• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //= ProgramState.cpp - Path-Sensitive "State" for tracking values --*- C++ -*--=
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 //  This file implements ProgramState and ProgramStateManager.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/StaticAnalyzer/Core/PathSensitive/ProgramState.h"
15 #include "clang/Analysis/CFG.h"
16 #include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h"
17 #include "clang/StaticAnalyzer/Core/PathSensitive/ProgramStateTrait.h"
18 #include "clang/StaticAnalyzer/Core/PathSensitive/SubEngine.h"
19 #include "clang/StaticAnalyzer/Core/PathSensitive/TaintManager.h"
20 #include "llvm/Support/raw_ostream.h"
21 
22 using namespace clang;
23 using namespace ento;
24 
25 namespace clang { namespace  ento {
26 /// Increments the number of times this state is referenced.
27 
ProgramStateRetain(const ProgramState * state)28 void ProgramStateRetain(const ProgramState *state) {
29   ++const_cast<ProgramState*>(state)->refCount;
30 }
31 
32 /// Decrement the number of times this state is referenced.
ProgramStateRelease(const ProgramState * state)33 void ProgramStateRelease(const ProgramState *state) {
34   assert(state->refCount > 0);
35   ProgramState *s = const_cast<ProgramState*>(state);
36   if (--s->refCount == 0) {
37     ProgramStateManager &Mgr = s->getStateManager();
38     Mgr.StateSet.RemoveNode(s);
39     s->~ProgramState();
40     Mgr.freeStates.push_back(s);
41   }
42 }
43 }}
44 
ProgramState(ProgramStateManager * mgr,const Environment & env,StoreRef st,GenericDataMap gdm)45 ProgramState::ProgramState(ProgramStateManager *mgr, const Environment& env,
46                  StoreRef st, GenericDataMap gdm)
47   : stateMgr(mgr),
48     Env(env),
49     store(st.getStore()),
50     GDM(gdm),
51     refCount(0) {
52   stateMgr->getStoreManager().incrementReferenceCount(store);
53 }
54 
ProgramState(const ProgramState & RHS)55 ProgramState::ProgramState(const ProgramState &RHS)
56     : llvm::FoldingSetNode(),
57       stateMgr(RHS.stateMgr),
58       Env(RHS.Env),
59       store(RHS.store),
60       GDM(RHS.GDM),
61       refCount(0) {
62   stateMgr->getStoreManager().incrementReferenceCount(store);
63 }
64 
~ProgramState()65 ProgramState::~ProgramState() {
66   if (store)
67     stateMgr->getStoreManager().decrementReferenceCount(store);
68 }
69 
ProgramStateManager(ASTContext & Ctx,StoreManagerCreator CreateSMgr,ConstraintManagerCreator CreateCMgr,llvm::BumpPtrAllocator & alloc,SubEngine * SubEng)70 ProgramStateManager::ProgramStateManager(ASTContext &Ctx,
71                                          StoreManagerCreator CreateSMgr,
72                                          ConstraintManagerCreator CreateCMgr,
73                                          llvm::BumpPtrAllocator &alloc,
74                                          SubEngine *SubEng)
75   : Eng(SubEng), EnvMgr(alloc), GDMFactory(alloc),
76     svalBuilder(createSimpleSValBuilder(alloc, Ctx, *this)),
77     CallEventMgr(new CallEventManager(alloc)), Alloc(alloc) {
78   StoreMgr.reset((*CreateSMgr)(*this));
79   ConstraintMgr.reset((*CreateCMgr)(*this, SubEng));
80 }
81 
82 
~ProgramStateManager()83 ProgramStateManager::~ProgramStateManager() {
84   for (GDMContextsTy::iterator I=GDMContexts.begin(), E=GDMContexts.end();
85        I!=E; ++I)
86     I->second.second(I->second.first);
87 }
88 
89 ProgramStateRef
removeDeadBindings(ProgramStateRef state,const StackFrameContext * LCtx,SymbolReaper & SymReaper)90 ProgramStateManager::removeDeadBindings(ProgramStateRef state,
91                                    const StackFrameContext *LCtx,
92                                    SymbolReaper& SymReaper) {
93 
94   // This code essentially performs a "mark-and-sweep" of the VariableBindings.
95   // The roots are any Block-level exprs and Decls that our liveness algorithm
96   // tells us are live.  We then see what Decls they may reference, and keep
97   // those around.  This code more than likely can be made faster, and the
98   // frequency of which this method is called should be experimented with
99   // for optimum performance.
