Refactor TypeHierarchy Interface (#721)

* Split TH and VTable into two interfaces + refactor TH interface

* Add vtable unittest back

* Add breaking changes

* minor fix

* Add ctor to LLVMBasedICFG that allows passing in an LLVMVFTableProvider

* minor

---------

Co-authored-by: Martin Mory <[email protected]>

Author: fabianbs96

BreakingChanges.md

@@ -2,6 +2,15 @@
 
 ## development HEAD
 
+- The API of the `TypeHierarchy` interface (and thus the `LLVMTypeHierarchy` and `DIBasedTypeHierarchy` as well) has changed:
+  - No handling of the super-type relation (only sub-types)
+  - No VTable handling anymore -- has been out-sourced into `LLVMVFTableProvider`
+  - minor API changes
+- The constructors of the call-graph resolvers have changed. They:
+  - take the `LLVMProjectIRDB` as pointer-to-const
+  - take an additional second parameter of type `const LLVMVFTableProvider *`
+  - not necessarily require a `LLVMTypeHierarchy` anymore
+- Some constructors of `LLVMBasedICFG` do not accept a `LLVMTypeHierarchy` pointer anymore
 - Removed IfdsFieldSensTaintAnalysis as it relies on LLVM's deprecated typed-pointers.
 
 ## v2403

include/phasar/PhasarLLVM/ControlFlow/LLVMBasedICFG.h

@@ -21,14 +21,12 @@
 #include "phasar/ControlFlow/CallGraphAnalysisType.h"
 #include "phasar/ControlFlow/ICFGBase.h"
 #include "phasar/PhasarLLVM/ControlFlow/LLVMBasedCFG.h"
+#include "phasar/PhasarLLVM/ControlFlow/LLVMVFTableProvider.h"
 #include "phasar/PhasarLLVM/Pointer/LLVMAliasInfo.h"
 #include "phasar/PhasarLLVM/Utils/LLVMBasedContainerConfig.h"
-#include "phasar/Utils/MaybeUniquePtr.h"
-#include "phasar/Utils/MemoryResource.h"
 #include "phasar/Utils/Soundness.h"
 
 #include "llvm/ADT/ArrayRef.h"
-#include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/Instruction.h"
@@ -90,17 +88,19 @@ class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
                          bool IncludeGlobals = true);
   explicit LLVMBasedICFG(LLVMProjectIRDB *IRDB, Resolver &CGResolver,
                          llvm::ArrayRef<std::string> EntryPoints = {},
-                         LLVMTypeHierarchy *TH = nullptr,
+                         Soundness S = Soundness::Soundy,
+                         bool IncludeGlobals = true);
+  explicit LLVMBasedICFG(LLVMProjectIRDB *IRDB, Resolver &CGResolver,
+                         LLVMVFTableProvider VTP,
+                         llvm::ArrayRef<std::string> EntryPoints = {},
                          Soundness S = Soundness::Soundy,
                          bool IncludeGlobals = true);
 
   /// Creates an ICFG with an already given call-graph
-  explicit LLVMBasedICFG(CallGraph<n_t, f_t> CG, LLVMProjectIRDB *IRDB,
-                         LLVMTypeHierarchy *TH = nullptr);
+  explicit LLVMBasedICFG(CallGraph<n_t, f_t> CG, LLVMProjectIRDB *IRDB);
 
   explicit LLVMBasedICFG(LLVMProjectIRDB *IRDB,
-                         const nlohmann::json &SerializedCG,
-                         LLVMTypeHierarchy *TH = nullptr);
+                         const nlohmann::json &SerializedCG);
 
   // Deleter of LLVMTypeHierarchy may be unknown here...
   ~LLVMBasedICFG();
@@ -165,13 +165,14 @@ class LLVMBasedICFG : public LLVMBasedCFG, public ICFGBase<LLVMBasedICFG> {
 
   void initialize(LLVMProjectIRDB *IRDB, Resolver &CGResolver,
                   llvm::ArrayRef<std::string> EntryPoints,
-                  LLVMTypeHierarchy *TH, Soundness S, bool IncludeGlobals);
+                  const LLVMVFTableProvider &VTP, Soundness S,
+                  bool IncludeGlobals);
 
   // ---
 
   CallGraph<const llvm::Instruction *, const llvm::Function *> CG;
   LLVMProjectIRDB *IRDB = nullptr;
-  MaybeUniquePtr<LLVMTypeHierarchy, true> TH;
+  LLVMVFTableProvider VTP;
 };
 
 extern template class ICFGBase<LLVMBasedICFG>;

