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/Decl.h"
16 #include "clang/AST/DeclCXX.h"
17 #include "clang/AST/Expr.h"
18 #include "clang/AST/Type.h"
19 #include "clang/Basic/JsonSupport.h"
20 #include "clang/Basic/LLVM.h"
21 #include "clang/StaticAnalyzer/Core/PathSensitive/BasicValueFactory.h"
22 #include "clang/StaticAnalyzer/Core/PathSensitive/MemRegion.h"
23 #include "clang/StaticAnalyzer/Core/PathSensitive/SValBuilder.h"
24 #include "clang/StaticAnalyzer/Core/PathSensitive/SymExpr.h"
25 #include "clang/StaticAnalyzer/Core/PathSensitive/SymbolManager.h"
26 #include "llvm/ADT/Optional.h"
27 #include "llvm/Support/Casting.h"
28 #include "llvm/Support/Compiler.h"
29 #include "llvm/Support/ErrorHandling.h"
30 #include "llvm/Support/raw_ostream.h"
31 #include <cassert>
32 
33 using namespace clang;
34 using namespace ento;
35 
36 //===----------------------------------------------------------------------===//
37 // Symbol iteration within an SVal.
38 //===----------------------------------------------------------------------===//
39 
40 //===----------------------------------------------------------------------===//
41 // Utility methods.
42 //===----------------------------------------------------------------------===//
43 
44 bool SVal::hasConjuredSymbol() const {
45   if (Optional<nonloc::SymbolVal> SV = getAs<nonloc::SymbolVal>()) {
46     SymbolRef sym = SV->getSymbol();
47     if (isa<SymbolConjured>(sym))
48       return true;
49   }
50 
51   if (Optional<loc::MemRegionVal> RV = getAs<loc::MemRegionVal>()) {
52     const MemRegion *R = RV->getRegion();
53     if (const auto *SR = dyn_cast<SymbolicRegion>(R)) {
54       SymbolRef sym = SR->getSymbol();
55       if (isa<SymbolConjured>(sym))
56         return true;
57     }
58   }
59 
60   return false;
61 }
62 
63 const FunctionDecl *SVal::getAsFunctionDecl() const {
64   if (Optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>()) {
65     const MemRegion* R = X->getRegion();
66     if (const FunctionCodeRegion *CTR = R->getAs<FunctionCodeRegion>())
67       if (const auto *FD = dyn_cast<FunctionDecl>(CTR->getDecl()))
68         return FD;
69   }
70 
71   if (auto X = getAs<nonloc::PointerToMember>()) {
72     if (const auto *MD = dyn_cast_or_null<CXXMethodDecl>(X->getDecl()))
73       return MD;
74   }
75   return nullptr;
76 }
77 
78 /// If this SVal is a location (subclasses Loc) and wraps a symbol,
79 /// return that SymbolRef.  Otherwise return 0.
80 ///
81 /// Implicit casts (ex: void* -> char*) can turn Symbolic region into Element
82 /// region. If that is the case, gets the underlining region.
83 /// When IncludeBaseRegions is set to true and the SubRegion is non-symbolic,
84 /// the first symbolic parent region is returned.
85 SymbolRef SVal::getAsLocSymbol(bool IncludeBaseRegions) const {
86   // FIXME: should we consider SymbolRef wrapped in CodeTextRegion?
87   if (Optional<nonloc::LocAsInteger> X = getAs<nonloc::LocAsInteger>())
88     return X->getLoc().getAsLocSymbol(IncludeBaseRegions);
89 
90   if (Optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>()) {
91     const MemRegion *R = X->getRegion();
92     if (const SymbolicRegion *SymR = IncludeBaseRegions ?
93                                       R->getSymbolicBase() :
94                                       dyn_cast<SymbolicRegion>(R->StripCasts()))
95       return SymR->getSymbol();
96   }
97   return nullptr;
98 }
99 
100 /// Get the symbol in the SVal or its base region.