100   ProgramState NewState = *state;
101 
102   NewState.Env = EnvMgr.removeDeadBindings(NewState.Env, SymReaper, state);
103 
104   // Clean up the store.
105   StoreRef newStore = StoreMgr->removeDeadBindings(NewState.getStore(), LCtx,
106                                                    SymReaper);
107   NewState.setStore(newStore);
108   SymReaper.setReapedStore(newStore);
109 
110   ProgramStateRef Result = getPersistentState(NewState);
111   return ConstraintMgr->removeDeadBindings(Result, SymReaper);
112 }
113 
bindLoc(Loc LV,SVal V,bool notifyChanges) const114 ProgramStateRef ProgramState::bindLoc(Loc LV, SVal V, bool notifyChanges) const {
115   ProgramStateManager &Mgr = getStateManager();
116   ProgramStateRef newState = makeWithStore(Mgr.StoreMgr->Bind(getStore(),
117                                                              LV, V));
118   const MemRegion *MR = LV.getAsRegion();
119   if (MR && Mgr.getOwningEngine() && notifyChanges)
120     return Mgr.getOwningEngine()->processRegionChange(newState, MR);
121 
122   return newState;
123 }
124 
bindDefault(SVal loc,SVal V) const125 ProgramStateRef ProgramState::bindDefault(SVal loc, SVal V) const {
126   ProgramStateManager &Mgr = getStateManager();
127   const MemRegion *R = loc.castAs<loc::MemRegionVal>().getRegion();
128   const StoreRef &newStore = Mgr.StoreMgr->BindDefault(getStore(), R, V);
129   ProgramStateRef new_state = makeWithStore(newStore);
130   return Mgr.getOwningEngine() ?
131            Mgr.getOwningEngine()->processRegionChange(new_state, R) :
132            new_state;
133 }
134 
135 typedef ArrayRef<const MemRegion *> RegionList;
136 typedef ArrayRef<SVal> ValueList;
137 
138 ProgramStateRef
invalidateRegions(RegionList Regions,const Expr * E,unsigned Count,const LocationContext * LCtx,bool CausedByPointerEscape,InvalidatedSymbols * IS,const CallEvent * Call,RegionList ConstRegions) const139 ProgramState::invalidateRegions(RegionList Regions,
140                                 const Expr *E, unsigned Count,
141                                 const LocationContext *LCtx,
142                                 bool CausedByPointerEscape,
143                                 InvalidatedSymbols *IS,
144                                 const CallEvent *Call,
145                                 RegionList ConstRegions) const {
146   SmallVector<SVal, 8> Values;
147   for (RegionList::const_iterator I = Regions.begin(),
148                                   End = Regions.end(); I != End; ++I)
149     Values.push_back(loc::MemRegionVal(*I));
150 
151   SmallVector<SVal, 8> ConstValues;
152   for (RegionList::const_iterator I = ConstRegions.begin(),
153                                   End = ConstRegions.end(); I != End; ++I)
154     ConstValues.push_back(loc::MemRegionVal(*I));
155 
156   if (!IS) {
157     InvalidatedSymbols invalidated;
158     return invalidateRegionsImpl(Values, E, Count, LCtx,
159                                  CausedByPointerEscape,
160                                  invalidated, Call, ConstValues);
161   }
162   return invalidateRegionsImpl(Values, E, Count, LCtx, CausedByPointerEscape,
163                                *IS, Call, ConstValues);
164 }
165 
166 ProgramStateRef
invalidateRegions(ValueList Values,const Expr * E,unsigned Count,const LocationContext * LCtx,bool CausedByPointerEscape,InvalidatedSymbols * IS,const CallEvent * Call,ValueList ConstValues) const167 ProgramState::invalidateRegions(ValueList Values,
168                                 const Expr *E, unsigned Count,
169                                 const LocationContext *LCtx,
170                                 bool CausedByPointerEscape,
171                                 InvalidatedSymbols *IS,
172                                 const CallEvent *Call,
173                                 ValueList ConstValues) const {
174   if (!IS) {
175     InvalidatedSymbols invalidated;
176     return invalidateRegionsImpl(Values, E, Count, LCtx,
177                                  CausedByPointerEscape,
178                                  invalidated, Call, ConstValues);
179   }
180   return invalidateRegionsImpl(Values, E, Count, LCtx, CausedByPointerEscape,
181                                *IS, Call, ConstValues);
182 }
183 
184 ProgramStateRef
invalidateRegionsImpl(ValueList Values,const Expr * E,unsigned Count,const LocationContext * LCtx,bool CausedByPointerEscape,InvalidatedSymbols & IS,const CallEvent * Call,ValueList ConstValues) const185 ProgramState::invalidateRegionsImpl(ValueList Values,
186                                     const Expr *E, unsigned Count,
