1 //== CStringSyntaxChecker.cpp - CoreFoundation containers API *- C++ -*-==// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // An AST checker that looks for common pitfalls when using C string APIs. 11 // - Identifies erroneous patterns in the last argument to strncat - the number 12 // of bytes to copy. 13 // 14 //===----------------------------------------------------------------------===// 15 #include "ClangSACheckers.h" 16 #include "clang/AST/Expr.h" 17 #include "clang/AST/OperationKinds.h" 18 #include "clang/AST/StmtVisitor.h" 19 #include "clang/Analysis/AnalysisDeclContext.h" 20 #include "clang/Basic/TargetInfo.h" 21 #include "clang/Basic/TypeTraits.h" 22 #include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.h" 23 #include "clang/StaticAnalyzer/Core/Checker.h" 24 #include "clang/StaticAnalyzer/Core/PathSensitive/AnalysisManager.h" 25 #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h" 26 #include "llvm/ADT/SmallString.h" 27 #include "llvm/Support/raw_ostream.h" 28 29 using namespace clang; 30 using namespace ento; 31 32 namespace { 33 class WalkAST: public StmtVisitor<WalkAST> { 34 const CheckerBase *Checker; 35 BugReporter &BR; 36 AnalysisDeclContext* AC; 37 38 /// Check if two expressions refer to the same declaration. 39 bool sameDecl(const Expr *A1, const Expr *A2) { 40 if (const auto *D1 = dyn_cast<DeclRefExpr>(A1->IgnoreParenCasts())) 41 if (const auto *D2 = dyn_cast<DeclRefExpr>(A2->IgnoreParenCasts())) 42 return D1->getDecl() == D2->getDecl(); 43 return false; 44 } 45 46 /// Check if the expression E is a sizeof(WithArg). 47 bool isSizeof(const Expr *E, const Expr *WithArg) { 48 if (const auto *UE = dyn_cast<UnaryExprOrTypeTraitExpr>(E)) 49 if (UE->getKind() == UETT_SizeOf && !UE->isArgumentType()) 50 return sameDecl(UE->getArgumentExpr(), WithArg); 51 return false; 52 } 53 54 /// Check if the expression E is a strlen(WithArg). 55 bool isStrlen(const Expr *E, const Expr *WithArg) { 56 if (const auto *CE = dyn_cast<CallExpr>(E)) { 57 const FunctionDecl *FD = CE->getDirectCallee(); 58 if (!FD) 59 return false; 60 return (CheckerContext::isCLibraryFunction(FD, "strlen") && 61 sameDecl(CE->getArg(0), WithArg)); 62 } 63 return false; 64 } 65 66 /// Check if the expression is an integer literal with value 1. 67 bool isOne(const Expr *E) { 68 if (const auto *IL = dyn_cast<IntegerLiteral>(E)) 69 return (IL->getValue().isIntN(1)); 70 return false; 71 } 72 73 StringRef getPrintableName(const Expr *E) { 74 if (const auto *D = dyn_cast<DeclRefExpr>(E->IgnoreParenCasts())) 75 return D->getDecl()->getName(); 76 return StringRef(); 77 } 78 79 /// Identify erroneous patterns in the last argument to strncat - the number 80 /// of bytes to copy. 81 bool containsBadStrncatPattern(const CallExpr *CE); 82 83 public: 84 WalkAST(const CheckerBase *Checker, BugReporter &BR, AnalysisDeclContext *AC) 85 : Checker(Checker), BR(BR), AC(AC) {} 86 87 // Statement visitor methods. 88 void VisitChildren(Stmt *S); 89 void VisitStmt(Stmt *S) { 90 VisitChildren(S); 91 } 92 void VisitCallExpr(CallExpr *CE); 93 }; 94 } // end anonymous namespace 95 96 // The correct size argument should look like following: 97 // strncat(dst, src, sizeof(dst) - strlen(dest) - 1); 98 // We look for the following anti-patterns: 99 // - strncat(dst, src, sizeof(dst) - strlen(dst)); 100 // - strncat(dst, src, sizeof(dst) - 1); 101 // - strncat(dst, src, sizeof(dst)); 102 bool WalkAST::containsBadStrncatPattern(const CallExpr *CE) { 103 if (CE->getNumArgs() != 3) 104 return false; 105 const Expr *DstArg = CE->getArg(0); 106 const Expr *SrcArg = CE->getArg(1); 107 const Expr *LenArg = CE->getArg(2); 108 109 // Identify wrong size expressions, which are commonly used instead. 110 if (const auto *BE = dyn_cast<BinaryOperator>(LenArg->IgnoreParenCasts())) { 111 // - sizeof(dst) - strlen(dst) 112 if (BE->getOpcode() == BO_Sub) { 113 const Expr *L = BE->getLHS(); 114 const Expr *R = BE->getRHS(); 115 if (isSizeof(L, DstArg) && isStrlen(R, DstArg)) 116 return true; 117 118 // - sizeof(dst) - 1 119 if (isSizeof(L, DstArg) && isOne(R->IgnoreParenCasts())) 120 return true; 121 } 122 } 123 // - sizeof(dst) 124 if (isSizeof(LenArg, DstArg)) 125 return true; 126 127 // - sizeof(src) 128 if (isSizeof(LenArg, SrcArg)) 129 return true; 130 return false; 131 } 132 133 void WalkAST::VisitCallExpr(CallExpr *CE) { 134 const FunctionDecl *FD = CE->getDirectCallee(); 135 if (!FD) 136 return; 137 138 if (CheckerContext::isCLibraryFunction(FD, "strncat")) { 139 if (containsBadStrncatPattern(CE)) { 140 const Expr *DstArg = CE->getArg(0); 141 const Expr *LenArg = CE->getArg(2); 142 PathDiagnosticLocation Loc = 143 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC); 144 145 StringRef DstName = getPrintableName(DstArg); 146 147 SmallString<256> S; 148 llvm::raw_svector_ostream os(S); 149 os << "Potential buffer overflow. "; 150 if (!DstName.empty()) { 151 os << "Replace with 'sizeof(" << DstName << ") " 152 "- strlen(" << DstName <<") - 1'"; 153 os << " or u"; 154 } else 155 os << "U"; 156 os << "se a safer 'strlcat' API"; 157 158 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument", 159 "C String API", os.str(), Loc, 160 LenArg->getSourceRange()); 161 } 162 } 163 164 // Recurse and check children. 165 VisitChildren(CE); 166 } 167 168 void WalkAST::VisitChildren(Stmt *S) { 169 for (Stmt *Child : S->children()) 170 if (Child) 171 Visit(Child); 172 } 173 174 namespace { 175 class CStringSyntaxChecker: public Checker<check::ASTCodeBody> { 176 public: 177 178 void checkASTCodeBody(const Decl *D, AnalysisManager& Mgr, 179 BugReporter &BR) const { 180 WalkAST walker(this, BR, Mgr.getAnalysisDeclContext(D)); 181 walker.Visit(D->getBody()); 182 } 183 }; 184 } 185 186 void ento::registerCStringSyntaxChecker(CheckerManager &mgr) { 187 mgr.registerChecker<CStringSyntaxChecker>(); 188 } 189 190