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 /// Identify erroneous patterns in the last argument to strlcpy - the number 84 /// of bytes to copy. 85 /// The bad pattern checked is when the size is known 86 /// to be larger than the destination can handle. 87 /// char dst[2]; 88 /// size_t cpy = 4; 89 /// strlcpy(dst, "abcd", sizeof("abcd") - 1); 90 /// strlcpy(dst, "abcd", 4); 91 /// strlcpy(dst + 3, "abcd", 2); 92 /// strlcpy(dst, "abcd", cpy); 93 /// Identify erroneous patterns in the last argument to strlcat - the number 94 /// of bytes to copy. 95 /// The bad pattern checked is when the last argument is basically 96 /// pointing to the destination buffer size or argument larger or 97 /// equal to. 98 /// char dst[2]; 99 /// strlcat(dst, src2, sizeof(dst)); 100 /// strlcat(dst, src2, 2); 101 /// strlcat(dst, src2, 10); 102 bool containsBadStrlcpyStrlcatPattern(const CallExpr *CE); 103 104 public: 105 WalkAST(const CheckerBase *Checker, BugReporter &BR, AnalysisDeclContext *AC) 106 : Checker(Checker), BR(BR), AC(AC) {} 107 108 // Statement visitor methods. 109 void VisitChildren(Stmt *S); 110 void VisitStmt(Stmt *S) { 111 VisitChildren(S); 112 } 113 void VisitCallExpr(CallExpr *CE); 114 }; 115 } // end anonymous namespace 116 117 // The correct size argument should look like following: 118 // strncat(dst, src, sizeof(dst) - strlen(dest) - 1); 119 // We look for the following anti-patterns: 120 // - strncat(dst, src, sizeof(dst) - strlen(dst)); 121 // - strncat(dst, src, sizeof(dst) - 1); 122 // - strncat(dst, src, sizeof(dst)); 123 bool WalkAST::containsBadStrncatPattern(const CallExpr *CE) { 124 if (CE->getNumArgs() != 3) 125 return false; 126 const Expr *DstArg = CE->getArg(0); 127 const Expr *SrcArg = CE->getArg(1); 128 const Expr *LenArg = CE->getArg(2); 129 130 // Identify wrong size expressions, which are commonly used instead. 131 if (const auto *BE = dyn_cast<BinaryOperator>(LenArg->IgnoreParenCasts())) { 132 // - sizeof(dst) - strlen(dst) 133 if (BE->getOpcode() == BO_Sub) { 134 const Expr *L = BE->getLHS(); 135 const Expr *R = BE->getRHS(); 136 if (isSizeof(L, DstArg) && isStrlen(R, DstArg)) 137 return true; 138 139 // - sizeof(dst) - 1 140 if (isSizeof(L, DstArg) && isOne(R->IgnoreParenCasts())) 141 return true; 142 } 143 } 144 // - sizeof(dst) 145 if (isSizeof(LenArg, DstArg)) 146 return true; 147 148 // - sizeof(src) 149 if (isSizeof(LenArg, SrcArg)) 150 return true; 151 return false; 152 } 153 154 bool WalkAST::containsBadStrlcpyStrlcatPattern(const CallExpr *CE) { 155 if (CE->getNumArgs() != 3) 156 return false; 157 const FunctionDecl *FD = CE->getDirectCallee(); 158 bool Append = CheckerContext::isCLibraryFunction(FD, "strlcat"); 159 const Expr *DstArg = CE->getArg(0); 160 const Expr *LenArg = CE->getArg(2); 161 162 const auto *DstArgDecl = dyn_cast<DeclRefExpr>(DstArg->IgnoreParenImpCasts()); 163 const auto *LenArgDecl = dyn_cast<DeclRefExpr>(LenArg->IgnoreParenLValueCasts()); 164 uint64_t DstOff = 0; 165 if (isSizeof(LenArg, DstArg)) 166 return false; 167 // - size_t dstlen = sizeof(dst) 168 if (LenArgDecl) { 169 const auto *LenArgVal = dyn_cast<VarDecl>(LenArgDecl->getDecl()); 170 if (LenArgVal->getInit()) 171 LenArg = LenArgVal->getInit(); 172 } 173 174 // - integral value 175 // We try to figure out if the last argument is possibly longer 176 // than the destination can possibly handle if its size can be defined. 