 auto-detected formula) ---- - // `select_dd8_formula_pe32` returns None when `.text` was not packer-dd8- - // encrypted (native DLLs leave it plaintext); applying dd8 there would - // scramble valid code, so skip it entirely in that case. - if text_size > 0 && text_off > 0 { - if let Some(big) = - primitives::select_dd8_formula_pe32(&self.decompressed, text_off, text_size) - { - if verbose { - println!( - "[7/9] Decrypting .text (PE32 dd8, formula={})...", - if big { "0x8000*(page+1)" } else { "page+1" } - ); - } - let num_pages = text_size / 0x1000; - for page in 0..num_pages { - let pk = if big { - 0x8000u32.wrapping_mul(page.wrapping_add(1)) - } else { - page.wrapping_add(1) - }; - let pa = text_off.wrapping_add(page.wrapping_mul(0x1000)); - let mut k = pk; - let rk = k.rotate_right(15); - k = rk; - for bi in 1..256u32 { - let rk = k.rotate_right(15); - let ri = rk.wrapping_add(bi); - k = ri.wrapping_add(bi); - let tidx = - pa.wrapping_add(bi.wrapping_mul(16)).wrapping_add(ri & 0xF) as usize; - self.decompressed[tidx] ^= k as u8; - } - } - } else if verbose { - println!("[7/9] Skipping .text dd8 (already plaintext)..."); - } - } - - // ---- Fix data directories (PE32: data dirs at pe+0x78) ---- - let exe_pe = get_u32(&self.decompressed, 60); - for i in 0..128u32 { - self.decompressed[(exe_pe + 0x78 + i) as usize] = metadata_dirs[i as usize]; - } - // DLL-aware reloc / DllCharacteristics handling. An EXE's packer rebuilds - // the relocation table and clears DllCharacteristics, so the loader needs - // no relocations. A DLL, by contrast, is almost always mapped at a - // non-preferred base, so it MUST keep its base-relocation directory - // (restored above from metadata_dirs) and a valid DllCharacteristics - // (DYNAMIC_BASE) — zeroing them leaves the DLL unrelocatable and its - // imports pinned to the packer stub, so it fails to load (which looks - // like a missing/broken export table). - let is_dll = (get_u16(&self.decompressed, exe_pe.wrapping_add(22)) & 0x2000) != 0; - if !is_dll { - // EXE: clear BaseReloc (index 5 = pe+0xA0) and DllCharacteristics (pe+0x5E). - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xA0), 0); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xA4), 0); - write_u16(&mut self.decompressed, exe_pe.wrapping_add(0x5E), 0); - } else { - // DLL: keep the BaseReloc dir from metadata; ensure DYNAMIC_BASE. - let mut dll_chars = get_u16(&self.decompressed, exe_pe.wrapping_add(0x5E)); - if dll_chars == 0 { - dll_chars = 0x0040; // IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE - } - write_u16( - &mut self.decompressed, - exe_pe.wrapping_add(0x5E), - dll_chars as u32, - ); - } - - // ---- TLS directory reconstruction (PE32: index 9 = pe+0xC0) ---- - let tls_dir_rva = get_u32(&self.decompressed, exe_pe.wrapping_add(0xC0)); - let tls_dir_sz = get_u32(&self.decompressed, exe_pe.wrapping_add(0xC4)); - if tls_dir_rva > 0 - && tls_dir_sz >= 24 - && (tls_dir_rva as usize + 24) <= self.decompressed.len() - { - let all_zero = (0..6u32) - .all(|i| get_u32(&self.decompressed, tls_dir_rva.wrapping_add(i * 4)) == 0); - if all_zero { - let image_base = get_u32(&self.decompressed, exe_pe.wrapping_add(52)); - // Prefer the module's real TLS directory, which survives in the - // loader stub's plaintext `.rdata`/`.tls`. Only when the stub - // cannot supply one does a placeholder get synthesized: it keeps - // the image loadable, but drops the initialized TLS template, - // `_tls_index` and the TLS callback array, so any module that - // actually uses `thread_local` faults once it runs. - if !self.restore_pe32_tls_from_stub(pe_off, tls_dir_rva, image_base) { - let mut tls_sec_va: u32 = 0; - let mut data_sec_va: u32 = 0; - let mut data_sec_sz: u32 = 0; - let sh = exe_pe - .wrapping_add(24) - .wrapping_add(get_u16(&self.decompressed, exe_pe.wrapping_add(20)) as u32); - let ns = get_u16(&self.decompressed, exe_pe.wrapping_add(6)) as u32; - for i in 0..ns { - let s = sh.wrapping_add(i * 40); - let nm = get_string_to_null(&self.decompressed, s); - let va = get_u32(&self.decompressed, s.wrapping_add(12)); - let sz = get_u32(&self.decompressed, s.wrapping_add(16)); - if nm.starts_with(".tls") { - tls_sec_va = va; - } - if nm.starts_with(".data") { - data_sec_va = va; - data_sec_sz = sz; - } - } - if tls_sec_va > 0 && data_sec_va > 0 { - let start_raw = image_base.wrapping_add(tls_sec_va); - let end_raw = start_raw; - let idx_addr = image_base - .wrapping_add(data_sec_va) - .wrapping_add(data_sec_sz) - .wrapping_sub(16); - let cb_addr = image_base - .wrapping_add(data_sec_va) - .wrapping_add(data_sec_sz) - .wrapping_sub(8); - let scratch = (data_sec_va + data_sec_sz - 16) as usize; - for b in &mut self.decompressed[scratch..scratch + 16] { - *b = 0; - } - write_u32(&mut self.decompressed, tls_dir_rva, start_raw); - write_u32(&mut self.decompressed, tls_dir_rva.wrapping_add(4), end_raw); - write_u32( - &mut self.decompressed, - tls_dir_rva.wrapping_add(8), - idx_addr, - ); - write_u32( - &mut self.decompressed, - tls_dir_rva.wrapping_add(12), - cb_addr, - ); - write_u32(&mut self.decompressed, tls_dir_rva.wrapping_add(16), 0); - write_u32( - &mut self.decompressed, - tls_dir_rva.wrapping_add(20), - 0x30_0000, - ); - } else { - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xC0), 0); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xC4), 0); - } - } - } - } - - // ---- Import table (PE32, 4-byte thunks) ---- - if verbose { - println!("[8/9] Decrypting import strings (PE32)..."); - } - let import_table_addr = eighth_start.wrapping_add(off_import_table); - let mut import_table_ptr = get_u32(&self.decompressed, import_table_addr); - let mut idt_size = get_u32(&self.decompressed, import_table_addr.wrapping_add(4)); - - let metadata_import_rva = get_u32(&metadata_dirs, 8); - let metadata_import_size = get_u32(&metadata_dirs, 12); - let dlen = self.decompressed.len() as u32; - - let mut eighth_import_valid = false; - if 0 < import_table_ptr && import_table_ptr < dlen && 0 < idt_size && idt_size < 0x10000 { - let test_name = if import_table_ptr + 20 <= dlen { - get_u32(&self.decompressed, import_table_ptr.wrapping_add(12)) - } else { - 0 - }; - let test_ilt = if import_table_ptr + 4 <= dlen { - get_u32(&self.decompressed, import_table_ptr) - } else { - 0 - }; - if 0x1000 < test_name && test_name < dlen && 0x1000 < test_ilt && test_ilt < dlen { - eighth_import_valid = true; - } - } - let mut metadata_import_valid = false; - if 0x1000 < metadata_import_rva && metadata_import_rva < dlen.wrapping_sub(20) { - let test_name2 = get_u32(&self.decompressed, metadata_import_rva.wrapping_add(12)); - let test_ilt2 = get_u32(&self.decompressed, metadata_import_rva); - if 0x1000 < test_name2 && test_name2 < dlen && 0x1000 < test_ilt2 && test_ilt2 < dlen { - metadata_import_valid = true; - } - } - if metadata_import_valid - && (!eighth_import_valid || metadata_import_rva != import_table_ptr) - { - import_table_ptr = metadata_import_rva; - idt_size = metadata_import_size; - } - - if 0 < import_table_ptr && import_table_ptr < dlen && 0 < idt_size && idt_size < 0x10000 { - let mut idt_pos = import_table_ptr; - let idt_end = import_table_ptr.wrapping_add(idt_size); - while idt_pos.wrapping_add(20) <= idt_end { - let ilt_rva = get_u32(&self.decompressed, idt_pos); - let name_rva = get_u32(&self.decompressed, idt_pos.wrapping_add(12)); - let iat_rva = get_u32(&self.decompressed, idt_pos.wrapping_add(16)); - if ilt_rva == 0 && name_rva == 0 && iat_rva == 0 { - break; - } - if 0 < name_rva && name_rva < dlen { - self.decrypt_data7(name_rva, name_rva as u8); - } - let thunk_base = if 0 < ilt_rva && ilt_rva < dlen { - ilt_rva - } else { - iat_rva - }; - if 0 < thunk_base && thunk_base < dlen.wrapping_sub(4) { - let mut thunk_pos = thunk_base; - while thunk_pos.wrapping_add(4) <= dlen { - let thunk_val = get_u32(&self.decompressed, thunk_pos); - if thunk_val == 0 { - break; - } - if thunk_val & 0x8000_0000 == 0 && thunk_val.wrapping_add(2) < dlen { - self.decrypt_data7(thunk_val.wrapping_add(2), thunk_val as u8); - write_u16(&mut self.decompressed, thunk_val, 0); - } - thunk_pos = thunk_pos.wrapping_add(4); - } - } - idt_pos = idt_pos.wrapping_add(20); - } - } - - // Update PE header: Import directory (index 1 = pe+0x80), clear IAT - // directory (index 12 = pe+0xD8). - write_u32( - &mut self.decompressed, - exe_pe.wrapping_add(0x80), - import_table_ptr, - ); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0x84), idt_size); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xD8), 0); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xDC), 0); - - // ---- EP (from metadata) ---- - if metadata_ep > 0 { - write_u32(&mut self.decompressed, exe_pe.wrapping_add(40), metadata_ep); - } else { - let real_ep = get_u32(self.file_data, pe_off.wrapping_add(40)); - write_u32(&mut self.decompressed, exe_pe.wrapping_add(40), real_ep); - } - - // ---- Output transforms ---- - if verbose { - println!("[9/9] Rebuilding PE file layout (PE32)..."); - } - let mut out = std::mem::take(&mut self.decompressed); - // kmiat import relocation is an EXE-only fixup: it discards the original - // import directory in favour of the loader-written IAT stub. A DLL keeps - // its real import table (restored above from metadata), so skip kmiat for - // DLLs (`!is_dll` guard). - let is_dll = (get_u16(&out, pe_off.wrapping_add(22)) & 0x2000) != 0; - if !is_dll && !primitives::pe32_imports_already_match_idata_layout(&mut out, pe_off) { - primitives::move_pe32_imports_to_kmiat(&mut out, pe_off); - } - let compact = primitives::compact_memory_image_to_pe(&out, pe_off) - .ok_or(UnpackError::Pe32OutputLayoutInvalid)?; - Ok(compact) - } } #[cfg(test)] diff --git a/senbei-pe/src/engine/exe/pipeline/pe32.rs b/senbei-pe/src/engine/exe/pipeline/pe32.rs new file mode 100644 index 0000000..0b319ba --- /dev/null +++ b/senbei-pe/src/engine/exe/pipeline/pe32.rs @@ -0,0 +1,1063 @@ +use super::super::super::layout; +use super::*; + +impl<'a> Unpacker<'a> { + /// PE32 (32-bit) unpack pipeline. The shared Stage 1/2 setup (info decrypt, + /// payload decrypt, raw copy, header restore) has already run in `run()` + /// before dispatch; this takes over from "Locating shell offsets". + pub(super) fn run_pe32(&mut self, pe_off: u32, verbose: bool) -> Result, UnpackError> { + let info = self.info; + let info3 = info[3]; + + // advance_key: replays the packer's per-iteration key walk. + let advance_key = |mut key: u32, iterations: u32| -> u32 { + for m in 0..iterations { + let bound = (m + 1).wrapping_mul(25) << 2; + let mut n: u32 = 1; + while n <= bound { + key = key.wrapping_add(n); + n += 1; + } + } + key + }; + + // ---- Locate tbl in shell ---- + let tbl = + layout::find_tbl_pe32(&self.decompressed, &info).ok_or(UnpackError::Pe32TblNotFound)?; + if verbose { + println!("[3/9] Locating config layout (PE32)..."); + println!(" tbl = 0x{:X}", tbl); + } + + // ---- PE header restore ---- + let val_bc = get_u32(&self.decompressed, tbl.wrapping_add(0xBC)); + let val_c8 = get_u32(&self.decompressed, tbl.wrapping_add(0xC8)); + let val_cc = get_u32(&self.decompressed, tbl.wrapping_add(0xCC)); + write_u32(&mut self.decompressed, pe_off.wrapping_add(0x80), val_bc); + write_u32(&mut self.decompressed, pe_off.wrapping_add(0x88), val_c8); + write_u32(&mut self.decompressed, pe_off.wrapping_add(0x8C), val_cc); + write_u32(&mut self.decompressed, pe_off.wrapping_add(0xB0), 0); + write_u32(&mut self.decompressed, pe_off.wrapping_add(0xB4), 0); + + // ---- Header-independent checksum inputs ---- + let first_stage_cs = self.calculate_checksum(tbl.wrapping_add(0xA8)); + let second_stage_key = get_u32(&self.decompressed, tbl.wrapping_add(0x40)); + + // ---- Stage 3: SecondStage ---- + // + // ss_key = headerChecksum ^ firstStageCS ^ secondStageKey, where the + // header checksum (a XOR of crc32(region)^size over the sub-regions at + // tbl+0x58) is taken over the *original* pre-pack PE header. For EXEs the + // import/resource restore above reconstructs that header exactly. Native + // DLLs additionally carry a packer-added BaseReloc data-directory entry + // (dir 5) that was absent from the checksummed original, so the header + // checksum only matches once that entry is treated as zero. EXEs have no + // dir-5 entry, so zeroing it is a no-op for them. + // + // Rather than branch on EXE-vs-DLL, try the header as-is and, on failure, + // with the BaseReloc entry zeroed; keep whichever ss_key decrypts a + // SecondStage whose ThirdStage (off,size) pair lands inside the image. + // This uses the same shift/key trial-and-validate the later stages + // already use, and keeps EXE output byte-identical (the as-is variant + // wins first). + let ss_pair = tbl.wrapping_add(0x98); + let ss = get_u32(&self.decompressed, ss_pair); + let ss_size = get_u32(&self.decompressed, ss_pair.wrapping_add(4)); + let ss_shift = ss_size.wrapping_sub(0xBC0); + // Back up the SecondStage ciphertext so a failed trial can be retried. + let ss_lo = ss as usize; + let ss_hi = ss_lo.wrapping_add(ss_size as usize); + if ss_hi < ss_lo || ss_hi > self.decompressed.len() { + return Err(UnpackError::Pe32SecondStageRangeInvalid { + offset: ss, + size: ss_size, + image_len: self.decompressed.len(), + }); + } + let ss_ct: Vec = self.decompressed[ss_lo..ss_hi].to_vec(); + // PE32 data dir 5 (BaseReloc) = optional_header(pe+24) + 0x60 + 5*8 = pe+0xA0. + let reloc_dir = pe_off.wrapping_add(0xA0); + let len = self.decompressed.len() as u64; + let pair_off = 0xB8Cu32.wrapping_add(ss_shift); + let mut found = false; + // Holds the winning variant's header checksum; the later stages + // (Forth/Fifth/Seven/Eighth) reuse it as a key component. + let mut header_checksum: u32 = 0; + for zero_reloc in [false, true] { + if zero_reloc { + write_u32(&mut self.decompressed, reloc_dir, 0); + write_u32(&mut self.decompressed, reloc_dir.wrapping_add(4), 0); + } + let mut hcs_addr = tbl.wrapping_add(0x58); + header_checksum = 0; + while get_u32(&self.decompressed, hcs_addr.wrapping_add(4)) != 0 { + header_checksum ^= self.calculate_checksum(hcs_addr); + hcs_addr = hcs_addr.wrapping_add(8); + } + let ss_key = header_checksum ^ first_stage_cs ^ second_stage_key; + self.decompressed[ss_lo..ss_hi].copy_from_slice(&ss_ct); + self.decrypt_data3(ss_pair, ss_key, 21); + // Validate: the ThirdStage (off,size) pair must reference the image. + let pair = ss.wrapping_add(pair_off); + let off = get_u32(&self.decompressed, pair) as u64; + let sz = get_u32(&self.decompressed, pair.wrapping_add(4)) as u64; + if off > 0x1000 && off < len && sz >= 4 && off.saturating_add(sz) <= len { + found = true; + break; + } + } + if !found { + return Err(UnpackError::Pe32RelocationDataNotFound); + } + if verbose { + println!( + " ss = 0x{:08X}, size = 0x{:X}, shift = 0x{:X}", + ss, ss_size, ss_shift + ); + } + + // ---- PE32 fixed offsets ---- + let third_key_off = 0x968u32.wrapping_add(ss_shift); + let forth_key_off = 0x964u32.wrapping_add(ss_shift); + let cs_base_off = 0x96Cu32.wrapping_add(ss_shift); + let dp_base_off = 0xA9Cu32.wrapping_add(ss_shift); + + // ---- Stage 4: ThirdStage (brute-force the rotate shift) ---- + let third_pair_off = 0xB8Cu32.wrapping_add(ss_shift); + let key = get_u32(&self.decompressed, ss.wrapping_add(third_key_off)); + let pair_addr = ss.wrapping_add(third_pair_off); + let ts_addr = get_u32(&self.decompressed, pair_addr); + let ts_size_raw = get_u32(&self.decompressed, pair_addr.wrapping_add(4)); + let backup: Vec = + self.decompressed[ts_addr as usize..(ts_addr + ts_size_raw) as usize].to_vec(); + let mut info_table: Option = None; + let mut keys_addr: u32 = 0; + let mut ts: u32 = 0; + for &shift in &[19u32, 21, 17, 23, 15, 25, 13, 11] { + self.decompressed[ts_addr as usize..(ts_addr + ts_size_raw) as usize] + .copy_from_slice(&backup); + write_u32(&mut self.decompressed, pair_addr, ts_addr); + write_u32( + &mut self.decompressed, + pair_addr.wrapping_add(4), + ts_size_raw, + ); + self.decrypt_data3(pair_addr, key, shift); + let mut off = 0u32; + while off + 32 < ts_size_raw { + let t0 = get_u32(&self.decompressed, ts_addr.wrapping_add(off)); + if t0 == 1 || t0 == 0x11 { + let t1 = get_u32(&self.decompressed, ts_addr.wrapping_add(off + 16)); + if t1 == 2 { + let addr0 = get_u32(&self.decompressed, ts_addr.wrapping_add(off + 4)); + if 0x1000 < addr0 && (addr0 as usize) < self.decompressed.len() { + let it = ts_addr.wrapping_add(off); + info_table = Some(it); + keys_addr = it.wrapping_sub(0x58); + ts = ts_addr; + break; + } + } + } + off = off.wrapping_add(4); + } + if info_table.is_some() { + break; + } + } + let info_table = info_table.ok_or(UnpackError::Pe32ThirdStageFailed)?; + if verbose { + println!("[4/9] Decrypting stages (PE32)..."); + println!( + " thirdStage start = 0x{:X}, infoTable = 0x{:X}", + ts, info_table + ); + } + + // ---- Process infoTable ---- + let mut it_addr = info_table; + for _ in 0..2 { + let tval = get_u32(&self.decompressed, it_addr); + if tval == 1 || tval == 0x11 { + self.decrypt_data4(it_addr.wrapping_add(4)); + } else if tval == 2 { + let mut copy_addr = get_u32(&self.decompressed, it_addr.wrapping_add(4)); + loop { + self.decrypt_data5(copy_addr, 16); + let s_a = get_u32(&self.decompressed, copy_addr); + let s_sz = get_u32(&self.decompressed, copy_addr.wrapping_add(4)); + let d_a = get_u32(&self.decompressed, copy_addr.wrapping_add(8)); + let d_sz = get_u32(&self.decompressed, copy_addr.wrapping_add(12)); + copy_addr = copy_addr.wrapping_add(16); + if s_sz == 0 { + break; + } + if s_a != 0 && d_a != 0 && d_sz == s_sz { + let sa = s_a as usize; + let da = d_a as usize; + let n = s_sz as usize; + self.decompressed.copy_within(sa..sa + n, da); + } + } + } + it_addr = it_addr.wrapping_add(16); + } + + // ---- keyOffsets ---- + let mut ka = keys_addr; + for k in 0..2usize { + let mut ka2 = ka; + for l in 0..2usize { + self.decrypt_data4(ka2); + self.key_offsets[k * 2 + l] = get_u32(&self.decompressed, ka2); + ka2 = ka2.wrapping_add(8); + } + ka = ka.wrapping_add(32); + } + + // ---- Checksum addresses ---- + let second_stage_cs_addr = tbl.wrapping_add(0xB0); + let forth_stage_cs_addr = ss.wrapping_add(cs_base_off); + let fifth_stage_cs_addr = ss.wrapping_add(cs_base_off).wrapping_add(0x08); + let seven_stage_cs_addr = ss.wrapping_add(cs_base_off).wrapping_add(0x10); + + // ---- ForthStage ---- + let second_stage_cs = self.calculate_checksum(second_stage_cs_addr); + let forth_stage_key = advance_key( + get_u32(&self.decompressed, ss.wrapping_add(forth_key_off)), + 4, + ); + let dp_base = ss.wrapping_add(dp_base_off); + let forth_addr = dp_base.wrapping_add(0x40); + let fk = header_checksum ^ second_stage_cs ^ forth_stage_key; + if let Err(reason) = self.decrypt_and_decompress_data(forth_addr, fk, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32FourthStage, + reason, + }); + } + + // ---- FifthStage ---- + let fifth_addr = dp_base.wrapping_add(0x50); + let forth_cs = self.calculate_checksum(forth_stage_cs_addr); + let forth_region_off = get_u32(&self.decompressed, forth_stage_cs_addr); + let forth_region_sz = get_u32(&self.decompressed, forth_stage_cs_addr.wrapping_add(4)); + let fifth_key = get_u32( + &self.decompressed, + forth_region_off + .wrapping_add(forth_region_sz) + .wrapping_sub(4), + ); + let fk5 = header_checksum ^ forth_cs ^ fifth_key; + if let Err(reason) = self.decrypt_and_decompress_data(fifth_addr, fk5, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32FifthStage, + reason, + }); + } + + // ---- SevenStage ---- + let seven_addr = dp_base.wrapping_add(0x70); + let seven_dsz = get_u32(&self.decompressed, seven_addr.wrapping_add(12)); + let fifth_cs = self.calculate_checksum(fifth_stage_cs_addr); + let cs1_addr = get_u32( + &self.decompressed, + ss.wrapping_add(cs_base_off).wrapping_add(0x08), + ); + let cs1_size = get_u32( + &self.decompressed, + ss.wrapping_add(cs_base_off) + .wrapping_add(0x08) + .wrapping_add(4), + ); + let seven_key = !get_u32( + &self.decompressed, + cs1_addr.wrapping_add(cs1_size).wrapping_sub(0x10), + ); + let fk7 = header_checksum ^ fifth_cs ^ seven_key; + if let Err(reason) = self.decrypt_and_decompress_data(seven_addr, fk7, None) { + return Err(UnpackError::StageDecompressionFailed { + stage: DecompressionStage::Pe32SeventhStage, + reason, + }); + } + + // ---- EighthStage ---- + let seven_start_actual = get_u32(&self.decompressed, seven_addr); + if verbose { + println!("[5/9] Decrypting eighthStage (PE32)..."); + println!( + " sevenStart = 0x{:X}, sevenDsz = 0x{:X}", + seven_start_actual, seven_dsz + ); + } + // Locate the customDecryptor LFSR block (scan backward from middle, then + // forward as fallback). + let scan_start = seven_dsz / 2; + let custom_dec_off = layout::find_lfsr_block( + &self.decompressed, + seven_start_actual, + seven_dsz, + scan_start, + true, + ) + .or_else(|| { + layout::find_lfsr_block(&self.decompressed, seven_start_actual, seven_dsz, 0, false) + }) + .ok_or(UnpackError::Pe32CustomDecryptorNotFound)?; + let custom_dec_addr = seven_start_actual.wrapping_add(custom_dec_off); + self.decrypt_data6(custom_dec_addr); + let custom_ops = generate(&self.decompressed, custom_dec_addr).ok_or( + UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32CustomDecryptor), + )?; + + let seven_cs = self.calculate_checksum(seven_stage_cs_addr); + let eighth_addr = dp_base.wrapping_add(0xC0); + let eighth_dsz = get_u32(&self.decompressed, eighth_addr.wrapping_add(12)); + let eighth_src = get_u32(&self.decompressed, eighth_addr); + let eighth_ssz = get_u32(&self.decompressed, eighth_addr.wrapping_add(4)); + let eighth_backup: Vec = + self.decompressed[eighth_src as usize..(eighth_src + eighth_ssz) as usize].to_vec(); + let eighth_pair_bak: Vec = + self.decompressed[eighth_addr as usize..(eighth_addr + 16) as usize].to_vec(); + let data_len = self.decompressed.len() as u32; + + // Build the eighthStageKey candidate list (offsets relative to + // sevenStart) using gap heuristics + scan. + let mut candidates: Vec = Vec::new(); + let push_cand = |c: &mut Vec, off: u32| { + if !c.contains(&off) { + c.push(off); + } + }; + for &end_gap in &[0xD0u32, 0xC0, 0xE0, 0xB0, 0xA0, 0xF0, 0x100] { + if end_gap <= seven_dsz { + let off = seven_dsz - end_gap; + if off < seven_dsz { + let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); + if val != 0 && val != 0xCCCC_CCCC { + push_cand(&mut candidates, off); + } + } + } + } + for &gap in &[ + 0x70u32, 0xD0, 0x28, 0x50, 0x48, 0x30, 0x40, 0x58, 0x60, 0x20, 0x38, 0x80, 0x90, 0xA0, + 0xB0, + ] { + if gap <= custom_dec_off { + let off = custom_dec_off - gap; + if off + 4 <= seven_dsz && !candidates.contains(&off) { + let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); + if val != 0 && val != 0xCCCC_CCCC { + push_cand(&mut candidates, off); + } + } + } + } + let scan_lo = custom_dec_off.saturating_sub(0x100); + let mut off = scan_lo; + while off < custom_dec_off { + if !candidates.contains(&off) { + let val = get_u32(&self.decompressed, seven_start_actual.wrapping_add(off)); + let all_printable = (0..4u32).all(|i| { + let b = (val >> (i * 8)) & 0xFF; + (32..127).contains(&b) + }); + if val != 0 && val != 0xCCCC_CCCC && !all_printable { + push_cand(&mut candidates, off); + } + } + off = off.wrapping_add(4); + } + + let k1 = self.key_offsets[1]; + let k3 = self.key_offsets[3]; + let mut eighth_ok = false; + for ek_off in candidates { + self.decompressed[eighth_src as usize..(eighth_src + eighth_ssz) as usize] + .copy_from_slice(&eighth_backup); + self.decompressed[eighth_addr as usize..(eighth_addr + 16) as usize] + .copy_from_slice(&eighth_pair_bak); + let raw = get_u32(&self.decompressed, seven_start_actual.wrapping_add(ek_off)); + let test_key = advance_key(raw, 3); + let fk8 = header_checksum ^ fifth_cs ^ seven_cs ^ test_key; + let result = primitives::decrypt_and_decompress_data( + &mut self.decompressed, + eighth_addr, + fk8, + k1, + k3, + Some(&custom_ops), + ); + if result { + let est = get_u32(&self.decompressed, eighth_addr); + if 0x1000 < est && est < data_len { + eighth_ok = true; + break; + } + } + } + if !eighth_ok { + return Err(UnpackError::Pe32EighthKeyNotFound); + } + let eighth_start = get_u32(&self.decompressed, eighth_addr); + if verbose { + println!( + " eighthStart = 0x{:08X}, dsz = 0x{:X}", + eighth_start, eighth_dsz + ); + } + + // ---- Final processing offsets (anchored on the eighthStage config cluster) ---- + // + // The eighthStage holds a config cluster — importTable, fileCS, + // compressedInfo, zeroList — at fixed offsets from a cluster base + // (base+0x18 / +0x30 / +0x40 / +0x48) with the fileLFSR at +0x4B4. + // Classic builds stamp a 0x00007679 dword at that base; native DLLs and + // some older PE32 EXEs (ss_size=0xBE8) omit the stamp. + // Locate the cluster by stamp when present (validated by fileCS at + // base+0x30 pointing past info[3]); otherwise fall back to finding the + // fileCS slot by shape — (addr, size) with addr just past info[3] and a + // small 16-aligned size — and back-derive base = fileCS_off - 0x30. + // Hardcoded eighthStart-relative constants remain as a last-resort + // fallback for builds where neither discovery path fires. + let marker = { + let mut m: Option = None; + let hi = eighth_dsz.saturating_sub(0x4C); + let mut o = 0u32; + while o < hi { + if get_u32(&self.decompressed, eighth_start.wrapping_add(o)) == 0x7679 { + let fc = get_u32(&self.decompressed, eighth_start.wrapping_add(o + 0x30)); + if fc > info3 && (fc as usize) < self.decompressed.len() { + m = Some(o); + break; + } + } + o = o.wrapping_add(4); + } + if m.is_none() { + // fileCS-shaped slot: addr in (info3, info3+0x2000], size in + // 0x10..=0x200 and 16-aligned. Prefer the candidate whose addr + // is closest to (but past) info3 — matches every observed + // build (one PE32 EXE family dist ~0x1C0, another ~0x1A0). + let mut best: Option<(u32 /*dist*/, u32 /*off*/)> = None; + let mut o = 0u32; + let dlen = self.decompressed.len() as u32; + while o + 8 <= eighth_dsz.saturating_sub(0x4B4u32.saturating_sub(0x30)) { + let fc = get_u32(&self.decompressed, eighth_start.wrapping_add(o)); + let sz = get_u32(&self.decompressed, eighth_start.wrapping_add(o + 4)); + if fc > info3 + && fc <= info3.wrapping_add(0x2000) + && fc < dlen + && (0x10..=0x200).contains(&sz) + && (sz & 0xF) == 0 + { + // Cluster base must leave room for the +0x4B4 LFSR slot + // (even if the exact LFSR is later adjusted by scan). + if o >= 0x30 { + let base = o - 0x30; + if base.wrapping_add(0x4C) <= eighth_dsz { + let dist = fc - info3; + match best { + None => best = Some((dist, base)), + Some((bd, _)) if dist < bd => best = Some((dist, base)), + _ => {} + } + } + } + } + o = o.wrapping_add(4); + } + if let Some((dist, base)) = best { + if verbose { + println!( + " pe32 cluster via fileCS (no 0x7679): base=+0x{:X} dist_info3=0x{:X}", + base, dist + ); + } + m = Some(base); + } + } + m + }; + let (off_import_table, off_file_cs, off_compressed_info, off_zero_list, off_file_lfsr) = + match marker { + Some(m) => (m + 0x18, m + 0x30, m + 0x40, m + 0x48, m + 0x4B4), + None => ( + 0x3C50u32.wrapping_add(ss_shift), + 0x3C68u32.wrapping_add(ss_shift), + 0x3C78u32.wrapping_add(ss_shift), + 0x3C80u32.wrapping_add(ss_shift), + 0x40ECu32.wrapping_add(ss_shift), + ), + }; + + // ---- File checksums (permanent decrypt) ---- + let file_cs_addr_ptr = eighth_start.wrapping_add(off_file_cs); + let mut file_cs_addr = get_u32(&self.decompressed, file_cs_addr_ptr); + let file_cs_size = get_u32(&self.decompressed, file_cs_addr_ptr.wrapping_add(4)); + if file_cs_size > 0 { + let file_cs_end = file_cs_addr.wrapping_add(file_cs_size); + while file_cs_addr < file_cs_end { + self.decrypt_data5(file_cs_addr, 16); + file_cs_addr = file_cs_addr.wrapping_add(16); + } + } else { + while get_u32(&self.decompressed, file_cs_addr.wrapping_add(4)) != 0 { + self.decrypt_data5(file_cs_addr, 16); + file_cs_addr = file_cs_addr.wrapping_add(16); + } + } + + // ---- File decryptor LFSR ---- + // + // When the marker-relative off_file_lfsr is in range, try that slot + // first (exact). If it is not a valid LFSR block, trial-and-validate + // candidates from off_zero_list forward — required for older PE32 EXEs + // without the 0x7679 stamp where the expected slot is empty and a loose + // decoded[0]+0xC3 nearest-hit picks the wrong decryptor. Fall back to + // the legacy loose scan only if no candidate trial-decompresses. Native + // DLLs have a smaller eighthStage where off_file_lfsr lands out of range + // and use the same trial-validate scan from just past the cluster (else + // branch). + let lfsr_off = if off_file_lfsr.wrapping_add(96) <= eighth_dsz { + let mut lfsr_off = off_file_lfsr; + let exact = layout::find_lfsr_block( + &self.decompressed, + eighth_start, + eighth_dsz, + off_file_lfsr, + false, + ); + if exact != Some(off_file_lfsr) { + // Prefer trial-and-validate (same as the DLL branch): a loose + // decoded[0]+0xC3 scan can land on coincidental LFSR-shaped + // blocks that decode to a wrong file_ops and scramble every + // compressed block. Observed on older PE32 EXEs without the + // 0x7679 cluster stamp: the expected slot is empty and the + // nearest loose hit is not the real decryptor. + let ci_slot = eighth_start.wrapping_add(off_compressed_info); + let mut scan = off_zero_list; + let mut chosen: Option = None; + let mut considered = 0u32; + while let Some(cand) = layout::find_lfsr_block( + &self.decompressed, + eighth_start, + eighth_dsz, + scan, + false, + ) { + considered = considered.wrapping_add(1); + if self.pe32_file_lfsr_validates(eighth_start.wrapping_add(cand), ci_slot) { + chosen = Some(cand); + break; + } + scan = cand + 1; + } + if let Some(c) = chosen { + if verbose { + println!( + " pe32 fileLFSR via trial-validate: +0x{:X} (expected +0x{:X}, considered {})", + c, off_file_lfsr, considered + ); + } + lfsr_off = c; + } else { + // No candidate trial-decompresses: fail loudly. The old + // "legacy loose scan" picked the nearest LFSR-shaped block + // by offset distance without any validation — that is + // exactly how a wrong file_ops got applied to every data + // block (uncompressed blocks never enter the decompressor), + // producing a plausible but fully wrong image (the PE32 + // .text scramble root cause). Trial-and-validate or error. + return Err(UnpackError::Pe32FileLfsrNotFound); + } + } + lfsr_off + } else { + // Native DLL: the marker-relative off_file_lfsr (EXE-tuned, marker + + // 0x4B4) overshoots the smaller DLL eighthStage, so the exact slot is + // unavailable. A plain forward scan returns the FIRST valid-opcode + // block, but the DLL eighthStage contains coincidental valid-opcode + // blocks that decode to trivial programs (e.g. a constant byte add) + // ahead of the real file decryptor. A wrong file_ops corrupts the + // per-block translate (applied before decompression), so every data + // block fails to decompress. Enumerate every candidate forward and + // keep the first whose decoded file_ops actually decompresses the + // first compressed data block — trial-and-validate, same idea as the + // D1/D2 fixes. Non-DLL (EXE) builds never reach this branch. + let ci_slot = eighth_start.wrapping_add(off_compressed_info); + let mut scan = off_zero_list.wrapping_add(8); + let mut chosen: Option = None; + while let Some(cand) = + layout::find_lfsr_block(&self.decompressed, eighth_start, eighth_dsz, scan, false) + { + if self.pe32_file_lfsr_validates(eighth_start.wrapping_add(cand), ci_slot) { + chosen = Some(cand); + break; + } + scan = cand + 1; + } + chosen.ok_or(UnpackError::Pe32FileLfsrNotFound)? + }; + let file_dec_addr = eighth_start.wrapping_add(lfsr_off); + self.decrypt_data6(file_dec_addr); + let file_ops = generate(&self.decompressed, file_dec_addr).ok_or( + UnpackError::BytecodeGenerationFailed(BytecodeStage::Pe32FileDecryptor), + )?; + + // ---- PE32 metadata: EP and data dirs from info[3] ---- + let test_val = get_u32(&self.decompressed, info3.wrapping_add(0x10)); + let metadata_ep: u32; + let mut metadata_dirs = [0u8; 128]; + if test_val > 0x10000 { + // Layout B + let s = info3.wrapping_add(0x10) as usize; + let backup_meta = self.decompressed[s..s + 0x290].to_vec(); + self.decrypt_data5(info3.wrapping_add(0x10), 0x290); + metadata_ep = get_u32(&self.decompressed, info3.wrapping_add(0x20)); + let d = info3.wrapping_add(0x30) as usize; + metadata_dirs.copy_from_slice(&self.decompressed[d..d + 128]); + self.decompressed[s..s + 0x290].copy_from_slice(&backup_meta); + } else { + // Layout A + let s = info3.wrapping_add(0x40) as usize; + let backup_meta = self.decompressed[s..s + 144].to_vec(); + self.decrypt_data5(info3.wrapping_add(0x40), 144); + metadata_ep = get_u32(&self.decompressed, info3.wrapping_add(0x40)); + let d = info3.wrapping_add(0x50) as usize; + metadata_dirs.copy_from_slice(&self.decompressed[d..d + 128]); + self.decompressed[s..s + 144].copy_from_slice(&backup_meta); + } + + // ---- Zero-out list (runs BEFORE decompression) ---- + let zero_list_addr = eighth_start.wrapping_add(off_zero_list); + let mut zero_ptr = get_u32(&self.decompressed, zero_list_addr); + loop { + self.decrypt_data5(zero_ptr, 16); + let src3 = get_u32(&self.decompressed, zero_ptr); + let s_sz3 = get_u32(&self.decompressed, zero_ptr.wrapping_add(4)); + zero_ptr = zero_ptr.wrapping_add(16); + if s_sz3 == 0 { + break; + } + if src3.wrapping_add(s_sz3) as usize > self.decompressed.len() { + break; + } + for b in &mut self.decompressed[src3 as usize..(src3 + s_sz3) as usize] { + *b = 0; + } + } + + // ---- File data decompression ---- + if verbose { + println!