Initial support for analyzing categories.

CMakeLists.txt

@@ -26,6 +26,7 @@ add_subdirectory(Vendor/BinaryNinjaAPI)
 # Core library -----------------------------------------------------------------
 
 set(CORE_SOURCE
+  Include/ObjectiveNinjaCore/Analyzers/CategoryAnalyzer.h
   Include/ObjectiveNinjaCore/Analyzers/CFStringAnalyzer.h
   Include/ObjectiveNinjaCore/Analyzers/ClassAnalyzer.h
   Include/ObjectiveNinjaCore/Analyzers/SelectorAnalyzer.h
@@ -36,6 +37,7 @@ set(CORE_SOURCE
   Include/ObjectiveNinjaCore/AnalysisProvider.h
   Include/ObjectiveNinjaCore/Analyzer.h
   Include/ObjectiveNinjaCore/TypeParser.h
+  Core/Analyzers/CategoryAnalyzer.cpp
   Core/Analyzers/CFStringAnalyzer.cpp
   Core/Analyzers/ClassAnalyzer.cpp
   Core/Analyzers/SelectorAnalyzer.cpp

Core/AnalysisProvider.cpp

@@ -9,6 +9,7 @@
 
 #include <ObjectiveNinjaCore/Analyzers/CFStringAnalyzer.h>
 #include <ObjectiveNinjaCore/Analyzers/ClassAnalyzer.h>
+#include <ObjectiveNinjaCore/Analyzers/CategoryAnalyzer.h>
 #include <ObjectiveNinjaCore/Analyzers/SelectorAnalyzer.h>
 
 namespace ObjectiveNinja {
@@ -20,6 +21,7 @@ SharedAnalysisInfo AnalysisProvider::infoForFile(SharedAbstractFile file)
     std::vector<std::unique_ptr<ObjectiveNinja::Analyzer>> analyzers;
     analyzers.emplace_back(new SelectorAnalyzer(info, file));
     analyzers.emplace_back(new ClassAnalyzer(info, file));
+    analyzers.emplace_back(new CategoryAnalyzer(info, file));
     analyzers.emplace_back(new CFStringAnalyzer(info, file));
 
     for (const auto& analyzer : analyzers)

Core/Analyzers/CategoryAnalyzer.cpp

@@ -0,0 +1,81 @@
+#include <ObjectiveNinjaCore/Analyzers/ClassAnalyzer.h>
+#include <ObjectiveNinjaCore/Analyzers/CategoryAnalyzer.h>
+
+#include <ObjectiveNinjaCore/TypeParser.h>
+
+using namespace ObjectiveNinja;
+
+
+CategoryAnalyzer::CategoryAnalyzer(SharedAnalysisInfo info,
+    SharedAbstractFile file)
+    : Analyzer(std::move(info), std::move(file))
+{
+}
+
+MethodListInfo CategoryAnalyzer::analyzeMethodList(uint64_t address)
+{
+    MethodListInfo mli;
+    mli.address = address;
+    mli.flags = m_file->readInt(mli.address);
+
+    auto methodCount = m_file->readInt(mli.address + 0x4);
+    auto methodSize = mli.hasRelativeOffsets() ? 12 : 24;
+
+    for (unsigned i = 0; i < methodCount; ++i) {
+        MethodInfo mi;
+        mi.address = mli.address + 8 + (i * methodSize);
+
+        m_file->seek(mi.address);
+
+        if (mli.hasRelativeOffsets()) {
+            mi.nameAddress = mi.address + static_cast<int32_t>(m_file->readInt());
+            mi.typeAddress = mi.address + 4 + static_cast<int32_t>(m_file->readInt());
+            mi.implAddress = mi.address + 8 + static_cast<int32_t>(m_file->readInt());
+        } else {
+            mi.nameAddress = arp(m_file->readLong());
+            mi.typeAddress = arp(m_file->readLong());
+            mi.implAddress = arp(m_file->readLong());
+        }
+
+        if (!mli.hasRelativeOffsets() || mli.hasDirectSelectors()) {
+            mi.selector = m_file->readStringAt(mi.nameAddress);
+        } else {
+            auto selectorNamePointer = arp(m_file->readLong(mi.nameAddress));
+            mi.selector = m_file->readStringAt(selectorNamePointer);
+        }
+
+        mi.type = m_file->readStringAt(mi.typeAddress);
+
+        m_info->methodImpls[mi.nameAddress] = mi.implAddress;
+
+        mli.methods.emplace_back(mi);
+    }
+
+    return mli;
+}
+
+void CategoryAnalyzer::run()
+{
+    const auto sectionStart = m_file->sectionStart("__objc_catlist");
+    const auto sectionEnd = m_file->sectionEnd("__objc_catlist");
+    if (sectionStart == 0 || sectionEnd == 0)
+        return;
+
+    for (auto address = sectionStart; address < sectionEnd; address += 8) {
+        CategoryInfo ci;
+        ci.listPointer = address;
+        ci.address = arp(m_file->readLong(address));
+        ci.nameAddress = arp(m_file->readLong(ci.address));
+        ci.name = m_file->readStringAt(ci.nameAddress);
+        ci.instanceMethodListAddress = arp(m_file->readLong(ci.address + 0x10));
+        ci.classMethodListAddress = arp(m_file->readLong(ci.address + 0x18));
+
+        if (ci.instanceMethodListAddress)
+            ci.instanceMethods = analyzeMethodList(ci.instanceMethodListAddress);
+
+        if (ci.classMethodListAddress)
+            ci.classMethods = analyzeMethodList(ci.classMethodListAddress);
+
+        m_info->categories.emplace_back(ci);
+    }
+}

