1 //== CStringSyntaxChecker.cpp - CoreFoundation containers API *- C++ -*-==//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
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
14 //===----------------------------------------------------------------------===//
15 #include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.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"
29 using namespace clang;
33 class WalkAST: public StmtVisitor<WalkAST> {
34 const CheckerBase *Checker;
36 AnalysisDeclContext* AC;
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();
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);
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();
60 return (CheckerContext::isCLibraryFunction(FD, "strlen") &&
61 sameDecl(CE->getArg(0), WithArg));
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));
73 StringRef getPrintableName(const Expr *E) {
74 if (const auto *D = dyn_cast<DeclRefExpr>(E->IgnoreParenCasts()))
75 return D->getDecl()->getName();
79 /// Identify erroneous patterns in the last argument to strncat - the number
81 bool containsBadStrncatPattern(const CallExpr *CE);
83 /// Identify erroneous patterns in the last argument to strlcpy - the number
85 /// The bad pattern checked is when the size is known
86 /// to be larger than the destination can handle.
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
95 /// The bad pattern checked is when the last argument is basically
96 /// pointing to the destination buffer size or argument larger or
99 /// strlcat(dst, src2, sizeof(dst));
100 /// strlcat(dst, src2, 2);
101 /// strlcat(dst, src2, 10);
102 bool containsBadStrlcpyStrlcatPattern(const CallExpr *CE);
105 WalkAST(const CheckerBase *Checker, BugReporter &BR, AnalysisDeclContext *AC)
106 : Checker(Checker), BR(BR), AC(AC) {}
108 // Statement visitor methods.
109 void VisitChildren(Stmt *S);
110 void VisitStmt(Stmt *S) {
113 void VisitCallExpr(CallExpr *CE);
115 } // end anonymous namespace
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)
126 const Expr *DstArg = CE->getArg(0);
127 const Expr *SrcArg = CE->getArg(1);
128 const Expr *LenArg = CE->getArg(2);
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))
140 if (isSizeof(L, DstArg) && isOne(R->IgnoreParenCasts()))
145 if (isSizeof(LenArg, DstArg))
149 if (isSizeof(LenArg, SrcArg))
154 bool WalkAST::containsBadStrlcpyStrlcatPattern(const CallExpr *CE) {
155 if (CE->getNumArgs() != 3)
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);
162 const auto *DstArgDecl = dyn_cast<DeclRefExpr>(DstArg->IgnoreParenImpCasts());
163 const auto *LenArgDecl = dyn_cast<DeclRefExpr>(LenArg->IgnoreParenLValueCasts());
165 if (isSizeof(LenArg, DstArg))
167 // - size_t dstlen = sizeof(dst)
169 const auto *LenArgVal = dyn_cast<VarDecl>(LenArgDecl->getDecl());
170 if (LenArgVal->getInit())
171 LenArg = LenArgVal->getInit();
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();
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.
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();
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;
199 if (RemainingBufferLen <= ILRawVal)
202 if (RemainingBufferLen < ILRawVal)
212 void WalkAST::VisitCallExpr(CallExpr *CE) {
213 const FunctionDecl *FD = CE->getDirectCallee();
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);
224 StringRef DstName = getPrintableName(DstArg);
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'";
235 os << "se a safer 'strlcat' API";
237 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument",
238 "C String API", os.str(), Loc,
239 LenArg->getSourceRange());
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);
249 StringRef DstName = getPrintableName(DstArg);
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 << ")";
258 os << "sizeof(<destination buffer>)";
261 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument",
262 "C String API", os.str(), Loc,
263 LenArg->getSourceRange());
267 // Recurse and check children.
271 void WalkAST::VisitChildren(Stmt *S) {
272 for (Stmt *Child : S->children())
278 class CStringSyntaxChecker: public Checker<check::ASTCodeBody> {
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());
289 void ento::registerCStringSyntaxChecker(CheckerManager &mgr) {
290 mgr.registerChecker<CStringSyntaxChecker>();