Complete PDG build revert

Author: ARISTODE

include/ProgramDependencyGraph.hh

@@ -4,7 +4,7 @@
 #include "Graph.hh"
 #include "PDGCallGraph.hh"
 #include "DataDependencyGraph.hh"
-// #include "ControlDependencyGraph.hh"
+#include "ControlDependencyGraph.hh"
 
 namespace pdg
 {

src/DataAccessAnalysis.cpp

@@ -39,7 +39,7 @@ bool pdg::DataAccessAnalysis::runOnModule(Module &M)
   _module = &M;
   _SDA = &getAnalysis<SharedDataAnalysis>();
   _PDG = _SDA->getPDG();
-  _callGraph = &PDGCallGraph::getInstance();
+  _callGraph = &PDGCallGraph::getInstance();;
   _ksplitStats = &KSplitStats::getInstance();
   computeExportedFuncsPtrNameMap();
   readDriverDefinedGlobalVarNames("driver_globalvar_names");
@@ -65,8 +65,8 @@ bool pdg::DataAccessAnalysis::runOnModule(Module &M)
     // compute data access for function arguments, used later for IDL generation
     computeDataAccessForFuncArgs(F);
     // generate json object for eBPF enforcement
-    if (_PDG->interfaceFuncs.find(f) != _PDG->interfaceFuncs.end())
-      generateJSONObjectForFunc(F, moduleJsonObj);
+    // if (_PDG->interfaceFuncs.find(f) != _PDG->interfaceFuncs.end())
+    //   generateJSONObjectForFunc(F, moduleJsonObj);
     total_num_funcs++;
   }
 
@@ -1007,11 +1007,14 @@ void pdg::DataAccessAnalysis::computeDataAccessForFuncArgs(Function &F)
   auto argRetTree = fw->getRetFormalInTree();
   computeDataAccessForTree(argRetTree, true);
   // compute arg access info
-  auto argTreeMap = fw->getArgFormalInTreeMap();
-  for (auto iter = argTreeMap.begin(); iter != argTreeMap.end(); iter++)
+  auto argFormalTreeMap = fw->getArgFormalInTreeMap();
+  for (auto iter = argFormalTreeMap.begin(); iter != argFormalTreeMap.end(); iter++)
   {
     Tree *argTree = iter->second;
     computeDataAccessForTree(argTree);
+    // auto arg = iter->first;
+    // if (arg->getType()->isFunctionTy())
+    //   errs() << "find function type in " << F.getName() << "\n";
   }
 }
 
@@ -1596,6 +1599,8 @@ void pdg::DataAccessAnalysis::generateRpcForFunc(Function &F, bool processExport
 {
   // need to use this wrapper to handle nescheck rewrittern funcs
   FunctionWrapper *fw = getNescheckFuncWrapper(F);
+  if (fw == nullptr)
+    return;
   std::string rpcStr = "";
   // generate function rpc stub
   // first generate for return value
@@ -1750,6 +1755,8 @@ void pdg::DataAccessAnalysis::generateIDLForFunc(Function &F, bool processingExp
   // assert(func_iter != func_wrapper_map.end() && "no function wrapper found (IDL-GEN)!");
   // auto fw = func_iter->second;
   FunctionWrapper *fw = getNescheckFuncWrapper(F);
+  if (fw == nullptr)
+    return;
   // process exported funcs later in special manner because of syntax requirement
   if (isExportedFunc(F) && !processingExportedFunc)
   {

src/DataDependencyGraph.cpp

@@ -7,12 +7,6 @@ char pdg::DataDependencyGraph::ID = 0;
 // cl::opt<bool, true> SFA("sf", cl::desc("enable mode for analyzing only one function"), cl::value_desc("single-func"), cl::location(pdg::SingleFuncAnalysis), cl::init(false));
 // cl::opt<std::string> TargetFuncName("tf", cl::desc("Target function to analyze"), cl::value_desc("string"), cl::location(pdg::TargetFuncNameStr));
 
-// specify interface function
-cl::opt<std::string> InterfaceFuncsPath("ifuncs",
-                                cl::desc("Specify the path of the instrumented binary"),
-                                cl::value_desc("target bin path to instrument"),
-                                cl::init(""));
-
 bool pdg::DataDependencyGraph::runOnModule(Module &M)
 {
   _module = &M;
@@ -31,29 +25,16 @@ bool pdg::DataDependencyGraph::runOnModule(Module &M)
 
   ProgramGraph &g = ProgramGraph::getInstance();
 
-  // read interface functions
-  std::ifstream file(InterfaceFuncsPath);
-  if (file.good())
-  {
-    pdgutils::readLinesFromFile(g.iFuncNames, InterfaceFuncsPath);
-    file.close();
-  }
-  else
-  {
-    std::cerr << "File not found: " << InterfaceFuncsPath << std::endl;
-    return false;
-  }
-
   if (!g.isBuild())
   {
     // setup the interface functions for PDG build
     g.build(M);
     g.bindDITypeToNodes(M);
   }
 