Include/ObjectiveNinjaCore/AnalysisInfo.h

@@ -92,6 +92,21 @@ struct ClassInfo {
     uint64_t methodListAddress {};
 };
 
+/**
+ * A description of an Objective-C category.
+ */
+struct CategoryInfo {
+    uint64_t address {};
+    std::string name {};
+    MethodListInfo instanceMethods {};
+    MethodListInfo classMethods {};
+
+    uint64_t listPointer {};
+    uint64_t nameAddress {};
+    uint64_t instanceMethodListAddress {};
+    uint64_t classMethodListAddress {};
+};
+
 /**
  * Analysis info storage.
  *
@@ -106,6 +121,7 @@ struct AnalysisInfo {
     std::unordered_map<uint64_t, SharedSelectorRefInfo> selectorRefsByKey {};
 
     std::vector<ClassInfo> classes {};
+    std::vector<CategoryInfo> categories {};
     std::unordered_map<uint64_t, uint64_t> methodImpls;
 
     std::string dump() const;
@@ -123,4 +139,8 @@ NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(MethodListInfo, address, flags, methods)
 
 NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(ClassInfo, listPointer, address, dataAddress,
     nameAddress, name, methodListAddress, methodList)
+
+NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(CategoryInfo, listPointer, address, nameAddress,
+    name, instanceMethodListAddress, instanceMethods, classMethodListAddress,
+    classMethods)
 }

Include/ObjectiveNinjaCore/Analyzers/CategoryAnalyzer.h

@@ -0,0 +1,29 @@
+/*
+ * Copyright (c) 2022 Fabian Freyer. All rights reserved.
+ *
+ * Use of this source code is governed by the BSD 3-Clause license; the full
+ * terms of the license can be found in the LICENSE.txt file.
+ */
+
+#pragma once
+
+#include <ObjectiveNinjaCore/Analyzer.h>
+
+namespace ObjectiveNinja {
+
+/**
+ * Analyzer for extracting Objective-C class information.
+ */
+class CategoryAnalyzer : public Analyzer {
+    /**
+     * Analyze a method list.
+     */
+    MethodListInfo analyzeMethodList(uint64_t);
+
+public:
+    CategoryAnalyzer(SharedAnalysisInfo, SharedAbstractFile);
+
+    void run() override;
+};
+
+}

Plugin/CustomTypes.cpp

@@ -73,6 +73,15 @@ struct objc_class_t {
     void* vtable;
     const fptr_t data;
 };
+
+struct objc_category_t {
+    const char* name;
+    void* class;
+    const tptr_t instance_methods;
+    const tptr_t class_methods;
+    const tptr_t protocols;
+    const tptr_t instance_properties;
+}
 )";
 
 namespace CustomTypes {

Plugin/CustomTypes.h

@@ -28,6 +28,7 @@ const std::string Method = "objc_method_t";
 const std::string MethodListEntry = "objc_method_entry_t";
 const std::string Class = "objc_class_t";
 const std::string ClassRO = "objc_class_ro_t";
+const std::string Category = "objc_category_t";
 
 /**
  * Define all Objective-C-related types for a view.

