| 1 | //===-- SVals.cpp - Abstract RValues for Path-Sens. Value Tracking --------===// |
| 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 | // This file defines SVal, Loc, and NonLoc, classes that represent |
| 10 | // abstract r-values for use with path-sensitive value tracking. |
| 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
| 14 | #include "clang/StaticAnalyzer/Core/PathSensitive/SVals.h" |
| 15 | #include "clang/AST/ASTContext.h" |
| 16 | #include "clang/AST/Decl.h" |
| 17 | #include "clang/AST/DeclCXX.h" |
| 18 | #include "clang/AST/Expr.h" |
| 19 | #include "clang/AST/Type.h" |
| 20 | #include "clang/Basic/JsonSupport.h" |
| 21 | #include "clang/Basic/LLVM.h" |
| 22 | #include "clang/StaticAnalyzer/Core/PathSensitive/BasicValueFactory.h" |
| 23 | #include "clang/StaticAnalyzer/Core/PathSensitive/MemRegion.h" |
| 24 | #include "clang/StaticAnalyzer/Core/PathSensitive/SValBuilder.h" |
| 25 | #include "clang/StaticAnalyzer/Core/PathSensitive/SValVisitor.h" |
| 26 | #include "clang/StaticAnalyzer/Core/PathSensitive/SymExpr.h" |
| 27 | #include "clang/StaticAnalyzer/Core/PathSensitive/SymbolManager.h" |
| 28 | #include "llvm/Support/Compiler.h" |
| 29 | #include "llvm/Support/ErrorHandling.h" |
| 30 | #include "llvm/Support/raw_ostream.h" |
| 31 | #include <cassert> |
| 32 | #include <optional> |
| 33 | |
| 34 | using namespace clang; |
| 35 | using namespace ento; |
| 36 | |
| 37 | //===----------------------------------------------------------------------===// |
| 38 | // Symbol iteration within an SVal. |
| 39 | //===----------------------------------------------------------------------===// |
| 40 | |
| 41 | //===----------------------------------------------------------------------===// |
| 42 | // Utility methods. |
| 43 | //===----------------------------------------------------------------------===// |
| 44 | |
| 45 | const FunctionDecl *SVal::getAsFunctionDecl() const { |
| 46 | if (std::optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>()) { |
| 47 | const MemRegion* R = X->getRegion(); |
| 48 | if (const FunctionCodeRegion *CTR = R->getAs<FunctionCodeRegion>()) |
| 49 | if (const auto *FD = dyn_cast<FunctionDecl>(Val: CTR->getDecl())) |
| 50 | return FD; |
| 51 | } |
| 52 | |
| 53 | if (auto X = getAs<nonloc::PointerToMember>()) { |
| 54 | if (const auto *MD = dyn_cast_or_null<CXXMethodDecl>(Val: X->getDecl())) |
| 55 | return MD; |
| 56 | } |
| 57 | return nullptr; |
| 58 | } |
| 59 | |
| 60 | /// If this SVal is a location (subclasses Loc) and wraps a symbol, |
| 61 | /// return that SymbolRef. Otherwise return 0. |
| 62 | /// |
| 63 | /// Implicit casts (ex: void* -> char*) can turn Symbolic region into Element |
| 64 | /// region. If that is the case, gets the underlining region. |
| 65 | /// When IncludeBaseRegions is set to true and the SubRegion is non-symbolic, |
| 66 | /// the first symbolic parent region is returned. |
| 67 | SymbolRef SVal::getAsLocSymbol(bool IncludeBaseRegions) const { |