187                                     const LocationContext *LCtx,
188                                     bool CausedByPointerEscape,
189                                     InvalidatedSymbols &IS,
190                                     const CallEvent *Call,
191                                     ValueList ConstValues) const {
192   ProgramStateManager &Mgr = getStateManager();
193   SubEngine* Eng = Mgr.getOwningEngine();
194   InvalidatedSymbols ConstIS;
195 
196   if (Eng) {
197     StoreManager::InvalidatedRegions TopLevelInvalidated;
198     StoreManager::InvalidatedRegions TopLevelConstInvalidated;
199     StoreManager::InvalidatedRegions Invalidated;
200     const StoreRef &newStore
201     = Mgr.StoreMgr->invalidateRegions(getStore(), Values, ConstValues,
202                                       E, Count, LCtx, Call,
203                                       IS, ConstIS,
204                                       &TopLevelInvalidated,
205                                       &TopLevelConstInvalidated,
206                                       &Invalidated);
207 
208     ProgramStateRef newState = makeWithStore(newStore);
209 
210     if (CausedByPointerEscape) {
211       newState = Eng->notifyCheckersOfPointerEscape(newState, &IS,
212                                                     TopLevelInvalidated,
213                                                     Invalidated, Call);
214       if (!ConstValues.empty()) {
215         StoreManager::InvalidatedRegions Empty;
216         newState = Eng->notifyCheckersOfPointerEscape(newState, &ConstIS,
217                                                       TopLevelConstInvalidated,
218                                                       Empty, Call,
219                                                       true);
220       }
221     }
222 
223     return Eng->processRegionChanges(newState, &IS,
224                                      TopLevelInvalidated, Invalidated,
225                                      Call);
226   }
227 
228   const StoreRef &newStore =
229   Mgr.StoreMgr->invalidateRegions(getStore(), Values, ConstValues,
230                                   E, Count, LCtx, Call,
231                                   IS, ConstIS, NULL, NULL, NULL);
232   return makeWithStore(newStore);
233 }
234 
killBinding(Loc LV) const235 ProgramStateRef ProgramState::killBinding(Loc LV) const {
236   assert(!LV.getAs<loc::MemRegionVal>() && "Use invalidateRegion instead.");
237 
238   Store OldStore = getStore();
239   const StoreRef &newStore =
240     getStateManager().StoreMgr->killBinding(OldStore, LV);
241 
242   if (newStore.getStore() == OldStore)
243     return this;
244 
245   return makeWithStore(newStore);
246 }
247 
248 ProgramStateRef
enterStackFrame(const CallEvent & Call,const StackFrameContext * CalleeCtx) const249 ProgramState::enterStackFrame(const CallEvent &Call,
250                               const StackFrameContext *CalleeCtx) const {
251   const StoreRef &NewStore =
252     getStateManager().StoreMgr->enterStackFrame(getStore(), Call, CalleeCtx);
253   return makeWithStore(NewStore);
254 }
255 
getSValAsScalarOrLoc(const MemRegion * R) const256 SVal ProgramState::getSValAsScalarOrLoc(const MemRegion *R) const {
257   // We only want to do fetches from regions that we can actually bind
258   // values.  For example, SymbolicRegions of type 'id<...>' cannot
259   // have direct bindings (but their can be bindings on their subregions).
260   if (!R->isBoundable())
261     return UnknownVal();
262 
263   if (const TypedValueRegion *TR = dyn_cast<TypedValueRegion>(R)) {
264     QualType T = TR->getValueType();
265     if (Loc::isLocType(T) || T->isIntegralOrEnumerationType())
266       return getSVal(R);
267   }
268 
269   return UnknownVal();
270 }
271 
getSVal(Loc location,QualType T) const272 SVal ProgramState::getSVal(Loc location, QualType T) const {
273   SVal V = getRawSVal(cast<Loc>(location), T);
274 
275   // If 'V' is a symbolic value that is *perfectly* constrained to
276   // be a constant value, use that value instead to lessen the burden
277   // on later analysis stages (so we have less symbolic values to reason
278   // about).
279   if (!T.isNull()) {
280     if (SymbolRef sym = V.getAsSymbol()) {
281       if (const llvm::APSInt *Int = getStateManager()
282                                     .getConstraintManager()
283                                     .getSymVal(this, sym)) {
284         // FIXME: Because we don't correctly model (yet) sign-extension
285         // and truncation of symbolic values, we need to convert
286         // the integer value to the correct signedness and bitwidth.
