1 //==- ExprInspectionChecker.cpp - Used for regression tests ------*- 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 #include "clang/Analysis/IssueHash.h"
10 #include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
11 #include "clang/StaticAnalyzer/Checkers/SValExplainer.h"
12 #include "clang/StaticAnalyzer/Checkers/Taint.h"
13 #include "clang/StaticAnalyzer/Core/BugReporter/BugType.h"
14 #include "clang/StaticAnalyzer/Core/Checker.h"
15 #include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h"
16 #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
17 #include "clang/StaticAnalyzer/Core/PathSensitive/DynamicExtent.h"
18 #include "llvm/ADT/StringSwitch.h"
19 #include "llvm/Support/ScopedPrinter.h"
20 #include <optional>
21 
22 using namespace clang;
23 using namespace ento;
24 
25 namespace {
26 class ExprInspectionChecker
27     : public Checker<eval::Call, check::DeadSymbols, check::EndAnalysis> {
28   mutable std::unique_ptr<BugType> BT;
29 
30   // These stats are per-analysis, not per-branch, hence they shouldn't
31   // stay inside the program state.
32   struct ReachedStat {
33     ExplodedNode *ExampleNode;
34     unsigned NumTimesReached;
35   };
36   mutable llvm::DenseMap<const CallExpr *, ReachedStat> ReachedStats;
37 
38   void analyzerEval(const CallExpr *CE, CheckerContext &C) const;
39   void analyzerCheckInlined(const CallExpr *CE, CheckerContext &C) const;
40   void analyzerWarnIfReached(const CallExpr *CE, CheckerContext &C) const;
41   void analyzerNumTimesReached(const CallExpr *CE, CheckerContext &C) const;
42   void analyzerCrash(const CallExpr *CE, CheckerContext &C) const;
43   void analyzerWarnOnDeadSymbol(const CallExpr *CE, CheckerContext &C) const;
44   void analyzerValue(const CallExpr *CE, CheckerContext &C) const;
45   void analyzerDumpSValType(const CallExpr *CE, CheckerContext &C) const;
46   void analyzerDump(const CallExpr *CE, CheckerContext &C) const;
47   void analyzerExplain(const CallExpr *CE, CheckerContext &C) const;
48   void analyzerPrintState(const CallExpr *CE, CheckerContext &C) const;
49   void analyzerGetExtent(const CallExpr *CE, CheckerContext &C) const;
50   void analyzerDumpExtent(const CallExpr *CE, CheckerContext &C) const;
51   void analyzerDumpElementCount(const CallExpr *CE, CheckerContext &C) const;
52   void analyzerHashDump(const CallExpr *CE, CheckerContext &C) const;
53   void analyzerDenote(const CallExpr *CE, CheckerContext &C) const;
54   void analyzerExpress(const CallExpr *CE, CheckerContext &C) const;
55   void analyzerIsTainted(const CallExpr *CE, CheckerContext &C) const;
56 
57   typedef void (ExprInspectionChecker::*FnCheck)(const CallExpr *,
58                                                  CheckerContext &C) const;
59 
60   // Optional parameter `ExprVal` for expression value to be marked interesting.
61   ExplodedNode *reportBug(llvm::StringRef Msg, CheckerContext &C,
62                           std::optional<SVal> ExprVal = std::nullopt) const;
63   ExplodedNode *reportBug(llvm::StringRef Msg, BugReporter &BR, ExplodedNode *N,
64                           std::optional<SVal> ExprVal = std::nullopt) const;
65   template <typename T> void printAndReport(CheckerContext &C, T What) const;
66 
67   const Expr *getArgExpr(const CallExpr *CE, CheckerContext &C) const;
68   const MemRegion *getArgRegion(const CallExpr *CE, CheckerContext &C) const;
69 
70 public:
71   bool evalCall(const CallEvent &Call, CheckerContext &C) const;
72   void checkDeadSymbols(SymbolReaper &SymReaper, CheckerContext &C) const;
73   void checkEndAnalysis(ExplodedGraph &G, BugReporter &BR,
74                         ExprEngine &Eng) const;
75 };
76 } // namespace
77 
78 REGISTER_SET_WITH_PROGRAMSTATE(MarkedSymbols, SymbolRef)
79 REGISTER_MAP_WITH_PROGRAMSTATE(DenotedSymbols, SymbolRef, const StringLiteral *)
80 
81 bool ExprInspectionChecker::evalCall(const CallEvent &Call,
82                                      CheckerContext &C) const {
83   const auto *CE = dyn_cast_or_null<CallExpr>(Call.getOriginExpr());
84   if (!CE)
85     return false;
86 
87   // These checks should have no effect on the surrounding environment
88   // (globals should not be invalidated, etc), hence the use of evalCall.