| 68 | // FIXME: should we consider SymbolRef wrapped in CodeTextRegion? |
| 69 | if (const MemRegion *R = getAsRegion()) |
| 70 | if (const SymbolicRegion *SymR = |
| 71 | IncludeBaseRegions ? R->getSymbolicBase() |
| 72 | : dyn_cast<SymbolicRegion>(Val: R->StripCasts())) |
| 73 | return SymR->getSymbol(); |
| 74 | |
| 75 | return nullptr; |
| 76 | } |
| 77 | |
| 78 | /// Get the symbol in the SVal or its base region. |
| 79 | SymbolRef SVal::getLocSymbolInBase() const { |
| 80 | std::optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>(); |
| 81 | |
| 82 | if (!X) |
| 83 | return nullptr; |
| 84 | |
| 85 | const MemRegion *R = X->getRegion(); |
| 86 | |
| 87 | while (const auto *SR = dyn_cast<SubRegion>(Val: R)) { |
| 88 | if (const auto *SymR = dyn_cast<SymbolicRegion>(Val: SR)) |
| 89 | return SymR->getSymbol(); |
| 90 | else |
| 91 | R = SR->getSuperRegion(); |
| 92 | } |
| 93 | |
| 94 | return nullptr; |
| 95 | } |
| 96 | |
| 97 | /// If this SVal wraps a symbol return that SymbolRef. |
| 98 | /// Otherwise, return 0. |
| 99 | /// |
| 100 | /// Casts are ignored during lookup. |
| 101 | /// \param IncludeBaseRegions The boolean that controls whether the search |
| 102 | /// should continue to the base regions if the region is not symbolic. |
| 103 | SymbolRef SVal::getAsSymbol(bool IncludeBaseRegions) const { |
| 104 | // FIXME: should we consider SymbolRef wrapped in CodeTextRegion? |
| 105 | if (std::optional<nonloc::SymbolVal> X = getAs<nonloc::SymbolVal>()) |
| 106 | return X->getSymbol(); |
| 107 | |
| 108 | return getAsLocSymbol(IncludeBaseRegions); |
| 109 | } |
| 110 | |
| 111 | const llvm::APSInt *SVal::getAsInteger() const { |
| 112 | if (auto CI = getAs<nonloc::ConcreteInt>()) |
| 113 | return CI->getValue().get(); |
| 114 | if (auto CI = getAs<loc::ConcreteInt>()) |
| 115 | return CI->getValue().get(); |
| 116 | return nullptr; |
| 117 | } |
| 118 | |
| 119 | const MemRegion *SVal::getAsRegion() const { |
| 120 | if (std::optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>()) |
| 121 | return X->getRegion(); |
| 122 | |
| 123 | if (std::optional<nonloc::LocAsInteger> X = getAs<nonloc::LocAsInteger>()) |
| 124 | return X->getLoc().getAsRegion(); |
| 125 | |
| 126 | return nullptr; |
| 127 | } |
| 128 | |
| 129 | namespace { |
| 130 | class TypeRetrievingVisitor |
| 131 | : public FullSValVisitor<TypeRetrievingVisitor, QualType> { |
| 132 | private: |
| 133 | const ASTContext &Context; |
| 134 | |
| 135 | public: |
| 136 | TypeRetrievingVisitor(const ASTContext &Context) : Context(Context) {} |
| 137 | |
| 138 | QualType VisitMemRegionVal(loc::MemRegionVal MRV) { |
| 139 | return Visit(R: MRV.getRegion()); |
| 140 | } |
| 141 | QualType VisitGotoLabel(loc::GotoLabel GL) { |
| 142 | return QualType{Context.VoidPtrTy}; |
| 143 | } |
| 144 | template <class ConcreteInt> QualType VisitConcreteInt(ConcreteInt CI) { |
| 145 | const llvm::APSInt &Value = CI.getValue(); |
| 146 | if (1 == Value.getBitWidth()) |