287         //
288         // This shows up in the following:
289         //
290         //   char foo();
291         //   unsigned x = foo();
292         //   if (x == 54)
293         //     ...
294         //
295         //  The symbolic value stored to 'x' is actually the conjured
296         //  symbol for the call to foo(); the type of that symbol is 'char',
297         //  not unsigned.
298         const llvm::APSInt &NewV = getBasicVals().Convert(T, *Int);
299 
300         if (V.getAs<Loc>())
301           return loc::ConcreteInt(NewV);
302         else
303           return nonloc::ConcreteInt(NewV);
304       }
305     }
306   }
307 
308   return V;
309 }
310 
BindExpr(const Stmt * S,const LocationContext * LCtx,SVal V,bool Invalidate) const311 ProgramStateRef ProgramState::BindExpr(const Stmt *S,
312                                            const LocationContext *LCtx,
313                                            SVal V, bool Invalidate) const{
314   Environment NewEnv =
315     getStateManager().EnvMgr.bindExpr(Env, EnvironmentEntry(S, LCtx), V,
316                                       Invalidate);
317   if (NewEnv == Env)
318     return this;
319 
320   ProgramState NewSt = *this;
321   NewSt.Env = NewEnv;
322   return getStateManager().getPersistentState(NewSt);
323 }
324 
assumeInBound(DefinedOrUnknownSVal Idx,DefinedOrUnknownSVal UpperBound,bool Assumption,QualType indexTy) const325 ProgramStateRef ProgramState::assumeInBound(DefinedOrUnknownSVal Idx,
326                                       DefinedOrUnknownSVal UpperBound,
327                                       bool Assumption,
328                                       QualType indexTy) const {
329   if (Idx.isUnknown() || UpperBound.isUnknown())
330     return this;
331 
332   // Build an expression for 0 <= Idx < UpperBound.
333   // This is the same as Idx + MIN < UpperBound + MIN, if overflow is allowed.
334   // FIXME: This should probably be part of SValBuilder.
335   ProgramStateManager &SM = getStateManager();
336   SValBuilder &svalBuilder = SM.getSValBuilder();
337   ASTContext &Ctx = svalBuilder.getContext();
338 
339   // Get the offset: the minimum value of the array index type.
340   BasicValueFactory &BVF = svalBuilder.getBasicValueFactory();
341   // FIXME: This should be using ValueManager::ArrayindexTy...somehow.
342   if (indexTy.isNull())
343     indexTy = Ctx.IntTy;
344   nonloc::ConcreteInt Min(BVF.getMinValue(indexTy));
345 
346   // Adjust the index.
347   SVal newIdx = svalBuilder.evalBinOpNN(this, BO_Add,
348                                         Idx.castAs<NonLoc>(), Min, indexTy);
349   if (newIdx.isUnknownOrUndef())
350     return this;
351 
352   // Adjust the upper bound.
353   SVal newBound =
354     svalBuilder.evalBinOpNN(this, BO_Add, UpperBound.castAs<NonLoc>(),
355                             Min, indexTy);
356 
357   if (newBound.isUnknownOrUndef())
358     return this;
359 
360   // Build the actual comparison.
361   SVal inBound = svalBuilder.evalBinOpNN(this, BO_LT, newIdx.castAs<NonLoc>(),
362                                          newBound.castAs<NonLoc>(), Ctx.IntTy);
363   if (inBound.isUnknownOrUndef())
364     return this;
365 
366   // Finally, let the constraint manager take care of it.