89   FnCheck Handler =
90       llvm::StringSwitch<FnCheck>(C.getCalleeName(CE))
91           .Case("clang_analyzer_eval", &ExprInspectionChecker::analyzerEval)
92           .Case("clang_analyzer_checkInlined",
93                 &ExprInspectionChecker::analyzerCheckInlined)
94           .Case("clang_analyzer_crash", &ExprInspectionChecker::analyzerCrash)
95           .Case("clang_analyzer_warnIfReached",
96                 &ExprInspectionChecker::analyzerWarnIfReached)
97           .Case("clang_analyzer_warnOnDeadSymbol",
98                 &ExprInspectionChecker::analyzerWarnOnDeadSymbol)
99           .StartsWith("clang_analyzer_explain",
100                       &ExprInspectionChecker::analyzerExplain)
101           .Case("clang_analyzer_dumpExtent",
102                 &ExprInspectionChecker::analyzerDumpExtent)
103           .Case("clang_analyzer_dumpElementCount",
104                 &ExprInspectionChecker::analyzerDumpElementCount)
105           .Case("clang_analyzer_value", &ExprInspectionChecker::analyzerValue)
106           .StartsWith("clang_analyzer_dumpSvalType",
107                       &ExprInspectionChecker::analyzerDumpSValType)
108           .StartsWith("clang_analyzer_dump",
109                       &ExprInspectionChecker::analyzerDump)
110           .Case("clang_analyzer_getExtent",
111                 &ExprInspectionChecker::analyzerGetExtent)
112           .Case("clang_analyzer_printState",
113                 &ExprInspectionChecker::analyzerPrintState)
114           .Case("clang_analyzer_numTimesReached",
115                 &ExprInspectionChecker::analyzerNumTimesReached)
116           .Case("clang_analyzer_hashDump",
117                 &ExprInspectionChecker::analyzerHashDump)
118           .Case("clang_analyzer_denote", &ExprInspectionChecker::analyzerDenote)
119           .Case("clang_analyzer_express", // This also marks the argument as
120                                           // interesting.