101 SymbolRef SVal::getLocSymbolInBase() const {
102   Optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>();
103 
104   if (!X)
105     return nullptr;
106 
107   const MemRegion *R = X->getRegion();
108 
109   while (const auto *SR = dyn_cast<SubRegion>(R)) {
110     if (const auto *SymR = dyn_cast<SymbolicRegion>(SR))
111       return SymR->getSymbol();
112     else
113       R = SR->getSuperRegion();
114   }
115 
116   return nullptr;
117 }
118 
119 // TODO: The next 3 functions have to be simplified.
120 
121 /// If this SVal wraps a symbol return that SymbolRef.
122 /// Otherwise, return 0.
123 ///
124 /// Casts are ignored during lookup.
125 /// \param IncludeBaseRegions The boolean that controls whether the search
126 /// should continue to the base regions if the region is not symbolic.
127 SymbolRef SVal::getAsSymbol(bool IncludeBaseRegions) const {
128   // FIXME: should we consider SymbolRef wrapped in CodeTextRegion?
129   if (Optional<nonloc::SymbolVal> X = getAs<nonloc::SymbolVal>())
130     return X->getSymbol();
131 
132   return getAsLocSymbol(IncludeBaseRegions);
133 }
134 
135 /// getAsSymbolicExpression - If this Sval wraps a symbolic expression then
136 ///  return that expression.  Otherwise return NULL.
137 const SymExpr *SVal::getAsSymbolicExpression() const {
138   if (Optional<nonloc::SymbolVal> X = getAs<nonloc::SymbolVal>())
139     return X->getSymbol();
140 
141   return getAsSymbol();
142 }
143 
144 const SymExpr* SVal::getAsSymExpr() const {
145   const SymExpr* Sym = getAsSymbol();
146   if (!Sym)
147     Sym = getAsSymbolicExpression();
148   return Sym;
149 }
150 
151 const MemRegion *SVal::getAsRegion() const {
152   if (Optional<loc::MemRegionVal> X = getAs<loc::MemRegionVal>())
153     return X->getRegion();
154 
155   if (Optional<nonloc::LocAsInteger> X = getAs<nonloc::LocAsInteger>())
156     return X->getLoc().getAsRegion();
157 
158   return nullptr;
159 }
160 
161 const MemRegion *loc::MemRegionVal::stripCasts(bool StripBaseCasts) const {
162   const MemRegion *R = getRegion();
163   return R ?  R->StripCasts(StripBaseCasts) : nullptr;
164 }
165 
166 const void *nonloc::LazyCompoundVal::getStore() const {
167   return static_cast<const LazyCompoundValData*>(Data)->getStore();
168 }
169 
170 const TypedValueRegion *nonloc::LazyCompoundVal::getRegion() const {
171   return static_cast<const LazyCompoundValData*>(Data)->getRegion();
172 }
173 
174 bool nonloc::PointerToMember::isNullMemberPointer() const {
175   return getPTMData().isNull();
176 }
177 
178 const DeclaratorDecl *nonloc::PointerToMember::getDecl() const {
179   const auto PTMD = this->getPTMData();
180   if (PTMD.isNull())
181     return nullptr;
182 
183   const DeclaratorDecl *DD = nullptr;
184   if (PTMD.is<const DeclaratorDecl *>())
185     DD = PTMD.get<const DeclaratorDecl *>();
186   else
187     DD = PTMD.get<const PointerToMemberData *>()->getDeclaratorDecl();
188 
189   return DD;
190 }
191 
192 //===----------------------------------------------------------------------===//
193 // Other Iterators.
194 //===----------------------------------------------------------------------===//
195 
196 nonloc::CompoundVal::iterator nonloc::CompoundVal::begin() const {
197   return getValue()->begin();
198 }
199 
200 nonloc::CompoundVal::iterator nonloc::CompoundVal::end() const {
201   return getValue()->end();
202 }
203 
204 nonloc::PointerToMember::iterator nonloc::PointerToMember::begin() const {
205   const PTMDataType PTMD = getPTMData();
206   if (PTMD.is<const DeclaratorDecl *>())
207     return {};
208   return PTMD.get<const PointerToMemberData *>()->begin();
209 }
210 
211 nonloc::PointerToMember::iterator nonloc::PointerToMember::end() const {
212   const PTMDataType PTMD = getPTMData();
213   if (PTMD.is<const DeclaratorDecl *>())
214     return {};
215   return PTMD.get<const PointerToMemberData *>()->end();
216 }
217 
218 //===----------------------------------------------------------------------===//
219 // Useful predicates.
