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