1 //== CStringSyntaxChecker.cpp - CoreFoundation containers API *- C++ -*-==//
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //===----------------------------------------------------------------------===//
9 // An AST checker that looks for common pitfalls when using C string APIs.
10 // - Identifies erroneous patterns in the last argument to strncat - the number
13 //===----------------------------------------------------------------------===//
14 #include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
15 #include "clang/AST/Expr.h"
16 #include "clang/AST/OperationKinds.h"
17 #include "clang/AST/StmtVisitor.h"
18 #include "clang/Analysis/AnalysisDeclContext.h"
19 #include "clang/Basic/TargetInfo.h"
20 #include "clang/Basic/TypeTraits.h"
21 #include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.h"
22 #include "clang/StaticAnalyzer/Core/Checker.h"
23 #include "clang/StaticAnalyzer/Core/PathSensitive/AnalysisManager.h"
24 #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
25 #include "llvm/ADT/SmallString.h"
26 #include "llvm/Support/raw_ostream.h"
28 using namespace clang;
32 class WalkAST: public StmtVisitor<WalkAST> {
33 const CheckerBase *Checker;
35 AnalysisDeclContext* AC;
37 /// Check if two expressions refer to the same declaration.
38 bool sameDecl(const Expr *A1, const Expr *A2) {
39 if (const auto *D1 = dyn_cast<DeclRefExpr>(A1->IgnoreParenCasts()))
40 if (const auto *D2 = dyn_cast<DeclRefExpr>(A2->IgnoreParenCasts()))
41 return D1->getDecl() == D2->getDecl();
45 /// Check if the expression E is a sizeof(WithArg).
46 bool isSizeof(const Expr *E, const Expr *WithArg) {
47 if (const auto *UE = dyn_cast<UnaryExprOrTypeTraitExpr>(E))
48 if (UE->getKind() == UETT_SizeOf && !UE->isArgumentType())
49 return sameDecl(UE->getArgumentExpr(), WithArg);
53 /// Check if the expression E is a strlen(WithArg).
54 bool isStrlen(const Expr *E, const Expr *WithArg) {
55 if (const auto *CE = dyn_cast<CallExpr>(E)) {
56 const FunctionDecl *FD = CE->getDirectCallee();
59 return (CheckerContext::isCLibraryFunction(FD, "strlen") &&
60 sameDecl(CE->getArg(0), WithArg));
65 /// Check if the expression is an integer literal with value 1.
66 bool isOne(const Expr *E) {
67 if (const auto *IL = dyn_cast<IntegerLiteral>(E))
68 return (IL->getValue().isIntN(1));
72 StringRef getPrintableName(const Expr *E) {
73 if (const auto *D = dyn_cast<DeclRefExpr>(E->IgnoreParenCasts()))
74 return D->getDecl()->getName();
78 /// Identify erroneous patterns in the last argument to strncat - the number
80 bool containsBadStrncatPattern(const CallExpr *CE);
82 /// Identify erroneous patterns in the last argument to strlcpy - the number
84 /// The bad pattern checked is when the size is known
85 /// to be larger than the destination can handle.
88 /// strlcpy(dst, "abcd", sizeof("abcd") - 1);
89 /// strlcpy(dst, "abcd", 4);
90 /// strlcpy(dst + 3, "abcd", 2);
91 /// strlcpy(dst, "abcd", cpy);
92 /// Identify erroneous patterns in the last argument to strlcat - the number
94 /// The bad pattern checked is when the last argument is basically
95 /// pointing to the destination buffer size or argument larger or
98 /// strlcat(dst, src2, sizeof(dst));
99 /// strlcat(dst, src2, 2);
100 /// strlcat(dst, src2, 10);
101 bool containsBadStrlcpyStrlcatPattern(const CallExpr *CE);
104 WalkAST(const CheckerBase *Checker, BugReporter &BR, AnalysisDeclContext *AC)
105 : Checker(Checker), BR(BR), AC(AC) {}
107 // Statement visitor methods.
108 void VisitChildren(Stmt *S);
109 void VisitStmt(Stmt *S) {
112 void VisitCallExpr(CallExpr *CE);
114 } // end anonymous namespace
116 // The correct size argument should look like following:
117 // strncat(dst, src, sizeof(dst) - strlen(dest) - 1);
118 // We look for the following anti-patterns:
119 // - strncat(dst, src, sizeof(dst) - strlen(dst));
120 // - strncat(dst, src, sizeof(dst) - 1);
121 // - strncat(dst, src, sizeof(dst));
122 bool WalkAST::containsBadStrncatPattern(const CallExpr *CE) {
123 if (CE->getNumArgs() != 3)
125 const Expr *DstArg = CE->getArg(0);
126 const Expr *SrcArg = CE->getArg(1);
127 const Expr *LenArg = CE->getArg(2);
129 // Identify wrong size expressions, which are commonly used instead.