("[6/9] Loading and decompressing file data (PE32)..."); + } + let compress_data_offset = (!get_u32(self.file_data, 0x1080)).wrapping_add(0x1000); + let compressed_info_addr = eighth_start.wrapping_add(off_compressed_info); + let mut compressed_info = get_u32(&self.decompressed, compressed_info_addr); + // Pass 1 (sequential): position-keyed descriptor chain (decrypt_data5), + // terminated by a zero source-size record. + struct Blk { + src: u32, + ssz: u32, + dst: u32, + dsz: u32, + } + let mut blocks: Vec = Vec::new(); + loop { + self.decrypt_data5(compressed_info, 16); + let src2 = get_u32(&self.decompressed, compressed_info); + let s_sz2 = get_u32(&self.decompressed, compressed_info.wrapping_add(4)); + let dst2 = get_u32(&self.decompressed, compressed_info.wrapping_add(8)); + let d_sz2 = get_u32(&self.decompressed, compressed_info.wrapping_add(12)); + compressed_info = compressed_info.wrapping_add(16); + if s_sz2 == 0 { + break; + } + blocks.push(Blk { + src: src2, + ssz: s_sz2, + dst: dst2, + dsz: d_sz2, + }); + } + // Pass 2: independent per-block work over disjoint dst spans (see + // `parallel_for` for how the spans are carved safely). + { + let lut = OpsLut::new(&file_ops); + let clean = &self.file_data; + let ko = self.key_offsets; + let ks_snap = primitives::aes_schedule_snapshot(&self.decompressed, ko[2]) + .ok_or(UnpackError::InvalidAesKeySchedule { offset: ko[2] })?; + let tab_snap = primitives::huffman_table_snapshot(&self.decompressed, ko[0]) + .ok_or(UnpackError::InvalidHuffmanTable { offset: ko[0] })?; + let spans: Vec<(usize, usize)> = blocks + .iter() + .map(|b| { + let s = b.dst as usize; + (s, s + b.ssz.max(b.dsz) as usize) + }) + .collect(); + let do_block = |i: usize, base: usize, span: &mut [u8]| -> Result<(), UnpackError> { + let b = &blocks[i]; + let file_src = b.src.wrapping_add(compress_data_offset) as usize; + let rel = b.dst as usize - base; + let n = b.ssz as usize; + span[rel..rel + n].copy_from_slice(&clean[file_src..file_src + n]); + primitives::aes_decrypt_ks(&ks_snap, span, rel as u32, b.ssz); + lut.map_region(span, rel, n); + if b.ssz != b.dsz { + // decompress reports corruption (after partial writes) via + // its bool; surface it instead of shipping a garbage block. + if !primitives::decompress_tbl( + &tab_snap, span, rel as u32, rel as u32, b.ssz, b.dsz, + ) { + return Err(UnpackError::SectionDecompressionFailed { + pipeline: SectionPipeline::ExePe32, + block: i, + }); + } + } + Ok(()) + }; + super::super::super::parallel::parallel_for( + &mut self.decompressed, + &spans, + 1, + do_block, + )?; + } + + // ---- Section fixup ---- + self.decompressed[..0x1000].copy_from_slice(&self.file_data[..0x1000]); + let opt_hdr_size = get_u16(self.file_data, pe_off.wrapping_add(20)) as u32; + let sec_hdr_table = pe_off.wrapping_add(24).wrapping_add(opt_hdr_size); + let export_va = get_u32( + self.file_data, + pe_off + .wrapping_add(24) + .wrapping_add(opt_hdr_size) + .wrapping_sub(128), + ); + let export_size = get_u32( + self.file_data, + pe_off + .wrapping_add(24) + .wrapping_add(opt_hdr_size) + .wrapping_sub(124), + ); + let mut export_file_off: u32 = 0; + let mut text_off: u32 = 0; + let mut text_size: u32 = 0; + // Walk by NumberOfSections (PE has no zero-VS sentinel; a real + // VirtualSize==0 section would truncate these fixups early), stopping + // at the all-zero padding in case NumberOfSections is overstated. + let num_sections = get_u16(self.file_data, pe_off.wrapping_add(6)) as u32; + for i in 0..num_sections.min(96) { + let sec_hdr = sec_hdr_table.wrapping_add(i.wrapping_mul(40)); + if self.file_data[sec_hdr as usize..sec_hdr as usize + 8] + .iter() + .all(|&b| b == 0) + { + break; + } + let va = get_u32(self.file_data, sec_hdr.wrapping_add(12)); + let sz = get_u32(self.file_data, sec_hdr.wrapping_add(8)); + let f_off = get_u32(self.file_data, sec_hdr.wrapping_add(20)); + let name = section_name(self.file_data, sec_hdr); + if name.starts_with(".text") { + text_size = sz; + text_off = va; + } + if export_size != 0 + && export_va >= va + && export_va.wrapping_add(export_size) <= va.wrapping_add(sz) + { + export_file_off = export_va.wrapping_sub(va).wrapping_add(f_off); + } + write_u32(&mut self.decompressed, sec_hdr.wrapping_add(16), sz); + write_u32(&mut self.decompressed, sec_hdr.wrapping_add(20), va); + if name.starts_with(".idata") { + write_u32( + &mut self.decompressed, + sec_hdr.wrapping_add(36), + 0xC000_0040, + ); + } + } + if export_size != 0 && export_file_off != 0 { + let d = export_va as usize; + let s = export_file_off as usize; + let n = export_size as usize; + self.decompressed[d..d + n].copy_from_slice(&self.file_data[s..s + n]); + } + + // ---- .text decrypt with decrypt_data8 (PE32 auto-detected formula) ---- + // `select_dd8_formula_pe32` returns None when `.text` was not packer-dd8- + // encrypted (native DLLs leave it plaintext); applying dd8 there would + // scramble valid code, so skip it entirely in that case. + if text_size > 0 && text_off > 0 { + if let Some(big) = + layout::select_dd8_formula_pe32(&self.decompressed, text_off, text_size) + { + if verbose { + println!( + "[7/9] Decrypting .text (PE32 dd8, formula={})...", + if big { "0x8000*(page+1)" } else { "page+1" } + ); + } + let num_pages = text_size / 0x1000; + for page in 0..num_pages { + let pk = if big { + 0x8000u32.wrapping_mul(page.wrapping_add(1)) + } else { + page.wrapping_add(1) + }; + let pa = text_off.wrapping_add(page.wrapping_mul(0x1000)); + let mut k = pk; + let rk = k.rotate_right(15); + k = rk; + for bi in 1..256u32 { + let rk = k.rotate_right(15); + let ri = rk.wrapping_add(bi); + k = ri.wrapping_add(bi); + let tidx = + pa.wrapping_add(bi.wrapping_mul(16)).wrapping_add(ri & 0xF) as usize; + self.decompressed[tidx] ^= k as u8; + } + } + } else if verbose { + println!("[7/9] Skipping .text dd8 (already plaintext)..."); + } + } + + // ---- Fix data directories (PE32: data dirs at pe+0x78) ---- + let exe_pe = get_u32(&self.decompressed, 60); + for i in 0..128u32 { + self.decompressed[(exe_pe + 0x78 + i) as usize] = metadata_dirs[i as usize]; + } + // DLL-aware reloc / DllCharacteristics handling. An EXE's packer rebuilds + // the relocation table and clears DllCharacteristics, so the loader needs + // no relocations. A DLL, by contrast, is almost always mapped at a + // non-preferred base, so it MUST keep its base-relocation directory + // (restored above from metadata_dirs) and a valid DllCharacteristics + // (DYNAMIC_BASE) — zeroing them leaves the DLL unrelocatable and its + // imports pinned to the packer stub, so it fails to load (which looks + // like a missing/broken export table). + let is_dll = (get_u16(&self.decompressed, exe_pe.wrapping_add(22)) & 0x2000) != 0; + if !is_dll { + // EXE: clear BaseReloc (index 5 = pe+0xA0) and DllCharacteristics (pe+0x5E). + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xA0), 0); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xA4), 0); + write_u16(&mut self.decompressed, exe_pe.wrapping_add(0x5E), 0); + } else { + // DLL: keep the BaseReloc dir from metadata; ensure DYNAMIC_BASE. + let mut dll_chars = get_u16(&self.decompressed, exe_pe.wrapping_add(0x5E)); + if dll_chars == 0 { + dll_chars = 0x0040; // IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE + } + write_u16( + &mut self.decompressed, + exe_pe.wrapping_add(0x5E), + dll_chars as u32, + ); + } + + // ---- TLS directory reconstruction (PE32: index 9 = pe+0xC0) ---- + let tls_dir_rva = get_u32(&self.decompressed, exe_pe.wrapping_add(0xC0)); + let tls_dir_sz = get_u32(&self.decompressed, exe_pe.wrapping_add(0xC4)); + if tls_dir_rva > 0 + && tls_dir_sz >= 24 + && (tls_dir_rva as usize + 24) <= self.decompressed.len() + { + let all_zero = (0..6u32) + .all(|i| get_u32(&self.decompressed, tls_dir_rva.wrapping_add(i * 4)) == 0); + if all_zero { + let image_base = get_u32(&self.decompressed, exe_pe.wrapping_add(52)); + // Prefer the module's real TLS directory, which survives in the + // loader stub's plaintext `.rdata`/`.tls`. Only when the stub + // cannot supply one does a placeholder get synthesized: it keeps + // the image loadable, but drops the initialized TLS template, + // `_tls_index` and the TLS callback array, so any module that + // actually uses `thread_local` faults once it runs. + if !self.restore_pe32_tls_from_stub(pe_off, tls_dir_rva, image_base) { + let mut tls_sec_va: u32 = 0; + let mut data_sec_va: u32 = 0; + let mut data_sec_sz: u32 = 0; + let sh = exe_pe + .wrapping_add(24) + .wrapping_add(get_u16(&self.decompressed, exe_pe.wrapping_add(20)) as u32); + let ns = get_u16(&self.decompressed, exe_pe.wrapping_add(6)) as u32; + for i in 0..ns { + let s = sh.wrapping_add(i * 40); + let nm = get_string_to_null(&self.decompressed, s); + let va = get_u32(&self.decompressed, s.wrapping_add(12)); + let sz = get_u32(&self.decompressed, s.wrapping_add(16)); + if nm.starts_with(".tls") { + tls_sec_va = va; + } + if nm.starts_with(".data") { + data_sec_va = va; + data_sec_sz = sz; + } + } + if tls_sec_va > 0 && data_sec_va > 0 { + let start_raw = image_base.wrapping_add(tls_sec_va); + let end_raw = start_raw; + let idx_addr = image_base + .wrapping_add(data_sec_va) + .wrapping_add(data_sec_sz) + .wrapping_sub(16); + let cb_addr = image_base + .wrapping_add(data_sec_va) + .wrapping_add(data_sec_sz) + .wrapping_sub(8); + let scratch = (data_sec_va + data_sec_sz - 16) as usize; + for b in &mut self.decompressed[scratch..scratch + 16] { + *b = 0; + } + write_u32(&mut self.decompressed, tls_dir_rva, start_raw); + write_u32(&mut self.decompressed, tls_dir_rva.wrapping_add(4), end_raw); + write_u32( + &mut self.decompressed, + tls_dir_rva.wrapping_add(8), + idx_addr, + ); + write_u32( + &mut self.decompressed, + tls_dir_rva.wrapping_add(12), + cb_addr, + ); + write_u32(&mut self.decompressed, tls_dir_rva.wrapping_add(16), 0); + write_u32( + &mut self.decompressed, + tls_dir_rva.wrapping_add(20), + 0x30_0000, + ); + } else { + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xC0), 0); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xC4), 0); + } + } + } + } + + // ---- Import table (PE32, 4-byte thunks) ---- + if verbose { + println!("[8/9] Decrypting import strings (PE32)..."); + } + let import_table_addr = eighth_start.wrapping_add(off_import_table); + let mut import_table_ptr = get_u32(&self.decompressed, import_table_addr); + let mut idt_size = get_u32(&self.decompressed, import_table_addr.wrapping_add(4)); + + let metadata_import_rva = get_u32(&metadata_dirs, 8); + let metadata_import_size = get_u32(&metadata_dirs, 12); + let dlen = self.decompressed.len() as u32; + + let mut eighth_import_valid = false; + if 0 < import_table_ptr && import_table_ptr < dlen && 0 < idt_size && idt_size < 0x10000 { + let test_name = if import_table_ptr + 20 <= dlen { + get_u32(&self.decompressed, import_table_ptr.wrapping_add(12)) + } else { + 0 + }; + let test_ilt = if import_table_ptr + 4 <= dlen { + get_u32(&self.decompressed, import_table_ptr) + } else { + 0 + }; + if 0x1000 < test_name && test_name < dlen && 0x1000 < test_ilt && test_ilt < dlen { + eighth_import_valid = true; + } + } + let mut metadata_import_valid = false; + if 0x1000 < metadata_import_rva && metadata_import_rva < dlen.wrapping_sub(20) { + let test_name2 = get_u32(&self.decompressed, metadata_import_rva.wrapping_add(12)); + let test_ilt2 = get_u32(&self.decompressed, metadata_import_rva); + if 0x1000 < test_name2 && test_name2 < dlen && 0x1000 < test_ilt2 && test_ilt2 < dlen { + metadata_import_valid = true; + } + } + if metadata_import_valid + && (!eighth_import_valid || metadata_import_rva != import_table_ptr) + { + import_table_ptr = metadata_import_rva; + idt_size = metadata_import_size; + } + + if 0 < import_table_ptr && import_table_ptr < dlen && 0 < idt_size && idt_size < 0x10000 { + let mut idt_pos = import_table_ptr; + let idt_end = import_table_ptr.wrapping_add(idt_size); + while idt_pos.wrapping_add(20) <= idt_end { + let ilt_rva = get_u32(&self.decompressed, idt_pos); + let name_rva = get_u32(&self.decompressed, idt_pos.wrapping_add(12)); + let iat_rva = get_u32(&self.decompressed, idt_pos.wrapping_add(16)); + if ilt_rva == 0 && name_rva == 0 && iat_rva == 0 { + break; + } + if 0 < name_rva && name_rva < dlen { + self.decrypt_data7(name_rva, name_rva as u8); + } + let thunk_base = if 0 < ilt_rva && ilt_rva < dlen { + ilt_rva + } else { + iat_rva + }; + if 0 < thunk_base && thunk_base < dlen.wrapping_sub(4) { + let mut thunk_pos = thunk_base; + while thunk_pos.wrapping_add(4) <= dlen { + let thunk_val = get_u32(&self.decompressed, thunk_pos); + if thunk_val == 0 { + break; + } + if thunk_val & 0x8000_0000 == 0 && thunk_val.wrapping_add(2) < dlen { + self.decrypt_data7(thunk_val.wrapping_add(2), thunk_val as u8); + write_u16(&mut self.decompressed, thunk_val, 0); + } + thunk_pos = thunk_pos.wrapping_add(4); + } + } + idt_pos = idt_pos.wrapping_add(20); + } + } + + // Update PE header: Import directory (index 1 = pe+0x80), clear IAT + // directory (index 12 = pe+0xD8). + write_u32( + &mut self.decompressed, + exe_pe.wrapping_add(0x80), + import_table_ptr, + ); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0x84), idt_size); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xD8), 0); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(0xDC), 0); + + // ---- EP (from metadata) ---- + if metadata_ep > 0 { + write_u32(&mut self.decompressed, exe_pe.wrapping_add(40), metadata_ep); + } else { + let real_ep = get_u32(self.file_data, pe_off.wrapping_add(40)); + write_u32(&mut self.decompressed, exe_pe.wrapping_add(40), real_ep); + } + + // ---- Output transforms ---- + if verbose { + println!("[9/9] Rebuilding PE file layout (PE32)..."); + } + let mut out = std::mem::take(&mut self.decompressed); + // kmiat import relocation is an EXE-only fixup: it discards the original + // import directory in favour of the loader-written IAT stub. A DLL keeps + // its real import table (restored above from metadata), so skip kmiat for + // DLLs (`!is_dll` guard). + let is_dll = (get_u16(&out, pe_off.wrapping_add(22)) & 0x2000) != 0; + if !is_dll && !layout::pe32_imports_already_match_idata_layout(&mut out, pe_off) { + layout::move_pe32_imports_to_kmiat(&mut out, pe_off); + } + let compact = layout::compact_memory_image_to_pe(&out, pe_off) + .ok_or(UnpackError::Pe32OutputLayoutInvalid)?; + Ok(compact) + } +} diff --git a/src/unpacker/integrity.rs b/senbei-pe/src/engine/integrity.rs similarity index 100% rename from src/unpacker/integrity.rs rename to senbei-pe/src/engine/integrity.rs diff --git a/senbei-pe/src/engine/layout.rs b/senbei-pe/src/engine/layout.rs new file mode 100644 index 0000000..a9ef00a --- /dev/null +++ b/senbei-pe/src/engine/layout.rs @@ -0,0 +1,14 @@ +//! Internal PE layout discovery and image reconstruction. + +mod dd8; +mod discovery; +mod image; + +pub(super) use dd8::{select_dd8_formula_pe32, select_dd8_shift}; +pub(super) use discovery::{ + discover_eighth_slots, find_bytecode_offset, find_lfsr_block, find_str_pos, find_tbl_pe32, + find_v_after_pad, find_v4_offset, get_string_to_null, section_name, trial_decrypt5_u32, +}; +pub(super) use image::{ + compact_memory_image_to_pe, move_pe32_imports_to_kmiat, pe32_imports_already_match_idata_layout, +}; diff --git a/senbei-pe/src/engine/layout/dd8.rs b/senbei-pe/src/engine/layout/dd8.rs new file mode 100644 index 0000000..fc5aca3 --- /dev/null +++ b/senbei-pe/src/engine/layout/dd8.rs @@ -0,0 +1,321 @@ +//! Validation-driven selection for per-page text transforms. + +/// PE32 `.text` dd8 key-formula selection with a skip decision. The packer keys +/// the per-page XOR either with `page+1` or `0x8000*(page+1)`; the formula is +/// not recorded. Replays the dd8 page pass on a scratch copy of sample pages +/// (25/50/75% of `.text`) under each formula and counts how many positions +/// decode to `0xCC` (int3 padding). +/// +/// Returns `Some(true)` for the `0x8000*(page+1)` formula, `Some(false)` for +/// `page+1`, or `None` when `.text` must NOT be dd8-decrypted at all. The packer +/// dd8-encrypts `.text` on EXEs (so unpacking must replay it) but leaves a native +/// DLL's `.text` plaintext; replaying dd8 there scrambles ~1 byte per 16-byte +/// block. The decision: dd8 only *restores* int3 padding when `.text` was +/// genuinely encrypted, so apply it only when the chosen formula's whole-page +/// 0xCC count rises *clearly* above the no-dd8 baseline; otherwise skip. +/// +/// "Clearly" matters: dd8 XORs 255 positions per page with pseudo-random bytes, +/// so on an already-plaintext `.text` it manufactures ~1 spurious `0xCC` per +/// sampled page for free (255/256 expected). A bare `best > baseline` test is +/// therefore biased towards *applying* dd8 on exactly the inputs that must skip +/// it — and a wrongly-applied dd8 is silent: it scrambles ~1 byte per 16 with no +/// error and nothing downstream (not even `integrity::check`, which only reads +/// 16 bytes at the entry point) notices. The [`MIN_DD8_NET_GAIN`] floor below is +/// the PE32 counterpart of the margin+floor `select_dd8_shift` already applies +/// on PE32+ for the same failure mode. +pub fn select_dd8_formula_pe32(data: &[u8], text_off: u32, text_size: u32) -> Option { + let num_pages_total = text_size / 0x1000; + let mut sample_pages: Vec = Vec::new(); + for frac in [0.25f64, 0.5, 0.75] { + let pg = (num_pages_total as f64 * frac) as u32; + if pg > 0 && pg < num_pages_total { + sample_pages.push(pg); + } + } + if sample_pages.is_empty() && num_pages_total > 1 { + sample_pages.push(num_pages_total / 2); + } + let score = |big: bool| -> i64 { + let mut total = 0i64; + for &sp in &sample_pages { + let pg_off = (text_off + sp * 0x1000) as usize; + if pg_off + 0x1000 > data.len() { + continue; + } + let mut buf = [0u8; 0x1000]; + buf.copy_from_slice(&data[pg_off..pg_off + 0x1000]); + let pk = if big { + 0x8000u32.wrapping_mul(sp.wrapping_add(1)) + } else { + sp.wrapping_add(1) + }; + let mut k = pk; + let rk = k.rotate_right(15); + k = rk; + for bi in 1..256u32 { + let rk = k.rotate_right(15); + let ri = rk.wrapping_add(bi); + k = ri.wrapping_add(bi); + let tidx = (bi.wrapping_mul(16).wrapping_add(ri & 0xF)) as usize; + if tidx < buf.len() { + buf[tidx] ^= k as u8; + } + } + total += buf.iter().filter(|&&b| b == 0xCC).count() as i64; + } + total + }; + let s_small = score(false); + let s_big = score(true); + // Baseline: whole-page 0xCC over the same sample pages with NO dd8. dd8 only + // rewrites 255 bytes per page, so comparing the chosen formula's whole-page + // 0xCC against this baseline reveals whether dd8 *restores* int3 padding + // (count rises -> .text was packer-encrypted, apply) or merely scrambles + // already-plaintext code (count falls -> native-DLL .text left intact, skip). + let mut baseline: i64 = 0; + for &sp in &sample_pages { + let pg_off = (text_off + sp * 0x1000) as usize; + if pg_off + 0x1000 > data.len() { + continue; + } + baseline += data[pg_off..pg_off + 0x1000] + .iter() + .filter(|&&b| b == 0xCC) + .count() as i64; + } + let big = s_big > s_small; + let best = s_small.max(s_big); + // Minimum net 0xCC gain over the baseline before dd8 is applied. Noise on an + // already-plaintext `.text` is ~1 manufactured 0xCC per sampled page (3 pages + // -> ~3); every corpus build that genuinely needs dd8 gains +154 or more + // (observed +154 and +312), and the one native DLL that must skip scores -18. + // A floor of 32 sits ~10x above the noise and ~5x below the smallest true + // positive, so it changes no existing decision. + const MIN_DD8_NET_GAIN: i64 = 32; + let apply = best.saturating_sub(baseline) >= MIN_DD8_NET_GAIN; + if std::env::var("SEL_DIAG").is_ok() { + eprintln!( + "SEL pe32 dd8 s_small={} s_big={} baseline={} gain={} big={} apply={}", + s_small, + s_big, + baseline, + best - baseline, + big, + apply + ); + } + // When no interior pages could be sampled (tiny .text) we cannot measure the + // effect; preserve the historical behavior of applying dd8. + if sample_pages.is_empty() || apply { + Some(big) + } else { + None + } +} + +// --------------------------------------------------------------------------- +// dd8 page-XOR shift selection. +// +// The packer scrambles ~1 byte per 16-byte block of .text via decrypt_data8, +// keyed by `page_idx << shift` (absolute page index = text_va >> 12). Observed +// shifts are 0 and 15. The shift is NOT stored in any header/config field: +// two otherwise-unrelated builds can carry byte-identical config-version stamps +// (0x40327253) yet require different shifts, so the only reliable discriminator +// is the .text content itself. +// +// Detection scoring formula: for each candidate shift, replay decrypt_data8 +// across a few sample pages (25/50/75% of .text) and count how many of the 255 +// mutated positions become 0xCC — the MSVC int3 padding byte. The correct shift +// hits int3 pads disproportionately often (~3-10x the baseline), so the +// highest-scoring shift wins. If neither shift clears 2x the baseline, .text +// is already plaintext → skip (return 99). +// +// This replaces an earlier entry-stub oracle that matched the 14 fixed CRT-stub +// bytes at the AEP. That oracle false-positived on a newer EXE-64 build: dd8 +// corrupted only the call rel32 (bytes 5-8, the wildcard region), so the stub +// matched under BOTH shifts and the selector defaulted to 0 when the truth was +// 15. The 0xCC statistic samples hundreds of positions per page and is not +// fooled by a stub whose fixed bytes happen to survive. +// --------------------------------------------------------------------------- +pub fn select_dd8_shift(data: &[u8], text_va: u32, text_size: u32, _info3: u32) -> u32 { + if text_size < 0x1000 { + return 0; + } + let text_off = text_va as usize; + let num_pages_total = text_size >> 12; + + // Sample pages at 25/50/75% of .text, falling back to the midpoint for tiny + // sections. + let mut sample_pages: Vec = Vec::new(); + for frac in [0.25f64, 0.5, 0.75] { + let pg = (num_pages_total as f64 * frac) as u32; + if pg > 0 && pg < num_pages_total { + sample_pages.push(pg); + } + } + if sample_pages.is_empty() && num_pages_total > 1 { + sample_pages.push(num_pages_total / 2); + } + if sample_pages.is_empty() { + return 0; + } + + let none_hits = score_dd8_baseline(data, text_off, &sample_pages); + let s0 = score_dd8_shift(data, text_off, text_va, &sample_pages, 0); + let s15 = score_dd8_shift(data, text_off, text_va, &sample_pages, 15); + let mut best_score = none_hits; + let mut best_shift = 99u32; // 99 == skip dd8 + for (shift, hits) in [(0u32, s0), (15u32, s15)] { + if hits > best_score { + best_score = hits; + best_shift = shift; + } + } + // Require a clear 2x margin over the already-plaintext baseline AND an + // absolute floor. The 2x test alone + // trips on noise when the counts are tiny: an external-companion DLL whose + // .text is already plaintext scores s15=4 vs none=1 — a spurious 4x — and + // gets dd8 wrongly applied, corrupting ~1 byte per 16. Across the whole + // golden corpus every build that genuinely needs dd8 scores >= 10 (lowest + // observed scores at 10-12; up to 107), so a floor of 8 rejects the noise + // while keeping every golden's shift selection unchanged. + const MIN_DD8_HITS: u32 = 8; + if best_shift != 99 && (best_score < none_hits * 2 || best_score < MIN_DD8_HITS) { + best_shift = 99; + } + if std::env::var("SEL_DIAG").is_ok() { + eprintln!( + "SEL dd8 best_shift={} s0={} s15={} none_hits={} samples={:?}", + best_shift, s0, s15, none_hits, sample_pages + ); + } + best_shift +} + +// Baseline: count int3 pads already present at the first byte of each 16-byte +// block, i.e. the positions dd8 would target if its in-block offset were 0. +fn score_dd8_baseline(data: &[u8], text_off: usize, sample_pages: &[u32]) -> u32 { + let mut hits = 0u32; + for &sp in sample_pages { + let pg_off = text_off + (sp as usize) * 0x1000; + if pg_off + 0x1000 > data.len() { + continue; + } + for bi in 1..256usize { + if data[pg_off + bi * 16] == 0xCC { + hits += 1; + } + } + } + hits +} + +// Replay decrypt_data8 on each sample page under `shift` and count how many of +// the 255 mutated positions decode to 0xCC. +fn score_dd8_shift( + data: &[u8], + text_off: usize, + text_va: u32, + sample_pages: &[u32], + shift: u32, +) -> u32 { + let abs_base = text_va >> 12; + let mut hits = 0u32; + for &sp in sample_pages { + let pg_off = text_off + (sp as usize) * 0x1000; + if pg_off + 0x1000 > data.len() { + continue; + } + let abs_page = abs_base.wrapping_add(sp); + let mut key = abs_page << shift; + for bi in 0..256u32 { + let mixed = key.rotate_right(15).wrapping_add(bi); + key = mixed.wrapping_add(bi); + if bi == 0 { + continue; + } + let tidx = (bi.wrapping_mul(16).wrapping_add(mixed & 0xF)) as usize; + if tidx < 0x1000 { + let mutated = data[pg_off + tidx] ^ (key as u8); + if mutated == 0xCC { + hits += 1; + } + } + } + } + hits +} + +#[cfg(test)] +mod tests { + use super::*; + + /// Seed the first `count` dd8-targeted positions of each sampled page with + /// the byte that decodes to `0xCC` under the `page+1` formula — i.e. an + /// encrypted `.text` whose plaintext is int3 padding. Positions whose key + /// byte would make the *ciphertext* itself `0xCC` are skipped so the + /// fixture contains no `0xCC` at all and every post-dd8 `0xCC` is a genuine + /// gain over a zero baseline. + fn seed_dd8_int3(data: &mut [u8], text_off: u32, pages: &[u32], count: u32) { + for &sp in pages { + let pg_off = (text_off + sp * 0x1000) as usize; + let mut k = sp.wrapping_add(1); + k = k.rotate_right(15); + let mut planted = 0u32; + for bi in 1..256u32 { + let ri = k.rotate_right(15).wrapping_add(bi); + k = ri.wrapping_add(bi); + if planted >= count { + continue; + } + let ct = 0xCCu8 ^ (k as u8); + if ct == 0xCC { + continue; + } + let tidx = (bi.wrapping_mul(16).wrapping_add(ri & 0xF)) as usize; + data[pg_off + tidx] = ct; + planted += 1; + } + } + } + + /// Review regression: a near-plaintext `.text` must NOT be dd8-decrypted. + /// dd8 XORs 255 positions per page with pseudo-random bytes, so it + /// manufactures a few `0xCC` for free — under the old bare + /// `best > baseline` test any positive gain was enough to "apply" dd8 and + /// scramble ~1 byte per 16 of a native DLL's already-plaintext code, + /// silently (nothing downstream, including the integrity check, notices). + /// Here the gain is real but small; the floor must still reject it. + #[test] + fn pe32_dd8_skips_text_whose_gain_is_only_noise_sized() { + let text_off: u32 = 0x1000; + let text_size: u32 = 8 * 0x1000; + let mut data = vec![0u8; (text_off + text_size) as usize]; + seed_dd8_int3(&mut data, text_off, &[2, 4, 6], 5); + assert!( + !data.contains(&0xCC), + "fixture must have a zero 0xCC baseline" + ); + assert_eq!( + select_dd8_formula_pe32(&data, text_off, text_size), + None, + "a gain this small is indistinguishable from dd8's own noise" + ); + } + + /// Control for the above: a `.text` whose dd8 pass restores a large amount + /// of int3 padding clears the floor and is decrypted. Same fixture shape, + /// only the amount of restored padding differs. + #[test] + fn pe32_dd8_applies_when_padding_is_restored() { + let text_off: u32 = 0x1000; + let text_size: u32 = 8 * 0x1000; + let mut data = vec![0u8; (text_off + text_size) as usize]; + seed_dd8_int3(&mut data, text_off, &[2, 4, 6], 255); + assert_eq!( + select_dd8_formula_pe32(&data, text_off, text_size), + Some(false), + "encrypted .text must be decrypted with the page+1 formula" + ); + } +} diff --git a/senbei-pe/src/engine/layout/discovery.rs b/senbei-pe/src/engine/layout/discovery.rs new file mode 100644 index 0000000..830b48a --- /dev/null +++ b/senbei-pe/src/engine/layout/discovery.rs @@ -0,0 +1,507 @@ +//! Structural locators for protected PE stages. + +use senbei_crypto::primitives::{get_u32, lfsr_keystream}; + +/// Find the 4-byte v_val that follows the LAST occurrence of `48 EB 01 B9` +/// (REX.W jmp+1; mov ecx,imm32) plus any 0xCC padding. Used to locate +/// stage4's accum2 seed. Works across builds even when API-name anchors are +/// absent. +pub fn find_v_after_pad(data: &[u8], base: u32, len: u32) -> Option { + let start = base as usize; + let end = (base.saturating_add(len)) as usize; + if end > data.len() { + return None; + } + let sig = [0x48u8, 0xEB, 0x01, 0xB9]; + let slice = &data[start..end]; + // last occurrence + let mut last = None; + let mut i = 0usize; + while i + sig.len() <= slice.len() { + if slice[i..i + sig.len()] == sig { + last = Some(i); + } + i += 1; + } + let pos = last?; + // skip CCs after the `48 EB 01 B9` + let mut after = pos + sig.len(); + while after < slice.len() && slice[after] == 0xCC { + after += 1; + } + if after + 4 > slice.len() { + return None; + } + Some((start + after) as u32) +} + +/// Predict the 4 bytes that DecryptData5(va, size) would produce at va+0..va+4 +/// without mutating the buffer. The cipher's per-byte transform depends only +/// on the byte itself and the low 8 bits of (va+i), with no cross-byte state, +/// so each byte can be decrypted in isolation. Used to detect the EP/DD layout +/// offset before committing to the actual call. +pub fn trial_decrypt5_u32(data: &[u8], va: u32) -> u32 { + let mut out = [0u8; 4]; + for i in 0..4u32 { + let b3 = data[(va + i) as usize]; + let b = (va + i) as u8; + let b2 = b.wrapping_add(1); + let b4 = b3.rotate_left(2) ^ b2; + let b5 = b4.rotate_left(2) ^ b; + out[i as usize] = b5.rotate_left(2); + } + u32::from_le_bytes(out) +} + +/// Scan stage4/stage5 for the encrypted custom-decryptor bytecode block. The +/// raw byte at p+95 is used by decrypt_data6 as the iteration count. We trial- +/// decrypt that many bytes with the LFSR keystream and accept the first +/// position where the byte stream parses as a valid opcode sequence ending in +/// 195 (ret). +pub fn find_bytecode_offset(data: &[u8], base: u32, len: u32) -> Option { + let start = base as usize; + let end = (base.saturating_add(len)) as usize; + if end > data.len() { + return None; + } + let mut ks = [0u8; 256]; + lfsr_keystream(&mut ks); + // Scan forward from `start+16` on 16-byte boundaries relative to `start`. + // The bytecode block is positioned a fixed offset into stage4/stage5; the + // lowest parseable candidate is the real one (later ones are coincidental + // parses of trailing filler bytes that happen to map to valid opcodes). + // The enclosing buffer isn't necessarily 16-aligned to its absolute + // address in newer builds, so we anchor the stride to `start`. + let mut p = start + 16; + while p + 96 <= end { + let count = data[p + 95] as usize; + if count >= 8 && p + count <= end { + let mut buf = [0u8; 256]; + let take = count.min(256); + for i in 0..take { + buf[i] = data[p + i] ^ ks[i]; + } + if let Some(nops) = parse_bytecode_check(&buf[..take]) + && nops >= 4 + { + return Some(p as u32); + } + } + p += 16; + } + None +} + +/// Validate bytecode structure without allocating a `Vec` of ops. Returns +/// `Some(non_nop_op_count)` if the byte stream parses successfully as a valid +/// opcode sequence ending in 195 (ret), `None` otherwise. Allows non-trivial +/// bytecode filtering by op count. +pub fn parse_bytecode_check(buf: &[u8]) -> Option { + let mut i = 0usize; + let mut nops: usize = 0; + while i < buf.len() { + let b = buf[i]; + i += 1; + match b { + 4 | 44 | 52 => { + if i >= buf.len() { + return None; + } + i += 1; + nops += 1; + } + 144 => {} + 192 | 254 => { + if i >= buf.len() { + return None; + } + let mb = buf[i]; + i += 1; + let rm = mb & 7; + let mod_ = (mb >> 6) & 3; + let reg = (mb >> 3) & 7; + if mod_ != 3 || rm != 0 { + return None; + } + if reg > 1 { + return None; + } + if b == 192 { + if i >= buf.len() { + return None; + } + i += 1; + } + nops += 1; + } + 195 => return Some(nops), + _ => return None, + } + } + None +} + +/// Locate stage3's v4_val: the last non-zero dword in the buffer, anchored +/// by the `C3 CC CC CC` (ret + 3 int3) immediately before it. +pub fn find_v4_offset(data: &[u8], base: u32, len: u32) -> Option { + let start = base as usize; + let end = (base.saturating_add(len)) as usize; + if end > data.len() || end < start + 4 { + return None; + } + // walk backwards looking for the first non-zero byte + let mut i = end; + while i > start && data[i - 1] == 0 { + i -= 1; + } + if i < start + 4 { + return None; + } + // v_val occupies the 4 bytes ending at i (rounded up to dword boundary) + let v_end = i; + let v_start = ((v_end + 3) & !3).saturating_sub(4); + // require that the 4 bytes preceding v_val match `C3 CC CC CC` + if v_start < start + 4 || data[v_start - 4..v_start] != [0xC3, 0xCC, 0xCC, 0xCC] { + return None; + } + Some(v_start as u32) +} + +/// Scan a sub-buffer for an ASCII needle; return its absolute position. +pub fn find_str_pos(data: &[u8], base: u32, len: u32, needle: &[u8]) -> Option { + let start = base as usize; + let end = (base.saturating_add(len)) as usize; + if end > data.len() || needle.is_empty() { + return None; + } + data[start..end] + .windows(needle.len()) + .position(|w| w == needle) + .map(|rel| (start + rel) as u32) +} + +pub fn get_string_to_null(data: &[u8], offset: u32) -> String { + let start = offset as usize; + if start >= data.len() { + return String::new(); + } + // Bounded: an unterminated run must never walk off the end of the buffer + // (panic) or scan unboundedly into unrelated data. + let limit = start.saturating_add(4096).min(data.len()); + let mut i = start; + while i < limit && data[i] != 0 { + i += 1; + } + String::from_utf8_lossy(&data[start..i]).into_owned() +} + +/// Read a PE section-name field: exactly 8 bytes, NOT necessarily +/// NUL-terminated (a full-width name like `.textbss` has no NUL at all). +/// Returns the name with trailing NULs stripped. Using `get_string_to_null` +/// here would run past the field into the VirtualSize/VirtualAddress dwords. +pub fn section_name(data: &[u8], offset: u32) -> String { + let start = offset as usize; + let Some(field) = data.get(start..start + 8) else { + return String::new(); + }; + let end = field.iter().position(|&b| b == 0).unwrap_or(8); + String::from_utf8_lossy(&field[..end]).into_owned() +} + +// --------------------------------------------------------------------------- +// PE32 (32-bit) helpers +// --------------------------------------------------------------------------- + +/// PE32 shell-table locator. Walks the shell region (`info[6]`) for a dword +/// equal to `info[6]` followed by a plausible shell size, returning the table +/// base (`candidate = off - 0x88`) when `candidate+0x58` holds a valid pointer. +pub fn find_tbl_pe32(data: &[u8], info: &[u32; 8]) -> Option { + let shell = info[6]; + if (data.len() as u64) < 0x100 { + return None; + } + let hi = (shell as u64) + .saturating_add(0x3000) + .min(data.len() as u64 - 0x100) as u32; + let mut off = shell; + while off < hi { + if off as usize + 8 <= data.len() { + let candidate = off.wrapping_sub(0x88); + if candidate >= shell && get_u32(data, off) == info[6] { + let shell_size_val = get_u32(data, off.wrapping_add(4)); + if shell_size_val > 0x1000 && shell_size_val < 0x100000 { + let v58_off = candidate.wrapping_add(0x58); + if (v58_off as usize + 4) <= data.len() { + let v58 = get_u32(data, v58_off); + if v58 > 0 && (v58 as usize) < data.len() { + return Some(candidate); + } + } + } + } + } + off = off.wrapping_add(4); + } + None +} + +/// Locate an LFSR-encrypted bytecode block (decrypt_data6 form) in a region. +/// `start_off` is the byte offset to begin scanning at, `scan_backward` +/// controls direction. Returns the relative offset of the block. Includes full +/// opcode-walk validation of candidate blocks. +pub fn find_lfsr_block( + data: &[u8], + base: u32, + size: u32, + start_off: u32, + scan_backward: bool, +) -> Option { + if size < 96 { + return None; + } + let mut ks = [0u8; 128]; + lfsr_keystream(&mut ks); + let check = |scan_off: u32| -> bool { + let abs_off = base.wrapping_add(scan_off) as usize; + if abs_off + 96 > data.len() { + return false; + } + let sz = data[abs_off + 95] as usize; + if !(10..=95).contains(&sz) { + return false; + } + let mut decoded = [0u8; 95]; + for bi in 0..sz { + decoded[bi] = data[abs_off + bi] ^ ks[bi]; + } + // Full bytecode validation (shared with the stage4/5 locator): every + // opcode must decode with a valid ModR/M and the stream must REACH a + // RET (0xC3) as an opcode. The previous check only required a 0xC3 + // byte *anywhere* in the window and accepted a walk that ran off the + // end without hitting RET — a `0x04 0xC3` (ADD 0xC3) tail passed, so + // coincidental LFSR-shaped garbage was accepted as a decryptor block. + parse_bytecode_check(&decoded[..sz]).is_some() + }; + if scan_backward { + let hi = size - 96; + if hi >= start_off { + let mut scan_off = hi; + loop { + if check(scan_off) { + return Some(scan_off); + } + if scan_off == start_off { + break; + } + scan_off -= 1; + } + } + } else { + let hi = size - 95; + let mut scan_off = start_off; + while scan_off < hi { + if check(scan_off) { + return Some(scan_off); + } + scan_off += 1; + } + } + None +} + +/// Slots discovered in the eighthStage for the marker-less layout. +pub struct EighthSlots { + /// Absolute address of the file-data decryptor LFSR bytecode block. The + /// fileCS chain pointer is derived downstream as `file_lfsr - 0x58`. + pub file_lfsr: u32, + /// Absolute address of the compressedInfo (ptr,size) table pointer slot. + pub compressed_info_ptr: u32, +} + +/// Marker-independent eighthStage slot discovery (PE32+ branch). +/// +/// Newer Crackproof builds (e.g. some native/managed DLLs) omit the +/// `pm\0\0cm\0\0` and `00 00 00 40 01 00 00 00` markers that the older layout's +/// walk3/walk4/walk5 slot derivation relies on. Instead this discovers the +/// slots structurally: +/// * Scan the eighthStage for every LFSR (decrypt_data6) bytecode block. +/// * The file decryptor is the LFSR block whose `fileCS = lfsr - 0x58` holds +/// a pointer sitting just past `info[3]` (smallest positive distance). +/// * `compressedInfo` is the pointer slot whose 16-byte target, after a +/// trial `decrypt_data5`, parses as a plausible (src,sSize,dst,dSize) +/// descriptor. +/// +/// Returns `None` if no plausible file LFSR is found. `eighth_start`/`eighth_dsz` +/// bound the search region; `info3` is `info[3]`; `compress_data_offset` is +/// `(!u32(file_data,0x1080)) + 0x1000`; `file_data_len` is the protected file +/// length. +#[allow(clippy::too_many_arguments)] +pub fn discover_eighth_slots( + data: &[u8], + eighth_start: u32, + eighth_dsz: u32, + info3: u32, + compress_data_offset: u32, + file_data_len: u32, +) -> Option { + // Collect all LFSR candidates (forward scan). + // + // Advance by 1 after each hit, NOT by 96. A false-positive LFSR match can sit + // just before the real file-decryptor block (observed on an il2cpp game + // assembly build, 2026-07-13: junk at rel=0x31C1, real block at 0x3210). + // Stepping by the LFSR body size then skips the real block and discovery + // fails. Byte-stepping is cheap: eighthStage is only a few KB. + let mut all_lfsrs: Vec = Vec::new(); + let mut scan_off: u32 = 0; + while scan_off + 95 < eighth_dsz { + match find_lfsr_block(data, eighth_start, eighth_dsz, scan_off, false) { + Some(found) => { + all_lfsrs.push(found); + scan_off = found + 1; + } + None => break, + } + } + + // Pick the file LFSR: prefer the candidate whose fileCS pointer sits the + // smallest positive distance past info[3]. + let mut off_file_lfsr: Option = None; + let mut best_dist: Option = None; + for &lfsr_off in &all_lfsrs { + if lfsr_off < 0x58 { + continue; + } + let cs_off = lfsr_off - 0x58; + let cs_val = get_u32(data, eighth_start.wrapping_add(cs_off)); + if !