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/Analysis/AnalysisContext.h" 17#include "clang/AST/Expr.h" 18#include "clang/AST/OperationKinds.h" 19#include "clang/AST/StmtVisitor.h" 20#include "clang/Basic/TargetInfo.h" 21#include "clang/Basic/TypeTraits.h" 22#include "clang/StaticAnalyzer/Core/Checker.h" 23#include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.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 29using namespace clang; 30using namespace ento; 31 32namespace { 33class WalkAST: public StmtVisitor<WalkAST> { 34 BugReporter &BR; 35 AnalysisDeclContext* AC; 36 ASTContext &ASTC; 37 38 /// Check if two expressions refer to the same declaration. 39 inline bool sameDecl(const Expr *A1, const Expr *A2) { 40 if (const DeclRefExpr *D1 = dyn_cast<DeclRefExpr>(A1->IgnoreParenCasts())) 41 if (const DeclRefExpr *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 inline bool isSizeof(const Expr *E, const Expr *WithArg) { 48 if (const UnaryExprOrTypeTraitExpr *UE = 49 dyn_cast<UnaryExprOrTypeTraitExpr>(E)) 50 if (UE->getKind() == UETT_SizeOf) 51 return sameDecl(UE->getArgumentExpr(), WithArg); 52 return false; 53 } 54 55 /// Check if the expression E is a strlen(WithArg). 56 inline bool isStrlen(const Expr *E, const Expr *WithArg) { 57 if (const CallExpr *CE = dyn_cast<CallExpr>(E)) { 58 const FunctionDecl *FD = CE->getDirectCallee(); 59 if (!FD) 60 return false; 61 return (CheckerContext::isCLibraryFunction(FD, "strlen", ASTC) 62 && sameDecl(CE->getArg(0), WithArg)); 63 } 64 return false; 65 } 66 67 /// Check if the expression is an integer literal with value 1. 68 inline bool isOne(const Expr *E) { 69 if (const IntegerLiteral *IL = dyn_cast<IntegerLiteral>(E)) 70 return (IL->getValue().isIntN(1)); 71 return false; 72 } 73 74 inline StringRef getPrintableName(const Expr *E) { 75 if (const DeclRefExpr *D = dyn_cast<DeclRefExpr>(E->IgnoreParenCasts())) 76 return D->getDecl()->getName(); 77 return StringRef(); 78 } 79 80 /// Identify erroneous patterns in the last argument to strncat - the number 81 /// of bytes to copy. 82 bool containsBadStrncatPattern(const CallExpr *CE); 83 84public: 85 WalkAST(BugReporter &br, AnalysisDeclContext* ac) : 86 BR(br), AC(ac), ASTC(AC->getASTContext()) { 87 } 88 89 // Statement visitor methods. 90 void VisitChildren(Stmt *S); 91 void VisitStmt(Stmt *S) { 92 VisitChildren(S); 93 } 94 void VisitCallExpr(CallExpr *CE); 95}; 96} // end anonymous namespace 97 98// The correct size argument should look like following: 99// strncat(dst, src, sizeof(dst) - strlen(dest) - 1); 100// We look for the following anti-patterns: 101// - strncat(dst, src, sizeof(dst) - strlen(dst)); 102// - strncat(dst, src, sizeof(dst) - 1); 103// - strncat(dst, src, sizeof(dst)); 104bool WalkAST::containsBadStrncatPattern(const CallExpr *CE) { 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 BinaryOperator *BE = 111 dyn_cast<BinaryOperator>(LenArg->IgnoreParenCasts())) { 112 // - sizeof(dst) - strlen(dst) 113 if (BE->getOpcode() == BO_Sub) { 114 const Expr *L = BE->getLHS(); 115 const Expr *R = BE->getRHS(); 116 if (isSizeof(L, DstArg) && isStrlen(R, DstArg)) 117 return true; 118 119 // - sizeof(dst) - 1 120 if (isSizeof(L, DstArg) && isOne(R->IgnoreParenCasts())) 121 return true; 122 } 123 } 124 // - sizeof(dst) 125 if (isSizeof(LenArg, DstArg)) 126 return true; 127 128 // - sizeof(src) 129 if (isSizeof(LenArg, SrcArg)) 130 return true; 131 return false; 132} 133 134void WalkAST::VisitCallExpr(CallExpr *CE) { 135 const FunctionDecl *FD = CE->getDirectCallee(); 136 if (!FD) 137 return; 138 139 if (CheckerContext::isCLibraryFunction(FD, "strncat", ASTC)) { 140 if (containsBadStrncatPattern(CE)) { 141 const Expr *DstArg = CE->getArg(0); 142 const Expr *LenArg = CE->getArg(2); 143 SourceRange R = LenArg->getSourceRange(); 144 PathDiagnosticLocation Loc = 145 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC); 146 147 StringRef DstName = getPrintableName(DstArg); 148 149 SmallString<256> S; 150 llvm::raw_svector_ostream os(S); 151 os << "Potential buffer overflow. "; 152 if (!DstName.empty()) { 153 os << "Replace with 'sizeof(" << DstName << ") " 154 "- strlen(" << DstName <<") - 1'"; 155 os << " or u"; 156 } else 157 os << "U"; 158 os << "se a safer 'strlcat' API"; 159 160 BR.EmitBasicReport(FD, "Anti-pattern in the argument", "C String API", 161 os.str(), Loc, &R, 1); 162 } 163 } 164 165 // Recurse and check children. 166 VisitChildren(CE); 167} 168 169void WalkAST::VisitChildren(Stmt *S) { 170 for (Stmt::child_iterator I = S->child_begin(), E = S->child_end(); I != E; 171 ++I) 172 if (Stmt *child = *I) 173 Visit(child); 174} 175 176namespace { 177class CStringSyntaxChecker: public Checker<check::ASTCodeBody> { 178public: 179 180 void checkASTCodeBody(const Decl *D, AnalysisManager& Mgr, 181 BugReporter &BR) const { 182 WalkAST walker(BR, Mgr.getAnalysisDeclContext(D)); 183 walker.Visit(D->getBody()); 184 } 185}; 186} 187 188void ento::registerCStringSyntaxChecker(CheckerManager &mgr) { 189 mgr.registerChecker<CStringSyntaxChecker>(); 190} 191 192