177 if (const auto *IL = dyn_cast<IntegerLiteral>(LenArg->IgnoreParenImpCasts())) { 178 uint64_t ILRawVal = IL->getValue().getZExtValue(); 179 180 // Case when there is pointer arithmetic on the destination buffer 181 // especially when we offset from the base decreasing the 182 // buffer length accordingly. 183 if (!DstArgDecl) { 184 if (const auto *BE = dyn_cast<BinaryOperator>(DstArg->IgnoreParenImpCasts())) { 185 DstArgDecl = dyn_cast<DeclRefExpr>(BE->getLHS()->IgnoreParenImpCasts()); 186 if (BE->getOpcode() == BO_Add) { 187 if ((IL = dyn_cast<IntegerLiteral>(BE->getRHS()->IgnoreParenImpCasts()))) { 188 DstOff = IL->getValue().getZExtValue(); 189 } 190 } 191 } 192 } 193 if (DstArgDecl) { 194 if (const auto *Buffer = dyn_cast<ConstantArrayType>(DstArgDecl->getType())) { 195 ASTContext &C = BR.getContext(); 196 uint64_t BufferLen = C.getTypeSize(Buffer) / 8; 197 auto RemainingBufferLen = BufferLen - DstOff; 198 if (Append) { 199 if (RemainingBufferLen <= ILRawVal) 200 return true; 201 } else { 202 if (RemainingBufferLen < ILRawVal) 203 return true; 204 } 205 } 206 } 207 } 208 209 return false; 210 } 211 212 void WalkAST::VisitCallExpr(CallExpr *CE) { 213 const FunctionDecl *FD = CE->getDirectCallee(); 214 if (!FD) 215 return; 216 217 if (CheckerContext::isCLibraryFunction(FD, "strncat")) { 218 if (containsBadStrncatPattern(CE)) { 219 const Expr *DstArg = CE->getArg(0); 220 const Expr *LenArg = CE->getArg(2); 221 PathDiagnosticLocation Loc = 222 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC); 223 224 StringRef DstName = getPrintableName(DstArg); 225 226 SmallString<256> S; 227 llvm::raw_svector_ostream os(S); 228 os << "Potential buffer overflow. "; 229 if (!DstName.empty()) { 230 os << "Replace with 'sizeof(" << DstName << ") " 231 "- strlen(" << DstName <<") - 1'"; 232 os << " or u"; 233 } else 234 os << "U"; 235 os << "se a safer 'strlcat' API"; 236 237 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument", 238 "C String API", os.str(), Loc, 239 LenArg->getSourceRange()); 240 } 241 } else if (CheckerContext::isCLibraryFunction(FD, "strlcpy") || 242 CheckerContext::isCLibraryFunction(FD, "strlcat")) { 243 if (containsBadStrlcpyStrlcatPattern(CE)) { 244 const Expr *DstArg = CE->getArg(0); 245 const Expr *LenArg = CE->getArg(2); 246 PathDiagnosticLocation Loc = 247 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC); 248 249 StringRef DstName = getPrintableName(DstArg); 250 251 SmallString<256> S; 252 llvm::raw_svector_ostream os(S); 253 os << "The third argument allows to potentially copy more bytes than it should. "; 254 os << "Replace with the value "; 255 if (!DstName.empty()) 256 os << "sizeof(" << DstName << ")"; 257 else 258 os << "sizeof(<destination buffer>)"; 259 os << " or lower"; 260 261 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument", 262 "C String API", os.str(), Loc, 263 LenArg->getSourceRange()); 264 } 265 } 266 267 // Recurse and check children. 268 VisitChildren(CE); 269 } 270 271 void WalkAST::VisitChildren(Stmt *S) { 272 for (Stmt *Child : S->children()) 273 if (Child) 274 Visit(Child); 275 } 276 277 namespace { 278 class CStringSyntaxChecker: public Checker<check::ASTCodeBody> { 279 public: 280 281 void checkASTCodeBody(const Decl *D, AnalysisManager& Mgr, 282 BugReporter &BR) const { 283 WalkAST walker(this, BR, Mgr.getAnalysisDeclContext(D)); 284 walker.Visit(D->getBody()); 285 } 286 }; 287 } 288 289 void ento::registerCStringSyntaxChecker(CheckerManager &mgr) { 290 mgr.registerChecker<CStringSyntaxChecker>(); 291 } 292 293