367   ConstraintManager &CM = SM.getConstraintManager();
368   return CM.assume(this, inBound.castAs<DefinedSVal>(), Assumption);
369 }
370 
isNull(SVal V) const371 ConditionTruthVal ProgramState::isNull(SVal V) const {
372   if (V.isZeroConstant())
373     return true;
374 
375   if (V.isConstant())
376     return false;
377 
378   SymbolRef Sym = V.getAsSymbol(/* IncludeBaseRegion */ true);
379   if (!Sym)
380     return ConditionTruthVal();
381 
382   return getStateManager().ConstraintMgr->isNull(this, Sym);
383 }
384 
getInitialState(const LocationContext * InitLoc)385 ProgramStateRef ProgramStateManager::getInitialState(const LocationContext *InitLoc) {
386   ProgramState State(this,
387                 EnvMgr.getInitialEnvironment(),
388                 StoreMgr->getInitialStore(InitLoc),
389                 GDMFactory.getEmptyMap());
390 
391   return getPersistentState(State);
392 }
393 
getPersistentStateWithGDM(ProgramStateRef FromState,ProgramStateRef GDMState)394 ProgramStateRef ProgramStateManager::getPersistentStateWithGDM(
395                                                      ProgramStateRef FromState,
396                                                      ProgramStateRef GDMState) {
397   ProgramState NewState(*FromState);
398   NewState.GDM = GDMState->GDM;
399   return getPersistentState(NewState);
400 }
401 
getPersistentState(ProgramState & State)402 ProgramStateRef ProgramStateManager::getPersistentState(ProgramState &State) {
403 
404   llvm::FoldingSetNodeID ID;
405   State.Profile(ID);
406   void *InsertPos;
407 
408   if (ProgramState *I = StateSet.FindNodeOrInsertPos(ID, InsertPos))
409     return I;
410 
411   ProgramState *newState = 0;
412   if (!freeStates.empty()) {
413     newState = freeStates.back();
414     freeStates.pop_back();
415   }
416   else {
417     newState = (ProgramState*) Alloc.Allocate<ProgramState>();
418   }
419   new (newState) ProgramState(State);
420   StateSet.InsertNode(newState, InsertPos);
421   return newState;
422 }
423 
makeWithStore(const StoreRef & store) const424 ProgramStateRef ProgramState::makeWithStore(const StoreRef &store) const {
425   ProgramState NewSt(*this);
426   NewSt.setStore(store);
427   return getStateManager().getPersistentState(NewSt);
428 }
429 
setStore(const StoreRef & newStore)430 void ProgramState::setStore(const StoreRef &newStore) {
431   Store newStoreStore = newStore.getStore();
432   if (newStoreStore)
433     stateMgr->getStoreManager().incrementReferenceCount(newStoreStore);
434   if (store)
435     stateMgr->getStoreManager().decrementReferenceCount(store);
436   store = newStoreStore;
437 }
438 
439 //===----------------------------------------------------------------------===//
440 //  State pretty-printing.
441 //===----------------------------------------------------------------------===//
442 
print(raw_ostream & Out,const char * NL,const char * Sep) const443 void ProgramState::print(raw_ostream &Out,
444                          const char *NL, const char *Sep) const {
445   // Print the store.
446   ProgramStateManager &Mgr = getStateManager();
447   Mgr.getStoreManager().print(getStore(), Out, NL, Sep);
448 
449   // Print out the environment.
450   Env.print(Out, NL, Sep);
451 
452   // Print out the constraints.
453   Mgr.getConstraintManager().print(this, Out, NL, Sep);
454 
455   // Print checker-specific data.
456   Mgr.getOwningEngine()->printState(Out, this, NL, Sep);
457 }
458 
printDOT(raw_ostream & Out) const459 void ProgramState::printDOT(raw_ostream &Out) const {
460   print(Out, "\\l", "\\|");
461 }
462 
dump() const463 void ProgramState::dump() const {
464   print(llvm::errs());
465 }
466 
printTaint(raw_ostream & Out,const char * NL,const char * Sep) const467 void ProgramState::printTaint(raw_ostream &Out,
468                               const char *NL, const char *Sep) const {
469   TaintMapImpl TM = get<TaintMap>();
470 
471   if (!TM.isEmpty())
472     Out <<"Tainted Symbols:" << NL;
473 
474   for (TaintMapImpl::iterator I = TM.begin(), E = TM.end(); I != E; ++I) {
475     Out << I->first << " : " << I->second << NL;
476   }
477 }
478 
dumpTaint() const479 void ProgramState::dumpTaint() const {
480   printTaint(llvm::errs());
481 }
482 
483 //===----------------------------------------------------------------------===//
484 // Generic Data Map.
