Added heuristics-based decompiler to dcb-tools.

This commit is contained in:
Filipe Rodrigues 2020-10-25 19:12:54 +00:00
parent 7d1ebd5fa3
commit 233eec4fa1
4 changed files with 396 additions and 2 deletions

View File

@ -12,6 +12,10 @@ path = "src/extractor/main.rs"
name = "patcher"
path = "src/patcher/main.rs"
[[bin]]
name = "decompiler"
path = "src/decompiler/main.rs"
[dependencies]
# Dcb
dcb = { path = "../dcb" }
@ -20,16 +24,19 @@ dcb = { path = "../dcb" }
ascii = "1.0"
# Helpers
int-conv = "0.1"
byteorder = "1.3"
float-ord = "0.2"
itertools = "0.9"
rand = "0.7"
ref-cast = "1.0"
# Cmd
clap = "2.33"
# Logging
log = "0.4"
simplelog = "0.7"
simplelog = "0.8"
# Error handling
anyhow = "1.0"

View File

@ -0,0 +1,41 @@
//! Cli manager
// Imports
use clap::{App as ClapApp, Arg as ClapArg};
use std::path::{Path, PathBuf};
/// Data from the command line
#[derive(PartialEq, Clone, Debug)]
pub struct CliData {
/// The game file
pub game_file_path: PathBuf,
}
impl CliData {
/// Constructs all of the cli data given and returns it
pub fn new() -> Self {
// Get all matches from cli
let matches = ClapApp::new("Dcb Decompiler")
.version("0.0")
.author("Filipe [...] <[...]@gmail.com>")
.about("Decompiles all code from the Digimon Digital Card Battle `.bin` game file")
.arg(
ClapArg::with_name("GAME_FILE")
.help("Sets the input game file to use")
.required(true)
.index(1),
)
.get_matches();
// Get the input filename
// Note: required
let game_file_path = matches
.value_of("GAME_FILE")
.map(Path::new)
.map(Path::to_path_buf)
.expect("Unable to get required argument `GAME_FILE`");
// Return the cli data
Self { game_file_path }
}
}