(0x1000 < cs_val && (cs_val as usize) < data.len()) { + continue; + } + if cs_val < info3 { + continue; + } + let dist = cs_val - info3; + if best_dist.is_none_or(|b| dist < b) { + best_dist = Some(dist); + off_file_lfsr = Some(lfsr_off); + } + } + // Fallback: last LFSR with any in-image fileCS pointer. + if off_file_lfsr.is_none() { + for &lfsr_off in all_lfsrs.iter().rev() { + if lfsr_off < 0x58 { + continue; + } + let cs_val = get_u32(data, eighth_start.wrapping_add(lfsr_off - 0x58)); + if 0x1000 < cs_val && (cs_val as usize) < data.len() { + off_file_lfsr = Some(lfsr_off); + break; + } + } + } + let off_file_lfsr = off_file_lfsr?; + let off_file_cs = off_file_lfsr - 0x58; + + // Trial-decrypt to find compressedInfo: the pointer slot in the data area + // (between fileCS region start and the LFSR) whose target parses as a valid + // (src,sSize,dst,dSize) descriptor after a transient decrypt_data5. + let scan_from = off_file_lfsr.saturating_sub(0x400); + let mut off_compressed_info: Option = None; + let mut doff = scan_from; + while doff < off_file_lfsr { + if doff == off_file_cs { + doff += 4; + continue; + } + let ptr_val = get_u32(data, eighth_start.wrapping_add(doff)); + if !(0x1000 < ptr_val && (ptr_val as usize) < data.len().saturating_sub(16)) { + doff += 4; + continue; + } + // Predict decrypt_data5(ptr_val, 16) without mutating: each dword is + // position-keyed and independent, so trial_decrypt5_u32 per dword. + let src2 = trial_decrypt5_u32(data, ptr_val); + let s_sz2 = trial_decrypt5_u32(data, ptr_val + 4); + let dst2 = trial_decrypt5_u32(data, ptr_val + 8); + let d_sz2 = trial_decrypt5_u32(data, ptr_val + 12); + let src_file_off = src2.wrapping_add(compress_data_offset); + let valid = s_sz2 > 0 + && s_sz2 < 0x200000 + && (src_file_off as u64 + s_sz2 as u64) <= file_data_len as u64 + && dst2 >= 0x1000 + && (dst2 as u64 + d_sz2 as u64) <= data.len() as u64 + && d_sz2 >= s_sz2 + && d_sz2 < 0x200000; + if valid { + off_compressed_info = Some(doff); + break; + } + doff += 4; + } + let off_compressed_info = off_compressed_info?; + + Some(EighthSlots { + file_lfsr: eighth_start.wrapping_add(off_file_lfsr), + compressed_info_ptr: eighth_start.wrapping_add(off_compressed_info), + }) +} + +#[cfg(test)] +mod tests { + use super::*; + + /// Task 4.1 regression: build a synthetic buffer whose valid bytecode block + /// sits PAST `len` but within `len*2`. Assert that the smaller window misses + /// it and the doubled window finds it. + #[test] + fn bytecode_locate_double_window_retry() { + // We place the block at offset (base + len + 16) which is inside + // the len*2 window but outside the len window. + let base: u32 = 0; + let len: u32 = 256; + // Block sits at base + len + 16 = 272, aligned to 16. + let block_pos: usize = (base + len + 16) as usize; // 272 + + // The buffer must be large enough for the block (block_pos + 96 bytes). + let buf_len = block_pos + 256; + let mut buf = vec![0u8; buf_len]; + + // Build a valid plaintext op stream: + // [4, 0, 4, 0, 4, 0, 4, 0, 195] (4 ADD-AL ops then RET) + // Padded to 10 bytes total; count >= 8. + let count: usize = 10; + let mut plain = [0u8; 256]; + plain[0] = 4; + plain[1] = 0; + plain[2] = 4; + plain[3] = 0; + plain[4] = 4; + plain[5] = 0; + plain[6] = 4; + plain[7] = 0; + plain[8] = 195; // ret + + // Compute the LFSR keystream and XOR the first `count` bytes to get the + // encrypted representation that the scanner would decrypt back. + let mut ks = [0u8; 256]; + lfsr_keystream(&mut ks); + for i in 0..count { + buf[block_pos + i] = plain[i] ^ ks[i]; + } + // Raw count byte at block_pos+95 (outside the XOR range since count=10 < 95). + buf[block_pos + 95] = count as u8; + + // Verify our construction: find_bytecode_offset with len should NOT find it. + assert_eq!( + find_bytecode_offset(&buf, base, len), + None, + "smaller window should not find the block" + ); + + // The doubled window should find it at block_pos. + assert_eq!( + find_bytecode_offset(&buf, base, len.saturating_mul(2)), + Some(block_pos as u32), + "doubled window should locate the block" + ); + } +} diff --git a/senbei-pe/src/engine/layout/image.rs b/senbei-pe/src/engine/layout/image.rs new file mode 100644 index 0000000..04065cd --- /dev/null +++ b/senbei-pe/src/engine/layout/image.rs @@ -0,0 +1,481 @@ +//! PE import reconstruction and memory-image compaction. + +use senbei_crypto::primitives::{get_u16, get_u32, write_u16, write_u32}; + +use super::super::MAX_IMAGE_SIZE; + +/// Read a NUL-terminated byte string starting at `off`, bounded to 512 bytes. +/// Returns the raw bytes up to the terminator (excluding it). +fn read_cstr_bounded(data: &[u8], off: u32) -> Vec { + let start = off as usize; + if start >= data.len() { + return Vec::new(); + } + let limit = (start + 512).min(data.len()); + let mut end = start; + while end < limit && data[end] != 0 { + end += 1; + } + data[start..end].to_vec() +} + +fn align_up_u32(value: u32, alignment: u32) -> u32 { + ((value.wrapping_add(alignment - 1)) / alignment).wrapping_mul(alignment) +} + +fn align_up_u64(value: u64, alignment: u64) -> u64 { + value.div_ceil(alignment) * alignment +} + +#[derive(Clone)] +enum ImportFunc { + Ordinal(u32), + Name(u16, Vec), +} + +struct ImportDesc { + time_date: u32, + fwd_chain: u32, + dll_name: Vec, + iat_rva: u32, + functions: Vec, +} + +/// Return true when PE32 imports already sit in the original `.idata` layout +/// (so no relocation to `.kmiat` is needed). May write the IAT data directory +/// (pe+0xD8). +pub fn pe32_imports_already_match_idata_layout(data: &mut [u8], pe_header: u32) -> bool { + let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; + let sec_table = pe_header.wrapping_add(24).wrapping_add(opt_hdr_size); + let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; + let import_rva = get_u32(data, pe_header.wrapping_add(0x80)); + let import_size = get_u32(data, pe_header.wrapping_add(0x84)); + let len = data.len() as u32; + if !(import_rva > 0 && import_size > 0) { + return false; + } + for idx in 0..num_sections { + let sec_off = sec_table.wrapping_add(idx * 40); + if (sec_off as usize + 40) > data.len() { + return false; + } + if &data[sec_off as usize..sec_off as usize + 6] != b".idata" { + continue; + } + let sec_va = get_u32(data, sec_off.wrapping_add(12)); + let sec_size = + get_u32(data, sec_off.wrapping_add(8)).max(get_u32(data, sec_off.wrapping_add(16))); + let sec_end = sec_va.wrapping_add(sec_size); + if !(sec_va <= import_rva + && import_rva < sec_end + && import_rva.wrapping_add(import_size) <= sec_end) + { + continue; + } + let first_oft = get_u32(data, import_rva); + let first_name = get_u32(data, import_rva.wrapping_add(12)); + let first_iat = get_u32(data, import_rva.wrapping_add(16)); + if !(sec_va <= first_oft + && first_oft < sec_end + && sec_va <= first_iat + && first_iat < sec_end) + { + return false; + } + if !(0x1000 < first_name && first_name < len) { + return false; + } + let dll_name = read_cstr_bounded(data, first_name); + let lower: Vec = dll_name.iter().map(|b| b.to_ascii_lowercase()).collect(); + if !lower.ends_with(b".dll") { + return false; + } + let mut iat_min = first_iat; + let mut iat_max = first_iat; + let mut idt_pos = import_rva; + while idt_pos.wrapping_add(20) <= len { + let oft_rva = get_u32(data, idt_pos); + let name_rva = get_u32(data, idt_pos.wrapping_add(12)); + let iat_rva = get_u32(data, idt_pos.wrapping_add(16)); + if oft_rva == 0 && name_rva == 0 && iat_rva == 0 { + break; + } + if !(sec_va <= oft_rva && oft_rva < sec_end && sec_va <= iat_rva && iat_rva < sec_end) { + return false; + } + let mut thunk = iat_rva; + while thunk.wrapping_add(4) <= sec_end { + let tv = get_u32(data, thunk); + thunk = thunk.wrapping_add(4); + if tv == 0 { + break; + } + } + iat_min = iat_min.min(iat_rva); + iat_max = iat_max.max(thunk); + idt_pos = idt_pos.wrapping_add(20); + } + if iat_max > iat_min { + write_u32(data, pe_header.wrapping_add(0xD8), iat_min); + write_u32(data, pe_header.wrapping_add(0xDC), iat_max - iat_min); + } + return true; + } + false +} + +/// Rebuild PE32 import metadata (descriptors, lookup tables, names) into the +/// last section as `.kmiat`, leaving the loader-written IAT in place. Mutates +/// `data` (may grow it). +pub fn move_pe32_imports_to_kmiat(data: &mut Vec, pe_header: u32) { + const SECTION_SIZE: u32 = 0x7000; + let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; + let opt_hdr = pe_header.wrapping_add(24); + let sec_table = opt_hdr.wrapping_add(opt_hdr_size); + let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; + if num_sections == 0 { + return; + } + let import_rva = get_u32(data, pe_header.wrapping_add(0x80)); + let import_size = get_u32(data, pe_header.wrapping_add(0x84)); + let len = data.len() as u32; + if !(0x1000 < import_rva && import_rva < len && import_size > 0 && import_size < SECTION_SIZE) { + return; + } + + let mut descriptors: Vec = Vec::new(); + let mut idt_pos = import_rva; + while idt_pos.wrapping_add(20) <= len { + let oft_rva = get_u32(data, idt_pos); + let time_date = get_u32(data, idt_pos.wrapping_add(4)); + let fwd_chain = get_u32(data, idt_pos.wrapping_add(8)); + let name_rva = get_u32(data, idt_pos.wrapping_add(12)); + let iat_rva = get_u32(data, idt_pos.wrapping_add(16)); + if oft_rva == 0 && name_rva == 0 && iat_rva == 0 { + break; + } + if !(0x1000 < name_rva && name_rva < len) { + break; + } + let dll_name = read_cstr_bounded(data, name_rva); + let thunk_rva = if 0x1000 < oft_rva && oft_rva < len { + oft_rva + } else { + iat_rva + }; + let mut functions: Vec = Vec::new(); + let mut thunk_pos = thunk_rva; + while 0x1000 < thunk_pos.wrapping_add(4) && thunk_pos.wrapping_add(4) <= len { + let thunk_val = get_u32(data, thunk_pos); + if thunk_val == 0 { + break; + } + if thunk_val & 0x8000_0000 != 0 { + functions.push(ImportFunc::Ordinal(thunk_val & 0xFFFF)); + } else { + let hint = if thunk_val.wrapping_add(2) <= len { + get_u16(data, thunk_val) + } else { + 0 + }; + let func_name = if thunk_val.wrapping_add(2) < len { + read_cstr_bounded(data, thunk_val.wrapping_add(2)) + } else { + Vec::new() + }; + functions.push(ImportFunc::Name(hint, func_name)); + } + thunk_pos = thunk_pos.wrapping_add(4); + } + descriptors.push(ImportDesc { + time_date, + fwd_chain, + dll_name, + iat_rva, + functions, + }); + idt_pos = idt_pos.wrapping_add(20); + } + if descriptors.is_empty() { + return; + } + + for desc in &mut descriptors { + let lower: Vec = desc + .dll_name + .iter() + .map(|b| b.to_ascii_lowercase()) + .collect(); + if lower.starts_with(b"api-ms-win-crt-") { + desc.dll_name = b"ucrtbase.dll".to_vec(); + } else { + desc.dll_name = lower; + } + } + descriptors.sort_by_key(|d| d.iat_rva); + + let last_sec = sec_table.wrapping_add((num_sections - 1) * 40); + let kmiat_rva = get_u32(data, last_sec.wrapping_add(12)); + // A zero last-section VA means a corrupt section table: building .kmiat at + // RVA 0 would zero the DOS/PE headers and emit a structurally broken image + // with no error. Bail and keep the original import table. + if kmiat_rva == 0 { + return; + } + // Grow the image when .kmiat overruns it, but cap the growth: a corrupt VA + // could otherwise request a multi-gigabyte allocation, which aborts the + // process (uncatchable). Use u64 math so a near-u32::MAX VA cannot wrap the + // end calculation the way the previous wrapping/plain-add mix could. + let kmiat_end = kmiat_rva as u64 + SECTION_SIZE as u64; + if kmiat_end > MAX_IMAGE_SIZE { + return; + } + if kmiat_end > data.len() as u64 { + data.resize(kmiat_end as usize, 0); + } + // Zero the .kmiat region. + for b in &mut data[kmiat_rva as usize..kmiat_end as usize] { + *b = 0; + } + + let idt_size = (descriptors.len() as u32 + 1) * 20; + let oft_start = kmiat_rva; + let mut idt_rva = oft_start; + for desc in &descriptors { + idt_rva = idt_rva.wrapping_add((desc.functions.len() as u32 + 1) * 4); + } + idt_rva = align_up_u32(idt_rva.wrapping_add(0x2C), 4); + + // Size check: compute the final name_pos and bail if it overruns .kmiat. + let mut name_pos_check = idt_rva.wrapping_add(idt_size); + for desc in &descriptors { + name_pos_check = name_pos_check.wrapping_add(desc.dll_name.len() as u32 + 1); + for func in &desc.functions { + if let ImportFunc::Name(_, fname) = func { + name_pos_check = name_pos_check.wrapping_add(2 + fname.len() as u32 + 1); + } + } + } + if name_pos_check > kmiat_rva.wrapping_add(SECTION_SIZE) { + // Section too small; keep existing import table untouched. + return; + } + + let mut oft_pos = oft_start; + let mut name_pos = idt_rva.wrapping_add(idt_size); + for (idx, desc) in descriptors.iter().enumerate() { + let idt_entry = idt_rva.wrapping_add(idx as u32 * 20); + let current_oft = oft_pos; + write_u32(data, idt_entry, current_oft); + write_u32(data, idt_entry.wrapping_add(4), desc.time_date); + write_u32(data, idt_entry.wrapping_add(8), desc.fwd_chain); + let dll_name_pos = name_pos; + write_u32(data, idt_entry.wrapping_add(12), dll_name_pos); + write_u32(data, idt_entry.wrapping_add(16), desc.iat_rva); + + let dnp = dll_name_pos as usize; + data[dnp..dnp + desc.dll_name.len()].copy_from_slice(&desc.dll_name); + data[dnp + desc.dll_name.len()] = 0; + name_pos = name_pos.wrapping_add(desc.dll_name.len() as u32 + 1); + + for func in &desc.functions { + match func { + ImportFunc::Ordinal(ord) => { + write_u32(data, oft_pos, 0x8000_0000 | ord); + } + ImportFunc::Name(hint, fname) => { + let hint_name_rva = name_pos; + write_u32(data, oft_pos, hint_name_rva); + write_u16(data, hint_name_rva, *hint as u32); + let fp = (hint_name_rva + 2) as usize; + data[fp..fp + fname.len()].copy_from_slice(fname); + data[fp + fname.len()] = 0; + name_pos = name_pos.wrapping_add(2 + fname.len() as u32 + 1); + } + } + oft_pos = oft_pos.wrapping_add(4); + } + write_u32(data, oft_pos, 0); + oft_pos = oft_pos.wrapping_add(4); + } + // Null-terminator IDT entry (20 zero bytes) after the last descriptor. + let term = idt_rva.wrapping_add(descriptors.len() as u32 * 20) as usize; + for b in &mut data[term..term + 20] { + *b = 0; + } + + let ls = last_sec as usize; + data[ls..ls + 8].copy_from_slice(b".kmiat\x00\x00"); + write_u32(data, last_sec.wrapping_add(8), SECTION_SIZE); + write_u32(data, last_sec.wrapping_add(16), SECTION_SIZE); + write_u32(data, last_sec.wrapping_add(36), 0xE000_0060); + write_u32(data, pe_header.wrapping_add(0x80), idt_rva); + write_u32(data, pe_header.wrapping_add(0x84), idt_size); + write_u32( + data, + pe_header.wrapping_add(80), + kmiat_rva.wrapping_add(SECTION_SIZE), + ); +} + +/// Convert the unpacked RVA-addressed image back to a compact PE file layout +/// (headers at 0x400, sections packed consecutively, FileAlignment 0x200). +/// Returns `None` if the accumulated output size wraps or exceeds +/// [`MAX_IMAGE_SIZE`]: the final allocation is sized from header-derived +/// section data, and an uncapped `vec![0; n]` from a corrupt header would abort +/// the process (which `catch_unpack` cannot trap). +pub fn compact_memory_image_to_pe(data: &[u8], pe_header: u32) -> Option> { + const FILE_ALIGNMENT: u32 = 0x200; + const HEADER_SIZE: u32 = 0x400; + let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; + let opt_hdr = pe_header.wrapping_add(24); + let sec_table = opt_hdr.wrapping_add(opt_hdr_size); + let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; + + struct SecLayout { + sec_off: u32, + va: u32, + vsize: u32, + raw_ptr: u32, + raw_size: u32, + } + + let mut raw_cursor: u64 = HEADER_SIZE as u64; + let mut raw_layout: Vec = Vec::new(); + for idx in 0..num_sections { + let sec_off = sec_table.wrapping_add(idx * 40); + let vsize = get_u32(data, sec_off.wrapping_add(8)); + let va = get_u32(data, sec_off.wrapping_add(12)); + let sd_start = va as usize; + let sd_end = if (va.wrapping_add(vsize) as usize) <= data.len() { + va.wrapping_add(vsize) as usize + } else { + data.len() + }; + let section_data: &[u8] = if sd_start <= sd_end && sd_start <= data.len() { + &data[sd_start..sd_end] + } else { + &[] + }; + + let mut last_nonzero: i64 = -1; + for pos in (0..section_data.len()).rev() { + if section_data[pos] != 0 { + last_nonzero = pos as i64; + break; + } + } + let meaningful = if last_nonzero >= 0 { + (last_nonzero + 1) as u32 + } else { + 0 + }; + let mut raw_size = if meaningful != 0 { + align_up_u32(meaningful, FILE_ALIGNMENT) + } else { + 0 + }; + if vsize != 0 && raw_size == 0 { + raw_size = FILE_ALIGNMENT; + } + raw_size = raw_size.min(align_up_u32(section_data.len() as u32, FILE_ALIGNMENT)); + + let raw_ptr = if raw_size != 0 { raw_cursor as u32 } else { 0 }; + raw_layout.push(SecLayout { + sec_off, + va, + vsize, + raw_ptr, + raw_size, + }); + if raw_size != 0 { + // Accumulate in u64 and cap: section sizes are header-derived, and + // a corrupt table could otherwise wrap raw_cursor (small alloc, + // huge recorded raw_ptrs → OOB panic) or request an abort-sized + // allocation. + raw_cursor = align_up_u64(raw_cursor + raw_size as u64, FILE_ALIGNMENT as u64); + if raw_cursor > MAX_IMAGE_SIZE { + return None; + } + } + } + + let mut compact = vec![0u8; raw_cursor as usize]; + let hdr_copy = (HEADER_SIZE as usize).min(data.len()); + compact[..hdr_copy].copy_from_slice(&data[..hdr_copy]); + write_u32(&mut compact, opt_hdr.wrapping_add(36), FILE_ALIGNMENT); + write_u32(&mut compact, opt_hdr.wrapping_add(60), HEADER_SIZE); + + for sl in &raw_layout { + write_u32(&mut compact, sl.sec_off.wrapping_add(16), sl.raw_size); + write_u32(&mut compact, sl.sec_off.wrapping_add(20), sl.raw_ptr); + if sl.raw_size != 0 { + let sd_start = sl.va as usize; + let sd_end = if (sl.va.wrapping_add(sl.vsize) as usize) <= data.len() { + sl.va.wrapping_add(sl.vsize) as usize + } else { + data.len() + }; + let section_data: &[u8] = if sd_start <= sd_end { + &data[sd_start..sd_end] + } else { + &[] + }; + let copy_size = (sl.raw_size as usize).min(section_data.len()); + let rp = sl.raw_ptr as usize; + compact[rp..rp + copy_size].copy_from_slice(§ion_data[..copy_size]); + } + } + Some(compact) +} + +#[cfg(test)] +mod tests { + use super::*; + + /// Review regression: a zero last-section VA (corrupt section table) must + /// bail instead of building .kmiat at RVA 0 — the old code zeroed + /// `[0, 0x7000)`, wiping the DOS/PE headers, and returned the broken image + /// as a success. A near-2 GiB VA must likewise refuse to grow the image + /// past [`MAX_IMAGE_SIZE`]. + #[test] + fn kmiat_bogus_section_va_bails_without_wiping_headers() { + for last_sec_va in [0u32, 0x5000_0000] { + let pe: u32 = 0x80; + let mut data = vec![0xAAu8; 0x8000]; + // COFF header: 1 section, optional header size 0xE0 (PE32). + write_u16(&mut data, pe + 6, 1); + write_u16(&mut data, pe + 20, 0xE0); + // Import directory at pe+0x80: one descriptor + null terminator. + write_u32(&mut data, pe + 0x80, 0x1100); + write_u32(&mut data, pe + 0x84, 0x28); + write_u32(&mut data, 0x1100, 0x1200); // OFT rva + write_u32(&mut data, 0x1100 + 12, 0x1300); // name rva + write_u32(&mut data, 0x1100 + 16, 0x1400); // IAT rva + for b in &mut data[0x1100 + 20..0x1100 + 40] { + *b = 0; // null terminator descriptor + } + data[0x1300..0x1300 + 13].copy_from_slice(b"KERNEL32.dll\0"); + write_u32(&mut data, 0x1200, 0x1500); // thunk -> hint/name + write_u32(&mut data, 0x1204, 0); // thunk terminator + data[0x1500..0x1502].copy_from_slice(&0u16.to_le_bytes()); + data[0x1502..0x1502 + 12].copy_from_slice(b"ExitProcess\0"); + // Section table at pe+24+0xE0 = 0x178; VA field at +12. + write_u32(&mut data, 0x178 + 12, last_sec_va); + + let head_before: Vec = data[..0x400].to_vec(); + let len_before = data.len(); + move_pe32_imports_to_kmiat(&mut data, pe); + assert_eq!( + data.len(), + len_before, + "VA 0x{last_sec_va:08X}: image must not grow" + ); + assert_eq!( + &data[..0x400], + &head_before[..], + "VA 0x{last_sec_va:08X}: headers must be untouched" + ); + } + } +} diff --git a/src/unpacker/mod.rs b/senbei-pe/src/engine/mod.rs similarity index 90% rename from src/unpacker/mod.rs rename to senbei-pe/src/engine/mod.rs index 35447a2..ad0ec6c 100644 --- a/src/unpacker/mod.rs +++ b/senbei-pe/src/engine/mod.rs @@ -1,30 +1,28 @@ -//! Pure, panic-free Crackproof unpacker core. No file I/O lives here. +//! PE detection, unpacking, and structural validation. -mod bytecode; -mod crc32; pub mod dll; +mod error; pub mod exe; pub mod integrity; +mod layout; pub(crate) mod parallel; -pub(crate) mod primitives; -mod tables; +use senbei_crypto::primitives; use std::cell::RefCell; use std::sync::{Arc, Mutex}; pub use dll::{unpack_dll, unpack_dll_v}; -pub use exe::{ - BufferOperation, BytecodeStage, DecompressionFailure, DecompressionStage, DescriptorTable, - SectionPipeline, UnpackError, unpack as unpack_exe, unpack_v as unpack_exe_v, -}; +pub use error::*; +pub use exe::{unpack as unpack_exe, unpack_v as unpack_exe_v}; pub use integrity::{IntegrityReport, check as check_integrity}; +pub use parallel::thread_cap; /// Maximum plausible PE `SizeOfImage` we are willing to allocate a zero buffer /// for. Guards against a corrupt/crafted header requesting a multi-gigabyte /// (or, as a sign-extended negative `i32`, multi-exabyte) allocation, which /// would abort the process — an abort that `catch_unpack` below cannot trap. /// Real protected binaries are far below this. -pub(crate) const MAX_IMAGE_SIZE: u64 = 1 << 30; // 1 GiB +pub(crate) const MAX_IMAGE_SIZE: u64 = senbei_crypto::MAX_IMAGE_SIZE; #[derive(Clone)] pub(crate) struct PanicCapture(Arc>>); @@ -56,7 +54,6 @@ impl PanicCapture { Self(Arc::new(Mutex::new(None))) } - #[cfg(not(target_arch = "wasm32"))] fn record(&self, info: &std::panic::PanicHookInfo<'_>) { let location = info.location(); let details = PanicDetails { @@ -123,7 +120,6 @@ fn panic_message(payload: &(dyn std::any::Any + Send)) -> String { } } -#[cfg(not(target_arch = "wasm32"))] fn install_panic_capture_hook() { static INSTALL: std::sync::Once = std::sync::Once::new(); INSTALL.call_once(|| { @@ -142,9 +138,6 @@ fn install_panic_capture_hook() { }); } -#[cfg(target_arch = "wasm32")] -fn install_panic_capture_hook() {} - pub(crate) fn current_panic_capture() -> Option { ACTIVE_PANIC_CAPTURE.with(|slot| slot.borrow().clone()) } @@ -166,8 +159,6 @@ pub(crate) fn catch_unpack(f: F) -> Result, UnpackError> where F: FnOnce() -> Result, UnpackError>, { - // Hook capture is skipped on wasm: the prebuilt std cannot unwind there, - // so a panic traps immediately. The Web Worker boundary reports that trap. install_panic_capture_hook(); let capture = PanicCapture::new(); let r = with_panic_capture(Some(capture.clone()), || { @@ -213,11 +204,8 @@ fn key_table(input: &[u8]) -> Option<[u32; 8]> { if input.len() < 4128 { return None; } - // Validate PE signature. `checked_add`, not `+`: `usize` is 32-bit on - // wasm32, where an `e_lfanew` of 0xFFFF_FFFC..=0xFFFF_FFFF wraps the bound - // check, and the slice below then panics with start > end. `detect` runs on - // the folder-scan threads and (in the web app) on the main thread outside - // the disposable-worker isolation, so it must not panic on any input. + // Validate the PE signature with checked arithmetic so a crafted offset + // cannot wrap the bounds check on a narrower target. let e_lfanew = primitives::get_u32(input, 0x3C); let pe_start = e_lfanew as usize; if pe_start.checked_add(4).is_none_or(|end| end > input.len()) { @@ -349,7 +337,6 @@ pub fn unpack_auto_v(input: &[u8], verbose: bool) -> Result<(Kind, Vec), Unp mod tests { use super::*; - #[cfg(not(target_arch = "wasm32"))] #[test] fn caught_panic_reports_location_and_message() { let error = catch_unpack(|| -> Result, UnpackError> { @@ -366,12 +353,14 @@ mod tests { panic!("unexpected error: {error}"); }; assert_eq!(message, "test panic"); - assert!(file.ends_with("src/unpacker/mod.rs") || file.ends_with("src\\unpacker\\mod.rs")); + assert!( + file.ends_with("senbei-pe/src/engine/mod.rs") + || file.ends_with("senbei-pe\\src\\engine\\mod.rs") + ); assert!(line > 0); assert!(column > 0); } - #[cfg(not(target_arch = "wasm32"))] #[test] fn worker_panic_keeps_the_worker_source_location() { let error = catch_unpack(|| -> Result, UnpackError> { @@ -396,7 +385,10 @@ mod tests { panic!("unexpected error: {error}"); }; assert_eq!(message, "worker panic"); - assert!(file.ends_with("src/unpacker/mod.rs") || file.ends_with("src\\unpacker\\mod.rs")); + assert!( + file.ends_with("senbei-pe/src/engine/mod.rs") + || file.ends_with("senbei-pe\\src\\engine\\mod.rs") + ); assert!(line > 0); assert!(column > 0); } diff --git a/src/unpacker/parallel.rs b/senbei-pe/src/engine/parallel.rs similarity index 99% rename from src/unpacker/parallel.rs rename to senbei-pe/src/engine/parallel.rs index 5f20d3d..fad1132 100644 --- a/src/unpacker/parallel.rs +++ b/senbei-pe/src/engine/parallel.rs @@ -21,7 +21,7 @@ use std::sync::atomic::{AtomicBool, Ordering}; /// Worker-thread cap. `SENBEI_THREADS` overrides it (`1` forces the sequential /// path); otherwise the host's available parallelism; otherwise 1. -pub(crate) fn thread_cap() -> usize { +pub fn thread_cap() -> usize { if let Ok(v) = std::env::var("SENBEI_THREADS") && let Ok(n) = v.trim().parse::() && n >= 1 diff --git a/senbei-pe/src/lib.rs b/senbei-pe/src/lib.rs new file mode 100644 index 0000000..5db4e9a --- /dev/null +++ b/senbei-pe/src/lib.rs @@ -0,0 +1,5 @@ +//! PE detection, unpacking, and structural validation. + +mod engine; + +pub use engine::*; diff --git a/src/unpacker/primitives.rs b/src/unpacker/primitives.rs deleted file mode 100644 index 886b159..0000000 --- a/src/unpacker/primitives.rs +++ /dev/null @@ -1,2394 +0,0 @@ -//! Shared crypto primitives and helper utilities. -//! -//! All functions here are `pub(crate)` so that both the EXE unpacker (`exe.rs`) -//! and the future DLL unpacker (`dll.rs`) can call them without duplication. -//! Each free function is self-contained: it takes the relevant byte buffer(s) -//! and parameters explicitly, with no coupling to the EXE `Unpacker` struct. - -use super::bytecode::{Op, OpsLut}; -use super::crc32; -use super::tables::{COLUMMIX1, COLUMMIX2, COLUMMIX3, COLUMMIX4, SBOX}; -use std::cell::RefCell; - -thread_local! { - /// Reusable scratch for `decompress`. A single unpack runs `decompress` - /// hundreds of times over small blocks; reusing one growable buffer avoids a - /// fresh allocation each call. Thread-local, so it stays correct (one buffer - /// per worker) under the parallel block fan-out. - static DECOMPRESS_SCRATCH: RefCell> = const { RefCell::new(Vec::new()) }; -} - -// --------------------------------------------------------------------------- -// Byte-order accessors -// --------------------------------------------------------------------------- - -pub(crate) fn get_u16(data: &[u8], offset: u32) -> u16 { - let i = offset as usize; - u16::from_le_bytes([data[i], data[i + 1]]) -} - -pub(crate) fn get_u32(data: &[u8], offset: u32) -> u32 { - let i = offset as usize; - u32::from_le_bytes([data[i], data[i + 1], data[i + 2], data[i + 3]]) -} - -pub(crate) fn get_u64(data: &[u8], offset: u32) -> u64 { - let i = offset as usize; - u64::from_le_bytes([ - data[i], - data[i + 1], - data[i + 2], - data[i + 3], - data[i + 4], - data[i + 5], - data[i + 6], - data[i + 7], - ]) -} - -pub(crate) fn write_u16(data: &mut [u8], offset: u32, value: u32) { - let i = offset as usize; - let v = value as u16; - let b = v.to_le_bytes(); - data[i] = b[0]; - data[i + 1] = b[1]; -} - -pub(crate) fn write_u32(data: &mut [u8], offset: u32, value: u32) { - let i = offset as usize; - let b = value.to_le_bytes(); - data[i] = b[0]; - data[i + 1] = b[1]; - data[i + 2] = b[2]; - data[i + 3] = b[3]; -} - -// --------------------------------------------------------------------------- -// Checked accessors (return Err instead of panicking on OOB) -// --------------------------------------------------------------------------- - -#[allow(dead_code)] -pub(crate) fn try_u32(d: &[u8], off: usize) -> Result { - let end = off - .checked_add(4) - .ok_or(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::Read, - offset: off, - size: 4, - buffer_len: d.len(), - })?; - d.get(off..end) - .map(|s| u32::from_le_bytes(s.try_into().unwrap())) - .ok_or(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::Read, - offset: off, - size: 4, - buffer_len: d.len(), - }) -} - -#[allow(dead_code)] -pub(crate) fn try_i32(d: &[u8], off: usize) -> Result { - try_u32(d, off).map(|v| v as i32) -} - -/// Checked copy with distinct source and destination range errors. -pub(crate) fn try_copy_from_slice( - dst: &mut [u8], - dst_off: usize, - dst_len: usize, - src: &[u8], - src_off: usize, -) -> Result<(), super::UnpackError> { - let dst_end = - dst_off - .checked_add(dst_len) - .ok_or(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::CopyDestination, - offset: dst_off, - size: dst_len, - buffer_len: dst.len(), - })?; - let src_end = - src_off - .checked_add(dst_len) - .ok_or(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::CopySource, - offset: src_off, - size: dst_len, - buffer_len: src.len(), - })?; - if dst_end > dst.len() { - return Err(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::CopyDestination, - offset: dst_off, - size: dst_len, - buffer_len: dst.len(), - }); - } - if src_end > src.len() { - return Err(super::UnpackError::BufferRangeOutOfBounds { - operation: super::BufferOperation::CopySource, - offset: src_off, - size: dst_len, - buffer_len: src.len(), - }); - } - dst[dst_off..dst_end].copy_from_slice(&src[src_off..src_end]); - Ok(()) -} - -// --------------------------------------------------------------------------- -// Locator helpers -// --------------------------------------------------------------------------- - -/// Find the 4-byte v_val that follows the LAST occurrence of `48 EB 01 B9` -/// (REX.W jmp+1; mov ecx,imm32) plus any 0xCC padding. Used to locate -/// stage4's accum2 seed. Works across builds even when API-name anchors are -/// absent. -pub(crate) fn find_v_after_pad(data: &[u8], base: u32, len: u32) -> Option { - let start = base as usize; - let end = (base.saturating_add(len)) as usize; - if end > data.len() { - return None; - } - let sig = [0x48u8, 0xEB, 0x01, 0xB9]; - let slice = &data[start..end]; - // last occurrence - let mut last = None; - let mut i = 0usize; - while i + sig.len() <= slice.len() { - if slice[i..i + sig.len()] == sig { - last = Some(i); - } - i += 1; - } - let pos = last?; - // skip CCs after the `48 EB 01 B9` - let mut after = pos + sig.len(); - while after < slice.len() && slice[after] == 0xCC { - after += 1; - } - if after + 4 > slice.len() { - return None; - } - Some((start + after) as u32) -} - -/// Predict the 4 bytes that DecryptData5(va, size) would produce at va+0..va+4 -/// without mutating the buffer. The cipher's per-byte transform depends only -/// on the byte itself and the low 8 bits of (va+i), with no cross-byte state, -/// so each byte can be decrypted in isolation. Used to detect the EP/DD layout -/// offset before committing to the actual call. -pub(crate) fn trial_decrypt5_u32(data: &[u8], va: u32) -> u32 { - let mut out = [0u8; 4]; - for i in 0..4u32 { - let b3 = data[(va + i) as usize]; - let b = (va + i) as u8; - let b2 = b.wrapping_add(1); - let b4 = b3.rotate_left(2) ^ b2; - let b5 = b4.rotate_left(2) ^ b; - out[i as usize] = b5.rotate_left(2); - } - u32::from_le_bytes(out) -} - -/// Reproduce the LFSR keystream that decrypt_data6 XORs in. Used to -/// trial-decrypt candidate bytecode positions without mutating the buffer. -pub(crate) fn lfsr_keystream(out: &mut [u8]) { - let mut state: u32 = 1; - for byte in out.iter_mut() { - let mut b: u8 = 0; - for k in 0..8u32 { - b |= ((state & 1) << k) as u8; - state <<= 1; - if state & 0x8000 != 0 { - state ^= 0x8003; - } - } - *byte = b; - } -} - -/// Scan stage4/stage5 for the encrypted custom-decryptor bytecode block. The -/// raw byte at p+95 is used by decrypt_data6 as the iteration count. We trial- -/// decrypt that many bytes with the LFSR keystream and accept the first -/// position where the byte stream parses as a valid opcode sequence ending in -/// 195 (ret). -pub(crate) fn find_bytecode_offset(data: &[u8], base: u32, len: u32) -> Option { - let start = base as usize; - let end = (base.saturating_add(len)) as usize; - if end > data.len() { - return None; - } - let mut ks = [0u8; 256]; - lfsr_keystream(&mut ks); - // Scan forward from `start+16` on 16-byte boundaries relative to `start`. - // The bytecode block is positioned a fixed offset into stage4/stage5; the - // lowest parseable candidate is the real one (later ones are coincidental - // parses of trailing filler bytes that happen to map to valid opcodes). - // The enclosing buffer isn't necessarily 16-aligned to its absolute - // address in newer builds, so we anchor the stride to `start`. - let mut p = start + 16; - while p + 96 <= end { - let count = data[p + 95] as usize; - if count >= 8 && p + count <= end { - let mut buf = [0u8; 256]; - let take = count.min(256); - for i in 0..take { - buf[i] = data[p + i] ^ ks[i]; - } - if let Some(nops) = parse_bytecode_check(&buf[..take]) - && nops >= 4 - { - return Some(p as u32); - } - } - p += 16; - } - None -} - -/// Validate bytecode structure without allocating a `Vec` of ops. Returns -/// `Some(non_nop_op_count)` if the byte stream parses successfully as a valid -/// opcode sequence ending in 195 (ret), `None` otherwise. Allows non-trivial -/// bytecode filtering by op count. -pub(crate) fn parse_bytecode_check(buf: &[u8]) -> Option { - let mut i = 0usize; - let mut nops: usize = 0; - while i < buf.len() { - let b = buf[i]; - i += 1; - match b { - 4 | 44 | 52 => { - if i >= buf.len() { - return None; - } - i += 1; - nops += 1; - } - 144 => {} - 192 | 254 => { - if i >= buf.len() { - return None; - } - let mb = buf[i]; - i += 1; - let rm = mb & 7; - let mod_ = (mb >> 6) & 3; - let reg = (mb >> 3) & 7; - if mod_ != 3 || rm != 0 { - return None; - } - if reg > 1 { - return None; - } - if b == 192 { - if i >= buf.len() { - return None; - } - i += 1; - } - nops += 1; - } - 195 => return Some(nops), - _ => return None, - } - } - None -} - -/// Locate stage3's v4_val: the last non-zero dword in the buffer, anchored -/// by the `C3 CC CC CC` (ret + 3 int3) immediately before it. -pub(crate) fn find_v4_offset(data: &[u8], base: u32, len: u32) -> Option { - let start = base as usize; - let end = (base.saturating_add(len)) as usize; - if end > data.len() || end < start + 4 { - return None; - } - // walk backwards looking for the first non-zero byte - let mut i = end; - while i > start && data[i - 1] == 0 { - i -= 1; - } - if i < start + 4 { - return None; - } - // v_val occupies the 4 bytes ending at i (rounded up