485 //===----------------------------------------------------------------------===//
486 
FindGDM(void * K) const487 void *const* ProgramState::FindGDM(void *K) const {
488   return GDM.lookup(K);
489 }
490 
491 void*
FindGDMContext(void * K,void * (* CreateContext)(llvm::BumpPtrAllocator &),void (* DeleteContext)(void *))492 ProgramStateManager::FindGDMContext(void *K,
493                                void *(*CreateContext)(llvm::BumpPtrAllocator&),
494                                void (*DeleteContext)(void*)) {
495 
496   std::pair<void*, void (*)(void*)>& p = GDMContexts[K];
497   if (!p.first) {
498     p.first = CreateContext(Alloc);
499     p.second = DeleteContext;
500   }
501 
502   return p.first;
503 }
504 
addGDM(ProgramStateRef St,void * Key,void * Data)505 ProgramStateRef ProgramStateManager::addGDM(ProgramStateRef St, void *Key, void *Data){
506   ProgramState::GenericDataMap M1 = St->getGDM();
507   ProgramState::GenericDataMap M2 = GDMFactory.add(M1, Key, Data);
508 
509   if (M1 == M2)
510     return St;
511 
512   ProgramState NewSt = *St;
513   NewSt.GDM = M2;
514   return getPersistentState(NewSt);
515 }
516 
removeGDM(ProgramStateRef state,void * Key)517 ProgramStateRef ProgramStateManager::removeGDM(ProgramStateRef state, void *Key) {
518   ProgramState::GenericDataMap OldM = state->getGDM();
519   ProgramState::GenericDataMap NewM = GDMFactory.remove(OldM, Key);
520 
521   if (NewM == OldM)
522     return state;
523 
524   ProgramState NewState = *state;
525   NewState.GDM = NewM;
526   return getPersistentState(NewState);
527 }
528 
scan(nonloc::CompoundVal val)529 bool ScanReachableSymbols::scan(nonloc::CompoundVal val) {
530   for (nonloc::CompoundVal::iterator I=val.begin(), E=val.end(); I!=E; ++I)
531     if (!scan(*I))
532       return false;
533 
534   return true;
535 }
536 
scan(const SymExpr * sym)537 bool ScanReachableSymbols::scan(const SymExpr *sym) {
538   unsigned &isVisited = visited[sym];
539   if (isVisited)
540     return true;
541   isVisited = 1;
542 
543   if (!visitor.VisitSymbol(sym))
544     return false;
545 
546   // TODO: should be rewritten using SymExpr::symbol_iterator.
547   switch (sym->getKind()) {
548     case SymExpr::RegionValueKind:
549     case SymExpr::ConjuredKind:
550     case SymExpr::DerivedKind:
551     case SymExpr::ExtentKind:
552     case SymExpr::MetadataKind:
553       break;
554     case SymExpr::CastSymbolKind:
555       return scan(cast<SymbolCast>(sym)->getOperand());
556     case SymExpr::SymIntKind:
557       return scan(cast<SymIntExpr>(sym)->getLHS());
558     case SymExpr::IntSymKind:
559       return scan(cast<IntSymExpr>(sym)->getRHS());
560     case SymExpr::SymSymKind: {
561       const SymSymExpr *x = cast<SymSymExpr>(sym);
562       return scan(x->getLHS()) && scan(x->getRHS());
563     }
564   }
565   return true;
566 }
567 
scan(SVal val)568 bool ScanReachableSymbols::scan(SVal val) {
569   if (Optional<loc::MemRegionVal> X = val.getAs<loc::MemRegionVal>())
570     return scan(X->getRegion());
571 
572   if (Optional<nonloc::LazyCompoundVal> X =
573           val.getAs<nonloc::LazyCompoundVal>()) {
574     StoreManager &StoreMgr = state->getStateManager().getStoreManager();
575     // FIXME: We don't really want to use getBaseRegion() here because pointer
576     // arithmetic doesn't apply, but scanReachableSymbols only accepts base
577     // regions right now.
578     if (!StoreMgr.scanReachableSymbols(X->getStore(),
579                                        X->getRegion()->getBaseRegion(),
580                                        *this))
581       return false;
582   }
583 
584   if (Optional<nonloc::LocAsInteger> X = val.getAs<nonloc::LocAsInteger>())
585     return scan(X->getLoc());
586 
587   if (SymbolRef Sym = val.getAsSymbol())
588     return scan(Sym);
589 
590   if (const SymExpr *Sym = val.getAsSymbolicExpression())
591     return scan(Sym);
592 
593   if (Optional<nonloc::CompoundVal> X = val.getAs<nonloc::CompoundVal>())
594     return scan(*X);
595 
596   return true;
597 }
598 
scan(const MemRegion * R)599 bool ScanReachableSymbols::scan(const MemRegion *R) {
600   if (isa<MemSpaceRegion>(R))
601     return true;
602 
603   unsigned &isVisited = visited[R];
604   if (isVisited)
605     return true;
606   isVisited = 1;
607 
608 
609   if (!visitor.VisitMemRegion(R))
610     return false;
611 
612   // If this is a symbolic region, visit the symbol for the region.
