encounter
/
decomp-toolkit
mirror of https://github.com/encounter/decomp-toolkit.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Analyzer fixes galore 

- Transparent NLZSS decompression (add `:nlzss` to path)
- Overhaul portions of the analyzer to support more games
- Reject some invalid data relocations automatically
- Jump table analysis fixes
This commit is contained in:
Luke Street 2023-09-13 02:08:51 -04:00
parent 50f913c4eb
commit d4ef1ce16a
29 changed files with 1533 additions and 669 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
Cargo.lock generated
View File

@ -37,6 +37,15 @@ dependencies = [
"memchr",
]
[[package]]
name = "ansi_term"
version = "0.12.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d52a9bb7ec0cf484c551830a7ce27bd20d67eac647e1befb56b0be4ee39a55d2"
dependencies = [
"winapi",
]
[[package]]
name = "anyhow"
version = "1.0.71"
@ -225,6 +234,21 @@ version = "1.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
[[package]]
name = "clap"
version = "2.34.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a0610544180c38b88101fecf2dd634b174a62eef6946f84dfc6a7127512b381c"
dependencies = [
"ansi_term",
"atty",
"bitflags",
"strsim",
"textwrap",
"unicode-width",
"vec_map",
]
[[package]]
name = "cpufeatures"
version = "0.2.5"
@ -307,7 +331,7 @@ dependencies = [
[[package]]
name = "decomp-toolkit"
version = "0.4.2"
version = "0.5.0"
dependencies = [
"anyhow",
"ar",
@ -330,6 +354,7 @@ dependencies = [
"memchr",
"memmap2",
"multimap",
"nintendo-lz",
"num_enum",
"object 0.31.1",
"once_cell",
@ -613,6 +638,16 @@ dependencies = [
"serde",
]
[[package]]
name = "nintendo-lz"
version = "0.1.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "768b95cd65a1a8b82d6c7e90a69d080b20996a876cd62987ab5bcb350c5ae646"
dependencies = [
"byteorder",
"clap",
]
[[package]]
name = "nu-ansi-term"
version = "0.46.0"
@ -985,6 +1020,12 @@ version = "1.11.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "62bb4feee49fdd9f707ef802e22365a35de4b7b299de4763d44bfea899442ff9"
[[package]]
name = "strsim"
version = "0.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8ea5119cdb4c55b55d432abb513a0429384878c15dde60cc77b1c99de1a95a6a"
[[package]]
name = "supports-color"
version = "1.3.1"
@ -1017,6 +1058,15 @@ dependencies = [
"unicode-ident",
]
[[package]]
name = "textwrap"
version = "0.11.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d326610f408c7a4eb6f51c37c330e496b08506c9457c9d34287ecc38809fb060"
dependencies = [
"unicode-width",
]
[[package]]
name = "thiserror"
version = "1.0.37"
@ -1157,6 +1207,12 @@ version = "0.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "830b7e5d4d90034032940e4ace0d9a9a057e7a45cd94e6c007832e39edb82f6d"
[[package]]
name = "vec_map"
version = "0.8.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f1bddf1187be692e79c5ffeab891132dfb0f236ed36a43c7ed39f1165ee20191"
[[package]]
name = "version_check"
version = "0.9.4"

3
Cargo.toml
View File

@ -3,7 +3,7 @@ name = "decomp-toolkit"
description = "Yet another GameCube/Wii decompilation toolkit."
authors = ["Luke Street <luke@street.dev>"]
license = "MIT OR Apache-2.0"
version = "0.4.2"
version = "0.5.0"
edition = "2021"
publish = false
build = "build.rs"
@ -42,6 +42,7 @@ log = "0.4.19"
memchr = "2.5.0"
memmap2 = "0.7.1"
multimap = "0.9.0"
nintendo-lz = "0.1.3"
num_enum = "0.6.1"
object = { version = "0.31.1", features = ["read_core", "std", "elf", "write_std"], default-features = false }
once_cell = "1.18.0"

3
assets/ldscript_partial.lcf
View File

@ -14,8 +14,5 @@ SECTIONS
FORCEACTIVE
{
_unresolved
_prolog
_epilog
$FORCEACTIVE
}

249
src/analysis/cfa.rs
View File

@ -1,10 +1,12 @@
use std::{
collections::{BTreeMap, BTreeSet},
cmp::min,
collections::BTreeMap,
fmt::{Debug, Display, Formatter, UpperHex},
ops::{Add, AddAssign, BitAnd, Sub},
};
use anyhow::{bail, ensure, Context, Result};
use itertools::Itertools;
use crate::{
analysis::{
@ -12,6 +14,7 @@ use crate::{
skip_alignment,
slices::{FunctionSlices, TailCallResult},
vm::{BranchTarget, GprValue, StepResult, VM},
RelocationTarget,
},
obj::{ObjInfo, ObjSectionKind, ObjSymbol, ObjSymbolFlagSet, ObjSymbolFlags, ObjSymbolKind},
};
@ -36,6 +39,20 @@ impl Display for SectionAddress {
impl SectionAddress {
pub fn new(section: usize, address: u32) -> Self { Self { section, address } }
pub fn offset(self, offset: i32) -> Self {
Self { section: self.section, address: self.address.wrapping_add_signed(offset) }
}
pub fn align_up(self, align: u32) -> Self {
Self { section: self.section, address: (self.address + align - 1) & !(align - 1) }
}
pub fn align_down(self, align: u32) -> Self {
Self { section: self.section, address: self.address & !(align - 1) }
}
pub fn is_aligned(self, align: u32) -> bool { self.address & (align - 1) == 0 }
}
impl Add<u32> for SectionAddress {
@ -70,16 +87,36 @@ impl BitAnd<u32> for SectionAddress {
fn bitand(self, rhs: u32) -> Self::Output { self.address & rhs }
}
#[derive(Default, Debug, Clone)]
pub struct FunctionInfo {
pub analyzed: bool,
pub end: Option<SectionAddress>,
pub slices: Option<FunctionSlices>,
}
impl FunctionInfo {
pub fn is_analyzed(&self) -> bool { self.analyzed }
pub fn is_function(&self) -> bool {
self.analyzed && self.end.is_some() && self.slices.is_some()
}
pub fn is_non_function(&self) -> bool {
self.analyzed && self.end.is_none() && self.slices.is_none()
}
pub fn is_unfinalized(&self) -> bool {
self.analyzed && self.end.is_none() && self.slices.is_some()
}
}
#[derive(Debug, Default)]
pub struct AnalyzerState {
pub sda_bases: Option<(u32, u32)>,
pub function_entries: BTreeSet<SectionAddress>,
pub function_bounds: BTreeMap<SectionAddress, Option<SectionAddress>>,
pub function_slices: BTreeMap<SectionAddress, FunctionSlices>,
pub functions: BTreeMap<SectionAddress, FunctionInfo>,
pub jump_tables: BTreeMap<SectionAddress, u32>,
pub known_symbols: BTreeMap<SectionAddress, ObjSymbol>,
pub known_sections: BTreeMap<usize, String>,
pub non_finalized_functions: BTreeMap<SectionAddress, FunctionSlices>,
}
impl AnalyzerState {
@ -87,7 +124,7 @@ impl AnalyzerState {
for (&section_index, section_name) in &self.known_sections {
obj.sections[section_index].rename(section_name.clone())?;
}
for (&start, &end) in &self.function_bounds {
for (&start, FunctionInfo { end, .. }) in self.functions.iter() {
let Some(end) = end else { continue };
let section = &obj.sections[start.section];
ensure!(
@ -120,7 +157,14 @@ impl AnalyzerState {
false,
)?;
}
for (&addr, &size) in &self.jump_tables {
let mut iter = self.jump_tables.iter().peekable();
while let Some((&addr, &(mut size))) = iter.next() {
// Truncate overlapping jump tables
if let Some((&next_addr, _)) = iter.peek() {
if next_addr.section == addr.section {
size = min(size, next_addr.address - addr.address);
}
}
let section = &obj.sections[addr.section];
ensure!(
section.contains_range(addr.address..addr.address + size),
@ -166,27 +210,31 @@ impl AnalyzerState {
pub fn detect_functions(&mut self, obj: &ObjInfo) -> Result<()> {
// Apply known functions from extab
for (&addr, &size) in &obj.known_functions {
self.function_entries.insert(addr);
self.function_bounds.insert(addr, Some(addr + size));
self.functions.insert(addr, FunctionInfo {
analyzed: false,
end: size.map(|size| addr + size),
slices: None,
});
}
// Apply known functions from symbols
for (_, symbol) in obj.symbols.by_kind(ObjSymbolKind::Function) {
let Some(section_index) = symbol.section else { continue };
let addr_ref = SectionAddress::new(section_index, symbol.address as u32);
self.function_entries.insert(addr_ref);
if symbol.size_known {
self.function_bounds.insert(addr_ref, Some(addr_ref + symbol.size as u32));
}
self.functions.insert(addr_ref, FunctionInfo {
analyzed: false,
end: if symbol.size_known { Some(addr_ref + symbol.size as u32) } else { None },
slices: None,
});
}
// Also check the beginning of every code section
for (section_index, section) in obj.sections.by_kind(ObjSectionKind::Code) {
self.function_entries
.insert(SectionAddress::new(section_index, section.address as u32));
self.functions
.entry(SectionAddress::new(section_index, section.address as u32))
.or_default();
}
// Process known functions first
let known_functions = self.function_entries.clone();
for addr in known_functions {
for addr in self.functions.keys().cloned().collect_vec() {
self.process_function_at(obj, addr)?;
}
if let Some(entry) = obj.entry.map(|n| n as u32) {
@ -203,26 +251,46 @@ impl AnalyzerState {
while self.finalize_functions(obj, true)? {
self.process_functions(obj)?;
}
if self.functions.iter().any(|(_, i)| i.is_unfinalized()) {
log::error!("Failed to finalize functions:");
for (addr, _) in self.functions.iter().filter(|(_, i)| i.is_unfinalized()) {
log::error!(" {:#010X}", addr);
}
bail!("Failed to finalize functions");
}
Ok(())
}
fn finalize_functions(&mut self, obj: &ObjInfo, finalize: bool) -> Result<bool> {
let mut finalized = Vec::new();
for (&addr, slices) in &mut self.non_finalized_functions {
let mut finalized_any = false;
let unfinalized = self
.functions
.iter()
.filter_map(|(&addr, info)| {
if info.is_unfinalized() {
info.slices.clone().map(|s| (addr, s))
} else {
None
}
})
.collect_vec();
for (addr, mut slices) in unfinalized {
// log::info!("Trying to finalize {:#010X}", addr);
let Some(function_start) = slices.start() else {
bail!("Function slice without start @ {:#010X}", addr);
};
let function_end = slices.end();
let mut current = SectionAddress::new(addr.section, 0);
while let Some(&block) = slices.possible_blocks.range(current + 4..).next() {
current = block;
while let Some((&block, vm)) = slices.possible_blocks.range(current..).next() {
current = block + 4;
let vm = vm.clone();
match slices.check_tail_call(
obj,
block,
function_start,
function_end,
&self.function_entries,
&self.functions,
Some(vm.clone()),
) {
TailCallResult::Not => {
log::trace!("Finalized block @ {:#010X}", block);
@ -232,7 +300,8 @@ impl AnalyzerState {
block,
function_start,
function_end,
&self.function_entries,
&self.functions,
Some(vm),
)?;
}
TailCallResult::Is => {
@ -252,7 +321,8 @@ impl AnalyzerState {
block,
function_start,
function_end,
&self.function_entries,
&self.functions,
Some(vm),
)?;
}
}
@ -261,55 +331,24 @@ impl AnalyzerState {
}
if slices.can_finalize() {
log::trace!("Finalizing {:#010X}", addr);
slices.finalize(obj, &self.function_entries)?;
self.function_entries.append(&mut slices.function_references.clone());
slices.finalize(obj, &self.functions)?;
for address in slices.function_references.iter().cloned() {
self.functions.entry(address).or_default();
}
self.jump_tables.append(&mut slices.jump_table_references.clone());
let end = slices.end();
self.function_bounds.insert(addr, end);
self.function_slices.insert(addr, slices.clone());
finalized.push(addr);
let info = self.functions.get_mut(&addr).unwrap();
info.analyzed = true;
info.end = end;
info.slices = Some(slices.clone());
finalized_any = true;
}
}
let finalized_new = !finalized.is_empty();
for addr in finalized {
self.non_finalized_functions.remove(&addr);
}
Ok(finalized_new)
Ok(finalized_any)
}
fn first_unbounded_function(&self) -> Option<SectionAddress> {
let mut entries_iter = self.function_entries.iter().cloned();
let mut bounds_iter = self.function_bounds.keys().cloned();
let mut entry = entries_iter.next();
let mut bound = bounds_iter.next();
loop {
match (entry, bound) {
(Some(a), Some(b)) => {
if b < a {
bound = bounds_iter.next();
continue;
} else if a != b {
if self.non_finalized_functions.contains_key(&a) {
entry = entries_iter.next();
continue;
} else {
break Some(a);
}
}
}
(Some(a), None) => {
if self.non_finalized_functions.contains_key(&a) {
entry = entries_iter.next();
continue;
} else {
break Some(a);
}
}
_ => break None,
}
entry = entries_iter.next();
bound = bounds_iter.next();
}
self.functions.iter().find(|(_, info)| !info.is_analyzed()).map(|(&addr, _)| addr)
}
fn process_functions(&mut self, obj: &ObjInfo) -> Result<()> {
@ -330,26 +369,29 @@ impl AnalyzerState {
}
pub fn process_function_at(&mut self, obj: &ObjInfo, addr: SectionAddress) -> Result<bool> {
// if addr == 0 || addr == 0xFFFFFFFF {
// log::warn!("Tried to detect @ {:#010X}", addr);
// self.function_bounds.insert(addr, 0);
// return Ok(false);
// }
Ok(if let Some(mut slices) = self.process_function(obj, addr)? {
self.function_entries.insert(addr);
self.function_entries.append(&mut slices.function_references.clone());
for address in slices.function_references.iter().cloned() {
self.functions.entry(address).or_default();
}
self.jump_tables.append(&mut slices.jump_table_references.clone());
if slices.can_finalize() {
slices.finalize(obj, &self.function_entries)?;
self.function_bounds.insert(addr, slices.end());
self.function_slices.insert(addr, slices);
slices.finalize(obj, &self.functions)?;
let info = self.functions.entry(addr).or_default();
info.analyzed = true;
info.end = slices.end();
info.slices = Some(slices);
} else {
self.non_finalized_functions.insert(addr, slices);
let info = self.functions.entry(addr).or_default();
info.analyzed = true;
info.end = None;
info.slices = Some(slices);
}
true
} else {
log::debug!("Not a function @ {:#010X}", addr);
self.function_bounds.insert(addr, None);
let info = self.functions.entry(addr).or_default();
info.analyzed = true;
info.end = None;
false
})
}
@ -360,58 +402,58 @@ impl AnalyzerState {
start: SectionAddress,
) -> Result<Option<FunctionSlices>> {
let mut slices = FunctionSlices::default();
let function_end = self.function_bounds.get(&start).cloned().flatten();
Ok(match slices.analyze(obj, start, start, function_end, &self.function_entries)? {
let function_end = self.functions.get(&start).and_then(|info| info.end);
Ok(match slices.analyze(obj, start, start, function_end, &self.functions, None)? {
true => Some(slices),
false => None,
})
}
fn detect_new_functions(&mut self, obj: &ObjInfo) -> Result<bool> {
let mut found_new = false;
let mut new_functions = vec![];
for (section_index, section) in obj.sections.by_kind(ObjSectionKind::Code) {
let section_start = SectionAddress::new(section_index, section.address as u32);
let section_end = section_start + section.size as u32;
let mut iter = self.function_bounds.range(section_start..section_end).peekable();
let mut iter = self.functions.range(section_start..section_end).peekable();
loop {
match (iter.next(), iter.peek()) {
(Some((&first_begin, &first_end)), Some(&(&second_begin, &second_end))) => {
let Some(first_end) = first_end else { continue };
if first_end > second_begin {
continue;
(Some((&first, first_info)), Some(&(&second, second_info))) => {
let Some(first_end) = first_info.end else { continue };
if first_end > second {
bail!("Overlapping functions {}-{} -> {}", first, first_end, second);
}
let addr = match skip_alignment(section, first_end, second_begin) {
let addr = match skip_alignment(section, first_end, second) {
Some(addr) => addr,
None => continue,
};
if second_begin > addr && self.function_entries.insert(addr) {
if second > addr {
log::trace!(
"Trying function @ {:#010X} (from {:#010X}-{:#010X} <-> {:#010X}-{:#010X?})",
addr,
first_begin,
first.address,
first_end,
second_begin,
second_end,
second.address,
second_info.end,
);
found_new = true;
new_functions.push(addr);
}
}
(Some((&last_begin, &last_end)), None) => {
let Some(last_end) = last_end else { continue };
(Some((last, last_info)), None) => {
let Some(last_end) = last_info.end else { continue };
if last_end < section_end {
let addr = match skip_alignment(section, last_end, section_end) {
Some(addr) => addr,
None => continue,
};
if addr < section_end && self.function_entries.insert(addr) {
log::debug!(
if addr < section_end {
log::trace!(
"Trying function @ {:#010X} (from {:#010X}-{:#010X} <-> {:#010X})",
addr,
last_begin,
last.address,
last_end,
section_end,
);
found_new = true;
new_functions.push(addr);
}
}
}
@ -419,6 +461,11 @@ impl AnalyzerState {
}
}
}
let found_new = !new_functions.is_empty();
for addr in new_functions {
let opt = self.functions.insert(addr, FunctionInfo::default());
ensure!(opt.is_none(), "Attempted to detect duplicate function @ {:#010X}", addr);
}
Ok(found_new)
}
}
@ -446,13 +493,15 @@ pub fn locate_sda_bases(obj: &mut ObjInfo) -> Result<bool> {
}
StepResult::Illegal => bail!("Illegal instruction @ {:#010X}", ins.addr),
StepResult::Jump(target) => {
if let BranchTarget::Address(addr) = target {
if let BranchTarget::Address(RelocationTarget::Address(addr)) = target {
return Ok(ExecCbResult::Jump(addr));
}
}
StepResult::Branch(branches) => {
for branch in branches {
if let BranchTarget::Address(addr) = branch.target {
if let BranchTarget::Address(RelocationTarget::Address(addr)) =
branch.target
{
executor.push(addr, branch.vm, false);
}
}