to dword boundary) - let v_end = i; - let v_start = ((v_end + 3) & !3).saturating_sub(4); - // require that the 4 bytes preceding v_val match `C3 CC CC CC` - if v_start < start + 4 || data[v_start - 4..v_start] != [0xC3, 0xCC, 0xCC, 0xCC] { - return None; - } - Some(v_start as u32) -} - -/// Scan a sub-buffer for an ASCII needle; return its absolute position. -pub(crate) fn find_str_pos(data: &[u8], base: u32, len: u32, needle: &[u8]) -> Option { - let start = base as usize; - let end = (base.saturating_add(len)) as usize; - if end > data.len() || needle.is_empty() { - return None; - } - data[start..end] - .windows(needle.len()) - .position(|w| w == needle) - .map(|rel| (start + rel) as u32) -} - -pub(crate) fn get_string_to_null(data: &[u8], offset: u32) -> String { - let start = offset as usize; - if start >= data.len() { - return String::new(); - } - // Bounded: an unterminated run must never walk off the end of the buffer - // (panic) or scan unboundedly into unrelated data. - let limit = start.saturating_add(4096).min(data.len()); - let mut i = start; - while i < limit && data[i] != 0 { - i += 1; - } - String::from_utf8_lossy(&data[start..i]).into_owned() -} - -/// Read a PE section-name field: exactly 8 bytes, NOT necessarily -/// NUL-terminated (a full-width name like `.textbss` has no NUL at all). -/// Returns the name with trailing NULs stripped. Using `get_string_to_null` -/// here would run past the field into the VirtualSize/VirtualAddress dwords. -pub(crate) fn section_name(data: &[u8], offset: u32) -> String { - let start = offset as usize; - let Some(field) = data.get(start..start + 8) else { - return String::new(); - }; - let end = field.iter().position(|&b| b == 0).unwrap_or(8); - String::from_utf8_lossy(&field[..end]).into_owned() -} - -// --------------------------------------------------------------------------- -// AES primitives -// --------------------------------------------------------------------------- - -/// One AES-CBC-like round over a 16-byte block in `d` at `pos`, using the -/// expanded key schedule stored in `d` at `key_offset`. Works entirely within -/// the single `d` buffer (both ciphertext and key schedule live there). -pub(crate) fn aes_round(d: &mut [u8], pos: u32, key_offset: u32, round: u32) { - let cm1 = &COLUMMIX1; - let cm2 = &COLUMMIX2; - let cm3 = &COLUMMIX3; - let cm4 = &COLUMMIX4; - let sbox = &SBOX; - - let mut n0 = get_u32(d, pos).swap_bytes() ^ get_u32(d, key_offset); - let mut n1 = - get_u32(d, pos.wrapping_add(4)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(4)); - let mut n2 = - get_u32(d, pos.wrapping_add(8)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(8)); - let mut n3 = - get_u32(d, pos.wrapping_add(12)).swap_bytes() ^ get_u32(d, key_offset.wrapping_add(12)); - - let mut r = 1u32; - while r < round { - let off = key_offset.wrapping_add(r.wrapping_mul(16)); - let a = get_u32(cm2, ((n3 >> 16) & 0xFF) * 4) - ^ get_u32(cm3, ((n2 >> 8) & 0xFF) * 4) - ^ get_u32(cm1, ((n0 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n1 & 0xFF) * 4) - ^ get_u32(d, off); - let b = get_u32(cm2, ((n0 >> 16) & 0xFF) * 4) - ^ get_u32(cm1, ((n1 >> 24) & 0xFF) * 4) - ^ get_u32(cm3, ((n3 >> 8) & 0xFF) * 4) - ^ get_u32(cm4, (n2 & 0xFF) * 4) - ^ get_u32(d, off.wrapping_add(4)); - let c = get_u32(cm2, ((n1 >> 16) & 0xFF) * 4) - ^ get_u32(cm3, ((n0 >> 8) & 0xFF) * 4) - ^ get_u32(cm1, ((n2 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n3 & 0xFF) * 4) - ^ get_u32(d, off.wrapping_add(8)); - let e = get_u32(cm3, ((n1 >> 8) & 0xFF) * 4) - ^ get_u32(cm2, ((n2 >> 16) & 0xFF) * 4) - ^ get_u32(cm1, ((n3 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n0 & 0xFF) * 4) - ^ get_u32(d, off.wrapping_add(12)); - n0 = a; - n1 = b; - n2 = c; - n3 = e; - r = r.wrapping_add(1); - } - - let s0 = (get_u32(sbox, ((n0 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n3 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n2 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n1 & 0xFF) * 4) & 0x0000_00FF); - let s1 = (get_u32(sbox, ((n1 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n0 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n3 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n2 & 0xFF) * 4) & 0x0000_00FF); - let s2 = (get_u32(sbox, ((n2 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n1 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n0 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n3 & 0xFF) * 4) & 0x0000_00FF); - let s3 = (get_u32(sbox, ((n3 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n2 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n1 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n0 & 0xFF) * 4) & 0x0000_00FF); - - let last = key_offset.wrapping_add(round.wrapping_mul(16)); - n0 = s0 ^ get_u32(d, last); - n1 = s1 ^ get_u32(d, last.wrapping_add(4)); - n2 = s2 ^ get_u32(d, last.wrapping_add(8)); - n3 = s3 ^ get_u32(d, last.wrapping_add(12)); - - write_u32(d, pos, n0.swap_bytes()); - write_u32(d, pos.wrapping_add(4), n1.swap_bytes()); - write_u32(d, pos.wrapping_add(8), n2.swap_bytes()); - write_u32(d, pos.wrapping_add(12), n3.swap_bytes()); -} - -/// AES-CBC-like decryption over `size` bytes starting at `pos` in `d`. -/// The key schedule lives at `key_offset` within the same buffer `d`. -pub(crate) fn aes_decrypt(d: &mut [u8], pos: u32, size: u32, key_offset: u32) { - let mut prev = [0u8; 16]; - let mut cur = [0u8; 16]; - let round = get_u16(d, key_offset.wrapping_add(2)) as u32; - let blocks = size >> 4; - for i in 0..blocks { - let p = pos.wrapping_add(i.wrapping_mul(16)); - let pi = p as usize; - cur.copy_from_slice(&d[pi..pi + 16]); - aes_round(d, p, key_offset.wrapping_add(4), round); - for j in 0..16 { - d[pi + j] ^= prev[j]; - } - prev = cur; - } -} - -/// [`aes_decrypt`] variant reading the key schedule from a separate snapshot -/// slice instead of the data buffer. `ks` is a snapshot of `d[key_offset..]` -/// taken by [`aes_schedule_snapshot`] (round count at `ks[2]`, round keys from -/// `ks[4]`), so the schedule extent is exactly right by construction. Used by -/// the parallel block fan-out, where each worker owns a disjoint `&mut` span -/// of the image and cannot read the schedule out of the shared buffer. -pub(crate) fn aes_decrypt_ks(ks: &[u8], d: &mut [u8], pos: u32, size: u32) { - let mut prev = [0u8; 16]; - let mut cur = [0u8; 16]; - let round = u16::from_le_bytes([ks[2], ks[3]]) as u32; - let sched = &ks[4..]; - let blocks = size >> 4; - for i in 0..blocks { - let p = pos.wrapping_add(i.wrapping_mul(16)); - let pi = p as usize; - cur.copy_from_slice(&d[pi..pi + 16]); - aes_round_ks(sched, d, p, round); - for j in 0..16 { - d[pi + j] ^= prev[j]; - } - prev = cur; - } -} - -/// [`aes_round`] with the round keys in a separate slice (see -/// [`aes_decrypt_ks`]). Identical math; only the key source differs. -fn aes_round_ks(ks: &[u8], d: &mut [u8], pos: u32, round: u32) { - let cm1 = &COLUMMIX1; - let cm2 = &COLUMMIX2; - let cm3 = &COLUMMIX3; - let cm4 = &COLUMMIX4; - let sbox = &SBOX; - let k = |i: u32| get_u32(ks, i); - - let mut n0 = get_u32(d, pos).swap_bytes() ^ k(0); - let mut n1 = get_u32(d, pos.wrapping_add(4)).swap_bytes() ^ k(4); - let mut n2 = get_u32(d, pos.wrapping_add(8)).swap_bytes() ^ k(8); - let mut n3 = get_u32(d, pos.wrapping_add(12)).swap_bytes() ^ k(12); - - let mut r = 1u32; - while r < round { - let off = r.wrapping_mul(16); - let a = get_u32(cm2, ((n3 >> 16) & 0xFF) * 4) - ^ get_u32(cm3, ((n2 >> 8) & 0xFF) * 4) - ^ get_u32(cm1, ((n0 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n1 & 0xFF) * 4) - ^ k(off); - let b = get_u32(cm2, ((n0 >> 16) & 0xFF) * 4) - ^ get_u32(cm1, ((n1 >> 24) & 0xFF) * 4) - ^ get_u32(cm3, ((n3 >> 8) & 0xFF) * 4) - ^ get_u32(cm4, (n2 & 0xFF) * 4) - ^ k(off.wrapping_add(4)); - let c = get_u32(cm2, ((n1 >> 16) & 0xFF) * 4) - ^ get_u32(cm3, ((n0 >> 8) & 0xFF) * 4) - ^ get_u32(cm1, ((n2 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n3 & 0xFF) * 4) - ^ k(off.wrapping_add(8)); - let e = get_u32(cm3, ((n1 >> 8) & 0xFF) * 4) - ^ get_u32(cm2, ((n2 >> 16) & 0xFF) * 4) - ^ get_u32(cm1, ((n3 >> 24) & 0xFF) * 4) - ^ get_u32(cm4, (n0 & 0xFF) * 4) - ^ k(off.wrapping_add(12)); - n0 = a; - n1 = b; - n2 = c; - n3 = e; - r = r.wrapping_add(1); - } - - let s0 = (get_u32(sbox, ((n0 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n3 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n2 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n1 & 0xFF) * 4) & 0x0000_00FF); - let s1 = (get_u32(sbox, ((n1 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n0 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n3 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n2 & 0xFF) * 4) & 0x0000_00FF); - let s2 = (get_u32(sbox, ((n2 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n1 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n0 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n3 & 0xFF) * 4) & 0x0000_00FF); - let s3 = (get_u32(sbox, ((n3 >> 24) & 0xFF) * 4) & 0xFF00_0000) - | (get_u32(sbox, ((n2 >> 16) & 0xFF) * 4) & 0x00FF_0000) - | (get_u32(sbox, ((n1 >> 8) & 0xFF) * 4) & 0x0000_FF00) - | (get_u32(sbox, (n0 & 0xFF) * 4) & 0x0000_00FF); - - let last = round.wrapping_mul(16); - n0 = s0 ^ k(last); - n1 = s1 ^ k(last.wrapping_add(4)); - n2 = s2 ^ k(last.wrapping_add(8)); - n3 = s3 ^ k(last.wrapping_add(12)); - - write_u32(d, pos, n0.swap_bytes()); - write_u32(d, pos.wrapping_add(4), n1.swap_bytes()); - write_u32(d, pos.wrapping_add(8), n2.swap_bytes()); - write_u32(d, pos.wrapping_add(12), n3.swap_bytes()); -} - -/// Snapshot the AES key schedule at `key_offset` for [`aes_decrypt_ks`]: -/// `d[key_offset .. key_offset + 4 + (round+1)*16]` where `round` is read from -/// the schedule header. Returns `None` when the header is truncated or the -/// round count is implausible (corrupt input — the same bytes would otherwise -/// drive reads past the buffer). -pub(crate) fn aes_schedule_snapshot(d: &[u8], key_offset: u32) -> Option> { - let base = key_offset as usize; - let round = u16::from_le_bytes([*d.get(base + 2)?, *d.get(base + 3)?]) as usize; - if round > 64 { - return None; - } - let end = base.checked_add(4 + (round + 1) * 16)?; - if end > d.len() { - return None; - } - Some(d[base..end].to_vec()) -} - -// --------------------------------------------------------------------------- -// Checksum primitives -// --------------------------------------------------------------------------- - -/// CRC32-based checksum over a (offset, length) descriptor pair embedded in -/// `d` at `pos`. Returns `crc32(d[offset..offset+length]) ^ length`. -pub(crate) fn calculate_checksum(d: &[u8], pos: u32) -> u32 { - let offset = get_u32(d, pos); - let length = get_u32(d, pos.wrapping_add(4)); - crc32::compute(&d[offset as usize..(offset + length) as usize]) ^ length -} - -/// CRC32 chained checksum. The (offset, length) descriptor at `pos` is read -/// from `d`; the bytes themselves are read from the separate `clean` buffer -/// (the original file image). `start` is the initial CRC accumulator. -pub(crate) fn calculate_checksum2(d: &[u8], clean: &[u8], pos: u32, start: u32) -> u32 { - let offset = get_u32(d, pos); - let length = get_u32(d, pos.wrapping_add(4)); - crc32::append(start, &clean[offset as usize..(offset + length) as usize]) -} - -// --------------------------------------------------------------------------- -// Decompression (Huffman/LZ) -// --------------------------------------------------------------------------- - -/// Huffman/LZ decompression operating entirely within a single `d` buffer. -/// Reads `s_size` bytes from `src`, writes `d_size` bytes to `dest`. -/// The Huffman table lives at `key_offset` within `d`. -/// -/// Returns a structured reason when the stream cannot produce exactly -/// `d_size` bytes. -pub(crate) fn decompress_detailed( - d: &mut [u8], - src: u32, - mut dest: u32, - key_offset: u32, - s_size: u32, - d_size: u32, -) -> Result<(), super::DecompressionFailure> { - use super::DecompressionFailure; - - // Bound the scratch allocation: a corrupt descriptor could request a - // multi-gigabyte source size, and an allocation failure aborts the process - // (uncatchable). Real payloads are far below this. - if s_size as u64 > super::MAX_IMAGE_SIZE { - return Err(DecompressionFailure::SourceTooLarge { - size: s_size, - max: super::MAX_IMAGE_SIZE, - }); - } - DECOMPRESS_SCRATCH.with_borrow_mut(|buf| -> Result<(), DecompressionFailure> { - let mut bit_pos: i32 = 0; - let need = (s_size as usize).saturating_add(3); - if buf.len() < need { - buf.resize(need, 0); - } - let mut buf_off: u32 = 0; - let mut src_consumed: i32 = 0; - let mut pending: u32 = 0; - let mut written: u32 = 0; - let src_u = src as usize; - let s_size_u = s_size as usize; - // The bit-reader's final get_u32 may read up to 3 bytes past s_size; those - // must be zero. Reused scratch can hold stale bytes there, so zero them - // before copying the (exactly s_size) source over the head. - buf[s_size_u] = 0; - buf[s_size_u + 1] = 0; - buf[s_size_u + 2] = 0; - buf[..s_size_u].copy_from_slice(&d[src_u..src_u + s_size_u]); - - while (src_consumed as u32) < s_size && written < d_size { - let word = get_u32(&buf[..], buf_off) >> bit_pos; - let tab_addr = key_offset.wrapping_add((word & 0xFF).wrapping_mul(3)); - let mut tab = get_u16(d, tab_addr); - let bits: u8; - if (tab & 0x8000) != 0 { - tab &= 0x7FFF; - bits = d[tab_addr as usize + 2]; - } else { - let mut b2 = d[tab_addr as usize + 2]; - // A Huffman code longer than 32 bits cannot exist; a larger - // length byte comes from a corrupt table, and `1 << b2` would - // panic (debug) or wrap (release) on it. - if b2 >= 32 { - return Err(DecompressionFailure::InvalidCodeLength { bits: b2 }); - } - let mut mask: u32 = 1u32 << b2; - b2 = b2.wrapping_add(1); - let mut idx = (tab & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; - let mut t2 = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); - // A corrupt table can form a non-terminal cycle; cap the walk so it - // fails instead of spinning forever. - let mut depth = 0u32; - while (t2 & 0x8000) == 0 { - depth += 1; - if depth > 64 { - return Err(DecompressionFailure::HuffmanTraversalLimit); - } - mask <<= 1; - b2 = b2.wrapping_add(1); - idx = (t2 & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; - t2 = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); - } - tab = t2 & 0x7FFF; - bits = b2; - } - bit_pos += bits as i32; - let advance = bit_pos / 8; - buf_off = buf_off.wrapping_add(advance as u32); - src_consumed += advance; - bit_pos %= 8; - - let mode = (tab as u32) & 0x300; - let payload = (tab as u32) & 0xFF; - let step: u32; - match mode { - 0 => { - step = 1; - d[dest as usize] = payload as u8; - } - 0x100 => { - step = 0; - if pending >= 256 { - return Err(DecompressionFailure::PendingLengthOverflow { pending }); - } - pending = if pending == 0 { - payload - } else { - (pending << 8) | payload - }; - } - 0x200 => { - if pending == 0 { - pending = 1; - } - step = pending.wrapping_mul(payload); - if step.wrapping_add(written) > d_size { - return Err(DecompressionFailure::OutputOverflow { - written, - step, - expected: d_size, - }); - } - // Run-fill replicates the unit just written before `dest`. A - // corrupt stream can emit one of these before anything has been - // written, so guard against reading before the buffer start - // (an unsigned underflow would index astronomically far OOB). - match payload { - 1 => { - if dest < 1 { - return Err(DecompressionFailure::RunFillBeforeOutput { - width: payload, - destination: dest, - }); - } - let v = d[(dest as usize) - 1]; - for k in 0..pending { - d[(dest + k) as usize] = v; - } - } - 2 => { - if dest < 2 { - return Err(DecompressionFailure::RunFillBeforeOutput { - width: payload, - destination: dest, - }); - } - let v = get_u16(d, dest.wrapping_sub(2)); - for k in 0..pending { - write_u16(d, dest.wrapping_add(k.wrapping_mul(2)), v as u32); - } - } - 4 => { - if dest < 4 { - return Err(DecompressionFailure::RunFillBeforeOutput { - width: payload, - destination: dest, - }); - } - let v = get_u32(d, dest.wrapping_sub(4)); - for k in 0..pending { - write_u32(d, dest.wrapping_add(k.wrapping_mul(4)), v); - } - } - _ => { - // Only unit widths 1/2/4 exist. Any other payload comes - // from a corrupt stream: previously this wrote nothing - // yet still counted `step` bytes as written, leaving - // stale-buffer holes that later stages treated as - // plaintext. Report corruption instead. - return Err(DecompressionFailure::InvalidRunFillWidth { - width: payload, - }); - } - } - pending = 0; - } - _ => { - step = payload; - if written.wrapping_add(payload) > d_size - || pending.wrapping_add(payload) > written - { - let distance = pending.wrapping_add(payload); - if distance > written { - return Err(DecompressionFailure::InvalidBackReference { - distance, - written, - }); - } - return Err(DecompressionFailure::OutputOverflow { - written, - step: payload, - expected: d_size, - }); - } - let back = pending.wrapping_add(payload); - for k in 0..payload { - d[(dest + k) as usize] = d[(dest + k - back) as usize]; - } - pending = 0; - } - } - - dest = dest.wrapping_add(step); - written = written.wrapping_add(step); - if bits == 0 && step == 0 { - // Corrupt table: no input bits consumed and no output bytes - // written, so the loop condition can never advance — an - // infinite loop (and `catch_unpack` traps panics, not hangs). - // Every real symbol consumes ≥ 1 bit, so a valid stream can - // never hit this. - return Err(DecompressionFailure::NoProgress); - } - } - src_consumed += if bit_pos != 0 { 1 } else { 0 }; - if written != d_size { - return Err(DecompressionFailure::OutputSizeMismatch { - written, - expected: d_size, - consumed: src_consumed.max(0) as u32, - source_size: s_size, - }); - } - Ok(()) - }) -} - -/// Boolean compatibility wrapper used by candidate searches and block fan-out. -pub(crate) fn decompress( - d: &mut [u8], - src: u32, - dest: u32, - key_offset: u32, - s_size: u32, - d_size: u32, -) -> bool { - decompress_detailed(d, src, dest, key_offset, s_size, d_size).is_ok() -} - -/// Walk the Huffman table at `key_offset` and snapshot its bytes for -/// [`decompress_tbl`]. The table is a forest of 256 root entries (3 bytes -/// each); non-terminal entries point at a child index pair. Returns `None` -/// when the table is truncated or self-referential past the buffer (corrupt -/// input — the same bytes would otherwise drive reads out of bounds). -pub(crate) fn huffman_table_snapshot(d: &[u8], key_offset: u32) -> Option> { - let mut visited = vec![false; 0x1_0000usize]; - let mut stack: Vec = (0..256).collect(); - let mut max_idx: u32 = 255; - while let Some(idx) = stack.pop() { - if idx >= 0x1_0000 || visited[idx as usize] { - continue; - } - visited[idx as usize] = true; - let off = key_offset as usize + idx as usize * 3; - if off + 3 > d.len() { - return None; - } - let t = get_u16(d, key_offset.wrapping_add(idx.wrapping_mul(3))); - if (t & 0x8000) == 0 { - let child = (t & 0x7FFF) as u32; - max_idx = max_idx.max(child).max(child.wrapping_add(1)); - stack.push(child); - stack.push(child.wrapping_add(1)); - } - } - let end = key_offset as usize + (max_idx as usize + 1) * 3; - if end > d.len() { - return None; - } - Some(d[key_offset as usize..end].to_vec()) -} - -/// [`decompress`] variant reading the Huffman table from a separate snapshot -/// slice (see [`huffman_table_snapshot`]) instead of the data buffer. Used by -/// the parallel block fan-out, where each worker owns a disjoint `&mut` span -/// and cannot read the table out of the shared image. Table reads are bounds -/// checked against the snapshot — past-the-end means corrupt table, reported -/// as `false` rather than a panic. -pub(crate) fn decompress_tbl( - tab: &[u8], - d: &mut [u8], - src: u32, - mut dest: u32, - s_size: u32, - d_size: u32, -) -> bool { - if s_size as u64 > super::MAX_IMAGE_SIZE { - return false; - } - DECOMPRESS_SCRATCH.with_borrow_mut(|buf| { - // Table reads, bounds-checked against the snapshot. - let tab16 = |addr: usize| -> Option { - let b = tab.get(addr..addr + 3)?; - Some(u16::from_le_bytes([b[0], b[1]])) - }; - let tab8 = |addr: usize| -> Option { tab.get(addr + 2).copied() }; - - let mut bit_pos: i32 = 0; - let need = (s_size as usize).saturating_add(3); - if buf.len() < need { - buf.resize(need, 0); - } - let mut buf_off: u32 = 0; - let mut src_consumed: i32 = 0; - let mut pending: u32 = 0; - let mut written: u32 = 0; - let src_u = src as usize; - let s_size_u = s_size as usize; - buf[s_size_u] = 0; - buf[s_size_u + 1] = 0; - buf[s_size_u + 2] = 0; - buf[..s_size_u].copy_from_slice(&d[src_u..src_u + s_size_u]); - - while (src_consumed as u32) < s_size && written < d_size { - let word = get_u32(&buf[..], buf_off) >> bit_pos; - let tab_addr = ((word & 0xFF).wrapping_mul(3)) as usize; - let mut tab = match tab16(tab_addr) { - Some(t) => t, - None => { - return false; - } - }; - let bits: u8; - if (tab & 0x8000) != 0 { - tab &= 0x7FFF; - bits = match tab8(tab_addr) { - Some(b) => b, - None => return false, - }; - } else { - let mut b2 = match tab8(tab_addr) { - Some(b) => b, - None => return false, - }; - if b2 >= 32 { - return false; - } - let mut mask: u32 = 1u32 << b2; - b2 = b2.wrapping_add(1); - let mut idx = (tab & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; - let mut t2 = match tab16(idx as usize * 3) { - Some(t) => t, - None => { - return false; - } - }; - // A corrupt table can form a non-terminal cycle; cap the walk so it - // fails instead of spinning forever. - let mut depth = 0u32; - while (t2 & 0x8000) == 0 { - depth += 1; - if depth > 64 { - return false; - } - mask <<= 1; - b2 = b2.wrapping_add(1); - idx = (t2 & 0x7FFF) as u32 + if (word & mask) != 0 { 1 } else { 0 }; - t2 = match tab16(idx as usize * 3) { - Some(t) => t, - None => { - return false; - } - }; - } - tab = t2 & 0x7FFF; - bits = b2; - } - bit_pos += bits as i32; - let advance = bit_pos / 8; - buf_off = buf_off.wrapping_add(advance as u32); - src_consumed += advance; - bit_pos %= 8; - - let mode = (tab as u32) & 0x300; - let payload = (tab as u32) & 0xFF; - let step: u32; - match mode { - 0 => { - step = 1; - d[dest as usize] = payload as u8; - } - 0x100 => { - step = 0; - if pending >= 256 { - return false; - } - pending = if pending == 0 { - payload - } else { - (pending << 8) | payload - }; - } - 0x200 => { - if pending == 0 { - pending = 1; - } - step = pending.wrapping_mul(payload); - if step.wrapping_add(written) > d_size { - return false; - } - // Run-fill replicates the unit just written before `dest` - // (see `decompress` for the underflow rationale). - match payload { - 1 => { - if dest < 1 { - return false; - } - let v = d[(dest as usize) - 1]; - for k in 0..pending { - d[(dest + k) as usize] = v; - } - } - 2 => { - if dest < 2 { - return false; - } - let v = get_u16(d, dest.wrapping_sub(2)); - for k in 0..pending { - write_u16(d, dest.wrapping_add(k.wrapping_mul(2)), v as u32); - } - } - 4 => { - if dest < 4 { - return false; - } - let v = get_u32(d, dest.wrapping_sub(4)); - for k in 0..pending { - write_u32(d, dest.wrapping_add(k.wrapping_mul(4)), v); - } - } - _ => { - return false; - } - } - pending = 0; - } - _ => { - step = payload; - if written.wrapping_add(payload) > d_size - || pending.wrapping_add(payload) > written - { - return false; - } - let back = pending.wrapping_add(payload); - for k in 0..payload { - d[(dest + k) as usize] = d[(dest + k - back) as usize]; - } - pending = 0; - } - } - - dest = dest.wrapping_add(step); - written = written.wrapping_add(step); - if bits == 0 && step == 0 { - return false; - } - } - src_consumed += if bit_pos != 0 { 1 } else { 0 }; - let _ = src_consumed; - written == d_size - }) -} -// Decrypt primitives (free-function wrappers) -// --------------------------------------------------------------------------- - -// --------------------------------------------------------------------------- -// PE32 (32-bit) helpers -// --------------------------------------------------------------------------- - -/// PE32 shell-table locator. Walks the shell region (`info[6]`) for a dword -/// equal to `info[6]` followed by a plausible shell size, returning the table -/// base (`candidate = off - 0x88`) when `candidate+0x58` holds a valid pointer. -pub(crate) fn find_tbl_pe32(data: &[u8], info: &[u32; 8]) -> Option { - let shell = info[6]; - if (data.len() as u64) < 0x100 { - return None; - } - let hi = (shell as u64) - .saturating_add(0x3000) - .min(data.len() as u64 - 0x100) as u32; - let mut off = shell; - while off < hi { - if off as usize + 8 <= data.len() { - let candidate = off.wrapping_sub(0x88); - if candidate >= shell && get_u32(data, off) == info[6] { - let shell_size_val = get_u32(data, off.wrapping_add(4)); - if shell_size_val > 0x1000 && shell_size_val < 0x100000 { - let v58_off = candidate.wrapping_add(0x58); - if (v58_off as usize + 4) <= data.len() { - let v58 = get_u32(data, v58_off); - if v58 > 0 && (v58 as usize) < data.len() { - return Some(candidate); - } - } - } - } - } - off = off.wrapping_add(4); - } - None -} - -/// Locate an LFSR-encrypted bytecode block (decrypt_data6 form) in a region. -/// `start_off` is the byte offset to begin scanning at, `scan_backward` -/// controls direction. Returns the relative offset of the block. Includes full -/// opcode-walk validation of candidate blocks. -pub(crate) fn find_lfsr_block( - data: &[u8], - base: u32, - size: u32, - start_off: u32, - scan_backward: bool, -) -> Option { - if size < 96 { - return None; - } - let mut ks = [0u8; 128]; - lfsr_keystream(&mut ks); - let check = |scan_off: u32| -> bool { - let abs_off = base.wrapping_add(scan_off) as usize; - if abs_off + 96 > data.len() { - return false; - } - let sz = data[abs_off + 95] as usize; - if !(10..=95).contains(&sz) { - return false; - } - let mut decoded = [0u8; 95]; - for bi in 0..sz { - decoded[bi] = data[abs_off + bi] ^ ks[bi]; - } - // Full bytecode validation (shared with the stage4/5 locator): every - // opcode must decode with a valid ModR/M and the stream must REACH a - // RET (0xC3) as an opcode. The previous check only required a 0xC3 - // byte *anywhere* in the window and accepted a walk that ran off the - // end without hitting RET — a `0x04 0xC3` (ADD 0xC3) tail passed, so - // coincidental LFSR-shaped garbage was accepted as a decryptor block. - parse_bytecode_check(&decoded[..sz]).is_some() - }; - if scan_backward { - let hi = size - 96; - if hi >= start_off { - let mut scan_off = hi; - loop { - if check(scan_off) { - return Some(scan_off); - } - if scan_off == start_off { - break; - } - scan_off -= 1; - } - } - } else { - let hi = size - 95; - let mut scan_off = start_off; - while scan_off < hi { - if check(scan_off) { - return Some(scan_off); - } - scan_off += 1; - } - } - None -} - -/// Slots discovered in the eighthStage for the marker-less layout. -pub(crate) struct EighthSlots { - /// Absolute address of the file-data decryptor LFSR bytecode block. The - /// fileCS chain pointer is derived downstream as `file_lfsr - 0x58`. - pub file_lfsr: u32, - /// Absolute address of the compressedInfo (ptr,size) table pointer slot. - pub compressed_info_ptr: u32, -} - -/// Marker-independent eighthStage slot discovery (PE32+ branch). -/// -/// Newer Crackproof builds (e.g. some native/managed DLLs) omit the -/// `pm\0\0cm\0\0` and `00 00 00 40 01 00 00 00` markers that the older layout's -/// walk3/walk4/walk5 slot derivation relies on. Instead this discovers the -/// slots structurally: -/// * Scan the eighthStage for every LFSR (decrypt_data6) bytecode block. -/// * The file decryptor is the LFSR block whose `fileCS = lfsr - 0x58` holds -/// a pointer sitting just past `info[3]` (smallest positive distance). -/// * `compressedInfo` is the pointer slot whose 16-byte target, after a -/// trial `decrypt_data5`, parses as a plausible (src,sSize,dst,dSize) -/// descriptor. -/// -/// Returns `None` if no plausible file LFSR is found. `eighth_start`/`eighth_dsz` -/// bound the search region; `info3` is `info[3]`; `compress_data_offset` is -/// `(!u32(file_data,0x1080)) + 0x1000`; `file_data_len` is the protected file -/// length. -#[allow(clippy::too_many_arguments)] -pub(crate) fn discover_eighth_slots( - data: &[u8], - eighth_start: u32, - eighth_dsz: u32, - info3: u32, - compress_data_offset: u32, - file_data_len: u32, -) -> Option { - // Collect all LFSR candidates (forward scan). - // - // Advance by 1 after each hit, NOT by 96. A false-positive LFSR match can sit - // just before the real file-decryptor block (observed on an il2cpp game - // assembly build, 2026-07-13: junk at rel=0x31C1, real block at 0x3210). - // Stepping by the LFSR body size then skips the real block and discovery - // fails. Byte-stepping is cheap: eighthStage is only a few KB. - let mut all_lfsrs: Vec = Vec::new(); - let mut scan_off: u32 = 0; - while scan_off + 95 < eighth_dsz { - match find_lfsr_block(data, eighth_start, eighth_dsz, scan_off, false) { - Some(found) => { - all_lfsrs.push(found); - scan_off = found + 1; - } - None => break, - } - } - - // Pick the file LFSR: prefer the candidate whose fileCS pointer sits the - // smallest positive distance past info[3]. - let mut off_file_lfsr: Option = None; - let mut best_dist: Option = None; - for &lfsr_off in &all_lfsrs { - if lfsr_off < 0x58 { - continue; - } - let cs_off = lfsr_off - 0x58; - let cs_val = get_u32(data, eighth_start.wrapping_add(cs_off)); - if !(0x1000 < cs_val && (cs_val as usize) < data.len()) { - continue; - } - if cs_val < info3 { - continue; - } - let dist = cs_val - info3; - if best_dist.is_none_or(|b| dist < b) { - best_dist = Some(dist); - off_file_lfsr = Some(lfsr_off); - } - } - // Fallback: last LFSR with any in-image fileCS pointer. - if off_file_lfsr.is_none() { - for &lfsr_off in all_lfsrs.iter().rev() { - if lfsr_off < 0x58 { - continue; - } - let cs_val = get_u32(data, eighth_start.wrapping_add(lfsr_off - 0x58)); - if 0x1000 < cs_val && (cs_val as usize) < data.len() { - off_file_lfsr = Some(lfsr_off); - break; - } - } - } - let off_file_lfsr = off_file_lfsr?; - let off_file_cs = off_file_lfsr - 0x58; - - // Trial-decrypt to find compressedInfo: the pointer slot in the data area - // (between fileCS region start and the LFSR) whose target parses as a valid - // (src,sSize,dst,dSize) descriptor after a transient decrypt_data5. - let scan_from = off_file_lfsr.saturating_sub(0x400); - let mut off_compressed_info: Option = None; - let mut doff = scan_from; - while doff < off_file_lfsr { - if doff == off_file_cs { - doff += 4; - continue; - } - let ptr_val = get_u32(data, eighth_start.wrapping_add(doff)); - if !(0x1000 < ptr_val && (ptr_val as usize) < data.len().saturating_sub(16)) { - doff += 4; - continue; - } - // Predict decrypt_data5(ptr_val, 16) without mutating: each dword is - // position-keyed and independent, so trial_decrypt5_u32 per dword. - let src2 = trial_decrypt5_u32(data, ptr_val); - let s_sz2 = trial_decrypt5_u32(data, ptr_val + 4); - let dst2 = trial_decrypt5_u32(data, ptr_val + 8); - let d_sz2 = trial_decrypt5_u32(data, ptr_val + 12); - let src_file_off = src2.wrapping_add(compress_data_offset); - let valid = s_sz2 > 0 - && s_sz2 < 0x200000 - && (src_file_off as u64 + s_sz2 as u64) <= file_data_len as u64 - && dst2 >= 0x1000 - && (dst2 as u64 + d_sz2 as u64) <= data.len() as u64 - && d_sz2 >= s_sz2 - && d_sz2 < 0x200000; - if valid { - off_compressed_info = Some(doff); - break; - } - doff += 4; - } - let off_compressed_info = off_compressed_info?; - - Some(EighthSlots { - file_lfsr: eighth_start.wrapping_add(off_file_lfsr), - compressed_info_ptr: eighth_start.wrapping_add(off_compressed_info), - }) -} - -/// PE32 `.text` dd8 key-formula selection with a skip decision. The packer keys -/// the per-page XOR either with `page+1` or `0x8000*(page+1)`; the formula is -/// not recorded. Replays the dd8 page pass on a scratch copy of sample pages -/// (25/50/75% of `.text`) under each formula and counts how many positions -/// decode to `0xCC` (int3 padding). -/// -/// Returns `Some(true)` for the `0x8000*(page+1)` formula, `Some(false)` for -/// `page+1`, or `None` when `.text` must NOT be dd8-decrypted at all. The packer -/// dd8-encrypts `.text` on EXEs (so unpacking must replay it) but leaves a native -/// DLL's `.text` plaintext; replaying dd8 there scrambles ~1 byte per 16-byte -/// block. The decision: dd8 only *restores* int3 padding when `.text` was -/// genuinely encrypted, so apply it only when the chosen formula's whole-page -/// 0xCC count rises *clearly* above the no-dd8 baseline; otherwise skip. -/// -/// "Clearly" matters: dd8 XORs 255 positions per page with pseudo-random bytes, -/// so on an already-plaintext `.text` it manufactures ~1 spurious `0xCC` per -/// sampled page for free (255/256 expected). A bare `best > baseline` test is -/// therefore biased towards *applying* dd8 on exactly the inputs that must skip -/// it — and a wrongly-applied dd8 is silent: it scrambles ~1 byte per 16 with no -/// error and nothing downstream (not even `integrity::check`, which only reads -/// 16 bytes at the entry point) notices. The [`MIN_DD8_NET_GAIN`] floor below is -/// the PE32 counterpart of the margin+floor `select_dd8_shift` already applies -/// on PE32+ for the same failure mode. -pub(crate) fn select_dd8_formula_pe32(data: &[u8], text_off: u32, text_size: u32) -> Option { - let num_pages_total = text_size / 0x1000; - let mut sample_pages: Vec = Vec::new(); - for frac in [0.25f64, 0.5, 0.75] { - let pg = (num_pages_total as f64 * frac) as u32; - if pg > 0 && pg < num_pages_total { - sample_pages.push(pg); - } - } - if sample_pages.is_empty() && num_pages_total > 1 { - sample_pages.push(num_pages_total / 2); - } - let score = |big: bool| -> i64 { - let mut total = 0i64; - for &sp in &sample_pages { - let pg_off = (text_off + sp * 0x1000) as usize; - if pg_off + 0x1000 > data.len() { - continue; - } - let mut buf = [0u8; 0x1000]; - buf.copy_from_slice(&data[pg_off..pg_off + 0x1000]); - let pk = if big { - 0x8000u32.wrapping_mul(sp.wrapping_add(1)) - } else { - sp.wrapping_add(1) - }; - let mut k = pk; - let rk = k.rotate_right(15); - k = rk; - for bi in 1..256u32 { - let rk = k.rotate_right(15); - let ri = rk.wrapping_add(bi); - k = ri.wrapping_add(bi); - let tidx = (bi.wrapping_mul(16).wrapping_add(ri & 0xF)) as usize; - if tidx < buf.len() { - buf[tidx] ^= k as u8; - } - } - total += buf.iter().filter(|&&b| b == 0xCC).count() as i64; - } - total - }; - let s_small = score(false); - let s_big = score(true); - // Baseline: whole-page 0xCC over the same sample pages with NO dd8. dd8 only - // rewrites 255 bytes per page, so comparing the chosen formula's whole-page - // 0xCC against this baseline reveals whether dd8 *restores* int3 padding - // (count rises -> .text was packer-encrypted, apply) or merely scrambles - // already-plaintext code (count falls -> native-DLL .text left intact, skip). - let mut baseline: i64 = 0; - for &sp in &sample_pages { - let pg_off = (text_off + sp * 0x1000) as usize; - if pg_off + 0x1000 > data.len() { - continue; - } - baseline += data[pg_off..pg_off + 0x1000] - .iter() - .filter(|&&b| b == 0xCC) - .count() as i64; - } - let big = s_big > s_small; - let best = s_small.max(s_big); - // Minimum net 0xCC gain over the baseline before dd8 is applied. Noise on an - // already-plaintext `.text` is ~1 manufactured 0xCC per sampled page (3 pages - // -> ~3); every corpus build that genuinely needs dd8 gains +154 or more - // (observed +154 and +312), and the one native DLL that must skip scores -18. - // A floor of 32 sits ~10x above the noise and ~5x below the smallest true - // positive, so it changes no existing decision. - const MIN_DD8_NET_GAIN: i64 = 32; - let apply = best.saturating_sub(baseline) >= MIN_DD8_NET_GAIN; - if std::env::var("SEL_DIAG").is_ok() { - eprintln!