Plugin/InfoHandler.cpp

@@ -125,6 +125,7 @@ void InfoHandler::applyInfoToView(SharedAnalysisInfo info, BinaryViewRef bv)
     BinaryReader reader(bv);
 
     auto taggedPointerType = namedType(bv, CustomTypes::TaggedPointer);
+    auto categoryType = namedType(bv, CustomTypes::Category);
     auto cfStringType = namedType(bv, CustomTypes::CFString);
     auto classType = namedType(bv, CustomTypes::Class);
     auto classDataType = namedType(bv, CustomTypes::ClassRO);
@@ -198,6 +199,62 @@ void InfoHandler::applyInfoToView(SharedAnalysisInfo info, BinaryViewRef bv)
         defineSymbol(bv, ci.methodListAddress, ci.name, "ml_");
     }
 
+    // Create data variables and symbols for the analyzed Categories.
+    for (const auto& ci : info->categories) {
+        defineVariable(bv, ci.listPointer, taggedPointerType);
+        defineVariable(bv, ci.address, categoryType);
+
+        defineSymbol(bv, ci.listPointer, ci.name, "catp_");
+        defineSymbol(bv, ci.address, ci.name, "cat_");
+
+        defineReference(bv, ci.listPointer, ci.address);
+
+        if (ci.instanceMethods.address && !ci.instanceMethods.methods.empty()) {
+            auto methodType = ci.instanceMethods.hasRelativeOffsets()
+                ? bv->GetTypeByName(CustomTypes::MethodListEntry)
+                : bv->GetTypeByName(CustomTypes::Method);
+
+            // Create data variables for each method in the method list.
+            for (const auto& mi : ci.instanceMethods.methods) {
+                defineVariable(bv, mi.address, methodType);
+                defineSymbol(bv, mi.address, sanitizeSelector(mi.selector), "mt_");
+                defineVariable(bv, mi.typeAddress, stringType(mi.type.size()));
+
+                defineReference(bv, ci.instanceMethods.address, mi.address);
+                defineReference(bv, mi.address, mi.nameAddress);
+                defineReference(bv, mi.address, mi.typeAddress);
+                defineReference(bv, mi.address, mi.implAddress);
+            }
+
+            // Create a data variable and symbol for the method list header.
+            defineVariable(bv, ci.instanceMethodListAddress, methodListType);
+            defineSymbol(bv, ci.instanceMethodListAddress, ci.name, "mli_");
+        }
+
+
+        if (ci.classMethods.address && !ci.classMethods.methods.empty()) {
+            auto methodType = ci.classMethods.hasRelativeOffsets()
+                ? bv->GetTypeByName(CustomTypes::MethodListEntry)
+                : bv->GetTypeByName(CustomTypes::Method);
+
+            // Create data variables for each method in the method list.
+            for (const auto& mi : ci.classMethods.methods) {
+                defineVariable(bv, mi.address, methodType);
+                defineSymbol(bv, mi.address, sanitizeSelector(mi.selector), "mt_");
+                defineVariable(bv, mi.typeAddress, stringType(mi.type.size()));
+
+                defineReference(bv, ci.classMethods.address, mi.address);
+                defineReference(bv, mi.address, mi.nameAddress);
+                defineReference(bv, mi.address, mi.typeAddress);
+                defineReference(bv, mi.address, mi.implAddress);
+            }
+
+            // Create a data variable and symbol for the method list header.
+            defineVariable(bv, ci.classMethodListAddress, methodListType);
+            defineSymbol(bv, ci.classMethodListAddress, ci.name, "mlc_");
+        }
+    }
+
     bv->CommitUndoActions();
     bv->UpdateAnalysis();
 }