include/phasar/PhasarLLVM/ControlFlow/LLVMVFTableProvider.h

@@ -0,0 +1,40 @@
+/******************************************************************************
+ * Copyright (c) 2024 Fabian Schiebel.
+ * All rights reserved. This program and the accompanying materials are made
+ * available under the terms of LICENSE.txt.
+ *
+ * Contributors:
+ *     Fabian Schiebel and others
+ *****************************************************************************/
+
+#ifndef PHASAR_PHASARLLVM_CONTROLFLOW_LLVMVFTABLEPROVIDER_H
+#define PHASAR_PHASARLLVM_CONTROLFLOW_LLVMVFTABLEPROVIDER_H
+
+#include "phasar/PhasarLLVM/TypeHierarchy/LLVMVFTable.h"
+
+#include <unordered_map>
+
+namespace llvm {
+class Module;
+class StructType;
+class GlobalVariable;
+} // namespace llvm
+
+namespace psr {
+class LLVMProjectIRDB;
+
+class LLVMVFTableProvider {
+public:
+  explicit LLVMVFTableProvider(const llvm::Module &Mod);
+  explicit LLVMVFTableProvider(const LLVMProjectIRDB &IRDB);
+
+  [[nodiscard]] bool hasVFTable(const llvm::StructType *Type) const;
+  [[nodiscard]] const LLVMVFTable *
+  getVFTableOrNull(const llvm::StructType *Type) const;
+
+private:
+  std::unordered_map<const llvm::StructType *, LLVMVFTable> TypeVFTMap;
+};
+} // namespace psr
+
+#endif // PHASAR_PHASARLLVM_CONTROLFLOW_LLVMVFTABLEPROVIDER_H

include/phasar/PhasarLLVM/ControlFlow/Resolver/CHAResolver.h

@@ -18,22 +18,28 @@
 #define PHASAR_PHASARLLVM_CONTROLFLOW_RESOLVER_CHARESOLVER_H_
 
 #include "phasar/PhasarLLVM/ControlFlow/Resolver/Resolver.h"
+#include "phasar/Utils/MaybeUniquePtr.h"
 
 namespace llvm {
 class CallBase;
 class Function;
 } // namespace llvm
 
 namespace psr {
+class LLVMTypeHierarchy;
 class CHAResolver : public Resolver {
 public:
-  CHAResolver(LLVMProjectIRDB &IRDB, LLVMTypeHierarchy &TH);
+  CHAResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
+              const LLVMTypeHierarchy *TH);
 
   ~CHAResolver() override = default;
 
   FunctionSetTy resolveVirtualCall(const llvm::CallBase *CallSite) override;
 
   [[nodiscard]] std::string str() const override;
+
+protected:
+  MaybeUniquePtr<const LLVMTypeHierarchy, true> TH;
 };
 } // namespace psr
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/DTAResolver.h

@@ -20,7 +20,7 @@
 #include "phasar/PhasarLLVM/ControlFlow/Resolver/CHAResolver.h"
 #include "phasar/PhasarLLVM/Pointer/TypeGraphs/CachedTypeGraph.h"
 // To switch the TypeGraph
-//#include "phasar/PhasarLLVM/Pointer/TypeGraphs/LazyTypeGraph.h"
+// #include "phasar/PhasarLLVM/Pointer/TypeGraphs/LazyTypeGraph.h"
 
 #include <string>
 
@@ -55,7 +55,8 @@ class DTAResolver : public CHAResolver {
   bool heuristicAntiConstructorVtablePos(const llvm::BitCastInst *BitCast);
 
 public:
-  DTAResolver(LLVMProjectIRDB &IRDB, LLVMTypeHierarchy &TH);
+  DTAResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
+              const LLVMTypeHierarchy *TH);
 
   ~DTAResolver() override = default;
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/NOResolver.h

@@ -28,7 +28,7 @@ class NOResolver final : public Resolver {
                             const llvm::CallBase *CallSite);
 
 public:
-  NOResolver(LLVMProjectIRDB &IRDB);
+  NOResolver(const LLVMProjectIRDB *IRDB);
 
   ~NOResolver() override = default;
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/OTFResolver.h

@@ -17,7 +17,7 @@
 #ifndef PHASAR_PHASARLLVM_CONTROLFLOW_RESOLVER_OTFRESOLVER_H_
 #define PHASAR_PHASARLLVM_CONTROLFLOW_RESOLVER_OTFRESOLVER_H_
 
-#include "phasar/PhasarLLVM/ControlFlow/Resolver/CHAResolver.h"
+#include "phasar/PhasarLLVM/ControlFlow/Resolver/Resolver.h"
 #include "phasar/PhasarLLVM/Pointer/LLVMAliasInfo.h"
 
 #include <set>
@@ -42,7 +42,7 @@ class OTFResolver : public Resolver {
   LLVMAliasInfoRef PT;
 
 public:
-  OTFResolver(LLVMProjectIRDB &IRDB, LLVMTypeHierarchy &TH,
+  OTFResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
               LLVMAliasInfoRef PT);
 
   ~OTFResolver() override = default;

include/phasar/PhasarLLVM/ControlFlow/Resolver/RTAResolver.h

@@ -31,7 +31,8 @@ class StructType;
 namespace psr {
 class RTAResolver : public CHAResolver {
 public:
-  RTAResolver(LLVMProjectIRDB &IRDB, LLVMTypeHierarchy &TH);
+  RTAResolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP,
+              const LLVMTypeHierarchy *TH);
 