( - "SEL pe32 dd8 s_small={} s_big={} baseline={} gain={} big={} apply={}", - s_small, - s_big, - baseline, - best - baseline, - big, - apply - ); - } - // When no interior pages could be sampled (tiny .text) we cannot measure the - // effect; preserve the historical behavior of applying dd8. - if sample_pages.is_empty() || apply { - Some(big) - } else { - None - } -} - -/// Read a NUL-terminated byte string starting at `off`, bounded to 512 bytes. -/// Returns the raw bytes up to the terminator (excluding it). -fn read_cstr_bounded(data: &[u8], off: u32) -> Vec { - let start = off as usize; - if start >= data.len() { - return Vec::new(); - } - let limit = (start + 512).min(data.len()); - let mut end = start; - while end < limit && data[end] != 0 { - end += 1; - } - data[start..end].to_vec() -} - -fn align_up_u32(value: u32, alignment: u32) -> u32 { - ((value.wrapping_add(alignment - 1)) / alignment).wrapping_mul(alignment) -} - -fn align_up_u64(value: u64, alignment: u64) -> u64 { - value.div_ceil(alignment) * alignment -} - -#[derive(Clone)] -enum ImportFunc { - Ordinal(u32), - Name(u16, Vec), -} - -struct ImportDesc { - time_date: u32, - fwd_chain: u32, - dll_name: Vec, - iat_rva: u32, - functions: Vec, -} - -/// Return true when PE32 imports already sit in the original `.idata` layout -/// (so no relocation to `.kmiat` is needed). May write the IAT data directory -/// (pe+0xD8). -pub(crate) fn pe32_imports_already_match_idata_layout(data: &mut [u8], pe_header: u32) -> bool { - let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; - let sec_table = pe_header.wrapping_add(24).wrapping_add(opt_hdr_size); - let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; - let import_rva = get_u32(data, pe_header.wrapping_add(0x80)); - let import_size = get_u32(data, pe_header.wrapping_add(0x84)); - let len = data.len() as u32; - if !(import_rva > 0 && import_size > 0) { - return false; - } - for idx in 0..num_sections { - let sec_off = sec_table.wrapping_add(idx * 40); - if (sec_off as usize + 40) > data.len() { - return false; - } - if &data[sec_off as usize..sec_off as usize + 6] != b".idata" { - continue; - } - let sec_va = get_u32(data, sec_off.wrapping_add(12)); - let sec_size = - get_u32(data, sec_off.wrapping_add(8)).max(get_u32(data, sec_off.wrapping_add(16))); - let sec_end = sec_va.wrapping_add(sec_size); - if !(sec_va <= import_rva - && import_rva < sec_end - && import_rva.wrapping_add(import_size) <= sec_end) - { - continue; - } - let first_oft = get_u32(data, import_rva); - let first_name = get_u32(data, import_rva.wrapping_add(12)); - let first_iat = get_u32(data, import_rva.wrapping_add(16)); - if !(sec_va <= first_oft - && first_oft < sec_end - && sec_va <= first_iat - && first_iat < sec_end) - { - return false; - } - if !(0x1000 < first_name && first_name < len) { - return false; - } - let dll_name = read_cstr_bounded(data, first_name); - let lower: Vec = dll_name.iter().map(|b| b.to_ascii_lowercase()).collect(); - if !lower.ends_with(b".dll") { - return false; - } - let mut iat_min = first_iat; - let mut iat_max = first_iat; - let mut idt_pos = import_rva; - while idt_pos.wrapping_add(20) <= len { - let oft_rva = get_u32(data, idt_pos); - let name_rva = get_u32(data, idt_pos.wrapping_add(12)); - let iat_rva = get_u32(data, idt_pos.wrapping_add(16)); - if oft_rva == 0 && name_rva == 0 && iat_rva == 0 { - break; - } - if !(sec_va <= oft_rva && oft_rva < sec_end && sec_va <= iat_rva && iat_rva < sec_end) { - return false; - } - let mut thunk = iat_rva; - while thunk.wrapping_add(4) <= sec_end { - let tv = get_u32(data, thunk); - thunk = thunk.wrapping_add(4); - if tv == 0 { - break; - } - } - iat_min = iat_min.min(iat_rva); - iat_max = iat_max.max(thunk); - idt_pos = idt_pos.wrapping_add(20); - } - if iat_max > iat_min { - write_u32(data, pe_header.wrapping_add(0xD8), iat_min); - write_u32(data, pe_header.wrapping_add(0xDC), iat_max - iat_min); - } - return true; - } - false -} - -/// Rebuild PE32 import metadata (descriptors, lookup tables, names) into the -/// last section as `.kmiat`, leaving the loader-written IAT in place. Mutates -/// `data` (may grow it). -pub(crate) fn move_pe32_imports_to_kmiat(data: &mut Vec, pe_header: u32) { - const SECTION_SIZE: u32 = 0x7000; - let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; - let opt_hdr = pe_header.wrapping_add(24); - let sec_table = opt_hdr.wrapping_add(opt_hdr_size); - let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; - if num_sections == 0 { - return; - } - let import_rva = get_u32(data, pe_header.wrapping_add(0x80)); - let import_size = get_u32(data, pe_header.wrapping_add(0x84)); - let len = data.len() as u32; - if !(0x1000 < import_rva && import_rva < len && import_size > 0 && import_size < SECTION_SIZE) { - return; - } - - let mut descriptors: Vec = Vec::new(); - let mut idt_pos = import_rva; - while idt_pos.wrapping_add(20) <= len { - let oft_rva = get_u32(data, idt_pos); - let time_date = get_u32(data, idt_pos.wrapping_add(4)); - let fwd_chain = get_u32(data, idt_pos.wrapping_add(8)); - let name_rva = get_u32(data, idt_pos.wrapping_add(12)); - let iat_rva = get_u32(data, idt_pos.wrapping_add(16)); - if oft_rva == 0 && name_rva == 0 && iat_rva == 0 { - break; - } - if !(0x1000 < name_rva && name_rva < len) { - break; - } - let dll_name = read_cstr_bounded(data, name_rva); - let thunk_rva = if 0x1000 < oft_rva && oft_rva < len { - oft_rva - } else { - iat_rva - }; - let mut functions: Vec = Vec::new(); - let mut thunk_pos = thunk_rva; - while 0x1000 < thunk_pos.wrapping_add(4) && thunk_pos.wrapping_add(4) <= len { - let thunk_val = get_u32(data, thunk_pos); - if thunk_val == 0 { - break; - } - if thunk_val & 0x8000_0000 != 0 { - functions.push(ImportFunc::Ordinal(thunk_val & 0xFFFF)); - } else { - let hint = if thunk_val.wrapping_add(2) <= len { - get_u16(data, thunk_val) - } else { - 0 - }; - let func_name = if thunk_val.wrapping_add(2) < len { - read_cstr_bounded(data, thunk_val.wrapping_add(2)) - } else { - Vec::new() - }; - functions.push(ImportFunc::Name(hint, func_name)); - } - thunk_pos = thunk_pos.wrapping_add(4); - } - descriptors.push(ImportDesc { - time_date, - fwd_chain, - dll_name, - iat_rva, - functions, - }); - idt_pos = idt_pos.wrapping_add(20); - } - if descriptors.is_empty() { - return; - } - - for desc in &mut descriptors { - let lower: Vec = desc - .dll_name - .iter() - .map(|b| b.to_ascii_lowercase()) - .collect(); - if lower.starts_with(b"api-ms-win-crt-") { - desc.dll_name = b"ucrtbase.dll".to_vec(); - } else { - desc.dll_name = lower; - } - } - descriptors.sort_by_key(|d| d.iat_rva); - - let last_sec = sec_table.wrapping_add((num_sections - 1) * 40); - let kmiat_rva = get_u32(data, last_sec.wrapping_add(12)); - // A zero last-section VA means a corrupt section table: building .kmiat at - // RVA 0 would zero the DOS/PE headers and emit a structurally broken image - // with no error. Bail and keep the original import table. - if kmiat_rva == 0 { - return; - } - // Grow the image when .kmiat overruns it, but cap the growth: a corrupt VA - // could otherwise request a multi-gigabyte allocation, which aborts the - // process (uncatchable). Use u64 math so a near-u32::MAX VA cannot wrap the - // end calculation the way the previous wrapping/plain-add mix could. - let kmiat_end = kmiat_rva as u64 + SECTION_SIZE as u64; - if kmiat_end > super::MAX_IMAGE_SIZE { - return; - } - if kmiat_end > data.len() as u64 { - data.resize(kmiat_end as usize, 0); - } - // Zero the .kmiat region. - for b in &mut data[kmiat_rva as usize..kmiat_end as usize] { - *b = 0; - } - - let idt_size = (descriptors.len() as u32 + 1) * 20; - let oft_start = kmiat_rva; - let mut idt_rva = oft_start; - for desc in &descriptors { - idt_rva = idt_rva.wrapping_add((desc.functions.len() as u32 + 1) * 4); - } - idt_rva = align_up_u32(idt_rva.wrapping_add(0x2C), 4); - - // Size check: compute the final name_pos and bail if it overruns .kmiat. - let mut name_pos_check = idt_rva.wrapping_add(idt_size); - for desc in &descriptors { - name_pos_check = name_pos_check.wrapping_add(desc.dll_name.len() as u32 + 1); - for func in &desc.functions { - if let ImportFunc::Name(_, fname) = func { - name_pos_check = name_pos_check.wrapping_add(2 + fname.len() as u32 + 1); - } - } - } - if name_pos_check > kmiat_rva.wrapping_add(SECTION_SIZE) { - // Section too small; keep existing import table untouched. - return; - } - - let mut oft_pos = oft_start; - let mut name_pos = idt_rva.wrapping_add(idt_size); - for (idx, desc) in descriptors.iter().enumerate() { - let idt_entry = idt_rva.wrapping_add(idx as u32 * 20); - let current_oft = oft_pos; - write_u32(data, idt_entry, current_oft); - write_u32(data, idt_entry.wrapping_add(4), desc.time_date); - write_u32(data, idt_entry.wrapping_add(8), desc.fwd_chain); - let dll_name_pos = name_pos; - write_u32(data, idt_entry.wrapping_add(12), dll_name_pos); - write_u32(data, idt_entry.wrapping_add(16), desc.iat_rva); - - let dnp = dll_name_pos as usize; - data[dnp..dnp + desc.dll_name.len()].copy_from_slice(&desc.dll_name); - data[dnp + desc.dll_name.len()] = 0; - name_pos = name_pos.wrapping_add(desc.dll_name.len() as u32 + 1); - - for func in &desc.functions { - match func { - ImportFunc::Ordinal(ord) => { - write_u32(data, oft_pos, 0x8000_0000 | ord); - } - ImportFunc::Name(hint, fname) => { - let hint_name_rva = name_pos; - write_u32(data, oft_pos, hint_name_rva); - write_u16(data, hint_name_rva, *hint as u32); - let fp = (hint_name_rva + 2) as usize; - data[fp..fp + fname.len()].copy_from_slice(fname); - data[fp + fname.len()] = 0; - name_pos = name_pos.wrapping_add(2 + fname.len() as u32 + 1); - } - } - oft_pos = oft_pos.wrapping_add(4); - } - write_u32(data, oft_pos, 0); - oft_pos = oft_pos.wrapping_add(4); - } - // Null-terminator IDT entry (20 zero bytes) after the last descriptor. - let term = idt_rva.wrapping_add(descriptors.len() as u32 * 20) as usize; - for b in &mut data[term..term + 20] { - *b = 0; - } - - let ls = last_sec as usize; - data[ls..ls + 8].copy_from_slice(b".kmiat\x00\x00"); - write_u32(data, last_sec.wrapping_add(8), SECTION_SIZE); - write_u32(data, last_sec.wrapping_add(16), SECTION_SIZE); - write_u32(data, last_sec.wrapping_add(36), 0xE000_0060); - write_u32(data, pe_header.wrapping_add(0x80), idt_rva); - write_u32(data, pe_header.wrapping_add(0x84), idt_size); - write_u32( - data, - pe_header.wrapping_add(80), - kmiat_rva.wrapping_add(SECTION_SIZE), - ); -} - -/// Convert the unpacked RVA-addressed image back to a compact PE file layout -/// (headers at 0x400, sections packed consecutively, FileAlignment 0x200). -/// Returns `None` if the accumulated output size wraps or exceeds -/// [`super::MAX_IMAGE_SIZE`]: the final allocation is sized from header-derived -/// section data, and an uncapped `vec![0; n]` from a corrupt header would abort -/// the process (which `catch_unpack` cannot trap). -pub(crate) fn compact_memory_image_to_pe(data: &[u8], pe_header: u32) -> Option> { - const FILE_ALIGNMENT: u32 = 0x200; - const HEADER_SIZE: u32 = 0x400; - let opt_hdr_size = get_u16(data, pe_header.wrapping_add(20)) as u32; - let opt_hdr = pe_header.wrapping_add(24); - let sec_table = opt_hdr.wrapping_add(opt_hdr_size); - let num_sections = get_u16(data, pe_header.wrapping_add(6)) as u32; - - struct SecLayout { - sec_off: u32, - va: u32, - vsize: u32, - raw_ptr: u32, - raw_size: u32, - } - - let mut raw_cursor: u64 = HEADER_SIZE as u64; - let mut raw_layout: Vec = Vec::new(); - for idx in 0..num_sections { - let sec_off = sec_table.wrapping_add(idx * 40); - let vsize = get_u32(data, sec_off.wrapping_add(8)); - let va = get_u32(data, sec_off.wrapping_add(12)); - let sd_start = va as usize; - let sd_end = if (va.wrapping_add(vsize) as usize) <= data.len() { - va.wrapping_add(vsize) as usize - } else { - data.len() - }; - let section_data: &[u8] = if sd_start <= sd_end && sd_start <= data.len() { - &data[sd_start..sd_end] - } else { - &[] - }; - - let mut last_nonzero: i64 = -1; - for pos in (0..section_data.len()).rev() { - if section_data[pos] != 0 { - last_nonzero = pos as i64; - break; - } - } - let meaningful = if last_nonzero >= 0 { - (last_nonzero + 1) as u32 - } else { - 0 - }; - let mut raw_size = if meaningful != 0 { - align_up_u32(meaningful, FILE_ALIGNMENT) - } else { - 0 - }; - if vsize != 0 && raw_size == 0 { - raw_size = FILE_ALIGNMENT; - } - raw_size = raw_size.min(align_up_u32(section_data.len() as u32, FILE_ALIGNMENT)); - - let raw_ptr = if raw_size != 0 { raw_cursor as u32 } else { 0 }; - raw_layout.push(SecLayout { - sec_off, - va, - vsize, - raw_ptr, - raw_size, - }); - if raw_size != 0 { - // Accumulate in u64 and cap: section sizes are header-derived, and - // a corrupt table could otherwise wrap raw_cursor (small alloc, - // huge recorded raw_ptrs → OOB panic) or request an abort-sized - // allocation. - raw_cursor = align_up_u64(raw_cursor + raw_size as u64, FILE_ALIGNMENT as u64); - if raw_cursor > super::MAX_IMAGE_SIZE { - return None; - } - } - } - - let mut compact = vec![0u8; raw_cursor as usize]; - let hdr_copy = (HEADER_SIZE as usize).min(data.len()); - compact[..hdr_copy].copy_from_slice(&data[..hdr_copy]); - write_u32(&mut compact, opt_hdr.wrapping_add(36), FILE_ALIGNMENT); - write_u32(&mut compact, opt_hdr.wrapping_add(60), HEADER_SIZE); - - for sl in &raw_layout { - write_u32(&mut compact, sl.sec_off.wrapping_add(16), sl.raw_size); - write_u32(&mut compact, sl.sec_off.wrapping_add(20), sl.raw_ptr); - if sl.raw_size != 0 { - let sd_start = sl.va as usize; - let sd_end = if (sl.va.wrapping_add(sl.vsize) as usize) <= data.len() { - sl.va.wrapping_add(sl.vsize) as usize - } else { - data.len() - }; - let section_data: &[u8] = if sd_start <= sd_end { - &data[sd_start..sd_end] - } else { - &[] - }; - let copy_size = (sl.raw_size as usize).min(section_data.len()); - let rp = sl.raw_ptr as usize; - compact[rp..rp + copy_size].copy_from_slice(§ion_data[..copy_size]); - } - } - Some(compact) -} - -/// decrypt_data3: XOR+rotate cipher. Reads/writes dwords in `d` starting at -/// the address stored at `d[pos]`, for `d[pos+4]>>2` words. `shift` is the -/// right-rotate amount (19 or 21 depending on caller). -pub(crate) fn decrypt_data3(d: &mut [u8], pos: u32, mut key: u32, shift: u32) { - let base_addr = get_u32(d, pos); - let length = get_u32(d, pos.wrapping_add(4)); - let words = length >> 2; - for i in 0..words { - let off = base_addr.wrapping_add(i.wrapping_mul(4)); - let v = get_u32(d, off) ^ key; - key = key.wrapping_add(i); - let rotated = v.rotate_right(shift); - write_u32(d, off, rotated.wrapping_sub(i)); - } -} - -/// decrypt_data1 (called `decrypt_data` in the original): decode the 8-dword -/// info header from `file_data` at offset 4096 and write results into `info`. -pub(crate) fn decrypt_data1(file_data: &[u8], info: &mut [u32; 8]) { - info[0] = get_u32(file_data, 4096); - let mut k = get_u32(file_data, 4096); - for i in 0..7u32 { - let off = i.wrapping_mul(4).wrapping_add(4); - let cell = get_u32(file_data, 4096u32.wrapping_add(off)); - info[(i + 1) as usize] = k ^ cell; - k = i.wrapping_mul(i) ^ (k.wrapping_add(cell).wrapping_sub(i)); - } -} - -/// decrypt_data6: LFSR XOR decryption of a bytecode block at `pos` in `d`. -/// The block length is read from `d[pos + 95]`. -pub(crate) fn decrypt_data6(d: &mut [u8], pos: u32) { - let len = d[(pos + 95) as usize] as usize; - // The keystream is exactly `lfsr_keystream`'s — generate it once (len is a - // byte, so 256 always covers it) instead of keeping a second copy of the - // LFSR that a future poly fix would have to update separately. - let mut ks = [0u8; 256]; - lfsr_keystream(&mut ks); - let pos = pos as usize; - for i in 0..len { - d[pos + i] ^= ks[i]; - } -} - -/// decrypt_data7: nibble-swap + key-rolling byte cipher applied to a -/// null-terminated string in `d` starting at `pos`. -pub(crate) fn decrypt_data7(d: &mut [u8], pos: u32, mut key: u8) { - let mut i: u32 = 0; - loop { - let idx = (pos + i) as usize; - if d[idx] == 0 { - break; - } - let mut b = d[idx]; - b = b.rotate_right(4); - b = b.wrapping_sub(key); - if b == 0 { - b = 0u8.wrapping_sub(key); - } - d[idx] = b; - key = key.wrapping_add(67); - i += 1; - } -} - -// --------------------------------------------------------------------------- -// Higher-level composite: AES + decrypt3 + optional bytecode + decompress -// --------------------------------------------------------------------------- - -/// Decrypt and optionally decompress a stage payload descriptor. -/// `pos` points to a (src, src_len, dest, dest_len) quad of dwords in `d`. -/// - AES-decrypts `src..src+src_len` using key at `key3_offset` -/// - XOR+rotate-decrypts with `decrypt_data3(pos, key, 19)` -/// - Applies optional custom `ops` bytecode per-byte -/// - If `src_len != dest_len`, Huffman/LZ-decompresses `src..` → `dest..` -/// -/// Returns the decompression success status (always `true` when no -/// decompression was needed). The PE32 eighth-stage key search relies on this. -pub(crate) fn decrypt_and_decompress_data_detailed( - d: &mut [u8], - pos: u32, - key: u32, - key1_offset: u32, - key3_offset: u32, - ops: Option<&[Op]>, -) -> Result<(), super::DecompressionFailure> { - let src = get_u32(d, pos); - let src_len = get_u32(d, pos.wrapping_add(4)); - aes_decrypt(d, src, src_len, key3_offset); - decrypt_data3(d, pos, key, 19); - if let Some(ops) = ops - && src_len != 0 - { - OpsLut::new(ops).map_region(d, src as usize, src_len as usize); - } - let dest = get_u32(d, pos.wrapping_add(8)); - let dest_len = get_u32(d, pos.wrapping_add(12)); - if src_len != dest_len { - return decompress_detailed(d, src, dest, key1_offset, src_len, dest_len); - } - Ok(()) -} - -/// Boolean compatibility wrapper used by key searches that trial candidates. -pub(crate) fn decrypt_and_decompress_data( - d: &mut [u8], - pos: u32, - key: u32, - key1_offset: u32, - key3_offset: u32, - ops: Option<&[Op]>, -) -> bool { - decrypt_and_decompress_data_detailed(d, pos, key, key1_offset, key3_offset, ops).is_ok() -} - -// --------------------------------------------------------------------------- -// dd8 page-XOR shift selection. -// -// The packer scrambles ~1 byte per 16-byte block of .text via decrypt_data8, -// keyed by `page_idx << shift` (absolute page index = text_va >> 12). Observed -// shifts are 0 and 15. The shift is NOT stored in any header/config field: -// two otherwise-unrelated builds can carry byte-identical config-version stamps -// (0x40327253) yet require different shifts, so the only reliable discriminator -// is the .text content itself. -// -// Detection scoring formula: for each candidate shift, replay decrypt_data8 -// across a few sample pages (25/50/75% of .text) and count how many of the 255 -// mutated positions become 0xCC — the MSVC int3 padding byte. The correct shift -// hits int3 pads disproportionately often (~3-10x the baseline), so the -// highest-scoring shift wins. If neither shift clears 2x the baseline, .text -// is already plaintext → skip (return 99). -// -// This replaces an earlier entry-stub oracle that matched the 14 fixed CRT-stub -// bytes at the AEP. That oracle false-positived on a newer EXE-64 build: dd8 -// corrupted only the call rel32 (bytes 5-8, the wildcard region), so the stub -// matched under BOTH shifts and the selector defaulted to 0 when the truth was -// 15. The 0xCC statistic samples hundreds of positions per page and is not -// fooled by a stub whose fixed bytes happen to survive. -// --------------------------------------------------------------------------- -pub(crate) fn select_dd8_shift(data: &[u8], text_va: u32, text_size: u32, _info3: u32) -> u32 { - if text_size < 0x1000 { - return 0; - } - let text_off = text_va as usize; - let num_pages_total = text_size >> 12; - - // Sample pages at 25/50/75% of .text, falling back to the midpoint for tiny - // sections. - let mut sample_pages: Vec = Vec::new(); - for frac in [0.25f64, 0.5, 0.75] { - let pg = (num_pages_total as f64 * frac) as u32; - if pg > 0 && pg < num_pages_total { - sample_pages.push(pg); - } - } - if sample_pages.is_empty() && num_pages_total > 1 { - sample_pages.push(num_pages_total / 2); - } - if sample_pages.is_empty() { - return 0; - } - - let none_hits = score_dd8_baseline(data, text_off, &sample_pages); - let s0 = score_dd8_shift(data, text_off, text_va, &sample_pages, 0); - let s15 = score_dd8_shift(data, text_off, text_va, &sample_pages, 15); - let mut best_score = none_hits; - let mut best_shift = 99u32; // 99 == skip dd8 - for (shift, hits) in [(0u32, s0), (15u32, s15)] { - if hits > best_score { - best_score = hits; - best_shift = shift; - } - } - // Require a clear 2x margin over the already-plaintext baseline AND an - // absolute floor. The 2x test alone - // trips on noise when the counts are tiny: an external-companion DLL whose - // .text is already plaintext scores s15=4 vs none=1 — a spurious 4x — and - // gets dd8 wrongly applied, corrupting ~1 byte per 16. Across the whole - // golden corpus every build that genuinely needs dd8 scores >= 10 (lowest - // observed scores at 10-12; up to 107), so a floor of 8 rejects the noise - // while keeping every golden's shift selection unchanged. - const MIN_DD8_HITS: u32 = 8; - if best_shift != 99 && (best_score < none_hits * 2 || best_score < MIN_DD8_HITS) { - best_shift = 99; - } - if std::env::var("SEL_DIAG").is_ok() { - eprintln!( - "SEL dd8 best_shift={} s0={} s15={} none_hits={} samples={:?}", - best_shift, s0, s15, none_hits, sample_pages - ); - } - best_shift -} - -// Baseline: count int3 pads already present at the first byte of each 16-byte -// block, i.e. the positions dd8 would target if its in-block offset were 0. -fn score_dd8_baseline(data: &[u8], text_off: usize, sample_pages: &[u32]) -> u32 { - let mut hits = 0u32; - for &sp in sample_pages { - let pg_off = text_off + (sp as usize) * 0x1000; - if pg_off + 0x1000 > data.len() { - continue; - } - for bi in 1..256usize { - if data[pg_off + bi * 16] == 0xCC { - hits += 1; - } - } - } - hits -} - -// Replay decrypt_data8 on each sample page under `shift` and count how many of -// the 255 mutated positions decode to 0xCC. -fn score_dd8_shift( - data: &[u8], - text_off: usize, - text_va: u32, - sample_pages: &[u32], - shift: u32, -) -> u32 { - let abs_base = text_va >> 12; - let mut hits = 0u32; - for &sp in sample_pages { - let pg_off = text_off + (sp as usize) * 0x1000; - if pg_off + 0x1000 > data.len() { - continue; - } - let abs_page = abs_base.wrapping_add(sp); - let mut key = abs_page << shift; - for bi in 0..256u32 { - let mixed = key.rotate_right(15).wrapping_add(bi); - key = mixed.wrapping_add(bi); - if bi == 0 { - continue; - } - let tidx = (bi.wrapping_mul(16).wrapping_add(mixed & 0xF)) as usize; - if tidx < 0x1000 { - let mutated = data[pg_off + tidx] ^ (key as u8); - if mutated == 0xCC { - hits += 1; - } - } - } - } - hits -} - -// --------------------------------------------------------------------------- -// Tests -// --------------------------------------------------------------------------- - -#[cfg(test)] -mod tests { - use super::*; - - #[test] - fn checked_copy_distinguishes_source_and_destination_ranges() { - let mut short_destination = [0u8; 2]; - let source = [1u8; 4]; - let error = try_copy_from_slice(&mut short_destination, 0, 3, &source, 0) - .expect_err("destination must be rejected"); - assert!(matches!( - error, - super::super::UnpackError::BufferRangeOutOfBounds { - operation: super::super::BufferOperation::CopyDestination, - offset: 0, - size: 3, - buffer_len: 2, - } - )); - - let mut destination = [0u8; 4]; - let short_source = [1u8; 2]; - let error = try_copy_from_slice(&mut destination, 0, 3, &short_source, 0) - .expect_err("source must be rejected"); - assert!(matches!( - error, - super::super::UnpackError::BufferRangeOutOfBounds { - operation: super::super::BufferOperation::CopySource, - offset: 0, - size: 3, - buffer_len: 2, - } - )); - } - - #[test] - fn aes_ks_variant_matches_single_buffer() { - // Random-ish key schedule at ko and data block; both variants must - // produce identical output. - let ko: usize = 0x40; - let mut d = vec![0u8; 0x400]; - let mut x: u32 = 0x12345678; - for b in d.iter_mut() { - x = x.wrapping_mul(1664525).wrapping_add(1013904223); - *b = (x >> 24) as u8; - } - d[ko + 2] = 10; // round count = 10 - d[ko + 3] = 0; - let snap = aes_schedule_snapshot(&d, ko as u32).expect("snapshot"); - - let mut a = d.clone(); - aes_decrypt(&mut a, 0x100, 0x80, ko as u32); - let mut b = d.clone(); - aes_decrypt_ks(&snap, &mut b, 0x100, 0x80); - if a != b { - let idx = (0..a.len()).find(|&i| a[i] != b[i]).unwrap(); - panic!( - "first diff at {idx:#x}: a={:02x} b={:02x}\n a[..]: {:02x?}\n b[..]: {:02x?}", - a[idx], - b[idx], - &a[idx..idx + 16], - &b[idx..idx + 16] - ); - } - } - - #[test] - fn dtbl_variant_matches_single_buffer() { - // Real table + real compressed block lifted from an actual unpack is - // covered by the golden suite; here we just check a trivial stream: - // build a table where every byte is a literal (mode 0, 8 bits), then - // a source stream of N bytes should expand to N identical bytes. - let ko: usize = 0x100; - let mut d = vec![0u8; 0x1000]; - for e in 0..256usize { - let off = ko + e * 3; - let sym = 0x8000u16 | (e as u16 & 0xFF); // terminal, mode 0, payload=e - d[off] = (sym & 0xFF) as u8; - d[off + 1] = (sym >> 8) as u8; - d[off + 2] = 8; // 8 bits per symbol - } - // Source: 16 bytes 0x00..0x0F at src. - let src = 0x600u32; - for i in 0..16u32 { - d[(src + i) as usize] = i as u8; - } - let snap = huffman_table_snapshot(&d, ko as u32).expect("table snapshot"); - - let mut a = vec![0u8; 0x1000]; - a[..d.len()].copy_from_slice(&d); - assert!(decompress(&mut a, src, 0x800, ko as u32, 16, 16)); - let mut b = d.clone(); - assert!(decompress_tbl(&snap, &mut b, src, 0x800, 16, 16)); - assert_eq!(&a[0x800..0x810], &b[0x800..0x810]); - assert_eq!(&b[0x800..0x810], &(0u8..16).collect::>()[..]); - } - - /// Task 4.1 regression: build a synthetic buffer whose valid bytecode block - /// sits PAST `len` but within `len*2`. Assert that the smaller window misses - /// it and the doubled window finds it. - #[test] - fn bytecode_locate_double_window_retry() { - // We place the block at offset (base + len + 16) which is inside - // the len*2 window but outside the len window. - let base: u32 = 0; - let len: u32 = 256; - // Block sits at base + len + 16 = 272, aligned to 16. - let block_pos: usize = (base + len + 16) as usize; // 272 - - // The buffer must be large enough for the block (block_pos + 96 bytes). - let buf_len = block_pos + 256; - let mut buf = vec![0u8; buf_len]; - - // Build a valid plaintext op stream: - // [4, 0, 4, 0, 4, 0, 4, 0, 195] (4 ADD-AL ops then RET) - // Padded to 10 bytes total; count >= 8. - let count: usize = 10; - let mut plain = [0u8; 256]; - plain[0] = 4; - plain[1] = 0; - plain[2] = 4; - plain[3] = 0; - plain[4] = 4; - plain[5] = 0; - plain[6] = 4; - plain[7] = 0; - plain[8] = 195; // ret - - // Compute the LFSR keystream and XOR the first `count` bytes to get the - // encrypted representation that the scanner would decrypt back. - let mut ks = [0u8; 256]; - lfsr_keystream(&mut ks); - for i in 0..count { - buf[block_pos + i] = plain[i] ^ ks[i]; - } - // Raw count byte at block_pos+95 (outside the XOR range since count=10 < 95). - buf[block_pos + 95] = count as u8; - - // Verify our construction: find_bytecode_offset with len should NOT find it. - assert_eq!( - find_bytecode_offset(&buf, base, len), - None, - "smaller window should not find the block" - ); - - // The doubled window should find it at block_pos. - assert_eq!( - find_bytecode_offset(&buf, base, len.saturating_mul(2)), - Some(block_pos as u32), - "doubled window should locate the block" - ); - } - - /// Review regression: a run-fill token with a unit width other than 1/2/4 - /// comes from a corrupt stream and must report failure — previously it - /// wrote nothing yet still counted the bytes as written, leaving stale - /// holes that later stages treated as plaintext. - #[test] - fn decompress_rejects_unknown_run_fill_width() { - // Huffman table at key_offset 0, entry 0: terminal symbol with - // mode 0x200 (run-fill), payload 3 (invalid width), code length 8. - let mut d = vec![0u8; 0x100]; - let sym: u16 = 0x8000 | 0x203; - d[0..2].copy_from_slice(&sym.to_le_bytes()); - d[2] = 8; - // All-zero source -> symbol index 0 -> the invalid run-fill. - assert_eq!( - decompress_detailed(&mut d, 0x40, 0x80, 0, 4, 3), - Err(super::super::DecompressionFailure::InvalidRunFillWidth { width: 3 }) - ); - } - - /// Control for the above: a width-1 run-fill is legal and succeeds. - #[test] - fn decompress_accepts_width1_run_fill() { - let mut d = vec![0u8; 0x100]; - d[0x7F] = 0x5A; // unit to replicate - let sym: u16 = 0x8000 | 0x201; - d[0..2].copy_from_slice(&sym.to_le_bytes()); - d[2] = 8; - assert!(decompress(&mut d, 0x40, 0x80, 0, 4, 3)); - assert_eq!(&d[0x80..0x83], &[0x5A, 0x5A, 0x5A]); - } - - /// Seed the first `count` dd8-targeted positions of each sampled page with - /// the byte that decodes to `0xCC` under the `page+1` formula — i.e. an - /// encrypted `.text` whose plaintext is int3 padding. Positions whose key - /// byte would make the *ciphertext* itself `0xCC` are skipped so the - /// fixture contains no `0xCC` at all and every post-dd8 `0xCC` is a genuine - /// gain over a zero baseline. - fn seed_dd8_int3(data: &mut [u8], text_off: u32, pages: &[u32], count: u32) { - for &sp in pages { - let pg_off = (text_off + sp * 0x1000) as usize; - let mut k = sp.wrapping_add(1); - k = k.rotate_right(15); - let mut planted = 0u32; - for bi in 1..256u32 { - let ri = k.rotate_right(15).wrapping_add(bi); - k = ri.wrapping_add(bi); - if planted >= count { - continue; - } - let ct = 0xCCu8 ^ (k as u8); - if ct == 0xCC { - continue; - } - let tidx = (bi.wrapping_mul(16).wrapping_add(ri & 0xF)) as usize; - data[pg_off + tidx] = ct; - planted += 1; - } - } - } - - /// Review regression: a near-plaintext `.text` must NOT be dd8-decrypted. - /// dd8 XORs 255 positions per page with pseudo-random bytes, so it - /// manufactures a few `0xCC` for free — under the old bare - /// `best > baseline` test any positive gain was enough to "apply" dd8 and - /// scramble ~1 byte per 16 of a native DLL's already-plaintext code, - /// silently (nothing downstream, including the integrity check, notices). - /// Here the gain is real but small; the floor must still reject it. - #[test] - fn pe32_dd8_skips_text_whose_gain_is_only_noise_sized() { - let text_off: u32 = 0x1000; - let text_size: u32 = 8 * 0x1000; - let mut data = vec![0u8; (text_off + text_size) as usize]; - seed_dd8_int3(&mut data, text_off, &[2, 4, 6], 5); - assert!( - !data.contains(&0xCC), - "fixture must have a zero 0xCC baseline" - ); - assert_eq!( - select_dd8_formula_pe32(&data, text_off, text_size), - None, - "a gain this small is indistinguishable from dd8's own noise" - ); - } - - /// Control for the above: a `.text` whose dd8 pass restores a large amount - /// of int3 padding clears the floor and is decrypted. Same fixture shape, - /// only the amount of restored padding differs. - #[test] - fn pe32_dd8_applies_when_padding_is_restored() { - let text_off: u32 = 0x1000; - let text_size: u32 = 8 * 0x1000; - let mut data = vec![0u8; (text_off + text_size) as usize]; - seed_dd8_int3(&mut data, text_off, &[2, 4, 6], 255); - assert_eq!( - select_dd8_formula_pe32(&data, text_off, text_size), - Some(false), - "encrypted .text must be decrypted with the page+1 formula" - ); - } - - /// Review regression: a zero last-section VA (corrupt section table) must - /// bail instead of building .kmiat at RVA 0 — the old code zeroed - /// `[0, 0x7000)`, wiping the DOS/PE headers, and returned the broken image - /// as a success. A near-2 GiB VA must likewise refuse to grow the image - /// past [`super::MAX_IMAGE_SIZE`]. - #[test] - fn kmiat_bogus_section_va_bails_without_wiping_headers() { - for last_sec_va in [0u32, 0x5000_0000] { - let pe: u32 = 0x80; - let mut data = vec![0xAAu8; 0x8000]; - // COFF header: 1 section, optional header size 0xE0 (PE32). - write_u16(&mut data, pe + 6, 1); - write_u16(&mut data, pe + 20, 0xE0); - // Import directory at pe+0x80: one descriptor + null terminator. - write_u32(&mut data, pe + 0x80, 0x1100); - write_u32(&mut data, pe + 0x84, 0x28); - write_u32(&mut data, 0x1100, 0x1200); // OFT rva - write_u32(&mut data, 0x1100 + 12, 0x1300); // name rva - write_u32(&mut data, 0x1100 + 16, 0x1400); // IAT rva - for b in &mut data[0x1100 + 20..0x1100 + 40] { - *b = 0; // null terminator descriptor - } - data[0x1300..0x1300 + 13].copy_from_slice(b"KERNEL32.dll\0"); - write_u32(&mut data, 0x1200, 0x1500); // thunk -> hint/name - write_u32(&mut data, 0x1204, 0); // thunk terminator - data[0x1500..0x1502].copy_from_slice(&0u16.to_le_bytes()); - data[0x1502..0x1502 + 12].copy_from_slice(b"ExitProcess\0"); - // Section table at pe+24+0xE0 = 0x178; VA field at +12. - write_u32(&mut data, 0x178 + 12, last_sec_va); - - let head_before: Vec = data[..0x400].to_vec(); - let len_before = data.len(); - move_pe32_imports_to_kmiat(&mut data, pe); - assert_eq!( - data.len(), - len_before, - "VA 0x{last_sec_va:08X}: image must not grow" - ); - assert_eq!( - &data[..0x400], - &head_before[..], - "VA 0x{last_sec_va:08X}: headers must be untouched" - ); - } - } -} diff --git a/tests/common/mod.rs b/tests/common/mod.rs deleted file mode 100644 index b598c3f..0000000 --- a/tests/common/mod.rs +++ /dev/null @@ -1,10 +0,0 @@ -//! Shared test fixtures. -#![allow(dead_code)] - -use std::path::PathBuf; - -/// Path to `senbei/samples` — the user-managed corpus dropped in by hand. -/// Git-ignored except its README; tests here run against whatever is present. -pub fn samples_dir() -> PathBuf { - PathBuf::from(env!("CARGO_MANIFEST_DIR")).join("samples") -} diff --git a/tests/job.rs b/tests/job.rs deleted file mode 100644 index 8fb8a9c..0000000 --- a/tests/job.rs +++ /dev/null @@ -1,31 +0,0 @@ -use senbei::job::{default_out_root_for_file, out_name}; -use std::path::Path; - -#[test] -fn out_name_inserts_unpack_before_last_dot() { - assert_eq!(out_name(Path::new("foo.exe")), Path::new("foo.unpack.exe")); - assert_eq!( - out_name(Path::new("a/b/bar.dll")), - Path::new("a/b/bar.unpack.dll") - ); - assert_eq!(out_name(Path::new("x.y.dll")), Path::new("x.y.unpack.dll")); -} - -#[test] -fn out_name_no_dot_appends_unpack() { - assert_eq!(out_name(Path::new("nodot")), Path::new("nodot.unpack")); -} - -#[test] -fn default_out_root_for_file_is_parent_unpack() { - assert_eq!( - default_out_root_for_file(Path::new("a/b/foo.exe")), - Path::new("a/b/unpack") - ); -} - -#[test] -fn default_out_root_for_file_cwd_when_no_parent() { - let p = default_out_root_for_file(Path::new("foo.exe")); - assert_eq!(p, Path::new(".").join("unpack")); -} diff --git a/tests/logfile.rs b/tests/logfile.rs deleted file mode 100644 index c8250ea..0000000 --- a/tests/logfile.rs +++ /dev/null @@ -1,47 +0,0 @@ -use senbei::logfile::{Log, local_stamp_compact, local_stamp_display}; - -#[test] -fn local_stamp_compact_matches_shape() { - let s = local_stamp_compact(); - // YYYYMMDD-HHMMSS → 15 chars, digit groups around dash - assert_eq!(s.len(), 15, "got {s}"); - assert_eq!(&s[8..9], "-"); - assert!(s.as_bytes().iter().enumerate().all(|(i, b)| { - if i == 8 { - *b == b'-' - } else { - b.is_ascii_digit() - } - })); -} - -#[test] -fn local_stamp_display_matches_shape() { - let s = local_stamp_display(); - // YYYY-MM-DD HH:MM:SS → 19 chars - assert_eq!(s.len(), 19, "got {s}"); - assert_eq!(&s[4..5], "-"); - assert_eq!(&s[7..8], "-"); - assert_eq!(&s[10..11], " "); - assert_eq!(&s[13..14], ":"); - assert_eq!(&s[16..17], ":"); -} - -#[test] -fn log_writes_timestamped_file_in_target_dir() { - let td = tempfile::tempdir().unwrap(); - let log = Log::create(td.path()).unwrap(); - log.step("hello"); - let path = log.path().to_path_buf(); - drop(log); - assert!(path.starts_with(td.path())); - let name = path.file_name().unwrap().to_string_lossy(); - assert!