68
src/analysis/mod.rs
View File

@ -1,6 +1,6 @@
use std::{collections::BTreeSet, num::NonZeroU32};
use anyhow::{anyhow, ensure, Context, Result};
use anyhow::{anyhow, bail, ensure, Context, Result};
use ppc750cl::Ins;
use crate::{
@ -35,13 +35,15 @@ fn read_unresolved_relocation_address(
section: &ObjSection,
address: u32,
reloc_kind: Option<ObjRelocKind>,
) -> Result<Option<SectionAddress>> {
) -> Result<Option<RelocationTarget>> {
if let Some(reloc) = obj
.unresolved_relocations
.iter()
.find(|reloc| reloc.section as usize == section.elf_index && reloc.address == address)
{
ensure!(reloc.module_id == obj.module_id);
if reloc.module_id != obj.module_id {
return Ok(Some(RelocationTarget::External));
}
if let Some(reloc_kind) = reloc_kind {
ensure!(reloc.kind == reloc_kind);
}
@ -52,10 +54,10 @@ fn read_unresolved_relocation_address(
reloc.target_section
)
})?;
Ok(Some(SectionAddress {
Ok(Some(RelocationTarget::Address(SectionAddress {
section: target_section_index,
address: target_section.address as u32 + reloc.addend,
}))
})))
} else {
Ok(None)
}
@ -66,7 +68,7 @@ fn read_relocation_address(
section: &ObjSection,
address: u32,
reloc_kind: Option<ObjRelocKind>,
) -> Result<Option<SectionAddress>> {
) -> Result<Option<RelocationTarget>> {
let Some(reloc) = section.relocations.at(address) else {
return Ok(None);
};
@ -74,13 +76,13 @@ fn read_relocation_address(
ensure!(reloc.kind == reloc_kind);
}
let symbol = &obj.symbols[reloc.target_symbol];
let section_index = symbol.section.with_context(|| {
format!("Symbol '{}' @ {:#010X} missing section", symbol.name, symbol.address)
})?;
Ok(Some(SectionAddress {
let Some(section_index) = symbol.section else {
return Ok(Some(RelocationTarget::External));
};
Ok(Some(RelocationTarget::Address(SectionAddress {
section: section_index,
address: (symbol.address as i64 + reloc.addend) as u32,
}))
})))
}
pub fn read_address(obj: &ObjInfo, section: &ObjSection, address: u32) -> Result<SectionAddress> {
@ -94,7 +96,11 @@ pub fn read_address(obj: &ObjInfo, section: &ObjSection, address: u32) -> Result
Some(ObjRelocKind::Absolute),
)?;
}
opt.with_context(|| {
opt.and_then(|t| match t {
RelocationTarget::Address(addr) => Some(addr),
RelocationTarget::External => None,
})
.with_context(|| {
format!("Failed to find relocation for {:#010X} in section {}", address, section.name)
})
} else {
@ -109,12 +115,18 @@ fn is_valid_jump_table_addr(obj: &ObjInfo, addr: SectionAddress) -> bool {
!matches!(obj.sections[addr.section].kind, ObjSectionKind::Code | ObjSectionKind::Bss)
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum RelocationTarget {
Address(SectionAddress),
External,
}
#[inline(never)]
pub fn relocation_target_for(
obj: &ObjInfo,
addr: SectionAddress,
reloc_kind: Option<ObjRelocKind>,
) -> Result<Option<SectionAddress>> {
) -> Result<Option<RelocationTarget>> {
let section = &obj.sections[addr.section];
let mut opt = read_relocation_address(obj, section, addr.address, reloc_kind)?;
if opt.is_none() {
@ -149,17 +161,24 @@ fn get_jump_table_entries(
if let Some(target) =
relocation_target_for(obj, cur_addr, Some(ObjRelocKind::Absolute))?
{
entries.push(target);
match target {
RelocationTarget::Address(addr) => entries.push(addr),
RelocationTarget::External => {
bail!("Jump table entry at {:#010X} points to external symbol", cur_addr)
}
}
} else {
let entry_addr = u32::from_be_bytes(*array_ref!(data, 0, 4));
let (section_index, _) =
obj.sections.at_address(entry_addr).with_context(|| {
format!(
"Invalid jump table entry {:#010X} at {:#010X}",
entry_addr, cur_addr
)
})?;
entries.push(SectionAddress::new(section_index, entry_addr));
if entry_addr > 0 {
let (section_index, _) =
obj.sections.at_address(entry_addr).with_context(|| {
format!(
"Invalid jump table entry {:#010X} at {:#010X}",
entry_addr, cur_addr
)
})?;
entries.push(SectionAddress::new(section_index, entry_addr));
}
}
data = &data[4..];
cur_addr += 4;
@ -172,7 +191,10 @@ fn get_jump_table_entries(
let target = if let Some(target) =
relocation_target_for(obj, cur_addr, Some(ObjRelocKind::Absolute))?
{
target
match target {
RelocationTarget::Address(addr) => addr,
RelocationTarget::External => break,
}
} else if obj.kind == ObjKind::Executable {
let Some(value) = read_u32(section, cur_addr.address) else {
break;

94
src/analysis/pass.rs
View File

@ -1,11 +1,10 @@
use std::ops::Range;
use anyhow::{bail, ensure, Result};
use flagset::FlagSet;
use itertools::Itertools;
use memchr::memmem;
use crate::{
analysis::cfa::{AnalyzerState, SectionAddress},
analysis::cfa::{AnalyzerState, FunctionInfo, SectionAddress},
obj::{
ObjInfo, ObjKind, ObjRelocKind, ObjSectionKind, ObjSymbol, ObjSymbolFlagSet,
ObjSymbolFlags, ObjSymbolKind,
@ -24,7 +23,9 @@ pub const TRK_TABLE_SIZE: u32 = 0x1F34; // always?
// TRK_MINNOW_DOLPHIN.a __exception.s
impl AnalysisPass for FindTRKInterruptVectorTable {
fn execute(state: &mut AnalyzerState, obj: &ObjInfo) -> Result<()> {
for (&start, _) in state.function_bounds.iter().filter(|&(_, &end)| end.is_none()) {
for (&start, _) in
state.functions.iter().filter(|(_, info)| info.analyzed && info.end.is_none())
{
let section = &obj.sections[start.section];
let data = match section.data_range(start.address, 0) {
Ok(ret) => ret,
@ -70,67 +71,58 @@ impl AnalysisPass for FindTRKInterruptVectorTable {
pub struct FindSaveRestSleds {}
const SLEDS: [([u8; 4], &str, &str); 4] = [
([0xd9, 0xcb, 0xff, 0x70], "__save_fpr", "_savefpr_"),
([0xc9, 0xcb, 0xff, 0x70], "__restore_fpr", "_restfpr_"),
([0x91, 0xcb, 0xff, 0xb8], "__save_gpr", "_savegpr_"),
([0x81, 0xcb, 0xff, 0xb8], "__restore_gpr", "_restgpr_"),
const SLEDS: [([u8; 8], &str, &str); 4] = [
([0xd9, 0xcb, 0xff, 0x70, 0xd9, 0xeb, 0xff, 0x78], "__save_fpr", "_savefpr_"),
([0xc9, 0xcb, 0xff, 0x70, 0xc9, 0xeb, 0xff, 0x78], "__restore_fpr", "_restfpr_"),
([0x91, 0xcb, 0xff, 0xb8, 0x91, 0xeb, 0xff, 0xbc], "__save_gpr", "_savegpr_"),
([0x81, 0xcb, 0xff, 0xb8, 0x81, 0xeb, 0xff, 0xbc], "__restore_gpr", "_restgpr_"),
];
// Runtime.PPCEABI.H.a runtime.c
impl AnalysisPass for FindSaveRestSleds {
fn execute(state: &mut AnalyzerState, obj: &ObjInfo) -> Result<()> {
const SLED_SIZE: usize = 19 * 4; // registers 14-31 + blr
let mut clear_ranges: Vec<Range<SectionAddress>> = vec![];
for (&start, _) in state.function_bounds.iter().filter(|&(_, &end)| end.is_some()) {
let section = &obj.sections[start.section];
let data = match section.data_range(start.address, 0) {
Ok(ret) => ret,
Err(_) => continue,
};
for (section_index, section) in obj.sections.by_kind(ObjSectionKind::Code) {
for (needle, func, label) in &SLEDS {
if data.starts_with(needle) {
log::debug!("Found {} @ {:#010X}", func, start);
clear_ranges.push(start + 4..start + SLED_SIZE as u32);
state.known_symbols.insert(start, ObjSymbol {
name: func.to_string(),
let Some(pos) = memmem::find(&section.data, needle) else {
continue;
};
let start = SectionAddress::new(section_index, section.address as u32 + pos as u32);
log::debug!("Found {} @ {:#010X}", func, start);
state.functions.insert(start, FunctionInfo {
analyzed: false,
end: Some(start + SLED_SIZE as u32),
slices: None,
});
state.known_symbols.insert(start, ObjSymbol {
name: func.to_string(),
demangled_name: None,
address: start.address as u64,
section: Some(start.section),
size: SLED_SIZE as u64,
size_known: true,
flags: ObjSymbolFlagSet(ObjSymbolFlags::Global.into()),
kind: ObjSymbolKind::Function,
align: None,
data_kind: Default::default(),
});
for i in 14..=31 {
let addr = start + (i - 14) * 4;
state.known_symbols.insert(addr, ObjSymbol {
name: format!("{}{}", label, i),
demangled_name: None,
address: start.address as u64,
address: addr.address as u64,
section: Some(start.section),
size: SLED_SIZE as u64,
size: 0,
size_known: true,
flags: ObjSymbolFlagSet(ObjSymbolFlags::Global.into()),
kind: ObjSymbolKind::Function,
kind: ObjSymbolKind::Unknown,
align: None,
data_kind: Default::default(),
});
for i in 14..=31 {
let addr = start + (i - 14) * 4;
state.known_symbols.insert(addr, ObjSymbol {
name: format!("{}{}", label, i),
demangled_name: None,
address: addr.address as u64,
section: Some(start.section),
size: 0,
size_known: true,
flags: ObjSymbolFlagSet(ObjSymbolFlags::Global.into()),
kind: ObjSymbolKind::Unknown,
align: None,
data_kind: Default::default(),
});
}
}
}
}
for range in clear_ranges {
let mut addr = range.start;
while addr < range.end {
state.function_entries.remove(&addr);
state.function_bounds.remove(&addr);
state.function_slices.remove(&addr);
addr += 4;
}
}
Ok(())
}
}
@ -179,7 +171,7 @@ impl AnalysisPass for FindRelCtorsDtors {
};
if target_section.kind != ObjSectionKind::Code
|| !state
.function_bounds
.functions
.contains_key(&SectionAddress::new(target_section_index, reloc.addend))
{
return false;
@ -197,7 +189,7 @@ impl AnalysisPass for FindRelCtorsDtors {
.collect_vec();
if possible_sections.len() != 2 {
log::warn!("Failed to find .ctors and .dtors");
log::debug!("Failed to find .ctors and .dtors");
return Ok(());
}
@ -311,7 +303,7 @@ impl AnalysisPass for FindRelRodataData {
.collect_vec();
if possible_sections.len() != 2 {
log::warn!("Failed to find .rodata and .data");
log::debug!("Failed to find .rodata and .data");
return Ok(());
}

2
src/analysis/signatures.rs
View File

@ -382,7 +382,7 @@ fn apply_init_user_signatures(obj: &mut ObjInfo) -> Result<()> {
// __init_user can be overridden, but we can still look for __init_cpp from it
let mut analyzer = AnalyzerState::default();
analyzer.process_function_at(obj, SectionAddress::new(section_index, symbol.address as u32))?;
for addr in analyzer.function_entries {
for (addr, _) in analyzer.functions {
let section = &obj.sections[addr.section];
if let Some(signature) = check_signatures_str(
section,

