CStringChecker.cpp source code [clang_source_code/lib/StaticAnalyzer/Checkers/CStringChecker.cpp]

1	//= CStringChecker.cpp - Checks calls to C string functions --------- 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	// This defines CStringChecker, which is an assortment of checks on calls
10	// to functions in <string.h>.
11	//
12	//===----------------------------------------------------------------------===//
13
14	#include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
15	#include "InterCheckerAPI.h"
16	#include "clang/Basic/CharInfo.h"
17	#include "clang/StaticAnalyzer/Core/BugReporter/BugType.h"
18	#include "clang/StaticAnalyzer/Core/Checker.h"
19	#include "clang/StaticAnalyzer/Core/CheckerManager.h"
20	#include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
21	#include "clang/StaticAnalyzer/Core/PathSensitive/ProgramStateTrait.h"
22	#include "llvm/ADT/STLExtras.h"
23	#include "llvm/ADT/SmallString.h"
24	#include "llvm/Support/raw_ostream.h"
25
26	using namespace clang;
27	using namespace ento;
28
29	namespace {
30	class CStringChecker : public Checker< eval::Call,
31	check::PreStmt<DeclStmt>,
32	check::LiveSymbols,
33	check::DeadSymbols,
34	check::RegionChanges
35	> {
36	mutable std::unique_ptr<BugType> BT_Null, BT_Bounds, BT_Overlap,
37	BT_NotCString, BT_AdditionOverflow;
38
39	mutable const char *CurrentFunctionDescription;
40
41	public:
42	/// The filter is used to filter out the diagnostics which are not enabled by
43	/// the user.
44	struct CStringChecksFilter {
45	DefaultBool CheckCStringNullArg;
46	DefaultBool CheckCStringOutOfBounds;
47	DefaultBool CheckCStringBufferOverlap;
48	DefaultBool CheckCStringNotNullTerm;
49
50	CheckName CheckNameCStringNullArg;
51	CheckName CheckNameCStringOutOfBounds;
52	CheckName CheckNameCStringBufferOverlap;
53	CheckName CheckNameCStringNotNullTerm;
54	};
55
56	CStringChecksFilter Filter;
57
58	static void *getTag() { static int tag; return &tag; }
59
60	bool evalCall(const CallExpr *CE, CheckerContext &C) const;
61	void checkPreStmt(const DeclStmt *DS, CheckerContext &C) const;
62	void checkLiveSymbols(ProgramStateRef state, SymbolReaper &SR) const;
63	void checkDeadSymbols(SymbolReaper &SR, CheckerContext &C) const;
64
65	ProgramStateRef
66	checkRegionChanges(ProgramStateRef state,
67	const InvalidatedSymbols *,
68	ArrayRef<const MemRegion *> ExplicitRegions,
69	ArrayRef<const MemRegion *> Regions,
70	const LocationContext *LCtx,
71	const CallEvent *Call) const;
72
73	typedef void (CStringChecker::*FnCheck)(CheckerContext &,
74	const CallExpr *) const;
75
76	void evalMemcpy(CheckerContext &C, const CallExpr *CE) const;
77	void evalMempcpy(CheckerContext &C, const CallExpr *CE) const;
78	void evalMemmove(CheckerContext &C, const CallExpr *CE) const;
79	void evalBcopy(CheckerContext &C, const CallExpr *CE) const;
80	void evalCopyCommon(CheckerContext &C, const CallExpr *CE,
81	ProgramStateRef state,
82	const Expr *Size,
83	const Expr *Source,
84	const Expr *Dest,
85	bool Restricted = false,
86	bool IsMempcpy = false) const;
87
88	void evalMemcmp(CheckerContext &C, const CallExpr *CE) const;
89
90	void evalstrLength(CheckerContext &C, const CallExpr *CE) const;
91	void evalstrnLength(CheckerContext &C, const CallExpr *CE) const;
92	void evalstrLengthCommon(CheckerContext &C,
93	const CallExpr *CE,
94	bool IsStrnlen = false) const;
95
96	void evalStrcpy(CheckerContext &C, const CallExpr *CE) const;
97	void evalStrncpy(CheckerContext &C, const CallExpr *CE) const;
98	void evalStpcpy(CheckerContext &C, const CallExpr *CE) const;
99	void evalStrlcpy(CheckerContext &C, const CallExpr *CE) const;
100	void evalStrcpyCommon(CheckerContext &C,
101	const CallExpr *CE,
102	bool returnEnd,
103	bool isBounded,
104	bool isAppending,
105	bool returnPtr = true) const;
106
107	void evalStrcat(CheckerContext &C, const CallExpr *CE) const;
108	void evalStrncat(CheckerContext &C, const CallExpr *CE) const;
109	void evalStrlcat(CheckerContext &C, const CallExpr *CE) const;
110
111	void evalStrcmp(CheckerContext &C, const CallExpr *CE) const;
112	void evalStrncmp(CheckerContext &C, const CallExpr *CE) const;
113	void evalStrcasecmp(CheckerContext &C, const CallExpr *CE) const;
114	void evalStrncasecmp(CheckerContext &C, const CallExpr *CE) const;
115	void evalStrcmpCommon(CheckerContext &C,
116	const CallExpr *CE,
117	bool isBounded = false,
118	bool ignoreCase = false) const;
119
120	void evalStrsep(CheckerContext &C, const CallExpr *CE) const;
121
122	void evalStdCopy(CheckerContext &C, const CallExpr *CE) const;
123	void evalStdCopyBackward(CheckerContext &C, const CallExpr *CE) const;
124	void evalStdCopyCommon(CheckerContext &C, const CallExpr *CE) const;
125	void evalMemset(CheckerContext &C, const CallExpr *CE) const;
126	void evalBzero(CheckerContext &C, const CallExpr *CE) const;
127
128	// Utility methods
129	std::pair<ProgramStateRef , ProgramStateRef >
130	static assumeZero(CheckerContext &C,
131	ProgramStateRef state, SVal V, QualType Ty);
132
133	static ProgramStateRef setCStringLength(ProgramStateRef state,
134	const MemRegion *MR,
135	SVal strLength);
136	static SVal getCStringLengthForRegion(CheckerContext &C,
137	ProgramStateRef &state,
138	const Expr *Ex,
139	const MemRegion *MR,
140	bool hypothetical);
141	SVal getCStringLength(CheckerContext &C,
142	ProgramStateRef &state,
143	const Expr *Ex,
144	SVal Buf,
145	bool hypothetical = false) const;
146
147	const StringLiteral *getCStringLiteral(CheckerContext &C,
148	ProgramStateRef &state,
149	const Expr *expr,
150	SVal val) const;
151
152	static ProgramStateRef InvalidateBuffer(CheckerContext &C,
153	ProgramStateRef state,
154	const Expr *Ex, SVal V,
155	bool IsSourceBuffer,
156	const Expr *Size);
157
158	static bool SummarizeRegion(raw_ostream &os, ASTContext &Ctx,
159	const MemRegion *MR);
160
161	static bool memsetAux(const Expr *DstBuffer, SVal CharE,
162	const Expr *Size, CheckerContext &C,
163	ProgramStateRef &State);
164
165	// Re-usable checks
166	ProgramStateRef checkNonNull(CheckerContext &C,
167	ProgramStateRef state,
168	const Expr *S,
169	SVal l) const;
170	ProgramStateRef CheckLocation(CheckerContext &C,
171	ProgramStateRef state,
172	const Expr *S,
173	SVal l,
174	const char *message = nullptr) const;
175	ProgramStateRef CheckBufferAccess(CheckerContext &C,
176	ProgramStateRef state,
177	const Expr *Size,
178	const Expr *FirstBuf,
179	const Expr *SecondBuf,
180	const char *firstMessage = nullptr,
181	const char *secondMessage = nullptr,
182	bool WarnAboutSize = false) const;
183
184	ProgramStateRef CheckBufferAccess(CheckerContext &C,
185	ProgramStateRef state,
186	const Expr *Size,
187	const Expr *Buf,
188	const char *message = nullptr,
189	bool WarnAboutSize = false) const {
190	// This is a convenience overload.
191	return CheckBufferAccess(C, state, Size, Buf, nullptr, message, nullptr,
192	WarnAboutSize);
193	}
194	ProgramStateRef CheckOverlap(CheckerContext &C,
195	ProgramStateRef state,
196	const Expr *Size,
197	const Expr *First,
198	const Expr *Second) const;
199	void emitOverlapBug(CheckerContext &C,
200	ProgramStateRef state,
201	const Stmt *First,
202	const Stmt *Second) const;
203
204	void emitNullArgBug(CheckerContext &C, ProgramStateRef State, const Stmt *S,
205	StringRef WarningMsg) const;
206	void emitOutOfBoundsBug(CheckerContext &C, ProgramStateRef State,
207	const Stmt *S, StringRef WarningMsg) const;
208	void emitNotCStringBug(CheckerContext &C, ProgramStateRef State,
209	const Stmt *S, StringRef WarningMsg) const;
210	void emitAdditionOverflowBug(CheckerContext &C, ProgramStateRef State) const;
211
212	ProgramStateRef checkAdditionOverflow(CheckerContext &C,
213	ProgramStateRef state,
214	NonLoc left,
215	NonLoc right) const;
216
217	// Return true if the destination buffer of the copy function may be in bound.
218	// Expects SVal of Size to be positive and unsigned.
219	// Expects SVal of FirstBuf to be a FieldRegion.
220	static bool IsFirstBufInBound(CheckerContext &C,
221	ProgramStateRef state,
222	const Expr *FirstBuf,
223	const Expr *Size);
224	};
225
226	} //end anonymous namespace
227
228	REGISTER_MAP_WITH_PROGRAMSTATE(CStringLength, const MemRegion *, SVal)
229
230	//===----------------------------------------------------------------------===//
231	// Individual checks and utility methods.
232	//===----------------------------------------------------------------------===//
233
234	std::pair<ProgramStateRef , ProgramStateRef >
235	CStringChecker::assumeZero(CheckerContext &C, ProgramStateRef state, SVal V,
236	QualType Ty) {
237	Optional<DefinedSVal> val = V.getAs<DefinedSVal>();
238	if (!val)
239	return std::pair<ProgramStateRef , ProgramStateRef >(state, state);
240
241	SValBuilder &svalBuilder = C.getSValBuilder();
242	DefinedOrUnknownSVal zero = svalBuilder.makeZeroVal(Ty);
243	return state->assume(svalBuilder.evalEQ(state, *val, zero));
244	}
245
246	ProgramStateRef CStringChecker::checkNonNull(CheckerContext &C,
247	ProgramStateRef state,
248	const Expr *S, SVal l) const {
249	// If a previous check has failed, propagate the failure.
250	if (!state)
251	return nullptr;
252
253	ProgramStateRef stateNull, stateNonNull;
254	std::tie(stateNull, stateNonNull) = assumeZero(C, state, l, S->getType());
255
256	if (stateNull && !stateNonNull) {
257	if (Filter.CheckCStringNullArg) {
258	SmallString<80> buf;
259	llvm::raw_svector_ostream os(buf);
260	assert(CurrentFunctionDescription);
261	os << "Null pointer argument in call to " << CurrentFunctionDescription;
262
263	emitNullArgBug(C, stateNull, S, os.str());
264	}
265	return nullptr;
266	}
267
268	// From here on, assume that the value is non-null.
269	assert(stateNonNull);
270	return stateNonNull;
271	}
272
273	// FIXME: This was originally copied from ArrayBoundChecker.cpp. Refactor?
274	ProgramStateRef CStringChecker::CheckLocation(CheckerContext &C,
275	ProgramStateRef state,
276	const Expr *S, SVal l,
277	const char *warningMsg) const {
278	// If a previous check has failed, propagate the failure.
279	if (!state)
280	return nullptr;
281
282	// Check for out of bound array element access.
283	const MemRegion *R = l.getAsRegion();
284	if (!R)
285	return state;
286
287	const ElementRegion *ER = dyn_cast<ElementRegion>(R);
288	if (!ER)
289	return state;
290
291	if (ER->getValueType() != C.getASTContext().CharTy)
292	return state;
293
294	// Get the size of the array.
295	const SubRegion *superReg = cast<SubRegion>(ER->getSuperRegion());
296	SValBuilder &svalBuilder = C.getSValBuilder();
297	SVal Extent =
298	svalBuilder.convertToArrayIndex(superReg->getExtent(svalBuilder));
299	DefinedOrUnknownSVal Size = Extent.castAs<DefinedOrUnknownSVal>();
300
301	// Get the index of the accessed element.
302	DefinedOrUnknownSVal Idx = ER->getIndex().castAs<DefinedOrUnknownSVal>();
303
304	ProgramStateRef StInBound = state->assumeInBound(Idx, Size, true);
305	ProgramStateRef StOutBound = state->assumeInBound(Idx, Size, false);
306	if (StOutBound && !StInBound) {
307	// These checks are either enabled by the CString out-of-bounds checker
308	// explicitly or implicitly by the Malloc checker.
309	// In the latter case we only do modeling but do not emit warning.
310	if (!Filter.CheckCStringOutOfBounds)
311	return nullptr;
312	// Emit a bug report.