-  for (auto f : g.funcToBuild)
+  // for (auto f : g.funcToBuild)
+  for (auto &F : M)
   {
-    Function &F = *f;
     if (F.isDeclaration() || F.empty())
       continue;
 

src/Graph.cpp

@@ -227,21 +227,6 @@ void pdg::ProgramGraph::build(Module &M)
 {
   auto &call_g = PDGCallGraph::getInstance();
   // setup functions to build
-  for (auto fName : iFuncNames)
-  {
-    auto func = M.getFunction(StringRef(fName));
-    interfaceFuncs.insert(func);
-    auto callNode = call_g.getNode(*func);
-    auto transFuncNodes = call_g.computeTransitiveClosure(*callNode);
-    for (auto n : transFuncNodes)
-    {
-      if (!n->getValue())
-        continue;
-      if (auto f = dyn_cast<Function>(n->getValue()))
-        funcToBuild.insert(f);
-    }
-  }
-
   buildGlobalVariables(M);
   buildFunctions(M);
   buildCallGraphAndCallSites(M);
@@ -274,9 +259,8 @@ void pdg::ProgramGraph::buildGlobalVariables(Module &M)
 
 void pdg::ProgramGraph::buildFunctions(Module &M)
 {
-  for (auto f : funcToBuild)
+  for (auto &F : M)
   {
-    Function &F = *f;
     if (F.isDeclaration() || F.empty())
       continue;
     FunctionWrapper *func_w = new FunctionWrapper(&F);
@@ -311,12 +295,10 @@ void pdg::ProgramGraph::buildFunctionInstructions(Function &F, FunctionWrapper *
 
 void pdg::ProgramGraph::buildCallGraphAndCallSites(Module &M)
 {
-  for (auto func : funcToBuild)
+  for (auto &F : M)
   {
-    Function &F = *func;
     if (F.isDeclaration() || F.empty() || !hasFuncWrapper(F))
       continue;
-
     FunctionWrapper *func_w = getFuncWrapper(F);
     auto call_insts = func_w->getCallInsts();
 
@@ -349,9 +331,8 @@ void pdg::ProgramGraph::handleCallSites(Module &M, CallInst *ci)
 
 void pdg::ProgramGraph::bindDITypeToNodes(Module &M)
 {
-  for (auto f : funcToBuild)
+  for (auto &F : M)
   {
-    Function &F = *f;
     if (F.isDeclaration())
       continue;
     FunctionWrapper *fw = _func_wrapper_map[&F];

src/ProgramDependencyGraph.cpp

@@ -15,7 +15,7 @@ bool pdg::DEBUG;
 void pdg::ProgramDependencyGraph::getAnalysisUsage(AnalysisUsage &AU) const
 {
   AU.addRequired<DataDependencyGraph>();
-  // AU.addRequired<ControlDependencyGraph>();
+  AU.addRequired<ControlDependencyGraph>();
   AU.setPreservesAll();
 }
 
@@ -48,9 +48,8 @@ bool pdg::ProgramDependencyGraph::runOnModule(Module &M)
   }
 
   unsigned func_size = 0;
-  for (auto f : _PDG->funcToBuild)
+  for (Function &F : M)
   {
-    Function &F = *f;
     if (F.isDeclaration() || !_PDG->hasFuncWrapper(F))
       continue;
     // if (!call_g.isBuildFuncNode(F))
@@ -206,7 +205,7 @@ void pdg::ProgramDependencyGraph::connectCallerAndCallee(CallWrapper &cw, Functi
 void pdg::ProgramDependencyGraph::connectIntraprocDependencies(Function &F)
 {
   // add control dependency edges
-  // getAnalysis<ControlDependencyGraph>(F); // add data dependencies for nodes in F
+  getAnalysis<ControlDependencyGraph>(F); // add data dependencies for nodes in F
   // connect formal tree with address variables
   FunctionWrapper *func_w = getFuncWrapper(F);
   Node *entry_node = func_w->getEntryNode();

src/RiskyBoundaryAPIAnalysis.cpp

@@ -257,6 +257,7 @@ void pdg::RiskyBoundaryAPIAnalysis::analyzeRiskyBoundaryKernelAPIs(nlohmann::ord
   unsigned caseID = 0;
 
   auto kernelFuncCalledUnderConditions = computeKernelInterfaceFuncCSUnderCondition();
+  errs() << "size of kernel interface call under condition: " << kernelFuncCalledUnderConditions.size() << "\n";
   for (auto boundaryFunc : kernelFuncCalledUnderConditions)
   {
     // only analyze kernel boundary functions