xref: /freebsd/contrib/llvm-project/clang/lib/StaticAnalyzer/Checkers/LLVMConventionsChecker.cpp (revision 81ad626541db97eb356e2c1d4a20eb2a26a766ab)
10b57cec5SDimitry Andric //=== LLVMConventionsChecker.cpp - Check LLVM codebase conventions ---*- C++ -*-
20b57cec5SDimitry Andric //
30b57cec5SDimitry Andric // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
40b57cec5SDimitry Andric // See https://llvm.org/LICENSE.txt for license information.
50b57cec5SDimitry Andric // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
60b57cec5SDimitry Andric //
70b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
80b57cec5SDimitry Andric //
90b57cec5SDimitry Andric // This defines LLVMConventionsChecker, a bunch of small little checks
100b57cec5SDimitry Andric // for checking specific coding conventions in the LLVM/Clang codebase.
110b57cec5SDimitry Andric //
120b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
130b57cec5SDimitry Andric 
140b57cec5SDimitry Andric #include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
150b57cec5SDimitry Andric #include "clang/AST/DeclTemplate.h"
160b57cec5SDimitry Andric #include "clang/AST/StmtVisitor.h"
170b57cec5SDimitry Andric #include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.h"
180b57cec5SDimitry Andric #include "clang/StaticAnalyzer/Core/Checker.h"
190b57cec5SDimitry Andric #include "llvm/ADT/SmallString.h"
200b57cec5SDimitry Andric #include "llvm/Support/raw_ostream.h"
210b57cec5SDimitry Andric 
220b57cec5SDimitry Andric using namespace clang;
230b57cec5SDimitry Andric using namespace ento;
240b57cec5SDimitry Andric 
250b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
260b57cec5SDimitry Andric // Generic type checking routines.
270b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
280b57cec5SDimitry Andric 
290b57cec5SDimitry Andric static bool IsLLVMStringRef(QualType T) {
300b57cec5SDimitry Andric   const RecordType *RT = T->getAs<RecordType>();
310b57cec5SDimitry Andric   if (!RT)
320b57cec5SDimitry Andric     return false;
330b57cec5SDimitry Andric 
340b57cec5SDimitry Andric   return StringRef(QualType(RT, 0).getAsString()) == "class StringRef";
350b57cec5SDimitry Andric }
360b57cec5SDimitry Andric 
370b57cec5SDimitry Andric /// Check whether the declaration is semantically inside the top-level
380b57cec5SDimitry Andric /// namespace named by ns.
390b57cec5SDimitry Andric static bool InNamespace(const Decl *D, StringRef NS) {
400b57cec5SDimitry Andric   const NamespaceDecl *ND = dyn_cast<NamespaceDecl>(D->getDeclContext());
410b57cec5SDimitry Andric   if (!ND)
420b57cec5SDimitry Andric     return false;
430b57cec5SDimitry Andric   const IdentifierInfo *II = ND->getIdentifier();
440b57cec5SDimitry Andric   if (!II || !II->getName().equals(NS))
450b57cec5SDimitry Andric     return false;
460b57cec5SDimitry Andric   return isa<TranslationUnitDecl>(ND->getDeclContext());
470b57cec5SDimitry Andric }
480b57cec5SDimitry Andric 
490b57cec5SDimitry Andric static bool IsStdString(QualType T) {
500b57cec5SDimitry Andric   if (const ElaboratedType *QT = T->getAs<ElaboratedType>())
510b57cec5SDimitry Andric     T = QT->getNamedType();
520b57cec5SDimitry Andric 
530b57cec5SDimitry Andric   const TypedefType *TT = T->getAs<TypedefType>();
540b57cec5SDimitry Andric   if (!TT)
550b57cec5SDimitry Andric     return false;
560b57cec5SDimitry Andric 
570b57cec5SDimitry Andric   const TypedefNameDecl *TD = TT->getDecl();
580b57cec5SDimitry Andric 
590b57cec5SDimitry Andric   if (!TD->isInStdNamespace())
600b57cec5SDimitry Andric     return false;
610b57cec5SDimitry Andric 