313	if (warningMsg) {
314	emitOutOfBoundsBug(C, StOutBound, S, warningMsg);
315	} else {
316	assert(CurrentFunctionDescription);
317	assert(CurrentFunctionDescription[0] != '\0');
318
319	SmallString<80> buf;
320	llvm::raw_svector_ostream os(buf);
321	os << toUppercase(CurrentFunctionDescription[0])
322	<< &CurrentFunctionDescription[1]
323	<< " accesses out-of-bound array element";
324	emitOutOfBoundsBug(C, StOutBound, S, os.str());
325	}
326	return nullptr;
327	}
328
329	// Array bound check succeeded. From this point forward the array bound
330	// should always succeed.
331	return StInBound;
332	}
333
334	ProgramStateRef CStringChecker::CheckBufferAccess(CheckerContext &C,
335	ProgramStateRef state,
336	const Expr *Size,
337	const Expr *FirstBuf,
338	const Expr *SecondBuf,
339	const char *firstMessage,
340	const char *secondMessage,
341	bool WarnAboutSize) const {
342	// If a previous check has failed, propagate the failure.
343	if (!state)
344	return nullptr;
345
346	SValBuilder &svalBuilder = C.getSValBuilder();
347	ASTContext &Ctx = svalBuilder.getContext();
348	const LocationContext *LCtx = C.getLocationContext();
349
350	QualType sizeTy = Size->getType();
351	QualType PtrTy = Ctx.getPointerType(Ctx.CharTy);
352
353	// Check that the first buffer is non-null.
354	SVal BufVal = C.getSVal(FirstBuf);
355	state = checkNonNull(C, state, FirstBuf, BufVal);
356	if (!state)
357	return nullptr;
358
359	// If out-of-bounds checking is turned off, skip the rest.
360	if (!Filter.CheckCStringOutOfBounds)
361	return state;
362
363	// Get the access length and make sure it is known.
364	// FIXME: This assumes the caller has already checked that the access length
365	// is positive. And that it's unsigned.
366	SVal LengthVal = C.getSVal(Size);
367	Optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
368	if (!Length)
369	return state;
370
371	// Compute the offset of the last element to be accessed: size-1.
372	NonLoc One = svalBuilder.makeIntVal(1, sizeTy).castAs<NonLoc>();
373	SVal Offset = svalBuilder.evalBinOpNN(state, BO_Sub, *Length, One, sizeTy);
374	if (Offset.isUnknown())
375	return nullptr;
376	NonLoc LastOffset = Offset.castAs<NonLoc>();
377
378	// Check that the first buffer is sufficiently long.
379	SVal BufStart = svalBuilder.evalCast(BufVal, PtrTy, FirstBuf->getType());
380	if (Optional<Loc> BufLoc = BufStart.getAs<Loc>()) {
381	const Expr *warningExpr = (WarnAboutSize ? Size : FirstBuf);
382
383	SVal BufEnd = svalBuilder.evalBinOpLN(state, BO_Add, *BufLoc,
384	LastOffset, PtrTy);
385	state = CheckLocation(C, state, warningExpr, BufEnd, firstMessage);
386
387	// If the buffer isn't large enough, abort.
388	if (!state)
389	return nullptr;
390	}
391
392	// If there's a second buffer, check it as well.
393	if (SecondBuf) {
394	BufVal = state->getSVal(SecondBuf, LCtx);
395	state = checkNonNull(C, state, SecondBuf, BufVal);
396	if (!state)
397	return nullptr;
398
399	BufStart = svalBuilder.evalCast(BufVal, PtrTy, SecondBuf->getType());
400	if (Optional<Loc> BufLoc = BufStart.getAs<Loc>()) {
401	const Expr *warningExpr = (WarnAboutSize ? Size : SecondBuf);
402
403	SVal BufEnd = svalBuilder.evalBinOpLN(state, BO_Add, *BufLoc,
404	LastOffset, PtrTy);
405	state = CheckLocation(C, state, warningExpr, BufEnd, secondMessage);
406	}
407	}
408
409	// Large enough or not, return this state!
410	return state;
411	}
412
413	ProgramStateRef CStringChecker::CheckOverlap(CheckerContext &C,
414	ProgramStateRef state,
415	const Expr *Size,
416	const Expr *First,
417	const Expr *Second) const {
418	if (!Filter.CheckCStringBufferOverlap)
419	return state;
420
421	// Do a simple check for overlap: if the two arguments are from the same
422	// buffer, see if the end of the first is greater than the start of the second
423	// or vice versa.
424
425	// If a previous check has failed, propagate the failure.
426	if (!state)
427	return nullptr;
428
429	ProgramStateRef stateTrue, stateFalse;
430
431	// Get the buffer values and make sure they're known locations.
432	const LocationContext *LCtx = C.getLocationContext();
433	SVal firstVal = state->getSVal(First, LCtx);
434	SVal secondVal = state->getSVal(Second, LCtx);
435
436	Optional<Loc> firstLoc = firstVal.getAs<Loc>();
437	if (!firstLoc)
438	return state;
439
440	Optional<Loc> secondLoc = secondVal.getAs<Loc>();
441	if (!secondLoc)
442	return state;
443
444	// Are the two values the same?
445	SValBuilder &svalBuilder = C.getSValBuilder();
446	std::tie(stateTrue, stateFalse) =
447	state->assume(svalBuilder.evalEQ(state, firstLoc, secondLoc));
448
449	if (stateTrue && !stateFalse) {
450	// If the values are known to be equal, that's automatically an overlap.
451	emitOverlapBug(C, stateTrue, First, Second);
452	return nullptr;
453	}
454
455	// assume the two expressions are not equal.
456	assert(stateFalse);
457	state = stateFalse;
458
459	// Which value comes first?
460	QualType cmpTy = svalBuilder.getConditionType();
461	SVal reverse = svalBuilder.evalBinOpLL(state, BO_GT,
462	firstLoc, secondLoc, cmpTy);
463	Optional<DefinedOrUnknownSVal> reverseTest =
464	reverse.getAs<DefinedOrUnknownSVal>();
465	if (!reverseTest)
466	return state;
467
468	std::tie(stateTrue, stateFalse) = state->assume(*reverseTest);
469	if (stateTrue) {
470	if (stateFalse) {
471	// If we don't know which one comes first, we can't perform this test.
472	return state;
473	} else {
474	// Switch the values so that firstVal is before secondVal.
475	std::swap(firstLoc, secondLoc);
476
477	// Switch the Exprs as well, so that they still correspond.
478	std::swap(First, Second);
479	}
480	}
481
482	// Get the length, and make sure it too is known.
483	SVal LengthVal = state->getSVal(Size, LCtx);
484	Optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
485	if (!Length)
486	return state;
487
488	// Convert the first buffer's start address to char*.
489	// Bail out if the cast fails.
490	ASTContext &Ctx = svalBuilder.getContext();
491	QualType CharPtrTy = Ctx.getPointerType(Ctx.CharTy);
492	SVal FirstStart = svalBuilder.evalCast(*firstLoc, CharPtrTy,
493	First->getType());
494	Optional<Loc> FirstStartLoc = FirstStart.getAs<Loc>();
495	if (!FirstStartLoc)
496	return state;
497
498	// Compute the end of the first buffer. Bail out if THAT fails.
499	SVal FirstEnd = svalBuilder.evalBinOpLN(state, BO_Add,
500	FirstStartLoc, Length, CharPtrTy);
501	Optional<Loc> FirstEndLoc = FirstEnd.getAs<Loc>();
502	if (!FirstEndLoc)
503	return state;
504
505	// Is the end of the first buffer past the start of the second buffer?
506	SVal Overlap = svalBuilder.evalBinOpLL(state, BO_GT,
507	FirstEndLoc, secondLoc, cmpTy);
508	Optional<DefinedOrUnknownSVal> OverlapTest =
509	Overlap.getAs<DefinedOrUnknownSVal>();
510	if (!OverlapTest)
511	return state;
512
513	std::tie(stateTrue, stateFalse) = state->assume(*OverlapTest);
514
515	if (stateTrue && !stateFalse) {
516	// Overlap!
517	emitOverlapBug(C, stateTrue, First, Second);
518	return nullptr;
519	}
520
521	// assume the two expressions don't overlap.
522	assert(stateFalse);
523	return stateFalse;
524	}
525
526	void CStringChecker::emitOverlapBug(CheckerContext &C, ProgramStateRef state,
527	const Stmt First, const Stmt Second) const {
528	ExplodedNode *N = C.generateErrorNode(state);
529	if (!N)
530	return;
531
532	if (!BT_Overlap)
533	BT_Overlap.reset(new BugType(Filter.CheckNameCStringBufferOverlap,
534	categories::UnixAPI, "Improper arguments"));
535
536	// Generate a report for this bug.
537	auto report = llvm::make_unique<BugReport>(
538	*BT_Overlap, "Arguments must not be overlapping buffers", N);
539	report->addRange(First->getSourceRange());
540	report->addRange(Second->getSourceRange());
541
542	C.emitReport(std::move(report));
543	}
544
545	void CStringChecker::emitNullArgBug(CheckerContext &C, ProgramStateRef State,
546	const Stmt *S, StringRef WarningMsg) const {
547	if (ExplodedNode *N = C.generateErrorNode(State)) {
548	if (!BT_Null)
549	BT_Null.reset(new BuiltinBug(
550	Filter.CheckNameCStringNullArg, categories::UnixAPI,
551	"Null pointer argument in call to byte string function"));
552
553	BuiltinBug BT = static_cast<BuiltinBug >(BT_Null.get());
554	auto Report = llvm::make_unique<BugReport>(*BT, WarningMsg, N);
555	Report->addRange(S->getSourceRange());
556	if (const auto *Ex = dyn_cast<Expr>(S))
557	bugreporter::trackExpressionValue(N, Ex, *Report);
558	C.emitReport(std::move(Report));
559	}
560	}
561
562	void CStringChecker::emitOutOfBoundsBug(CheckerContext &C,
563	ProgramStateRef State, const Stmt *S,
564	StringRef WarningMsg) const {
565	if (ExplodedNode *N = C.generateErrorNode(State)) {
566	if (!BT_Bounds)
567	BT_Bounds.reset(new BuiltinBug(
568	Filter.CheckCStringOutOfBounds ? Filter.CheckNameCStringOutOfBounds
569	: Filter.CheckNameCStringNullArg,
570	"Out-of-bound array access",
571	"Byte string function accesses out-of-bound array element"));
572
573	BuiltinBug BT = static_cast<BuiltinBug >(BT_Bounds.get());
574
575	// FIXME: It would be nice to eventually make this diagnostic more clear,
576	// e.g., by referencing the original declaration or by saying why this
577	// reference is outside the range.
578	auto Report = llvm::make_unique<BugReport>(*BT, WarningMsg, N);
579	Report->addRange(S->getSourceRange());
580	C.emitReport(std::move(Report));
581	}
582	}
583
584	void CStringChecker::emitNotCStringBug(CheckerContext &C, ProgramStateRef State,
585	const Stmt *S,
586	StringRef WarningMsg) const {
587	if (ExplodedNode *N = C.generateNonFatalErrorNode(State)) {
588	if (!BT_NotCString)
589	BT_NotCString.reset(new BuiltinBug(
590	Filter.CheckNameCStringNotNullTerm, categories::UnixAPI,
591	"Argument is not a null-terminated string."));
592
593	auto Report = llvm::make_unique<BugReport>(*BT_NotCString, WarningMsg, N);
594
595	Report->addRange(S->getSourceRange());
596	C.emitReport(std::move(Report));
597	}
598	}
599
600	void CStringChecker::emitAdditionOverflowBug(CheckerContext &C,
601	ProgramStateRef State) const {
602	if (ExplodedNode *N = C.generateErrorNode(State)) {
603	if (!BT_NotCString)
604	BT_NotCString.reset(
605	new BuiltinBug(Filter.CheckNameCStringOutOfBounds, "API",
606	"Sum of expressions causes overflow."));
607
608	// This isn't a great error message, but this should never occur in real
609	// code anyway -- you'd have to create a buffer longer than a size_t can
610	// represent, which is sort of a contradiction.
611	const char *WarningMsg =
612	"This expression will create a string whose length is too big to "
613	"be represented as a size_t";
614
615	auto Report = llvm::make_unique<BugReport>(*BT_NotCString, WarningMsg, N);
616	C.emitReport(std::move(Report));
617	}
618	}
619
620	ProgramStateRef CStringChecker::checkAdditionOverflow(CheckerContext &C,
621	ProgramStateRef state,
622	NonLoc left,
623	NonLoc right) const {
624	// If out-of-bounds checking is turned off, skip the rest.
625	if (!Filter.CheckCStringOutOfBounds)
626	return state;
627
628	// If a previous check has failed, propagate the failure.
629	if (!state)
630	return nullptr;
631
632	SValBuilder &svalBuilder = C.getSValBuilder();
633	BasicValueFactory &BVF = svalBuilder.getBasicValueFactory();
634
635	QualType sizeTy = svalBuilder.getContext().getSizeType();
636	const llvm::APSInt &maxValInt = BVF.getMaxValue(sizeTy);
637	NonLoc maxVal = svalBuilder.makeIntVal(maxValInt);
638
639	SVal maxMinusRight;
640	if (right.getAs<nonloc::ConcreteInt>()) {
641	maxMinusRight = svalBuilder.evalBinOpNN(state, BO_Sub, maxVal, right,
642	sizeTy);
643	} else {
644	// Try switching the operands. (The order of these two assignments is
645	// important!)
646	maxMinusRight = svalBuilder.evalBinOpNN(state, BO_Sub, maxVal, left,
647	sizeTy);
648	left = right;
649	}
650
651	if (Optional<NonLoc> maxMinusRightNL = maxMinusRight.getAs<NonLoc>()) {
652	QualType cmpTy = svalBuilder.getConditionType();
653	// If left > max - right, we have an overflow.