218
src/analysis/slices.rs
View File

@ -8,11 +8,12 @@ use ppc750cl::{Ins, Opcode};
use crate::{
analysis::{
cfa::SectionAddress,
cfa::{FunctionInfo, SectionAddress},
disassemble,
executor::{ExecCbData, ExecCbResult, Executor},
uniq_jump_table_entries,
vm::{section_address_for, BranchTarget, StepResult, VM},
RelocationTarget,
},
obj::{ObjInfo, ObjKind, ObjSection},
};
@ -26,10 +27,11 @@ pub struct FunctionSlices {
pub prologue: Option<SectionAddress>,
pub epilogue: Option<SectionAddress>,
// Either a block or tail call
pub possible_blocks: BTreeSet<SectionAddress>,
pub possible_blocks: BTreeMap<SectionAddress, Box<VM>>,
pub has_conditional_blr: bool,
pub has_rfi: bool,
pub finalized: bool,
pub has_r1_load: bool, // Possibly instead of a prologue
}
pub enum TailCallResult {
@ -72,6 +74,25 @@ fn check_sequence(
Ok(found)
}
fn check_prologue_sequence(section: &ObjSection, ins: &Ins) -> Result<bool> {
#[inline(always)]
fn is_mflr(ins: &Ins) -> bool {
// mfspr r0, LR
ins.op == Opcode::Mfspr && ins.field_rD() == 0 && ins.field_spr() == 8
}
#[inline(always)]
fn is_stwu(ins: &Ins) -> bool {
// stwu r1, d(r1)
ins.op == Opcode::Stwu && ins.field_rS() == 1 && ins.field_rA() == 1
}
#[inline(always)]
fn is_stw(ins: &Ins) -> bool {
// stw r0, d(r1)
ins.op == Opcode::Stw && ins.field_rS() == 0 && ins.field_rA() == 1
}
check_sequence(section, ins, &[(&is_stwu, &is_mflr), (&is_mflr, &is_stw)])
}
impl FunctionSlices {
pub fn end(&self) -> Option<SectionAddress> {
self.blocks.last_key_value().and_then(|(_, &end)| end)
@ -109,22 +130,16 @@ impl FunctionSlices {
ins: &Ins,
) -> Result<()> {
#[inline(always)]
fn is_mflr(ins: &Ins) -> bool {
// mfspr r0, LR
ins.op == Opcode::Mfspr && ins.field_rD() == 0 && ins.field_spr() == 8
}
#[inline(always)]
fn is_stwu(ins: &Ins) -> bool {
// stwu r1, d(r1)
ins.op == Opcode::Stwu && ins.field_rS() == 1 && ins.field_rA() == 1
}
#[inline(always)]
fn is_stw(ins: &Ins) -> bool {
// stw r0, d(r1)
ins.op == Opcode::Stw && ins.field_rS() == 0 && ins.field_rA() == 1
fn is_lwz(ins: &Ins) -> bool {
// lwz r1, d(r)
ins.op == Opcode::Lwz && ins.field_rD() == 1
}
if check_sequence(section, ins, &[(&is_stwu, &is_mflr), (&is_mflr, &is_stw)])? {
if is_lwz(ins) {
self.has_r1_load = true;
return Ok(()); // Possibly instead of a prologue
}
if check_prologue_sequence(section, ins)? {
if let Some(prologue) = self.prologue {
if prologue != addr && prologue != addr - 4 {
bail!("Found duplicate prologue: {:#010X} and {:#010X}", prologue, addr)
@ -170,21 +185,37 @@ impl FunctionSlices {
Ok(())
}
fn is_known_function(
&self,
known_functions: &BTreeMap<SectionAddress, FunctionInfo>,
addr: SectionAddress,
) -> Option<SectionAddress> {
if self.function_references.contains(&addr) {
return Some(addr);
}
if let Some((&fn_addr, info)) = known_functions.range(..=addr).next_back() {
if fn_addr == addr || info.end.is_some_and(|end| addr < end) {
return Some(fn_addr);
}
}
None
}
fn instruction_callback(
&mut self,
data: ExecCbData,
obj: &ObjInfo,
function_start: SectionAddress,
function_end: Option<SectionAddress>,
known_functions: &BTreeSet<SectionAddress>,
known_functions: &BTreeMap<SectionAddress, FunctionInfo>,
) -> Result<ExecCbResult<bool>> {
let ExecCbData { executor, vm, result, ins_addr, section, ins, block_start } = data;
// Track discovered prologue(s) and epilogue(s)
self.check_prologue(section, ins_addr, ins)
.with_context(|| format!("While processing {:#010X}", function_start))?;
.with_context(|| format!("While processing {:#010X}: {:#?}", function_start, self))?;
self.check_epilogue(section, ins_addr, ins)
.with_context(|| format!("While processing {:#010X}", function_start))?;
.with_context(|| format!("While processing {:#010X}: {:#?}", function_start, self))?;
if !self.has_conditional_blr && is_conditional_blr(ins) {
self.has_conditional_blr = true;
}
@ -193,9 +224,20 @@ impl FunctionSlices {
}
// If control flow hits a block we thought may be a tail call,
// we know it isn't.
if self.possible_blocks.contains(&ins_addr) {
if self.possible_blocks.contains_key(&ins_addr) {
self.possible_blocks.remove(&ins_addr);
}
if let Some(fn_addr) = self.is_known_function(known_functions, ins_addr) {
if fn_addr != function_start {
log::warn!(
"Control flow from {} hit known function {} (instruction: {})",
function_start,
fn_addr,
ins_addr
);
return Ok(ExecCbResult::End(false));
}
}
match result {
StepResult::Continue | StepResult::LoadStore { .. } => {
@ -214,7 +256,9 @@ impl FunctionSlices {
Ok(ExecCbResult::End(false))
}
StepResult::Jump(target) => match target {
BranchTarget::Unknown => {
BranchTarget::Unknown
| BranchTarget::Address(RelocationTarget::External)
| BranchTarget::JumpTable { address: RelocationTarget::External, .. } => {
// Likely end of function
let next_addr = ins_addr + 4;
self.blocks.insert(block_start, Some(next_addr));
@ -234,34 +278,41 @@ impl FunctionSlices {
self.blocks.insert(block_start, Some(ins_addr + 4));
Ok(ExecCbResult::EndBlock)
}
BranchTarget::Address(addr) => {
BranchTarget::Address(RelocationTarget::Address(addr)) => {
// End of block
self.blocks.insert(block_start, Some(ins_addr + 4));
self.branches.insert(ins_addr, vec![addr]);
if addr == ins_addr {
// Infinite loop
} else if addr >= function_start
&& matches!(function_end, Some(known_end) if addr < known_end)
&& (matches!(function_end, Some(known_end) if addr < known_end)
|| matches!(self.end(), Some(end) if addr < end)
|| addr < ins_addr)
{
// If target is within known function bounds, jump
if self.add_block_start(addr) {
return Ok(ExecCbResult::Jump(addr));
}
} else if matches!(section.data_range(ins_addr.address, ins_addr.address + 4), Ok(data) if data == [0u8; 4])
{
} else if let Some(fn_addr) = self.is_known_function(known_functions, addr) {
ensure!(fn_addr != function_start); // Sanity check
self.function_references.insert(fn_addr);
} else if addr.section != ins_addr.section
// If this branch has zeroed padding after it, assume tail call.
|| matches!(section.data_range(ins_addr.address, ins_addr.address + 4), Ok(data) if data == [0u8; 4])
{
self.function_references.insert(addr);
} else {
self.possible_blocks.insert(addr);
self.possible_blocks.insert(addr, vm.clone_all());
}
Ok(ExecCbResult::EndBlock)
}
BranchTarget::JumpTable { address, size } => {
BranchTarget::JumpTable { address: RelocationTarget::Address(address), size } => {
// End of block
let next_address = ins_addr + 4;
self.blocks.insert(block_start, Some(next_address));
let (mut entries, size) = uniq_jump_table_entries(
log::debug!("Fetching jump table entries @ {} with size {:?}", address, size);
let (entries, size) = uniq_jump_table_entries(
obj,
address,
size,
@ -269,8 +320,12 @@ impl FunctionSlices {
function_start,
function_end.or_else(|| self.end()),
)?;
log::debug!("-> size {}: {:?}", size, entries);
if entries.contains(&next_address)
&& !entries.iter().any(|addr| known_functions.contains(addr))
&& !entries.iter().any(|&addr| {
self.is_known_function(known_functions, addr)
.is_some_and(|fn_addr| fn_addr != function_start)
})
{
self.jump_table_references.insert(address, size);
let mut branches = vec![];
@ -284,7 +339,8 @@ impl FunctionSlices {
} else {
// If the table doesn't contain the next address,
// it could be a function jump table instead
self.possible_blocks.append(&mut entries);
self.possible_blocks
.extend(entries.into_iter().map(|addr| (addr, vm.clone_all())));
}
Ok(ExecCbResult::EndBlock)
}
@ -296,11 +352,15 @@ impl FunctionSlices {
let mut out_branches = vec![];
for branch in branches {
match branch.target {
BranchTarget::Unknown | BranchTarget::Return => {
continue;
}
BranchTarget::Address(addr) => {
if branch.link || known_functions.contains(&addr) {
BranchTarget::Address(RelocationTarget::Address(addr)) => {
let known = self.is_known_function(known_functions, addr);
if let Some(fn_addr) = known {
if fn_addr != function_start {
self.function_references.insert(fn_addr);
continue;
}
}
if branch.link {
self.function_references.insert(addr);
} else {
out_branches.push(addr);
@ -310,8 +370,14 @@ impl FunctionSlices {
}
}
BranchTarget::JumpTable { address, size } => {
bail!("Conditional jump table unsupported @ {:#010X} -> {:#010X} size {:#X?}", ins_addr, address, size);
bail!(
"Conditional jump table unsupported @ {:#010X} -> {:?} size {:#X?}",
ins_addr,
address,
size
);
}
_ => continue,
}
}
if !out_branches.is_empty() {
@ -328,14 +394,15 @@ impl FunctionSlices {
start: SectionAddress,
function_start: SectionAddress,
function_end: Option<SectionAddress>,
known_functions: &BTreeSet<SectionAddress>,
known_functions: &BTreeMap<SectionAddress, FunctionInfo>,
vm: Option<Box<VM>>,
) -> Result<bool> {
if !self.add_block_start(start) {
return Ok(true);
}
let mut executor = Executor::new(obj);
executor.push(start, VM::new_from_obj(obj), false);
executor.push(start, vm.unwrap_or_else(|| VM::new_from_obj(obj)), false);
let result = executor.run(obj, |data| {
self.instruction_callback(data, obj, function_start, function_end, known_functions)
})?;
@ -345,7 +412,8 @@ impl FunctionSlices {
// Visit unreachable blocks
while let Some((first, _)) = self.first_disconnected_block() {
executor.push(first.end, VM::new_from_obj(obj), true);
let vm = self.possible_blocks.remove(&first.start);
executor.push(first.end, vm.unwrap_or_else(|| VM::new_from_obj(obj)), true);
let result = executor.run(obj, |data| {
self.instruction_callback(data, obj, function_start, function_end, known_functions)
})?;
@ -356,13 +424,25 @@ impl FunctionSlices {
// Visit trailing blocks
if let Some(known_end) = function_end {
loop {
let Some(end) = self.end() else {
'outer: loop {
let Some(mut end) = self.end() else {
log::warn!("Trailing block analysis failed @ {:#010X}", function_start);
break;
};
if end >= known_end {
break;
loop {
if end >= known_end {
break 'outer;
}
// Skip nops
match disassemble(&obj.sections[end.section], end.address) {
Some(ins) => {
if !is_nop(&ins) {
break;
}
}
_ => break,
}
end += 4;
}
executor.push(end, VM::new_from_obj(obj), true);
let result = executor.run(obj, |data| {
@ -393,20 +473,23 @@ impl FunctionSlices {
pub fn finalize(
&mut self,
obj: &ObjInfo,
known_functions: &BTreeSet<SectionAddress>,
known_functions: &BTreeMap<SectionAddress, FunctionInfo>,
) -> Result<()> {
ensure!(!self.finalized, "Already finalized");
ensure!(self.can_finalize(), "Can't finalize");
match (self.prologue, self.epilogue) {
(Some(_), Some(_)) | (None, None) => {}
(Some(_), None) => {
match (self.prologue, self.epilogue, self.has_r1_load) {
(Some(_), Some(_), _) | (None, None, _) => {}
(Some(_), None, _) => {
// Likely __noreturn
}
(None, Some(e)) => {
(None, Some(e), false) => {
log::warn!("{:#010X?}", self);
bail!("Unpaired epilogue {:#010X}", e);
}
(None, Some(_), true) => {
// Possible stack setup
}
}
let Some(end) = self.end() else {
@ -425,7 +508,7 @@ impl FunctionSlices {
if !self.has_conditional_blr {
if let Some(ins) = disassemble(section, end.address - 4) {
if ins.op == Opcode::B {
if let Some(target) = ins
if let Some(RelocationTarget::Address(target)) = ins
.branch_dest()
.and_then(|addr| section_address_for(obj, end - 4, addr))
{
@ -450,7 +533,7 @@ impl FunctionSlices {
if self.has_conditional_blr
&& matches!(disassemble(section, end.address - 4), Some(ins) if !ins.is_blr())
&& matches!(disassemble(section, end.address), Some(ins) if ins.is_blr())
&& !known_functions.contains(&end)
&& !known_functions.contains_key(&end)
{
log::trace!("Found trailing blr @ {:#010X}, merging with function", end);
self.blocks.insert(end, Some(end + 4));
@ -459,7 +542,7 @@ impl FunctionSlices {
// Some functions with rfi also include a trailing nop
if self.has_rfi
&& matches!(disassemble(section, end.address), Some(ins) if is_nop(&ins))
&& !known_functions.contains(&end)
&& !known_functions.contains_key(&end)
{
log::trace!("Found trailing nop @ {:#010X}, merging with function", end);
self.blocks.insert(end, Some(end + 4));
@ -480,7 +563,8 @@ impl FunctionSlices {
addr: SectionAddress,
function_start: SectionAddress,
function_end: Option<SectionAddress>,
known_functions: &BTreeSet<SectionAddress>,
known_functions: &BTreeMap<SectionAddress, FunctionInfo>,
vm: Option<Box<VM>>,
) -> TailCallResult {
// If jump target is already a known block or within known function bounds, not a tail call.
if self.blocks.contains_key(&addr) {
@ -521,12 +605,37 @@ impl FunctionSlices {
{
return TailCallResult::Is;
}
// If we haven't discovered a prologue yet, and one exists between the function
// start and the jump target, known tail call.
if self.prologue.is_none() {
let mut current_address = function_start;
while current_address < addr {
let ins = disassemble(target_section, current_address.address).unwrap();
match check_prologue_sequence(target_section, &ins) {
Ok(true) => {
log::debug!(
"Prologue discovered @ {}; known tail call: {}",
current_address,
addr
);
return TailCallResult::Is;
}
Ok(false) => {}
Err(e) => {
log::warn!("Error while checking prologue sequence: {}", e);
return TailCallResult::Error(e);
}
}
current_address += 4;
}
}
// Perform CFA on jump target to determine more
let mut slices = FunctionSlices {
function_references: self.function_references.clone(),
..Default::default()
};
if let Ok(result) = slices.analyze(obj, addr, function_start, function_end, known_functions)
if let Ok(result) =
slices.analyze(obj, addr, function_start, function_end, known_functions, vm)
{
// If analysis failed, assume tail call.
if !result {
@ -545,7 +654,7 @@ impl FunctionSlices {
let other_blocks = self
.possible_blocks
.range(start + 4..end)
.cloned()
.map(|(&addr, _)| addr)
.collect::<Vec<SectionAddress>>();
if !other_blocks.is_empty() {
for other_addr in other_blocks {
@ -563,6 +672,7 @@ impl FunctionSlices {
return TailCallResult::Is;
}
}
// If all else fails, try again later.
TailCallResult::Possible
}