| 147 | return Context.BoolTy; |
| 148 | return Context.getIntTypeForBitwidth(DestWidth: Value.getBitWidth(), Signed: Value.isSigned()); |
| 149 | } |
| 150 | QualType VisitLocAsInteger(nonloc::LocAsInteger LI) { |
| 151 | QualType NestedType = Visit(V: LI.getLoc()); |
| 152 | if (NestedType.isNull()) |
| 153 | return NestedType; |
| 154 | |
| 155 | return Context.getIntTypeForBitwidth(DestWidth: LI.getNumBits(), |
| 156 | Signed: NestedType->isSignedIntegerType()); |
| 157 | } |
| 158 | QualType VisitCompoundVal(nonloc::CompoundVal CV) { |
| 159 | return CV.getValue()->getType(); |
| 160 | } |
| 161 | QualType VisitLazyCompoundVal(nonloc::LazyCompoundVal LCV) { |
| 162 | return LCV.getRegion()->getValueType(); |
| 163 | } |
| 164 | QualType VisitSymbolVal(nonloc::SymbolVal SV) { |
| 165 | return Visit(S: SV.getSymbol()); |
| 166 | } |
| 167 | QualType VisitSymbolicRegion(const SymbolicRegion *SR) { |
| 168 | return Visit(S: SR->getSymbol()); |
| 169 | } |
| 170 | QualType VisitAllocaRegion(const AllocaRegion *) { |
| 171 | return QualType{Context.VoidPtrTy}; |
| 172 | } |
| 173 | QualType VisitTypedRegion(const TypedRegion *TR) { |
| 174 | return TR->getLocationType(); |
| 175 | } |
| 176 | QualType VisitSymExpr(const SymExpr *SE) { return SE->getType(); } |
| 177 | }; |
| 178 | } // end anonymous namespace |
| 179 | |
| 180 | QualType SVal::getType(const ASTContext &Context) const { |
| 181 | TypeRetrievingVisitor TRV{Context}; |
| 182 | return TRV.Visit(V: *this); |
| 183 | } |
| 184 | |
| 185 | const MemRegion *loc::MemRegionVal::stripCasts(bool StripBaseCasts) const { |
| 186 | return getRegion()->StripCasts(StripBaseAndDerivedCasts: StripBaseCasts); |
| 187 | } |
| 188 | |
| 189 | const void *nonloc::LazyCompoundVal::getStore() const { |
| 190 | return static_cast<const LazyCompoundValData*>(Data)->getStore(); |
| 191 | } |
| 192 | |
| 193 | const TypedValueRegion *nonloc::LazyCompoundVal::getRegion() const { |
| 194 | return static_cast<const LazyCompoundValData*>(Data)->getRegion(); |
| 195 | } |
| 196 | |
| 197 | bool nonloc::PointerToMember::isNullMemberPointer() const { |
| 198 | return getPTMData().isNull(); |
| 199 | } |
| 200 | |
| 201 | const NamedDecl *nonloc::PointerToMember::getDecl() const { |
| 202 | const auto PTMD = this->getPTMData(); |
| 203 | if (PTMD.isNull()) |
| 204 | return nullptr; |
| 205 | |
| 206 | const NamedDecl *ND = nullptr; |
| 207 | if (const auto *NDP = dyn_cast<const NamedDecl *>(Val: PTMD)) |
| 208 | ND = NDP; |
| 209 | else |
| 210 | ND = cast<const PointerToMemberData *>(Val: PTMD)->getDeclaratorDecl(); |
| 211 | |
| 212 | return ND; |
| 213 | } |
| 214 | |
| 215 | //===----------------------------------------------------------------------===// |
| 216 | // Other Iterators. |
| 217 | //===----------------------------------------------------------------------===// |
| 218 | |
| 219 | nonloc::CompoundVal::iterator nonloc::CompoundVal::begin() const { |
| 220 | return getValue()->begin(); |