( - name.starts_with("senbei-") && name.ends_with(".log"), - "unexpected log name: {name}" - ); - // senbei-YYYYMMDD-HHMMSS.log - let core = name.trim_start_matches("senbei-").trim_end_matches(".log"); - assert_eq!(core.len(), 15, "stamp in name: {name}"); - assert!(std::fs::read_to_string(&path).unwrap().contains("hello")); -} diff --git a/tests/run_log.rs b/tests/run_log.rs deleted file mode 100644 index 3a909e7..0000000 --- a/tests/run_log.rs +++ /dev/null @@ -1,80 +0,0 @@ -use senbei::job; -use std::path::Path; - -fn list_logs(dir: &Path) -> Vec { - std::fs::read_dir(dir) - .into_iter() - .flatten() - .filter_map(|e| e.ok()) - .map(|e| e.path()) - .filter(|p| { - p.file_name() - .and_then(|n| n.to_str()) - .map(|n| n.starts_with("senbei-") && n.ends_with(".log")) - .unwrap_or(false) - }) - .collect() -} - -#[test] -fn run_file_no_log_creates_no_logfile() { - let td = tempfile::tempdir().unwrap(); - let input = td.path().join("not_crackproof.bin"); - std::fs::write(&input, b"not a pe").unwrap(); - let out = td.path().join("out"); - let s = job::run_file_v(&input, Some(&out), 2, false, true).unwrap(); - assert_eq!(s.errors, 1); - // With no_log, no senbei-*.log under out (even if the dir was created). - assert!(list_logs(&out).is_empty()); -} - -#[test] -fn run_file_writes_log_under_out_with_header_footer() { - let td = tempfile::tempdir().unwrap(); - let input = td.path().join("not_crackproof.bin"); - std::fs::write(&input, b"not a pe").unwrap(); - let out = td.path().join("out"); - let s = job::run_file_v(&input, Some(&out), 2, false, false).unwrap(); - assert_eq!(s.errors, 1); - let logs = list_logs(&out); - assert_eq!(logs.len(), 1, "expected one log under out, got {logs:?}"); - let text = std::fs::read_to_string(&logs[0]).unwrap(); - assert!(text.contains("Senbei "), "header version: {text}"); - assert!(text.contains("started "), "{text}"); - assert!(text.contains("input "), "{text}"); - assert!(text.contains("out "), "{text}"); - assert!(text.contains("ERR "), "{text}"); - assert!(text.contains("done in "), "{text}"); - assert!(text.contains("summary:"), "{text}"); -} - -#[test] -fn run_file_default_out_root_is_parent_unpack() { - let td = tempfile::tempdir().unwrap(); - let input = td.path().join("not_crackproof.bin"); - std::fs::write(&input, b"not a pe").unwrap(); - let _ = job::run_file_v(&input, None, 2, false, false).unwrap(); - let unpack = td.path().join("unpack"); - assert!(unpack.is_dir()); - assert_eq!(list_logs(&unpack).len(), 1); - // log must NOT be next to input's parent root without unpack - assert!(list_logs(td.path()).is_empty()); -} - -#[test] -fn run_folder_log_lives_under_out_not_root() { - let td = tempfile::tempdir().unwrap(); - // empty tree: 0 candidates still creates log under unpack - let s = job::run_folder_v(td.path(), None, 2, false, false).unwrap(); - assert_eq!(s.unpacked, 0); - let unpack = td.path().join("unpack"); - assert!(unpack.is_dir()); - assert_eq!(list_logs(&unpack).len(), 1); - assert!( - list_logs(td.path()).is_empty(), - "log must not sit on input root" - ); - let text = std::fs::read_to_string(&list_logs(&unpack)[0]).unwrap(); - assert!(text.contains("done in ")); - assert!(text.contains("summary:")); -} diff --git a/tests/samples.rs b/tests/samples.rs deleted file mode 100644 index 782c413..0000000 --- a/tests/samples.rs +++ /dev/null @@ -1,229 +0,0 @@ -//! Corpus test over the user-managed `senbei/samples` folder. -//! -//! Drop real Crackproof `*.exe` / `*.dll` inputs in there (and/or il2cpp -//! `*.dat` metadata blobs), optionally alongside a byte-exact golden named -//! `.golden.`. Each input is processed and classified: -//! -//! - golden present, bytes identical -> pass (silent) -//! - golden present, bytes differ -> FAIL (the test fails) -//! - no golden -> WARNING (printed; needs a manual check) -//! -//! Inputs go through [`senbei::job::unpack_bytes`], the same routing the CLI -//! uses, **not** `unpack_auto` directly. That matters: `unpack_auto` alone -//! cannot reach the external-companion layout, whose stub is meaningless -//! without its `._` payload — a corpus wired to `unpack_auto` silently -//! covers none of the splice / export-overlay / TLS-restore code, nor the -//! marker-less "new layout" those builds use. A `._` sibling in the -//! samples folder is picked up automatically, exactly as it is on disk. -//! -//! An input whose bytes carry the il2cpp metadata magic is routed through -//! [`senbei::metadata::deobfuscate`] instead, giving the method-token remap -//! real-world coverage (its unit tests only build synthetic layouts). -//! -//! The folder is git-ignored (see `senbei/samples/README.md`), so the set of -//! samples is whatever happens to be on the machine. An empty/absent folder is -//! a no-op pass. - -mod common; -use common::samples_dir; -use std::path::Path; -use walkdir::WalkDir; - -/// An input is a `.exe`/`.dll`/`.dat` whose name doesn't carry the `.golden.` -/// marker — those are goldens, not inputs. External companions (`._`) -/// have extension `_` and are therefore never inputs in their own right; they -/// are consumed by their base module. -fn is_input(path: &Path) -> bool { - let Some(ext) = path.extension().and_then(|e| e.to_str()) else { - return false; - }; - let ext = ext.to_ascii_lowercase(); - if ext != "exe" && ext != "dll" && ext != "dat" { - return false; - } - // Reject goldens like `foo.golden.exe`. - !path - .file_name() - .and_then(|n| n.to_str()) - .map(|n| n.to_ascii_lowercase().contains(".golden.")) - .unwrap_or(false) -} - -/// Golden path for an input: `.golden.` next to it. -fn golden_for(input: &Path) -> std::path::PathBuf { - let ext = input.extension().and_then(|e| e.to_str()).unwrap_or(""); - let stem = input.file_stem().and_then(|s| s.to_str()).unwrap_or(""); - input.with_file_name(format!("{stem}.golden.{ext}")) -} - -/// External-companion path for an input: `._` next to it, -/// matching what the CLI looks for on disk. -fn companion_for(input: &Path) -> Option { - let name = input.file_name()?; - let mut n = name.to_os_string(); - n.push("._"); - let p = input.with_file_name(n); - p.is_file().then_some(p) -} - -#[test] -fn samples_unpack_against_goldens() { - let dir = samples_dir(); - // An absent/empty corpus fails only when explicitly required — a green - // run that unpacked nothing hides every unpack regression, but on public - // CI there is no corpus at all (binaries are never committed), so the - // gate is opt-in via SENBEI_REQUIRE_SAMPLES rather than implied by CI. - // Locally the corpus is the user-managed samples/ folder (see - // samples/README.md). - let require = std::env::var_os("SENBEI_REQUIRE_SAMPLES").is_some(); - if !dir.is_dir() { - assert!( - !require, - "samples: {} does not exist — corpus required (CI)", - dir.display() - ); - eprintln!("samples: {} does not exist, nothing to test", dir.display()); - return; - } - - let mut inputs: Vec<_> = WalkDir::new(&dir) - .follow_links(false) - .into_iter() - .filter_entry(|entry| { - entry.depth() == 0 - || !entry - .file_name() - .to_str() - .is_some_and(|name| name.eq_ignore_ascii_case("unpack")) - }) - .map(|entry| entry.unwrap_or_else(|e| panic!("walk {}: {e}", dir.display()))) - .filter(|entry| entry.file_type().is_file() && is_input(entry.path())) - .map(|entry| entry.into_path()) - .collect(); - inputs.sort(); - - if inputs.is_empty() { - assert!( - !require, - "samples: no .exe/.dll inputs in {} — corpus required (CI)", - dir.display() - ); - eprintln!("samples: no .exe/.dll inputs in {}", dir.display()); - return; - } - - let mut passed = 0usize; - let mut warnings: Vec = Vec::new(); - let mut failures: Vec = Vec::new(); - - for input in &inputs { - let name = input - .strip_prefix(&dir) - .unwrap_or(input) - .to_string_lossy() - .to_string(); - let bytes = match std::fs::read(input) { - Ok(b) => b, - Err(e) => { - failures.push(format!("{name}: read error: {e}")); - continue; - } - }; - - let got = if senbei::metadata::is_metadata(&bytes) { - // il2cpp metadata: method-token de-obfuscation, no PE pipeline and - // no integrity check (the output is not a PE image). - match senbei::metadata::deobfuscate(&bytes) { - Ok((out, _report)) => out, - Err(e) => { - failures.push(format!("{name}: de-obfuscation failed: {e}")); - continue; - } - } - } else { - // Splice in the external companion when one sits next to the input, - // then run the CLI's routing (which also overlays the stub's export - // table and TLS directory for spliced inputs). - let companion = match companion_for(input) { - Some(p) => match std::fs::read(&p) { - Ok(b) => Some(b), - Err(e) => { - failures.push(format!("{name}: companion read error: {e}")); - continue; - } - }, - None => None, - }; - let image = match senbei::job::unpack_bytes(&bytes, companion.as_deref()) { - Ok(img) => img, - Err(e) => { - failures.push(format!("{name}: unpack failed: {e:?}")); - continue; - } - }; - // The static integrity check is a second, golden-independent gate: - // it catches an output that is structurally plausible but would - // crash at runtime (0xC0000005) even when a stale golden still - // byte-matches. (Goldens are byte comparisons only — "matches - // golden" ≠ runs.) - if !image.integrity.ok() { - failures.push(format!( - "{name}: integrity check failed: {}", - image.integrity.issues.join("; ") - )); - continue; - } - image.bytes - }; - - let golden = golden_for(input); - if !golden.exists() { - warnings.push(format!( - "{name}: unpacked OK ({} bytes) but no golden ({}) — MANUAL CHECK", - got.len(), - golden.file_name().unwrap().to_string_lossy() - )); - continue; - } - - let want = match std::fs::read(&golden) { - Ok(b) => b, - Err(e) => { - failures.push(format!("{name}: golden read error: {e}")); - continue; - } - }; - - if got.len() != want.len() { - failures.push(format!( - "{name}: length differs: got {} want {}", - got.len(), - want.len() - )); - continue; - } - if let Some((i, (a, b))) = got.iter().zip(&want).enumerate().find(|(_, (a, b))| a != b) { - failures.push(format!( - "{name}: first diff at 0x{i:X}: got {a:02X} want {b:02X}" - )); - continue; - } - passed += 1; - } - - eprintln!( - "samples: {} input(s) — {} pass, {} warning(s), {} failure(s)", - inputs.len(), - passed, - warnings.len(), - failures.len() - ); - for w in &warnings { - eprintln!(" WARN {w}"); - } - for f in &failures { - eprintln!(" FAIL {f}"); - } - - assert!(failures.is_empty(), "{} sample(s) failed", failures.len()); -} diff --git a/web/Cargo.lock b/web/Cargo.lock deleted file mode 100644 index 5d0237d..0000000 --- a/web/Cargo.lock +++ /dev/null @@ -1,441 +0,0 @@ -# This file is automatically @generated by Cargo. -# It is not intended for manual editing. -version = 4 - -[[package]] -name = "anyhow" -version = "1.0.104" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "330a5ed07fa54e4702c9d6c4174f74427fc0ef6e214bbd677ae50a5099946470" - -[[package]] -name = "bumpalo" -version = "3.20.3" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "72f5acc6cb2ba439de613abc23857ec3d78374d8ed5ac84e9d11336e87da8649" - -[[package]] -name = "cfg-if" -version = "1.0.4" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "9330f8b2ff13f34540b44e946ef35111825727b38d33286ef986142615121801" - -[[package]] -name = "console" -version = "0.16.4" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "4fe5f465a4f6fee88fad41b85d990f84c835335e85b5d9e6e63e0d06d28cba7c" -dependencies = [ - "encode_unicode", - "libc", - "unicode-width", - "windows-sys", -] - -[[package]] -name = "console_error_panic_hook" -version = "0.1.7" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "a06aeb73f470f66dcdbf7223caeebb85984942f22f1adb2a088cf9668146bbbc" -dependencies = [ - "cfg-if", - "wasm-bindgen", -] - -[[package]] -name = "encode_unicode" -version = "1.0.0" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "34aa73646ffb006b8f5147f3dc182bd4bcb190227ce861fc4a4844bf8e3cb2c0" - -[[package]] -name = "futures-core" -version = "0.3.33" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "2cd50c473c80f6d7c3670a752354b8e569b1a7cbfdc0419ec88e5edad85e0dc7" - -[[package]] -name = "futures-task" -version = "0.3.33" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "b231ed28831efb4a61a08580c4bc233ec56bc009f4cd8f52da2c3cb97df0c109" - -[[package]] -name = "futures-util" -version = "0.3.33" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "a77a90a256fce34da66415271e30f94ee91c57b04b8a2c042d9cf3220179deaa" -dependencies = [ - "futures-core", - "futures-task", - "pin-project-lite", - "slab", -] - -[[package]] -name = "indicatif" -version = "0.18.6" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "9433806cd6b4ec1aba79c021c7e4c58fb4c3b9977c085062e611ac929998fb0c" -dependencies = [ - "console", - "portable-atomic", - "unicode-width", - "unit-prefix", - "web-time", -] - -[[package]] -name = "js-sys" -version = "0.3.103" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "53b44bfcdb3f8d5837a46dae1ca9660a837176eee74a28b229bc626816589102" -dependencies = [ - "cfg-if", - "futures-util", - "wasm-bindgen", -] - -[[package]] -name = "libc" -version = "0.2.189" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "3eaf3ede3fee6db1a4c2ee091bf8a8b4dccdc6d17f656fb07896ee72867612f2" - -[[package]] -name = "once_cell" -version = "1.21.4" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50" - -[[package]] -name = "owo-colors" -version = "4.3.0" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "d211803b9b6b570f68772237e415a029d5a50c65d382910b879fb19d3271f94d" - -[[package]] -name = "pin-project-lite" -version = "0.2.17" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "a89322df9ebe1c1578d689c92318e070967d1042b512afbe49518723f4e6d5cd" - -[[package]] -name = "portable-atomic" -version = "1.14.0" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "3d20d5497ef88037a52ff98267d066e7f11fcc5e99bbfbd58a42336193aacec3" - -[[package]] -name = "proc-macro2" -version = "1.0.107" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "985e7ec9bb745e6ce6535b544d84d6cd6f7ad8bd711c398938ae983b91a766d9" -dependencies = [ - "unicode-ident", -] - -[[package]] -name = "quote" -version = "1.0.47" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "1fbf4db142a473a8d80c26bbf18454ed458bf8d26c8219c331daecfdbd079001" -dependencies = [ - "proc-macro2", -] - -[[package]] -name = "rustversion" -version = "1.0.23" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "cf54715a573b99ac80df0bc206da022bcd442c974952c7b9720069370852e21f" - -[[package]] -name = "same-file" -version = "1.0.6" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502" -dependencies = [ - "winapi-util", -] - -[[package]] -name = "senbei" -version = "1.0.0" -dependencies = [ - "anyhow", - "indicatif", - "libc", - "owo-colors", - "thiserror", - "walkdir", - "windows", -] - -[[package]] -name = "senbei-web" -version = "1.0.0" -dependencies = [ - "console_error_panic_hook", - "senbei", - "wasm-bindgen", -] - -[[package]] -name = "slab" -version = "0.4.12" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "0c790de23124f9ab44544d7ac05d60440adc586479ce501c1d6d7da3cd8c9cf5" - -[[package]] -name = "syn" -version = "2.0.119" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "872831b642d1a07999a962a351ed35b955ea2cfc8f3862091e2a240a84f17297" -dependencies = [ - "proc-macro2", - "quote", - "unicode-ident", -] - -[[package]] -name = "syn" -version = "3.0.3" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "53e9bae58849f64dfa4f5d5ae372c8341f7305f82a3868709269343628b659a3" -dependencies = [ - "proc-macro2", - "quote", - "unicode-ident", -] - -[[package]] -name = "thiserror" -version = "2.0.19" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "09a43598840e33d5b0331f38c5e30d13bb11c11210a4b58f0d9b18a5a5eefcd9" -dependencies = [ - "thiserror-impl", -] - -[[package]] -name = "thiserror-impl" -version = "2.0.19" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "43cbfe0cf76104d42a574802844187e84a305e531ed54455f11fbde0f10541cd" -dependencies = [ - "proc-macro2", - "quote", - "syn 3.0.3", -] - -[[package]] -name = "unicode-ident" -version = "1.0.24" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "e6e4313cd5fcd3dad5cafa179702e2b244f760991f45397d14d4ebf38247da75" - -[[package]] -name = "unicode-width" -version = "0.2.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "b4ac048d71ede7ee76d585517add45da530660ef4390e49b098733c6e897f254" - -[[package]] -name = "unit-prefix" -version = "0.5.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "81e544489bf3d8ef66c953931f56617f423cd4b5494be343d9b9d3dda037b9a3" - -[[package]] -name = "walkdir" -version = "2.5.0" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "29790946404f91d9c5d06f9874efddea1dc06c5efe94541a7d6863108e3a5e4b" -dependencies = [ - "same-file", - "winapi-util", -] - -[[package]] -name = "wasm-bindgen" -version = "0.2.126" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "4b067c0c11094aef6b7a801c1e34a26affafdf3d051dba08456b868789aaf9a4" -dependencies = [ - "cfg-if", - "once_cell", - "rustversion", - "wasm-bindgen-macro", - "wasm-bindgen-shared", -] - -[[package]] -name = "wasm-bindgen-macro" -version = "0.2.126" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "167ce5e579f6bcf889c4f7175a8a5a585de84e8ff93976ce393efa5f2837aab1" -dependencies = [ - "quote", - "wasm-bindgen-macro-support", -] - -[[package]] -name = "wasm-bindgen-macro-support" -version = "0.2.126" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "f3997c7839262f4ef12cf90b818d6340c18e80f263f1a94bf157d0ec4420380e" -dependencies = [ - "bumpalo", - "proc-macro2", - "quote", - "syn 2.0.119", - "wasm-bindgen-shared", -] - -[[package]] -name = "wasm-bindgen-shared" -version = "0.2.126" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "dc1b4cb0cc549fcf58d7dfc081778139b3d283a081644e833e84682ad71cea24" -dependencies = [ - "unicode-ident", -] - -[[package]] -name = "web-time" -version = "1.1.0" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "5a6580f308b1fad9207618087a65c04e7a10bc77e02c8e84e9b00dd4b12fa0bb" -dependencies = [ - "js-sys", - "wasm-bindgen", -] - -[[package]] -name = "winapi-util" -version = "0.1.11" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22" -dependencies = [ - "windows-sys", -] - -[[package]] -name = "windows" -version = "0.62.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "527fadee13e0c05939a6a05d5bd6eec6cd2e3dbd648b9f8e447c6518133d8580" -dependencies = [ - "windows-collections", - "windows-core", - "windows-future", - "windows-numerics", -] - -[[package]] -name = "windows-collections" -version = "0.3.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "23b2d95af1a8a14a3c7367e1ed4fc9c20e0a26e79551b1454d72583c97cc6610" -dependencies = [ - "windows-core", -] - -[[package]] -name = "windows-core" -version = "0.62.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "b8e83a14d34d0623b51dce9581199302a221863196a1dde71a7663a4c2be9deb" -dependencies = [ - "windows-implement", - "windows-interface", - "windows-link", - "windows-result", - "windows-strings", -] - -[[package]] -name = "windows-future" -version = "0.3.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "e1d6f90251fe18a279739e78025bd6ddc52a7e22f921070ccdc67dde84c605cb" -dependencies = [ - "windows-core", - "windows-link", - "windows-threading", -] - -[[package]] -name = "windows-implement" -version = "0.60.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "053e2e040ab57b9dc951b72c264860db7eb3b0200ba345b4e4c3b14f67855ddf" -dependencies = [ - "proc-macro2", - "quote", - "syn 2.0.119", -] - -[[package]] -name = "windows-interface" -version = "0.59.3" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "3f316c4a2570ba26bbec722032c4099d8c8bc095efccdc15688708623367e358" -dependencies = [ - "proc-macro2", - "quote", - "syn 2.0.119", -] - -[[package]] -name = "windows-link" -version = "0.2.1" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "f0805222e57f7521d6a62e36fa9163bc891acd422f971defe97d64e70d0a4fe5" - -[[package]] -name = "windows-numerics" -version = "0.3.1" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "6e2e40844ac143cdb44aead537bbf727de9b044e107a0f1220392177d15b0f26" -dependencies = [ - "windows-core", - "windows-link", -] - -[[package]] -name = "windows-result" -version = "0.4.1" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "7781fa89eaf60850ac3d2da7af8e5242a5ea78d1a11c49bf2910bb5a73853eb5" -dependencies = [ - "windows-link", -] - -[[package]] -name = "windows-strings" -version = "0.5.1" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "7837d08f69c77cf6b07689544538e017c1bfcf57e34b4c0ff58e6c2cd3b37091" -dependencies = [ - "windows-link", -] - -[[package]] -name = "windows-sys" -version = "0.61.2" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "ae137229bcbd6cdf0f7b80a31df61766145077ddf49416a728b02cb3921ff3fc" -dependencies = [ - "windows-link", -] - -[[package]] -name = "windows-threading" -version = "0.2.1" -source = "registry+https://github.com/rust-lang/crates.io-index" -checksum = "3949bd5b99cafdf1c7ca86b43ca564028dfe27d66958f2470940f73d86d75b37" -dependencies = [ - "windows-link", -] diff --git a/web/Cargo.toml b/web/Cargo.toml deleted file mode 100644 index 1ce74ec..0000000 --- a/web/Cargo.toml +++ /dev/null @@ -1,19 +0,0 @@ -[package] -name = "senbei-web" -version = "1.0.0" -edition = "2024" -description = "WebAssembly browser frontend for senbei" -license = "AGPL-3.0-only" - -[lib] -crate-type = ["cdylib"] - -[dependencies] -senbei = { path = ".." } -wasm-bindgen = "0.2" -console_error_panic_hook = "0.1" - -[profile.release] -opt-level = "z" -lto = true -codegen-units = 1 diff --git a/web/LICENSE b/web/LICENSE deleted file mode 100644 index fe6b903..0000000 --- a/web/LICENSE +++ /dev/null @@ -1,662 +0,0 @@ - GNU AFFERO GENERAL PUBLIC LICENSE - Version 3, 19 November 2007 - - Copyright (C) 2007 Free Software Foundation, Inc. - Everyone is permitted to copy and distribute verbatim copies - of this license document, but changing it is not allowed. - - Preamble - - The GNU Affero General Public License is a free, copyleft license for -software and other kinds of works, specifically designed to ensure -cooperation with the community in the case of network server software. - - The licenses for most software and other practical works are designed -to take away your freedom to share and change the works. By contrast, -our General Public Licenses are intended to guarantee your freedom to -share and change all versions of a program--to make sure it remains free -software for all its users. - - When we speak of free software, we are referring to freedom, not -price. Our General Public Licenses are designed to make sure that you -have the freedom to distribute copies of free software (and charge for -them if you wish), that you receive source code or can get it if you -want it, that you can change the software or use pieces of it in new -free programs, and that you know you can do these things. - - Developers that use our General Public Licenses protect your rights -with two steps: (1) assert copyright on the software, and (2) offer -you this License which gives you legal permission to copy, distribute -and/or modify the software. - - A secondary benefit of defending all users' freedom is that -improvements made in alternate versions of the program, if they -receive widespread use, become available for other developers to -incorporate. Many developers of free software are heartened and -encouraged by the resulting cooperation. However, in the case of -software used on network servers, this result may fail to come about. -The GNU General Public License permits making a modified version and -letting the public access it on a server without ever releasing its -source code to the public. - - The GNU Affero General Public License is designed specifically to -ensure that, in such cases, the modified source code becomes available -to the community. It requires the operator of a network server to -provide the source code of the modified version running there to the -users of that server. Therefore, public use of a modified version, on -a publicly accessible server, gives the public access to the source -code of the modified version. - - An older license, called the Affero General Public License and -published by Affero, was designed to accomplish similar goals. This is -a different license, not a version of the Affero GPL, but Affero has -released a new version of the Affero GPL which permits relicensing under -this license. - - The precise terms and conditions for copying, distribution and -modification follow. - - TERMS AND CONDITIONS - - 0. Definitions. - - "This License" refers to version 3 of the GNU Affero General Public License. - - "Copyright" also means copyright-like laws that apply to other kinds of -works, such as semiconductor masks. - - "The Program" refers to any copyrightable work licensed under this -License. Each licensee is addressed as "you". "Licensees" and -"recipients" may be individuals or organizations. - - To "modify" a work means to copy from or adapt all or part of the work -in a fashion requiring copyright permission, other than the making of an -exact copy. The resulting work is called a "modified version" of the -earlier work or a work "based on" the earlier work. - - A "covered work" means either the unmodified Program or a work based -on the Program. - - To "propagate" a work means to do anything with it that, without -permission, would make you directly or secondarily liable for -infringement under applicable copyright law, except executing it on a -computer or modifying a private copy. Propagation includes copying, -distribution (with or without modification), making available to the -public, and in some countries other activities as well. - - To "convey" a work means any kind of propagation that enables other -parties to make or receive copies. Mere interaction with a user through -a computer network, with no transfer of a copy, is not conveying. - - An interactive user interface displays "Appropriate Legal Notices" -to the extent that it includes a convenient and prominently visible -feature that (1) displays an appropriate copyright notice, and (2) -tells the user that there is no warranty for the work (except to the -extent that warranties are provided), that licensees may convey the -work under this License, and how to view a copy of this License. If -the interface presents a list of user commands or options, such as a -menu, a prominent item in the list meets this criterion. - - 1. Source Code. - - The "source code" for a work means the preferred form of the work -for making modifications to it. "Object code" means any non-source -form of a work. - - A "Standard Interface" means an interface that either is an official -standard defined by a recognized standards body, or, in the case of -interfaces specified for a particular programming language, one that -is widely used among developers working in that language. - - The "System Libraries" of an executable work include anything, other -than the work as a whole, that (a) is included in the normal form of -packaging a Major Component, but which is not part of that Major -Component, and (b) serves only to enable use of the work with that -Major Component, or to implement a Standard Interface for which an -implementation is available to the public in source code form. A -"Major Component", in this context, means a major essential component -(kernel, window system, and so on) of the specific operating system -(if any) on which the executable work runs, or a compiler used to -produce the work, or an object code interpreter used to run it. - - The "Corresponding Source" for a work in object code form means all -the source code needed to generate, install, and (for an executable -work) run the object code and to modify the work, including scripts to -control those activities. However, it does not include the work's -System Libraries, or general-purpose tools or generally available free -programs which are used unmodified in performing those activities but -which are not part of the work. For example, Corresponding Source -includes interface definition files associated with source files for -the work, and the source code for shared libraries and dynamically -linked subprograms that the work is specifically designed to require, -such as by intimate data communication or control flow between those -subprograms and other parts of the work. - - The Corresponding Source need not include anything that users -can regenerate automatically from other parts of the Corresponding -Source. - - The Corresponding Source for a work in source code form is that -same work. - - 2. Basic Permissions. - - All rights granted under this License are granted for the term of -copyright on the Program, and are irrevocable provided the stated -conditions are met. This License explicitly affirms your unlimited -permission to run the unmodified Program. The output from running a -covered work is covered by this License only if the output, given its -content, constitutes a covered work. This License acknowledges your -rights of fair use or other equivalent, as provided by copyright law. - - You may make, run and propagate covered works that you do not -convey, without conditions so long as your license otherwise remains -in force. You may convey covered works to others for the sole purpose -of having them make modifications exclusively for you, or provide you -with facilities for running those works, provided that you comply with -the terms of this License in conveying all material for which you do -not control copyright. Those thus making or running the covered works -for you must do so exclusively on your behalf, under your direction -and control, on terms that prohibit them from making any copies of -your copyrighted material outside their relationship with you. - - Conveying under any other circumstances is permitted solely under -the conditions stated below. Sublicensing is not allowed; section 10 -makes it unnecessary. - - 3. Protecting Users' Legal Rights From Anti-Circumvention Law. - - No covered work shall be deemed part of an effective technological -measure under any applicable law fulfilling obligations under article -11 of the WIPO copyright treaty adopted on 20 December 1996, or -similar laws prohibiting or restricting circumvention of such -measures. - - When you convey a covered work, you waive any legal power to forbid -circumvention of technological measures to the extent such circumvention -is effected by exercising rights under this License with respect to -the covered work, and you disclaim any intention to limit operation or -modification of the work as a means of enforcing, against the work's -users, your or third parties' legal rights to forbid circumvention of -technological measures. - - 4. Conveying Verbatim Copies. - - You may convey verbatim copies of the Program's source code as you -receive it, in any medium, provided that you conspicuously and -appropriately publish on each copy an appropriate copyright notice; -keep intact all notices stating that this License and any -non-permissive terms added in accord with section 7 apply to the code; -keep intact all notices of the absence of any warranty; and give all -recipients a copy of this License along with the Program. - - You may charge any price or no price for each copy that you convey, -and you may offer support or warranty protection for a fee. - - 5. Conveying Modified Source Versions. - - You may convey a work based on the Program, or the modifications to -produce it from the Program, in the form of source code under the -terms of section 4, provided that you also meet all of these conditions: - - a) The work must carry prominent notices stating that you modified - it, and giving a relevant date. - - b) The work must carry prominent notices stating that it is - released under this License and any conditions added under section - 7. This requirement modifies the requirement in section 4 to - "keep intact all notices". - - c) You must license the entire work, as a whole, under this - License to anyone who comes into possession of a copy. This - License will therefore apply, along with any applicable section 7 - additional terms, to the whole of the work, and all its parts, - regardless of how they are packaged. This License gives no - permission to license the work in any other way, but it does not - invalidate such permission if you have separately received it. - - d) If the work has interactive user interfaces, each must display - Appropriate Legal Notices; however, if the Program has interactive - interfaces that do not display Appropriate Legal Notices, your - work need not make them do so. - - A compilation of a covered work with other separate and independent -works, which are not by their nature extensions of the covered work, -and which are not combined with it such as to form a larger program, -in or on a volume of a storage or distribution medium, is called an -"aggregate" if the compilation and its resulting copyright are not -used to limit the access or legal rights of the compilation's users -beyond what the individual works permit. Inclusion of a covered work -in an aggregate does not cause this License to apply to the other -parts of the aggregate. - - 6. Conveying Non-Source Forms. - - You may convey a covered work in object code form under the terms -of sections 4 and 5, provided that you also convey the -machine-readable Corresponding Source under the terms of this License, -in one of these ways: - - a) Convey the object code in, or embodied in, a physical product - (including a physical distribution medium), accompanied by the - Corresponding Source fixed on a durable physical medium - customarily used for software interchange. - - b) Convey the object code in, or embodied in, a physical product - (including a physical distribution medium), accompanied by a - written offer, valid for at least three years and valid for as - long as you offer spare parts or customer support for that product - model, to give anyone who possesses the object code either (1) a - copy of the Corresponding Source for all the software in the - product that is covered by this License, on a durable physical - medium customarily used for software interchange, for a price no - more than your reasonable cost of physically performing this - conveying of source, or (2) access to copy the - Corresponding Source from a network server at no charge. - - c) Convey individual copies of the object code with a copy of the - written offer to provide the Corresponding Source. This - alternative is allowed only occasionally and noncommercially, and - only if you received the object code with such an offer, in accord - with subsection 6b. - - d) Convey the object code by offering access from a designated - place (gratis or for a charge), and offer equivalent access to the - Corresponding Source in the same way through the same place at no - further charge. You need not require recipients to copy the - Corresponding Source along with the object code. If the place to - copy the object code is a network server, the Corresponding Source - may be on a different server (operated by you or a third party) - that supports equivalent copying facilities, provided you maintain - clear directions next to the object code saying where to find the - Corresponding Source. Regardless of what server hosts the - Corresponding Source, you remain obligated to ensure that it is - available for as long as needed to satisfy these requirements. - - e) Convey the object code using peer-to-peer transmission, provided - you inform other peers where the object code and Corresponding - Source of the work are being offered to the general public at no - charge under subsection 6d. - - A separable portion of the object code, whose source code is excluded -from the Corresponding Source as a System Library, need not be -included in conveying the object code work. - - A "User Product" is either (1) a "consumer