311
src/analysis/tracker.rs
View File

@ -14,6 +14,7 @@ use crate::{
executor::{ExecCbData, ExecCbResult, Executor},
relocation_target_for, uniq_jump_table_entries,
vm::{is_store_op, BranchTarget, GprValue, StepResult, VM},
RelocationTarget,
},
obj::{
ObjDataKind, ObjInfo, ObjKind, ObjReloc, ObjRelocKind, ObjSection, ObjSectionKind,
@ -23,13 +24,31 @@ use crate::{
#[derive(Debug, Copy, Clone)]
pub enum Relocation {
Ha(SectionAddress),
Hi(SectionAddress),
Lo(SectionAddress),
Sda21(SectionAddress),
Rel14(SectionAddress),
Rel24(SectionAddress),
Absolute(SectionAddress),
Ha(RelocationTarget),
Hi(RelocationTarget),
Lo(RelocationTarget),
Sda21(RelocationTarget),
Rel14(RelocationTarget),
Rel24(RelocationTarget),
Absolute(RelocationTarget),
}
impl Relocation {
fn kind_and_address(&self) -> Option<(ObjRelocKind, SectionAddress)> {
let (reloc_kind, target) = match self {
Relocation::Ha(v) => (ObjRelocKind::PpcAddr16Ha, v),
Relocation::Hi(v) => (ObjRelocKind::PpcAddr16Hi, v),
Relocation::Lo(v) => (ObjRelocKind::PpcAddr16Lo, v),
Relocation::Sda21(v) => (ObjRelocKind::PpcEmbSda21, v),
Relocation::Rel14(v) => (ObjRelocKind::PpcRel14, v),
Relocation::Rel24(v) => (ObjRelocKind::PpcRel24, v),
Relocation::Absolute(v) => (ObjRelocKind::Absolute, v),
};
match *target {
RelocationTarget::Address(address) => Some((reloc_kind, address)),
RelocationTarget::External => None,
}
}
}
#[derive(Debug)]
@ -93,13 +112,37 @@ impl Tracker {
#[instrument(name = "tracker", skip(self, obj))]
pub fn process(&mut self, obj: &ObjInfo) -> Result<()> {
self.process_code(obj)?;
for (section_index, section) in obj
.sections
.iter()
.filter(|(_, s)| matches!(s.kind, ObjSectionKind::Data | ObjSectionKind::ReadOnlyData))
{
log::debug!("Processing section {}, address {:#X}", section_index, section.address);
self.process_data(obj, section_index, section)?;
if obj.kind == ObjKind::Executable {
for (section_index, section) in obj.sections.iter().filter(|(_, s)| {
matches!(s.kind, ObjSectionKind::Data | ObjSectionKind::ReadOnlyData)
}) {
log::debug!("Processing section {}, address {:#X}", section_index, section.address);
self.process_data(obj, section_index, section)?;
}
}
self.reject_invalid_relocations(obj)?;
Ok(())
}
/// Remove data relocations that point to an unaligned address if the aligned address has a
/// relocation. A relocation will never point to the middle of an address.
fn reject_invalid_relocations(&mut self, obj: &ObjInfo) -> Result<()> {
let mut to_reject = vec![];
for (&address, reloc) in &self.relocations {
let section = &obj.sections[address.section];
if !matches!(section.kind, ObjSectionKind::Data | ObjSectionKind::ReadOnlyData) {
continue;
}
let Some((_, target)) = reloc.kind_and_address() else {
continue;
};
if !target.is_aligned(4) && self.relocations.contains_key(&target.align_down(4)) {
log::debug!("Rejecting invalid relocation @ {} -> {}", address, target);
to_reject.push(address);
}
}
for address in to_reject {
self.relocations.remove(&address);
}
Ok(())
}
@ -143,6 +186,22 @@ impl Tracker {
Ok(())
}
#[inline]
fn gpr_address(
&self,
obj: &ObjInfo,
ins_addr: SectionAddress,
value: &GprValue,
) -> Option<RelocationTarget> {
match *value {
GprValue::Constant(value) => {
self.is_valid_address(obj, ins_addr, value).map(RelocationTarget::Address)
}
GprValue::Address(address) => Some(address),
_ => None,
}
}
fn instruction_callback(
&mut self,
data: ExecCbData,
@ -162,28 +221,37 @@ impl Tracker {
Opcode::Addi | Opcode::Addic | Opcode::Addic_ => {
let source = ins.field_rA();
let target = ins.field_rD();
if let GprValue::Constant(value) = vm.gpr[target].value {
if let Some(value) = self.is_valid_address(obj, ins_addr, value) {
if (source == 2
&& matches!(self.sda2_base, Some(v) if vm.gpr[2].value == GprValue::Constant(v)))
|| (source == 13
&& matches!(self.sda_base, Some(v) if vm.gpr[13].value == GprValue::Constant(v)))
{
self.relocations.insert(ins_addr, Relocation::Sda21(value));
self.sda_to.insert(value);
} else if let (Some(hi_addr), Some(lo_addr)) =
(vm.gpr[target].hi_addr, vm.gpr[target].lo_addr)
{
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
debug_assert_ne!(value, SectionAddress::new(usize::MAX, 0));
self.relocations.insert(hi_addr, Relocation::Ha(value));
}
let lo_reloc = self.relocations.get(&lo_addr).cloned();
if lo_reloc.is_none() {
self.relocations.insert(lo_addr, Relocation::Lo(value));
}
self.hal_to.insert(value);
if let Some(value) = self.gpr_address(obj, ins_addr, &vm.gpr[target].value)
{
if (source == 2
&& matches!(self.sda2_base, Some(v) if vm.gpr[2].value == GprValue::Constant(v)))
|| (source == 13
&& matches!(self.sda_base, Some(v) if vm.gpr[13].value == GprValue::Constant(v)))
{
self.relocations.insert(ins_addr, Relocation::Sda21(value));
if let RelocationTarget::Address(address) = value {
self.sda_to.insert(address);
}
} else if let (Some(hi_addr), Some(lo_addr)) =
(vm.gpr[target].hi_addr, vm.gpr[target].lo_addr)
{
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
debug_assert_ne!(
value,
RelocationTarget::Address(SectionAddress::new(
usize::MAX,
0
))
);
self.relocations.insert(hi_addr, Relocation::Ha(value));
}
let lo_reloc = self.relocations.get(&lo_addr).cloned();
if lo_reloc.is_none() {
self.relocations.insert(lo_addr, Relocation::Lo(value));
}
if let RelocationTarget::Address(address) = value {
self.hal_to.insert(address);
}
}
}
@ -191,20 +259,21 @@ impl Tracker {
// ori rA, rS, UIMM
Opcode::Ori => {
let target = ins.field_rA();
if let GprValue::Constant(value) = vm.gpr[target].value {
if let Some(value) = self.is_valid_address(obj, ins_addr, value) {
if let (Some(hi_addr), Some(lo_addr)) =
(vm.gpr[target].hi_addr, vm.gpr[target].lo_addr)
{
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
self.relocations.insert(hi_addr, Relocation::Hi(value));
}
let lo_reloc = self.relocations.get(&lo_addr).cloned();
if lo_reloc.is_none() {
self.relocations.insert(lo_addr, Relocation::Lo(value));
}
self.hal_to.insert(value);
if let Some(value) = self.gpr_address(obj, ins_addr, &vm.gpr[target].value)
{
if let (Some(hi_addr), Some(lo_addr)) =
(vm.gpr[target].hi_addr, vm.gpr[target].lo_addr)
{
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
self.relocations.insert(hi_addr, Relocation::Hi(value));
}
let lo_reloc = self.relocations.get(&lo_addr).cloned();
if lo_reloc.is_none() {
self.relocations.insert(lo_addr, Relocation::Lo(value));
}
if let RelocationTarget::Address(address) = value {
self.hal_to.insert(address);
}
}
}
@ -214,20 +283,28 @@ impl Tracker {
Ok(ExecCbResult::Continue)
}
StepResult::LoadStore { address, source, source_reg } => {
if let Some(address) = self.is_valid_address(obj, ins_addr, address.address) {
if self.is_valid_section_address(obj, ins_addr) {
if (source_reg == 2
&& matches!(self.sda2_base, Some(v) if source.value == GprValue::Constant(v)))
|| (source_reg == 13
&& matches!(self.sda_base, Some(v) if source.value == GprValue::Constant(v)))
{
self.relocations.insert(ins_addr, Relocation::Sda21(address));
self.sda_to.insert(address);
if let RelocationTarget::Address(address) = address {
self.sda_to.insert(address);
}
} else {
match (source.hi_addr, source.lo_addr) {
(Some(hi_addr), None) => {
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
debug_assert_ne!(address, SectionAddress::new(usize::MAX, 0));
debug_assert_ne!(
address,
RelocationTarget::Address(SectionAddress::new(
usize::MAX,
0
))
);
self.relocations.insert(hi_addr, Relocation::Ha(address));
}
if hi_reloc.is_none()
@ -235,26 +312,38 @@ impl Tracker {
{
self.relocations.insert(ins_addr, Relocation::Lo(address));
}
self.hal_to.insert(address);
if let RelocationTarget::Address(address) = address {
self.hal_to.insert(address);
}
}
(Some(hi_addr), Some(lo_addr)) => {
let hi_reloc = self.relocations.get(&hi_addr).cloned();
if hi_reloc.is_none() {
debug_assert_ne!(address, SectionAddress::new(usize::MAX, 0));
debug_assert_ne!(
address,
RelocationTarget::Address(SectionAddress::new(
usize::MAX,
0
))
);
self.relocations.insert(hi_addr, Relocation::Ha(address));
}
let lo_reloc = self.relocations.get(&lo_addr).cloned();
if lo_reloc.is_none() {
self.relocations.insert(lo_addr, Relocation::Lo(address));
}
self.hal_to.insert(address);
if let RelocationTarget::Address(address) = address {
self.hal_to.insert(address);
}
}
_ => {}
}
}
self.data_types.insert(address, data_kind_from_op(ins.op));
if is_store_op(ins.op) {
self.stores_to.insert(address);
if let RelocationTarget::Address(address) = address {
self.data_types.insert(address, data_kind_from_op(ins.op));
if is_store_op(ins.op) {
self.stores_to.insert(address);
}
}
}
Ok(ExecCbResult::Continue)
@ -266,22 +355,27 @@ impl Tracker {
function_end
),
StepResult::Jump(target) => match target {
BranchTarget::Unknown | BranchTarget::Return => Ok(ExecCbResult::EndBlock),
BranchTarget::Unknown
| BranchTarget::Return
| BranchTarget::JumpTable { address: RelocationTarget::External, .. } => {
Ok(ExecCbResult::EndBlock)
}
BranchTarget::Address(addr) => {
let next_addr = ins_addr + 4;
if next_addr < function_end {
possible_missed_branches.insert(ins_addr + 4, vm.clone_all());
}
if is_function_addr(addr) {
Ok(ExecCbResult::Jump(addr))
} else {
if ins.is_direct_branch() {
self.relocations.insert(ins_addr, Relocation::Rel24(addr));
if let RelocationTarget::Address(addr) = addr {
if is_function_addr(addr) {
return Ok(ExecCbResult::Jump(addr));
}
Ok(ExecCbResult::EndBlock)
}
if ins.is_direct_branch() {
self.relocations.insert(ins_addr, Relocation::Rel24(addr));
}
Ok(ExecCbResult::EndBlock)
}
BranchTarget::JumpTable { address, size } => {
BranchTarget::JumpTable { address: RelocationTarget::Address(address), size } => {
let (entries, _) = uniq_jump_table_entries(
obj,
address,
@ -301,19 +395,30 @@ impl Tracker {
StepResult::Branch(branches) => {
for branch in branches {
match branch.target {
BranchTarget::Unknown | BranchTarget::Return => {}
BranchTarget::Address(addr) => {
if branch.link || !is_function_addr(addr) {
BranchTarget::Unknown
| BranchTarget::Return
| BranchTarget::JumpTable { address: RelocationTarget::External, .. } => {}
BranchTarget::Address(target) => {
let (addr, is_fn_addr) = if let RelocationTarget::Address(addr) = target
{
(addr, is_function_addr(addr))
} else {
(SectionAddress::new(usize::MAX, 0), false)
};
if branch.link || !is_fn_addr {
self.relocations.insert(ins_addr, match ins.op {
Opcode::B => Relocation::Rel24(addr),
Opcode::Bc => Relocation::Rel14(addr),
Opcode::B => Relocation::Rel24(target),
Opcode::Bc => Relocation::Rel14(target),
_ => continue,
});
} else if is_function_addr(addr) {
} else if is_fn_addr {
executor.push(addr, branch.vm, true);
}
}
BranchTarget::JumpTable { address, size } => {
BranchTarget::JumpTable {
address: RelocationTarget::Address(address),
size,
} => {
let (entries, _) = uniq_jump_table_entries(
obj,
address,
@ -390,13 +495,24 @@ impl Tracker {
for chunk in section.data.chunks_exact(4) {
let value = u32::from_be_bytes(chunk.try_into()?);
if let Some(value) = self.is_valid_address(obj, addr, value) {
self.relocations.insert(addr, Relocation::Absolute(value));
self.relocations
.insert(addr, Relocation::Absolute(RelocationTarget::Address(value)));
}
addr += 4;
}
Ok(())
}
fn is_valid_section_address(&self, obj: &ObjInfo, from: SectionAddress) -> bool {
if let Some((&start, &end)) = obj.blocked_ranges.range(..=from).next_back() {
if from.section == start.section && from.address >= start.address && from.address < end
{
return false;
}
}
true
}
fn is_valid_address(
&self,
obj: &ObjInfo,
@ -410,11 +526,12 @@ impl Tracker {
}
}
// Check for an existing relocation
if let Some(target) = relocation_target_for(obj, from, None).ok().flatten() {
if obj.kind == ObjKind::Executable {
debug_assert_eq!(target.address, addr);
if cfg!(debug_assertions) {
let relocation_target = relocation_target_for(obj, from, None).ok().flatten();
if !matches!(relocation_target, None | Some(RelocationTarget::External)) {
// VM should have already handled this
panic!("Relocation already exists for {:#010X} (from {:#010X})", addr, from);
}
return Some(target);
}
// Remainder of this function is for executable objects only
if obj.kind == ObjKind::Relocatable {
@ -530,17 +647,27 @@ impl Tracker {
}
}
for (addr, reloc) in &self.relocations {
let addr = *addr;
let (reloc_kind, target) = match *reloc {
Relocation::Ha(v) => (ObjRelocKind::PpcAddr16Ha, v),
Relocation::Hi(v) => (ObjRelocKind::PpcAddr16Hi, v),
Relocation::Lo(v) => (ObjRelocKind::PpcAddr16Lo, v),
Relocation::Sda21(v) => (ObjRelocKind::PpcEmbSda21, v),
Relocation::Rel14(v) => (ObjRelocKind::PpcRel14, v),
Relocation::Rel24(v) => (ObjRelocKind::PpcRel24, v),
Relocation::Absolute(v) => (ObjRelocKind::Absolute, v),
for (&addr, reloc) in &self.relocations {
let Some((reloc_kind, target)) = reloc.kind_and_address() else {
// Skip external relocations, they already exist
continue;
};
if obj.kind == ObjKind::Relocatable {
// Sanity check: relocatable objects already have relocations,
// did our analyzer find one that isn't real?
let section = &obj.sections[addr.section];
if section.relocations.at(addr.address).is_none()
// We _do_ want to rebuild missing R_PPC_REL24 relocations
&& !matches!(reloc_kind, ObjRelocKind::PpcRel24)
{
bail!(
"Found invalid relocation {} {:#?} (target {}) in relocatable object",
addr,
reloc,
target
);
}
}
let data_kind = self
.data_types
.get(&target)
@ -607,12 +734,14 @@ impl Tracker {
!= reloc_symbol.address as i64 + addend
{
bail!(
"Conflicting relocations (target {:#010X}): {:#010X?} ({}) != {:#010X?} ({})",
"Conflicting relocations (target {:#010X}): {:#010X?} ({} {:#X}) != {:#010X?} ({} {:#X})",
target,
e.value,
iter_symbol.name,
iter_symbol.address as i64 + e.value.addend,
reloc,
reloc_symbol.name
reloc_symbol.name,
reloc_symbol.address as i64 + addend,
);
}
}

200
src/analysis/vm.rs
View File

@ -3,7 +3,7 @@ use std::num::NonZeroU32;
use ppc750cl::{Argument, Ins, Opcode, GPR};
use crate::{
analysis::{cfa::SectionAddress, relocation_target_for},
analysis::{cfa::SectionAddress, relocation_target_for, RelocationTarget},
obj::{ObjInfo, ObjKind},
};
@ -15,13 +15,13 @@ pub enum GprValue {
/// GPR value is a constant
Constant(u32),
/// GPR value is a known relocated address
Address(SectionAddress),
Address(RelocationTarget),
/// Comparison result (CR field)
ComparisonResult(u8),
/// GPR value is within a range
Range { min: u32, max: u32, step: u32 },
/// GPR value is loaded from an address with a max offset (jump table)
LoadIndexed { address: u32, max_offset: Option<NonZeroU32> },
LoadIndexed { address: RelocationTarget, max_offset: Option<NonZeroU32> },
}
#[derive(Default, Debug, Copy, Clone, Eq, PartialEq)]
@ -52,6 +52,14 @@ impl Gpr {
self.hi_addr = hi_gpr.hi_addr;
self.lo_addr = Some(hi_gpr.lo_addr.unwrap_or(addr));
}
fn address(&self, obj: &ObjInfo, ins_addr: SectionAddress) -> Option<RelocationTarget> {
match self.value {
GprValue::Constant(value) => section_address_for(obj, ins_addr, value),
GprValue::Address(target) => Some(target),
_ => None,
}
}
}
#[derive(Default, Debug, Clone, Eq, PartialEq)]
@ -85,9 +93,9 @@ pub enum BranchTarget {
/// Branch to LR
Return,
/// Branch to address
Address(SectionAddress),
Address(RelocationTarget),
/// Branch to jump table
JumpTable { address: SectionAddress, size: Option<NonZeroU32> },
JumpTable { address: RelocationTarget, size: Option<NonZeroU32> },
}
#[derive(Debug, Clone, Eq, PartialEq)]
@ -105,7 +113,7 @@ pub enum StepResult {
/// Continue normally
Continue,
/// Load from / store to
LoadStore { address: SectionAddress, source: Gpr, source_reg: u8 },
LoadStore { address: RelocationTarget, source: Gpr, source_reg: u8 },
/// Hit illegal instruction
Illegal,
/// Jump without affecting VM state
@ -118,16 +126,16 @@ pub fn section_address_for(
obj: &ObjInfo,
ins_addr: SectionAddress,
target_addr: u32,
) -> Option<SectionAddress> {
) -> Option<RelocationTarget> {
if let Some(target) = relocation_target_for(obj, ins_addr, None).ok().flatten() {
return Some(target);
}
if obj.kind == ObjKind::Executable {
let (section_index, _) = obj.sections.at_address(target_addr).ok()?;
return Some(SectionAddress::new(section_index, target_addr));
return Some(RelocationTarget::Address(SectionAddress::new(section_index, target_addr)));
}
if obj.sections[ins_addr.section].contains(target_addr) {
Some(SectionAddress::new(ins_addr.section, target_addr))
Some(RelocationTarget::Address(SectionAddress::new(ins_addr.section, target_addr)))
} else {
None
}
@ -183,12 +191,6 @@ impl VM {
pub fn clone_all(&self) -> Box<Self> { Box::new(self.clone()) }
pub fn step(&mut self, obj: &ObjInfo, ins_addr: SectionAddress, ins: &Ins) -> StepResult {
// let relocation_target = relocation_target_for(obj, ins_addr, None).ok().flatten();
// if let Some(_target) = relocation_target {
// let _defs = ins.defs();
// // TODO
// }
match ins.op {
Opcode::Illegal => {
return StepResult::Illegal;
@ -201,61 +203,99 @@ impl VM {
(GprValue::Constant(left), GprValue::Constant(right)) => {
GprValue::Constant(left.wrapping_add(right))
}
(
GprValue::Address(RelocationTarget::Address(left)),
GprValue::Constant(right),
) => GprValue::Address(RelocationTarget::Address(left + right)),
(
GprValue::Constant(left),
GprValue::Address(RelocationTarget::Address(right)),
) => GprValue::Address(RelocationTarget::Address(right + left)),
_ => GprValue::Unknown,
};
self.gpr[ins.field_rD()].set_direct(value);
}
// addis rD, rA, SIMM
Opcode::Addis => {
let left = if ins.field_rA() == 0 {
GprValue::Constant(0)
if let Some(target) =
relocation_target_for(obj, ins_addr, None /* TODO */).ok().flatten()
{
debug_assert_eq!(ins.field_rA(), 0);
self.gpr[ins.field_rD()].set_hi(GprValue::Address(target), ins_addr);
} else {
self.gpr[ins.field_rA()].value
};
let value = match left {
GprValue::Constant(value) => {
GprValue::Constant(value.wrapping_add((ins.field_simm() as u32) << 16))
let left = if ins.field_rA() == 0 {
GprValue::Constant(0)
} else {
self.gpr[ins.field_rA()].value
};
let value = match left {
GprValue::Constant(value) => {
GprValue::Constant(value.wrapping_add((ins.field_simm() as u32) << 16))
}
_ => GprValue::Unknown,
};
if ins.field_rA() == 0 {
// lis rD, SIMM
self.gpr[ins.field_rD()].set_hi(value, ins_addr);
} else {
self.gpr[ins.field_rD()].set_direct(value);
}
_ => GprValue::Unknown,
};
if ins.field_rA() == 0 {
// lis rD, SIMM
self.gpr[ins.field_rD()].set_hi(value, ins_addr);
} else {
self.gpr[ins.field_rD()].set_direct(value);
}
}
// addi rD, rA, SIMM
// addic rD, rA, SIMM
// addic. rD, rA, SIMM
Opcode::Addi | Opcode::Addic | Opcode::Addic_ => {
let left = if ins.field_rA() == 0 && ins.op == Opcode::Addi {
GprValue::Constant(0)
if let Some(target) =
relocation_target_for(obj, ins_addr, None /* TODO */).ok().flatten()
{
self.gpr[ins.field_rD()].set_lo(
GprValue::Address(target),
ins_addr,
self.gpr[ins.field_rA()],
);
} else {
self.gpr[ins.field_rA()].value
};
let value = match left {
GprValue::Constant(value) => {
GprValue::Constant(value.wrapping_add(ins.field_simm() as u32))
let left = if ins.field_rA() == 0 && ins.op == Opcode::Addi {
GprValue::Constant(0)
} else {
self.gpr[ins.field_rA()].value
};
let value = match left {
GprValue::Constant(value) => {
GprValue::Constant(value.wrapping_add(ins.field_simm() as u32))
}
GprValue::Address(RelocationTarget::Address(address)) => GprValue::Address(
RelocationTarget::Address(address.offset(ins.field_simm() as i32)),
),
_ => GprValue::Unknown,
};
if ins.field_rA() == 0 {
// li rD, SIMM
self.gpr[ins.field_rD()].set_direct(value);
} else {
self.gpr[ins.field_rD()].set_lo(value, ins_addr, self.gpr[ins.field_rA()]);
}
_ => GprValue::Unknown,
};
if ins.field_rA() == 0 {
// li rD, SIMM
self.gpr[ins.field_rD()].set_direct(value);
} else {
self.gpr[ins.field_rD()].set_lo(value, ins_addr, self.gpr[ins.field_rA()]);
}
}
// ori rA, rS, UIMM
Opcode::Ori => {
let value = match self.gpr[ins.field_rS()].value {
GprValue::Constant(value) => {
GprValue::Constant(value | ins.field_uimm() as u32)
}
_ => GprValue::Unknown,
};
self.gpr[ins.field_rA()].set_lo(value, ins_addr, self.gpr[ins.field_rS()]);
if let Some(target) =
relocation_target_for(obj, ins_addr, None /* TODO */).ok().flatten()
{
self.gpr[ins.field_rA()].set_lo(
GprValue::Address(target),
ins_addr,
self.gpr[ins.field_rS()],
);
} else {
let value = match self.gpr[ins.field_rS()].value {
GprValue::Constant(value) => {
GprValue::Constant(value | ins.field_uimm() as u32)
}
_ => GprValue::Unknown,
};
self.gpr[ins.field_rA()].set_lo(value, ins_addr, self.gpr[ins.field_rS()]);
}
}
// or rA, rS, rB
Opcode::Or => {
@ -336,20 +376,18 @@ impl VM {
Opcode::Bcctr => {
match self.ctr {
GprValue::Constant(value) => {
// TODO only check valid target?
if let Some(target) = section_address_for(obj, ins_addr, value) {
BranchTarget::Address(target)
} else {
BranchTarget::Unknown
}
},
GprValue::Address(target) => BranchTarget::Address(target),
GprValue::LoadIndexed { address, max_offset }
// FIXME: avoids treating bctrl indirect calls as jump tables
if !ins.field_LK() => {
if let Some(target) = section_address_for(obj, ins_addr, address) {
BranchTarget::JumpTable { address: target, size: max_offset.and_then(|n| n.checked_add(4)) }
} else {
BranchTarget::Unknown
}
BranchTarget::JumpTable { address, size: max_offset.and_then(|n| n.checked_add(4)) }
}
_ => BranchTarget::Unknown,
}
@ -369,7 +407,7 @@ impl VM {
if ins.field_LK() {
return StepResult::Branch(vec![
Branch {
target: BranchTarget::Address(ins_addr + 4),
target: BranchTarget::Address(RelocationTarget::Address(ins_addr + 4)),
link: false,
vm: self.clone_for_return(),
},
@ -386,7 +424,7 @@ impl VM {
let mut branches = vec![
// Branch not taken
Branch {
target: BranchTarget::Address(ins_addr + 4),
target: BranchTarget::Address(RelocationTarget::Address(ins_addr + 4)),
link: false,
vm: self.clone_all(),
},
@ -413,15 +451,20 @@ impl VM {
}
// lwzx rD, rA, rB
Opcode::Lwzx => {
let left = self.gpr[ins.field_rA()].value;
let left = self.gpr[ins.field_rA()].address(obj, ins_addr);
let right = self.gpr[ins.field_rB()].value;
let value = match (left, right) {
(GprValue::Constant(address), GprValue::Range { min: _, max, .. })
(Some(address), GprValue::Range { min: _, max, .. })
if /*min == 0 &&*/ max < u32::MAX - 4 && max & 3 == 0 =>
{
GprValue::LoadIndexed { address, max_offset: NonZeroU32::new(max) }
}
(GprValue::Constant(address), _) => {
(Some(address), GprValue::Range { min: _, max, .. })
if /*min == 0 &&*/ max < u32::MAX - 4 && max & 3 == 0 =>
{
GprValue::LoadIndexed { address, max_offset: NonZeroU32::new(max) }
}
(Some(address), _) => {
GprValue::LoadIndexed { address, max_offset: None }
}
_ => GprValue::Unknown,
@ -452,16 +495,29 @@ impl VM {
op if is_load_store_op(op) => {
let source = ins.field_rA();
let mut result = StepResult::Continue;
if let GprValue::Constant(base) = self.gpr[source].value {
let address = base.wrapping_add(ins.field_simm() as u32);
if let GprValue::Address(target) = self.gpr[source].value {
if is_update_op(op) {
self.gpr[source].set_lo(
GprValue::Constant(address),
GprValue::Address(target),
ins_addr,
self.gpr[source],
);
}
result = StepResult::LoadStore {
address: target,
source: self.gpr[source],
source_reg: source as u8,
};
} else if let GprValue::Constant(base) = self.gpr[source].value {
let address = base.wrapping_add(ins.field_simm() as u32);
if let Some(target) = section_address_for(obj, ins_addr, address) {
if is_update_op(op) {
self.gpr[source].set_lo(
GprValue::Address(target),
ins_addr,
self.gpr[source],
);
}
result = StepResult::LoadStore {
address: target,
source: self.gpr[source],
@ -573,11 +629,21 @@ fn split_values_by_crb(crb: u8, left: GprValue, right: GprValue) -> (GprValue, G
#[inline]
fn mask_value(begin: u32, end: u32) -> u32 {
let mut mask = 0u32;
for bit in begin..=end {
mask |= 1 << (31 - bit);
if begin <= end {
let mut mask = 0u32;
for bit in begin..=end {
mask |= 1 << (31 - bit);
}
mask
} else if begin == end + 1 {
u32::MAX
} else {
let mut mask = u32::MAX;
for bit in end + 1..begin {
mask &= !(1 << (31 - bit));
}
mask
}
mask
}
#[inline]

