| 1 | //=== ConversionChecker.cpp -------------------------------------*- C++ -*-===// |
| 2 | // |
| 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 |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | // |
| 9 | // Check that there is no loss of sign/precision in assignments, comparisons |
| 10 | // and multiplications. |
| 11 | // |
| 12 | // ConversionChecker uses path sensitive analysis to determine possible values |
| 13 | // of expressions. A warning is reported when: |
| 14 | // * a negative value is implicitly converted to an unsigned value in an |
| 15 | // assignment, comparison or multiplication. |
| 16 | // * assignment / initialization when the source value is greater than the max |
| 17 | // value of the target integer type |
| 18 | // * assignment / initialization when the source integer is above the range |
| 19 | // where the target floating point type can represent all integers |
| 20 | // |
| 21 | // Many compilers and tools have similar checks that are based on semantic |
| 22 | // analysis. Those checks are sound but have poor precision. ConversionChecker |
| 23 | // is an alternative to those checks. |
| 24 | // |
| 25 | //===----------------------------------------------------------------------===// |
| 26 | #include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h" |
| 27 | #include "clang/AST/ParentMap.h" |
| 28 | #include "clang/StaticAnalyzer/Core/BugReporter/BugType.h" |
| 29 | #include "clang/StaticAnalyzer/Core/Checker.h" |
| 30 | #include "clang/StaticAnalyzer/Core/CheckerManager.h" |
| 31 | #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h" |
| 32 | #include "llvm/ADT/APFloat.h" |
| 33 | |
| 34 | #include <climits> |
| 35 | |
| 36 | using namespace clang; |
| 37 | using namespace ento; |
| 38 | |
| 39 | namespace { |
| 40 | class ConversionChecker : public Checker<check::PreStmt<ImplicitCastExpr>> { |
| 41 | public: |
| 42 | void checkPreStmt(const ImplicitCastExpr *Cast, CheckerContext &C) const; |
| 43 | |
| 44 | private: |
| 45 | const BugType BT{this, "Conversion" }; |
| 46 | |
| 47 | bool isLossOfPrecision(const ImplicitCastExpr *Cast, QualType DestType, |
| 48 | CheckerContext &C) const; |
| 49 | |
| 50 | bool isLossOfSign(const ImplicitCastExpr *Cast, CheckerContext &C) const; |
| 51 | |
| 52 | void reportBug(ExplodedNode *N, const Expr *E, CheckerContext &C, |
| 53 | const char Msg[]) const; |
| 54 | }; |
| 55 | } |
| 56 | |
| 57 | void ConversionChecker::checkPreStmt(const ImplicitCastExpr *Cast, |
| 58 | CheckerContext &C) const { |
| 59 | // Don't warn for implicit conversions to bool |
| 60 | if (Cast->getType()->isBooleanType()) |
| 61 | return; |
| 62 | |
| 63 | // Don't warn for loss of sign/precision in macros. |
| 64 | if (Cast->getExprLoc().isMacroID()) |
| 65 | return; |
| 66 | |
| 67 | // Get Parent. |
| 68 | const ParentMap &PM = C.getLocationContext()->getParentMap(); |
| 69 | const Stmt *Parent = PM.getParent(Cast); |
| 70 | if (!Parent) |
| 71 | return; |
| 72 | // Dont warn if this is part of an explicit cast |
| 73 | if (isa<ExplicitCastExpr>(Val: Parent)) |
| 74 | return; |
| 75 | |
| 76 | bool LossOfSign = false; |
| 77 | bool LossOfPrecision = false; |
| 78 | |
| 79 | // Loss of sign/precision in binary operation. |
| 80 | if (const auto *B = dyn_cast<BinaryOperator>(Parent)) { |
| 81 | BinaryOperator::Opcode Opc = B->getOpcode(); |
| 82 | if (Opc == BO_Assign) { |
| 83 | if (!Cast->IgnoreParenImpCasts()->isEvaluatable(C.getASTContext())) { |
| 84 | LossOfSign = isLossOfSign(Cast, C); |
| 85 | LossOfPrecision = isLossOfPrecision(Cast, DestType: Cast->getType(), C); |
| 86 | } |
| 87 | } else if (Opc == BO_AddAssign || Opc == BO_SubAssign) { |
| 88 | // No loss of sign. |
| 89 | LossOfPrecision = isLossOfPrecision(Cast, DestType: B->getLHS()->getType(), C); |
| 90 | } else if (Opc == BO_MulAssign) { |
| 91 | LossOfSign = isLossOfSign(Cast, C); |
| 92 | LossOfPrecision = isLossOfPrecision(Cast, DestType: B->getLHS()->getType(), C); |
| 93 | } else if (Opc == BO_DivAssign || Opc == BO_RemAssign) { |
| 94 | LossOfSign = isLossOfSign(Cast, C); |
| 95 | // No loss of precision. |
| 96 | } else if (Opc == BO_AndAssign) { |
| 97 | LossOfSign = isLossOfSign(Cast, C); |
| 98 | // No loss of precision. |
| 99 | } else if (Opc == BO_OrAssign || Opc == BO_XorAssign) { |
| 100 | LossOfSign = isLossOfSign(Cast, C); |