130 if (const auto *BE = dyn_cast<BinaryOperator>(LenArg->IgnoreParenCasts())) {
131 // - sizeof(dst) - strlen(dst)
132 if (BE->getOpcode() == BO_Sub) {
133 const Expr *L = BE->getLHS();
134 const Expr *R = BE->getRHS();
135 if (isSizeof(L, DstArg) && isStrlen(R, DstArg))
139 if (isSizeof(L, DstArg) && isOne(R->IgnoreParenCasts()))
144 if (isSizeof(LenArg, DstArg))
148 if (isSizeof(LenArg, SrcArg))
153 bool WalkAST::containsBadStrlcpyStrlcatPattern(const CallExpr *CE) {
154 if (CE->getNumArgs() != 3)
156 const Expr *DstArg = CE->getArg(0);
157 const Expr *LenArg = CE->getArg(2);
159 const auto *DstArgDecl = dyn_cast<DeclRefExpr>(DstArg->IgnoreParenImpCasts());
160 const auto *LenArgDecl = dyn_cast<DeclRefExpr>(LenArg->IgnoreParenLValueCasts());
162 if (isSizeof(LenArg, DstArg))
164 // - size_t dstlen = sizeof(dst)
166 const auto *LenArgVal = dyn_cast<VarDecl>(LenArgDecl->getDecl());
167 if (LenArgVal->getInit())
168 LenArg = LenArgVal->getInit();
172 // We try to figure out if the last argument is possibly longer
173 // than the destination can possibly handle if its size can be defined.
174 if (const auto *IL = dyn_cast<IntegerLiteral>(LenArg->IgnoreParenImpCasts())) {
175 uint64_t ILRawVal = IL->getValue().getZExtValue();
177 // Case when there is pointer arithmetic on the destination buffer
178 // especially when we offset from the base decreasing the
179 // buffer length accordingly.
181 if (const auto *BE = dyn_cast<BinaryOperator>(DstArg->IgnoreParenImpCasts())) {
182 DstArgDecl = dyn_cast<DeclRefExpr>(BE->getLHS()->IgnoreParenImpCasts());
183 if (BE->getOpcode() == BO_Add) {
184 if ((IL = dyn_cast<IntegerLiteral>(BE->getRHS()->IgnoreParenImpCasts()))) {
185 DstOff = IL->getValue().getZExtValue();
191 if (const auto *Buffer = dyn_cast<ConstantArrayType>(DstArgDecl->getType())) {
192 ASTContext &C = BR.getContext();
193 uint64_t BufferLen = C.getTypeSize(Buffer) / 8;
194 auto RemainingBufferLen = BufferLen - DstOff;
195 if (RemainingBufferLen < ILRawVal)
204 void WalkAST::VisitCallExpr(CallExpr *CE) {
205 const FunctionDecl *FD = CE->getDirectCallee();
209 if (CheckerContext::isCLibraryFunction(FD, "strncat")) {
210 if (containsBadStrncatPattern(CE)) {
211 const Expr *DstArg = CE->getArg(0);
212 const Expr *LenArg = CE->getArg(2);
213 PathDiagnosticLocation Loc =
214 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC);
216 StringRef DstName = getPrintableName(DstArg);
219 llvm::raw_svector_ostream os(S);
220 os << "Potential buffer overflow. ";
221 if (!DstName.empty()) {
222 os << "Replace with 'sizeof(" << DstName << ") "
223 "- strlen(" << DstName <<") - 1'";
227 os << "se a safer 'strlcat' API";
229 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument",
230 "C String API", os.str(), Loc,
231 LenArg->getSourceRange());
233 } else if (CheckerContext::isCLibraryFunction(FD, "strlcpy") ||
234 CheckerContext::isCLibraryFunction(FD, "strlcat")) {
235 if (containsBadStrlcpyStrlcatPattern(CE)) {
236 const Expr *DstArg = CE->getArg(0);
237 const Expr *LenArg = CE->getArg(2);
238 PathDiagnosticLocation Loc =
239 PathDiagnosticLocation::createBegin(LenArg, BR.getSourceManager(), AC);
241 StringRef DstName = getPrintableName(DstArg);
244 llvm::raw_svector_ostream os(S);
245 os << "The third argument allows to potentially copy more bytes than it should. ";
246 os << "Replace with the value ";
247 if (!DstName.empty())
248 os << "sizeof(" << DstName << ")";
250 os << "sizeof(<destination buffer>)";
253 BR.EmitBasicReport(FD, Checker, "Anti-pattern in the argument",
254 "C String API", os.str(), Loc,
255 LenArg->getSourceRange());
259 // Recurse and check children.
263 void WalkAST::VisitChildren(Stmt *S) {
264 for (Stmt *Child : S->children())
270 class CStringSyntaxChecker: public Checker<check::ASTCodeBody> {
273 void checkASTCodeBody(const Decl *D, AnalysisManager& Mgr,
274 BugReporter &BR) const {
275 WalkAST walker(this, BR, Mgr.getAnalysisDeclContext(D));
276 walker.Visit(D->getBody());
281 void ento::registerCStringSyntaxChecker(CheckerManager &mgr) {
282 mgr.registerChecker<CStringSyntaxChecker>();
285 bool ento::shouldRegisterCStringSyntaxChecker(const LangOptions &LO) {