151
src/cmd/dol.rs
View File

@ -5,7 +5,7 @@ use std::{
ffi::OsStr,
fs,
fs::DirBuilder,
io::Write,
io::{Cursor, Write},
mem::take,
path::{Path, PathBuf},
time::Instant,
@ -31,8 +31,9 @@ use crate::{
},
cmd::shasum::file_sha1_string,
obj::{
ObjDataKind, ObjInfo, ObjReloc, ObjRelocKind, ObjSectionKind, ObjSymbol, ObjSymbolFlagSet,
ObjSymbolFlags, ObjSymbolKind, ObjSymbolScope, SymbolIndex,
best_match_for_reloc, ObjDataKind, ObjInfo, ObjKind, ObjReloc, ObjRelocKind,
ObjSectionKind, ObjSymbol, ObjSymbolFlagSet, ObjSymbolFlags, ObjSymbolKind, ObjSymbolScope,
SymbolIndex,
},
util::{
asm::write_asm,
@ -44,10 +45,7 @@ use crate::{
dep::DepFile,
dol::process_dol,
elf::{process_elf, write_elf},
file::{
buf_reader, buf_writer, decompress_if_needed, map_file, touch, verify_hash,
FileIterator, Reader,
},
file::{buf_reader, buf_writer, map_file, touch, verify_hash, FileIterator},
lcf::{asm_path_for_unit, generate_ldscript, obj_path_for_unit},
map::apply_map_file,
rel::{process_rel, process_rel_header},
@ -272,6 +270,7 @@ pub struct OutputModule {
pub module_id: u32,
#[serde(with = "path_slash_serde")]
pub ldscript: PathBuf,
pub entry: Option<String>,
pub units: Vec<OutputUnit>,
}
@ -293,7 +292,7 @@ pub fn run(args: Args) -> Result<()> {
}
fn apply_selfile(obj: &mut ObjInfo, buf: &[u8]) -> Result<()> {
let rso = process_rso(&mut Reader::new(buf))?;
let rso = process_rso(&mut Cursor::new(buf))?;
for symbol in rso.symbols.iter() {
let dol_section_index = match symbol.section {
Some(section) => section,
@ -373,25 +372,26 @@ fn apply_selfile(obj: &mut ObjInfo, buf: &[u8]) -> Result<()> {
fn info(args: InfoArgs) -> Result<()> {
let mut obj = {
let file = map_file(&args.dol_file)?;
let data = decompress_if_needed(file.as_slice())?;
process_dol(data.as_ref(), "")?
process_dol(file.as_slice(), "")?
};
apply_signatures(&mut obj)?;
let mut state = AnalyzerState::default();
FindSaveRestSleds::execute(&mut state, &obj)?;
state.detect_functions(&obj)?;
log::debug!("Discovered {} functions", state.function_slices.len());
log::debug!(
"Discovered {} functions",
state.functions.iter().filter(|(_, i)| i.end.is_some()).count()
);
FindTRKInterruptVectorTable::execute(&mut state, &obj)?;
FindSaveRestSleds::execute(&mut state, &obj)?;
state.apply(&mut obj)?;
apply_signatures_post(&mut obj)?;
if let Some(selfile) = &args.selfile {
let file = map_file(selfile)?;
let data = decompress_if_needed(file.as_slice())?;
apply_selfile(&mut obj, data.as_ref())?;
apply_selfile(&mut obj, file.as_slice())?;
}
println!("{}:", obj.name);
@ -450,19 +450,31 @@ fn update_symbols(obj: &mut ObjInfo, modules: &ModuleMap<'_>, create_symbols: bo
{
if source_module_id == obj.module_id {
// Skip if already resolved
let (_, source_section) = obj
.sections
.get_elf_index(rel_reloc.section as usize)
.ok_or_else(|| anyhow!("Failed to locate REL section {}", rel_reloc.section))?;
let (_, source_section) =
obj.sections.get_elf_index(rel_reloc.section as usize).ok_or_else(|| {
anyhow!(
"Failed to locate REL section {} in module ID {}: source module {}, {:?}",
rel_reloc.section,
obj.module_id,
source_module_id,
rel_reloc
)
})?;
if source_section.relocations.contains(rel_reloc.address) {
continue;
}
}
let (target_section_index, target_section) = obj
.sections
.get_elf_index(rel_reloc.target_section as usize)
.ok_or_else(|| anyhow!("Failed to locate REL section {}", rel_reloc.target_section))?;
let (target_section_index, target_section) =
obj.sections.get_elf_index(rel_reloc.target_section as usize).ok_or_else(|| {
anyhow!(
"Failed to locate REL section {} in module ID {}: source module {}, {:?}",
rel_reloc.target_section,
obj.module_id,
source_module_id,
rel_reloc
)
})?;
if let Some((symbol_index, symbol)) = obj.symbols.for_relocation(
SectionAddress::new(target_section_index, rel_reloc.addend),
@ -517,10 +529,15 @@ fn create_relocations(obj: &mut ObjInfo, modules: &ModuleMap<'_>, dol_obj: &ObjI
// Resolve all relocations in this module
for rel_reloc in take(&mut obj.unresolved_relocations) {
// Skip if already resolved
let (_, source_section) = obj
.sections
.get_elf_index(rel_reloc.section as usize)
.ok_or_else(|| anyhow!("Failed to locate REL section {}", rel_reloc.section))?;
let (_, source_section) =
obj.sections.get_elf_index(rel_reloc.section as usize).ok_or_else(|| {
anyhow!(
"Failed to locate REL section {} in module ID {}: {:?}",
rel_reloc.section,
obj.module_id,
rel_reloc
)
})?;
if source_section.relocations.contains(rel_reloc.address) {
continue;
}
@ -575,10 +592,8 @@ fn create_relocations(obj: &mut ObjInfo, modules: &ModuleMap<'_>, dol_obj: &ObjI
Some(rel_reloc.module_id)
},
};
let (_, source_section) = obj
.sections
.get_elf_index_mut(rel_reloc.section as usize)
.ok_or_else(|| anyhow!("Failed to locate REL section {}", rel_reloc.section))?;
let (_, source_section) =
obj.sections.get_elf_index_mut(rel_reloc.section as usize).unwrap();
source_section.relocations.insert(rel_reloc.address, reloc)?;
}
@ -653,11 +668,10 @@ fn load_analyze_dol(config: &ProjectConfig) -> Result<AnalyzeResult> {
log::debug!("Loading {}", config.base.object.display());
let mut obj = {
let file = map_file(&config.base.object)?;
let data = decompress_if_needed(file.as_slice())?;
if let Some(hash_str) = &config.base.hash {
verify_hash(data.as_ref(), hash_str)?;
verify_hash(file.as_slice(), hash_str)?;
}
process_dol(data.as_ref(), config.base.name().as_ref())?
process_dol(file.as_slice(), config.base.name().as_ref())?
};
let mut dep = vec![config.base.object.clone()];
@ -688,10 +702,9 @@ fn load_analyze_dol(config: &ProjectConfig) -> Result<AnalyzeResult> {
if !config.quick_analysis {
let mut state = AnalyzerState::default();
debug!("Detecting function boundaries");
state.detect_functions(&obj)?;
FindTRKInterruptVectorTable::execute(&mut state, &obj)?;
FindSaveRestSleds::execute(&mut state, &obj)?;
state.detect_functions(&obj)?;
FindTRKInterruptVectorTable::execute(&mut state, &obj)?;
state.apply(&mut obj)?;
}
@ -701,11 +714,10 @@ fn load_analyze_dol(config: &ProjectConfig) -> Result<AnalyzeResult> {
if let Some(selfile) = &config.selfile {
log::info!("Loading {}", selfile.display());
let file = map_file(selfile)?;
let data = decompress_if_needed(file.as_slice())?;
if let Some(hash) = &config.selfile_hash {
verify_hash(data.as_ref(), hash)?;
verify_hash(file.as_slice(), hash)?;
}
apply_selfile(&mut obj, data.as_ref())?;
apply_selfile(&mut obj, file.as_slice())?;
dep.push(selfile.clone());
}
@ -761,11 +773,21 @@ fn split_write_obj(
debug!("Writing object files");
let obj_dir = out_dir.join("obj");
let entry = if obj.kind == ObjKind::Executable {
obj.entry.and_then(|e| {
let (section_index, _) = obj.sections.at_address(e as u32).ok()?;
let symbols = obj.symbols.at_section_address(section_index, e as u32).collect_vec();
best_match_for_reloc(symbols, ObjRelocKind::PpcRel24).map(|(_, s)| s.name.clone())
})
} else {
obj.symbols.by_name("_prolog")?.map(|(_, s)| s.name.clone())
};
let mut out_config = OutputModule {
name: module_config.name().to_string(),
module_id: obj.module_id,
ldscript: out_dir.join("ldscript.lcf"),
units: Vec::with_capacity(split_objs.len()),
entry,
};
for (unit, split_obj) in obj.link_order.iter().zip(&split_objs) {
let out_obj = write_elf(split_obj)?;
@ -802,13 +824,12 @@ fn split_write_obj(
fn load_analyze_rel(config: &ProjectConfig, module_config: &ModuleConfig) -> Result<AnalyzeResult> {
debug!("Loading {}", module_config.object.display());
let map = map_file(&module_config.object)?;
let buf = decompress_if_needed(map.as_slice())?;
let file = map_file(&module_config.object)?;
if let Some(hash_str) = &module_config.hash {
verify_hash(buf.as_ref(), hash_str)?;
verify_hash(file.as_slice(), hash_str)?;
}
let (_, mut module_obj) =
process_rel(&mut Reader::new(buf.as_ref()), module_config.name().as_ref())?;
process_rel(&mut Cursor::new(file.as_slice()), module_config.name().as_ref())?;
if let Some(comment_version) = config.mw_comment_version {
module_obj.mw_comment = Some(MWComment::new(comment_version)?);
@ -863,11 +884,10 @@ fn split(args: SplitArgs) -> Result<()> {
for module_config in config.modules.iter_mut() {
let file = map_file(&module_config.object)?;
let buf = decompress_if_needed(file.as_slice())?;
if let Some(hash_str) = &module_config.hash {
verify_hash(buf.as_ref(), hash_str)?;
verify_hash(file.as_slice(), hash_str)?;
} else {
module_config.hash = Some(file_sha1_string(&mut Reader::new(buf.as_ref()))?);
module_config.hash = Some(file_sha1_string(&mut file.as_reader())?);
}
}
@ -1080,21 +1100,26 @@ fn validate<P: AsRef<Path>>(obj: &ObjInfo, elf_file: P, state: &AnalyzerState) -
for (_symbol_idx, symbol) in real_obj.symbols.for_section(section_index) {
let symbol_addr = SectionAddress::new(section_index, symbol.address as u32);
real_functions.insert(symbol_addr, symbol.name.clone());
match state.function_bounds.get(&symbol_addr) {
Some(&Some(end)) => {
if symbol.size > 0 && end != (symbol_addr + symbol.size as u32) {
match state.functions.get(&symbol_addr) {
Some(info) => {
if let Some(end) = info.end {
if symbol.size > 0 && end != (symbol_addr + symbol.size as u32) {
log::warn!(
"Function {:#010X} ({}) ends at {:#010X}, expected {:#010X}",
symbol.address,
symbol.name,
end,
symbol.address + symbol.size
);
}
} else {
log::warn!(
"Function {:#010X} ({}) ends at {:#010X}, expected {:#010X}",
"Function {:#010X} ({}) has no end",
symbol.address,
symbol.name,
end,
symbol.address + symbol.size
symbol.name
);
}
}
Some(_) => {
log::warn!("Function {:#010X} ({}) has no end", symbol.address, symbol.name);
}
None => {
log::warn!(
"Function {:#010X} ({}) not discovered!",
@ -1105,8 +1130,8 @@ fn validate<P: AsRef<Path>>(obj: &ObjInfo, elf_file: P, state: &AnalyzerState) -
}
}
}
for (&start, &end) in &state.function_bounds {
let Some(end) = end else {
for (&start, info) in &state.functions {
let Some(end) = info.end else {
continue;
};
if !real_functions.contains_key(&start) {
@ -1206,11 +1231,10 @@ fn diff(args: DiffArgs) -> Result<()> {
log::info!("Loading {}", config.base.object.display());
let mut obj = {
let file = map_file(&config.base.object)?;
let data = decompress_if_needed(file.as_slice())?;
if let Some(hash_str) = &config.base.hash {
verify_hash(data.as_ref(), hash_str)?;
verify_hash(file.as_slice(), hash_str)?;
}
process_dol(data.as_ref(), config.base.name().as_ref())?
process_dol(file.as_slice(), config.base.name().as_ref())?
};
if let Some(symbols_path) = &config.base.symbols {
@ -1353,11 +1377,10 @@ fn apply(args: ApplyArgs) -> Result<()> {
log::info!("Loading {}", config.base.object.display());
let mut obj = {
let file = map_file(&config.base.object)?;
let data = decompress_if_needed(file.as_slice())?;
if let Some(hash_str) = &config.base.hash {
verify_hash(data.as_ref(), hash_str)?;
verify_hash(file.as_slice(), hash_str)?;
}
process_dol(data.as_ref(), config.base.name().as_ref())?
process_dol(file.as_slice(), config.base.name().as_ref())?
};
if let Some(symbols_path) = &config.base.symbols {

6
src/cmd/elf.rs
View File

@ -2,7 +2,7 @@ use std::{
collections::{btree_map, hash_map, BTreeMap, HashMap},
fs,
fs::DirBuilder,
io::Write,
io::{Cursor, Write},
path::PathBuf,
};
@ -22,7 +22,7 @@ use crate::{
comment::{read_comment_sym, MWComment},
config::{write_splits_file, write_symbols_file},
elf::{process_elf, write_elf},
file::{buf_writer, process_rsp, Reader},
file::{buf_writer, process_rsp},
signatures::{compare_signature, generate_signature, FunctionSignature},
split::split_obj,
IntoCow, ToCow,
@ -544,7 +544,7 @@ fn info(args: InfoArgs) -> Result<()> {
if let Some(comment_section) = in_file.section_by_name(".comment") {
let data = comment_section.uncompressed_data()?;
if !data.is_empty() {
let mut reader = Reader::new(&*data);
let mut reader = Cursor::new(&*data);
let header =
MWComment::parse_header(&mut reader).context("While reading .comment section")?;
println!("\nMetrowerks metadata (.comment):");

1
src/cmd/mod.rs
View File

@ -6,6 +6,7 @@ pub mod elf;
pub mod elf2dol;
pub mod map;
pub mod metroidbuildinfo;
pub mod nlzss;
pub mod rarc;
pub mod rel;
pub mod rso;