620b57cec5SDimitry Andric   return TD->getName() == "string";
630b57cec5SDimitry Andric }
640b57cec5SDimitry Andric 
650b57cec5SDimitry Andric static bool IsClangType(const RecordDecl *RD) {
660b57cec5SDimitry Andric   return RD->getName() == "Type" && InNamespace(RD, "clang");
670b57cec5SDimitry Andric }
680b57cec5SDimitry Andric 
690b57cec5SDimitry Andric static bool IsClangDecl(const RecordDecl *RD) {
700b57cec5SDimitry Andric   return RD->getName() == "Decl" && InNamespace(RD, "clang");
710b57cec5SDimitry Andric }
720b57cec5SDimitry Andric 
730b57cec5SDimitry Andric static bool IsClangStmt(const RecordDecl *RD) {
740b57cec5SDimitry Andric   return RD->getName() == "Stmt" && InNamespace(RD, "clang");
750b57cec5SDimitry Andric }
760b57cec5SDimitry Andric 
770b57cec5SDimitry Andric static bool IsClangAttr(const RecordDecl *RD) {
780b57cec5SDimitry Andric   return RD->getName() == "Attr" && InNamespace(RD, "clang");
790b57cec5SDimitry Andric }
800b57cec5SDimitry Andric 
810b57cec5SDimitry Andric static bool IsStdVector(QualType T) {
820b57cec5SDimitry Andric   const TemplateSpecializationType *TS = T->getAs<TemplateSpecializationType>();
830b57cec5SDimitry Andric   if (!TS)
840b57cec5SDimitry Andric     return false;
850b57cec5SDimitry Andric 
860b57cec5SDimitry Andric   TemplateName TM = TS->getTemplateName();
870b57cec5SDimitry Andric   TemplateDecl *TD = TM.getAsTemplateDecl();
880b57cec5SDimitry Andric 
890b57cec5SDimitry Andric   if (!TD || !InNamespace(TD, "std"))
900b57cec5SDimitry Andric     return false;
910b57cec5SDimitry Andric 
920b57cec5SDimitry Andric   return TD->getName() == "vector";
930b57cec5SDimitry Andric }
940b57cec5SDimitry Andric 
950b57cec5SDimitry Andric static bool IsSmallVector(QualType T) {
960b57cec5SDimitry Andric   const TemplateSpecializationType *TS = T->getAs<TemplateSpecializationType>();
970b57cec5SDimitry Andric   if (!TS)
980b57cec5SDimitry Andric     return false;
990b57cec5SDimitry Andric 
1000b57cec5SDimitry Andric   TemplateName TM = TS->getTemplateName();
1010b57cec5SDimitry Andric   TemplateDecl *TD = TM.getAsTemplateDecl();
1020b57cec5SDimitry Andric 
1030b57cec5SDimitry Andric   if (!TD || !InNamespace(TD, "llvm"))
1040b57cec5SDimitry Andric     return false;
1050b57cec5SDimitry Andric 
1060b57cec5SDimitry Andric   return TD->getName() == "SmallVector";
1070b57cec5SDimitry Andric }
1080b57cec5SDimitry Andric 
1090b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
1100b57cec5SDimitry Andric // CHECK: a StringRef should not be bound to a temporary std::string whose
1110b57cec5SDimitry Andric // lifetime is shorter than the StringRef's.
1120b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
1130b57cec5SDimitry Andric 
1140b57cec5SDimitry Andric namespace {
1150b57cec5SDimitry Andric class StringRefCheckerVisitor : public StmtVisitor<StringRefCheckerVisitor> {
1160b57cec5SDimitry Andric   const Decl *DeclWithIssue;
1170b57cec5SDimitry Andric   BugReporter &BR;
1180b57cec5SDimitry Andric   const CheckerBase *Checker;
1190b57cec5SDimitry Andric 
1200b57cec5SDimitry Andric public:
1210b57cec5SDimitry Andric   StringRefCheckerVisitor(const Decl *declWithIssue, BugReporter &br,
1220b57cec5SDimitry Andric                           const CheckerBase *checker)
1230b57cec5SDimitry Andric       : DeclWithIssue(declWithIssue), BR(br), Checker(checker) {}
1240b57cec5SDimitry Andric   void VisitChildren(Stmt *S) {
1250b57cec5SDimitry Andric     for (Stmt *Child : S->children())
1260b57cec5SDimitry Andric       if (Child)
1270b57cec5SDimitry Andric         Visit(Child);
1280b57cec5SDimitry Andric   }