654	SVal willOverflow = svalBuilder.evalBinOpNN(state, BO_GT, left,
655	*maxMinusRightNL, cmpTy);
656
657	ProgramStateRef stateOverflow, stateOkay;
658	std::tie(stateOverflow, stateOkay) =
659	state->assume(willOverflow.castAs<DefinedOrUnknownSVal>());
660
661	if (stateOverflow && !stateOkay) {
662	// We have an overflow. Emit a bug report.
663	emitAdditionOverflowBug(C, stateOverflow);
664	return nullptr;
665	}
666
667	// From now on, assume an overflow didn't occur.
668	assert(stateOkay);
669	state = stateOkay;
670	}
671
672	return state;
673	}
674
675	ProgramStateRef CStringChecker::setCStringLength(ProgramStateRef state,
676	const MemRegion *MR,
677	SVal strLength) {
678	(0) . __assert_fail ("!strLength.isUndef() && \"Attempt to set an undefined string length\"", "/home/seafit/code_projects/clang_source/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp", 678, __PRETTY_FUNCTION__))" file_link="../../../../include/assert.h.html#88" macro="true">assert(!strLength.isUndef() && "Attempt to set an undefined string length");
679
680	MR = MR->StripCasts();
681
682	switch (MR->getKind()) {
683	case MemRegion::StringRegionKind:
684	// FIXME: This can happen if we strcpy() into a string region. This is
685	// undefined [C99 6.4.5p6], but we should still warn about it.
686	return state;
687
688	case MemRegion::SymbolicRegionKind:
689	case MemRegion::AllocaRegionKind:
690	case MemRegion::VarRegionKind:
691	case MemRegion::FieldRegionKind:
692	case MemRegion::ObjCIvarRegionKind:
693	// These are the types we can currently track string lengths for.
694	break;
695
696	case MemRegion::ElementRegionKind:
697	// FIXME: Handle element regions by upper-bounding the parent region's
698	// string length.
699	return state;
700
701	default:
702	// Other regions (mostly non-data) can't have a reliable C string length.
703	// For now, just ignore the change.
704	// FIXME: These are rare but not impossible. We should output some kind of
705	// warning for things like strcpy((char[]){'a', 0}, "b");
706	return state;
707	}
708
709	if (strLength.isUnknown())
710	return state->remove<CStringLength>(MR);
711
712	return state->set<CStringLength>(MR, strLength);
713	}
714
715	SVal CStringChecker::getCStringLengthForRegion(CheckerContext &C,
716	ProgramStateRef &state,
717	const Expr *Ex,
718	const MemRegion *MR,
719	bool hypothetical) {
720	if (!hypothetical) {
721	// If there's a recorded length, go ahead and return it.
722	const SVal *Recorded = state->get<CStringLength>(MR);
723	if (Recorded)
724	return *Recorded;
725	}
726
727	// Otherwise, get a new symbol and update the state.
728	SValBuilder &svalBuilder = C.getSValBuilder();
729	QualType sizeTy = svalBuilder.getContext().getSizeType();
730	SVal strLength = svalBuilder.getMetadataSymbolVal(CStringChecker::getTag(),
731	MR, Ex, sizeTy,
732	C.getLocationContext(),
733	C.blockCount());
734
735	if (!hypothetical) {
736	if (Optional<NonLoc> strLn = strLength.getAs<NonLoc>()) {
737	// In case of unbounded calls strlen etc bound the range to SIZE_MAX/4
738	BasicValueFactory &BVF = svalBuilder.getBasicValueFactory();
739	const llvm::APSInt &maxValInt = BVF.getMaxValue(sizeTy);
740	llvm::APSInt fourInt = APSIntType(maxValInt).getValue(4);
741	const llvm::APSInt *maxLengthInt = BVF.evalAPSInt(BO_Div, maxValInt,
742	fourInt);
743	NonLoc maxLength = svalBuilder.makeIntVal(*maxLengthInt);
744	SVal evalLength = svalBuilder.evalBinOpNN(state, BO_LE, *strLn,
745	maxLength, sizeTy);
746	state = state->assume(evalLength.castAs<DefinedOrUnknownSVal>(), true);
747	}
748	state = state->set<CStringLength>(MR, strLength);
749	}
750
751	return strLength;
752	}
753
754	SVal CStringChecker::getCStringLength(CheckerContext &C, ProgramStateRef &state,
755	const Expr *Ex, SVal Buf,
756	bool hypothetical) const {
757	const MemRegion *MR = Buf.getAsRegion();
758	if (!MR) {
759	// If we can't get a region, see if it's something we /know/ isn't a
760	// C string. In the context of locations, the only time we can issue such
761	// a warning is for labels.
762	if (Optional<loc::GotoLabel> Label = Buf.getAs<loc::GotoLabel>()) {
763	if (Filter.CheckCStringNotNullTerm) {
764	SmallString<120> buf;
765	llvm::raw_svector_ostream os(buf);
766	assert(CurrentFunctionDescription);
767	os << "Argument to " << CurrentFunctionDescription
768	<< " is the address of the label '" << Label->getLabel()->getName()
769	<< "', which is not a null-terminated string";
770
771	emitNotCStringBug(C, state, Ex, os.str());
772	}
773	return UndefinedVal();
774	}
775
776	// If it's not a region and not a label, give up.
777	return UnknownVal();
778	}
779
780	// If we have a region, strip casts from it and see if we can figure out
781	// its length. For anything we can't figure out, just return UnknownVal.
782	MR = MR->StripCasts();
783
784	switch (MR->getKind()) {
785	case MemRegion::StringRegionKind: {
786	// Modifying the contents of string regions is undefined [C99 6.4.5p6],
787	// so we can assume that the byte length is the correct C string length.
788	SValBuilder &svalBuilder = C.getSValBuilder();
789	QualType sizeTy = svalBuilder.getContext().getSizeType();
790	const StringLiteral *strLit = cast<StringRegion>(MR)->getStringLiteral();
791	return svalBuilder.makeIntVal(strLit->getByteLength(), sizeTy);
792	}
793	case MemRegion::SymbolicRegionKind:
794	case MemRegion::AllocaRegionKind:
795	case MemRegion::VarRegionKind:
796	case MemRegion::FieldRegionKind:
797	case MemRegion::ObjCIvarRegionKind:
798	return getCStringLengthForRegion(C, state, Ex, MR, hypothetical);
799	case MemRegion::CompoundLiteralRegionKind:
800	// FIXME: Can we track this? Is it necessary?
801	return UnknownVal();
802	case MemRegion::ElementRegionKind:
803	// FIXME: How can we handle this? It's not good enough to subtract the
804	// offset from the base string length; consider "123\x00567" and &a[5].
805	return UnknownVal();
806	default:
807	// Other regions (mostly non-data) can't have a reliable C string length.
808	// In this case, an error is emitted and UndefinedVal is returned.
809	// The caller should always be prepared to handle this case.
810	if (Filter.CheckCStringNotNullTerm) {
811	SmallString<120> buf;
812	llvm::raw_svector_ostream os(buf);
813
814	assert(CurrentFunctionDescription);
815	os << "Argument to " << CurrentFunctionDescription << " is ";
816
817	if (SummarizeRegion(os, C.getASTContext(), MR))
818	os << ", which is not a null-terminated string";
819	else
820	os << "not a null-terminated string";
821
822	emitNotCStringBug(C, state, Ex, os.str());
823	}
824	return UndefinedVal();
825	}
826	}
827
828	const StringLiteral *CStringChecker::getCStringLiteral(CheckerContext &C,
829	ProgramStateRef &state, const Expr *expr, SVal val) const {
830
831	// Get the memory region pointed to by the val.
832	const MemRegion *bufRegion = val.getAsRegion();
833	if (!bufRegion)
834	return nullptr;
835
836	// Strip casts off the memory region.
837	bufRegion = bufRegion->StripCasts();
838
839	// Cast the memory region to a string region.
840	const StringRegion *strRegion= dyn_cast<StringRegion>(bufRegion);
841	if (!strRegion)
842	return nullptr;
843
844	// Return the actual string in the string region.
845	return strRegion->getStringLiteral();
846	}
847
848	bool CStringChecker::IsFirstBufInBound(CheckerContext &C,
849	ProgramStateRef state,
850	const Expr *FirstBuf,
851	const Expr *Size) {
852	// If we do not know that the buffer is long enough we return 'true'.
853	// Otherwise the parent region of this field region would also get
854	// invalidated, which would lead to warnings based on an unknown state.
855
856	// Originally copied from CheckBufferAccess and CheckLocation.
857	SValBuilder &svalBuilder = C.getSValBuilder();
858	ASTContext &Ctx = svalBuilder.getContext();
859	const LocationContext *LCtx = C.getLocationContext();
860
861	QualType sizeTy = Size->getType();
862	QualType PtrTy = Ctx.getPointerType(Ctx.CharTy);
863	SVal BufVal = state->getSVal(FirstBuf, LCtx);
864
865	SVal LengthVal = state->getSVal(Size, LCtx);
866	Optional<NonLoc> Length = LengthVal.getAs<NonLoc>();
867	if (!Length)
868	return true; // cf top comment.
869
870	// Compute the offset of the last element to be accessed: size-1.
871	NonLoc One = svalBuilder.makeIntVal(1, sizeTy).castAs<NonLoc>();
872	SVal Offset = svalBuilder.evalBinOpNN(state, BO_Sub, *Length, One, sizeTy);
873	if (Offset.isUnknown())
874	return true; // cf top comment
875	NonLoc LastOffset = Offset.castAs<NonLoc>();
876
877	// Check that the first buffer is sufficiently long.
878	SVal BufStart = svalBuilder.evalCast(BufVal, PtrTy, FirstBuf->getType());
879	Optional<Loc> BufLoc = BufStart.getAs<Loc>();
880	if (!BufLoc)
881	return true; // cf top comment.
882
883	SVal BufEnd =
884	svalBuilder.evalBinOpLN(state, BO_Add, *BufLoc, LastOffset, PtrTy);
885
886	// Check for out of bound array element access.
887	const MemRegion *R = BufEnd.getAsRegion();
888	if (!R)
889	return true; // cf top comment.
890
891	const ElementRegion *ER = dyn_cast<ElementRegion>(R);
892	if (!ER)
893	return true; // cf top comment.
894
895	// FIXME: Does this crash when a non-standard definition
896	// of a library function is encountered?
897	(0) . __assert_fail ("ER->getValueType() == C.getASTContext().CharTy && \"IsFirstBufInBound should only be called with char* ElementRegions\"", "/home/seafit/code_projects/clang_source/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp", 898, __PRETTY_FUNCTION__))" file_link="../../../../include/assert.h.html#88" macro="true">assert(ER->getValueType() == C.getASTContext().CharTy &&
898	(0) . __assert_fail ("ER->getValueType() == C.getASTContext().CharTy && \"IsFirstBufInBound should only be called with char* ElementRegions\"", "/home/seafit/code_projects/clang_source/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp", 898, __PRETTY_FUNCTION__))" file_link="../../../../include/assert.h.html#88" macro="true"> "IsFirstBufInBound should only be called with char* ElementRegions");
899
900	// Get the size of the array.
901	const SubRegion *superReg = cast<SubRegion>(ER->getSuperRegion());
902	SVal Extent =
903	svalBuilder.convertToArrayIndex(superReg->getExtent(svalBuilder));
904	DefinedOrUnknownSVal ExtentSize = Extent.castAs<DefinedOrUnknownSVal>();
905
906	// Get the index of the accessed element.
907	DefinedOrUnknownSVal Idx = ER->getIndex().castAs<DefinedOrUnknownSVal>();
908
909	ProgramStateRef StInBound = state->assumeInBound(Idx, ExtentSize, true);
910
911	return static_cast<bool>(StInBound);
912	}
913
914	ProgramStateRef CStringChecker::InvalidateBuffer(CheckerContext &C,
915	ProgramStateRef state,
916	const Expr *E, SVal V,
917	bool IsSourceBuffer,
918	const Expr *Size) {
919	Optional<Loc> L = V.getAs<Loc>();
920	if (!L)
921	return state;
922
923	// FIXME: This is a simplified version of what's in CFRefCount.cpp -- it makes
924	// some assumptions about the value that CFRefCount can't. Even so, it should
925	// probably be refactored.
926	if (Optional<loc::MemRegionVal> MR = L->getAs<loc::MemRegionVal>()) {
927	const MemRegion *R = MR->getRegion()->StripCasts();
928
929	// Are we dealing with an ElementRegion? If so, we should be invalidating
930	// the super-region.
931	if (const ElementRegion *ER = dyn_cast<ElementRegion>(R)) {
932	R = ER->getSuperRegion();
933	// FIXME: What about layers of ElementRegions?
934	}
935
936	// Invalidate this region.
937	const LocationContext *LCtx = C.getPredecessor()->getLocationContext();
938
939	bool CausesPointerEscape = false;
940	RegionAndSymbolInvalidationTraits ITraits;
941	// Invalidate and escape only indirect regions accessible through the source
942	// buffer.