63
src/cmd/nlzss.rs Normal file
View File

@ -0,0 +1,63 @@
use std::{fs, path::PathBuf};
use anyhow::{anyhow, Context, Result};
use argp::FromArgs;
use crate::util::{
file::{open_file, process_rsp},
IntoCow, ToCow,
};
#[derive(FromArgs, PartialEq, Debug)]
/// Commands for processing NLZSS-compressed files.
#[argp(subcommand, name = "nlzss")]
pub struct Args {
#[argp(subcommand)]
command: SubCommand,
}
#[derive(FromArgs, PartialEq, Debug)]
#[argp(subcommand)]
enum SubCommand {
Decompress(DecompressArgs),
}
#[derive(FromArgs, PartialEq, Eq, Debug)]
/// Decompresses NLZSS-compressed files.
#[argp(subcommand, name = "decompress")]
pub struct DecompressArgs {
#[argp(positional)]
/// NLZSS-compressed file(s)
files: Vec<PathBuf>,
#[argp(option, short = 'o')]
/// Output file (or directory, if multiple files are specified).
/// If not specified, decompresses in-place.
output: Option<PathBuf>,
}
pub fn run(args: Args) -> Result<()> {
match args.command {
SubCommand::Decompress(args) => decompress(args),
}
}
fn decompress(args: DecompressArgs) -> Result<()> {
let files = process_rsp(&args.files)?;
let single_file = files.len() == 1;
for path in files {
let data = nintendo_lz::decompress(&mut open_file(&path)?)
.map_err(|e| anyhow!("Failed to decompress '{}' with NLZSS: {}", path.display(), e))?;
let out_path = if let Some(output) = &args.output {
if single_file {
output.as_path().to_cow()
} else {
output.join(path.file_name().unwrap()).into_cow()
}
} else {
path.as_path().to_cow()
};
fs::write(out_path.as_ref(), data)
.with_context(|| format!("Failed to write '{}'", out_path.display()))?;
}
Ok(())
}

238
src/cmd/rel.rs
View File

@ -1,8 +1,7 @@
use std::{
collections::{btree_map, BTreeMap},
ffi::OsStr,
fs,
io::Write,
io::{ Write},
path::PathBuf,
time::Instant,
};
@ -10,7 +9,8 @@ use std::{
use anyhow::{anyhow, bail, ensure, Context, Result};
use argp::FromArgs;
use object::{
Architecture, Endianness, Object, ObjectSection, ObjectSymbol, RelocationTarget, SymbolIndex,
Architecture, Endianness, File, Object, ObjectSection, ObjectSymbol, RelocationTarget,
SectionIndex, SymbolIndex,
};
use rayon::prelude::*;
use rustc_hash::FxHashMap;
@ -27,18 +27,18 @@ use crate::{
tracker::Tracker,
},
array_ref_mut,
cmd::dol::ProjectConfig,
cmd::dol::{ModuleConfig, ProjectConfig},
obj::{ObjInfo, ObjReloc, ObjRelocKind, ObjSection, ObjSectionKind, ObjSymbol},
util::{
config::is_auto_symbol,
dol::process_dol,
elf::{to_obj_reloc_kind, write_elf},
file::{
buf_reader, buf_writer, decompress_if_needed, map_file, process_rsp, verify_hash,
FileIterator, Reader,
},
file::{buf_reader, buf_writer, map_file, process_rsp, verify_hash, FileIterator},
nested::NestedMap,
rel::{process_rel, process_rel_header, write_rel, RelHeader, RelReloc, RelWriteInfo},
rel::{
process_rel, process_rel_header, process_rel_sections, write_rel, RelHeader, RelReloc,
RelSectionHeader, RelWriteInfo, PERMITTED_SECTIONS,
},
IntoCow, ToCow,
},
};
@ -106,8 +106,8 @@ pub fn run(args: Args) -> Result<()> {
}
}
fn load_obj(buf: &[u8]) -> Result<object::File> {
let obj = object::read::File::parse(buf)?;
fn load_obj(buf: &[u8]) -> Result<File> {
let obj = File::parse(buf)?;
match obj.architecture() {
Architecture::PowerPc => {}
arch => bail!("Unexpected architecture: {arch:?}"),
@ -116,21 +116,134 @@ fn load_obj(buf: &[u8]) -> Result<object::File> {
Ok(obj)
}
/// Attempt to match the section index from the ELF to the original REL.
/// Our built ELFs may be missing sections that were present in the original RELs.
fn match_section_index(
obj: &File,
section_index: SectionIndex,
rel_sections: &[RelSectionHeader],
) -> Result<usize> {
let (_, _) = (obj, rel_sections);
Ok(section_index.0)
// TODO
// rel_sections
// .iter()
// .enumerate()
// .filter(|(_, s)| s.size() > 0)
// .zip(obj.sections().filter(|s| s.size() > 0))
// .find_map(
// |((rel_section_index, _), obj_section)| {
// if obj_section.index() == section_index {
// Some(rel_section_index)
// } else {
// None
// }
// },
// )
// .ok_or_else(|| {
// anyhow!(
// "Failed to find matching section index for {} ({}), REL section count: {}",
// obj.section_by_index(section_index)
// .ok()
// .and_then(|s| s.name().ok().map(|s| s.to_string()))
// .unwrap_or("[invalid]".to_string()),
// section_index.0,
// rel_sections.len()
// )
// })
}
fn load_rel(module_config: &ModuleConfig) -> Result<(RelHeader, Vec<RelSectionHeader>)> {
let file = map_file(&module_config.object)?;
if let Some(hash_str) = &module_config.hash {
verify_hash(file.as_slice(), hash_str)?;
}
let mut reader = file.as_reader();
let header = process_rel_header(&mut reader)?;
let sections = process_rel_sections(&mut reader, &header)?;
Ok((header, sections))
}
fn resolve_relocations(
module: &File,
existing_headers: &BTreeMap<u32, (RelHeader, Vec<RelSectionHeader>)>,
module_id: usize,
symbol_map: &FxHashMap<&[u8], (usize, SymbolIndex)>,
modules: &[(File, PathBuf)],
relocations: &mut Vec<RelReloc>,
) -> Result<usize> {
let mut resolved = 0usize;
for section in module.sections() {
if !matches!(section.name(), Ok(name) if PERMITTED_SECTIONS.contains(&name)) {
continue;
}
let section_index = if let Some((_, sections)) = existing_headers.get(&(module_id as u32)) {
match_section_index(module, section.index(), sections)?
} else {
section.index().0
} as u8;
for (address, reloc) in section.relocations() {
let reloc_target = match reloc.target() {
RelocationTarget::Symbol(idx) => {
module.symbol_by_index(idx).with_context(|| {
format!("Relocation against invalid symbol index {}", idx.0)
})?
}
reloc_target => bail!("Unsupported relocation target: {reloc_target:?}"),
};
let (target_module_id, target_symbol) = if reloc_target.is_undefined() {
resolved += 1;
symbol_map
.get(reloc_target.name_bytes()?)
.map(|&(module_id, symbol_idx)| {
(module_id, modules[module_id].0.symbol_by_index(symbol_idx).unwrap())
})
.ok_or_else(|| {
anyhow!(
"Failed to find symbol {} in any module",
reloc_target.name().unwrap_or("[invalid]")
)
})?
} else {
(module_id, reloc_target)
};
let target_section_index = target_symbol.section_index().unwrap();
let target_section = if let Some((_, sections)) =
existing_headers.get(&(target_module_id as u32))
{
match_section_index(&modules[target_module_id].0, target_section_index, sections)?
} else {
target_section_index.0
} as u8;
relocations.push(RelReloc {
kind: to_obj_reloc_kind(reloc.kind())?,
section: section_index,
address: address as u32,
module_id: target_module_id as u32,
target_section,
addend: (target_symbol.address() as i64 + reloc.addend()) as u32,
// Extra
original_section: section.index().0 as u8,
original_target_section: target_section_index.0 as u8,
});
}
}
Ok(resolved)
}
fn make(args: MakeArgs) -> Result<()> {
let total = Instant::now();
// Load existing REL headers (if specified)
let mut existing_headers = BTreeMap::<u32, RelHeader>::new();
let mut existing_headers = BTreeMap::<u32, (RelHeader, Vec<RelSectionHeader>)>::new();
if let Some(config_path) = &args.config {
let config: ProjectConfig = serde_yaml::from_reader(&mut buf_reader(config_path)?)?;
for module_config in &config.modules {
let map = map_file(&module_config.object)?;
let buf = decompress_if_needed(map.as_slice())?;
if let Some(hash_str) = &module_config.hash {
verify_hash(buf.as_ref(), hash_str)?;
}
let header = process_rel_header(&mut Reader::new(buf.as_ref()))?;
existing_headers.insert(header.module_id, header);
let _span = info_span!("module", name = %module_config.name()).entered();
let (header, sections) = load_rel(module_config).with_context(|| {
format!("While loading REL '{}'", module_config.object.display())
})?;
existing_headers.insert(header.module_id, (header, sections));
}
}
@ -144,6 +257,7 @@ fn make(args: MakeArgs) -> Result<()> {
.zip(&paths)
.map(|(file, path)| {
load_obj(file.as_slice())
.map(|o| (o, path.clone()))
.with_context(|| format!("Failed to load '{}'", path.display()))
})
.collect::<Result<Vec<_>>>()?;
@ -151,7 +265,8 @@ fn make(args: MakeArgs) -> Result<()> {
// Create symbol map
let start = Instant::now();
let mut symbol_map = FxHashMap::<&[u8], (usize, SymbolIndex)>::default();
for (module_id, module) in modules.iter().enumerate() {
for (module_id, (module, path)) in modules.iter().enumerate() {
let _span = info_span!("file", path = %path.display()).entered();
for symbol in module.symbols() {
if symbol.is_definition() && symbol.scope() == object::SymbolScope::Dynamic {
symbol_map.entry(symbol.name_bytes()?).or_insert((module_id, symbol.index()));
@ -163,45 +278,19 @@ fn make(args: MakeArgs) -> Result<()> {
let mut resolved = 0usize;
let mut relocations = Vec::<Vec<RelReloc>>::with_capacity(modules.len() - 1);
relocations.resize_with(modules.len() - 1, Vec::new);
for ((module_id, module), relocations) in
for ((module_id, (module, path)), relocations) in
modules.iter().enumerate().skip(1).zip(&mut relocations)
{
for section in module.sections() {
for (address, reloc) in section.relocations() {
let reloc_target = match reloc.target() {
RelocationTarget::Symbol(idx) => {
module.symbol_by_index(idx).with_context(|| {
format!("Relocation against invalid symbol index {}", idx.0)
})?
}
reloc_target => bail!("Unsupported relocation target: {reloc_target:?}"),
};
let (target_module_id, target_symbol) = if reloc_target.is_undefined() {
resolved += 1;
symbol_map
.get(reloc_target.name_bytes()?)
.map(|&(module_id, symbol_idx)| {
(module_id, modules[module_id].symbol_by_index(symbol_idx).unwrap())
})
.ok_or_else(|| {
anyhow!(
"Failed to find symbol {} in any module",
reloc_target.name().unwrap_or("[invalid]")
)
})?
} else {
(module_id, reloc_target)
};
relocations.push(RelReloc {
kind: to_obj_reloc_kind(reloc.kind())?,
section: section.index().0 as u8,
address: address as u32,
module_id: target_module_id as u32,
target_section: target_symbol.section_index().unwrap().0 as u8,
addend: (target_symbol.address() as i64 + reloc.addend()) as u32,
});
}
}
let _span = info_span!("file", path = %path.display()).entered();
resolved += resolve_relocations(
module,
&existing_headers,
module_id,
&symbol_map,
&modules,
relocations,
)
.with_context(|| format!("While resolving relocations in '{}'", path.display()))?;
}
let duration = start.elapsed();
@ -214,12 +303,10 @@ fn make(args: MakeArgs) -> Result<()> {
// Write RELs
let start = Instant::now();
for (((module_id, module), path), relocations) in
modules.iter().enumerate().zip(&paths).skip(1).zip(relocations)
for ((module_id, (module, path)), relocations) in
modules.iter().enumerate().skip(1).zip(relocations)
{
let name =
path.file_stem().unwrap_or(OsStr::new("[unknown]")).to_str().unwrap_or("[invalid]");
let _span = info_span!("module", name = %name).entered();
let _span = info_span!("file", path = %path.display()).entered();
let mut info = RelWriteInfo {
module_id: module_id as u32,
version: 3,
@ -230,13 +317,13 @@ fn make(args: MakeArgs) -> Result<()> {
section_count: None,
quiet: args.no_warn,
};
if let Some(existing_module) = existing_headers.get(&(module_id as u32)) {
info.version = existing_module.version;
info.name_offset = Some(existing_module.name_offset);
info.name_size = Some(existing_module.name_size);
info.align = existing_module.align;
info.bss_align = existing_module.bss_align;
info.section_count = Some(existing_module.num_sections as usize);
if let Some((header, _)) = existing_headers.get(&(module_id as u32)) {
info.version = header.version;
info.name_offset = Some(header.name_offset);
info.name_size = Some(header.name_size);
info.align = header.align;
info.bss_align = header.bss_align;
info.section_count = Some(header.num_sections as usize);
}
let rel_path = path.with_extension("rel");
let mut w = buf_writer(&rel_path)?;
@ -254,8 +341,7 @@ fn make(args: MakeArgs) -> Result<()> {
fn info(args: InfoArgs) -> Result<()> {
let file = map_file(args.rel_file)?;
let buf = decompress_if_needed(file.as_slice())?;
let (header, mut module_obj) = process_rel(&mut Reader::new(buf.as_ref()), "")?;
let (header, mut module_obj) = process_rel(&mut file.as_reader(), "")?;
let mut state = AnalyzerState::default();
state.detect_functions(&module_obj)?;
@ -319,9 +405,8 @@ fn merge(args: MergeArgs) -> Result<()> {
log::info!("Loading {}", args.dol_file.display());
let mut obj = {
let file = map_file(&args.dol_file)?;
let buf = decompress_if_needed(file.as_slice())?;
let name = args.dol_file.file_stem().map(|s| s.to_string_lossy()).unwrap_or_default();
process_dol(buf.as_ref(), name.as_ref())?
process_dol(file.as_slice(), name.as_ref())?
};
log::info!("Performing signature analysis");
@ -434,11 +519,14 @@ fn merge(args: MergeArgs) -> Result<()> {
log::info!("Detecting function boundaries");
let mut state = AnalyzerState::default();
FindSaveRestSleds::execute(&mut state, &obj)?;
state.detect_functions(&obj)?;
log::info!("Discovered {} functions", state.function_slices.len());
log::info!(
"Discovered {} functions",
state.functions.iter().filter(|(_, i)| i.is_function()).count()
);
FindTRKInterruptVectorTable::execute(&mut state, &obj)?;
FindSaveRestSleds::execute(&mut state, &obj)?;
state.apply(&mut obj)?;
apply_signatures_post(&mut obj)?;

15
src/cmd/rso.rs
View File

@ -1,12 +1,9 @@
use std::path::PathBuf;
use anyhow::{Context, Result};
use anyhow::Result;
use argp::FromArgs;
use crate::util::{
file::{decompress_if_needed, map_file, Reader},
rso::process_rso,
};
use crate::util::{file::map_file, rso::process_rso};
#[derive(FromArgs, PartialEq, Debug)]
/// Commands for processing RSO files.
@ -39,10 +36,10 @@ pub fn run(args: Args) -> Result<()> {
fn info(args: InfoArgs) -> Result<()> {
let rso = {
let file = map_file(&args.rso_file)?;
let data = decompress_if_needed(file.as_slice())
.with_context(|| format!("Failed to decompress '{}'", args.rso_file.display()))?;
process_rso(&mut Reader::new(data.as_ref()))?
let file = map_file(args.rso_file)?;
let obj = process_rso(&mut file.as_reader())?;
#[allow(clippy::let_and_return)]
obj
};
println!("Read RSO module {}", rso.name);
Ok(())

13
src/cmd/yaz0.rs
View File

@ -30,7 +30,8 @@ pub struct DecompressArgs {
/// YAZ0-compressed files
files: Vec<PathBuf>,
#[argp(option, short = 'o')]
/// Output directory. If not specified, decompresses in-place.
/// Output file (or directory, if multiple files are specified).
/// If not specified, decompresses in-place.
output: Option<PathBuf>,
}
@ -41,10 +42,16 @@ pub fn run(args: Args) -> Result<()> {
}
fn decompress(args: DecompressArgs) -> Result<()> {
for path in process_rsp(&args.files)? {
let files = process_rsp(&args.files)?;
let single_file = files.len() == 1;
for path in files {
let data = decompress_reader(&mut open_file(&path)?)?;
let out_path = if let Some(output) = &args.output {
output.join(path.file_name().unwrap()).into_cow()
if single_file {
output.as_path().to_cow()
} else {
output.join(path.file_name().unwrap()).into_cow()
}
} else {
path.as_path().to_cow()
};

2
src/main.rs
View File

@ -82,6 +82,7 @@ enum SubCommand {
Elf2Dol(cmd::elf2dol::Args),
// Map(cmd::map::Args),
MetroidBuildInfo(cmd::metroidbuildinfo::Args),
Nlzss(cmd::nlzss::Args),
Rarc(cmd::rarc::Args),
Rel(cmd::rel::Args),
Rso(cmd::rso::Args),
@ -129,6 +130,7 @@ fn main() {
SubCommand::Elf2Dol(c_args) => cmd::elf2dol::run(c_args),
// SubCommand::Map(c_args) => cmd::map::run(c_args),
SubCommand::MetroidBuildInfo(c_args) => cmd::metroidbuildinfo::run(c_args),
SubCommand::Nlzss(c_args) => cmd::nlzss::run(c_args),
SubCommand::Rarc(c_args) => cmd::rarc::run(c_args),
SubCommand::Rel(c_args) => cmd::rel::run(c_args),
SubCommand::Rso(c_args) => cmd::rso::run(c_args),

4
src/obj/mod.rs
View File

@ -69,8 +69,8 @@ pub struct ObjInfo {
pub link_order: Vec<ObjUnit>,
pub blocked_ranges: BTreeMap<SectionAddress, u32>, // start -> end
// From extab
pub known_functions: BTreeMap<SectionAddress, u32>,
// From .ctors, .dtors and extab
pub known_functions: BTreeMap<SectionAddress, Option<u32>>,
// REL
/// Module ID (0 for main)