1290b57cec5SDimitry Andric   void VisitStmt(Stmt *S) { VisitChildren(S); }
1300b57cec5SDimitry Andric   void VisitDeclStmt(DeclStmt *DS);
1310b57cec5SDimitry Andric private:
1320b57cec5SDimitry Andric   void VisitVarDecl(VarDecl *VD);
1330b57cec5SDimitry Andric };
1340b57cec5SDimitry Andric } // end anonymous namespace
1350b57cec5SDimitry Andric 
1360b57cec5SDimitry Andric static void CheckStringRefAssignedTemporary(const Decl *D, BugReporter &BR,
1370b57cec5SDimitry Andric                                             const CheckerBase *Checker) {
1380b57cec5SDimitry Andric   StringRefCheckerVisitor walker(D, BR, Checker);
1390b57cec5SDimitry Andric   walker.Visit(D->getBody());
1400b57cec5SDimitry Andric }
1410b57cec5SDimitry Andric 
1420b57cec5SDimitry Andric void StringRefCheckerVisitor::VisitDeclStmt(DeclStmt *S) {
1430b57cec5SDimitry Andric   VisitChildren(S);
1440b57cec5SDimitry Andric 
1450b57cec5SDimitry Andric   for (auto *I : S->decls())
1460b57cec5SDimitry Andric     if (VarDecl *VD = dyn_cast<VarDecl>(I))
1470b57cec5SDimitry Andric       VisitVarDecl(VD);
1480b57cec5SDimitry Andric }
1490b57cec5SDimitry Andric 
1500b57cec5SDimitry Andric void StringRefCheckerVisitor::VisitVarDecl(VarDecl *VD) {
1510b57cec5SDimitry Andric   Expr *Init = VD->getInit();
1520b57cec5SDimitry Andric   if (!Init)
1530b57cec5SDimitry Andric     return;
1540b57cec5SDimitry Andric 
1550b57cec5SDimitry Andric   // Pattern match for:
1560b57cec5SDimitry Andric   // StringRef x = call() (where call returns std::string)
1570b57cec5SDimitry Andric   if (!IsLLVMStringRef(VD->getType()))
1580b57cec5SDimitry Andric     return;
1590b57cec5SDimitry Andric   ExprWithCleanups *Ex1 = dyn_cast<ExprWithCleanups>(Init);
1600b57cec5SDimitry Andric   if (!Ex1)
1610b57cec5SDimitry Andric     return;
1620b57cec5SDimitry Andric   CXXConstructExpr *Ex2 = dyn_cast<CXXConstructExpr>(Ex1->getSubExpr());
1630b57cec5SDimitry Andric   if (!Ex2 || Ex2->getNumArgs() != 1)
1640b57cec5SDimitry Andric     return;
1650b57cec5SDimitry Andric   ImplicitCastExpr *Ex3 = dyn_cast<ImplicitCastExpr>(Ex2->getArg(0));
1660b57cec5SDimitry Andric   if (!Ex3)
1670b57cec5SDimitry Andric     return;
1680b57cec5SDimitry Andric   CXXConstructExpr *Ex4 = dyn_cast<CXXConstructExpr>(Ex3->getSubExpr());
1690b57cec5SDimitry Andric   if (!Ex4 || Ex4->getNumArgs() != 1)
1700b57cec5SDimitry Andric     return;
1710b57cec5SDimitry Andric   ImplicitCastExpr *Ex5 = dyn_cast<ImplicitCastExpr>(Ex4->getArg(0));
1720b57cec5SDimitry Andric   if (!Ex5)
1730b57cec5SDimitry Andric     return;
1740b57cec5SDimitry Andric   CXXBindTemporaryExpr *Ex6 = dyn_cast<CXXBindTemporaryExpr>(Ex5->getSubExpr());
1750b57cec5SDimitry Andric   if (!Ex6 || !IsStdString(Ex6->getType()))
1760b57cec5SDimitry Andric     return;
1770b57cec5SDimitry Andric 
1780b57cec5SDimitry Andric   // Okay, badness!  Report an error.
1790b57cec5SDimitry Andric   const char *desc = "StringRef should not be bound to temporary "
1800b57cec5SDimitry Andric                      "std::string that it outlives";
1810b57cec5SDimitry Andric   PathDiagnosticLocation VDLoc =
1820b57cec5SDimitry Andric     PathDiagnosticLocation::createBegin(VD, BR.getSourceManager());
1830b57cec5SDimitry Andric   BR.EmitBasicReport(DeclWithIssue, Checker, desc, "LLVM Conventions", desc,
1840b57cec5SDimitry Andric                      VDLoc, Init->getSourceRange());
1850b57cec5SDimitry Andric }
1860b57cec5SDimitry Andric 
1870b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
1880b57cec5SDimitry Andric // CHECK: Clang AST nodes should not have fields that can allocate
1890b57cec5SDimitry Andric //   memory.