View File

@ -0,0 +1,346 @@
//! Decompiler
#![feature(
box_syntax,
backtrace,
panic_info_message,
unsafe_block_in_unsafe_fn,
array_value_iter,
array_chunks,
format_args_capture,
or_patterns
)]
// Lints
#![warn(clippy::restriction, clippy::pedantic, clippy::nursery)]
// Instead of `unwrap`, we must use `expect` and provide a reason
#![forbid(clippy::unwrap_used)]
// We must use `unsafe` in unsafe `fn`s and specify if the guarantee is
// made by the caller or by us.
#![forbid(unsafe_op_in_unsafe_fn)]
// We'll disable the ones we don't need
#![allow(clippy::blanket_clippy_restriction_lints)]
// Necessary items may be inlined using `LTO`, so we don't need to mark them as inline
#![allow(clippy::missing_inline_in_public_items)]
// We prefer tail returns where possible, as they help with code readability in most cases.
#![allow(clippy::implicit_return)]
// We're fine with shadowing, as long as the variable is used for the same purpose.
// Hence why `clippy::shadow_unrelated` isn't allowed.
#![allow(clippy::shadow_reuse, clippy::shadow_same)]
// We panic when we know it won't happen, or if it does happen, then a panic is the best option
#![allow(clippy::panic, clippy::expect_used, clippy::unreachable, clippy::todo)]
// We use `expect` even in functions that return a `Result` / `Option` if there is a logic error
#![allow(clippy::unwrap_in_result)]
// We find it more important to be able to copy paste literals such as `0xabcd1234` than
// being able to read them, which does not provide many benefits
#![allow(clippy::unreadable_literal, clippy::unseparated_literal_suffix)]
// We separate implementations per their functionality usually, such as constructors, getters, setters, and others.
#![allow(clippy::multiple_inherent_impl)]
// Many operations we need to repeat, and to keep symmetry
#![allow(clippy::identity_op)]
// We only introduce items before their first usage, which sometimes is half-way through the code.
// We make sure that we only use the item after introduced, however.
#![allow(clippy::items_after_statements)]
// Useful for when they either change a lot with new variants / data,
// or for symmetry purposes
#![allow(clippy::match_same_arms)]
// In this library we have very grain-level error types, each function
// will have it's own error type ideally, so any errors are explicit
// by the type, without needing a section for them
#![allow(clippy::missing_errors_doc)]
// Although we generally try to avoid this, this can happen due to our module organization.
// In the future, this lint should be removed globally and only enabled for modules which
// actually require the use of it.
#![allow(clippy::module_inception, clippy::module_name_repetitions)]
// We use integer arithmetic and operations with the correct intent
#![allow(clippy::integer_arithmetic, clippy::integer_division)]
// We prefer using match ergonomic where possible
#![allow(clippy::pattern_type_mismatch)]
// Sometimes the blocks make it easier to invert their order
#![allow(clippy::if_not_else)]
// This lint triggers when using `assert`s and `todo`s, which is unsuitable for this project
#![allow(clippy::panic_in_result_fn)]
// We want to print the resulting instructions to stdout in this binary.
#![allow(clippy::print_stdout)]
// Lint goes off when going byte by byte in binary, not useful
#![allow(clippy::large_digit_groups)]
// We don't put the final `else` if it's empty
#![allow(clippy::else_if_without_else)]
// Modules
mod cli;
#[path = "../logger.rs"]
mod logger;
// Exports
use std::collections::{HashMap, HashSet};
// Imports
use anyhow::Context;
use byteorder::{ByteOrder, LittleEndian};
use dcb::{
game::exe::{
instruction::{
Directive, Pos,
PseudoInstruction::{self, Nop},
Raw, Register, SimpleInstruction,
},
Instruction,
},
GameFile,
};
use itertools::Itertools;
use ref_cast::RefCast;
#[allow(clippy::too_many_lines)] // TODO: Refactor
fn main() -> Result<(), anyhow::Error> {
// Initialize the logger and set the panic handler
logger::init();