38
src/obj/splits.rs
View File

@ -1,9 +1,12 @@
use std::{collections::BTreeMap, ops::RangeBounds};
use std::{cmp::max, collections::BTreeMap, ops::RangeBounds};
use anyhow::{anyhow, Result};
use itertools::Itertools;
use crate::util::nested::NestedVec;
use crate::{
obj::{ObjInfo, ObjSection},
util::{nested::NestedVec, split::default_section_align},
};
/// Marks a split point within a section.
#[derive(Debug, Clone, Eq, PartialEq)]
@ -21,6 +24,30 @@ pub struct ObjSplit {
pub rename: Option<String>,
}
impl ObjSplit {
pub fn alignment(
&self,
obj: &ObjInfo,
section_index: usize,
section: &ObjSection,
split_addr: u32,
) -> u32 {
self.align.unwrap_or_else(|| {
let default_align = default_section_align(section) as u32;
max(
// Maximum alignment of any symbol in this split
obj.symbols
.for_section_range(section_index, split_addr..self.end)
.filter(|&(_, s)| s.size_known && s.size > 0)
.filter_map(|(_, s)| s.align)
.max()
.unwrap_or(default_align),
default_align,
)
})
}
}
/// Splits within a section.
#[derive(Debug, Clone, Default)]
pub struct ObjSplits {
@ -46,6 +73,13 @@ impl ObjSplits {
}
}
pub fn at_mut(&mut self, address: u32) -> Option<&mut ObjSplit> {
match self.for_range_mut(..=address).next_back() {
Some((_, split)) if split.end == 0 || split.end > address => Some(split),
_ => None,
}
}
/// Locate existing splits within the given address range.
pub fn for_range<R>(&self, range: R) -> impl DoubleEndedIterator<Item = (u32, &ObjSplit)>
where R: RangeBounds<u32> {

17
src/obj/symbols.rs
View File

@ -112,6 +112,15 @@ impl ObjSymbolFlagSet {
self.0 &= !ObjSymbolFlags::ForceActive;
}
}
/// Special flags to keep when merging symbols.
#[inline]
pub fn keep_flags(&self) -> FlagSet<ObjSymbolFlags> {
self.0
& (ObjSymbolFlags::ForceActive
| ObjSymbolFlags::NoWrite
| ObjSymbolFlags::RelocationIgnore)
}
}
#[allow(clippy::derived_hash_with_manual_eq)]
@ -213,8 +222,7 @@ impl ObjSymbols {
let replace = replace || (is_auto_symbol(existing) && !is_auto_symbol(&in_symbol));
let size =
if existing.size_known && in_symbol.size_known && existing.size != in_symbol.size {
// TODO fix and promote back to warning
log::debug!(
log::warn!(
"Conflicting size for {}: was {:#X}, now {:#X}",
existing.name,
existing.size,
@ -248,7 +256,7 @@ impl ObjSymbols {
section: in_symbol.section,
size,
size_known: existing.size_known || in_symbol.size != 0,
flags: in_symbol.flags,
flags: ObjSymbolFlagSet(in_symbol.flags.0 | existing.flags.keep_flags()),
kind: in_symbol.kind,
align: in_symbol.align.or(existing.align),
data_kind: match in_symbol.data_kind {
@ -511,7 +519,8 @@ impl ObjSymbol {
ObjSymbolKind::Unknown => true,
ObjSymbolKind::Function => !matches!(reloc_kind, ObjRelocKind::PpcEmbSda21),
ObjSymbolKind::Object => {
!matches!(reloc_kind, ObjRelocKind::PpcRel14 | ObjRelocKind::PpcRel24)
// !matches!(reloc_kind, ObjRelocKind::PpcRel14 | ObjRelocKind::PpcRel24)
true // SADX has bugged relocations that jump from .text to .bss, how awful
}
ObjSymbolKind::Section => {
matches!(

2
src/util/dep.rs
View File

@ -32,7 +32,7 @@ impl DepFile {
pub fn write<W: Write>(&self, w: &mut W) -> std::io::Result<()> {
write!(w, "{}:", self.name.to_slash_lossy())?;
for dep in self.dependencies.iter().unique() {
write!(w, " \\\n {}", dep.to_slash_lossy())?;
write!(w, " \\\n {}", dep.to_slash_lossy().replace(' ', "\\ "))?;
}
Ok(())
}

39
src/util/dol.rs
View File

@ -1,4 +1,4 @@
use std::collections::BTreeMap;
use std::{collections::BTreeMap, io::Cursor};
use anyhow::{anyhow, bail, ensure, Result};
use dol::{Dol, DolSection, DolSectionType};
@ -9,7 +9,6 @@ use crate::{
ObjArchitecture, ObjInfo, ObjKind, ObjSection, ObjSectionKind, ObjSymbol, ObjSymbolFlagSet,
ObjSymbolFlags, ObjSymbolKind,
},
util::file::Reader,
};
const MAX_TEXT_SECTIONS: usize = 7;
@ -23,7 +22,7 @@ fn read_u32(dol: &Dol, addr: u32) -> Result<u32> {
}
pub fn process_dol(buf: &[u8], name: &str) -> Result<ObjInfo> {
let dol = Dol::read_from(Reader::new(buf))?;
let dol = Dol::read_from(Cursor::new(buf))?;
// Locate _rom_copy_info
let first_rom_section = dol
@ -410,7 +409,9 @@ pub fn process_dol(buf: &[u8], name: &str) -> Result<ObjInfo> {
)
})?;
let addr = SectionAddress::new(section_index, entry.function);
if let Some(old_value) = obj.known_functions.insert(addr, entry.function_size) {
if let Some(Some(old_value)) =
obj.known_functions.insert(addr, Some(entry.function_size))
{
if old_value != entry.function_size {
log::warn!(
"Conflicting sizes for {:#010X}: {:#X} != {:#X}",
@ -465,6 +466,36 @@ pub fn process_dol(buf: &[u8], name: &str) -> Result<ObjInfo> {
}
}
// Add .ctors and .dtors functions to known functions if they exist
for (_, section) in obj.sections.iter() {
if section.size & 3 != 0 {
continue;
}
let mut entries = vec![];
let mut current_addr = section.address as u32;
for chunk in section.data.chunks_exact(4) {
let addr = u32::from_be_bytes(chunk.try_into()?);
if addr == 0 || addr & 3 != 0 {
break;
}
let Ok((section_index, section)) = obj.sections.at_address(addr) else {
break;
};
if section.kind != ObjSectionKind::Code {
break;
}
entries.push(SectionAddress::new(section_index, addr));
current_addr += 4;
}
// .ctors and .dtors end with a null pointer
if current_addr != (section.address + section.size) as u32 - 4
|| section.data_range(current_addr, 0)?.iter().any(|&b| b != 0)
{
continue;
}
obj.known_functions.extend(entries.into_iter().map(|addr| (addr, None)));
}
// Locate _SDA2_BASE_ & _SDA_BASE_
match locate_sda_bases(&mut obj) {
Ok(true) => {

78
src/util/file.rs
View File

@ -23,9 +23,7 @@ pub struct MappedFile {
}
impl MappedFile {
pub fn new(mmap: Mmap, offset: u64, len: u64) -> Self { Self { mmap, offset, len } }
pub fn as_reader(&self) -> Reader { Reader::new(self.as_slice()) }
pub fn as_reader(&self) -> Cursor<&[u8]> { Cursor::new(self.as_slice()) }
pub fn as_slice(&self) -> &[u8] {
&self.mmap[self.offset as usize..self.offset as usize + self.len as usize]
@ -59,28 +57,47 @@ pub fn split_path<P: AsRef<Path>>(path: P) -> Result<(PathBuf, Option<PathBuf>)>
Ok((base_path, sub_path))
}
/// Opens a memory mapped file.
pub fn map_file<P: AsRef<Path>>(path: P) -> Result<MappedFile> {
let (base_path, sub_path) = split_path(path)?;
/// Opens a memory mapped file, and decompresses it if needed.
pub fn map_file<P: AsRef<Path>>(path: P) -> Result<FileEntry> {
let (base_path, sub_path) = split_path(path.as_ref())?;
let file = File::open(&base_path)
.with_context(|| format!("Failed to open file '{}'", base_path.display()))?;
let mmap = unsafe { MmapOptions::new().map(&file) }
.with_context(|| format!("Failed to mmap file: '{}'", base_path.display()))?;
let (offset, len) = if let Some(sub_path) = sub_path {
let rarc = rarc::RarcReader::new(&mut Reader::new(&*mmap))
.with_context(|| format!("Failed to read RARC '{}'", base_path.display()))?;
let mut reader = Cursor::new(&*mmap);
if sub_path.as_os_str() == OsStr::new("nlzss") {
return Ok(FileEntry::Buffer(nintendo_lz::decompress(&mut reader).map_err(|e| {
anyhow!("Failed to decompress '{}' with NLZSS: {}", path.as_ref().display(), e)
})?));
} else if sub_path.as_os_str() == OsStr::new("yaz0") {
return Ok(FileEntry::Buffer(yaz0::decompress_file(&mut reader).with_context(
|| format!("Failed to decompress '{}' with Yaz0", path.as_ref().display()),
)?));
}
let rarc = rarc::RarcReader::new(&mut reader)
.with_context(|| format!("Failed to open '{}' as RARC archive", base_path.display()))?;
rarc.find_file(&sub_path)?.map(|(o, s)| (o, s as u64)).ok_or_else(|| {
anyhow!("File '{}' not found in '{}'", sub_path.display(), base_path.display())
})?
} else {
(0, mmap.len() as u64)
};
Ok(MappedFile { mmap, offset, len })
let map = MappedFile { mmap, offset, len };
let buf = map.as_slice();
// Auto-detect compression if there's a magic number.
if buf.len() > 4 && buf[0..4] == *b"Yaz0" {
return Ok(FileEntry::Buffer(yaz0::decompress_file(&mut map.as_reader()).with_context(
|| format!("Failed to decompress '{}' with Yaz0", path.as_ref().display()),
)?));
}
Ok(FileEntry::MappedFile(map))
}
pub type OpenedFile = TakeSeek<File>;
/// Opens a file (not memory mapped).
/// Opens a file (not memory mapped). No decompression is performed.
pub fn open_file<P: AsRef<Path>>(path: P) -> Result<OpenedFile> {
let (base_path, sub_path) = split_path(path)?;
let mut file = File::open(&base_path)
@ -98,7 +115,10 @@ pub fn open_file<P: AsRef<Path>>(path: P) -> Result<OpenedFile> {
Ok(file.take_seek(size))
}
pub type Reader<'a> = Cursor<&'a [u8]>;
pub trait Reader: BufRead + Seek {}
impl Reader for Cursor<&[u8]> {}
// impl Reader for &mut OpenedFile {}
/// Creates a buffered reader around a file (not memory mapped).
pub fn buf_reader<P: AsRef<Path>>(path: P) -> Result<BufReader<File>> {
@ -178,7 +198,7 @@ struct RarcIterator {
impl RarcIterator {
pub fn new(file: Mmap, base_path: &Path) -> Result<Self> {
let reader = rarc::RarcReader::new(&mut Reader::new(&*file))?;
let reader = rarc::RarcReader::new(&mut Cursor::new(&*file))?;
let paths = Self::collect_paths(&reader, base_path);
Ok(Self { file, base_path: base_path.to_owned(), paths, index: 0 })
}
@ -233,10 +253,31 @@ pub enum FileEntry {
impl FileEntry {
/// Creates a reader for the file.
pub fn as_reader(&self) -> Reader {
pub fn as_reader(&self) -> Box<dyn Reader + '_> {
match self {
Self::MappedFile(file) => file.as_reader(),
Self::Buffer(slice) => Reader::new(slice),
Self::MappedFile(file) => Box::new(file.as_reader()),
Self::Buffer(slice) => Box::new(Cursor::new(slice.as_slice())),
}
}
pub fn as_slice(&self) -> &[u8] {
match self {
Self::MappedFile(file) => file.as_slice(),
Self::Buffer(slice) => slice.as_slice(),
}
}
pub fn len(&self) -> u64 {
match self {
Self::MappedFile(file) => file.len(),
Self::Buffer(slice) => slice.len() as u64,
}
}
pub fn is_empty(&self) -> bool {
match self {
Self::MappedFile(file) => file.is_empty(),
Self::Buffer(slice) => slice.is_empty(),
}
}
}
@ -279,7 +320,8 @@ impl FileIterator {
let path = self.paths[self.index].clone();
self.index += 1;
match map_file(&path) {
Ok(map) => self.handle_file(map, path),
Ok(FileEntry::MappedFile(map)) => self.handle_file(map, path),
Ok(FileEntry::Buffer(_)) => todo!(),
Err(err) => Some(Err(err)),
}
}
@ -303,7 +345,7 @@ impl FileIterator {
fn handle_yaz0(
&mut self,
mut reader: Reader,
mut reader: Cursor<&[u8]>,
path: PathBuf,
) -> Option<Result<(PathBuf, FileEntry)>> {
Some(match yaz0::decompress_file(&mut reader) {
@ -340,7 +382,7 @@ pub fn touch<P: AsRef<Path>>(path: P) -> std::io::Result<()> {
pub fn decompress_if_needed(buf: &[u8]) -> Result<Cow<[u8]>> {
Ok(if buf.len() > 4 && buf[0..4] == *b"Yaz0" {
yaz0::decompress_file(&mut Reader::new(buf))?.into_cow()
yaz0::decompress_file(&mut Cursor::new(buf))?.into_cow()
} else {
buf.to_cow()
})

4
src/util/lcf.rs
View File

@ -78,9 +78,6 @@ pub fn generate_ldscript(obj: &ObjInfo, force_active: &[String]) -> Result<Strin
}
pub fn generate_ldscript_partial(obj: &ObjInfo, force_active: &[String]) -> Result<String> {
let section_defs =
obj.sections.iter().map(|(_, s)| format!("{} :{{}}", s.name)).join("\n ");
let mut force_files = Vec::with_capacity(obj.link_order.len());
for unit in &obj.link_order {
let obj_path = obj_path_for_unit(&unit.name);
@ -96,7 +93,6 @@ pub fn generate_ldscript_partial(obj: &ObjInfo, force_active: &[String]) -> Resu
}
let out = include_str!("../../assets/ldscript_partial.lcf")
.replace("$SECTIONS", &section_defs)
.replace("$FORCEACTIVE", &force_active.join("\n "));
Ok(out)
}

2
src/util/map.rs
View File

@ -564,7 +564,7 @@ impl StateMachine {
}
}
pub fn process_map<R: BufRead>(reader: &mut R) -> Result<MapInfo> {
pub fn process_map<R: BufRead + ?Sized>(reader: &mut R) -> Result<MapInfo> {
let mut sm = StateMachine {
state: ProcessMapState::None,
result: Default::default(),