121                 &ExprInspectionChecker::analyzerExpress)
122           .StartsWith("clang_analyzer_isTainted",
123                       &ExprInspectionChecker::analyzerIsTainted)
124           .Default(nullptr);
125 
126   if (!Handler)
127     return false;
128 
129   (this->*Handler)(CE, C);
130   return true;
131 }
132 
133 static const char *getArgumentValueString(const CallExpr *CE,
134                                           CheckerContext &C) {
135   if (CE->getNumArgs() == 0)
136     return "Missing assertion argument";
137 
138   ExplodedNode *N = C.getPredecessor();
139   const LocationContext *LC = N->getLocationContext();
140   ProgramStateRef State = N->getState();
141 
142   const Expr *Assertion = CE->getArg(0);
143   SVal AssertionVal = State->getSVal(Assertion, LC);
144 
145   if (AssertionVal.isUndef())
146     return "UNDEFINED";
147 
148   ProgramStateRef StTrue, StFalse;
149   std::tie(StTrue, StFalse) =
150       State->assume(AssertionVal.castAs<DefinedOrUnknownSVal>());
151 
152   if (StTrue) {
153     if (StFalse)
154       return "UNKNOWN";
155     else
156       return "TRUE";
157   } else {
158     if (StFalse)
159       return "FALSE";
160     else
161       llvm_unreachable("Invalid constraint; neither true or false.");
162   }
163 }
164 
165 ExplodedNode *
166 ExprInspectionChecker::reportBug(llvm::StringRef Msg, CheckerContext &C,
167                                  std::optional<SVal> ExprVal) const {
168   ExplodedNode *N = C.generateNonFatalErrorNode();
169   reportBug(Msg, C.getBugReporter(), N, ExprVal);
170   return N;
171 }
172 
173 ExplodedNode *
174 ExprInspectionChecker::reportBug(llvm::StringRef Msg, BugReporter &BR,
175                                  ExplodedNode *N,
176                                  std::optional<SVal> ExprVal) const {
177   if (!N)
178     return nullptr;
179 
180   if (!BT)
181     BT.reset(new BugType(this, "Checking analyzer assumptions", "debug"));
182 
183   auto R = std::make_unique<PathSensitiveBugReport>(*BT, Msg, N);
184   if (ExprVal) {
185     R->markInteresting(*ExprVal);
186   }
187   BR.emitReport(std::move(R));
188   return N;
189 }
190 
191 const Expr *ExprInspectionChecker::getArgExpr(const CallExpr *CE,
192                                               CheckerContext &C) const {
193   if (CE->getNumArgs() == 0) {
194     reportBug("Missing argument", C);
195     return nullptr;
196   }
197   return CE->getArg(0);
198 }
199 
200 const MemRegion *ExprInspectionChecker::getArgRegion(const CallExpr *CE,
201                                                      CheckerContext &C) const {
202   const Expr *Arg = getArgExpr(CE, C);
203   if (!Arg)
204     return nullptr;
205 
206   const MemRegion *MR = C.getSVal(Arg).getAsRegion();
207   if (!MR) {
208     reportBug("Cannot obtain the region", C);
209     return nullptr;
210   }
211 
212   return MR;
213 }
214 
215 void ExprInspectionChecker::analyzerEval(const CallExpr *CE,
216                                          CheckerContext &C) const {
217   const LocationContext *LC = C.getPredecessor()->getLocationContext();
218 
219   // A specific instantiation of an inlined function may have more constrained
220   // values than can generally be assumed. Skip the check.
221   if (LC->getStackFrame()->getParent() != nullptr)
222     return;
223 
224   reportBug(getArgumentValueString(CE, C), C);
225 }
226 
227 void ExprInspectionChecker::analyzerWarnIfReached(const CallExpr *CE,
228                                                   CheckerContext &C) const {
229   reportBug("REACHABLE", C);
230 }
231 
232 void ExprInspectionChecker::analyzerNumTimesReached(const CallExpr *CE,
233                                                     CheckerContext &C) const {
234   ++ReachedStats[CE].NumTimesReached;
235   if (!ReachedStats[CE].ExampleNode) {
236     // Later, in checkEndAnalysis, we'd throw a report against it.
237     ReachedStats[CE].ExampleNode = C.generateNonFatalErrorNode();
238   }
239 }
240 
241 void ExprInspectionChecker::analyzerCheckInlined(const CallExpr *CE,
242                                                  CheckerContext &C) const {
243   const LocationContext *LC = C.getPredecessor()->getLocationContext();
244 
245   // An inlined function could conceivably also be analyzed as a top-level
246   // function. We ignore this case and only emit a message (TRUE or FALSE)
247   // when we are analyzing it as an inlined function. This means that
248   // clang_analyzer_checkInlined(true) should always print TRUE, but
249   // clang_analyzer_checkInlined(false) should never actually print anything.