613   if (const SymbolicRegion *SR = dyn_cast<SymbolicRegion>(R))
614     if (!visitor.VisitSymbol(SR->getSymbol()))
615       return false;
616 
617   // If this is a subregion, also visit the parent regions.
618   if (const SubRegion *SR = dyn_cast<SubRegion>(R)) {
619     const MemRegion *Super = SR->getSuperRegion();
620     if (!scan(Super))
621       return false;
622 
623     // When we reach the topmost region, scan all symbols in it.
624     if (isa<MemSpaceRegion>(Super)) {
625       StoreManager &StoreMgr = state->getStateManager().getStoreManager();
626       if (!StoreMgr.scanReachableSymbols(state->getStore(), SR, *this))
627         return false;
628     }
629   }
630 
631   // Regions captured by a block are also implicitly reachable.
632   if (const BlockDataRegion *BDR = dyn_cast<BlockDataRegion>(R)) {
633     BlockDataRegion::referenced_vars_iterator I = BDR->referenced_vars_begin(),
634                                               E = BDR->referenced_vars_end();
635     for ( ; I != E; ++I) {
636       if (!scan(I.getCapturedRegion()))
637         return false;
638     }
639   }
640 
641   return true;
642 }
643 
scanReachableSymbols(SVal val,SymbolVisitor & visitor) const644 bool ProgramState::scanReachableSymbols(SVal val, SymbolVisitor& visitor) const {
645   ScanReachableSymbols S(this, visitor);
646   return S.scan(val);
647 }
648 
scanReachableSymbols(const SVal * I,const SVal * E,SymbolVisitor & visitor) const649 bool ProgramState::scanReachableSymbols(const SVal *I, const SVal *E,
650                                    SymbolVisitor &visitor) const {
651   ScanReachableSymbols S(this, visitor);
652   for ( ; I != E; ++I) {
653     if (!S.scan(*I))
654       return false;
655   }
656   return true;
657 }
658 
scanReachableSymbols(const MemRegion * const * I,const MemRegion * const * E,SymbolVisitor & visitor) const659 bool ProgramState::scanReachableSymbols(const MemRegion * const *I,
660                                    const MemRegion * const *E,
661                                    SymbolVisitor &visitor) const {
662   ScanReachableSymbols S(this, visitor);
663   for ( ; I != E; ++I) {
664     if (!S.scan(*I))
665       return false;
666   }
667   return true;
668 }
669 
addTaint(const Stmt * S,const LocationContext * LCtx,TaintTagType Kind) const670 ProgramStateRef ProgramState::addTaint(const Stmt *S,
671                                            const LocationContext *LCtx,
672                                            TaintTagType Kind) const {
673   if (const Expr *E = dyn_cast_or_null<Expr>(S))
674     S = E->IgnoreParens();
675 
676   SymbolRef Sym = getSVal(S, LCtx).getAsSymbol();
677   if (Sym)
678     return addTaint(Sym, Kind);
679 
680   const MemRegion *R = getSVal(S, LCtx).getAsRegion();
681   addTaint(R, Kind);
682 
683   // Cannot add taint, so just return the state.
684   return this;
685 }
686 
addTaint(const MemRegion * R,TaintTagType Kind) const687 ProgramStateRef ProgramState::addTaint(const MemRegion *R,
688                                            TaintTagType Kind) const {
689   if (const SymbolicRegion *SR = dyn_cast_or_null<SymbolicRegion>(R))
690     return addTaint(SR->getSymbol(), Kind);
691   return this;
692 }
693 
addTaint(SymbolRef Sym,TaintTagType Kind) const694 ProgramStateRef ProgramState::addTaint(SymbolRef Sym,
695                                            TaintTagType Kind) const {
696   // If this is a symbol cast, remove the cast before adding the taint. Taint
697   // is cast agnostic.
698   while (const SymbolCast *SC = dyn_cast<SymbolCast>(Sym))
699     Sym = SC->getOperand();
700 
701   ProgramStateRef NewState = set<TaintMap>(Sym, Kind);
702   assert(NewState);
703   return NewState;
704 }
705 
isTainted(const Stmt * S,const LocationContext * LCtx,TaintTagType Kind) const706 bool ProgramState::isTainted(const Stmt *S, const LocationContext *LCtx,
707                              TaintTagType Kind) const {
708   if (const Expr *E = dyn_cast_or_null<Expr>(S))
709     S = E->IgnoreParens();
710 
711   SVal val = getSVal(S, LCtx);
712   return isTainted(val, Kind);
713 }
714 
isTainted(SVal V,TaintTagType Kind) const715 bool ProgramState::isTainted(SVal V, TaintTagType Kind) const {
716   if (const SymExpr *Sym = V.getAsSymExpr())
717     return isTainted(Sym, Kind);
718   if (const MemRegion *Reg = V.getAsRegion())
719     return isTainted(Reg, Kind);
720   return false;
721 }
722 
isTainted(const MemRegion * Reg,TaintTagType K) const723 bool ProgramState::isTainted(const MemRegion *Reg, TaintTagType K) const {
724   if (!Reg)
725     return false;
726 
727   // Element region (array element) is tainted if either the base or the offset
728   // are tainted.