112
src/util/rel.rs
View File

@ -110,7 +110,7 @@ struct RelImport {
#[binrw]
#[derive(Copy, Clone, Debug)]
struct RelSectionHeader {
pub struct RelSectionHeader {
offset_and_flags: u32,
size: u32,
}
@ -120,11 +120,11 @@ impl RelSectionHeader {
Self { offset_and_flags: offset | (exec as u32), size }
}
fn offset(&self) -> u32 { self.offset_and_flags & !1 }
pub fn offset(&self) -> u32 { self.offset_and_flags & !1 }
fn size(&self) -> u32 { self.size }
pub fn size(&self) -> u32 { self.size }
fn exec(&self) -> bool { self.offset_and_flags & 1 != 0 }
pub fn exec(&self) -> bool { self.offset_and_flags & 1 != 0 }
}
#[binrw]
@ -140,15 +140,26 @@ pub fn process_rel_header<R: Read + Seek>(reader: &mut R) -> Result<RelHeader> {
RelHeader::read_be(reader).context("Failed to read REL header")
}
pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHeader, ObjInfo)> {
let header = process_rel_header(reader)?;
pub fn process_rel_sections<R: Read + Seek>(
reader: &mut R,
header: &RelHeader,
) -> Result<Vec<RelSectionHeader>> {
let mut sections = Vec::with_capacity(header.num_sections as usize);
reader.seek(SeekFrom::Start(header.section_info_offset as u64))?;
let mut found_text = false;
let mut total_bss_size = 0;
for idx in 0..header.num_sections {
let section = RelSectionHeader::read_be(reader)
.with_context(|| format!("Failed to read REL section header {}", idx))?;
sections.push(section);
}
Ok(sections)
}
pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHeader, ObjInfo)> {
let header = process_rel_header(reader)?;
let mut sections = Vec::with_capacity(header.num_sections as usize);
let mut text_section = None;
let mut total_bss_size = 0;
for (idx, section) in process_rel_sections(reader, &header)?.iter().enumerate() {
let offset = section.offset();
let size = section.size();
if size == 0 {
@ -173,8 +184,8 @@ pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHea
total_bss_size = size;
(".bss".to_string(), ObjSectionKind::Bss, true)
} else if section.exec() {
ensure!(!found_text, "Multiple text sections in REL");
found_text = true;
ensure!(text_section.is_none(), "Multiple text sections in REL");
text_section = Some(idx as u8);
(".text".to_string(), ObjSectionKind::Code, true)
} else {
(format!(".section{}", idx), ObjSectionKind::Data, false)
@ -190,7 +201,7 @@ pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHea
_ => header.align,
}
.unwrap_or_default() as u64,
elf_index: idx as usize,
elf_index: idx,
relocations: Default::default(),
original_address: 0,
file_offset: offset as u64,
@ -206,32 +217,37 @@ pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHea
);
let mut symbols = Vec::new();
let mut add_symbol = |rel_section_idx: u8, offset: u32, name: &str| -> Result<()> {
if rel_section_idx > 0 {
let (section_index, _) = sections
.iter()
.enumerate()
.find(|&(_, section)| section.elf_index == rel_section_idx as usize)
.ok_or_else(|| anyhow!("Failed to locate {name} section {rel_section_idx}"))?;
log::debug!("Adding {name} section {rel_section_idx} offset {offset:#X}");
symbols.push(ObjSymbol {
name: name.to_string(),
demangled_name: None,
address: offset as u64,
section: Some(section_index),
size: 0,
size_known: false,
flags: ObjSymbolFlagSet(ObjSymbolFlags::Global.into()),
kind: ObjSymbolKind::Function,
align: None,
data_kind: Default::default(),
});
}
Ok(())
};
add_symbol(header.prolog_section, header.prolog_offset, "_prolog")?;
add_symbol(header.epilog_section, header.epilog_offset, "_epilog")?;
add_symbol(header.unresolved_section, header.unresolved_offset, "_unresolved")?;
let mut add_symbol =
|rel_section_idx: u8, offset: u32, name: &str, force_active: bool| -> Result<()> {
if rel_section_idx > 0 {
let (section_index, _) = sections
.iter()
.enumerate()
.find(|&(_, section)| section.elf_index == rel_section_idx as usize)
.ok_or_else(|| anyhow!("Failed to locate {name} section {rel_section_idx}"))?;
log::debug!("Adding {name} section {rel_section_idx} offset {offset:#X}");
let mut flags = ObjSymbolFlagSet(ObjSymbolFlags::Global.into());
if force_active {
flags.set_force_active(true);
}
symbols.push(ObjSymbol {
name: name.to_string(),
demangled_name: None,
address: offset as u64,
section: Some(section_index),
size: 0,
size_known: false,
flags,
kind: ObjSymbolKind::Function,
align: None,
data_kind: Default::default(),
});
}
Ok(())
};
add_symbol(header.prolog_section, header.prolog_offset, "_prolog", true)?;
add_symbol(header.epilog_section, header.epilog_offset, "_epilog", true)?;
add_symbol(header.unresolved_section, header.unresolved_offset, "_unresolved", true)?;
let mut unresolved_relocations = Vec::new();
let mut imp_idx = 0;
@ -303,6 +319,8 @@ pub fn process_rel<R: Read + Seek>(reader: &mut R, name: &str) -> Result<(RelHea
module_id: import.module_id,
target_section: reloc.section,
addend: reloc.addend,
original_section: section,
original_target_section: reloc.section,
};
unresolved_relocations.push(reloc);
}
@ -338,6 +356,10 @@ pub struct RelReloc {
/// Target addend within section.
/// If target module ID is 0 (DOL), this is an absolute address.
pub addend: u32,
// EXTRA for matching
pub original_section: u8,
pub original_target_section: u8,
}
#[inline]
@ -356,13 +378,15 @@ fn apply_relocation(
data: &mut [u8],
module_id: u32,
rel_reloc: &RelReloc,
unresolved: u32,
header: &RelHeader,
) -> Result<()> {
let diff = if rel_reloc.module_id == module_id && rel_reloc.section == rel_reloc.target_section
{
rel_reloc.addend as i32 - rel_reloc.address as i32
} else if header.unresolved_section == rel_reloc.section {
header.unresolved_offset as i32 - rel_reloc.address as i32
} else {
unresolved as i32 - rel_reloc.address as i32
return Ok(());
};
let ins_ref = array_ref_mut!(data, rel_reloc.address as usize, 4);
let mut ins = u32::from_be_bytes(*ins_ref);
@ -404,7 +428,7 @@ pub struct RelWriteInfo {
pub quiet: bool,
}
const PERMITTED_SECTIONS: [&str; 7] =
pub const PERMITTED_SECTIONS: [&str; 7] =
[".init", ".text", ".ctors", ".dtors", ".rodata", ".data", ".bss"];
pub fn should_write_section(section: &object::Section) -> bool {
@ -445,7 +469,7 @@ pub fn write_rel<W: Write>(
let mut apply_relocations = vec![];
relocations.retain(|r| {
if !should_write_section(
&file.section_by_index(object::SectionIndex(r.section as usize)).unwrap(),
&file.section_by_index(object::SectionIndex(r.original_section as usize)).unwrap(),
) {
return false;
}
@ -660,10 +684,10 @@ pub fn write_rel<W: Write>(
let section_index = section.index().0 as u8;
let mut section_data = section.uncompressed_data()?;
if apply_relocations.iter().any(|r| r.section == section_index) {
if apply_relocations.iter().any(|r| r.original_section == section_index) {
let mut data = section_data.into_owned();
for reloc in apply_relocations.iter().filter(|r| r.section == section_index) {
apply_relocation(&mut data, info.module_id, reloc, header.unresolved_offset)?;
for reloc in apply_relocations.iter().filter(|r| r.original_section == section_index) {
apply_relocation(&mut data, info.module_id, reloc, &header)?;
}
section_data = data.into_cow();
}

23
src/util/signatures.rs
View File

@ -13,6 +13,7 @@ use crate::{
analysis::{
cfa::SectionAddress,
tracker::{Relocation, Tracker},
RelocationTarget,
},
array_ref,
obj::{
@ -171,16 +172,18 @@ pub fn apply_signature(
None => continue,
};
let target = match (reloc, sig_reloc.kind) {
(&Relocation::Absolute(addr), ObjRelocKind::Absolute)
| (&Relocation::Hi(addr), ObjRelocKind::PpcAddr16Hi)
| (&Relocation::Ha(addr), ObjRelocKind::PpcAddr16Ha)
| (&Relocation::Lo(addr), ObjRelocKind::PpcAddr16Lo)
| (&Relocation::Rel24(addr), ObjRelocKind::PpcRel24)
| (&Relocation::Rel14(addr), ObjRelocKind::PpcRel14)
| (&Relocation::Sda21(addr), ObjRelocKind::PpcEmbSda21) => SectionAddress::new(
addr.section,
(addr.address as i64 - sig_reloc.addend as i64) as u32,
),
(&Relocation::Absolute(RelocationTarget::Address(addr)), ObjRelocKind::Absolute)
| (&Relocation::Hi(RelocationTarget::Address(addr)), ObjRelocKind::PpcAddr16Hi)
| (&Relocation::Ha(RelocationTarget::Address(addr)), ObjRelocKind::PpcAddr16Ha)
| (&Relocation::Lo(RelocationTarget::Address(addr)), ObjRelocKind::PpcAddr16Lo)
| (&Relocation::Rel24(RelocationTarget::Address(addr)), ObjRelocKind::PpcRel24)
| (&Relocation::Rel14(RelocationTarget::Address(addr)), ObjRelocKind::PpcRel14)
| (&Relocation::Sda21(RelocationTarget::Address(addr)), ObjRelocKind::PpcEmbSda21) => {
SectionAddress::new(
addr.section,
(addr.address as i64 - sig_reloc.addend as i64) as u32,
)
}
_ => bail!("Relocation mismatch: {:?} != {:?}", reloc, sig_reloc.kind),
};
let sig_symbol = &signature.symbols[sig_reloc.symbol];

188
src/util/split.rs
View File

@ -69,7 +69,7 @@ fn split_ctors_dtors(obj: &mut ObjInfo, start: SectionAddress, end: SectionAddre
.section
.and_then(|idx| obj.sections.get(idx).map(|s| s.name.clone()))
.unwrap_or_else(|| "unknown".to_string());
format!("{}_{}", function_symbol.name, section_name.trim_start_matches('.'))
format!("auto_{}_{}", function_symbol.name, section_name.trim_start_matches('.'))
});
log::debug!("Adding splits to unit {}", unit);
@ -250,7 +250,7 @@ fn split_extabindex(obj: &mut ObjInfo, start: SectionAddress) -> Result<()> {
.section
.and_then(|idx| obj.sections.get(idx).map(|s| s.name.clone()))
.unwrap_or_else(|| "unknown".to_string());
format!("{}_{}", function_symbol.name, section_name.trim_start_matches('.'))
format!("auto_{}_{}", function_symbol.name, section_name.trim_start_matches('.'))
});
log::debug!("Adding splits to unit {}", unit);
@ -333,12 +333,20 @@ fn create_gap_splits(obj: &mut ObjInfo) -> Result<()> {
break;
}
let (split_start, split_end) = match file_iter.peek() {
let (split_start, split_end, split_align) = match file_iter.peek() {
Some(&(addr, split)) => {
log::debug!("Found split {} ({:#010X}..{:#010X})", split.unit, addr, split.end);
(addr, SectionAddress::new(section_index, split.end))
(
addr,
SectionAddress::new(section_index, split.end),
split.alignment(obj, section_index, section, addr.address),
)
}
None => (section_end, SectionAddress::new(section_index, 0)),
None => (
section_end,
SectionAddress::new(section_index, 0),
default_section_align(section) as u32,
),
};
ensure!(
split_start >= current_address,
@ -347,7 +355,15 @@ fn create_gap_splits(obj: &mut ObjInfo) -> Result<()> {
split_end
);
if split_start > current_address {
let aligned_addr = current_address.align_up(split_align);
if split_start > aligned_addr {
log::debug!(
"Creating auto split: {} > {} (orig: {}, align {})",
split_start,
aligned_addr,
current_address,
split_align
);
// Find any duplicate symbols in this range
let mut new_split_end = split_start;
let symbols = obj
@ -355,7 +371,7 @@ fn create_gap_splits(obj: &mut ObjInfo) -> Result<()> {
.for_section_range(section_index, current_address.address..split_start.address)
.collect_vec();
let mut existing_symbols = HashSet::new();
for (_, symbol) in symbols {
for &(_, symbol) in &symbols {
if !existing_symbols.insert(symbol.name.clone()) {
log::debug!(
"Found duplicate symbol {} at {:#010X}",
@ -367,13 +383,22 @@ fn create_gap_splits(obj: &mut ObjInfo) -> Result<()> {
}
}
ensure!(
new_split_end > current_address,
"Duplicate symbols at {:#010X}: {:?}",
current_address,
symbols
.iter()
.filter(|(_, s)| s.address == current_address.address as u64)
.collect_vec(),
);
log::debug!(
"Creating split from {:#010X}..{:#010X}",
current_address,
new_split_end
);
let unit = format!(
"{:02}_{:08X}_{}",
"auto_{:02}_{:08X}_{}",
current_address.section,
current_address.address,
section.name.trim_start_matches('.')
@ -622,6 +647,85 @@ fn add_padding_symbols(obj: &mut ObjInfo) -> Result<()> {
#[inline]
const fn align_up(value: u32, align: u32) -> u32 { (value + (align - 1)) & !(align - 1) }
#[allow(dead_code)]
fn trim_split_alignment(obj: &mut ObjInfo) -> Result<()> {
// For each split, set the end of split to the end of the last symbol in the split.
let mut split_updates = vec![];
let mut iter = obj.sections.all_splits().peekable();
while let Some((section_index, section, addr, split)) = iter.next() {
let next_split = iter
.peek()
.filter(|&&(idx, _, _, _)| section_index == idx)
.map(|&(_, _, addr, split)| (addr, split));
let mut split_end = split.end;
if let Some((_, symbol)) = obj
.symbols
.for_section_range(section_index, addr..split.end)
.filter(|&(_, s)| s.size_known && s.size > 0)
.next_back()
{
split_end = symbol.address as u32 + symbol.size as u32;
}
split_end = align_up(split_end, split.alignment(obj, section_index, section, addr));
if split_end < split.end {
if let Some((next_addr, next_split)) = next_split {
let next_split_align = next_split.alignment(obj, section_index, section, addr);
if align_up(split_end, next_split_align) < next_addr {
log::warn!(
"Tried to trim {} split {} {:#010X}..{:#010X} to {:#010X}, but next split {} starts at {:#010X} with alignment {}",
section.name,
split.unit,
addr,
split.end,
split_end,
next_split.unit,
next_addr,
next_split_align
);
}
}
log::info!(
"Trimming {} split {} {:#010X}..{:#010X} to {:#010X}",
section.name,
split.unit,
addr,
split.end,
split_end
);
split_updates.push((section_index, addr, split_end));
}
}
drop(iter);
for (section_index, addr, split_end) in split_updates {
obj.sections[section_index].splits.at_mut(addr).unwrap().end = split_end;
}
Ok(())
}
/// Trim splits if they contain linker generated symbols.
fn trim_linker_generated_symbols(obj: &mut ObjInfo) -> Result<()> {
for section_index in 0..obj.sections.count() {
let section_end = end_for_section(obj, section_index)?;
let section = &mut obj.sections[section_index];
if section.address as u32 + section.size as u32 == section_end.address {
continue;
}
if let Some((addr, split)) = section.splits.iter_mut().next_back() {
if split.end > section_end.address {
log::debug!(
"Trimming split {} {:#010X}..{:#010X} to {:#010X}",
split.unit,
addr,
split.end,
section_end.address
);
split.end = section_end.address;
}
}
}
Ok(())
}
/// Perform any necessary adjustments to allow relinking.
/// This includes:
/// - Ensuring .ctors & .dtors entries are split with their associated function
@ -654,6 +758,9 @@ pub fn update_splits(obj: &mut ObjInfo, common_start: Option<u32>, fill_gaps: bo
split_ctors_dtors(obj, start, end)?;
}
// Remove linker generated symbols from splits
trim_linker_generated_symbols(obj)?;
// Create gap splits
create_gap_splits(obj)?;
@ -663,6 +770,10 @@ pub fn update_splits(obj: &mut ObjInfo, common_start: Option<u32>, fill_gaps: bo
// Ensure splits don't overlap symbols or each other
validate_splits(obj)?;
// Trim alignment from splits
// TODO figure out mwld pooled data alignment
// trim_split_alignment(obj)?;
if fill_gaps {
// Add symbols to beginning of any split that doesn't start with a symbol
add_padding_symbols(obj)?;
@ -793,7 +904,7 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
for (section_index, section) in obj.sections.iter() {
let mut current_address = SectionAddress::new(section_index, section.address as u32);
let section_end = end_for_section(obj, section_index)?;
let mut file_iter = section
let mut split_iter = section
.splits
.for_range(current_address.address..section_end.address)
.map(|(addr, split)| (SectionAddress::new(section_index, addr), split))
@ -804,30 +915,38 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
break;
}
let (file_addr, split) = match file_iter.next() {
let (split_addr, split) = match split_iter.next() {
Some((addr, split)) => (addr, split),
None => bail!("No file found"),
None => bail!("No split found"),
};
ensure!(
file_addr <= current_address,
"Gap in files: {} @ {:#010X}, {} @ {:#010X}",
split_addr == current_address,
"Split @ {} {} not found",
section.name,
section.address,
split.unit,
file_addr
current_address
);
let mut file_end = section_end;
if let Some(&(next_addr, _next_split)) = file_iter.peek() {
file_end = min(next_addr, section_end);
let split_end = SectionAddress::new(section_index, split.end);
let next_addr = split_iter.peek().map(|&(addr, _)| addr).unwrap_or(section_end);
if next_addr > split_end
&& section.data_range(split_end.address, next_addr.address)?.iter().any(|&b| b != 0)
{
bail!(
"Unsplit data in {} from {} {} to next split {}",
section.name,
split.unit,
split_end,
next_addr
);
}
// Skip over this data
if split.skip {
current_address = file_end;
current_address = next_addr;
continue;
}
let file = name_to_obj
let split_obj = name_to_obj
.get(&split.unit)
.and_then(|&idx| objects.get_mut(idx))
.ok_or_else(|| anyhow!("Unit '{}' not in link order", split.unit))?;
@ -842,7 +961,10 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
max(
// Maximum alignment of any symbol in this split
obj.symbols
.for_section_range(section_index, current_address.address..file_end.address)
.for_section_range(
section_index,
current_address.address..split_end.address,
)
.filter(|&(_, s)| s.size_known && s.size > 0)
.filter_map(|(_, s)| s.align)
.max()
@ -877,7 +999,7 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
// Collect relocations; target_symbol will be updated later
let out_relocations = section
.relocations
.range(current_address.address..file_end.address)
.range(current_address.address..split_end.address)
.map(|(addr, o)| {
(addr - current_address.address, ObjReloc {
kind: o.kind,
@ -889,10 +1011,10 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
.collect_vec();
// Add section symbols
let out_section_idx = file.sections.next_section_index();
let out_section_idx = split_obj.sections.next_section_index();
for (symbol_idx, symbol) in obj
.symbols
.for_section_range(section_index, current_address.address..=file_end.address)
.for_section_range(section_index, current_address.address..=split_end.address)
.filter(|&(_, s)| {
s.section == Some(section_index) && !is_linker_generated_label(&s.name)
})
@ -902,7 +1024,7 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
}
// TODO hack for gTRKInterruptVectorTableEnd
if (symbol.address == file_end.address as u64
if (symbol.address == split_end.address as u64
&& symbol.name != "gTRKInterruptVectorTableEnd")
|| (symbol.address == current_address.address as u64
&& symbol.name == "gTRKInterruptVectorTableEnd")
@ -910,7 +1032,7 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
continue;
}
symbol_idxs[symbol_idx] = Some(file.symbols.add_direct(ObjSymbol {
symbol_idxs[symbol_idx] = Some(split_obj.symbols.add_direct(ObjSymbol {
name: symbol.name.clone(),
demangled_name: symbol.demangled_name.clone(),
address: if split.common {
@ -934,22 +1056,22 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
// For mwldeppc 2.7 and above, a .comment section is required to link without error
// when common symbols are present. Automatically add one if needed.
if split.common && file.mw_comment.is_none() {
file.mw_comment = Some(MWComment::new(8)?);
if split.common && split_obj.mw_comment.is_none() {
split_obj.mw_comment = Some(MWComment::new(8)?);
}
if !split.common {
let data = match section.kind {
ObjSectionKind::Bss => vec![],
_ => section.data[(current_address.address as u64 - section.address) as usize
..(file_end.address as u64 - section.address) as usize]
..(split_end.address as u64 - section.address) as usize]
.to_vec(),
};
file.sections.push(ObjSection {
split_obj.sections.push(ObjSection {
name: split.rename.as_ref().unwrap_or(&section.name).clone(),
kind: section.kind,
address: 0,
size: file_end.address as u64 - current_address.address as u64,
size: split_end.address as u64 - current_address.address as u64,
data,
align,
elf_index: out_section_idx + 1,
@ -962,7 +1084,7 @@ pub fn split_obj(obj: &ObjInfo) -> Result<Vec<ObjInfo>> {
});
}
current_address = file_end;
current_address = next_addr;
}
}