1900b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
1910b57cec5SDimitry Andric 
1920b57cec5SDimitry Andric static bool AllocatesMemory(QualType T) {
1930b57cec5SDimitry Andric   return IsStdVector(T) || IsStdString(T) || IsSmallVector(T);
1940b57cec5SDimitry Andric }
1950b57cec5SDimitry Andric 
1960b57cec5SDimitry Andric // This type checking could be sped up via dynamic programming.
1970b57cec5SDimitry Andric static bool IsPartOfAST(const CXXRecordDecl *R) {
1980b57cec5SDimitry Andric   if (IsClangStmt(R) || IsClangType(R) || IsClangDecl(R) || IsClangAttr(R))
1990b57cec5SDimitry Andric     return true;
2000b57cec5SDimitry Andric 
2010b57cec5SDimitry Andric   for (const auto &BS : R->bases()) {
2020b57cec5SDimitry Andric     QualType T = BS.getType();
2030b57cec5SDimitry Andric     if (const RecordType *baseT = T->getAs<RecordType>()) {
2040b57cec5SDimitry Andric       CXXRecordDecl *baseD = cast<CXXRecordDecl>(baseT->getDecl());
2050b57cec5SDimitry Andric       if (IsPartOfAST(baseD))
2060b57cec5SDimitry Andric         return true;
2070b57cec5SDimitry Andric     }
2080b57cec5SDimitry Andric   }
2090b57cec5SDimitry Andric 
2100b57cec5SDimitry Andric   return false;
2110b57cec5SDimitry Andric }
2120b57cec5SDimitry Andric 
2130b57cec5SDimitry Andric namespace {
2140b57cec5SDimitry Andric class ASTFieldVisitor {
2150b57cec5SDimitry Andric   SmallVector<FieldDecl*, 10> FieldChain;
2160b57cec5SDimitry Andric   const CXXRecordDecl *Root;
2170b57cec5SDimitry Andric   BugReporter &BR;
2180b57cec5SDimitry Andric   const CheckerBase *Checker;
2190b57cec5SDimitry Andric 
2200b57cec5SDimitry Andric public:
2210b57cec5SDimitry Andric   ASTFieldVisitor(const CXXRecordDecl *root, BugReporter &br,
2220b57cec5SDimitry Andric                   const CheckerBase *checker)
2230b57cec5SDimitry Andric       : Root(root), BR(br), Checker(checker) {}
2240b57cec5SDimitry Andric 
2250b57cec5SDimitry Andric   void Visit(FieldDecl *D);
2260b57cec5SDimitry Andric   void ReportError(QualType T);
2270b57cec5SDimitry Andric };
2280b57cec5SDimitry Andric } // end anonymous namespace
2290b57cec5SDimitry Andric 
2300b57cec5SDimitry Andric static void CheckASTMemory(const CXXRecordDecl *R, BugReporter &BR,
2310b57cec5SDimitry Andric                            const CheckerBase *Checker) {
2320b57cec5SDimitry Andric   if (!IsPartOfAST(R))
2330b57cec5SDimitry Andric     return;
2340b57cec5SDimitry Andric 
2350b57cec5SDimitry Andric   for (auto *I : R->fields()) {
2360b57cec5SDimitry Andric     ASTFieldVisitor walker(R, BR, Checker);
2370b57cec5SDimitry Andric     walker.Visit(I);
2380b57cec5SDimitry Andric   }
2390b57cec5SDimitry Andric }
2400b57cec5SDimitry Andric 
2410b57cec5SDimitry Andric void ASTFieldVisitor::Visit(FieldDecl *D) {
2420b57cec5SDimitry Andric   FieldChain.push_back(D);
2430b57cec5SDimitry Andric 
2440b57cec5SDimitry Andric   QualType T = D->getType();
2450b57cec5SDimitry Andric 
2460b57cec5SDimitry Andric   if (AllocatesMemory(T))
2470b57cec5SDimitry Andric     ReportError(T);
2480b57cec5SDimitry Andric 
2490b57cec5SDimitry Andric   if (const RecordType *RT = T->getAs<RecordType>()) {
2500b57cec5SDimitry Andric     const RecordDecl *RD = RT->getDecl()->getDefinition();
2510b57cec5SDimitry Andric     for (auto *I : RD->fields())
2520b57cec5SDimitry Andric       Visit(I);
2530b57cec5SDimitry Andric   }
2540b57cec5SDimitry Andric 
2550b57cec5SDimitry Andric   FieldChain.pop_back();
2560b57cec5SDimitry Andric }
2570b57cec5SDimitry Andric 
2580b57cec5SDimitry Andric void ASTFieldVisitor::ReportError(QualType T) {
2590b57cec5SDimitry Andric   SmallString<1024> buf;
2600b57cec5SDimitry Andric   llvm::raw_svector_ostream os(buf);
2610b57cec5SDimitry Andric 