// Get all data from cli
let cli::CliData { game_file_path } = cli::CliData::new();
// Open the game file
let input_file = std::fs::File::open(&game_file_path).context("Unable to open input file")?;
let mut game_file = GameFile::from_reader(input_file).context("Unable to parse input file as dcb")?;
// Read the executable
let exe = dcb::game::Exe::deserialize(&mut game_file).context("Unable to parse game executable")?;
// Get all instructions
let instructions: Vec<(Pos, Instruction)> = Instruction::new_iter(
exe.data
.array_chunks::<4>()
.map(|bytes| LittleEndian::read_u32(bytes))
.zip(0..)
.map(|(word, offset)| Raw {
repr: word,
pos: Pos(exe.header.dest + 4 * offset),
}),
)
.collect();
// All instruction offsets
let offsets: HashSet<Pos> = instructions.iter().map(|(offset, _)| offset).copied().collect();
// All data / string addresses
let data_string_addresses: HashSet<Pos> = instructions
.iter()
.filter_map(|(_, instruction)| match instruction {
Instruction::Pseudo(
PseudoInstruction::La { target: offset, .. } |
PseudoInstruction::Li32 { imm: offset, .. } |
PseudoInstruction::LbImm { offset, .. } |
PseudoInstruction::LbuImm { offset, .. } |
PseudoInstruction::LhImm { offset, .. } |
PseudoInstruction::LhuImm { offset, .. } |
PseudoInstruction::LwlImm { offset, .. } |
PseudoInstruction::LwImm { offset, .. } |
PseudoInstruction::LwrImm { offset, .. } |
PseudoInstruction::SbImm { offset, .. } |
PseudoInstruction::ShImm { offset, .. } |
PseudoInstruction::SwlImm { offset, .. } |
PseudoInstruction::SwImm { offset, .. } |
PseudoInstruction::SwrImm { offset, .. },
) |
Instruction::Directive(Directive::Dw(offset) | Directive::DwRepeated { value: offset, .. }) => Some(Pos(*offset)),
_ => None,
})
.collect();
// Get all function jumps
let funcs_pos: HashMap<Pos, usize> = instructions
.iter()
.filter_map(|(_, instruction)| match instruction {
Instruction::Simple(SimpleInstruction::Jal { target }) |
Instruction::Directive(Directive::Dw(target) | Directive::DwRepeated { value: target, .. }) => Some(Pos(*target)),
_ => None,
})
.filter(|target| (Instruction::CODE_START..Instruction::CODE_END).contains(target) && offsets.contains(target))
.unique()
.zip(0..)
.collect();
// Get all local jumps
let locals_pos: HashMap<Pos, usize> = instructions
.iter()
.filter_map(|(_, instruction)| match instruction {
Instruction::Simple(
SimpleInstruction::J { target } |
SimpleInstruction::Beq { target, .. } |
SimpleInstruction::Bne { target, .. } |
SimpleInstruction::Bltz { target, .. } |
SimpleInstruction::Bgez { target, .. } |
SimpleInstruction::Bgtz { target, .. } |
SimpleInstruction::Blez { target, .. } |
SimpleInstruction::Bltzal { target, .. } |
SimpleInstruction::Bgezal { target, .. },
) |
Instruction::Pseudo(
PseudoInstruction::Beqz { target, .. } | PseudoInstruction::Bnez { target, .. } | PseudoInstruction::B { target },
) => Some(Pos(*target)),
_ => None,
})
.filter(|target| (Instruction::CODE_START..Instruction::CODE_END).contains(target) && offsets.contains(target))
.unique()
.zip(0..)
.collect();
// Get all returns
let return_pos: HashSet<Pos> = instructions
.iter()
.filter_map(|(cur_pos, instruction)| match instruction {
Instruction::Simple(SimpleInstruction::Jr { rs: Register::Ra }) => Some(*cur_pos),
_ => None,
})
.collect();
// Get all strings
let strings_pos: HashMap<Pos, usize> = instructions
.iter()
.filter_map(|(cur_pos, instruction)| match instruction {
Instruction::Directive(Directive::Ascii(_)) => Some(*cur_pos),
_ => None,
})
.filter(|cur_pos| data_string_addresses.contains(cur_pos))
.unique()
.zip(0..)
.collect();
// Get all data
let data_pos: HashMap<Pos, usize> = instructions
.iter()
.filter_map(|(cur_pos, instruction)| match instruction {
Instruction::Directive(Directive::Dw(_) | Directive::DwRepeated { .. }) => Some(*cur_pos),
_ => None,
})
.filter(|cur_pos| data_string_addresses.contains(cur_pos))
.unique()
.zip(0..)
.collect();
// Read all instructions
let mut last_instruction = None;
let mut skipped_nops = 0;