250   if (LC->getStackFrame()->getParent() == nullptr)
251     return;
252 
253   reportBug(getArgumentValueString(CE, C), C);
254 }
255 
256 void ExprInspectionChecker::analyzerExplain(const CallExpr *CE,
257                                             CheckerContext &C) const {
258   const Expr *Arg = getArgExpr(CE, C);
259   if (!Arg)
260     return;
261 
262   SVal V = C.getSVal(Arg);
263   SValExplainer Ex(C.getASTContext());
264   reportBug(Ex.Visit(V), C);
265 }
266 
267 static void printHelper(llvm::raw_svector_ostream &Out, CheckerContext &C,
268                         const llvm::APSInt &I) {
269   Out << I.getBitWidth() << (I.isUnsigned() ? "u:" : "s:");
270   Out << I;
271 }
272 
273 static void printHelper(llvm::raw_svector_ostream &Out, CheckerContext &C,
274                         SymbolRef Sym) {
275   C.getConstraintManager().printValue(Out, C.getState(), Sym);
276 }
277 
278 static void printHelper(llvm::raw_svector_ostream &Out, CheckerContext &C,
279                         SVal V) {
280   Out << V;
281 }
282 
283 template <typename T>
284 void ExprInspectionChecker::printAndReport(CheckerContext &C, T What) const {
285   llvm::SmallString<64> Str;
286   llvm::raw_svector_ostream OS(Str);
287   printHelper(OS, C, What);
288   reportBug(OS.str(), C);
289 }
290 
291 void ExprInspectionChecker::analyzerValue(const CallExpr *CE,
292                                           CheckerContext &C) const {
293   const Expr *Arg = getArgExpr(CE, C);
294   if (!Arg)
295     return;
296 
297   SVal V = C.getSVal(Arg);
298   if (const SymbolRef Sym = V.getAsSymbol())
299     printAndReport(C, Sym);
300   else if (const llvm::APSInt *I = V.getAsInteger())
301     printAndReport(C, *I);
302   else
303     reportBug("n/a", C);
304 }
305 
306 void ExprInspectionChecker::analyzerDumpSValType(const CallExpr *CE,
307                                                  CheckerContext &C) const {
308   const Expr *Arg = getArgExpr(CE, C);
309   if (!Arg)
310     return;
311 
312   QualType Ty = C.getSVal(Arg).getType(C.getASTContext());
313   reportBug(Ty.getAsString(), C);
314 }
315 
316 void ExprInspectionChecker::analyzerDump(const CallExpr *CE,
317                                          CheckerContext &C) const {
318   const Expr *Arg = getArgExpr(CE, C);
319   if (!Arg)
320     return;
321 
322   SVal V = C.getSVal(Arg);
323   printAndReport(C, V);
324 }
325 
326 void ExprInspectionChecker::analyzerGetExtent(const CallExpr *CE,
327                                               CheckerContext &C) const {
328   const MemRegion *MR = getArgRegion(CE, C);
329   if (!MR)
330     return;
331 
332   ProgramStateRef State = C.getState();
333   DefinedOrUnknownSVal Size = getDynamicExtent(State, MR, C.getSValBuilder());
334 
335   State = State->BindExpr(CE, C.getLocationContext(), Size);
336   C.addTransition(State);
337 }
338 
339 void ExprInspectionChecker::analyzerDumpExtent(const CallExpr *CE,
340                                                CheckerContext &C) const {
341   const MemRegion *MR = getArgRegion(CE, C);
342   if (!MR)
343     return;
344 
345   DefinedOrUnknownSVal Size =
346       getDynamicExtent(C.getState(), MR, C.getSValBuilder());
347   printAndReport(C, Size);
348 }
349 
350 void ExprInspectionChecker::analyzerDumpElementCount(const CallExpr *CE,
351                                                      CheckerContext &C) const {
352   const MemRegion *MR = getArgRegion(CE, C);
353   if (!MR)
354     return;
355 
356   QualType ElementTy;
357   if (const auto *TVR = MR->getAs<TypedValueRegion>()) {
358     ElementTy = TVR->getValueType();
359   } else {