| 221 | } |
| 222 | |
| 223 | nonloc::CompoundVal::iterator nonloc::CompoundVal::end() const { |
| 224 | return getValue()->end(); |
| 225 | } |
| 226 | |
| 227 | nonloc::PointerToMember::iterator nonloc::PointerToMember::begin() const { |
| 228 | const PTMDataType PTMD = getPTMData(); |
| 229 | if (isa<const NamedDecl *>(Val: PTMD)) |
| 230 | return {}; |
| 231 | return cast<const PointerToMemberData *>(Val: PTMD)->begin(); |
| 232 | } |
| 233 | |
| 234 | nonloc::PointerToMember::iterator nonloc::PointerToMember::end() const { |
| 235 | const PTMDataType PTMD = getPTMData(); |
| 236 | if (isa<const NamedDecl *>(Val: PTMD)) |
| 237 | return {}; |
| 238 | return cast<const PointerToMemberData *>(Val: PTMD)->end(); |
| 239 | } |
| 240 | |
| 241 | //===----------------------------------------------------------------------===// |
| 242 | // Useful predicates. |
| 243 | //===----------------------------------------------------------------------===// |
| 244 | |
| 245 | bool SVal::isConstant() const { |
| 246 | return getAs<nonloc::ConcreteInt>() || getAs<loc::ConcreteInt>(); |
| 247 | } |
| 248 | |
| 249 | bool SVal::isConstant(int I) const { |
| 250 | if (std::optional<loc::ConcreteInt> LV = getAs<loc::ConcreteInt>()) |
| 251 | return *LV->getValue() == I; |
| 252 | if (std::optional<nonloc::ConcreteInt> NV = getAs<nonloc::ConcreteInt>()) |
| 253 | return *NV->getValue() == I; |
| 254 | return false; |
| 255 | } |
| 256 | |
| 257 | bool SVal::isZeroConstant() const { |
| 258 | return isConstant(I: 0); |
| 259 | } |
| 260 | |
| 261 | //===----------------------------------------------------------------------===// |
| 262 | // Pretty-Printing. |
| 263 | //===----------------------------------------------------------------------===// |
| 264 | |
| 265 | StringRef SVal::getKindStr() const { |
| 266 | switch (getKind()) { |
| 267 | #define BASIC_SVAL(Id, Parent) \ |
| 268 | case Id##Kind: \ |
| 269 | return #Id; |
| 270 | #define LOC_SVAL(Id, Parent) \ |
| 271 | case Loc##Id##Kind: \ |
| 272 | return #Id; |
| 273 | #define NONLOC_SVAL(Id, Parent) \ |
| 274 | case NonLoc##Id##Kind: \ |
| 275 | return #Id; |
| 276 | #include "clang/StaticAnalyzer/Core/PathSensitive/SVals.def" |
| 277 | #undef REGION |
| 278 | } |
| 279 | llvm_unreachable("Unkown kind!" ); |
| 280 | } |
| 281 | |
| 282 | LLVM_DUMP_METHOD void SVal::dump() const { dumpToStream(OS&: llvm::errs()); } |
| 283 | |
| 284 | void SVal::printJson(raw_ostream &Out, bool AddQuotes) const { |
| 285 | std::string Buf; |
| 286 | llvm::raw_string_ostream TempOut(Buf); |
| 287 | |
| 288 | dumpToStream(OS&: TempOut); |
| 289 | |
| 290 | Out << JsonFormat(RawSR: Buf, AddQuotes); |
| 291 | } |
| 292 | |
| 293 | void SVal::dumpToStream(raw_ostream &os) const { |
| 294 | if (isUndef()) { |
| 295 | os << "Undefined" ; |
| 296 | return; |
| 297 | } |
| 298 | if (isUnknown()) { |
| 299 | os << "Unknown" ; |
| 300 | return; |
| 301 | } |
| 302 | if (NonLoc::classof(V: *this)) { |