2620b57cec5SDimitry Andric   os << "AST class '" << Root->getName() << "' has a field '"
2630b57cec5SDimitry Andric      << FieldChain.front()->getName() << "' that allocates heap memory";
2640b57cec5SDimitry Andric   if (FieldChain.size() > 1) {
2650b57cec5SDimitry Andric     os << " via the following chain: ";
2660b57cec5SDimitry Andric     bool isFirst = true;
2670b57cec5SDimitry Andric     for (SmallVectorImpl<FieldDecl*>::iterator I=FieldChain.begin(),
2680b57cec5SDimitry Andric          E=FieldChain.end(); I!=E; ++I) {
2690b57cec5SDimitry Andric       if (!isFirst)
2700b57cec5SDimitry Andric         os << '.';
2710b57cec5SDimitry Andric       else
2720b57cec5SDimitry Andric         isFirst = false;
2730b57cec5SDimitry Andric       os << (*I)->getName();
2740b57cec5SDimitry Andric     }
2750b57cec5SDimitry Andric   }
276*81ad6265SDimitry Andric   os << " (type " << FieldChain.back()->getType() << ")";
2770b57cec5SDimitry Andric 
2780b57cec5SDimitry Andric   // Note that this will fire for every translation unit that uses this
2790b57cec5SDimitry Andric   // class.  This is suboptimal, but at least scan-build will merge
2800b57cec5SDimitry Andric   // duplicate HTML reports.  In the future we need a unified way of merging
2810b57cec5SDimitry Andric   // duplicate reports across translation units.  For C++ classes we cannot
2820b57cec5SDimitry Andric   // just report warnings when we see an out-of-line method definition for a
2830b57cec5SDimitry Andric   // class, as that heuristic doesn't always work (the complete definition of
2840b57cec5SDimitry Andric   // the class may be in the header file, for example).
2850b57cec5SDimitry Andric   PathDiagnosticLocation L = PathDiagnosticLocation::createBegin(
2860b57cec5SDimitry Andric                                FieldChain.front(), BR.getSourceManager());
2870b57cec5SDimitry Andric   BR.EmitBasicReport(Root, Checker, "AST node allocates heap memory",
2880b57cec5SDimitry Andric                      "LLVM Conventions", os.str(), L);
2890b57cec5SDimitry Andric }
2900b57cec5SDimitry Andric 
2910b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
2920b57cec5SDimitry Andric // LLVMConventionsChecker
2930b57cec5SDimitry Andric //===----------------------------------------------------------------------===//
2940b57cec5SDimitry Andric 
2950b57cec5SDimitry Andric namespace {
2960b57cec5SDimitry Andric class LLVMConventionsChecker : public Checker<
2970b57cec5SDimitry Andric                                                 check::ASTDecl<CXXRecordDecl>,
2980b57cec5SDimitry Andric                                                 check::ASTCodeBody > {
2990b57cec5SDimitry Andric public:
3000b57cec5SDimitry Andric   void checkASTDecl(const CXXRecordDecl *R, AnalysisManager& mgr,
3010b57cec5SDimitry Andric                     BugReporter &BR) const {
3020b57cec5SDimitry Andric     if (R->isCompleteDefinition())
3030b57cec5SDimitry Andric       CheckASTMemory(R, BR, this);
3040b57cec5SDimitry Andric   }
3050b57cec5SDimitry Andric 
3060b57cec5SDimitry Andric   void checkASTCodeBody(const Decl *D, AnalysisManager& mgr,
3070b57cec5SDimitry Andric                         BugReporter &BR) const {
3080b57cec5SDimitry Andric     CheckStringRefAssignedTemporary(D, BR, this);
3090b57cec5SDimitry Andric   }
3100b57cec5SDimitry Andric };
3110b57cec5SDimitry Andric }
3120b57cec5SDimitry Andric 
3130b57cec5SDimitry Andric void ento::registerLLVMConventionsChecker(CheckerManager &mgr) {
3140b57cec5SDimitry Andric   mgr.registerChecker<LLVMConventionsChecker>();
3150b57cec5SDimitry Andric }
3160b57cec5SDimitry Andric 
3175ffd83dbSDimitry Andric bool ento::shouldRegisterLLVMConventionsChecker(const CheckerManager &mgr) {
3180b57cec5SDimitry Andric   return true;
3190b57cec5SDimitry Andric }
320