360     ElementTy = MR->castAs<SymbolicRegion>()->getPointeeStaticType();
361   }
362 
363   assert(!ElementTy->isPointerType());
364 
365   DefinedOrUnknownSVal ElementCount =
366       getDynamicElementCount(C.getState(), MR, C.getSValBuilder(), ElementTy);
367   printAndReport(C, ElementCount);
368 }
369 
370 void ExprInspectionChecker::analyzerPrintState(const CallExpr *CE,
371                                                CheckerContext &C) const {
372   C.getState()->dump();
373 }
374 
375 void ExprInspectionChecker::analyzerWarnOnDeadSymbol(const CallExpr *CE,
376                                                      CheckerContext &C) const {
377   const Expr *Arg = getArgExpr(CE, C);
378   if (!Arg)
379     return;
380 
381   SVal Val = C.getSVal(Arg);
382   SymbolRef Sym = Val.getAsSymbol();
383   if (!Sym)
384     return;
385 
386   ProgramStateRef State = C.getState();
387   State = State->add<MarkedSymbols>(Sym);
388   C.addTransition(State);
389 }
390 
391 void ExprInspectionChecker::checkDeadSymbols(SymbolReaper &SymReaper,
392                                              CheckerContext &C) const {
393   ProgramStateRef State = C.getState();
394   const MarkedSymbolsTy &Syms = State->get<MarkedSymbols>();
395   ExplodedNode *N = C.getPredecessor();
396   for (SymbolRef Sym : Syms) {
397     if (!SymReaper.isDead(Sym))
398       continue;
399 
400     // The non-fatal error node should be the same for all reports.
401     if (ExplodedNode *BugNode = reportBug("SYMBOL DEAD", C))
402       N = BugNode;
403     State = State->remove<MarkedSymbols>(Sym);
404   }
405 
406   for (auto I : State->get<DenotedSymbols>()) {
407     SymbolRef Sym = I.first;
408     if (!SymReaper.isLive(Sym))
409       State = State->remove<DenotedSymbols>(Sym);
410   }
411 
412   C.addTransition(State, N);
413 }
414 
415 void ExprInspectionChecker::checkEndAnalysis(ExplodedGraph &G, BugReporter &BR,
416                                              ExprEngine &Eng) const {
417   for (auto Item : ReachedStats) {
418     unsigned NumTimesReached = Item.second.NumTimesReached;
419     ExplodedNode *N = Item.second.ExampleNode;
420 
421     reportBug(llvm::to_string(NumTimesReached), BR, N);
422   }
423   ReachedStats.clear();
424 }
425 
426 void ExprInspectionChecker::analyzerCrash(const CallExpr *CE,
427                                           CheckerContext &C) const {
428   LLVM_BUILTIN_TRAP;
429 }
430 
431 void ExprInspectionChecker::analyzerHashDump(const CallExpr *CE,
432                                              CheckerContext &C) const {
433   const LangOptions &Opts = C.getLangOpts();
434   const SourceManager &SM = C.getSourceManager();
435   FullSourceLoc FL(CE->getArg(0)->getBeginLoc(), SM);
436   std::string HashContent =
437       getIssueString(FL, getCheckerName().getName(), "Category",
438                      C.getLocationContext()->getDecl(), Opts);
439 
440   reportBug(HashContent, C);
441 }
442 
443 void ExprInspectionChecker::analyzerDenote(const CallExpr *CE,
444                                            CheckerContext &C) const {
445   if (CE->getNumArgs() < 2) {
446     reportBug("clang_analyzer_denote() requires a symbol and a string literal",
447               C);
448     return;
449   }
450 
451   SymbolRef Sym = C.getSVal(CE->getArg(0)).getAsSymbol();
452   if (!Sym) {
453     reportBug("Not a symbol", C);
454     return;
455   }
456 
457   const auto *E = dyn_cast<StringLiteral>(CE->getArg(1)->IgnoreParenCasts());
458   if (!E) {
459     reportBug("Not a string literal", C);
460     return;
461   }
462 
463   ProgramStateRef State = C.getState();
464 