| 101 | LossOfPrecision = isLossOfPrecision(Cast, DestType: B->getLHS()->getType(), C); |
| 102 | } else if (B->isRelationalOp() || B->isMultiplicativeOp()) { |
| 103 | LossOfSign = isLossOfSign(Cast, C); |
| 104 | } |
| 105 | } else if (isa<DeclStmt, ReturnStmt>(Val: Parent)) { |
| 106 | if (!Cast->IgnoreParenImpCasts()->isEvaluatable(C.getASTContext())) { |
| 107 | LossOfSign = isLossOfSign(Cast, C); |
| 108 | LossOfPrecision = isLossOfPrecision(Cast, DestType: Cast->getType(), C); |
| 109 | } |
| 110 | } else { |
| 111 | LossOfSign = isLossOfSign(Cast, C); |
| 112 | LossOfPrecision = isLossOfPrecision(Cast, DestType: Cast->getType(), C); |
| 113 | } |
| 114 | |
| 115 | if (LossOfSign || LossOfPrecision) { |
| 116 | // Generate an error node. |
| 117 | ExplodedNode *N = C.generateNonFatalErrorNode(State: C.getState()); |
| 118 | if (!N) |
| 119 | return; |
| 120 | if (LossOfSign) |
| 121 | reportBug(N, Cast, C, "Loss of sign in implicit conversion" ); |
| 122 | if (LossOfPrecision) |
| 123 | reportBug(N, Cast, C, "Loss of precision in implicit conversion" ); |
| 124 | } |
| 125 | } |
| 126 | |
| 127 | void ConversionChecker::reportBug(ExplodedNode *N, const Expr *E, |
| 128 | CheckerContext &C, const char Msg[]) const { |
| 129 | // Generate a report for this bug. |
| 130 | auto R = std::make_unique<PathSensitiveBugReport>(args: BT, args&: Msg, args&: N); |
| 131 | bugreporter::trackExpressionValue(N, E, R&: *R); |
| 132 | C.emitReport(R: std::move(R)); |
| 133 | } |
| 134 | |
| 135 | bool ConversionChecker::isLossOfPrecision(const ImplicitCastExpr *Cast, |
| 136 | QualType DestType, |
| 137 | CheckerContext &C) const { |
| 138 | // Don't warn about explicit loss of precision. |
| 139 | if (Cast->isEvaluatable(C.getASTContext())) |
| 140 | return false; |
| 141 | |
| 142 | QualType SubType = Cast->IgnoreParenImpCasts()->getType(); |
| 143 | |
| 144 | if (!DestType->isRealType() || !SubType->isIntegerType()) |
| 145 | return false; |
| 146 | |
| 147 | const bool isFloat = DestType->isFloatingType(); |
| 148 | |
| 149 | const auto &AC = C.getASTContext(); |
| 150 | |
| 151 | // We will find the largest RepresentsUntilExp value such that the DestType |
| 152 | // can exactly represent all nonnegative integers below 2^RepresentsUntilExp. |
| 153 | unsigned RepresentsUntilExp; |
| 154 | |
| 155 | if (isFloat) { |
| 156 | const llvm::fltSemantics &Sema = AC.getFloatTypeSemantics(T: DestType); |
| 157 | RepresentsUntilExp = llvm::APFloat::semanticsPrecision(Sema); |
| 158 | } else { |
| 159 | RepresentsUntilExp = AC.getIntWidth(T: DestType); |
| 160 | if (RepresentsUntilExp == 1) { |
| 161 | // This is just casting a number to bool, probably not a bug. |
| 162 | return false; |
| 163 | } |
| 164 | if (DestType->isSignedIntegerType()) |
| 165 | RepresentsUntilExp--; |
| 166 | } |
| 167 | |
| 168 | if (RepresentsUntilExp >= sizeof(unsigned long long) * CHAR_BIT) { |
| 169 | // Avoid overflow in our later calculations. |
| 170 | return false; |
| 171 | } |
| 172 | |
| 173 | unsigned CorrectedSrcWidth = AC.getIntWidth(T: SubType); |
| 174 | if (SubType->isSignedIntegerType()) |
| 175 | CorrectedSrcWidth--; |
| 176 | |
| 177 | if (RepresentsUntilExp >= CorrectedSrcWidth) { |
| 178 | // Simple case: the destination can store all values of the source type. |
| 179 | return false; |
| 180 | } |
| 181 | |
| 182 | unsigned long long MaxVal = 1ULL << RepresentsUntilExp; |
| 183 | if (isFloat) { |
| 184 | // If this is a floating point type, it can also represent MaxVal exactly. |
| 185 | MaxVal++; |
| 186 | } |
| 187 | return C.isGreaterOrEqual(E: Cast->getSubExpr(), Val: MaxVal); |
| 188 | // TODO: maybe also check negative values with too large magnitude. |
| 189 | } |
| 190 | |
| 191 | bool ConversionChecker::isLossOfSign(const ImplicitCastExpr *Cast, |
| 192 | CheckerContext &C) const { |
| 193 | QualType CastType = Cast->getType(); |
| 194 | QualType SubType = Cast->IgnoreParenImpCasts()->getType(); |
| 195 | |
| 196 | if (!CastType->isUnsignedIntegerType() || !SubType->isSignedIntegerType()) |
| 197 | return false; |
| 198 | |
| 199 | return C.isNegative(E: Cast->getSubExpr()); |
| 200 | } |
| 201 | |
| 202 | void ento::registerConversionChecker(CheckerManager &mgr) { |
| 203 | mgr.registerChecker<ConversionChecker>(); |
| 204 | } |
| 205 | |
| 206 | bool ento::shouldRegisterConversionChecker(const CheckerManager &mgr) { |
| 207 | return true; |
| 208 | } |
| 209 | |