220 //===----------------------------------------------------------------------===//
221 
222 bool SVal::isConstant() const {
223   return getAs<nonloc::ConcreteInt>() || getAs<loc::ConcreteInt>();
224 }
225 
226 bool SVal::isConstant(int I) const {
227   if (Optional<loc::ConcreteInt> LV = getAs<loc::ConcreteInt>())
228     return LV->getValue() == I;
229   if (Optional<nonloc::ConcreteInt> NV = getAs<nonloc::ConcreteInt>())
230     return NV->getValue() == I;
231   return false;
232 }
233 
234 bool SVal::isZeroConstant() const {
235   return isConstant(0);
236 }
237 
238 //===----------------------------------------------------------------------===//
239 // Transfer function dispatch for Non-Locs.
240 //===----------------------------------------------------------------------===//
241 
242 SVal nonloc::ConcreteInt::evalBinOp(SValBuilder &svalBuilder,
243                                     BinaryOperator::Opcode Op,
244                                     const nonloc::ConcreteInt& R) const {
245   const llvm::APSInt* X =
246     svalBuilder.getBasicValueFactory().evalAPSInt(Op, getValue(), R.getValue());
247 
248   if (X)
249     return nonloc::ConcreteInt(*X);
250   else
251     return UndefinedVal();
252 }
253 
254 nonloc::ConcreteInt
255 nonloc::ConcreteInt::evalComplement(SValBuilder &svalBuilder) const {
256   return svalBuilder.makeIntVal(~getValue());
257 }
258 
259 nonloc::ConcreteInt
260 nonloc::ConcreteInt::evalMinus(SValBuilder &svalBuilder) const {
261   return svalBuilder.makeIntVal(-getValue());
262 }
263 
264 //===----------------------------------------------------------------------===//
265 // Transfer function dispatch for Locs.
266 //===----------------------------------------------------------------------===//
267 
268 SVal loc::ConcreteInt::evalBinOp(BasicValueFactory& BasicVals,
269                                  BinaryOperator::Opcode Op,
270                                  const loc::ConcreteInt& R) const {
271   assert(BinaryOperator::isComparisonOp(Op) || Op == BO_Sub);
272 
273   const llvm::APSInt *X = BasicVals.evalAPSInt(Op, getValue(), R.getValue());
274 
275   if (X)
276     return nonloc::ConcreteInt(*X);
277   else
278     return UndefinedVal();
279 }
280 
281 //===----------------------------------------------------------------------===//
282 // Pretty-Printing.
283 //===----------------------------------------------------------------------===//
284 
285 LLVM_DUMP_METHOD void SVal::dump() const { dumpToStream(llvm::errs()); }
286 
287 void SVal::printJson(raw_ostream &Out, bool AddQuotes) const {
288   std::string Buf;
289   llvm::raw_string_ostream TempOut(Buf);
290 
291   dumpToStream(TempOut);
292 
293   Out << JsonFormat(TempOut.str(), AddQuotes);
294 }
295 
296 void SVal::dumpToStream(raw_ostream &os) const {
297   switch (getBaseKind()) {
298     case UnknownValKind:
299       os << "Unknown";
300       break;
301     case NonLocKind:
302       castAs<NonLoc>().dumpToStream(os);
303       break;
304     case LocKind:
305       castAs<Loc>().dumpToStream(os);
306       break;
307     case UndefinedValKind:
308       os << "Undefined";
309       break;
310   }
311 }
312 
313 void NonLoc::dumpToStream(raw_ostream &os) const {
314   switch (getSubKind()) {
315     case nonloc::ConcreteIntKind: {
316       const auto &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().getAsString();
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 (getSubKind()) {
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