943	if (IsSourceBuffer) {
944	ITraits.setTrait(R->getBaseRegion(),
945	RegionAndSymbolInvalidationTraits::TK_PreserveContents);
946	ITraits.setTrait(R, RegionAndSymbolInvalidationTraits::TK_SuppressEscape);
947	CausesPointerEscape = true;
948	} else {
949	const MemRegion::Kind& K = R->getKind();
950	if (K == MemRegion::FieldRegionKind)
951	if (Size && IsFirstBufInBound(C, state, E, Size)) {
952	// If destination buffer is a field region and access is in bound,
953	// do not invalidate its super region.
954	ITraits.setTrait(
955	R,
956	RegionAndSymbolInvalidationTraits::TK_DoNotInvalidateSuperRegion);
957	}
958	}
959
960	return state->invalidateRegions(R, E, C.blockCount(), LCtx,
961	CausesPointerEscape, nullptr, nullptr,
962	&ITraits);
963	}
964
965	// If we have a non-region value by chance, just remove the binding.
966	// FIXME: is this necessary or correct? This handles the non-Region
967	// cases. Is it ever valid to store to these?
968	return state->killBinding(*L);
969	}
970
971	bool CStringChecker::SummarizeRegion(raw_ostream &os, ASTContext &Ctx,
972	const MemRegion *MR) {
973	const TypedValueRegion *TVR = dyn_cast<TypedValueRegion>(MR);
974
975	switch (MR->getKind()) {
976	case MemRegion::FunctionCodeRegionKind: {
977	const NamedDecl *FD = cast<FunctionCodeRegion>(MR)->getDecl();
978	if (FD)
979	os << "the address of the function '" << *FD << '\'';
980	else
981	os << "the address of a function";
982	return true;
983	}
984	case MemRegion::BlockCodeRegionKind:
985	os << "block text";
986	return true;
987	case MemRegion::BlockDataRegionKind:
988	os << "a block";
989	return true;
990	case MemRegion::CXXThisRegionKind:
991	case MemRegion::CXXTempObjectRegionKind:
992	os << "a C++ temp object of type " << TVR->getValueType().getAsString();
993	return true;
994	case MemRegion::VarRegionKind:
995	os << "a variable of type" << TVR->getValueType().getAsString();
996	return true;
997	case MemRegion::FieldRegionKind:
998	os << "a field of type " << TVR->getValueType().getAsString();
999	return true;
1000	case MemRegion::ObjCIvarRegionKind:
1001	os << "an instance variable of type " << TVR->getValueType().getAsString();
1002	return true;
1003	default:
1004	return false;
1005	}
1006	}
1007
1008	bool CStringChecker::memsetAux(const Expr *DstBuffer, SVal CharVal,
1009	const Expr *Size, CheckerContext &C,
1010	ProgramStateRef &State) {
1011	SVal MemVal = C.getSVal(DstBuffer);
1012	SVal SizeVal = C.getSVal(Size);
1013	const MemRegion *MR = MemVal.getAsRegion();
1014	if (!MR)
1015	return false;
1016
1017	// We're about to model memset by producing a "default binding" in the Store.
1018	// Our current implementation - RegionStore - doesn't support default bindings
1019	// that don't cover the whole base region. So we should first get the offset
1020	// and the base region to figure out whether the offset of buffer is 0.
1021	RegionOffset Offset = MR->getAsOffset();
1022	const MemRegion *BR = Offset.getRegion();
1023
1024	Optional<NonLoc> SizeNL = SizeVal.getAs<NonLoc>();
1025	if (!SizeNL)
1026	return false;
1027
1028	SValBuilder &svalBuilder = C.getSValBuilder();
1029	ASTContext &Ctx = C.getASTContext();
1030
1031	// void memset(void dest, int ch, size_t count);
1032	// For now we can only handle the case of offset is 0 and concrete char value.
1033	if (Offset.isValid() && !Offset.hasSymbolicOffset() &&
1034	Offset.getOffset() == 0) {
1035	// Get the base region's extent.
1036	auto *SubReg = cast<SubRegion>(BR);
1037	DefinedOrUnknownSVal Extent = SubReg->getExtent(svalBuilder);
1038
1039	ProgramStateRef StateWholeReg, StateNotWholeReg;
1040	std::tie(StateWholeReg, StateNotWholeReg) =
1041	State->assume(svalBuilder.evalEQ(State, Extent, *SizeNL));
1042
1043	// With the semantic of 'memset()', we should convert the CharVal to
1044	// unsigned char.
1045	CharVal = svalBuilder.evalCast(CharVal, Ctx.UnsignedCharTy, Ctx.IntTy);
1046
1047	ProgramStateRef StateNullChar, StateNonNullChar;
1048	std::tie(StateNullChar, StateNonNullChar) =
1049	assumeZero(C, State, CharVal, Ctx.UnsignedCharTy);
1050
1051	if (StateWholeReg && !StateNotWholeReg && StateNullChar &&
1052	!StateNonNullChar) {
1053	// If the 'memset()' acts on the whole region of destination buffer and
1054	// the value of the second argument of 'memset()' is zero, bind the second
1055	// argument's value to the destination buffer with 'default binding'.
1056	// FIXME: Since there is no perfect way to bind the non-zero character, we
1057	// can only deal with zero value here. In the future, we need to deal with
1058	// the binding of non-zero value in the case of whole region.
1059	State = State->bindDefaultZero(svalBuilder.makeLoc(BR),
1060	C.getLocationContext());
1061	} else {
1062	// If the destination buffer's extent is not equal to the value of
1063	// third argument, just invalidate buffer.
1064	State = InvalidateBuffer(C, State, DstBuffer, MemVal,
1065	/IsSourceBuffer/ false, Size);
1066	}
1067
1068	if (StateNullChar && !StateNonNullChar) {
1069	// If the value of the second argument of 'memset()' is zero, set the
1070	// string length of destination buffer to 0 directly.
1071	State = setCStringLength(State, MR,
1072	svalBuilder.makeZeroVal(Ctx.getSizeType()));
1073	} else if (!StateNullChar && StateNonNullChar) {
1074	SVal NewStrLen = svalBuilder.getMetadataSymbolVal(
1075	CStringChecker::getTag(), MR, DstBuffer, Ctx.getSizeType(),
1076	C.getLocationContext(), C.blockCount());
1077
1078	// If the value of second argument is not zero, then the string length
1079	// is at least the size argument.
1080	SVal NewStrLenGESize = svalBuilder.evalBinOp(
1081	State, BO_GE, NewStrLen, SizeVal, svalBuilder.getConditionType());
1082
1083	State = setCStringLength(
1084	State->assume(NewStrLenGESize.castAs<DefinedOrUnknownSVal>(), true),
1085	MR, NewStrLen);
1086	}
1087	} else {
1088	// If the offset is not zero and char value is not concrete, we can do
1089	// nothing but invalidate the buffer.
1090	State = InvalidateBuffer(C, State, DstBuffer, MemVal,
1091	/IsSourceBuffer/ false, Size);
1092	}
1093	return true;
1094	}
1095
1096	//===----------------------------------------------------------------------===//
1097	// evaluation of individual function calls.
1098	//===----------------------------------------------------------------------===//
1099
1100	void CStringChecker::evalCopyCommon(CheckerContext &C,
1101	const CallExpr *CE,
1102	ProgramStateRef state,
1103	const Expr Size, const Expr Dest,
1104	const Expr *Source, bool Restricted,
1105	bool IsMempcpy) const {
1106	CurrentFunctionDescription = "memory copy function";
1107
1108	// See if the size argument is zero.
1109	const LocationContext *LCtx = C.getLocationContext();
1110	SVal sizeVal = state->getSVal(Size, LCtx);
1111	QualType sizeTy = Size->getType();
1112
1113	ProgramStateRef stateZeroSize, stateNonZeroSize;
1114	std::tie(stateZeroSize, stateNonZeroSize) =
1115	assumeZero(C, state, sizeVal, sizeTy);
1116
1117	// Get the value of the Dest.
1118	SVal destVal = state->getSVal(Dest, LCtx);
1119
1120	// If the size is zero, there won't be any actual memory access, so
1121	// just bind the return value to the destination buffer and return.
1122	if (stateZeroSize && !stateNonZeroSize) {
1123	stateZeroSize = stateZeroSize->BindExpr(CE, LCtx, destVal);
1124	C.addTransition(stateZeroSize);
1125	return;
1126	}
1127
1128	// If the size can be nonzero, we have to check the other arguments.
1129	if (stateNonZeroSize) {
1130	state = stateNonZeroSize;
1131
1132	// Ensure the destination is not null. If it is NULL there will be a
1133	// NULL pointer dereference.
1134	state = checkNonNull(C, state, Dest, destVal);
1135	if (!state)
1136	return;
1137
1138	// Get the value of the Src.
1139	SVal srcVal = state->getSVal(Source, LCtx);
1140
1141	// Ensure the source is not null. If it is NULL there will be a
1142	// NULL pointer dereference.
1143	state = checkNonNull(C, state, Source, srcVal);
1144	if (!state)
1145	return;
1146
1147	// Ensure the accesses are valid and that the buffers do not overlap.
1148	const char * const writeWarning =
1149	"Memory copy function overflows destination buffer";
1150	state = CheckBufferAccess(C, state, Size, Dest, Source,
1151	writeWarning, /* sourceWarning = */ nullptr);
1152	if (Restricted)
1153	state = CheckOverlap(C, state, Size, Dest, Source);
1154
1155	if (!state)
1156	return;
1157
1158	// If this is mempcpy, get the byte after the last byte copied and
1159	// bind the expr.
1160	if (IsMempcpy) {
1161	// Get the byte after the last byte copied.
1162	SValBuilder &SvalBuilder = C.getSValBuilder();
1163	ASTContext &Ctx = SvalBuilder.getContext();
1164	QualType CharPtrTy = Ctx.getPointerType(Ctx.CharTy);
1165	SVal DestRegCharVal =
1166	SvalBuilder.evalCast(destVal, CharPtrTy, Dest->getType());
1167	SVal lastElement = C.getSValBuilder().evalBinOp(
1168	state, BO_Add, DestRegCharVal, sizeVal, Dest->getType());
1169	// If we don't know how much we copied, we can at least
1170	// conjure a return value for later.
1171	if (lastElement.isUnknown())
1172	lastElement = C.getSValBuilder().conjureSymbolVal(nullptr, CE, LCtx,
1173	C.blockCount());
1174
1175	// The byte after the last byte copied is the return value.
1176	state = state->BindExpr(CE, LCtx, lastElement);
1177	} else {
1178	// All other copies return the destination buffer.
1179	// (Well, bcopy() has a void return type, but this won't hurt.)
1180	state = state->BindExpr(CE, LCtx, destVal);
1181	}
1182
1183	// Invalidate the destination (regular invalidation without pointer-escaping
1184	// the address of the top-level region).
1185	// FIXME: Even if we can't perfectly model the copy, we should see if we
1186	// can use LazyCompoundVals to copy the source values into the destination.
1187	// This would probably remove any existing bindings past the end of the
1188	// copied region, but that's still an improvement over blank invalidation.
1189	state = InvalidateBuffer(C, state, Dest, C.getSVal(Dest),
1190	/IsSourceBuffer/false, Size);
1191
1192	// Invalidate the source (const-invalidation without const-pointer-escaping
1193	// the address of the top-level region).
1194	state = InvalidateBuffer(C, state, Source, C.getSVal(Source),
1195	/IsSourceBuffer/true, nullptr);
1196
1197	C.addTransition(state);
1198	}
1199	}
1200
1201
1202	void CStringChecker::evalMemcpy(CheckerContext &C, const CallExpr *CE) const {
1203	if (CE->getNumArgs() < 3)
1204	return;
1205
1206	// void memcpy(void restrict dst, const void *restrict src, size_t n);
1207	// The return value is the address of the destination buffer.
1208	const Expr *Dest = CE->getArg(0);
1209	ProgramStateRef state = C.getState();
1210
1211	evalCopyCommon(C, CE, state, CE->getArg(2), Dest, CE->getArg(1), true);
1212	}
1213
1214	void CStringChecker::evalMempcpy(CheckerContext &C, const CallExpr *CE) const {
1215	if (CE->getNumArgs() < 3)
1216	return;
1217
1218	// void mempcpy(void restrict dst, const void *restrict src, size_t n);
1219	// The return value is a pointer to the byte following the last written byte.
1220	const Expr *Dest = CE->getArg(0);
1221	ProgramStateRef state = C.getState();
1222
1223	evalCopyCommon(C, CE, state, CE->getArg(2), Dest, CE->getArg(1), true, true);
1224	}
1225
1226	void CStringChecker::evalMemmove(CheckerContext &C, const CallExpr *CE) const {
1227	if (CE->getNumArgs() < 3)
1228	return;
1229
1230	// void memmove(void dst, const void *src, size_t n);
1231	// The return value is the address of the destination buffer.
1232	const Expr *Dest = CE->getArg(0);
1233	ProgramStateRef state = C.getState();
1234
1235	evalCopyCommon(C, CE, state, CE->getArg(2), Dest, CE->getArg(1));
1236	}
1237
1238	void CStringChecker::evalBcopy(CheckerContext &C, const CallExpr *CE) const {
1239	if (CE->getNumArgs() < 3)
1240	return;
1241
1242	// void bcopy(const void src, void dst, size_t n);
1243	evalCopyCommon(C, CE, C.getState(),
1244	CE->getArg(2), CE->getArg(1), CE->getArg(0));
1245	}
1246
1247	void CStringChecker::evalMemcmp(CheckerContext &C, const CallExpr *CE) const {
1248	if (CE->getNumArgs() < 3)
1249	return;
1250
1251	// int memcmp(const void s1, const void s2, size_t n);
1252	CurrentFunctionDescription = "memory comparison function";
1253
1254	const Expr *Left = CE->getArg(0);
1255	const Expr *Right = CE->getArg(1);
1256	const Expr *Size = CE->getArg(2);
1257
1258	ProgramStateRef state = C.getState();
1259	SValBuilder &svalBuilder = C.getSValBuilder();
1260
1261	// See if the size argument is zero.