   ~RTAResolver() override = default;
 

include/phasar/PhasarLLVM/ControlFlow/Resolver/Resolver.h

@@ -34,9 +34,9 @@ class StructType;
 
 namespace psr {
 class LLVMProjectIRDB;
+class LLVMVFTableProvider;
 class LLVMTypeHierarchy;
 enum class CallGraphAnalysisType;
-class LLVMPointsToInfo;
 
 [[nodiscard]] std::optional<unsigned>
 getVFTIndex(const llvm::CallBase *CallSite);
@@ -51,10 +51,10 @@ getReceiverType(const llvm::CallBase *CallSite);
 
 class Resolver {
 protected:
-  LLVMProjectIRDB &IRDB;
-  LLVMTypeHierarchy *TH;
+  const LLVMProjectIRDB *IRDB;
+  const LLVMVFTableProvider *VTP;
 
-  Resolver(LLVMProjectIRDB &IRDB);
+  Resolver(const LLVMProjectIRDB *IRDB);
 
   const llvm::Function *
   getNonPureVirtualVFTEntry(const llvm::StructType *T, unsigned Idx,
@@ -63,7 +63,7 @@ class Resolver {
 public:
   using FunctionSetTy = llvm::SmallDenseSet<const llvm::Function *, 4>;
 
-  Resolver(LLVMProjectIRDB &IRDB, LLVMTypeHierarchy &TH);
+  Resolver(const LLVMProjectIRDB *IRDB, const LLVMVFTableProvider *VTP);
 
   virtual ~Resolver() = default;
 
@@ -83,8 +83,9 @@ class Resolver {
   [[nodiscard]] virtual std::string str() const = 0;
 
   static std::unique_ptr<Resolver> create(CallGraphAnalysisType Ty,
-                                          LLVMProjectIRDB *IRDB,
-                                          LLVMTypeHierarchy *TH,
+                                          const LLVMProjectIRDB *IRDB,
+                                          const LLVMVFTableProvider *VTP,
+                                          const LLVMTypeHierarchy *TH,
                                           LLVMAliasInfoRef PT = nullptr);
 };
 } // namespace psr

include/phasar/PhasarLLVM/TypeHierarchy/DIBasedTypeHierarchy.h

@@ -38,11 +38,12 @@ class DIBasedTypeHierarchy
     return TypeToVertex.count(Type);
   }
 
-  [[nodiscard]] bool isSubType(ClassType Type, ClassType SubType) override {
+  [[nodiscard]] bool isSubType(ClassType Type,
+                               ClassType SubType) const override {
     return llvm::is_contained(subTypesOf(Type), SubType);
   }
 
-  [[nodiscard]] std::set<ClassType> getSubTypes(ClassType Type) override {
+  [[nodiscard]] std::set<ClassType> getSubTypes(ClassType Type) const override {
     const auto &Range = subTypesOf(Type);
     return {Range.begin(), Range.end()};
   }
@@ -51,39 +52,27 @@ class DIBasedTypeHierarchy
   [[nodiscard]] llvm::iterator_range<const ClassType *>
   subTypesOf(ClassType Ty) const noexcept;
 
-  [[nodiscard]] bool isSuperType(ClassType Type, ClassType SuperType) override;
-
-  /// Not supported yet
-  [[nodiscard]] std::set<ClassType> getSuperTypes(ClassType Type) override;
-
   [[nodiscard]] ClassType
-  getType(std::string TypeName) const noexcept override {
+  getType(llvm::StringRef TypeName) const noexcept override {
     return NameToType.lookup(TypeName);
   }
 
-  [[nodiscard]] std::set<ClassType> getAllTypes() const override {
+  [[nodiscard]] std::vector<ClassType> getAllTypes() const override {
     return {VertexTypes.begin(), VertexTypes.end()};
   }
 
   [[nodiscard]] const auto &getAllVTables() const noexcept { return VTables; }
 
-  [[nodiscard]] std::string getTypeName(ClassType Type) const override {
-    return Type->getName().str();
+  [[nodiscard]] llvm::StringRef getTypeName(ClassType Type) const override {
+    return Type->getName();
   }
 
-  [[nodiscard]] bool hasVFTable(ClassType Type) const override;
-
-  [[nodiscard]] const VFTable<f_t> *getVFTable(ClassType Type) const override {
-    auto It = TypeToVertex.find(Type);
-    if (It == TypeToVertex.end()) {
-      return nullptr;
-    }
-    return &VTables[It->second];
+  [[nodiscard]] size_t size() const noexcept override {
+    return VertexTypes.size();
+  }
+  [[nodiscard]] bool empty() const noexcept override {
+    return VertexTypes.empty();
   }
-
-  [[nodiscard]] size_t size() const override { return VertexTypes.size(); }
-
-  [[nodiscard]] bool empty() const override { return VertexTypes.empty(); }
 
   void print(llvm::raw_ostream &OS = llvm::outs()) const override;