From 26f52f65b756d9a5b0d65368ce6c2ef98c239da1 Mon Sep 17 00:00:00 2001
From: Amber Brault <celestialamber1@gmail.com>
Date: Thu, 3 Oct 2024 03:13:23 -0400
Subject: [PATCH] Automatically check for invalid extab relocations (#75)

* Begin work on extab reloc analysis code

* Refactoring

* Make format happy

* Address feedback + improvements
---
 Cargo.lock              | 12 ++++++++-
 Cargo.toml              |  1 +
 src/analysis/tracker.rs | 58 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 70 insertions(+), 1 deletion(-)

diff --git a/Cargo.lock b/Cargo.lock
index 2d1368d..ddd0584 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -393,6 +393,15 @@ dependencies = [
  "thiserror",
 ]
 
+[[package]]
+name = "cwextab"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e5aa7f13cc2fcb2bcfd3abc51bdbbf8f1fb729a69ed8c05ecbaa1a42197d1842"
+dependencies = [
+ "thiserror",
+]
+
 [[package]]
 name = "decomp-toolkit"
 version = "1.0.0"
@@ -404,6 +413,7 @@ dependencies = [
  "base64",
  "crossterm",
  "cwdemangle",
+ "cwextab 1.0.2",
  "enable-ansi-support",
  "filetime",
  "fixedbitset 0.5.7",
@@ -1018,7 +1028,7 @@ dependencies = [
  "anyhow",
  "byteorder",
  "cwdemangle",
- "cwextab",
+ "cwextab 0.3.1",
  "filetime",
  "flagset",
  "log",
diff --git a/Cargo.toml b/Cargo.toml
index 41262b6..73089c4 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -32,6 +32,7 @@ base16ct = "0.2"
 base64 = "0.22"
 crossterm = "0.28"
 cwdemangle = "1.0"
+cwextab = "1.0.2"
 enable-ansi-support = "0.2"
 filetime = "0.2"
 fixedbitset = "0.5"
diff --git a/src/analysis/tracker.rs b/src/analysis/tracker.rs
index af4cc5f..5e5fbee 100644
--- a/src/analysis/tracker.rs
+++ b/src/analysis/tracker.rs
@@ -4,6 +4,7 @@ use std::{
 };
 
 use anyhow::{bail, Result};
+use cwextab::decode_extab;
 use ppc750cl::Opcode;
 use tracing::{debug_span, info_span};
 use tracing_attributes::instrument;
@@ -120,6 +121,7 @@ impl Tracker {
                 self.process_data(obj, section_index, section)?;
             }
         }
+        self.check_extab_relocations(obj)?;
         self.reject_invalid_relocations(obj)?;
         Ok(())
     }
@@ -147,6 +149,62 @@ impl Tracker {
         Ok(())
     }
 
+    /// Check all of the extab relocations, and reject any invalid ones by checking against the decoded table data
+    /// of each table.
+    fn check_extab_relocations(&mut self, obj: &ObjInfo) -> Result<()> {
+        let mut to_reject = vec![];
+        let Some((section_index, section)) = obj.sections.by_name("extab")? else {
+            // No extab section found, return
+            return Ok(());
+        };
+        let mut decoded_reloc_addrs: BTreeSet<u32> = BTreeSet::new();
+
+        // Decode each exception table, and collect all of the relocations from the decoded data for each
+        for (_, symbol) in obj.symbols.for_section(section_index) {
+            let extab_name = &symbol.name;
+            let extab_start_addr: u32 = symbol.address as u32;
+            let extab_end_addr: u32 = extab_start_addr + symbol.size as u32;
+            let Ok(extab_data) = section.data_range(extab_start_addr, extab_end_addr) else {
+                log::warn!("Failed to get extab data for symbol {}", extab_name);
+                continue;
+            };
+            let data = match decode_extab(extab_data) {
+                Ok(decoded_data) => decoded_data,
+                Err(e) => {
+                    log::warn!(
+                        "Exception table decoding failed for symbol {}, reason: {}",
+                        extab_name,
+                        e
+                    );
+                    continue;
+                }
+            };
+
+            for reloc in data.relocations {
+                let reloc_addr = extab_start_addr + reloc.offset;
+                decoded_reloc_addrs.insert(reloc_addr);
+            }
+        }
+
+        let section_start_addr = SectionAddress::new(section_index, section.address as u32);
+        let section_end_addr = section_start_addr + (section.size as u32);
+
+        // Check all the extab relocations against the list of relocations from the decoded tables. Any
+        // relocations that aren't in the list are invalid, and are removed (if a table fails to decode,
+        // however, its relocations are all removed).
+        for (&address, _) in self.relocations.range(section_start_addr..section_end_addr) {
+            if !decoded_reloc_addrs.contains(&address.address) {
+                log::debug!("Rejecting invalid extab relocation @ {}", address);
+                to_reject.push(address);
+            }
+        }
+
+        for address in to_reject {
+            self.relocations.remove(&address);
+        }
+        Ok(())
+    }
+
     fn process_code(&mut self, obj: &ObjInfo) -> Result<()> {
         if let Some(entry) = obj.entry {
             let (section_index, _) = obj.sections.at_address(entry as u32)?;