1262	const LocationContext *LCtx = C.getLocationContext();
1263	SVal sizeVal = state->getSVal(Size, LCtx);
1264	QualType sizeTy = Size->getType();
1265
1266	ProgramStateRef stateZeroSize, stateNonZeroSize;
1267	std::tie(stateZeroSize, stateNonZeroSize) =
1268	assumeZero(C, state, sizeVal, sizeTy);
1269
1270	// If the size can be zero, the result will be 0 in that case, and we don't
1271	// have to check either of the buffers.
1272	if (stateZeroSize) {
1273	state = stateZeroSize;
1274	state = state->BindExpr(CE, LCtx,
1275	svalBuilder.makeZeroVal(CE->getType()));
1276	C.addTransition(state);
1277	}
1278
1279	// If the size can be nonzero, we have to check the other arguments.
1280	if (stateNonZeroSize) {
1281	state = stateNonZeroSize;
1282	// If we know the two buffers are the same, we know the result is 0.
1283	// First, get the two buffers' addresses. Another checker will have already
1284	// made sure they're not undefined.
1285	DefinedOrUnknownSVal LV =
1286	state->getSVal(Left, LCtx).castAs<DefinedOrUnknownSVal>();
1287	DefinedOrUnknownSVal RV =
1288	state->getSVal(Right, LCtx).castAs<DefinedOrUnknownSVal>();
1289
1290	// See if they are the same.
1291	DefinedOrUnknownSVal SameBuf = svalBuilder.evalEQ(state, LV, RV);
1292	ProgramStateRef StSameBuf, StNotSameBuf;
1293	std::tie(StSameBuf, StNotSameBuf) = state->assume(SameBuf);
1294
1295	// If the two arguments might be the same buffer, we know the result is 0,
1296	// and we only need to check one size.
1297	if (StSameBuf) {
1298	state = StSameBuf;
1299	state = CheckBufferAccess(C, state, Size, Left);
1300	if (state) {
1301	state = StSameBuf->BindExpr(CE, LCtx,
1302	svalBuilder.makeZeroVal(CE->getType()));
1303	C.addTransition(state);
1304	}
1305	}
1306
1307	// If the two arguments might be different buffers, we have to check the
1308	// size of both of them.
1309	if (StNotSameBuf) {
1310	state = StNotSameBuf;
1311	state = CheckBufferAccess(C, state, Size, Left, Right);
1312	if (state) {
1313	// The return value is the comparison result, which we don't know.
1314	SVal CmpV = svalBuilder.conjureSymbolVal(nullptr, CE, LCtx,
1315	C.blockCount());
1316	state = state->BindExpr(CE, LCtx, CmpV);
1317	C.addTransition(state);
1318	}
1319	}
1320	}
1321	}
1322
1323	void CStringChecker::evalstrLength(CheckerContext &C,
1324	const CallExpr *CE) const {
1325	if (CE->getNumArgs() < 1)
1326	return;
1327
1328	// size_t strlen(const char *s);
1329	evalstrLengthCommon(C, CE, /* IsStrnlen = */ false);
1330	}
1331
1332	void CStringChecker::evalstrnLength(CheckerContext &C,
1333	const CallExpr *CE) const {
1334	if (CE->getNumArgs() < 2)
1335	return;
1336
1337	// size_t strnlen(const char *s, size_t maxlen);
1338	evalstrLengthCommon(C, CE, /* IsStrnlen = */ true);
1339	}
1340
1341	void CStringChecker::evalstrLengthCommon(CheckerContext &C, const CallExpr *CE,
1342	bool IsStrnlen) const {
1343	CurrentFunctionDescription = "string length function";
1344	ProgramStateRef state = C.getState();
1345	const LocationContext *LCtx = C.getLocationContext();
1346
1347	if (IsStrnlen) {
1348	const Expr *maxlenExpr = CE->getArg(1);
1349	SVal maxlenVal = state->getSVal(maxlenExpr, LCtx);
1350
1351	ProgramStateRef stateZeroSize, stateNonZeroSize;
1352	std::tie(stateZeroSize, stateNonZeroSize) =
1353	assumeZero(C, state, maxlenVal, maxlenExpr->getType());
1354
1355	// If the size can be zero, the result will be 0 in that case, and we don't
1356	// have to check the string itself.
1357	if (stateZeroSize) {
1358	SVal zero = C.getSValBuilder().makeZeroVal(CE->getType());
1359	stateZeroSize = stateZeroSize->BindExpr(CE, LCtx, zero);
1360	C.addTransition(stateZeroSize);
1361	}
1362
1363	// If the size is GUARANTEED to be zero, we're done!
1364	if (!stateNonZeroSize)
1365	return;
1366
1367	// Otherwise, record the assumption that the size is nonzero.
1368	state = stateNonZeroSize;
1369	}
1370
1371	// Check that the string argument is non-null.
1372	const Expr *Arg = CE->getArg(0);
1373	SVal ArgVal = state->getSVal(Arg, LCtx);
1374
1375	state = checkNonNull(C, state, Arg, ArgVal);
1376
1377	if (!state)
1378	return;
1379
1380	SVal strLength = getCStringLength(C, state, Arg, ArgVal);
1381
1382	// If the argument isn't a valid C string, there's no valid state to
1383	// transition to.
1384	if (strLength.isUndef())
1385	return;
1386
1387	DefinedOrUnknownSVal result = UnknownVal();
1388
1389	// If the check is for strnlen() then bind the return value to no more than
1390	// the maxlen value.
1391	if (IsStrnlen) {
1392	QualType cmpTy = C.getSValBuilder().getConditionType();
1393
1394	// It's a little unfortunate to be getting this again,
1395	// but it's not that expensive...
1396	const Expr *maxlenExpr = CE->getArg(1);
1397	SVal maxlenVal = state->getSVal(maxlenExpr, LCtx);
1398
1399	Optional<NonLoc> strLengthNL = strLength.getAs<NonLoc>();
1400	Optional<NonLoc> maxlenValNL = maxlenVal.getAs<NonLoc>();
1401
1402	if (strLengthNL && maxlenValNL) {
1403	ProgramStateRef stateStringTooLong, stateStringNotTooLong;
1404
1405	// Check if the strLength is greater than the maxlen.
1406	std::tie(stateStringTooLong, stateStringNotTooLong) = state->assume(
1407	C.getSValBuilder()
1408	.evalBinOpNN(state, BO_GT, strLengthNL, maxlenValNL, cmpTy)
1409	.castAs<DefinedOrUnknownSVal>());
1410
1411	if (stateStringTooLong && !stateStringNotTooLong) {
1412	// If the string is longer than maxlen, return maxlen.
1413	result = *maxlenValNL;
1414	} else if (stateStringNotTooLong && !stateStringTooLong) {
1415	// If the string is shorter than maxlen, return its length.
1416	result = *strLengthNL;
1417	}
1418	}
1419
1420	if (result.isUnknown()) {
1421	// If we don't have enough information for a comparison, there's
1422	// no guarantee the full string length will actually be returned.
1423	// All we know is the return value is the min of the string length
1424	// and the limit. This is better than nothing.
1425	result = C.getSValBuilder().conjureSymbolVal(nullptr, CE, LCtx,
1426	C.blockCount());
1427	NonLoc resultNL = result.castAs<NonLoc>();
1428
1429	if (strLengthNL) {
1430	state = state->assume(C.getSValBuilder().evalBinOpNN(
1431	state, BO_LE, resultNL, *strLengthNL, cmpTy)
1432	.castAs<DefinedOrUnknownSVal>(), true);
1433	}
1434
1435	if (maxlenValNL) {
1436	state = state->assume(C.getSValBuilder().evalBinOpNN(
1437	state, BO_LE, resultNL, *maxlenValNL, cmpTy)
1438	.castAs<DefinedOrUnknownSVal>(), true);
1439	}
1440	}
1441
1442	} else {
1443	// This is a plain strlen(), not strnlen().
1444	result = strLength.castAs<DefinedOrUnknownSVal>();
1445
1446	// If we don't know the length of the string, conjure a return
1447	// value, so it can be used in constraints, at least.
1448	if (result.isUnknown()) {
1449	result = C.getSValBuilder().conjureSymbolVal(nullptr, CE, LCtx,
1450	C.blockCount());
1451	}
1452	}
1453
1454	// Bind the return value.
1455	(0) . __assert_fail ("!result.isUnknown() && \"Should have conjured a value by now\"", "/home/seafit/code_projects/clang_source/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp", 1455, __PRETTY_FUNCTION__))" file_link="../../../../include/assert.h.html#88" macro="true">assert(!result.isUnknown() && "Should have conjured a value by now");
1456	state = state->BindExpr(CE, LCtx, result);
1457	C.addTransition(state);
1458	}
1459
1460	void CStringChecker::evalStrcpy(CheckerContext &C, const CallExpr *CE) const {
1461	if (CE->getNumArgs() < 2)
1462	return;
1463
1464	// char strcpy(char restrict dst, const char *restrict src);
1465	evalStrcpyCommon(C, CE,
1466	/* returnEnd = */ false,
1467	/* isBounded = */ false,
1468	/* isAppending = */ false);
1469	}
1470
1471	void CStringChecker::evalStrncpy(CheckerContext &C, const CallExpr *CE) const {
1472	if (CE->getNumArgs() < 3)
1473	return;
1474
1475	// char strncpy(char restrict dst, const char *restrict src, size_t n);
1476	evalStrcpyCommon(C, CE,
1477	/* returnEnd = */ false,
1478	/* isBounded = */ true,
1479	/* isAppending = */ false);
1480	}
1481
1482	void CStringChecker::evalStpcpy(CheckerContext &C, const CallExpr *CE) const {
1483	if (CE->getNumArgs() < 2)
1484	return;
1485
1486	// char stpcpy(char restrict dst, const char *restrict src);
1487	evalStrcpyCommon(C, CE,
1488	/* returnEnd = */ true,
1489	/* isBounded = */ false,
1490	/* isAppending = */ false);
1491	}
1492
1493	void CStringChecker::evalStrlcpy(CheckerContext &C, const CallExpr *CE) const {
1494	if (CE->getNumArgs() < 3)
1495	return;
1496
1497	// char strlcpy(char dst, const char *src, size_t n);
1498	evalStrcpyCommon(C, CE,
1499	/* returnEnd = */ true,
1500	/* isBounded = */ true,
1501	/* isAppending = */ false,
1502	/* returnPtr = */ false);
1503	}
1504
1505	void CStringChecker::evalStrcat(CheckerContext &C, const CallExpr *CE) const {
1506	if (CE->getNumArgs() < 2)
1507	return;
1508
1509	//char strcat(char restrict s1, const char *restrict s2);
1510	evalStrcpyCommon(C, CE,
1511	/* returnEnd = */ false,
1512	/* isBounded = */ false,
1513	/* isAppending = */ true);
1514	}
1515
1516	void CStringChecker::evalStrncat(CheckerContext &C, const CallExpr *CE) const {
1517	if (CE->getNumArgs() < 3)
1518	return;
1519
1520	//char strncat(char restrict s1, const char *restrict s2, size_t n);
1521	evalStrcpyCommon(C, CE,
1522	/* returnEnd = */ false,
1523	/* isBounded = */ true,
1524	/* isAppending = */ true);
1525	}
1526
1527	void CStringChecker::evalStrlcat(CheckerContext &C, const CallExpr *CE) const {
1528	if (CE->getNumArgs() < 3)
1529	return;
1530
1531	//char strlcat(char s1, const char *s2, size_t n);
1532	evalStrcpyCommon(C, CE,
1533	/* returnEnd = */ false,
1534	/* isBounded = */ true,
1535	/* isAppending = */ true,
1536	/* returnPtr = */ false);
1537	}
1538
1539	void CStringChecker::evalStrcpyCommon(CheckerContext &C, const CallExpr *CE,
1540	bool returnEnd, bool isBounded,
1541	bool isAppending, bool returnPtr) const {
1542	CurrentFunctionDescription = "string copy function";
1543	ProgramStateRef state = C.getState();
1544	const LocationContext *LCtx = C.getLocationContext();
1545
1546	// Check that the destination is non-null.
1547	const Expr *Dst = CE->getArg(0);
1548	SVal DstVal = state->getSVal(Dst, LCtx);
1549
1550	state = checkNonNull(C, state, Dst, DstVal);
1551	if (!state)
1552	return;
1553
1554	// Check that the source is non-null.
1555	const Expr *srcExpr = CE->getArg(1);
1556	SVal srcVal = state->getSVal(srcExpr, LCtx);
1557	state = checkNonNull(C, state, srcExpr, srcVal);
1558	if (!state)
1559	return;
1560
1561	// Get the string length of the source.
1562	SVal strLength = getCStringLength(C, state, srcExpr, srcVal);
1563
1564	// If the source isn't a valid C string, give up.