for (offset, instruction) in &instructions {
// If both last and current instructions are nops, skip
if let (Some(&Instruction::Pseudo(Nop)), Instruction::Pseudo(Nop)) = (last_instruction, instruction) {
skipped_nops += 1;
continue;
}
// If we skipped any nops, output the number of skipped nops
// TODO: Merge nops in `Pseudo` or something.
if skipped_nops != 0 {
println!("# + {skipped_nops} x nop");
skipped_nops = 0;
}
// Check if we need to prefix
if let Some(func_idx) = funcs_pos.get(offset) {
println!("\n\tfunc_{func_idx}:");
}
if let Some(local_idx) = locals_pos.get(offset) {
println!("\t.{local_idx}:");
}
if let Some(string_idx) = strings_pos.get(offset) {
println!("\tstring_{string_idx}:");
}
if let Some(data_idx) = data_pos.get(offset) {
println!("\tdata_{data_idx}:");
}
// Print the instruction
print!("{offset:#010x}: {instruction}");
// Check if we should have any comments with this instruction
// TODO: Add Pseudo jumps too
match instruction {
// If we have a jump, make a comment with it's target
Instruction::Simple(
SimpleInstruction::J { target } |
SimpleInstruction::Jal { target } |
SimpleInstruction::Beq { target, .. } |
SimpleInstruction::Bne { target, .. } |
SimpleInstruction::Bltz { target, .. } |
SimpleInstruction::Bgez { target, .. } |
SimpleInstruction::Bgtz { target, .. } |
SimpleInstruction::Blez { target, .. } |
SimpleInstruction::Bltzal { target, .. } |
SimpleInstruction::Bgezal { target, .. },
) => {
print!(" #");
if let Some(func_idx) = funcs_pos.get(Pos::ref_cast(target)) {
print!(" func_{func_idx}");
}
if let Some(local_idx) = locals_pos.get(Pos::ref_cast(target)) {
print!(" .{local_idx}");
}
},
// Comment returns
Instruction::Simple(SimpleInstruction::Jr { rs: Register::Ra }) => {
print!(" # Return");
},
// Comment loading address, loading and writing values of string and data
// TODO: Maybe check loads / writes to halfway between
// the strings / data.
Instruction::Pseudo(
PseudoInstruction::La { target: offset, .. } |
PseudoInstruction::Li32 { imm: offset, .. } |
PseudoInstruction::LbImm { offset, .. } |
PseudoInstruction::LbuImm { offset, .. } |
PseudoInstruction::LhImm { offset, .. } |
PseudoInstruction::LhuImm { offset, .. } |
PseudoInstruction::LwlImm { offset, .. } |
PseudoInstruction::LwImm { offset, .. } |
PseudoInstruction::LwrImm { offset, .. } |
PseudoInstruction::SbImm { offset, .. } |
PseudoInstruction::ShImm { offset, .. } |
PseudoInstruction::SwlImm { offset, .. } |
PseudoInstruction::SwImm { offset, .. } |
PseudoInstruction::SwrImm { offset, .. },
) => {
print!(" #");
if let Some(string_idx) = strings_pos.get(Pos::ref_cast(offset)) {
print!(" string_{string_idx}");
}
if let Some(data_idx) = data_pos.get(Pos::ref_cast(offset)) {
print!(" data_{data_idx}");
}
},
// Comment `dw`s with both function and data
Instruction::Directive(Directive::Dw(offset) | Directive::DwRepeated { value: offset, .. }) => {
print!(" #");
if let Some(func_idx) = funcs_pos.get(Pos::ref_cast(offset)) {
print!(" func_{func_idx}");
}
if let Some(local_idx) = locals_pos.get(Pos::ref_cast(offset)) {
print!(" .{local_idx}");
}
if let Some(string_idx) = strings_pos.get(Pos::ref_cast(offset)) {
print!(" string_{string_idx}");
}
if let Some(data_idx) = data_pos.get(Pos::ref_cast(offset)) {
print!(" data_{data_idx}");
}
},
_ => (),
}
// And finish the line
println!();
// If the _last_ instruction was a return, print a newline after this one
if return_pos.contains(&(offset - 4)) {
println!();
}
last_instruction = Some(instruction);
}
Ok(())
}

View File

@ -11,7 +11,7 @@ type BoxedLogger = Box<dyn SharedLogger>;
pub fn init() {
// All loggers to try and initialize
let loggers = [
TermLogger::new(LevelFilter::Warn, Config::default(), TerminalMode::Mixed).map(|logger| BoxedLogger::from(logger)),
Some(TermLogger::new(LevelFilter::Warn, Config::default(), TerminalMode::Mixed)).map(|logger| BoxedLogger::from(logger)),
std::fs::File::create("latest.log")
.ok()
.map(|file| WriteLogger::new(LevelFilter::Trace, Config::default(), file))