product", which means any -tangible personal property which is normally used for personal, family, -or household purposes, or (2) anything designed or sold for incorporation -into a dwelling. In determining whether a product is a consumer product, -doubtful cases shall be resolved in favor of coverage. For a particular -product received by a particular user, "normally used" refers to a -typical or common use of that class of product, regardless of the status -of the particular user or of the way in which the particular user -actually uses, or expects or is expected to use, the product. A product -is a consumer product regardless of whether the product has substantial -commercial, industrial or non-consumer uses, unless such uses represent -the only significant mode of use of the product. - - "Installation Information" for a User Product means any methods, -procedures, authorization keys, or other information required to install -and execute modified versions of a covered work in that User Product from -a modified version of its Corresponding Source. The information must -suffice to ensure that the continued functioning of the modified object -code is in no case prevented or interfered with solely because -modification has been made. - - If you convey an object code work under this section in, or with, or -specifically for use in, a User Product, and the conveying occurs as -part of a transaction in which the right of possession and use of the -User Product is transferred to the recipient in perpetuity or for a -fixed term (regardless of how the transaction is characterized), the -Corresponding Source conveyed under this section must be accompanied -by the Installation Information. But this requirement does not apply -if neither you nor any third party retains the ability to install -modified object code on the User Product (for example, the work has -been installed in ROM). - - The requirement to provide Installation Information does not include a -requirement to continue to provide support service, warranty, or updates -for a work that has been modified or installed by the recipient, or for -the User Product in which it has been modified or installed. Access to a -network may be denied when the modification itself materially and -adversely affects the operation of the network or violates the rules and -protocols for communication across the network. - - Corresponding Source conveyed, and Installation Information provided, -in accord with this section must be in a format that is publicly -documented (and with an implementation available to the public in -source code form), and must require no special password or key for -unpacking, reading or copying. - - 7. Additional Terms. - - "Additional permissions" are terms that supplement the terms of this -License by making exceptions from one or more of its conditions. -Additional permissions that are applicable to the entire Program shall -be treated as though they were included in this License, to the extent -that they are valid under applicable law. If additional permissions -apply only to part of the Program, that part may be used separately -under those permissions, but the entire Program remains governed by -this License without regard to the additional permissions. - - When you convey a copy of a covered work, you may at your option -remove any additional permissions from that copy, or from any part of -it. (Additional permissions may be written to require their own -removal in certain cases when you modify the work.) You may place -additional permissions on material, added by you to a covered work, -for which you have or can give appropriate copyright permission. - - Notwithstanding any other provision of this License, for material you -add to a covered work, you may (if authorized by the copyright holders of -that material) supplement the terms of this License with terms: - - a) Disclaiming warranty or limiting liability differently from the - terms of sections 15 and 16 of this License; or - - b) Requiring preservation of specified reasonable legal notices or - author attributions in that material or in the Appropriate Legal - Notices displayed by works containing it; or - - c) Prohibiting misrepresentation of the origin of that material, or - requiring that modified versions of such material be marked in - reasonable ways as different from the original version; or - - d) Limiting the use for publicity purposes of names of licensors or - authors of the material; or - - e) Declining to grant rights under trademark law for use of some - trade names, trademarks, or service marks; or - - f) Requiring indemnification of licensors and authors of that - material by anyone who conveys the material (or modified versions of - it) with contractual assumptions of liability to the recipient, for - any liability that these contractual assumptions directly impose on - those licensors and authors. - - All other non-permissive additional terms are considered "further -restrictions" within the meaning of section 10. If the Program as you -received it, or any part of it, contains a notice stating that it is -governed by this License along with a term that is a further -restriction, you may remove that term. If a license document contains -a further restriction but permits relicensing or conveying under this -License, you may add to a covered work material governed by the terms -of that license document, provided that the further restriction does -not survive such relicensing or conveying. - - If you add terms to a covered work in accord with this section, you -must place, in the relevant source files, a statement of the -additional terms that apply to those files, or a notice indicating -where to find the applicable terms. - - Additional terms, permissive or non-permissive, may be stated in the -form of a separately written license, or stated as exceptions; -the above requirements apply either way. - - 8. Termination. - - You may not propagate or modify a covered work except as expressly -provided under this License. Any attempt otherwise to propagate or -modify it is void, and will automatically terminate your rights under -this License (including any patent licenses granted under the third -paragraph of section 11). - - However, if you cease all violation of this License, then your -license from a particular copyright holder is reinstated (a) -provisionally, unless and until the copyright holder explicitly and -finally terminates your license, and (b) permanently, if the copyright -holder fails to notify you of the violation by some reasonable means -prior to 60 days after the cessation. - - Moreover, your license from a particular copyright holder is -reinstated permanently if the copyright holder notifies you of the -violation by some reasonable means, this is the first time you have -received notice of violation of this License (for any work) from that -copyright holder, and you cure the violation prior to 30 days after -your receipt of the notice. - - Termination of your rights under this section does not terminate the -licenses of parties who have received copies or rights from you under -this License. If your rights have been terminated and not permanently -reinstated, you do not qualify to receive new licenses for the same -material under section 10. - - 9. Acceptance Not Required for Having Copies. - - You are not required to accept this License in order to receive or -run a copy of the Program. Ancillary propagation of a covered work -occurring solely as a consequence of using peer-to-peer transmission -to receive a copy likewise does not require acceptance. However, -nothing other than this License grants you permission to propagate or -modify any covered work. These actions infringe copyright if you do -not accept this License. Therefore, by modifying or propagating a -covered work, you indicate your acceptance of this License to do so. - - 10. Automatic Licensing of Downstream Recipients. - - Each time you convey a covered work, the recipient automatically -receives a license from the original licensors, to run, modify and -propagate that work, subject to this License. You are not responsible -for enforcing compliance by third parties with this License. - - An "entity transaction" is a transaction transferring control of an -organization, or substantially all assets of one, or subdividing an -organization, or merging organizations. If propagation of a covered -work results from an entity transaction, each party to that -transaction who receives a copy of the work also receives whatever -licenses to the work the party's predecessor in interest had or could -give under the previous paragraph, plus a right to possession of the -Corresponding Source of the work from the predecessor in interest, if -the predecessor has it or can get it with reasonable efforts. - - You may not impose any further restrictions on the exercise of the -rights granted or affirmed under this License. For example, you may -not impose a license fee, royalty, or other charge for exercise of -rights granted under this License, and you may not initiate litigation -(including a cross-claim or counterclaim in a lawsuit) alleging that -any patent claim is infringed by making, using, selling, offering for -sale, or importing the Program or any portion of it. - - 11. Patents. - - A "contributor" is a copyright holder who authorizes use under this -License of the Program or a work on which the Program is based. The -work thus licensed is called the contributor's "contributor version". - - A contributor's "essential patent claims" are all patent claims -owned or controlled by the contributor, whether already acquired or -hereafter acquired, that would be infringed by some manner, permitted -by this License, of making, using, or selling its contributor version, -but do not include claims that would be infringed only as a -consequence of further modification of the contributor version. For -purposes of this definition, "control" includes the right to grant -patent sublicenses in a manner consistent with the requirements of -this License. - - Each contributor grants you a non-exclusive, worldwide, royalty-free -patent license under the contributor's essential patent claims, to -make, use, sell, offer for sale, import and otherwise run, modify and -propagate the contents of its contributor version. - - In the following three paragraphs, a "patent license" is any express -agreement or commitment, however denominated, not to enforce a patent -(such as an express permission to practice a patent or covenant not to -sue for patent infringement). To "grant" such a patent license to a -party means to make such an agreement or commitment not to enforce a -patent against the party. - - If you convey a covered work, knowingly relying on a patent license, -and the Corresponding Source of the work is not available for anyone -to copy, free of charge and under the terms of this License, through a -publicly available network server or other readily accessible means, -then you must either (1) cause the Corresponding Source to be so -available, or (2) arrange to deprive yourself of the benefit of the -patent license for this particular work, or (3) arrange, in a manner -consistent with the requirements of this License, to extend the patent -license to downstream recipients. "Knowingly relying" means you have -actual knowledge that, but for the patent license, your conveying the -covered work in a country, or your recipient's use of the covered work -in a country, would infringe one or more identifiable patents in that -country that you have reason to believe are valid. - - If, pursuant to or in connection with a single transaction or -arrangement, you convey, or propagate by procuring conveyance of, a -covered work, and grant a patent license to some of the parties -receiving the covered work authorizing them to use, propagate, modify -or convey a specific copy of the covered work, then the patent license -you grant is automatically extended to all recipients of the covered -work and works based on it. - - A patent license is "discriminatory" if it does not include within -the scope of its coverage, prohibits the exercise of, or is -conditioned on the non-exercise of one or more of the rights that are -specifically granted under this License. You may not convey a covered -work if you are a party to an arrangement with a third party that is -in the business of distributing software, under which you make payment -to the third party based on the extent of your activity of conveying -the work, and under which the third party grants, to any of the -parties who would receive the covered work from you, a discriminatory -patent license (a) in connection with copies of the covered work -conveyed by you (or copies made from those copies), or (b) primarily -for and in connection with specific products or compilations that -contain the covered work, unless you entered into that arrangement, -or that patent license was granted, prior to 28 March 2007. - - Nothing in this License shall be construed as excluding or limiting -any implied license or other defenses to infringement that may -otherwise be available to you under applicable patent law. - - 12. No Surrender of Others' Freedom. - - If conditions are imposed on you (whether by court order, agreement or -otherwise) that contradict the conditions of this License, they do not -excuse you from the conditions of this License. If you cannot convey a -covered work so as to satisfy simultaneously your obligations under this -License and any other pertinent obligations, then as a consequence you may -not convey it at all. For example, if you agree to terms that obligate you -to collect a royalty for further conveying from those to whom you convey -the Program, the only way you could satisfy both those terms and this -License would be to refrain entirely from conveying the Program. - - 13. Remote Network Interaction; Use with the GNU General Public License. - - Notwithstanding any other provision of this License, if you modify the -Program, your modified version must prominently offer all users -interacting with it remotely through a computer network (if your version -supports such interaction) an opportunity to receive the Corresponding -Source of your version by providing access to the Corresponding Source -from a network server at no charge, through some standard or customary -means of facilitating copying of software. This Corresponding Source -shall include the Corresponding Source for any work covered by version 3 -of the GNU General Public License that is incorporated pursuant to the -following paragraph. - - Notwithstanding any other provision of this License, you have -permission to link or combine any covered work with a work licensed -under version 3 of the GNU General Public License into a single -combined work, and to convey the resulting work. The terms of this -License will continue to apply to the part which is the covered work, -but the work with which it is combined will remain governed by version -3 of the GNU General Public License. - - 14. Revised Versions of this License. - - The Free Software Foundation may publish revised and/or new versions of -the GNU Affero General Public License from time to time. Such new versions -will be similar in spirit to the present version, but may differ in detail to -address new problems or concerns. - - Each version is given a distinguishing version number. If the -Program specifies that a certain numbered version of the GNU Affero General -Public License "or any later version" applies to it, you have the -option of following the terms and conditions either of that numbered -version or of any later version published by the Free Software -Foundation. If the Program does not specify a version number of the -GNU Affero General Public License, you may choose any version ever published -by the Free Software Foundation. - - If the Program specifies that a proxy can decide which future -versions of the GNU Affero General Public License can be used, that proxy's -public statement of acceptance of a version permanently authorizes you -to choose that version for the Program. - - Later license versions may give you additional or different -permissions. However, no additional obligations are imposed on any -author or copyright holder as a result of your choosing to follow a -later version. - - 15. Disclaimer of Warranty. - - THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY -APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT -HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY -OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, -THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR -PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM -IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF -ALL NECESSARY SERVICING, REPAIR OR CORRECTION. - - 16. Limitation of Liability. - - IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING -WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS -THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY -GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE -USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF -DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD -PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), -EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF -SUCH DAMAGES. - - 17. Interpretation of Sections 15 and 16. - - If the disclaimer of warranty and limitation of liability provided -above cannot be given local legal effect according to their terms, -reviewing courts shall apply local law that most closely approximates -an absolute waiver of all civil liability in connection with the -Program, unless a warranty or assumption of liability accompanies a -copy of the Program in return for a fee. - - END OF TERMS AND CONDITIONS - - How to Apply These Terms to Your New Programs - - If you develop a new program, and you want it to be of the greatest -possible use to the public, the best way to achieve this is to make it -free software which everyone can redistribute and change under these terms. - - To do so, attach the following notices to the program. It is safest -to attach them to the start of each source file to most effectively -state the exclusion of warranty; and each file should have at least -the "copyright" line and a pointer to where the full notice is found. - - - Copyright (C) - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU Affero General Public License as published - by the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU Affero General Public License for more details. - - You should have received a copy of the GNU Affero General Public License - along with this program. If not, see . - -Also add information on how to contact you by electronic and paper mail. - - If your software can interact with users remotely through a computer -network, you should also make sure that it provides a way for users to -get its source. For example, if your program is a web application, its -interface could display a "Source" link that leads users to an archive -of the code. There are many ways you could offer source, and different -solutions will be better for different programs; see section 13 for the -specific requirements. - - You should also get your employer (if you work as a programmer) or school, -if any, to sign a "copyright disclaimer" for the program, if necessary. -For more information on this, and how to apply and follow the GNU AGPL, see -. - diff --git a/web/README.md b/web/README.md deleted file mode 100644 index c284174..0000000 --- a/web/README.md +++ /dev/null @@ -1,75 +0,0 @@ -# Senbei web - -Senbei running in the browser: the unpacker core compiled to WebAssembly, -wrapped in a small static page. Everything is client-side — files are read -into the page, unpacked locally, and offered back as downloads. Nothing is -uploaded; there is no server component. - -## Features - -- A legal notice is shown as a blocking dialog on page open; the tool is - unusable until it is acknowledged. -- Dropped files land in a file list, not unpacked immediately: review the - batch, remove mistakes, then press **Unpack**. A module and its `._` - companion can be dropped in any order (or in separate drops) — companions - auto-pair by name (`Foo.dll._` → `Foo.dll`) and show as a badge on the - module's row; removing a module removes its companion too. -- Rows show state at a glance: black while staged, an animated blue bar - while unpacking, green on success (with a download button) and red on - failure. -- Drop one or more protected `.exe` / `.dll` modules → get `.unpack.*` - downloads. -- Drop an il2cpp `global-metadata.dat` → de-obfuscated - `global-metadata.unpack.dat` (only when tokens actually change). -- Each output passes the same static integrity check as the CLI; suspect - outputs are flagged with the specific defects found. - -## Architecture notes - -- Every unpack runs in a **disposable Web Worker** (fresh wasm instance per - file): the UI stays responsive on 100 MB+ modules, and a wasm trap is - isolated to that worker. -- **Why workers matter for correctness:** the DLL-first routing probe relies - on `catch_unwind` to reject EXE-shell-layout DLLs, and panics cannot be - caught in WebAssembly — the probe traps the whole call. When a DLL unpack - traps, the app retries once in a new worker with the forced-EXE pipeline - (`unpack_file_force_exe`), reproducing the CLI's dll-first/exe-fallback - outcome. Spliced companion inputs skip the probe entirely (they are always - EXE-shell layout), exactly like the CLI. -- Rust panic messages are forwarded to the browser console - (`console_error_panic_hook`) — check devtools when reporting an issue. - -## Building - -Requires a Rust toolchain (`rust-toolchain.toml` in the repo root pins one, -including the `wasm32-unknown-unknown` target) and -[wasm-pack](https://rustwasm.github.io/wasm-pack/installer/). - -```cmd -cd web -wasm-pack build --target web --release -``` - -This produces `web/pkg/` (git-ignored). Then serve the `web/` directory with -any static file server and open `index.html`: - -```cmd -python -m http.server -d web 8000 -:: -> http://localhost:8000 -``` - -(Opening `index.html` via `file://` won't work — ES modules require HTTP.) - -## Layout - -``` -web/ -├── Cargo.toml senbei-web cdylib crate (depends on the senbei lib) -├── src/lib.rs #[wasm_bindgen] bindings: detect / unpack_file / -│ unpack_file_force_exe / deobfuscate_metadata -├── index.html the page -├── app.js dropzone, file list, worker orchestration, downloads -├── worker.js one-shot unpack worker (fresh wasm instance per file) -├── style.css -└── pkg/ wasm-pack output (git-ignored) -``` diff --git a/web/app.js b/web/app.js deleted file mode 100644 index a5c9df6..0000000 --- a/web/app.js +++ /dev/null @@ -1,392 +0,0 @@ -import init, { detect, deobfuscate_metadata } from './pkg/senbei_web.js'; - -const dropzone = document.getElementById('dropzone'); -const picker = document.getElementById('picker'); -const fileList = document.getElementById('files'); -const actions = document.getElementById('actions'); -const unpackBtn = document.getElementById('unpack-btn'); -const clearBtn = document.getElementById('clear-btn'); -const legalOverlay = document.getElementById('legal-overlay'); -const legalAccept = document.getElementById('legal-accept'); -const legalLink = document.getElementById('legal-link'); - -await init(); - -// --- Legal gate: the page is unusable until the notice is acknowledged. --- -legalAccept.addEventListener('click', () => legalOverlay.remove()); -legalLink.addEventListener('click', (e) => { - e.preventDefault(); - if (!document.getElementById('legal-overlay')) { - document.body.appendChild(legalOverlay); - } -}); - -// --- File list: one row per module. Companions (`X._`) never get their own --- -// --- row once their base module `X` is present — they show as a badge on --- -// --- the base row. Rows are black while staged, show an animated blue --- -// --- progress bar while unpacking, and turn green (success) or red --- -// --- (failure) at the end; success rows gain a download button. --- - -// --- Row DOM is updated INCREMENTALLY: rows are created once and patched --- -// --- in place. Rebuilding the list on every change would restart the --- -// --- entrance animation of every row and reset the unpack shimmer. --- - -/** name -> { - * file: File, - * kind: string|undefined, // detect() result; undefined for `._` files - * state: 'staged'|'working'|'ok'|'err', - * bytes: Uint8Array|null, // unpacked output (state 'ok') - * note: string, // status line (kind, suspect issues, error) - * suspect: boolean, - * } */ -const files = new Map(); - -/** name -> row
  • element (companions merged into their base have none) */ -const rowEls = new Map(); - -const KIND_LABEL = { - exe: 'protected EXE', - 'native-dll': 'protected native DLL', - 'managed-dll': 'protected managed DLL', - metadata: 'il2cpp metadata', -}; - -const COMPANION_SVG = - ''; - -const DOWNLOAD_SVG = - ''; - -dropzone.addEventListener('click', () => picker.click()); -dropzone.addEventListener('keydown', (e) => { - if (e.key === 'Enter' || e.key === ' ') picker.click(); -}); -picker.addEventListener('change', () => { - stageFiles(picker.files); - picker.value = ''; -}); -dropzone.addEventListener('dragover', (e) => { - e.preventDefault(); - dropzone.classList.add('over'); -}); -dropzone.addEventListener('dragleave', () => dropzone.classList.remove('over')); -dropzone.addEventListener('drop', (e) => { - e.preventDefault(); - dropzone.classList.remove('over'); - stageFiles(e.dataTransfer.files); -}); - -async function stageFiles(list) { - // Snapshot synchronously: `picker.files` and `dataTransfer.files` are LIVE - // lists — clearing the picker or returning from the drop event empties - // them, so an await before this point silently drops every file after the - // first. - const snapshot = [...list]; - for (const file of snapshot) { - // Detection only needs the file header (key table at offset 4096 plus - // the PE header fields); read a small slice, not the whole file. - const head = new Uint8Array(await file.slice(0, 65536).arrayBuffer()); - // Companions are ciphertext fragments; detect() only makes sense on the - // base module, so skip it for `._` files. - const kind = file.name.endsWith('._') ? undefined : detect(head); - const old = files.get(file.name); - files.set(file.name, { - file, - kind, - state: 'staged', - bytes: null, - note: '', - suspect: false, - }); // same name re-dropped: replace - // A re-dropped file restarts as staged; drop any stale row/output. - if (old) removeRow(file.name, true); - } - render(); -} - -/** Insert `.unpack` before the final extension: `app.exe` -> `app.unpack.exe`. */ -function outName(name) { - const dot = name.lastIndexOf('.'); - return dot > 0 ? `${name.slice(0, dot)}.unpack${name.slice(dot)}` : `${name}.unpack`; -} - -function statusText(name, entry) { - if (name.endsWith('._')) { - return `companion — needs ${name.slice(0, -2)}`; - } - switch (entry.state) { - case 'staged': - return entry.kind === undefined - ? 'not recognized — will be skipped' - : KIND_LABEL[entry.kind] ?? entry.kind; - case 'working': - return 'unpacking…'; - case 'ok': - case 'err': - return entry.note; - } -} - -function buildRow(name) { - const li = document.createElement('li'); - li.className = 'file staged'; - li.dataset.name = name; - - const bar = document.createElement('div'); - bar.className = 'bar'; - li.appendChild(bar); - - const row = document.createElement('div'); - row.className = 'row'; - - const label = document.createElement('span'); - label.className = 'name'; - label.textContent = name; - row.appendChild(label); - - const badge = document.createElement('span'); - badge.className = 'badge companion'; - badge.innerHTML = COMPANION_SVG; - badge.hidden = true; - row.appendChild(badge); - - const status = document.createElement('span'); - status.className = 'status'; - row.appendChild(status); - - const dl = document.createElement('a'); - dl.className = 'dl'; - dl.innerHTML = DOWNLOAD_SVG; - dl.hidden = true; - row.appendChild(dl); - - const rm = document.createElement('button'); - rm.type = 'button'; - rm.className = 'remove'; - rm.textContent = '×'; - rm.title = `Remove ${name}`; - rm.addEventListener('click', () => { - // A companion belongs to its base module: removing the base removes the - // companion too. - files.delete(name); - if (!name.endsWith('._')) files.delete(`${name}._`); - render(); - }); - row.appendChild(rm); - - li.appendChild(row); - return li; -} - -function updateRow(li, name, entry) { - const isCompanion = name.endsWith('._'); - li.className = - `file ${entry.state}` + (entry.suspect && entry.state === 'ok' ? ' suspect' : ''); - - const badge = li.querySelector('.badge'); - const hasCompanion = isCompanion || files.has(`${name}._`); - badge.hidden = !hasCompanion; - if (hasCompanion) { - badge.title = isCompanion - ? 'external companion (._)' - : `companion loaded: ${name}._`; - } - - li.querySelector('.status').textContent = statusText(name, entry); - - const dl = li.querySelector('.dl'); - const downloadable = entry.state === 'ok' && entry.bytes; - dl.hidden = !downloadable; - if (downloadable) { - if (dl._bytesFor !== entry.bytes) { - if (dl.href) URL.revokeObjectURL(dl.href); - dl.href = URL.createObjectURL( - new Blob([entry.bytes], { type: 'application/octet-stream' }), - ); - dl._bytesFor = entry.bytes; - } - dl.download = outName(name); - dl.title = `Download ${outName(name)}`; - } -} - -function removeRow(name, instant) { - const li = rowEls.get(name); - if (!li) return; - rowEls.delete(name); - // Release the download blob. Object URLs are roots: without this an unpacked - // 100 MB image stays resident for the life of the page every time a row is - // removed or the list is cleared. - const dl = li.querySelector('.dl'); - if (dl?.href) { - URL.revokeObjectURL(dl.href); - dl.removeAttribute('href'); - dl._bytesFor = null; - } - if (instant) { - li.remove(); - return; - } - // Fade AND collapse: without the height/margin transition the rows below - // would hold position during the fade and then snap up on removal. The - // end state must be inline too — an inline start value would otherwise - // beat the stylesheet's `.leaving { max-height: 0 }`. - li.style.maxHeight = `${li.offsetHeight}px`; - void li.offsetHeight; // reflow: give the transition a concrete start value - li.classList.add('leaving'); - li.style.maxHeight = '0px'; - setTimeout(() => li.remove(), 230); -} - -function render() { - // Create/update rows in Map order; companions whose base is staged merge - // into the base row (no row of their own). - const wanted = []; - for (const [name] of files) { - if (name.endsWith('._') && files.has(name.slice(0, -2))) continue; - wanted.push(name); - } - const wantedSet = new Set(wanted); - - // Removals first: departed rows are marked leaving (and dropped from - // rowEls) BEFORE the ordering loop, so the loop treats them as transparent - // and never reorders siblings around them (that would snap, not slide). - for (const name of [...rowEls.keys()]) { - if (!wantedSet.has(name)) removeRow(name, false); - } - - // In-place ordering: only rows that are out of position are moved, so - // running animations (entrance, shimmer) are never restarted by a render. - // Rows mid-leave-animation (no longer in rowEls) are skipped and keep - // their spot — reordering siblings around them would make them snap - // instead of sliding with the collapse. - let cursor = fileList.firstChild; - for (const name of wanted) { - let li = rowEls.get(name); - if (!li) { - li = buildRow(name); - rowEls.set(name, li); - } - updateRow(li, name, files.get(name)); - while (cursor && !rowEls.has(cursor.dataset.name)) { - cursor = cursor.nextSibling; - } - if (li === cursor) { - cursor = cursor.nextSibling; - } else { - fileList.insertBefore(li, cursor); - } - } - - const unpackable = [...files].some( - ([name, e]) => - !name.endsWith('._') && e.kind !== undefined && e.state === 'staged', - ); - unpackBtn.disabled = !unpackable; - actions.hidden = files.size === 0; -} - -clearBtn.addEventListener('click', () => { - files.clear(); - render(); -}); - -/** - * Run one unpack in a disposable Web Worker (fresh wasm instance per call — - * see worker.js). Buffers are transferred, so the inputs are neutered on the - * main thread afterwards; callers re-read from the File for a retry. - */ -function runUnpack(inputBytes, compBytes, forceExe) { - return new Promise((resolve) => { - const w = new Worker('worker.js', { type: 'module' }); - w.onmessage = (e) => { - w.terminate(); - resolve(e.data); - }; - w.onerror = (e) => { - w.terminate(); - resolve({ ok: false, trap: true, message: e.message || 'worker error' }); - }; - const transfer = [inputBytes.buffer, ...(compBytes ? [compBytes.buffer] : [])]; - w.postMessage({ input: inputBytes, companion: compBytes ?? null, forceExe }, transfer); - }); -} - -async function unpackModule(name, entry) { - const compEntry = files.get(`${name}._`); - const read = (f) => f.arrayBuffer().then((b) => new Uint8Array(b)); - - let input = await read(entry.file); - let comp = compEntry ? await read(compEntry.file) : undefined; - let r = await runUnpack(input, comp, false); - - if (!r.ok && r.trap && entry.kind !== 'exe') { - // The DLL-routing probe trapped (panics can't be caught in wasm). Retry - // once with the forced-EXE pipeline in a fresh worker — this mirrors the - // CLI's dll-first/exe-fallback outcome for EXE-shell-layout DLLs. - input = await read(entry.file); - comp = compEntry ? await read(compEntry.file) : undefined; - r = await runUnpack(input, comp, true); - } - - if (!r.ok) { - entry.state = 'err'; - entry.note = r.trap - ? 'unpack failed (internal trap) — this Crackproof layout may be unsupported' - : r.message; - return; - } - entry.state = 'ok'; - entry.bytes = r.bytes; - entry.suspect = r.suspect; - entry.note = - (r.companion ? 'spliced from ._ companion; ' : '') + - `kind: ${r.kind}` + - (r.suspect ? ` — SUSPECT: ${r.issues.join('; ')}` : ''); -} - -unpackBtn.addEventListener('click', async () => { - unpackBtn.disabled = true; - clearBtn.disabled = true; - try { - for (const [name, entry] of files) { - if (name.endsWith('._') || entry.state !== 'staged') continue; - - if (entry.kind === undefined) { - entry.state = 'err'; - entry.note = 'not recognized as Crackproof-protected — skipped'; - render(); - continue; - } - - entry.state = 'working'; - render(); - try { - if (entry.kind === 'metadata') { - const bytes = new Uint8Array(await entry.file.arrayBuffer()); - const r = deobfuscate_metadata(bytes); - if (r.remapped === 0) { - entry.state = 'err'; - entry.note = `metadata already clean (v${r.version}, ${r.methods} methods) — nothing to do`; - } else { - entry.state = 'ok'; - entry.bytes = r.bytes; - entry.note = `${r.remapped}/${r.methods} method tokens remapped across ${r.modules} modules`; - } - } else { - await unpackModule(name, entry); - } - } catch (e) { - entry.state = 'err'; - entry.note = e instanceof Error ? e.message : String(e); - } - render(); - } - } finally { - clearBtn.disabled = false; - render(); - } -}); diff --git a/web/index.html b/web/index.html deleted file mode 100644 index 59e9c2c..0000000 --- a/web/index.html +++ /dev/null @@ -1,78 +0,0 @@ - - - - - -Senbei — static Crackproof unpacker - - - - - -
    -
    -

    Senbei web

    - - - -
    -

    - Static unpacker for Crackproof-protected PE files, running entirely in - your browser. No file ever leaves your device. -

    - -
    -

    Drop files here or click to browse

    -

    - Protected .exe / .dll modules, optional - ._ companions, or an il2cpp - global-metadata.dat. -

    - -
    - -
      - - - -
      -

      Legal notice · - Senbei is free software under the AGPL-3.0 license.