1565	if (strLength.isUndef())
1566	return;
1567
1568	SValBuilder &svalBuilder = C.getSValBuilder();
1569	QualType cmpTy = svalBuilder.getConditionType();
1570	QualType sizeTy = svalBuilder.getContext().getSizeType();
1571
1572	// These two values allow checking two kinds of errors:
1573	// - actual overflows caused by a source that doesn't fit in the destination
1574	// - potential overflows caused by a bound that could exceed the destination
1575	SVal amountCopied = UnknownVal();
1576	SVal maxLastElementIndex = UnknownVal();
1577	const char *boundWarning = nullptr;
1578
1579	state = CheckOverlap(C, state, isBounded ? CE->getArg(2) : CE->getArg(1), Dst, srcExpr);
1580
1581	if (!state)
1582	return;
1583
1584	// If the function is strncpy, strncat, etc... it is bounded.
1585	if (isBounded) {
1586	// Get the max number of characters to copy.
1587	const Expr *lenExpr = CE->getArg(2);
1588	SVal lenVal = state->getSVal(lenExpr, LCtx);
1589
1590	// Protect against misdeclared strncpy().
1591	lenVal = svalBuilder.evalCast(lenVal, sizeTy, lenExpr->getType());
1592
1593	Optional<NonLoc> strLengthNL = strLength.getAs<NonLoc>();
1594	Optional<NonLoc> lenValNL = lenVal.getAs<NonLoc>();
1595
1596	// If we know both values, we might be able to figure out how much
1597	// we're copying.
1598	if (strLengthNL && lenValNL) {
1599	ProgramStateRef stateSourceTooLong, stateSourceNotTooLong;
1600
1601	// Check if the max number to copy is less than the length of the src.
1602	// If the bound is equal to the source length, strncpy won't null-
1603	// terminate the result!
1604	std::tie(stateSourceTooLong, stateSourceNotTooLong) = state->assume(
1605	svalBuilder.evalBinOpNN(state, BO_GE, strLengthNL, lenValNL, cmpTy)
1606	.castAs<DefinedOrUnknownSVal>());
1607
1608	if (stateSourceTooLong && !stateSourceNotTooLong) {
1609	// Max number to copy is less than the length of the src, so the actual
1610	// strLength copied is the max number arg.
1611	state = stateSourceTooLong;
1612	amountCopied = lenVal;
1613
1614	} else if (!stateSourceTooLong && stateSourceNotTooLong) {
1615	// The source buffer entirely fits in the bound.
1616	state = stateSourceNotTooLong;
1617	amountCopied = strLength;
1618	}
1619	}
1620
1621	// We still want to know if the bound is known to be too large.
1622	if (lenValNL) {
1623	if (isAppending) {
1624	// For strncat, the check is strlen(dst) + lenVal < sizeof(dst)
1625
1626	// Get the string length of the destination. If the destination is
1627	// memory that can't have a string length, we shouldn't be copying
1628	// into it anyway.
1629	SVal dstStrLength = getCStringLength(C, state, Dst, DstVal);
1630	if (dstStrLength.isUndef())
1631	return;
1632
1633	if (Optional<NonLoc> dstStrLengthNL = dstStrLength.getAs<NonLoc>()) {
1634	maxLastElementIndex = svalBuilder.evalBinOpNN(state, BO_Add,
1635	*lenValNL,
1636	*dstStrLengthNL,
1637	sizeTy);
1638	boundWarning = "Size argument is greater than the free space in the "
1639	"destination buffer";
1640	}
1641
1642	} else {
1643	// For strncpy, this is just checking that lenVal <= sizeof(dst)
1644	// (Yes, strncpy and strncat differ in how they treat termination.
1645	// strncat ALWAYS terminates, but strncpy doesn't.)
1646
1647	// We need a special case for when the copy size is zero, in which
1648	// case strncpy will do no work at all. Our bounds check uses n-1
1649	// as the last element accessed, so n == 0 is problematic.
1650	ProgramStateRef StateZeroSize, StateNonZeroSize;
1651	std::tie(StateZeroSize, StateNonZeroSize) =
1652	assumeZero(C, state, *lenValNL, sizeTy);
1653
1654	// If the size is known to be zero, we're done.
1655	if (StateZeroSize && !StateNonZeroSize) {
1656	if (returnPtr) {
1657	StateZeroSize = StateZeroSize->BindExpr(CE, LCtx, DstVal);
1658	} else {
1659	StateZeroSize = StateZeroSize->BindExpr(CE, LCtx, *lenValNL);
1660	}
1661	C.addTransition(StateZeroSize);
1662	return;
1663	}
1664
1665	// Otherwise, go ahead and figure out the last element we'll touch.
1666	// We don't record the non-zero assumption here because we can't
1667	// be sure. We won't warn on a possible zero.
1668	NonLoc one = svalBuilder.makeIntVal(1, sizeTy).castAs<NonLoc>();
1669	maxLastElementIndex = svalBuilder.evalBinOpNN(state, BO_Sub, *lenValNL,
1670	one, sizeTy);
1671	boundWarning = "Size argument is greater than the length of the "
1672	"destination buffer";
1673	}
1674	}
1675
1676	// If we couldn't pin down the copy length, at least bound it.
1677	// FIXME: We should actually run this code path for append as well, but
1678	// right now it creates problems with constraints (since we can end up
1679	// trying to pass constraints from symbol to symbol).
1680	if (amountCopied.isUnknown() && !isAppending) {
1681	// Try to get a "hypothetical" string length symbol, which we can later
1682	// set as a real value if that turns out to be the case.
1683	amountCopied = getCStringLength(C, state, lenExpr, srcVal, true);
1684	assert(!amountCopied.isUndef());
1685
1686	if (Optional<NonLoc> amountCopiedNL = amountCopied.getAs<NonLoc>()) {
1687	if (lenValNL) {
1688	// amountCopied <= lenVal
1689	SVal copiedLessThanBound = svalBuilder.evalBinOpNN(state, BO_LE,
1690	*amountCopiedNL,
1691	*lenValNL,
1692	cmpTy);
1693	state = state->assume(
1694	copiedLessThanBound.castAs<DefinedOrUnknownSVal>(), true);
1695	if (!state)
1696	return;
1697	}
1698
1699	if (strLengthNL) {
1700	// amountCopied <= strlen(source)
1701	SVal copiedLessThanSrc = svalBuilder.evalBinOpNN(state, BO_LE,
1702	*amountCopiedNL,
1703	*strLengthNL,
1704	cmpTy);
1705	state = state->assume(
1706	copiedLessThanSrc.castAs<DefinedOrUnknownSVal>(), true);
1707	if (!state)
1708	return;
1709	}
1710	}
1711	}
1712
1713	} else {
1714	// The function isn't bounded. The amount copied should match the length
1715	// of the source buffer.
1716	amountCopied = strLength;
1717	}
1718
1719	assert(state);
1720
1721	// This represents the number of characters copied into the destination
1722	// buffer. (It may not actually be the strlen if the destination buffer
1723	// is not terminated.)
1724	SVal finalStrLength = UnknownVal();
1725
1726	// If this is an appending function (strcat, strncat...) then set the
1727	// string length to strlen(src) + strlen(dst) since the buffer will
1728	// ultimately contain both.
1729	if (isAppending) {
1730	// Get the string length of the destination. If the destination is memory
1731	// that can't have a string length, we shouldn't be copying into it anyway.
1732	SVal dstStrLength = getCStringLength(C, state, Dst, DstVal);
1733	if (dstStrLength.isUndef())
1734	return;
1735
1736	Optional<NonLoc> srcStrLengthNL = amountCopied.getAs<NonLoc>();
1737	Optional<NonLoc> dstStrLengthNL = dstStrLength.getAs<NonLoc>();
1738
1739	// If we know both string lengths, we might know the final string length.
1740	if (srcStrLengthNL && dstStrLengthNL) {
1741	// Make sure the two lengths together don't overflow a size_t.
1742	state = checkAdditionOverflow(C, state, srcStrLengthNL, dstStrLengthNL);
1743	if (!state)
1744	return;
1745
1746	finalStrLength = svalBuilder.evalBinOpNN(state, BO_Add, *srcStrLengthNL,
1747	*dstStrLengthNL, sizeTy);
1748	}
1749
1750	// If we couldn't get a single value for the final string length,
1751	// we can at least bound it by the individual lengths.
1752	if (finalStrLength.isUnknown()) {
1753	// Try to get a "hypothetical" string length symbol, which we can later
1754	// set as a real value if that turns out to be the case.
1755	finalStrLength = getCStringLength(C, state, CE, DstVal, true);
1756	assert(!finalStrLength.isUndef());
1757
1758	if (Optional<NonLoc> finalStrLengthNL = finalStrLength.getAs<NonLoc>()) {
1759	if (srcStrLengthNL) {
1760	// finalStrLength >= srcStrLength
1761	SVal sourceInResult = svalBuilder.evalBinOpNN(state, BO_GE,
1762	*finalStrLengthNL,
1763	*srcStrLengthNL,
1764	cmpTy);
1765	state = state->assume(sourceInResult.castAs<DefinedOrUnknownSVal>(),
1766	true);
1767	if (!state)
1768	return;
1769	}
1770
1771	if (dstStrLengthNL) {
1772	// finalStrLength >= dstStrLength
1773	SVal destInResult = svalBuilder.evalBinOpNN(state, BO_GE,
1774	*finalStrLengthNL,
1775	*dstStrLengthNL,
1776	cmpTy);
1777	state =
1778	state->assume(destInResult.castAs<DefinedOrUnknownSVal>(), true);
1779	if (!state)
1780	return;
1781	}
1782	}
1783	}
1784
1785	} else {
1786	// Otherwise, this is a copy-over function (strcpy, strncpy, ...), and
1787	// the final string length will match the input string length.
1788	finalStrLength = amountCopied;
1789	}
1790
1791	SVal Result;
1792
1793	if (returnPtr) {
1794	// The final result of the function will either be a pointer past the last
1795	// copied element, or a pointer to the start of the destination buffer.
1796	Result = (returnEnd ? UnknownVal() : DstVal);
1797	} else {
1798	Result = finalStrLength;
1799	}
1800
1801	assert(state);
1802
1803	// If the destination is a MemRegion, try to check for a buffer overflow and
1804	// record the new string length.
1805	if (Optional<loc::MemRegionVal> dstRegVal =
1806	DstVal.getAs<loc::MemRegionVal>()) {
1807	QualType ptrTy = Dst->getType();
1808
1809	// If we have an exact value on a bounded copy, use that to check for
1810	// overflows, rather than our estimate about how much is actually copied.
1811	if (boundWarning) {
1812	if (Optional<NonLoc> maxLastNL = maxLastElementIndex.getAs<NonLoc>()) {
1813	SVal maxLastElement = svalBuilder.evalBinOpLN(state, BO_Add, *dstRegVal,
1814	*maxLastNL, ptrTy);
1815	state = CheckLocation(C, state, CE->getArg(2), maxLastElement,
1816	boundWarning);
1817	if (!state)
1818	return;
1819	}
1820	}
1821
1822	// Then, if the final length is known...
1823	if (Optional<NonLoc> knownStrLength = finalStrLength.getAs<NonLoc>()) {
1824	SVal lastElement = svalBuilder.evalBinOpLN(state, BO_Add, *dstRegVal,
1825	*knownStrLength, ptrTy);
1826
1827	// ...and we haven't checked the bound, we'll check the actual copy.
1828	if (!boundWarning) {
1829	const char * const warningMsg =
1830	"String copy function overflows destination buffer";
1831	state = CheckLocation(C, state, Dst, lastElement, warningMsg);
1832	if (!state)
1833	return;
1834	}
1835
1836	// If this is a stpcpy-style copy, the last element is the return value.
1837	if (returnPtr && returnEnd)
1838	Result = lastElement;
1839	}
1840
1841	// Invalidate the destination (regular invalidation without pointer-escaping
1842	// the address of the top-level region). This must happen before we set the
1843	// C string length because invalidation will clear the length.
1844	// FIXME: Even if we can't perfectly model the copy, we should see if we
1845	// can use LazyCompoundVals to copy the source values into the destination.
1846	// This would probably remove any existing bindings past the end of the
1847	// string, but that's still an improvement over blank invalidation.
1848	state = InvalidateBuffer(C, state, Dst, *dstRegVal,
1849	/IsSourceBuffer/false, nullptr);
1850
1851	// Invalidate the source (const-invalidation without const-pointer-escaping
1852	// the address of the top-level region).
1853	state = InvalidateBuffer(C, state, srcExpr, srcVal, /IsSourceBuffer/true,
1854	nullptr);
1855
1856	// Set the C string length of the destination, if we know it.
1857	if (isBounded && !isAppending) {
1858	// strncpy is annoying in that it doesn't guarantee to null-terminate
1859	// the result string. If the original string didn't fit entirely inside
1860	// the bound (including the null-terminator), we don't know how long the
1861	// result is.
1862	if (amountCopied != strLength)
1863	finalStrLength = UnknownVal();
1864	}
1865	state = setCStringLength(state, dstRegVal->getRegion(), finalStrLength);
1866	}
1867
1868	assert(state);
1869
1870	if (returnPtr) {
1871	// If this is a stpcpy-style copy, but we were unable to check for a buffer
1872	// overflow, we still need a result. Conjure a return value.
1873	if (returnEnd && Result.isUnknown()) {
1874	Result = svalBuilder.conjureSymbolVal(nullptr, CE, LCtx, C.blockCount());
1875	}
1876	}
1877	// Set the return value.