465   C.addTransition(C.getState()->set<DenotedSymbols>(Sym, E));
466 }
467 
468 namespace {
469 class SymbolExpressor
470     : public SymExprVisitor<SymbolExpressor, std::optional<std::string>> {
471   ProgramStateRef State;
472 
473 public:
474   SymbolExpressor(ProgramStateRef State) : State(State) {}
475 
476   std::optional<std::string> lookup(const SymExpr *S) {
477     if (const StringLiteral *const *SLPtr = State->get<DenotedSymbols>(S)) {
478       const StringLiteral *SL = *SLPtr;
479       return std::string(SL->getBytes());
480     }
481     return std::nullopt;
482   }
483 
484   std::optional<std::string> VisitSymExpr(const SymExpr *S) {
485     return lookup(S);
486   }
487 
488   std::optional<std::string> VisitSymIntExpr(const SymIntExpr *S) {
489     if (std::optional<std::string> Str = lookup(S))
490       return Str;
491     if (std::optional<std::string> Str = Visit(S->getLHS()))
492       return (*Str + " " + BinaryOperator::getOpcodeStr(S->getOpcode()) + " " +
493               std::to_string(S->getRHS().getLimitedValue()) +
494               (S->getRHS().isUnsigned() ? "U" : ""))
495           .str();
496     return std::nullopt;
497   }
498 
499   std::optional<std::string> VisitSymSymExpr(const SymSymExpr *S) {
500     if (std::optional<std::string> Str = lookup(S))
501       return Str;
502     if (std::optional<std::string> Str1 = Visit(S->getLHS()))
503       if (std::optional<std::string> Str2 = Visit(S->getRHS()))
504         return (*Str1 + " " + BinaryOperator::getOpcodeStr(S->getOpcode()) +
505                 " " + *Str2)
506             .str();
507     return std::nullopt;
508   }
509 
510   std::optional<std::string> VisitUnarySymExpr(const UnarySymExpr *S) {
511     if (std::optional<std::string> Str = lookup(S))
512       return Str;
513     if (std::optional<std::string> Str = Visit(S->getOperand()))
514       return (UnaryOperator::getOpcodeStr(S->getOpcode()) + *Str).str();
515     return std::nullopt;
516   }
517 
518   std::optional<std::string> VisitSymbolCast(const SymbolCast *S) {
519     if (std::optional<std::string> Str = lookup(S))
520       return Str;
521     if (std::optional<std::string> Str = Visit(S->getOperand()))
522       return (Twine("(") + S->getType().getAsString() + ")" + *Str).str();
523     return std::nullopt;
524   }
525 };
526 } // namespace
527 
528 void ExprInspectionChecker::analyzerExpress(const CallExpr *CE,
529                                             CheckerContext &C) const {
530   const Expr *Arg = getArgExpr(CE, C);
531   if (!Arg)
532     return;
533 
534   SVal ArgVal = C.getSVal(CE->getArg(0));
535   SymbolRef Sym = ArgVal.getAsSymbol();
536   if (!Sym) {
537     reportBug("Not a symbol", C, ArgVal);
538     return;
539   }
540 
541   SymbolExpressor V(C.getState());
542   auto Str = V.Visit(Sym);
543   if (!Str) {
544     reportBug("Unable to express", C, ArgVal);
545     return;
546   }
547 
548   reportBug(*Str, C, ArgVal);
549 }
550 
551 void ExprInspectionChecker::analyzerIsTainted(const CallExpr *CE,
552                                               CheckerContext &C) const {
553   if (CE->getNumArgs() != 1) {
554     reportBug("clang_analyzer_isTainted() requires exactly one argument", C);
555     return;
556   }
557   const bool IsTainted =
558       taint::isTainted(C.getState(), CE->getArg(0), C.getLocationContext());
559   reportBug(IsTainted ? "YES" : "NO", C);
560 }
561 
562 void ento::registerExprInspectionChecker(CheckerManager &Mgr) {
563   Mgr.registerChecker<ExprInspectionChecker>();
564 }
565 
566 bool ento::shouldRegisterExprInspectionChecker(const CheckerManager &mgr) {
567   return true;
568 }
569