729   if (const ElementRegion *ER = dyn_cast<ElementRegion>(Reg))
730     return isTainted(ER->getSuperRegion(), K) || isTainted(ER->getIndex(), K);
731 
732   if (const SymbolicRegion *SR = dyn_cast<SymbolicRegion>(Reg))
733     return isTainted(SR->getSymbol(), K);
734 
735   if (const SubRegion *ER = dyn_cast<SubRegion>(Reg))
736     return isTainted(ER->getSuperRegion(), K);
737 
738   return false;
739 }
740 
isTainted(SymbolRef Sym,TaintTagType Kind) const741 bool ProgramState::isTainted(SymbolRef Sym, TaintTagType Kind) const {
742   if (!Sym)
743     return false;
744 
745   // Traverse all the symbols this symbol depends on to see if any are tainted.
746   bool Tainted = false;
747   for (SymExpr::symbol_iterator SI = Sym->symbol_begin(), SE =Sym->symbol_end();
748        SI != SE; ++SI) {
749     if (!isa<SymbolData>(*SI))
750       continue;
751 
752     const TaintTagType *Tag = get<TaintMap>(*SI);
753     Tainted = (Tag && *Tag == Kind);
754 
755     // If this is a SymbolDerived with a tainted parent, it's also tainted.
756     if (const SymbolDerived *SD = dyn_cast<SymbolDerived>(*SI))
757       Tainted = Tainted || isTainted(SD->getParentSymbol(), Kind);
758 
759     // If memory region is tainted, data is also tainted.
760     if (const SymbolRegionValue *SRV = dyn_cast<SymbolRegionValue>(*SI))
761       Tainted = Tainted || isTainted(SRV->getRegion(), Kind);
762 
763     // If If this is a SymbolCast from a tainted value, it's also tainted.
764     if (const SymbolCast *SC = dyn_cast<SymbolCast>(*SI))
765       Tainted = Tainted || isTainted(SC->getOperand(), Kind);
766 
767     if (Tainted)
768       return true;
769   }
770 
771   return Tainted;
772 }
773 
774 /// The GDM component containing the dynamic type info. This is a map from a
775 /// symbol to its most likely type.
REGISTER_TRAIT_WITH_PROGRAMSTATE(DynamicTypeMap,CLANG_ENTO_PROGRAMSTATE_MAP (const MemRegion *,DynamicTypeInfo))776 REGISTER_TRAIT_WITH_PROGRAMSTATE(DynamicTypeMap,
777                                  CLANG_ENTO_PROGRAMSTATE_MAP(const MemRegion *,
778                                                              DynamicTypeInfo))
779 
780 DynamicTypeInfo ProgramState::getDynamicTypeInfo(const MemRegion *Reg) const {
781   Reg = Reg->StripCasts();
782 
783   // Look up the dynamic type in the GDM.
784   const DynamicTypeInfo *GDMType = get<DynamicTypeMap>(Reg);
785   if (GDMType)
786     return *GDMType;
787 
788   // Otherwise, fall back to what we know about the region.
789   if (const TypedRegion *TR = dyn_cast<TypedRegion>(Reg))
790     return DynamicTypeInfo(TR->getLocationType(), /*CanBeSubclass=*/false);
791 
792   if (const SymbolicRegion *SR = dyn_cast<SymbolicRegion>(Reg)) {
793     SymbolRef Sym = SR->getSymbol();
794     return DynamicTypeInfo(Sym->getType());
795   }
796 
797   return DynamicTypeInfo();
798 }
799 
setDynamicTypeInfo(const MemRegion * Reg,DynamicTypeInfo NewTy) const800 ProgramStateRef ProgramState::setDynamicTypeInfo(const MemRegion *Reg,
801                                                  DynamicTypeInfo NewTy) const {
802   Reg = Reg->StripCasts();
803   ProgramStateRef NewState = set<DynamicTypeMap>(Reg, NewTy);
804   assert(NewState);
805   return NewState;
806 }
807