1878	state = state->BindExpr(CE, LCtx, Result);
1879	C.addTransition(state);
1880	}
1881
1882	void CStringChecker::evalStrcmp(CheckerContext &C, const CallExpr *CE) const {
1883	if (CE->getNumArgs() < 2)
1884	return;
1885
1886	//int strcmp(const char s1, const char s2);
1887	evalStrcmpCommon(C, CE, /* isBounded = / false, / ignoreCase = */ false);
1888	}
1889
1890	void CStringChecker::evalStrncmp(CheckerContext &C, const CallExpr *CE) const {
1891	if (CE->getNumArgs() < 3)
1892	return;
1893
1894	//int strncmp(const char s1, const char s2, size_t n);
1895	evalStrcmpCommon(C, CE, /* isBounded = / true, / ignoreCase = */ false);
1896	}
1897
1898	void CStringChecker::evalStrcasecmp(CheckerContext &C,
1899	const CallExpr *CE) const {
1900	if (CE->getNumArgs() < 2)
1901	return;
1902
1903	//int strcasecmp(const char s1, const char s2);
1904	evalStrcmpCommon(C, CE, /* isBounded = / false, / ignoreCase = */ true);
1905	}
1906
1907	void CStringChecker::evalStrncasecmp(CheckerContext &C,
1908	const CallExpr *CE) const {
1909	if (CE->getNumArgs() < 3)
1910	return;
1911
1912	//int strncasecmp(const char s1, const char s2, size_t n);
1913	evalStrcmpCommon(C, CE, /* isBounded = / true, / ignoreCase = */ true);
1914	}
1915
1916	void CStringChecker::evalStrcmpCommon(CheckerContext &C, const CallExpr *CE,
1917	bool isBounded, bool ignoreCase) const {
1918	CurrentFunctionDescription = "string comparison function";
1919	ProgramStateRef state = C.getState();
1920	const LocationContext *LCtx = C.getLocationContext();
1921
1922	// Check that the first string is non-null
1923	const Expr *s1 = CE->getArg(0);
1924	SVal s1Val = state->getSVal(s1, LCtx);
1925	state = checkNonNull(C, state, s1, s1Val);
1926	if (!state)
1927	return;
1928
1929	// Check that the second string is non-null.
1930	const Expr *s2 = CE->getArg(1);
1931	SVal s2Val = state->getSVal(s2, LCtx);
1932	state = checkNonNull(C, state, s2, s2Val);
1933	if (!state)
1934	return;
1935
1936	// Get the string length of the first string or give up.
1937	SVal s1Length = getCStringLength(C, state, s1, s1Val);
1938	if (s1Length.isUndef())
1939	return;
1940
1941	// Get the string length of the second string or give up.
1942	SVal s2Length = getCStringLength(C, state, s2, s2Val);
1943	if (s2Length.isUndef())
1944	return;
1945
1946	// If we know the two buffers are the same, we know the result is 0.
1947	// First, get the two buffers' addresses. Another checker will have already
1948	// made sure they're not undefined.
1949	DefinedOrUnknownSVal LV = s1Val.castAs<DefinedOrUnknownSVal>();
1950	DefinedOrUnknownSVal RV = s2Val.castAs<DefinedOrUnknownSVal>();
1951
1952	// See if they are the same.
1953	SValBuilder &svalBuilder = C.getSValBuilder();
1954	DefinedOrUnknownSVal SameBuf = svalBuilder.evalEQ(state, LV, RV);
1955	ProgramStateRef StSameBuf, StNotSameBuf;
1956	std::tie(StSameBuf, StNotSameBuf) = state->assume(SameBuf);
1957
1958	// If the two arguments might be the same buffer, we know the result is 0,
1959	// and we only need to check one size.
1960	if (StSameBuf) {
1961	StSameBuf = StSameBuf->BindExpr(CE, LCtx,
1962	svalBuilder.makeZeroVal(CE->getType()));
1963	C.addTransition(StSameBuf);
1964
1965	// If the two arguments are GUARANTEED to be the same, we're done!
1966	if (!StNotSameBuf)
1967	return;
1968	}
1969
1970	assert(StNotSameBuf);
1971	state = StNotSameBuf;
1972
1973	// At this point we can go about comparing the two buffers.
1974	// For now, we only do this if they're both known string literals.
1975
1976	// Attempt to extract string literals from both expressions.
1977	const StringLiteral *s1StrLiteral = getCStringLiteral(C, state, s1, s1Val);
1978	const StringLiteral *s2StrLiteral = getCStringLiteral(C, state, s2, s2Val);
1979	bool canComputeResult = false;
1980	SVal resultVal = svalBuilder.conjureSymbolVal(nullptr, CE, LCtx,
1981	C.blockCount());
1982
1983	if (s1StrLiteral && s2StrLiteral) {
1984	StringRef s1StrRef = s1StrLiteral->getString();
1985	StringRef s2StrRef = s2StrLiteral->getString();
1986
1987	if (isBounded) {
1988	// Get the max number of characters to compare.
1989	const Expr *lenExpr = CE->getArg(2);
1990	SVal lenVal = state->getSVal(lenExpr, LCtx);
1991
1992	// If the length is known, we can get the right substrings.
1993	if (const llvm::APSInt *len = svalBuilder.getKnownValue(state, lenVal)) {
1994	// Create substrings of each to compare the prefix.
1995	s1StrRef = s1StrRef.substr(0, (size_t)len->getZExtValue());
1996	s2StrRef = s2StrRef.substr(0, (size_t)len->getZExtValue());
1997	canComputeResult = true;
1998	}
1999	} else {
2000	// This is a normal, unbounded strcmp.
2001	canComputeResult = true;
2002	}
2003
2004	if (canComputeResult) {
2005	// Real strcmp stops at null characters.
2006	size_t s1Term = s1StrRef.find('\0');
2007	if (s1Term != StringRef::npos)
2008	s1StrRef = s1StrRef.substr(0, s1Term);
2009
2010	size_t s2Term = s2StrRef.find('\0');
2011	if (s2Term != StringRef::npos)
2012	s2StrRef = s2StrRef.substr(0, s2Term);
2013
2014	// Use StringRef's comparison methods to compute the actual result.
2015	int compareRes = ignoreCase ? s1StrRef.compare_lower(s2StrRef)
2016	: s1StrRef.compare(s2StrRef);
2017
2018	// The strcmp function returns an integer greater than, equal to, or less
2019	// than zero, [c11, p7.24.4.2].
2020	if (compareRes == 0) {
2021	resultVal = svalBuilder.makeIntVal(compareRes, CE->getType());
2022	}
2023	else {
2024	DefinedSVal zeroVal = svalBuilder.makeIntVal(0, CE->getType());
2025	// Constrain strcmp's result range based on the result of StringRef's
2026	// comparison methods.
2027	BinaryOperatorKind op = (compareRes == 1) ? BO_GT : BO_LT;
2028	SVal compareWithZero =
2029	svalBuilder.evalBinOp(state, op, resultVal, zeroVal,
2030	svalBuilder.getConditionType());
2031	DefinedSVal compareWithZeroVal = compareWithZero.castAs<DefinedSVal>();
2032	state = state->assume(compareWithZeroVal, true);
2033	}
2034	}
2035	}
2036
2037	state = state->BindExpr(CE, LCtx, resultVal);
2038
2039	// Record this as a possible path.
2040	C.addTransition(state);
2041	}
2042
2043	void CStringChecker::evalStrsep(CheckerContext &C, const CallExpr *CE) const {
2044	//char strsep(char stringp, const char delim);
2045	if (CE->getNumArgs() < 2)
2046	return;
2047
2048	// Sanity: does the search string parameter match the return type?
2049	const Expr *SearchStrPtr = CE->getArg(0);
2050	QualType CharPtrTy = SearchStrPtr->getType()->getPointeeType();
2051	if (CharPtrTy.isNull() \|\|
2052	CE->getType().getUnqualifiedType() != CharPtrTy.getUnqualifiedType())
2053	return;
2054
2055	CurrentFunctionDescription = "strsep()";
2056	ProgramStateRef State = C.getState();
2057	const LocationContext *LCtx = C.getLocationContext();
2058
2059	// Check that the search string pointer is non-null (though it may point to
2060	// a null string).
2061	SVal SearchStrVal = State->getSVal(SearchStrPtr, LCtx);
2062	State = checkNonNull(C, State, SearchStrPtr, SearchStrVal);
2063	if (!State)
2064	return;
2065
2066	// Check that the delimiter string is non-null.
2067	const Expr *DelimStr = CE->getArg(1);
2068	SVal DelimStrVal = State->getSVal(DelimStr, LCtx);
2069	State = checkNonNull(C, State, DelimStr, DelimStrVal);
2070	if (!State)
2071	return;
2072
2073	SValBuilder &SVB = C.getSValBuilder();
2074	SVal Result;
2075	if (Optional<Loc> SearchStrLoc = SearchStrVal.getAs<Loc>()) {
2076	// Get the current value of the search string pointer, as a char*.
2077	Result = State->getSVal(*SearchStrLoc, CharPtrTy);
2078
2079	// Invalidate the search string, representing the change of one delimiter
2080	// character to NUL.
2081	State = InvalidateBuffer(C, State, SearchStrPtr, Result,
2082	/IsSourceBuffer/false, nullptr);
2083
2084	// Overwrite the search string pointer. The new value is either an address
2085	// further along in the same string, or NULL if there are no more tokens.
2086	State = State->bindLoc(*SearchStrLoc,
2087	SVB.conjureSymbolVal(getTag(),
2088	CE,
2089	LCtx,
2090	CharPtrTy,
2091	C.blockCount()),
2092	LCtx);
2093	} else {
2094	assert(SearchStrVal.isUnknown());
2095	// Conjure a symbolic value. It's the best we can do.
2096	Result = SVB.conjureSymbolVal(nullptr, CE, LCtx, C.blockCount());
2097	}
2098
2099	// Set the return value, and finish.
2100	State = State->BindExpr(CE, LCtx, Result);
2101	C.addTransition(State);
2102	}
2103
2104	// These should probably be moved into a C++ standard library checker.
2105	void CStringChecker::evalStdCopy(CheckerContext &C, const CallExpr *CE) const {
2106	evalStdCopyCommon(C, CE);
2107	}
2108
2109	void CStringChecker::evalStdCopyBackward(CheckerContext &C,
2110	const CallExpr *CE) const {
2111	evalStdCopyCommon(C, CE);
2112	}
2113
2114	void CStringChecker::evalStdCopyCommon(CheckerContext &C,
2115	const CallExpr *CE) const {
2116	if (CE->getNumArgs() < 3)
2117	return;
2118
2119	ProgramStateRef State = C.getState();
2120
2121	const LocationContext *LCtx = C.getLocationContext();
2122
2123	// template <class _InputIterator, class _OutputIterator>
2124	// _OutputIterator
2125	// copy(_InputIterator __first, _InputIterator __last,
2126	// _OutputIterator __result)
2127
2128	// Invalidate the destination buffer
2129	const Expr *Dst = CE->getArg(2);
2130	SVal DstVal = State->getSVal(Dst, LCtx);
2131	State = InvalidateBuffer(C, State, Dst, DstVal, /IsSource=/false,
2132	/Size=/nullptr);
2133
2134	SValBuilder &SVB = C.getSValBuilder();
2135
2136	SVal ResultVal = SVB.conjureSymbolVal(nullptr, CE, LCtx, C.blockCount());
2137	State = State->BindExpr(CE, LCtx, ResultVal);
2138
2139	C.addTransition(State);
2140	}
2141
2142	void CStringChecker::evalMemset(CheckerContext &C, const CallExpr *CE) const {
2143	if (CE->getNumArgs() != 3)
2144	return;
2145
2146	CurrentFunctionDescription = "memory set function";
2147
2148	const Expr *Mem = CE->getArg(0);
2149	const Expr *CharE = CE->getArg(1);
2150	const Expr *Size = CE->getArg(2);
2151	ProgramStateRef State = C.getState();
2152
2153	// See if the size argument is zero.
2154	const LocationContext *LCtx = C.getLocationContext();
2155	SVal SizeVal = State->getSVal(Size, LCtx);
2156	QualType SizeTy = Size->getType();
2157
2158	ProgramStateRef StateZeroSize, StateNonZeroSize;
2159	std::tie(StateZeroSize, StateNonZeroSize) =
2160	assumeZero(C, State, SizeVal, SizeTy);
2161
2162	// Get the value of the memory area.
2163	SVal MemVal = State->getSVal(Mem, LCtx);
2164
2165	// If the size is zero, there won't be any actual memory access, so
2166	// just bind the return value to the Mem buffer and return.
2167	if (StateZeroSize && !StateNonZeroSize) {
2168	StateZeroSize = StateZeroSize->BindExpr(CE, LCtx, MemVal);
2169	C.addTransition(StateZeroSize);
2170	return;
2171	}
2172
2173	// Ensure the memory area is not null.
2174	// If it is NULL there will be a NULL pointer dereference.
2175	State = checkNonNull(C, StateNonZeroSize, Mem, MemVal);
2176	if (!State)
2177	return;
2178
2179	State = CheckBufferAccess(C, State, Size, Mem);
2180	if (!State)
2181	return;
2182
2183	// According to the values of the arguments, bind the value of the second
2184	// argument to the destination buffer and set string length, or just
2185	// invalidate the destination buffer.