| 303 | castAs<NonLoc>().dumpToStream(Out&: os); |
| 304 | return; |
| 305 | } |
| 306 | if (Loc::classof(V: *this)) { |
| 307 | castAs<Loc>().dumpToStream(Out&: os); |
| 308 | return; |
| 309 | } |
| 310 | llvm_unreachable("Unhandled SVal kind!" ); |
| 311 | } |
| 312 | |
| 313 | void NonLoc::dumpToStream(raw_ostream &os) const { |
| 314 | switch (getKind()) { |
| 315 | case nonloc::ConcreteIntKind: { |
| 316 | APSIntPtr Value = castAs<nonloc::ConcreteInt>().getValue(); |
| 317 | os << Value << ' ' << (Value->isSigned() ? 'S' : 'U') |
| 318 | << Value->getBitWidth() << 'b'; |
| 319 | break; |
| 320 | } |
| 321 | case nonloc::SymbolValKind: |
| 322 | os << castAs<nonloc::SymbolVal>().getSymbol(); |
| 323 | break; |
| 324 | |
| 325 | case nonloc::LocAsIntegerKind: { |
| 326 | const nonloc::LocAsInteger& C = castAs<nonloc::LocAsInteger>(); |
| 327 | os << C.getLoc() << " [as " << C.getNumBits() << " bit integer]" ; |
| 328 | break; |
| 329 | } |
| 330 | case nonloc::CompoundValKind: { |
| 331 | const nonloc::CompoundVal& C = castAs<nonloc::CompoundVal>(); |
| 332 | os << "compoundVal{" ; |
| 333 | bool first = true; |
| 334 | for (const auto &I : C) { |
| 335 | if (first) { |
| 336 | os << ' '; first = false; |
| 337 | } |
| 338 | else |
| 339 | os << ", " ; |
| 340 | |
| 341 | I.dumpToStream(os); |
| 342 | } |
| 343 | os << "}" ; |
| 344 | break; |
| 345 | } |
| 346 | case nonloc::LazyCompoundValKind: { |
| 347 | const nonloc::LazyCompoundVal &C = castAs<nonloc::LazyCompoundVal>(); |
| 348 | os << "lazyCompoundVal{" << const_cast<void *>(C.getStore()) |
| 349 | << ',' << C.getRegion() |
| 350 | << '}'; |
| 351 | break; |
| 352 | } |
| 353 | case nonloc::PointerToMemberKind: { |
| 354 | os << "pointerToMember{" ; |
| 355 | const nonloc::PointerToMember &CastRes = |
| 356 | castAs<nonloc::PointerToMember>(); |
| 357 | if (CastRes.getDecl()) |
| 358 | os << "|" << CastRes.getDecl()->getQualifiedNameAsString() << "|" ; |
| 359 | bool first = true; |
| 360 | for (const auto &I : CastRes) { |
| 361 | if (first) { |
| 362 | os << ' '; first = false; |
| 363 | } |
| 364 | else |
| 365 | os << ", " ; |
| 366 | |
| 367 | os << I->getType(); |
| 368 | } |
| 369 | |
| 370 | os << '}'; |
| 371 | break; |
| 372 | } |
| 373 | default: |
| 374 | assert(false && "Pretty-printed not implemented for this NonLoc." ); |
| 375 | break; |
| 376 | } |
| 377 | } |
| 378 | |
| 379 | void Loc::dumpToStream(raw_ostream &os) const { |
| 380 | switch (getKind()) { |
| 381 | case loc::ConcreteIntKind: |
| 382 | os << castAs<loc::ConcreteInt>().getValue()->getZExtValue() << " (Loc)" ; |
| 383 | break; |
| 384 | case loc::GotoLabelKind: |
| 385 | os << "&&" << castAs<loc::GotoLabel>().getLabel()->getName(); |
| 386 | break; |
| 387 | case loc::MemRegionValKind: |
| 388 | os << '&' << castAs<loc::MemRegionVal>().getRegion()->getString(); |
| 389 | break; |
| 390 | default: |
| 391 | llvm_unreachable("Pretty-printing not implemented for this Loc." ); |
| 392 | } |
| 393 | } |
| 394 | |