      -
      -
      - - - diff --git a/web/src/lib.rs b/web/src/lib.rs deleted file mode 100644 index 4b98583..0000000 --- a/web/src/lib.rs +++ /dev/null @@ -1,180 +0,0 @@ -//! WebAssembly bindings for the senbei unpacker core. -//! -//! Everything here is I/O-free: the browser hands in file bytes and gets -//! unpacked file bytes back. No network, no filesystem, no uploads. - -use wasm_bindgen::prelude::*; - -/// Install a panic hook that forwards Rust panic messages to the browser -/// console (and to the JS error), instead of a bare `unreachable` trap. -#[wasm_bindgen(start)] -pub fn init_panic_hook() { - console_error_panic_hook::set_once(); -} - -/// Result of unpacking one protected module. -#[wasm_bindgen] -pub struct UnpackResult { - kind: String, - bytes: Vec, - suspect: bool, - issues: Vec, - companion: bool, -} - -#[wasm_bindgen] -impl UnpackResult { - /// Detected module kind: `"exe"`, `"native-dll"`, or `"managed-dll"`. - #[wasm_bindgen(getter)] - pub fn kind(&self) -> String { - self.kind.clone() - } - - /// The unpacked image bytes. - #[wasm_bindgen(getter)] - pub fn bytes(&self) -> Vec { - self.bytes.clone() - } - - /// True when the static integrity check flagged the output as likely - /// broken at runtime. The bytes are still the best available. - #[wasm_bindgen(getter)] - pub fn suspect(&self) -> bool { - self.suspect - } - - /// Human-readable integrity defects (empty when the check is clean). - #[wasm_bindgen(getter)] - pub fn issues(&self) -> Vec { - self.issues.clone() - } - - /// True when the input was spliced from an external-companion (`._`) - /// payload. - #[wasm_bindgen(getter)] - pub fn companion(&self) -> bool { - self.companion - } -} - -/// Result of de-obfuscating an il2cpp `global-metadata.dat`. -#[wasm_bindgen] -pub struct MetadataResult { - bytes: Vec, - version: u32, - methods: usize, - remapped: usize, - modules: usize, -} - -#[wasm_bindgen] -impl MetadataResult { - /// The (possibly rewritten) metadata bytes. - #[wasm_bindgen(getter)] - pub fn bytes(&self) -> Vec { - self.bytes.clone() - } - - #[wasm_bindgen(getter)] - pub fn version(&self) -> u32 { - self.version - } - - /// Total method-definition entries in the metadata. - #[wasm_bindgen(getter)] - pub fn methods(&self) -> usize { - self.methods - } - - /// Method tokens actually rewritten (0 means the input was already - /// de-obfuscated and the bytes are unchanged). - #[wasm_bindgen(getter)] - pub fn remapped(&self) -> usize { - self.remapped - } - - /// Modules (images) owning at least one method. - #[wasm_bindgen(getter)] - pub fn modules(&self) -> usize { - self.modules - } -} - -fn kind_str(kind: senbei::unpacker::Kind) -> &'static str { - match kind { - senbei::unpacker::Kind::Exe => "exe", - senbei::unpacker::Kind::NativeDll => "native-dll", - senbei::unpacker::Kind::ManagedDll => "managed-dll", - } -} - -/// Classify a file's bytes without unpacking. -/// -/// Returns `"exe"`, `"native-dll"`, `"managed-dll"`, `"metadata"` (an il2cpp -/// `global-metadata.dat`), or `undefined` for anything unrecognized. -#[wasm_bindgen] -pub fn detect(input: &[u8]) -> Option { - if senbei::metadata::is_metadata(input) { - return Some("metadata".to_string()); - } - senbei::unpacker::detect(input).map(|d| kind_str(d.kind).to_string()) -} - -/// Unpack a protected module. -/// -/// `input` is the protected `.exe`/`.dll`; `companion` is the optional -/// `._` external-companion payload (pass `null`/`undefined` when there -/// is none). Throws a string error when the input is not a supported -/// Crackproof file or is corrupt. -#[wasm_bindgen] -pub fn unpack_file( - input: &[u8], - companion: Option>, -) -> Result { - let r = senbei::job::unpack_bytes(input, companion.as_deref()) - .map_err(|e| JsError::new(&e.to_string()))?; - Ok(UnpackResult { - kind: kind_str(r.kind).to_string(), - bytes: r.bytes, - suspect: !r.integrity.ok(), - issues: r.integrity.issues, - companion: r.companion, - }) -} - -/// De-obfuscate the method tokens of an il2cpp `global-metadata.dat`. -/// -/// The transform is idempotent: an already-clean metadata comes back -/// byte-identical with `remapped == 0`. Throws a string error for non-metadata -/// input, an unsupported format version, or a malformed layout. -#[wasm_bindgen] -pub fn deobfuscate_metadata(data: &[u8]) -> Result { - let (bytes, report) = - senbei::metadata::deobfuscate(data).map_err(|e| JsError::new(&e.to_string()))?; - Ok(MetadataResult { - bytes, - version: report.version, - methods: report.methods, - remapped: report.remapped, - modules: report.modules, - }) -} - -/// Unpack a protected module, forcing the EXE pipeline (no DLL-pipeline -/// probe). See [`senbei::job::unpack_bytes_force_exe`] for why the web app -/// needs this recovery path. -#[wasm_bindgen] -pub fn unpack_file_force_exe( - input: &[u8], - companion: Option>, -) -> Result { - let r = senbei::job::unpack_bytes_force_exe(input, companion.as_deref()) - .map_err(|e| JsError::new(&e.to_string()))?; - Ok(UnpackResult { - kind: kind_str(r.kind).to_string(), - bytes: r.bytes, - suspect: !r.integrity.ok(), - issues: r.integrity.issues, - companion: r.companion, - }) -} diff --git a/web/style.css b/web/style.css deleted file mode 100644 index 8a2da81..0000000 --- a/web/style.css +++ /dev/null @@ -1,369 +0,0 @@ -:root { - color-scheme: dark; - --bg: #14161a; - --panel: #1d2026; - --black: #0c0e11; - --border: #2e323b; - --text: #e4e7ec; - --dim: #9aa3b0; - --accent: #e8b64c; - --blue: #3b82f6; - --blue-deep: #1d4ed8; - --ok: #1e6b34; - --ok-bright: #6fcf7c; - --err: #7a2828; - --err-bright: #e06c6c; - --warn: #e8b64c; -} - -* { box-sizing: border-box; } - -/* display rules below (inline-flex etc.) would otherwise beat the hidden - attribute's UA display:none — the badge/download icons must stay hidden. */ -[hidden] { display: none !important; } - -body { - margin: 0; - background: var(--bg); - color: var(--text); - font: 16px/1.55 system-ui, "Segoe UI", sans-serif; -} - -main { - max-width: 720px; - margin: 0 auto; - padding: 2.5rem 1.25rem 3rem; -} - -.topbar { - display: flex; - align-items: center; - justify-content: space-between; -} - -h1 { margin-bottom: 0.25rem; } - -.tag { - font-size: 0.45em; - vertical-align: super; - color: var(--accent); - letter-spacing: 0.08em; -} - -.github-link { - color: var(--dim); - transition: color 0.2s, transform 0.2s; -} - -.github-link:hover { - color: var(--text); - transform: scale(1.12); -} - -.lede { color: var(--dim); } -.lede strong { color: var(--text); } - -/* --- legal modal --- */ - -#legal-overlay { - position: fixed; - inset: 0; - z-index: 10; - display: flex; - align-items: center; - justify-content: center; - padding: 1.25rem; - background: rgba(10, 11, 13, 0.82); - backdrop-filter: blur(3px); - animation: fadeIn 0.25s ease-out; -} - -.legal-box { - max-width: 560px; - max-height: 85vh; - overflow-y: auto; - padding: 1.75rem 2rem; - background: var(--panel); - border: 1px solid var(--border); - border-radius: 12px; - animation: popIn 0.3s cubic-bezier(0.2, 1.4, 0.4, 1); -} - -.legal-box h2 { margin-top: 0; } -.legal-box p { color: var(--dim); font-size: 0.95rem; } -.legal-box strong { color: var(--text); } - -button { - font: inherit; - padding: 0.55rem 1.2rem; - border: 1px solid var(--accent); - border-radius: 8px; - background: var(--accent); - color: #14161a; - font-weight: 600; - cursor: pointer; - transition: transform 0.15s, box-shadow 0.15s, opacity 0.15s; -} - -button:not(:disabled):hover { - transform: translateY(-1px); - box-shadow: 0 3px 12px rgba(232, 182, 76, 0.25); -} - -button:not(:disabled):active { transform: translateY(0); } - -button:disabled { - opacity: 0.45; - cursor: default; -} - -button.secondary { - background: transparent; - color: var(--dim); - border-color: var(--border); -} - -button.secondary:not(:disabled):hover { - box-shadow: none; - color: var(--text); -} - -#legal-accept { width: 100%; margin-top: 0.5rem; } - -/* --- dropzone --- */ - -#dropzone { - margin: 1.5rem 0; - padding: 2.25rem 1.5rem; - text-align: center; - background: var(--panel); - border: 2px dashed var(--border); - border-radius: 12px; - cursor: pointer; - transition: border-color 0.2s, background 0.2s, transform 0.2s; -} - -#dropzone:hover, #dropzone:focus-visible { - border-color: var(--accent); - outline: none; -} - -#dropzone.over { - border-color: var(--accent); - background: #232730; - transform: scale(1.01); - animation: pulse 1s ease-in-out infinite; -} - -#dropzone p { margin: 0.25rem 0; } -.hint { color: var(--dim); font-size: 0.9rem; } -code { - background: var(--bg); - padding: 0.1em 0.35em; - border-radius: 4px; - font-size: 0.9em; -} - -/* --- file list --- */ - -#files { - list-style: none; - margin: 0 0 0.75rem; - padding: 0; -} - -.file { - position: relative; - margin-bottom: 0.45rem; - border: 1px solid var(--border); - border-radius: 8px; - overflow: hidden; - background: var(--black); /* staged */ - transition: background-color 0.5s ease; - animation: slideIn 0.25s ease-out; -} - -.file.leaving { - opacity: 0; - transform: translateX(12px); - margin-bottom: 0; - border-width: 0; - transition: - opacity 0.13s ease-in, - transform 0.13s ease-in, - max-height 0.17s ease-in 0.03s, - margin-bottom 0.17s ease-in 0.03s, - border-width 0.17s ease-in 0.03s; -} - -.file .bar { - position: absolute; - inset: 0; - opacity: 0; - transition: opacity 0.3s; -} - -.file.working { background: var(--black); } - -/* The gradient's left and right edge colors match, so the 200%-sized image - tiles seamlessly and the position loop has no visible restart. */ -.file.working .bar { - opacity: 1; - background: linear-gradient( - 100deg, - var(--blue-deep) 0%, - var(--blue) 25%, - #7fb3ff 50%, - var(--blue) 75%, - var(--blue-deep) 100% - ); - background-size: 200% 100%; - animation: shimmer 1.6s linear infinite; -} - -.file.ok { background: var(--ok); } -.file.ok.suspect { background: #6b5a1e; } -.file.err { background: var(--err); } - -.file .row { - position: relative; - display: flex; - align-items: center; - gap: 0.6rem; - padding: 0.55rem 0.85rem; -} - -.file .name { - overflow-wrap: anywhere; - font-weight: 600; -} - -.file.working .name { color: #fff; } -.file.ok .name, .file.err .name { color: #fff; } - -.badge.companion { - flex: none; - display: inline-flex; - align-items: center; - padding: 0.15rem 0.35rem; - border-radius: 5px; - background: rgba(59, 130, 246, 0.18); - color: #7fb3ff; -} - -.file.ok .badge.companion, -.file.err .badge.companion, -.file.working .badge.companion { - background: rgba(255, 255, 255, 0.15); - color: #fff; -} - -.file .status { - flex: 1; - text-align: right; - font-size: 0.85rem; - color: var(--dim); - overflow-wrap: anywhere; -} - -.file.ok .status { color: #cfe9d5; } -.file.ok.suspect .status { color: #f0e3b2; } -.file.err .status { color: #f0c8c8; } -.file.working .status { color: #dbe7ff; } - -.file .dl { - flex: none; - display: inline-flex; - align-items: center; - justify-content: center; - width: 1.9rem; - height: 1.9rem; - border-radius: 6px; - background: rgba(255, 255, 255, 0.16); - color: #fff; - transition: background 0.15s, transform 0.15s; - animation: popIn 0.3s cubic-bezier(0.2, 1.4, 0.4, 1); -} - -.file .dl:hover { - background: rgba(255, 255, 255, 0.32); - transform: scale(1.1); -} - -.file .remove { - flex: none; - padding: 0.1rem 0.55rem; - background: transparent; - border: 1px solid var(--border); - border-radius: 6px; - color: var(--dim); - font-weight: 400; -} - -.file .remove:hover { - color: var(--err-bright); - border-color: var(--err-bright); - box-shadow: none; - transform: none; -} - -.file.ok .remove, -.file.err .remove, -.file.working .remove { - border-color: rgba(255, 255, 255, 0.3); - color: rgba(255, 255, 255, 0.75); -} - -.file.ok .remove:hover, -.file.err .remove:hover { - color: #fff; - border-color: #fff; -} - -.actions { - display: flex; - gap: 0.6rem; - animation: fadeIn 0.25s ease-out; -} - -footer { - margin-top: 2rem; - color: var(--dim); - font-size: 0.85rem; -} -footer a { color: var(--dim); } - -/* --- animations --- */ - -@keyframes fadeIn { - from { opacity: 0; } - to { opacity: 1; } -} - -@keyframes popIn { - from { opacity: 0; transform: scale(0.85); } - to { opacity: 1; transform: scale(1); } -} - -@keyframes slideIn { - from { opacity: 0; transform: translateY(-6px); } - to { opacity: 1; transform: translateY(0); } -} - -@keyframes shimmer { - from { background-position: 0 0; } - to { background-position: -200% 0; } -} - -@keyframes pulse { - 0%, 100% { box-shadow: 0 0 0 0 rgba(232, 182, 76, 0.25); } - 50% { box-shadow: 0 0 0 6px rgba(232, 182, 76, 0); } -} - -@media (prefers-reduced-motion: reduce) { - *, *::before, *::after { - animation-duration: 0.01ms !important; - animation-iteration-count: 1 !important; - transition-duration: 0.01ms !important; - } -} diff --git a/web/worker.js b/web/worker.js deleted file mode 100644 index 439d0a6..0000000 --- a/web/worker.js +++ /dev/null @@ -1,41 +0,0 @@ -// One-shot unpack worker: each unpack runs in a fresh worker with its own -// wasm instance. Two reasons: -// -// 1. UI stays responsive — unpacking a 100 MB+ module blocks for seconds. -// 2. Trap isolation — the senbei DLL-first routing probe relies on -// catch_unwind to reject EXE-shell-layout DLLs, and panics cannot be -// caught in WebAssembly: the probe traps the whole call. A trap kills -// this worker's message handler, which the main thread observes and -// retries with the forced-EXE pipeline in a NEW worker (the trapped -// instance is never reused). That reproduces the CLI's -// dll-first/exe-fallback routing without a catchable panic. - -import init, { unpack_file, unpack_file_force_exe } from './pkg/senbei_web.js'; - -let ready = null; - -self.onmessage = async (e) => { - const { input, companion, forceExe } = e.data; - try { - ready ??= init(); - await ready; - const r = forceExe - ? unpack_file_force_exe(input, companion ?? undefined) - : unpack_file(input, companion ?? undefined); - const bytes = r.bytes; - self.postMessage( - { - ok: true, - kind: r.kind, - suspect: r.suspect, - issues: r.issues, - companion: r.companion, - bytes, - }, - [bytes.buffer], - ); - } catch (err) { - const trap = err instanceof WebAssembly.RuntimeError; - self.postMessage({ ok: false, trap, message: String(err?.message ?? err) }); - } -}; From ab1a14c9d11808d387cb3646e8faa552c46b7cf9 Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Tue, 11 Aug 2026 19:24:32 +0800 Subject: [PATCH 4/8] perf(scan): skip extensionless files by default --- senbei-cli/src/main.rs | 4 +-- senbei-io/src/job.rs | 6 ++--- senbei-io/src/scan.rs | 59 ++++++++++++++++++++++++++++++------------ 3 files changed, 47 insertions(+), 22 deletions(-) diff --git a/senbei-cli/src/main.rs b/senbei-cli/src/main.rs index f861f10..2c226bd 100644 --- a/senbei-cli/src/main.rs +++ b/senbei-cli/src/main.rs @@ -106,7 +106,7 @@ fn print_help() { ); println!( " --scan-all probe every file in a folder, including ones the scan\n\ - \x20 pre-filter skips (under 4128 bytes, or a bulk-asset\n\ - \x20 extension like .ab/.xml/.acb). Much slower on game trees." + \x20 pre-filter skips (under 4128 bytes, extensionless,\n\ + \x20 or a bulk-asset extension). Much slower on large trees." ); } diff --git a/senbei-io/src/job.rs b/senbei-io/src/job.rs index 32d315a..bf51120 100644 --- a/senbei-io/src/job.rs +++ b/senbei-io/src/job.rs @@ -387,9 +387,9 @@ pub fn run_folder_v( /// /// When `scan_all` is true every regular file under `root` is opened and /// content-probed, instead of skipping ones the free directory metadata already -/// rules out (too small to hold a Crackproof key table, or a bulk-asset -/// extension). See [`crate::scan::find_targets_opts`] — exhaustive scanning is -/// dramatically slower on asset-heavy game trees and finds the same targets. +/// rules out (extensionless, too small to hold a Crackproof key table, or a +/// bulk-asset extension). See [`crate::scan::find_targets_opts`] — exhaustive +/// scanning is dramatically slower on asset-heavy trees. pub fn run_folder_opts( root: &Path, out_dir: Option<&Path>, diff --git a/senbei-io/src/scan.rs b/senbei-io/src/scan.rs index 43f46ee..8f3bbed 100644 --- a/senbei-io/src/scan.rs +++ b/senbei-io/src/scan.rs @@ -28,12 +28,11 @@ const MIN_SIZE: u64 = 4128; /// File extensions that are bulk data by construction and can never be a PE /// image or an il2cpp metadata blob. /// -/// This is deliberately a **deny**-list, not an allow-list: the default is to -/// probe, so anything unrecognised is still opened. Targets are recognised by -/// content, not extension, and can carry arbitrary names — there is no closed -/// set of target extensions an allow-list of `exe`/`dll` could enumerate. -/// Only extensions that are bulk asset or text formats by construction appear -/// here. +/// This is deliberately a **deny**-list, not an executable allow-list: unknown +/// extensions are still probed. Extensionless files are handled separately by +/// [`denied_name`] because asset stores commonly contain tens of thousands of +/// extensionless chunks; exhaustive probing remains available through +/// `--scan-all`. /// /// Set `SENBEI_SCAN_ALL=1` (or pass `--scan-all`) to probe every file regardless. const DENY_EXT: &[&str] = &[ @@ -90,11 +89,12 @@ const DENY_EXT: &[&str] = &[ "sr", ]; -/// Whether `path`'s extension is on [`DENY_EXT`]. Extensionless files are never -/// denied (they could be anything). -fn denied_ext(path: &Path) -> bool { +/// Whether `path` can be skipped from its name alone. Extensionless files and +/// files whose extension is on [`DENY_EXT`] are not opened during a default +/// scan. `--scan-all` remains available when exhaustive probing is required. +fn denied_name(path: &Path) -> bool { let Some(ext) = path.extension() else { - return false; + return true; }; let Some(ext) = ext.to_str() else { return false; @@ -206,12 +206,17 @@ pub fn find_targets_opts(root: &Path, scan_all: bool) -> (Vec, Vec Date: Tue, 11 Aug 2026 19:24:45 +0800 Subject: [PATCH 5/8] fix(unpacker): validate executable entry transforms --- senbei-pe/src/engine/exe/pipeline.rs | 58 ++++++-- senbei-pe/src/engine/integrity.rs | 83 ++++++++++++ senbei-pe/src/engine/layout/dd8.rs | 192 ++++++++++++++++++++++++--- 3 files changed, 308 insertions(+), 25 deletions(-) diff --git a/senbei-pe/src/engine/exe/pipeline.rs b/senbei-pe/src/engine/exe/pipeline.rs index cac25da..ffcb61f 100644 --- a/senbei-pe/src/engine/exe/pipeline.rs +++ b/senbei-pe/src/engine/exe/pipeline.rs @@ -831,14 +831,17 @@ impl<'a> Unpacker<'a> { // (i.e., the 4 bytes immediately after the last instance of that pattern). let v6 = find_v_after_pad(&u.decompressed, stage4_field, stage4_dlen) .unwrap_or_else(|| idb_pos.wrapping_sub(24)); - let mut accum2 = get_u32(&u.decompressed, v6); - for m in 0..3u32 { + let accum2_seed = get_u32(&u.decompressed, v6); + let mut accum2 = accum2_seed; + let mut accum2_candidates = vec![(0u32, accum2_seed)]; + for m in 0..8u32 { let bound = (m + 1).wrapping_mul(25) << 2; let mut i: u32 = 1; while i <= bound { accum2 = accum2.wrapping_add(i); i = i.wrapping_add(1); } + accum2_candidates.push((m + 1, accum2)); } let ops1 = match generate(&u.decompressed, data_offset) { @@ -852,19 +855,58 @@ impl<'a> Unpacker<'a> { let at4 = stage1.wrapping_add(stage2_off.wrapping_add(216)); let stage5_field = get_u32(&u.decompressed, at4); - // at4 is a (src, src_len, dest, dest_len) quad; only src and dest_len - // are needed here, the other two are consumed by the decrypt below. + let stage5_slen = get_u32(&u.decompressed, at4.wrapping_add(4)); + let stage5_dest = get_u32(&u.decompressed, at4.wrapping_add(8)); let stage5_dlen = get_u32(&u.decompressed, at4.wrapping_add(12)); if verbose { println!(" stage5 = 0x{:08X}", stage5_field); + println!( + " stage5 descriptor = [0x{stage5_field:08X}, 0x{stage5_slen:08X}, 0x{stage5_dest:08X}, 0x{stage5_dlen:08X}]" + ); + println!(" stage5 bytecode = 0x{data_offset:08X}"); + println!(" stage5 accumulator seed = 0x{accum2_seed:08X} at 0x{v6:08X}"); } - if let Err(reason) = - u.decrypt_and_decompress_data(at4, xor_acc ^ chk4 ^ chk5 ^ accum2, Some(&ops1)) - { + let stage5_lo = stage5_field.min(stage5_dest) as usize; + let stage5_hi = stage5_field + .checked_add(stage5_slen) + .zip(stage5_dest.checked_add(stage5_dlen)) + .map(|(source_end, dest_end)| source_end.max(dest_end) as usize) + .filter(|&end| stage5_lo <= end && end <= u.decompressed.len()) + .ok_or(UnpackError::BufferRangeOutOfBounds { + operation: BufferOperation::Read, + offset: stage5_lo, + size: stage5_slen.max(stage5_dlen) as usize, + buffer_len: u.decompressed.len(), + })?; + let stage5_backup = u.decompressed[stage5_lo..stage5_hi].to_vec(); + let mut first_failure = None; + let mut selected_rounds = None; + for rounds in std::iter::once(3u32).chain((0..=8).filter(|&rounds| rounds != 3)) { + u.decompressed[stage5_lo..stage5_hi].copy_from_slice(&stage5_backup); + let candidate_accum = accum2_candidates[rounds as usize].1; + match u.decrypt_and_decompress_data( + at4, + xor_acc ^ chk4 ^ chk5 ^ candidate_accum, + Some(&ops1), + ) { + Ok(()) => { + selected_rounds = Some(rounds); + break; + } + Err(reason) => { + first_failure.get_or_insert(reason); + } + } + } + let Some(selected_rounds) = selected_rounds else { + u.decompressed[stage5_lo..stage5_hi].copy_from_slice(&stage5_backup); return Err(UnpackError::StageDecompressionFailed { stage: DecompressionStage::ExeStage5, - reason, + reason: first_failure.expect("at least one Stage5 candidate was tried"), }); + }; + if verbose { + println!(" selected stage5 accumulator rounds = {selected_rounds}"); } // Inside stage5, the loader stores a table of (ptr, size) pairs at a diff --git a/senbei-pe/src/engine/integrity.rs b/senbei-pe/src/engine/integrity.rs index b49cecb..c768511 100644 --- a/senbei-pe/src/engine/integrity.rs +++ b/senbei-pe/src/engine/integrity.rs @@ -77,6 +77,49 @@ fn rva_to_off(secs: &[Section], file_len: usize, rva: u32, need: u32) -> Option< None } +fn is_executable_rva(secs: &[Section], rva: u32) -> bool { + secs.iter().any(|section| { + let span = section.vsize.max(section.raw_size); + rva >= section.va + && rva < section.va.wrapping_add(span) + && (section.chars & 0x2000_0000) != 0 + }) +} + +fn check_common_entry_branches( + stub: &[u8], + ep: u32, + secs: &[Section], + report: &mut IntegrityReport, +) { + if stub.len() < 18 + || stub[0..3] != [0x48, 0x83, 0xEC] + || stub[4] != 0xE8 + || stub[9..12] != [0x48, 0x83, 0xC4] + || stub[12] != stub[3] + || stub[13] != 0xE9 + { + return; + } + for (name, rel_off, instruction_len) in [("call", 5usize, 9i64), ("jump", 14usize, 18i64)] { + let rel = i32::from_le_bytes([ + stub[rel_off], + stub[rel_off + 1], + stub[rel_off + 2], + stub[rel_off + 3], + ]) as i64; + let target = i64::from(ep) + instruction_len + rel; + let valid = u32::try_from(target) + .ok() + .is_some_and(|rva| is_executable_rva(secs, rva)); + if !valid { + report.issues.push(format!( + "entry point {name} target 0x{target:X} is outside executable sections (DD8 selection is likely wrong)" + )); + } + } +} + /// Inspect an unpacked PE image and report any defect that would make the OS /// loader fault at runtime. `out` is the bytes the unpacker produced. pub fn check(out: &[u8]) -> IntegrityReport { @@ -253,6 +296,9 @@ pub fn check(out: &[u8]) -> IntegrityReport { "entry point RVA 0x{ep:X} is not in an executable section" )); } + if let Some(entry_stub) = out.get(off as usize..off as usize + 18) { + check_common_entry_branches(entry_stub, ep, &secs, &mut r); + } } } } @@ -363,3 +409,40 @@ fn looks_like_dll_name(d: &[u8], off: u32) -> bool { } d[start..end].iter().all(|&b| (0x20..0x7F).contains(&b)) } + +#[cfg(test)] +mod tests { + use super::*; + + fn executable_text() -> Vec
      { + vec![Section { + va: 0x1000, + vsize: 0x4000, + raw_ptr: 0x1000, + raw_size: 0x4000, + chars: 0x6000_0020, + }] + } + + #[test] + fn common_entry_stub_rejects_out_of_image_branches() { + let stub = [ + 0x48, 0x83, 0xEC, 0x28, 0xE8, 0x5B, 0x02, 0x41, 0x00, 0x48, 0x83, 0xC4, 0x28, 0xE9, + 0x7A, 0xFE, 0x54, 0xFF, + ]; + let mut report = IntegrityReport::default(); + check_common_entry_branches(&stub, 0x1264, &executable_text(), &mut report); + assert_eq!(report.issues.len(), 2); + } + + #[test] + fn common_entry_stub_accepts_executable_branches() { + let stub = [ + 0x48, 0x83, 0xEC, 0x28, 0xE8, 0x5B, 0x02, 0x00, 0x00, 0x48, 0x83, 0xC4, 0x28, 0xE9, + 0x7A, 0xFE, 0xFF, 0xFF, + ]; + let mut report = IntegrityReport::default(); + check_common_entry_branches(&stub, 0x1264, &executable_text(), &mut report); + assert!(report.ok()); + } +} diff --git a/senbei-pe/src/engine/layout/dd8.rs b/senbei-pe/src/engine/layout/dd8.rs index fc5aca3..b3faa2b 100644 --- a/senbei-pe/src/engine/layout/dd8.rs +++ b/senbei-pe/src/engine/layout/dd8.rs @@ -1,5 +1,7 @@ //! Validation-driven selection for per-page text transforms. +use super::discovery::trial_decrypt5_u32; + /// PE32 `.text` dd8 key-formula selection with a skip decision. The packer keys /// the per-page XOR either with `page+1` or `0x8000*(page+1)`; the formula is /// not recorded. Replays the dd8 page pass on a scratch copy of sample pages @@ -118,26 +120,29 @@ pub fn select_dd8_formula_pe32(data: &[u8], text_off: u32, text_size: u32) -> Op // // The packer scrambles ~1 byte per 16-byte block of .text via decrypt_data8, // keyed by `page_idx << shift` (absolute page index = text_va >> 12). Observed -// shifts are 0 and 15. The shift is NOT stored in any header/config field: -// two otherwise-unrelated builds can carry byte-identical config-version stamps -// (0x40327253) yet require different shifts, so the only reliable discriminator -// is the .text content itself. +// shifts are 0 and 15. The shift is NOT stored in any header/config field, so +// the decision must be validated against the resulting .text content. // -// Detection scoring formula: for each candidate shift, replay decrypt_data8 -// across a few sample pages (25/50/75% of .text) and count how many of the 255 -// mutated positions become 0xCC — the MSVC int3 padding byte. The correct shift -// hits int3 pads disproportionately often (~3-10x the baseline), so the -// highest-scoring shift wins. If neither shift clears 2x the baseline, .text -// is already plaintext → skip (return 99). +// A recognised CRT entry stub is the strongest oracle: decode skip/0/15 and +// require both of its direct rel32 branches to land in executable .text. This +// includes the call/jump displacement bytes themselves; an older entry oracle +// wildcarded those bytes and could accept a stub whose opcodes looked right but +// whose branch targets were outside the image. // -// This replaces an earlier entry-stub oracle that matched the 14 fixed CRT-stub -// bytes at the AEP. That oracle false-positived on a newer EXE-64 build: dd8 -// corrupted only the call rel32 (bytes 5-8, the wildcard region), so the stub -// matched under BOTH shifts and the selector defaulted to 0 when the truth was -// 15. The 0xCC statistic samples hundreds of positions per page and is not -// fooled by a stub whose fixed bytes happen to survive. +// Other entry shapes fall back to padding statistics: replay each shift across +// sample pages and count positions restored to the MSVC int3 padding byte. A +// clear gain selects the shift; otherwise .text is treated as already plain. // --------------------------------------------------------------------------- -pub fn select_dd8_shift(data: &[u8], text_va: u32, text_size: u32, _info3: u32) -> u32 { +pub fn select_dd8_shift(data: &[u8], text_va: u32, text_size: u32, info3: u32) -> u32 { + if let Some((shift, scores)) = select_dd8_by_entry_stub(data, text_va, text_size, info3) { + if std::env::var("SEL_DIAG").is_ok() { + eprintln!( + "SEL dd8 entry best_shift={} none={} s0={} s15={}", + shift, scores[0], scores[1], scores[2] + ); + } + return shift; + } if text_size < 0x1000 { return 0; } @@ -192,6 +197,123 @@ pub fn select_dd8_shift(data: &[u8], text_va: u32, text_size: u32, _info3: u32) best_shift } +/// Select DD8 from the common CRT entry stub when its direct call and jump +/// provide a stronger oracle than sparse padding statistics. The candidate is +/// accepted only when it is the sole one whose two branch targets stay inside +/// `.text`; unrecognised entry code falls through to the padding selector. +fn select_dd8_by_entry_stub( + data: &[u8], + text_va: u32, + text_size: u32, + info3: u32, +) -> Option<(u32, [u8; 3])> { + for entry in entry_candidates(data, text_va, text_size, info3) { + let [Some(none), Some(s0), Some(s15)] = [None, Some(0), Some(15)] + .map(|shift| entry_stub_branch_score(data, text_va, text_size, entry, shift)) + else { + continue; + }; + let scores = [none, s0, s15]; + let best = scores.iter().copied().max()?; + if best == 2 && scores.iter().filter(|&&score| score == best).count() == 1 { + let index = scores.iter().position(|&score| score == best)?; + return Some(([99, 0, 15][index], scores)); + } + } + None +} + +fn entry_candidates(data: &[u8], text_va: u32, text_size: u32, info3: u32) -> Vec { + let text_end = text_va.saturating_add(text_size); + let mut entries = Vec::with_capacity(3); + if let Some(pe) = read_u32(data, 0x3C) + && let Some(entry) = pe.checked_add(40).and_then(|offset| read_u32(data, offset)) + && (text_va..text_end).contains(&entry) + { + entries.push(entry); + } + for metadata_off in [32u32, 64] { + let Some(end) = info3 + .checked_add(metadata_off) + .and_then(|offset| offset.checked_add(8)) + else { + continue; + }; + if end as usize > data.len() { + continue; + } + let entry = trial_decrypt5_u32(data, info3 + metadata_off); + let image_base = trial_decrypt5_u32(data, info3 + metadata_off + 4); + if image_base == info3 && (text_va..text_end).contains(&entry) && !entries.contains(&entry) + { + entries.push(entry); + } + } + entries +} + +fn entry_stub_branch_score( + data: &[u8], + text_va: u32, + text_size: u32, + entry: u32, + shift: Option, +) -> Option { + let text_end = text_va.checked_add(text_size)?; + if entry < text_va || entry.checked_add(18)? > text_end { + return None; + } + + let mut stub = [0u8; 18]; + for (offset, byte) in stub.iter_mut().enumerate() { + *byte = dd8_candidate_byte(data, entry + offset as u32, shift)?; + } + if stub[0..3] != [0x48, 0x83, 0xEC] + || stub[4] != 0xE8 + || stub[9..12] != [0x48, 0x83, 0xC4] + || stub[12] != stub[3] + || stub[13] != 0xE9 + { + return None; + } + + let call_rel = i32::from_le_bytes(stub[5..9].try_into().ok()?) as i64; + let jump_rel = i32::from_le_bytes(stub[14..18].try_into().ok()?) as i64; + let call_target = i64::from(entry) + 9 + call_rel; + let jump_target = i64::from(entry) + 18 + jump_rel; + let in_text = |target: i64| target >= i64::from(text_va) && target < i64::from(text_end); + Some(u8::from(in_text(call_target)) + u8::from(in_text(jump_target))) +} + +fn dd8_candidate_byte(data: &[u8], rva: u32, shift: Option) -> Option { + let mut byte = *data.get(rva as usize)?; + let Some(shift) = shift else { + return Some(byte); + }; + let page = rva >> 12; + let block = (rva & 0xFFF) >> 4; + let mut key = page << shift; + for index in 0..=block { + let mixed = key.rotate_right(15).wrapping_add(index); + key = mixed.wrapping_add(index); + if index != 0 { + let target = (page << 12) + .wrapping_add(index << 4) + .wrapping_add(mixed & 0xF); + if target == rva { + byte ^= key as u8; + } + } + } + Some(byte) +} + +fn read_u32(data: &[u8], offset: u32) -> Option { + let start = offset as usize; + let bytes = data.get(start..start.checked_add(4)?)?; + Some(u32::from_le_bytes(bytes.try_into().ok()?)) +} + // Baseline: count int3 pads already present at the first byte of each 16-byte // block, i.e. the positions dd8 would target if its in-block offset were 0. fn score_dd8_baseline(data: &[u8], text_off: usize, sample_pages: &[u32]) -> u32 { @@ -250,6 +372,42 @@ fn score_dd8_shift( mod tests { use super::*; + fn entry_stub_fixture() -> Vec { + let mut data = vec![0u8; 0x5000]; + data[0x3C..0x40].copy_from_slice(&0x100u32.to_le_bytes()); + data[0x128..0x12C].copy_from_slice(&0x1264u32.to_le_bytes()); + data[0x1264..0x1276].copy_from_slice(&[ + 0x48, 0x83, 0xEC, 0x28, 0xE8, 0x5B, 0x02, 0x00, 0x00, 0x48, 0x83, 0xC4, 0x28, 0xE9, + 0x7A, 0xFE, 0xFF, 0xFF, + ]); + data + } + + fn apply_dd8_page(data: &mut [u8], page_rva: u32, shift: u32) { + let mut key = (page_rva >> 12) << shift; + for index in 0..256u32 { + let mixed = key.rotate_right(15).wrapping_add(index); + key = mixed.wrapping_add(index); + if index == 0 { + continue; + } + let target = page_rva.wrapping_add(index << 4).wrapping_add(mixed & 0xF) as usize; + data[target] ^= key as u8; + } + } + + #[test] + fn entry_stub_selects_plaintext_and_both_dd8_shifts() { + let plain = entry_stub_fixture(); + assert_eq!(select_dd8_shift(&plain, 0x1000, 0x4000, 0), 99); + + for expected in [0u32, 15] { + let mut encrypted = plain.clone(); + apply_dd8_page(&mut encrypted, 0x1000, expected); + assert_eq!(select_dd8_shift(&encrypted, 0x1000, 0x4000, 0), expected); + } + } + /// Seed the first `count` dd8-targeted positions of each sampled page with /// the byte that decodes to `0xCC` under the `page+1` formula — i.e. an /// encrypted `.text` whose plaintext is int3 padding. Positions whose key From 55a31a2371f851ed2be6f8ed164c9d10b14e5df7 Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Wed, 12 Aug 2026 23:57:00 +0800 Subject: [PATCH 6/8] doc(none): add README.md --- .gitignore | 2 ++ README.md | 15 +++++++++++++++ 2 files changed, 17 insertions(+) create mode 100644 README.md diff --git a/.gitignore b/.gitignore index 24a516a..187978d 100644 --- a/.gitignore +++ b/.gitignore @@ -151,3 +151,5 @@ target/ # descend into it to find the re-included README). See senbei/samples/README.md. /samples/* !/samples/README.md + +/test \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000..2ad5dbb --- /dev/null +++ b/README.md @@ -0,0 +1,15 @@ +# Senbei + +## 兼容性 + +| 名称 | 平台 | 版本 | 状态 | +|---------|------|------|:----:| +| プリンセスコネクト！Re:Dive | DMM | 12.6.0 | ✅ | +| ウマ娘 プリティーダービー | DMM | 2.29.5 | ✅ | +| 学園アイドルマスター | DMM | 3.2.3 | ✅ | +| 呪術廻戦 ファントムパレード | DMM | 3.8.0 | ✅ | +| 地獄楽 パラダイスバトル | DMM | 1.6.10 | ❌ | +| アサルトリリィ Last Bullet | DMM | 9.1.1 | ❌ | +| Heaven Burns Red | Steam | 6070195521799211081 | ✅ | +| Madoka Magica Magia Exedra | Steam | 2933870524569652965 | ✅ | +| MaiMai DX | Arcade | SDEZ 1.66 | ✅ | \ No newline at end of file From c5982a64b9b0b275ae4768b1f51c6e13a2dea93f Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Thu, 13 Aug 2026 10:49:07 +0800 Subject: [PATCH 7/8] fix(unpacker): classify checksum range failures --- senbei-crypto/src/primitives.rs | 41 ++++++++++++++++++++++++++-- senbei-pe/src/engine/error.rs | 10 +++++++ senbei-pe/src/engine/exe/pipeline.rs | 32 ++++++++++++++++++++-- 3 files changed, 77 insertions(+), 6 deletions(-) diff --git a/senbei-crypto/src/primitives.rs b/senbei-crypto/src/primitives.rs index 236f137..99b0774 100644 --- a/senbei-crypto/src/primitives.rs +++ b/senbei-crypto/src/primitives.rs @@ -382,11 +382,28 @@ pub fn calculate_checksum(d: &[u8], pos: u32) -> u32 { /// CRC32 chained checksum. The (offset, length) descriptor at `pos` is read /// from `d`; the bytes themselves are read from the separate `clean` buffer -/// (the original file image). `start` is the initial CRC accumulator. -pub fn calculate_checksum2(d: &[u8], clean: &[u8], pos: u32, start: u32) -> u32 { +/// (the original file image). `start` is the initial CRC accumulator. Returns +/// a range error instead of panicking when a descriptor points past `clean`. +pub fn calculate_checksum2( + d: &[u8], + clean: &[u8], + pos: u32, + start: u32, +) -> Result { let offset = get_u32(d, pos); let length = get_u32(d, pos.wrapping_add(4)); - crc32::append(start, &clean[offset as usize..(offset + length) as usize]) + let data_start = offset as usize; + let size = length as usize; + let end = data_start.checked_add(size); + let Some(end) = end.filter(|&end| end <= clean.len()) else { + return Err(crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::Read, + offset: data_start, + size, + buffer_len: clean.len(), + }); + }; + Ok(crc32::append(start, &clean[data_start..end])) } // --------------------------------------------------------------------------- @@ -1095,4 +1112,22 @@ mod tests { assert!(decompress(&mut d, 0x40, 0x80, 0, 4, 3)); assert_eq!(&d[0x80..0x83], &[0x5A, 0x5A, 0x5A]); } + + #[test] + fn checksum2_rejects_source_range_outside_clean_image() { + let mut descriptor = [0u8; 8]; + descriptor[0..4].copy_from_slice(&448u32.to_le_bytes()); + descriptor[4..8].copy_from_slice(&634_432u32.to_le_bytes()); + let clean = vec![0u8; 590_896]; + let error = calculate_checksum2(&descriptor, &clean, 0, 0).expect_err("range must fail"); + assert!(matches!( + error, + crate::Error::BufferRangeOutOfBounds { + operation: crate::BufferOperation::Read, + offset: 448, + size: 634_432, + buffer_len: 590_896, + } + )); + } } diff --git a/senbei-pe/src/engine/error.rs b/senbei-pe/src/engine/error.rs index 3e7cfa6..11ef75c 100644 --- a/senbei-pe/src/engine/error.rs +++ b/senbei-pe/src/engine/error.rs @@ -149,6 +149,16 @@ pub enum UnpackError { buffer_len: usize, }, + #[error( + "EXE checksum descriptor at 0x{descriptor:08X} points outside input (offset {offset}, size {size}, input length {image_len})" + )] + ExeChecksumRangeOutOfBounds { + descriptor: u32, + offset: usize, + size: usize, + image_len: usize, + }, + #[error( "{table} descriptor out of bounds (offset {offset}, size 16, image length {image_len})" )] diff --git a/senbei-pe/src/engine/exe/pipeline.rs b/senbei-pe/src/engine/exe/pipeline.rs index ffcb61f..fbbf5d5 100644 --- a/senbei-pe/src/engine/exe/pipeline.rs +++ b/senbei-pe/src/engine/exe/pipeline.rs @@ -73,8 +73,22 @@ impl<'a> Unpacker<'a> { } // Strategy (a): delegate to primitives::calculate_checksum2 - fn calculate_checksum2(&self, pos: u32, start: u32) -> u32 { - primitives::calculate_checksum2(&self.decompressed, self.file_data, pos, start) + fn calculate_checksum2(&self, pos: u32, start: u32) -> Result { + primitives::calculate_checksum2(&self.decompressed, self.file_data, pos, start).map_err( + |error| match error { + senbei_crypto::Error::BufferRangeOutOfBounds { + offset, + size, + buffer_len, + .. + } => UnpackError::ExeChecksumRangeOutOfBounds { + descriptor: pos, + offset, + size, + image_len: buffer_len, + }, + }, + ) } // Strategy (a): delegate to primitives::decrypt_data1 @@ -1018,7 +1032,19 @@ impl<'a> Unpacker<'a> { let n = get_u32(&u.decompressed, p.wrapping_add(4)); walk3 = walk3.wrapping_add(16); if n != 0 { - chain_crc = u.calculate_checksum2(walk3.wrapping_sub(16), chain_crc); + match u.calculate_checksum2(walk3.wrapping_sub(16), chain_crc) { + Ok(next) => chain_crc = next, + Err(UnpackError::ExeChecksumRangeOutOfBounds { .. }) => { + if verbose { + println!( + " checksum chain terminates at 0x{:08X}: descriptor payload is outside protected input", + walk3.wrapping_sub(16) + ); + } + break; + } + Err(error) => return Err(error), + } } if get_u32(&u.decompressed, walk3.wrapping_sub(16).wrapping_add(4)) == 0 { break; From 763bdbb21fb136a2c50af73ce56e1b0f6ffae5fe Mon Sep 17 00:00:00 2001 From: bfloat16 Date: Thu, 13 Aug 2026 10:49:44 +0800 Subject: [PATCH 8/8] docs(readme): organize compatibility matrix by game --- README.md | 65 +++++++++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 54 insertions(+), 11 deletions(-) diff --git a/README.md b/README.md index 2ad5dbb..3b36bd0 100644 --- a/README.md +++ b/README.md @@ -2,14 +2,57 @@ ## 兼容性 -| 名称 | 平台 | 版本 | 状态 | -|---------|------|------|:----:| -| プリンセスコネクト！Re:Dive | DMM | 12.6.0 | ✅ | -| ウマ娘 プリティーダービー | DMM | 2.29.5 | ✅ | -| 学園アイドルマスター | DMM | 3.2.3 | ✅ | -| 呪術廻戦 ファントムパレード | DMM | 3.8.0 | ✅ | -| 地獄楽 パラダイスバトル | DMM | 1.6.10 | ❌ | -| アサルトリリィ Last Bullet | DMM | 9.1.1 | ❌ | -| Heaven Burns Red | Steam | 6070195521799211081 | ✅ | -| Madoka Magica Magia Exedra | Steam | 2933870524569652965 | ✅ | -| MaiMai DX | Arcade | SDEZ 1.66 | ✅ | \ No newline at end of file +### プリンセスコネクト！Re:Dive + +| 平台 | 版本 | PrincessConnectReDive.exe | GameAssembly.dll | coneshell.dll | +|------|------|--------------------------|------------------|---------------| +| DMM | 12.6.0 | ✅ | ✅ | ✅ | + +### ウマ娘 プリティーダービー + +| 平台 | 版本 | umamusume.exe | GameAssembly.dll | CySpringPlugin.dll | lib_burst_generated.dll | libnative.dll | +|------|------|---------------|------------------|-------------------|------------------------|----------------| +| DMM | 2.29.5 | ✅ | ✅ | ✅ | ✅ | ✅ | + +### 学園アイドルマスター + +| 平台 | 版本 | gakumas.exe | GameAssembly.dll | +|------|------|------------|------------------| +| DMM | 3.2.3 | ✅ | ✅ | + +### 呪術廻戦 ファントムパレード + +| 平台 | 版本 | Jujutsuphanpara.exe | GameAssembly.dll | +|------|------|---------------------|------------------| +| DMM | 3.8.0 | ✅ | ✅ | + +### 地獄楽 パラダイスバトル + +| 平台 | 版本 | paradisebattle_cl.exe | GameAssembly.dll | +|------|------|----------------------|------------------| +| DMM | 1.6.10 | ✅ | ✅ | + +### アサルトリリィ Last Bullet + +| 平台 | 版本 | Assaultlily.exe | GameAssembly.dll | +|------|------|-----------------|------------------| +| DMM | 9.1.1 | ✅ | ✅ | + +### Heaven Burns Red + +| 平台 | 版本 | HeavenBurnsRed.exe | GameAssembly.dll | cpp_gamelib_global_api.dll | cpp_gamelib_socialkit.dll | cpp_gamelib_steam.dll | unity_gamelib_wrapper.dll | +|------|------|--------------------|------------------|-----------------------------|-----------------------------|-----------------------|--------------------------| +| Steam | 6070195521799211081 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | + +### Madoka Magica Magia Exedra + +| 平台 | 版本 | MadokaExedra.exe | GameAssembly.dll | baselib.dll | +|------|------|-----------------|------------------|--------------| +| Steam | 2933870524569652965 | ✅ | ✅ | ✅ | +| Steam | 2619304243038353808 | ✅ | ✅ | ✅ | + +### MaiMai DX + +| 平台 | 版本 | amdaemon.exe | Sinmai.exe | Assembly-CSharp.dll | amdaemon_api.dll | Cake.dll | +|------|------|---------------|-----------|---------------------|------------------|----------| +| Arcade | SDEZ 1.66 | ✅ | ✅ | ✅ | ✅ | ✅ |