2186	if (!memsetAux(Mem, C.getSVal(CharE), Size, C, State))
2187	return;
2188
2189	State = State->BindExpr(CE, LCtx, MemVal);
2190	C.addTransition(State);
2191	}
2192
2193	void CStringChecker::evalBzero(CheckerContext &C, const CallExpr *CE) const {
2194	if (CE->getNumArgs() != 2)
2195	return;
2196
2197	CurrentFunctionDescription = "memory clearance function";
2198
2199	const Expr *Mem = CE->getArg(0);
2200	const Expr *Size = CE->getArg(1);
2201	SVal Zero = C.getSValBuilder().makeZeroVal(C.getASTContext().IntTy);
2202
2203	ProgramStateRef State = C.getState();
2204
2205	// See if the size argument is zero.
2206	SVal SizeVal = C.getSVal(Size);
2207	QualType SizeTy = Size->getType();
2208
2209	ProgramStateRef StateZeroSize, StateNonZeroSize;
2210	std::tie(StateZeroSize, StateNonZeroSize) =
2211	assumeZero(C, State, SizeVal, SizeTy);
2212
2213	// If the size is zero, there won't be any actual memory access,
2214	// In this case we just return.
2215	if (StateZeroSize && !StateNonZeroSize) {
2216	C.addTransition(StateZeroSize);
2217	return;
2218	}
2219
2220	// Get the value of the memory area.
2221	SVal MemVal = C.getSVal(Mem);
2222
2223	// Ensure the memory area is not null.
2224	// If it is NULL there will be a NULL pointer dereference.
2225	State = checkNonNull(C, StateNonZeroSize, Mem, MemVal);
2226	if (!State)
2227	return;
2228
2229	State = CheckBufferAccess(C, State, Size, Mem);
2230	if (!State)
2231	return;
2232
2233	if (!memsetAux(Mem, Zero, Size, C, State))
2234	return;
2235
2236	C.addTransition(State);
2237	}
2238
2239	static bool isCPPStdLibraryFunction(const FunctionDecl *FD, StringRef Name) {
2240	IdentifierInfo *II = FD->getIdentifier();
2241	if (!II)
2242	return false;
2243
2244	if (!AnalysisDeclContext::isInStdNamespace(FD))
2245	return false;
2246
2247	if (II->getName().equals(Name))
2248	return true;
2249
2250	return false;
2251	}
2252	//===----------------------------------------------------------------------===//
2253	// The driver method, and other Checker callbacks.
2254	//===----------------------------------------------------------------------===//
2255
2256	static CStringChecker::FnCheck identifyCall(const CallExpr *CE,
2257	CheckerContext &C) {
2258	const FunctionDecl *FDecl = C.getCalleeDecl(CE);
2259	if (!FDecl)
2260	return nullptr;
2261
2262	// Pro-actively check that argument types are safe to do arithmetic upon.
2263	// We do not want to crash if someone accidentally passes a structure
2264	// into, say, a C++ overload of any of these functions.
2265	if (isCPPStdLibraryFunction(FDecl, "copy")) {
2266	if (CE->getNumArgs() < 3 \|\| !CE->getArg(2)->getType()->isPointerType())
2267	return nullptr;
2268	return &CStringChecker::evalStdCopy;
2269	} else if (isCPPStdLibraryFunction(FDecl, "copy_backward")) {
2270	if (CE->getNumArgs() < 3 \|\| !CE->getArg(2)->getType()->isPointerType())
2271	return nullptr;
2272	return &CStringChecker::evalStdCopyBackward;
2273	} else {
2274	// An umbrella check for all C library functions.
2275	for (auto I: CE->arguments()) {
2276	QualType T = I->getType();
2277	if (!T->isIntegralOrEnumerationType() && !T->isPointerType())
2278	return nullptr;
2279	}
2280	}
2281
2282	// FIXME: Poorly-factored string switches are slow.
2283	if (C.isCLibraryFunction(FDecl, "memcpy"))
2284	return &CStringChecker::evalMemcpy;
2285	else if (C.isCLibraryFunction(FDecl, "mempcpy"))
2286	return &CStringChecker::evalMempcpy;
2287	else if (C.isCLibraryFunction(FDecl, "memcmp"))
2288	return &CStringChecker::evalMemcmp;
2289	else if (C.isCLibraryFunction(FDecl, "memmove"))
2290	return &CStringChecker::evalMemmove;
2291	else if (C.isCLibraryFunction(FDecl, "memset") \|\|
2292	C.isCLibraryFunction(FDecl, "explicit_memset"))
2293	return &CStringChecker::evalMemset;
2294	else if (C.isCLibraryFunction(FDecl, "strcpy"))
2295	return &CStringChecker::evalStrcpy;
2296	else if (C.isCLibraryFunction(FDecl, "strncpy"))
2297	return &CStringChecker::evalStrncpy;
2298	else if (C.isCLibraryFunction(FDecl, "stpcpy"))
2299	return &CStringChecker::evalStpcpy;
2300	else if (C.isCLibraryFunction(FDecl, "strlcpy"))
2301	return &CStringChecker::evalStrlcpy;
2302	else if (C.isCLibraryFunction(FDecl, "strcat"))
2303	return &CStringChecker::evalStrcat;
2304	else if (C.isCLibraryFunction(FDecl, "strncat"))
2305	return &CStringChecker::evalStrncat;
2306	else if (C.isCLibraryFunction(FDecl, "strlcat"))
2307	return &CStringChecker::evalStrlcat;
2308	else if (C.isCLibraryFunction(FDecl, "strlen"))
2309	return &CStringChecker::evalstrLength;
2310	else if (C.isCLibraryFunction(FDecl, "strnlen"))
2311	return &CStringChecker::evalstrnLength;
2312	else if (C.isCLibraryFunction(FDecl, "strcmp"))
2313	return &CStringChecker::evalStrcmp;
2314	else if (C.isCLibraryFunction(FDecl, "strncmp"))
2315	return &CStringChecker::evalStrncmp;
2316	else if (C.isCLibraryFunction(FDecl, "strcasecmp"))
2317	return &CStringChecker::evalStrcasecmp;
2318	else if (C.isCLibraryFunction(FDecl, "strncasecmp"))
2319	return &CStringChecker::evalStrncasecmp;
2320	else if (C.isCLibraryFunction(FDecl, "strsep"))
2321	return &CStringChecker::evalStrsep;
2322	else if (C.isCLibraryFunction(FDecl, "bcopy"))
2323	return &CStringChecker::evalBcopy;
2324	else if (C.isCLibraryFunction(FDecl, "bcmp"))
2325	return &CStringChecker::evalMemcmp;
2326	else if (C.isCLibraryFunction(FDecl, "bzero") \|\|
2327	C.isCLibraryFunction(FDecl, "explicit_bzero"))
2328	return &CStringChecker::evalBzero;
2329
2330	return nullptr;
2331	}
2332
2333	bool CStringChecker::evalCall(const CallExpr *CE, CheckerContext &C) const {
2334
2335	FnCheck evalFunction = identifyCall(CE, C);
2336
2337	// If the callee isn't a string function, let another checker handle it.
2338	if (!evalFunction)
2339	return false;
2340
2341	// Check and evaluate the call.
2342	(this->*evalFunction)(C, CE);
2343
2344	// If the evaluate call resulted in no change, chain to the next eval call
2345	// handler.
2346	// Note, the custom CString evaluation calls assume that basic safety
2347	// properties are held. However, if the user chooses to turn off some of these
2348	// checks, we ignore the issues and leave the call evaluation to a generic
2349	// handler.
2350	return C.isDifferent();
2351	}
2352
2353	void CStringChecker::checkPreStmt(const DeclStmt *DS, CheckerContext &C) const {
2354	// Record string length for char a[] = "abc";
2355	ProgramStateRef state = C.getState();
2356
2357	for (const auto *I : DS->decls()) {
2358	const VarDecl *D = dyn_cast<VarDecl>(I);
2359	if (!D)
2360	continue;
2361
2362	// FIXME: Handle array fields of structs.
2363	if (!D->getType()->isArrayType())
2364	continue;
2365
2366	const Expr *Init = D->getInit();
2367	if (!Init)
2368	continue;
2369	if (!isa<StringLiteral>(Init))
2370	continue;
2371
2372	Loc VarLoc = state->getLValue(D, C.getLocationContext());
2373	const MemRegion *MR = VarLoc.getAsRegion();
2374	if (!MR)
2375	continue;
2376
2377	SVal StrVal = C.getSVal(Init);
2378	(0) . __assert_fail ("StrVal.isValid() && \"Initializer string is unknown or undefined\"", "/home/seafit/code_projects/clang_source/clang/lib/StaticAnalyzer/Checkers/CStringChecker.cpp", 2378, __PRETTY_FUNCTION__))" file_link="../../../../include/assert.h.html#88" macro="true">assert(StrVal.isValid() && "Initializer string is unknown or undefined");
2379	DefinedOrUnknownSVal strLength =
2380	getCStringLength(C, state, Init, StrVal).castAs<DefinedOrUnknownSVal>();
2381
2382	state = state->set<CStringLength>(MR, strLength);
2383	}
2384
2385	C.addTransition(state);
2386	}
2387
2388	ProgramStateRef
2389	CStringChecker::checkRegionChanges(ProgramStateRef state,
2390	const InvalidatedSymbols *,
2391	ArrayRef<const MemRegion *> ExplicitRegions,
2392	ArrayRef<const MemRegion *> Regions,
2393	const LocationContext *LCtx,
2394	const CallEvent *Call) const {
2395	CStringLengthTy Entries = state->get<CStringLength>();
2396	if (Entries.isEmpty())
2397	return state;
2398
2399	llvm::SmallPtrSet<const MemRegion *, 8> Invalidated;
2400	llvm::SmallPtrSet<const MemRegion *, 32> SuperRegions;
2401
2402	// First build sets for the changed regions and their super-regions.
2403	for (ArrayRef<const MemRegion *>::iterator
2404	I = Regions.begin(), E = Regions.end(); I != E; ++I) {
2405	const MemRegion MR = I;
2406	Invalidated.insert(MR);
2407
2408	SuperRegions.insert(MR);
2409	while (const SubRegion *SR = dyn_cast<SubRegion>(MR)) {
2410	MR = SR->getSuperRegion();
2411	SuperRegions.insert(MR);
2412	}
2413	}
2414
2415	CStringLengthTy::Factory &F = state->get_context<CStringLength>();
2416
2417	// Then loop over the entries in the current state.
2418	for (CStringLengthTy::iterator I = Entries.begin(),
2419	E = Entries.end(); I != E; ++I) {
2420	const MemRegion *MR = I.getKey();
2421
2422	// Is this entry for a super-region of a changed region?
2423	if (SuperRegions.count(MR)) {
2424	Entries = F.remove(Entries, MR);
2425	continue;
2426	}
2427
2428	// Is this entry for a sub-region of a changed region?
2429	const MemRegion *Super = MR;
2430	while (const SubRegion *SR = dyn_cast<SubRegion>(Super)) {
2431	Super = SR->getSuperRegion();
2432	if (Invalidated.count(Super)) {
2433	Entries = F.remove(Entries, MR);
2434	break;
2435	}
2436	}
2437	}
2438
2439	return state->set<CStringLength>(Entries);
2440	}
2441
2442	void CStringChecker::checkLiveSymbols(ProgramStateRef state,
2443	SymbolReaper &SR) const {
2444	// Mark all symbols in our string length map as valid.
2445	CStringLengthTy Entries = state->get<CStringLength>();
2446
2447	for (CStringLengthTy::iterator I = Entries.begin(), E = Entries.end();
2448	I != E; ++I) {
2449	SVal Len = I.getData();
2450
2451	for (SymExpr::symbol_iterator si = Len.symbol_begin(),
2452	se = Len.symbol_end(); si != se; ++si)
2453	SR.markInUse(*si);
2454	}
2455	}
2456
2457	void CStringChecker::checkDeadSymbols(SymbolReaper &SR,
2458	CheckerContext &C) const {
2459	ProgramStateRef state = C.getState();
2460	CStringLengthTy Entries = state->get<CStringLength>();
2461	if (Entries.isEmpty())
2462	return;
2463
2464	CStringLengthTy::Factory &F = state->get_context<CStringLength>();
2465	for (CStringLengthTy::iterator I = Entries.begin(), E = Entries.end();
2466	I != E; ++I) {
2467	SVal Len = I.getData();
2468	if (SymbolRef Sym = Len.getAsSymbol()) {
2469	if (SR.isDead(Sym))
2470	Entries = F.remove(Entries, I.getKey());
2471	}
2472	}
2473
2474	state = state->set<CStringLength>(Entries);
2475	C.addTransition(state);
2476	}
2477
2478	void ento::registerCStringModeling(CheckerManager &Mgr) {
2479	Mgr.registerChecker<CStringChecker>();
2480	}
2481
2482	bool ento::shouldRegisterCStringModeling(const LangOptions &LO) {
2483	return true;
2484	}
2485
2486	#define REGISTER_CHECKER(name) \
2487	void ento::register##name(CheckerManager &mgr) { \
2488	CStringChecker *checker = mgr.getChecker<CStringChecker>(); \
2489	checker->Filter.Check##name = true; \
2490	checker->Filter.CheckName##name = mgr.getCurrentCheckName(); \
2491	} \
2492	\
2493	bool ento::shouldRegister##name(const LangOptions &LO) { \
2494	return true; \
2495	}
2496
2497	REGISTER_CHECKER(CStringNullArg)
2498	REGISTER_CHECKER(CStringOutOfBounds)
2499	REGISTER_CHECKER(CStringBufferOverlap)
2500	REGISTER_CHECKER(CStringNotNullTerm)
2501

Clang Project